Kirrel2 is differentially required in populations of olfactory sensory neurons for the targeting of axons in the olfactory bulb

The elaboration of precise neural maps in developing sensory systems allows organisms to obtain an accurate representation of signals that are detected in their environment. Sensory coding of olfactory cues in mice relies on the formation of stereotypic glomerular maps within the olfactory bulb (OB). Olfactory sensory neurons (OSNs) of the olfactory epithelium (OE) project their axons to the OB where they synapse with dendrites of second-order neurons in spatially conserved neuropil structures termed glomeruli. Each OSN expresses a single functional olfactory receptor (OR), and OSN axons expressing the same OR coalesce together in one or two stereotypically conserved locations of the OB. This occurs through a two-step process that involves the coarse targeting of these axons within defined regions of the OB, and their subsequent coalescence into specific glomeruli (reviewed by Takeuchi and Sakano, 2014).

Targeting of OSN axons to broad regions of the OB is mediated by axon guidance molecules and by their receptors that are expressed on the growing axons (Assens et al., 2016; Cho et al., 2007, 2011, 2012; Col et al., 2007; Cutforth et al., 2003; Imai et al., 2006, 2009; Nakashima et al., 2013; Nguyen-Ba-Charvet et al., 2008; Schwarting et al., 2000, 2004; Takeuchi et al., 2010; Walz et al., 2002; Zapiec et al., 2016). In contrast, coalescence of OSN axons expressing the same OR into a specific glomerulus has been suggested to rely on selective interactions of OSN axons with similar identities. Axonal identity may be established through the combinatorial expression of multiple families of cell-surface molecules that promote adhesion of similar axons or repulsion of dissimilar axons (Ihara et al., 2016; Kaneko-Goto et al., 2008; Mountoufaris et al., 2017; Serizawa et al., 2006). In addition, expression of the OR itself may provide a unique axonal identity that contributes to the coalescence of like axons through an as yet unidentified mechanism (Feinstein and Mombaerts, 2004; Feinstein et al., 2004; Movahedi et al., 2016; Rodriguez-Gil et al., 2015).

The differential expression of Kirrel family members on OSN axons has been suggested to serve as an axonal identity code to regulate coalescence of OSNs into glomeruli (Serizawa et al., 2006). Kirrel2 and Kirrel3 are expressed in a complementary and OR-correlated manner in OSN axons, whereby axons that express high levels of Kirrel2 express low levels of Kirrel3, and vice versa (Serizawa et al., 2006), and all axons of a single OR population will express the same levels of each Kirrel family member. Furthermore, their differential expression is regulated by OR-evoked activity (Serizawa et al., 2006). The main evidence implicating Kirrels in olfactory map formation comes from the mosaic overexpression of Kirrel2 or Kirrel3 in OSNs that express MOR28 (also known as Olfr1507), a manipulation that leads to the duplication of MOR28 glomeruli in the OB (Serizawa et al., 2006). Although these experiments suggest that the levels of expression of these proteins in OSNs can regulate axonal coalescence, the requirement of Kirrels in olfactory map formation needs to be addressed by analysis of genetic loss-of-function mutations.

Here, we use a genetic approach to ablate expression of Kirrel2 alone or in combination with Kirrel3 and study the projections of multiple OSN populations to their glomerular targets in different regions of the OB. We provide evidence that the differential expression of Kirrels in OSNs does not represent a universal mechanism to promote the coalescence of OSN axons, but rather that their expression contributes to the coalescence of subsets of OSN axonal populations, likely in a region-specific manner.

