A Genetic-Pathophysiological Framework for Craniosynostosis

Stem Cell Specification and Migration

It has long been presumed that sutures contain a population of undifferentiated osteogenic cells with stem cell-like properties,^54,55 and this has recently received further experimental support with the demonstration that expression of Gli1 (a classical marker of hedgehog [HH] signaling) probably defines sutural stem cells at postnatal stages; cells emanating from the suture populate the periosteum on both surfaces of the growing bones, and genetic ablation of Gli1-expressing cells in 1-month-old mice caused coronal synostosis within a further month.⁵⁶ An elegant study (which also exploited Gli1 as the key marker of cell identity) pinpoints that the precursors of these cells originate from cephalic paraxial mesoderm, in the region of the rostral mesencephalon/caudal diencephalon and located immediately adjacent to the neural tube, at embryonic day (E)7.5 in response to sonic hedgehog (SHH) accumulation in the adjacent notochord (Figure 2A).⁵⁷ Lineage tracing using genetically marked cells shows that this population migrates laterally during E8.5–E9.5 to locate above the developing eye.⁵⁷ Of note, expression of Gli1 is transient because calvarial structures do not label after E8, indicating that the influence of SHH on these cells is rapidly lost.

Lineage Commitment

During a critical period from E9.5 to E11.5, the cells above the developing eye pattern the future coronal suture. Deckelbaum et al.⁵⁷ have proposed the term “supraorbital regulatory center” for this region, in recognition of its role as an organizing center. Put simply, groups of cells with osteogenic potential, originating from neural crest (future frontal bone) and mesoderm (future parietal bone), are separated by the undifferentiated stem cell population originating from a localized section of paraxial mesoderm (Figure 2B). During E11.5–E13.5, and co-ordinated with growth of the underlying brain, cells from the supraorbital region extend apically, centered above the diencephalic-telencephalic boundary, to overlay the surface of the rapidly growing brain.^57,62,63 Descendants of these cells can still be identified in the mid-sutural mesenchyme of the definitive coronal suture at birth (P0), demonstrating that they make a permanent contribution to the population of undifferentiated cells.⁵⁷ A separate population of descendants becomes integrated into the parietal bone mesenchyme and adopts an osteogenic fate; the future frontal bone originates mostly from cells of neural crest origin (next section).

Key transcription factors involved in orchestrating this early lineage commitment are EN1 (engrailed 1), MSX2, and TWIST1, all of which are present in the supraorbital regulatory center at E11.⁵⁷ Recently, individuals with severe bilateral coronal synostosis were described with heterozygous truncations or missense substitution of another transcription factor, ZIC1.^23,52 Constructs containing these ZIC1 mutations caused altered or enhanced expression of a target gene, engrailed-2, in a Xenopus embryo assay, consistent with a gain of function.⁵² Given the epistatic relationship of ZIC homologs (odd-paired in Drosophila) upstream of engrailed in both Drosophila⁶⁴ and Xenopus,^65,66 and the expression of murine Zic1 in the territory corresponding to supraorbital regulatory center,⁵²ZIC1 (MIM: 600470) mutations are likely to disrupt early lineage commitment in the coronal suture.

A complex and incompletely understood set of positive and negative feedback loops links these transcription factors to the major signaling pathways required to commit cells to an osteogenic fate, including bone morphogenetic proteins (BMPs), wingless-related family members (WNTs), and fibroblast growth factors (FGFs); there is evidence for activity of each of these pathways at this time (reviewed in Mishina and Snider⁶⁷). However, expression studies do not enable disentanglement of the exact sequence of events or signaling hierarchies involved; neither do classical knockout approaches answer the question, because the multiple roles of these pathways in organogenesis at earlier stages of embryonic development tend to lead to lethality before the cranial sutures are established.⁶⁷ The availability of suitable drivers for suture expression, together with temporal cell labeling and sorting, should enable considerable progress in teasing apart these processes over the coming decade. Within the next 24–48 hr of development (E12.0–E13.5), the expression of master regulators of osteogenic differentiation Runx2 and Sp7 is initiated.

