Otolith-based species identification in the killifish Aphaniops (Teleostei; Cyprinodontiformes; Aphaniidae) using both morphometry and wavelet analysis

1.1 Otoliths

Otoliths are calcium carbonate increments inside membranous sacs of the inner ear of teleosts (Panfili et al., 2002). They form three pairs of mineralized structures called sagittae, lapilli and asterisci, which are each located in a distinct otolithic end organ (Parmentier et al., 2007). Otoliths serve as a tool to sense balance and sound and are associated with important biological functions such as feeding, swimming or intraspecific communication (Ghanbarifardi et al., 2020; Lombarte et al., 2010; Popper et al., 2005; Schulz-Mirbach et al., 2019). Among the three pairs, sagittae are mostly used in research, as they are usually the larger and easier to recover pair (Assis, 2005). Comparative study of sagitta morphology is a well-established taxonomic approach for species and genus identification (Nolf, 1985, 2013), even in the case of closely related sympatric species (Reichenbacher & Reichard, 2014; Reichenbacher et al., 2023; Zarei, Esmaeili, et al., 2023). Furthermore, the use of otolith morphometrics, geometric landmarks and otolith shape analysis can produce quantifiable and comparable results and thus represents a robust complementary tool in taxonomy and also in evaluating genetic results or life history hypotheses (Castonguay et al., 1991; Tracey et al., 2006; Tuset et al., 2016, 2019).

Otolith morphology (and morphometry) has been applied successfully to separate between species of Aphaniidae suggesting that their otoliths are species-specific and can assist in species discrimination (e.g. Gholami et al., 2014; Reichenbacher et al., 2007; Reichenbacher, Kamrani, et al., 2009; Teimori et al., 2014, 2018). However, quantitative otolith data for closely related species of Aphaniops, specifically focusing on interspecies comparisons, are almost non-existent. Most studies have focused on employing such methods to explore intraspecific otolith variation, that is, variation within and between populations of individual Aphaniops species (Bidaye et al., 2023; Herbert Mainero et al., 2023; Masoumi et al., 2024; Motamedi et al., 2021; Reichenbacher, Feulner, & Schulz-Mirbach, 2009; Teimori, Iranmanesh, et al., 2021; Teimori, Motamedi, & Zeinali, 2021).

2.1 Fish sampling, otolith extraction, SEM imaging

We used a total of 117 specimens from southern Iran, Saudi Arabia and Oman, encompassing Aphaniops dispar, A. hormuzensis, A. kruppi and A. stoliczkanus, for the extraction of their otolith pairs (see Table 1 and Figure 1 for details on sampling points, habitats, sample sizes, sex and specimen sizes). To test the utility of morphometry, particularly outline analysis—which has not yet been applied to comparisons of Aphaniops species—in distinguishing between species, single populations were used for each species to minimize the known impact of geographic variation on otolith morphology (see above). The sampling was done in the years 2017–2019 and complied with protocols approved by the responsible governmental authorities (according to the countries where the sampling was conducted). Species identification was additionally confirmed for three captive-bred A. kruppi based on cyt b (see Charmpila et al., 2020) and for the wild-caught A. kruppi based on CO1 gene markers (Freyhof et al., 2017; Freyhof pers. comm.). Furthermore, 48 left otoliths of A. ginaonis were available from the previous work of Reichenbacher, Kamrani, et al. (2009).

Left and right saccular otoliths (termed otoliths in the following text) were extracted dorsally. Following the protocol of Reichenbacher et al. (2007), residual tissues were removed by immersing the otoliths in a 1% KOH solution for 3 hours, followed by rinsing in distilled water for 4 hours. If necessary, the procedure was repeated, and the otoliths were then stored in distilled water overnight. To reduce possible effects of variation associated with ontogeny or right–left otolith asymmetry, otoliths were sorted into size classes of standard length (Table 1) and only left otoliths were used (exceptionally two right otoliths were used because their left counterparts could not be retrieved).

