Synchronization by Distal Dendrite-Targeting Interneurons

BRAIN INSTITUTE Federal University of Rio Grande do Norte

Synchronization by Distal Dendrite-targeting Interneurons

- phenomenon, concept and reality -

THESIS

for the degree of

Doctor of Philosophy

Neuroscience

submitted by

Markus Michael Hilscher Matriculation: 2013116277

at the Brain Institute of the Federal University of Rio Grande do Norte

Advisor: Dr. Emelie Katarina Svahn Leão Brain Institute, Federal University of Rio Grande do Norte, Brazil Co-Advisor: Dr. Richardson Naves Leão Brain Institute, Federal University of Rio Grande do Norte, Brazil Co-Advisor: Dr. Klas Kullander Department of Neuroscience, Uppsala University, Sweden

Opponent: Dr. Christopher Kushmerick Institute of Biological Sciences, Federal University of Minas Gerais, Brazil Opponent: Dr. Gilad Silberberg Department of Neuroscience, Karolinska Institute, Sweden

Natal, 01.04.2017 Federal University of Rio Grande do Norte 59056-450 Natal Av. Nascimento de Castro, 2155 www.neuro.ufrn.br

BRAIN INSTITUTE Federal University of Rio Grande do Norte

Sincronização através de interneurônios que inervam dendritos distais

- fenômeno, conceito e realidade -

TESE

para a obtenção do título de

Doutor

Neurociências

apresentada por

Markus Michael Hilscher Matrícula 2013116277 no Instituto do Cérebro da Universidade Federal do Rio Grande do Norte

Orientadora: Dra. Emelie Katarina Svahn Leão Instituto do Cérebro, Universidade Federal do Rio Grande do Norte, Brasil Co-Orientador: Dr. Richardson Naves Leão Instituto do Cérebro, Universidade Federal do Rio Grande do Norte, Brasil Co-Orientador: Dr. Klas Kullander Departamento de Neurociências, Universidade de Uppsala, Suécia

Avaliador: Dr. Christopher Kushmerick Instituto de Ciências Biológicas, Universidade Federal de Minas Gerais, Brasil Avaliador: Dr. Gilad Silberberg Departamento de Neurociências, Instituto Karolinska, Suécia

Natal, 01.04.2017 Federal University of Rio Grande do Norte 59056-450 Natal Av. Nascimento de Castro, 2155 www.neuro.ufrn.br

Universidade Federal do Rio Grande do Norte - UFRN Sistema de Bibliotecas - SISBI Catalogação de Publicação na Fonte. UFRN - Biblioteca Setorial Árvore do Conhecimento - Instituto do Cérebro - ICE

Hilscher, Markus Michael. Synchronization by Distal Dendrite-targeting Interneurons / Markus Michael Hilscher. - Natal, 2017. 152f.: il.

Universidade Federal do Rio Grande do Norte, Instituto do Cérebro, PGNeuro. Orientador: Dr. Emelie Katarina Svahn Leão. Coorientador: Dr. Richardson Naves Leão.

1. Synchronization. 2. Hippocampus. 3. OLM cells. 4. Neocortex. 5. Martinotti cells. I. Leão, Emelie Katarina Svahn. II. Leão, Richardson Naves. III. Título.

RN/UF/BSICe CDU 612.8

Abstract

Synchronization among neurons arises from the cooperative interaction of various cell types through excitation and inhibition. The mechanisms behind this type of neuronal orchestration are as versatile as almost no other coordination task in the brain, making its comprehension heavily challenging. Among many others, the high number of involved cell types, the diversity of synaptic processes as well as the contribution of a multitude of ion channels and currents span the plurality of neuronal synchronization mechanisms in our brains. Focusing on two brain areas, the hippocampus and the neocortex, this thesis aims to understand the role of distal dendrite- targeting interneurons in shaping pyramidal cell activity and the timing of their action potentials.

The distribution of ion channels and synaptic receptors in pyramidal cell dendrites is extremely anisotropic. Thus, interneurons innervating the proximal or distal areas of the dendrites cause different effects in the target cell when activated. For example, the distal portions of the pyramidal cell dendrites contain one of the most prominent pacemaker channels: the hyperpolarization- activated cyclic nucleotide-gated channels. These channels produce a cationic depolarizing current (Ih) and play an essential role in the regulation of neuronal excitability. Using computational modeling, this thesis shows how the amount of Ih in certain cell types determines their spike rate, synchrony as well as power and frequency of ongoing network oscillations. Moreover, since Ih differs between brain regions as well as cell types and location, this thesis electrophysiologically explores how Ih differs along the dorsoventral axis of hippocampus in oriens-lacunosum moleculare (OLM) cells, the main distal dendrite-targeting interneurons of that region.

Utilizing the main distal dendrite-targeting interneuron of the neocortex, the Martinotti cell, this thesis also shows how a defined population of interneurons can be manipulated in order to control and align pyramidal cell firing. By providing the right amount and frequency of inhibition, Martinotti cells are able to synchronize trains of subtype-specific pyramidal cells. Using optogenetic approaches to activate/inactivate populations of Martinotti cells, these dendrite-targeting interneurons are shown to trigger rebound action potentials in pyramidal cells when temporally aligned. The rebound action potentials in turn are the result of strong inhibition by Martinotti cells, giving these distal dendrite-targeting interneurons the power to reset pyramidal cell firing.

Overall, Martinotti cells and OLM cells show quite striking similarities in morphological, neurochemical and electrophysiological properties. Especially, their long axonal projections to upper layers as well as their low-threshold, slow spiking fashion and the accommodating ﬁring make these distal dendrite-targeting interneurons so special for neuronal synchronization.

Resumo

A sincronização neuronal surge de uma interação cooperativa de vários tipos celulares através de excitação e inibição. Os mecanismos por trás desse tipo de coordenação neuronal são, prova- velmente, os mais dinâmicos entre as funções cerebrais, dificultando sua compreensão. Entre os fatores que dificultam o estudo da sincronia, pode-se citar: o vasto número de tipos de celulares, a diversidade de processos sinápticos, a contribuição de uma multiplicidade de canais e corren- tes iônicas, entre outros. Essa tese tem como objetivo entender o papel de interneurônios que especificamente inervam o domínio distal dos dendritos de células piramidais do hipocampo e neocórtex, na sincronização de neurônios em suas respectivas redes.

A distribuição de canais iônicos e receptors sinápticos em dendritos de células piramidais é ex- tremamente anisotrópica. Assim, interneurônios que inervam domínios proximais e distais dos dendritos causam efeitos distintos na célula alvo quando ativados. Por exemplo, porções distais dos dendritos contém em abundância um dos principais canais marcapassos em neurônios: o canal regulado por nucleotídeo cíclico ativado por hiperpolarização. Esses canais produzem uma corrente catiônica despolarizante (Ih) e tem um papel importante na regulação da excita- bilidade neuronal alterando dramaticamente as propriedades de disparo de neurônios. Usando modelagem computacional, essa tese mostra como a amplitude de Ih em certos tipos celulares muda a taxa de disparo de um neurônio, sua sincronia além da energia espectral e frequência de oscilações. Além disso, como a expressão de Ih difere entre regiões cerebrais, localização e tipos celulares, essa tese, fazendo o uso de patch clamp, explora como Ih difere ao longo do eixo dorsoventral do hipocampo em células oriens-lacunosum moleculare (OLM), que são os principais interneurônios que inervam dendritos distais dessa região.

Ademais, estudou-se aqui as células Martinotti, interneurônios que inervam os dendritos distais do neocórtex. Nesse estudo, mostrou-se como uma população deﬁnida de interneurônios pode ser manipulada com o objetivo de controlar e coordenar o disparo de células piramidais. Ao fornecer inibição com energia e frequência adequada, as células Martinotti afetam especiﬁca- mente um único tipo de célula piramidal. Usando optogenética para ativar/desativar populações de células Martinotti, é possível gerar potenciais de ação rebote em células piramidais quando alinhadas temporalmente. Os potenciais de ação rebote, por sua vez, são resultado de uma forte inibição produzida pelas células Martinotti, o que faz com que esses esses interneurônios pos- sam resetar o disparo de células piramidais.

iii De forma geral, células Martinotti e células OLM mostram similaridades surpreendentes em propriedades morfológicas, neuroquímicas e eletroﬁsiológicas. Especialmente, suas longas pro- jeções axonais para camadas superiores assim como seus modos de disparo lentos, com baixos limiares e acomodativos tornam esses neurônios singulares em suas capacidades de sincronizar os circuitos nos quais estão inseridos. Kurzfassung

Synchronisation zwischen Neuronen entsteht durch das kooperative Zusammenspiel von ver- schiedenen Zelltypen und deren Erregung und Hemmung. Die Mechanismen hinter dieser Art der neuronalen Orchestrierung sind so vielseitig wie kaum ein anderer Koordinierungsprozess im Gehirn. Die große Anzahl an beteiligten Zelltypen, die Vielfalt der synaptischen Prozesse sowie der Beitrag einer Vielzahl von Ionenkanälen und Strömen charakterisieren die neuronalen Synchronisationsmechanismen in unserem Gehirn. Der Schwerpunkt dieser Arbeit liegt auf zwei Bereichen des Gehirns, dem Hippocampus und dem Neocortex, und der Rolle von distalen Dendriten-innervierenden Interneuronen, die Pyramidenzellen steuern und den Zeitpunkt ihrer Aktionspotentiale beeinﬂussen können.

Die Verteilung der Ionenkanäle und synaptischen Rezeptoren in den Dendriten von Pyrami- denzellen ist sehr anisotrop. Daher haben Interneurone, welche die proximalen oder distalen Bereiche der Dendriten innervieren, verschiedene Effekte in Pyramidalzellen. Beispielsweise, enthalten die distalen Abschnitte der Dendriten bestimmte Schrittmacherkanäle: die hyperpola- risations-aktivierten, zyklischen, Nukleotid-gesteuerten Kanäle. Diese Kanäle erzeugen einen kationischen, depolarisierenden Strom (Ih) und spielen eine wichtige Rolle bei der Regulati- on der neuronalen Erregbarkeit. Die vorliegende Arbeit zeigt mit Computermodellen, wie die Menge von Ih in bestimmten Zelltypen die neuronale Aktivität, deren Synchronität sowie die Leistung und Frequenz von Netzwerkoszillationen verändert. Da darüber hinaus Ih zwischen Hirnregionen und Zelltypen verschieden ist, untersucht diese Arbeit elektrophysiologisch wie sich Ih entlang der dorsoventralen Achse des Hippocampus in oriens-lacunosum moleculare (OLM) Zellen, den wichtigsten distalen Dendriten-innervierenden Zellen dieser Hirnstruktur, unterscheidet.

Durch Manipulation der wichtigsten Dendriten-innervierenden Zellen des Neocortex, den Mar- tinotti Zellen, zeigt diese Arbeit auch, wie eine deﬁnierte Population von Interneuronen die Aktivität von Pyramidenzellen steuern kann. Mit der richtigen Menge und Frequenz können Martinotti Zellen distale Dendriten von Pyramidenzellen hemmen und subtyp-speziﬁsche Py- ramidenzellen synchronisieren. Mittels optogenetischer Verfahren werden Populationen dieser Dendriten-innervierenden Interneurone aktiviert bzw. inaktiviert und gezeigt, dass sie Rebound- Aktionspotentiale in Pyramidenzellen auslösen können, wenn sie zeitlich koordiniert sind. Die Rebound-Aktionspotentiale wiederum sind das Ergebnis einer kräftigen Martinotti Zell-Hem- mung, welche den distalen Dendriten-innervierenden Interneuronen erlaubt das Potenzial von Pyramidenzellen zurückzusetzen.

v Insgesamt besitzen Martinotti Zellen und OLM Zellen ziemlich auffallende Ähnlichkeiten in morphologischer, neurochemischer und elektrophysiologischer Hinsicht. Insbesondere ihre lan- gen axonalen Projektionen zu den oberen Schichten sowie deren vergleichsweise langsames und adaptierendes Feuern von Aktionspotentialen machen diese distalen Dendriten-innervierenden Interneurone so besonders für die neuronale Synchronisation. The danger of computers becoming like humans is not as great as the danger of humans becoming like computers.

Konrad Zuse

Contents

Abstract i

Abstract in Portuguese iv

Abstract in German vi

1 Introduction 1 1.1 Synchronization in the Brain ...... 2 The hippocampus ...... 2 The neocortex ...... 3 1.2 Models and Theories of Synchronization ...... 5 Synﬁre chains ...... 5 Coupled oscillators ...... 5 Binding theory ...... 6 Conductance-based models ...... 6 1.3 Hippocampal Interneurons and their Synchronization features ...... 9 Pacemaker currents in distal dendrites ...... 10 Cellular subthreshold oscillators in theta ...... 11 1.4 Neocortical Microcircuits partly resemble that of hippocampus ...... 13 Cortical pyramidal cell diversity ...... 15 Cholinergic modulation and beta frequency ...... 17 1.5 Aims of Study I, II and III ...... 17 Studies ...... 18 Additional publications ...... 18 Poster presentation ...... 18

2 General Methods 19 2.1 Computer simulations ...... 19 2.2 Animal model ...... 20 2.3 Optogenetics ...... 20 2.4 Electrophysiology ...... 20 2.5 Imaging and Tracing ...... 21 2.6 Data analysis ...... 21

ix 2.7 Ethical statement ...... 23 2.8 Methodological considerations ...... 23

3 Studies and Results 25 3.1 Ih Tunes Theta/Gamma Oscillations and Cross-Frequency Coupling In an In Sil- ico CA3 Model ...... 25 Background ...... 25 Experimental Question (Hypothesis) ...... 25 Outcome ...... 26 Conclusion and Future Perspectives ...... 26 3.2 Dorsoventral gradient of hyperpolarization-activated current (Ih) in hippocampal CA1 OLM interneurons ...... 29 Background ...... 29 Experimental Question (Hypothesis) ...... 29 Outcome ...... 29 Conclusion and Future Perspectives ...... 30 3.3 Chrna2-Martinotti cells Synchronize layer 5 type A Pyramidal Cells via Re- bound Excitation ...... 33 Background ...... 33 Experimental Question (Hypothesis) ...... 33 Outcome ...... 33 Conclusion and Future Perspectives ...... 35

4 Discussion 37 4.1 Conclusion ...... 40

Abbreviations 44

Acknowledgments 45

Bibliography 49

A Study I 73

B Study II 87

C Study III 101

x CHAPTER 1 Introduction

One of the most ubiquitous phenomena in nature is synchronization. Continuously we are ex- posed to a multitude of naturally given or artiﬁcially generated synchronization processes, such as adjusting our sleep to the day-night cycle, sharing data between multiple devices or simply meeting colleagues to exchange knowledge - we synchronize our activities in time and space. Especially communication processes and information transfer rely on the temporal synchronization of events. Derived from the Greek words syn (meaning the same) and chronos (meaning time) synchronization describes the coordination of two or more systems in time and univer- sally characterizes phenomena in natural sciences, formal sciences and social sciences. While synchronization characterizes the act of the coordination, the word synchrony outlines the state of being temporally aligned - often in a certain frequency. The adjectives synchronized and synchronous are applied accordingly, but because of their close relationship often used synonymously to emphasize both, coordination aspects at a certain point of time as well as continuous alignment of events over a period of time.

Historically, probably the ﬁrst description of synchronized systems goes back to the Dutch mathematician Christiaan Huygens who showed that two pendulum clocks, hung from the same wooden structure will come to oscillate in synchrony (Huygens, 1664; Peña Ramirez et al., 2016). There are many examples from various disciplines utilizing the principle of Huygens’ pendulum clock; among many others metronomes - to keep a steady beat, seismometers - to measure eruptions, or mass dampers - to stabilize high buildings such as the Taipei 101 tower, are exemplary mentioned here. Overall, synchronization mechanisms may induce periodic oscillations (repeated variations) to a system and produce stability via rhythmicity and precise timing.

In neuroscience one of the most famous principles where precise timing of neuronal events is of great importance, comes from the Canadian psychologist Donald Olding Hebb. His theory postulates that connections between neurons increase or decrease in efﬁcacy depending on the timing of neuronal events. Sometimes summarized as “neurons wire together if they ﬁre together” (Löwel and Singer, 1992), Hebbian theory describes the principles of synaptic plas-

1 ticity, i.e. the ability of synapses to strengthen or weaken over time in response to increases or decreases in their activities (Hebb, 1949). Hebbian learning is fundamental for the formation of cell assemblies - that are groups of neurons which ﬁre simultaneously; and provides a biological basis for learning and memory. However, the mechanism behind learning and memory are very complex and still remain an enigma.

1.1 Synchronization in the Brain

Neuroscientists study the formation and propagation of synchronized activities using a multitude of distinct strategies. For example, top-down approaches investigate the role of synchronized processes during behaviors associated with learning and memory, bottom-up approaches explore the cellular mechanisms of memory encoding brain regions by measuring the firing of local groups of neurons. A comprehensible way to study basic mechanisms of synchronization is to examine small networks, so-called microcircuits, consisting of a minimal number of interacting neurons that collectively produce a functional output (Grillner et al., 2005). A microcircuit is so to say base and interface for higher-order functions; synchronization created by a unique combination of temporal characteristics its condition and output. Depending on the scientific question, electrophysiological advances, computational methods and more recently genetic techniques are used to identify individual neurons and study their contribution within a specific microcircuit. Higher-order functions, such as memory, spatial orientation or consciousness are very often linked to the synchronized output of hippocampal or neocortical microcircuits, placing these two brain regions among the most extensively studied when aiming to decode the mechanisms behind neuronal synchronization (Engel et al., 1991; Buzsáki et al., 1994; Lisman and Idiart, 1995; Grillner et al., 2005).

The hippocampus The hippocampal formation, comprising cornu ammonis (CA), dentate gyrus (DG) and subiculum (Sub) (Fig. 1.1), is a widely used brain structure for studying the role of neuronal synchronization and brain rhythms, often described as oscillations (Berger, 1929; Buzsáki, 2002). Going back to the Spanish histologist Santiago Ramón y Cajal (Cajal, 1893; Cajal, 1911) and his disciple Rafael Lorente de Nó (Lorente de Nó, 1933; Lorente de Nó, 1934), the hippocampus proper has been divided into the three principal subdivisions CA1, CA2 and CA3 with a vast number of different cell types situated in clearly distinguishable strata (layers). The largest and most extensively studied hippocampal region is CA1, receiving its main inputs from the temporoammonic pathway originating in the entorhinal cortex (EC) and the Schaffer collaterals of CA3 pyramidal cells (Buzsáki, 2002). More than 20 different interneuron subtypes can be distinguished (Freund and Buzsáki, 1996; Klausberger and Somogyi, 2008) and following neurochemical as well as morphological properties roughly divided into parvalbumin-expressing interneurons, cholecystokinin-expressing interneurons, interneurons with densely packed axons and gamma-aminobutyric acid (GABA)ergic long-range projecting neurons (Klausberger, 2009). This categorization is not mutually exclusive but rather a distinction into groups of morphologically and functionally different neurons, similar to classiﬁcations into perisomatic and

2 dendritic neurons or proximal and distal dendrite-targeting cells. Together with excitatory pyramidal cells these distinct interneuron classes give rise to a multitude of mechanisms, such as feedback and feedforward inhibition, which are required for the recruitment of synchronous activities, the generation of rhythms and the temporal coupling of oscillations (Buzsáki et al., 1994; Tort et al., 2010; Hyaﬁl et al., 2015).

# CA1

CA2

Sub

DG EC CA3

CA1 DG

CA3

Figure 1.1: Anatomy of the hippocampal formation. top: Drawing of Santiago Ramón y Cajal, with cornu ammonis (CA), dentate gyrus (DG), subiculum (Sub) and entorhinal cortex (EC). Note the densely packed pyramidal cell layer stratum pyramidale (#). bottom: Coronal section (left) highlighting stratum oriens (4)& stratum lacunosum moleculare (◦) and projection (right) of hippocampus from a Chrna2-Cre/tdTomato mouse (transgenic animal where cells containing cholinergic receptor nicotinic alpha 2 also express red ﬂuorescent protein). Images obtained using CLARITY (Clear Lipid-exchanged Acrylamide-hybridized Rigid Imaging Tissue hYdrogel).

The neocortex Like the hippocampus, the neocortex also consists of a vast amount of interconnected neurons with distinct cell types, residing in six well distinguishable lamina (layers) (Cajal, 1899; Brod- mann, 1909; Lorente de Nó, 1934). First described by the German neuroanatomist Korbinian Brodmann, the cortex can be divided into sensory, motor and association areas including the primary somatosensory cortex, motor cortex, visual cortex, and auditory cortex (Economo and Koskinas, 1925). Classically, the main input to the neocortex arrives in layer 4 from the tha- lamus, the main output to other cortical and subcortical regions emerges in layer 5 (Fig. 1.2), although recent studies show that the input/output layers are much more diverse (Huang, 2014;

3 Harris and Shepherd, 2015). Especially, these densely packed layers with numerous pyramidal cell-interneuronal connectivity patterns make the neocortex well suited to study the role of interlaminar top-down and bottom-up rhythms as well as intra- and intercortical oscillations (Buzsáki and Draguhn, 2004). Inhibitory cells of the neocortex are often classified by their neurochemical properties into parvalbumin-, somatostatin- and ionotropic serotonin receptor 5HT3a-expressing interneurons (Markram et al., 2004; Ascoli et al., 2008; Rudy et al., 2011; Tremblay et al., 2016). However, overlapping morphological and electrophysiological properties result in a high number of morpho-electrical subtypes, targeting pyramidal cells on proximal and distal compartments (Markram et al., 2004). Functionally different, these perisomatic and dendritic targeting interneurons take specific roles during higher-order functions and allow the formation of various synchronous activities (Diesmann et al., 1999), poetically referred to as cortical songs (Ikegaya et al., 2004), spike avalanches (Plenz and Thiagarajan, 2007; Ribeiro et al., 2016) and most recently blankets of inhibition (Karnani et al., 2014; Karnani et al., 2016a). Spike sequences have been found in various preparations (Harvey et al., 2012; Okubo et al., 2015), however, the underlying mechanisms are still not clear and particularly the question of how single cells can fire synchronously with one another remains a challenging and recurrent topic in neuroscience.

V VC2

AC1 Hip.

Figure 1.2: Anatomy of the cortex. top: Drawing of Rafael Lorente de Nó, with pyramidal cells spread over the six layers (I-VI). bottom: CLARITY image (left) including primary auditory cortex (AC1), secondary visual cortex (VC2) & hippocampus (Hip.) and 3D projection (right) showing the distribution of speciﬁcally labeled interneurons in Chrna2-Cre/tdTomato mice.

4 1.2 Models and Theories of Synchronization

Synfire chains An elegant way to explore synchronization mechanisms is the use of computational models and the simulation of selected issues under isolated, predefined conditions. In 1982 the Israelian neuroscientist Moshe Abeles (Abeles, 1982) introduced so-called synfire chains - models to investigate the propagation of synchronized neuronal impulses within defined microcircuits. Mainly utilizing integrate-and-fire models (Lapicque, 1907; Abbott, 1999; Brunel and Van Rossum, 2007) these networks consist of excitatory neurons linked via feedforward connections distributed over many layers - also named pools (Fig. 1.3). Each pool was interconnected with a high degree of divergent and convergent connections (Abeles et al., 2004). Thereby, divergent connections (synapses from one neuron onto many other neurons) create alternatives while convergent connections (synapses from many neurons onto one neuron) make choices. Driven by local excitation and its hierarchical architecture, synfire chains showed that synchronous firing can propagate from one pool to the other, even when single neurons within pools fired at random times (Abeles, 1991).

Figure 1.3: Architecture of synﬁre chains. Two synﬁre chains with seven pools (P): Each neuron is connected to four neurons in the next layer (pool) from the same chain and to one neuron from the other chain (Abeles et al., 2004).

Coupled oscillators If synchronous activity is rhythmic and repetitive, oscillations can occur. These repetitive variations do not necessarily need to be in-phase (constant time), but can also be anti-phase (180 degrees shifted) or with an arbitrary phase shift (an arbitrary difference in degrees or time). A simple model for studying synchrony and oscillations goes back to the Japanese physicist Yoshiki Kuramoto (Kuramoto, 1984), who extended Winfree’s model of coupled oscillators (Winfree, 1967) and mathematically described the phase transition from incoherent states (random drifts between oscillators) to coherent states (non-random drifts). These models are theoretical formu- lations to investigate the coupling of oscillators as a basis for synchrony and are often used to study oscillators with ﬁxed vs. variable frequencies or critical vs. ’all-to-all’ coupling in order

5 to ﬁnd a certain coupling value above which synchronization occurs. Many other models exist, usually favoring excitatory coupling over inhibitory coupling for the generation of synchrony (Ermentrout, 1985; Abbott, 1990; Mirollo and Strogatz, 1990). Nowadays, such models are for example used to describe the coupling strength between oscillations of different brain areas, i.e. to identify functional connectivity between whole-brain networks (Cumin et al., 2007; Cabral et al., 2014).

Binding theory The coupling of oscillations is also central subject of the binding-by-synchrony hypothesis but attempts to find the link between synchrony and cognition. The so-called binding model explores the question of how spatially separated processes can be coordinated and bound together in order to form coherent outputs (Milner, 1974; Grossberg, 1976a; Grossberg, 1976b; van der Malsburg, 1981). In contrast to the aforementioned model, this theory does not put the amount of coupling as the pivotal element for synchrony but rather questions if neuronal binding is a consequence of synchronized states. Giving the model its name, the main postulate is that synchronously firing neurons are bound through oscillations while when firing out of synchrony neurons are unbound, i.e. implying that neuronal signals are paired through synchronized oscillations. Especially in the visual field certain evidence exists for context-dependent synchronous oscillatory discharges (Neuenschwander and Singer, 1996; Singer, 1999; Fries et al., 2002; Womelsdorf et al., 2007), although the derived theory that features of individual objects are bound via the synchronization of different neurons is heavily debated (Shadlen and Movshon, 1999; Dong et al., 2008). Ad- ditional to unit synchronization, the coordination of oscillations between different brain areas has also been investigated, developing in a theory known as communication-through-coherence hypothesis (Fries, 2005; Fries, 2009). This model suggests that neuronal communication is achieved by the coherence of neuronal groups, i.e. that synchronization is efficient to transmit information between neurons and optimal, when groups of separated neurons transmit synchrony via coherent oscillations (Fries, 2005; Fries, 2009). However, although experimental evidence supports the existence of coherent communication during attention and learning, this model is mainly of phenomenological nature and leaves the underlying cellular mechanisms unanswered (for a review see Deco et al., 2011).

Conductance-based models In order to convey the cellular processes behind synchronous ﬁring and oscillations, conductance- based models going back to the famous Hodgkin-Huxley model are often used (Hodgkin and Huxley, 1952). Besides ’all-to-all’ and purely excitatory networks, inhibitory networks as well as mixed excitatory-inhibitory networks were simulated to prove the role of so-called pacemaker neurons for the formation of oscillations (Van Vreeswijk et al., 1994; Hansel et al., 1995; Brunel, 2000). Thereby, similar to telecommunication networks where switches control the clock rate of a system, interneurons are often thought to have pacemaker qualities and can control rhythmicity and synchronization via inhibition (Lytton and Sejnowski, 1991; Wang and Rinzel, 1992; Bush and Sejnowski, 1996). Simulations of randomly connected, mutually inhibiting fast-spiking basket cells showed that purely inhibitory networks are indeed able to synchronize spikes in time

6 (Wang and Buzsáki, 1996). Moreover, these simulated interneurons were able to generate network oscillations at gamma frequency (20-80 Hz) (Wang and Buzsáki, 1996), oscillations which have been reported in vitro (Frégnac et al., 1994; Whittington et al., 1995) and in vivo (Bragin et al., 1995; Penttonen et al., 1998) and require a critical sparseness of activity, i.e. a minimal number of synaptic contacts per cell, independent of network size (Börgers and Kopell, 2003).

Combining inhibitory basket cells with excitatory pyramidal cells (Traub and Miles, 1991; Traub et al., 1991; Ermentrout and Kopell, 1998) mixed excitatory-inhibitory network models resulted in so-called pyramidal-interneuronal network gamma (PING) oscillations (Börgers et al., 2005; Börgers et al., 2008). Thereby, the inhibitory cell population is driven by pyramidal excitation and provides feedback inhibition to the pyramidal cells. This means that increased excitation of interneurons also increases inhibition onto pyramidal cells and as a consequence temporarily turns off the pyramidal cells. After the inhibition decays, the pyramidal cells get excitable again and the cycle starts over. These networks then result in the generation of rhythmic changes between excitation and inhibition. In a simulation of 160 excitatory pyramidal cells and 40 inhibitory basket cells, PING was shown to be highly sensitive to the connection probabilities (Börgers et al., 2005). If either the excitatory drive was too strong or the inhibitory drive too weak, PING was lost - a phenomenon sometimes described as balance of excitation and inhibition (Shadlen and Newsome, 1994; Van Vreeswijk and Sompolinsky, 1996). Balance of excitation and inhibition has been reported in in vitro (Shu et al., 2003) and in vivo (Haider et al., 2006) and describes the well-adjusted acting of a huge number of different types of pyramidal cells and interneurons in our brains. In order to increase pyramidal cell ﬁring, models often include recurrent excitatory connections between pyramidal cells (Börgers et al., 2005), which is especially apparent in the CA3 region of hippocampus (Wong and Traub, 1983; Le Duigou et al., 2014) and adds more excitatory drive to the network. This drive in turn suppresses acceler- ating inhibition and is able to facilitate synchrony in the network (Börgers et al., 2005).

Introducing a second type of interneuron to the pyramidal cell-basket cell model in form of the distal dendrite-targeting oriens-lacunosum moleculare (OLM) cell, revealed that these interneurons are able form network oscillations in theta frequency (4-12 Hz) via their feedback connections (Saraga et al., 2003; Tort et al., 2007). Theta oscillations were ﬁrst described in rabbits (Petsche et al., 1962) and have also been reported in in vitro (Ylinen et al., 1995) and in vivo studies in rats (Kamondi et al., 1998) as well as other rodents, primates and humans (Buzsáki, 2002). However, while fast-spiking basket cells ﬁre quite regularly at each gamma cycle, OLM cells primarily spike slower at each theta cycle and are able to nest ongoing gamma within theta oscillations, i.e. OLM cells promote the coexistence of multiple rhythms and thus coordinate pyramidal cell activity in multiple cell assemblies (Tort et al., 2007). The nesting of gamma within theta has been reported in both, in vitro and in vivo preparations (Fisahn et al., 1998; Gillies et al., 2002; Gloveli et al., 2005; Dugladze et al., 2007). Modeling studies further revealed that nested gamma and theta rhythms emerge from reciprocally connected basket cells and mutual feedback connections between basket cells and OLM cells (Fig. 1.4A, top) (Tort et al., 2007; Kopell et al., 2010). Although it has not been shown yet whether OLM cells and basket cells are connected in such manner, some evidence exists (Rotstein et al., 2005). When

7 removing the feedback connection from OLM cells to basket cells in the model, the OLM cells get unsynchronized and the nesting immediately is lost (Fig. 1.4A, bottom). The same holds for pyramidal cell-basket cell-OLM cell networks where the nesting of gamma within theta also is carried by the feedback connections from OLM cells to basket cells (Fig. 1.4B, top and bottom). Besides the distinct nesting properties, the basket cell-OLM cell network models further show that the fast and slow oscillations are differently altered by conduction delays and that the underlying cell properties inﬂuence synchrony as well (White et al., 2000; Kopell et al., 2000; Kopell et al., 2010). While gamma oscillations are merely inﬂuenced by the decay time of the inhibition, theta oscillations seem to be additionally affected by the kinetics of the OLM cells, i.e. their subthreshold oscillations or resonance in the theta frequency range (Kopell et al., 2010).

Overall, what these models impressively showed on different level of detail, is that neuronal oscillations can emerge from the spike-to-spike synchronization of neurons or populations thereof. Aptly formulated by Xiao-Jing Wang: “Neural correlation implies synchronization on some time scale, which can occur with or without oscillations.” (Wang, 2010). Moreover, different to classical oscillations as e.g. described by Huygens, the individual neurons that participate in an oscillatory network usually do not fire at each cycle of the oscillation and can vary in phase. Neuronal oscillations are the result of a group of individual cells firing synchronously with an inherent tolerance to allow spatiotemporal integration (Ermentrout and Kopell, 1998; Bartos et al., 2007; Brunel and Hakim, 2008). Sometimes also referred to as sparsely synchronized oscillations, neuronal oscillations can only appear at the population level in multi-unit or local field potential recordings but are not apparent in paired recordings (Brunel and Hakim, 2008). In silico, especially interneurons fire quite consistently during on-going oscillations, whereas pyramidal cells tend to skip cycles (Wang, 2002; Olufsen et al. 2003; Wang, 2010), prompting to rule out their synchronization features under experimental conditions.

Figure 1.4: Simulations of nested gamma and theta. A) Spiking of OLM (O) and basket cells (I) with (top) and without (bottom) feedback connections between OLM and basket cells. B) Same as in [A] for a pyramidal cell-basket cell-OLM cell (E-I-O) microcircuit (Kopell et al., 2010).

8 1.3 Hippocampal Interneurons and their Synchronization features

While computer simulations showed that interneurons can have certain pacemaker qualities and differentially mediate synchrony in theta and gamma (Wang and Buzsáki, 1996; Wang, 2002; Tort et al., 2007), in vitro and in vivo studies revealed that neuronal oscillations also depend on the behavioral condition or brain-state, with gamma being connected to attention and theta to movement or rapid eye movement sleep (Wang, 2010). Further, morphological differences in genetically distinguishable interneuron populations can give rise to the different synchronization properties (Klausberger and Somogyi, 2008). Among many others, preferably phase-locked to gamma oscillations are basket cells, bistratified cells and ivy cells (Klausberger et al., 2003; Fuentealba et al., 2008). While basket cells show strong coupling to gamma under urethane anesthesia in rats, bistratified cells display more prominent gamma-coupling in freely moving animals (Klausberger et al., 2003; Klausberger et al., 2005; Tukker et al., 2007). Both cell types express parvalbumin and together with axo-axonic cells are labeled in parvalbumin-Cre mice. Parvalbumin interneurons exhibit a distance-dependent gradient of connectivity to pyramidal cell dendrites of CA1, i.e. the synapses with CA1 pyramidal cells are distributed in an organized fashion, with highest density on proximal dendrites and lowest density on distal dendrites (Fig. 1.5) (Klausberger and Somogyi, 2008; Bloss et al., 2016). Axo-axonic cells exclusively inhibit pyramidal cells on the axon-initial segments, bistratified cells inhibit the basal domain and the apical oblique dendrites in stratum radiatum, the main region receiving glutamatergic inputs from CA3 (Klausberger and Somogyi, 2008). However, the majority of parvalbumin-expressing interneurons in the hippocampus are basket cells (Miles et al., 1996; Gulyás et al., 1999). Some basket cells can also express cholecystokinin and vasoactive intestinal protein (VIP) (Acsády et al., 1996; Lawrence, 2008). Parvalbumin+ basket cells preferably innervate cell bodies and very proximal dendrites of pyramidal cells in stratum pyramidale and are named after their basket- like appearance around pyramidal cell somas (Klausberger and Somogyi, 2008). They are fast- spiking and capable to fine-tune pyramidal cell firing in high frequencies, making this cell type one of the most studied perisomatic interneuron and a prominent model for proximal inhibition and gamma synchrony (Cobb et al., 1995; Wang and Buzsáki, 1996; Freund and Katona, 2007).

somatostatin+ temporoammonic pathway

lacunosum moleculare parvalbumin+

cholecystokinin+ Schaffer collateral (SC) pathway radiatum vasoactive intestinal peptide+ nitric oxide synthase+ pyramidale / neuropeptide Y+ oriens excitatory Pyramidal cell Bistratified cell Ivy cell SC-associated cell inhibitory Basket cell Axo-axonic cell Neurogliaform cell OLM cell axonal arborization

Figure 1.5: Hippocampal interneurons and their axonal distribution. The differently colored cell bodies show the main but not exclusive neurochemical content for each interneuron.

9 On the contrary, distal inhibition is mainly supplied by OLM cells and neurogliaform cells in stratum lacunosum moleculare (Fig. 1.5) (Klausberger and Somogyi, 2008). Neurogliaform cells in the hippocampus look fairly similar to neurogliaform cells in the neocortex (Tamás et al., 2003), possessing a very small cell body with radiating dendrites and very dense axons (Klaus- berger, 2009). Sometimes grouped together with ivy cells as interneurons with densely packed axons (Klausberger, 2009), the molecular properties of neurogliaform cells and ivy cells can be characterized by the expression of neuronal nitric oxide synthase and the neuropeptide Y (Price et al., 2005; Fuentealba et al., 2008; Bloss et al., 2016). While ivy cells mainly target apical dendrites and neurogliaform cells preferably target tuft dendrites, both cell types innervate sub- branches randomly, i.e. do not show a connectivity preference to the branch origin, center or end (Bloss et al., 2016). Differentially, OLM cells are somatostatin+ (but can also contain parvalbumin) and target terminal tuft dendrites of pyramidal cells in a non-random way, i.e. OLM cells preferably target the branch ends and to a lesser degree intermediate or branch origins of pyramidal cell dendrites (Müller and Remy, 2014; Bloss et al., 2016). Named after their anatomical location with the cell body residing in stratum oriens and extensive axonal projections to stratum lacunosum moleculare, OLM cells ﬁre strongly phase-locked to theta oscillations but do not show coupling to gamma (Klausberger et al., 2003; Klausberger and Somogyi, 2008). The main input to OLM cells comes from local pyramidal cells and medial septum, the main output goes back to neighboring pyramidal cells providing feedback inhibition on distal dendrites (Sik et al., 1995; Pastoll et al., 2013; Müller and Remy, 2014). With this mechanism OLM cells appear to counteract pyramidal cell overexcitation. Recently, OLM cells have been shown to gate CA1 inputs by inhibiting temporoammonic inputs while disinhibiting Schaffer collateral (SC) synapses via the inhibition of SC-associated cells in stratum radiatum (Leão et al., 2012), making this cell type a prominent model for theta synchrony and distal dendrite inhibition.

Pacemaker currents in distal dendrites Being the main target of OLM cells, the distal dendrites of pyramidal cells are important for synaptic integration and consist of a wide range of voltage-gated ion channels (Häusser et al., 2000; Magee, 2000; Spruston, 2008). Besides sodium channels, calcium channels and NMDA receptors, pyramidal cell dendrites express a high density of hyperpolarization-activated cyclic nucleotide-gated (HCN) channels. HCN channels are voltage-gated ion channels, which are activated by hyperpolarization and render spatiotemporal integration and dendritic ﬁltering of postsynaptic potentials (Magee, 2000; Lörincz et al., 2002; Nolan et al., 2004; Vaidya and Johnston, 2013). These channels are further involved in subthreshold resonance and are able to change neuronal excitability, by regulating resting membrane potential and input resistance (Maccaferri and McBain, 1996; Magee, 2000; Lupica et al., 2001; Cobb et al., 2003; Santoro and Baram, 2003; Zemankovics et al., 2010). The current through HCN channels is a mixed Na+ and K+, depolarizing current, named Ih and sometimes also referred to as pacemaker current because of its ability to generate rhythmic activities, e.g. in the heart (Pape, 1996; Lüthi and McCormick, 1998; Robinson and Siegelbaum, 2003). Four different HCN isoforms can be distinguished, with HCN1 being the fastest and activated 10-20 mV more positive than HCN2, and HCN4 being the slowest (Chen et al., 2001). HCN1 is mainly found in hippocampal and cortical regions, whereas HCN4 is more expressed in subcortical regions. HCN2 and HCN3 are

10 expressed in various brain regions, but at a lower level (Moosmang et al., 1999; Santoro, 2000). All isoforms are facilitated by the direct binding of the second messengers cAMP, cCMP, cGMP and cUMP (cyclic adenosine/cytosine/guanosine/uridine monophosphate), but HCN4 shows the highest sensitivity to cAMP and cGMP, HCN1 the least (Zong et al., 2012). The activation of HCN channels causes changes in the spike timing and alters synchronization properties (Magee, 1999; Rotstein et al., 2005). While in pyramidal cells unsynchronized excitatory inputs cause only weak dendritic spikes and diminish with sufficient inhibition, synchronized excitatory inputs can form strong dendritic spikes, tuned by the recruitment of the HCN channels (Fan et al., 2005; Müller et al., 2012). Thereby, Ih alters dendritic excitability and shortens excitatory postsynaptic potentials by counteracting the spatiotemporal filtering of neuronal dendrites (Magee, 2000; Cobb et al., 2003). When blocking Ih this compensatory mechanism gets lost, causing increased synaptic summation in the pyramidal cell dendrites (Magee, 1999). Interestingly, in computer simulations investigating theta synchrony, mutually coupled, single-compartment neurons lacking Ih appear to synchronize preferably with inhibition while cells containing Ih favor cell synchronization through excitation (Acker et al., 2003; Netoff et al., 2005; Rotstein et al., 2005). How the amount of Ih influences synchrony in a neuronal network with multiple cell types and the frequency preference of these cells was subject of Study I.

In Study I we employed a computational model to individually investigate how the different scaling of Ih can alter the spike timing of pyramidal cells, basket cells and OLM cells as well as showed how Ih ﬁne-tunes the power and frequencies of theta and gamma oscillations in a hippocampal network.

Cellular subthreshold oscillators in theta The question arising next is, if theta and gamma synchrony are pure network phenomena or produced from resonating cells, i.e. cellular oscillators? While inhibition-based gamma is stable in various cell types and mainly dependent on conduction delays (White et al., 2000; Kopell et al., 2000; Sohal and Huguenard, 2005), theta synchrony is often attributed to theta resonance in pyramidal cells and OLM cells (Pike et al., 2000; Gloveli et al., 2005; Zemankovics et al., 2010). Interestingly, theta oscillations can be generated within the hippocampus itself and do not need any external input from medial septum or entorhinal cortex (Goutagny et al., 2009; Huh et al., 2016), suggesting that this internal theta could be produced by neuronal oscillators. Sub- threshold resonance may be reflected in the spiking output and stems from rhythmic fluctuations between the interior and the exterior of a cell, the dendritic filtering of Ih as well as inductive and capacitive properties of the cell membrane (Pike et al., 2000; Hu et al., 2009; Wang, 2010).

The membrane of a neuron can be described as a capacitance (C in farad [F]), representing the lipid bilayer, and some conductances (G in siemens [S]), representing the voltage-gated and leak ion channels (Hodgkin and Huxley, 1952). Inductance (L in henry [H]) is sometimes ne- glected in neuronal simulations but for example can be modeled as signal transmission along an electrical cable. Resistance, the inverse of conductance (R=1/G in Ohm [Ω]) is a concept used for DC (direct current) while impedance is the AC (alternating current) equivalent. Providing an input current in form of a linearly increasing (e.g. 1 Hz/s) sine wave with constant amplitude

11 (chirp stimulus) to a resistor will result in a constant impedance profile, equal to the resistance of the resistor (Fig. 1.6a). Applying the same input to a parallel circuit of a resistor and a capacitor (RC circuit) results in a more complex impedance, which declines with increasing frequency, similar to a low-pass filter (Fig. 1.6b). Adding an inductor to the circuit, finally results in an impedance profile similar to a bandpass because of the low-pass filtering of the RC circuit and the high-pass filtering of the RL circuit, which has an impedance profile that rises with increasing frequency (Fig. 1.6c). Impedance (Z=R+jX) is a complex number consisting of resistance (real part) and reactance (imaginary part) with reactance being either positive or negative, depending on the presence of inductive or capacitive elements (Narayanan and Johnston, 2008). If the output voltage in response to the sine wave leads the input current, reactance is positive, if it lags the current, reactance is negative (Narayanan and Johnston, 2008). A reactance of zero means that the voltage and the current are in phase, i.e. that there are no inductive or capacitive elements (Narayanan and Johnston, 2008). The impedance for certain frequencies can be calculated by dividing the Fourier spectrum of the output (V per frequency) by the Fourier spectrum of the input (I per frequency). A peak in the impedance profile highlights the frequency at which the output voltage to the input current is highest. Thus, in response to a chirp stimulus resonant neurons exhibit a peak in the impedance profile close to their resonance frequency and smaller responses at other frequencies (Fig. 1.6d; Hutcheon and Yarom, 2000).

Figure 1.6: Schematic explaining how the same type of input current (chirp, from low to high frequency) alters the voltage output depending on the type of circuit (resistor (R), capacitor (C), inductor (L)) for an R (a), RC (b), RLC (c) circuit or a neuron (d) (Hutcheon and Yarom, 2000).

12 While fast-spiking interneurons show varying and comparably little resonance (c.f. impedance proﬁles), pyramidal cells and OLM cells preferably resonate at theta frequency (Pike et al., 2000; Zemankovics et al., 2010). In pyramidal cells two forms of electrical resonance at theta frequencies coexist: the somatic M-resonance produced by a non-inactivating K+ current (M- current IM) and persistent Na+ current as well as the dendritic H-resonance generated by Ih (Hu et al., 2002; Hu et al., 2009). Somatic M-resonance is strong at depolarized potentials, dendritic H-resonance during hyperpolarized potentials (Hu et al., 2002). Blocking of IM with the K channel blocker XE991 or Ih with the HCN channel blocker ZD7288 abolishes M-resonance or H-resonance respectively, indicating two spatially separated resonator mechanisms (Hu et al., 2002; Hu et al., 2009). Additionally to the resonance difference along the somatodendritic axis, pyramidal cells also differ in H-resonance along the dorsoventral axis of hippocampus, with ventral pyramidal cells displaying a higher subthreshold resonance frequency and impedance during hyperpolarized potentials than dorsal pyramidal cells (Marcelin et al., 2012a; Dougherty et al., 2013). Compared to pyramidal cells very little is known about OLM cells and their contribution of Ih. Recently, we found that the OLM cell marker Chrna2 (cholinergic receptor nicotinic alpha 2), displays a higher density in ventral hippocampus than in dorsal hippocampus (Leão et al., 2012; Mikulovic et al., 2015), suggesting that the OLM cell population along the dorsoventral axis is not homogeneous. How Ih differs in dorsal and ventral OLM cells is subject of Study II.

In Study II we patch-clamped OLM cells of Chrna2-Cre mice to investigate the differential generation of Ih along the dorsoventral axis and found that dorsal OLM cells have higher subthreshold resonance frequency and impedance than ventral OLM cells.

1.4 Neocortical Microcircuits partly resemble that of hippocampus

Similar to the hippocampus, the neocortex possesses a huge variety of GABAergic interneurons (Fig. 1.7) (Markram et al., 2004; Ascoli et al., 2008). Although it remains unknown if there are any homologs of the hippocampal interneurons in the neocortex, striking similarities exist. The main interneuron class targeting proximal dendrites and soma of pyramidal cells are parvalbumin+ basket cells, which can be further subdivided into small, nest and the classical large basket cells (Markram et al., 2004). Analog to their hippocampal counterpart, cortical basket cells are fast-spiking, show little spike frequency adaptation and facilitate gamma oscillations (Whittington et al., 1997). Other proximal dendrite-targeting interneurons in the neocortex are neurogliaform cells, double-bouquet, bitufted and bipolar cells (Fig. 1.7) (Markram et al., 2004). Interestingly, neurogliaform cells of the neocortex are morphologically and chemically similar to hippocampal neurogliaform cells, but developmentally they are derived from different ganglionic eminence, suggesting more subtypes of nitric oxide synthase+ and neuropeptide Y+ interneurons which are distributed in deeper layers as well as superﬁcial layers (Tamás et al., 2003; Fishell and Rudy, 2011; Jaglin et al., 2012). Also hippocampal bistratiﬁed cells do not appear to have a homologous population in the neocortex (Fishell and Rudy, 2011). More striking are the similarities between axo-axonic cells in the hippocampus and Chandelier cells of the neocortex which are both parvalbumin+, derived from the medial ganglionic eminence and exclusively inhibit pyramidal cells on the axon-initial segments (Fishell and Rudy, 2011). Finally,

13 the group of cholecystokinin+ interneurons in the hippocampus is quite diverse (Klausberger et al., 2005), thus comparisons with cortical cholecystokinin+ interneurons are not straightforward. However, cortical cholecystokinin+ cells can be divided into vasoactive intestinal protein (VIP) cells and non-VIP cells (Gonchar et al., 2008; Xu et al., 2010; Naka and Adesnik, 2016). VIP cells are sometimes also called bipolar cells, they express the 5HT3a receptor and make most of their contacts with somatostatin+ Martinotti cells, forming a strong disinhibitory circuit in the neocortex, i.e. VIP cells effectively inhibit Martinotti cells and therefore can switch off Martinotti cell inhibition onto pyramidal cells (Karnani et al., 2016b).

somatostatin+

L1 parvalbumin+

cholecystokinin+ L2/3 thalamo-cortical input vasoactive intestinal peptide+ L4 cortical output nitric oxide synthase+ L5 / neuropeptide Y+

L6 excitatory

Pyramidal cells Bitufted cell Bipolar cell Double bouquet cell inhibitory Basket cell Chandelier cell Neurogliaform cell Martinotti cell axonal arborization

Figure 1.7: Neocortical interneurons and their axonal distribution. The differently colored cell bodies show the main but not exclusive neurochemical content for each interneuron. The focus of this ﬁgure lies on layer 5 interneurons, i.e. layer 2/3 Martinotti cells are omitted for simplicity.

