The Transcriptomic Profile and Synaptic Excitability of Vasoactive Intestinal Peptide-Expressing Interneurons in the Mouse Hippocampus

The transcriptomic profile and synaptic excitability of vasoactive intestinal peptide-expressing interneurons in the mouse hippocampus

Thèse

Xiao Luo

Doctorat en biochimie Philosophiæ doctor (Ph. D.)

Québec, Canada

The transcriptomic profile and synaptic excitability of vasoactive intestinal peptide-expressing interneurons in the mouse hippocampus

Thèse

Xiao Luo

Sous la direction de :

Lisa Topolnik, directeur de recherche

Résumé

Les neurones sont les éléments constitutifs du système nerveux. Dans le cortex, les neurones peuvent être divisés en cellules principales qui effectuent des calculs excitateurs via des connexions synaptiques locales et à longue distance et des interneurones qui contrôlent tous les domaines subcellulaires des cellules principales. Les interneurones inhibent les cellules principales en hyperpolarisant la membrane postsynaptique via les récepteurs GABA. En plus de contrôler le niveau d’excitabilité de cellules isolées via une inhibition transitoire ou à long terme, elles coordonnent le déclenchement des ensembles cellulaires principaux pour générer des oscillations de réseau qui traversent les zones du cerveau. Le dysfonctionnement des interneurones entraîne des troubles cérébraux comme la schizophrénie, l'autisme et l'épilepsie. Contrairement aux cellules principales, les interneurones présentent un haut niveau de diversité, cohérent avec leurs différents rôles fonctionnels dans les circuits cérébraux. Pour comprendre leurs fonctions de réseau, les neuroscientifiques ont développé plusieurs critères pour classer les interneurones, notamment la cytomorphologie, la connectivité, les propriétés électrophysiologiques et les marqueurs moléculaires. En général, trois types d’interneurones représentent la majorité des interneurones corticaux: les cellules somatostatine (SOM)+ ciblant la dendrite, les cellules parvalbumine (PV)+ ciblant le soma, et les cellules l'interneurone specifique peptide vasoactif intestinal (VIP)+. Dans l'hippocampe, les cellules VIP+ (hormis les cellules VIP+) jouent un rôle unique dans le réseau, car elles innervent de préférence les interneurones mais évitent les cellules principales. Cependant, leur taxonomie et leurs propriétés physiologiques sont moins claires que celles des autres types d’interneurones. Mon projet de thèse a porté sur deux sous-types de cellules VIP: la cellule VIP+ à cellule longue et la cellule à interneurone de type 3 (IS3).

Une nouvelle cellule VIP+ à longue projection (VIP-LRP) a été identifiée dans l'hippocampe CA1 Oriens/Alveus (Francavilla et al., 2018). Ces cellules ciblent de manière sélective les interneurones dans la zone hippocampique CA1, mais se projettent également dans le subiculum de la zone voisine. En outre, ils sont plus actifs pendant la période stationnaire d'éveil et silencieux pendant les oscillations thêta ou d'ondulation. Cependant, les marqueurs moléculaires qu'ils expriment n'étaient pas clairs. Pour examiner leurs marqueurs moléculaires et développer des lignées de souris spécifiques de type cellulaire en utilisant une approche génétique combinatoire, j'ai d'abord procédé à une immunohistochimie pour déterminer les marqueurs couramment exprimés, notamment le récepteur muscarinique 2, la cholécystokinine, la calbidine nNOS), calrétinine (CR) et SOM dans les cellules VIP dans les striatum oriens (SO) de CA1. Nous avons constaté que les cellules VIP-LRP étaient négatives pour SOM et nNOS, mais que la moitié d'entre elles exprimaient M2R. De plus, une petite fraction des cellules GFP expriment CCK, CB et CR. La proportion de cellules M2R + VIP-LRP était différente entre différentes souches de souris.

Ensuite, nous avons effectué le profilage transcriptomique de VIP-LRP identifiés anatomiquement en utilisant le séquençage d'ARN à cellule unique. J'ai identifié plusieurs

iii

marqueurs moléculaires, tels que la proenképhaline, le neuropeptide Y et la nétrine G1, ainsi que de nombreux autres gènes appartenant à plusieurs familles de gènes importantes: canaux ioniques, récepteurs de neurotransmetteurs, neuromodulateurs, molécules d'adhésion cellulaire et de myélinisation. De plus, les LRP VIP partagent des gènes communs lors de la comparaison avec les types de cellules VIP, CR et VIP, CCK dans le néocortex. Ensemble, ces données suggèrent que même si les LRP-VIP représentent un groupe intermédiaire dans le sous-type VIP, ils peuvent exprimer des gènes liés à des caractéristiques spécifiques permettant une coordination à longue distance des activités neuronales dans le CA1 et le subiculum.

Après cela, j'ai examiné les propriétés synaptiques excitatrices d'un autre sous-type VIP +, la cellule IS3. Des études antérieures ont montré qu’ils fabriquaient des synapses sur les interneurones dans le SO, qui contrôlent à leur tour l’intégration des apports excitateurs reçus par les dendrites proximales et distales des cellules pyramidales. Cependant, les propriétés des entrées excitatrices véhiculant les cellules IS3 restent inconnues. En utilisant l'enregistrement par patch clamp et le recalage du glutamate à deux photons, nous avons évalué les propriétés synaptiques de deux entrées excitatrices formées sur les cellules IS3 par les collatérales de Schaffer (CA) et la voie temporoammonique (TA) du cortex entorhinal. Les résultats ont montré que les courants postsynaptiques excitateurs (EPSC) évoqués dans les cellules IS3 par stimulation électrique de la voie TA avaient une amplitude, une élévation et un temps de décroissance inférieurs à ceux des synapses SC. De plus, la transmission synaptique TA-IS3 était médiée par les récepteurs AMPA et NMDA. En outre, les deux AT et SC- EPSC ont montré une facilitation synaptique à court terme en réponse à une stimulation répétitive. Enfin, les voies TA et SC ont montré un degré similaire d'intégration spatiale. Lorsque ces propriétés synaptiques ont été incorporées dans le modèle de calcul IS3 in vivo (Guet-McCreight et al., 2016), l'activation des cellules IS3 peut être provoquée par ces entrées excitatrices au cours des oscillations du thêta et de l'ondulation hippocampique. L'imagerie in vivo à deux photons chez des souris éveillées a montré que le déclenchement des cellules IS3 augmentait pendant le rythme thêta, alors que leurs activités n'étaient pas associées

Summary

Neurons are the building blocks of nervous system. In the cortex, neurons can be divided into principal cells that perform excitatory computations through local and long-range synaptic connections and interneurons that controls all subcellular domains of principal cells. Interneurons inhibit principal cells by hyperpolarizing the postsynaptic membrane via GABA receptors. In addition to controlling the level of excitability of single cells via transient or long-lasting inhibition, they coordinate the firing of principal cell ensembles to generate network oscillations that travel across brain areas. The malfunction of interneurons leads to severe brain disorders such as schizophrenia, autism and epilepsy. In contrast to principal cells, interneurons display high level of diversity, consistant with their various functional roles in the brain circuitry. To understand their network functions, neuroscientists have developed several criteria to classify interneurons, including cytomorphology, connectivity, electrophysiological properties, and molecular markers. In general, three interneuron types account for the majority of cortical interneurons: dendrite- targeting somatostatin (SOM)+ cells, soma-targeting parvalbumin (PV)+ cells, and interneuron-specific vasoactive intestinal peptide (VIP)+ cells. In the hippocampus, VIP+ cells (excluding VIP+ basket cell) play a unique role in the network, since they preferentially innervate interneurons but avoid principal cells. However, their taxonomy and physiological properties are less clear compared to other interneuron types. My PhD project focused on two subtype of VIP cells: VIP+ long-projecting cell and type 3 interneuron-specific (IS3) cell.

A novel long-projecting VIP+ cell (VIP-LRP) has been identified in the hippocampal CA1 Oriens/Alveus (Francavilla et al., 2018). These cells selectively target interneurons in the hippocampal CA1 area, but also project to the neighbouring area subiculum. In addition, they are more active during the stationary period of wakefulness, and silent during theta or ripple oscillations. However, the molecular markers they express were unclear. To examine their molecular markers and develop cell-type specific mouse lines using combinatorial genetic approach, I first performed immunohistochemistry to profile commonly expressed markers including muscarinic receptor 2 (M2R), cholecystokinin (CCK), calbidin (CB), neuronal nitric oxide synthase (nNOS), calretinin (CR), and SOM in VIP cells in the striatum oriens (SO) of CA1. We found that VIP-LRP cells were negative for SOM and nNOS but half of them expressed M2R. Moreover, a small fraction of GFP cells express CCK, CB, and CR. The proportion of M2R+ VIP-LRP cells was different between different mouse strains.

Next, we performed transcriptomic profiling of anatomically identified VIP-LRPs using single-cell RNA sequencing. I identified several molecular markers, such as proenkephalin, neuropeptide Y and netrin G1, as well as many other genes that belong to several important gene families including ion channels, neurotransmitter receptors, neuromodulators, cell adhesion and myelination molecules. In addition, VIP-LRPs share common genes when comparing with VIP;CR and VIP;CCK cell types in the neocortex. Together, these data suggest that although VIP-LRPs represent an intermediate group

within the VIP subtype, they may express genes related to specific features that allow for long-distance coordination of neuronal activities in the CA1 and the subiculum.

After that, I examined the excitatory synaptic properties of another VIP+ subtype-the IS3 cell. Previous studies showed that they make synapses on interneurons in SO, which in turn control the integration of excitatory inputs received by the proximal and distal dendrites of pyramidal cells. However, the properties of excitatory inputs conveying on IS3 cells remain unknown. Using patch clamp recording and two-photon glutamate uncaging, we evaluated the synaptic properties of two excitatory inputs formed on the IS3 cells by the Schaffer collaterals (SC) from CA3 and the Temporoammonic (TA) pathway from entorhinal cortex. The results showed that the excitatory postsynaptic currents (EPSCs) evoked in IS3 cells by electrical stimulation of the TA pathway had a smaller amplitude, slower rise and decay time compared to that of the SC synapses. In addition, TA-IS3 synaptic transmission was mediated by AMPA and NMDA receptors. Furthermore, both TA and SC-EPSCs showed short-term synaptic facilitation in response to repetitive stimulation. Finally, TA and SC pathways displayed similar degree of spatial integration. When these synaptic properties were incorporated into the in vivo-like IS3 computational model (Guet-McCreight et al., 2016), The activation of IS3 cells can be driven by these excitatory inputs during hippocampal theta and ripple oscillations. In vivo two-photon imaging in awake mice showed that the firing of IS3 cells increased during theta rhythm, whereas their activites were not associated with ripples. Together, these data shows that while excitatory inputs are able to drive the firing of IS3 cells during theta, additional mechanisms, such as local inhibition and subcortical modulation may account for the silence of IS3 cells during ripples.

Table of Contents

Résumé ...... iii

Summary ...... v

List of figures and tables ...... xiii

Abbreviations ...... xiv

Acknowledgements ...... xvi

Preface ...... xvii

Chapter 1: Introduction ...... 1

1.The anatomy of the hippocampus ...... 1

1.1 Connectivity ...... 1

1.2 Cellular architecture of the hippocampal CA1 area ...... 10

1.2.1 Distal dendrite targeting interneurons ...... 11

1.2.2 Proximal and basal dendrite targeting interneurons ...... 12

1.2.3 Soma targeting cells ...... 14

1.2.4 Axonal initial segment targeting interneurons ...... 15

1.2.5 Interneuron-specific interneurons ...... 16

1.2.6 IS3 cells ...... 17

2. Single-cell RNA sequencing and its application in identifying neuronal markers ... 19

2.1 Neurochemical properties of hippocampal interneurons ...... 19

2.2 Pipeline of scRNA-seq ...... 21

2.2.1 Single cell harvesting ...... 22

2.2.2 cDNA library construction and sequencing ...... 24

2.2.3 Qualitiy control and data processing ...... 25

vii

2.3 Cell-type specific neuronal markers identified with scRNA-seq ...... 26

2.3.1 scRNA-seq in cortical neuron classification ...... 27

2.3.2 The application of patch-seq ...... 32

3. Glutamatergic synaptic transmission in hippocampal interneurons ...... 34

3.1 Glutamate receptors in hippocampal interneurons ...... 35

3.1.1 AMPA receptors ...... 35

3.1.2 NMDA receptors ...... 36

3.1.3 Kainate and metabotropic glutamate receptors ...... 39

3.2 Comparison of TA and SC synapses in the CA1 region ...... 40

4. Temporal and spatial summation on dendrites of interneurons...... 44

4.1 Short-term plasticity at excitatory synapses of interneurons ...... 44

4.2 Dendritic spatial integration in PCs and interneurons ...... 48

4.2.1 Dendritic properties in PCs ...... 48

4.2.2 Dendritic integration in interneurons ...... 52

General hypothesis 1 ...... 59

General hypothesis 2 ...... 60

Chapter 2 Integrated article 1: Connectivity and network state-dependent recruitment of long-range VIP-GABAergic neurons in the mouse hippocampus ...... 61

Résumé...... 61

Abstract ...... 62

Introduction ...... 62

Results ...... 64

VIP-LRP neuron in the CA1 hippocampus ...... 64

Local connectivity of VIP-LRP cells ...... 66

Distant connectivity of VIP-LRP cells ...... 68

viii

Activity of VIP-LRP cells in awake mice ...... 69

Diversity of subiculum-projecting VIP-LRPs ...... 72

Discussion...... 74

Methods ...... 77

Mouse lines ...... 77

Viral constructs ...... 78

Slice preparation and patch-clamp recordings ...... 78

Two-photon laser scanning photostimulation by glutamate uncaging ...... 80

ChR2-based mapping of VIP-LRP targets ...... 81

In vitro patch-clamp data analysis ...... 82

Cell reconstruction and immunohistochemistry ...... 83

Retrograde labeling ...... 84

Electron microscopy ...... 85

Two-photon imaging in awake mice ...... 86

Analysis of two-photon Ca2+ imaging data ...... 88

Statistics ...... 90

References ...... 90

Figure Legends ...... 95

Figures ...... 101

Chapter 3 Integrated article 2: Transcriptomic profile of hippocampal long-range VIP- GABAergic neurons...... 108

Résumé...... 108

Abstract ...... 109

Introduction ...... 110

Materials and Methods ...... 111

Results ...... 116

Discussion...... 120

References ...... 122

Figures and tables ...... 126

Chapter 4 Integrated article 3: Synaptic properties and the network state-dependent recruitment of the VIP interneuron-specific interneurons in the CA1 hippocampus ..... 132

Résumé...... 132

Abstract ...... 133

Introduction ...... 134

Results ...... 136

Synaptic properties of IS3 interneurons ...... 136

Synaptic properties of IS3 cells predict that they fire during rising/peak phases of theta oscillations and during SWRs in vivo ...... 139

IS3 cells are preferentially activated during locomotion ...... 142

IS3 cells prefer to ﬁre near the rising/peak phases of theta oscillations ...... 145

IS3 cells are silent during ripples ...... 147

Discussion...... 147

Methods and Materials ...... 152

Animals ...... 152

Whole cell patch-clamp recordings in hippocampal slices in vitro...... 152

Two-photon glutamate uncaging...... 154

Computational model ...... 154

Stereotaxic injections ...... 159

In vivo two-photon Ca2+-imaging in awake mice ...... 160

Immunohistochemistry and morphological reconstructions ...... 161

Data analysis ...... 162

Statistical analysis ...... 166

Acknowledgments ...... 167

Reference ...... 167

Figures and legends ...... 179

Chapter 5: Discussion ...... 198

1. VIP-LRPs as an intermediate subpopulation of VIP interneurons ...... 198

2. Molecular profiling of VIP+ cells in CA1 SO ...... 202

3. Input-specific synaptic transmission in IS3 cells ...... 202

4. Short-term facilitation at excitatory synapses to IS3 cells ...... 204

5. Spatial summation of excitatory inputs in IS3s ...... 205

6. Comparison of IS3s with PV-BCs ...... 206

7. The potential role of IS3 cells as network gate-keepers in behavior ...... 209

8. Limitations and perspectives ...... 212

Chapter 6: Conclusion ...... 214

Bibliography ...... 217

Appendices ...... 231

Article 4: Coordination of dendritic inhibition through local disinhibitory circuits ...... 231

Résumé ...... 231

Abstract: ...... 231

Introduction ...... 232

Properties and connectivity of is3 cells ...... 232

Morphological and neurochemical features ...... 232

Physiological properties ...... 233

Connectivity ...... 233

Properties of IS3 synapses ...... 233

Comparison with VIP+ interneurons in the neocortex ...... 233

Functional role of disinhibitory circuits ...... 234

Acknowledgments ...... 235

References ...... 236

Figures and legends ...... 239

Tables in “Transcriptomic profile of hippocampal long-range VIP-GABAergic neurons” ...... 240

xii

List of figures and tables

Figure 1 Comparative views of the hippocampal system for the human (left), monkey (middle), and rat (right)...... 3

Figure 2 Drawing of the circuitry and major excitatory inputs to the hippocampal formation...... 9

Figure 3 Three types of pyramidal cell are accompanied by at least 21 classes of interneuron in the hippocampal CA1 area...... 10

Figure 4 Dendrite targeting interneurons and interneuron selective interneurons (ISIs)...... 13

Figure 5 Perisomatic targeting interneurons ...... 15

Figure 6 VIP-positive interneurons at the PYR/RAD border target O–LM interneurons. 15

Figure 7 Immunohistochemical markers primarily associated with CGE-derived interneurons...... 20

Figure 8 Work flow of scRNA-seq...... 20

Figure 9 Neuron subclasses in the somatosensory cortex ...... 29

Figure 10 Schematic of patch-seq approach...... 33

Figure 11 MGE- and CGE-dependent expression of synaptic glutamate receptors...... 38

Figure 12 Similar current–voltage curves for isolated AMPA EPSCs and different voltage dependence of the isolated NMDA response in the SC and PP (equal to TA) inputs. .. 42

Figure 13 Short-term plasticity at synapses with different I–V properties.……….………46

Figure 14 Figure 14 Dendritic excitability of pyramidal neurons..…………….……..……51

Figure 15 Biophysical properties of PV and O-LM neuron dendrites..…………………..57

Figure 16 Simplified disinhibitory network in the hippocampal CA1 area mediated by IS3 and VIPLRPs.....…………………..……………………………………………….…………211

Table 1 Electrophysiological properties of PV-BC and IS3.…..………………………..207

xiii

Abbreviations

5-HT3AR 5-hydroxytryptamine receptor 3A AHP Afterhyperpolarization AMPAR α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor AP Action potential AUROC Mean area under the receiver operator characteristic curve bAP back-propagating action potential BC Basket cell BIS Bistratified cell CA Cornu ammonis CAM Cell adhesion molecule CB Calbindin CB1R Cannabinoid receptor 1 CCK Cholecystokinin cDNA Complementary DNA CGE Caudal ganglionic eminence CNV Gene copy number variants COUPTF2 COUP transcription factor 2 DG Dentate gyrus DNQX 6,7-dinitroquinoxaline-2,3(1H,4H)-dione EC Entorhinal cortex EPSC Excitatory postsynaptic current EPSP Excitatory postsynaptic potential FACS Fluorescence-activated cell sorting FPKM Fragments per kilobase per million mapped reads GABA Gamma-aminobutyric acid GAD Glutamate decarboxylase GSEA Gene set enrichment analysis gw Gestational week 63 HCN Hyperpolarization-activated cation channel IPSP Inhibitory postsynaptic potentials IS Interneuron specific KR Kainate receptor Kv Voltage-gated potassium channel LCM Laser capture microdissection LEC Lateral entorhinal cortex LRP Long-range projecting LTP Long-term potentiation M2R Muscarinic receptor type 2 MACS Magnetic-activated cell sorting MEC Medial entorhinal cortex MGE Medial ganglionic eminence mGluR Metabotropic glutamate receptor mRNA Messenger RNA

xiv

NGFC Neurogliaform cell NGL-1 Human netrin-G1 ligand NMDAR N-methyl-D-aspartate receptor nNOS Neuronal nitric oxide synthase NPY Neuropeptide Y O-LM Oriens lacunosum-moleculare interneuron PC Pyramidal cell PCA Principle component analysis RT-PCR Real-time polymerase chain reaction PPA Perforant path-associated cell ProMMT Probabilistic Mixture Modeling for Transcriptomics PV Parvalbumin Rin Input resistance RPKM Reads per kilobase per million mapped reads S1 Primary somatosensory cortex SC Schaffer collaterals SCA Schaffer collateral-associated cell scRNA-seq Single-cell RNA sequencing SLM Stratum lacunosum-moleculare SO Stratum oriens SOM Somatostatin SP Stratum pyramidale SR Stratum radiatum STD Short-term depression STF Short-term facilitation STP Short-term plasticity SVZ Subventricular zone TA Temporoammonic TPM Transcripts per million mapped reads t-SNE T-distributed stochastic neighbor embedding UMI Unique molecular identifiers VGCC Voltage-gated calcium channels VGIC Voltage-gated ion channels VGLUT Vesicular glutamate transporter VIP Vesoactive intestinal peptide Vm Resting membrane potential αCaMKII Calcium/calmodulin-dependent protein kinase IIα

Acknowledgements

First, I would like to express my sincere gratitude to my advisor Lisa Topolnik, she helped me to build the knowledge base in the field of interneuron study and made a detailed plan for my project. She showed me some experimental skills in person, so that I can quickly adapt to the laboratory work. She is always patient and providing useful advices whenever I run into trouble or have a question about my experiment. Finally, she guided me earnestly in writing thesis and research ariticles. Without her help, I couldn’t have achieved all this.

Next, I would like to offer my great thank my former and current lab collegues including Ruggiero Francavilla, Olivier Camiré, Sona Amalyan, Einer Munoz Pino, Linda Suzanne David, Vincent Villette, Ekaterina Martianova, Elise Magnin, Étienne Gervais, and Alfonsa Zamora Moratalla. In particular, Ruggiero, Olivier and I have worked in collaboration for the longest time in the lab. We published many articles as co-authors. I would like to thank former and current lab technicians Sarah Coté, Stéphanie Racine, Émilie Pic and Juliette Tremblay. All of them helped me a lot in doing experiment, data processing and other lab works. They provided support and suggestions during ups and downs of my life. I become good friends with many of them. I’m grateful for the proofreading of the thesis performed by Qian Li.

I also appreciate the efforts made by external collaborators, including Alexandre Guet- McCreight, Frances Skinner, Peter Somogyi, Maxime Vallée, Arnaud Droit, Chin Wai Hui and Marie-Eve Tremblay. We have achieved high quality research with sincere cooperation.

Last but not least, I would like to thank my parents for their spiritual support far away from China and their good advices for my life and health. I wish them long lives full of happiness.

xvi

Preface

I included four articles in my thesis. The first three articles represent my major works in the lab. They are integrated in the thesis as chapter 2, 3 and 4. I also contributed to the review articles (article 4) attached to the appendix. The information for each article is listed below: Integrated article 1: Connectivity and network state-dependent recruitment of long-range VIP-GABAergic neurons in the mouse hippocampus Publication status: Published online in the journal “Nature communication” on November the 28th, 2018.

Author status: Second author.

Authors: Ruggiero Francavilla1, 2, 4, Vincent Villette1, 2, 4, Xiao Luo1, 2, Simon Chamberland2, Einer Muñoz-Pino1, 2, Olivier Camiré1, 2, Kristina Wagner3, Viktor Kis3, Peter Somogyi3, Lisa Topolnik1, 2, *

Affiliations of each author:

1 Neuroscience Axis, CHU de Québec Research Center – Université Laval; Québec, PQ, G1V 4G2, Canada

2 Dept. Biochemistry, Microbiology and Bio-informatics, Université Laval, Québec, PQ, G1V 0A6, Canada

3 Dept. Pharmacology, Oxford University, Oxford, OX1 3QT, UK

4 Co-first author

Author Contributions:

L.T. and P.S. supervised the whole study; R.F., X.L., S.C., E.M.-P., O.C. and L.T. performed in vitro recordings; V.V., R.F. and L.T. performed in vivo recordings; K.W., V.K. and P.S. conducted EM studies; R.F. and L.T. wrote the manuscript with comments from V.V., X.L., S.C., E.M.-P., O.C. and P.S.; R.F., L.T. and X.L. prepared the figures.

The authors declare no conflict of interest.

Correspondence should be addressed to:

Dr. Lisa Topolnik [email protected]

xvii

Integrated article 2: Transcriptomic profile of hippocampal long-range VIP- GABAergic neurons Publication status: Submitted to the journal “Scientific Reports” in November of 2018.

Author status: First author.

1, 2 1, 2 1, 2 3 Authors: Xiao Luo , Einer Muñoz-Pino , Ruggiero Francavilla , Maxime Vallée , 3 1, 2 Arnaud Droit , Lisa Topolnik

Affiliations of each author:

1 Dept. Biochemistry, Microbiology and Bio-informatics, Université Laval, Québec, PQ, 2 G1V 0A6, Canada Neuroscience Axis, CHU de Québec Research Center – Université 3 Laval; Québec, PQ, G1V 4G2, Canada CHU de Québec Research Center – Université 4 Laval; Dept. Molecular Medicine, Laval University, Québec, PQ, G1V 4G2, Canada Co- first author.

Contribution: I developed the experimental protocol for single cell harvesting. I collected three single cell samples for RNA sequencing. I also contributed to the data anaylsis and writing of the manuscripts and provided suggestions for the data processing.

Correspondence: Dr. Lisa Topolnik [email protected]

Integrated article 3: Synaptic properties and the network state-dependent recruitment of the VIP interneuron-speci1c interneurons in the CA1 hippocampus Publication status: Submitted to the journal “eLife” in August of 2018, currently under revision.

Authors: Xiao Luo1,2 *, Alexandre Guet-McCreight3,4 *, Ruggiero Francavilla1,2, Vincent Villette1,2, Frances K Skinner3,5, Lisa Topolnik1,2

Affiliations of each author: 1Dept. Biochemistry, Microbiology and Bio-informatics, Université Laval, Québec, PQ, G1V 0A6, Canada; 2Neuroscience Axis, CHU de Québec Research Center – Université Laval; Québec, PQ, G1V 4G2, Canada; 3Krembil Research Institute, University Health Network, Toronto, ON, Canada; 4Department of Physiology, University of Toronto, Toronto, ON, Canada; 5Departments of Medicine (Neurology) and Physiology, University of Toronto, Toronto, ON, Canada

*the two authors contribute equally to this work

Author status: Co-first author.

xviii

Contribution: I conducted experiments for examining synaptic transmissions, pharmacology, short-term plasticity at SC-IS3 and TA-IS3 synapses, as well as TA and SC stimulation induced AP firing in IS3 cells. I also performed VIP cell molecular profiling in the CA1, and data analysis for the experiments mentioned above. I contributed to the analysis and writing of the major chapters of the manuscript and provided suggestions for in vivo data processing and analysis.

Correspondence: Dr. Lisa Topolnik [email protected]

Attached article 4: Coordination of dendritic inhibition through local disinhibitory circuits Publication status: Received: 28 November 2014; accepted: 11 February 2015; published online: 26 February 2015 in the journal Frontiers in Synaptic Neuroscience.

Author status: Second author.

Authors: Ruggiero Francavilla, Xiao Luo, Elise Magnin, Leonid Tyanand Lisa Topolnik*

Affiliations of each author: Department of Biochemistry, Microbiology and Bio- informatics, Université Laval; Axis of Cellular and Molecular Neuroscience, IUSMQ, Québec, PQ, Canada

Contribution: I wrote the chapter on functional role of disinhibitory circuits.

*Correspondence:

Lisa Topolnik, Department of Biochemistry, Microbiology and Bio-informatics, Université Laval; Axis of Cellular and Molecular Neuroscience, IUSMQ, 2601 Ch. De La Canardière, CRULRG, Québec, PQ G1J 2G3, Canada

E-mail: [email protected]

This article is published in the journal Frontiers in Synaptic Neuroscience. Copyright © 2015 Francavilla, Luo, Magnin, Tyan and Topolnik. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution and reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

xix

Chapter 1: Introduction

1.The anatomy of the hippocampus

In 1957, Scoville and Milner reported memory deficit following the medial temporal lobe exsection from a patient HM (Scoville and Milner 1957). Since then, the critical role of the hippocampus and neighbouring areas in memory formation has been well established. The progress made in the last 50 years in the connectivity, neuroanatomy, cellular architectures, and physiology of the hippocampus have deepened our understanding of this simple yet complicated structure. In this chapter, we will focus on the anatomy and connectivity of parahippocampal regions and the hippocampal formation, as well as the cellular diversity in the CA1 area.

1.1 Connectivity

1.1.1 Parahippocampal regions

Despite the controversies in terminology, the hippocampal region can be divided into the hippocampal formation and parahippocampal regions. It is generally accepted that the hippocampal region including the hippocampal formation and neighbouring cortical areas should be considered as an integrated system involved in episodic memory. Parahippocampal (or retrohippocampal) regions include perirhinal, postrhinal, entorhinal, presubicullar, and parasubicular cortices. The hippocampal formation consists of the dentate gyrus (DG), Cornu Ammonis (CA) areas: CA3, CA2, CA1, and subiculum. In rodents, the hippocampal region occupies a large proportion of the brain, indicating a crucial role of this structure. In primates, this region is located in the medial temporal lobe and accounts for a small part of the brain, although its absolute size is 100 times larger in human than in the rats. Despite the size difference, the fundamental structure of the hippocampal region is essentially similar across species. The most significant species differences are reported for the entorhinal cortex (EC). In rodents, EC can be segregated into two cytoarchitectonically different areas: the medial and lateral area (MEC and LEC), 1

whereas in monkeys, 7 subregions are found, and in human, there are 9 subregions (Hevner and Wong-Riley, 1992; Krimer et al., 1997).

The postrhinal cortex is a six-layer cortical area located near the caudal pole of the brain (Figure 1). According to cellular architecture, this area can be divided into two subregions. The major afferents of this cortex arise from visual-association areas, such as the posterior parietal cortex and the dorsal retrosplenial cortex. Besides, ventrolateral orbital frontal cortex, caudal, and ventral temporal areas also contribute to the cortical afferents. Other inputs include the hippocampus and the subcortical nuclei such as the thalamus. Postrhinal cortex projects to another cortical area reciprocally, with strong innervation to the perirhinal cortex and MEC. In addition, there are strong inter-connections between CA1, subiculum and postrhinal cortex (Burwell and Amaral, 1998).

The perirhinal cortex is bordered rostrally by the insular cortex and is associated with the fundus and both banks of the rhinal sulcus (Figure 1). Similarly, the perirhinal cortex includes two subareas: Brodmann areas 35 and 36. Brodmann area 35 is characterized by the soma size: big, heart-shaped pyramidal cells (PC) are located in layer V, small cells are located in superficial layers. In contrast, Area 36 has a narrow granular layer IV and smaller cells located in layer V. Area 35 receives major input from the piriform cortex; the other inputs are from the entorhinal, and insular cortex, temporal association, and frontal regions. Area 36 receives afferents mainly from the ventral temporal association cortex as well as insular, entorhinal, and frontal areas. Area 36 also receives subcortical inputs from the amygdala and the thalamus, whereas area 35 receives inputs from the endopiriform nucleus and piriform transition area. In particular, area 35 projects to temporal CA1, temporal subiculum, and presubiculum (Burwell and Amaral, 1998).

The EC has drawn more attention due to its unique structural and functional connection with the hippocampal formation. As mentioned before, EC is divided into MEC and LEC in rodents. Dorsal LEC is bordered by perirhinal and postrhinal

Figure 1 Comparative views of the hippocampal system for the human (left), monkey (middle), and rat (right).

The upper panel shows the relevant structures in lateral views of the human brain (a), the monkey brain (b), and the rodent brain (c). The lower panel shows unfolded maps of the relevant cortical structures for the human brain (d), the monkey brain (e), and the rodent brain (f). Shown for the human and monkey brain are unfolded layer IV maps of the perirhinal (PER) areas 35 and 36, parahippocampal (PH) areas TF and TH, and entorhinal cortex (EC). Figures adapted from Burwell RD, Witter MP, and Amaral DG (1995) The perirhinal and postrhinal cortices of the rat: A review of the neuroanatomical literature and comparison with findings from the monkey brain. Hippocampus 5: 390–408; Insausti R, Tuñón T, Sobreviela T, Insausti AM, and Gonsalo LM (1995) The human entorhinal cortex: Acytoarchitectonic analysis. J. Comp. Neurol. 355: 171–198. Shown for the rodent brain are unfolded surface maps of the PER areas 35 and 36, the postrhinal cortex (POR), and the lateral and medial entorhinal areas (LEA and MEA). The rodent POR is the homolog of the primate PH (see text for details). In the monkey and the rat brain, the parasubiculum (ParaS) is interposed between the entorhinal and POR/PH (arrows). The pre- and parasubiculum, which are components of the parahippocampal region, are not shown. Abbreviations: cs, collateral sulcus; rs, rhinal sulcus; DG, dentate gyrus; D, dorsal; L, lateral; M, medial; ParaS, parasubiculum; PreS, presubiculum; S, septal; Sub, subiculum; T, temporal; V, ventral. Adapted from Byrne (2008).

cortices, rostral LEC is located closely to the piriform cortex, and the caudal portion of LEC is next to the dorsal MEC. The MEC is relatively smaller and narrower than the LEC, rostral MEC is bordered by ventral LEC, and caudal MEC is next to the ventral occipital pole (see figure 1). The cytoarchitectonical study has shown that Lamina dissecans of layer IV is considered as the hallmark of EC, which is more 3

significant in MEC (Figure 1). Moreover, LEC has a narrower layer II compared to MEC (Dolorfo and Amaral, 1998). EC receives inputs from various cortical and subcortical areas. Thus, it is considered as a node linking the sensory information with the hippocampus. The most prominent inputs of EC come from parahippocampal regions. In turn, EC provides strong feedback projections to these cortices. Other strong afferents of EC are from pre- and parasubiculum. However, the feedback projection form EC is less significant (Witter and Amaral, 2004).

The neocortical inputs of LEC come from piriform, agranular insular, medial and orbital frontal regions. MEC also receives inputs from piriform, frontal, cingulate, parietal, and occipital cortices. Subcortical inputs of EC include claustrum, olfactory structures, the amygdala, septal nuclei and dorsal thalamus (Pereira et al., 2016). EC projects to all regions of the hippocampal formation, including DG, CA3, CA2, CA1, and subiculum. In general, the lateral EC innervates dorsal, septal hippocampus, whereas medial EC project to ventral, temporal hippocampus. The details of these projections will be discussed in chapter 1.1.2.

The dorsal parasubiculum is bordered by the presubiculum, and the ventral parasubiculum is located near the MEC. The layer I and II/III of this cortex can be stained by acetylcholinesterase. The strongest input to this region is from MEC. The superficial layers of parasubiculum receive inputs from subiculum, while projecting back to the hippocampal formation. The major output region is the postrhinal cortex, followed by the layer II of MEC (Witter and Amaral, 2004).

Dorsoventrally, the presubiculum is located in between retrosplenial cortex and parasubiculum. Layers can be differentially stained by acetylcholinesterase. Presubiculum projects intensively to the contralateral part, and is interconnected with hippocampal formation, especially the subiculum. Besides, it is reciprocally connected with MEC, parasubiculum, and postrhinal cortex (Witter and Amaral, 2004).

1.1.2 Hippocampal formation

The hippocampal formation is a banana-shaped structure in rodents. Its rostrodorsal end is near the septal nuclei, and the caudalventral end is located next to the temporal lobe. DG and CA fields are present along the septotemporal axis, whereas subiculum is present at ventral two-third of the axis. Growing evidence has shown that dorsal and ventral hippocampi are anatomically and functionally distinct compartments. Dorsal hippocampus is involved in cognition related processes such as spatial navigation and episodic memory, whereas ventral hippocampus is responsible for stress, mood, emotion and motivation (Fanselow and Dong, 2010). Here we focus on the dorsal portion of the rodent hippocampus.

The DG consist of three layers (Figure 2). The innermost layer is called polymorphic layer. The beginning of CA3 principal cell layer extends to this layer. It contains various interneuron cell types and the axons of dentate principal cells. One of the major cell types is the mossy cell, characterized by their large dendritic spines that receive inputs from dentate principal cell. Other interneurons include hilar perforant- path associated cells, hilar commissural-associational pathway related cells, and molecular layer perforant-path associated cells (Amaral, 1978). The granule layer contains the somas of dentate excitatory principal cells, thus also named the granule cells. These cells have small and densely packed somas. Their spiny dendrites extend to the molecular layer, the outermost layer close to the fissure. Their axons target interneurons in the polymorphic layer and project all the way to principal cells in the CA3, forming the mossy fiber pathway. In addition to granule cell dendrites, some interneurons are located in the molecular layer, such as molecular layer perforant path-associated cells (Han et al., 1993) and axo-axonic cell (Soriano and Frotscher, 1989). The EC is the only cortical region that projects to DG through the perforant path (Figure 2). This pathway originates predominately form the EC layer II. In particular, the EC-DG projection is organized topographically through septotemporal and transversal axis. LEC projects to septal DG, whereas MEC terminates on temporal DG. Moreover, LEC fibers innervate the outer third of the

molecular layer, whereas MEC fibers target the middle third of this layer (Steward, 1976; Wyss, 1981). Other major inputs of DG come from subcortical areas. The medial septal and diagonal band of Broca area provide cholinergic input to septal DG, and temporal DG receives lateral septal projection. The supramamillary body of hypothalamus innervates the inner molecular layer next to the granule layer. Noradrenergic input comes from the pontine nucleus of the locus coeruleus and serotoninergic input is from raphe nuclei innervate polymorphic layer. CA3 is the only region that DG sends outputs to. The mossy fibers formed by granule cell axons pass through the polymorphic layer and terminate in CA3 PC layer and stratum lucidum.

The hippocampus proper is also called Ammon’s horn. This region can be divided into CA3, CA2, and CA1 based on their cellular architecture and connectivity (Figure 2). The principal cells in this region have pyramid-like somas. All subfields show a similar laminar structure. The superficial layer next to the neocortex is called stratum oriens (SO) which contains the basal dendrites of PCs. The axonal bundles of PCs form alveus in the outer SO. Deep to SO is the stratum pyramidale (SP), where the densely packed somas of PCs are located. Stratum radiatum (SR) contains the proximal part of the PC apical dendrites. Stratum lacunosum-moleculare (SLM) is located next to DG. Distal dendrites of PCs reside in this layer. CA3 has an additional layer stratum lucidum between SP and SR.

The CA3 contains PCs with larger somas and dendritic arborisation. The axons of CA3 PCs are highly collateralized, which project to other ipsilateral CA fields as well as the contralateral hippocampus. The axonal collaterals that project to CA1 are called the Schaffer Collaterals (SCs). This projection is organized topographically, with the proximal part of CA3 next to DG innervating septal CA1, and distal CA3 next to CA2 projecting to temporal CA1. Moreover, proximal CA3 cells preferentially target superficial CA1 SP close to SR, whereas distal CA3 cells target deep CA1 SP and SO. The CA3-CA3/CA2 projection is termed as the associational projections. The CA3-contralateral projection is termed commissural projection. Unlike SCs,

these projections travel along the septotemporal axis and form complex topographical connections (Witter and Amaral, 2004). Similar to DG, CA3 receives input from EC, medial septum, diagonal band of Broca area, locus coeruleus, and raphe nucleus. Additionally, CA3 receives inputs from the basal nucleus of the amygdala.

The connectivity of CA2, a relatively small region has only been fully characterized recently (Kohara et al., 2014). CA2 PCs have larger somas similar to CA3, but lack complex spines on apical dendrites. Proximal apical dendrites of CA2 PC receive strong mossy fiber innervation (Häussler et al., 2016), whereas CA3 innervation terminates on distal parts of apical dendrites. The cortical input of CA2 derives from EC layer II stellate cells. CA2 also receives strong subcortical input from the supramammillary body of the hypothalamus (Haglund et al., 1984). CA2 axons selectively target the basal dendrites of deep CA1 PCs in SO.

Unlike CA3, CA1 field lacks internal and contralateral connections. However, efferent projections are substantial. CA1 is innervated by all parahippocampal regions, including perirhinal, postrhinal, and entorhinal cortices. In particular, EC inputs to CA1 region mainly arise bilaterally from layer III, with minor projections also from layer II and deep layers are also found. These projections form the temporoammonic (TA) pathway (also called perforant path in some literatures), which terminates in SLM, with intensive projections located in the deeper part (close to the fissure). EC- CA1 projections are topographically organized along the transversal axis. LEC preferentially target lateral CA1 close to subiculum, whereas MEC selectively innervates medial CA1 close to CA2. Moreover, a recent study suggests that MEC projection is biased toward deep PCs and LEC tends to excite superficial PCs (Masurkar et al., 2017).

Subcortical inputs to the CA1 region are similar to those to CA3. Septal inputs terminate in SO and exclusively target interneurons. Basal and accessory basal nuclei of amygdala target CA1 intensively. Substantial inputs from the nucleus reuniens of thalamus innervate SLM. The noradrenergic input from the locus 7

coeruleus, serotonergic input from the raphe nucleus, and dopaminergic input from the ventral tegmental area and Substantia nigra are relatively weaker.

The major target region of CA1 axons is the subiculum. The distal CA1 projects to the proximal subiculum, whereas the proximal CA1 projects to distal subiculum. In addition to mutual connections with parahippocampal areas, the CA1 also project to the retrosplenial, preinfralimbic, and medial prefrontal cortices. Subcortical output regions are the anterior olfactory nucleus, the hypothalamus, nucleus accumbens, and the basal nucleus of the amygdala.

The subiculum is considered as the output region of the hippocampus. It can be divided into three layers: the polymorphic layer, the pyramidal layer, and the molecular layer. The extensive afferents come from CA1. Parahippocampal regions are reciprocally connected with the subiculum. The major neocortical targets of subiculum are retrosplenial and prefrontal cortices. The subcortical efferent targets of subiculum include the nucleus accumbens, the amygdala, septal complex, the hypothalamus, and the thalamus.

Figure 2 Drawing of the circuitry and major excitatory inputs to the hippocampal formation.

Left: Cajal proposed that the nervous system is made up of countless separate units, or nerve cells composed of dendrites, soma, and axons, each of which is a conductive device. He further proposed that information is received on the cell bodies and dendrites and conducted to distant locations through axons. Abbreviations: A, retrosplenial area; B, subiculum; C, Ammon’s horn; D, dentate gyrus; E, fimbria; F, cingulum; G, angular bundle; H, corpus callosum; K, recurrent collaterals; a, axon entering the cingulum; b, cingulum fibers; c-e, perforant path fibers; g, subicular cell; h, CA1 pyramidal cells; i, Schaffer collaterals; collaterals of alvear fibers. Adapted from Byrne (2008). Right: a: An illustration of the hippocampal circuitry. b: Diagram of the hippocampal neural network. The traditional excitatory trisynaptic pathway (entorhinal cortex (EC)–dentate gyrus–CA3–CA1–EC) is depicted by solid arrows. The axons of layer II neurons in the entorhinal cortex project to the dentate gyrus through the perforant pathway (PP), including the lateral perforant pathway (LPP) and medial perforant pathway (MPP). The dentate gyrus sends projections to the pyramidal cells in CA3 through mossy fibres. CA3 pyramidal neurons relay the information to CA1 pyramidal neurons through Schaffer collaterals. CA1 pyramidal neurons send back-projections into deep-layer neurons of the EC. CA3 also receives direct projections from EC layer II neurons through the PP. CA1 receives direct input from EC layer III neurons through the temporoammonic pathway (TA). The dentate granule cells also project to the mossy cells in the hilus and hilar interneurons, which send excitatory and inhibitory projections, respectively, back to the granule cells. Adapted from Deng et al. (2010).

1.2 Cellular architecture of the hippocampal CA1 area

As mentioned previously, cortical neurons can be divided into excitatory principal cells and inhibitory interneurons (Ramón y Cajal, 1893; Andersen et al., 1963). Compared to principal cells, interneurons have a variety of morphologies and connectivity patterns, allowing them to implement complex tasks. Experimental data and computer simulation results have shown that the inhibition is crucial for controlling the excitatory information flow along the dendrites of PCs during

Figure 3 Three types of pyramidal cell are accompanied by at least 21 classes of interneuron in the hippocampal CA1 area.

The main termination of five glutamatergic inputs are indicated on the left. The somata and dendrites of interneurons innervating pyramidal cells (blue) are orange, and those innervating mainly other interneurons are pink. Axons are purple; the main synaptic terminations are yellow. Note the association of the output synapses of different interneuron types with the perisomatic region (left) and either the Schaffer collateral/commissural or the entorhinal pathway termination zones (right), respectively. VIP, vasoactive intestinal polypeptide; VGLUT, vesicular glutamate transporter; O-LM, oriens lacunosum moleculare. Adapted from Klausberger and Somogyi, (2008).

behavioral tasks (Constantinidis et al., 2002; Yizhar et al., 2011). Moreover, interneurons can coordinate the activation of large cell ensembles and modulate network oscillations (Traub et al., 1996). Thus, it is essential to understand the classification of interneurons in terms of cellular anatomy and connectivity.

Interneurons can be categorized according to different criteria, such as morphology, connectivity, molecular markers, electrophysiological properties etc. Here, we classify hippocampal CA1 interneurons according to their postsynaptic targets. Different compartments of CA1 PCs are targeted by distinct populations of interneurons (Figure 3).

1.2.1 Distal dendrite targeting interneurons

The neurogliaform cells (NGFCs) have small and round somas located mainly in SLM. Their dendrites ramify around the soma, forming a dense cloud, which resembles the morphology of glial cells (Figure 4D) (Elfant et al., 2008). The axons are highly collateralized in SLM, forming dense axonal plexus, but also project to the molecular layer of DG, as well as SR. The dense axonal cloud of NGFCs is believed to be responsible for slow inhibitory synaptic transmission mediated by gamma- aminobutyric acid (GABA)A and GABAB receptors (Price et al., 2005). The neurochemical markers expressed by NGFCs include reelin, neuronal nitric oxide synthase (nNOS), neuropeptide Y (NPY), actinin2, and COUP transcription factor 2 (COUPTF2). Interestingly, NGFCs originate from different embryonic structures during development. nNOS-expressing NGFCs derive from the medial ganglionic eminence (MGE), whereas nNOS-lacking NGFCs originate from the caudal ganglionic eminence (CGE).

The somas and dendrites of the perforant path-associated cells (PPAs) reside in SR and SLM. Their axons are densely ramified in SLM, but also target the dendrites of DG granule cells by crossing the fissure. These cells belong to CGE-derived interneurons that express cholecystokinin (CCK) (Pawelzik et al., 2002). Together,

PPAs and NGFCs receive excitatory input mainly from EC and reuniens nucleus while providing feed-forward inhibitory control to the distal dendrites of PCs.

Oriens lacunosum-moleculare interneurons (O-LMs) are named by their morphologies (Figure 4C). Their horizontally oriented somas and spiny dendrites are exclusively located in SO and alveus, indicating that they may receive excitatory inputs from axonal collaterals of CA1 PCs. Their axons extend into SP and SR with few branches and ramify intensively when they reach SLM (Gulyas et al., 1993). Thus, O-LMs are thought to be responsible for the feedback control of PC distal dendrites, gating the information coming from EC. O-LMs express somatostatin (SOM), parvalbumin (PV) and reelin. This cell type has dual origins, which is similar to NGFCs, with the serotonin 3A receptor (5-HT3AR)-expressing subgroup deriving from CGE, whereas the 5-HT3A-lacking subgroup is derived from MGE (Chittajallu et al., 2013).

1.2.2 Proximal and basal dendrite targeting interneurons

Schaffer collateral-associated cells (SCAs) have somas and dendrites mainly located in SR but with a smaller extent in SLM (Figure 4B). Their axons target the PC oblique dendrites that receive CA3 inputs in SR and SO. They are a part of the CCK-expressing interneuron groups mediating the feed-forward inhibition of PCs (Pawelzik et al., 2002).

Bistratified cells (BISs) have horizontally oriented somas exclusively located in SO and alveus. Their aspiny multipolar dendrites branch in all layers except in SLM (Figure 4A). Their axons selectively arborize in SO and the proximal part of SR but avoid SP. They co-express the markers SOM, NPY and calbindin (CB). Trilaminar cells have similar soma and dendrite locations with BISs. However, their dendrites extend to SLM. Unlike BISs, the axons of trilaminar cells also target SP. In addition, they have long-projecting axonal collaterals innervating the medial septum and subiculum (Tóth and Freund, 1992; Ferraguti et al., 2005).

The somas of ivy cells are mainly located in SP, but also in SO and SR. They represent the largest interneuron group in CA1. They have aspiny multipolar dendrites extending to SO and SR (Figure 4E). The word “ivy” is used to describe their dense fine axons targeting the proximal and basal dendrites of PCs. Ivy cells represent the largest interneuron subtype originated in MGE (Bezaire and Soltesz, 2013). They express the markers nNOS, NPY, and COUPTF2. Together, these cells with different embryonic origins are positioned in controlling the CA3 input arriving at the proximal and base dendrites of CA1 PCs.

Figure 4 Dendrite targeting interneurons and interneuron selective interneurons (ISIs).

A: morphological reconstruction of a representative bistratiﬁed cell (BiC). At right, the reconstructed cell is illustrated to be immunopositive for SST and NPY while a different BiC highlights PV expression in this interneuron subtype. B: morphological reconstruction of a representative SCA with top inset showing that the cell is CCK immunopositive. Also shown at right (bottom) are STORM images illustrating strong CB1R immunolabeling within terminals of a separate dendrite targeting CCK interneuron. (Bottom right panel modiﬁed with permission from Dudok et al. (2015). C: morphological reconstruction of a representative O-LM with insets illustrating SST and mGluR1 immunoreactivity in

the soma and along a dendritic segment, respectively. D–F: Morphological reconstructions of representative NGFC (D), IvC (E), and ISI3 (F) cells along with single cell RT-PCR profiles probing for mRNA expression of the indicated markers. [A modified with permission from Klausberger et al. (2004). B modified with permission from Lee et al. (2010). C modified with permission from Katona et al. (2014). D–F modified with permission from Tricoire et al. (2010) Adapted from Pelkey et al. (2017).

1.2.3 Soma targeting cells

Basket cells (BCs) are named by their “basket-like” axonal boutons formed on somas and proximal dendrites of PCs (Figure 5B). BCs can be further classified by their developmental origins. MGE-derived BCs express PV (PV-BCs). These cells preferentially innervate deeply located PCs near to SO (Lee et al., 2014). They have large pyramidal-like somas located predominately in SP, but also in SO. Their aspiny dendrites extend to all layers of CA1, allowing them to integrate excitatory inputs in different layers. Their densely arborized axons are mainly distributed in SP, but also in SO and SR to a smaller degree. Beside chemical synapses formed on PCs, they are highly interconnected via gap junctions (Galarreta et al., 1999). PV-BCs are the members of MGE-derived interneurons expressing the transcription factor Nkx2.1 (Butt et al., 2005).

Another subtype of BCs expresses CCK (CCK-BCs) (Figure 5C). They differ from PV-BCs in several aspects. Although their somas are mainly located in SR, other layers such as SO, SP and SR/SLM border are also populated by this cell type. Compared to PV-BC, CCK-BCs innervate a smaller number of PCs, and a small proportion of them target interneurons, including other CCK interneurons and PV- BCs. Moreover, CCK-BCs can be further divided into two subtypes according to their neurochemical markers: vasoactive intestinal peptide (VIP)-expressing and vesicular glutamate transporter 3 (VGLUT3)-expressing subgroups. Besides, CCK- BC terminals are decorated with cannabinoid type 1 receptors (CB1Rs), which suppress the GABA release from the terminals of these cells.

Figure 5 Perisomatic targeting interneurons

A: morphological reconstruction of a representative axo-axonic cell (AAC). Inset shows that the dye- filled AAC is immunopositive for parvalbumin (PV). B: morphological reconstruction of a representative PV-BC with inset confirming PV immunoreactivity within a dendritic segment of the dye filled PV-BC. C: morphological reconstruction of a representative CCK-BC with inset showing CB1R immunolabeling within a segment of dye filled axon. Adapted from Nissen et al. (2010).

Figure 6 VIP-positive interneurons at the PYR/RAD border target O–LM interneurons.Figure 7 1.2.4Perisomatic Axonal targeting initial interneuronssegment targeting interneurons

TheA: morphological axo-axonic reconstruction cells are also of a called representative chandelier axo-axonic cells. cellTheir (AAC). somas Inset areshows largely that the found dye- filled AAC is immunopositive for parvalbumin (PV). B: morphological reconstruction of a inrepresentative SP, SO, andPV-BC SR with areas inset c adjacentonfirming PV to immunoreactivity SP. A subtype within of thesea dendritic cells segment posses of sethes dye filled PV-BC. C: morphological reconstruction of a representative CCK-BC with inset showing dendriticCB1R immunolabeling tufts extending within a tosegment all layers of dye offilled CA1 axon. (Figure Adapted 5A)from. Nissen Another et al. subtype (2010). has horizontally oriented dendritic trees dominantly occupying SO (Ganter et al., 2004). . Their axonal arborisation is selectively found within SP and proximal SO. Their axonal branches travel horizontally along the SP and either vertically or obliquely penetrate the SP, forming bouton rows exclusively targeting initial segments of PCs. Each row contains 5-12 boutons, which resemble the shape of a chandelier. 15

Approximately 1200 PCs are targeted by a single axo-axonic cell. This cell type express PV. Like other MGE-derived cells, they also express Nkx2.1 but lack the transcription factor SATB1, which is expressed in most MGE-derived cell types (Close et al., 2012).

1.2.5 Interneuron-specific interneurons

PCs are not the only target of interneurons. Several interneuron subtypes are contacting each other as well as other subtypes (Sik et al., 1995). For instance, O- LMs innervate other dendritic targeting interneurons such as NGFCs, BCs, PPAs and SCAs (Elfant et al., 2008). Besides, there are interneuron-specific (IS) interneurons that specialize in controlling other interneurons but avoid PCs. These cells account for 20% of total interneurons in the CA1 and express the marker calretinin (CR) and/or VIP. Early studies showed that VIP cells in the rat hippocampus are heterogeneous in morphology, and form symmetrical synapses onto PCs (Léránth et al., 1984). Detailed morphological characterization and immunostaining have confirmed the GABAergic nature of VIP and CR cell projections. Despite a subgroup of VIP+ basket cells that target the perisomatic area of PCs, the rest of them are subdivided into three subtypes based on their soma, dendrite, and axon locations (IS1-3, Acsády et al., 1996a, b; Gulyás et al., 1996).

The somas of IS1s are evenly distributed in all layers of CA1. Their multipolar spine- free dendrites extend to all layers but appear denser in SR. An interesting feature of IS1 is that several IS1s often form dendro-dendritic connections, such that 2-7 dendrites approach each other and travel in parallel for more than 100 µm. Their axons spread in SP and SR, and rarely in SO. The axonal boutons are unevenly distributed; high-density segments are separated by long bouton-free segments. The postsynaptic targets of IS1s include calbindin (CB)-expressing interneurons, such as SCAs PPAs and CCK-BCs, as well as other, IS1 interneurons expressing CR.

IS2s have somas residing at the border between SLM and SR. Their sparsely spiny dendritic arbors are restricted in SLM. The axons of IS2s travel vertically in SR,

innervating CB+, CR+ and VIP+ interneurons in this layer. However, another subtype of IS2s have somas in SP or SR with their dendrites spreading to all layers. Their axons arborize vertically in SR, but sparsely in SP and SO (Ascády et al., 1996a). The first type of IS2 mentioned above express VIP but not CR, whereas the second type co-express VIP and CR. As IS3 cells were studied in details, here in the following chapter I will present their properties in details.

1.2.6 IS3 cells

IS3s have round or oval somas located in SP and proximal SR. Their dendrites are aspiny, with all synapses forming directly on the shaft. They form 2-4 primary bidirectional or single directional dendrites extending to all layers, with dense collaterals residing in SLM, indicating that they may receive EC excitatory input. Their axons originate from somas or primary dendrites, then go directly down to SO and the alveus border and branches horizontally in this region, indicating that they target interneurons in this layer (Figure 4F). Immunostaining shows that IS3 cells co- express VIP and CR (Figure 6B).

CR+ cells are also found in the hippocampal formation of humans and monkeys. In the CA1 area, these cells have small bipolar or fusiform somas frequently located in SLM and SP but rarely in SO and SP. Their dendrites are sparsely spiny and vertically oriented towards the hippocampal fissure, which resembles the morphology of IS3 cells in rodents. However, a striking difference is that CR+ cells in primate hippocampus target both principal cells and interneurons, as shown by electron microscopy (Seress et al., 1993).

What are the postsynaptic targets of IS3s? Using two photon uncaging and patch- clamp recording techniques, Chamberland et al (2010) studied the local and long- range inhibitory inputs of O-LMs located in SO. Focal uncaging on the somas of SP/SR interneurons evoked slow IPSCs with large amplitude and low failure rate in O-LMs. Post hoc reconstruction showed that one of the major inhibitory inputs of O-

LMs were from IS3s. In addition, local inhibitory inputs exhibited short-term depression, but no long-term plasticity was observed (Chamberland et al 2010).

Figure 6 VIP-positive interneurons at the PYR/RAD border target O–LM interneurons.

(A) Maximal projection of a two-photon z-stack acquired in the CA1 region of the hippocampus of a VIP-eGFP mouse, showing bipolarly oriented VIP-positive cell bodies located at the PYR/RAD border and a dense axonal arborization in the O/A. (B) Immunofluorescence images of neurons located in PYR positive for calretinin (top) and VIP (middle) as well as their superimposition (bottom). Scale bar: 20 μ m. (C) Reconstruction of a bipolarly oriented VIP-positive cell, showing anatomical features of IS-IIIs (soma and dendrites are shown in black and axon is shown in red) and its irregularly spiking firing pattern typical for these cells. (D) Neurolucida reconstruction of a connected pair of interneurons: presynaptic IS-III (soma and dendrites are in black and axon is in red) and postsynaptic O–LM (soma and dendrites are in green, axon is in blue) and examples of unitary IPSCs evoked by two-photon glutamate uncaging (bottom left) and presynaptic spikes during paired recordings (bottom right). Black arrows indicate three putative contact sites onto O– LM dendrites. Modified from Chamberland et al. (2010).

Tyan et al. (2014) further examined the postsynaptic targets of IS3s through paired recordingFigure 8 andImmunohistochemical optogenetics. Consistent markers with primarily previous associated study, O -LMs with represent CGE-derived the interneurons.Figure 9 VIP-positive interneurons at the PYR/RAD border target O–LM majorinterneurons. targets of IS3s (Figure 6D). The IS3-O-LM synapses have low release probability(A) Maximal andprojection form ofmultiple a two-photon release z-stack sites. acquired Moreover, in the CA1the inhibitionregion of the provided hippocampus by IS3s of a VIP-eGFP mouse, showing bipolarly oriented VIP-positive cell bodies located at the PYR/RAD isborder able and to a controldense axonal the firingarborization of O in-LMs the O/A. through (B) Immunofluorescence rebound spikes images following of neurons after - located in PYR positive for calretinin (top) and VIP (middle) as well as their superimposition hyperpolarizations. IS3s innervate other SO interneurons, such as BIS, BC and (bottom). Scale bar: 20 μ m. (C) Reconstruction of a bipolarly oriented VIP-positive cell, showing anatomical features of IS-IIIs (soma and dendrites18 are shown in black and axon is shown in red) and its irregularly spiking firing pattern typical for these cells. (D) Neurolucida reconstruction of a connected pair of interneurons: presynaptic IS-III (soma and dendrites are in black and axon is in red) and postsynaptic O–LM (soma and dendrites are in green, axon is in blue) and examples of unitary IPSCs evoked by two-photon glutamate uncaging (bottom left) and presynaptic spikes during paired recordings (bottom right). Black arrows indicate three putative contact sites onto O– LM dendrites. Modified from Chamberland et al. (2010).

Oriens-Oriens (O-O) cells to a smaller degree. In addition, IS3 cells exhibit high input resistance, low membrane capacitance, and irregular firing pattern. Apart from the target cells of IS3s, their excitatory inputs remain unknown. The morphology of their dendrites indicates that they may receive excitatory inputs located in SR and SLM, including SC from CA3 PCs and TA from EC principal cells. In this study, we compared the synaptic properties, along with the spatial and temporal summation of these two inputs in IS3 cells.

2. Single-cell RNA sequencing and its application in identifying neuronal markers

2.1 Neurochemical properties of hippocampal interneurons

Cortical interneurons can be classified based on the expression of molecular markers identified by immunostaining. The Petilla terminology classified these specifically expressed proteins into several functional groups including transcription factors, neurotransmitters or their synthesizing enzymes, neuropeptides, Ca2+- binding proteins, neurotransmitter receptors, structural proteins, ion channels, connexins, pannexins and membrane transporters (The Petilla interneuron nomenclature Group, 2008). Like neocortical interneurons, hippocampal interneurons derive from their precursor cells in two embryonic structures: MGE and CGE in the basal telencephalon. Cell type differentiation is regulated by various transcription factors. Accordingly, MGE-derived cells can be immunolabeled by the Thyroid transcription factor Nkx2.1, which is specifically expressed in MGE. MGE- derived interneurons account for 60% of total hippocampal interneurons, and can be labelled by several markers including PV, SOM, nNOS and NPY. For example, PV- BCs, BISs, and AACs express the marker PV, while SOM is present in O-LMs, BISs and a subset of long-projecting cells (Tricoire et al., 2011). On the other hand, no common transcription factor has been identified in CGE-derived cells, but they are often labelled with 5-HT3AR. The molecular markers for CGE-derived hippocampal cells include CCK, COUPTF2, Muscarinic receptor type 2 (M2R), CR and VIP

Figure 7 Immunohistochemical markers primarily associated with CGE-derived interneurons.

A, B, Representative images illustrating the co-expression of GFP with M2R (i), CoupTFII (ii), CCK (iii), VIP (iv), and CR (v) in the Nkx2-1 Cre:RCE (A) and GAD65-GFP (B) lines. The filled arrowheads indicate interneurons co-expressing GFP and the indicated marker. The open arrowheads indicate cells expressing the indicated marker but not GFP. C, Histogram showing the contribution of GFP+ cells from Nkx2-1Cre: RCE (warm colors) and GAD65-GFP (cool colors) lines to the populations of M2R-, CR-, CCK-, VIP-, and CoupTFII- immunolabeled interneurons in CA1 (n= 76, 151, 242, 144, and 216, respectively, in the Nkx2-1Cre: RCE; n =133, 854, 281, 163, and 1376, respectively, in the GAD65-GFP). D, Number of cells co-expressing GFP with M2R, CR, CCK, VIP, and CoupTFII in the GAD65-GFP line presented as a percentage of the total number of GFP+ cells [n= 867, 798, 545, 556, 767, respectively, in the GAD65-GFP; note that group data concerning GFP+ /VIP + cells in GAD65-GFP mice includes counts previously reported as supplemental data in Cea-del Rio et al. (2010)]. Scale bar: 25 µm. Adapted from Tricoire et al. (2011).

(Figure 7). For instance, CCK-BCs, SCAs, and PPAs express the marker CCK, whereas CR and/or VIP label IS cells. Some markers are expressed in both CGE Figure 10 Work flow of scRNA-seq.Figure 11 Immunohistochemical markers primarily andassociated MGE with cells, CGE such-derived as interneurons. SOM and reelin. It should be noticed that some morphologically A, B, Representative identified images illustrating cell types the coexpression have dual of origin GFP ws.ith ForM2R example,(i), CoupTFII 5 (-ii),HT CCK3AR - (iii), VIP (iv), and CR (v) in the Nkx2-1Cre:RCE (A) and GAD65-GFP (B) lines. The filled arrowheads expressing O-LMs originate from CGE, whereas 5-HT3AR-lacking O-LMs has indicate interneurons coexpressing GFP and the indicated marker. The open arrowheads indicate originatecells expressing from the MGE indicated (Chittajallu marker butet notal., GFP. 2013). C, HistogramSimilarly showing, NGFCs the can contribution be divided of GFP+ into cells from Nkx2-1Cre: RCE (warm colors) and GAD65-GFP (cool colors) lines to the populations of MGEM2R-, CR (nNOS+)-, CCK-, VIP and-, and CGE CoupTFII (nNOS- imm-) unolabeled subtypes interneurons based on in theCA1ir (n= nNOS 76, 151, expression 242, 144, and 216, respectively, in the Nkx2-1Cre: RCE; n =133, 854, 281, 163, and 1376, respectively, in the GAD65-GFP). D, Number of cells coexpressing20 GFP with M2R, CR, CCK, VIP, and CoupTFII in the GAD65-GFP line presented as a percentage of the total number of GFP+ cells [n= 867, 798, 545, 556, 767, respectively, in the GAD65-GFP; note that group data concerning GFP+ /VIP + cells in GAD65-GFP mice includes counts previously reported as supplemental data in Cea-del Rio et al. (2010)]. Scale bar: 25 µm. Adapted from Tricoire et al. (2011).

(Tricoire et al., 2010). Nevertheless, with the fast-developing RNA sequencing techniques, more and more numbers of cell-type specific molecules can be identified efficiently, which may update the traditional view of interneuron classification.

2.2 Pipeline of scRNA-seq

Mammalian brain comprises a huge number of neurons conducting diverse functions. Understanding the functional role of single cells is a fundamental goal in neuroscience. Traditionally, neuron subtypes are characterized by their morphologies, connectivity pattern, electrophysiological properties, and the presence of certain molecular markers. All these features are determined by spatially and temporally regulated gene expressions in single cells. Several approaches have been applied to examine gene expression patterns in neurons, such as in situ hybridization, quantitative real-time polymerase chain reaction (RT-PCR), and microarrays. However, these techniques are limited by their depth (the number of detected genes) and dynamic range (the difference of expression levels between genes). The whole transcriptomic analysis based on next generation RNA- sequencing has been proven to be a powerful tool for unveiling subtle differences of gene expression among tissue samples. However, some cell types are hardly accessible due to their low density. Moreover, each single cell is located in a unique micro environment, and thus may display different features of gene expression in response to external stimuli (Pfisterer and Khodosevich, 2017). In addition, even identical cells may display stochastic gene expressions due to random nature of transcription and translation regulations (Raj and van Oudenaarden, 2008). Fortunately, recent technical progress has enabled the sampling and analysing of the whole transcriptome at the single cell level. Single-cell RNA sequencing (scRNA- seq) has been widely applied in brain research, such as for understanding neural stem cell differentiation, identifying novel splicing variants, neuronal subtype classification, investigating brain development regulation by long-non-coding RNAs and developing drugs. The pipeline of scRNA-seq will be introduced in general as follows (Figure 8).

Figure 8 Work flow of scRNA-seq.

Left: Diagram of the work flow of scRNA-seq. Adapted from Potter (2018). Right: Illustration of different approaches to isolate and capture single cells for single-cell RNA Sequencing. Adapted from Hedlund and Deng (2017).

2.2.1 Single cell harvesting

The first step of scRNA-seq is to isolate single cells from brain tissues. A critical factor at this step is to keep the maximum quality of the cells and/or the preservation of mRNA molecules, since RNAs are vulnerable to the degradation by RNases in the environment. The techniques commonly used to achieve this goal include Fluorescence-activated cell sorting (FACS), Magnetic-activated cell sorting (MACS), Laser capture microdissection (LCM), microfluid and manual sorting. FACS is performed on automatic flow cytometers. One of the advantages of FACS is that it allows cells to be pre-labelled with fluorophore-conjugated antibodies or genetic markers and cell sorting is based on different levels of fluorescence intensity.

However, the availability of antibodies and the requirement of large starting volume render this technique unsuitable for rare cell types. MACS is another probe-based sorting technique. Magnetic nanoparticles are coated with antibodies that bind to cell surface antigens, and the cells attached to nanoparticles can then be isolated by a strong magnetic field. Moreover, LCM makes use of lasers to dissect the cells directly from the tissue under the microscope. As the cell can be selected by the researchers according to various morphological criteria, the spatial location of the targeted cells is preserved. However, this semi-automatic approach limited the efficiency of sorting larger number of cells. In contrast, microfluidic chips can isolate and capture cells according to their various physical properties such as size, electric and magnetic properties in nanoliter-scale and high throughput manner. Nevertheless, the application of this technique is hindered by its high cost. Manual sorting is a relatively economical approach. Following tissue digestion, single cells are resuspended in solution and picked up manually using micropipette under a dissection microscope. This technique results in more purified cells, but is less efficient and requires practice.

In particular, an emerging technique called patch-seq utilizes patch-clamp recording to obtain a combination of morphological, electrophysiological and transcriptomic information from single neurons (Cadwell et al., 2016; Fuzik et al., 2016). In this method, neurons of interest are patched in acute brain slices. Electrophysiological properties are recorded after membrane rupture. In parallel, biocytin is diffused into the target cell with the patch solution. At the end of the recording, the cytoplasm containing RNA molecules is aspirated into the patch pipette and then transferred into lysis buffer for subsequent processing. The morphology of processes of recorded cells is recovered through biocytin labelling. This technique enables to access to multiple properties of a single cell. However, caution needs to be taken when searching for high quality cells, and potential contamination from the neighbouring neuropil should be avoided.

2.2.2 cDNA library construction and sequencing

After cell harvesting, total RNAs need to be released from the soma with detergents. Purification of RNA using magnetic beads or nucleic acids after cell lysis is optimal to obtain a high yield. Because it is impossible to sequence RNA directly at present, two subsequent critical steps are needed following RNA purification: the reverse transcription of mRNAs and the amplification of complementary DNAs (cDNAs). Two alternative methods are used: PCR and in vitro transcription. In the case of PCR based approach, the selective reverse transcription of mRNAs is achieved using oligo(dT) primers and engineered viral reverse transcriptase. Since most eukaryotic mRNAs contain a polyA tail, oligo(dT) primers selectively bind to the 3’ ends of mRNAs but avoid tRNAs and rRNAs. After the first strand cDNA synthesis, the second strand cDNA is obtained initially by adding a polyA tail to the 3’ end of first strand cDNAs using terminal deoxynucleotidyl transferase. However, this method often causes the early abolishment of cDNA synthesis and non-specific binding of primers. Later, it has been replaced by a template-switching technique called SMART-seq. In this approach, three to four cytosines instead of polyA, are added into the 3’ end of first strand cDNAs, resulting in full-length coverage of cDNA molecules and higher strand selectivity (Ramskold et al., 2012). PCR based amplification yields high quantity of cDNAs, yet with less fidelity. On the contrary, in vitro transcription amplify cDNAs with high selectivity utilizing T7 RNA polymerase. However, its low-throughput and labour-intensive nature limits its massive application.

After cDNA amplification, samples can be pooled and sequenced simultaneously on the same lane of the sequencer. cDNA fragments from the same sample are tagged with short sequences. After sequencing, samples can be recovered according to the specific tags (demultiplexing). Several sequencing systems such as SOLiD and Illumina are commercially available. Reads are generated during the sequencing that represents small segments of cDNAs. They are aligned to the reference genome using software like Bowtie, STAR, and Kallisto. The expression level of a given

transcript is normalized to the sequencing depth and the length of the molecule. The three commonly used expression metrics are RPKM (reads per kilobase per million mapped reads), FPKM (fragments per kilobase per million mapped reads) and TPM (transcripts per million mapped reads).

2.2.3 Quality control and data processing

Since scRNA-seq is a multi-step process, technical noise can occur at any step and may severely interfere with the result. Thus, quality control is critical prior to further analysis of the sequencing result. Quality control is introduced before and after sequencing. Before sequencing, unhealthy cells can be identified and excluded according to morphological and physiological criteria. Cell samples with low cDNA yield after library construction should be excluded from sequencing. After sequencing, low-quality samples can be readout from summary statistics, such as the percentage of uniquely mapped reads, sequence duplication levels, the percentage of GC content, following demultiplexing and reads alignment. In addition, broken cells often show the downregulation of cytoplasm, metabolism, mitochondrion and membrane genes (Ilicic et al., 2016). Moreover, principle component analysis (PCA) is also used to count out low-quality cells, since good cells are often clustered together, excluding bad cells.

Because gene expression levels are highly variable across samples and sequencing batches, it is important to estimate the absolute number of mRNA molecules in a given sample and normalize expression levels across all samples. Two strategies have been developed to address this issue. Unique molecular identifiers (UMI) are short random DNA sequences that are tagged to each mRNA molecule during reverse transcription. After sequencing, reads that carry the same UMI tag can be traced back to the same original mRNA molecule. This method provides a simple way to count an absolute number of mRNA copies regardless of amplification bias (Kivioja et al., 2011). Alternatively, spike-ins are added into lysis buffers together with cell samples as an internal control. Spike-ins are bacterial nucleotides with known sequences and quantity. When the same volume of spike-ins is added, the 25

absolute number of spike-in molecules is assumed identical across samples. Thus, the ratio between reads mapped to the genome and reads mapped to spike-ins can be used to estimate the amount of mRNAs between samples (External RNA Control Consortium, 2005).

