Hedgehog signaling mediates adaptive variation in a dynamic functional system in the cichlid feeding apparatus

Significance

Although evolutionary developmental biology has traditionally focused on linking genes to simple shifts in morphology, adaptation to new environments involves divergence in resource use, which often involves variation in complex functional systems. Thus, for evo-devo research to explain broad evolutionary processes, we must examine more dynamic traits within this context. Here we combine traditional quantitative trait locus mapping, population genetics, and experimental embryology to show how differences in gene expression during larval development lead to adaptive morphological variation in adults. By functional modeling of the skull, we show further how such anatomical shifts in response to genetic variation predict discrete differences in feeding mechanics in African cichlids.

Abstract

Adaptive variation in the craniofacial skeleton is a key component of resource specialization and habitat divergence in vertebrates, but the proximate genetic mechanisms that underlie complex patterns of craniofacial variation are largely unknown. Here we demonstrate that the Hedgehog (Hh) signaling pathway mediates widespread variation across a complex functional system that affects the kinematics of lower jaw depression—the opercular four-bar linkage apparatus—among Lake Malawi cichlids. By using a combined quantitative trait locus mapping and population genetics approach, we show that allelic variation in the Hh receptor, ptch1, affects the development of distinct bony elements in the head that represent two of three movable links in this functional system. The evolutionarily derived allele is found in species that feed from the water column, and is associated with shifts in anatomy that translate to a four-bar system capable of faster jaw rotation. Alternatively, the ancestral allele is found in species that feed on attached algae, and is associated with the development of a four-bar system that predicts slower jaw movement. Experimental manipulation of the Hh pathway during cichlid development recapitulates functionally salient natural variation in craniofacial geometry. In all, these results significantly extend our understanding of the mechanisms that fine-tune the craniofacial skeletal complex during adaptation to new foraging niches.

The diversification of craniofacial morphology has played a key role during vertebrate evolution. Adaptive variation in craniofacial structure facilitates specialization to different food sources and habitats, which in turn contributes to niche partitioning and speciation. In fact, most of the morphological and functional divergence between vertebrates can be found in the craniofacial region (1), which reflects adaptations to a wide variety of environments. It is therefore not surprising that myriad studies have investigated patterns of craniofacial divergence in various animals, including dogs (2, 3), bats (4), birds (5–7), and cichlids (8–10). Although recent efforts have started to associate variation in trophic morphology with differences in gene expression (5, 6, 10–13), the causative loci that underlie these differences remain largely unknown, especially in terms of complex and functionally relevant patterns of craniofacial divergence.

The explosive radiation of East African cichlids has produced a large degree of variation in craniofacial morphology (14, 15), providing an excellent system to study the genetic and developmental mechanisms that promote such diversity. In Lake Malawi cichlids, we have previously associated relatively simple shifts in trophic morphology with genetic variation in the receptor of the Hedgehog (Hh) signaling pathway, patched 1 (ptch1) (10). Specifically, we found two functionally distinct ptch1 alleles that produce quantitatively different amounts of transcript during early development (10) and affect the development of a specific bony process on the mandible, the retroarticular process (RA). The ancestral ptch1 allele, as inferred via the evolutionary history of the ptch1 locus among Lake Malawi cichlids (10), leads to a higher level of expression that results in more bone deposition and a longer RA. Alternatively, the derived allele is associated with lower levels of expression and the development of a shorter RA. In functional terms, variation in RA length changes the mechanical advantage of jaw opening (MAo), which in turn affects the speed of jaw rotation during opening (16). The predicted differences caused by ptch1 also coincide with variation in feeding strategies. For example, Labeotropheus fuelleborni (LF) populations are fixed for the ancestral ptch1 allele, and this species feeds almost exclusively in benthic habitats by scraping algae from the surface of rocks. To accommodate this task, LF possesses a relatively long RA, which translates to higher MAo. Maylandia zebra (MZ; formerly genus Metriaclima, a junior synonym of Maylandia), on the contrary, are fixed for the derived ptch1 allele, and this species is one of the few members in the rock-dwelling clade of cichlids (i.e., mbuna) that routinely forage in the water column via suction feeding. Consistent with this function, MZ possesses a relatively short RA, which translates to lower MAo.

