Altered FGF Signaling Pathways Impair Cell Proliferation and Elevation of Palate Shelves

In palatogenesis, palatal shelves are patterned along the mediolateral axis as well as the anteroposterior axis before the onset of palatal fusion. Fgf10 specifically expressed in lateral mesenchyme of palate maintains Shh transcription in lateral epithelium, while Fgf7 activated in medial mesenchyme by Dlx5, suppressed the expansion of Shh expression to medial epithelium. How FGF signaling pathways regulate the cell behaviors of developing palate remains elusive. In our study, we found that when Fgf8 is ectopically expressed in the embryonic palatal mesenchyme, the elevation of palatal shelves is impaired and the posterior palatal shelves are enlarged, especially in the medial side. The palatal deformity results from the drastic increase of cell proliferation in posterior mesenchyme and decrease of cell proliferation in epithelium. The expression of mesenchymal Fgf10 and epithelial Shh in the lateral palate, as well as the Dlx5 and Fgf7 transcription in the medial mesenchyme are all interrupted, indicating that the epithelial-mesenchymal interactions during palatogenesis are disrupted by the ectopic activation of mesenchymal Fgf8. Besides the altered Fgf7, Fgf10, Dlx5 and Shh expression pattern, the reduced Osr2 expression domain in the lateral mesenchyme also suggests an impaired mediolateral patterning of posterior palate. Moreover, the ectopic Fgf8 expression up-regulates pJak1 throughout the palatal mesenchyme and pErk in the medial mesenchyme, but down-regulates pJak2 in the epithelium, suggesting that during normal palatogenesis, the medial mesenchymal cell proliferation is stimulated by FGF/Erk pathway, while the epithelial cell proliferation is maintained through FGF/Jak2 pathway.

Introduction palate development have been reported, their exact intracellular mechanism pathways remain elusive. In our study, we conditionally activated Fgf8 in the developing palatal mesenchyme by using Rosa26R-Fgf8 and Osr2-cre KI alleles [20,21]. We found that the ectopically expressed Fgf8 disturbed the expression of Fgf7 and Fgf10, which provided us a useful tool to explore the role of FGF signaling in the palate development.

Ethics Statement
All animals used in this study were bred and treated in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health and in compliance with animal protocol approved by the Committee on the Ethics of Animal Experiments at Tulane University (Protocol Number: 0184-R4).

Mouse lines
The Osr2-cre KI mouse was obtained from Dr. Jiang's lab at Cincinnati Children's Hospital Medical Center [20]. The Rosa26R-Fgf8 line, generated by Dr. Ma at Washington University School of Medicine, has been described previously [21]. We crossed the Osr2-cre KI mouse with Rosa26R-Fgf8 mouse to get the E12.5-18.5 Osr2-cre KI ;Rosa26R-Fgf8 mutants. The K14-cre and Shh-cre lines, previously applied in craniofacial research, were purchased from Jackson Laboratories [22,23]. For the analysis of Cre activity, the Osr2-cre KI or Shh-cre mouse was mated with Rosa26R-LacZ mouse to get the embryos from E11.5-E13.5 [24]. The mated mice were checked for vaginal plugs every morning and the noon of plug day was recorded as 0.5 day of pregnancy (E0.5).

Cell proliferation assay
Cell proliferation in the E13.5 embryos was performed by the detection of incorporated BrdU on 10 μm paraffin-embedded sections as described in the previous report [27]. The proliferation index of palatal mesenchyme was defined as the number of BrdU-positive nuclei in a fixed area of the medial or lateral palatal shelves, while the proliferation index of palatal epithelium was calculated as the ratio of the BrdU positive nuclei in 100 counted nuclei. Three samples and at least four sections of each samples were counted and statistically analyzed by student ttest.