In order to identify cell types based on scRNA-seq data, clustering algorithms are used to group cells with a similar gene expression pattern. Unsupervised clustering methods, such as PCA and hierarchical clustering are often used if prior knowledge of molecular markers is unavailable. T-distributed stochastic neighbor embedding (t- SNE) is applied for dimensional reduction and data visualization. To identify genes that separate different clusters, differential gene expression analysis can be applied. Several tools have been designed for single cell analysis, such as Single-cell differential expression (Kharchenko et al., 2014), Model-based analysis of single-cell transcriptomics (Finak et al., 2015) and Monocle (Trapnell et al., 2014). Moreover, to understand the biological functions of differentially expressed genes, gene set enrichment analysis (GSEA) can be applied. This approach determines whether differentially expressed genes between clusters are preferentially enriched in known gene families (Subramanian et al., 2005). Available software include GSEA (Mootha et al. 2003) DAVID (Huang et al., 2009) and Ingenuity (Krämer et al., 2014).

2.3 Cell-type specific neuronal markers identified with scRNA-seq

The development of novel scRNA-seq techniques as well as data analysing tools during recent years has greatly improved our understanding of neuronal classification based on the levels of specific gene expression. Despite early studies using RT-PCR and microarray, scRNA-seq and its variants have been performed in different brain areas including the brain stem medial vestibular nucleus, olfactory mucosa, dorsal root ganglion, lumbar spinal cord, hypothalamus, amygdala, dorsal striatum, hippocampus, neocortex and human temporal lobe and fetal cortex. Here, we will focus on the latest progress in mouse cortical neuron classification and novel marker identification.

2.3.1 scRNA-seq in cortical neuron classification

The hallmark study of scRNA-seq was performed by Zeisel et al. (2015). They provided the first complete census of single cell transcriptomes of the mouse neocortex and hippocampus. Single cells are obtained from brain tissues of wild type and 5HT3-EGFP mice between postnatal day 21 and 31. Using FACS sorting followed by microfluidic cell capturing, 3005 single transcriptomes are recovered from the primary somatosensory cortex (S1) and hippocampal CA1 area. An unsupervised biclustering method is performed to group both genes and cells. 9 classes and 47 subclasses are identified including neurons and non-neuronal cells. Several cell-type specific markers are reported, such Tbr1 (transcription factor, marks S1 PC), Spink8 (a serine protease, marks CA1 PC) and Pnoc (prepronociceptin, marks interneuron). In particular, S1 PCs can be sub-classified in a layer-dependent manner, with each layer expressing specific markers (Figure 9A). Similarly, two layer-specific markers of CA1 PC are found: Calb1 and Nov. Moreover, S1 and CA1 share the same 16 interneuron subclasses, except that Vip, Penk, Calb2, and Crh-marked cells (interneuron-specific) are concentrated in S1; Lhx6, Reln, and Gabrd-marked cells (Ivy and NGFC) are enriched in CA1 (Figure 9B).

Tasic et al. (2016) characterized cell taxonomy in the adult primary visual cortex with similar sampling methods. The transcriptomes of 1679 cells passed quality control and were processed for clustering. Two types of iterative clustering algorithms are used: PCA and iteratively weighted gene co-expression network analysis. Next, the random forest approach is used to validate clusters generated in the last step. The cells always grouped in the same cluster are deemed ‘core cells’, whereas cells belonging to multiple clusters are referred to as ‘intermediate’ cells. The number of ‘intermediate’ cells between ‘core cell’ groups reflect the continuity of gene expression across cell types. As a result, 23 interneuron groups, 19 glutamatergic groups, and 7 non-neuronal groups are identified. Consistent with previous findings, three major groups of interneurons marked by Vip, Sst, and Pvalb respectively are found, in addition to NGFCs labelled by Ndnf. In particular, 5 sub-clusters are

identified within the Vip group, with each group expressing specific secondary markers. Four of them are located in superficial layers (I to IV), except one located in deep layers (V to VI) is Vip and Gpc3-positive. It is unknown yet whether these transcriptomic groups correspond to morphologically or electrophysiologically identified cell types.

Figure 9 Neuron subclasses in the somatosensory cortex

(A) Subclasses of pyramidal neurons in the somatosensory cortex (S1) identified by BackSPIN clustering. Bar plots show mean expression of selected known and novel markers (error bars show standard deviations). Layer-specific expression shown by in situ hybridization (Allen Brain Atlas). S1PyrL23, layer II-III; S1PyrL4, layer IV; S1PyrL5a, layer Va; S1PyrL5, layer V; S1PyrL6, layer VI; S1PyrL6b, layer VIb; S1PyrDL, deep layers; ClauPyr, claustrum. (B) Identification of interneuron subclasses. Bar plots show selected known and novel markers. Fraction of S1/CA1 cells is depicted at bottom: blue, S1; yellow, CA1; white, flow-sorted Htr3a+ cells from S1. (C) Immunohistochemistry demonstrating the existence and localization of novel PAX6+/5HT3aEGFP+ interneurons, Int11. Bar plots show the layer distribution of these neurons. (D) Intrinsic electrophysiology and morphology of PAX6+ interneurons in S1 layer I, identified by post hoc staining. Adapted from Zeisel et al. (2015).

Figure 12 Schematic of patch-seq approach.Figure 13 Neuron subclasses in the somatosensory cortex

(A) Subclasses of pyramidal neurons in the somatosensory cortex (S1) identified by BackSPIN clustering. Bar plots show mean expression of selected known and novel markers (error bars show

Paul et al. (2017) concentrated on interneurons in motor and somatosensory cortices of adult mice. Intersectional approaches are employed to label interneuron major cell types with one or two markers. Manual sorting accumulated 584 cells of different identities. In search of gene families that best differentiate interneuron cell types, the authors developed the supervised MetaNeighbour algorithm based on Spearman comparison. A single cell network is constructed for a given gene family. Cells that show similar gene expression are linked as neighbours. Genetically labelled cells are first applied to train the network, and then the true identities of some cells are hidden and predicted by the network. The degree of true detection is scored as the mean area under the receiver operator characteristic curve (AUROC; Crow et al., 2018). Among ~620 HGNC (Human Genome Nomenclature Committee) gene families, those involved in synaptic functions receive the highest AUROC score (0.91-0.98). In general, 40 gene families responsible for cell-cell communication, including cell-adhesion, receptors for neurotransmitters and modulators, voltage- gated ion channels (VGICs), regulatory signaling proteins, neuropeptides and vesicle release machinery, and transcription factors are most distinguishing among interneuron cell types.

Besides, Hu et al. (2017) combined sucrose-gradient based single-nucleus purification and droplet microfluidics to separate 18,000 nuclei from single cells of the neocortex. The t-SNE method identified 27 excitatory neuron groups, 7 interneuron groups, and 6 non-neuron groups. Interestingly, activity-dependent cell- type specific expression of non-coding RNAs are detected in some neuron groups. For example, 1700016P03Rik, a non-coding precursor RNA that generates two microRNAs is specifically Transcripted in Ex17 and Ex27 clusters and is associated with neuronal activities (Aten et al., 2016). Non-coding RNAs are also detected in some interneuron and non-neuronal clusters. In addition, activity-associated gene expression also correlates with cell-type specific gene expression. Although only a small percentage, the activity-associated genes are highly expressed in two excitatory clusters Ex17 and Ex24. These findings have raised the possibility of

detecting cellular activity-associated gene expression and non-coding RNA transcription with scRNA-seq.

A recent study has focused on the classification of hippocampal CA1 interneurons. Employing FACS cell sorting and UMI-based tagging methods, Harris et al. (2018) obtained 3663 healthy interneurons from 6 Slc32a1-tdTomato mice ((The Slc32a1 gene encodes for the Solute Carrier Family 32 (GABA Vesicular Transporter) Member 1, thus labels GABAergic interneurons). The clustering algorithm ProMMT (Probabilistic Mixture Modeling for Transcriptomics) identified 10 major clusters and 49 sub-clusters in total. Specifically, a sub-cluster in cluster 1 co-expressing Sst and Npy resembled the hippocampal-septal cells; cluster 6 was considered as Sst- long projecting interneurons that expressed the common marker Ntng1 (Netrin G1); Cluster 10 represent another long-projecting subtype co-expressing Sst and Nos1; Cluster 9 was deemed as IS interneurons that expressed Penk. The putative IS3 cells expressed several markers such as Nos1, Myl1 (Myosin Light Chain 1) Gpd1 (Glycerol-3-Phosphate Dehydrogenase 1), and Igfbp4 (Insulin Like Growth Factor Binding Protein 4). Interestingly, the discrete interneuron classes identified by clustering showed a certain degree of continuous gene expression. Then the latent factor analysis based on the assumption of a continuous distribution of gene expression was applied to further understand the continuity of gene expression across clusters. As with cluster analysis, latent factor analysis also attempts to predict the expression of all genes using only a single variable (the “latent factor”), but now with a continuous (negative binomial distribution) rather than discrete distribution. The authors found that the mean latent factor for each cluster is associated with the subdomains of single cells targeted by axons. For instance, the somatic-targeting cells showed the largest mean latent factor values, whereas IS cells that target other interneurons displayed the smallest values. However, the conventional “cell types” assigned to subgroups identified by clustering were not correlated with cell morphologies and anatomical locations due to tissue dissociation prior to FACS sorting.

2.3.2 The application of patch-seq

Given the high throughput nature of scRNA-seq, discrepancies in cell type classification may be observed compared to traditional methods. Although attempts have been made by studies listed above to combine scRNA-seq data with morphological and electrophysiological properties, these properties cannot be obtained directly from the same cell due to the limitation of sampling methods. Thus, patch-seq was introduced to overcome this issue (Figure 10). Fuzik et al. (2016) recorded the electrophysiological properties of 120 genetically labelled CCK+ and glutamate decarboxylase 1 (GAD67)+ cells in layer I/II. Recorded cells were clustered into 5 groups solely by their membrane properties (I-type 1-5). The morphologies of recorded cells were identified by post hoc reconstruction. Next, 83 cells were aspirated into a pipette following recording for RNA-seq. The molecular markers of CCK+ cells and the correlation between electrophysiological properties and gene expression were evaluated. Intriguingly, the expression level of 24 genes including ion channels, receptors and neuropeptides were closely related to membrane properties recorded in these cells. For instance, the expression of Atp1a1-Atp1b3 (ATPase subunits) transcripts was strongly and positively coupled with resting membrane potentials in all I-types. Moreover, Gria2 (GluR2 subunit) mRNA was preferentially expressed in PCs rather than CCK+ interneurons, whereas Cnr1 (cannabinoid receptor type 1) and Htr3a (5-HT3A receptor) mRNAs showed differential expression levels across I-types. In a parallel study, Cadwell et al. (2016) used 50-250 variable genes obtained from patch cells to predict their electrophysiological properties. The after-hyperpolarization (AHP) amplitude, after- depolarization amplitude, and AP amplitude were the most predictable among all the properties tested. In addition, GSEA analysis indicated that cell adhesion molecules (CAMs) and synaptic regulatory proteins were enriched in bouquet cells in layer I, whereas RNA processing and mitochondrial genes were abundantly expressed in elongated NGFCs. Accordingly, CAMs had also shown higher expression level in SP interneurons than in PCs in the hippocampal CA1 area (Földy et al., 2016). In particular, Sncg and Syt6 genes were selectively expressed in fast-spiking

interneurons recorded near SP. Moreover, VGICs, such as Kcnc1 (Kv3.1) and Kcnc2 (Kv3.2) were more enriched in interneurons compared to PCs, and their expression levels were positively correlated with certain electrophysiological parameters, like AP firing rate, threshold, symmetry, and AHP. Together, these findings highlight the power of scRNA-seq in estimating electrophysiological properties of neurons, especially AP firing properties and AHP, which are directly associated with the cell- type specific expression of ion channel subtypes. Furthermore, the abundance and diversity of CAMs indicate their important yet unknown functions in regulating intercellular communication. Further study is required to evaluate these molecules as cell-type specific markers.

Figure 10 Schematic of patch-seq approach.

Briefly, standard patch-clamp recording is applied; the membrane properties and morphology are obtained. At the end of the recording, cell contents are aspirated into the recording pipette, then processed for RNA-seq. The mRNA expression levels, firing properties and morphologies of single cells are analysed together to classify neurons. Adapted from Cadwell et al. (2015)

InFigure addition 14 MGE to -local and CGEprojecting-dependent interneurons, expression thereof synaptic are longglutamate-range receptors. projectingFigure (LRP 15) Schematic of patch-seq approach. interneurons that target neighbouring or distant areas. For instance, a subgroup of Briefly, standard patch-clamp recording is applied; the membrane properties and morphology are BISobtained. and trilaminarAt the end ofcells the locatedrecording, in cell SO contents of the areCA1 aspirated project into to subiculumthe recording (Ferraguti pipette, then et processed for RNA-seq. The mRNA expression levels, firing properties and morphologies of single cells are analysed together to classify neurons. Adapted33 from Cadwell et al. (2015)

al., 2005). Besides, CA1 interneurons projecting to CA3, DG, medial septum, EC and retrosplenial cortex are also found. Immunohistochemistry results have shown that these cells express SOM, PV, NPY, CB, CR, M2R, metabotropic glutamate receptor 1α (mGluR1α), and Enkephalin. However, none of these markers is unique for LRP cells. Thus, single-cell transcriptomic profiling is required to elucidate cell- type specific markers for hippocampal LRP cells. A scRNA-seq study has identified three subtypes of hippocampal long-projecting cells that showed distinct transcriptomic profiles (Harris et al., 2018). Recently, our laboratory discovered a novel type of VIP+ cell in CA1 that sends their axons to the subiculum. These cells are sparsely distributed in different layers of the CA1 area. They selectively contact CA1 interneurons, whereas innervate both PCs and interneurons in the subiculum. To identify cell-type specific molecular markers of the VIP-LRPs, we performed molecular profiling of common markers with immunohistochemistry. The preliminary results showed that they may express M2R, and they are decorated with mGluR8- expressing boutons. However, none of these markers are unique for VIP-LRPs. Here, we used the patch-seq technique to study the transcriptomic profile and identify molecular makers expressed in this rare cell type with confirmed morphology. We also performed immunohistochemistry to profile commonly used molecular markers in all VIP+ cells in SO of CA1.

3. Glutamatergic synaptic transmission in hippocampal interneurons

As in principal cells, excitatory synaptic transmission in interneurons is primarily mediated by glutamate receptors. Glutamate receptors include ionotropic receptors that mediate fast post-synaptic depolarization and metabotropic receptors that function by triggering intracellular signal pathways. However, the subunit composition and synaptic distribution of these receptors differ substantially between principal cells and interneurons, which may have profound effects on neuronal excitability and plasticity.

3.1 Glutamate receptors in hippocampal interneurons

3.1.1 AMPA receptors

Α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs) are glutamate-gated cation ion channels and mediate fast synaptic transmission in the central nerve system. The opening of the channels allows positive ions to enter the cell and induce excitatory post-synaptic currents (EPSCs). Most endogenous AMPARs form tetramers consisting of four subunits: GluA1-4. Distinct from other subunits, GluA2 contains an edited arginine at the transmembrane domain, which has a remarkable effect on channel conductance and Ca2+ permeability. AMPARs containing GluA2 subunits are impermeable to Ca2+, and their single channel membrane current-voltage (I-V) curve shows a linear relationship (Hollmann et al., 1991). In contrast, GluA2-lacking AMPARs are highly permeable to Ca2+. However, these ion channels are blocked by cytoplasmic polyamines at depolarized membrane potentials, resulting in much smaller currents, as shown by the inward- rectifying I-V relationship. Thus, Ca2+ influx through GluA2- lacking AMPARs can only occur at hyperpolarized membrane potentials (Bowie and Mayer, 1995).

In hippocampal PCs, the most abundant AMPAR forms are GluA1/2 or GluA2/3 heterotetramers that are impermeable to Ca2+ (CI-AMPAR). On contrary, many interneurons essentially express the GluA2-lacking Ca2+ permeable AMPARs (CP- AMPARs). In particular, the expression of AMPAR subunits occurs in the input- specific and cell type-specific manner. For instance, CA3 interneurons receive inputs from local CA3 PCs as well as DG inputs through mossy fibers. CP-AMPARs are found predominately expressed at mossy fiber synapses, whereas synapses formed by CA3 PC associate collaterals are dominated by CI-AMPARs (Galván et al., 2015). Moreover, quantitative studies showed that CA1 PV interneurons express a higher level of GluA4 and a lower level of GluA2 compared to PCs, whereas CCK, NPY and NOS-expressing cells possess lower levels of all AMPAR subunits than PCs (Yamasaki et al., 2016). Matta et al. (2013) summarized the distribution of AMPARs at SC synapses in CA1 MGE and CGE-derived interneurons (Figure 11). Their 35

results show that MGE-derived cells genetically labelled by Nkx2.1 largely contain

CP-AMPARs, whereas 5-HT3ARexpressing CGE-derived cells contain CI-AMPARs. Together, these data imply that the subunit composition of AMPAR in interneurons is determined by their embryonic origin and presynaptic partners.

In addition, CP-AMPARs mediate a form of synaptic plasticity termed “anti-Hebbian” long-term potentiation (LTP), which is typically observed in interneurons. LTP refers to the long-lasting enhancement of synaptic strength and action potential (AP) firing induced by presynaptic and postsynaptic neuronal activities. Traditionally, LTP induction requires the pairing of presynaptic cells firing and postsynaptic membrane depolarization, whereas “anti-Hebbian” LTP induction requires the coincidence of presynaptic stimulation and hyperpolarized postsynaptic membrane potentials. This phenomenon was initially reported in CA3 interneurons (Laezza et al., 1999), and subsequently in CA1 O-LM cells and other PV+ cells (Nissen et al., 2010). In this case, CP-AMPARs play the role of coincidence detector and mediate the Ca2+ influx, which is essential for the induction of long-term plasticity.

3.1.2 NMDA receptors

The N-methyl-D-aspartate receptor (NMDAR) represents another major group of ligand-gated ionotropic glutamate receptor. The activation and deactivation of NMDAR is associated with slower kinetics of EPSCs compared to AMPAR, which allows it to detect and integrate more synaptic inputs in a longer time window. Similarly, NMDARs are composed of four subunits, two homogeneous GluN1s, and two GluN2s. GluN1 is universally expressed in all NMDARs, whereas GluN2s contains four subtypes: GluN2A-D. The opening of NMDAR requires simultaneous binding of glycine with GluN1 and glutamate with GluN2. Moreover, the blocking of NMDARs with bivalent cations, such as Zn2+ and Mg2+ occurs in a voltage-dependent manner, that is hyperpolarization strengthen the blocking, whereas depolarization unleashes the blocking. These unique features place NMDAR in a position of coincidence detector and initiator of synaptic plasticity.

The subunit distribution of NMDAR in hippocampal interneuron subtypes is less clear. Early studies found that NMDAR subunit expression is coupled with the expression of AMPARs. Synapses containing CI-AMPARs have lower 2B subunit, which generates large and fast EPSCs (Lei and McBain, 2002). A recent study shows that MGE-derived cells co-express GluN2B and CP-AMPAR at the same synapses in neonate mice, and the NMDA/AMPA ratio is low (Matta et al. 2013). Moreover, the 2B subunit is replaced by 2A in MGE-derived interneurons during postnatal development as well as after low-frequency stimulation. The 2A subunit displays faster rise and decay kinetics than 2B, indicating that the subunit switching may significantly change the properties of synaptic transmission and dendritic integration in these cells in adulthood. On contrary, CGE-derived cells contain 2B subunits co- expressed with CI-AMPAR and don’t show subunit switching throughout a lifetime. The higher NMDA/AMPA ratio in this cell group indicates a higher contribution of NMDAR in excitatory synapses. In addition, the functional GluN2D subunits and their mRNAs have been detected in hippocampal interneurons, largely in PV-BCs and AACs, but their roles in synaptic transmission are not clear (von Engelhardt et al., 2015).

NMDAR-dependent LTP is induced in CA1 SO and SR interneurons by pairing high- frequency stimulation with postsynaptic depolarization. This type of LTP is similar with that in PCs in terms of induction protocol and expression loci. However, Ca2+/calmodulin-dependent protein kinase IIα (αCaMKII) is required for the induction of LTP in PCs, which is absent in interneurons. A study indicates that other CaMK isoforms may underlie the NMDAR-dependent LTP in interneurons (Lamsa et al., 2007).

Figure 11 MGE- and CGE-dependent expression of synaptic glutamate receptors.

(a,b) MGE- and CGE-derived cohorts of inhibitory interneurons were targeted using hippocampal slices derived from the Nkx2-1-cre:RCE GFP and Htr3a-GFP reporter mouse lines, respectively. Scale bars, 100 μm). (c,d) Top, representative total glutamate receptor (AMPAR and NMDAR)-mediated EPSCs evoked between −60 mV and +40 mV in 20-mV increments triggered by Schaffer collateral stimulation in MGE-derived (c) and CGE-derived (d) interneurons located in CA1 stratum radiatum. Bottom, I-V relationships of the AMPAR- mediated component measured at the time point of the EPSC peak obtained at −60 mV (indicated by dotted lines). Lines are the extrapolated linear fit of the data between −60 mV and 0 mV to reveal deviations from linearity at positive potentials. (e,f) AMPAR-mediated EPSCs before (black) and after (red) application of 2 μM philanthotoxin (PhTx) for representative recordings from MGE-derived (e) and CGE-derived (f) interneurons. EPSCs were evoked as pairs (with a 50-ms interstimulus interval), and PhTx did not alter the PPRs. (g,h) Representative EPSC traces from an MGE-derived (g) and CGE-derived (h) interneuron measured at −60 mV and +40 mV to extract the NMDAR-to-AMPAR amplitude ratio. (i–k) Summary plots of the AMPAR rectification index (RI) (i; MGE, 79 cells from 79 slices from 57 mice; CGE, 59 cells from 59 slices from 51 mice), philanthotoxin sensitivity (j; MGE, 8 cells from 8 slices from 5 mice; CGE, 8 cells from 8 slices from 6 mice) and NMDAR-to-AMPAR ratios (k; MGE, 85 cells from 85 slices from 57 mice; CGE, 88 cells from 88 slices from 51 mice) measured from all MGE- and CGE-derived interneurons regardless of their developmental age or anatomical identity. MGE-derived interneurons typically had AMPARs with significantly lower rectification indices (P = 2.73 × 10−25, degrees of freedom (d.f.) = 136, t = −12.9), higher philanthotoxin sensitivity (P = 7.9 × 10−5, d.f. = 14, t = −5.5) and lower NMDA-to-AMPA ratios (P = 1.87 × 10−20, d.f. = 171, t = −10.6) than their CGE-derived counterparts (P values were determined by unpaired t tests). Group data are presented as mean ± s.e.m, with results from individual experiments represented by open circles. Adapted from Matta et al. (2013).

38 Figure 16 Similar current–voltage curves for isolated AMPA EPSCs and different voltage dependence of the isolated NMDA response in the SC and PP (equal to TA) inputsFigure 17 MGE- and CGE-dependent expression of synaptic glutamate receptors.

3.1.3 Kainate and metabotropic glutamate receptors

Kainate receptors (KR) are ionotropic receptors named after its agonist Kainic acid. Five subunits of KR (GluK1-5) form hetero or homotetramers. GluK1 and GluK2 are the most abundant isoforms in interneurons. KRs-mediate slow-decaying synaptic transmission, and account for a small proportion of evoked and spontaneous EPSCs (sEPSCs) in PCs. However, in CA1 SO interneurons, spontaneous and evoked EPSC are mediated by KR along or by KR/AMPAR in combination. The spontaneous events mediated by KRs represent 31% of total sEPSCs, implying a larger contribution of KR in glutamatergic transmission in SO interneurons (Goldin et al., 2007). Events involving KR are also found in SR and SLM interneurons. In addition, KRs can be expressed at both presynaptic and postsynaptic sites. Presynaptically expressed KRs regulate GABA release at inhibitory synapses between interneurons and PCs. mGluRs are a subfamily of G protein-coupled receptors. Depending on the receptor structure and G protein binding properties, they can be categorized into 3 groups:

Group I including mGluR1 and 5 are coupled with Gq/G11 and subsequently activate phospholipase C signaling. Group II consists of mGluR 2 and 3, Group III comprises mGluR 4, 6, 7, and 8. They are coupled with Gi/o and inhibit the activation of adenylyl cyclase. In particular, the mGluR1α isoform is found in SOM+, VIP+, and CCK+ interneurons, as well as in PV-BCs (Ferraguti et al., 2004; Hainmüller et al., 2014). The activation of Group I mGluRs depolarizes postsynaptic membrane by mediating Ca2+ influx through voltage-gated Ca2+ channels (VGCCs) or intracellular Ca2+ store release. In particular, the activation of mGluR1α leads to the Ca2+ influx passing through transient receptor potential (TRP) channels in SO interneurons, whereas the activation of mGluR5 induces Ca2+ elevation through intracellular Ca2+ release (Topolnik et al., 2006). Besides, group I mGluRs modulate Na+, K+ and non-selective cation channels, which may have an impact on dendritic integration. Moreover, Group I mGluRs are involved in Hebbian and anti-Hebbian LTP in hippocampal 39

interneurons (Perez et al., 2001; Topolnik et al., 2005; Lamsa et al., 2007). On the other hand, presynapticially expressed Group II and III mGluRs regulate GABA release at the active zone of axonal terminals by controlling Ca2+ influx through P/Q- or N-type (VGCCs).

3.2 Comparison of TA and SC synapses in the CA1 region

As we discussed in chapter 1.1.2, TA and SC pathways represent the two major excitatory afferents to the CA1 region, the direct cortical input from entorhinal excitatory cells and the indirect local input from CA3 pyramidal cells. Morphological studies showed that the major targets of TA and SC pathways are CA1 PCs. 92% of SC synapses in CA1 SO and SR make contact with PCs, whereas 7.1% of the SC synapses contact aspiny interneurons. In contrast, the distribution of TA synapses displays laminar differences. In SLM where the majority of TA projections are located, more than 90% of TA synapses are made on PC dendritic spines and shafts. In SO, SP and SR, TA projections have a higher preference for interneurons (21%; Takács et al., 2011). Intriguingly, a group of CB- expressing PCs in EC layer II predominately target CA1 interneurons located in SLM (Kitamura et al., 2014), indicating a cell-type specific connection.

As we described in chapter one, different dendritic compartments of PCs are controlled by different types of interneurons. In particular, distal dendrites of PC receive feed-forward and feedback inhibitory controls. The feed-forward inhibition is mediated by interneurons located in SLM or SR, such as NGFCs, PPAs, and other CCK+ cells that receive input from the TA projection. The feedback control is mediated by O-LMs located in SO, which are excited by the alveus from CA1 PCs and in turn inhibit distal apical dendrites. Accordingly, electrical stimulation of the TA pathway induces moderate excitatory postsynaptic potentials (EPSPs) followed by fast and slow inhibitory postsynaptic potentials (IPSPs) in 90% CA1 PCs (Empson and Heinemann, 1995). The remaining 10% of recorded cells showing EPSPs only are presumably interneurons. Pharmacological studies indicate that the TA-PC

EPSPs are mediated by NMDARs and AMPARs (Figure 12), whereas inhibitory transmissions in PCs are mediated by GABAARs and GABABRs. Focal application of bicuculline, a GABAAR antagonist reveals that the TA-activated inhibition preferentially targets PC dendrites in SR. Moreover, pairing TA and SC stimulation with short time intervals (<30 ms) with intact inhibition leads to the sublinear summation of somatic responses, whereas simultaneous activation of TA and SC synapses in the presence of GABAR blockers results in a superlinear summation (Enoki et al., 2001). On the other hand, optogenetic activation of O-LMs induces inhibition of PC distal dendrites and facilitate SC-PC transmission (Leão et al., 2012). Together, these studies highlight the role of feed-forward and feedback inhibition in pathway interaction and dendritic integration in apical dendrites of PCs.

Moreover, several pathway-specific properties are found between TA and SC synapses in CA1 PCs. It is well known that the density of APMAR and hyperpolarization-activated cation channels (HCN) increases with distance in the apical dendritic trunk (100~350 µm from the soma), which compensates the dendritic filtering effect. However, the NMDAR mediated current remains unchanged within this range. In contrast, the NMDA/AMPA ratio increases significantly at the dendritic tuft region, where TA pathway terminates, whereas the densities of HCN and AMPAR are similar between tuft and apical dendritic areas (Bittner et al., 2012). Further studies show that in the adult rats, the NMDA component of TA-PC EPSCs exhibits a larger degree of inward rectification at depolarized voltage levels than SC- PC EPSCs, while SC-PC EPSCs are more sensitive to the GluN2B subunit antagonist ifenprodil. As a result, the ifenprodil insensitive component shows faster decay kinetics at both pathways (Arrigoni and Greene, 2004), indicating a larger contribution of 2A subunit to the synaptic transmission at TA-synapses.

Figure 12 Similar current–voltage curves for isolated AMPA EPSCs and different voltage dependence of the isolated NMDA response in the SC and PP (equal to TA) inputs.

A, Voltage dependence (the average of 10 cells) of the SC EPSC area and a linear ﬁt (gray line) using SEM as a weight (r=0.99; p <0.001). B, Voltage dependence (10 cells) of the PP EPSC area and a linear ﬁt (gray line; r = 0.99; p < 0.001). C, The average (10 traces) SC EPSC at -65 and +40 mV. D, The average (n=10) PP EPSC from the same cell. Vertical lines mark the rising phase of EPSC; horizontal arrows mark the half-width. Calibration is shown in the middle. E, Voltage dependence of the isolated NMDA EPSC area in the SC and PP inputs (average of 10 cells). F, The same data normalized by a maximal inward current in each cell. G, The SC and PP NMDA EPSC (average of 5 traces) at two voltages (-20 and +60 mV) in standard ACSF. H, The SC and PP NMDA EPSC (average of 5 traces) at -20 and +60 mV in ACSF without Mg2+. The early component of the SC response is drawn in black; the late polysynaptic component is marked in gray. I, The summated SC and PP AMPA EPSC to four stimuli at 200 Hz (average of 5 traces, artifact is truncated) at -20 and +60 mV in standard ACSF containing 100 µM ±APV and 200 µM cyclothiazide. For easier visualization, the EPSC traces in C–E were scaled by the peak currents at -20 mV. The individual calibration bars (50 pA) are shown to the right of each pair of traces; the common time scale is 50 ms (C, top right). Adapted from Otmakhova et al. (2002).

As discussed before, TA and SC innervate both PCs and interneurons. Tracing studies indicate that interneurons with somas located near to the interface of SLM and SR and dendrites extending in SLM and SR are targeted by both TA and SC projections. Compared to SC synapses, synaptic transmission at TA synapses remains unstudied in most interneuron cell types. Price et al. (2005) compared the synaptic transmissions evoked by TA and SC stimulation in NGFCs. TA stimulation elicits short latency excitatory EPSCs that are mediated solely by AMPA receptors at -70 mV, whereas NMDA component is observed at +40 mV in the presence of the AMPAR/KR antagonist 6,7-dinitroquinoxaline-2,3(1H,4H)-dione (DNQX). Similar voltage-dependence of NMDA component is also observed in SC-NGFC synapses. Moreover, optogenetic experiments show that MEC-NGFC synaptic transmission displays increased NMDA/AMPA ratio throughout development, and a switching from GluN2B to GluN2A subunit in CGE-derived NGFCs. This observation is in contrast with the SC synapses in other CGE-derived interneurons, where no subunit switching is found during development (Matta et al. 2013). Together, TA and SC synapses display major differences in NMDA/AMPA ratio and NMDA components, which may influence dendritic integration and plasticity along the apical dendrites of PCs and interneurons.

Synaptic plasticity can be induced at both TA and SC synapses in PCs. Based on the Hebb’s postulation, pairing presynaptic stimulation and postsynaptic depolarization provided by back-propagating APs (bAPs) robustly evokes LTP at SC-PC synapses in CA1. However, bAP is inefficient in invading distal dendrites of PCs that receive TA input, suggesting alternative mechanisms for inducing LTP in distal dendrites. Indeed, tetanic stimulation is required to elicit TA-LTP in CA1 PCs in the absence of bAP, whereas low frequency stimulation (1Hz) induces long-term depression (LTD) of synaptic transmission at TA synapses. Further studies show that TA-LTP requires the initial activation of NMDAR and subsequent opening of VGCCs. The Ca2+ influx produced by these receptors induces Ca2+ spikes at distal

dendrites, which in turn activate multiple synapses and facilitate LTP induction at distal sites (Golding et al., 2002). In addition, slow inhibition mediated by GABABRs is also needed for the induction of TA-LTP in PCs. Furthermore, the interaction between TA and SC pathways in inducing synaptic plasticity has been reported in PCs. TA stimulation either potentiates (SLM stimulation arrives 20 ms before SR stimulation) or depresses (SR stimulation precedes SLM stimulation 20 ms) SC synapses depending on the relative timing of the two inputs. Such heterosynaptic LTP requires the opening of NMDARs, AMPARs, as well as Ca2+ release from intracellular Ca2+ stores. However, little is known about the synaptic plasticity at TA synapses in hippocampal interneurons. In this study, we compared the excitatory synaptic transmission at SC-IS3 and TA-IS3 synapses. We also explored the glutamate receptors that are involved in excitatory transmissions in these cells.

4. Temporal and spatial summation on dendrites of interneurons

4.1 Short-term plasticity at excitatory synapses of interneurons

In addition to long-term plasticity, the synaptic transmission can be modified on a faster time scale. Short-term plasticity (STP) refers to the change of synaptic strength within the period ranging from a few milliseconds to tens of seconds (Zucker and Regehr, 2002). Practically, STP can be observed when a train of stimulating pulses is delivered. The postsynaptic responses induced by the second and/or subsequent pulses can be either facilitated (short-term facilitation, STF) or depressed (short-term depression, STD), compared with the first response. STP has been observed at excitatory synapses between PCs or between PC and interneurons in the neocortex as well as in the hippocampus.

STP has a profound role in controlling information flow within the short time scale. Specifically, enhanced synaptic transmission in response to repetitive activities may remove the Mg2+ blockage of NMDAR. The activation of NMDAR may facilitate synaptic integration and the induction of LTP/LTD. Also, STF quickly depolarizes the membrane, which facilitates AP firing of the postsynaptic cells. Furthermore, STF is 44

more likely to be induced by high frequency burst firing pattern of the presynaptic cell, which endows STF with the function of the high-pass filter. On the contrary, STD narrows the time window for synaptic integration. Only strictly timed presynaptic pulses can induce AP in the postsynaptic cells. In addition, STD diminishes the probability of AP firing during repetitive presynaptic activities, which serves as a low- pass filter.

The direction of STP depends on both presynpatic and postsynaptic cells. In the neocortex, layer II/III and layer V PC to PC connection displays only STD. PCs also make synapses onto SOM+ bitufted interneurons and PV+ interneurons. PC to bitufted SOM+ synapses show STF, whereas PC to PV+ synapses show STD (Reyes et al., 1998; Buchanan et al., 2012). In the hippocampus, interneurons often show mixed facilitation/depression (Figure 13). For instance, 75% of O-LMs have facilitating excitatory synapses in SO, whereas the rest of them have depressing synapses in SO (Losonczy et al., 2002). Furthermore, the majority of BCs and BISs show combined facilitation and depression. In SR, the excitatory synapses on PCs have stronger paired-pulse facilitation than that of interneurons. Low frequency stimulation (5, 10, 20 Hz) pulses induced weak facilitation in interneurons, and high frequency stimulation (50 Hz) induced depression (Sun et al., 2005). In NGFCs located in SLM, 40 Hz stimulation of the TA pathway induced STD in some cells, whereas other displayed initial STP followed by STD (Price et al., 2005). Together, the STP of excitatory synapses is target-selective and depend on stimulation frequencies.

What are the synaptic mechanisms that determine the degree and direction of STP? It has been shown that both presynaptic and postsynaptic mechanisms are involved. First of all, several presynaptic factors should be taken into account: initial release probability of presynaptic vesicles, distribution of Ca2+ and K+ channels, the distance between Ca2+ entry site and the vesicle releasing complex, and the distribution of presynaptic NMDARs. Among these factors, the initial release probability plays an essential role. It is generally accepted that synapses with low initial release

probability show STF and high initial release probability synapses display STD. In addition, evidence has shown that the amplitude of Ca2+ transient is positively associated with the release probability at axonal boutons formed on layer II/III PCs, indicating the potential role of presynaptic Ca2+ dynamics in controlling the release probability (Koester and Sakmann, 2000).

Figure 13 Short-term plasticity at synapses with different I–V properties.

A, EPSCs from a representative OA-IN showing inward rectification (RI = 0.18; left) and facilitation during repetitive stimulation (5 pulses at 20 Hz) (average of 15 traces; right). B, graph of EPSC amplitude during the train (normalized to EPSC1) for all cells with inward/intermediate rectification (n=12), showing facilitation at these synapses. C, summary histogram of synapses with inward/intermediate rectification, showing most often facilitation (Fac.) versus depression (Dep.). D, example of EPSCs with linear I–V relation (RI =1.4; left) which showed initial facilitation followed by depression during repetitive stimulation (right). E, graph of EPSC amplitude during the train for all OA-INs with linear I–V relation (n = 6), illustrating the initial facilitation and subsequent depression at these synapses. F, summary histogram of synapses with linear I–V relation, showing usually depression over facilitation at the end of the train. Adapted from Croce et al. (2010).

Furthermore, several genes have been identified as key regulators that determine the release probability of presynaptic vesicles. For instance, the transsynaptic regulator Elfn1 is typically expressed in O-LMs, which regulates release probability 46

by directly interacting with presynaptic mGluR7 or GluK2. The knockdown of Elfn1 gene results in an increase of release probability and diminution of facilitation in O- LM excitatory synapses formed by CA1 PCs. In contrast, the up regulation of Elfn1 expression switched the STD in PV cells to STF (Sylwestrak and Ghosh, 2012).

Besides, CAMs also play a role in regulating release probability, although the mechanisms are not fully understood. Cadherins are molecules that attached to both pre and postsynaptic membranes. They are actively involved in synapse formation, plasticity and vesicle recycling (Bozdagi et al., 2004; Jüngling et al., 2006). The manipulation of N-cadherin expression alters the release probability in a similar way as Elfn1. Neuroligins represent another type of adhesion molecule that mainly attach to postsynaptic membrane proteins like PSD-95. However, it retrogradely regulates vesicle release at presynaptic terminals by contacting the presynaptic membrane protein neurexins.

Despite various presynaptic mechanisms that regulate the STP, STP in interneurons can also have a postsynaptic origin. As described in chapter 3.1.1, the polyamine blockage of CP-AMPAR leads to small currents at the depolarized level. However, polyamine can be removed from the receptor during repetitive stimulation. In multipolar interneurons of neocortical layer II/III, paired-pulse stimulation either reduces the degree of STD or induces STF at a certain frequency (Rozov and Burnashev, 1999). On the other hand, CGE-derived VIP interneurons typically lack CP-AMPAR, and thus may have STP primarily expressed at presynaptic terminals, although the mechanism of STP at excitatory synapses made onto VIP interneurons remains unstudied.

4.2 Dendritic spatial integration in PCs and interneurons

4.2.1 Dendritic properties in PCs

Neurons communicate with each other by AP firing. The AP firing of a single PC is influenced differentially by tens of thousands of excitatory and inhibitory synapses located on its soma and dendrites. Understanding the mechanisms of dendritic integration and its impact on AP initiation is a key mission in neuroscience. Numerus studies have focused on the dendritic properties of PCs in the hippocampus and neocortex. Several critical factors should be considered when discussing dendritic integration in PCs. First of all, dendritic morphologies may have an impact on synaptic integration. PC dendritic trees are generally composed of an apical dendrite and several basal dendrites. The apical dendrite bifurcates at a certain distance from the soma and forms a tuft at the distal part. Several oblique branches are connected with the main trunk. These oblique dendrites are distinct compartments in synaptic integration, since active currents integrate efficiently in a single branch (Losonczy and Magee, 2006).

Secondly, as indicated in cable theory, passive membrane properties shape the AP initiation in different ways. The more hyperpolarized the Vm is, the higher the driving force is needed to induce the AP. Also, input resistance (Rin) is determined by the number of active channels on the membrane. High Rin means that a lower current is required to recruit the cell. Next, the membrane capacitance reflects the surface area of the cell, which implies the number of inputs one cell can receive. Thus, a cell with higher Cm may be more excitable than others.

Thirdly, the impact of a given synaptic input on AP firing depends on its distance from the soma. Dendritic patch-clamp recordings have proved that distal synaptic events attenuate dramatically when they approach the soma (Figure 14B) (Nevian et al., 2007; Larkum et al., 2009). However, distal dendrites developed several mechanisms to compensate for the filtering effect, such as synaptic scaling (Magee and Cook, 2000). One of these mechanisms refers to dendritic spike (Figure 14C).

How dendritic spikes can be generated? PC dendrites are well equipped with VGICs. For instance, the density of A-type K+ channels increases from the soma to distal dendrites, whereas it’s evenly distributed on dendrites of neocortical layer V PCs (Bekkers, 2000). The HCNs display similar gradients in both cell types (Magee, 1998; Berger et al., 2001). In contrast, the Na+ channels are uniformly distributed in cortical PCs (Bittner et al., 2012). Additionally, different types of VGCCs are also discovered in cortical PCs. These channels are distributed throughout somas and dendrites, with higher densities in specific dendritic domains (Westenbroek et al., 1990).

The existence of VGICs endows PC dendrites the capability of generating spikes. The Na+ spikes are short-lasting (< 5ms) and are mediated by Na+ channels. The initiation of these spikes is independent of somatic AP. Nevertheless, they attenuate sharply when they propagate from the small-diameter terminal to wider branches due to a higher impedance. Ca2+ spikes are usually mediated by the L-type Ca2+ channels, which last longer (> 10 ms) and have a strong impact on soma. The third type of dendritic spike is named the NMDA spike. The duration of NMDA spike is even longer, and often generating plateau-like potentials in dendrites. However, NMDAR is a ligand-gated ion channel, and the generation of NMDA spike is restricted to synaptic regions. Together, these long-lasting dendritic spikes provide a long time window that allows synaptic events to summate spatially and temporally and overcome passive membrane filtering effect along the dendrite. Moreover, dendritic spike facilitates the induction of long-term plasticity at distal dendrites (Golding et al., 2002).

In PCs, APs are primarily initiated at the axonal initial segments. However, early studies show that the propagation of AP is bidirectional: they back-propagate into the dendrites (Figure 14A). bAP is also observed in awake mice, with equal or larger amplitude than in vitro bAPs, despite background synaptic activities (Bereshpolova et al., 2007). The distance that bAP can travel depends on dendritic morphology and the distribution of VGICs. The high density of K+ and HCN channels at distal dendrites prevents bAP from propagating beyond the proximal region (Hoffman et

al., 1997). Nevertheless, bAP interacts with synaptic events and promotes the generation of dendritic spikes (Larkum et al., 2001). Properly timed bAP and presynaptic APs at excitatory synapses induce spike timing-dependent plasticity.

The properties of PC dendrites described above enable them to conduct arithmetical summation of excitatory inputs. Both linear and non-linear summation of synaptic inputs has been observed in PC dendrites. Early studies using stimulation pipettes and microiontophoresis to induce subthreshold events on a different location of CA1 PC dendrites indicate that the linearity of excitatory input summation depends on resting membrane potentials, event amplitude, VGICs, the distance between stimulation sites and the amount of inhibition. Hyperpolarized membrane, small event amplitudes, the opening of VGICs and the small distance between stimulation sites improves non-linearity of inputs summation. Switching the balance of excitation/inhibition may even change the direction of non-linearity (Langmoen and Andersen, 1983).

The technique of two-photon glutamate uncaging allows the activation of synaptic inputs with high spatial resolution stimulations, as the stimulating area can be focused on individual dendritic spines. The Ca2+ dynamics can be monitored simultaneously with two-photon imaging. Araya et al. (2006) reported that EPSPs evoked at individual spines on basal dendrites of layer V PCs summate linearly despite the amplitude and distance between spines, whereas uncaging on dendritic shafts induced EPSPs to summate sublinearly. Near synchronous (0.1 ms interval) uncaging on spines of CA1 PC oblique dendrites results in initial linear summation of EPSPs. As the number of activated spine increases (6-20 spines), dendritic Na+ spike occurs, which in turn induce supralinear summation of somatic EPSPs (Figure 14D). Synchronously recorded Ca2+ transients show a similar transition of linearity and are mediated by NMDAR and VGCCs (Losonczy and Magee, 2006). Together, these studies indicate that dendritic spines are functional compartments for dendritic integration and Ca2+ influx.

Figure 14 Dendritic excitability of pyramidal neurons.

a. A backpropagating action potential recorded simultaneously from the soma and dendrite of a layer II/III pyramidal neuron in a slice from rat somatosensory cortex. Note the attenuation of the dendritically recorded action potential. b. Somatic and dendritic action potentials (left-hand image and lower plot; right-hand image and upper plot, respectively) recorded from rat somatosensory cortex in vivo. The images are two-dimensional projections of two-photon scans through the neurons. Note that the dendritically recorded action potentials have smaller amplitudes, suggestive of the attenuated backpropagating action potentials that are observed in vitro. c. Dendritically initiated spikes observed in a CA1 pyramidal neuron in response to stimulation of the perforant path (Temporoammonic pathway). In the top pair of traces (thick trace, dendritic membrane voltage; thin trace, somatic membrane voltage) the dendritic spike (*) fails to evoke a somatic action potential: it evokes only a spikelet. Usually the spikelet (an attenuated dendritic spike) would trigger an action potential in the axon, but in this case action- potential initiation was prevented by local application of tetrodotoxin (TTX) to the soma, the axon and the proximal dendrites. In the middle and bottom pairs of traces, which were recorded without TTX application, a small dendritic spike is virtually undetectable in the soma (middle pair) but a larger dendritic spike evokes a full-sized action potential (bottom pair; the action potential is truncated). d. Glutamate uncaging evokes local dendritic spikes in a CA1 pyramidal neuron. Uncaging was performed at several spots that were either distributed along the dendritic branch (red circles and lines) or clustered in one spot (green circles and lines). The voltage responses recorded in the soma exhibited a nonlinear increase when multiple spots were activated in rapid succession. Associated with this nonlinear increase in voltage was a spike-like increase in dV/dt, which suggests that a dendritic spike was generated. At the soma, however, the spike increased the response by only approximately 2 mV. The threshold for evoking a dendritic spike was lower for the distributed input than for the clustered input. EPSP, excitatory postsynaptic potential; L1, layer 1. Parts a and b reproduced, with permission, from Waters et al. (2003). Part c reproduced, with permission, from Golding et al. (2002). Part d reproduced, with permission, from Losonczy and Magee (2006). Adapted from Spruston (2008).

in vivoin vitro

4.2.2 Dendritic integration in interneurons

Compared to PCs, interneurons are less abundant yet more diverse, which complicate their dendritic integration properties and AP firing initiation. Interneurons differ from PCs in many aspects. Firstly, the axons of interneurons often originate directly from their dendrites, in contrast to the soma origin of PC axons. The distance from interneuron axons to soma can be as long as few hundred micrometers. This

implies that dendritic events can be converted into AP firing in axons without passing through the soma. Electron microscopy has shown that the density of excitatory and inhibitory synapses is higher in dendrites than in soma or axonal initial segment (Gulyás et al., 1999), further supporting this proposition.

Secondly, in addition to chemical synapses, cortical interneurons are often coupled electrically via gap junctions, which is uncommon in PCs. Electrical coupling can be found between the same interneuron subtypes, or even across different cell types (Simon et al., 2005). The fast current transduction through gap junctions enables signal synchronization among a large group of cells, and the propagation of network oscillations (Galarreta and Hestrin, 1999). Also, gap junctions substantially affect dendritic integration in interneurons. Experimental data have shown that gap junctions account for ~50% of the total membrane conductance (Deans et al., 2001; Cruikshank et al., 2004). As predicted by cable theory, the presence of electrical coupling decreases input resistance and time constant of EPSPs, resulting in less excitability and shorter time window for dendritic integration at the level of individual cells. However, the spreading of currents enhance the excitability of cell ensembles interconnected with gap junctions, regulating network activity in a larger spatial scale.

Recent progress in understanding dendritic integration focused on a few cell types. Thus, other properties will be discussed in a cell-type specific manner. PV-BC is a large interneuronal group in the cortex. The high frequency firing pattern allows it to transfer information with high temporal precision and to synchronize network activity, which requires fast input-output conversion in PV-BCs. Several morphological and physiological properties account for this unique role of PV-BCs. Firstly, PV-BCs lack dendritic spines, their excitatory synapses are located directly on dendritic shafts, leading to fast propagation of synaptic events. Moreover, evidence has shown that excitatory synapses in PV-BCs and dendritic-targeting cells are distributed in clusters. These clusters form “hot-spots” (~10 µm) on the dendrites. Using 3D two- photon Ca2+ imaging, Katona et al., (2011) and Chiovini et al., (2014) found that excitatory inputs on the hot spot integrate supralinearly and triggers dendritic spikes

when a certain number of inputs are activated simultaneously. These dendritic spikes are initially generated by CP-AMPAR and NMDAR activation, and subsequently, activate the L-type Ca2+ channels in the lateral area. In turn, Ca2+ spikes propagate into neighbouring segments and generate ripple-like oscillations on the dendrites. These findings indicate that hot spots are functional compartments for dendritic integration in PV-BCs and other interneurons. In contrast, when the single synapse is activated, the Ca2+ signal is restricted to 1 µm around the synapse. This fast influx and extrusion of Ca2+ are mediated by CP-AMPAR and Na+/Ca2+ exchanger, respectively (Goldberg et al., 2003a). Thus, distinct mechanisms are involved in subthreshold events at single synapse of interneurons.

Secondly, the Rm in PV-BCs is much lower than that in PCs. This feature may reduce the probability of AP firing induced by background dendritic events and enhance the impact of synchronized inputs. However, further evidence shows that the Rm shows a gradient on the dendrites, with the Rm at distal dendrites nine times higher than that at proximal ones (Stuart and Spruston, 1998). This feature indicates that the distal and proximal dendrites of PV-BCs are distinct computational compartments. Distal dendrites facilitate input integration, whereas proximal dendrites improve temporal precision.

Thirdly, multiple studies show that bAP can hardly propagate into the dendrites of PV-BCs (Figure 15B) (Goldberg et al., 2003a; Evstratova et al., 2011; Camiré and Topolnik, 2014). This is due to the low density of voltage-gated Na+ channels and highly expressed Kv3 K+ channels. In fact, the Na+ current density decreases dramatically from the soma, whereas the density of Kv3 K+ channels only declines slightly at distal dendrites (Figure 15 C). Pharmacological studies indicate that the presence of Kv3 K+ channels reduces the decay time of EPSPs and promotes coincidence detection in PV-BCs (Figure 15D) (Hu et al., 2010). In addition, HCN channels are also expressed in PV-BCs. However, the details of their distributions are unknown.

SOM+ cells are a large group of dendrite-targeting cells in the cortex. The dendritic properties of two types of SOM+ cells, the Martinotti cell in the neocortex and O-LMs in the hippocampus have been characterized. Martinotti cells have multipolar or bitufted dendrites spreading in different layers. Their axons project to layer I where distal dendrites of PCs are located. In contrast, O-LMs have horizontally oriented dendrites restricted in SO (Figure 15E). Unlike PV+ cells, the dendrites of SOM+ cells contain spines, although detailed studies are still required to understand their functional roles. In addition, the Rm in SOM+ cells is higher than that of PV+ cells (Gloveli et al., 2005), meaning that they are more excitable than PV+ cells. Moreover, the density of voltage-gated Na+ channels is two times higher than that of PCs. Indeed, bAPs has been observed to propagate throughout the distal dendrites of cortical Martinotti cells. The presence of Na+ channels also facilitates the propagation of dendritic spikes (Figure 15F) (Goldberg et al., 2004). Furthermore, HCN channels have been found in Martinotti cells and O-LMs with less details about their distributions. Sekulić et al. (2015) estimated that HCN channels decrease with distance from the soma along the dendrites of O-LMs. Together, these dendritic features allow SOM+ cells to integrate synaptic inputs within a large time window and subsequently induce APs with relatively weak and sparse presynaptic inputs.

CGE-derived interneurons account for a large cell assembly. However, little is known about their dendritic integration properties. Goldberg et al. (2003b) studied the Ca2+ dynamics evoked by bAPs and dendritic stimulation in the dendrites of layer II/III irregular-spiking CR+ cells. bAPs induced Ca2+ influx through VGCCs and subsequently activated internal Ca2+ stores. However, bAP is unable to propagate to distal dendrites despite that VGCCs are present throughout the dendritic tree. This is due to the rapid activation of A-type K+ channels, which quickly reduce the amplitude of bAPs. In addition, pairing dendritic stimulation with a bAP induced supralinear summation of Ca2+ close to the stimulation site.

The distribution of VGICs on dendrites of IS3 cells is initially studied by Guet- McCreight et al. (2016). The Ca2+ transients evoked by bAPs decease to ~40 % at

~150 µm from the soma, indicating that bAPs have less influence on distal dendrites. Also, immunohistochemistry shows that Kv3.1 but not Kv2.1 channels are expressed at least in the soma and proximal dendrites of IS3 cells. The distribution of other VGICs at distal dendrites of IS3s is unknown. However, computational modelling showed that this restricted distribution of VGICs at soma and proximal dendrites better captures the experimental features of these cells, such as the strong attenuation of bAPs along the dendrites (Guet-McCreight et al., 2016).

Figure 15 Biophysical properties of PV and O-LM neuron dendrites.

(A) Top, reconstruction of a PV-expressing basket cell in the hippocampal dentate gyrus (soma and dendrites in black; axon in red). Bottom, fast-spiking phenotype of a hippocampal PV- expressing basket cell in vitro, in response to a 1-s-long current pulse at the soma. (B) Top, confocal image of simultaneous patch-clamp recording from the soma and dendrite of a PV neuron in vitro. Bottom, strong attenuation of back-propagating action potential amplitude in PV neuron dendrites. (C) Left panels, current responses to voltage steps using outside-out patches from the soma (black) or the dendrites (red). Right panels, Na+ and K+ current densities in PV neuron somata (black) and dendrites (red). (D) Top, blocking Kv3 channels with tetraethylammonium (TEA, blue trace) reduced the decay speed of synaptic potentials in PV neurons. Middle: Cable modeling demonstrating local synaptic activation of Kv3-type K+ channels in PV-expressing interneuron dendrites (a synapse was activated in the apical dendrite, arrow). Pseudocolor code indicates the activated K+ conductance. Bottom: dendritic Kv3-type channels reduce the half-duration of EPSPs in a PV Neuron cable model. Data in (A) is adapted from Nörenberg et al. (2010), and data in (B–D) is adapted from Hu et al. (2010). (E) Camera lucida reconstruction of a biocytin-filled oriens-alveus interneuron. Axon, red; soma and dendrites, black. Str. l.-m., stratum lacunosum-moleculare; str. rad., stratum radiatum; str. pyr., stratum pyramidale; str. ori., stratum oriens. (F) Na+ (solid symbols and continuous lines) and K+(open symbols and dashed lines) current densities (I Na and I K, respectively), calculated from maximal inward or outward current at −10 mV (15) and plotted against distance from the center of the soma (positive values indicate the axon-bearing dendrite). Four patches in which neither Na+ nor K+ current could be evoked (probably due to vesicle formation) were excluded. (G) Ratio of dendritic and somatic action potential (AP) amplitudes plotted versus the distance from the soma for the axon-bearing dendrite (open circles and continuous line at 22° to 25°C; triangles at 33° to 36°C) and the axon-lacking dendrite (squares and dashed line at 22° to 25°C; inverted triangles at 33° to 36°C). In the control, the ratio is close to unity, almost independent of location. In TTX, the ratio is smaller than unity and decreases with distance (solid circles and lower continuous line at 22° to 25°C). Points at a distance of 0 μm represent double-soma recordings. Mean somatic action potential amplitude is 115 ± 2 mV at 22° to 25°C and 98 ± 3 mV at 33° to 36°C. E-G: Adapted from Martina et al. (2000).

Figure 18 Biophysical properties of PV and O-LM neuron dendrites.

(A) Top, reconstruction of a PV-expressing basket cell in the hippocampal dentate gyrus (soma and dendrites in black; axon in red). Bottom, fast-spiking phenotype of a hippocampal PV- expressing basket cell in vitro, in response to a 1-s-long current pulse at the soma. (B) Top, confocal image of simultaneous patch-clamp recording from the soma and dendrite of a PV neuron in vitro. Bottom, strong attenuation of backpagating action potential amplitude in PV neuron dendrites. (C) Left panels, current responses to voltage steps using outside-out patches from the soma (black) or the dendrites (red). Right panels, Na+ and K+ current densities in PV neuron somata (black) and dendrites (red). (D) Top, blocking Kv3 channels with tetraethylammonium (TEA, blue trace) reduced the decay speed of synaptic potentials in PV neurons. Middle: Cable modeling demonstrating local synaptic activation of Kv3-type K+ channels in PV-expressing interneuron dendrites (a synapse was activated in the apical dendrite, arrow). Pseudocolor code indicates the activated K+ conductance. Bottom: dendritic Kv3-type channels reduce the half-duration of EPSPs in a PV Neuron cable model. Data in (A) is adapted from Nörenberg et al. (2010), and data in (B–D) is adapted from Hu et al. (2010). (E) Camera lucida reconstruction of a biocytin-filled oriens-alveus interneuron. Axon, red; soma and dendrites, black. Str. l.-m., stratum lacunosum-moleculare; str. rad., stratum radiatum; str. pyr., stratum pyramidale; str. ori., stratum oriens. (F) Na+ (solid symbols and continuous lines) and K+(open symbols and dashed lines) current densities (I 58Na and I K, respectively), calculated from maximal inward or outward current at −10 mV (15) and plotted against distance from the center of the soma + (positive values indicate the axon-bearing dendrite). Four patches in which neither Na nor K+ current could be evoked (probably due to vesicle formation) were excluded. (G) Ratio of dendritic and somatic action potential (AP) amplitudes plotted versus the distance from the soma for the axon-bearing dendrite (open circles and continuous line at 22° to 25°C; triangles at 33° to 36°C) and the axon-lacking dendrite (squares and dashed line at 22° to 25°C; inverted triangles at 33° to 36°C). In the control, the ratio is close to unity, almost independent of location. In TTX,

General hypothesis 1

To summarize, recent studies have shown that hippocampal VIP+ cells form heterogenous group. Electrophysiological experiments revealed that VIP-LRPs contacted local CA1 interneurons and both excitatory cells and interneurons in the neighbouring structure subiculum . In addition, in vivo recordings showed that the activity of these cells was supressed during theta oscillations and during ripples, but that they were active during quiet wakefulness. Our first general hypothesis is that VIP-LRPs control the information flow along the hippocampal-subicular axis to regulate the memory formation during quiet states. Testing this hypothesis requires the cell-type specific manipulation of VIP-LRPs in behaving animals. Thus, the first objective of my PhD thesis was to study molecular markers expressed by VIP-LRPs that could be used for cell-specific targeting of these cells.

The specific aims were:

1. To identify molecular markers of VIP-LRPs using immunohistochemistry in brain slices obtained from different transgenic mouse models (VIP-eGFP & VIP-tdTomato).

2. To study the transcriptomic profile of morphologically identified VIP-LRPs with scRNA-seq.

3. To compare the gene expression in VIP-LRP cell with CA1 GABAergic neurons, neocortical VIP:CR and VIP:CCK interneurons.

General hypothesis 2

The second part of my thesis has focused on IS3 cells. These cells have been well characterized at anatomical and molecular levels (Ascády et al., 1996a). In addition, their post-synaptic targets were examined by Chamberland et al. (2010) and Tyan et al. (2014). However, the properties of the excitatory inputs converging onto these cells and important for their recruitment in vivo remained unknown. Based on the morphological properties of dendrites of these cells, I tested the hypothesis that they can be driven via the SC and TA inputs with the last one playing a more important role in IS3 cell recruitment due to high density of dendritic arborization in SLM layer.

The specific aims were:

1. To study the synaptic properties of TA and SC excitatory inputs to IS3 cells.

1.1 To examine the synaptic mechanisms (ionotropic glutamate receptors) underlying the TA-IS3 transmission.

1.2 To examine the short-term plasticity at TA and SC-IS3 synapses.

2.To study the activity of IS3 cells during network oscillations in awake animals.

Chapter 2 Integrated article 1: Connectivity and network state- dependent recruitment of long-range VIP-GABAergic neurons in the mouse hippocampus

Résumé

Les interneurones GABAergiques de l'hippocampe assurent la coordination locale et à distance des neurones dans les zones fonctionnellement connectées. Les interneurones exprimant le peptide intestinal vasoactif (VIP+) occupent une place particulière dans le circuit, car nombre d’entre eux sont spécialisés dans l’innervation des cellules GABAergiques ce qui permet une désinhibition du réseau. Dans la région CA1 de l'hippocampe, les interneurones VIP+ spécifiques aux interneurones ciblent les interneurones locaux. Ici, nous avons découvert un nouveau type de neurone VIP + dont l'axone innerve la région CA1 et se projette également dans le subiculum (VIP-LRP). Les VIP-LRPs présentent des propriétés moléculaires spécifiques et ciblent les interneurones dans la région CA1 mais aussi bien des interneurones que des cellules pyramidales localisées dans le subiculum. Ces neurones sont interconnectés par des jonctions lacunaires, mais ont exprimé, in vitro, un couplage dispersé de signaux. Chez les souris éveillées, les VIP-LRPs ont diminué leur activité pendant les oscillations thêta, mais ont été plus actives pendant les périodes calmes d’éveil, mais non couplées aux ondes à pics aigus. Ensemble, les résultats fournissent de nouvelles preuves de la diversité moléculaire des interneurones VIP+ et de la spécialisation fonctionnelle dans le contrôle des ensembles cellulaires le long de l'axe hippocampo-subiculaire.

Abstract

GABAergic interneurons in the hippocampus provide for local and long-distance coordination of neurons in functionally connected areas. Vasoactive intestinal peptide-expressing (VIP+) interneurons occupy a distinct niche in circuitry as many of them specialize in innervating GABAergic cells, thus providing network disinhibition. In the CA1 hippocampus, VIP+ interneuron-selective cells target local interneurons. Here, we discover a type of VIP+ neuron whose axon innervates CA1 and also projects to the subiculum (VIP-LRPs). VIP-LRPs show specific molecular properties and target interneurons within the CA1 area but both interneurons and pyramidal cells within subiculum. They are interconnected through gap junctions but demonstrated sparse spike coupling in vitro. In awake mice, VIP-LRPs decrease their activity during theta-run epochs and were more active during quiet wakefulness but not coupled to sharp-wave ripples. Together, the data provide evidence for VIP interneuron molecular diversity and functional specialization in controlling cell ensembles along the hippocampo-subicular axis.

Key words: Hippocampal formation, circuit, connectome, network oscillations, disinhibition, long-range GABAergic cell, subiculum.

Introduction

Understanding brain computations during different cognitive states requires identifying cell types, their connectivity motifs and the recruitment patterns under different behavioural conditions. GABAergic inhibitory neurons play a pivotal role in cortical computations through gain control, sensory tuning and oscillatory binding of cell ensembles1–4. However, understanding cortical inhibition has been a challenging task as this process is executed through a diverse group of local and long-range projecting (LRP) GABAergic neurons5. Many types of GABAergic cells that have been identified by earlier investigations remain functionally uncharacterized. This is

especially the case for sparse cell types, which represent a minority of the cortical neuronal population and, therefore, have not been frequently sampled in blind electrophysiological recordings. In particular, until recently, very little has been known about the functional organization of GABAergic cell types that are specialized in the selective coordination of inhibitory interneurons. These so-called interneuron- selective (IS) cells express vasoactive intestinal peptide (VIP) alone or in combination with calretinin6,7. They originate from the caudal ganglionic eminence and are the last cells to integrate into the cortical habitat8,9, where they innervate many different types of local interneurons, including the somatostatin (SOM+), calbindin (CB+), parvalbumin (PV+), VIP (VIP+) and calretinin (CR+) expressing GABAergic cells6,7,10,11. Development of novel transgenic and optogenetic technologies allowed to investigate how these cells can coordinate the operation of cortical microcircuits12–17. A common finding between different cortical regions is that VIP+ IS cells suppress some local interneuron activity during complex behaviours, including visual processing12,14,16, locomotion13 and reward-associated learning17, thus leading to network disinhibition. However, similar to other GABAergic cells, VIP+ neurons are diverse in properties6,7,18–20 and, likely, in circuit function. Yet, no attempt has been made for a detailed physiological and functional analysis of morphologically defined subtypes of VIP+ interneurons.

The hippocampal CA1 inhibitory circuitry can be considered one of the best characterized so far. Indeed, over the last three decades, the findings of multiple laboratories have culminated in a detailed wiring diagram of hippocampal CA1 GABAergic circuitry, with at least 21 inhibitory cell types identified to date21. Hippocampal CA1 VIP+ interneurons constitute two functionally different GABAergic cell populations: basket cells (BCs22) and IS interneurons (IS2 and IS3 cells6), which can modulate the activity of principal cells (PCs) or of different types of CA1 interneurons with a different degree of preference23,24. VIP+ BCs (VIP-BCs) can co- express cholecystokinin (CCK) and, in addition to targeting PC somata, can contact PV-positive BCs, indicating that VIP-BCs can exert both inhibitory and disinhibitory network influences23. In contrast, the VIP+ IS interneurons prefer to contact inhibitory 63

interneurons6, and modulate interneuron firing properties24. Although disinhibition can be a common mechanism of hippocampal computations necessary for the induction of synaptic plasticity and memory trace formation and consolidation25, current findings indicate that its effect is mostly local due to the local innervation of hippocampal inhibitory microcircuits through VIP+ interneurons24. Interestingly, anatomical data point to the existence of long-range circuit elements that could account for cross-regional disinhibition between the hippocampus and functionally connected areas: CA1 SOM- or muscarinic receptor 2 (M2R)-expressing GABAergic cells innervate hippocampal inhibitory interneurons and can project to several cortical and sub-cortical areas, including the rhinal and retrosplenial cortices, subiculum (SUB) and medial septum (MS)26–30. Despite the considerable recent interest in LRP GABAergic neurons, very little is currently known about the connectivity and function of these cells during different network states in awake animals.

Here, we reveal a subtype of VIP-expressing LRP (VIP-LRP) GABAergic neuron that exhibits a specific molecular profile and innervates, in addition to the hippocampal CA1, the SUB, with region-specific target preference. Functionally, VIP-LRP cells correspond to theta-off cells31,32 as they decrease their activity during theta-run epochs associated with locomotion and exhibit high activity during quiet wakefulness. The identification of this circuit element reveals an additional mechanism for the behaviour- and network-state-dependent inter-regional coordination of activity within the hippocampal formation.

Results

VIP-LRP neuron in the CA1 hippocampus

To characterize the electrophysiological and morphological properties of VIP+ interneurons in the hippocampal CA1 area, we first performed patch-clamp recordings from VIP+ cells in acute slices obtained from VIP-eGFP mice (Fig. 1a; Supplementary Figs. 1; 2g, i; the expression of GFP in this mouse strain was

characterized previously24). Following biocytin labelling, 97 VIP-GFP+ interneurons were visualized and identified as BCs, IS3 cells or as novel LRP neurons (VIP-LRP; Fig. 1b; Supplementary Fig. 1b; Supplementary Table 1). The VIP-LRP neurons typically occurred at the oriens/alveus (O/A) border and had horizontal sparsely spiny dendrites, which were mostly restricted to the stratum oriens (Fig. 1b). Their axon formed a local arbor in O/A and extended slightly into strata pyramidale (PYR) and RAD of the CA1 region. The main axon was partially myelinated and travelled outside the CA1, giving rise to a large axon cloud in the proximal SUB (Fig. 1b; n = 40 cells out of 78 biocytin-filled O/A VIP-GFP+ interneurons). Thus, in contrast to VIP-BCs and to IS3 cells, the axon of VIP-LRP neurons occupied two major areas: CA1 O/A and SUB (Fig. 1b, e). The total length of the axonal arbor in a 300-µm slice was between 7704 and 36,167 µm (no shrinkage correction). The axon length occupying the CA1 vs SUB was 2000–20,000 (median ± SD: 10,184 ± 5002) and 1000–26,000 (median ± SD: 5041 ± 6740) µm, respectively; the large variability was likely due to a different degree of the axon preservation in slices (Fig. 1e; n = 10 cells). These cells showed a regularly spiking firing pattern and a membrane potential ‘sag’ in response to a hyperpolarizing step to −100 mV (Fig. 1c; Supplementary Fig. 1a). Furthermore, their intrinsic membrane properties were similar to those of VIP-BCs [except for the sag and the fast afterhyperpolarization (AHP) amplitude] but differed in many parameters from IS3 cells (Supplementary Table 1). To provide additional evidence for the presence of SUB-projecting VIP+ neurons in the CA1 O/A, we injected red RetroBeads in the SUB of VIP-eGFP mice. In addition to pyramidal cells, bead labelled VIP-GFP-positive neurons were detected in the CA1 O/A area (Fig. 1d, left), thus confirming that a population of O/A VIP+ neurons sends long-range axons to the SUB. As to their molecular profile, all VIP-LRP cells tested with an axon reaching SUB were immunopositive for muscarinic receptor 2 (M2R; n = 7/7 cells; Fig. 1d, right), thus identifying the M2R as an additional molecular marker of VIP- LRPs. Furthermore, a large fraction of M2R+/VIP-GFP+ cells co-expressed CB (29/53 cells; Supplementary Fig. 2a, i) but were negative for CCK (Supplementary Fig. 2d, i), nitric oxide synthase (NOS; Supplementary Fig. 2e, i), CR (Supplementary Fig. 2f) and SOM (Supplementary Fig. 2h, i). This was in contrast to VIP-BCs and 65

IS3 cells, which co-expressed CCK or CR, respectively, and were negative for M2R and CB (Supplementary Fig. 2b, c). Overall, ~50% of VIP-GFP+ cells in the CA1 O/A of VIP-eGFP mouse were co-expressing M2R (Supplementary Fig. 2i), corresponding to the VIP-LRP population. In the rat, trilaminar cells projecting to the subiculum are rich in M2R in the somato-dendritic membrane and are innervated by presynaptic mGluR8-positive terminals33. We tested if the VIP-GFP-M2R-positive cells in the mouse received mGluR8+ input and found that most VIP-GFP-M2R- positive cells in O/A were decorated by mGluR8+ terminals, some of which were themselves VIP-positive as in the rat (Fig. 1f). Some M2R+ neurons in O/A were not immunoreactive for VIP and GFP, and not all VIP-GFP+ neurons showed M2R immunoreactivity (Supplementary Fig. 2i), pointing to additional molecular diversity within M2R+ and VIP+ neuronal populations.

Local connectivity of VIP-LRP cells

To determine the VIP-LRP physiological function, we next examined its local connectivity using simultaneous paired recordings and electron microscopic (EM) analysis (Fig. 2). Dual whole-cell patch clamp recordings showed that out of 118 attempts, 33 pairs of VIP-LRPs and CA1 O/A interneurons were connected synaptically, and no connection was found with CA1 PCs (Fig. 2i). Among the VIP- LRP targets, we identified different types of dendritic inhibitory cells (DT-INs), such as O-LM (Fig. 2a–e) and bistratified cells (BIS; Fig. 2f), and the perisomatic terminating interneurons (ST-INs), such as BCs (Fig. 2g). As O-LM cells were the most frequent target, we characterized the VIP-LRP to O-LM cell connection in more detail (Fig. 2a–e). The VIP-LRP synapses occurred on dendritic shafts of OLMs (distance from soma: 63.4 ± 19.8 µm, n = 7 pairs; Fig. 2a), had mean unitary inhibitory postsynaptic current (uIPSC) amplitude of 16.3 ± 2.4 pA (0 mV holding potential) and a failure rate of 60.1 ± 4.1% (n = 8 pairs). uIPSCs had slow kinetics consistent with the dendritic location of synapses (Fig. 2d). During repetitive VIP- LRP firing, uIPSC showed no change at 10–50 Hz but summated efficiently at 100 Hz (Fig. 2c, e). Apart from the uIPSC amplitude, which was higher in BISs (32.1 ± 66

6.2 pA, n = 5 pairs; Fig. 2j), the properties of VIP-LRP synapses were similar among different postsynaptic targets (Supplementary Table 3), and, also, did not differ significantly from synapses made by IS3 cells on O/A interneurons24.