Although simple mechanical advantage is important for lower jaw depression, this action is mediated by many other functional systems including (but not limited to) head lifting, hyoid depression, and the opercle four-bar linkage chain (17, 18). Here we show that the same ptch1 polymorphism is associated with more widespread variation in craniofacial architecture. Specifically, we show that allelic variation in ptch1 is associated with variation in the shape of the interopercle (IOP) bone, which is oriented just posterior to the RA. Together, these two bones contribute to a complex functional system, the opercular four-bar linkage chain (Fig. 1), which is necessary for proper jaw opening in teleosts (19). The action of this four-bar system is powered by the levator opercula, a muscle that originates on the skull and inserts along the dorsal aspect of the operculum. As it contracts, it rotates the operculum (input link) posterodorsally. Then, through ligamentous connections, the IOP serves as a coupler link that transmits the posterior motion to the RA, which is effectively the output link of this system that directly opens the lower jaw. Variation in the relative length of the RA or IOP is predicted to significantly affect the kinematics of the system (20). Thus, the RA and IOP represent functionally integrated elements in the teleost skull. We show here that they are also integrated at the genetic and developmental levels. Finally, we demonstrate that RA and IOP dimensions covary across multiple Lake Malawi cichlid species. In all, we propose that the Hh signaling pathway has played a critical role in promoting functional divergence in the cichlid feeding apparatus.

Fig. 1.

The opercular four-bar linkage system in LF (A) and MZ (B). (A) LF is an algae scraper, which has a relatively longer RA and shorter IOP that results in slower jaw rotation. (B) MZ is a suction feeder that has a relatively shorter RA and longer IOP, which leads to faster jaw rotation. The black bar represents the fixed link, which extends from the opercle–neurocranium joint posteriorly to the mandible–quadrate joint anteriorly. The red bar is the input link, which extends from the opercle–neurocranium joint dorsally to the posterior most edge of the IOP bone ventrally. The orange bar (labeled IOP) is the coupler link, which extends from the posterior edge of the IOP bone to the insertion of the IOP ligament onto the ventral tip of the RA. The green bar (RA) is the output link, which extends from the mandible–quadrate joint to the ventral tip of the RA. The blue bars represent the length of the lower jaw from the mandible–quadrate joint and the RA. Black circles represent fixed joints, whereas white circles are mobile joints. Arrowheads represent the direction of movement during jaw opening.

Results and Discussion

Differential ptch1 Expression Surrounding the IOP Precedes Differential Bone Development.

To explore potential roles for Hh signaling during cichlid craniofacial bone development, we examined the expression of ptch1, GLI-Kruppel family member 1 (gli1), and two osteogenic markers, collagen 1a1 (col1a1) and collagen 10a1 (col10a1), via in situ hybridization in two cichlid species, LF and MZ (Fig. 2 A–F and Fig. S1). At stage 17–18 [5.5–6 days postfertilization (dpf)], when bone development initiates in the head, we found considerable overlap among these genes. Specifically, ptch1 and gli1 are broadly expressed in the hyoid region of the skull, surrounding the IOP–mandibular ligament (Fig. 2 C and D and Fig. S1 A and B). Within this ligament, col1a1 is expressed in the anterior region (Fig. S1C), whereas col10a1 is expressed in the posterior region (Fig. 2 E and F and Fig. S1D), indicating the onset of IOP bone deposition.

Fig. 2.

