THE SPATIAL AND TEMPORAL ROLE OF EZH2 IN SKULL BONE FORMATION

by

JAMES W. FERGUSON

Submitted in partial fulfillment of the requirements of the degree of Doctor of Philosophy

Department of Biology

CASE WESTERN RESERVE UNIVERSITY

August 2018

CASE WESTERN RESERVE UNIVERSITY SCHOOL OF GRADUATE STUDIES

We hereby approve the dissertation of James Ferguson candidate for the degree of Doctor of

Philosophy.

Committee Chair: Brian McDermott, PhD

Radhika Atit, PhD

Emmitt K. Jolly, PhD

Stephen E. Haynesworth, PhD

Veronique Lefebvre, PhD

Peter Harte, PhD

Date of Defense: April 18, 2018

*We also certify that written approval has been obtained for any proprietary material contained therein.

Completion of this dissertation was not feat I accomplished alone. I would like to dedicate this

to all the people who made this possible through both emotional and intellectual support.

i

Table of Contents

Abstract: Chapter 1: Introduction and significance 1.1: Anatomy and development of bone: 1.1a: Structure and function of bone 1.1b: Skeletal bone formation 1.1c: Anatomy and structure of the skull bones 1.2: Cellular and molecular mechanisms of skull bone development: 1.2a: Stem cell origins of the skull bones 1.2b: Embryonic developmental timeline of cranial mesenchyme bone 1.3 Transcriptional regulation of the bone initiation program: 1.3a: MshHomeobox 1 (Msx1) and MshHomeobox 2 (Msx2) mark bone precursors 1.3b: Runt Related Transcription Factor 2 (Runx2) and Osterix (Sp7/OSX) are the key determinants of skull bone formation: 1.4: Regulation of molecular signal during of murine skull bone formation: 1.4a: Fibroblast Growth Factor signaling is preferentially active in the CNCC: 1.4b: The Bone Morphogenic Protein signaling pathway is primarily active in the PM 1.4c: Wnt/β-catenin signaling pathway is required throughout skull bone formation 1.4d: Twist Family BHLH Transcription Factor 1 (Twist1) is regulated by Wnt/β-catenin signaling and required for skull bone formation: 1.4e: Retinoic Acid signaling affects many different signaling pathways required for skull bone formation 1.4f: Epigenetic regulation of skull bone development 1.4g: The in vivo role of PRC2 in craniofacial development 1.5: Functional interaction and regulation of bone transcriptional program by BMP, FGF, Wnt/β- catenin, RA, and PRC2: 1.6: Clinical significance:

Chapter 2: PRC2 is Dispensable in Vivo for β-Catenin-Mediated Repression of Chondrogenesis in Mouse Embryonic Cranial Mesenchyme Abstract: 2.1: Introduction: 2.2: Results: 2.2a: Genes dysregulated upon loss of β-catenin are enriched for the PRC2-associated H3K27me3 histone mark 2.2b: Chondrocyte fate genes are enriched for H3K27me3 in the embryonic CM 2.2c: Endogenous β-catenin and EZH2 may physically interact in the CM 2.2d: β-catenin is not required for PRC2 component expression or bulk H3K27me3 levels 2.2e: Loss of Ezh2 does not lead to ectopic cell type fate selection or chondrogenesis in the CM 2.2f: Loss of β-catenin does not significantly alter H3K27me3 enrichment genome-wide 2.2g: H3K27me3 enrichment is not depleted on ectopically expressed chondrocytic gene determinants in β-catenin mutants ii

2.3: Discussion: 2.4: Materials and methods: 2.4a: Mice and genotyping 2.4b: Cranial mesenchyme isolation 2.4c: RT-qPCR 2.4d: Immunofluorescence 2.4e: Co-immunoprecipitation 2.4f: Protein Isolation and Immunoblotting 2.4g: RNA sequencing 2.4h: ChIP-sequencing 2.4i: Cell Culture 2.4j: Statistics 2.4k: Data Availability 2.5: Acknowledgments:

Chapter 3: Ezh2 is required for skull bone formation in a tissue- and developmental-stage specific manner Abstract 3.1: Introduction: 3.2: Results: 3.2a: Conditional deletion of Ezh2in cranial mesenchyme stem cells 3.2b: E8.5-CMEzh2mutants have decreased craniofacial bone volume and size 3.2c: The effect of Ezh2 on skull bone formation is developmental-stage specific 3.2d: Diminished cranial bones in E8.5-CMEzh2 mutants is not due to defects in cell survival and proliferation at E10.5 3.2e: E8.5-CMEzh2 mutants exhibit defects in the differentiation of the skull bone progenitors 3.2f: In vivo inhibition of Retinoic Acid signaling partially restores skull bones 3.3: Discussion: 3.3a: The role of Ezh2 in the lineage-restriction of cranial bone progenitors 3.3b: The genetic mechanism involved in lineage selection of the cranial bones by Ezh2 3.3c: Ectopic expression of Hox genes may lead to a reduction of skull bone in E8.5- CMEzh2 mutants 3.3d: An upregulation of Hedgehog signaling could lead to a reduction of the skull bones E8.5-CMEzh2 mutants 3.4: Materials and methods:

3.4a: Mice and genotyping 3.4b: Histology, β-Galactosidase, and Immunohistochemistry 3.4c: RT-qPCR 3.4d: Protein Isolation and immunoblotting 3.4e: MicroCT: 3.4f: Whole mount skeletal preparation: 3.4g: Cell Proliferation/Death Assay iii

3.4h: Retinoic acid inhibition 3.4i: Statistics 3.5: Acknowledgements: 3.6 Conflict of interest:

Chapter 4: Discussion: Discussion outline: 4.1: β-catenin and its role in the suppression of chondrogenesis 4.1a: Possible repression of chondrogenesis mediated by TWIST1 4.1b: Possible suppression of Sox9 through unliganded RARs 4.1c: Possible inhibition of chondrogensis through β-catenin-SOX9 protein-protein interactions 4.1d: Possible repression of Sox9 by direct β-catenin binding 4.2: Insights into the repressive functions of β-catenin: 4.3: The mechanisms governing the developmental stage and cell type specific role of Ezh2: 4.3a: Ezh2 is required for lineage selection of the cranial bone prior to the expression of Msx genes 4.3b: Canonical vs. non-canonical function of EZH2 4.4: Ezh2 regulates skull bone formation by inhibiting a target of the RA signaling pathway: 4.4a: Hedgehog signaling and Hox genes are potential targets of RA signaling 4.5: The correlation of H3K27me3 enrichment with mRNA expression: 4.6: Conclusion:

iv

List of Figures:

Chapter 1: Figure 1.1: Comparative anatomy of the mouse and human skull. Figure 1.2: The migration of cranial neural crest cells from the neural tube. Figure 1.3: Spatial organization of the mesenchymal stem cell origins and bone primordia. Figure 1.4: The bone initiation program and cell fate selection of the skull bone. Figure 1.5: Temporal differences in the bone initiation program between the CNCC-CM and PM- CM Figure 1.6: Signaling mechanisms regulating the bone initiation program in the CNCC-CM and PM-CM. Figure 1.7: Canonical Wnt/β-catenin signaling: Figure 1.8: Enrichment of H3K27me3 by PRC2 leads to transcriptional repression.

Chapter 2: Figure 2.1: β-catenin regulates known PRC2 targets and can physically interact with PRC2. Figure S2.1: Genomic Regions Enrichment of Annotations Tool (GREAT) analysis of differentially expressed genes upon deletion of β-catenin. Figure S2.2: Genes regulated by β-catenin are enriched for PRC2 targets in dorsal dermal fibroblasts. Figure 2.2: β-catenin activity is not required for the expression of PRC2 components and bulk H3K27me3 levels. Figure S2.3: Knockout of Ezh2 with En1Cre does not lead to a change in H3K27me3 enrichment in the CM. Figure 2.3: Knockdown of Ezh2 in the cranial mesenchyme does not lead to changes in cell type fate selection. Figure 2.4: Chemical inhibition of EZH2 methyltransferase does not lead to an up-regulation of early chondrocyte markers in CM+ectoderm. Figure 2.5: Loss of β-catenin does not significantly alter H3K27me3 enrichment genome wide or on cartilage differentiation determinants. Figure S2.4: Analysis of strong, medium, and weak H3K27me3 enrichment in the CM+ectoderm. Figure S2.5: Gene ontology analysis of strong, medium, and weak H3K27me3 enrichment. Figure S2.6: Strong, medium, and weak H3K27me3 peak groups have similar representation in genes regulated by β-catenin. Figure S2.7: H3K27me3 peaks are associated with genes actively transcribed and repressed in CM.

Chapter 3: Figure 3.1: Tamoxifen induced knockout of Ezh2at E8.5 in both the mesoderm- and neural crest-derived mesenchymal stem cells is sufficient to lead to craniofacial defects. Figure S3.1: Whole mount analysis of E13.5 and E17.5 E8.5-CMEzh2 mutants. Figure 3.2: e8.5-CMEzh2leads to a reduction of CNCC-derived bones and a severe reduction in PM-derived bones. Figure S3.2: Quantification of the mandible and snout in e8.5-CMEzh2 mutants. v

Figure 3.3: Ezh2 is required for skull bone formation in a developmental stage-dependent manner. Figure S3.3: Ezh2 is lost in the same tissue in the E8.5-CMEzh2 and E9.5-CMEzh2 mutant. Figure 3.4: Increased cell death in E8.5-CMEzh2 mutants is insufficient to account for loss of skull bones. Figure S3.4: Significant increase in cell death, but no change in cell proliferation, in the frontonasal process in E8.5-CMEzh2 mutants. Figure 3.5: Loss of Ezh2 at E8.5 leads to defects in bone progenitor differentiation. Figure S3.5: No change in bone marker, alkaline phosphatase (AP) inE8.5-CMEzh2 mutants and OSX in E9.5-CMEzh2mutants. Figure S3.6: E8.5-CMEzh2 mutants exhibit an ectopic expression of the Hox genes. Figure S3.7: The effects of RA signaling disruptions in the CM. Figure 3.6: Chemical inhibition of RA signaling partially rescues the in E8.5-CMEzh2 mutant phenotype and restores the PM-derived bones.

Chapter 4: Figure 4.1: Proposed mechanisms by which β-catenin suppresses chondrogenesis in the cranial mesenchyme. Figure 4.2: Ezh2 is required for the lineage selection of the skull bone precursors prior to the activation of the bone initiation program.

vi

List of Tables:

Chapter 1: Table 1.1: Signaling pathway mutants leading to skull bone defects

Chapter 2: Table 2.1: Summary of publications demonstrating a biological interaction between β-catenin and PRC2.

Chapter 3: Table 3.1: Fold expression changes in E8.5-CMEzh2 mutants relative to controls.

vii

List of Abbreviations:

CM = cranial mesenchyme

CNCC = cranial neural crest

PM = paraxial mesoderm

CNCC-CM = cranial mesenchyme derived from cranial neural crest cells

PM-CM = cranial mesenchyme derived from paraxial mesoderm

viii

Abstract:

The Spatial and Temporal Role of Ezh2 In Skull Bone Formation

JAMES W. FERGUSON

The skull bones are vital anatomical structures providing shape and function to the face, protection to the brain, and housing multiple sensory organs. Derived from two mesenchymal stem cell populations, the cranial neural crest and paraxial mesoderm, the skull bones are some of the few bones that ossify through intramembranous ossification. Intramembranous ossification begins with the bone initiation program; a transcription factor cascade consisting of

Msx 1 and 2, Runx2, and Osx. Coordination and regulation of the bone initiation program has been associated with multiple signaling pathways such as Wnt/β-catenin signaling and PRC2.

Wnt/ β-catenin is a major transcriptional regulatory pathway that primarily acts as a transcriptional activator. In the skull bones, Wnt/β-catenin signaling is required for progression through the bone initiation program and the repression of chondrogensis. PRC2 is an epigenetic modifier required for the repressive H3K27me3 histone modification. Disruption in PRC2 function can lead to various skull bone defects depending on the specific mesenchyme stem cell population and developmental stage in which disruption occurs. Here I examine the spatial and temporal role of PRC2 in skull bone formation and the connection with Wnt/β-catenin to ensure the repression of chondrogenesis. Using in vivo, conditional mouse mutants, we demonstrate that β-catenin does not require PRC2 for the repression of chondrogensis.

Additionally, we show that Ezh2 is required in a spatial and temporal manner to ensure skull bone formation by regulating a target of the retinoic acid signaling pathway. Our data ix demonstrates the dynamic genetic requirements governing skull bone formation at specific developmental stages, and highlights the differential signaling requirements for skull bone formation between the CNCC and PM. These results provide insights into the genetic and epigenetic mechanisms governing skull bone formation.

1

Chapter 1: Introduction and Significance

1.1: Anatomy and development of bone:

1.1a: Structure and function of bone

Bone is a unique and multifunctional tissue. As one of the hardest tissues in the body, it is the basis of the skeletal framework which provides structural form and support. In addition, it is a production center for multiple critical and life sustaining cell types including blood, immune, and fat cells (Zhao et al. 2012). Throughout the body, there are two primary types of bones; long bones and flat bones. The long bones, such as the femur and humerus, are the sources of yellow bone marrow which produces fat cells, cartilage, and additional bone. The flat bones, such as the skull, ribs, and pelvis, are the sources of red bone marrow which primarily produces platelets, red blood cells, and white blood cells. Considering the vital role of bone in human physiology, it is critical to understand the mechanisms behind the formation of bone during embryonic development.

1.1b: Skeletal bone formation

In mammals, bone is formed through two different ossification processes; endochondral and intramembranous. Endochondral ossification is characterized by the formation of a cartilage scaffold prior to ossification. As the bone begins to form, the cartilage acts as a structural support for the osteoprogenitor cells. As the bone develops, the cartilage hypertrophies leaving only fully ossified bone. Endochondral ossification primarily occurs in the trunk and long bones. The skull bones are a unique 2 subset of bone in the body as they are the primary bones that undergo intramembranous ossification. Intramembranous ossification is characterized by the direct conversion of bone progenitors into ossified bone without the presence of cartilage. The formation and ossification of the skull bones requires the coordination of complex genetic mechanisms and has been the primary focus of my graduate studies.

1.1c: Anatomy and structure of the skull bones

The human skull is composed of 22 dense, compact flat bones containing red marrow. In human anatomy, the skull is divided into two primary regions; the neurocranium and the viscerocranium. The bones of the face such as the mandible, maxilla, and nasal bones constitute the viscerocranium. The bones surrounding the brain including the frontal, parietal, interparietal, and occipital bones, constitute the neurocranium (calvarium). In humans, the interparietal bone becomes fused with the occipital bone, but is maintained as a separate bone in the mouse (Fig. 1.1). It is worth noting that these human anatomical 3

Figure 1.1: Comparative anatomy of the mouse and human skull. (A,B) The bones of the mouse (A) and human (B) developing skull. (C,D) The bones of the adult mouse (C) and human (D) skull. The bones of the viscerocranium are outlined in green. The bones of the neurocranium are outlined in magenta. Note: the interparietal bone is fused to the occipital bone in humans. definitions vary slightly relative to the comparative anatomy definitions which divides the skull into three regions; the viscerocranium, neurocranium, and dermatocranium

(Hanken and Thorogood 1993).

During embryonic development, the bones of the viscero- and neurocranium form from different stem cell origins into stereotyped and complex shapes within a small spatial area. Strict spatial and temporal genetic organization is required for the formation of these complex bones resulting in a susceptibility to congenital and 4 environmental perturbations (Trainor 2007). A greater understanding of the cellular and molecular mechanisms guiding skull bone development may provide insights contributing to the advancement of therapeutic and diagnostic applications.

1.2: Cellular and molecular mechanisms of skull bone development:

1.2a: Stem cell origins of the skull bones

The bones of the skull are derived from two distinct stem cell origins during embryonic development; the cranial neural crest cells (CNCC) and the paraxial mesoderm (PM) (Hanken and Thorogood 1993; Jiang et al. 2002; Yoshida et al. 2008;

Fan et al. 2016). The neural tube, flanked by the paraxial mesoderm, gives rise to the

CNCC which then migrate to the head and face prior to embryonic day (E) 8.5 (Fig. 1.2).

Figure 1.2: The migration of cranial neural crest cells from the neural tube. The neural tube and the flanking paraxial mesoderm are located along the dorsal aspect of the developing embryo. Prior to E8.5, the neural crest cells delaminate from the neural tube and migrate ventrally to populate the branchial arches and ventral aspect of the head. 5

After the CNCC and PM populate the head, these two mesenchymal stem cell populations form the head mesenchyme and give rise to the bones of the skull and the overlying dermis. As the embryo develops, the head mesenchyme can be divided into two regions. The facial mesenchyme primarily consists of the 1st branchial arch and frontonasal process which forms the bones of the viscerocranium and is comprised of the CNCC. The cranial mesenchyme consists of the tissue surrounding the brain that give rise to the bones of the neurocranium and is comprised of the CNCC and PM (Fig 1.2, Fig

1.3). The CNCC in the cranial mesenchyme (CNCC-CM) forms the frontal bone and medial portion of the interparietal bone. The PM in the cranial mesenchyme (PM-CM) forms the parietal, occipital, and lateral portion of the interparietal bone (Fig 1.3).

Figure 1.3: Spatial organization of the mesenchymal stem cell origins and bone primordia. (A) Diagram of the spatial organization of the CNCC- and PM-derived mesenchymal stem cells, and the calvarial bone primordia in the mouse embryo. The regions covered by the frontal and parietal bone primordia constitute the supraorbital arch. The yellow dashed regions mark the bone primordia and ossification centers. (B) Skeletal preparation of E14.5 mouse embryo. Bones are stained with alizarin red staining (photo from Paul Trainor lab, www.stowers.org/faculty/trainor-lab). F=frontal bone, P=parietal bone, IP =interparietal bone. The frontal and parietal bone primordia make up the supraorbital arch mesenchyme.

6

An added layer of complexity is added to the genetic mechanisms governing the formation and the characteristics of the skull due to the contribution from two stem cell origins during skull bone development. Despite both the CNCC and PM ossifying through intramembranous ossification, differences in their osteogenic potential has been identified. In an in vivo calvarial healing model, the CNCC-CM has increased osteogenic potential compared to the PM-CM (Quarto et al. 2010). Furthermore, multiple differences in the genetic signaling between the CNCC and PM have been identified demonstrating two unique molecular environments in which the bones of the skull form. Previous studies have primarily focused on the CNCC-derived structures, and the developmental dynamics between the CNCC- and PM-derived bone in the cranial mesenchyme are poorly understood.

1.2b: Embryonic developmental timeline of cranial mesenchyme bone

In mice, between embryonic day E8.5 and E9.5, the CNCC and PM populate the cranial mesenchyme. Between E9.5 and E11.5, the cranial bone primordia condense in two small patches directly above the eye called the supraorbital arch (SOA) mesenchyme (Fig 1.3) (Jiang et al. 2002; Yoshida et al. 2008; Roybal et al. 2010). Based on lineage tracing experiments, after E12.5, the SOA mesenchyme must migrate apically over the brain and form the skull bones (Jiang et al. 2002; Yoshida et al. 2008). The anterior patch gives rise to the frontal bone and is CNCC-derived. The posterior patch gives rise to the parietal bone and is PM-derived. The formation of the neurocranial bones involves two major processes; cell fate selection and migration. The exact genetic mechanisms regulating both of these processes have not been fully elucidated. The 7 genetic mechanisms governing the cell fate selection and differentiation of the skull bones has been the focus of my graduate work and will be the topic of my dissertation.

The transcription factor cascade, which I refer to as the bone initiation program, critical for initiating bone formation consists of three primary factors; MshHomeobox 1

(Msx1) and 2 (Msx2), Runt Related Transcription Factor 2 (Runx2), and Osterix (Osx/Sp7)

(Fig. 1.4).

Figure 1.4: The bone initiation program and cell fate selection of the skull bone. Osteoblast precursors may become bone or dermal cells. The bone transcriptional program begins with the expression of Runx2 from MSX positive cells. The expression of Runx2 demarcates "cell fate selection" of the bone cells. The expression of Sp7 (OSX) demarcates fate "commitment" to bone. The OSX positive cells can then further differentiate into bone. The gene regulatory networks governing these processes is currently under investigation.

Beginning around E10.5, the differences in the bone transcriptional program can be observed between the CNCC-CM and PM-CM (Han et al. 2007). In the CNCC-CM frontal bone primordia, the bone precursor markers Msx1 and Msx2 can be detected at E10.5.

By E12.5, the bone master regulator, Runx2 and its downstream target Osx can be 8 detected. Ossification of the frontal bone, as demonstrated by alizarin red staining, occurs by E14.0. In contrast, activation of the bone initiation program in the PM-CM parietal bone primordia is not detected until E11.5-E12.5 after which Msx2 and Runx2 are expressed. Osx is expressed a day later by E13.5 (Deckelbaum et al. 2012). By E14.5,

Figure 1.5: Temporal differences in the bone initiation program between the CNCC-CM and PM- CM. The CNCC-CM progresses through the bone transcriptional prior to the PM-CM. At E10.5, the expression of Msx 1 and 2 can be detected in only the CNCC-CM. At E12.5, the PM-CM begins to express Msx 2 only. In between E12.5 and E13.5 the CNCC-CM begins to express Runx2 and Osx. At E13.5, the PM-CM begins to express Runx2 and Osx. Ossification of the CNCC-CM can be observed at E14.0 and in the PM-CM at E14.5. ossification of the parietal bone can be detected by alizarin red staining (Fig. 1.5). The specific mechanisms controlling the temporal and spatial expression of these critical transcription factors in the CNCC-CM and PM-CM remain elusive. Current in vivo studies are beginning to reveal the signaling and epigenetics involved in guiding the spatial and temporal dynamics of skull development.

1.3 Transcriptional regulation of the bone initiation program: 9

The bone initiation program begins with bone precursors expressing Msx1 and

Msx2. The bone precursors then undergo cell fate selection into bone progenitors with the expression of Runx2. Expression of Runx2 leads to further bone cell fate commitment with the expression of Sp7 (OSX). Analysis of the bone initiation program in vivo has demonstrated the complex spatial and temporal dynamics behind the genetic mechanisms regulating skull bone formation (Fig. 1.6).

Figure 1.6: Signaling mechanisms regulating the bone initiation program in the CNCC-CM and PM- CM.

1.3a: MshHomeobox 1 (Msx1) and MshHomeobox 2 (Msx2) mark bone precursors:

Msx1 and Msx2 are members of the muscle homeobox family and are transcriptional repressors. Both Msx1 and Msx2 are required for skull bone formation in a temporal and spatial manner. During cell fate selection of the bone precursors, Msx1 and Msx2 are required for the expression of Runx2 and Osx (Ishii et al. 2003; Han et al.

2007). Individual deletions of Msx1 or Msx2, or heterozygous deletions of Msx1 and 10

Msx2 lead to defects in skull bone patterning but never a complete loss of bones.

Homozygous deletion of Msx1 and Msx2, however, leads to complete agenesis of the skull demonstrating a requirement in bone formation (Satokata and Maas 1994;

Satokata et al. 2000; Han et al. 2007). Loss of Msx1 and Msx2 at E9.5 using an inducible

Cre in all β-actin expressing cells (Cagg-CreER;Msxfl/fl ;Msx2fl/fl) results in severely diminished bone. In contrast, loss at E13.5 in Cagg-CreER;Msxfl/fl ;Msx2fl/fl mutants results in no phenotype (Roybal et al. 2010). These results further identify Msx1 and

Msx2 as the initial step in the bone initiation program and highlight an early role in bone development.

The loss of Msx1 and Msx2 in only the CNCC-CM adds a layer of complexity to the bone initiation program. Loss of Msx1 and Msx2 using Wnt1Cre, in which Cre is active at E8.5 in only the CNCC, does not lead to a reduction in either the CNCC- and

PM-derived bones. These mutants exhibit an increased fontanel between the parietal bones, and, unexpectedly, increased bone in the frontal bone (Han et al. 2007; Roybal et al. 2010). These results, when compared to E9.5 Cagg-CreER;Msxfl/fl ;Msx2fl/fl mutants, demonstrates the critical role of PM-CM expression of Msx1 and Msx2, and demonstrates cross-talk between the CNCC-CM and PM-CM during skull bone formation.

Considering the PM-CM does express Msx2 until E12.5, however, it is unclear as to the specific mechanism by which loss of Msx gene expression at E9.5 with Cagg-

CreER;Msxfl/fl ;Msx2fl/fl leads to a loss of skull bones (Fig 1.5). It is possible the bone initiation program is active in the PM-CM earlier in development and a more sensitive 11 readout is required. Further studies conditionally knocking out Msx1 and Msx2 in the

PM-CM only are required to fully elucidate the spatial and temporal requirements for the Msx genes in skull bone formation.

1.3b: Runt Related Transcription Factor 2 (Runx2) and Osterix (Sp7/OSX) are the key determinants of skull bone formation:

Downstream of MSX1 and MSX2 are the bone transcription factors Runx2 and

Osx. Deletion of Runx2 in mice results in a complete loss of bone throughout the entire body indicating that RUNX2 is a master regulator of bone formation (Komori et al.

1997). Downstream of RUNX2 is Osx (Nakashima et al. 2002; Nishio et al. 2006; Baek et al. 2014). In vitro, RUNX2 has been shown to directly bind to Osx and activate its transcription (Nishio et al. 2006). OSX is required for intramembranous ossification.

Unlike the Runx2 deleted mice, which completely lack bone, deletion of Osx leads to a complete loss of ossified bone in the head and face while the trunk and limbs still maintain ossification (Nakashima et al. 2002; Fan et al. 2016). The role of OSX between the CNCC-derived and the PM-derived CM is cell autonomous. Wnt1Cre;Osxfl/fl leads to a complete loss of CNCC-derived bone with almost no effect on the PM-derived bone

(Baek et al. 2013). Considering Runx2 and Osx are downstream of Msx1 and Msx2, which have distinct spatial and temporal requirements in skull bone formation, further studies examining the roles of Runx2 and Osx between the CNCC-CM and the PM-CM are required.

12

1.4: Regulation of molecular signal during of murine skull bone formation:

The bone initiation program describes the specific transcription factors required for the formation of intramembranous bone. These factors must be coordinated and expressed in a spatial and temporal manner between the CNCC-CM and PM-CM. In this regard, multiple signaling pathways critical for embryonic development, such as Bone

Morphogenic Protein (BMP), Fibroblast Growth Factor (FGF), and Wnt/β-catenin signaling, have been identified as regulatory mechanisms in bone formation (Table 1). 13

Table 1: Signaling pathway mutants leading to skull bone defects 14

1)(Cadigan and Nusse 1997; Wang et al. 2014; Ornitz and Itoh 2015). Current in vivo

Studies are beginning to identify the roles of each pathway in the CNCC-CM and the PM-

CM, however, data in the PM-CM is currently lacking.

1.4a: Fibroblast Growth Factor signaling is preferentially active in the CNCC:

FGF signaling involves a large family of signaling factors that play many roles during development (Ornitz and Itoh 2015). In the cranial mesenchyme, many members of the pathway are expressed in both the CNCC-CM and PM-CM (Fan et al. 2016). FGF signaling has been shown to affect the bone initiation program through the regulation of Msx1, Msx2, and Runx2 expression (Kettunen and Thesleff 1998; Abu-Issa et al. 2002;

Kim et al. 2003; Omoteyama and Takagi 2009; Griffin et al. 2013). Overexpression of

Fibroblast growth factor receptor 2 (Fgfr2) at E8.5 in the PM-CM using Mesp1Cre leads to suture fusion but normal looking bone (Holmes and Basilico 2012; Fan et al. 2016).

Overexpression of the FGF signaling antagonist, Sprouty1, at E8.5 in the CNCC-CM using

Wnt1Cre leads to loss of CNCC-derived skull bone and diminished PM-derived bone

(Yang et al. 2010). These results demonstrate a cross-talk between the CNCC-CM and the PM-CM and a sensitivity to FGF signaling in early bone formation. It is difficult, however, to fully implicate FGF signaling in the Sprouty1 mutants requiring further studies.

1.4b: The Bone Morphogenetic Protein signaling pathway is primarily active in the PM:

Bone Morphogenetic Proteins (BMP)s are part of the Transforming Growth

Factor Beta (TGF-β) family. They were initially discovered for their role in endochondral 15 bone formation, but play roles in many developmental processes (Wang et al. 2014). In the cranial mesenchyme, multiple BMP signaling components, such as BMP4, are preferentially expressed in the PM-CM compared to the CNCC-CM (Fan et al. 2016).

However, loss of multiple BMP ligands in CNCC-CM using Wnt1Cre leads to a near complete loss of CNCC-CM bones with little impact on the PM-CM bones indicating the

CNCC-CM is responsive to BMP in a cell autonomous manner (Bonilla-Claudio et al.

2012). In these mutants, a reduction in Msx1 and Msx2 expression was observed indicating disruptions in the bone initiation program. These results demonstrate a cell/tissue autonomous role for BMP signaling in bone formation and indicate it is required for the induction of the bone initiation program. Current data examining BMP activity in the PM-CM is lacking, and further studies utilizing a PM-CM specific knockout of BMP signaling are required.

1.4c: Wnt/β-catenin signaling pathway is required throughout skull bone formation

The Wnt/β-catenin signaling pathway is highly conserved from in the animal kingdom and is critical for embryonic development (Cadigan and Nusse 1997). In the

CM, defects in Wnt/β-catenin signaling lead to dramatic disruptions in skull bone formation (Fan et al. 2016). Defects ranging from a complete loss of head to a total cell fate switch has been observed depending on the timing and tissue in which Wnt/β- catenin is disrupted (Brault et al. 2001; Goodnough et al. 2012).