Martinotti cells are probably the most prominent distal dendrite-targeting interneurons of the neocortex (Martinotti, 1889; Martinotti, 1890) and could be seen as the cortical “cousin” of OLM cells. Besides their neurochemical similarity, Martinotti cells resemble OLM cells in their low-threshold spiking fashion, adaptive ﬁring and inhibitory connections to distal dendrites of pyramidal cells (Fig. 1.7) (Wang et al., 2004). Due to missing morphological certainty and the lack of distinct markers or simply the focus on the ﬁring pattern, Martinotti cells are sometimes synonymously called low-threshold spiking cells or slow-inhibitory interneurons (Goldberg et al., 2004; Lee et al., 2013). Especially, the long axonal projections to layer 1, similar to the axonal projection of OLM cells to stratum lacunosum moleculare, distinguishes Martinotti cells from other neocortical interneurons (DeFelipe and Jones, 1988; Wang et al., 2004; Ma et al., 2006). The most frequently used genetic marker to identify Martinotti cells has been somatostatin, however varying morphological and electrophysiological properties suggest a multitude of somatostatin-expressing neurons, including certain Martinotti cells as well as non-Martinotti cells in layer 2/3 and 4 (Wang et al., 2004; Ma et al., 2006; Kapfer et al., 2007; Xu et al., 2013). The Martinotti cells that reside in the main cortical output layer 5 provide a motif called frequency-dependent disynaptic inhibition (FDDI) on neighboring pyramidal cells (Silberberg and Markram, 2007; Kapfer et al., 2007; Berger et al., 2009). This is an inhibitory mechanism where two or more pyramidal cells synchronize via one or a few intermediate Martinotti cells (Berger et al., 2010). FDDI can be triggered by a brief, high frequency burst in the presynap-

14 tic pyramidal cell (Silberberg and Markram, 2007; Berger et al., 2010). Thereby, burst ﬁring in a few pyramidal cells is sufﬁcient to evoke Martinotti cell spikes due to the high degree of convergent pyramidal cell-Martinotti cell connections as well as the facilitating nature of the synapse (Silberberg and Markram, 2007; Berger et al., 2010). At the same time, Martinotti cells are able to synchronize a multitude of postsynaptic pyramidal cells because of their high degree of divergent and reciprocal connections (Silberberg and Markram, 2007; Berger et al., 2009). Repetitive activation of Martinotti cells causes brief and small hyperpolarizing potentials in the pyramidal cell dendrites, inactivating the calcium spike initiation zone of pyramidal cells via GABAA-mediated dendritic inhibition (Murayama et al., 2009; Palmer et al., 2012). Although Martinotti cells have been shown to also contact apical oblique dendrites of pyramidal cells, simultaneous, precisely timed Martinotti cell activity especially on the tuft dendrites seems to powerfully shape pyramidal cell spiking (Silberberg and Markram, 2007; Berger et al., 2010). Moreover, the synchronization in the pyramidal cell-Martinotti cell-pyramidal cell microcircuit can be disrupted when blocking Ih with ZD7288 due to the increased synaptic summation of inhibitory postsynaptic potentials in the pyramidal cell dendrites (Berger et al., 2010). While Martinotti cells have been reported to express Ih (Wang et al., 2004), the Martinotti cells mediating FDDI did not show a prominent Ih-mediated sag in response to negative current inputs (Berger et al., 2010). The pyramidal cells participating in FDDI however, exhibit a prominent Ih-mediated sag and increased HCN channel density on distal dendrites, suggesting that Mar- tinotti cells, similar to OLM cells, could synchronize pyramidal cell spiking via the activation of Ih (Berger et al., 2001; Kole et al., 2006; Berger et al., 2010).

Cortical pyramidal cell diversity While originally considered much less diverse than interneurons, recent studies have shown that not all pyramidal cells mediate a disynaptic circuit with Martinotti cells (Le Bé et al., 2007) and suggest the existence of different types of pyramidal cells (Hattox and Nelson, 2007; Brown and Hestrin, 2009; Harris and Shepherd, 2015; Kim et al., 2015; Jiang et al., 2015; Barth et al., 2016). Layer 5 pyramidal cells for example, can be electrophysiologically distinguished as type A and type B pyramidal cells by their expression of Ih (Lee et al., 2014). Upon negative current injections, type A pyramidal cells respond with a prominent hyperpolarization sag and pronounced rebound afterdepolarization (Lee et al., 2014). Positive current injections cause burst-regular spiking and large afterhyperpolarizations in type A pyramidal cells (Lee et al., 2014). Differentially, type B pyramidal cells show little afterhyper- and afterdepolarization as well as small a hyperpolarization sag in response to positive or negative current steps respectively (Lee et al., 2014). Morphologically, type A pyramidal cells show a large cell body and thick-tufted basal dendrites with apical dendrites extensively branching in layer 1, while type B pyramidal cells have a smaller soma, thin-tufted basal dendrites and limited spreading apical dendrites (Lee et al., 2014). Thick-tufted pyramidal cells are also named tall-tufted and sometimes referred to as L5B pyramidal cells, while thin-tufted pyramidal cells are synonymously called slender-tufted and usually named L5A pyramidal cells (Kim et al., 2015; Ramaswamy and Markram, 2015). Since different pyramidal cell morphologies are often associated with different connectivity patterns, thick-tufted pyramidal cells are also named cortico-subcortical neurons, subcerebral projection neurons or pyramidal tract neurons whereas thin-tufted pyramidal cells

15 are sometimes named cortico-cortical neurons, callosal projection neurons or intratelencephalic neurons (Hattox and Nelson, 2007; Brown and Hestrin, 2009; Harris and Shepherd, 2015; Kim et al., 2015; Jiang et al., 2015; Barth et al., 2016). Further distinctions are cortico-tectal, cortico- collicular or cortico-spinal (e.g. in motor areas) for type A pyramidal cells and cortico-striatal or cortico-non-striatal for type B pyramidal cells (Fig. 1.8, top; Hattox and Nelson, 2007; Brown and Hestrin, 2009; Harris and Shepherd, 2015; Kim et al., 2015; Jiang et al., 2015; Barth et al., 2016).

Since electrophysiological differentiation alone is not sufficient to distinguish pyramidal cell subtypes, e.g. L5B pyramidal cells can be regular spiking as well as bursting, genetic markers in combination with connectivity patterns are needed for clear definitions of pyramidal cell subgroups (Connors and Gutnick, 1990; Kim et al., 2015; Harris and Shepherd, 2015). To date, a multitude of genetic markers has been examined, however, a clear correlation between spiking properties is still lacking, as seen for example in firing properties, ranging from bursting (subset) to invariably intrinsically bursting for cortico-subcortical pyramidal cells (c.f. Harris and Shepherd, 2015; Kim et al., 2015). Especially the classical markers COUP-TF-interacting protein 2 (Ctip2) for subcerebral projection neurons and special AT-rich sequence binding protein 2 (Satb2) for callosal projection neurons are postnatally co-expressed in layer 5 pyramidal cells, making genetic distinctions cumbersome (Harb et al., 2016). However, considering morphological, electrophysiological and synaptic properties together revealed that thick-tufted, cortico- subcortically projecting pyramidal cells mediate a disynaptic feedback loop (a combination of feedback and feedforward inhibition; Fig. 1.8, bottom), while thin-tufted, cortico-callosally projecting pyramidal cells do not show this disynaptic motif (Le Bé et al., 2007). Thus, these findings suggest that interneurons might not non-specifically target pyramidal cells and disynaptic inhibition of Martinotti cells preferentially occurs in thick-tufted, putative type A pyramidal cells.

Type A Pyramidal cells Type B Pyramidal cells

cortical collicular tectal

striatal spinal

Feedback Inhibition Feedforward Inhibition Disynaptic Feedback Loop

Figure 1.8: Pyramidal cell classes and their connectivity motifs. The schematic drawings (top) show the projections of type A (left) and type B (right) pyramidal cells in the sagittal plane. Putative type A pyramidal cells appear to mediate a disynaptic feedback loop (bottom, right), consisting of a feedback inhibition (bottom, left) and a feedforward inhibition (bottom, middle) motif.

16 Cholinergic modulation and beta frequency

The neocortex is modulated by several types of neuromodulatory signals, such as noradrenaline, dopamine, histamine, serotonin and acetylcholine. Thereby, acetylcholine is particularly interesting because disynaptic inhibition in the cortex has been shown to be mediated by cholinergic axons (Arroyo et al., 2012). In particular, the cholinergic modulation of low-threshold spiking interneurons in layer 5 has been shown to generate oscillations in the beta frequency range (beta2, around 25 Hz) in layer 5 pyramidal cells with subcortical origin. (Roopun et al., 2006; Roopun et al., 2010). Interestingly, these layer 5 beta2 oscillations are insensitive to the muscarinic antagonist atropine, but sensitive to the nicotinic receptor antagonist d-Tubocurarine, suggesting a specific role for interneurons expressing nicotinic receptors (Roopun et al., 2010). The deep layer beta2 rhythm further coexists with superficial gamma (around 40 Hz) and can combine to form beta1 oscillations (around 15 Hz) through interlaminar interactions (Kramer et al., 2008; Roopun et al., 2008). Computational models suggest that beta1 activity results from the period concatenation of the two frequencies, i.e. gamma period (25 ms)+beta2 period (40 ms)=beta1 period (65 ms) (Kramer et al., 2008; Kopell et al., 2011). With the resulting beta1 frequency Martinotti cells are able to coordinate pyramidal cell activity in multiple cell assemblies and nest gamma activity within beta1 oscillations via feedback connections, similar to the nesting of gamma within theta oscillations by OLM cells (Kramer et al., 2008; Kopell et al., 2011; Li et al., 2013; Lee et al., 2015). Finally, computational studies also suggest that distal oscillatory inhibition in beta can change the firing pattern of layer 5 pyramidal cell from bursting to non-bursting (Li et al., 2013). However, to date the missing genetic marker to exclusively target layer 5 Martinotti cells, as well as optogenetic approaches to drive populations of Martinotti cells have hindered to investigate these computational outcomes experimentally.

In Study III we used Chrna2-Cre mice to show that Martinotti cells express the cholinergic receptor nicotinic alpha 2 and by using optogenetics we demonstrated that Martinotti cells reciprocally connect to one type of pyramidal cells and synchronize their action potentials via brief bursts in beta.

1.5 Aims of Study I, II and III

Overall, the goals of Study I, II and III are to investigate how distal dendrite-targeting interneurons promote the generation of synchrony. Because of their unique anatomical position, distal dendrite-targeting interneurons are prone to coordinate pyramidal cells and synchronize their activities. Study I, II and III provide information on how distal dendrite-targeting interneurons of hippocampus and neocortex can control neuronal spike timing in order to form a functional output - either by providing a certain level of Ih (Study I), generating subthreshold resonance (Study II) or mediating burst ﬁring in a certain frequency (Study III). In detail, Study I shows how the scaling of Ih alters the spike timing of hippocampal neurons and tunes theta/gamma oscillations. Study II reports how subthreshold resonance and Ih differ in dorsal and ventral OLM cells. Study III shows how Martinotti cells can synchronize pyramidal cells via brief beta bursts.

17 Studies This thesis is based on the following studies, which are referred to by their Roman numerals. I. Neymotin SA, Hilscher MM, Moulin TC, Skolnick Y, Lazarewicz MT, Lytton WW. (2013) Ih Tunes Theta/Gamma Oscillations and Cross-Frequency Coupling In an In Silico CA3 Model. PLoS One, Oct 18;8(10):e76285; (Neymotin et al., 2013)

II. Hilscher MM, Leão RN, Mikulovic S, Kullander K, Leão KE. A dorsoventral gradient of hyperpolarization-activated current (Ih) in hippocampal CA1 OLM interneurons. Manuscript

III. Hilscher MM, Leão RN, Edwards SJ, Leão KE, Kullander K. (2017) Chrna2-Martinotti Cells Synchronize layer 5 type A Pyramidal Cells via Rebound Excitation. PLoS Biol, Feb 9;15(2):e2001392; (Hilscher et al., 2017) Reprints were made with permission from the respective publishers.

Additional publications i. Hilscher MM, Leão KE, Leão RN. (2013) Synchronization through nonreciprocal connections in a hybrid hippocampus microcircuit. Front. Neural Circuits, Jul 23;7:120; (Hilscher et al., 2013)

ii. Patra K*, Lyons DJ*, Bauer P, Hilscher MM, Sharma S, Leão RN, Kullander K. (2015) A role for solute carrier family 10 member 4, or vesicular aminergic-associated transporter, in structural remodelling and transmitter release at the mouse neuromuscular junction. Eur. J. Neurosci., Feb;41(3):316-27 *Equal contribution; (Patra et al., 2015)

iii. Perry S, Gezelius H, Larhammar M, Hilscher MM, Lamotte d’Incamps B, Leão KE, Kullander K. (2015) Firing properties of Renshaw cells deﬁned by Chrna2 are modulated by hyperpolarizing and small conductance ion currents Ih and ISK. Eur. J. Neurosci., Apr;41(7):889-900; (Perry et al., 2015)

iv. Mikulovic S, Restrepo CE, Hilscher MM, Kullander K, Leão RN. (2015) Novel markers for OLM interneurons in the hippocampus. Front. Cell. Neurosci., Jun 2;9:201; (Mikulovic et al., 2015)

v. Dagostin AA, Lovell PV, Hilscher MM, Mello CV, Leão RM. (2015) Control of Phasic Firing by a Background Leak Current in Avian Forebrain Auditory Neurons. Front. Cell. Neurosci., Dec 10;9:471; (Dagostin et al., 2015)

Poster presentation i. Hilscher MM, Moulin T, Skolnick Y, Lytton WW, Neymotin, SA. (2012) Synchronization through nonreciprocal connections in a hybrid hippocampus microcircuit. BMC Neurosci., 13 (Suppl 1), P80; (Hilscher et al., 2012)

18 CHAPTER 2 General Methods

2.1 Computer simulations

In Study I we performed computer simulations in order to individually adjust the levels of Ih conductance for specific hippocampal cell populations. We used conductance-based models which exhibit a good trade-off between beeing biophysically realistic and not excessively computa- tionally intensive. The conductance-based model consisted of 800 five-compartment pyramidal cells, 200 one-compartment basket cells and 200 one-compartment OLM cell in order to account for an appropriate ratio of cell types (Wang and Buzsáki, 1996; Tort et al., 2007; Neymotin et al., 2011). All cells contained leak current, transient sodium current INa and delayed rectifier current IKDR, to allow action potential generation. Pyramidal cells further had a potassium type A current IKA for rapid inactivation in all compartments. OLM cells additionally contained a simple calcium-activated potassium current IKCa to allow long lasting inactivation after bursting, high-threshold calcium current IL to activate IKCa, and intracellular calcium concentration dynamics (McCormick and Huguenard, 1992; Tort et al., 2007; Stacey et al., 2009). The hyperpolarization-activated current Ih in pyramidal cells and interneurons was based on HCN1 and HCN2 isoform parameterization (Bender et al., 2001; Santoro and Baram, 2003; Migliore et al., 2004; Aponte et al., 2006).

The simulations were run on a Linux system with eight 2.27 GHz quad-core Intel Xeon CPUs using NEURON (Carnevale and Hines, 2006). Eight seconds of simulation took about 2.2 minutes to run. In order to assess the robustness of the results, we ran each simulation condition with six different randomizations of synaptic inputs, and six different randomizations of network connectivity. The simulations were run in the NEURON simulation environment with python interpreter, multithreaded over 16–32 threads (Carnevale and Hines, 2006; Hines et al., 2009). Analysis of the simulation data was done with the Neural Query System (Lyt- ton, 2006) and Matlab (Mathworks, Inc.). The full model is available online on ModelDB: https://senselab.med.yale.edu/modeldb/showModel.cshtml?model=151282.

19 2.2 Animal model

In Study II and Study III we used transgenic mice (both males and females, P19-29), with cyclization recombinase (Cre) coupled to the Chrna2 gene (Chrna2-Cre) (Leão et al., 2012; Perry et al., 2015; Mikulovic et al., 2015). Chrna2-Cre mice were routinely crossed with a tdTomato fluorescence reporter line Gt(ROSA)26Sortm14(CAG-tdTomato)Hze (R26tom; Allen Brain Institute) to mark OLM cells and Martinotti cells with a red fluorescent protein and simplify the identification process. For Study III Chrna2-Cre mice were crossed with a Halorhodopsin- expressing line Rosa26-eNphR-EYFP (HaloR; Jackson Laboratory Stock No. 014539) to inhibit HaloR-expressing Chrna2 cells. HaloR acts as a light-driven chloride pump and green-yellow light causes the transport of Cl- into the cell, changing the cell’s membrane potential towards the chloride potential. Age matched, Cre-negative littermates were routinely used as controls.

2.3 Optogenetics

In Study III we injected Chrna2-Cre/R26tom mice (1-2 months old) with Channelrhodopsin (ChR2). ChR2 is a light-gated ion channel and blue light transports H+, Na+,K+ or Ca2+ into the cytosol which depolarizes the neuron. Chrna2-Cre/R26tom mice were anesthetized with Isofluran (1-4%), placed on a heat pad and the head fixed with a nose holder and ear bars in a stereotaxic frame (Stoelting Co.). The skin was cleaned with iodine, opened with a straight incision, and the bregma identified using small amount of peroxide. The coordinates for bilateral virus injection were: AP: -2.46 mm, ML: +/-4.00 mm and DV: 2.00/2.50 mm. A small hole was drilled in the skull using a dental micro drill causing minimal bleeding during the process. Viral vectors (pAAV-EF1a-double floxed-hChR2(H134R)-EYFP-WPRE-HGHpA, University of Pennsylva- nia Vector Core Facility), in solution (6.2x1012 / 1.6x1013 particles/ml) of 0.50-1.00 µl was slowly infused (0.10 µl/min) into the auditory cortex at two depths (2.00 and 2.50 mm) using a Hamilton 10 µl syringe and an infusion/withdraw pump (World Precision Instruments / KD Scientific). After infusion, the needle was left in place for 1-5 min to allow complete diffusion of the virus. Next the scalp was rehydrated with saline and sutured with 4-5 stitches and local anesthesia (a drop of Lidocaine/Marcaine) applied onto the sutured skin. The animal was monitored and kept warm until fully awake (moving and starting to eat and drink water). Mice were sac- rificed after approx. 3-6 weeks for in vitro electrophysiological experiments and/or histological procedures.

2.4 Electrophysiology

For Study II and III 300-µm thick coronal brain slices from Chrna2-Cre/R26tom transgenic mice were obtained containing hippocampus (Study II) or primary auditory cortex (Study III). In general, the Chrna2-Cre mice brains were rapidly removed and placed in ice-cold sucrose/artiﬁcial cerebrospinal ﬂuid (ACSF) consisting of the following (in mM): KCl, 2.49; NaH2PO4, 1.43; NaHCO3, 26; glucose, 10; sucrose, 252; CaCl2, 1; MgCl2, 4. Slices were cut 300-µm thick using a vibratome (VT1200, Leica, Microsystems) and subsequently moved to a submerged holding chamber containing normal ACSF (in mM): NaCl, 124; KCl, 3.5; NaH2PO4, 1.25;

20 MgCl2, 1.5; CaCl2, 1.5; NaHCO3, 30; glucose,10. Solutions were constantly bubbled with 95% ◦ O2 and 5% CO2. Slices were kept at 35 C for 1 h and then maintained at room temperature. The slices were transferred to submerged chamber under an upright microscope equipped with DIC optics (Olympus) and perfused with at 30◦C oxygenated ASCF (1–1.25 ml/min). For some experiments, carbachol (10 µM, Sigma-Aldrich; Study III), ZD7288 (20 µM, Tocris Cookson Inc; Study II and Study III) or tetrodotoxin (1 µM, Tocris Cookson Inc; Study II), were added to the perfusate. Patch pipettes from borosilicate glass capillaries (GC150F-10 Harvard Apparatus) were pulled on a vertical puller (Narishige, Japan) with resistance around 7 MΩ. Pipettes were ﬁlled with internal solution containing (in mM): K-gluconate,130; NaCl, 7; MgCl2, 2; ATP, 2; GTP 0.5; HEPES, 10; EGTA, 0.1 (pH was adjusted to 7.2 using KOH). Whole-cell current clamp recordings were acquired using a Multiclamp 700B ampliﬁer (Molecular Devices) or a Dagan BVC700 (Dagan) and digitized with a Digidata 1440A or a National Instruments data acquisition card. WinWCP and WinEDR softwares implemented by Dr.J. Dempster (University of Strathclyde, Glasgow, UK) were used to record electrophysiological signals. Light pulses (488 nm or 555 nm, Study III) were applied using a CoolLED pE-1 light source.

2.5 Imaging and Tracing

Cells targeted by electrophysiological experiments in Study II and III were routinely ﬁlled with biocytin (Sigma Aldrich; 1.5 mg/ml) and counterstained with Streptavidin-Alexa-Fluor (Invitro- gen; 488 nm) for post-hoc analysis. Images were collected on a Zeiss LSM 510 Meta confocal microscope and stacked and 2D-stitched using ImageJ 1.50a (NIH), where the color palette was adjusted for consistency (tdTomato - red, biocytin - green). Soma detection and neurite tracings were done semi-automated with NeuronJ 1.4.2 (ImageJ Plugin) or fully automated with Imaris 8.1 FilamentTracer (Bitplane, UK) using “Autopath” in the algorithm settings and the threshold mechanism to correct for over-/under-sampled tracings following the image intensity. The reconstruction of a Chrna2+ Martinotti cell from Study III is available online on NeuroMorpho: http://www.neuromorpho.org/neuron_info.jsp?neuron_name=c043.

2.6 Data analysis

Action potentials (APs) in Study II and III were usually triggered by 500 ms or 1000 ms depolarizing current injections (10 - 100 pA). The first fired AP in response to minimal current injection was analyzed for AP amplitude (peak to afterhyperpolarization voltage), threshold (where the change in membrane potential exceeds 20 mV/ms), half-width (halfway between threshold voltage and peak) and first spike latency (time between stimulus onset and the AP threshold of the first spike). Afterhyperpolarizations were analyzed for magnitude (AP threshold - minimum of voltage trough between the first and the second AP in a spike train) and latency (distance to the AP threshold) of the negative peak. Spike rate was defined as the number of APs per 1000 ms. Spike-frequency adaptation was calculated as the inverse of the mean of the last three interspike intervals (steady-state frequency) divided by the inverse of the first interspike interval (maximum frequency) in response to 100 pA current injections and subtracted from 100% (no adaptation = 0%). The instantaneous spike frequency (reciprocal of consecutive interspike intervals) was

21 plotted for each AP vs. the time after onset of the current pulse. Matlab (Mathworks, Inc.) was used for data analysis.

For the experiments using optogenetic methods in Study III, ChR2 stimulation frequencies (2, 5, 15, 25, 40 and 70 Hz) and HaloR-evoked hyperpolarization (5, 10, 100, 250, 500 and 1000 ms) were applied in randomized order to avoid statistical dependencies between cases. ChR2- triggered inhibitory postsynaptic potentials and frequency-dependent disynaptic inhibition were measured as amplitude, time to peak and half decay time where the onset was defined as the time at which the potential exceeded three times the standard deviation of the preceding baseline. HaloR-evoked hyperpolarization amplitudes were quantified as the difference between resting membrane potential and the peak of hyperpolarization. Rebound APs were quantified by number, maximum frequency and duration (time of last - time of first rebound AP). Burst- spiking pyramidal cells were distinguished from single-spiking pyramidal cells by obtaining the interspike intervals, where burst-spiking pyramidal cells showed an increased amount of short (≤ 20 ms) interspike intervals at near-threshold potentials.

APs from cells were detected based on threshold (> -20 mV). Pyramidal cell spike trains were transformed into a series of 0 (no spike) and 1 (spike) with 0.1 ms-precision (binning). Accord- ingly, kernel density estimates were computed on the population data (all patched pyramidal cells) with 0.1 ‘bandwidth’, i.e. smoothing (“ksdensity” command in Matlab). Power spectral density analysis of the binary spike series was made using Welch’s method (“pwelch” command in Matlab). Coherence was calculated pairwise (i.e. for simultaneously patched pyramidal cells the similarity between the binary sequences of the two pyramidal cells was calculated) and plot as mean over all recordings (“cohere” command in Matlab). Cross-correlograms were calculated using the “coeff” option of the cross-correlation command in Matlab (to scale the cross- correlation values from -1 to 1 and prevent dependency of the cross-correlation on the number of spikes) and then smoothed by a moving average filter with a span of 10 ms. We displayed cross-correlations over a lag range of ±1 s. The synchrony index was defined as the maximum peak of the normalized cross-correlograms between -50 ms and 50 ms, as previously described (Hilscher et al., 2013) with 0 =b no synchronization, 1 =b full synchronization. Mutual information of simultaneously recorded pyramidal cells (PCs) MI(PCleft, PCright) was determined by calculating the distributions of the spike trains (univariate distribution of each spike train separately as well as bivariate distribution of both spike trains) and expressing them as entropies H(PCleft) and H(PCright), meaning how “diverse” the spike trains were. The sum of the two entropies H(PCleft) and H(PCright) minus the joint entropy H(PCleft, PCright) quantified the conditional entropy, i.e. the mutual information MI(PCleft, PCright). The MI was formulated as a mutual information index with a scaling factor of 1000 (Hilscher et al., 2013).

Statistical analysis was performed using R version 3.2.3 (Foundation for Statistical Computing, Vienna, Austria). Data is reported as mean ± Standard Error of the Mean (SEM) and displayed as bar plots or box plots. Data larger than q3 + 1.5*(q3 – q1) or smaller than q1 – 1.5*(q3 – q1), with q1 and q3 denoting the 25th and 75th percentiles (see box plots), was considered as outlier and discarded. Statistical comparisons were determined using two-tailed Student’s t-test and to

22 account for multiple comparisons, the data was analyzed using ANOVA and post-hoc test with Tukey correction (* =b p < 0.05, ** =b p < 0.01, *** =b p < 0.001 and **** =b p < 0.0001).

2.7 Ethical statement

This thesis is the result of a cooperation between the Brain Institute, Federal University of Rio Grande do Norte, Natal, Brazil and the Department of Neuroscience, Uppsala University, Up- psala, Sweden. All animal experiments have been approved by the Comite de Etica no Uso de Animais da UFRN (CEUA 013/2012) and the Swedish Animal Welfare authorities, following the European Communities Council Directive (86/609/EEC) for the care and usage of laboratory animals. Ethics permits at Uppsala University: C132/13 and C135/14. Efforts were made to minimize the numbers of animals used.

2.8 Methodological considerations

We focused on in vitro preparations in Study II and Study III in order to investigate the cellular mechanisms of selected neurons and/or populations of neurons isolated from the intact nervous system. This approach is advantageous when determining the electrophysiological properties of single cells and studying their connectivity as well as mechanisms for spike-to-spike synchronization. However, some considerations have to be taken into account. For example, slicing angle and tissue thickness are of central interest when studying connectivity. In Study III we used 300-µm thick coronal brain slices in order to increase the likelihood of connected pyramidal cells in layer 5 and preserved distal inhibition by Martinotti cells in layer 1. Initially, the same set of experiments were also carried out in coronal slices of primary motor cortex, secondary visual cortex and medial prefrontal cortex (c.f. Berger et al., 2009). However, experiments in primary somatosensory cortex were done using parasagittal slices (10◦) in order to maintain a high probability of Martinotti cell-pyramidal cell connectivity. Thus, we optimized brain sectioning to the speciﬁc research question.

Another important factor for in vitro studies is the determination of recording temperature (Kush- merick, 2006). Changing temperature from near-physiological conditions (35-37◦C) to room temperature (22-24◦C) inﬂuences AP shape and size, as well as shape and size of postsynaptic potentials. As a consequence, lowering the recording temperature usually causes a decrease in ﬁring frequency. In our experiments in Study II and Study III, we adjusted the temperature of the in-line heater to 35-37◦C. However, due to the distance from the heater to the submerged holding chamber, the temperature usually drops ∼5◦, resulting in a temperature of the perfusate of ∼30◦C, as measured by the bath thermometer. This temperature on the other hand, is also a trade-off since recording conditions at near-physiological values increase cell metabolism, whereas recording conditions lower than near-physiological values have been shown to postpone the onset of cell apoptosis as well as lower swelling and damage of the brain slices (Thompson et al., 1985). Reduction of swelling and damage can also be achieved by adapting the solutions in which the brain slices are kept during cutting and recording, such using glycerol-based ACSF (Ye et al., 2006) or N-methyl-D-glucamine-based ACSF (Ting et al., 2014). This is particularly

23 important when preparing slices from adult animals. However, here we did not need to utilize these improved brain slice methods since we focused on P19-29 mice in Study II and Study III as well as 1-2 months old mice in Study III, still allowing straightforward patch clamp experiments.

In Study III we performed optogenetic experiments on 1-2 months old mice due to the application of viral vectors in order to express ChR2 and optogenetically stimulate neurons. Viral vectors driving ChR2 can be directly delivered into the brain region of interest and cell-type- specific expression is achieved by the Cre-loxP system, i.e. in Study III only Chrna2-Cre cells near the injection site (primary auditory cortex) expressed ChR2 due to the utilization of loxP- flanked/inverse ChR2. This ChR2 construct depends on Cre flipping the ChR2 reading frame in the correct direction before it can be transcribed and thereafter the ChR2 protein is expressed in a Cre-dependent manner. Especially double-floxed inverted open reading frame (DIO) viral vectors are “state of the art”, achieving high specificity and expression levels (Cardin et al., 2010). Due to steadily improving expression profiles and the raising availability of transgenic animals for optogenetic manipulation, we decided to use a HaloR-expressing mouse line for the optogenetic silencing of neurons (Zhao, 2008; Ting and Feng, 2013). Here, HaloR is also expressed in a Cre-dependent manner, but differently to the viral strategy not spatially specific, i.e. all Chrna2-Cre cells expressed HaloR (not only Chrna2+ cells in primary auditory cortex). To account for the differences of the ChR2 and HalorR experiments, two control experiments were performed. First, since HaloR-expressing Chrna2+ cells can also be found in the hippocampus, we separated neocortex from hippocampus in our slice preparation and performed experiments on brain slices without hippocampus. Second, we injected ChR2 only in hippocampal Chrna2+ cells and investigated whether the activation of hippocampal Chrna2+ cells alters neuronal firing in the neocortex (c.f. Study III).

Furthermore, it is important to note that inhibition via HaloR can cause a stimulation artifact after the offset of the stimulation (Ferenczi and Deisseroth, 2012). This artifact can result in a rebound-like depolarization generated by a shift in the GABAA reversal potential following prolonged (up to 15 s) HaloR activity (Raimondo et al., 2012). In Study III we made use of rebound APs following HaloR inhibition, however, the currents underlying the rebound APs have not been determined as HaloR was used as a tool to make Chrna2+ cells burst. Recently, it has been shown that the HaloR-induced rebound APs can be attenuated by ramp-like light termination (Mahn et al., 2016). Another option to optogenetically silence neurons is the usage of Archaerhodopsin (Arch). Arch acts as a light-driven proton pump and green-yellow light causes the transport of H+ out of the cell to hyperpolarize it. But similar to HaloR, Arch also shows some stimulation artifacts. Besides the relatively the strong inhibitory photocurrents, sustained activation of Arch has been shown to trigger pH-dependent Ca2+ inﬂux and as a result increased neurotransmitter release at presynaptic terminal. So far, HaloR has been reported to be the “most suitable existing tool for synaptic terminal silencing, although its use should be carefully controlled to account for changes in chloride reversal potential and strong light-off rebound responses” (Mahn et al., 2016). Thus, the usage of optogenetic tools in general and the choice of optogenetic inhibition in particular, should to be taken with caution and the consideration of alternatives should always be taken into account.

24 CHAPTER 3 Studies and Results

3.1 Ih Tunes Theta/Gamma Oscillations and Cross-Frequency Coupling In an In Silico CA3 Model

Background

The hyperpolarization-activated current Ih provides certain features which are able to change the synchronization properties of neurons (Acker et al., 2003; Rotstein et al., 2005). Involved in a neuron’s resonance frequency (Pike et al., 2000; Hu et al., 2002), Ih has been shown to alter membrane input resistance, resting membrane potential and dendritic excitability (Maccaferri and McBain, 1996; Magee, 2000; Lupica et al., 2001; Cobb et al., 2003; Santoro and Baram, 2003; Dyhrfjeld-Johnsen et al., 2009; Zemankovics et al., 2010). Dendritic ﬁltering of postsynaptic potentials inﬂuences spike timing and as a consequence the formation of synchronous activities and oscillations. Especially in pyramidal cells the differential expression of HCN channels along dendrites plays an important role for rhythmogenesis and suggests distinct contributions of perisomatic region-targeting interneurons and dendritic region-targeting interneurons (Bender et al., 2001; Accili et al., 2002; Santoro and Baram, 2003). Additionally, computational models studying theta synchrony propose that mutually coupled, single-compartment neurons containing Ih seem to favor synchronization through excitation while cells without Ih appear to synchronize better with inhibition (Acker et al., 2003; Netoff et al., 2005; Rotstein et al., 2005). In Study I, we explored the role of Ih in a network of 800 pyramidal cells, 200 perisomatic region-targeting basket cells as well as 200 dendritic region-targeting OLM cells and hypothesized that the amount of Ih alters power and frequencies of theta and gamma oscillations.

Experimental Question (Hypothesis)

Is hippocampal synchronization dependent on the amount of Ih and how do distal dendrite- targeting interneurons shape theta synchrony?

25 Outcome

We simulated a hippocampal model of CA3 and found that Ih altered theta and gamma oscillations, mostly by influencing the firing rate of the inhibitory cells. In general, reducing Ih in OLM cells decreases theta synchrony because of less depolarizing drive from Ih, lower excitability and lower firing in OLM cells. Furthermore, the decreased firing of OLM cells and the reduced OLM cell-pyramidal cell inhibition increases pyramidal cell spiking, which in turn results in a higher pyramidal cell-basket cell drive and as a consequence larger gamma synchrony due to the enhanced PING. Looking at frequency and power of theta and gamma separately, shows that increasing Ih in OLM cells causes increased OLM cell firing, concurrently decreased pyramidal cell spiking and as a result lower spectral power in the model. Power was measured as the sum of differences in membrane potential between the most distal apical and the basal dendritic compartment over all pyramidal cells. Although theta frequency rises with increasing Ih in OLM cells theta and gamma power can be moderately reduced or gradually decline (Fig. 3.1 left). On the contrary, increasing Ih in basket cells mainly affects gamma, with increases in gamma power but modest decreases in gamma frequency because of prolonged inhibition on pyramidal cells (Fig. 3.1 middle). Changing Ih in pyramidal cells mainly influences theta, with increasing Ih conductance resulting in increased theta frequency and power. Also gamma power is modulated to some extend via the enhancement of PING, but compared to theta power negligible (Fig. 3.1 right). Finally, increasing Ih across the cell types simultaneously increases gamma and theta power of the network but shifts gamma towards lower frequencies and theta towards higher frequencies.

OLM cells Basket cells Pyramidal cells

Ih conductance Ih conductance Ih conductance

Firing frequency Firing frequency Firing frequency

Theta power over Theta power over Theta power over pyramidal cells pyramidal cells pyramidal cells

Gamma power over Gamma power over Gamma power over pyramidal cells pyramidal cells pyramidal cells

Figure 3.1: Illustrative figure of the outcome of Study I. Altering Ih conductance individually for OLM cells (left), basket cells (middle) and pyramidal cells (right) shows how Ih changes the firing frequency of each cell type and theta or gamma power in the model. The increases and decreases in the figure are not proportional but rather highlight the direction of change.

Conclusion and Future Perspectives

Study I predicts that the modulation of Ih has several consequences, such as the regulation of cell excitability, the tuning of gamma and theta power and frequency, as well as the modulation

26 of cross-frequency coupling. Cross-frequency coupling characterizes the ability of theta inﬂu- encing gamma, i.e. by the low-frequency oscillation providing an envelope that modulates the amplitude of the superimposed high-frequency oscillation (Tort et al., 2010). Phase-amplitude coupling or nesting of gamma within theta is an important mechanism for memory organization, encoding and storage (Lisman and Idiart, 1995; Lisman, 2005), suggesting an essential role of Ih during memory formation. Indeed, recently a study showed that mechanisms for learning and memory underlie the regulation of HCN channels (Maroso et al., 2016). Another experimental study reported that decreased Ih activity, as seen for example in temporal lobe epilepsy, goes together with the reduction of theta resonance and phase lead in pyramidal cells (Marcelin et al., 2009), similar to our observations where decreased Ih shifted gamma towards later phases of the theta cycle.

Although computational models offer elegant possibilities to study the mechanisms for theta and gamma synchrony, these models often follow the principle of keeping it as simple as possible but as accurate as needed and contain various simplifications. Focusing on the network dynamics rather than the single neuron precision, our model included only reduced morphological descriptions of cell types. Furthermore, certain connectivity patterns were set to a fixed number of connections and activated randomly with each simulation. While these type of random- ization increases reliability of repeated simulations, synchronization in networks with sparse, random connectivity have certain constraints. Although not applicable here, it is noteworthy that in networks of mixed excitatory and inhibitory cells, the desynchronization effect of random connectivity can be reduced by simply strengthening the synapses between the excitatory cell class and the inhibitory cell class, while the opposite does not hold (Börgers and Kopell, 2003). Strengthening the synapses between the inhibitory cell class and the excitatory cell class does not reduce the desynchronization, the synchronization rather depends on the decay time constant of the inhibition (Börgers and Kopell, 2003). Different, but critical in our model is the fact that Ih changes the resting membrane potential of the cells, i.e. more Ih leads to a more depolarized resting membrane potential and consequently more spikes. While this is consistent with experimental findings (Lupica et al., 2001; Dyhrfjeld-Johnsen et al., 2009), the depolarized resting membrane potential is influencing the drive to OLM cells, basket cells and pyramidal cells respectively and as a consequence the theta and gamma oscillations. An interesting future experiment would be to investigate the role of Ih while compensating for the depolarizing drive, e.g. by adding extra hyperpolarizing current such as IM which appears to be complementary to Ih (Hu et al., 2002; Hu et al., 2009). For computational studies with greater morphological details than the current model, the inclusion of somatodendritic gradients of Ih and IM as well as the investigation of the non-random targeting of pyramidal cell dendrites by OLM cells (Bloss et al., 2016) would be exciting, in order to study how random targeting and non-random targeting in dendrites influences theta and gamma synchrony.

Another restricting factor for modeling studies is the translation of experimental data into math- ematical descriptions and availability of enough detail for a model. For example, we did not include second messenger mechanisms behind Ih in our model, simply because studying of second messenger cascades is challenging and information about their exact functions is scarce

27 (Maccaferri and Lacaille, 2003; Poolos et al., 2006; Fogle et al., 2007; Wahl-Schott and Biel, 2009; Migliore and Migliore, 2012; Zong et al., 2012; Maroso et al., 2016). Moreover, due to the lack of HCN channel subtypes for the different cell classes, we only simulated two subtypes of HCN. While pyramidal cells have been extensively studied (Brandt et al., 2009; Matt et al., 2011; Dougherty et al., 2012; Dougherty et al., 2013; Malik et al., 2016) and are known now to express HCN1 channels as well as HCN2 channels in their dendrites, the somatodendritic distribution of HCN subtypes in basket cells and OLM cells is poorly understood. Basket cells do not exhibit much sag in response to hyperpolarizing current injections, indicating very little expression of HCN channels (Pike et al., 2000; Zemankovics et al., 2010; Hughes et al., 2013). On the contrary, OLM cells seem to generate more Ih, by expressing HCN2 channels but not HCN1 channels on the soma (Matt et al., 2011) and HCN1 channels on the dendrites (Lopez et al., unpublished observations). Lastly, our model did not consider regional differences of HCN channel expression and their effects on synchrony, i.e. that ventral pyramidal cells in CA1 produce more Ih than dorsal pyramidal cells and thus potentially shape synchrony differently. The variation of Ih in OLM cells along the dorsoventral axis is the subject of Study II.

28 3.2 Dorsoventral gradient of hyperpolarization-activated current (Ih) in hippocampal CA1 OLM interneurons

Background Early electrophysiology experiments revealed two distinct types of theta oscillations in the hippocampus; theta1 being the movement related theta and theta2 being the immobility associated theta (Vanderwolf, 1969; Kramis et al., 1975; Sainsbury et al., 1987). A frequency of 7-12 Hz was assigned to type 1 theta, type 2 theta was deﬁned as a slightly lower frequency of 4-9 Hz. However, the mechanisms that generate theta oscillations are quite diverse and probably a combination of multiple processes: Among many others, the mechanisms most likely involved in the formation of theta are: 1) theta resonance of pyramidal cells and OLM cells (Pike et al., 2000), 2) the coupling and synchronization between pyramidal cells and OLM cells (Tort et al., 2007), 3) recurrent connections between pyramidal cells (Börgers et al., 2005) and 4) the activation of the pacemaker current Ih (Rotstein et al., 2005). Especially the subthreshold resonance of pyramidal cells and OLM cells governed by Ih is commonly linked to the emergence of theta oscillations (Pike et al., 2000; Gloveli et al., 2005; Zemankovics et al., 2010). Recently, patch clamp-experiments of pyramidal cells along the dorsoventral axis of hippocampus have shown that theta resonance in pyramidal cells is location-dependent (Marcelin et al., 2012b; Dougherty et al., 2013). Thereby, ventral pyramidal cells displayed a higher resonance frequency during hyperpolarized potentials and possessed more Ih than dorsal pyramidal cells (Dougherty et al., 2013). If the dorsoventral difference can be linked to the distinct theta1 and theta2 is not completely clear yet. Since pyramidal cells are strongly coupled with OLM cells, we opted to similarly investigate if OLM cells also show a dorsoventral difference of Ih and if subthreshold resonance varies between dorsal and ventral OLM cells. In Study II, we patched OLM cells of Chrna2-Cre mice (Leão et al., 2012; Perry et al., 2015; Mikulovic et al., 2015) and hypothesized that active as well as passive properties differ between dorsal and ventral OLM cells.

Experimental Question (Hypothesis) Are distal dendrite-targeting interneurons of dorsal and ventral hippocampus electrophysiologically different and how does the amount of Ih differ along the dorsoventral axis?

Outcome We performed whole-cell current- and voltage-clamp recordings in the hippocampus of Chrna2- Cre mice and found that OLM cells display different Ih-dependent features along the dorsoventral axis, opposite to the pyramidal cell trend. Passive and active electrophysiological properties suggest that dorsal OLM cells generate more Ih than ventral OLM cells. Ventral OLM cells expose a lower resting membrane potential as well as a lower membrane input resistance than dorsal OLM cells (Fig. 3.2). Furthermore, ventral OLM cells only produce a small membrane ‘sag’ in response to hyperpolarizing current injections, indicating that these cells have minimal Ih compared to dorsal OLM cells. On the contrary, dorsal OLM cells display a strong depolarizing membrane sag during negative current injections and pronounced rebound afterdepolarization

29 upon termination of the current pulse, which is able to generate rebound action potentials. Addi- tionally, both dorsal and ventral OLM cells show electrical subthreshold resonance close to the theta range, with dorsal OLM cells resonating at higher frequency bands than ventral OLM cells (Fig. 3.2). While at near-threshold potentials Ih is not active, at hyperpolarized potentials Ih facilitates theta resonance in the OLM cells. Finally, resonance and hyperpolarization-activated sag are suppressed with ZD7288, suggesting that the differential expression of HCN channels accounts for the dorsal and ventral differences in OLM cells during hyperpolarized potentials.

Ih Dorsal OLM cells

V rest + -

Rinp Ventral OLM cells

Dorsal Ventral

Figure 3.2: Illustrative ﬁgure of the outcome of Study II. Dorsal OLM cells exhibit more Ih as well as a higher resting membrane potential, input resistance and resonance frequency than ventral OLM cells.

Conclusion and Future Perspectives

Study II shows that Ih differs within the OLM cell population and predicts a differential expression of HCN channels along the dorsoventral axis. This is of special interest because in vivo data is mostly derived from the dorsal hippocampus while in vitro data usually comes from more ventral hippocampus (Maccaferri and McBain, 1996; Zemankovics et al., 2010; Dougherty et al., 2012). Dorsoventral gradients of Ih often go together with different roles of synaptic integration along the axis (Garden et al., 2008). For example, in the entorhinal cortex, the time constants of Ih differ along the dorsoventral axis which has been shown to account for grid ﬁeld organization and spacing (Garden et al., 2008; Giocomo and Hasselmo, 2008; Giocomo et al., 2011). Also in the auditory system dorsoventral gradients have been reported, such as in the medial superior olive (Baumann et al., 2013) or the nucleus laminaris (Yamada et al., 2005), coincid- ing with the tonotopic axes of these areas. Our study suggests that additionally to the pyramidal cells of the hippocampus, also the OLM cell population possesses a dorsoventral difference in Ih.

Thereby, ventral OLM cells show less pronounced Ih-dependent features than dorsal OLM cells, including a smaller hyperpolarization-activated sag and a more hyperpolarized resting membrane potential. Furthermore, the input resistance of ventral OLM cells appears to be lower than the input resistance of dorsal OLM cells. This is quite puzzling, since the downregulation of Ih has been reported to generally decrease the resting membrane potential of a cell, but (different

30 to our findings) Ih-downregulation usually increases the cell’s input resistance (Maccaferri and McBain, 1996; Lupica et al., 2001; Cobb et al., 2003; Poolos et al., 2006; Dyhrfjeld-Johnsen et al., 2009). Interestingly, also the pyramidal cells of the hippocampus have been shown to possess varying Ih-dependent features. Compared to ventral pyramidal cells, dorsal pyramidal cells have been reported to express a smaller hyperpolarization-activated sag and a more hyperpolarized resting membrane potential on one hand, but a lower input resistance on the other hand (Dougherty et al., 2012). However, this discrepancy of having less Ih and at the same time a lower membrane input resistance could be explained by the bigger dendritic length and the bigger dendritic surface area of dorsal pyramidal cells using morphological characterization and post hoc Sholl analysis (Sholl, 1953; Dougherty et al., 2012). In general, a lower input resistance can be attributed to neurons with larger cell body and/or better preserved dendritic arbor. Accordingly, ventral OLM cells could be bigger than dorsal OLM cells. Especially considering the total length of CA1 in the radial axis of transverse slices, the CA1 subfield has been shown to be significantly longer in slices containing ventral hippocampus than in slices containing dorsal hippocampus (Dougherty et al., 2012). Thus, a profound morphological examination of OLM cells should be very helpful to further unveil the differences between dorsal and ventral OLM cells. Additionally, the membrane capacitance could be further examined, since this property is proportional to the area of the cell membrane and thus another indicator for cell size.

While the membrane input resistance has been further reported to decrease with increasing hyperpolarization, the usage of ZD7288 to block Ih has been shown to hyperpolarize hippocampal CA1 pyramidal cells but at the same time to increase input resistance within a potential range between -65 mV and -80 mV (Surges at al., 2004; Dyhrfjeld-Johnsen et al., 2009). Thus, these seemingly opposing characteristics suggest to not only investigate membrane input resistance at resting membrane potential but also at hyperpolarized potentials (e.g. at -73 mV, similar to Dougherty et al., 2012) as well as before and after ZD7288 application. ZD7288 appears to have some side effects. Although commonly considered as a selective blocker for HCN channels, ZD7288 has been shown to not only inhibit HCN channel currents, but also Na+ currents (Wu at al., 2012). Furthermore, the widely used concentration of 20µM ZD7288 has been reported to block about 75% of Ih, suggesting to investigate the efﬁcacy of different concentrations when applying ZD7288 (Magee, 1999). Alternatively CsCl can be used to block Ih (c.f. Ma et al., 2006 or Dagostin et al., 2015). Finally, the measurement of the input resistance itself is of central importance for comparative studies. While other studies have calculated the input resistance as the slope of the linear ﬁt of the voltage-current plot (c.f. Dougherty et al., 2012; Malik et al., 2016), here the input resistance values have only been reported directly after break-in, i.e. at resting membrane potential, and thus urge for further examination.

Another interesting aspect in the present study could be the comparison between somatic and dendritic input resistance as well as the differential expression of HCN channels within the cells. Thus, a neat addition to the present study would be to provide insights into the dendritic distribution of HCN channels for dorsal and ventral OLM cells, e.g. by using immunohistochemistry or recent advances in in situ sequencing methods (Shah et al., 2016). Additionally, databases of morphologically detailed multi-compartment OLM cells exist to study the compartmental dis-

31 tribution of Ih in OLM cell dendrites with computational modeling approaches (Lawrence et al., 2006; Sekulic´ et al., 2014). Thereby, reverse engineering strategies which employ electrophysiological and morphological data, can mathematically predict the distribution of HCN channels (Sekulic´ et al., 2014; Sekulic´ et al., 2015). Based on such an approach, OLM cell models were reported to better ﬁt with experimental data when the dendritic HCN channel conductance density non-uniformly decreased with increasing distance from the soma (Sekulic´ et al., 2015), suggesting that OLM cells could render more somatic than dendritic Ih.

Confronting computational models with experimental data further revealed that there are striking differences in the kinetics of Ih itself. While OLM cells have been shown to exhibit subthreshold resonance at low frequencies, the contribution of Ih at depolarized potentials appears less than originally assumed (Kispersky et al., 2012). Even with extra, simulated Ih supplied by dynamic clamp in order to artiﬁcially depolarize the half-activation of the innate Ih, a recent study found no preference of OLM cell spiking output in theta (Kispersky et al., 2012). Similar to our results, this study suggests that Ih is inactive near threshold potentials (Kispersky et al., 2012). Addition- ally, OLM cells have been shown to depend on the input frequency of presynaptic sources, i.e. strongly phase-lock to theta-modulated inputs, proposing attractor dynamics rather than pacemaker qualities (Wang, 2002; Tort et al., 2007; Neymotin et al., 2011; Kispersky et al., 2012). Attractor dynamics usually refer to the properties of neuronal networks that have one or more stable states and depending on the network’s initial conditions and the dynamics of underlying synapses, the network will end up in one of the states. Thus, taken together these recent ﬁndings, the studies suggest that the activation curves of Ih in computational models should be moved to more hyperpolarized potentials as well as the pacing properties of OLM cells could be reduced (Chapman and Lacaille, 1999; Zemankovics et al., 2010; Kispersky et al., 2012).

The question concerning pacemaker duties of OLM cells during theta is a non-trivial. On one hand, stands the classical theta model, which suggests that the medial septum and the diagonal band of broca drive hippocampal theta activity (Stewart and Fox, 1990; Hangya et al., 2009). Particularly experiments where septal lesions resulted in memory deﬁcits due to the loss of theta activity argue for this classical model (Vinogradova, 1995). On the other hand, there are recent studies showing that theta oscillations can exist autonomously in hippocampal preparations, i.e. without external inputs from medial septum (Goutagny et al., 2009; Huh et al., 2016). It appears that the hippocampus possesses at least two systems underlying theta synchrony, a minimal in- terhippocampal circuit capable to generate theta in absence of external drive and an external generator for the modulation of theta oscillations. Especially the strong phase-locking capabili- ties of OLM cells, which receive inputs from local pyramidal cells and medial septum, as well as their extensive feedback connections, suggest that OLM cells together with pyramidal cells could cooperatively manage internal and external activities in theta. The role of the opposing Ih-dependent features of OLM cells and pyramidal cells remains to be studied though. It could account for different synchronization aspects, frequency regimes or in/output functions. E.g. in the neocortex is a similar relationship between pyramidal cells and Martinotti cells, where layer 5 Martinotti cells, opposite to the layer trend in pyramidal cells, display a smaller Ih-mediated sag than layer 2/3 Martinotti cells and are pivotal for cortical output (Wang et al., 2004).

