The Slit‐binding Ig1 domain is required for multiple axon guidance activities of Drosophila Robo2

Summary Drosophila Robo2 is a member of the evolutionarily conserved Roundabout (Robo) family of axon guidance receptors. Robo receptors signal midline repulsion in response to Slit ligands, which bind to the N‐terminal Ig1 domain in most family members. In the Drosophila embryonic ventral nerve cord, Robo1 and Robo2 signal Slit‐dependent midline repulsion, while Robo2 also regulates the medial‐lateral position of longitudinal axon pathways and acts non‐autonomously to promote midline crossing of commissural axons. While Robo2 signals midline repulsion in response to Slit, it is less clear whether Robo2's other activities are also Slit‐dependent. To determine which of Robo2's axon guidance roles depend on its Slit‐binding Ig1 domain, we used a clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9‐based strategy to replace the endogenous robo2 gene with a robo2 variant lacking the Ig1 domain (robo2∆Ig1). We compare the expression and localization of Robo2∆Ig1 protein with full‐length Robo2 in embryonic neurons in vivo and examine its ability to substitute for Robo2 to mediate midline repulsion and lateral axon pathway formation. We find that the removal of the Ig1 domain from Robo2∆Ig1 disrupts both of these axon guidance activities. In addition, we find that the Ig1 domain of Robo2 is required for its proper subcellular localization in embryonic neurons, a role that is not shared by the Ig1 domain of Robo1. Finally, we report that although FasII‐positive lateral axons are misguided in embryos expressing Robo2∆Ig1, the axons that normally express Robo2 are correctly guided to the lateral zone, suggesting that Robo2 may guide lateral longitudinal axons through a cell non‐autonomous mechanism.


Summary
Drosophila Robo2 is a member of the evolutionarily conserved Roundabout (Robo) family of axon guidance receptors. Robo receptors signal midline repulsion in response to Slit ligands, which bind to the N-terminal Ig1 domain in most family members. In the Drosophila embryonic ventral nerve cord, Robo1 and Robo2 signal Slit-dependent midline repulsion, while Robo2 also regulates the medial-lateral position of longitudinal axon pathways and acts non-autonomously to promote midline crossing of commissural axons. While Robo2 signals midline repulsion in response to Slit, it is less clear whether Robo2's other activities are also Slit-dependent. To determine which of Robo2's axon guidance roles depend on its Slit-binding Ig1 domain, we used a clustered regularly interspaced short palindromic repeats (CRISPR)/ Cas9-based strategy to replace the endogenous robo2 gene with a robo2 variant lacking the Ig1 domain (robo2ΔIg1). We compare the expression and localization of Robo2ΔIg1 protein with full-length Robo2 in embryonic neurons in vivo and examine its ability to substitute for Robo2 to mediate midline repulsion and lateral axon pathway formation. We find that the removal of the Ig1 domain from Robo2ΔIg1 disrupts both of these axon guidance activities. In addition, we find that the Ig1 domain of Robo2 is required for its proper subcellular localization in embryonic neurons, a role that is not shared by the Ig1 domain of Robo1. Finally, we report that although FasIIpositive lateral axons are misguided in embryos expressing Robo2ΔIg1, the axons that normally express Robo2 are correctly guided to the lateral zone, suggesting that Robo2 may guide lateral longitudinal axons through a cell non-autonomous mechanism.