Canonical Wnt-signaling involves transcriptional regulation through β-catenin. In the absence of Wnt ligands, β-catenin is phosphorylated by the β-catenin degradation 16 complex, composed of Axin2 (AXIN2), Glycogen Synthase Kinase 3 Beta (GSK-3β), Casein

Kinase I (CKI) and Adenomatosis Polyposis Coli (APC), and degraded. Binding of Wnt ligands, which are secreted factors, to LDL Receptor Related Protein 5 (LRP5) and LDL

Receptor Related Protein 6 (LRP6) and Frizzled (FZD) receptors, inhibits the β-catenin degradation complex. β-catenin can then enter the nucleus and act as a transcriptional coactivator of the TCF4/LEf1 transcription factors to regulate gene transcription (Fig.

1.7) (Bhanot et al. 1996; Korinek et al. 1998; Liu et al. 1999; He et al. 2000).

Figure 1.7: Canonical Wnt/β-catenin signaling: The binding of Wnt ligands to the LRP5/6 and FZD receptors inhibits the β-catenin degradation complex (composed of GSK3, APC, CKI, and AXIN2). β- catenin then enters the nucleus to interact with TCF/LEF to regulate gene transcription.

17

Throughout the body, Wnt/β-catenin signaling plays multiple roles in skeletogenesis. Activation of Wnt signaling through the loss of the Wnt inhibitor, Sfrp1, leads to an upregulation of Runx2 in long bones (Gaur et al. 2005). In humans, mutations in the Wnt3 gene lead to loss of limbs and craniofacial defects including cleft lip and palate (Niemann et al. 2004). Lef1-/-;TCF4-/- mutants form bone, but exhibit truncations of multiple structures (Brugmann et al. 2007).

During craniofacial development, Wnt/β-catenin signaling is required for proliferation and cell fate selection of the skull bone primordia (Brault et al. 2001; Hill et al. 2005; Goodnough et al. 2012; Fan et al. 2016). Early loss of β-catenin in CNCC-CM at

E8.5 using Wnt1Cre leads to a complete loss of head (Brault et al. 2001). Loss of β- catenin in both the CNCC and the PM at E10.5 using Engrailed1Cre (En1Cre) leads to a complete loss of bone and a conversion to cartilage in both the CNCC-CM and PM-CM

(Hill et al. 2005; Goodnough et al. 2012). Unlike the long bones, the skull bones ossify through intramembranous ossification which lacks a cartilage intermediate stage. The cartilage conversion is a result of a loss of Osx expression and the ectopic expression of the chondrogenic gene, SRY-Box 9 (Sox9). Sox9 is a transcription factor and a key chondrogenic determinant required to establish the chondrocyte lineage (Bi et al. 1999).

These results indicate that β-catenin is required to initiate head formation, and expression of β-catenin must be maintained during intramembranous bone differentiation to ensure the bone cell fate.

Previous studies have established the requirement of Wnt/β-catenin signaling in both the CNCC-CM and PM-CM. However, in vivo analysis of the spatial regulation of 18

Wnt/β-catenin has also revealed complex signaling dynamics between multiple tissues.

Compared to the PM-CM, the CNCC-CM has higher expression of multiple Wnt ligands and receptors (Fan et al. 2016). However, the primary source of Wnt ligands in both the

CNCC-CM and the PM-CM is the surrounding ectoderm. Inactivation of Wls, which is required cell-autonomously for the secretion of Wnt ligands, in the ectoderm using

Crect (active in the head around E8.5) results in a complete loss of both the CNCC- and

PM-derived skull bone as a result of diminished RUNX2 expression (Carpenter et al.

2010; Goodnough et al. 2014). These results identify Wnt/β-catenin signaling as a critical factor throughout the entire process of skull bone development including RUNX2 dependent cell fate selection and OSX dependent bone commitment (Fig. 1.6). They demonstrate that multiple tissues must coordinate to promote Wnt/β-catenin signaling to ensure cell fate selection of the bone precursors and differentiation of the bone progenitors.

1.4d: Twist Family BHLH Transcription Factor 1 (Twist1) is regulated by Wnt/β-catenin signaling and required for skull bone formation

One of the factors by which Wnt/β-catenin signaling may regulate skull bone formation is through Twist1. Twist1 is a basic helix loop helix transcription factor that is positively regulated by β-catenin (Goodnough et al. 2014). In both the CNCC-CM and

PM-CM, Twist1 is required for the cell fate selection and differentiation of the bone progenitors spatially and temporally. Inactivation of Twist 1 at E8.5 (Wnt1Cre) in just the

CNCC-CM or E9.5 in both the CMCC-CM and PM-CM (Twist2Cre) leads to a reduction in

Runx2 expression and a complete loss of facial structures (Bildsoe et al. 2009; 19

Goodnough et al. 2016). In contrast, loss of Twist1 from E10.5 using En1Cre leads to a reduction of the frontal and parietal bone and ectopic cartilage formation in the posterior region of the skull (Tran et al. 2010; Goodnough et al. 2012, 2016; Fan et al.

2016). These data demonstrate that, similar to β-catenin, Twist1 is required for both bone cell fate selection and commitment. Twist1 mutants also demonstrate a cross-talk between the CNCC-CM and the PM-CM. Loss of Twist1 at E8.5 in the CNCC-CM using

Wnt1Cre leads to loss of CNCC-derived bones, and almost complete loss of PM-derived bones (Bildsoe et al. 2009). Similarly, loss of Twist1 at E8.5 in the PM-CM using

Mesp1Cre leads a loss of the PM-derived bones as well as the frontal bone (Bildsoe et al.

2013). All together, these data indicate that Twist1 is a potential mediator by which

Wnt/β-catenin affects the cell fate selection of the bone precursors and differentiation of the bone progenitors.

1.4e: Retinoic Acid signaling affects many different signaling pathways required for skull bone formation

Retinoic acid is derived from Vitamin A (retinol) and plays multiple roles during embryonic development (Mark et al. 2009; Rhinn and Dolle 2012). Directly and indirectly, RA signaling has been associated with the regulation of over 500 genes, including members of the BMP, FGF, and Wnt/β-catenin pathways and affecting a wide range of biological processes including differentiation and proliferation (Balmer and

Blomhoff 2002; Das et al. 2014). The diversity of RA signaling targets originates primarily from the dimerization of RA with two families of receptors (Chambont 1996; Balmer and

Blomhoff 2002). Typical RA signaling involves the binding of all-trans RA (atRA) to a 20 retinoic acid receptor (RAR) which then binds to a specific DNA element called a retinoic acid response element (RARE). Retinoid x receptors (RXR) are another type of retinoid receptor that is activated by 9-cis retinoic acid (9cRA). Both RARs and RXRs can be classified into three specific types; alpha, beta, and gamma. The specific expression of each individual receptors varies from tissue to tissue (Lohnes et al. 1994; Mark et al.

2009; Rhinn and Dolle 2012). However, a level of functional redundancy during development exists between the receptor types as embryos with null mutations in individual receptors are born healthy.

During craniofacial development in mice, tight regulation of RA signaling is required as too much or too little RA signaling leads to craniofacial defects. Null mutations in Rarα and Rarγ lead to a severe reduction of the nasal bones and embryonic lethality near E18.5 (Lohnes et al. 1994). In addition, some embryos exhibit a severe reduction in both the frontal and parietal bone, with the parietal bone becoming almost completely lost. Null mutations in Cyp26b1, which is required to break down RA, leads to a disruption in the formation of the bones of the skull including reductions in the frontal and parietal bones (Maclean et al. 2009). It is currently unclear as to the exact mechanism by which RA regulates skull bone development as RA signaling is a potential mediator of multiple singling pathways. Considering that RA signaling affects many different targets directly and indirectly, multiple pathways could be working synergistically.

1.4f: Epigenetic regulation of skull bone development 21

The formation of the skull bone requires strict spatial and temporal control of multiple transcription factors and signaling pathways within a short window of development. Epigenetic modifications are a means by which transcription can be finely tuned during development to allow for modular control of transcription in a reversible and heritable manner (Rothbart and Strahl 2014; Schübeler 2015). Such transcriptional regulation primarily occurs through DNA methylation or histone modifications.

DNA methylation plays an important role during early embryonic development to guide cell fate selection (Messerschmidt et al. 2014). The methylation primarily occurs on cytosine residues in cytosine-guanine dense regions (CpG islands), and is linked to transcriptional repression. However, DNA methylation is thought to be established early in development, and more transient transcriptional regulation throughout development is achieved through histone modifications (Tanay et al. 2007;

Lynch et al. 2012; Schübeler 2015).

Histone modifications are post-translational modifications that affect chromatin structure. Multiple different histone modifications have been identified such as methylation, acetylation, ubiquitination, and sumoylation, with the most understood modifications being methylation and acetylation. Gene regulation by histone modifications has been linked to multiple functions such as transcriptional activation, repression, or poising. (Barski et al. 2007; Rothbart and Strahl 2014; Minoux et al. 2017).

One major histone modification critical in development is the tri-methylation of the 27th lysine on the third histone (H3K27me3). H3K27me3 is a repressive modification 22 catalyzed by the Polycomb Repressive Complex 2 (PRC2). The Polycomb group (Pc-G) genes were originally discovered in Drosophila melanogaster as repressors of the body segmentation gene Bithorax. Male flies with a mutation in Pc-Gs had extra sex comb teeth on their second and third leg indicating a transformation into the first legs (Paro

1990). In mammals, PRC2 is a multi-protein Pc-G complex primarily composed of

Enhancer of Zeste 2 (Ezh2), Suppressor of Zeste 12 (Suz12), and Embryonic Ectoderm

Development (Eed). Ezh2 is the catalytic component of PRC2, and contains the methyltransferase activity required for the H3K27me3 modification (Fig. 1.8)

(Margueron and Reinberg 2011). Ezh2 null mutations in mice result in embryonic lethality demonstrating the critical role of PRC2 during embryonic development

(O’Carroll et al. 2001).

Figure 1.8: Enrichment of H3K27me3 by PRC2 leads to transcriptional repression. PRC2 binds to the third histone and tri-methylates lysine 27. Can lead to transcriptional repression or poising.

A major role of PRC2 during development is thought to be the regulation of cell fate selection (Boyer et al. 2006; Bracken et al. 2006a; Lee et al. 2006; Asp et al. 2011;

Bardot et al. 2013). In C2C12 mouse myoblast cells, knockdown of Suz12 promotes myogenic differentiation (Asp et al. 2011). In human embryonic stem cells (ESCs), genes 23 known to regulate cell fate selection and differentiation are enriched for H3K27me3

(Bracken et al. 2006a). In addition, both EZH2 and SUZ12 are bound to genes known for cell fate selection indicating transcriptional repression by PRC2. These data associate

PRC2 with cell fate selection, however, many of the studies connecting PRC2 to cell fate selection are in vitro or only correlative. It is becoming evident that PRC2 biology in mammals is complex requiring more rigorous studies.

In mammals, a homologue of EZH2, EZH1, has also been shown to compose the

PRC2 complex and possess methyltransferase activity (Shen et al. 2008). In mouse embryonic stem cells (ESCs), H3K27me3 is still detectable by chromatin immunoprecipitation followed by quantitative polymerase chain reaction (ChIP-qPCR) and western blot in Ezh2 null mutants. In contrast, no H3K27me3 is detectable following

Eed deletion, demonstrating the methyltransferase capabilities of EZH1. In Ezh2 null cells, EZH1 can co-localize to regions of H3K27me3. Furthermore, inhibition of Ezh1 with siRNA in Ezh2 null cells leads to a loss of H3K27me3 on specific genes. These results indicate functional redundancy between Ezh2 and Ezh1. However, in vivo, EZH1 null mice exhibit no phenotype and are viable, indicating that EZH2 is the primary methyltransferase during embryonic development (Ezhkova et al. 2011).

The H3K27me3 histone modification is typically thought to be a repressive mark.

In multiple cell types, however, genes enriched for H3K27me3 have been shown to be actively transcribed (Bracken et al. 2006a; Asp et al. 2011). It has also been demonstratd that, in mouse ESCs, PRC2 does not actively silence transcription but is recruited to previously silenced CpG islands to maintain transcriptional repression (Riising et al. 24

2014). These results demonstrate a more nuanced function of the H3K27me3 histone modification beyond explicit gene silencing. However, these results must be considered with a level of scrutiny. In vitro, epigenetic modifiers can vary with culture conditions

(McEwen et al. 2013). Therefore, in vivo studies on PRC2 are required to further elucidate its role in embryonic development.

1.4g: The in vivo role of PRC2 in craniofacial development

In vivo analysis of PRC2 function has revealed multiple functions, beyond cell fate selection, in various tissues in both a spatial and a temporal manner. Because Ezh2 is the methyltransferase component of PRC2, most in vivo studies investigating PRC2 have focused on Ezh2 mutants. In skeletal muscle cells, Ezh2 is required to maintain cell fate and promote regeneration (Juan et al. 2011). In the developing embryonic skin, Ezh2 and Ezh1 inhibit merkel cell differentiation (Bardot et al. 2013). Alternatively, the hair follicle lineage exhibits proliferation defects while the epidermal lineage exhibits hyperproliferation in Ezh2 and Ezh1 knockout mutants (Ezhkova et al. 2011). It is worth noting that both Ezh1 and Ezh2 need to be lost to affect the hair follicles and the epidermis indicating a tissue specific compensatory role for Ezh1. In long bones, loss of

Ezh2 and Ezh1 leads to decreased chondrocyte proliferation and hypertrophy (Lui et al.

2016).

During craniofacial development, PRC2 promotes the spatial identity of multiple structures. In early post-migratory CNCC cells, EZH2 is not required for transcriptional silencing, but for transcriptionally poising specific genes required to establish positional 25 identity within the head and face (Minoux et al. 2017). Transcriptional poising involves the pausing transcription rather than complete repression. Isolation of frontonasal, maxillary, mandibular (1st branchial arch), and 2nd branchial arch process CNCC-CM by

Wnt1Cre-driven florescence sorting revealed differential expression of multiple

"positional genes" linked to the identity of each structure. ATAQ-sequencing of each structure revealed these positional genes had open, accessible promoters but were not expressed. ChIP-sequencing further revealed these promoters were bound by both

H3K27me3 and H3K4me2 histone marks indicating transcriptional poising. Furthermore,

Ezh2 knockout mutants using Wnt1Cre displayed very little changes in gene expression, with no changes in expression of 92.6% of the genes. However, of the 7.4% upregulated genes, 57% were "positional genes" which had both H3K27me3 and H3K4me2 bound.

Genes that were only bound by H3K27me3 had no transcriptional changes indicating that, in vivo, PRC2 may be primarily required for short term repression/poising rather than long-term repression.

The effect of PRC2 on skull bone formation is also both spatially and temporally specific. Loss of Ezh2 in the CNCC-CM using Wnt1Cre leads to almost a complete loss of the CNCC derived skull bones. In contrast, loss of Ezh2 in PM-CM using Prx1Cre leads to no loss of skull bone and a fusion of multiple sutures (Schwarz et al. 2014; Dudakovic et al. 2015). These results could be due to the spatial and/or temporal differences between the two Cre drivers. Wnt1Cre is expressed in the CNCC-CM around E8.5 and Prx1Cre is expressed in the PM-CM around E9.5. Further evidence for a temporal requirement for

PRC2, however, comes from mutants lacking Ezh2 in both the CNCC-and PM-derived 26 cranial mesenchyme. Twist2Cre (Dermo1Cre) is expressed in both tissues around E9.5 and Ezh2 mutants have no craniofacial phenotype with only a change in the smooth muscle of the lung (Snitow et al. 2016). The spatial and temporal requirement for Ezh2 mirrors the spatial and temporal requirements seen in other signaling pathways during skull bone formation. Future studies examining the connection between the spatial and temporal organization of PRC2 and the known signaling pathways critical for skull bone formation will provide insights into craniofacial defects and tissue engineering.

1.5: Functional interaction and regulation of bone transcriptional program by BMP,

FGF, Wnt/β-catenin, RA, and PRC2:

During skull bone formation, the bone initiation program requires both spatial and temporal regulation. To what capacity the major signaling pathways and epigenetics functionally interact to regulate the bone initiation program is of great interest. With the use of conditional, in vivo, genetic mutants, the current work is beginning to establish the mechanisms behind the spatial and temporal regulation of bone cell fate selection, commitment, and differentiation (Fig. 1.4). The first step of the bone initiation programs is the expression of Msx1 and Msx2 in post migratory CNCC. Both FGF- and

BMP-signaling have been shown to affect the expression of the Msx genes. BMP- signaling is required for the induction of the pre-migratory CNCC prior to E8.5 and directly regulates Msx1 and Msx2 expression (Suzuki et al. 1997; Knecht and Bronner-

Fraser 2002; Tribulo et al. 2003). Loss of Bmp2, 4, and 7 at E8.5 in the post-migratory 27

CNCC using Wnt1Cre resulted in a decrease in Msx gene expression demonstrating that

BMP-signaling is required for establishing the skull bone precursors (Bonilla-Claudio et al. 2012). Additionally, FGF has also been shown to affect Msx gene expression, however, this regulation is often associated with BMP signaling interactions (Kettunen and Thesleff 1998; Yang et al. 2010).

In the bone initiation program, Runx2 and Osx are expressed after Msx genes and represent cell fate selection and commitment. Loss of Wnt signaling in the ectoderm or Twist1 in the CNCC-CM and PM-CM results in a loss of Runx2 and Osx expression with little effect on the expression of the Msx genes (Bildsoe et al. 2009;

Tran et al. 2010; Goodnough et al. 2012, 2014, 2016). Based on these in vivo data, it appears that BMP signaling is required to establish the bone precursors and Wnt/β- catenin signaling is required for the cell fate selection and commitment to bone. In addition, both RA signaling and PRC2 activity have been shown to regulate many factors, including Wnt, BMP, and FGF factors, involved in these processes adding an extra layer of regulation (Drissi et al. 2003; Iwamoto et al. 2003; Bracken et al. 2006a; Shimono et al. 2011; Schwarz et al. 2014; Dudakovic et al. 2015).

Many of the transcription factors involved in the bone initiation program perform different functions based on the developmental stage of bone highlighting the temporal dynamics governing skull bone formation (Fan et al. 2016). For example, after bone commitment, the Msx genes are required to suppress ectopic bone formation near the sutures (Roybal et al. 2010). Loss of β-catenin in pre-migratory neural crest results in a near complete loss of facial structures, while loss of β-catenin in post-migratory neural 28 crest leads to disruptions in the bone fate commitment (Brault et al. 2001; Goodnough et al. 2012). In addition to the temporal requirements, many of the factors involved in skull bone formation maintain spatial requirements as well. However, current in vivo data has primarily focused in the CNCC-CM. With the predominant use of Wnt1Cre to study the bone initiation program, relatively little is known about the roles of Wnt, BMP,

FGF, RA, and PRC2 in the PM-CM. Future studies utilizing multiple Cre drivers to systematically examine the roles of the signaling pathways between the CNCC-CM and the PM-CM will provide a valuable piece of the puzzle to help uncover the spatial and temporal regulation of the skull bones. A more functional understanding of these processes may lead to advancements in tissue engineering along with detection, prevention, and treatment of birth defects.

1.6: Clinical significance:

Approximately one third of babies born with birth defects display craniofacial anomalies (CA) (Trainor 2007). CAs of the skull bone can range from an increase or reduction in overall bone volume to missing or incomplete structures. Defects such as cleft lip and palate, resulting from incomplete palate fusion, craniosynostosis, resulting from fusion of the skull bone, orosteoporosis-pseudoglioma syndrome, which has a decrease in bone mass, can result from both environmental or genetic disturbances

(CDC.gov). A greater understanding of the genetics coordinating craniofacial development will provide insights into the pathogenesis of CAs and allow for early 29 detection of such defects and identification of environmental factors. In addition, CAs are often associated with other congenital defects such as heart defects, polydactyly, and hydrocephaly, further underscoring the importance of understanding the genetic mechanisms behind CAs (States 2001).

In humans, mutations in many these factors involved in the bone initiation program lead to CAs. Mutations in Runx2, for example, leads cleidocranial dysplasia, which is characterized by absent clavicles, open fontanelles, and tooth defects (Zelzer and Olsen 2003). Weaver's syndrome, which is characterized by skeletal defects including craniofacial defects, is cause by mutations in the Ezh2 gene (Tatton-Brown et al. 2011; Gibson et al. 2012). Multiple diseases have also been associated with defects in

Wnt/β-catenin signaling. For example, Saethre-Chotzen syndrome is the second most common form of craniosynostosis and is defined by mutations in Twist genes (Behr et al.

2011). Loss of function mutants in LRP5 lead to OPPG syndrome, and mutations in LRP6 are associated with osteoporosis. Robinow syndrome, which exhibits multiple craniofacial defects, is associated with Wnt target gene ROR2. Mutations in Wnt receptor Frizzled Class Receptor 9 (FZD9) leads to Williams-Beuren Syndrome which also exhibit multiple craniofacial defects (Clevers and Nusse 2012).

Considering the complexity of the bone initiation program, and the birth defects associated with its disruption, my graduate work has focused on understanding the in vivo interactions between multiple regulatory mechanisms. By taking a basic science approach to understanding how the skull bone develops, we hope to apply our results to many aspects of biology and gain a greater understanding of CAs. 30

Chapter 2: PRC2 is Dispensable in Vivo for β-Catenin-Mediated Repression of

Chondrogenesis in Mouse Embryonic Cranial Mesenchyme

I would like Mahima Devarajan, Gregg DiNuoscio, Alina Saikhova, Chia-Feng Liu,

Veronique Lefebvre, Peter Scacheri, and Radhika Atit for the valuable work and feedback on the work presented in chapter 2.

Chapter 2 was published in the journal of G3: Genes, Genomes, Genetics on December

2017.

Abstract:

A hallmark of craniofacial development is the differentiation of multiple cell lineages in close proximity to one another. The mouse skull bones and overlying dermis are derived from the cranial mesenchyme (CM). Specification of the embryonic cranial bone and dermis cell fates in the CM requires Wnt/β-catenin signaling. Loss of β-catenin leads to a switch to a chondrogenic cell fate. The mechanism by which Wnt/β-catenin activity suppresses the cartilage fate is unclear. Upon conditional deletion of β-catenin in the CM, several key determinants of the cartilage differentiation program, including Sox9, become differentially expressed. Many of these differentially expressed genes are known targets of the Polycomb Repressive Complex 2 (PRC2), suggesting that PRC2 may be required for

Wnt/β-catenin-mediated repression of chondrogenesis in the embryonic CM. Consistent with this possibility, we find that β-catenin can physically interact with PRC2 components in the CM in vivo. However, upon genetic deletion of Enhancer of Zeste homolog 2 (EZH2), 31 the catalytic subunit of PRC2, chondrogenesis remains repressed and the bone and dermis cell fate is preserved in the CM. Furthermore, loss of β-catenin does not alter either the

H3K27me3 enrichment levels genome-wide or on cartilage differentiation determinants, including Sox9. Our results indicate that EZH2 is not required to repress chondrogenesis in the CM downstream of Wnt/β-catenin signaling.

2.1: Introduction:

Embryonic craniofacial development involves the formation of complex structures from two progenitor stem cell pools; the cranial neural crest (CNC) and cephalic paraxial mesoderm (PM) (Jiang et al. 2002; Yoshida et al. 2008). The CNC gives rise to the more anterior tissues of the head and the PM gives rise to the posterior tissues. Both the CNC and PM contribute to the mesenchymal stem cells surrounding the brain called the cranial mesenchyme (CM). Both the cranial bones and the overlying dermis are derived from the CM(Yoshida et al. 2008; Tran et al. 2010). For proper patterning and development of these tissues, specific signaling pathways are required to regulate cell fate selection and tissue morphogenesis (Fan et al. 2016). Disruption of these pathways can lead to craniofacial malformations such as craniosynostosis and focal dermal hypoplasia (Wilkie 1997; Wang et al. 2007; Hsu et al. 2010). Thus, understanding the genetic mechanisms directing skull and dermal cell fate selection is critical to elucidating the etiology of such malformations.

The Wnt/β-catenin pathway is instructive for cranial bone and dermal fibroblast cell fate selection in the developing embryo (Atit et al. 2006; Ohtola et al. 2008; Tran et 32 al. 2010; Goodnough et al. 2012; Fan et al. 2016). Conditional loss of ectodermal Wnt- ligand secretion or mesenchymal β-catenin leads to a loss of cranial bones and dermis.

Instead, an ectopic formation of cartilage, marked by the upregulation of a key chondrocyte marker geneSox9, is observed (Tran et al. 2010; Goodnough et al. 2012,

2014; Budnick et al. 2016). In intramembranous bone, loss-of-function mutations in other known signaling pathways important in craniofacial development, such as those driven by Fibroblast Growth Factors (FGF) and Bone Morphogenetic Proteins (BMP), do not result in ectopic chondrogenesis (O’Rourke and Tam 2002; Fan et al. 2016). In craniofacial development, Wnt/β-catenin signaling seems to have a unique role in the repression of chondrogenesis in the CM.

β-catenin is a central transducer of the canonical Wnt signaling pathway, where it acts as a transcriptional co-activator of context-specific target genes to regulate cell fate selection in many cell types during development (Bhanot et al. 1996; Korinek et al.

1998; Liu et al. 1999; Haegele et al. 2003; Verani et al. 2007). While β-catenin is typically known as a transcriptional activator, a stabilized or post-translationally methylated form of β-catenin has been shown to function as a transcriptional repressor in vitro (Delmas et al. 2007; Hoffmeyer et al. 2017). However, the mechanism by which Wnt/β-catenin signaling in the CM prevents chondrogenesis, while ensuring proper cranial bone and dermal fibroblast cell fate selection in vivo, is unknown.

Recent in vitro studies have highlighted epigenetic histone modifications, by the

Polycomb Repressive Complex 2 (PRC2) specifically, as a possible mechanism by which

Wnt/β-catenin signaling represses chondrogenesis. PRC2 is a multi-protein complex 33 required for the repressive histone modification H3K27me3 (Jiang et al. 2002; Lund and

Van Lohuizen 2004; Peng et al. 2009). In multiple cell types and organisms, numerous connections between the Wnt/β-catenin pathway and PRC2 have been demonstrated.

First, like Wnt/β-catenin signaling, PRC2 is required for the regulation of cell fate selection (Lee et al. 2006; Sparmann and van Lohuizen 2006; Asp et al. 2011; Margueron and Reinberg 2011). Second, Sox9 and other chondrocyte differentiation determinants are known targets of PRC2 by H3K27me3 enrichment in multiple cell types ranging from mouse embryonic stem cells (mESC) to chick limb bud micromass cultures (Peng et al.

2009; Kim et al. 2013b; Kumar and Lassar 2014; Tien et al. 2015). Third, PRC2 regulates components of the Wnt/β-catenin pathway and vice-versa (Wang et al. 2010; Zemke et al. 2015; Mirzamohammadi et al. 2016; Yi et al. 2016). Fourth, β-catenin can physically interact with PRC2 components(Shi et al. 2007; Li et al. 2009; Jung et al. 2013;

Hoffmeyer et al. 2017). Fifth, β-catenin and PRC2 can cooperate with one another to enhance either Wnt signaling or PRC2 activity(Shi et al. 2007; Jung et al. 2013; Kumar and Lassar 2014; Hoffmeyer et al. 2017). It is important to note that these studies were all performed in cell culture models with one or more overexpressed proteins. Follow- up studies in vivo are therefore required. Understanding how Wnt/β-catenin signaling intersects with PRC2 to direct cell fate selection in vivo will provide new insights into the genetic mechanisms governing cranial bone and dermal development.

Here, we test the hypothesis that repression of chondrogenesis in the CM by

Wnt/β-catenin signaling requires PRC2-mediated epigenetic repression. In a conditional

β-catenin loss-of-function model, among the genes dysregulated in both mutant CM and 34 mutant dorsal mesenchyme, we found an overrepresentation of known targets of the

PRC2 pathway. Conditional deletion of Ezh2 in the CM does not phenocopy the ectopic cartilage in the β-catenin mutants, nor do H3K27me3 levels change upon complete loss of β-catenin in the CM mesenchyme. Our results suggest that the repression of chondrogenesis in CM is not reliant on PRC2, indicating that repressive mechanisms besides PRC2 are likely involved. We propose that either the “off” state of chondrogenic genes is not actively maintained by PRC2, and β-catenin represses chondrogenesis by regulating an unidentified inhibitory pathway.

2.2: Results:

2.2a: Genes dysregulated upon loss of β-catenin are enriched for the PRC2-associated

H3K27me3 histone mark

In an effort to determine if there is a functional link between β-catenin and PRC2 in vivo, we conditionally deleted β-catenin in the CM using Engrailed1Cre (En1Cre), and manually dissected the CM, along with the ectoderm (CM+ectoderm), and collected all the CNC- and PM-derived mesenchyme surrounding the brain (Fig. 2.1A) (Kimmel et al.

2000; Tran et al. 2010). 35

Figure 2.1: β-catenin regulates known PRC2 targets and can physically interact with PRC2

36

Figure 2.1 cont. (A) Schematic demonstrating the isolation of CM+ectoderm by manual dissection.

