COUP-TFII is required for morphogenesis of the neural crest-derived tympanic ring

Chicken Ovalbumin Upstream Promoter-Transcription Factor II (COUP-TFII) plays pivotal roles in cell growth, cell differentiation, and cell fate determination. Although genome-wide studies have identified COUP-TFII binding on gene sets mainly involved in neural crest cell (NCC) development and craniofacial morphogenesis, the direct functional connection between COUP-TFII and NCCs in vivo has not been well characterized. In this study, we show that COUP-TFII is expressed in the subpopulation of NCCs and its derivatives, and targeted ablation of COUP-TFII in mouse NCCs results in markedly shortened and bifurcated tympanic rings, which in turn disturb the caudal direction of external acoustic meatus invagination. However, formation of the manubrium of the malleus (MM) in Wnt1-Cre/+;COUP-TFII flox/flox mice is not perturbed, suggesting that the rostral half of the tympanic ring is sufficient to support proper MM development. Interestingly, we found that loss of COUP-TFII up-regulates Sox9 in the tympanic ring primordium and affects the distribution of preosteoblasts before mesenchymal condensation. Together, our results demonstrate that COUP-TFII plays an essential role in regulating the patterning of the NCC-derived tympanic ring.


COUP-TFII is expressed in the NCC-derived subpopulation.
To explore whether COUP-TFII plays a physiological role in regulating development of the NCC lineage, we examined the spatiotemporal expression of COUP-TFII. First, we monitored COUP-TFII expression by performing whole-mount X-Gal staining of COUP-TFII Z/+ embryos 31 , in which the LacZ reporter expression is driven by the COUP-TFII promoter. In E10.5 COUP-TFII Z/+ embryos, X-Gal staining revealed relatively high levels of COUP-TFII expression in several embryonic regions, including the caudal telencephalon, posterior two-thirds of the diencephalon, diencephalon-mesencephalon boundary, eye, cephalic flexure, hindbrain, and craniofacial ganglia (Fig. 1a). This expression pattern recapitulates the expression profile of endogenous COUP-TFII 32,33 . We also observed moderate X-Gal-positive signals in pharyngeal arches (PAs), especially in the posterior part of PA1 and the anterior part of PA2 (Fig. 1a). Dual immunofluorescence staining of E10.5 Wnt1-Cre/+;R26R mTmG/+ embryonic sections further showed COUP-TFII expression in the cephalic mesenchyme and pharyngeal endoderm as well as in the pharyngeal mesenchyme subpopulation ( Fig. 1b and e). This COUP-TFII expression in GFP-positive mesenchymal cells of the PAs confirmed that COUP-TFII is expressed in specific tissues of NCC origin ( Fig. 1b-g). Next, we performed whole-mount X-Gal staining of COUP-TFII Z/+ embryos to further characterize the expression of COUP-TFII during craniofacial development. We observed spatiotemporal changes in the distribution of strongly X-Gal-stained cells as development proceeded (Fig. 1h-l), indicating that COUP-TFII exhibits a dynamic expression pattern during craniofacial morphogenesis. Also, we noticed that strongly X-Gal-stained cells located within PAs were gradually confined to the corner of the mouth and showed a C-shaped ring rostral to the ear pinna at E13.5 (Fig. 1l).
Next, to characterize the COUP-TFII expression profile in NCCs and their derivatives, we performed lineage-tracing experiments using Wnt1-Cre/+;COUP-TFII flox/+ mice, in which LacZ reporter expression is under control of the COUP-TFII promoter upon Cre-mediated COUP-TFII deletion (Fig. 1h'-l'). The dynamic distribution of strongly X-Gal-stained cells in Wnt1-Cre/+;COUP-TFII flox/+ embryos was comparable to that of COUP-TFII Z/+ embryos from E10.5 to E13.5. These results show that craniofacial cells with high COUP-TFII expression mainly reside within the NCC lineage during development, and therefore COUP-TFII may play an important role in guiding the development of a specific NCC subpopulation.
