Ontogenetic shifts in olfactory rosette morphology of the sockeye salmon, Oncorhynchus nerka

Abstract Sockeye salmon, Oncorhynchus nerka, are anadromous, semelparous fish that breed in freshwater—typically in streams, and juveniles in most populations feed in lakes for 1 or 2 years, then migrate to sea to feed for 2 or 3 additional years, before returning to their natal sites to spawn and die. This species undergoes important changes in behavior, habitat, and morphology through these multiple life history stages. However, the sensory systems that mediate these migratory patterns are not fully understood, and few studies have explored changes in sensory function and specialization throughout ontogeny. This study investigates changes in the olfactory rosette of sockeye salmon across four different life stages (fry, parr, smolt, and adult). Development of the olfactory rosette was assessed by comparing total rosette size (RS), lamellae number, and lamellae complexity from scanning electron microscopy images across life stages, as a proxy for olfactory capacity. Olfactory RS increased linearly with lamellae number and body size (p < .001). The complexity of the rosette, including the distribution of sensory and nonsensory epithelia and the appearance of secondary lamellar folding, varied between fry and adult life stages. These differences in epithelial structure may indicate variation in odor‐processing capacity between juveniles imprinting on their natal stream and adults using those odor memories in the final stages of homing to natal breeding sites. These findings improve our understanding of the development of the olfactory system throughout life in this species, highlighting that ontogenetic shifts in behavior and habitat may coincide with shifts in nervous system development.

The olfactory system of salmonids has been given particular attention, given their unusual life history: embryos hatch in freshwater, emerge as fry from their gravel nests, and then rear in freshwater (parr stage) until undergoing a series of changes in physiology, morphology, and energetics, termed smolt transformation, that prepare them to migrate downstream from the lake and make the transition from fresh to salt water (Hoar, 1976). During their adult life stages, they migrate long distances back to their natal stream, relying on odor memories to return to the natal site for spawning, years after they left it as juveniles. Thus, the olfactory system must be sufficiently developed at early juvenile stages to acquire chemosensory information needed for learning (imprinting to) natal stream odors that are used later to guide homing migrations (Barnett et al., 2019;Brannon, 1972;Quinn et al., 2006;Zielinski & Hara, 1988).
For Oncorhynchus spp., it is proposed that individuals imprint during early development on odors, including dissolved free amino acids, specific to their natal streams (Shoji et al., 2000(Shoji et al., , 2003Ueda, 2012), which leads to a memory-associated migration (Ueda, 2019). This imprinting period likely occurs during embryonic development (Dittman et al., 2015;Tilson et al., 1994), before the parr-smolt transformation (Shrimpton et al., 2014), and again during the smolt life stage (Havey et al., 2017). Later, during the homeward migratory phase, maturing adults use these imprinted odors, perhaps augmented with odors from conspecifics (Bett & Hinch, 2016), to discriminate their natal stream from others with high precision Hasler & Scholz, 1983).
In salmonids, the detection of chemosensory cues is governed by a well-developed peripheral sensory organ, the olfactory rosette, and a relatively large olfactory bulb (the first-order olfactory processing center in the brain; Døving et al., 1985;Wisby & Hasler, 1954). The olfactory organs in teleost fishes are found in paired nasal cavities (nares), situated on the dorso-rostral end of the snout (Hara, 1994;Kermen et al., 2013) (Figure 1). There are two olfactory rosettes (left and right), each positioned internally between the anterior and posterior nares. Each olfactory rosette is comprised of a central raphe that is surrounded by olfactory lamellae on either side (Figure 1; Hansen & Zeiske, 1998, Hansen & Zielinski, 2005Zeiske et al., 1992).
An individual lamella is made up of a sensory epithelium that develops to create two layers of olfactory lamellae: primary and secondary (Calvo-Ochoa & Byrd-Jacobs, 2019;Fishelson et al., 2010;Zeiske et al., 1992). In teleosts, primary lamellae comprise the olfactory epithelium, which contain the olfactory receptor neurons (ORNs), responsible for the recognition of biologically relevant, water-borne odorants (Zeiske et al., 1992). Secondary lamellae, present in some groups of teleosts (Kasumyan, 2004), are described as the lamellar folds present on the primary lamellae.
In salmonids, the importance of odors for homing suggests that changes in the size and complexity of the olfactory system at key life stages may parallel variation in olfactory processing and/or capacity. Early olfactory rosette development may occur before hatching (Brannon, 1972;Zielinski & Hara, 1988), but continues throughout early life stages, with variation in rosette morphology across ontogeny. Lamellae count increases markedly across ontogeny in chum salmon (O. keta) (Kudo et al., 2009) and the absolute number of ORNs in the sensory epithelium changes from hundreds of thousands in fry to tens of millions of cells in mature adults of this species (Kalinina et al., 2005;Kudo et al., 2009;Yamamoto & Ueda, 1977). Surprisingly, there have been no studies to quantify how olfactory function covaries with anatomical differences in salmon. However, morphological changes in the olfactory system may reflect differences in olfactory capacity (as part of the imprinting or odorrecall process) in salmon, corresponding to discrete life stages and varying habitats across ontogeny.
This study assessed morphological differences in the olfactory rosette of sockeye salmon across a series of life stages: fry that had just entered the lake, parr feeding in the lake, smolts leaving the lake a year later, adults that have left the ocean and are migrating into their natal river system, maturing adults, and fully mature fish entering the spawning stream itself. The gross morphology and ultrastructure of the olfactory rosette and the sensory portions of the olfactory epithelium was assessed using scanning electron microscopy (SEM). This study tested the hypotheses (1) that the relative size of the olfactory rosette increases throughout ontogeny, (2) that there is an increase in primary lamellae number across the life stages, and (3) that there is an increase in lamellar complexity, as defined by anatomical changes to the sensory epithelium and the presence of secondary folding.

