Heparan sulfate molecules mediate synapse formation and function of male mating neural circuits in C. elegans

The nervous system regulates complex behaviors through a network of neurons interconnected by synapses. How specific synaptic connections are genetically determined is still unclear. Male mating is the most complex behavior in C. elegans. It is composed of sequential steps that are governed by more than 3,000 chemical connections. Here we show that heparan sulfates (HS) play a role in the formation and function of the male neural network. Cell-autonomous and non-autonomous 3-O sulfation by the HS modification enzyme HST-3.1/HS 3-O-sulfotransferase, localized to the HSPG glypicans LON-2/glypican and GPN-1/glypican, was specifically required for response to hermaphrodite contact during mating. Loss of 3-O sulfation resulted in the presynaptic accumulation of RAB-3, a molecule that localizes to synaptic vesicles, disrupting the formation of synapses in a component of the mating circuits. We also show that neural cell adhesion protein neurexin promotes and neural cell adhesion protein neuroligin inhibits formation of the same set of synapses in a parallel pathway. Thus, neural cell adhesion proteins and extracellular matrix components act together in the formation of synaptic connections. Author Summary The formation of the nervous system requires the function of several genetically-encoded proteins to form complex networks. Enzymatically-generated modifications of these proteins play a crucial role during this process. These authors analyzed the role of heparan sulfates in the process of synaptogenesis in the male tail of C. elegans. A modification of heparan sulfate is required for the formation of specific synapses between neurons by acting cell-autonomously and non-autonomously. Could it be that heparan sulfates and their diverse modifications are a component of the specification factor that neurons use to make such large numbers of connections unique?


Introduction
Behaviors are the result of a combination of signaling pathways coordinated at various cellular and tissue levels. Male mating is the most complex behavior of C.
elegans. This behavior is governed by 144 neurons and 64 muscles in the posterior part of the male worm that are extensively inter-connected to each other resulting in approximately 3,200 connected cell pairs (JARRELL et al. 2012). How these synapses are determined so that they are formed in reproducible patterns is still unknown. One hypothesis is the chemo-affinity hypothesis proposed by Sperry (SPERRY 1963), which postulates that matching pairs of cell adhesions molecules between the presynaptic and postsynaptic neurons will promote synaptogenesis. These interactions occur in the context of the surrounding extracellular matrix (ECM), the influence of which needs to be taken into account.
The ECM plays an important role in the development of the nervous system (PORCIONATTO 2006;ZIMMERMANN AND DOURS-ZIMMERMANN 2008;MYERS et al. 2011). Heparan sulfate proteoglycans (HSPG) are components of the ECM that function in several processes such as neurogenesis, cell migration, axon guidance, dendritic branching and synapse formation (BERNFIELD et al. 1999;BÜLOW AND HOBERT 2006;POULAIN AND YOST 2015). HSPGs exist in three different forms: (1) transmembrane proteins such as the syndecans; (2) glycosylphosphatidylinositol (GPI) anchored proteins such as the glypicans; and (3) secreted forms such as perlecan, agrin and collagen XVIII (BERNFIELD et al. 1999;BÜLOW AND HOBERT 2006;POULAIN AND YOST 2015). A special feature of HSPG is the presence of heparan sulfate (HS) chains that are attached to the core proteins. These HS chains are linear glycosaminoglycan polysaccharides composed of varying numbers of hexuronic acid and glucosamine repeating units (n = 50-150).
These polysaccharide chains are subjected to modifications including deacetylation, sulfation, epimerization, and acetylation. The formation of these modifications is catalyzed by heparan sulfate modification enzymes (HSME) (LINDAHL AND LI 2009). Deacetylation is mediated by N-deacetylase/N-sulfotransferase (Ndst) enzymes, sulfation is mediated by 2-O, 6-O and 3-O HS sulfotransferases, and epimerization is mediated by the glucuronic C-5 epimerase. These modifications form domains of distinctly modified HS within the HS chains that serve as binding sites for ligands and receptors. In this way they mediate specific biological functions, such as cell migration and axon guidance (BENNETT et al. 1997;PRATT et al. 2006;MATSUMOTO et al. 2007;KASTENHUBER et al. 2009). Some of these domains may be conserved throughout evolution (ATTREED et al. 2016).
The role of HSPG in synapse formation and function is not well-understood. In vertebrates, the HSPG agrin plays a role in the development of neuromuscular junctions by promoting the aggregation of acetylcholine receptors in skeletal muscle and activating the receptor tyrosine kinase Musk on the muscle surface (GLASS et al. 1996). The presence of agrin has also been detected in the central nervous system, where the suppression of agrin expression in cultured hippocampal neurons and in the cortex of mice results in the formation of fewer synapses (FERREIRA 1999;BOSE et al. 2000;KSIAZEK et al. 2007). The intracellular domains of syndecan-2 interact with various cellular components to promote filopodia formation and induce dendritic spine formation (ETHELL et al. 2001;LIN et al. 2007). In Drosophila, the extracellular and cytoplasmic domains of syndecan act postsynaptically to regulate synapse growth of neuromuscular junctions (NGUYEN et al. 2016). Glypicans have also been implicated in the formation of synapses. Glypican-4 (Gpc4) and glypican-6 (Gpc6) secretion from astrocytes is sufficient to induce functional synapses in retinal ganglion cells, while their elimination reduces the postsynaptic activity induced by these molecules (ALLEN et al. 2012). In addition, Gpc4 interaction with a postsynapatic leucine-rich repeat transmembrane protein Lrrtm4 is required to induce excitatory synapse formation (DE WIT et al. 2013). This interaction is mediated by an HS-dependent interaction between Gpc4 and the receptor protein tyrosine phosphatase PTPσ in the presynaptic site (KO et al. 2015).
Additional studies have demonstrated the involvement of HS chain modifications in the process of synapse formation. An RNAi screen in Drosophila, specifically directed to glycan genes, revealed that the functionally paired HS 6-O sulfotrasferase (hs6st) and HS 6-O endosulfatase (sulf1) have opposite effects in synaptic functional development of neuromuscular junctions (DANI et al. 2012). In mammals, the elimination of Ext1, a gene encoding an enzyme essential for heparan sulfate synthesis, causes the attenuation of excitatory synaptic transmission in amygdala pyramidal neurons. In humans, this results in autism-like behavioral deficits (IRIE et al. 2012).
Here we investigated the role of HS molecules in the development and function of the posterior male nervous system in C. elegans. We show that loss of the HS 3-O sulfotransferase HST-3.1 and the glypicans LON-2/glypican and GPN-1/glypican result in defects in response to hermaphrodite contact during male mating behavior, suggesting that 3-O-sulfated HS attached to LON-2/glypican and GPN-1/glypican is required for this process. In addition, HS molecules and their modifications, with the exception of 3-O sulfation, were required for the dorsoventral axonal migration of male-specific sensory neurons that are essential for male mating behavior and function. Loss of 3-O sulfation in the postsynaptic cell resulted in accumulation of a presynaptic vesicle marker in the presynaptic cell of a mating circuit. Further, synapse formation between male-specific sensory neurons and target interneurons was disrupted, possibly accounting for the observed behavioral defect.