The isolated tissue is comprised of the supraorbital cranial mesenchyme and overlying ectoderm. (B)

Gene ontology analysis of differentially expressed genes in E13.5 En1Cre/+;R26R/+;β-cateninfl/+ control (n=5) and En1Cre/+;R26R/+;β-cateninfl/∆ mutant CM+ectoderm (n=4). Overlapping gene sets from MsigDB Perturbations ontology ranked by binomial P-value, which accounts for varying sizes of gene regulatory domains, from GREAT. Green bars represent gene sets regulated by PRC2. The ranked list is as follows: 1) Genes with high-CpG-density promoters bearing H3K4me2 and

H3K27me3 in the brain (MsigDBM1941). 2) Genes with H3K27me3 on their promoters in human embryonic stem cells identified by ChIP on chip (MsigDBM10371). 3) Genes identified as targets of

SUZ12 in human embryonic stem cells by ChIP on chip (MsigDBM2291). 4) Genes identified as targets of EED in human embryonic stem cells by ChIP on chip (MsigDB M2736). 5) Genes coordinately up-regulated in a compendium of adult tissue stem cells (M1999). 6) Genes up- regulated in uterus upon knockout of BMP2 in the uterus (MsigDB M2324). 7) Genes down- regulated in mouse embryonic stem cells upon deletion of SUZ12 (MsigDB M2291). 8) Genes up- regulated in human lung fibroblasts (IMR90) after knockdown of RB1 by RNAi (MsigDB M2129). 9)

Genes down-regulated in breast cancer (MsigDBM13072). 10) Genes down-regulated in immortalized non-. transformed mammary epithelium (HMLE) after knockdown by RNAi or expression of a dominant negative form of CDH1 (MsigDBM11790). (C) Integrated Genome Viewer

(IGV) representation of H3K27me3 ChIP-sequencing on E13.5 En1Cre/+;R26R/+;β-cateninfl/+ control

CM+ectoderm of Sox9(chr11:112,641,524-112,651,071), Col2a1 (chr15:97,804,033-97,837,155),

Col9a2 (chr4:120,710,171-120,729,930), Col11a2 (chr17:34,174,382-34,205,187). (D) RT-qPCR of enriched CM after bead purification with an antibody specific to PDGFRα. A reduction in K14 expression levels demonstrates purification of the mesenchyme. (E) Co-immunoprecipitation of β- catenin with components of PRC2. CM+ectoderm was isolated by manual dissection and the mesenchyme was purified using a PDGFRɑ antibody. IgG was used as a negative immunoprecipitation control.

37

En1Cre is expressed in both the CNC- and PM-derived CM. In order to analyze in vivo tissues with minimal manipulation, the ectoderm was isolated with the CM. We then profiled the whole transcriptome on three litter-matched E13.5

Figure S2.1: Genomic Regions Enrichment of Annotations Tool (GREAT) analysis of differentially

expressed genes upon deletion of β-catenin. (A) Identification of expected gene expression changes

between En1Cre/+;R26R/+;β-cateninfl/∆ mutant and En1Cre/+;R26R/+;β-cateninfl/+ control RNA-seq.

(B) The top biological pathways represented in the differentially expressed genes in

En1Cre/+;R26R/+;β-cateninfl/∆ mutant and En1Cre/+;R26R/+;β-cateninfl/+ control tissue. (B) The top

biological pathways from the Panther Pathway gene ontology. (C) The top biological pathways from

the MsigDB ontology. All ontologies ranked by binomial P-value.

38

En1Cre/+;R26R/+;β-cateninfl/+controls and four En1Cre/+;R26R/+;β-cateninfl/Δ mutants using the RNA-seq approach (GSE96872). The analysis of the data revealed 521 genes which were differential expression levels of at least 1.4 fold in the two experimental groups (p<0.05). Of the 521 differentially expressed genes, 322 were down-regulated and 199 were up-regulated in the mutants relative to the controls. Validating the approach, changes in expression of known Wnt/β-catenin targets were observed despite the presence ectodermal cells, where canonical Wnt signaling is known to be active (Fig. S2.1A) (Budnick et al. 2016).

To gain some initial insights into the biological role of the 521 differentially expressed genes, we performed a gene ontology analysis using the GREAT program which queries multiple ontology databases (McLean et al. 2010). As a comparison, we also analyzed RNA-seq data from E13.5 En1Cre/+;R26R/+;β-cateninfl/Δ dorsal dermal mesenchyme (GSE75944) (Budnick et al. 2016). The top five ontologies of the differentially expressed genes in both the mutant CM+ectoderm and mutant dorsal dermal fibroblasts included the Wnt signaling pathway, along with Cadherin signaling,

Integrin signaling, and ECM-receptor interactions (Fig. S2.1B,C) (Thomas et al. 2003).

Using the Molecular Signatures Database (MsigDB) Perturbations ontology, we also found that both data sets were highly enriched for genes regulated by PRC2 (Fig. 2.1B;

Fig. S2.2A)(Subramanian et al. 2005). 39

Figure S2.2: Genes regulated by β-catenin are enriched for PRC2 targets in dorsal dermal fibroblasts.

40

Figure S2.1 cont. (A) Differentially expressed genes in En1Cre/+;R26R/+;β-cateninfl/∆ mutant and

En1Cre/+;R26R/+;β-cateninfl/+ control dorsal dermal fibroblasts were run through GREAT. Overlapping

gene sets from the MsigDB Perturbations ontology were ranked by binomial P-value. Green bars

represent gene sets regulated PRC2. The ranked list is as follows: 1) Genes with CpG enriched

promoters bearing H3K4me2 and H3K27me3 in brain. 2) Genes up-regulated in ES (human embryonic

stem cells) with deficient SUZ12 [GeneID=23512]. 3) Genes up-regulated in TIG3 cells (human

fibroblasts) upon knockdown of EED [GeneID=8726]. 4) H3K27me3 target genes in hESC identified by

ChIP on chip. 5) Genes up-regulated in the normal-like subtype of breast cancer. 6) Genes up-

regulated in in vitro stromal cells from adipose tissue compared to in vivo cells. 7) Genes possessing

H3K27me3, SUZ12 [GeneID=23512], and EED [GeneID=8726] in hESC by ChIP on ChIP. 8) Genes

coordinately up-regulated in a compendium of adult tissue stem cells. 9) Genes up-regulated in

uterus upon knockout of BMP2 [GeneID=650]. 10) Genes consistently up-regulated in mammary cells

in both humans and mice. (B) All genes identified in the En1Cre/+;R26R/+;β-cateninfl/+ CM+ectoderm

were run through GREAT. The ranked list is as follows: 1) Genes down-regulated in NB4 cells (acute

promyelocytic leukemia, APL) in response to tretinoin [PubChem=444795]. 2) Genes with promoters

bound by E2F4 [GeneID=1874] in unstimulated hybridoma cells. 3) Genes up-regulated in ME-A cells

(breast cancer) undergoing apoptosis in response to doxorubicin [PubChem=31703]. 4) Genes whose

promoters are bound by MYC [GeneID=4609]. 5) Up-regulated genes in colon carcinoma tumors

compared to the matched normal mucosa samples. 6) Targets of c-Myc [GeneID=4609] and Max

[GeneID=4149] identified by ChIP on chip in a Burkitt's lymphoma cell line. 7) Genes up-regulated in

liver tumor compared to the normal adjacent tissue. 8) Genes up-regulated in lymphoblastoid cells

from the European population compared to those from the Asian population. 9) Genes up-regulated

in robust Cluster 2 (rC2) of hepatoblastoma samples compared to those in the robust Cluster 1 (rC1).

10) Housekeeping genes identified as expressed across 19 normal tissues.

Interestingly, enrichment for targets of PRC2 can be found in both the up- and down- 41 regulated genes. However, this enrichment is unique only to the genes differentially expressed in our β-catenin mutants. GREAT analysis on all genes expressed in the

CM+ectoderm (FPKM >1) did not result in enrichment for targets of PRC2 (Fig S2.2B).

Thus, the differential expression of PRC2 targets in the β-catenin mutant CM+ectoderm and dorsal mesenchyme reveals a potential functional link between the two pathways.

2.2b: Chondrocyte fate genes are enriched for H3K27me3 in the embryonic CM

To investigate a role for PRC2 in the repression of chondrogenesis in the CM in vivo, we queried for H3K27me3 enrichment in the loci of individual chondrocyte marker genes. We manually dissected the CM+ectoderm in E13.5 En1Cre/+;R26R/+;β- cateninfl/+controls and performed chromatin immunoprecipitation using an antibody against H3K27me3 followed by massive parallel DNA sequencing (ChIP-seq). This assay allowed unbiased comprehensive mapping of the genome-wide distribution of the

H3K27me3 modification (Active Motif Technology) (GSE96872). In the CM+ectoderm of

E13.5 controls, the transcriptional start sites of multiple cartilage markers, such as Sox9,

Col2a1, Col9a2 and Col11a2 (Fig. 2.1C) were enriched for H3K27me3 relative to the rest of the genome, indicating they are targets of PRC2 in the CM.

2.2c: Endogenous β-catenin and EZH2 may physically interact in the CM

Given the emerging reported connections made between the Wnt/β-catenin pathway and PRC2 in various systems in vitro (summarized Table 2.1), we tested whether β-catenin and PRC2 components physically interact at the native protein levels 42 present in the mouse CM extracts.

Table 2.1: Summary of publications demonstrating a biological interaction between β-catenin and

PRC2.

43

We manually dissected the cranial mesenchyme, made a cell suspension, and used a

PDGFRɑ antibody bound to magnetic beads to enrich for the CM population

(Goodnough et al. 2016). Relative to the unpurified sample, we found comparable levels of mRNA for mesenchyme progenitor markers, Pdgfrɑ, Twist2, and diminished ectoderm marker, Keratin 14 (K14) in the purified sample, confirming enrichment for

CM (Fig. 2.1D). We then prepared cell extracts from sorted CM and used them in a co- immunoprecipitation assay for β-catenin and EZH2. EZH2 is the methlytransferase component of PRC2, and is required for the H3K27me3 modification (Margueron and

Reinberg 2011). In line with our hypothesis, β-catenin co-immunoprecipitated with the

EZH2 antibody. In addition, we also observed reciprocal co-immunoprecipitation of

EZH2 and another major PRC2 component, SUZ12, by the β-catenin antibody (Fig. 2.1E).

These results indicate that PRC2 components and β-catenin physically interact when present at wild-type expression levels in the CM. Thus, these data provide a potential molecular link between Wnt/β-catenin signaling and PRC2 in the mouse embryo.

2.2d: β-catenin is not required for PRC2 component expression or bulk H3K27me3 levels

To determine if β-catenin is required for the formation of the PRC2 complex itself, we first examined the expression of the main PRC2 components: Ezh2, Suz12, and

Eed. Based on our RNA-seq data set (FPKM values), we found no significant changes in the PRC2 component mRNA levels (Fig. 2.2A). 44

Figure 2.2: β-catenin activity is not required for the expression of PRC2 components and bulk

H3K27me3 levels.

To validate this result, we manually dissected E13.5 En1Cre/+;R26R/+;β-cateninfl/+ control and En1Cre/+;R26R/+;β-cateninfl/Δ mutant CM+ectoderm (Fig. 2.1A) and 45 determined the mRNA levels of Ezh2, Suz12, and Eed by RT-qPCR.

Figure 2.2 cont. (A) FPKM values obtained from RNA-seq assays in CM+ectoderm from

En1Cre/+;R26R/+;β-cateninfl/∆ mutant and En1Cre/+;R26R/+;β-cateninfl/+ controls. (B) RT-qPCR analysis

of the CM+ectoderm in En1Cre/+;R26R/+;β-cateninfl/+ (n=7) and En1Cre/+;R26R/+;β-cateninfl/∆ mutant

(n=9). The data are represented as fold change in mutants over controls. The dotted line represents

the β-catenin controls. Axin2 and Sox9 are known targets regulated by β-catenin. (C,D) Western blot

analysis (n=5) of EZH2 and H3K27me3 in CM+ectoderm from En1Cre/+;R26R/+;β-cateninfl/∆mutants

and En1Cre/+;R26R/+;β-cateninfl/+ controls. Band intensities were quantified using ImageJ. (E,F)

Schematics representing the supraorbital CM in coronal sections near the frontal bone primordia. (G)

Indirect immunofluorescence of SOX9, H3K27me3, and DAPI in the supraorbital mesenchyme (n=2

controls; 3 mutants). Images were taken near the frontal bone primordia (plane I). Dashed lines

indicate the brain and ectoderm boundaries. (*) indicates region of ectopic cartilage. Scale bars:

200µm

Similar to the RNA-seq dataset, the relative mRNA levels of the individual PRC2 components were comparable in the control and mutant samples (Fig. 2.2B). In comparison, the expected changes in mRNA levels was observed in known β-catenin responsive genes, Axin2 and Sox9 (Fig. 2.2A,B) (Jho et al. 2002; Goodnough et al. 2012).

Evaluation of the total H3K27me3 and EZH2 protein levels using Western blot assays also revealed comparable protein levels between control and β-catenin mutant

CM+ectoderm (Fig. 2.2C,D). To obtain spatial information and account for levels in the ectoderm between our control and mutants, we performed indirect immunofluorescence for H3K27me3 in the E13.5 En1Cre/+;R26R/+;β-cateninfl/+ control and En1Cre/+;R26R/+;β-cateninfl/Δ mutants (Fig. 2.2E,F,G). While we consider the CM to 46 include the entire cranial mesenchyme surrounding the brain (Fig. 2.1A), we focused our indirect immunofluorescence analysis on the region directly above the eye (supraorbital

CM) (Fig. 2.2E) due to easily identified histological landmarks such as the eye and brain ventricles. Considering that knockout of β-catenin results in ectopic chondrogenesis throughout the CM, we expect the supraorbital CM to be representative of the entire

CM. In the supraorbital CM, both the control and β-catenin mutants are positive for

H3K27me3 demonstrating PRC2 is still active without β-catenin. Furthermore,

H3K27me3 can still be found in the expanded SOX9 domain in the β-catenin mutants compared to the controls. We concluded that β-catenin is not required cell autonomously in the CM to regulate the relative mRNA levels of core PRC2 subunits, the

EZH2 protein levels, and bulk H3K27me3 levels. However, these results leave open the possibility that it may be required to recruit PRC2 to specific loci in the genome.

2.2e: Loss of Ezh2 does not lead to ectopic cell type fate selection or chondrogenesis in the CM

We next determined if PRC2 is required for the repression of chondrogenesis in the CM in vivo. In order to remove PRC2 function in the cranial mesenchyme, we conditionally deleted Ezh2using a floxed allele (Shen et al. 2008). Surprisingly, conditional deletion of Ezh2 at E10.5 using En1Cre did not lead to the expected loss of

H3K27me3 at E13.5 in the CM as determined by indirect immunofluorescence (Fig. 47

Figure S2.3: Knockout of Ezh2 with En1Cre does not lead to a change in H3K27me3 enrichment in the CM. Indirect immunofluorescence in En1Cre;Ezh2fl/+ and En1Cre;Ezh2fl/fl supraorbital CM of the

PRC2 repressive mark H3K27me3. No qualitative difference in between (A) En1Cre;Ezh2fl/+ and (B)

En1Cre;Ezh2fl/fl is observed. Scale bars: 200µm.

48

S2.3). We then conditionally deleted Ezh2 in the CM using Dermo1Cre which is expressed in the CM by E10.0(Yu et al. 2003; Goodnough et al. 2012). Loss of Ezh2 was sufficient to lead to an upregulation of Cdkn2a, a known target of PRC2 (Fig 2.3A) (Shen et al. 2008; Lui et al. 2016). 49

Figure 2.3: Knockdown of Ezh2 in the cranial mesenchyme does not lead to changes in cell type

fate selection. (A) Known downstream target Cdkn2a expression relative to controls in

Dermo1Cre;Ezh2fl/fl CM+ectoderm (n=2). (B) Percent H3K27me3 positive cells in the supraorbital CM.

(C) Schematic representing the coronal sections near the frontal bone primordia. (D-G) Indirect

immunofluorescence in Dermo1Cre;Ezh2fl/fl supraorbital CM. (D) Qualitative loss of the PRC2

repressive mark H3K27me3. (E-G) Domains of similar size and location were observed for the

osteoblast (OSX), dermal fibroblast (LEF1), and chondrocyte (SOX9) markers between controls and

mutants. Dashed lines indicate the brain and ectoderm boundaries. (D-F) Arrows mark the tissue

domains. Scale bars: 200µm.

We also found depletion of H3K27me3 in the supraorbital CM (Fig. 2.3C) by indirect immunofluorescence between the Dermo1Cre; Ezh2fl/fl mutants to Dermo1Cre; Ezh2fl/+ 50 controls (Fig. 2.3B,D). H3K27me3 signal was maintained in both the ectoderm and the brain, where Dermo1Cre is not expressed. After confirming the reduced H3K27me3 levels, we then examined the protein levels of cell fate markers for bone, dermis, and cartilage progenitors by indirect immunofluorescence. Conditional deletion of Ezh2 in the supraorbital CM did not lead to change in the location and size of the dermal domain as indicated by LEF1, and the bone domain as indicated by Osterix (OSX) (Fig.

2.3E,F). Consistent with this, we did not observe ectopic expression beyond the cartilage base of the key cartilage differentiation determinant, SOX9 (Fig 2.2G). Based on these data, we conclude that Ezh2 has little effect on the patterning of the tissue domains and minimal effect on the levels their respective proteins by immunofluorescence.

To further test wither the PRC2-dependent H3K27me3 plays a role in the repression of chondrogenesis in the CM, we chemically inhibited EZH2 function in primary E13.5 CM+ectoderm cells in vitro (Fig. 2.4A,E). 51

Figure 2.4: Chemical inhibition of EZH2 methyltransferase does not lead to an up-regulation of early

chondrocyte markers in CM+ectoderm. (A,E) Schematic demonstrating the isolation of primary

CM+ectoderm fibroblasts. GSK126 is specific to EZH2, and UNC1999 is specific to both EZH2 and EZH1.

GSK126 and UNC1999 inhibit EZH2’s methyltransferase activity. (B, F) Western blots demonstrating

reduction/loss of H3K27me3 level following incubation with GSK126 ((IC50=75-100nm) or UNC1999

(IC50< 10 nM for EZH2 and 45 nM for EZH1). (C,D,G) qPCR analysis of the expression of Sox9 and Col2a1

following inhibition of EZH2. GSK126: n=5 mutants and 6 controls for Sox9 and n=3 mutants and

controls for Col2a1. UNC1999: n=7 mutants and 9 controls.

Incubation with small molecule methyltransferase inhibitors GSK126, which is specific for EZH2, or with UNC1999, which inhibits EZH2 and EZH1,led to a considerable reduction in bulk H3K27me3 protein levels (Fig. 2.4B,F). Upon treatment with GSK126 or

UNC1999, the Sox9 and Col2a1 mRNA levels were not significantly increased (Fig.

2.4C,D,G). Taken together, these data indicate that EZH2 and H3K27me3 are dispensable for regulating the mRNA level of chondrocyte differentiation markers in the

CM+ectoderm. 52

2.2f: Loss of β-catenin does not significantly alter H3K27me3 enrichment genome- wide

Next, we tested to what extent β-catenin is required for the recruitment of PRC2 to the genome in a site-specific manner. We performed ChIP-seq assays, as described in

Figure 2.1, to map the genome-wide distribution of H3K27me3 in the CM in vivo between En1Cre/+;R26R/+;β-cateninfl/+ controls and En1Cre/+;R26R/+;β-cateninfl/Δ mutants (GSE96872). Sequencing of the CM+ectoderm revealed 14,337 peaks in the control and 10,752 peaks in the mutant, thus 25% fewer peaks in the mutant.

Surprisingly, genome-wide comparisons between individual mutant and control

H3K27me3 peaks revealed modest differences in fold enrichment between the two 53

Figure 2.5: Loss of β-catenin does not significantly alter H3K27me3 enrichment genome wide or on

cartilage differentiation determinants.

samples (Fig. 2.5A). The differences in peak numbers between β-catenin controls and mutants were associated with changes in smaller H3K27me3 peaks (Fig. 2.5A’, A’’’, B’,

B’’’). Furthermore, any gains and losses of strength of H3K27me3 peaks was not 54

Figure S2.4: Analysis of strong, medium, and weak H3K27me3 enrichment in the CM+ectoderm.

H3K27me3 ChIP-sequencing reads were grouped based on fold enrichment (strong peaks = >20, medium peaks = 10-20, weak peaks = <10), and were mapped to the single nearest gene using

GREAT. (A) IGV representation of each class of peak. (B) Distribution of H3K27me3 strong, medium, and weak peaks in relation to the transcription start site of the single nearest gene to the peak. (C)

Number of peaks in each class between En1Cre/+;R26R/+;β-cateninfl/+ controls and

En1Cre/+;R26R/+;β-cateninfl/∆ mutants.

55

Figure 2.5 cont. (A) Treeview representation of H3K27me3 peak strength genome wide in the

CM+ectoderm between E13.5 En1Cre/+;R26R/+;β-cateninfl/+ controls and En1Cre/+;R26R/+;β-

cateninfl/∆mutants. H3K27me3 ChIP-sequencing signal strength was mapped 5kb -up and down-

stream from each peak. A change from high intensity to low intensity identifies a peak is lost. (B)

Intersection of changes in H3K27me3 enrichment from (A) with En1Cre/+;R26R/+;β-cateninfl/∆

mutant and En1Cre/+;R26R/+;β-cateninfl/+ control RNA-seq. B’, B’’, and B’’’ correspond with A’, A’’,

and A’’’ respectively. (C,D) Intersection of En1Cre/+;R26R/+;β-cateninfl/+ control and

En1Cre/+;R26R/+;β-cateninfl/∆ mutant ChIP-sequencing and RNA-sequencing. H3K27me3 ChIP-

sequencing signal strength was measured across all genes bound by H3K27me3 or genes identified

to be differentially expressed in β-catenin mutant CM+ectoderm. The X-axis demarcates the

percent distance across a gene between the transcription start site and the transcription end site.

(E) IGV representation of H3K27me3 signal peaks between En1Cre/+;R26R/+;β-cateninfl/+ control

(n=1) and En1Cre/+;R26R/+;β-cateninfl/∆ mutant (n=1) CM+ectoderm. Cdkn2a is a known target of

PRC2. Sox9 and Col2a1 are chondrocyte marker genes.

associated with gene expression changes (Fig. 2.5B). In addition, on all genes containing

H3K27me3, the signal intensity of the peaks was also comparable in the mutant and control across the gene body (Fig. 2.5C). Next, we examined changes in H3K27me3 peak signal across the gene body of the differentially expressed genes identified in our RNA- sequencing data. In both the up- and down-regulated genes, the H3K27me3 enrichment was comparable between β-catenin controls and mutants (Fig. 2.5D).

From our ChIP-seq dataset, we observed variations of the H3K27me3 enrichment throughout the genome ranging from large peaks blanketing an entire gene body to smaller peaks located just on the promoter. To further investigate the connection between H3K27me3 peak enrichment strength and gene expression, we divided the 56

H3K27me3 peaks into three categories based on the level of enrichment: strong (>20- fold enrichment), medium (10-20-fold enrichment), and weak (<5-fold enrichment) (Fig.

S4). Representative enrichment for strong, medium, and weak peaks can be found on the HoxA cluster, Sept9, and Lmtk3 respectively (Fig. S2.4A). The genomic location of each class of peak using GREAT revealed that the large majority of strong and medium peaks were within 5kb of the TSS, and the weak peaks had a more even distribution spanning out 500kb from the TSS (Fig. S2.4B). Between controls and β-catenin mutants, the number of strong and medium peaks was comparable with most of the variation found in the weak peaks (Fig. S2.4C). To further characterize each class of peak, we performed gene ontology analysis on the control H3K27me3 ChIP-seq dataset. Gene ontology analysis revealed distinct functions for the strong peaks such as DNA- binding/transcription regulation and conserved homeobox sites, and the medium and weak peaks shared functions such as Wnt signaling and ion transport (Fig. S2.5). 57

Figure S2.5: Gene ontology analysis of strong, medium, and weak H3K27me3 enrichment. Gene

ontology analysis on a list of genes pertaining to each class of H3K27me3 peaks generated by

GREAT. The gene set is limited to include only the genes with a H3K27me3 peak +/- 5kb from their

transcription start site. (A) Top five biological pathways identified in the Panther and MsigDB

ontology databases for genes containing each class of peaks. Analysis was performed using

GREAT. (B) DAVID ontology analysis of the gene list from each class of H3K27me3 peak.

Furthermore, comparisons between each class of peak found near a TSS (+/- 5kb) and the genes identified in our RNA-seq dataset revealed that 70% of strong peaks, 53% of 58 medium peaks, and 47% of weak peaks were associated with transcriptional repression

(<1FPKM) (Fig. S2.6A).

Figure S2.6: Strong, medium, and weak H3K27me3 peak groups have similar representation in

genes regulated by β-catenin. Intersection of En1Cre/+;R26R/+;β-cateninfl/+ control and

En1Cre/+;R26R/+;β-cateninfl/∆ mutant RNA-seq data with genes bound each class of H3K27me3 peak.

(A) Number of each class of peak found in all expressed genes (>1 FPKM) and repressed genes (<1

FPKM) identified in En1Cre/+;R26R/+;β-cateninfl/+ control CM+ectoderm. (B) Overlap of genes up- and

down-regulated in En1Cre/+;R26R/+;β-cateninfl/∆ mutants with all, strong, medium, and weak peaks.

These results indicate the level of H3K27me3 enrichment may be predictive of its transcriptional repressive function. When we intersected genes bound by each class of

H3K27me3 peak and differentially expressed genes, we found each class of peak had similar enrichment between the down- and up-regulated genes indicating H3K27me3 enrichment does not predict transcriptional repression by β-catenin (Fig. S2.6B).

59

2.2g: H3K27me3 enrichment is not depleted on ectopically expressed chondrocytic gene determinants in β-catenin mutants

Figure S2.7: H3K27me3 peaks are associated with genes actively transcribed and repressed in CM.

Integrated genome viewer representation of H3K27me3 peaks in En1Cre/+;R26R/+;β-cateninfl/+

control and En1Cre/+;R26R/+;β-cateninfl/∆ mutant CM+ectoderm. (A-B) The HoxA locus and T are

known targets of PRC2. (C-F) Lef1, Axin2, Twist1, and Twist2 are down-regulated upon the deletion of

β-catenin. (G) Runx2 is an osteoblast marker. (H,I) Col11a2 and Col9a2 are chondrocyte markers. (J)

Mcm6 is not bound by H3K27me3 in both the controls and mutants.

60

To determine if the loss of β-catenin resulted in depletion of H3K27me3 on chondrocyte differentiation determinants, we examined the enrichment of H3K27me3 on Sox9 and its downstream target Col2a1, which have higher mRNA levels in the β- catenin mutant CM (Fig. 2.2) (Goodnough et al. 2012). Cdkn2a, HoxA, T/Brachyury are known targets of PRC2 containing strong H3K27me3 peaks and serve as a controls.

Upon deletion of β-catenin, we did not observe a change in H3K27me3 enrichment on known PRC2 target genes (Fig. 2.5E, S2.7A,B). More importantly, H3K27me3 enrichment did not change on Sox9, Col2a1, Col9a2, Col11a2 (Fig. 2.5E, S2.7H,I). Furthermore,

H3K27me3 enrichment was similar between β-catenin controls and mutants on the TSS of critical bone and dermal marker genes, such as Runx2, Twist1, Twist2, Axin2, Lef1,

(Fig. S2.7C-G). Mcm6 serves as a negative control and lacks H3K27me3 enrichment in either the control or mutant (Fig. S2.7J). It is worth noting that Cdkn2a, HoxA, and T loci had strong H3K27me3 enrichment peaks, while Sox9, Col2a1 loci had medium enrichment peaks in both controls and β-catenin mutants (Fig. 2.5E, S2.7A,B). Our results showed that H3K27me3 enrichment is not depleted from the TSS of chondrocyte differentiation determinants in β-catenin mutants and remains enriched in actively transcribed genes.

2.3: Discussion:

Based on in vivo evidence that β-catenin is required to repress chondrogenesis in the CM, and on reported in vitro evidence connecting β-catenin and PRC2 in other processes, we tested the hypothesis that repression of chondrogenesis by Wnt/β- catenin signaling requires epigenetic repression by PRC2 in vivo. Consistent with the 61 findings from previous studies, our results demonstrate that an in vivo loss of β-catenin in the CM and dorsal mesenchyme leads to the activation of chondrogenic marker genes such as Sox9, Col2a1, and Col11a2 as well as other known PRC2 targets genes. Further, we find that β-catenin can physically interact with PRC2 components at native protein levels in CM-enriched protein extracts. In contrast to findings from in vitro studies, we observe that β-catenin is not required for expression of major PRC2 components in vivo, and that PRC2 is dispensable for the repression of chondrogenic marker genes in CM cells. Conditional deletion of β-catenin in the CM does not alter H3K27me3 enrichment around differentially expressed genes nor genome-wide in vivo. Our data in genetic mutants in vivo are consistent with a model whereby EZH2 and H3K27me3 are not required in the CM for guiding cell fate selection. Interrogating mixed cell populations is unlikely to account for our major finding, given that our CM restricted deletion of β- catenin did not lead to changes in H3K27me3 profiles, and Ezh2 mutants in vivo did not show changes in cell fate selection in the supraorbital CM.