For scanning electron microscopy (SEM) imaging, otoliths were mounted on aluminium pin stubs (12.5 mm in diameter, 3.2 × 8 mm), to which adhesive tabs had been applied. A thin (20 nm) gold coating was applied to the stubs (sputter coating) in a high vacuum coater. The pin stubs were then inserted into the HITACHI SU 5000 Schottky FE-SEM at the Department of Earth and Environmental Sciences (LMU Munich), and a current of 15 kV was applied. Images were processed using Adobe Photoshop.

2.2 Method 1—Otolith morphology and morphometry

Morphological and morphometric otolith analyses were conducted based on the SEM images of the otoliths. For morphological descriptions of otoliths, the established otolith terminology, as shown in Figure 2, was used (Nolf, 1985; Reichenbacher et al., 2007; Smale et al., 1995; Tuset et al., 2008).

Morphometric analysis was conducted with ImageJ (Schneider et al., 2012) and followed the methodology introduced by Reichenbacher et al. (2007) for otolith variables, and Tuset et al. (2003) and Bani et al. (2013) for otolith shape indices. SEM images were oriented so that the ventral margin and the sulcus acusticus of the otolith were essentially horizontal (Figure 2). According to Reichenbacher et al. (2007), eight linear distances and three angles were measured for each of the otoliths (Figure 2, Table 2). Linear measurements were standardized as a function of either otolith length (OL) or height (OH), which resulted in 10 variables that were used in the statistical analyses. Following Bani et al. (2013), four morphometric measurements (otolith perimeter, otolith area, and maximum and minimum value of ferret diameter calculated at angles of 0, 30, 60, 90, 120 and 150 degrees) were calculated and combined in morphometric equations to produce five shape indices (Table 2).

Univariate statistical analysis was performed based on the morphometric data obtained from the otoliths of size class 1 (20 mm ≤ SL ≤ 32 mm) using RStudio v 2023.3 (Posit Team, 2023). Violin plots for otolith variables, shape indices, and standard lengths were plotted for each species using Ggplot2 (Wickham, 2016). Normality of distribution was tested for each of the otolith variables and shape indices with Shapiro–Wilk's test for each species and for each sex within a species (p < 0.01, if non-normal). All variables and indices were normally distributed, so there was no need for transformation of values to fit a normal distribution. Sex differences for each species were tested with Student's t-test (p < 0.01) (no sex comparisons were possible for A. dispar and A. kruppi due to low sample sizes). As sex differences were not significant (p > 0.01) for A. ginaonis, A. hormuzensis (except antirostrum length and excisura angle) and A. stoliczkanus, female and male specimens were pooled for each species. One-way ANOVA with post-hoc tests was used to test the significance of differences in individual otolith variables and shape indices between groups. The homogeneity of variances was tested with Levene's test. In the case of heterogeneity (p < 0.01), individual variables were tested for differences with Tamhane's-T2 test (p < 0.01), and in the case of homogeneity, Duncan's (p < 0.01) post-hoc test was conducted.

2.3 Method 2—Otolith shape analysis

Otolith outlines were extracted and generated with the ShapeR 0.1–5 package for R (Libungan & Pálsson, 2015) from the SEM images. Images were previously stored in JPEG format (*.jpg) with increased contrast between the otolith and the background, increased brightness of the otolith, and rotated with the rostrum to the left according to Libungan and Pálsson (2015). Equally spaced radii were drawn out starting horizontally from the otolith centroid to the otolith outline with the first drawn radius representing the 0° angle of the otolith outline and the following radii continuing counter clockwise until 360°; X and Y coordinates (polar coordinates) were collected from the drawn radii (Libungan & Pálsson, 2015). To reduce pixel noise, the “smoothout” function of ShapeR was set to 100 iterations. The polar coordinates were returned as wavelet coefficients after implementation of the wavelet transformations with the R package Wavethresh 4.6.8 (Nason, 2016). As no significant sex differences were indicated through method (1), samples were pooled for each species. Mean otolith shapes were plotted for each species using wavelet coefficients. The built-in function in the packages ShapeR and Wavethresh 4.6.8 employs a normalization technique based on regression (Lleonart et al., 2000) to remove the effect of allometric growth on otolith size. This involves scaling the wavelet coefficients with standard fish length (Libungan & Pálsson, 2015). Additionally, coefficients showing a significant interaction between species and fish length (p < 0.01) were automatically omitted from subsequent analyses. Finally, otolith outlines were reconstructed based on standardized wavelet coefficients (Libungan & Pálsson, 2015). The variation of the standardized wavelet coefficients along the otolith outline was shown by plotting the mean and standard deviation of coefficients against the otolith outline angle—see above (Libungan & Pálsson, 2015). Canonical Analysis of Principal Coordinates (CAP) with the standardized wavelet coefficients as variables was used to uncover patterns among species on canonical plots with the use of the function capscale in the R package Vegan 2.5–7 (Oksanen et al., 2020). The constrained ordination of species averages along the first two canonical axes from CAP was assessed graphically (plot not shown). The significance of the joint effect of wavelet coefficients on species ordination was examined with an ANOVA-like permutation test, also implemented in Vegan.