| Specimen collection
A total of 26 specimens of Oncorhynchus nerka (Walbaum, 1792) across different life stages were collected from Lake Aleknagik, Alaska, and Hansen Creek, one of its tributaries, according to the ethical guidelines of the University of Washington (IACUC protocol #3142-01). Specimens were categorized based on their developmental stage (Groot & Margolis, 1991) and included fry (n = 3), parr (n = 4), smolt (n = 5), and adults (n = 12). Samples (fry-maturing adults) were collected from Lake Aleknagik, whereas spawning adults were collected at Hansen Creek, a tributary of the lake, in late July (Table 1). Fry, parr, and smolt specimens were collected using a seine net, tow net, and fyke net, respectively. Adults (six males and six females) represented three points along their migratory continuum: "migrating" (just entered freshwater) and "maturing" (transitioning in freshwater to maturity), both collected with gill nets, and "spawning" adult (fully mature, ready to spawn, and entering their natal stream), collected in Hansen Creek itself. Small fishes (fry, parr, smolt) were euthanized via an overdose of MS-222 (m-aminobenzoic acid ethyl ester, methansulfate salt), buffered to neutral pH, and adults were euthanized with a sharp blow to the head. Upon collection, morphometric measurements for each individual were recorded, including body mass (g; Table 1). Immediately after euthanasia,

| Olfactory rosette preparation
Based on previous research suggesting no bilateral differences between the left and right olfactory structures in fishes (e.g., Camilieri-Asch, Shaw, et al., 2020), the left rosette was arbitrarily dissected from each specimen to assess its morphology. Following postfixation, the olfactory rosette was separated from the olfactory nerve (nI) and immersed in a 0.1 mol L −1 phosphate buffer solution (PBS) for 24 h. Rosettes were then submerged in 1:1 ratio of osmium tetroxide in a 0.1 mol L −1 PBS for 1 h. Samples were then washed with deionized (DI) water to remove excess fixative and dehydrated through a classic graded ethanol series (20 min for each series). After dehydration, samples were placed in 100% ethanol, critical point dried, and sputter coated with 13.1 nm platinum.

