Adaptive Diversification of the Lateral Line System during Cichlid Fish Radiation.

The mechanosensory lateral line system is used by fishes to sense hydrodynamic stimuli in their environment. It provides information about flow regimes, proximity to substrate, and the presence and identity of prey and predators and represents a means of receiving communication signals from other fish. Thus we may expect lateral line system structures to be under strong divergent selection during adaptive radiation. Here, we used X-ray micro-computed tomography scans to quantify variation in cranial lateral line canal morphology within the adaptive radiation of Lake Malawi cichlids. We report that cranial lateral line canal morphology is strongly correlated with diet and other aspects of craniofacial morphology, including the shape of oral jaws. These results indicate an adaptive role for the lateral line system in prey detection and suggest that diversification of this system has taken an important role in the spectacular evolution of Lake Malawi's cichlid fish diversity.

Introduction

The lateral line system is an important mechanosensory system used to detect hydrodynamic stimuli in aquatic environments (Webb, 2014, Klein and Bleckmann, 2015). It is found in all fishes, including lampreys, and some amphibians (Gelman et al., 2007, Schlosser, 2012), indicating that it may be a primitive vertebrate character. In teleost fishes, it comprises two key components with separate receptive abilities: the superficial neuromasts and the canal neuromasts (Bleckmann and Zelick, 2009). The superficial neuromasts, present on the surface of the fish, are thought to assess direct current and are used for sensing the direction and speed of water flow (Bleckmann and Zelick, 2009, Wark and Peichel, 2010). Canal neuromasts are situated within fluid-filled canals between skeletal openings (pores) and are thought to be more important for detecting pulses in water movement, such as those associated with the movement of other organisms (Montgomery et al., 1994, Coombs et al., 2001), against background noise (Engelmann et al., 2000) (Figure 1). A widened lateral line canal phenotype, accompanied with increased pore size, is thought to convey increased sensitivity to certain hydrodynamic stimuli (Webb, 2014, Klein and Bleckmann, 2015, Schwalbe and Webb, 2015). Collectively, the transduction of flow stimuli through the superficial and canal neuromasts is thought to inform multiple key behaviors in fishes, including rheotaxis (Montgomery et al., 1997, Kulpa et al., 2015), prey detection (Montgomery and McDonald, 1987, Janssen, 1996, Pohlmann et al., 2004, Schwalbe et al., 2012), predator avoidance (Stewart et al., 2014), shoaling behavior (Faucher et al., 2010), and male-male competition (Butler and Maruska, 2015). Thus we may expect them to have been subject to strong divergent selection during adaptive radiation into multiple ecologically and behaviorally distinct species.

Figure 1

Overview of the Cranial Canal Lateral Line System of the Lake Malawi Cichlid Astatotilapia calliptera

(A and B) (A) Ventral view and (B) lateral view. The positions of canals (blue) and pores (red) are shown. Approximate position of canal neuromasts are shown as asterisks. MC, mandibular canal; PR, preopercular canal; SO, supraorbital canal; IO, infraorbital canal; OT, otic canal; PO, post-otic canal; ST, supratemporal canal; TC, trunk canal. Positioning of canal neuromasts is from staining of A. calliptera with DASPEI, corroborated with evidence from Butler and Maruska (2015) who studied the congeneric Astatotilapia burtoni.

Cichlid fishes are one of the largest and most diverse of all vertebrate families, comprising over 3,000 species (Seehausen, 2006, Salzburger, 2018) and reaching their highest diversity in the Great Rift Valley lakes of East Africa (Seehausen, 2006, Seehausen, 2015). The Lake Malawi haplochromine radiation alone comprises over 500 species that have evolved from common ancestry within the last 1.1 Ma (Malinsky and Salzburger, 2016, Malinsky et al., 2018). These cichlids have diversified extensively in habitat preferences, diet, body shape, craniofacial morphology related to prey capture and processing (Turner, 1996, Albertson et al., 2003, Konings, 2016), breeding behaviors (Konings, 2016, York et al., 2018), and sensory abilities, notably vision (Parry et al., 2005). However, knowledge of their lateral line system diversification is limited to (1) anatomical work that has shown variation among genera in the size and development of cranial canal morphology (Eccles and Trewavas, 1989, Schwalbe et al., 2012, Bird and Webb, 2014, Becker et al., 2016) and (2) behavioral work showing that two Lake Malawi species with differences in cranial canal morphology differ in their ability to locate live prey in dark environments (Schwalbe et al., 2012, Schwalbe and Webb, 2015). Thus we conducted the first large-scale comparative study of lateral line system morphology across the phylogenetic and ecomorphological diversity of an African cichlid radiation and tested for interspecific associations between morphology and ecological variables.