In situ hybridization analysis
For section in situ hybridization, samples were harvested in ice-cold PBS and fixed in 4% PFA/ PBS at 4°C overnight prior to dehydration through graded ethanol and embedding in paraffin. Samples were sectioned at 10 μm and subjected to non-radioactive in situ hybridization as described previously [11]. The validity of the anti-sense probes for Dlx5, Fgf7, Fgf10, Osr2 and Wnt5a has been reported previously [4,7,16]. The whole mount in situ hybridization with anti-sense Shh probe was performed as described previously [28]. Both the section and whole mount in situ hybridization were repeated in the samples collected from three litters.

Immunohistochemistry
For immunohistochemical staining, samples were fixed in Z-fix (Anatech) at room temperature for 2 hours, dehydrated with gradient alcohol, cleared with xylene, embedded in paraffin, and sectioned at 10 μm. The rabbit anti-mouse polyclonal antibodies against pAkt, pErk1/2, pJak1, pJak2 and pPLCγ1 were all produced by Santa Cruz Biotechnology, Inc. The horseradish peroxidase-coupled goat anti-rabbit IgG was purchased from Vector Laboratories, Inc. Immunohistochemical staining was conducted according to the manufacturer's instruction using above antibodies. Each antibody was applied on at least three sections from different samples.