Loss of COUP-TFII leads to malformation of the tympanic ring and external acoustic meatus (EAM). Cranial NCCs and their derivatives can differentiate not only into cranial ganglia and connective tissues but also into bone and cartilage in the head and neck 1 . Therefore, we stained cranial skeletal preparations with Alcian Blue and Alizarin Red to examine skull architecture ( Fig. 2a-b). We observed that the tympanic ring, which develops to provide physical support for the tympanic membrane, was slightly thicker in E15.5 Wnt1-Cre/+;COUP-TFII flox/flox mutants (Fig. 2c') than in littermate controls (Fig. 2c). From E16.5 onward, the normal C-shaped tympanic ring gradually developed in control embryos ( Fig. 2d-g). In contrast, the tympanic ring in Wnt1-Cre/+;COUP-TFII flox/flox mutants was noticeably shortened. As development proceeded, abnormal bifurcation of the tympanic ring was also seen in Wnt1-Cre/+;COUP-TFII flox/flox mutants with full penetrance (Fig. 2d'-g'). Comparative histological analyses further confirmed the shortening and abnormal bifurcation of the tympanic ring in Wnt1-Cre/+;COUP-TFII flox/flox mutants ( Fig. 2i-k').
which was a likely consequence of the premature ending and bifurcation of the mutant tympanic ring. Proper EAM invagination is crucial for coordinating the development and positioning of the manubrium of the malleus (MM) 34 . Therefore, it was surprising that the anomalies in the tympanic ring and EAM invagination of Wnt1-Cre/+;COUP-TFII flox/flox mutants did not perturb the coordinated development of the MM (compare Fig. 3k' with k), indicating that the rostral half of the tympanic ring is sufficient to guide MM formation.

Loss of COUP-TFII up-regulates Sox9 in the tympanic ring primordium. Sox9 expression is
progressively down-regulated in the osteogenic cell compartment during intramembranous ossification, indicating that Sox9 may be involved in incipient intramembranous bone formation 35,36 . However, the role of Sox9 in the development of intramembranous bones remains poorly understood. In mice, the primordium for the tympanic ring becomes apparent by E13.5 as a condensation ventral to the first pharyngeal cleft (a-j') Serial frontal head sections through the middle ear region from E16.5 controls (n = 5) (a-j) and Wnt1-Cre/+;COUP-TFII flox/flox mutants (n = 4) (a'-j') were stained with the epithelial cell marker K5 (green) and osteogenic lineage marker Runx2 (red) to identify the spatial location of the EAM (red arrows) and tympanic ring, respectively. Compared with controls, Wnt1-Cre/+;COUP-TFII flox/flox mutants showed defective morphogenesis of EAM invagination (yellow asterisk in c and c'). In both control and mutant mice, in the presence of the tympanic ring, the extremity of the EAM pointed to the position of the tympanic ring. However, the tympanic ring ended rostrally so that the direction of EAM invagination was misaligned in the caudal region (i' and j', yellow arrow) of Wnt1-Cre/+;COUP-TFII flox/flox mutants. (k-k') Alcian Blue and Alizarin Redstained middle ear. MM structure was comparable between controls (n = 3) and Wnt1-Cre/+;COUP-TFII flox/ flox mutants (n = 3). TR, tympanic ring; i, incus; m, malleus; MM, manubrium of the malleus; s, stapes. Scale bar = 200 μm in a-j', 500 μm in k and k'.
At a later stage (E16.5) when the bony structure of the tympanic ring can be clearly identified by histological analysis, the level of Sox9 expression was lower in the bony part of the tympanic ring than in the surrounding mesenchyme in both control and mutant embryos (Fig. 5a-j'), consistent with previous studies reporting the down-regulation of Sox9 expression during intramembranous ossification 35,36 . We also noticed that high levels of Sox9 were detected in the mesenchyme around the caudal end of developing tympanic ring in both control and COUP-TFII mutant embryos (control - Fig. 5i and j; mutant -Fig. 5h' and i'). Nevertheless, the Sox9 expression in Wnt1-Cre/+;COUP-TFII flox/flox mutants was consistently higher in the surrounding mesenchyme of the rostral part of tympanic ring than that of the control mice (compare Fig. 5f-h with f '-g'). The master regulator of osteogenesis, Runx2, was highly expressed in the bony part of the tympanic ring, and more moderate levels were observed in Sox9-positive surrounding mesenchymal cells in both control and Wnt1-Cre/+;COUP-TFII flox/ flox mutant mice (Fig. 5k-o'). These results suggest that COUP-TFII may coordinate the tympanic ring development by inhibiting Sox9 expression in the osteogenic mesenchyme. Given that Sox9 is required for determining chondrogenic cell lineage and is sufficient for cartilage formation 8,14,39,40 , we next investigated whether malformations of the tympanic ring in Wnt1-Cre/+;COUP-TFII flox/flox mutants result from inaccurate cell fate decision (i.e., chondrogenic cell fate) due to abnormal up-regulation of Sox9. However, skeletal preparations revealed that corresponding regions of the Alizarin Red-stained tympanic ring showed no Alcian Blue (i.e., cartilage) staining in either E18.5 controls or Wnt1-Cre/+;COUP-TFII flox/flox mutants (Fig. 5p-q'). Thus, aberrant Sox9 up-regulation did not result in the diversion of osteogenic mesenchymal cells toward cartilage differentiation. Collectively, our findings indicate that COUP-TFII-null mesenchymal cells are able to condense, differentiate into osteoblasts, and form a truncated tympanic ring in Wnt1-Cre/+;COUP-TFII flox/flox mutants through intramembranous ossification.

