CXCL12-CXCR4 Interplay Facilitates Palatal Osteogenesis in Mice

Cranial neural crest cells (CNCCs), identified by expression of transcription factor Sox9, migrate to the first branchial arch and undergo proliferation and differentiation to form the cartilage and bone structures of the orofacial region, including the palatal bone. Sox9 promotes osteogenic differentiation and stimulates CXCL12-CXCR4 chemokine-receptor signaling, which elevates alkaline phosphatase (ALP)-activity in osteoblasts to initiate bone mineralization. Disintegration of the midline epithelial seam (MES) is crucial for palatal fusion. Since we earlier demonstrated chemokine-receptor mediated signaling by the MES, we hypothesized that chemokine CXCL12 is expressed by the disintegrating MES to promote the formation of an osteogenic center by CXCR4-positive osteoblasts. Disturbed migration of CNCCs by excess oxidative and inflammatory stress is associated with increased risk of cleft lip and palate (CLP). The cytoprotective heme oxygenase (HO) enzymes are powerful guardians harnessing injurious oxidative and inflammatory stressors and enhances osteogenic ALP-activity. By contrast, abrogation of HO-1 or HO-2 expression promotes pregnancy pathologies. We postulate that Sox9, CXCR4, and HO-1 are expressed in the ALP-activity positive osteogenic regions within the CNCCs-derived palatal mesenchyme. To investigate these hypotheses, we studied expression of Sox9, CXCL12, CXCR4, and HO-1 in relation to palatal osteogenesis between E15 and E16 using (immuno)histochemical staining of coronal palatal sections in wild-type (wt) mice. In addition, the effects of abrogated HO-2 expression in HO-2 KO mice and inhibited HO-1 and HO-2 activity by administrating HO-enzyme activity inhibitor SnMP at E11 in wt mice were investigated at E15 or E16 following palatal fusion. Overexpression of Sox9, CXCL12, CXCR4, and HO-1 was detected in the ALP-activity positive osteogenic regions within the palatal mesenchyme. Overexpression of Sox9 and CXCL12 by the disintegrating MES was detected. Neither palatal fusion nor MES disintegration seemed affected by either HO-2 abrogation or inhibition of HO-activity. Sox9 progenitors seem important to maintain the CXCR4-positive osteoblast pool to drive osteogenesis. Sox9 expression may facilitate MES disintegration and palatal fusion by promoting epithelial-to-mesenchymal transformation (EMT). CXCL12 expression by the MES and the palatal mesenchyme may promote osteogenic differentiation to create osteogenic centers. This study provides novel evidence that CXCL12-CXCR4 interplay facilitates palatal osteogenesis and palatal fusion in mice.