Boundary Formation and Integrity

Coordinated with the processes described above, guidance cues are required to ensure that cells migrate along the correct path. Of particular significance for the coronal suture is that it lies at the boundary between tissues with distinct embryological origins: trigeminal neural crest (frontal bone) and cephalic mesoderm (parietal bone) (reviewed in Morriss-Kay and Wilkie⁵³). Lineage tracing using the Wnt1-Cre/R26R system to label cells of neural crest origin and their progeny shows that the coronal suture itself arises from mesoderm, and this was confirmed using a reciprocal mesodermal driver, Mesp1-Cre.^58,62 The neural crest/mesoderm boundary can be identified from E9.5, and so includes the critical period during which the definitive coronal sutures are forming. The characteristic overlapping of the frontal bone by the parietal bone, forming the oblique cross-section of the coronal suture (Figure 2C), can be understood from the expansion of the underlying cerebral hemispheres toward the hindbrain, taking the neural crest underneath the mesoderm.⁵⁸ The dura mater underlying both bones is of neural crest origin (Figure 2C). Interestingly, the other major sutures have different tissue origins and relationships: the metopic suture uniquely forms within neural crest and the sagittal and lambdoid sutures, like the coronal, have dual neural crest/mesodermal origins, but unlike the coronal, this is not the case along their entire length (Figure 2C).⁵³ For this reason, in the subset of cases of metopic synostosis that are associated with neurocognitive abnormalities and/or dysmorphic features, a disturbance of neural crest development is the likely mechanism.⁶⁸

Although it was originally thought that the coronal suture boundary did not permit migration of sutural cells either into or from the neural crest, evidence has emerged that the barrier is in fact unidirectional, in that cells of neural crest origin cannot normally cross into the suture, but the reverse does not apply.⁵⁷ This latter conclusion is consistent with the evidence from Zhao et al.⁵⁶ that the coronal suture contains stem cells; the progeny of these cells can be expected to fuel the growth of both the adjacent bones, parietal and frontal (Figures 2B and 2D). Molecules thought to be important in the maintenance of the suture boundary include the transcription factors EN1, TWIST1, and MSX2 (previous section) and members of the EPHRIN/EPH receptor and JAGGED/NOTCH families; all these molecules have been implicated in tissue boundary formation in multiple other contexts.^63,69,70

Osteogenic Proliferation and Differentiation

The importance of SHH in the earliest stages of coronal suture development was described above. Gli1, a general marker of HH signaling, is also expressed in the undifferentiated mesenchyme of established postnatal sutures;⁵⁶ however, in this context the paralogous molecule Indian hedgehog (IHH) is the instructive ligand. Mice lacking IHH (Ihh^−/−) have reduced sizes of the developing frontal and parietal bones, reduced Bmp2/Bmp4 expression, and correspondingly widened midline sutures at E15.5; this phenotype is thought to result from a differentiation defect, because proliferation was found to be unaffected.^56,71 Secretion of IHH by differentiating cells is proposed to maintain the recruitment of undifferentiated osteoprogenitors from mid-sutural mesenchyme⁵⁶ (Figure 2D). For calvarial expansion to occur, this differentiation process must be exquisitely tuned to ongoing cell proliferation.⁵⁵ Notably, the observation that craniosynostosis is frequently associated with defects in certain key cell division genes (discussed later) points to the need for rapid mitotic turnover in the cranial sutures. Twist1, expressed in mid-sutural mesenchyme in the established coronal suture (E16),⁵⁹ is directly antagonistic to RUNX2, potentially providing an important mechanism to prevent the mid-sutural cells from undergoing osteogenesis.⁷² TWIST1 also has inhibitory activity on levels of bone sialoprotein and osteocalcin, two of the downstream targets of RUNX2: this requires heterodimerization with a type I basic-helix-loop-helix partner.⁷³ In vivo, the strong epistatic genetic interaction between Twist1 and Tcf12 in the murine coronal suture shows that TCF12/HEB is the key partner protein.⁴⁸