To further validate the results of paired recordings, EM analysis of 39 synaptic junctions within the CA1 area made by two VIP-LRPs filled with biocytin was performed. The data showed that interneuron dendrites were frequent synaptic targets of VIP-LRPs. Of the 18 postsynaptic dendrites tested from the targets of one VIP-LRP, 16 originated from interneurons and 2 were unidentified (Fig. 2h, top). The other VIP-LRP cell made synapses with 6 interneuron dendrites and 15 unidentified dendrites, some of which emitted spines (Fig. 2h, bottom), suggesting that spiny interneurons or PCs could be among the VIP-LRP targets. Taken together, these data indicate that, locally, VIP-LRPs prefer to target interneurons and constitute a novel type of IS cells in the hippocampus.

To investigate the possible connectivity within the VIP-LRP population, we performed dual whole-cell patch-clamp recordings from VIP-GFP+ pairs (Fig. 3a). Our data showed that out of 36 attempts, 16 VIP-LRP pairs were connected through symmetric gap junctions with a coupling coefficient of 0.11 ± 0.1 (Fig. 3d, e), and one pair was connected synaptically. Electrotonic coupling between VIP-LRPs was blocked by the gap-junction inhibitor mefloquine34 (to 21.05 ± 11.5% of control, n = 8 pairs; p < 0.01; paired t test; Fig. 3d, e) or the broad-spectrum gap-junction blocker carbenoxolone (to 14.9 ± 10.3% of control, n = 6 pairs; p < 0.01; paired t test; Fig. 3e). Furthermore, the electrotonic signal conduction exhibited low-pass filter properties. As such, fast action potentials (APs) generated in cell 1 were strongly attenuated in cell 2, whereas the slow AHPs were better conducted, leading to a substantial hyperpolarization of cell 2 (Fig. 3b, c, g, bottom right insets). We then examined how a sinusoidal excitatory input modulated at theta-like frequency can be integrated by the electrically coupled VIP-LRP neurons (Fig. 3f–h). We found that, when both cells were kept at rest, a sinusoidal input applied to cell 1 induced subthreshold synchronous fluctuations of the membrane potential of cell 2, but was

not able to drive its firing (Fig. 3f). When cell 2 was slightly depolarized to allow for spontaneous firing, the two VIPLRPs could occasionally fire together but their coupling remained weak, and most spikes occurred asynchronously due to the AHP associated inhibition of cell 2 (Fig. 3g). The voltage fluctuation in cell 2 was significantly higher when spikes were not generated in cell 1 [voltage peak without AP in cell 1: 8.5 ± 0.1 mV (inset black trace) vs voltage peak with AP in cell 1: 6.9 ± 0.1 mV (inset red trace), n = 6 pairs; p < 0.01; paired t test; Fig. 3g, bottom right inset]. When both cells received a synchronous theta-modulated excitatory input, their firing increased (to 112%, n = 6 pairs; Fig. 3h), but the spike synchrony remained weak (Fig. 3h, bottom). Together, these data indicate that electrotonic coupling between VIP-LRPs is unlikely to synchronize their recruitment in response to the theta-like input. Whether this may be the case at a different firing frequency35,36, activity-dependent state of the gap junctions or network size will need to be explored using computational modelling37.

Distant connectivity of VIP-LRP cells

To determine the distant targets of VIP-LRPs in SUB, we conducted single-cell two -photon glutamate uncaging-based mapping of connections by combining the photoactivation of CA1 O/A VIP-GFP+ cells and patch-clamp recordings of interneuron and PC targets (Fig. 4). As reported previously38, two-photon uncaging (730 nm, 20–30 mW/180 ms laser pulses) of locally delivered MNI-Glu [micropressure pulses (5 psi, 5 ms) via a glass pipette with tip diameter of 2–3 μm positioned ∼10 μm above the cell of interest] triggered single spikes in VIP-GFP+ interneurons, resulting in fast uncaging-evoked IPSCs (glu-IPSCs; delay onset: 4.8 ± 1.8 ms) in target cells in case of connection (Fig. 4a–e). Depending on the number of VIP-GFP+ cells per slice, this approach allowed us the testing of several VIP- GFP+ connections to a given target (1–5; Fig. 4a–e). First, consistent with our findings using paired recordings and ultrastructural analysis (Fig. 2), in the CA1 area, VIP-LRPs were connected to O/A interneurons (n = 9 connections out of 31 tested/11 cells; Fig. 4a, f). Glu-IPSCs were not detected in CA1 PCs (n = 0 connections out of

18 tested/10 cells; Fig. 4b, f), confirming the interneuron preference of VIP-LRPs. In addition, post hoc immunohistochemical examination of connected neurons revealed that VIP-GFP+ cells that were connected to O/A interneurons were positive for M2R (Fig. 4g), consistent with the neurochemical profile of VIP-LRPs (Figs. 1d and 2b). Surprisingly however, in the SUB, VIP-LRPs innervated both interneurons (n = 8 connections out of 45 tested/26 cells; Fig. 4c, f) and PCs (n = 11 connections out of 37 tested/24 cells; similar between all CA1 and SUB targets (Fig. 4h, bottom), Fig. 4d, f) identified based on their dendritic and axonal although some target-specific differences were observed within properties. These data highlight the region-specific target pre- a population of CA1 interneurons (Fig. 2j; Supplementary ference of VIP- LRPs (Fig. 4h, top). The glu-IPSC amplitude was Table 3).

Activity of VIP-LRP cells in awake mice

To understand the functional role of VIP-LRP cells, we performed in vivo two-photon calcium (Ca2+) imaging of VIP+ interneuron activity in head-restrained awake mice running on a treadmill (Fig. 5a–d)39. The Cre-dependent viral vector adeno- associated virus (AAV)1. Syn.Flex.GCaMP6f.WPRE.SV40 was delivered to the CA1 hippocampus of VIP-Cre mice to express Ca2+-sensitive protein GCaMP6f selectively in VIP+ neurons (Fig. 5b). The immunohistochemical analysis of VIP+ O/A neurons in VIP-Cre mice (Supplementary Fig. 3a–d) confirmed that VIP/M2R- coexpressing cells were present in the CA1 O/A (albeit at a significantly lower fraction when compared with VIP-eGFP mice: 7% in VIP-Cre vs 50% in VIP-GFP out of total VIP+ O/A cells; Supplementary Figs. 2i and 3e–h), likely due to the mouse strain differences (CD1 for VIP-eGFP vs C57BL/6J for VIP-Cre mice)40–42. The M2R+ VIP O/A cells in VIP-Cre mice exhibited virus-driven GCaMP6f expression, showed a normal morphological appearance (Fig. 5e) and were examined for the activity dependent recruitment during different behavioural states.

During the experiment, habituated mice showed spontaneous alternations in their behaviour between locomotion (running speed median and interquartile range: 8.8 and 5–30 cm/s; Fig. 5c), immobility and flickering39, a transitional state associated 69

with brief random movements (Fig. 5a, n = 6 mice). As flickering periods were very short (<1 s) and variable in occurrence, they were excluded from further analysis. Among 54 CA1 VIP+ interneurons imaged, 28 VIP+ cells were located within O/A and analyzed in detail during 409 locomotion and 356 immobility periods (10–15 periods/cell pooled from two independent imaging sessions of 5 min each/ 3 mice; Table S4). Consistent with previous observations in neocortical circuits13, as a population, the majority of VIP+ O/A cells were more active during animal locomotion (Supplementary Table 4). By applying Otsu’s method43 to somatic Ca2+-activity, we could segregate these cells into two distinct sub-types (Fig. 5d): type I VIP cells (n = 8), on average, exhibited higher somatic Ca2+-signals during immobility than during locomotion (locomotion: 30.0 ± 13.2% ΔF/F vs immobility: 100.4 ± 28.0% ΔF/F; p < 0.05; Mann–Whitney test; Fig. 5d), while type II cells (n = 20) were more active during locomotion than during quiet states (locomotion: 94.9 ± 10.1% ΔF/F vs immobility: 31.7 ± 4.0% ΔF/F; p < 0.001; Mann–Whitney test; Fig. 5d). Post hoc immunohistochemical analysis of recorded neurons (6 out of 28 VIP+ O/A interneurons recorded in vivo were found after brain resectioning and processed for markers; Fig. 5e) revealed that type I cells which were more active during quiet states express M2R but not CR (2 cells out of 2 tested) and, therefore, correspond to VIP- LRP neurons (Figs. 5e and 6d). The type II VIP cells tested were negative for both M2R and CR (4 cells out of 4 tested; Fig. 6e) and, therefore, could correspond to VIP-BCs or other VIP+ interneurons.

To further validate these data, we performed Ca2+ imaging of CA1 O/A VIP+ interneurons that were retrogradely labelled with red RetroBeads injected in SUB in addition to Cre-driven GCaMP6f expression (Fig. 5f; n = 5 cells/2 mice). Local delivery of a small volume of RetroBeads in SUB (25 nL; Fig. 5f, left top) resulted in labelling of CA1 PCs as well as of a few O/A interneurons, which is consistent with previous observations of the several distinct types of SUB-projecting GABAergic cells in the CA1 O/A27,29,33 (Fig. 5f, left bottom). Out of total 11 interneurons labelled with RetroBeads, 5 were VIP+ cells, which thus corresponded to VIP-LRPs (Fig. 5f, right). We confirmed that bead-labelled interneurons exhibit normal physiological 70

properties using patch-clamp current clamp recordings in vitro (Supplementary Fig. 4a, 4b). In vivo, Ca2+ transients detected in retrogradely labelled VIP-LRPs had larger peak amplitude during immobility than during locomotion periods (immobility: 55.5 ± 6.7% ΔF/F, n = 81 periods/5 cells; locomotion: 21.0 ± 3.2% ΔF/F, n = 63 periods/5 cells; p < 0.01, one-way ANOVA followed by Tukey’s test; Fig. 5g; Supplementary Fig. 4c, d). Taken together, these data indicate that VIP-LRPs are more active at rest than during locomotion.

As VIP-LRP cells showed different levels of somatic activity during behavioural states, we next examined their recruitment during network oscillations through parallel recordings of the local field potential (LFP) from the contralateral CA1 hippocampus39,44. The locomotion periods were associated with prominent theta oscillations (7.1 ± 0.3 Hz; n = 6 mice; Fig. 6a–c), while high-frequency ripples were observed during the animal quiet state (144.5 ± 2.6 Hz; n = 6 mice; Fig. 6a–c). For VIP-LRP population (type I VIP+ cells), we combined the cells segregated based on their behaviour activity pattern (n = 6 cells with LFP recorded; Fig. 5d) with those that were labelled retrogradely (n = 5; Fig. 5f, g). In these cells, the peak somatic Ca2+ signal decreased significantly, from 80.80 ± 9.03% ΔF/F during quiet states to 38.61 ± 5.15% ΔF/F during theta-run epochs (p < 0.001; Mann–Whitney test; 154 stationary and 116 theta-run periods, n = 11 cells; Fig. 6b), indicating that VIP-LRPs decrease their activity during theta oscillations. In contrast, the M2R-/CR- type II VIP+ cells increased their activity from 31.8 ± 3.3% ΔF/F during quiet state to 96.8 ± 6.1% ΔF/F during theta-run episodes (p < 0.001; Mann–Whitney test; 162 stationary and 190 theta-run periods, n = 14 cells; Fig. 6c), pointing to the on-going recruitment of these cells during theta. As the quiet state in awake rodents is associated with recurrent ripple oscillations2,45,46 and these events may co-occur in the two hippocampi44,47–49 (but see recent findings for rats during sleep50), we investigated the potential recruitment of VIP+ neurons during ripple episodes. The VIP-LRPs showed no change in somatic activity in relation to ripple episodes (57.2 ± 34.1% ΔF/ F before vs 52.7 ± 29.3% ΔF/F after ripple episode, n = 56 episodes/5 cells; p > 0.05; Mann–Whitney test; Fig. 6b). Similar data was obtained for the type II M2R- 71

/CR- VIP+ cells (22.2 ± 4.5% ΔF/F before vs 23.5 ± 6.2% ΔF/F after ripple episode, n = 13; p > 0.05; Mann–Whitney test; Fig. 6c), although strong ripple coupling was observed in some M2R-/CR- VIP-expressing cells located within PYR (Supplementary Fig. 5). Collectively, these data reveal the preferential recruitment of VIP-LRPs during quiet wakefulness and the suppression in their activity during theta-run epochs, pointing also to functional segregation of VIP interneuron subtypes during different behavioural and network states.

Diversity of subiculum-projecting VIP-LRPs

How diverse is the population of SUB-projecting VIP-LRPs? To address this question, we conducted retrograde tracing by (1) injecting a small volume (20 nL) of Cre- dependent hEf1-LS1L-GFP herpes simplex virus (HSV) into the SUB of VIP-Cre;Ai9 mice (Fig. 7a, b) or (2) using a combinatorial VIP-LRP targeting via injection of retrograde Cav2-Cre into SUB of VIP-flp;Ai65 mice (Fig. 7a, c). The reporter Ai65D (B6;129S-Gt(ROSA)26Sortm65.1(CAG-tdTomato)Hze/J) mouse line expresses tdTomato under the control of cre and flp. Crossing this reporter line with VIP-flp mice and injecting the Cav2-Cre in the SUB allows selective targeting of VIP+ SUB- projecting neurons. With both strategies, prior calibration experiments were performed to control for the virus spread from SUB to CA1 (Fig. 7a; see Methods for details). In addition to a small population of local SUB VIP+ cells (Fig. 7b, left), CA1 VIP+ interneurons with somata located within O/A, PYR, RAD or LM were sparsely labelled (VIP-Cre;Ai9 mice + HSV-GFP: 6.7 ± 0.4% of total CA1 VIP+ population, 103/1522 cells from 3 animals, Fig. 7a, b; VIP-flp;Ai65 + Cav2-Cre: 7.3 ± 0.6% of total CA1 VIP+ population, 100/1364 cells from 3 animals, Fig. 7a, c). O/A VIP-LRPs made ~20% of the total VIP-LRP population (22 out of 103 cells in VIP-Cre;Ai9 mice). Consistent with our findings of a low fraction of M2R+ VIP O/A cells in VIP-Cre mice (Supplementary Fig. 3e), some O/A VIP-LRPs labelled with an HSV-GFP in VIP-Cre;Ai9 mice co-expressed M2R (Fig. 7d, left; 2 out of 13 O/A VIP-LRPs tested). In addition, those with soma located within PYR, RAD or LM co-expressed CR (10 out of 31 cells tested; Fig. 7d, middle) or 72

proenkephalin (Penk, 2 out of 28 cells tested; Fig. 7d, right), revealing further molecular diversity within the SUB-projecting VIP-LRP population.

To examine the local connectivity of the entire SUB-projecting VIP+ population in the CA1 area, we next employed a ChR2assisted circuit mapping approach based on the antidromic activation of VIP-LRP cells through wide-field stimulation of their axons in the SUB of VIP-Cre;Ai32 mice (Fig. 7e, f). Importantly, we found no evidence for the existence of subiculo-hippocampal VIP+ projecting neurons that could be activated by light stimulation in SUB and contact CA1 interneurons (Supplementary Fig. 6). In addition, no antidromic spikes were evoked in VIPBCs (n = 3) or IS3 cells (n = 3) by light stimulation in SUB (data not shown), thus validating our photostimulation approach for antidromic activation of hippocampo-subicular VIP-LRP neurons.

In total, 59 CA1 interneurons and 49 CA1 PCs were examined as potential VIP-LRP targets. In CA1 O/A, 33 out of 36 interneurons tested were connected and 26 were visualized with biocytin (Fig. 7g), including O-LM (n = 21) and BIS (n = 7) cells (Fig. 7g). Moreover, a putative LRP cell, that was negative for SOM and M2R, with a partially myelinated axon travelling outside the hippocampus, received inhibitory input from VIPLRP neurons (Supplementary Fig. 7a). In CA1 RAD, 12 out of 23 interneurons tested received input from VIP-LRPs, including CCK-expressing Schaffer-collateral-associated cells (n = 5) and BCs (n = 4; Supplementary Fig. 7b, c). The amplitude of light evoked IPSCs (lIPSCs) was substantially higher in O/A (88. 6 ± 18.3 pA, n = 26) than in RAD interneurons (30.5 ± 7.6 pA, n = 8; p < 0.01, Mann– Whitney test), with the SOM+ O-LM and BIS cells demonstrating the largest amplitude of lIPSCs. In contrast, out of 49 PCs tested, only 2 cells with soma within O/A received input from VIP-LRPs (Supplementary Fig. 7d, e). These data further support the preferential interneuron innervation by VIPLRP population, revealing the CA1 circuit disinhibition as a local function of SUB-projecting VIP-LRPs.

Subicular targets were also examined using a ChR2-assisted mapping strategy but through wide-field photostimulation in CA1 and patch-clamp recordings in subiculum. 73

Out of 43 attempts, 15 subicular neurons were connected to VIP-LRPs (Fig. 7h, i). Morphological analysis of cells filled with biocytin showed that SUB targets of VIP- LRPs included both PCs (n = 10 out of 32 attempts; Fig. 7h, left; i) and interneurons (n = 5 out of 11 attempts; one was identified as SOM+, Fig. 7h, right; i), thus pointing to a shared VIP-LRP input by PCs and interneurons in the distant projection area (Fig. 7j). Taken together, these data highlight the region-specific target preference of VIP-LRP population and suggest their potential functional role in setting up CA1 disinhibition concurrently with an inhibitory reset in the subiculum.

Discussion

We discovered a novel population of hippocampal VIP-expressing GABAergic neurons that exhibit specific molecular properties and, in addition to local innervation of CA1, also make long-range projections to the subiculum with region-specific connectivity patterns. These cells are only weakly active during theta oscillations associated with locomotion but maintain high activity level during quiet state. The latter may promote disinhibition of CA1 PCs in parallel with inhibition–disinhibition periods in the subiculum due to a concomitant innervation of both subicular PCs and interneurons. The likely role of VIP-LRP neurons is therefore to synchronize PC ensembles along the hippocampo-subicular axis that may be necessary for memory consolidation during animal quiet state.

We show that VIP-expressing neurons in the mouse CA1 hippocampus form several functionally and molecularly distinct populations, including BCs, local circuit IS cells and LRP neurons. The VIP-LRP cell is a novel circuit element to be included in the CA1 connectome. Here, we focused on O/A VIP-LRPs, which correspond to ~20% of the total VIP-LRP population. This cell type is different from IS3 interneurons, which express CR, exhibit distinct electrophysiological parameters and have a similar axon distribution within the CA1 but target local inhibitory neurons24,38. Indeed, the major and, perhaps, the most striking feature of VIP-LRPs is their distant projection and innervation of both interneurons and PCs in the distal projection area.

We demonstrate that, locally, VIP-LRP neurons prefer to make synapses with different classes of inhibitory interneurons, either in the O/A or in the RAD. Interneurons that are known to innervate the PC dendrites, including the O-LM, the BIS and the SC-AC cells, were among the targets of VIP-LRP axons. In addition, perisomatic terminating BCs were also innervated. As the activity of VIP-LRPs was strongly decreased during theta-run epochs, these cells are unlikely to modulate CA1 interneuron firing during theta oscillations associated with locomotion. Indeed, in agreement with our observations, most interneurons exhibit a time-locked maximal activity during theta oscillations necessary for the temporal sequence representation in PC firing2,45,51,52. Interestingly, Buzsáki et al.31 identified some rare cells in the hilus as ‘anti-theta’ cells, which were later found in the CA1, subiculum and entorhinal cortex, and classified as theta-off cells32,53,54. Despite the remarkable network behaviour of theta-off cells, their cellular identity and connectivity patterns have remained unknown. We provide evidence that, at least in the CA1 hippocampus, theta-off cells include a population of VIP-LRP GABAergic neurons that mediate local disinhibition. Importantly, the theta-off cells can display tonic firing by inactivation of the MS54, pointing to critical MS suppressive influences in their network motif. In particular, the activation of the M2Rs expressed in the somato-dendritic membrane of these cells or at presynaptic excitatory terminals may be responsible for suppression in VIPLRP activity during theta oscillations55–57. Furthermore, the activation of local or long-range GABAergic projections7,58,59 that likely converge onto VIP-LRP neurons may prevent these cells from firing during theta-run epochs. Our data also indicate that, in response to the theta-modulated input, VIP-LRPs show weak synchrony due to surround inhibition resulting from the preferential propagation of the spike AHP through gap junctions60. Therefore, on the contrary to other cell types61,62, electrotonic coupling between VIP-LRPs does not promote their synchrony; at least this is not the case in response to the theta-like input. How the different firing frequency, the state of gap junctions or the number of coupled neurons participating in the network activity35–37 may shape the cell recruitment remains to be determined.

Are VIP-LRPs discovered here similar to other subiculum-projecting hippocampal GABAergic neurons? One candidate is the trilaminar cell identified previously in the rat hippocampus63, which was shown to express M2R in the somato-dendritic membrane, was decorated with mGluR8a-containing terminals and projected to the subiculum33. This cell had a large soma with horizontally running dendrites at the O/A border and an axon innervating the CA1 from O/A through PYR to RAD. However, the distribution of the trilaminar cell axon in the rat was biased toward proximal RAD (~70%63), which is not the case for VIPLRP cells in the mouse. The presence of VIP has not been reported in the trilaminar cell, and, in contrast to VIP- LRPs, this cell shows complex spike bursts during theta oscillations and strong discharges during ripples33. The other populations of subiculum-projecting GABAergic neurons were described in the outer molecular layer of the dentate gyrus64 and in the CA127,29. The latter express COUP-TFII alone or in combination with enkephalin or calretinin, but do not express VIP. Moreover, in contrast to VIP- LRPs, these subiculum-projecting GABAergic neurons are strongly modulated during theta oscillations.

In conclusion, our data identify the VIP-LRP neuron as a novel circuit element, which, through its region- and target-specific GABAergic interactions, controls the information flow along the hippocampo-subicular axis. As activation of VIP-LRPs occurred preferentially during animal quiet state, this cell type constitutes a good candidate for hippocampo-subicular mnemonic processing associated with episode recollection and comparison65,66. Indeed, the coherency between the two regions increases during quiet network states67,68 and, in addition to other mechanisms, may require the involvement of hippocampo-subicular VIP-LRP GABAergic neurons.

Methods

Mouse lines

Nine mouse lines were used in this study: the previously characterized VIP/enhanced green fluorescent protein (VIP-eGFP24); mice [BAC line with multiple gene copies; MMRRC strain #31009, STOCK Tg(Vip-EGFP) 37Gsat, University of California, Davis, CA]; the previously described VIP-Cre mice (stock #010908, The Jackson Laboratory69); the previously characterized VIP-Cre;Ai32 mice70, which were obtained by breeding the VIP-Cre with the Ai32 line (B6;129SGt(ROSA)26Sortm32(CAG-COP4*H134R/EYFP)Hze/J; stock #012569, The Jackson Lab); Vip-Cre;Ai9 mice obtained by breeding the VIP-Cre mice with the reporter line Ai9-(RCL-tdTomato)line(B6.Cg-Gt(ROSA)26Sortm9(CAG-tdTomato) Hze/J, stock #007909, The Jackson Laboratory) and Vip-Flp;Ai65 mice obtained by breeding the VIP-FlpO mice (kindly provided by Dr. Ed Callaway under agreement with Dr. Josh Huang, CSHL) with a combinatorial reporter Ai65D line (B6;129S-Gt (ROSA)26Sortm65.1(CAG-tdTomato)Hze/J, stock #021875, The Jackson Laboratory). In VIP-eGFP mice, virtually all interneurons that were immunoreactive for VIP endogenously were confirmed to express eGFP (Supplementary Fig. 2g, i; see also ref. 24). In VIP-Cre;Ai9 mice hippocampus (Supplementary Fig. 3), we also confirmed the presence of molecular cell type markers found in VIP-eGFP mouse interneurons, albeit at a different proportion. Mice had access to food and water ad libitum and were housed in groups of two to four. Mice of either sex, of CD1 and C57BL/6J genetic backgrounds, at postnatal days 19–50 were used for all experiments. Mice were randomly assigned to experimental groups, which were matched in terms of numbers of males and females in each group. Mice undergoing surgery were housed separately (1/cage). All experiments were approved by the Animal Protection Committee of Université Laval and the Canadian Council on Animal Care. The minimal number of animals necessary for the appropriate sample size was used.

Viral constructs

The pAAV-Ef1a-DIO-hChR2(H134R)-EYFP-WPRE-pA virus was acquired from the University of North Carolina (UNC) at Chapel Hill Vector Core. The AAV1.Syn.Flex.GCaMP6f.WPRE.SV40 was acquired from the University of Pennsylvania Vector Core. The hEf1-LS1L-GFP HSV vector was provided by Dr. Rachael Neve at the MIT Viral Gene Transfer Core and packaged at the University of Massachusetts Medical School Gene Therapy Center and Vector Core. The Cav2- Cre virus was acquired from the Plateforme de Vectorologie de Montpelier (PVM) at Bio-Campus Montpelier.

Slice preparation and patch-clamp recordings

Transverse hippocampal slices (thickness, 300 µm) were prepared from VIP-eGFP or VIP-Cre;Ai32 mice of either sex as described previously24,38. Briefly, animals (P15- 30) were anaesthetized deeply with isoflurane or ketamine–xylazine (ketamine: 100 mg/kg, xylazine: 10 mg/kg) and decapitated. The brain was dissected carefully and transferred rapidly into an ice-cold (0–4 °C) solution containing the following (in mM): 250 sucrose, 2 KCl,

1.25 NaH2PO4, 26 NaHCO3, 7 MgSO4, 0.5 CaCl2 and 10 glucose oxygenated continuously with 95% O2 and 5% CO2, pH 7.4, 330–340 mOsm/L. Transverse hippocampal slices (thickness, 300 µm) were cut using a vibratome (VT1000S; Leica Microsystems or Microm; Fisher Scientific), transferred to a heated (37.5 °C) oxygenated recovery solution containing the following (in mM): 124 NaCl, 2.5 KCl,

1.25 NaH2PO4, 26 NaHCO3, 3 MgSO4, 1 CaCl2 and 10 glucose; pH 7.4; 300 mOsm/ L and allowed to recover for 1 h. Subsequently, they were kept at room temperature until use. During experiments, slices were continuously perfused (2 mL/min) with standard artificial cerebrospinal fluid (ACSF) at physiological temperature (30–33 °C) containing the following (in mM): 124 NaCl, 2.5 KCl, 1.25 NaH2PO4,

26 NaHCO3, 2 MgSO4, 2CaCl2 and 10 glucose, pH 7.4 saturated with 95% O2 and

5% CO2. VIP-positive O/A interneurons were visually identified as GFP-expressing 78

cells upon illumination with blue light (filter set: 450–490 nm). Two-photon images of GFP-expressing interneurons in acute slices were obtained using a two-photon microscope (TCS SP5; Leica Microsystems) based on a Ti-Sapphire laser tuned to 900 nm. Images were acquired with a 25× water-immersion objective (NA 0.95). Whole-cell patch-clamp recordings were obtained from single cells or pairs of neurons in voltage- or current-clamp mode. Recording pipettes (3.5–6 MΩ) were filled with a Cs-based solution for voltage-clamp recordings (in mM): 130 CsMeSO4, 2CsCl, 10 diNa-phosphocreatine, 10 HEPES, 4 ATP-Tris, 0.4 GTP-Tris, 0.3% biocytin, 2 QX-314, pH 7.2–7.3, 280–290 mOsm/L; or a K+-based intracellular solution for current-clamp recordings (in mM): 130 KMeSO4, 2 MgCl2, 10 diNaphosphocreatine, 10 HEPES, 4 ATP-Tris, 0.4 GTP-Tris and 0.3% biocytin (Sigma), pH 7.2–7.3, 280–290 mOsm/L. Data acquisition (filtered at 2–3 kHz and digitized at 10 kHz; Digidata 1440, Molecular Devices, CA, USA) was performed using the Multiclamp 700B amplifier and the Clampex 10.5 software (Molecular Devices). Active membrane properties were recorded in current-clamp mode by subjecting cells to multiple current step injections of varying amplitudes (−240 to +280 pA).

To assess synaptic connectivity between VIP-LRPs and O/A interneurons, two neurons were recorded simultaneously, with the presynaptic interneuron (VIPLRP) kept in current-clamp mode at −60 mV and the postsynaptic cell (O/A interneuron) held in voltage-clamp mode at 0 mV. The junction potential was not corrected. APs were evoked in the presynaptic interneuron via two brief somatic current injections (2 ms, 1–1.5 nA) at 20 Hz. In case of synaptic connection, this protocol evoked short- latency (<5 ms) unitary IPSCs (uIPSCs) in the postsynaptic cell. Although individual uIPSCs were small (~10 pA) and close to the noise level in our experiments (5–9 pA), we could detect them based on a constant latency (Fig. 2c, f, g). The pipette capacitance and series resistance (in voltage-clamp configuration) were compensated and bridge balance (in current-clamp

configuration) was adjusted. The series resistance (Rser) before compensation was 15–20 MΩ and was monitored continuously by applying a −5 mV step at the end of every sweep. Recordings with changes in Rser > 15% were removed from the analysis. To detect changes in uIPSCs amplitude during different frequencies of firing of VIP-LRPs, APs were generated in VIP-LRPs at 10, 50 and 100 Hz. To examine electrical coupling between VIP-LRPs, two neurons were recorded simultaneously as mentioned above in the presence of synaptic blockers: gabazine (1 µM), NBQX

(10 µM) and AP5 (100 µM). The coupling coefficient (CC12) was calculated as the ratio of voltage responses of the receiving cell (here, cell 2) to the stimulated cell (here, cell 1) with a hyperpolarizing current step (−140 pA, 1000 ms) applied to the cell 1. Pairs were considered to be electrically coupled if their coupling coefficient was ≥0.01. Gap junctions were tested with a connexin-36, connexin-50 and connexin-43 gap-junction blocker mefloquine (100 µM, #M2319, Sigma) or a broad- spectrum gap-junction blocker carbenoxolone (100 µM, #C4790, Sigma). A sinusoidal excitatory input modulated at theta-frequency (5-Hz) was applied to the electrically coupled pair in three different conditions: to the cell 1 only with both cells held at resting membrane potential (Fig. 3f), to the cell 1 only when cell 2 was depolarized to allow for spontaneous firing (Fig. 3g), and to both cells at more depolarized membrane potential (Fig. 3h).

Two-photon laser scanning photostimulation by glutamate uncaging

Two photon glutamate uncaging experiments were performed as described previously38. Briefly, acute hippocampal slices (300 µm) were obtained from VIP- eGFP mice (P15–25) and perfused during experiment with ACSF containing high Ca2+ (4 mM), high Mg2+ (4 mM) and DL-AP5 (50 μM) to reduce the spontaneous synaptic activity and the confounding polysynaptic effects. To avoid non-specific effects of 4-methoxy-7-nitroindolinyl (MNI)-caged glutamate (Glu) on inhibitory synaptic transmission, the MNI-Glu (5 mM; Tocris) was applied locally by fast micropressure pulses (5 psi, 5 ms) via a glass pipette (tip diameter of 2–3 μm) connected to a

pressure application system (PicoSpritzer II; Parker Instrumentation, Fairfield, NJ, USA) and positioned ∼10 μm above the putative VIP-LRP. The putative VIP-LRPs were selected for photoactivation based on their soma location in the CA1 O/A and expression of eGFP. They were visualized for puff-pipette positioning and two- photon somatic glutamate uncaging with a two-photon Dodt infrared scanning gradient contrast technique (Dodt-IRSGC38) using a two-photon laser scanning system (Leica TCS SP5 microscope with a 40×, 0.8 NA water-immersion objective; Leica Microsystems) based on a Ti-Sapphire laser tuned to 730 nm (laser power measured under the objective, 5–10 mW). Focal release of glutamate was accomplished by illuminating the somatic region for ~180 ms (laser power, 25–30 mW) immediately after puff application of the caged compound. These settings were reliable in evoking a single spike in VIP+ O/A interneurons. To prevent photodamage, the stimulations were repeated once every 30 s and the laser power did not exceed 40 mW (measured under the objective). Control experiments included the application of MNI-Glu without subsequent uncaging, and uncaging without prior application of MNI-Glu38.

ChR2-based mapping of VIP-LRP targets

Optogenetic activation of VIP-LRPs was achieved through wide-field low-intensity stimulation with blue light (filter set: 450–590 nm; average power at the sample, 1.30 mW; pulse duration 2.5 or 5 ms, which was corresponded to the minimal duration able to evoke the response) using a 40× water-immersion objective (NA 0.8), which was applied to an area of ~0.2 mm2 within the SUB 1.0–1.2 mm away from the CA1 border to generate antidromic spikes in VIP-LRPs while avoiding the activation of VIP-BCs and IS3 cells in CA1. The generation of antidromic spikes was confirmed using current-clamp recordings from VIP-positive O/A neurons in VIP-Cre;Ai32 mice and photostimulation in SUB. The antidromic spike generated in this case differed from the somatic one evoked by light illumination in the CA1 area (Fig. 7f). The opposite experimental paradigm, with patch-clamp recordings in the SUB and photostimulation in the CA1, was applied to investigate the targets of VIP-LRPs in

subiculum. The light-evoked IPSCs (IPSCLs) were recorded at 0 mV. Using a low- light stimulation paradigm allowed for spatially localized excitation, which generated both successful lIPSCs and failures (Fig. 7g, h; Supplementary Fig. 7a, c, d).

In vitro patch-clamp data analysis

Analysis of electrophysiological recordings was performed using Clampfit 10.6 (Molecular Devices) and Igor Pro 6.2 (WaveMetrics). For the analysis of the AP properties, the first AP appearing at current pulse of +40 to 60 pA within a 50-ms time window was analyzed. The AP amplitude was measured from the threshold to the peak. The AP latency was measured from the beginning of the current pulse to the AP threshold level. The AP half-width was measured at the voltage level of the half of AP amplitude. The fast AHP amplitude was measured from the AP threshold.

Ih-associated voltage rectification was determined as the amplitude of the membrane potential sag from the peak hyperpolarized level to the stable level when hyperpolarized to −100 mV.

To analyze the properties of uIPSCs, 100 sweeps were acquired. Sweeps with spontaneous activity occurring right before or during uIPSCs were removed. The failures were identified from individual sweeps as traces that did not contain any time- dependent signal after the end of the presynaptic AP. The failure rate was calculated as the number of failures divided by the total number of traces. After this step, all sweeps containing failures were removed and successful uIPSCs were averaged to obtain uIPSC potency for further analysis. The uIPSC latency was determined as the time interval between the peak of a presynaptic AP and the onset of the uIPSC in the postsynaptic cell. The rise time of uIPSC was taken at 20–80% and a monoexponential decay time constant was determined. We did not attempt to calculate the uIPSC synaptic conductance since the GABA reversal potential can be different at different targets and was not examined in this study. The paired-pulse ratio was determined as the ratio between the mean peak amplitude of the second response and the mean peak amplitude of the first response, which were obtained 50 ms apart, including failures. During repetitive stimulation (10–100 Hz; Fig. 2c), the 82

peak amplitudes of individual uIPSCs were extrapolated from the baseline by fitting the decay of the preceding uIPSC at the average trace. The connection ratio for specific postsynaptic targets (OLM, BIS and BC; Fig. 2i) was determined as a ratio between the number of connected cells of a specific type to the total number of recording attempts (n = 118).For ChR2-based mapping analysis, the potency of the light-evoked IPSCs (lIPSCs) was determined as the average lIPSC obtained after removal of all sweeps containing failures. The connection ratio for each specific target was determined as described above for paired recordings.

For electrical coupling analysis, cross-correlation functions in Clampfit were used to explore synchrony in voltage fluctuations (Fig. 3f) or firing (Fig. 3g, h) between electrically coupled VIP-LRPs. For spike synchrony, cross-correlation analysis was performed on high-pass (at 125 Hz) filtered voltage traces following the spike detection algorithm to correlate spike start times between the two connected cells.

Cell reconstruction and immunohistochemistry

For post hoc reconstruction, neurons were filled with biocytin (Sigma) during whole- cell recordings. Slices with recorded cells were fixed overnight with 4% paraformaldehyde (PFA) at 4 °C. To reveal biocytin, the slices were permeabilized with 0.3% Triton X-100 and incubated at 4 °C with streptavidin-conjugated Alexa-488 or Alexa-546 (1:1000) in TBS. For combined morphological and immunohistochemical analysis of recorded cells, the duration of whole-cell recordings was reduced to 10 min, and the concentration of biocytin was increased to 0.5% for reliable axonal labelling. This procedure was not required for the analysis of the expression of the membrane-bound proteins (e.g. M2R, mGluR1).

All immunohistochemical tests were performed on free-floating sections (40 or 70 µm thick) obtained with Leica VT1000S or PELCO EasySlicer vibratomes from 3–4 mice (20 sections/animal) per condition. VIP-eGFP, VIP-Cre (injected with GCaMP6f) or VIP-Cre;Ai9 mice were perfused with 4% PFA and the brains were sectioned. Sections were permeabilized with 0.25% Triton X-100 in PBS and incubated

overnight at 4 °C with primary antibodies followed by the secondary antibodies. The list of primary and secondary antibodies used is provided in Supplementary Table 2. For proenkephalin immunoreaction, biotinylation was performed to enhance the labelling specificity. Briefly, following overnight incubation of sections with rabbit proenkephalin primary antibody, biotinylated anti-rabbit antibody was applied for 24 h followed by streptavidin-conjugated AlexaFluor (1:1000; Supplementary Table 2). For controlling method specificity, the primary antibodies were omitted and sections incubated in the full mixture of secondary antibodies. Under such conditions no selective cell labelling was detected. Confocal images were acquired sequentially using a Leica TCS SP5 imaging system coupled with a 488-nm argon, a 543-nm HeNe and a 633-nm HeNe lasers. Z-stacks of biocytin-filled cells were acquired with a 1-μm step and merged for detailed reconstruction in Neurolucida 8.26.2. The axon length was measured without shrinkage correction. For Fig. 1f, an LSM710 confocal microscope (Axio Imager.Z1, Carl Zeiss) with ZEN 2008 software v5.0 (Zeiss) was used to acquire multi-channel fluorescence images sequentially with a DIC M27 Plan-Apochromat 63× (NA 1.4) objective. The cells were considered immunopositive when the corresponding fluorescence intensity was at least twice of that of the background. For representation only, the overall brightness and contrast of images were adjusted manually. Portions of images were not modified separately in any way. As the antibody to detect immunoreactivity for mGluR8 is sensitive to fixation conditions, we used sections from one well-reacting mouse.

Retrograde labeling

VIP-eGFP, VIP-Cre;Ai9 or VIP-flp;Ai65 mice (P30–100) were anesthetised deeply via the intraperitoneal injection of ketamine/xylazine (ketamine: 100 mg/kg, xylazine: 10 mg/kg). After receiving a subcutaneous injection of Buprenorphine SR (0.6 mg/mL, 0.05/30 g), animals were placed in a stereotaxic frame (Kopf Instruments) and craniotomy was performed on the right hemisphere. For subicular injections, the following bregma coordinates were used: AP, −3.62 mm; ML, ±2.4 mm; and DV, −1.4 mm or AP, −2.54 mm; ML, ±0.75–0.85 mm; and DV, −1.65 mm. For injections in

hippocampal CA1, the coordinates were: AP, −2.44 mm; ML, ± 2.4 mm; and DV, −1.3 mm. The injection pipette, which was attached to a microprocessor-controlled nanoliter injector (Nanoliter 2000; World Precision Instruments), was lowered at a speed of 1 mm/ min, and the injection of red IX RetroBeads (Luma Fluor, Inc., a total volume of 25–30 nL) or retrograde viruses (HSV-hEf1-LS1L-GFP, 20 nL; or Cav2- Cre; 50 nL) was performed at a rate of 1 nL/s. For both retrograde viruses tested (HSV and Cav2-Cre), we detected sparse labelling of local subicular reporter-VIP+ interneurons (Fig. 7b) in addition to CA1 VIP-LRPs. To restrict virus spread, in prior experiments, we estimated the minimal volume of virus required to infect subicular VIP+ cells within a maximum distance of 200 µm from the injection site. The virus spatial labelling efficacy was estimated from the number of subicular cells infected in consecutive coronal sections (50-µm thickness), with ‘zero’ distance corresponding to the injection site (Fig. 7a, left). Ten minutes after the injection, the pipette was slowly withdrawn, the scalp was sutured and the animals were allowed to recover. Two days (for RetroBeads) or 2–3 weeks (for HSV and Cav2-Cre) after the injection, the animals were intracardially perfused with 4%-PFA and hippocampal slices were prepared. For all retrograde labelling estimates, only sections from animals with a highly localized subicular injection without spread to the adjoining CA1 area were included in the analysis.

Electron microscopy

Slices from VIP-eGFP mice containing recorded cells filled with biocytin were re- sectioned to 70 μm, cryoprotected in 20% sucrose solution in 0.1 M PB for ≥3 h and freeze-thawed. Sections were washed in 0.1 M PB and incubated with Streptavidin Alexa 488 (1:1000) in TBS for 48 h at 4 °C. After revealing the biocytin in recorded cells, sections were incubated in biotin (1:100, Vector Labs) overnight followed by the avidin/biotin complex (1:100; Vector Labs) in TBS at 4 °C for 48 h. Sections were reacted with a solution of 0.05% diaminobenzidine and 0.002% hydrogen peroxide (HRP reaction) in Tris buffer for ~10 min After washing in 0.1 M PB, sections were treated with 1% osmium tetroxide solution in 0.1 M PB for 1 h, washed in PB and

dehydrated in a graded series of alcohol (70, 90, 95 and 100%) followed by propylene oxide. Uranyl acetate (1%) was added to the 70% alcohol for 35 min for contrast enhancement. Dehydrated sections were embedded in Durcupan resin (Fluka) and polymerized at 60 °C for 2 days. Target areas were cut out from the resin-embedded 70-µm-thick sections and re-embedded for ultramicrotome sectioning. Serial 60-nm-thick sections were cut and mounted on single-slot, pioloform-coated copper grids. Sections were observed with a Philips CM100 transmission electron microscope and electron micrographs were acquired with a Gatan UltraScan 1000 CCD camera. Synaptic junctions were examined in CA1 O/A and RAD. The postsynaptic target identity was determined using published criteria. Briefly, postsynaptic interneuron dendrites receive type 1 (asymmetrical) synapses on the dendritic shafts and show no spines or low spine density. In contrast, postsynaptic PCs receive type 1 synapses on their spines and type 2 symmetrical synapses on their shafts.

Two-photon imaging in awake mice

Two-photon somatic Ca2+-imaging of VIP interneuron activity was performed in head- restrained awake mice running on the treadmill, which consisted of a shock absorber free rotating wheel with minimized brain motion artifacts. The running wheel was equipped with lateral walls for increased animal contentment and coupled with an optical encoder allowing for acquisition of running speed synchronously with electrophysiological signal39. Male adult VIP-Cre mice (25–35 g body weight; P40– 100) were injected stereotaxically with AAV1.Syn.Flex.GCaMP6f.WPRE.SV40 (stock diluted 1:4 in PBS; total injection volume 100 nL) into two sites of the CA1 hippocampus using the following coordinates: AP, −2.54 mm, ML, −2.1 mm, DV, −1.3 mm and AP, −2.0, ML, −1.6, DV, −1.3 mm. At 7–10 days after viral injection, mice were anaesthetized deeply with a ketamine–xylazine mixture (ketamine: 100 mg/kg, xylazine: 10 mg/kg), and fixed in a stereotaxic frame. For hippocampal imaging window, a glass-bottomed cannula was inserted on top of the dorsal hippocampus after the cortex aspiration, and secured with Kwik-Sil at the tissue interface and

Superbond at the skull level39. For Ca2+ imaging from the retrogradely labelled VIP+ O/A interneurons, VIP-Cre mice injected with AAV-GCaMP6f in the CA1 were receiving an injection of red RetroBeads (25 nL) in the subiculum (AP, −2.54 mm, ML, −0.85 mm, DV, −1.65 mm) before hippocampal imaging window preparation. A single tungsten electrode for LFP recordings was implanted in the contralateral CA1 hippocampus and a reference electrode was implanted above the cerebellum39,44. The head plate was oriented medio-laterally at 7–13° using a fouraxis micromanipulator (MX10L, Siskiyou) and fixed with several layers of Superbond and dental cement. Mice were allowed to recover for several days with postoperative pain killer treatment for 3 consecutive days (buprenorphine, 0.1 mg kg−1; 48 h). Behavioural habituation involved progressive handling by the experimenter for ~5– 15 min twice per day for a total of 3 days, with the animal fixation in the apparatus starting from the 3rd day. During experiment, the LFP signal acquisition was performed simultaneously with the optical encoder signal and imaging trigger at a sampling frequency of 10 kHz using the DigiData 1440 (Molecular Devices), AM Systems amplifier and the AxoScope software (v10.5, Molecular Devices). Imaging was performed using a Leica SP5 TCS two-photon microscope equipped with two external photomultiplier tubes (PMTs) for simultaneous detection of green (GCaMP6f) and red (RetroBeads) fluorescence and coupled with a Ti:sapphire femtosecond laser (Chameleon Ultra II, Coherent), which was mode-locked at 900 nm. A long- range water-immersion 25× objective (0.95 NA, 2.5 mm working distance) was used for excitation and light collection to PMTs at 12 bits. Image series were acquired at axial resolutions of 2 μm/pixel and temporal resolutions of 30–48 images/s. Two 5- min long recording sessions were acquired for each cell. The experiment lasted up to 1 h, after which the mouse was placed back in its home cage. The locomotion wheel between different animals was cleaned with tap water. The image and LFP analyses were performed off-line using Leica LAS, Igor Pro (Wavemetrics, Lake Oswego, USA), Clampfit 10.6 and Statistica (StatSoft).

For post hoc immunohistochemical analysis of VIP-OA interneurons recorded in vivo, a 3D reconstruction of the hippocampal window imaged in vivo was performed using 87

sequential confocal acquisition and automatic stitching. Following in vivo experiments, animals were perfused with 4% PFA, the brains were removed, re- sectioned to 70 µm and processed for GFP, M2R and CR. Sequential confocal Z stacks (120–150 stacks in total/imaging window, 2-µm step, 500–700-µm depth from the alveus surface) were acquired using Nikon AR1 MP+ multiphoton microscope equipped with a 20× objective (NA 1.1), and automatic stitching of individual Z-stacks was applied using NIS Elements AR 4.51.00 software (Nikon Instruments).

Analysis of two-photon Ca2+ imaging data

For the analysis of spontaneous behaviour, three behavioural phases were identified: locomotion, flickering and immobility. Locomotion epochs were defined as the periods when the instantaneous speed was higher than 2 cm/s for a minimal distance of 2 cm, thereby pooling together the walking and running periods. The periods with small random movements, when the speed was above 0.25 cm/s but below the locomotion threshold, were defined as flickering. Immobility periods were defined as the times without wheel rotation.

The image analysis was performed off-line using Leica LAS, Igor Pro (Wavemetrics, Lake Oswego, USA) and Statistica (StatSoft). Movies were motion corrected along the x–y plane, no neuropil subtraction was performed39. For extraction of somatic Ca2+-transients, a region of interest was drawn around individual soma to generate the relative fluorescence change (F) vs time trace. The baseline fluorescence level

(F0) was determined as the average fluorescence signal derived from three 1-s time intervals corresponding to the lowest fluorescence level in the absence of Ca2+- transients irrespective of the behaviour state. Somatic Ca2+transients were 2+ expressed as %ΔF/F = (F − F0)/F0 × 100%. Peak Ca -signals were determined as averaged signals derived from 315-ms windows around the peak of individual Ca2+- transients over a total period of locomotion or immobility. A total of 7–10 individual Ca2+-transients were analyzed per animal behavioural state with a total of 10–15 states per cell to calculate the average peak Ca2+-transient/state for a given cell. This analysis was performed for two independent imaging sessions (S1 and S2, AVG— 88

average between the two; Fig. 5d). Ca2+-transient peak amplitudes recorded during two behavioural states (locomotion and immobility) were tested for normality in their distribution using the Shapiro–Wilcoxon test. Otsu’s method based on the discrimination criterion

2 2 2+ (η¼σ B=σ T) was applied to somatic Ca -activity recorded in VIP O/A neurons during locomotion and immobility to identify two types of cells (Fig. 5d). The discrimination criterion between the two groups of cells was 0.86, indicating a good separability. Furthermore, the Mann–Whitney test was used to determine whether obtained groups of neurons exhibit statistically different properties. The results of neuron classification are illustrated as a heat-map with the amplitude of somatic Ca2+- fluctuations during different behavioural states colour-coded (Fig. 5d).

To examine somatic Ca2+-fluctuations in relation to network oscillations, LFP traces were band-pass filtered to obtain theta oscillations (5–10 Hz) or ripples (125–250 Hz). The frequency of theta oscillations during locomotion or ripple events that were detected during quiet state was determined using the power spectrum analysis in Clampfit. The onset of the theta-run epoch, which was always associated with an increase in theta power, was defined by the beginning of the locomotion period based on the animal speed trace acquired simultaneously with LFP (Fig. 6b, c). Ripple events (9–30 events/cell; minimal event duration: 50 ms; Fig. 6b, c) were selected semi-automatically at 5 SDs above the signal background using Clampfit event search algorithm with a minimal 50-ms spacing interval between individual events. The frequency at the spectral peak of each selected event was confirmed using power spectrum analysis. Ca2+-trace segmentation triggered by the event onset (theta-run epoch or ripple) as well as event-triggered averages were conducted in Igor Pro.

Statistics

The sample size was chosen based on the pilot studies. For statistical analysis, distributions of data were first tested for normality with a Kolmogorov–Smirnov or Shapiro–Wilcoxon test (Figs. 1e, 2e and 5d; Supplementary Tables 1, 3, 4). If data were normally distributed, standard parametric statistics were used: unpaired or paired t tests for comparisons of two groups and one-way or repeated-measures ANOVA for comparisons of multiple groups followed by Tukey, Kruskal–Wallis or Chi2 tests (Figs. 1e and 2e; Supplementary Tables 1, 3, 4). If data were not normally distributed, non-parametric statistics were used: Mann–Whitney or Wilcoxon’s matched pairs test for comparisons of two groups and Kruskal–Wallis test or Dunn’s test for comparisons of multiple groups (Supplementary Table 1). All statistical analysis was conducted in Sigma Plot 11.0, Igor Pro 4.0 or Statistica. p-Values < 0.05 were considered significant. Error bars correspond to SEM.

References

Atallah, B. V. & Scanziani, M. Instantaneous modulation of gamma oscillation frequency by balancing excitation with inhibition. Neuron 62, 566–577 (2009).

Lapray, D. et al. Behavior-dependent specialization of identified hippocampal interneurons. Nat. Neurosci. 15, 1265–1271 (2012).

Lovett-Barron, M. et al. Regulation of neuronal input transformations by tunable dendritic inhibition. Nat. Neurosci. 15, 423–430 (2012).

Royer, S. et al. Control of timing, rate and bursts of hippocampal place cells by dendritic and somatic inhibition. Nat. Neurosci. 15, 769–775 (2012).

Soltesz, I. Diversity in the Neuronal Machine: Order and Variability in Interneuronal Microcircuits. (Oxford University Press, Oxford, 2006).

Acsády, L., Arabadzisz, D. & Freund, T. F. Correlated morphological and neurochemical features identify different subsets of vasoactive intestinal polypeptide-immunoreactive interneurons in rat hippocampus. Neuroscience 73, 299–315 (1996).

Acsády, L., Görcs, T. J. & Freund, T. F. Different populations of vasoactive intestinal polypeptide- immunoreactive interneurons are specialized to control pyramidal cells or interneurons in the hippocampus. Neuroscience 73, 317–334 (1996).

De Marco García, N. V., Karayannis, T. & Fishell, G. Neuronal activity is required for the development of specific cortical interneuron subtypes. Nature 472, 351–355 (2011).

Miyoshi, G. et al. Prox1 regulates the subtype-specific development of caudal ganglionic eminence- derived GABAergic cortical interneurons. J. Neurosci. 35, 12869–12889 (2015).

Dávid, C., Schleicher, A., Zuschratter, W. & Staiger, J. F. The innervation of parvalbumin-containing interneurons by VIP-immunopositive interneurons in the primary somatosensory cortex of the adult rat. Eur. J. Neurosci. 25, 2329–2340 (2007).

Staiger, J. F., Masanneck, C., Schleicher, A. & Zuschratter, W. Calbindincontaining interneurons are a target for VIP-immunoreactive synapses in rat primary somatosensory cortex. J. Comp. Neurol. 468, 179–189 (2004).

Ayzenshtat, I., Karnani, M. M., Jackson, J. & Yuste, R. Cortical control of spatial resolution by VIP+ interneurons. J. Neurosci. 36, 11498–11509 (2016).

Fu, Y. et al. A cortical circuit for gain control by behavioral state. Cell 156, 1139–1152 (2014).

Jackson, J., Ayzenshtat, I., Karnani, M. M. & Yuste, R. VIP+ interneurons control neocortical activity across brain states. J. Neurophysiol. 115, 3008–3017 (2016).

Lee, S., Kruglikov, I., Huang, Z. J., Fishell, G. & Rudy, B. A disinhibitory circuit mediates motor integration in the somatosensory cortex. Nat. Neurosci. 16, 1662–1670 (2013).

Pfeffer, C. K., Xue, M., He, M., Huang, Z. J. & Scanziani, M. Inhibition of inhibition in visual cortex: the logic of connections between molecularly distinct interneurons. Nat. Neurosci. 16, 1068–1076 (2013).

Pi, H.-J. et al. Cortical interneurons that specialize in disinhibitory control. Nature 503, 521–524 (2013).

Bayraktar, T., Welker, E., Freund, T. F., Zilles, K. & Staiger, J. F. Neurons immunoreactive for vasoactive intestinal polypeptide in the rat primary somatosensory cortex: morphology and spatial relationship to barrel-related columns. J. Comp. Neurol. 420, 291–304 (2000).

Porter, J. T. et al. Properties of bipolar VIPergic interneurons and their excitation by pyramidal neurons in the rat neocortex. Eur. J. Neurosci. 10, 3617–3628 (1998).

He, M. et al. Strategies and tools for combinatorial targeting of GABAergic neurons in mouse cerebral cortex. Neuron 91, 1228–1243 (2016).

Somogyi, P. Hippocampus: intrinsic organisation. In: Handbook of Brain Microcircuits (eds Shepherd, G. & Grillner, S.) 148–664 (Oxford University Press, Oxford, 2010).

Somogyi, J. et al. GABAergic basket cells expressing cholecystokinin contain vesicular glutamate transporter type 3 (VGLUT3) in their synaptic terminals in hippocampus and isocortex of the rat. Eur. J. Neurosci. 19, 552–569 (2004).

Karson, M. A., Tang, A., Milner, T. A. & Alger, B. E. Synaptic cross-talk between perisomatic- targeting interneuron classes expressing cholecystokinin and parvalbumin in hippocampus. J. Neurosci. 29, 4140–4154 (2009). 91

Tyan, L. et al. Dendritic inhibition provided by interneuron-specific cells controls the firing rate and timing of the hippocampal feedback inhibitory circuitry. J. Neurosci. 34, 4534–4547 (2014).

Letzkus, J. J., Wolff, S. B. E. & Lüthi, A. Disinhibition, a circuit mechanism for associative learning and memory. Neuron 88, 264–276 (2015).

Gulyas, A. I., Hajos, N., Katona, I. & Freund, T. F. Interneurons are the local targets of hippocampal inhibitory cells which project to the medial septum. Eur. J. Neurosci. 17, 1861–1872 (2003).

Jinno, S. et al. Neuronal diversity in GABAergic long-range projections from the hippocampus. J. Neurosci. 27, 8790–8804 (2007).

Miyashita, T. & Rockland, K. S. GABAergic projections from the hippocampus to the retrosplenial cortex in the rat. Eur. J. Neurosci. 26, 1193–1204 (2007).

Fuentealba, P. et al. Rhythmically active enkephalin-expressing GABAergic cells in the CA1 area of the hippocampus project to the subiculum and preferentially innervate interneurons. J. Neurosci. 28, 10017–10022 (2008).

Melzer, S. et al. Long-range-projecting GABAergic neurons modulate inhibition in hippocampus and entorhinal cortex. Science 335, 1506–1510 (2012).

Buzsáki, G., Lai-Wo, S. L. & Vanderwolf, C. H. Cellular bases of hippocampal EEG in the behaving rat. Brain Res. Rev. 6, 139–171 (1983).

Colom, L. V. & Bland, B. H. State-dependent spike train dynamics of hippocampal formation neurons: evidence for theta-on and theta-off cells. Brain Res. 422, 277–286 (1987).

Ferraguti, F. et al. Metabotropic glutamate receptor 8-expressing nerve terminals target subsets of GABAergic neurons in the hippocampus. J. Neurosci. 25, 10520–10536 (2005).

Cruikshank, S. J. et al. Potent block of Cx36 and Cx50 gap junction channels by mefloquine. Proc. Natl Acad. Sci. USA 101, 12364–12369 (2004).

Alvarez, V. A., Chow, C. C., Van Bockstaele, E. J. & Williams, J. T. Frequencydependent synchrony in locus ceruleus: role of electrotonic coupling. Proc. Natl Acad. Sci. USA 99, 4032–4036 (2002).

Trenholm, S. et al. Nonlinear dendritic integration of electrical and chemical synaptic inputs drives fine-scale correlations. Nat. Neurosci. 17, 1759–1766 (2014).

Pernelle, G., Nicola, W. & Clopath, C. Gap junction plasticity as a mechanism to regulate network- wide oscillations. PLoS Comput. Biol. 14, e1006025 (2018).

Chamberland, S., Salesse, C., Topolnik, D. & Topolnik, L. Synapse-specific inhibitory control of hippocampal feedback inhibitory circuit. Front. Cell Neurosci. 4, 130 (2010).

Villette, V., Levesque, M., Miled, A., Gosselin, B. & Topolnik, L. Simple platform for chronic imaging of hippocampal activity during spontaneous behaviour in an awake mouse. Sci. Rep. 7, 43388 (2017).

Chia, R., Achilli, F., Festing, M. F. W. & Fisher, E. M. C. The origins and uses of mouse outbred stocks. Nat. Genet. 37, 1181 (2005).

Chen, X. J. et al. Neuroanatomical differences between mouse strains as shown by high-resolution 3D MRI. NeuroImage 29, 99–105 (2006).

Mekada, K. et al. Genetic differences among C57BL/6 substrains. Exp. Anim. 58, 141–149 (2009).

Otsu, N. A threshold selection method from gray-level histogram. IEEE Trans. Syst. Man Cybern. 9, 62–66 (1979).

Malvache, A., Reichinnek, S., Villette, V., Haimerl, C. & Cossart, R. Awake hippocampal reactivations project onto orthogonal neuronal assemblies. Science 353, 1280–1283 (2016).

Katona, L. et al. Sleep and movement differentiates actions of two types of somatostatin-expressing GABAergic interneuron in rat hippocampus. Neuron 82, 872–886 (2014).

Varga, C., Golshani, P. & Soltesz, I. Frequency-invariant temporal ordering of interneuronal discharges during hippocampal oscillations in awake mice. Proc. Natl Acad. Sci. USA 109, 2726– 2734 (2012).

Buzsáki, G., Horvath, Z., Urioste, R., Hetke, J. & Wise, K. High-frequency network oscillation in the hippocampus. Science 256, 1025–1027 (1992).

Buzsáki, G. et al. Hippocampal network patterns of activity in the mouse. Neuroscience 116, 201– 211 (2003).

Chrobak, J. & Buzsaki, G. High-frequency oscillations in the output networks of the hippocampal- entorhinal axis of the freely behaving rat. Neuroscience 16, 3056–3066 (1996).

Villalobos, C., Maldonado, P. E. & Valdés, J. L. Asynchronous ripple oscillations between left and right hippocampi during slow-wave sleep. PLoS ONE 2017, 12 (2017).

Klausberger, T., Magill, P. J., Cobden, P. M. & Somogyi, P. Brain-state- and cell-type-specific firing of hippocampal interneurons in vivo. Nature 421, 844–848 (2003).

Klausberger, T. et al. Spike timing of dendrite-targeting bistratified cells during hippocampal network oscillations in vivo. Nat. Neurosci. 7, 41–47 (2004).

Colom, L. V., Ford, R. D. & Bland, B. H. Hippocampal formation neurons code the level of activation of the cholinergic septohippocampal pathway. Brain Res. 410, 12–20 (1987).

Mizumori, S. J., Barnes, C. A. & McNaughton, B. L. Behavioral correlates of theta-on and theta-off cells recorded from hippocampal formation of mature young and aged rats. Exp. Brain Res. 80, 365–373 (1990).

Apostol, G. & Creutzfeldt, O. D. Crosscorrelation between the activity of septal units and hippocampal EEG during arousal. Brain Res. 67, 65–75 (1974).

Fukudome, Y. et al. Two distinct classes of muscarinic action on hippocampal inhibitory synapses: M2-mediated direct suppression and M1/M3-mediated indirect suppression through endocannabinoid signalling. Eur. J. Neurosci. 19, 2682–2692 (2004).

Lawrence, J. J., Haario, H. & Stone, E. F. Presynaptic cholinergic neuromodulation alters the temporal dynamics of short-term depression at parvalbumin-positive basket cell synapses from juvenile CA1 mouse hippocampus. J. Neurophysiol. 113, 2408–2419 (2015).

Kaifosh, P., Lovett-Barron, M., Turi, G. F., Reardon, T. R. & Losonczy, A. Septohippocampal GABAergic signaling across multiple modalities in awake mice. Nat. Neurosci. 16, 1182–1184 (2013).

Unal, G., Joshi, A., Viney, T. J., Kis, V. & Somogyi, P. Synaptic targets of medial septal projections in the hippocampus and extrahippocampal cortices of the mouse. J. Neurosci. 35, 15812–15826 (2015).

Vervaeke, K. et al. Rapid desynchronization of an electrically coupled interneuron network with sparse excitatory synaptic input. Neuron 67, 435–451 (2010).

Mann-Metzer, P. & Yarom, Y. Electrotonic coupling interacts with intrinsic properties to generate synchronized activity in cerebellar networks of inhibitory interneurons. J. Neurosci. 19, 3298–3306 (1999). van Welie, I., Roth, A., Ho, S. S. N., Komai, S. & Häusser, M. Conditional spike transmission mediated by electrical coupling ensures millisecond precision-correlated activity among interneurons in vivo. Neuron 90, 810–823 (2016).

Sik, A., Penttonen, M., Ylinen, A. & Buzsaki, G. Hippocampal CA1 interneurons: an in vivo intracellular labeling study. J. Neurosci. 15, 6651–6665 (1995).

Ceranik, K. et al. A novel type of GABAergic interneuron connecting the input and the output regions of the hippocampus. J. Neurosci. 17, 5380–5394 (1997).

Deadwyler, S. A. & Hampson, R. E. Differential but complementary mnemonic functions of the hippocampus and subiculum. Neuron 42, 465–476 (2004).

Kim, S. M., Ganguli, S. & Frank, L. M. Spatial information outflow from the hippocampal circuit: distributed spatial coding and phase precession in the subiculum. J. Neurosci. 32, 11539–11558 (2012).

Jackson, J., Goutagny, R. & Williams, S. Fast and slow gamma rhythms are intrinsically and independently generated in the subiculum. J. Neurosci. 31, 12104–12117 (2011).

Jackson, J. et al. Reversal of theta rhythm flow through intact hippocampal circuits. Nat. Neurosci. 17, 1362–1370 (2014).

Taniguchi, H. et al. A resource of Cre driver lines for genetic targeting of GABAergic neurons in cerebral cortex. Neuron 71, 995–1013 (2011).

David, L. S. & Topolnik, L. Target-specific alterations in the VIP-inhibitory drive to hippocampal GABAergic cells after status epilepticus. Exp. Neurol. 292, 102–112 (2017).

Figure Legends

Figure 1. Identification of VIP-LRPs in the VIP-eGFP mouse. a Two-photon image (maximal projection of a z-stack of 200 µm height) of the CA1 area from an acute hippocampal slice (300 µm) of a VIP-eGFP mouse showing the location of GFP cell bodies, axons and dendrites in the O/A area of CA1. Scale bar: 100 μm. b Reconstruction (the axon is shown in red, the dendrites are shown in green) of a VIP-LRP cell that was recorded and ﬁ lled with biocytin in a slice obtained from a VIP-eGFP mouse. Scale bar: 100 µm. c Representative voltage responses of a VIP-LRP to hyperpolarizing (-240 pA), and depolarizing (+80 pA and +280 pA) current injections, with an inset illustrating the first spike evoked by +80-pA current pulse. d Confocal images showing RetroBeads labelling of a VIP-LRP soma (left) after injection in subiculum and immunoreactivity for M2R in a VIP-positive neuron labelled with biocytin (single focal plane, right). Scale bar: 10 μm. e Pie charts illustrating the mean axonal distribution in different layers (based on axon length obtained following reconstruction in Neurolucida) for groups of cells corresponding to 3 different cell types: VIP-LRP (n = 10), VIP-BC (n = 5) and IS3 cell (n = 6). OA, oriens- alveus; PYR, stratum pyramidale; RAD, stratum radiatum and SUB, subiculum. No axon was detected within stratum lacunosum moleculare (LM) for the three cell types. Statistically significant differences in the axon distribution between VIP-LRP and VIP-BCs, VIP-LRP and IS3, and VIPBCs and IS3 at **p< 0.01, one-way ANOVA followed by Tukey’s test. f VIP-LRP cells, identified by somato- dendritic membrane M2R immunoreactivity (blue), are innervated by terminals rich in presynaptic mGluR8 (purple). Single optical slices (0.45 mm thick) of confocal images of quadruple immunoreactions as indicated: i, right, framed area at higher magnification; ii, four VIP+ terminals (arrows) show mGluR8 immunoreactivity.

Figure 2. VIP-LRPs provide inhibition to CA1 O/A interneurons. a Reconstruction of a connected pair of VIP-LRP and oriens-lacunosum moleculare (O-LM) interneurons. The axon of the presynaptic VIPLRP cell is shown in red and its dendrites are shown in green. The axon of the postsynaptic O/A interneuron is shown in blue, with its dendrites shown in black. The inset shown on top illustrates schematically the configuration of the recording. b Post hoc immunohistochemical analysis of a different synaptically connected pair of VIP- LRP (immunoreactive for M2R, arrow) and O-LM cell (arrowhead) with insets showing the O-LM cell expanded. Scale bars: 20 µm, 10 µm (inset).

c Representative traces of uIPSCs in an O-LM cell at 0 mV (100 consecutive traces with the average shown in red) evoked by two APs through a current injection in a presynaptic VIP-LRP cell (top), and examples of uIPSCs during different frequencies of VIP-LRP firing: 10 Hz (middle) and 100 Hz (bottom). Insets at the bottom show expanded uIPSCs during 100-Hz fi ring of VIP-LRPs. d Cumulative histograms of uIPSC rise time and decay time constant in O-LM cells (n = 11 pairs). e Summary plot (mean ± SEM) showing changes in uIPSC amplitude in O-LM cells during different fi ring frequencies of VIP-LRPs (** p < 0.01, one-way ANOVA/Tukey ’ s test). f-g Reconstructions of connected pairs of VIP-LRP and a bistrati fi ed (BIS; f) or basket (BC; g) cells. Scale bars: 100 µm. The insets on the right show corresponding uIPSCs. h Electron micrograph (EM) images of biocytin labelled boutons (b) of VIP-LRP cells. Top, the VIP-LRP boutons, which form type-2 synapses (arrowheads) with a small (left d1) and a large diameter (right) dendritic shafts (d) of interneurons receiving type-1 synapses (arrows) in stratum oriens. Bottom, EM images illustrating dendrites (d1, d2) in CA1 RAD as postsynaptic targets of a VIP-LRP. The VIP-LRP bouton (white ‘b’ ) makes two type-2 synapses (solid arrows): with a spiny (d1, s) and aspiny dendrite (d2) receiving a type-1 synapse (open arrow). Scale bars: 0.5 µm (top), 200 nm (bottom). i Summary graph illustrating the connection probability. j Boxplots representing the uIPSC amplitude for different targets (OLM: n = 8, BIS: n = 5, BC: n = 4; * p < 0.05, unpaired t test)

Figure 3. Electrical coupling between VIP-LRPs. a Schematic of the simultaneous recording of two VIP-LRPs and their reconstruction. Scale bar: 100 µm. b Representative example of the simultaneous recording of two VIP-LRPs with an AP initiated in the presynaptic cell (IN1, black trace) and a corresponding voltage response in the postsynaptic cell (IN2, red trace). The positive and negative components of the spikelet are shown with a star and square symbols, respectively. c Summary plots for a group of cells (n = 17) indicating changes in the postsynaptic Vm as a function of the AP amplitude (upper) or sAHP amplitude (bottom). Red line is a linear fit to the data points (r = 0.88, Pearson correlation) for a slow negative spikelet component associated with sAHP in the presynaptic cell. d Representative examples of voltage traces recorded in the VIP-LRP pair before and after the application of mefloquine. e Summary plots for the coupling coefficient (CC) exhibiting symmetry between different pairs (left; m, slope of the regression line ± SE), and for the gap-junction blockers’ effect

(Mefl, Mefloquine; Cbx, Carbenoxolone) for a group of cells (right; Mefl: n = 6; Cbx: n = 4; p< 0.05; paired t test). f Voltage responses recorded in a VIP-LRP pair (at a subthreshold level for AP generation Vm) to a sinusoidal current (red trace, 5 Hz) applied to the presynaptic cell (IN1). Plot below shows cross-correlations in Vm fluctuations between the two cells for a group of pairs (n = 6), with red trace corresponding to the average data. g Voltage responses (ﬁve consecutive traces of different colours superimposed) recorded in the VIP-LRP pair with the postsynaptic cell (IN2) being depolarized, and a sinusoidal current (red trace) applied to the IN1. Stars of different colours above the IN2 traces indicate APs generated synchronously in two cells. The plot below shows cross-correlations in the AP occurrence (n = 6 pairs), with red trace illustrating the average data. Insets on the right show voltage responses in two cells with (red trace) and without (black trace) an AP generated in the presynaptic cell. h Voltage responses recorded in a VIP-LRP pair to a sinusoidal current (red trace, 5 Hz) applied to both cells. Stars above the IN2 trace indicate APs generated synchronously in two cells. Plot below shows cross-correlations in the AP occurrence for a group of pairs (n = 6 pairs), with red trace corresponding to the average data.

Figure 4. Two-photon glutamate uncaging-based mapping of local and distant axonal targets of VIP-LRPs. a-d Average traces of glu-IPSCs (Vhold: 0 mV) evoked by uncaging of MNI-Glu on VIP+ O/A interneuron somata (left) and the corresponding connection probability (right) in CA1 O/A interneurons a, CA1 PCs b, SUB interneurons c and SUB PCs d. Each row corresponds to a single cell with the ratio of connections indicated at bar graphs. Each recorded cell was tested for receiving input from 1 to 5 VIP-GFP+ O/A interneurons. Traces with shadow area correspond to examples of glu-IPSCs: CA1 interneurons (n = 9 connections out of 31 tested/11 cells), CA1 PCs (n = 0 connections out of 18 tested/10 cells), SUB interneurons (n = 8 connections out of 45 tested/26 cells), SUB PCs (n = 11 connections out of 37 tested/24 cells). Scale bars (shown in A): 20 pA, 10 ms. e Schematic of simultaneous patch-clamp recordings from different CA1 and SUB targets and two-photon MNI-Glu uncaging on somata of VIP-GFP+ cells in CA1 O/A. f Summary spatial maps illustrating the density of connections within the CA1 (top) and SUB (bottom). Connected cells are shown as shaded symbols. Scales are in µm. g Post hoc immunohistochemical validation of connected interneurons confirmed that VIP- GFP+ cells innervating CA1 O/A interneurons were positive for M2R and, thus corresponded to VIP-LRPs. Scale bar: 20 µm. h Summary bar graphs showing the connection ratio (top) and the peak amplitude of glu- IPSCs (bottom) for different postsynaptic targets in CA1 and subiculum. The connection ratio is a ratio between the number of connections over the total number of tests for a given target type.