Results and Discussion

Lateral Line Morphology and Head Shape Covary

We focused our analysis on the lateral line canal components of the head and used X-ray micro-computed tomography scans to study lateral and ventral views of cranial morphology, including the size and position of cranial canal pores (Figure S2; Table S2). We quantified cranial lateral line system variation among 52 species of Lake Malawi cichlids using landmark-based geometric morphometrics (Figures 2, S2, and S3; Table S1) and tested for associations with gross morphology and ecology using distance-based redundancy analysis (Legendre and Anderson, 1999). We initially found that variation in cranial lateral line pore morphology was strongly coupled with variation in gross cranial morphology (Figures 3A–3C). For example, larger pores were associated with longer and broader jaw bones (Figures 4A–4C, 5A, and 5B). Previous work has demonstrated considerable modularity of the cichlid skull, including a major preorbital module that encompasses both the upper and lower oral jaws (Parsons et al., 2011). Our findings suggest that the cranial lateral line canal system is an intrinsic part of this more complex preorbital module. Although oral jaws have typically been considered to have a primary role in the handling and processing of prey in cichlid fishes (Turner, 1996, Albertson et al., 2003, Hulsey and García de León, 2005, Muschick et al., 2011, Parsons et al., 2012, Konings, 2016), these results additionally highlight a role for the lateral line canals of the lower jaw in hydrodynamic sensing of prey.

Figure 2

The Diversity of Cranial Lateral Line Canal Morphology Among the Lake Malawi Cichlid Species Included in Our Study

Individuals are grouped into the six major evolutionary lineages (Malinsky et al., 2018) and color coded by dietary preference. Micro-computed tomographic images are not to scale. The positions of pores (red) and approximate position of canals (blue) are shown on the left side of each fish. R, Rhamphochromis; D, Diplotaxodon; U, utaka; SB, shallow benthic; DB, deep benthic; M, mbuna. A full list of species is provided in Table S1 and Figure S1.

Figure 3

The Proportion of Variance in Morphology Explained by Each Morphological or Ecological Explanatory Variable, Calculated through Distance-Based Redundancy Analysis and Analysis of Variance

(A–F, left) Sets of landmarks used as independent variables for our analysis, illustrated on micro-computed tomographic images of Astatotilapia calliptera. Red dots are landmarks, and yellow dots are semi-landmarks. Semi-landmarks slide along blue lines between landmarks. Approximate canal positioning is illustrated in blue. Full explanation of landmarks is provided in Figure S2.

(A–F, right) Proportion of variance in morphology explained by each explanatory variable as calculated through distance-based redundancy analysis. †Variable omitted from model; *p < 0.05; **p < 0.01; ***p < 0.001.

(A) Preopercular lateral line pore morphology.

(B) Orbital (infra- and supraorbital) lateral line pore morphology.

(C) Ventral head lateral line pore morphology.

(D) Gross lateral head cranial morphology.

(E) Gross ventral head cranial morphology.

(F) Lower pharyngeal jaw morphology.

Figure 4

Variation in Cranial Lateral Line Pore Morphology Observed Among Dietary Groupings within the Lake Malawi Haplochromine Radiation

(A–C, left) Arrangements of landmarks used in our analysis illustrated on micro-computed tomographic images of Astatotilapia calliptera. Red dots represent landmarks, and the approximate positioning of the relevant lateral line canals are shown in blue. Full explanation of landmark positioning is provided in Figure S2.

(A–C, right) Variation in morphology of cranial lateral line pores among our 52 study specimens. Variation is shown for each landmark set, calculated using canonical variates analysis, displaying canonical variate (CV) scores on the first two axes. Individuals are grouped by diet, and for CV scores, 95% confidence ellipses are shown for all dietary groupings with n > 2. Changes in landmark arrangement associated with each axis are illustrated on wireframe graphs (Klingenberg, 2013). Wireframe graphs along each axis show consensus landmark arrangement for the 52 species (gray dots and lines) and landmark position shifts associated with each CV axis (black dots and lines). The two wireframe graphs on each axis illustrate changes in landmark positioning associated with the highest and lowest values for each CV on that axis (black dots and lines).

(A) Preopercular lateral line pore morphology.

(B) Orbital (infraorbital and supraorbital) lateral line pore morphology.

(C) Ventral head lateral line pore morphology.

Figure 5

Variation in Gross Cranial Morphology Observed Among Dietary Groupings within the Lake Malawi Haplochromine Radiation

(A–C, left) Arrangements of landmarks used in our analysis illustrated on micro-computed tomographic images of Astatotilapia calliptera. Red dots represent landmarks, and yellow dots represent semi-landmarks. Semi-landmarks were placed at equal distances between landmarks along blue lines. Full explanation of landmark positioning is provided in Figure S2.

