Anti-osteogenic function of a LIM-homeodomain transcription factor LMX1B is essential to early patterning of the calvaria
Introduction
The normal development of the calvaria (top part of the skull) requires a fine balance between osseous and non-osseous components. At birth, a human calvaria is made of plates of bone and fibrous joints called sutures. Larger areas of soft tissue at the intersections of the sutures are called fontanelles. While the bone protects the brain, the sutures and fontanelles provide flexibility. The sutures also contain progenitors and stem cells for osteoblasts, and thus they enable the skull to grow coordinately with the expanding brain of a child (Ishii et al., 2015; Lenton et al., 2005; Morriss-Kay and Wilkie, 2005; Rice, 2008; Twigg and Wilkie, 2015).
Calvarial abnormalities are a major class of birth defects in humans. Craniosynostosis is a condition in which one or more sutures are lost prematurely due to a fusion of the bone plates, and it occurs in 1 out of 2,000~2,500 live births (Levi et al., 2012; Twigg and Wilkie, 2015). Craniosynostosis often results in a dysmorphic skull, for which the only treatment option currently is invasive surgeries (Derderian and Seaward, 2012; Garza and Khosla, 2012; Levi et al., 2012). Enlarged parietal foramina are another congenital anomaly affecting the calvaria, and it is caused by deficient ossification (Rice, 2005).
The structure of the calvaria at birth is very similar between humans and mice (Holmes, 2012; Rice, 2008). As such, mice have been used as a main model system for investigating the molecular and cellular underpinnings of calvarial development (Holmes, 2012; Ishii et al., 2015). They have a pair of frontal bones separated by the interfrontal (metopic in humans) suture, a pair of parietal bones separated by the sagittal suture, and one interparietal (occipital in humans) bone at the posterior end (Fig 1A,B). Coronal sutures are present between a frontal bone and a parietal bone, and lambdoidal sutures are located between a parietal bone and the interparietal bone. While research in the past couple of decades have identified a growing list of genes and signaling pathways that are essential to normal development of the calvaria (Ishii et al., 2015; Twigg and Wilkie, 2015), our knowledge of the process is far from complete. In particular, little is known about the molecular genetic regulation of early steps, such as establishing the initial layout of the calvaria so that the bones and the sutures arise at specific locations (Rice, 2008).
A) A lateral view of a postnatal day 0 (P0) mouse skull, stained with Alizarin red for bone. The bracket indicates the calvaria. B) A schematic for the dorsal view of P0 calvaria. Nose is to the left. Bones and sutures are indicated in red and gray, respectively. af: anterior fontanelle, co: coronal suture, Fr: frontal bone, if: interfrontal suture, IP: interparietal bone, la: lambdoidal suture, Pa: parietal bone, sa: sagittal suture. C) A schematic of an embryonic day (E) 12.5 head showing the position of the section in D. D) A schematic for a coronal section of the head showing the brain (Br, light purple), head mesenchyme (orange), and the surface ectoderm (blue outline). Also indicated are the locations of the early migrating mesenchyme (EMM) and the supra-orbital mesenchyme (SOM), and their contribution to the calvaria according to (Roybal et al., 2010; Yoshida et al., 2008). The dark brown spot represents the bone rudiment appearing in SOM at this stage. E) Developing bone is shown in isolation from coronal sections of the head at various stages, to illustrate the apical growth of the frontal bone and the parietal bone over time, and the subsequent formation of the sutures at the dorsal midline of the head. Apical (Ap) – basal (Ba) axis is indicated. E is modified from (Rice et al., 2000). Bar: 0.5 mm.
The calvaria develops from the head mesenchyme of two embryonic origins, neural crest and mesoderm (Jiang et al., 2002; Yoshida et al., 2008). These cells form a layer between the brain and the ectoderm, and surround the brain from base to apex by mouse embryonic day (E) 10.5 (mouse gestation is 19 days) (Jiang et al., 2002; Yoshida et al., 2008). At ~E12, the mesenchyme just above the eye, or supra-orbital mesenchyme (SOM), turns on osteogenic genes to form the rudiments of the frontal bone and the parietal bone, intervened by the prospective coronal suture (Deckelbaum et al., 2012; Jiang et al., 2002; Yoshida et al., 2008) (Fig 1C,D). The frontal bone and the parietal bone expand apically toward the vertex over the following days. When the bones from left and right approximate late in gestation, the interfrontal and the sagittal sutures are established along the dorsal midline of the head (Rice, 2008) (Fig 1E).
