Sorcs2 Is Required for BDNF-Dependent Plasticity in the Hippocampus

ORIGINAL ARTICLE SorCS2 is required for BDNF-dependent plasticity in the hippocampus

S Glerup1,2,9, U Bolcho1,2,9, S Mølgaard1,2, S Bøggild1,2, CB Vaegter1,2, AH Smith3,4, JL Nieto-Gonzalez2, PL Ovesen1,2, LF Pedersen1,2, AN Fjorback2, M Kjolby1,2, H Login1,2, MM Holm2, OM Andersen1,2, JR Nyengaard5, TE Willnow6, K Jensen7 and A Nykjaer1,2,8

SorCS2 is a member of the Vps10p-domain receptor gene family receptors with critical roles in the control of neuronal viability and function. Several genetic studies have suggested SORCS2 to confer risk of bipolar disorder, schizophrenia and attention deﬁcit- hyperactivity disorder. Here we report that hippocampal N-methyl-D-aspartate receptor-dependent synaptic plasticity is eliminated in SorCS2-deﬁcient mice. This defect was traced to the ability of SorCS2 to form complexes with the neurotrophin receptor p75NTR, required for pro-brain-derived neurotrophic factor (BDNF) to induce long-term depression, and with the BDNF receptor tyrosine kinase TrkB to elicit long-term potentiation. Although the interaction with p75NTR was static, SorCS2 bound to TrkB in an activity- dependent manner to facilitate its translocation to postsynaptic densities for synaptic tagging and maintenance of synaptic potentiation. Neurons lacking SorCS2 failed to respond to BDNF by TrkB autophosphorylation, and activation of downstream signaling cascades, impacting neurite outgrowth and spine formation. Accordingly, Sorcs2–/– mice displayed impaired formation of long-term memory, increased risk taking and stimulus seeking behavior, enhanced susceptibility to stress and impaired prepulse inhibition. Our results identify SorCS2 as an indispensable coreceptor for p75NTR and TrkB in hippocampal neurons and suggest SORCS2 as the link between proBDNF/BDNF signaling and mental disorders.

Molecular Psychiatry (2016) 21, 1740–1751; doi:10.1038/mp.2016.108; published online 26 July 2016

INTRODUCTION proBDNF, and the subsequent conversion of proBNDF into mature Mood disorders are common, severe, chronic, often life- BDNF by extracellular proteinases, is required for the late phase 6 threatening illnesses, and are frequently accompanied by of LTP (L-LTP) and synapse consolidation. In contrast, weakening comorbidities including substance abuse.1–3 The Global Burden of synapses is facilitated by a different mechanism, involving the interaction of proBDNF with the postsynaptic p75NTR, which of Disease Study conducted by the World Health Organization 7 identified schizophrenia, major depressive disorder and bipolar results in LTD. disorder as being among the leading causes of disability world- The molecular mechanisms underlying synaptic plasticity are wide, and as illnesses likely to take an even greater toll on health, incompletely understood. The so-called synaptic tagging and capture hypothesis8 has been proposed to explain how some society and the economy in the coming years.4 Impairment in synapses are strengthened while others are weakened. According synaptic plasticity is critical to the pathophysiology of mood to this model, high-frequency neuronal stimulation induces disorders, and current treatments such as antidepressants and synthesis of plasticity-related proteins that are then captured lithium exert effects on the pathways that regulate synaptic specifically at selected synapses by a ‘tag’ inserted into the plasticity, in part, by increasing signaling by neurotrophic factors, 2,3 postsynaptic membrane. BDNF/proBDNF is a plasticity-related especially brain-derived neurotrophic factor (BDNF). protein, and TrkB was recently described as having properties — Synaptic plasticity the ability of a synapse to change in that could enable it to serve as a synaptic tag. Neuronal activity — strength is considered the cellular basis of most cognitive and elevated cyclic AMP results in the recruitment of TrkB into processes, including memory formation. Strengthening or weak- postsynaptic densities (PSDs), thereby allowing BDNF signaling ening of synapses occurs via long-term potentiation (LTP) and and consequent L-LTP at stimulated synapses.9–11 However, the long-term depression (LTD), respectively. Glutamatergic CA1–3 molecular mechanism by which TrkB is trafficked into the PSD synapses in the hippocampus can be used as a system to study remains unknown. synaptic plasticity in depth. At the CA1–CA3 junction, neuronal We and others recently established the receptor SorCS2 as activity induces presynaptic release of BDNF, and the subsequent a p75NTR coreceptor required for binding of proBDNF and interaction of BNDF with the postsynaptic tyrosine kinase receptor growth cone collapse of, for example, developing dopaminergic TrkB is required for the induction of the early phase of LTP neurons.12,13 SorCS2 belongs to the Vps10p-domain/sortilin family (E-LTP).5 In addition, the release of the precursor of BDNF, of sorting and signaling receptors, a protein family that also