Differential expression of ptch1 in the pharyngeal skeleton precedes differential development of the IOP in LF and MZ. (A–F) In situ hybridization results showing gene expression at stage 17–18 (5.5–6 dpf). (A and B) Lateral view of the whole mount. (C–F) Flat-mount preparations of the pharyngeal skeleton. (G–L) Flat mount of cleared and stained (alizarin red and Alcian blue) pharyngeal skeletons. Dashed line outlines the IOP. (M and N) Dissected adult IOP. bsr, branchiostegal rays; ch, ceratohyal; e, eye; iopl, interopercular–mandibular ligament; m, Meckel cartilage; pq, palatoquadrate; ra, retroarticular.

It was previously shown, via quantitative PCR, that ptch1 is differentially expressed in the craniofacial region between LF vs. MZ (10), with LF exhibiting relatively higher levels of expression. This can be visualized in the hyoid region of the pharyngeal skeleton via whole-mount in situ hybridization (Fig. 2 A–D) (10). Immediately after this stage of differential Hh expression, patterns of col10a1 expression also differ between these two species. Compared with MZ, LF has a relatively wider and shorter expression domain of col10a1 in the IOP (Fig. 2 E and F and Table 1), as well as an expanded expression domain at the base of the RA process (Fig. 2 E and F, asterisk). For the IOP, differences in col10a1 gene expression predict differences in IOP shape across multiple stages of larval and juvenile development (Fig. 2 E–L and Table 1). Thus, different IOP shapes observed in adult fish (Fig. 2 M and N) can be traced to differential gene expression at the earliest stages of IOP bone development.

Table 1.

Width/length ratio of the IOP at different stages in LF and MZ

Age	Mean width/length ratio of IOP ± SE		P value
	LF	MZ
6 dpf*	0.152 ± 0.012	0.120 ± 0.005	0.040
9 dpf	0.157 ± 0.016	0.023 ± 0.004	< 0.001
12 dpf	0.243 ± 0.005	0.185 ± 0.008	< 0.001
26 dpf	0.315 ± 0.003	0.265 ± 0.007	< 0.001
Adult	0.405 ± 0.007	0.312 ± 0.007	< 0.001

^*At this stage, mineralization in the IOP cannot be visualized by Alizarin red staining. Dimensions of the IOP were therefore measured via the expression domain of Col10a1.

The IOP is a specialized sesamoid bone that is not present in basal fish groups (21). Shared among halecostome fishes, the IOP provides an novel biomechanical pathway of lower jaw depression, and is thought to be an evolutionary innovation that promotes the versatility of mouth opening mechanisms (22, 23). Here we show that early bone development within this ligament is nested within the expression domain of Hh signaling pathway members, ptch1 and gli1 (Fig. 2 A–F and Fig. S1). We therefore hypothesized that ptch1/Hh signaling might regulate the development of species-specific IOP shapes.

Single Quantitative Trait Locus for IOP Shape Maps to ptch1.

As a first step toward testing this hypothesis, we conducted a quantitative trait locus (QTL) analysis in an F2 population derived from a cross between LF and MZ. We detected a single QTL interval on linkage group (LG) 12 that affected the width/length ratio of the IOP (Fig. S2). The QTL peak was located squarely over the ptch1 locus (i.e., LOD peak of 6.25 at 25 cM, 95% CI, 18–42 cM). Although the CI associated with the QTL peak is large, our previous work has demonstrated, via association mapping among natural cichlid populations, that variation in jaw morphology is associated with a relatively narrow ∼10-kb genomic region located ∼15 kb upstream of the first ptch1 exon (10). As expected, the allelic effects of the IOP QTL on LG12 were consistent with the interspecific variation observed between parental species: inheritance of the ancestral (i.e., LF) ptch1 allele was associated with the development of a relatively wider and shorter IOP, whereas the derived (i.e., MZ) allele was associated with a relatively narrower and longer IOP. The QTL exhibited an additive mode of inheritance (Table S1), and accounted for 17% of the phenotypic variance in the F2 population.

