Species specificity and sexual dimorphism in tooth shape among the three sympatric haplochromine species in Lake Kivu cichlids

Abstract Tooth shape is used to differentiate between morphologically similar species of vertebrates, including fish. This study aimed to quantify tooth shape of three sympatric species: Haplochromis kamiranzovu, H. insidiae, and H. astatodon endemic to Lake Kivu, whose existing identification criteria are currently only qualitative. A quantitative tooth shape analysis was performed based on digitized tooth outline data with a subsequent elliptic Fourier analysis to test for differences among the three species. We looked at crown shape and size differences within H. kamiranzovu and H. insidiae at geographical, habitat, and gender levels. No comparison at habitat level was done for H. astatodon because it is found only in littoral zone. The analysis revealed significant tooth shape differences among the three species. Haplochromis astatodon had a significantly longer major cusp height and a longer and larger minor cusp than that of H. insidiae. It had also a longer major cusp height and a longer and larger minor cusp than that of H. kamiranzovu. Tooth shape differences of H. kamiranzovu and H. insidiae species were not significantly different between littoral and pelagic fish (p > .05) while differences were significant between southern and northern Lake Kivu populations (p < .05). Tooth sizes in H. kamiranzovu and H. insidiae were significantly different, both in height and width as well as in their ratios, and this was true at sex and geographic levels (p < .05), but not at habitat level (p > .05). Tooth shape was also significantly different with sharp teeth for males compared with females of southern populations versus northern ones. These shape‐ and size‐related differences between sexes suggest differences in the foraging strategies toward available food resources in the lake habitat. Further research should explain the genetic basis of the observed pattern.


| INTRODUC TI ON
Tooth shape is often used as a taxonomic tool in vertebrates. For example, the enamel-dentine junction morphology and enamel thickness in early hominid molars proved to be useful for taxonomic studies (Ayvazyan, Vasilyan, & Böhme, 2019;Delgado et al., 2015) and the cranial dental morphology was used to assess the taxonomic status of the hoolock species in China (Fan et al., 2017). Crown height and crown base were found to be key identification features for different phytosaur teeth from the upper Triassic of India (Datta, Kumar, & Ray, 2019) and the upper molar morphology appeared to be taxonomically diagnostic within the genus Pongo (Ortiz et al., 2019).
For example, among Lake Victoria cichlids, having unicuspid and obliquely truncated teeth is one of the most distinctive features to identify Haplochromis obliquidens Hilgendorf, 1888 (Witte & Oijen, 1990). However, the precise characterization of the species-specific differences sometimes remains ambiguous. In Lake Kivu, identification of Haplochromis astatodon Regan, 1921, H. insidiae andH. kamiranzovu Snoeks, 1987 proved to be very difficult. This confusion is due to the fact that some specimens of H. kamiranzovu have a tooth profile that might be confused with that of the somewhat truncated teeth with an expanded crown found in H. insidiae (Snoeks, 1994).
For Haplochromis astatodon, it was noted that this species can unambiguously be distinguished from all other Lake Kivu haplochromines except H. insidiae by the characteristic shape of its outer oral jaw teeth. Teeth of H. astatodon are more oblique and also more truncated than those from H. insidiae (Snoeks, 1994), but without a clear difference between them. In this scenario, the use of outlines, combined with an Elliptic Fourier Analysis (EFA) would show clear differences in tooth shape between Haplochromis astatodon, H. insidiae and H. kamiranzovu.
Dentition was also found to be an excellent indicator of feeding type in cichlid fishes (Albertson & Kocher, 2006;Hulsey et al., 2006). Considering local variation in niche traits within a lake environment, studying tooth morphology of specimens from heterogeneous local habitats might reflect micro-level trophic variation. For example, fish from pelagic and littoral habitats in Lake Kivu, Rwanda, were found to represent two distinct environments (Masilya, Darchambeau, Isumbisho, & Descy, 2011). In Lake Kivu, littoral substrates comprise submerged rocks, macrophytes, and sandy, muddy, rock-sandy, and mixed habitats where zooplankton density is 15-50 times lower than in the pelagic zone (Isumbisho, Kaningini, Descy, & Baras, 2004;Isumbisho, Sarmento, Kaningini, Micha, & Descy, 2006). The coastal area is abundant with Cladophora on the rock surface (Verbeke, 1957) and nauplii , while the pelagic zone is dominated by a complex microbial community, including bacteria and phytoplankton (with taxa like Nitzschia bacata and Fragilaria danica and cyanobacteria like Planktolyngbya limnetica and Synechococcus sp., Wimba, 2009;Sarmento et al., 2007)  .
The adaptive value of tooth size in the three Lake Kivu haplochromines has not yet been studied in relation to the habitat (littoral vs. pelagic). Studying tooth size variation of the haplochromines species offers an opportunity to explore the relationship between crown size and habitat gradient from the littoral versus pelagic zone haplochromines of Lake Kivu.
We tested the hypothesis that tooth shape indeed varies among Haplochromis astatodon, H. insidiae and H. kamiranzovu, and hence is a taxonomically reliable trait among these Lake Kivu cichlids. We also examined the hypothesis that tooth size vary among Haplochromis astatodon, H. insidiae and H. kamiranzovu.
Specifically, we examined whether Haplochromis astatodon, H. insidiae, and H. kamiranzovu can be identified using tooth shape, and whether tooth size (i.e., height and width) of major cusps can be another tool to aid in diagnosis of the three species. We also tested for tooth shape differences between sexes, habitat, and region in major cusp heights and widths of the three species.