Apart from the importance of maintaining the undifferentiated fate of mid-sutural stem cells, a second key element of cranial vault growth is provided by positive differentiation signals emanating from osteoid, the collagenous unmineralized matrix produced by osteoblasts along the expanding osteogenic fronts (reviewed in Morriss-Kay and Wilkie⁵³). Redundant signaling by several fibroblast growth factors (there is evidence to implicate particularly FGF2, FGF9, FGF10, and FGF18) is crucial to orchestrating this process.^60,74–80 The receptor Fgfr2 is expressed in the rapidly proliferating osteoprogenitor cells, whereas Fgfr1 is associated with a more differentiated state (Figure 2D); increased FGF-signaling flux drives a switch from Fgfr2 to Fgfr1 expression, associated with the onset of osteogenic differentiation.^60,61 The dura mater underlying the sutures also provides a source of growth factors including FGF2, BMP4, and TGFβ1⁵ (reviewed in Levi et al.⁸¹).

As is the case for cartilaginous bones, the transcription factor Runx2 represents a well-established positive regulator of commitment to terminal osteogenic differentiation. However, for reasons that remain poorly understood, bones undergoing intramembraneous ossification (such as the skull vault) are dependent on a higher RUNX2 dosage than those undergoing endochondral ossification, so that Runx2/RUNX2 haploinsufficiency is associated with pathologically widened cranial sutures in both mice and humans.⁸² Conversely, craniosynostosis occurs in most individuals with RUNX2 duplication.^45,46

Genetic studies have illuminated the role of another molecule in osteoblast differentiation, retinoic acid. Recessive mutations in POR (MIM: 124015), the flavoprotein that donates electrons to all microsomal P450 enzymes, cause Antley-Bixler syndrome (MIM: 201750), in which craniosynostosis is associated with abnormal steroidogenesis.⁴³ A genotype-phenotype correlation is apparent, such that individuals with skeletal malformations are usually compound heterozygous for variants encoding a combination of a missense and a null allele; moreover, no subjects harboring homozygous null mutations have been encountered, supporting the conclusion from mouse studies that complete loss of function is lethal.^83,84 Residual POR activity could differentially affect the function of the many P450 enzymes, raising the question as to which aspect of steroidogenesis is related to craniosynostosis. Homozygous mutations of CYP26B1 (MIM: 605207), which encodes one of the P450 enzymes required for degradation of retinoic acid, lead to mineralization defects of the skull and/or craniosynostosis (MIM: 614416), related to increased differentiation of osteoblasts to terminal osteocytes.⁸⁵ This evidence, together with the observation of elevated retinoic acid levels in blood or tissues of POR mutants in both mice^86,87 and humans,⁸⁸ suggests that disturbed retinoic acid metabolism is a key contributor to the pathology.⁸⁹ Whether this manifests with craniosynostosis or mineralization defects might depend on the precise extent of osteoblast/osteocyte imbalance.⁸⁵

How mechanical forces are integrated into the proliferation/differentiation response is uncertain, but primary cilia are likely to play a key role.⁹⁰ In calvarial cultures, stretched sutures released FGF2 and demonstrated an immediate increase in permeability to Ca²⁺ ions.⁹¹ Potentially linked to these observations, mice with a neural crest-driven knockout of Pkd2, encoding a calcium channel that is a key component of the ciliary mechanotransduction system, exhibited postnatal fusions of sutures in the facial bones (causing a bent snout) and features such as fractured molar roots that were suggestive of increased trauma, attributed to failure of normal feedback linking tissue strength to mechanical loading.⁹² Furthermore, mice with a mutation in Fuz, a key regulator of ciliogenesis, have coronal synostosis.⁹³ Cilia are also involved in integrating HH and WNT signaling, both of which are implicated in suture biogenesis, and mice with a Wnt1-driven neural crest conditional mutation of Kif3a have metopic synostosis.^68,94 Given the proposed role of the IHH-Gli1 feedback loop in suture maintenance (Figure 2D), it is tempting to speculate that mechanotransduction in the suture could feed directly into this pathway; evidence is lacking at present but this would be a fruitful field for future investigation.

