Overlapping but distinct topology for zebrafish V2R-like olfactory receptors reminiscent of odorant receptor spatial expression zones

Abstract

Background

The sense of smell is unrivaled in terms of molecular complexity of its input channels. Even zebrafish, a model vertebrate system in many research fields including olfaction, possesses several hundred different olfactory receptor genes, organized in four different gene families. For one of these families, the initially discovered odorant receptors proper, segregation of expression into distinct spatial subdomains within a common sensory surface has been observed both in teleost fish and in mammals. However, for the remaining three families, little to nothing was known about their spatial coding logic. Here we wished to investigate, whether the principle of spatial segregation observed for odorant receptors extends to another olfactory receptor family, the V2R-related OlfC genes. Furthermore we thought to examine, how expression of OlfC genes is integrated into expression zones of odorant receptor genes, which in fish share a single sensory surface with OlfC genes.

Results

To select representative genes, we performed a comprehensive phylogenetic study of the zebrafish OlfC family, which identified a novel OlfC gene, reduced the number of pseudogenes to 1, and brought the total family size to 60 intact OlfC receptors. We analyzed the spatial pattern of OlfC-expressing cells for seven representative receptors in three dimensions (height within the epithelial layer, horizontal distance from the center of the olfactory organ, and height within the olfactory organ). We report non-random distributions of labeled neurons for all OlfC genes analysed. Distributions for sparsely expressed OlfC genes are significantly different from each other in nearly all cases, broad overlap notwithstanding. For two of the three coordinates analyzed, OlfC expression zones are intercalated with those of odorant receptor zones, whereas in the third dimension some segregation is observed.

Conclusion

Our results show that V2R-related OlfC genes follow the same spatial logic of expression as odorant receptors and their expression zones intermingle with those of odorant receptor genes. Thus, distinctly different expression zones for individual receptor genes constitute a general feature shared by teleost and tetrapod V2R/OlfC and odorant receptor families alike.

Background

The sense of smell is unrivaled in terms of molecular complexity of its input channels. Several hundred to over two thousand different receptor genes convey olfactory signals in mammals [1,2,3]. Nearly all of these olfactory receptors belong to one of four different gene families, the initially discovered odorant receptors proper (ORs, [4]), two types of vomeronasal receptors (V1Rs and V2R, respectively), and the trace amine-associated receptors (TAARs), all of which have counterparts in the teleost olfactory system (ORs, ORAs, OlfCs, and TAARs, respectively [5]. Generally, out of this large repertoire only a single olfactory receptor gene is expressed in any particular olfactory sensory neuron [6, 7]. Olfaction is different from most other senses in that neurons with the same sensivity, i.e. expressing the same sensory receptor, are scattered within the sensory surface [8,9,10]. Nevertheless, qualitative as well as quantitative analysis of expression patterns of rodent and zebrafish odorant receptors has shown that this scattering is not completely random, but that different ORs segregate into distinct spatial subdomains within a common sensory surface [8, 10, 11].

Some borders between subdomains appear to be rather sharp, e.g. between zone I and II in the mammalian olfactory epithelium [8, 9, 12], but in many cases expression zones of different genes overlap widely [10, 11]. It has been suggested that sharply delineated subdivisions serve to segregate receptor groups with different biological functions or with different target regions in the olfactory bulb [13]. However it is difficult to make the same argument for receptor genes with extensively overlapping expression zones. Interestingly chromosomal location maps to some extent to expression zone [14] suggesting potential ways how a spatial segregation of expression might be generated by the olfactory system.

So far it is not known, whether other olfactory receptor families besides ORs show similar segregation into expression zones, although an initial report in rodents points to this possibility for V2Rs [15]. Moreover, in the fish olfactory system all four olfactory receptor gene families are expressed in a single sensory surface, but it is not known, how the other receptor gene families integrate into the spatial expression patterns found for odorant receptors [10]. The cell types expressing ORs and V2R-related OlfC receptors (ciliated and microvillous neurons, respectively) are intermingled in the zebrafish olfactory epithelium [16], although initial qualitative assessment suggested microvillous neurons to lie more apical (closer to the lumen) than ciliated neurons [17, 18]. We have recently developed a thorough analysis method to quantify and compare three-dimensional spatial distribution patterns observed for different receptor neuron populations and receptor genes [19,20,21]. Here we use this method to analyse expression patterns for seven OlfC genes, chosen as representative based on a rigorous phylogenetic analysis of the zebrafish OlfC gene family. We have performed in situ hybridization for all seven genes, and have quantified the positions of cells expressing these genes in three dimensions. We observe non-random, distinctly different expression zones for different zebrafish V2R-related OlfC receptors that intercalate into those described for odorant receptors. This spatial logic thus constitutes a general feature shared by teleost and tetrapod V2R/OlfC and OR receptor families.