2.4 Further statistical analysis (for both methods 1 and 2)

Linear Discriminant Analysis (LDA) with leave–one–out cross validation was performed for each dataset of the two methods to test for the significance of each dataset in species discrimination. The following packages in R and their functions were used for this: the lda function in the MASS 7.3–60.0.1 (Venables & Ripley, 2002) package and the confusion Matrix function in the Caret (Kuhn, 2008) package. Linear discriminant functions were plotted against each other with Ggplot2.

3.1 Results of otolith morphology

The otoliths of the studied Aphaniops species display the general characteristics of the genus in the following traits: (i) the shape is generally rounded-trapezoid to elongated-trapezoid and the margins are smooth; (ii) the well-formed heterosulcoid sulcus acusticus is in median position and extends more than ¾ of the length of the otolith; (iii) the sulcus opening is ostial and the cauda is markedly curved in its posteroventral end; (iv) the slightly deepened tubular ostium is usually shorter compared to the cauda (Figures 3 and 4). In addition, differences in specific otolith characters are identifiable in each species (Figures 3 and 4) and are summarized in Table 3. In short, A. dispar otoliths display two dorsal tips and a posteroventral projection, a short but well-formed rostrum and antirostrum and a deep excisura angle (Figure 3a–d). None of the other studied species displays a clear dorsal tip (Figure 3e–t), with very few exceptions, especially among the otoliths of A. hormuzensis (Figure S2). Aphaniops ginaonis and A. hormuzensis otoliths exhibit a medium-sized, more or less pointed rostrum and antirostrum, a V-shaped deep excisura angle (Figure 3e–l, Table 3), while the posterior margin is slightly more rounded and more symmetrically shaped in A. ginaonis. Additionally, high overall variability was noted in the A. hormuzensis subset (n = 22) (Figure S2). The otoliths of A. kruppi display a rounded-triangular outline with a short, rounded or truncated broad rostrum and U-shaped excisura angle (Figure 3m–p). Notably, the otoliths from the wild-caught A. kruppi specimens display a less developed antirostrum and shallower excisura angle as seen in the captive-bred A. kruppi (Figure 3m–p vs. Figure S1, Table 3). Aphaniops stoliczkanus exhibits a short to medium-sized round rostrum, a well-defined antirostrum and a U-shaped, moderately deep excisura (Figure 3q–t, Table 3).

3.2 Results of otolith morphometry

Violin plots for otolith variables, shape indices, and standard lengths of each species are shown in Figure 4. The importance of each otolith variable in the separation among species was tested using one-way ANOVA with post-hoc tests (Duncan's test, if variance was homogeneous; Tamhane-T2 test in case of heterogeneous variances). Pairwise differences between species in otolith variables and indices are summarized in Table 4. In short, A. ginanonis differed significantly from all other species in the length–height ratio, circularity and ellipticity, which are all measures of overall shape. Furthermore, A. hormuzensis was significantly different vs. all species in the length–height ratio, and A. kruppi vs. all species in the antirostrum height and length and the excisura angle (Table 4). Moreover, otolith variables associated with the rostrum and antirostrum (heights and lengths), and overall otolith shape (length–height ratio, form factor, circularity) were significantly different between several species. The highest differentiation was shown for A. ginaonis and A. kruppi which differed vs. all other species respectively in several variables and indices (Table 4).

