Facial Shape Variation in Humans

University of Calgary PRISM: University of Calgary's Digital Repository

Graduate Studies The Vault: Electronic Theses and Dissertations

2017 Facial Shape Variation in Humans

Larson, Jacinda R.

Larson, J. R. (2017). Facial Shape Variation in Humans (Unpublished doctoral thesis). University of Calgary, Calgary, AB. doi:10.11575/PRISM/26141 http://hdl.handle.net/11023/4208 doctoral thesis

University of Calgary graduate students retain copyright ownership and moral rights for their thesis. You may use this material in any way that is permitted by the Copyright Act or through licensing that has been assigned to the document. For uses that are not allowable under copyright legislation or licensing, you are required to seek permission. Downloaded from PRISM: https://prism.ucalgary.ca UNIVERSITY OF CALGARY

Facial Shape Variation in Humans

Jacinda R. Larson

A THESIS

SUBMITTED TO THE FACULTY OF GRADUATE STUDIES

IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE

DEGREE OF DOCTOR OF PHILOSOPHY

GRADUATE PROGRAM IN MEDICAL SCIENCES

CALGARY, ALBERTA

October, 2017

The human face is a highly variable feature. Facial shape variation is seen between and within multiple populations. However, the source of this variation is mostly unknown.

Furthermore, phenotypic variation is observed within syndromes that display a unique craniofacial phenotype. Because of this inherent variation, facial shape has a large clinical important and diagnostic significance. To explore facial shape variation, I have conducted several studies that examine the role of morphological integration in producing coordinated shape changes in the face.

In these studies, I hypothesized that the size of the face and skull are correlated with facial shape; and that dysmorphic patients will display different morphological relationships between parts of the face, when compared to controls. These hypotheses were tested in several groups; namely healthy controls, patients with non-syndromic cleft lip and/or palate, and patients with ectodermal dysplasia. While these conditions have different etiologies, genetics, inheritance, mechanism of development, and resultant facial shape, using both of these patient groups has allowed me to explore these hypotheses in starkly different phenotypic groups. My studies have shown that the allometric factors of facial shape variation are complex and overlapping. Size represents a relatively small proportion of facial shape variation in humans. Furthermore, different classifications of cleft phenotypes are crucial when deciphering covariance structure in cleft individuals. Collectively, these studies have furthered the knowledge of complex craniofacial phenotypes, and have extended the knowledge on how allometry affects human facial shapes.

Keywords: phenotypic variation, birth defects, cleft lip, cleft palate, ectodermal dysplasia, geometric morphometrics, Tanzania, Tanzanians, bantu, covariation, covariance

ii Acknowledgements

Drs. Benedikt Hallgrimsson, Heather Jamniczky, Robertson Harrop, and Campbell Rolian; thank you for the academic and scholarly guidance and assistance over the years.

Thank you to Dr. Mange Manyama for his camaraderie and friendship, and for his leadership in organizing our field work seasons in Mwanza, Tanzania. Thank you to Dr. Charles Roseman for teaching me how to write code using R software. I appreciate your patience and kindness through my many struggles with learning R. Thank you to the Cleft Palate Clinic at the Alberta Children’s

Hospital for facilitating my research data collection. Thank you to Edimer Pharmaceuticals for facilitating my research data collection in Houston, Texas. Thank you to my lab colleagues Drs.

Hayley Britz, Rebecca Green and Chris Percival; and J. David Aponte. You have all made me a better scientist. Finally, thank you to all of the study participants. I would not have been able to complete my research without your participation.

Thank you to my parents Claire Jack and Leander Larson for always encouraging me to keep going; I would have given up without your support. Thank you to my late Gramma, Ada Emard, for always insisting that I was allowed to be “bossy”; and for teaching me that women can accomplish anything that men can – but that it doesn’t hurt to look like a lady while doing so.

Thank you to my rock, Leanne. You came at a time that I needed you the most.

Thank you to the Alberta Children’s Hospital Research Institute, and to Alberta Innovates Health

Solutions, for the generous funding to complete my research.

iii Dedication

"And to all the little girls who are watching this, never doubt that you are valuable and powerful and deserving of every chance and opportunity in the world to pursue and achieve your own dreams."

-Hillary Rodham Clinton

Table of Contents Abstract ...... ii Acknowledgements ...... iii Dedication ...... iv Table of Contents ...... v List of Tables ...... viii List of Figures and Illustrations ...... ix List of Symbols, Abbreviations and Nomenclature ...... xiii

CHAPTER 1: INTRODUCTION ...... 1 Introduction ...... 1 1.1.1 Thematic Overview ...... 1 1.1.2 Craniofacial Development ...... 3 1.1.3 Allometry and Variation ...... 8 1.1.4 Congenital Craniofacial Anomalies and Syndromic Phenotypes ...... 10 1.1.5 Synthesis of Background Information ...... 11 Organization of the Dissertation ...... 12 1.2.1 General Information ...... 12 1.2.2 Chapter Two ...... 12 1.2.3 Chapter Three ...... 13 1.2.4 Chapter Four ...... 14 1.2.5 Chapter Five ...... 15

CHAPTER 2: BODY SIZE AND ALLOMETRIC VARIATION IN FACIAL SHAPE IN CHILDREN ...... 16 Abstract ...... 16 Introduction ...... 17 Materials and Methods ...... 19 2.1.1 Sample Collection...... 19 2.1.2 Three-dimensional Imaging and automated 3D Landmarking...... 21 2.1.3 Morphometric Analysis ...... 23

v Results ...... 26 2.1.4 Distribution of variation across PCs ...... 26 2.1.5 Analysis of allometric variance components ...... 30 2.1.6 Variation due to Sex and Population ...... 34 2.1.7 Comparison of linear and polynomial regressions ...... 37 2.1.8 Comparisons of the allometric trajectories associated with different measures of size and age ...... 38 Discussion ...... 44 Acknowledgements ...... 49

CHAPTER 3: VARIATION IN NON-SYNDROMIC CLEFT LIP AND/OR PALATE PATIENTS ...... 50 Abstract ...... 50 Introduction ...... 51 Materials and Methods ...... 54 3.1.1 Study Subject Demographics ...... 54 3.1.2 Three-Dimensional Imaging of Subjects and Automated 3-D Landmarking .56 3.1.2 Morphometric Analysis ...... 60 Results ...... 62 3.1.3 Facial Shape Differences Between Groups ...... 62 3.1.4 Allometric Variation ...... 66 3.1.5 Morphological Integration ...... 78 Discussion ...... 82 Acknowledgements ...... 87

CHAPTER 4: CRANIOFACIAL MORPHOMETRIC ANALYSIS OF INDIVIDUALS WITH X-LINKED HYPOHIDROTIC ECTODERMAL DYSPLASIA ...... 88 Abstract ...... 88 Introduction ...... 89 Material and Methods ...... 90 4.1.1 Study Subject Demographics ...... 90 4.1.2 3D Imaging and Landmarking ...... 93

vi 4.1.3 Statistical Shape Analyses ...... 95 Results ...... 98 4.1.4 Facial shape of XLHED individuals differs from controls ...... 98 4.1.5 Characteristic midfacial shape in XLHED individuals ...... 102 Discussion ...... 103 Acknowledgments ...... 105 Conflict of Interest ...... 105

CHAPTER 5: CONCLUSION AND FUTURE DIRECTIONS ...... 106 Introduction ...... 106 Synthesis ...... 107 Future Directions ...... 111 Concluding Statement ...... 113

REFERENCES ...... 115

APPENDIX ...... 136

vii List of Tables Table 2-1: Anatomical descriptions and variable names of 29 landmarks used in the study. .22

Table 2-2: Eigenvalues and variances for Principal Components 1-10, Tanzanian and European-Derived North American results are shown...... 27

Table 2-3: Anatomical shape changes across Principal Component axes...... 28

Table 2-4: Tanzanian Procrustes ANOVA Model. Relative proportion of variation attributable to several allometric measures...... 31

Table 2-5: Procrustes ANOVA with permutation for face shape by population, sex and age...... 35

Table 3-1: Three-dimensional anatomical landmark descriptions. Landmark name, abbreviation and anatomical description. As referenced from http://www.facebase.org...... 59

Table 3-2: Procrustes ANOVA model results. Combined Sample, nsCL/P Sample, and Control Sample. Significant results are highlighted...... 67

Table 3-3: Matrix correlation between covariance matrices. Separate covariance matrices derived for bilateral, left, right, complete, incomplete, cleft and control groups...... 81

Table 3-4: Supplementary Table 1: Counts of patients for each classification grouping, by laterality...... 87

Table 4-1: Gene mutations in our cohort of 23 XLHED individuals. Asterisk (*) denotes mutation of brother pair...... 92

Table 4-2: Facial landmarks utilized in morphometric analysis ...... 94

Table 4-3: Magnitude of shape change by landmark and principal component. Calculated from principal component loadings. Highlighted columns indicate the greatest five magnitudes for each PC...... 101

viii

List of Figures and Illustrations Figure 2-1: Anatomical Landmarks. 29 landmarks as placed on the 3-D facial photo scans. .21

Figure 2-2: Thin-plate spline warps for PC 1-5 of Tanzanian and North American sample. Negative and positive PC scores represented. Color map diagrams represent areas of greatest difference (red) and least difference (blue)...... 29

Figure 2-3: Thin-plate spline warps of Tanzanian allometric variation. Thin-plate spline warps showing variation across age, centroid size (CS), head circumference (HC), height (HT), and weight (WT). Negative end of the axis of variation is displayed in the left...... 32

Figure 2-4: Thin-plate spline warps of North American allometric variation. Thin-plate spline warps showing variation across age, centroid size (CS). Negative end of the axis of variation is displayed in the left column, while the positive is displayed on the right. ...33

Figure 2-5: Facial shape effects by population and sex. A shows the mean face shapes for the Tanzanian and North American samples and the differences between those means as a heatmap. C shows exaggerated morphs for Males and Females. These were calculated as 2.5X the Procrustes distance between the sexes after standardizing for age. D shows the regression of face shape on age by sex...... 36

Figure 2-6: Variance explained by linear versus polynomial regressions (two to five terms) for each size variable and age...... 37

Figure 2-7: Three-dimensional morphs showing the facial shape variation that corresponds to each size measure and age (A). The morphs are scaled to 2 standard deviation departures from the mean in each direction. B shows heatmaps that correspond to these morphs. .40

Figure 2-8: Regressions of Conditional Variation. A) Regression of the conditional variation for each variable against the first PC of the size measures and age. B) Visualization of the correlation matrix for the size measure and age. C) The shape variances explained by each variable for the regressions of the conditional variation...... 41

ix Figure 2-9: Vectors in Size Measure Regressions. A shows the shape vectors that correspond to the regressions of face shape on each size measure and age. Some vectors point inwards from the surface of the face. B shows the distributions of angles among resampled vectors. Since all angles are positive, a mean angle of 0 is not possible. The Null distribution shows the expected distribution of angles when the angles between the vectors for the same regression is resamples. The blue line shows the expected mean when the angles are orthogonal (random) while the red line shows 0 (completely parallel)...... 43

Figure 3-1: Anatomical positions of the head used in three-dimensional photography capture with the Gemini. Top row: frontal, half right, right. Bottom row: inferior chin, half left, left...... 57

Figure 3-2: Anatomical landmarks (29) used on digitally rendered 3-D surface. Frontal and Lateral views shown. Corresponds to Table 3-1...... 58

Figure 3-3: 3-D Morphs (thin plate spline warps) of the average facial shapes of nsCL/P and controls. Also shown is the average grand mean facial shape of the entire sample...... 64

Figure 3-4: Heatmap showing localized mean shape differences between the nsCL/P and control groups. Colour map showing distance measurement (mm) and colour correlation, with blue showing the greatest distance and purple showing the smallest...... 65

Figure 3-5: 3-D thin plate spline warps (morphs) showing the resultant shape variation that corresponds to each group and allometric measure. The morphs are scaled to +/- 2 Standard Deviations from the mean...... 68

Figure 3-6: Regression models for age and centroid size with cleft classifications. A: Control/Cleft, B: Cleft Type, C: Laterality, D: Severity. (*** denotes significant interaction term between centroid size and classification factor)...... 69

Figure 3-7: Morphed mean shapes for laterality classification. Combined group, left, right and bilateral morphs shown...... 71

Figure 3-8: Morphed mean shapes for severity classification. Combined group, incomplete and complete morphs shown...... 72 x Figure 3-9: Growth trajectory morphs for centroid size (CS), by laterality. Frontal view. Trajectory progressions shown (from minimum to maximum centroid size): Left, right, and bilateral clefts. Also Shown is entire combined nsCL/P group...... 74

Figure 3-10: Growth trajectory morphs for centroid size (CS), by laterality. Lateral view. Trajectory progressions shown (from minimum to maximum centroid size): Left, right, and bilateral clefts. Also Shown is entire combined nsCL/P group...... 75

Figure 3-11: Growth trajectory morphs for centroid size (CS), by severity. Frontal view. Trajectory progressions shown (from minimum to maximum centroid size): Incomplete and complete clefts. Also shown is entire combined nsCL/P group...... 76

Figure 3-12: Growth trajectory morphs for centroid size (CS), by severity. Lateral view. Trajectory progressions shown (from minimum to maximum centroid size): Incomplete and complete clefts. Also shown is entire combined nsCL/P group...... 77

Figure 3-13: Comparisons of the Integratedness (SVE) and Multivariate Variances, grouped by various diagnostic criteria. A shows laterality, B shows severity, and C shows cleft vs control. Error bars obtained represent standard deviations in the variable, through resampling 1000 times. Significance level calculated via the overlap between the curves, 1000 repetitions...... 79

Figure 3-14: Covariance Distance vs Procrustes Distance. Groups shown include all laterality comparisons, severity, and cleft/control...... 81

Figure 4-1: Landmarks collected from digitized 3D facial photographs. Correspond to landmarks in Table 4-2...... 93

Figure 4-2: Multivariate pooled within-group regression of shape on centroid size (A) and age (B)...... 96

Figure 4-3: Multivariate shape analyses of craniofacial features of XLHED subjects compared to controls. (A) PC1 versus PC2, showing shape distribution of XLHED and control individuals. Ellipses correspond to 95% confidence intervals. Thin-plate spline warps illustrate the shape changes in PC1, corresponding to the observed zero, positive, and xi negative extreme values. (B) Canonical variate (CV) analysis histogram showing shape distribution of XLHED and control individuals. Thin-plate spline warps illustrate the shape changes in CV1, corresponding to the observed zero, positive, and negative extreme values...... 100

Figure 4-4: Magnitude of shape change by principal component. Magnitude of shape change for PCs 1–3, as calculated from PC loadings. Magnitudes are magnified by 2X...... 101

xii List of Symbols, Abbreviations and Nomenclature

Symbol Definition

ED Ectodermal dysplasia

HED Hypohidrotic ectodermal dysplasia

XL X-linked recessive

AR Autosomal recessive

AD Autosomal dominant

EDAR Ectodysplasin associated receptor

EDARADD EDAR-associated death domain

XLHED X-linked hypohidrotic ectodermal dysplasia

3D Three-dimensional

GM Geometric morphometrics

PCA Principal components analysis

CVA Canonical variates analysis

TNF Tumor necrosis factor

NIH National Institutes of Health

PC Principal component

ANOVA Analysis of variance

CS Centroid size

HC Head circumference

HT Height

xiii WT Weight

Eur.der.NA European Derived North American

CL/P Cleft lip and/or cleft palate

CL Cleft lip

CP Cleft palate nsCL/P Non-syndromic cleft lip and/or cleft palate

SCHIP1 Schwannomin interacting protein 1

PDE8A Phosphodiesterase 8A

Gemini Inspeck© 3-D Gemini Camera System

3dMD 3dMD© 3dMDtrio Camera System mm Millimeter

SVE Scaled variance of the eigenvalues

DS Down’s Syndrome

FASD Fetal Alcohol Spectrum Disorder

NF1 Neurofibromatosis Type 1

xiv

Chapter 1: Introduction

Introduction

1.1.1 Thematic Overview

The study of facial shape variation is important for several reasons. First, the face is a highly variable human feature, and facial variation is observed both between and within various population groups. The source of this variation is largely unknown, but recent genetic association studies have shown that certain gene sequences correlate with specific aspects of facial size and shape (Cole et al., 2016). Secondly, facial shape and by extension, facial shape variation, has a large clinical importance and diagnostic significance. Many previously characterized and novel genetic syndromes display a distinct facial phenotype (Allanson, O'Hara, Farkas, & Nair, 1993;

Bhuiyan et al., 2006; de Campinas, 2009; Goodwin et al., 2014; Hill et al., 2013; Richtsmeier,

Zumwalt, Carlson, Epstein, & Reeves, 2002). While phenotypic variation does occur within these syndromic groups, the shared common features specific to the syndrome define the syndromic facial phenotype. It is important to recognize that while these syndromic patient groups share a commonality of phenotype, these individuals also display strong familial resemblance. In some cases, a subclinical phenotype of the syndrome may even be present in the parents (Weinberg et al., 2008). To explore facial shape variation, I have conducted several experiments that converge on a central hypothesis; that the size of the skull and face are correlated with facial shape.

Specifically, we hypothesize that non-syndromic cleft lip and/or palate (nsCL/P) patients and control groups will have vastly different morphological relationships between parts of the face, and that the relationship of size and shape will play a significant role in the morphological covariation of the human face. Furthermore, I examine an ectodermal dysplasia (ED) patient cohort as a case study of a syndrome with a craniofacial phenotype, and hypothesize that the

syndromic phenotype is quantifiable, and significantly differs from that of controls. The rationale behind choosing patients that display nsCL/P and also patients with ED, is to explore these hypotheses in two vastly different phenotypic groups. nsCL/P represents a patient group with a severely dysmorphic phenotype resulting from an extreme developmental perturbation, while ED is an inherited genetic disorder that does not primarily affect the face. However, patients with ED do display a characteristic facial phenotype. By examining different categories of syndromes (i.e.: congenital and inherited genetic), our hypotheses are able to be tested across a broad patient group, and will further the understanding of morphological integration in the skull.

In this chapter, I will discuss the development of the craniofacial complex. As development shapes the face, an understanding of the complexities of development is necessary. Furthermore,

I will discuss the generalities of craniofacial syndromes and their implications on facial shape and facial shape variation. Lastly, a discussion of morphological integration and how integration is important in development will follow. The integration between physical structures and the modular nature of development influence the evolvability of a trait. Morphological integration is an important tendency that is observed in mammals (Hallgrimsson et al., 2009). The results of the studies contained within this dissertation will ultimately further the understanding of complex craniofacial phenotypes, and extend the understanding of how phenotypic variation arises in the craniofacial complex. This knowledge could potentially be applied to clinical studies, and used in the clinical management of complex dysmorphic craniofacial phenotypes.

1.1.2 Craniofacial Development

The growth and development of the craniofacial complex is a highly regulated process that requires the organized integration of various specialized tissues, such as the surface ectoderm, neural crest cells, mesoderm, and pharyngeal endoderm. Historically, the skull has been divided into two main bone groups; the neurocranium—consisting of the calvaria or the bones which surround and protect the brain, and the viscerocranium formed via intramembranous neural crest- derived bones (G. H. Sperber, Sperber, & Guttmann, 2010). During the development of the head, a concentration of mesenchyme surrounding the hindbrain begins the formation of a floor for the brain. The conversion of this mesenchyme into cartilage marks the initial development of the chondrocranium, beginning at approximately 40 days post conception. The chondrocranium represents a unique portion of the skull; the anterior portion is derived via neural crest cells, while the posterior portion is of mesodermal origin (G. H. Sperber et al., 2010). Both the ossifying calvaria and the ossifying chondrocranium together form the neurocranium (G. H. Sperber et al.,

2010).

Beginning in the 4th week of normal development, the formation of the head is initiated through the migration of rhombomere-derived neural crest cells (Levaillant, Bault, Benoit, &

Couly, 2017). These neural crest cells then form the facial prominences. The face develops via the five primordial facial prominences; the single median frontonasal prominence, and the bilaterally paired maxillary and mandibular prominences (Jiang, Bush, & Lidral, 2006). Tissues from the frontonasal process develop separately from the first pharyngeal arch derived mandibular and maxillary prominences (Jiang et al., 2006). Differentiation of the facial skeleton bones is

facilitated via epithelial-mesenchymal interaction (G. H. Sperber et al., 2010). The facial bones develop intramembranously via ossification centers in the facial prominences. The distal tips of the paired maxillary prominences merge in the midline to form the upper lip and the primary palate

(Jiang et al., 2006). The space between the paired prominences “fills in” via rapid proliferation of the tissue in the midline(Jiang et al., 2006). The initial contact and fusion occurs between the medial and lateral nasal processes, forming the rudimentary olfactory pits (Mossey, Little, Munger,

Dixon, & Shaw, 2009). Following this contact, fusion between the maxillary prominences and the medial nasal prominences forms the structural basis of the primary palate (Hinrichsen, 1985a). In fusion, the epithelium, and by extension the epithelial seam between the two prominences, is broken down through epithelial mesenchymal transition (Hinrichsen, 1985a).

The development of the primary palate occurs during the 5th and 6th week of gestation(G.

H. Sperber et al., 2010). The medial and lateral nasal processes first come into contact, followed by the contact of the medial nasal and maxillary prominences. This site of contact forms an epithelial sheet, which will function as the inferior nasal epithelium and the superior oral epithelium. These layers are then separated, which establishes the structural basis of the primary palate. Posterior to the primary palate, the nasal and oral epithelium are still in contact but soon undergo separation. This separation then connects the nasal cavity to the oral cavity (G. H. Sperber et al., 2010). The development of the secondary palate occurs through the outgrowth of the palatal processes from the bilateral maxillary processes. During the 7th and 8th week, the palatal processes grow downwards; parallel to the developing tongue. Concurrently, the mandibular prominences also grow and lower, which contracts and lowers the tongue. The head also lifts. Following the

lowering of the tongue and mandibular prominences, the palatal processes elevate and begin to horizontally position themselves with the primary palate for fusion in the midline(Diewert, 1985).

This is followed by midline fusion to form an epithelial seam that degrades into mesenchyme, which then subsequently differentiates into the hard and soft palate (Mossey et al., 2009). The further development and separation of the oronasal space into the nasal and oral cavities allows for future mastication and respiration to occur simultaneously (G. H. Sperber, 2002). Following the compartmentalization of the nasal and oral cavities, the anterior portion of the primitive primary palate grows and extends to form the anterior portion of the palate (Hinrichsen, 1985b).

The posterior portion of the primitive palate subsequently fuses with the maxillary-derived secondary palate to form the palate (Hinrichsen, 1985b).