Chemical Manipulation of the Hh Signaling Pathway Recapitulates Natural Interspecific Variation in IOP Shape.

We next tested the prediction that modulation of the Hh pathway will change the shape of the IOP in a way that mimics natural variation in this structure. To accomplish this, we treated LF larvae with cyclopamine at stage 17–18, the same period when LF showed higher levels of ptch1 expression around the IOP than MZ (Fig. 2 A–D) (10). The results of this experiment were consistent with our previous work (Fig. 3): cyclopamine-treated LF larvae developed shorter RAs whereas the length of the mandible remained essentially unaffected (10). We also found that treated LF larvae exhibited IOP width/length ratios that were statistically indistinguishable from those of MZ (Fig. 3). Although the specific cellular mechanism of how ptch1 affects IOP bone development remains to be investigated, these results suggest that IOP shape is influenced by the Hh signaling pathway.

Fig. 3.

Cyclopamine-treated LF larvae recapitulates an MZ-like IOP phenotype. LF larvae were treated with 50 μM cyclopamine (CyA, n = 7) or 0.5% ethanol (EtOH control, n = 7) for 6 h at stage 17 (6 dpf). IOP length and width (black arrows) measured at stage 25 (12 dpf). (A and B) Flat mount of cleared and stained pharyngeal skeletons. (C) Bar plot showing the width/length ratio of IOP was significantly reduced in the cyclopamine-treated group compared with the EtOH control group, but was not significantly different from untreated MZ larvae (n = 8). Larvae treated with ethanol were not distinguishable from untreated siblings (n = 5). ch, ceratohyal; pop, preopercle; pq, palatoquadrate; ra, retroarticular; P values determined by Tukey's honestly significant difference test. (Scale bar, 200 μm.)

Combined with our previous work, we propose that natural variation in IOP shape and RA length between LF and MZ is caused, at least in part, by two alternatively fixed ptch1 alleles. The functional difference between these two alleles results in different transcript levels being produced in an area of the skull where the IOP and RA develop (Fig. 2 A–F and Fig. S1) (10). The LF allele is associated with elevated ptch1 expression, a relatively long RA, and a relatively wider/shorter IOP. The MZ allele is associated with reduced ptch1 expression, a shorter RA, and narrower/longer IOP.

IOP and Jaw Shapes Covary Among Natural Populations of Cichlids.

Our results suggest that a single locus alters two functionally related bones in the cichlid skull. We next set out to assess whether this genetic association is reflected in patterns of variation among natural populations of cichlids. To this end, we measured IOP and RA dimensions in several closely related wild-caught cichlid species from Lake Malawi that exhibit a range of foraging modes. Besides LF and MZ, an additional four species from the Tropheops species complex were included. Whereas LF and MZ represent opposite ends of the biting-suction feeding continuum among mbuna, Tropheops species were chosen that represent various points along this continuum. We show that patterns of variation in the relative length of the RA, measured as the MAo, precisely matches that of the width/length ratio of the IOP in these six species (Fig. 4). Moreover, patterns of covariation are consistent with the frequency of ptch1 alleles across species. In LF, the ancestral allele is fixed, and they show the highest MAo and width/length ratio of the IOP. In MZ, the derived allele is fixed, and they show the lowest MAo and width/length ratio of IOP. Among Tropheops species, the two ptch1 alleles are still segregating, and they show a range of MAo and IOP ratios. Notably, Tropheops species fixed for the derived allele exhibit MAo and IOP phenotypes that match those of MZ, whereas species with higher frequencies of the ancestral allele have phenotypes closer to LF (Fig. 4). These results support the assertion that the Hh signaling pathway contributes to ongoing trophic adaptations in Malawi cichlids (10).

Fig. 4.