Results

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).

Fig. 1

Selection of seven representative genes from a phylogenetic tree of the zebrafish OlfC family. A phylogenetic tree of 60 full length OlfC gene sequences was constructed using a maximum likelihood (ML) algorithm (see Methods for details). Bootstrap support in percent is indicated at relevant nodes. As in previous analyses, the OlfC family appears polyphyletic, with the calcium sensor gene CaSR intercalating between OlfCa1, OlfCb1, and the remainder of the OlfC family. Mouse and zebrafish T1R taste receptors were used as outgroup. Single asterisk, gene was predicted in [23] as pseudogene (OlfCb1, e1, q10) or fragment (OlfCf1, m2), but is intact and full length in the current prediction; double asterisk, gene is lost in GRCz10; triple asterisk, novel gene. OlfC genes highlighted with yellow were selected for expression analysis. To the right of the tree the core cell distribution parameters for the genes analysed are shown as bar graphs. Light blue, median radial position; dark blue, median height position, all values normalized to the respective maximal values. No correlation is apparent between position in the phylogenetic tree and median radius or height

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).

Fig. 2

v2r-related OlfC genes generally are expressed in small subsets of scattered olfactory sensory neurons. Horizontal sections of adult zebrafish olfactory epithelium were hybridized with probes for OlfCg1, OlfCn1, OlfCq1, OlfCj1, OlfCu1, and OlfCt1. Column a shows representative complete sections labeled with the respective probes. The scale bars correspond to 40 μm. Column b, higher magnifications from different sections. The hybridization signal was observed in sparse cells within the sensory region of the olfactory epithelium, as expected; arrowheads point to some labeled neurons. The scale bars correspond to 20 μm. c Bar graphs representing number of labeled cells per section for each OlfC gene (mean +/− SEM, n = 78–265 sections/gene)

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 (h_rel = 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.

Fig. 3

Quantitative assessment of laminar height distributions of v2r-related OlfC-expressing neurons. Laminar height of OlfC-expressing cells was quantified for seven OlfC genes, including OlfCc1. Complete series of sections from three to five olfactory epithelia were evaluated for each OlfC gene. Height within the lamina was normalized to maximal laminar thickness. a The resulting distributions of relative laminar height (from 0, most basal to 1, most apical, i.e. bordering to the lumen) are shown binned (histogram, top row) and unbinned (empirical cumulative distribution function, ECDF, bottom row). The color code for the OlfC genes is the same as for later figures to facilitate comparisons between different positional parameters. Light grey curves in the ECDF plot for each of the tested OlfC genes represent the distribution for individual olfactory organs. Due to the smaller number of cells, scatter is increased. b Overlay of the seven distributions shown individually in panel a), both as histogram (left panel) and ECDF (right panel)

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.

Fig. 4

Quantitative assessment of the radial distribution of v2r-related OlfC-expressing neurons. The distribution of radial positions of OlfC-expressing cells was quantified for seven OlfC genes, including OlfCc1, using the same set of sections, for which laminar height was determined, except the very first sections, where the sensory surface does not yet extend toward the median raphe. Radial position within the section was normalized to maximal radius, i.e. length of the lamella containing the respective labeled cell. In other words, radial distance was measured from the apex of the lamellar ‘curve’, i.e. closest to the median raphe, to the cell soma center, and normalized to the distance between this apex position (most central) and the border of the epithelial section (most peripheral). a The resulting distributions of relative radius (from 0, innermost to 1, outermost) are shown binned (histogram, top row) and unbinned (ECDF, bottom row). The color code for the OlfC genes is the same as in Fig. 3 to facilitate comparisons between different positional parameters. Light grey curves in the ECDF plot for each of the tested OlfC genes represent the distribution for individual olfactory organs. Due to the smaller number of cells, scatter is increased. b Overlay of the seven distributions shown individually in panel a), both as histogram (left panel) and ECDF (right panel)

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.

