Molecular Vision 2018; 24:443-458
Received 29 November 2017 | Accepted 17 July 2018 | Published 19 July 2018
1University of Idaho, Department of Biological Sciences, Moscow, ID; 2University of Washington School of Medicine, WWAMI Medical Education Program, Moscow, ID
Correspondence to: Deborah L. Stenkamp, Department of Biological Sciences University of Idaho, Moscow, Idaho 83844; Phone: (208) 885-8963; FAX: (208) 885-7905; email: firstname.lastname@example.org
Purpose: The differential adhesion hypothesis states that a cell adhesion code provides cues that direct the specificity of nervous system development. The Down syndrome cell adhesion molecule (DSCAM) and sidekick (SDK) proteins belong to the immunoglobulin superfamily of cell adhesion molecules (CAMs) and provide both attractive and repulsive cues that help to organize the nervous system during development, according to the differential adhesion hypothesis. The zebrafish genome is enriched in dscam and sdk genes, making the zebrafish an excellent model system to further test this hypothesis. The goal of this study is to describe the phylogenetic relationships of the paralogous CAM genes and their spatial expression and co-expression patterns in the embryonic zebrafish retina.
Methods: Exon–intron structures, karyotypic locations, genomic context, and amino acid sequences of the zebrafish CAM genes (dscama, dscamb, dscaml1, sdk1a, sdk1b, sdk2a, and sdk2b) were obtained from the Ensembl genome database. The Prosite and SMART programs were used to determine the number and identity of protein domains for each CAM gene. The randomized axelerated maximum likelihood (RaxML) program was used to perform a phylogenetic analysis of the zebrafish CAM genes and orthologs in other vertebrates. A synteny analysis of regions surrounding zebrafish CAM paralogs was performed. Digoxigenin (dig)-labeled cRNA probes for each CAM gene were generated to perform in situ hybridization of retinal cryosections from zebrafish embryos and larvae. Dual in situ hybridization of retinal cryosections from zebrafish larvae was performed with dig- and fluorescein-labeled cRNA probes.
Results: We found the studied zebrafish CAM genes encode similar protein domain structures as their corresponding orthologs in mammals and possess similar intron–exon organizations. CAM paralogs were located on different chromosomes. Phylogenetic and synteny analyses provided support for zebrafish dscam and sdk2 paralogs having originated during the teleost genome duplication. We found that dscama and dscamb are co-expressed in the ganglion cell layer (GCL) and the basal portion of the inner nuclear layer (INL), with weak expression in the photoreceptor-containing outer nuclear layer (ONL). Of the dscam genes, only dscamb was strongly expressed in ONL. Sdk1a and sdk1b were co-expressed in the GCL and the basal portion of the INL. Sdk2a and sdk2b also showed co-expression in the GCL and basal portion of the INL. All Sdk genes were expressed in the ciliary marginal zone (CMZ). Dual in situ hybridizations revealed alternating patterns of co-expression and exclusive expression for the dscam and sdk1 paralogs in cells of the GCL and the INL. The same alternating pattern was observed between dscam and sdk2 paralogs and between sdk1 and sdk2 paralogs. The expression of dscaml1 was observed in the INL and the GCL, with some cells in the basal portion of the INL showing co-expression of dscaml1 and dscama.
Conclusions: These findings suggest that zebrafish dscam and sdk2 paralogs were likely the result of the teleost whole genome duplication and that all CAM duplicates show some differential expression patterns. We also demonstrate that the comparative expression patterns of CAM genes in the zebrafish are distinct from the exclusive expression patterns observed in chick retina, in which retinal ganglion cells express one of the four chick Dscam or Sdk genes only. The patterns in zebrafish are more similar to those of mice, in which co-expression of Dscam and Sdk genes is observed. These findings provide the groundwork for future functional analysis of the roles of the CAM paralogs in zebrafish.