Considering that loss of β-catenin at E10.5 leads to ectopic chondrogenesis but loss of Ezh2 at E10.5 did not phenocopy the β-catenin mutant, the function of the physical interaction between β-catenin and PRC2 remains unclear. A recent study in human colon cancer cells demonstrated that EZH2 alone, independent of H3K27me3, was sufficient to repress transcription (O’Geen et al. 2017). While we did not observe genome-wide changes in H3K27me3 enrichment upon loss of β-catenin, it is possible β- catenin is required to recruit EZH2 itself to the genome. Alternatively, EZH2 was recently shown to bind to β-catenin in mouse embryonic stem cells and trimethylate lysine 49 62

(K49me3) on the β-catenin protein itself (β-catMe3)(Hoffmeyer et al. 2017). The β- catMe3 protein could then function as a transcriptional repressor at defined loci in ES cells to govern neuronal versus mesoderm fate. However, loss of Ezh2 in the CM did not lead to alteration in cell fate selection indicating K49me3 modification of β-catenin does not play a role in cell fate selection in the CM. Future studies examining DNA-binding by

EZH2 and β-catenin could provide a biological function to the physical interaction between β-catenin and EZH2.

The lack of cell fate changes in the supraorbital CM of Ezh2 mutants could indicate that the role of EZH2, and by extension PRC2, is dependent on the developmental stage and cell type. Most mammalian studies linking PRC2 and cell fate selection were performed in embryonic stem (ES) cells in vitro. Differences in the role of

EZH2 between in vivo CM and in vitro ES cells may indicate the cell fate selection role of

PRC2 is unique to ES cells or linked to specific cell types. In addition, previous studies deleting Ezh2 at similar developmental stages in the mouse embryo found varying craniofacial phenotypes and defects(Schwarz et al. 2014; Dudakovic et al. 2015).

Deletion of Ezh2 in the pre-migratory cranial neural crest cells with Wnt1Cre by E8.5 resulted in severe reduction of facial and skull bones and embryonic lethality(Schwarz et al. 2014). Conditional deletion of Ezh2 at E9.5 in CM and facial mesenchyme with

Prx1Cre resulted predominantly in postnatal craniosynostosis. We did not find gross changes in embryonic craniofacial morphology upon deletion of Ezh2 in the CM at E10.0 with Dermo1Cre (data not shown). These results suggest that the role of PRC2 in embryonic development may be cell type- and developmental stage-specific. Future 63 studies in vivo are required to tease out the timing and dynamics of developmental gene regulation by PRC2.

There are recent data from several groups refining the role of PRC2 and

H3K27me3 enrichment. According to the histone code hypothesis, H3K27me3 is often found on transcriptionally repressed genes and is widely considered a sign of transcriptional repression (Boyer et al. 2006; Lee et al. 2006; Roh et al. 2006; Barski et al. 2007; Heintzman et al. 2007, 2009). In early post-migratory mouse neural crest cells,

H3K27me3 was shown to mark bivalent domains that also contain the activating mark

H3K4me3, indicating transcriptional poising rather than repression (Minoux et al. 2017).

Recently, the histone code model has been refined to show that H3K27me3 enrichment is not just predictive of transcriptional repression in mouse embryonic stem cells, but also indicative of a past-transcriptional repressed state (Riising et al. 2014; Comet et al.

2016). Furthermore, in human colon cancer cells, ectopic deposition of H3K27me3 with an EZH2-dCas9 fusion construct was not sufficient for transcriptional repression (O’Geen et al. 2017). In mouse rib chondrocytes, an intersection of RNA-seq data with H3K27me3

ChIP-seq data also suggested that H3K27me3 enrichment on TSS was not sufficient for transcriptional repression. When compared to genes dysregulated upon knockout of

EED, an essential PRC2 subunit, only 11% of the genes dysregulated were enriched for

H3K27me3. Thus, the biological role of the remaining 89% of H3K27me3 peaks is unclear

(Mirzamohammadi et al. 2016). Our data are entirely consistent with these recent findings. Intersecting our in vivo RNA-seq and ChIP-seq studies revealed legitimate

H3K27me3 peaks in genes that did not correlate with transcription levels. We found that 64 both expressed and repressed genes in control CM+ectoderm can be enriched for

H3K27me3 demonstrating H3K27me3 is not sufficient to indicate repression. The

H3K27me3 marks remain at Sox9 and other cartilage marker genes in β-catenin mutants, suggesting that these marks may be carried over and reflective of a past transcriptional “off” state.

If PRC2 is not required for repressing chondrogenesis in CM, the question remains as to what factors exert this function. We propose three other models that will require further testing. The first model calls for other epigenetic-related mechanisms such as direct covalent modifications of DNA (DNA methylation), or other histone modification-related mechanisms, such as G9a-associated K9me3 repression. A study in chick limb bud micromass cultures showed that addition of exogenous Wnt3a led to an increase in DNA methylation by DNMT3a on the Sox9 promoter (Kumar and Lassar

2014). In our hands, however, the addition of DNMT inhibitors did not alter Sox9 and

Col2a1 mRNA levels in primary CM+ectoderm cells cultured in vitro (data not shown).

Further studies in vivo will be required to investigate this model. The second model postulates that β-catenin activates yet-to-be identified signaling pathways or transcription factors that would be directly involved in repression. For example, Twist1 is positively regulated by Wnt/β-catenin signaling, and conditional deletion of

Twist1partially phenocopies the ectopic chondrogenesis found in the

En1Cre/+;R26R/+;β-cateninfl/Δ mutants(Komori et al. 1997; Goodnough et al. 2012). The retinoic acid (RA) signaling pathway can interact with Wnt/β-catenin signaling and it can promote chondrocyte development and function in vitro (Yasuhara et al. 2010; Uchibe 65 et al. 2017). RA signaling pathway components are robustly expressed in the control

CM. Their interaction with Wnt/β-catenin signaling and role in the CM remain to be tested. A third model is that β-catenin does not directly control the transcription of cartilage determinants and marker genes, but may control the expression or activity of factors involved in the post transcriptional modification of cell fate determination and chondrocyte differentiation genes.

Overall, our data suggested a model whereby the repression of the chondrogenic fate by Wnt/β-catenin signaling does not rely on EZH2 and H3K27me3, but implies other yet-to-be identified transcriptional or post-transcriptional mechanisms.

2.4: Materials and methods:

2.4a: Mice and genotyping

The following strains were used in this study: Engrailed1Cre (En1Cre) (Kimmel et al. 2000), Rosa26 Reporter (R26R) (Soriano 1999), β-catenin null (β-catenin∆) (Brault et al. 2001), conditional β-catenin floxed (β-cateninfl) (Haegel et al. 1995),

Twist2Cre(Dermo1Cre)(Yu et al. 2003), and conditional Ezh2floxed (Ezh2fl)(Shen et al.

2008). Mice were maintained in mixed genetic backgrounds. For timed matings,

En1Cre;β-catenin+/∆ males were crossed with R26R/R26R;β-cateninfl/fl females, and

Dermo1Cre;Ezh2fl/+ males were crossed with Ezh2fl/fl females. Vaginal plugs were checked every morning and assigned as embryonic (E) 0.5. For each experiment, a minimum of three mutants with litter-matched controls were studied unless otherwise noted. Animals of both sexes were randomly assigned to all the studies. Case Western 66

Reserve Institutional Animal Care and Use Committee approved all animal procedures in accordance with AVMA guidelines (Protocol 2013-0156, approved 21 November 2014,

Animal Welfare Assurance No. A3145-01).

2.4b: Cranial mesenchyme isolation

At E13.5, the CM was isolated by manual dissection. An incision was made around the circumference of the CM, and the tissue covering the brain was manually dissociated. The CM samples were a mixed cell population comprised of the CNC- and

PM-derived cranial mesenchyme, which is En1Cre positive, and also contained the overlying ectoderm, which is negative for En1Cre (CM+ectoderm). Each embryo yielded

~500,000 CM cells for the controls and 250,000-500,000 CM cells for the mutants.

Individual embryos were kept separate and considered single biological replicates. The wild-type samples isolated for co-immunoprecipitation were dissociated by incubating the tissue in 0.25% Trypsin-EDTA (Thermo Fisher Scientific 25200056) at 37oC for 5-7 minutes, and the CM was selectively enriched from the ectoderm using an Invitrogen

FlowComp Flexi Kit (Invitrogen 11060D) and a PDGFRα antibody (5-10 µg/2.5 million cells) (R&D Systems AF1062) (Goodnough et al. 2016) according to manufacturer’s guidelines.

2.4c: RT-qPCR

The CM+ectoderm was manually dissected from E13.5 embryos (described above). RNA was isolated as previously described (Hamburg-Shields et al. 2015). Relative mRNA expression levels were quantified using 5ng of cDNA on a StepOne Plus Real-Time

PCR System (Life Technologies) and the ΔΔCT method. Commercially available TaqMan 67 probes (Life Technologies) specific to each gene were used: Ezh2(Mm00468464_m1),

Suz12 (Mm01304152_m1), Eed (Mm00469660_m1), Sox9 (Mm00448840_m1), Axin2

(Mm00443610_m1) Twist2 (Mm00492147_m1), Keratin14(mm00516876_m1), and

Pdgfrα (Mm00440701_m1). CT values obtained for specific genes were normalized to those of β-actin (Invitrogen 4352663).

2.4d: Immunofluorescence

Heads of E13.5 embryos were fixed in 4% paraformaldehyde (PFA) for 30 min at

4ͦ C and cryopreserved as previously described (Atit et al. 2006). Rabbit polyclonal antibodies against H3K27me3 (1:1000; Cell Signaling 9733), LEF1 (1:100; Cell Signaling

2286), SP7/OSX (1:1000; Abcam ab94744), and SOX9 (1:1000; Millipore ab5535) were used for indirect immunofluorescence assays. Appropriate species-specific Alexafluor

594 secondary antibodies were used (1:500; Invitrogen). Images were captured using an

Olympus BX60 microscope and an Olympus DP70 digital camera using DC controller software. Confocal images were captured on a Leica TCS SP8 (Leica Biosystems) using

Application Suite X software (Leica Biosystems). Images were processed using

ImageJ/Fiji(Schindelin et al. 2012; Schneider et al. 2012) and Adobe Photoshop software. Images were prepared for cell counting in ImageJ/Fiji by subtracting background and thresholding the signal across all replicates. The percent of the cells which were H3K27me3 positive compared to DAPI was determined using the “analyze particles” feature in ImageJ/Fiji. Counting was performed on the supraorbital CM directly above the eye.

2.4e: Co-immunoprecipitation 68

E13.5 CM was collected by manual dissection, and the CM was enriched from the ectoderm (described above). At least 2 million cells from multiple embryos were incubated with 1 mL lysis buffer (50 mMTris pH 7.5, 250 mMNaCl, 2 mM EGTA, 1%

Triton X-100 in H2O) and disrupted with a 23-gauge needle and syringe. The precipitating antibody was added to the lysate and incubated overnight at 4oC.

Immunoprecipitation was performed by incubating 100 µL of Dynabeads Protein G beads (Invitrogen 10003D) with the lysate and antibody for 1h at 4oC. Beads were then washed 4 times with 1 mL wash buffer (40 mM HEPES, 300 mMNaCl, 10% Glycerol, 0.2%

NP40 in H2O). The sample was collected in NuPage LDS Sample Buffer (Thermo Fisher

Scientific NP0007) supplemented with β-mercaptoethanol, and was heated at 90oC for 5 min. Rabbit polyclonal antibodies against non-phospho β-catenin, 3.43 µg (Cell Signaling

D13A1 #8814); EZH2,10 µg (Diagenode C15410039); and IgG, 6 µg (Abcam ab46540) were used for immunoprecipitation. Protein species were separated by SDS-PAGE using

Mini-PROTEAN TGC gels (BioRad #456-1084). Western Blots were performed with primary antibodies against β-catenin (1:1000; Millipore 06-734), EZH2 (1:500, Cell

Signaling #5246). Clean-Blot IP Detection Reagent (HRP) (1:250; Thermo Fisher Scientific

21232) was used as secondary antibody.

2.4f: Protein Isolation and Immunoblotting

E13.5 CM+ectoderm was collected by manual dissection. Protein was isolated using RIPA buffer. Protein species were separated by SDS-PAGE using Mini-PROTEAN

TGC gels (BioRad #456-1084). Western Blots were performed with polyclonal rabbit primary antibodies against H3K27me3 (1:1000, Cell Signaling 9733), EZH2 (1:500, Cell 69

Signaling #5246), and SUZ12 (1:1000, Cell Signaling 3737). Species-specific HRP- conjugated secondary antibodies were used at a 1:10,000 dilution. Immunoblots were probed with a rabbit anti-β-tubulin antibody (1:400, Santa Cruz 9104) as a loading control. Signals were detected using an Amersham ECL Western Blotting Analysis

System (GE Healthcare RPN2109), and imaged using an Odyssey FC Imaging System (Li-

Cor). Relative protein levels were quantified using ImageJ/Fiji software(Schindelin et al.

2012; Schneider et al. 2012).

2.4g: RNA sequencing

E13.5 CM+ectoderm was collected by manual dissection (described above). Total

RNA was isolated from individual embryos as previously described (Hamburg-Shields et al., 2015). Libraries were prepared in the CWRU Genomics sequencing core using the

Illumina TruSeq Stranded Total RNA kit-with Ribo Zero Gold. Paired-end sequencing was performed on an Illumina HiSeq 2500 v2 Rapid Run flow cell. The resulting 100 bp reads were aligned to the mouse mm9 assembly using TopHat (Trapnell et al. 2009; Kim and

Salzberg 2011; Langmead and Salzberg 2012; Kim et al. 2013a). Genomic assembly was completed using Cufflinks v1.3 (Trapnell et al. 2010, 2013, Roberts et al. 2011b; a). mm9_reFlat was used to annotate the data with a maximum intron length of 20,000 bp and genomic bias correction. Cufflinks FPKMs below 0.3 were floored to 0.3. Differential gene expression was determined with CuffDiff using the default settings plus genomic bias correction. Gene ontology analysis examining all differentially expressed genes was performed using Genomic Regions Enrichment of Annotations Tool (GREAT) by 70 associating reads to the single nearest gene located within 5kb(McLean et al. 2010).

(Deposited in GEO, GSE96872)

2.4h: ChIP-sequencing

E13.5 CM+ectoderm was manually dissected from three En1Cre;β-cateninfl/+ and four En1Cre;β-cateninfl/∆ embryos, pooled, and H3K27me3 immunoprecipitation and sequencing was performed by Active Motif (www.activemotif.com) (Deposited in

GEO, GSE96872). 14 µg chromatin was immunoprecipitated with 4 µg rabbit anti-

H3K27me3 (Millipore #07-449). Sequencing was performed on an Illumina NextSeq 500 producing 75-nucleotide, single end reads. Drosophila DNA was “spiked in”. The ratio of aligned Drosophila reads in the mutant versus control samples (calculated to be 1.3) was used to normalize the number of reads in the mouse samples by downsampling the larger sample (mutant, in this case).

Sequences were aligned and analyzed twice independently. The first time using a custom pipeline consisting of Bowtie2for genome alignment to the mouse mm9 genome and Macs 1.4 at default settings for peak calling (Zhang et al. 2008; Langmead and

Salzberg 2012). To generate the windowed heat map from this analysis, the genome was divided into 40 windows of equal size 5 Kb up- and down-stream of each H3K27me3 peak genome wide or on peaks located within 1Kb of known promoters. The median peak signal in each window was then converted to a z-score and mapped using Java

TreeView (Saldanha 2004). Analysis was then performed second time using the NGS 2.8 pipeline (Strand NGS Manual, Version 2.8, Build 230243. © Strand Life Sciences,

Bangalore, India) and aligning to the mm10 genome. Peaks were called using Macs 1.4 71 at default settings. Association of peaks with specific genes was performed using

PAVIS(Huang et al. 2013). Specific H3K27me3 peaks were visualized using the Integrated

Genome Viewer (IGV)(Robinson et al. 2011; Thorvaldsdóttir et al. 2013). Ngs.plot was used to generate the average fold enrichment of H3K27me3 overview across the gene bodies (Shen et al. 2014).

2.4i: Cell Culture

The CM+ectoderm was manually isolated and dissociated by incubating the tissue in 0.25% Trypsin-EDTA (Thermo Fisher Scientific 25200056) at 37oC for 5-7 minutes, and then plated in DMEM. Fibroblasts were allowed to adhere to the plate for

1-2 hours, after which the media was removed and fresh media was added. Chemical inhibition was performed at no later than passage 3. Ten percent Wnt3a conditioned media and the chemical inhibitor, UNC1999 (Sigma SML0778) or GSK126 (Cayman

Medical CAS1346574-57-9) were added simultaneously. The cells were incubated for the indicated amount of time. Following incubation, the cells were trypsinized and processed for protein or mRNA analysis.

2.4j: Statistics

Graphs and statistical analysis was generated using Prism 6 (GraphPad Software).

Data are presented as mean ± SEM in all graphs unless otherwise stated. All pairwise sample comparisons were performed using a Mann-Whitney test. The p-values for statistical tests in all figures are represented as: * = P < 0.05 and ** = P < 0.01.

2.4k: Data Availability 72

Strains are publicly available at Jackson Laboratory. Sequencing data is available at GEO with the accession number GSE96872.

2.5: Acknowledgments:

I would like to thank the previous and current members of the Atit laboratory for excellent discussion and advice. Thank you to the Case Microscopy and Genomics Cores

Services. J. Ferguson and R. Atit conceived experiments; J. Ferguson, M. Devarajan, and

R. Atit. carried out the experiments; J. Ferguson, A. Saiakhova, C-F. Liu, and R. Atit analyzed the data; G. DiNuoscio performed genotyping and westerns; J. Ferguson, A.

Saiakhova, C-F. Liu generated figures; J. Ferguson, V. Lefevbre, P. Scacheri, and R. Atit interpreted the data and wrote the manuscript. The authors declare that they have no competing interests. This work was supported by the following grants: NIH-

T32AR007505 (J. Ferguson), NIH-NIDCR-R01DE01870 (R. Atit) NIH-R01DE01870; NIH-

R01CA160356 (P. Scacheri.), NIH-R01CA193677 (P. Scacheri), NIH-NIAMS R01 AR46249

(V. Lefebvre), NIH-NIAMS R01-AR68308 (V. Lefebvre), Case Western Reserve University

ENGAGE (M. Devarajan) and SOURCE Programs (M. Devarajan).

2.6: Conflict of interest:

We declare no conflicts of interest.

73

Chapter 3: Ezh2 is required for skull bone formation in a tissue- and developmental- stage specific manner

I would like Mahima Devarajan and Radhika Atit for their work and feedback on chapter

3.

Abstract

The bones of the skull are derived from cranial mesenchyme stem cell progenitors (CM) which originate from the cranial neural crest cells (CNCC) and head paraxial mesoderm

(PM). Formation of the skull bones requires the cross-talk between multiple signaling pathways in a spatial and temporal manner. The Polycomb Repressive Complex (PRC2) is an important epigenetic regulator of many signaling pathways that direct cell fate selection. Its epigenetic function is mediated by the H3K27me3 histone modification which is catalyzed by the methyltransferase, Enhancer of Zeste Homolog 2 (EZH2). The current conditional Ezh2 mutant models indicate a spatial and temporal requirement during craniofacial development. The specific functions of Ezh2 in vivo between CNCC- and PM-derived cranial bone development remains unclear. Here, using a temporally- inducible conditional knockout of Ezh2 in the CM, we investigate a tissue and stage- specific role of Ezh2 in cranial bone development. We find the timing of Ezh2 is critical early in skull bone development to establish craniofacial bone lineage commitment. We demonstrate that loss of Ezh2 in the CM at embryonic day (E) 8.5 (E8.5-CMEzh2) results in a reduction of the CNCC-derived bones and a near complete loss of the PM-derived bones. In contrast, loss of Ezh2 at E9.5 (E9.5-CMEzh2) does not lead to a loss of PM- 74 derived bones. In E8.5-CMEzh2 mutants, the number of OSX-positive skeletogenic progenitors was diminished in CNCC-derived frontal bone primordia and nearly absent in the PM-derived parietal bone primordia. Furthermore, loss of Ezh2 led to an increase in retinoic acid (RA) signaling target genes, and administration of a pan retinoic acid receptor (RAR) antagonist, BMS-453, in E8.5-CMEzh2 mutants partially restored OSX expression and rescued the parietal and occipital bone development. Thus, our results show that EZH2 has a stage-specific role and inhibits targets of the RA signaling pathway during skull bone progenitor allocation in the CM.

3.1: Introduction:

The cranial mesenchymal stem cells (CM) which give rise to the bones of the head and face originate from two different cell origins, the cranial neural crest cells

(CNCC) and the paraxial mesoderm (PM) (Jiang et al. 2002; Yoshida et al. 2008). The

CNCC gives rise to the frontal, medial portion of the interparietal, and facial bones. The

PM primarily gives rise to the parietal, lateral portion of the interparietal, and occipital skull bones. Both the CNCC- and the PM-derived bones ossify through intramembranous ossification. The calvarial bones, which are the bones surrounding the brain, present a unique developmental scenario in which to study skull bone development. The calvarial bone primordia originates from a population of CM located directly above the eye in the supraorbital arch (SOA). From E12.5 onwards, the SOA-CM migrates apically over the brain and differentiates into the calvarial bones. These bones are derived with contributions from the CNCC, PM, or both. (Jiang et al. 2002; Karsenty 2008; Yoshida et 75 al. 2008; Roybal et al. 2010; Tran et al. 2010; Ishii et al. 2015). A greater understanding of how multiple stem cell populations can be organized and coordinated to form the bones of the calvaria will help provide insights into the intricacies of skull bone formation.

The initiation of bone formation involves the sequential expression of multiple factors; Msh Homeobox 1 (Msx1) and 2 (Msx2), then Related Transcription Factor 2

(Runx2), and finally Osterix (Osx/Sp7). Msx1 and Msx2 represent bone precursors. Runx2 expression leads to the cell fate selection into bone progenitors, and Osx expression leads to a commitment of the bone progenitors to a skeletogenic fate. The genetic signals and mechanisms governing the specification of the CM into committed skull bone progenitors is not fully elucidated.

Multiple signaling pathways have been shown to be required for the formation and differentiation of the calvarial bone including Fibroblast Growth Factor (FGF), Bone

Morphogenetic Protein (BMP), Wnt/β-catenin, and retinoic acid (RA) (Ishii et al. 2015;

Liu et al. 2016). For example, mutations in Wnt/β-catenin signaling leads loss of calvarial bone and ectopic chondrogenesis (Goodnough et al. 2012). Disruptions in various BMP signaling components can lead to numerous skull bone defects ranging from loss of bone structures to fused bones (Graf et al. 2016). FGF signaling promotes skull bone ossification with gain-of-function mutants exhibiting fused skull bones (Ornitz and Marie

2002; Moosa and Wollnik 2016; Pfaff et al. 2016). Increases or decreases in RA signaling has been shown to lead to an inhibition of bone growth and reduced skull bones in vivo

(Lohnes et al. 1994; Kochhar et al. 1998; Abe et al. 2008; Maclean et al. 2009). The 76 spatial organization and developmental timing by which these signaling pathways are orchestrated in vivo resulting in the formation of the skull bones is unclear (Abe et al.

2008; Graf et al. 2016).

The Polycomb Repressive Complex 2 (PRC2) is an important epigenetic regulator in the head and face. PRC2 is a multi-protein complex which catalyzes the trimethylation of histone 3 on lysine 27 (H3K27me3) and plays a role in many aspects of development

(Lund and Van Lohuizen 2004; Simon and Kingston 2009; Minoux et al. 2017;

Schuettengruber et al. 2017). In humans, mutations in Ezh2, the catalytic component required for H3K27me3, leads to Weavers’s syndrome which is characterized by craniofacial defects (Tatton-Brown et al. 2011; Gibson et al. 2012). In mice, conditional deletion of Ezh2 in embryos leads to varying craniofacial bone defects depending on the timing and tissue of deletion. Loss of Ezh2 in the pre-migratory CNCC using Wnt1Cre leads to a dramatic loss of the facial bones along with the frontal bone (Schwarz et al.

2014). Surprisingly, mice lacking Ezh2 in PM derived tissues using Prx1Cre, display craniosynostosis of multiple sutures during post-natal development and potentially increased bone (Dudakovic et al. 2015). These data indicate a potential spatial and developmental-stage specific role for Ezh2 in skull bone formation. However, the mechanisms governing the stage and lineage-specific role of Ezh2 in calvarial bone formation is unclear.

In order to address the spatiotemporal role of Ezh2 in skull bone formation, we created an inducible, conditional knockout of Ezh2 that targets the post-migratory CNCC and PM derived CM. Loss of Ezh2 around E8.5 resulted in defect in the fate 77 determination of the PM-derived bone progenitors leading to a loss of bone. In contrast, no such defect was observed after loss of Ezh2 around E9.5. These results are striking because defects in bone differentiation were not observed until E13.5 indicating Ezh2 is required at E8.5 to for lineage selection prior to the expression of Msx genes.

Furthermore, inhibition of RA signaling in Ezh2 mutants rescues the skull bone defect.

Our results indicate that loss of Ezh2 expression is required to maintain suppression of

RA target genes ensuring bone fate commitment.

3.2: Results:

3.2a: Conditional deletion of Ezh2 in cranial mesenchyme stem cells

In order to address the stage-specific roles of Ezh2 during CNCC- and PM-derived skull bone development, we used PdgfrαCreER, a tamoxifen-inducible Cre, to conditionally activate the Rosa26Reporter (R26R) and delete Ezh2 in the post-migratory

CM prior to bone cell fate selection (PdgfrαCreER;Ezh2fl/fl). Administration of tamoxifen by oral gavage to pregnant females at E8.5 and E9.5 (E8.5-CMEzh2) was sufficient to induce β-galactosidase expression in the majority of the CNCC-derived (plane I) and PM- derived (plane II) CM by E10.5 (Fig. 3.1A-C). By E13.5, a whole mount phenotype was apparent in the E8.5-CMEzh2 mutants in the form of a truncated mandible and frontonasal prominence, and extensive β-galactosidase expression was maintained 78

Figure 3.1: Tamoxifen induced knockout of Ezh2at E8.5 in both the mesoderm- and neural crest- derived mesenchymal stem cells is sufficient to lead to craniofacial defects. (A) Mating strategy and gavage regimen for Ezh2 knockout for E8.5-CMEzh2 mutants. Tamoxifen was administered by oral gavage at E8.5 and E9.5 at a concentration of 25µg/g mouse body weight. (B) Anatomy of E9.5 mouse embryo. PdgfrαCreER is active in the cranial-CM, frontonasal prominence, maxillary process, and BA1. Plane I corresponds to the future frontal bone, and plane II corresponds to the future parietal bone. (C) Rosa-reporter staining in E10.5 coronal sections. E8.5+E9.5 gavage is sufficient to induce Cre-recombination in plane I and plane II. (D) Schematic representing manual isolation of cranial mesenchyme (CM). The ectoderm was manually removed and all the CM above the eye was collected. (E) RT-qPCR for Ezh2 in the isolated CM. (F) Western blot for EZH2 in the isolated cranial mesenchyme. Band intensities were quantified using ImageJ. (G) Western blot for

H3K27me3 in isolated CM. (H) Whole mount images of E8.5-CMEzh2 mutants at E17.5. Embryos exhibit domed skulls, reduced snout, and reduced mandible. Scale bar = 200µm

79 throughout the entire facial, dentary, and cranial mesenchyme (Fig. S3.1A,B).

Figure S3.1: Whole mount analysis of E13.5 and E17.5 E8.5-CMEzh2 mutants. (A) Gross phenotype of

e8.5-CMEzh2 mutant embryos at E13.5. The shortened snout and limbs are beginning to be apparent

(B) Rosa-reporter staining at E13.5 coronal sections. (C) E8.5-CMEzh2 mutants also exhibit truncated

limbs, defects in vasculature, and omphalocele. Vasculature defects are apparent by bleeding under

the skin and few identifiable major vessels.

80

To quantify the extent of Ezh2 loss, we assayed for Ezh2 mRNA and protein in

E8.5-CMEzh2 mutants. Fluorescent reporters are too sensitive for PdgfrαCreER and thus preventing isolation of the CM by fluorescence. As a result, we isolated the CM by manual dissection. E13.5 marks the time point by which a sufficiently pure CM population can be isolated by manual dissection (Fig. 3.1D). At E13.5, we found an 80% reduction in relative amounts of Ezh2 mRNA and a 94% reduction in EZH2 protein in

E8.5-CMEzh2 mutants compared to Cre-negative controls (Fig. 3.1E,F). As a readout to further confirm our deletion of Ezh2, we examined the bulk level of H3K27me3, the repressive histone modification catalyzed by EZH2. Compared to the Cre-negative control E13.5 manually isolated CM, we found a near complete loss of bulk H3K27me3 protein in E8.5-CMEzh2 mutants (Fig. 3.1G).

At E17.5, the shortened mandible and maxilla were very pronounced along with a domed head (Fig. 3.1H). In addition, E8.5-CMEzh2 mutants exhibit shortened limb, defects in the vasculature, and omphalocele (Fig. S3.1C). It is worth noting that we obtained the expected ratio of mutants by Mendelian genetics for our genetic cross, suggesting lack of embryonic lethality (data not shown, (n=39)). These results show that

E8.5-CMEzh2 leads to pleiotropic effects on multiple structures and craniofacial deformities of CNCC- and PM-derived structures.

3.2b: E8.5-CMEzh2 mutants have decreased craniofacial bone volume and size

To further characterize the skull bone defects following loss of Ezh2, we performed micro-computed tomography (microCT) and morphometric analysis on the skulls of E17.5 controls and E8.5-CMEzh2 mutants. We focused our analysis on the bones 81 of the skull vault and the dentary bones of the face which are the most readily identifiable by E17.5 (Fig. 3.2A).