Resorption/Homeostasis

The steady state in the mature suture represents a balance between osteogenesis and resorption, the latter of which is mediated by osteoclasts of hematopoietic origin. Many of the genes associated with generalized skeletal dysplasias but with only low frequency of craniosynostosis (Table S1) are likely to affect this balance; for example, it is not surprising that osteopetroses, characterized by excessively dense bones and deficient osteoclast function, sometimes feature craniosynostosis in the clinical presentation. A more suture-specific mechanism is likely in the case of interleukin 11 (IL11) signaling, because loss-of-function mutations in the co-receptor IL11RA are predictably associated with craniosynostosis (MIM: 614188) in humans (and fusion of facial sutures in the case of mouse Il11ra^−/− mutants).⁴¹ However, interpretation is complicated because of evidence that both osteoblast and osteoclast function are compromised. Although IL11 has a stimulatory role in osteoblasts,⁹⁵ predicting that deficient signaling would be associated with osteopenia, the opposite turns out to be the case; trabecular bone volume was increased in knockout mice and osteoclast precursors showed a cell-autonomous defect in differentiation.⁹⁶ This suggests that the osteoclast defect might predominate in the context of loss of IL11 function. Circumstantial support is provided by the observation that individuals homozygous for IL11RA (MIM: 600939) mutations often have delayed eruption of the secondary dentition, because localized resorption of the jaw bone by osteoclasts is a prerequisite for tooth eruption.⁴¹

The role and importance of cartilage intermediates in craniosynostosis is unclear. Although the skull bones are classically described as forming by intramembranous ossification, transient cartilage intermediates are well known to occur.⁶⁰ Behr et al.^97,98 have identified cartilage both in the normally fusing posterior frontal suture of wild-type mice (equivalent to the human metopic suture) and in the fusing coronal sutures of Twist1^+/− mice. Of note, Twist1 inhibits chondrogenesis and is regulated by β-catenin (canonical WNT signaling).⁹⁹ Consistent with this, the appearance of cartilage is accompanied by downregulation of canonical WNT signaling (revealed, for example, by loss of AXIN2 protein). However, these events were described during postnatal stages, long after the initiating processes leading to craniosynostosis, consistent with the possibility that these represent secondary, downstream consequences of altered signaling arising from earlier events.

Apert-Fgfr2^Ser252Trp/+ Mutant and Encroachment by Osteogenic Fronts

An exemplar for a carefully conducted disease model study is that on the Apert-Fgfr2^Ser252Trp/+ mutant.¹³¹ Although no abnormality could be detected in mutant embryos at E12.5, by E13.5 multiple changes in expression of osteogenic genes were apparent, associated with narrowing of the nascent gap at the base of the coronal suture. Although more apical parts of the coronal suture remained patent for several further embryonic days, effectively its fate was already sealed because the bones could not grow separately at their bases. Having identified this critical E13.5 time point, studies of sutures from 1 day earlier (E12.5) were conducted in an attempt to identify even earlier markers for these events. A small but consistent increase in proliferation (assessed by bromodeoxyuridine incorporation and detection of Ki67 antigen) was found, demonstrating a biological abnormality at E12.5, corresponding to the later stages in organization of the supraorbital regulatory center. By further analysis of osteoblast cultures (postnatal day 1), which showed both accelerated proliferation and differentiation, it was proposed that the primary defect involved failure of the osteogenic fronts to halt their progress across the undifferentiated mesenchyme at the base of the cranial suture. In other words, the delicate balance that maintains the existence of a population of undifferentiated cells in the mid-sutural mesenchyme in the face of rapid proliferation and differentiation to each side is disturbed, and the stem cells are overwhelmed (Figure 3A).