1The Lundbeck Foundation Research Center MIND, Danish Research Institute of Translational Neuroscience DANDRITE— Nordic EMBL Partnership for Molecular Medicine, Aarhus C, Denmark; 2Department of Biomedicine, Aarhus University, Aarhus C, Denmark; 3Yale School of Medicine, Interdepartmental Neuroscience Program and Medical Scientist Training Program, New Haven, CT, USA; 4Department of Psychiatry, VAT CT Healthcare Center, and Yale School of Medicine, New Haven, CT, USA; 5MIND Center, Stereology and Electron Microscopy Laboratory, Aarhus University, Aarhus C, Denmark; 6Max-Delbrück-Center for Molecular Medicine, Berlin, Germany; 7Institute of Neuroscience and Physiology, The Sahlgrenska Academy, Gothenburg University, Gothenburg, Sweden and 8Department of Neuroscience, Mayo Clinic, Jacksonville, FL, USA. Correspondence: Dr S Glerup or Professor A Nykjaer, MIND Center and DANDRITE, Department of Biomedicine, Aarhus University, Ole Worms Allé, Building 1170, Aarhus 8000C, Denmark. E-mail: [email protected] or [email protected] 9These authors share first authorship. Received 21 August 2015; revised 6 April 2016; accepted 18 April 2016; published online 26 July 2016 Role of SorCS2 in BDNF-dependent plasticity S Glerup et al 1741 includes sortilin, SorLA and SorCS-1 and -3.14,15 All five receptors there was no paired-pulse facilitation when stimuli were separated by are highly expressed in both the developing and adult nervous 50 ms. Slices were perfused at 33 °C in artificial cerebral spinal fluid system. They share a characteristic N-terminal Vps10p domain containing 0.5% bovine serum albumin. Hippocampal slices were perfused 13 that has high homology to Vps10p, a vacuolar protein-sorting with rabbit anti-sorCS2 immunoglobulin G (IgG), rabbit control IgG, cleavage-resistant proBDNF, BDNF or vehicle control during experiments. protein in yeast. In fact, sortilins are capable of Golgi-endosome − fi BDNF were perfused into slices from an initial concentration of 0.5 × 10 4 traf cking, internalization, polarized anterograde transport and − 1 fi − 1 to engage in signaling, suggesting key roles in regulating (2.5 pg ml )toa nal concentration of 0.5 nM (25 ng ml ), with a 10-fold 16–18 increase every 30 min. The evoked fEPSPs recorded in current-clamp mode neuronal function. Interestingly, genome-wide association were analyzed with Clampfit 10 (Molecular Devices). We used a Student's studies have implicated SORCS2 in the etiology of BD and the rate t-test for statistical comparison of mean fEPSP slopes. In LTD experiments, of episodes of polarity,19,20 schizophrenia (SZ)21 and attention 22 we compared fEPSP slopes at 30 min after application of LFS. In LTP deficit-hyperactivity disorder (ADHD), although no damaging experiments, we compared fEPSPs at 30 min after application of tetanus, mutations so far have been identified. The observed pleiotropy is defined as E-LTP expression. We also compared fEPSPs 180 min after consistent with multiple studies reporting a high degree of application of tetanus, defined as L-LTP expression. In LTP summary genetic overlap between BD, SZ and ADHD.23,24 In particular, the graphs, each point represents the average of eight consecutive responses. genetic associations of SORCS2 make it similar to BDNF, as variants 25–31 in BDNF have also been associated with BD, SZ and ADHD. Dendritic complexity While we can understand the BDNF genetic effects in terms of the ’ Intracellular labeling was performed by including 0.5% biocytin (Sigma- protein s known functions, our understanding of SORCS2 and its Aldrich) in the internal pipette solution and facilitated by depolarizing relationship to behavior is far less well developed. We sought to current steps (0.2–1 nA) of 500 ms at 1 Hz for 10–20 min. After injection, explore if faulty proBDNF and/or BDNF signaling may explain the slices were transferred to a solution of 4% paraformaldehyde at 4 °C connection between SORCS2 and psychiatric illness. We find that overnight. Slices were subsequently washed with phosphate-buffered SorCS2 is essential for proBDNF/BDNF-induced synaptic plasticity, saline (PBS) (0.1 M, pH 7.3), and incubated overnight with FITC-conjugated and for activity-dependent recruitment of TrkB into PSDs. In avidin-D (Vector Laboratories, Burlingame, CA, USA) in PBS and 0.3% Triton addition, SorCS2 knockout mice display a behavioral phenotype X-100. Slices were then washed with PBS and embedded with fluorescent that recapitulates key symptoms of serious mental illnesses. mounting medium (Dako, Glostrup, Denmark) and labeled neurons visualized on a confocal laser microscope (LSM710, Zeiss, Jena, Germany). For morphological analysis, injected granule cells with no evidence of MATERIALS AND METHODS somatic membrane damage, minimal extracellular dye leakage and a good contrast with low background labeling of the tissue were included. Each Reagents dendritic branch was assigned a branch order following a centrifugal The Fc-tagged extracellular domain of TrkB in addition to goat NTR ordering method from cell body to terminal branch. anti-TrkB, goat-anti-p75 and goat anti-SorCS2 was from R&D Systems To determine dendritic branching in vitro, hippocampal neurons were (Minneapolis, MN, USA). Anti-synaptophysin and anti-PSD95 monoclonal isolated from P3 mice and grown in culture at a density of 5000 neurons per antibodies were from Millipore (Darmstadt, Germany). Rabbit anti- coverslip. After 24 h, the medium was changed to medium containing either neurofilament 200 was from Sigma Aldrich (St. Louis, MO, USA). Rabbit 1nM of BDNF or same volume of sterile PBS. Hereafter, the cells were left in γ anti-phospho-TrkB directed against the phospholipase C- phosphoryla- the incubator at 37 °C for 72 h before being fixed in 4% paraformaldehyde tion site was from Epitomics (Cambridge, UK). Anti-phosphotyrosine for 20 min at room temperature. To determine cell morphology, the neurons monoclonal antibody was from Millipore. Anti-phosphoTrkA/B directed were stained against β-tubulin (β-tubulin mouse monoclonal antibody against the autophosphorylation site was from Cell Signaling (Danvers, MA, MAB3408; Millipore) and confocal images were obtained. The dendritic USA). Anti-N-methyl-D-aspartate receptor 1 (NMDAR1) (no. 5704) was from complexity was subsequently assessed by image analysis using Imaris Cell Signaling. Purified BDNF and cleavage-resistant proBDNF was from software (Bitplane, Zürich, Switzerland). Alomone (Jerusalem, Israel). The purified extracellular domain of human To determine spine density in vitro, hippocampal neurons were isolated SorCS2 was used to generate rabbit polyclonal anti-SorCS2 antibodies as at P0 and grown in culture at a density of 100 000 neurons per coverslip. well as mouse monoclonal antibodies.13 Further details on our use of After 7 days in vitro, the neurons were transfected with a plasmid research antibodies can be found at http://www.pabmabs.com. containing GFP using Lipofectamine LTX reagent (15338-100, Thermo- Fischer Scientific, Waltham, MA, USA) using the supplier's protocol. At Electrophysiological recordings 13 days in vitro, the medium was changed to medium containing either 1 We prepared coronal hippocampal slices (400 μm) from P18–P22 mice for or 0 nM BDNF. Hereafter, the cells were left in the incubator at 37 °C for LTD experiments and P30–P40 for LTP experiments. In each experiment we 24 h before being fixed in 4% paraformaldehyde for 20 min at room used 5–14 slices from three to eight animals. Slices were maintained in a temperature. To determine spine density confocal images were obtained storage chamber bubbled with 95% O2 and 5% CO2. After a minimum and subsequently assessed using Imaris software. recovery period of at least 2 h, the slices were moved to an interface fi chamber exposed to a humidi ed atmosphere of 95% O2 and 5% CO2 and Animal experiments we recorded field excitatory postsynaptic potential (fEPSP) using a Multiclamp 700B amplifier (Molecular Devices, Sunnyvale, CA, USA) with The SorCS2 knockout mouse has been backcrossed for 10 generations into C57/BL6J. Behavioral studies were carried out with the backcrossed an artificial cerebral spinal fluid (in mM: 126 NaCl, 2.5 KCl, 1.25 NaH2PO4,26 D fi – homozygous mice compared with the same C57/BL6J substrain that was NaHCO3, 2.5 CaCl2, 1.3 MgCl2,10 -glucose)- lled glass microelectrode (10 13 15 MΩ) positioned in the stratum radiatum of hippocampal area CA1. used for backcrossing. All experiments were approved by the Danish Synaptic responses were evoked by stimulating Schaffer collateral with Animal Experiments Inspectorate under the Ministry of Justice (permit 0.1 ms pulses with a concentric bipolar electrode (FHC, Bowdoinham, 2011/561-119) and carried according to institutional and national guide- ME, USA). We applied high-frequency tetanic stimulation (HFS) or low- lines. All animals were bred and housed at the Animal Facility at Aarhus frequency stimulation (LFS) after stable baseline was established. Stimulus University. Animals were housed in groups of up to five mice per plastic 3 intensity was adjusted to evoke fEPSP ~ 40% of the maximum. We induced cage (42 × 25 × 15 cm ) under pathogen-free conditions with a 12-h light/ NMDAR-dependent LTP by two trains of HFS (100 Hz, 1 s, separated by 30 s), 12-h dark schedule and fed standard chow (Altromin, Lage, Germany; no. whereas we induced LTD by either single (for NMDAR-dependent LTD) or 1324) and water ad libitum. Cages were cleaned every week and supplied twin pulses (for NMDAR-independent LTD, 50 ms interstimulus interval) with bedding and nesting material, a wooden stick and a metal tunnel. applied at a rate of 1 Hz for 15 min. ‘Two pathway’ experiments were Behavioral experiments were carried out using 12–16-week-old male mice performed in LTP experiments by placing two stimulation electrodes on during their light cycle between 0900 to 1700 hours. Each of the behavioral both sides of the recording electrode in area CA1. The two stimulation test described below were carried out using naïve animals tested in a electrodes were activated alternately once every 15 s with stimuli randomized order by an investigator blinded to the mouse genotype. No separated by 7.5 s to stimulate two separate synaptic populations. animals were excluded from the subsequent analysis. At the end of the Crosstalk between synaptic populations was minimized by assuring that experiment, animals were killed by cervical dislocation.