| Study area and specimen preparation
Teeth were obtained from adult fish (Figure 1) caught using 15 × 1 m gillnets made from monofilament nylon (10 mm mesh size) in littoral and pelagic zones of northern and southern regions of Lake Kivu. In the north, we caught fish from the Brewery bay of Gisenyi, Berries of Paradise motel, Kigufi bay, and Mouth of Sebeya River. After several unfruitful fishing of the targeted species along the south shore, sampling was systematically done only at three sites where both haplochromines species occurred in Nyamasheke (Nyamasheke 1, Nyamesheke 2, and Nyamasheke 3). These sampling locations were chosen in order to get a disparate representation of tooth shape variation across the lake. In addition, one sampling was done in the open waters of the northern part of the lake and another sampling was done in the open waters of the southern part ( Figure 2).
Haplochromines samples were collected during the dry season (from June to September 2018). Since the haplochromine species we investigated are very rare (Kaningini et al., 1999), fish samples were collected during early morning, late morning, and in the afternoon (05 00-07 00 hr; 10 00-12 00 hr and 15 00-17 00 hr) using eight mm mesh nets (15 × 2 m) to get a representative number of fish samples for statistical analyses.
These are periods of intense activity of H. kamiranzovu (Ulyel, Ollevier, Ceuster, & Audenaerde, 1990). On the first days of the month, only littoral samples were collected in the morning, late morning, and in the afternoon, while pelagic samples were collected at night. The easy way to get the haplochromines in the pelagic zone was to follow Limnothrissa miodon fishermen at night when they are allowed to operate. At night, nets were set to a maximum depth of 80 m in the pelagic zone once a month. The night samples were also collected at the following times: 21 00-22 00 hr; 23 00-01 00 hr; and 02 00-05 00 hr.
Haplochromis insidiae and H. kamiranzovu contribute 0.01% and 16.62% of total fish catches in littoral zone of Lake Kivu, respectively (Kaningini et al., 1999). No catch statistics have been reported for H. astatodon, which is known only from a specific littoral habitat. We did not catch any H. astatodon specimens in our 3 months of fish sampling campaigns, and we thus borrowed specimens from Institut The fish were identified under a binocular microscope (at the magnification of 10 × 4) based on quantitative and qualitative characteristics described by Snoeks (1994), which are overall tooth shape, body shape, caudal peduncle length, and number of gill rakes.
Fish were given an identification number (ID) indicating sex and sampling location.
Immediately after identification of the fish at the capture site, formalin solution (5%) was injected into the fish esophagus. Thereafter, F I G U R E 1 Haplochromines species studied. (The above are the photographs of the organisms studied) the whole fish was submerged in formalin solution (5%). Fish remained in this solution for 2 weeks in the Ichthyology Laboratory of the Bio-Technology Complex at the University of Rwanda. After 2 weeks, fishes were washed overnight with running water, then transferred into alcohol of 30%, 60% then in 70% as recommended by Wanink, Witte, and Kishe-Machumu (2008). After that, they were cleared and stained according to Potthoff (1984); Taylor & Van Dyke protocol (1985). The Alcian blue gives the cartilage a blue hue and Alizarin Red S, a red dye, acts on the bone. Clearing and staining was done in 10 consecutive steps: (a) dissection and removal of skin,   Teeth were labeled with the same ID as the fish from which they were extracted.
The teeth from the collection of the IRST in collaboration with RMCA were also included and are referred to as IRST group and taken as part of our samples. The IRST group had a similar tooth shape with the recently collected specimens despite the time factor of their sampling period. The purpose was to increase the teeth sample size. The H. astatodon, H. insidiae, and H. kamiranzovu specimens of the IRST group were collected during different sampling campaigns in 1984 and 1987 in the littoral zone of Lake Kivu (Snoeks, 1994). The sex ratio of the specimens from our fish sample and the IRST group are reported in Table 1.
There were 48 bicuspid teeth extracted from females and 35 from males. Fish specimens used in this study were stored individually in 70% ethanol in the Bio-Technology Complex of University of Rwanda.

