Cranium growth, patterning and homeostasis

Skull shapes have fascinated humans for millennia. Starting in ancient cultures and lasting until the 20th century, skull shaping, known as artificial cranial deformation, was commonly used for rites of passage, instilling communal unity, and to symbolize beauty and social status (Alfonso-Durrruty et al., 2015). Among the earliest written records of artificial cranial deformation are by Hippocrates in ‘On air, waters, and places’ (∼400 BCE). He describes a tribe known as the ‘Macrocephali’ that practiced artificial head lengthening by tightly wrapping newborn heads with cloth. By applying pressure during infancy and early childhood when the skull is malleable, mechanical deformation modulates physiological bone growth and permanently alters skull morphology. Similarly, the shape of the modern human skull, which is more dome-shaped compared to ancient hominids, evolved in response to internal forces and the expansion of the frontal and temporal lobes (Lieberman et al., 2002). These cultural practices and evolutionary adaptations exemplify the remarkable ability of the skull to grow and adapt in response to environmental and genetic factors, raising many questions about the signaling pathways and cell types that are involved.

The skull is a complex collection of flat bones composed of two main anatomical divisions. One aspect, the viscerocranium, forms the facial skeleton and is derived from the neural crest (Jin et al., 2016). The other division, the neurocranium, surrounds the brain (Kaucka and Adameyko, 2019). The neurocranium is further subdivided into the chondrocranium and calvarium, which form the skull base and skull vault, respectively. The calvarium includes the frontal and parietal bones, the squamous portion of the temporal bone, and the interparietal portion of the occipital bone (Holmes, 2012; Carter and Anslow, 2009; Ferguson et al., 2018) (Fig. 1). At the junction of these bones lie sutures, strips of fibrous tissue that act as synarthrotic joints. Sutures permit skull compliance, assist with shock absorption and are major signaling centers that regulate the shape of calvarial bones and their respective boundaries (Alaqeel et al., 2006; Siismets and Hatch, 2020; Opperman et al., 2000). Four major sutures separate the calvarial bones. The metopic/frontal suture separates the frontal bones, the coronal suture separates the frontal and parietal bones, and the sagittal suture lies on the dorsal midline between the parietal bones. Finally, the lambdoid suture lies between the parietal and occipital bones, which corresponds to the parietal-interparietal boundary in rodents (Fig. 1) (Li et al., 2021).

Some anatomical and embryological differences between human and animal calvaria exist. For example, cranial sutures typically fuse and ossify in adult humans, whereas most sutures remain patent in rodents and zebrafish (Quarto and Longaker, 2005). Also, comparative anatomical studies suggest the zebrafish ‘coronal suture’ is the anatomical equivalent of the lambdoidal suture in mammals (Teng et al., 2019). Despite these differences, the molecular mechanisms and major developmental processes in skull morphogenesis remain remarkably consistent between vertebrates (Cubbage and Mabee, 1996; Kanther et al., 2019; Topczewska et al., 2016; Ferguson and Atit, 2019). Thus, vertebrate models allow for the characterization of cellular interactions and signaling processes that control skull morphogenesis, and how these are affected in human craniofacial disorders. Here, primarily focusing on human studies and mouse models, we review key studies and recent insights that have shaped our understanding of the cellular and molecular regulation of calvarial patterning in normal development and disease states. We also discuss recently described populations of sutural stem cells that mediate postnatal skull growth and injury repair, as well as cutting-edge regenerative medicine approaches that may transform how we treat skull malformations. We conclude by describing how skull development is intertwined with that of the meningeal vasculature and then subsequently explain how craniosynostosis affects these processes, with a focus on the potential implications for brain health and function.

Skull development requires the regulated commitment, differentiation and maturation of osteoblasts from undifferentiated mesenchymal precursor cells (Fig. 2A). Starting around mouse mid-gestation, between embryonic (E) days E8.5 and E9.5, neural crest cells from the mid-hindbrain domain along with the paraxial mesoderm migrate into the supraorbital ridge, forming a pool of calvarial progenitors in the supraorbital mesenchyme (SOM) (Ferguson and Atit, 2019; Jiang et al., 2002). This process is dependent upon canonical Wnt/β-catenin signaling and downstream factors, such as Apc (Hasegawa et al., 2002) and Twist1, a transcription factor required for cell migration and osteoprogenitor cell (OPC) specification from mesenchymal precursors (Tischfield et al., 2017; Goodnough et al., 2012). As cells migrate into the supraorbital ridge, they spatially segregate and form a distinct neural crest-mesoderm boundary (NeuCMesB) (Morriss-Kay and Wilkie, 2005; Holmes, 2012). Anterior to the eye, the neural crest comprises the rostral domain and gives rise to the frontal bone (Doro et al., 2019; Jiang et al., 2002). The caudal domain is composed of mesoderm-derived cells that form the parietal bone (Yoshida et al., 2008) (Fig. 2B).

After migration, from approximately E10.5 to E12.5, mesenchymal cells within the SOM undergo a bone lineage program to become specified into OPCs (Ferguson and Atit, 2019) (Fig. 2A). This process is dependent upon expression of Msx1 and Msx2, genes encoding homeobox transcription factors that control both OPC proliferation and differentiation (Liu et al., 1999; Satokata et al., 2000). The expression of these transcription factors requires fibroblast growth factors (FGFs) (Kim et al., 1998; Ignelzi et al., 2003) and the secretion of bone morphogenic proteins (BMPs) from the surrounding calvarial mesenchyme and meninges (Bonilla-Claudio et al., 2012; Rice et al., 2003). In mice, Msx1 and Msx2 are co-expressed in the frontal bone anlage and exhibit redundant functions during this period, whereas the parietal bone anlage solely expresses Msx2 at E12.5 (Han et al., 2007), as some molecular differences exist given their cell ontology.