© 2016 Macmillan Publishers Limited, part of Springer Nature. Molecular Psychiatry (2016), 1740 – 1751 Role of SorCS2 in BDNF-dependent plasticity S Glerup et al 1742 Memory Statistics The Barnes maze paradigm of spatial learning consisted of a circular Unless otherwise mentioned, two-tailed Student's t-test was used to assess platform (92 cm in diameter) elevated 105 cm above the floor. The maze significance. Error bars indicate s.e.m. contains 20 equally spaced holes (5 cm in diameter; 7.5 cm distance between the individual holes). The centers of all holes are 2 cm away from the perimeter of the maze. A hidden escape box is placed under the target RESULTS hole. To facilitate spatial orientation, visual cues were placed on the walls. Postsynaptic expression of SorCS2 On day 1, mice were allowed to freely explore the maze for 3 min and were fi subsequently placed in the escape box for 2 min. Mice were then placed We rst examined expression of SorCS2 in the adult murine in the center of the maze under a dark circular box for 10 s to ensure brain by western blot analysis and found that the receptor was random orientation at the beginning of the test. A fan was turned on as enriched in the hippocampus and striatum (Figure 1a). In the CA1, reinforcing stimulus during the test phase. Mice were tested for four SorCS2 costained with the neuronal marker NeuN but not with consecutive days and their latency to enter the target hole was measured GFAP that defines glia cells (Figure 1b). In neurons SorCS2 was by video recording and analysis using the Any-Maze tracking software restricted to dendrites as neurites positive for the axonal marker (Stoelting, Dublin, Ireland). The inhibitory avoidance experiment (Gemini NF200 were devoid of SorCS2 immunoreactivity (Figure 1c). Avoidance System; San Diego Instruments, San Diego, CA, USA) was used Subcellular fractionation experiments revealed enrichment of for assessing behavioral tagging and was performed essentially as described in Lu et al.11 and Moncada and Viola.32 The test takes advantage SorCS2 in fractions positive for the PSD marker PSD95 but of the natural preference of mice for the dark, and consists of a brightly lit negative for the presynaptic marker synaptophysin (Figure 1d). room and a dark room separated by a guillotine door. On the training day, Double immunostaining confirmed that as ~ 23% of SorCS2 the mouse is placed in the bright room. When entering the dark room, the colocalized with PSD95 (Supplementary Figures S1a and b), a door closes and the mouse receives an electric shock (0.1 or 0.8 mA for 1 s). finding substantiated by immunoelectron microscopy, which After 1 or 24 h, the mouse is returned to the bright room and the latency showed that gold particles labeling SorCS2 were found in PSDs to enter the dark room indicates memory of the shock. Behavioral tagging (Figure 1e). was achieved by exposing mice to an open field containing four novel objects for 15 min. Mice were subsequently returned to their home cage NTR for 1 h before subjecting them to inhibitory avoidance training. SorCS2 is essential for proBDNF- and p75 -dependent hippocampal LTD NMDA receptor-dependent hippocampal LTD requires proBDNF Anxiety NTR 34 For locomotor activity, mice were tested in an open field test consisting of and p75 . To test whether SorCS2 is also required, we 3 characterized the electrophysiological properties of the CA1 a (40 × 40 × 35 cm ) clear Plexiglas arena. Mice were placed in the corner of –/– 13 the arena and their activity was recorded over a 20 min session. For synapse in receptor-deficient (Sorcs2 ) mice. Schaffer collaterals anxiety-related behavior, mice were tested in an elevated plus maze raised of 20-day-old mice were subjected to LFS and fEPSPs were 40 cm above the floor, and consisting of two opposite enclosed arms with measured over 30 min (Figure 1f). fEPSP slopes decreased to 15 cm high opaque walls and two opposite open arms of the same size 78 ± 1% (P = 0.005) in wild-type mice but receptor-deficient 2 (35 × 5 cm ). The elevated plus maze was set up in a dim lit room under a animals were unresponsive (98 ± 1%). Notably, addition of video camera connected to a computer under the control of the Any-maze proBDNF failed to rescue LTD in the knockout slices (100 ± 4%, tracking system. Testing sessions of 10 min were carried out for each P = 0.70) even though proBDNF was capable of increasing LTD in mouse and measured the number of entries and the time spent in the open arms. The percentage of entries into the open arms was calculated as wild-type slices (Figure 1g and Supplementary Figure S2). follows: open arm entries/(open arm entries+closed arm entries) × 100. SorCS2 is required for induction of growth cone collapse in hippocampal cultures.35 However, lack of LTD in Sorcs2–/– mice was not due to developmental deficits. First, at E15-P5 when Depressive behavior neurons send out projections SorCS2 was not expressed in the Mice were also tested in a forced swim test, a well-described test hippocampus (Supplementary Figures S1c–f). In contrast, it for depressive behavior in rodents. Mice were dropped individually into was abundant at P30 when synaptogenesis predominates36 glass cylinders (height: 30 cm; diameter: 15 cm) containing 20 cm water, maintained at 23–25 °C and remained there for 6 min while their (Supplementary Figure S1f). Second, basal synaptic properties, movements were recorded using a digital video camera. A mouse was determined by input–output and paired-pulse responses (Supple- judged to be immobile when it floated in an upright position and made mentary Figure S3), and metabotropic LTD were unaffected in – – only small movements to keep its head above water. The time of Sorcs2 / mice (78 ± 5% versus 82 ± 4%; P = 0.54) (Figure 1h). immobility was scored during the last 4 min of the 6 min testing period, Third, NMDAR-dependent LTD was restored in the knockouts by after 2 min habituation. Depressive behavior was also assessed in the tail lentiviral-mediated overexpression of SorCS2 in the CA1 (83 ± 2% suspension test, which was carried out by hanging mice by the tails and versus 101 ± 4%; P = 0.001) (Figure 1i and Supplementary scoring their immobility during 6 min using the Any-maze tracking Figure S4). Last, antibodies blocking the interaction between software. p75NTR and SorCS2 abolished LTD in wild-type animals (99 ± 2% versus 80 ± 3%; P = 0.001), but control IgG had no effect (Figure 1j Startle and Supplementary Figures S5a and b). We concluded that SorCS2 The prepulse inhibition (PPI) experiment was performed on the StartFear in complex with p75NTR is required for proBDNF to induce instrument (Panlab, Barcelona, Spain) and initiated by an exploratory NMDAR-dependent LTD in CA1. period of 5 min with background noise (60 db), followed by 10 presentations of the pulse (115 db, 8 kHz, 40 ms) with an intertrial interval of 30 s. The mice then receives 40 trials in pseudorandom order (30 s BDNF-dependent hippocampal LTP is reliant on SorCS2 intertrial interval) consisting of no stimulation, presentation of the pulse Next, we studied NMDAR-dependent LTP, which is reliant on alone, the prepulse alone (80 db, 10 kHz, 20 ms) and trials where the BDNF binding to TrkB.37 LTP was triggered in P35 mice by HFS, prepulse precedes the pulse by an interstimulus interval of 100 ms. PPI was and fEPSP was monitored. HFS evoked robust E-LTP at 30 min in − calculated as follows: %PPI = 100 (startle response for prepulse+pulse)/ wild-type mice (157 ± 6%), but potentiation was substantially (startle response for pulse alone) × 100. Experiments measuring startle –/– response to sound pulse was performed as described in Fadok et al.33 decreased in the Sorcs2 mutants (125 ± 3%, P = 0.0006) using the StartFear apparatus from Panlab. Restrained mice were first (Figure 2a). In addition to reduced E-LTP, potentiation further habituate to the apparatus for 5 min and then exposed to seven series of declined over time, reaching baseline after 180 min (104 ± 5% sound pulses (0, 80, 90, 100, 105, 110 and 120 dB) presented in a random versus 141 ± 6% for wild-type animals; P = 0.0001), demonstrating order with an intertrial interval of 30 s. that L-LTP is also reliant on SorCS2 (Figure 2a). In agreement,

Molecular Psychiatry (2016), 1740 – 1751 © 2016 Macmillan Publishers Limited, part of Springer Nature. 1743 – / – 1751 )LTD f – Sorcs2 b ) Western / a _ _ b a Sorcs2 ) Inhibition of LTD in j Time (min) Time a mGluR-dependent b m a WT µ 10 SorCS2 NF200 0 10203040506070 0 80 60 40 20 (DIV)) stained for NF200 (red) and SorCS2 ) Normal mGluR-dependent LTD in 140 120 100

h fEPSP slope (%) slope fEPSP 0.001), but not LacZ. ( vesicle preparation (SVP) and synaptic plasma -dependent plasticity = P in vitro BDNF b b synaptic / _ _ a et al a b b WT anti-SorCS2 Sorcs2 proBDNF 0.001). ) Immuno-EM gold labeling of SoCS2 in the hippocampus. Particles = SorCS2 GFAP e Role of SorCS2 in S Glerup P / Time (min) Time _ _ m Time (min) Time µ a uorescence microscopy of wild-type hippocampal slices for SorCS2 (green) a a CA1 b ﬂ a b WT control IgG 100 Sorcs2 vehicle 4 ng/ml proBDNF 10 µg/ml control IgG -aspartate receptor (NMDAR)-dependent long-term depression (LTD). ( 10 µg/ml anti-SorCS2 IgG D 0 10203040506070 ) Hippocampal neurons (14 days 0 10203040506070 c 0 0

80 60 40 20 40 20 80 60

140 120 100 140 120 100

fEPSP slope (%) slope fEPSP fEPSP slope (%) slope fEPSP -methyl- N ) Double immuno b SorCS2 NeuN

PSD2 0.005). Sample traces are shown at indicated time points labeled 'a' and 'b'. Scale bars, 5 mV and b b = / / P Spinal cord PSD1 _ _ _ _ mice by pro-brain-derived neurotrophic factor (pro BDNF). ( a – b

Striatum SPM / a – b Sorcs2

SVP , mice ( Sorcs2 lenti lacZ

Cerebellum – / –

Midbrain Sorcs2 / Time (min) Time Time (min) Time Sorcs2 Cortex _ _ a PSD95 - a a SorCS2 - b a

Hippocampus NMDAR-dependent b WT Sorcs2 lenti SorCS2 Synaptophysin - 0 10203040506070 0 10203040506070 actin - ) Subcellular fractionation of hippocampal extracts followed by immunoblotting for SorCS2, postsynaptic density 95 (PSD95) and 0 0 ) Rescue of LTD in knockout mice by lentiviral-mediated expression of SorCS2 ( SorCS2 is a postsynaptic regulator of i 20 80 60 40 80 60 40 20