The mesenchyme lying apical to SOM at the beginning of calvarial bone development (E12~E13) has been termed early migrating mesenchyme (Roybal et al., 2010) (EMM) (Fig 1D). Interestingly, unlike SOM, EMM does not generate any ossification center during normal development of an embryo (Kaufman, 1994; Rice, 2008). Dye labeling and tracing have confirmed that the cells from SOM, but not EMM, contributed to the calvarial bone; cells from EMM were found in the suture or the soft tissue layers flanking the bone (Roybal et al., 2010; Ting et al., 2009; Yoshida et al., 2008). The absence of osteogenesis from EMM is highly significant in determining the final structure of the calvaria because it allows the dorsal midline of the skull to be occupied by sutures and fontanelles as opposed to being capped by a bone plate (Fig. 1E). However, the molecular basis of this difference between SOM and EMM had been unknown.
In the current study, we present evidence that Lmx1b (LIM homeobox transcription factor 1 beta), encoding a LIM-domain homeodomain transcription factor, plays a key role in specifying the different regions within the head mesenchyme with different osteogenic potential. Our findings provide the first molecular insight into the fundamental question of how the calvarial primordium gets patterned into some areas that will make the bone and other areas that will become the sutures and fontanelles.
Material and methods
Animals
All animal work was conducted with the approval from New York University Institutional Animal Care and Use Committee (approved protocols #100905 and #160711). Mice were euthanized by cervical dislocation after carbon dioxide-induced narcosis, or by decapitation with sharp scissors in case of neonates. Prrx1-Cre, Lmx1b, Lmx1b, and R26Lmx1b lines have been described previously (Chen et al., 1998a; Li et al., 2010; Logan et al., 2002; Zhao et al., 2006). Cre-reporter (R26R) lines R26R-LacZ and R26R− YFP have been described in (Soriano, 1999; Srinivas et al., 2001). Lmx1b LOF mutant embryos (Prrx1-Cre;Lmx1b;R26R/+) were obtained from crosses of Prrx1-Cre;Lmx1b+/− males and Lmx1b;R26R/R or Lmx1b;R26R/R females. The embryos were genotyped by PCR using DNA from the tail. Littermates with two copies of functional Lmx1b (Lmx1b;R26R/+ or Lmx1b;R26R/+) were used as controls. Lmx1b GOF mutant embryos (Prrx1-Cre;R26Lmx1b/+ or Prrx1-Cre;R26Lmx1b/R) were obtained from crosses of Prrx1-Cre;R26R-LacZ/+ males and R26Lmx1b/Lmx1b females. Littermates without the Cre (R26Lmx1b/+ or R26Lmx1b/R-LacZ) were used as controls. The mice were maintained in a mixed background, predominantly C57Bl6 and CD-1. The embryos were stage-matched using combinations of overall size, facial morphology, and the degree of limb digitation. For experiments using sections, morphologies of the tongue, palate, and teeth were used as additional criteria for staging.
Skeletal staining, silver nitrate staining, β-galactosidase staining
Skeletal staining was performed as described (Wallin et al., 1994) using Alizarin red for bone and Alcian blue for cartilage. Frozen sections were prepared as described (Jeong et al., 2012), and silver nitrate staining for the bone was performed using Von Kossa Histology Stain Kit (Diagnostic Biosystems, KT 028) following the manufacturer’s instructions. β-galactosidase staining was performed as described (Jeong et al., 2004).
Micro-CT
P0 heads were scanned with Skyscan at 40 kV, 250 µA, and 7.47 µm resolution, and reconstructed using NRecon program.
RNA in situ hybridization
Whole mount RNA in situ hybridization was performed as described using a digoxigenin-labeled probe (Wilkinson and Nieto, 1993). Section RNA in situ hybridization was performed on frozen sections as described (Schaerenwiemers and Gerfinmoser, 1993). The templates for RNA probes for Acan, Bmp4, Dlx5, Runx2, Sp7, Sox9, and Wnt7b were obtained from other researchers, and further information is available upon request. The template for Lmx1b probe was PCR-amplified from a commercial full-length cDNA clone (Open Biosystems, ID 40129984), adding a T3 polymerase-binding site. Primer sequences are provided in S1 Table.
Immunofluorescence and immunohistochemistry
Immunofluorescence was performed on frozen sections as described (Jeong and McMahon, 2005). The primary antibodies were rabbit anti-Ki67 (Abcam, ab15580, 1:400), rabbit anti-cleaved caspase-3 (Cell Signaling Technology, #9661, 1:300), and rabbit anti-Sp7 (Abcam, ab209484, 1:1000). Alexa Fluor secondary antibodies (Invitrogen) were used for detection. 4',6-diamidino-2-phenylindole (DAPI) was used to stain nuclei. The numbers of cells were counted manually from two sections per embryo. Two-tailed Student’s t-test was used for comparison of cell proliferation rates.