The specification of osteogenic mesenchyme is dependent on paracrine Wnt signaling from the adjacent surface ectoderm. In mice, calvarial OPCs must be exposed to a gradient of Wnt10 and Wnt7b in order to sustain their osteogenic identity (Goodnough et al., 2014). Twist1 is required to transduce Wnt signaling and maintain Wnt responsiveness in the SOM (Goodnough et al., 2016). Subsequently, OPCs produce Wnt3a, Wnt5a, Wnt11 and Wnt16 in an autocrine manner to specify committed preosteoblasts and prevent cartilage formation through canonical Wnt and extracellular signal-regulated kinase (ERK) 1/2 signaling (Goodnough et al., 2014; Ibarra et al., 2021) (Fig. 2C).

As OPCs undergo specification, they express Twist1, along with Dlx5, low levels of alkaline phosphatase (ALP, encoded by Alpl) and Runx2, which is a master regulator and one of the earliest known markers for OPC specification (Komori et al., 1997; Otto et al., 1997; Ishii et al., 2003; Robledo et al., 2002). Regulating Runx2 activity is important to prevent premature differentiation into postmitotic osteoblasts. Twist1 is crucial for this process as it heterodimerizes with Runx2, inhibiting its ability to bind to DNA (Bialek et al., 2004). Once Twist1 expression declines, the brake on Runx2 function is relieved, permitting the differentiation of Osx1 (Sp7)/Runx2⁺ preosteoblasts towards more mature bone sialoprotein (Bsp or Ibsp)- and osteocalcin (Bglap)-expressing osteoblasts (Bialek et al., 2004) (Fig. 2A,D). Thus, Twist1 is necessary for specifying OPCs from mesenchymal progenitors and also controls the timing of osteoblast differentiation, ensuring the pool of proliferating OPCs is not prematurely depleted.

Once OPCs begin to upregulate Osx1 expression, which is activated by Runx2, they transition into a more mature preosteoblast state (Nishio et al., 2006). In both mice and zebrafish, Osx1 is essential for early osteoblast lineage commitment and the eventual differentiation of preosteoblasts into mature bone matrix-producing osteoblasts (Nakashima et al., 2002; Kague et al., 2016; Xu et al., 2018) (Fig. 2D). Loss of Osx1 in zebrafish causes excessive proliferation of OPCs and impedes their differentiation, leading to the formation of ectopic bones that take on the appearance of Wormian bones (Kague et al., 2016). In the absence of mouse Osx1, Runx2⁺ preosteoblasts begin to express chondrocyte-specific markers, such as Sox9, suggesting that Osx1 represses the differentiation of cartilage (Nakashima et al., 2002).

By E12.5 in mice, OPCs aggregate and form condensation centers in the SOM (Hall and Miyake, 2000) (Fig. 2B), which subsequently expand to form the flat bones. The initiation of condensation is influenced by the timing of OPC differentiation and maintenance of the progenitor pool. To facilitate the growth and maturation of the skull vault, OPCs must self-renew and gradually differentiate to ensure the progenitor pool is not prematurely depleted. Maintaining this balance requires the activities of multiple signaling molecules, including BMPs, transforming growth factor β (TGFβ) family proteins, Hedgehog (Hh) proteins, retinoic acid (RA) and Notch, which crosstalk to pattern the skull (Bonilla-Claudio et al., 2012; Li et al., 2015; Wurdak et al., 2005; Dong et al., 2010; Jeong et al., 2004; Pakvasa et al., 2021). Together, these signals set the equilibrium between cell proliferation and differentiation. For example, studies in mice have shown that BMP signaling is required for proper Dlx5, Runx2 and Osx1 expression, and failure to upregulate these transcription factors impedes osteoblast differentiation and calvarial growth (Nakashima et al., 2002; Miyama et al., 1999; Phimphilai et al., 2006; Baek et al., 2014) (Fig. 2D). Conversely, although the Notch mediator RBP-Jκ (Rbpj) and downstream effector Hey1 promote the proliferation of murine preosteoblasts via upregulation of Osx1 expression, they also suppress Runx2 activity to prevent premature differentiation into osteocalcin-expressing osteoblasts (Engin et al., 2008). Thus, there is likely crosstalk between Notch and BMP signaling (Pakvasa et al., 2021).

Dysregulation of these signaling pathways can impede mouse calvarial development by inducing premature OPC differentiation, thereby depleting the pools of OPCs and preventing SOM condensation (Singh et al., 2013; Maeno et al., 2011; Sun et al., 2013). For example, although murine BMP signaling is necessary to initiate the expression of Msx1, Msx2 and Runx2 (Rice et al., 2000, 2005; Beederman et al., 2013), preosteoblasts require Foxc1 to maintain the appropriate level of BMP responsiveness (Rice et al., 2003) (Fig. 2A). When Foxc1 is lost in mice, OPCs become overly sensitive to BMP signaling and upregulate Msx2 expression (Rice et al., 2003). Consequently, OPCs prematurely differentiate, which depletes the progenitor pool and inhibits condensation in SOM (Sun et al., 2013).

As mesenchymal condensations form, they expand bilaterally from the base of the skull to the apex of the head such that by ∼E15, the opposite edges of the paired frontal and parietal bones rudiments are juxtaposed (Fig. 3A). During expansion, growth factors and morphogens, including Hh and autocrine BMP signaling, control the proliferation of preosteoblasts at the leading edge, called the osteogenic front (OF) (Kim et al., 1998; Lenton et al., 2011; Veistinen et al., 2017; Jiang et al., 2019). Proliferating preosteoblasts express Osx1 and high levels of Runx2 (Ishii et al., 2015; Yoshida et al., 2008). They are responsive to FGF signaling, which is required for their proliferation and differentiation towards osteoblasts. Fgfr2 is required for proliferation at the leading edge (Iseki et al., 1999), whereas Fgfr1 controls the differentiation of the more mature osteoid-laying osteoblasts behind the OF (Johnson et al., 2000; Iseki et al., 1999) (Fig. 2A, Fig. 3B). Bordering these groups lies a population of cells that transiently express FGFR3 and are transitioning to the mature osteoblast state (Iseki et al., 1999). Activating mutations in FGFR3 affect calvarial development in mice and humans (Di Rocco et al., 2014; Nah et al., 2012), although less is known about the developmental functions of FGFR3 versus those of FGFR2. In zebrafish, the OF expresses Fgfr3 in a manner similar to that of mice (Ledwon et al., 2018) and FGFR3 is necessary for calvarial growth, osteoblast differentiation and suture formation (Dambroise et al., 2020).