How do these findings relate to knowledge of craniosynostosis in humans? Embryonic day 13.0 in the mouse is equivalent to 6 weeks post-conception in terms of human skull development (8 weeks’ gestational age).⁴ At this stage it would be technically challenging to identify a causative mutation and, if a decision was made to terminate an affected pregnancy, study the embryonic material. The earliest available human pathological fetal material is from around 19 weeks,¹³² illustrating the critical role of the mouse studies in understanding the earliest stages of pathogenesis of craniosynostosis. However, Mathijssen et al.,¹³³ by calibration against the separation of the frontal and parietal bone ossification centers in a series of human fetal skulls, found that the distance between ossification centers in two skulls of neonates who had Apert syndrome corresponded to 15 weeks’ gestation, not long after coronal sutures are first apparent. This corroborates the mouse study described above,¹³¹ that in Apert syndrome there is a primary failure in the function of the coronal suture around the time of its formation. Unfortunately, most functional studies of clinical material from craniosynostosis cases utilize cells obtained months or years after disease onset, and from poorly controlled tissue sources; hence, their pathogenic interpretation should be treated with particular caution.

Further analysis of the Apert-Fgfr2^Ser252Trp/+ mutant was undertaken using tissue-specific drivers to enable selective expression of the mutation in cells of mesodermal or neural crest origin.¹³⁴ This demonstrated that craniosynostosis still occurs in the mesoderm-driven Fgfr2^Ser252Trp/+ mutant, typically with bony bridging from the frontal bone edge to a point behind the parietal bone edge. This suggests that Fgfr2^Ser252Trp/+ expressed in early mesodermal osteoprogenitors is sufficiently sensitive to allow paracrine osteogenic induction by FGFs secreted from neural crest-derived tissue.¹³⁴ By contrast, the reciprocal neural-crest-driven Fgfr2^Ser252Trp/+ mutant does not have craniosynostosis, eliminating a deterministic role in pathology for either the nascent frontal bone or the underlying dura mater. However, the onset of the earliest abnormalities in the mesoderm-conditional mutant was 2–3 days later (E15.5) than in the germline mutant, indicating that the presence of mutant cells of neural crest origin does have a synergistic effect.

The mechanism outlined above falls into the general category of altered proliferation/differentiation balance, which at this early time point severely compromises coronal suture function. As indicated in Table 2, Apert substitutions result in both increased ligand affinity and broadened specificity of mutant FGFR2 receptors. Illegitimate binding to FGF10 is strongly implicated in the coronal (and facial) suture pathology, because reduction in Fgf10 dosage in the Apert-Fgfr2^ΔIIIc/+ mouse model prevents the occurrence of craniosynostosis.⁷⁴ The selective early coronal closure, combined with rapid expansion of the brain, might act to keep open the other major sutures even in a pro-osteogenic environment, leading to the widely open metopic and sagittal sutures seen in most individuals with Apert syndrome.¹³⁵ A similar FGF10-mediated mechanism is likely to explain the selective coronal synostosis that occurs in Muenke syndrome (FGFR3-p.Pro250Arg heterozygote). Although the equivalent Fgfr3^Pro244Arg/+ mouse model does not develop craniosynostosis, impeding a direct test of this idea, characteristic hearing loss in this model is rescued by reduced Fgf10 expression.¹⁰⁵

Unlike in Apert syndrome, the activation mechanisms occurring for the FGFR2 mutations associated with Crouzon (MIM: 123500) and Pfeiffer syndromes are typically constitutive rather than ligand dependent (Table 2). Perhaps reflecting this distinct biochemical abnormality, these syndromes show a different pattern of craniosynostosis with less marked predilection for the coronal suture; instead, sagittal or multisuture synostoses are also common presentations. In severe Pfeiffer syndrome, all sutures either fail to form or fuse before birth, so the skull can grow only by remodelling; this leads to cloverleaf skull, a very severe disorder.^136,137 A milder manifestation occurs in some Crouzon syndrome and also ERF-related craniosynostosis, whereby delayed closure of all sutures is associated with a relatively normal skull shape but high risk of raised intracranial pressure.^26,138,139 Interestingly, some of the mouse mutants initially show paradoxically delayed markers of early osteogenesis, followed by later craniosynostosis,^26,126 or differences in ossification potential between bones of differing origins,¹⁴⁰ illustrating a shift of the proliferation-differentiation balance occurring over time. Neben et al.¹⁴¹ propose that this shift might reflect differential utilization of non-canonical nucleolar FGF receptor signaling.