Genetic elimination of HS-modification enzymes and HS proteoglycans causes behavioral defects in male mating
Behavior provides a sensitive readout of developmental or functional disruptions.
We conducted a response to hermaphrodite contact behavioral assay of male worms carrying null mutations for the HS-modification enzymes. Response to hermaphrodite contact is the decision of a male worm to start mating after contact by pressing its tail against the hermaphrodite body while moving backward searching for the vulva ( Figure   1A). We observed that males carrying a mutation in hst-2/HS 2-O-sulfotransferase, hse-5/HS C-5-epimerase and hst-3.1/HS 3-O-sulfotransferase showed deficiency in response to contact. Loss of hst-2/HS 2-O-sulfotransferase or hse-5/HS C-5-epimerase, which respectively introduce 2-O-sulfation and C-5 epimerization in the hexuronic acid of the linear glycosaminoglycan heparan sulfate chains, mildly affected the response to hermaphrodite contact. Forty five percent of hst-2(ok595) and 55% of hse-5(tm472) mutant males failed to respond after tail contact compared to a 10% failed response in wild type worms ( Figure 1B). For mutants in hst-3.1(tm734), which introduces 3-O-8 sulfation, 82% of the hst-3.1(tm734) mutant males failed to respond ( Figure 1B). This reduction in response to contact results in a reduced number of cross progeny ( Figure   S1A), demonstrating that the hst-3.1(tm734) mutant defect in response reduced the overall ability to succeed in mating. Both response and mating potency defects were rescued by a fosmid containing the hst-3.1 locus ( Figure 1C and S1A). The hst-3.1(tm734) mutant deficiencies in response behavior involved a failure in backward movement after mate contact; all non-responsive mutant worms either failed to backup after hermaphrodite contact (52%) or showed a discontinued backward movement losing contact with the hermaphrodite (30%) ( Figure S1B). This defect in backward locomotion is not caused by a general defect in backing behavior as hst-3.1(tm734) mutant males did not show defects in the backing response to nose touch ( Figure S1C). Since hst-3.1(tm734) mutant worms showed severe defects in response to contact, subsequent steps of male mating behavior were not examined.
The abnormal response to mate contact due to loss of 3-O-sulfation was specific for 3-O-sulfation modifications catalyzed by hst-3.1/HS 3-O-sulfotransferase, since male worms of hst-3.2(tm3006), a null allele of the other 3-O-sulfatransferase, did not show defects in response after tail contact during mating ( Figure 1B). A double mutant for hst-3.1 and hst-3.2 did not enhance the abnormal response phenotype of hst-3.1(tm734) single mutant, suggesting that 3-O-sulfation by hst-3.1/HS 3-O-sulfotransferase, but not hst-3.2/HS 3-O-sulfotransferase, is required to mediate male mating behavior. Finally, hst-6/HS 6-O-sulfotransferase, which introduces 6-O-sulfation, is not necessary for response to hermaphrodite contact, as 90% of hst-6(ok273) mutant males showed a response after tail contact. Taken together, these results indicate that HS molecules modified by hst-3.1/HS 3-O-sulfotransferase, hst-2/HS 2-O-sulfotransferase and hse-5/HS C-5-epimerase, but not hst-6/HS 6-O-sulfotransferase, are required for response to hermaphrodite contact by the male worm.
Since hst-3.1 is a 3-O-sulfotransferase that modifies the HS chains on HSPG, we wanted to determine which HSPG may contain the epitope with 3-O sulfation that is required for response to hermaphrodite contact. We tested mutants of sdn-1/syndecan and the two forms of glypican, lon-2/glypican and gpn-1/glypican. The sdn-1(zh20) mutant males were not defective in response to hermaphrodite contact ( Figure 1D). Males containing the lon-2(e678) mutation, a hypomorphic allele, were significantly different from control worms as 60% showed a response to contact compared to approximately 90% for control males. This lon-2 defect is not due to the anatomical Lon phenotype as lon-1(e185) mutant males, which have an elongated body similar to lon-2, did not show defects in response to hermaphrodite contact ( Figure S1D). gpn-1(ok377) (a null mutation) mutant males were not significantly different from control worms in their response to hermaphrodites. However, the lon-2(e678) gpn-1(ok377) double mutant worms had a stronger mutant phenotype than lon-2 single mutants, indicating that gpn-1 also has a function that promotes response ( Figure 1D).