Cell proliferation drives calvarial growth as mesenchymal condensations expand. Studies in mice suggest that the rate of expansion is partially determined by the number of OPCs in the condensed SOM, and subsequently by the numbers of proliferating preosteoblasts within the OF (Lana-Elola et al., 2007). Preosteoblasts proliferate through asymmetric cell division. In response to FGF2 signaling, one daughter cell continues to divide at the leading edge, whereas the other remains behind, exiting the cell cycle to differentiate into a more mature FGFR1⁺ osteoblast (Lana-Elola et al., 2007; Iseki et al., 1999). Additionally, FGF2 upregulates the expression of Twist1 in murine OPCs (Rice et al., 2000). In a positive feedback loop, Twist1 upregulates Fgfr2 expression to prime the sensitivity of OPCs to FGF2 (Johnson et al., 2000) (Fig. 3B). Interestingly, the protocadherin proteins Fat4 and Dchs1 also influence the proliferation of OPCs directly, as well as indirectly by controlling the relative balance of Yap (Yap1) and Taz (Tafazzin) (Crespo-Enriquez et al., 2019). In the absence of Fat4-Dchs1 signaling, levels of Yap become aberrantly high in OPCs, leading to excess formation of Yap-Runx2 complexes. This alters Runx2 activity, which expands the osteoprogenitor pool and increases proliferation at both the frontal and parietal bone OFs, while delaying the differentiation of Osx1⁺ preosteoblasts (Crespo-Enriquez et al., 2019).

Despite these data, the various mechanisms that control expansion, including cell proliferation and migration, are still somewhat unclear. Notably, lineage tracing in the SOM using an inducible En1-CreER driver found some labeled progenitors beyond the OF (Tran et al., 2010; Deckelbaum et al., 2012). These findings are supported by lipophilic tracer experiments tracking the migration of SOM progenitors (Yoshida et al., 2008). Additionally, although administration of nocodazole inhibited proliferation at the parietal bone OF by 44%, the reduction of parietal bone size was only 19% (Lana-Elola et al., 2007), suggesting that other unidentified mechanisms contribute to calvarial growth.

Calvarial OPCs are technically osteochondroprogenitor cells, possessing the ability to differentiate into both osteoblasts and chondrocytes in mice (Hill et al., 2005; Nakashima et al., 2002). However, calvarial OPCs directly differentiate into osteoblasts through intramembranous ossification, a process distinct from endochondral ossification of the long bones, which requires a cartilage intermediate. Canonical Wnt/β-catenin signaling in mice is required to repress chondrocyte differentiation (Hill et al., 2005; Day et al., 2005) by activating Twist1, which suppresses Sox9 function by binding to the 3′ UTR of the Sox9 gene (Goodnough et al., 2012) (Fig. 2C). Sox9 is a master regulator for chondrocyte differentiation and an inhibitor of Runx2 (Mori-Akiyama et al., 2003; Hill et al., 2005). Ablating β-catenin in murine OPCs using En1-Cre results in loss of Twist1 expression, upregulation of Sox9 activity and the conversion of flat bones to cartilage (Goodnough et al., 2012). Notably, balancing the levels of β-catenin activity is also essential for intramembranous ossification. For example, β-catenin upregulates Runx2 and Osx1 expression in committed preosteoblasts, and abolishing its activity perturbs osteoblast differentiation and skull development (Rodda and McMahon, 2006). Conversely, expressing a constitutively activated form of β-catenin inhibits osteoblast differentiation in mice by prolonging Twist1 expression, and ablating Twist1 expression restores normal skull development (Goodnough et al., 2012).

Intramembranous ossification is also sensitive to the levels of FGF signaling because activating mutations in FGFR1/2 can antagonize canonical Wnt/β-catenin signaling (Ambrosetti et al., 2008). Sox2, a downstream target of FGF signaling, is expressed by preosteoblasts in the OF and can inhibit β-catenin activity to downregulate target genes (Mansukhani et al., 2005). Forced Sox2 expression inhibits the differentiation of osteoblasts and causes parietal bones to become thin and immature (Holmes et al., 2011). Additionally, an excess of certain FGFs, such as FGF8 and FGF9, induces endochondral ossification and severe calvarial dysplasia (Schmidt et al., 2018; Govindarajan and Overbeek, 2006). Ectopic FGF9-driven endochondral ossification is partially rescued by reducing the activity of the Wnt inhibitor Axin2 (Schmidt et al., 2018). Although the exact mechanisms behind ectopic calvarial endochondral ossification are not well understood, excess platelet-derived growth factor α (PDGFRα, encoded by Pdgfra) signaling may drive this process. Humans with Apert syndrome and activating mutations in FGFR2 have increased levels of PDGFRα and epidermal growth factor (EGF) signaling in the coronal suture (Miraoui et al., 2010). Furthermore, recent work in Apert mice has revealed that ectopic cartilage forms in place of coronal sutures and constitutive activation of PDGFRα signaling can induce suture closure through endochondral ossification (Peskett et al., 2017; He and Soriano, 2017).