COUP-TFII governs the distribution of preosteoblasts during tympanic ring development. Mesenchymal condensation is a crucial stage for skeletal patterning during bone development.
Therefore, we analyzed mesenchymal condensation to verify whether the COUP-TFII-null mesenchyme exhibits impaired skeletal patterning for the future tympanic ring. Tenascin-C, a glycoprotein of extracellular matrix components, is involved in establishing the condensation boundary of the osteogenic population from non-skeletogenic mesenchymes [41][42][43] . At E13.5, tenascin-C expression was detected in the condensed mesenchyme as evidenced by cell aggregation but not in the non-condensed mesenchyme of the developing tympanic ring in both control and mutant embryos (Fig. 6a-e'). Loss of COUP-TFII in NCCs did not affect the level of tenascin-C expression, but tenascin-C-positive condensed mesenchyme within the developing tympanic ring of Wnt1-Cre/+;COUP-TFII flox/flox mutants did not extend caudally as far as the condensed mesenchyme did in control mice (compare Fig. 6c' with c). This result revealed that the abnormal shortening of condensed mesenchymal structure contributes to anomalies of the tympanic ring in Wnt1-Cre/+;COUP-TFII flox/flox mutants. Similarly, expression of the osteogenic markers Runx2 and Osterix (Osx) at the single-cell level were comparable between the corresponding regions within the developing tympanic rings of control and COUP-TFII mutant embryos. However, both Runx2-and Osx-expressing mesenchymal cells were not detected in the caudal region of Wnt1-Cre/+;COUP-TFII flox/flox mutants, while such cells could easily be seen in the counterpart of control mice (Fig. 6f-o'). According to previous studies, specified preosteoblasts express Runx2, and at a later stage, preosteoblast committed to an osteoblast lineage express both Runx2 and Osx 44 . Therefore, these results revealed that, instead of extending into more caudal region as in control mice, the distributions of both specified and committed preostroblasts were rostrally restricted in Wnt1-Cre/+;COUP-TFII flox/flox mutants. Interestingly, we noted that Runx2-and Osx-positive signals were detected not only in condensed mesenchyme but also in non-condensed mesenchyme at the caudal part of the developing tympanic ring (Fig. 6a-o'). Thus, the distributions of Runx2and Osx-expressing cells were much broader than that of tenascin-C-expressing cells, which indicates that, before mesenchymal condensation, these Runx2-and Osx-expressing preosteoblasts already exist during tympanic ring development. Taken together, our results suggest that loss of COUP-TFII results in aberrant up-regulation of Sox9 in the subpopulation of NCC-derived mesenchyme and affects the distribution of preosteoblasts, which in turn leads to improper pattern of mesenchymal condensation and consequently truncated tympanic ring formation.

Discussion
The work presented here unveils the differential expression of COUP-TFII and its biological function in cranial NCCs and their derivatives in vivo. We showed that targeted inactivation of COUP-TFII in the NCC lineage leads to tympanic ring abnormalities. Genetic evidence showed that COUP-TFII deficiency results in the aberrant expression/regulation of Sox9 in the primordium of the tympanic ring and affects the distribution of preosteoblasts before mesenchymal condensation.