Fig. 5

Quantitative assessment of distribution of v2r-related OlfC expressing neurons along the vertical z-axis (height within the organ). Height within the olfactory organ was quantified as section number in a series of horizontal sections, and normalized to the total number of sections containing sensory epithelium, using the same set of cells, for which laminar height was determined. Relative height within the organ ranges from 0 (top section, near to the opening of the bowl-shaped olfactory organ) to 1 (bottommost section). a The resulting distributions are shown binned (histogram, top row) and unbinned (ECDF, bottom row). The color code for the OlfC genes is the same as in Figs. 3 and 4 to facilitate comparisons between different positional parameters. Light grey curves in the ECDF plot for each of the tested OlfC genes represent the distribution for individual olfactory organs. Due to the smaller number of cells, scatter is increased. b Overlay of the distributions shown individually in panel a), both as histogram (left panel) and ECDF (right panel)

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.

Fig. 6

Simultaneous labeling of two OlfC genes confirms distinctly different distributions. a Representative micrograph of two-color in situ hybridization, depicting sparse expression of OlfCg1 (in magenta) and OlfCu1 (in green), within a single horizontal section of the olfactory epithelium. b-d Quantitative assessment of the distributions of the labeled cells for laminar height (b), radial distance (c) and along the vertical z-axis (d). The resulting distributions of relative laminar height (from 0, most basal to 1, most apical, i.e. bordering to the lumen), relative radius (from 0, innermost to 1, outermost) and relative height within the organ (from 0, top section to 1, bottommost section) are shown unbinned as empirical cumulative distribution function (ECDF). Color of ECDF graphs corresponds to the color employed in panel a. KS-test with a p-value cutoff of < 0.01 was used to evaluate the significance of differences between the distributions, if any. Significance is indicated by asterisks, **, p ≤ 0.01; ****, p ≤ 0.0001

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.

Fig. 7

Comparison of spatial distribution parameters between OlfC genes. Schematic representation of spatial distributions for different OlfC genes and two marker genes, omp and trpc2, by ellipses ranging from the 1st to the 3rd quartile value for x and y parameter. Color code for OlfC genes as before. a Radius (x axis) is depicted vs. laminar height (y axis). Note the distributions for all OlfC genes and trpc2, the marker for microvillous neurons, center on rather apical positions within the lamina (large values for laminar height), clearly segregated from the much more basal positions for omp-positive, ciliated neurons. No correlation is apparent between radius and laminar height values. b Radius (x axis) is depicted vs. height within the organ (y axis). No correlation is apparent between radius and height within organ values

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.

Discussion

Spatial segregation of olfactory receptor gene expression within a family appears to be a conserved feature in the tetrapod lineage [8, 9, 11, 15, 25, 27]. In teleost fish a single study has shown a corresponding organisation for ORs [10]. Fish possess only one olfactory organ, in which all their olfactory receptor families are expressed. We were interested to find out, how the expression of another major olfactory receptor family, the V2R-related OlfCs is integrated into the spatial pattern of OR-expressing neurons. We have examined six different sparsely expressed zebrafish OlfC genes, chosen as representative by their position in the phylogenetic tree, as well as the broadly expressed [17] gene OlfCc1, which lies ancestral to the main group in the phylogenetic tree. Earlier expression studies had not analysed spatial positions of OlfC-expressing cells beyond noting a sparse expression pattern [17]. We have performed a thorough quantitative analysis in three dimensions, radial distance (central to peripheral), laminar height (basal to apical) and along the vertical z-axis of the olfactory organ (top to bottom), to establish the spatial pattern of OlfC gene expression.

We show here that several distinctly different, if broadly overlapping expression zones can be distinguished for zebrafish OlfC receptors. Preferred position appears to be independently specified for each coordinate analysed, similar to results obtained for the amphibian main olfactory epithelium [25]. Furthermore we have re-analysed the raw data from two earlier publications covering frog V2R and zebrafish OR expression. These re-analyses show highly similar segregation into several distinct, and always broadly overlapping expression zones.

The spread of distributions for zebrafish OlfC genes appears to be somewhat more narrow than those observed for both OlfC-related v2r genes [15] and or genes [10, 11, 26]. Although we cannot exclude this to be a consequence of our selection of OlfC genes analysed, we consider it unlikely due to our sampling a large swath of the phylogenetic tree of OlfC receptors. For the or genes one could argue that their larger evolutionary age [5] would have allowed a correspondingly larger diversification in position-specifying mechanisms. However the apparently larger spread of tetrapod V2R expression zones would need to be explained differently. Dedicated experiments to quantify the tetrapod v2r distributions will be required to unambiguously solve this issue.