The differential adhesion hypothesis proposes that the organization of an organism’s body plan is mediated by the expression of cell adhesion molecules that results in the aggregation of cells into different tissues and organs . Experimental evidence for the differential adhesion hypothesis in the nervous system, where it is invoked to explain complex connectivity patterns, is robust. For example, neuroligins and neurexins, interacting families of cell adhesion molecules, have been experimentally demonstrated to provide specificity during neural development by partnering pre- and post-synaptic cells . Other cell adhesion molecules, including cadherins, contactins, Dscams, semaphorins, and plexins, have also been implicated in patterning the nervous system by promoting adhesion or avoidance [3-8].
Many studies on the role of differential adhesion in neural development have been conducted vis-à-vis the retina. Unlike other neural tissues, retinal neurons locate and form synapses with the appropriate partners in the absence of synaptic input , which then play a more local role in sculpting precise synaptic structures in the retina [10,11]. The retina is organized into three layers—the outer nuclear layer (ONL), which contains rod and cone photoreceptors; the inner nuclear layer (INL), which contains bipolar, amacrine, and horizontal cells; and the retinal ganglion cell layer (GCL), which contains ganglion and displaced amacrine cells . Retinal synapses are localized to two synaptic layers located in between the cellular layers, the outer plexiform layer (OPL) and the inner plexiform layer (IPL). The abundance of tools to label and manipulate retinal neurons and the importance of the tissue in our species dominant sense, vision has made the retina a valuable model to study the mechanisms of neural connectivity.
The zebrafish retina is an exceptionally good model to study the mechanisms by which differential adhesion sculpts neural architecture. This is in part because zebrafish, as members of the teleost lineage of fishes, underwent a whole genome duplication (WGD) event approximately 400 million years ago (mya) . As a result, many genes represented by a single copy in tetrapods have two homologs in zebrafish. Following duplication, a gene may be neofunctionalized (a new function emerges), subfunctionalized (each of the gene products adopts some of the functions of the original gene, possibly tissue- or developmental stage-dependent), or, most commonly, pseudogenized (loss of expression and/or function of one gene copy) . Potential subfunctionalized and neofunctionalized gene paralogs in the zebrafish genome present the opportunity to dissect the function(s) of genes that play multiple critical roles in neural development. For example, a loss of function of only one paralog in zebrafish may generate a phenotype that is simpler to interpret than a corresponding mouse or human phenotype . For cell adhesion molecules, the study of subfunctionalized paralogs would further our understanding of the differential adhesion hypothesis.
As a first step, we have characterized the expression of Down Syndrome cell adhesion molecule (dscam), and sidekick (sdk) paralogs in the zebrafish retina. DSCAM and SDK proteins in other model organisms have been implicated in multiple developmental roles, including axon guidance, avoidance, and synaptic targeting [16-20]. The zebrafish genome has three dscam genes (dscama, dscamb, and dscaml1) and four sdk genes (sdk1a, sdk1b, sdk2a, and sdk2b), while the chick and mouse genomes each contain only two Dscam genes (Dscam and Dscaml) and two Sdk genes (Sdk1 and Sdk2). In this study, we examine the comparative predicted protein structure and phylogenetic relationships of the zebrafish dscam and sdk genes and determine their respective expression patterns in the developing zebrafish retina.
Wild-type strains of zebrafish (Danio rerio) were maintained on a 14 h:10 h light:dark cycle in monitored, recirculating system water according to standard protocols, as previously described . All procedures using animals were approved by the University of Idaho’s Institutional Animal Care and Use Committee. Embryos were collected and maintained at 28.5 °C until collecting at 48, 72, and 96 h post fertilization (hpf). The time of spawn was considered as 0 hpf. Embryos/larvae were fixed in a solution of 4% paraformaldehyde in 5% buffered (pH 7.4) sucrose. Following fixation, embryos/larvae were transferred in sequential steps of increasing the percentage of sucrose solution in phosphate buffer (PB) to 20% sucrose for overnight cryoprotection at 4 °C. Tissues were embedded in a mixture consisting of a 1:2 ratio of 20% sucrose in PB and optimal cutting temperature (OCT) embedding medium (Sakura Finetek USA, Torrance, CA). Tissues were frozen and sectioned into 5 μm samples using a Microm HM 550 cryostat or a Leica CM3050 cryostat.