Figure 5. Imaging VIP-LRP activity in awake mice. a Schematic of simultaneous two-photon Ca2+-imaging and LFP recordings in awake head- restrained mice (left) and a pie-chart illustrating the time distribution of different behavioural states (right). b Two-photon image of the GCaMP6f-expressing VIP cells in CA1 O/A (maximal projection of a 100-µm Z-stack). Scale bar: 50 µm. c Representative traces of simultaneous LFP recording (upper raw trace and ﬁ ltered for theta: 5 – 10 Hz), somatic Ca2+-imaging from a VIP cell, and animal speed (red). d VIP O/A cell segregation based on the cell recruitment determined from the peak Ca2+- transient during locomotion and immobility. Two types of neurons were revealed: type I VIP cells (n = 8) were positive for M2R and negative for CR (2 out of 2 cells tested) whereas type II VIP-expressing cells (n = 20) were negative for both M2R and CR (4 out of 4 cells tested). S1 and S2 represent two independent imaging sessions and AVG is the average peak Ca2+-signal of the two sessions. e Confocal stitching of the hippocampal CA1 O/A imaging window used for two-photon image acquisition in vivo and processed for post hoc immunohistochemistry illustrating GFP (left) and M2R (middle) immunoreactivity as well as the overlay of the two (right). White arrows point to M2R-positive VIP cells targeted with GCaMP6f, which are illustrated as insets. Anatomical landmarks are indicated as following: A – anterior, P – posterior, M – medial, L – lateral. Scale bar: 100 µm. f Representative two-photon images obtained from hippocampal slices of VIP-Cre mice in vitro (left) and from the CA1 O/A VIP + interneuron in vivo (right) illustrating the RetroBead injection site in SUB (left top, Dodt image superimposed with epiﬂuorescence image for Retrobeads), the retrogradely labelled CA1 PCs and an O/A interneuron (white arrow; left bottom) and a CA1 O/A VIP-LRP labelled with RetroBeads and expressing GCaMP6f (right). Scale bars: 500 µm (top), 50 µm (bottom left), 10 µm (bottom right). g Histograms of Ca2+-transient peak amplitude (top) and its ratio during immobility to that during locomotion (bottom) obtained from VIP-LRPs labelled with RetroBeads ( n = 5 cells, 25–30 locomotion/immobility periods/cell) and showing a larger amplitude of Ca2+-signals during immobility than during locomotion ( p < 0.01, one-way ANOVA/Tukey’s test)

Figure 6. Network state-dependent recruitment of VIP-OA interneurons in awake mice. a Representative traces of simultaneous LFP (raw trace and filtered for theta and ripples) and Ca2+-transient (∆F/F) recordings from a putative VIP-LRP cell identified post hoc as M2R-positive (d). Red trace illustrates the animal locomotion speed (dotted line indicates the threshold for the locomotion state at 2 cm/s). b Individual traces from the event-triggered Ca2+-trace segmentation and corresponding average (red trace) generated by the theta-run epochs (left) and ripple episodes (right; with inset showing an expanded view of the ripple event.) from the cell illustrated in (A) with heat- maps showing the group data for all VIPLRPs (n = xx events/6 cells for theta-run; n = xx 98

events/5 cells for ripples). Decrease in somatic Ca2+-signals was significant during theta-run epochs for a group of cells (n = 6) at p< 0.001; Mann-Whitney test. c Representative traces from the event-triggered Ca2+-trace segmentation with corresponding average (red trace) generated by the theta-run epochs (left) and ripples (right) with heat-maps showing the group data for type II M2R-/CR- VIP-expressing cells (n = 14 cells for theta-run; n = 13 cells for ripples). Increase in somatic Ca2+-signals during theta-run epochs was significant for a group of cells (n = 14) at p < 0.001; Mann-Whitney test. d Post hoc immunohistochemical analysis of the recorded VIP-LRP showing that cells of this sub-type (type I) express M2R. GFP was revealed with Alexa488, CR with Cy3 and M2R with CF-633 secondary antibodies. Scale bar: 10 µm. e Post hoc immunohistochemical analysis of the recorded type II VIP-expressing cells showing that cells of this type do not express M2R or CR. Scale bar: 10 µm.

Figure 7. Cellular diversity and connectivity of subiculum-projecting VIPLRPs. a Summary distributions (left) of infected neurons within SUB for HSV-hEf1-LS1L-GFP and Cav2-Cre retrograde viruses following injection in SUB (the ‘zero’ distance indicates the virus injection focus; data points indicate the number of cells targeted in the adjacent slices), and a pie-chart (right) illustrating the VIP-LRP fraction (green) out of the total VIP+ population (orange) in the CA1 area of VIP- Cre ;Ai9 mice using a Cre-inducible herpes simplex virus (HSV-hEf1-LS1L-GFP). b Representative confocal images showing retrograde labelling of subiculum-projecting VIP- LRPs (right) using an HSV-hEf1-LS1L-GFP injection in the subiculum of VIP-Cre;Ai9 mice (left). Red signal – tdTomato-expressing VIP+ neurons; yellow signal – a subpopulation of GFP-VIP+ cells (indicated with white arrowheads) that were labelled with the virus within the subiculum (left) or retrogradely in the CA1 (right). Scale bar: 100 µm. c Combinatorial genetic labeling of VIP-LRPs using Cav2-Cre virus injections in the subiculum of VIP-Flp;Ai65 mice confirms the location of VIP-LRPs in different CA1 layers. Left, white arrows point to tdTomato-VIP somata expressing tdTomato under cre and flp control. The area indicated with a white rectangle on the left is shown expanded on the right. Scale bar: 20 µm. d Representative confocal images illustrating the markers expressed by VIPLRPs: M2R (left), CR (center) and Penk (right) in VIP-Cre;Ai9 mice. Scale bars: 10 µm. e Schematic illustration of the optogenetic activation of VIP-LRPs through light stimulation in the subiculum. f Example traces and summary phase plots for antidromic spike (red) evoked in VIP-LRP by light stimulation vs somatically evoked spike (top, black). g, h Light-evoked IPSCs in response to antidromic VIP-LRP activation in different CA1 (g) and subicular (h) targets, including an O-LM (g, left) and a BIS (g, right) cells in the CA1

area, as well as pyramidal cells (h, left) and a SOM-positive interneuron (h, right) in subiculum (scale bars: 100 µm for reconstructions, 20 µm (g bottom), 10 µm (h, bottom)). i Pie-chart (left) illustrating the distribution of VIP-LRP CA1 targets (DT-IN dendrite-targeting interneuron, ST-IN soma-targeting interneuron, PYR pyramidal cell) and summary bar graphs (right) showing the connection ratio for subicular targets (PYR pyramidal cell, subIN subicular interneuron). j Summary bar graphs illustrating the connection ratio and the lIPSC amplitude (mean ± SEM) for CA1 targets (O-LM n = 18, BIS n = 5, BC n = 4, SC-AC n = 4; O-LM vs Bis, p > 0.05, O-LM vs BC, * p < 0.05, O-LM vs SC-AC, ** p < 0.01, Mann – Whitney test).

100

Figures

Figure 1.

101

Figure 2.

102

Figure 3.

103

Figure 4.

104

Figure 5.

105

Figure 6.

106

Figure 7.

107

Chapter 3 Integrated article 2: Transcriptomic profile of hippocampal long-range VIP-GABAergic neurons

Résumé

Dans les circuits corticaux, les interneurones GABAergiques exprimant le peptide intestinal vasoactif (VIP+) représentent un groupe hétérogène mais unique d'interneurones principalement spécialisé dans la désinhibition des réseaux. Les répertoires moléculaires spécifiques aux cellules importants pour le ciblage sélectif des sous-types de cellules VIP+ et la compréhension de leurs fonctions restent inconnus. En utilisant une approche de séquençage par patch-clamp, nous avons analysé le profil transcriptomique de neurones, identifiés anatomiquement, VIP+ GABAergiques à projection longue portée (VIP-LRPs) localisés dans la région CA1 de l'hippocampe de la souris qui coordonnent les interactions hippocampo- subiculaire. Nous avons exploré l'expression du gène VIP-LRP dans les principales familles de gènes : les canaux ioniques, les récepteurs de neurotransmetteurs, les neuromodulateurs, les molécules d'adhésion cellulaire et de myélinisation. De nombreux gènes qui ont été enrichis dans d'autres sous-types de cellules VIP+, y compris les interneurones spécifiques aux interneurones et les cellules en panier co- exprimant la cholécystokinine, ont été détectés dans les VIP-LRPs. De plus, le neuropeptide Y et la nétrine G1 ont été identifiés comme des marqueurs moléculaires potentiels pouvant être utilisés pour le ciblage combinatoire des VIP- LRPs. En somme, nos résultats suggèrent que les VIP-LRPs représentent un sous- type spécifique à l’intérieur de la population des neurones VIP+ qui partagent une identité moléculaire avec d'autres interneurones VIP+, mais qui possèdent également des gènes associés à des caractéristiques spécifiques leur permettant une coordination à distance.

108

Abstract

In cortical circuits, the vasoactive intestinal peptide (VIP+)-expressing GABAergic interneurons represent a heterogeneous but unique group of interneurons that is mainly specialized in network disinhibition. The cell-specific molecular repertoires important for selective targeting of VIP+ cell subtypes and understanding their functions remain unknown. Using patch-sequencing approach, we analyzed the transcriptomic profile of anatomically identified long-range projecting VIP+ GABAergic neurons (VIPLRP) that reside in the mouse hippocampal CA1 area and coordinate the hippocampo-subicular interactions. We explored the VIP-LRP gene expression within major gene families including ion channels, neurotransmitter receptors, neuromodulators, cell adhesion and myelination molecules. Many genes that were enriched in other local VIP+ cell subtypes, including the interneuron- selective interneurons and the cholecystokinin-co-expressing basket cells, were detected in VIP-LRPs. Moreover, the neuropeptide Y and netrin G1 were identified as potential molecular markers that can be used for combinatorial targeting of VIP- LRPs. Together, our data suggest that VIP-LRPs represent a specific subtype within a continuum of VIP+ population, which shares molecular identity with other VIP+ interneurons, but also has genes associated with specific features that allow for long- distance coordination.

109

Introduction

Understanding brain computations requires a detailed analysis of single neurons, from gene expression to cell-specific network and behaviour function. GABAergic inhibitory interneurons constitute one of the most heterogeneous neuronal populations in cortical networks, as dozens of different interneuron types with specific morphological and physiological properties have been described so far (Petilla Interneuron Nomenclature et al., 2008; Somogyi, 2010). While some interneuron types (e.g., parvalbumin-expressing basket or somatostatin-positive Martinotti cells) have been well characterized at a structural and functional level, the other rare types remain unstudied. This is particularly the case for the vasoactive intestinal peptide-expressing (VIP+) interneurons that derive from the caudal ganglionic eminence (CGE) (Miyoshi et al., 2015) and account for about∼10-15% of GABAergic interneurons in the neocortex (Gonchar et al., 2007). Recent findings highlighted the critical role of VIP+ interneurons in regulating complex behavioural tasks such as reward-associated learning, visual processing and locomotion through local network disinhibition (Lee et al., 2013; Pi et al., 2013; Fu et al., 2014, Ayzenshtat et al., 2016). However, these cells are also heterogeneous and can exhibit distinct morphological and neurochemical properties as well as cell-specific connectivity patterns and physiological roles (Acsády et al., 1996a; 1996b; Porter et al., 1998; Bayraktar et al., 2000; Chamberland et al., 2010; Tyan et al., 2014). In addition to VIP+ interneuron-selective cells targeting local interneurons (Acsády et al., 1996a, 1996b; Chamberland et al., 2010; Tyan et al., 2014) and VIP+ basket cells (Somogyi et al., 2004), a novel type of VIP/muscarinic receptor 2 (M2R)-co- expressing long range-projecting (VIP-LRP) GABAergic neuron has been identified in the CA1 hippocampus. These cells target interneurons in the CA1 and both interneurons and principal cells in subiculum (Francavilla et al., 2018). Unlike other GABAergic cells in the hippocampus (Klausberger and Somogyi, 2008), VIP-LRP cells are more active during quiet wakefulness, and are not involved in theta and ripple oscillations. However, the molecular profile of VIP-LRPs, which would help to target these cells for selective manipulations and functional studies remains

110

unknown.

Recent advances in opto- and pharmacogenetic technologies and the development of cell-specific transgenic targeting strategies allow the manipulation of specific cellular populations. For example, a large set of mouse lines based on the Cre/Flp and Cre/Dre double recombinase systems has been generated to allow for highly selective manipulations (Taniguchi et al., 2011; Madisen et al., 2015; Paul et al., 2017). However, the ongoing development of cell-specific mouse lines requires identification of cell-specific markers. Next generation single-cell RNA sequencing (scRNA-seq) has proved to be a powerful tool in neuronal transcriptomic profiling (Zeisel et al., 2015; Tasic et al., 2016; Paul et al., 2017). Importantly, when combined with patch-clamp recordings (Patch-seq), it allows the acquisition of single-cell transcriptomes from morphologically and electrophysiologically identified neurons, thus providing a unique opportunity for multimodal sampling of rare cell types (Cadwell et al., 2016; Fuzik et al., 2016; Földy et al., 2016). Here, we used this approach to explore the transcriptomic profile of the hippocampal CA1 VIP-LRP cells, a sparse GABAergic population, which is positioned to control the information flow along the hippocampo-subicular axis and to coordinate the two functionally related areas. We provide VIP-LRP single-cell transcriptomes and identify some molecular markers that can be used for their combinatorial targeting.

Materials and Methods

Experimental subjects and housing conditions

We used the previously characterized VIP/enhanced green fluorescent protein (VIP- eGFP; Tyan et al., 2014) mice [BAC line with multiple gene copies; MMRRC strain #31009, STOCK Tg(Vip-EGFP) 37Gsat, University of California, Davis, CA]. Mice had access to food and water ad libitum and were housed in groups of two to four. All experiments were approved by the Animal Protection Committee of Université Laval and the Canadian Council on Animal Care.

Patch RNA-sequencing in acute hippocampal slices

111

Transverse hippocampal slices (thickness, 300 µm) were prepared from VIP-eGFP mice (P15–P25) of either sex as described previously (Tyan et al., 2014). Briefly, animals were anaesthetized deeply with isoflurane and decapitated. The brain was dissected and transferred into an ice-cold (0 to 4°C) solution containing the following

(in mM): 250 sucrose, 2 KCl, 1.25 NaH2PO4, 26 NaHCO3, 7 MgSO4, 0.5 CaCl2, and

10 glucose oxygenated continuously with 95% O2 and 5% CO2, pH 7.4, 330–340 mOsm/L. Slices were cut using a vibratome (VT1000S; Leica Microsystems or Microm; Fisher Scientific), transferred to a heated (37.5°C) oxygenated recovery solution containing the following (in mM): 124 NaCl, 2.5 KCl, 1.25 NaH2PO4, 26

NaHCO3, 3 MgSO4, 1 CaCl2, and 10 glucose; pH 7.4; 300 mOsm/L and allowed to recover for 1 h. Subsequently, they were kept at room temperature until use. During experiments, slices were continuously perfused (speed: 2 mL/min; temperature: 30– 33ºC) with standard artificial cerebrospinal fluid (ACSF) containing the following (in mM): 124 NaCl, 2.5 KCl, 1.25 NaH2PO4, 26 NaHCO3, 2 MgSO4, 2CaCl2, and 10 glucose, pH 7.4 saturated with 95% O2 and 5% CO2. VIP-positive O/A interneurons were visually identified as eGFP-expressing cells upon illumination with blue light (filter set: 450–490 nm). Glass capillaries and small tools were autoclaved before the experiment. All working surfaces were cleaned with DNA-OFF (Takara, Cat.No. 9036) and RNase Zap (Life Technologies, Cat.No. AM9780). The patch-clamp protocol was optimized to perform high-quality RNAseq of single morphologically identified neurons (Fuzik et al., 2016; Cadwell et al., 2016). In particular, to provide for optimal RNA yield, we kept the same pipette tip size, volume of patch-solution and a modified intracellular solution as in Cadwell et al., 2016. Patch pipettes with low resistance (2–4 MΩ) were filled with ~1µl of intracellular solution containing (in mM): 130 KMeSO4, 2 MgCl2, 10 diNa-phosphocreatine, 10 HEPES, 4 ATP-Tris, 0.4 GTPTris, glycogen (an inert carrier to increase the RNA recovery, 20 µg/ml; ThermoFisher Scientific), and 1.3 mg/ml biocytin, pH 7.2–7.4, 275–290 mOsm/L. Before approaching the cell of interest, 0.1–0.2 ml positive pressure was applied to the patch-pipette, and the pipette was quickly advanced through the slice until it was touching the cell membrane. Membrane properties were recorded in current-clamp mode within the first 5 min by subjecting cells to multiple current step injections of

112

varying amplitudes (-400 to +280 pA). Data acquisition (filtered at 2-3 kHz and digitized at 10kHz; Digidata 1440, Molecular Devices, CA, USA) was performed using the Multiclamp 700B amplifier and the Clampex 10.5 software (Molecular Devices). The whole-cell configuration was kept for 15–20 min to allow the diffusion of biocytin. The cell quality was evaluated according to electrophysiological (stable membrane potential of more than –50 mV, and AP amplitude ≥ 50 mV measured from the threshold) criteria. Only cells of good quality were processed for complementary DNA (cDNA) library construction. RNA was collected at the end of the recording by slowly applying light suction of 0.2–0.3 ml through a syringe connected to the patch-pipette until the cell was shrunk or the giga-seal was lost. Both the cytoplasm and the nucleus were sampled, resulting in a high cDNA yield (Fig. 1D). Then the tip of the pipette was broken, and the cell content was ejected into an RNase-free PCR tube containing 10 µl 1X lysis buffer and 1µl RNase inhibitor, spun down to the bottom and stored at –80 °C. If any tissue debris were observed on the pipette tip when withdrawn from the slice, the sample was discarded. The slices were fixed with 4% PFA and processed for biocytin labeling to recover the cell morphology. In negative control (Ctl”-“) experiments, the samples were obtained by aspirating tissue in O/A using the same approach and suction procedures. Positive control (Ctl”+“) was obtained by sequential aspiration of 3 GFP+ O/A cells into the same patch-pipette (Fig. 1F).

The cDNA libraries were constructed using SMART-Seq v4 Ultra low input RNA Kit (Clontech Laboratories, Takara Bio; Mountain View, CA, USA). The first-strand cDNA was synthesized from cell contents by the 3’ SMART-Seq CDS Primer II A and template switching was performed by SMART-Seq v4 Oligonucleotide at the 5’ end of the transcript. Then cDNA from SMART sequences was amplified by PCR Primer II A. After 17 cycles of long-distance PCR, amplified cDNA was purified using the Agencourt AMPure XP Kit (Beckman Coulter, Cat.No. A63882). The quality of cDNA library construction was validated using the Agilent Tapestation 2200 system. The threshold cDNA concentration of 200 pg/ml (Cadwell et al., 2016) was achieved in all samples. Subsequently, cDNA libraries were prepared for Illumina Next Generation sequencing using Nextera XT DNA Library Preparation Kits (Illumina Inc.,

113

San Diego, CA, USA). Libraries with unique index were then pooled together in equimolar ratio and sequenced for paired-end sequencing using both lanes of a rapid run flow-cell on the Illumina HiSeq 2500 system. The average insert size for the paired-end libraries was 225 base pairs. Investigators conducting the cDNA library construction were blind to the experimental condition (cell vs Ctl”-“ vs Ctl”+”).

Analysis of patch RNA-seq data

Given that the External RNA Controls Consortium (ERCC) spike-in-based approach may not be appropriate for quantifying technical noise, quality control assessment and absolute transcript normalization (Brennecke et al., 2013; Grun and van Oudenaarden, 2015; Risso et al., 2014), this method was not used in our analysis of single-cell samples. Sequencing raw data were de-multiplexed to discriminate reads from different samples and then trimmed to remove sequencing adapters, low confidence bases, sequencing-specific bias and PCR artifacts. We computed reads alignments as a measure of quality, specifically the fraction of reads that could be mapped back to the mouse genome (mm10 assembly) as indicated by the aligner STAR (v2.5.2a). For reads quantification we computed pseudoalignments using Kallisto (Pachter Lab, UC Berkeley). The average number of pseudo-aligned reads per cell was 2.1 million. Expression levels in individual cells were presented in TPM (transcripts per million; with TPM threshold set to 1.0 to filter out noise from the expression data). The number of genes with TPM values ≥ 1.0 varied from 4,000 to 16,000 (Fig. 1E) and was in the range of the previously reported with Patch-seq (e.g., 7,000 genes/cell in Cadwell et al., 2016; 4,000–10,000 genes/cell in Fuzik et al., 2016), thus validating our protocol. Out of 32 cells processed in these experiments, only 7 cells could be identified post hoc as projecting to subiculum (Fig. 1B) and showing no contamination by glial and excitatory neuronal transcripts (Fig. 1F). We also confirmed the VIP-LRP phenotype by verifying the M2R+ expression in dendrites of recorded cells used for RNA-seq (Fig 1C). These cells were included in subsequent gene expression analysis (Fig. 2-4). The gene expression data are presented in TPM values using a base-2 log scale. Investigators were blind to the condition (sample vs control) during analysis. Accession numbers: The RNA-

114

sequencing raw data associated with this manuscript have been uploaded to GEO (accession # GSE109755).

Cell reconstruction and immunohistochemistry

For patch-seq validation, immunohistochemical tests were performed on free- floating sections (40 µm thick) obtained with Leica VT1000S vibratome from 3–4 mice (20 sections/animal) per condition. VIP-eGFP mice were perfused with 4% PFA and the brains were sectioned. Sections were permeabilized with 0.25% Triton X- 100 in PBS containing normal donkey serum (10%) and incubated overnight at 4 °C with primary antibodies (1:1000, chicken anti-eGFP, GFP-1020, Aves; 1:2000, rat anti-M2R, MAB367, Millipore; 1:500, goat anti-mGluR1a, mGluR1a-Go-Af1220, Frontier Institute; 1:250, rabbit anti-NPY, 22940, Immunostar; 1:500, rabbit anti- Netrin G1, GTX115637, Genetex; 1:500, rabbit anti-proenkephalin, LS-C23084, Lifespan Biosciences) followed by the incubation with secondary antibodies (donkey anti-chicken Alexa-488; donkey anti-rat CF-633; donkey anti-goat Cy3; donkey anti- rabbit Alexa-546). For proenkephalin and Netrin G1 immunoreactions, biotinylation was performed to enhance the labeling specificity. Briefly, following overnight incubation of sections with rabbit proenkephalin or Netrin G1 primary antibodies, biotinylated anti-rabbit antibody was applied for 24h followed by streptavidin conjugated AlexaFluor-546 (1:1000). For controlling method specificity, the primary antibodies were omitted, and sections incubated in the full mixture of secondary antibodies. Under such conditions no selective cell labeling was detected. Confocal images were acquired sequentially using a Leica TCS SP5 imaging system coupled with a 488-nm argon, a 543-nm HeNe and a 633-nm HeNe lasers. Z-stacks of

115

biocytin-filled cells were acquired with a 1-μm step and merged for reconstruction in Neurolucida 8.26.2.

Results

This study focused on a population of the M2R-expressing VIP-LRP GABAergic neurons that reside in the oriens/alveus (O/A) of CA1 hippocampus and, in addition to CA1, innervate subiculum (Francavilla et al., 2018). Through preferential synaptic contacts onto interneurons located in the CA1 and conjoint innervation of interneurons and principal cells in subiculum, these cells control the information flow along the hippocampo-subicular axis. Yet, the molecular identity and functional role of VIP-LRPs remain unknown. To explore the genetic basis of the input-output transformations in VIP-LRPs and identify additional discriminant molecular markers that can be used to target these cells selectively, we obtained their transcriptomic profile using single-cell patch RNA-sequencing (Patch-seq) in combination with electrophysiological recordings and post hoc morphological reconstruction (Cadwell et al., 2016; Fuzik et al., 2016). Following patch-clamp recordings of membrane properties and filling the recorded cells with biocytin, the cell cytoplasm was gently aspirated for transcriptomic analysis (Fig. 1A). From the dataset obtained (n = 32 cells), 7 single cell samples satisfied the morphological (Fig. 1B), electrophysiological (Fig. 1B), neurochemical (Fig. 1C) and transcriptomic (Fig. 1D) selection criteria (Materials and Methods) and were used for gene expression analysis. Selected neurons had a resting membrane potential of –60.0 ± 1.0 mV, an input resistance of 209 ± 41.1 MΩ and a membrane capacitance of 76.5 ± 11.2 pF.

They exhibited a regularly spiking firing pattern and a large amplitude Ih current (Ih rectification: 16.1 ± 1.5 mV; Fig. 1B). Post hoc morphological analysis of selected VIP-LRPs revealed horizontally oriented dendrites within CA1 O/A with a single dendritic branch often projecting to the LM, and an axonal arborization within the CA1 O/A and proximal subiculum (Fig. 1B), while neurochemical analysis of these cells confirmed that in addition to VIP, they co-express M2R (Fig. 1C) in line with VIP-LRP phenotype (see also Francavilla et al., 2018, Fig. 1).

116

Following the cDNA library construction and next-generation sequencing, the transcriptomic profiles of selected VIP-LRPs were analyzed (Fig. 1E, Table 1). We found that these cells expressed the common VIP+ interneuron genes (Gad1, 7 of 7 cells; Gad2, 5 of 7 cells; Vip, 7 of 7 cells; Cck, 4 of 7 cells; and Chrm2, 4 of 7 cells) but not the excitatory neuronal (Emx1, Tbr1, 0 of 7 cells) or glial (Gfap, 0 of 7 cells) transcripts (Fig. 1F), consistent with their GABAergic VIP+ interneuron identity. Furthermore, VIPLRPs expressed several genes specific for CGE-derived interneurons (Fig. 1G), consistent with their CGE origin. In searching for commonly expressed genes, we examined the transcripts with TPM>1 across 7 cell samples (Fig. 1E). We extracted the list of common genes between the first two samples, and then explored which genes from this gene list can be found in the next sample. By repeating this process for all samples, we identified 604 common genes that were expressed in all 7 cells (Fig 3A, Table 1). We compared the frequency distribution of all detected genes (total genes) with common genes (Fig 2A, 2B, 2C). The median expression levels of common genes (median of medians: 6.5 log2 TPM) were lager than median expression level of total genes (median of medians: 3.8 log2 TPM, P<0.001 in all cells, one-way ANOVA), indicating the higher expression levels of common genes in all samples.

The ontology of common genes was analyzed using the PANTHER Gene List Analysis (Mi et al., 2017). The products of many of these transcripts belong to important protein families, such as calcium biding proteins (PANTHER protein class: PC00060), cell adhesion molecules (PC00069), cell junction proteins (PC00070), receptors (PC00197), transcription factors (PC00218), extracellular matrix proteins (PC00102), and transmembrane receptor regulatory/adaptor proteins (PC00226, Fig 3B, C). Arranging the common genes by their mean expression values across samples, we identified several highly expressed genes (Fig 3B). For example, these included the calcium biding proteins Myl6 (Myosin Light Chain 6), Calm1 (Calmodulin 1), Myl12b (Myosin Light Chain 12B), and Calr (Calreticulin); the cell adhesion molecule Sparcl1 (SPARC Like 1); the extracellular matrix proteins Mbp (Myelin Basic Protein) and Bsg (Basigin); the transcription factors Tceb2 (Transcription Elongation Factor B) and Park7 (Parkinsonism associated deglycase); and the

117

axonal growth regulator Olmf1 (Olfactomedin).

We next explored the similarity of the transcriptomic profile of VIP-LRPs with other cortical VIP+ interneurons, such as VIP/CR-co-expressing (VIP:CR) interneuron- selective cells and VIP/CCK-co-expressing (VIP:CCK) basket cells (Paul et al., 2017), or hippocampal CA1 interneurons (Harris et al. 2018). We found that many genes that were enriched in cortical VIP:CR interneurons (e.g., Chrna4, Unc5a) or in VIP:CCK basket cells (e.g., Tac2, Cnr1) (Paul et al., 2017) were detected in VIP- LRPs (see table 2). Two VIP:CR-specific genes (Dlgap3, DLG Associated Protein 3 and Ptms, Parathymosin) and one gene specific for long-range projecting somatostatin/nitric oxide synthase-co-expressing (Sst:Nos1) GABAergic neurons (Nnat, Neuronatin) were detected in all 7 VIP-LRP samples. In addition, the Pcp4 (Purkinje Cell Protein 4) mRNAs, which were highly enriched in the long-range projecting Sst:Nos1 cells in hippocampal CA1 (Harris et al., 2018), were also present in all VIP-LRP cell samples. The expression levels of these genes in VIP-LRP samples are shown in Table 1. These data indicate that VIP-LRPs may correspond to an intermediate VIP+ population with transcriptomic profile similar to other VIP+ and long-range projecting cell types.

To predict the intrinsic, synaptic and neuromodulatory properties of VIP-LRPs, we next examined the expression of genes within the major functional families, including ion channels, excitatory and inhibitory inputs, neuromodulatory molecules and receptors and the axon guidance/ CAMs (Table 2). We compared the frequency distribution of these selected genes within functional gene families with total genes in each cell sample (Fig. 2A, 2C, 2D). The distributions followed similar trends in each cell except cell 7. The median expression levels of selected genes (median of medians: 3.4 log2 TPM) were also similar to those of total genes (median of medians:

3.8 log2 TPM, P>0.05 in all cells, one-way ANOVA), pointing to the overall variable expression levels of selected functional genes between samples. Furthermore, our data showed that VIP-LRP cells exhibit the common ion channel genes within the Kv, Nav and Cav families as well as the cyclic nucleotide-regulated HCN ion channel genes, consistent with a prominent Ih current (Fig. 1B, Fig. 4A). In line with previous

118

findings for cortical VIP+ interneurons (Paul et al., 2017), VIP-LRPs showed roughly similar GluA1 and GluA2 content and expressed GluA4, alone with genes encoding for stargazin and the SHISA family (SHISA4 and SHISA9) AMPA receptor (AMPAR) auxiliary subunits (Fig. 4B). In addition, they expressed genes for NR2B and NR2D NMDA receptor (NMDAR) subunits, and both metabotropic glutamate receptor 1 (mGluR1) (Grm1, Fig. 4B) and mGluR5 (Grm5, Fig. 4B). Among the constituents of the GABAA receptor (GABAAR), the α1, α2, α3 and α5 as well as β1 and β3 subunits were detected (Fig. 4B). They also expressed the pleiotropin and connexin29 genes (Ptn and Gjc3, Fig. 4C), both involved in axon myelination (Altevogt et al., 2002; Kuboyama et al., 2015), consistent with a partial myelination of the VIP-LRP axon. Furthermore, VIP-LRPs expressed a large variety of genes involved in neuromodulatory signaling, including acetylcholine (Chrna4), norepinephrin (Adrb1), dopamine (Drd1), serotonin (Htr1d), cannabinoid (Cnr1), opioid (Oprd1, Oprl1), neuropeptide Y (Npy1r) and neurotensin (Ntsr2) receptors (Fig. 4D), indicating that VIP-LRPs are likely modulated via mood, reward and stress-activated neural pathways. In addition to VIP, these cells expressed neuropeptide Y (NPY) and proenkephalin (Penk) (Fig. 4D), suggesting that VIP, NPY and enkephalin peptides can be co-released by VIP-LRPs under certain conditions. Finally, they expressed the common axon guidance and CAM genes (Fig. 4E), with Ntng1 and Cdh8 involved in the formation of long-range projections, among others (Lin et al., 2003).

To validate our patch-seq results, we took advantage of a strong M2R expression in VIP-LRPs (Francavilla et al., 2018) and performed a triple immunolabeling for VIP- eGFP, M2R and some proteins of interest predicted from the cell transcriptome (Fig. 5A). In addition to Ntng1 identified previously as a long-range axonal marker (Lin et al., 2003), we focused on several proteins detected in hippocampal long-range projecting GABAergic neurons, including the mGluR1a, NPY and Penk (Jinno et al., 2007; Fuentealba et al., 2008). We calculated the percentage of VIP-GFP+ cells located in CA1 O/A that co-express M2R (VIP-GFP+/M2R+) and the marker of interest (mGluR1, NPY, Ntng1 or Penk) and compared these data with VIP- GFP+/M2R– cells. We found that mGluR1a, Ntng1, NPY and Penk were all detected in M2R+ VIP-LRPs as predicted from their transcriptomes (Fig. 5A, 5C). However,

119

these molecules were also expressed in VIP-GFP+/M2R– cells (Fig. 5B, 5C), indicating that they are not specific for VIP-LRP neurons. In contrast, a large proportion of VIP-GFP+/M2R+ cells co-expressed Penk, which was detected in only a small number of VIP-GFP+/M2R– cells, indicating that Penk may be more frequently expressed in VIP-LRP neurons (Fig. 5A, 5B, 5C). Collectively, these data validate results obtained with transcriptomic analysis and identify the VIP-LRPs as a member of the VIP+ population continuum, which shares molecular identity and, likely, physiological function with other VIP interneurons, but also has genes associated with specific features (long-range projections and axon myelination) that allow for long-distance coordination.

Discussion

We show that VIP-LRPs in the mouse CA1 hippocampus form a distinct population of VIP+ cells that can be distinguished based on the M2R and Penk co-expression. Using a transcriptomic analysis based on the patch-sequencing of morphologically identified cells, we found that VIP-LRPs share molecular identity with both VIP+ interneuron-selective cells and the CCK-co-expressing basket cells (Paul et al., 2017). The additional molecular marker identified here for VIP-LRPs, the proenkephalin, can be used for combinatorial targeting of this VIP+ cell type, which will be necessary to provide insights into the physiological organization and behaviour-related functions of VIP-LRPs.

Different types of the LRP GABAergic neurons have been described in neocortical and hippocampal networks (Jinno et al., 2007; Miyashita et al., 2007; Melzer et al., 2012; Basu et al., 2016). Still, very little is known regarding the physiological properties and network function of these cells due to their low density and absence of specific markers that can be used to target the GABAergic LRPs selectively. Our transcriptome analysis can predict some physiological properties and network-state dependent recruitment of these cells. In particular, VIP-LRPs exhibit the GluA1, GluA2 and GluA4 of AMPAR core subunits, as well as the NR1, NR2B and NR2D NMDAR subunits, suggesting that their excitatory synapses contain the Ca2+-

120

impermeable AMPARs and NMDARs. The latter, when validated, may indicate the generation of the NMDAR-dependent dendritic spikes and Hebbian forms of plasticity at excitatory inputs to these cells (Topolnik, 2012). VIP-LRPs also contain both the mGluR1 and mGluR5, which may be involved in the regulation of the NMDAR-independent Hebbian LTP (Perez et al., 2001; Topolnik, 2012). They express the α1, α2, α3 and α5 as well as β1 and β3-containing GABAARs, pointing to input-specific GABAAR composition, with slow α3 and α5-containing synapses at some inhibitory inputs (Ali and Thomson, 2008; Vargas-Caballero et al., 2010; Salesse et al., 2011; Magnin et al., 2018). Thus, VIP-LRPs can be sensitive to α3- and the α5-GABAAR-specific pharmacological manipulations and contribute to the associated cognitive and anxiogenic effects (Navarro et al., 2002; Mohler and Rudolf, 2017). In addition, these cells express a large variety of neuromodulatory receptors, including acetylcholine, norepinephrine, dopamine and 5-HT receptors, indicating that VIP-LRP recruitment can be controlled by the brain state-dependent modulatory subcortical projections. In terms of output, VIP-LRPs express NPY and Penk and, therefore, may release these neuropeptides in addition to VIP and GABA. Furthermore, these cells feature numerous neuropeptide receptors, including cannabinoid, opioid, NPY and neurotensin receptors, highlighting that they are themselves under control of multiple peptidergic influences. In contrast to most GABAergic interneurons, VIP-LRPs exhibit a partially myelinated axon and, consistent with this feature, express genes involved in axon myelination, such as Ptn and Gjc3 (Altevogt et al., 2002; Kuboyama et al., 2015). The latter genes together with Ntng1 and Cdh8, which have been involved in long-distance axon guidance, position the VIP-LRPs apart from their VIP+ counterparts, providing for longdistance coordination. Collectively, this study reveals a molecular profile of a rare subtype of VIP+ projecting neuron that can be recruited through multiple cortical and subcortical inputs to modulate the hippocampo-subicular dialogue.

Finally, emerging evidence has linked the dysfunction of VIP+ interneurons with several neurological and psychiatric disorders. The reduction in NPY and VIP gene expression, as well as the duplication of the VIP receptor Vipr2 gene copies has been associated with schizophrenia and bipolar disorder in humans (Vacic et al.,

121

2011; Fung et al., 2014). Moreover, Cunniff and Sohal (2017) reported that VIP+ interneurons in the prefrontal cortex of two mouse models of the autism spectrum disorder show abnormal cholinergic responses. In addition, the reduction in NPY+ and VIP+ interneurons has been detected in the hippocampus of human subjects with temporal lobe epilepsy (TLE; De Lanerolle et al., 1989; Sundstrom et al., 2001) and in the pilocarpine-induced experimental chronic TLE (David and Topolnik, 2017). Taken together, these findings suggest that impaired circuit coordination by VIP+ interneurons may be associated with several CNS disorders. Among VIP+ cells, the subiculum-projecting CA1 VIP-LRPs are well positioned to coordinate the fear memory encoding and retrieval (Roy et al., 2017), the episode recollection (Eldridge et al., 2005), and the initiation and maintenance of epileptic discharges in TLE (Stafstrom, 2005). The transcriptomic profile of VIP-LRPs identified in our study may be useful for functional analysis of this cell type, necessary for understanding their functional role and the development of the cell- and circuit-specific therapy.

References

Acsády L, Arabadzisz D, Freund T (1996a) Correlated morphological and neurochemical features identify different subsets of vasoactive intestinal polypeptide-immunoreactive interneurons in rat hippocampus. Neuroscience 73:299–315.

Acsady L, Gorcs TJ, Freund TF (1996b) Different populations of vasoactive intestinal polypeptide- immunoreactive interneurons are specialized to control pyramidal cells or interneurons in the hippocampus Neuroscience 73:317-334

Ali AB, Thomson AM (2008) Synaptic alpha 5 subunit-containing GABAA receptors mediate IPSPs elicited by dendrite-preferring cells in rat neocortex Cereb Cortex 18:1260-1271 doi:10.1093/cercor/bhm160

Altevogt BM, Kleopa KA, Postma FR, Scherer SS, Paul DL (2002) Connexin29 is uniquely distributed within myelinating glial cells of the central and peripheral nervous systems J Neurosci 22:6458-6470

Ayzenshtat I, Karnani MM, Jackson J, Yuste R (2016) Cortical Control of Spatial Resolution by VIP+ Interneurons J Neurosci 36:11498-11509 doi:10.1523/JNEUROSCI.1920-16.2016

Basu J, Zaremba JD, Cheung SK, Hitti FL, Zemelman BV, Losonczy A, Siegelbaum SA (2016) Gating of hippocampal activity, plasticity, and memory by entorhinal cortex long-range inhibition Science 351:aaa5694 doi:10.1126/science.aaa5694

Bayraktar T, Welker E, Freund TF, Zilles K, Staiger JF (2000) Neurons immunoreactive for vasoactive intestinal polypeptide in the rat primary somatosensory cortex: morphology and spatial relationship to barrel-related columns J Comp Neurol 420:291-304

122

Brennecke P et al. (2013) Accounting for technical noise in single-cell RNA-seq experiments Nat Methods 10:1093-1095 doi:10.1038/nmeth.2645

Cadwell CR et al. (2016) Electrophysiological, transcriptomic and morphologic profiling of single neurons using Patch-seq Nat Biotechnol 34:199-203 doi:10.1038/nbt.3445

Chamberland S, Salesse C, Topolnik D, Topolnik L (2010) Synapse-Specific Inhibitory Control of Hippocampal Feedback Inhibitory Circuit Front. Cell. Neurosci.,

Cunniff M, Sohal V (2017) 336. VIP interneuron cholinergic responses are altered in multiple models of autism. Biol Psychiatry 81:S138 doi:10.1016/j.biopsych.2017.02.351

David LS, Topolnik L (2017) Target-specific alterations in the VIP inhibitory drive to hippocampal GABAergic cells after status epilepticus. Exp Neurol 292:102-112 doi:10.1016/j.expneurol.2017.03.007 de Lanerolle NC, Kim JH, Robbins RJ, Spencer DD (1989) Hippocampal interneuron loss and plasticity in human temporal lobe epilepsy. Brain Res 495:387-395

Distinct Neural Circuits for the Formation and Retrieval of Episodic Memories. Cell 170:1000-1012 e1019 doi:10.1016/j.cell.2017.07.013

Eldridge LL, Engel SA, Zeineh MM, Bookheimer SY, Knowlton BJ (2005) A dissociation of encoding and retrieval processes in the human hippocampus. J Neurosci 25:3280-3286 doi:10.1523/JNEUROSCI.3420-04.2005

Foldy C, Darmanis S, Aoto J, Malenka RC, Quake SR, Sudhof TC (2016) Single-cell RNAseq reveals cell adhesion molecule profiles in electrophysiologically defined neurons Proc Natl Acad Sci U S A 113:E5222-5231 doi:10.1073/pnas.1610155113

Francavilla R, Villette V, Luo X, Chamberland S, Munoz-Pino E, Camire O, Wagner K, Kis V, Somogyi P, Topolnik L (2018) Connectivity and network state-dependent recruitment of longrange VIP- GABAergic neurons in the mouse hippocampus. bioRxiv doi: 10.1101/364547

Fu Y, Tucciarone JM, Espinosa JS, Sheng N, Darcy DP, Nicoll RA, Huang ZJ, Stryker MP (2014) A cortical circuit for gain control by behavioral state. Cell 156:1139-1152 doi:10.1016/j.cell.2014.01.050

Fuentealba P, Tomioka R, Dalezios Y, Marton LF, Studer M, Rockland K, Klausberger T, Somogyi P (2008) Rhythmically active enkephalin-expressing GABAergic cells in the CA1 area of the hippocampus project to the subiculum and preferentially innervate interneurons. J Neurosci 28:10017-10022 doi:10.1523/JNEUROSCI.2052-08.2008

Fung SJ, Fillman SG, Webster MJ, Shannon Weickert C (2014) Schizophrenia and bipolar disorder show both common and distinct changes in cortical interneuron markers. Schizophr Res 155:26-30 doi:10.1016/j.schres.2014.02.021

Fuzik J, Zeisel A, Máté Z, Calvigioni D, Yanagawa Y, Szabó G, Linnarsson S, Harkany T (2016) Integration of electrophysiological recordings with single-cell RNA-seq data identifies neuronal subtypes. Nat Biotechnol. 34:175-183.

Gonchar Y, Wang Q, Burkhalter A (2007) Multiple distinct subtypes of GABAergic neurons in mouse visual cortex identified by triple immunostaining Front Neuroanat 1:3 doi:10.3389/neuro.05.003.2007

123

Grun D, van Oudenaarden A (2015) Design and Analysis of Single-Cell Sequencing Experiments Cell 163:799-810 doi:10.1016/j.cell.2015.10.039

Harris KD, Hochgerner H, Skene NG, Magno L, Katona L, Bengtsson Gonzales C, Somogyi P, Kessaris N, Linnarsson S, Hjerling-Leffler J (2018) Classes and continua of hippocampal CA1 inhibitory neurons revealed by single-cell transcriptomics. PLoS Biol 16:e2006387 doi:10.1371/journal.pbio.2006387

He M, Tucciarone J, Lee S, Nigro MJ, Kim Y, Levine JM, Kelly SM, Krugikov I, Wu P, Chen Y, Gong L, Hou Y, Osten P, Rudy B, Huang ZJ (2016) Strategies and Tools for Combinatorial Targeting of GABAergic Neurons in Mouse Cerebral Cortex. Neuron 92:555 doi:10.1016/j.neuron.2016.10.009

Jinno S et al. (2007) Neuronal diversity in GABAergic long-range projections from the hippocampus J Neurosci 27:8790-8804 doi:10.1523/JNEUROSCI.1847-07.2007

Klausberger T, Somogyi P (2008) Neuronal diversity and temporal dynamics: the unity of hippocampal circuit operations. Science 321:53-57 doi:10.1126/science.1149381321/5885/53 [pii] 27.

Kuboyama K, Fujikawa A, Suzuki R, Noda M (2015) Inactivation of Protein Tyrosine Phosphatase Receptor Type Z by Pleiotrophin Promotes Remyelination through Activation of Differentiation of Oligodendrocyte Precursor Cells J Neurosci 35:12162-12171 doi:10.1523/JNEUROSCI.2127- 15.2015

Lee S, Kruglikov I, Huang ZJ, Fishell G, Rudy B (2013) A disinhibitory circuit mediates motor integration in the somatosensory cortex Nat Neurosci 16:1662-1670 doi:10.1038/nn.3544

Lin JC, Ho WH, Gurney A, Rosenthal A (2003) The netrin-G1 ligand NGL-1 promotes the outgrowth of thalamocortical axons Nat Neurosci 6:1270-1276 doi:10.1038/nn1148

Madisen L et al. (2015) Transgenic mice for intersectional targeting of neural sensors and effectors with high specificity and performance Neuron 85:942-958 doi:10.1016/j.neuron.2015.02.022

Magnin E, Francavilla R, Amalyan S, Gervais E, David L, Luo X, Topolnik L (2018) Input specific synaptic location and function of the alpha5 GABAA receptor subunit in the mouse CA1 hippocampal neurons J Neurosci in press

Melzer S, Michael M, Caputi A, Eliava M, Fuchs EC, Whittington MA, Monyer H (2012) Longrange- projecting GABAergic neurons modulate inhibition in hippocampus and entorhinal cortex Science 335:1506-1510 doi:10.1126/science.1217139

Mi H, Huang X, Muruganujan A, Tang H, Mills C, Kang D, Thomas PD (2017) PANTHER version 11: expanded annotation data from Gene Ontology and Reactome pathways, and data analysis tool enhancements. Nucleic Acids Res 45:D183-D189 doi:10.1093/nar/gkw1138

Miyashita T, Rockland KS (2007) GABAergic projections from the hippocampus to the retrosplenial cortex in the rat Eur J Neurosci 26:1193-1204 doi:10.1111/j.1460-9568.2007.05745.x

Miyoshi G et al. (2015) Prox1 Regulates the Subtype-Specific Development of Caudal Ganglionic Eminence-Derived GABAergic Cortical Interneurons J Neurosci 35:12869-12889 doi:10.1523/JNEUROSCI.1164-15.2015

Mohler H, Rudolph U (2017) Disinhibition, an emerging pharmacology of learning and memory F1000Res 6 doi:10.12688/f1000research.9947.1

124

Navarro JF, Buron E, Martin-Lopez M (2002) Anxiogenic-like activity of L-655,708, a selective ligand for the benzodiazepine site of GABA(A) receptors which contain the alpha-5 subunit, in the elevated plus-maze test Prog Neuropsychopharmacol Biol Psychiatry 26:1389-1392

Paul A, Crow M, Raudales R, He M, Gillis J, Huang ZJ (2017) Transcriptional Architecture of Synaptic Communication Delineates GABAergic Neuron Identity Cell 171:522-539 e520 doi:10.1016/j.cell.2017.08.032

Perez Y, Morin F, Lacaille JC (2001) A hebbian form of long-term potentiation dependent on mGluR1a in hippocampal inhibitory interneurons Proc Natl Acad Sci U S A 98:9401-9406 doi:10.1073/pnas.161493498

Petilla Interneuron Nomenclature G et al. (2008) Petilla terminology: nomenclature of features of GABAergic interneurons of the cerebral cortex Nat Rev Neurosci 9:557-568 doi:10.1038/nrn2402

Pi HJ, Hangya B, Kvitsiani D, Sanders JI, Huang ZJ, Kepecs A (2013) Cortical interneurons that specialize in disinhibitory control Nature 503:521-524 doi:10.1038/nature12676

Porter JT, Cauli B, Staiger JF, Lambolez B, Rossier J, Audinat E (1998) Properties of bipolar VIPergic interneurons and their excitation by pyramidal neurons in the rat neocortex Eur J Neurosci 10:3617- 3628

Risso D, Ngai J, Speed TP, Dudoit S (2014) Normalization of RNA-seq data using factor analysis of control genes or samples Nat Biotechnol 32:896-902 doi:10.1038/nbt.2931

Roy DS, Kitamura T, Okuyama T, Ogawa SK, Sun C, Obata Y, Yoshiki A, Tonegawa S (2017)

Salesse C, Mueller CL, Chamberland S, Topolnik L (2011) Age-dependent remodelling of inhibitory synapses onto hippocampal CA1 oriens-lacunosum moleculare interneurons J Physiol 589:4885- 4901 doi:10.1113/jphysiol.2011.215244

Somogyi J, Baude A, Omori Y, Shimizu H, El Mestikawy S, Fukaya M, Shigemoto R, Watanabe M, Somogyi P (2004) GABAergic basket cells expressing cholecystokinin contain vesicular glutamate transporter type 3 (VGLUT3) in their synaptic terminals in hippocampus and isocortex of the rat. Eur J Neurosci 19:552-569 doi.org/10.1111/j.0953-816X.2003.03091.x

Somogyi P, (2010) Hippocampus: Intrinsic Organisation. In: Shepherd G, Grillner S (eds) Handbook of Brain Microcircuits, Oxford University Press, Oxford, pp 148-664.

Stafstrom CE (2005) The role of the subiculum in epilepsy and epileptogenesis. Epilepsy Curr

Sundstrom LE, Brana C, Gatherer M, Mepham J, Rougier A (2001) Somatostatin- and neuropeptide Y-synthesizing neurones in the fascia dentata of humans with temporal lobe epilepsy. Brain 124:688- 697.

Taniguchi H, He M, Wu P, Kim S, Paik R, Sugino K, Kvitsiani D, Fu Y, Lu J, Lin Y, Miyoshi G, Shima Y, Fishell G, Nelson SB, Huang ZJ (2011) A resource of Cre driver lines for genetic targeting of GABAergic neurons in cerebral cortex. Neuron 71:995-101 doi:10.1016/j.neuron.2011.07.026

Tasic B et al. (2016) Adult mouse cortical cell taxonomy revealed by single cell transcriptomics Nat Neurosci 19:335-346 doi:10.1038/nn.4216

Topolnik L (2012) Dendritic calcium mechanisms and long-term potentiation in cortical inhibitory

125

interneurons Eur J Neurosci 35:496-506 doi:10.1111/j.1460-9568.2011.07988.x

Tyan L et al. (2014) Dendritic inhibition provided by interneuron-specific cells controls the firing rate and timing of the hippocampal feedback inhibitory circuitry J Neurosci 34:4534-4547 doi:10.1523/JNEUROSCI.3813-13.2014

Vacic V et al. (2011) Duplications of the neuropeptide receptor gene VIPR2 confer significant risk for schizophrenia. Nature 471:499-503 doi:10.1038/nature09884

Vargas-Caballero M, Martin LJ, Salter MW, Orser BA, Paulsen O (2010) alpha5 Subunitcontaining GABA(A) receptors mediate a slowly decaying inhibitory synaptic current in CA1 pyramidal neurons following Schaffer collateral activation Neuropharmacology 58:668-675 doi:10.1016/j.neuropharm.2009.11.005

Zeisel A et al. (2015) Brain structure. Cell types in the mouse cortex and hippocampus revealed by single-cell RNA-seq Science 347:1138-1142 doi:10.1126/science.aaa1934

Figures and tables

Figure 1.

126

Figure 2.

127

Figure 3.

Figure 4.

128

Figure 5.

Figure Caption

Fig. 1 Patch-sequencing of VIP-LRPs

A. Schematic of patch RNAseq experiment. B. Example of Neurolucida reconstruction and membrane potential responses to current injections of a processed VIP-LRP. Dendrites and soma are shown in green, axon in red; current steps of–400 pA and +60 pA were applied to evoke membrane potential responses. C. Immunolabelling for M2R in dendrites of a VIP-LRP cell used for patch-seq analysis. Streptavidin-Alexa Fluor 546 labelling shows the morphology of dendrites filled during recording with biocytin. D. Bar graph showing the concentration of cDNA (with red line indicating the 200 pg/ml threshold used as a criterion for sample inclusion as in Cadwell et al., 2016). E. Bar graph showing the number of transcripts per sample (with TPM > 1) generated in single cell sequencing experiments. F. Heat-map showing the expression of some VIP interneuron genes in VIP-LRP neurons (Gad1, Gad2; Vip, Cck, Chrm2) and the absence of excitatory (Emx1) and glial (Gfap) transcripts in single- cell samples selected for further analysis (n = 7 cells; based on post hoc identification of axonal 129

projections and quality control criteria). Negative control (Ctl”-“) line illustrates the expression of the same transcripts in sample obtained by aspiration of extracellular matrix. Positive control (Ctl”+“) line shows the expression of genes following the sequential aspiration of three GFP+ interneurons (without post hoc anatomical analysis) in the same patch-pipette for combined analysis. G. Summary boxplots of some MGE- and CGE-specific genes show preferential expression of CGE-genes in VIP- LRP cells.

Fig. 2 Frequency distribution of total genes, common genes and genes selected based on functional families for each single cell sample

A. Histograms of gene frequencies (blue: total genes; green: common genes expressed in all 7 samples and shown in table1; red: genes selected based on functional families and shown in table 2) at different expression levels for each single cell (c1–c7). Horizontal axis: expression level shown in log2 TPM. Vertical axis: gene frequencies that fall into each bin of the expression level. The bin size was determined automatically in Igor Pro 4.0. The frequency distribution of total genes, selected genes, and common genes follow the same trends from low to high expression values. The medians of gene expression levels are depicted as dashed lines and were similar between total genes and selected genes within the same cell. The common genes (green) showed a significantly higher expression level in all samples (P<0.001 for all cells, one-way ANOVA). B–D. Superimposed frequency distribution of common genes (B) total genes (C) and selected genes (D). Different colors represent different single cell samples. Each dot represents the number of genes within a 0.1 log2 TPM bin.

Fig. 3 Heat maps showing the expression values of common genes in all single cell samples

A. Venn diagram of 604 common genes shared by 7 single cell samples. Each circle indicates the total genes within a single cell sample. The number in the overlapped area indicates the number of common genes (C1–C7 correspond to individual cells). B. Heat-map indicating the expression levels of common genes. The mean expression values across samples were calculated for each gene and then arranged from top to bottom (high to low). The gene names are indicated on the right. The single cell samples are shown on the bottom (C1–C7). C. Some common genes selected according to protein families. The protein family names were obtained from PANTHER classification system (Mi et al., 2017). The gene names are shown on the right of the heat map. The protein family names are shown on the left side of the heat map. The single cell samples are shown on the bottom (C1–C7). The color scale indicates the logarithm of TPM values to base 2 from low (black) to high (yellow).

Figure 4: Boxplots showing selected genes with various expression levels across single cell samples.

Summary boxplots showing the expression levels of selected genes for ion channels (A, including Kv channels, Kv regulatory elements, Cv channels and HCN channels), glutamate and GABA receptors (B, including AMPARs and auxiliary elements, NMDARs, mGluRs and GABARs), myelination factors (C), neuromodulatory/ neuropeptide signalling, (D) and axon guidance/ cell adhesion molecules (CAMs, E). The upper and lower whiskers show the maximum and minimum values respectively, the lower border and upper border of the box show the 1st and 3rd quartile, the line in the middle of the box indicates the median. The gene expression values are expressed as the logarithm of TPM values to base 2. Gene names are in italic.

Figure 5 Novel markers for VIP-LRP cells obtained following gene expression validation with immunohistochemical analysis.

A. Representative confocal images showing triple immunolabelling in CA1 O/A for VIP-GFP (green),

130

M2R (cyan) and other markers: mGluR1a, NPY, netrin G1, proenkephalin (all in red) as well as the overlay of all. In A, GFP cells co-expressed M2R and other markers. B. Representative confocal images showing that mGluR1, NPY and Ntng1 were also present in some GFP+M2R– cells. However, the expression of Penk+ in GFP+M2R– cells was not frequent. The colors representing different markers are the same as in A. C. Bar graphs showing the percentage of GFP cells co-expressing M2R and other markers. mGluR1, NPY and Ntng1 were present in both GFP+M2R+ and GFP+M2R– cells, whereas Penk was preferentially co-expressed with GFP and M2R.

Table 1 Gene expression levels of common gene transcripts presented in 7 single cell samples. (see Appendices)

Table 2 Gene expression levels by functional gene categories in VIP-LRPs (see Appendices)

131

Chapter 4 Integrated article 3: Synaptic properties and the network state-dependent recruitment of the VIP interneuron- specific interneurons in the CA1 hippocampus

Résumé

Dans l'hippocampe, il existe une population hautement spécialisée de cellules inhibitrices spécifiques aux interneurones (IS) qui coordonnent l'activité des circuits inhibiteurs locaux. Bien que l'on pense que la désinhibition joue un rôle critique dans l'apprentissage, la contribution des cellules IS dans l'activité du réseau reste incertaine. Ici, nous révélons les propriétés synaptiques des cellules IS de type 3 situées dans la région CA1 qui coexpriment le peptide intestinal vasoactif et la calrétinine (IS3) et nous démontrons comment elles sont recrutées au cours des différents états du réseau chez la souris éveillée. À l'aide d'enregistrements par patch-clamp et de libération du glutamate par photostimulation à deux photons, nous avons constaté que les cellules IS3 envoient des pics de courant en réponse à la fois aux voies collatérales de Schaffer et aux voies temporoammoniques. En utilisant des modèles synaptiques de ces signaux et un modèle computationnel de cellules IS3 avec des niveaux d'activité synaptique répliquant ceux vus in vivo, nous avons prédit que les cellules IS3 pourraient être stimulées de manière rythmée pendant les oscillations thêta et de manière transitoire. Cependant, l'imagerie calcique à deux photons chez des souris éveillées a révélé une grande variabilité dans l’activité des cellules IS3 dans les différents états comportementaux. En règle générale, nos résultats montrent que des signaux transitoires somatiques de calcium (CaTs) sont survenus au cours des oscillations thêta associés à la locomotion et ont suggéré le déclenchement préférentiel des IS3 autour de la phase ascendante et du pic de l'onde thêta. De plus, les CaTs ont été détectés pendant l'immobilité, mais non couplés aux ondes à pics aigus. Ainsi, alors que les propriétés synaptiques des cellules IS3 peuvent prédire un signal particulier, des facteurs supplémentaires

132

peuvent moduler leur recrutement au cours des différents comportements et états du réseau.

Abstract

In the hippocampus, a highly specialized population of vasoactive intestinal peptide- expressing interneuron-specific (IS) inhibitory cells coordinates the activity of GABAergic interneurons. While disinhibition may play a critical role in hippocampal learning, the synaptic properties and recruitment of IS cells remain unknown. Here, we found that the CA1 type 3 IS cells (IS3) receive excitatory inputs through the Schaffer collateral and the temporoammonic pathways with synapse-specific properties. The computational models predicted that these synaptic inputs could rhythmically drive the IS3 cell firing during theta oscillations and sharp-wave ripples. However, recordings in awake mice revealed a large variability in IS3 cell recruitment, with preferential recruitment of these cells in relation to theta-run epochs and silence during sharp-wave ripples. Taken together, these data indicate that, while synaptic properties of IS3 cells predict a particular output, additional factors may modulate the cell recruitment during different behavioral and network states.

133

Introduction

In the hippocampus, GABAergic interneurons coordinate the activity of principal cells (PCs) and are themselves controlled through a variety of local and long-range inhibitory mechanisms (Chamberland and Topolnik, 2012; Klausberger and Somogyi, 2008). Interneuron-specific (IS) inhibitory interneurons play a unique role in this circuitry as they provide synaptic inhibition exclusively to GABAergic cells. Three distinct types of IS interneurons have been described in the CA1 hippocampus on the basis of their axonal projections and neurochemical features, with vasoactive intestinal peptide (VIP) being identified as a major IS cell marker (Acsády et al., 1996a,b; Gulyás et al., 1996). These cells reside in different CA1 layers and may integrate major excitatory inputs converging onto the CA1 area, including the Schaffer collaterals (SC) from CA3 pyramidal cells, local axon collaterals from CA1 pyramidal cells, and the temporoammonic (TA) path afferents from the entorhinal cortex. Integrated differentially by individual interneuron types (Ali and Thomson, 1998; Glickfeld and Scanziani, 2006; Losonczy et al., 2002; Pouille and Scanziani, 2004), CA1 afferent inputs determine the cell type-specific contribution of a given interneuron type to ongoing network activity during specific behavioral states (Katona et al., 2014; Klausberger et al., 2003; Lapray et al., 2012; Varga et al., 2012). However, the synaptic and integrative properties of the CA1 IS interneurons remain unknown.

The VIP/calretinin-co-expressing type 3 IS cells (IS3s) innervate preferentially the somatostatin-expressing (SOM+) oriens lacunosum-moleculare (OLM) interneurons (Acsády et al., 1996b; Chamberland et al., 2010; Tyan et al., 2014). Synchronous activation of IS3 cells is capable of coordinating the ﬁring rate and timing of OLM interneurons (Tyan et al., 2014), which in vivo ﬁre preferentially at the trough of theta oscillations and remain silent during sharp wave-associated ripples (SWRs) (Katona et al., 2014; Klausberger et al., 2003; Lapray et al., 2012; Varga et al., 2012). While SOM+ OLM cells have been considered essential for generation of theta oscillations (Gillies et al., 2002; Gloveli et al., 2005), selective manipulation with CA1 SOM+ interneuron population had almost no impact on the presence of intrinsic theta

134

rhythm in the isolated hippocampus in situ. Yet, SOM+ interneurons were important in modulating the TA input integration by PCs (Amilhon et al., 2015). Moreover, silencing of CA1 SOM+ cells increased the burst ﬁring of PCs (Lovett-Barron et al., 2012; Royer et al., 2012) likely through facilitated integration of the SC and TA inputs (Amilhon et al., 2015; Lovett-Barron et al., 2014). In addition to distal dendrites of PCs, OLM cells also target the local stratum radiatum (RAD) interneurons (Leão et al., 2012; Sik et al., 1995). The OLM-to-interneuron circuit thus sub-serves a role in CA1 PC disinhibition that may facilitate selection of the intrahippocampal information and enhance the CA3 gating (Leão et al., 2012; Siwani et al., 2018). Interestingly, recent modeling work has shown that the OLM-to-interneuron pathway is critical in producing a robust theta output in the isolated hippocampus (Chatzikalymniou and Skinner, 2018). Thus, IS3 cell input to SOM+ interneurons, and OLM cells in particular, may act as a higher-level control mechanism regulating the gating of the SC vs TA inputs in the CA1 area. Whether IS3 cells are recruited during in vivo theta oscillations and ﬁre synchronously remains unknown. But if this is indeed the case, IS3 cells’ role in the CA1 will be to pace the OLM cell activity for rhythmic gating of CA1 excitation during hippocampus-dependent learning.

In this study, we examined the synaptic and integrative properties of CA1 IS3 cells in hippocampal slices and investigated their contribution to network oscillations in awake mice. We found that IS3 cells can fire spikes in response to both the SC and TA input activation. Moreover, the TA-specific NMDA receptor expression may allow for the significant increase in the IS3 cell firing probability during repetitive TA activation. Integrating intrinsic and synaptic properties of IS3 cells into computational models and simulating an in vivo-like state, we could predict rhythmic IS3 cell recruitment during theta-oscillations and SWRs. However, in vivo Ca2+ imaging in awake mice showed variable recruitment of these cells during theta-oscillations and no correlation with SWRs recorded during immobility. Together, these results indicate that while IS3 cells are able to fire in response to both the SC and TA inputs and are preferentially activated during theta oscillations, additional factors may modulate their activity.

135

Results

Synaptic properties of IS3 interneurons

We took advantage of two transgenic mouse lines (VIP-eGFP and VIP-Cre) that express enhanced green fluorescent protein (eGFP) or Cre recombinase in VIP- expressing (VIP+) interneurons (David and Topolnik, 2017; Tyan et al., 2014) (Figure 1A). The VIP-Cre mice were crossed with a reporter Ai9 line to achieve the tdTomato expression in VIP+ cells (Madisen et al., 2010). Consistent with previous findings on VIP expression in the hippocampal interneurons (Acsády et al., 1996a,b; Köhler, 1982; Léránth et al., 1984), in both mouse models, fluorescently-labeled cells were located throughout the CA1, with half of these having a vertical bipolar orientation that resembled the IS3 cell morphology (Figure 1A) (Tyan et al., 2014). Indeed, 48.8% of VIP+ interneurons in the CA1 area (375/683 cells, n = 3 VIP-eGFP mice; 319/739 cells, n = 2 VIP-tdTomato mice) were co-expressing calretinin (CR) consistent with the IS3 neurochemical profile (Acsády et al., 1996a). The remainder were VIP+/CR- cells, including a sparsely distributed population of cholecystokinin-co-expressing (CCK+) basket cells (oriens/alveus: 15.2%, pyramidal layer: 10.2%, RAD: 25.8%, LM: 22.2% of total VIP+ interneuron number/layer; average data for two mouse lines; Figure 1, Figure Supplement 1). We characterized the morphological and electrophysiological properties of VIP+ cells by making targeted patch-clamp recordings from CA1 RAD VIP+ interneurons in slices from both mouse lines. All biocytin-filled VIP+ cells included in this analysis (n = 68) had soma located in the RAD, bipolar or monopolar dendrites extending to the LM and an axon arborizing within the oriens/alveus (Figure 1B-C), consistent with the IS3 cell morphology (Acsády et al., 1996a; Chamberland et al., 2010; Tyan et al., 2014). These cells had high input resistance (VIP-eGFP mouse line: 569 ± 38 MOhm; VIP-tdTomato mouse line: 505 ± 91 MOhm) and slow membrane time constant (VIP-eGFP: 25.5 ± 1.8 ms, VIP-tdTomato: 23.7 ± 1.3 ms; mean ± SEM here and in all subsequent text) (n = 19, Figure 1, Figure Supplement 2C). Depolarizing steps elicited slow action potentials (half-width, VIP-eGFP mouse line: 1.2 ± 0.06 ms; VIP-tdTomato mouse line: 1.3 ± 0.2 ms) with regularly spiking or stuttering firing pattern (Figure 1, Figure Supplement

136

2A–C). The IS3 cell membrane and morphological properties were similar between the two mouse lines except the membrane capacitance, which was lower in VIP- tdTomato mice likely due to a smaller soma size (Figure 1, Figure Supplement 2) (David and Topolnik, 2017; Tyan et al., 2014). Therefore, both mouse models were used for studying the synaptic properties of IS3 cells.

Spontaneous excitatory synaptic drive was very low in these cells (Figure 1D). In voltage clamp recordings, spontaneous excitatory postsynaptic currents (EPSCs) acquired at -70 mV in the presence of gabazine had the average frequency of 0.07 ± 0.01 Hz and the average amplitude of 19.9 ± 2.5 pA (n = 6). In contrast, the frequency of spontaneous inhibitory postsynaptic currents (IPSCs) recorded at the reversal potential for EPSCs (0 mV) was signiﬁcantly higher (1.64 ± 0.13 Hz, p < 0.001; ANOVA; average amplitude of IPSCs: 20.3 ± 2.8 pA, n = 6), indicating that under basal conditions in slices the IS3 cells receive a dominant inhibitory drive.

The laminar location of IS3 cell dendrites suggests that these cells may be preferentially driven via SC and TA inputs from CA3 pyramidal cells and entorhinal cortex, respectively. Accordingly, electrical stimulation in the RAD to activate the SC input elicited kinetically fast and large amplitude EPSCs in IS3 cells (amplitude = 24.0 ± 3.8 pA, rise time = 1.0 ± 0.3 ms, decay tau = 3.9 ± 1.0 ms, n = 9; Figure 1E- F), which exhibited a linear current-voltage (I-V) relationship (Figure 1F), in line with a major role of the GluA2-containing Ca2+-impermeable AMPARs at SC-IS3 cell synapses. In contrast to electrical stimulation within the RAD, stimulation within the LM to activate the TA input evoked slow, small amplitude EPSCs in IS3 cells (Figure 1E,G). Depolarizing the membrane above -50 mV revealed a second slow component to the TA-EPSC, indicating that NMDA receptors (NMDARs) may be present at TA-IS3 cell synapses. Indeed, the current-voltage relationship of the fast EPSC was linear whereas that of the slow component was outwardly rectifying (Figure 1G, n = 13), suggesting that NMDARs contribute to the TA path transmission onto IS3 cells. Blocking the AMPA receptors with NBQX (12.5 M) removed the fast component (peak EPSC amplitude decrease to 13 ± 5% of control at -70 mV, n= 10; Figure 1H), while application of AP5 (100 M), an NMDAR antagonist, blocked the

137

slow component of TA-EPSCs (peak EPSC amplitude decrease to 23.2 ± 5.9% of control at -30 mV). Together, these data point to the synapse-speciﬁc contribution of NMDARs in IS3 cells, in line with previous reports on the input-speciﬁc organization of excitation in other interneuron types (Lei and McBain, 2002; Le Roux et al., 2013; Glickfeld and Scanziani, 2006; Hainmüller et al., 2014; Sambandan et al., 2010).

We next explored the temporal and spatial integration of SC and TA inputs by IS3 cells using patch-clamp recordings, and electrical stimulation or two-photon glutamate uncaging. We found that both inputs show facilitation in response to repetitive stimulation at 10 to 40 Hz (Figure 2A-C). However, the facilitation degree at TA-IS3 synapses was significantly higher than that at SC-IS3 inputs for all frequencies tested (10 Hz: p < 0.05; 20 Hz: p < 0.01; 40 Hz: p < 0.01; n = 18, one- way ANOVA). Indeed, compared with SC-EPSCs, TA-EPSCs increased stronger in response to trains of stimuli (5 pulses; 10 Hz: TA-EPSC5/EPSC1 ratio = 2.3 ± 0.08; SC-EPSC5/EPSC1 ratio= 1.6 ± 0.2; 20 Hz: TA-EPSC5/EPSC1 ratio = 2.6 ± 0.09; SC-EPSC5/EPSC1 ratio = 1.7 ± 0.1; 40 Hz: TA-EPSC5/EPSC1 ratio = 3.4 ± 0.1; SC- EPSC5/EPSC1 ratio = 1.8 ± 0.2; n = 18; Figure 2C). These data indicate that both SC- and TA-IS3 cell synapses have a low release probability, with TA-IS3 input possessing a higher “detonating” capacity (Silver, 2010; Zucker and Regehr, 2002). To examine the synapse-specific recruitment of IS3 cells, we combined current- clamp recordings with repetitive stimulation. At both SC-IS3 (n = 6) and TA-IS3 (n = 5) synapses, repetitive stimulation at 20-40 Hz resulted in strong summation of excitatory postsynaptic potentials (EPSPs) and increased cell firing, although activation of TA-IS3 synapses evoked significantly less of action potentials (APs) at 20 Hz (AP number; SC-IS3-20Hz: 8.7 ± 2, n = 5; TA-IS3-20Hz: 3.3 ± 0.7, n = 6; p < 0.05; SC-IS3-40Hz: 11 ± 3.1, n = 5; TA-IS3-40Hz: 8 ± 1.4, n = 6; p > 0.05; one-way ANOVA, Figure 2D). These results indicate that IS3 cells can be driven via both SC and TA inputs during high frequency activity patterns, and that the CA3 input will be more efficient in recruiting these cells during lower activity levels.

What can be the minimal number of synapses that need to be activated to drive IS3 ﬁring via SC vs TA inputs? To address this question, we used two-photon glutamate

138

uncaging along the IS3 dendrite to mimic the activation of different number of excitatory synapses. As IS3 dendrites exhibit a very low density of spines, to estimate the synapse number activated in our experiments we considered as a reference point the average synapse density reported for dendrites of CR+ cells (RAD: 8.5-10 synapses/10 m; LM: 7.5 synapses/10 m) (Gulyás et al., 1999). In hippocampal slices from VIP-eGFP mice, two-photon uncaging (730 nm, 9 ms) of MNI-glutamate (5 mM) along the IS3 dendrite (2–8 m length) elicited EPSPs (Figure 2E-G) that were completely abolished by the combination of AMPA and NMDA receptor antagonists (NBQX, 12.5 M; AP5, 100 M; n = 3, not shown). We then explored how many synapses were potentially required for triggering APs at SC vs TA inputs by uncaging glutamate on proximal (SC input, < 50 m from the soma, n = 6) vs distal (TA input, > 100 m from the soma, n = 5) dendritic segments of increasing length. The data showed that at segments receiving SC and TA synapses APs can be evoked within the same dendritic length (SC-IS3: 6.5 ± 0.1 m; Figure 2F,H; TA- IS3: 5.1 ± 0.1 m; Figure 2G,H; p > 0.05, unpaired t-test). We performed simple approximations of the number of synapses necessary for AP generation using the anatomical data for CR+ dendrites of Gulyás et al. (1999). We could predict that, although the dendritic lengths that were triggering a single AP at SC- vs TA-IS3 synapses were similar, the number of synapses activated could slightly differ due to the dendrite-specific synapse densities (SC-IS3: 5-7 synapses; TA-IS3: 4-5 synapses) (Gulyás et al., 1999). Collectively, these results indicate that, when activated synchronously, clusters of ∼5 synapses at different dendritic locations should be able to drive the IS3 cell firing.