The key effector pathway perturbed by FGF receptor mutations is the RAS-ERK pathway, as demonstrated by the marked reduction in abnormal phenotypes associated with genetic inhibition of FRS2-mediated signal transduction,¹⁴² in vivo inhibition of MEK1/2,¹⁴³ and the craniosynostosis associated with Erf mutations.²⁶ However, the complex relationship of RAS-ERK activation with craniosynostosis is highlighted by the fact that only a minority of individuals harboring RASopathy mutations develop craniosynostosis.¹⁴⁴

Twist1^+/− Mutant, Neural Crest Migration, and Boundary Integrity

The second process that has been subject to the most rigorous investigation in mouse mutants has been that of boundary formation at the coronal suture. The starting point for this work was the observation that in Saethre-Chotzen syndrome, caused by TWIST1 halpoinsufficiency, the coronal sutures are typically fused. By use of Wnt1-Cre/R26R labeling of cells of neural crest origin, Merrill et al.⁷⁰ showed that in Twist1^+/− mouse mutants, abnormal migration of cells derived from neural crest could be detected within the coronal suture and extending into the parietal bone territory, first observed at E14.5 (Figure 3B). Importantly, this cell-mixing process was shown to correlate with craniosynostosis. Given the frequent role of EPHRIN-EPH interactions in tissue boundary formation, protein levels of family members were examined and this revealed disturbed distribution of EPHRINs-A2 and -A4 and their receptor EPHA4 at E14.5.⁷⁰ Consistent with this, EphA4^−/− mutant mice were found to have craniosynostosis and Twist1^+/−EphA4^+/− embryos exhibited abnormal partitioning of DiI-labeled osteogenic cells, which occupied the coronal suture territory.⁶³ Postnatally, the number of Gli1-expressing, presumptive stem cells was reduced in the Twist1^+/− mutants.⁵⁶

Mutations in another ligand, the X-chromosome-encoded EPHRIN-B1, cause craniofrontonasal syndrome (CFNS), which is characteristically associated with coronal synostosis; orthologous Efnb1 is expressed in neural crest but not cephalic mesoderm in mice.¹⁵ Based on studies of mouse limb and craniofacial development,^115,116 there is strong evidence to implicate the female-restricted process of X inactivation in disrupting the normal coronal suture boundary, although this could not be directly proved because Efnb1^+/− mice do not develop craniosynostosis.¹²¹ This unusual mechanism has been termed cellular interference (Table 2).²⁵

Twist1^+/− Mutant and Mesoderm Specification

NOTCH signaling also contributes to boundary formation, although no genes from this pathway have been identified as being frequently mutated in clinical craniosynostosis. In mice, the earliest abnormality identified in the sutures of Twist1^+/− mutants, observed at E12.5 (the identical time point as for the Fgfr2^Ser252Trp/+ mutant reviewed above, but 48 hr prior to observation of abnormal cell mixing), was spreading of NOTCH2 protein to include the mid-sutural mesenchyme (Figure 3C).⁶⁹ Also of note, crossing of a conditional transgenic allele of Spry1 (encoding an inhibitor of FGF receptor signaling) with Twist1^+/− mutant mice largely prevented the craniosynostosis from occurring.¹⁴⁵ Together, these observations show that TWIST1 lies primarily upstream of FGF receptor signaling (as in nematode worms and fruit flies)^146,147 and implicate TWIST1 in maintaining the identity of sutural cells, which might be important to prevent later cell mixing. Importantly, however, the conditional Spry1 experiment demonstrates that these abnormalities can be overridden in an environment that reduces the overall osteogenic drive. Hence the relative contributions that loss of boundary integrity versus increased osteogenic differentiation make to the craniosynostosis associated with TWIST1 mutation are difficult to separate.