Amino acid sequences of proteins encoded by the zebrafish (dscama, dscamb, dscaml1 sdk1a, sdk1b, sdk2a, and sdk2b genes) were obtained from the Ensembl genome database (ensembl.org release 90). The longest protein coding transcript was used in all analyses. The amino acid sequence of the genes was analyzed using the SMART  and Prosite  programs to determine the number and identity of protein domains (Figure 1A). The intron–exon structure and the position of genes in chromosomes were obtained from Ensembl (release 90; Figure 1B,C).
Amino acid sequences for zebrafish Dscama, Dscamb, Dscaml1, Sdk1a, Sdk1b, Sdk2a, Sdk2b and their orthologs in medaka (Oryzias latipes), tilapia (Oreochromis niloticus), threespine stickleback (Gasterosteus aculeatus), cod (Gadus morhua), humans (Homo sapiens), mice (Mus musculus), chickens (Gallus gallus), and spotted gar (Lepisosteus oculatos) were obtained from Ensembl (release 90). Alignments for the amino acid sequences were generated with ClustalW  using a BLOSUM cost matrix. Gaps were removed from the alignments. Gene trees were generated with randomized axelerated maximum likelihood (RaxML) 8.2.11  using an algorithm for rapid bootstrapping and search for the best score maximum likelihood tree under the model gamma blosum62.
Images of regions of the zebrafish genome with annotated genes near the locations of the dscam and sdk genes were obtained from Ensembl (release 92). The identities of the annotated genes and their paralogs were confirmed through a BLAST alignment.
Zebrafish larvae (96 hpf) were homogenized in 1.0 mL of TRIzol reagent (Invitrogen, Waltham, MA, USA), and RNA was extracted using a PureLink RNA Micro Kit (Invitrogen). The extracted RNA was used to generate cDNA with qScript cDNA SuperMix (Quanta Bioscience, Saskatoon, SK, Canada). The resulting cDNA library was subjected to PCR using gene-specific primers. The resulting PCR products were run and isolated on 0.75% agarose gel and purified using a GeneJet Gel Extraction Kit (Fermentas Life Technologies, Waltham, MA). The purified DNA was TA cloned into a T-easy vector (Promega, Madison, WI) and used to generate cRNA probes through in vitro transcription using DIG RNA labeling mix (Roche, Indianapolis, IN) and T7 or SP6 RNA polymerase (Roche). The primers and promoter sequences corresponding with each gene target are summarized in Table 1.
In situ hybridization was performed as previously described in [26-28]. Sections were vacuum dried for 24 h before storage at -20 °C. They were then hydrated using sequential ethanol treatments at concentrations of 100%, 95%, 70%, and 50%. Proteinase K was used to permeabilize the tissue, and this was followed by acetylation with triethanolamine (TEA) and acetic anhydride. Sections were then dehydrated with sequential ethanol treatments at concentrations of 50%, 75%, 95%, and 100% before air drying for 1 h. Tissues were hybridized with probes for dscama, dscamb, and dscaml1 at 68 °C. Probes for sdk1a, sdk1b, sdk2a, and sdk2b were hybridized at 70 °C. These hybridization temperatures were selected according to Polypro software . Following stringency washes of 1:1 formamide and 2xSSC performed at 65 °C, tissue was treated with RNase A and then incubated overnight at room temperature in anti-digoxigenin-AP, Fab fragments or anti-fluorescein-AP, Fab fragments (Roche). NBT/BCIP solution and/or Fast Red were used to generate colored precipitates. Dual in situ hybridization was performed as previously described . Upon completion of color development, slides were washed in alkaline phosphatase (AP) buffer and mounted with 80% glycerol. Images were collected using a Leica DMR microscope and a SPOT camera or a Leica DM2500 upright microscope with a Leica DFC700T camera using bright-field or differential interference contrast (DIC) optics. Each probe was examined using at least four slides containing tissue from at least three different embryos/larvae for each sampling time. Sense probes were prepared and did not generate a detectable signal.