In vivo fate mapping and comparative histological analyses showed that loss of COUP-TFII affects neither NCC migration from neural folds to the ear region nor survival during this period. Before mesenchymal condensation, we detected COUP-TFII loss-of-function and increased Sox9 protein levels in the mesenchyme that presumably gives rise to the tympanic ring in E13.5 Wnt1-Cre/+;COUP-TFII flox/flox mutants, suggesting that COUP-TFII may negatively regulate Sox9 expression in the osteogenic mesenchyme. This result stands in contrast to the previous finding showing that COUP-TFII is recruited to the Sox9 regulatory region by Sp1 to activate  (a-e'). At E16.5, the bony structure of the tympanic ring (indicated by red arrow) can be clearly identified from surrounding mesenchyme. Mesenchymal Sox9 expression (red, indicated by yellow arrow) was persistently up-regulated in E16.5 Wnt1-Cre/+;COUP-TFII flox/flox mutants (n = 5) (f '-j'), and its spatial distribution was also different from that of controls (n = 6) (f-j). Runx2 was expressed in both the bony part and the surrounding mesenchyme of the tympanic ring (k-o'). Histological analysis revealed that Sox9 expression in the surrounding mesenchyme was localized to the corresponding lower Runx2-expressing osteogenic cells (white arrow) in both control and mutant embryos (n = 2). (p-q') Lateral view of E18.5 middle ear. Alcian Blue and Alizarin Red staining of the middle ear revealed that the ossified tympanic ring was shorter in Wnt1-Cre/+;COUP-TFII flox/flox mutants (n = 13) (p') compared with the expression of Sox9, leading to the chondrogenic commitment of mesenchymal precursors 27 . The underlying basis for the inverse relationship between COUP-TFII and Sox9 during osteogenesis versus chondrogenesis is currently not clear. One possibility is that COUP-TFII interacts with other transcription factors to cooperatively inhibit Sox9 expression, whereas another possibility is that Sox9 is an indirect target of COUP-TFII-mediated repression. Future studies are needed to delineate how COUP-TFII regulates Sox9 expression in osteogenesis. In addition, COUP-TFII physically interacts with Runx2 and interferes with its binding on the target promoter, which inhibits the transcriptional activity, but not the level of Runx2, in directed osteoblast differentiation 27,38 . However, we found that loss of COUP-TFII does not affect the expression level of Runx2 or its direct targets, Osx and osteocalcin (OC) (Supplementary Fig. S3), suggesting that COUP-TFII does not regulate Runx2 transactivity during osteoblast differentiation in the developing tympanic ring.
We detected Runx2-and Osx-positive signals, which indicate commitment of preosteoblasts to an osteoblast lineage, in dispersed mesenchyme, as well as in condensed mesenchyme (i.e., tenascin-C-expressing cells or the cell aggregation in tissues), in the developing tympanic ring, which reveals that preosteoblast commitment precedes mesenchymal condensation during tympanic ring development. A similar process has been observed during the development of the chicken mandible 45 , which also occurs through intramembranous ossification. Therefore, our findings provide further evidence supporting that osteogenesis is different from chondrogenesis, in which mesenchymal condensation triggers prechondroblast commitment 46,47 .
In skeletal development, Sox9 and Runx2 are the major transcription factors for chondrogenesis and osteogenesis, respectively. The spatiotemporal overlap of Sox9 and Runx2 expression unveils lineage-specific roles in orchestrating transcription of downstream targets during bone formation. Loss of Sox9 in NCCs and growth plate result in ectopic expression of osteoblast marker genes, such as Runx2, Osx and Col1a1, in the presumptive cartilage existing in wild-type embryos and differentiated growth plate chondrocytes, respectively 8,9 . In addition to directly activating the expression of major cartilage-specific extracellular matrix genes, Sox9 is also capable of suppressing Runx2 expression to specify and maintain the chondrogenic lineage fate [48][49][50] . By contrast, the role of Sox9 in osteogenesis is not well understood. Sox9 expression is gradually down-regulated during intramembranous bone formation 35,36 . In cranial skeletogenesis, misexpression of Sox9 results in down-regulation of Runx2 in chondrocytes of the ceratobranchial cartilage, whereas the expression of Runx2 in osteoblasts and osteogenic mesenchymal condensation are not disrupted in the intramembranous surangular bone 14 . Further mechanistic investigation showed that Sox9 suppresses the transactivity of Runx2 in the established osteoblast lineage, which hinders osteoblast differentiation 51 . Our study provides additional in vivo evidence that the regulation of Runx2 by Sox9 during osteogenesis is different from that during chondrogenesis. Despite the shortened primordium of the tympanic ring in Wnt1-Cre/+;COUP-TFII flox/flox mutants, abnormal Sox9 up-regulation did not affect the expression of Runx2, and Runx2-positive cells still resided within their anatomical locations, consistent with previous observations in surangular bone 14 . Notably, the expression levels of Osx and OC, the direct targets of Runx2, were not altered, indicating that abnormally increased Sox9 in the Wnt1-Cre/+;COUP-TFII flox/flox tympanic ring primordium does not inhibit the transactivity of Runx2. However, we did not rule out the possibility that the level of Sox9 might not have been sufficient to inhibit Runx2 transactivity in mutants. The inconsistencies among studies may also be due to different experimental approaches under varying physiological and cellular contexts.