SorCS2 - 140 120 100 120 100 140

) No rescue of LTD in

fEPSP slope (%) slope fEPSP fEPSP slope (%) slope fEPSP g synaptophysin. PSD1 and 2membrane designate (SPM) postsynaptic synaptic vesicles fractions andare of plasma found increasing membranes, in purity, respectively. dendrites ( and and in postsynaptic densities (shown). No labeling was observed in tissue from knockout mice (data not shown). ( blotting for SorCS2 in differentwith brain NeuN regions. (red, ( left(green). ( panel) or GFAP (red, right panel). ( © 2016 Macmillan Publishers Limited, part of Springer Nature. Molecular Psychiatry (2016), 1740 animals. ( Figure 1. 2 ms. ( in wild-type, but not in wild-type slices by antibodies blocking binding of proBDNF to SorCS2 ( Role of SorCS2 in BDNF-dependent plasticity S Glerup et al 1744 anti-SorCS2 antibodies abolished L-LTP in wild-type mice (97 ± 6% reduced (Figure 2i). Neurite branching and spine formation versus 137 ± 7%; P = 0.005) (Figure 2b), whereas in the converse involves phosphorylation of Akt and this was diminished in situation lentiviral-mediated reintroduction of SorCS2 rescued mutant neurons too (Figure 2j). As a consequence, Sorcs2–/– L-LTP in knockout slices (127 ± 8% versus 99 ± 9%; P = 0.04) neurons failed to respond to BDNF by increased neurite growth, (Figure 2c). Finally, supplementation with BDNF (25 ng ml − 1) also complexity and spine density, whereas wild-type neurons were restored L-LTP in Sorcs2–/– samples (140 ± 11% versus 141 ± 6%; highly responsive (Figures 2l and m). To investigate whether the P = 0.92) (Figure 2d). Extending these studies, we investigated cytoplasmic tail of SorCS2 was required, we measured sprouting in whether SorCS2 was required for CA1 neurons to respond to BDNF knockout neurons after transfection with wild-type SorCS2 or a when present at low concentrations. In brief, LTP was monitored mutant lacking the intracellular receptor domain (SorCS2-tailless). after the addition of increasing concentrations of BDNF to slices While BDNF stimulated neurite branching by 136% (Po0.04) in from Sorcs2–/– animals and mice heterozygous for BDNF (Bdnf+/–), cells expressing wild-type SorCS2, neurons expressing SorCS2- which are also impaired in LTP.38 Remarkably, 1 ng ml − 1 BDNF tailless failed to respond (Figure 2m). We also assessed neuron was sufficient to restore LTP in the Bdnf+/– mutants, but 20-fold morphology by microinjection of biocytin into dentate gyrus higher concentrations were needed in SorCS2-deficient mice granule neurons followed by assessment of dendritic branch order (Figure 2e). (Supplementary Figure S8). In 8-week-old Sorcs2–/– mice, dendritic complexity of the fourth branch order was decreased from SorCS2 binds TrkB to facilitate BDNF signaling 7.5 ± 0.7 to 4.9 ± 0.9 (P = 0.04) branch points. Given the reduced BDNF sensitivity, we hypothesized that SorCS2 might interact with TrkB to increase its signaling capacity. To evaluate SorCS2 is required for synaptic targeting of TrkB this, non-permeabilized hippocampal neurons were subjected to PSDs form the structural scaffold in excitatory synapses and are fluorescence resonance energy transfer and in situ proximity ligation important for efficient TrkB signaling.39 Because the SorCS2 (PLA). PLA is a method based on hybridization of circular DNA probes cytoplasmic tail contains a tyrosine-based sorting signal (Yxx∅; coupled to antibodies. When probes are localized o40 nm apart ∅ = large hydrophobic residue) and two modified dileucine they will hybridize, and juxtapositioned proteins can be visualized internalization motifs, and because SorCS2-tailless failed to rescue by PCR using fluorescently labeled nucleotides. We measured a BDNF activity, we speculated that SorCS2 might target TrkB to polarized-fluorescence resonance energy transfer efficiency (Eapp%) PSDs. To address this question, we subjected hippocampal of ~ 18%, indicating colocalization of SorCS2 and TrkB in neurites neurons to PLA for TrkB and PSD95 followed by immunostaining (Supplementary Figures S6a and b). Consistent with this observation, for SorCS2. Using z-stack images of distal neurites we observed punctate staining was evident on the surface of hippocampal that SorCS2 frequently appeared in the TrkB-enriched PSDs neurons when analyzed by PLA (Figure 2f). In co-immuno- (Figure 3a). Hence, we compared the synaptic distribution of TrkB precipitation experiments with hippocampal neurons, SorCS2 not in wild-type and Sorcs2–/– neurons. Remarkably, by subcellular only interacted with full-length TrkB but also with the truncated fractionationing TrkB was abundant in PSD95-positive fractions variant TrkB.T1 lacking the tyrosine kinase domain, indicating that the from wild-type hippocampi but not in the corresponding knock- interaction takes place between the extracellular receptor domains out fractions (Figure 3b). (Figure 2g). This hypothesis was confirmed by surface plasmon We next explored how SorCS2 regulates TrkB trafficking. resonance analysis where we measured a Kd of ~ 10 nM for binding Hippocampal cultures were subjected to HFS, and complex formation of the TrkB ectodomain to the extracellular part of SorCS2 of SorCS2 with TrkB and targeting of TrkB to PSD, respectively, were (Supplementary Figure S6c). Because anti-SorCS2 antibodies visualized by PLA followed by quantification of the fluorescence prevented this interaction (Supplementary Figures S5c and d) but intensity. Intriguingly, HFS induced ~ 2-fold increase in receptor not the interaction of BDNF with TrkB (data not shown), the deficient heterodimerization that was accompanied by an equivalent translo- L-LTP in anti-SorCS2-treated slices (cf. Figure 2b) likely reflects cation of TrkB to PSD95-positive microdomains (Figures 3c–e and g). that receptor heterodimerization has been disrupted. Finally, The synaptic translocation of TrkB was strictly SorCS2 dependent as ionotropic NMDA receptors, required for induction of LTP, failed to targeting to PSD was completely abolished in knockout neurons co-immunoprecipitation with SorCS2, supporting a function of (Figures 3d and h) As opposed to the high-frequency stimulation, LFS SorCS2 upstream of glutamate (Supplementary Figure S7). Taken did not impact on TrkB distribution (Figures 3d and g). In marked together, our data demonstrated that SorCS2 regulates LTP through contrast to the activity-dependent regulation of SorCS2/TrkB a direct and specific interaction with TrkB. heterodimerization and synaptic targeting, complex formation of Next, we studied BDNF signaling in SorCS2-deficient hippo- SorCS2 with p75NTR was static, that is, unresponsive to both HFS and campal neurons. TrkB autoactivation and total phosphorylation LFS (Figures 3c and f). Finally, we found that LTP in SorCS2-deficient were robustly induced in wild-type neurons, but substantially mice was rescued by lentivirus encoding full-length SorCS2 but not compromised in Sorcs2–/– cells (Figures 2h and i). Activation of the the tailless and sorting incompetent variant (Figure 2n). Taken phospholipase-Cγ pathway, required for induction of LTP, was also together, our data suggest that sorting signals in SorCS2 control

–/– Figure 2. Impaired N-methyl-D-aspartate receptor (NMDAR)-dependent long-term potentiation (LTP) and neuronal branching in Sorcs2 mice. (a) LTP in CA1 of wild type but not Sorcs2–/– (P = 0.0001). Scale bars, 5 mV and 2 ms. (b) Inhibition of LTP by inhibitory anti-SorCS2 antibodies (P = 0.005) but not by control immunoglobulin G (IgG). (c) Rescue of LTP by lentiviral-mediated expression of SorCS2, but not LacZ, in SorCS2 knockouts. (d) Rescue of LTP in Sorcs2–/– mice by the addition of 25 ng ml − 1 brain-derived neurotrophic factor (BDNF). Wild-type slices treated with vehicle are used for comparison. (e) LTP in Bdnf+/– and Sorcs2 − / − mice following the addition of increasing concentrations of BDNF. Field excitatory postsynaptic potential (fEPSP) slopes measured 240 min after high-frequency tetanic stimulation (HFS) are plotted against the BDNF concentration. (f) Proximity ligation (PLA) of surface TrkB with SorCS2 in wild-type, but not in knockout, neurons. Cells were not permeabilized to exclusively visualize surface TrkB/SorCS2 complex. (g) Immunoprecipitation of TrkB and SorCS2 from hippocampal neurons. (h–j) Impaired BDNF-induced tyrosine phosphorylation and signaling of TrkB in Sorcs2–/– hippocampal neurons. (h) TrkB immunoprecipitation followed by western blot (WB) for phosphotyrosine. (i) WB for the TrkB autophosphorylation site, and for the phosphorylated phospholipase C-γ (PLCγ) binding site. (j) Immunoﬂuorescence microscopy of phosphorylated Akt. (k–l) Dendritic branching and spine density in hippocampal neurons after stimulation with 1 nM BDNF. (m) Rescue of BDNF-induced branching in knockout neurons by full-length SorCS2, but not the tailless mutant. (n) Full-length SorCS2, but not SorCS2-tailless, rescues LTP in SorCS2 knockouts.

Molecular Psychiatry (2016), 1740 – 1751 © 2016 Macmillan Publishers Limited, part of Springer Nature. Role of SorCS2 in BDNF-dependent plasticity S Glerup et al 1745 synaptic targeting of TrkB and effectuate the selective responses to L-LTP.8,40 TrkB was recently suggested as a tag for expression of high- and low-frequency stimulation. L-LTP because of its ability to translocate to PSD and capture HFS creates a short-lived 'synaptic tag' at the potentiated BDNF.41 We speculated that SorCS2 was required for TrkB- synapse, which is required to sequester de novo synthesized mediated synaptic tagging, and performed a two-pathway plasticity-related proteins such as BDNF in the PSD and establish experiment in which a single electrode records fEPSPs from a

WT WT Sorcs2-/- Sorcs2-/- a Sorcs2-/- b c WT control IgG anti-SorCS2 IgG lenti SorCS2 lenti lacZ 250 250 250 a a 200 a 200 a 200 a a b b b b b b 150 150 150

100 100 a 100 a b 10 µg/ml control IgG b a b 50 50 50 fEPSP slope (%) fEPSP fEPSP slope (%) fEPSP 10 µg/ml anti-SorCS2 IgG slope (%) fEPSP 0 0 0 -20 0 20 40 60 80 100 120 140 160 180 -20 0 20 40 60 80 100 120 140 160 180 -20 0 20 40 60 80 100 120 140 160 180 Time (min) Time (min) Time (min)

deWT Sorcs2-/- f 250 vehicle BDNF Bdnf+/- Sorcs2-/- + BDNF + BDNF _ _ a 160 Sorcs2 / 200 a WT b b 150 150 140