DSCAM and SDK proteins belong to the immunoglobulin superfamily (IgSF) of CAMs. The structure of these proteins includes a variable number of immunoglobulin-like domains, a variable number of fibronectin domains (type III), a single-pass transmembrane domain, and an intracellular domain with a C-terminus that binds to the PDZ domain of an interacting protein [31–34] (Figure 1A). Genome analysis reveals that zebrafish dscama, dscamb, dscaml1, sdk1a, sdk1b, sdk2a, and sdk2b each encode similar protein domain patterns, which corresponds to what has been shown previously in their respective orthologs in other vertebrates  (Figure 1A). The presence of introns in all CAM genes analyzed indicates that paralogous genes were not duplicated by retrotransposition events, as these events would result in intronless genes . Dscaml1 and dscamb were found to be located on chromosome 15, dscama was found to be located on chromosome 10, sdk1a and sdk2a were found to be located on chromosome 3, sdk1b was found to be located on chromosome 1, and sdk2b was found to be located on chromosome 12 (Figure 1B). Data obtained from Ensembl (GRCz10) indicated that dscama is predicted to have 32–33 exons, dscamb is predicted to have 30–33 exons, and dscaml1 is predicted to have four alternative splice products with a variable number of exons (Figure 1C). Sdk1a is predicted to have a single splice product of 44 exons, and sdk1b is predicted to have four alternative splice products with a variable number of exons (Figure 1D). Sdk2a and sdk2b are predicted to possess 43 and 36 or 46 exons, respectively (Figure 1E). Unlike Drosophila Dscam1, which has an estimated 38,000 splice variants, tetrapod dscam and sdk genes undergo only limited alternative splicing [16,37-39]. The limited number of alternative splice forms in zebrafish is consistent with the splicing patterns of tetrapod dscam and sdk genes (Figure 1C–E).
Individual phylogenetic trees were generated using amino acid sequences for Dscaml1 and each pair of Dscam, Sdk1 and Sdk2 paralogs using RaxML. The phylogenetic analysis for each set of genes included amino acid sequences corresponding to orthologs from the teleost fish—medaka (Oryzias latipes), tilapia (Oreochromis niloticus), threespine stickleback (Gasterosteus aculeatus), and cod (Gadus morhua). In addition, we included orthologs from humans (Homo sapiens), mice (Mus musculus), chickens (Gallus gallus), and spotted gar (Lepisosteus oculatos) that diverged from the teleost lineage before the teleost genome duplication (TGD) .
The resulting tree for Dscam showed Dscama and Dscamb orthologs from teleost fish grouped into two different branches. Spotted gar Dscam was placed as diverging before the duplication that gave rise to the two branches of Dscam paralogs in teleost fish (Figure 2A), consistent with these paralogs emerging with the TGD.
The DSCAML1 tree grouped teleost Dscaml1 duplicates into two branches (albeit with low confidence); however, zebrafish Dscaml1 was placed diverging before the branches of Dscaml1 paralogs in teleosts, and the spotted gar Dscaml1 was placed diverging before the zebrafish Dscaml1 and teleost Dscaml1 paralogs diverged (Figure 2B). The branch point for zebrafish Dscaml1 versus Dscaml1 of the other teleosts examined likely diverged before speciation events, leading to the cyprinid lineage (such as zebrafish). Cyprinids, or specifically zebrafish, may not have retained any duplicated Dscaml1 gene(s) arising from the TGD.
