Introduction

The specificity of synaptic connections is essential for the function of neural circuits(Yogev and Shen, 2014). Cell adhesion molecules play crucial roles in the specificity of synapse formation(Duan et al., 2014, Serizawa et al., 2006, Tan et al., 2015, Rawson et al., 2017, Sytnyk et al., 2017, Berns et al., 2018, Jang et al., 2017, Courgeon and Desplan, 2019), but how to achieve specificity at the microcircuit level remains to be determined. The unique expression pattern of clustered protocadherins (cPCDH) leads to millions of possible combinations of cPCDH isoforms on the neuron surface(Kaneko et al., 2006, Esumi et al., 2005), suggesting cPCDHs function as a unique barcode for each neuron(Yagi, 2012). The absence of γ-PCDH does not cause general abnormality in the development of the cerebral cortex, such as cell differentiation, migration, and survival(Wang et al., 2002). However, the presence of γ-PCDH has been detected at the synaptic contacts(Fernandez-Monreal et al., 2009, Phillips et al., 2003, LaMassa et al., 2021), and its absence has substantial effects on neuronal connectivity(Tarusawa et al., 2016, Kostadinov and Sanes, 2015, Lv et al., 2022). The homophilic property of γ-PCDH allows them to promote exuberant dendritic complexity(Molumby et al., 2016). However, more evidence suggests that it might be detrimental to synapse formation. Previous studies indicated that homophilic interactions mediated by large overlapping patterns of cPCDH isoforms on opposing cell surfaces might lead to intercellular repulsion(Rubinstein et al., 2015, Brasch et al., 2019, Honig and Shapiro, 2020, Lefebvre et al., 2012), in which the interaction with the neuroligin family may be involved(Molumby et al., 2017, Steffen et al., 2021).

Following this repulsion concept, the absence of γ-PCDH caused significantly more dendritic spines and inhibitory synapse densities in neocortical neurons(Molumby et al., 2017, Steffen et al., 2021). Consistently, neurons with overexpressed one of the γ-PCDH isoforms had significantly fewer dendritic spines from the same condition(Molumby et al., 2017). Furthermore, the absence of the clustered PCDH also increased local reciprocal neural connection between lineage-related neurons in the neocortex(Tarusawa et al., 2016, Lv et al., 2022), even though sister cells had more similar expression patterns of γ-PCDH isoforms(Lv et al., 2022).

Although γ-PCDH has these “contradictory” effects on dendrites’ complexity and dendritic spines, it is detrimental to synapse formation in the forebrain. However, each neuron expresses multiple isoforms of γ-PCDH(Kaneko et al., 2006, Lv et al., 2022). What could be the impact of this combinatorial expression on synapse formation? Here, using 5-prime end (5’-end) single-cell sequencing, we revealed the diversified combinatorial expression of γ-PCDH isoforms in neocortical neurons. With multiple whole-cell patch-clamp recordings after the sequential in utero electroporation, we discovered that the combinatorial expression of γ-PCDHs plays a vital role in synaptic specificity to assemble the neural microcircuit in the mouse neocortex. This is achieved by allowing a neuron to choose which partners not to form a synapse with rather than choosing which ones to make synapses with.

Results

The diversified combinatorial expression pattern of γ-PCDHs in neocortical neurons revealed by 5’-end single-cell sequencing