| SEM and morphological assessment
Variations in olfactory rosette morphometrics throughout ontogeny were assessed using a ApreoS HiVac SEM, at a working distance of approximately 10 nm, a beam strength of 5 kEv, and a 13-spot size.
Low (×10-×100) and high (×500-×10,000) magnification images were acquired to estimate RS and assess its organization and composition (as proxies for complexity) across individuals and life stages. Data on RS and lamellar number were acquired from low magnification images for fry, parr, smolt, and adults. However, comparisons of rosette complexity from high magnification images are restricted to fry and adult life stages only, as higher magnification imaging of the sensory epithelium for parr and smolt was not possible, due to damage to the epithelia during tissue processing.
Assessment of lamellae at low and high magnification was conducted on an intact rosette. The digital images acquired were saved/ downloaded as (1536 × 1094, RBG Gray, uncompressed), exported as.TIFF files, and assessed in Adobe Photoshop ® .
For all samples, the gross morphology of the rosette was assessed and lamellae number and total olfactory RS (in mm 2 ) were measured. For all samples, rosette ultrastructure was assessed from digital images in Adobe Photoshop ® to count the number of lamellae in each specimen and calculate size of the total olfactory rosette (in mm 2 ). As the olfactory organ is cylindrical and dorsoventrally flattened in shape, total RS was calculated from an image of the dorsal face of the rosette (i.e., the portion exposed to the incurrent opening of the nare), captured with SEM. RS was calculated by approximating the shape of the rosette as an oval. A digital overlay of the oval shape was delineated in Adobe Photoshop ® and included all epithelial tissue across the multiple lamellae, but excluded bony structures and epineurium in the olfactory cavity. Using the Ruler Tool, a measurement scale was set by assigning a specified number of pixels per number of scale units (mm) for each image. RS was then extrapolated by counting the total number of pixels within the overlain oval, converted to area (mm 2 ).

| Statistical analysis
To test for differences in the olfactory rosette across different life stages, the number of lamellae and total olfactory RS were averaged, and standard deviations (±SD) reported for each life stage (fry, parr, smolt, and adult). These parameters were compared between life stage categories using an analysis of variance. For purposes of statistical analyses, all three adult stages were pooled because they did not differ in RS (F(1, 10) = 0.185, p = .676) and lamellae count (F(2, 9) = 0.321, p = .733) across adult life stages (Table 1).
To assess the scaling relationship between RS and body size throughout ontogeny, RS was scaled against body mass using an ordinary least squares (OLS) linear regression (y = ax b ). Before analysis, body mass (x) and RS (y) were log 10 -transformed. A qualitative description of the sensory epithelium was then performed on SEM images, to identify the boundary between sensory versus nonsensory epithelia (×10-×1000 magnification), the presence of ORNs (apical olfactory knobs [OKs]; ×2500), and, when possible, morphological differences between the OKs at the epithelial surface (at ≥×2500) of fry and adult individuals. Note that individual lamellae T A B L E 1 The collection date (all in 2016), sample size, average body mass (g ± SD), average lamellae count (±SD), and rosette size (mm 2 ± SD) for fry, parr, smolt, and adult life history stages were not able to be nondestructively removed from the central raphe and assessment of the lamellae (quantitative and qualitative) were performed while positioned on the intact rosette. All quantitative statistical analyses were conducted in R (R Core Team, 2020).

| Gross morphology of the olfactory rosette
Sockeye salmon (O. nerka) have paired olfactory organs located inside the olfactory cavity on either side of the head. The arrangement and shape of the lamellae forming the rosette was consistent across all life stages and thus did not change throughout ontogeny. This species exhibits an "arrow-like" olfactory organ, similar to other salmonids (see review by Kasumyan, 2004). Specifically, in an oval-shaped rosette, lamellae are positioned radially around a central raphe, and the "origin" of the radiation was slightly anterior to the oval center; further, lamellae appeared to increase in size incrementally from the anterior to posterior ends of the rosette (Figures 1 and 2). For each lamella, it is larger at the origin from the median raphe and tapers toward the edges of the rosette, at the outer lamellar margin (Figure 2). There was a visible increase in lamellar complexity and RS at the adult stage compared to earlier life stages. Adult specimens displayed prominent secondary folding (secondary lamellae) that have both sensory and nonsensory epithelial regions (Figures 2d and 5c); but, as individual lamellae could not be isolated, percent cover could not be quantified. In contrast, juveniles (fry, parr, and smolt) showed no visible secondary lamellae (Figure 2a-c). In fry, distinct regions of sensory versus nonsensory epithelia could not be visualized on the lamellar surface at high magnification. Rather, only sensory epithelia were observed across the exposed portions of the lamellae (Figure 2a).