We have previously carried out a comprehensive structure/ function study of the ectodomain elements within the Drosophila Robo1 receptor, and we found that while the midline repulsive activity of Drosophila Robo1 is strictly dependent on its Slit-binding Ig1 domain, each of its other seven ectodomain elements (Ig2-5, Fn1-3) is individually dispensable for midline repulsion (Brown et al., 2015;Brown, Reichert, & Evans, 2018;Reichert, Brown, & Evans, 2016).
Although not required for midline repulsive signaling, Robo1's Fn domains are necessary for its negative regulation by Commissureless (Comm) and Robo2 (Brown et al., 2018;Brown & Evans, 2020  show that this activity of Robo2 is Slit-dependent, as robo1,robo2 double mutants display more severe midline crossing defects than robo1 or robo2 single mutants, and the robo1,robo2 double mutants phenocopy Slit null mutants Simpson, Kidd, et al., 2000). However, Robo1 and Robo2 signaling mechanisms are not entirely the same, as robo1 can rescue robo2's midline repulsive role, but robo2 cannot substitute for robo1 in this context (Spitzweck et al., 2010).
In addition to its canonical role in midline repulsion, Robo2 also acts non-autonomously to inhibit Slit-Robo1 repulsion in trans to promote midline crossing of commissural axons in the embryonic VNC. We have previously shown that deletion of the Slit-binding Ig1 domain decreases, but does not eliminate Robo2's ability to promote midline crossing in gain-of-function experiments .
The mechanism by which Robo2 promotes lateral pathway formation has not been characterized, although it has been proposed that the three Drosophila Robo receptors act (either alone or in combination) to specify the medial-lateral distance of longitudinal axon pathways from the midline in response to a midline-secreted Slit gradient . We have previously shown that Robo2's ability to induce lateral shifting of medial longitudinal neurons in gain of function experiments is disrupted when Ig1 + Ig2 of Robo2 is deleted, consistent with the hypothesis that this activity is Slit-dependent (Evans & Bashaw, 2010). If this is the case, we should expect that Slit binding via the Ig1 domain will be required for Robo2's endogenous lateral positioning activity.
To determine the requirements for individual ectodomain elements for the various axon guidance roles of Drosophila Robo2 and to distinguish between its Slit-dependent and Slit-independent activities (if any), we have begun a systematic structure/function analysis of ectodomain elements within Robo2. Here, we describe our initial set of experiments using a clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9-based gene replacement approach to examine the requirement for the Robo2 Ig1 domain for the receptor's endogenous roles in midline repulsion and lateral pathway formation. We show that each of these activities is disrupted by deletion of the Robo2 Ig1 domain, and we also show that, in contrast to Robo1, the Ig1 domain of Robo2 is also important for proper axonal localization of the Robo2 protein in embryonic neurons in vivo.