In the SDK1 tree, zebrafish Sdk1a was grouped with spotted gar Sdk1, while zebrafish Sdk1b was grouped with the Sdk1 genes of other teleost fish (Figure 2C). The SDK1 tree suggests that the sdk1a paralog that was retained in zebrafish could have emerged before the TGD, followed by the loss of both resulting genes’ paralogs. Alternatively, this could be a result of the differential divergence of the paralogous sdk1 genes in the teleost lineages, which is a more parsimonious explanation. Interestingly, the non-cyprinid teleosts that were sampled did not show evidence of duplicated sdk1 genes in their genomes.
In the SDK2 gene tree, Sdk2a and Sdk2b were grouped into two different branches, and spotted gar Sdk2 was placed diverging previous to the duplication of Sdk2 in teleosts (Figure 2C). These results are consistent with the Sdk2 paralogs emerging with the TGD.
The locations of paralogous genes on different chromosomes (Figure 1B) suggest that none of the paralogs were generated by local, tandem duplications. In addition, the presence of introns in all CAM genes analyzed (Figure 1C-E) indicates that paralogous genes were not duplicated by retrotransposition events, as these events would result in intronless genes . Synteny analysis of dscama/dscamb and of sdk2a/sdk2b revealed the presence of multiple genes possessing corresponding paralogs around the duplicated portions of genome where the CAM paralogs are located, supporting duplications related to the TGD  (Figure 3). However, synteny analysis of sdk1a/sdk1b revealed nearby genes with no corresponding paralogs around the duplicated sdk1, leaving open the possibility of these paralogs having originated with events not related to the TGD (Figure 3).
We used in situ hybridization to examine expression patterns for the dscam paralogs in the developing zebrafish retina. Transcript-specific probes were applied to cryosectioned retinal tissue obtained from zebrafish embryos/larvae sacrificed at 48, 72 and 96 h post fertilization (hpf). These times correspond to the emergence of the OPL (48 hpf), the onset of cone-mediated visual function (72 hpf), and continued retinal growth beyond the embryonic period (96 hpf) . At 48, 72, and 96 hpf, dscama mRNA was expressed by a subset of cells located in the ganglion cell layer (GCL) and the basal portion of the INL, which corresponds to the location of amacrine cells, with weak expression in some cells of the ONL, where photoreceptors are located (Figure 4A–C). Likewise, dscamb mRNA was also detected in subsets of cells located in the GCL and basal portions of the INL. However, the dscamb signal was observed to be strongly expressed in the ONL (Figure 4D–F); this expression pattern was consistent for 48, 72 and 96 hpf samples. We previously found dscamb to be expressed in the ONL of adult zebrafish by in situ hybridization and to be enriched in the rod photoreceptors of adult zebrafish by RNA-Seq and qPCR of purified rods .
Expression of dscaml1 mRNA was observed in patches of cells scattered throughout the INL and the GCL in 48, 72, and 96 hpf zebrafish retina (Figure 4G-I). The expression domains of dscamb mRNA and dscaml1 mRNA appear to increase in size over developmental time, while the expression domain of dscama in the INL and the GCL appears to decrease in thickness over developmental time (Figure 4).
To ensure that the expression patterns observed were not products of artifacts produced by the microscopy method used, cryosections were also imaged under bright-field and DIC optics for direct comparison (Figure 5A,B,E,F,I,J). The bright-field images show patterns similar to those viewed under DIC optics. In addition, in situ hybridization controls using sense probes were performed for each dscam gene and imaged under bright-field and DIC optics (Figure 5C,D,G,H,K,L). Sense probes did not produce a detectable reaction product.