HS 3-O sulfation in the EF interneurons regulates male response to hermaphrodite contact
To define the focus of action of the hst-3.1/HS 3-O-sulfotransferase HSME in its role in response to contact during mating, we expressed the hst-3.1/HS 3-Osulfotransferase cDNA tissue-specifically in neurons, muscles and hypodermis. When we expressed hst-3.1/HS 3-O-sulfotransferase in neurons using the rgef-1 pan-neuronal promoter ( p rgef-1::hst-3.1) we observed rescue of the abnormal response to contact phenotype (Figure 2A-C). Similarly, when we expressed a hst-3.1/HS 3-Osulfotransferase cDNA in hypodermal tissue using the dpy-7 promoter ( p dpy-7::hst-3.1), we observed rescue of the mutant phenotype. However, when we expressed hst-3.1/HS 3-O-sulfotransferase in muscles using myo-3 promoter ( p myo-3::hst-3.1), we were not able to rescue the response defect. These results suggest that the expression of hst-3.1/HS 3-O-sulfotransferase in neurons or hypodermis, but not in muscles, is sufficient to regulate response behavior in the male worms.
To further delineate the neuronal focus of action, we used cell-specific promoters that drove expression of the hst-3.1/HS 3-O-sulfotransferase cDNA in subsets of neurons in the male tale. Cell ablation experiments indicated that the B-type ray sensory neurons are essential for male mating behavior, particularly for the response to hermaphrodite contact and vulva location steps (LIU AND STERNBERG 1995;BARR AND STERNBERG 1999;KOO et al. 2011). By examining double mutants, we found that genes such as pkd-2/Polycystin-2, lov-1/Polycystin-1 and klp-6/kinesin that act cell-autonomously in RnB neurons to mediate the response to hermaphrodite contact (BARR et al. 2001;PEDEN AND BARR 2005), act genetically in the same genetic pathway as hst-3.1/HS 3-Osulfotransferase ( Figure S2). However, expression of hst-3.1/HS 3-O-sulfotransferase in B-type ray neurons by using the pkd-2/Polycystin-2 cell-specific promoter ( p pkd-2::hst-3.1) was not sufficient to rescue the abnormal response phenotype ( Figure 2E). These results suggested that mutation in hst-3.1 compromised the function of B-type ray neurons but did not act in these neurons. Therefore, to determine whether hst-3.1/HS 3- We further tested whether hst-3.1/HS 3-O-sulfotransferase could act in serotonergic, dopaminergic, glutamatergic and GABAergic neurons by expressing it with the tph-1 promoter ( p tph-1::hst-3.1), cat-2 promoter ( p cat-2::hst-3.1), eat-4 promoter ( p eat-4::hst-3.1) or unc-47 promoter ( p unc-47::hst-3.1), respectively. Expression of hst-3.1 cDNA in serotonergic, dopaminergic and glutamatergic neurons did not rescue the abnormal response phenotype ( Figure 2I-K). However, expression of hst-3.1 cDNA in GABAergic neurons rescued the response to hermaphrodite contact defects ( Figure 2L). This is consistent with rescue in GABAergic EF interneurons (GENDREL et al. 2016).