Both in fish (our results, [10]), and in mammals [11] large overlaps are observed between all neighboring distributions, suggesting the (partial) segregation of olfactory receptor gene expression not to be relevant for segregation of function. Alternatively the differences in center of gravity between distributions might constitute a ‘trace record’ of ontogenetic processes. For some zebrafish odorant receptor genes different onsets of expression have been reported, from 24 h to 3 dpf, see e.g. [28]. However, considering the complex morphological remodelling of the sensory surface in the following weeks (formation of median raphe and of lamella) and the huge growth in absolute size of the olfactory organ from larval stage to adulthood, it appears very unlikely that any potential initial asymmetries would have propagated to the adult stage.

One could speculate that age differences in the individual epithelia examined would represent ‘snapshots’ of different migration or maturation stages of sensory neurons or their precursors resulting in differently positioned expression domains. However, this is highly unlikely for several reasons: Firstly, we examined a narrow age range, between 8 and 10 months of age. Secondly, we also find the distribution differences in double-labeling experiments, i.e. within single olfactory organs. Lastly, any potential mechanism has to take into account the constant renewal of olfactory sensory neurons during adult life. The average lifetime of zebrafish OSN has been estimated as around one month [29], so at the age examined, several rounds of neurogenesis have already occurred blurring any potential relation to organismal age.

It is noteworthy that the zebrafish olfactory organ possesses two proliferative zones, one at the apex of the lamella, the other at the outer border of the sensory region [30] . A recent publication [29] examining three zebrafish odorant receptor genes has suggested that different olfactory progenitor cells might migrate differentially from these proliferative zones, which could result in a pattern of broad expression zones with different centers-of-gravity. While beyond the scope of the current investigation, it will be interesting to see whether OlfC genes follow a similar pattern. As an aside, at least for the mammalian system tangential migration appears to be less relevant, cf. [31]. Alternatively, the molecular mechanisms underlying olfactory receptor gene choice might rely on gradients of signalling molecules, which are read out differently in different precursor cells.

Conclusions

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, some quantitative differences between olfactory receptor gene families and species notwithstanding. A hallmark of this organisational principle is the large overlap even of distinctly different distributions, suggesting that these expression domains may constitute a byproduct of ontogenetic processes rather than being relevant for segregation of function.

Methods

Sequence data mining and phylogenetic analysis

TblastN searches were performed on the GRCz10 assembly of the zebrafish genome, using representative zebrafish OlfC and mouse V2R sequences as queries. Candidate sequences were aligned by MAFFT algorithm (E-INS-i setting, [32] and sequence positions with over 90% gaps were removed using Gapstreeze (http://hcv.lanl.gov/content/sequence/GAPSTREEZE/gap.html). A phylogenetic tree was constructed using a modified Maximum Likelihood method (PhyML-aLRT) with SPR setting for tree optimization and chi square-based aLRT for branch support [33]. Trees were drawn using Treedyn [34]. Zebrafish and mouse T1Rs served as outgroup. Candidate sequences had to fulfill three criteria to be accepted as bona fide unique OlfCs: the gene had to be located inside the corresponding phylogenetic tree with branch support over 80%; the sequence had to map to a unique, non-overlapping genomic position; the amino acid sequence had to show at least 2% difference to other OlfC receptors, and the variable amino acids had to be distributed along the sequence. Sequences that are > 98% identical in amino acid sequence are considered allelic variants [14].

Sequences were named according to the OlfC nomenclature introduced by [23] since this study previously had named most OlfC genes; one novel sequence (OlfCe2) was named according to phylogenetic relationship. OlfCb1p, OlfCe1p, and OlfCq10p were assumed to be pseudo genes [23], but have intact full length ORFs in GRCz10 as predicted by Genewise [35], and have been renamed as OlfCb1, OlfCe1, and OlfCq10, respectively. Two fragment predictions (OlfCf1 and OlfCm2) were extended, OlfCm2 to full length using Genewise [35].

Animal handling and probe generation

Zebrafish used in this study are of Ab/Tü genetic background and were raised in the local fish facility. Adult wild type zebrafish (8–10 months old) were anesthetized with MS-222 (ethyl 3-aminobenzoate, Sigma) and decapitated. Olfactory epithelia were dissected out, embedded in TissueTek O.C.T. compound (Tissue-Tek; Sakura Finetek USA), and frozen at − 20 °C. Ten micrometer-thick horizontal cryosections were thaw mounted onto Superfrost Plus slide glasses (Fisher Scientific, Pittsburgh, PA).