The gamma isoform of cPCDHs (γ-PCDHs) is critical for synaptic connectivity(Kostadinov and Sanes, 2015, Tarusawa et al., 2016, Lv et al., 2022). To determine the role of γ-PCDH in the neocortex, we examined their expression in the neocortical neurons of postnatal mice. Previous studies have suggested that cPCDHs were stochastically expressed in neurons(Hirayama et al., 2012, Toyoda et al., 2014). Here, we applied a 5’-end single-cell RNA sequencing to identify γ-PCDH isoforms by sequencing their variable exon (exon 1), where they differ from each other(Kohmura et al., 1998, Wu and Maniatis, 1999). Since the second postnatal week is the critical stage for synapse formation in the rodent neocortex(Lendvai et al., 2000, Holtmaat and Svoboda, 2009), we chose postnatal day 11 (P11) to examine their expression. After reverse transcription and cDNA amplification (Fig. 1-S1A, B), we split cDNA into two parts: one for specific amplification of pcdhg mRNAs and the other for the 5’ gene expression library construction (Fig. 1-S1C). After the cluster analysis (Fig. 1-S2-4), we collected 6505 neurons out of 17438 cells (Fig. 1A and Fig. 1-S1D). Neurons with more than 10 UMI (cutoff>1 for each type of individual isoform) for all γ-PCDH (Fig. 1B, C) were further analyzed. We found that significantly more cells expressed “C-type” isoforms (C3, C4, and C5) compared to other “variable” isoforms (Fig. 1D and Fig. 1-S1E), which is consistent with a previous study(Toyoda et al., 2014). We performed the pair-wise analysis on similarity between neurons in the single-cell expression pattern of γ-PCDH variable isoforms, which revealed that most neocortical neurons had distinct combinatorial expression patterns (Fig. 1E). Based on the variance analysis(Wada et al., 2018) of fraction distribution of the number of the expressed isoforms per cell for all neurons (Fig. 1F and Fig. 1-S5), we found a weak but significant co-occurrence for the expression of γ-PCDH isoforms in most neurons (Fig. 1G). We did not detect any apparent differences among clusters except for cluster 0, which had much less expression (Fig. 1C) and no co-occurrence of γ-PCDH isoforms (Fig. 1G and Fig. 1-S5). Since all of our electrophysiological recordings were carried out on pyramidal neurons in the layer 2/3 of the neocortex, we analyzed the corresponding cluster 7 in more detail. They also showed distinct expression patterns as a general population of neocortical neurons (Fig. 1-S6). In summary, the data from 5’-end single-cell sequencing reveal that γ-PCDHs are diversely expressed in most neocortical neurons with a combinatorial pattern.

Figure 1:

The absence of γ-PCDHs led to an increase in local synaptic connectivity among pyramidal neurons

To investigate the function of γ-PCDH in the synaptic formation between neocortical neurons, we conducted paired recordings on pyramidal neurons in the layer 2/3 of the neocortex from γ-PCDH conditional knockout (cKO) mice. These mice were obtained by crossing Pcdhg ^flox/flox mice(Lefebvre et al., 2008) with Nex-cre mice(Goebbels et al., 2006), in which all variable and C-type γ-PCDH isoforms were specifically removed in pyramidal neurons. Using multiple whole-cell patch-clamp recordings from cortical slices of P9-32 mice, we measured the connectivity among nearby pyramidal cells (<50 μm between cell somas) in the layer 2/3 of the neocortex by the presence of evoked monosynaptic responses (Fig. 2 and Fig. 2-S1). As shown in the sample traces in the connectivity matrix from six recorded neurons (Fig. 2A and Fig. 2-S1), two neuronal pairs (the neuronal pairs 4→3 and 5→6) exhibited unidirectional monosynaptic connections (indicated by orange arrows), and one pair (the neuronal pair 1↔3) with bidirectional connection (indicated by green arrows) out of 15 possibilities. Overall, we found that the percentage of connected pairs was significantly higher in Pcdhg cKO mice (20.2%, 103/511) than in wild-type (WT) mice (15.0%, 122/813). Due to their potential different roles in the synaptic organizations in the neocortex(Douglas and Martin, 2004), we performed an extra set of recordings in the layer 2/3 of the neocortex from P10-20 mice to separate neuron pairs along their axes vertical vs. horizontal to the pial surface in Pcdhg cKO mice. We found a more significant difference in the connectivity for vertically aligned cells (18.3%, 94/515 in Pcdhg cKO mice vs. 11.2%, 73/651 in WT mice for vertically aligned neuron pairs; 12.2%, 27/221 in Pcdhg cKO mice vs. 9.5%, 28/294 in WT mice for horizontally aligned neuron pairs) (Fig. 2C). We also generated Pcdha cKO mice (Fig. 2-S2-4) and performed the same set of experiments for vertically aligned neurons. We did not find a significant difference (11.3%, 38/337 in Pcdha cKO mice vs. 14.3%, 26/182 in WT mice, Fig. 2D) in the mice without α-PCDH. In a more detailed analysis of synaptic connections among vertically aligned neurons in Pcdhg cKO mice, we found that the absence of γ-PCDH expression appeared to significantly increase synapse formation between cells separated vertically by 50 to 100 μm (24.0%, 50/208 in Pcdhg cKO mice vs. 9.6%, 20/208 in WT mice) (Fig. 2E). Furthermore, we found that the absence of γ-PCDH increased synapse formation starting from P10 (16.1%, 32/199 in Pcdhg cKO mice vs. 8.7%, 20/230 in WT mice for P10-12; 21.6%, 22/102 in Pcdhg cKO mice vs. 12.8%, 29/227 in WT mice for P13-15), when chemical synapses between cortical neurons become more detectable, and such tendency maintained for as long as measurements were performed (up to P20) (Fig. 2F). Thus, γ-PCDH may play a role in preventing synapse formation starting from the early stage of neural development.