HS 3-O sulfation is not required for axon guidance of the B-type ray sensory neurons
Since HS molecules are mediators of axon guidance in many neurons in C. elegans (BÜLOW AND HOBERT 2004;BÜLOW et al. 2008), we wanted to determine whether the observed behavioral defects in hst-3.1/HS 3-O-sulfotransferase mutant worms were the result of guidance defects of neurons involved in response behavior. As previously mentioned, the B-type ray sensory neurons are mediators of response to contact behavior and our findings suggest that hst-3.1/HS 3-O-sulfotransferase is acting from their main postsynaptic target, the EF interneurons, to mediate this behavior. To determine whether 3-O sulfation by HST-3.1/HS 3-O-sulfotransferase in the EF interneurons is acting as a guidance cue to regulate axon migration, we examined B-type ray neurons axonal processes in hst-3.1/HS 3-O-sulfotransferase mutant worms. To visualize B-type ray neuron processes, we used a cytoplasmic GFP reporter driven by the pkd-2/Polycystin-2 promoter. pkd-2/Polycystin-2 encodes a TRPP cation channel that is expressed in three types of male-specific sensory neurons: the RnBs, HOB and the CEMs.
In wild type worms, during mid to late L4 stage, ray neuron cell bodies migrate away from the posterior tail hypodermis and enter the lumbar ganglion (SULSTON et al. 1980) ( Figure S4A). The axons migrate out of the lumbar ganglion in a DV pathway, forming four to five circumferential commissures. On the ventral side, the axon terminals enter the pre-anal ganglion (PAG) where they contact their postsynaptic partners. In the case of R1B and sometimes R2B, the axons first migrate anteriorly before changing to the DV migration. Therefore, defects in AP migration of B-type ray neurons result in an anterior overextension of processes, while defects in DV migration result in the absence of commissures. As previously reported (JIA AND EMMONS 2006), these migratory defects are observed in mutant worms for the netrin signaling pathway, where unc-6(ev400) and unc-40(e271) single mutants showed 30% and 47% (respectively) of defects in AP migration, while both mutants showed 100% of defects in DV migration (Figure 3). We found that HS-modifications 2-O sulfation, 6-O sulfation and 5-C epimerization are required for B-type ray neurons axon guidance as defects are observed in both AP and DV migratory pathways of hst-2(ok595), hst-6(ok273) and hse-5(tm472) single mutant worms ( Figure S4C-D). In addition, SDN-1 is a regulator of B-type ray neurons axon guidance as defects in both AP and DV migration were observed in sdn-1(zh20) single mutants. The hst-2(ok595); hst-6(ok273) and sdn-1(zh20); lon-2(e678) double mutants showed highly penetrant AP and DV migration defects. In fact, the DV migration defects of these double mutants are comparable to those observed in netrin signaling mutants, suggesting that HS molecules and their distinct HS-modification patterns regulate B-type ray neurons axon guidance through the unc-6/netrin ligand system (Figure 3 and S4C-F).
In contrast, the axon morphology of B-type ray neurons in hst-3.1(tm734) single mutant worms was indistinguishable from control worms, with no defects observed in the AP or DV axon migration of RnB neurons (Figure 3). The same results were obtained when we examined the lon-2 gpn-1 double mutants, which are also defective for response behavior, where no defects in RnB neurons AP or DV axon migration were observed.
These results suggest that the defects in response to hermaphrodite behavior in hst-3.1/HS 3-O-sulfotransferase and glypican mutants are not a consequence of defective axonal projections of B-type ray neurons function. Taken together, our analysis of the roles of HS molecules in axon guidance suggests that specific HS-modifications are required for axonal migration of B-type ray neurons, whereas, 3-O sulfation by HST-3.1/HS 3-O-sulfotransferase is not required for this process, but may rather serve different functions.

HS 3-O sulfation mediates synapse formation of B-type ray sensory neurons
The observation that hst-3.1/HS 3-O-sulfotransferase mutants are defective in response to contact behavior during male mating, but do not display any obvious defects in axonal projections, prompted us to investigate whether HS 3-O sulfation is regulating synaptic function of the response circuits. To examine the possibility that hst-3.1/HS 3-Osulfotransferase is affecting synapse formation of B-type ray neurons, we investigated the presynaptic densities of mutant worms using an mCherry::RAB-3 reporter expressed under the control of pkd-2/Polycystin-2 promoter. rab-3 encodes a member of the Ras GTPase superfamily that localizes to presynaptic sites. To study RnB synapses, we used the confocal microscope and analyzed the synaptic pattern of compressed z-stacks containing all the RnBs synapses located in the preanal ganglion. We looked at the synapses of young adult male worms (64 hours after hatching) because by this time most B-type ray neuron synapses are formed ( Figure S5A). Based on the presynaptic puncta distribution in the preanal ganglion, the presynaptic density patterning of hst-3.1/HS 3-Osulfotransferase mutant males was similar in morphology to that of control worms, suggesting that the elimination of hst-3.1/HS 3-O-sulfotransferase does not affect the distribution of B-type ray neurons presynaptic sites ( Figure 4B). However, we found that the mCherry::RAB-3 levels are accumulated above control levels ( Figure 4C). To quantify the levels of mCherry::RAB-3 in synapses we performed a densitometry analysis of compressed z-stack images containing all RnB synapses in the preanal ganglion. To control for variability in transgene expression, we normalized by using a cytoplasmic GFP expressed from the same transgene and promoter. We observed that hst-3.1/HS 3-O-sulfotransferase mutants have a mCherry relative fluorescence of 1.47 ± 0.11 compared to 1 ± 0.03 in control worms ( Figure 4D).
The mCherry::RAB-3 levels in the B-type neuron synapses of lon-2 gpn-1 double mutant worms have a similar accumulation of RAB-3 with a relative fluorescence of 1.71 ± 0.13 compared to 1 ± 0.03 in control worms ( Figure 4D). The lon-2(e678) single mutant worms, which are defective for response to contact, also showed increased levels of RAB-3 in B-type ray neuron synapses when compared to control ( Figure 1D and S6B).
For the EF interneurons, we expressed an acceptor peptide AP fused N-terminally to nlg-1/neuroligin driven by the nlg-1/neuroligin promoter. We found that hst-3.1(tm734) mutants show less synaptic labeling compared to control worms, suggesting smaller or fewer synapses between RnBs and the EFs neurons. Control worms have a relative fluorescence of 1 ± 0.05, while the hst-3.1(tm734) mutants have a relative fluoresce of 0.26 ± 0.03 ( Figure 4E and 4F). Moreover, lon-2 gpn-1 double mutant worms also showed less synaptic labeling than control worms with a relative fluorescence of 0.75 ± 0.05. Altogether, our synaptic and behavioral results suggest that HS 3-O sulfation is involved in the process of synapse formation, which in turn may affect the function of the connection between B-type ray neurons and EF interneurons, resulting in the defects in response behavior during male mating.