Intramembranous ossification also requires protection from excessive exposure to all-trans RA, a vitamin A-derived metabolite that controls cell fate decisions (Morriss-Kay, 1993). As the meninges begin to differentiate and baso-apically expand at E12.5, skull progenitors are proximal to cells expressing Rdh10/Raldh2 to produce RA in the future arachnoid membrane (Siegenthaler et al., 2009). Exogenously applied RA inhibits calvarial development and induces the formation of ectopic cartilage (Jiang et al., 2002). In addition, humans, zebrafish and mice with inactivating mutations in the Cyp26b1 gene, which encodes a RA-degrading enzyme, exhibit skull dysplasia and abnormal calvarial development (Spoorendonk et al., 2008; Maclean et al., 2009; Laue et al., 2011). Taken together, these findings suggest that OPCs may fail to become committed preosteoblasts when exposed to elevated levels of RA. In support, in vitro studies have shown that RA antagonizes Wnt signaling and inhibits osteoblast differentiation (Lind et al., 2017; Roa et al., 2019).

Periosteal stem cells (PSCs) have recently been shown to be a source of calvarial osteoblasts in normal development (Debnath et al., 2018) (Fig. 3B). The periosteum has traditionally been viewed to be a source of osteoblasts in the context of fractures and other bone injuries (Utvåg et al., 1996; van Gastel et al., 2012), and the molecular identities of periosteal cells contributing to bone healing are unclear. However, cathepsin K (Ctsk)-positive PSCs have been identified in the calvarial periosteum and in cranial sutures of mice, and directly form bone through intramembranous ossification (Debnath et al., 2018). When Osx1 is deleted via Ctsk-Cre, these cells fail to differentiate into osteoblasts, leading to hypoplastic and hypomineralized calvarial bones, and enlarged cranial sutures (Debnath et al., 2018).

The dura, the outermost of the three membranes of the brain and spinal cord, produces morphogens that regulate calvarial growth. Starting at E9.5 in mice, the cranial mesenchyme begins to express Foxc1 (Inman et al., 2013), which is required for the growth and expansion of the dura and the calvarium (Mishra et al., 2016; Vivatbutsiri et al., 2008). In parallel with the apical expansion of skull progenitors, the dura differentiates in a baso-apical manner through a process dependent on Foxc1 and Wnt signaling (Zarbalis et al., 2007; DiNuoscio and Atit, 2019). The dura secretes growth factors, such as FGF2, BMP4 and TGFβ1 (Tgfb1), all of which contribute to osteogenesis (Machida et al., 2014; Mehrara et al., 1999). TGFβ signaling acts upstream of FGF signaling and induces FGFR2 expression in calvarial OPCs (Sasaki et al., 2006). When canonical (SMAD-dependent) TGFβ signaling is lost, calvarial OPCs shift to non-canonical (SMAD-independent) signaling and fail to properly develop (Ito et al., 2003; Iwata et al., 2012). Shifting the balance back to canonical TGFβ signaling can recover calvarial development (Ho et al., 2015). In postnatal mice, FGF2 is enriched in the dura and promotes osteogenesis (Li et al., 2007, 2010; Ogle et al., 2004). Upon FGF2 stimulation, dural fibroblasts proliferate, secrete TGFβ1/TGFβ3 (Tgfb3) and produce even more FGF2 to induce osteogenesis (Spector et al., 2002).

The dura also regulates suture patency during both embryonic and postnatal development. In vitro studies with E19 rat calvaria showed that coronal suture patency is dependent on dural-derived heparin-binding factors (Opperman et al., 1996). When the dura is removed from fetal rat calvaria, coronal sutures fuse (Opperman et al., 1993, 1995; Opperman, 2000; Ogle et al., 2004). Additional experiments have induced coronal suture fusion by removing TGFβ3 activity, whereas fusion can be prevented by removing TGFβ2 (Tgfb2) activity (Opperman et al., 2000). In a mouse model of intrauterine constraint, elevated levels of TGFβ1 and TGFβ3 correlate with increased incidence of suture closure (Kirschner et al., 2002). Additionally, the application of exogenous TGFβ2 in E15.5 mouse calvarial explants induces sagittal suture fusion via activation of ERK-MAPK signaling (Lee et al., 2006). In postnatal rodents, TGFβ and FGF signaling from the dura is regionally specified and the relative levels dictate suture fusion. For example, the coronal, lambdoid and sagittal sutures remain patent throughout adulthood, but the posterior frontal suture fuses in rodents by the third week (Hermann et al., 2013; Most et al., 1998). In comparison with the sagittal suture, the dura associated with the posterior suture shows higher levels of TGFβ1-3 and FGF2 expression (Roth et al., 1997; Most et al., 1998; Greenwald et al., 2000). Co-culture experiments have shown that the dura derived from the posterior frontal suture has a greater capacity to induce osteogenic gene expression versus that of sagittal suture dura (Warren et al., 2003b). Furthermore, autologous transplantation experiments in which the posterior frontal and sagittal sutures were surgically resected and swapped caused sagittal suture fusion, whereas the posterior frontal suture remained patent (Levine et al., 1998).

Coordinated growth of the calvarial anlagen is necessary to form and position cranial sutures, which develop between the neighboring OFs of the paired frontal and parietal bones (Teng et al., 2019; Richtsmeier and Flaherty, 2013). The coronal suture serves as a model for understanding sutural development and patency, and the cell-autonomous and non-cell-autonomous mechanisms that regulate these processes. The coronal suture develops from a subpopulation of mesoderm-derived Engrailed1 (En1)-positive cells, which appears in the SOM by E11 in mice (Deckelbaum et al., 2012). These precursors proliferate and apically migrate between E11.5 and E13.5 to form the coronal suture, which gives rise to osteoblasts that make major and minor contributions to the parietal and frontal bones, respectively (Deckelbaum et al., 2012). In the sutural mesenchyme, En1 is necessary to maintain Twist1 and repress Msx2 expression and FGFR2 signaling to prevent sutural progenitor cells from aberrantly adopting an osteogenic phenotype (Deckelbaum et al., 2012). The murine sutural mesenchyme also contains a population of Prx1 (Prrx1)/Sca1 (Ly6a) double-positive stem cells that influence intramembranous ossification (Takarada et al., 2015). Recently, single-cell sequencing experiments have identified previously unreported populations of sutural progenitor cells and markers for specific subpopulations at the OF (Box 1). For intramembranous ossification to occur, these cells upregulate Runx2 expression and become recruited into the OF (Takarada et al., 2015).