| CRISPR/Cas9-based gene replacement of robo2
To begin our functional analysis of the Robo2 ectodomain, we used a CRISPR/Cas9-based gene modification approach (Gratz et al., 2014;Port, Chen, Lee, & Bullock, 2014) to modify the robo2 locus to express structural variants of Robo2 ( Figure 1). We first used this approach to create a full-length robo2 robo2 allele, in which exons 2-14 of the robo2 locus are replaced by an hemagglutinin (HA)-tagged full-length robo2 cDNA. In our robo2 modified alleles, the endogenous robo2 promoter, transcriptional start site, first exon (including the start codon and signal sequence), and first intron remain unmodified. Spitzweck F I G U R E 1 CRISPR/Cas9-based gene replacement of robo2. (a) Schematics of the full-length Robo2 protein and the Robo2ΔIg1 variant, from which the Slit-binding Ig1 domain has been deleted. (b) Schematic of the Drosophila robo2 gene showing intron/exon structure, location of gRNA target sites, robo2 robo2 homologous donor plasmid, and the resulting robo2 robo2 HDR allele. Endogenous robo2 coding exons are shown as purple boxes; 5 0 and 3 0 untranslated regions are shown as light gray boxes. The start of transcription is indicated by the bent arrow. Introns and exons are shown to scale, with the exception of the first intron, from which approximately 19 kb has been omitted. Red arrows indicate the location of upstream (gRNA 1) and downstream (gRNA 2) gRNA target sites. Gray brackets demarcate the region to be replaced by sequences from the donor plasmid. Arrows indicate the position and orientation of PCR primers. The same two gRNAs were combined with a robo2 robo2ΔIg1 donor plasmid to create the robo2 robo2ΔIg1 HDR allele. (c) Partial DNA sequences of the unmodified robo2 gene and the modified robo2 robo2 and robo2 robo2ΔIg1 HDR alleles. Black letters indicated endogenous DNA sequence; red letters indicate exogenous sequence. Both DNA strands are illustrated. The gRNA protospacer and PAM sequences are indicated for both gRNAs. The first five base pairs of robo2 exon 2 are unaltered in both modified alleles, and the robo2 coding sequence beginning with codon N90 is replaced by the HA-tagged full-length robo2 (for robo2 robo2 ) or robo2ΔIg1 (for robo2 robo2ΔIg1 ) cDNAs. The endogenous robo2 transcription start site, ATG start codon, and signal peptide are retained unmodified in exon 1. The PAM sequences for both gRNA targets and the protospacer sequence for the gRNA2 target are modified in the donor plasmids, ensuring that the robo2 robo2 and robo2 robo2ΔIg1 donor plasmids and modified alleles are not cleaved by Cas9. The gray box shows robo2 sequence polymorphisms present in the nos-Cas9 injection stock compared with the reference genome sequence, which are predicted to interfere with Cas9 cleavage at the gRNA2 target site. 3 0 H, 3 0 homology region; 5 0 H, 5 0 homology region; gRNA, guide RNA; HA, hemagglutinin epitope tag; HDR, homology directed repair; PAM, protospacer adjacent motif; UTR, untranslated regions et al. (2010) used a knock-in approach to similarly replace robo2 with a full-length robo2 cDNA, and showed that HA-tagged Robo2 protein expressed from this modified locus was properly expressed and could fully rescue Robo2's roles in midline repulsion, lateral pathway formation, and promotion of midline crossing (Spitzweck et al., 2010).
We generated a guide RNA (gRNA) expression plasmid using the pCFD4 gRNA backbone (Port et al., 2014), containing two gRNA sequences targeting the first intron (~50 bp upstream of exon 2) and exon 14 (3 0 UTR) of robo2. We also created a robo2 robo2 homologous donor plasmid containing the HA-tagged robo2 coding sequence along with 1-kb upstream (5 0 H) and downstream (3 0 H) flanking sequences to serve as a template for homology-directed repair (HDR). The robo2 coding sequence in this donor construct is flanked by restriction sites, allowing us to swap out the full-length robo2 sequence for any alternative coding sequence. Using this approach, we should be able to generate many different robo2 gene replacement variants using the same set of gRNAs and the same homologous donor backbone. For the CRISPR-modified alleles described here (robo2 robo2 and robo2 robo2ΔIg1 ), each donor construct was co-injected along with the pCFD4 gRNA construct into Drosophila embryos expressing Cas9 under the control of the germline-specific nanos promoter (nos-Cas9. P) (Port et al., 2014), and F1 progeny from the injected flies were screened by PCR to identify those carrying the expected modification.
We generated stable lines from positive F1 flies and sequenced the modified locus fully from at least two lines for each modified allele.
Additional details are provided in the Methods.