Figure 2:

Overexpression of γ-PCDHs decreases local synaptic connectivity in the mouse neocortex

To further determine the effect of γ-PCDHs on synapse formation, we overexpressed randomly selected single or multiple γ-PCDH isoforms tagged with fluorescent proteins through in utero electroporation in mice. We then performed multiple whole-cell patch-clamp recordings to address the effect of γ-PCDHs on pair-wise synaptic connectivity for neurons in the layer 2/3 of the neocortex.

To confirm the overexpression effect, we performed the assays of RT-qPCR (Fig. 3A, B), single-cell RT-PCR (Fig. 3C, D), and immunohistochemistry (Fig. 3-S1A to C). The RT-qPCR assay revealed that the electroporated isoforms, but not the non-overexpressed ones, from the surgical side indeed had significantly higher expression levels than those from the contralateral side (Fig. 3A, B). To estimate how many isoforms were expressed in a given neuron when multiple plasmids were electroporated, we tagged first five isoforms with mNeongreen and the sixth one with mCherry (Fig. 3-S1A). Based on the rates of yellow and red-only cells in the total electroporated neuron population (Fig. 3-S1B), the probability analysis revealed that each positive neuron expressed an average of 5.6 types of isoforms (Fig. 3-S1C and the top panel of D). This result was consistent with the data from single-cell RT-PCR analysis in which an average of 5.3 types out of 6 electroporated isoforms was detected from 19 neurons (Fig. 3C, D and the bottom panel of Fig. 3-S1D). The results from single-cell RT-PCR also suggest that electroporation-introduced isoforms are the dominant ones in these neurons (Fig. 3D).

Figure 3:

Since the layer distribution of neocortical neurons plays a critical role in their synaptic connection(He et al., 2015), we examined their positions relatively to the pial surface after the overexpression. We found that overexpressing γ-PCDH isoforms did not affect cell positions in the neocortex compared to the control plasmids (Fig. 3-S1E, F).

We then used multiple recordings to examine the impact of γ-PCDHs on synapse formation. Overexpressing of one or six γ-PCDH variable isoforms in neurons significantly reduced the synaptic connection rate among them (10.3%, 15/146 in expressing control plasmids; 1.9%, 4/216 in overexpressing one isoform; and 4.4%, 8/181 in overexpressing six isoforms, red bars in Fig. 3E). However, overexpressing γ-PCDH C4 did not affect the synaptic connection rate (11.3%, 12/106 in overexpressing γ-PCDH C4, Fig. 3E). To exclude further the potential effect of C-types of γ-PCDHs’ impacts in synapse formation, we electroporated six variable isoforms in pcdhg cKO mice. The overexpression also led to the reduction of the connection rate in pcdhg cKO mice (6.1%, 12/198 in overexpressing six isoforms vs. 16.5%, 31/188 in expressing control plasmids), similar to what we observed in WT littermate mice (6.6%, 8/121 in overexpressing six isoforms vs. 16.5%, 18/109 in expressing control plasmids) (Fig. 3F). These observations reveal that overexpressing the variables, but not the C-types of γ-PCDHs isoforms, decrease synaptic connectivity in the mouse neocortex.

Combinatorial expression of γ-PCDHs regulates synaptic specificity in the mouse neocortex