100 b 130 a 120 50 25 ng/ml BDNF

fEPSP slope (%) fEPSP 110 fEPSP slope (%) fEPSP 4 h after tetanus 0 100 SorCS2/TrkB PLA -20 0 20 40 60 80 100 120 140 160 180 0.01 0.1 1 10 100 Time (min) BDNF (ng/ml) α SorCS2 g h ij IgG BDNF ng/ml 0105001050 WT BDNF WB: ng/ml - TrkB 0 5 10 25 50 0 5 10 25 50 pTrkB auto - TrkB T1 IP: TrkB - SorCS2 WB: pTyr WB: pAkt BDNF TrkB IP IP: TrkB Sorcs2-/- WB: TrkB WB: - TrkB pTrkB PLCγ - TrkB T1 WT Sorcs2-/- Sorcs2-/- pAkt BDNF Input WT

l k 4 4 P=0.002 4 No BDNF No BDNF 3 BDNF 3 3 BDNF P=0.003 2 2 2 Spines/uM 1 1 1 P=0.005 Number of branches Number of branches 0 0 0 I II III I II III WT Sorcs2-/- WT Branch order Sorcs2-/- Branch order m n Sorcs2-/- Sorcs2-/- lenti SorCS2 lenti SorCS2 10 P=0.004 10 tailless No BDNF 8 8 a a BDNF 250 bb 6 P=0.02 6 200

4 P=0.04 4 150 2 2 100 a b fPSP slope (%) fPSP Number of branches 50 Number of branches 0 0 I II III I II III 0 Sorcs2-/- Branch order Sorcs2-/- Branch order -20 0 20 40 60 80 100 120 140 160 180 + SorCS2 cDNA + SorCS2 tailless cDNA Time (min)

© 2016 Macmillan Publishers Limited, part of Springer Nature. Molecular Psychiatry (2016), 1740 – 1751 Role of SorCS2 in BDNF-dependent plasticity S Glerup et al 1746 S SPM PS SV P S P S V M D P D P PSD95/TrkB/SorCS2 - TrkB - TrkB - TrkB T1 - TrkB T1

_ _ WT Sorcs2 /

Control LFS HFS Control LFS HFS

SorCS2/TrkB PLA PSD95/TrkB PLA Control LFS HFS Control-/- LFS -/- HFS -/-

SorCS2/p75 PLA PSD95/TrkB PLA

50 P =0.03 16 25 25 14 P =0.01 40 20 20 12

30 10 15 15 8 20 6 10 10 4 10 5 5 Intensity (SorCS2/p75) Intensity (TrkB/PSD95) P Intensity (TrkB/PSD95) Intensity (SorCS2/TrkB) =0.0002 2 0 0 0 0 LFS LFS LFS HFS HFS HFS LFS HFS Control Control Control Control WT Sorcs2 -/- Figure 3. SorCS2 targets TrkB to postsynaptic densities and regulates signaling. (a) Proximity ligation (PLA) for postsynaptic density 95 (PSD95) and TrkB (red) combined with immunoﬂuorescense staining for SorCS2 (green). Colocalization (yellow) is indicated by arrowheads. (b) Subcellular fractionation of hippocampal extracts analyzed by western blot (WB) for TrkB. (c) Heterodimerization between SorCS2 and TrkB, but not SorCS2 and p75NTR, is differentially regulated by low-frequency stimulation (LFS) and high-frequency tetanic stimulation (HFS). (d) Impaired targeting of TrkB to PSD in Sorcs2–/– neurons (lower panel) subjected to HFS compared to WT neurons (upper panel). (e–h) Quantiﬁcation of the dendritic PLA signal 10 μm from the soma. Twelve to 15 neurons were chosen at random and evaluated per coverslip (n=3).

neuronal population innervated by two separate populations of Behavioral deficits in SorCS2 knockout mice Schaffer collaterals (Figure 4a). As paired-pulse facilitation is a We further studied this process using a behavioral tagging basal property of CA3–CA1 synapses (Supplementary Figure S3), paradigm thought to be the functional analog of synaptic we tested if two separate populations of synapses were activated. tagging.11,32,42 In this task, a short-term contextual memory induced Activation of one population of Schaffer collaterals (S1) did not by a weak foot shock in an inhibitory avoidance experiment can be facilitate fEPSP induced by activation of the second population of converted into long-term memory (LTM) if there has been prior fibers (S2) 50 ms earlier, and vice versa, demonstrating that the exploration of a novel environment consisting of an open field two pathways are independent (Figure 4a). We then stimulated containing novel objects. During training, latency to enter the dark with weak HFS (wHFS; 100 Hz, 1 s) to induce a synaptic tag in the box was identical between genotypes, and stimulation with a weak S2 pathway. Weak stimulation to S2 failed to induce L-LTP (fEPSPs: electric shock (0.1 mA) resulted in a higher latency in both geno- 101 ± 3%) (Figure 4b) but HFS to S1 60 min before wHFS at S2 types after 1 h, reflecting intact short-term memory (Figure 4f). rescued L-LTP of S2 (Figure 4c) (S1: 151 ± 4%; S2: 136 ± 5%). Next, When tested 24 h later, none of the animals recalled the stimulus we applied anti-SorCS2 IgG to slices from wild-type mice, 40 min (Figure 4g). Remarkably, exposure for 15 min to the novel after HFS in S1. Strikingly, while anti-SorCS2 IgG had no effect on environment before the weak stimulus induced LTM in wild-type L-LTP in S1 (S1: 150 ± 5%), L-LTP in S2 was now abolished as but not in Sorcs2–/– mice (Figure 4h). When subjected to a strong potentiation decayed to baseline within 180 min after wHFS (S2: stimulus (0.8 mA) both genotypes responded by LTM (Figure 4i), 98 ± 3%, P = 0.01) (Figure 4d). For comparison, control IgG had no demonstrating that SorCS2 mutants can consolidate memory, a effect on L-LTP (Figure 4e) (S2: 140 ± 5%). These data confirmed finding also reported for mice with impaired TrkB activity.11,43 that SorCS2 targets TrkB to the PSD of potentiated synapses Spatial learning and memory are considered reliant on L-LTP securing their persistence. and are substantially reduced in Bdnf+/– mice.44 We trained mice

Molecular Psychiatry (2016), 1740 – 1751 © 2016 Macmillan Publishers Limited, part of Springer Nature. Role of SorCS2 in BDNF-dependent plasticity S Glerup et al 1747 HFS wHFS WT WT 250 wHFS 250 S1 S1 Recording S2 S2 200 a 200 b 150 150

100 100 a b 50 50 fEPSP slope (%) fEPSP fEPSP slope (%) fEPSP 0 0 -20 0 20 40 60 80 100 120140 160 180 0 40 80 120 160 200 240 S2 alone Time (min) Time (min)

HFS wHFS HFS wHFS S1 S2 WT WT 250 S1 250 S1 S1 alone S2 S2 200 200

S2 S1 150 150

100 100

50 10 µg/ml anti-sorCS2 IgG 50 10 µg/ml control IgG fEPSP slope (%) fEPSP fEPSP slope (%) fEPSP 0 0 0 40 80 120 160 200240 0 40 80 120 160 200 240 Time (min) Time (min)

300 300 100 500 P=0.01 P=0.04

250 250 80 400 200 200 60 300 150 150 40 200 100 100 100 50 50 20 Latency to cross (sec) Latency to cross (sec) Latency to cross (sec) Latency to cross (sec) 0 0 0 0 / / / / / / / / ______WT WT WT WT WT WT WT WT Sorcs2 Sorcs2 Sorcs2 Sorcs2 Sorcs2 Sorcs2 Sorcs2 Sorcs2 Training 1 h test Training 24 h test Novelty + 24 h test Training 24 h test 0.1 mA 0.1 mA Training 0.8 mA 0.1 mA Figure 4. SorCS2 is essential for synaptic tagging in vitro and in vivo.(a) Illustration of a two-pathway experiment (upper panel), and demonstration of independence between two pathways (lower panel). (b) Weak high-frequency tetanic stimulation (wHFS; 100 Hz, 1 s) to S2 failed to induce late phase of long-term potentiation (L-LTP) (101 ± 3%). (c) HFS to S1 before wHFS at S2 rescued L-LTP. (d and e) Anti-SorCS2 immunoglobulin G (IgG), but not preimmune IgG, applied 40 min after HFS at S1 has no effect on L-LTP in S1 but abolishes synaptic potentiation induced by wHFS at S2. (f and g) Normal short-term memory (STM) but not long-term memory (LTM) in knockout mice after weak electrical shock in the inhibitory avoidance test (wild type: n=4; SorCS2–/–: n=5). (h) Subjection of mice to novel environment converts STM to LTM in wild-type mice (n=9) but not in SorCS2–/– mice (n=12). (i) A strong stimulus (0.8 mA) leads to LTM in SorCS2 knockouts (n=4) and wild types (n=7).