| Dissection and image acquisition
A maximum number of undamaged teeth were carefully extracted at the median part of the premaxillary bone using a fine pin. Teeth were selected from this median part as earlier observations during species identification of the specimens showed that this median part, in most of the cases, bore undamaged teeth. Another reason was that at this part of the premaxillary, the teeth were found to display a consistent shape trait that facilitated species identification ( Figure 3). Unfortunately, their size and positions on the premaxillary were not consistent. The tooth position trait analysis was not considered for haplochromine differentiation. That is why, even its variance within individuals of each species was not considered for the analysis. However, we performed the crown size measurements and the values obtained were normalized to the standard length of the fish. worried that this few sample size could affect the statistical power of the study. However, it was adequate because the sample size represents a species-specific morphospace.
The number of teeth examined differed among samples for shape analysis. This depended on the availability of the undamaged teeth on the premaxillary bone of the haplochromine species. These teeth were also used to facilitate the work flow in performing the outline based shape analysis in the SHAPE software. We realized that working with many teeth impeded the SHAPE software to run, and we had to limit the sample size for this analysis which represents a limitation in this study. We ensured that the teeth selected for analysis were distributed across the lake fish sampling locations.
As teeth were curved along their long axis (Figure 3), the tooth crown was cut at its base on both sides where the enamel-dentine junction ( Figure 4) reaches the tooth outline (see points 1, 3, and 5 in Figure 5) under the stereoscopic microscope using a fine surgical blade. This allowed the teeth to be positioned consistently flat on the surface for standardized imaging. However, before the pictures were taken, each tooth crown cusp was immobilized between a glass and cover slip, while submerged in dis- required to describe the teeth with a relative high degree of precision and closer to the original digitized outline (Crampton, 1995).
We did not use the sliding semilandmarks based geometric morphometrics (SSBG) because it can lead to serious statistical and visualization artifacts (Gunz & Mitteroecker, 2013).