To investigate the impact of the combinatorial expression pattern of γ-PCDH isoforms on synapse formation, we conducted a sequential in utero electroporation at E14.5 and E15.5 (Fig. 4A). By electroporating the same or different combinations of γ-PCDH variable isoforms on these two days, we aimed to manipulate the expression similarity level between two groups of neurons. Isoforms were randomly selected based on their comparable expression, except for isoforms A5 and B8, which had low expression levels according to the single cell sequencing data (Fig. 1D and Fig. 1-S1E). We assigned their combinations to achieve similarity levels ranging from 0% (no overlap, completely different) to 100% (complete overlap, all identical) (Fig. 4B). We used fluorescent proteins mNeongreen and mRuby3(Bajar et al., 2016) to tag isoforms for electroporation at E14.5 and E15.5 respectively (Fig. 4A). Subsequently, whole-cell patch-clamp recordings were conducted on layer 2/3 neurons in the neocortex from acute brain slices containing both green and red cells from P10-14 pups. Each set of recordings included at least one mNeongreen⁺, one mRuby3⁺, and one nearby control neuron without fluorescence (Fig. 4C). Our results showed that neurons with the same color exhibited significantly lower connectivity (Fig. 4-S1A), which aligns with the findings from single electroporation (as shown in Fig. 3E). In the complete overlap (100%) group, the connectivity rate between neurons electroporated at different days was also significantly lower compared to the control (3.7%, 5/134 in the complete-overlap group; 12.5%, 19/151 in control pairs, 100% in Fig. 4D). However, as the similarity levels decreased from 100% to 0%, the connectivity probabilities progressively returned to the control level. The likelihood recovered to 8.4% (14/165) for the pairs with a 33% similarity level and 10.3% (12/117) and 10.5% (12/114) for pairs with 11% or 0% similarity level, respectively (Fig. 4D). These observations suggested that it is the similarity level of γ-PCDH isoforms between neurons, but not the absolute expression of the protein in individual neurons, that regulate the synaptic formation. Similar to the single electroporation experiment (gray bars in Fig. 3E), there were no significant changes in the synaptic connectivity between electroporated and nearby control neurons (Fig. 4-S1B). Overall, our findings demonstrate a negative correlation between the probability of forming synaptic connections and the similarity level of γ-PCDH isoforms expressed in neuron pairs (Fig. 4E). These findings strongly suggest that the diversified combinatorial expression of γ-PCDH isoforms is crucial for synapse formation between nearby pyramidal cells. The more similar the γ-PCDH isoforms patterns expressed in neurons, the lower the probability of forming synapses between them (Fig. 4F).

Figure 4:

Discussion

Homophilic proteins cPCDHs are strong candidates for promoting synaptic specificity due to their combinatorial and stochastic expression pattern(Yagi, 2012, Kohmura et al., 1998, Toyoda et al., 2014). Our 5’-end single-cell sequencing data provided insights into the combinatorial expression pattern of γ-PCDH isoforms in neocortical neurons. We further demonstrated the critical role of this diversity in synaptic specificity through three lines of evidence. Firstly, the absence of γ-PCDH significantly increased functional connectivity between adjacent neocortical neurons. Secondly, electroporation-induced overexpression of identical γ-PCDH variable isoforms in developing neurons markedly decreased their connectivity. Lastly, using sequential in utero electroporation with different batchs of isoforms, we found that increasing the similarity level of γ-PCDH variable isoforms expressed in neurons led to a reduction in their synaptic connectivity. These findings suggest that γ-PCDHs regulate the specificity of synapse formation by preventing synapse formation with specific cells, rather than by selectively choosing particular targets. It remains to be studied whether the diversified patterns of γ-PCDH isoforms expressed in different neurons have additional coding functions for neurons beyond their homophilic interaction.

Previous studies by Molumby et al. demonstrated that neurons from the neocortex of pcdhg knockout mice exhibited significantly more dendritic spines, while neurons overexpressing a single γ-PCDH isoforms had fewer dendritic spines in (Molumby et al., 2017). Our recordings are consistent with these previous morphology studies. Tarusawa et al. revealed that the absence of the whole cluster of cPCDH affected synaptic connections among lineage-related cells(Tarusawa et al., 2016). More recently, overexpression of the C-type γ-PCDH isoform C3 also showed a negative effect on synapse formation within a defined clone(Lv et al., 2022). In our study, we further demonstrated an increased synaptic connection rate between adjacent pyramidal neurons in the neocortex of pcdhg knockout mice, while it decreased between neurons overexpressing single or multiple identical γ-PCDH variable isoforms. These effects were not just limited to lineage-dependent cells. Together with previous findings(Molumby et al., 2017, Tarusawa et al., 2016), our observations solidify the repulsion effect of γ-PCDH on synapse formation among neocortical neurons. Some subtle differences exist between our findings and previous recordings(Tarusawa et al., 2016, Lv et al., 2022). Tarusawa et al. demonstrated that the connection probability between excitatory neurons lacking the entire cPCDH cluster in the layer 4 was approximately twofold higher at early stage P9-11, significantly lower at P13-16, and similar to control cells at P18-20 compared (Tarusawa et al., 2016). In our study, pcdhg^-/- pyramidal neuronal pairs consistently exhibited a higher connection probability from P10 to P20. Two potential reasons could explain these differences. Firstly, we only removed γ-PCDH instead of the entire cPCDH cluster, which includes α, β, and γ isoforms. Secondly, γ-PCDH might have different functions in neurons located in the layer 2/3 compared to the layer 4. Lv et al. found that overexpression of C-type γ-PCDH C3 decreased the preferential connection between sister cells(Lv et al., 2022). However, our study demonstrated that only variable, but not C-type isoforms had a negative impact on synapse formation. This discrepancy might be attributed to the lineage relationship, which could have an unknown impact on synapse formation. Since a single neuron can express multiple isoforms, deleting all γ-PCDH isoforms might mask the role of this combination. In this study, we manipulated the combinatorial expression patterns of γ-PCDH isoforms in nearby neocortical neurons through sequential in utero electroporation, expressing different batches of isoforms with adjustable similarities. We observed that when two neurons expressed identical variable isoforms (100% group), the likelihood of synapse formation between them was lowest. As the similarity level between two cells decreased, with fewer shared isoforms, the connectivity probability increased. The connectivity probability between neurons with different variable isoforms (0% group) did not differ from the control pairs (without overexpression). However, the connections between overexpressed and control neurons were not affected under both 100% and 0% similarity conditions. These observations suggest that the similarity level, rather than the absolute expression of the protein, affects synapse formation between neurons.