Results
Cre activity was specifically distributed in the palatal mesenchyme of Osr2-cre KI mouse Previous studies have reported that the expression of Osr2 gene in the developing palate was specifically restricted in the mesenchyme with an increasing gradient from medial to lateral side [15,29,30]. The Cre following IRES sequence targeted into the 3'-untranslated region (UTR) of Osr2 locus (Osr2 IresCre ) has been applied for the study of mouse palatal development as a valid tool [7,31]. However, the Cre activity of Osr2 IresCre mouse was detected all-over the palatal mesenchyme from E11.5 without a medial-lateral gradient [29,32]. Moreover, a subset of Osr2 IresCre mouse embryos exhibited an ectopic Cre activity, which might affect the accuracy of the interpretation of phenotypes [29]. To circumvent the influence of the ectopic Cre activity, we adopted the Osr2-cre KI line replacing the Osr2 coding sequence with Cre cDNA in our following study [20]. When we crossed the Osr2-cre KI line with the Rosa26R-LacZ reporter line, Cre-mediated activation of LacZ expression showed that at E11.5, the Cre activity only could be detected in the mesenchyme connecting the mandible and maxillary processes, but excluded from anterior and posterior palate shelves (Fig 1A and 1B). In the E12.5 palatal shelves, Cre activity was distributed throughout the anterior and posterior mesenchyme without a medial to lateral gradient (Fig 1C and 1D). In the E13.5 palatal shelves, although the LacZ pattern showed an all-over distribution of Cre activity in the anterior palatal mesenchyme, but in the posterior palate, Cre activity was only detected in the medial and lateral mesenchyme with the same intensity, and absent from the central mesenchyme (Fig 1E and 1F).
Over-expression of Fgf8 by Osr2-Cre KI results in complete cleft palate To investigate the influence of over-dosage FGF8 on the palate development, we crossed the Osr2-Cre KI line with the Rosa26R-Fgf8 line [21]. The Osr2-Cre KI ;Rosa26R-Fgf8 embryos displayed normal palatal outgrowth up to E12.5 (data not shown). At E13.5, the anterior palatal shelves of Osr2-Cre KI ; Rosa26R-Fgf8 embryos were comparable to the wild type littermates (Fig 2A and 2B), but the posterior palatal shelves were enlarged, especially in the medial side (Fig 2C and 2D). At E14.5, the anterior palatal shelves had elevated to horizontal position in the wild type embryos, while still being kept in the vertical position in the mutant littermates (Fig 2E and 2F). The E14.5 wild type posterior palatal shelves had contacted with each other and formed midline epithelial seam after elevation ( Fig 2G). Whereas, the posterior shelves of the E14.5 Osr2-Cre KI ;Rosa26R-Fgf8 embryo failed to elevate (Fig 2H). The secondary palate of wild type embryos completely fused and separated oral from nasal cavity in both the anterior and posterior halves at E18.5 (Fig 2I and 2K), but the enlarged mutant palatal shelves only elevated slightly in the anterior, and stayed vertical in the posterior half (Fig 2J and 2L), resulting in a complete cleft palate which was lethal to the newborn mice.  Abnormal cell proliferation in Osr2-Cre KI ; Rosa26R-Fgf8 palatal shelves To explore how the constitutively activated Fgf8 enlarged the posterior palatal shelves, we compared the cell proliferation by BrdU labelling test between wild type and Osr2-Cre KI ; Rosa26R-Fgf8 posterior palatal shelves. Compared with their wild type littermates, the amount of BrdU labelled cells of E13.5 Osr2-Cre KI ; Rosa26R-Fgf8 posterior palatal shelves increased significantly in both medial and lateral mesenchyme (Fig 3A, 3B and 3E), but decreased drastically in the epithelium (from 19.364 to 13.222; Fig 3C, 3D and 3F). The number of BrdU positive cells of the medial mesenchyme (64.667) was higher than that of the lateral mesenchyme (61.556) in the E13.5 Osr2-Cre KI ; Rosa26R-Fgf8 palatal shelves (Fig 3E). In contrast, the proliferating medial mesenchymal cells (48.455) were less than those of lateral side (50.455) in the wild type palatal shelves ( Fig 3E). However, these differences had no statistical significance ( Fig 3E). The BrdU-labelling examination indicated that the enlarged palatal shelves of Osr2--Cre KI ; Rosa26R-Fgf8 mice resulted from the increased mesenchymal cell proliferation stimulated by the mesenchyme derived FGF8. It also implied that the palatal mesenchyme derived FGF8 had an inhibitory effect on the cell proliferation of palatal epithelium.
FGF8 had no direct effects on the Shh expression and cell proliferation in palatal epithelium SHH directly stimulates the mesenchymal cell proliferation of the developing palatal shelves and indirectly promotes epithelial cell proliferation by activating mesenchymal Fgf10 [5,7]. Whole mount in situ hybridization with Shh probe was performed to examine if FGF8 up-regulated the Shh transcription to stimulate mesenchymal cell proliferation. Surprisingly, when Shh was activated in the most anterior 3 rugae of the WT palatal shelves at E13.5, only a slight expression of Shh was detected in the most anterior ruga of the Osr2-Cre KI ; Rosa26R-Fgf8 palatal shelves (Fig 4A and 4B). At E14.5, when Shh expression was detected in all 8 rugae of the fused wild type palate, there were only two short Shh-expressing rugae found in the mutant littermates (Fig 4C and 4D). These phenomena suggested that the epithelial Shh transcription was inhibited by ectopic FGF8. To examine if FGF8 exerted a direct inhibition on Shh transcription, the Rosa26R-Fgf8 allele was constitutively activated by K14-Cre and Shh-Cre. Intriguingly, cleft palate was only detected in the Shh-Cre;Rosa26R-Fgf8embryos with an unaffected Shh expression in palatal shelves (Fig 4E, 4F and 4G), and not detected in the K14-Cre; Rosa26R-Fgf8 mice at all (data not shown). Furthermore, although the cell proliferation exhibited a marked decrease in both the E13.5 medial and lateral Shh-Cre;Rosa26R-Fgf8 palatal mesenchyme compared with their wild type controls (Fig 4H and 4J), there was no significant difference in the palatal epithelia between the E13.5 WT and Shh-Cre;Rosa26R-Fgf8 littermates (Fig 4H, 4I and 4K). Therefore, FGF8 does not directly suppress Shh expression and cell proliferation in palatal epithelium. Taken together, the disrupted Shh expression in the Osr2-Cre KI ; Rosa26R-Fgf8 palatal epithelium was supposed to be attributed to altered gene expression in palatal mesenchyme, instead of the direct inhibition by FGF8.
The altered gene expression in the Osr2-Cre KI ;Rosa26R-Fgf8 palatal mesenchyme Then, we examined the transcription of a series of mesenchymal markers by in situ hybridization. Consistent with the loss of Shh expression in epithelial rugae, the expression of Fgf10 diminished from the lateral mesenchyme in the E13.5 Osr2-Cre KI ;Rosa26R-Fgf8 palatal shelves (Fig 5A and 5B), indicating that the increased mesenchymal cell proliferation was attributed to FGF8, instead of FGF10 as in normal palatal development. Identical to medial Fgf10  Fig  5C and 5D). As the upstream activator of Fgf7, the transcription of Dlx5 was suppressed in the E13.5 Osr2-Cre KI ;Rosa26R-Fgf8 medial palatal mesenchyme (Fig 5E and 5F). Another transcription factor, Osr2 which was reported to be dominantly expressed in the lateral mesenchyme [15,33], displayed a reduced expression domain along the Osr2-Cre KI ;Rosa26R-Fgf8 lateral palatal mesenchyme (Fig 5G and 5H). However, Wnt5a, a growth factor which was also expressed with a priority in lateral palate [4,34], showed no difference between wild type and mutant palatal shelves (Fig 5I and 5L). These results indicated that the ectopically activated Fgf8 expression disrupted the Shh-Fgf10 feedback loop required for cell proliferation in normal palate development, and stimulated the proliferation of medial and lateral palatal mesenchymal cells indiscriminately.
The altered FGF signaling pathways inthe Osr2-Cre KI ;Rosa26R-Fgf8 palatal shelves To reveal how the ectopic FGF8 exerted the function on the developing palatal mesenchyme, antibodies against pivotal transducers of FGF signaling pathways were applied in immunohistochemistry. There are four intracellular pathways mediating the biological function of FGF ligands, including PI3K/Akt pathway, Erk pathway, Jak/Stat pathway and PLCγ pathway [17]. There was no difference detected in the four signaling pathways between the E13.5 wild type and Osr2-Cre KI ; Rosa26R-Fgf8 anterior palatal shelves (data not shown). However, an elevation of Phosphorylated Erk (pErk) was detected in the medial side of Osr2-Cre KI ; Rosa26R-Fgf8 posterior palatal shelves, but not in the wild type posterior palatal shelves (Fig 6A, 6A', 6B and 6B'). For the Jak-Stat pathway, the level of phosphorylated Jak1 (pJak1) was drastically up-regulated in the mutant palatal mesenchyme compared with the wild type control (Fig 6C, 6C', 6D and 6D'), while the phosphorylated Jak2 (pJak2), which is restricted in the wild type lateral palatal mesenchyme, diminished from the mutant palatal mesenchyme, but was drastically upregulated in the mutant epithelium ( Fig 6E, 6E', 6F and 6F'). In contrast, the level of the phosphorylated PLCγ1 (pPLCγ1) and the phosphorylated Akt (pAkt) exhibited no discrepancy in the palatal mesenchyme between the Osr2-Cre KI ; Rosa26R-Fgf8 mice and their wild type littermates (Fig 6G-6J). These alterations on FGF signaling pathways suggested that the ectopically activated FGF8 stimulated mesenchymal cell proliferation indiscriminately in all of the mesenchyme through pJak1/Stat pathway and specifically in medial mesenchyme through pErk pathway. Moreover, the cell proliferation of palatal epithelium could be disabled by pJak2-Stat pathway.