Figure 3.2: e8.5-CMEzh2 leads to a reduction of CNCC-derived bones and a severe reduction in PM-

derived bones.

82

Figure 3.2 cont. (A) Schematic and key representing the primary bones examined in e8.5-CMEzh2

embryos. (B) 3D images from microCT at e17.5 of e8.5-CMEzh2 embryos.* indicates ear bones which

are lost in the mutants. Arrows point to the reduced/lost PM-derived bones. Numbers represent

points of measurements morphometric measurements. (C) Quantification of changes in combine

bone volume of the calvaria, mandible, maxilla, premaxilla, and nasal bones. (D) Quantification of

changes in bone volume in the bones of the calvaria. (E) Morphometric analysis of the frontal and

interparietal bone. Both the left and right bone was measured and plotted. Colored dots

correspond with each left-right pair.

The microCT analysis revealed malformations or losses in the majority of the head and face bones and a 65% decrease in overall bone volume in the mutants (control:

1.013+0.09, mutant: 0.357+0.11; n=3 controls, 4 mutants) (Fig. 3.2B,C). 83

Figure S3.2: Quantification of the mandible and snout in e8.5-CMEzh2 mutants. (A) Quantification of the length and volume from the mandible microCT 3D images. (B) Quantification of the length and volume of the pre-maxilla, maxilla, and nasal bones from the microCT 3D images. (C) Representative

3D reconstruction images of the interparietal bone.

84

To quantify the craniofacial bone deformities in more detail, we used the recently described landmarks in embryonic craniofacial bones (Ho et al. 2015). In the face, which includes the mandible, maxilla, pre-maxilla, and nasal bones, the morphology of the bone was compromised, but the bones were still present (Fig. 3.2B, Fig.S3.2A,B). We observed the overall length of the mandible from the posterior point of the condylar process to the most anterior point of the mandible was decreased by 60% (control:

1.063+0.11, mutant: 0.4157+0.046p=0.0357), and the bone volume was reduced by 83% relative to the control (control: 1.022+0.04, mutant: 0.17+0.12; n=3 controls, 4 mutants)

(Fig.S3.2A). The combined length of maxilla and premaxilla from the posterior-medial point of the palatine process of the maxilla to the most distal point was decreased by

28% in the mutant (control: 1.035+0.08, mutant: 0.7419+0.09; p=0.0357). The combined volume of the nasal, premaxilla and maxilla region was reduced by 42% in the E8.5-

CMEzh2mutant (control: 1.059+0.28, mutant: 0.6097+0.11; n=3 controls, 4 mutants) (Fig.

S3.2B). It is worth noting, we could not clearly identify the premaxilla bone, and it was unclear if the premaxilla was missing or it was fused with the maxilla.

Interestingly, in the calvaria, the CNCC- and PM-derived bones were differentially sensitive to the deletion of Ezh2. The morphology of the CNCC-derived frontal bone was compromised, and the relative volume was diminished by 39% (control:0.9355+0.0.43, mutant:0.5687+0.171; n=3 controls, 4 mutants). In contrast, the PM-derived parietal bone volume was decreased by 85% (control: 1.158+0.33, mutant:0.1643+0.14; n= 3 controls, 4 mutants) and the PM-derived occipital bone was nearly absent in the mutant

(Fig. 3.2B,D). The interparietal bone, which has contributions from both the CNCC and 85

PM, has similar morphology between the controls and mutants, but the relative volume was decreased by 45% (control: 1.117+0.0.66, mutant: 0.611+0.30; n=3 controls, 4 mutants) (Fig. 3.2D, Fig. S3.2C). In addition, the temporal bone, inner ear bones, and the tympanic ring were absent in theE8.5-CMEzh2 (Fig. 3.2B). These results indicate differential sensitivities to Ezh2 expression among the bones of the skull.

To further analyze the morphology of the CNCC-derived frontal bone and interparietal bone, we performed morphometric measurements to quantify changes in the relative dimensions of the frontal and interparietal bones (Fig. 3.2B,E). Relative to the controls, both the length and height of the frontal bone was reduced by roughly

25% (length: control:1.012+0.06, mutant: 0.7596+0.10; p=0.0005; height: control:

1.054+0.16, mutant: 0.7897+0.2; p=0.03) in the E8.5-CMEzh2 mutants. In addition, the distance (width) between the most posterior-superior points of the left and right frontal bones was increased by 1.5 fold (control: 0.9683+0.12, mutant: 1.521+0.08; p=0.036) in the mutants indicating larger fontanel. Despite an increase in overall volume (Fig. 2D), the interparietal bone showed no major changes in morphology with only a 4% decrease in length in the mutant (control: 0.9939+0.0.09, mutant: 0.9584+0.05; p=n.s.) (distance between points 7-8).

Varying levels of deformities in the mandible, the snout region, and the skull vault in the E8.5-CMEzh2 mutant demonstrates the differential requirement for Ezh2 in bone formation between the CNCC and PM derived bone, and a greater disruption of the PM-derived parietal bone and occipital bone indicates the PM is more sensitive to the loss of Ezh2. 86

3.2c: The effect of Ezh2 on skull bone formation is developmental-stage specific

Figure 3.3: Ezh2 is required for skull bone formation in a developmental stage-dependent manner.

(A) Gavage regimen for E9.5-CMEzh2 mutants. Tamoxifen was administered by oral gavage at E9.5 and

E10.5 at a concentration of 25µg/g mouse body weight. (B) RT-qPCR for Ezh2 in E8.5-CMEzh2 and

E9.5-CMEzh2 mutants from manually isolated CM. (C) Skeletal preparations on Ezh2 mutants. E8.5-

CMEzh2 mutants lead to a loss of PM-derived structures, but the E9.5-CMEzh2 mutants do not. Alcian

blue marks cartilage and alizarin red marks bone. Arrows mark unclear sutures in E9.5-CMEzh2

mutants. F=frontal bone, P=parietal bone, IP=interparietal bone, O=occipital bone. Scale bars = 500

pixels

To address the stage-specific role of Ezh2 in the CM, we conditionally deleted

Ezh2 by administration of tamoxifen through oral gavage at E9.5 and E10.5 (E9.5-CMEzh2) 87

Figure S3.3: Ezh2 is lost in the same tissue in the E8.5-CMEzh2 and E9.5-CMEzh2 mutant. (A) Rosa- reporter staining in E9.5-CMEzh2 mutants. (B) Schematic of coronal sections at the frontal bone primordia. (C) H3K27me3 immunofluorescence in coronal sections in the frontal bone primordia.

88

(Fig. 3.3A). Similar to the E8.5-CMEzh2 mutants, a significant reduction in Ezh2 mRNA levels was observed in the E9.5-CMEzh2 mutant CM (Fig. 3.3B). Furthermore, the β- galactosidase recombination (Fig. S3.3A compare to Fig. S3.1B) and loss of H3K27me3

(Fig. S3.3C) was comparable between the E8.5-CMEzh2 and E9.5-CMEzh2 mutants.

Surprisingly, unlike the E8.5-CMEzh2 mutants, the skeletal staining showed the PM- derived skull bones in the E9.5-CMEzh2 mutant were not lost. Ossification in what appeared to be the parietal and occipital bone was found suggesting that Ezh2 is not required for the ossification of the PM-derived bones after E9.5 (Fig. 3.3C). The coronal and lamboidal suture were not readily visible in the E9.5-CMEzh2 mutants, as a result, we were unable to accurately identify and quantify the area of the frontal, parietal, and interparietal bones individually. It is worth noting, the squamous portion of the temporal bone and the tympanic ring were still absent in the E9.5-CMEzh2 mutants. In addition, the facial bones and mandible remained truncated and the maxilla and premaxilla bones remained fused as previously seen in the E8.5-CMEzh2 mutants. These results together suggest a developmental-stage specific role for Ezh2 in the formation of the skull bones and implies a defined temporal window around E8.5 for Ezh2 is required for the formation of PM-derived skull bone formation.

3.2d: Diminished cranial bones in E8.5-CMEzh2 mutants is not due to defects in cell survival and proliferation at E10.5

Because mutants lacking Ezh2 exhibit an overall decrease in bone and other structures in late fetal stages, we next examined the effect on cell survival and 89 proliferation at E10.5 in E8.5-CMEzh2 mutants. We focused our analysis on the regions corresponding to the future frontal and parietal bone primordia; plane I and plane II respectively (Fig. 3.4A). 90

Figure 3.4: The increased cell death in E8.5-CMEzh2 mutants is insufficient to account for loss of

skull bones. (A) Schematic of E10.5 mouse embryo and coronal sections. Plane I refers to future

frontal bone and plane II refers to the future parietal bone. Blue region on the coronal

representations indicates the area of cell counts. (B) Immunofluorescence for cell death by activated

Caspase-3 on E10.5 coronal sections in plane I and plane II in E8.5-CMEzh2 mutants. (C) Quantification

of activated Caspase-3 immunofluorescence, total number of DAPI positive cells, and EdU staining in

E8.5-CMEzh2 mutants. (D) Quantification of activated Caspase-3 in E9.5-CMEzh2 mutants.

Quantification of the cells positive for activated Caspase-3 in the CM revealed a 2.5% 91 increase in plane I and 5.3% increase that didn’t achieve statistical significance in plane II in E8.5-CMEzh2 mutants (Fig. 3.4B,C). In addition, at E10.5, a comparable number of total

DAPI positive cells between controls and mutants was seen at all positions showing, the total cell number was consistent (Fig. 3.4C). We also did not find changes in cell survival by activated Caspase-3 at E11.5 and 13.5 (data not shown). In addition, no changes in the cell proliferation index was detected by 5-ethynyl-2´-deoxyuridine (EdU) staining in plane I and a non-statistically significant decrease was observed in plane II CM (Fig. 3.4C;

Fig. S3.4A). 92

Figure S3.4: Significant increase in cell death, but no change in cell proliferation, in the frontonasal process in E8.5-CMEzh2 mutants. (A) Fluorescence staining for cell proliferation by EdU in E8.5-CMEzh2 mutants. Plane I corresponds with future frontal bone and plane II corresponds with future parietal bone. (B) Quantification of cell death by activated Caspase-3, total number of cells, and proliferation by EdU in the frontonasal process in E8.5-CMEzh2 mutants. (C) Immunofluorescence for H3K27me3 in plane I and plane II in E8.5-CMEzh2 mutants.

93

Cell death and proliferation results in the frontonasal process were similar to plane I with an 8% increase in activated Caspase-3 cells and no change in EdU or total cell number (Fig. S3.4B). Strikingly, these results occur in a tissue which is still primarily positive for H3K27me3 indicating 48 hours after initial gavage is not sufficient to fully de-methylate the CM to detectable levels (Fig. S3.4C).

To compare the cell survival between E8.5-CMEzh2 and the E9.5-CMEzh2, we also examined the number of cells with activated Caspase-3 at an equivalent time point in the E9.5-CMEzh2 mutants. At E11.5, we found a marginal increase in number of activated

Caspase-3 expressing cells (Fig 3.4D). Considering the differences in cell death between

E8.5-CMEzh2, which has a dramatic reduction of bone, and the E9.5-CMEzh2, which does not have a dramatic loss of bone in the CM, is only 2-3%, it is unlikely that changes in cell survival and proliferation account for the loss of PM-derived skull bone in the E8.5-

CMEzh2 mutants.

3.2e: E8.5-CMEzh2 mutants exhibit defects in the differentiation of the skull bone progenitors

We next wanted to determine if the defects in the skull bones result from an arrest in cell fate selection and commitment in E8.5-CMEzh2 mutants. Therefore, we examined expression of genes in the cranial bone initiation program in the frontal (Plane

I) and parietal bone (Plane II) primordia (Fig. 3.5A) (Karsenty 2008). 94

Figure 3.5: Loss of Ezh2 at E8.5 leads to defects in bone progenitor differentiation.

At E11.5, the bone precursor marker, MSX1/2, had comparable expression domains in both plane I and plane II, and a similar number of total cells expressing MSX1/2 between controls and E8.5-CMEzh2 mutants (Fig 3.5B). At E13.5, the expression of bone progenitor 95 markers RUNX2 and alkaline phosphatase (AP) was comparable in both plane I and plane II in the controls and E8.5-CMEzh2 mutants (Fig. 3.5C, Fig. S3.5A).

Figure 3.5 cont. (A) Schematic of E13.5 mouse embryo and coronal sections. Plane I refers to future

frontal bone and plane II refers to future parietal bone. (B) Immunofluorescence of bone precursor

marker MSX1/2 in E11.5 coronal sections. Quantification of total number of cells positive for MSX1/2

in plane I and plane II. (C) Immunofluorescence for bone progenitor marker RUNX2 in E13.5 coronal

sections. Arrows indicate expanded domains. Quantification of total number of cells positive for

RUNX2 in plane I and plane II. (D) Immunofluorescence for bone progenitor marker OSX in E13.5

coronal sections. Arrows indicate lost expression in plane II. Quantification for total number of cells

positive for OSX in plane I. Quantification in plane II was not performed due to lost expression in the

mutant.

96

Figure S3.5: No change in bone marker, alkaline phosphatase (AP) in E8.5-CMEzh2 mutants and OSX

in E9.5-CMEzh2 mutants. (A) AP staining in plane I and plane II in E13.5 coronal sections. (B)

Immunofluorescence on E13.5 coronal sections for OSX in E9.5-CMEzh2 mutants.

However, the morphology of the RUNX2 domain was shifted ventrally in the frontal bone and less compacted in the parietal bone primordia in the E8.5-CMEzh2 mutants. The number of RUNX2 positive cells in plane I and plane II was slightly diminished in the mutants but did not approach statistical significance (Fig. 3.5C). RUNX2 is first step in lineage selection and the establishment of bone progenitors (Komori et al. 1997;

Karsenty 2008). These results suggest the disruption in bone formation in E8.5-CMEzh2 97 mutants is not due to failure to establish the bone progenitors. Next, we examined the expression OSX, which is downstream of RUNX2, and is required for the cell fate commitment of the bone progenitors and intramembranous ossification (Nakashima et al. 2002; Karsenty 2008). At E13.5, we found the OSX positive domain and the number of OSX positive cells at E13.5 were diminished in plane I in the E8.5-CMEzh2 mutants, and the OSX expression is nearly absent in plane II (Fig. 3.5D). In comparison to the E8.5-

CMEzh2 mutants, the E9.5-CMEzh2 mutants do not show a noticeable disruption in the establishment of the OSX-positive domain (Fig. S3.5B). The difference in bone defects in the E8.5-CMEzh2 mutants and E9.5-CMEzh2 mutants indicates that Ezh2 is required in a defined temporal window for the expression of OSX in the bone progenitor domain, and failure to express OSX may lead to the bone ossification defects seen in the E8.5-CMEzh2 mutants at E17.5.

3.2f: In vivo inhibition of Retinoic Acid signaling partially restores skull bones

Because Ezh2 is typically known for its repressive function, the loss of OSX expression in E8.5-CMEzh2 mutants most likely involves an intermediate signaling factor.

To identify such a factor, we examined readouts of various signaling pathways important in early bone development known to be regulated by Ezh2 (Bracken et al.

2006a; Schwarz et al. 2014; Mirzamohammadi et al. 2016). In manually isolated CM of

E13.5 E8.5-CMEzh2 mutants (Fig. S3.6A), we did not find significant changes in mRNA levels of Sprouty2, Serpine1, Id1, and Axin2 in which are readouts for FGF, TGF-, BMP, and Wnt signaling pathways respectively (Fig. S3.6B). 98

Figure S3.6: E8.5-CMEzh2 mutants exhibit an ectopic expression of the Hox genes. (A) Expression

analysis was performed on E13.5 manually isolated CM. (B) RT-qPCR for signaling pathways known

to play a role in skull bone formation. (C) RT-qPCR for a subset of Hox genes. (D) RT-qPCR for HoxC8.

Surprisingly, only a subset of the genes queried exhibited considerable increase (>10 fold) in expression levels. These genes consisted of the Hox genes, the to cell cycle regulator, Cdkn2a, and Hand2 (Table 3.1). 99

Table 3.1: Fold expression changes in E8.5-CMEzh2 mutants relative to controls. 100

Previous published studies have identified an upregulation of various Hox genes, depending on the tissue and genetic conditions, in Ezh2 mutants (Schwarz et al. 2014;

Dudakovic et al. 2015; Minoux et al. 2017). However, it is not clear if the dysregulation of the Hox genes results in the skeletal defects observed. Hox genes are positional transcription factors that are normally absent in the cranial region (Krumlauf 1994;

Creuzet et al. 2002). In our E8.5-CMEzh2 mutants, we found a subset of the known Ezh2- targeted Hox genes were variably upregulated with HoxC8 as the most upregulated with a 354 fold increase in mRNA levels (Fig. S3.6C,D). In vivo, overexpression of leads to a near complete loss of skull bone and facial structures identifying HoxC8 as an anti- osteogenic factor (Carroll and Capecchi 2015). In order to determine if an upregulation of HoxC8, or other Hox genes, results in the loss of skull bones in our E8.5-CMEzh2 mutants, we attempted to inhibit the ectopic Hox gene expression. The Hox genes are known to be positively regulated by retinoic acid (RA) signaling (Krumlauf 1994; Balmer and Blomhoff 2002; Lee et al. 2014; Savory et al. 2014). In addition, mice lacking

Cyp26b1, which is required for breakdown of RA, have craniofacial and limb defects with striking similarity to E8.5-CMEzh2 mutants (Maclean et al. 2009). Based on these data, we identified RA signaling as a potential target through which to inhibit Hox genes expression. BMS-453 is a pan-RAR antagonist highly specific to RAR-γ (Chung et al.

2011). In the CM, retinoic acid receptor gamma (RAR-γ) was the most abundant RAR in both controls and E8.5-CMEzh2 mutants (Fig. S3.7A). 101

Figure S3.7: The effects of RA signaling disruptions in the CM. (A) RT-qPCR for the three RARs in

E13.5 manually isolated CM. (B) Gross phenotypes of the different doses of BMS-453. (C) Skeletal preparations examining the effect on skull bones following administration of exogenous all-trans RA.

Alcian blue marks cartilage and alizarin red marks bone. (D,E) RT-qPCR for changes in HoxC8 and

HoxA1 expression levels following administration of BMS-453 in E13.5 manually isolated CM. (F) RT- qPCR for changes in Ezh2 expression levels following administration of BMS-453 in E13.5 manually isolated CM.

102

Figure 3.6: Chemical inhibition of RA signaling partially rescues the in E8.5-CMEzh2 mutant

phenotype and restores the PM-derived bones. (A) Gavage regimen for tamoxifen and RA-

antagonist BMS-453 in E8.5-CMEzh2 mutants. (B) RT-qPCR for Crabp2 in E8.5-CMEzh2 and E8.5-CMEzh2 +

BMS-453 mutants from E13.5 manually isolated CM. (C) Skeletal preparations on 3.5µg/g BMS-453

treated Ezh2 mutants. C' and C'' represent two different litters. Alcian blue marks cartilage and

alizarin red marks bone. (D) Quantification of the Alizarin Red area. Quantification was performed

using ImageJ. (E) Immunofluorescence of OSX in the parietal bone at E13.5. Arrows indicate partial

restoration of OSX. Scale bar on immunofluorescence = 200µm.

We administered BMS-453 at 3.5µg/g and 5µg/g body weight at E8.5, E9.5, and E11.5 by

103 oral gavage to pregnant dams and manually collected the CM at E13.5 (Fig. 3.6A). We found a down-regulation of RA signaling target gene Crabp2 in both controls and E8.5-

CMEzh2 mutants demonstrating the efficiency of BMS-453 (Fig. 3.6B). At E17.5, the 5µg/g body weight dose resulted in a drug-induced phenotype in the controls. There was no detectable phenotype in the controls at the 3.5µg/g body weight dose and what appeared to be a reduced phenotype in the E8.5-CMEzh2 mutants treated with BMS-453

(Fig. S3.7B). Analysis of the skull bones in the 3.5µg/g body weight dose further revealed a dramatic rescue of the parietal and occipital bone in the E8.5-CMEzh2 mutants as compared to untreated mutants (Fig. 3.6C compared to Fig. 3.3E). We also observed improvement of mandible and maxilla bones and partial recovery of the inner ear bones in some treated mutants. In the three E17.5 litters analyzed, two out of three showed partial rescue of parietal and occipital bone in E8.5-CMEzh2 mutants (Figure 3.6C’, C’’).

Quantification of the relative area of alizarin red stained bone also showed a recovery of the parietal bone (Fig. 3.6D). In addition, the quantification revealed a small, but statistically significant, decrease in area of the frontal bone in both the control and E8.5-

CMEzh2 mutant embryos dosed with BMS-453. Surprisingly, there was no change in

HoxC8 or HoxA1, another ectopically expressed Hox gene, expression at E13.5 in the

3.5µg/g body weight dose with only a decrease observed at the 5µg/g dose (Fig.

S3.7D,E).

To determine if BMS-453 treatment restored differentiation of the bone progenitors at E13.5, we examined OSX expression in the parietal bone primordia.

Compared to non-treated E8.5-CMEzh2 mutants which expressed no OSX in the parietal 104 bone primordia (Fig. 3.5D), E8.5-CMEzh2 mutants treated with BMS-453 showed a partial restoration of OSX expression (Fig. 3.6E). While the OSX expression does not appear to reach the levels of the control, any expression indicates a potential restoration of parietal bone cell fate commitment. It is worth noting, shorter treatment with BMS-453 only at E8.5-9.5 did not rescue the parietal bone phenotype as well as the full five doses

(n=1, data not shown). In addition, administration of exogenous all-trans RA to pregnant dams does not lead to a skull bone phenotype or exacerbate the Ezh2 phenotype. These results indicate that Ezh2 does not directly regulate RA signaling, but regulates a downstream target of RA (Fig. S3.7C). Strikingly, Ezh2 expression is also affected by

BMS-453, and treated controls also have a down-regulation of Ezh2 despite having no phenotype further demonstrating reliance on RA signaling as a mediator of the skull bone phenotype (Fig. S3.7E). These results together indicate that a downstream target of RA signaling regulated by Ezh2 in skull bone formation, and exquisite timing and levels of RA signaling is required to rescue the E8.5-CMEzh2 mutants.

3.3: Discussion:

In this study, by conditional inactivation of Ezh2 in post-migratory CNCC and PM, we provide evidence for a stage-specific requirement of Ezh2 for osteoblast fate commitment. Conditional deletion of Ezh2 in the CM leads to normal lineage selection of the bone progenitors in both the CNCC- and PM-derived bone, but an interruption in the further commitment towards osteoblast lineage primarily in the PM-derived bone.

In addition, we show that Ezh2 is required to restrict a population of mesenchymal stem 105 cells to the skull bone lineage prior to the establishment of the skull bone precursors.

Finally, we show a partial restoration of calvarial bone formation in Ezh2 mutants by inhibiting RA signaling during the cranial bone formation.

3.3a: The role of Ezh2 in the lineage-restriction of cranial bone progenitors

Based on the expression of the early cranial bone progenitor markers, the CNCC progresses through the bone initiation program roughly one to two days before the PM

(Han et al. 2007). Expression of bone precursor markers, Msx1 and Msx2, can first be detected in the CNCC at E10.5. Expression of Msx2, but not Msx1, can first be detected in the PM at between E11.5 and E12.5. Through inducible, temporal-specific deletions, our data comparing E8.5-CMEzh2 and E9.5-CMEzh2 mutants suggests that lineage- restriction of parietal bone progenitors occurs between E8.5 and E9.5 in the PM prior to

Msx2 expression. Furthermore, a loss of lineage restriction does not result in observable defects until E13.5 when the bone progenitors fail to express Osx. Thus, Ezh2 is required at a specific developmental stage to ensure the bone progenitors commit to an osteoblast fate at E13.5.

Other genetic mutants of Ezh2 further demonstrate its role in lineage restriction at specific developmental-stages. Loss of Ezh2 in pre-migratory neural crest at E8.5 using

Wnt1Cre leads to a near complete loss of the CNCC derived bones (Schwarz et al. 2014).

In contrast, loss of Ezh2 in the CNCC and PM around E9.5 using Dermo1Cre does not result in a loss of skull bones (Ferguson et al. 2017). These data indicate that Ezh2 may 106 be required to lineage restrict the CNCC-derived bone progenitors prior to E9.5 ensuring they will proceed down the bone initiation pathway at E10.5.

In addition to the cranial bones, the concept of lineage restriction at specific developmental stages has also been observed in the trunk. In multiple organisms, pre- migratory CNCC have been shown to be lineage restricted in the neural tube (Schilling and Kimmel 1994; Dorsky et al. 2000; Krispin et al. 2010). When cells are lineage marked prior to migration, the progeny of a single cell often populate the same body segment after migration and differentiate into the same cell type indicating that cell had some level of lineage determination in the neural tube. In light of our results, a similar process of lineage restriction, regulated by Ezh2, may be occurring to ensure formation of the skull bones.

3.3b: The genetic mechanism involved in lineage selection of the cranial bones by Ezh2

Ezh2 has been associated with many different signaling pathways required for skull bone development (Bracken et al. 2006b). Signaling pathways such as Wnt/β- catenin, BMP, FGF, and RA signaling have all been shown to affect cell fate selection and differentiation of the skull bones (Lohnes et al. 1994; Li and Cao 2006; Abzhanov et al.

2007; Maclean et al. 2009; Yang et al. 2010; Goodnough et al. 2012; Fan et al. 2016; Li et al. 2017). In our E8.5-CMEzh2 mutants, we did not see any dramatic changes to these signaling pathways (Fig. S6). Our data demonstrates, however, that inhibition of RA signaling can partially rescue the phenotype and restore the parietal bone indicating 107 that Ezh2 inhibits RA signaling or a downstream target of RA signaling to ensure skull bone formation.

Previous studies have demonstrated a wide range of craniofacial phenotypes associated with RA signaling defects. Retinoic acid receptor α and γ (Rarα; Rarγ) compound null mutants exhibit phenotypes such as exencephaly and underdeveloped skull (Lohnes et al. 1994). Loss of Cyp26B1, which negatively regulates RA signaling, leads to craniofacial defects similar to those observed in the E8.5-CMEzh2 mutants

(Maclean et al. 2009). Considering the RA signaling has been shown to regulate many different genes, directly or indirectly, the specific mechanism by which Ezh2 interacts with RA signaling is unclear (Balmer and Blomhoff 2002).

In the E8.5-CMEzh2 mutants, we did not find evidence of enhanced RA signaling

(Fig S7). In addition, exogenous all-trans RA (at-RA) lead to wider fontanelles in the wild- type and Ezh2 heterozygous embryos, but it did not phenocopy E8.5-CMEzh2 mutants.

Our small-molecule based approach suggests that EZH2 may not function in a linear pathway to modulate RA signaling in the CM, but indirectly by inhibiting a downstream target of RA. The exact target inhibited by EZH2 is unclear. However, our data has identified two potential targets by which Ezh2 inhibits RA signaling, specifically, HoxC8 and Hedgehog signaling.

3.3c: Ectopic expression of Hox genes may lead to a reduction of skull bone in E8.5-

CMEzh2 mutants 108

The Hox genes are a well-established target of H3K27me3 in multiple cell types in vitro and in the CM in vivo (Schwarz et al. 2014; Dudakovic et al. 2015; Ferguson et al. 2017; Minoux et al. 2017). In addition, the Hox genes are well known targets of Retinoic acid signaling

(http://www.uniprot.org/uniprot/Q15910) (Krumlauf 1994; Balmer and Blomhoff 2002). In our

E8.5-CMEzh2 mutants, we identified an upregulation of multiple the Hox genes. It is worth noting that the CM is normally negative for Hox gene expression, thus any Hox gene expression can be considered ectopic. We hypothesize that the ectopic expression of a specific Hox gene or multiple Hox genes may lead to the truncation of the CNCC bones and loss of PM-derived bones.

Of the Hox genes queried, our data revealed that HoxC8 had the most significant increase in expression in the CM experiencing a 350-fold increase (Fig. S6). In vivo, ectopic expression of HoxC8 in the head and face leads to a near complete loss of skull bones and a truncation of the face resembling our E8.5-CMEzh2 mutants placing HoxC8 as a promising candidate as the mediator between Ezh2 and RA signaling (Carroll and

Capecchi 2015). Along with HoxC8, we also saw an upregulation of multiple other Hox genes that could negatively affect skull bone development in our E8.5-CMEzh2 mutants.

For example, HoxA1 can lead to craniofacial defects and is responsive to RA (Fig. S6)

(Boylan et al. 1993, 1995; Balmer and Blomhoff 2002). In zebrafish, HoxA1 overexpression or the administration of RA leads to an identity switch of the first branchial arch (Hill et al. 1995; Alexandre et al. 1996).

It is currently unclear as to which specific Hox gene, or group of Hox genes, acts as a bone inhibitor in our E8.5-CMEzh2 mutants. We did not observe changes in HoxC8 or

HoxA1 expression following treatment with the 3.5µg/g dose of BMS-453. However, the changes in protein levels are unknown. Examination of the spatial and temporal 109 expression of HOXC8 or HOXA1 in the CM is required to determine if BMS-453 inhibits

HOXC8 leading a restoration of the parietal bone. It is worth noting, the Hox gene often associated with bone inhibition, HoxA2, was not significantly upregulated in our E8.5-

CMEzh2 mutants (Fig. S6) (Grammatopoulos et al. 2000; Trainor and Krumlauf 2001).