32 3.3 Chrna2-Martinotti cells Synchronize layer 5 type A Pyramidal Cells via Rebound Excitation

Background Deep layer Martinotti cells hold a close relationship with thick-tufted pyramidal cells, establish- ing a disynaptic inhibitory circuit in layer 5, the main output layer of the neocortex (Markram et al., 2004; Le Bé et al., 2007). Activated by a few pyramidal cells with a brief, high frequency burst, Martinotti cells provide strong inhibition onto a multitude of postsynaptic pyramidal cells and can synchronize their first few spikes (Silberberg and Markram, 2007; Berger et al., 2010). The classical marker of Martinotti cells is somatostatin, although this marker appears unspecific for layer 5 Martinotti cells since it also labels non-Martinotti cells in layer 2/3 and 4 (Wang et al., 2004; Ma et al., 2006). Approaches to isolate layer 5 Martinotti cells from other somatostatin- expressing neurons, mainly rely on the consulting of morphological, electrophysiological or neurochemical properties (Wang et al., 2004; Markram et al., 2004). Several strategies of transgenic lines have been used to separate Martinotti cells, but still depend on the morphological reconstruction to identify a homogeneous population of Martinotti cells (Oliva et al., 2000; Ma et al., 2006; Xu et al., 2013). For example, the SOM-Cre/RFP/X94 mouse line marks layer 1-projecting Martinotti cells with cell bodies in both layer 5 and layer 2/3 (infragranular and supragranular layers) as well as layer 4-non Martinotti cells (Xu et al., 2013). On the other hand, ChAT-Cre (choline acetyltransferase-Cre) mice have been shown to send cholinergic axons from the basal forebrain to cortical layer 1 and layer 2/3 late-spiking interneurons (Arroyo et al., 2012; Arroyo et al., 2014). Considering that cholinergic modulation of low-threshold spiking interneurons in layer 5 have been reported to generate beta oscillations in layer 5 pyramidal cells via dendritic events, kinetically similar to Martinotti cells (Roopun et al., 2008; Roopun et al., 2010), we hypothesized that Martinotti cells could express the cholinergic receptor nicotinic alpha 2 subunit (Chrna2), similar to OLM cells (Leão et al., 2012). Thus, in Study III we used Chrna2-Cre mice (Leão et al., 2012; Perry et al., 2015; Mikulovic et al., 2015) to investigate if Chrna2 is a specific marker to exclusively target Martinotti cells and optogenetically triggered oscillatory inhibition to synchronize pyramidal cell firing over longer periods of time.

Experimental Question (Hypothesis) Do distal dendrite-targeting interneurons of the neocortex express Chrna2 and can their rhythmic activation synchronize the spike timing of cortical pyramidal cells?

Outcome We crossed Chrna2-Cre mice with a tdTomato reporter line and patched Chrna2-Cre/tdTomato cells in coronal slices of primary auditory cortex, as this cortical region was shown to possess the highest connectivity between pyramidal cells and Martinotti cells compared to other cortical areas (Berger et al., 2009). We found that Chrna2 specifically labels Martinotti cells in layer 5. Chrna2+ Martinotti cells have an ovoid cell body in layer 5, bipolar dendrites, the defining characteristic of a long axonal projection to layer 1 and extensive ramifications in layer 4. Whole-cell

33 current- and voltage-clamp recordings further suggest that passive and active electrophysiological properties of Chrna2+ Martinotti cells resemble the classical low-threshold spiking patterns of Martinotti cells. Chrna2+ Martinotti cells usually exhibit spike frequency adaptation and burst discharge when depolarized. On a circuitry level, Chrna2+ Martinotti cells reciprocally connect to type A pyramidal cells projecting to layer 1 but bypass type B pyramidal cells (Fig. 3.3). While the stimulation of type A pyramidal cell generates excitatory postsynaptic potentials in Chrna2+ Martinotti cells, type B pyramidal cells do not cause depolarizations in Chrna2+ Martinotti cells. In turn, activating Chrna2+ Martinotti cells with depolarizing current injections results in inhibitory postsynaptic potentials with long rise times and synaptic depression in postsynaptic type A pyramidal cells but no inhibitory responses in type B pyramidal cells. Moreover, Chrna2+ Martinotti cells contribute to the frequency-dependent disynaptic inhibition (FDDI) circuit. When activating presynaptic type A pyramidal cells with high frequency, inhibitory postsynaptic potentials are apparent in postsynaptic type A pyramidal cells. These inhibitory postsynaptic potentials do not appear when deactivating the Chrna2+ Martinotti cells, e.g. with Halorhodopsin, suggesting that Chrna2+ Martinotti cells account for FDDI.

Connectivity of Martinotti cells

Martinotti cells - Type A Martinotti cells - Type B Pyramidal cells Pyramidal cells

Stimulation of Martinotti cells

Tonic Bursts Synchronization of Type A Pyramidal cells

Figure 3.3: Illustrative ﬁgure of the outcome of Study III. Left: Experimental setup. Right: Chrna2+ Martinotti cells are reciprocally connected to type A but not type B pyramidal cells (top). Optogenetically stimulating Chrna2+ Martinotti cells in bursts, e.g. with Channel- rhodopsin, results in a higher pyramidal cell synchronization than tonic activation (bottom).

Additionally, after strong hyperpolarization, Chrna2+ Martinotti cells usually respond with rebound action potentials. If a population of Chrna2+ Martinotti cells rebounds together, the resulting rebound action potentials are able to generate strong inhibition in type A pyramidal cells. The inhibition then causes type A pyramidal cells to rebound as well and form subsequent contemporaneous rebound action potentials, i.e. Martinotti cell rebound action potentials are able to temporally align the pyramidal cell ﬁring. On the other hand, driving Chrna2+ Martinotti cells with brief bursts, e.g. via optogenetic depolarization with Channelrhodopsin (ChR2), also robustly aligns type A pyramidal cell action potentials. While oscillatory activation in beta (15 Hz) is able to change burst spiking to regular spiking in type A pyramidal cells, beta burst activation of Chrna2+ Martinotti cells (3-4 pulses in 15 Hz, repeated with slow frequency around 2 Hz) can continuously synchronize type A pyramidal cell spiking (Fig. 3.3). Thereby, the burst

34 ﬁring of Chrna2+ Martinotti cells appears more effective to synchronize type A than tonic/continuous spiking because of the frequency-dependent, depressing Martinotti cell-pyramidal cell synapse, i.e. increasing stimulation frequency in Martinotti cells causes bigger amplitudes in pyramidal cell inhibitory postsynaptic potentials which reduce in size with every repetition.

Conclusion and Future Perspectives Study III shows that Chrna2 is a highly specific marker for layer 5 Martinotti cells and predicts that Chrna2+ Martinotti cells are selective when targeting pyramidal cells by specifically connecting to type A pyramidal cells. This specificity is quite surprising considering the highly divergent Martinotti cell connections but gives the layer 5 Martinotti cell the power to modulate the firing of the main output neurons of the neocortex (Markram et al., 2004; Fino and Yuste, 2011). Belonging mainly to the group of somatostatin-expressing interneurons, Martinotti cells have been suggested to provide a dense but rather unspecific blanket of inhibition onto neocortical pyramidal cells and serve synchronous activities (Karnani et al., 2014). Utilizing the analogy of a blanket of inhibition, studies have shown different interneuron populations to create this unspecific inhibition at different time points of activity (Karnani et al., 2014). Thereby, parvalbumin-expressing interneurons provide an early blanket, i.e. the depressing synapses between pyramidal cells and parvalbumin-expressing cells suggest an early and fast action on local pyramidal cells, whereas somatostatin-expressing interneurons provide a late blanket of inhibition (Karnani et al., 2014). Furthermore, recent studies using vasoactive intestinal protein (VIP)- expressing interneurons have shown that the blanket is more specific than originally expected (Karnani et al., 2016a; Karnani et al., 2016b). VIP cells preferentially inhibit somatostatin- expressing cells, most likely Martinotti cells, and thereby metaphorically make holes into the somatostatin-blanket (Karnani et al., 2014; Karnani et al., 2016a). A close relationship between VIP cells and somatostatin-expressing cells can also be seen in the hippocampus, where distinct axonal bands can be found in stratum pyramidale and stratum oriens (Taniguchi, 2014). If VIP cells selectively target Chrna2+ cells, is not entirely clear and remains to be shown. Similarly, the relatively low overlap between Cre+ cells in Chrna2-Cre/R26tom mice and somatostatin in the present study (probably due to the extra-somatic location of the peptide) could be further examined, e.g. by using single-cell RNA sequencing (Zeisel et al., 2015).

Whether the subtype-specific pyramidal cells in Study III comprise pyramidal cells with the same functionality is another interesting question. The preparation used in this study does not permit the investigation of the cellular targets of type A pyramidal cells. Here, the pyramidal cells were distinguished according to electrophysiological and morphological properties rather than their targets. However, the thick-tufted morphology with extensively branching distal dendrites and the large triangular-shaped soma as well as their Ih-dependent features suggest that type A pyramidal cells are subcortically projecting pyramidal cells (Hattox and Nelson, 2007; Brown and Hestrin, 2009; Harris and Shepherd, 2015; Kim et al., 2015). In the auditory cortex, subcortically projecting pyramidal cells include cortico-collicular neurons which mainly target the inferior colliculus and have distinct roles in auditory processing than cortico-cortical neurons, suggesting a specific role for the here identified type A pyramidal cells during sound presentation (Turner et al., 2005; Sun et al., 2013; Joshi et al., 2015). Also from an immuno-

35 histochemical point of view, pyramidal cells have been categorized into subgroups but show overlapping results using the classical pyramidal cell markers Ctip2 and Satb2 due to developmental processes (Harris and Shepherd, 2015; Harb et al., 2016). Notably, a recent study using RNA sequencing has found two types of genetically distinct pyramidal cells in cortex layer 5 (Zeisel et al., 2015), providing evidence of genetically different pyramidal cell populations. Thus, a fascinating addition to the current study would be to investigate if Chrna2+ Martinotti cells speciﬁcally target one of the two populations and determine their targets, e.g. via retrograde tracing. Interestingly, also in the hippocampal CA1 region at least two pyramidal cell populations could be distinguished (Zeisel et al., 2015), urging the investigation of whether Chrna2+ OLM cells could sub-speciﬁcally target pyramidal cells in the hippocampus as well.

Another interesting finding in this study is the ability of Chrna2+ Martinotti cells to cause strong hyperpolarizations in postsynaptic pyramidal cells which are able to trigger synchronized rebound action potentials. Rebound action potentials have been shown to occur in vivo and are usually dependent on T-type calcium channels which are inactive at resting membrane potential and get activated by prolonged inhibitory inputs (Llinas and Yarom, 1981; Bean, 2007; Tsuno et al., 2015). Studies have shown that subthreshold oscillations and the hyperpolarization- activated current Ih play a crucial role for the recruitment of rebound action potentials and as a consequence the temporal tuning of a cell’s firing and its synchronization abilities (Desmaisons et al., 1999; Degenetais et al., 2002; Funahashi et al., 2003; Leão et al., 2006; Zemankovics et al., 2010). For example, in the entorhinal cortex Ih-dependent resonance frequency and rebound spiking together with theta cycle skipping, i.e. synchronous and anti-synchronous firing to the theta cycle, are able to generate grid cell firing (Hasselmo, 2013). The mechanisms of Ih during rebound action potentials of pyramidal cells of both neocortex and hippocampus have been largely studied, however, the generation of rebound action potentials in interneurons is not completely ruled out yet (Ascoli et al., 2010; Lee et al., 2014). While the involved currents underlying the rebound action potentials in Chrna2+ Martinotti cells have not been determined in our study, an initial attempt to decipher these currents by bath-applying ZD7288 showed that Chrna2+ Martinotti cell rebound action potentials are resistant to Ih inhibition. This is analog to X98 cells, which show resemblance to Martinotti cells in another mouse line and where application of CsCl blocked the hyperpolarization-induced sag but did not block the rebound spikes (Ma et al., 2006). Accordingly, the rebound action potentials of Chrna2+ Renshaw cells were insensitive to ZD7288 (Perry et al., 2015). How Ih is shaping the rebound action potentials of Chrna2+ OLM cells remains to be shown, but the increased expression of Ih in dorsal OLM cells suggests that dorsal OLM cells could rebound as shown in Study II. Additionally, it would be interesting to investigate if ventral pyramidal cells in the hippocampus, which possess more Ih-dependent features than dorsal ones, respond with rebound action potentials upon the termination of OLM cell inhibition as it is the case for layer 5 pyramidal cells which rebound upon the termination of Martinotti cell inhibition. Subsequently, the role of rebound action potentials triggered by a defined interneuronal subpopulation (OLM cells or Martinotti cells) could be examined in vivo using the specificity of the Chrna2-Cre mouse line. These and other open questions concerning the generation of synchrony and the emergence of oscillations by the activation of distal dendrite-targeting interneurons remain addressed in further studies.

36 CHAPTER 4 Discussion

What the present studies jointly convey is that distal dendrite-targeting interneurons play an important role for the formation of synchronous activities in hippocampus and neocortex. Although distal dendrite-targeting interneurons generate relatively small IPSPs compared to proximally innervating interneurons, they are potent enough to synchronize pyramidal cells when a population fires together. Interestingly, both OLM cells and Martinotti cells have been shown to be electrically coupled with one another through gap junctions which synchronize their activities and as a result support the synchronous inhibition of pyramidal cells (Gibson et al., 1999; Berger et al., 2010; Leão et al., 2012). Because of their unique anatomical location, the here presented cell types are perfectly suited to act as switches for inputs and firing patterns of pyramidal cells. With focus on the inhibition of the very distal portion, both OLM cells as well as Martinotti cells are able to shape pyramidal cell firing and can even organize their spikes in synchrony, with the latter sometimes also being called community organizer (Lee and Huguenard, 2011). While OLM cells largely exhibit characteristics of an input gating machinery by inhibiting inputs from entorhinal cortex onto distal dendrites and disinhibiting inputs from CA3 onto proximal dendrites, Martinotti cells mostly employ mechanisms for frequency gating, e.g. by changing firing from high-frequency bursting to non-bursting. Both gating features of OLM cells and Martinotti cells are able to generate cell assemblies and further produce oscillations in the theta and beta frequency range respectively (Womelsdorf et al., 2014). Whether and to what extend these features are utilized comparably and if they are interchangeable, however remains to be studied in further experiments. An interesting addition for future experiments with OLM cells could be the question if OLM cells can change the firing of hippocampal pyramidal cells, e.g. from bursting to non-bursting as Martinotti cells can do with cortical pyramidal cells. Likewise, a potential extension for future experiments with Martinotti cells could be the question if Martinotti cells can inhibit supragranular inputs while disinhibiting infragranular inputs, as OLM cells are suggested to do with temporoammonic and Schaffer collateral inputs. Moreover, the inactivation of OLM cells during aversive stimuli has recently been shown to prevent fear learning in mice (Lovett-Barron et al., 2012), highlighting the central role of these dendrite-targeting cells and motivating a multitude of further studies of OLM cells/Martinotti cells during cognitive tasks.

37 Furthermore, it would be highly interesting to compare the functional output of OLM cells and Martinotti cells. If morphological, neurochemical and electrophysiological alikeness automat- ically implies functional equality is a complex question and still not known. This is not even clear within cell types where behavioral states or anatomical variabilities might differentially affect the actions of a cell. For example, our results in Study I suggest that theta synchrony is a property of OLM cells and that the amount of Ih can change theta synchrony (c.f. Study I). How- ever, Study II provides evidence that dorsal and ventral OLM cells differ in the amount of Ih and therefore suggests different roles for dorsal and ventral OLM cells (c.f. Study II). This would mean that theta synchrony could be location-dependent and different between the dorsal and ventral hippocampus. A multitude of experiments already suggests so (for a review see Strange et al., 2014), though the mechanisms and reasons behind such diversity remain unknown. Dif- ferences depending on the cell location or behavioral state have also been reported in other brain regions (Giocomo et al., 2007), where seemingly versatile cells within a population can form a collective and functionally meaningful output. The question arises if such difference within a cell population promotes diversity or harmony. In other words: how much does the individu- ality of each cell matter? While in vitro studies often try to find criteria which are statistically supported in order to fit a single randomly chosen sample to a group of entities, in silico studies often use reduced complexity by focusing on a unit model with only a few parameters, aiming to find universal conclusions for certain questions. In that regard, a very recent and extensive study is the Blue Brain Project, launching the exploration of the neocortex (Markram et al., 2015). This combinatory approach includes reverse-engineering methods to create models from experimental data with the mission to build a computer simulation, i.e. a unit model, of the entire brain (Markram et al., 2015). The main criticism however, is that this approach suffers from the lack of a detailed connectome. Although the project includes a richer set of experimental data and more profound conductance-based models than preceding approaches, information about synaptic plasticity, possible neuronal changes during development and their underlying genetic aspects still appear confined (Markram et al., 2015). On the other hand, such simulations as the Blue Brain Project are highly advantageous when representing the point of view that one understands a complex system by building it step by step, e.g. to identify the role of cells and their synaptic connections under various predefined conditions.

Whether synapses or cells are the main building blocks of cognition and memory is another important question in neuroscience and goes back to the neuron doctrine which, since the famous Golgi stains of Santiago Ramón y Cajal, assumes that neurons are autonomous identities and are intimately interconnected by synapses (for a review see Poo et al., 2016). If information is stored in the potentiation (increasing number) and depression (decreasing number) of connections between neurons and the precise timing of neuronal events as proposed by Donald Hebb, it appears to be one of the multiple keys for understanding the brain, to rule out the synaptic dynamics of each cell type and the various aspects of neuroplasticity during development and behavior. That is one of a multitude of arguments why simulating a brain in a computer can help us to obtain snapshots or even make predictions on certain brain states and functionality but from a today’s perspective will not be able to entirely explain the brain. Of note, and because

38 the brain is often compared with a supercomputer, a highly interesting work on the similarities and dissimilarities between brains and computers including processing speed and paralleliza- tion, was published by the mathematician John von Neumann in 1958 (Von Neumann, 1958). To date, it appears that the brain is much more than a supercomputer and that a well-adjusted combination of multiple techniques is needed to eventually be able to uncover the mysteries of the brain. Thereby, in addition to recent advances in technology, including increased computational power and the ability to manipulate speciﬁc cells, a profound understanding of genetics seems particularly important.

In Study III we focused on the genetic profiling of a specific cell type and used this specificity to study its underlying circuitry. Following the question if interneurons are not non-specifically targeting pyramidal cells, we identified a marker for Martinotti cells in the neocortex. Identifying marker genes for specific cell populations is a widely used strategy in neuroscience nowadays. Especially the development of a multitude of mouse models and Cre-lines appears to be a powerful strategy for cell identification (Madisen et al., 2010; Madisen et al., 2012; Huang, 2014; Taniguchi, 2014). But as every identification strategy, also this strategy should be taken with caution. Genetic similarity does not necessarily mean cell type homogeneity. Some marker genes are better candidates than others, e.g. the classical markers parvalbumin and somatostatin lack the possibility of addressing homogeneous cell types, and thus their Cre driver lines appear to label cell populations with overlapping features (Taniguchi et al., 2011; Hu et al., 2013; Lovett- Barron et al., 2014; Katona et al., 2016). For example, these Cre lines have been shown to label cell populations with different developmental origin, functionality or morphology (Taniguchi et al., 2011; Hu et al., 2013; Mikulovic et al., 2015). In order to target subtype-specific interneuronal populations, we focused on our Chrna2-Cre mouse line (Leão et al., 2012) in Study II and Study III and examined if Martinotti cells similar to OLM cells could be genetically identified by Chrna2.

Including the expression of Chrna2, OLM cells and Martinotti cells appear to share quite striking similarities. In addition to their anatomical similarities (especially the long axonal projections to the upper layers and the targeting of pyramidal cell distal dendrites), Chrna2+ OLM cells and Chrna2+ Martinotti cells possess a high neurochemical alikeness. Immunohistochemical data suggests that both, Chrna2+ OLM cells and Chrna2+ Martinotti cells can express somatostatin, but neither contain parvalbumin nor calretinin. If these results hold, Chrna2 could be considered as more specific than most other markers available for targeting OLM cells and Martinotti cells so far (Leão et al., 2012; Mikulovic et al., 2015). Hereby of speculative note, but considering their immunohistochemical profile, Chrna2+ OLM cells and Chrna2+ Martinotti cells could be derived from the caudal ganglionic eminence (Cauli et al., 2014; Liguz-Lecznar et al., 2016). Recently, functionally distinct OLM cells have been shown to originate from different ganglionic eminence, identified by the presence/absence of 5-HT3A receptors (Chittajallu et al., 2013). Thus, additional investigations including 5-HT3A receptors would be interesting to further identify the origin of Chrna2+ OLM cells and Chrna2+ Martinotti cells. Moreover, none of the here presented studies has addressed the actual role of the Chrna2 gene yet. Our Chrna2-Cre mice were created by a random insertion of a transgene, i.e. the Cre gene is expressed under

39 the control of a bacterial artificial chromosome (BAC) (Gong et al., 2007). On one hand, this technique seems advantageous because the endogenous gene is not modified itself as it is the case for knock-in transgenes, on the other hand the additional, artificial sequences might cause overexpression of certain mRNA and possible functional alterations are not completely clear yet and remain to be studied (Madisen et al., 2010; Madisen et al., 2012). A profound study of the Chrna2 gene therefore could be beneficial to learn more about the genetic identity of OLM cells and Martinotti cells.

Finally, considering the electrophysiological alikeness of Chrna2+ OLM cells and Chrna2+ Martinotti cells, both cell types can be described as low-threshold, slow spiking and accommodating interneurons. These features go together with the facilitating dynamics of the afferent excitatory synapses and allows single presynaptic pyramidal cells to trigger action potentials via high-frequency bursts (Silberberg and Markram, 2007; Berger et al., 2010). Burst firing in Chrna2+ Martinotti cells does not necessarily require the synchronous firing of many presynaptic cells, but rather three or more action potentials generated in quick succession. If repeated in slow frequency and separated by pauses between the bursts, Chrna2+ Martinotti cells are able to synchronize postsynaptic pyramidal cells (c.f. Study III). Burst synchronization is a common feature in the brain and has been described in various brain regions (Maeda et al., 1995; Sherman, 2001), most recently in the dorsal cochlear nucleus (Wu et al., 2016). The dorsal cochlear nucleus takes a specific role during auditory perception and is thought to contribute to the emergence of tinnitus (Kaltenbach et al., 2000; Brozoski et al., 2002; Shore et al., 2008). Evidence exists that tinnitus especially correlates with increased synchrony and bursting (Wu et al., 2016). However, not only under pathological but also physiological conditions, burst firing and synchronized activities occur together, both in vivo as well as in vitro (Sanchez-Vives and McCormick, 2000; Okun and Lampl, 2008). For example, during slow-wave sleep, Up and Down states in the neocortex have been shown to occur synchronized (Volgushev et al., 2006). Up and Down states describe subthreshold membrane potentials of neurons which can be more depolarized (Up state) or more hyperpolarized (Down state) and are often connected to the bursting subtype of pyramidal cells (Sanchez-Vives and McCormick, 2000; Timofeev et al., 2000). Taken into account that disinhibitory circuit motifs (such as the one Martinotti cells are part of), are ubiquitous in the brain and can control integrative tasks in various cortical areas (Lee et al., 2013; Pfeffer et al., 2013; Pi et al., 2013), Chrna2+ Martinotti cells could therefore play an essential role during slow oscillations in the neocortex and modulate synchronous activities under in vivo conditions. Analogously, Chrna2+ OLM cells could promote synchronous activities during slow oscillations in the hippocampus.

4.1 Conclusion

Using a combination of multiscale computational modeling (c.f. Study I), patch-clamp electrophysiology (c.f. Study II and Study III) and the latest optogenetic advances (c.f. Study III), the aim of this PhD thesis was to investigate how distal dendrite-targeting interneurons can take central roles for the synchronization of pyramidal cells and thus the rhythmogenesis of certain brain structures. Achieved through various ways, such as direct coupling or common inputs, synchro-

40 nization by OLM cells and Martinotti cells stands out due to the inhibition of distal dendrites, the recruitment of hyperpolarization-activated cyclic nucleotide-gated ion channels and the induction of rhythmicity in a slow frequency range. The ﬁndings of this thesis also encourage the observation that synchronization mechanisms may introduce periodic oscillations and stability to a neuronal system. All in all, OLM cells in the hippocampus as well as Martinotti cells in the neocortex cooperatively work together with their surrounding cells and achieve performances that neither a supercomputer nor us have completely understood yet. This and other aspects are making the brain unique and provide us researchers with the daily motivation on our exploratory journeys in order to successively challenge and discover the brain.

Abbreviations

AC1 primary auditory cortex

ACSF artiﬁcial cerebrospinal ﬂuid

AP action potential

Arch archaerhodopsin

BAC bacterial artiﬁcial chromosome

CA cornu ammonis cAMP cyclic adenosine monophosphate cCMP cyclic cytosine monophosphate cGMP cyclic guanosine monophosphate

ChAT choline acetyltransferase

ChR2 Channelrhodopsin

Chrna2 cholinergic receptor nicotinic alpha 2

CLARITY Clear lipid-exchanged acrylamide-hybridized rigid imaging tissue hydrogel

Cre cyclization recombinase

Ctip2 COUP-TF-interacting protein 2 cUMP cyclic uridine monophosphate

DG dentate gyrus

DIO double-ﬂoxed inverted open reading frame

EC entorhinal cortex

FDDI frequency-dependent disynaptic inhibition

43 GABA gamma-aminobutyric acid

HaloR Halorhodopsin

HCN hyperpolarization-activated cyclic nucleotide-gated

Hip. hippocampus

Ih hyperpolarisation-activated current

IM non-inactivating K+ current IT intra-telencephalic

MI mutual information

OLM oriens-lacunosum moleculare

PC pyramidal cell

PING pyramidal-interneuronal network gamma

PT pyramidal-tract

R circuit resistor circuit

RC circuit resistor-capacitor circuit

RLC circuit resistor-inductor-capacitor circuit

Satb2 special AT-rich sequence binding protein 2

SEM standard error of the mean

Sub subiculum

VC2 secondary visual cortex

VIP vasoactive intestinal protein

44 Acknowledgements

First and foremost I would like to thank my supervisor Kia Leão for the continuous support, the daily challenges and the steady feedback, making my PhD an amazing journey of unforgettable moments. In one of our first meetings you were drawing a stick figure of me holding a Swedish and a Brazilian flag in my hands, accompanied by a big smile and writing “Dr. Hilscher” on top. This is where I am now: having experienced all facets of neuroscience in these two countries - from creating ideas to the resourceful planning of experiments and being confident in publishing papers, ending up with a big smile on my face. Thank you very much for that. A big contribution to my success also comes from Richardson Leão. Thank you for providing me with all kinds of techniques, tricks and experimental knowledge to survive the wilderness of neuroscience. I feel like Bear Grylls on his survival trips in the jungle - prepared for every eventuality. I am also stronlgy encouraged to thank Klas Kullander. Thank you for hosting me during my time in Sweden and fascinating me with your enthusiasm for genetics and the Martinotti cell story. Thank you all, for providing me the perfect combination of knowledge, creativity and innovation during my joint PhD between Brazil and Sweden.

That I felt home in both countries was furthermore the contribution of and fun with wonderful friends and colleagues, who I want to mention here. From the Brain Institute, I would like to say thanks to Adriano Tort. Thank you for impressing me with your analytical competence, your preciseness when discussing methods and your football skills. I would also like to thank Diego Laplagne. Thank you for all your questions on my electrophysiology experiments and your participation in my scientific committee. Further, Kerstin Schmidt thank you for participating in my scientific committee and the inputs to my presentations as well as Sergio Neuenschwander, Martin Cammarota, Marcos Costa and Sidarta Ribeiro, thank you for your great support and very helpful ideas. In this regard, I would also like to thank Gilad Silberberg and Christopher Kushmerick for paricipating in my examination, the 3 hours of questioning and your inspiring inputs. I of course cannot forget my friends in Ribeirao Preto, where I learned more about auditory circuits, songbirds, Sertanejo and the good Cervejaria Colorado. Ricardo Leão, thank you for hosting me in your lab and introducing me to such cool people as André, Ana, João Baiano, Alexandra, Paula and Teddy. Besides Natal, Ribeirao Preto became a neuroscience capital for me, where I also got to know Antonio Roque, Sam Neymotin and William Lytton. LASCON (Latin American School on Computational Neuroscience), which I first attended just before starting my PhD and two years later for a series of invited lectures, was truly a great experience. It provided me with broad knowledge in neurocomputation and amazing friends all over the world, including Giseli, Jean, Gary, Thiaginho and Olavo, to name the closest.

45 Furthermore, I would like to thank the best “Pastelanche gang” in the world. Lampião Helton, thank you for our great times in João Pessoa and Uppsala together with Maria Bonita Rossana. Aron, the Malandro of Paraiba, Hindiael, my personal gym professor from Minas and Ernesto, the greatest Portuguese explorer together with his companion João, thank you for the amazing times inside and outside the lab including some cultural events and for your participation in my scientiﬁc committee. Also a big thank you to my lab members, for all that fun we had together. Thawann and his boss Barbara, thank you for all the jokes during work and lunch and the Por- tuguese translations (which btw do not necessarily sound more melodic than the Austro-German versions...in my ears). Also obrigado to, “it’s dorsal” Margareth, “Blacky” Annara, the rock musician Franzon and our lab book designer Rafael Lima for the great discussions and events together. I would also like to thank Robson, the Chuck Norris of the Brain Institute for inspiring me with all your publications, Zezito for your spontaneity and surprises and Vitor for your strict and sceptic appearance when discussing science with me.

Furthermore, I would like to thank the most laughing persons around me. Pavão and Malek, thank you for all our great trips including the warm “cachaça experience” in Passa e Fica. Lorenzo, thank you for all our jokes and trips, especially the stuttering one at this famous fast food restaurant in Natal. Renzo, another Malandro, thanks for our great times at RU haha, as well as Conde, Annie, Bryan, Rafael Pedrosa, Nathalia and Fernanda thanks for our beau- tiful trips to various places outside of Natal. A big thanks also goes to the ICe Girls. Geissy, thanks for all the pedreiragem and the trips to Johnny Camara, Yay-Juliana thanks for our coisa fofas and all the “elegance and comfort”, Kung fu-Dani thanks for the endless help with mobility in Natal and Roberta thanks for the seemingly endless speeches to me. Moreover, I would like to thank Bruna, Ana Raquel, Jessica, Juliana Martins, Kelly, Carol, Olga, Renata, Janine, Marina, Andressa and the Argentinian Carolina for making my time in Natal so so nice. The same accounts for Dardo, Anderson, “Dudu” Eduardo, Eduardo Diehl-Fleig, Daniel, André, Hjalmar, Wilfredo and Vincent and Ana Maria who I actually had my ﬁrst buggy ride with. Special thanks also go to Rai and Bela, who I spent a lot of time with outside the lab to switch off from work and the fastest jardineiro in the world Jailson, thank you for your kindness and language courses all the time.

Next, I would like to thank my friends and colleagues in Sweden, who made me feel welcome all the time and always had a place for me when I needed. Especially people from the actual and former Kullander Lab, Lagerström Lab, Boije Lab and Mackenzie Lab, including Malin, Hen- rik, Åsa, Lina, Christiane, Fatima, Anders Enjin, Jörgen, Fabio C., Atieh, Amilcar, Dave and Sharn I would like to thank for sharing such great experiences with me. I am happy that our paths crossed in Uppsala. Additionally, I would like to thank the “dhinka chika gang” Chetan, Kali, the “real” Emil and “Mart” Martin - especially for our fun times, our common dedication to the Stars and our great trips, movies and beach volleyball games. I would also like to thank Sören and Ana del Pozo Cano, who introduced me to zebraﬁsh electrophysiology. A big thank you as well to Bejan, who taught me the importance of mouse genetics and that she made me admire her Cre-ationism (I am talking about Cre lines), as well as thanks to Elin M., Marina F.,

46 Jon and William for our good times. I would also like to thank Kasia and Adriano, providing me the first roof in Uppsala, for teaching me “Non sparare! Sono un ricercatore austriaco.” and all the fun we had together. Futhermore, I would like to thank Emma for her positive crazi- ness and playing the weekly “meeting bell” before our group meetings for us to be in time as well as Casey for her extensive lectures on PCRs, Anna A. for our awesome parties and Carly for your enthusiasm in singing everything. Also big thanks to Nadine, the best tango teacher and cultural-event-spotter in the world, the Japanese cat-lovers Thomas and Nina, “Louca” (in a positive way) Bianca, the fastest-running neuroscientist Åsa K., my favourite duck-supplier Hanna and Julia for exchanging opto-knowledge in- and outside the lab. Tack så mycket! Also thanks to the present and former Neurodynamics branch in Sweden: Sanja, Pavel, Ida, Ali and Martina, especially for always providing me with sweets from Austria and the Sacher cake for my birthday as well as George, the master of electronics and cockroaches in Uppsala. I would further like to thank Masha, Sveta, Carl, Anongnad and Niclas, Lill-Anders and Linda, Sofie and Emelie as well as Belen and Pamela from the Department of Neuroscience, who made my time there as amazing as possible with various adventures and swedish fikas.

This leads me to thank my favorite secretaries of the Department of Neuroscience in Uppsala as well as the Brain Institute in Natal for all the help and the great mind-refreshing talks around lunch time. Without them I would have been so much more into bureaucracy. Special thanks go to my two “Mother Theresas” Mariana and Akaline. Furthermore, I would like to thank Siv, Berit, both Cecilias, Emma, Karin Nygren and Neil for helping me out of every possible situation, especially when I was lost in admistrative work. Also, I would like to thank Karin Nordström, especially for the Humbug at the very beginning of my experiments, Erika for the great talks and discussions in the corridors and recently at the dopamine conference in Vienna, Michael for his enthusiasm in ﬂies, Christian for his amazing advices and inspiring presence in media, as well as Jenny and Eva from the animal facility, as representers for all the other great co-workers, who made my experiments run so smoothly and awesome.

This awesomeness I could feel on a daily basis because of such great “Amigos” and “Komp- isar”: Fabiano, or should I say Geraldo or even Janaina haha, thanks for sharing the apartment, rig and many other things with me. Samer, thanks for an amazing Norway trip, for adapting so quickly to my Portuguese-Swedish-English language-mix and for sharing CLARITY material (especially for Figure 1.1). Steven, thanks for showing me how to clear brains, spinal cords and minds and for sharing that knowledge with me. Arthur, thanks for introducing me to the best laughing ever and your “AUUUU” as well as inspiring me with your great work in the hippocampus. Ninnie, thanks for an amazing time, also for going to Gothenburg and for introducing me to the most famous facade in Uppsala. Nina, thanks for discovering parts of the world with me and for teaching me all the very details about dentistry. Hermany, who I met during my early days in Uppsala, thanks for always being so close, for sharing the passion for Martinotti cells with me, for discovering so many new countries together and for making me experience Christmas and New Years from another perspective - the Brazilian perspective with your family (just to name a few of our seemingly endless adventures). Carro, thanks for the cutest Portuguese mes- sages - fofura all over. Malte, my special German friend, thanks for all the silly jokes (mostly

47 when comparing Germany and Austria), all the Strunkelbärchenmarmelade and the motivation for me to do more sports - especially when thinking about our half marathon together. The 42k race is still on our to-do list. Annika, thanks for introducing me to wakeboarding (and not to Pokemon Go :D) and all our good times. Pieter, thanks for inviting me to all your parties and Oktoberfests and climing the highest mountain in Austria with me. Elin, thanks for our great times, especially shaping our New Years in Berlin. Jorge Ivan, thanks to the great Colombian always making me laugh, being so chilled and being the one introducing me to Jenny García, the most powerful and smiling lady I have met. Simon-Pierre, my favourite gigantisch Cana- dian beer-brewer and ice hockey player and Anna, thank you for our amazing events and I will never forget how you two ﬁlled my car with beers after lending it to you. This tradition could be continued haha. German and Oscar, thanks for being so so funny to hang out with and thanks for your inspiring words all the time. Thanks also to the great organizers and friends from the “neuroscience pub” and all the “after work” clubs, Chus Xesus, Maestro Vanni, Igor, Leonor, Heidur and and and...as well as the bands Ghost and Fat Freddy’s Drop, who I wrote this thesis with and who I got to see twice during my PhD, even with my developing thesis in the backpack.

Back in Austria I would like to thank my friends who have been so patient with me all the time, especially when postponing catch-ups because of work in Brazil or Sweden. Thanks for being so loyal to me Emanuel, Daniel, Bernhard O., Lukas, Szymi, Morten Martin as well as Paul and Simon, my colleagues and friends at the Computational Neuroscience and Biomedical Engineering research group from the Vienna University of Technology. A huge thank you also goes to Frank Rattay, my mentor and professor, who followed me since my early steps in science, who understood and supported my talents in every direction possible and who made this “fun of a PhD” possible by sending me to Sweden.

Finally, I would like to thank my family. My brother Bernhard, my mum Silvia, my dad Wil- fried and my grandmothers and grandfathers, thank you for always supporting me during my life and espcially during this PhD. Thank you for all the calls, mails, emails, chats and visits. Thank you for making me feel so special when coming home and also when being away. Although my visits usually were quite short you were so understanding, supportive and fun. Thank you for keeping me free of bureaucracy. Thank you for forwarding me all the wishes and greetings of my family, neighbors and friends. Thank you for all the kind words. Thank you for all the good food, the spontaneous trips, the nice surprises and many many things more. You were so far away in distance but so close in support and love. Thank you very much for all and everything! Danke! Danke! Danke!

Figure by Kia Leão, 2013

48 Bibliography

Abbott, L.F. 1990. A network of oscillators. J. Phys. A: Math. Gen. 23:3835-3859.

Abbott, L.F. 1999. Lapicque’s introduction of the integrate-and-ﬁre model neuron (1907). Brain Res. Bull. 50:303-304.

Abeles, M. 1982. Local Cortical Circuits: An Electrophysiological study. Springer, Berlin.

Abeles, M. 1991. Corticonics: Neural Circuits of the Cerebral Cortex. Cambridge Univeristy Press, New York.

Abeles, M., Hayon, G. and Lehmann, D. 2004. Modeling compositionality by dynamic binding of synﬁre chains. J. Comput. Neurosci. 17:179-201.

Accili E.A., Proenza, C., Baruscotti, M. and DiFrancesco D. 2002. From funny current to HCN channels: 20 years of excitation. News Physiol. Sci. 17:32-37.

Acker, C. D., Kopell, N. and White, J. A. 2003. Synchronization of strongly coupled excitatory neurons: relating network behavior to biophysics. J. Comput. Neurosci. 15:71-90.

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

Aponte, Y., Lien, C.C., Reisinger, E. and Jonas, P. 2006. Hyperpolarization-activated cation channels in fast-spiking interneurons of rat hippocampus. J. Physiol. 574:229-243.

Arroyo, S., Bennett, C., Aziz, D., Brown, S.P. and Hestrin, S. 2012. Prolonged disynaptic inhibition in the cortex mediated by slow, non-α7 nicotinic excitation of a speciﬁc subset of cortical interneurons. J. Neurosci. 32:3859-3864.

Arroyo, S., Bennett, C. and Hestrin, S. 2014. Nicotinic modulation of cortical circuits. Front. Neural Circuits. 8:30.

Ascoli, G.A., Alonso-Nanclares, L., Anderson, S.A., Barrionuevo, G., Benavides-Piccione, R., Burkhalter, A., Buzsáki, G., Cauli, B., Defelipe, J., Fairén, A., Feldmeyer, D., Fishell, G., Fregnac, Y., Freund, T.F., Gardner, D., Gardner, E.P., Goldberg, J.H., Helmstaedter, M., Hestrin, S., Karube, F., Kisvárday, Z.F., Lambolez, B., Lewis, D.A., Marin, O., Markram, H.,

49 Muñoz, A., Packer, A., Petersen, C.C., Rockland, K.S., Rossier, J., Rudy, B., Somogyi, P., Staiger, J.F., Tamas, G., Thomson, A.M., Toledo-Rodriguez, M., Wang, Y., West, D.C. and Yuste R. 2008. Petilla terminology: nomenclature of features of GABAergic interneurons of the cerebral cortex. Nat. Rev. Neurosci. 9:557-568.

Ascoli, G.A., Gasparini, S., Medinilla, V. and Migliore, M. 2010. Local control of postin- hibitory rebound spiking in CA1 pyramidal neuron dendrites. J. Neurosci. 30:6434-6442.

Barth, A., Burkhalter, A., Callaway, E.M., Connors, B.W., Cauli, B., DeFelipe, J., Feldmeyer, D., Freund, T., Kawaguchi, Y., Kisvarday, Z., Kubota, Y., McBain, C., Oberlaender, M., Rossier, J., Rudy, B., Staiger, J.F., Somogyi, P., Tamas, G. and Yuste, R. 2016. Comment on “Principles of connectivity among morphologically deﬁned cell types in adult neocortex”. Science. 353:1108.

Bartos, M., Vida, I. and Jonas, P. 2007. Synaptic mechanisms of synchronized gamma oscillations in inhibitory interneuron networks. Nat. Rev. Neurosci. 8:45-56.

Baumann, V.J., Lehnert, S., Leibold, C. and Koch, U. 2013. Tonotopic organization of the hyperpolarization-activated current (Ih) in the mammalian medial superior olive. Front. Neural Circuits. 7:117.

Bean, B.P. 2007. The action potential in mammalian central neurons. Nat. Rev. Neurosci. 8:451-465.

Bender, R.A., Brewster, A., Santoro, B., Ludwig, A., Hofmann, F., Biel, M. and Baram, T.Z. 2001. Differential and age-dependent expression of hyperpolarization-activated, cyclic nucleotide-gated cation channel isoforms 1-4 suggests evolving roles in the developing rat hippocampus. Neuroscience. 106:689-698.

Berger, H. 1929. Über das Elektroenkephalogramm des Menschen. Arch. Psychiat. Nervenkr. 87:527-570.

Berger, T., Larkum, M.E. and Lüscher, 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.

Berger, T.K., Perin, R., Silberberg, G. and Markram, H. 2009. Frequency-dependent disynaptic inhibition in the pyramidal network: a ubiquitous pathway in the developing rat neocortex. J. Physiol. 587:5411-5425.

Berger, T.K., Silberberg, G., Perin, R. and Markram, H. 2010. Brief bursts self-inhibit and correlate the pyramidal network. PLoS Biol. 8:e1000473.

Bloss, E.B., Cembrowski, M.S., Karsh, B., Colonell, J., Fetter, R.D. and Spruston, N. 2016. Structured Dendritic Inhibition Supports Branch-Selective Integration in CA1 Pyramidal Cells. Neuron. 89:1016-1030.

50 Börgers, C. and Kopell, N. 2003. Synchronization in networks of excitatory and inhibitory Neurons with sparse, random connectivity. Neural Comput. 15:509-538.

Börgers, C., Epstein, S. and Kopell, N. 2005. Background gamma rhythmicity and attention in cortical local circuits: a computational study. Proc. Natl. Acad. Sci. USA. 102:7002-7007.

Börgers, C., Epstein, S. and Kopell, N. 2008. Gamma oscillations mediate stimulus competition and attentional selection in a cortical network model. Proc. Natl. Acad. Sci. USA. 105:18023- 18028.

Bragin, A., Jandó, G., Nádasdy, Z., Hetke, J., Wise, K. and Buzsáki, G. 1995. Gamma (40-100 Hz) oscillation in the hippocampus of the behaving rat. J. Neurosci. 15:47-60.

Brandt, M.C., Endres-Becker, J., Zagidullin, N., Motloch, L.J., Er, F., Rottlaender, D., Michels, G., Herzig, S., Hoppe, U.C. 2009. Effects of KCNE2 on HCN isoforms: distinct modulation of membrane expression and single channel properties. Am. J. Physiol. Heart Circ. Physiol. 297:H355-H363.

Brodmann, K. 1909. Vergleichende Lokalisationslehre der Großhirnrinde. Leipzig.

Brown, S.P. and Hestrin, S. 2009. Intracortical circuits of pyramidal neurons reﬂect their long-range axonal targets. Nature. 457:1133-1136.

Brozoski, T.J., Bauer, C.A. and Caspary, D.M. 2002. Elevated fusiform cell activity in the dorsal cochlear nucleus of chinchillas with psychophysical evidence of tinnitus. J. Neurosci. 22:2383-2390.

Brunel, N. 2000. Dynamics of sparsely connected networks of excitatory and inhibitory spiking neurons. J. Comput. Neurosci. 8:183-208.

Brunel, N. and Van Rossum, M.C.W. 2007. Quantitative investigations of electrical nerve excitation treated as polarization. Biol. Cybern. 97:341-349.

Brunel, N. and Hakim, V. 2008. Sparsely synchronized neuronal oscillations. Chaos. 18:015113.

Buzsáki, G., Bragin, A., Chrobak, J.J., Nádasdy, Z., Sik, A., Hsu, M. and Ylinen A 1994. Os- cillatory and intermittent synchrony in the hippocampus: relevance to memory trace formation. Temporal Coding in the Brain. 145-172.

Buzsáki, G., 2002. Theta oscillations in the hippocampus. Neuron. 33:325-340.

Buzsáki, G. and Draguhn, A. 2004. Neuronal oscillations in cortical networks. Science. 25:1926-1929.

Lytton, W. W. and Sejnowski, T.J. 1996. Inhibition synchronizes sparsely connected cortical neurons within and between columns in realistic network models. J. Comput. Neurosci. 3:91- 110.

51 Cabral, J., Luckhoo, H., Woolrich, M., Joensson, M., Mohseni, H., Baker, A., Kringelbach, M.L. and Deco, G. 2014. Exploring mechanisms of spontaneous functional connectivity in MEG: how delayed network interactions lead to structured amplitude envelopes of band-pass ﬁltered oscillations. Neuroimage. 90:423-435.

Cajal, R. 1893. Histologie du Systéme Nerveux de l’Homme et des Vertébrés. Consejo Superior Investigaciones Cientíﬁcas.

Cajal, R. 1899. Histologie du Systéme Nerveux de l’Homme et des Vertébrés. Consejo Superior Investigaciones Cientíﬁcas.

Cajal, R. 1911. Histologie du Systéme Nerveux de l’Homme et des Vertébrés. Consejo Superior Investigaciones Cientíﬁcas.

Cardin, J.A., Carlén, M., Meletis, K., Knoblich, U., Zhang, F., Deisseroth, K., Tsai, L.H. and Moore, C.I. 2010. Targeted optogenetic stimulation and recording of neurons in vivo using cell-type-speciﬁc expression of Channelrhodopsin-2. Nat. Protoc. 5:247-254.

Carnevale, N. and Hines, M. 2006. The NEURON Book. New York: Cambridge Neuroinform. 3:1.

Cauli, B., Zhou, X., Tricoire, L., Toussay, X. and Staiger, J.F. 2014. Revisiting enigmatic cortical calretinin-expressing interneurons. Front. Neuroanat. 8:52.

Chapman, C.A. and Lacaille, J.C. 1999. Cholinergic induction of theta-frequency oscillations in hippocampal inhibitory interneurons and pacing of pyramidal cell ﬁring. J. Neurosci. 19:8637-8645.

Chen, S., Wang, J. and Siegelbaum, S.A. 2001. Properties of hyperpolarization-activated pacemaker current deﬁned by coassembly of HCN1 and HCN2 subunits and basal modulation by cyclic nucleotide. J. Gen. Physiol. 117:491-504.

Chittajallu, R., Craig, M.T., McFarland, A., Yuan, X., Gerfen, S., Tricoire, L., Erkkila, B., Barron, S.C., Lopez, C.M., Liang, B.J., Jeffries, B.W., Pelkey, K.A. and McBain, C.J. 2013. Dual origins of functionally distinct O-LM interneurons revealed by differential 5-HT(3A)R expression. Nat. Neurosci. 16:1598-1607.

Cobb, S. R., Buhl, E. H., Halasy, K., Paulsen, O. and Somogyi, P. 1995. Synchronization of neuronal activity in hippocampus by individual GABAergic interneurons. Nature. 378:75-78.

Cobb, S.R., Larkman, P.M., Bulters, D.O., Oliver, L., Gill, C.H. and Davies, C.H. 2003. Activation of Ih is necessary for patterning of mGluR and mAChR induced network activity in the hippocampal CA3 region. Neuropharmacology. 44:293-303.

Connors, B.W. and Gutnick, M.J. 1990. Intrinsic ﬁring patterns of diverse neocortical neurons. Trends Neurosci. 13:99-104.

52 Cumin, D. and Unsworth, C.P. 2007. Generalising the Kuramoto model for the study of neuronal synchronisation in the brain. Physica D: Nonlinear Phenomena. 226:181-196.

Dagostin, A.A., Lovell, P.V., Hilscher, M.M., Mello, C.V., Leão, R.M. 2015. Control of Phasic Firing by a Background Leak Current in Avian Forebrain Auditory Neurons. Front. Cell. Neurosci. 9:471.

DeFelipe, J. and Jones, E.G. 1988. Cajal on the Cerebral Cortex. Oxford University Press.

Deco, G., Buehlmann, A., Masquelier, T. and Hugues, E. 2011. The role of rhythmic neural synchronization in rest and task conditions. Front. Hum. Neurosci. 5:4.

Dégenétais, E., Thierry, A.M., Glowinski, J. and Gioanni, Y. 2002. Electrophysiological properties of pyramidal neurons in the rat prefrontal cortex: an in vivo intracellular recording study. Cereb. Cortex. 12:1-16.

Desmaisons, D., Vincent, J.D. and Lledo, P.M. 1999. Control of action potential timing by intrinsic subthreshold oscillations in olfactory bulb output neurons. J. Neurosci. 19:10727- 10737.

Diesmann, M., Gewaltig, M.O. and Aertsen, A. 1999. Stable propagation of synchronous spiking in cortical neural networks. Nature. 402:529-533.

Dong, Y., Mihalas, S., Qiu, F., von der Heydt, R. and Niebur, E. 2008. Synchrony and the binding problem in macaque visual cortex. J. Vis. 8:30 1-16.

Dougherty, K.A., Islam, T. and Johnston, D. 2012. Intrinsic excitability of CA1 pyramidal neurones from the rat dorsal and ventral hippocampus. J. Physiol. 590:5707-5722.