to navigate to a specific position in the Barnes maze when guided elevated plus maze, Sorcs2–/– mice were hyperactive, but also less by spatial cues. At day 1, mutant and wild-type mice used anxious and more risk taking as evidenced by augmented mobility, a combination of random and serial search strategies, and and more entries and time spent in the open arms (Figures 5e–g). acquired the task equally well (Figures 5a–d). When tested on Inattentiveness typified the mutants as nearly 60% of the knockouts day 4 Sorcs2–/– mice did not remember the location of the escape (4 out of 7), but none of the wild-type animals (0 out of 16) fell tunnel as their search pattern continued to be random and the down from the maze. In the tail suspension and Porsolt forced time to reach the hole was unchanged. In contrast, wild-type swim tests, Sorcs2–/– mice displayed a depression-like response in animals soon learned to use an efficient search strategy leading both paradigms. Immobility time was increased by 72% and 40%, them directly to the target hole (Figures 5a–d). respectively (Figures 5h and i), suggesting increased susceptibility Periods of mania and depression are hallmarks of BD but to stress. Furthermore, SorCS2 knockouts displayed altered sensory symptoms including anxiety, hallucinations and impulsivity are motor gating, as demonstrated by ~ 50% attenuated PPI compared common in SZ as well.45 We tested knockouts in behavioral with controls (Figures 5j, P=0.0001) despite normal startle response paradigms modeling these symptoms. In an anxiety paradigm, the (Figure 5k), a common feature of both BD and SZ in humans.46,47

© 2016 Macmillan Publishers Limited, part of Springer Nature. Molecular Psychiatry (2016), 1740 – 1751 Role of SorCS2 in BDNF-dependent plasticity S Glerup et al 1748 P =0.007 60 P WT _ _ =0.004 WT Sorcs2 / 0.4 50 Day 1 180 160 40 0.3 140 120 30 Day 4 100 0.2 80 * 20 * Mean errors _ _ 60 0.1 Sorcs2 / 40 10 Time to target (s) Time 20

Mean distance to target (m) Day 1 0 0 0.0 Day 1 Day 4 Day 1 Day 4 Day 1 Day 2 Day 3 Day 4 Day 1 Day 4 Day 1 Day 4 _ _ _ _ Day 4 WT Sorcs2 / WT Sorcs2 /

P =1.1E-4 50 16 P =7.0E-4 P =0.02 P 18 P =0.02 14 250 =1.1E-4 16 40 12 200 14 200 10 12 30 150 10 8 150 8 20 6 100 100

6 Immobile (s) Floating (s)

Distance (m) 4

4 10 Open arm time (%) 50 50 2 Open arm entries (%) 2 0 0 0 0 0 / / / / / _ _ _ _ _ WT WT WT WT WT Sorcs2 Sorcs2 Sorcs2 Sorcs2 Sorcs2

60 P =2.0E-4 70 WT _ /_ 50 60 Sorcs2 40 50 30 40 30 20 20

10 Startle response 10 Prepulse inhibition (%) 0 0

/ 0 80 90 100 105 110 120 _ Sound (dB) WT Sorcs2 Figure 5. Altered behavior in SorCS2 knockouts. (a–d) Impaired spatial memory of SorCS2–/– mice in the Barnes maze; time to reach target (a), mean number of errors (b) and distance to target (c)(n = 8 and 5 animals per group). (d) Traces of the search patterns. (e–g) Increased motor activity (e; distance traveled) and decreased anxiety (f–g; open arm entries, time spent in open arms) of SorCS2–/– mice in the elevated plus maze (n=7 animals per group). (h–i) Increased immobility of SorCS2 knockouts in the tail suspension test (n = 6 of each genotype) and Porsolt forced swim test (n=9 per group). (j–k) Attenuated prepulse inhibition in SorCS2–/– animals (j)(n = 7 and 11 animals per group), despite normal startle response to increasing sound intensities (k)(n = 7 per group).

DISCUSSION response of BD patients to pharmacological treatment was Although it is ﬁrmly established that genetic risk factors are critical suggestively associated with markers at SORCS2, among other in the pathogenesis of mental disorders, the genes involved and genes.49 In a recent genome-wide association studies of attention- their modes of action are less well understood. A suggestive related problems in ADHD cases, the strongest association was association between BD and an intronic SORCS2 single-nucleotide with a single-nucleotide polymorphism in SORCS2.22 The human polymorphism was reported in three independent samples, and genetic evidence suggests that variation in SORCS2 confers risk to one independent sample then highlighted this region in a a broad range of psychiatric illnesses, in line with genetic overlap combined analysis of BD and SZ cases.19–21,48 Notably, long-term and comorbidity between mental disorders. However, to date no

Molecular Psychiatry (2016), 1740 – 1751 © 2016 Macmillan Publishers Limited, part of Springer Nature. Role of SorCS2 in BDNF-dependent plasticity S Glerup et al 1749 predicted damaging coding mutations in SORCS2 have been transport along neurites.18 The Slit and Trk family member Slitrk5 reported. expressed postsynatically can bind protein tyrosine phosphatase-δ Many psychiatric disorders have an important neurodevelop- (PTP-δ) in a transconfiguration. However, in the presence of BDNF mental component involving altered neurogenesis and neuronal it will dissociate from presynaptic PTP-δ and shift to a cis- wiring.2,50,51 However, acute alterations in synaptic function are interaction with TrkB allowing efficient recycling in Rab11-positive also implicated, as highlighted by the behavioral response of endosomes and BDNF signaling. Hence, we speculate that sortilin, psychiatric patients to mood-stabilizing drugs and central nervous SorCS2 and Slitrk5 work in concert to effectuate TrkB signaling by system stimulants. The importance of proper synaptic function to facilitating its long-range transport, local synaptic targeting and mental health has received further support from patient-derived recycling back to the plasma membrane.57 inducible pluripotent stem cell models,52 and genome-wide Binding of proBDNF to SorCS2 is accounted for by its association studies and sequencing studies that have connected prodomain.13 Recently, Anastasia et al.58 reported that not only 53 multiple synaptic proteins to psychiatric disorders. We and proBDNF and mature BDNF but also the isolated prodomain others recently identified SorCS2 as a novel proBDNF receptor can be secreted in an activity-dependent manner. A common critical for proBDNF-induced growth cone collapse in cultures of single-nucleotide polymorphism in BDNF resulting in a Val66Met 13,35 CNS neurons and for correct dopaminergic innervation of the substitution in the BDNF prodomain is associated with impaired 13 frontal cortex during development. We report here that SorCS2 memory and linked to risk of mental disorders.59 Intriguingly, the also exhibits acute synaptic effects, regulating BDNF-induced Met66 variant binds more strongly to SorCS2 than the Val66 synaptic and morphological plasticity in the mature hippocampus. prodomain58 raising the possibility that the risk variant might lead We show that LTP and LTD are completely missing in hippocampal –/– to displacement of proBDNF from the receptor to inhibit LTD and slices from Sorcs2 mice. Plasticity could be fully rescued by LTP. Indeed, the purified Met66 prodomain abrogates LTD in vitro transfection with lentivirus encoding SorCS2 or, in the case of LTP, and transgenic BDNF Met/Met mice exhibit reduced NMDAR- by application of high concentrations of exogenous BDNF. dependent LTP and LTD but preserved metabotropic LTD,60,61 a Remarkably, the BDNF concentration required for full rescue of phenotype equivalent to the perturbations observed in SorCS2 LTP was ~ 20-fold higher than the concentration that rescued knockout mice. fi plasticity de cits in hippocampal slices from BDNF heterozygous Obligate TrkB knockout mice die shortly after birth due to mice, pointing to the critical role of SorCS2 in BDNF-induced impaired development of the peripheral and sympathetic nervous signaling. SorCS2 was required for BDNF-induced changes in system. SorCS2 knockout mice, on the other hand, are viable and neuronal morphology, as evidenced by the failure of BDNF to show normal growth and development. However, several stimulate increases in spine density and dendritic complexity in behavioral phenotypes of Sorcs2–/– animals resemble those of hippocampal cultures lacking SorCS2. This effect appeared to be mice with postnatal conditional forebrain-specific TrkB deletion, caused by a direct impact on TrkB, because BDNF-stimulated TrkB namely reduced spatial memory, hyperactivity and increased risk- transphosphorylation and downstream signaling were markedly taking behavior.43,62 Thus, it appears that SorCS2 is selectively reduced in SorCS2 knockout neurons compared with WT controls. required for postnatal TrkB activity in the CNS. In the present Because antibodies raised against the extracellular domain of study, we show that SorCS2-deficient mice also display a number SorCS2 completely abolished L-LTP in WT hippocampal slices of additional important behavioral phenotypes, modeling symp- SorCS2 likely functions on the cell surface of neurons to regulate toms of psychiatric disorders including BD and SZ. These findings plasticity. add to results establishing the role of SorCS2 in behaviors related Notably, HFS, but not LFS, increased SorCS2 and TrkB hetero- to ADHD.13 The behavioral consequences of SorCS2 loss are dimerization in the PSD. This was in contrast to the SorCS2/p75NTR pleiotropic, but they overlap substantially the behavior pheno- interaction, which was not influenced by stimulation. Neuronal types reported in mice with mutant copies of Disc1 (disrupted-in- activity did not increase total SorCS2 levels in the PSD but did schizophrenia-1), Drd2, Grik2, Clock, Grin1, Grm3 and -6, and Shank, markedly increase TrkB/PSD95 colocalization. This effect of – genes that have all been implicated in BD and/or SZ.63 69 Hence, neuronal activity on TrkB/PSD95 colocalization was observed –/– only in wild-type neurons, and its absence in SorCS2 knockout the phenotype observed in Sorcs2 mice is consistent with the neurons indicates that synaptic SorCS2 is instrumental for activity- multiple reported associations between SORCS2 polymorphisms and psychiatric diagnoses in humans. dependent recruitment of TrkB into the PSD. fi The synaptic tagging and capture hypothesis is an attractive In conclusion, we have identi ed SorCS2 as an indispensable regulator of NMDAR-dependent plasticity (see model in Supple- model because of its ability to explain how LTP and memory NTR consolidation can be restricted to selected synapses.8 Importantly, mentary Figure S9). SorCS2 facilitates LTD by joining with p75 it has been shown that TrkB could function as a synaptic tag both to form a composite binding site for proBDNF, and SorCS2 fi in vitro and in vivo.11 Our findings suggest that SorCS2 is required regulates LTP by acting as an activity-dependent traf cking fi for several activities of TrkB, including its role in synaptic tagging. receptor that captures TrkB in the activated synapse. Our ndings Hence, it is possible that interactions between SorCS2, p75NTR and identify SorCS2 as a novel component in proBDNF- and BDNF- TrkB help determine the type of plasticity, which will be induced dependent signaling, and suggest a mechanistic link between by neuronal activity. Under this model, proBDNF released by SORCS2 and severe mental illnesses. LFS54 induces LTD through formation of a ternary complex with NTR the static SorCS2/p75 complex, whereas HFS enables SorCS2 to CONFLICT OF INTEREST capture TrkB in the PSD allowing BDNF to elicit and sustain LTP. fl Future studies should clarify whether SorCS2 also complies with The authors declare no con ict of interest. the defined criteria for being a synaptic tag.11,55,56 The combined data suggests a model in which SorCS2 is ACKNOWLEDGMENTS distributed in a steady-state manner between synaptic and extrasynaptic compartments, and is constantly being recycled. This study was funded by the Lundbeck Foundation (to AN, SG), the Rochester fi Epidemiology Project (grant number R01 AG034676) (to AN), Danish Agency for Supporting such a model, we found that a sorting-de cient Science Technology and Innovation (DAGMAR) (to AN) and Danish Council for SorCS2 mutant failed to rescue LTP in SorCS2 knockout hippo- Independent Research Sapere Aude starting grant (grant number DFF 4183-00604) campus, demonstrating that BDNF-induced plasticity is reliant on (to SG). Professor Joel Gelernter (Yale University School of Medicine) provided helpful SorCS2 trafficking. Curiously, the structurally related protein comments. Anja Aagaard and Benedicte Vestergaard are thanked for excellent sortilin associates with Trk receptors to enhance their anterograde technical assistance.