During development, mesenchymal cells differentiate into osteoblasts that will give rise to the fetal skull. Osteogenesis of these cells occurs via development of an osteoid matrix that covers the developing brain with intramembranous bone (G. H. Sperber et al., 2010). In the absence of the brain, as seen in anencephaly, no discernable calvaria forms. The mesoderm-derived tissue differentiates and gives rise to the parietal, sphenoid, temporal and occipital bones. Neural crest cell derived mesenchyme forms the lacrimal, nasal, zygomatic, maxillary and mandibular bones

(G. H. Sperber et al., 2010). These bones form via primary ossification centers that form as early as the beginning of the eighth week of pregnancy. It is important to note that the calvaria are not fully ossified during fetal development. The cranial vault continues to grow throughout childhood and adolescence. At birth, the cranium must compress and deform to pass through the pelvis and birth canal; this is facilitated via the soft fibrous sutures (or fontanelles) occurring between the

junctions of the cranial bones (Jiang et al., 2006). However, in craniosynostosis syndromes such as Cruzon or Apert syndrome, the sutures prematurely ossify and cause characteristic deformation of the brain and skull (Hill et al., 2013). While the majority of the sutures close by two years of age, the frontal bones do not unite into a single bone typically until seven years of age (G. H.

Sperber et al., 2010). The differences between the human neonatal skull and the adult skull are largely summarized by difference in size and in the number of bones (G. H. Sperber, 2002). During postnatal growth, the sutures become narrowed and fontanelles disappear. Another important feature of skull growth is the predominance of the neurocranium over the facial skeleton. The developing brain grows very rapidly, which in turn creates neurocranium growth. However, neurocranium predominance is most marked in the early years; by birth the neurocranium has already achieved 25% of it’s future growth, and by 10 years of age, 95% of that growth is complete.

Conversely, at 10 years of age, the facial skeleton has only achieved 65% of its total growth (G.

H. Sperber et al., 2010).

The growth of the skull and the face are important concepts in craniofacial development.

Growth of the calvaria occurs via structural growth, bone remodeling and through functional matrix growth, or the displacement of the bone from the rapidly expanding brain (G. H. Sperber et al., 2010). Adult features of the cranium such as the temporal and nuchal lines, the superciliary arches, and the external occipital protuberance all occur via growth of the external cranium.

Throughout life, the bones of the cranium continue to thicken even while not actively growing

(Schoenwolf & Bleyl, 2009). Furthermore, cranial shape may be distorted during development resulting in conditions such as brachycephaly or platycephaly, but the cranial volume remains

largely unaffected (G. H. Sperber et al., 2010). During growth, the concept of allometry becomes crucial. Allometry, or the change of shape in relation to size, is an important feature of cranial growth and development. As the skull changes in size, a coordinated shape change in the face may also occur.

Many recent studies have shown that there is a developmental integration between the brain and face (Parsons et al., 2011; Weinberg et al., 2013), occurring in “normal” and dysmorphic individuals; meaning that the size and shape of the developing brain has direct structural and shape effects on the developing face. This is seen in conditions such as craniosynostosis, where the expanding brain is constrained by the prematurely ossified sutures, and associated facial defects occur. This strongly suggests that the integration between the brain and face is of critical relevance in syndromes of facial malformations. (Parsons et al., 2011; Weinberg et al., 2013).

It is clear that the development of the face is an extraordinarily complex process with a substantial amount of phenotypic variation. This phenotypic variation is likely derived through highly complex genomic variation involving a very large number of genes and alleles. In addition, genomic factors and gene-gene interactions, combined with an overall high tolerance for subtle differences contribute to the development of an aesthetically “normal” human face.

1.1.3 Allometry and Variation

Morphological variation tends to be integrated, meaning that physical traits exhibit varying tendencies for coordinated variation (Hallgrimsson et al., 2009). The face and craniofacial complex of humans and other vertebrates display morphological variation (Bastir, 2008; Lieberman, Carlo,

Ponce de León, & Zollikofer, 2007). In model organisms such as the mouse and chick, recent work has shown how modulation of developmental processes drives integrated axes of variation, resulting in covariation among different parts of the face (Chong et al., 2012; Hallgrimsson et al.,

2009; Lieberman et al., 2007; Young, Chong, Hu, Hallgrimsson, & Marcucio, 2010). Studies on the impact of variation in the growth of the chondrocranium (Bastir, 2008; Parsons et al., 2011;

Parsons, Downey, Jirik, Hallgrimsson, & Jamniczky, 2015) or the brain (Lieberman,

Hallgrimsson, Liu, Parsons, & Jamniczky, 2008) on craniofacial morphology strongly suggest that there are major developmental determinants of covariation patterns for craniofacial morphology.

For facial shape, somatic growth is likely to be one such determinant. This is because morphology tends not to be independent of size – a phenomenon known as allometry.

Allometry can be defined as the relationship between shape and size (Klingenberg, 1998a;

Klingenberg & Zimmerman, 1992). For most anatomical structures, shapes and proportions have some regular relationship to size (Jolicoeur, 1963). Small mammals, for example, tend to have brains that are larger, relative to body size, than large mammals (Jerison, 1973). Allometry is typically separated into ontogenetic and static components. Ontogenetic allometry is the shape variation, or proportion of variation, that correlates with age or developmental stage; while static allometry is the shape variation that correlates with size, whilst controlling for age or stage

(German & Meyers, 1989). While allometry is a frequently observed pattern of variation, how size

influences the developmental determinants of shape variation is poorly understood. Allometry can be characterized as a unique case of morphological integration, the tendency for structures to covary (Hallgrimsson et al., 2009). This covariation occurs due to developmental processes affecting multiple traits. In allometric variation, it is assumed that the growth process is driving integration. Studies of allometry often assume that size can be related unambiguously to shape. It is common, for example, to use the centroid size of a landmark configuration as measure of size.

This approach is based on the assumption that the size and shape of a particular structure should be separated in data analyses (Klingenberg, 2016). However, the size of a structure, such as the face, will also vary in relation to organismal size as well as to age or stage. In fact, there are multiple ways to quantify growth and size and it is rarely intuitively obvious what is the most relevant measure of size for a particular analysis. These various measures of size may relate differently to variation in shape. Quantifying this complexity is important for framing hypotheses about the mechanistic basis for allometry.

Understanding the mechanistic basis for allometry is important for the study of facial shape, and for this dissertation, for several reasons. Many genetic perturbations influence size as well morphology and it may be necessary to disentangle those effects. Furthermore, allometric variation can confound studies of the genetics of facial shape when facial shape variation that occurs as a consequence of size masks other genetic influences on facial shape (Cole et al., 2016).

Finally, allometry represents an important axis of integration that constrains and is altered by evolution (Klingenberg, 1998a; Klingenberg & Marugán-Lobón, 2013).

1.1.4 Congenital Craniofacial Anomalies and Syndromic Phenotypes

Many genetic syndromes display a subtle but clinically distinctive craniofacial phenotype

(Basel Vanagaite et al., 2016; Chehadeh-Djebbar et al., 2013), or syndromic facies. In facial clefts, including nsCL/P, some of the facial prominences fail to fuse. In addition to overt craniofacial abnormalities such as nsCL/P, Some syndromes, such as achondroplasia, can produce size and shape related changes; the observed allometry, or the relationship of body size to shape, of these syndromes then becomes crucial in describing the coordinated phenotypic outcome. Clinicians are trained to recognize these subtle syndromic phenotypes, and may use this information as an informal differential diagnostic tool (Biesecker & Carey, 2011). These phenotypes contain characteristic facial shape morphologies that may provide clues in determining any possible underlying genetic syndromes (de Campinas, 2009). These non-dysmorphic subtle syndromic phenotypes are often thought of as occurring at either extreme end of the spectrum of normal human facial morphology, as the variation in the disease is superimposed upon general biological consistency. The need for a standard classification system of dysmorphology and dysmorphic phenotypes is a recurring issue in clinical medicine and the medical sciences (Biesecker & Carey,

2011).

Congenital anomalies of the craniofacial complex are among the most prevalent congenital birth defects, occurring from 1:500 to 1:2500 live births (Sivertsen et al., 2008). Orofacial clefts may present as isolated cleft lip, isolated cleft palate (secondary palate), or a combination of both cleft lip and palate. Clinically, facial clefts present as either unilateral or bilateral fissures of the upper lip and/or palate, and consist of an incomplete cleft isolated to the lip and/or palate, or a complete cleft extending through the upper lip and into the nostril (Berkowitz, 1996). The etiology

of cleft lip and palate is highly complex. The development of CL/P is thought to likely involve genetic, environmental and genotype-phenotype interactions (Hallgrimsson, Dorval, Zelditch, &

German, 2004). While multiple candidate genes have been identified through chromosomal linkage studies, epistasis and multiple gene-gene interactions are hypothesized to contribute strongly to the development of non-syndromic CL/P (nsCL/P) (Nikopensius et al., 2011). nsCL/P has been shown to be heritable, and tends to run in families (Kumar S, Gopalkrishnan, Bhasker

Rao, & Ganeshkar, 2010; McIntyre & Mossey, 2002; Mladina, Skitarelić, & Skitarelić, 2009;

Mladina et al., 2008). Furthermore, a recent study has shown significant association between clefting and single nucleotide polymorphisms at IRF6 and PAX7, critical developmental genes

(Moreno Uribe et al., 2017). Additionally, several lines of morphology-based research have also indicated that the morphological integration of the developing brain and face may related to the etiology of craniofacial malformations such as CL/P (Boughner et al., 2008; Parsons et al., 2011).

1.1.5 Synthesis of Background Information

This dissertation aims to explore facial shape variation in control individuals, non- syndromic cleft lip and/or palate patients, and ectodermal dysplasia (ED) patients. Because of the ubiquitous nature of growth and size changes occurring during growth, I hypothesize that the size of the skull and face are correlated with facial shape. In most morphometric studies, allometry is typically divided into static (size) and ontogenetic (age) components. However, size is typically represented with a single measure — body weight, stature or circumferential measurements. The crux of my thesis work is to explore how these different measures of size may affect facial shape, and how these patterns of allometry may overlap or converge. By examining the hypothesis in

three distinct groups (controls, nsCL/P and ED), I am able to explore my hypothesis in a broad population of normal controls, a developmental disorder, and an inherited condition. This work is important because variation in facial shape due to size may be important in diagnosing craniofacial conditions and developing clinical treatment plans.

Organization of the Dissertation

1.2.1 General Information

The research described and presented within this dissertation consists of a series of topically related studies that have been conducted over the past six years (September 2011-June

2017). This thesis is written in a “manuscript chapter” presentation style.

1.2.2 Chapter Two

Chapter Two uses data collected within Tanzania and North America to conduct an analysis with the aims of understanding morphological integration in the human face. Through this study, we explored the effects of growth on shape. Allometry is commonly misunderstood as a single measure that can be simply removed from morphometric data as a covariate. Here we show the extent to which age and other measures of body size influence patterns of allometry in human facial shape. For this chapter, I participated in data collection in Tanzania, imaging approximately

4000 individuals. Additionally, I performed the analyses and wrote the manuscript, along with my supervisor Dr. Benedikt Hallgrimsson.

This study was part of a larger collaboration effort known as “The FaceBase Consortium”

(http://www.facebase.org). The partnership with Dr. Mange Manyama of the Catholic University of Health and Allied Science in Mwanza, Tanzania generously allowed us to collect a large sample size of Bantu children. The findings of Chapter Two allowed a better understanding of morphological integration and allometry, and were used as a basis for the development of Chapter

Three. Chapter Two is currently a re-submitted manuscript under review, submitted to the

American Journal of Physical Anthropology (submission date: March 23, 2017, resubmission date:

August 20, 2017).

1.2.3 Chapter Three

Chapter Three aims to use the findings of Chapter Two as a basis for the study of a dysmorphic syndrome, non-syndromic cleft lip and/or palate (nsCL/P). This study allowed a continued investigation into facial shape integration. I also evaluated allometry as an axis of morphological integration, and explored the how confounding allometric measures can be used in analyses of allometry. This study was conducted at the Alberta Children’s Hospital Cleft Palate

Clinic, under the tutelage of Dr. Robertson Harrop MD, MSc, FRCSC. Dr. Harrop’s expertise was instrumental to the success of this project, as he provided key clinical insight into the data collection and study design. For this study, I designed the research question and data collection protocol. Additionally, I collected all of the data with the occasional assistance of an

Undergraduate student. I performed all data analyses, and wrote the manuscript. Chapter Three is currently a manuscript in preparation for submission to the Journal of Anatomy or the American

Cleft Palate Craniofacial Journal.

1.2.4 Chapter Four

Chapter Four is a descriptive study of a syndromic patient group. In Chapter Four, we performed a quantitative analysis of the craniofacial morphology of subjects with X-linked hypohidrotic ectodermal dysplasia (XLHED). XLHED is an X-linked form of hypohidrotic ectodermal dysplasia (HED), and is one of three forms of HED. HED also can occur as Autosomal

Recessive or Autosomal Dominant HED. HED is a syndrome that doesn’t directly affect the face, and is characterized by a reduced ability to sweat, fine and thin hair, and hypodontia. However, despite HED not directly affecting the face, it is widely clinically known that patients with hypohidrotic ectodermal dysplasia have a distinct facial appearance. At the time of publication there were no known clinical reports of a 3-D quantification of HED craniofacial morphology. The

XLHED patients were enrolled in the study at the University of California, San Francisco in May

2011 by the collaborating lab (Dr. Ophir Klein’s group); I also enrolled patients at the National

Foundation for Ectodermal Dysplasia Family Conference in Houston, Texas in July 2013. This study was funded by Edimer Pharmaceuticals, a biotechnology company developing innovative replacement protein therapies for newborn treatment of XLHED. Chapter Four is a published manuscript, of which I am the joint first-author along with Dr. Alice F. Goodwin MD, DDS.

Chapter 4 is published in: Molecular Genetics & Genomic Medicine 2014; 2(5): 422–429, doi:

10.1002/mgg3.84. For this project, I collected a portion of patient data in Texas. I also performed all landmarking and morphometric analyses. I wrote the manuscript with input from Dr. Alice F

Goodwin. Please see Appendix for Copyright information from the journal of Molecular Genetics

& Genomic Medicine.

1.2.5 Chapter Five

Chapter Five concludes the dissertation with an overview of the major findings and results from Chapters Two, Three and Four. Also included is a discussion of the significance of this work and future research directions.

Chapter 2: Body Size and Allometric Variation in Facial Shape in Children

Abstract

Morphological integration, or the tendency for covariation, is commonly seen in complex phenotypes such as the human face. The effects of growth on shape, or allometry, represent a ubiquitous but poorly understood axis of integration. I address the question of to what extent age and measures of size converge on a single pattern of allometry for human facial shape. Our study is based on two large cross-sectional cohorts of children, one from Tanzania and the other from the USA (n=7173). I employ 3D facial imaging and geometric morphometrics to relate facial shape to age and anthropometric measures. The two populations differ significantly in facial shape, but the magnitude of this difference is small relative to the variation within each group. Allometric variation for facial shape is similar in both populations, representing a small but significant proportion of total variation in facial shape. Different measures of size are associated with overlapping but statistically distinct aspects of shape variation. Only half of the size-related variation in facial shape can be associated with the covariation of four size measures and age while the remainder associates distinctly with individual measures. Allometric variation in the human face is complex and should not be regarded as a singular effect. This finding has important implications for how size is treated in studies of human facial shape and for the developmental basis for allometric variation more generally.

Introduction

Most morphological structures are integrated (Olson & Miller, 1958), meaning that they tend to co-vary among anatomical regions (Hallgrimsson et al., 2009). This is the case for the face and craniofacial complex in humans and other vertebrates (Bastir, 2008; Porto, de Oliveira, Shirai,

De Conto, & Marroig, 2009). Variation in the growth of the chondrocranium (Bastir & Rosas,

2006; Bastir, Rosas, & O’Higgins, 2006; Hallgrimsson et al., 2006; Parsons et al., 2015) or the brain (Lieberman et al., 2008; Marcucio, Young, Hu, & Hallgrimsson, 2011; Parsons et al., 2011) are major developmental determinants of covariation patterns for craniofacial morphology.

Somatic growth is likely to be another such determinant, because morphology tends to be related to size. For most anatomical structures, shapes and proportions have a regular relationship to size

(Jolicoeur, 1963). Taller people tend to have longer, more prognathic faces (Baume, Buschang, &

Weinstein, 1983; Mitteroecker, Gunz, & Windhager, 2013). The specific relationship between shape and size is termed “allometry” (Klingenberg, 2016; Klingenberg & Zimmerman, 1992).

Many genetic and environmental influences affect growth as well as other aspects of facial development. To disentangle the correlated effects of size from other more specific effects, it is necessary to understand the role of allometric variation.

Allometry is typically divided into ontogenetic versus static components. Ontogenetic allometry is the shape variation that correlates with age or developmental stage. Static allometry is the shape variation that correlates with size, controlling for age or stage (German & Meyers,

1989). It is not well understood how these shape correlates of size and age arise in development.

Allometry is a special case of morphological integration (Magwene, 2001) , which refers to the tendency for structures to co-vary because developmental processes tend to affect multiple traits

(Hallgrimsson et al., 2009) . For allometric variation, the process assumed to produce these correlated effects is growth. A challenge is that there are multiple ways to quantify both growth and size for most anatomical structures. For the human face, is the appropriate measure the length, width, or area of the face, head circumference, or some measure of body size? If allometric variation reflects the shape consequences of variation in a single underlying growth parameter, then the shape correlates of different size measures should converge on a single covariation pattern.

Here, I address the relationship between various measures of size and age to facial shape in two previously-described cross-sectional cohorts of children, one comprised of Bantu speaking groups in northwest Tanzania (Cole et al., 2016) and the other comprised of Americans of predominantly European ancestry (Shaffer et al., 2016) . I compare the shape correlates of age, two measures of somatic size (stature and body mass), and two local measures of head size (face size and head circumference). These measures are selected to capture disparate growth-related effects. Age represents ontogenetic effects. Stature reflects longitudinal growth while body mass captures overall somatic growth. Face size is related to local growth of the face while head circumference is influenced by overall head and brain size. Brain growth is known to influence craniofacial shape (Aldridge et al., 2005; Hill et al., 2013; Marcucio et al., 2011; Marcucio,

Hallgrimsson, & Young, 2015). I find that facial shape variation is highly structured, with most variation falling along a few axes of morphological co-variation. Allometric variation represents a relatively small fraction of total variance in facial shape (5%). The allometric component of variation is complex with age and the various measures of size correlating with overlapping but distinct patterns of covariation in facial shape.

Materials and Methods

2.1.1 Sample Collection.

The cohort of 5961 Tanzanian children of self-identified Bantu origin (3342 female, 2619 male) has been described in detail previously (Cole et al., 2016; 2017). Subject ages were 3 to 23 years (mean 10.8 ± 2.8 years; females, 10.7 ± 2.7 years; males, 10.9 ± 2.8 years). All participants were examined by a physician (M.M.) to exclude subjects with any birth defects, history of facial surgery or interventionist orthodontic treatment, or first-degree relatives with craniofacial abnormalities.

The cohort of 1212 North American children (609 female, 603 male) has also been described in detail previously (Shaffer et al. 2016). In this sample, we included only children who self-identified as “white” using the NIH “racial and ethnic categories” for recruitment and consenting of research subjects (https://grants.nih.gov/grants/guide/notice-files/NOT-OD-15-

089.html). European ancestry was confirmed from genomic data (Shaffer et al., 2016). Ages were

3 to 18 years (mean 9.0 ± 4.1 years; females, 9.2 ± 4.2 years; males, 8.9 3.9 years). Participants in the North American sample were screened using exclusion criteria similar to the Tanzanian cohort.

Anthropometric measurements of height, weight, and head circumference were obtained for a subsample of 4814 Tanzanian subjects (2675 female, 2139 male) by one of the investigators

(J.M.), taking the average of two sequential measurements. Height was measured in centimeters using a standard stadiometer; body weight was measured in kilograms using a digital body weight scale. Head circumference was measured in centimeters using a standard ribbon measuring tape,

with the measuring tape placed approximately two centimeters above the supraorbital ridge. The same anthropometric measurements were obtained for the North American subjects from Denver and San Francisco (n=706) while head circumference was not obtained for the Pittsburgh subjects

(n=506).

Ethical approval was granted by the Tanzania National Institute for Medical Research

(NIMR/HQ/R.8a/Vol.IX/845), and the University of Calgary (CHREB 21741), the University of

Colorado (09-0731) and the University of Pittsburgh (#PRO09060553 and #RB0405013).

Informed written consent was obtained from the parents and guardians of all participants, prior to participation in the study. Neither cohort is assumed to be representative of any biologically definable racial or ethnic category, nor do I assume that such categories exist in a biological sense

(Edgar & Hunley, 2009; Hunley, Healy, & Long, 2009). Rather, these cohorts sample two populations in which within-sample heterogeneity due to ancestry has been minimized, likely to a greater extent in the Tanzanian sample.

2.1.2 Three-dimensional Imaging and automated 3D Landmarking.

3D facial surface images were captured and processed as described previously (Cole et al.,

2016; Shaffer et al., 2016). Analysis of facial shape and size were based on 29 landmarks that were obtained using a novel automated landmarking method (Figure 2-1 and Table 2-1) (Li et al., 2017)

Figure 2-1: Anatomical Landmarks. 29 landmarks as placed on the 3-D facial photo scans.

Corresponds to anatomical descriptions in Table 2-1.

Table 2-1: Anatomical descriptions and variable names of 29 landmarks used in the study. As referenced from http://www.facebase.org

3D Landmark Abbrev- Definition Name iation Nasion n Midline point in where the frontal and nasal bones contact (nasofrontal suture). Corresponds to the underlying bony landmark. Pronasale prn Midline point marking the maximum protrusion of the nasal tip. Subnasale sn Midline point marking the junction between the inferior border of the nasal septum and the cutaneous upper lip. It is the apex of the nasolabial angle. Labiale Superius ls Midline point of the vermilion border of the upper lip, at the base of the philtrum. Stomion sto Midpoint of the labial fissure. Labiale Inferius li Midline point of the vermilion border of the lower lip. Sublabiale sl Midpoint along the inferior margin of the cutaneous lower lip (labiomental sulcus). Gnathion gn Midline point on the inferior border of the mandible. Corresponds to the underlying bony landmark. Endocanthion en_r Apex of the angle formed at the inner corner of the palpebral (Right) fissure where the upper and lower eyelids meet. Endocanthion (Left) en_l Same as above Exocanthion (Right) ex_r Apex of the angle formed at the outer corner of the palpebral fissure where the upper and lower eyelids meet. Exocanthion (Left) ex_l Same as above Alare (Right) al_r Most lateral point on the nasal ala. Alare (Left) al_l Same as above Alar Curvature Point ac_r Most posterolateral point on the alar cartilage, located within the (Right) crease formed by the union of the alar cartilage and the skin of the cheek. Alar Curvature Point ac_l Same as above (Left) Subalare (Right) sbal_r Point located at the lower margin of the nasal ala, where the cartilage insterts in the cutaneous upper lip. Subalare (Left) sbal_l Same as above Crista Philtri (Right) cph_r Point marking the lateral crest of the philtrum at the vermilion border of the upper lip Crista Philtri (Left) cph_l Same as above Chelion (Right) ch_r Point marking the lateral extent of the labial fissure. Chelion (Left) ch_l Same as above Tragion (Right) t_r Point marking the notch at the superior margin of the tragus, where the cartilage meets the skin of the face.