Synaptic properties of IS3 cells predict that they fire during rising/peak phases of theta oscillations and during SWRs in vivo

We then took advantage of our previous IS3 cell multi-compartment models (Guet- McCreight et al., 2016) developed in the NEURON software environment (Carnevale and Hines, 2006) to predict input-speciﬁc synaptic recruitment of IS3 cells. Using SC- and TA-EPSCs, we ﬁt the weights, rise times, and decay times for excitatory synapses (reversal potential = 0 mV) onto each dendritic compartment of the model.

139

We took the synaptic weights from the best fit on the proximal dendrites and the best fit on the distal dendrites, and extrapolated to obtain a linear distance-dependent weight rule. All synapses on the proximal and distal dendrites were then also assigned the rise and decay times of their corresponding best fits. We did a similar process for inhibitory synapses (reversal potential = -70 mV), where we fit the synaptic parameters to a spontaneous IPSC (amplitude: 20.3 ± 2.8 pA; rise time: 0.52 ± 0.07 ms; decay time: 2.8 ± 0.5; n = 6). Here we extrapolated a linear distance- dependent weight rule according to the weights of the two most optimal fits, and fixed the time constants according to the time constants of the most optimal fit (Guet- McCreight et al., 2017). Altogether, these extrapolations generated realistic EPSCs and IPSCs across the entire dendritic arbor of the IS3 cell model from the soma to the distal dendrites (Figure 3). The weights used here predicted a biologically realistic range of receptors per synapse (Figure 3, Figure Supplement 1A). The model also predicted an increase in the measured reversal potential with distance from soma (Figure 3, Figure Supplement 1B1), which is in line with what was seen experimentally due to the signal decay from the synapse on the dendrite to the soma (Figure 1F-G; Figure 3, Figure Supplement 1B2).

We next used these realistic synaptic parameters (Guet-McCreight et al., 2017) to simulate in vivo-like conditions in our IS3 cell model (Guet-McCreight and Skinner, 2018), where the model was bombarded with synaptic inputs. In our computational explorations, we tested two IS3 cell model variants that did (SDprox1) or did not (SDprox2) have dendritic A-type potassium channels (Guet-McCreight et al., 2016). Using our baseline in vivo-like models, we applied theta- and SWR-timed inputs to examine when IS3 cells are likely to be recruited to spike in vivo.

For theta-timed inputs in an in vivo-like scenario (Figure 4), we assumed phasic excitatory and inhibitory inputs that only spiked once per cycle (Figure 4A3), with a small amount of noise in their exact timing since this was found to enhance model recruitment during theta-timed inputs (Figure 4, Figure Supplement 2). We included diﬀerent proximal and distal dendritic excitatory and inhibitory populations based on the relative timing of diﬀerent hippocampal populations during theta (Mizuseki et al.,

140

2009; Klausberger and Somogyi, 2008) (Figure 4A1-A2). We found that both CA3 and EC inputs could mediate the phasic timing of IS3 cell recruitment (Figure 4B–D), with EC inputs driving the model to fire during the rising phase of theta, and CA3 inputs driving the model to fire near the peak of theta (Figure 4, Figure Supplement 1 and Figure Supplement 2). Considering various proximal and distal inhibitory inputs at several phases of theta (i.e. the peak, the falling phase, and the trough), we found that inhibitory inputs mostly reduced spike recruitment during theta (Figure 4, Figure Supplement 1). This was likely due to there being too many inhibitory inputs as this effect was pronounced when any of the inhibitory input subgroup numbers were doubled (Figure 4, Figure Supplement 1C-D). In other words, significant inhibition could counteract excitatory inputs and thus reduce recruitment during theta. Notably though, this effect was seen less with inhibitory inputs occurring near the falling phase and the trough. Overall, these results suggest that IS3 cells would spike near the rising phase and peak of a theta cycle in vivo (Figure 4), and that their recruitment is driven by excitatory inputs arriving near the rising phase and peak (Figure 4, Figure Supplement 1). It also predicts that inhibitory inputs near the falling phase and trough will modulate IS3 recruitment during theta, but that too much inhibition will counteract IS3 cell recruitment.

Considering SWR-timed inputs in an in vivo-like scenario (i.e. transient inputs that last for 50 ms) (Katona et al., 2014; Varga et al., 2014), we ﬁrst looked at proximal excitatory inputs alone with a burst frequency mimicking CA3 PCs (Frerking et al., 2005) (Figure 5A–B, left). We found that IS3 cells could exhibit an increase in activity above baseline when receiving CA3 inputs alone (Figure 5C-D). We then included a delayed feed-forward inhibitory input onto the proximal dendrites with a frequency that could correspond to bistratiﬁed (BIS) cell spiking (Katona et al., 2014) (Figure 5B, center). In increasing the number of SWR-timed inhibitory inputs we found a substantial reduction in spike recruitment (Figure 5C–D, center). Finally, we also included a delayed feed-forward inhibitory input onto the distal dendrites with a frequency that could correspond to OLM cell spiking (Katona et al., 2014). This did not generally show much more of a decrease in activity back to baseline (Figure 5C– D, right). Note that the SDprox1 and SDprox2 models showed similar results and re- 141

randomizing the synapse locations and presynaptic spike times also showed similar results (Figure 5C-D). From the simulations done for Figure 4 and Figure 5, we did not observe any consequential differences between our two model variants (SDprox1 and SDprox2) with theta and SWR timed recruitment. This could be because the addition of synaptic inputs needed to generate baseline in vivo-like electrophysiological outputs was dependent on the model variant (Guet-McCreight and Skinner, 2018). Thus, this lack of difference in recruitment cannot be directly explained by the inclusion (SDprox1) or exclusion (SDprox2) of dendritic A-type potassium channels. Taken together, these results suggest that IS3 cells could spike during SWRs upon receiving CA3 input. However, the presence of feed-forward proximal inhibitory SWR-correlated inputs (e.g. from BIS cells or possibly other VIP+ or CR+ interneuron types) would have a strong dampening effect on IS3 cell SWR- recruitment. Thus, IS3 cells can be silenced during SWRs depending on the excitatory/inhibitory input balance.

IS3 cells are preferentially activated during locomotion

To test the model predictions, we next examined the recruitment of IS3 cells during different brain states in vivo using two-photon calcium imaging in awake, head- restrained mice trained to run on the treadmill (Villette et al., 2017). To express the genetically-encoded calcium indicator GCaMP6f (Chen et al., 2013) in VIP+ cells for calcium imaging in vivo, a Cre-dependent viral vector AAV1.Syn.Flex.GCaMP6f.WPRE.SV40 was injected into the CA1 area of VIP-Cre mice (n = 3 mice). Two-photon somatic calcium imaging was performed within the hippocampal CA1 PYR and RAD, where most IS3 cells are located (Figure 1A; Figure 1, Figure Supplement 1; Figure 6C). This was combined with local field potential (LFP) recordings from the contralateral CA1 area, as well as recordings of the animal’s speed (Figure 6A-B). In this simple behavioral paradigm, the animals can voluntarily alternate between immobility (speed < 1 cm/s), and locomotion (speed > 1 cm/s; Figure 6B), allowing us to record and analyze hippocampal theta oscillations during locomotion and SWRs during quiet states (Figure 6D-E). To separate IS3 cells from VIP+ basket cells (BCs), we first developed a set of

142

morphological criteria based on the soma properties derived from the anatomically reconstructed neurons recorded in vitro and identified as IS3 cells or BCs (Figure 6, Figure Supplement 1A1-B1; IS3: n = 36; BC: n = 11). The 3D rendering of somatic surface was used to develop the soma areas and extract the soma diameters in medio-lateral X- and rostro-caudal Y-dimensions (Figure 6, Figure Supplement 1A2- A3,B2-B3). Similarly, the X- and Y- parameters corresponding to different soma diameters were extracted from VIP+ cells imaged in vivo (Figure 6, Figure Supplement 1D1-D2,E1-E2). Based on the data distribution for soma diameter and area in anatomically confirmed IS3 cells and VIP+ BCs recorded in vitro (Figure 6, Figure Supplement 1C), the cut-off parameters (X ≤ 16μm, Y ≤ 11μm, soma area size ≤ 152 μm2) were established to separate putative IS3 cells (pIS3) from the putative BCs (pBC) imaged in vivo (Figure 6, Figure Supplement 1F; pIS3 n = 9; pBC, n = 12). To validate the appropriateness of the selected cut-off parameters for identification of the in vivo recorded VIP+ cells, following in vivo imaging, the mouse brains were processed for immunolabelling for GFP and CR, the IS3 cell marker (Figure 6, Figure Supplement 1G). The results showed that CR- (pBC) and CR+ (pIS3) cells in PYR/RAD separate well into two clusters using our cut-off criteria (Figure 6, Figure Supplement 1H; CR+: n = 17; CR-: n = 13).

Using the established morphological criteria for pIS3 cell separation, we then explored pIS3 cell activity during locomotion and immobility (n = 18 cells, 221 running periods and 556 stationary states from 2 imaging sessions of 5 min each). Post hoc immunohistochemical analysis of recorded neurons (2 out of 18 VIP+ PYR/RAD interneurons recorded in vivo were found after brain re-sectioning and processed for CR) confirmed that VIP+ cells identified as pIS3 were positive for CR (2 cells out of 2 tested; Figure 6F). To analyze calcium transients during run epochs (Figure 7A), we identified periods in the speed traces of > 1 cm/s that lasted for at least 3 consecutive seconds. We then further selected run epochs that were preceded by 5 seconds of average speed below 1 cm/s (i.e. run-starts; 64 epochs selected across 17/18 cells) or followed by 5 seconds of average speed below 1 cm/s (i.e. run-stops; 69 epochs selected across 18/18 cells; Figure 7B). In correspondence with run

143

epochs, there was an increase in the time-varying max theta power (Figure 7C). In looking at the mean calcium responses across the pIS3 cells, we mainly observed calcium transients during run-stop periods (Figure 7D), pointing to the delayed recruitment of pIS3 cells during theta-run epochs. The timing of this change was so variable across diﬀerent pIS3s, however, that it was imperceptible when looking only at the median of mean calcium traces across run epochs (Figure 7D, black line).

We also calculated cross-correlations between calcium signals and running speed or the time- varying theta power (Figure 7E-F). Specifically we looked at whether each set of time series were significantly cross-correlated in time by looking at their cross-correlation magnitudes at lags of zero seconds (i.e. zeroth peaks) as well as their maximal peak magnitudes and corresponding lag times (Figure 7, Figure Supplement 1A-D, see methods). Significant zeroth peaks indicated that events (e.g. run epochs and calcium transients) from the two time series co-occurred in time. The lag time of the maximal peaks (i.e. if the maximal peak magnitude was significant) indicated the approximate delay with which the events from one signal lagged behind the events from another signal. Specifically, if maximal peak lag times were negative, this indicated that calcium transients followed run-theta epochs (and the opposite if positive). For both speed and theta power, we found that almost all cells had significant zeroth peaks and maximal peaks (Figure 7, Figure Supplement 1A-D). Most cells (N = 15/18 for speed; N = 15/18 for theta power) also had slightly negative (i.e. -5s to 0s) maximal peak lag times. This indicated that calcium transients occurred with a slight delay relative to run-theta epochs. In fact, this can be seen in the example shown in Figure 6D, as well as the general observation that calcium transients tended to occur around run-stop periods (Figure 7D).

As well, we computed Pearson’s correlations between speed and calcium traces or estimated spike rate traces from three different spike estimation algorithms (see next section). We obtained significant relationships between speed and the calcium signal across all pIS3 cells, though different cells varied in whether this relationship was positive, negative, or flat (Figure 7G). The variability in the sign of these relationships is at least partly due to variability in cross-correlation lag times between

144

signals (i.e. see Figure 7, Figure Supplement 1A-D). For example, larger negative lag times between speed traces and calcium signals meant that calcium transients were occurring closer to periods of immobility following run epochs, which meant more negative relationships. Also, relationships between speed and estimated spike rates were not always signiﬁcant, but there was consistency of whether the relationship was positive or negative for the diﬀerent spike estimation algorithms. Similarly, Pearson’s correlations between time-varying theta power and calcium traces or estimated spike rate traces equally showed consistently similar results when compared to the Pearson’s correlations for speed (Figure 7, Figure Supplement 1E). Together, we conclude that the activity of pIS3 cells correlates with running speed and theta oscillation power, however they exhibit a delayed recruitment.

IS3 cells prefer to ﬁre near the rising/peak phases of theta oscillations

To further examine pIS3 activites during diﬀerent network states and test model predictions, we used several methods to estimate spike times from GCaMP6f calcium signals. This included the Matlab toolboxes MLspike (Deneux et al., 2016) and UFARSA (Rahmati et al., 2018), as well as custom Matlab code for estimating spike times using the gradient of the calcium signal (DF/Dt; similar to what was used in Jackson et al. (2016)).

Across the spike times estimated from all methods (Figure 8A), we observed a larger mean normalized percent of spikes during locomotion compared to immobility, as shown in DF/Dt and MLspike methods (Figure 8B), indicating an overall higher activity of pIS3s during locomotion. However, some cells exhibited a larger normalized percent of spikes during immobility (3/18 cells), indicating that this state- dependent preference was variable across the cells (Figure 8B).

We then analyzed the theta phasic distributions of spikes where the speed was greater than 1 cm/s and the time-varying theta power was greater than its mean (i.e. periods of running and high-theta activity). In pooling all of the spike phases across cells into a single distribution for each method (Figure 8C), we observed signiﬁcant

145

non-uniform distributions (p < 0.001, Rayleigh’s Tests), and signiﬁcant polarities towards the circular means across all spike phases (Table 1; p < 0.001, V-Tests). However, these distributions also exhibited large amounts of dispersion. To characterize this dispersion, we looked at vector length and angular deviation. Vector lengths closer to 1 indicate more clustering of angles around the mean (i.e. less dispersion) and larger angular deviation values (with a maximum possible value of 81.03) indicate larger amounts of dispersion. We found vector lengths much smaller than 1 (Table 1) and angular deviations close to the maximal possible value of 81.03° (Table 1), which indicates broad distributions of spikes across theta phases.

To further investigate this result, we performed second-order analyses in which the mean of the mean angles across all the cells was examined using the three spike estimation methods. We found signiﬁcant mean spike phase preferences with most vector lengths towards the rising/peak phase of the theta cycle (p < 0.05, Rayleigh’s Test; p < 0.01, V-Test; Figure 8D), indicating that IS3 cells may be preferentially recruited during this phase of the theta oscillation. In comparing the three methods, we note that in the mean of means analyses, the UFARSA method generated the largest amount of phasic dispersion, that is, a large variability. Speciﬁcally, several of the cells exhibited large vector lengths towards the falling phase and trough of theta (Figure 8D, middle). This can be explained by noticing that the individual UFARSA spike phase histograms (not shown) for some particular cells (i.e. cells #6, #11, #13, and #51) possessed low spike number estimates during running/high theta (i.e. 4, 4, 16, and 9 spikes, respectively) relative to the other two methods. This is possibly due to the UFARSA method underestimating the number of spikes compared to the other two methods. Low estimates using the UFARSA method was not limited to these particular cells, as two other cells (cells #12 and #52) were found to not have any spikes during periods of running and high theta. It is possible that the low spike number estimates with the UFARSA method increased the probability of outliers. Another outlier of note was the large vector length of 1 seen for one of the cells in the DF/Dt mean-of-means polar plot (Figure 8D, left). Upon closer inspection, this cell (i.e. cell #12) was found to have only one DF/Dt spike estimated

146

during running/high theta, thus explaining why it generated a vector length of 1 (i.e. meaning no phasic dispersion).

IS3 cells are silent during ripples

We did not observe any apparent calcium transients that coincided with ripple activity (Figure 9A), indicating that pIS3 cells are likely silent during ripples. Indeed, using DF/Dt and MLspike algorithms, we extracted a few spikes co-localized in time with ripples (Figure 9B), which represented a small proportion of the total number of spikes. As well, in looking at the peri-stimulus time histograms of estimated spiking relative to sharp-wave ripple occurrence, we did not observe any increase in spike probability density during ripples (Figure 9B, top subplots). This suggests that IS3 cells are silent during SWRs despite the model predictions. In fact, as the model suggests, they can be silenced through activation of diﬀerent inhibitory inputs (Figure 5C–D, center), though these predictions remain to be examined in detail.

Discussion

Here, we demonstrate that IS3 cells receive TA and SC excitatory inputs with input- specific glutamate receptor composition and spatio-temporal integration properties. In particular, TA-IS3 synapses express NMDA receptors postsynaptically and show a larger degree of short-term facilitation during repetitive synaptic activity compared to that of SC-IS3 synapses. Despite a different dendritic location, both TA and SC- IS3 synapses were able to drive the IS3 cell spiking when a cluster of at least five closely spaced inputs was activated synchronously. Furthermore, model simulations predicted that the synaptic properties of IS3 cells allow them to fire during rising/peak phases of theta oscillations and during SWRs in simulated in vivo-like conditions. The recruitment of these cells during theta oscillations was controlled through rhythmic excitatory inputs from the CA3, EC3, and local inhibitory populations. Moreover, the CA3 and EC3 inputs provided the largest degree of modulation in terms of the firing timing of IS3 cells. During SWRs, the model predicted that CA3 input alone would be sufficient to recruit these cells. However, experimental data using two-photon calcium imaging of pIS3 cells in awake mice revealed a large

147

degree of variability in their activity. Moreover, they exhibited delayed recruitment during theta-run epochs and were silent during SWRs. Consistent with the model predictions, however, the activity of pIS3 cells correlated well with theta power and animal speed, and they could preferentially ﬁre during the rising/peak phases of theta-run episodes. Taken together, these data indicate that additional local and long-range synaptic mechanisms may modulate the recruitment of IS3 cells during diﬀerent behavior and network states.

The TA vs SC pathway-specific properties of excitatory synaptic transmission have been well characterized in PCs (Otmakhova et al., 2002; Arrigoni and Greene, 2004). Compared with the SC-PC synapses, TA-PC input exhibits slow kinetics and a larger NMDA/AMPA receptor ratio (Otmakhova et al., 2002). In hippocampal interneurons, the distribution and subunit composition of ionotropic glutamate receptors is input- and cell type-specific. This has a profound impact on the input integration and induction of synaptic plasticity (NyÍri et al., 2003; Lamsa et al., 2007; Croce et al., 2010; Szabo et al., 2012). Specifically, the excitatory synaptic transmission at TA and SC pathways was compared in neurogliaform cells (Price et al., 2005), and NMDA and AMPA receptors were found to be present at both pathways. Our data indicate that TA-IS3 EPSCs in IS3 cells have slower kinetics than those evoked at the SC synapses, which in addition to dendritic filtering, can be explained in part by contribution of the NMDA receptors. The I-V curves for the fast component of both TA- and SC-EPSCs showed linear relationships, indicating the involvement of the Glu2A-containing AMPA receptors. These data are consistent with the transcriptomic analysis of excitatory inputs’ composition in VIP+ neocortical interneurons (Paul et al., 2017). In particular, the Gria2 (GluR2 subunit) mRNAs were highly enriched in VIP+/CR+ neocortical interneurons. In addition, the Grin2b (GluN2B subunit) transcripts were preferentially expressed in VIP+/CR+ cells. The larger NMDA/AMPA receptor ratio at distal dendrites may have impact on dendritic integration and induction of synaptic plasticity in IS3 cells, such that the slow kinetics of the TA-EPSC may provide a longer time window for excitatory input integration, along with the induction of Hebbian LTP (Regehr and Tank, 1990; Golding et al., 2002; Dudman et al., 2007; Kullmann and Lamsa, 2007). In addition, NMDAR spikes

148

generated at distal dendrites have been shown to facilitate AP ﬁring and the induction of associative LTP in PCs (Brandalise et al., 2016). Whether these mechanisms may operate in interneurons, and in IS3 cells in particular, remains to be examined.

Similarly, the short-term plasticity at excitatory synapses in hippocampal interneurons displays aﬀerent and cell type-speciﬁc features (Thomson, 1997; Ali et al., 1998; Sun et al., 2005; Croce et al., 2010). Such complexity in input processing might arise from the highly variable pre- and post-synaptic distribution of ionotropic and metabotropic glutamate receptors (Shigemoto et al., 1996; Sun and Dobrunz, 2006; Chittajallu et al., 2017). The fact that IS3 cells exhibited exclusively short-term facilitation at both inputs may imply similar presynaptic mechanisms operating at the SC- and TA-IS3 synapses with low initial release probability (Katz and Miledi, 1968). In addition, a larger facilitation at TA-IS3 synapses may involve the NMDA receptors, especially if it contains the kinetically slow GluN2B-subunit (Chamberlain et al., 2008).

Furthermore, our data showed that it is likely that a similar number of synchronously activated synapses in proximal or distal dendrites were required for eliciting the AP firing of IS3 cells. It has been documented that the density of excitatory synapses on thick proximal dendrites of CR+ cell is 85.36/100 μm (0.85/μm) while on distal dendrites it is 75.31/100 μm (0.75/μm) (Gulyás et al., 1999). Accordingly, based on the dendritic length required to evoke a single AP in different dendritic domains (proximal: 6.5 μm, distal: 5.1 μm), the minimal number of synapses leading to AP generation may correspond to ∼5 on proximal and ∼4 synapses on distal dendrites. This observation indicates that additional mechanisms, such as increased NMDA/AMPA receptor ratio at distal dendrites (Otmakhova et al., 2002), synaptic scaling of AMPA receptors (Magee and Cook, 2000) or dendrite site-specific distribution of potassium channels in IS3 cells (Guet-McCreight et al., 2016) may compensate for dendritic filtering (Stuart and Spruston, 1998; Emri et al., 2001).

149

How can these synaptic mechanisms control IS3 cell recruitment in vivo? Using computational simulations, we demonstrate the impact of different excitatory and inhibitory inputs on IS3 cell firing. First, we note that the relative timing and level of activity along the CA3 and EC pathways can modulate the timing of IS3 cell recruitment during theta oscillations. If both were to strongly influence IS3 cell spiking, this could enhance the variability of cell recruitment during theta rhythms. Second, removing inhibition occurring at the falling phase and trough of theta oscillation could increase the spike initiation, pointing to the importance for these phase-specific inputs in IS3 cell recruitment during theta. However, doubling inhibition would decrease IS3 cell firing, regardless of the phase, which indicates that too much inhibition can hinder recruitment of these cells. This suggests that during theta, IS3 cells are likely modulated through specific local and long-range GABAergic projections. Specific timing and levels of inhibition provided by these inputs may be responsible for delayed recruitment of IS3 cells during theta-run epochs. The candidate local inhibitory inputs that could fit the spatio-temporal patterns of theta activation shown here include OLM, bistratified, basket, and neurogliaform cells (Mizuseki et al., 2009; Klausberger and Somogyi, 2008), although it is still unknown whether they are connected to IS3 cells. It is also possible that IS3 cells receive inputs from IS1 and IS2 cells (though the timing of these cells during theta remains unknown), as well as the medial septum (MS) GABAergic projections. The morphological studies showed that long-projecting MS GABAergic neurons make multiple synaptic contacts with hippocampal VIP+ cells (Papp et al., 1999), raising the possibility that they may also play a role in modulating IS3 cells during network oscillations (Dragoi et al., 1999; Hangya et al., 2009; Viney et al., 2013; Unal et al., 2015). Furthermore, in line with model predictions, the in vivo calcium imaging of pIS3 cells with estimated spike times demonstrated that these cells appear to preferentially fire during the rising/peak phases of theta, though with a large degree of phasic dispersion. The large variability in the experimental data relative to the model could be due to a number of factors including errors in spike time estimation by the different methods, impact of additional excitatory inputs that were not included in the model (e.g. from local CA1 PYR and subcortical neuronal populations), or

150

cellular diversity within the pIS3 cell population (Francavilla et al., 2018), which remain to be explored in details.

The model also predicted that IS3 cells would spike during SWRs since they receive excitatory input from CA3 PCs. However, the presence of feed-forward inhibition could dampen the IS3 cell recruitment considerably. Indeed, in the in vivo calcium imaging experiments, we did not observe any correlation between the IS3 cell activity and SWR appearance. In addition to the feed-forward inhibition, this can be explained by the impact of subcortical neuromodulatory inputs, for example serotonergic projections from the median raphe nucleus (Papp et al., 1999) that can modulate hippocampal theta (Prisco et al., 2002; Domonkos et al., 2016) and ripple (Wang et al., 2015) oscillations via specific cellular mechanisms due to the cell type- specific expression of different types of 5-HT receptors (Freund et al., 1990; Helboe et al., 2015; Wyskiel and Andrade, 2016). Whether the IS3 cells are targeted by these long-range projections and can be modulated during theta oscillations or ripples remains to be explored.

In conclusion, our data indicate that IS3 cell operation and function may be more complex than can be predicted from their synaptic properties. And, in agreement with previous observations on VIP+ interneurons in cortical circuits, these cells may be tuned by speciﬁc activity patterns and learning paradigms in relation to mood state and animal experience. For example, in the auditory cortex, VIP+ cells are strongly recruited by positive or negative reinforcement stimuli during auditory discrimination tasks (Pi et al., 2013). In the prefrontal cortex, the activation of VIP+ cells is coupled with the outcome of goal-directed behavior (Pinto and Dan, 2015). These data point to the role of modulatory systems in controlling the cortical disinhibition mediated by VIP+ cells. Thus, we cannot exclude that hippocampal IS3 cells can be preferentially involved in learning and memory-related behaviors and exhibit variable recruitment under basal conditions, such as our behavior paradigm.

151

Methods and Materials

Animals

Four mouse lines of both sexes were used in this study: VIP/enhanced green ﬂuorescent protein (VIP-eGFP), VIP-IRES-Cre (Jackson #010908), ROSA9 and VIP-IRES-Cre-ROSA9. VIP-eGFP line [MMRRC strain #31009, STOCK Tg(Vip- EGFP) 37Gsat] was purchased from the MMRRC facility at the University of California, Davis, CA, in which eGFP was targeted selectively to VIP+ interneurons (Tyan et al., 2014). VIP-tdTomato line was generated by breeding VIP-Cre mice with the reporter line Ai9-(RCL- tdTomato) (B6.Cg-Gt(ROSA)26Sortm9(CAG- tdTomato)Hze/J, stock #007909, The Jackson Laboratory). All experiments were performed in accordance with animal welfare guidelines of the Animal Protection Committee of Université Laval and the Canadian Council on Animal Care. Animals were housed in groups of 2–4 per cage with standard light/dark cycle and ad libitum access to food and water.

Whole cell patch-clamp recordings in hippocampal slices in vitro

Slices were prepared from P14-30 mice. Mice were deeply anaesthetized with iso2urane or ketamine- xylazine (10 mg/ml-1mg/ml). Mice older than one month were intracardially perfused with ice cold arti1cial cerebrospinal fluid (ACSF) sucrose solution containing the following (in mM): 250 sucrose, 2 KCl, 1.25 NaH2PO4, 26

NaHCO3, 7 MgSO4, 0.5 CaCl2, and 10 glucose (osmolarity: 290-310 mOsm/L), oxygenated constantly with 95% O2 and 5% CO2, to improve the quality of acute brain slices. Transversal hippocampal slices were obtained with a vibratome (VT1000S; Leica Microsystems or Microm; Fisher Scientific) in ice cold ASCF sucrose solution and transferred to pre-heated recovery solution at 37.5°C containing (in mM) 124 NaCl, 2.5 KCl, 1.25 NaH2PO4, 26 NaHCO3, 3 MgSO4, 1

CaCl2, and 10 glucose, oxygenated constantly with 95% O2 and 5% CO2, with the osmolarity 290-310 mOsm/L.

152

After 1 hour of recovery, slices were transferred to the recording chamber and continuously perfused with ACSF solution containing (in mM) 124 NaCl, 2.5 KCl,

1.25 NaH2PO4, 26 NaHCO3, 2 MgSO4, 2 CaCl2, and 10 glucose; osmolarity 290-310 mOsm/L, which were oxygenated constantly with 95% O2 and 5% CO2. eGFP cells and tdTomato cells were identified with blue light and green light illumination, respectively using upright electrophysiological microscope (Nikon Eclipse FN1), and visualized with differential interference contrast during patch-clamp recordings. Two- photon images of eGFP and tdTomato cells in acute slices were acquired with a 25x water-immersion objective (NA 0.95, 2.5 mm working distance) using a two-photon microscope (TCS SP5; Leica Microsystems) with Ti-sapphire laser (Chameleon Ultra II; Coherent; > 3W, 140 fs pulses, 80 Hz repetition rate) and laser wavelength tuned to 800 nm. For whole-cell patch clamp recording, a concentric bipolar electrode (FHC) or theta glass pipette (Sutter instrument) was placed at stratum lacunosum moleculare or stratum radiatum to deliver electrical stimuli. Borosilicate glass capillaries (World precision Instruments Inc.) were used to make pipettes for whole-cell recording (tip resistance 3–6 MΩ). For voltage-clamp recordings, the pipette was filled with a Cs-based solution (in mM): 130 CsMeSO4, 2CsCl, 10 diNa- phosphocreatine, 10 HEPES, 4 ATP-Tris, 0.4 GTP-Tris, 0.3% biocytin, 2 QX-314, 0.1 spermine, pH 7.2–7.3, 280–290 mOsm/L. For current-clamp recordings, a K+- based intracellular solution was used containing (in mM):130 KMeSO4, 2 MgCl2, 10 diNa-phosphocreatine, 10 HEPES, 4 ATP-Tris, 0.4 GTP-Tris and 0.3% biocytin (Sigma), pH 7.2–7.3, 280–290 mOsm/L. Passive membrane properties were recorded immediately after cell membrane rupture, active membrane properties were recorded in current-clamp mode with ramping current steps. Pharmacological experiments were performed in voltage-clamp mode at the holding potential of - 70mV. The following pharmacological agents were used: gabazine (10 µM, Ascent Scientific), CGP-55845 (2 µM, Abcam Biochemicals), NBQX (12.5 µM, Abcam Biochemicals), AP5 (50 µM, Abcam Biochemicals). Gabazine and CGP were applied into the bath 10 min before electrical or optical stimulations to block GABAergic transmissions. To record the miniature EPSCs and IPSCs, TTX (1 µM, Alomon labs) was applied to the bath 10 min before recording. The EPSCs were evoked by

153

stimulation (1- 5 pulses, 5-40 Hz, pulse width 0.2 ms) every 30 seconds. The series resistance (Rser) was monitored throughout the experiments by applying a -5 mV voltage step at the end of each sweep. Recordings with changes in Rser> 15% were discarded from the analysis. Data was filtered at 2–3 kHz and digitized at 10 kHz with Multiclamp 700B ampli1er and the Clampex 10.5 software (Molecular Devices).

Two-photon glutamate uncaging

The protocol for two-photon glutamate uncaging was previously described in details (Chamberland et al., 2010). The Alexa Fluor 594 (20 μM) was included in the intracellular solution to visualize the cell morphology. The caged compound MNI-Glu (5 mM; Tocris) was introduced in the recording chamber and focal photolysis was performed by illuminating the proximal (<50 μm from soma) or distal (>100 μm from soma) dendritic segments of diﬀerent length (2-8 μm) for 9 ms (the smallest photostimulation duration that can be achieved in xyt-mode with our two-photon system) with a 730-nm laser pulse (laser power, 25–30 mW, Leica TCS SP5 microscope with a 40×, 0.8 NA water-immersion objective; Leica Microsystems). The size of the uncaging region was increased gradually until an action potential was evoked. Two-photon images for dendritic visualization were acquired using two- photon excitation at 800 nm (laser power 5-10 mW). To avoid photodamage, the uncaging-based stimulation was performed every 30 seconds, and the laser power did not exceed 40 mW. Uncaging-evoked EPSPs or action potentials were recorded in current clamp mode.

Computational model

Model Details and Code

The models used in this study were previously custom-built to specifically simulate the somatic and dendritic electrophysiological characteristics of IS3 cells (Guet- McCreight et al.,2016) (ModelDB accession #: 223031). Models were simulated in the NEURON software environment (Carnevale and Hines,2006) and results were plotted using customized Python and Matlab scripts. Note that all code and scripts

154

can be accessed online (synaptic optimization code: https://github.com/ FKSkinnerLab/IS3-Cell-Model/tree/master/LayerSpecificInputTests; rhythmic input simulations code: https://github.com/FKSkinnerLab/IS3-Cell- Model/tree/master/RhythmTests).

Model Synapse Details

For the synapse model, we used NEURON’s built-in Exp2Syn function, which models synaptic current as a two-state kinetic scheme.

Where i is the synaptic current, G is the synaptic conductance, E is the reversal potential, V is the membrane potential, weight is the synaptic weight, factor is a NEURON process that is used to normalize the peak synaptic conductance to the value of weight, t is time, τr is the rise time, and τd is the decay time.

We used NEURON’s Praxis Optimizer, to fit the synaptic parameters for each compartment to an experimental EPSC or IPSC. Excitatory synapses less than 300 µm from soma were fit to a Schaffer Collateral-evoked EPSC, and excitatory synapses greater than 300 µm from soma were fit to a temporoammonic pathway-

155

evoked EPSC. All inhibitory synapses were fit to a spontaneous IPSC. We set the weight, τr, and τd, as the free parameters in the optimization and assumed that all voltage-gated channels were blocked since sodium and potassium channel blockers were used during the experimental recordings. To remove any driving forces caused by leak conductances, we also set the leak reversal potential to the voltage-clamped holding potential of the model (0 mV when fitting IPSC recordings, and -70 mV when fitting to EPSC recordings). During the optimizations, we used experimental τr and

τd individual trace measurements as seed values (IPSC:= 0.25 ms and τd = 4.14 ms;

EPSCProximal: τr = 0.45 ms and = 1.41 ms; EPSCDistal: τr = 1.71 ms; τd = 5.04 ms) and used parameter bounds based on literature values (AMPA: Swanson et al. (1997); Masugi-Tokita et al. (2007); Antal et al. (2008); Nusser (1999); Spruston et al. (1995); Andrásfalvy and Magee (2001); Andersen et al. (2006); GABAA: Andersen et al. (2006); Ernst et al. (2003); Nusser et al. (1997); Nusser (1999)). These chosen parameter ranges and bounds were the following:

Using the weight value obtained from the most optimal fit for proximal dendritic synapses as well as the weight value obtained from the most optimal fit for distal dendritic synapses, we applied a linear distance-dependent weight rule for AMPA synapses:

−6 −6 Weight AMPA = (2.308 × 10 µS/µm) × Distance + (220.2 × 10 µS) (3)

Similarly, using the weight values obtained from the two most optimal fits for inhibitory synapses, we applied a linear distance-dependent weight rule for GABAA synapses:

−6 −6 weightGABAA = (4.691 × 10 µS/µm) × Distance + (269.6 × 10 µS) (4)

Respectively, the rise times and decay times of all the excitatory synapses were fixed to the rise and decay times of their corresponding layer-specific most optimal fit. The rise times and decay times of all the inhibitory synapses were fixed to the

156

rise and decay times of the most optimal inhibitory fit. Excitatory SR: τr = 2.9936 ×

10−4 ms, τd = 2.4216 ms; Excitatory SLM: τr = 6.1871 × 10−4 ms, τd = 3.1975 ms;

Inhibitory: τr = 0.1013 ms, τd = 4.8216 ms.

To get the ranges of numbers of receptors per synapse for each synaptic location (Figure 4 S1A), we computed the following:

(Receptorsmin,Receptorsmax)=Weight/(UnitaryConductancemax,UnitaryConductance min) (5)

To simulate the experimental protocol for estimating the distance-dependent reversal potential of each excitatory synapse (Figure 4 S1B), we held the soma in voltage-clamp mode and incrementally increased the holding potential. The reversal potential of each synapse was recorded as the holding potential when the polarity of the postsynaptic current at the soma reversed.

Baseline in vivo-Like Inputs

In vivo-like excitatory and inhibitory inputs (i.e. spike rates and numbers of active synapses) were estimated previously (Guet-McCreight and Skinner, in submission). In all simulations, the baseline in vivo-like synaptic locations and presynaptic spike trains are chosen randomly according to a set of random seeds. For SDprox1, baseline in vivo-like inputs have the following parameters: 144 excitatory synapses with 5 Hz rates each, and 8 inhibitory synapses with 50 Hz rates each. For SDprox2, baseline in vivo-like inputs have the following parameters: 144 excitatory synapses with 5 Hz rates each, and 24 inhibitory synapses with 10 Hz rates each.

Theta-Timed Inputs

Note that in all in vivo-like simulations, all synaptic locations (i.e. for baseline, theta- timed, and SWR- timed synapses) are chosen randomly such that they are placed anywhere within their designated proximal or distal (or both) dendritic sections. For estimated numbers of X1 theta-timed excitatory inputs, we increased the number of inputs on either proximal or distal dendrites, until a large enough power spectral 157

density (PSD) was obtained (i.e. 50 spikes2/Hz). For estimated numbers of X1 theta- timed inhibitory inputs, we first increased a current input at the soma until a base spike rate of 35 Hz was attained. We then increased the number of inhibitory inputs on either proximal or distal dendrites until a large enough PSD was obtained (i.e. 80 spikes2/Hz). Between the proximal and distal dendrites for the SDprox1 and SDprox2 models, we estimated approximately 27 synapses per excitatory presynaptic population and approximately 8 synapses per inhibitory presynaptic population. Note that we used these same estimates for SWR-timed presynaptic populations.

Unless otherwise noted, presynaptic spiking for both theta-timed inputs and SWR- timed inputs were idealized, such that they spiked perfectly on time at their prescribed frequencies. For theta, this was 8 Hz for all presynaptic populations. For SWR inputs this varied across diﬀerent populations (excitatory: 100 Hz, feedforward proximal inhibitory: 120 Hz, feedforward distal inhibitory: 20 Hz; Figure 5B), and the duration was limited to 50 ms.

To simulate theta-timed inputs, we used NEURON’s NetStim function to simulate spike trains with 8 Hz frequencies. We adjusted the start time of the presynaptic populations depending on their phase preferences. Finally, we also analyzed of the effects of fractional randomness on theta-recruitment (Figure 4, Figure Supplement 2). The fractional randomness (i.e. noise values between 0 and 1), controls the proportion by which each spike interval is influenced by the set interval value versus a random interval value sampled from a negative exponential distribution with a mean duration of noise × interval.

t0 = (1 − noise) × interval (6)

Where t is time, interval is the set inter-spike interval, and noise is the fractional randomness. This probability density function ensures that t0 is the minimum

158

possible interval, that the maximum possible interval is infinite, and that the mean expected interval is (1−noise)×interval +noise×interval. Here we found that theta- recruitment was optimal with a 0.01 level of fractional randomness (noise; Figure 4, Figure Supplement 2).

Short-term Plasticity

We decided not to explicitly model short term facilitation (i.e., a form of short-term plasticity, STP) as observed experimentally since we felt that we were already extrapolating enough from the existing data. Further, explicit data on STP for inhibitory synapses is not yet available to consider, and our in vivo-like representations were necessarily approximate (Guet-McCreight and Skinner, 2018). Also, since both pathways were found to be facilitating, it is unlikely that incorporating STP facilitation in our synaptic models would cause any major changes in the model output and predictions at this stage. Speciﬁcally, as we used just enough synapses for recruitment considerations and explorations (see ’theta-timed inputs’ above), adding STP would likely just reduce the number of synapses needed for this. Also, the pre-synaptic frequencies may not be high enough to fully engage STP dynamics. For the models presented here, it is the relative contributions of diﬀerent pathways and presynaptic inhibitory populations in considering IS3 cell recruitment given their morphological and biophysical characteristics together with a reasonable capturing of their relative synaptic properties (see Figure 3).

Stereotaxic injections

For in vivo two-photon calcium imaging, AAV1.Syn.Flex.GCaMP6f.WPRE.SV40 (Penn Vector Core) was injected in the CA1 area of VIP-Cre mice to express GCaMP6f selectively in VIP+ interneurons.

For stereotaxic injection, mice were deeply anaesthetized with intraperitoneal injection of ketamine–xylasine mixture (100 /10 mg kg−1) and fixed on a stereotaxic frame (Kopf Instruments). Viral vector was delivered to the areas of interest using a microprocessor-controlled nanoliter injector according to the following coordinates:

159

CA1 (two sites were injected): site 1: AP + 2.5 mm, ML ± 2.1 mm, DV - 1.3 mm; site 2: AP, -2.0, ML, -1.6, DV, -1.3 mm. Viral vector was injected with the speed of 1 mm/min. The total volume was 100 nl. The pipette was kept for 5 min after injection and then withdrawn slowly and the scalp was sutured. Mice were treated with postoperative pain killer (buprenorphine, 0.1 mg/kg−1; 48 h) for 3 consecutive days after injection; or once (buprenorphine slow release, 0.1mg/kg−1) after injection. Animals were recovered for 7-10 days before implantation of hippocampal imaging window.

In vivo two-photon Ca2+-imaging in awake mice

Two-photon somatic Ca2+-imaging of neuronal activity was performed simultaneously with contralateral local field potential (LFP) recording in head-fixed awake mice running on a treadmill. One week after viral injection, a glass-bottomed cannula was inserted on top of the dorsal hippocampus and secured with Kwik-Sil at the tissue interface and Superbond at the skull level (Dombeck et al., 2010). A single tungsten electrode (33 Ω–CM/F, California Fine Wire) for LFP recordings was implanted in the stratum pyramidale of the contralateral CA1 region and a reference electrode was implanted above the cerebellum (Villette et al., 2017). The head plate was oriented medio-laterally at 7–13° using a four-axis micromanipulator (MX10L, Siskiyou) and fixed with several layers of Superbond and dental cement. Mice were allowed to recover for several days with postoperative pain killer treatment for 3 consecutive days (buprenorphine, 0.1 mg/ kg−1; 48 h).

The head ﬁxation platform was described previously by Villette et al. (2017). Brieﬂy, a head plate was clamped to a XY moveable metal frame, which minimized brain motion artifacts during recording. The treadmill was equipped with lateral walls to increase animal contentment and coupled with an optical encoder allowing for synchronous acquisition of running speed and electrophysiological signal. After recovery from surgery, animals were handled for 5-15 min twice per day for two days prior to the beginning of the experiment. During experiments, animals alternated between two behavior states: immobility and walking-running periods. The LFP

160

signal and animal running speed were acquired at a sampling frequency of 10 kHz using the DigiData1440 (Molecular Devices), AM Systems ampliﬁer and the AxoScope software (v10.5, Molecular Devices). Two-photon imaging was performed using a Leica SP5 TCS two-photon system coupled with a Ti:sapphire femtosecond laser (Chameleon Ultra II, Coherent), tuned to 900 nm. A long-range water- immersion 25× objective (0.95 NA, 2.5 mm working distance) was used for ﬂuorophore excitation and light collection to external photomultiplier tubes at 12 bits. Image series were acquired at axial resolutions of 2 m/pixel and temporal resolutions of 48 images/sec.

For each cell recorded, two 5-min imaging sessions were acquired. For each animal, the total length of imaging did not exceed 1 h. After imaging session, animals were returned to home cage. Between animals, the treadmill was cleaned with tap water. IS3 cells were identiﬁed post hoc using criteria described in the main text (Figure 7, Figure Supplement 1A-B, top).

Immunohistochemistry and morphological reconstructions

For immunohistochemical analysis of molecular markers expressed in hippocampal CA1, animals were intracardially perfused with 4% paraformaldehyde (PFA) followed by sucrose ACSF, then the brain was extracted and fixed in 4% PFA overnight at 4 °C. On the next day, fixed brains were embedded in 4% agar. Hippocampal sections (thickness, 40–70 µm) were acquired using a vibratome (VT1000; Leica Microsystems or PELCO EasySlicer) and stored in PB sodium azide (0.5mg/ml) solutions. Sections were permeabilized with 0.1–0.3% Triton X-100 in TBS/PBS and incubated in blocking solution containing 20% normal serum for 1 h. Then sections were incubated with primary antibodies at 4°C for 24–48 h. After that, slices were incubated with conjugated secondary antibodies for 2–4 h, rinsed, and mounted on microscope slides. Confocal images were obtained using a Leica TCS SP5 imaging system or Zeiss confocal laser scanning system equipped with a 488 nm argon, a 543 nm HeNe, or a 633 nm HeNe laser and a 20 × (NA 0.8), a 40 × (NA 1.4) or a 63 × (NA 1.4) oil-immersion objective (Leica Microsystems). The primary antibodies

161

used were chicken-GFP (1:1000; Aves Labs Inc., Cat: GFP-1020), goat-CR (1:1000; Santa Cruz Biotechnology, Cat: 705-165-003), rabbit-CCK (1:800; Sigma, Cat: C2581). The secondary antibodies used were: donkey anti chicken-Alexa Flour 488 (1:1000, Jackson Immunoresearch, Cat: 703-545-155), donkey anti goat-Dylight 650 (1:250; Thermo scientific, Cat: SA5-10089), donkey anti rabbit-Alexa Flour 647 (1:250; Invitrogen, Cat: A31573).

For post hoc morphological reconstruction, patched cells were filled with biocytin (0.6 mg/ml, Sigma) during patch-clamp recording. After recording, 300 µm-thick slices were fixed in 4% PFA at 4°C overnight. To reveal biocytin, the slices were permeabilized with 0.3% Triton X-100 and incubated at 4 °C with streptavidin- conjugated Alexa-488 (1:1000, Jackson Immunoresearch, Cat: 016-540-084) or Alexa-546 (1:1000, Invitrogen, Cat: S11225) in TBS. Z-stacks of biocytin-filled cells were acquired with a 1.5 µm step and merged for detailed reconstruction in Neurolucida 8.26.2 (MBF Bioscience).

Data analysis

Data analysis acquired in electrophysiological experiments was analyzed in Clampfit 10.5 (Molecular Devices) and Igor Pro 4.0 (WaveMetrics). For the analysis of the AP properties, the first AP appearing at current pulse step of +40 to +60 pA within 50 ms from the beginning of current step was analyzed. The AP amplitude was measured from the threshold to the peak. The AP half-width was measured at the voltage level of the half of AP amplitude. Input resistance (Rin) and capacitance (Cm) were obtained using a membrane test in Clampex (Molecular Devices). The membrane time constant (τ) was measured offline using an exponential fit of membrane voltage at -100 mV voltage level in Clampfit 10.5 (Molecular Devices). Soma area, dendritic length and dendritic surface were determined using 3D reconstructions of recorded cells in neurolucida 8.26.2.

For the analysis of spontaneous behavior, three behavioral phases were identified: locomotion, flickering, and immobility. Spontaneous locomotion activity was defined as time points where the speed was higher than 2 cm/s. As the mice were placed on

162

a non-flat surface, the search for the most balanced position translated into flickering, which was detected as periods when the speed was above 0.25 cm/s but below the locomotion threshold. Immobility episodes were defined as the times when the speed was below 0.25 cm/s. We also identified run epochs, which were were defined as the contiguous periods when the instantaneous speed was larger than 1 cm/s for a minimal time period of 3s (221 epochs of mean duration 8.97s ± 0.50s SEM). Stationary epochs were defined as contiguous periods when the instantaneous speed was smaller than 1 cm/s for a minimal time period of 3s (556 epochs of mean duration 9.31s ± 0.31s SEM).

The image analysis was performed offline using Leica LAS, Igor Pro (Wavemetrics, Lake Oswego, USA) and Matlab. For extraction of somatic Ca2+-transients, a region of interest was drawn around individual soma to generate the relative fluorescence change (F) versus time trace. For the purposes of automating the analysis, the 2+ baseline fluorescence level (F0) was taken to be the median of Ca signal. Somatic 2+ Ca -transients were expressed as ΔF/F = (F - F0)/F0. This analysis was performed for 18 different pIS3 cells, of which 14 were recorded across two independent imaging sessions, which were separated by 1 min interval. Analyses between these sessions were pooled together for each cell. Cross-correlations were computed using Matlab’s xcorr(_ _ _ _ ,’coeff’) function.

To extract estimated spike times from Ca2+ events, we used three different methods. The first method was custom-written using Matlab, whereas the other two are previously developed algorithms, which were also available in Matlab. For the first method, we take the gradient of the calcium signal (DF/Dt), and choose times where the DF/Dt trace crosses a threshold value during rises in the calcium signal (i.e. similar to Jackson et al. (2016)). For the threshold value, we used a moving threshold that depends on a normalized version of the amplitude of the raw calcium signal (i.e. calcium signal normalized to have values ranging from zero to one, or .

163

where the threshold at time t, depends on the normalized signal at time t ( , shifted to a value between -0.5 and 0.5, which is added to the mean of the DF/Dt of the calcium signal to produce a value above or below the mean, depending on whether the shifted normalized calcium signal is positive or negative. These time- dependent values are scaled by the standard deviation of the DF/Dt of the calcium signal and then shifted to a higher threshold value by adding 2 times the standard deviation of the DF/Dt of the calcium signal at each time point. We also add the condition that for any spike to be recognized, the corresponding data point in the processed calcium signal must be greater than 1.5 (similar to Jackson et al. (2016), though they use 0.15 instead). This adds accuracy for only identifying times that are during calcium events. While this method does a reasonable job at identifying periods of calcium activation, it is not as precise at this as the other two methods.

Secondly, we use the UFARSA algorithm (Ultra-Fast Accurate Reconstruction of Spiking Activity) (Rahmati et al., 2018). While this algorithm appears to be much more selective in its spike time estimates, it is likely that it often underestimates the number of spikes within a trace compared to the other methods used. This is possibly because it appears to only have been tested against OGB calcium traces. Because of this, we adjusted several parameters past their recommended values to increase the number of within-event spikes (though not to the point where its accuracy in identifying periods of calcium activation begins to drop).

Thirdly, we use the MLspike algorithm (Deneux et al., 2016), of which we tuned the parameters to be optimal for GCaMP6f data, as well as the levels of noise seen in the calcium recordings. To optimize the accuracy for each calcium trace, we use MLspike’s functions for estimating the noise levels. While this algorithm was very good at identifying periods of calcium activation, it is possible that it over-estimated the number of spikes within these periods, compared to the other two methods used, given the parameters that we tuned it to.

164

To examine somatic Ca2+-fluctuations in relation to network oscillations, LFP traces were bandpass filtered in the forward and reverse directions to obtain theta-filtered (5–12 Hz, 4th order) or ripples-filtered (125–250 Hz, 8th order) signals in Matlab as follows:

[z,p,g] = butter(order,[min_frequency max_frequency]/nyquist,’bandpass’);

[sos,g] = zp2sos(z,p,g); filtered_signal = filtfilt(sos,g,LFP_signal);

The onset of the theta-run epoch, which was always associated with an increase in theta power, was defined by the beginning of the locomotion period based on the animal speed trace acquired simultaneously with LFP. To examine time-varying relationships with theta power, we extracted the maximum powers from the spectrograms of the theta-filtered LFP traces between 5-12 Hz in Matlab, as follows: window_theta = 20000; overlap = []; freqrange_theta = 0:0.001:20;

[s,f,t] = spectrogram(filtered_signal,window_theta,overlap,freqrange_theta,fs) a = abs(s); for l = 1:length(t)

Theta_area(l) = max(a((f>5 & f<12),l)); end

We labeled periods where the time-varying theta power was greater than its mean as periods of “high-theta activity”. For obtaining spike phases from the estimated spike times, we took the Hilbert transform of the theta-filtered LFP signal and interpolated the spike phases from this trace. In this case, we set 0° to the rising phase of the theta oscillation.

165

Ripple events were detected automatically using custom-written Matlab code. Specifically, we computed the envelope of the ripple-ﬁltered trace by taking the absolute (abs( )) of its Hilbert transform (hilbert( )). We then set a threshold criteria to identify time periods where the envelope was greater than its mean plus 5 times its standard deviation for at least 12.5 ms. We chose this threshold criteria manually to ensure that ripple events were both large enough in amplitude and long enough in duration. These time periods were then marked as ripple periods.

Statistical analysis

Data was initially tested for normality using a Kolmogorov–Smirnov or Shapiro- Wilcoxon test. If data was normally distributed, standard parametric statistics such as one-way ANOVA were used (*p < 0.05, **p < 0.01, ***p < 0.001) to evaluate statistical signiﬁcance. If data was not normally distributed, Dunn’s test or Mann– Whitney test were used for comparisons of multiple groups. The data are presented as means ± SEM.

Statistical signiﬁcance of cross-correlations were evaluated by doing a structured reshuﬄing of 10s segments of the speed traces or time-varying theta power traces to generate 1,000 surrogate cross-correlation traces. Within this surrogate dataset, the 95th and 99th percentile traces were computed using Matlab’s prctile( ) function, in order to evaluate whether the peak or zeroth cross-correlation magnitudes fell above these percentiles.

We computed Pearson’s linear correlation coefficient between the different spike rate estimates and calcium signals using Matlab’s corr(X,Y,’Type’,’Pearson’) function. Note that this function was also used to compute Pearson’s linear correlation coefficient between the calcium signals and speed traces. For comparisons of percent of normalized numbers of spikes between immobile and mobile states, we used Matlab’s ttest(x,y,’Alpha’, ) function, to perform paired-sample t-tests.

All circular statistical and descriptive measurements were computed using the CircStat toolbox in Matlab. Signiﬁcance of the non-uniformity of circular spike

166

estimate distributions relative to theta phase were tested using Rayleigh’s method (circ_rtest( )), and directional circular statistics were calculated using V-Tests (circ_vtest( )). Mean theta-phase modulation of spike estimates were computed using the circular mean (circ_mean( )) and measures of dispersion included the vector length (circ_r( )) and the angular deviation (circ_std( )).

Acknowledgments

This work was supported by the Canadian Institutes of Health Research (CIHR) and the Natural Sciences and Engineering Research Council of Canada (NSERC). Vincent Villette was supported by the PDF fellowship from Savoy Foundation. Alexandre Guet-McCreight was supported by an NSERC CGS-D2 award. We thank Sarah Côté for technical assistance and Dimitry Topolnik for equipment calibration and maintenance.

Reference

Acsády L, Arabadzisz D, Freund TF. Correlated morphological and neurochemical features identify diﬀerent subsets of vasoactive intestinal polypeptide-immunoreactive interneurons in rat hippocampus. Neuroscience. 1996 Jul; 73(2):299–315. http://www.sciencedirect.com/science/article/pii/0306452295006109, doi:10.1016/0306- 4522(95)00610-9.

Acsády L, Görcs TJ, Freund TF. Diﬀerent populations of vasoactive intestinal polypeptide- immunoreactive interneurons are specialized to control pyramidal cells or interneurons in the hippocampus. Neuroscience. 1996 Jul; 73(2):317–334. http://www.sciencedirect.com/science/article/pii/0306452295006095, doi: 10.1016/0306 4522(95)00609-5.

Ali AB, Deuchars J, Pawelzik H, Thomson AM. CA1 pyramidal to basket and bistratiﬁed cell EPSPs: dual intracellular recordings in rat hippocampal slices. The Journal of Physiology. 1998 Feb; 507(Pt 1):201–217.https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2230771/, doi: 10.1111/j.1469- 7793.1998.201bu.x.

Ali AB, Thomson AM. Facilitating pyramid to horizontal oriens-alveus interneurone inputs: dual intracellular recordings in slices of rat hippocampus. The Journal of Physiology. 1998 Feb; 507(Pt

167

1):185–199. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2230767/, doi: 10.1111/j.1469- 7793.1998.185bu.x.

Amilhon B, Huh CYL, Manseau F, Ducharme G, Nichol H, Adamantidis A, Williams S. Parvalbumin Interneurons of Hippocampus Tune Population Activity at Theta Frequency. Neuron. 2015 Mar; 86(5):1277–1289. http://www.cell.com/article/S0896627315004341/abstract, doi: 10.1016/j.neuron.2015.05.027.

Andersen P, Morris R, Amaral D, Bliss T, O’Keefe J. The Hippocampus Book. Oxford Neuroscience Series, Oxford, New York: Oxford University Press; 2006.

Andrásfalvy BK, Magee JC. Distance-Dependent Increase in AMPA Receptor Number in the Dendrites of Adult Hippocampal CA1 Pyramidal Neurons. Journal of Neuroscience. 2001 Dec; 21(23):9151–9159. http://www.jneurosci.org/content/21/23/9151, doi: 10.1523/JNEUROSCI.21-23- 09151.2001.

Antal M, Fukazawa Y, Eördögh M, Muszil D, Molnár E, Itakura M, Takahashi M, Shigemoto R. Numbers, Densities, and Colocalization of AMPA- and NMDA-Type Glutamate Receptors at Individual Synapses in the Superﬁcial Spinal Dorsal Horn of Rats. Journal of Neuroscience. 2008 Sep; 28(39):9692–9701. http://www.jneurosci.org/content/28/39/9692, doi: 10.1523/JNEUROSCI.1551-08.2008.

Arrigoni E, Greene RW. Schaﬀer collateral and perforant path inputs activate diﬀerent subtypes of NMDA receptors on the same CA1 pyramidal cell. British Journal of Pharmacology. 2004; 142(2):317–322. https://bpspubs.onlinelibrary.wiley.com/doi/abs/10.1038/sj.bjp.0705744, doi: 10.1038/sj.bjp.0705744.

Brandalise F, Carta S, Helmchen F, Lisman J, Gerber U. Dendritic NMDA spikes are necessary for timing-dependent associative LTP in CA3 pyramidal cells. Nature Communications. 2016 Nov; 7:13480. https://www.nature.com/articles/ncomms13480, doi: 10.1038/ncomms13480.

Carnevale NT, Hines ML. The NEURON Book. 1 edition ed. Cambridge, UK ; New York: Cambridge University Press; 2006.

Chamberlain SEL, Yang J, Jones RSG. The Role of NMDA Receptor Subtypes in Short-Term Plasticity in the Rat Entorhinal Cortex. Neural Plasticity. 2008; 2008. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2577183/, doi: 10.1155/2008/872456.

Chamberland S, Salesse C, Topolnik D, Topolnik L. Synapse-Speciﬁc Inhibitory Control of

168

Hippocampal Feedback Inhibitory Circuit. Frontiers in Cellular Neuroscience. 2010; 4. https://www.frontiersin.org/articles/10.3389/fncel.2010.00130/full, doi: 10.3389/fncel.2010.00130.

Chamberland S, Topolnik L. Inhibitory control of hippocampal inhibitory neurons. Frontiers in Neuroscience. 2012; 6. https://www.frontiersin.org/articles/10.3389/fnins.2012.00165/full, doi: 10.3389/fnins.2012.00165.

Chatzikalymniou AP, Skinner FK. Deciphering the Contribution of Oriens-Lacunosum/Moleculare (OLM) Cells to Intrinsic Theta Rhythms Using Biophysical Local Field Potential (LFP) Models. eNeuro. 2018 Aug;p. ENEURO.0146–18.2018. http://www.eneuro.org/content/early/2018/08/22/ENEURO.0146-18.2018, doi:10.1523/ENEURO.0146-18.2018.

Chen TW, Wardill TJ, Sun Y, Pulver SR, Renninger SL, Baohan A, Schreiter ER, Kerr RA, Orger MB, Jayaraman V, Looger LL, Svoboda K, Kim DS. Ultrasensitive ﬂuorescent proteins for imaging neuronal activity. Nature. 2013 Jul; 499(7458):295–300. https://www.nature.com/articles/nature12354, doi: 10.1038/nature12354.

Chittajallu R, Wester JC, Craig MT, Barksdale E, Yuan XQ, Akgül G, Fang C, Collins D, Hunt S, Pelkey KA, McBain CJ. Aﬀerent speciﬁc role of NMDA receptors for the circuit integration of hippocampal neurogliaform cells. Nature Communications. 2017 Jul; 8(1):152. https://www.nature.com/articles/s41467-017-00218-y, doi:10.1038/s41467-017-00218-y.

Croce A, Pelletier JG, Tartas M, Lacaille JC. Aﬀerent-speciﬁc properties of interneuron synapses underlie selective long-term regulation of feedback inhibitory circuits in CA1 hippocampus. The Journal of Physiology. 2010; 588(12):2091–2107. https://physoc.onlinelibrary.wiley.com/doi/abs/10.1113/jphysiol.2010.189316, doi: 10.1113/jphysiol.2010.189316.

David LS, Topolnik L. Target-speciﬁc alterations in the VIP inhibitory drive to hippocampal GABAergic cells after status epilepticus. Experimental Neurology. 2017 Jun; 292:102–112. http://www.sciencedirect.com/science/article/pii/S0014488617300687, doi: 10.1016/j.expneurol.2017.03.007.

Deneux T, Kaszas A, Szalay G, Katona G, Lakner T, Grinvald A, Rózsa B, Vanzetta I. Accurate spike estimation from noisy calcium signals for ultrafast three-dimensional imaging of large neuronal populations in vivo. Nature Communications. 2016 Jul; 7:12190. https://www.nature.com/articles/ncomms12190, doi:10.1038/ncomms12190.

169

Dombeck DA, Harvey CD, Tian L, Looger LL, Tank DW. Functional imaging of hippocampal place cells at cellular resolution during virtual navigation. Nature Neuroscience. 2010 Nov; 13(11):1433– 1440. https://www.nature.com/articles/nn.2648, doi: 10.1038/nn.2648.

Domonkos A, Ledri LN, Laszlovszky T, Cserép C, Borhegyi Z, Papp E, Nyiri G, Freund TF, Varga V. Divergent in vivo activity of non-serotonergic and serotonergic VGluT3–neurones in the median raphe region. The Journal of Physiology. 2016 Jul; 594(13):3775–3790. https://physoc.onlinelibrary.wiley.com/doi/abs/10.1113/JP272036, doi: 10.1113/JP272036.

Dragoi G, Carpi D, Recce M, Csicsvari J, Buzsáki G. Interactions between Hippocampus and Medial Septum during Sharp Waves and Theta Oscillation in the Behaving Rat. Journal of Neuroscience. 1999 Jul; 19(14):6191–6199. http://www.jneurosci.org/content/19/14/6191, doi: 10.1523/JNEUROSCI.19-14-06191.1999.

Dudman JT, Tsay D, Siegelbaum SA. A Role for Synaptic Inputs at Distal Dendrites: Instructive Signals for Hippocampal Long-Term Plasticity. Neuron. 2007 Dec; 56(5):866–879. http://www.sciencedirect.com/science/article/pii/S089662730700815X, doi: 10.1016/j.neuron.2007.10.020.

Emri Z, Antal K, Gulyás AI, Megias M, Freund TF. Electrotonic proﬁle and passive propagation of synaptic potentials in three subpopulations of hippocampal CA1 interneurons. Neuroscience. 2001 Jul; 104(4):1013–1026. http://www.sciencedirect.com/science/article/pii/S0306452201001361, doi: 10.1016/S0306-4522(01)00136-1.

Ernst M, Brauchart D, Boresch S, Sieghart W. Comparative modeling of GABAA receptors: limits, insights, future developments. Neuroscience. 2003 Jul; 119(4):933–943. http://www.sciencedirect.com/science/article/pii/S0306452203002884, doi: 10.1016/S0306- 4522(03)00288-4.

Francavilla R, Villette V, Luo X, Chamberland S, Munoz-Pino E, Camire O, Wagner K, Kis V, Somogyi P, Topolnik L. Connectivity and network state-dependent recruitment of long-range VIP- GABAergic neurons in the mouse hippocampus. bioRxiv. 2018 Jul; p. 364547. https://www.biorxiv.org/content/early/2018/07/15/364547, doi:10.1101/364547.

Frerking M, Schulte J, Wiebe SP, Stäubli U. Spike Timing in CA3 Pyramidal Cells During Behavior: Implications for Synaptic Transmission. Journal of Neurophysiology. 2005 Aug; 94(2):1528–1540. http://jn.physiology.org/content/94/2/1528, doi: 10.1152/jn.00108.2005.

Freund TF, Gulyás AI, Acsády L, Görcs T, Tóth K. Serotonergic control of the hippocampus via

170

local inhibitory interneurons. Proceedings of the National Academy of Sciences. 1990 Nov; 87(21):8501–8505. http://www.pnas.org/content/87/21/8501, doi: 10.1073/pnas.87.21.8501.

Fu Y, Tucciarone JM, Espinosa JS, Sheng N, Darcy DP, Nicoll RA, Huang ZJ, Stryker MP. A Cortical Circuit for Gain Control by Behavioral State. Cell. 2014 Mar; 156(6):1139–1152. https://www.cell.com/cell/abstract/S0092-8674(14)00144-5, doi: 10.1016/j.cell.2014.01.050.

Gillies MJ, Traub RD, LeBeau FEN, Davies CH, Gloveli T, Buhl EH, Whittington MA. A model of atropine-resistant theta oscillations in rat hippocampal area CA1. The Journal of Physiology. 2002 Sep; 543(Pt 3):779–793. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2290530/, doi: 10.1113/jphysiol.2002.024588.

Glickfeld LL, Scanziani M. Distinct timing in the activity of cannabinoid-sensitive and cannabinoid- insensitive basket cells. Nature Neuroscience. 2006 Jun; 9(6):807–815. https://www.nature.com/articles/nn1688, doi:10.1038/nn1688.

Gloveli T, Dugladze T, Saha S, Monyer H, Heinemann U, Traub RD, Whittington MA, Buhl EH. Diﬀerential involvement of oriens/pyramidale interneurones in hippocampal network oscillations in vitro. The Journal of Physiology. 2005; 562(1):131–147. https://physoc.onlinelibrary.wiley.com/doi/abs/10.1113/jphysiol.2004.073007, doi: 10.1113/jphysiol.2004.073007.

Golding NL, Staﬀ NP, Spruston N. Dendritic spikes as a mechanism for cooperative long-term potentiation. Na-ture. 2002 Jul; 418(6895):326–331. https://www.nature.com/articles/nature00854, doi: 10.1038/nature00854.

Guet-McCreight A, Camiré O, Topolnik L, Skinner FK. Using a Semi-Automated Strategy to Develop Multi- Compartment Models That Predict Biophysical Properties of Interneuron-Speciﬁc 3 (IS3) Cells in Hippocampus. eNeuro. 2016 Aug; 3(4). doi: 10.1523/ENEURO.0087-16.2016.

Guet-McCreight A, Luo X, Francavilla R, Topolnik L, Skinner FK. Using computational modeling to estimate synaptic receptor densities along hippocampal CA1 interneuron speciﬁc 3 cell dendrites (Poster). F1000Research. 2017 Aug; 6. https://f1000research.com/posters/6-1552, doi: 10.7490/f1000research.1114780.1.

Guet-McCreight A, Skinner FK. Weak and equally balanced synaptic inputs to interneurons in the CA1 hippocampus characterize in vivo rhythmic states. bioRxiv. 2018 Mar; p. 286369. https://www.biorxiv.org/content/early/2018/03/22/286369, doi: 10.1101/286369.

171

Gulyás AI, Hájos N, Freund TF. Interneurons Containing Calretinin Are Specialized to Control Other Interneurons in the Rat Hippocampus. Journal of Neuroscience. 1996 May; 16(10):3397–3411. http://www.jneurosci.org/content/16/10/3397.

Gulyás AI, Megias M, Emri Z, Freund TF. Total Number and Ratio of Excitatory and Inhibitory Synapses Converging onto Single Interneurons of Diﬀerent Types in the CA1 Area of the Rat Hippocampus. Journal of Neuroscience. 1999 Nov; 19(22):10082–10097. http://www.jneurosci.org/content/19/22/10082.

Hainmüller T, Krieglstein K, Kulik A, Bartos M. Joint CP-AMPA and group I mGlu receptor activation is required for synaptic plasticity in dentate gyrus fast-spiking interneurons. Proceedings of the National Academy of Sciences. 2014 Sep; 111(36):13211–13216. http://www.pnas.org/content/111/36/13211, doi: 10.1073/pnas.1409394111.

Hangya B, Borhegyi Z, Szilágyi N, Freund TF, Varga V. GABAergic Neurons of the Medial Septum Lead the Hippocampal Network during Theta Activity. Journal of Neuroscience. 2009 Jun; 29(25):8094–8102. http://www.jneurosci.org/content/29/25/8094, doi: 10.1523/JNEUROSCI.5665- 08.2009.

Helboe L, Egebjerg J, de Jong IEM. Distribution of serotonin receptor 5-HT6 mRNA in rat neuronal subpopulations: A double in situ hybridization study. Neuroscience. 2015 Dec; 310:442–454. http://www.sciencedirect.com/science/article/pii/S030645221500891X, doi: 10.1016/j.neuroscience.2015.09.064.

Jackson J, Ayzenshtat I, Karnani MM, Yuste R. VIP+ interneurons control neocortical activity across brain states. Journal of Neurophysiology. 2016 Jun; 115(6):3008–3017. http://jn.physiology.org/content/115/6/3008, doi:10.1152/jn.01124.2015.

Katona L, Lapray D, Viney TJ, Oulhaj A, Borhegyi Z, Micklem BR, Klausberger T, Somogyi P. Sleep and Movement Diﬀerentiates Actions of Two Types of Somatostatin-Expressing GABAergic Interneuron in Rat Hippocampus. Neuron. 2014 May; 82(4):872–886. http://www.cell.com/neuron/abstract/S0896-6273(14)00297-9, doi:10.1016/j.neuron.2014.04.007.

Katz B, Miledi R. The role of calcium in neuromuscular facilitation. The Journal of Physiology. 1968 Mar; 195(2):481–492. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1351674/.

Klausberger T, Magill PJ, Márton LF, Roberts JDB, Cobden PM, Buzsáki G, Somogyi P. Brain- state- and cell type-speciﬁc ﬁring of hippocampal interneurons in vivo. Nature. 2003 Feb; 421(6925):844–848. https://www.nature.com/nature/journal/v421/n6925/full/nature01374.html, doi:

172

10.1038/nature01374.

Klausberger T, Somogyi P. Neuronal Diversity and Temporal Dynamics: The Unity of Hippocampal Circuit Operations. Science. 2008 Jul; 321(5885):53–57. http://www.sciencemag.org/content/321/5885/53, doi:10.1126/science.1149381.

Kullmann DM, Lamsa KP. Long-term synaptic plasticity in hippocampal interneurons. Nature Reviews Neuroscience. 2007 Sep; 8(9):687–699. https://www.nature.com/articles/nrn2207, doi: 10.1038/nrn2207.

Köhler C. Distribution and morphology of vasoactive intestinal polypeptide-like immunoreactive neurons in regio superior of the rat hippocampal formation. Neuroscience Letters. 1982 Dec; 33(3):265–270. http://www.sciencedirect.com/science/article/pii/0304394082903822, doi: 10.1016/0304-3940(82)90382-2.

Lamsa K, Irvine EE, Giese KP, Kullmann DM. NMDA receptor-dependent long-term potentiation in mousehippocampal interneurons shows a unique dependence on Ca2+/calmodulin-dependent kinases. The Journal of Physiology. 2007; 584(3):885–894. https://physoc.onlinelibrary.wiley.com/doi/abs/10.1113/jphysiol.2007.137380, doi: 10.1113/jphysiol.2007.137380.

Lapray D, Lasztoczi B, Lagler M, Viney TJ, Katona L, Valenti O, Hartwich K, Borhegyi Z, Somogyi P, Klausberger T. Behavior-dependent specialization of identiﬁed hippocampal interneurons. Nature Neuroscience. 2012 Sep;15(9):1265–1271. https://www.nature.com/neuro/journal/v15/n9/full/nn.3176.html, doi: 10.1038/nn.3176.

Le Roux N, Cabezas C, Böhm UL, Poncer JC. Input-speciﬁc learning rules at excitatory synapses onto hippocampal parvalbumin-expressing interneurons. The Journal of Physiology. 2013 Apr; 591(7):1809–1822. doi:10.1113/jphysiol.2012.245852.

Lei S, McBain CJ. Distinct NMDA Receptors Provide Diﬀerential Modes of Transmission at Mossy Fiber-Interneuron Synapses. Neuron. 2002 Mar; 33(6):921–933. http://www.sciencedirect.com/science/article/pii/S0896627302006086, doi: 10.1016/S0896- 6273(02)00608-6.

Leão RN, Mikulovic S, Leão KE, Munguba H, Gezelius H, Enjin A, Patra K, Eriksson A, Loew LM, Tort ABL, Kullander K. OLM interneurons diﬀerentially modulate CA3 and entorhinal inputs to hippocampal CA1 neurons. Nature Neuroscience. 2012 Nov; 15(11):1524–1530. https://www.nature.com/articles/nn.3235, doi: 10.1038/nn.3235.

173

Losonczy A, Zhang L, Shigemoto R, Somogyi P, Nusser Z. Cell type dependence and variability in the short-term plasticity of EPSCs in identiﬁed mouse hippocampal interneurones. The Journal of Physiology. 2002 Jul; 542(Pt1):193–210. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2290398/, doi: 10.1113/jphysiol.2002.020024.