3.3d: An upregulation of Hedgehog signaling could lead to a reduction of the skull bones E8.5-CMEzh2 mutants

In multiple organisms, Hedgehog signaling has been shown to be a negative regulator of bone formation. In mice, similar to our E8.5-CMEzh2 mutants, an upregulation of Shh signaling in the Suppressor of Fused (Sufu) mutant mouse (Sufu-/-) leads to arrest in early specification and absence of Runx2 and Osx expression in a spatial and temporal manner (Li et al. 2017). In the chick embryo, overexpression of

Indian Hedgehog (Ihh) signaling and its downstream target, Pthrp negatively regulates bone differentiation (Abzhanov et al. 2007). Ihh-/- mouse embryos have diminished expression of osteoblast progenitor and early osteoblast markers such as Runx2 and Osx and later stage markers such as Osteopontin (Opn) were increased, suggesting IHH may function in blocking differentiation (Abzhanov et al. 2007). In addition, parathyroid- hormone related protein (PTHrP) is activated by RA through a retinoic acid response element (RARE).

In the CM, hedgehog signaling is a potential target of Ezh2. Genes such as Ihh and Pthrp are heavily enriched for H3K27me3 (Ferguson et al. 2017). We also see a dramatic increase in the Hedgehog signaling regulator, Hand2, in our E8.5-CMEzh2 110 mutants (Table S1). In addition, we do see an increasing, but not statistically significant, trend in Pthrp expression levels. We hypothesize that Ezh2 is required to inhibit

Hedgehog signaling in the CM to ensure skull bone formation. Further studies examining spatial and temporal changes in the CM in Ezh2 mutants is required.

3.3e: Conclusion

We show here that Ezh2 is required transiently between E8.5-9.5 to ensure a bone fate commitment through the expression of OSX in the parietal bone. Our data supports previous findings in craniofacial development indicating that Ezh2 is required in craniofacial bone formation in a stage specific manner. Considering OSX is primarily required for intramembranous ossification, our results highlight a specific role of Ezh2 in intramembranous ossification (Nakashima et al. 2002). Finally, our data demonstrates a window during craniofacial development at which RA signaling can indirectly inhibit skull bone formation. Further studies examining the role of ectopic expression of various Hox genes and Hedgehog signaling in the CNCC and PM could provide insights into the intermediate target between Ezh2 and RA signaling and mechanism by which Ezh2 promotes skull bone commitment. These results provide insights into the dynamic formation of skull bone between the CNCC and PM highlighting the spatial and temporal differences in osteogenic potential during development.

3.4: Materials and methods:

3.4a: Mice and genotyping 111

PdgfrɑCreER (JAX stock #018280)(Rivers et al. 2008), Rosa26 Reporter (AX stock

#003309) (Soriano,1999), and Ezh2 floxed (Ezh2fl) (JAX stock #022616)(Shen et al. 2008).

For timed matings, PdgfrɑCreER;Ezh2fl/fl males were crossed with R26R/R26R; Ezh2fl/fl females overnight. Mice were checked for vaginal plugs and then separated in the morning. Vaginal plug day was assigned as embryonic (E) 0.5.CreER recombination was induced by oral gavage at 25ug Tamoxifen/g body weight (BW) (Sigma T5648) to pregnant dams. Tamoxifen was dissolved in corn oil and administered at 5p.m. of the designated day. For each experiment, a minimum of five mutants with litter-matched

Cre negative controls from two to three litters were studied unless otherwise noted.

Case Western Reserve Institutional Animal Care and Use Committee approved all animal procedures in accordance with AVMA guidelines (Protocol 2013-0156, approved 21

November 2014, Animal Welfare Assurance No. A3145-01).

3.4b: Histology, β-Galactosidase, and Immunohistochemistry

Heads of E10.5-13.5 embryos were drop-fixed in 4% paraformaldehyde (PFA) for

20-30 min, respectively at 40C and cryopreserved as previously described and sectioned at 8-10 microns in the coronal plane in the frontal and parietal bone primordia (Atit et al., 2006). β-galactosidase staining on cryosections was performed as previously described (Rivera-Perez et al. 1999).

For immunofluorescence on cryo sections, sections were dried at room temperature, washed in 1x PBS and blocked in goat serum or donkey serum. For mouse raised antibodies, block buffer from the Vector M.O.M Kit (BMK-2202) was used. 112

Primary antibodies were incubated overnight at 4°C, washed next day in 1x PBS, incubated with species- specific secondary antibody (below) for one hour at room temperature, washed with DAPI 0.5µg/mL, and mounted with Fluoroshield (Sigma

F6057). The following primary antibodies were used for immunofluorescence: rabbit anti-H3K27me3 (1:1000; Cell Signaling 9733) , Rabbit anti-Caspase3 (1:250; Abcam ab13847), Rabbit anti-OSX (1:1000 Abcam ab94744), Anti-MSX1/2 (DSHB 4g1), rabbit anti-RUNX2 (1:250 Cell Signaling 8486S), rabbit anti-Sox9 (1:1000; Millipore ab5535).

Appropriate species-specific Alexafluor 594 secondary antibodies were used (1:500;

Invitrogen). Images were captured using the Olympus BX60 microscope and Olympus

DP70 digital camera using DC controller software. Confocal images were captured on the Leica TCS SP8 (Leica Biosystems) using Application Suite X software (Leica

Biosystems). Images were processed in Adobe Photoshop and Fiji (Schindelin et al. 2012;

Schneider et al. 2012).

For alkaline phosphatase staining, sections were then washed in PBST (0.1%

Tween-20 (Bioworld 42030016-1) in PBS) for 10 minutes, then TBST (0.1% Tween in TBS) for 10 minutes, then washed in NTMT (100mM Tris, pH 9.4, 100mM NaCl, 60mM NgCl2, and 0.02% Tween-20) for 10 minutes. Embryos were stained 20ul/mL NBT/BCIP (Roche

11681451001) in the dark O/N at RT. Slides were then washed in PBS and mounted with aqueous mounting medium.

3.4c: RT-qPCR 113

At E13.5, the supraorbital cranial mesenchyme was isolated by manual dissection. Following manual removal of the ectoderm, an incision was made around the circumference of the neurocranium, and the tissue covering the brain was manually disassociated. The CM isolated consists of the neural crest and mesoderm derived cranial mesenchyme. RNA was isolated as previously described (Hamburg-Shields et al.,

2015). Relative mRNA expression was quantified using 5 ng of cDNA on a StepOne Plus

Real-Time PCR System (Life Technologies) and the ΔΔCT method. Commercially available

TaqMan probes (Life Technologies) specific to each gene were used (Table S1). CT values were normalized to β-actin (Invitrogen 4352663). ΔΔCT values were obtained by normalizing the ΔCT values to the average ΔCT values of the controls. Relative mRNA fold change was determined using the ΔΔCT values.

3.4d: Protein Isolation and immunoblotting

E13.5 cranial mesenchyme was collected by manual dissection. Protein was isolated using RIPA buffer. Protein species were was separated by SDS-PAGE using Mini-

PROTEAN TGC gels (BioRad #456-1084). Western Blots were performed with the following primary antibodies: rabbit anti-H3K27me3 (1:1000, Cell Signaling 9733), rabbit anti-EZH2 (1:500, Cell Signaling #5246). Species-specific HRP-conjugated secondary antibodies were used at 1:10,000. Immunoblots were probed with anti-β tubulin (1:400,

Santa Cruz 9104) as a loading control. Protein was detected using an Amersham ECL

Western Blotting Analysis System (GE Healthcare RPN2109), and imaged using an 114

Odyssey FC Imaging System (Li-Cor). Relative protein levels were quantified using ImageJ

(Schneider et al., 2012).

3.4e: MicroCT:

E17.5 heads were fixed and stored in 95% ethanol for at least 24 hours. Heads were then re-hydrated overnight in PBS for 24 hours prior to imaging. MicroCT images were acquired using……. at a 20.2 voxel (20.13µm) resolution and a power of 50kV,

100µA, 5W. MicroCT 3D images were visualized using CTvox (BrukerMicroCT).

Quantification was performed using Amira 6.01 (Thermo Fisher Scientific) using the

“Materials Statistics” tool.

3.4f: Whole mount skeletal preparation:

All steps were performed at room temperature. Embryos were fixed in 95 %

EtOH overnight. Samples were then placed in acetone overnight. Embryos were then placed in alcian blue (Sigma A5268) dissolved in 80 % EtOH, 20 % (glacial) acetic acid at a concentration of 0.03 % overnight. The embryos were then de-stained by two thirty minute washes in 70 % EtOH and then incubating them in 95 % EtOH overnight. The embryos were pre-cleared in 1 % potassium hydroxide (KOH) (Fisher Scientific 1310-58-

3) solution for 1 h at room temperature. Embryos were then place in a 0.005 % alizarin red (Sigma A5533) dissolved in 1% KOH overnight. The embryos were then placed in a

50 % glycerol (Fisher 56-81-5): 50 % (1 %) KOH solution until clear. The average time was one week to become fully cleared. Once cleared, the embryos were then placed in 100% 115 glycerol for long-term storage and imaging. Skeletal preparations were imaged Leica

MZ16F stereoscope and Leica MC120 HD camera with Lecia software.

3.4g: Cell Proliferation/Death Assay

Mice were administered 250µg EdU in PBS/10g mouse weight by intraperitoneal injection one hour prior to sacrifice. Embryos were then collected and prepared for cryopreservation as stated above. EdU was detected using Click-iTEdU Alexa Fluor 488

Imaging Kit (Invitrogen #C10337) according to manufacturer's protocol. Images were captured using the Olympus BX60 microscope and Olympus DP70 digital camera using

DC controller software. The percent of EdU positive cells was quantified using ImageJ.

Cell death was detected using activated Caspase-3 (stated previously) by immunofluorescence. For quantification, images were converted to 8-bit and background was removed with a 10 pixel “rolling ball”. A signal threshold was set.

Individual cells were determined using “watershed” and cells were then counted using

“count particles”.

3.4h: Retinoic acid inhibition

BMS-453 (Caymen Chemical #19076) was administered by oral gavage at E8.5,

E9.5, E11.5, E13.5, and E15.5. BMS-453 was reconstituted in DMSO at 10µg/µl, diluted in corn oil, and administered at 3.5µg/g mouse. For oral gavage on day of tamoxifen administration, the corn oil used to dilute also contained tamoxifen.

116

3.4i: Statistics

Graphs and statistical analysis was generated using Prism 6 (GraphPad Software).

Data are presented as mean ± SEM in all graphs unless otherwise stated. All pairwise sample comparisons were performed using a Mann-Whitney test. The p-values for statistical tests in all figures are represented as: * = P < 0.05, ** = P < 0.01, and *** = P <

0.001.

3.5: Acknowledgements:

I would like to thank previous and current members of the Atit laboratory for excellent discussion and advice. Thank you to the undergraduate students, Samuel Pan,

Samhitha Cinthala for their assistance with sectioning tissues and Gregg DiNuoscio for genotyping. We thank Dr. Brian Hausman in the Orthopedics department for microCT imaging. Thank you to the CWRU SOM Light Microscopy Core Facility. This work was supported by the following grants: National Institutes of Health (NIH) T32 AR-007505

(J.F.), NIH National Institute of Dental and Craniofacial Research R01 DE-01870 (R.P.A.), the Case Western Reserve University ENGAGE (M.D.) and SOURCE Programs (M.D.), and

NIH Grant S10-OD016164 (CWRU SOM Light Microscopy Core Facility).

3.6 Conflict of interest:

We declare no conflicts of interest

117

Discussion outline:

4.1: β-catenin and its role in the suppression of chondrogenesis

4.1a: Possible repression of chondrogenesis mediated by TWIST1

4.1b: Possible suppression of Sox9 through unliganded RARs

4.1c: Possible inhibition of chondrogensis through β-catenin-SOX9 protein-

protein interactions

4.1d: Possible repression of Sox9 by direct genomic binding of β-catenin

4.2: Additional insights into the repressive functions of β-catenin:

4.3: The mechanisms governing the developmental stage and cell type specific role of

Ezh2:

4.3a: Ezh2 is required for lineage selection of the cranial bone prior to the

expression of Msx genes

4.3b: Canonical vs. non-canonical function of EZH2

4.4: Ezh2 regulates skull bone formation by inhibiting a target of the RA signaling pathway:

4.4a: Hedgehog signaling and Hox genes are potential targets of RA signaling

4.5: The correlation of H3K27me3 enrichment with mRNA expression:

4.6: Conclusion: 118

Chapter 4: Discussion:

The skull is composed of multiple bones with complex shapes that must fit together to form the head and face. During development, coordination of complex cellular and molecular processes, in different cell populations, is required to ensure the formation of these structures. Throughout my graduate work, we have investigated the genetic and epigenetic mechanisms governing the formation of the skull bones in a spatial and temporal manner. Using CNCC-CM and PM-CM conditional mouse mutants in vivo, we have 1) shown a physical interaction between two major signaling pathways;

Wnt/β-catenin and PRC2, 2) investigated the mechanism governing the inhibition of chondrogenesis by β-catenin, 3) demonstrated a developmental-stage specific requirement for Ezh2 between the CNCC-CM and PM-CM, and 4) examined the association of H3K27me3 enrichment levels with transcriptional status.

These results have provided insights into the in vivo dynamics of skull bone formation. They demonstrate the heterogeneity of the cranial mesenchyme and the developmental differences between the CNCC-CM and PM-CM. In addition, they highlight the complex regulatory role of Ezh2 in vivo during cell fate selection and bone differentiation. Future studies examining the molecular mechanism by which β-catenin suppresses Sox9, and the specific genes regulated by Ezh2 during skull bone commitment will help piece together the complex signaling pathways that must be coordinated to ensure skull bone formation. 119

4.1: β-catenin and its role in the suppression of chondrogenesis:

Wnt/β-catenin signaling is well known for its role in the inhibition of chondrogenesis in both intramembranous and endochondral bone formation (Day et al.

2005; Goodnough et al. 2012; Fan et al. 2016). During intramembranous bone formation, loss of β-catenin in the CNCC-CM and PM-CM leads to a disruption in bone commitment and a complete conversion of bone into cartilage. Sox9 is a key determinant of chondrogenesis and is upregulated within 36 hours following loss of β- catenin (Bi et al. 1999; Goodnough et al. 2012). Our data demonstrates that, despite the multiple genetic connections between β-catenin and PRC2 in vitro, PRC2 is dispensable for β-catenin-mediated repression of chondrogenesis. The exact mechanism, or mechanisms, by which β-catenin suppresses Sox9, and thus chondrogenesis, is still under investigation. The following are four proposed mechanisms by which β-catenin 120 can suppress chondrogenesis (Fig. 4.1).

Figure 4.1: Proposed mechanisms by which β-catenin suppresses chondrogenesis in the cranial mesenchyme. (A) TWIST1, which is positively regulated by β-catenin, homodimerizes or heterodimerizes with an unknown factor to suppress Sox9. Loss of β-catenin results in a loss of TWIST1 and the ectopic expression of Sox9. (B) β-catenin can bind to an unliganded RAR preventing the activation of Sox9. Loss of β-catenin leads to free unliganded RAR and the activation of Sox9. (C) β-catenin physically interacts with SOX9 protein preventing its autoregulation. Loss of β-catenin leads to free SOX9 enabling binding to the Sox9 enhancer and activating expression. (D) A methylated form of β-catenin may bind to and inhibit the Sox9 gene.

4.1a: Possible repression of chondrogenesis mediated by TWIST1

TWIST1 is a potential regulator of Sox9 expression in the skull. TWIST1 is a transcription factor that is positively regulated by Wnt/β-catenin signaling, can inhibit chondrogenesis, and has been shown to bind to the 3'UTR of Sox9 in vitro (Reinhold et al. 2006; Goodnough et al. 2012). Based on its dimerization status, TWIST1 can act as a 121 transcriptional activator or repressor. In addition to forming a homodimer, Twist1 can heterodimerize with HAND2 and RUNX2 to regulate bone formation and craniofacial development (Bialek et al. 2004; Firulli et al. 2005; Connerney et al. 2006; Bildsoe et al.

2016). These data place TWIST1 as a promising candidate as a mediator by which β- catenin represses chondrogenesis. Loss of Twist1 in the CNCC-CM and PM-CM at E10.5 using En1Cre leads to ectopic Sox9 expression in the PM-CM, but not the CNCC-CM.

These results may be due to the different timepoints at which the CNCC-CM and the

PM-CM progress through the bone initiation program. Highlighting the timepoint- specific role of Twist1, loss of Twist1 at E8.5 using Wnt1Cre or at E9.5 using Dermo1Cre

(Twist2Cre) leads to almost complete agenesis of the skull bones, rather than ectopic chondrogenesis (Goodnough et al. 2012, 2016). These results also correspond with the variation in β-catenin mutant phenotypes. Loss of β-catenin using Wnt1Cre leads to almost complete agenesis of the head and loss using Dermo1Cre leads to ectopic chondrogenesis (Brault et al. 2001; Tran et al. 2010).

Future Directions: In order to address the relationship between β-catenin and

Twist1 in the repression of chondrogenesis, I propose using an inducible Cre driver that is active in both the CNCC-CM and PM-CM, such as PdgfrαCreER, to knockout β-catenin or Twist1. An inducible Cre driver will provide the ability to compare the temporal requirements and gain insights into the spatial requirements for β-catenin and Twist1 in the repression of chondrogenesis. The CNCC-CM progresses through the bone initiation program before the PM-CM (Han et al. 2007). The CNCC-CM begins the bone initiation program at a similar timepoint that Dermo1Cre becomes active. The PM-CM begins the 122 bone initiation program around the time En1Cre is active. In β-catenin mutants, the ectopic expression of Sox9 occurs prior to Osx expression (Goodnough et al. 2012).

Based on these data, I would expect induction of Cre near the Dermo1Cre timepoint

(E9.5-E10.5) would lead to ectopic chondrogenesis in the CNCC-CM, and a loss near the

En1Cre timepoint (E10.5-E11.5) would lead to cartilage in the PM-CM in PdgfrαCreER;β- cateninfl/Δ and PdgfrαCreER;Twist1fl/fl mutants.

4.1b: Possible suppression of Sox9 through unliganded RARs

Wnt/β-catenin signaling is known to interact with multiple other signaling pathways that can regulate Sox9 expression including RA signaling (Yasuhara et al. 2010;

Goodnough et al. 2016; Uchibe et al. 2017). In E11.5 primary limb mesenchyme cultures, unliganded RARγ, which is the highest expressed RA receptor in the CM (Fig. S3.7), can stimulate Sox9 expression. In addition, unliganded RARγ can also bind to β-catenin and inhibit Wnt/β-catenin signaling (Weston et al. 2002; Yasuhara et al. 2010). Based on these data, it is possible that loss of β-catenin leads to an increase in available RARγ which then stimulates Sox9. It has been shown that only a 20% increase in Sox9 expression is sufficient to induce phenotypes in chondrocytes (Akiyama et al. 2004).

Thus, only a small increase in free RARγ could potentially be sufficient to activate Sox9.

Because RARγ must be unliganded to stimulate Sox9, it may explain as to why administration of RA to wild-type mice does not lead to ectopic chondrogenesis in the skull (Fig. S3.7). 123

Future directions: If unliganded RARγ can stimulate Sox9 expression, knocking out both RARγ and β-catenin using En1Cre should rescue the cartilage phenotype. In addition, overexpression of RARγ, ideally a form that can no longer bind to its ligand, using En1Cre should lead to ectopic chondrogenesis and phenocopy the β-catenin mutants.

4.1c: Possible inhibition of chondrogenesis through β-catenin-SOX9 protein-protein interactions

Beyond transcriptional regulation, inhibition of chondrogenesis by β-catenin could be at the protein level. In chondrocytes, β-catenin and SOX9 can genetically and physically interact, negatively regulating the function of one another (Akiyama et al.

2004). In addition, SOX9 can bind to its own enhancer sequences and positively regulate itself (Mead et al. 2013; Liu and Lefebvre 2015; Yao et al. 2015). It is possible that, in normal conditions, β-catenin interacts with SOX9 protein preventing it from activating its enhancers. In β-catenin mutants, SOX9 is no longer sequestered by β-catenin, and the increases in Sox9 expression are due to increased autoregulation. As mentioned previously, only a small change in Sox9 expression is required to induce a cartilage phenotype (Akiyama et al. 2004). In addition, Sox9Cre;Osxfl/lacZ leads to a loss or reduction of the skull bones indicating the intramembranous bones are derived from

Sox9 positive precursors experiencing some level of Sox9 expression (Akiyama et al. 124

2005). It is possible, in the absence of β-catenin, the autoregulation from base level expression of Sox9 is sufficient to induce ectopic chondrogenesis.

Future directions: Flouorscence resonance energy transfer (FRET) is a technique in which two proteins that are in close proximity can be detected in vivo using confocal microscopy. Using FRET, I would expect to find a physical interaction between β-catenin and SOX9 in the CM. Alternatively, through co-immunoprecipitation (co-IP) experiments, physical interaction between β-catenin and SOX9 can be detected. In order to determine the functional role between β-catenin and SOX9 interaction, mutant proteins lacking the specific binding sites are required. β-catenin and SOX9 interact at the armadillo repeat of β-catenin and the C-terminal transactivation domain of SOX9

(Akiyama et al. 2004). I would expect generating conditional mutants lacking the C- terminal transactivation domain of SOX9 using En1Cre would lead to ectopic chondrogenesis in the presence of β-catenin.

4.1d: Possible repression of Sox9 by direct genomic binding of β-catenin

In vitro, EZH2 has been shown to bind to β-catenin and trimethylate the β- catenin protein itself. The trimethylated β-catenin can then act as a transcriptional repressor (Hoffmeyer et al. 2017). In the cranial mesenchyme, we have shown that EZH2 can physically interact with β-catenin by co-IP (Fig. 2.1). It is possible that a methylated form of β-catenin can bind to the Sox9 promoter or an enhancer to repress chondrogenesis. 125

Future Directions: To determine if β-catenin can directly repress Sox9, I would propose ChIP-sequencing to determine the genome binding profile of β-catenin in the cranial mesenchyme. I would expect to find β-catenin bound to the enhancer or promoter of Sox9. In relation to this point, our data has also shown that loss of Ezh2 in the cranial mesenchyme does not lead to ectopic chondrogenesis. However, this does not rule out direct repression of Sox9 by β-catenin. In specific tissues, the loss of both

Ezh2 and Ezh1 are required to induce a phenotype (Ezhkova et al. 2011). In the cranial mesenchyme, it is possible that EZH1 compensates for the loss of EZH2 in methylating β- catenin. Examining the β-catenin binding by ChIP-sequencing in an Ezh1/Ezh2 double mutant background would help determine if methylation of β-catenin leads a repression of Sox9. Alternatively, an unmethylated form of β-catenin could bind to Sox9 and act as a competitive antagonist to Sox9 activation.

4.2: Additional insights into the repressive functions of β-catenin:

Wnt/β-catenin signaling is typically thought of as a transcriptional activator and regulator of cell fate selection. The canonical pathway involves nuclear β-catenin interacting with TCF/LEF to bind to specific DNA sites to activate transcription (Clevers and Nusse 2012). In multiple organisms and cell types, however, Wnt/β-catenin signaling is not always associated with transcriptional activation. In Xenopus embryos, β- catenin and TCF/LEF binding to the genome alone is not sufficient to activate transcription. After the β-catenin-TCF/LEF complex binds to the DNA, additional tissue- 126 specific signaling factors, such as BMP or FGF, are then required to induce transcription

(Nakamura et al. 2016). In Drosophila, Arm (the Drosophila form of β-catenin) and TCF can bind to novel sites in the genome to repress transcriptional activity (Blauwkamp et al. 2008). Furthermore, in multiple cell types in vitro, β-catenin can bind to DNA in regions relatively devoid of TCF4 that are not associated with transcriptional activation.

A dominant negative form of TCF4, however, abrogates the β-catenin binding indicating that it is still dependent on TCF4 (Schuijers et al. 2014). Similar to Drosophila, the low

TCF4 binding sites could represent novel binding sites at which β-catenin represses transcription.

Another potential avenue of repression could be through epigenetics such as

DNA methylation and PRC2. In colon cancer cells, β-catenin can physically bind to DNA methyltransferase 1 (DNMT1), which maintains DNA methylation (Song et al. 2015).

Furthermore, in chick limb bud cultures, activation of Wnt signaling leads to increased

DNA methylation over the Sox9 promoter (Kumar and Lassar 2014). In the context of

PRC2, in both liver cancer cells and mouse ESCs, EZH2 can bind and methylate the β- catenin protein. The methylated form of β-catenin can then inhibit cell fate selection and maintain stem cell renewal (Zhu et al. 2016; Hoffmeyer et al. 2017). In support of these data, our RNA-sequencing data identified numerous genes upregulated following loss of β-catenin indicating a repressive function of β-catenin. In addition, as stated above, we have also shown that β-catenin can physically bind to PRC2 in wild-type, non- cultured cranial mesenchyme (Fig. 2.1). Our data helps support the growing literature that β-catenin may function in a transcriptional repressor. However, the exact 127 mechanism by which β-catenin represses transcription and the biological functions of the β-catenin/EZH2 complex during development are unclear. Unlike Drosophila, very few Polycomb Response Elements (PRE) have been identified (Bauer et al. 2015). β- catenin could be required to recruit PRC2 to specific genes without modifying the

H3K27me3 status (Mirzamohammadi et al. 2016). It is also possible that the β- catenin/EZH2 complex can repress transcription independent of PRC2. I would be remiss to mention, however, that β-catenin could positively regulate a transcriptional inhibitor.

Thus, the genes upregulated in β-catenin mutants are simply due to a loss of their upstream inhibitor. In addition, it is possible that a combination of multiple mechanism leads to the repression of chondrogenesis by β-catenin.

Future directions: As technology advances, the ability to gather and analyze genome-wide data is increasing. Correlating and comparing RNA-sequencing, ChIP- sequencing, and ATAQ-sequencing studies on in vivo cranial mesenchyme will provide insights into the target genes and the mechanisms by which β-catenin represses transcription. For example, ATAQ-sequencing in β-catenin single or β-catenin/EZH2 (or

β-catenin and another epigenetic regulator such as DNMT3) double mutant backgrounds will associate β-catenin, in both EZH2-dependent and independent manners, with changes in open chromatin. ATAQ-sequencing is designed for mapping open and closed chromatin using a small number of cells making it ideal for in vivo tissues during development. Alternatively, ChIP-sequencing studies on β-catenin (using antibodies specific to wild-type and methylated β-catenin) and EZH2 will identify genomic locations bound by both β-catenin and EZH2. In addition, ChIP-sequencing for 128

TCF4 will provide even further insight into the function of the β-catenin/EZH2 complex by associating it with TCF4 binding sites. Furthermore, cross-referencing the ATAQ- sequencing and ChIP-sequencing with RNA-sequencing from β-catenin mutants will provide insights into the transcriptional changes associated with the β-catenin/EZH2 complex.

4.3: The mechanisms governing the developmental stage and cell type specific role of

Ezh2:

In vivo, the exact function of Ezh2 is cell-type specific (Ezhkova et al. 2011; Juan et al. 2011; Bardot et al. 2013; Lui et al. 2016). In skull bone formation, the role of Ezh2 is dependent on the developmental stage as well as on the specific cell type. Loss of

Ezh2 at E8.5 with Wnt1Cre (Wnt1Cre;Ezh2 fl/fl) results in a complete loss of CNCC-derived bone (Schwarz et al. 2014). In contrast, loss of Ezh2 at E9.5 in the PM-CM using Prx1Cre

(Prx1Cre;Ezh2 fl/fl ) does not result in a loss of bone and exhibits a potential increase in bone (Dudakovic et al. 2015). Our temporal Ezh2 knockout mutants using in the CNCC-

CM and PM-CM PdgfrαCreER (E8.5-CMEzh2) demonstrates that the phenotypic variations observed in various Ezh2 mutants is most likely due developmental stage-specific roles of Ezh2 (Fig. 4.2, Fig. 3.3). 129

Figure 4.2: Ezh2 is required for the lineage selection of the skull bone precursors prior to the activation of the bone initiation program. (A) Loss of Ezh2 at E8.5 leads to a near complete loss of the facial bones and PM-CM derived bones and a failure to express OSX. (B) Loss of Ezh2 at E9.5 leads to only a reduction of the facial bones and the expression of OSX. Note that Msx genes are not expressed until E12.5 in the PM-CM. Thus, Ezh2 is required only in a window between E8.5 and E9.5, roughly 3 days prior to the bone initiation program to ensure OSX expression at E13.5. (C) Inhibition of RA signaling with BMS-453 rescues the PM-CM bone phenotype.

Loss of Ezh2 in the early CNCC-CM bone progenitors and the PM-CM bone precursors by oral gavage at E8.5 in our E8.5-CM Ezh2 mutants leads to a near complete loss of PM- derived parietal bone and a misshapen CNCC-derived frontal bone. In contrast, our E9.5-

CM Ezh2 mutants do no exhibit a loss of the calvarial bones. Considering that Ezh2 is required for the H3K27me3 repressive histone modification throughout the entire genome, it is intriguing as to why there is such a stage- and tissue-specific requirement for Ezh2 expression. As mentioned throughout my dissertation, previous studies in various cell types, often in vitro, have associated EZH2 and PRC2 with specific functions

(Margueron and Reinberg 2011). Based on our data, however, careful interpretations of these studies is required. The variations in phenotypes we see in vivo, which appear to be dependent on specific developmental-stages, demonstrates that the tissue source must be heavily weighed when determining the function of Ezh2. 130

4.3a: Ezh2 is required for lineage selection of the cranial bone prior to the expression of Msx genes

The CNCC-CM initiates the bone initiation program roughly two days prior to the

PM-CM (Han et al. 2007). Bone initiation, as marked by the expression of Msx genes, occurs in the CNCC-CM at E10.5 and in the PM-CM between E11.5 and E12.5. Loss of

Ezh2 in E8.5-CMEzh2 mutants leads to defects in the bone commitment in the PM-CM and ultimately the loss of the PM-derived bones. In E8.5-CMEzh2 mutants, the RUNX2- expressing cells in the parietal bone primordia fail to express OSX at E13.5 (Fig. 3.5). In contrast, loss of Ezh2 around E9.5 (E9.5-CMEzh2) results in normal OSX expression and the formation of the PM-derived bones (Fig. S3.5). Based on these data, expression of

Ezh2 in the developmental window between E8.5 and E9.5 is critical for the expression of OSX. These results are striking as the PM-CM does not begin to express Msx2 until

E12.5. Ezh2 appears to be required to establish lineage selection of the cranial bone roughly three to four days prior to the onset of the bone initiation program.