| robo2 gRNA target polymorphisms and variable HDR replacements
In addition to the correctly modified robo2 alleles we recovered for the robo2 robo2 and robo2 robo2ΔIg1 HDR gene replacements, we also recovered lines that tested positive in our initial PCR screening but deviated from the expected HDR replacements in several ways. These

| Expression and localization of Robo2ΔIg1 in embryonic neurons
We have previously shown that deleting the Ig1 domain from Drosophila Robo1 does not affect its expression pattern, axonal localization, clearance from commissural axon segments, or regulation by the endosomal sorting protein Commissureless (Comm) (Brown et al., 2015). To examine whether the Ig1 domain of Robo2 is similarly dispensable for its expression and localization in embryonic  . We are unable to directly compare our HA-tagged robo2 robo2 CRISPR allele expression with endogenous Robo2 expression, as there is no monoclonal anti-Robo2 antibody (unlike for Drosophila Robo1 and Robo3), and the original polyclonal anti-Robo2 antibodies  are no longer available. However, the expression pattern we observe closely matches previous descriptions of Robo2's endogenous protein expression throughout embryogenesis, both in the VNC and other embryonic tissues Spitzweck et al., 2010). This result is also consistent with Spitzweck et al.'s description of a similar knock-in robo2 robo2 allele and confirms that removing most of the introns from robo2 and adding an N-terminal 4 Â HA tag does not interfere with the normal transcription or translation of robo2, or the stability, trafficking, or localization of the Robo2 protein (Spitzweck et al., 2010).
In homozygous robo2 robo2ΔIg1 embryos, HA-tagged Robo2ΔIg1 protein was present on longitudinal axons and restricted to the lateral-most region of the neuropile, similar to full-length Robo2 (Figure 2b, arrowhead). We also observed an increased degree of HA staining in neuronal cell bodies in the cortex surrounding the neuropile compared with robo2 robo2 embryos (Figure 2b, asterisk), suggesting that some portion of the Robo2ΔIg1 protein may not be trafficked correctly in embryonic neurons and instead retained at elevated levels in neuronal cell bodies. To quantify this, we compared the ratio of anti-HA pixel intensity in cortical cell bodies versus lateral axons in robo2 robo2 and robo2 robo2ΔIg1 embryos. We found that the ratio of cell body:axon HA levels was 2.8-fold higher in the VNC cortex of robo2 robo2ΔIg1 embryos compared with robo2 robo2 (Figure 2c).
We also noted that the overall architecture of the axon scaffold appears affected in robo2 robo2ΔIg1 homozygous embryos, with an overall decrease in the width of the scaffold along with irregularly shaped segmental neuromeres similar to robo2 loss-of-function mutants, suggesting that replacing Robo2 with Robo2ΔIg1 may interfere with one or more aspects of neural development in robo2 robo2ΔIg1 embryos.
To quantify this, we measured the width of the axon scaffold in robo2 robo2 and robo2 robo2ΔIg1 embryos and found that the average width of the axon scaffold in individual segments in robo2 robo2ΔIg1 embryos was 80.6% of the average width recorded in robo2 robo2 embryos (a decrease of 19.4%; Figure 2d).
In addition to these nerve cord defects, we observed embryonic head development defects in robo2 robo2ΔIg1 embryos (not shown), similar to those present in robo2/leak mutant embryos (Schimmelpfeng, Gögel, & Klämbt, 2001). We also note that the robo2 robo2ΔIg1 allele is homozygous lethal, similar to previously characterized null alleles of robo2 (Simpson, Kidd, et al., 2000), while stocks carrying the robo2 robo2 allele produce homozygous viable adult flies. While it is not known exactly which function(s) of robo2 are essential for viability, these observations suggest that multiple functions of robo2 may be disrupted by deletion of its Ig1 domain.
F I G U R E 2 Expression of HA-tagged robo2 alleles in the embryonic CNS. (a,b) Stage 16 Drosophila embryos stained with anti-HRP (magenta; labels all axons) and anti-HA (green) antibodies. (a 0 ,b 0 ) Anti-HA channels alone from the same embryos in (a,b). Lower panels show optical cross sections of the regions outlined in dashed boxes. (a,a 0 ) In robo2 robo2 homozygous embryos, HA-tagged Robo2 protein reproduces Robo2's endogenous expression pattern. At stage 16, Robo2 protein is primarily localized to longitudinal axons and restricted to the lateral-most region of the ventral nerve cord neuropile (arrowhead). (b,b 0 ) In robo2 robo2ΔIg1 homozygous embryos, HA-tagged Robo2ΔIg1 protein is detectable on lateral longitudinal axons (arrowhead) and also present at elevated levels on or in neuronal cell bodies within the cortex (asterisk). (c) Bar graph quantifies relative HA pixel intensities in the VNC cortex (area marked by asterisk in panel b). The ratio of HA pixel intensity between lateral axons and cell bodies in the cortex was measured for the genotypes shown in (a,b), and the ratios were normalized to robo2 robo2 (*p < .000001 by Student's twotailed t test). Raw quantification data are provided in Table S1. (d) Bar graph quantifies relative width of the axon scaffold in robo2 robo2 and robo2 robo2ΔIg1 embryos. The width of the axon scaffold was measured for the genotypes shown in (a,b) and normalized to robo2 robo2 (*p < .000001 by Student's two-tailed t test). Raw quantification data are provided in Table S2 In this initial report, we will focus on the midline repulsion and lateral pathway formation activities of robo2.