Tragion (Left) t_l Same as above Superior Alar supa_r Most superior portion of alar groove. Groove(Right) Superior Alar supa_l Same as above Groove (Left) Zygion (Right) z_r Most prominent portion of zygomatic arch Zygion (Left) Z_l Same as above Pogonion P Most prominent portion of chin, anatomical pogonion.

2.1.3 Morphometric Analysis

Following data corrections, I performed geometric morphometric analyses using R and

MorphoJ v1.04a (Klingenberg, 2011; R Development Core Team, 2015). Landmarks were subjected to Procrustes Superimposition to rescale to unit centroid size, translate to standard position, and rotate to standard orientation (Rohlf, 1999). To describe variation in facial shape in both populations, I used principal components analysis. Thin-plate-spline warps were constructed using Landmark software (Wiley et al., 2005) for the first five principal component (PC) axes.

Heat maps to visualize the areas of greatest shape differences were constructed using the Hausdorff function in MeshLab (Callieri, Ponchio, & Cignoni, 2008). Separate principal component analyses

(PCA) were performed for the Tanzanian and North American samples.

To estimate the relative proportions of shape variation attributable to size and growth measures and their interactions within each cohort, I used a linear model built in R (R Development

Core Team, 2015)with the Geomorph R package, version 3.0.3 (Adams & Otárola Castillo, 2013).

I used Geomorph’s procD.lm function to perform Procrustes ANOVA with permutation

procedures, to assess shape variation and patterns of co-variation, and thereby built a statistical model to quantify the relative amount of facial shape variation attributable to age, centroid size, height, weight, head circumference, sex, and the interactions between these variables. Since procD.lm uses type 1 error, I alternated the position of the last variable in sequential regressions to accurately estimate percentage variance attributable to each variable. To visualize the shape effects of age, centroid size, height, weight, head circumference and sex, regression scores were computed for the model using the RegScore function from the R Morpho package, version 2.4.1.1

(S. Schlager, Jefferis, & Schlager, 2016). All analyses of allometry were performed with data not corrected for age or size.

I also tested the effects of random error in the age values in the Tanzanian data on our estimates of allometric variation in face shape. To do this, we used three models: the first using the exact age, and then transforming the exact age to the nearest whole year, and finally to the closest 3-year interval.

To analyze allometric variation in the combined Tanzanian and North American cohorts, I first determined the variation due to sex and population using a multiple analysis of variance model for these factors and their interactions. This was implemented in the Geomorph package for R

(Adams & Felice, 2014; Collyer, Sekora, & Adams, 2015). I then removed ethnic group and sex from the analysis by centering the residuals from this model on the sample grand mean. I quantified and visualized the shape variation associated with each size measure and age using multiple linear regression implemented in Geomorph. Morphs and heatmaps were created in R using the Morpho package (S. Schlager et al., 2016).

In the combined Tanzanian and North American cohorts, I tested whether measures of size and age were associated with distinct effects on face shape. First, I used regressions for 1st, 2nd and 3rd, polynomials to determine the amount of variation captured by each variable using linear versus polynomial regressions. I then took three distinct approaches. In the first, I used multiple regression (procD.allometry) implemented in Geomorph for R to compare the slopes for each variable. In the second, I calculated the conditional variation for each variable, first removing the shape effects of all size variables and age, and then projecting those data using the coefficients obtained from a separate regression of each variable on the original data. This conditional variation is intended to compare, to the extent possible, the individual contribution of each variable to allometric variation. These coefficients were scaled to 1x the variance for each variable so that the method compares the same magnitude of effect across variables. The conditional variation for each dataset was regressed on the first PC obtained from the size measures. In the third approach, I performed regressions of the sex and population corrected data on each variable separately. I then obtained the vectors that correspond to these regressions from the regression coefficients and calculated the angles among them. To compare these vectors, I resampled the sex and population adjusted data with replacement and obtained the full set of vectors at each resampling iteration.

This approach thus compared the specific shape variation associated with each independent variable.

Finally, I asked to what extent the five measures converge on a single underlying allometry factor. To determine the proportion of variation in facial shape explained jointly by all five measures, I performed a PCA for age and the four measures of size. I then used the data corrected

for sex and population and estimated the proportion of variation in facial shape explained by each size/age PC.

Results

2.1.4 Distribution of variation across PCs

A principal component analysis (PCA) of the size and age corrected Procrustes coordinates revealed that facial shape variation is highly structured. In the Tanzanian cohort, along the PC1 axis (22.7% variance) individuals varied in relative upper facial height, overall facial width, nasal base width, and interorbital distance. Similar to PC1, PC2 (21.3% variance) described shape changes in total facial height and width, while also capturing the relative degree of maxillary prognathism. PC3 (14.7% variance) captured mandibular prognathism, upper facial depth, and midfacial length. In the North American cohort, along the PC1 axis (22.7%) subjects varied in facial height, facial width, philtrum height and chin protrusion. PC2 (18.9%) described variation in maxillary prognathism, chin protrusion and nasal projection. PC3 (16.5%) captured degree of retrognathia, nasal projection and interorbital distance. PC 1-10 Eigenvalue variances for both study populations are listed in Table 2-2. Figure 2-2 shows visualization of the shape changes associated with PCs 1-5 for both study populations, constructed via thin-plate-spline warps and

Hausdorff distance color maps. In conjunction with Figure 2-2, Table 3 anatomically describes these shape changes.

Table 2-2: Eigenvalues and variances for Principal Components 1-10, Tanzanian and

European-Derived North American results are shown.

Tanzanian European-Derived North American

Principal % Cumulative Principal % Cumulative Eigenvalues Eigenvalues Component Variance Variance Component Variance Variance 1 0.00046 22.67 22.67 1 0.00051 22.74 22.74 2 0.00043 21.34 44.01 2 0.00041 18.89 40.83 3 0.00030 14.72 58.72 3 0.00037 16.51 57.34 4 0.00015 7.28 66.01 4 0.00018 8.09 65.43 5 0.00013 6.28 72.29 5 0.00012 5.52 70.95 6 0.00009 4.15 76.44 6 0.00009 3.91 74.86 7 0.00006 3.12 79.56 7 0.00008 3.57 78.43 8 0.00006 2.93 82.49 8 0.00007 3.21 81.63 9 0.00004 2.07 84.56 9 0.00004 1.92 83.55 10 0.00003 1.38 85.94 10 0.00004 1.81 85.36

Table 2-3: Anatomical shape changes across Principal Component axes.

Tanzanian European-Derived North American

Shape Change Across Principal Shape Change Across Principal PC PC Component Axis Component Axis Facial Height, Facial Width, Facial Height, Facial Width, Philtrum PC1 PC1 Interorbital Distance Height, Chin Protrusion Facial Height, Maxillary Maxillary Prognathism, Chin PC2 PC2 Prognathism, Facial Width Protrusion, Nasal Projection Retrognathia, Nasal Projection, PC3 Facial Width, Upper Facial Depth PC3 Interorbital Distance Facial Height, Maxillary Nasal Width, Philtrum Length, PC4 PC4 Prognathism, Chin Protrusion Maxillary Retrognathism Facial Height, Nasal Tip Projection, Nasal Tip Projection, Nasal Width, PC5 PC5 Width of Mouth Philtrum Length and Width Nasal Cavity Width, Maxillary Facial Width, Facial Height, Zygomatic PC6 PC6 Prognathism Projection Width of Mouth, Facial Nasal Width, Width of Mouth, PC7 PC7 Prognathism Zygomatic Projection Facial Width, Zygomatic Projection, PC8 Zygomatic Projection PC8 Interorbital Distance Maxillary Prognathism, Upper Width of Mouth, Nasal Width, PC9 PC9 Facial Height, Nostril Width Retrognathia Maxillary Retrusion, Philtrum Width, PC10 Length of Chin, Nasion Positioning PC10 Nasal Width

Figure 2-2: Thin-plate spline warps for PC 1-5 of Tanzanian and North American sample.

Negative and positive PC scores represented. Color map diagrams represent areas of greatest difference (red) and least difference (blue).

2.1.5 Analysis of allometric variance components

To assess the relative proportion of facial shape variation attributable to age, centroid size, height, weight, head circumference, sex and any interactions between them, I analyzed the results from the linear model and Procrustes ANOVA in the Tanzanian cohort, randomizing the order of factors to obtain unbiased estimates of the variances explained by each one (Table 2-4). I found that age, centroid size, height, weight, head circumference and sex all had significant effects (p =

0.001 for all) on facial shape variation. Centroid size captured approximately 4.5% (r2 = 0.045) of the shape variation, while weight captured 1.2% (r2 = 0.012) and head circumference captured approximately 1% (r2 = 0.010). Height (r2 = 0.003) and age (r2 = 0.001) by this model explained <

1% of the total variation. The extended results of this Procrustes ANOVA model are summarized in Table 2-4, and Figure 2-3 shows visualization of the related shape variation. Similar patterns were observed in our North American cohort (Figure 2-4). In my exact-age model, age explained less than 1% of the total variation (r2 = 0.01), while centroid size explained 3.9% (r2 = 0.039).

These values did not change significantly when transforming the exact age to the nearest whole- year (r2 = 0.001) or the nearest 3-year interval (r2 = 0.0006).

Table 2-4: Tanzanian Procrustes ANOVA Model. Relative proportion of variation attributable to several allometric measures.

Measurement r2 p Centroid Size 0.0456 0.001

Weight 0.012 0.001

Head Circumference 0.0108 0.001

Height 0.0032 0.001

Height and Weight Interaction 0.0014 0.001

Age 0.001 0.001

Age and Centroid Size Interaction 0.0009 0.001 Head Circumference and Centroid Size 0.0006 0.004 interaction Height and Age Interaction 0.0005 0.012

Head Circumference, Height and Age 0.0004 0.014 Interaction

Figure 2-3: Thin-plate spline warps of Tanzanian allometric variation. Thin-plate spline warps showing variation across age, centroid size (CS), head circumference (HC), height

(HT), and weight (WT). Negative end of the axis of variation is displayed in the left.

2.1.6 Variation due to Sex and Population

Multiple analyses of variance revealed significant effects for both sex and population. Both factors interact significantly with age (Table 2-5). Figure 2-5 shows 3D morphs and heatmaps corresponding to the sex and population differences in the sample. All effects and interactions are significant (Table 2-5). However, the magnitudes of these effects are fairly small compared to variation within each population. The difference between the two populations explains less than

4% of the total variation in the combined sample. The extensively overlapping variation in facial shape is evident in scatterplots of the first four PCs. Here, the ranges of variation overlap almost entirely with the most separation evident on PC4 which explains 6% of the combined sample variance.

The interaction effects between population and age or population and the size measures, while statistically significant, explain very little variance compared to the main effects (Table 2-

5). The variances explained by the interaction terms are an order of magnitude lower than the main effects. This shows that both the allometric trajectories and sexual dimorphism, while detectable in this large sample, are actually very similar in the two populations.

The interaction effect for age and sex, is significant but also quite small. The regression score plot (Figure 2-5D) shows a reversal in this effect at 14-15, likely reflecting somewhat altered ontogenetic trajectories between the sexes after puberty. The magnitude of this interaction effect, however, is small compared to the overall relationship between age and face shape.

Table 2-5: Procrustes ANOVA with permutation for face shape by population, sex and age.

2 Factor MS r F Z p Age Population 0.407 0.037 184.0 30.9 0.001 Sex 0.080 0.007 36.1 21.8 0.001 Age 0.341 0.031 154.4 30.7 0.001 Population * Sex 0.008 0.001 3.5 3.2 0.003

Population * Age 0.040 0.004 18.1 14.5 0.001 Sex * Age 0.012 0.001 5.3 4.7 0.001 Centroid Size

CS 0.553 0.050 254.3 31.3 0.001 Population * CS 0.008 0.001 3.5 3.1 0.001 Sex * CS 0.009 0.001 4.0 3.6 0.001

Height Height 0.364 0.033 164.8 30.8 0.001 Population * Height 0.017 0.002 7.5 6.5 0.001

Sex * Height 0.015 0.001 6.7 5.8 0.001 Weight Weight 0.292 0.027 131.4 30.4 0.001

Population * Weight 0.017 0.002 7.6 6.4 0.001 Sex * Weight 0.017 0.002 7.8 6.8 0.001 Head Circumference

HC 0.057 0.005 25.0 18.0 0.001 Population * HC 0.007 0.001 3.2 2.8 0.006 Sex * HC 0.007 0.001 2.9 2.6 0.003

Size/Age PC1 Size/Age PC1 0.452 0.041 206.2 31.1 0.001 Population * Size/Age PC1 0.013 0.001 5.9 5.1 0.001 Sex * Size/Age PC1 0.014 0.001 6.5 5.7 0.001

2.1.7 Comparison of linear and polynomial regressions

Comparison of regressions with up to five polynomial terms revealed only small increases in the variance explained with the addition of polynomial terms. This suggests that while all variables potentially have nonlinear relationships to face shape, the vast majority of associated variance is captured using linear regression. More importantly, the relative magnitudes of the variances explained by the size variables and age is not altered by polynomial regression. As performing nonlinear regressions would substantially complicate the analysis and this would also risk overfitting the data, I performed all subsequent analyses based on linear models. Figure 2-6 shows the change in variance explained by polynomial regressions for all variables.

Figure 2-6: Variance explained by linear versus polynomial regressions (two to five

terms) for each size variable and age.

2.1.8 Comparisons of the allometric trajectories associated with different measures of size and age

To compare allometric trajectories across size measures, I removed the effects of population, sex and the interaction between the two using Procrustes ANOVA and centered the resulting residuals on the average of the sex and population averages. Figure 2-7 shows 3D morphs and heatmaps that correspond to the regressions of each variable separately after removing the effects of sex and population. These morphs show that the shape effects associated with all factors except head circumference are fairly similar, with higher values associating with narrower, taller and more prognathic faces while lower values are associated with rounder and less prognathic faces.

To determine how similar the effects of the five factors are, I first removed the effects of all size factors, age and their interactions from the data with multiple multivariate regression and then projected the data on to each factor individually. Figure 2-7A shows the regressions of these projected data onto the first PC of the five measures of size and age. These plots show common allometric component scores (Mitteroecker et al., 2004) plotted against the common size-age axis.

This axis is estimated as PC1 of the size variables and age. The slopes of these regressions are significantly different as determined by homogeneity of slopes test (F = 35, p < 0.01). Further, the shape variation produced by projecting each variable on to the allometry free data is also significantly different (MANOVA, p < 0.001). Figure 2-7B shows the correlation matrix for the size measures of age, from which the common measure of size was obtained. Figure 2-7C shows the resampled variances explained (r2) that correspond to the regressions of each conditional dataset on the common measure of size. These values estimate the proportion of size-related or

allometric variation that corresponds to a standardized amount of variation (1x variance) in each independent variable. As in the variance component estimates above, there results show that centroid size is associated with the most allometric variation, head circumference the least, while weight, height and age fall in between. The resampled distributions in Figure 2-7C show that centroid size is associated with significantly more variation and head circumference with significantly less variation than the other three variables.

To compare the directions of size-related shape variation, I resampled the regression coefficients obtained from separate regressions of the sex and population adjusted data on each variable. Figure 2-9A shows the 3D vectors that correspond to these regressions, scaled to 3x variance for each variable. Figure 2-9B shows the results of a resampling test to compare these vectors. These results show that all of the vectors are significantly different from one another

(p<0.001) for all comparisons except for age and height (p = 0.103).

However, all of the vectors are also significantly more similar than expected by chance (p

< 0.001). Age and height are associated with the most similar shape changes, followed by centroid size and age. These results show that the shape changes associated with age and the various measures of size are closer to being parallel than random. However, most of the comparisons among these measures show shape transformations that differ significantly in both direction and magnitude.

Finally, I estimated the face shape variation related to the covariance of age and the four measures of size and compared this to the total face shape variation explained by all factors and their covariation. After adjusting the data for population and sex, size/age PC1 explains, 3.5% of the variation in facial shape. Adding PC2 brings this to 5%. All size/age PCs together explain

6.1% while a linear model for all size factors and their interactions explains 7.1% of the variation in facial shape. Thus, half of “allometric” variation is shared among size and age measures and half is distinctly associated with individual measures or subsets of measures.

Figure 2-9: Vectors in Size Measure Regressions. A shows the shape vectors that correspond to the regressions of face shape on each size measure and age. Some vectors point inwards from the surface of the face. B shows the distributions of angles among resampled vectors. Since all angles are positive, a mean angle of 0 is not possible. The Null distribution shows the expected distribution of angles when the angles between the vectors for the same regression is resamples. The blue line shows the expected mean when the angles are orthogonal (random) while the red line shows 0 (completely parallel).

Discussion

Allometric variation results from the correlated effects of variation in size and is special because variation in size is so ubiquitous and functionally important. Here, I have analyzed the facial shape correlates of four measures of size as well as age in a large sample of 3D facial images for Tanzanian and North American children of European descent. I report that facial shape is significantly related to height, weight, face size (centroid size), head circumference and age. The patterns of shape variation associated with these variables are broadly similar with the shape correlates of head circumference differing the most. Age and height are associated with the most similar shape variation (Figure 2-7). However, the patterns of shape variation also vary significantly in both direction and magnitude, showing that different measures of size are associated with overlapping but distinct patterns of variation in facial shape. To what extent do age and different size measures converge on a common axis of allometric variation in the face?

Roughly half of the size-related variation is associated with a common allometric component while the remaining half is distinctly associated with particular measures or subsets of measures. Aside for head circumference, the proportion related to a common underlying size effect is likely higher than 50%. However, each measure also adds a component of variation that is distinct from the others.

My results also show that variation in the face is highly structured, with the majority of the shape variation for this high (128) dimensional dataset falling on the first 10 PCs. This is consistent with other studies of morphological variation, including studies of human facial variation

(Bugaighis, Mattick, Tiddeman, & Hobson, 2013; Gonzalez, Perez, & Bernal, 2011; Jonke et al.,

2008; Young et al., 2016). The major axes of covariation involve the facial width, midfacial shape,

as well as orbital shape and orientation. These aspects of facial variation also have significant heritabilities and genetic correlations (Cole et al., 2017).

The two populations studied here differ significantly in facial shape. This is unsurprising, as facial morphology is known to vary geographically or by ancestry (Buck & Viđarsdóttir, 2012;

Hopman, Merks, Suttie, Hennekam, & Hammond, 2014; Klimentidis & Shriver, 2009). This effect is fairly small, however, compared to the variation within each population. This is consistent with the many studies of human variation that show much greater magnitudes of variation within than among populations (Lewontin, 1972; Relethford, 2002). Importantly, the interactions between age and population or measures of size and population of origin is much smaller still. This shows that the age-related shape changes are have similar trajectories in our two populations. Population differences in facial allometry have been reported (Freidline, Gunz, & Hublin, 2015; Viðarsdóttir,

O’Higgins, & Stringer, 2002). The populations in this study are genetically and geographically very different, and furthermore likely experience very different environmental influences, such as nutrition. Yet, their patterns of facial shape allometry are very similar.

These findings also offer clues to the developmental basis for size-related variation in shape. Integration is the tendency for variation in developmental processes to produce covariation in morphological traits (Hallgrimsson et al., 2009). Allometric variation occurs when variation in a process that affects size produces correlated effects on shape. These correlated effects often relate to function, as in the case of scaling relationships (Schmidt-Nielsen, 1984) and may be shaped by selection (Cheverud, 1996). Here, face size emerges as the largest contributor to allometric variation in the face. Facial size co-varies with height and weight but also varies independently of

overall somatic size. Face size interacts with age as well as with height and weight. The head, including the face, grows earlier than stature and smaller individuals have relatively larger heads and faces. I show that the size of the face contributes significantly to face shape in a manner that differs from the effects of overall growth, suggesting that variation in facial growth is a source of allometric variation over and above the effects of overall somatic growth. The developmental basis for variation in head size or face size is not well understood. It is interesting, however, that one of the strongest signals in our genome-wide association study of facial form is for face centroid size and its allometric consequences (Cole et al., 2016). The similarity of the shape vectors associated with age and height suggests an underlying commonality in ontogenetic and static allometry.

(Mitteroecker et al., 2013) report a similar finding in a small cross-sectional sample of facial images of male children (n=19) and adults (n=25). Investigating the precise relationship between ontogenetic and static allometry, however, would require a study with a longitudinal design.

Weight also correlates significantly with facial shape when other size factors are considered. Weight can relate to face shape through adiposity (C. Mayer, Windhager, Schaefer, &

Mitteroecker, 2017; Windhager, Patocka, & Schaefer, 2013). Further, lean body mass may also be related to face shape independently of stature. Face shape correlates with 2D/4D ratio in boys, which is related to testosterone level (Meindl, Windhager, Wallner, & Schaefer, 2012), and lean body mass is associated with facial morphology in adult males (Holzleitner & Perrett, 2016).

Variation in lean mass, skeletal robusticity, or adiposity would translate to covariation between weight and face shape that departs from the allometric pattern associated with other size measures.

Finally, the distinctiveness of the facial shape effects of head circumference suggest a fairly independent role for brain size in determining facial shape. Brain size correlates only weakly with stature and weight in humans (Heymsfield, Gallagher, Mayer, Beetsch, & Pietrobelli, 2007) and brain size relative to cranial base length and width have been shown to influence facial shape in birds, mouse models and humans (Bright, Marugán-Lobón, Cobb, & Rayfield, 2016; Gonzalez,

Kristensen, Morck, Boyd, & Hallgrimsson, 2013; Hallgrimsson, Lieberman, Liu, Ford-

Hutchinson, & Jirik, 2007; Lieberman et al., 2008; Marcucio et al., 2011; Martínez-Abadías et al.,

2012; Marugán-Lobón, Watanabe, & Kawabe, 2016; Parsons et al., 2011). It is not surprising, therefore, that head circumference, influenced largely by brain size, relates to facial shape differently than other measures of size. Size-related variation accounts for less than 5% of overall variation in face shape in my study. This low value is consistent with other studies of human facial variation (Gonzalez et al., 2011; Mydlová, Dupej, Koudelová, & Velemínská, 2015; Velemínská et al., 2012) and it contrasts with other primates species in which the allometric correlates of head size tend to account for much larger greater proportions of shape variance (Ito, Nishimura, &

Takai, 2011; Lieberman et al., 2007). This may reflect a tendency for co-variances among craniofacial traits to be lower in humans overall as has been shown in large comparative study of craniofacial integration (Porto et al., 2009). This may also relate to a tendency for individually features of human facial shape to appear quite early in ontogeny (Viđarsdóttir & O’Higgins, 2003).

My finding that allometric variation in the face is complex is important for two reasons.

First, allometry is a central concept in the study of evolution and development. Growth changes proportions and shape as well as size, and variation in size influences most morphological traits.