Cranial neural crest cells (CNCCs), identified by expression of transcription factor Sox9, migrate to the first branchial arch and undergo proliferation and differentiation to form the cartilage and bone structures of the orofacial region, including the palatal bone. Sox9 promotes osteogenic differentiation and stimulates CXCL12-CXCR4 chemokinereceptor signaling, which elevates alkaline phosphatase (ALP)-activity in osteoblasts to initiate bone mineralization. Disintegration of the midline epithelial seam (MES) is crucial for palatal fusion. Since we earlier demonstrated chemokine-receptor mediated signaling by the MES, we hypothesized that chemokine CXCL12 is expressed by the disintegrating MES to promote the formation of an osteogenic center by CXCR4-positive osteoblasts. Disturbed migration of CNCCs by excess oxidative and inflammatory stress is associated with increased risk of cleft lip and palate (CLP). The cytoprotective heme oxygenase (HO) enzymes are powerful guardians harnessing injurious oxidative and inflammatory stressors and enhances osteogenic ALP-activity. By contrast, abrogation of HO-1 or HO-2 expression promotes pregnancy pathologies. We postulate that Sox9, CXCR4, and HO-1 are expressed in the ALP-activity positive osteogenic regions within the CNCCs-derived palatal mesenchyme. To investigate these hypotheses, we studied expression of Sox9, CXCL12, CXCR4, and HO-1 in relation to palatal osteogenesis between E15 and E16 using (immuno)histochemical staining of coronal palatal sections in wild-type (wt) mice. In addition, the effects of abrogated HO-2 expression in HO-2 KO mice and inhibited HO-1 and HO-2 activity by administrating HO-enzyme activity inhibitor SnMP at E11 in wt mice were investigated at E15 or E16 following palatal fusion. Overexpression of Sox9, CXCL12, CXCR4, and HO-1 was detected in the ALP-activity positive osteogenic regions within the palatal mesenchyme. Overexpression of Sox9 and CXCL12 by the disintegrating MES was detected. Neither palatal fusion nor MES disintegration seemed affected by either HO-2 abrogation or inhibition of HO-activity. Sox9 progenitors seem important to maintain the CXCR4-positive osteoblast pool to drive osteogenesis. Sox9 expression may facilitate MES disintegration and palatal fusion by promoting epithelial-to-mesenchymal transformation (EMT). CXCL12 expression by
CXCL12-CXCR4 signaling regulates osteoblast formation by promoting alkaline phosphatase (ALP)-activity in human (Hosogane et al., 2010;Li et al., 2017), and murine osteoblasts  in vitro. ALP-activity is often used as a marker of osteoblastic development (Stein and Lian, 1993), and ALP KO mice demonstrate defects in bone mineralization (Wennberg et al., 2000). Mesenchymal stem cells exposed to osteogenic medium with CXCL12 demonstrated higher ALP-activity, supporting a role for CXCL12 in osteogenic differentiation in vitro (Kortesidis et al., 2005).
Since CXCL12-CXCR4 expression was found to promote bone formation (Zhu et al., 2011;Shahnazari et al., 2013) and ALP-activity (Kortesidis et al., 2005), we postulate that both CXCL12 and CXCR4 are expressed within the palatal mesenchyme to facilitate palatal osteogenesis. Transcription factor Sox9 may promote CXCL12-CXCR4 signaling to initiate ALP-activity, facilitating osteoblast and chondrocyte formation within the CNCCs derived mesenchyme, starting osteogenesis in the developing orofacial region.
Disturbed migration of CNCCs to the orofacial region can lead to craniofacial abnormalities . Exposure to reactive oxygen species (ROS), generated by diabetes, infections, or social poisons, including maternal smoking, alcohol consumption, and exposure to teratogens, was found to harm fetal growth and development and could even lead to craniofacial anomalies in rats (Al Ghafli et al., 2004), zebrafish (Kim et al., 2014;de Peralta et al., 2016), and humans (Simpson, 1957;Andres, 1996;Shah and Bracken, 2000;Ion and Bernal, 2015;Parks, 2015;Zadzinska et al., 2016). ROS can contribute to the etiology of congenital malformations by disrupting migration of CNCCs to the orofacial region (Sakai and Trainor, 2016). In chick embryo, ROS production generated by caffeine exposure was found to disturb CNCC migration, leading to asymmetrical microphthalmia and abnormal orbital bone development (Ma et al., 2014).
During palatogenesis, failure of elevation, horizontal growth, adherence of the palatal shelves, or disruption of MES formation and MES disintegration results in cleft palate (Nakajima et al., 2018). Palatal clefting can be caused by disturbed migration of CNCCs to the orofacial region . In humans, disruption of CNCC migration to the oral region can lead to cleft lip and palate (CLP) (Ito et al., 2003) and Pierre Robin sequence, a congenital craniofacial anomaly characterized by mandibular micrognathia, glossoptosis, and cleft palate (Jakobsen et al., 2006;Chen et al., 2019).
A combination of genetic and environmental factors is thought to play a role in the etiology of orofacial clefting (Spilson et al., 2001;Mossey et al., 2009;Brocardo et al., 2011). Mice lacking transcription factor AP-2a show hampered CNCC migration resulting in congenital anomalies, including cleft palate (Nottoli et al., 1998). Repression of CNCC proliferation by inhibiting the transcription factor family Zeb prevents fusion between cultured mice palatal shelves in vitro (Shin et al., 2018). Failure of palatal bone formation between E14.5 and E16.5 in mice is associated with cleft palate (Oka et al., 2012). Maternal smoking (Shi et al., 2008), alcohol consumption (DeRoo et al., 2008, and diabetes (Correa et al., 2008), associated with oxidative and inflammatory stress (Chen and Scholl, 2005;Albano, 2006;Kamceva et al., 2016), were found to increase the risk at babies with CLP. Gestational treatment with nicotine inhibited palatal fusion by persistence of the MES in mice fetuses (Kang and Svoboda, 2003;Ozturk et al., 2016). In pregnant CL/Fr mice, having an incidence of spontaneous CLP of 35-40% in the offspring, exposure to hypoxia at E11 doubled the incidence of CLP at E18, indicating that a combination of genetic susceptibility and oxidative stress can result in CLP (Millicovsky and Johnston, 1981).
We postulate that Sox9, CXCR4, and HO-1 are expressed in the ALP-activity positive osteogenic regions within the CNCCs derived mesenchyme during palatal fusion in mice. Furthermore, we hypothesize that chemokine CXCL12 is expressed by the disintegrating MES to promote the formation of an osteogenic center by CXCR4-positive osteoblasts. In addition, we expect that increased levels of oxidative and inflammatory stress in HO-2 KO mice, and further increase of these stress levels, obtained by administration of HO-1 and HO-2 activity inhibitor SnMP at E11 in wt mice disrupts the migration of CNCCs to the orofacial region, increasing the risk of cleft palate.
In the present study, expression of Sox9, CXCL12, CXCR4, and HO-1 was studied in relation to MES disintegration at E15 and palatal osteogenesis between E15 and E16 marked by ALPactivity, in wild-type (wt) mice using (immuno)histochemical staining of coronal palatal sections. Additionally, the effects of absence of HO-2 in HO-2 KO mice, or inhibition of HO-1 and HO-2 activity using SnMP from E11 in wt mice on palatal bone formation at E15 and E16, was investigated.

