Zebrafish OlfC family consists of 60 intact genes and 1 pseudogene
In previous studies the size of the zebrafish OlfC gene family was given as 46 [22] or 54 intact genes [23], additionally several incomplete and pseudogenes were reported. We performed extensive Blast searches in the latest zebrafish genome assembly (GRCz10), using representative zebrafish OlfC amino acid sequences as templates. OlfC genes were identified by their position in the phylogenetic tree, using the closely related calcium sensor and t1r taste receptor genes as outgroup (Fig. 1).
We report 60 intact zebrafish OlfC genes and one pseudo gene (Fig. 1, Additional file 1). One gene was newly identified, and named for its close homology to OlfCe1 as OlfCe2. Four sequences formerly reported as pseudogenes or fragments were identified as intact and full length and were renamed accordingly. The sequences of the novel gene and the corrected predictions are given in (Additional file 2). The higher number of pseudogenes and fragments in previous studies may reflect inadequacies of the earlier versions of the genome assembly used in those studies. Unexpectedly, one gene, OlfCt1, is absent from the current assembly, although it was present in earlier versions. Since we cloned the gene from zebrafish DNA and could demonstrate specific in situ hybridization signals (Fig. 2), we assume an erroneous curation of the current assembly as most likely cause. In the phylogenetic analysis the OlfC family is paraphyletic, with OlfCa1 and OlfCb1 ancestral to the calcium sensor, which itself is ancestral to the main group of OlfC genes (Fig. 1). All subfamilies suggested by [23] were confirmed with very high branch support (Fig. 1). Based on the phylogenetic analysis we selected seven representative genes for analysis of expression patterns, including the hypothesized co-receptor OlfCc1 [18], two genes that are members of large gene expansions (g1 and q1), and four relatively isolated genes (j1, n1, u1, t1).
Main group zebrafish OlfC genes are expressed in sparse populations of sensory neurons
To analyse the spatial distribution of OlfC-expressing cells, we have performed in situ hybridization with cRNA probes on complete series of horizontal cryostat sections from adult zebrafish olfactory epithelia. Cross-reactivity of probes to other genes is not expected, since nucleotide identity of probes to the closest neighbor gene was below 80%, cf. [24, 25] in all but one case (OlfCg1 probe 80.7% identical to OlfCg7). For each gene, 3–5 complete epithelia were taken from different adult zebrafishes of similar age (8–10 months).
OlfCc1 is expressed in a large population of olfactory sensory neurons (Additional file 3), consistent with earlier observations [18]. OlfCc1 is an ortholog of murine vmn2r1, a hypothesized co-receptor, and indeed recently has been shown to co-express with other OlfC receptors in microvillous sensory neurons [18]. In contrast, all six genes from the main OlfC clade labeled sparse populations of olfactory sensory neurons within the sensory surface of the olfactory epithelium (Fig. 2). The frequency of labeled cells ranged between 60 to 180 per olfactory organ, similar to frequencies observed for expression of ORs [10] (Additional file 3). The average number of labeled cells per section ranges between 0.8 and 3.8 for different OlfC genes (Fig. 2c), again within the range of frequencies reported for other olfactory receptor genes in zebrafish [19, 21]. At first glance the spatial expression pattern of all tested OlfC genes from the main clade looked rather similar. We therefore performed quantitative analysis of spatial expression patterns in three dimensions (height within lamella, radial distance, height within organ) to examine if and how distributions for different genes differ.
Three different expression zones distinguishable in analysis of laminar height
We quantified the laminar height (height of the neuronal soma within the epithelial layer) as relative height (hrel = h soma center/thickness of sensory layer; 0, basal;1, apical, see Additional file 3 for a graphical visualization of the parameter) for all tested OlfC genes as well as omp and trpc2, which serve as cell type markers for ciliated and microvillous receptor neurons, respectively. The laminar height is characteristically different between all four populations of olfactory sensory neurons, ciliated, microvillous, crypt and kappe neurons [20]. However, it is not known, whether the distribution for individual receptors is identical to that of their corresponding cell type, as the distribution for the cell type might be composed of several different and more narrow distributions for individual receptors. To evaluate similarity and dissimilarity of distributions, we evaluated median values, half width, and maximal vertical distance in pairwise comparisons. The median value is less sensitive to outliers and skewed data than the arithmetic mean and constitutes thus a more robust measure of the center-of-gravity of the respective distribution within the sensory surface. The half width is a measure for the broadness of the respective distribution, and was estimated as difference between 1st and 3rd quartile. The maximal vertical distance between two distributions is a measure for the degree of difference between two distributions, and is measured as the maximal difference in their respective cumulative distribution functions.