Dougherty, K.A., Nicholson, D.A., Diaz, L., Buss, E.W., Neuman, K.M., Chetkovich, D.M. and Johnston, D. 2013. Differential expression of HCN subunits alters voltage-dependent gating of h-channels in CA1 pyramidal neurons from dorsal and ventral hippocampus. J. Neurophysiol. 109:1940-1953.

Dugladze, T., Vida, I., Tort, A.B., Gross, A., Otahal, J., Heinemann, U., Kopell, N.J. and Gloveli, T. 2007. Impaired hippocampal rhythmogenesis in a mouse model of mesial temporal lobe epilepsy. Proc. Natl. Acad. Sci. USA. 104:17530-17535.

Dyhrfjeld-Johnsen, J., Morgan, R.J. and Soltesz, I. 2009. Double Trouble? Potential for Hyperexcitability Following Both Channelopathic up- and Downregulation of I(h) in Epilepsy. Front. Neurosci. 3:25-33.

Economo, C. and Koskinas, G.N. 1925. Die Cytoarchitektonik der Hirnrinde des erwachsenen Menschen. Atlas mit 112 mikrophotographischen Tafeln in besonderer Mappe. Julius Springer.

Engel, A.K., Kreiter, A.K., Kónig, P. and Singer, W. 1991. Synchronization of oscillatory neuronal responses between striate and extrastriate visual cortical areas of the cat. Proc. Natl. Acad. Sci. USA. 88:6048-6052.

53 Ermentrout, G.B. 1985. Synchronization in a pool of mutually coupled oscillators with random frequencies. J. Math. Biology. 22:1-9.

Ermentrout, G.B. and Kopell, N. 1998. Fine structure of neural spiking and synchronization in the presence of conduction delays. Proc. Natl. Acad. Sci. USA. 95:1259-1264.

Fan, Y., Fricker, D., Brager, D.H., Chen, X., Lu, H.C., Chitwood, R.A. and Johnston, D. 2005. Activity-dependent decrease of excitability in rat hippocampal neurons through increases in I(h). Nat. Neurosci. 8:1542-1551.

Ferenczi, E. and Deisseroth, K. 2012. When the electricity (and the lights) go out: transient changes in excitability. Nat. Neurosci. 15:1058-1060.

Fino, E. and Yuste, R. 2011. Dense inhibitory connectivity in neocortex. Neuron. 69:1188- 1203.

Fisahn, A., Pike, F.G., Buhl, E.H. and Paulsen, O. 1998. Cholinergic induction of network oscillations at 40 Hz in the hippocampus in vitro. Nature. 394:186-189.

Fishell, G. and Rudy, B. 2011. Mechanisms of inhibition within the telencephalon: “where the wild things are”. Annu. Rev. Neurosci. 34:535-567.

Fogle, K.J., Lyashchenko, A.K., Turbendian, H.K. and Tibbs, G.R. 2007. HCN pacemaker channel activation is controlled by acidic lipids downstream of diacylglycerol kinase and phos- pholipase A2. J. Neurosci. 27:2802-2814.

Frégnac, Y., Bringuier, V. and Baranyi, A. 1994. Oscillatory neuronal activity in the visual cortex: a critical re-evaluation. Temporal Coding in the Brain. 81-102.

Freund, T. F. and Buzsáki, G. 1996. Interneurons of the hippocampus. Hippocampus. 6:347- 470.

Freund, T. F. and Katona, I. 2007. Perisomatic inhibition. Neuron. 56:33-42.

Fries, P., Schröder, J.H., Roelfsema, P.R., Singer, W. and Engel, A.K. 2002. Oscillatory neuronal synchronization in primary visual cortex as a correlate of stimulus selection. J. Neurosci. 22:3739-3754.

Fries, P. 2005. A mechanism for cognitive dynamics: neuronal communication through neuronal coherence. Trends Cogn. Sci. 9:474-480.

Fries, P. 2009. Neuronal gamma-band synchronization as a fundamental process in cortical computation. Annu. Rev. Neurosci. 32:209-224.

Fuentealba P., Begum, R., Capogna, M., Jinno, S., Márton, L.F., Csicsvari, J., Thomson, A., Somogyi, P. and Klausberger, T. 2008. Ivy cells: a population of nitric-oxide-producing, slow- spiking GABAergic neurons and their involvement in hippocampal network activity. Neuron. 57:917-929.

54 Funahashi, M., Mitoh, Y., Kohjitani, A. and Matsuo, R. 2003. Role of the hyperpolarization- activated cation current (Ih) in pacemaker activity in area postrema neurons of rat brain slices. J. Physiol. 552:135-148.

Garden, D.L., Dodson, P.D., O’Donnell, C., White, M.D. and Nolan, M.F. 2008. Tuning of synaptic integration in the medial entorhinal cortex to the organization of grid cell ﬁring ﬁelds. Neuron. 60:875-889.

Gibson, J.R., Beierlein, M. and Connors, B.W. 1999. Two networks of electrically coupled inhibitory neurons in neocortex. Nature. 402:75-79.

Gillies, M.J., Traub, R.D., LeBeau, F.E., Davies, C.H., Gloveli, T., Buhl, E.H. and Whittington, M.A. 2002. A model of atropine-resistant theta oscillations in rat hippocampal area CA1. J. Physiol. 543:779-793.

Giocomo, L.M., Zilli, E.A., Fransén, E. and Hasselmo, M.E. 2007. Temporal frequency of subthreshold oscillations scales with entorhinal grid cell ﬁeld spacing. Science. 315:1719- 1722.

Giocomo, L.M. and Hasselmo, M.E. 2008. Time constants of h current in layer ii stellate cells differ along the dorsal to ventral axis of medial entorhinal cortex. J. Neurosci. 28:9414-9425.

Giocomo, L.M., Hussaini, S.A., Zheng, F., Kandel, E.R., Moser, M.B. and Moser, E.I. 2011. Grid cells use HCN1 channels for spatial scaling. Cell. 147:1159-1170.

Gloveli, T., Dugladze, T., Rotstein, H.G., Traub, R.D., Monyer, H., Heinemann, U., Whitting- ton, M.A. and Kopell, N.J. 2005. Orthogonal arrangement of rhythm-generating microcircuits in the hippocampus. Proc. Natl. Acad. Sci. USA. 102:13295-13300.

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.

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.

Gong, S., Doughty, M., Harbaugh, C.R., Cummins, A., Hatten, M.E., Heintz, N. and Gerfen, C.R. 2007. Targeting Cre recombinase to speciﬁc neuron populations with bacterial artiﬁcial chromosome constructs. J. Neurosci. 27:9817-9823.

Goutagny, R., Jackson, J. and Williams, S. 2009. Self-generated theta oscillations in the hippocampus. Nat. Neurosci. 12:1491-1493.

Grillner, S., Markram, H., De Schutter, E., Silberberg, G. and LeBeau F.E. 2005. Microcircuits in action-from CPGs to neocortex. Trends Neurosci. 28:525-33.

Grossberg, S. 1976a. Adaptive pattern classiﬁcation and universal recoding, I: Parallel development and coding of neural feature detectors. Biol. Cyber. 23:121-134.

55 Grossberg, S. 1976b. Adaptive pattern classiﬁcation and universal recoding, II: Feedback, expectations, olfaction, and illusions. Biol. Cyber. 23:187-202.

Gulyás, A. I., Megías, 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.

Haider, B., Duque, A., Hasenstaub, A.R. and McCormick, D.A. 2006. Neocortical network activity in vivo is generated through a dynamic balance of excitation and inhibition. J. Neurosci. 26:4535-4545.

Hangya, B., Borhegyi, Z., Szilágyi, N., Freund, T.F. and Varga, V. 2009. GABAergic neurons of the medial septum lead the hippocampal network during theta activity. J. Neurosci. 29:8094- 8102.

Hansel, D., Mato, G. and Meunier, C. 1995. Synchrony in Excitatory Neural Networks. Neural Computation. 7:307-337.

Harb, K., Magrinelli, E., Nicolas, C.S., Lukianets, N., Frangeul, L., Pietri, M., Sun, T., Sandoz, G., Grammont, F., Jabaudon, D., Studer, M., Alfano, C. 2016. Area-speciﬁc development of distinct projection neuron subclasses is regulated by postnatal epigenetic modiﬁcations. Elife. 5:e09531.

Harris, K.D. and Shepherd, G.M. 2015. The neocortical circuit: themes and variations. Nat. Neurosci. 18:170-181.

Harvey, C.D., Coen, P. and Tank, D.W. 2012. Choice-speciﬁc sequences in parietal cortex during a virtual-navigation decision task. Nature. 484:62-68.

Hasselmo, M.E. 2013. Neuronal rebound spiking, resonance frequency and theta cycle skipping may contribute to grid cell ﬁring in medial entorhinal cortex. Philos. Trans. R. Soc. Lond. B. Biol. Sci. 369:20120523.

Hattox, A.M. and Nelson, S.B. 2007. Layer V neurons in mouse cortex projecting to different targets have distinct physiological properties. J. Neurophysiol. 98:3330-3340.

Häusser, M., Spruston, N. and Stuart, G.J. 2000. Diversity and dynamics of dendritic signaling. Science. 290:739-744.

Hebb D.O. 1949. The Organization of Behavior. New York: Wiley & Sons.

Hilscher, M.M., Moulin, T., Skolnick, Y., Lytton, W.W. and Neymotin, S.A. 2012. Ih modulates theta rhythm and synchrony in computer model of CA3. BMC Neurosci. 13 (Suppl 1), P80.

Hilscher, M.M., Leão, K.E. and Leão, R.N. 2013. Synchronization through nonreciprocal connections in a hybrid hippocampus microcircuit. Front. Neural Circuits. 7:120.

56 Hilscher, M.M., Leão, R.N., Edwards, S.J., Leão, K.E. and Kullander, K. 2017. Chrna2- Martinotti Cells Synchronize layer 5 type A Pyramidal Cells via Rebound Excitation. PLoS Biol. 15:e2001392.

Hines, M., Davison, A. and Muller, E. 2009. NEURON and Python. Front. Neuroinform. 3:1.

Hodgkin, A. L. and Huxley, A. F. 1952. A quantitative description of membrane current and its application to conduction and excitation in nerve. J. Physiol. 117:500-544.

Hu, H. Vervaeke, K. and Storm, J.F. 2002. Two forms of electrical resonance at theta frequencies, generated by M-current, h-current and persistent Na+ current in rat hippocampal pyramidal cells. J. Physiol. 545:783-805.

Hu, H. Vervaeke, K., Graham, L.J. and Storm, J.F. 2009. Complementary theta resonance ﬁltering by two spatially segregated mechanisms in CA1 hippocampal pyramidal neurons. J. Neurosci. 29:14472-14483.

Hu, H., Cavendish, J.Z. and Agmon, A. 2013. Not all that glitters is gold: off-target recombination in the somatostatin–IRES-Cre mouse line labels a subset of fast-spiking interneurons. Front. Neural Circuits. 7:195.

Huang, Z.J. 2014. Toward a genetic dissection of cortical circuits in the mouse. Neuron. 83:1284-1302.

Hughes, D.I., Boyle, K.A., Kinnon, C.M., Bilsland, C., Quayle, J.A., Callister, R.J. and Gra- ham, B.A. 2013. HCN4 subunit expression in fast-spiking interneurons of the rat spinal cord and hippocampus. Neuroscience. 237:7-18.

Huh, C.Y., Amilhon, B., Ferguson, K.A., Manseau, F., Torres-Platas, S.G., Peach, J.P., Sco- dras, S., Mechawar, N., Skinner, F.K. and Williams, S. 2016. Excitatory Inputs Determine Phase-Locking Strength and Spike-Timing of CA1 Stratum Oriens/Alveus Parvalbumin and Somatostatin Interneurons during Intrinsically Generated Hippocampal Theta Rhythm. J. Neu- rosci. 36:6605-6622.

Hutcheon, B. and Yarom, Y. 2000. Resonance, oscillation and the intrinsic frequency preferences of neurons. Trends Neurosci. 23:216-222.

Huygens, C. 1664. Oeuvres complétes de Christiaan Huygens. La Societe Hollandaise des Sciences.

Hyaﬁl, A., Giraud, A.L., Fontolan, L. and Gutkin, B. 2015. Neural cross-frequency coupling: connecting architectures, mechanisms, and functions. Trends Neurosci. 38:725-740.

Ikegaya, Y., Aaron, G., Cossart, R., Aronov, D., Lampl, I., Ferster, D. and Yuste, R. 2004. Synﬁre chains and cortical songs: temporal modules of cortical activity. Science. 304:559-564.

Jaglin, X.H., Hjerling-Lefﬂer, J., Fishell, G. and Batista-Brito, R. 2012. The origin of neocortical nitric oxide synthase-expressing inhibitory neurons. Front. Neural Circuits. 6:44.

57 Jiang, X., Shen, S., Cadwell, C.R., Berens, P., Sinz, F., Ecker, A.S., Patel, S. and Tolias, A.S. 2015. Principles of connectivity among morphologically deﬁned cell types in adult neocortex. Science. 27:350.

Joshi, A., Middleton, J.W., Anderson, C.T., Borges, K., Suter, B.A., Shepherd, G.M. and Tzounopoulos, T. 2015. Cell-speciﬁc activity-dependent fractionation of layer 2/3 → 5B excitatory signaling in mouse auditory cortex. J. Neurosci. 35:3112:3123.

Kaltenbach, J.A., Zhang, J. and Afman, C.E. 2000. Plasticity of spontaneous neural activity in the dorsal cochlear nucleus after intense sound exposure. Hear. Res. 147:282-292.

Kamondi, A., Acsády, L., Wang, X.J. and Buzsáki, G 1998. Theta oscillations in somata and dendrites of hippocampal pyramidal cells in vivo: activity-dependent phase-precession of action potentials. Hippocampus. 8:244-261.

Kapfer, C., Glickfeld, L.L., Atallah, B.V. and Scanziani, M. 2007. Supralinear increase of recurrent inhibition during sparse activity in the somatosensory cortex. Nat. Neurosci. 10:743- 753.

Karnani, M.M., Agetsuma, M. and Yuste, R. 2014. A blanket of inhibition: functional inferences from dense inhibitory connectivity. Curr. Opin. Neurobiol. 26:96-102.

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.

Karnani, M.M., Jackson, J., Ayzenshtat, I., Tucciarone, J., Manoocheri, K., Snider, W.G. and Yuste, R. 2016. Cooperative Subnetworks of Molecularly Similar Interneurons in Mouse Neocortex. Neuron. 90:86-100.

Katona, L., Lapray, D., Viney, T.J., Oulhaj, A., Borhegyi, Z., Micklem, B.R., Klausberger, T. and Somogyi, P. 2016. Sleep and Movement Differentiates Actions of Two Types of Somatostatin-Expressing GABAergic Interneuron in Rat Hippocampus. Neuron. 91:1183.

Kim, E.J., Juavinett, A.L., Kyubwa, E.M., Jacobs, M.W. and Callaway, E.M. 2015. Three Types of Cortical Layer 5 Neurons That Differ in Brain-wide Connectivity and Function. Neuron. 88:1253-1267.

Kispersky, T.J., Fernandez, F.R., Economo, M.N. and White, J.A. 2012. Spike resonance properties in hippocampal O-LM cells are dependent on refractory dynamics. J. Neurosci. 32:3637-3651.

Klausberger, T., Magill, P.J., Márton, L.F., Roberts. J.D., Cobden, P.M., Buzsáki, G. and Som- ogyi, P. 2003. Brain-state- and cell-type-speciﬁc ﬁring of hippocampal interneurons in vivo. Nature. 421:844-848.

58 Klausberger, T., Marton, L.F., O’Neill, J., Huck, J.H., Dalezios, Y., Fuentealba, P., Suen, W.Y., Papp, E., Kaneko, T., Watanabe, M., Csicsvari, J. and Somogyi, P. 2005. Complementary roles of cholecystokinin- and parvalbumin-expressing GABAergic neurons in hippocampal network oscillations. J. Neurosci. 25:9783-9793.

Klausberger, T. and Somogyi, P. 2008. Neuronal diversity and temporal dynamics: the unity of hippocampal circuit operations. Science. 321:53-57.

Klausberger, T. 2009. GABAergic interneurons targeting dendrites of pyramidal cells in the CA1 area of the hippocampus. Eur. J. Neurosci. 30:947-957.

Kole, M.H., Hallermann, S. and Stuart, G.J. 2006. Single Ih channels in pyramidal neuron dendrites: properties, distribution, and impact on action potential output. J. Neurosci. 26:1677- 1687.

Kopell, N., Ermentrout, G.B., Whittington, M.A. and Traub, R.D. 2000. Gamma rhythms and beta rhythms have different synchronization properties. Proc. Natl. Acad. Sci. USA. 97:1867- 1872.

Kopell, N., Börgers, C., Pervouchine, D., Malerba, P. and Tort, A. 2010. Gamma and Theta Rhythms in Biophysical Models of Hippocampal Circuits. Hippocampal Microcircuits. 423- 457.

Kopell, N., Whittington, M.A. and Kramer, M.A. 2011. Neuronal assembly dynamics in the beta1 frequency range permits short-term memory. Proc. Natl. Acad. Sci. USA. 108:3779-3784.

Kramer, M.A., Roopun, A.K., Carracedo, L.M., Traub, R.D., Whittington, M.A. and Kopell, N.J. 2008. Rhythm generation through period concatenation in rat somatosensory cortex. PLoS Comput. Biol. 4:e1000169.

Kramis, R., Vanderwolf, C.H. and Bland, B.H. 1975. Two types of hippocampal rhythmical slow activity in both the rabbit and the rat: relations to behavior and effects of atropine, diethyl ether, urethane, and pentobarbital. Exp. Neurol. 49:58-85.

Kuramoto, Y. 1984. Chemical Oscillations, Waves, and Turbulence. New York, Spinger.

Kushmerick, C., Renden, R. and von Gersdorff, H. 2006. Physiological temperatures reduce the rate of vesicle pool depletion and short-term depression via an acceleration of vesicle recruitment. J. Neurosci. 26:1366-1377.

Lapicque, L. 1907. Recherches quantitatives sur l’excitation électrique des nerfs traitée comme une polarisation. J. Physiol. Pathol. Gen. 9:620-635.

Lawrence, J.J., Saraga, F., Churchill, J.F., Statland, J.M., Travis, K.E., Skinner, F.K. and McBain, C.J. 2006. Somatodendritic Kv7/KCNQ/M channels control interspike interval in hippocampal interneurons. J. Neurosci. 26:12325-12338.

59 Lawrence, J.J. 2008. Cholinergic control of GABA release: emerging parallels between neocortex and hippocampus. Trends Neurosci. 31:317-327.

Le Bé J.V., Silberberg, G., Wang, Y. and Markram, H. 2007. Morphological, electrophysiological, and synaptic properties of corticocallosal pyramidal cells in the neonatal rat neocortex. Cereb. Cortex 17:2204-2213.

Le Duigou, C., Simonnet, J., Teleñczuk, M.T., Fricker, D. and Miles, R. 2014. Recurrent synapses and circuits in the CA3 region of the hippocampus: an associative network. Front. Cell. Neurosci. 7:262.

Leão, K.E., Leão, R.N., Sun, H., Fyffe, R.E. and Walmsley, B. 2006. Hyperpolarization- activated currents are differentially expressed in mice brainstem auditory nuclei. J. Physiol. 576:849-864.

Leão, R.N., Mikulovic, S., Leão, K.E., Munguba, H., Gezelius, H., Enjin, A., Patra, K., Eriks- son, 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, C.K., Huguenard, J.R. 2011. Martinotti cells: community organizers. Neuron. 69:1042- 1045.

Lee, J.H., Whittington, M.A. and Kopell, N.J. 2013. Top-down beta rhythms support selective attention via interlaminar interaction: a model. PLoS Comput. Biol. 9:e1003164.

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. 11:1662-1670.

Lee, A.T., Gee, S.M., Vogt, D., Patel, T., Rubenstein, J.L. and Sohal, V.S. 2014. Pyramidal neurons in prefrontal cortex receive subtype-speciﬁc forms of excitation and inhibition. Neuron. 81:61-68.

Lee, J.H., Whittington, M.A. and Kopell, N.J. 2015. Potential Mechanisms Underlying Inter- cortical Signal Regulation via Cholinergic Neuromodulators. J. Neurosci. 35:15000-15014.

Li, X., Morita, K., Robinson, H.P. and Small, M. 2013. Control of layer 5 pyramidal cell spiking by oscillatory inhibition in the distal apical dendrites: a computational modeling study. J. Neurophysiol. 109:2739-2756.

Liguz-Lecznar, M., Urban-Ciecko, J. and Kossut, M. 2016. Somatostatin and Somatostatin- Containing Neurons in Shaping Neuronal Activity and Plasticity. Front. Neural Circuits. 10:48.

Lisman, J.E. and Idiart, M.A. 1995. Storage of 7 +/- 2 short-term memories in oscillatory subcycles. Science. 267:1512-1515.

Lisman, J. 2005. The theta/gamma discrete phase code occuring during the hippocampal phase precession may be a more general brain coding scheme. Hippocampus. 15:913-922.

60 Llinás, R. and Yarom, Y. 1981. Electrophysiology of mammalian inferior olivary neurones in vitro. Different types of voltage-dependent ionic conductances. J. Physiol. 315:549-567.

Lorente de Nó, R. 1933. Studies on the structure of the cerebral cortex. I. Area entorhinalis. J. Psychol. Neurol. 45:381–438.

Lorente de Nó, R. 1934. Studies on the structure of the cerebral cortex. II. Continuation of the study of the Ammonic system. J. Psychol. Neurol. 46:113-177.

Lörincz, A., Notomi, T., Tamás, G., Shigemoto, R. and Nusser, Z. 2002. Polarized and compartment-dependent distribution of HCN1 in pyramidal cell dendrites. Nat. Neurosci. 5:1185-1193.

Lovett-Barron, M., Turi, G.F., Kaifosh, P., Lee, P.H., Bolze, F., Sun, X.H., Nicoud, J.F., Zemel- man, B.V., Sternson, S.M. and Losonczy A. 2012. Regulation of neuronal input transformations by tunable dendritic inhibition. Nat. Neurosci. 15:423-430.

Lovett-Barron, M., Kaifosh, P., Kheirbek, M.A., Danielson, N., Zaremba, J.D., Reardon, T.R., Turi, G.F., Hen, R., Zemelman, B.V. and Losonczy A. 2014. Dendritic inhibition in the hippocampus supports fear learning. Science. 343:857:863.

Löwel, S. and Singer, W. 1992. Selection of intrinsic horizontal connections in the visual cortex by correlated neuronal activity. Science. 255:209-212.

Lupica, C.R., Bell, J.A., Hoffman, A.F. and Watson, P.L. 2001. Contribution of the hyperpolarization-activated current (I(h)) to membrane potential and GABA release in hippocampal interneurons. J. Neurophysiol. 86:261-268.

Lüthi, A. and McCormick, D.A. 1996. H-current: properties of a neuronal and network pacemaker. Neuron. 21:9-12.

Lytton, W. W. and Sejnowski, T.J. 1991. Simulations of cortical pyramidal neurons synchronized by inhibitory interneurons. J. Neurophysiol. 66:1059-1079.

Lytton, W. W. 2006. Neural query system: data-mining from within the NEURON simulator. Neuroinformatics. 4:163-176.

Ma, Y., Hu, H., Berrebi, A.S., Mathers, P.H. and Agmon, A. 2006. Distinct subtypes of somatostatin-containing neocortical interneurons revealed in transgenic mice. J. Neurosci. 26:5069-5082.

Maccaferri, G. and McBain, C.J. 1996. The hyperpolarization-activated current (Ih) and its contribution to pacemaker activity in rat CA1 hippocampal stratum oriens-alveus interneurones. J. Physiol. 497:119-130.

Maccaferri, G. and Lacaille, J.C. 2003. Interneuron Diversity series: Hippocampal interneuron classiﬁcations–making things as simple as possible, not simpler. Trends Neurosci. 26:564-571.

61 Madisen, L., Zwingman, T.A., Sunkin, S.M., Oh, S.W., Zariwala, H.A., Gu, H., Ng, L.L., Palmiter, R.D., Hawrylycz, M.J., Jones, A.R., Lein, E.S. and Zeng, H. 2010. A robust and high-throughput Cre reporting and characterization system for the whole mouse brain. Nat. Neurosci. 13:133-140.

Madisen, L., Mao, T., Koch, H., Zhuo, J.M., Berenyi, A., Fujisawa, S., Hsu, Y.W., Garcia. A.J. 3rd, Gu, X., Zanella, S., Kidney, J., Gu, H., Mao, Y., Hooks, B.M., Boyden, E.S., Buzsáki, G., Ramirez, J.M., Jones, A.R., Svoboda, K., Han, X., Turner, E.E. and Zeng, H. 2012. A toolbox of Cre-dependent optogenetic transgenic mice for light-induced activation and silencing. Nat. Neurosci. 15:793-802.

Maeda, E., Robinson, H.P. and Kawana, A. 1995. The mechanisms of generation and propagation of synchronized bursting in developing networks of cortical neurons. J. Neurosci. 15:6834-6845.

Magee, J.C. 1999. Dendritic Ih normalizes temporal summation in hippocampal CA1 neurons. Nat. Neurosci. 2:848.

Magee, J.C. 2000. Dendritic integration of excitatory synaptic input. Nat. Rev. Neurosci. 1:181-190.

Mahn, M., Prigge, M., Ron, S., Levy, R. and Yizhar, O. 2016. Biophysical constraints of optogenetic inhibition at presynaptic terminals. Nat. Neurosci. 19:554-556.

Malik, R., Dougherty, K.A., Parikh, K., Byrne, C. and Johnston D. 2016. Mapping the electrophysiological and morphological properties of CA1 pyramidal neurons along the longitudinal hippocampal axis. Hippocampus. 26:341-361.

Marcelin, B., Chauviére, L., Becker, A., Migliore, M., Esclapez, M. and Bernard, C. 2009. h channel-dependent deﬁcit of theta oscillation resonance and phase shift in temporal lobe epilepsy. Neurobiol. Dis. 33:436-447.

Marcelin, B., Liu, Z., Chen, Y., Lewis, A.S., Becker, A., McClelland, S., Chetkovich, D.M., Migliore, M., Baram, T.Z., Esclapez, M. and Bernard, C. 2012a. Dorsoventral differences in intrinsic properties in developing CA1 pyramidal cells. J. Neurosci. 32:3736-3747.

Marcelin, B., Lugo, J.N., Brewster, A.L., Liu, Z., Lewis, A.S., McClelland, S., Chetkovich, D.M., Baram, T.Z., Anderson, A.E., Becker, A., Esclapez, M. and Bernard C. 2012b. Differen- tial dorso-ventral distributions of Kv4.2 and HCN proteins confer distinct integrative properties to hippocampal CA1 pyramidal cell distal dendrites. J. Biol. Chem. 287:17656-17661.

Markram, H., Toledo-Rodriguez, M., Wang, Y., Gupta, A., Silberberg, G. and Wu, C. 2004. Interneurons of the neocortical inhibitory system. Nat. Rev. Neurosci. 5:793-807.

Markram, H., Muller, E., Ramaswamy, S., Reimann, M.W., Abdellah, M., Sanchez, C.A., Ailamaki, A., Alonso-Nanclares, L., Antille, N., Arsever, S., Kahou, G.A., Berger, T.K., Bilgili, A., Buncic, N., Chalimourda, A., Chindemi, G., Courcol, J.D., Delalondre, F., Delattre, V.,

62 Druckmann, S., Dumusc, R., Dynes, J., Eilemann, S., Gal, E., Gevaert, M.E., Ghobril, J.P., Gidon, A., Graham, J.W., Gupta, A., Haenel, V., Hay, E., Heinis, T., Hernando, J.B., Hines, M., Kanari, L., Keller, D., Kenyon, J., Khazen, G., Kim, Y., King, J.G., Kisvarday, Z., Kumbhar, P., Lasserre, S., Le Bé, J.V., Magalhães, B.R., Merchán-Pérez, A., Meystre, J., Morrice, B.R., Muller, J., Muñoz-Céspedes, A., Muralidhar, S., Muthurasa, K., Nachbaur, D., Newton, T.H., Nolte, M., Ovcharenko, A., Palacios, J., Pastor, L., Perin, R., Ranjan, R., Riachi, I., Rodríguez, J.R., Riquelme, J.L., Rössert, C., Sfyrakis, K., Shi, Y., Shillcock, J.C., Silberberg, G., Silva, R., Tauheed, F., Telefont, M., Toledo-Rodriguez, M., Tränkler, T., Van Geit, W., Díaz, J.V., Walker, R., Wang, Y., Zaninetta, S.M., DeFelipe, J., Hill, S.L., Segev, I., Schürmann, F. 2015. Reconstruction and Simulation of Neocortical Microcircuitry. Cell. 163:456-492.

Maroso, M., Szabo, G.G., Kim, H.K., Alexander, A., Bui, A.D., Lee, S.H., Lutz, B. and Soltesz, I. 2016. Cannabinoid Control of Learning and Memory through HCN Channels. Neuron. 89:1059-1073.

Martinotti, C. 1889. Contributo allo studio della corteccia cerebrale, ed all’origine centrale dei nervi. Ann. Freniatr. Sci. Afﬁni. 1:14-381.

Martinotti, C. 1890. Beitrag zum Studium der Hirnrinde und dem Centralursprung der Nerven. Int. Monatsschr. Anat. Physiol. 7:69-90.

Matt, L., Michalakis, S., Hofmann, F., Hammelmann, V., Ludwig, A., Biel, M. and Kleppisch, T. 2011. HCN2 channels in local inhibitory interneurons constrain LTP in the hippocampal direct perforant path. Cell. Mol. Life Sci. 68:125-137.

McCormick, D.A. and Huguenard, J.R. 1992. A model of the electrophysiological properties of thalamocortical relay neurons. J. Neurophysiol. 68:1384-1400.

Migliore, M., Messineo, L. and Ferrante, M. 2004. Dendritic Ih selectively blocks temporal summation of unsynchronized distal inputs in CA1 pyramidal neurons. J. Comput. Neurosci. 16:5-13.

Migliore, M. and Migliore, R. 2012. Know Your Current Ih: Interaction with a Shunting Current Explains the Puzzling Effects of Its Pharmacological or Pathological Modulations. PLoS One. 7:e36867.

Mikulovic, S., Restrepo, C.E., Hilscher, M.M., Kullander, K. and Leão, R.N. 2015. Novel markers for OLM interneurons in the hippocampus. Front. Cell. Neurosci. 9:201.

Miles, R., Tóth, K., Gulyás, A. I., Hájos, N. and Freund, T. F. 1996. Differences between somatic and dendritic inhibition in the hippocampus. Neuron. 16:815–823.

Milner, P.M. 1974. A model for visual shape recognition. Psychol. Review. 81:521-535.

Mirollo, R.E. and Strogatz, S.H. 1990. Synchronization of pulse-coupled biological oscillators. SIAM J. Appl. Math. 50:1645-1662.

63 Moosmang, S., Biel, M., Hofmann, F. and Ludwig, A. 1999. Differential distribution of four hyperpolarization-activated cation channels in mouse brain. Biol. Chem. 380:975-980.

Müller, C., Beck, H., Coulter, D. and Remy, S. 2012. Inhibitory control of linear and supralinear dendritic excitation in CA1 pyramidal neurons. Neuron. 75:851-864.

Müller, C. and Remy, S. 2014. Dendritic inhibition mediated by O-LM and bistratiﬁed interneurons in the hippocampus. Front. Synaptic Neurosci. 6:23.

Murayama, M., Pérez-Garci, E., Nevian, T., Bock, T., Senn, W. and Larkum, M.E. 2009. Den- dritic encoding of sensory stimuli controlled by deep cortical interneurons. Nature. 457:1137- 1141.

Naka, A. and Adesnik, H. 2016. Inhibitory Circuits in Cortical Layer 5. Front. Neural Circuits. 10:35.

Narayanan, R. and Johnston, D. 2008. The h channel mediates location dependence and plasticity of intrinsic phase response in rat hippocampal neurons. J. Neurosci. 28:5846-5860.

Netoff, T.I., Banks, M.I., Dorval, A.D., Acker, C.D., Haas, J.S., Kopell, N. and White, J.A. 2005. Synchronization in hybrid neuronal networks of the hippocampal formation. J. Neuro- physiol. 93:1197-1208.

Neuenschwander, S. and Singer, W. 1996. Long-range synchronization of oscillatory light responses in the cat retina and lateral geniculate nucleus. Nature. 379:728-732.

Neymotin, S.A., Lazarewicz, M.T., Sherif, M., Contreras, D., Finkel, L.H. and Lytton, W.W. 2011. Ketamine Disrupts Theta Modulation of Gamma in a Computer Model of Hippocampus. J. Neurosci. 31:11733-11743.

Neymotin, S. A., Hilscher, M.M., Moulin, T.C., Skolnick, Y., Lazarewicz, M.T. and Lytton, W.W. 2013. Ih tunes theta/gamma oscillations and cross-frequency coupling in an in silico CA3 model. PLoS One. 8:e76285.

Nolan, M.F., Malleret, G., Dudman, J.T., Buhl, D.L., Santoro, B., Gibbs, E., Vronskaya, S., Buzsáki, G., Siegelbaum, S.A., Kandel, E.R. and Morozov, A. 2004. A behavioral role for dendritic integration: HCN1 channels constrain spatial memory and plasticity at inputs to distal dendrites of CA1 pyramidal neurons. Cell. 119:719-732.

Okubo, T.S., Mackevicius, E.L., Payne, H.L., Lynch, G.F. and Fee M.S. 2015. Growth and splitting of neural sequences in songbird vocal development. Nature. 528:352-357.

Okun, M. and Lampl, I. 2008. Instantaneous correlation of excitation and inhibition during ongoing and sensory-evoked activities. Nat. Neurosci. 11:535-537.

Oliva, A.A. Jr., Jiang, M., Lam, T., Smith, K.L. and Swann, J.W. 2000. Novel hippocampal interneuronal subtypes identiﬁed using transgenic mice that express green ﬂuorescent protein in GABAergic interneurons. J. Neurosci. 20:3354-3368.

64 Olufsen, M. S., Whittington, M. A., Camperi, M. and Kopell, N. 2003. New roles for the gamma rhythm: population tuning and preprocessing for the beta rhythm. J. Comput. Neurosci. 14:33-54.

Palmer, L., Murayama, M. and Larkum, M. 2012. Inhibitory Regulation of Dendritic Activity in vivo. Front. Neural Circuits. 6:26.

Pape, H.C 1996. Queer current and pacemaker: the hyperpolarization-activated cation current in neurons. Annu. Rev. Physiol. 58:299-327.

Pastoll, H., Solanka, L., van Rossum, M.C. and Nolan, M.F. 2013. Feedback inhibition enables θ-nested γ oscillations and grid ﬁring ﬁelds. Neuron. 77:141-154.

Patra, K., Lyons, D.J., Bauer, P., Hilscher, M.M., Sharma, S., Leão, R.N. and Kullander, K. 2015. A role for solute carrier family 10 member 4, or vesicular aminergic-associated transporter, in structural remodelling and transmitter release at the mouse neuromuscular junction. Eur. J. Neurosci. 41:316-327.

Peña Ramirez, J., Olvera, L. A., Nijmeijer, H. and Alvarez, J. 2016. The sympathy of two pendulum clocks: beyond Huygens0 observations. Sci. Rep. 6:23580.

Penttonen, M., Kamondi, A., Acsády, L., Buzsáki, G. 1998. Gamma frequency oscillation in the hippocampus of the rat: intracellular analysis in vivo. Eur. J. Neurosci. 10:718-728.

Perry, S., Gezelius, H., Larhammar, M., Hilscher, M.M., Lamotte d’Incamps. B., Leão, K.E. and Kullander, K. 2015. Firing properties of Renshaw cells deﬁned by Chrna2 are modulated by hyperpolarizing and small conductance ion currents Ih and ISK. Eur. J. Neurosci. 41:889-900.

Petsche, H., Stumpf, C., and Gogolak, G. 1962. The signiﬁcance of the rabbit’s septum as a relay station between the midbrain and the hippocampus. I. The control of hippocampus arousal activity by the septum cells. Electroencephalogr Clin Neurophysiol. 14:202-211.

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.

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.

Pike, F.G., Goddard, R.S., Suckling, J.M., Ganter, P., Kasthuri, N. and Paulsen, O. 2000. Distinct frequency preferences of different types of rat hippocampal neurones in response to oscillatory input currents. J. Physiol. 529:205-213.

Plenz, D. and Thiagarajan, T.C. 2007. The organizing principles of neuronal avalanches: cell assemblies in the cortex? Trends Neurosci. 30:101-110.

65 Poo, M.M., Pignatelli, M., Ryan, T.J., Tonegawa, S., Bonhoeffer, T., Martin, K.C., Rudenko, A., Tsai, L.H., Tsien, R.W., Fishell, G., Mullins, C., Gonçalves, J.T., Shtrahman, M., Johnston, S.T., Gage, F.H., Dan, Y., Long, J., Buzsáki, G. and Stevens, C. 2016. What is memory? The present state of the engram. BMC Biol. 14:40.

Poolos, N.P., Bullis, J.B. and Roth, M.K. 2006. Modulation of h-channels in hippocampal pyramidal neurons by p38 mitogen-activated protein kinase. J. Neurosci. 26:7995-8003.

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.

Raimondo, J.V., Kay, L., Ellender, T.J. and Akerman, C.J. 2012. Optogenetic silencing strategies differ in their effects on inhibitory synaptic transmission. Nat. Neurosci. 15:1102-1104.

Ramaswamy, S. and Markram, H. 2015. Anatomy and physiology of the thick-tufted layer 5 pyramidal neuron. Front. Cell. Neurosci. 9:233.

Ribeiro, T.L., Ribeiro, S. and Copelli, M. 2016. Repertoires of spike avalanches are modulated by behavior and novelty. Front. Neural Circuits. 10:16.

Robinson, R.B. and Siegelbaum, S.A. 2003. Hyperpolarization-activated cation currents: from molecules to physiological function. Annu. Rev. Physiol. 65:453-480.

Roopun, A.K., Middleton, S.J., Cunningham, M.O,. LeBeau, F.E., Bibbig, A., Whittington, M.A. and Traub, R.D. 2006. A beta2-frequency (20-30 Hz) oscillation in nonsynaptic networks of somatosensory cortex. Proc. Natl. Acad. Sci. USA. 103:15646-15650.

Roopun, A.K., Kramer, M.A., Carracedo, L.M., Kaiser, M., Davies, C.H., Traub, R.D., Kopell, N.J. and Whittington, M.A. 2008. Period concatenation underlies interactions between gamma and beta rhythms in neocortex. Front. Cell. Neurosci. 2:1.

Roopun, A.K., Lebeau, F.E., Rammell, J., Cunningham, M.O., Traub, R.D. and Whittington, M.A. 2010. Cholinergic neuromodulation controls directed temporal communication in neocortex in vitro. Front. Neural Circuits. 4:8.

Rotstein, H.G., Pervouchine, D.D., Acker, C.D., Gillies, M.J., White, J.A., Buhl, E.H., Whit- tington, M.A. and Kopell, N. 2005. Slow and fast inhibition and an H-current interact to create a theta rhythm in a model of CA1 interneuron network. J. Neurophysiol. 94:1509-1518.

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.

Sainsbury, R.S., Heynen, A. and Montoya, C.P. 1987. Behavioral correlates of hippocampal type 2 theta in the rat. Physiol. Behav. 39:513-519.

Sanchez-Vives, M.V. and McCormick, D.A. 2000. Cellular and network mechanisms of rhythmic recurrent activity in neocortex. Nat. Neurosci. 3:1027-1034.

66 Santoro, B., Chen, S., Lüthi, A., Pavlidis, P., Shumyatsky, G. P., Tibbs, G. R. and Siegelbaum, S. A. 2000. Molecular and functional heterogeneity of hyperpolarization-activated pacemaker channels in the mouse CNS. J. Neurosci. 20:5264-5275.

Santoro, B. and Baram, T.Z. 2003. The multiple personalities of h-channels. Trends Neurosci. 26:550-554.

Saraga, F., Wu, C.P., Zhang, L. and Skinner, F.K. 2003. Active dendrites and spike propagation in multi-compartment models of oriens-lacunosum/moleculare hippocampal interneurons. J. Physiol. 552:673-689.

Sekulic,´ V., Lawrence, J.J. and Skinner, F.K. 2014. Using multi-compartment ensemble modeling as an investigative tool of spatially distributed biophysical balances: application to hippocampal oriens-lacunosum/moleculare (O-LM) cells. PLoS One. 9:e106567.

Sekulic,´ V., Chen, T.C., Lawrence, J.J. and Skinner, F.K. 2014. 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.

Shadlen, M.N. and Newsome, W.T. 1994. Noise, neural codes and cortical organization. Curr. Opin. Neurobiol. 4:569-579.

Shadlen, M.N. and Movshon, J.A. 1999. Synchrony unbound: a critical evaluation of the temporal binding hypothesis. Neuron. 24:67-77.

Shah, S., Lubeck, E., Zhou, W. and Cai, L. 2016. In Situ Transcription Proﬁling of Single Cells Reveals Spatial Organization of Cells in the Mouse Hippocampus. Neuron. 92:342-357.

Sherman, S.M. 2001. Tonic and burst ﬁring: dual modes of thalamocortical relay. Trends Neurosci. 24:122-126.

Sholl, D.A. 1953. Dendritic organization in the neurons of the visual and motor cortices of the cat. J. Anat. 87:387-406.

Shore, S.E., Koehler, S., Oldakowski, M., Hughes, L.F. and Syed, S. 2008. Dorsal cochlear nucleus responses to somatosensory stimulation are enhanced after noise-induced hearing loss. Eur. J. Neurosci. 27:155-168.

Shu, Y., Hasenstaub, A. and McCormick, D.A. 2003. Turning on and off recurrent balanced cortical activity. Nature. 423:288-293.

Sik, A., Penttonen, M., Ylinen, A. and Buzsáki, 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.

Singer, W. 1999. Neuronal synchrony: a versatile code for the deﬁnition of relations? Neuron. 24:49-65.

67 Sohal, V.S. and Huguenard, J.R. 2005. Inhibitory coupling speciﬁcally generates emergent gamma oscillations in diverse cell types. Proc. Natl. Acad. Sci. USA. 102:18638-18643.

Spruston, N. 2008. Pyramidal neurons: dendritic structure and synaptic integration. Nat. Rev. Neurosci. 9:206-221.

Stacey, W.C., Lazarewicz, M.T. and Litt, B. 2009. Synaptic noise and physiological coupling generate high-frequency oscillations in a hippocampal computational model. J. Neurophysiol. 102:2342-2357.

Stewart, M. and Fox, S.E. 1990. Do septal neurons pace the hippocampal theta rhythm? Trends Neurosci. 13:163-168.

Strange, B.A., Witter, M.P., Lein, E.S. and Moser, E.I. 2014. Functional organization of the hippocampal longitudinal axis. Nat. Rev. Neurosci. 15:655-669.

Sun, Y.J., Kim. Y.J., Ibrahim, L.A., Tao, H.W. and Zhang, L.I. 2013. Synaptic mechanisms underlying functional dichotomy between intrinsic-bursting and regular-spiking neurons in auditory cortical layer 5. J. Neurosci. 33:5326-5339.

Surges, R., Freiman, T.M. and Feuerstein, T.J. 2004. Input resistance is voltage dependent due to activation of Ih channels in rat CA1 pyramidal cells. J. Neurosci. Res. 76:475:480.

Tamás, G., Lörincz, A., Simon, A. and Szabadics, J. 2003. Identiﬁed sources and targets of slow inhibition in the neocortex. Science. 299:1902-1905.

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, S.B. and Huang, Z.J. 2011. A resource of Cre driver lines for genetic targeting of GABAergic neurons in cerebral cortex. Neuron. 71:995-1013.

Taniguchi, H. 2014. Genetic dissection of GABAergic neural circuits in mouse neocortex. Front. Cell. Neurosci. 8:8.

Thompson, S.M., Masukawa, L.M. and Prince, D.A. 1985. Temperature dependence of intrinsic membrane properties and synaptic potentials in hippocampal CA1 neurons in vitro. J. Neurosci. 5:817-824.

Timofeev, I., Grenier, F., Bazhenov, M., Sejnowski, T.J. and Steriade, M. 2000. Origin of slow cortical oscillations in deafferented cortical slabs. Cereb. Cortex. 10:1185-1199.

Ting, J.T. and Feng, G. 2013. Development of transgenic animals for optogenetic manipulation of mammalian nervous system function: progress and prospects for behavioral neuroscience. Behav. Brain Res. 255:3-18.

Ting, J.T., Daigle, T.L., Chen, Q. and Feng, G. 2014. Acute brain slice methods for adult and aging animals: application of targeted patch clamp analysis and optogenetics. Methods. Mol. Biol. 1183:221-242.

68 Tort, A. B. L., Rotstein, H. G., Dugladze, T., Gloveli, T. and Kopell, N. J. 2007. On the For- mation of Gamma-Coherent Cell Assemblies by Oriens Lacunosum-Moleculare Interneurons in the Hippocampus. Proc. Natl. Acad. Sci. USA. 104:13490-13495.

Tort, A. B. L., Komorowski, R., Eichenbaum, H. and Kopell, N. 2010. Measuring phase- amplitude coupling between neuronal oscillations of different frequencies. J. Neurophysiol. 104:1195-1210.

Traub, R. D., and Miles, R. 1991. Neuronal Networks of the Hippocampus. Cambridge Univ. Press.

Traub, R. D., Wong, R.K., Miles, R. and Michelson, H. 1991. A model of a CA3 hippocampal pyramidal neuron incorporating voltage-clamp data on intrinsic conductances. J. Neurophysiol. 66:635-650.

Tremblay, R., Lee, S. and Rudy, B. 2016. GABAergic Interneurons in the Neocortex: From Cellular Properties to Circuits. Neuron. 91:260-292.

Tsuno, Y., Chapman, G.W. and Hasselmo, M.E. 2015. Rebound spiking properties of mouse medial entorhinal cortex neurons in vivo. Eur. J. Neurosci. 42:2974-2984.

Tukker, J.J., Fuentealba, P., Hartwich, K., Somogyi, P. and Klausberger, T. 2007. Cell type- speciﬁc tuning of hippocampal interneuron ﬁring during gamma oscillations in vivo. J. Neu- rosci. 27:8184-8189.

Turner, J.G., Hughes, L.F. and Caspary, D.M. 2005. Divergent response properties of layer-V neurons in rat primary auditory cortex. Hear. Res. 202:129-140.

Vaidya, S.P. and Johnston, D. 2013. Temporal synchrony and gamma-to-theta power conver- sion in the dendrites of CA1 pyramidal neurons. Nat. Neurosci. 16:1812-1920. van der Malsburg, C. 1981. The correlation theory of brain function. Max-Planck-Institute for Biophysical Chemistry.

Van Vreeswijk, C., Abbott, L. F. and Ermentrout, G. B. 1994. When inhibition not excitation synchronizes neural ﬁring. J. Comput. Neurosci. 1:313-321.

Van Vreeswijk, C. and Sompolinsky, H. 1996. Chaos in neuronal networks with balanced excitatory and inhibitory activity. Science. 274:1724-1726.

Vanderwolf, C.H. 1969. Hippocampal electrical activity and voluntary movement in the rat. Electroencephalogr. Clin. Neurophysiol. 26:407-418.

Vinogradova, O.S. 1995. Expression, control, and probable functional signiﬁcance of the neuronal theta-rhythm. Prog. Neurobiol. 45:523-583.

Volgushev, M., Chauvette, S., Mukovski, M. and Timofeev, I. 2006. Precise long-range synchronization of activity and silence in neocortical neurons during slow-wave oscillations. J. Neurosci. 26:5665-5672.

69 Von Neumann, J. 1958. The computer and the brain. Yale University Press.

Wahl-Schott, C. and Biel, M. 2009. HCN channels: structure, cellular regulation and physiological function. Cell. Mol. Life Sci. 66:470-494.

Wang, X.J. and Rinzel, J. 1992. Alternating and Synchronous Rhythms in Reciprocally In- hibitory Model Neurons. Neural Comput. 4:84-97.

Wang, X.J. and Buzsáki, G. 1996. Gamma oscillation by synaptic inhibition in a hippocampal interneuronal network model. J. Neurosci. 16:6402-6413.

Wang, X.J. 2002. Pacemaker neurons for the theta rhythm and their synchronization in the septohippocampal reciprocal loop. J. Neurophysiol. 87:889-900.

Wang, Y., Toledo-Rodriguez, M., Gupta, A., Wu, C., Silberberg, G., Luo, J. and Markram, H. 2004. Anatomical, physiological and molecular properties of Martinotti cells in the somatosensory cortex of the juvenile rat. J. Physiol. 56:65-90.

Wang, X.-J. 2010. Neurophysiological and computational principles of cortical rhythms in cognition. Physiol. Rev. 90:1195-1268.

White, J.A., Banks, M.I., Pearce, R.A. and Kopell, N.J. 2000. Networks of interneurons with fast and slow gamma-aminobutyric acid type A (GABAA) kinetics provide substrate for mixed gamma-theta rhythm. Proc. Natl. Acad. Sci. USA. 97:8128-8133.

Whittington, M.A., Traub, R.D. and Jefferys, J.G.R. 1995. Synchronized oscillations in interneuron networks driven by metabotropic glutamate receptor activation. Nature. 373:612- 615.

Whittington, M.A., Traub, R.D., Faulkner, H.J., Stanford, I.M. and Jefferys, J.G. 1997. Recur- rent excitatory postsynaptic potentials induced by synchronized fast cortical oscillations. Proc. Natl. Acad. Sci. USA. 94:12198-12203.

Winfree, A.T. 1967. Biological rhythms and the behavior of populations of coupled oscillators. J. Theor. Biol. 16:15-42.

Womelsdorf, T., Schoffelen, J.M., Oostenveld, R., Singer, W., Desimone, R., Engel, A.K. and Fries. P. 2007. Modulation of neuronal interactions through neuronal synchronization. Science. 316:1609-1612.

Womelsdorf, T., Valiante, T.A., Sahin, N.T., Miller, K.J. and Tiesinga, P. 2014. Dynamic circuit motifs underlying rhythmic gain control, gating and integration. Nat. Neurosci. 17:1031-1039.

Wong, R.K. and Traub, R.D. 1983. Synchronized burst discharge in disinhibited hippocampal slice. I. Initiation in CA2-CA3 region. J. Neurophysiol. 49:442-458.

Wu, X., Liao, L., Liu, X., Luo, F., Yang, T. and Li, C. 2012. Is ZD7288 a selective blocker of hyperpolarization-activated cyclic nucleotide-gated channel currents? Channels (Austin) 76:438-442.