| Image analysis
Tooth contours were traced using the Chain Coder module of the SHAPE software package ver. 1.3 (Iwata & Ukai, 2002). This involved semi-automated image processing (converting to black and white images) and contour tracing (describing the outer contour as a chain code). The shape descriptors were obtained after SHAPE models of the tooth specimens were superimposed and scaled to unit size based on the first harmonic (by keeping three of the four descriptors of this harmonic constant). The obtained shape descriptors (elliptic Fourier coefficients-EFCs) were then used for further statistical analyses (Kuhl & Giardina, 1982).
Shape variation across and within species (between sexes, habitats, and regions) was then analyzed by performing a PCA on the variance-covariance matrix of the normalized EFCs. Outlines reflecting the variation explained by the individual PC's were generated using the PrinComp module of the SHAPE package.
As the test for normality showed that tooth shape variation was not normally distributed (p < .05), the data were subjected to a pair-wise comparison of a nonparametric MANOVA (npMANOVA) using PAST (Hammer, Harper, & Ryan, 2001) with the Hotelling Bonferroni correction.
To examine species-specific differences, only bicuspid teeth of the three species were pooled together in both analyses (npMANOVA  (Table 1). However, the shape trait of teeth was consistent in each haplochromine species. This facilitated their identification.
For this reason, tooth shape was the only selected trait for their identification with a purely visual tooth inspection using the taxonomic description available. The major assumption in this analysis was that the position of the tooth on the premaxillary bone did not affect the tooth morphology. Using quantitative methods like shape analysis could confirm objectively the species identification. With regard to the size of the crown, we took into account this aspect using traditional measurement of height and width of the major and minor cusps. The variances within and among haplochromine species were separated in statistical analyses (considering males and females individuals within a species and among haplochromine species studied).
Only bicuspid teeth were used to test whether teeth vary between males and females within each species, as indicated in Table 2. The normality test showed that tooth shape variation was not normally distributed among males and females (p < .05). Thus, a one-way npMANOVA on teeth specimens was performed for sex.
To check whether tooth shape of the studied species was related to the habitat (as a proxy for ecological niche), we compared teeth from the littoral versus pelagic specimens for H. kamiranzovu and H. insidiae. Haplochromis astatodon was excluded as they were not found in the pelagic zone (Table 3).
Furthermore, for the analysis of geographical variation, teeth of H. kamiranzovu and H. insidiae were subjected to a south versus north pair-wise comparison in a one-way npMANOVA within species across habitats (Table 4). All the above pair-wise comparisons were subjected to Bonferroni corrections.

| Measurements
Traditional length measurements on teeth crown were used to provide dental metric variations in tooth shape/ tooth size among three studied haplochromine species. The following four measurements were taken on the digital images using ImageJ, ver. 1.44 (Collins, 2007, Figure 5): height of major and minor cusps (measured from point 7 to 2 and from 6 to 4, respectively) and the width of the major and minor cusps (measured from point 1 to 3 and from 3 to 5, respectively). This approach was adopted from premaxillary tooth and crown length and width measurements of cichlids (Casciotta & Arratia, 1993;Dieleman et al., 2015;Wautier et al., 2002).
The height against width measurements on the major cusp of littoral specimens of the three haplochromines species was tested using the nonparametric Kruskal-Wallis rank test to discern differences among the species.
The heights and widths of the major cusp in H. insidiae and H. kamiranzovu species were tested for normality. They showed that they were not normally distributed (p < .05); pair-wise comparisons between all the groups male versus female, littoral versus pelagic and south versus north of Lake Kivu were then performed for each of the major cusp measurements using a one-way npMANOVA. The pairwise comparisons were also subjected to Bonferroni corrections.
A post hoc test with one-way ANOVA was done on data from teeth measurements to see which pair-wise comparisons showed significant differences (Table 5).