(A–C, right) Variation in gross cranial morphology among our 52 study specimens. Variation is shown for each landmark set using canonical variate analysis, displaying canonical variate (CV) scores on the first two axes. Individuals are grouped by diet, and for CV scores, 95% confidence ellipses are shown for all dietary groupings with n > 2. Changes in landmark arrangement associated with each axis are illustrated on wireframe graphs (Klingenberg, 2011). Wireframe graphs along each axis show consensus landmark arrangement for the 52 species (gray dots and lines) and landmark position shifts associated with each CV (black dots and lines). The two wireframe graphs on each axis illustrate changes in landmark positioning associated with the highest and lowest values for each CV on that axis (black dots and lines).

(A) Gross lateral head cranial morphology.

(B) Gross ventral head cranial morphology.

(C) Lower pharyngeal jaw morphology.

Diet Predicts Cranial Lateral Line Morphology

Our results demonstrated dietary grouping to be a significant predictor of both ventral and orbital canal morphologies (Figures 3B and 3C), but not preopercular canal morphology (Figure 3A). This is notable as canals in the ventral and orbital regions can be considered to have ventral- (or anterior-) facing pores (Figures S2A and S2C). Specifically, the ventral view of the head encompassed the ventral-facing mandibular canal and ventral arm of the preopercular canal (Figures 1A and S2A), whereas the orbital lateral line pore morphology we measured encompassed primarily anterior-/ventral-facing pores, including the infraorbital canal within the lacrimal bone (Figures 1B and S2C). In contrast, the pores of the preopercular canal can be considered to be more lateral facing (Figures 1B and S2B). Anterior- or ventral-facing pores have been proposed to be functional during feeding behavior, specifically in relation to the observed “sonar” feeding by wild Aulonocara and other benthic genera, where foraging fish angle their body and probe for cryptic buried prey (Turner, 1996, Schwalbe et al., 2012, Konings, 2016). Our results are therefore supportive of this proposed role for the cranial lateral line in prey detection.

Benthic invertebrate feeders and molluscivores appear to have the largest cranial canal pores (Figures 4A–4C), consistent with widened canals and enlarged pores associating with detection of cryptic prey within the substrate (Turner, 1996, Schwalbe et al., 2012, Konings, 2016). Behavioral trials in light-limited environments have shown that Aulonocara stuartgranti, a species with large pores, is able to forage more effectively on live arthropod prey than Tramitichromis sp., which possesses comparatively smaller pores and narrower canals (Schwalbe and Webb, 2015). Our results emphasize the need to further explore the limits of benthic prey detection associated with the range of lateral line morphologies of these cichlids, ideally focusing on natural prey items and substrates mirroring those present in Lake Malawi. Our results also reveal variation among other trophic groups, with variation in pore size and positioning present in piscivores, zooplanktivores, and herbivores (Figures 4A–4C).

Diet was also a significant predictor of the variation in gross cranial morphology, which covaried extensively with the cranial lateral line canal morphology (Figures 3D–3F). Species in each dietary grouping had a combination of traits specific to those diets. For example, molluscivores tended to have shorter downward-facing oral jaws and more robust pharyngeal jaws, piscivores had more elongated forward-facing oral jaws and narrower pharyngeal jaws, and benthic algivores had short downward-facing oral jaws coupled with relatively robust pharyngeal jaws (Figures 5A–5C). The fin biter and scale eater Genyochromis mento was unique in possessing wide, but short oral jaws (Figures 5A–5C). We note that the elongated jaw phenotype of piscivores and the broad jaw phenotype of algivores (Figures 5A and 5B) were both paired with a relatively small pore size (Figures 4A–4C), whereas the relatively large pores of benthic invertebrate feeders tended to be coupled with widened preopercular bones in particular (Figures 5A and 5B). Our results confirmed our expectations of an association between lower pharyngeal jaw morphology and diet in Malawi cichlids (Figures 5C and 3F), as has been shown in Tanganyika cichlid fishes (Muschick et al., 2011). This is important in the context of this study as it demonstrates that our methods are reliably able to recover functional associations between diet and morphology. Taken together, our results are supportive of natural selection being a major driver of cranial ecomorphological diversification in cichlids, with selection on traits related to the detection of hydrodynamic stimuli produced by prey being an important yet largely overlooked component of this diversification process.