| Robo2ΔIg1 cannot substitute for Robo2 to promote midline repulsion or lateral pathway formation
To examine specific axon guidance outcomes in robo2 robo2 and robo2 robo2ΔIg1 embryos, we used an anti-FasII antibody to label a subset of longitudinal axon pathways in the VNC. Robo2 is required for guidance of FasII-positive axons in the contexts of midline repulsion and longitudinal pathway formation: in robo2 mutants, medial FasII-positive axons ectopically cross the midline (reflecting a lack of midline repulsion), and FasII-positive lateral axon pathways fail to form correctly Simpson, Kidd, et al., 2000). We quantified ectopic midline crossing and lateral pathway defects in robo2 robo2 and robo2 robo2ΔIg1  (Figure 3c). This result is consistent with a previous study by Spitzweck et al. (2010), which reported that a robo2 robo2 allele created via an ends-in knock-in approach could also fully rescue robo2-dependent axon guidance outcomes (Spitzweck et al., 2010).
In contrast, we observed ectopic midline crossing (24.2% of segments) and lateral pathway defects (37.6% of hemisegments) in robo2 robo2ΔIg1 homozygous embryos at frequencies that were statistically indistinguishable from those in robo2 amorphic mutants (p = .42 and p = .62 by t test, respectively), suggesting that Robo2ΔIg1 is not able to substitute for full-length Robo2 in the contexts of midline repulsion or lateral pathway formation. The observation that lateral

| DISCUSSION
Here, we have described a CRISPR/Cas9-based gene replacement approach to characterize the functional importance of structural elements in the Drosophila Robo2 axon guidance receptor and used this approach to show that the Slit-binding Ig1 domain of Robo2 is required for two distinct axon guidance roles of Robo2 during development of the Drosophila embryonic VNC (midline repulsion and longitudinal pathway formation). We have also shown that the Ig1 domain contributes to the proper localization of Robo2 in embryonic neurons, suggesting a possible role for Robo2 Ig1 in protein trafficking to and/or retention in neuronal axons that is not conserved in Drosophila Robo1. The tools and approach we describe here will facilitate additional structure/function and gene replacement studies of Drosophila robo2.

| CRISPR gene replacement versus rescue transgene studies of Drosophila Robo receptors
We have previously used a transgene-based approach to characterize the functional importance of individual ectodomain elements in the Drosophila Robo1 protein (Brown et al., 2015;Brown et al., 2018;Brown & Evans, 2020;Reichert et al., 2016). This approach relies on a rescue transgene carrying a small region of genomic DNA (~4.5 kb) containing regulatory sequences sufficient to recapitulate the full expression pattern of robo1. Equivalent regulatory sequences have not been identified for robo2 or robo3, so we could not use a similar rescue transgene approach for structure/function studies of Robo2.
We have previously used a bacterial artificial chromosome (BAC) rescue approach employing a large (83.9 kb) robo2-containing BAC to examine the role of Robo2's Ig2 domain in promoting midline crossing .
The CRISPR/Cas9-based strategy described here has a number of advantages over the above approaches, including: (a) identifying/ isolating regulatory sequences is not required, as endogenous regulatory sequences are used instead; (b) the genetics of introducing markers and/or other mutations into the modified background is simplified, as there is no need to track an inactivating mutation plus a separate rescue transgene; and (c) the laborious recombineering and difficult transgenesis with very large BAC DNA fragments can be avoided. This CRISPR/Cas9 gene replacement approach could also be used to replace robo2 with other coding sequences, including its paralogs from Drosophila (robo1 and robo3), orthologs from other species, or chimeric/variant receptors, which would facilitate further structure/function or comparative/evo-devo studies. For example, we have used an equivalent approach to replace Drosophila Robo3 with its Tribolium ortholog Robo2/3 to compare their axon guidance activities (Evans, 2017).