Covariation of the width/length ration of the IOP (A) and MAo (B) across wild caught species. LF, 0% MZ (derived) ptch1 allele (n = 20); MZ, 100% MZ ptch1 allele (n = 20); T.gra, Tropheops gracilior, 100% MZ ptch1 allele (n = 10); T.int, Tropheops intermedius, 100% MZ ptch1 allele (n = 10); T.kwazi, Tropheops sp. from chinyankwazi, 59% MZ ptch1 allele (n = 10); T.wezi, Tropheops sp. from chinyamwezi, 27% MZ ptch1 allele (n = 10).

Covariation, or integration, of traits is believed to be a major factor that determines evolvability (24). In particular, coordinated changes in multiple traits can promote patterns of variability that, when aligned with the vector of selection, can result in rapid evolutionary responses (25). As two of three movable links in the opercular four-bar system, the RA and IOP represent functionally integrated elements of the teleost head. Here we show that these elements are also integrated at the evolutionary (i.e., the covariation across species), developmental (i.e., responsiveness to Hh signaling), and genetic (i.e., associated with the same genetic polymorphism) levels. This widespread integration may provide greater insights to the outstanding diversity in cichlid trophic morphology: instead of two independent mutations, these fish can generate morphological changes in two bones that operate in a common function via a single mutation that affects the Hh signaling pathway (e.g., ptch1).

Ptch1 Induced Changes to the RA and IOP Are Predicted to Influence the Mechanics of the Opercular Four-Bar Linkage System.

To investigate the potential biomechanical outcome of Hh-induced shape variation in the IOP and RA, we built digital models and simulated the movement of the opercular four-bar linkage system under different scenarios (Fig. 5) and monitored changes in kinematic transmission ratio (KT) during jaw opening. First, we simulated the model with parameters that depicted the parental phenotypes, and showed that, during jaw opening, KT is higher in the suction-feeding species (MZ) than in the biting species (LF). This is consistent with the general observation that suction-feeding species usually possess four-bar systems that are capable of faster jaw movements, i.e., higher KT (26–28). Next, we manipulated parameters in the LF model (i.e., representing the ancestral condition) to simulate ptch1-mediated phenotypic changes in the IOP and RA toward a more MZ-like condition (i.e., representing the derived condition). Note that the IOP was used as a proxy for the coupler link because it contributes to ∼85% of the length of the coupler link in both species. As expected, we see that, with a progressively longer IOP and progressively shorter RA, KT shifts away from the LF/ancestral model toward the MZ model (Fig. 5).

Fig. 5.

KT during jaw opening in digital models of the opercular four-bar linkage system. x axis shows the rotation of the input link from starting position; y axis shows KT. LF, model that represents the LF morphology; MZ, model that represents the MZ morphology; Ptch1 10%, modified LF model with 10% longer IOP and 10% shorter RA; Ptch1 15%, modified LF model with 15% longer IOP and 15% shorter RA; Ptch1 20%, modified LF model with 20% longer IOP and 20% shorter RA.

Conclusion

One of the key questions in evolutionary studies is how genetic variation translates into ecomorphological adaptation and ultimately fitness (29–31). Here we present empirical evidence that variation at a single locus affects multiple components in a dynamic mechanical system characterized by several distinct moving elements. We show that, in cichlids, ptch1 mediates morphological variation in the IOP and RA. We demonstrate further that the mode of action of this effect can be traced to early stages of bone development—e.g., expression patterns of the osteogenic marker col10a1 differ between LF and MZ. Ultimately, these molecular, cellular, and anatomical differences translate to variation in the mechanical properties of the opercular four-bar linkage model, a complex functional system that is predicted to play important roles in the ecological divergence among closely related teleost species (26–28). In all, this work offers an integrative view on how adaptive radiations can occur at the genetic, developmental, and functional levels.

Materials and Methods

Cichlid Maintenance.