Lateral Line Diversification Is Partially Independent of Phylogenetic Constraint

Recent work has confirmed that phylogeny corresponds closely with previously defined ecomorphological groupings (Genner and Turner, 2012) across the endemic Lake Malawi haplochromines (Malinsky et al., 2018) (Figure S1). We included representatives of all the major lineages known to comprise the radiation, including open water piscivores (Rhamphochromis), deep water predators (Diplotaxodon-Pallidochromis), open water zooplanktivores (utaka), shallow water rock cichlids (mbuna), the typically shallow water sediment-associated cichlids (shallow benthic), and the typically deep water sediment-associated cichlids (deep benthic). For our analyses we placed the shallow water generalist Astatotilapia calliptera within the mbuna grouping, given their recent shared evolutionary history (Malinsky et al., 2018) (Figure S1).

When previous adaptation limits future evolutionary pathways despite the presence of strong selection pressures we may consider an evolutionary lineage to be phylogenetically constrained (McKitrick, 1993). For example, in the context of cichlid fishes, it may be possible for some taxa possessing broad oral jaw bones to develop larger pores as a sensory specialization for feeding on motile benthic prey. Other taxa with narrow oral jaw bones may be unable to follow this evolutionary trajectory. It is notable that no species within the mbuna, Rhamphochromis, or Diplotaxodon lineages have widened lateral line canal phenotypes (Figures 2 and 4). We suggest that constraints conferred by head shape including the thin, laterally compressed piscivore phenotype, and the flat, anteroposteriorly compressed algivore phenotype (Figures 5A and 5B), may both prevent the evolution of broad preopercular and mandibular bones and hence also prevent the evolution of widened canals and pores. Such phylogenetic constraints would manifest in traits associating more strongly with phylogenetic grouping than with ecology. Our results showed that both gross cranial morphology (Figures 3D and 3E and S2D) and preopercular lateral line pore morphology (Figures 3A and S2B) were significantly associated with phylogenetic grouping. However, this was not the case for both the diet-associated ventral and infraorbital cranial canal pore morphologies (Figures 3A, 3B, S2A, and S2C). These results indicate that diet-associated traits are able to diversify somewhat independent of phylogenetic constraint and is further suggestive of the lateral line system being under selection during rapid adaptive radiation.

Associations with Depth and Habitat

Research on the lateral line systems within the Lake Malawi radiation has highlighted a contrast between the “widened” canals or pores of a dark-adapted benthic genus (Aulonocara) and the relatively “narrow” canals and pores of a shallow water genus (Tramitochromis) (Schwalbe et al., 2012). Our analyses show for the first time that depth (and hence light level) does not consistently predict this morphology (Figures 4B and 4C). For example, shallow-living species such as the benthic arthropod feeding Fossorochromis rostratus and the mollusc-specialist Trematocranus placodon also have relatively large mandibular and preopercular canal pores (Figures 2 and 4A–4C). Given that these species are commonplace in clear water soft-sediment littoral habitats, we suggest that a range of benthic invertebrate feeders may use motion cues for location of prey within the sediment, independent of depth and light levels.

Habitat was not a consistent predictor of either gross craniofacial morphology or cranial lateral line canal morphology in our analyses. This may in part be related to the cranial lateral line system being useful for sensing benthic prey multiple habitat types, as indicated by cichlid genera with enlarged cranial lateral line canal pores occupying a range of habitats. For example, Aulonocara are found in both deep water soft-sediment habitats (e.g., Aulonocara sp. “copper”) (Turner, 1996) and in shallow water cave-like habitats (e.g., Aulonocara jacobfreibergi) (Konings, 2016). The decoupling of lateral line morphology and habitat may also reflect the diversity of trophic-specialist taxa present within each of the habitats. For example, a rich diversity of piscivores and zooplanktivores are found in soft-sediment habitats along with benthic invertebrate feeding species (Turner, 1996, Konings, 2016). Given the range of potential diets within each habitat type, an equivalent diversity of corresponding craniofacial morphologies would be predicted to be present within each to enable the detection, capture, and processing of available prey items.

Modularity and the Cranial Lateral Line System

The degree of modularity within a skeletal system is thought to either constrain or facilitate rapid evolutionary divergence (Pugliucci, 2008, Parsons et al., 2011, Parsons et al., 2012, Bird and Webb, 2014). The bones of the preorbital region of the cichlid head, which comprises the oral jaws and supporting structures, form a functional module (Cooper et al., 2010, Parsons et al., 2011). This module has limited axes of variation available during adaptation to specific trophic resources (Parsons et al., 2012), which is perhaps one contributing factor to the striking parallel evolution of craniofacial morphology present among the Lake Malawi, Lake Victoria, and Lake Tanganyika cichlid radiations (Young et al., 2009).