Mice Selection and Mating
Because HO-1 KO mice demonstrated severe pregnancy complications with a fetal loss rate of more than 85% (Zenclussen et al., 2011), and HO-2 mice demonstrated to be viable FIGURE 1 | Isolation of the uteri and mice fetuses. After sacrifice of the plugged mice the uteri were isolated: (A) wt E15 mouse. (B) wt E15 mouse, (C) wt CD1 SnMP E16 mouse, (D,E) HO-2 KO E15 mouse. (F) Isolation of the fetuses from the uterus of a HO-2 KO E15 mouse. Left panel: Uterus containing 6 fetuses (green asterisks) in total. Middle panel: After removal of the fetuses five resorptions (red asterisks) were found (centimeter ruler). Right panel: Overview of the six isolated fetuses with placenta (green asterisks). , the HO-2 mouse model was considered more suitable to collect fetuses for this study. Homozygote HO-2 KO mice were generated by targeted disruption of the HO-2 gene with mixed 129Sv × C57BL/6 background (Poss et al., 1995;Bellner et al., 2009). By quantitative realtime PCR, we previously confirmed the genotypes of mice by showing that HO-2 mRNA was only present in samples from wt fetuses, and not in HO-2 KO fetuses (Suttorp et al., 2017). The wt mice strain with mixed 129Sv × C57BL/6 background were used to obtain fetuses for the control group. Both the wt (n = 7) and HO-2 KO mice (n = 7) of 8 weeks old were bred and maintained in our animal facility. Female wt mice were mated with wt males, and female HO-2 KO mice were mated with HO-2 KO males. The next morning the mice were checked for the presence of a vaginal copulation plug, taken as day 0 of pregnancy (gestational/embryonic day 0; E0) (Behringer et al., 2016). Preliminary experiments (Ethical permission # RU-DEC 2009-160) demonstrated that both wt and HO-2 KO animals with mixed 129Sv × C57BL/6 background often did not carry fetuses. The hormones Folligonan (Gonadotropin serum, Intervet Nederland BV, Boxmeer, Netherlands) and Pregnyl (Human chorionic gonadotropin, NV Organon, Oss, Netherlands) were used to enhance the chance of pregnancy. At day -3 at 16.00 h Folligonan (6E in 30 µl) and at day -1 at 16.00 h Pregnyl (6E in 30 µl) was administered by intraperitoneal (i.p.) injection. At first, 1 wt mouse and two HO-2 KO mice demonstrated no plugged status. These animals were mated again 4 weeks later, and all demonstrated a plugged status the next morning.
Female CD1 mice of 12-17 weeks old (n = 11) were mated with male mice from the same strain at the animal facility of the animal suppliers Envigo (Venray, Netherlands). The plugged CD1 mice were transported to our animal facility and could acclimatize for at least 1 week before the start of the experiment. Absence of HO expression (both isoforms HO-1 and HO-2) was found to severely affect embryonic implantation (Zenclussen et al., 2011). SnMP can bind to the enzymes HO-1 and HO-2, but cannot be broken down by both isotypes, acting as a competitive inhibitor of the HO system (Stevenson et al., 1989). Instead of heme (iron protoporphyrin IX dichloride), SnMP (Sn mesoporphyrin IX dichloride) consists of a protoporphyrin IX ring with tin in its center (Lutton et al., 1999). SnMP was purchased from Frontier Scientific (Carnforth, United Kingdom). SnMP was freshly dissolved with Trizma base. The pH was adjusted to pH 7.6-8.0 with HCl, and further diluted till 10 mL with H 2 O. The SnMP solution was filter sterilized before administration. To be able to study palatal fusion in the absence of HO-1 and HO-2 activity (later referred to as HO-activity), pharmacological blocking in CD1 wt mice was obtained by administration of SnMP at the start of palatogenesis at E11.

Sample Size, Housing, Ethical Permission
To detect an effect size of 0.40 (generalized estimation of the reduction in fetal body weight following HO-2 abrogation/HOactivity inhibition) with power of 0.80 and a significance level of 0.05 for the four groups (wt E15, HO-2 KO E15, wt CD1 E16, and wt CD1 SnMP E16) a total sample size of n = 76 fetuses for this study was calculated by the one-way ANOVA power analysis a priori (G * Power 3.1 software) (Faul et al., 2007). This indicated that the mean sample size per group should comprise approximately 19 fetuses. We estimated that the litter size could range from 8 to 18 fetuses, suggesting that each group should contain three pregnant mice. The chance of conception was assumed to be around 70%, indicating that for each group at least five mice should be mated.
The animals of both models were housed under specific pathogen-free housing conditions with 12 h light/dark cycle and ad libitum access to water and powdered rodent chow.
Ethical permission for the study was obtained according to the guidelines of the Board for Animal Experiments of the Radboud University Nijmegen (Ethical permission # RU-DEC 2012-166).

Studying Chemokine Expression by the MES in the Presence or Absence of HO-2
To be able to study chemokine expression by the MES, fetuses should be studied before the palatal shelves are completely fused. In wt mice the palatal shelves fuse between E14.5 and E15.5. In this period, the MES disintegrates, allowing the formation of mesenchymal confluence (Dudas et al., 2007). Therefore, fetuses of E15 were considered to be suitable to study chemokine expression by the disintegrating MES. In the absence of HO-2 expression chemokine expression by the MES was examined in HO-2 KO fetuses of E15 (later referred to as HO-2 KO E15). The E15 wt fetuses served as controls (later referred to as wt E15).

Studying Palatal Fusion in the Presence or Absence of HO-Activity
To be able to study palatal clefting, fetuses should be studied beyond the time point of palatal fusion. Since the epithelium of the palatal shelves loses its capacity to fuse at E16, the absence of palatal fusion at E16 is diagnosed as palatal clefting (Dudas et al., 2007). Therefore, fetuses of E16 were considered to be suitable to study palatal fusion. In the absence of HO-activity, obtained by SnMP administration, palatal fusion was examined in wt CD1 fetuses of E16 (later referred to as wt CD1 SnMP E16). In the control group, no SnMP was administered (later referred to as wt CD1 E16). All CD1 mice were randomly assigned to the wt CD1 E16 or wt CD1 SnMP E16 group.

Isolation of Mice Fetuses
The plugged animals of both models were sacrificed by CO 2 /O 2 inhalation for 10 min, and the uteri and fetuses were isolated and photographed (see Figure 1). The wt and HO-2 KO mice were sacrificed at E15. Only three out of seven plugged wt mice, and four out of seven plugged HO-2 KO mice carried fetuses. In total, 16 wt E15 fetuses and 11 HO-2 KO E15 fetuses were obtained. The CD1 wt mice, the controls, and SnMP administered mice, were sacrificed at E16. In total of six out of six plugged wt mice of the control group, and four out of five plugged wt mice of the SnMP group carried fetuses. In total, 91 wt CD1 E16 fetuses and 56 wt CD1 SnMP E16 fetuses were obtained.

Fetal Loss Rate Calculation
For the fetus-carrying mice of both models, the fetal loss rate, the percentage of non-viable and hemorrhagic embryonic implantations to the total number of embryonic implantations (non-viable or hemorrhagic embryonic implantations + fetuses) was calculated.