1/glypican, genetically interact with synaptic molecules
The process of synapse formation is thought to be mediated by the interaction of cell adhesion molecules located in the presynaptic and postsynaptic sites. Such is the case for presynaptic neurexin and postsynaptic neuroligin adhesion molecules, which have been implicated in synaptogenic activity and synapse maturation (SCHEIFFELE et al. 2000;GRAF et al. 2004;SÜDHOF 2008). We found nrx-1/neurexin null mutants exhibited defects in response to contact during male mating similar to the defects observed in hst-3.1/HS 3-O-sulfotransferase mutant worms ( Figure 5A). nlg-1/neuroligin mutants, by contrast, did not show defects in response to contact behavior. To investigate whether hst-3.1 still retained a function in the absence of nrx-1 function, we constructed hst-3.1; nrx-1 double mutants. We found that the hst-3.1; nrx-1 double mutant further enhanced the defects observed in both single mutants, thus suggesting that these genes act in parallel genetic pathways, while not excluding the possibility that they act in the same pathway given the strong phenotype observed in the hst-3.1/HS 3-O-sulfotransferase single mutant ( Figure 5A).
The hst-3.1; nlg-1 double mutant suppressed the defect in response to contact during male mating observed in the hst-3.1 single mutant ( Figure 5A). Thus there is an hst-3.1-independent pathway that is suppressed by nlg-1. This pathway depends on nrx-1 function. We examined hst-3.1; nrx-1; nlg-1 triple mutant worms and observed that in the absence of nrx-1/neurexin function, a mutation in nlg-1/neuroligin no longer suppresses the response defects due to mutation in hst-3.1. Thus nrx-1/neurexin is epistatic and suggesting that nlg-1 acts to suppress the activity of nrx-1 that promotes response behavior. To further investigate the nlg-1/neuroligin suppression of defects in response behavior due to mutation in hst-3.1, we constructed triple mutants for nlg-1/neuroligin and the two glypicans, lon-2/glypican and gpn-1/glypican. The nlg-1 lon-2 gpn-1 triple mutants did not show defects in response to contact, thus indicating that nlg-1/neuroligin also suppresses the defects of lon-2 gpn-1 double mutants ( Figure 5A). Together, these results suggest that nrx-1/neurexin and nlg-1/neuroligin adhesion, possibly through opposing roles, are involved in regulating response behavior in a pathway parallel to that in which hst-3.1/HS 3-O-sulfotransferase and the glypicans act.
To determine whether the genetic interaction of these molecules is also regulating synapse formation, we examined the presynaptic levels of mCherry::RAB-3 in B-type ray sensory neurons in the same series of nrx-1/neurexin and nlg-1/neuroligin double and triple mutant worms. The hst-3.1; nrx-1 mutants showed an increased accumulation of mCherry::RAB-3, however the increase in accumulation was not significantly different from hst-3.1/HS 3-O-sulfotransferase single mutant worms ( Figure 5B). Considering that hst-3.1; nrx-1 double mutants' defects in behavior are significantly more severe than the defects in hst-3.1/HS 3-O-sulfotransferase single mutants, we conclude that either it is difficult to detect an even higher accumulation of mCherry::RAB-3 as measured by the relative fluorescence densities, or that behavioral defects due to nrx-1 mutation result from a disruption in another synaptic connection.
Consistent with the behavioral results, the hst-3.1; nlg-1 double mutant suppressed the mCherry::RAB-3 accumulation in the presynaptic cell, decreasing it to control levels ( Figure 5B). As we observed in response behavior, the hst-3.1; nrx-1; nlg-1 triple mutants showed an increased accumulation of mCherry::RAB-3 that was not suppressed by nlg-1/neuroligin, consistently indicating that in the absence of nrx-1, nlg-1 does not suppress the hst-3.1/HS 3-O-sulfotransferase induced defects in B-type ray neuron presynaptic sites. Lastly, nlg-1 lon-2 gpn-1 triple mutants showed mCherry::RAB-3 levels comparable to control worms, again mirroring the genetic interactions observed for the male mating behavior.

Discussion
How neuronal connectivity determines behavioral output in an organism remains one of the biggest questions in the neuroscience field. To address this question, one important aspect is to investigate the roles of molecules in the extracellular matrix that are involved in establishing and making these connections functional. In this work, we used male mating behavior in C. elegans as a read-out of synaptic function together with fluorescence labeling of synapses to study the role of HS molecules in formation of the male nervous system and its synaptic connectivity.