Our study has shown covariance between jaw morphology and aspects of lateral line morphology, such as the positioning of pores, when the Lake Malawi radiation is considered as a whole. However, our study also confirms observations that species with very distinct jaw phenotypes can possess similar narrow canal phenotypes, demonstrating that cranial lateral line morphology can decouple from gross jaw morphology (Bird and Webb, 2014). Detailed comparative analysis of cichlid canal phenotypes has demonstrated that the phenotypic differences observed among large-pored and small-pored species can largely be attributed to rate heterochrony (Bird and Webb, 2014, Webb et al., 2014). Thus future studies concentrating on the evolution of interspecific variation in lateral line morphology should have a strong focus on the relative rate of phenotypic development among taxa.

Concluding Remarks

Our study provides insights into the scale of cranial lateral line system variation across a major vertebrate adaptive radiation. We provide quantitative evidence showing that cranial canal lateral line pore morphology covaries with gross oral jaw morphology, and that additional variation is explained by species ecology, most strikingly by diet (Figures 3A–3C). Collectively these data demonstrate the importance of ecological variables as predictors of both the gross craniofacial morphology and subtle variation in the size and positioning of cranial lateral line canal pores. Our evidence suggests that the cranial lateral line system can adapt readily, like other aspects of trophic morphology, including the lower pharyngeal jaw (Muschick et al., 2012) and oral jaws (Albertson et al., 2003, Albertson and Kocher, 2006, Parsons et al., 2012). We suggest that our findings indicate an important role for the system in facilitating trophic segregation within the wider adaptive radiation context.

Taking an integrated approach will be important for a more robust understanding of the role of the lateral line system in trophic diversification within the radiation. This will include identifying the genetic basis of lateral line system diversity and examining superficial neuromasts alongside canal morphology (Wark and Peichel, 2010, Wark et al., 2012, Becker et al., 2016). To date, research on the genetic basis of lateral line system diversity has focused on superficial neuromasts of the trunk (Becker et al., 2016). Research has identified genes associated with both superficial neuromast morphology and schooling behavior (Wark et al., 2012, Mills et al., 2014, Greenwood et al., 2016). However, identifying the genetic basis for the phenotypic diversity of a system as varied and complex as the cichlid lateral line will be challenging, due to potential epistatic and pleiotropic effects of associated loci (Albertson et al., 2003, Wark et al., 2012, Mills et al., 2014, Greenwood et al., 2016).

In addition, understanding the system's influence on adaptive radiation will require investigating aspects of cichlid behavior and ecology not considered within the scope of this study (Faucher et al., 2010, Stewart et al., 2014, Butler and Maruska, 2015, Butler and Maruska, 2016, York et al., 2018), and evaluating lateral line systems in a multimodal sensory context (Parry et al., 2005, Schwalbe and Webb, 2015). Our findings could extend to other teleost fish radiations, including the cichlid fish radiations of Lake Victoria and Lake Tanganyika. We note that at least one genus in the Lake Tanganyika radiation, Trematocara, contains species with enlarged lateral line pores comparable to those of the Lake Malawi Aulonocara genus (Takahashi, 2002).

Limitations of the Study

Our study focused on interspecific morphological variability in cranial lateral line morphology, yet ontogenetic variation in cichlid lateral line morphology has been shown in some cichlid species (Bird and Webb, 2014, Webb et al., 2014), and it is possible that sexual dimorphism in morphology may be present if this is associated with ecological factors such as diet. We included one specimen per species in this study, which enabled us to capture broader patterns of interspecific variation. However, a more detailed understanding of intraspecific ontogenetic or sex-associated morphological variation is required.

Our study considered phylogenetic relationships from the perspective of membership of monophyletic clades, as resolved through analyses of whole-genome data (Malinsky et al., 2018). However, whole-genome data are currently not available for all 52 species considered in this study, so we assumed clade membership of some species based on other phylogenetic studies, or knowledge of the phylogenetic placement of congeneric species (Figure S1). In practice, it may not be possible to fully account for phylogenetic history during interspecific analyses across the diversity of cichlids in the Lake Malawi radiation: even with whole-genome data it is not possible to generate a single tree that consistently and adequately resolves all species relationships due to hybridization and incomplete lineage sorting (Malinsky et al., 2018). Finally, the lateral line system is thought to be important for several aspects of species behavior and ecology we did not consider here, including shoaling behavior (Faucher et al., 2010), aggressive intraspecific interactions (Butler and Maruska, 2015), and habitat light level (Schwalbe et al., 2012). Further comparative work will help to establish whether these factors have additionally contributed to the remarkable diversity we observed.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.