| Lamellae count and RS
The average number of lamellae increased throughout ontogeny, from fry to adult stages (   F I G U R E 4 Scaling relationship between and log 10 -transformed rosette size (mm 2 ) and log 10transformed body mass (g) of sockeye salmon across four different life stages indicated by color: white = fry, light gray = parr, dark gray = smolt, and black = adult. Gray area represents 95% CI (0.530, 0.600).

| Sensory epithelia in fry and adult stages
Two types of epithelia were found on the rosette of fry and adult specimens: a nonsensory epithelium, populated with large supporting cells that bear microvilli or cilia ("kinociliated cells"; e.g., Cox, 2008), and a sensory epithelium, populated by ciliated OKs. However, in fry, the sensory epithelium was densely packed with OKs of a similar appearance that did not appear homogeneously distributed across the lamellar surface, while the remaining sensory area was covered by cilia-bearing supporting cells (×2500; Figure 6a). The adult sensory epithelium comprised more dispersed OKs, which qualitatively appeared more evenly distributed than in fry. Although they could not be empirically identified, apparent variation in OKs was suggestive of multiple cell types across the epithelial surface (Figure 6b).

| DISCUSSION
The peripheral olfactory system of fishes is well-developed and highly variable in shape and organization across species (Cox, 2008;Hara, 2011;Kasumyan, 2004). Considerable interspecific variation has been documented in the gross morphology, anatomy, and ultrastructure of the fish peripheral olfactory system. These include differences in the presence and size of the nasal bridge between nostrils, total rosette size, lamellar arrangement within the rosette, lamellae number and lamellar surface area, and the total number, density, distribution, and type of ORNs across the sensory epithelium (Døving et al., 1985;Hansen & Zielinski, 2005;Kasumyan, 2004;Yamamoto, 1982;Zeiske et al., 1992), as well as variation in the size and shape of the olfactory bulbs themselves (Kotrschal et al., 1998;Wagner, 2001;Yopak et al., 2015Yopak et al., , 2019 within and across fish groups. However, while interspecific variation in olfactory anatomy has been reasonably well explored in fishes, intraspecific diversity in this system is less understood (e.g., Bauchot et al., 1979;Brandstätter & Kotrschal, 1989;Kihslinger et al., 2006;Näslund et al., 2017;Tomoda & Uematsu, 1996).
Consistent with our predictions, morphological changes occur in the peripheral olfactory organ of sockeye salmon throughout ontogeny, which may underlie imprinting, migration, and/or natal homing in this species. In accordance with the assumption that larger individuals have larger olfactory organs (Hansen & Zielinski, 2005), both RS and complexity (lamellar number and morphology, epithelial distribution, and composition) change throughout life in this species.
The relationship between lamellar growth in the peripheral rosette, olfactory processing in the brain, and memory formation is poorly understood. Evidence suggests that salmon imprint to chemosensory patterns during juvenile stages Dittman et al., 2015;Yamamoto et al., 2013) that can then be utilized later in life to identify their natal stream as migrating adults.
Therefore, the presence of specific lamellae as early as the fry stage may be critical for imprinting. In this case, as mature adult sockeye complete their homing migrations, only~7 of the 14-15 lamellae within the rosette would have been present at the time of imprinting, similarly proposed for chum salmon (Kudo et al., 2009). However, there is currently no empirical evidence that links specific lamellae to imprinting mechanisms, so this requires further study. In addition to changes in lamellae number throughout ontogeny, lamellar size also varies in sockeye salmon individuals. For all life stages, a qualitative assessment of visible portions of the rosette suggests the posterior lamellae are the most well-developed (i.e., the largest), as compared to those positioned more anterior in the nare (Figure 2), which is consistent with olfactory rosette anatomy in other teleost fishes (Chakrabarti & Ghosh, 2011;Dymek et al., 2021;Ferrando et al., 2019;Hansen & Zeiske, 1998;Hansen & Zielinski, 2005;Kudo et al., 2009).
Distinct apical OKs of the ORNs were identified within the sensory epithelium in sockeye salmon (Figure 7). Variation in the F I G U R E 6 Oncorhynchus nerka, scanning electron micrographs, presence of olfactory knobs (OKs) on the sensory epithelia of a representative (a) fry and (b) adult specimen. OKs are indicated by white arrows.