Ectopic activation of Fgf8 results in increased cell proliferation and failure of elevation of palatal shelves
Cleft palate is one of the most common congenital defects in human newborns. The high morbidity is attributed to the susceptibility of the finely regulated events to various influences during palatogenesis. The mouse models mimicking human cleft palate revealed that cell proliferation is the most subtle event to be affected. Most cases of cleft palate resulted from a reduced cell proliferation, which retarded outgrowth from maxillary process and horizontal growth toward mid-line of palatal shelves [7,10,11,30,35,36]. The hypothesis that cell proliferation is involved in the re-orientation and elevation of palatal shelves has been reviewed in detail [1]. In our study, the mesenchymal cell proliferation drastically increased in both the medial and lateral sides by the over-expressed Fgf8 in mesenchyme, which in turn, altered the palate size and morphology, but cleft palate still happened due to the failure of elevation of the palatal shelves, indicating that the sole horizontal expansion of palatal shelves is insufficient to complete the palatal elevation. During normal palatogenesis, palate elevation is initiated from the anterior part in a "flip-up" manner and extends to posterior palate [1]. In the Osr2-cre KI ; Rosa26R-Fgf8 palate, although the elevation has been initiated in the anterior, it failed in the posterior half. Therefore, the ectopic FGF8 not only enhances mesenchymal cell proliferation, but also impairs the reorientation of the posterior palatal shelves.
During palatogenesis, Fgf7 and Fgf10 are confined to the medial and lateral mesenchyme, respectively. Fgf7 knockout mouse survives to adulthood without cleft palate [37]. Even in the Dlx5 null mice, in which Fgf7 expression is greatly decreased, the palatal shelves are able to elevate to horizontal position [16]. By contrast, the palatal shelves of Fgf10 deficient mouse stay vertical at E15 with a reduced size [8]. Therefore, although the ectopic activation of Fgf8 suppressed the expression of Dlx5 and Fgf7, the ablation of Fgf10 expression in the developing palate is regarded as the cause of the failure of palatal elevation.