Body size varies among past and present human populations for both genetic and environmental

reasons, producing correlated changes in facial shape. To understand size-related variation in human facial shape, it is important to know whether allometry is a single axis of integration or whether the relationship between shape and size is more complex. Second, many genetic syndromes appear to influence the shape of the face (Gorlin, Cohen, & Hennekam, 2001). Many such syndromes also influence stature, body mass and brain size. For these reasons, it can be difficult to disentangle facial shape effects that are produced as a side-effect of the alteration in growth from those that result from other, more distinctive, perturbations to development.

Conclusion

(Klingenberg, 2016) distinguishes two concepts of allometry. In the Gould-Mosimann

(Gould, 1966; Mosimann, 1958) approach, allometry is the covariation of shape with size, while the Huxley-Jolicoeur approach (Huxley, 1932; Jolicoeur, 1963) approach defines allometry as the covariation among traits that contain information about size (Klingenberg, 1998b). In the former approach, an a priori assumption is made about what constitutes size while in the latter, size is assumed to be a single latent variable that can be teased out of covariation patterns. My finding that allometry is complex has implications for both approaches, and underscore the need to fully explore how size-parameters influence morphology in the context of questions where allometry is either a factor of interest or a factor that must be quantified and controlled in an analysis.

The structure of phenotypic variation is determined by multiple developmental processes acting at different times, scales and locations in development (Hallgrimsson et al., 2009). Here, I have shown that allometric variation is complex, determined by variation in incompletely

overlapping mechanisms that contribute to growth. This is important for understanding the genetic and developmental basis for allometry. Due to the importance of size variation in most populations and the fact that allometry is generally the largest component of variation for any morphological trait, determining how size produces variation in shape is a question of central importance for evolutionary morphology and for understanding the structure of morphological variation in humans.

Acknowledgements

Supported by NIH-NIDCR (1U01DE020054) to RS and BH, NSERC Grant #238992-12 to BH and U01DE020078 to SW, and the University of Calgary (UIRG to MM, KL and BH).

Many people participated in various aspects of the Facebase Tanzania project, of which this study is part. Megan Wright, Maria Finnsdóttir, and Kimia Ghavani, contribute to reconstruction of the facial image data. Sariko Matari and Saleh Seleman Mganzil, Mitzi Murray, John Humphreys,

Kai Lukowiak, Kris, Kannon, Hayley Britz, Rebecca Green, Kim McKenney, Kimani Leyaro,

Margaret Kaisoe, Diana Dills, Nicola Hahn, assisted in the field. We also thank the teachers in the many schools in which this work was conducted and, most importantly, the children who participated as well as their parents. Primary data were deposited at FaceBase https://www.facebase.org/, accession: FB00000667.01.

Chapter 3: Variation in Non-Syndromic Cleft Lip and/or Palate Patients

Abstract

Non-syndromic cleft lip and/or palate (nsCL/P) is a developmental disease, occurring in isolation of other defects. nsCL/P is thought to behave as a complex phenotype resulting from substantial genetic interactions and environmental risk factors. Because of the extensive variation seen in the condition, a furthered understanding of the structure of the phenotypic variation of nsCL/P may help to improve clinical assessment and treatment. This work converges on two central hypotheses; that facial shape variation due to size-related factors is significant in nsCL/P patients; and that the growth trajectories of nsCL/P subtypes will significantly differ. Here, I have shown that in nsCL/P patients, variation in facial shape due to facial size is significant. The proportion of variation due to facial size is lower in nsCL/P patients than that in controls. Additionally, shape variation due to severity (complete vs. incomplete clefting) and laterality of clefting is significantly associated with allometric shape variation. I also have examined differences in covariance structure of the diagnosis classification groups, and found that bilateral and right-sided clefts are more similar in covariance structure than that of right and left-sided clefts. This suggests that unilateral clefts are not simply mirrored images of each other, and rather they likely represent distinct etiologies. These results are important when considering the developmental determinants of integration — a furthered understanding of the link between covariation patterns and the developmental processes that generate integration in nsCL/P may help improve clinical assessment and treatment.

Introduction

Phenotypic variation is a prominent feature common to many developmental disorders of the craniofacial complex. However, the mechanisms responsible for this variation are largely unknown. Non-syndromic cleft lip and/or cleft palate (nsCL/P) is a classic example of such a disease, occurring in isolation from other phenotypic anomalies (S. Banerjee, Banerjee, Radke, &

Mundhe, 2011; Nakasima & Ichinose, 1983). nsCL/P is associated with aspects of orofacial shape in humans; morphometric studies have shown that relatives of nsCL/P patients tend to have relatively wide but short mid-faces, and obligate normal and carrier individuals in a large nsCL/P kindred cohort differ from non-carriers in these same features (Weinberg et al., 2008). There is also considerable evidence that non-syndromic CL/P (nsCL/P) cases represent extremes of the normal range of phenotypic variation in orofacial size and shape (Weinberg et al., 2008). nsCL/P as a condition presents in varying degrees of severity, and as such, displays considerable phenotypic variation. Orofacial clefts, principally cleft lip (CL), cleft palate (CP), and cleft lip and palate (CL/P) are among the most common major birth defects, occurring in ~1/500 to 1/2500 live births in various populations around the world (Gorlin et al., 2001). Non-syndromic CL/P (nsCL/P) occurs in approximately 70% of all orofacial clefts, and is thought to behave as a complex trait resulting from multiple genetic interactions and various environmental risk factors (Carinci,

Scapoli, Palmieri, Zollino, & Pezzetti, 2007; Nakasima & Ichinose, 1983). Because of etiological complexity, elucidating the structure of phenotypic variation seen in nsCL/P becomes critical in clinical assessment and treatment. For example, reparative surgeries can become more targeted to the specific nsCL/P phenotype by accounting for the growth of the face, relative to the clinical classification of the cleft and/or the severity. By establishing separate growth trajectories for nsCL/P subtypes, the optimization of the surgical procedure may improve the aesthetic result, the

functional capacity of the oronasal cavity, and improve the mental health and psychological condition of the patient (R. L. B. de Oliveira, de Santana Santos, de Almeida Teixeira, Martins-

Filho, & da Silva, 2015; Pedersen, Wehby, Murray, & Christensen, 2016; Roberts, 2014; Q. Wang,

Shang, & Fang, 2012).

The development of the face is an extraordinarily complex process with substantial variation occurring during development (Hallgrimsson et al., 2009). This phenotypic variation is likely derived through highly complex genomic variation involving a very large number of genes and alleles. In addition, genomic factors and gene-gene interactions, combined with an overall high tolerance for subtle differences contribute to variation in the development of an aesthetically

“normal” human face. Therefore, it seems likely that genetic susceptibility to orofacial clefts reflects an additive genetic liability, attributable to the many genes involved in normal orofacial morphogenesis, particularly genes responsible for facial prominence outgrowth and fusion.

Orofacial clefts and midfacial abnormalities arise during fetal orofacial development. These phenotypic abnormalities may be attributed to the manner in which the medial and lateral nasal processes interact with the maxillary prominences. When phenotypic anomalies such as CL/P occur, the perturbation of the normal developmental process is hypothesized to occur through prominence fusion obstruction (Sözen et al., 2001), or incorrect timing of prominence fusion

(Fraser, 1970). These perturbations may be the result of increased cellular apoptosis, disruption of cell signaling, or increased cellular density (Boughner et al., 2008; Young, Wat, Diewert, Browder,

& Hallgrimsson, 2007).

The morphological integration of the vertebrate skull is well-established (Cheverud, 1982;

Hallgrimsson et al., 2004; 2009; Porto et al., 2009). Recently, there have been studies that examine cranial morphological integration in patients with craniofacial anomalies (Richtsmeier & DeLeon,

2009). However, these studies emphasize internal osseous integration patterns, primarily focusing on the cranial base and its integration with other cranial structures. While osteological procedures are critical in cleft interventionist care, the majority of surgical procedures throughout an nsCL/P individual’s lifetime focus on soft tissue manipulations and cosmetic corrections. Procedures are often repeated several times throughout an individual’s life, due to the growth the face. As the postnatal face grows and develops, the soft tissue manipulations and cosmetic corrections that were previously performed can become aesthetically and functionally compromised, necessitating further corrections. Through soft tissue growth analyses, the growth trajectories of nsCL/P faces can be quantified and may help to inform surgical care and practice.

Here, I have examined patterns of morphological integration of the facial soft tissue, utilizing 3-D photography to capture 3-D surface images. Furthermore, I use a three-factor classification system of nsCL/P phenotypic presentations (K. H. Wang et al., 2014), which allows more precise examination of the variation seen between nsCL/P patients. This work converges on two central hypotheses; that facial shape variation due to allometry is significant in nsCL/P patients; and that the growth trajectories of nsCL/P subtypes will significantly differ. This work examines and compares facial structure among nsCL/P subtypes. I also show that variation in

facial shape due to age and size is significant and that the shape variation associated with these allometric factors is complex, with overlapping effects between facial shape and size. Furthermore,

I show that the growth trajectories of facial shape in nsCL/P differ among laterality and severity subtypes. Lastly, I show that facial shape integration significantly differs among various cleft classifications and subtypes.

Materials and Methods

3.1.1 Study Subject Demographics

This study received ethical approval, granted by the University of Calgary (CHREB

21741), the University of Colorado (09-0731), and the University of Pittsburgh (#PRO09060553 and #RB0405013). Prior to participation in the study, informed written consent was obtained from the parents and/or legal guardians of all study participants. nsCL/P patients were collected at the

Alberta Children’s Hospital Cleft Palate Clinic, in Calgary, Alberta from 2013-2016. A total of

1008 individuals participated in the study; 99 nsCL/P patients and 909 healthy control subjects.

The healthy control subjects, present without any birth defects or family history of craniofacial disorders, were obtained from the University of Colorado and the University of Pittsburgh.

The cohort of 99 nsCL/P patients of self-identified Caucasian origin (29 females, 70 males) were 0.1 to 33.6 years of age (grand mean = 11.5 ± 5.5 years; females: 11.8 ± 4.6 years; males:

11.4 ± 5.9 years). In this cohort, 86 patients had cleft lip and palate, while 13 patients had only cleft lip. This patient group contains patients with left-sided, right-sided, and bilateral clefting (left

= 43, right = 30, bilateral = 26); and with both complete and incomplete clefts (complete = 74,

incomplete = 25). Due to the small size of the sample, only three discrete categorical groups were used in clinical classification (Cleft Type, Laterality, Severity); further sub-groupings of these categories were not defined due to low sample size. Please see Table 3-4 (Supplementary Table 1) for complete counts of categorical patient groupings.

For this study, cleft type was categorized as CL/P or CL. Cleft lip and palate was defined as a cleft that affects the lip and extends through the palate to some degree; whereas a cleft lip was defined as a cleft that affects the lip and may also include an alveolar notch, or a cleft in the bone that surrounds the teeth anterior to the palate. Any individuals with an alveolar notch that also had any degree of cleft palate were classified as CL/P. The laterality of the cleft was defined as the anatomical position in which the cleft occurred on the face; either bilateral, left or right. Finally, the third classification grouping that was employed was severity of the cleft. A complete cleft lip extends directly though the nasal floor, while an incomplete cleft lip has a degree of intact tissue in the nasal floor. A complete cleft palate involves a cleft through both the primary and secondary palates, while an incomplete cleft palate is isolated to the secondary palate only (Kosowski,

Weathers, Wolfswinkel, & Ridgway, 2013).

The cohort of 909 healthy control participants (456 female, 453 male) were 3.0 to 18.0 years of age (grand mean = 9.5 ± 4.4 years; females: 9.8 ± 4.5 years; males: 9.3 ± 4.2 years). This sample only includes children who self-identify as “white” using the NIH “Racial and Ethnic

Categories” for recruitment and consenting of research subjects

(https://grants.nih.gov/grants/guide/notice-files/NOT-OD-15-089.html). This cohort is not representative of “white” individuals as a discrete ethnic group, and these discrete groups are not

assumed to be definable in a biological sense (Edgar & Hunley, 2009; Hunley et al., 2009). This cohort represents a sample population where within-group heterogeneity (due to ancestry) has been minimized.

3.1.2 Three-Dimensional Imaging of Subjects and Automated 3-D Landmarking

Study participants were imaged using a non-invasive 3-D imager, following an established protocol (Weinberg & Kolar, 2005). However, due to the relative long length of the collection period (2013-2016), the multiple international collection sites, and the speed of current technological advances, two separate 3-D imagers were used. The Inspeck© 3-D Gemini System can acquire a 3-D surface map of the head and face in ~0.4 seconds, with a 640 x 480 mm field of view. In the two study cohorts, 71 nsCL/P patients and 252 healthy control subjects were imaged using the Gemini system. The Gemini camera functions by taking four photographs of an individual using standardized head positions: frontal, right, left and inferior chin (Figure 3-1).

Subjects were imaged with the eyes and mouth closed, while maintaining a distance of 1 meter to the imager. After the four images were spliced together using InSpeck© Photo Crystal Software

(PCS), InSpeck© Editing and Merging Software (EM) software was used to produce a texture map of the 3-D surface map.

The 3dMD© 3dMDtrio System uses an ultra-fast capture speed of ~1.5 milliseconds to capture a 200° ear-to-ear surface capture. In this sample, 28 nsCL/P patients and 657 healthy control subjects were imaged using the 3dMDtrio system. The 3dMDtrio system utilizes multiple

imagers, positioned directly in front of, and to both sides, of the study participant. These imagers work together to capture a 3-D image of the head and neck. The 3dMDtrio software automatically renders the image data into a seamless 3-D surface model. Because of the extremely high quality of the 3dMDtrio images, extraneous details such as hair or the back of the chair were captured.

Additional cropping of the 3dMDtrio images was performed in Meshlab (Cignoni, Corsini, &

Ranzuglia, 2008), to remove any extraneous image information. All 1008 3-D facial surfaces were landmarked using a novel landmarking method (Figure 3-2, Table 3-1) (Bannister et al., 2017).

Figure 3-1: Anatomical positions of the head used in three-dimensional photography capture

with the Gemini. Top row: frontal, half right, right. Bottom row: inferior chin, half left, left.

Figure 3-2: Anatomical landmarks (29) used on digitally rendered 3-D surface. Frontal and

Lateral views shown. Corresponds to Table 3-1.

Table 3-1: Three-dimensional anatomical landmark descriptions. Landmark name, abbreviation and anatomical description. As referenced from http://www.facebase.org.

3D Landmark Abbrev- Definition Name iation Nasion n Midline point in where the frontal and nasal bones contact (nasofrontal suture). Corresponds to the underlying bony landmark. Pronasale prn Midline point marking the maximum protrusion of the nasal tip. Subnasale sn Midline point marking the junction between the inferior border of the nasal septum and the cutaneous upper lip. It is the apex of the nasolabial angle. Labiale Superius ls Midline point of the vermilion border of the upper lip, at the base of the philtrum. Stomion sto Midpoint of the labial fissure. Labiale Inferius li Midline point of the vermilion border of the lower lip. Sublabiale sl Midpoint along the inferior margin of the cutaneous lower lip (labiomental sulcus). Gnathion gn Midline point on the inferior border of the mandible. Corresponds to the underlying bony landmark. Endocanthion en_r Apex of the angle formed at the inner corner of the palpebral (Right) fissure where the upper and lower eyelids meet. Endocanthion en_l Same as above (Left) Exocanthion ex_r Apex of the angle formed at the outer corner of the palpebral (Right) fissure where the upper and lower eyelids meet. Exocanthion (Left) ex_l Same as above Alare (Right) al_r Most lateral point on the nasal ala. Alare (Left) al_l Same as above Alar Curvature ac_r Most posterolateral point on the alar cartilage, located within Point (Right) the crease formed by the union of the alar cartilage and the skin of the cheek. Alar Curvature ac_l Same as above Point (Left) Subalare (Right) sbal_r Point located at the lower margin of the nasal ala, where the cartilage insterts in the cutaneous upper lip. Subalare (Left) sbal_l Same as above Crista Philtri cph_r Point marking the lateral crest of the philtrum at the (Right) vermilion border of the upper lip Crista Philtri (Left) cph_l Same as above

Chelion (Right) ch_r Point marking the lateral extent of the labial fissure. Chelion (Left) ch_l Same as above Tragion (Right) t_r Point marking the notch at the superior margin of the tragus, where the cartilage meets the skin of the face. Tragion (Left) t_l Same as above Superior Alar supa_r Most superior portion of alar groove. Groove(Right) Superior Alar supa_l Same as above Groove (Left) Zygion (Right) z_r Most prominent portion of zygomatic arch Zygion (Left) Z_l Same as above Pogonion P Most prominent portion of chin, anatomical pogonion.

3.1.2 Morphometric Analysis

All analyses were performed using R Software (R Development Core Team, 2015). To be able to compare unilateral versus bilateral clefts, all right-sided nsCL/P patients were mirrored to have a “proxy” left-sided cleft, by inverting the sign of the x-coordinate bilateral landmarks

(Manyama et al., 2011). This procedure was utilized so that the resultant shape variation was not simply due to laterality of the cleft. It also allowed an examination of the biology of left and right sided clefts; to explore whether left and right sided clefts are simply mirrored images of each other, or if there is an underlying phenotypic difference in these unilateral clefts. The imager used in the collection had a significant effect on the total variation in the dataset (r2=0.07, p<0.001). To account for the two imagers used, the Inspeck©Gemini and 3dMD© 3dMDtrio, a statistical correction was applied using Procrustes ANOVA with permutation testing, via the procD.lm function of the geomorph R package (Adams & Otárola Castillo, 2013). Further calculations were based upon the residuals of this linear model. The variation due to camera and sex was removed

by centering the model’s residuals on the sample’s grand mean. Following this, residuals were used to build a Procrustes ANOVA model of allometric variation; a model to quantify relative proportions of shape variation attributable to age, centroid size, cleft type, laterality, and severity.

Since procD.lm uses type 1 error calculations, we alternated the position of the last variable in the model while performing sequential regressions. This allowed a more accurate estimate of percentage variance attributable to each variable. Shape variation morphs and heatmaps were created and visualized in R, using Morpho (S. Schlager et al., 2016). This model was used to examine allometric facial shape variation, and to determine resultant facial shapes and growth trajectories in nsCL/P subtypes. Following this, separate models were built for the nsCL/P and control groups. This allowed an estimation of the different allometries with respect to cleft subtype, laterality and severity. With this procedure, mean shape morphs and shape variation morphs along growth trajectories were constructed in R, using Morpho (S. Schlager et al., 2016). All morphs were constructed using the asymmetric component of variation, following the laterality x- coordinate reflection as described above. Homogeneity of slope tests were performed using the procD.allometry function for cleft subtype, laterality and severity, combined with age and centroid size.

Integration, or the tendency of a developmental system to produce covariation

(Hallgrimsson et al., 2009), can be relatively measured as the scaled variance of the eigenvalues

(Pavlicev, Cheverud, & Wagner, 2009; Wagner, 1984). To examine the dimensionality of the covariance structure of our two patient cohorts, we calculated the scaled variance of the eigenvalues and the trace of the matrix, of several different group iterations: Cleft/Control,

Left/Right/Bilateral, Incomplete/Complete, Cleft Lip/Cleft Palate. These groups were analyzed

based upon the clinical classification groups as described previously. To perform these calculations, first a covariance matrix was obtained using the residual coordinates from the

Procrustes ANOVA model. Then, each eigenvalue was divided by the sum of the eigenvalues; and then its variance calculated.

The covariance structure was examined by calculating matrix correlations between the covariance matrices of several different clinical criterion groups (Status: Cleft, Control; Laterality:

Bilateral, Left, Right; Severity: Complete, Incomplete). Following (Hallgrimsson et al., 2009), to examine the covariance structure, covariance matrices were calculated using camera, age, and sex corrected residual data, for each criterion group. The covariance matrices were then used in a matrix correlation calculation. The matrix correlation result (ranging from 0 to 1.0) describes the mathematical correlation between the covariance matrices; and thus reveals how similar or dissimilar the covariance structure is between the groups being examined. Additionally, following

(Jamniczky & Hallgrimsson, 2009), “covariance distances” were calculated from each comparison between covariance matrices, and then regressed onto the Procrustes distance between the mean shape of each group/subtype. This displays the relationship between morphological distance

(Procrustes) and covariance distance for each subtype comparison.

Results

3.1.3 Facial Shape Differences Between Groups

Individuals with nsCL/P differed significantly in facial shape from controls (r2=0.09, p<0.001). Average shapes of the whole sample, and of the nsCL/P and control groups were calculated via the mshape function in the Momocs R Package (Bonhomme, Picq, Gaucherel, &

Claude, 2014). Figure 3-3 shows the mean shapes of these groups. Furthermore, Figure 3-4 shows the heatmap visualization of the shape differences between the nsCL/P and control mean shapes.

The largest shape difference observed between the groups is in the philtrum and nasal tip. In nsCL/P individuals, the philtrum tends to be longer and flatter; while the nasal tip lacks the anterior projection as seen in the control group (Figure 3-3). Furthermore, the nsCL/P group have apparent shape differences in nasal width and facial width – with both measures tending to be wider in these individuals when compared against control subjects.

Figure 3-3: 3-D Morphs (thin plate spline warps) of the average facial shapes of nsCL/P and controls. Also shown is the average grand mean facial shape of the entire sample.

3.1.4 Allometric Variation

To estimate the proportion of facial shape variation attributable to centroid size and age, we used a linear model using Procrustes ANOVA across the entire sample, and then also separately for both study cohorts. In the combined sample, two separate models using only centroid size or only age were built, and showed both measures capturing significant proportions of shape variation

(centroid size, r2=0.07, p<0.0001; age, r2=0.04, p<0.0001), totalling over 10% of the total variation. Performing a combined factor Procrustes ANOVA, we alternated the position of the last variable in sequential regressions to accurately estimate the percentage of variance attributable to each variable (Table 3-2; ‘Combined Sample’). To visualize the shape variation associated with these allometric factors, the mean shape was calculated from the camera and sex-corrected residuals for each within-group (Cleft, Control) measure. The mean shape was morphed to +/- 2

Standard Deviations along the scaled vectors. 3-D morphs were constructed to visualize this variation (Figure 3-5). Separate models were then built using cleft classifications (Table 3-2; ‘Cleft

Sample’) . In the cleft model, centroid size and severity captured significant proportions of shape variation (centroid size, r2=0.05, p<0.0001; severity, r2=0.02, p=0.01). Shape variation due to age, and all age interaction terms, were not significant in the cleft sample. Additionally, variable interactions between centroid size and laterality (r2=0.03, p=0.02), and centroid size and severity

(r2=0.02, p=0.01) showed significant effects on facial shape. To visualize the regression models with the interaction term between centroid size and age with cleft classifications, regression plots were constructed in R using the procD.allometry function in Geomorph (Figure 3-6) (Adams &

Otárola Castillo, 2013). For both of these groups (centroid size with laterality, and centroid size with severity), the null hypothesis of parallel slopes was rejected based on a significance criterion

of alpha =0.05 (centroid size with laterality, r2=0.06, p=0.01; centroid size with severity, r2=0.05,

p=0.01).