En1 is necessary to maintain the coronal suture by establishing a NeuCMesB (Fig. 4A), thereby preventing the invasion of neural crest cells into the sutural mesenchyme and mesoderm (Deckelbaum et al., 2012; Merrill et al., 2006). Neural crest-derived cells secrete pro-osteogenic growth factors (i.e. FGFs) and/or have the propensity to differentiate into bone, which can cause fusion. The NeuCMesB is also enforced by Twist1 and Eph-ephrin signaling. Through genetic interactions with Twist1, ephrin A2 (Efna2)- and ephrin A4 (Efna4)-expressing cells in the non-osteogenic mesenchyme lining the frontal bone maintain the separation of the cranial neural crest and mesoderm via repulsive interactions with EphA4⁺ cells in the sutural mesenchyme (Ting et al., 2009; Merrill et al., 2006). Twist1 also maintains the NeuCMesB by regulating jagged 1 (Jag1) expression. In turn, Jag1 maintains the mesenchymal architecture of the suture and inhibits osteogenesis by attenuating the levels of SMAD, ERK, Notch and Wnt signaling (Yen et al., 2010).

Twist1 also maintains suture patency through its ability to heterodimerize with other basic helix-loop-helix (bHLH) proteins. In mice and human cell lines, Twist1 can signal by forming homodimers (T/T) or heterodimers with bHLH E proteins (T/E) (Fan et al., 2020; Mikheeva et al., 2019). T/T homodimers stimulate osteogenesis by upregulating the levels of Fgfr2 and periostin (Postn) and are preferentially localized to the murine OF, whereas T/E heterodimers are enriched in the mid-sutural mesenchyme (Connerney et al., 2006). Thus, Twist1 helps maintain a cell-fate border within the sutural mesenchyme – distinct from the NeuCMesB – that separates the OF from the mid-sutural mesenchyme (Fig. 4A). In mice, dimerization is influenced by the relative abundance of Twist1 and the bHLH E proteins, but also by Ids, which are bHLH proteins that preferentially bind to E proteins and disrupt the formation of T/E heterodimers by outcompeting Twist1 (Connerney et al., 2008). Thus, at sites of Id enrichment, such as the OF, T/T homodimers predominate and promote the proliferation of preosteoblasts by upregulating Fgfr2 expression (Connerney et al., 2008). By contrast, because Id expression is low in the murine sutural mesenchyme, T/E dimers form and repress responsiveness to pro-osteogenic signaling (Connerney et al., 2006). Thus, controlling the balance of Twist1, bHLH E proteins and Ids is crucial for maintaining suture patency and growth at the OF.

Twist1 expression in the sutural mesenchyme is controlled by BCL11B (Kyrylkova et al., 2016), a zinc finger protein that is enriched in the cranial sutures throughout mid-to-late calvarial development (Holmes et al., 2015). BCL11B is switched on at E14.5 in the coronal suture, whereas its expression is initiated at E16.5 in the sagittal, interfrontal and lambdoid sutures (Holmes et al., 2015). Similar to Twist1 haploinsufficient animals discussed below, Bcl11b knockout mice develop craniosynostosis and show upregulated expression of Runx2 and Fgfr2 in the sutural mesenchyme, and also increased cell proliferation at OFs (Kyrylkova et al., 2016; Rice et al., 2000; Bialek et al., 2004).

Postnatal sutures contain heterogeneous populations of multipotent mesenchymal stem cells (MSCs) (Doro et al., 2017). Sutural MSCs have overlapping, but also distinct, molecular signatures and show various capacities to differentiate into bone, cartilage and dura. The ontological relationships between sutural MSCs are unclear and it is unknown whether residing in spatially distinct locations within the sutural mesenchyme is important for the determination of cell fate. Cell ablation studies and findings from craniosynostosis models, however, have shown that sutural MSCs are required for postnatal growth of the calvarium, injury repair and suture patency (Wilk et al., 2017; Zhao et al., 2015). Thus, sutural MSCs promote skull expansion while the brain grows in adolescents, and later help maintain bone homeostasis and turnover in adults.

Sutural MSCs are characterized by unique or overlapping expression of Gli1, Axin2, Prx1 and CD51 (Itgav)/CD200 (Maruyama et al., 2016; Zhao et al., 2015; Di Pietro et al., 2020; Wilk et al., 2017; Menon et al., 2021). Several populations demonstrate clonal expansion, long-term self-renewal and the ability to differentiate (Maruyama et al., 2016; Wilk et al., 2017; Zhao et al., 2015). Most sutural MSCs express Gli1, whereas subsets also co-express Axin2 and Prx1 (Zhao et al., 2015; Di Pietro et al., 2020; Wilk et al., 2017). A landmark study in 2015 showed that Gli1-lacZ mice exhibit lacZ expression in the periosteum, dura and sutural mesenchyme of newborns (Zhao et al., 2015). Gli1-lacZ expression gradually becomes restricted to the sutural mesenchyme, such that by 1 month of age, lacZ-expressing cells are mainly detected at the midline, with expression absent in Runx2⁺/Osx1⁺ preosteoblasts at the OF (Zhao et al., 2015). This pattern holds consistent for most adult sutures. Interestingly, Gli1-lacZ cells are absent in the posterior frontal suture, the only suture that fuses in rodents. Lineage labeling has further revealed that Gli1⁺ sutural MSCs give rise to the sutural mesenchyme, as well as osteoblasts, the periosteal dura and osteoclasts, with the entire skull and surrounding tissue labeled after several months. Following injury and in response to IHH signaling, Gli1⁺ cells can mobilize to the region of repair and differentiate into the periosteum, dura and osteocytes (Zhao et al., 2015). Thus, Gli1⁺ sutural MSCs represent an essential pool of multipotent MSCs responsible for calvarial growth and turnover.