Lovett-Barron M, Kaifosh P, Kheirbek MA, Danielson N, Zaremba JD, Reardon TR, Turi GF, Hen R, Zemelman BV, Losonczy A. Dendritic Inhibition in the Hippocampus Supports Fear Learning. Science. 2014 Feb; 343(6173):857–863. http://science.sciencemag.org/content/343/6173/857, doi: 10.1126/science.1247485.

Lovett-Barron M, Turi GF, Kaifosh P, Lee PH, Bolze F, Sun XH, Nicoud JF, Zemelman BV, Sternson SM, Losonczy A. Regulation of neuronal input transformations by tunable dendritic inhibition. Nature Neuroscience. 2012 Mar; 15(3):423–430. https://www.nature.com/articles/nn.3024, doi: 10.1038/nn.3024.

Léránth C, Frotscher M, To¨mbo¨l T, Palkovits M. Ultrastructure and synaptic connections of vasoactive intestinal polypeptide-like immunoreactive non-pyramidal neurons and axon terminals in the rat hippocampus. Neuroscience. 1984 Jun; 12(2):531–542. http://www.sciencedirect.com/science/article/pii/030645228490071X, doi: 10.1016/0306- 4522(84)90071-X.

Madisen L, Zwingman TA, Sunkin SM, Oh SW, Zariwala HA, Gu H, Ng LL, Palmiter RD, Hawrylycz MJ, Jones AR, Lein ES, Zeng H. A robust and high-throughput Cre reporting and characterization system for the whole mouse brain. Nature Neuroscience. 2010 Jan; 13(1):133–140. https://www.nature.com/articles/nn.2467, doi:10.1038/nn.2467.

Magee JC, Cook EP. Somatic EPSP amplitude is independent of synapse location in hippocampal pyramidal neurons. Nature Neuroscience. 2000 Sep; 3(9):895–903. https://www.nature.com/articles/nn0900_895, doi:10.1038/78800.

Masugi-Tokita M, Tarusawa E, Watanabe M, Molnár E, Fujimoto K, Shigemoto R. Number and Density of AMPA Receptors in Individual Synapses in the Rat Cerebellum as Revealed by SDS Digested Freeze-Fracture Replica Labeling. Journal of Neuroscience. 2007 Feb; 27(8):2135–2144. http://www.jneurosci.org/content/27/8/2135, doi: 10.1523/JNEUROSCI.2861-06.2007.

Mizuseki K, Sirota A, Pastalkova E, Buzsáki G. Theta Oscillations Provide Temporal Windows for Local Circuit Computation in the Entorhinal-Hippocampal Loop. Neuron. 2009 Oct; 64(2):267–280. http://www.cell.com/neuron/abstract/S0896-6273(09)00673-4, doi: 10.1016/j.neuron.2009.08.037.

174

Nusser Z. A new approach to estimate the number, density and variability of receptors at central synapses. European Journal of Neuroscience. 1999; 11(3):745–752. https://onlinelibrary.wiley.com/doi/abs/10.1046/j.1460-9568.1999.00535.x, doi: 10.1046/j.1460- 9568.1999.00535.x.

Nusser Z, Cull-Candy S, Farrant M. Diﬀerences in Synaptic GABAA Receptor Number Underlie Variation in GABA Mini Amplitude. Neuron. 1997 Sep; 19(3):697–709. http://www.sciencedirect.com/science/article/pii/S0896627300803827, doi: 10.1016/S0896- 6273(00)80382-7.

NyÍri G, Stephenson FA, Freund TF, Somogyi P. Large variability in synaptic n-methyl-d-aspartate receptor density on interneurons and a comparison with pyramidal-cell spines in the rat hippocampus. Neuroscience. 2003 Jun;119(2):347–363. http://www.sciencedirect.com/science/article/pii/S030645220300157X, doi: 10.1016/S0306- 4522(03)00157-X.

Otmakhova NA, Otmakhov N, Lisman JE. Pathway-Speciﬁc Properties of AMPA and NMDA- Mediated Transmission in CA1 Hippocampal Pyramidal Cells. Journal of Neuroscience. 2002 Feb; 22(4):1199–1207. http://www.jneurosci.org/content/22/4/1199, doi: 10.1523/JNEUROSCI.22-04- 01199.2002.

Papp EC, Hájos N, Acsády L, Freund TF. Medial septal and median raphe innervation of vasoactive intestinal polypeptide-containing interneurons in the hippocampus. Neuroscience. 1999 Feb; 90(2):369–382. http://www.sciencedirect.com/science/article/pii/S0306452298004552, doi: 10.1016/S0306-4522(98)00455-2.

Paul A, Crow M, Raudales R, He M, Gillis J, Huang ZJ. Transcriptional Architecture of Synaptic Communication Delineates GABAergic Neuron Identity. Cell. 2017 Oct; 171(3):522–539.e20. doi: 10.1016/j.cell.2017.08.032.

Pi HJ, Hangya B, Kvitsiani D, Sanders JI, Huang ZJ, Kepecs A. Cortical interneurons that specialize in disinhibitory control. Nature. 2013 Nov; 503(7477):521–524. https://www.nature.com/articles/nature12676, doi:10.1038/nature12676.

Pinto L, Dan Y. Cell-Type-Speciﬁc Activity in Prefrontal Cortex during Goal-Directed Behavior. Neuron. 2015 Jul; 87(2):437–450. http://www.sciencedirect.com/science/article/pii/S0896627315005644, doi:10.1016/j.neuron.2015.06.021.

175

Pouille F, Scanziani M. Routing of spike series by dynamic circuits in the hippocampus. Nature. 2004 Jun;429(6993):717–723. https://www.nature.com/articles/nature02615, doi: 10.1038/nature02615.

Price CJ, Cauli B, Kovacs ER, Kulik A, Lambolez B, Shigemoto R, Capogna M. Neurogliaform Neurons Form a Novel Inhibitory Network in the Hippocampal CA1 Area. Journal of Neuroscience. 2005 Jul; 25(29):6775–6786. http://www.jneurosci.org/content/25/29/6775, doi: 10.1523/JNEUROSCI.1135-05.2005.

Prisco GVD, Albo Z, Vertes RP, Kocsis B. Discharge properties of neurons of the median raphe nucleus during hippocampal theta rhythm in the rat. Experimental Brain Research. 2002 Aug; 145(3):383–394. https://link.springer.com/article/10.1007/s00221-002-1123-8, doi: 10.1007/s00221- 002-1123-8.

Rahmati V, Kirmse K, Holthoﬀ K, Kiebel SJ. Ultra-fast accurate reconstruction of spiking activity from calcium imaging data. Journal of Neurophysiology. 2018 Feb; 119(5):1863–1878. https://www.physiology.org/doi/abs/10.1152/jn.00934.2017, doi: 10.1152/jn.00934.2017.

Regehr WG, Tank DW. Postsynaptic NMDA receptor-mediated calcium accumulation in hippocampal CA1 pyramidal cell dendrites. Nature. 1990 Jun; 345(6278):807–810. https://www.nature.com/articles/345807a0, doi: 10.1038/345807a0.

Royer S, Zemelman BV, Losonczy A, Kim J, Chance F, Magee JC, Buzsáki G. Control of timing, rate and bursts of hippocampal place cells by dendritic and somatic inhibition. Nature Neuroscience. 2012 May; 15(5):769–775. https://www.nature.com/articles/nn.3077, doi: 10.1038/nn.3077.

Sambandan S, Sauer JF, Vida I, Bartos M. Associative Plasticity at Excitatory Synapses Facilitates Recruitment of Fast-Spiking Interneurons in the Dentate Gyrus. Journal of Neuroscience. 2010 Sep; 30(35):11826–11837. http://www.jneurosci.org/content/30/35/11826, doi: 10.1523/JNEUROSCI.2012-10.2010.

Shigemoto R, Kulik A, Roberts JDB, Ohishi H, Nusser Z, Kaneko T, Somogyi P. Target-cell-speciﬁc concentration of a metabotropic glutamate receptor in the presynaptic active zone. Nature. 1996 Jun; 381(6582):523–525. https://www.nature.com/articles/381523a0, doi: 10.1038/381523a0.

Sik A, Penttonen M, Ylinen A, Buzsaki G. Hippocampal CA1 interneurons: an in vivo intracellular labeling study. Journal of Neuroscience. 1995 Oct; 15(10):6651–6665. http://www.jneurosci.org/content/15/10/6651, doi:10.1523/JNEUROSCI.15-10-06651.1995.

176

Silver RA. Neuronal arithmetic. Nature Reviews Neuroscience. 2010 Jul; 11(7):474–489. https://www.nature.com/articles/nrn2864, doi: 10.1038/nrn2864.

Siwani S, França ASC, Mikulovic S, Reis A, Hilscher MM, Edwards SJ, Leão RN, Tort ABL, Kullander K. OLMa2 Cells Bidirectionally Modulate Learning. Neuron. 2018 Jul; 99(2):404–412.e3. doi: 10.1016/j.neuron.2018.06.022.

Spruston N, Jonas P, Sakmann B. Dendritic glutamate receptor channels in rat hippocampal CA3 and CA1 pyramidal neurons. The Journal of Physiology. 1995 Jan; 482(Pt 2):325–352. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1157732/.

Stuart G, Spruston N. Determinants of Voltage Attenuation in Neocortical Pyramidal Neuron Dendrites. Journal of Neuroscience. 1998 May; 18(10):3501–3510. http://www.jneurosci.org/content/18/10/3501, doi:10.1523/JNEUROSCI.18-10-03501.1998.

Sun HY, Dobrunz LE. Presynaptic Kainate Receptor Activation Is a Novel Mechanism for Target Cell-Speciﬁc Short-Term Facilitation at Schaﬀer Collateral Synapses. Journal of Neuroscience. 2006 Oct; 26(42):10796–10807. http://www.jneurosci.org/content/26/42/10796, doi: 10.1523/JNEUROSCI.2746-06.2006.

Sun HY, Lyons SA, Dobrunz LE. Mechanisms of target-cell speciﬁc short-term plasticity at Schaﬀer collateral synapses onto interneurones versus pyramidal cells in juvenile rats. The Journal of Physiology. 2005; 568(3):815–840. https://physoc.onlinelibrary.wiley.com/doi/abs/10.1113/jphysiol.2005.093948, doi:10.1113/jphysiol.2005.093948.

Swanson GT, Kamboj SK, Cull-Candy SG. Single-Channel Properties of Recombinant AMPA Receptors Depend on RNA Editing, Splice Variation, and Subunit Composition. Journal of Neuroscience. 1997 Jan; 17(1):58–69. http://www.jneurosci.org/content/17/1/58, doi: 10.1523/JNEUROSCI.17-01-00058.1997.

Szabo A, Somogyi J, Cauli B, Lambolez B, Somogyi P, Lamsa KP. Calcium-Permeable AMPA Receptors Provide a Common Mechanism for LTP in Glutamatergic Synapses of Distinct Hippocampal Interneuron Types. Journal of Neuroscience. 2012 May; 32(19):6511–6516. http://www.jneurosci.org/content/32/19/6511, doi:10.1523/JNEUROSCI.0206-12.2012.

Thomson AM. Activity-dependent properties of synaptic transmission at two classes of connections made by rat neocortical pyramidal axons in vitro. The Journal of Physiology. 1997 Jul; 502(Pt 1):131–147. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1159577/.

177

Tyan L, Chamberland S, Magnin E, Camiré O, Francavilla R, David LS, Deisseroth K, Topolnik L. Dendritic Inhibition Provided by Interneuron-Speciﬁc Cells Controls the Firing Rate and Timing of the Hippocampal Feedback Inhibitory Circuitry. Journal of Neuroscience. 2014 Mar; 34(13):4534– 4547. http://www.jneurosci.org/content/34/13/4534, doi: 10.1523/JNEUROSCI.3813-13.2014.

Unal G, Joshi A, Viney TJ, Kis V, Somogyi P. Synaptic Targets of Medial Septal Projections in the Hippocampus and Extrahippocampal Cortices of the Mouse. Journal of Neuroscience. 2015 Dec; 35(48):15812–15826.http://www.jneurosci.org/content/35/48/15812, doi: 10.1523/JNEUROSCI.2639-15.2015.

Varga C, Golshani P, Soltesz I. Frequency-invariant temporal ordering of interneuronal discharges during hippocampal oscillations in awake mice. Proceedings of the National Academy of Sciences. 2012 Oct; 109(40):E2726–E2734. http://www.pnas.org/content/109/40/E2726, doi: 10.1073/pnas.1210929109.

Varga C, Oijala M, Lish J, Szabo GG, Bezaire M, Marchionni I, Golshani P, Soltesz I. Functional ﬁssion of parvalbumin interneuron classes during fast network events. eLife. 2014 Nov; 3:e04006. https://elifesciences.org/articles/04006, doi: 10.7554/eLife.04006.

Villette V, Levesque M, Miled A, Gosselin B, Topolnik L. Simple platform for chronic imaging of hippocampalactivity during spontaneous behaviour in an awake mouse. Scientiﬁc Reports. 2017 Feb; 7:43388. https://www.nature.com/articles/srep43388, doi: 10.1038/srep43388.

Viney TJ, Lasztoczi B, Katona L, Crump MG, Tukker JJ, Klausberger T, Somogyi P. Network state- dependent inhibition of identiﬁed hippocampal CA3 axo-axonic cells in vivo. Nature Neuroscience. 2013 Dec; 16(12):1802–1811. https://www.nature.com/articles/nn.3550, doi: 10.1038/nn.3550.

Wang DV, Yau HJ, Broker CJ, Tsou JH, Bonci A, Ikemoto S. Mesopontine median raphe regulates hippocampal ripple oscillation and memory consolidation. Nature Neuroscience. 2015 May; 18(5):728–735. https://www.nature.com/articles/nn.3998, doi: 10.1038/nn.3998.

Wyskiel DR, Andrade R. Serotonin excites hippocampal CA1 GABAergic interneurons at the stratum radiatum-stratum lacunosum moleculare border. Hippocampus. 2016 Sep; 26(9):1107– 1114. https://onlinelibrary.wiley.com/doi/abs/10.1002/hipo.22611, doi: 10.1002/hipo.22611.

Zucker RS, Regehr WG. Short-Term Synaptic Plasticity. Annual Review of Physiology. 2002; 64(1):355–405. https://doi.org/10.1146/annurev.physiol.64.092501.114547, doi: 10.1146/annurev.physiol.64.092501.114547.

178

Figures and legends

Figure 1. Properties of SC- and TA-EPSCs in IS3 cells. A: Two-photon images of CA1 area in slices from VIP-eGFP (green) and VIP-tdTomato (red) mice. B-C: Anatomical reconstructions of IS3 cells in VIP-eGFP (B) and VIP-tdTomato (C) mice. D: Examples of recordings of sEPSCS and sIPSCs (top) and a summary cumulative distribution for sEPSC and sIPSC frequency (bottom). E: Summary bar graphs showing the basic properties of SC-IS3 and TA-IS3 EPSCs. The individual data points (each point is a single cell) are shown for SC (circles) and TA (squares) EPSCs, respectively. The bars indicate the means (*: P<0.05, one-way ANOVA). F: Left top, schematics illustrating the location of the stimulation electrode for SC pathway activation and an example trace of SC-EPSC recorded at -70 mV. Left bottom, example traces of SC-EPSCs recorded at different voltage levels. Right, I-V curve of SC-EPSCs. G: Left bottom, schematics illustrating the location of stimulation for TA pathway and an example trace of TA-EPSC. Left bottom, example traces of TA-EPSCs obtained at different voltage levels. Black and red vertical dotted lines indicate the peak levels for fast and slow EPSC components, respectively. Right, I-V curves of TA-EPSC fast (black) and slow (red) components. H: Example traces of TA-EPSCs at different voltage levels (left) and an I-V curve (right) obtained in the presence of NBQX.

179

A B C

E F G

Figure 2. Input-specific temporal and spatial summation. A: Schematics illustrating the stimulating electrode location for the SC pathway activation (upper), and the example traces in response to repetitive stimulation (10, 20 and 40 Hz) in voltage (middle) and current (bottom) clamp mode. B: Same as panel A for TA-stimulation. C: Normalized EPSC amplitudes in response to repetitive stimulation of SC vs TA inputs. Solid lines and closed symbols indicate TA-EPSC, dashed lines and open symbols indicate SC-EPSC (*: P<0.05, one-way ANOVA. The EPSCs induced by the last three pulses were compared between TA and SC pathway). D: The average number of APs generated in response to different stimulation frequencies at TA vs SC synapses. The statistical analysis was performed on the APs induced by the fifth pulse at 20Hz. (*: P<0.05, one-way ANOVA). E: Two-photon image (single focal plane) showing the IS3 cell filled with Alexa-594 during recording and the location of the target areas in proximal (SC) and distal (TA) dendrites for two-photon glutamate uncaging; F-G: Representative responses of IS3s when glutamate was released at shorter (bottom, 2-4 μm) and longer dendritic areas (top, 5-8 m) at SC (F) vs TA (G) inputs. H: Top, representative example of the AP probability in relation to photostimulation extend at two inputs. Bottom, summary data for a group of cells illustrating the transition to AP initiation when the length of stimulated dendrite was increased. Circles correspond to the SC and squares - to the TA inputs, respectively. Red lines indicate the mean threshold for dendritic length, at which it was possible to induce spikes at SC (dotted line) vs TA (continuous line) inputs. The difference in the dendritic length was not significant between the two inputs. P > 0.05; Mann-Whitney test.

180

Figure 3. EPSCs and IPSCs along model dendritic tree. Simulated somatically-recorded EPSCs (blue) and IPSCs (red) generated from synapses at selected locations along the dendritic arbor of the IS3 cell model (i.e. with linearly rectiﬁed weights). Note that both the IPSCs and EPSCs exhibit a waveform broadening and amplitude deterioration with synaptic distance from soma. Red arrows denote the approximate synaptic locations. Dashed brown line shows the border between proximal and distal dendrites at 300 μm from the soma. Labels on the y-axis denote the dendritic subtree labels. RAD = Stratum Radiatum; LM = Stratum Lacunosum Moleculare; AIS = axon initial segment.

181

182

Figure 4. Recruitment of IS3 cell models with theta-timed inputs. A1. Relative timing of different theta-timed presynaptic populations. A2. Same as A1, but in polar plot format with an example plot of SDprox1 phase preference when theta-inputs are perfectly timed. A3. Example raster plot of the presynaptic theta-timed populations. B. Example simulated SDprox1 and SDprox2 voltage traces of resulting baseline in vivo-like conditions, X1 theta inputs, X2 theta inputs and X3 theta inputs. C. Top subplot: Power spectral density (PSD) of the SDprox1 and SDprox2 IS3 cell model spike trains during baseline in vivo-like conditions, X1 theta inputs, X2 theta inputs, and X3 theta inputs. Bottom subplot: first four bars show the area under the PSD between 4-12 Hz, and the last four bars show the PSD magnitude at 8 Hz. D. Polar plots showing phase preference of the SDprox1 and SDprox2 IS3 cell models. Inner traces show the mean frequencies of the interspike intervals, binned according to phase (binned standard deviations shown in the shaded areas). Outer histograms show the binned spike phases (bin width = 14.4◦).

Figure 5. Recruitment of IS3 cell models with SWRs. A: Example synaptic locations of SWR-timed inputs. Note that synaptic locations are chosen randomly in all simulations. B: Raster plots of SWR- timed presynaptic populations. C: Example voltage traces during SWR-timed inputs. Dashed black line coincides with the start of the SWR event. Note that we show two examples for both SDprox1 and SDprox2 where we resample synaptic locations and baseline presynaptic spike times. D: Example peri-stimulus histogram of the voltage traces shown in C (bin size is 67ms - 30 bins). Results show that delayed proximal inhibition is suﬃcient to bring the model spiking back to baseline levels.

183

Acronyms: CA3-Cornu Ammonis 3, SRinh-Proximal Feedforward Inhibition, SLMinh-Distal Feedforward Inhibition.

Figure 6. Imaging activity of pIS3 cells in awake mice. A: Schematic of simultaneous calcium imaging and LFP recordings in awake mice. B: Bar graph showing the average running speed during diﬀerent behavioral states. C: Representative two-photon images showing VIP+ cells located within O/A, PYR and RAD. Note a smaller size round soma of VIP+ cells within PYR and RAD. D: Representative traces of simultaneous LFP recording [from top to bottom, raw trace and ﬁltered for

184

theta: 5–12 Hz (black) or ripples: 125-250 Hz (red, with black stars showing individual ripple episodes)] and somatic calcium transients (blue) from a pIS3 cell during different behavioural states (grey trace: the higher animal speed during locomotion is reported as a high-frequency step pattern). The post hoc identification of this cell as CR+ is illustrated in F. E: Left, expanded theta-run period indicated by bracket in D (red line indicates the duration of the running period). Right, expanded ripple event indicated by an arrow in D. F: Left, two-photon image showing an IS3 soma labeled with GCaMP6f and recorded in vivo. Right, post hoc verification of CR expression in the same cell.

185

Figure 7. Delayed recruitment of pIS3 cells during theta-run epochs. A: Representative simultaneously recorded traces of a calcium signal from a pIS3 cell (red), a theta-ﬁltered (5-12Hz) LFP recording (black), and animal speed (blue trace). B-D: Average traces of speed (B), theta power (C), and calcium activity (D) during run-starts (left panels; averages are across 64 run-starts selected across 17/18 cells) and run-stops (right panels; averages are across 69 run-stops selected across 18/18 cells). The vertical dashed lines show the transition into (left panels) or out of (right panels) run epochs. "S" denotes the start time of a run epoch, and "E" denotes the end time of a run epoch. For the run epoch portions in the time axes, the time vector is normalized (i.e. by using Matlab’s interp1( ) function) such that the run epochs occupy the same span across the x-axis, despite having variable durations. The times outside of these run epochs are shown normally (i.e. ±5s pre- and post- run epochs). Blue lines show the mean traces for each cell. The mean of means across cells are shown in black, except for the calcium signal in D, which shows the median of means. E-F: Cross-correlations between the calcium signal and speed (E), and the time-varying theta power (F). Average traces across all recordings are shown in black, and individual session traces are shown in grey. Speed cross-correlations (E) were quite variable but generally tended to exhibit large peaks near zero. Theta-ﬁltered cross-correlations (F) also exhibited a mean large peak near zero. G: Pearson correlations of speed with either calcium signal magnitude (top subplot), DF/Dt estimated spike rates (upper middle subplot), UFARSA estimated spike rates (lower middle subplot), or MLspike estimated spike rates (bottom subplot; * < 0.05; ** < 0.001). Red area indicates the mean ± standard deviation.

186

Figure 8. Estimations of spike times from calcium signals suggest spike preferences during movement and near the rising/peak phases of theta, with a high degree of spike phase dispersion. Three different methods were used: DF/Dt, UFARSA, and MLspike. A: Raster plots for each cell, where blue dots are estimated spikes during immobile states, and red dots are estimated spikes during mobile states. Above each of the raster plots is a calcium trace from cell #66, above which are the estimated spike extractions using each method. The dashed line in the calcium traces indicates the time point where there was a 1 min break between recording sessions. B: Percent of spikes for each state normalized relative to the states where the animal was spending most of their time (statistical significance across states was evaluated using paired-sample t-tests). The dashed lines indicate which data points belong to the same cell (black dots). Bars indicate the mean, and the error bars indicate the standard deviation. Statistical significance is indicated by "*" (*p < 0.05, **p < 0.01, ***p < 0.001). C: Pooled distribution of spike phases (i.e. across all cells) relative to the theta- filtered LFP, where spike count is shown on the radial axis (bin width = 14.4◦). These pooled

187

distributions were significantly non-uniform (Rayleigh’s test), with significant non-uniform polarities (v- test) towards the circular means of the distributions (see Table 1). D: Mean of means test. Each line shows the mean spike phase preference of a cell along the polar axis, where the vector length is shown along the radial axis. The red lines indicate the mean of means angles (i.e. polar axis) and the mean vector lengths (i.e. radial axis). Again, these were significantly non-uniform (Rayleigh’s test), with significant non-uniform polarities (V-Test) towards the circular mean of the means (see Table 1).

Figure 9. pIS3 cells are not activated during SWRs. A: Peri-stimulus time histogram of all calcium traces (blue lines) that occur during SWRs. The red line shows the calcium trace that corresponds to the example SWR shown above the plot. The black line shows the mean across all calcium traces. The dashed lines highlight the ±30ms time window surrounding the SWR, which is centered at 0ms along the time axis. B: Peri-stimulus time histogram of the spike times estimated from the MLspike and DF/Dt algorithms. Note that UFARSA is not shown since only two estimated spikes occurred within this ±200ms time window (i.e. across all ripples observed), and both of these spikes were not within the ±30ms time window of any ripple. Note that the ripple index on the y-axis highlights how many ripples were analyzed, and the diﬀerent dot colors corresponds to estimated spike times estimated from diﬀerent cells. Bin size in top subplots is 8 ms (i.e. 50 bins).

188

Figure 1–Figure supplement 1. Immunolabelling for CR and CCK in the CA1 in VIP-eGFP (green) and VIP-tdTomato (red) mouse lines. A,B. GFP (green), CR (red) and CCK (red) immunoreactivity in the hippocampus. Insets in the lower right corner show examples of soma co-expressing GFP and CR (B) or GFP and CCK (D). C,D. tdTomato (red), CR (green) and CCK (green) immunolabelling in the hippocampus. Insets in the lower right corner show examples of soma co-expressing tdTomato and CR (C) or tdTomato and CCK (D).

189

Figure 1–Figure supplement 2. Membrane and morphological properties of IS3 cells in VIP- eGFP (green) and VIP-tdTomato (red) mouse lines. A,B. Example traces of voltage response (top) to depolarizing and hyperpolarizing current steps (bottom). C. Comparison of membrane properties and morphological parameters of IS3 cells in VIP-eGFP (green) and VIP-tdTomato (red) mouse lines.

190

Figure 3–Figure supplement 1. Receptor numbers, reversal potentials and electrotonic distances. A. Predicted ranges of receptors per synapse, calculated according to as the fitted weights (blue) and the linearly recti1ed distance-dependent weights (red). Note that each x-axis data point shows a range which is larger or smaller depending on the thickness (along the y-axis) of the plotted line. Dashed red line denotes the maximal limits as per previous 1ndings from other cell types in the literature. Line bifurcations in this plot correspond to bifurcations in the dendritic morphology. B1. Estimated reversal potential of each synapse along the dendritic arbor of the model using a simulated protocol that is similar to what is used experimentally. Note that the true reversal potential of each excitatory synapse is 0 mV, and does not change during these simulations. Line bifurcations in this plot correspond to bifurcations in the dendritic morphology. B2. Electrotonic distance (i.e. decay of a 1 mV signal) for voltage 2owing into the soma (left) and voltage 2owing out of the soma (right). Note that the subtree organization approximately follows the same organization as in B1, suggesting that reversal potential recordings in IS3 cells are sensitive to distance-dependent voltage decay.

191

Figure 4–Figure supplement 1. Excitatory presynaptic populations can modulate timing of IS3 cell spikes during theta rhythms. A1/B1. Same as in Figure 4C for X1 theta-timed inputs, but during removal of speci1ed presynaptic theta-timed populations. Bottom subplots show only the PSD at 8 Hz. A2/B2. Same as in the Figure 4D theta polar plots, but during removal of either CA3 or EC3. C/D. Same as in A/B, but with doubled presynaptic populations instead of removed. Acronyms: CA3 - Cornu Ammonis 3, EC3 - Entorhinal Cortex Layer 3, SR90/180/270-Theta Peak/Falling Phase/Trough -Timed Proximal Inhibition, SLM90/180/270 - Theta Peak/Falling Phase/Trough -Timed Distal Inhibition.

192

193

Figure 4–Figure supplement 2. Effect of fractional randomness (noise) on theta-recruitment in the SDprox1 model. A. Example X1 theta raster plots of the presynaptic theta-timed populations. B. Same as in Figure 4C. C-E. Polar plots (same as in Figure 4D) showing phase preference of the IS3 cell model for X1 (C), X2 (D), and X3 (E) theta inputs (bin width = 14.4◦).

Figure 6–Figure supplement 1. Morphological criteria used for the VIP cell identification. A- B: Neurolucida 3D reconstructions (A1, B1) and confocal images (A2-A3, B2-B3; maximal projections of 10–32 sections) illustrating the analysis of somatic parameters in anatomically con1rmed IS3 and BCs filled during in vitro patch-clamp recordings with biocytin. 3D rendering of somatic surface (A2, B2) was used to derive somatic diameters in medio-lateral (X-axis) and rostro-caudal (Y-axis) dimensions (A3, B3). C: Summary bar graphs illustrating the distributions of somatic diameters in a group of cells (IS3, blue, n = 36; BC, black, n = 11). D-E: Two-photon raw (D1, E1) and mask (D2, E2) images of VIP interneurons recorded in vivo showing the extraction of somatic X- and Y- parameters for cell identification. F: Summary data showing the clustering of in vivo recorded VIP cells based on the distribution of X- and Y-parameters (pIS3, blue, n = 9; pBC, black, n = 12). Dotted lines indicate the cut-off levels obtained from in vitro analysis and used for in vivo cell segregation (X = 15.9 µm, Y= 11.3 µm). G: Representative confocal images of horizontal hippocampal sections processed for post hoc immunohistochemical identification following in vivo two-photon experiments illustrating the expression of GFP (G1), CR (G2) and both (G3). H: Summary group data showing the

194

clustering of CR+ and CR- VIP cells based on the distribution of X- and Y-parameters (pIS3, blue, n = 17; pBC, black, n = 13). Dotted lines indicate the same cut-off levels.

195

Figure 7–Figure supplement 1. Summary statistics of cross-correlations and Pearson correlations between theta power and calcium signal. A-B: Analysis of speed x calcium signal cross-correlations (for raw traces, see Figure 7E). Dots indicate the lag time of the maximum peaks versus the magnitude of the maximum (A) or zeroth (B) peaks seen in the cross-correlations. Histograms along the ordinate axes show the distributions of cross-correlation maximum peak amplitudes and zeroth peak amplitudes. The color of dots indicates the signiﬁcance of the cross- correlation using a 10s structured reshuﬄing surrogate analysis. Black dots indicate p-value > 0.05, blue dots indicate p-value < 0.05, and red dots indicate p-value < 0.01. C-D: Same as A-B but for analysis of theta power x calcium signal cross-correlations (for raw traces, see Figure 7F). E: Same as in Figure 7G but with time-varying theta power instead of speed.

196

197

Chapter 5: Discussion

1. VIP-LRPs as an intermediate subpopulation of VIP interneurons

Paul et al. (2017) found that 40 gene families mainly involved in cell-cell communication are most distinguishable among interneuron subtypes. Thus, we took advantage of this study by choosing similar functional gene families, including ion channels, receptors, neuromodulation, axonal guidance, CAMs, and myelin. Using patch-scRNA seq, we examined the expression level of several genes within these families in morphologically identified VIP-LRPs in SO. First, we confirmed that VIP-LRPs express the genes that denote their CGE origin, including Npas3 (Neuronal PAS Domain Protein 3, a transcription factor regulating genes involved in neurogenesis), Nfia (Nuclear Factor I A, a transcription factor), Nfib (Nuclear Factor I B, a transcription factor), Nfix (Nuclear Factor I X, a transcription factor), and Htr3 (5-Hydroxytryptamine Receptor 3), whereas MGE marker genes, such as Lhx6 (LIM Homeobox 6, a transcription factor involved in embryogenesis), Sox6 (SRY-Box 6, a transcriptional activator that is required for the normal development of the central nervous system), and Satb1 (Special AT-Rich Sequence Binding Protein 1 , a matrix protein which regulate chromatin structure and gene expression) were poorly expressed. In the hippocampus, several non-PC cell types were found projecting to the subiculum, including Oriens-BIS, trilaminar cells, Subiculum-medial septum double-projecting cells and a subset of enkephalin+ cells (Jinno et al., 2007; Fuentealba et al., 2008). These diverse cell types have different origins: Oriens-BISs appear to derive from MGE (Tricoire et al., 2011), whereas trilaminar cells and enkephalin+ cells are developed from CGE. Together, Subiculum-projecting cells including VIP-LRPs represent a complex group that may exert distinct circuit functions. For instance, these cells are differentially involved in network oscillations. SOM+/mGluR1+ projecting cells and Oriens-trilaminar cells are active during sharp wave ripples, whereas a group SOM-subiculum-projecting cells fired rhythmically with theta but are quiet during ripples (Jinno et al., 2007).

198

Next, we examined the mRNA expression levels of ion channels in VIP-LRPs. In particular, Kcnc1, the gene encoding the delayed rectifier K+ channel Kv3.1 was found in VIP-LRPs. Kcnc1 (Kv3.1) and Kcnc2 (Kv3.2) are highly enriched in PV-BCs (Fuzik et al. 2016; Földy et al., 2016), enabling high-frequency firing of these cells. The maximum firing frequency of VIP-LRPs was not studied. However, the presence of Kcnc1 transcript indicates that the presence of Kv3.1 channel may shape their firings. In addition, VIP-LRPs express mRNAs encoding Kv regulatory proteins, like Dpp10 and Kchip1. Also, these cells express the transcripts of voltage-gated Na+ channel (Nav) subunits Scn1a, Scn1b, and Scn2b. Although the distribution of Nav channels is unknown in VIP+ cells, their somatic and dendritic expression may facilitate the propagation of bAPs and dendritic spikes. Moreover, VGCCs are important in dendritic integration, generating dendritic spikes and regulating neurotransmitter release at the presynaptic terminals. VGCCS are widely expressed in interneurons (Tsien et al., 1988; Bean, 1989). Specifically, Cav2.1 (P/Q type) and Cav3.1 (T type) are universally expressed in almost all interneurons, and are also expressed in VIP-LRPs. Furthermore, the mRNAs of three types of HCN channels (Ncn1-3) were detected in these cells. Accordingly, a voltage sag in response to hyperpolarization current was observed in current-clamp recordings in VIP-LRPs, corresponding to the activation of HCN channels. It has been shown that Ih current in interneurons is mainly mediated by HCN1 and HCN2 channels. It would be interesting to see which subtypes of HCN channels underlie the Ih current recorded in VIP-LRPs.

Glutamate and GABA receptors play fundamental roles in synaptic transmission and plasticity in interneurons. Paul et al. (2017) reported that the expression level of CI and CP-AMPAR were highly variable across neocortical interneurons, with Gria2 (GluR2 subunit) enriched in VIP;CCK groups. Besides, several AMPAR auxiliary subunits also showed cell-type specific expressions. For instance, TARPγ3 and SHISA9 were highly expressed in VIP;CCKs, whereas SHISA7 was enriched in VIP;CR cells. Additionally, Grin2b (GluN2B subunit) was preferentially expressed in VIP;CR cells. On the other hand, VIP-LRPs expressed Gria1 and Gria2 with similar levels, indicating that GluR2 subunits may play a role in synaptic transmission and

199

plasticity in these cells. These cells also expressed SHISA4 and SHISA9, which is similar to VIP;CCK cells in the neocortex. Furthermore, in the CA1 area, 40% of subiculum projecting interneurons express mGluR1α (Jinno et al., 2007). VIP-LRPs expressed Grm1 (mGluR1) and Grm5 (mGluR5), consistent with other subtypes of the long-projecting cell ensemble such as hippocampal-septum projecting and hippocampal-Sub projecting cells. In addition, Grin1 and Grin2b were highly expressed in VIP-LRPs, indicating that GluN2B mediated slow synaptic transmission may shape the dendritic integration and plasticity in these cells.

GABAAR are heteropentamers that consist of three major types of subunits: α,β, and γ. Paul et al. (2017) found that γ2 and γ3 were highly enriched in all six interneuron groups in the neocortex. The expression of other subunits was less distinguishable between VIP+ and other groups. Similarly, VIP-LRPs expressed a variety of these subunits including α1, α2, α3, β1, and β3. This feature suggests that the GABAergic transmission mediated by these cells may display target-specific kinetics due to the differential distribution of GABAAR subunits. Together VIP-LRPs share similar expression levels of excitatory and inhibitory receptors with both VIP; CCK and VIP; CR groups in the neocortex.

In the neocortex, CGE-derived interneurons are characterized by a wider spectrum of neuromodulatory receptors compared to MGE-derived cells, especially for the VIP;CCK group (Paul et al. 2017). One exception is the SOM;NOS1 group, which represents a subtype of the long-projecting GABAergic cells originating from the MGE (Kilduff et al., 2011). Several neuromodulators and hormone receptors were enriched in this group: Chrm1 (Cholinergic Receptor Muscarinic 1), Hcrtr1 (Hypocretin Receptor 1), Oxtr (Oxytocin Receptor), Tacr1 (Tachykinin Receptor 1), Crhbp (Corticotropin Releasing Hormone Binding Protein), and Adra1a (Alpha-1- adrenergic receptors). Most of these receptors function by coupling with G proteins. The expression of these receptors is in line with the speculation that SOM;NOS1 cells are inhibited during wakefulness by a variety of modulators and activated by a different set of modulating substances during sleep and synchronize the EEG across brain areas (Kilduff et al., 2011). Similarly, VIP-LRPs expressed a series of receptors

200

for neuromodulators including Chrna4 (cholinergic receptor nicotinic α4 subunit), Adrb1 (Adrenoceptor β1), Drd1 (Dopamine Receptor D1), Htr1d (5- Hydroxytryptamine Receptor 1D), Cnr1 (Cannabinoid Receptor 1), Oprd1 (Opioid Receptor δ1), Oprl1 (Opioid-Related Nociceptin Receptor 1), Npy1r (Neuropeptide Y Receptor Y1), and Ntsr2 (Neurotensin Receptor 2). Francavilla et al. (2018) showed that VIP-LRPs are more active during the stationary period, rather than locomotion of awake mice, and the recruitment of these cells was not associated with theta nor ripple oscillations. The special behavior of VIP-LRPs indicates that they are inhibited during locomotion presumably by the global modulatory signals that are typically activated during running period, such as acetylcholine (Fu et al., 2014).

In addition, VIP-LRPs express many CAM genes and genes responsible for axonal guidance. Among these genes, Ntng1 has a higher expression level. Netrin G1 is a plasma membrane anchoring protein that binds to a CAM human netrin-G1 ligand (NGL-1). It is involved in axonal growth and laminar distribution of axons in the cortex (Nishimura-Akiyoshi et al., 2007). In particular, the surface binding of NGL-1with Netrin G1 promotes the growth of embryonic thalamic axons (Lin et al., 2003). In the neocortex, Ntng1 is enriched in SOM;NOS1 cells and thus can be considered as a marker for long-projecting cells. The expression of Ntng1 further proves the long- projecting feature of VIP-LRPs.

What is the taxonomic position of VIP-LRPs in major interneuron cell types? Our data showed that these cells expressed high levels of Vip mRNA, indicating that they belong to the VIP+ cell subtype. However, the expression of Cck varied across samples. Besides, they expressed the transcripts of Ntng1 and Crhm2 (M2R), which are markers for the long-projecting cells (Jinno, 2009). We explored this issue initially by comparing enriched genes in neocortical VIP;CR and VIP;CCK groups (Paul et al., 2017) with genes detected in VIP-LRPs. Intriguingly, VIP-LRPs shared ~50% of the genes with VIP;CR and ~50% of the genes with VIP;CCK. Thus, we conclude that this cell type represents an intermediate subtype between VIP;CR and VIP;CCK groups, but also share genes with long-projecting cells.

201

2. Molecular profiling of VIP+ cells in CA1 SO

As discussed in the introduction 2.3.2, subicular-projecting interneurons are a heterogeneous group of cells that express various molecular markers. Previous studies showed that half of the hippocampal-subicular projecting interneurons in SO express SOM, mGluR1α, and M2R. A smaller proportion of them express CB and NPY. In general, two molecular groups were found in SO: SOM+/mGluR1α+/M2R- and M2R+/SOM-/mGluR1α-. However, no VIP immunoreactivity was detected in these cells (Jinno et al., 2007). To identify the molecular markers expressed in VIP cells in SO, we tested several commonly used markers in SO VIP cells in VIP-GFP and VIP-tdTomato mouse lines, including M2R, CCK, CB, nNOS, CR, and SOM. The results showed that GFP+ cells were negative for SOM and nNOS indicating that VIP-LRPs are not a subpopulation of SOM+ nor nNOS cells. Moreover, half of GFP+ cells in SO express M2R, suggesting that they may belong to the M2R+/SOM- long-projecting cell group. Moreover, a smaller proportion of GFP+ cells express CCK, CB, and CR, pointing to the distribution of IS3 cells and CCK-BCs in SO. In contrast, M2R+ cells accounted for a much smaller proportion in VIP-tdTomato+ cells (6.7%). Such a difference between the two mouse lines suggests altered gene expression or cell type differentiation during development in the Cre mouse line.

3. Input-specific synaptic transmission in IS3 cells

In this study, we examined the basic properties of excitatory synaptic transmissions on the apical dendrites of IS3 cells, an interneuron-specific cell derived from CGE. By comparing proximal (SC) and distal (TA) inputs, we found that when the cells were recorded at -70 mV, EPSCs evoked at SC synapses had larger amplitude, faster rise and decay time compared to TA synapses. The pathway specific features were also reported in CA1 PCs: when recorded at -65 or +40mV, SC-PC EPSCs displayed larger amplitude and faster kinetics than TA-PC EPSCs (Otmakhova et al., 2002). However, we observed two components of TA-IS3 EPSCs at -70mV and the depolarized voltage levels (-50~+70 mV). The smaller current amplitude and slower rise time of TA-EPSCs recorded at the somatic level of IS3 cells can be explained

202

by passive dendritic filtering effect, whereas the distribution of glutamate receptors may also account for the different rise and decay kinetics at SC and TA synapses. At -70 mV, which is close to the resting potential of IS3s, AMPAR mediates the fast synaptic transmission at the proximal site, whereas distal dendrites may not be clamped at the same level as the soma due to the presence of VGICs along the dendritic tree (Bar-Yehuda and Korngreen, 2008). Thus, the activation of NMDAR may contribute to the slow component of distal synaptic events. Nevertheless, the I- V curves recorded at TA and SC synapses unveiled the slow component at TA-IS3 responses at different levels of Vm, which was absent at SC-IS3 responses. This finding indicates that TA input may have a larger impact on AP firing when the cell is in active states. That is, enhanced network activities modulating TA transmission during specific behavior states may favor the EC-IS3 information flow and subsequently enhance the recruitment of IS3 cells (Fu et al., 2014).

Next, pharmacological experiments showed that TA-IS3 synaptic transmission was mediated by AMPARs and NMDARs, whereas the I-V curve recorded at SC-IS3 synapses indicates the more prominent contribution of AMPARs. Again, this observation resembles the increased NMDA/AMPA ratio in the apical tuft area of CA1 PCs compared to proximal dendrites (Otmakhova et al., 2002; Bittner et al., 2012), indicating that similar mechanism is used by IS3 to amplify distal signals. However, it is not the case in other interneurons. In contrast to IS3s, profound NMDA component was evoked at SC-NGFC synapses recorded at +40 mV, which is comparable with AMPA component recorded at -70mV (Price et al., 2005). Also, a large NMDA/AMPA ratio was found at SC synapses in other CGE-derived interneurons (Matta et al. 2013). Such a differential distribution of NMDARs indicates cell-type specific features at SC synapses. In addition, the linear relationship of I-V curve of the AMPA component recorded in IS3 cells implies the presence of CI- AMPAR at both pathways, which is in line with the lack of CP-AMPAR in other CGE- derived cells (Matta et al. 2013). Together, these data indicate that the distribution of NMDARs and AMPARs in interneurons depends on both cell-type and inputs. Further study is required to understand the spatial distribution of NMDARs and AMPARs along the apical dendrites of IS3 cells.

203

4. Short-term facilitation at excitatory synapses to IS3 cells

We studied the short-term plasticity of excitatory synaptic inputs at TA and SC-IS3 synapses by delivering repetitive electrical stimulation at the distal and proximal dendrites respectively. The results showed that TA and SC-IS3 synapses showed consistent facilitation regardless of the stimulation location and pulse numbers. The degree of potentiation showed frequency dependence: higher frequency induced larger potentiation at both pathways. This unique feature observed in IS3s indicates a very low initial release probability of vesicles at presynaptic terminals, which can be explained by large Ca2+ buffering reservoirs. The initial Ca2+ influx may lead to the binding of Ca2+ with internal buffering proteins, and facilitate the subsequent vesicle release. For instance, high concentrations of Ca2+-binding proteins at presynaptic terminals were discovered in the neocortex (Blatow et al., 2003). Moreover, the consistent STF in IS3s is in contrast to other hippocampal interneurons, whose direction of STP displays the input and frequency dependence (see introduction 4.1). It means that IS3 cells are preferentially recruited by high frequency burst firing. In particular, the magnitude of LTF at TA-IS3 synapses is significantly higher than that of SC-IS3 synapses, indicating that in addition to the presynaptic facilitation, the repetitive stimulation provided the depolarization necessary for the activation of postsynaptic NMDARs. In turn, the slow kinetics of NMDAR facilitated the temporal integration of synaptic inputs at distal dendrites of IS3 cells. Thus, the STF at distal dendrites may serve as a candidate mechanism in facilitating the recruitment of IS3 cells by EC inputs.

What are the patterns of information transmitted from EC? In vivo studies showed that the firing pattern of principal cells in MEC displays cell-type specific properties. Layer II stellate cells and layer III neurons had no burst firing, whereas layer II PCs fired in bursts (Laut Ebbesen et al., 2016). Moreover, the burst firing of layer II PCs was locked with theta rhythm and associated with grid-like firing pattern in spatial navigation (Tang et al., 2014). Intriguingly, layer II PCs selectively targeted interneurons on the SR/SLM border of the CA1 (Kitamura et al., 2014), indicating that the feedforward inhibition mediated by these cells is important in spatial learning.

204

Thus, it would be interesting to see if distal dendrites of IS3 cells at the SR/SLM border are innervated by layer II PCs as well. In addition to EC projection, other long- range projections also terminate in SLM, like excitatory inputs from the median raphe nucleus and the nucleus reuniens of the thalamus. In the median raphe nucleus, a subtype of VGLUT3+ neurons fired with high frequency and phase-coupled with hippocampal theta, indicating their potential role in regulating hippocampal oscillations (Viana Di Prisco et al., 2002; Domonkos et al., 2016). It is unclear whether neurons in nucleus reuniens fire in bursts. However, they displayed burst firing in vitro in response to depolarizing current injection (Walsh et al., 2017). In summary, many extrinsic inputs innervating the SLM layer of CA1 convey information with burst firing. The synaptic properties of IS3s are well suited for processing these inputs.

Apart from distal inputs, IS3s also received local CA3 projections via Schaffer collateral. Current clamp experiments showed that SC and TA synapses were equally efficient in inducing APs in IS3s. Burst firing of CA3 PCs has been observed in both awake and anesthetized rodents (Tropp Sneider et al., 2006; Kowalski et al., 2016). The relative weight of SC and TA synapses in activating IS3 cells may depend on the activity-induced synaptic plasticity and behavior states.

5. Spatial summation of excitatory inputs in IS3s

Using glutamate uncaging, we showed that a similar length of scanning areas at the proximal and distal dendrites of IS3s was required for inducing APs at the soma. An early study showed that the excitatory synaptic density on the distal and proximal dendrites of CR+ cells were 0.75/µm and 0.85/µm respectively (Gulyás et al., 1999). That is, the simultaneous activation of 4-5 synapses at either proximal or distal dendrites is sufficient to trigger somatic firing of IS3s. This feature indicates that compensating mechanisms at distal dendrites are capable of counteracting passive filtering effects. First, a morphological study showed that distal dendrites of CR+ cells are small in diameter (0.4 µm) compared to proximal dendrites (0.6-1.2 µm, Gulyás et al., 1999). Small dendrites have large input impedance, which results in larger

205

currents at distal sites. Second, the cooperative activation of nearby synapses may induce local NMDA spikes. The plateau-like slow kinetics of NMDA spikes efficiently integrate dendritic events and propagate into the soma with less attenuation. Third, it is unknown whether a gradient of VGICs are distributed in apical dendrites of IS3s. But they may positively contribute to the amplification of distal synaptic events as observed in PCs.

The dendritic properties that underlie synaptic integration in IS3 cells are largely unknown. Using computational simulation, Guet-McCreight et al. (2016) explored the distribution of VGICs on IS3 dendrites and their impacts on AP generation. In the case that VGICs are restricted to the soma and proximal dendrites, the model fitted the experimental data better than others. Moreover, a single presynaptic spike at distal dendrites was unable to elicit the somatic firing of IS3 cells in the model. Thus, temporally or spatially clustered inputs at distal dendrites are necessary for AP induction. In addition, the number of branching points and dendritic surface area also have a large influence on signal propagation along the dendritic trees. Nevertheless, limited types of Na+ and K+ channels were considered in this study, and the role of NMDARs and other VGICs were not examined in their models.

6. Comparison of IS3s with PV-BCs

Can we predict the in vivo activation of IS3 cells from the in vitro properties mentioned above? In order to do so, let’s compare IS3 with PV-BCs, which are one of the most characterized interneuron subtypes. When compared to PV-BCs, IS3s exhibit several distinct properties. First, the input resistance of PV-BCs at the somatic level is 5 times smaller than that of IS3s. This feature weakens the influence of trivial input on the somatic firing in PV-BC, whereas promoting cell recruitment in IS3s by moderate presynaptic activities. Second, the dendritic thickness, dendritic surface area and dendritic length of PV-BCs were larger than those of CR+ cells (Gulyás et al., 1999). These morphological features allow PV-BCs to integrate large numbers of synaptic inputs and enable quick signal propagation along the dendrites. On the other hand, the firing of IS3s are affected by a smaller number of inputs. Third,

206

bAP contributes poorly to the synaptic integration at distal dendrites of both cell types. Nevertheless, localized VGCCs and NMDARs on PV-BC dendrites enabled the generation of the dendritic spike (Chiovini et al., 2014). Similarly, the presence of NMDAR at distal dendrites counteract the dendritic filtering effect in IS3 cells and might be responsible for triggering local supralinear events. Fourth, the excitatory inputs on PV-BCs show STD, whereas STF was constantly observed in IS3s. The STD in PV-BCs improves the precision in detecting simultaneously arriving synaptic events. In contrast, IS3 cells are more likely to be activated by high-frequency presynaptic spike trains amplified by STF from different inputs with less temporal precision. Last but not least, PV-BCs were able to fire at high frequencies without adaption in response to depolarizing current injection (650Hz; McCormick et al., 1985), whereas the maximum firing frequency of IS3 cells was relatively low (171 Hz, Tyan et al., 2014) and was adapted strongly at depolarized levels. The fast- spiking character endows PV-BCs the ability to coordinate fast network oscillations such as sharp wave ripples (Stark et al., 2014). On the contrary, the irregular and adaptive firing patterns of IS3s suggest that their firings are unlikely to be phase- locked with high-frequency network oscillations. In conclusion, the morphological and electrophysiological properties of PV-BCs are well designed for fast input-output conversion, and therefore are capable of transferring properly timed inputs and synchronizing network activity of PCs. On the other hand, IS3s are broadly tuned by their inputs due to their slow integration properties. Their activities in network oscillations, if any, are less precise in timing compared to PV-BCs.

PV-BC IS3 -63.6 ± 0.9 (Camiré and Topolnik, Vm (mV) -74.2 ± 0.7 (Tyan et al., 2014) 2014) Rm (MΩ) 82.6 ± 11.8 (Nörenberg et al., 2010) 496.9 ± 28.6 (Tyan et al., 2014) ~40% at 150 µm (Guet- bAP ~15% at 250 µm (Hu et al., 2010) McCreight et al., 2016) Depression/facilitation (Losonczy et STP facilitation al., 2002)

207

transient gNa :Dend < 70 µm: Dend < 100 µm: 22 ± 5 pS µm-2 0.07 S cm-2 ; persistant Dendritic Dend > 100 µm: 18 ± 6 pS µm-2 gNa :Dend < 70 µm: 0.000075 gNa density S cm-2 (predicted, Guet- (Hu et al., 2010) McCreight et al. 2016) Dend < 100 µm: 81 ± 13 pS µm-2 Dend < 70 µm: 0.07 S cm-2 Dendritic gK Dend > 100 µm: 93 ± 22 pS µm-2 (predicted, Guet-McCreight et density al. 2016) (Hu et al., 2010) Irregular spiking (Tyan et al., Firing pattern Fast spiking (McCormick et al., 1985) 2014) CP-AMPAR and NMDAR (Jones and iGluRs CI-AMPAR and NMDAR Bühl, 1993; Geiger et al., 1995)

Table 1 Electrophysiological properties of PV-BC and IS3

Network oscillations are synchronized neuronal activities that travel across brain areas. Coordinated neuronal firing enhances cognitive functions such as spatial navigation and memory consolidation. Indeed, In vivo studies have shown that PV- BCs and other fast-spiking cells in the hippocampus vitally participate in network oscillations at different frequencies, such as theta (4-10 Hz), gamma (30-80 Hz) and ripple (100-200 Hz) oscillations (Klausberger and Somogyi, 2008). While regular or adapting firing cells, like CCK-BCs and O-LMs were active during theta phase, they were silent during ripples. One possible scenario would be strong phasic inhibition of O-LMs and CCK-BCs provided by PV-BCs or BISs during ripples (Karson et al., 2009). PV-BCs are also interconnected by inhibitory synapses (Ali et al., 1999). However, given that their activities are highly synchronized by gap junctions, the excitatory current flow during ripples may exceed inhibition and generate spikes in the majority of this cell ensemble. Additionally, the low firing rate of CCK-BCs and O-LMs during ripple may arise from their intrinsic properties: they are unable to fire with high temporal precision. Because they have strong facilitating synapses made by local PCs and large membrane time constant (Silberberg and Markram, 2007),

208

these features enable synaptic events to summate in a long time window.. Although the source of inhibition received by IS3s is unclear, these feature are in line with the fact that IS3s are involved in theta oscillations, but silent during ripples.

7. The potential role of IS3 cells as network gate-keepers in behavior

What are the potential roles of IS3s in different behavior states? It has been suggested that IS3s are responsible for the disinhibition of the distal dendrites of CA1 PCs. Experimental data have shown that O-LM cells in SO are the primary targets of IS3s (Chamberland et al., 2010; Tyan et al., 2014). In turn, O-LMs inhibit the distal dendrites of PCs. Hence, the activation of IS3s releases the distal excitation from EC in PCs. A similar disinhibitory network in the neocortex is mediated by VIP+/CR+ interneurons that control dendrite-targeting SOM+ cells, presumably Martinotti cells. Accumulating data have unveiled that these VIP+ cells are activated during locomotion, whisking, reward, and aversive stimuli (Lee et al., 2013; Fu et al., 2014; Pinto and Dan, 2015). These behavior states associate the activation of VIP+ cells with global modulatory signals such as cholinergic, serotoninergic and dopaminergic projections. Morphological evidence has proven that IS3 cells in the hippocampus also receive these modulatory inputs (Papp et al., 1999). Hence, they are likely to be recruited during those behaviors that trigger the global release of neuromodulators.

In particular, the activation of VIP+ cells in the neocortex is closely related to network states. Karnani et al. (2016) found that the activation of VIP+ cells and nearby PCs are strongly correlated during the awake stationary period and visual stimuli. The strength of correlation between VIP+ cells and PCs decreased with distance, up to 120 µm from the soma of a single VIP+ cell. However, locomotion abolished the distance dependence of VIP+cell-PC coupling, probably due to the massive recruitment of VIP+ cells by the global modulatory signals. Accordingly, Jackson et al. (2016) reported that the correlation between cortical VIP+ cells and network activity was stronger than the correlation between VIP+ cells and locomotion. These findings indicate that VIP+ cells in visual cortex are largely driven by local excitatory

209

inputs at the stationary state. During locomotion, local and long-range inputs converge onto VIP+ cells. In return, the activation of VIP+ cell enhances the overall excitation of V1 PCs via disinhibitory circuits. Computational simulation estimated that during network oscillations, excitatory and inhibitory inputs with different spike timings converge onto the apical dendrites of IS3s and push their firings toward the rising phase of theta. Moreover, our simulation data shows that burst firing of CA3 PCs drives the firing of IS3s during sharp wave-ripples. Using two-photon in vivo Ca2+ imaging, my colleagues found that the activation of IS3s in free moving animals is moderately coupled with running speed and theta power: some cells showed stronger coupling while the correlation was weak in others. This result is similar to the VIP cells in the neocortex described above, indicating that IS3 cells may receive cholinergic input during locomotion. Furthermore, the spike timing of IS3 cells during theta oscillations also displays a high level of variability, but aggregates near the rising phase of theta. On the contrary, the activation of IS3 cells is not coupled with sharp wave ripples. These results can be explained by the properties studied in vitro, such as the facilitating synapses, intrinsic membrane properties, and putative strong inhibition of IS3s during ripples as discussed in chapter 6. Together, the in vivo data suggest that although low in cell numbers, IS3 cells are a heterogeneous group that display different degrees of coupling with behavioral states and theta power/phases. Such heterogeneity may arise from their pre and postsynaptic partners. Thus, further study is required to understand their cell-type specific connectivity.

Analogously, I propose the function of IS3s as a network gate-keeper. In the hippocampus, the dendritic tree of IS3s arborizes intensively in SLM, indicating that they are largely activated by EC projections, although CA3 input may also contribute to their recruitment in specific behavior states. Given that many IS3s possess bipolar dendrites, they may receive axonal innervation from local CA1 PCs in SO/Alveus. Hence, they are able to sense the level of excitation from CA1 PCs as well as EC PCs. In addition, modulatory inputs may strongly influence the firing rate of IS3 cells, as shown in their neocortical analogs. On the other hand, the axons of IS3s widely spread along the SO and alveus border, where they are able to coordinate the activities of large numbers of interneurons within SO of CA1, including O-LMs, BISs,

210

BCs and O-O cells (Tyan et al., 2014). Subsequently, the inhibition of these cell types by IS3s facilitate excitation arriving at different dendritic domains of CA1 PCs, but mostly biased to distal dendrites, since O-LMs are the main targets of IS3s (Figure 16). In conclusion, IS3s enhance the network excitability of CA1 by targeting SO interneurons. Moreover, they may promote the information flow from the EC to CA1 in specific behavior states, such as spatial navigation (Brun et al., 2002) and olfactory associative learning (Igarashi et al., 2014; Li et al., 2017).

TA pathway

St. LM

PC SC pathway

IS3 St. RAD

St. PYR

O-LM St. O/A

VIP-LRP

Figure 16 Simplified disinhibitory network in the hippocampal CA1 area mediated by IS3 and VIP-LRPs. O-LMs are inhibited by IS3 and VIP-LRPs. O-LMs in turn inhibit the distal dendrites of PCs. Dashed lines indicate the borders of different layers. Thick lines depict the dendrites of different cell types. Thick arrows indicate the excitatory inputs converging onto IS3s. Rad arrows indicate the targets of axons originate from IS3, VIP-LRP and O-LM. The local CA1 PC input and long-range projection of VIP-LRP in the subiculum are not indicated.

211

8. Limitations and perspectives

We have discussed the local (CA3) and long-range (EC) excitatory inputs targeting IS3 cells and their impacts in cell recruitment. However, inhibition may also have a strong influence in controlling the excitability of IS3 cells. An in vivo intracellular study showed that the responses in VIP cells in the auditory cortex are dominated by broadly tuned inhibitions during auditory perception (Kato et al., 2017). In the hippocampal CA1, CR+ and VIP+ cells often form dendro-dendritic and axon- dendritic connections on each other, indicating that they inhibit each other. Computational simulation implies that mutual inhibition network is able to operate complex behavior tasks such as multiple-choice selection (Koyama and Pujala, 2018) In addition; the CA1 area is innervated by long-range inhibitory projections from medial septum and MEC. The disinhibition provided by septal GABAergic neurons is responsible for modulating hippocampal oscillations (Yoder and Pang, 2005). On the other hand, the axons of PV+ cells in MEC selectively target hippocampal interneurons in the SLM, thus releasing the feed-forward inhibition to the dendrites of CA1 PCs (Melzer et al., 2012). It is unknown whether these long-range inhibitory inputs target IS3 cells. Thus, further study is required to elucidate the inhibitory control of IS3 cells.

In this study, we identified several molecular markers using patch-seq from 7 single cell samples. The sample size in our transcriptomic study was limited due to the difficulty in obtaining both morphological and transcriptomic data from the same cells. According to a study, a minimum sample size of 6 single cells is required for gene differential expression analysis (Schurch et al., 2016), although 30 sample cells can be used to distinguish two cell classes with higher confidence (Cadwell et al., 2016). A larger sample size may generate results with higher statistical power and reduces the technical noise. Thus, more sample will be required for advanced analysis.

When discussing cell-type specific markers, we assume that the mRNA expression levels in a single cell correspond to their protein expression levels. However, in most species studied, the correlation between the mRNA and protein levels is fairly weak.

212

In general, the producing rate of mRNAs is much lower than that of proteins. Also, the half-life of transcripts is shorter than proteins (2.6-7h vs 46h; Schwanhäusser et al., 2011), which means that the copy number of proteins at a given time is higher than that of mRNAs. Using RNA-seq and proteomic approaches, Bauernfeind and Babbitt (2017) studied the predictive nature of transcripts on protein expression levels in human and chimpanzee brain tissues. For most transcripts, the predictive values were within one standard deviation of randomly correlated mRNA-protein pairs. However, those transcripts encoding synaptic proteins, kinases, and phosphatases had higher predictive values than others. Hence, caution should be taken when discussing the results of RNA-seq in a gene-specific manner unless the protein expression level is also measured.

213

Chapter 6: Conclusion

Compared to rodents, the human cortex contains a larger number of interneurons, indicating a crucial role of inhibition in regulating neocortical circuits. What are the roles of VIP cells in human cortex? VIP cells are barely studied in human due to the inaccessibility of brain tissues. Poorthuis et al. (2018) examined the distribution of VIP mRNAs in the anterior medial temporal cortex of adult patients. In consensus with VIP-Cre mouse line, VIP mRNA expressing cells in humans accounted for a small proportion of layer I interneurons, whereas the density of cells peaked in upper layer II and declined gradually in deeper layers. In addition, acetylcholine -induced nicotinic currents in most layer I interneurons in acute slices, which were confirmed by the presence of α7 and β2-containing nicotinic receptors. However, serotonin currents were not observed in human layer I interneurons, consistent with a smaller proportion of HT3AR-expressing VIP cells in this layer (Poorthuis et al., 2018).

Moreover, the development of VIP+ and CR+ cells has been studied in fetal brain tissues. Using scRNA-seq and pseudo-time analysis, Zhong et al. (2018) found that VIP+ cells appeared at late stages of human prefrontal cortex development, after the emergence of CR+, and SOM+ cells. Similarly, several immunolabeling studies confirmed the dual origin of CR+ cells in the human brain (Jakovcevski et al., 2011) In rodents, most interneurons originate in GE, whereas the subventricular zone (SVZ) provides an additional proliferation area for late-born neurons in primates. During early developmental stages (gestational week 6, gw 6), the highest density of CR+ cells appeared in GE and showed a gradient toward the neocortex (Zecevic et al., 2010). This suggests that CR+ cells were born in GE and migrated tangentially to the neocortex. At gw 15-20, the density of CR+ cells in GE declined, whereas the density in ventricular zone /SVZ increased six times, indicating a second proliferation zone for CR+ cells. However, CR+ cells in human neocortex and hippocampus target both interneuron and PCs (Urban et al., 2002; del Río and DeFelipe, 1997). Thus, functional study will be required to further characterize these cell types.

214

GABAergic interneurons delicately control the balance of excitation and inhibition in the circuitry, as well as the synchronicity of network oscillations. This is in addition to the fact that the dysfunction of interneurons may cause a series of neurological and psychiatric disorders. During the postnatal development, the proliferation and migration of interneurons are regulated by various factors. For instance, the Erb-B2 Receptor Tyrosine Kinase 4 (ErbB4) is responsible for interneuron migration and synapse development. The selective deletion of ErbB4 gene in VIP cells in mouse primary visual cortex led to an increased firing rate and reduced synchronicity of excitatory neurons. In addition, the state dependent recruitment of VIP interneurons decreased in mutant mice, resulted in a general reduction of network activity in behavioral up-states (Batista-Brito et al., 2017).

What can be the role of VIP+ interneurons in brain pathologies? Schizophrenia is a severe mental disorder occurring in early adulthood. The etiology of this disorder has been partially associated with altered gene copies or gene mutations. For example, gene copy number variants (CNVs) is a common phenomenon in the genome. It is caused by microdeletion or duplication of small chromosomal fragments. However, rare CNVs occurring at 22q11.2 and 16p11.2 chromosomal regions are highly correlated with the onset of autism and schizophrenia. Vacic et al. (2011) reported that microduplications at the 7q36.3 region, upstream of the VIP receptor gene Vipr2 were found in a small proportion of Schizophrenia patients, but the proportion was significantly higher than in normal subjects. This finding indicates that VIP mediated signaling may play a role in schizophrenia pathology. Moreover, a reduction in VIP mRNA was detected in the dorsolateral prefrontal cortex of schizophrenic and bipolar disorder patients (Fung et al., 2014).

Autism spectrum disorder (ASD) is a series of developmental disorders that happens at early developmental stages. The onset of ASD is often linked with gene mutations and exposure to environmental factors. Cunniff and Sohal (2017) examined the electrophysiological properties of VIP interneurons in the prefrontal cortex of two autistic mouse models. VIP cells showed increased excitability with smaller degrees

215

and longer halfwidth of APs in response to acetylcholine application compared to control mice.

Epilepsy is a neurological disorder that is primarily associated with the hyperexcitability of hippocampal circuits. It is well documented that CR+ cells in the human hippocampus are sensitive to epilepsy. Both the cell number and the dendritic morphology of CR+ cells undergo a dramatic change in the DG of patients with temporal lobe epilepsy compared with normal subjects (Maglóczky et al., 2000; Tóth et al., 2010). In particular, David and Topolnik (2017) examined the properties of VIP/CR interneurons in the CA1 area of an epilepsy mouse model. The number of VIP/CR boutons significantly decreased in SO after seizure induction. As a result, all targets in SO exhibited reduced frequency of spontaneous IPSCs and smaller amplitudes of evoked IPSCs. Together, these data indicate that VIP/CR interneurons play important roles in normal brain as wells as in the pathology of various diseases. A better understanding of these cell types may provide significant insights into the development of new therapies and drugs.

216

Bibliography

(2008). Learning and Memory: A Comprehensive Reference (Academic Press).

Acsady, L., Arabadzisz, D., and Freund, T.F. (1996a). Correlated morphological and neurochemical features identify different subsets of vasoactive intestinal polypeptide-immunoreactive interneurons in rat hippocampus. Neuroscience 73, 299-315.

Acsady, L., Gorcs, T.J., and Freund, T.F. (1996b). Different populations of vasoactive intestinal polypeptide-immunoreactive interneurons are specialized to control pyramidal cells or interneurons in the hippocampus. Neuroscience 73, 317-334.

Ali, A.B., Bannister, A.P., and Thomson, A.M. (1999). IPSPs elicited in CA1 pyramidal cells by putative basket cells in slices of adult rat hippocampus. Eur J Neurosci 11, 1741-1753.

Amaral, D.G. (1978). A Golgi study of cell types in the hilar region of the hippocampus in the rat. J Comp Neurol 182, 851-914.

Andersen, P., Eccles, J., and Voorhoeve, P.E. (1963). Inhibitory Synapses on Somas of Purkinje Cells in the Cerebellum. Nature 199, 655-656.

Araya, R., Eisenthal, K.B., and Yuste, R. (2006). Dendritic spines linearize the summation of excitatory potentials. Proc Natl Acad Sci U S A 103, 18799-18804.

Arrigoni, E., and Greene, R.W. (2004). Schaffer collateral and perforant path inputs activate different subtypes of NMDA receptors on the same CA1 pyramidal cell. Br J Pharmacol 142, 317-322.

Aten, S., Hansen, K.F., Hoyt, K.R., and Obrietan, K. (2016). The miR-132/212 locus: a complex regulator of neuronal plasticity, gene expression and cognition. RNA Dis 3.Bar-Yehuda, D., and Korngreen, A. (2008). Space-clamp problems when voltage clamping neurons expressing voltage- gated conductances. J Neurophysiol 99, 1127-1136.

Batista-Brito, R., Vinck, M., Ferguson, K.A., Chang, J.T., Laubender, D., Lur, G., Mossner, J.M., Hernandez, V.G., Ramakrishnan, C., Deisseroth, K., et al. (2017). Developmental Dysfunction of VIP Interneurons Impairs Cortical Circuits. Neuron 95, 884-895 e889.

Bauernfeind, A.L., and Babbitt, C.C. (2017). The predictive nature of transcript expression levels on protein expression in adult human brain. BMC Genomics 18, 322.

Bean, B.P. (1989). Classes of calcium channels in vertebrate cells. Annu Rev Physiol 51, 367-384.

Bekkers, J.M. (2000). Distribution and activation of voltage-gated potassium channels in cell-attached and outside-out patches from large layer 5 cortical pyramidal neurons of the rat. J Physiol 525 Pt 3, 611-620.

Bereshpolova, Y., Amitai, Y., Gusev, A.G., Stoelzel, C.R., and Swadlow, H.A. (2007). Dendritic backpropagation and the state of the awake neocortex. J Neurosci 27, 9392-9399.

Berger, T., Larkum, M.E., and Luscher, H.R. (2001). High I(h) channel density in the distal apical dendrite of layer V pyramidal cells increases bidirectional attenuation of EPSPs. J Neurophysiol 85, 855-868.

217

Bezaire, M.J., and Soltesz, I. (2013). Quantitative assessment of CA1 local circuits: knowledge base for interneuron-pyramidal cell connectivity. Hippocampus 23, 751-785.

Bittner, K.C., Andrasfalvy, B.K., and Magee, J.C. (2012). Ion channel gradients in the apical tuft region of CA1 pyramidal neurons. PLoS One 7, e46652.

Blatow, M., Caputi, A., Burnashev, N., Monyer, H., and Rozov, A. (2003). Ca2+ buffer saturation underlies paired pulse facilitation in calbindin-D28k-containing terminals. Neuron 38, 79-88.

Bowie, D., and Mayer, M.L. (1995). Inward rectification of both AMPA and kainate subtype glutamate receptors generated by polyamine-mediated ion channel block. Neuron 15, 453-462.

Bozdagi, O., Valcin, M., Poskanzer, K., Tanaka, H., and Benson, D. L. (2004). Temporally distinct demands for classic cadherins in synapse formation and maturation. Mol. Cell. Neurosci. 27, 509– 521.

Brun, V.H., Otnass, M.K., Molden, S., Steffenach, H.A., Witter, M.P., Moser, M.B., and Moser, E.I. (2002). Place cells and place recognition maintained by direct entorhinal-hippocampal circuitry. Science 296, 2243-2246.

Buchanan, K.A., Blackman, A.V., Moreau, A.W., Elgar, D., Costa, R.P., Lalanne, T., Tudor Jones, A.A., Oyrer, J., and Sjostrom, P.J. (2012). Target-specific expression of presynaptic NMDA receptors in neocortical microcircuits. Neuron 75, 451-466.

Burwell, R.D., and Amaral, D.G. (1998). Perirhinal and postrhinal cortices of the rat: interconnectivity and connections with the entorhinal cortex. J Comp Neurol 391, 293-321.

Butt, S.J., Fuccillo, M., Nery, S., Noctor, S., Kriegstein, A., Corbin, J.G., and Fishell, G. (2005). The temporal and spatial origins of cortical interneurons predict their physiological subtype. Neuron 48, 591-604.

Cadwell, C.R., Palasantza, A., Jiang, X., Berens, P., Deng, Q., Yilmaz, M., Reimer, J., Shen, S., Bethge, M., Tolias, K.F., et al. (2016). Electrophysiological, transcriptomic and morphologic profiling of single neurons using Patch-seq. Nat Biotechnol 34, 199-203.

Camire, O., and Topolnik, L. (2014). Dendritic calcium nonlinearities switch the direction of synaptic plasticity in fast-spiking interneurons. J Neurosci 34, 3864-3877.

Chamberland, S., Salesse, C., Topolnik, D., and Topolnik, L. (2010). Synapse-specific inhibitory control of hippocampal feedback inhibitory circuit. Front Cell Neurosci 4, 130.

Chiovini, B., Turi, G.F., Katona, G., Kaszas, A., Palfi, D., Maak, P., Szalay, G., Szabo, M.F., Szabo, G., Szadai, Z., et al. (2014). Dendritic spikes induce ripples in parvalbumin interneurons during hippocampal sharp waves. Neuron 82, 908-924.

Chittajallu, R., Craig, M.T., McFarland, A., Yuan, X., Gerfen, S., Tricoire, L., Erkkila, B., Barron, S.C., Lopez, C.M., Liang, B.J., et al. (2013). Dual origins of functionally distinct O-LM interneurons revealed by differential 5-HT(3A)R expression. Nat Neurosci 16, 1598-1607.

