Epigenetic interactions and the structure of phenotypic variation in thecraniumB. Hallgrı´mssona,,1, D.E. Liebermanb,1,W.Liua, A.F. Ford-Hutchinsonc, and F.R. JirikcaDepartment of Cell Biology and Anatomy, Alberta Bone and Joint Institute, Faculty of Medicine, University of Calgary,3330 Hospital Dr. NW, Calgary, AB, Canada T2N 4N1bDepartments of Anthropology and Organismic and Evolutionary Biology, Harvard University, 11 Divinity Avenue, Cambridge,MA 02138, USAcDepartment of Biochemistry and Molecular Biology, Alberta Bone and Joint Institute, Faculty of Medicine, University ofCalgary, 3330 Hospital Dr. NW, Calgary, AB, Canada T2N 4N1Author for correspondence (email: [email protected])1These authors contributed equally to this work.SUMMAR Y Understanding the developmental and geneticbasis for evolutionarily significant morphological variation incomplex phenotypes such as the mammalian skull is a challengebecause of the sheer complexity of the factors involved. Wehypothesize that even in this complex system, the expression ofphenotypic variation is structured by the interaction of a few keydevelopmenta l processes. To test this hypothesis, we created ahighly variable sample of crania using four mouse mutants andtheir wild-type controls from similar genetic backgrounds withdevelopmental perturbations to particular cranial regions. Usinggeometric morphometric method s we compared patterns of size,shape, and integration in the sample within and between thebasicranium, neurocranium, and face. The results highlightregular and predictable patterns of covariation among regions ofthe skull that presumably reflect the epigenetic influences of thegenetic perturbations in the sample. Covariation betweenrelative widths of adjoining regions is the most dominant factor,but there are other significant axes of covariation such as therelationship between neurocranial size and basicranial flexion.Although there are other sources of variation related todevelopmental perturbations not analyzed in this study, thepatterns of covariation created by the epigenetic interactionsevident in this sample may underlie larger scale evolutionarypatterns in mammalian craniofacial form.INTRODUCTIONIn order to test hypotheses about evolutionary changes inphenotype, it is often necessary to understand their underlyinggenotypic and developmental bases. This is a challenge be-cause complex phenotypes arise from many sequential inter-actions among genes, cells, tissues, organs and theenvironment. Yet, in spite of this complexity, simple modi-fications to developmental pathways may often generate noveladaptive phenotypes by using basic ‘‘toolkit’’ genes that func-tion in diverse developmental contexts (Atchley and Hall1991; Davidson et al. 2002; Wilkins 2002; Carroll et al. 2005).Such tinkering permits evolvability, but also leads to integra-tion, as manifested through covariation among structures. Itfollows that mutations that produce evolutionarily significantvariation can generate seemingly unrelated suites of integratedphenotypic change that may obfuscate where selection acted(e.g., Kangas et al. 2004).Here, we focus on the problem of how modifications tocomplex developmental pathways cause integrated phenotyp-ic change in the skull. The skull is arguably the most com-plexly integrated region of the skeleton because it comprisesmany closely packed capsules surrounding organs and spacessuch as the brain and the pharynx, in which most bony wallsare shared between capsules (Moss and Young 1960; Enlow1990; Moss 1997a, b, c, d). In addition, the skull performsmany dynamic functions, some of which involve substantialmechanical forces that affect multiple regions. These functionsare maintained as the skull changes enormously in size andshape during ontogeny. Without multiple mechanisms of in-tegration, such a complex structure would probably fail togrow and function correctly, and evolve. Finally, the skull hasbeen the locus of many key evolutionary transformations,including changes associated with the origins of our ownspecies, Homo sapiens (Lieberman et al. 2002).The skull’s complexity poses many challenges for under-standing its evolutionary developmental bases. However,there may be some simplifying principles. The skull compris-es three partially independent and embryologically distinctunits that surround a few dominant organs and spacesEVO LUTION & DEVELOPMENT 9:1, 76 –91 (2007)& 2007 The Author(s)Journal compil ation & 2007 Blackwell Publishing Ltd.76(Fig. 1): the basicranium (derived from the chondrocranium),the neurocranium (the dermatocranial bones of the cranialvault), and the face (derived initially from the splanchnocra-nium with subsequent development of dermatocranial elem-ents). These regions behave as modules by varying somewhatindependently (Cheverud 1982b, 1989, 1995; Lieberman et al.2000b; Hallgrı´msson et al. 2004b). In addition, these modulesmay interact unequally via epigenetic interactions (we use theterm epigenetic in a general sense to refer to interactions be-tween a given cell and its environment, includ ing other cells,that influences the cells activity).There are several reasons to hypothesize that the basi-cranium acts as the skull’s central integrator (De Beer 1937;Lieberman et al. 2000a). First, the basicranium is located inthe center of the skull, below the brain and neurocraniumbut above and behind the face. Thus, variation is transmit-ted between the face and neurocranium indirectly via thebasicranium. In addition, the basicranium is the first part ofthe cranium to attain adult size and shape, slightly beforethe neurocranium, and long before the face (Stamrud 1959;Moore and Lavelle 1974; Baughan et al. 1979; Farkas et al.1992). Second, the basicranium grows mostly via endo-chondral ossification in synchondroses (the spheno-occipi-tal, mid-sphenoidal, and spheno-ethmoid). The face andneurocranium, by contrast, grow via intramembranous os-sification in sutures. This distinction is relevant because en-dochondral ossification may be less subject to epigeneticinteractions with nearby organs than intramembranous os-sification. Intramembranous ossification in the neurocrani-um and face is driven almost completely by organ growthwithin capsules in which mechanical forces upregulate tran-scription factors in sutures to induce osteogenesis (e.g.,Opperman 2000; Wilkie and Morriss-Kay 2001; Yu et al.2001; Spector et al. 2002), but synchondroses elongate muchlike