Table 3-2: Procrustes ANOVA model results. Combined Sample, nsCL/P Sample, and

Control Sample. Significant results are highlighted.

Type I (Sequential) Sums of Squares and Cross-products Randomized Residual Permutation Procedure Used 1001 Permutations

Factor Df SS MS r2 F Z p Procrustes ANOVA Combined Sample age 1 0.7769 0.77687 0.041999 48.0611 24.2487 0.000999 *** cs 1 1.3398 1.28762 0.072434 82.8887 28.3926 0.000999 *** age:cs 1 0.0362 0.03617 0.001956 2.1633 1.8181 0.057942 Procrustes ANOVA Cleft Sample cleft type 2 0.02107 0.010534 0.014283 0.7728 0.7033 0.624376 age 1 0.01961 0.019614 0.013297 1.4389 1.2250 0.142857 cs 1 0.08086 0.080858 0.054818 5.9317 5.3364 0.000999 *** laterality 2 0.03922 0.019611 0.028008 1.5364 1.4483 0.058941 severity 1 0.02961 0.029607 0.021142 2.2817 1.0550 0.018981*** age:cs 1 0.02246 0.022457 0.015225 1.6474 1.4609 0.096903 cs:cleft type 2 0.02442 0.012211 0.016557 0.8958 0.7615 0.373626 cs:laterality 2 0.05000 0.025000 0.035706 1.9587 1.8447 0.023976 *** cs:severity 1 0.02961 0.029607 0.021142 2.2799 2.0550 0.018981*** age:cleft type 2 0.01988 0.009939 0.014195 0.7247 0.6681 0.710290 age:laterality 2 0.04033 0.020164 0.028799 1.4937 1.4390 0.060939 age:severity 1 0.01396 0.013964 0.009972 1.0761 0.9794 0.316683 Procrustes ANOVA Control Sample age 1 0.7926 0.79262 0.048069 48.9536 24.4464 0.000999 *** cs 1 1.4495 1.44955 0.087909 89.5266 28.4479 0.000999 *** age:cs 1 0.0389 0.03890 0.002359 2.4024 1.9304 0.051948

Figure 3-6: Regression models for age and centroid size with cleft classifications. A: Control/Cleft,

B: Cleft Type, C: Laterality, D: Severity. (*** denotes significant interaction term between centroid size and classification factor).

To visualize the mean shapes of these significant classification groups, mean coordinates were calculated, and morphs constructed, for all factors of laterality (Figure 3-7) and severity

(Figure 3-8). These mean shape morphs inform us of the generalized shape differences between each group, but are not a statistical comparison point. In Figure 3-7, it appears that left-sided and bilateral clefts have a more severe maxillary phenotype than that of right-sided clefts; the maxilla displays retrusion and the philtrum is wider and flatter. Right-sided clefts appear to have a more anteriorly positioned chin when compared to left-sided clefts. In Figure 3-8, it appears that complete clefts generally are also more severe in the maxillary phenotype; with complete clefts showing a shorter and more compact maxilla, displaying retrusion.

Figure 3-7: Morphed mean shapes for laterality classification. Combined group, left, right and bilateral morphs shown.

Figure 3-8: Morphed mean shapes for severity classification. Combined group, incomplete and complete morphs shown.

Following the regression models (Figure 3-6), regression trajectory morphs were calculated and constructed for each significant classification group, using Geomorph (Adams &

Otárola Castillo, 2013). These morphs display the trajectory of facial growth for each group. While the trajectories are similar in terms of pattern, there are distinct differences in how each group progresses through their unique growth curve. Unilateral clefts appear to produce a narrower facial shape at the end of the growth curve, when compared to bilateral clefts. Left-sided clefts and bilateral clefts produce a wider philtrum, and a retruded maxilla, which in turn alters the positioning of the nose and nasal cavities. The right-sided cleft growth curve is unique; these individuals show an initially wide and flattened facial profile that grows into a rounded profile, narrower width, and increases in vertical length. The left-sided and bilateral growth curves show an initially rounded midface and zygomatic region that grow to result in a flattened midface displaying midfacial hypoplasia. While all of these growth curves show an initially wide face, the unilateral clefts grow to produce a narrower face. In bilateral cleft growth, there doesn’t appear to be as stark of a reduction in facial width (Figure 3-9 and Figure 3-10). In severity of cleft (Figure

3-11 and Figure 3-12), the incomplete cleft growth trajectory is quite different than that of the complete cleft trajectory. Incomplete clefts show initial midfacial hypoplasia with widened midfacial features, and progress through the growth curve to gradually decrease facial width and philtrum width. This results in an extremely narrow facial phenotype, with a narrowed mouth and wide and flattened philtrum with thin upper lip. The complete cleft trajectory is less drastic, initially showing a wider face that gradually narrows as it grows, with minor maxillary retrusion.

The philtrum appears to shorten in length as the trajectory progresses.

Figure 3-9: Growth trajectory morphs for centroid size (CS), by laterality. Frontal view.

Trajectory progressions shown (from minimum to maximum centroid size): Left, right, and bilateral clefts. Also Shown is entire combined nsCL/P group.

Figure 3-10: Growth trajectory morphs for centroid size (CS), by laterality. Lateral view.

Trajectory progressions shown (from minimum to maximum centroid size): Left, right, and bilateral clefts. Also Shown is entire combined nsCL/P group.

Figure 3-11: Growth trajectory morphs for centroid size (CS), by severity. Frontal view.

Trajectory progressions shown (from minimum to maximum centroid size): Incomplete and complete clefts. Also shown is entire combined nsCL/P group.

Figure 3-12: Growth trajectory morphs for centroid size (CS), by severity. Lateral view.

Trajectory progressions shown (from minimum to maximum centroid size): Incomplete and complete clefts. Also shown is entire combined nsCL/P group.

3.1.5 Morphological Integration

I investigated the degree of integratedness by looking at the scaled variance of the eigenvalues (SVE), and the multivariate variances for various diagnostic classifications. Namely, by laterality (bilateral, left, right); severity (complete, incomplete) and finally by patient group

(cleft, control). Bilateral clefts showed greater variation in SVE than in left or right sided clefts.

Additionally, incomplete clefts showed greater variation in SVE (and thus, increased integration) than complete clefts, and nsCL/P subjects showed greater variation in SVE than control subjects.

The SVE difference between bilateral clefts and left and right-sided clefts, and also between nsCL/P and control subjects, is significant when resampling SVE 1000 times (p<0.05). The multivariate variances are significant between bilateral clefts and left and right-sided clefts, and also between incomplete and complete clefts. Results are summarized in Figure 3-13.

I also compared the covariance structure between several groups by calculating matrix correlations (Table 3-4). These results demonstrate the vast difference of the covariance structure in the cleft patient group when compared to that of healthy controls (matrix correlation = 0.285).

The correlation between the covariance matrices of the bilateral and left-sided cleft groups (matrix correlation = 0.422) is significantly different from that of the bilateral and right-sided cleft groups

(matrix correlation = 0.607). Interestingly, the correlation between the bilateral and right-sided cleft groups is larger than the correlation between left and right-sided clefts (matrix correlation =

0.562), even following the cleft-side-mirroring as described in statistical methods. I also calculated

‘Covariance Distance’ between these groups (Jamniczky & Hallgrimsson, 2009), and regressed it onto Procrustes distance. This shows the relationship between morphological and covariance distance (Figure 3-14). Most obviously, the cleft and control group comparison showed both the greatest Procrustes and covariance distance. The laterality and severity comparisons show varying results, with the complete and incomplete cleft group comparison showing low morphological and covariance distances.

Table 3-3: Matrix correlation between covariance matrices. Separate covariance matrices derived for bilateral, left, right, complete, incomplete, cleft and control groups.

Covariance Matrices Used for Factor Matrix Correlation Correlation

Bilateral – Left 0.422

Laterality Bilateral – Right 0.607

Left - Right 0.562

Severity Complete – Incomplete 0.730

Status Cleft – Control 0.285

Figure 3-14: Covariance Distance vs Procrustes Distance. Groups shown include all laterality comparisons, severity, and cleft/control.

Discussion

Here, I have used a three-factor classification system of non-syndromic cleft lip and/or palate (nsCL/P) patients to precisely examine the morphological variation, and morphological integration, present in a cohort of nsCL/P patients. I have shown that variation in facial shape due to age and size are significant in both nsCL/P patients and also control individuals. These allometric variables have overlapping effects on the variation in facial shape. I also show that facial shape integration, and the strength of integration, differs significantly among various cleft classifications and subtypes.

My results show that there is a significant shape difference between nsCL/P patients and healthy controls. Given the massive defect, it is not surprising that nsCL/P patients differ significantly from controls in facial shape. Even following surgery, most nsCL/P patients face a battle of social issues due in part to their outwardly different facial appearance (Stock, Feragen, &

Rumsey, 2015). The primary axis of variation shows morphological differences in the philtrum length and projection, and the anterior projection of the nasal tip. Also, separating the average shape of nsCL/P and control groups is the width of the nasal ala, and to a lesser extent the total width of the face. I have shown that in nsCL/P patients, variation due to centroid size is significant, while age has a non-significant effect. However, in control patients, variation due to both age and centroid size are significant, and in larger proportions. In addition to centroid size, the shape variation due to severity of a cleft is also significant. The severity of the cleft, being complete or incomplete, is associated with allometric shape variation. Also, the laterality of the cleft is significantly associated with allometric shape variation. As previously reported, there is no correlation between laterality and severity (Sivertsen et al., 2008).

Previous studies have shown correlations between cleft laterality and various other measures. Degree of cleft severity has recently been shown to be associated with IRF6 mutations, a protein coding gene important in early development (Kerameddin, Namipashaki, Ebrahimi, &

Ansari-Pour, 2015). It has also been shown that nsCL/P laterality may be under genetic control

(Farina et al., 2002). However, despite the vast cleft literature, this present study is the first to report that nsCL/P laterality is significantly associated with allometric shape variation.

Additionally, I report that cleft severity is also significantly associated with allometric shape variation.

Growth trajectories of this allometric shape variation show how facial growth occurs differently in different cleft subtypes. I have demonstrated that while the patterning of the growth trajectories is similar, left-sided and bilateral clefts initially progress through growth trajectories that produce a phenotype indicative of maxillary retrusion and midfacial hypoplasia; while right- sided clefts initially begin with these features, and progress through growth to reach a more prominent maxillary and zygomatic appearance. Similarly, complete and incomplete cleft trajectories begin with rounded and wide facial phenotypes. However, as growth progresses, incomplete clefts result in a narrower face, narrower features, and facial flattening with maxillary retrusion. Complete clefts progress through a trajectory that shows less facial narrowing, and overall maintains a mesognathic relationship. These findings suggest that left-sided and bilateral clefts have a similar growth pattern and trajectory, and that growth in right-sided clefts may follow an opposite pattern. In addition, growth in incomplete clefts appear to produce a more severe phenotype than growth in complete clefts. This could suggest that individuals with incomplete clefts may have underlying issues with tissue or bone growth, or perhaps the increased integration

seen in nsCL/P with incomplete clefts produces a more severe phenotype. These trajectories may be useful in surgical planning and care; by understanding the growth trajectory that an individual’s face will undergo, surgical adjustments to account for growth may be made.

These findings also estimate the degree of integratedness of the face, as measured by the scaled variance of the eigenvalues (Pavlicev et al., 2009; Wagner, 1984). Previous studies have shown that inbred mutant strains of mice have increased eigenvalue variance when compared to wild type strains (Hallgrimsson et al., 2009). When grouped by laterality, the most severe phenotype (bilateral nsCL/P) shows the highest eigenvalue variance (Figure 3-13A), or the strongest integrated phenotype. Incomplete nsCL/P patients also show an increased eigenvalue variance than that of complete nsCL/P, suggesting that as the variance in facial shape due to growth and size increase, a correlated response is occurring in areas of the face not overtly affected by the cleft (Figure 3-13B) (Hallgrimsson et al., 2009). The developmental determinants of this covariance structure must be fully studied to be able to interpret the significance of these covariation patterns. By studying the developmental determinants of integration, a furthered understanding of the link between covariation patterns and the developmental processes that generate integration in nsCL/P patients can be developed.

Previous studies have shown that the bilateral nsCL/P patient does have a more integrated facial skeleton (Starbuck, Ghoneima, & Kula, 2015). My study supports those findings and extends their integration hypothesis to the soft tissue of the face. As previously stated, I have found that bilateral nsCL/P patients have an increase in their eigenvalue variance, suggesting morphological integration. Other previous studies have shown that the magnitude and pattern of morphological

integration of soft tissue is similar in pre and post-operative unilateral nsCL/P individuals; that the surgeries don’t appear to either increase or decrease the magnitude of integration (Richtsmeier &

DeLeon, 2009). This finding supports our usage of both pre and post-op nsCL/P individuals in our investigation.

The covariance structure of the various groups was also assessed by examining the correlation between the respective covariance matrices. Interestingly, the matrix correlations between the bilateral and left nsCL/P subtypes, and between the bilateral and right-side transformed to left nsCL/P subtypes are different. There is a higher matrix correlation between bilateral and right-side transformed clefts than that of truly biological left-sided clefts. The hypothesis that right and left-sided nsCL/P individuals are simply mirrored biologically is a common occurrence in the literature, where comparisons are typically made between unilateral and bilateral clefts (Manyama et al., 2011; Richtsmeier & DeLeon, 2009). My findings however show that the covariance structure of right-sided clefts, even following mirroring, is more similar to that of bilateral clefts than truly biological left-sided clefts. This finding broadly suggests that there are biological differences in the right and left-sided clefts. It appears that the left-sided cleft is different than the right-sided cleft, suggesting an underlying difference in etiology. This may be due to altered genetic control of spatial patterning.

My findings show the underlying covariance structure of cleft types and subtypes to be complex. It is important to consider this complexity during surgical planning, and in cosmetic outcomes following surgery. As previous studies have shown, the degree of integratedness in unilateral clefts is not affected by surgery (Richtsmeier & DeLeon, 2009). Our findings support

this result, and show that the degree of integration is more significantly related to cleft subtype than previously hypothesized.

Conclusion

Phenotypic variation is a predominant feature common to many dysmorphic disorders of the craniofacial complex. My finding that allometric factors have significant affect on the observed variance show that measures of size are important covariates in variance.

The covariance structure of the face is almost exclusively used in developmental integration studies of the craniofacial complex. There exists a difficulty in distinguishing between covariance structure and integration; or the conflation of pattern and process (Hallgrimsson et al.,

2009). Through this work, I have shown that structural integration occurs so that function and facial aesthetics are maintained over a vast range of allometric, or size, variation. One must be cognizant that the properties of development, and the underlying developmental architecture of a system, are much more complex than simple phenotypic correlations. The covariation of parts of the nsCL/P face is directly influenced by nsCL/P, and the developmental mechanism that produces the condition. It is important to be able to use the phenotypic correlations to make reasonable conjectures about how covariance structure may be affected through dysmorphology. These findings show the differences in covariance structure of several categories of nsCL/P patients.

Acknowledgements

This research was supported by studentships from the Alberta Children’s Hospital

Research Institute and by Alberta Innovates Health Solutions, to JRL. Dr. Robertson Harrop and

Dr. Donald McPhalen were critical initiators of this study, and facilitated patient access and acquisition.

Table 3-4: Supplementary Table 1: Counts of patients for each classification grouping, by laterality.

Right Left Bilateral Cleft Type Classification Cleft Lip 6 5 2 Cleft Lip and Palate 24 38 24 Cleft Severity Classification Complete 19 33 22 Incomplete 11 10 4 Cleft Type, Severity and Laterality Classification Cleft Lip, Complete 4 2 1 Cleft Lip, Incomplete 2 3 1

Cleft Lip and Palate, Complete 15 31 21

Cleft Lip and Palate, Incomplete 9 7 3

Chapter 4: Craniofacial Morphometric Analysis of Individuals with X-linked

Hypohidrotic Ectodermal Dysplasia

Abstract

Hypohidrotic ectodermal dysplasia (HED) is the most prevalent type of ectodermal dysplasia (ED).

ED is an umbrella term for a group of syndromes characterized by missing or malformed ectodermal structures, including skin, hair, sweat glands, and teeth. The X-linked recessive (XL), autosomal recessive (AR), and autosomal dominant (AD) types of HED are caused by mutations in the genes encoding ectodysplasin (EDA1), EDA receptor (EDAR), or EDAR-associated death domain (EDARADD). Patients with HED have a distinctive facial appearance, yet a quantitative analysis of the HED craniofacial phenotype using advanced three-dimensional (3D) technologies has not been reported. In this study, we characterized craniofacial morphology in subjects with X- linked hypohidrotic ectodermal dysplasia (XLHED) by use of 3D imaging and geometric morphometrics (GM), a technique that uses defined landmarks to quantify size and shape in complex craniofacial morphologies. We found that the XLHED craniofacial phenotype differed significantly from controls. Patients had a smaller and shorter face with a proportionally longer chin and midface, prominent midfacial hypoplasia, a more protrusive chin and mandible, a narrower and more pointed nose, shorter philtrum, a narrower mouth, and a fuller and more rounded lower lip. Our findings refine the phenotype of XLHED and may be useful both for clinical diagnosis of XLHED and to extend understanding of the role of EDA in craniofacial development.

Introduction

Ectodermal dysplasia (ED) encompasses more than 150 clinically distinct syndromes, all of which exhibit defects in the morphogenesis of ectodermal structures, including skin, hair, sweat glands, and teeth (Clauss et al., 2008). Hypohidrotic ectodermal dysplasia (HED) is the most prevalent type of ED and can be inherited in an X-linked (XL) recessive, autosomal recessive

(AR), or autosomal dominant (AD) manner. X-linked hypohidrotic ectodermal dysplasia

(XLHED) (OMIM #305100) is caused by mutations in EDA1, encoding ectodysplasin (Mikkola,

2009). AR-HED and AD-HED are caused by mutations in EDAR, encoding the EDA receptor, or

EDARADD, encoding EDAR-associated death domain (EDARADD) (Mikkola, 2009). In humans, EDA1 is expressed in multiple tissues including various epithelia, neuroectoderm, thymus, and bone during embryonic and fetal development and in adulthood (Montonen et al.,

1998). The clinical features of HED include sparse hair and eyebrows, wrinkled and dry skin, missing and malformed teeth, hypoplasia of sweat, sebaceous, meibomian, lacrimal, and mammary glands, and severe hypohidrosis (Mikkola, 2009). Mice with spontaneous mutations in Eda

(tabby), Edar (downless), or Edaradd (crinkled) exhibit abnormal phenotypes similar to humans with HED, including missing teeth, teeth with abnormal cusp morphology, absent hair types, and missing sweat glands (Courtney, Blackburn, & Sharpe, 2005).

Previous studies of individuals with HED using clinical dysmorphologic and cephalometric evaluations have identified the following craniofacial characteristics in these patients: maxillary hypoplasia, mandibular prognathism, facial concavity, frontal prominence, and depressed nasal bridge (Clauss et al., 2008). In this study, we extended the craniofacial phenotype of HED using three-dimensional (3D) imaging and geometric morphometric (GM) analysis, which applies

multivariate statistical techniques to defined landmarks to precisely quantify shape and size variation in complex morphologies (Zelditch, Swiderski, & Sheets, 2012). By contrast, cephalometric techniques capture only dimensional differences rather than changes in overall morphology or shape. 3D morphometric analysis has great potential in clinical diagnosis of syndromes associated with craniofacial dysmorphologies and has been applied to a number of syndromes, including Noonan syndrome, fragile X syndrome, and others (Hammond et al., 2004;

Heulens et al., 2013). Morphometric analysis has been utilized to identify subtle changes in craniofacial features that are difficult to observe by clinical examination, and this can help to define phenotypically distinct subgroups within a syndrome (Hammond et al., 2012) and to discover genotype–phenotype correlations (Bhuiyan et al., 2006; Hammond & Suttie, 2012). Here, in a cohort of 23 male subjects with XLHED, we characterize facial morphology using 3D GM analysis.

Material and Methods

4.1.1 Study Subject Demographics

This study received Institutional Review Board approval. Patients were enrolled in the study at the University of California, San Francisco in May 2011 or the National Foundation for

Ectodermal Dysplasias (NFED) Family Conference in Houston, TX in July 2013. All study subjects, or their legal guardians if subjects were under 18 years of age, provided written informed consent prior to participation in the study. A total of 59 healthy male control subjects with no family history of XLHED and 23 male case subjects with a genetically proven diagnosis of

XLHED participated in the study. EDA1 mutations are listed in Table 4-1. The 23 XLHED subjects

consisted of three pairs of brothers and 17 unrelated individuals. Control subjects were all unrelated. The age range of the XLHED cohort was 4–29 years (mean 15.83 years), and ethnic backgrounds included Caucasian (n = 19), Hispanic (n = 2), and African American (n = 2). Ages of the control subjects ranged from 4 to 31 years (mean=12.22 years), with all controls having

Caucasian ethnic background (n = 59). The age mismatch between the two groups is due to a larger number of younger subjects in the control group, which were included to better show phenotypic variation in the control sample. When cases and controls are matched one-to-one, the mean age difference between the two groups is very small (control mean age = 16.00 years, XLHED mean age = 15.82).

Table 4-1: Gene mutations in our cohort of 23 XLHED individuals. Asterisk (*) denotes mutation of brother pair.

Region in Ectodysplasin EDA1 Gene Sequence Mutation Type Affected Exon 01 R156H Missense Transmembrane Exon 01 164T>A (Leu55Gln)* Missense Transmembrane Exon 02 463C>T (Arg155Cys)* Missense Furin Exon 02 467G>A (Arg156His) Missense Furin Exon 02 C332Y Missense TNF Exon 02 novel R384S Missense Furin Exon 03 463C>T (Arg155Cys) Missense Furin Exons 03-08 del Deletion TNF Exon 04 553_588 del36 (185-196 del Deletion, in frame Collagen (GlyXY)X4) Exon 05 546_581 del36 Deletion Furin Exon 06 766C>T (Gln256X)* Nonsense TNF Exon 07 794A>G (Asp265Gly) Missense TNF Exon 07 822G>T (Trp274Cys) Missense TNF Exon 07 822 delG Deletion, truncating TNF Exon 07 895G>A (Gly299Ser) Missense TNF Exon 07 809 delT (Val270GlyfsX10) Deletion TNF Exon 08 ?_925 1176_? del Deletion TNF Exon 08 1070G>C (Arg357Pro) Missense TNF Exon 08 1087A>G (Lys363Glu) Missense TNF E67X mutation in EDA1 gene Nonsense Extracellular

4.1.2 3D Imaging and Landmarking

3D facial images were created using the 3D Capturor II camera system (InSpeck, Montreal,

Canada), utilizing white light 3D photogrammetry to create a 3D surface map in ~0.4 sec with a

640x480-mm field of view. Following digital reconstruction of the 3D images, 3D landmarks were

determined using MeshLab software (Cignoni et al., 2008). Figure 4-1 and Table 4-2 show the 24

discrete anatomical landmarks that were utilized to define and measure the shape of the

craniofacial and midfacial complexes. The landmarking protocol included the use of type 1 and

type 2 landmarks(Bookstein, 1997a).