The extent at which Ezh2 is required for lineage restriction of other cell populations and tissues during craniofacial development is unclear. Craniofacial development involves the organization of multiple tissues for the formation of complex structures such as the skull bones, teeth and dermis (Chai et al. 2000; Fan et al. 2016;

Goodnough et al. 2016). It is possible that the role of Ezh2 in lineage fate selection expands beyond the PM-CM depending on the developmental timepoint. For example, 131 based on the development stage at which the CNCC-CM derived bone begins the bone initiation program, the loss of the bones in the Wnt1Cre;Ezh2 fl/fl mutants may be due to an early lineage selection defect during neural crest migration around E8.5. An understanding of the developmental timing and the role of Ezh2 in lineage selection different tissue precursors will help provide a clearer picture of the dynamics of craniofacial development and skull bone formation.

Future directions: Inducible, fluorescent lineage reporters provide the ability to trace all the progeny of a cell in vivo (Kretzschmar and Watt 2012). In order to determine the extent in which Ezh2 is required for lineage restriction at different developmental stages in craniofacial development, I propose the use of the confetti fluorescent reporter with an inducible Cre driver (Livet et al. 2007; Snippert et al. 2010).

With the confetti reporter, cells experiencing Cre recombination are randomly marked with one of four colors, and the progeny of those cells will also be marked with the same color of reporter.

Using an inducible Cre-line that hits all of the cranial mesenchyme, such as

PdgfrαCreER or Cagg-CreER (Roybal et al. 2010), the developmental stages at which the various tissues begin lineage selection can be determined. Furthermore, utilizing the confetti reporter in combination with the Ezh2fl/fl mutants will provide insights into the role of Ezh2 in lineage selection.

4.3b: Canonical vs. non-canonical function of EZH2 132

It is unclear if the defects in bone formation in our E8.5-CMEzh2 mutants are associated with PRC2 or a non-canonical function of EZH2. It has been shown that EZH2 alone can repress transcription independent of its methyltransferase activity and PRC2

(Xu et al. 2012; O’Geen et al. 2017). In addition, our H3K27me3 ChIP-sequencing and

RNA-sequencing in the CM (Chapter 2) has demonstrated little correlation between the

H3K27me3 status and transcriptional activity indicating that Ezh2 may function independent of PRC2.

Future directions: Examine the phenotype of Suz12 or Eed knockout mutants using PdgfrαCreER. If the loss of the parietal bone in E8.5-CMEzh2 mutants is due to a loss of canonical PRC2 function, I would expect a phenocopy in Suz12 or Eed knockout mutants.

4.3c: Developmental stage specific role of Ezh2 could be due to compensation by Ezh1:

In mammals, a homologue of Ezh2 is Ezh1 in which EZH1 also contains methyltransferase activity. However, the specifics of if, when, and where EZH1 can compensate for EZH2 is unclear. In vitro, Ezh2 and Ezh1 have been shown to have reciprocal expression as cells differentiate (Xu et al. 2015). In addition, Ezh1 null mutants are healthy and viable indicating that Ezh2 is primarily required for early development compared to Ezh1 (Ezhkova et al. 2011). The reduction in phenotype observed in our E9.5-CMEzh2 mutants (Fig. 3.3), and the lack of phenotype in the

Dermo1Cre;Ezh2fl/fl and En1Cre;Ezh2fl/fl mutants (Fig. S2.3) may be due to temporal 133 compensation by Ezh1. At E14.5 in the ectoderm, loss of both Ezh2 and Ezh1 are required to induce a phenotype (Ezhkova et al. 2011). It is possible that in between E9.5 and E14.5 in the mouse embryo, Ezh2 expression becomes downregulated and Ezh1 becomes upregulated, or EZH1 is capable of compensating for a loss of EZH2.

Future directions: Further analysis of the changes in expression between Ezh2 and Ezh1 at E8.5, E9.5, and E10.5 could provide insights into the dynamics between Ezh2 and Ezh1 in the CM. In addition, double knockouts of Ezh2 and Ezh1 using PdgfrαCreER would determine the compensatory role of Ezh1. If Ezh1 compensates for Ezh2 after

E9.5, I would expect a skull bone phenotype in the E9.5 induction of PdgfrαCreER;Ezh2fl/fl

;Ezh1fl/fl , Dermo1Cre;Ezh2fl/fl ;Ezh1fl/fl, and En1Cre;Ezh2fl/fl ;Ezh1 fl/fl mutants.

4.4: Ezh2 regulates skull bone formation by inhibiting a target of the RA signaling pathway:

Using the RAR antagonist BMS-453, we observe a partial rescue of the E8.5-

CMEzh2 mutant phenotype (Fig. 3.6). These results indicate that RA-signaling negatively regulates the bone transcriptional program and is inhibited by Ezh2, however, the exact mechanism is unclear. It is possible that Ezh2 is required to inhibit molecules of the RA pathway such as the RARs or RA itself. Alternatively, Ezh2 could suppress a downstream or indirect target of RA. In the E8.5-CMEzh2 mutants, we didn't observe changes in the expression of the RARs (α, β, or γ) or the direct target Crabp2. In addition, administration of all-trans RA in wild-type embryos or Ezh2 heterozygous embryos (E8.5 134 tamoxifen induction) does not lead to a loss of skull bones similar to the E8.5-CMEzh2 mutants (Fig. S3.7). These results indicate that Ezh2 inhibits a specific downstream or indirect target of RA signaling rather than RA signaling itself. The specific target inhibited by Ezh2 is currently unclear. Based on our data, we have identified two potential downstream targets.

4.4a: Hedgehog signaling and Hox genes are potential targets of RA signaling

Using a gene candidate approach in manually dissected cranial mesenchyme from our E8.5-CMEzh2 mutants, we examined the expression of numerous genes known to be targets of PRC2. Of the genes queried, the Hox genes, Cdkn2a, and an upstream regulator of Hedgehog signaling, Hand2, had the largest upregulation of expression levels in Ezh2 mutants. Cdkn2a is a regulator of the cell cycle and cell death (Stott et al.

1998). Based on our data, changes in cell death and proliferation are not sufficient to account for the skull bone defects observed in our E8.5-CMEzh2 mutants. Therefore, dysregulation of Cdkn2a is most likely not the cause of the phenotype.

Both the Hox genes and Hedgehog signaling have been associated with craniofacial and bone development. The Hox genes are body patterning genes during embryonic development and are normally not expressed in the CM. In the 1st branchial arch of Wnt1Cre;Ezh2fl/fl mutants, a homeotic transformation into the 2nd branchial arch was attributed to ectopic Hox gene expression (Schwarz et al. 2014). HoxC8 overexpression mutants also exhibit craniofacial defects including a the complete loss of 135 skull bone and truncated face (Carroll and Capecchi 2015). Out of all the Hox genes queried, HoxC8 exhibits the highest level of ectopic expression in our E8.5-CMEzh2 mutants highlighting it as a strong candidate as a target of EZH2.

Hedgehog signaling is another set of factors that could potentially lead a loss of the parietal bone in our E8.5-CMEzh2 mutants. HAND2 is upstream of Hedgehog- signaling, can bind to RUNX2 protein, and inhibit osteoblast differentiation (Charité et al. 2000; Funato et al. 2009; Osterwalder et al. 2014). Parathyroid hormone-related peptide (Pthrp), a downstream target of Indian Hedgehog (Ihh), is also a target of RA signaling and has a retinoic acid response element (RARE) near its promoter (Karperien et al. 1999; Abzhanov et al. 2007). In addition, Indian Hedgehog (Ihh) and Pthrp are highly enriched for H3K27me3 in the cranial mesenchyme (Chapter 2-observed data from H3K7me3 ChIP-sequencing). Furthermore, in the CNCC-CM and PM-CM, Hedgehog signaling also appears to have a developmental stage-dependent phenotype. Loss of

Suppressor of fused (Sufu), a negative regulator of Shh, at E8.5 using Wnt1Cre leads to a complete loss of the CNCC-derived bones. Loss of Sufu with Dermo1Cre, which hits both the CNCC-CM and PM-CM at E9.5, leads to only a loss of the PM-derived bones with little effect in the CNCC-derived bones bone (Li et al. 2017). Similar to the E8.5-CMEzh2 and E9.5-CMEzh2 mutants (Fig. 3.3), the Sufu knockout mutants display a temporal sensitivity to Sufu signaling during skull bone formation.

The dramatic increase in Hand2 expression and the established role of Hedgehog signaling highlight Hedgehog signaling as a potential target of Ezh2 in regulating lineage selection. Alternatively, HAND2 binding to RUNX2, independent of Hedgehog signaling, 136 could be sufficient to cause the inhibition of Osx expression we see in our E8.5-CMEzh2 mutants. Further studies are required to parse out the roles of Hand2 and Hedgehog signaling.

Future directions: To identify specific Hox genes regulated by Ezh2, ChIP- sequencing or ChIP-qPCR for EZH2 would identify which Hox genes are directly bound by

EZH2. In contrast to H3K27me3, which has broad enrichment over the entire Hox locus, I expect EZH2 to have a smaller peak over specific Hox genes allowing for the identification of the binding to specific Hox genes. After identification of candidate Hox genes, I would expect a double knockout of Ezh2 and the candidate Hox gene using

PdgfrαCreER at E8.5 would rescue the E8.5-CMEzh2 mutant phenotype.

However, it is also possible that a disruption of multiple Hox genes leads to a loss of the parietal bone. Unfortunately, genetic mutants for multiple Hox genes would be difficult to obtain. To account for multiple Hox genes, inhibition of other Hox activating mechanisms in E8.5-CMEzh2 mutants could further implicate the Hox genes. For example,

I expect inhibition or mutations in the PI3K/Akt signaling or the Mixed Lineage Leukemia

(MLL/HRX/ALL1) gene in an Ezh2 mutant background, could potentially rescue the phenotype (Hsieh et al. 2003; Costa et al. 2010; Lee et al. 2014).

In order to parse out the role of Hand2 and Hedgehog signaling in lineage selection by Ezh2, I would generate double knockouts with Ezh2 and Pthrp or Ezh2 and

Hand2 using PdgfrαCreER. If an upregulation of Hedgehog signaling leads to a loss of the 137 parietal bone, I would expect a double knockout of Ezh2 and a member of Hedgehog signaling such as Pthrp or Hand2 would rescue the phenotype.

4.5: The correlation of H3K27me3 enrichment with mRNA expression:

It has been shown in multiple cell types that H3K27me3 enrichment is not predictive of the transcriptional status. It has been proposed that H3K27me3 does not induce transcriptional repression, but maintains an already repressed state (Comet et al.

2016). In addition, in human colon cancer cells, targeted disposition of H3K27me3 using the EZH2 methyltransferase enzyme attached to dCas9 is insufficient to induce transcriptional repression (O’Geen et al. 2017). In the manually isolated cranial mesenchyme, we also observed relatively few correlations between the enrichment of all H3K27me3 peaks with gene expression (Chapter 2). However, after dividing the enrichment peaks three categories, strong, medium, and weak, we did observe some level of correlation (Fig. S2.4). We found the strong peaks correlated to transcriptional repression more than the medium and weak peaks.

These results indicate that H3K27me3 enrichment is not binary, and the level of enrichment may be predictive of its function. It is possible that the peak sizes correspond to specific states of transcription such as repressed or poised. It has been shown in the developing head, H3K27me3 can co-occupy gene loci with H3K4me2 indicating poising (Minoux et al. 2017). Comparing the sizes of peaks to the extent at which they are repressed, poised, and expressed could provide insights into how to 138 better predict a histone modification's function. Alternatively, considering the dynamic activity of Ezh2, it is possible the medium and weak peaks are just a reflection of active methylation and demethylation and could explain the hypothesis that H3K27me3 is only indicative of past expression (Riising et al. 2014). Correlating the medium and weak peaks to the past transcriptional state could provide insights into the dynamics of histone modifications. In addition, ChIP-sequencing on in vivo tissue at sequential development times could provide a greater picture of the epigenetic landscape during development of a tissue.

4.6: Conclusion:

During my graduate work, we examined, in vivo, the genetic mechanisms governing the formation of the CNCC- and PM-derived skull bones. We have investigated the genetic and temporal differences governing lineage selection and the formation of the skull bones. Our mutant mouse models have demonstrated the dynamic factors required during development, and highlighted the importance of in vivo studies. During craniofacial development, many transcription factors and signaling pathways have been shown to be involved in skull bone formation (Fan et al. 2016).

However, primary focus has been on the CNCC-CM. A full picture of how all these different genetic pieces fit together across the CNCC-CM and PM-CM in vivo is currently lacking. A greater understanding of the genetic signaling and osteogenic potential 139 between the CNCC-CM and PM-CM will identify candidates for tissue engineering and provide insights into birth defects.

140

Bibliography:

Abe M., Maeda T., Wakisaka S., 2008 Retinoic acid affects craniofacial patterning by

changing Fgf8 expression in the pharyngeal ectoderm. Dev. Growth Differ. 50: 717–

729.

Abu-Issa R., Smyth G., Smoak I., Yamamura K., Meyers E. N., 2002 Fgf8 is required for

pharyngeal arch and cardiovascular development in the mouse. Development 129:

4613–4625.

Abzhanov A., Rodda S. J., McMahon A. P., Tabin C. J., 2007 Regulation of skeletogenic

differentiation in cranial dermal bone. Development 134: 3133–3144.

Akiyama H., Kim J.-E., Nakashima K., Balmes G., Iwai N., et al., 2005 Osteo-

chondroprogenitor cells are derived from Sox9 expressing precursors. Proc. Natl.

Acad. Sci. 102: 14665–14670.

Akiyama H., Lyons J. P., Mori-Akiyama Y., Yang X., Zhang R., et al., 2004 Interactions

between Sox9 and β-catenin control chondrocyte differentiation. Genes Dev. 18:

1072–1087.

Alexandre D., Clarke J. D. W., Oxtoby E., Yan Y.-L., Jowett T., et al., 1996 Ectopic

expression of Hoxa-1 in the zebrafish alters the fate of the mandibular arch neural

crest and phenocopies a retinoic acid-induced phenotype. Development 122: 735–

746.

Asp P., Blum R., Vethantham V., Parisi F., Micsinai M., et al., 2011 Genome-wide 141

remodeling of the epigenetic landscape during myogenic differentiation. Proc. Natl.

Acad. Sci. U. S. A. 108: E149-58.

Atit R., Sgaier S. K., Mohamed O. A., Taketo M. M., Dufort D., et al., 2006 β-Catenin

Activation Is Necessary and Sufficient To Specify the Dorsal Dermal Fate in the

Mouse. Dev. Biol. 296: 164–176.

Baek W. Y., Kim Y. J., Crombrugghe B. de, Kim J. E., 2013 Osterix is required for cranial

neural crest-derived craniofacial bone formation. Biochem. Biophys. Res. Commun.

432: 188–192.

Baek J. E., Choi J. Y., Kim J. E., 2014 Skeletal analysis and differential gene expression in

Runx2/Osterix double heterozygous embryos. Biochem. Biophys. Res. Commun.

451: 442–448.

Balmer J. E., Blomhoff R., 2002 Gene expression regulation by retinoic acid. J. Lipid Res.

43: 1773–808.

Bardot E. S., Valdes V. J., Zhang J., Perdigoto C. N., Nicolis S., et al., 2013 Polycomb

subunits Ezh1 and Ezh2 regulate the Merkel cell differentiation program in skin

stem cells. EMBO J. 32: 1990–2000.

Barski A., Cuddapah S., Cui K., Roh T. Y., Schones D. E., et al., 2007 High-Resolution

Profiling of Histone Methylations in the Human Genome. Cell 129: 823–837.

Bauer M., Trupke J., Ringrose L., 2015 The quest for mammalian Polycomb response

elements: are we there yet? Chromosoma. 142

Behr B., Longaker M. T., Quarto N., 2011 Craniosynostosis of coronal suture in Twist1 +/-

mice occurs through endochondral ossification recapitulating the physiological

closure of posterior frontal suture. Front. Physiol. JUL: 1–7.

Bhanot P., Brink M., Samos C. H., Hsieh J.-C., Want A., et al., 1996 A new member of the

frizzled family from Drosophila functions as a Wingless receptor. Nature 382: 225–

230.

Bi W., Deng J. M., Zhang Z., Behringer R. R., Crombrugghe B. De, 1999 Sox9 is required

for cartilage formation. Nat. Genet. 22: 85–89.

Bialek P., Kern B., Yang X., Schrock M., Sosic D., et al., 2004 A Twist Code Determines the

Onset of Osteoblast Differentiation. Dev. Cell 6: 423–435.

Bildsoe H., Loebel D. A. F., Jones V. J., Chen Y. T., Behringer R. R., et al., 2009

Requirement for Twist1 in frontonasal and skull vault development in the mouse

embryo. Dev. Biol. 331: 176–188.

Bildsoe H., Loebel D. A. F., Jones V. J., Hor A. C. C., Braithwaite A. W., et al., 2013 The

mesenchymal architecture of the cranial mesoderm of mouse embryos is disrupted

by the loss of Twist1 function. Dev. Biol. 374: 295–307.

Bildsoe H., Fan X., Wilkie E. E., Ashoti A., Jones V. J., et al., 2016 Transcriptional targets

of TWIST1 in the cranial mesoderm regulate cell-matrix interactions and

mesenchyme maintenance. Dev. Biol.: 1–15.

Blauwkamp T. a, Chang M. V, Cadigan K. M., 2008 Novel TCF-binding sites specify 143

transcriptional repression by Wnt signalling. EMBO J. 27: 1436–1446.

Bonilla-Claudio M., Wang J., Bai Y., Klysik E., Selever J., et al., 2012 Bmp signaling

regulates a dose-dependent transcriptional program to control facial skeletal

development. Development 139: 709–719.

Boyer L. A., Plath K., Zeitlinger J., Brambrink T., Medeiros L. A., et al., 2006 Polycomb

complexes repress developmental regulators in murine embryonic stem cells.

Nature 441: 349–53.

Boylan J. F., Lohnes D., Taneja R., Chambon P., Gudas L. J., 1993 Loss of retinoic acid

receptor gamma function in F9 cells by gene disruption results in aberrant Hoxa-1

expression and differentiation upon retinoic acid treatment. Proc. Natl. Acad. Sci.

U. S. A. 90: 9601–5.

Boylan J. F., Lufkin T., Achkar C. C., Taneja R., Chambon P., et al., 1995 Targeted

disruption of retinoic acid receptor alpha (RAR alpha) and RAR gamma results in

receptor-specific alterations in retinoic acid-mediated differentiation and retinoic

acid metabolism. Mol. Cell. Biol. 15: 843–51.

Bracken A. P., Dietrich N., Pasini D., Hansen K. H., Helin K., 2006a Genome-wide

mapping of polycomb target genes unravels their roles in cell fate transitions.

Genes Dev. 20: 1123–1136.

Bracken A. P., Dietrich N., Pasini D., Hansen K. H., Helin K., 2006b roles in cell fate

transitions Genome-wide mapping of Polycomb target genes unravels their roles in 144

cell fate transitions. Genes Dev.: 1123–1136.

Brault V., Moore R., Kutsch S., Ishibashi M., Rowitch D. H., et al., 2001 Inactivation of the

β-catenin gene by Wnt1-Cre-mediated deletion results in dramatic brain

malformation and failure of craniofacial development. Development 128: 1253–

1264.

Brugmann S. A., Goodnough L. H., Gregorieff A., Leucht P., Berge D. ten, et al., 2007 Wnt

signaling mediates regional specification in the vertebrate face. Development 134:

3283–3295.

Budnick I., Hamburg-Shields E., Chen D., Torre E., Jarrell A., et al., 2016 Defining the

identity of mouse embryonic dermal fibroblasts. Genesis 54: 415–430.

Cadigan K. M., Nusse R., 1997 Wnt signalling: a common theme in animal development.

Genes Dev. 11: 3286–3305.

Carpenter A. C., Rao S., Wells J. M., Campbell K., Lang R. A., 2010 Generation of mice

with a conditional null allele for Wntless. Genesis 48: 554–558.

Carroll L. S., Capecchi M. R., 2015 HOXC8 initiates an ectopic mammary program by

regulating Fgf10 and Tbx3 expression, and Wnt/β-catenin signaling. Development:

dev.128298-.

Chai Y., Jiang X., Ito Y., Bringas P., Han J., et al., 2000 Fate of the mammalian cranial

neural crest during tooth and mandibular morphogenesis. Development 127:

1671–1679. 145

Chambont P., 1996 A decade receptors biology of retinoic acid. Faseb 10: 940–954.

Charité J., McFadden D. G., Olson E. N., 2000 The bHLH transcription factor dHAND

controls Sonic hedgehog expression and establishment of the zone of polarizing

activity during limb development. Development 127: 2461–2470.

Chung S. S. W., Wang X., Roberts S. S., Griffey S. M., Reczek P. R., et al., 2011 Oral

administration of a retinoic acid receptor antagonist reversibly inhibits

spermatogenesis in mice. Endocrinology 152: 2492–2502.

Clevers H., Nusse R., 2012 Wnt/β-catenin signaling and disease. Cell 149: 1192–1205.

Comet I., Riising E. M., Leblanc B., Helin K., 2016 Maintaining cell identity: PRC2-

mediated regulation of transcription and cancer. Nature 16: 803–810.

Connerney J., Andreeva V., Leshem Y., Muentener C., Mercado M. A., et al., 2006 Twist1

dimer selection regulates cranial suture patterning and fusion. Dev. Dyn. 235:

1334–1346.

Costa B. M., Smith J. S., Chen Y., Chen J., Phillips H. S., et al., 2010 Reversing HOXA9

oncogene activation by PI3K inhibition: Epigenetic mechanism and prognostic

significance in human glioblastoma. Cancer Res. 70: 453–462.

Creuzet S., Couly G., Vincent C., Douarin N. M. Le, 2002 Negative effect of Hox gene

expression on the development of the neural crest-derived facial skeleton.

Development 129: 4301–4313.

Das B. C., Thapa P., Karki R., Das S., Mahapatra S., et al., 2014 Retinoic acid signaling 146

pathways in development and diseases. Bioorganic Med. Chem. 22: 673–683.

Day T. F., Guo X., Garrett-Beal L., Yang Y., 2005 Wnt/β-catenin signaling in mesenchymal

progenitors controls osteoblast and chondrocyte differentiation during vertebrate

skeletogenesis. Dev. Cell 8: 739–750.

Deckelbaum R. A., Holmes G., Zhao Z., Tong C., Basilico C., et al., 2012 Regulation of

cranial morphogenesis and cell fate at the neural crest-mesoderm boundary by

engrailed 1. Development 139: 1346–1358.

Delmas V., Beermann F., Martinozzi S., Carreira S., Ackermann J., et al., 2007 β-Catenin

induces immortalization of melanocytes by suppressing. Genes Dev. 21: 2923–

2935.

Dorsky R. I., Moon R. T., Raible D. W., 2000 Environmental signals and cell fate

specification in premigratory neural crest. BioEssays 22: 708–16.

Drissi M. H., Li X., Sheu T. J., Zuscik M. J., Schwarz E. M., et al., 2003 Runx2/Cbfa1

Stimulation by Retinoic Acid Is Potentiated by BMP2 Signaling Through Interaction

with Smad1 on the Collagen X Promoter in Chondrocytes. J. Cell. Biochem. 90:

1287–1298.

Dudakovic A., Camilleri E. T., Xu F., Riester S. M., McGee-Lawrence M. E., et al., 2015

Epigenetic control of skeletal development by the histone methyltransferase Ezh2.

J. Biol. Chem. 290: 27604–27617.

Ezhkova E., Lien W. H., Stokes N., Pasolli H. A., Silva J. M., et al., 2011 EZH1 and EZH2 147

cogovern histone H3K27 trimethylation and are essential for hair follicle

homeostasis and wound repair. Genes Dev. 25: 485–498.

Fan X., A F Loebel D., Bildsoe H., E Wilkie E., P L Tam P., et al., 2016 Tissue interactions,

cell signaling and transcriptional control in the cranial mesoderm during

craniofacial development. AIMS Genet. 3: 74–98.

Ferguson J., Devarajan M., DiNuoscio G., Saiakhova A., Liu C.-F., et al., 2017 PRC2 is

Dispensable in Vivo for β-Catenin-Mediated Repression of Chondrogenesis in

Mouse Embryonic Cranial Mesenchyme. G3&#58; Genes|Genomes|Genetics

8: g3.300311.2017.

Firulli B. A., Krawchuk D., Centonze V. E., Vargesson N., Virshup D. M., et al., 2005

Altered Twist1 and Hand2 dimerization is associated with Saethre-Chotzen

syndrome and limb abnormalities. Nat. Genet. 37: 373–81.

Funato N., Chapman S. L., McKee M. D., Funato H., Morris J. a, et al., 2009 Hand2

controls osteoblast differentiation in the branchial arch by inhibiting DNA binding

of Runx2. Development 136: 615–25.

Gaur T., Lengner C. J., Hovhannisyan H., Bhat R. A., Bodine P. V. N., et al., 2005 Canonical

WNT signaling promotes osteogenesis by directly stimulating Runx2 gene

expression. J. Biol. Chem. 280: 33132–33140.

Gibson W. T., Hood R. L., Zhan S. H., Bulman D. E., Fejes A. P., et al., 2012 Mutations in

EZH2 cause weaver syndrome. Am. J. Hum. Genet. 90: 110–118. 148

Goodnough L. H., Chang A., Treloar C., Yang J., Scacheri P. C., et al., 2012 Twist1

mediates repression of chondrogenesis by β-catenin to promote cranial bone

progenitor specification. Development 139: 4428–4438.

Goodnough L. H., DiNuoscio G. J., Ferguson J. W., Williams T., Lang R. A., et al., 2014

Distinct Requirements for Cranial Ectoderm and Mesenchyme-Derived Wnts in

Specification and Differentiation of Osteoblast and Dermal Progenitors. PLoS

Genet. 10: 12–14.

Goodnough L. H., Dinuoscio G. J., Atit R. P., 2016 Twist1 contributes to cranial bone

initiation and dermal condensation by maintaining wnt signaling responsiveness.

Dev. Dyn. 245: 144–156.

Graf D., Malik Z., Hayano S., Mishina Y., 2016 Common mechanisms in development and

disease: BMP signaling in craniofacial development. Cytokine Growth Factor Rev.

27: 129–139.

Grammatopoulos G. a, Bell E., Toole L., Lumsden a, Tucker a S., 2000 Homeotic

transformation of branchial arch identity after Hoxa2 overexpression. Development

127: 5355–5365.

Griffin J. N., Compagnucci C., Hu D., Fish J., Klein O., et al., 2013 Fgf8 dosage determines

midfacial integration and polarity within the nasal and optic capsules. Dev. Biol.

374: 185–197.

Haegel H., Larue L., Ohsugi M., Fedorov L., Herrenknecht K., et al., 1995 Lack of β- 149

catenin affects mouse development at gastrulation. Development 121: 3529–3537.

Haegele L., Ingold B., Naumann H., Tabatabai G., Ledermann B., et al., 2003 Wnt

signalling inhibits neural differentiation of embryonic stem cells by controlling bone

morphogenetic protein expression. Mol. Cell. Neurosci. 24: 696–708.

Hamburg-Shields E., Dinuoscio G. J., Mullin N. K., Lafayatis R., Atit R. P., 2015 Sustained

β-catenin activity in dermal fibroblasts promotes fibrosis by up-regulating

expression of extracellular matrix protein-coding genes. J. Pathol. 235: 686–697.

Han J., Ishii M., Bringas P., Maas R. L., Maxson R. E., et al., 2007 Concerted action of

Msx1 and Msx2 in regulating cranial neural crest cell differentiation during frontal

bone development. Mech. Dev. 124: 729–745.

Hanken J., Thorogood P., 1993 Evolution and development of the veterbrate skull: the

role of pattern formation. Tree 8: 9–15.

He X., Tamai K., Semenov M., Kato Y., Spokony R., et al., 2000 LDL-receptor-related

proteins in Wnt signal transduction. Nature 407: 530–535.

Heintzman N. D., Stuart R. K., Hon G., Fu Y., Ching C. W., et al., 2007 Distinct and

predictive chromatin signatures of transcriptional promoters and enhancers in the

human genome. Nat. Genet. 39: 311–8.

Heintzman N. D., Hon G. C., Hawkins R. D., Kheradpour P., Stark A., et al., 2009 Histone

modification at human enhancers reflect global cell-type specific gene expression.

Nature 459: 108–112. 150

Hill J., Clarke J. D. W., Vargesson N., Jowett T., Holder N., 1995 Exogenous retinoic acid

causes specific alterations in the development of the midbrain and hindbrain of the

zebrafish embryo including positional respecification of the Mauthner neuron.