HS molecules mediate male mating behavior in C. elegans
We found that hst-3.1/HS 3-O-sulfotransferase is acting in the same genetic pathway as pkd-2/Polycystin-2, lov-1/Polycystin-1 and klp-6/Kinesin to regulate response to contact behavior during male mating, implicating compromise of B-type ray sensory neuron function as the basis of the phenotype. However, hst-3.1/HS 3-O-sulfotransferase focus of action is not the B-type ray neurons, but rather the downstream EFs malespecific interneurons or the hypodermis. Based on the double mutant defects, the HS proteoglycans LON-2/glypican and GPN-1/glypican act in parallel to mediate response behavior. Since it has been shown that gpn-1/glypican is expressed in neurons (HUDSON et al. 2006) and lon-2/glypican acts in the hypodermis to mediate different aspects of neuronal development (PEDERSEN et al. 2013), we propose that response to contact behavior is mediated by 3-O sulfation to the HS attached to GPN-1/glypican in neurons, and 3-O sulfation to the HS attached to LON-2/glypican in hypodermis. Even though glypicans possess a GPI anchor, it has been shown that LON-2/glypican can be shed from epidermal cells, secreted and diffuse into the extracellular matrix where they interact with components of Netrin signaling to mediate axon guidance (BLANCHETTE et al. 2015). It has been shown that the EFs interneurons are important for male exploratory behavior, which is essential for males to locate and contact their mating partners (BARRIOS et al. 2008). In terms of connectivity, their main synaptic input is from the Btype ray sensory neurons, while their main synaptic output is onto the AVB pre-motor interneuron (S.J. Cook, C.A. Brittin, T.A. Jarrell, Y. Wang, A.E. Bloniarz, et al., manuscript in preparation). Previous cell ablation studies have shown that the EFs interneurons mediate backward locomotion after mate contact (SHERLEKAR 2015), which is necessary for response to contact behavior. Because AVB is the pre-motor interneuron that promotes forward movement in the locomotion circuit, while EF ablation promotes forward movement after mate contact, it is thought that the EFs -> AVB connections are inhibitory synapses. Our findings support this hypothesis given that hst-3.1/HS 3-Osulfotransferase is acting in the EFs to regulate response behavior, and the hst-3.1/HS 3-O-sulfotransferase mutants have defects in backward locomotion after contacting the hermaphrodite.

HS 3-O sulfation regulates synapse formation of male mating neurons
Our findings, together with previously published studies, demonstrate that heparan sulfates are mediators of synapse development and function. In this work, we identify a HS motif with 3-O-sulfation likely attached to the HSPG LON-2/glypican and/or GPN-1/glypicans that is required for synaptogenesis. Our observations conceptually extend studies in cultured hippocampal neurons, where the LRRTM4's synaptogenic activity requires the presence of heparan sulfates (de Wit et al., 2013).
Moreover, they showed that glypican acts as a receptor for LRRTM4, and their interaction is important for the development of excitatory synapses.
Through the analysis of null mutations, we demonstrate that the elimination of hst-3.1/HS 3-O-sulfotransferase and the two glypican forms, lon-2/glypican and gpn-1/glypican, induces defects in the presynaptic specialization of B-type ray neurons, while it reduces the iBLINC synaptic labeling of RnBs -> EFs synapses, indicating their role in the process of synapse formation. Higher levels of the presynaptic marker RAB-3 in B-type ray neurons in hst-3.1/HS 3-O-sulfotransferase mutants could be because vesicle fusion is not occurring properly at the synapse. The higher levels of RAB-3 can be explained by a change in the size and morphology of the synaptic puncta rather than an increase in their number. This is suggested by the fact that the presynaptic pattern of Btype ray neurons in the pre-anal ganglion of mutant worms is comparable to the one observed in control worms. In addition, using the iBLINC labeling system, we showed that hst-3.1/HS 3-O-sulfotransferase mutants and lon-2 gpn-1 double mutants form fewer RnB -> EFs connections. In terms of synaptic function, from our results, we argue that the observed hst-3.1/HS 3-O-sulfotransferase and lon-2gpn-1 defect in response behavior during mating is a reflection of the observed defects in synapse formation.
The HS 3-O sulfation regulation during neuronal development of B-type ray sensory neurons seems to be specific for the process of synapse formation as no defects were observed in the axon morphology of these mating neurons. A similar restricted role of HS molecules in the process of synapse formation has been previously reported in mammals, where ext1 conditional knockout in mice results in autism-like behavioral phenotypes due to abnormal functioning of glutamatergic synapses while no detectable morphological defects were observed in the brain (IRIE et al. 2012). Even though there is no involvement of HS 3-O sulfation in the process of B-type ray neurons axon guidance, we found that other HS modifications such as 2-O sulfation, 6-O sulfation and C-5 epimerization mediate anteroposterior and dorsoventral axon guidance pathways by acting in parallel genetic pathways. This is consistent with previous findings that demonstrate the various functions of distinct HS modification patterns in the development of different neuronal cell types (SAIED-SANTIAGO et al. 2017). The simultaneous knockdown of three proteoglycans, sdn-1/syndecan, lon-2/glypican and gpn-1/glypican, severely affected axon guidance of B-type ray neurons demonstrating they act redundantly in this process, as is the case for other processes such as KAL-1/Anosmin 1 induced neurite branching in the AIY interneuron (DIAZ-BALZAC et al. 2014). In the context of dorsoventral migration of B-type neurons, genetic elimination of HS molecules causes defects similar to unc-6/netrin ligand and its unc-40/DCC surface receptor; possibly HS may regulate axon guidance through the netrin signaling pathway.