Fetal Body Weight
To study the effects of the absence of HO-2 expression on fetal growth wt E15 fetuses (n = 15) and HO-2 KO E15 fetuses (n = 4) were weighed. Unfortunately, when the isolation of the uterus was performed for the first time, we weighed the fetuses in their amniotic sacs together with their placentas. During preparation we noticed that in some cases the amniotic fluid was partially leaked away caused by a perforation of the amniotic membrane, making this method unreliable. Subsequently, it was decided to perform the weighing of the fetuses separately from their amniotic sacs and placentas. Therefore, the weight of 1 wt E15 and 7 HO-2 KO E15 fetuses was not collected. To study the effects of HO-activity inhibition on fetal growth the wt CD1 E16 fetuses (n = 91) and wt CD1 SnMP E16 fetuses (n = 56) were weighed.

Palatogenesis Classification of HE-Stained Palatal Sections
Serial coronal palatal sections from the wt E15, HO-2 KO E15, wt CD1 E16 and wt CD1 SnMP E16 fetuses were routinely stained with hematoxylin and eosin (HE). Because the palate is not fused in once, the four stages according to Dudas et al. (2007) and Suttorp et al. (2017) can all be present on the same embryonic day. Therefore, the sections were classified into the four stages of palatogenesis based on the anatomy of the secondary palatal shelves: elevation, horizontal growth, midline adhesion, and fusion. Per individual fetus palatogenesis was studied on multiple sections. A minimum of five sections from five to seven head samples per group were essayed. Microscopic photographs were taken using a Carl Zeiss Imager Z.1 system (Carl Zeiss  Microimaging Gmbh, Jena, Germany) with AxioVision (4.8 v) software (Zeiss, Göttingen, Germany). HE series from the wt E15 and HO-2 KO fetuses were used as reference to obtain coronal palatal sections containing the MES for (immuno)histochemical staining. HE series from the wt CD1 E16 and wt CD1 SnMP E16 fetuses were used as reference to obtain coronal palatal sections containing the middle region of the fusing palatal shelves for (immuno)histochemistry.

Palatal Osteogenesis Identification by ALP-Activity
ALP-activity is often used as a marker of osteoblastic development (Stein and Lian, 1993). In the coronal palatal sections, increased expression of ALP activity was found to indicate enhanced differentiation of mesenchymal stem cells into osteoblasts (Rosa et al., 2003;George et al., 2006). Therefore, in this study ALP-positive stained regions were judged sites of initiated osteogenic differentiation. Paraffin embedded sections were selected from the wt E15, HO-2 KO E15, wt CD1 E16, and wt CD1 SnMP E16 fetuses. The sections were rinsed in demineralized water. Then, the sections were incubated for 60 min with TRIS buffer at 37 • C, followed by incubation with Medium Alkaline Phosphatase at 37 • C for half an hour. After rinsing with demineralized water, the sections were counterstained by Na-acetate buffer 0.1M pH 5.1, continued with 0.1% Methyl green in Na-acetate buffer with pH 5.1. The sections were briefly rinsed before mounting in Kaisers gelatin. Microscopic photographs of the ALP-stained palatal sections were taken.

Immunofluorescence Staining for Sox9, CXCR4, and HO-1-Expression in the Palate
For the Sox9 immunofluorescence staining, paraffin-embedded coronal palatal sections from the wt E15 and HO-2 KO E15 fetuses containing a complete or partial MES were selected. For the CXCR4 and HO-1, and the double-staining Sox9-CXCR4 and Sox9-HO-1 paraffin embedded sections from the wt E15, HO-2 KO E15, wt CD1 E16, and wt CD1 SnMP E16 fetuses were selected. Antigens were retrieved with citrate buffer at 70 • C for 10 min, following by incubation in 0.015% trypsin in PBS at 35 • C for 5 min. Next, the sections were pre-incubated with 10% normal donkey serum (NDS) in phosphate-buffered saline with glycine (PBSG). First, antibodies were applied overnight (see Table 1), after rinsing, followed by the secondary antibodies (see Table  1) for 1 h. For the double staining, the second first antibodies were also applied overnight. Nuclear staining was performed with DAPI. The sections were mounted with a glycerol based mounting medium containing 1, 4-Diazobicyclo-(2,2,2-octane (DABCO). Microscopic photographs of the immunofluorescence stained sections were taken.

Immunohistochemical Staining of Palatal CXCL12 Expression
Paraffin-embedded coronal palatal sections from the wt E15 and HO-2 KO E15 fetuses, containing a full or partial MES, were selected. Endogenous peroxidase activity was quenched with 3% H 2 O 2 in methanol for 20 min. No antigen retrieval was used. Next, the sections were pre-incubated with 10% NDS in PBSG. First antibody for CXCL12 (see Table 1) was diluted in 2% NDS in PBSG and incubated overnight at 4 • C. After washing with PBSG, sections were incubated for 60 min with a biotin-labeled second antibody against host species (see Table 1) as previously described (Tan et al., 2009). Next, the sections were washed with PBSG and treated with avidin-biotin peroxidase complex (ABC) FIGURE 4 | MES disintegration despite HO-2 abrogation and palatal fusion despite HO-activity inhibition from E11. In the HE-stained coronal palatal sections of the wt E15 and HO-2 KO E15 fetuses, the palatal shelves were attached and disintegration of the MES was found. In the HE-stained coronal palatal sections from wt CD1 E16 and wt CD1 SnMP E16 fetuses, the palatal shelves were fused. Representative palatal section per group was shown.
for 45 min in the dark. After extensive washing with PBSG, diaminobenzidine-peroxidase (DAB) staining was performed for 10 min for the CXCL12 staining. Finally, the nuclei were stained with Hematoxylin for 10 s and sections were rinsed for 10 min in water, dehydrated, and embedded in distyrene plasticizer xylene (DPX). Microscopic photographs of the CXCL12 stained sections were taken.