We used in situ hybridization to determine the mRNA expression patterns for the gene paralog pairs sdk1a and sdk1b and sdk2a and sdk2b. Antisense probes detected the expression of transcripts for all four genes in the basal region of the INL and in the GCL at 48, 72, and 96 hpf (Figure 6A–L). This predicts their expression by amacrine cells (AC) and retinal ganglion cells (RGCs). Sections from 96 hpf retina showed the expression of sdk2a and sdk2b in subsets of cells in the ONL (Figure 6I, L) and possibly weak expression of sdk1a and sdk1b in some cells of the ONL (Figure 6C–F). The patterns for sdk2 transcripts are similar to the expression pattern seen for dscamb (Figure 6F). All four zebrafish sdk paralogs were detected in the ciliary marginal zone (CMZ) of far peripheral retina (Figure 6C, F, I, L). The CMZ contains stem and progenitor cells that generate new retinal neurons and glia as the retina grows . The extent of the expression domains of sdk1a and sdk1b appear to decrease over developmental time, while those of sdk2a and sdk2b appear to increase (Figure 6).
To ensure that the expression patterns observed were not a product of artifacts produced by the microscopy method used, cryosections were also imaged under bright-field and DIC for direct comparison (Figure 7A,B,E,F,I,JM,N). In addition, in situ hybridization controls using sense probes were performed for each sdk gene (Figure 7C,D,G,H,K,L,O,P).
To determine whether our target CAM paralogs are co-expressed in the same cell populations, we used dual in situ hybridization in retinal sections obtained from 96 hpf zebrafish larvae. In these sections, co-expression of dscama and dscamb was observed within a subset of cells located in the GCL and INL (Figure 8A–C). However, only dscamb was strongly expressed in the ONL (Figure 8B). We also used dual in situ hybridization for the sdk1 gene paralogs to determine if they are co-expressed during development. We found sdk1a and sdk1b were co-expressed by a subset of cells in the GCL and the basal portion of the INL (Figure 8D–F). Similarly, sdk2a and sdk2b were also found to be co-expressed by a subset of cells in the GCL, the basal portion of the INL, and the ONL (Figure 8 G–I).
Dscam and Sdk genes are expressed in a mutually exclusive pattern in the chick retinal GCL  but are co-expressed in at least some common sets of neurons in mice [34,45]. We next sought to determine whether the expression of different CAM types was mutually exclusive in zebrafish retina. We performed dual in situ hybridization in 96 hpf retinal sections with the following CAM gene pairs: dscama with dscaml1, sdk1b with dscamb, sdk2b with dscamb, and sdk2b with sdk1b. Dscama (Figure 9A) and dscaml1 (Figure 9B) showed co-expression in a small number of cells located in the basal part of the INL (Figure 9C), while other areas of the INL only showed the expression of dscaml1 or dscama. Sdk1b (Figure 9D) and dscamb (Figure 9E) were co-expressed in some cells located in the INL and the GCL (Figure 9F). Similarly, sdk2b (Figure 9G) and dscamb (Figure 9H) were observed to have a similar pattern of sporadic co-expression within neurons of the INL and the GCL (Figure 9I); sdk2b (Figure 9J) and sdk1b (Figure 9 K), were observed to have occasional co-expression within the INL and the GCL (Figure 9L).
We report the predicted protein structures, gene structures, phylogenetic relationships, and developmental retinal expression patterns of the zebrafish genes encoding the Dscam and Sdk CAMs. The zebrafish has a richer repertoire of these CAM genes in their genomes, with duplicates (paralogs) of dscam, sdk1, and sdk2 but not of dscaml1, while other vertebrate model organisms lack these duplicates. The paralogs likely did not arise through tandem duplication or retrotransposition because the duplicated genes are found in different chromosomes and because all the pairs of paralogs show the presence of introns (Figure 1B–E). Phylogenetic trees provide good support for the timing of duplication of the dscam genes (dscama and dscamb) and of the sdk2 genes (sdk2a and sdk2b) as part of the TGD.