| External teeth morphology
The teeth of H. insidiae were all bicuspid (Figure 6a,b). The most distinctive feature was an oblique distal edge (cutting edge formed by the enamel) of the major cusp. In H. astatodon, both bicuspid and TA B L E 3 Teeth sample size used in pair-wise comparisons of tooth shape between the fish from the littoral and pelagic zone unicuspid teeth were present (Figure 6c,d). The major cusp of H. astatodon was tilted to the left with respect to the crown. Its distal cutting edge was longer than in H. insidiae in absolute and relative measurements. In H. kamiranzovu, bicuspid and tricuspid teeth were present ( Figures 5, 6e,f). In the tricuspid teeth, there was a supplementary minor cusp at the posterior edge of the major cusp (Figure 6f).

| Tooth shape in morphospace
From principal component analysis, we found significant differences in bicuspid teeth height and width of major cusps, as well as in tooth shape among H. kamiranzovu, H. insidiae, and H. astatodon (p < .05, Figure 7a). The results of Wilks' Lambda test and Pillai trace test were 0.06 and 1.36, respectively. The degrees of freedom and F values of the above tests were Df1 = 10; Df2 = 146; and the F = 44.99 and Df1 = 10; Df2 = 148; and F = 32, respectively. This suggests that the majority of the differences between the three species studied is due to differences in height and width of major cusps ( Figure 7b and Table 6).
The SHAPE Software package ver.1.3 performed well using PC2 to visualize tooth shape of the three species. Unfortunately, it failed only to compute the teeth size of H. kamiranzovu in PC1 and displayed a very weird size (not a shape) of the species. This might be due to too much intraspecific variability observed in H. kamiranzovu teeth.
The heights and widths of the major and minor cusps of different species are compared in Table 6. The three studied haplochromines species are classified differently according to their teeth shape as described above. Based on PCA results and the above results, the following diagnostic features are suggested for tooth shape differentiation of the three species: the height of the major cusp is larger in H. astatodon than that of H. insidiae, and they are larger than that of H. kamiranzovu. The distal enamel cutting edge of the major cusp of H. astatodon is longer than the one of H. insidiae. In the latter species, it is shorter than the one of H. kamiranzovu. The widths of H. astatodon and that of H. insidiae are higher than that of H. kamiranzovu.
The height of the minor cusp in H. kamiranzovu is larger than that of H. insidiae. In the latter species, it is larger than that of H. astatodon.
Plotting the major cusp height against the cusp width highlighted the relationship between teeth length differences and haplochromine size as species-specific (Figure 8).

| Sexual shape linked dentition
We uncovered significant sexual differences in tooth shape in H. insidiae only (p < .05), as well as in height and width of major cusps. The males had longer and wider cusps than females (Table 7).

| Environmental linked dentition
Our data showed that tooth height and widths of major cusps, as well as tooth shape in H. insidiae is also significant between the northern and southern regions of the lake (p < .05), (Table 8). The height and width of the major cusp in haplochromine female were consistently higher in the south than in north of the lake, and the same trend was observed in height and width of the minor cusp for haplochromine female being higher in the north than in south. The opposite pattern was found in male of H. insidiae, meaning that the height and width of the major and minor cusp were higher in the north than in the south of the lake.
In contrast, no significant differences were found in tooth height and width of major cusps, as well as in tooth shape in H. insidiae and H. kamiranzovu between the littoral and the pelagic zones of Lake Kivu (p > .05,   Dieleman et al. (2015) in