70 Wu, C., Martel, D.T. and Shore, S.E. 2016. Increased Synchrony and Bursting of Dorsal Cochlear Nucleus Fusiform Cells Correlate with Tinnitus. J. Neurosci. 36:2068-2073.

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.

Xu, H., Jeong, H.Y., Tremblay, R. and Rudy, B. 2013. Neocortical somatostatin-expressing GABAergic interneurons disinhibit the thalamorecipient layer 4. Neuron. 77:155-167.

Yamada, R., Kuba, H., Ishii, T.M. and Ohmori, H. 2005. Hyperpolarization-activated cyclic nucleotide-gated cation channels regulate auditory coincidence detection in nucleus laminaris of the chick. J. Neurosci. 25:8867-8877.

Ye, J.H., Zhang, J., Xiao, C. and Kong, J.Q. 2006. Patch-clamp studies in the CNS illustrate a simple new method for obtaining viable neurons in rat brain slices: glycerol replacement of NaCl protects CNS neurons. J. Neurosci. Methods 158:251-259.

Ylinen, A., Soltész, I., Bragin, A., Penttonen, M., Sik, A. and Buzsáki, G 1995. Intracellular correlates of hippocampal theta rhythm in identiﬁed pyramidal cells, granule cells, and basket cells. Hippocampus. 5:78-90.

Zeisel, A., Muñoz-Manchado, A.B., Codeluppi, S., Lönnerberg, P., La Manno, G., Juréus, A., Marques, S., Munguba, H., He, L., Betsholtz, C., Rolny, C., Castelo-Branco, G., Hjerling- Lefﬂer, J. and Linnarsson, S. 2015. Brain structure. Cell types in the mouse cortex and hippocampus revealed by single-cell RNA-seq. Science. 347:1138-1142.

Zemankovics, R., Káli, S., Paulsen, O., Freund, T.F. and Hájos, N. 2010. Differences in subthreshold resonance of hippocampal pyramidal cells and interneurons: the role of h-current and passive membrane characteristics. J. Physiol. 588:2109-2132.

Zhao, S., Cunha, C., Zhang, F., Liu, Q., Gloss, B., Deisseroth, K., Augustine, G.J. and Feng, G. 2008. Improved expression of halorhodopsin for light-induced silencing of neuronal activity. Brain Cell Bio. 36:141-154.

Zong, X., Krause, S., Chen, C.-C., Krüger, J., Gruner, C., Cao-Ehlker, X., Fenske, S., Wahl- Schott, C. and Biel, M. 2012. Regulation of Hyperpolarization-activated Cyclic Nucleotide- gated (HCN) Channel Activity by cCMP. Biol. Chem. 287:26506-26512.

APPENDIX A Study I

73 Ih Tunes Theta/Gamma Oscillations and Cross-Frequency Coupling In an In Silico CA3 Model

Samuel A. Neymotin1,2*, Markus M. Hilscher3,4, Thiago C. Moulin5, Yosef Skolnick1,6, Maciej T. Lazarewicz7¤, William W. Lytton1,8,9 1 Department of Physiology & Pharmacology, State University of New York Downstate, Brooklyn, New York, United States of America, 2 Department of Neurobiology, Yale University School of Medicine, New Haven, Connecticut, United States of America, 3 Vienna University of Technology, Vienna, Austria, 4 Department of Neuroscience, Uppsala University, Uppsala, Sweden, 5 Medical Biochemistry Institute, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil, 6 Computer Science Program, Brooklyn College, City University of New York, Brooklyn, New York, United States of America, 7 Department of Bioengineering, University of Pennsylvania, Philadelphia, Pennsylvania, United States of America, 8 Department of Neurology, State University of New York Downstate, Brooklyn, New York, United States of America, 9 Kings County Hospital, Brooklyn, New York, United States of America

Abstract

Ih channels are uniquely positioned to act as neuromodulatory control points for tuning hippocampal theta (4–12 Hz) and gamma (w25 Hz) oscillations, oscillations which are thought to have importance for organization of information flow. Ih contributes to neuronal membrane resonance and resting membrane potential, and is modulated by second messengers. We investigated Ih oscillatory control using a multiscale computer model of hippocampal CA3, where each cell class (pyramidal, basket, and oriens-lacunosum moleculare cells), contained type-appropriate isoforms of Ih. Our model demonstrated that modulation of pyramidal and basket Ih allows tuning theta and gamma oscillation frequency and amplitude. Pyramidal Ih also controlled cross-frequency coupling (CFC) and allowed shifting gamma generation towards particular phases of the theta cycle, effected via Ih ’s ability to set pyramidal excitability. Our model predicts that in vivo neuromodulatory control of Ih allows flexibly controlling CFC and the timing of gamma discharges at particular theta phases.

Citation: Neymotin SA, Hilscher MM, Moulin TC, Skolnick Y, Lazarewicz MT, et al. (2013) Ih Tunes Theta/Gamma Oscillations and Cross-Frequency Coupling In an In Silico CA3 Model. PLoS ONE 8(10): e76285. doi:10.1371/journal.pone.0076285 Editor: Gennady Cymbalyuk, Georgia State University, United States of America Received June 13, 2013; Accepted August 22, 2013; Published October 18, 2013 Copyright: ß 2013 Neymotin et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: Research supported by National Institutes of Health grant R01MH086638 (http://nih.gov/). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: The authors have no conflicts of interest to disclose. The authors would also like to state that WWL is a PLOS ONE Editorial Board member. This does not alter the authors’ adherence to all the PLOS ONE policies on sharing data and materials. * E-mail: [email protected] ¤ Current address: Medtronic Neuromodulation, Minneapolis, Minnesota, United States of America

Introduction In addition to its contribution to cell resonance, the HCN channel has a number of properties that suggest Ih might play a The hyperpolarization-activated cyclic-nucleotide gated (HCN) major role in control of oscillations in hippocampus and other channel is a voltage-gated ion channel involved in sub-threshold brain areas: 1. It is one determinant of a critical cell-excitability resonance [1–4]. Additionally, HCN plays an important role in control, RMP [5,8]. 2. It is differentially expressed in different cell regulating neuronal excitability by setting resting membrane types by virtue of inhomogeneous isoform distributions [1,3,7,9]. potential (RMP) [5,6]. HCN produces the current known as Ih 3. It is differentially modulated in different cell types by virtue of (h for hyperpolarization-activated), also known as If (f for funny), targeting of particular excitatory or inhibitory cell types by Iq (q for queer), and as ‘‘the anomalous rectifier’’. Ih is peculiar/ particular neurotransmitters and neuromodulators projecting from funny/queer/anomalous because, unlike most channels, it inacti- different brain areas [2,10213]. Because it is modulated through vates with depolarization (hyperpolarization-activated). Another second messengers, these neurotransmitters and neuromodulators peculiarity is its mixed permeability, which gives it an intermediate will be expected to have complex interactions within the cell 2 reversal potential (Erev) near 30 mV, unlike many channels chemistry prior to interacting with the membrane properties via Ih which are dominated by a major permeability to Naz,Kz,or zz [14]. Ca . Hippocampus contains many classes of pyramidal and inhibi- HCN channels are modulated by cyclic nucleotide second tory cells, with differing contributions to network dynamics messengers. HCN has four isoforms which are differentially [15,16]. We hypothesized that differential modulation of Ih expressed in different cell types and differ in intrinsic properties, currents in different cell classes would fine-tune the power and kinetics, and pharmacological sensitivities [1,7]. HCN1 and frequencies of network-generated oscillations. We therefore HCN2 isoforms are the dominant forms in hippocampus, and investigated the effects of altering Ih conductance [14,17] in a are present in varying proportions in all cell types studied. Of the computer model of hippocampal CA3, consisting of 800 pyrami- two, HCN1 is faster (shorter time-constant). dal cells, 200 basket interneurons, and 200 oriens-lacunosum

PLOS ONE | www.plosone.org 1 October 2013 | Volume 8 | Issue 10 | e76285 Ih Tunes Oscillations in an In Silico CA3 Model moleculare cells [18], using different isoform combinations based terization [3,7,9]. The OLM cells had a simple calcium-activated on the literature [4,7,9]. We found that tuning Ih in different cell potassium current IKCa to allow long lasting inactivation after classes altered network rhythms, providing independent control for bursting, high-threshold calcium current IL to activate IKCa, gamma and theta oscillations. Ih modulation also set the level of hyperpolarization-activated current Ih, and intracellular calcium cross-frequency coupling and timing of gamma generation relative concentration dynamics. Selection of currents was based on prior to the theta cycle. Ih modulation may therefore be an important published models [25,28–30] and basket interneuron Ih currents control point with functional consequences, since these dynamics were based on the literature [3,7,9]. : are hypothesized to contribute to learning and cognitive function For all cell types the Ih current was defined as Ih~g (v{eh), [19–21]. where g is the instantaneous conductance, v is the membrane potential, and eh is the reversal potential (230 mV for BAS and Materials and Methods PYR cells; 240 mV for OLM cells). Each Ih channel had a parameter, g, which represented the maximal conductance density Simulations (0.0002 S/cm2 for BAS, 0.0001 S/cm2 for PYR, and 0.00015 This model is an extension of a model of hippocampal CA3 that cm2 was previously published [18]. Simulations were performed on a S/ for OLM cells). To simulate neuromodulatory scaling of I g f Linux system with eight 2.27 GHz quad-core Intel Xeon CPUs the h conductance values, was multiplied by another factor, , using NEURON [22]. Eight seconds of simulation ran in about which varied between 0.0 and 2.0, and was set to 1.0 for the baseline simulations. Instantaneous conductance was then set to 2.2 minutes. In order to assess the robustness of the results, we ran : : each simulation condition with six different randomizations of g h f , where h is the Ih gating variable which activated at synaptic inputs, and six different randomizations of network hyperpolarized voltages. The evolution of the h state variable in _ connectivity. Simulations were run in the NEURON simulation time followed h~(h?{h)=th, where h? was the voltage- environment with python interpreter, multithreaded over 16–32 dependent steady-state value of h, and th was the voltage- threads [22,23]. Analysis of simulation data was done with the dependent time-constant of h (in milliseconds). 0:151:(v{v ) Neural Query System [24] and Matlab (Mathworks, Inc.). The full For BAS cells, h? was set to 1=(1ze 50 ), where v was model is available on ModelDB (https://senselab.med.yale.edu/ 1 the membrance voltage, and v , the -maximal voltage level, modeldb). 50 2 0:033:(vz75) was set to 273 mV. BAS cell th followed e = : : 0:083 (vz75) (v{v50)=10:5 Cells and connections (0:011 (1ze )). PYR h? followed 1=(1ze ), {14:59{0:086:v The network consisted of 800 five-compartment pyramidal with v50 at 282 mV. PYR th was set to 1=(e z (PYR) cells, 200 one-compartment basket (BAS) interneurons, and {1:87z0:0701:v ((vz80)=10) e ). OLM h? followed 1=(e z1), and OLM 200 one-compartment oriens lacunosum-moleculare (OLM) t followed 200=(e(vz70)=20ze{(vz70)=20)z5. interneurons [25–27] (Fig. 1). Current injections (pyramidal cell h I -static. To test the effect that I had on individual neurons, s: 50 pA; OLM cells 225 pA) were added to get baseline activity. h h we isolated the dynamic component, which had the voltage- This was a simplification to substitute for absence of external dependent conductance (g) described above. To do this, we first inputs from other areas, and to compensate for the small size of the ran a set of 7 second simulations, varying the f parameter from 0.0 model, which did not allow for much self-activation. to 2.0 (with increments of 0.5) and measured the I conductance All cells contained leak current, transient sodium current I , h Na (g) at the end of each simulation. This conductance (g) was saved and delayed rectifier current I { , to allow for action potential K DR for each compartment of each cell type. I -static was then defined generation. Additionally, pyramidal cells contained in all com- h as the current from a leak channel with conductance equal to g partments potassium type A current IK{A for rapid inactivation, and hyperpolarization-activated current I based on HCN2 measured in the previous step, and with the same reversal h e I I isoform parameterization [3,7]. Interneurons contained hyperpo- potential ( ) as the original h channel. h-static followed i~g:(v{e). larization-activated Ih current based on HCN1 isoform parame- The network contained 152,000 synapses. Pyramidal cell projections were mixed alpha-amino-3-hydroxy-5-methyl-4-isoxa- zolepropionic acid (AMPA) and N-methyl-D-aspartic acid (NMDA) response. Basket cells synapsed on the soma of both pyramidal cells and other basket cells via gamma-aminobutyric acid A (GABAA) receptors. OLM cells connected to distal dendrites of pyramidal cells via GABAA receptors. AMPA and NMDA receptors had reversal potentials of 0 mV, while GABAA receptors had reversal potentials of 280 mV. Connections in the network were set up based on fixed convergences (Table 1). However, connectivity was random and specific divergence could therefore vary. All synaptic delays between cells were 2 ms, to simulate axonal propagation and neurotransmitter diffusion and binding, which were not explicitly modeled. Parameters were based on the literature where available, Figure 1. Schematic representation of the network. Each symbol represents a population: 800 pyramidal cells (P), 200 basket cells (B), 200 as well as on previous models [25,31]. OLM cells. Convergence values (number of inputs for an individual synapse) are shown near synapses: GABAA receptors (filled circles), Synapses AMPA receptors (open circles), NMDA receptors (open squares). Synapses were modeled by a standard NEURON double- External stimulation from other areas was modeled by synaptic bombardment (synapses with truncated lines). exponential mechanism with parameters based on Tort et al., 2007 doi:10.1371/journal.pone.0076285.g001 [25] (Table 1). Magnesium block in NMDA receptors used the

PLOS ONE | www.plosone.org 2 October 2013 | Volume 8 | Issue 10 | e76285 Ih Tunes Oscillations in an In Silico CA3 Model

Table 1. Synaptic parameters. removed [34]. In addition, the first and last 200 ms of simulated data were removed to avoid artifacts associated with endpoints in the data. The spectral power was calculated using the multitaper method (MatLab pmtm() function; Mathworks, Inc.). Peak values Presy Postsy t1 naptic naptic Receptor (ms) in the power spectra are reported for theta (4{12 Hz) and low gamma (25{55 Hz) frequency bands. All r-values reported were t2 Conductance Conver (ms) (nS) gence calculated using the Pearson correlation coefficient. To determine cross-frequency-coupling (CFC) between theta and gamma Pyramidal Pyramidal AMPA 0.05 5.3 0.02 25 oscillations, we used a modified version of the modulation index Pyramidal Pyramidal NMDA 15 150 0.004 25 [35] to reduce artifacts in CFC measures associated with sharp Pyramidal Basket AMPA 0.05 5.3 0.36 100 spikes [36]. Theta oscillations were extracted by filtering LFPs Pyramidal Basket NMDA 15 150 1.38 100 between 6–10 Hz using a zero phase distortion band-pass filter. Pyramidal OLM AMPA 0.05 5.3 0.36 10 Gamma spikes (duty cycle between 18–40 ms, corresponding to 55–25 Hz) were extracted using a time-domain feature-extraction Pyramidal OLM NMDA 15 150 0.7 10 method [37]. Theta phases at times of gamma spike peaks were Basket Pyramidal GABAA 0.07 9.1 0.72 50 then used to form the gamma-amplitude/theta-phase measure, Basket Basket GABAA 0.07 9.1 4.5 60 which consisted of 100 equally-spaced phase bins, and were then

OLM Pyramidal GABAA 0.2 20 72 20 used to calculate the modulation index [35]. Final evaluations to produce the results presented here were doi:10.1371/journal.pone.0076285.t001 made over the course of 1044 network simulations, using six different random wirings, six different input streams, and : : {0:062:V experimental scaling factor 1=(1z0:28 Mg e ); Mg~ variations in maximal Ih conductance level (relative to baseline: 1mM [32]. 0.0, 0.5, 1.0, 1.5, 2.0) at the different cell types, where baseline is the Ih density estimated from the literature. A typical network Background activity simulation (8 s; 1200 neurons) took approximately 2.2 minutes Throughout the simulation duration, background activity was using 16 threads on a 2.27 GHz Intel Xeon quad core CPU. simulated by synaptic excitatory and inhibitory inputs following a A long-duration simulation set (900 seconds for each simulation) Poisson process, sent to somata of all cells and dendrites of was run using 5 Ih levels for the pyramidal and basket cells. These pyramidal cell s (Table 2). Fast background activity consisted of simulations all had identical wiring and input streams. The data AMPA and GABA-ergic bombardment at 1000 Hz. Slow activity obtained were used to evaluate theta/gamma cross-frequency- used activation of the NMDA receptors at a mean frequency of coupling and phase relationships as a function of Ih level. 10 Hz. These inputs represented the influence of surrounding An additional set of simulations of isolated cells was run, varying excitatory and inhibitory cells not explicitly modeled in the Ih conductance level in the same amounts as in the network. These simulation and produced a high conductance state similar to that simulations were used to assess Ih effects on resting membrane observed in vivo [33]. In addition, we placed slow excitatory potential (RMP) and synaptic integration. These simulations were inputs in the last distal apical compartment of pyramidal cells, in run for 7 s to allow the cells to reach a steady-state where net order to model input from the entorhinal cortex. This input was transmembrane currents were zero. Then, Ih conductance was capable of simulating calcium-spike-like activity in the dendritic measured and was used to set a fixed conductance with equivalent compartment and driving sparse firing of pyramidal cells. Erev to Ih, to separate Ih dynamics from its static features. In these Synapses were activated randomly according to a Poisson simulations, AMPA and GABAA inputs (0.5 nS) were provided at distribution. 5.5 s to assess post-synaptic-potential amplitude and temporal Local field potential (LFP) was simulated by a sum of differences integration. in membrane potential between the most distal apical and the basal dendritic compartment over all pyramidal cells. Before Results calculating spectral power, the DC component of the signal was This study involved over 1000 eight-second network simulations, testing six different input streams, and variations in maximal Table 2. Parameters for modeling background activity. Ih conductance level for the different cell types. These are presented as 0.0|Ih, 0.5|Ih, 1.0|Ih, 1.5|Ih, 2.0|Ih, relative to a baseline set to a standard Ih density estimated from the t1 t2 Conductance literature. In order to ensure robustness of the results shown, each Cell Section Synapse (ms) (ms) (nS) simulation was tested with six different wirings (wiring density is Pyramidal Soma AMPA 0.05 5.3 0.05 parameterized but specific point-to-point wiring is random). An

Pyramidal Soma GABAA 0.07 9.1 0.012 additional set of 25 long-term (900 second) simulations were run to Pyramidal Dend AMPA 0.05 5.3 0.05 evaluate theta/gamma cross-frequency-coupling and phase relationships as a function of I level. Simulations were run using the Pyramidal Dend NMDA 15 150 6.5 h NEURON simulator on Linux on a 2.27 GHz quad-core Intel Pyramidal Dend GABAA 0.07 9.1 0.012 XEON CPU. Eight seconds of network simulation ran in Basket Soma AMPA 0.05 5.3 0.02 *2.2 minutes.

Basket Soma GABAA 0.07 9.1 0.2 Ih is a prominent part of resting conductance, contributing to OLM Soma AMPA 0.05 5.3 0.0625 resting membrane potential (RMP), due to the presence of non- zero Ih conductance at RMP, and to a relatively depolarized OLM Soma GABAA 0.07 9.1 0.2 reversal potential (Erev). The isolated model oriens-lacunosum doi:10.1371/journal.pone.0076285.t002 moleculare (OLM) cell was depolarized with increasing Ih, from

PLOS ONE | www.plosone.org 3 October 2013 | Volume 8 | Issue 10 | e76285 Ih Tunes Oscillations in an In Silico CA3 Model

268.1 mV without Ih,to264.3 mV at 0.5|,to261.8 mV at frequency (4{12 Hz). Interactions between cells in the network 1|Ih. Increasing Ih past baseline produced further depolarization led to the generation of theta and gamma (w25 Hz) oscillations and cell firing. At 1.5|Ih, the OLM produced a single action (Fig. 3). These emergent rhythms were generated through the potential and then stabilized with an RMP of 259.5 mV. Further different synaptic time constants in the network and through the increase to 2| Ih produced rhythmic firing at 6 Hz, a low theta cellular interactions of pyramidal-interneuron network gamma frequency. Pyramidal (PYR) and basket (BAS) cells displayed (PING) and interneuron network gamma (ING) [15,31,38,39]. monotonic RMP dependence on Ih, with RMP ranging from Baseline oscillations were similar to those described in earlier 265.6 – –57.5 mV and 265 – 261.7 mV, respectively. PYR cells versions of this model, which contained Ih currents in PYR but not emitted one and two transient spikes at 1.5 and 2|Ih, respectively, in BAS cells [18]. Briefly, strong periodic OLM firing shut down while BAS cells did not exhibit any spontaneous firing. PYR activity resulting in lower PYR ? BAS drive. PING Altering Ih altered both the magnitude and time-course of interactions between PYR and BAS cells contributed to gamma excitatory and inhibitory postsynaptic potentials (EPSPs and oscillations: lower PYR to BAS drive led to lower gamma IPSPs). Both types of PSP showed increasing amplitude with amplitude during periods of OLM ? PYR inhibition. As the PYR increasing Ih. IPSP amplitude increase can be directly explained as cells recovered from OLM inhibition, their activity gradually built a consequence of the greater driving force at the more depolarized up providing increased drive to BAS cells and increasing gamma RMP. With 0{2|Ih, BAS and PYR increases were from 0.30– amplitudes, accounting for nesting of gamma within the theta 0.44 mV and 0.49–1.1 mV, respectively, while OLM increased cycle (Fig. 3a,b). ING also contributed to the strength of gamma in from 0.22–0.65 mV, with 0{1:5|Ih (at 2| Ih, OLM fired this model due to strong BAS ? BAS connectivity. The presence rhythmically, precluding accurate IPSP measurement). of Ih led to a slightly higher gamma amplitude than in the prior EPSP amplitude was also generally augmented with Ih increase model due to the stronger repolarization enhancing the ING (Fig. 2a). This is a paradoxical effect, given that the direct RMP mechanism. Individual cell voltages showed multiple rhythms as depolarizing shift that augmented IPSP driving force decreased well, with both the PYR and BAS cells reflecting the network EPSP driving force. In addition to reducing driving force, oscillation in their postsynaptic potentials (Fig. 3c). increased Ih also increased shunting, an effect that would reduce Given the complex of RMP shifts and temporal integration amplitude of both EPSPs and IPSPs. Both of these static factors properties through PSP alterations in the individual cells, we predict EPSP amplitude decrease. We therefore predicted that hypothesized that Ih changes would substantially alter frequency replacement of the dynamical Ih with a static version (fixed and power in network rhythms. Testing Ih modulation across conductances of equivalent magnitudes and Erev; see Materials different cell types within the full network demonstrated consistent and Methods) would reduce EPSP amplitude. Instead, we found but dramatically different effects depending on which cell type was even larger increases in EPSP magnitude. Examination of targeted. We started by looking at OLM cells because they provide transmembrane current activations of both Naz and Kz currents, a central modulating role for theta activity (Fig. 4, Fig. 5) [18]. revealed a larger depolarizing effect of INa (Fig. 2b), which Reducing or eliminating Ih from OLM cells abolished theta by dominated over the hyperpolarizing effect of IK (Fig. 2c). With eliminating the depolarizing influence of Ih. The resulting block of Naz and Kz channels, EPSP amplitudes decreased with hyperpolarization reduced OLM firing rate (1.2{0.2 Hz) which depolarized RMP, as originally predicted. The dynamics of Ih reduced theta modulation throughout the network (red and black itself worked to reduce this amplitude increase: Ih turns off during in Fig. 4a,b; Fig. 5a,b). The reduced inhibition coming from OLM the EPSP, reducing the degree of depolarization and reducing the cells resulted in higher firing rates of PYR cells (1.8{3.5 Hz), INa boost (Fig. 2d). The combination of these 5 effects (driving which then strengthened BAS activity (10.8{28.7 Hz). The force,shunting, INa, IK, Ih dynamics) produced a mild overall increased dominance of PYR and BAS populations produced a EPSP amplitude increase, that was far less pronounced than the large increase in gamma power (inset in Fig. 4b right) created via increase in IPSP: BAS: 0.99–1.17 mV with 0{2|Ih; PYR 1.78– the PING mechanism. Increasing OLM Ih conductance from 2.21 mV with 0{1:5|Ih,2|Ih produced spiking; OLM 0.96– baseline increased OLM firing rate (1.2{2.8 Hz) and caused the 1.07 mV, with 0{1|Ih, 1.5|Ih produced spiking. In one case, a OLM inhibition of the network to dominate, gradually reducing slight decrease in EPSP amplitude was seen: 1.78 to 1.72 mV with both theta and gamma power as PYR and BAS rates went towards increase of Ih from 0 to 0.5| baseline in the PYR cell. zero (PYR:1.8{0.4 Hz; BAS:10.8{2.0 Hz; Fig. 5e). Time to peak PSP was delayed by increasing Ih. These effects Increasing Ih conductance across all cellular locations produced were again a result of multiple conflicting tendencies. We therefore effects primarily similar to the effects on OLM, with reduced theta looked separately at the effects of the conductance change, effects power and augmented gamma at reduced Ih amplitudes. These of other channels, and effects of Ih dynamics themselves. The effects were brought about via the strong governing inhibitory conductance change alone lowered Rin which reduced membrane influence of OLM cells, which increased at heightened Ih levels. As time-constant, which reduced the duration of synaptic response, with OLM Ih enhancement, higher Ih values showed decrease in leading to an earlier peak. Returning Naz and Kz currents to the gamma power and frequency with increase in theta. simulation moved PSP peaks to slightly later times. Adding back BAS cells are particularly involved in both ING (BAS-BAS) and the dynamics of Ih moved the PSPs to earlier times again. With all PING (PYR-BAS) mechanisms of gamma generation [38,39]. these dynamical factors in place, IPSP delays had noticeably Hence, it was not surprising that variation of BAS Ih altered increasing values: BAS 10.8–12.5 ms with 0{2|Ih; PYR 6.8– gamma power and frequency consistently with no consistent effect 9.5 ms with 0{2|Ih; OLM 10.0–16.5 ms with 0{1:5|Ih since on theta (Fig. 6, Fig. 7). Increased BAS Ih augmented gamma 2| Ih produced rhythmic spiking. Similar effects were observed power (r~0:88; pv1e{9; n~180) and reduced gamma frequen- for EPSP delays (BAS: 8.6–10.7 ms with 0{2|Ih; PYR: 4.9– cy (r~{0:56; pv1e{9; n~180). The increased power corre- 8.1 ms with 0{1:5|Ih since at 2|Ih the synaptic input produced sponded to increase in the BAS population firing rates a spike; OLM: 7.6–9.5 ms with 0{1:5|Ih). (9.1{12.6 Hz with Ih 0{2|) due to the depolarizing effect of In the network, baseline firing rates of PYR, BAS, and OLM Ih. These increases in BAS firing also dampened PYR firing cells were 1.8 Hz, 10.8 Hz, and 1.2 Hz, respectively. As a (1.8{1.7 Hz), which secondarily reduced OLM activity population, OLM cells tended to fire rhythmically at theta (1.3{1.2 Hz). The decreased gamma frequency was due to the

PLOS ONE | www.plosone.org 4 October 2013 | Volume 8 | Issue 10 | e76285 Ih Tunes Oscillations in an In Silico CA3 Model

Figure 2. BAS cell response to AMPA stimulus at different levels of Ih conductance. Solid lines represent responses with dynamic Ih and dotted lines represent responses with static Ih. Note that only BAS cell is displayed since it did not fire action potentials in response to AMPA-ergic stimulation. Time axes are relative to AMPA input at t~0 ms. (a) EPSP (starting voltage levels aligned vertically for easier comparison of EPSPs), (b) INa,(c) IK , and (d) Ih at BAS cell soma. doi:10.1371/journal.pone.0076285.g002 longer synaptic integration times that the BAS cells displayed with BAS Ih levels (r~0:93; pv1e{9; n~180) due to direct enhance- enhanced Ih. ment to BAS population activity via Ih (8.0{14.1 Hz) and also By contrast with BAS Ih effects, PYR Ih effect was primarily on secondarily due to PING mechanisms. Gamma peak frequency theta, progressively increasing theta peak (r~0:89; pv had a clear trend of reduction with increases in PYR and BAS Ih 1e{9; n~180) and power (r~0:73; pv1e{9; n~180; Fig. 8, (r~{0:69; pv1e{9; n~180), due to the extended delays to peak Fig. 9; consistent with experiment [40]). Increases in theta peak IPSPs and EPSPs that PYR and BAS cells exhibited with and power were effected through increased PYR firing increasing Ih. (1.6{1.9 Hz) which produced increased OLM firing HCN1 and HCN2 have different molecular modulators: cAMP (0.9{1.6 Hz). Unlike in Fig. 4, OLM firing did not suppress selectively modulates HCN2, [41,42] while p38 MAP kinase PYR firing since PYR activity was the driving force and was modulates HCN1 [43]. However, the complexity of linkages from supported by the PYR Ih. Due to PING interplay, gamma neuromodulators to expression of second and third messengers, oscillation power was positively correlated with PYR Ih level and the consequent control in HCN isoforms by these messengers, (r~0:73; pv1e{9; n~180; BAS rates: 9.2{12.1 Hz). Although is currently inaccessible to simulation. We therefore assessed all gamma peak frequency was not significantly shifted, there was combinations of Ih modulation at PYR and BAS cells in order to some broadening with increasing PYR Ih. Overall PYR Ih observe the patterns of gamma-theta relations that could be modulation tuned both theta and gamma power together, distinct expressed through HCN modulation in this system. As expected from other pharmacological effects where theta and gamma are from the relative independence of gamma and theta control from inversely correlated [18]. the cell types, we found that these patterns were highly constrained The contrast of a nearly orthogonal arrangement of strong (Fig. 12). Both theta amplitude and frequency increased with PYR influence of PYR Ih on theta and strong influence of BAS Ih on Ih level with effectively no effect of BAS Ih levels (Fig. 12a,b). gamma led us to hypothesize that detailed control of network Although gamma frequency (Fig. 12c) and amplitude (Fig. 12d) oscillation could be effected through comodulation of Ih in both. showed primary control by BAS Ih as expected, there was also a This comodulation could involve simultaneous control where Ih in prominent effect of PYR Ih, producing the greatest overall gamma both cell types were altered together. Alternatively, more complex amplitude augmentation with coordinated increase in both BAS modulation could occur via activation through different second and PYR Ih. Hence the highest gamma amplitude and highest messengers, or different isoform second-messenger sensitivity, gamma frequency also showed correlation with the highest theta through activation by a neuromodulator with divergent down- amplitude and frequency. stream effects. Simultaneous Ih modulation of both PYR and BAS Cross-frequency-coupling (CFC) measures the ability of the cells produced an additive effect, with changes in both theta and slower theta wave to provide an envelope that modulates the gamma rhythms (Fig. 10, Fig. 11). There was a clear trend of amplitude of the superimposed faster gamma. Since the strong progressively increasing theta peak (r~0:91; pv1e{9; n~180) OLM inhibition only allowed co-expression of theta and gamma and a similar trend for increasing theta power (r~0:72; pv oscillations in a relatively narrow range of OLM Ih, we only 1e{9; n~180). The changes in theta power were brought about measured CFC as a function of PYR and BAS Ih. Substantial by increased PYR firing (1.6{1.9 Hz) which drove increases in CFC was only present with high PYR Ih, corresponding to large OLM firing (0.9{1.6 Hz). Similar to the simulations where PYR theta (Fig. 12e). The difference between low and high CFC can Ih was modulated independently, OLM firing did not suppress be seen in Fig. 10a. The black trace demonstrates low CFC: at PYR firing due to Ih increases supporting PYR activity. Gamma left only a little alteration of gamma amplitude with theta is seen; oscillation power had a large positive correlation with PYR and at right there is almost no gamma hence no coupling. By

PLOS ONE | www.plosone.org 5 October 2013 | Volume 8 | Issue 10 | e76285 Ih Tunes Oscillations in an In Silico CA3 Model

Figure 3. Activity of network at baseline. (a) Raster plot showing firing times of cells within the network. Cell types are color-coded. (b) Local field potential (LFP) generated by PYR cells. (c) Voltage traces from soma of different cell types. (d) Average (n~36) local field potential power spectrum + standard error of the mean (SEM; dotted lines). doi:10.1371/journal.pone.0076285.g003 contrast the orange trace shows substantial coupling, most readily cell firing, and hence earlier production of gamma via PING. seen in the 4th theta cycle. Note that these cycle-to-cycle Reduced phase lag was therefore associated with stronger CFC differences make the overall CFC difficult to calculate. In this (r~{0:81; pv1e{5; n~25). high PYR Ih regime, coupling was highest at low values of BAS At baseline, PYR spiking tended to occur near the peak of theta p Ih, where average gamma activity, reflecting this modulation (+ 16 radians), earlier than the theta phase for maximum gamma. from low to high, was low (Fig. 12d). By contrast high BAS Ih This delay from peak PYR firing to peak local field gamma is corresponded to a strong continuous gamma which was not as consistent with a PING mechanism: peak PYR firing engages a readily modulated. Peak coupling corresponded to oscillations larger number of inhibitory cells. This then leads to a subsequent with gamma frequency of 33.5 Hz and theta frequency of peak gamma cycle, representing the maximum proximal/distal 8.6 Hz. synaptic-activation differences, which then occurs on the subse- Across Ih levels, the peak gamma amplitude always occurred quent cycle. during the positive portion of the theta cycle (Fig. 12f), slightly p p after the theta peak from 6 to 4 radians (*0.5–0.8, where 0 is Discussion theta peak). This is consistent with experimental data, which shows peak amplitude of gamma occurring on the positive but Our modeling predicts that neuromodulation of Ih conductance descending portion of the theta oscillation [44]. Increased PYR could have several functional roles in in vivo neuronal dynamics Ih shifted peak gamma amplitude towards earlier phases of the including: 1) tuning of theta and gamma oscillation amplitude and theta cycle. This was due to the depolarizing effects of PYR Ih frequency, 2) modulation of cross-frequency coupling (CFC) levels, producing heightened PYR excitability, leading to earlier PYR and 3) enhanced excitability of cells within a circuit, expressed as

PLOS ONE | www.plosone.org 6 October 2013 | Volume 8 | Issue 10 | e76285 Ih Tunes Oscillations in an In Silico CA3 Model

Figure 6. Activity (from 180 simulations) with Ih scaling in Figure 4. Activity (from 180 simulations) with Ih scaling in OLM interneurons. (a) Local field potentials (LFPs). Blue LFP is from basket (BAS) interneurons. (a) Local field potentials (LFPs). Blue LFP baseline simulation. Up (down) arrows indicate directions of increase is from baseline simulation. Up (down) arrows indicate directions of increase (decrease) of Ih.(b) Scatter plots of theta and gamma peak (decrease) of Ih.(b) Scatter plots of theta (left) and gamma (right) peak frequencies and power (normalized); color code as in (a); each point frequencies and power (normalized). from a single simulation with different random activation and wiring. doi:10.1371/journal.pone.0076285.g006 Gamma: main panel shows zoom-in of subset of values. Inset shows full set. doi:10.1371/journal.pone.0076285.g004 in different cell types known to contribute to different oscillation frequencies, and 4) neuromodulators allow precise control of the conductance of specific HCN isoforms via second-messenger increased gamma oscillation amplitude at earlier phases of the signaling cascades [7,43]. These functions of theta and gamma theta cycle. I is uniquely positioned for these roles for several h oscillations are linked to different aspects of cognition and reasons: 1) I enhances resonance in individual neurons, 2) I h h behavior: CFC level is correlated with hippocampal-dependent contributes to resting membrane potential, and hence neuronal learning performance [21,45] and attentional modulation [46], excitability, 3) multiple HCN isoforms are differentially expressed

Figure 5. Activity from a single network after modulating OLM Ih levels (OLM Ih increases left to right). Top shows spike rasters (PYR:red; BAS:green; OLM:blue). Bottom displays somatic voltage from a single OLM cell (blue) and average somatic voltage from 200 OLM cells (black). doi:10.1371/journal.pone.0076285.g005

PLOS ONE | www.plosone.org 7 October 2013 | Volume 8 | Issue 10 | e76285 Ih Tunes Oscillations in an In Silico CA3 Model

Figure 7. Activity from a single network after modulating BAS Ih levels (BAS Ih increases left to right). Top shows spike rasters (PYR:red; BAS:green; OLM:blue). Bottom displays somatic voltage from a single BAS cell (green) and average somatic voltage from 200 BAS cells (black). doi:10.1371/journal.pone.0076285.g007 and gamma nesting within theta oscillations is a hypothesized on where in the circuit the modulation occurred. As we have mechanism for encoding information dynamically [20]. previously shown, OLM provides control over theta activation in We investigated Ih channel function in a multiscale model the network due to its long time constants [18]. Reduced OLM Ih across levels from ion channel population to the neuronal network. eliminated theta by removing this influence (Fig. 4, Fig. 5). This Emergent predictions arose at the levels of channel interactions in then allowed the PYR and BAS interactions to create strong, dendrites, of dendritic signal interactions in cells and of neurons continuous gamma through ING and PING mechanisms. forming the network. At the dendritic and cellular level, Ih Increased OLM Ih eliminated all activity by causing increased generally increased both EPSP and IPSP magnitude and duration OLM activity which shut down activity in the other cells, OLM with some variation by cell type. At the cell level, excitability being an inhibitory cell type. Modulating Ih across all cell types increased due to cell depolarization. At the network level, Ih had effects similar to those seen with OLM modulation, due to this modulation altered both theta and gamma, with effects depending strong governing influence of OLM. Different neurotransmitters are likely to have differential effects on different cell types through effects on different receptors on the different cell types. Our modeling suggests likely cellular locations of neuromodulation targets for changing oscillation power and frequency. These could be tested by using immunohistochemistry to correlate the location of neurotransmitter receptor types with particular cell types. For example, it is known that noradrenaline is involved in Ih regulation [47]. In addition, recent experimental evidence demonstrates that acetylcholine modulates different features of Ih activity, including its sag amplitude [11,12]. Interestingly, acetylcholine has also been shown to contribute to modulation of theta frequency over a range similar to that observed in our model [48]. The BAS cell is particularly involved in the genesis of gamma oscillations through the ING (BAS-BAS) and PING (PYR-BAS) mechanisms. Increased BAS cell Ih increased BAS activity and raised gamma power (Fig. 6, Fig. 7). This increase also slightly lowered gamma frequency, due to the increased duration of synaptic responses. The PYR cell is the only excitatory cell in the network and therefore plays a role in maintaining firing of all cell types. Increased PYR Ih increased PYR ? OLM activation and produced a monotonically increasing effect on both power and frequency of theta (Fig. 8, Fig. 9). Note that this apparent PYR ?

Figure 8. Activity (from 180 simulations) with Ih scaling in OLM effect was quite different than the more direct activation pyramidal (PYR) cells. (a) Local field potentials (LFPs). Blue LFP is provided by increasing OLM Ih. At the same time, the increased from baseline simulation. Up (down) arrows indicate directions of PYR ? BAS activation produced a tendency to increased gamma increase (decrease) of Ih.(b) Scatter plots of theta and gamma peak frequencies and power (normalized). power without consistent effect on frequency. The overall PYR doi:10.1371/journal.pone.0076285.g008 effect was to tune both theta and gamma power together, distinct

PLOS ONE | www.plosone.org 8 October 2013 | Volume 8 | Issue 10 | e76285 Ih Tunes Oscillations in an In Silico CA3 Model

Figure 9. Activity from a single network after modulating PYR Ih levels (PYR Ih increases left to right). Top shows spike rasters (PYR:red; BAS:green; OLM:blue). Bottom displays somatic voltage from a single PYR cell (red) and average somatic voltage from 800 PYR cells (black). doi:10.1371/journal.pone.0076285.g009 from other pharmacological effects where theta and gamma trade modulations of theta and gamma oscillations could be utilized by off [18]. functional mechanisms that are postulated to utilize linkages Simultaneous modulation of PYR and BAS Ih similarly between theta and gamma to provide encodings such as phase comodulated power, while now shifting both frequencies consis- precession in place cells [49], cross-frequency coupling (CFC) tently: gamma tuning towards lower frequency while theta tuned [35,50–52], and gamma on theta phase for memory [20]. Indeed, towards higher frequency with increased Ih (Fig. 10, Fig. 11). our model demonstrated that shifting oscillatory modulations were Independent modulation of PYR and BAS Ih allowed flexible effective in setting the CFC level, with increases evident at high control of the frequencies and amplitudes of theta and theta PYR Ih levels (Fig. 12e). We therefore predict the presence of gamma oscillations (Fig. 12a,b,c,d). We hypothesized that these distinct neurotransmitter receptor types in PYR and BAS cells which would allow Ih to be tuned independently, and therefore support flexible shifting of the CFC level. Our model demonstrated that increased PYR Ih would increase PYR excitability, augment PYR ? BAS feedforward activation via a PING mechanism, and thereby shift gamma activation to an earlier phase within the theta cycle. In the context of neural coding, the timing of pyramidal cell firing within a theta cycle has been hypothesized to allow the most relevant neurons for a particular stimulus to fire at earlier phases and then inhibit firing of other ensembles [53]. Our model suggests how modulation of Ih could enhance this contrast sensitivity by enhancing this initial activation. This is also consistent with recent experimental work that demonstrates the contribution of Ih currents to hippocampal pyramidal neuron synchronization [54], which could cause downstream neurons to fire earlier, thereby modulating timing of gamma spikes. Intracellular signalling can be used to modulate the degree to which Ih is regulated. This has been demonstrated experimentally in the heart [14,55], and similar mechanisms may take place in neurons via neuromodulatory control [11]. This mechanism has been demonstrated in computer models of prefrontal cortex neurons [17]. In this process, the neuron is initially activated via feedforward excitatory inputs. With sufficiently strong activation, Figure 10. Activity (from 180 simulations) with Ih scaling in calcium is admitted. Subsequently, calcium binds to protein both pyramidal (PYR) and basket (BAS) cells. (a) Local field kinases (e.g., cAMP) which bind to HCN and increase Ih potentials (LFPs). Blue LFP is from baseline simulation. Up (down) conductance, leading to increased excitability. Our model shows arrows indicate directions of increase (decrease) of Ih.(b) Scatter plots of theta and gamma peak frequencies and power (normalized). that in the neuronal network context, this process leads to doi:10.1371/journal.pone.0076285.g010 frequency tuning, increased CFC, and earlier generation of

PLOS ONE | www.plosone.org 9 October 2013 | Volume 8 | Issue 10 | e76285 Ih Tunes Oscillations in an In Silico CA3 Model

Figure 11. Activity from a single network after modulating PYR and BAS Ih levels (PYR and BAS Ih increases left to right). Top shows spike rasters (PYR:red; BAS:green; OLM:blue). Bottom displays average somatic voltage from PYR (red; n~800) and BAS (green; n~200) cells. doi:10.1371/journal.pone.0076285.g011

Figure 12. Amplitudes and coupling of oscillations with variation of Ih density in BAS and PYR cells (x- and y-axes, respectively). (a) Theta frequency and (b) amplitude are controlled by PYR Ih, while (c) Gamma frequency and (d) amplitude are largely controlled by BAS Ih.(e) Cross- frequency coupling (gamma amplitude modulation by theta phase) is greatest when theta is strong (high PYR Ih) with gamma relatively weak. Units p p are scaled up by 1e3 for readability. (f) Gamma amplitude peaks in the region between 6 (0.5) and 4 (0.8) radians in a complex pattern. (a,b,c,d: average of 900 8s simulations; e,f: average of 25 900 s simulations). doi:10.1371/journal.pone.0076285.g012

PLOS ONE | www.plosone.org 10 October 2013 | Volume 8 | Issue 10 | e76285 Ih Tunes Oscillations in an In Silico CA3 Model gamma spikes by the activated cells. Due to long time constants of Acknowledgments protein kinase binding with HCN, the effects of this initial activation could be used to prime a circuit’s response to The authors would like to thank Antonio Carlos Roque da Silva Filho Latin American School of Computational subsequent inputs. (University of Sao Paulo) for organizing Neuroscience IV (University of Sao Paulo, Ribeirao Preto, Brazil), where this Our current model remains limited by lack of explicit second research was begun; Herman Moreno (SUNY Downstate) for discussions; messenger modeling and lack of detailed information about Michael Hines and Ted Carnevale (Yale) for NEURON support; Tom differences between HCN isoforms. In particular, cAMP, the Morse (Yale) for ModelDB support; Larry Eberle and Amy Delman z second messenger which acts on Ih, also has effects on K [56] or (SUNY Downstate) for Neurosim lab support; the anonymous reviewers for leak [6,57] channels, which would also tend to change cell and their helpful comments. network dynamics. Our HCN isoform modeling also remains limited, since we only included electrophysiological, and not Author Contributions second messenger, differences. Inclusion of second messenger Conceived and designed the experiments: SAN WWL. Performed the signaling pathways will be of greatest value once further details are experiments: SAN MH TM YS. Analyzed the data: SAN MH TM YS available concerning differences in second messenger responsitivity MTL WWL. Contributed reagents/materials/analysis tools: MTL. Wrote between the two major isoforms studied here. Further detail might the paper: SAN WWL. also consider differences in phosphorylation states which provide further modulation of these channels [58].

References 1. Accili E, Proenza C, Baruscotti M, DiFrancesco D (2002) From funny current to 21. Tort A, Komorowski R, Manns J, Kopell N, Eichenbaum H (2009) Theta- HCN channels: 20 years of excitation. Physiology 17: 32–37. gamma coupling increases during the learning of item-context associations. Proc 2. Chen S, Wang J, Siegelbaum S (2001) Properties of hyperpolarization-activated Nat Acad Sci 106: 20942–20947. pacemaker current defined by coassembly of HCN1 and HCN2 subunits and 22. Carnevale N, Hines M (2006) The NEURON Book. New York: Cambridge basal modulation by cyclic nucleotide. Journal Gen Physiol 117: 491–504. University Press. 3. Santoro B, Baram T (2003) The multiple personalities of h-channels. Trends 23. Hines M, Davison A, Muller E (2009) NEURON and Python. Front Neurosci 26: 550–554. Neuroinform 3: 1. 4. Zemankovics R, Ka´li S, Paulsen O, Freund T, Ha´jos N (2010) Differences in 24. Lytton W (2006) Neural query system: data-mining from within the NEURON subthreshold resonance of hippocampal pyramidal cells and interneurons: the simulator. Neuroinformatics 4: 163–176. role of h-current and passive membrane characteristics. J Physiol 588: 2109– 25. Tort A, Rotstein H, Dugladze T, Gloveli T, Kopell N (2007) On the formation 2132. of gamma-coherent cell assemblies by oriens lacunosum-moleculare interneu- 5. Dyhrfjeld-Johnsen J, Morgan R, Soltesz I (2009) Double trouble? potential for rons in the hippocampus. Proc Natl Acad Sci U S A 104: 13490–13495. hyperexcitability following both channelopathic up-and downregulation of Ih in 26. Wang X, Buzsaki G (1996) Gamma oscillation by synaptic inhibition in a epilepsy. Front Neurosci 3: 25. hippocampal interneuronal network model. J Neurosci 16: 6402–6413. 6. Poolos N, Migliore M, Johnston D (2002) Pharmacological upregulation of h- 27. Wang X (2002) Pacemaker neurons for the theta rhythm and their channels reduces the excitability of pyramidal neuron dendrites. Nat Neurosci 5: synchronization in the septohippocampal reciprocal loop. J Neurophysiol 87: 767–774. 889–900. 7. Bender R, Brewster A, Santoro B, Ludwig A, Hofmann F, et al. (2001) 28. McCormick D, Huguenard J (1992) A model of the electrophysiological Differential and age-dependent expression of hyperpolarization-activated, cyclic properties of thalamocortical relay neurons. J Neurophysiol 68: 1384–1400. nucleotide-gated cation channel isoforms 1-4 suggests evolving roles in the 29. Migliore M, Messineo L, Ferrante M (2004) Dendritic Ih selectively blocks developing rat hippocampus. Neuroscience 106: 689–698. temporal summation of unsynchronized distal inputs in CA1 pyramidal neurons. 8. Dyhrfjeld-Johnsen J, Morgan R, Fo¨ldy C, Soltesz I (2008) Upregulated H- J Comput Neurosci 16: 5–13. Current in hyperexcitable CA1 dendrites after febrile seizures. Front Cell 30. Stacey W, Lazarewicz M, Litt B (2009) Synaptic noise and physiological Neurosci 2. coupling generate high-frequency oscillations in a hippocampal computational 9. Aponte Y, Lien C, Reisinger E, Jonas P (2006) Hyperpolarization-activated model. J Neurophysiol 102: 2342–2357. cation channels in fast-spiking interneurons of rat hippocampus. J Physiol 574: 31. White J, Banks M, Pearce R, Kopell N (2000) Networks of interneurons with fast 229–243. and slow c–aminobutyric acid type A (GABAA) kinetics provide substrate for 10. Carr D, Andrews G, Glen W, Lavin A (2007) a2-noradrenergic receptors mixed gamma-theta rhythm. Proc Nat Acad Sci 97: 8128–8133. activation enhances excitability and synaptic integration in rat prefrontal cortex 32. Jahr C, Stevens C (1990) Voltage dependence of NMDA-activated macroscopic pyramidal neurons via inhibition of hcn currents. J Physiol 584: 437–450. conductances predicted by single-channel kinetics. J Neurosci 10: 3178–3182. 11. Heys J, Hasselmo M (2012) Neuromodulation of Ih in layer II medial entorhinal 33. Destexhe A, Rudolph M, Pare´ D (2003) The high-conductance state of cortex stellate cells: a voltage-clamp study. J Neurosci 32: 9066–9072. neocortical neurons in vivo. Nat Rev Neurosci 4: 739–751. 12. Tsuno Y, Schultheiss N, Hasselmo M (2013) In vivo cholinergic modulation of 34. Oppenheim A, Schafer R, Buck J (1999) Discrete-time signal processing. the cellular properties of medial entorhinal cortex neurons. J Physiol 591: 2611– Prentice Hall, Upper Saddle River, N.J., 2nd ed. 2627. 35. Tort A, Kramer M, Thorn C, Gibson D, Kubota Y, et al. (2008) Dynamic cross- 13. Wang M, Ramos B, Paspalas C, Shu Y, Simen A, et al. (2007) a2a- frequency couplings of local field potential oscillations in rat striatum and adrenoceptors strengthen working memory networks by inhibiting cAMP- hippocampus during performance of a t-maze task. Proc Nat Acad Sci 105: HCN channel signaling in prefrontal cortex. Cell 129: 397–410. 20517–20522. 14. Hagiwara N, Irisawa H (1989) Modulation by intracellular ca2+ of the 36. Kramer M, Tort A, Kopell N (2008) Sharp edge artifacts and spurious coupling hyperpolarization-activated inward current in rabbit single sino-atrial node cells. in EEG frequency comodulation measures. J Neurosci Methods 170: 352–357. J Physiol 409: 121–141. 37. Neymotin S, Uhlrich D, Manning K, Lytton W (2008) Data mining of time- 15. Buzsa´ki G, Wang X (2012) Mechanisms of gamma oscillations. Annu Rev domain features from neural extracellular field data. Studies in Computational Neurosci 35: 203–225. Intelligence 151: 119–140. 16. Freund T, Buzsa´ki G (1996) Interneurons of the hippocampus. Hippocampus 6: 38. Bo¨rgers C, Kopell N (2003) Synchronization in networks of excitatory and 347–470. inhibitory neurons with sparse, random connectivity. Neural Comput 15: 509– 17. Winograd M, Destexhe A, Sanchez-Vives M (2008) Hyperpolarization-activated 538. graded persistent activity in the prefrontal cortex. Proc Natl Acad Sci USA 105: 39. Lytton W, Sejnowski T (1991) Simulations of cortical pyramidal neurons 7298–7303. synchronized by inhibitory interneurons. J Neurophysiol 66: 1059–1079. 18. Neymotin S, Lazarewicz M, Sherif M, Contreras D, Finkel L, et al. (2011) 40. Marcelin B, Chauvie`re L, Becker A, Migliore M, Esclapez M, et al. (2009) H Ketamine disrupts theta modulation of gamma in a computer model of channel-dependent deficit of theta oscillation resonance and phase shift in hippocampus. J Neurosci 31: 11733–11743. temporal lobe epilepsy. Neurobiol Dis 33: 436–447. 19. Lakatos P, Karmos G, Mehta A, Ulbert I, Schroeder C (2008) Entrainment of 41. Wahl-Schott C, Biel M (2009) HCN channels: structure, cellular regulation and neuronal oscillations as a mechanism of attentional selection. Science 320: 110– physiological function. Cel Mol Life Sci 66: 470–494. 113. 42. Zong X, Krause S, Chen C, Kru¨ger J, Gruner C, et al. (2012) Regulation of 20. Lisman J, Idiart M (1995) Storage of 7 6 2 short-term memories in oscillatory hyperpolarization-activated cyclic nucleotide-gated (HCN) channel activity by subcycles. Science 267: 10. cCMP. J Biol Chem 287: 26506–26512.