In mice, a subset of Gli1⁺ sutural MSCs also expresses Prx1 and differentiates into both chondrocytes and osteoblasts in vitro (Ouyang et al., 2014; Wilk et al., 2017). During mouse embryogenesis, Prx1 is expressed in the osteogenic mesenchyme and is required for the development of both the cranial neural crest-derived and paraxial mesoderm-derived calvarial bones as well as the intervening sutures (Ouyang et al., 2014; Wilk et al., 2017). Ablating these cells in the mouse postnatal sutural mesenchyme, however, does not stunt calvarial growth, but rather the ability of the skull to repair itself following injury. Furthermore, these cells are absent in the calvarial periosteum and dura and are distinct from Osx1-expressing preosteoblasts, and their abundance declines with age (Wilk et al., 2017). Prx1⁺/Gli1⁺ MSCs, therefore, appear to constitute a distinct lineage that can differentiate into osteoblasts under specific contexts (i.e. injury) and environmental conditions (i.e. exposure to Wnt signaling).

Similar to Gli1⁺ MSCs, Axin2⁺ MSCs are found at the midline of all sutures except the posterior frontal suture in mice (Maruyama et al., 2016). Axin2⁺ MSCs have much in common with Gli1⁺ MSCs, although they are a more restricted subset, making up ∼15% of mesenchymal cells. Axin2 expression is localized to the sutural midline by the first postnatal week and is maintained in adults, matching the spatiotemporal profile of Gli1⁺ MSCs (Maruyama et al., 2016; Zhao et al., 2015). Lineage labeling has shown that Axin2⁺ MSCs can differentiate into osteoblasts that are incorporated throughout the calvarium, as well as osteocytes (Maruyama et al., 2016). Following injury, Axin2-lacZ cells surround the repair site, and their descendants express markers for both preosteoblasts (Osx1) and osteoclasts (Sost), suggesting their involvement in bone homeostasis and repair. Transplantation of purified Axin2-lacZ-expressing MSCs into the site of injury hastens the healing process and results in bone engraftment (Maruyama et al., 2016). Thus, Axin2⁺ MSCs appear to be closely related to Gli1⁺ MSCs and constitute a long-lasting source of sutural MSCs that build and maintain the calvarium.

Premature fusion of the cranial sutures is called craniosynostosis, a disorder that affects one in ∼2000-2500 live births (Lajeunie et al., 1995; Boulet et al., 2008; Miller et al., 2017). Craniosynostosis can be idiopathic or occur as part of a defined genetic syndrome. Although all sutures have the potential to fuse prematurely, the sagittal suture is the most commonly affected in idiopathic craniosynostosis (Stanton et al., 2022). Close to 80 genes have been implicated in various forms of craniosynostosis (Twigg and Wilkie, 2015; Goos and Mathijssen, 2019). Although clinical manifestations of craniosynostosis are most commonly identified after birth, many – but not all – of the pathophysiological mechanisms seem to affect processes that regulate early suture patterning. As such, genes involved in craniosynostosis tend to be enriched in signaling pathways that control the proliferation and differentiation of OPCs, as well as the maintenance of the NeuCMesB (Stanton et al., 2022).

Saethre–Chotzen syndrome, caused by heterozygous missense and loss-of-function mutations (haploinsufficiency) in Twist1, is the second leading cause of syndromic craniosynostosis and primarily affects the coronal suture (el Ghouzzi et al., 1997). Findings suggest a dual paradigm wherein the coronal suture first assumes a partially osteogenic character, affecting the cell-fate border in the mesoderm, before allowing the invasion of neural crest cells into the mid-sutural mesoderm, obliterating the NeuCMesB (Ishii et al., 2015) (Fig. 4A,Ba). In Twist1^+/− mice, the coronal suture displays ectopic pro-osteogenic gene expression, including Fgfr2, Alpl and Msx2 (Merrill et al., 2006; Rice et al., 2000). As early as E12.5, an aberrant osteogenic character is evident, as detected by Notch2 expression (Yen et al., 2010). Twist1 haploinsufficiency is postulated to shift the balance between T/E and T/T dimers, causing T/T dimers to form in the mid-sutural mesenchyme, thereby promoting pro-osteogenic gene expression and cell fate and depleting uncommitted MSCs (Firulli et al., 2005; Connerney et al., 2006, 2008) (Fig. 4Bb). Inhibiting the expression of Fgfr2 in the sutural mesenchyme can rescue craniosynostosis, suggesting that this counteracts pro-osteogenic gene induction. Furthermore, it is suggested that missense mutations in Twist1 may affect its ability to repress Runx2 activity, promoting premature osteoblast differentiation and suture fusion (Bialek et al., 2004; Yoshida et al., 2005).

Several Twist1-associated genes and downstream effectors have also been implicated in craniosynostosis, pointing to the loss of the NeuCMesB as another underlying mechanism (Ting et al., 2009; Kamath et al., 2002; Sharma et al., 2013). Mutations in EFNA4 cause sagittal synostosis in humans (Merrill et al., 2006) and mutation of the gene encoding the binding partner of EFNA4, Epha4, causes craniosynostosis in mice (Ting et al., 2009). Epha4 acts downstream from Twist1 to prevent migratory neural crest-derived OPCs from invading the sutural mesenchyme, which is observed in Twist1^+/− mice, and Twist1^+/−;Epha4^+/− compound heterozygotes display more severe craniosynostosis (Ting et al., 2009). Furthermore, it is suggested that Twist1 regulates the Notch pathway, functionally linking their respective roles in NeuCMesB formation (Yen et al., 2010). Mutations in Jag1, a constituent of the Notch pathway, cause Alagille syndrome, which can be associated with craniosynostosis (Oda et al., 1997; Kamath et al., 2002). Jag1 is expressed in the embryonic coronal suture and its inactivation leads to fusion (Yen et al., 2010).