© 2016 Macmillan Publishers Limited, part of Springer Nature. Molecular Psychiatry (2016), 1740 – 1751 Role of SorCS2 in BDNF-dependent plasticity S Glerup et al 1750 REFERENCES gene confers susceptibility to bipolar disorder and affects transcriptional activity. 1 Hartz SM, Pato CN, Medeiros H, Cavazos-Rehg P, Sobell JL, Knowles JA et al. Mol Psychiatry 2006; 11:695–703. Comorbidity of severe psychotic disorders with measures of substance use. 29 Hosang GM, Uher R, Keers R, Cohen-Woods S, Craig I, Korszun A et al. Stressful JAMA Psychiatry 2014; 71:248–254. life events and the brain-derived neurotrophic factor gene in bipolar disorder. 2 Manji HK, Drevets WC, Charney DS. The cellular neurobiology of depression. J Affect Disord 2010; 125: 345–349. Nat Med 2001; 7:541–547. 30 Cheah SY, Lawford BR, Young RM, Connor JP, Phillip Morris C, Voisey J. BDNF SNPs 3 Schloesser RJ, Martinowich K, Manji HK. Mood-stabilizing drugs: mechanisms are implicated in comorbid alcohol dependence in schizophrenia but not in of action. Trends Neurosci 2012; 35:36–46. alcohol-dependent patients without schizophrenia. Alcohol Alcohol 2014; 49: 4 Lopez AD, Murray CC. The global burden of disease, 1990–2020. Nat Med 1998; 4: 491–497. 1241–1243. 31 Nees F, Witt SH, Dinu-Biringer R, Lourdusamy A, Tzschoppe J, Vollstadt-Klein S 5 Nagappan G, Lu B. Activity-dependent modulation of the BDNF receptor TrkB: et al. BDNF Val66Met and reward-related brain function in adolescents: role for mechanisms and implications. Trends Neurosci 2005; 28:464–471. early alcohol consumption. Alcohol 2015; 49:103–110. 6 Pang PT, Teng HK, Zaitsev E, Woo NT, Sakata K, Zhen S et al. Cleavage of proBDNF 32 Moncada D, Viola H. Induction of long-term memory by exposure to novelty by tPA/plasmin is essential for long-term hippocampal plasticity. Science 2004; requires protein synthesis: evidence for a behavioral tagging. J Neurosci 2007; 27: 306:487–491. 7476–7481. 7 Woo NH, Teng HK, Siao CJ, Chiaruttini C, Pang PT, Milner TA et al. Activation of 33 Fadok JP, Dickerson TM, Palmiter RD. Dopamine is necessary for cue-dependent p75NTR by proBDNF facilitates hippocampal long-term depression. Nat Neurosci fear conditioning. J Neurosci 2009; 29: 11089–11097. 2005; 8: 1069–1077. 34 Lu B, Pang PT, Woo NH. The yin and yang of neurotrophin action. Nat Rev Neurosci 8 Redondo RL, Morris RG. Making memories last: the synaptic tagging and capture 2005; 6: 603–614. hypothesis. Nat Rev Neurosci 2011; 12:17–30. 35 Deinhardt K, Kim T, Spellman DS, Mains RE, Eipper BA, Neubert TA et al. Neuronal 9 Meyer-Franke A, Wilkinson GA, Kruttgen A, Hu M, Munro E, Hanson MG Jr et al. growth cone retraction relies on proneurotrophin receptor signaling through Rac. Depolarization and cAMP elevation rapidly recruit TrkB to the plasma membrane Sci Signal 2011; 4: ra82. of CNS neurons. Neuron 1998; 21: 681–693. 36 Zhang D, Zhang J, Bian C, Deng Q. Postnatal and ovariectomic regulation of 10 Ji Y, Pang PT, Feng L, Lu B. Cyclic AMP controls BDNF-induced TrkB phosphory- postsynaptic density protein-95 in the hippocampus of female Sprague– lation and dendritic spine formation in mature hippocampal neurons. Nat Neu- Dawley rats. Synapse 2010; 64: 875–878. rosci 2005; 8: 164–172. 37 Messaoudi E, Ying SW, Kanhema T, Croll SD, Bramham CR. Brain-derived neuro- 11 Lu Y, Ji Y, Ganesan S, Schloesser R, Martinowich K, Sun M et al. TrkB as a potential trophic factor triggers transcription-dependent, late phase long-term potentiation synaptic and behavioral tag. J Neurosci 2011; 31: 11762–11771. in vivo. J Neurosci 2002; 22:7453–7461. 12 Deinhardt K, Kim T, Spellman DS, Mains RE, Eipper BA, Neubert TA et al. Neuronal 38 Pang PT, Teng HK, Zaitsev E, Woo NT, Sakata K, Zhen SH et al. Cleavage of growth cone retraction relies on proneurotrophin receptor signaling through Rac. proBDNF by tPA/plasmin is essential for long-term hippocampal plasticity. Sci Signal 2011; 4: ra82. Science 2004; 306: 487–491. 13 Glerup S, Olsen D, Vaegter CB, Gustafsen C, Sjoegaard SS, Hermey G et al. SorCS2 39 Suzuki S, Numakawa T, Shimazu K, Koshimizu H, Hara T, Hatanaka H et al. BDNF- regulates dopaminergic wiring and is processed into an apoptotic two-chain induced recruitment of TrkB receptor into neuronal lipid rafts: roles in synaptic receptor in peripheral glia. Neuron 2014; 82: 1074–1087. modulation. J Cell Biol 2004; 167:1205–1215. 14 Willnow TE, Petersen CM, Nykjaer A. VPS10P-domain receptors—regulators of 40 Frey U, Morris RGM. Synaptic tagging and long-term potentiation. Nature 1997; neuronal viability and function. Nat Rev Neurosci 2008; 9:899–909. 385:533–536. 15 Glerup S, Nykjaer A, Vaegter CB. Sortilins in neurotrophic factor signaling. Handb 41 Lu Y, Ji YY, Ganesan S, Schloesser R, Martinowich K, Sun M et al. TrkB as a potential Exp Pharmacol 2014; 220: 165–189. synaptic and behavioral tag. J Neurosci 2011; 31: 11762–11771. 16 Nielsen MS, Keat SJ, Hamati JW, Madsen P, Gutzmann JJ, Engelsberg A et al. 42 Wang SH, Redondo RL, Morris RGM. Relevance of synaptic tagging and capture to Different motifs regulate trafficking of SorCS1 isoforms. Traffic 2008; 9:980–994. the persistence of long-term potentiation and everyday spatial memory. Proc Natl 17 Nielsen MS, Madsen P, Christensen EI, Nykjaer A, Gliemann J, Kasper D et al. The Acad Sci USA 2010; 107: 19537–19542. sortilin cytoplasmic tail conveys Golgi-endosome transport and binds the VHS 43 Minichiello L, Korte M, Wolfer D, Kuhn R, Unsicker K, Cestari V et al. Essential role domain of the GGA2 sorting protein. EMBO J 2001; 20: 2180–2190. for TrkB receptors in hippocampus-mediated learning. Neuron 1999; 24: 401–414. 18 Vaegter CB, Jansen P, Fjorback AW, Glerup S, Skeldal S, Kjolby M et al. Sortilin 44 Linnarsson S, Bjorklund A, Ernfors P. Learning deficit in BDNF mutant mice. Eur J associates with Trk receptors to enhance anterograde transport and neurotrophin Neurosci 1997; 9: 2581–2587. signaling. Nat Neurosci 2011; 14:54–61. 45 Kurnianingsih YA, Kuswanto CN, McIntyre RS, Qiu A, Ho BC, Sim K. Neuro- 19 Baum AE, Akula N, Cabanero M, Cardona I, Corona W, Klemens B et al. A genome- cognitive-genetic and neuroimaging–genetic research paradigms in schizo- wide association study implicates diacylglycerol kinase eta (DGKH) and several phrenia and bipolar disorder. J Neural Transm 2011; 118: 1621–1639. other genes in the etiology of bipolar disorder. Mol Psychiatry 2008; 13: 46 Perry W, Minassian A, Feifel D, Braff DL. Sensorimotor gating deficits in bipolar 197–207. disorder patients with acute psychotic mania. Biol Psychiatry 2001; 50: 418–424. 20 Ollila HM, Soronen P, Silander K, Palo OM, Kieseppa T, Kaunisto MA et al. Findings 47 Braff DL, Grillon C, Geyer MA. Gating and habituation of the startle reflex in from bipolar disorder genome-wide association studies replicate in a Finnish schizophrenic patients. Archiv Gen Psychiatry 1992; 49:206–215. bipolar family cohort. Mol Psychiatry 2009; 14:351–353. 48 Takata A, Kawasaki H, Iwayama Y, Yamada K, Gotoh L, Mitsuyasu H et al. Nominal 21 Christoforou A, McGhee KA, Morris SW, Thomson PA, Anderson S, McLean A et al. association between a polymorphism in DGKH and bipolar disorder detected in a Convergence of linkage, association and GWAS findings for a candidate region for meta-analysis of East Asian case–control samples. Psychiatry Clin Neurosci 2011; bipolar disorder and schizophrenia on chromosome 4p. Mol Psychiatry 2011; 16: 65:280–285. 240–242. 49 Fabbri C, Serretti A. Genetics of long-term treatment outcome in bipolar disorder. 22 Alemany S, Ribases M, Vilor-Tejedor N, Bustamante M, Sanchez-Mora C, Bosch R Prog Neuropsychopharmacol Biol Psychiatry 2016; 65:17–24. et al. New suggestive genetic loci and biological pathways for attention function in 50 Martinowich K, Schloesser RJ, Manji HK. Bipolar disorder: from genes to behavior adult attention-deficit/hyperactivity disorder. Am J Med Genet B 2015; 168:459–470. pathways. J Clin Invest 2009; 119: 726–736. 23 Cross-Disorder Group of the Psychiatric Genomics C. Identification of risk loci with 51 Ross CA, Margolis RL, Reading SA, Pletnikov M, Coyle JT. Neurobiology of schi- shared effects on five major psychiatric disorders: a genome-wide analysis. Lancet zophrenia. Neuron 2006; 52:139–153. 2013; 381: 1371–1379. 52 Wen Z, Nguyen HN, Guo Z, Lalli MA, Wang X, Su Y et al. Synaptic dysregulation in 24 Cross-Disorder Group of the Psychiatric Genomics C, Lee SH, Ripke S, Neale BM, a human iPS cell model of mental disorders. Nature 2014; 515:414–418. Faraone SV, Purcell SM et al. Genetic relationship between five psychiatric dis- 53 Schizophrenia Working Group of the Psychiatric Genomics C. Biological insights orders estimated from genome-wide SNPs. Nat Genet 2013; 45:984–994. from 108 schizophrenia-associated genetic loci. Nature 2014; 511: 421–427. 25 Castren E, Rantamaki T. The role of BDNF and its receptors in depression and 54 Nagappan G, Zaitsev E, Senatorov VV Jr, Yang J, Hempstead BL, Lu B. Control of antidepressant drug action: Reactivation of developmental plasticity. Dev Neuro- extracellular cleavage of ProBDNF by high frequency neuronal activity. Proc Natl biol 2010; 70: 289–297. Acad Sci USA 2009; 106: 1267–1272. 26 Martinowich K, Manji H, Lu B. New insights into BDNF function in depression and 55 Martin KC, Kosik KS. Synaptic tagging—who's it? Nat Rev Neurosci 2002; 3: anxiety. Nat Neurosci 2007; 10: 1089–1093. 813–820. 27 Rybakowski JK. BDNF gene: functional Val66Met polymorphism in mood disorders 56 Kelleher RJ III, Govindarajan A, Tonegawa S. Translational regulatory mechanisms and schizophrenia. Pharmacogenomics 2008; 9: 1589–1593. in persistent forms of synaptic plasticity. Neuron 2004; 44:59–73. 28 Okada T, Hashimoto R, Numakawa T, Iijima Y, Kosuga A, Tatsumi M et al. A 57 Song M, Giza J, Proenca CC, Jing D, Elliott M, Dincheva I et al. Slitrk5 mediates BDNF- complex polymorphic region in the brain-derived neurotrophic factor (BDNF) dependent TrkB receptor trafficking and signaling. Dev Cell 2015; 33:690–702.