Quantification of CXCL12 Immunoreactivity in the Palatal Epithelium
The CXCL12 immunostained coronal palatal sections from the wt E15 and HO-2 KO E15 fetuses were examined. Within each section the epithelium of the palatal shelves was subdivided into three separate regions of interest (Inside, Outside, and Lateral nasal wall) according to morphological characteristics (epithelium region classification) as previously described (Suttorp et al., 2017). The epithelial regions per single section were semi-quantitatively scored according to the immunoreactivity scoring scale in three categories: HIGH, MODERATE and LOW (for details see Figure 2). CXCL12 immunoreactivity was evaluated by two observers, independently and blinded for the groups. The inter-and intra-examiner reliability was determined. For each individual fetus the modus of the scoring per epithelial region was calculated.

Quantification of CXCL12-Positive Cells in the Palatal Mesenchyme
Because the coronal palatal sections showed a significant variance in size of the palatal shelves, CXCL12 staining was adjusted to the surface area. The individual cross-sectional surface of the inner and outer half of each pair of shelves was measured using Fiji Image J 1.51n software (Zeiss, Göttingen, Germany). Then, the number of CXCL12-positive cells within the outline of the inner and outer half of the mesenchyme of the palatal shelves was counted (for details see Figure 3). The number of mesenchymal CXCL12-positive cells/mm 2 per inner and outer half of the palatal mesenchyme per section was calculated. Cell counting was performed by two observers, independently and blinded for the groups. The inter-and intra-examiner reliability was determined. For each individual fetus the mean number of mesenchymal CXCL12-positive cells/mm 2 per inner and outer half of the mesenchyme of the palatal shelves was calculated.

Statistical Analysis
The fetal loss rate of the wt E15 and HO-2 KO E15 fetuses, and the wt CD1 E16 and wt CD1 SnMP E16 fetuses, showed a non-normal distribution as evaluated by the Kolmogorov-Smirnov test (KS-test). The differences in fetal loss rate between those groups were compared using the non-parametric Kruskal-Wallis ANOVA on ranks test and Dunn's Multiple Comparison post hoc test. The fetal body weight of the wt E15 and HO-2 KO E15 fetuses showed a normal distribution as evaluated by the KS-test. To compare differences between those groups the Independent-Samples T-test was performed. The fetal body weight of the wt CD1 E16 and wt CD1 SnMP E16 fetuses showed a nonnormal distribution as evaluated by the KS-test. The nonparametric Mann-Whitney test was used to compare differences between both groups.
The CXCL12 immunoreactivity in the three regions of interest of the palatal epithelium of the wt E15 and HO-2 KO E15 fetuses was semi-quantitatively scored and analyzed using the nonparametric Kruskal-Wallis ANOVA on ranks test and Dunn's Multiple Comparison post hoc test to compare differences between both groups.
The number of CXCL12-positive cells/mm 2 within the outline of the inner and outer half of the palatal mesenchyme of the wt E15 and HO-2 KO E15 fetuses was quantitatively scored and showed a normal distribution as evaluated by the KS-test. The data was analyzed using the ANOVA and Tukey's multiple comparison post hoc test to compare differences between the inner and outer half of the palatal mesenchyme of both groups.
To determine the inter-and intra-examiner reliability for the semi-quantitative data of the CXCL12 immunoreactivity in the epithelium, the Cohen's kappa coefficient was calculated. Acceptable scores >0.80 were obtained for the semi-quantitatively scoring.
To determine the inter-and intra-examiner reliability, the coefficient of determination (R 2 ) was calculated by the square of the Pearson correlation coefficient for the quantitative data of the CXCL12-positive cells/mm 2 in the palatal mesenchyme. Acceptable scores >0.80 were obtained for the counting.
Differences were considered to be significant if p < 0.05. All statistical analyses were performed using Graphpad Prism 5.03 software (GraphPad Software, San Diego, CA, United States).

MES Disintegration Despite HO-2 Abrogation and Palatal Fusion Despite HO-Activity Inhibition From E11
In the HE-stained coronal palatal sections of the wt E15 and HO-2 KO E15 fetuses, the tips of the palatal shelves were attached and disintegration of the MES was found. In the HE-stained sections from the wt CD1 E16 and wt CD1 SnMP E16 fetuses, the palatal shelves were fused. Representative palatal section per group was shown in Figure 4.

Palatal Osteogenesis During MES Disintegration Is Normal Despite HO-Activity Inhibition
In the coronal palatal sections of the wt E15 and HO-2 KO E15 fetuses, clusters of ALP-positive-stained mesenchymal cells including small bone matrix depositions were found in the regions lateral of the fusing palatal shelves and lateral near the nasal septum. No disruption of ALP expression due to HO-2 abrogation was found. In the coronal palatal sections of the wt CD1 E16 and wt CD1 SnMP E16 fetuses, clusters of ALP-positive-stained mesenchymal cells surrounding large areas of bone depositions were found in both the center and lateral regions of the palatal shelves, and lateral near the nasal septum. Because of their ALP-activity, and their location around bone depositions, these ALP-positive clustered cells were regarded as osteoblast progenitors and osteoblasts. No disruption of ALP activity due to HO-activity inhibition was found. Representative section per group are shown in Figure 5. The clusters of ALP-positive-stained mesenchymal cells of the fusing palate were considered as palatal osteogenic centers and determined as region of interest for studying CXCL12, CXCR4, and Sox9 expression.

Sox9 Expressing Cells Present in the MES and Palatal Osteogenic Centers
In both wt E15 and HO-2 KO E15 fetuses, Sox9 highexpressing cells were found in the disintegrating MES. Moreover, clusters of Sox9-positive mesenchymal cells were found in the osteogenic centers in the lateral parts of the fusing palate and at the oral side of the forming nasal septum. Regarding their clustered arrangement in the osteogenic centers it was assumed that these Sox9-positive cells are predominantly osteoblast progenitors. Furthermore, Sox9 expressing mesenchymal cells were present in the cartilage of the nasal septum, and are likely chondroblasts because of their specific position (see Figure 6). The non-specific autofluorescence is shown in Supplementary Figure 1A.