3.3 Results of otolith shape analysis

Figure 5 shows the mean otolith outline of each species. Clear variation along the otolith outline was detected between 170° and 220°, corresponding roughly to the rostrum, excisura and antirostrum areas, and also between 10° and 70°, representing the posteroventral portion of the otoliths (Figure 5). CAP of standardized wavelet coefficients from all five species showed four principal coordinates with CAP1 (61.9%) and CAP2 (28.7%) explaining most of the variation in otolith shape, but species were overlapping when plotted against CAP1 and CAP2 (plot not shown). The results of the overall and pairwise ANOVA-like permutation tests to evaluate the significance of otolith shape variation among species are summarized in Table 5. The outcome indicated significant differences in otolith shape (p < 0.01) between almost all species pairs, but not for A. ginaonis versus A. hormuzensis, A. hormuzensis versus A. kruppi and A. kruppi versus A. stoliczkanus.

3.4 Results of multivariate analysis

Overall classification of otoliths after cross validation for the dataset of otolith variables and shape indices yielded 78.7% success (Table 6, Figure S3). In contrast, overall classification success after cross validation for the dataset of standardized wavelet coefficients was only 40% (Table 6). In both datasets, A. dispar had the highest score of correct classification (Table 6).

4.1 Distinctions among Aphaniops species using otoliths

Previous studies on the otolith morphology of Aphaniops have suggested that the otoliths of species within this genus possess species-specific characteristics, as well as features that differentiate among populations (Reichenbacher et al., 2007; Reichenbacher, Feulner, & Schulz-Mirbach, 2009; Reichenbacher, Kamrani, et al., 2009; Teimori, Schulz-Mirbach, et al., 2012). Specifically, the otolith length–height ratio, the relative height of both the antirostrum and rostrum, and the relative length of the rostrum are characters that vary among species. In contrast, characters associated with the excisura and posterior angle, along with the relative medial length and relative antirostrum length, are suitable for distinguishing among populations. Notably, in our study, the excisura angle and antirostrum length were variables that contributed to the distinction of the otoliths of A. kruppi from the other species (see below).

4.2 Distinction of Aphaniops dispar using otoliths

Among the here studied species, the otoliths of A. dispar were the most distinguishable in otolith morphology (Table 3). Although there is no otolith variable or shape index that distinguishes all species from one another (Table 4), multivariate analyses for each method successfully separated the otoliths of A. dispar based on their overall otolith shape (Table 6, Figure S3). According to the time calibrated phylogeny in Teimori et al. (2018), A. dispar diverged prior to the other Aphaniops species studied here in the Late Miocene–Early Pliocene (95% age credibility interval 9–4 million years ago [MYA]). The older age of A. dispar may be reflected in the more distinct morphology of its otoliths compared to the other Aphaniops species studied here. Another Aphaniops species, A. furcatus (not included in this study), which diverged from A. dispar in the middle Miocene (15.2–11.7 MYA) according to Teimori et al. (2018), also exhibits unique otolith morphology (Motamedi et al., 2021; Teimori et al., 2014). The idea that the geological age of a species can increase its otolith distinctness is further supported by the distinctive otolith morphology of two other species of the Aphaniidae, Esmaeilius vladykovi (Coad, 1988) and E. shirini (Gholami et al., 2014). Their otoliths are shown in Gholami et al. (2014: figure 4, as Aphanius). These two species may have diverged as early as the Middle or Late Miocene (Gholami et al., 2014). Additionally, E. vladykovi has been identified as the oldest species among the Iranian inland clade of the Aphaniidae (Esmaeili et al., 2014, as Aphanius).

4.3 Distinction among Aphaniops species from southern Iran using otoliths

According to Teimori et al. (2018), the divergence of A. ginaonis, A. hormuzensis and A. stoliczkanus occurred in the Late Pliocene–Early Pleistocene (3.2–2.0 MYA). Their relatively young geological age may account for the similarity of their otoliths, as revealed in our analyses (Figures 4 and 5). Nevertheless, otoliths of A. ginaonis were clearly distinguished from those of the other species due to their elongated shape (see shape index differences in Table 4 and otolith contour in Figure 5a). Aphaniops hormuzensis and A. stoliczkanus were less clearly distinguishable, particularly A. hormuzensis, whose classification based on the two methods did not yield high success; only the length–height ratio clearly separated it from all other species (Table 4). Visual inspection of the complete otolith sample of A. hormuzensis (n = 22) revealed considerable variability in otolith morphology (Figure S2), potentially posing a challenge for both methods in reliably distinguishing A. hormuzensis otoliths from those of other species.