The differential expression of Kirrel family members between OSN populations has been proposed to contribute to a molecular code whereby OSN axons of a given population would be similarly coded and can find each other to coalesce into the same glomeruli (Serizawa et al., 2006). To assess whether Kirrel family members are necessary for OSN axonal coalescence, we examined the targeting of five populations of OSN axons that express either tau-lacZ or tau-GFP in mice carrying a targeted deletion of exons 3 and 4 in the Kirrel2 gene. We have previously shown that Kirrel2 protein expression is ablated and that coalescence of vomeronasal sensory neuron axons is altered in these mice (Brignall et al., 2018). We selected these five populations of OSN axons as they target to well-defined regions of the OB (Fig. 1W). Axons of S50 (Olfr545)- and MOR1-3 (Olfr66)-expressing OSNs innervate glomeruli in the most dorsal region of the OB, termed DI. M72 (Olfr160)- and MOR174-9 (Olfr73)-expressing axons project to glomeruli located slightly more ventrally, in a region designated DII. In addition, we examined the targeting of ventrally projecting MOR28-expressing axons, as overexpression of Kirrel2 or Kirrel3 in a mosaic pattern among these axons leads to glomeruli duplication (Serizawa et al., 2006).

We first examined the levels of Kirrel2 and Kirrel3 expression on OSN axons expressing specific ORs, as both Kirrel members have been reported to be differentially expressed in populations of OSNs (Serizawa et al., 2006). Although high levels of Kirrel2 were observed on S50- and MOR174-9-positive axons, medium and low Kirrel2 expression was observed on MOR1-3-, and M72- and MOR28-expressing axons, respectively (Fig. 1A-J,U). The pattern of Kirrel3 expression between these glomeruli was different to that of Kirrel2: MOR28 axons expressed high levels of Kirrel3 whereas S50-, MOR1-3-, MOR174-9- and M72-positive axons expressed lower levels of Kirrel3 (Fig. 1K-T,V).

To determine whether Kirrel2 is required for the targeting of OSN axons, we first examined the targeting of DI-projecting S50- and MOR1-3-positive axons in control and Kirrel2^−/− mice by whole-mount GFP epifluorescence and Xgal staining, respectively. Whole-mount assessment did not reveal a significant difference in the positioning of these glomeruli within the OB of Kirrel2^−/− mice (Fig. 2A-L). Furthermore, there was no significant increase in the number of S50 and MOR1-3 glomeruli in the OB of these mice, demonstrating that Kirrel2 is dispensable for the targeting of these two populations of axons in the OB (Fig. 2O-S). Similarly, the position and number of MOR28 glomeruli in the OB was also unchanged in Kirrel2^−/− mice (Fig. 2M-O,T,U).

Although ablation of Kirrel2 expression did not affect the coalescence of S50- and MOR1-3-positive axons, it remained possible that expression of Kirrel3 on these axons, or Kirrel3 upregulation in the absence of Kirrel2, might compensate for Kirrel2 loss in these OSNs. To test for potential compensation by Kirrel3, we crossed the Kirrel2 mice with mice carrying a targeted insertion of a GFP cassette in the first exon of the Kirrel3 gene, which lack Kirrel3 expression and show defects in the coalescence of vomeronasal sensory neuron axons (Prince et al., 2013). Assessment of the targeting of MOR1-3-positive axons in Kirrel2^−/−; Kirrel3^−/− mice revealed that both the location and number of MOR1-3 glomeruli were unchanged in these mice, demonstrating that both Kirrel2 and Kirrel3 are dispensable for the coalescence of these axons (Fig. 3A-D,I). Furthermore, ablation of both Kirrel2 and Kirrel3 did not affect the targeting of MOR28-positive axons (Fig. 3E-I).

In addition to axonal coalescence, Kirrel family members have been implicated in synapse formation in the hippocampus and accessory olfactory bulb (Shen and Bargmann, 2003; Shen et al., 2004; Martin et al., 2015; Brignall et al., 2018). We therefore assessed the number of excitatory and inhibitory synapses in glomeruli of the OB in Kirrel2^−/−; Kirrel3^−/− mice. In contrast to our previous findings in the accessory olfactory bulb of these mice, which showed reduced numbers of excitatory synapses, we could not detect a significant change in the number of synapses in the OB (Fig. 3J-Q). Taken together, these results demonstrate that Kirrel2 and Kirrel3 are dispensable for the targeting of specific populations of OSNs that project to the most dorsal DI region of the OB and for the formation of synapses in the OB. Furthermore, unlike mosaic overexpression of Kirrel2 or Kirrel3, which leads to MOR28 glomeruli duplication (Serizawa et al., 2006), loss of these receptors does not affect the location and number of MOR28-positive glomeruli in the ventral region of the OB.