density and distribution of these OKs along the epithelial surface may provide insight into changes in olfactory capacity at these distinct life history stages (Ahuja et al., 2014;Oka et al., 2012;Shoji et al., 2000).
There are five main ORN morphotypes currently described in teleosts: ciliated, microvillous, crypt (Fishelson et al., 2010;Kermen et al., 2013;Muller, 1984), kappe (Ahuja et al., 2014), and pear (Calvo-Ochoa & Byrd-Jacobs, 2019). Although empirical identification of individual subtypes was outside of the scope of this study, as they cannot be confirmed without immunohistochemistry (e.g., Camilieri-Asch, , the presence of OKs with apparent morphological differences between fry and adult specimens suggests there may be ontogenetic changes in ORN subtype number and density. However, it is important to approach this with caution, as morphological differences from SEM can be due to fixation artifacts or tissue preparation, rather than true anatomical variation. In fry, the sensory epithelium was densely packed with OKs of a similar appearance that did not appear to be homogeneously distributed across the lamellar surface, while the remaining sensory area was covered by cilia-bearing supporting cells (×2500; Figure 6a).
Additional OKs may have been present, but potentially obscured by the dense cilia of the supporting cells (Figures 6 and 7). In contrast, the sensory epithelium in adults presented a more dispersed and morphologically diverse OK distribution, which qualitatively appeared more evenly distributed than in fry ( Figure 6b). Critically, future work should empirically identify and quantify ORN subtype number and density throughout ontogeny, which may inform the relative capacity to bind different odorant classes (e.g., amino acids, bile salts, pheromones) at these key life stages.
Not all teleost fishes exhibit secondary lamellar folds (currently only described in Salmoniformes, Gadiformes, Esociformes, many species of Anabantiformes, and some species of Perciformes (Kasumyan, 2004). In sockeye, during the fry, parr, and smolt life stages, no secondary lamellae were observed (Figure 2a-c), but they were evident in the adults (Figures 2d and 5). This is similar to chum salmon (Kudo et al., 2009), in which there is a distinct shift in lamellar complexity, with the appearance of secondary lamellae in mature adults. The lack of secondary lamellae in juvenile sockeye salmon suggests a lesser total lamellar surface area compared to adults, which may reflect differences in ORN number (Kudo et al., 2009) compared to mature individuals (14.2 million; Kudo et al., 2009), though overall ORN density may not change (Kalinina et al., 2005).
However, while secondary lamellae may increase total lamellar surface area, ORN distribution is not homogenous and sensory epithelium was not identified on the peaks of the secondary folds in adult sockeye. Similarly, previous work in salmonids showed that the convex areas of the secondary folds lack ORNs (Kudo et al., 2009;Olsen, 1993;Yamamoto & Ueda, 1977). This is in contrast to many elasmobranch fishes, where the sensory epithelium covers the secondary lamellae (both troughs and peaks), with areas of nonsensory epithelium along the inner margins of the lamellae and/ or as intermittent projections into the sensory regions in some species (e.g., Camilieri-Asch, Shaw, et al., 2020;Schluessel et al., 2008;Theiss et al., 2009;Simonitis & Marshall, 2022). In adult sockeye salmon, as in other salmonids, the sensory epithelium lies only along the troughs of the secondary folds, while the nonsensory epithelium covers the secondary lamellar peaks (Figure 5c). Thus, secondary lamellae in adults may not serve to expand olfactory capacity, but rather may facilitate water dynamics in the olfactory capsule for efficient odor sampling (Kudo et al., 2009 conceptualization; funding acquisition; data curation.

ACKNOWLEDGMENTS
The authors thank the field sampling team and especially Jackie Carter and Curry Cunningham for specimen collection in Alaska. We also gratefully acknowledge the UNCW Richard M. Dillaman Bioimaging facility at UNCW for access to microscopy equipment, and specifically Dr. Alison Taylor for her expertise and guidance during sample preparation and imaging. We are also very grateful to E. Peele, V. Camilieri-Asch, and C. Stehr for helpful comments on the manuscript and analyses. Special thanks to C. Connor for assistance with Figure 1e. KEY acknowledges startup funds, a Charles L. Cahill Grant, and a CAS Research Initiative Award from UNCW, which funded this study.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.