A role for neurexin and neuroligin
We have shown that neural cell adhesion proteins neurexin and neuroligin play a role in the formation of synapses by the RnB neurons. Neurexin appears to promote synapse formation while neuroligin opposes the function of neurexin. Interestingly, Hart and Hobert (HART AND HOBERT 2018) recently found a similar novel, mutually antagonistic role for these proteins in synapse formation elsewhere in the C. elegans male mating circuits. This neurexin/neuroligin pathway is to some extent independent of the pathway in which hst-3.1/ HS 3-O-sulfotransferase functions as each gene retains some function in null mutants of the other (Figure 6). In fact, in the absence of the inhibitory function of neuroligin, it appears that the synapse-promoting function of neurexin can fully restore synapse formation and function in the absence of hst-3.1/ HS 3-Osulfotransferase ( Figure 5). These results indicating multiple independent pathways promoting synapse formation point up the complexity of the process and help to understand how specificity and robustness are achieved.

Acknowledgements
We thank the laboratory members of the Emmons and Bülow lab for comments on the manuscript and helpful discussions; the Caenorhabditis Genetics Center (which is funded by the NIH, P40 OD010440) for strains; Oliver Hobert Lab for DNA clones. This

C. elegans strains and imaging
All strains were maintained using standard methods (BRENNER 1974). All strains used contain the him-5(e1490) mutation on chromosome V to increase the male population (Broverman 1994). We refer to him-5 as control worms. All experiments were performed at 20°C, and animals were scored as 1-day-old adults unless otherwise specified. The strains and mutant alleles used in this study are listed in the Supplemental Experimental Procedures. Fluorescent images were captured in live C. elegans using a Plan-Apochromat 40×/1.4 or 63x/1.4 objective on a Zeiss Axioimager Z1 Apotome.
Worms were immobilized using 10 mM sodium azide and Z stacks were collected.
Maximum intensity projections were used for further analysis.

Molecular biology and transgenesis
To assemble tissue specific expression constructs used for rescue experiments, the

Behavioral scoring
Response to hermaphrodite contact assays were performed with one day old adult, virgin males isolated at the L4 stage. Male worms to be tested were placed in a 10-mm food lawn with ten unc-31(e169) hermaphrodites. The mating behavior of the males was observed for 5 minutes and annotations were made every time a male responded to hermaphrodite contact. For this assay, a male is considered to have responded to contact if after mate contact it starts backward locomotion, scanning the hermaphrodite body followed by turning behavior. Males that failed to response to contact did not start backward locomotion to scan the hermaphrodite body, or they lost tail contact right after starting the backward locomotion. The quantitation of response to contact was performed by dividing the number of males with response by the total number of male tested [Response % = (# of males with response/total # of males evaluated) x 100. For statistical analysis, we performed a t-test to calculate the significant difference between control worms (him-5) and mutant worms (in a him-5 background).
For the mating potency assay, we placed one virgin young adult male worm with one pha-1 mutant hermaphrodite in a 10-mm food lawn for 4 hours. We scored a total of 50 male worms (50 plates). After 4 hours, males were removed from the plate and the plate was placed at 25°C. The pha-1 mutant worms are temperature sensitive and are not viable at 25°C, so only the crossed progeny grows at 25°C. Three days later, we counted the plates with worms that survived at 25°C. To calculate mating potency, we divided the number of cross progeny plates by the total number of plates/males tested [Mating potency %= (# plates/males with cross progeny/total # of plates/males tested) x 100]. For statistical analysis, we performed a t-test to calculate the significant difference of mating potency between wild type worms (him-5) and mutant worms (in a him-5 background).
For backing response after nose touch, we placed one virgin young adult male worm in a clean bacterial lawn and gently touched its nose 10 times with an eyelash, waiting 10 seconds between each touch. The number of times the worm showed response to nose touch by backward movement was scored and the backing response was calculated [Backing response= (# of times the worm backed up after nose touch/10 nose touches) x 100]. For statistical analysis, we performed a t-test to calculate the significant difference of backing response between wild type worms (him-5) and mutant worms (in a him-5 background).

