Charles Kimmel

Professor Emeritus (Active), Department of Biology
Member, ION

Ph.D.,  John Hopkins
B.A., Swarthmore 
Lab Website


Research Interests: Morphogenesis and evolutionary developmental biology of the skull

Overview: Our lab focuses on the cellular and genetic mechanisms that control the shaping of skull cartilages and bones during development, and on how development changes during evolution to produce the marvelous diversity that we observe the skulls in different kinds of animals. We use teleost fish for these studies. We take advantage of zebrafish as a model system for development and genetics, of stickleback as a model for evolutionary biology and evo devo, and of teleosts more generally, the most species rich of all vertebrate groups, for their outstanding phenotypic variation.

Cellular and genetic basis of skeletal shaping. 

A general hypothesis that guides much of our work is that the key cellular determinant of skeletal shaping is the patterning of the local spatial arrangements of skeleton forming cells. Our work on development of the zebrafish opercle, a bone that provides support for the gill cover, shows aspects of this control remarkably clearly (Fig. 1). The bone lineage arises in a specific preskeletal condensation, under control of upstream extracellular signals including Endothelin1, which we have studied for many years [1,2]. 

Fig. 1. The zebrafish opercle (Alizarin red) acquires a new form as it develops by reorganizing osteoblast (green, sp7:eGFP) recruitment [3]. Live imaging from 3 to 5 days post fertilization. From our web-based developmental Atlas FishFace. 


Structured phenotypic variation in evoluationary change. 

We see outstanding diversity of the shape of this very same (homologous) bone among teleosts. For example in many neoteleosts the opercle takes on a protective function, developing bony spines in a variety of configurations (Fig. 2).


Fig. 2. Varying arrangements of protective opercle spines.
A. hake, B. rockfish, C. paradox fish, D. weak fish.

We hypothesize that developmental modularity, which we discovered to underlie patterning of the bone’s shape, facilitates shape evolution. In this scenario, the development can evolve to produce a seemingly inexhaustible array of bone shapes because each local region of a bone (that we consider to be a developmental module) can be modified nearly independently of the others. Subsequently, osteoblasts differentiate at precisely regulated positions within the condensation, first appearing in an elongating linear row to produce a bony strut (A). Then, at the growing end of this row, osteoblast addition abruptly switches to a new pattern to locally broaden the region of bone outgrowth at this region (B), resulting in a fan-shaped(C). Other similar reorganizations occur subsequently [3]. These first two phases are under separate genetic control: Mutation of the transcription-factor encoding gene mef2ca specifically disrupts the first [4]. Mutation of Indian hedgehog_a gene, encoding a local extracellular protein signal, disrupts the second [5]. This temporal and spatial specificity in morphogenesis and its control reveals modularity, meaning that separate local regions of the bone develop largely autonomously from one another. Modularity is an important concept in our understanding of developmental patterning, and particularly, of how development evolves [3-7]. Such dissociability in phenotype is a hallmark of modularity. We can use stickleback to test this hypothesis. The opercle and associated facial bones all evolve new shapes as stickleback evolve from ocean-dwelling to freshwater forms [6,7]. Dorsally the bone expands and ventrally it contracts. We predicted that the dorsal and ventral regions, showing separate kinds of evolutionary modification, are separate developmental modules. In the example shown, our hypothetical boundary between the modules, shown by the thick line crossing the two bones examined (Fig. 3A), is supported by multivariate statistical analysis (Fig. 3B; see citations [7-9]). These results are important because they indicate that the way development structures phenotype into modules can impact morphological evolutionary change. The findings provide an answer, at least for this system, to long-standing central question in the field of evo devo: Can development bias evolution? Our evidence suggests that, indeed, it can. 

Fig. 3. Evolutionary divergence in stickleback predicts a module boundary crossing the opercle & subopercle (A). Partial least squares analysis reveals that the covariance between these hypothetical modules is among the lowest of all possible subdivisions of the two bone configuration (B, arrow), supporting the hypothesis. This is because separate models, by definition, are semiautonomous, and hence show only low covariance between them [8].

'Unstructured' phenotypic variation in development and disease. 

Phenotypic variation in evolving natural populations (as just described) clearly is highly structured by Darwinian selection, and likely by developmental modularity as well. In contrast, there are plenty of examples of rather ‘unstructured’ variation showing up nonadaptively in genetic mutants. This variation is thought to be due to loss of a process of buffering of stochastic noise associated with developmental processes, and understanding the buffering is important because its loss occurs in many kinds of human diseases. In ongoing collaborative studies with John Postlethwait’s lab we are studying mutations in two zebrafish genes that show ‘noisy’ variation in their phenotypes, to try to learn more about the mechanism of phenotypic buffering. A first step is to find out where in the developmental process the variation seems to arise. One of these zebrafish genes is fras1, encoding an extracellular matrix protein that is a critical component of a signaling/adhesion complex. Mutations in the human FRAS1 ortholog underlie Fraser’s syndrome, a rare but serious disorder that also shows a good deal of phenotypic noise. Our time-lapse studies in zebrafish reveal that the very early stages of cartilage morphogenesis, when the cartilage primordia outgrow in association with pharyngeal epithelia, appear variably disrupted [10]. We hypothesize that the Fras1 protein complex is an essential component of morphogenesis, mediating mesenchymal-epithelial interaction, as we are testing with conditional mutant alleles. The second gene, already mentioned above, is mef2ca. The level of noisy phenotypic variation in mef2ca mutants is remarkable (Fig. 4) [4]. We carried out a systematic study to see when we could first detect the variation, and pinpointed this time to the earliest stage of morphogenesis (i.e. corresponding to the stage shown in Fig. 1A, above). It may be more than coincidence that both fras1 and mef2ca appear to be required for buffering morphogenesis specifically (cartilage on one hand and bone on the other). Morphogenesis may be a particularly intricate process, especially prone to disturbance by stochastic noise.

Fig 4. Loss of developmental buffering in mef2ca mutants produces outstanding phenotypic variation, here shown in eight individual mutant larvae. The first example shows an essentially wild-type morphology of both the opercle (Op) and branchiostegal ray (BR). Live imaging, Alizarin red staining.