Morphogenesis of the tympanic ring, EAM, and MM are spatially and temporally associated 37,52 . This close relationship is also observed through clinical experience with humans 53,54 . Analyses of genetically modified (Gsc −/− and Prx1 −/− mutants) and retinoic acid-treated mouse embryos support the deduction that the tympanic ring is essential for inducing the invagination of the first pharyngeal cleft to form the EAM, which then provides signals that act on the underling mesenchymal cells to coordinate proper MM development 34,52,[55][56][57][58] . It is noteworthy that the severity of tympanic ring anomaly is tightly associated with the level of MM deformity. Concurrent disappearance of the tympanic ring and MM is also found in endothelin A receptor-deficient mice 59,60 . Both Gas1-deficient and Tshz1-deficient mice have shorter and thicker tympanic rings accompanied by the absence of a MM 61,62 . Bapx1 −/− embryos show relatively normal MM formation but exhibit hypoplasia (i.e., thinness) of the tympanic ring only in the anterior/rostral part 63 . Although EAM development has not been clearly described in every aforementioned mutant, studies consistently indicate that the MM forms in a coordinated manner alongside tympanic ring development. In this study, we found that the loss of COUP-TFII in the NCC lineage results in a shortened and thicker tympanic ring. Although this truncated tympanic ring was able to induce and guide the morphogenesis of EAM invagination, the extremity of caudal EAM invagination went astray. However, the MM was well developed and properly positioned in Wnt1-Cre/+;COUP-TFII flox/flox mutants, suggesting that the rostral half of the tympanic ring-rather than the entire structure-is sufficient to coordinate the development and positioning of the MM. Nevertheless, this hypothesis is inconsistent with phenotypes observed in Tshz1 −/− mutants. The morphologies of the shortened tympanic rings in Wnt1-Cre/+;COUP-TFII flox/ flox and Tshz1 −/− mutants are similar to each other and to that of the rostral half of the control tympanic ring; however, MM development differs between these two mutants. Coré et al. showed that Tshz1 is expressed in the mesenchyme surrounding the malleal primordium, implying that Tshz1 may play a cell-autonomous role in MM development. By contrast, down-regulation of COUP-TFII in the Sox9-positive developing malleus and neighboring mesenchyme was noted in control embryos from E12.5 ( Supplementary Fig. S4). Therefore, the controls (n = 11) (p). Single Alcian Blue staining showed no cartilage formation in the tympanic ring of controls (n = 8) (q) or Wnt1-Cre/+;COUP-TFII flox/flox mutants (n = 5) (q'). The red arrow denotes the position of the ossified tympanic ring, and the yellow arrow indicates the endpoint of the tympanic ring. Scale bar = 100 μm in a-j', 500 μm in k-l'. discrepancy in MM development between Wnt1-Cre/+;COUP-TFII flox/flox and Tshz1 −/− mutants could result from a difference between the cell-autonomous roles of COUP-TFII and Tshz1.
In summary, our study establishes that COUP-TFII is an essential transcription factor in NCCs and their derivatives for tympanic ring development in vivo. Our results provide new insights into the differential roles of COUP-TFII in fine-tuning the expression of Sox9 and regulating the distribution of preosteoblasts at the beginning of intramembranous bone development. As previous studies report that COUP-TFII can specify the lineage commitment of mesenchymal stem/stromal cells (MSCs), prospective studies aimed at understanding the precise spatiotemporal control and plasticity of MSCs in vivo will be of great benefit to future MSCs-based therapy. Figure 6. The distribution of preosteoblasts is rostrally restricted in the ear region of E13.5 Wnt1-Cre/+;COUP-TFII flox/flox mutants. Adjacent frontal sections, through the otic region from late E13.5 embryos (n = 5 for control and n = 3 for mutant), were immunostained with antibodies against tenascin-C, Runx2, and Osx. (a-e') At late E13.5, condensed mesenchymal cells in the rostral part of the tympanic ring primordium exhibited tenascin-C expression (indicated by red arrow) in controls (a-e) and Wnt1-Cre/+;COUP-TFII flox/flox mutants (a'-e'). However, examination of the serial sections through the otic region revealed that tenascin-C-expressing condensed mesenchyme was detected in the first three sections (a-c) of control mice, while that was detected only in the rostralmost section (a') of Wnt1-Cre/+;COUP-TFII flox/flox mutants. (f-o') Both Runx2-(f-j', indicated by yellow arrow) and Osx-expressing (k-o', indicated by white arrow) mesenchymal cells were not detected in the caudal region of Wnt1-Cre /+ ;COUP-TFII flox/flox mutants, while such cells could easily be seen in the counterpart of control mice. In addition to condensed mesenchyme, Runx2 and Osx were also expressed in non-condensed mesenchyme in the caudal part of the tympanic ring primordium, which showed no tenascin-C expression. Scale bar = 100 μm.