Reverse effects of ectopic FGF8 on the cell proliferation of palatal epithelium and mesenchyme
In normal palate development, FGFR2b is distributed in the epithelium and FGFR2c in mesenchyme [8]. FGF10 signals epithelial cells by binding the FGFR2b to maintain the epithelial cell proliferation [7]. Previous studies showed that FGFR2c had a much higher affinity to FGF8 binding than FGFR2b, while FGF10 had the highest affinity to FGFR2b [38]. Therefore, we assume that the ectopic activated Fgf8 directly enhanced both the lateral and medial mesenchymal proliferation through FGFR2c. Previous genetic screening demonstrates that the mutations of Fgfr1 and Fgfr3 are also associated with human cleft palate [39,40]. Recently, ablation of Fgfr1 from neural crest cells was reported to compromise the cell proliferation, elevation and fusion of palate [41]. These results suggest that FGF8 can stimulate the cell proliferation through FGFRs, especially FGFR1, more than FGFR2c. On the other hand, the reduced cell proliferation in palatal epithelium is attributed to the diminished Fgf10 expression in Osr2cre KI ;Rosa26R-Fgf8 palate, instead of the direct inhibition by FGF8, because ectopically activated Fgf8 in palatal epithelium has no effect on epithelial proliferation and Shh transcription. Based on the unaffected epithelial cell proliferation, the normal Shh expression and the elevation of palate, it is suggested that the FGF8 did not disrupt the Fgf10-Shh interaction loop in Shh-cre;Rosa26R-Fgf8 palate. Interestingly, cell proliferation was decreased in both the medial and lateral mesenchyme of Shh-cre;Rosa26R-Fgf8 palate. The controversial effect on cell proliferation by FGF8 may result from the earlier activation of Osr2-cre than Shh-cre [20,23], and /or the different dosage of FGF8 produced by mesenchyme and epithelium, considering the larger amount of mesenchymal cells than that of epithelial cells.