| D ISCUSS I ON
Oreochromis hunteri and O. korogwe showed significant differences in tooth shape between species. Referring to the Lake Kivu haplochromines taxonomic study (Snoeks, 1994), it is evident that these species can also be distinguished based on other traits such as mor- Van Bocxlaer & Schultheiß, 2010). Thus, tooth shape is a reliable trait to identify haplochromines species in Lake Kivu as suggested by previous observations (Wautier et al., 2002), as it is controlled genetically Kocher, 2004;van Rijssel et al., 2015).
In this study, H. kamiranzovu and H. insidiae were collected in the same foraging areas of Lake Kivu. Their tooth shape differences are proposed to be a consequence of resource partitioning as suggested by others for cichlid fish (Chinsembu, 2009;Golcher-Benavides & Wagner, 2019a, 2019bRuber et al., 1999). It may be considered as an indicator of different species feeding specialization (Hulsey et al., 2006;Ruber et al., 1999) or a behavior and food types they consume (Kocher, 2004 indicating that they are efficient algal scrapers (Fraser, Bloomquist, & Streelman, 2008). In another study of algae feeding species in Lake Tanganyika, the shape of the jaw teeth differed according to the resources they exploit (Takashi & Koblmüller, 2011). Species of the genus Petrochromis were observed to have tricuspid teeth in high density that make a brash-like structure to comb unicellular algae from filamentous algae on rocks, while species of Tropheus were observed to have large bicuspid teeth in the most anterior row that allowed ripping filamentous algae from rocks (Yamaoka, Hori, & Kuratani, 1986). It is commonly known that cichlids tooth shape often resembles their feeding behavior and diet. Thus, tooth shape differences enhance the usefulness of the teeth as a taxonomic tool to distinguish these species (Hulme-Beaman et al., 2019).
Diagnostic value of tooth shape in the identification of Lake Kivu specimens was also determined by the results of species-specific differences in tooth shape. These results agree with the findings of Stauffer, Bowers, Kellogg, and McKaye (1997) who reported that tooth shape was one of the morphological characters used to define the genus Metriaclima and to discern between the indigenous Oreochromis hunteri and the recently introduced Oreochromis korogwe in Crater Lake Chala in East Africa (Dieleman et al., 2015).
In H. kamiranzovu, two bicuspids teeth groups may suggest an existence of two morphotypes within H. kamiranzovu. Our results indicated an interspecific tooth shape variation in the three studied Lake Kivu haplochromines as earlier studies showed that tooth shape can vary among species and characterized diverging teeth morphs (Tichy & Seegers, 1999). This result suggests that tooth shape as well as other traits analyzed by Elliptic Fourier is a good predictor to discern species differences at interspecific level (Alvarez & Pérez, 2016;Chuang & Bonhomme, 2019;Oliveira, Panthee, & Silva, 2010).
The PCA reconstructed outlines from the Elliptic Fourier Descriptors (EFDs) showed that the height of the major cusp was the key component in distinguishing the three studied species. As reported by Carlo, Barbeitos, and Lasker (2011), it may be possible to develop diagnostic feature based on the reconstructed images from the EFDs. It is evident that the use of Fourier coefficients yielded a number of characters that could never have been achieved by a purely qualitative description of form as observed in previous study (Carlo et al., 2011;Kenyhercz, Klales, & Kenyhercz, 2013). Tooth shape showed that bicuspid teeth almost differentiated the three species. This result also confirms that the tooth shape is species-specific trait (Clauss, 2019;Mclean, Helgen, Goodwin, & Cook, 2018;Talebi, Sheidai, & Ariyanejad, 2019;Wautier et al., 2002).