CXCR4 Expressing Cells in the Palatal Osteogenic Centers
In both wt E15 and HO-2 KO E15 fetuses, clusters of CXCR4positive mesenchymal cells were found in the osteogenic centers in the lateral parts of the fusing palate. No CXCR4 expression was found within the disintegrating MES. In the wt CD1 E16 and wt CD1 SnMP E16 fetuses, CXCR4positive stained cells in the osteogenic centers in both the central and lateral parts of the forming palate were found. Furthermore, at the oral side of the forming nasal septum CXCR4 expressing cells were found at E15 and E16 (see Figure 7). Regarding their clustered arrangement in the osteogenic centers, these CXCR4-positive mesenchymal cells were supposed to be primarily osteoblasts. In the cartilage of the nasal septum, the observed CXCR4-positive cells were considered to be generally chondroblasts. The non-specific staining of the second antibody for CXCR4 is shown in Supplementary Figure 1B.

Most CXCR4-Positive Cells in the Palatal Osteogenic Centers Are Not Positive for Sox9
In both wt and HO-2 KO E15 fetuses, clusters of Sox9 expressing cells surrounding CXCR4-positive cells in the osteogenic centers in the lateral parts of the forming palate were found (see Figure 8). Only a few Sox9-CXCR4 double-positive stained cells were found near the remnants of the MES. In the wt CD1 E16 and wt CD1 SnMP E16 fetuses, almost no Sox9-CXCR4 doublepositive stained cells were found in the osteogenic centers in the central part of the forming palate. In general, less Sox9 expressing cells were found in the palatal mesenchyme in the wt CD1 E16 and wt CD1 SnMP E16 fetuses compared to the wt E15 and HO-2 KO E15 fetuses. In the osteogenic centers, Sox9 expressing cells were surrounding CXCR4-positive cells. At the oral side of the forming nasal septum CXCR4 expressing cells near Sox9 expressing cells were found at E15 and E16. The observation that Sox9-positive cells were located close to CXCR4-positive cells surrounding the osteogenic centers supports the suggestion that these Sox9-positive cells could be osteoblast progenitors to maintain the osteoblast pool to drive osteogenesis.

HO-1 Expressing Cells in the Palatal Osteogenic Centers
In both wt E15 and HO-2 KO E15 fetuses, HO-1-positive-stained cells were found in the osteogenic centers in the central part but mainly in the osteogenic centers in the lateral parts of the forming palate. In the wt CD1 E16 and wt CD1 SnMP E16 fetuses, a few HO-1-positive stained cells were found in the osteogenic centers in the central part, but clusters of HO-1positive stained cells were dominantly found in the osteogenic centers in the lateral parts of the forming palate. At the oral side of the forming nasal septum, HO-1 expressing cells were found at E15 and E16 (see Figure 9). Since many HO-1-positive cells were found in the osteogenic centers, they were considered to be predominantly osteoblasts.

Most HO-1-Positive Cells in the Palatal Osteogenic Centers Are Not Positive for Sox9
In both wt and HO-2 KO E15 fetuses, only a few Sox9-HO-1 double-positive stained cells were found near the remnants of the MES (see Figure 10). In the wt CD1 E16 and wt CD1 SnMP E16 fetuses, HO-1 expressing cells were found in the osteogenic centers in the central part of the forming palate, but almost no Sox9-HO-1 double-positive stained cells were found. At the oral side of the forming nasal septum, clusters of HO-1 expressing cells were present (see Figure 10). The examination that Sox9-positive cells surrounded the osteogenic centers containing multiple HO-1positive cells possibly means that osteoblast progenitors are not HO-1 positive.

CXCL12 Expression Present in the MES and Palatal Osteogenic Centers
Chemokine CXCL12 was significantly higher expressed in the inside epithelial region including the MES (Inside) of the wt E15 and HO-2 KO E15 fetuses compared to the other epithelial layers of the fusing palatal shelves (Outside), and the lateral epithelium of the nasal cavity (Lateral nasal wall) (p < 0.001) (see Figures 11A,B).
CXCL12 expressing cells were also observed in the mesenchyme of the palatal shelves, but the number of CXCL12positive cells/mm 2 in palatal section showed no significant difference between the wt E15 and HO-2 KO E15 fetuses, and between the inner half and outer half (p = 0.85) (see Figure 11C). Based on their uniformly distribution, a majority of these CXCL12-positive cells in the palatal mesenchymal were considered to be fibroblasts. Furthermore, the osteogenic centers in the lateral parts of the fusing palate demonstrated clusters of strong CXCL12 expressing cells. Considering their clustered arrangement in the osteogenic centers, these CXCL12positive cells were thought to be osteoblast progenitors or osteoblasts. Also within the cartilage of the forming nasal septum CXCL12 expressing cells were present, and these cells were regarded as chondroblast because of their specific location (see Figure 11A).

Fetal Resorption Independent of HO-2 Expression and HO-Activity
Approximately half of the wt E15 and HO-2 KO E15 fetuses were resorbed. Contrarily, in the wt CD1 E16 and wt CD1 SnMP E16 fetuses, the number of fetal resorptions was low (less than 7%). Higher fetal loss rate was found in the HO-2 KO E15 group compared to both the wt CD1 E16 and wt CD1 SnMP groups (p < 0.05) (see Figure 12). Disruption of HO-2 expression and HO-activity inhibition both did not result in increased fetal loss (p > 0.05).

Fetal Body Weight Is Decreased by Absence of HO-2 Expression but Increased by HO-Activity Inhibition From E11
The HO-2 KO E15 fetuses demonstrated a reduced fetal body weight compared to the wt E15 fetuses (p = 0.0416) (Figures 13A,B). On the other hand, in CD1 fetuses of E16, inhibition of the HO-activity by SnMP from E11 increased the fetal body weight (p < 0.0001) (Figures 13C,D). A representative fetus per group is shown in Figures 13B,D.