4.4 Otolith variation among A. stoliczkanus from inland and coastal sites

The studied otoliths of A. stoliczkanus from the Mirahmad hot mineral spring, a land-locked site in the Helleh Basin in Southern Iran, have a discoidal shape and lack a dorsal tip, which aligns with previous descriptions of this species from the same or other inland sites in Iran and Iraq (see Teimori, Jawad, et al., 2012, Teimori, Schulz-Mirbach, et al., 2012 [as A. dispar]; Teimori et al., 2018). Conversely, the coastal and coastal-associated populations of A. stoliczkanus in Reichenbacher et al. (2009, as A. dispar), Teimori et al. (2012, as A. dispar), Teimori et al. (2018) and Bidaye et al. (2023) always present a more triangular otolith shape with a more or less pronounced dorsal tip.

The formation of a dorsal tip on otoliths has been associated with environmental factors such as distance from the sea and salinity level (Bidaye et al., 2023; Reichenbacher, Feulner, & Schulz-Mirbach, 2009; Teimori, Jawad, et al., 2012; Teimori, Schulz-Mirbach, et al., 2012; Teimori et al., 2018). However, an experimental study on reared specimens of purebred Atlantic herring and hybrid Atlantic/Baltic herring strongly suggests that salinity has minimal influence on otolith morphology, and that genetic factors are the primary driver of otolith variation (Berg et al., 2018). This suggests that coastal and coastal-associated populations of A. stoliczkanus from the Gulf of Oman may represent a distinct phylogenetic lineage within the species, as reflected by the consistent presence of a dorsal tip in their otolith morphology. On the other hand, COI barcoding data of A. stoliczkanus from Oman by Zarei, Masoumi, et al. (2023) detected only closely related haplotypes, separated by single mutational steps, between coastal and freshwater populations of this species.

4.5 Distinction of Aphaniops kruppi using otoliths

The otoliths of A. kruppi have not been previously examined in comparative studies. According to Teimori et al. (2022), A. kruppi is sister to the clade containing A. ginaonis, A. hormuzensis and A. stoliczkanus, suggesting that A. kruppi is also a relatively young species with an age dating back to the Late Pliocene–Early Pleistocene (3.2–2.0 MYA). Despite its young geological age, the otoliths of A. kruppi studied here can be distinguished from those of its congeners by a unique combination of characteristics, including the U-shaped excisura, short rostrum and weakly developed antirostrum (Figure 3m–p, Figure S1). Notably, otolith contour analysis (method 2) achieved minimal success in distinguishing between A. kruppi, A. hormuzensis and A. stoliczkanus (Table 6), likely attributed to the similar rounded shape of their otoliths, causing overlap and confusion. Conversely, the results of otolith morphometry (method 1) clearly reinforce that A. kruppi otoliths possess unique characteristics, with significant differences observed in three otolith variables (excisura angle, relative antirostrum height and relative antirostrum length) compared to all other studied species (Table 4). Remarkably, two of these variables, namely excisura angle and antirostrum height, are known to indicate differences between isolated populations rather than species (Reichenbacher et al., 2007). The significance of these variables in separating A. kruppi likely stems from the species' relatively young geological age.

Another interesting outcome was that otoliths from wild specimens of A. kruppi exhibit attenuated features compared to those from captive-bred specimens (see Figure 3m–p vs. Suppl. Figure S1, Table 3). These differences may be attributed to the distinct size difference between wild-captured (SL 29–32 mm) and captive-bred specimens (SL > 39 mm). Other reasons could be differences in water temperature, chemistry and food availability between the natural and the aquarium environment as such differences have been shown to influence otolith growth and morphology across multiple fish groups (e.g. D'Iglio et al., 2021; Gagliano & McCormick, 2004; Katayama & Isshiki, 2007; Lombarte & Lleonart, 1993; Mérigot et al., 2007; Tuset et al., 2003).