The height distributions for OlfCg1, OlfCn1 and OlfCq1 show a more narrow peak in the histogram representation, corresponding to a steeper slope in the empirical cumulative distribution function, ECDF, compared to the broader peaks seen for OlfCu1, OlfCj1 and OlfCt1 (Fig. 3, Additional file 4). Kolmogorov-Smirnov test suggests the distributions to fall into three groups, which are significantly different from each other (Fig. 3, Additional file 4). The first group is formed by OlfCg1, which is different from all five other receptors. The second group contains OlfCn1 and OlfCq1, which are also different from OlfCj1 and OlfCt1 that form the third group. OlfCu1 is similar to both the second and the third group.
The overall spread of the distributions is moderate (Fig. 3). Median values occupy a narrow range between 0.603 and 0.714, i.e. the difference in median height between the most apical (OlfCg1) and the most basal (OlfCj1) receptor amounts to only 11% of the total height of the epithelial layer, and the maximal vertical difference between two distributions ranges between 5 and 31% (Additional file 4).
All six distributions are significantly different from the more basal distribution of omp-expressing cells (Fig. 7 Additional file 4). Interestingly, the most basal and most apical distributions are also different from the distribution for trpc2, the microvillous marker, whereas the intermediate distributions (OlfCn1, q1, u1) are undistinguishable from trpc2 (Additional file 4). This may seem surprising, but is in fact expected, if the trpc2 distribution results from summing over a heterogenous group of distributions for individual OlfC genes. In such a case, only distributions for individual receptors from the middle range would be expected to be undistinguishable from the trpc2 distribution.
Analysis of the radial coordinate yields two further subdivisions in expression zones
The horizontal distance of the labeled cell from the center of the lamella has been shown to be characteristically different for several or genes [10]. We quantified this coordinate as relative radius (r rel = r soma center/length of the lamella; 0, innermost; 1, outermost, cf. [10], see also Additional file 3 for a graphical representation of the coordinate. For each gene, cells were found over a wide range, with a half width (hw = r rel 3rd quartile - r rel 1st quartile) in the range of 0.225 to 0.262 (Fig. 4, Additional file 4). On first glance distributions for the seven OlfC genes analysed looked rather similar, with a maximal difference in median radius (which indicates the center of the distribution) between innermost and outermost distribution of 13.2% of the total lamellar length (Fig. 4, Additional file 4). The range for maximal vertical distance between radial distributions of sparsely expressed OlfC genes was between 6.2 and 20.2% of all cells (Additional file 4). Indeed the Kolmogorov–Smirnov test showed few significant differences for the radial distributions (Additional file 4). However, those differences found resulted in two further subdivisions of the three expression zones identified in analysis of laminar height: radial patterns for OlfCj1 and OlfCt1 are significantly different, and the radial pattern for OlfCu1 could be distinguished from OlfCq1 (Additional file 4). omp and trpc2 distributions are rather similar for the radial coordinate (Additional file 4), thus no distinction between ciliated and microvillous neurons exists for this coordinate.
Two further subdivisions of OlfC spatial distributions become apparent in the analysis of height within the olfactory organ
As third dimension we analysed the position of OlfC-expressing cells with respect to height within the organ (z coordinate, see Additional file 3 for a graphical representation). This parameter was quantified as (horizontal) section number and normalized to total section number of the whole olfactory organ. Z distributions for all but one OlfC receptor are rather similar, and are centered within the middle region of the z axis, with median values between 0.44 and 0.56, and maximal vertical distance between distributions in the range of 5.6 to 22.3% of all cells (Additional file 4). The exception is OlfCu1, which is found closer to the top, i.e. closer to the opening of the cup-shaped olfactory epithelium (Fig. 5, Additional file 4). Indeed the distribution for OlfCu1 is significantly different from that of OlfCj1, n1, q1, and thus the fuzzy border between the two major domains distinguishable in analysis of laminar height can be resolved into an additional expression domain for OlfCu1.
When combining the results for all three coordinates analysed, in total five significantly different expression zones can be distinguished for six OlfC genes analysed. Preferred positions in the radial coordinate do not covary with preferred laminar height (Figs. 1 and 6), i.e. radius and laminar height appear to be specified independently. Furthermore, no obvious correlation between preferred radius or laminar height and position of the corresponding gene in the phylogenetic tree is visible (Fig. 1). We note that we use the term ‘preferred position’ strictly to indicate the location of the corresponding distribution, not to imply potential causative mechanisms.
Double-labeling experiment confirms distinctly different, if broadly overlapping distributions
Due to the broad overlap in the observed distributions we wished to investigate, whether the differences found to be significant in single gene analyses would also hold up, when two genes were analysed by double-labeling in the same olfactory organ. For this analysis we chose OlfCg1 and OlfCu1, which differ in laminar height (g1 is more apical), height within organ (g1 is closer to the bottom), but not in the radial coordinate (Figs. 3, 4 and 5). We performed fluorescent in situ hybridization in complete series of sections for three different epithelia, and determined laminar height, radial parameter, and height within the organ as described in the preceding paragraphs. We report that for each olfactory organ the differences between OlfCg1 and OlfCu1 distributions determined in single labeling experiments are faithfully reproduced in the double-labeling experiment. For laminar height the OlfCg1 distribution was always more apical than that of OlfCu1; for radius both distributions were always extremely similar; and for height within the organ (z axis) OlfCg1 was closer to the bottom than OlfCu1 in all three cases (Fig. 6).