DISCUSSION
We found clusters of Sox9 expressing cells surrounding CXCR4positive cells in the ALP-positive osteogenic centers of the palatal shelves at E15 and E16 in wt mice. However, most Sox9 positive cells in the palate were not positive for CXCR4 expression. Furthermore, we showed that chemokine CXCL12, the ligand of receptor CXCR4, was overexpressed within these palatal osteogenic centers. Sox9, CXCR4, and CXCL12 expressing cells were also observed within the cartilage of developing nasal septum, suggesting their involvement in chondrogenesis. Several in vitro and in vivo studies on the role of CXCL12-CXCR4 signaling in relation to bone formation have been performed (Gilbert et al., 2019). CXCL12 expression in non-differentiated mesenchymal stem cells in bone was demonstrated, but toward the end of osteogenic differentiation CXCL12 was downregulated (Ito, 2011). Deletion of CXCL12 in osteoprogenitor cells resulted in decreased bone mass in the femur bone in mice (Tzeng et al., 2018). CXCL12 antibody administration inhibited new bone formation during the repair of femoral bone graft in mice (Kitaori et al., 2009). CXCR4 KO fetuses of E18.5 were smaller and showed deficient bone marrow development compared to the controls (Ma et al., 1998). CXCR4-deficient bone marrow-derived mesenchymal stromal stem cells from mice exhibited impaired osteogenic differentiation in response to BMP2 stimulation in vitro (Guang et al., 2013). Primary cultures for osteoblastic cells derived from CXCR4 KO mice showed decreased proliferation and impaired osteoblast differentiation in response to BMP2 or BMP6 stimulation in vitro (Zhu et al., 2011). Additionally, disruption of CXCR4 receptor in mouse hematopoietic stem cells led to increased endogenous ROS production (Zhang et al., 2016). Since our results showed that Sox9 and CXCR4 do not have overlapping expression, we suggest that Sox9-positive cells are osteoblast progenitors FIGURE 12 | Fetal resorption independent of HO-2 KO expression and HO-activity. (A) Box-and-whisker plot with 10-90 percentiles of quantitative assessment of the fetal loss ratio in the wt (n = 3) and HO-2 KO pregnant mice (n = 4) at E15, and wt CD1 mice (n = 6) and wt CD1 SnMP mice (n = 4) at E16, * p < 0.05. (B) The uteri were photographed, the green asterisk indicated a fetus, the red asterisk indicated a fetal resorption. Representative uterus per group was shown. (C) Isolated fetus with placenta and isolated fetal resorption from the uterus. maintaining the osteoblast pool to drive osteogenesis, and that both CXCL12 and CXCR4 are later expressed by the mature osteoblasts. Furthermore, we propose that Sox9 expression and CXCL12-CXCR4 signaling in the cartilage of the nasal septum develop chondroblast formation. Since CXCL12-CXCR4 signaling promotes osteoblast formation, it is suggesting that CXCL12 expression by the disintegrating MES is a major initiator of palatal osteogenesis.
Studies on CXCL12-CXCR4 signaling in the developing palate are scarce. We provided novel evidence that chemokine CXCL12, the ligand of receptor CXCR4, was overexpressed by the epithelial cells of the MES, possibly to activate osteoblasts progenitors to facilitate palatal bone formation. In human palatal sections, CXCR4 expression was observed within the MES (Jakobsen et al., 2009). On the contrary, in our study we did not observe CXCR4 overexpression within the disintegrating MES. We could not find other studies that demonstrated CXCR4 expression in the MES. It is suggesting that CXCL12 expression by the disintegrating MES is a major initiator of palatal osteogenesis. Besides chemokine CXCL12, the disintegrating MES demonstrated strong expression of Sox9. Although Sox9 expression is demonstrated in the developing palatal shelves in mice fetuses (Potter and Potter, 2015;Watanabe et al., 2016;Xu et al., 2018), to our knowledge we are the first who demonstrated Sox9 expression specifically in the disintegrating MES. Sox9 signaling showed to be essential in the EMT mechanism in non-small-cell lung carcinoma cell Huang et al., 2019), mouse embryonic mammary cells (Kogata et al., 2018), mouse gastric cancer cells , human colorectal cancer cells (Choi et al., 2017), human hepatocellular carcinoma cells (Kawai et al., 2016), thyroid cancer cells (Huang and Guo, 2017), and avian neural crest cells (Sakai et al., 2006) in vitro. Therefore, it's tempting to speculate that Sox9 expression near the MES is involved in EMT processes, turning the basal epithelial layer of the MES into mesenchyme during palatal fusion. Our novel hypothetical model (Figure 14) proposes that CXCL12-CXCR4 signaling facilitates osteogenesis during palatal fusion in mice. Expression of Sox9 in osteoblast progenitors was thought to initiate osteogenic differentiation by promoting CXCL12-CXCR4 signaling. CXCL12 expression by the MES and osteogenic centers in the fusing palatal shelves could promote maturation of immature CXCR4-positive osteoblasts. Sox9 progenitors thus seem important to maintain the CXCR4positive osteoblast pool to drive osteogenesis. Furthermore, Sox9 expression found in the MES may in addition regulate MES disintegration by the process of EMT.
Only a few Sox9-positive cells in the palatal mesenchyme were also positive for HO-1. Studies on the role of HOactivity on embryonic craniofacial development in mice are limited (Suttorp et al., 2017). As far as we know, this is the first study on the effects of inhibition of HO-1 and HO-2 by SnMP administration during palatogenesis. Adult HO-1 KO mice demonstrate growth retardation (Suliman et al., 2017), but HO-2 KO mice were found to be morphologically indistinguishable FIGURE 13 | Fetal body weight decreases by disruption of HO-2, but increases by HO-activity inhibition from E11. (A) Bar chart of the fetal body weight of the wt E15 fetuses (n = 15) and HO-2 KO E15 fetuses (n = 4), * p < 0.05. Data are shown as mean ± SD. (B) Representative fetus per group was shown (centimeter ruler). (C) Box-and-whisker plot with 10-90 percentiles of quantitative assessment of the body weight of the wt CD1 E16 fetuses (n = 91) and wt CD1 SnMP E16 fetuses (n = 56), * * * p < 0.001. (D) Representative fetus per group was shown (centimeter ruler).