PLOS ONE | www.plosone.org 11 October 2013 | Volume 8 | Issue 10 | e76285 Ih Tunes Oscillations in an In Silico CA3 Model

43. Poolos NP, Bullis JB, Roth MK (2006) Modulation of h-channels in 51. Canolty R, Edwards E, Dalal S, Soltani M, Nagarajan S, et al. (2006) High hippocampal pyramidal neurons by p38 mitogen-activated protein kinase. gamma power is phase-locked to theta oscillations in human neocortex. Science J Neurosci 26: 7995–8003. 313: 1626–1628. 44. Belluscio M, Mizuseki K, Schmidt R, Kempter R, Buzsaki G (2012) Cross- 52. Lee H, Dvorak D, Kao H, Duffy A´ , Scharfman H, et al. (2012) Early cognitive frequency phase-phase coupling between theta and gamma oscillations in the experience prevents adult deficits in a neurodevelopmental schizophrenia model. hippocampus. J Neurosci 32: 423–435. Neuron 75: 714–724. 45. Canolty R, Knight R (2010) The functional role of cross-frequency coupling. 53. de Almeida L, Idiart M, Lisman J (2009) A second function of gamma frequency Trends Cogn Sci 14. oscillations: an E%-max winner-take-all mechanism selects which cells fire. 46. Schroeder C, Lakatos P (2009) The gamma oscillation: master or slave. Brain J Neurosci 29: 7497–7503. Topogr 22: 24–26. 54. Hilscher M, Lea˜o K, Lea˜o R (2013) Synchronization through nonreciprocal 47. Wang J, Chen S, Nolan M, Siegelbaum S (2002) Activity-dependent regulation connections in a hybrid hippocampus microcircuit. Front Neural Circuits 7: 120. of HCN pacemaker channels by cyclic AMP: signaling through dynamic 55. El Khoury N, Mathieu S, Marger L, Ross J, El Gebeily G, et al. (2013) allosteric coupling. Neuron 36: 451–461. Upregulation of the hyperpolarizationactivated current increases pacemaker activity of the sinoatrial node and heart rate during pregnancy in mice. 48. Newman E, Gillet S, Climer J, Hasselmo M (2012) Effects of cholinergic Circulation 127: 2009–2020. modulation on interactions of entorhinal cortex and hippocampus as measured 56. Arnsten A, Paspalas C, Gamo N, Yang Y, Wang M (2010) Dynamic Network by theta modulation of high and low gamma in the rat. Society for Neuroscience Connectivity: a new form of neuroplasticity. Trends Cogn Sci 14: 365–375. Abstracts 42. 57. Migliore M, Migliore R (2012) Know your current Ih: Interaction with a 49. O’Keefe J, Recce M (1993) Phase relationship between hippocampal place units shunting current explains the puzzling effects of its pharmacological or and the EEG theta rhythm. Hippocampus 3: 317–330. pathological modulations. PloS One 7: e36867. 50. Caixeta F, Corne´lio A, Scheffer-Teixeira R, Ribeiro S, Tort A (2013) Ketamine 58. Robinson R, Siegelbaum S (2003) Hyperpolarization-activated cation currents: alters oscillatory coupling in the hippocampus. Sci Rep 3: 2348. from molecules to physiological function. Annu Rev Physiol 65: 453–480.

PLOS ONE | www.plosone.org 12 October 2013 | Volume 8 | Issue 10 | e76285

APPENDIX B Study II

87 A dorsoventral gradient of hyperpolarization-activated current (Ih) in hippocampal CA1 OLM interneurons

Authors: Markus M. Hilscher1,2, Richardson N. Leão1,2, Sanja Mikulovic2, Klas Kullander2, Katarina E. Leão1,*

Author affiliations: 1. Brain Institute, Federal University of Rio Grande do Norte, Natal-RN, Brazil. 2. Unit of Developmental Genetics, Department of Neuroscience, Uppsala University, Uppsala, Sweden.

*Corresponding author Katarina E. Leao The Brain institute, Federal University of Rio Grande do Norte Av. Nascimento de Castro 2155, CEP: 59056-045, Natal-RN, Brazil Email: [email protected]

Key points

• Hyperpolarization-activated current (Ih) is present in neurons with pacemaker activity that participate in low frequency rhythm generation, such as theta oscillations. • There are strong indications that oriens lacunosum moleculare (OLM) cells contribute to hippocampal theta oscillations.

• We show a difference in Ih in dorsal and ventral OLM cells, opposite to the Ih gradient of pyramidal cells, as well as a difference in OLM cell membrane resonance behavior. • The different membrane properties of OLM cells across the dorsoventral axis of the hippocampus help to explain variation in theta oscillations across the hippocampus. • These results supports computational models of theta generation and are important for understanding how hippocampal theta oscillations correlate to different behaviors such as locomotion, sleep and emotional states.

Abstract

The hyperpolarization-activated current (Ih) is generally present in neurons with pacemaker activity that participate in low frequency rhythm generation, such as theta oscillations. Computational studies suggest that Ih in oriens-lacunosum moleculare (OLM) cells of the hippocampus is essential for theta generation in hippocampal circuits. However, the hippocampus is not a homogeneous structure and there are significant differences in membrane properties of given neuron types along its dorsoventral axis. Yet, it is not known if OLM neurons display differences in membrane properties across the dorsoventral hippocampal axis. Here, we investigate if CA1 OLM cell Ih differs between the dorsal and ventral hippocampus. We found that dorsal OLM cells show greater Ih than ventral counterparts. Application of the HCN channel blocker ZD7288 decreased the firing rate of dorsal OLM cells, while mainly leaving the firing rate of ventral OLM cells unaltered. Both dorsal and ventral OLM cells showed resonance in membrane potential. However, dorsal OLM cells showed higher membrane resonance frequency than ventral OLM cells. Our results demonstrate that dorsal and ventral OLM cells differentially express Ih and that they are electrophysiologically heterogeneous. Hence, dorsal and ventral OLM cells could differently participate in the generation of theta rhythms.

Keywords: OLM cells, dorsal ventral hippocampus, Chrna2, hyperpolarization-activated current, H- resonance Introduction Recently, PC membrane properties have been shown Theta oscillations (4-12 Hz) are thought to ’travel’ to differ along the dorsoventral axis (Marcelin et al., along the dorsoventral (or septotemporal) axis of the 2012a, 2012b; Dougherty et al., 2012; Malik et al., hippocampus (Lubenov & Siapas, 2009). The 2016; Milior et al., 2016). An ion current shaping travelling theta wave hypothesis would imply a similar active membrane properties of PCs, the cytoarchitecture between the dorsal and ventral hyperpolarizing activated current (Ih) has been shown hippocampus (dHPC and vHPC, respectively), as to be differentially expressed in PCs along the CA1 neural elements in these two portions have to maintain dorsoventral axis (Marcelin et al., 2012a, 2012b; oscillations in similar frequencies. Still, there are Dougherty et al., 2012). The presence of Ih in evidences of a large functional and genetic variability interneurons of the stratum oriens has been studied along the hippocampal dorsoventral axis (Jung et al., (Maccaferri & McBain, 1996; Lupica et al., 2001; 1994; Fanselow & Dong, 2010; Shah et al., 2016; Matt et al., 2011) but whether Ih differs in OLM cells Cembrowski et al., 2016). For example, genetic along the dorsoventral axis has not yet been shown. analysis of single cells in the mouse hippocampus has Ih time constants approximate theta periods shown the vHPC CA1 to be heterogeneous, while the and in PCs these currents are known to contribute to dorsal CA1 region appears more homogeneous on a subthreshold oscillations at theta frequencies single cell level (Shah et al., 2016). The travelling (Zemankovics et al., 2010; Dougherty et al., 2013). theta wave has been shown to rely on temporal shifts Moreover, Ih shapes membrane resonance frequency, between several groups of different interneurons causing cells to preferably fire in response to inputs during sharp wave-ripple activity (Forro et al., 2015) around theta frequency (Pike et al., 2000; especially between the intermediate CA1 and the Zemankovics et al., 2010). Computational studies dorsal CA1 region, indicating that information is have shown that Ih tunes both theta and gamma flowing through the whole hippocampal structure. oscillations in a CA3 circuit model (Neymotin et al., Nevertheless, the hippocampus also appears to be a 2013). On the contrary, a recent study found CA1 structure with topographical task division (e.g. dHPC OLM cells to follow theta-frequency inputs is involved with cognitive functions while the vHPC independently of Ih (Kispersky et al., 2012). Hence, relates to emotions and stress) (Fanselow & Dong, the role of OLM cell Ih in generating theta oscillations 2010). This functional divergence mirrors the is not well understood. differential connectivity patterns along the Here, we investigate active and passive hippocampus. The dHPC receive inputs from the membrane properties of dHPC and vHPC CA1 OLM medial entorhinal cortex, and the vHPC afferents arise cells of the mouse hippocampus. We demonstrate that mostly from the lateral entorhinal cortex. The vHPC, dorsal CA1 OLM cells have larger Ih than ventrally but not the dHPC, is connected to the medial located OLM cells. Besides, we show that Ih prefrontal cortex and basolateral amygdala, thereby contributes to setting the firing rate of dorsal but not more associated with emotions than its dorsal ventral OLM cells. We also show that the larger Ih in counterpart (Felix-Ortiz et al., 2013; Padilla-Coreano dorsal OLM cells is responsible for setting a et al., 2016). Thereby, the hippocampal structure does subthreshold resonance close to theta frequency, while not appear homogenous and this leaves questions to ventral OLM cells resonate in a significantly lower how hippocampal theta oscillations are produced. frequency. Modeling studies have suggested that oriens lacunosum moleculare (OLM) interneurons are Methods essential to hippocampal theta rhythmogenesis ETHICAL APPROVAL through feedback inhibition (Tort et al., 2007; In this study we used transgenic mice, with cyclization Neymotin et al., 2013). OLM cells are intimately recombinase (Cre) coupled to the Chrna2 gene connected with pyramidal cells (PCs), providing (Chrna2-Cre) (Hilscher et al., 2017; Perry et al., 2015; inhibition specifically to PC distal dendrites (Leão et Leão et al., 2012) which were crossed with a al., 2012). OLM cells together with bistratified cells tdTomato fluorescence reporter line (targeting different dendritic compartments of PCs) Gt(ROSA)26Sortm14(CAG-tdTomato)Hze (R26tom; Allen Brain (Klausberger, 2009) contribute to behaviors where Institute). All experiments were approved by the gating of information is important for functional Swedish Animal Welfare authorities and follow output of the hippocampus (Katona et al., 2014; Uppsala University guidelines for the care and usage Lovett-Barron et al., 2014; Leão et al., 2012). of laboratory animals (ethics permits C132/13 and

C135/14). Efforts were made to minimize the numbers hippocampus in a coronal section. Images were of animals used. collected on a Zeiss LSM 510 Meta confocal microscope and stacked. ELECTROPHYSIOLOGY Coronal slices from 10 Chrna2-Cre/R26tom transgenic DATA ANALYSIS AND STATISTICS mice (P19-29) were obtained. In summary, brains Patch-clamp data were analyzed with custom routines were rapidly removed and placed in ice-cold in MATLAB (version 2016b, MathWorks, MA, sucrose/artificial cerebrospinal fluid (ACSF) USA). Action potentials were triggered by 500 ms consisting of the following (in mM): KCl, 2.49; depolarizing current injections from 10 - 100 pA. The NaH2PO4, 1.43; NaHCO3, 26; glucose, 10; sucrose, first generated AP was analyzed for AP amplitude 252; CaCl2, 1; MgCl2, 4. Slices were cut using a (peak to AHP voltage), threshold (where the change in vibratome (VT1200, Leica, Microsystems) and were membrane potential exceeds 20 mV/ms), half-width subsequently moved to a submerged holding chamber (halfway between threshold voltage and peak) and containing normal ACSF (in mM): NaCl, 124; KCl, first spike latency (time between stimulus onset and 3.5; NaH2PO4, 1.25; MgCl2, 1.5; CaCl2, 1.5; the beginning/threshold of the first spike). AHPs were NaHCO3, 30; glucose,10, constantly bubbled with analyzed for magnitude (AP threshold - minimum of 95% O2 and 5% CO2 and kept at 35°C for 1 h then voltage trough between the first and the second AP in maintained at room temperature. The slices were a spike train). transferred to submerged chamber under an upright H-current was measured as steady-state microscope equipped with DIC optics (Olympus) and current following a 1s voltage step. The steady-state perfused with at 30°C oxygenated ASCF (1–1.25 current was the average current between 500-1000 ms. ml/min). Patch pipettes from borosilicate glass Tail currents were analysed to examine voltage capillaries (GC150F-10 Harvard Apparatus) were dependency of Ih. Normalized tail currents (I/Imin) pulled on a vertical puller (Narishige, Japan) with were fitted to a Boltzmann function (I/Imin = [1 + exp −1 resistance around 7 MΩ. Pipettes were filled with (V1/2 − V)/k] , where V1/2 is the half-activation internal solution containing (in mM): K-gluconate, voltage and k is the slope factor). Firing frequency 130; NaCl, 7; MgCl2, 2; ATP, 2; GTP, 0.5; HEPES, was calculated in response to 1000 ms depolarizing 10; EGTA, 0.1 (pH was adjusted to 7.2 using KOH). current injections from 0 - 200 pA, in 10 pA Whole cell current and voltage clamp recordings were increments. Maximum frequency was measured as the acquired using a Multiclamp 700B amplifier (Axon inverse of the first interspike interval. Steady-state Instruments, CA, USA) and digitized with a Digidata frequency was calculated as the inverse of the mean of 1440A data acquisition card (Axon Instruments, CA, the last three interspike intervals. USA). WinWCP and WinEDR softwares implemented To characterize the electrical resonance and by Dr.J. Dempster (University of Strathclyde, filtering properties of dorsal and ventral OLM cells, Glasgow, UK) were used to record the impedance (Z) amplitude profile (ZAP) was electrophysiological signals. Some experiments obtained (Puil et al., 1986; Hutcheon & Yarom, 2000). required bath application of the Ih blocker ZD7288 Therefore, a sinusoidal current with constant (Tocris Cookson Inc., Bristol, UK; 20 μM) and amplitude (three conditions of peak to peak value: ± tetrodotoxin (TTX, Tocris Cookson Inc., Bristol, UK; 25 pA, ± 50 pA and ± 100 pA) and linearly increasing 1 μM). When ZD7288 was applied to the bath only frequency (0 – 20 Hz over 20 s) was injected to dorsal one cell per slice was patched, as ZD7288 binds and ventral OLM cells and their voltage response was intracellularly and is therefore hard to wash out from measured. TTX (1 μM) was routinely applied to the the cytosol. perfusate to prevent cells from spiking. To measure the magnitude of the cell impedance as a function of IMAGING AND TRACING frequency, the magnitude of the fast Fourier transform Single images of OLM cells targeted by (FFT) of the voltage response was divided by the electrophysiological experiments and routinely filled magnitude of the FFT of the input current (Hutcheon with Alexa Fluor (Invitrogen, MA, USA; 488 nm), or & Yarom, 2000). biocytin (Sigma Aldrich, MO, USA; 1.5 mg/ml) Data is reported as mean ± Standard Error of counterstained with Streptavidin-Alexa-Fluor the Mean (SEM) and plot as bar plots or box plots. (Invitrogen, MA, USA; 488 nm), were classified as Data larger than q3 + 1.5*(q3 – q1) or smaller dorsal or ventral OLM cells based on the presence in than q1 – 1.5*(q3 – q1), with q1 and q3 denoting the the upper half or lower half of the visible length of the 25th and 75th percentiles, was considered as outlier and

3 discarded. Statistical comparisons were determined Visual inspection confirmed that there were less using two-tailed Student’s t-test and to account for Chrna2+ cell bodies in the far most rostral dHPC multiple comparisons, the data was analyzed using slices compared to the vHPC (Mikulovic et al., 2015). ANOVA and post-hoc test with Tukey correction (*p Current clamp recordings showed a mean input < 0.05, **p < 0.01). resistance of 642.70 ± 113.58 MΩ for dorsal OLM cells (n = 9 cells) and 326.94 ± 73.34 MΩ for ventral Results OLM cells (n = 9 cells; p = 0.0329; two-tailed Firing and intrinsic properties differ between dorsal Student’s t-test). Mean resting membrane potential and ventral OLM cells was -59.87 ± 0.62 mV for dorsal OLM cells (n = 9 To investigate whether OLM cell membrane cells) compared to -65.30 ± 2.19 mV for ventral OLM properties differ along the dorsoventral axis of the cells (n = 9 cells; p = 0.0298). Depolarizing current hippocampus we recorded electrophysiological injections in dorsal OLM cells resulted in a low properties of Chrna2+ OLM cells (Leão et al., 2012; frequency discharge (Figure 1B, left) with mean Perry et al., 2015; Mikulovic et al., 2015) in coronal action potential (AP) amplitude of 84.12 ± 7.65 mV, a brain slices of Chrna2-Cre/tdTomato mice (Hilscher et mean AP threshold of -41.03 ± 3.55 mV, a mean AP al., 2017; Perry et al., 2015; Leão et al., 2012). Whole- half-width of 1.24 ± 0.21 ms and a mean first spike cell patch clamp recordings from visually identified latency of 67.98 ± 24.44 ms (analyzed in response to OLM cells, expressing red fluorescent protein the first AP generated upon positive current injections (Tomato), were grouped as dorsal or ventral in post- of 10 - 100 pA, 500 ms, with 10 pA increments; n = 9 hoc analysis of images of Alexa488 or biocytin-filled cells). Ventral OLM cells had a mean AP amplitude of OLM cells (Figure 1A). 86.29 ± 4.58 mV (p = 0.8103), a mean AP threshold of -44.45 ± 1.76 mV (p = 0.4013), a mean AP half-width of 1.21 ± 0.12 ms (p = 0.9228) and a mean first spike latency of 75.22 ± 24.35 ms (n = 9 cells; p = 0.8366) (Figure 1B, right). Afterhyperpolarizations (AHPs) measured after the first AP had a mean magnitude of - 8.62 ± 1.65 mV in dorsal OLM cells and a mean magnitude of -8.10 ± 1.43 mV in ventral OLM cells (n = 9 cells; p = 0.8150). Both, dorsal and ventral OLM cells revealed AHPs waveforms with both a fast and a slow component as well as showed a gradual depolarization of AHP over repeated spikes. Passive and active membrane properties of dorsal and ventral OLM cells are summarized in Table 1. Hyperpolarizing current injections are known to generate a characteristic membrane ‘sag’ in OLM cells (Kispersky et al., 2012). Here we show dorsal OLM cells to produce a prominent membrane sag in response to negative current injections, while ventral OLM cell hardly display membrane sags (Figure 1B) suggesting that the dorsal portion of OLM cells

Figure 1. Hyperpolarization-induced responses of OLM cells produce larger Ih. Comparing instantaneous and are different in the dorsal and ventral CA1 hippocampal region. steady-state voltage for dorsal OLM cells in response (A) Schematic 3D images (middle) show the approximate location of patched dorsal and ventral OLM cells (dashed line marks the separation line). Representative images of Chrna2-Cre/tdTomato positive OLM cells (red) with examples of biocytin-filled OLM cells (green) for identification of spatial location in a 300 µm thick slice are shown. Dorsal (left) and ventral (right). Scale bar 50 µm. (B) Current clamp traces of a dorsal (left) and a ventral (right) OLM cell in response to depolarizing (30 pA) and hyperpolarizing (0 pA to -100 pA; 20 pA decrements) current injections are shown. (C) Current–voltage relationships between instantaneous membrane voltage (Inst, o) and steady-state membrane voltage (SS, ■) before (black) and after (grey) ZD7288 (20 µM) application for dorsal OLM cells (left) and ventral OLM cells (right). Data is presented as Table 1. Summary of membrane properties of dorsal and ventral mean ± SEM. *p < 0.05 OLM cells. 4 to negative current injections (0 to - 100 pA, 500 ms, OLM cells along the dorsoventral axis. In response to with 10 pA decrements) showed a signiﬁcant negative voltage steps (- 10 mV to -70 mV, 1000 ms, difference between instantaneous and steady-state with 10 mV decrements), dorsal OLM cells generated values in response to -60 pA to - 100 pA steps (n = 9 greater inward current compared to ventral OLM cells cells; p < 0.05; ANOVA; Figure 1C, left, black line) (Figure 2A-B, top), and this difference was abolished but no difference for ventral OLM cells (Figure 1C, by ZD7288 (20 µM) application (Figure 2A, bottom). right, black line). When applying the On average, dorsal OLM cells generated a steady-state HCN channel blocker ZD7288 (20 μM) to the current of -317.99 ± 14.91 pA in response to -70 mV artificial cerebrospinal fluid perfusate the sag-response holding potential which significantly diminished to - of dorsal OLM cells diminished (Figure 1C, left, grey) 217.43 ± 16.15 pA following ZD7288 application, whereas the response in ventral OLM cells in general resulting in an average Ih-dependent current of -100 remained unaltered (Figure 1C, right, grey). While the pA (subtracted current; Figure 2C). Dorsal OLM cell average resting membrane potential of dorsal and ZD7288-sensitive current was active at voltage steps ventral OLM cells was significantly different before more negative than -30 mV as shown as significant ZD7288 (p = 0.0322), this difference was abolished difference between current responses before and following ZD7288-application (dorsal OLM cells: - following ZD7288 application (n = 9 cells; p < 0.05; 64.4 ± 2.8 mV; ventral OLM cells: -66.1 ± 3.1 mV; p ANOVA; Figure 2C). Ventral OLM cells did not show = 0.7603). a significantly smaller current upon h-current inhibition, possibly masked by the small size of

Dorsal OLM cells generate larger Ih than ventral current and variability between cells (ventral OLM OLM cells cell steady state current: -238.11 ± 20.05 pA, in As PCs have been shown a dorsoventral gradient response to a -70 mV voltage step, n = 9 cells). expression of the hyperpolarization-activated current We next investigated tail currents generated

(Dougherty et al., 2012), we aimed to quantify Ih of by hyperpolarizing current steps. At -90 mV, the magnitude of dorsal OLM cells tail current was equal to -52.31 ± 3.40 mV (n = 9 cells), while ventral OLM cells current was equal to -22.92 ± 10.37 mV (n = 9 cells, p = 0.0160, two-tailed Student’s t-test). On average, dorsal OLM cells generated a twice as large tail current than ventral OLM cells, significantly

Figure 2. Dorsal OLM cells display more hyperpolarization- Figure 3. Ih of dorsal and ventral OLM cell has similar voltage- activated current (Ih) than ventral OLM cells. (A) Example dependency. (A) Graph showing difference in magnitude of tail traces of a dorsal OLM cell response to negative voltage steps currents generated following termination of negative voltage steps before (top) and after (bottom) ZD7288 application. (B) The same between dorsal OLM cells (black) and ventral OLM cells (red). as presented in [A] is shown for a ventral OLM cell. (C) Graph of Inset: example of tail currents (arrow) from a dorsal OLM cell. average current responses to negative voltage steps showing the Scale bar 100 pA, 100 ms. (B) Boltzmann curves illustrating the current before (control, ●) and the current remaining after ZD7288 normalized current in response to voltage for dorsal (black) and application (■) for dorsal OLM cells. Digital subtraction of these ventral (red) OLM cells. (C) Boxplot of half-activation voltage of currents shows the ZD7288-sensitive current (Δ). (D) The same as dorsal (left) and ventral (right) OLM cells. (D) Boxplot comparing in [C] shown for ventral OLM cells. Data is presented as mean ± slope factor (k) for dorsal (left) and ventral (right) OLM cells. *p < SEM. *p < 0.05 0.05, **p < 0.01 5 different at potentials equal or lower than -90 mV (Figure 3A). Fitting a Boltzmann equation to quantify the voltage-dependency (Figure 3B) of Ih showed dorsal OLM cells to have a half-activation voltage

(V1/2) of -92.72 ± 1.12 mV, and a slope factor (k) of -

7.52 ± 0.35, while for ventral OLM cells the V1/2 was - 94.85 ± 1.21 mV and k was -8.10 ± 0.66 (Figure 3C- D).

Ih modulates ﬁring frequency of dorsal OLM cells

As Ih has been shown to also influences the firing frequencies of OLM cells (Maccaferri & McBain, 1996), we quantified the firing across depolarizing current steps (10 - 200 pA, 1000 ms, with 10 pA increments) for dorsal and ventral OLM cells separately. Here, examining spike frequency adaptation we extrapolated the firing frequency from the interspike intervals (ISIs) for the first two (maximum frequency) and the last four (steady-state frequency) action potentials in OLM cell spike trains in response to 150 pA (1000 ms) before and after ZD7288 application (Figure 4A-B). At current steps ≥ 140 pA, ZD7288 significantly decreased the maximum frequency (by increased the initial ISIs) of Figure 4. Dorsal OLM cell firing frequency is modulated by Ih. (A) Current clamp traces of a dorsal OLM cell (control in black, dorsal OLM cells (at 150 pA: from 34.36 ± 4.21 Hz to after ZD7288 application in grey) in response to a 150 pA, 1000 ms 26.72 ± 1.52 Hz; n = 9 cells; p < 0.05; ANOVA; long stimulus. Grey boxes highlight maximum frequency (inverse of Figure 4C, top) while steady-state frequency was the first interspike interval) and steady-state frequency (inverse of lowered at current steps ≥ 70 pA with ZD7288 present the mean of the last three interspike intervals.) resp. (B) Same as presented in [A] shown for a ventral OLM cell. (C) Top: The (at 150 pA: from 19.62 ± 1.64 Hz to 15.54 ± 1.60 Hz; maximum frequency before (control, ●) and after ZD7288 n = 9 cells; p < 0.05; ANOVA; Figure 4C, bottom). application (■) for dorsal OLM cells. Bottom: The steady-state Differently, ZD7288 did not significantly influence firing frequency before (control, ●) and after ZD7288 application maximum (at 150 pA: from 40.73 ± 1.21 Hz to 37.47 (■) for dorsal OLM cells. (D) The same as presented in [C] is shown for ventral OLM cells. Data is presented as mean ± SEM. ± 4.41 Hz; n = 9 cells; p ≥ 0.05; Figure 4D, top) or *p<0.05 steady-state frequencies (at 150 pA: from 19.47 ± 1.75 Hz to 17.95 ± 0.64 Hz; n = 9 cells; p ≥ 0.05; Figure increasing (1 Hz/s) sine wave with constant amplitude

4D, bottom) in ventral OLM cells. Thereby, Ih during 20 s. Initially voltage responses during control supports dorsal OLM cells firing at higher frequencies conditions (black voltage traces) and following by shortening spike latency. ZD7288 application (grey voltage traces) for dorsal (Figure 5A) and ventral (Figure 5B) OLM cells were

Resonance frequencies are different for dorsal and recorded. Blocking Ih with ZD7288 caused the ventral OLM cells membrane to act like a low-pass filter (Figure 5A-B), Next, we investigated resonance frequency of dorsal with passive membrane properties shaping the cell and ventral OLM cells in the presence of the voltage- response, as previously reported for pyramidal cells gated sodium channel blocker tetrodotoxin (TTX, 1 (Hu et al., 2009). Instead, leaving Ih intact, we μM) which was applied to prevent action potential compared the magnitude of the impedances as a generation. Membrane resonance is shaped by function of frequency. We performed fast Fourier interactions between passive membrane properties and transform (FFT) of the voltage responses (Figure 5C- additional currents with slow time constants, such as Ih D, top) as well as FFT of the ZAP input current (Hu et al., 2009). OLM cells were stimulated with a (Figure 5C-D, middle), then the impedance magnitude sinusoidal current (the ZAP current giving the (Z) was calculated by dividing FFT(V) by FFT(I) (Hu impedance (Z) Amplitude Profile) in the form of a et al., 2002). The impedance profile peak value chirp stimulus of three different amplitudes (± 25 pA, highlights the resonance frequency (fR) of OLM cells. ± 50 pA and ±100 pA). Chirps consisted of a linearly Here, we found that the peak impedance value, in

Figure 5. Dorsal OLM cells have a higher resonance frequency than ventral OLM cells. (A) One example of voltage responses of a dorsal OLM cell in response to a chirp stimulus (top: ± 25 pA, middle: ± 50 pA, bottom: ± 100 pA) before (black) and after ZD7288 (grey), showing ZD7288 augmenting low-pass filtering of injected 20 s long chirp current (frequency changing linearly from 0-20 Hz). (B) The same as presented in [A] is shown for a ventral OLM cell. (C) Top: Fast Fourier transform (FFT) of membrane potential responses as shown in [A] for dorsal OLM cells. Middle: Fast Fourier transform of the impedance amplitude function (ZAP) input for ± 25 pA (blue), ± 50 pA (red), ± 100 pA (black). Bottom: Impedance amplitude proﬁles for dorsal OLM cells in response to ± 25 pA (blue), ± 50 pA (red), ± 100 pA (black). The vertical line shows the mean resonance frequency (fR) in response to each chirp frequency. The resonance frequency value in response to a 100 pA chirp is stated in the figure (black line). (D) The same as presented in [C] is shown for ventral OLM cells from [B]. Close-up is shown in the inset.

response to a 100 pA chirp signal, was significantly different between dorsal (fR = 2.75 ± 0.12 Hz, n = 5 cells) and ventral (fR = 1.85 ± 0.10 Hz, n = 5 cells, p = 0.0372; two-tailed Student’s t-test; Figure 5C-D, bottom) OLM cells. Together, these data provide electrophysiological evidence that dorsal and ventral

OLM cells differently express Ih and further have distinct membrane resonance properties.

OLM cell Ih oppose the PC Ih gradient along the dorsoventral axis of the hippocampucells Studies focusing on membrane properties of pyramidal cells (PCs) along the dorsoventral CA1 axis has shown ventral PCs to be more excitable than dorsal PCs, due to a higher input resistance of ventral PCs (Marcelin et al., 2012a; Dougherty et al., 2012; Malik et al., 2016; Milior et al., 2016). When comparing our data with data from Dougherty et al., Figure 6. Hippocampal OLM cells and PC show opposite (2012 and 2013) we found opposite yet membrane resting and resonance properties. Data summarizing properties of dorsal and ventral OLM cells compared to previously complementary basic membrane properties and published PC values (Vrest and Rinp were taken from Dougherty et resonance frequency between the two cell types al., 2012 (*) and fR was taken from Dougherty et al., 2013 (^)) to depending on dorsal or ventral location within the plot complementary differences between the two cell populations hippocampal CA1 region (Figure 6). depending on location. Bar graphs represent percentage (maximum absolute value for each condition; Vrest, Rinp and fR as 100%) for both groups.

Discussion magnitude of the current. This might suggest that Recent studies have suggested that the dorsal and similar types of HCN channels are contributing to ventral hippocampus are genetically, generating the current, although dorsal OLM cells electrophysiologically and morphologically distinct having stronger membrane expression than ventral areas (Fanselow & Dong, 2010; Dougherty et al., OLM cells. Studies of HCN2 knockout mice suggest 2013). Most electrophysiology studies addressing OLM cell Ih to mostly consist of HCN2 subunits and differential membrane properties of dorsal and ventral that this current regulates resting membrane potential hippocampus cells have targeted the pyramidal cells (Matt et al., 2011). Interestingly, the resting (Marcelin et al., 2012a; Dougherty et al., 2012; Malik membrane potential of OLM cells in Matt et al. was et al., 2016; Milior et al., 2016) while little is known similar to dorsal OLM cells in our study, and HCN2 about the properties of interneurons. Chittajallu et al. knockout mice showed an Vrest similar to after showed that OLM cells can be separated based on ZD7288 application for dorsal OLM cells reported their expression of either 5-HT3A receptors here (Matt et al., 2011). Bender et al. show overlap (5HT3AR) or the transcription factor Nkx2-1, but it is between somatostatin and HCN2 mRNA of stratum not known whether there is a differential distribution oriens interneurons, while a lack of overlap between of 5-HT3AR or Nkx2-1 positive OLM cells along the HCN1 and somatostatin-expression in stratum oriens dorsoventral axis (Chittajallu et al., 2013). However, interneurons from rats (Bender et al., 2001). This is in 5-HT3AR and Nkx2-1 expressing OLM cells were agreement with OLM cells recordings from HCN1 differentially involved in network oscillations, with knockout mice not showing significant difference in Nkx2-1-expressing OLM cells firing phase locked to hyperpolarization-activated currents compared to kainate-induced gamma oscillations in vitro normal mice (Matt et al., 2011). HCN4 subunits have (Chittajallu et al., 2013). Here, we find that our ventral been shown to label cell somas of interneurons of the OLM cell recordings showing a resting membrane oriens layer of the CA1 (Hughes et al., 2013; Bender potential and input resistance similar to both the 5- et al., 2001) but it is unclear if OLM cells express HT3AR and Nkx2-1 positive OLM cells populations HCN4 subunit. Furthermore, differential expression of (Chittajallu et al., 2013), yet further studies are HCN channels on OLM cell dendrites has been necessary for addressing whether the ventral Chrna2+ suggested by a multi-compartment computational OLM cells express 5-HT3AR and/or Nkx2-1. Another model of OLM cells (Sekulić et al., 2014, 2015) but study (Kipiani, 2009) has shown that parvalbumin remains to be studied in vitro. For PCs, Ih in dendritic (PV) expressing OLM cells fire phase-locked to the in recordings shows a gradient along the dorsoventral vitro induced gamma oscillations, while parvalbumin axis (Marcelin et al., 2012b). These findings are in negative OLM cells do not. Interestingly, Chrna2+ agreement with immunohistochemical studies of HCN OLM cells do not express PV, nor do they fire phase- subunit expression on PC dendrites (Dougherty et al., locked to the carbachol-induced gamma activity in 2012). In addition, Marcelin et al. showed vHPC CA1 vitro (Mikulovic et al, submitted 2017). This indicates PCs to be more sensitive to cyclic AMP (cAMP) that Chrna2 positive OLM cells, prevalent in the (Marcelin et al., 2012a) that shift voltage-dependency vCA1, form a distinct subpopulation, different from of HCN4 channels, but less so of HCN2 and HCN1 dorsal OLM cells. As a consequence, a distinct genetic channels, to more positive potentials (Santoro & marker for dorsal OLM cells may still be lacking. Baram, 2003; Robinson & Siegelbaum, 2003). If OLM cell Ih is modulated by cAMP remains to be studied. Dorsal OLM cells express more hyperpolarization- activated current than ventral OLM cells Dorsal OLM cells have Ih-dependent spike-frequency We found that the resting membrane potential of adaptation dorsal OLM cells was significantly more depolarized Spike-frequency adaptation or firing rate gain is than of ventral OLM cells, and this difference was due important for balancing excitatory and inhibitory input to the larger Ih as application of the Ih -blocker and is a common feature of firing of e.g. motor ZD7288 abolished this difference. The small neurons and primary afferents of the auditory nerve hyperpolarization induced current of ventral OLM (Hultborn et al., 2004; Benda & Hennig, 2008), and cells did not contribute significantly to setting the spike frequency adaptation has been shown to be resting membrane potential. Analyzing tail-currents shaped by conductance and voltage fluctuations in showed dorsal and ventral OLM cells to share similar CA1 PCs (Fernandez et al., 2011). Many inhibitory voltage-dependency of Ih despite the difference in interneurons contributing to rhythmic networks, such

8 as locomotor networks, also demonstrate spike- hippocampus (Lubenov & Siapas, 2009; Patel et al., frequency adaptation, with some shown to be 2012) and it is associated with cognition and spatial modulated by Ih (Perry et al., 2015; Borowska et al., information processing. Differently, lower frequency, 2013). Here, we found dorsal OLM cell spike cholinergic-dependent type 2 theta is involved in frequency to be modulated by Ih. Since dorsal and emotional processing and suggested to originate in the ventral OLM cells display drastically different input ventral hippocampus (Ciocchi et al., 2015). A recent resistance, with dorsal OLM cells expressing higher study (Fuhrmann et al., 2015) has shown that type 1 input resistance compared to ventral OLM cells, this theta is driven by the glutamatergic input from the could make dorsal OLM cells more sensitive to small medial septum onto putative OLM cells in dHPC, current input, generating rapid responses. Thereby while recent work from our own laboratory dorsal OLM cells may respond collectively more (Mikulovic et al., submitted 2017) revealed that homogenous to a variety of inputs. Instead, ventral cholinergic input onto vHPC OLM cells drive type 2 OLM cells, with a low input resistance, may generate theta activity. In addition, one recent modeling study more graded responses to various types of inputs. (Sekulić & Skinner, 2017) demonstrated that OLM Whether such properties of OLM cells have functional cells can be recruited at high or low theta frequencies output for memory computational task remains to be depending on presence and absence of h channels on investigated, yet, interestingly a recent study show their dendrites and co-regulation with the slow how ventral, but not dorsal, inactivation of the delayed rectifier channel. The authors suggest that this hippocampus impairs reward memory (Riaz et al., system of channels could represent the “switch” that 2017). allows two different OLM cells subpopulations to be type 1 or type 2 theta spiking resonators. Hence, our Dorsal and ventral OLM cells have different data support the emerging evidence that OLM cells resonance frequencies play distinct roles in dorsal and ventral theta Oscillations can be generated by synchronized pacemaking. rhythmic synaptic input to populations of neurons, and is influenced by membrane properties and The larger Ih in dorsal OLM cells opposes the lower subthreshold currents expressed by postsynaptic Ih in dorsal PCs neurons, which can make neurons likely to resonate in It is interesting that dorsal OLM cells and dorsal PCs certain frequencies. Impedance represents the show similar resonance frequency, while several frequency-domain extension of membrane resistance membrane properties are strikingly opposing along the for neurons and adding HCN channels to a membrane dorsoventral axis of the CA1. Dougherty et al. will add high-pass filter properties to cells (Kase & investigated properties along the dorsoventral axis and

Imoto, 2012). Thereby, slow activating, subthreshold found a greater Ih in ventral PCs compared to dorsal active HCN channels can influence resonance PCs (Dougherty et al., 2012). Also, in the entorhinal properties of neurons. cortex, the time constants of Ih of PCs differ along the Here we show that both dorsal and ventral dorsoventral axis, a finding shown to account for grid OLM cells show membrane resonance behavior. field organization and spacing (Giocomo & Hasselmo, However, dorsal OLM cells showed a higher 2008; Garden et al., 2008; Giocomo et al., 2011). resonance frequency, possibly due to the larger Ih in Here, comparing features related to Ih, we show that, dorsal OLM cells. Two studies from horizontal slices dorsal OLM cell have a resting membrane potential ~5 of the rat CA1 has examined whether OLM cells show mV more depolarized than ventral OLM cell while resonance behavior. Zemankovics et al. reported ten resting membrane potential of dorsal PCs is ~5 mV out of 15 OLM cells to resonate in a frequency range more hyperpolarized than ventral PCs. Dorsal OLM of 1 – 3 Hz (Zemankovics et al., 2010) while cells have input resistance (measured at resting Kispersky et al. found no resonance peak in membrane potential) almost twice as higher than impedance measurements of OLM cells (Kispersky et ventral OLM cells, whereas the input resistance of al., 2012). In our preparation, we found dorsal OLM dorsal PCs is only half of the input resistance of cells of mice to resonate at close to 3 Hz while ventral ventral PCs (see (Dougherty et al., 2012)). Likewise, OLM cells to resonate at close to 2 Hz. The existence the resonance frequency of dorsal OLM cells is higher of two different types of theta oscillations in vivo has than the resonance frequency of dorsal OLM cells, been postulated for decades (Kramis et al., 1975). The while PCs show opposite features with dorsal PCs higher frequency, movement-related, cholinergic- having a lower resonance frequency than ventral PCs. independent type 1 theta originates in the dorsal Finally, firing frequency of PCs and OLM cells along

9 the dorsoventral axis is differently influenced by Ih. ventral hippocampal CA1 projection neurons. Science 348, 560–563. While dorsal PCs do not fire at current inputs of 200 pA, ventral PCs fire regularly with action potentials Dougherty KA, Islam T & Johnston D (2012). Intrinsic excitability generated in ~21 Hz at 200 pA current injections. of CA1 pyramidal neurones from the rat dorsal and ventral hippocampus. J Physiol 590, 5707–5722. OLM cell on the other hand, are accommodating interneurons, resulting in a high-frequency discharge Dougherty KA, Nicholson DA, Diaz L, Buss EW, Neuman KM, of the initial action potentials in ~48 Hz in ventral Chetkovich DM & Johnston D (2013). Differential expression of HCN subunits alters voltage-dependent gating OLM cell and ~36 Hz in dorsal OLM cells. Due to the of h-channels in CA1 pyramidal neurons from dorsal and stronger spike-frequency adaptation in ventral OLM ventral hippocampus. J Neurophysiol 109, 1940–1953. cells, the steady-state firing of dorsal and ventral OLM Fanselow MS & Dong H-W (2010). Are The Dorsal and Ventral cells results in a similar firing frequency of ~20 Hz at Hippocampus functionally distinct structures? Neuron 65, 7. 200 pA current injections. Hence, opposing membrane features of OLM cells and PCs may contribute to the Felix-Ortiz AC, Beyeler A, Seo C, Leppla CA, Wildes CP & Tye KM (2013). BLA to vHPC inputs modulate anxiety-related generation of travelling waves of theta through the behaviors. Neuron 79, 658–664. hippocampal structure, despite the hippocampus not being a homogenous structure on a cellular level. Fernandez FR, Broicher T, Truong A & White JA (2011). Membrane voltage fluctuations reduce spike frequency In summary, we show that, similarly to PCs, adaptation and preserve output gain in CA1 pyramidal OLM cells are also electrophysiologically distinct in neurons in a high-conductance state. J Neurosci Off J Soc the dorsal and ventral hippocampus. These differences Neurosci 31, 3880–3893. may have important implications in pacemaking Forro T, Valenti O, Lasztoczi B & Klausberger T (2015). Temporal properties of dorsal and ventral PC/OLM circuits. It is organization of GABAergic interneurons in the intermediate CA1 hippocampus during network oscillations. Cereb known that Ih is distributed differentially in soma and Cortex N Y N 1991 25, 1228–1240. dendrites of PCs and future studies should address whether there are differences in compartmentalization Fuhrmann F, Justus D, Sosulina L, Kaneko H, Beutel T, Friedrichs D, Schoch S, Schwarz MK, Fuhrmann M & Remy S (2015). of dorsal and ventral OLM cell Ih, as the current Locomotion, Theta Oscillations, and the Speed-Correlated distribution between soma and dendrites have Firing of Hippocampal Neurons Are Controlled by a Medial important implications in the function of this current Septal Glutamatergic Circuit. Neuron 86, 1253–1264.

(Sekulić et al., 2014, 2015). Garden DLF, Dodson PD, O’Donnell C, White MD & Nolan MF (2008). Tuning of synaptic integration in the medial References entorhinal cortex to the organization of grid cell firing fields. Neuron 60, 875–889. Benda J & Hennig RM (2008). Spike-frequency adaptation generates intensity invariance in a primary auditory Giocomo LM & Hasselmo ME (2008). Time constants of h current interneuron. J Comput Neurosci 24, 113–136. in layer ii stellate cells differ along the dorsal to ventral axis of medial entorhinal cortex. J Neurosci Off J Soc Neurosci Bender R., Brewster A, Santoro B, Ludwig A, Hofmann F, Biel M 28, 9414–9425. & Baram T. (2001). Differential and age-dependent expression of hyperpolarization-activated, cyclic nucleotide- Giocomo LM, Hussaini SA, Zheng F, Kandel ER, Moser M-B & gated cation channel isoforms 1–4 suggests evolving roles Moser EI (2011). Grid cells use HCN1 channels for spatial in the developing rat hippocampus. Neuroscience 106, 689– scaling. Cell 147, 1159–1170. 698. Hilscher MM, Leão RN, Edwards SJ, Leão KE & Kullander K Borowska J, Jones CT, Zhang H, Blacklaws J, Goulding M & (2017). Chrna2-Martinotti Cells Synchronize Layer 5 Type Zhang Y (2013). Functional subpopulations of V3 A Pyramidal Cells via Rebound Excitation. PLOS Biol 15, interneurons in the mature mouse spinal cord. J Neurosci e2001392. Off J Soc Neurosci 33, 18553–18565. Hu H, Vervaeke K, Graham LJ & Storm JF (2009). Complementary Cembrowski MS, Bachman JL, Wang L, Sugino K, Shields BC & theta resonance filtering by two spatially segregated Spruston N (2016). Spatial Gene-Expression Gradients mechanisms in CA1 hippocampal pyramidal neurons. J Underlie Prominent Heterogeneity of CA1 Pyramidal Neurosci Off J Soc Neurosci 29, 14472–14483. Neurons. Neuron 89, 351–368. Hu H, Vervaeke K & Storm JF (2002). Two forms of electrical Chittajallu R, Craig MT, McFarland A, Yuan X, Gerfen S, Tricoire resonance at theta frequencies, generated by M-current, h- L, Erkkila B, Barron SC, Lopez CM, Liang BJ, Jeffries BW, current and persistent Na+ current in rat hippocampal Pelkey KA & McBain CJ (2013). Dual embryonic origins of pyramidal cells. J Physiol 545, 783–805. functionally distinct hippocampal O-LM cells revealed by differential 5-HT3AR expression. Nat Neurosci 16, 1598– Hughes DI, Boyle KA, Kinnon CM, Bilsland C, Quayle JA, 1607. Callister RJ & Graham BA (2013). HCN4 subunit expression in fast-spiking interneurons of the rat spinal cord Ciocchi S, Passecker J, Malagon-Vina H, Mikus N & Klausberger T and hippocampus. Neuroscience 237, 7–18. (2015). Brain computation. Selective information routing by

Hultborn H, Brownstone RB, Toth TI & Gossard J-P (2004). Key Marcelin B, Liu Z, Chen Y, Lewis AS, Becker A, McClelland S, mechanisms for setting the input–output gain across the Chetkovich DM, Migliore M, Baram TZ, Esclapez M & motoneuron pool. Prog Brain Res 143, 75–95. Bernard C (2012a). Dorsal-ventral differences in intrinsic properties in developing CA1 pyramidal cells. J Neurosci Hutcheon B & Yarom Y (2000). Resonance, oscillation and the 32, 3736–3747. intrinsic frequency preferences of neurons. Trends Neurosci 23, 216–222. Marcelin B, Lugo joaquin N, Brewster AL, Liu Z, Lewis AS, McClelland S, Chetkovich DM, Baram TZ, Anderson AE, Jung MW, Wiener SI & McNaughton BL (1994). Comparison of Becker A, Esclapez M & Bernard C (2012b). Differential spatial firing characteristics of units in dorsal and ventral dorso-ventral distributions of Kv4.2 and hyperpolarization- hippocampus of the rat. J Neurosci Off J Soc Neurosci 14, activated cyclic adenosine monophosphate gated channel 7347–7356. (HCN) proteins confer distinct integrative properties to hippocampal CA1 pyramidal cell distal dendrites. J Biol Chemjbc.C112.367110. Kase D & Imoto K (2012). The Role of HCN Channels on Membrane Excitability in the Nervous System. J Signal Transduct 2012, 619747. Matt L, Michalakis S, Hofmann F, Hammelmann V, Ludwig A, Biel M & Kleppisch T (2011). HCN2 channels in local inhibitory interneurons constrain LTP in the hippocampal direct Katona L, Lapray D, Viney TJ, Oulhaj A, Borhegyi Z, Micklem BR, perforant path. Cell Mol Life Sci CMLS 68, 125–137. Klausberger T & Somogyi P (2014). Sleep and movement differentiates actions of two types of somatostatin- expressing GABAergic interneuron in rat hippocampus. Mikulovic S, Restrepo CE, Hilscher MM, Kullander K & Leão RN Neuron 82, 872–886. (2015). Novel markers for OLM interneurons in the hippocampus. Front Cell Neurosci; DOI: 10.3389/fncel.2015.00201. Kipiani E (2009). OLM interneurons are transiently recruited into field gamma oscillations evoked by brief kainate pressure ejections onto area CA1 in mice hippocampal slices. Milior G, Castro MAD, Sciarria LP, Garofalo S, Branchi I, Georgian Med News63–68. Ragozzino D, Limatola C & Maggi L (2016). Electrophysiological Properties of CA1 Pyramidal Neurons along the Longitudinal Axis of the Mouse Hippocampus. Sci Kispersky TJ, Fernandez FR, Economo MN & White JA (2012). Rep 6, 38242. Spike resonance properties in hippocampal O-LM cells are dependent on refractory dynamics. J Neurosci Off J Soc Neurosci 32, 3637–3651. Neymotin SA, Hilscher MM, Moulin TC, Skolnick Y, Lazarewicz MT & Lytton WW (2013). Ih tunes theta/gamma oscillations and cross-frequency coupling in an in silico Klausberger T (2009). GABAergic interneurons targeting dendrites CA3 model. PloS One 8, e76285. of pyramidal cells in the CA1 area of the hippocampus. Eur J Neurosci 30, 947–957. Padilla-Coreano N, Bolkan SS, Pierce GM, Blackman DR, Hardin WD, Garcia-Garcia AL, Spellman TJ & Gordon JA (2016). Kramis R, Vanderwolf CH & Bland BH (1975). Two types of Direct Ventral Hippocampal-Prefrontal Input Is Required hippocampal rhythmical slow activity in both the rabbit and for Anxiety-Related Neural Activity and Behavior. Neuron the rat: Relations to behavior and effects of atropine, diethyl 89, 857–866. ether, urethane, and pentobarbital. Exp Neurol 49, 58–85. Patel J, Fujisawa S, Berényi A, Royer S & Buzsáki G (2012). Leão RN, Mikulovic S, Leão KE, Munguba H, Gezelius H, Enjin A, Traveling theta waves along the entire septotemporal axis of Patra K, Eriksson A, Loew LM, Tort ABL & Kullander K the hippocampus. Neuron 75, 410–417. (2012). OLM interneurons differentially modulate CA3 and entorhinal inputs to hippocampal CA1 neurons. Nat Neurosci 15, 1524–1530. Perry S, Gezelius H, Larhammar M, Hilscher MM, Lamotte d’Incamps B, Leao KE & Kullander K (2015). Firing properties of Renshaw cells defined by Chrna2 are Lovett-Barron M, Kaifosh P, Kheirbek MA, Danielson N, Zaremba modulated by hyperpolarizing and small conductance ion JD, Reardon TR, Turi GF, Hen R, Zemelman BV & currents Ih and ISK. Eur J Neurosci 41, 889–900. Losonczy A (2014). Dendritic Inhibition in the Hippocampus Supports Fear Learning. Science 343, 857– 863. Pike FG, Goddard RS, Suckling JM, Ganter P, Kasthuri N & Paulsen O (2000). Distinct frequency preferences of different types of rat hippocampal neurones in response to Lubenov EV & Siapas AG (2009). Hippocampal theta oscillations oscillatory input currents. J Physiol 529 Pt 1, 205–213. are travelling waves. Nature 459, 534–539. Puil E, Gimbarzevsky B & Miura RM (1986). Quantification of Lupica CR, Bell JA, Hoffman AF & Watson PL (2001). membrane properties of trigeminal root ganglion neurons in Contribution of the Hyperpolarization-Activated Current (I guinea pigs. J Neurophysiol 55, 995–1016. h) to Membrane Potential and GABA Release in Hippocampal Interneurons. J Neurophysiol 86, 261–268. Riaz S, Schumacher A, Sivagurunathan S, Van Der Meer M & Ito R (2017). Ventral, but not dorsal, hippocampus inactivation Maccaferri & McBain CJ (1996). The hyperpolarization-activated impairs reward memory expression and retrieval in contexts current (Ih) and its contribution to pacemaker activity in rat defined by proximal cues. Hippocampus; DOI: CA1 hippocampal stratum oriens-alveus interneurones. Hum 10.1002/hipo.22734. Dev 497 ( Pt 1, 119–130. Robinson RB & Siegelbaum SA (2003). Hyperpolarization- Malik R, Dougherty KA, Parikh K, Byrne C & Johnston D (2016). activated cation currents: from molecules to physiological Mapping the electrophysiological and morphological function. Annu Rev Physiol 65, 453–480. properties of CA1 pyramidal neurons along the longitudinal hippocampal axis. Hippocampus 26, 341–361.