All the zebrafish CAM genes examined in this study are expressed within the developing retina, as summarized diagrammatically in Figure 10. The dscams are all expressed in the GCL and the INL, with occasional co-expression. Dscamb is also strongly expressed in the ONL, and dscaml1 is also found in the outer INL, consistent with some degree of subfunctionalization of the paralogs (Figure 10A,B). Sdk1a and sdk1b show very similar expression domains in the INL, the GCL, and the CMZ. Sdk2a and sdk2b also show similar expression domains in the INL, the GCL, and the CMZ of all developmental stages examined, along with expression in the ONL of 96 hpf larvae (Figure 10C,D,E,F). However, in each case (sdk1a versus sdk1b and sdk2a versus sdk2b), these expression domains were not completely overlapping, again consistent with some degree of subfunctionalization (Figure 10C,D,E,F,G,H). We also note that the in situ probes used did not discriminate among predicted alternative splice products for dscamb, dscaml1 (Figure 1C), sdk1b (Figure 1D), or sdk2a (Figure 1E); therefore, we may have undersampled the diversity of expression patterns of these CAMs. In addition, some of the CAM genes appeared to expand their expression domains, while others appeared to restrict these domains over developmental time. This finding is consistent with developmentally specific roles for the paralogs, for example, in cell–cell spacing within laminae versus later events, such as synaptogenesis.
Our results indicate that cellular expression patterns of Dscam and Sdk genes in the retina are not strictly conserved among vertebrates, with different vertebrate model species expressing different Dscam and Sdk genes in identified retinal cell types. Dscama is expressed in the GCL and the INL of the zebrafish retina (Figure 10A,B), which is consistent with the expression pattern of Dscam in the developing mouse retina  but differs from the expression pattern of Dscam in the chick retina, where subsets of cells in the ONL, all regions of the INL, and the GCL express Dscam . Likewise, Dscamb is expressed in rods in the zebrafish ONL (43; Figure 10A,B), which differs from the mouse retina in which Dscaml1, but not Dscam, is expressed in rods . Interestingly, Dscam is observed in the ONL of the developing chick retina, but in this case expression is enriched in green-sensitive cones [35,46].
The expression of zebrafish sdk1 paralogs in the GCL and the INL (Figure 10C,D,E,F) is consistent with the expression of Sdk1 in mouse retina, where there is expression in subsets of cells in the GCL and throughout the INL but very little expression in the ONL . The expression of sdk2a and sdk2b in 96 hpf zebrafish retina (Figure 10C,D,E,F) appears similar to the pattern of Sdk expression in the developing chick retina, in subsets of cells in the ONL, throughout the INL, and in the GCL [35,39].
Our results map the expression patterns of dscam and sdk genes in the zebrafish retina. According to the differential cell adhesion hypothesis, these combinatorial expression “codes” for retinal cell types may be important for the development of their spatial patterns and/or of their synaptic connections. Although the apparently distinct codes for zebrafish as compared to mice or chickens generate the same general retinal structure, the differences among species may underlie patterning features important for environmentally adaptive retinal functions, such as those related to color vision or high-acuity specializations. Why multiple proteins that mediate homotypic binding are co-expressed in individual cells is an open question. In mice co-expression of Dscaml1 and Sdk genes in vglut3+ amacrine cells serve different roles, with Dscaml1 preventing excessive adhesion and Sdk genes mediating synaptic lamination and pairing [34,45]. Compartment-specific roles for different CAMs may also explain the co-expression of multiple homotypic CAMs; for example, Dscam prevents adhesion in mouse RGC dendrites but promotes axon growth in RGC axons [18,20]. Differential binding of ligands, such as slit to Dscam but not Sdk proteins, may also explain the co-expression of multiple seemingly similar proteins . In the future, we will perform functional studies to determine how the expression and co-expression of these genes influence the organization and circuitry of retinal neurons.
This research was supported by NSF Cooperative Agreement DBI #093945 (BEACON; PGF), and NIH R21EY026814 (DLS). J. Sukeena was supported partially by a teaching assistantship from the University of Idaho. We thank Ruth Frey for assisting with in situ hybridization procedures training and Celeste Brown for assistance with phylogenetic analysis. Dr. Deborah L. Stenkamp (email@example.com) and Dr. Peter G Fuerst (firstname.lastname@example.org) are co-corresponding authors for this study.