Digoxigenin (DIG)-labeled probes for seven V2R-related OlfC genes were generated as described [36]. Templates for probes were amplified either from genomic DNA or in some cases from cloned cDNA fragments, with T3 promoter site (TAT TAACCCTCACTAAAGGGAA) attached to the 5′ end of the primers. In situ probes were generated for the following genes: OlfCn1, OlfCq1, OlfCg1, OlfCu1, OlfCj1, OlfCt1, OlfCc1, omp and trpc2, using the following primer pairs: OlfCn1 Fwd: 5´ GACTTGGATTGGAGCTTTGC 3′; Rev.: 5´ TTGCAGATGGCTCACAGTTC 3′; OlfCq1 Fwd: 5´ GAGATCCAGGGACTTCGTGA 3′; Rev.: 5´ CCAGGGCATAAACTGCCTTA; OlfCg1 Fwd: 5´ AGTCAAGCACTTTGGCTGGT 3′; Rev.: 5’CCTCCCAGCACATGAAAACT 3′; OlfCu1 Fwd: 5´ GCTCCTGGTTGAAGTTGCTC 3′; OlfCu1 Rev.: 5´ ACA GGC TCTCCATTGGTG TC 3′; OlfCj1 Fwd: 5′ TGAGGGTTG GATCACGTACA 3′; Rev.: 5´ ATGCGTCATACAAGCCAATG 3′; OlfCt1 Fwd: 5´ GCAGCA ATTCTCTCCACTCC 3′; Rev.: 5´ TCTTGTTTTGCCACTGAGCTG 3′; OlfCc1 Fwd: 5´ GGGCCTTTTGAGAACGACACATG 3′; Rev.: 5´ CAGATTTGCCCATTAGCGAAGAGAG 3′. The primer sets used for trpc2 and omp are described in [37].

In situ hybridization

Ten micrometer-thick horizontal cryosections were thaw-mounted onto Superfrost Plus slide glasses (Thermo). Pretreatment of sections, probe hybridization, and stringent washing were performed as described [10], except omitting Proteinase K digestion. After stringent washing at 65 °C, sections were blocked in 1% blocking reagent (Roche) in PBS for 1 h. The slides were then incubated at 37 °C for 2 h with sheep anti-DIG Fab fragments conjugated with alkaline phosphatase (Roche), dilution 1:500 in blocking solution. After washing 3 times in PBS, hybridized probes were visualized by enzymatic reaction with NBT-BCIP (Roche). After evaluating the success of the staining, slides were washed 2 times in PBS for 5 mins each, mounted with VectaMount (Vector Laboratories, Burlingame, CA, USA) and photographed with a wide field microscope (Keyence BZ-9000). In case of the two-color double in situ hybridization, probes for OlfCg1 and OlfCu1 were labeled with dig and flu antigen respectively and similar in situ conditions were applied as described above except the detection steps. OlfCg1 probe was detected with anti-dig conjugated with alkaline phosphate, followed by HNPP detection method (Sigma Aldrich) whereas the OlfCu1 probe was detected by peroxidase labeled anti-flu antibody, followed by treatment with biotin-tyramide. Biotin was detected with Alexa-488 conjugated streptavidin.

Measurement and analysis of spatial coordinates

The distribution of receptor neurons labeled with a DIG-labeled probe was assessed in complete series of sections of olfactory epithelium. Three spatial coordinates were evaluated: radial distance (center of the lamella to cell position), height within the lamella (basal border of the lamella to cell position; laminar height), and height within the organ (Number of horizontal section from top to bottom; z axis). A graphical description of the measurements is given in Additional file 3. No differences in frequency were observed between left and right side of the center line (the median raphe). Spatial coordinates were measured in arbitrary units and normalized as described [20]. For example, apical-to-basal position within an lamella (laminar height) was measured as the shortest distance between center of the cell and basal border of the epithelial layer, and normalized to the thickness of the epithelial layer at the position of the cell [20]. Thus the range of values is between 0 (most basal) and 1 (most apical). Unbinned distributions were represented as the corresponding empirical cumulative distribution function (ECDF) [38, 39]. In this presentation, data points are sorted by their parameter value (x axis), with their ordinal number (normalized) as y axis. Each data point results in a curve point, thus no information about the distribution is lost in the representation as ECDF, in contrast to the usual histogram representation. To estimate, whether two spatial distributions were significantly different, we have performed Kolmogorov-Smirnov tests on the unbinned distributions using R [40] with the following command: laply(inputfile.csv, function(x) llply(inputfile.csv, function(y) ks.test(x, y)$p.value)). The Kolmogorov-Smirnov test makes no assumptions about the nature of the distributions investigated, which is essential since the skewness of many distributions showed that these are not Gaussian. Due to the sensitive nature of the test on large distributions (n > 100) we selected p < 0.01 as cutoff criterion for significant difference, cf. [25].