Regarding sexual differences, several studies have been addressed in vertebrates focusing only on tooth size difference (Blanckenhorn, 2005;Garcia & Zuanon, 2019). Few authors have studied sexual differences in tooth shape in general, particularly in cichlids. Sexual difference in tooth shape at intraspecific level existed only in H. insidiae among the three studied species. A similar sexual difference in outer tooth shape has been reported for H. flaviijosephi (a species from the Jordan Lake), in which adult males had conical and females had bicuspid teeth (Spataru & Gophen, 1985). Similar differences were observed in H. megalops in Lake Mwanza Gulf (Witte & Witte-Maas, 1987). In Lake Kivu, sexual dimorphism in tooth shape was reported previously only in H. gracilior (Boulenger 1914) andin H. graueri Boulenger (1914), which were among twelve of the fifteen described haplochromines by Snoeks (1994). For both H. gracilior and H. graueri, it was observed that the occurrence of unicuspid teeth was more abundant in males than in females (Snoeks, 1994). The same author did not observe any sexual dimorphism related to the tooth shape in H. kamiranzovu and H. astatodon (Snoeks, 1994), and this agrees with our results. No information regarding tooth shape dimorphism in H. insidiae was provided by the same author. However, Snoeks (1994) reported sexual dimorphism related to tooth number in the three Lake Kivu haplochromines. The meristic data showing sexual differences in tooth number in fifteen Lake Kivu haplochromines were reported in the published work on taxonomy of Lake Kivu cichlids (Snoeks, 1994).
Tooth shape in H. insidiae was associated with geographic differences and not habitat differences. The results indicate that tooth shape variation of H. insidiae is influenced by geographic distance suggesting that gene flow may be limited from south to north of the lake or vice versa. Similar results in an Arctic fox, Vulpes (Alopex) lagopus showed that tooth shape was significantly correlated with geographic distance (Daitch & Guralnick, 2007). Tooth shape similarity of H. kamiranzovu and H. insidiae in pelagic zone suggested that food items consumed by these haplochromines species were the same. Similar diet induces similar tooth shape, and this has been reported in other vertebrates (Christensen, 2014). Tooth shape differences observed in H. insidiae from littoral south versus littoral north might be linked to its gender differences between male and female individuals. It is known that during spawning, teeth of males may become larger than the ones of females as tools used in territoriality to protect the nests (Auld, Noakes, & Banks, 2019;Johnson et al., 2006). Differences of dental morphology in cichlids have usually been correlated with dietary and behavioral differences (Poll, 1952;Yamaoka et al., 1986) but not with geographic distance. For example, teeth of Eretmodus are spatula-shaped with a slender neck region, those of Spathodus are cylindrical with a flattened and truncated crown, and those of Tanganicodus are slender and pointed (Ruber et al., 1999). They range from widely spaced and sharply pointed unicuspids in zooplanktivorous and insectivorous species to closely packed tricuspids in algal scrapers. Their tooth shape differences with their associated mtDNA lineages were sampled and Note: The males of H. insidiae had a larger tooth height and a larger width of the major cusp. Tooth shape of males was also significantly different with more sharp teeth for males than for females. The south of H. insidiae had a longer tooth height and a larger width of the major cusp. Their shape was also significantly different with more sharp teeth for south than for north.