Santoro B & Baram TZ (2003). The multiple personalities of h- Zemankovics R, Káli S, Paulsen O, Freund TF & Hájos N (2010). channels. Trends Neurosci 26, 550–554. Differences in Subthreshold Resonance of Hippocampal Pyramidal Cells and Interneurons: The Role of H-Current Sekulić V, Chen T-C, Lawrence JJ & Skinner FK (2015). Dendritic and Passive Membrane Characteristics. J Physiol; DOI: distributions of I h channels in experimentally-derived 10.1113/jphysiol.2009.185975. multi-compartment models of oriens-lacunosum/moleculare (O-LM) hippocampal interneurons. Front Synaptic Neurosci Competing interests 7, 2. The authors have no conflicts of interest to disclose. Sekulić V, Lawrence JJ & Skinner FK (2014). Using multi- compartment ensemble modeling as an investigative tool of Author contributions spatially distributed biophysical balances: application to hippocampal oriens-lacunosum/moleculare (O-LM) cells. M.M.H and R.N.L conceived and conducted the PloS One 9, e106567. experiments. M.M.H, R.N.L and K.E.L analyzed the results and wrote the paper with inputs from S.M and Sekulić V & Skinner FK (2017). Computational models of O-LM cells are recruited by low or high theta frequency inputs K.K. depending on h-channel distributions. eLife; DOI: 10.7554/eLife.22962. Acknowledgments Shah S, Lubeck E, Zhou W & Cai L (2016). In Situ Transcription This work was supported by the Swedish Foundation Profiling of Single Cells Reveals Spatial Organization of for International Cooperation in Research and Higher Cells in the Mouse Hippocampus. Neuron 92, 342–357. Education (STINT) www.stint.se, the Brazilian Tort ABL, Rotstein HG, Dugladze T, Gloveli T & Kopell NJ National Council of Technological and Scientific (2007). On the Formation of Gamma-Coherent Cell Development (CNPq) www.cnpq.br and the Brazilian Assemblies by Oriens Lacunosum-Moleculare Interneurons in the Hippocampus. Proc Natl Acad Sci 104, 13490–13495. Federal Agency for Support and Evaluation of Graduate Education (CAPES) www.capes.gov.br.

APPENDIX C Study III

101 RESEARCH ARTICLE Chrna2-Martinotti Cells Synchronize Layer 5 Type A Pyramidal Cells via Rebound Excitation

Markus M. Hilscher1,2*, Richardson N. Leão1,2, Steven J. Edwards1, Katarina E. Leão2☯, Klas Kullander1☯*

1 Unit of Developmental Genetics, Department of Neuroscience, Uppsala University, Uppsala, Sweden, 2 Brain Institute, Federal University of Rio Grande do Norte, Natal, Rio Grande do Norte, Brazil

☯ These authors contributed equally to this work. * [email protected] (MMH); [email protected] (KK) a1111111111 a1111111111 a1111111111 a1111111111 Abstract a1111111111 Martinotti cells are the most prominent distal dendrite±targeting interneurons in the cortex, but their role in controlling pyramidal cell (PC) activity is largely unknown. Here, we show that the nicotinic acetylcholine receptor α2 subunit (Chrna2) specifically marks layer 5 (L5) Martinotti cells projecting to layer 1. Furthermore, we confirm that Chrna2-expressing Marti- OPEN ACCESS notti cells selectively target L5 thick-tufted type A PCs but not thin-tufted type B PCs. Using Citation: Hilscher MM, Leão RN, Edwards SJ, Leão optogenetic activation and inhibition, we demonstrate how Chrna2-Martinotti cells robustly KE, Kullander K (2017) Chrna2-Martinotti Cells reset and synchronize type A PCs via slow rhythmic burst activity and rebound excitation. Synchronize Layer 5 Type A Pyramidal Cells via Rebound Excitation. PLoS Biol 15(2): e2001392. Moreover, using optical feedback inhibition, in which PC spikes controlled the firing of sur- doi:10.1371/journal.pbio.2001392 rounding Chrna2-Martinotti cells, we found that neighboring PC spike trains became syn-

Academic Editor: Alberto Bacci, ICM - Institut du chronized by Martinotti cell inhibition. Together, our results show that L5 Martinotti cells Cerveau et de la Moelle e´pinière, France participate in defined cortical circuits and can synchronize PCs in a frequency-dependent

Received: October 21, 2016 manner. These findings suggest that Martinotti cells are pivotal for coordinated PC activity, which is involved in cortical information processing and cognitive control. Accepted: January 5, 2017

Published: February 9, 2017

Copyright: © 2017 Hilscher et al. This is an open access article distributed under the terms of the Author Summary Creative Commons Attribution License, which permits unrestricted use, distribution, and Cognitive functions and information processing are linked to the coordination of neuro- reproduction in any medium, provided the original nal events and activities. This coordination is achieved through the synchronization of author and source are credited. neuronal signals within subnetworks. Local networks contain different types of nerve Data Availability Statement: All relevant data are cells, each of them playing distinct roles in the synchronization mechanism. To under- within the paper and its Supporting Information stand how synchronization is initiated and maintained, we have identified one of the files. key players using genetic strategies; we have identified a subtype of nicotine receptors Funding: Swedish Foundation for International uniquely expressed in cortical Martinotti cells. Because of their architecture and connec- Cooperation in Research and Higher Education tion properties, Martinotti cells are able to synchronize ongoing activity of unconnected (STINT) www.stint.se. Received by RNL and KK. pyramidal cells (PCs). We show that this mechanism only applies to one subtype of PCs, The funder had no role in study design, data collection and analysis, decision to publish, or thereby demonstrating that Martinotti cell inhibition is not spread randomly. By testing preparation of the manuscript. American Tinnitus optimal firing patterns of Martinotti cells, we are able to coordinate the firing of this spe- Association www.ata.org. Received by MMH, RNL, cific PC subtype over longer periods of time, showing how one unique interneuron is con- and KEL. The funder had no role in study design, tributing to information processing. data collection and analysis, decision to publish, or

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 1 / 26 Synchronization by Chrna2-Martinotti Cells preparation of the manuscript. Brazilian National Introduction Council of Technological and Scientific Development (CNPq) www.cnpq.br. Received by Martinotti cells, ubiquitous to the cortex [1], are the most prominent cross-laminar interneu- MMH, RNL, and KEL. The funder had no role in ron subtype forming synapses in layer 1 onto the distal dendrites of cortical pyramidal cells study design, data collection and analysis, decision (PCs) [1–3]. Despite this close structural relationship, the role of Martinotti cell inhibition is to publish, or preparation of the manuscript. not clear. Studies identifying Martinotti cells by various markers have found different mor- Brazilian Federal Agency for Support and phologies and microcircuit connectivity depending on the cortical layer in which their cell Evaluation of Graduate Education (CAPES) www. capes.gov.br. Received by MMH, RNL, and KEL. bodies reside [2]. In general, the division of neocortical interneurons into either parvalbumin-, The funder had no role in study design, data somatostatin (SOM)-, or 5HT3aR-expressing cells [4–6] has been helpful for dissecting neural collection and analysis, decision to publish, or functionality; yet, these groups can be further subdivided and show partial overlap between preparation of the manuscript. Swedish Research interneuron markers. Martinotti cells are a subclass of SOM+ cells [7,2,4], and several combi- Council www.vr.se. Received by KK. The funder nations of transgenic lines have been created to try to genetically and morphologically isolate had no role in study design, data collection and Martinotti cells [8–10]. For example, the SOM-cyclization recombinase (Cre) mouse line analysis, decision to publish, or preparation of the manuscript. Swedish Brain Foundation www. marks layer 1–projecting Martinotti cells with cell bodies in both layer 5 (L5) and layer 2/3 hjarnfonden.se. Received by KK. The funder had no (infragranular and supragranular layers) but also labels non-Martinotti cells in layer 4 [10]. role in study design, data collection and analysis, Although electrophysiologically, SOM+ Martinotti cells are often referred to as low-threshold decision to publish, or preparation of the spiking (LTS) neurons [3] or slow-inhibitory interneurons [11], early studies have shown up manuscript. to four different firing patterns for Martinotti cells [2,9]. Competing Interests: The authors have declared Functionally, cortical SOM+ interneurons have been suggested to provide a “blanket of that no competing interests exist. inhibition” [12], a dense and nonspecific spread of inhibition on nearby PCs. Whether Marti- Abbreviations: ADP, afterdepolarization; AHP, notti cells are capable of generating such indiscriminate inhibition when firing simultaneously afterhyperpolarization; AP, action potential; ChR2, in large groups has not been tested. Martinotti cells that reside in the main cortical output L5 Channelrhodopsin-2; Chrna2, nicotinic provide frequency-dependent disynaptic inhibition (FDDI) on neighboring PCs [13,14], an acetylcholine receptor 2 subunit; Cre, cyclization α inhibitory mechanism that synchronizes two or more PCs by one or a few Martinotti cells recombinase; EPSC, excitatory postsynaptic current; EPSP, excitatory postsynaptic potential; [15]. Synchronized activities in the cortex have been reported in vivo [16] as well as in vitro, FDDI, frequency-dependent disynaptic inhibition; where slow oscillations appear to be initiated in L5 [17]. Moreover, computational studies sug- GAD1, Glutamate decarboxylase 1; GIN cell, a type gest that Martinotti cell activity can synchronize L5 PC spiking through distal inhibition [18]; of somatostatin-expressing interneuron defined by however, this has not been tested experimentally. It is intriguing that distal dendrite–targeting its green fluorescent protein expression in a interneurons, generating attenuated inhibitory currents, can affect PC spike time output. transgenic mouse; HaloR, Halorhodopsin; IPSP, Here, we genetically targeted the L5 Martinotti cell population using a nicotinic acetylcholine inhibitory postsynaptic potential; L5, layer 5; LTS, low-threshold spiking; MCsα2, L5-specific Chrna2- receptor α2 subunit (Chrna2)-Cre mouse line to investigate how Martinotti cell inhibition can Martinotti cells; nAChR, nicotinic acetylcholine synchronize L5 PC firing. Our results show that Chrna2-Cre–labeled L5 Martinotti cells were receptor; OLM, oriens-lacunosum moleculare; PC, preferentially and reciprocally connected with thick-tufted PCs. Furthermore, we found that pyramidal cell; RT-PCR, reverse transcription PCR; short burst firing of L5 Martinotti cells was able to reset L5 PC spiking and that controlling SEM, standard error of the mean; SOM, Martinotti cell activity to rapid bursts repeated in a slow rhythm was the most efficient inhibi- somatostatin; VIP, vasoactive intestinal peptide. tion to synchronize unconnected PCs. Finally, we show that L5 PC microcircuits could synchronize their own action potentials (APs) when coupled by L5 Martinotti cells and that inhibition was crucial for PC synchronization over prolonged periods.

Results L5 Martinotti Cells Express the Nicotinic Acetylcholine Receptor Subunit Alpha 2 To test whether Chrna2 can be used as a marker of Martinotti cells, we crossed Chrna2-Cre mice with a tdTomato reporter line (R26tom, Fig 1 and S1 Fig) [19,20]. L5 Chrna2-Cre/R26tom cells were found in all cortices (S1A–S1C Fig, S1 Movie). Only very few Chrna2-Cre/R26tom cells were detected in layer 2/3 (supragranular: 27 [2.4%] versus infragranular: 1102 [97.6%] cells in an 800-μm-thick section), suggesting that Chrna2-Cre/R26tom specifically labels L5 neurons. Reconstructions of patched biocytin-filled Chrna2-Cre/R26tom cells showed that 36

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 2 / 26 Synchronization by Chrna2-Martinotti Cells

Fig 1. L5 Chrna2-Cre/R26tom cells show Martinotti cell morphology and are low-threshold, slow accommodating firing. (A) Confocal image (20 μm, coronal slice) of primary auditory cortex of a Chrna2-Cre/R26tom mouse showing tdTomato+ somas (red) in L5 with dense axonal arborizations in layer 1 (arrow in corner, scale bar = 100 μm). (B) Confocal image and tracing of a biocytin-filled tdTomato+ neuron (green). Reconstruction of soma and dendrites (black) and axon (red; scale bar = 20 μm) shows long axonal projections to layer 1. (C) Confocal images of a biocytin-filled (green) tdTomato+ neuron among several tdTomato+ neurons (red) show that cells have an ovoid cell body in L5, bipolar dendritic morphology, and proximal axonal arborizations. (D) Image illustrating how the Chrna2-Cre/R26tom axons emerge from the main dendrite (circle). Scale bars = 50 μm. (E) Image showing the long axonal arborizations (arrows) from one biocytin-filled Chrna2-Cre/R26tom cell (yellow) to layer 1 and the dense axonal ramifications (asterisk) in layer 1 from all Chrna2-Cre/R26tom cells expressing tdTomato (red). Scale bars = 50 μm. (F) Example from another biocytin-filled Chrna2-Cre/R26tom cell to emphasize axonal arborization extending laterally in layer 1, seen as a thin yellow axon at the border of the axonal plexus of Chrna2-Cre/R26tom cell in layer 1. (G) Top: Example current clamp traces from a tdTomato+ cell showing low-threshold, accommodating firing (20 pA response in red, 100 pA in black, 500 ms) and rebound APs (−20 to −80 pA, 500 ms) typical for Martinotti cells. Bottom: Current clamp trace in response to a 200-pA, 1,000-ms-long stimulus used for analysis in (H). (H) Left: The frequency/current (f/I) curve of MCsα2 shows an average firing rate around 20 Hz (at 200 pA, 1,000 ms) indicating slow spiking properties. Middle: Difference in maximum frequency and steady-state frequency for each neuron to a 200 pA, 1,000-ms-long current step highlights an accommodating discharge. The black line depicts the mean adaptation. Right: Spike-frequency adaptation is shown as a function of time. Data (n = 36 cells) are presented as mean ± standard error of the mean (SEM) and shown in S1 Data. doi:10.1371/journal.pbio.2001392.g001

out of 37 (97.3%) cells met the criteria of deep layer Martinotti cells by having an ovoid cell body in L5, bipolar dendritic morphology, axons emerging from the main dendrite, proximal axonal arborizations, and long axonal projections to layer 1 with a dense arborization around PC distal dendrites (Fig 1B–1F and S2 Fig) [2,3,10,13]. The excluded cell had its cell body outside of L5. At higher magnifications, the axonal plexus of L1 is highlighted by red fluorescent signal (tdTomato) of Chrna2-Cre/R26tom cell axonal ramifications (see star in Fig 1E, 1F and

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 3 / 26 Synchronization by Chrna2-Martinotti Cells

S1C, S1D and S2B Figs). Immunohistochemistry revealed that 30.3% of Chrna2-Cre/R26tom cells were SOM+ (n = 3 mice, 2–3 mo old, 8 sections of 35-μm thickness, of a total of 792 cells (L1–L6); 297 cells were Chrna2+, 495 cells were SOM+, and 90 of these were double labelled for both genetically expressed tdTomato and SOM antibodies; S1D and S1E Fig). Counting cells only in L5, we found a total of 549 cells, of which 292 cells were Chrna2+, 257 were SOM +, and 85 of these were double labelled for both Chrna2 and SOM (29.1% of L5 Chrna2+ cells were SOM+; S1D and S1E Fig). Single-cell reverse transcription PCR (RT-PCR) of individually picked Chrna2-Cre/R26tom cells (n = 7 cells, n = 2 animals) found 6/7 collected neurons to be positive for Glutamate decarboxylase 1 (GAD1; S1F Fig), while no cell was positive for the vesicular glutamate transporter subtype 1 or 2, indicating an inhibitory nature of Chrna2+ neurons. Membrane properties of Chrna2-Cre/R26tom cells measured by whole-cell patch clamp revealed a mean input resistance of 337.28 ± 11.42 MO and a mean resting membrane potential of −63.69 ± 1.02 mV (n = 36 cells, S1 Data). The first AP generated upon 500 ms depolarizing current injections with 1-pA increments (on average, the first spike was reached in response to 18.11 ± 1.97 pA) had a mean AP amplitude of 72.61 ± 2.20 mV, an AP threshold of −43.28 ± 0.53 mV, an AP half-width of 1.92 ± 0.08 ms, and a first spike latency of 254.75 ± 24.24 ms (n = 36 cells, S1 Data). Afterhyperpolarizations (AHPs) measured after the first AP had a mean magnitude of −8.28 ± 0.72 mV with a gradual depolarization over repeated spikes, and each AHP displayed both a fast and a slow component (Fig 1G, S1 Data) [2]. Hyperpolarizing currents generated rebound afterdepolarizations (ADPs) and, on average, 2.50 ± 0.25 rebound APs (at −80 pA, 500 ms) upon termination of current steps but did not produce a sizable membrane “sag,” suggesting that these cells have a minimal hyperpolariza- tom tion-activated current (Ih, Fig 1G, top, S1 Data). The Chrna2-Cre/R26 cell firing frequency versus current relationship showed a linear increase of average firing rate with increasing current towards a frequency of 22.4 ± 2.49 Hz (at 200 pA, n = 36 cells; Fig 1H, left, S1 Data), indicating that Chrna2-Cre/R26tom cells are slow spiking interneurons. The relationship between maximum frequency (52.87 ± 2.44 Hz, n = 36 cells, S1 Data) and steady-state frequency (21.87 ± 1.02 Hz, n = 36 cells, S1 Data) revealed a spike-frequency adaptation ratio (see Materi- als and Methods) of ~59% in response to a 200-pA step (Fig 1G, bottom and Fig 1H, middle). The spike-frequency adaption shows how firing frequency of Chrna2-Cre/R26tom cells decreases as a function of time (20.16 ± 2.44 Hz at 416 ms, n = 36 cells; Fig 1H, right). In summary, both morphological and electrophysiological characteristics of infragranular Chrna2-Cre/R26tom cells (~97%) are similar to those reported in previous studies of Martinotti cells [2,13,21], thus we hereafter refer to these L5-specific Chrna2-Martinotti cells as MCsα2.

MCsα2 Are Reciprocally Connected with Thick-Tufted Type A PCs Previous studies have speculated that SOM+ cells (including Martinotti cells) predominantly contact specific subpopulations of PCs [21,22]. Thus, we patched pairs of a MCα2 and its neighboring PC in L5 (60 μm). Next, we categorized PCs into type A and type B cells based on morphological and electrophysiological criteria [23]. Cells with a large cell body, thick- tufted basal dendrites with apical dendrites extensively branching in layer 1, burst-regular spiking, responding with large AHPs, prominent hyperpolarization sags, and pronounced rebound ADP were classified as type A PCs (Fig 2A, left). Cells with small soma, thin-tufted basal dendrites with limited spreading apical dendrites, absence of AHP or ADP, and small hyperpolarization sags were classified as type B PCs (Fig 2A, right). We expected L5-specific MCsα2 to be locally connected to PCs [2,13] and found, amongst morphologically recon- structed pairs of patched cells, that 77% of type A PCs–MCsα2 were connected (n = 7/9), while none of the patched type B PCs–MCsα2 were connected (n = 0/9). Out of the paired

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 4 / 26 Synchronization by Chrna2-Martinotti Cells

Fig 2. MCsα2 connect to local type A PCs but not type B PCs. (A) Left: The reconstruction of a typical type A PC showing a thick-tufted dendrite (scale bar = 40 μm) and its response to a 500-ms-long depolarizing (100 pA) and hyperpolarizing (−60 pA) stimulus. Right: A representative type B PC with a thin-tufted apical dendrite (scale bar = 40 μm) and its current clamp response (as for left). Note the deeper AHP (following a depolarizing current pulse), the more prominent sag (during a hyperpolarizing current pulse), as well as the pronounced rebound ADP (following a hyperpolarizing current pulse) in the type A PC compared to type B PC (see arrows). (B) Type A PCs can excite postsynaptic MCsα2 (inset) and generate facilitating EPSPs (left, n = 7/9 pairs, 12 repetitions from one example pair are shown) when stimulated with high frequency (70 Hz), whereas type B PCs do not trigger EPSPs in MCsα2 (right, n = 0/9 pairs, 12 repetitions). Inset shows experimental setup. (C) Typical MCα2 discharges (top) to a 500-ms-long (25 pA) stimulus are shown. Inset shows experimental setup. MCα2 spikes cause inhibition in postsynaptic type A PCs (inset) displaying synaptic depression (middle left, n = 7/9 pairs), whereas type B PCs do not receive MCα2 inhibition (middle right, n = 0/ 9 pairs). Grey dashed lines highlight timing of presumably individually generated IPSPs for type A PCs, whereas for type B PCs, dashed line shows the lack of response. Example IPSP responses of 12 repetitions are shown in grey, mean response in black (bottom). doi:10.1371/journal.pbio.2001392.g002

MCα2–type A PC recordings, 55% (n = 5/9) of pairs were reciprocally connected. Paired recordings of PCs and MCsα2 revealed that high-frequency stimulation (70 Hz) of type A PCs generated excitatory postsynaptic potentials (EPSPs), or an occasional spike, in MCsα2 (example shows 12 repetitions from one patched type A PC–MCα2 pair; Fig 2B, left), whereas type B PC stimulation did not result in EPSPs in local MCsα2 (12 repetitions; Fig 2B, right).

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 5 / 26 Synchronization by Chrna2-Martinotti Cells

Additionally, inhibitory postsynaptic potentials (IPSPs) were generated in type A PCs (IPSP amplitude: −1.08 ± 0.12 mV; example shows 12 repetitions from one patched MCα2–type A PC pair; Fig 2C, left) after MCα2 stimulation, whereas no inhibitory responses were observed in type B PCs (12 repetitions; Fig 2C, right). Thus, our data suggest that MCsα2 connect with type A PCs and not type B PCs.

Optogenetic Activation of MCsα2 Shows Frequency-Dependency of MC± PC Inhibition We next investigated the influence of MC inhibition on PCs when simultaneously activating a large group of MCsα2 in Chrna2-Cre mice (1–2 mo old) previously injected with floxed Chan- nelrhodopsin-2 (ChR2; Fig 3A). Compared to electrical stimulation of single MCsα2, light activation of MCα2 groups produced IPSPs in type A PCs with a higher mean amplitude (from −0.96 ± 0.05 mV to −1.41 ± 0.04 mV), a smaller mean time to peak (from 29.53 ± 1.24 ms to 20.54 ± 0.97 ms), and a decreased mean half decay time (from 63.46 ± 2.04 ms 51.20 ± 2.08 ms [all comparisons: n = 12 cells; a total of 54 IPSPs, p < 0.001, Fig 3B, left and S3A Fig, S2 Data]). We also recorded from type B PCs (n = 12 cells), but no IPSPs were observed in type B PCs in response to blue light stimulation of ChR2+ MCsα2 (Fig 3B, right). We next tested different stimulation frequencies (2, 5, 15, 25, 40, and 70 Hz) [13,18] for ChR2+ MCsα2 to investigate the role of MCα2 firing frequency on IPSP amplitude in type A PCs (n = 12 cells; Fig 3C). We found a nonlinear relationship between IPSP amplitude and MCα2 stimulation frequency in which, at higher frequencies (>15 Hz), IPSPs summed into smooth compound IPSPs, probably due to the depressing synaptic properties of the MCα2-to-PC connection [13]. To further characterize the frequency-dependency of the MCα2–PC IPSPs, we stimulated MCsα2 with continuous light, which generated accommodating firing in MCsα2 (Fig 3D, bottom and S3B– S3D Fig) but still large IPSPs in PCs (n = 12 cells, Fig 3D and 3E, S3 Data). This suggests that large compound IPSPs can be generated in type A PCs when MCsα2 fire at high frequencies and that the compound IPSP amplitude mainly depends on the firing frequency during the first 300 ms (Fig 3D and 3E).

Bursts of MCsα2 Can Reset Type A PC Firing Martinotti cells have been shown to provide FDDI [13,24,15] onto PCs. To examine whether L5-specific MCsα2 generate FDDI, we patched pairs of type A PCs with cell bodies next to MCsα2 and provided high-frequency (70 Hz) current injection to one PC (Fig 4A, top). This led to an early EPSP in the other PC, presumably due to monosynaptic PC–PC connections, followed by a delayed inhibition (FDDI, amplitude: −1.02 ± 0.23 mV, time to peak: 57.16 ± 2.43 ms, half decay time: 95.34 ± 5.64 ms, n = 12 cells; Fig 4A, bottom, S4 Data). In some cases, patched type A PC pairs were not monosynaptically connected and only the FDDI was observed (amplitude: −1.05 ± 0.21 mV, time to peak: 91.33 ± 2.83 ms, half decay time: 123.34 ± 5.87 ms, n = 12 cells; Fig 4B, top and Fig 4C, S4 Data). To confirm that the FDDI response was mediated by MCsα2, we silenced MCsα2 in slices from Chrna2-Cre/Halorhodop- sin (HaloR)-floxed mice (1–2 mo old) with green light. Indeed, green light abolished the delayed inhibition (n = 12 cells; Fig 4B, bottom and Fig 4C, S4 Data). However, we also noted that large IPSPs occurred in both patched type A PCs upon termination of the light pulse (amplitude: −1.66 ± 0.16 mV, time to peak: 53.45 ± 2.23 ms, half decay time: 88.45 ± 4.53 ms, n = 24 IPSPs, all comparisons: p < 0.001; Fig 4B, bottom and Fig 4C, S4 Data). Current clamp recordings from MCsα2 showed that the green light generated strong hyperpolarization of MCsα2 and that, subsequently, bursts of rebound spikes were generated in HaloR-expressing MCsα2 upon light termination (Fig 4D, top and S4A and S4B Fig, S5 Data). HaloR-activation

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 6 / 26 Synchronization by Chrna2-Martinotti Cells

Fig 3. MC±PC inhibition is frequency dependent. (A) Expression of AAV-DIO-ChR2-EYFP (green) in MCsα2 (red) in a primary auditory cortical slice used for optogenetic stimulation, with inset on L5 showing overlap of membrane expression in yellow (scale bars 50 μm). (B) Optogenetic activation (3-ms light pulses, 488 nm) of a group of MCsα2 induced IPSPs in type A PCs (left) but not type B PCs (right) (n = 12 cells, single examples in grey, mean in black). (C) Example traces show MCsα2 responses to blue light stimulation at various frequencies (3-ms blue light pulses at 2, 5, 15, 25, 40, and 70 Hz) and the corresponding IPSPs in a nearby type A PC (n = 12 cells, single examples in grey, mean in black). At higher frequencies (15Hz), the MCα2±PC synapse showed depression. Note that MCsα2 could not follow 70-Hz light stimulation for prolonged time. (D) Top: Continuous light stimulation of MCsα2 (500 ms) generated large type A IPSP amplitudes (middle; n = 12 cells, single examples in grey, mean in black) similar in magnitude to IPSPs generated by high-frequency stimulation at 70 Hz. Bottom: Spike-frequency adaptation of MCsα2 is shown as a function of time (see also S3B±S3D Fig). (E) Mean IPSP amplitudes in type A PCs following stimulation of MCsα2 at different frequencies (from (C) and (D); 2 Hz: −1.57 ± 0.13 mV, 5 Hz: −1.70 ± 0.08 mV, 15 Hz: −3.05 ± 0.12 mV, 25 Hz: −3.20 ± 0.13 mV, 40 Hz: −3.76 ± 0.10 mV, 70 Hz: −4.37 ± 0.10 mV, 500 ms: −4.41 ± 0.07 mV; 2, 5 Hz versus 15, 25 Hz ≙ p < 0.0001; 15, 25 Hz versus 40 Hz ≙ p < 0.0001; 40 Hz versus 70 Hz, 500 ms ≙ p < 0.001; mean ± SEM, ANOVA, n = 12 cells, S3 Data). doi:10.1371/journal.pbio.2001392.g003

for 500 ms consistently evoked one or more rebound APs with varying frequency in MCsα2

that could not be blocked by the Ih blocker ZD7288 (20 μM, S4C Fig), similar to Ih-independent rebound APs in distal dendrite–targeting X98 cells [9]. Additionally, MCsα2 and type A PCs were patched in the presence of carbachol (10 μM) to further examine how MCsα2 could modulate L5 PC spontaneous firing. Carbachol depolarized PCs and MCsα2 by 10 to 15 mV and did not result in a specific oscillatory frequency as seen with high concentrations of carbachol;

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 7 / 26 Synchronization by Chrna2-Martinotti Cells

Fig 4. MCsα2 contribute to FDDI, and MCα2 burst firing can reset type A PC spikes. (A) High-frequency stimulation (70 Hz, see arrow) of a presynaptic PC (▲) generates delayed IPSPs on a neighboring PC (▲) via intermediate MCsα2 (О). A mixed excitation (due to a monosynaptic PC±PC connection) followed by a disynaptic inhibition is shown (n = 12 cells, single examples in grey, mean in black). (B) An example of disynaptic inhibition alone (top) is shown (n = 12 cells, single examples in grey, mean in black). Silencing of HaloR-expressing MCsα2 via green light (555 nm) prevents FDDI, although IPSPs are generated following termination of green light stimulation (bottom, n = 12 cells, single examples in grey, mean in black). (C) Mean IPSP amplitudes with (white) and without (green) FDDI at two different time points. (D) Responses from HaloR-expressing MCsα2 (top) and local type A PCs (single-spiking and burst-spiking; middle) and type B PCs (bottom) are shown in presence of carbachol (10 μM). Green light stimulation (500 ms) hyperpolarizes HaloR- expressing MCsα2 and upon termination MCα2 rebound APs are triggered. This burst of APs generates robust inhibition in local postsynaptic type A PCs that synchronizes the timing of PC (rebound) APs. Kernel density estimates (orange) highlight increased (peaks) and decreased (valleys) co-occurance of APs. (E) Example of voltage clamp responses for type A (top) and type B (bottom) PCs in response to MCsα2 burst firing (single examples in grey, mean in black). Values are shown in S4 Data. doi:10.1371/journal.pbio.2001392.g004

instead, it showed broad peaks in the power spectral density plots (S5 Fig). Two firing patterns could be distinguished for type A PCs [25–28]: single-spiking (n = 42 cells; Fig 4D, middle) and burst-spiking (n = 18 cells; Fig 4D, middle). Independent of the type A PC firing type, bursts of HaloR-induced rebound spikes from MCsα2 caused large, compound IPSPs (S6 Fig, S6 Data) that resulted in a resetting of type A PC firing (Fig 4D, middle). We define resetting as temporally aligning spiking after a period of inhibition. APs from type A PCs aligned around 500 ms (single-spiking: 556.19 ± 9.42 ms, n = 42 cells; burst-spiking: 524.72 ± 10.17 ms, n = 18 cells, S4 Data) after light-off for HaloR-inhibition of MCsα2. The shape of the summed type A PC IPSP trace corresponded well to the first 3–4 rebound APs of MCsα2 (S6A Fig) and steadily evoked subsequent post-inhibitory rebound APs in the type A PCs (n = 60 cells; Fig 4D, middle) but not type B PCs (n = 12 cells; Fig 4D, bottom). Note that the IPSP amplitude more than doubled with carbachol present (Vm = −60 mV; IPSP amplitude = −1.66 ± 0.26 mV, Fig 4B, bottom; Vm = −48 mV [single-spiking PC]; IPSP amplitude = −3.51 ± 0.48; Vm = −48 mV [burst-spiking PC]; IPSP amplitude = −6.30 ± 0.91 mV, all comparisons: p < 0.0001; S6B and S6C Fig, S6 Data). Voltage clamp experiments (holding at −60 mV) further highlighted the presence of

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 8 / 26 Synchronization by Chrna2-Martinotti Cells

large IPSCs in type A PCs (107.40 ± 3.54 pA, n = 3 cells; Fig 4E, top, S4 Data) and the absence of IPSCs in type B PCs (n = 3 cells; Fig 4E, bottom). Together, these results show that a rapid burst of MCα2 APs can abruptly halt the firing of type A PCs and also reset PC firing by temporal coupling rebound APs of PCs while leaving type B PC firing unaffected. Therefore, we only aimed for type A PCs for the remainder of the study.

Bursts of MCα2 APs Synchronize Firing of Type A PCs at Slow Frequencies It is unknown how PCs synchronize their firing, although computational studies have suggested a role for distal dendrite–targeting interneurons in synchronizations [11,18,29]. Thus, we only patched unconnected type A PCs and recorded spontaneous firing, in the presence of carbachol, from two PCs simultaneously (n = 24 cells) while optogenetically stimulating the MCα2 population at various frequencies (2, 5, 15, 25, 40, and 70 Hz). To identify the frequency of MCα2 activity that best temporally aligns unconnected, randomly firing, type A PCs, we recorded repeats of 4-s sweeps (2 s light-off, 2 s light-on). When pairwise superimposing simultaneous recordings from two type A PCs, we observed that light stimulation of 2 Hz or 15 Hz (n = 24 cells; 12 black and 12 grey PC spike trains; Fig 5A) created a rhythmical firing pattern of type A PCs highlighted by kernel density estimates, which show the distribution of APs over time (orange traces). Although mean power spectral density plots from both 2 Hz and 15 Hz MCα2 stimulation revealed peaks around 2 Hz (1.99 ± 0.09 versus 1.87 ± 0.14 Hz, n = 24 cells), other frequencies tested (5, 25, 40, and 70 Hz) did not result in any clear peaks (Fig 5B, top and S7 Fig, S7 Data). This indicates that MCsα2 preferentially give rise to slow frequencies in a group of type A PCs. However, flat mean coherence plots of pairwise analyzed PCs did not show any correlation between simultaneously recorded type A PCs in specific frequency bands plotted up to 20 Hz. This suggests that type A PCs as a population can produce an oscillatory firing rhythm, but individual cells are mostly out of phase and not synchronized with each other (n = 24 cells; Fig 5B, bottom). To generate complete synchrony between type A PCs, we hypothesized that a rapid burst of MCα2 activity could reset/align type A PC spiking (3–4 APs as seen in Fig 4D and S6A Fig) and a slow rhythm could maintain in-phase synchronous firing. To test this, we patched two unconnected type A PCs and stimulated MCsα2 with 15-Hz bursts every 500 ms (2 Hz). This stimulation protocol resulted in high AP synchronization (Fig 5C) directly after MCsα2 were paused. Mean power spectral density (peak at 2.02 ± 0.04 Hz, n = 24 cells) and mean coherence examination showed that type A PCs followed MCα2 stimulation frequency and were pairwise aligned in that frequency (Fig 5D, S7 Data), suggesting synchronized firing of type A PCs at slow frequencies. This shows that MCsα2 have means to both initiate and maintain prolonged type A PC synchronous firing.

Slow Frequency Burst Stimulation of MCsα2 Is Synchronizing Type A PC Spike Timing via Minimally Depressing IPSPs Next, we sought to quantify the synchrony (provide a synchrony index [30]) between type A PC firing when MCsα2 were activated in bursts of 15 Hz. A representative recording of two simultaneously captured type A PCs is shown in Fig 6A, where initially unsynchronized type A PCs aligned during MCα2 stimulation (orange rectangles highlight synchronized APs). In the absence of MCα2 stimulation, the mean cross-correlograms of pairwise analyzed recordings showed only low magnitude peaks, while light stimulation organized firing of both type A PCs in cohorts every 500 ms (n = 24 cells; Fig 6B). A 3-fold increase in the synchrony index could be extracted from the cross-correlograms when MCsα2 were light stimulated (control: 0.21 ± 0.03, burst stimulation: 0.61 ± 0.04, n = 12 dual recordings, n = 24 cells, p < 0.0001; Fig 6C, S8 Data).

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 9 / 26 Synchronization by Chrna2-Martinotti Cells

Fig 5. MCα2 bursts synchronize type A PC firing in slow frequencies. (A) Population response of 12 dual recordings of unconnected type A PCs (n = 24 cells; 12 black and 12 grey PC spike trains) before and during pulsed light stimulation of ChR2-expressing MCsα2 (2 Hz [left] and 15 Hz [right]). Kernel density estimates (orange) show increased (peaks) and decreased (valleys) co-occurrences of APs. (B) Mean power spectral density plots for both cases revealed peaks around 2 Hz (top) but flat mean coherence plots (bottom), suggesting no synchronization (correlation) between firing of type A PCs within a frequency range of 0±20 Hz (n = 24 cells). (C) Pairwise overlayed type A PC voltage traces are shown in response to combined light stimulation (i.e., 15-Hz bursts in 2 Hz). Kernel density estimates (orange) highlight co-occurring PC APs. (D) Mean power spectral density plot (left) and mean coherence plot (right) (from [C], n = 24 cells), show a peak at 2 Hz corresponding to the rhythmic activation of MCsα2. Peak values are shown in S7 Data. doi:10.1371/journal.pbio.2001392.g005

Thus, we conclude that delivery of MCα2 inhibition in bursts of 15 Hz indeed synchronized type A PC firing. This is likely due to burst firing repeated in slow frequency, causing inhibition in type A PCs with little depression compared to continuous 15 Hz stimulation of MCsα2 showing apparent synaptic depression (examples in grey, mean in black, red dashed lines for visual guidance, Fig 6D). Interestingly, the 15-Hz continuous light stimulation revealed that bursting PCs can switch firing patterns from burst-spiking into single-spiking [25,18] during continuous low-magnitude inhibition (S8A Fig). This change in firing was observed only at near-threshold potentials; the physiological role remains to be studied (S8B Fig).

Type A PCs Can Auto-Synchronize Their APs When Intercoupled by MCsα2 To test if type A PC circuits can self-synchronize their firing through Martinotti cell activation, we designed a closed-loop system (optical feedback inhibition [31]) for paired recordings that

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 10 / 26 Synchronization by Chrna2-Martinotti Cells

Fig 6. Repeated bursts of MCα2 inhibition synchronize type A PC spike trains via minimally depressing IPSPs. (A) Voltage traces from an unconnected pair of type A PCs (black and grey) with MCsα2 stimulated in 15-Hz bursts at 2 Hz (blue dots). Orange rectangles highlight synchronous APs. (B) Mean cross- correlograms (n = 24 cells) show little synchrony before light stimulation and increased synchrony during light stimulation (15-Hz bursts in 2 Hz) of MCsα2 as shown by a prominent peak around zero (and recurring peaks at every 500 ms). (C) Box plots of the synchrony indices for control and 15-Hz bursts show the significant increase of synchrony (0 ≙ no synchronization, 1 ≙ full synchronization) when MCsα2 are stimulated by blue light in brief bursts (n = 12 dual recordings, p < 0.0001, two-tailed Student's paired t test). Values are shown in S8 Data. (D) IPSPs in type A PCs (n = 24 cells, single examples in grey, mean in black) following burst protocol of 15-Hz stimulation in 2 Hz (top) and constant 15-Hz light stimulation (bottom). Note minimal- depressing inhibition in the top and the depression of IPSPs leading to a rapid diminution of inhibition in the bottom (red dashed lines for improved visualization). doi:10.1371/journal.pbio.2001392.g006

delivered four blue light pulses (15 Hz) to the MCsα2 when one PC fired in the presence of carbachol (n = 24 cells; 12 black and 12 grey PC spike trains; Fig 7A, inset) in tissue from Chrna2-Cre mice previously injected with floxed ChR2. Kernel density estimates showed increased co-occurrences of type A PC APs during optical feedback inhibition (Fig 7A). More- over, because these experiments involved a leading and a following type A PC, it was possible to calculate the statistical dependency between the spike trains of two PCs and to express this with a mutual information index (see Materials and Methods). This index gives an estimate of how well one signal can predict the other and is helpful to interpret to what extent one PC can drive another, e.g., via recurrent or feedforward inhibition. In controls, the mutual information index was low (2 ± 1, n = 12 dual recordings), whereas turning on the optical feedback inhibition resulted in immediate auto-alignment of type A PC APs with an increased index (11 ± 5, n = 12 dual recordings, p < 0.05; Fig 7B, S9 Data). Venn diagrams show the mutual information as the degree of overlap between two circles, representing each PC train as entropy (Fig 7B, inset). The overlap demonstrates the predictive value (mutual dependency)

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 11 / 26 Synchronization by Chrna2-Martinotti Cells

Fig 7. Type A PCs auto-synchronize via MCα2 inhibition. (A) Coupling one PC (black) out of two unconnected type A PCs to the light source/optical feedback inhibition system shows unsynchronized activity before and synchronized APs during optical feedback inhibition. A total of 24 PC discharges pairwise aligned to the first PC AP with optical feedback inhibition are shown (n = 24 cells; 12 black and 12 grey spike trains). Kernel density estimates (orange) highlight increased (peaks) and decreased (valleys) co-occurrence of APs. Note that the time points of the blue light depend on the PC APs during the optical feedback inhibition and therefore vary between PC pairs. (B) Top: One pair of simultaneously recorded unconnected type A PCs (black and grey) showing discharges before and during the optical feedback inhibition. Bottom: Pairwise mutual information index versus time lag from recordings in (A) showing low mutual information for unconnected PCs and high mutual information around 0-ms lag for PCs coupled by optical feedback inhibition (n = 12 dual recordings, 24 cells). Inset: Amount of overlap in Venn diagrams (black and grey circles) shows low mutual information for unconnected (left) and significantly higher mutual information for coupled (right) PCs (p < 0.05, two-tailed Student's paired t test). Values are shown in S9 Data. doi:10.1371/journal.pbio.2001392.g007

between a known PC train and a following PC train. When shifting one spike train relative to the other, the incremental mutual information index plot (mutual information index as a function of time lag, Fig 7B) showed that the mutual dependency was largest around 0-ms lag, suggesting high synchronization of the two PCs directly when coupled by optical feedback inhibition. Peaks around ± 400–600 ms indicate that this activity-dependent inhibition causes repeated synchronization every 400–600 ms.

Discussion We found that MCsα2 were exclusively synaptically connected to large, thick-tufted PCs, often referred to as L5B PCs [32] or type A PCs [23]. Different PC morphologies seem to be associated with different connectivity patterns in the brain, e.g., large, thick-tufted PCs are usually synonymously named subcerebral projection neurons or pyramidal tract neurons, whereas thin-tufted PCs, or type B PCs [23] are callosal projection neurons or intratelencephalic neurons [26–28]. Our in vitro preparation could not define PCs according to connectivity patterns; however, based on the extensive branching of the distal dendrites and the large triangular- shaped cell bodies, we find it likely that the type A PCs correspond to the thick-tufted PCs [22] and are probably subcortically projecting [26–28]. Thick-tufted type A PCs can further be described by firing properties as single-spiking or burst-spiking [25–28]. Typically, at near- threshold potentials, bursting cells respond with two or more bursts, of two or more APs, generated in quick succession with short interspike intervals [25]. Burst properties of PCs disap- pear with increasing current injections [25] and may be dependent on the size of the dendritic tree [33]. In addition to morphological variances, such as a smaller soma compared to type A PCs, type B PCs had characteristic electrophysiological differences. Our pair-recordings between type A PCs and MCsα2 confirmed that type A PCs provided facilitating synaptic responses in MCsα2. We also found depressing synaptic connections from MCsα2 to type A

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 12 / 26 Synchronization by Chrna2-Martinotti Cells

PCs [13], while no type B PC connectivity with MCsα2 was observed. Lack of IPSPs in type B PC was not likely due to shunting of inhibition, as voltage clamp recordings also failed to find synaptic connectivity between MCsα2 and type B PCs. However, due to difference in thin- ver- sus thick-tufted morphology, it is possible that the internal solution creates less dialysis of the chloride ion Cl- concentration in type B PCs compared to type A PCs. Therefore, perforated patch recordings would be needed to firmly rule out the possibility of shunting of inhibition. Still, we found that FDDI, a Martinotti cell–dependent feature [13,24,15] was relayed by MCsα2 and consequently was specific for type A PCs. This is in agreement with a previous study showing FDDI between thick-tufted PCs but not between corticocallosally projecting cells [22]. Distal inhibition by individual MCs is important for shaping local dendritic voltage–activated responses. FDDI combined with dendritic depolarization has shown that MCs can atten- uate back-propagating AP-activated Ca2+ spike firing and thereby reduce burst firing of PCs [34]. On the network level, collective and precisely timed Martinotti cell activity can further be potent enough to affect somatic spike generation. Here, we first used HaloR to examine if blocking MCα2 activity could eliminate the appearance of FDDI in thick-tufted PCs [13,24,15]. This led to the observation that on the termination of green light MCsα2 fired bursts of HaloR- induced rebound spikes, inhibiting PCs and subsequently causing hyperpolarization-induced rebound APs that could reset PC firing. The HaloR-induced rebound in MCsα2 is a methodological artifact and has little physiological relevance; however, it is interesting to speculate whether Martinotti cells receive inhibition that could generate rebound spikes. Recently, the vasoactive intestinal peptide (VIP) interneuron has been shown to densely inhibit Martinotti cells [35]. The high connection probabilities between VIP cells and Martinotti cells [36] suggest that VIP cells could provide strong hyperpolarization in Martinotti cells for the possible generation of rebound excitation. Rebound spikes have been previously demonstrated to occur in entorhinal cortex neurons in vivo and are attributed to play a role in generating grid cell fields that usually arises when grid cells fire synchronized [37,38]. A similar role could be applicable to rebound spikes in the neocortex, where a “blanket of inhibition” [12] evolves through the synchronized spread of inhibition, serving to coordinate PC firing. Thereby, VIP cells appear to make "holes in the blanket of inhibition" [35] by inhibiting Martinotti cells [35,36]. In other words, VIP cell activity might regionally disrupt coordinated PC firing while local Martinotti cell activity could reset and rescue PC synchronous firing. Second, focusing on the combined activity of ChR2-expressing MCsα2, our data show that bursts of MCsα2 were able to reset type A PC firing and, if repeated, could synchronize PC activity. In a computer model, oscillatory inhibition of the distal PC dendrite at 10–20 Hz, presumably by LTS SOM+ Martinotti cells, was shown to control L5 PC firing [18]. Our findings support that 15 Hz firing of MCsα2 can align type A PC firing but also show that 15 Hz firing in short bursts more reliably synchronizes PCs compared to continuous 15 Hz firing. In other computational models, the importance of a beta rhythm in regulating gamma oscillations and intercortical signaling has been demonstrated and, furthermore, that the beta frequency is regulated by cholinergic modulators [11,29,39]. In this respect, the exclusive expression of the alpha 2 cholinergic receptors in MCsα2 is noteworthy and may suggest a specific role for MCsα2 in transmitting the modulatory action of cholinergic signaling. Cholinergic modulation of LTS cells has been suggested to generate beta oscillatory activity (beta2) in L5 of the primary auditory cortex [40]. These oscillations were insensitive to the muscarinic antagonist atropine but sensitive to the nicotinic receptor antagonist d-Tubocurarine [40]. Thus, computational and experimental studies indicate that the beta rhythm is important for network properties [40,41]; however, beta activity in bursts repeated in slow frequency has not been reported previously. At this slow frequency MCα2–PC inhibition shows minimal depression, similar to the minimal depression of slow firing SOM+ interneurons defined by their green fluorescent

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 13 / 26 Synchronization by Chrna2-Martinotti Cells

protein expression in a transgenic mouse (GIN-cells) [42], and therefore, a combination of rapid bursting and slow rhythmical inhibition seems most effective to synchronize PCs. Genetic targeting and optical feedback inhibition are a potent technique to study how PCs can drive a population of interneurons by their innate rhythm. A previous study used a closed- loop system to optogenetically produce feedback inhibition onto PCs from parvalbumin + interneurons [31]. Sohal et al. used synthetic excitatory post synaptic currents (EPSCs, dynamic clamp) in a single PC triggering parvalbumin+ interneuron excitation with light [31]. Differently, in our study, we depolarized optically stimulated MCsα2 in the presence of carbachol and measured synchronization of simultaneously recorded PCs using mutual information. Analogously to our optogenetic stimulation, gap junctions could provide a physiological mechanism for the synchronization of interneuron populations [42–44]. Berger et al. have shown the existence of electrical coupling between L5 MCs [15]. It will be interesting in the future to explore the existence of gap junctions between MCsα2. The Chrna2-Cre/R26tom mouse line simplifies identification and characterization of L5 Martinotti cells. MCsα2 are morphologically and electrophysiologically homogenous, further evincing the specificity of our marker. The dense axonal plexus observed in L1 and the near absence of Cre+ cell bodies in L2 (2.4%) in Chrna2-Cre/R26tom mice also indicate that Cre + cells are, in fact, L5 MCs. Still, we found a high proportion of Cre+ cells in Chrna2-Cre/ R26tom mice that were not labelled with the antibody against SOM, and this could be due to extra-somatic location of the peptide. So far, SOM-Cre is the most widely used transgenic mouse line for targeting MCs together with the GIN mouse [8,45], but still SOM-Cre has been shown to label all cell layers [46]. Here, we provide a layer-specific, single genetic marker for MCs across the cortex and confirmed their inhibitory nature using single cell RT-PCR. Although we did not explicitly block optogenetically evoked, inhibitory postsynaptic currents of MCsα2 (e.g., with Gabazine), Martinotti cell dendritic inhibition in vivo has been shown to α2 be GABAA-mediated [34]. The specific expression of Chrna2 in inhibitory L5 MCs raises questions of how important the α2 subunit is for cholinergic inputs. Several cortical interneurons express nicotinic acetylcholine receptors (nAChRs) [47–49], suggesting cholinergic modulation of inhibition in the cortex, most likely from the basal forebrain [50]. Cortical LTS cells, such as Martinotti cells [51,52], are excited by acetylcholine via nicotinic receptors and alter cortical circuit processing [53]. Cholinergic input is most likely mediated by additional nicotinic subunits that together form high affinity receptors for acetylcholine [54]. Several candidate subunits exist, but perhaps the more promising ones, judged from their specific expression in cortical L5, response to nicotine, and known co-expression with α2 subunits, include α6-nAChRs, β2-nAChRs, and β4-nAChRs [55–57]. The focus of our work has been on the functionality of Martinotti cells, not the nAChR subunits; however, earlier studies of cholinergic subunits can provide potential clues to Martinotti cell function. A deletion of α2-nAChRs has shown a normal phenotype but altered responses during nicotine-associated behaviors [58]. Deletion of α2-nAChRs has also shown reduced nicotine-induced hippocampal LTP in the temperoammonic path, most likely via oriens-lacunosum moleculare (OLM) interneurons [59,19]. Interestingly, Chrna2 is expressed in OLM cells, which target the distal dendrites of hippocampal PCs in a comparable manner as MCsα2 target the distal dendrites of cortical PCs. In some similarity to the suggested role for LTS cells in directing the flow of information in the cortex [53], OLM cells have been suggested to gate internal and external signals to the hippocampus [19]. In addition, MCsα2 might modulate cortical states, because SOM+ interneurons have recently been implied to be involved in transitions between UP and DOWN states [60]. Furthermore, studies of β2-nAChR KO mice have suggested a role for β2-containing nAChR in restricting cortical UP states and might be interesting for future studies on how nAChR are distributed in cortical interneurons such as Martinotti cells [13,61,62].