RnB synapses imaging
To visualize the presynaptic pattern of RnB neurons, z-stack images of young adult males containing Ppkd-2::GFP and Ppkd-2::mCherry::RAB-3 reporters (bxIs30) were acquired using a Leica SP5 confocal microscope. The z-stack images were analyzed one by one from ventral to dorsal and compressed by looking at the synaptic puncta located in the preanal ganglion. To quantify the protein levels of mCherry::RAB-3, we performed a fluorescent densitometry analysis of compressed z-stacks images acquired by using a Zeiss Axioimager Z1 Apotome with a 63x/1.4 objective. For this analysis, we used the same exposure time for all control and mutant samples. The relative fluorescence values were measured by dividing mCherry densitometry (corresponding to synapses) by GFP densitometry (corresponding to axon terminals). In this way, we corrected for missing synapses that are a product of defects in axonal migration. For statistical analysis, we performed a t-test to calculate the significant difference between control worms and mutant worms.
To visualize the synapses between RnB ray sensory neurons and EF interneurons, we used the iBLINC trans-synaptic biotin transfer system. For the presynaptic RnB sensory neurons labeling, we expressed the biotin ligase with nrx-1 fusion protein (BirA::nrx-1) driven by the pkd-2/Polycystin-2 promoter. For the postsynaptic EF interneurons labeling, we expressed the biotinylated acceptor peptide with nlg-1/neuroligin fusion protein (AP::nlg-1) driven by the nlg-1/neuroligin promoter. We coexpressed these pre-and postsynaptic fusion proteins with streptavidin::RFP fusion protein driven by the unc-122 coelomocytes promoter. To quantify RnBs>EFs biotinylated synapses, we performed a fluorescent densitometry analysis of compressed z-stack images acquired from mutant and wild type worms by using a Zeiss Axioimager Z1 Apotome with a 63x/1.4 objective. Only those worms completely oriented in a dorsoventral position were imaged in order to obtained a clear view of the synaptic ring located in the preanal ganglion. The relative fluorescent (a.u.) values were the measurement of the RFP densitometry subtracted by the background (noise). The measured area consists of a circle with a 5 m radio. To compare the difference in synaptic densities between mutant and wild type worms, we divided the relative fluorescence (a.u.) value of mutants by the average of the relative fluorescence (a.u.) of wild type worms measured on the same experimental day. For statistical analysis, we performed a t-test to calculate the significant difference between wild type worms and mutant worms.

Data Statement
All strains and reagents are available upon request. Figure S1 gives analysis of male potency, response defects and general backing response in the hst-3.1/HS 3-Osulfotransferase. Figure S2 provides additional genetic data pertaining to the genetic interactions between hst-3.1/HS 3-O sulfotransferase, pkd-2/polycystin-2, lov-1/polycystin-1 and klp-6/kinesin. Figure S3 shows expression data of the hst-3.1/HS 3-Osulfotransferase transcriptional reporter and nlg-1/neuroligin transcriptional GFP fusion. Figure S4 provides additional genetic data pertaining to the role of HSME and HSPG in the axon guidance of B-type ray neurons. Figure S5 provides additional information about the visualization of the B-type ray neuron synapses with the EF interneurons. Figure S6 provides additional genetic data pertaining to the quantification of mCherry::RAB-3 in the HSME and HSPG mutants. Table S1 provides a complete list of strains created for this study. Tables S2 and S3 provides a complete list of transgenic strains, and the respective constructs, created for this study. File S1-S3 comprises all data used to create figures. (C) Quantification of a hst-3.1-containing fosmid rescue of the response to hermaphrodite contact during male mating behavior in the hst-3.1(tm734) mutant. Error bars denote the SEM; statistical significance is shown as follows: * p < 0.05; * * p < 0.005; * * * p < 0.0005; * * ** p < 0.00005; and ns, not significant. The data for control and hst-3.1 are identical to (B) and shown for comparison only. (E-L) Cell-specific rescue of response to hermaphrodite contact during male mating behavior in hst-3.1(tm734) mutants with hst-3.1 cDNA under heterologous promoters as indicated. Rescue was defined as restoration of response to hermaphrodite contact during male mating in transgenic animals (darker shade) and had to be statistically significant (p < 0.05) compared to non-transgenic siblings (lighter shade) (n ≥ 12). (B and C) Quantification of B-type ray neurons anteroposterior and dorsoventral axon guidance in the genotypes indicated. Error bars denote the SEM; statistical significance is shown as follows: * p < 0.05; * * p < 0.005; * * * p < 0.0005; * * ** p < 0.00005; and ns, not significant. (D) Quantification of mCherry::RAB-3 fluorescence in the preanal ganglion synaptic ring in the genotypes indicated. Error bars denote the SEM; statistical significance is shown as follows: * p < 0.05; * * p < 0.005; * * * p < 0.0005; * * ** p < 0.00005; and ns, not significant. The data presented is a ratio of mCherry::RAB-3 to GFP and control value.
(F) The data presented is the normalized RFP density in the synaptic area. Error bars denote the SEM; statistical significance is shown as follows: *p < 0.05; **p < 0.005; ***p < 0.0005; ****p < 0.00005; and ns, not significant. A) Quantification of response to hermaphrodite contact during male mating behavior in the genotypes indicated. Error bars denote the SEM; statistical significance is shown as follows: * p < 0.05; * * p < 0.005; * * * p < 0.0005; * * ** p < 0.00005; and ns, not significant. The data for control and hst-3.1/ HS 3-O-sulfotransferase are identical to figure 1 and shown for comparison only. B) Quantification of mCherry::RAB-3 fluorescence in the preanal ganglion synaptic ring in the genotypes indicated. The data presented is a ratio of mCherry::RAB-3 to GFP and control value. Error bars denote the SEM; statistical significance is shown as follows: * p < 0.05; * * p < 0.005; * * * p < 0.0005; * * ** p < 0.00005; and ns, not significant. The