We next considered the possibility that Kirrel2 may be involved in the coalescence of OSN axons that project to the DII region of the OB, which is located more ventrally than the DI region. Analysis of the projections of MOR174-9-positive OSNs revealed that, although the location of the main MOR174-9 glomerulus was unaltered in Kirrel2^−/− mice (Fig. 4A-J), the number of MOR174-9-positive glomeruli was increased (Fig. 4M-O). To determine whether other populations of axons that innervate the DII region were also affected by loss of Kirrel2, we assessed the targeting of M72-positive axons in Kirrel2^−/− mice. Whole-mount and immunohistochemical analyses of M72-positive projections revealed the formation of additional glomeruli in the OB of Kirrel2^−/ mice (Fig. 4K,L,M,P,Q). Furthermore, the additional MOR174-9- and M72-positive glomeruli observed in the OB of Kirrel2^−/− mice appeared to be heterogeneous and innervated by MOR174-9- and M72-negative populations of axons, indicating that their coalescence into glomeruli has been disturbed in the absence of Kirrel2 (Fig. 4N-Q; glomeruli marked by asterisks).

This study is the first loss-of-function demonstration of a requirement for a Kirrel family member in the formation of the olfactory glomerular map. Our results show that, although the differential expression of Kirrel2 and Kirrel3 provide axonal identity information in some populations of OSNs, Kirrel2/3-independent mechanisms also exist to ensure that accurate coalescence of populations of OSN axons in the DI region of the OB takes place. Taken together, these results indicate that different populations of OSN axons rely on different families of cell-surface receptors to regulate their coalescence during formation of the olfactory map.

Our results support the model that the differential expression of Kirrel2 between subsets of OSN axons is necessary for their precise coalescence into OB glomeruli. Although it remains possible, we think it is unlikely that expression of Kirrel2 in cells of the OB contributes to OSN axonal coalescence, as Kirrel2 is expressed at very low levels in the OB and its specific ablation in vomeronasal sensory neurons leads to axonal coalescence defects in the accessory olfactory bulb (Brignall et al., 2018). The previous demonstration that mosaic overexpression of Kirrel2 in subsets of MOR28-expressing axons leads to the splitting of these axons into two adjacent glomeruli, further supports an axonal role for Kirrel2 in this process (Serizawa et al., 2006). Interestingly, loss of Kirrel3 (Völker et al., 2018) or the combined loss of Kirrel2 and Kirrel3 (Fig. 3) in MOR28-positive neurons does not affect glomerular targeting of their axons. It remains possible that ablating their expression in a mosaic fashion within MOR28-positive OSNs could recapitulate the phenotypes observed in the overexpression experiments.

The overall contribution of Kirrels to axonal identity in a specific type of OSN may be determined by the levels of expression of various other cell-surface adhesion molecules in these axons, such as BIG-2 (Contactin4) and Pcdh10, which have been implicated in OSN axonal targeting (Kaneko-Goto et al., 2008; Williams et al., 2011). OSN axons expressing higher levels or a more diversified set of such proteins may be less reliant on the presence of Kirrel2 and Kirrel3 at their surface to coalesce accurately into glomeruli. Alternatively, the relative contribution of Kirrel2 to the coalescence of axons projecting to the DI and DII regions may be determined by the type of OR expressed by these axons. Class I ORs expressed in DI-projecting axons may be sufficient to impart axonal identity to these axons, whereas expression of class II ORs on axons innervating the DII region also requires Kirrel2/3 expression to impart axonal identity and promote coalescence (Bozza et al., 2009; Dewan et al., 2018; Kobayakawa et al., 2007; Miyamichi et al., 2005; Tsuboi et al., 2006; Zhang et al., 2004). However, our observation that the coalescence of OSN axons expressing the Type II MOR28 OR is unaffected by the ablation of Kirrel2 and Kirrel3 suggests that the type of OR expressed does not represent the main difference between OSN populations that do or do not require Kirrel expression for their coalescence. Future identification of cell-surface receptors that are differentially expressed in DI- and DII-innervating axons should provide further insight into the mechanistic differences underlying the coalescence of axons from these two populations of OSNs.