Differential FGF signaling pathways are involved in the cell proliferation and Fgf10 expression in palatal shelves
In the Fgfr2 C342Y/C342Y mutants, the transcription of intracellular FGF signaling pathway suppressors, Sprouty2 and Sprouty4 is up-regulated [39,40]. Deficiency of Sprouty2 increases mesenchymal cell proliferation and the level of pErk in the developing palate [42]. Therefore, we speculate that the imbalanced cell proliferation, delayed elevation and reduced size of palatal shelves in the Fgfr2 C342Y/C342Y mutants result from the up-regulated Erk signaling pathway. Our study found that the up-regulated pErk was concentrated in the Osr2-cre KI ;Rosa26R-Fgf8 medial mesenchyme, instead of the entire palatal mesenchyme. Combined with the ablation of Dlx5 and Fgf7 in the same region, we assume that FGF/pErk signaling pathway promotes the cell proliferation, but is unable to maintain Fgf10 expression in the medial mesenchyme. Meanwhile, the down-regulated pJak2 in the Osr2-cre KI ;Rosa26R-Fgf8 lateral mesenchyme might also be involved in the loss of Fgf10 transcription. Similarly, the elevated pJak1 level throughout the palatal mesenchyme may be regarded as the activator for mesenchymal cell proliferation.
The altered gene expression along the mediolateral axis of the developing palate shelves In the E13.5 Osr2-cre KI ;Rosa26R-Fgf8 palate, the enlarged medial and lateral mesenchyme are accompanied with the diminished Dlx5 and Fgf7 expression, and the disappeared Fgf10 expression, respectively. The Osr2-expressing domain was also reduced along lateral mesenchyme. Although these alterations suggest the attenuation of lateral identity of Osr2-cre KI ; Rosa26R-Fgf8 palate, it is hard to conclude if the mediolateral patterning of the Osr2-cre KI ; Rosa26R-Fgf8 palatal shelves has been impaired because of the lack of a functional marker identifying the medial or lateral mesenchyme of palate. For this reason, whether the fate of the medial and lateral mesenchyme is changed by the ectopic FGF8, still requires further investigation. Although Wnt5a plays an essential role in the cell migration along the anterior-posterior axis [4], its unaffected pattern in Osr2-cre KI ;Rosa26R-Fgf8 palate suggests that Wnt5a is not involved in mediolateral identity.
Walker and Fraser hypothesized that horizontal outgrowth of the medial mesenchyme remodeled the vertical posterior palatal shelves into elevating shelves [1]. The horizontal outgrowth is at least partially mimicked by increased medial cell proliferation in the Osr2-cre KI ; In situ hybridization for mesenchymal markers of E13.5 palatal shelves. Fgf10 was activated in the lateral mesenchyme of WT palate (arrow in A), but silenced in Osr2-Cre KI ;Rosa26R-Fgf8 palate (B). Similarly, Fgf7 activated in WT medial palatal mesenchyme (arrow in C) disappeared in mutant palate (D). The expression of Dlx5 acting as the activator of Fgf7 was detected in medial side of WT palatal mesenchyme (arrow in E), but not in the palate of mutant littermate (F). Compared with the Osr2 expression in WT lateral palate (arrow in G), the domain of Osr2 in the Osr2-Cre KI ;Rosa26R-Fgf8 palate reduced from the distal mesenchyme (arrow in H). The expression of Wnt5a detected in lateral mesenchyme of the Osr2-Cre KI ;Rosa26R-Fgf8 palate (arrow in J) was comparable to that in the WT (arrow in I). doi:10.1371/journal.pone.0136951.g005 Rosa26R-Fgf8 palate. As demonstrated in our study, the excessive proliferation in medial mesenchyme by over-expressed Fgf8 failed to compensate for the failure of elevation. Considering the alteration on the gene expression along the mediolateral axis of palatal shelves, the mediolateral patterning may be impaired. The cleft palate caused by loss of mediolateral patterning most likely results from disrupted cell migration. This proposal is supported by the decreased glycosaminoglycan accumulation and synthesis in the gain-of-function of Fgfr2 mutation [39]. Although the existence of cell migration along the mediolateral axis during palate elevation is still in debate, the cell migration along the mediolateral axis is the one of the most preferred events for further investigation. Furthermore, there is no impact on FGF signaling pathway in the Osr2-Cre KI ; Rosa26R-Fgf8 anterior palate, implying that the responsiveness to FGF signaling also varies along the A-P axis of palatal shelves. Further study on the intracellular FGF signaling pathway that regulates gene expression and cell behavior will provide us a thorough understanding of its role in the palatogenesis.