Twist1 forms heterodimers with the bHLH E protein Tcf12. Mutations in TCF12 cause coronal craniosynostosis in humans, closely resembling Saethre–Chotzen syndrome (Fan et al., 2020; Sharma et al., 2013). Twist1^+/−;Tcf12^+/− compound heterozygotes display severe craniosynostosis, suggesting that these proteins physically interact to maintain the cell-fate boundary within the coronal suture (Sharma et al., 2013; Ting et al., 2022). In both mice and zebrafish, Twist1-Tcf12 heterodimers control OPC proliferation at the OF (Teng et al., 2018). When heterodimers are lost, progenitor pools at the OF proliferate excessively, whereas sutural MSCs become depleted (Teng et al., 2018) (Fig. 4Bb,Bc). Interestingly, these mice also show neural crest- and mesoderm-cell mixing within the suture, suggesting that the balance of these dimers also maintains the NeuCMesB.

Gain-of-function mutations in the FGF family of receptors are among the most extensively studied, causing several forms of syndromic craniosynostosis including Apert (FGFR2), Crouzon (FGFR2), Muenke (FGFR3) and Pfeiffer syndromes (FGFR1 and FGFR2) (Sawh-Martinez and Steinbacher, 2019). Two different amino acid substitutions in Fgfr2 (S252W and P253R), which increase affinity of the mutant receptor for FGF ligands (Anderson, 1998; Chen et al., 2003; Wang et al., 2005; Yin et al., 2008), account for 99% of the cases of Apert syndrome, characterized by coronal suture fusion (Park et al., 1995; Wilkie et al., 1995). FGFR2 activates ERK1/2 signaling at the OF, controlling the proliferation and differentiation of preosteoblasts (Neben et al., 2014). In mice expressing constitutively active Fgfr2^S252W, cell proliferation is increased and even extends beyond the OF to encroach the mid-sutural mesenchyme (Wang et al., 2005) (Fig. 4Bc). Using RNA interference to silence the mutant Fgfr2^S252W transcript reduces ERK1/2 activity and restores patency (Shukla et al., 2007). Notably, limiting activation of FGFR2^S252W to the mesoderm is sufficient to induce craniosynostosis, as neural crest-specific activation has no effect. Thus, it is proposed that coronal synostosis in Apert syndrome is not caused by failure to maintain the NeuCMesB (as cell mixing is not observed), but rather by the cell-fate border within the suture that separates the mid-sutural mesenchyme from OPCs closer to the OF (Holmes and Basilico, 2012).

BMP signaling and control over osteoblast differentiation is also implicated in craniosynostosis. Loss of Bmp2/4 in the mouse cranial mesoderm affects the NeuCMesB and causes narrowing of the coronal suture (Tischfield et al., 2017). Conversely, constitutive activation of BMPR1A and downstream effectors (SMADs) in the neural crest promotes premature osteoblast differentiation and suture fusion, which is partially rescued by BMP receptor kinase inhibitors (Komatsu et al., 2013). BMP9, a potent inducer of osteogenesis, is also involved as it is more highly expressed in fused versus patent sutures (Luther et al., 2011; Song et al., 2020). Fused sutures are more responsive to BMP9 signaling, and forced expression increases osteogenic gene expression and matrix mineralization in sutural MSCs (Song et al., 2020). In addition, noggin (Nog), a BMP antagonist that is repressed by FGF2 signaling, is expressed in the dura of patent sutures. Gain-of-function mutations in FGFR2 receptors may attenuate noggin signaling, leading to upregulation of BMP activity and osteogenesis of the sutural mesenchyme (Warren et al., 2003a). Interestingly, sensory nerves, which innervate the periosteum of the coronal, sagittal and lambdoid sutures (Kosaras et al., 2009), may also play a role in maintaining suture patency. These nerves secrete BMP inhibitors, such as Fstl1, preventing excess BMP/TGFβ signaling and depletion of sutural MSCs (Tower et al., 2021).

Finally, Hh signaling and activation of the Gli family of transcription factors have been implicated in craniosynostosis. Humans with Greig cephalopolysyndactyly syndrome have disrupting mutations in Gli3 and some show synostosis of the lambdoid suture (Vortkamp et al., 1991; Johnston et al., 2005). Gli3-null mice exhibit lambdoid suture fusion, ectopic expression of Runx2, and imbalances between OPC proliferation and differentiation (Rice et al., 2010). Furthermore, Carpenter syndrome, an autosomal-recessive form of syndromic craniosynostosis marked by multi-suture fusion, is associated with mutations in Rab23 (Jenkins et al., 2007). A recent study showed that Rab23 is required to maintain embryonic suture patency (Hasan et al., 2020). In mice, Rab23 normally antagonizes Hh signaling through the regulation of the Smoothened (Smo) and Gli proteins, which modulate the expression of Hh target genes (Eggenschwiler et al., 2001, 2006). By repressing Gli1 and FGF signaling, Rab23 maintains normal levels of OPC proliferation and prevents excessive preosteoblast formation in sutures (Hasan et al., 2020).

Skulls are subjected to mechanical forces that are transmitted to the sutures. These forces alter gene expression, affect the equilibrium between differentiation and proliferation, and influence the maintenance of sutural MSCs. For example, tensional forces in rodent calvaria can upregulate canonical TGFβ signaling and maintain patency of the posterior frontal suture (Tholpady et al., 2007). Furthermore, compressive forces can induce aberrant fusion of the sagittal suture (Oppenheimer et al., 2009, 2012; Heller et al., 2007). During embryonic development, fetal constraint can alter the expression of morphogens, such as IHH, at the OF and induce craniosynostosis (Jacob et al., 2007). Although little is known about how mechanical forces influence sutural gene expression, studies suggest that cilia might play a role in sutural mechanotransduction. Primary cilia are important for maintaining the expression of osteogenic genes in human MSCs (Hoey et al., 2012; Chen et al., 2016), and ciliopathies, such as Sensenbrenner syndrome, have been associated with craniosynostosis (Tiberio et al., 2021). Interestingly, polycystin-1 (PC1, encoded by Pkd1), a mechanotransduction protein in primary cilia, is overexpressed in patients with craniosynostosis, and PC1 induces Runx2 expression through ERK signaling (Katsianou et al., 2021).