Kirrel3 (Prince et al., 2013), Kirrel2 (Brignall et al., 2018), S50-ires-tau-GFP (Bozza et al., 2009), M72-ires-tau-lacZ (Zheng et al., 2000), MOR28-ires-tau-lacZ (Barnea et al., 2004), MOR28-ires-tau-GFP (Shykind et al., 2004), MOR174-9-ires-tau-GFP (Cho et al., 2011) and MOR1-3-ires-tau-lacZ (Cho et al., 2011) mice have been previously described and were maintained in a mixed background. All animal procedures have been approved by the Montreal Neurological Institute Animal Care Committee, in accordance with the guidelines of the Canadian Council on Animal Care. Most analyses included male and female mice.

Two to three-month old adult mice were anesthetized and perfused transcardially with ice-cold PBS containing 4% paraformaldehyde, dissected and processed as previously described (Cho et al., 2012). For Kirrel3 staining, an additional step of antigen retrieval was performed on a hot plate at 95°C, for 2 min with ddH₂0 and then 10 min with 0.01M sodium citrate (pH 6.0). The sections were incubated overnight with primary antibody at 4°C using the following dilutions: anti-Kirrel2, 1:500 (R&D Systems, AF2930; characterized in Brignall et al., 2018), anti-Kirrel3, 1:100 (NeuroMab, clone N321C/49; characterized in Fig. S1), anti-GFP, 1:500 (Abcam, ab13970), anti-βgal, 1:500 (Abcam, ab9361), anti-OMP, 1:1000 (Wako Chemicals, 544-10001) and anti-vesicular transporter 2 (VGLUT2), 1:500 (Synaptic Systems, 135402). After rinsing in Tris-buffered saline, primary antibody was detected with Alexa-488- or Alexa-546-conjugated secondary antibodies, 1:500 (Molecular Probes, A11056, A10036, A10040).

Whole-mount OBs were stained with X-gal or processed for GFP fluorescence as previously described (Cho et al., 2011). The location of glomeruli along the three axes of the OB was determined as described in the Supplementary Materials and Methods (Fig. S2). To count the number of GFP- or β-galactosidase (β-gal)-positive glomeruli in each OB, consecutive sections (20 µm) in the coronal plane through the whole OB of adult mice were immunostained with GFP, β-gal, VGLUT2 or OMP antibodies and counter-stained with Hoechst. The number of GFP- or β-gal-positive glomeruli was counted on all sections of each OB from each mouse examined using ImageJ software. Student's t-tests (unpaired) were used to evaluate differences in glomerular location and numbers between control and mutant animals. Control animals included both wild-type and heterozygous littermates, and the number of OBs analyzed for each experiment is stated in the figure legends. Each OB was considered as an independent sample as the number of glomeruli between each OB sometimes varied in individual control and mutant mice.

OBs from perfused 3 month old mice were sectioned and stained with uranyl acetate and Reynold's lead as described in detail in Brignall et al. (2018). Images were acquired using an FEI Tecnai 12 120 kV transmission electron microscope equipped with an AMT XR80C CCD camera system. At least ten micrographs from the glomerular layer of the OB from three mice of each genotype were analyzed as described in Supplementary Materials and Methods. Student's t-tests (unpaired) were used to evaluate differences between control and mutant animals.

We thank Drs Louis Hermo and Thomas Stroh, as well as the Facility for Electron Microscopy Research of McGill University for invaluable help and advice with synapse analyses. We thank members of the Cloutier lab for helpful discussions and technical advice, and Dr Don Van Meyel for comments on the manuscript.