| Differential requirement for Ig1 in axonal localization of Robo1 and Robo2
We have previously reported that the Ig1 domain of Drosophila Robo1 is not required for proper expression or axonal localization of the Robo1 protein in the embryonic VNC (Brown et al., 2015). The results presented here indicate that this is not true for Drosophila Robo2; instead, deleting Ig1 from Robo2 does appear to alter its subcellular localization in embryonic neurons. We see a similar effect on protein localization when the Ig1 domain is deleted from Drosophila Robo3 (Abigail Carranza and T.A.E., unpublished), suggesting that the Ig1 domains in Robo2 and Robo3 play a role in protein localization that is F I G U R E 3 The Robo2 Ig1 domain is required for midline repulsion and lateral pathway formation. (a-d) Stage 16 Drosophila embryos stained with anti-HRP (magenta) and anti-FasII (green) antibodies. Lower images show anti-FasII channel alone from the same embryos. (a) In robo2/+ heterozygous embryos, FasII-positive axons form three distinct longitudinal pathways on either side of the midline, one each in the medial, intermediate, and lateral zones of the neuropile. FasII-positive axons do not cross the midline in these embryos. Arrowhead points to the lateral FasII pathway. (b) In robo2 loss of function mutants (robo2 123 /robo2 135 ), FasII-positive axons cross the midline inappropriately in around 20% of segments (arrow with asterisk), and the lateral FasII pathway fails to form correctly in around 44% of hemisegments (arrowhead with asterisk). (c) In homozygous robo2 robo2 embryos, midline repulsion and lateral pathway formation occur normally. (d) Homozygous robo2 robo2ΔIg1 embryos display ectopic midline crossing (arrow with asterisk) and lateral pathway defects (arrowhead with asterisk) equivalent to robo2 mutants. (e,f) Quantification of ectopic midline crossing defects (e) and lateral pathway defects (f) in the genotypes shown in (a-d). Number of embryos scored for each genotype (n) is shown. Error bars indicate s.e.m. Percent defects for the two modified alleles were compared to robo2 mutants by two-tailed Student's t test with a Bonferroni correction for multiple comparisons (*p < .01; **p < .001; n.s., not significant). Raw quantification data are provided in Table S3 not shared by the Ig1 domain in Robo1. Importantly, deleting the Ig1 domain from Robo2 does not appear to affect its translation or protein stability, as Robo2 and Robo2ΔIg1 proteins are expressed at equivalent levels and detectable at the cell surface in cultured Drosophila S2R+ cells , and both proteins are detectable both on neuronal axons and in neuronal cell bodies in vivo, with the main difference being the relative levels in/on axons versus cell bodies (Figure 2).
We have reported that deleting the Ig3 or Fn1 domains from Robo1 also resulted in elevated cell body expression of Robo1 (Brown et al., 2018;Reichert et al., 2016), but this effect (increased punctate staining in neuronal cell bodies for both Robo1ΔIg3 and Robo1ΔFn1) appears qualitatively distinct from what we observe with Robo2ΔIg1, where the increased cell body signal appears more membrane-localized (see Figure 2b, where circular staining patterns presumably reflecting outlines of individual cell bodies can be seen). Whether this reflects differential roles for Robo2 Ig1 versus Robo1 Ig3/Fn1 in protein localization or instead reflects underlying differences in the normal expression of Robo1 (which normally does not appear to reach the membrane in neuronal cell bodies) versus Robo2 (which is normally detectable at low levels on cell body membranes; see Figure 2a), is unclear.

| Slit-dependent versus Ig1-dependent roles of Robo2
Although it is clear that Robo2's normal roles in midline repulsion and lateral pathway formation are deficient in robo2 robo2ΔIg1 embryos, we cannot distinguish between a direct requirement for Ig1 in each of these roles versus a secondary effect of altering Robo2's subcellular distribution when Ig1 is deleted. In other words, perhaps Robo2ΔIg1 would be able to rescue some or all of these roles if it were primarily localized to axons similar to full-length Robo2.
We also note that while deleting the Ig1 domain from Robo2 does strongly or completely abrogate Slit binding , we cannot formally rule out the possibility that Ig1 may have other, Slit-independent activities that would also be disrupted by deleting the entire Ig1 domain.
Slit binding in Drosophila and human Robo receptors can be strongly disrupted in vitro through targeted point mutations in Ig1 (Fukuhara et al., 2008;Morlot et al., 2007); a similar strategy might allow targeted disruption of Slit binding without altering other putative functions of Robo2 Ig1, which may, in turn, allow a more precise dissection of Slit-dependent versus Slit-independent roles of Ig1 in vivo. The CRISPR/Cas9-based gene replacement approach described here could be used to engineer a robo2 locus expressing a cDNA carrying one or more targeted point mutations in Ig1 to test this possibility and may also help disentangle the functional importance of Slit-binding versus axonal-localization activities of Robo2 Ig1.