TA B L E 7 (Continued)
TA B L E 8 Size sexual dimorphism in Haplochromis kamiranzovu littoral versus pelagic, south versus north measurements of the major and minor cusp of the height and width as well as their ratios in Lake Kivu, tooth shape may be adapted to the regional differences of the Lake Kivu (south vs. north) as food for fish proved to be different between southern and northern parts of the Lake Kivu in terms of food availability of diatoms and cyanobacteria . Apparently, one might think that population of H. insidiae in the south were isolated by physical barriers to gene flow. The only physical barrier existing in Lake Kivu could be the deeply cut vegetative bays found along the shore line from south to the north. Further research will enable an understanding of the genetic control of tooth shape differences between southern and northern populations of H. insidiae. While tooth shape in H. kamiranzovu was neither linked to habitat nor geographic differences, it suggests that selective pressure on its tooth shape might be minimal when diet was not significantly different (Daitch, 2008). Tooth shape variation was also observed at interspecific level. This suggests that variation in size and shape of teeth among these haplochromine species may be due to their differences in foraging strategies. This finding is similar to that of Kocher (2004) who observed that tooth morphology was highly correlated with the feeding ecology of cichlid fishes.
Size of the major cusp in the three species was observed to be species-specific. These results suggest that cusp morphology as well as cusp number, topology, and orientation are species-specific traits as was previously noticed in different vertebrates (Matalova, Tucker, & Sharpe, 2004). Size of the major cusp in the three studied species was found to be dimorphic. Male individual fish were found to have larger major cusp than female fish. These results are similar to the ones found within each of the breeding habitats of the Eretmodini tribe where male individuals had longer teeth than females (Wautier et al., 2002). Species-specific size of the major cusp of the three species and their ratios showed taxonomic significance of morphological difference that supports the idea of Wood (1981). He observed that without any prior assumptions about taxonomic groups, the analysis of first cusp mandibular molars measurements, has demonstrated that the major axis of variation separates the pooled sample into morphological subgroups.
Crown height and width differences observed between littoral versus pelagic as well as southern versus northern populations of Lake Kivu support the idea that tooth size trait is an adaptive response to environmental changes or gradients Therry et al., 2019;van Rijssel et al., 2015). Environmental gradients from north to south in the lake probably involve the presence of methane gas and the current extraction of the gas in the northern basin (Boehrer, Tümpling, Mugisha, Rogemont, & Umutoni, 2019) while in south, there is no methane extraction activities.
The observed tooth size changes suggest the existence of phenotypic difference between littoral versus pelagic, and southern versus northern populations as food texture in those habitats might be different that may contribute to the differences in crown cusp size. For instance, Astatoreochromis alluaudi that feed on soft and hard foods showed plasticity in phenotype of the teeth of the pharyngeal bone (Huysseune, 1995). The height of the major cusp in the both species H. kamiranzovu and H. insidiae, in the both regions (south and north) was found to be higher in littoral than in pelagic zone. Similar results were observed in Haplochromis argens sp. caught in sand, mud, and rocky habitats of the Lake Victoria, Uganda. The outer-row teeth were found to be relatively long and slender in those habitats (De Zeeuw et al., 2013).
These shape and size teeth phenotypes suggest species differences in foraging behavior based on available food resources in the lake habitats (Ahirwal, Abidi, Kumar, Singh, & Bavithra, 2019;Golcher-Benavides & Wagner, 2019a, 2019b. The results of this study suggest resource partitioning between the studied haplochromine species as was observed in other cichlids (Ahirwal et al., 2019). The species demarcation was noticed at the longest teeth.
This result is similar to findings of Dieleman et al., (2015) who demonstrated that Oreochromis subspecies differences were most pronounced in the longest teeth.
The patterns identified in this study give insight into evolutionary pathway of these haplochromines species and suggest that they may reflect a genetic signal (Torres et al., 2013;Wolpoff, 1971). The EFA results showed that the three species can be distinguished quantitatively based on tooth shape. All taxonomic features mentioned above should be considered to differentiate them. Further revision on Lake Kivu haplochromines taxonomy could consider the tooth shape analysis using the EFA approach to distinguish species. Note: The males of H. kamiranzovu had a larger tooth height and a larger width of the major cusp. Tooth shape of males was also significantly different with more pointy teeth for males than for females. The south of H. kamiranzovu had a longer tooth height and a larger width of the major cusp. Their shape was also significantly different with more pointy teeth for south than for north.

TA B L E 8 (Continued)
The current study is the first to address the question of tooth shape analysis using EFA on the three Lake Kivu haplochromines species. The crown size analysis revealed phenotypic variation of the studied haplochromines. Many questions remain to be answered, particularly sexual dimorphism related to each habitat. Further research should also focus on the genetic basis of the observed patterns of dental size as it relates to sexual dimorphism and should be associated with the habitat.

ACK N OWLED G M ENTS
The sampling campaigns in different habitats and regions of Lake Kivu were generously facilitated by the Monitoring of Lake Kivu. My gratitude is addressed to Africa Centre of Excellence in Aquaculture and Fisheries at Lilongwe University of Agriculture and Natural Resources for a partial scholarship that enabled this work to be conducted. Thanks to Ngely Susan, Charles Muhayimana, and Dr.
Christine Gasingirwa for several administrative assistances during this research.

CO N FLI C T O F I NTE R E S T
The authors have no conflict of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
Sampling locations and fish morphological data will be archived in dryad https://doi.org/10.5521/dryad.12311.