Mech. Dev. 50: 3–16.

Hill T. P., Später D., Taketo M. M., Birchmeier W., Hartmann C., 2005 Canonical Wnt/β-

Catenin Signaling Prevents Osteoblasts from Differentiating into Chondrocytes.

Dev. Cell 8: 727–738.

Ho T. V., Iwata J., Ho H. A., Grimes W. C., Park S., et al., 2015 Integration of

comprehensive 3D microCT and signaling analysis reveals differential regulatory

mechanisms of craniofacial bone development. Dev. Biol. 400: 180–190.

Hoffmeyer K., Junghans D., Kanzler B., Kemler R., 2017 Trimethylation and Acetylation of

β-Catenin at Lysine 49 Represent Key Elements in ESC Pluripotency-Supp. Cell Rep.

18: 2815–2824.

Holmes G., Basilico C., 2012 Mesodermal expression of Fgfr2S252W is necessary and

sufficient to induce craniosynostosis in a mouse model of Apert syndrome. Dev.

Biol. 368: 283–293.

Hsieh J. J. D., Cheng E. H. Y., Korsmeyer S. J., 2003 Taspase1: A threonine aspartase

required for cleavage of MLL and proper HOX gene expression. Cell 115: 293–303.

Hsu Y. H., Zillikens M. C., Wilson S. G., Farber C. R., Demissie S., et al., 2010 An

integration of genome-wide association study and gene expression profiling to 151

prioritize the discovery of novel susceptibility loci for osteoporosis-related traits.

PLoS Genet. 6: 1–16.

Huang W., Loganantharaj R., Schroeder B., Fargo D., Li L., 2013 PAVIS: A tool for Peak

Annotation and Visualization. Bioinformatics 29: 3097–3099.

Ishii M., Merrill A. E., Chan Y.-S., Gitelman I., Rice D. P. C., et al., 2003 Msx2 and Twist

cooperatively control the development of the neural crest-derived skeletogenic

mesenchyme of the murine skull vault. Development 130: 6131–42.

Ishii M., Sun J., Ting M. C., Maxson R. E., 2015 The Development of the Calvarial Bones

and Sutures and the Pathophysiology of Craniosynostosis. Elsevier Inc.

Iwamoto M., Kitagaki J., Tamamura Y., Gentili C., Koyama E., et al., 2003 Runx2

expression and action in chondrocytes are regulated by retinoid signaling and

parathyroid hormone-related peptide (PTHrP). Osteoarthr. Cartil. 11: 6–15.

Jho E., Zhang T., Domon C., Joo C.-K., Freund J.-N., et al., 2002 Wnt/β-catenin/Tcf

signaling induces the transcription of Axin2, a negative regulator of the signaling

pathway. Mol. Cell. Biol. 22: 1172–83.

Jiang X., Iseki S., Maxson R. E., Sucov H. M., Morriss-Kay G. M., 2002 Tissue origins and

interactions in the mammalian skull vault. Dev. Biol. 241: 106–16.

Juan A. H., Derfoul A., Feng X., Conerly M. L., Macquarrie K. L., et al., 2011 Polycomb

EZH2 controls self-renewal and safeguards the transcriptional identity of skeletal

muscle stem cells Polycomb EZH2 controls self-renewal and safeguards the 152

transcriptional identity of skeletal muscle stem cells. Genes Dev.: 789–794.

Jung H. Y., Jun S., Lee M., Kim H. C., Wang X., et al., 2013 PAF and EZH2 induce wnt/β-

catenin signaling hyperactivation. Mol. Cell 52: 193–205.

Karperien M., Farih-sips H., Hendriks J. A. A., Lanske B., Papapoulos S. E., et al., 1999

Identification of a Retinoic Acid- Inducible Element in the Murine PTH / PTHrP

(Parathyroid Hormone/Parathyroid Hormone-Related Peptide) Receptor Gene.

Mol. Endocrinol. 13: 1183–1196.

Karsenty G., 2008 Transcriptional Control of Skeletogenesis. Annu. Rev. Genomics Hum.

Genet. 9: 183–196.

Kettunen P., Thesleff I., 1998 Expression and function of FGFs-4, -8, and -9 suggest

functional redundancy and repetitive use as epithelial signals during tooth

morphogenesis. Dev. Dyn. 211: 256–268.

Kim H. J., Kim J. H., Bae S. C., Choi J. Y., Kim H. J., et al., 2003 The protein kinase C

pathway plays a central role in the fibroblast growth factor-stimulated expression

and transactivation activity of Runx2. J. Biol. Chem. 278: 319–326.

Kim D., Salzberg S. L., 2011 TopHat-Fusion: an algorithm for discovery of novel fusion

transcripts. Genome Biol. 12: 15.

Kim D., Pertea G., Trapnell C., Pimentel H., Kelley R., et al., 2013a TopHat2: accurate

alignment of transcriptomes in the presence of insertions, deletions and gene

fusions. Genome Biol. 14: R36. 153

Kim K. Il, Park Y. S., Im G. Il, 2013b Changes in the epigenetic status of the SOX-9

promoter in human osteoarthritic cartilage. J. Bone Miner. Res. 28: 1050–1060.

Kimmel R. A., Turnbull D. H., Blanquet V., Wurst W., Loomis C. A., et al., 2000 Two

lineage boundaries coordinate vertebrate apical ectodermal ridge formation.

Genes Dev. 14: 1377–1389.

Knecht A. K., Bronner-Fraser M., 2002 Induction of the neural crest: a multigene

process. Nat. Rev. Genet. 3: 453–461.

Kochhar D. M., Jiang H., Penner J. D., Johnson A. T., Chandraratna R. A. ., 1998 The use

of a retinoid receptor antagonist in a new model to study vitamin A-dependent

developmental events. Int. J. Dev. Biol. 42: 601–608.

Komori T., Yagi H., Nomura S., Yamaguchi A., Sasaki K., et al., 1997 Targeted Disruption

of Cbfa1 Results in a Complete Lack of Bone Formation owing to Maturational

Arrest of Osteoblasts. Cell 89: 755–764.

Korinek V., Barker N., Willert K., Molenaar M., Roose J., et al., 1998 Two members of the

Tcf family implicated in Wnt/β-catenin signaling during embryogenesis in the

mouse. Mol. Cell. Biol. 18: 1248–56.

Kretzschmar K., Watt F. M., 2012 Lineage tracing. Cell 148: 33–45.

Krispin S., Nitzan E., Kassem Y., Kalcheim C., 2010 Evidence for a dynamic

spatiotemporal fate map and early fate restrictions of premigratory avian neural

crest. Development 137: 585–595. 154

Krumlauf R., 1994 Hox genes in vertebrate development. Cell 78: 191–201.

Kumar D., Lassar A. B., 2014 Fibroblast Growth Factor Maintains Chondrogenic Potential

of Limb Bud Mesenchymal Cells by Modulating DNMT3A Recruitment. Cell Rep. 8:

1419–1431.

Langmead B., Salzberg S. L., 2012 Fast gapped-read alignment with Bowtie 2. Nat

Methods 9: 357–359.

Lee T. I., Jenner R. G., Boyer L. A., Guenther M. G., Levine S. S., et al., 2006 Control of

Developmental Regulators by Polycomb in Human Embryonic Stem Cells. Cell 125:

301–313.

Lee Y., Lee J.-Y., Kim M. H., 2014 PI3K/Akt pathway regulates retinoic acid-induced Hox

gene expression in F9 cells. Dev. Growth Differ. 56: 518–25.

Li X., Cao X., 2006 BMP signaling and skeletogenesis. Ann. N. Y. Acad. Sci. 1068: 26–40.

Li X., Gonzalez M. E., Toy K., Filzen T., Merajver S. D., et al., 2009 Targeted

Overexpression of EZH2 in the Mammary Gland Disrupts Ductal Morphogenesis

and Causes Epithelial Hyperplasia. Am. J. Pathol. 175: 1246–1254.

Li J., Cui Y., Xu J., Wang Q., Yang X., et al., 2017 Suppressor of Fused restraint of

Hedgehog activity level is critical for osteogenic proliferation and differentiation

during calvarial bone development. J. Biol. Chem. 292: 15814–15825.

Liu C., Kato Y., Zhang Z., Do V. M., Yankner B. a, et al., 1999 β-Trcp couples β-catenin

phosphorylation-degradation and regulates Xenopus axis formation. Proc. Natl. 155

Acad. Sci. U. S. A. 96: 6273–6278.

Liu C. F., Lefebvre V., 2015 The transcription factors SOX9 and SOX5/SOX6 cooperate

genome-wide through super-enhancers to drive chondrogenesis. Nucleic Acids Res.

43: 8183–8203.

Liu L., Liu X., Ren X., Tian Y., Chen Z., et al., 2016 Smad2 and Smad3 have differential

sensitivity in relaying TGFβ signaling and inversely regulate early lineage

specification. Sci. Rep. 6: 21602.

Livet J., Weissman T. A., Kang H., Draft R. W., Lu J., et al., 2007 Transgenic strategies for

combinatorial expression of fluorescent proteins in the nervous system. Nature

450: 56–62.

Lohnes D., Mark M., Mendelsohn C., Dolle P., Dierich A., et al., 1994 Function of the

retinoic acid receptors (RARs) during development (I). Craniofacial and skeletal

abnormalities in RAR double mutants. Development 120: 2723–2748.

Lui J. C., Garrison P., Nguyen Q., Ad M., Keembiyehetty C., et al., 2016 EZH1 and EZH2

promote skeletal growth by repressing inhibitors of chondrocyte proliferation and

hypertrophy. Nat. Commun. 7: 13685.

Lund A. H., Lohuizen M. Van, 2004 Polycomb complexes and silencing mechanisms. Curr.

Opin. Cell Biol. 16: 239–246.

Lynch M. D., Smith A. J. H., Gobbi M. De, Flenley M., Hughes J. R., et al., 2012 An

interspecies analysis reveals a key role for unmethylated CpG dinucleotides in 156

vertebrate Polycomb complex recruitment. EMBO J. 31: 317–329.

Maclean G., Dollé P., Petkovich M., 2009 Genetic disruption of CYP26B1 severely affects

development of neural crest derived head structures, but does not compromise

hindbrain patterning. Dev. Dyn. 238: 732–745.

Margueron R., Reinberg D., 2011 The Polycomb complex PRC2 and its mark in life.

Nature 469: 343–9.

Mark M., Ghyselinck N. B., Chambon P., 2009 Function of retinoic acid receptors during

embryonic development. Nucl.Recept.Signal. 7: e002.

McEwen K. R., Leitch H. G., Amouroux R., Hajkova P., 2013 The impact of culture on

epigenetic properties of pluripotent stem cells and pre-implantation embryos.

Biochem Soc Trans 41: 711–719.

McLean C. Y., Bristor D., Hiller M., Clarke S. L., Schaar B. T., et al., 2010 GREAT improves

functional interpretation of cis-regulatory regions. Nat. Biotechnol. 28: 495–501.

Mead T. J., Wang Q., Bhattaram P., Dy P., Afelik S., et al., 2013 A far-upstream (-70 kb)

enhancer mediates Sox9 auto-regulation in somatic tissues during development

and adult regeneration. Nucleic Acids Res. 41: 4459–4469.

Messerschmidt D. M., Knowles B. B., Solter D., 2014 DNA methylation dynamics during

epigenetic reprogramming in the germline and preimplantation embryos. Genes

Dev. 28: 812–828.

Minoux M., Holwerda S., Vitobello A., Kitazawa T., Kohler H., et al., 2017 Gene bivalency 157

at Polycomb domains regulates cranial neural crest positional identity. Science (80-.

). 355: eaal2913.

Mirzamohammadi F., Papaioannou G., Inloes J. B., Rankin E. B., Xie H., et al., 2016

Polycomb repressive complex 2 regulates skeletal growth by suppressing Wnt and

TGF-β signalling. Nat. Commun. 7: 12047.

Moosa S., Wollnik B., 2016 Altered FGF signalling in congenital craniofacial and skeletal

disorders. Semin. Cell Dev. Biol. 53: 115–125.

Nakamura Y., Paiva Alves E. de, Veenstra G. J., Hoppler S., 2016 Tissue- and stage-

specific Wnt target gene expression is controlled subsequent to β-catenin

recruitment. Development: 1914–1925.

Nakashima K., Zhou X., Al. G. K. et, 2002 The novel zinc fingercontaining transcription

factor Osterix is required for osteoblast 12 BioMed Research International

differentiation and bone formation. Cell 108: 17–29.

Niemann S., Zhao C., Pascu F., Stahl U., Aulepp U., et al., 2004 Homozygous WNT3

mutation causes tetra-amelia in a large consanguineous family. Am. J. Hum. Genet.

74: 558–63.

Nishio Y., Dong Y., Paris M., O'Keefe R. J., Schwarz E. M., et al., 2006 Runx2-

mediated regulation of the zinc finger Osterix/Sp7 gene. Gene 372: 62–70.

O’Carroll D., Erhardt S., Pagani M., Barton S. C., 2001 The Polycomb -Group Gene Ezh2 Is

Required for Early Mouse Development. Mol. Cell. Biol. 21: 4330–4336. 158

O’Geen H., Ren C., Nicolet C. M., Perez A. A., Halmai J., et al., 2017 dCas9-based

epigenome editing suggests acquisition of histone methylation is not sufficient for

target gene repression. Nucleic Acids Res. 44: 4123–4133.

O’Rourke M. P., Tam P. P. L., 2002 Twist functions in mouse development. Int. J. Dev.

Biol. 46: 401–413.

Ohtola J., Myers J., Akhtar-Zaidi B., Zuzindlak D., Sandesara P., et al., 2008 β-Catenin has

sequential roles in the survival and specification of ventral dermis. Development

135: 2321–2329.

Omoteyama K., Takagi M., 2009 FGF8 regulates myogenesis and induces Runx2

expression and osteoblast differentiation in cultured cells. J. Cell. Biochem. 106:

546–552.

Ornitz D., Marie P., 2002 FGF signaling pathways in endochondral and intramembranous

bone development and human genetic disease. Genes Dev.: 1446–1465.

Ornitz D. M., Itoh N., 2015 The fibroblast growth factor signaling pathway. Wiley

Interdiscip. Rev. Dev. Biol. 4: 215–266.

Osterwalder M., Speziale D., Shoukry M., Mohan R., Ivanek R., et al., 2014 HAND2

targets define a network of transcriptional regulators that compartmentalize the

early limb bud mesenchyme. Dev. Cell 31: 345–357.

Paro R., 1990 Imprinting a determined state into the chromatin of Drosophila. Trends

Genet. 6: 416–421. 159

Peng J. C., Valouev A., Swigut T., Zhang J., Zhao Y., et al., 2009 Jarid2/Jumonji

Coordinates Control of PRC2 Enzymatic Activity and Target Gene Occupancy in

Pluripotent Cells. Cell 139: 1290–1302.

Pfaff M. J., Xue K., Li L., Horowitz M. C., Steinbacher D. M., et al., 2016 FGFR2c-mediated

ERK-MAPK activity regulates coronal suture development. Dev. Biol. 415: 242–250.

Quarto N., Wan D. C., Kwan M. D., Panetta N. J., Li S., et al., 2010 Origin matters:

Differences in embryonic tissue origin and Wnt signaling determine the osteogenic

potential and healing capacity of frontal and parietal calvarial bones. J. Bone Miner.

Res. 25: 1680–1694.

Reinhold M. I., Kapadia R. M., Liao Z., Naski M. C., 2006 The Wnt-inducible transcription

factor Twist1 inhibits chondrogenesis. J. Biol. Chem. 281: 1381–1388.

Rhinn M., Dolle P., 2012 Retinoic acid signalling during development. Development 139:

843–858.

Riising E. M., Comet I., Leblanc B., Wu X., Johansen J. V., et al., 2014 Gene silencing

triggers polycomb repressive complex 2 recruitment to CpG Islands genome wide.

Mol. Cell 55: 347–360.

Rivera-Perez J. A., Wakamiya M., Behringer R. R., 1999 Goosecoid acts cell

autonomously in mesenchyme-derived tissues during craniofacial development.

Development 126: 3811–3821.

Rivers L. E., Young K. M., Rizzi M., Jamen F., Psachoulia K., et al., 2008 PDGFRA/NG2 glia 160

generate myelinating oligodendrocytes and piriform projection neurons in adult

mice. Nat. Neurosci. 11: 1392–1401.

Roberts A., Pimentel H., Trapnell C., Pachter L., 2011a Identification of novel transcripts

in annotated genomes using RNA-seq. Bioinformatics 27: 2325–2329.

Roberts A., Trapnell C., Donaghey J., Rinn J. L., Pachter L., 2011b Improving RNA-Seq

expression estimates by correcting for fragment bias. Genome Biol. 12: R22.

Robinson J., Thorvaldsdóttir H., Winckler W., Guttman M., Lander E. S., et al., 2011

Integrative Genomics Viewer. Nat. Biotechnol. 29: 24–26.

Roh T.-Y., Cuddapah S., Cui K., Zhao K., 2006 The genomic landscape of histone

modifications in human T cells. Proc. Natl. Acad. Sci. U. S. A. 103: 15782–7.

Rothbart S. B., Strahl B. D., 2014 Interpreting the language of histone and DNA

modifications. Biochim. Biophys. Acta - Gene Regul. Mech. 1839: 627–643.

Roybal P. G., Wu N. L., Sun J., Ting M. chun, Schafer C. A., et al., 2010 Inactivation of

Msx1 and Msx2 in neural crest reveals an unexpected role in suppressing

heterotopic bone formation in the head. Dev. Biol. 343: 28–39.

Saldanha A. J., 2004 Java Treeview - Extensible visualization of microarray data.

Bioinformatics 20: 3246–3248.

Satokata I., Maas R., 1994 Msx1 deficient mice exhibit cleft palate and abnormalities of

craniofacial and tooth development. Nat. Genet. 6: 348–356.

Satokata I., Ma L., Ohshima H., Bei M., Ian W., et al., 2000 Msx2 deficiency in mice 161

causes pleiotropic defects in bone growth and ectodermal organ formation. Nat.

Genet. 24: 391–395.

Savory J. G. A., Edey C., Hess B., Mears A. J., Lohnes D., 2014 Identification of novel

retinoic acid target genes. Dev. Biol. 395: 199–208.

Schilling T. F., Kimmel C. B., 1994 Segment and cell type lineage restrictions during

pharyngeal arch development in the zebrafish embryo. Development 120: 483–94.

Schindelin J., Arganda-Carreras I., Frise E., Kaynig V., Longair M., et al., 2012 Fiji: an

open-source platform for biological-image analysis. Nat. Methods 9: 676–82.

Schneider C. a, Rasband W. S., Eliceiri K. W., 2012 NIH Image to ImageJ: 25 years of

image analysis. Nat. Methods 9: 671–675.

Schübeler D., 2015 Function and information content of DNA methylation. Nature 517:

321–326.

Schuettengruber B., Bourbon H. M., Croce L. Di, Cavalli G., 2017 Genome Regulation by

Polycomb and Trithorax: 70 Years and Counting. Cell 171: 34–57.

Schuijers J., Mokry M., Hatzis P., Cuppen E., Clevers H., 2014 Wnt-induced

transcriptional activation is exclusively mediated by TCF/LEF. EMBO J. 33: 146–156.

Schwarz D., Varum S., Zemke M., Schöler A., Baggiolini A., et al., 2014 Ezh2 is required

for neural crest-derived cartilage and bone formation. Development 141: 867–77.

Shen X., Liu Y., Hsu Y. J., Fujiwara Y., Kim J., et al., 2008 EZH1 Mediates Methylation on

Histone H3 Lysine 27 and Complements EZH2 in Maintaining Stem Cell Identity and 162

Executing Pluripotency. Mol. Cell 32: 491–502.

Shen L., Shao N., Liu X., Nestler E., 2014 ngs.plot: Quick mining and visualization of next-

generation sequencing data by integrating genomic databases. BMC Genomics 15:

284.

Shi B., Liang J., Yang X., Wang Y., Zhao Y., et al., 2007 Integration of estrogen and Wnt

signaling circuits by the polycomb group protein EZH2 in breast cancer cells. Mol.

Cell. Biol. 27: 5105–5119.

Shimono K., Tung W., Macolino C., Chi A. H.-T., Didizian J. H., et al., 2011 Potent

inhibition of heterotopic ossification by nuclear retinoic acid receptor-γ agonists.

Nat. Med. 17: 454–460.

Simon J. a, Kingston R. E., 2009 Mechanisms of polycomb gene silencing: knowns and

unknowns. Nat. Rev. Mol. Cell Biol. 10: 697–708.

Snippert H. J., Flier L. G. van der, Sato T., Es J. H. van, Born M. van den, et al., 2010

Intestinal crypt homeostasis results from neutral competition between

symmetrically dividing Lgr5 stem cells. Cell 143: 134–144.

Snitow M., Lu M., Cheng L., Zhou S., Morrisey E. E., 2016 Ezh2 restricts the smooth

muscle lineage during mouse lung mesothelial development. Development:

dev.134932.

Song J., Du Z., Ravasz M., Dong B., Wang Z., et al., 2015 A Protein Interaction between -

Catenin and Dnmt1 Regulates Wnt Signaling and DNA Methylation in Colorectal 163

Cancer Cells. Mol. Cancer Res. 13: 969–981.

Soriano P., 1999 Generalized lacZ expression with the ROSA26 Cre reporter strain. Nat.

Genet. 21: 70–71.

Sparmann A., Lohuizen M. van, 2006 Polycomb silencers control cell fate, development

and cancer. Nat. Rev. Cancer 6: 846–856.

States U., 2001 Craniofacial anomalies and associated birth defects. Glob. Regist.

Database Craniofacial Anomalies: 15–33.

Stott F. J., Bates S., James M. C., McConnell B. B., Starborg M., et al., 1998 The

alternative product from the human CDKN2A locus, p14(ARF), participates in a

regulatory feedback loop with p53 and MDM2. EMBO J. 17: 5001–14.

Subramanian A., Tamayo P., Mootha V. K., Mukherjee S., Ebert B. L., et al., 2005 Gene

set enrichment analysis: a knowledge-based approach for interpreting genome-

wide expression profiles. Proc. Natl. Acad. Sci. U. S. A. 102: 15545–50.

Suzuki A., Ueno N., Hemmati-Brivanlou A., 1997 Xenopus msx1 mediates epidermal

induction and neural inhibition by BMP4. Development 124: 3037–44.

Tanay A., O’Donnell A. H., Damelin M., Bestor T. H., 2007 Hyperconserved CpG domains

underlie Polycomb-binding sites. Proc. Natl. Acad. Sci. 104: 5521–5526.

Tatton-Brown K., Hanks S., Ruark E., Zachariou A., Duarte S. D. V., et al., 2011 Germline

mutations in the oncogene EZH2 cause Weaver syndrome and increased human

height. Oncotarget 2: 1127–33. 164

Thomas P. D., Campbell M. J., Kejariwal A., Mi H., Karlak B., 2003 PANTHER: A Library of

Protein Families and Subfamilies Indexed by Function. Genome Res. 13: 2129–

2141.

Thorvaldsdóttir H., Robinson J. T., Mesirov J. P., 2013 Integrative Genomics Viewer

(IGV): High-performance genomics data visualization and exploration. Brief.

Bioinform. 14: 178–192.

Tien C.-L., Jones A., Wang H., Gerigk M., Nozell S., et al., 2015 Snail2/Slug cooperates

with Polycomb repressive complex 2 (PRC2) to regulate neural crest development.

Development 142: 722–731.

Trainor P. A., Krumlauf R., 2001 Hox genes, neural crest cells and branchial arch

patterning. Curr. Opin. Cell Biol. 13: 698–705.

Trainor P. A., 2007 Craniofacial birth defects: the role of neural crest cells in the etiology

and pathogenesis of treacher collins syndrome and the potential for prevention.

Am. J. Med. Genet. 2: 1–6.

Tran T. H., Jarrell A., Zentner G. E., Welsh A., Brownell I., et al., 2010 Role of canonical

Wnt signaling/β-catenin via Dermo1 in cranial dermal cell development.

Development 137: 3973–3984.

Trapnell C., Pachter L., Salzberg S. L., 2009 TopHat: Discovering splice junctions with

RNA-Seq. Bioinformatics 25: 1105–1111.

Trapnell C., Williams B. A., Pertea G., Mortazavi A., Kwan G., et al., 2010 Transcript 165

assembly and quantification by RNA-Seq reveals unannotated transcripts and

isoform switching during cell differentiation. Nat. Biotechnol. 28: 511–515.

Trapnell C., Hendrickson D. G., Sauvageau M., Goff L., Rinn J. L., et al., 2013 Differential

analysis of gene regulation at transcript resolution with RNA-seq. Nat. Biotechnol.

31: 46–53.

Tribulo C., Aybar M. J., Nguyen V. H., Mullins M. C., Mayor R., 2003 Regulation of Msx

genes by a Bmp gradient is essential for neural crest specification. Development

130: 6441–52.

Uchibe K., Son J., Larmour C., Pacifici M., Enomoto-iwamoto M., et al., 2017 Genetic and

pharmacological inhibition of retinoic acid receptor γ function promotes

endochondral bone formation. J. Orthop. Res. 35: 1096–1105.

Verani R., Cappuccio I., Spinsanti P., Gradini R., Caruso A., et al., 2007 Expression of the

Wnt inhibitor Dickkopf-1 is required for the induction of neural markers in mouse

embryonic stem cells differentiating in response to retinoic acid. J. Neurochem.

100: 242–250.

Wang X., Reid Sutton V., Omar Peraza-Llanes J., Yu Z., Rosetta R., et al., 2007 Mutations

in X-linked PORCN, a putative regulator of Wnt signaling, cause focal dermal

hypoplasia. Nat. Genet. 39: 836–838.

Wang L., Jin Q., Lee J.-E., Su I., Ge K., 2010 Histone H3K27 methyltransferase Ezh2

represses Wnt genes to facilitate adipogenesis. Proc. Natl. Acad. Sci. U. S. A. 107: 166

7317–7322.

Wang R. N., Green J., Wang Z., Deng Y., Qiao M., et al., 2014 Bone Morphogenetic

Protein (BMP) signaling in development and human diseases. Genes Dis. 1: 87–105.

Weston A. D., Chandraratna R. A. S., Torchia J., Underhill T. M., 2002 Requirement for

RAR-mediated gene repression in skeletal progenitor differentiation. J. Cell Biol.

158: 39–51.

Wilkie A. O. M., 1997 Craniosynostosis: Genes and mechanisms. Hum. Mol. Genet. 6:

1647–1656.

Xu K., Wu Z. J., Groner A. C., He H. H., Cai C., et al., 2012 EZH2 oncogenic activity in

castration-resistant prostate cancer cells is Polycomb-independent. Science 338:

1465–9.

Xu J., Shao Z., Li D., Xie H., Kim W., et al., 2015 Developmental control of polycomb

subunit composition by GATA factors mediates a switch to non-canonical functions.

Mol. Cell 57: 304–316.

Yang X., Kilgallen S., Andreeva V., Spicer D. B., Pinz I., et al., 2010 Conditional expression

of Spry1 in neural crest causes craniofacial and cardiac defects. BMC Dev. Biol. 10:

1–12.

Yao B., Wang Q., Liu C. F., Bhattaram P., Li W., et al., 2015 The SOX9 upstream region

prone to chromosomal aberrations causing campomelic dysplasia contains multiple

cartilage enhancers. Nucleic Acids Res. 43: 5394–5408. 167

Yasuhara R., Yuasa T., Williams J. A., Byers S. W., Shah S., et al., 2010 Wnt/β-Catenin and

retinoic acid receptor signaling pathways interact to regulate chondrocyte function

and matrix turnover. J. Biol. Chem. 285: 317–327.

Yi S. A., Um S. H., Lee J., Yoo J. H., Bang S. Y., et al., 2016 S6K1 Phosphorylation of H2B

Mediates EZH2 Trimethylation of H3: A Determinant of Early Adipogenesis. Mol.

Cell 62: 1–10.

Yoshida T., Vivatbutsiri P., Morriss-Kay G., Saga Y., Iseki S., 2008 Cell lineage in

mammalian craniofacial mesenchyme. Mech. Dev. 125: 797–808.

Yu K., Xu J., Liu Z., Sosic D., Shao J., et al., 2003 Conditional inactivation of FGF receptor 2

reveals an essential role for FGF signaling in the regulation of osteoblast function

and bone growth. Development 130: 3063–74.

Zelzer E., Olsen B. R., 2003 The Genetic Basis for Skeletal Disease. Nature 423: 343–348.

Zemke M., Draganova K., Klug A., Schöler A., Zurkirchen L., et al., 2015 Loss of Ezh2

promotes a midbrain-to-forebrain identity switch by direct gene derepression and

Wnt-dependent regulation. BMC Biol. 13: 103.

Zhang Y., Liu T., Meyer C. A., Eeckhoute J., Johnson D. S., et al., 2008 Model-based

Analysis of ChIP-Seq (MACS). Genome Biol. 9: R137.

Zhao E., Xu H., Wang L., Kryczek I., Wu K., et al., 2012 Bone marrow and the control of

immunity. Cell. Mol. Immunol. 9: 11–19.

Zhu P., Wang Y., Huang G., Ye B., Liu B., et al., 2016 Lnc-β-Catm elicits EZH2-dependent 168

β-catenin stabilization and sustains liver CSC self-renewal. Nat. Struct. Mol. Biol. 23:

631–639.