Molecular Psychiatry (2016), 1740 – 1751 © 2016 Macmillan Publishers Limited, part of Springer Nature. Role of SorCS2 in BDNF-dependent plasticity S Glerup et al 1751 58 Anastasia A, Deinhardt K, Chao MV, Will NE, Irmady K, Lee FS et al. Val66Met 64 O'Tuathaigh CM, Kirby BP, Moran PM, Waddington JL. Mutant mouse models: polymorphism of BDNF alters prodomain structure to induce neuronal growth genotype–phenotype relationships to negative symptoms in schizophrenia. cone retraction. Nat Commun 2013; 4: 2490. Schizophr Bull 2010; 36:271–288. 59 Dincheva I, Glatt CE, Lee FS. Impact of the BDNF Val66Met polymorphism 65 Chen G, Henter ID, Manji HK. Partial rodent genetic models for bipolar disorder. on cognition: implications for behavioral genetics. Neuroscientist 2012; 18: Curr Top Behav Neurosci 2011; 5:89–106. 439–451. 66 Umemori J, Takao K, Koshimizu H, Hattori S, Furuse T, Wakana S et al. ENU- 60 Mizui T, Ishikawa Y, Kumanogoh H, Lume M, Matsumoto T, Hara T et al. BDNF mutagenesis mice with a non-synonymous mutation in Grin1 exhibit abnormal pro-peptide actions facilitate hippocampal LTD and are altered by the common anxiety-like behaviors, impaired fear memory, and decreased acoustic startle BDNF polymorphism Val66Met. Proc Natl Acad Sci USA 2015; 112: E3067–E3074. response. BMC Res Notes 2013; 6: 203. 61 Ninan I, Bath KG, Dagar K, Perez-Castro R, Plummer MR, Lee FS et al. The BDNF 67 Schmeisser MJ. Translational neurobiology in Shank mutant mice—model Val66Met polymorphism impairs NMDA receptor-dependent synaptic plasticity in systems for neuropsychiatric disorders. Ann Anat 2015; 200:115–117. the hippocampus. J Neurosci 2010; 30: 8866–8870. 68 Shaltiel G, Maeng S, Malkesman O, Pearson B, Schloesser RJ, Tragon T et al. 62 Zorner B, Wolfer DP, Brandis D, Kretz O, Zacher C, Madani R et al. Forebrain- Evidence for the involvement of the kainate receptor subunit GluR6 (GRIK2) speciﬁc trkB-receptor knockout mice: behaviorally more hyperactive than in mediating behavioral displays related to behavioral symptoms of mania. ‘depressive’. Biol Psychiatry 2003; 54: 972–982. Mol Psychiatry 2008; 13: 858–872. 63 Cosgrove VE, Kelsoe JR, Suppes T. Toward a valid animal model of bipolar 69 Han K, Holder JL Jr, Schaaf CP, Lu H, Chen H, Kang H et al. SHANK3 overexpression disorder: how the research domain criteria help bridge the clinical-basic causes manic-like behaviour with unique pharmacogenetic properties. Nature science divide. Biol Psychiatry 2016; 79:62–70. 2013; 503:72–77.

Supplementary Information accompanies the paper on the Molecular Psychiatry website (http://www.nature.com/mp)