Close, J., Xu, H., De Marco Garcia, N., Batista-Brito, R., Rossignol, E., Rudy, B., and Fishell, G. (2012). Satb1 is an activity-modulated transcription factor required for the terminal differentiation and connectivity of medial ganglionic eminence-derived cortical interneurons. J Neurosci 32, 17690- 17705.

218

Constantinidis, C., Williams, G.V., and Goldman-Rakic, P.S. (2002). A role for inhibition in shaping the temporal flow of information in prefrontal cortex. Nat Neurosci 5, 175-180.

Croce, A., Pelletier, J.G., Tartas, M., and Lacaille, J.C. (2010). Afferent-specific properties of interneuron synapses underlie selective long-term regulation of feedback inhibitory circuits in CA1 hippocampus. J Physiol 588, 2091-2107.

Crow, M., Paul, A., Ballouz, S., Huang, Z.J., and Gillis, J. (2018). Characterizing the replicability of cell types defined by single cell RNA-sequencing data using MetaNeighbor. Nat Commun 9, 884.

Cruikshank, S.J., Hopperstad, M., Younger, M., Connors, B.W., Spray, D.C., and Srinivas, M. (2004). Potent block of Cx36 and Cx50 gap junction channels by mefloquine. Proc Natl Acad Sci U S A 101, 12364-12369.

Cunniff, M., and Sohal, V. (2017). 336. VIP Interneuron Cholinergic Responses are Altered in Multiple Models of Autism. Biol Psychiatry 81, S138.

David, L.S., and Topolnik, L. (2017). Target-specific alterations in the VIP inhibitory drive to hippocampal GABAergic cells after status epilepticus. Exp Neurol 292, 102-112.

Deans, M.R., Gibson, J.R., Sellitto, C., Connors, B.W., and Paul, D.L. (2001). Synchronous activity of inhibitory networks in neocortex requires electrical synapses containing connexin36. Neuron 31, 477-485. del Rio, M.R., and DeFelipe, J. (1997). Synaptic connections of calretinin-immunoreactive neurons in the human neocortex. J Neurosci 17, 5143-5154.

Deng, W., Aimone, J.B., and Gage, F.H. (2010). New neurons and new memories: how does adult hippocampal neurogenesis affect learning and memory? Nat Rev Neurosci 11, 339-350.

Dolorfo, C.L., and Amaral, D.G. (1998). Entorhinal cortex of the rat: organization of intrinsic connections. J Comp Neurol 398, 49-82.

Domonkos, A., Nikitidou Ledri, L., Laszlovszky, T., Cserep, C., Borhegyi, Z., Papp, E., Nyiri, G., Freund, T.F., and Varga, V. (2016). Divergent in vivo activity of non-serotonergic and serotonergic VGluT3-neurones in the median raphe region. J Physiol 594, 3775-3790.

Dudok, B., Barna, L., Ledri, M., Szabo, S.I., Szabadits, E., Pinter, B., Woodhams, S.G., Henstridge, C.M., Balla, G.Y., Nyilas, R., et al. (2015). Cell-specific STORM super-resolution imaging reveals nanoscale organization of cannabinoid signaling. Nat Neurosci 18, 75-86.

Ebbesen, C.L., Reifenstein, E.T., Tang, Q., Burgalossi, A., Ray, S., Schreiber, S., Kempter, R., and Brecht, M. (2016). Cell Type-Specific Differences in Spike Timing and Spike Shape in the Rat Parasubiculum and Superficial Medial Entorhinal Cortex. Cell Rep 16, 1005-1015.

Elfant, D., Pal, B.Z., Emptage, N., and Capogna, M. (2008). Specific inhibitory synapses shift the balance from feedforward to feedback inhibition of hippocampal CA1 pyramidal cells. Eur J Neurosci 27, 104-113.

219

Empson, R.M., and Heinemann, U. (1995). The perforant path projection to hippocampal area CA1 in the rat hippocampal-entorhinal cortex combined slice. J Physiol 484 (Pt 3), 707-720.

English, D.F., Peyrache, A., Stark, E., Roux, L., Vallentin, D., Long, M.A., and Buzsaki, G. (2014). Excitation and inhibition compete to control spiking during hippocampal ripples: intracellular study in behaving mice. J Neurosci 34, 16509-16517.

Enoki, R., Inoue, M., Hashimoto, Y., Kudo, Y., and Miyakawa, H. (2001). GABAergic control of synaptic summation in hippocampal CA1 pyramidal neurons. Hippocampus 11, 683-689.

Evstratova, A., Chamberland, S., and Topolnik, L. (2011). Cell type-specific and activity-dependent dynamics of action potential-evoked Ca2+ signals in dendrites of hippocampal inhibitory interneurons. J Physiol 589, 1957-1977.

External, R.N.A.C.C. (2005). Proposed methods for testing and selecting the ERCC external RNA controls. BMC Genomics 6, 150.

Fanselow, M.S., and Dong, H.W. (2010). Are the dorsal and ventral hippocampus functionally distinct structures? Neuron 65, 7-19.

Ferraguti, F., Cobden, P., Pollard, M., Cope, D., Shigemoto, R., Watanabe, M., and Somogyi, P. (2004). Immunolocalization of metabotropic glutamate receptor 1alpha (mGluR1alpha) in distinct classes of interneuron in the CA1 region of the rat hippocampus. Hippocampus 14, 193-215.

Ferraguti, F., Klausberger, T., Cobden, P., Baude, A., Roberts, J.D., Szucs, P., Kinoshita, A., Shigemoto, R., Somogyi, P., and Dalezios, Y. (2005). Metabotropic glutamate receptor 8-expressing nerve terminals target subsets of GABAergic neurons in the hippocampus. J Neurosci 25, 10520- 10536.

Finak, G., McDavid, A., Yajima, M., Deng, J., Gersuk, V., Shalek, A.K., Slichter, C.K., Miller, H.W., McElrath, M.J., Prlic, M., et al. (2015). MAST: a flexible statistical framework for assessing transcriptional changes and characterizing heterogeneity in single-cell RNA sequencing data. Genome Biol 16, 278.

Foldy, C., Darmanis, S., Aoto, J., Malenka, R.C., Quake, S.R., and Sudhof, T.C. (2016). Single-cell RNAseq reveals cell adhesion molecule profiles in electrophysiologically defined neurons. Proc Natl Acad Sci U S A 113, E5222-5231.

Francavilla, R., Villette, V., Luo, X., Chamberland, S., Munoz-Pino, E., Camire, O., Wagner, K., Kis, V., Somogyi, P., and Topolnik, L. (2018). Connectivity and network state-dependent recruitment of long-range VIP-GABAergic neurons in the mouse hippocampus. bioRxiv.

Fu, Y., Tucciarone, J.M., Espinosa, J.S., Sheng, N., Darcy, D.P., Nicoll, R.A., Huang, Z.J., and Stryker, M.P. (2014). A cortical circuit for gain control by behavioral state. Cell 156, 1139-1152.

Fuentealba, P., Tomioka, R., Dalezios, Y., Marton, L.F., Studer, M., Rockland, K., Klausberger, T., and Somogyi, P. (2008). Rhythmically active enkephalin-expressing GABAergic cells in the CA1 area of the hippocampus project to the subiculum and preferentially innervate interneurons. J Neurosci 28, 10017-10022.

Fung, S.J., Fillman, S.G., Webster, M.J., and Shannon Weickert, C. (2014). Schizophrenia and bipolar disorder show both common and distinct changes in cortical interneuron markers. Schizophr Res 155, 26-30.

220

Fuzik, J., Zeisel, A., Mate, Z., Calvigioni, D., Yanagawa, Y., Szabo, G., Linnarsson, S., and Harkany, T. (2016). Integration of electrophysiological recordings with single-cell RNA-seq data identifies neuronal subtypes. Nat Biotechnol 34, 175-183.

Galarreta, M., and Hestrin, S. (1999). A network of fast-spiking cells in the neocortex connected by electrical synapses. Nature 402, 72-75.

Galvan, E.J., Perez-Rosello, T., Gomez-Lira, G., Lara, E., Gutierrez, R., and Barrionuevo, G. (2015). Synapse-specific compartmentalization of signaling cascades for LTP induction in CA3 interneurons. Neuroscience 290, 332-345.

Ganter, P., Szucs, P., Paulsen, O., and Somogyi, P. (2004). Properties of horizontal axo-axonic cells in stratum oriens of the hippocampal CA1 area of rats in vitro. Hippocampus 14, 232-243.

Gloveli, T., Dugladze, T., Saha, S., Monyer, H., Heinemann, U., Traub, R.D., Whittington, M.A., and Buhl, E.H. (2005). Differential involvement of oriens/pyramidale interneurones in hippocampal network oscillations in vitro. J Physiol 562, 131-147.

Goldberg, J.H., Lacefield, C.O., and Yuste, R. (2004). Global dendritic calcium spikes in mouse layer 5 low threshold spiking interneurones: implications for control of pyramidal cell bursting. J Physiol 558, 465-478.

Goldberg, J.H., Tamas, G., and Yuste, R. (2003b). Ca2+ imaging of mouse neocortical interneurone dendrites: Ia-type K+ channels control action potential backpropagation. J Physiol 551, 49-65.

Goldberg, J.H., Tamas, G., Aronov, D., and Yuste, R. (2003a). Calcium microdomains in aspiny dendrites. Neuron 40, 807-821.

Goldin, M., Epsztein, J., Jorquera, I., Represa, A., Ben-Ari, Y., Crepel, V., and Cossart, R. (2007). Synaptic kainate receptors tune oriens-lacunosum moleculare interneurons to operate at theta frequency. J Neurosci 27, 9560-9572.

Golding, N.L., Staff, N.P., and Spruston, N. (2002). Dendritic spikes as a mechanism for cooperative long-term potentiation. Nature 418, 326-331.

Guet-McCreight, A., Camire, O., Topolnik, L., and Skinner, F.K. (2016). Using a Semi-Automated Strategy to Develop Multi-Compartment Models That Predict Biophysical Properties of Interneuron- Specific 3 (IS3) Cells in Hippocampus. eNeuro 3.

Gulyas, A.I., Hajos, N., and Freund, T.F. (1996). Interneurons containing calretinin are specialized to control other interneurons in the rat hippocampus. J Neurosci 16, 3397-3411.

Gulyas, A.I., Megias, M., Emri, Z., and Freund, T.F. (1999). Total number and ratio of excitatory and inhibitory synapses converging onto single interneurons of different types in the CA1 area of the rat hippocampus. J Neurosci 19, 10082-10097.

Gulyas, A.I., Miles, R., Hajos, N., and Freund, T.F. (1993). Precision and variability in postsynaptic target selection of inhibitory cells in the hippocampal CA3 region. Eur J Neurosci 5, 1729-1751.

221

Haglund, L., Swanson, L.W., and Kohler, C. (1984). The projection of the supramammillary nucleus to the hippocampal formation: an immunohistochemical and anterograde transport study with the lectin PHA-L in the rat. J Comp Neurol 229, 171-185.

Hainmueller, T., Krieglstein, K., Kulik, A., and Bartos, M. (2014). Joint CP-AMPA and group I mGlu receptor activation is required for synaptic plasticity in dentate gyrus fast-spiking interneurons. Proc Natl Acad Sci U S A 111, 13211-13216.

Han, Z.S., Buhl, E.H., Lorinczi, Z., and Somogyi, P. (1993). A high degree of spatial selectivity in the axonal and dendritic domains of physiologically identified local-circuit neurons in the dentate gyrus of the rat hippocampus. Eur J Neurosci 5, 395-410.

Harris, K.D., Hochgerner, H., Skene, N.G., Magno, L., Katona, L., Bengtsson Gonzales, C., Somogyi, P., Kessaris, N., Linnarsson, S., and Hjerling-Leffler, J. (2018). Classes and continua of hippocampal CA1 inhibitory neurons revealed by single-cell transcriptomics. PLoS Biol 16, e2006387.

Haussler, U., Rinas, K., Kilias, A., Egert, U., and Haas, C.A. (2016). Mossy fiber sprouting and pyramidal cell dispersion in the hippocampal CA2 region in a mouse model of temporal lobe epilepsy. Hippocampus 26, 577-588.

Hedlund, E., and Deng, Q. (2018). Single-cell RNA sequencing: Technical advancements and biological applications. Mol Aspects Med 59, 36-46.

Hevner, R.F., and Wong-Riley, M.T. (1992). Entorhinal cortex of the human, monkey, and rat: metabolic map as revealed by cytochrome oxidase. J Comp Neurol 326, 451-469.

Hoffman, D.A., Magee, J.C., Colbert, C.M., and Johnston, D. (1997). K+ channel regulation of signal propagation in dendrites of hippocampal pyramidal neurons. Nature 387, 869-875.

Hollmann, M., Hartley, M., and Heinemann, S. (1991). Ca2+ permeability of KA-AMPA--gated glutamate receptor channels depends on subunit composition. Science 252, 851-853.

Hu, H., Martina, M., and Jonas, P. (2010). Dendritic mechanisms underlying rapid synaptic activation of fast-spiking hippocampal interneurons. Science 327, 52-58.

Hu, P., Fabyanic, E., Kwon, D.Y., Tang, S., Zhou, Z., and Wu, H. (2017). Dissecting Cell-Type Composition and Activity-Dependent Transcriptional State in Mammalian Brains by Massively Parallel Single-Nucleus RNA-Seq. Mol Cell 68, 1006-1015 e1007.

Huang da, W., Sherman, B.T., and Lempicki, R.A. (2009). Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc 4, 44-57.

Igarashi, K.M., Lu, L., Colgin, L.L., Moser, M.B., and Moser, E.I. (2014). Coordination of entorhinal- hippocampal ensemble activity during associative learning. Nature 510, 143-147.

Ilicic, T., Kim, J.K., Kolodziejczyk, A.A., Bagger, F.O., McCarthy, D.J., Marioni, J.C., and Teichmann, S.A. (2016). Classification of low quality cells from single-cell RNA-seq data. Genome Biol 17, 29.

Jackson, J., Ayzenshtat, I., Karnani, M.M., and Yuste, R. (2016). VIP+ interneurons control neocortical activity across brain states. J Neurophysiol 115, 3008-3017.

Jakovcevski, I., Mayer, N., and Zecevic, N. (2011). Multiple origins of human neocortical interneurons are supported by distinct expression of transcription factors. Cereb Cortex 21, 1771-1782.

222

Jinno, S. (2009). Structural organization of long-range GABAergic projection system of the hippocampus. Front Neuroanat 3, 13.

Jinno, S., Klausberger, T., Marton, L.F., Dalezios, Y., Roberts, J.D., Fuentealba, P., Bushong, E.A., Henze, D., Buzsaki, G., and Somogyi, P. (2007). Neuronal diversity in GABAergic long-range projections from the hippocampus. J Neurosci 27, 8790-8804.

Jones, R.S., and Buhl, E.H. (1993). Basket-like interneurones in layer II of the entorhinal cortex exhibit a powerful NMDA-mediated synaptic excitation. Neurosci Lett 149, 35-39.

Jüngling, K., Eulenburg, V., Moore, R., Kemler, R., Lessmann, V., and Gottmann, K. (2006). N- cadherin transsynaptically regulates short-term plasticity at glutamatergic synapses in embryonic stem cell-derived neurons. J. Neurosci. 26, 6968–6978.

Karnani, M.M., Jackson, J., Ayzenshtat, I., Hamzehei Sichani, A., Manoocheri, K., Kim, S., and Yuste, R. (2016). Opening Holes in the Blanket of Inhibition: Localized Lateral Disinhibition by VIP Interneurons. J Neurosci 36, 3471-3480.

Karson, M.A., Tang, A.H., Milner, T.A., and Alger, B.E. (2009). Synaptic cross talk between perisomatic-targeting interneuron classes expressing cholecystokinin and parvalbumin in hippocampus. J Neurosci 29, 4140-4154.

Kato, H.K., Asinof, S.K., and Isaacson, J.S. (2017). Network-Level Control of Frequency Tuning in Auditory Cortex. Neuron 95, 412-423 e414.

Katona, G., Kaszas, A., Turi, G.F., Hajos, N., Tamas, G., Vizi, E.S., and Rozsa, B. (2011). Roller Coaster Scanning reveals spontaneous triggering of dendritic spikes in CA1 interneurons. Proc Natl Acad Sci U S A 108, 2148-2153.

Katona, L., Lapray, D., Viney, T.J., Oulhaj, A., Borhegyi, Z., Micklem, B.R., Klausberger, T., and Somogyi, P. (2014). Sleep and movement differentiates actions of two types of somatostatin- expressing GABAergic interneuron in rat hippocampus. Neuron 82, 872-886.

Kharchenko, P.V., Silberstein, L., and Scadden, D.T. (2014). Bayesian approach to single-cell differential expression analysis. Nat Methods 11, 740-742.

Kilduff, T.S., Cauli, B., and Gerashchenko, D. (2011). Activation of cortical interneurons during sleep: an anatomical link to homeostatic sleep regulation? Trends Neurosci 34, 10-19.

Kitamura, T., Pignatelli, M., Suh, J., Kohara, K., Yoshiki, A., Abe, K., and Tonegawa, S. (2014). Island cells control temporal association memory. Science 343, 896-901.

Kivioja, T., Vaharautio, A., Karlsson, K., Bonke, M., Enge, M., Linnarsson, S., and Taipale, J. (2011). Counting absolute numbers of molecules using unique molecular identifiers. Nat Methods 9, 72-74.

Klausberger, T., and Somogyi, P. (2008). Neuronal diversity and temporal dynamics: the unity of hippocampal circuit operations. Science 321, 53-57.

Klausberger, T., Marton, L.F., Baude, A., Roberts, J.D., Magill, P.J., and Somogyi, P. (2004). Spike timing of dendrite-targeting bistratified cells during hippocampal network oscillations in vivo. Nat Neurosci 7, 41-47.

223

Koester, H.J., and Sakmann, B. (2000). Calcium dynamics associated with action potentials in single nerve terminals of pyramidal cells in layer 2/3 of the young rat neocortex. J Physiol 529 Pt 3, 625- 646.

Koh, D.S., Geiger, J.R., Jonas, P., and Sakmann, B. (1995). Ca(2+)-permeable AMPA and NMDA receptor channels in basket cells of rat hippocampal dentate gyrus. J Physiol 485 ( Pt 2), 383-402.

Kohara, K., Pignatelli, M., Rivest, A.J., Jung, H.Y., Kitamura, T., Suh, J., Frank, D., Kajikawa, K., Mise, N., Obata, Y., et al. (2014). Cell type-specific genetic and optogenetic tools reveal hippocampal CA2 circuits. Nat Neurosci 17, 269-279.

Kowalski, J., Gan, J., Jonas, P., and Pernia-Andrade, A.J. (2016). Intrinsic membrane properties determine hippocampal differential firing pattern in vivo in anesthetized rats. Hippocampus 26, 668- 682.

Koyama, M., and Pujala, A. (2018). Mutual inhibition of lateral inhibition: a network motif for an elementary computation in the brain. Curr Opin Neurobiol 49, 69-74.

Kramer, A., Green, J., Pollard, J., Jr., and Tugendreich, S. (2014). Causal analysis approaches in Ingenuity Pathway Analysis. Bioinformatics 30, 523-530.

Krimer, L.S., Hyde, T.M., Herman, M.M., and Saunders, R.C. (1997). The entorhinal cortex: an examination of cyto- and myeloarchitectonic organization in humans. Cereb Cortex 7, 722-731.

Laezza, F., Doherty, J.J., and Dingledine, R. (1999). Long-term depression in hippocampal interneurons: joint requirement for pre- and postsynaptic events. Science 285, 1411-1414.

Lamsa, K., Irvine, E.E., Giese, K.P., and Kullmann, D.M. (2007). NMDA receptor-dependent long- term potentiation in mouse hippocampal interneurons shows a unique dependence on Ca(2+)/calmodulin-dependent kinases. J Physiol 584, 885-894.

Langmoen, I.A., and Andersen, P. (1983). Summation of excitatory postsynaptic potentials in hippocampal pyramidal cells. J Neurophysiol 50, 1320-1329.

Larkum, M.E., Nevian, T., Sandler, M., Polsky, A., and Schiller, J. (2009). Synaptic integration in tuft dendrites of layer 5 pyramidal neurons: a new unifying principle. Science 325, 756-760.

Larkum, M.E., Zhu, J.J., and Sakmann, B. (2001). Dendritic mechanisms underlying the coupling of the dendritic with the axonal action potential initiation zone of adult rat layer 5 pyramidal neurons. J Physiol 533, 447-466.

Leao, R.N., Mikulovic, S., Leao, K.E., Munguba, H., Gezelius, H., Enjin, A., Patra, K., Eriksson, A., Loew, L.M., Tort, A.B., and Kullander, K. (2012). OLM interneurons differentially modulate CA3 and entorhinal inputs to hippocampal CA1 neurons. Nat Neurosci 15, 1524-1530.

Lee, S., Hjerling-Leffler, J., Zagha, E., Fishell, G., and Rudy, B. (2010). The largest group of superficial neocortical GABAergic interneurons expresses ionotropic serotonin receptors. J Neurosci 30, 16796- 16808.

Lee, S., Kruglikov, I., Huang, Z.J., Fishell, G., and Rudy, B. (2013). A disinhibitory circuit mediates motor integration in the somatosensory cortex. Nat Neurosci 16, 1662-1670.

224

Lee, S.H., Marchionni, I., Bezaire, M., Varga, C., Danielson, N., Lovett-Barron, M., Losonczy, A., and Soltesz, I. (2014). Parvalbumin-positive basket cells differentiate among hippocampal pyramidal cells. Neuron 82, 1129-1144.

Lei, S., and McBain, C.J. (2002). Distinct NMDA receptors provide differential modes of transmission at mossy fiber-interneuron synapses. Neuron 33, 921-933.

Leranth, C., Frotscher, M., Tombol, T., and Palkovits, M. (1984). Ultrastructure and synaptic connections of vasoactive intestinal polypeptide-like immunoreactive non-pyramidal neurons and axon terminals in the rat hippocampus. Neuroscience 12, 531-542.

Li, Y., Xu, J., Liu, Y., Zhu, J., Liu, N., Zeng, W., Huang, N., Rasch, M.J., Jiang, H., Gu, X., et al. (2017). A distinct entorhinal cortex to hippocampal CA1 direct circuit for olfactory associative learning. Nat Neurosci 20, 559-570.

Lin, J.C., Ho, W.H., Gurney, A., and Rosenthal, A. (2003). The netrin-G1 ligand NGL-1 promotes the outgrowth of thalamocortical axons. Nat Neurosci 6, 1270-1276.

Losonczy, A., and Magee, J.C. (2006). Integrative properties of radial oblique dendrites in hippocampal CA1 pyramidal neurons. Neuron 50, 291-307.

Losonczy, A., Zhang, L., Shigemoto, R., Somogyi, P., and Nusser, Z. (2002). Cell type dependence and variability in the short-term plasticity of EPSCs in identified mouse hippocampal interneurones. J Physiol 542, 193-210.

Magee, J.C. (1998). Dendritic hyperpolarization-activated currents modify the integrative properties of hippocampal CA1 pyramidal neurons. J Neurosci 18, 7613-7624.

Magee, J.C., and Cook, E.P. (2000). Somatic EPSP amplitude is independent of synapse location in hippocampal pyramidal neurons. Nat Neurosci 3, 895-903.

Magloczky, Z., Wittner, L., Borhegyi, Z., Halasz, P., Vajda, J., Czirjak, S., and Freund, T.F. (2000). Changes in the distribution and connectivity of interneurons in the epileptic human dentate gyrus. Neuroscience 96, 7-25.

Martina, M., Vida, I., and Jonas, P. (2000). Distal initiation and active propagation of action potentials in interneuron dendrites. Science 287, 295-300.

Matta, J.A., Pelkey, K.A., Craig, M.T., Chittajallu, R., Jeffries, B.W., and McBain, C.J. (2013). Developmental origin dictates interneuron AMPA and NMDA receptor subunit composition and plasticity. Nat Neurosci 16, 1032-1041.

McCormick, D.A., Connors, B.W., Lighthall, J.W., and Prince, D.A. (1985). Comparative electrophysiology of pyramidal and sparsely spiny stellate neurons of the neocortex. J Neurophysiol 54, 782-806.

Melzer, S., Michael, M., Caputi, A., Eliava, M., Fuchs, E.C., Whittington, M.A., and Monyer, H. (2012). Long-range-projecting GABAergic neurons modulate inhibition in hippocampus and entorhinal cortex. Science 335, 1506-1510.

Mootha, V.K., Bunkenborg, J., Olsen, J.V., Hjerrild, M., Wisniewski, J.R., Stahl, E., Bolouri, M.S., Ray, H.N., Sihag, S., Kamal, M., et al. (2003). Integrated analysis of protein composition, tissue diversity, and gene regulation in mouse mitochondria. Cell 115, 629-640.

225

Nevian, T., Larkum, M.E., Polsky, A., and Schiller, J. (2007). Properties of basal dendrites of layer 5 pyramidal neurons: a direct patch-clamp recording study. Nat Neurosci 10, 206-214.

Nishimura-Akiyoshi, S., Niimi, K., Nakashiba, T., and Itohara, S. (2007). Axonal netrin-Gs transneuronally determine lamina-specific subdendritic segments. Proc Natl Acad Sci U S A 104, 14801-14806.

Nissen, W., Szabo, A., Somogyi, J., Somogyi, P., and Lamsa, K.P. (2010). Cell type-specific long- term plasticity at glutamatergic synapses onto hippocampal interneurons expressing either parvalbumin or CB1 cannabinoid receptor. J Neurosci 30, 1337-1347.

Norenberg, A., Hu, H., Vida, I., Bartos, M., and Jonas, P. (2010). Distinct nonuniform cable properties optimize rapid and efficient activation of fast-spiking GABAergic interneurons. Proc Natl Acad Sci U S A 107, 894-899.

Otmakhova, N.A., Otmakhov, N., and Lisman, J.E. (2002). Pathway-specific properties of AMPA and NMDA-mediated transmission in CA1 hippocampal pyramidal cells. J Neurosci 22, 1199-1207.

Papp, E.C., Hajos, N., Acsady, L., and Freund, T.F. (1999). Medial septal and median raphe innervation of vasoactive intestinal polypeptide-containing interneurons in the hippocampus. Neuroscience 90, 369-382.

Paul, A., Crow, M., Raudales, R., He, M., Gillis, J., and Huang, Z.J. (2017). Transcriptional Architecture of Synaptic Communication Delineates GABAergic Neuron Identity. Cell 171, 522-539 e520.

Pawelzik, H., Hughes, D.I., and Thomson, A.M. (2002). Physiological and morphological diversity of immunocytochemically defined parvalbumin- and cholecystokinin-positive interneurones in CA1 of the adult rat hippocampus. J Comp Neurol 443, 346-367.

Pelkey, K.A., Chittajallu, R., Craig, M.T., Tricoire, L., Wester, J.C., and McBain, C.J. (2017). Hippocampal GABAergic Inhibitory Interneurons. Physiol Rev 97, 1619-1747.

Pereira, J.B., Mijalkov, M., Kakaei, E., Mecocci, P., Vellas, B., Tsolaki, M., Kloszewska, I., Soininen, H., Spenger, C., Lovestone, S., et al. (2016). Disrupted Network Topology in Patients with Stable and Progressive Mild Cognitive Impairment and Alzheimer's Disease. Cereb Cortex 26, 3476-3493.

Perez, Y., Morin, F., and Lacaille, J.C. (2001). A hebbian form of long-term potentiation dependent on mGluR1a in hippocampal inhibitory interneurons. Proc Natl Acad Sci U S A 98, 9401-9406.

Petilla Interneuron Nomenclature, G., Ascoli, G.A., Alonso-Nanclares, L., Anderson, S.A., Barrionuevo, G., Benavides-Piccione, R., Burkhalter, A., Buzsaki, G., Cauli, B., Defelipe, J., et al. (2008). Petilla terminology: nomenclature of features of GABAergic interneurons of the cerebral cortex. Nat Rev Neurosci 9, 557-568.

Pfisterer, U., and Khodosevich, K. (2017). Neuronal survival in the brain: neuron type-specific mechanisms. Cell Death Dis 8, e2643.

Pinto, L., and Dan, Y. (2015). Cell-Type-Specific Activity in Prefrontal Cortex during Goal-Directed Behavior. Neuron 87, 437-450.

226

Poorthuis, R.B., Muhammad, K., Wang, M., Verhoog, M.B., Junek, S., Wrana, A., Mansvelder, H.D., and Letzkus, J.J. (2018). Rapid Neuromodulation of Layer 1 Interneurons in Human Neocortex. Cell Rep 23, 951-958.

Potter, S.S. (2018). Single-cell RNA sequencing for the study of development, physiology and disease. Nat Rev Nephrol 14, 479-492.

Price, C.J., Cauli, B., Kovacs, E.R., Kulik, A., Lambolez, B., Shigemoto, R., and Capogna, M. (2005). Neurogliaform neurons form a novel inhibitory network in the hippocampal CA1 area. J Neurosci 25, 6775-6786.

Raj, A., and van Oudenaarden, A. (2008). Nature, nurture, or chance: stochastic gene expression and its consequences. Cell 135, 216-226.

Ramón y Cajal, S. Estuctura del asta de Ammon. 1893.

Ramskold, D., Luo, S., Wang, Y.C., Li, R., Deng, Q., Faridani, O.R., Daniels, G.A., Khrebtukova, I., Loring, J.F., Laurent, L.C., et al. (2012). Full-length mRNA-Seq from single-cell levels of RNA and individual circulating tumor cells. Nat Biotechnol 30, 777-782.

Reyes, A., Lujan, R., Rozov, A., Burnashev, N., Somogyi, P., and Sakmann, B. (1998). Target-cell- specific facilitation and depression in neocortical circuits. Nat Neurosci 1, 279-285.

Rozov, A., and Burnashev, N. (1999). Polyamine-dependent facilitation of postsynaptic AMPA receptors counteracts paired-pulse depression. Nature 401, 594-598.

Schurch, N.J., Schofield, P., Gierlinski, M., Cole, C., Sherstnev, A., Singh, V., Wrobel, N., Gharbi, K., Simpson, G.G., Owen-Hughes, T., et al. (2016). How many biological replicates are needed in an RNA-seq experiment and which differential expression tool should you use? Rna 22, 839-851.

Schwanhausser, B., Busse, D., Li, N., Dittmar, G., Schuchhardt, J., Wolf, J., Chen, W., and Selbach, M. (2011). Global quantification of mammalian gene expression control. Nature 473, 337-342.

Scoville, W.B., and Milner, B. (1957). Loss of recent memory after bilateral hippocampal lesions. J Neurol Neurosurg Psychiatry 20, 11-21.

Sekulic, V., Chen, T.C., Lawrence, J.J., and Skinner, F.K. (2015). Dendritic distributions of I h channels in experimentally-derived multi-compartment models of oriens-lacunosum/moleculare (O- LM) hippocampal interneurons. Front Synaptic Neurosci 7, 2.

Seroogy, K.B., Seress, L., and Ribak, C.E. (1983). Ultrastructure of commissural neurons of the hilar region in the hippocampal dentate gyrus. Exp Neurol 82, 594-608.

Sik, A., Penttonen, M., Ylinen, A., and Buzsaki, G. (1995). Hippocampal CA1 interneurons: an in vivo intracellular labeling study. J Neurosci 15, 6651-6665.

Silberberg, G., and Markram, H. (2007). Disynaptic inhibition between neocortical pyramidal cells mediated by Martinotti cells. Neuron 53, 735-746.

Simon, A., Olah, S., Molnar, G., Szabadics, J., and Tamas, G. (2005). Gap-junctional coupling between neurogliaform cells and various interneuron types in the neocortex. J Neurosci 25, 6278- 6285.

227

Soriano, E., and Frotscher, M. (1989). A GABAergic axo-axonic cell in the fascia dentata controls the main excitatory hippocampal pathway. Brain Res 503, 170-174.

Spruston, N. (2008). Pyramidal neurons: dendritic structure and synaptic integration. Nat Rev Neurosci 9, 206-221.

Steward, O. (1976). Topographic organization of the projections from the entorhinal area to the hippocampal formation of the rat. J Comp Neurol 167, 285-314.

Stuart, G., and Spruston, N. (1998). Determinants of voltage attenuation in neocortical pyramidal neuron dendrites. J Neurosci 18, 3501-3510.

Subramanian, A., Tamayo, P., Mootha, V.K., Mukherjee, S., Ebert, B.L., Gillette, M.A., Paulovich, A., Pomeroy, S.L., Golub, T.R., Lander, E.S., and Mesirov, J.P. (2005). Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A 102, 15545-15550.

Sun, H.Y., Lyons, S.A., and Dobrunz, L.E. (2005). Mechanisms of target-cell specific short-term plasticity at Schaffer collateral synapses onto interneurones versus pyramidal cells in juvenile rats. J Physiol 568, 815-840.

Sylwestrak, E.L., and Ghosh, A. (2012). Elfn1 regulates target-specific release probability at CA1- interneuron synapses. Science 338, 536-540.

Takacs, V.T., Klausberger, T., Somogyi, P., Freund, T.F., and Gulyas, A.I. (2012). Extrinsic and local glutamatergic inputs of the rat hippocampal CA1 area differentially innervate pyramidal cells and interneurons. Hippocampus 22, 1379-1391.

Tang, Q., Burgalossi, A., Ebbesen, C.L., Ray, S., Naumann, R., Schmidt, H., Spicher, D., and Brecht, M. (2014). Pyramidal and stellate cell specificity of grid and border representations in layer 2 of medial entorhinal cortex. Neuron 84, 1191-1197.

Tasic, B., Menon, V., Nguyen, T.N., Kim, T.K., Jarsky, T., Yao, Z., Levi, B., Gray, L.T., Sorensen, S.A., Dolbeare, T., et al. (2016). Adult mouse cortical cell taxonomy revealed by single cell transcriptomics. Nat Neurosci 19, 335-346.

Topolnik, L., Azzi, M., Morin, F., Kougioumoutzakis, A., and Lacaille, J.C. (2006). mGluR1/5 subtype- specific calcium signalling and induction of long-term potentiation in rat hippocampal oriens/alveus interneurones. J Physiol 575, 115-131.

Topolnik, L., Congar, P., and Lacaille, J.C. (2005). Differential regulation of metabotropic glutamate receptor- and AMPA receptor-mediated dendritic Ca2+ signals by presynaptic and postsynaptic activity in hippocampal interneurons. J Neurosci 25, 990-1001.

Toth, K., and Freund, T.F. (1992). Calbindin D28k-containing nonpyramidal cells in the rat hippocampus: their immunoreactivity for GABA and projection to the medial septum. Neuroscience 49, 793-805.

Toth, K., Eross, L., Vajda, J., Halasz, P., Freund, T.F., and Magloczky, Z. (2010). Loss and reorganization of calretinin-containing interneurons in the epileptic human hippocampus. Brain 133, 2763-2777.

228

Trapnell, C., Cacchiarelli, D., Grimsby, J., Pokharel, P., Li, S., Morse, M., Lennon, N.J., Livak, K.J., Mikkelsen, T.S., and Rinn, J.L. (2014). The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells. Nat Biotechnol 32, 381-386.

Traub, R.D., Whittington, M.A., Stanford, I.M., and Jefferys, J.G. (1996). A mechanism for generation of long-range synchronous fast oscillations in the cortex. Nature 383, 621-624.

Tricoire, L., Pelkey, K.A., Daw, M.I., Sousa, V.H., Miyoshi, G., Jeffries, B., Cauli, B., Fishell, G., and McBain, C.J. (2010). Common origins of hippocampal Ivy and nitric oxide synthase expressing neurogliaform cells. J Neurosci 30, 2165-2176.

Tricoire, L., Pelkey, K.A., Erkkila, B.E., Jeffries, B.W., Yuan, X., and McBain, C.J. (2011). A blueprint for the spatiotemporal origins of mouse hippocampal interneuron diversity. J Neurosci 31, 10948- 10970.

Tropp Sneider, J., Chrobak, J.J., Quirk, M.C., Oler, J.A., and Markus, E.J. (2006). Differential behavioral state-dependence in the burst properties of CA3 and CA1 neurons. Neuroscience 141, 1665-1677.

Tsien, R.W., Lipscombe, D., Madison, D.V., Bley, K.R., and Fox, A.P. (1988). Multiple types of neuronal calcium channels and their selective modulation. Trends Neurosci 11, 431-438.Tyan, L., Chamberland, S., Magnin, E., Camire, O., Francavilla, R., David, L.S., Deisseroth, K., and Topolnik, L. (2014). Dendritic inhibition provided by interneuron-specific cells controls the firing rate and timing of the hippocampal feedback inhibitory circuitry. J Neurosci 34, 4534-4547.

Urban, Z., Magloczky, Z., and Freund, T.F. (2002). Calretinin-containing interneurons innervate both principal cells and interneurons in the CA1 region of the human hippocampus. Acta Biol Hung 53, 205-220.

Vacic, V., McCarthy, S., Malhotra, D., Murray, F., Chou, H.H., Peoples, A., Makarov, V., Yoon, S., Bhandari, A., Corominas, R., et al. (2011). Duplications of the neuropeptide receptor gene VIPR2 confer significant risk for schizophrenia. Nature 471, 499-503.

Viana Di Prisco, G., Albo, Z., Vertes, R.P., and Kocsis, B. (2002). Discharge properties of neurons of the median raphe nucleus during hippocampal theta rhythm in the rat. Exp Brain Res 145, 383-394. von Engelhardt, J., Bocklisch, C., Tonges, L., Herb, A., Mishina, M., and Monyer, H. (2015). GluN2D- containing NMDA receptors-mediate synaptic currents in hippocampal interneurons and pyramidal cells in juvenile mice. Front Cell Neurosci 9, 95.

Walsh, D.A., Brown, J.T., and Randall, A.D. (2017). In vitro characterization of cell-level neurophysiological diversity in the rostral nucleus reuniens of adult mice. J Physiol 595, 3549-3572.

Westenbroek, R.E., Ahlijanian, M.K., and Catterall, W.A. (1990). Clustering of L-type Ca2+ channels at the base of major dendrites in hippocampal pyramidal neurons. Nature 347, 281-284.

Witter, M.P., and Amaral, D.G. (2004). CHAPTER 21 - Hippocampal Formation. In The Rat Nervous System (Third Edition), G. Paxinos, ed. (Burlington: Academic Press), pp. 635-704.

Wyss, J.M. (1981). An autoradiographic study of the efferent connections of the entorhinal cortex in the rat. J Comp Neurol 199, 495-512.

229

Yamasaki, M., Fukaya, M., Yamazaki, M., Azechi, H., Natsume, R., Abe, M., Sakimura, K., and Watanabe, M. (2016). TARP gamma-2 and gamma-8 Differentially Control AMPAR Density Across Schaffer Collateral/Commissural Synapses in the Hippocampal CA1 Area. J Neurosci 36, 4296-4312.

Yizhar, O., Fenno, L.E., Prigge, M., Schneider, F., Davidson, T.J., O'Shea, D.J., Sohal, V.S., Goshen, I., Finkelstein, J., Paz, J.T., et al. (2011). Neocortical excitation/inhibition balance in information processing and social dysfunction. Nature 477, 171-178.

Yoder, R.M., and Pang, K.C. (2005). Involvement of GABAergic and cholinergic medial septal neurons in hippocampal theta rhythm. Hippocampus 15, 381-392.

Zecevic, N., Hu, F., and Jakovcevski, I. (2011). Interneurons in the developing human neocortex. Dev Neurobiol 71, 18-33.

Zeisel, A., Munoz-Manchado, A.B., Codeluppi, S., Lonnerberg, P., La Manno, G., Jureus, A., Marques, S., Munguba, H., He, L., Betsholtz, C., et al. (2015). Brain structure. Cell types in the mouse cortex and hippocampus revealed by single-cell RNA-seq. Science 347, 1138-1142.

Zhong, S., Zhang, S., Fan, X., Wu, Q., Yan, L., Dong, J., Zhang, H., Li, L., Sun, L., Pan, N., et al. (2018). A single-cell RNA-seq survey of the developmental landscape of the human prefrontal cortex. Nature 555, 524-528.

Zucker, R.S., and Regehr, W.G. (2002). Short-term synaptic plasticity. Annu Rev Physiol 64, 355- 405.

230

Appendices

Article 4: Coordination of dendritic inhibition through local disinhibitory circuits

Résumé

Il est connu depuis un certain temps que différents sous-types d'interneurones inhibiteurs corticaux innervent des domaines dendritiques spécifiques des cellules principales et libèrent du GABA à des moments particuliers au cours d'oscillations du réseau associées aux comportements. Cependant, le manque d'informations sur la manière dont l'activité des interneurones peut être contrôlée par le GABA pendant les différents états comportementaux a gêné notre compréhension des règles qui régissent l'organisation spatio-temporelle et la fonction de l'inhibition dendritique. Semblable aux cellules principales, tout interneurone donné peut recevoir plusieurs signaux inhibiteurs fonctionnellement distincts qui ciblent ses domaines subcellulaires spécifiques. Nous avons récemment découvert que les circuits locaux des interneurones (IS) spécifiques aux interneurones sont responsables de l'inhibition dendritique de différents sous-types d'interneurones hippocampiques avec un impact important sur la production des signaux cellulaires. Ici, nous allons examiner les propriétés et la spécificité des connexions des interneurones IS dans la région CA1 de l'hippocampe et le néocortex et discuter de leur rôle possible dans la régulation dépendante de l'activité de l'inhibition dendritique reçue par les neurones pyramidaux.

Abstract:

It has been recognized for some time that different subtypes of cortical inhibitory interneurons innervate specific dendritic domains of principal cells and release GABA at particular times during behaviorally relevant network oscillations. However, the lack of basic information on how the activity of interneurons can be controlled by GABA released in particular behavioral states has hindered our understanding of the rules that govern the spatio-temporal organization and function of dendritic inhibition. Similar to principal cells, any given interneuron may receive several functionally distinct inhibitory inputs that target its specific subcellular domains. We recently found that local circuitry of the so-called interneuron-specific (IS) interneurons is responsible for dendritic inhibition of different subtypes of hippocampal interneurons with a great impact on cell output. Here, we will review the properties and the specificity of connections of IS interneurons in the CA1 hippocampus and neocortex, and discuss their possible role in the activity-dependent regulation of dendritic inhibition received by pyramidal neurons.

231

Introduction

In neocortical and hippocampal networks, a large diversity of GABAergic inhibitory inputs converges onto the dendrites of glutamatergic principal cells. Many of them may overlap within the same dendritic domain but remain segregated temporally due to specific inhibitory mechanisms that evolved to control the activity of dendrite-targeting interneurons. The vasoactive intestinal polypeptide (VIP) and/or calretinin (CR) expressing interneurons have been consistently associated with cortical dendritic disinhibition. For example, in the CA1 hippocampal area, three types of the so-called interneuron-specific (IS) interneurons have been shown to make symmetric contacts with interneurons selectively (Acsády et al., 1996; Gulyás et al., 1996). Type 1 (IS1) cells express CR and have a soma located in the oriens/alveus (O/A), stratum pyramidale (PYR) or radiatum (RAD). Type 2 (IS2) interneurons express VIP but lack CR: they have a soma located between the RAD and lacunosum-moleculare (LM), a dendritic arbor restricted to LM and axonal projections in the RAD targeting cholecystokinin (CCK)/VIP coexpressing basket cells. Type 3 (IS3) interneurons coexpress CR and VIP and may also express enkephalins with a soma located at the PYR and RAD border and dendrites extending into LM (Blasco-Ibáñez et al., 1998). IS3 cells have been reported to contact preferentially somatostatin (SOM)- and metabotropic glutamate receptor 1a (mGluR1a)positive oriens–lacunosum moleculare (OLM) cells that are responsible for distal dendritic inhibition of CA1 pyramidal neurons (Acsády et al., 1996). In a second example, the majority of VIP+ terminals in the somatosensory cortex are made onto SOM-/mGluR1a- and calbindin (CB)-expressing interneurons that provide dendritic inhibition to the layer II/III and layer V pyramidal cells (Dalezios et al., 2002; Staiger et al., 2004). However, the physiological properties, functional connectivity, recruitment during network activity, and role of IS interneurons in cortical computations remained until recently unknown.

In the past several years, advances in transgenic and optical technologies have converged to enable researchers to target and manipulate specific cell types within highly heterogeneous inhibitory circuits. Using VIP-GFP mice, it became possible to characterize the properties and connectivity of VIP+ interneurons in acute hippocampal slices (Chamberland et al., 2010; Tyan et al., 2014), while mice expressing Cre recombinase and channelrhodopsin (ChR) or halorhodopsin under the control of the VIP or CR promoters have been successfully used to manipulate VIP+ and CR+ interneurons in slices and in awake mice (Lee et al., 2013; Pfeffer et al., 2013; Pi et al., 2013; Tyan et al., 2014). Here, we summarize current knowledge about the properties, connectivity and function of VIP+ interneurons in the hippocampus and neocortex. In particular, we concentrate on hippocampal CA1 IS3 cells that control the level of dendritic inhibition received by CA1 pyramidal neurons. It is not our intention to discuss CR+ cells in cortical circuits, as they represent a highly heterogeneous population of interneurons and have been thoroughly discussed in a recent review (Cauli et al., 2014).

Properties and connectivity of is3 cells

Morphological and neurochemical features

The hippocampal CA1 IS3 cells have small round somata (13–18 µm) located in PYR or RAD with 2– 3 primary dendrites of unipolar or bipolar orientation extending towards LM or LM and O/A, respectively (Figures 1A,B). In most cells, the primary dendrite extending to LM is particularly thick resembling that of pyramidal neurons (Figure 1A upper panel). Dendritic spines can be observed occasionally on proximal and distal branches. These cells send their axon primarily to the O/A but random collaterals can be found in PYR or RAD (Figure 1B upper panel). Accordingly, the major postsynaptic targets of IS3 cells reside in the O/A and correspond to O/A interneurons (Figure 1C). On the basis of immunohistochemistry, IS3 interneurons are defined as GABAergic cells that co- express the Ca2+-binding protein CR and neuropeptide VIP (Figure 1A lower panel) (Acsády et al., 1996; Freund and Buzsáki, 1996; Gulyás et al., 1996).

232

Physiological properties

In acute hippocampal slices, IS3 cells have a resting membrane potential of −64 to −75 mV, suggesting that these cells are likely silent under basal conditions. However, compared with other interneuron subtypes, IS3 interneurons have a particularly high input resistance (400–600 M•) and a small rheobase (30–50 pA), which makes them one of the most excitable interneuron subtypes in the hippocampus. The properties of the action potential, including the spike threshold, the amplitude and the halfwidth are similar to those in other types of neurons (Tyan et al., 2014). Nevertheless, IS3 interneurons can distinguish themselves by a characteristic “irregularly spiking” firing pattern with an inter-spike interval varying broadly upon membrane depolarization (Figure 1B lower panel) (Chamberland et al., 2010).

Connectivity

The IS3 axon shows extensive arborization within O/A with a cumulative axonal length of a single interneuron going up to 11 mm. Data from anatomical analysis (Acsády et al., 1996) and paired electrophysiological recordings (Tyan et al., 2014) showed that IS3 cells contact several distinct subtypes of O/A interneurons, including OLM, bistratified and basket cells as well as some other interneurons with somata, dendrites and axon located within stratum oriens [the so-called oriens– oriens cells]. The OLM cell is the preferential target of IS3 interneurons while the oriens–oriens and bistratified cells share most of the remaining IS3 inputs with a minor proportion of inputs made onto basket cells (Tyan et al., 2014). Taken together, these data indicate that the major role of IS3 interneurons is in coordinating the level of inhibition converging onto different dendritic domains of CA1 pyramidal neurons.

Properties of IS3 synapses

The properties of IS3 synapses made on different targets have been examined using paired patch- clamp recordings (Tyan et al., 2014). In all dendrite-targeting interneurons, unitary inhibitory postsynaptic currents (uIPSCs) recorded at 0 mV at nearphysiological temperature (32 ± 1◦C) had a high failure rate of ∼60%, small amplitude (10–25 pA) and varying kinetics (uIPSC rise time: 0.7–1.3 ms; uIPSC decay τ: 5–12 ms). The latter could result from the different dendritic location of IS3 synapses in distinct targets, although a target-specific GABAA receptor composition cannot be excluded (Salesse et al., 2011). Variance-mean analysis has been performed at IS3–OLM synapses (Tyan et al., 2014). It revealed that an IS3 cell contacts an OLM through multiple release sites and produces uIPSCs with a quantal size of 5–6 pA. Repetitive firing of IS3 cells at 10–100 Hz does not result in any form of short-term plasticity at IS3–OLM synapses. However, efficient summation of slow uIPSCs occurring during 100-Hz firing of IS3 cells leads to a large inhibitory response in OLMs with a potential impact on their firing.

Comparison with VIP+ interneurons in the neocortex

In neocortical regions, VIP+ interneurons have been classified as a sub-group of interneurons that express the 5-hydroxytryptamine 3a receptor (5HT3aR+), making up ∼40% of 5HT3aR+ interneurons (Rudy et al., 2011). Similar to hippocampal IS3 interneurons, most neocortical VIP+ interneurons have a bipolar/ bitufted orientation with soma and dendrites located primarily in layers II/III or V (Figure 1D; Pi et al., 2013). These cells have dendrites located perpendicularly to the pial surface and branching within layers I and V. The axon of bipolar VIP+ interneurons originates from a primary dendrite and makes an extensive arborisation within layers V/VI (Figure 1E; Porter et al., 1998). A sub-population of neocortical VIP+ interneurons coexpress CR (35% of VIP+ cells) and, therefore, may be similar to the hippocampal IS3 interneurons (Figure 1G; Kawaguchi and Kubota, 1997; Porter et al., 1998; Gonchar et al., 2008; Xu et al., 2010). It is to be noted that a fraction of neocortical VIP+ interneurons

233

may also co-express CCK and, therefore, may correspond to VIP+ basket cells (Galarreta et al., 2004; Sugino et al., 2006).

Similar to hippocampal IS3 interneurons, neocortical CR+/VIP+ interneurons recorded in slices in vitro are hyperpolarized with a resting membrane potential of −62 to −74 mV (Porter et al., 1998). These cells have a high input resistance (240–2200 MΩ) and exhibit “irregularly spiking” firing pattern (Figure 1F; Cauli et al., 1997, 2000; Porter et al., 1998; Galarreta et al., 2004; Lee et al., 2010; Miyoshi et al., 2010). In support of their interneuron-selectivity, the results of ultrastructural, physiological and optogenetic analysis revealed that VIP+ interneurons prefer to contact several distinct subtypes of neocortical interneurons, including CB+, SOM+, VIP+ and parvalbumin-positive (PV+) cells (Figure 1H). In particular, electron microscopy studies have shown that VIP+ boutons onto PV+, CB+, SOM and VIP+ interneurons are homogeneously distributed across layers II to VI (Dalezios et al., 2002; Staiger et al., 2004; Dávid et al., 2007). Moreover, paired whole-cell recordings from neocortical layer II/III CR+/VIP+ interneurons showed that these cells prefer to contact several types of interneurons rather than pyramidal cells, including the multipolar CR+/VIP– cells (with a connectivity rate of 80%), fast spiking cells (30%), and PV+ multipolar bursting cells (27%) (Caputi et al., 2009). Furthermore, optogenetic studies using a VIP-Cre mouse model have shown that SOM+ interneurons represent the major target of VIP+ interneurons; in particular, the inhibition provided by VIP+ interneurons was much larger in SOM+ cells compared with PV+ interneurons in the visual and somatosensory cortices (Lee et al., 2013; Pfeffer et al., 2013). A similar observation was reported in the auditory and medial prefrontal areas (Pi et al., 2013), where activation of ChR2-expressing VIP+ interneurons elicited IPSCs primarily in SOM+ cells; albeit no difference in the amplitude of the ChR2-evoked IPSCs appeared between SOM+ and PV+ interneurons. In addition, optogenetic silencing of VIP+ interneurons strongly reduced the IPSCs recorded in neocortical SOM+ cells (Lee et al., 2013). Finally, CR+/VIP+ interneurons are coupled through gap junctions (with a connectivity rate of 63%) (Caputi et al., 2009), which may play an important role in synchronizing the activity of CR+/VIP+ interneurons with a great impact on the output of SOM+ interneurons. Taken together, these studies show that VIP+/CR+ IS interneurons are well positioned to modulate primarily the activity of local SOM+ circuits, providing dendritic disinhibition to cortical pyramidal neurons.

Functional role of disinhibitory circuits

In the CA1 hippocampus, dendritic inhibition provided by the IS3 interneurons controls the firing rate and timing of OLM cells. The latter may be possible because of the dendritic initiation of the action potential in OLM interneurons (Martina et al., 2000). Furthermore, it has been shown that SOM+ dendrite-targeting OLM as well as bistratified cells may be responsible for gating the active dendritic conductances and burst firing of pyramidal cells through initiation of dendritic spikes (Lovett-Barron et al., 2012; Müller and Remy, 2014). From this perspective, IS3 inhibition of SOM+ cells appears to be crucial in coordination of dendritic inhibition of pyramidal neurons with a direct impact on their input-output conversion and firing behavior.

Under what network conditions might this happen in vivo? Based on anatomical data, IS3 cells are likely to be driven by the three major excitatory pathways in the CA1 area: the perforant path, the Schaffer collaterals and the CA1 local collaterals. Additionally, inhibitory input from the CR+ type 1 IS cells may control the activity of IS3 interneurons as CR+ terminals make numerous contacts with CR+ and VIP+ cells (Gulyás et al., 1996). Therefore, the dynamic properties and the relative weight of excitatory and inhibitory inputs converging onto IS3 cells will determine their state-dependent recruitment during ongoing network activity and, accordingly, their role in the recruitment of OLM interneurons in vivo. OLM interneurons demonstrate state-dependent fluctuations in activity during network oscillations. In particular, the firing of OLM cells can vary during different episodes or phases of sharp wave ripples (SWRs). For example, in anesthetized animals, OLM cells were quiet during SWRs (Klausberger et al., 2003), whereas in awake, head-fixed animals, OLM cells could fire with a low probability during some SWR episodes (Varga et al., 2012). In freely moving rats, the firing rate of OLM interneurons decreased significantly during sleep compared to awake states and was low during the sleep-associated SWRs (Katona et al., 2014). In addition, OLM cells recorded in slices in

234

vitro could fire at a later phase of SWRs (Pangalos et al., 2013). Moreover, both OLM and bistratified cells are strongly modulated during theta oscillations in anesthetized as well as freely-moving animals (Klausberger et al., 2003, 2004; Royer et al., 2012; Varga et al., 2012). In particular, optogenetic experiments revealed that SOM+ dendrite-targeting CA1 interneurons fire at the decay phase of place field during spatial learning, and reduce the firing rate of pyramidal cells without changing the theta phase (Royer et al., 2012). Interestingly, firing of IS3 cells at theta frequency resulted in theta synchronization of OLM cells (Tyan et al., 2014). It is therefore plausible to suggest that IS3 interneurons may increase their firing at specific stages of SWRs and/or theta oscillations in vivo and, subsequently, modulate the activity of OLM interneurons.

Recent experimental observations obtained from different neocortical regions highlight the idea that disinhibitory VIP+ interneurons may be engaged in network activity during specific behavioral states (Lee et al., 2013; Pi et al., 2013; Fu et al., 2014). For example, in the somatosensory cortex, the activation of VIP+ interneurons was increased during whisking (Lee et al., 2013). In addition, in the auditory cortex, VIP+ interneurons were strongly recruited by positive and negative reinforcement signals during discrimination tasks (Pi et al., 2013). Furthermore, in the primary sensory cortex, the activity of VIP+ cells was highly increased during locomotion (Fu et al., 2014). Together, these data indicate that VIP+ interneurons may be specialized in controlling the intracortical gating of information during specific behavioral states. Such brain-state-dependent recruitment of VIP+ interneurons points to the important role of the modulatory systems in the regulation of cortical disinhibition. The neuromodulatory effects of dopamine (DA), acetylcholine, and serotonin on pyramidal cells as well as different types of interneurons have been explored in details in different cortical areas. Recent studies have focused on the role of modulators in controlling the recruitment of interneurons in specific behavioral states (Letzkus et al., 2011; Leão et al., 2012; Kimura et al., 2014; Lovett-Barron et al., 2014). For example, it has been reported that dendrite-targeting SOM+ interneurons in the hippocampal CA1 area were recruited by aversive stimuli during contextual fear conditioning through activation of the cholinergic input (Lovett-Barron et al., 2014). VIP+ interneurons in the neocortex express nicotinic acetylcholine receptors (nAChRs; Alitto and Dan, 2013), indicating the potential role of acetylcholine in regulating VIP+ interneuron activity. Indeed, nAChR antagonists strongly attenuated the activation of VIP+ interneurons during behavioral tasks (Fu et al., 2014). Considering the involvement of dopaminergic and cholinergic systems in the reward-associated circuitry (Fukuda et al., 1990; Morris et al., 2004), it is possible that the phasic release of DA and/or acetylcholine, through modulation of VIP+ interneuron activity, may increase their recruitment during reinforcement tasks. Furthermore, in the hippocampus, the dopamine 1 receptor is expressed by CR+ interneurons (Gangarossa et al., 2012). Yet, the role of DA as well as acetylcholine in the recruitment of hippocampal IS3 interneurons remains unexplored.

In conclusion, recent studies from several laboratories provided direct experimental evidence that cortical IS interneurons may play a major role in the state-dependent gating of information flow across cortical regions primarily through dendritic disinhibition of principal neurons. By controlling dendritic electrogenesis and firing mode of principal cells, IS interneurons may determine the functional output of intracortical processing during specific brain states. Future experiments studying the impact of specific connections and their modulation will be required to understand the role of VIP+ interneurons in gating and consolidation of cortical information.

Acknowledgments

This work was supported by the Canadian Institute of Health Research and the Natural Sciences and Engineering Research Council of Canada (NSERC). LT is the recipient of the NSERC Faculty Award for Women. We thank Olivier Camiré for providing valuable comments on the manuscript.

235

References

Acsády, L., Görcs, T. J., and Freund, T. F. (1996). Different populations of vasoactive intestinal polypeptide-immunoreactive interneurons are specialized to control pyramidal cells or interneurons in the hippocampus. Neuroscience 73, 317–334. doi: 10.1016/0306-4522(95)00609-5

Alitto, H. J., and Dan, Y. (2013). Cell-type-specific modulation of neocortical activity by basal forebrain input. Front. Syst. Neurosci. 6:79. doi: 10.3389/fnsys. 2012.00079

Blasco-Ibáñez, J. M., Martínez-Guijarro, F. J., and Freund, T. F. (1998). Enkephalincontaining interneurons are specialized to innervate other interneurons in the hippocampal CA1 region of the rat and guinea-pig. Eur. J. Neurosci. 10, 1784–1795. doi: 10.1046/j.1460-9568.1998.00190.x

Caputi, A., Rozov, A., Blatow, M., and Monyer, H. (2009). Two calretininpositive GABAergic cell types in layer 2/3 of the mouse neocortex provide different forms of inhibition. Cereb. Cortex 19, 1345–1359. doi: 10.1093/cercor/bhn175

Cauli, B., Audinat, E., Lambolez, B., Angulo, M. C., Ropert, N., Tsuzuki, K., et al. (1997). Molecular and physiological diversity of cortical nonpyramidal cells. J. Neurosci. 17, 3894–3906.

Cauli, B., Porter, J. T., Tsuzuki, K., Lambolez, B., Rossier, J., Quenet, B., et al. (2000). Classification of fusiform neocortical interneurons based on unsupervised clustering. Proc. Natl. Acad. Sci. U S A 97, 6144–6149. doi: 10.1073/pnas.97. 11.6144

Cauli, B., Zhou, X., Tricoire, L., Toussay, X., and Staiger, J. F. (2014). Revisiting enigmatic cortical calretinin-expressing interneurons. Front. Neuroanat. 8:52. doi: 10.3389/fnana.2014.00052

Chamberland, S., Salesse, C., Topolnik, D., and Topolnik, L. (2010). Synapsespecific inhibitory control of hippocampal feedback inhibitory circuit. Front. Cell. Neurosci. 4:130. doi: 10.3389/fncel.2010.00130

Dalezios, Y., Luján, R., Shigemoto, R., Roberts, J. D. B., and Somogyi, P. (2002). Enrichment of mGluR7a in the presynaptic active zones of GABAergic and nonGABAergic terminals on interneurons in the rat somatosensory cortex. Cereb. Cortex 12, 961–974. doi: 10.1093/cercor/12.9.961

Dávid, C., Schleicher, A., Zuschratter, W., and Staiger, J. F. (2007). The innervation of parvalbumin- containing interneurons by VIP-immunopositive interneurons in the primary somatosensory cortex of the adult rat. Eur. J. Neurosci. 25, 2329–2340. doi: 10.1111/j.1460-9568.2007.05496.x

Freund, T. F., and Buzsáki, G. (1996). Interneurons of the hippocampus. Hippocampus 6, 347–470. doi: 10.1002/(sici)1098-1063(1996)6:4<347::aidhipo1>3.0.co;2-i

Fu, Y., Tucciarone, J. M., Espinosa, J. S., Sheng, N., Darcy, D. P., Nicoll, R. A., et al. (2014). A cortical circuit for gain control by behavioral state. Cell 156, 1139–1152. doi: 10.1016/j.cell.2014.01.050

Fukuda, M., Ono, T., Nakamura, K., and Tamura, R. (1990). Dopamine and ACh involvement in plastic learning by hypothalamic neurons in rats. Brain Res. Bull. 25, 109–114. doi: 10.1016/0361- 9230(90)90260-7

236

Galarreta, M., Erdélyi, F., Szabó, G., and Hestrin, S. (2004). Electrical coupling among irregular- spiking GABAergic interneurons expressing cannabinoid receptors. J. Neurosci. 24, 9770–9778. doi: 10.1523/jneurosci.3027-04. 2004

Gangarossa, G., Longueville, S., De Bundel, D., Perroy, J., Hervé, D., Girault, J.-A., et al. (2012). Characterization of dopamine D1 and D2 receptor-expressing neurons in the mouse hippocampus. Hippocampus 22, 2199–2207. doi: 10. 1002/hipo.22044

Gonchar, Y., Wang, Q., and Burkhalter, A. (2008). Multiple distinct subtypes of GABAergic neurons in mouse visual cortex identified by triple immunostaining. Front. Neuroanat. 1:3. doi: 10.3389/neuro.05.003.2007

Gulyás, A. I., Hájos, N., and Freund, T. F. (1996). Interneurons containing calretinin are specialized to control other interneurons in the rat hippocampus. J. Neurosci. 16, 3397–3411.

Katona, L., Lapray, D., Viney, T. J., Oulhaj, A., Borhegyi, Z., Micklem, B. R., et al. (2014). Sleep and movement differentiates actions of two types of somatostatin-expressing GABAergic interneuron in rat hippocampus. Neuron 82, 872–886. doi: 10.1016/j.neuron.2014.04.007

Kawaguchi, Y., and Kubota, Y. (1997). GABAergic cell subtypes and their synaptic connections in rat frontal cortex. Cereb. Cortex 7, 476–486. doi: 10. 1093/cercor/7.6.476

Kimura, R., Safari, M. S., Mirnajafi-Zadeh, J., Ebina, T., Yanagawa, Y., Sohya, K., et al. (2014). Curtailing effect of awakening on visual responses of cortical neurons by cholinergic activation of inhibitory circuits. J. Neurosci. 34, 10122–10133. doi: 10.1523/JNEUROSCI.0863-14.2014

Klausberger, T., Magill, P. J., Márton, L. F., Roberts, J. D., Cobden, P. M., Buzsáki, G., et al. (2003). Brain-state- and cell-type-specific firing of hippocampal interneurons in vivo. Nature 421, 844–848. doi: 10.1038/nature 01374

Klausberger, T., Márton, L. F., Baude, A., Roberts, J. D. B., Magill, P. J., and Somogyi, P. (2004). Spike timing of dendrite-targeting bistratified cells during hippocampal network oscillations in vivo. Nat. Neurosci. 7, 41–47. doi: 10. 1038/nn1159

Leão, R. N., Mikulovic, S., Leão, K. E., Munguba, H., Gezelius, H., Enjin, A., et al. (2012). OLM interneurons differentially modulate CA3 and entorhinal inputs to hippocampal CA1 neurons. Nat. Neurosci. 15, 1524–1530. doi: 10.1038/nn. 3235

Lee, S., Kruglikov, I., Huang, Z. J., Fishell, G., and Rudy, B. (2013). A disinhibitory circuit mediates motor integration in the somatosensory cortex. Nat. Neurosci. 16, 1662–1670. doi: 10.1038/nn.3544

Letzkus, J. J., Wolff, S. B. E., Meyer, E. M. M., Tovote, P., Courtin, J., Herry, C., et al. (2011). A disinhibitory microcircuit for associative fear learning in the auditory cortex. Nature 480, 331–335. doi: 10.1038/nature10674

Lovett-Barron, M., Kaifosh, P., Kheirbek, M. A., Danielson, N., Zaremba, J. D., Reardon, T. R., et al. (2014). Dendritic inhibition in the hippocampus supports fear learning. Science 343, 857–863. doi: 10.1126/science.1247485

237

Lovett-Barron, M., Turi, G. F., Kaifosh, P., Lee, P. H., Bolze, F., Sun, X.-H., et al. (2012). Regulation of neuronal input transformations by tunable dendritic inhibition. Nat. Neurosci. 15, 423–430, S1– S3. doi: 10.1038/nn.3024

Martina, M., Vida, I., and Jonas, P. (2000). Distal initiation and active propagation of action potentials in interneuron dendrites. Science 287, 295–300. doi: 10. 1126/science.287.5451.295

Miyoshi, G., Hjerling-leffler, J., Karayannis, T., Sousa, V. H., Butt, J. B., Battiste, J., et al. (2010). Genetic fate mapping reveals that the caudal ganglionic eminence produces a large and diverse population of superficial cortical interneurons. J. Neurosci. 30, 1582–1594. doi: 10.1523/JNEUROSCI.4515-09.2010

Morris, G., Arkadir, D., Nevet, A., Vaadia, E., and Bergman, H. (2004). Coincident but distinct messages of midbrain dopamine and striatal tonically active neurons. Neuron 43, 133–143. doi: 10.1016/j.neuron.2004.06.012

Müller, C., and Remy, S. (2014). Dendritic inhibition mediated by O-LM and bistratified interneurons in the hippocampus. Front. Synaptic Neurosci. 6:23. doi: 10.3389/fnsyn.2014.00023

Pangalos, M., Donoso, J. R., Winterer, J., Zivkovic, A. R., Kempter, R., Maier, N., et al. (2013). Recruitment of oriens-lacunosum-moleculare interneurons during hippocampal ripples. Proc. Natl. Acad. Sci. U S A 110, 4398–4403. doi: 10. 1073/pnas.1215496110

Pfeffer, C. K., Xue, M., He, M., Huang, Z. J., and Scanziani, M. (2013). Inhibition of inhibition in visual cortex: the logic of connections between molecularly distinct interneurons. Nat. Neurosci. 16, 1068–1076. doi: 10.1038/nn.3446

Pi, H.-J., Hangya, B., Kvitsiani, D., Sanders, J. I., Huang, Z. J., and Kepecs, A. (2013). Cortical interneurons that specialize in disinhibitory control. Nature 503, 521–524. doi: 10.1038/nature12676

Porter, J. T., Cauli, B., Staiger, J. F., Lambolez, B., Rossier, J., and Audinat, E. (1998). Properties of bipolar VIPergic interneurons and their excitation by pyramidal neurons in the rat neocortex. Eur. J. Neurosci. 10, 3617–3628. doi: 10.1046/j. 1460-9568.1998.00367.x

Royer, S., Zemelman, B. V., Losonczy, A., Kim, J., Chance, F., Magee, J. C., et al. (2012). Control of timing, rate and bursts of hippocampal place cells by dendritic and somatic inhibition. Nat. Neurosci. 15, 769–775. doi: 10.1038/ nn.3077

Rudy, B., Fishell, G., Lee, S., and Hjerling-Leffler, J. (2011). Three groups of interneurons account for nearly 100% of neocortical GABAergic neurons. Dev. Neurobiol. 71, 45–61. doi: 10.1002/dneu.20853

Salesse, C., Mueller, C. L., Chamberland, S., and Topolnik, L. (2011). Agedependent remodelling of inhibitory synapses onto hippocampal CA1 orienslacunosum moleculare interneurons. J. Physiol. 589, 4885–4901. doi: 10. 1113/jphysiol.2011.215244

Staiger, J. F., Masanneck, C., Schleicher, A., and Zuschratter, W. (2004). Calbindincontaining interneurons are a target for VIP-immunoreactive synapses in rat primary somatosensory cortex. J. Comp. Neurol. 468, 179–189. doi: 10.1002/cne. 10953

Sugino, K., Hempel, C. M., Miller, M. N., Hattox, A. M., Shapiro, P., Wu, C., et al. (2006). Molecular taxonomy of major neuronal classes in the adult mouse forebrain. Nat. Neurosci. 9, 99–107. doi: 10.1038/nn1618

238

Tyan, L., Chamberland, S., Magnin, E., Camiré, O., Francavilla, R., David, L. S., et al. (2014). Dendritic inhibition provided by interneuron-specific cells controls the firing rate and timing of the hippocampal feedback inhibitory circuitry. J. Neurosci. 34, 4534–4547. doi: 10.1523/JNEUROSCI.3813-13. 2014

Varga, C., Golshani, P., and Soltesz, I. (2012). Frequency-invariant temporal ordering of interneuronal discharges during hippocampal oscillations in awake mice. Proc. Natl. Acad. Sci. U S A 109, E2726–E2734. doi: 10.1073/pnas. 1210929109

Xu, X., Roby, K. D., and Callaway, E. M. (2010). Immunochemical characterization of inhibitory mouse cortical neurons: three chemically distinct classes of inhibitory cells. J. Comp. Neurol. 518, 389–404. doi: 10.1002/cne.22229

Figures and legends

Figure 1 Properties of IS3 interneurons in the hippocampus and CR/VIP co-expressing interneurons in the neocortex. (A) Two-photon image (maximal projection of a z-stack) of the CA1 area from an acute hippocampal slice (300 µ m) of a VIP-eGFP mouse showing the morphological features of VIP-positive interneurons in the CA1 area. Lower panel represents confocal images showing CR expression by IS3 interneurons in the CA1 area. (B) Anatomical reconstruction (the axon is shown in red, the dendrites are shown in black) of an IS3 cell that was recorded and ﬁlled with

239

biocytin. Inset illustrates representative voltage responses of an IS3 interneuron to positive (50 pA) and negative (−100 pA) current injections (Modiﬁed from Tyan et al., 2014). (C) EM image showing VIP+ boutons of two neurons (b1 and b2) forming symmetrical synaptic contacts (arrows) on the same dendrite which is shown to be immunoreactive for GABA by the accumulation of gold particles (small arrows) (Data are from Acsády et al., 1996). (D) Confocal image of the auditory cortex (ACx) with the morphological features and layer distribution of VIP+ somata; arrow indicates a VIP+ interneuron in the ﬁrst layer. Scale bar, 100 µ m (Data are from Pi et al., 2013). (E) Anatomical reconstruction of a VIP+ interneuron with dendrites shown in white and axon shown in black. The arrow indicates the initiation point of a descending axon arborizing in the sixth cortical layer (Data are from Porter et al., 1998). (F,G) Electrophysiological and molecular properties of CR/VIP coexpressing interneurons in the neocortex. (F) Voltage responses to depolarizing pulses of 50 (lower trace) and 200 (upper trace) pA. The response consists of an initial burst followed by intermittent action potentials at an irregular frequency. (G) Single-cell RT-mPCR analysis showing the expression of ChAT, GAD65 and GAD67 mRNAs in CR/VIP coexpressing neocortical interneurons (Data are from Porter et al., 1998). (H) EM images of symmetric synapses (indicated by arrows) formed by VIP- presynaptic boutons (b 1 and b 2 ) with the soma of CB+ interneuron. Scale bars, 0.5 µ m (Data are from Staiger et al., 2004).