FIGURE 14 | Hypothetical model: CXCL12-CXCR4 interplay facilitates palatal osteogenesis. We propose that the CXCL12-CXCR4 interplay facilitates osteogenesis during palatal fusion in mice. Expression of transcription factor Sox9 in osteoblast progenitors was thought to initiate differentiation into osteoblasts by promoting CXCL12-CXCR4 signaling. CXCL12 expression in the midline epithelial seam could promote maturation of immature CXCR4-positive osteoblast. Sox9 progenitors maintain the CXCR4-positive osteoblast pool to drive osteogenesis. Furthermore, Sox9 expression found in the MES, possibly regulates MES disintegration by the process of epithelial-to-mesenchymal transformation (EMT). from wt mice (Poss et al., 1995). In rat, HO-1 inhibition by zinc deuteroporphyrin IX 2,4 bis glycol resulted in a significant decrease in pup size (Kreiser et al., 2002). This study showed that neither adhesion of the palatal shelves nor Sox9 and CXCL12 expression by the MES was affected by HO-2 KO disruption. We found that pharmacologically blocking of HO-activity by SnMP did not disturb palatal fusion or expression of Sox9, CXCR4, HO-1, and ALP-activity in the osteogenic regions.
Besides the absence of HO-activity, we did not further disturb the pregnancy by additional oxidative and inflammatory stress. Interestingly, HO-1 expression, but not HO-2, was found to rescue mesenchymal stem cells from H 2 O 2 -induced cell death in vitro . It is likely that the role of HO is underestimated when its function is studied only during normal physiological conditions. For future experiments we consider to combine HO-activity inhibition with heme administration in different dose to study the effects of higher oxidative and inflammatory stress levels at palatal fusion. More research is needed to elucidate the relation between HO-activity and palatogenesis during stress-induced pathological pregnancy.
In our study, HO-2 abrogation demonstrated no increased fetal loss since both the HO-2 KO mice and wt mice showed a high fetal loss rate. Others found that HO-2 KO mice were fertile (Poss et al., 1995;Lundvig et al., 2014), although, no information was provided about the litter size. In another study, an increased abortion rate in mice was experimentally generated by lipopolysaccharide (LPS) injection at E18 in a dosedependent fashion (Toyama et al., 2015). A limitation of the present study was the relatively small number of wt E15 and HO-2 KO E15 fetuses. Although we tried to enhance the fertility by administration of the hormones Folligonan and Pregnyl before mating, only three out of seven plugged wt mice, and four out of seven plugged HO-2 KO mice carried fetuses. It is suggested that the low fertility and high fetal loss rate in our mice with mixed 129Sv × C57BL/6 background possibly have dominated the effects of HO-2 abrogation in this study. However, our data could not provide evidence for this statement. The low fertility could possibly be explained by the excessive inbreeding within the colony. For future studies this problem could be solved by breeding homozygous HO-2 knockouts with hybrid 129Sv × C57BL/6 wt mice. By breeding the resulting heterozygotes, wt and homozygous HO-2 KO mice would then be obtained. HO-activity inhibition from E11 was not found to increase the fetal loss rate compared to the controls. HOactivity inhibition by administration of zinc mesoporphyrin early in pregnancy in mice between E0-E6 was also found to increase the abortion rate (Sollwedel et al., 2005;Schumacher et al., 2012). We demonstrated that HO-2 KO fetuses showed a reduced fetal body weight, but, interestingly, pharmacological blocking of HO-activity by SnMP administration at E11 resulted into a higher fetal body weight. This suggests that during implantation HO-activity promotes fetal growth, whereas HOactivity at later moments (E11-E16) inhibits embryonal growth. In fact, others showed that adult HO-2 KO mice were obese, induced by disrupted metabolic homeostasis, caused by insulin resistance and elevated blood pressure (Cao et al., 2012). It would then be interesting to investigate whether induction of HO-1 could indeed counteract this obesity. However, no difference in phenotype was found between HO-2 KO and wt mice by others (Qu et al., 2005). We hypothesize that HO-activity is especially essential during implantation and early embryonic development, but stress-level-dependent in the later stages of fetal development.
In conclusion, our results support the hypothesis that CXCL12-CXCR4 interplay facilitates palatal osteogenesis during palatal fusion in mice. To the best of our knowledge, this is the first study demonstrating Sox9 and CXCL12 expression in the disintegrating MES and other palatal osteogenic centers. Neither CXCL12 expression during the MES disintegration nor palatal fusion was affected by HO-2 abrogation or inhibition of HO-activity. Further research is needed to unravel the role of cytoprotective HO-activity in the presence of additional oxidative and inflammatory stresses in relation to the development of craniofacial malformations, including palatal clefting.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The animal study was reviewed and approved by Board for Animal Experiments of the Radboud University Nijmegen (Ethical permission # RU-DEC 2012-166).

AUTHOR CONTRIBUTIONS
NV and CS designed the experiments, performed the experiments, analyzed the data, and wrote the manuscript. RER performed the experiments. RFR provided the mice. MH performed the experiments. AK-J wrote the manuscript and supervised the research. FW designed the experiments, analyzed the data, wrote the manuscript, and supervised research. All authors contributed to the article and approved the submitted version.

FUNDING
This study was supported by the Radboud University Medical Centre, Nijmegen, Netherlands.