Figure 4-1: Landmarks collected from digitized 3D facial photographs. Correspond to landmarks in Table 4-

Table 4-2: Facial landmarks utilized in morphometric analysis

Number Name Landmark Description Midline point where the frontal and nasal bones contact 1 (M) Naison (nasofrontal suture). Midline point marking the maximum protrusion of the 2 (M) Pronasale nasal tip. Midline point marking the junction between the inferior 3 (M) Subnasale border of the nasal septum and the cutaneous upper lip. Apex of the nasolabial angle. Midline point of the vermilion border of the upper lip, at 4 (M) Labiale Superius the base of the philtrum. 5 (M) Stomion Midpoint of the labial fissure. 6 (M) Labiale Inferius Midline point of the vermilion border of the lower lip. Midpoint along the inferior margin of the cutaneous lower 7 (M) Sublabiale lip. 8 (M) Gnathion Midline point on the inferior border of the mandible. Apex of the angle formed at the inner corner of the 9 (R/L) Endocanthion palpebral fissure where the upper and lower eyelids meet. Apex of the angle formed at the outer corner of the 11 (R/L) Exocanthion palpebral fissure where the upper and lower eyelids meet. 13 (R/L) Alare Most lateral point on the nasal ala. Most posterolateral point on the alar cartilage, located 15 (R/L) Alare Curvature Point within the crease formed by the union of the alar cartilage and the skin of the cheek. Point located at the lower margin of the nasal ala, where 17 (R/L) Subalare the cartilage inserts in the cutaneous upper lip. Point marking the lateral crest of the philtrum of the upper 19 (R/L) Christa Philtri lip. 21 (R/L) Chelion Point marking the lateral extent of the labial fissure. 23 (R/L) Zygion Most prominent portion of the zygomatic arch.

4.1.3 Statistical Shape Analyses

The shape analyses tested the null hypothesis that XLHED subjects did not have statistically different facial shape compared to control subjects. We used GM methods, based on

Procrustes superimposition, to quantify the shape and size of XLHED and control subjects(Bookstein, 1997b). Procrustes coordinates were calculated using the Procrustes generalized least squares superimposition method in MorphoJ software (Klingenberg, 2011) which removes isometric scaling, rotational, and translational data from the landmark coordinates (Rohlf

& Slice, 1990). The symmetric component of each coordinate was extracted from the landmark coordinates, and the resulting coordinates were used as shape variables in subsequent analyses. As a measure of size, for each subject we computed centroid size, which is the square root of the sum of the squared distances of each landmark coordinate from the centroid, or the mean x, y, z coordinate (Bookstein, 1997b). In virtually all complex morphological traits, a substantial component of the variation in shape is directly correlated with size (Hallgrimsson et al., 2009;

Klingenberg, 1998a).This variation, or allometry, can confound comparisons in which there is both a size and a shape effect. Even if the groups do not differ in size, removing the allometric component of variation will sharpen the focus on the morphological differences between the groups. Here, we removed both size (static allometry) and age (ontogenetic allometry) related variation from the coordinates using pooled within-group multivariate regression of shape on centroid size and age in years (Figure 4-2). The residuals of this regression were used in subsequent statistical shape analyses.

Figure 4-2: Multivariate pooled within-group regression of shape on centroid size (A) and age (B).

To examine the effects of the mutation on size, we regressed centroid size against age, and performed a t-test on the residuals. To visualize shape variation within the entire sample, we performed principal components analysis (PCA). PCA is a multivariate data reduction technique that summarizes patterns of variation and covariation by extracting independent and orthogonal axes of covariation, termed principal components (PCs), from a multivariate dataset. Each PC describes an axis of shape variation that explains a progressively smaller proportion of the total variation in the data (Zelditch et al., 2012). The shape variation described by each PC can be visualized as a 3D morphing of facial shape. 3D morphings of shape axes were generated by warping the 3D surface of an unaffected control, using the thin-plate spline procedure in the

Landmark software (Wiley et al., 2005). We additionally compared the eigenvector lengths of each landmark associated with PCs 1–3 to identify which landmarks were strongly associated with each

PC.

To visualize shape variation among XLHED and control individuals, we performed canonical variates analysis (CVA). CVA is similar to PCA in that canonical variates (CVs) are a linear combination of the original variables, constrained to be mutually orthogonal (Zelditch et al.,

2012), which scales the shape variation to the pooled within-group covariance matrix to maximize among-group shape variation (Zelditch et al., 2012). In addition to testing for differences between affected XLHED and control groups, we also performed CVA to test for differences based on ethnicity (Caucasian, African American, and Hispanic), type of EDA1 mutation (nonsense, missense, or deletion), and region of the EDA protein affected (tumor necrosis factor [TNF], furin, or transmembrane domain). As there were three pairs of siblings present in the subject sample, one sibling from each pair was removed before performing the PCA and CVA so as not to artificially

reduce variation in the sample due to shared facial similarity among related individuals. The eigenvalues for the PCs and CVs were exported, and the PC and CV scores for these three individuals were imputed into this shape space by summing the eigenvectors across the regression residuals for each individual using the statistical software R. Therefore, while they are depicted in the plots and analyses, there was no loss of power due to relatedness as it was based on the variation in the unrelated sample only.

Results

4.1.4 Facial shape of XLHED individuals differs from controls

Pooled within-group multivariate regression of shape on centroid size and age revealed that

22.91% of shape variation within the dataset was due to static and ontogenetic allometry combined.

These sources of allometry were removed by using the residuals to examine variation within the sample. Furthermore, an additional regression of centroid size on age revealed a size effect of the

XLHED mutation, such that XLHED individuals have a significantly smaller face than healthy controls (Control mean=0.009, XLHED mean=-0.0254, p=0.003). XLHED and control individuals moderately differed from each other across PCs 1, 4, and 6. Together, PCs 1 through 6 accounted for 74% of the total shape variance.

The first PC (32% of the total variance) captured shape variation concentrated in the nose and mouth (midfacial complex) and zygomatic region (Figure 4-3A). Shape variance was also evident in the mandible, with the positive end of PC1 displaying a degree of mandibular prognathism, as deciphered from an anterior translation of midline landmarks 7 and 8 (Table 4-2).

Compared to controls, individuals with XLHED displayed more protruding chin and mandible, high zygomatic arches, a narrower and more pointed nose, and a narrower mouth. PC1 also separated XLHED and control individuals in terms of facial height. The XLHED shape described at the positive end of PC1 was an overall shorter face with relatively longer chin and shortened philtrum compared to control individuals, who scatter toward the negative end and zero of PC1.

Furthermore, in comparison of the eigenvector lengths, we found strong association of landmarks located in the midface and upper face, and the mandible/chin in PCs 1–3 (Figure 4-4, Table 4-3).

CVA of the XLHED cohort found no significant differences between mean facial shape according to ethnicity (Caucasian-African American p= 0.3339; Hispanic-African American p= 0.3456,

Caucasian-Hispanic p= 0.0568). These data indicate that the predominant shape effects observed result from XLHED, rather than ethnicity.

Figure 4-3: Multivariate shape analyses of craniofacial features of XLHED subjects compared to controls. (A) PC1 versus PC2, showing shape distribution of XLHED and control individuals. Ellipses correspond to 95% confidence intervals. Thin-plate spline warps illustrate the shape changes in PC1, corresponding to the observed zero, positive, and negative extreme values. (B) Canonical variate (CV) analysis histogram showing shape distribution of XLHED and control individuals. Thin-plate spline warps illustrate the shape changes in CV1, corresponding to the observed zero, positive, and negative extreme values.

100

Figure 4-4: Magnitude of shape change by principal component. Magnitude of shape change for PCs 1–3, as calculated from PC loadings. Magnitudes are magnified by 2X.

Table 4-3: Magnitude of shape change by landmark and principal component. Calculated from principal component loadings. Highlighted columns indicate the greatest five magnitudes for each PC.

Landmark PC1 PC2 PC3 Naison 0.143117456 0.116092848 0.119137884 Pronasale 0.241410529 0.15089658 0.310678676 Subnasale 0.171016368 0.193307205 0.143618644 Labiale Superius 0.02574881 0.196070283 0.074043264 Stomion 0.020125978 0.182290454 0.10970578 Labiale Inferius 0.049268372 0.101872805 0.280908153 Sublabiale 0.083130685 0.049530123 0.214015768 Gnathion 0.400064161 0.435640186 0.31867488 Endocanthion 0.113563547 0.17500632 0.037266746 Exocanthion 0.102852151 0.200614585 0.314381179 Alare 0.086500173 0.074737828 0.142510365 Alare Curvature Point 0.07772929 0.073955088 0.148630331 Subalare 0.088548816 0.148272502 0.171547156 Christa Philtri 0.03416358 0.185010957 0.06549604 Chelion 0.072636326 0.147697372 0.224723712 Zygion 0.523884875 0.352113085 0.23239031

101

4.1.5 Characteristic midfacial shape in XLHED individuals

Permutation tests (10,000 permutation rounds) using the Procrustes distance between groups defined by affected status revealed a significant midfacial shape difference between

XLHED and control individuals (Procrustes distance = 0.0650, p<0.0001) with a characteristic midfacial shape in XLHED individuals. We then performed an additional permutation test, using the dataset with the three related individuals imputed into the Procrustes space. This analysis resulted in slightly but not significantly altered Procrustes distances (Procrustes distance = 0.0706 p<0.0001). Since including the other half of the sibling pairs did not appreciably alter the resultant shapes, all individuals were included in the final analyses due to the small sample size of the

XLHED cohort. We performed a CVA using the unrelated subjects and projected the related individuals into this space using the CVA eigenvectors. CV1 showed that XLHED individuals had a relatively shorter face with a shortened philtrum and nasal columella, and displayed a degree of mandibular prognathism (Figure 4-3B). XLHED individuals also had altered labium inferius oris shape, with a fuller and more rounded lower lip than controls. Narrower nasal ala and a more pointed nasal tip were also observed in XLHED individuals. We found no significant shape differences within the XLHED group when a permutation test was performed based on type of mutation (nonsense, missense, or deletion) or region of the EDA protein affected (TNF, furin, or transmembrane domain). Together, these findings show that individuals with XLHED have a characteristic craniofacial phenotype, statistically different from controls.

102

Discussion

GM analysis on individuals with XLHED showed that, compared to control individuals, subjects with XLHED exhibit a quantitatively distinct set of craniofacial characteristics, including an overall reduction in size of the face, a shorter face, high zygomatic arches, relatively long chin, shortened philtrum, midface hypoplasia, fuller and more rounded lower lip, more protrusive chin and prognathic mandible, narrower and more pointed nose, and narrower mouth.

Previous reports that utilized anthropormorphic and cephalometric measurements have shown that male patients with XLHED exhibit decreased total facial height (Lexner et al., 2007), and this finding agrees with our study, in which XLHED individuals had relatively shorter and narrower facial shape than controls. Additionally, previous studies in XLHED individuals have reported the following: midfacial hypoplasia, with a retroclined nasal bone; short, retrognathic, anteriorly inclined maxilla; and a prognathic mandible, all of which are in agreement with our findings (Johnson et al., 2002; Lexner et al., 2007; Saksena & Bixler, 1990).

Currently, cephalometric measurement of skeletal structures are standard diagnostic tools in orthodontics and oral surgery. Our study shows that 3D craniofacial morphometric analysis provides a more detailed, more efficient, and more accurate tool than 2D cephalometrics in diagnosis and treatment of individuals with XLHED and potentially other syndromes as well.

103

In addition, we attempted to determine phenotype–genotype correlations in the XLHED cohort based on 3D morphometric analysis. Overall, we found no significant differences in craniofacial structures based on type of EDA1 mutation (nonsense, missense, or deletion) or region of the EDA protein affected (TNF, Furin, or transmembrane domain). This absence of genotype– phenotype correlation suggests that any mutations in EDA that cause significant loss of function can result in a similar craniofacial appearance, consistent with published reports that have not observed genotype–phenotype correlations in XLHED (Clauss et al., 2010; Kobielak et al., 2001;

Zhang et al., 2011). Nevertheless, it remains possible that, with analysis of a larger cohort in the future, the utilization of 3D morphometrics might accurately distinguish subtle morphological variations that may highlight correspondingly subtle genotype–phenotype correlations.

Precise 3D craniofacial morphometric analysis thus is a powerful tool for rapid clinical diagnosis of XLHED, and may serve as a useful adjunct to genetic testing. Genetic diagnoses can be difficult to obtain, patients may wait several years before receiving a molecular diagnosis. In addition, the same technology may be applicable to diagnosis of female carriers of XLHED.

Indeed, previous cephalometric analyses of female carriers of XLHED have reported a relatively short, retrognathic maxilla and retruded lips (Saksena & Bixler, 1990). Thus, 3D craniofacial morphometric analysis may become an important tool for the rapid identification of syndromes in the future. The ability to quantitatively define craniofacial phenotypes will improve the speed and accuracy of diagnosis, and as molecular therapies for conditions such as XLHED are developed,

3D morphometrics may help to pave the way for early identification and treatment.

104

Acknowledgments

The authors thank all of the participants in this study and their families.

Conflict of Interest

Edimer Pharmaceuticals provided funding for the study as well as technical assistance. The academic authors independently analyzed all data. Additionally, the decision about where and what to publish was made solely by the primary authors.

105

Chapter 5: Conclusion and Future Directions

Introduction

This dissertation work converges on the theme of morphological integration. I have explored the role of morphological integration in craniofacial variation in humans via three related hypotheses. Specifically, I hypothesized that the size of the skull and face will have significant effects in facial shape; that non-syndromic cleft lip and/or palate (nsCL/P) patients and control groups will have vastly different morphological relationships between parts of the face, and that the relationship of size and shape will play a significant role in the morphological covariation of the human face. Furthermore, I examined an ectodermal dysplasia (ED) patient cohort as a case study of a syndrome with a craniofacial phenotype, and hypothesized that the syndromic phenotype is quantifiable, and significantly differs from that of controls. This study design allowed me to examine my hypothesis in two etiologically distinct dysmorphic groups; one with a severely dysmorphic phenotype (nsCL/P) resulting from a profound developmental perturbation, and the other an inherited condition (XLHED) that does not affect the face. By utilizing patients with both congenital and inherited syndromes, I was able to test my hypotheses across a broad patient group.

My findings have shown that the allometric factors of facial shape variation are complex and overlapping. I have also shown that different subtypes and classifications of cleft phenotype are important in deciphering covariance structure. Furthermore, a quantitative analysis of XLHED patients showed distinct phenotypes unique to the syndrome including a size effect of the XLHED mutation, as XLHED individuals have significantly smaller faces than controls. Ultimately, our studies have furthered the understanding of complex craniofacial phenotypes, both in dysmorphic

106

and syndromic conditions, and in normal individuals. Extending the knowledge of the growth of the craniofacial complex and how soft tissue groups interact during development may help inform future clinical practice.

Synthesis

Facial shape and facial size are important biological traits influenced by numerous factors including familial inheritance, gene perturbations, natural selection and sexual selection. The findings of Chapter Two show that facial shape variation is also significantly affected by various measures of biological size. These axes of variation are complex and overlapping, and emphasize the overlapping nature of measures of size and its effect on facial shape variation. Through this

Chapter, it is emphasized that the magnitude and direction of variation are important when looking at variation in morphology. Furthermore, this Chapter shows that shape differences between

Tanzanian and European-Derived North American populations only explain a small portion of the total variation, less than the variation present within each population. Despite the likely different environmental influences, the facial shape allometry patterns are very similar.

Why should the medical sciences field continue to explore variation in facial morphology?

Firstly, because the human face, and by extension the human skull, are a defining characteristic of our species. Secondly, variation in facial shape displays hints and clues about one’s health, ethnic background, disease status, and genetic health. Diagnostic facies are established for several syndromes, with certain diseases having more distinct pathognomonic features than others.

However, despite established diagnostic facies, variation within these groups is a little-studied

107

feature — despite the fact that phenotypic variation is a predominant feature common to many dysmorphic disorders of the craniofacial complex. Chapter Three of this dissertation focused on an allometric examination of facial shape in a dysmorphic group, a cohort of patients with non- syndromic cleft lip and/or palate. Following my results from Chapter Two, allometric factors have significant effects on the observed variance in nsCL/P patients. This finding only further emphasizes that measures of size are important covariates in shape variance. Furthermore, Chapter

Three showed that structural integration occurs in the face so that necessary biological functions and facial aesthetics are maintained over a vast range of allometric, or size, variation. These findings emphasize the importance of understanding the range of phenotypic variation that is displayed in a dysmorphic syndrome.

Why is the face, and facial appearance, so important in diagnostic medicine? The answer to this questions converges on a central concept; that the face is a complex trait that is sensitive to genetic and environmental perturbations, and as such can reveal important underlying health markers about an individual. The connection between prenatal exposure to alcohol and facial appearance is strongly established (Wilhoit, Scott, & Simecka, 2017). While Fetal Alcohol

Spectrum Disorder (FASD) is characterized as a syndrome with inherent variation, the FASD phenotype is almost always universally described as displaying small eyes, a wide and flat philtrum with a thin upper lip, and low-set ears. A recent study has examined the effect of the variation in alcohol exposure on facial shape (Muggli et al., 2017). It was shown that even at low exposure levels, alcohol influences normal craniofacial development. At low levels of alcohol exposure, a marginally altered forehead phenotype was produced. At high and binge-level exposure levels, the phenotype becomes more severe and affects more substantial regions of the face. Many research

108

groups are also developing tools to assist clinicians in recognizing FASD phenotypes with respect to ethnic facial variation, as the variation seen in FASD becomes superimposed upon the general underlying ethnic facial shape variation (Hoyme et al., 2015; Suttie et al., 2017). These findings are important when one considers FASD phenotypes; the dose-dependent effects of the alcohol exposure are realized in the resultant severity of phenotype. In the case of FASD, increased alcohol exposure worsens the phenotype.

Variation in phenotype can also be seen in syndromes such as Neurofibromatosis Type 1

(NF1). The majority of NF1 patients have a 1.4-Mb deletion encompassing the NF1 gene.

However, there is a distinct group of individuals that have NF1 microdeletions in the flanking regions of the gene. The microdeletion group displays a much more severe facial phenotype, with up to 90% displaying facial dysmorphism, which is not seen in the general NF1 population.

Furthermore, 28% of microdeletion patients display severe facial asymmetry, while only 8% of the general NF1 patients display asymmetry (Kehrer-Sawatzki, Mautner, & Cooper, 2017). This is an interesting case study of facial shape variation due to genetic variation, especially when one compares these results to the dose-dependent variation as seen in the FASD study. In the NF1 patients, it seems that the microdeletions in the gene flanking regions have more deleterious phenotypic effects than would a massive 1.4-Mb deletion in the NF1 gene. As clearly evidenced by contrasting FASD and NF1 variation, variation patterns are different across syndromes.

In light of these differences, it is crucial that the understanding of the complexity of genotype-phenotype relationships be expanded upon. In particular, studies involving the face are promising for their clinical merit; as the human face displays variation not only due to ethnic

109

background, but also due to genetic perturbations, environmental conditions, and developmental determinants. Narrow-sense heritability of facial features is estimated to range from 28 to 67%

(Cole et al., 2017), furthering emphasizing that genetic factors are an important point of inquiry for future facial shape investigations.

By exploring these genotype-phenotype relationships, it is possible to characterize additional diagnostic facies for many syndromes and conditions. In this dissertation, Chapter Four described the previously unreported 3-D phenotype of a patient cohort with X-linked hypohidrotic ectodermal dysplasia. This chapter emphasized the importance of classifying syndromic phenotypes, and underscored the need for furthered understanding of subtle facial phenotypes connected to underlying genetic syndromes. In the future, it has been hypothesized that the diagnostic value of the face will be tantamount to DNA analysis. Because the human face can reveal such a vast amount of information, from health status (Phalane, Tribe, Steel, Cholo, &

Coetzee, 2017) to susceptibility of producing offspring with nsCL/P (Weinberg et al., 2008), we must continue to study many facets surrounding craniofacial shape and appearance — the role of developmental insults, how malformation occurs in the fetus, if dysmorphology may be rescued during development (as in the case of microform cleft lip), why certain areas of the face are affected more than others, and how all of these factors converge together to form a functional craniofacial complex. It is important to continue characterizing facial phenotypes in syndromic individuals, as the variation within the syndrome presentation can actually inform clinicians about additional health-related issues. In Down Syndrome, facial shape can be predictive of obstructive sleep apnea (Jayaratne et al., 2017). It is also important to consider that subtle facial dysmorphologies are found in patients with psychiatric conditions such as bipolar disorder and

110

schizophrenia (Hennessy, Baldwin, Browne, Kinsella, & Waddington, 2010). This once again demonstrates the diagnostic value in facial phenotyping from a health-related standpoint. If we can diagnose diseases via facial morphology before that disease progresses, not only would the affected individual benefit from early treatment, but the clinical outcome of that patient likely would improve. Lending credence to this paradigm is the topic of prenatal diagnosis via facial phenotyping, as seen in syndromes such as nsCL/P, Down syndrome, Pierre-Robin sequence, and

Binder phenotype (Katsube et al., 2017; Werner, Lopes Dos Santos, Ribeiro, & Araujo Júnior,

2017). There is great clinical significance in early detection and diagnosis of these syndromes, as treatment plans are often coordinated while the fetal patient is still in utero—improving their postnatal clinical outcome.

Future Directions

A study examining the effects of growth on nsCL/P patients, in relation to facial integration would be a useful extension of this dissertation. Ideally, this study would involve following patients long-term; beginning at birth, and following them through childhood, at which point several reparative surgeries would have been performed. This would allow a longitudinal, as opposed to cross-sectional, analysis of the allometric affects of size on facial shape. In general, there are many benefits to longitudinal studies including the ability to evaluate the relationship between risk factors and disease development, to evaluate treatment outcomes over time and to study the affect of treatments in terms of timing and chronicity. A cross-sectional study would not allow a study of cause-and-effect, because the study design would not include the influence of time on the measured variables (Caruana, Roman, Hernández-Sánchez, & Solli, 2015).

111

Specifically, in a craniofacial study, a longitudinal study would allow investigators the ability to follow facial shape change over time and to examine the facial shape affects attributable to specific surgical interventions. Additionally, by longitudinally examining healthy control individuals, knowledge about normal craniofacial growth patterns can be established and applied to the study of growth patterns in dysmorphic individuals. A longitudinal study would allow questions surrounding allometric relationships over time to be thoroughly investigated and further our knowledge surrounding craniofacial growth patterns in both normal and dysmorphic individuals.