Sutural MSCs also maintain patency. Ablating postnatal Gli1⁺ MSCs with diphtheria toxin causes craniosynostosis and a dramatic reduction in skull growth, suggesting loss of Gli1⁺ MSCs may contribute to suture fusion (Zhao et al., 2015). Indeed, these cells are depleted in sutures undergoing fusion in Twist1^+/− mice and are undetectable in fused sutures (Zhao et al., 2015). Similar findings have been obtained with CD51⁺/CD200⁺ MSCs, which are present in sutures of postnatal rodents, with the exception of the posterior frontal suture. In Twist1^+/− mice, these cells are depleted in coronal sutures undergoing fusion (Menon et al., 2021). Furthermore, by inducing sagittal suture fusion via small molecule inhibition of TGFβ signaling, a significant decrease of CD51⁺/CD200⁺ MSCs is observed. By contrast, activating Wnt signaling via exogenous Wnt3a application in the posterior frontal suture, which normally has low levels of Wnt signaling and is devoid of CD51⁺/CD200⁺ MSCs, stimulates proliferation and increases the number of CD51⁺/CD200⁺ cells, leading to suture patency (Menon et al., 2021). This may explain why the posterior frontal suture, which normally fuses by endochondral ossification several weeks after birth, is naturally devoid of Gli1⁺ and CD51⁺/200⁺ MSCs (Menon et al., 2021). Thus, both cell ablation studies and findings in craniosynostosis models show that MSCs are crucial for suture patency. Moreover, sutural stem cell transplantation can restore patency in craniosynostosis mouse models (Box 2).

Dural venous sinuses are large superficial cerebral veins located in the periosteal dura attached to the skull. They develop in proximity to the coronal suture and parietal bone anlagen (Tischfield et al., 2017) (Fig. 5A). As the coronal suture develops and the parietal bone expands, the dural venous sinuses are actively growing and remodeling. In mice, this process requires paracrine BMP4 signaling from preosteoblasts and dura. Inactivating Twist1 in the neural crest and the cranial mesoderm via Sm22a-Cre shifts the NeuCMesB boundary caudally and perturbs the development of the coronal suture and parietal bone by E12.75. Concurrently, the dural venous sinuses fail to properly grow and remodel, leading to vessel regression and/or hypoplasia by E14.5 (Tischfield et al., 2017) (Fig. 5B). In addition, BMP4 expression is reduced in the sutural mesenchyme and the OF. Conditional inactivation of Bmp2/Bmp4 via Sm22a-Cre also shifts the NeuCMesB boundary caudally and affects the development of the coronal suture and parietal bone, causing similar effects on the growth and remodeling of the dural venous sinuses. Moreover, inactivating BMPR2 in venous endothelial cells results in hypoplastic vessels (Tischfield et al., 2017). Thus, the skull and dura crosstalk with endothelial cells to regulate venous angiogenesis, suggesting that the growth and patterning of the calvarium and underlying blood vessels co-evolved in mammals.

Recent insights into pathways that govern the maintenance, proliferation and cell-fate decisions of the osteoblast lineage have advanced our understanding of normal skull development. In particular, single-cell RNA-sequencing has identified previously unreported subpopulations of osteoblasts, sutural stem cells and dural cells (Holmes et al., 2021, 2020; Farmer et al., 2021). Understanding how these cells function and interact to regulate calvarial development and sutural homeostasis may enhance our ability to treat craniofacial disorders. Notably, a recent human genome-wide association study identified different loci that influence skull shape and are shared between humans and mice (Roosenboom et al., 2018). Expanding on these types of analyses will be beneficial, as variations in skull patterning may also affect vascular development, central nervous system fluid balance and/or waste clearance.

The effects of craniosynostosis and elevated intracranial pressure (ICP) on cerebrospinal fluid (CSF) balance and brain waste clearance also need to be investigated. Elevated ICP impedes CSF influx along perivascular channels in the brain (i.e. the glymphatic system), which is necessary to clear waste and transport it to dural lymphatic vessels (Iliff et al., 2012; Bolte et al., 2020). Affecting this process is associated with cognitive deficits and the buildup of tau and β-amyloid plaques in mice (Da Mesquita et al., 2018; Iliff et al., 2014). Importantly, cognitive deficits are present in craniosynostosis but, in most cases, the underlying causes are unknown (Brooks et al., 2018). Thus, it will be important to test whether brain waste clearance is impeded by craniosynostosis and whether animal models have combined deficits in the glymphatic and lymphatic systems (Box 3). Although craniosynostosis is currently treated through surgical interventions (Kyutoku and Inagaki, 2017), surgeries are not always effective, especially in syndromic craniosynostosis. In addition to sutural stem cell transplantation that can restore suture patency and normalize ICP in mouse models (Box 2), recent work has shown that applying cyclical mechanical loading to the frontal bone, which generates tensile force on the coronal suture, contributes to the normalization of skull growth in Crouzon mice (Moazen et al., 2022). As opening the skull vault can impair glymphatic function (Plog et al., 2020; Hablitz and Nedergaard, 2021), these minimally invasive methods may also reduce the risk of harmful side effects.

The figures were adapted from the following BioRender templates: ‘skull (lateral view)', ‘skull (superior view)’, ‘mouse skull (with mandible)’, ‘mouse skull (dorsal)’, ‘mouse brain (dorsal, with veins)’, ‘mouse embryo (E11.5)’, ‘mouse embryo (E12.5)’, ‘mouse embryo (E15.5)’, ‘mesenchymal stromal cell', ‘osteoprogenitor cell' and ‘osteoblast', by BioRender.com (2022). Retrieved from https://app.biorender.com/biorender-templates.