Thus the differences in expression domains found in the single labeling experiments for OlfCg1 and OlfCu1 are validated by the double-labeling experiment. Despite it being practically impossible to test all possible combinations of seven genes in such double-labeling experiments the close concordance of both measuring methods in the chosen example suggests the single labeling experiments in general to deliver robust results.
The distribution of OlfCc1-expressing cells is not identical to that of trpc2-expressing cells
OlfCc1, the ancestral gene of the main OlfC group, is expressed broadly in microvillous neurons, suggesting a co-receptor function for this gene [18]. We have quantified the distribution of OlfCc1-expressing cells for two coordinates, radius and laminar height. For laminar height the distribution is very similar to that of the microvillous neuron marker trpc2 (Additional file 4), and lies central within the distributions observed for the six sparsely expressed OlfC genes analysed here (Fig. 7). However, for the radial coordinate OlfCc1-expressing cells show the largest preferred radius of all genes analysed, and their distribution is significantly different from that of trpc2 (p < 0.001). It has been reported that a subset of OlfCc1-positive cells do not co-localize with trpc2 expression [18], and our data are consistent with this observation. Differences in onset of expression of these two genes during maturation of microvillous neurons could conceivably play a role, alternatively OlfCc1 might be expressed in potential microvillous neurons negative for trpc2 expression.
Distributions of OlfC-expressing cells are similarly broad as those of OR-expressing cells, but more narrowly clustered
To compare the position and shape of OlfC expression zones with those of OR-expressing cells we have re-analysed the raw data from [10]. The width of OlfC and OR expression zones, estimated as half-width, was found to be similar for individual OlfC and OR genes, both for radial distance and height within the organ (Additional file 4, and data not shown). The median value for height within the organ of OlfC and OR expression zones was rather similar for all but one gene (Fig. 7, Additional file 4, and data not shown). However, median values for the radial coordinate of ORs were more divergent, and covered more than double the range than what we observed here for the OlfC genes, even though more OlfC genes were examined (6 OlfCs vs 4 ORs). Notably, the OlfC distributions are located between inner and intermediate OR radial distributions, whereas no OlfC receptors with distributions close to the outer OR radial distributions were found. However, we cannot exclude the potential presence of such distributions for other, untested OlfC receptors.
Our results show that (partial) spatial segregation of olfactory receptor gene expression extends to zebrafish OlfC receptors and thus constitutes a general feature shared by teleost OlfC and OR receptor families.
Segregation of teleost expression zones similar to that observed in tetrapods
Recently some OlfC-related v2r genes of an amphibian were found to be expressed in the main olfactory epithelium, i.e. sharing a common sensory surface with, among others, ORs [25]. This corresponds to the situation of fish OlfC receptors, which are intermingled with ORs in the single sensory surface. We were therefore interested to evaluate the degree of similarity in the spatial representation strategies of these two species. To enable comparison with results reported here, we have re-analysed the raw data of [25]. Indeed the extent of segregation of ciliated and OlfC/V2R-expressing neurons with respect to laminar height is very similar in the clawed frog and zebrafish although interestingly the order is inverse: zebrafish OlfC-expressing cells are located apical to ciliated cells, whereas V2R-expressing cells in the clawed frog lie basal to ciliated neurons ([25], Additional file 5). Furthermore, different amphibian v2r genes exhibit broadly overlapping, but significantly different expression domains: The two sparsely expressed v2r genes in the amphibian olfactory organ show a significantly more basal distribution than the broadly expressed V2R-C (p < 0.003), the ortholog of the potential co-receptor OlfCc1 (Additional file 5). This amounts to the same type of laminar sub-segregation within the amphibian V2R domain, as shown here for the zebrafish OlfC family.
Furthermore, two initial analyses showed pronounced segregation according to laminar height in the rodent olfactory system, both for v2r gene expression in the vomeronasal organ (seven rat v2r genes, [15]) and for or gene expression in the main olfactory epithelium (four rat or genes, [26]).
A thorough investigation of OR expression patterns in mice [11] has suggested that different ORs have slightly, but distinctly different preferred positions in the unrolled olfactory sensory surface, measured as zonal index along the dorsomedial/ventrolateral axis of the olfactory epithelium. This coordinate corresponds to a radial coordinate in the coronal cross section, the usually chosen representation [11].
Taken together, although some quantitative differences exist between olfactory receptor gene families and species, the basic feature of gradually changing preferred positions for different genes from the same family appears to be conserved across a large swath of vertebrate evolution and several different olfactory receptor families. In all cases large overlap occurs between neighboring distributions.