Another avenue of interest would be to conduct a study to characterize the phenotypes of additional syndromes in anticipation of using shape data as a diagnostic tool. This classification of syndromes is an endless pursuit; unreported novel genetic syndromes are discovered regularly. A future direction of the XLHED study presented in Chapter 4, which is already underway by several research groups, is to quantify the syndromic phenotypes of multiple genetic syndromes. By quantifying phenotypes, a diagnostic tool can be built to allow clinicians to explore quick and minimally invasive phenotype-driven diagnoses. Currently, there is a rudimentary analysis tool which utilizes 2D photographs to diagnose XLHED in males (Hadj-Rabia et al., 2017). By developing these sorts of diagnostic tools, we can improve patient outcomes and advance the field of diagnostics.

A modern, and novel, way of conducting future facial phenotype studies would be to utilize the vast population of individuals who are submitting their DNA to for-profit companies such as

23andme, AncestryDNA, and Family Tree DNA. These services provide customers with complete reports about their ancestry, including things such as their ancestry composition, maternal

112

haplogroups and even Neanderthal ancestry. Health-related genetic risk reports are produced for conditions such as macular degeneration, Parkinson’s disease and Alzheimer’s disease.

Additionally, these services will provide you with trait analysis such as genetic eye color, freckles, hair texture and hair color. By coordinating collection sites in large metropolitan areas, it would be a relatively easy way to rapidly collect data on facial phenotypes and to explore how these phenotypes may be influenced via genetic variants. This could also be applied in the realm of ancestry and facial phenotype. Because of the sheer volume of participants, and the company’s ability to test for multiple genetic variants, this would be an extremely cost-effective way to procure large volumes of data on facial phenotypes and population variation.

Concluding Statement

The study of facial morphology is a small but emerging field, punctuated by rapidly evolving technology and the discovery of new rare genetic syndromes. The work presented in this dissertation remains relevant to the field, as allometry and shape variation due to allometry are not widely studied topics, but are important when characterizing facial shape variation in a group. This dissertation clearly shows the importance of considering allometry in facial shape investigations

— while comprising a relatively small proportion of the total variation, as shown in Chapters 2 and 3, allometry is an important source of integration that is poorly understood. Commonly regarded as a singular effect, this dissertation has shown the importance of characterizing the effects of growth on shape, and how allometric patterns in the human face are complex and overlapping. Additionally, the role of allometry in complex syndromes such as nsCL/P is important when considering the structural integration and resulting functionality of the craniofacial

113

complex. By exploring these issues, treatments may be made more precise, clinical outcomes will improve, and most importantly, patients will experience less negative consequences arising from their condition.

114

References

Adams, D. C., & Felice, R. N. (2014). Assessing Trait Covariation and Morphological Integration

on Phylogenies Using Evolutionary Covariance Matrices. PLoS ONE, 9(4), e94335.

http://doi.org/10.1371/journal.pone.0094335

Adams, D. C., & Otárola Castillo, E. (2013). geomorph: an r package for the collection and

analysis of geometric morphometric shape data. Methods in Ecology and Evolution, 4(4), 393–

399. http://doi.org/10.1111/2041-210X.12035

Aldridge, K., Kane, A. A., Marsh, J. L., Yan, P., Govier, D., & Richtsmeier, J. T. (2005).

Relationship of brain and skull in pre- and postoperative sagittal synostosis. Journal of

Anatomy, 206(4), 373–385. http://doi.org/10.1111/j.1469-7580.2005.00397.x

Allanson, J. E., O'Hara, P., Farkas, L. G., & Nair, R. C. (1993). Anthropometric craniofacial

pattern profiles in Down syndrome. American Journal of Medical Genetics, 47(5), 748–752.

http://doi.org/10.1002/ajmg.1320470530

Banerjee, S., Banerjee, R., Radke, U. M., & Mundhe, D. (2011). Pre-Surgical Management of

Unilateral Cleft Lip and Palate in a Neonate: A Clinical Report. The Journal of Indian

Prosthodontic Society, 11(1), 71–76. http://doi.org/10.1007/s13191-011-0052-2

Bannister, J., Larson, J. R., Hallgrimsson, B., Lim, P. H., Spritz, R. A., Klein, O. D., et al. (2017).

CARS 2017—Computer Assisted Radiology and Surgery Proceedings of the 31st

International Congress and Exhibition Barcelona, Spain, June 20–24, 2017

. International Journal of Computer Assisted Radiology and Surgery, 12(Suppl 1), S216. Retrieved

from 10.1007/s11548-017-1588-3

Basel Vanagaite, L., Wolf, L., Orin, M., Larizza, L., Gervasini, C., Krantz, I. D., & Deardoff, M.

A. (2016). Recognition of the Cornelia de Lange syndrome phenotype with facial 115

dysmorphology novel analysis. Clinical Genetics, 89(5), 557–563.

http://doi.org/10.1111/cge.12716

Bastir, M. (2008). A systems-model for the morphological analysis of integration and modularity

in human craniofacial evolution. Journal of Anthropological Sciences = Rivista Di

Antropologia : JASS / Istituto Italiano Di Antropologia, 86, 37–58.

Bastir, M., & Rosas, A. (2006). Correlated variation between the lateral basicranium and the face:

A geometric morphometric study in different human groups. Archives of Oral Biology, 51(9),

814–824. http://doi.org/10.1016/j.archoralbio.2006.03.009

Bastir, M., Rosas, A., & O’Higgins, P. (2006). Craniofacial levels and the morphological

maturation of the human skull. Journal of Anatomy, 209(5), 637–654.

http://doi.org/10.1111/j.1469-7580.2006.00644.x

Baume, R. M., Buschang, P. H., & Weinstein, S. (1983). Stature, head height, and growth of the

vertical face. American Journal of Orthodontics, 83(6), 477–484.

http://doi.org/10.1016/S0002-9416(83)90246-4

Berkowitz, S. (1996). A comparison of treatment results in complete bilateral cleft lip and palate

using a conservative approach versus millard-latham PSOT procedure. Seminars in

Orthodontics, 2(3), 169–184. http://doi.org/10.1016/S1073-8746(96)80012-0

Bhuiyan, Z. A., Klein, M., Hammond, P., van Haeringen, A., Mannens, M. M. A. M., Van

Berckelaer-Onnes, I., & Hennekam, R. C. M. (2006). Genotype-phenotype correlations of 39

patients with Cornelia De Lange syndrome: the Dutch experience. Journal of Medical

Genetics, 43(7), 568–575. http://doi.org/10.1136/jmg.2005.038240

Biesecker, L. G., & Carey, J. C. (2011). Implementing the elements of morphology in the

American Journal of Medical Genetics. American Journal of Medical Genetics Part A, 155(5),

116

969–971. http://doi.org/10.1002/ajmg.a.33772

Bonhomme, V., Picq, S., Gaucherel, C., & Claude, J. (2014). Momocs: Outline Analysis Using R.

Journal of Statistical Software, 56(13), 1–24. http://doi.org/10.18637/jss.v056.i13

Bookstein, F. L. (1997a). Landmark methods for forms without landmarks: morphometrics of

group differences in outline shape. Medical Image Analysis, 1(3), 225–243.

http://doi.org/10.1016/S1361-8415(97)85012-8

Bookstein, F. L. (1997b). Morphometric Tools for Landmark Data. Cambridge University Press.

Boughner, J. C., Wat, S., Diewert, V. M., Young, N. M., Browder, L. W., & Hallgrimsson, B.

(2008). Short-faced mice and developmental interactions between the brain and the face.

Journal of Anatomy, 213(6), 646–662. http://doi.org/10.1111/j.1469-7580.2008.00999.x

Bright, J. A., Marugán-Lobón, J., Cobb, S. N., & Rayfield, E. J. (2016). The shapes of bird beaks

are highly controlled by nondietary factors. Proceedings of the National Academy of Sciences

of the United States of America, 113(19), 5352–5357. http://doi.org/10.1073/pnas.1602683113

Buck, T., & Viđarsdóttir, U. S. (2012). Craniofacial evolution in polynesia: A geometric

morphometric study of population diversity. American Journal of Human Biology, 24(6), 776–

785. http://doi.org/10.1002/ajhb.22315

Bugaighis, I., Mattick, C. R., Tiddeman, B., & Hobson, R. (2013). Three-dimensional gender

differences in facial form of children in the North East of England. The European Journal of

Orthodontics, 35(3), 295–304. http://doi.org/10.1093/ejo/cjr033

Callieri, M., Ponchio, F., & Cignoni, P. (2008). Virtual Inspector: a flexible visualizer for dense

3D scanned models. … Computer Graphics and ….

http://doi.org/10.1109/MCG.2008.20","publicationTitle“:”IEEE

Carinci, F., Scapoli, L., Palmieri, A., Zollino, I., & Pezzetti, F. (2007). Human genetic factors in

117

nonsyndromic cleft lip and palate: An update. International Journal of Pediatric

Otorhinolaryngology, 71(10), 1509–1519. http://doi.org/10.1016/j.ijporl.2007.06.007

Caruana, E. J., Roman, M., Hernández-Sánchez, J., & Solli, P. (2015). Longitudinal studies.

Journal of Thoracic Disease, 7(11), E537–40. http://doi.org/10.3978/j.issn.2072-

1439.2015.10.63

Chehadeh-Djebbar, El, S., Blair, E., Holder-Espinasse, M., Moncla, A., Frances, A.-M., Rio, M.,

et al. (2013). Changing facial phenotype in Cohen syndrome: towards clues for an earlier

diagnosis. European Journal of Human Genetics, 21(7), 736–742.

http://doi.org/10.1038/ejhg.2012.251

Cheverud, J. M. (1982). Phenotypic, Genetic, and Environmental Morphological Integration in the

Cranium. Evolution, 36(3), 499–516. http://doi.org/10.2307/2408096

Cheverud, J. M. (1996). Developmental integration and the evolution of pleiotropy. American

Zoologist.

Chong, H. J., Young, N. M., Hu, D., Jeong, J., McMahon, A. P., Hallgrimsson, B., & Marcucio,

R. S. (2012). Signaling by SHH rescues facial defects following blockade in the brain.

Developmental Dynamics, 241(2), 247–256. http://doi.org/10.1002/dvdy.23726

Cignoni, P., Corsini, M., & Ranzuglia, G. (2008). Meshlab: an open-source 3d mesh processing

system. Ercim News.

Clauss, F., Chassaing, N., Smahi, A., Vincent, M. C., Calvas, P., Molla, M., et al. (2010). X-linked

and autosomal recessive Hypohidrotic Ectodermal Dysplasia: genotypic-dental phenotypic

findings. Clinical Genetics, 78(3), 257–266. http://doi.org/10.1111/j.1399-0004.2010.01376.x

Clauss, F., Manière, M. C., Obry, F., Waltmann, E., Hadj-Rabia, S., Bodemer, C., et al. (2008).

Dento-Craniofacial Phenotypes and underlying Molecular Mechanisms in Hypohidrotic

118

Ectodermal Dysplasia (HED): a Review. Journal of Dental Research, 87(12), 1089–1099.

http://doi.org/10.1177/154405910808701205

Cole, J. B., Manyama, M., Kimwaga, E., Mathayo, J., Larson, J. R., Liberton, D. K., et al. (2016).

Genomewide Association Study of African Children Identifies Association of SCHIP1 and

PDE8A with Facial Size and Shape. PLoS Genetics, 12(8), e1006174.

http://doi.org/10.1371/journal.pgen.1006174

Cole, J. B., Manyama, M., Larson, J. R., Liberton, D. K., Ferrara, T. M., Riccardi, S. L., et al.

(2017). Human Facial Shape and Size Heritability and Genetic Correlations. Genetics, 205(2),

967–978. http://doi.org/10.1534/genetics.116.193185

Collyer, M. L., Sekora, D. J., & Adams, D. C. (2015). A method for analysis of phenotypic change

for phenotypes described by high-dimensional data. Heredity, 115(4), 357–365.

http://doi.org/10.1038/hdy.2014.75

Courtney, J.-M., Blackburn, J., & Sharpe, P. T. (2005). The Ectodysplasin and NF?B signalling

pathways in odontogenesis. Archives of Oral Biology, 50(2), 159–163.

http://doi.org/10.1016/j.archoralbio.2004.11.019 de Campinas, U. E. (2009). Orofacial features of Treacher Collins syndrome. Web of Science. de Oliveira, R. L. B., de Santana Santos, T., de Almeida Teixeira, J. L., Martins-Filho, P. R. S., &

da Silva, L. C. F. (2015). Health-Related Quality of Life Patients With a Cleft Lip and/or

Palate. The Journal of Craniofacial Surgery, 26(8), 2315–2319.

http://doi.org/10.1097/SCS.0000000000001992

Diewert, V. M. (1985). Development of human craniofacial morphology during the late embryonic

and early fetal periods. American Journal of Orthodontics, 88(1), 64–76.

Edgar, H. J. H., & Hunley, K. L. (2009). Race reconciled?: How biological anthropologists view

119

human variation. American Journal of Physical Anthropology, 139(1), 1–4.

http://doi.org/10.1002/ajpa.20995

Farina, A., Wyszynski, D. F., Pezzetti, F., Scapoli, L., Martinelli, M., Carinci, F., et al. (2002).

Classification of oral clefts by affection site and laterality: a genotype-phenotype correlation

study. Orthodontics & Craniofacial Research, 5(3), 185–191.

Fraser, F. C. (1970). The genetics of cleft lip and cleft palate. American Journal of Human

Genetics, 22(3), 336–352.

Freidline, S. E., Gunz, P., & Hublin, J.-J. (2015). Ontogenetic and static allometry in the human

face: Contrasting Khoisan and Inuit. Am J Phys Anthropol, 158(1), 116–131.

http://doi.org/10.1002/ajpa.22759

German, R. Z., & Meyers, L. L. (1989). The role of time and size in ontogenetic allometry: I.

Review. Growth, Development, and Aging : GDA, 53(3), 101–106.

Gonzalez, P. N., Kristensen, E., Morck, D. W., Boyd, S., & Hallgrimsson, B. (2013). Effects of

growth hormone on the ontogenetic allometry of craniofacial bones. Evolution &

Development, 15(2), 133–145. http://doi.org/10.1111/ede.12025

Gonzalez, P. N., Perez, S. I., & Bernal, V. (2011). Ontogenetic Allometry and Cranial Shape

Diversification Among Human Populations From South America. The Anatomical Record:

Advances in Integrative Anatomy and Evolutionary Biology, 294(11), 1864–1874.

http://doi.org/10.1002/ar.21454

Goodwin, A. F., Larson, J. R., Jones, K. B., Liberton, D. K., Landan, M., Wang, Z., et al. (2014).

Craniofacial morphometric analysis of individuals with X-linked hypohidrotic ectodermal

dysplasia. Molecular Genetics & Genomic Medicine, 2(5), 422–429.

http://doi.org/10.1002/mgg3.84

120

Gorlin, R. J., Cohen, M. M., Jr, & Hennekam, R. (2001). Syndromes of the Head and Neck. New

York: Oxford University Press.

Gould, S. J. (1966). ALLOMETRY AND SIZE IN ONTOGENY AND PHYLOGENY. Biological

Reviews, 41(4), 587–638. http://doi.org/10.1111/j.1469-185X.1966.tb01624.x

Hadj-Rabia, S., Schneider, H., Navarro, E., Klein, O., Kirby, N., Huttner, K., et al. (2017).

Automatic recognition of the XLHED phenotype from facial images. American Journal of

Medical Genetics Part A, 173(9), 2408–2414. http://doi.org/10.1002/ajmg.a.38343

Hallgrimsson, B., Brown, J. J. Y., Ford Hutchinson, A. F., Sheets, H. D., Zelditch, M. L., & Jirik,

F. R. (2006). The brachymorph mouse and the developmental-genetic basis for canalization

and morphological integration. Evolution & Development, 8(1), 61–73.

http://doi.org/10.1111/j.1525-142X.2006.05075.x

Hallgrimsson, B., Dorval, C. J., Zelditch, M. L., & German, R. Z. (2004). Craniofacial variability

and morphological integration in mice susceptible to cleft lip and palate. Journal of Anatomy,

205(6), 501–517.

Hallgrimsson, B., Jamniczky, H., Young, N. M., Rolian, C., Parsons, T. E., Boughner, J. C., &

Marcucio, R. S. (2009). Deciphering the Palimpsest: Studying the Relationship Between

Morphological Integration and Phenotypic Covariation. Evolutionary Biology, 36(4), 355–

376. http://doi.org/10.1007/s11692-009-9076-5

Hallgrimsson, B., Lieberman, D. E., Liu, W., Ford-Hutchinson, A. F., & Jirik, F. R. (2007).

Epigenetic interactions and the structure of phenotypic variation in the cranium. Evolution &

Development, 9(1), 76–91. http://doi.org/10.1111/j.1525-142X.2006.00139.x

Hammond, P., & Suttie, M. (2012). Large-scale objective phenotyping of 3D facial morphology.

Human Mutation, 33(5), 817–825. http://doi.org/10.1002/humu.22054

121

Hammond, P., Hannes, F., Suttie, M., Devriendt, K., Vermeesch, J. R., Faravelli, F., et al. (2012).

Fine-grained facial phenotype|[ndash]|genotype analysis in Wolf|[ndash]|Hirschhorn

syndrome. European Journal of Human Genetics, 20(1), 33–40.

http://doi.org/10.1038/ejhg.2011.135

Hammond, P., Hutton, T. J., Allanson, J. E., Campbell, L. E., Hennekam, R. C. M., Holden, S., et

al. (2004). 3D analysis of facial morphology. American Journal of Medical Genetics, 126A(4),

339–348. http://doi.org/10.1002/ajmg.a.20665

Hennessy, R. J., Baldwin, P. A., Browne, D. J., Kinsella, A., & Waddington, J. L. (2010).

Frontonasal dysmorphology in bipolar disorder by 3D laser surface imaging and geometric

morphometrics: comparisons with schizophrenia. Schizophrenia Research, 122(1-3), 63–71.

http://doi.org/10.1016/j.schres.2010.05.001

Heulens, I., Suttie, M., Postnov, A., De Clerck, N., Perrotta, C. S., Mattina, T., et al. (2013).

Craniofacial characteristics of fragile X syndrome in mouse and man. European Journal of

Human Genetics, 21(8), 816–823. http://doi.org/10.1038/ejhg.2012.265

Heymsfield, S. B., Gallagher, D., Mayer, L., Beetsch, J., & Pietrobelli, A. (2007). Scaling of

human body composition to stature: new insights into body mass index. The American Journal

of Clinical Nutrition, 86(1), 82–91.

Hill, C. A., Martínez-Abadías, N., Motch, S. M., Austin, J. R., Wang, Y., Jabs, E. W., et al. (2013).

Postnatal brain and skull growth in an Apert syndrome mouse model. American Journal of

Medical Genetics Part A, 161(4), 745–757. http://doi.org/10.1002/ajmg.a.35805

Hinrichsen, K. (1985a). The Early Development of Morphology and Patterns of the Face in the

Human Embryo. Springer.

Hinrichsen, K. (1985b). The early development of morphology and patterns of the face in the

122

human embryo. Advances in Anatomy, Embryology, and Cell Biology, 98.

Holzleitner, I. J., & Perrett, D. I. (2016). Perception of strength from 3D faces is linked to facial

cues of physique. Evolution and Human Behavior, 37(3), 217–229.

http://doi.org/10.1016/j.evolhumbehav.2015.11.004

Hopman, S. M. J., Merks, J. H. M., Suttie, M., Hennekam, R. C. M., & Hammond, P. (2014). Face

shape differs in phylogenetically related populations. European Journal of Human Genetics,

22(11), 1268–1271. http://doi.org/10.1038/ejhg.2013.289

Hoyme, H. E., Hoyme, D. B., Elliott, A. J., Blankenship, J., Kalberg, W. O., Buckley, D., et al.

(2015). A South African mixed race lip/philtrum guide for diagnosis of fetal alcohol spectrum

disorders. American Journal of Medical Genetics Part A, 167A(4), 752–755.

http://doi.org/10.1002/ajmg.a.37023

Hunley, K. L., Healy, M. E., & Long, J. C. (2009). The global pattern of gene identity variation

reveals a history of long-range migrations, bottlenecks, and local mate exchange: implications

for biological race. American Journal of Physical Anthropology, 139(1), 35–46.

http://doi.org/10.1002/ajpa.20932

Huxley, J. S. (1932). Problems of Relative Growth. New York: Dial Press.

Ito, T., Nishimura, T., & Takai, M. (2011). Allometry and Interspecific Differences in the Facial

Cranium of Two Closely Related Macaque Species. Anatomy Research International, 2011(2-

3), 1–7. http://doi.org/10.1155/2011/849751

Jamniczky, H. A., & Hallgrimsson, B. (2009). A comparison of covariance structure in wild and

laboratory muroid crania. Evolution, 63(6), 1540–1556. http://doi.org/10.1111/j.1558-

5646.2009.00651.x

Jayaratne, Y. S. N., Elsharkawi, I., Macklin, E. A., Voelz, L., Weintraub, G., Rosen, D., & Skotko,

123

B. G. (2017). The facial morphology in Down syndrome: A 3D comparison of patients with

and without obstructive sleep apnea. American Journal of Medical Genetics Part A, 138a(3),

247. http://doi.org/10.1002/ajmg.a.38399

Jerison, H. J. (1973). Evolution of the Brain and Intelligence. New York and London: Academic

Press. Retrieved from h

Jiang, R., Bush, J. O., & Lidral, A. C. (2006). Development of the upper lip: Morphogenetic and

molecular mechanisms. Developmental Dynamics, 235(5), 1152–1166.

http://doi.org/10.1002/dvdy.20646

Johnson, E. L., Roberts, M. W., Guckes, A. D., Bailey, L. J., Phillips, C. L., & Wright, J. T. (2002).

Analysis of craniofacial development in children with hypohidrotic ectodermal dysplasia.

American Journal of Medical Genetics Part A, 112(4), 327–334.

http://doi.org/10.1002/ajmg.10654

Jolicoeur, P. (1963). The Multivariate Generalization of the Allometry Equation. Biometrics,

19(3), 497–499.

Jonke, E., Prossinger, H., Bookstein, F. L., Schaefer, K., Bernhard, M., & Freudenthaler, J. W.

(2008). Secular trends in the European male facial skull from the Migration Period to the

present: a cephalometric study. The European Journal of Orthodontics, 30(6), 614–620.

http://doi.org/10.1093/ejo/cjn065

Katsube, M., Yamada, S., Miyazaki, R., Yamaguchi, Y., Makishima, H., Takakuwa, T., et al.