Cichlid species were collected from Lake Malawi and reared in 40-gallon glass aquaria at 28.5 ±1 °C on a 14 h light/10 h dark cycle. All larvae used for this project were F1 or F2 derived from wild-caught stock, and obtained by natural matings. Embryos were extracted from mouth-brooding females between 3 and 4 dpf (stages 10–14), and incubated in 1-L glass beakers with ∼900 mL of system water plus two or three drops of methylene blue at 28.5 ±1 °C. An aeration stone was placed at the bottom of the flask to provide enough air to keep the embryos vigorously swirling at the bottom of the flask. Embryo medium was changed every 2 d. Cichlid staging was according to a previous work (32). Cichlid husbandry follows a protocol approved by the institutional animal care and use committee at the University of Massachusetts.

In Situ Hybridization.

Ptch1 (NCBI-GenBank accession no. JN037690) and col1a1 (NCBI-GenBank accession no. JN116727) riboprobes were made according to a previous work (10). A gli1 riboprobe was similarly made from a cichlid clone (NCBI-GenBank accession no. JN037689). A col10a1 riboprobe was made directly from cichlid cDNA using primers that contained T3 (sense) and T7 (antisense) RNA polymerase binding sequences: Col10a1_T3F1 CATTAACCCTCACTAAAGGGAACAGGAGCACCAGGTAAAAGC; Col10a1_T7R1 TAATACGACTCACTATAGGGAGAAGGACCTGGGAGACCAT. Polymerase recognition sequences are underlined. Whole-mount in situ hybridization was performed as previously described (13) on stage 17-18 embryos. The lower jaws were dissected and flat-mounted before imaging.

QTL Mapping.

We photographed and measured the length and width of the IOP in 114 F2 individuals derived from a cross between LF and MZ as described previously (8, 9). The calculated length/width ratio was used for QTL mapping, which was done in R by using Multiple-QTL Mapping routines described in a previous study (33). Genome-wide significance threshold (α = 0.05) was calculated by permutation tests with 1,000 repeats.

Cyclopamine Treatment.

LF larvae (stage 17) from the same brood were divided into three treatment groups: cyclopamine treatment, ethanol control, and untreated control, with the same amount of larval fish water. In the cyclopamine treatment group, cyclopamine stock solution (10 mM cyclopamine in ethanol) was added to reach a final concentration of 50 μM, based on a previous study (10). The same volume of ethanol or larval fish water was added to the ethanol control and untreated control groups, respectively. Animals were treated for 6 h in the dark at 28.5 °C, and then washed with larval fish water several times before returning to standard culture flasks. At stage 25 (12 dpf), they were euthanized and stained with alizarin red and Alcian blue for bone and cartilage (34), then imaged with a Leica DFC450 C digital microscope camera mounted to a Leica M165 FC microscope. An MZ brood was stained and imaged with the same procedure for comparison. Measurements of bone development were taken from images with ImageJ 1.47.

Digital Modeling of the Opercular Four-Bar Linkage System.

Models were built in GeoGebra (www.geogebra.org/cms/en/). Lateral images of LF and MZ were imported as background to locate joint positions in the default state (i.e., mouth closed). During simulations of mouth opening, the input link (i.e., opercle) was rotated posteriorly until the coupler link (i.e., IOP) and output link (i.e., RA) were aligned, which prevented further rotation of the input link. The rotation of the input link was done stepwise in increments of 0.5 °. At each increment, the corresponding output link rotation was recorded (ΔO), which was measured as the change in the angle between the output link and the fixed link. The KT ratio was then calculated at each increment as the output link rotation (ΔO) divided by the input link rotation (0.5 °). When simulating ptch1-induced phenotypic changes, we increased the length of coupler link (i.e., IOP) and decreased the length of output link (i.e., RA) from the LF model by 10%, 15%, and 20% (e.g., model “ptch1 10%” is an LF model with 10% longer IOP and 10% shorter RA). These three models were chosen because they roughly approximate the magnitude of QTL effects of ptch1 on IOP (17%) and RA (11%) shapes.