| Does Robo2 guide FasII-positive longitudinal axons non-autonomously?
When robo2's roles in embryonic axon guidance were first described two decades ago, the initial models posited that it acted as a cell-autonomous Slit receptor to signal midline repulsion and to guide longitudinal axons to lateral pathways Simpson, Kidd, et al., 2000). Subsequent studies of Robo2's unique pro-crossing function revealed that Robo2 acts non-autonomously to guide commissural axons across the midline . We note that although FasII-positive lateral axons are misguided in robo2 robo2ΔIg1 embryos, HA-positive axons in these embryos (that is, the axons that normally express Robo2) are still tightly restricted to the lateral region of the neuropile and are not apparently misguided into the intermediate or medial zones. We have observed that HA-positive axons similarly remain restricted to the lateral zone in robo2 robo1 embryos, in which the robo2 coding sequence has been replaced by robo1, even though these embryos also display lateral FasII pathway defects similar to robo2 null mutants (Spitzweck et al., 2010) (T.A.E., unpublished). These observations suggest that the FasIIpositive axons that are misguided in robo2 robo1 and robo2 robo2ΔIg1 embryos (and robo2 null mutants) may not be the same as the lateral axons that normally express Robo2. In other words, Robo2 may also act non-autonomously to regulate lateral position of FasII-positive axon pathways. Distinguishing between these possibilities will require examining both FasII and HA expression in the same embryos (which technical limitations have thus far prevented us from doing), and/or generating additional markers to label Robo2-expressing lateral axons independently of Robo2 expression to examine their lateral positions in wild-type, robo2 mutant, and robo2 gene replacement embryos.
These results also demonstrate that the Ig1 domain of Robo2 is not required for guidance of Robo2-expressing longitudinal axons to lateral pathways, which indicates that the lateral axons that normally express Robo2 may not require its activity to form and/or join lateral pathways, or that Robo2 directs the medial-lateral positioning of these axons through an Ig1-independent mechanism.

| Construction of robo2 donor plasmids
The robo2 robo2 donor construct was assembled from four PCR fragments via Gibson assembly (New England Biolabs #E2611). The four fragments were derived from pBluescript (plasmid backbone), the wild-type robo2 genomic locus (5 0 and 3 0 homology regions), and an HA-tagged robo2 cDNA plasmid (4 Â HA epitope tag and robo2 coding region). The robo2 coding sequence in the robo2 robo2 donor construct is flanked by NheI and NotI restriction sites. To make the robo2 robo2ΔIg1 donor, the full-length robo2 coding sequence was excised with NheI-NotI and replaced with the robo2ΔIg1 coding sequence. Donor plasmids contain engineered mutations in PAM and/or protospacer sequences to prevent cleavage by Cas9. Modified robo2 HDR alleles include the following amino acid residues after the N-terminal 4 Â HA epitope tag, relative to Genbank reference sequence AAF51375: robo2 robo2 (E89-V1463), robo2 robo2ΔIg1 (E89-N90/L187-V1463). The entire donor regions including coding sequences and robo2 flanking regions were sequenced prior to injection. 4.1.2 | Construction of robo2 gRNA plasmid robo2 gRNA sequences were cloned into the tandem expression vector pCFD4 (Port et al., 2014) via PCR followed by Gibson assembly using the PCR product and BbsI-digested pCFD4 backbone. For gRNA 2, an additional G nucleotide was added to the 5 0 end of the gRNA target sequence to facilitate transcription from the U6-1 promoter.

| Generation and recovery of CRISPRmodified alleles
The robo2 gRNA plasmid was co-injected with the robo2 robo2 or robo2 robo2ΔIg1 homologous donor plasmids into nos-Cas9.P embryos (Bloomington Drosophila Stock Center stock #54591) (Port et al., 2014) by BestGene Inc. (Chino Hills, CA). Injected individuals (G0) were crossed as adults to Sco/CyOwg. Founders (G0 flies producing F1 progeny carrying modified robo2 alleles) were identified by testing two pools of three F1 females per G0 cross by genomic PCR with primers 458 and 325 (for robo2 robo2 ) or 458 and 327 (for robo2 robo2ΔIg1 ), which produce 1.5-kb products only when the respective HDR alleles are present. From each identified founder, 5-10 F1 males were then crossed individually to Sco/ CyOwg virgin females. After 3 days, the F1 males were removed from the crosses and tested by PCR with the same set of primers to determine if they carried the modified allele. F2 flies from positive F1 crosses were used to generate balanced stocks, and the modified alleles were fully sequenced by amplifying the entire modified locus (approximately 6 kb) from genomic DNA using primers 458 and 545 or 458 and 594, then sequencing the PCR product after cloning via CloneJET PCR cloning kit (Thermo Scientific).

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.