(2017). Quantitation of nasal development in the early prenatal period using geometric

morphometrics and MRI: a new insight into the critical period of Binder phenotype. Prenatal

Diagnosis, 17, 438. http://doi.org/10.1002/pd.5106

Kehrer-Sawatzki, H., Mautner, V.-F., & Cooper, D. N. (2017). Emerging genotype-phenotype

124

relationships in patients with large NF1 deletions. Human Genetics, 136(4), 349–376.

http://doi.org/10.1007/s00439-017-1766-y

Kerameddin, S., Namipashaki, A., Ebrahimi, S., & Ansari-Pour, N. (2015). IRF6 Is a Marker of

Severity in Nonsyndromic Cleft Lip/Palate. Journal of Dental Research, 94(9 Suppl), 226S–

32S. http://doi.org/10.1177/0022034515581013

Klimentidis, Y. C., & Shriver, M. D. (2009). Estimating Genetic Ancestry Proportions from Faces.

PLoS ONE, 4(2), e4460. http://doi.org/10.1371/journal.pone.0004460

Klingenberg, C. P. (1998a). Heterochrony and allometry: the analysis of evolutionary change in

ontogeny. Biological Reviews, 73(1), 79–123. http://doi.org/10.1111/j.1469-

185X.1997.tb00026.x

Klingenberg, C. P. (1998b). Heterochrony and allometry: the analysis of evolutionary change in

ontogeny. Biol Rev Camb Philos Soc, 73(1), 79–123.

Klingenberg, C. P. (2011). MorphoJ: an integrated software package for geometric morphometrics.

Molecular Ecology Resources, 11(2), 353–357. http://doi.org/10.1111/j.1755-

0998.2010.02924.x

Klingenberg, C. P. (2016). Size, shape, and form: concepts of allometry in geometric

morphometrics. Development Genes and Evolution, 226(3), 113–137.

http://doi.org/10.1007/s00427-016-0539-2

Klingenberg, C. P., & Marugán-Lobón, J. (2013). Evolutionary covariation in geometric

morphometric data: analyzing integration, modularity, and allometry in a phylogenetic

context. Systematic Biology, 62(4), 591–610. http://doi.org/10.1093/sysbio/syt025

Klingenberg, C. P., & Zimmerman, M. (1992). Static, Ontogenetic, and Evolutionary Allometry:

A Multivariate Comparison in Nine Species of Water Striders. The American Naturalist,

125

140(4), 601–620.

Kobielak, K., Kobielak, A., Roszkiewicz, J., Wierzba, J., Limon, J., & Trzeciak, W. H. (2001).

Mutations in the EDA gene in three unrelated families reveal no apparent correlation between

phenotype and genotype in the patients with an X-linked anhidrotic ectodermal dysplasia.

American Journal of Medical Genetics Part A, 100(3), 191–197.

http://doi.org/10.1002/ajmg.1225

Kosowski, T., Weathers, W., Wolfswinkel, E., & Ridgway, E. (2013). Cleft Palate. Seminars in

Plastic Surgery, 26(04), 164–169. http://doi.org/10.1055/s-0033-1333883

Kumar S, D., Gopalkrishnan, K., Bhasker Rao, C., & Ganeshkar, S. V. (2010). Evaluation of

Parental Nasomaxillary Asymmetry as a Risk Factor for Development of Palatal Clefts in their

Offsprings. International Journal of Clinical Pediatric Dentistry, 3(3), 163–171.

http://doi.org/10.5005/jp-journals-10005-1072

Levaillant, J. M., Bault, J. P., Benoit, B., & Couly, G. (2017). Normal and Abnormal Fetal Face

Atlas: Ultrasonographic Features.

Lewontin, R. C. (1972). The Apportionment of Human Diversity. Evolutionary Biology, (Chapter

14), 381–398. http://doi.org/10.1007/978-1-4684-9063-3_14

Lexner, M. O., Bardow, A., Bjorn Jorgensen, J., Hertz, J. M., Almer, L., & Kreiborg, S. (2007).

Anthropometric and cephalometric measurements in X-linked hypohidrotic ectodermal

dysplasia. Orthodontics & Craniofacial Research, 10(4), 203–215.

http://doi.org/10.1111/j.1601-6343.2007.00402.x

Li, M., Cole, J. B., Manyama, M., Larson, J. R., Liberton, D. K., Riccardi, S. L., et al. (2017).

Rapid automated landmarking for morphometric analysis of three-dimensional facial scans.

Journal of Anatomy, 230(4), 607–618. http://doi.org/10.1111/joa.12576

126

Lieberman, D. E., Carlo, J., Ponce de León, M., & Zollikofer, C. P. E. (2007). A geometric

morphometric analysis of heterochrony in the cranium of chimpanzees and bonobos. Journal

of Human Evolution, 52(6), 647–662. http://doi.org/10.1016/j.jhevol.2006.12.005

Lieberman, D. E., Hallgrimsson, B., Liu, W., Parsons, T. E., & Jamniczky, H. A. (2008). Spatial

packing, cranial base angulation, and craniofacial shape variation in the mammalian skull:

testing a new model using mice. Journal of Anatomy, 212(6), 720–735.

http://doi.org/10.1111/j.1469-7580.2008.00900.x

Magwene, P. M. (2001). Comparing ontogenetic trajectories using growth process data. Systematic

Biology.

Manyama, M., Rolian, C., Gilyoma, J., Magori, C. C., Mjema, K., Mazyala, E., et al. (2011). An

assessment of orofacial clefts in Tanzania. BMC Oral Health, 11(1), 5.

http://doi.org/10.1186/1472-6831-11-5

Marcucio, R. S., Young, N. M., Hu, D., & Hallgrimsson, B. (2011). Mechanisms that underlie co-

variation of the brain and face. Genesis (New York, N.Y. : 2000), 49(4), 177–189.

http://doi.org/10.1002/dvg.20710

Marcucio, R., Hallgrimsson, B., & Young, N. M. (2015). Facial Morphogenesis: Physical and

Molecular Interactions Between the Brain and the Face. Current Topics in Developmental

Biology, 115, 299–320. http://doi.org/10.1016/bs.ctdb.2015.09.001

Martínez-Abadías, N., Mitteroecker, P., Parsons, T. E., Esparza, M., Sjøvold, T., Rolian, C., et al.

(2012). The Developmental Basis of Quantitative Craniofacial Variation in Humans and Mice.

Evolutionary Biology, 39(4), 554–567. http://doi.org/10.1007/s11692-012-9210-7

Marugán-Lobón, J., Watanabe, A., & Kawabe, S. (2016). Studying avian encephalization with

geometric morphometrics. Journal of Anatomy, 229(2), 191–203.

127

http://doi.org/10.1111/joa.12476

Mayer, C., Windhager, S., Schaefer, K., & Mitteroecker, P. (2017). BMI and WHR Are Reflected

in Female Facial Shape and Texture: A Geometric Morphometric Image Analysis. PLoS ONE,

12(1), e0169336. http://doi.org/10.1371/journal.pone.0169336

McIntyre, G. T., & Mossey, P. A. (2002). Asymmetry of the parental craniofacial skeleton in

orofacial clefting. Journal of Orthodontics, 29(4), 299–305– discussion 278–9.

http://doi.org/10.1093/ortho/29.4.299

Meindl, K., Windhager, S., Wallner, B., & Schaefer, K. (2012). Second-to-fourth digit ratio and

facial shape in boys: the lower the digit ratio, the more robust the face. Proceedings of the

Royal Society B: Biological Sciences, 279(1737), rspb20112351–2463.

http://doi.org/10.1098/rspb.2011.2351

Mikkola, M. L. (2009). Molecular aspects of hypohidrotic ectodermal dysplasia. American Journal

of Medical Genetics Part A, 149A(9), 2031–2036. http://doi.org/10.1002/ajmg.a.32855

Mitteroecker, P., Gunz, P., & Windhager, S. (2013). A brief review of shape, form, and allometry

in geometric morphometrics, with applications to human facial morphology. Hystrix.

Mladina, R., Skitarelić, N. B., & Skitarelić, N. P. (2009). Could we prevent unilateral cleft

lip/palate in the future? Medical Hypotheses, 73(4), 601–603.

http://doi.org/10.1016/j.mehy.2009.05.028

Mladina, R., Skitarelić, N., Vuković, K., Subarić, M., Carić, T., & Orihovac, Z. (2008). Unilateral

cleft lip/palate children: the incidence of type 6 septal deformities in their parents. Journal of

Cranio-Maxillo-Facial Surgery : Official Publication of the European Association for Cranio-

Maxillo-Facial Surgery, 36(6), 335–340. http://doi.org/10.1016/j.jcms.2008.03.003

Montonen, O., Ezer, S., Saarialho-Kere, U. K., Herva, R., Karjalainen-Lindsberg, M.-L., Kaitila,

128

I., et al. (1998). The Gene Defective in Anhidrotic Ectodermal Dysplasia Is Expressed in the

Developing Epithelium, Neuroectoderm, Thymus, and Bone. Journal of Histochemistry &

Cytochemistry, 46(3), 281–289. http://doi.org/10.1177/002215549804600301

Moreno Uribe, L. M., Fomina, T., Munger, R. G., Romitti, P. A., Jenkins, M. M., Gjessing, H. K.,

et al. (2017). A Population-Based Study of Effects of Genetic Loci on Orofacial Clefts.

Journal of Dental Research, 35(6), 22034517716914.

http://doi.org/10.1177/0022034517716914

Mosimann, J. E. (1958). An analysis of allometry in the chelonian shell. Revue Canadienne De

Biologie, 17(2), 137–228.

Mossey, P. A., Little, J., Munger, R. G., Dixon, M. J., & Shaw, W. C. (2009). Cleft lip and palate.

Lancet (London, England), 374(9703), 1773–1785. http://doi.org/10.1016/S0140-

6736(09)60695-4

Muggli, E., Matthews, H., Penington, A., Claes, P., O'Leary, C., Forster, D., et al. (2017).

Association Between Prenatal Alcohol Exposure and Craniofacial Shape of Children at 12

Months of Age. JAMA Pediatrics, 171(8), 771–780.

http://doi.org/10.1001/jamapediatrics.2017.0778

Mydlová, M., Dupej, J., Koudelová, J., & Velemínská, J. (2015). Sexual dimorphism of facial

appearance in ageing human adults: A cross-sectional study. Forensic Science International,

257, 519.e1–519.e9. http://doi.org/10.1016/j.forsciint.2015.09.008

Nakasima, A., & Ichinose, M. (1983). Characteristics of craniofacial structures of parents of

children with cleft lip and/or palate. American Journal of Orthodontics, 84(2), 140–146.

http://doi.org/10.1016/0002-9416(83)90178-1

Nikopensius, T., Kempa, I., Ambrozaitytė, L., Jagomägi, T., Saag, M., Matulevičienė, A., et al.

129

(2011). Variation in FGF1, FOXE1, and TIMP2 genes is associated with nonsyndromic cleft

lip with or without cleft palate. Birth Defects Research Part a: Clinical and Molecular

Teratology, 91(4), 218–225. http://doi.org/10.1002/bdra.20791

Olson, E. C., & Miller, R. L. (1958). Morphological Integration. Chicago: University of Chicago

Press.

Parsons, T. E., Downey, C. M., Jirik, F. R., Hallgrimsson, B., & Jamniczky, H. A. (2015). Mind

the gap: genetic manipulation of basicranial growth within synchondroses modulates calvarial

and facial shape in mice through epigenetic interactions. PLoS ONE, 10(2), e0118355.

http://doi.org/10.1371/journal.pone.0118355

Parsons, T. E., Schmidt, E. J., Boughner, J. C., Jamniczky, H. A., Marcucio, R. S., & Hallgrimsson,

B. (2011). Epigenetic integration of the developing brain and face. Developmental Dynamics,

240(10), 2233–2244. http://doi.org/10.1002/dvdy.22729

Pavlicev, M., Cheverud, J. M., & Wagner, G. P. (2009). Measuring Morphological Integration

Using Eigenvalue Variance. Evolutionary Biology, 36(1), 157–170.

http://doi.org/10.1007/s11692-008-9042-7

Pedersen, D. A., Wehby, G. L., Murray, J. C., & Christensen, K. (2016). Psychiatric Diagnoses in

Individuals with Non-Syndromic Oral Clefts: A Danish Population-Based Cohort Study. PLoS

ONE, 11(5), e0156261. http://doi.org/10.1371/journal.pone.0156261

Phalane, K. G., Tribe, C., Steel, H. C., Cholo, M. C., & Coetzee, V. (2017). Facial appearance

reveals immunity in African men. Scientific Reports, 7(1), 7443.

http://doi.org/10.1038/s41598-017-08015-9

Porto, A., de Oliveira, F. B., Shirai, L. T., De Conto, V., & Marroig, G. (2009). The Evolution of

Modularity in the Mammalian Skull I: Morphological Integration Patterns and Magnitudes.

130

Evolutionary Biology, 36(1), 118–135. http://doi.org/10.1007/s11692-008-9038-3

R Development Core Team. (2015). R: A language and environment for statistical computing.

Retrieved December 19, 2016, from http://www.R-project.org

Relethford, J. H. (2002). Apportionment of global human genetic diversity based on craniometrics

and skin color. American Journal of Physical Anthropology, 118(4), 393–398.

http://doi.org/10.1002/ajpa.10079

Richtsmeier, J. T., & DeLeon, V. B. (2009). Morphological integration of the skull in craniofacial

anomalies. Orthodontics & Craniofacial Research, 12(3), 149–158.

http://doi.org/10.1111/j.1601-6343.2009.01448.x

Richtsmeier, J. T., Zumwalt, A., Carlson, E. J., Epstein, C. J., & Reeves, R. H. (2002). Craniofacial

phenotypes in segmentally trisomic mouse models for Down syndrome. American Journal of

Medical Genetics, 107(4), 317–324. http://doi.org/10.1002/ajmg.10175

Roberts, R. M. (2014). Discrimination among adults with craniofacial conditions. The Journal of

Craniofacial Surgery, 25(1), 77–81. http://doi.org/10.1097/SCS.0b013e3182a2e005

Rohlf, F. J. (1999). Shape Statistics: Procrustes Superimpositions and Tangent Spaces. Journal of

Classification, 16(2), 197–223. http://doi.org/10.1007/s003579900054

Rohlf, F. J., & Slice, D. (1990). Extensions of the Procrustes method for the optimal

superimposition of landmarks. Systematic Biology.

Saksena, S. S., & Bixler, D. (1990). Facial morphometrics in the identification of gene carriers of

X-linked hypohidrotic ectodermal dysplasia. American Journal of Medical Genetics Part A,

35(1), 105–114. http://doi.org/10.1002/ajmg.1320350120

Schlager, S., Jefferis, G., & Schlager, M. S. (2016). Package “Morpho.”

Schmidt-Nielsen, K. (1984). Scaling: Why is Animal Size So Important? Cambridge and New

131

York: Cambridge University Press. Retrieved from

https://books.google.ca/books?hl=en&lr=&id=8WkjD3L_avQC&oi=fnd&pg=PR11&ots=oS

m86kqXZR&sig=J1BLvSgbSQnNTtsBPqa64Wn6p8w&redir_esc=y#v=onepage&q&f=fals

Schoenwolf, G. C., & Bleyl, S. B. (2009). Development of the pharyngeal apparatus and face.

Larsen's human ….

Shaffer, J. R., Orlova, E., Lee, M. K., Leslie, E. J., Raffensperger, Z. D., Heike, C. L., et al. (2016).

Genome-Wide Association Study Reveals Multiple Loci Influencing Normal Human Facial

Morphology. PLoS Genetics, 12(8), e1006149. http://doi.org/10.1371/journal.pgen.1006149

Sivertsen, A., Wilcox, A., Johnson, G. E., Abyholm, F., Vindenes, H. A., & Lie, R. T. (2008).

Prevalence of major anatomic variations in oral clefts. Plastic and Reconstructive Surgery,

121(2), 587–595. http://doi.org/10.1097/01.prs.0000297839.78179.f5

Sözen, M. A., Suzuki, K., Tolarova, M. M., Bustos, T., Fernández Iglesias, J. E., & Spritz, R. A.

(2001). Mutation of PVRL1 is associated with sporadic, non-syndromic cleft lip/palate in

northern Venezuela. Nature Genetics, 29(2), 141–142. http://doi.org/10.1038/ng740

Sperber, G. H. (2002). Formation of The Primary and Secondary Palate. In D. F. Wyszynski (Ed.),

Cleft lip and palate: from origin to treatment. (pp. 5–24). New York.

Sperber, G. H., Sperber, S. M., & Guttmann, G. D. (2010). Craniofacial Embryogenetics and

Development. PMPH-USA.

Starbuck, J. M., Ghoneima, A., & Kula, K. (2015). Bilateral cleft lip and palate: A morphometric

analysis of facial skeletal form using cone beam computed tomography. Clinical Anatomy,

28(5), 584–592. http://doi.org/10.1002/ca.22530

Stock, N. M., Feragen, K. B., & Rumsey, N. (2015). “It Doesn't All Just Stop at 18”: Psychological

132

Adjustment and Support Needs of Adults Born With Cleft Lip and/or Palate. Dx.Doi.org,

52(5), 543–554. http://doi.org/10.1597/14-178

Suttie, M., Wetherill, L., Jacobson, S. W., Jacobson, J. L., Hoyme, H. E., Sowell, E. R., et al.

(2017). Facial Curvature Detects and Explicates Ethnic Differences in Effects of Prenatal

Alcohol Exposure. Alcoholism, Clinical and Experimental Research, 41(8), 1471–1483.

http://doi.org/10.1111/acer.13429

Velemínská, J., Bigoni, L., Krajíček, V., Borský, J., Šmahelová, D., Cagáňová, V., & Peterka, M.

(2012). Surface facial modelling and allometry in relation to sexual dimorphism. HOMO-

Journal of Comparative Human Biology, 63(2), 81–93.

http://doi.org/10.1016/j.jchb.2012.02.002

Viðarsdóttir, U. S., O’Higgins, P., & Stringer, C. (2002). A geometric morphometric study of

regional differences in the ontogeny of the modern human facial skeleton. Journal of Anatomy,

201(3), 211–229. http://doi.org/10.1046/j.1469-7580.2002.00092.x

Viđarsdóttir, U. S., & O’Higgins, P. (2003). Developmental variation in the facial skeleton of

anatomically modern Homo sapiens. In Patterns of Growth and Development in the Genus

Homo (pp. 114–143).

Wagner, G. P. (1984). On the eigenvalue distribution of genetic and phenotypic dispersion

matrices: Evidence for a nonrandom organization of quantitative character variation. Journal

of Mathematical Biology, 21(1), 77–95. http://doi.org/10.1007/BF00275224

Wang, K. H., Heike, C. L., Clarkson, M. D., Mejino, J. L. V., Brinkley, J. F., Tse, R. W., et al.

(2014). Evaluation and integration of disparate classification systems for clefts of the lip.

Frontiers in Physiology, 5, 2. http://doi.org/10.3389/fphys.2014.00163

Wang, Q., Shang, L., & Fang, Y. (2012). [Research on the mental health of the parents of cleft lip

133

and palate patients and social impact factor from the perspective of social network]. Hua Xi

Kou Qiang Yi Xue Za Zhi = Huaxi Kouqiang Yixue Zazhi = West China Journal of

Stomatology, 30(4), 374–379.

Weinberg, S. M., & Kolar, J. C. (2005). Three-dimensional surface imaging: limitations and

considerations from the anthropometric perspective. Journal of Craniofacial Surgery, 16(5),

847–851.

Weinberg, S. M., Neiswanger, K., Richtsmeier, J. T., Maher, B. S., Mooney, M. P., Siegel, M. I.,

& Marazita, M. L. (2008). Three-dimensional morphometric analysis of craniofacial shape in

the unaffected relatives of individuals with nonsyndromic orofacial clefts: a possible marker

for genetic susceptibility. American Journal of Medical Genetics Part A, 146A(4), 409–420.

http://doi.org/10.1002/ajmg.a.32177

Weinberg, S. M., Parsons, T. E., Fogel, M. R., Walter, C. P., Conrad, A. L., & Nopoulos, P. (2013).

Corpus callosum shape is altered in individuals with nonsyndromic cleft lip and palate.

American Journal of Medical Genetics Part A, 161A(5), 1002–1007.

http://doi.org/10.1002/ajmg.a.35835

Werner, H., Lopes Dos Santos, J. R., Ribeiro, G., & Araujo Júnior, E. (2017). Prenatal Phenotype

of Down Syndrome Using Three-Dimensional Virtual Reality. Journal of Obstetrics and

Gynaecology Canada : JOGC = Journal D'obstetrique Et Gynecologie Du Canada : JOGC.

http://doi.org/10.1016/j.jogc.2017.03.100

Wiley, D. F., Amenta, N., Alcantara, D. A., Ghosh, D., Kil, Y. J., Delson, E., et al. (2005).

Evolutionary Morphing (pp. 431–438). Presented at the VIS 05. IEEE Visualization, 2005.,

IEEE. http://doi.org/10.1109/VISUAL.2005.1532826

Wilhoit, L. F., Scott, D. A., & Simecka, B. A. (2017). Fetal Alcohol Spectrum Disorders:

134

Characteristics, Complications, and Treatment. Community Mental Health Journal, 53(6),

711–718. http://doi.org/10.1007/s10597-017-0104-0

Windhager, S., Patocka, K., & Schaefer, K. (2013). Body fat and facial shape are correlated in

female adolescents. American Journal of Human Biology, 25(6), 847–850.

http://doi.org/10.1002/ajhb.22444

Young, N. M., Chong, H. J., Hu, D., Hallgrimsson, B., & Marcucio, R. S. (2010). Quantitative

analyses link modulation of sonic hedgehog signaling to continuous variation in facial growth

and shape. Development, 137(20), 3405–3409. http://doi.org/10.1242/dev.052340

Young, N. M., Sherathiya, K., Gutierrez, L., Nguyen, E., Bekmezian, S., Huang, J. C., et al. (2016).

Facial surface morphology predicts variation in internal skeletal shape. American Journal of

Orthodontics and Dentofacial Orthopedics : Official Publication of the American Association

of Orthodontists, Its Constituent Societies, and the American Board of Orthodontics, 149(4),

501–508. http://doi.org/10.1016/j.ajodo.2015.09.028

Young, N. M., Wat, S., Diewert, V. M., Browder, L. W., & Hallgrimsson, B. (2007). Comparative

morphometrics of embryonic facial morphogenesis: implications for cleft-lip etiology.

Anatomical Record (Hoboken, N.J. : 2007), 290(1), 123–139. http://doi.org/10.1002/ar.20415

Zelditch, M. L., Swiderski, D. L., & Sheets, H. D. (2012). Geometric Morphometrics for

Biologists. Academic Press.

Zhang, J., Han, D., Song, S., Wang, Y., Zhao, H., Pan, S., et al. (2011). Correlation between the

phenotypes and genotypes of X-linked hypohidrotic ectodermal dysplasia and non-syndromic

hypodontia caused by ectodysplasin-A mutations. European Journal of Medical Genetics,

54(4), e377–e382. http://doi.org/10.1016/j.ejmg.2011.03.005

135

Appendix

“Permission is therefore not required for academic or commercial reuse, provided that full attribution is included in the new work”.

136