Our findings also demonstrated that the overexpression of multiple γ-PCDH variable isoforms in one neuron only affected its connection if the other neuron overexpressed an identical combination of PCDHg isoforms. This indicates that the expression of multiple PCDHg variable isoforms in a neuron assists in selecting its synapse partner, emphasizing the essential role of diversified combinatorial expression of γ-PCDHs in synaptic specificity in the mouse neocortex. Moreover, the overexpression of the γ-PCDH C4 isoform did not affect synaptic connections, while overexpression of six variable isoforms led to a reduction in the connection rate in pcdhg cKO mice. These findings suggest that the variables, but not the C-type isoforms of γ-PCDHs, play crucial roles in the synapse formation in the mouse neocortex.

While the absence of γ-PCDHs causes significantly more synaptic formation among neocortical pyramidal neurons, evidence supports that their absence also leads to a significant reduction of synapse formation in other brain regions. For example, mice lacking γ-PCDHs exhibit fewer synapses in spinal cord interneurons(Weiner et al., 2005). Knocking down γ-PCDHs causes a decline in dendritic spines in cultured hippocampal neurons(Suo et al., 2012) and decrease the astrocyte-neuron contacts in the coculture from the developing spinal cord(Garrett and Weiner, 2009). The absence of γ-PCDHs leads to the reduction of dendritic arborization and dendritic spines in olfactory granule cells(Ledderose et al., 2013). Additionally, immuno-positive signals for γ-PCDHs are more frequently detected in mushroom spines than in thin spines(LaMassa et al., 2021). Moreover, we observed that the absence of γ-PCDHs impacted vertical-aligned neurons more than horizontal-situated pairs in the neocortex. These findings suggest that different mechanisms may be employed by synapses in different brain regions to achieve their specificity. Notably, different mechanisms have already been proposed for targeting specific inhibitory neural circuits in the neocortex, including “on-target” synapse formation for targeting apical dendrites and “off-target” synapse selective removal for somatic innervations(Gour et al., 2021).

In summary, our data demonstrate that the similarity level of γ-PCDH isoforms between neocortical neurons is critical for their synapse formation, with neurons expressing more similar γ-PCDH isoforms patterns exhibiting a lower probability of forming synapses between them. This suggests that the presence of γ-PCDHs enables neocortical neurons to choose which neurons to avoid synapsing with, rather than selecting specific neurons to form synapses with. Whether there are specific attractive forces between cells to promote synaptic specificity remains an open question.

Acknowledgements

We thank Dr. Yifeng Zhang for providing the Pcdhg flox/flox mouse line and Dr. Zilong Qiu for Nex-cre mouse line, Dr. Jun Chu for sharing plasmids with mNeongreen and mRuby3, and other members in Xu lab for their discussions and technique supports. We are grateful to Prof. Mu-ming Poo and Song-hai Shi for critical reading of the manuscript. This work was supported by grants from the Training Program of the Major Research Plan of the National Natural Science Foundation of China, grant No. 91632101; Strategic Priority Research Program of the Chinese Academy of Sciences, grant No. XDB32010100; National Natural Science Foundation of China project 31671113; Shanghai Municipal Science and Technology Major Project,grant No. 2018SHZDZX05,the State Key Laboratory of Neuroscience and the Lingang Laboratory, grant No. LG-GG-202201-01.

Author contributions

H. -T. X. conceived the project. Y. -J. Z. performed most recordings and single-cell analysis. C. -Y. D. and L. F. performed part of recordings. Y. -Q. W. prepared PCDH plasmids. H. Z. prepared mice. Y. -J. Z. and H. -T. X. wrote the manuscript.