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 14 / 26 Synchronization by Chrna2-Martinotti Cells

Our preparation did not examine cortical UP and DOWN states; instead, we depolarized neurons with a low concentration of the cholinergic agonist carbachol. As the IPSP amplitude generated by MCsα2 is dependent on the membrane potential of the postsynaptic PCs, this illustrates how MCα2 inhibition (amplitude of IPSPs) could alter in a state-dependent manner, thereby exerting a state-dependent modulation of PC excitability. In summary, we report the identification of a marker specific for L5 Martinotti cells projecting to layer 1. These Martinotti cells were synaptically connected to large, thick-tufted PCs with prominent AHP and ADP, demonstrating a distinctive microcircuit between one type of interneuron and one subtype of PCs. Furthermore, we demonstrate that Martinotti cell–mediated inhibition can initiate and also maintain synchronous firing between PCs. We also show that this inhibition is frequency dependent and, when repeated in beta bursts, can continuously align firing of PCs. Lastly, using a closed-loop system in which PCs auto-synchronized their firing, we show that Martinotti cells were able to bridge the communication between unconnected PCs via activity-dependent inhibition. Thus, via their feedback and feedforward connections, Martinotti cells are important for regulating thick-tufted type A PC output in L5, most likely altering voltage-dependent dendritic properties and actively influencing somatic spike generation and synchronization.

Materials and Methods Ethics Statement All experiments were approved by the Swedish Animal Welfare authorities and followed Upp- sala University guidelines for the care and usage of laboratory animals (ethics permits C132/13 and C135/14). Efforts were made to minimize the numbers of animals used.

Mice In this study, we used transgenic mice (both males and females), with Chrna2-Cre [19,20] that were crossed with a tdTomato fluorescence reporter line Gt(ROSA)26Sortm14(CAG-tdTomato)Hze (R26tom; Allen Brain Institute) or with a HaloR-expressing line Rosa26-eNphR-EYFP (HaloR; Jackson Laboratory Stock No. 014539). Cre-negative littermates were routinely used as controls.

CLARITY The CLARITY procedure followed a standard protocol [63]. In summary, 2–3-mo-old Chrna2-Cre/R26tom mice (n = 2) were transcardially perfused with 20 ml of ice-cold 1x PBS solution followed by 20 ml of a hydrogel monomer solution consisting of 4% acrylamide, 0.05% bis-acrylamide, 0.25% VA-044 initiator, and 4% paraformaldehyde in PBS. Brains were quickly dissected and placed in hydrogel monomer solution for 3 d at 4˚C. Prior to polymeri- zation of the hydrogel monomer solution, samples were placed in a desiccation chamber attached to a vacuum pump. With the sample lid ajar, air was removed from the chamber for 10 min and replaced with nitrogen gas, after which the sample lid was tightly shut. The hydrogel monomer solution was polymerized by heating the samples to 37˚C for 3 h in a water bath whilst shaking. Embedded tissue was extracted from the gel, and brains were sliced to 3-mm coronal sections using a brain matrix. Passive clearing of slices was achieved by repeated, 3-d washes in a 4% Sodium Dodecyl Sulphate (SDS) sodium borate buffer (200 mM, pH 8.5) solution at 45˚C on a shaker plate for 6 wk. SDS was removed from the samples by incubating in

PBST0.1 (1x PBS and 0.1% Triton X-100) on a shaker plate for two consecutive 1-d washes.

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 15 / 26 Synchronization by Chrna2-Martinotti Cells

Clear tissue was refractive index-matched through serial, 1-d incubations in 20%, 40%, and 63% 2,20-Thiodiethanol (TDE, Sigma-Aldrich) in 1x PBS solution. Light sheet fluorescence images were acquired using Zeiss light sheet Z1 with a 5x/0.16 objective. Individual image tiles were 3D-stitched using Arivis Vision4D (Arivis). Imaris 8.1 (Bitplane) was used for analysis, volume rendering, soma detection, and data visualization. Chrna2-Cre/R26tom cells were counted using Imaris 8.1 (Bitplane) and Matlab (version 2013a, MathWorks) in an 800-μM section (AP: −2.40 to −3.20 mm, ML: 2.00 to 5.00 mm, and DV: 0.50 to 3.50 mm). Cells were divided into infragranular (roughly corresponding to L5/6) and supragranular (roughly corresponding to L2/3) cells by fitting a curve between ML: 4.25 mm, DV: 3.50 mm and ML: 2.00 mm, DV: 0.75 mm, roughly dorsolateral to L5.

Immunohistochemistry Chrna2-Cre/R26tom mice (2–3 mo, n = 3) were anesthetized with isoflurane and decapitated before dissection. Immunohistochemistry (IHC) was performed as previously described in [19]. The following dilutions of antibodies were used: SOM antibody 1:150 (MAB354, Anti- SOM Antibody, clone YC7 [Merck Millipore Corporation]), Anti-rat Cy5, 1:500 (Invitrogen).

Single-Cell Reverse Transcriptase PCR Following whole-cell recordings, we used strong negative pressure to suck the cytoplasm and organelles of the cells into the recording pipette tip, similar to [19]. Buffers and cDNA conver- sion are further described in [19]. A two-round PCR (nested) to detect GAD1, Vglut1, or Vglut2 cDNA was done. The following mix was used for the first and second round of PCR:

1.5 mM MgCl2, 10 pmol of each primer, 1.0 U of platinum Taq-DNA polymerase (Invitrogen), 20 mM TrisÁHCl, and 50 mM KCl pH 8.4; thermal cycle: 94˚C/2-min denaturation step followed by 35 cycles of 94˚C/50 s, 55˚C/45 s, and 72˚C/45 s. In the second round, instead of mixing the original template, 10% of the first PCR reaction as template was used. Second- round PCR products were visualized on 2% agarose gels. Primers were designed based upon sequences deposited in the GenBank database (www.ncbi.nlm.nih.gov/nucleotide). The primers used were as follows: first round: GAD1: CCAATAGCCTGGAAGAGAAGAG (forward), TCCCATCACCATCTTTATTTGA (reverse); Vglut1: CGCTACATCATCGCCATCATGAG (forward), GGAGGGGCCCATTTGCTCCA (reverse); Vglut2: GCCGCTACATCATAGCC ATC (forward), GCTCTCTCCAATGCTCTCCTC (reverse); second round: GAD1: CCAATA GCCTGGAAGAGAAGAG (forward), TCCCATCACCATCTTTATTTGA (reverse); Vglut1: CTGGAGGATTTATCTGCCAAAAAT (forward), GGTATGTGACCCCCTCCACCAAT (reverse); Vglut2: ACATGGTCAACAACAGCACTATC (forward), ATAAGACACCAGAAG CCAGAACA (reverse).

Electrophysiology Coronal slices from Chrna2-Cre/R26tom transgenic mice (P19–29, n = 12) were obtained similar to [24]. In summary, brains were rapidly removed and placed in ice-cold sucrose/artificial cerebrospinal fluid (ACSF) consisting of the following (in mM): KCl, 2.49; NaH2PO4, 1.43; NaHCO3, 26; glucose, 10; sucrose, 252; CaCl2, 1; MgCl2, 4. Coronal 300-μm-thick slices containing the primary auditory cortex were cut using a vibratome (VT1200, Leica, Microsystems) and were subsequently moved to a submerged holding chamber containing normal ACSF (in mM): NaCl, 124; KCl, 3.5; NaH2PO4, 1.25; MgCl2, 1.5; CaCl2, 1.5; NaHCO3, 30; glucose,10,

constantly bubbled with 95% O2 and 5% CO2 and kept at 35˚C for 1 h then maintained at room temperature. The slices were transferred to submerged chamber under an upright microscope equipped with DIC optics (Olympus) and perfused with oxygenated ASCF (1–

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 16 / 26 Synchronization by Chrna2-Martinotti Cells

1.25 ml/min) at 30˚C. For some experiments, carbachol (10 μM, Sigma-Aldrich) and ZD7288 (20 μM, Tocris Cookson Inc.) were added to the perfusate. Patch pipettes from borosilicate glass capillaries (GC150F-10 Harvard Apparatus) were pulled on a vertical puller (Narishige, Japan) with resistance around 7 MΩ. Pipettes were filled with internal solution containing (in mM) the following: K-gluconate, 130; NaCl, 7; MgCl2, 2; ATP, 2; GTP, 0.5; HEPES, 10; EGTA, 0.1 (pH was adjusted to 7.2 using KOH). Whole-cell current clamp recordings were acquired using a Multiclamp 700B amplifier (Axon Instruments, CA, USA) and digitized with a Digi- data 1440A data acquisition card (Axon Instruments, CA, USA). WinWCP and WinEDR softwares implemented by Dr. J. Dempster (University of Strathclyde, Glasgow, UK) were used to record electrophysiological signals. Patch-clamp data were analyzed with custom routines in MATLAB. APs were triggered by 500-ms depolarizing current injections from 10–100 pA. The first fired AP in response to minimal current injection was analyzed for AP amplitude (peak to AHP voltage), threshold (where the change in membrane potential exceeds 20 mV/ms), half-width (halfway between threshold voltage and peak), and first spike latency (time between stimulus onset and the AP threshold of the first spike). AHPs were analyzed for magnitude (AP threshold—minimum of voltage trough between the first and the second AP in a spike train). Spike rate was calculated as the number of APs per 1,000 ms. Spike-frequency adaptation was measured as the inverse of the mean of the last three interspike intervals (steady-state frequency) divided by the inverse of the first interspike interval (maximum frequency) in response to 100-pA current injections and subtracted from 100% (no adaptation). In spike-frequency adaptation plots, the reciprocal of consecutive interspike intervals is shown for each AP versus the time after onset of the current pulse. ChR2 stimulation frequencies (2, 5, 15, 25, 40, and 70 Hz) and HaloR-evoked hyperpolarization (5, 10, 100, 250, 500, and 1000 ms) were applied in randomized order to avoid statistical dependencies between cases. ChR2-triggered IPSPs and FDDI were measured as amplitude, time to peak, and half decay time, for which the onset was defined as the time at which the potential exceeded three times the standard deviation of the preceding baseline. HaloR-evoked hyperpolarization amplitudes were quantified as the difference between resting membrane potential and the peak of hyperpolarization. Rebound APs were quantified by number, maximum frequency, and duration (time of last minus time of first rebound AP). Burst-spiking PCs were distinguished from single-spiking PCs by obtaining the interspike intervals, at which burst-spiking PCs showed an increased amount of short (20 ms) interspike intervals at near- threshold potentials.

Imaging and Tracing MCsα2 were identified by cortical tdTomato-expression in Chrna2-Cre/R26tom mice that were perfused as previously described [19], and 20-μm- and 60-μm-thick coronal slices were imaged using a Zeiss LSM 510 Meta confocal microscope. Cells targeted by electrophysiology experiments were routinely filled with biocytin and stained with streptavidin-488 nm for post-hoc analysis. Images were collected on a Zeiss LSM 510 Meta confocal microscope and stacked and 2D-stichted using ImageJ 1.50a (NIH), where the color palette was adjusted for consistency (tdTomato—red, biocytin/ChR2—green). Soma detection and Neurite tracings were done semi-automated with NeuronJ 1.4.2 (ImageJ Plugin) or fully automated with Imaris 8.1 Fila- mentTracer (Bitplane) using “Autopath” in the algorithm settings and the threshold mechanism to correct for over-/under-sampled tracings following the image intensity.

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 17 / 26 Synchronization by Chrna2-Martinotti Cells

Optogenetics Chrna2-Cre/R26tom mice (1–2 mo old, n = 33) were anesthetized with Isofluran (1%–4%) and placed on a heat pad, with the head fixed with a nose holder and ear bars in a stereotaxic frame (Stoelting Co.). The skin was cleaned with iodine and opened with a straight incision, and the bregma was identified using small amount of peroxide. The coordinates for bilateral virus injection were as follows: AP: −2.46 mm, ML: +/−4.00 mm, and DV: 2.00/2.50 mm. We used bilateral injections to obtain the maximum number of slices containing the primary auditory cortex, with preserved dendritic trees of the type A PCs in layer 1 (on average two slices, 300 μm thick, per hemisphere) per animal. A small hole was drilled in the skull using a dental micro drill, causing minimal bleeding during the process. Viral vectors (pAAV-EF1a-double floxed-hChR2(H134R)-EYFP-WPRE-HGHpA, University of Pennsylvania Vector Core Facil- ity) in solution (6.2 x1012 / 1.6×1013 particles/ml) of 0.50–1.00 μl were slowly infused (0.10 μl/ min) into the auditory cortex at two depths (2.00 and 2.50 mm) using a Hamilton 10-μl syringe and an infusion/withdraw pump (World Precision Instruments/KD Scientific). After infusion, the needle was left in place for 1–5 min to allow complete diffusion of the virus. Next, the scalp was rehydrated with saline and sutured with 4–5 stitches and local anesthesia (a drop of Lido- caine/Marcaine) applied onto the sutured skin before the mouse was allowed to wake up. The animal was monitored and kept warm until fully awake (moving and starting to eat and drink water). Mice were killed after approximately 3–6 wk for in vitro electrophysiological experiments and/or histological procedures.

Optical Feedback Inhibition We modified our dynamic clamp system [30] running the Real Time Application Interface for Linux-based (RTAI) from the Politecnico di Milano Institute-Dipartimento di Ingegneria Aerospaziale (Mantegazza, http://www.rtai.org/) on a Dell Precision T1500 with a Quad-Core Intel Core i7 with 2.80 Ghz, 5.8 Gigabyte memory, and a National Instruments DAQ card (NI PCI-6251). Routines for data acquisition were programmed in GNU-C using the Linux Con- trol and Measurement Device Interface (COMEDI). The membrane potentials of the patched cells were acquired in 20 kHz, and APs were detected based on threshold (>−20 mV) in real time, triggering the LED (CoolLED pE-1) via 3-ms (ChR2 experiments) or 500-ms (HaloR experiments) TTL pulses.

Data Analysis Matlab (version 2013a, MathWorks) was used for data analysis. APs from MCsα2 and PCs were detected based on threshold (>−20 mV). PC spike trains were transformed into a series of 0 (no spike) and 1 (spike), with 0.1-ms precision (binning). Accordingly, kernel density estimates are probability densities and were computed on population data (all patched PCs) with 0.1 “bandwidth” (“ksdensity” command in Matlab). The kernel density estimates show the distribution of APs over time and highlight increased (peaks) and decreased (valleys) co-occurrences of spikes. Power spectral density analysis of the binary spike series was made using Welch’s method (“pwelch” command in Matlab) to find the frequency components with highest power. Coherence was calculated pairwise (i.e., for simultaneously patched PCs, the similarity between the binary sequences of the two PCs was calculated) and plotted as mean over all recordings (“cohere” command in Matlab) to investigate the dependence of two cells as a function of frequency. Cross-correlograms were calculated using the “coeff” option of the cross-correlation command in Matlab (to scale the cross-correlation values from −1 to 1 and prevent dependency of the cross-correlation on the number of spikes) and then smoothed by a moving average filter with a span of 10 ms to find the functional dependence between the APs

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 18 / 26 Synchronization by Chrna2-Martinotti Cells

of simultaneously recorded cells over time. We displayed cross-correlations over a lag range of ±1 s. The synchrony index was defined as the maximum peak of the normalized cross-correlograms between −50 ms and 50 ms, as previously described [30], with 0 ≙ no synchronization and 1 ≙ full synchronization. To investigate statistical dependence, mutual information MI (PCleft, PCright) was determined on the binary spike data (binning = 20 ms) by calculating the distributions of the spike trains (univariate distribution of each spike train separately as well as bivariate distribution of both spike trains) and expressing them as entropies H(PCleft) and H(PCright), meaning how “diverse” the spike trains were. The sum of the two entropies H (PCleft) and H(PCright) minus the joint entropy H(PCleft, PCright) quantified the conditional entropy, i.e., the mutual information MI(PCleft, PCright). The MI was formulated as a mutual information index with a scaling factor of 1,000 [30].

Statistical Analysis All statistical analysis was performed using R version 3.2.3 (Foundation for Statistical Comput- ing, Vienna, Austria). Data are reported as mean ± standard error of the mean (SEM) and plotted as bar plots or box plots. Data larger than q3 + 1.5Ã(q3–q1) or smaller than q1–1.5Ã(q3– q1), with q1 and q3 denoting the 25th and 75th percentiles (see box plots), were considered as outlier and discarded. Statistical comparisons were determined using two-tailed Student’s paired t test, and to account for multiple comparisons, the data were analyzed using ANOVA and post-hoc test with Tukey correction (Ã ≙ p < 0.05, ÃÃ ≙ p < 0.01, ÃÃÃ ≙ p < 0.001, and ÃÃÃÃ ≙ p < 0.0001). The order of stimulations of different frequencies (e.g. 2, 5, 15, 25, 40, 70 Hz) was systematically varied to avoid statistical dependencies between the timing of recordings and the frequency investigated.

Supporting Information S1 Fig. Tomato+ cells of Chrna2-Cre/R26tom mice across cortical areas. (A) Left; 4x coronal images (60 μm thick) of primary auditory cortex and (and parts of secondary visual cortex— left image; top). Cell bodies of tdTomato+ neurons (red) appear in layer 5 and dense axonal arborizations are shown in layer 1 (image at approx. bregma -2.46 mm). The star highlights the oriens layer of the hippocampus and arrowhead shows dense axonal projections of oriens lacunosum-moleculare cells [19]. Right; 10x magnification of square area outlined in left. (B) Coronal slices (4x (left) and 10x (right) magnification) of the medial prefrontal cortex where the corpus callosum was not yet joined (around bregma +1.78 mm). (C) Parasagittal slices (10˚ angle), approximately 1.92 mm lateral to the midline (4x (left) and 10x (right) magnification), showing distribution of tdTomato+ cell bodies (red) in the primary somatosensory cortex, primary motor cortex and secondary visual cortex. Red squares show the approximate location of the 4x image (inset), white dashed squares for the 10x images. Note the dense axonal ramifications of Chrna2-Cre/R26tom cells in layer 1 (star). Scale bars = 400 μm (left) and 200 μm (right) resp. (D) Immunohistochemistry for somatostatin (left) in a cortical section from a Chrna2-Cre/R26tom mouse (middle, star highlighting the dense axonal ramifications of Chrna2-Cre/R26tom cells in layer 1) to visualize co-expression with chrna2 (right, arrowheads). Scale bar = 100 μm. (E) A total 792 cells were counted; 297 cells were Chrna2+, 495 were somatostatin+ and 90 of these were double labelled for both Chrna2 and somatostatin (n = 3 mice, 8 sections of 35 μm thickness). Venn diagrams for all layers (layer1-6) and layer 5 visualize the overlap (somatostatin -grey; chrna2 –red and co-expression -pink). Insets in top corner of all panels show mouse brain atlas schematics of area show. (F) Electrophoresis gel image from the single cell analysis showing 6 positive cells (columns 2–4 and 6–8) for GAD1+, 1

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 19 / 26 Synchronization by Chrna2-Martinotti Cells

negative cell (column 5) and the negative control (column 1). (TIF) S2 Fig. Chrna2-Cre/R26tom cells show typical Martinotti cell morphology and putative interconnections with PCs. (A) Example of biocytin-filled (green) Chrna2-Cre/R26tom cell highlighting the long axonal projection to layer 1 (left, à) emerging from the main dendrite (right, circle). Note thick main trunks of dendrites of other red Chrna2-Cre/R26tom cell in the vicinity also pointing in the direction of layer 1. (B) Overview of the long axonal projection (à) of a biocytin filled (green) Chrna2-Cre/R26tom cell, showing proximal axonal arborizations (à) with main axons extending to layer 1. Note the dense axonal ramifications in layer 1 (star). (C) i) High magnification image (63x) of layer 1 (showing biocytin-filled (green) projections from one filled thick-tufted PC and a MCα2 cell, also green-yellow. The thin green-yellow MCα2 axon (highlighted with à) could be followed visually and the high magnification image shows that it passes in close proximity to the thick dendrite of the PC, which was shown to be synaptically coupled with the recorded MCα2. The image is a collapsed z-stack composed of 40 (1 μm sections). ii) Close-up of the image in (i) but only showing collapsed z-stack of 10 images, to give a higher resolution, and still provide a pseudo 3D image of putative connections between the thin axon of the MCα2 and the thick dendrite of the PC. iii) Image showing the corresponding cell bodies of the PC and MCα2 (yellow) in the images on the left. Note also putative connections (arrow) from the PC to the red (not patched) chrna2-Cre/R26tom cell in the lower part of the image. Scale bars = 20 μm. (TIF) S3 Fig. MCsα2 are consistently activated by short duration blue light pulses and accommodating during continuous blue light stimulation. (A). Comparison of evoked IPSPs in type A PCs following (left) action potentials generated by brief current injection (50 pA, 3 ms) in a connected MCα2, and (right) brief light stimulation (488 nm, 3 ms) of a population of ChR2+ MCα2. Graphs show comparison between amplitude, time to peak and half decay time of electrically (white) and optogenetically (blue) evoked IPSPs (n = 12 cells, n = 54 IPSPs with outliers removed, see methods). Values are shown in S2 Data. All comparisons: ÃÃÃ p < 0.001, mean ± SEM, two-tailed Student’s paired t-test. (B) Continuous blue light stimulation (500 ms) generates adaptation in MCα2 firing. (C) Continuous blue light of 1000 ms fails to generate prolonged firing in ChR2-expressing MCsα2. (D) Increasing light intensities (500 ms continuous blue light between 0.5 and 6 mW) does not improve spike capability of ChR2-expressing MCsα2. (TIF) S4 Fig. Properties of HaloR+ MCsα2 rebound spikes. (A) Rebound APs of HaloR+ MCsα2 following different durations (100, 250, 500 and 1000 ms) of continuous green light (12 repetitions)

in the presence of carbachol (Vm = -48 mV). Note that 100 ms of green light fails to consistently evoke rebound APs (failures highlighted by arrow), while 500ms light was sufficient to generate a burst of rebound spikes. (B) Bar graphs show quantifications of number of rebound spikes, the peak hyperpolarization amplitude, rebound maximum frequency, rebound duration and time to peak of hyperpolarization. A 500 ms light stimulation was necessary to generate a burst of rebound spikes (>1), the hyperpolarization amplitude (during light) reached a plateau of -26.66 ± 0.17 mV (top right) (n = 12 cells, mean ± SEM, ANOVA). 1000 ms light does not increase maximum frequency of rebound APs but increases the rebound duration (bottom left) (n = 12 cells, mean ± SEM, two-tailed Student’s paired t-test). Quantification of hyperpolarization time to peak shows that there is no difference 250ms of light stimulation. All comparisons: Ã p < 0.05, ÃÃ p < 0.01, ÃÃÃ p < 0.001 and ÃÃÃÃ p < 0.0001. (C) MCα2 rebound APs generated by either negative current steps (0–100 pA) or by 500 ms green light are both resistant

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 20 / 26 Synchronization by Chrna2-Martinotti Cells

to ZD7288 (20 μM), a blocker of the hyperpolarization-activated cation current (Ih). (TIF) S5 Fig. Comparison of depolarization induced and carbachol induced firing of type A PCs and MCsα2. Power spectral density plots (95% confidence interval in grey, mean in black) of type A PCs (top) and MCsα2 (bottom) are shown in response to a continuous (+40 pA) current injection (left) and following (right) carbachol (10 μM) bath application. Continuously adding carbachol to the perfusate increased the spontaneous firing frequency of both cells, with a broad peak (therefore not at any specific frequency) in the power spectrum at 4.90 Hz (range: 0.62 to 9.39 Hz, n = 60 cells) for type A PCs and 16.02 Hz (range: 9.16 to 23.65 Hz, n = 24 cells) for MCsα2. (TIF) S6 Fig. Characteristics of compound IPSPs generated by a burst of APs of HaloR-expressing MCsα2. (A) Top; Schematic of circuit. Green light stimulation (500 or 1000 ms) hyperpolarizes HaloR-expressing MCsα2 and upon termination of light the MCsα2 fired a burst of rebound APs. The corresponding compound IPSP in type A PCs represent the response to the first 3–4 MCsα2 spikes. Grey dashed lines highlight nicks in the trace where individual IPSPs are summed. (B) Example traces of type A PC IPSPs (grey traces, mean response in black) evoked by rebound APs following silencing a population of HaloR-expressing MCsα2 by green light of different duration (left: 500 ms; right: 1000 ms) for burst-spiking and (top) single-spiking (bottom) type A PCs. (C) IPSP amplitudes (left), time to peak (middle) and half decay time (right) in single-spiking and burst-spiking type A PCs vary depending on the light-duration (i.e. time of MCα2 inhibition and the subsequent rebound APs; n = 12 IPSPs, mean ± SEM, ANOVA). All comparisons: Ã p < 0.05, ÃÃ p < 0.01, ÃÃÃ p < 0.001 and ÃÃÃÃ p < 0.0001. Val- ues are shown in S6 Data. (TIF) S7 Fig. Stimulation frequencies for MCsα2 that do not lead to synchronization of type A PCs. Population response of 12 dual recordings of unconnected type A PCs (left, n = 24 cells; 12 black and 12 grey PC spike trains) during different frequencies of pulsed light stimulation (5 Hz, 25 Hz, 40 Hz and 70 Hz) of ChR2-expressing MCsα2 and without light stimulation (second half of spike train), to compare if MC activity could synchronize PC firing. Kernel density estimates (orange trace) show increased (peaks) and decreased (valleys) co-occurances of APs. Mean power spectral density plots for each tested frequency (right) revealed no particular peak that indicated increased synchronization for 5, 25, 40 or 70 Hz. (TIF) S8 Fig. Firing pattern of depolarized layer 5 PCs change from bursting to non-bursting upon inhibition by MCsα2 at 15 Hz. (A) Inset: Experimental set up and indication of how the PC firing pattern is altered/not altered by 15Hz MCsα2 activity. Voltage traces of a single-spik-

ing (top) and a burst-spiking (bottom) type A PC at Vm = -48 mV (bath application of carbachol). Frames highlight typical APs. Bottom orange box; Note the change from doublet-spiking to single-spiking shortly after the initiation of MCsα2 inhibition at 15 Hz (by activating ChR2-expressing MCsα2 as indicated by blue dots above traces). (B) Voltage traces in (A) but

at Vm = -53 mV shows that the change of firing pattern only happens at depolarized potentials, probably due to the stronger inhibition (larger Cl- drive at depolarized potentials). Frames highlight examples of APs at a zoomed in timescale. (TIF) S1 Data. Data for Fig 1G and 1H. (XLSX)

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 21 / 26 Synchronization by Chrna2-Martinotti Cells

S2 Data. Data for S3A Fig. (XLSX) S3 Data. Data for Fig 3D and 3E. (XLSX) S4 Data. Data for Fig 4A, 4B, 4D and 4E. (XLSX) S5 Data. Data for S4B Fig. (XLSX) S6 Data. Data for S6C Fig. (XLSX) S7 Data. Data for Fig 5A and 5B. (XLSX) S8 Data. Data for Fig 6C. (XLSX) S9 Data. Data for Fig 7B. (XLSX) S1 Movie. Tomato+ cells of Chrna2-Cre/R26tom mice visualized across cortical areas. A series of images from adult (2 months old) Chrna2-Cre/R26tom mouse cortex (coronal slice, 1300 μm thickness) after CLARITY processing is shown. Please note the second band of tomato+ cells highlighted in the stratum oriens of hippocampus [19] and the dense axonal arborisation in stratum lacunosum-moleculare, highlighted as a grey dense mass. (MP4) S1 Text. Supporting Information. (DOCX)

Acknowledgments We thank H. Munguba for comments on earlier versions of this manuscript and S. Perry for technical assistance.

Author Contributions Conceptualization: Markus M. Hilscher. Data curation: Katarina E. Leão. Formal analysis: Markus M. Hilscher. Funding acquisition: Richardson N. Leão, Katarina E. Leão, Klas Kullander. Investigation: Markus M. Hilscher. Methodology: Markus M. Hilscher. Project administration: Markus M. Hilscher, Katarina E. Leão. Resources: Richardson N. Leão, Klas Kullander. Software: Richardson N. Leão, Klas Kullander. Supervision: Katarina E. Leão, Klas Kullander.

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 22 / 26 Synchronization by Chrna2-Martinotti Cells

Validation: Katarina E. Leão. Visualization: Markus M. Hilscher, Steven J. Edwards. Writing – original draft: Markus M. Hilscher, Katarina E. Leão. Writing – review & editing: Markus M. Hilscher, Richardson N. Leão, Katarina E. Leão, Klas Kullander.

References 1. Markram H, Toledo-Rodriguez M, Wang Y, Gupta A, Silberberg G, Wu C. Interneurons of the neocortical inhibitory system. Nat Rev Neurosci. 2004; 5: 793±807. doi: 10.1038/nrn1519 PMID: 15378039 2. Wang Y, Toledo-Rodriguez M, Gupta A, Wu C, Silberberg G, Luo J, et al. Anatomical, physiological and molecular properties of Martinotti cells in the somatosensory cortex of the juvenile rat. J Physiol. 2004; 561: 65±90. doi: 10.1113/jphysiol.2004.073353 PMID: 15331670 3. Goldberg JH, Lacefield CO, Yuste R. Global dendritic calcium spikes in mouse layer 5 low threshold spiking interneurones: implications for control of pyramidal cell bursting. J Physiol. 2004; 558: 465±478. doi: 10.1113/jphysiol.2004.064519 PMID: 15146046 4. Rudy B, Fishell G, Lee S, Hjerling-Leffler J. Three groups of interneurons account for nearly 100% of neocortical GABAergic neurons. Dev Neurobiol. 2011; 71: 45±61. doi: 10.1002/dneu.20853 PMID: 21154909 5. Huang ZJ. Toward a genetic dissection of cortical circuits in the mouse. Neuron. 2014; 83: 1284±1302. doi: 10.1016/j.neuron.2014.08.041 PMID: 25233312 6. Taniguchi H. Genetic dissection of GABAergic neural circuits in mouse neocortex. Front Cell Neurosci. 2014; 8: 8. doi: 10.3389/fncel.2014.00008 PMID: 24478631 7. Kawaguchi Y, Kubota Y. GABAergic cell subtypes and their synaptic connections in rat frontal cortex. Cereb Cortex N Y N 1991. 1997; 7: 476±486. 8. Oliva AA, Jiang M, Lam T, Smith KL, Swann JW. Novel hippocampal interneuronal subtypes identified using transgenic mice that express green fluorescent protein in GABAergic interneurons. J Neurosci Off J Soc Neurosci. 2000; 20: 3354±3368. 9. Ma Y, Hu H, Berrebi AS, Mathers PH, Agmon A. Distinct Subtypes of Somatostatin-Containing Neocor- tical Interneurons Revealed in Transgenic Mice. J Neurosci Off J Soc Neurosci. 2006; 26: 5069±5082. 10. Xu H, Jeong H-Y, Tremblay R, Rudy B. Neocortical somatostatin-expressing GABAergic interneurons disinhibit the thalamorecipient layer 4. Neuron. 2013; 77: 155±167. doi: 10.1016/j.neuron.2012.11.004 PMID: 23312523 11. Lee JH, Whittington MA, Kopell NJ. Top-Down Beta Rhythms Support Selective Attention via Interlami- nar Interaction: A Model. PLoS Comput Biol. 2013; 9: e1003164. doi: 10.1371/journal.pcbi.1003164 PMID: 23950699 12. Karnani MM, Agetsuma M, Yuste R. A blanket of inhibition: functional inferences from dense inhibitory connectivity. Curr Opin Neurobiol. 2014; 26: 96±102. doi: 10.1016/j.conb.2013.12.015 PMID: 24440415 13. Silberberg G, Markram H. Disynaptic inhibition between neocortical pyramidal cells mediated by Marti- notti cells. Neuron. 2007; 53: 735±746. doi: 10.1016/j.neuron.2007.02.012 PMID: 17329212 14. Kapfer C, Glickfeld LL, Atallah BV, Scanziani M. Supralinear increase of recurrent inhibition during sparse activity in the somatosensory cortex. Nat Neurosci. 2007; 10: 743±753. doi: 10.1038/nn1909 PMID: 17515899 15. Berger TK, Silberberg G, Perin R, Markram H. Brief bursts self-inhibit and correlate the pyramidal network. PLoS Biol. 2010; 8. 16. Okun M, Lampl I. Instantaneous correlation of excitation and inhibition during ongoing and sensory- evoked activities. Nat Neurosci. 2008; 11: 535±537. doi: 10.1038/nn.2105 PMID: 18376400 17. Sanchez-Vives MV, McCormick DA. Cellular and network mechanisms of rhythmic recurrent activity in neocortex. Nat Neurosci. 2000; 3: 1027±1034. doi: 10.1038/79848 PMID: 11017176 18. Li X, Morita K, Robinson HPC, Small M. Control of layer 5 pyramidal cell spiking by oscillatory inhibition in the distal apical dendrites: a computational modeling study. J Neurophysiol. 2013; 109: 2739±2756. doi: 10.1152/jn.00397.2012 PMID: 23486202

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 23 / 26 Synchronization by Chrna2-Martinotti Cells

19. Leão RN, Mikulovic S, Leão KE, Munguba H, Gezelius H, Enjin A, et al. OLM interneurons differentially modulate CA3 and entorhinal inputs to hippocampal CA1 neurons. Nat Neurosci. 2012; 15: 1524±1530. doi: 10.1038/nn.3235 PMID: 23042082 20. Perry S, Gezelius H, Larhammar M, Hilscher MM, Lamotte d'Incamps B, Leao KE, et al. Firing properties of Renshaw cells defined by Chrna2 are modulated by hyperpolarizing and small conductance ion currents Ih and ISK. Eur J Neurosci. 2015; 41: 889±900. doi: 10.1111/ejn.12852 PMID: 25712471 21. Fino E, Yuste R. Dense inhibitory connectivity in neocortex. Neuron. 2011; 69: 1188±1203. doi: 10. 1016/j.neuron.2011.02.025 PMID: 21435562 22. BeÂ J-VL, Silberberg G, Wang Y, Markram H. Morphological, Electrophysiological, and Synaptic Proper- ties of Corticocallosal Pyramidal Cells in the Neonatal Rat Neocortex. Cereb Cortex. 2007; 17: 2204± 2213. doi: 10.1093/cercor/bhl127 PMID: 17124287 23. Lee AT, Gee SM, Vogt D, Patel T, Rubenstein JL, Sohal VS. Pyramidal neurons in prefrontal cortex receive subtype-specific forms of excitation and inhibition. Neuron. 2014; 81: 61±68. doi: 10.1016/j. neuron.2013.10.031 PMID: 24361076 24. Berger TK, Perin R, Silberberg G, Markram H. Frequency-dependent disynaptic inhibition in the pyramidal network: a ubiquitous pathway in the developing rat neocortex. J Physiol. 2009; 587: 5411±5425. doi: 10.1113/jphysiol.2009.176552 PMID: 19770187 25. Connors BW, Gutnick MJ. Intrinsic firing patterns of diverse neocortical neurons. Trends Neurosci. 1990; 13: 99±104. PMID: 1691879 26. Brown SP, Hestrin S. Intracortical circuits of pyramidal neurons reflect their long-range axonal targets. Nature. 2009; 457: 1133±1136. doi: 10.1038/nature07658 PMID: 19151698 27. Harris KD, Shepherd GMG. The neocortical circuit: themes and variations. Nat Neurosci. 2015; 18: 170±181. doi: 10.1038/nn.3917 PMID: 25622573 28. Kim EJ, Juavinett AL, Kyubwa EM, Jacobs MW, Callaway EM. Three Types of Cortical Layer 5 Neurons That Differ in Brain-wide Connectivity and Function. Neuron. 2015; 88: 1253±1267. doi: 10.1016/j. neuron.2015.11.002 PMID: 26671462 29. Kopell N, Whittington MA, Kramer MA. Neuronal assembly dynamics in the beta1 frequency range permits short-term memory. Proc Natl Acad Sci U S A. 2011; 108: 3779±3784. doi: 10.1073/pnas. 1019676108 PMID: 21321198 30. Hilscher MM, Leão KE, Leão RN. Synchronization through nonreciprocal connections in a hybrid hippocampus microcircuit. Front Neural Circuits. 2013; 7: 120. doi: 10.3389/fncir.2013.00120 PMID: 23888129 31. Sohal VS, Zhang F, Yizhar O, Deisseroth K. Parvalbumin neurons and gamma rhythms enhance cortical circuit performance. Nature. 2009; 459: 698±702. doi: 10.1038/nature07991 PMID: 19396159 32. Ramaswamy S, Markram H. Anatomy and physiology of the thick-tufted layer 5 pyramidal neuron. Front Cell Neurosci. 2015; 9: 233. doi: 10.3389/fncel.2015.00233 PMID: 26167146 33. Mainen ZF, Sejnowski TJ. Influence of dendritic structure on firing pattern in model neocortical neurons. Nature. 1996; 382: 363±366. doi: 10.1038/382363a0 PMID: 8684467 34. Murayama M, PeÂrez-Garci E, Nevian T, Bock T, Senn W, Larkum ME. Dendritic encoding of sensory stimuli controlled by deep cortical interneurons. Nature. 2009; 457: 1137±1141. doi: 10.1038/nature07663 PMID: 19151696 35. Karnani MM, Jackson J, Ayzenshtat I, Hamzehei Sichani A, Manoocheri K, Kim S, et al. Opening Holes in the Blanket of Inhibition: Localized Lateral Disinhibition by VIP Interneurons. J Neurosci Off J Soc Neurosci. 2016; 36: 3471±3480. 36. Karnani MM, Jackson J, Ayzenshtat I, Tucciarone J, Manoocheri K, Snider WG, et al. Cooperative Sub- networks of Molecularly Similar Interneurons in Mouse Neocortex. Neuron. 2016; 90: 86±100. doi: 10. 1016/j.neuron.2016.02.037 PMID: 27021171 37. Tsuno Y, Chapman GW, Hasselmo ME. Rebound spiking properties of mouse medial entorhinal cortex neurons in vivo. Eur J Neurosci. 2015; 42: 2974±2984. doi: 10.1111/ejn.13097 PMID: 26454151 38. Ferrante M, Shay CF, Tsuno Y, Chapman GW, Hasselmo ME. Post-Inhibitory Rebound Spikes in Rat Medial Entorhinal Layer II/III Principal Cells: In Vivo, In Vitro, and Computational Modeling Characteri- zation. Cereb Cortex. 2016; bhw058. doi: 10.1093/cercor/bhw058 PMID: 26965902 39. Lee JH, Whittington MA, Kopell NJ. Potential Mechanisms Underlying Intercortical Signal Regulation via Cholinergic Neuromodulators. J Neurosci. 2015; 35: 15000±15014. doi: 10.1523/JNEUROSCI. 0629-15.2015 PMID: 26558772 40. Roopun AK, LeBeau FEN, Rammell J, Cunningham MO, Traub RD, Whittington MA. Cholinergic neuromodulation controls directed temporal communication in neocortex in vitro. Front Neural Circuits. 2010; 4: 8. doi: 10.3389/fncir.2010.00008 PMID: 20407636

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 24 / 26 Synchronization by Chrna2-Martinotti Cells

41. Lacey MG, Gooding-Williams G, Prokic EJ, Yamawaki N, Hall SD, Stanford IM, et al. Spike Firing and IPSPs in Layer V Pyramidal Neurons during Beta Oscillations in Rat Primary Motor Cortex (M1) In Vitro. PLoS ONE. 2014; 9: e85109. doi: 10.1371/journal.pone.0085109 PMID: 24465488 42. Fanselow EE, Richardson KA, Connors BW. Selective, State-Dependent Activation of Somatostatin- Expressing Inhibitory Interneurons in Mouse Neocortex. J Neurophysiol. 2008; 100: 2640±2652. doi: 10.1152/jn.90691.2008 PMID: 18799598 43. Gibson JR, Beierlein M, Connors BW. Two networks of electrically coupled inhibitory neurons in neocortex. Nature. 1999; 402: 75±79. doi: 10.1038/47035 PMID: 10573419 44. Ma Y, Hioki H, Konno M, Pan S, Nakamura H, Nakamura KC, et al. Expression of Gap Junction Protein Connexin36 in Multiple Subtypes of GABAergic Neurons in Adult Rat Somatosensory Cortex. Cereb Cortex. 2011; 21: 2639±2649. doi: 10.1093/cercor/bhr051 PMID: 21467210 45. Taniguchi H, He M, Wu P, Kim S, Paik R, Sugino K, et al. A Resource of Cre Driver Lines for Genetic Targeting of GABAergic Neurons in Cerebral Cortex. Neuron. 2011; 71: 995±1013. doi: 10.1016/j. neuron.2011.07.026 PMID: 21943598 46. Hu H, Cavendish JZ, Agmon A. Not all that glitters is gold: off-target recombination in the somatostatin- IRES-Cre mouse line labels a subset of fast-spiking interneurons. Front Neural Circuits. 2013; 7: 195. doi: 10.3389/fncir.2013.00195 PMID: 24339803 47. Christophe E, Roebuck A, Staiger JF, Lavery DJ, Charpak S, Audinat E. Two types of nicotinic receptors mediate an excitation of neocortical layer I interneurons. J Neurophysiol. 2002; 88: 1318±1327. PMID: 12205153 48. Gulledge AT, Park SB, Kawaguchi Y, Stuart GJ. Heterogeneity of phasic cholinergic signaling in neocortical neurons. J Neurophysiol. 2007; 97: 2215±2229. doi: 10.1152/jn.00493.2006 PMID: 17122323 49. Arroyo S, Bennett C, Aziz D, Brown SP, Hestrin S. Prolonged disynaptic inhibition in the cortex mediated by slow, non-α7 nicotinic excitation of a specific subset of cortical interneurons. J Neurosci Off J Soc Neurosci. 2012; 32: 3859±3864. 50. Eckenstein FP, Baughman RW, Quinn J. An anatomical study of cholinergic innervation in rat cerebral cortex. Neuroscience. 1988; 25: 457±474. PMID: 2456488 51. Couey JJ, Meredith RM, Spijker S, Poorthuis RB, Smit AB, Brussaard AB, et al. Distributed Network Actions by Nicotine Increase the Threshold for Spike-Timing-Dependent Plasticity in Prefrontal Cortex. Neuron. 2007; 54: 73±87. doi: 10.1016/j.neuron.2007.03.006 PMID: 17408579 52. von Engelhardt J, Eliava M, Meyer AH, Rozov A, Monyer H. Functional Characterization of Intrinsic Cholinergic Interneurons in the Cortex. J Neurosci. 2007; 27: 5633±5642. doi: 10.1523/JNEUROSCI. 4647-06.2007 PMID: 17522308 53. Xiang Z, Huguenard JR, Prince DA. Cholinergic Switching Within Neocortical Inhibitory Networks. Sci- ence. 1998; 281: 985±988. PMID: 9703513 54. McGehee DS. Molecular diversity of neuronal nicotinic acetylcholine receptors. Ann N Y Acad Sci. 1999; 868: 565±577. PMID: 10414338 55. Lein ES, Hawrylycz MJ, Ao N, Ayres M, Bensinger A, Bernard A, et al. Genome-wide atlas of gene expression in the adult mouse brain. Nature. 2007; 445: 168±176. doi: 10.1038/nature05453 PMID: 17151600 56. Walsh H, Govind AP, Mastro R, Hoda JC, Bertrand D, Vallejo Y, et al. Up-regulation of nicotinic receptors by nicotine varies with receptor subtype. J Biol Chem. 2008; 283: 6022±6032. doi: 10.1074/jbc. M703432200 PMID: 18174175 57. Lamotte d'Incamps B, Ascher P. Subunit composition and kinetics of the Renshaw cell heteromeric nicotinic receptors. Biochem Pharmacol. 2013; 86: 1114±1121. doi: 10.1016/j.bcp.2013.06.017 PMID: 23811311 58. Lotfipour S, Byun JS, Leach P, Fowler CD, Murphy NP, Kenny PJ, et al. Targeted deletion of the mouse α2 nicotinic acetylcholine receptor subunit gene (Chrna2) potentiates nicotine-modulated behaviors. J Neurosci Off J Soc Neurosci. 2013; 33: 7728±7741. 59. Nakauchi S, Brennan RJ, Boulter J, Sumikawa K. Nicotine gates long-term potentiation in the hippocampal CA1 region via the activation of alpha2* nicotinic ACh receptors. Eur J Neurosci. 2007; 25: 2666±2681. doi: 10.1111/j.1460-9568.2007.05513.x PMID: 17466021 60. Fanselow EE, Connors BW. The roles of somatostatin-expressing (GIN) and fast-spiking inhibitory interneurons in UP-DOWN states of mouse neocortex. J Neurophysiol. 2010; 104: 596±606. doi: 10. 1152/jn.00206.2010 PMID: 20538767 61. Urban-Ciecko J, Fanselow EE, Barth AL. Neocortical somatostatin neurons reversibly silence excitatory transmission via GABAb receptors. Curr Biol CB. 2015; 25: 722±731. doi: 10.1016/j.cub.2015.01.035 PMID: 25728691

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 25 / 26 Synchronization by Chrna2-Martinotti Cells

62. Sigalas C, Rigas P, Tsakanikas P, Skaliora I. High-Affinity Nicotinic Receptors Modulate Spontaneous Cortical Up States In Vitro. J Neurosci. 2015; 35: 11196±11208. doi: 10.1523/JNEUROSCI.5222-14. 2015 PMID: 26269630 63. Tomer R, Ye L, Hsueh B, Deisseroth K. Advanced CLARITY for rapid and high-resolution imaging of intact tissues. Nat Protoc. 2014; 9: 1682±1697. doi: 10.1038/nprot.2014.123 PMID: 24945384

PLOS Biology | DOI:10.1371/journal.pbio.2001392 February 9, 2017 26 / 26 Supporting Information Chrna2-Martinotti Cells Synchronize layer 5 type A Pyramidal Cells via Rebound Excitation Markus M. Hilscher, Richardson N. Leão, Steven J. Edwards, Katarina E. Leão, Klas Kullander

S1 Fig.

S2 Fig.

S3 Fig.

S4 Fig.

S5 Fig.

S6 Fig.

S7 Fig.

S8 Fig.

S1 Movie.