Tables in “Transcriptomic profile of hippocampal long-range VIP-

GABAergic neurons”

Table 1 Gene expression levels of common gene transcripts presented in 7 single cell samples.

gene name TPM c1 TPM c2 TPM c3 TPM c4 TPM c5 TPM c6 TPM c7 1700128F08Ri k 2.2895 38.8093 5.49452 1.83085 17.6447 1.4843 1.28342 2010107E04Ri 493.964 1932.99 1007.33 k 2084 454.703 1 6 851.414 6 4844.63 4930402H24Ri 11.6206 12.0733 k 99.0258 41.1894 6.72574 2 5 3.79855 1.19217 5730455P16Ri 14.8449 k 92.4122 21.81485 8 20.9765 13.6942 1.36517 22.8822 105.709 7-Sep 215.708 125.912 13.2313 30.0545 4 6.53547 273.175 A230046K03Ri 59.2629 k 10.0639 102.885 22.7269 40.565 3 49.7291 30.5165 53.1188 64.3821 Abcc5 2.73535 54.88213 5.04773 2 9.01796 2 12.8169 Acnat1 4.49524 5.04155 5.97431 3.05732 4.54339 1.28307 42.6573 245.969 Acot7 252.426 33.8934 44.541 6 83.0153 78.9742 917.966 Actb 42.0215 193.099 341.533 275.138 112.991 132.592 350.852 742.974 Actg1 244.685 828.944 110.278 297.063 118.822 347.596 5 57.3243 388.189 Adam22 5.41394 68.84776 6 14.5261 11.2347 4.79631 4 Adam23 8.69959 10.3467 24.0941 2.96646 3.71501 1.17946 3.35569 104.3336 Adarb2 9 314.3312 28.3639 11.1408 53.1587 9.4474 283.936

240

Aes 134.814 108.0265 102.827 46.1215 37.4371 74.3955 175.226 33.1422 117.276 Ahi1 50.44049 220.9715 2 47.9755 6 41.9732 327.542 28.3419 45.7052 Akap5 10.4717 3.86078 2 9.87084 6.19346 2 38.9363 531.758 Akr1a1 256.488 205.945 216.457 450.846 3.68416 27.1424 8 1397.05 298.325 1852.11 Aldoa 944.425 422.544 709.696 3 492.988 4 7 2149.57 Aldoc 375.1723 153.7077 134.319 120.476 27.3509 36.463 1 18.7107 144.822 Ank2 13.6173 38.10296 6.94978 4.25261 6 5.94689 2 Anp32a 54.8082 3.17812 56.6695 7.26983 18.3249 2.98238 2.16556 22.0563 Ap2a2 28.3545 66.575 8.75181 12.0838 6 8.92743 2.77285 97.6212 89.2354 28.1788 Ap2m1 358.1 276.904 71.8898 58.0071 3 5 8 Aplp2 20.9328 292.2482 80.3763 62.6996 60.4841 13.7867 9.02888 11139.96 16.0710 3233.79 Apoe 5 355.1117 555.797 430.005 81.0225 3 9 176.712 App 454.088 308.1586 4.60981 180.003 4 96.7553 144.384 56.3595 32.9180 Arglu1 15.2456 177.8933 6 22.6875 7 14.8945 49.5815 52.6764 57.5930 Arhgdia 3.22521 71.481 3 1 96.8052 67.7994 142.505 Arl6ip1 69.3628 258.979 36.7611 129.929 27.893 103.683 542.438 Arl6ip4 1.59646 253.188 8.33043 167.848 6.12567 13.1796 2.04888 471.657 Arpc1a 1.5245 30.7935 10.4001 32.8584 55.3821 18.5416 5 Ash1l 6.92309 19.1493 4.70862 2.02951 3.48246 8.34644 2.89191 14.7139 Atg12 425.519 39.6361 22.5156 1 25.5338 28.8182 813.063 2851.25 Atox1 271.939 122.767 141.157 150.462 43.2715 245.929 8 Atp1a2 144.99 2.32376 22.5619 39.3431 3.62828 5.51855 10.4509 134.469 204.112 Atp1a3 319.9585 394.9501 41.516 6 8 80.0055 9.53835 Atp1b1 468.22 951.64 95.7963 533.929 273.102 112.881 465.806 106.834 Atp2a2 164.3905 86.5987 20.3802 10.9188 83.3159 10.2078 8 Atp5a1 251.911 326.435 115.961 304.572 89.7405 34.422 66.5755 1086.09 Atp5b 392.079 773.82 268.821 400.565 339.444 45.9547 4 230.299 287.267 Atp5c1 198.881 30.6321 9 9 120.068 127.374 547.79 65.1631 Atp5d 361.544 498.0171 366.013 408.727 168.886 182.351 4 Atp5e 168.97 733.681 765.423 1653.77 594.707 544.68 1790.55 690.493 1094.18 1911.88 Atp5g1 452.735 1119.785 4 9 393.737 389.929 8

241

1794.37 282.002 Atp5g3 517.807 1272.202 838.817 4 463.185 2 425.894 876.446 485.589 Atp5h 50.9558 959.283 2 962.414 8 7.54242 678.56 1200.84 574.712 1004.22 1843.18 Atp5j 725.109 632.4362 395.731 8 7 5 2 2203.18 Atp5j2 1016.97 790.663 755.694 952.724 906.9 96.6862 2 3439.75 Atp5k 526.005 574.712 1098.23 827.631 532.315 548.205 3 291.769 Atp5l 190.22 234.279 949.107 1169.61 62.1751 107.678 5 751.028 Atp5o 494.928 386.066 455.169 546.112 342.058 129.042 8 Atp6ap1 22.8121 239.4841 43.2832 3.80729 22.3932 36.614 475.239 327.739 Atp6v0b 456.432 1480.05 89.6714 308.699 651.666 5.35486 5 145.898 178.448 Atp6v0e2 20.1452 337.574 63.6043 1 7 48.3539 4.00699 11.3639 Atp6v1a 65.6245 101.276 17.1563 46.9542 8.7589 3 3.89574 Atp6v1b2 32.9089 262.389 77.6276 4.60399 60.0868 31.7716 1.20634 Atp6v1d 26.5525 408.885 41.5386 266.502 125.257 2.92546 591.678 Atp6v1e1 240.881 101.8 271.342 551.012 492.551 161.713 6.78241 Atp6v1f 1.14047 446.687 407.869 284.052 542.65 20.3589 96.5999 Atp6v1g1 247.526 193.872 70.1523 1.87607 147.581 76.5073 197.847 203.905 195.731 Atp6v1g2 16.7824 656.666 7 357.373 6 27.4574 344.954 38.9404 Atp9a 40.9664 351.0555 1.68395 5 18.5171 24.4081 1.29047 Atpif1 86.8139 136.74 371.403 292.122 172.321 146.586 3622.73 B4galt6 28.1875 5.55276 9.31397 3.60709 16.5073 2.4289 29.3864 Bag1 62.3081 83.9618 212.61 126.145 64.6383 19.6009 239.575 Basp1 269.15 252.867 180.273 45.565 87.3317 140.966 138.859 171.010 121.667 Bcas2 218.075 146.824 5.96406 9 1 30.7369 1.1323 Bex2 363.75 262.438 234.735 176.373 230.896 295.255 1274.92 Bhlhe41 272.99 64.0359 109.457 108.376 62.7002 89.9875 8.93271 575.673 Bloc1s1 277.136 104.725 262.89 1.515 59.8765 76.741 7 Bod1 1.30795 196.665 3.26893 161.775 15.0549 20.5107 233.162 25.5290 18.0182 734.593 Brd8 20.0336 65.45214 12.4527 6 4 3.75467 1 856.9212 557.044 Bsg 5 737.925 27.4273 582.349 168.338 49.5178 3 10.6208 21.8457 60.8873 19.2713 Cadps 37.212 189.7158 2 5 1 1.13908 4 1508.69 Calm1 161.683 522.4731 314.331 330.893 143.858 139.352 9 2840.20 Calm2 633.1013 703.377 565.951 859.497 771.491 317.687 1

242

243.039 Calm3 321.267 148.61 123.697 52.7453 35.7119 48.6524 9 333.729 Calr 261.076 228.287 11.6789 116.295 219.428 19.943 6 20.7144 Camk2a 62.193 9.18919 24.6081 1.81664 3.87965 3 4.54493 45.2840 17.0068 Camk2b 36.88447 22.76942 1.17008 8 16.3168 2 22.2166 51.0116 Camk4 11.9891 23.14029 24.7673 16.1552 6 19.3712 9.48116 227.180 Canx 75.6629 74.422 37.2064 98.568 52.7341 20.9904 2 136.696 431.940 Capzb 194.052 82.948 93.1071 2 76.1916 17.1527 5 27.4802 Casc4 1.47108 29.3955 13.6934 6 29.0813 1.56617 4.79788 33.6849 85.4437 Cbx3 31.22363 15.97881 94.8794 4 24.447 60.649 2 52.0266 112.321 Ccdc136 86.7177 49.9256 6 163.552 2 6.93074 5.56281 78.0685 Ccdc36 59.19672 11.00548 9 52.8564 52.3313 82.8335 111.904 Cct8 107.162 135.711 57.6794 3.19655 55.4843 17.7364 1519.64 Cdip1 97.8673 104.0985 1.78009 33.5769 42.7319 9.2143 4.07871 31.8737 32.8034 Celf4 27.65091 236.2156 7 1 7.60917 50.0653 202.975 Cfdp1 31.4484 58.3924 108.804 5.06395 77.8897 16.2319 447.914 Cfl1 267.74 485.923 425.603 289.917 187.713 99.1539 1328.02 Chchd2 184.729 757.034 379.512 623.131 343.504 425.6 1687.58 Chmp2a 695.912 188.0794 79.3499 20.8791 159.036 184.679 463.53 Churc1 969.638 10.3473 83.9034 10.2901 194.274 5.12949 373.392 Ckap5 70.9257 45.2727 4.52093 2.43058 2.56254 2.67643 7.42911 Ckb 473.152 1109.85 885.39 497.893 226.918 212.477 1033.77 Clstn3 52.1597 66.1141 4.10797 1.34316 52.6658 4.56004 256.613 229.7484 127.072 169.952 Clta 6 141.0501 6 11.3121 6 80.3474 199.526 10.6107 Cltc 17.5687 75.85528 25.8724 9 16.8513 9.61965 2.9164 126.353 Cnbp 74.2898 61.6604 28.1148 210.174 91.3266 4 443.56 Cntn1 1.77981 20.0463 4.81019 12.5155 6.54153 4.06645 2.12234 86.3728 Comt 2.94224 55.74122 70.5187 4.78964 3 1.17072 5.54166 123.213 307.844 92.5373 Copg1 6.60765 162.344 5 2 2 61.1216 583.56 Cox14 254.058 208.671 30.9634 15.295 52.4186 167.796 876.887 1523.52 Cox17 195.823 174.92 346.831 159.461 362.41 296.706 5 438.065 5469.73 Cox4i1 738.44 910.052 9 1085.65 556.589 395.719 7 Cox5a 459.151 567.543 257.435 978.758 363.196 154.132 1.17273

243

1611.55 Cox5b 2030.17 419.344 929.258 719.579 584.35 416.918 2 3526.61 Cox6a1 1533.34 719.001 747.491 1617.86 1381.5 642.328 8 Cox6b1 744.033 1138.31 484.833 1371.16 177.863 263.785 1881.8 4005.01 Cox6c 2165.41 1573.17 865.137 2115.19 713.028 612.88 6 Cox7a2 342.646 655.296 548.477 994.451 464.786 259.176 825.392 Cox7c 196.726 737.582 333.006 1424.93 386.905 344.102 2475.82 Cox8a 3970.39 740.196 1061.92 1193.68 1425.93 1373.7 2505.8 Cpe 2448.93 425.146 257.154 353.071 153.732 82.3798 1614.14 Cplx1 852.161 156.391 214.752 398.085 123.011 58.6417 4.03267 29.4381 Cplx2 71.5984 177.5928 79.4919 26.3542 7 7.4247 49.3502 Cs 36.749 19.9251 14.3847 8.02028 86.1994 1.48462 3.10762 Cst3 14147.8 1458.42 5392.42 2371.92 4290.07 69.9357 10151.7 Ctbp1 142.888 4.55924 31.5369 9.03057 52.3317 14.2453 167.928 195.254 158.379 Ctsb 7.03482 330.772 5 85.3382 56.2634 9 137.866 Ctsd 947.039 294.806 672.844 80.3526 642.686 5.74124 363.643 Cttnbp2 90.0416 104.5254 2.76442 6.06611 15.3979 2.15405 8.41044 Cxx1a 2.027 43.924 24.2301 58.9259 10.3083 105.126 285.063 Cyc1 304.55 818.394 157.535 95.8686 162.342 54.8288 13.6329 Cycs 72.5323 207.977 35.1642 55.3628 43.8115 12.6471 41.4394 Cyfip2 34.9326 84.6805 26.9972 14.902 1.00745 4.42155 213.21 Dad1 4.90491 175.727 61.021 547.3 133.334 10.5349 418.104 Dazap2 489.003 47.0627 71.7235 250.221 60.584 130.432 658.892 45.2004 Dctn1 16.1565 74.7707 9.66166 1 4.5528 13.239 3.1443 138.951 Dctn2 70.2266 141.305 65.8525 11.0468 2 18.2365 80.2609 Ddit3 43.1272 75.6397 15.7687 101.825 61.0383 138.79 447.199 56.2252 Ddx24 69.4791 165.229 62.7547 15.5967 76.2353 46.0953 9 Ddx5 45.88722 160.0992 26.6372 61.5729 92.249 29.5867 7.49473 294.436 Dennd5a 5.66023 43.2036 6.36202 35.0137 7.12153 1.64858 3 Dennd5b 16.9109 30.36818 2.7443 1.91778 15.6356 6.71487 3.77382 Dhx36 68.8327 85.7305 12.7918 69.4033 44.7383 7.89171 1.15005 Dlgap3 1.07287 32.7475 58.7456 9.37909 30.4447 7.18957 115.629 62.4147 451.025 Dnaja1 31.0713 181.8065 32.3545 4 74.2698 19.9536 8 618.311 Dnajc8 279.454 216.407 10.6814 6.65067 170.036 20.2532 5 98.8300 Dnm1 59.44996 137.8071 84.0697 73.4869 5 38.5142 34.4887 23.5203 109.209 44.0751 Dnm3 1.50492 73.07262 3 14.3435 9 4 7.27607

244

113.920 91.1372 10.2919 Dpcd 85.8328 57.79152 1 1 64.0709 3 375.59 13.2262 Dpp10 26.4759 12.2369 2 5.56444 47.2648 1.26399 220.699 Dpy30 31.9069 118.3724 59.3143 1.08863 117.664 16.3117 546.68 Dpysl2 126.576 20.1019 40.8702 54.0607 14.5219 65.5176 78.2006 350.441 Drap1 307.739 42.9098 55.4617 51.7426 82.5079 4.33551 5 Dstn 17.3425 55.8562 20.6151 32.7753 109.97 14.767 127.832 Dusp23 13.4189 5.94859 6.19639 9.44815 4.04822 2.68555 88.878 Dusp3 8.39945 40.5996 37.1455 11.44 1.75519 43.5651 56.4912 29.4742 21.1915 28.7395 Dync1i2 74.8501 99.8585 6 11.2556 29.7548 9 7 100.658 Dynll2 58.554 24.4749 144.383 70.0551 29.7696 29.7273 5 565.924 Dynlrb1 145.109 124.562 1 563.802 246.995 159.216 499.662 Eef1a1 513.31 248.673 478.11 268.24 420.174 119.723 952.999 Eef1a2 66.9822 3.82963 53.0131 36.5476 20.2762 54.7131 6.48614 354.371 Eef1g 234.736 31.6116 62.0931 205.868 50.4356 45.3015 4 Eef2 98.6288 103.529 25.0098 19.2402 3.43471 10.7846 29.1847 Efhd2 202.538 130.583 48.2764 39.0662 40.7422 10.8167 354.483 1240.76 Eif1 955.191 186.758 395.426 341.672 390.3 61.51 4 220.855 381.935 83.2393 567.051 Eif4a2 1.84236 361.3508 31.7427 4 8 7 7 49.2225 191.112 Eif4g2 57.3728 65.0762 6 6.77295 2.80307 2.88182 3 Eif4g3 6.7328 18.50424 7.7839 2.71724 4.06171 1.05633 36.9451 312.793 Eif5 68.7157 96.3578 11.1897 44.2082 106.157 1.31345 5 101.941 198.440 129.235 1466.67 Eif5a 5.88268 250.1631 6 4 6 2.25987 1 223.637 Emc10 32.9306 109.9713 8.9464 7.70078 137.004 13.9256 4 115.182 734.637 Eno1 103.614 112.9819 8 211.419 160.261 131.588 2 Eno1b 109.099 90.8531 70.8584 135.617 79.7454 74.0812 393.815 110.684 126.654 Eno2 1.66186 323.5101 96.9999 5 6 49.0593 8.02018 26.8891 16.4528 Epn1 1.03226 44.81794 6 1.8775 7 4.45537 9.12875 64.8616 Eprs 46.5919 90.45344 1.67134 5.87763 2 11.3984 7.30899 568.529 Erdr1 289.458 1923.19 302.023 161.359 453.873 1.35231 1 192.442 Erh 16.6562 80.4349 31.9949 200.917 6 8.96862 208.819 Exosc4 14.0482 2.98757 11.6875 29.9101 37.92 35.3449 2.80327 Fabp3 41.5359 121.815 16.5453 328.486 124.464 56.0175 884.722

245

Fam213b 141.587 181.948 84.1146 192.495 4.37488 216.331 16.8948 561.690 Fau 263.925 127.87 650.34 381.253 89.3709 3 217.897 Fhl3 38.2967 4.10193 6.18352 7.54572 1.67285 20.1933 7.17614 264.412 492.862 Fkbp2 2.13005 347.4605 8 6 293.107 28.4441 925.639 41.7026 Fn3krp 56.0831 11.2508 51.4283 1 16.1824 16.992 5.09697 Fth1 6002.4 2300.13 3566.42 1123.36 2353.91 549.668 15779.2 Ftl1 1737.43 1273.16 2861.26 718.938 2117.74 375.026 3354.62 G3bp2 18.2253 74.9778 22.7587 2.65748 11.3728 10.6266 25.9279 406.226 Gabarap 15.8271 274.057 330.694 124.002 46.7681 74.3221 7 Gabarapl1 158.617 116.927 69.5759 121.905 105.36 30.9429 15.3034 Gabarapl2 584.519 45.8235 68.0266 153.994 117.421 151.609 866.817 Gabbr1 4.70366 112.3943 11.217 26.074 48.5706 3.05832 77.1251 183.444 Gad1 489.213 402.3984 5 32.8326 111.917 8.98145 136.045 12.8259 Gap43 7.05906 83.3429 4.16785 119.669 215.662 21.3945 5 2292.531 1370.66 2634.17 1183.50 263.198 2390.49 Gapdh 4 1896.977 7 9 1 1 9 Gars 10.9897 74.8803 43.7418 73.5535 145.624 31.9398 3.30002 63.7411 21.3840 Ggnbp2 10.4112 87.0187 4 14.7906 8 9.89869 1.15549 87.0371 130.651 464.954 Glrx2 1.10038 133.1956 3.66484 5 5.18804 9 2 Gm10045 291.04 241.123 89.7646 160.716 205.055 42.8666 1.41111 122.368 Gm14295 107.029 111.4371 53.4947 1.02015 6 1.93174 11.7388 17.3260 Gm14391 14.0559 17.9098 2.3133 21.7529 4 5.73884 50.7926 Gm15772 149.984 149.928 297.936 145.068 70.8305 147.176 927.555 155.946 Gm1673 136.187 110.9686 22.257 4 80.4651 102.141 355.068 Gm1821 623.01 103.852 170.141 277.991 61.9996 82.8838 1389.56 Gm20594 7933.43 1301.03 6367.72 846.554 1793.75 1483.53 12331.9 Gm21596 129.014 78.4344 61.0659 21.7363 34.0349 109.797 87.0126 210.2636 71.2259 183.889 Gnai2 4 104.3684 5 6.48139 46.4109 3.2209 6 737.431 865.350 367.420 75.1941 1202.33 Gnas 742.0341 827.8687 2 9 9 3 5 Gnaz 12.5466 15.869 6.34966 5.56282 31.6765 2.32568 2.85888 105.820 28.6353 Gnb1 146.733 151.5307 48.5283 45.9698 3 4.42342 1 Gnb2 230.538 17.4909 17.1007 36.4742 44.5338 36.7585 312.352 Gng13 187.467 89.3125 55.3479 430.265 85.826 63.0806 365.468 Gng3 287.898 103.704 129.14 151.688 59.8814 27.4725 549.534 Gpi1 171.871 385.604 65.721 4.78757 149.054 61.1153 103.115

246

108.102 12.7808 Gpm6a 312.574 119.897 14.0088 9 59.5969 8 232.478 702.5404 Gpm6b 3 144.1817 8.34006 27.0588 3.99516 4.37517 2.6462 214.136 Gpx1 491.478 151.597 43.6571 84.49 71.2398 65.8608 8 255.426 27.0027 1299.72 Gpx4 859.271 503.661 4 167.619 276.235 9 5 885.160 Grcc10 35.7862 368.736 244.51 599.795 193.749 289.15 4 Gria1 17.61657 21.5147 46.7334 46.46 59.0416 1.28811 1.02516 41.7231 185.697 18.9730 95.9143 Gria2 1.87627 688.6857 3 8.85686 6 5 2 18.2176 67.0894 Grin2b 19.41118 117.6695 3 6.21811 2 4.40665 6.15556 15.9931 120.567 167.019 Grina 139.1858 434.716 8 2 1 1.47089 88.6961 46.2471 Gripap1 35.1402 38.96638 5.76208 27.2262 9 40.7484 8.08817 Grk4 13.7459 260.872 50.2877 8.44352 107.574 23.5607 1.03918 25.2491 Grm1 3.22797 3.10992 2.49267 7.344 2 5.12981 1.51574 19.0155 12.3999 Gsk3b 22.0917 9.10416 9 1.99901 4 16.3455 50.3661 Gstp1 578.989 72.4694 58.8207 448.932 327.671 229.878 334.573 1237.01 H3f3a 173.802 294.232 115.521 227.502 120.265 38.4334 6 Hba-a1 54.2102 6764.84 1430.07 2438.27 626.488 701.329 252.164 Hcfc1r1 254.139 240.923 102.257 148.547 26.7387 40.1919 496.608 Herc1 12.6663 38.16865 4.97265 8.07566 13.4449 2.219 1.39633 Higd2a 90.842 219.122 292.581 292.518 418.181 12.1814 1024.91 Hint1 265.267 488.892 128.732 896.767 408.232 183.888 17.674 27.9232 26.7284 Hk1 12.3749 66.55797 8.99574 6 8.35439 28.2911 3 31.9842 19.0007 285.521 Hmgb1 65.019 102.846 8 21.7231 9 65.4194 3 137.380 114.643 257.101 Hnrnpa2b1 101.9398 518.4441 1 3 192.395 36.9243 6 Hnrnpk 59.00694 227.2649 24.8749 88.7175 46.9049 3.8293 103.075 23.0481 73.0795 Hnrnpl 9.6155 94.0888 8.82323 29.5559 5 28.3566 9 Hnrnpm 8.69438 66.22287 40.6174 6.85221 29.8932 13.2869 45.0575 18.2410 185.148 Hp1bp3 68.22221 19.70166 5 4.27526 5 11.7163 2.14555 1011.88 Hsp90aa1 384.939 188.638 377.189 344.705 264.932 194.164 2 Hsp90ab1 157.1 344.108 180.566 490.499 226.835 110.947 230.437 Hsp90b1 44.8493 288.225 28.3832 8.46171 46.278 2.61227 2.25777 1395.03 Hspa8 254.607 1189.2 278.645 434.341 439.823 161.966 7 Igf2bp2 15.6731 3.43076 10.5026 6.19829 3.40819 6.69853 4.0004

247

36.9002 18.0594 Impact 119.882 83.8711 7.40762 7 55.4231 8 18.1874 Iqsec1 8.45659 24.8143 4.23161 2.32894 5.80692 2.71159 5.0914 503.463 Iscu 1.49453 205.418 343.716 180.273 180.658 42.2307 9 Itm2b 915.591 877.907 371.839 330.359 810.351 5.24456 451.206 21.4840 Jkamp 22.63141 279.0722 97.3279 7.43557 31.1359 6.63844 1 Junb 1.62022 68.2958 3.9826 97.2469 29.3701 20.82 284.916 Jund 38.652 31.0938 26.9191 34.3511 12.7542 7.55526 12.4818 203.204 Khdrbs3 347.0128 84.8891 14.7718 26.1974 8 59.8799 282.649 384.426 Kif1bp 10.0901 58.8329 55.6666 134.881 76.0576 4.8811 7 Kif5a 146.357 90.6763 41.9207 15.1223 68.3121 10.3407 79.3885 Kif5b 27.7188 28.3319 6.99869 3.14062 23.9599 31.7932 129.616 Kif5c 36.8112 95.6438 2.39477 42.295 52.6576 3.13122 144.508 Klf13 15.2122 21.978 5.53667 7.46597 6.68501 10.2552 1.90945 60.6847 Ktn1 3.5306 70.74513 2.45667 23.1094 1 2.57007 6.5741 Lars2 1454.38 1466.68 5181.63 4274.14 4096.6 4380.47 204.101 Ldhb 200.898 228.687 180.149 271.166 90.5373 80.7767 967.977 Lrrc4b 15.7545 2.54158 7.04031 1.37534 6.49692 1.74552 3.71397 83.6924 12.5665 Luc7l2 4.92961 174.2765 37.1238 2.11542 9 6 59.9371 Maged1 71.0278 668.863 40.677 27.2881 291.265 24.8832 5.19372 50.0169 Magi2 9.38561 3.98455 9.08367 9.10294 2 2.66816 8.79216 19.1174 Map1b 58.4201 53.0534 67.2641 61.9613 70.7821 4 76.0687 105.823 325.732 Map1lc3a 59.0428 161.2591 6 124.198 79.7064 141.459 3 24.6867 67.0346 535.810 Map2 5.92433 105.045 5 1.18769 7 7.88853 6 84.7354 39.8451 Map7d2 56.6912 72.08298 7 66.3245 58.299 5 66.1317 37.1887 Mapt 5.41165 67.0251 44.0292 7.83619 5 20.4799 16.5487 62.1580 90.8796 16128.9 Mbp 683.2467 163.0631 2 194.036 873.201 9 2 Mdh1 232.405 223.324 101.447 226.687 166.848 74.7949 56.4663 Mif 138.556 209.394 664.586 1040.87 270.722 529.828 2176.09 54.9032 56.4520 Mllt11 97.7004 285.3691 8 272.765 1 23.3479 587.692 Mob4 2.68382 42.051 6.70352 60.3052 7.39169 11.4644 9.87254 Morf4l2 88.0887 93.8291 41.9175 114.132 115.951 39.984 2.52159 17.4694 Mpnd 221.472 127.536 2.89286 1.92737 69.7348 6.58493 7 Mrfap1 458.747 204 86.6517 211.729 158.743 7.85838 353.045

248

Mrpl19 9.2903 1.51493 6.75978 2.33801 18.759 2.46962 2.15848 Mt1 1534.925 1977.631 211.094 371.817 81.9643 180.948 4425.08 Mt3 1620.48 862.382 486.164 399.974 125.541 505.728 1855.64 61.9263 227.816 Mtch1 288.6261 99.7553 5 4 288.796 5.80932 54.3991 50.3940 Myef2 90.2675 73.091 3.72635 8 24.993 2.04317 1.88062 565.455 1162.21 Myeov2 5.40594 612.36 2 589.907 394.93 389.835 3 Myl12b 199.16 196.706 157.71 370.805 331.705 104.38 316.823 583.340 779.983 217.482 Myl6 489.097 711.3561 3 9 408.401 2 1728.75 Myo5a 4.88894 93.4843 1.22145 5.31474 27.308 2.91415 5.22138 37.3234 43.5652 Myt1l 15.56598 69.23631 3 40.8078 4 1.12613 48.7897 527.317 Naa20 1.22153 234.959 2.38957 68.3728 5.75656 96.1203 2 141.722 Nagk 14.4376 77.20888 4.4017 9 35.4488 94.9209 37.0826 114.356 110.080 Nap1l1 182.462 202.7449 10.0104 6 66.1944 6 2.45614 Nap1l5 898.974 2029.75 333.115 347.505 379.978 254.902 1873.07 32.6422 Ncam1 55.90497 20.0051 3.52788 7.421 7 6.87345 27.604 Ncdn 64.9304 36.16657 22.7518 73.2025 38.2935 21.2002 7.47579 12.6726 368.235 Ncl 59.3461 36.9057 4.32715 10.6509 40.1466 4 1 Ndfip1 109.205 1174.89 17.4494 728.757 297.115 57.5347 1380.87 343.210 118.370 Ndrg2 267.494 145.0794 6 8 93.3143 8.42782 790.317 80.8081 415.154 246.280 Ndrg4 205.919 727.4455 4 494.259 9 43.1573 3 Ndufa1 642.71 335.633 171.061 376.263 170.288 25.3645 1091.21 Ndufa10 5.63374 191.007 91.7056 74.3418 2.21078 14.54 61.814 Ndufa11 39.4181 154.07 148.942 146.153 67.3571 66.7209 289.086 Ndufa12 417.959 81.4181 262.238 785.566 168.938 12.3368 280.727 989.8053 Ndufa13 1 716.9459 358.024 289.636 490.03 402.864 743.428 Ndufa2 374.738 73.8653 174.19 484.428 328.878 249.506 494.525 3881.17 Ndufa3 968.291 913.435 924.629 764.695 512.211 556.83 8 Ndufa4 675.102 709.85 425.777 1464.51 313.28 243.244 3899.25 2534.72 Ndufa5 313.962 345.4157 293.453 489.06 286.113 439.314 3 Ndufa6 224.365 603.347 375.653 143.315 444.808 182.752 1721.58 2985.47 Ndufab1 2.14252 146.448 153.384 298.35 249.043 25.7089 5 1420.65 Ndufb10 32.7112 167.736 192.91 783.733 304.55 33.9546 7 Ndufb11 920.851 277.789 76.161 209.304 71.8874 2.76601 179.466

249

309.538 406.572 Ndufb2 343.843 480.038 539.103 1258.04 6 8 953.354 Ndufb3 135.48 140.424 256.442 523.493 216.111 99.5302 1511.04 278.344 Ndufb4 259.698 174.622 201.563 70.422 221.67 221.208 2 222.918 154.187 1289.57 Ndufb5 98.13372 597.284 403.49 337.656 1 1 2 Ndufb6 94.5277 214.609 184.578 353.904 128.58 160.261 355.893 Ndufb7 561.79 456.471 16.546 340.352 175.163 99.2154 426.478 654.026 498.693 1911.99 Ndufb8 305.247 581.9739 266.833 2 4 388.534 8 Ndufb9 19.8651 424.914 280.146 830.634 365.431 216.851 575.154 2557.97 Ndufc1 4.19814 466.719 778.145 1387.97 360.553 184.191 4 Ndufc2 376.497 327.669 281.24 239.977 262.861 97.7585 1262 Ndufs5 254.456 184.093 252.363 567.347 564.203 480.091 1860.19 Ndufs6 1.47505 655.857 299.417 412.882 2.12469 421.536 538.401 277.609 374.725 142.597 16.4263 625.178 Ndufs7 383.502 232.0536 6 8 1 2 5 Ndufv2 8.28152 89.2138 122.115 261.666 124.256 44.2777 452.673 389.060 Ndufv3 682.639 108.25 703.788 71.796 324.742 366.521 7 568.832 Nedd8 2.44927 202.63 785.092 25.4258 470.834 182.14 3 Negr1 1.77951 9.25078 5.15433 1.35668 2.10455 3.21928 248.945 Nfe2l1 16.9755 8.59946 3.90336 79.2403 2.52907 42.4577 7.0468 163.849 1622.35 Ngfrap1 642.595 389.576 89.1329 1 367.418 52.0974 3 363.804 Nme1 167.479 112.151 211.878 412.505 130.682 20.113 8 Nnat 67.1086 52.2837 84.3095 14.6484 21.6299 58.4431 40.2872 Nrgn 18.7253 119.21 90.9197 10.4801 14.853 26.2876 458.157 Nrip3 28.3097 267.529 33.0008 5.06935 99.8236 6.71932 150.106 105.677 Nrsn1 190.2769 68.7086 5 15.9434 109.608 88.736 25.1587 251.8593 Nsfl1c 7 27.197 51.7701 13.3758 52.5499 16.6809 2.12999 53.3659 196.886 87.1578 Ntm 144.371 241.1283 8 1 8 7.33849 124.653 290.529 1071.40 Oaz1 392.7272 478.3003 263.449 648.196 3 70.394 4 Ogfrl1 4.62432 42.533 21.2704 22.1528 1.97225 13.2543 3.29 134.753 151.392 Olfm1 125.3718 485.186 475.278 15.4101 1 82.5413 8 Otub1 11.4817 76.8067 99.1676 189.125 56.5253 34.5556 57.9507 204.141 Pabpn1 3.33356 510.685 18.3166 66.8042 3 3.06507 5.41959 22.0155 Pafah1b1 8.61005 33.0561 30.7447 2.18856 9.25233 3.16345 5

250

40.2578 Park2 83.092 14.66442 9 142.955 176.878 8.92105 466.684 995.146 Park7 248.795 217.917 68.909 100.672 91.5654 139.666 1 Pcbp1 1.05285 197.296 118.806 45.5762 95.4285 42.165 137.965 38.9160 Pcmt1 216.4862 191.2045 28.3096 56.4707 9 3.64804 382.599 Pcnp 1.53868 33.3267 98.7305 14.272 96.4735 27.1507 91.7276 Pcp4 367.433 21.6644 1173.28 241.933 1027.89 585.736 1712.57 Pcp4l1 298.851 18.9569 76.0881 198.539 56.1668 45.7806 746.728 Pcsk1n 67.5362 399.993 184.436 261.139 89.8904 130.347 237.084 Pdcd4 36.38 77.2974 4.74171 6.85747 13.227 26.4432 10.5974 1005.57 Pea15a 513.36 111.3663 48.6964 98.7531 69.1054 35.1493 3 777.711 Pebp1 232.073 176.51 88.2814 342.014 86.9977 7.17117 9 45.8266 177.315 Pfdn2 45.4564 212.6674 129.589 83.0435 1 6 34.9673 570.624 Pfdn5 366.258 68.3987 173.324 86.4584 225.072 71.3616 1 116.601 123.238 Pfdn6 1.48302 286.437 6 6 93.993 30.0942 1.20411 105.490 139.028 128.019 Pfkp 5.57589 224.0996 6 9 6 18.6651 3.03191 1217.88 Pfn1 183.338 88.3511 151.896 91.8639 35.1415 65.2339 2 Pgrmc1 64.512 635.238 82.3072 210.515 21.7313 6.68663 436.919 58.7157 Phf20l1 27.711 248.7622 4.37033 6.85132 3 2.64348 121.628 33.8410 103.532 168.778 Phyhipl 141.548 63.5968 34.8992 3 7 5.84935 5 351.789 Pitpna 3.896 116.812 39.4967 44.9657 51.6247 44.2576 4 300.481 Pja2 24.3708 28.1027 51.7857 9.27344 23.1911 21.4566 9 266.212 307.176 103.570 47.9394 Pkm 414.58 529.491 9 4 7 9 909.727 426.138 Pld3 44.8796 580.4499 14.0883 10.5353 109.045 11.2359 4 Plp1 8.75239 5.20404 7.51243 23.1997 95.9278 18.907 1751.41 61.0950 66.7958 Pnisr 20.8601 316.3481 1 67.296 8 1.01663 2.76492 Pnrc1 3.67039 26.1764 33.4243 10.3473 9.02069 3.17085 7.70833 31.9643 Pola2 18.7389 84.0875 31.3035 61.25 150.562 4 171.293 Poldip2 14.8955 134.74 4.77541 5.97126 1.312 7.40608 12.2309 Polr2l 1.842 87.5097 23.1192 46.7004 71.616 38.2353 8.42167 Pomp 780.849 169.75 280.086 92.8846 360.503 34.8295 12.8497 3274.24 Ppia 1962.18 2014.48 1980.98 2503.33 2588.96 835.16 6 Ppib 211.694 358.269 132.529 70.9854 183.817 54.3003 428.88

251

Ppp1r16b 5.132 20.51116 7.18354 6.95919 43.8484 3.09049 111.867 Ppp1r1a 86.2681 39.8887 36.3685 46.3256 80.4942 13.1209 501.72 Ppp2ca 91.3546 287.086 22.8981 91.6372 141.605 19.7695 87.3849 64.1603 Ppp2r2d 41.9936 123.2086 14.7331 5.83811 6 23.9067 6.73102 13.7037 Ppp3ca 21.5522 95.0622 40.109 2 19.2015 6.58993 6.73151 Ppp3r1 63.8785 113.58 60.3095 159.9 119.929 52.8962 4.9581 214.7255 1932.92 Prdx2 3 430.81 215.723 423.024 185.805 187.497 1 Prdx5 709.702 461.097 299.117 322.985 184.037 77.1441 289.43 Prepl 18.4154 167.9822 36.193 52.8848 40.5451 33.1089 240.504 16.1928 178.151 125.435 Prkar1a 81.9554 128.4487 3 3 3 28.0094 599.344 46.3766 153.349 12.3931 177.443 Prkar1b 188.5014 146.459 2 8 83.3199 8 8 14.2914 142.134 Prnp 79.0934 263.383 4 109.425 3 34.2619 61.0465 Prr15 21.9062 5.42413 5.26768 8.26726 3.8468 6.85275 32.6779 19.2663 Prrc2c 30.39623 51.09161 2.51322 1.49062 5 2.42766 4.88013 85.2473 29.4307 61.5194 185.374 Psip1 77.98772 111.568 3 4 14.0939 3 7 Psmb3 66.5196 684.901 46.2788 101.67 28.033 90.4377 7.29706 Psmb4 2.06622 62.5173 17.772 102.245 22.3001 9.56541 461.496 1229.94 Psmb6 207.276 201.535 190.237 285.834 244.239 73.1957 4 Psmd7 182.419 63.4204 103.801 86.0759 57.2765 5.8048 2.16651 13.4572 Ptges3 66.1147 62.57254 16.8321 56.8554 4 13.233 343.179 876.668 Ptma 299.341 212.351 204.829 200.736 201.33 17.9486 8 194.199 Ptms 25.6881 109.25 23.8025 90.0734 25.9526 9.30349 9 Qpct 3.10747 86.3743 68.8715 1.25157 45.6035 4.50958 393.493 60.6853 139.404 87.9513 Rab2a 270.3138 86.6597 5 1 43.9765 4 159.055 Rabac1 227.168 386.333 89.1531 210.658 185.321 69.8524 396.419 13.0718 22.5616 Rabgap1l 3.79329 8.45733 1 5.96801 3 1.35655 1.61399 140.657 Rap1gds1 84.8392 116.759 6 21.8476 2.54527 13.1289 14.453 157.3881 47.9689 Rbfox3 2 35.89792 5 13.2186 6.35011 1.54138 4.46817 Rbm45 4.37038 1.00821 3.41099 2.72532 7.41025 1.94148 61.2046 41.1842 Rbm4b 22.1123 2.85437 7.025 8.50935 1 4.28749 10.7001 11.8251 224.937 Rbms3 1.72931 120.4348 4 2.31947 27.8721 1.9736 3 1882.58 Rnasek 186.381 822.143 199.258 892.626 521.478 179.869 5

252

Rnf187 26.7567 160.919 106.986 83.0037 268.805 93.1977 366.308 Rnf214 5.3168 13.18606 1.80949 1.28236 12.3716 6.11467 14.1789 330.203 492.150 Romo1 3.48017 93.9814 8 266.605 206.904 72.1628 3 44.6143 29.7453 29.2298 Rph3a 35.93756 126.0516 5 20.7941 8 7 3.83975 Rpl10 336.012 140.206 370.07 34.373 121.833 257.778 464.871 Rpl11 775.9128 511.1467 184.128 54.5088 178.052 11.8306 534.958 Rpl12 622.565 265.642 228.654 361.502 76.2799 168.261 18.3262 224.241 1302.37 Rpl13 962.541 139.73 612.487 446.854 5 162.848 3 46.5081 Rpl13a 402.008 102.176 323.671 140.365 71.1977 280.312 4 Rpl14 658.369 105.706 174.697 280.881 156.895 291.335 410.31 227.901 333.239 197.112 1495.85 Rpl15 248.3337 190.6381 1.14056 2 2 9 9 292.222 557.632 230.094 Rpl17 154.869 373.8442 311.262 3 3 71.1099 4 749.071 Rpl18a 298.209 232.618 90.7514 118.109 370.249 141.605 7 721.557 Rpl19 39.755 65.2785 594.774 553.092 133.855 116.088 2 22.7205 Rpl21 406.3038 86.57 37.7429 58.2328 88.401 1.35062 8 Rpl23 138.63 11.0435 49.3961 6.49605 17.9889 8.04332 25.2102 398.656 Rpl24 1380.72 129.961 537.918 550.478 33.3855 289.177 3 910.243 Rpl26 456.057 88.6545 247.85 26.1975 576.346 71.8855 6 124.872 358.138 316.870 1035.84 Rpl28 394.5976 283.222 4 328.102 3 7 8 187.289 Rpl3 49.368 287.0084 244.182 4 293.745 265.599 437.657 119.246 1460.30 Rpl31 167.936 250.7135 9 86.9641 226.811 264.817 5 Rpl32 573.737 714.915 873.47 5.14381 517.04 217.654 16.3871 209.677 930.335 103.254 138.134 1224.76 Rpl34 436.2202 214.4959 6 9 9 1 6 Rpl34-ps1 406.831 200.053 180.722 843.63 88.3749 111.451 988.486 Rpl35 1488.32 415.455 770.789 541.439 408.114 212.624 455.258 139.756 522.231 Rpl35a 127.312 454.155 374.828 1 87.0625 306.274 4 919.744 762.498 421.577 Rpl36 191.4984 178.0254 599.263 6 7 564.083 3 Rpl37 406.866 22.0472 259.254 175.297 278.149 203.992 1077.77 Rpl37a 375.567 456.968 794.275 756.429 533.754 310.068 786.087 648.023 302.335 196.211 1758.75 Rpl38 1581.268 388.5801 378.123 8 4 6 3 Rpl4 52.6011 184.72 194.273 152.925 140.197 92.2883 1032.42 3486.90 Rpl41 1329.12 1011.87 1734.64 300.908 1127.82 587.425 4

253

Rpl8 208.138 112.343 114.916 92.066 86.8545 302.364 6.58137 688.669 Rpl9 157.054 137.536 209.311 297.745 575.171 613.159 7 Rplp0 404.873 234.975 190.625 193.468 212.987 5.38904 518.735 Rplp1 2756.39 92.8499 1487.35 336.237 815.174 476.693 480.243 771.512 822.425 606.628 289.978 638.429 Rplp2 740.4262 456.4224 8 6 2 1 2 Rps11 486.18 154.718 594.392 286.374 559.005 71.1093 805.027 340.387 25.5738 Rps13 1.27305 669.57 2 23.6516 374.001 82.2866 8 1951.21 Rps14 802.306 270.055 797.009 594.128 534.797 401.303 9 426.798 1094.01 Rps15 803.687 327.712 264.501 449.989 5 487.85 9 425.087 Rps16 641.029 94.1936 245.732 115.1 131.901 207.345 4 2142.02 Rps17 415.883 183.385 986.76 195.234 318.312 375.079 2 343.840 Rps18 394.482 175.689 336.849 63.2408 126.036 379.186 8 659.782 254.862 372.453 Rps19 669.826 490.1923 3 3 508.751 313.889 3 1522.99 Rps2 130.173 330.223 203.599 99.0305 261.632 57.8986 5 297.426 Rps20 124.271 14.1371 22.4343 33.0353 31.0534 21.473 5 Rps21 637.182 271.49 316.066 373.72 105.612 359.367 51.8067 Rps23 1046.33 501.047 764.129 328.57 132.617 385.029 24.2528 318.265 232.499 214.184 Rps24 42.2372 61.47421 305.374 4 9 2 14.6829 Rps25 734.807 306.743 238.296 226.458 264.655 316.623 782.545 Rps26 99.426 132.008 295.602 712.537 95.1415 35.6353 318.478 628.424 Rps27a 342.382 69.3316 143.281 367.418 266.233 300.135 2 Rps29 1781.67 1003.62 1083.11 1498.44 864.637 1012.62 1995.5 Rps3a1 319.271 104.593 49.3633 131.245 64.5188 83.1775 96.0856 Rps6 2.81585 6.79922 7.08446 2.0481 17.5518 103.796 1.22316 Rps7 512.911 290.844 154.358 165.888 135.266 155.33 101.305 Rps8 384.495 94.2547 216.689 214.498 190.512 161.853 88.5954 591.119 433.983 Rps9 2025.871 241.7273 3 78.5259 2 68.4277 476.448 43.6476 Rpsa 329.331 70.811 45.0196 49.9449 50.8791 45.5648 6 332.117 Rsrp1 54.1945 1416.27 13.299 263.549 425.74 78.6761 4 354.022 1230.59 Rtn1 834.508 585.1538 99.2686 347.984 2 88.0644 4 146.071 Rtn3 72.1466 382.2644 53.5797 157.801 2 15.2119 193.742 24.3205 114.611 Rufy3 1.08642 62.67397 10.3464 2 1 7.62004 82.9111

254

S100b 3.23136 68.4466 89.9218 9.40372 1.54966 50.1941 512.737 Scg5 856.403 1253.65 168.33 684.863 621.992 7.6311 976.2 70.3003 50.7435 113.741 Sec62 3.62945 127.7364 54.9114 4 7 9.77487 3 Selm 139.547 302.559 163.525 140.111 93.4271 108.885 551.364 Sepw1 1310.95 647.256 1120.3 985.452 452.872 339.256 3115.9 Serbp1 11.6684 29.52377 8.14983 15.418 7.32281 4.93858 5.90568 Serinc1 409.383 388.782 12.699 159.001 185.202 22.1887 145.206 170.327 35.5800 45.4365 Serinc3 140.8017 190.4327 9 1 2 3.77465 3.1401 454.071 Serpine2 1.46443 36.3391 35.4135 41.5455 133.833 77.142 9 212.243 Set 83.5496 56.5406 12.6851 1.40512 20.2234 5.32687 9 21.2357 Sgip1 26.63263 63.99911 7 16.8586 37.0994 16.86 27.497 537.581 Sh3bgrl3 392.627 191.774 147.149 167.877 106.461 2.96809 6 Shank1 18.26977 44.90077 29.8663 5.52049 17.5016 3.51193 21.562 Skp1a 192.966 138.335 32.0802 6.42493 43.8643 25.0083 7.78128 129.620 129.123 Slc22a17 351.366 162.7125 39.5946 6 9 24.7607 2.58568 671.176 Slc25a3 156.297 942.97 175.955 387.23 276.835 46.2169 9 Slc25a4 577.031 805.828 466.08 505.271 251.798 414.413 1660.43 Slc25a5 8.46357 306.086 267.475 236.565 113.996 25.0199 747.496 102.599 Slc3a2 215.0029 437.0108 9.87852 1.86975 4 1.9904 2.15285 Slc6a1 36.8239 116.185 53.4872 140.851 168.43 26.6539 3.5213 Slx1b 5.56904 5.77887 6.32932 6.54826 30.5641 5.04994 11.7803 19.6563 39.8603 Smarcc2 55.56588 88.855 8 9 35.5729 56.8668 3.98615 Smdt1 638.058 481.569 303.904 323.493 274.05 317.682 484.341 503.658 546.454 376.333 193.945 Snap25 240.116 491.375 7 2 7 7 844.103 270.094 179.759 100.242 112.024 Snap47 78.3845 251.0143 4 8 8 25.6382 3 Snca 4.75529 236.122 87.9717 251.184 93.7361 4.87866 95.1045 194.7562 47.5566 100.601 54.5194 Sncb 3 111.6804 4 99.5869 3 2 82.2 1695.07 219.607 Snhg11 39.0999 3001.876 380.902 426.017 2 38.9765 1 Snrnp27 40.1679 65.1831 95.2965 5.2621 109.254 31.7451 468.626 Snrpd1 66.5806 94.3753 28.8193 6.13008 210.106 146.557 290.735 Snrpd3 2.43832 185.814 4.9315 142.546 76.3654 116.756 10.2099 268.626 52.6327 Snrpn 112.877 402.1896 5 264.706 5 58.4541 2122.09 Sparcl1 318.7862 398.896 58.2629 72.8772 40.9141 22.1171 403.153

255

1154.86 Spcs1 134.325 385.892 81.8578 189.097 142.02 103.268 5 12.4740 Spop 10.2058 24.53287 27.6534 3.65065 47.2754 9 8.44674 Spry2 19.5498 3.87085 23.6535 9.67938 5.92597 5.95408 11.0755 14.7533 Sptbn1 3.35514 21.70483 4.51772 8.58533 3 2.62707 1.24611 18.2426 Sptbn2 39.4273 1.13449 1 1.64858 4.91172 3.96031 4.11382 1080.97 Srp14 6.03237 218.976 166.333 115.713 395.611 43.7253 8 180.486 112.435 Srsf2 1.79175 688.488 5.33271 110.627 9 2.64168 3 St3gal5 26.7069 78.2123 55.1446 6.22582 2.13774 10.1254 2.93307 75.7638 Stmn1 255.117 362.81 125.121 791.202 176.244 1 1205.66 Stmn2 41.2065 586.021 77.6 257.883 36.3776 44.6539 72.6509 Stmn3 747.72 692.198 460.574 1267.28 684.667 181.022 486.276 29.9499 Strn3 43.7102 46.5901 32.0153 10.4583 8 1.35932 1.86497 115.296 226.511 Stxbp1 20.0981 153.1445 24.8088 1 40.3607 13.2741 1 12.3427 97.5222 Stxbp2 8.33461 2.19975 9 10.4507 9 1.82642 8.10084 30.3127 1024.64 Sub1 12.0286 138.8081 82.8018 66.9509 31.8173 7 7 Sumo1 35.5836 242.624 122.177 93.7596 202.971 8.65963 400.609 Sv2a 40.1395 110.854 14.8823 1.10686 119.232 3.7679 4.99004 347.723 142.024 59.3314 Swi5 80.9731 214.5217 1 75.9964 7 9 152.197 11.7589 163.287 Syn2 74.1045 211.2256 1 30.8163 4 9.3128 2.00164 23.8137 106.193 14.3653 Syne1 14.7025 212.2412 2 30.3774 6 2 29.7364 Syt1 279.632 176.1305 54.1113 76.122 104.722 45.5118 7.19213 72.0565 51.8152 Syt11 178.745 259.0783 21.0364 19.6677 69.5013 7 9 Tagln3 622.448 48.0043 55.7876 146.678 2.47015 137.54 716.859 12.0387 Tanc2 23.5932 41.7694 3.1544 9 11.6814 1.28475 3.49387 Tbca 372.575 177.319 56.9885 362.803 563.845 105.946 20.0005 21.9645 20.0059 Tcea1 3.1271 185.6274 2 17.1555 3 3.70146 276.182 Tceal3 283.0223 117.6296 68.7784 13.6613 4.7517 22.0763 1.21754 Tceb2 1695.57 509.23 752.453 1466.32 419.668 295.379 2674.96 166.021 104.021 119.356 42.2755 Tcf25 89.2663 409.8473 5 5 3 9 2.85667 38.3969 20.6837 136.293 59.0558 109.809 Tcf4 26.92741 329.7528 1 3 2 2 2 57.0745 36.7161 147.506 Thoc7 4.57862 159.811 38.6348 8 8 5 55.8534

256

Timm17b 135.063 3.30185 7.6731 51.3399 7.68837 6.05183 2.35554 194.668 Timm23 127.003 103.214 179.132 158.159 50.7624 91.2738 4 Timm8b 483.98 587.685 297.607 492.204 389.887 494.932 1023.87 1326.71 Tma7 583.757 371.749 305.042 148.878 506.066 111.807 8 13.9136 Tmem136 13.1229 50.69533 9 1.22962 3.0013 1.17098 1.58137 3024.512 417.313 687.106 Tmem170 2 228.547 2 6 149.383 32.9334 4.18479 33.1750 488.079 Tmem234 93.2165 190.7317 24.6184 8 41.0295 7.16756 2 1809.53 Tmem256 157.853 341.554 590.784 261.809 93.9675 26.8968 7 133.679 Tmem50a 196.458 11.971 13.6361 9.30801 8 17.0916 392.937 Tmod2 6.70094 37.2015 6.12627 11.4069 26.7523 4.34963 61.7873 837.588 337.559 811.900 Tmsb10 949.059 554.5694 362.154 8 336.507 5 4 7870.661 7864.74 2250.91 3963.66 Tmsb4x 4 3490.609 7 9 4 827.802 6335.14 Tomm20 6.88792 177.349 36.9732 50.6401 27.0437 7.99129 185.537 154.012 Tomm6 197.441 103.042 17.2753 1 122.214 68.5368 326.702 Tomm7 139.604 95.6811 45.5714 65.7384 50.6249 50.7971 167.529 Tpi1 232.659 293.861 155.378 664.438 195.895 162.256 682.508 24.2434 46.7646 468.602 Tpm1 54.7583 96.204 6 2.99096 7 9.10422 1 152.961 160.824 147.954 43.8884 21.0783 Tpm3 187.163 148.2081 9 9 5 2 4 623.824 Tpt1 573.47 201.388 164.124 407.63 97.2601 160.855 2 21.0071 Trim2 19.52518 69.5623 7.09336 4 27.1508 8.61956 26.6237 67.5674 Tsc22d1 2.06124 144.352 9 132.384 14.2814 4.69007 297.091 Tspan3 220.208 791.199 2.10719 32.94 115.967 1.24361 618.786 600.965 Tspan7 368.476 156.336 116.663 134.038 39.8549 37.0018 9 19.1358 Tspyl2 8.32647 27.669 10.5232 27.4924 91.2062 9 1.25156 Tspyl4 40.9075 69.6938 24.7309 35.4571 7.40776 21.198 3.48843 111.000 225.230 107.971 56.7853 808.696 Ttc3 210.7973 321.3469 7 6 2 9 7 18.3285 Ttf1 1.05637 39.74955 4.34228 1.13999 8 1.19934 10.618 Ttr 13.1311 38.812 36.7955 276.147 62.4948 2.34656 13.7618 Ttyh1 234.275 64.3546 93.3819 45.7571 71.4194 14.42 13.0561 Tuba1a 469.631 1252.16 384.261 555.631 372.935 203.705 690.232 166.603 Tuba1b 90.8169 632.05 418.816 720.784 1 143.653 82.4962

257

104.023 Tuba4a 30.399 127.9082 65.4023 62.6982 7.43303 9.29699 1 Tufm 18.8195 14.9006 8.98766 4.92911 52.5266 5.45493 562.952 207.168 Uba1 24.88841 11.40781 1.59524 3.2364 6.29444 3.83835 7 2506.20 Uba52 863.115 723.05 412.59 770.241 434.717 593.607 6 11.0300 57.5040 Ubap2l 1.12396 30.03512 1 20.7353 1 8.09689 9.01431 1346.91 112.264 Ubb 927.272 1668.751 548.873 7 592.693 9 3619.88 72.0435 Ubc 206.874 431.852 159.524 379.313 193.963 17.854 5 370.672 200.833 Ube2m 182.2608 202.0214 61.7059 8 90.6454 84.9475 5 211.984 514.622 389.974 241.784 1856.60 Ubl5 1.37184 273.8005 1 5 8 3 5 15.2187 Ubxn4 23.7966 66.3806 10.7552 9 6.54292 38.1169 9.31507 Uchl1 471.206 656.602 406.098 861.693 720.041 131.246 475.461 Uqcc2 1739.17 71.7096 447.099 364.826 162.214 505.917 1064.41 Uqcr10 26.4181 228.637 547.508 1298.6 386.598 110.567 466.974 4738.22 Uqcr11 812.374 734.663 1449.68 1322.46 345.619 365.252 4 Uqcrb 639.823 490.839 117.143 1305.06 405.334 322.617 1012.29 239.883 Uqcrc1 45.9308 267.319 13.327 213.523 72.0647 40.2589 4 Uqcrh 264.951 788.001 198.239 1040.91 425.177 680.314 394.094 5824.06 Uqcrq 45.9114 177.197 196.589 337.632 130.495 33.8274 5 Usmg5 1306.34 1138.71 1299.29 3615.98 2197.84 870.668 7386.62 Vamp2 311.12 310.757 237.854 200.446 53.5535 19.168 712.868 Vcp 4.14415 116.397 48.2154 9.23267 5.09203 17.7697 124.804 811.288 Vdac2 1.08714 200.232 93.2268 144.346 52.6574 125.706 4 323.013 577.501 Vip 678.386 4299.632 7 24.6098 9 319.041 5006.11 Vsnl1 193.03 48.3639 77.6497 153.516 65.0664 9.16669 204.702 25.9926 Vstm2a 69.5049 382.6117 7 82.3478 78.7921 16.416 236.62 Wdr61 12.2572 30.8054 13.2133 79.8221 49.2353 34.806 502.748 21.6945 23.0810 36.1766 Ypel5 2.86199 135.5952 7 54.0765 1 1 170.169 137.2834 90.6860 100.448 Ywhab 3 135.05 3 85.6909 1 27.8028 209.895 Ywhae 83.1792 211.738 83.4014 25.0936 123.544 54.0284 488.051 Ywhag 79.6394 63.6626 44.1444 145.119 10.5424 40.9086 11.5417 Ywhah 15.8749 385.936 111.995 263.471 183.425 153.866 378.068 18.3580 419.173 981.000 Ywhaz 10.5812 428.641 3 2 130.466 34.0661 6

258

13.5035 Zc3h13 1.61885 76.56147 24.3604 6.20207 5 8.57592 1.38874 200.250 Zcchc18 80.656 409.9553 39.4255 5 31.2538 15.9285 80.4108 Zfand5 205.039 141.019 51.3033 93.2466 76.164 17.8392 353.218 Zfp488 4.48364 33.1909 8.95718 4.08676 15.0177 3.34945 11.3519 Zfp691 4.4322 56.26218 3.56359 46.6795 2.75204 2.86518 8.56242 Zmat2 4.25649 26.4055 18.9911 71.979 36.9848 18.4582 4.00515 115.052 210.360 14.4625 Zmynd8 56.97817 149.9101 42.3268 1 2 4 154.296 196.694 Znhit1 73.0979 11.13187 51.392 4 25.7328 54.3487 281.35 182.712 428.740 Zwint 63.744 623.0542 8 649.956 401.563 414.714 9

259

Table 2 Gene expression levels by functional gene categories in VIP-LRPs

Gene Gene expression Gene Approved name AVG Median Range Symbo TPM TPM TPM l (log2) (log2) (log2) Gad1 Glutamate decarboxylase 1 5.86 5.8 2.8 - 8.9 Gad2 Glutamate decarboxylase 2 4.11 6.05 0 - 7.3 VIP Vasoactive intestinal peptide 6.80 7 1.6 - 12 Cck Cholecystokinin 5.37 5.4 0 - 13

Chrm2 Cholinergic receptor muscarinic 2 3.65 3.1 0 - 7.5

Tac2 Tachykinin 2 3.94 0 0 - 12.2 IN genes IN

- Calb1 Calbindin 1 1.10 0 0 - 6.2 VIP Lhx6 LIM homeobox 6 0.07 0 0 - 0.5 Sox6 SRY-box 6 0.67 0 0 - 4.7 Satb1 SATB homeobox 1 0.57 0 0 - 3.7 Npas3 Neuronal PAS domain protein 3 1.17 0 0 - 4 Nfia Nuclear factor I A 1.42 0.69 0 - 5.1

Nfib Nuclear factor I B 0.98 0.41 0 - 3.7

Nfix Nuclear factor I X 0.91 0 0 - 3.3

Htr3a 5-Hydroxytryptamine receptor 3A 3.24 2.9 0 - 8.8 MGE/CGE Neo1 Neogenin 1 0.30 0 0 - 1.2 Robo1 Roundabout guidance receptor 1 1.24 0 0 - 5.3 Robo2 Roundabout guidance receptor 2 0.69 0 0 - 4.7 Ntng1 Netrin G1 4.17 4.23 0 - 6.9 Dscam DS cell adhesion molecule 1.03 0.81 0 - 3.7 Sdk2 Sidekick cell adhesion molecule 2 0.56 0 0 - 3 Kirrel3 Kirre like nephrin family adhesion molecule 3 1.16 0 0 - 5.6

Cttnbp Cortactin binding protein 2 3.30 3.07 1.1 - 6.5

2 CAMs

260

Lamp5 Lysosomal associated membrane protein 4.23 6.4 0 - 7.7 family member 5 Cdh8 Cadherin 8 2.03 1.62 0 - 4.9 Cdh2 Cadherin 2 2.57 3.22 0 - 4 Cdh11 Cadherin 11 3.24 4.21 0 - 6.3 Plxna1 Plexin A1 1.98 0 0 - 5.4 Sema6 Semaphorin 6B 1.68 0 0 - 5.7 b Nrp1 Neuropilin 1 1.05 0.27 0 - 3.3 Sema3 Semaphorin 3E 0.77 0 0 - 2.9 e Nlgn2 Neuroligin 2 2.69 1.77 0 - 6.8 Nrxn1 Neurexin 1 2.49 2.86 0 - 5 Nrxn2 Neurexin 2 1.82 2.22 0 - 4.4 Nrxn3 Neurexin 3 3.54 3.8 0 - 5.7 Erbb4 Erb-b2 receptor tyrosine kinase 4 1.45 1.52 0 - 3.7 Ptprn2 Protein tyrosine phosphatase, receptor type 2.17 2.66 0 - 3 N2 Ptpra Protein tyrosine phosphatase, receptor type 2.29 3.5 0 - 4.5 A Ptpro Protein tyrosine phosphatase, receptor type 1.00 0 0 - 2.9 O Ptprz1 Protein tyrosine phosphatase, receptor type 1.39 0 0 - 4.7 Z1 Ptprj Protein tyrosine phosphatase, receptor type 0.73 0 0 - 3.5 J Lrrc4b Leucine rich repeat containing 4B 2.00 1.89 0.8 - 4 Lrrc49 Leucine rich repeat containing 49 1.79 0 0 - 5.1 Lrrc61 Leucine rich repeat containing 61 1.32 0 0 - 5.6 Lrrc8c Leucine rich repeat containing 8 VRAC 0.98 0 0 - 4.7 subunit C

261

Lrrcc1 Leucine rich repeat and coiled-coil 0.56 0 0 - 2.2 centrosomal protein 1 Hs6st2 Heparan sulfate 6-O-sulfotransferase 2 1.14 0 0 - 3.6 Ust Uronyl 2-sulfotransferase 1.64 0 0 - 4.8 St6gal ST6 N-acetylgalactosaminide alpha-2,6- 1.07 0 0 - 3.8 nac4 sialyltransferase 4 St8sia5 ST8 alpha-N-acetyl-neuraminide alpha-2,8- 1.94 0 0 - 5.5 sialyltransferase 5 St3gal ST3 beta-galactoside alpha-2,3- 1.95 1.1 0 - 6.3 5 sialyltransferase 5 St3gal ST3 beta-galactoside alpha-2,3- 1.15 0 0 - 4.3 3 sialyltransferase 3 Gria1 Glutamate ionotropic receptor AMPA type 2.56 3.17 0 - 5.5 subunit 1 Gria2 Glutamate ionotropic receptor AMPA type 3.86 3.15 0.9 - 7.6 subunit 2 Gria4 Glutamate ionotropic receptor AMPA type 2.01 2.56 0 - 3.6 subunit 4 Cacng2 Calcium voltage-gated channel auxiliary 2.40 3.6 0 - 4 subunit gamma 2

Shisa4 Shisa family member 4 3.72 4.72 0 - 8.3

Shisa9 Shisa family member 9 3.08 3.83 0 - 6.2 AMPAR Grin1 Glutamate ionotropic receptor NMDA type 3.59 3.54 0 - 7.1 subunit 1 Grin2b Glutamate ionotropic receptor NMDA type 2.32 2.23 0.6 - 4.4 subunit 2B

Grin2d Glutamate ionotropic receptor NMDA type 0.75 0 0 - 3.2

subunit 2D NMDAR

Grm1 Glutamate metabotropic receptor 1 2.09 1.69 0.6 - 4.2

Grm5 Glutamate metabotropic receptor 5 1.87 0.91 0 - 4.8 mGluR

262

Gabra1 Gamma-aminobutyric acid type A receptor 3.20 3.16 0 - 8.1 alpha1 subunit Gabra2 Gamma-aminobutyric acid type A receptor 2.77 3.18 0 - 6.7 alpha2 subunit Gabra3 Gamma-aminobutyric acid type A receptor 1.51 0.42 0 - 5.2 alpha3 subunit Gabra5 Gamma-aminobutyric acid type A receptor 1.48 1.28 0 - 6.6 alpha5 subunit Gabrb1 Gamma-aminobutyric acid type A receptor 2.97 0.76 0 - 7.5 beta1 subunit Gabrb3 Gamma-aminobutyric acid type A receptor 2.49 2.54 0 - 6.3 beta3 subunit

Gabbr2 Gamma-aminobutyric acid type B receptor 1.23 0 0 - 4.4

subunit 2 GABA Chrna4 Cholinergic receptor nicotinic alpha 4 0.94 0.17 0 - 3.5 subunit Adrb1 Adrenoceptor beta 1 2.23 0 0 - 7.5 Adrbk2 Adrenergic, Beta, Receptor kinase 2 2.16 0.38 0 - 5.1 Drd1 Dopamine receptor D1 0.80 0 0 - 3.3 Htr1d 5-hydroxytryptamine receptor 1D 0.29 0 0 - 1.5 Hrh2 Histamine receptor H2 1.57 0 0 - 5.7 Cnr1 Cannabinoid receptor 1 4.23 4.39 0 - 9.3 Oprd1 Opioid receptor delta 1 1.14 0 0 - 3.7 Oprl1 Opioid related nociceptin receptor 1 0.80 0 0 - 3.3 Ogfr Opioid growth factor receptor 1.18 0 0 - 5.7 Npy Neuropeptide Y 4.05 2.95 0 - 10.4

Npy1r Neuropeptide Y receptor Y1 2.58 1.95 0 - 6.4

Penk Proenkephalin 0.45 0 0 - 2.5 Sstr1 Somatostatin receptor 1 0.58 0 0 - 3.3 Sstr2 Somatostatin receptor 2 2.47 0 0 - 8.7

Ntsr2 Neurotensin receptor 2 3.40 0 0 - 9.04 Neuromodulation

263

Rgs8 Regulator of G protein signaling 8 1.93 1.09 0 - 4.2 Rgs10 Regulator of G protein signaling 10 4.29 5.53 0 - 8 Rgs11 Regulator of G protein signaling 11 1.57 0 0 - 5.7 Pde10a Phosphodiesterase 10A 1.68 1.58 0 - 3 Pde4b Phosphodiesterase 4B 1.12 1 0 - 2.5 Pde9a Phosphodiesterase 9A 1.30 0 0 - 6.8 Pde1b Phosphodiesterase 1B 1.21 0.25 0 - 5.3 Pde7a Phosphodiesterase 7A 1.40 0 0 - 5.1 Tesc Tescalcin 1.44 0 0 - 5.2 Hpcal1 Hippocalcin like 1 4.60 5.28 0 - 6.3 Caln1 Calneuron 1 2.38 1.39 0 - 6.3 Necab N-terminal EF-hand calcium binding protein 1.87 0 0 - 6.5 2 2 Kcnip3 Potassium voltage-gated channel interacting 1.62 0 0 - 6.1 protein 3 Fstl5 Follistatin like 5 1.06 0 0 - 4 Hpcal4 Hippocalcin like 4 4.05 5.26 0 - 7.5 Nras NRAS proto-oncogene, GTPase 1.29 1.34 0 - 2.8 Rap1a RAP1A, member of RAS oncogene family 2.30 0.91 0 - 5.4 Tiam1 T cell lymphoma invasion and metastasis 1 1.67 1.47 0 - 4.3 Als2 ALS2, alsin Rho guanine nucleotide exchange 1.85 1.41 0 - 4.6 factor Rasgrf Ras protein specific guanine nucleotide 2.86 3.03 0 - 4.7 1 releasing factor 1 Trio Trio Rho guanine nucleotide exchange factor 0.83 0 0 - 3 Bcr BCR, RhoGEF and GTPase activating protein 1.67 0 0 - 4.7 Itsn2 Intersectin 2 0.65 0 0 - 3.3 Rnd1 Rho family GTPase 1 1.03 0.32 0 - 4.5

Rhob Ras homolog family member B 2.19 1.82 0 - 6

Rhoa Ras homolog family member A 2.28 3.1 0 - 4.3

Igf1 Insulin like growth factor 1 1.30 0 0 - 4.6 Signaling

264

Nos1 Nitric oxide synthase 1 0.79 0 0 - 3.9 Vav3 Vav guanine nucleotide exchange factor 3 1.45 0 0 - 5.5 Cplx2 Complexin 2 4.82 5.63 2.2 - 6.3 Kcna1 Potassium voltage-gated channel subfamily 1.83 1.1 0 - 6 A member 1 Kcna2 Potassium voltage-gated channel subfamily 2.84 2.59 0 - 6.2 A member 2 Kcna6 potassium voltage-gated channel subfamily 1.72 1.63 0 - 8.8 A member 6 Kcnb1 Potassium voltage-gated channel subfamily 0.97 0 0 - 4.3 B member 1 Kcnc1 potassium voltage-gated channel subfamily 2.24 2.59 0 - 11.9 C member 1 Kcnc2 Potassium voltage-gated channel subfamily 2.86 0.54 0 - 8.5 C member 2 Kcnc3 Potassium voltage-gated channel subfamily 1.91 0 0 - 5.1 C member 3 Kcnc4 Potassium voltage-gated channel subfamily 0.72 0 0 - 3.1 C member 4 Kcnq2 Potassium voltage-gated channel subfamily 1.45 0 0 - 4.7 Q member 2 Kcnh1 Potassium voltage-gated channel subfamily 2.23 1.34 0 - 5.7 H member 1 Kcnh6 Potassium voltage-gated channel subfamily 4.79 5 0 - 9.5 H member 6 Kcnq3 Potassium voltage-gated channel subfamily 0.78 0 0 - 3.4 Q member 3 Kcnab1 Potassium voltage-gated channel subfamily 0.39 0 0 - 2.1

A member regulatory beta subunit 1 Kcnab3 Potassium voltage-gated channel subfamily 1.59 0 0 - 5.9

A regulatory beta subunit 3 Kv channels Kv

265

Dpp4 Dipeptidyl peptidase 4 0.43 0 0 - 1.8 Dpp6 Dipeptidyl peptidase like 6 1.69 0.58 0 - 5.8 Dpp10 Dipeptidyl peptidase like 10 3.58 3.61 0.3 - 7.8 Kcnip1 Potassium voltage-gated channel interacting 3.60 4.71 0 - 5.9 protein 1 Cacna1 Calcium voltage-gated channel subunit 0.73 1.06 0 - 1.5

a alpha1 A

Cacna1 Calcium voltage-gated channel subunit 1.40 0 0 - 5.4 b alpha1 B Cacna1 Calcium voltage-gated channel subunit 1.00 0.16 0 - 3.1 c alpha1 C Cacna1 Calcium voltage-gated channel subunit 0.94 0 0 - 6.3 g alpha1 G Cacnb2 Calcium voltage-gated channel auxiliary 1.14 0 0 - 3.5 subunit beta 2

Cacnb3 Calcium voltage-gated channel auxiliary 1.20 0.38 0 - 5.1

subunit beta 3 Cacna2 Calcium voltage-gated channel auxiliary 1.93 0 0 - 5.1

d3 subunit alpha2delta 3 Cav channels Cav Scn3b Sodium voltage-gated channel beta subunit 2.34 2.9 0 - 5 3 Scn8a Sodium voltage-gated channel alpha subunit 1.27 1.64 0 - 3.6 8 Scn2a1 Sodium voltage-gated channel alpha subunit 1.51 1.23 0 - 3.4 2 Scn2b Sodium voltage-gated channel beta subunit 2.65 2.45 0 - 5.2 2

Scn1b Sodium voltage-gated channel beta subunit 4.28 5.21 0 - 6.4

1 Scn1a Sodium voltage-gated channel alpha subunit 3.02 3.08 0 - 5

1 Nav channels Nav

266

Hcn1 Hyperpolarization activated cyclic nucleotide 2.25 1.4 0 - 4.9 gated potassium channel 1

Hcn2 Hyperpolarization activated cyclic nucleotide 3.17 3.35 0 - 6.1

gated potassium and sodium channel 2 Hcn3 Hyperpolarization activated cyclic nucleotide 1.70 0.45 0 - 5

gated potassium channel 3 HCN channels HCN

Ptn Pleiotrophin 3.55 3.96 0 - 7.6

Gjc3 Gap junction protein gamma 3 1.16 1.38 0 - 2.6 Myelin

Emx1 Empty spiracles homeobox 1 0.00 0 0

Tbr1 T-Box, Brain 1 0.00 0 0

Gfap Glial fibrillary acidic protein 0.00 0 0 Contamination

267