Immunohistochemisty for p-Smad1/5/9 was performed using rabbit anti-phospho-Smad1(Ser463/465)/ Smad5(Ser463/465)/ Smad9(Ser465/467) antibody from Cell Signaling Technology (#13820, 1:800). TSA Biotin Kit (Perkin-Elmer, NEL700A001KT) was used to amplify the signal and 3,3’-Diaminobenzidine tablets (DAB) (Sigma-Aldrich, D4293) were used for detection.
Western blot
For western blot comparing EMM and SOM of wild type embryos, the tissue was dissected from 11~16 of E12.5 wild type CD-1 embryos, as illustrated in S1 Fig, to prepare one lysate sample for each region. Three samples of EMM and SOM each were used for the quantitative analysis described below. For western blot comparing EMM of control and Lmx1b LOF mutant embryos, EMM was dissected as in S1 Fig and three samples per genotype were analyzed, where each sample consisted of EMM from 4 embryos.
The tissue was homogenized and lysed in radioimmunoprecipitation assay (RIPA) buffer containing Halt phosphatase inhibitor (Thermo Scientific, 78420) and Complete mini protease inhibitor (Sigma-Aldrich, 11836153001). Protein concentration of the lysate was determined by Bradford assay (Bradford, 1976), and 2~20 ug of total protein was loaded to each lane of a gel, depending on the marker to be examined. The primary antibodies were rabbit anti-pSMAD1/5/9 (Cell signaling #13820, 1:1000), mouse anti-active β-catenin (ABC) clone 8E7 (Millipore 05-665, 1:2000), rabbit anti-pErk1/2 (p44/42 MAPK) (Cell signaling #4370, 1:2000), rabbit anti-p38 MAPK (Cell signaling #9211, 1:1000) and rabbit anti-β-actin (Cell signaling #8457, 1:1000). The secondary antibodies were goat anti-mouse IgG horseradish peroxidase (HRP) (Thermo Fisher A10551, 1:1000) and donkey anti-rabbit IgG-HRP (Novex A16023, 1:10,000). Following enhanced chemiluminescence reaction with SuperSignal West Dura Extended Duration Substrate (Thermo Fisher, 37071), the signals were detected by Gel Doc (BioRad). The intensities of the bands were quantified using Image Lab (BioRad), and they were normalized against the intensity of the β-actin band from each sample. Two-tailed Student’s t-test was used for statistical comparison.
Reverse transcription and real-time quantitative PCR (RT-qPCR)
EMM and/or SOM were dissected from E12.5 embryos as in S1 Fig. EMM from E11.5 embryos was dissected as in S1 Fig except that the ectoderm was not removed due to a technical difficulty. The tissue was homogenized using a pestle and QIAshredder (Qiagen), and total RNA was extracted using RNeasy Mini Kit (Qiagen). Reverse-transcription was performed using Transcriptor First Strand cDNA Synthesis Kit (Roche, 04379012001). qPCR was performed using Power SYBR Green PCR Master Mix (Applied Biosystems, 4367659), and a housekeeping gene Peptidylpropyl isomerase (Ppib) was used as an internal standard (Pachot et al., 2004). mRNA levels of all the genes were normalized against Ppib level within each sample, using a formula 2^(−ΔCt(Gene of interest − Ppib))×1000, and compared across different samples. For all the genes, >5 samples per genotype were analyzed and two-tailed Student’s t-test was used for a statistical comparison. The primers used for qPCR are listed in S1 Table. Most primer sequences were from PrimerBank (Spandidos et al., 2010).
Animals
All animal work was conducted with the approval from New York University Institutional Animal Care and Use Committee (approved protocols #100905 and #160711). Mice were euthanized by cervical dislocation after carbon dioxide-induced narcosis, or by decapitation with sharp scissors in case of neonates. Prrx1-Cre, Lmx1b, Lmx1b, and R26Lmx1b lines have been described previously (Chen et al., 1998a; Li et al., 2010; Logan et al., 2002; Zhao et al., 2006). Cre-reporter (R26R) lines R26R-LacZ and R26R− YFP have been described in (Soriano, 1999; Srinivas et al., 2001). Lmx1b LOF mutant embryos (Prrx1-Cre;Lmx1b;R26R/+) were obtained from crosses of Prrx1-Cre;Lmx1b+/− males and Lmx1b;R26R/R or Lmx1b;R26R/R females. The embryos were genotyped by PCR using DNA from the tail. Littermates with two copies of functional Lmx1b (Lmx1b;R26R/+ or Lmx1b;R26R/+) were used as controls. Lmx1b GOF mutant embryos (Prrx1-Cre;R26Lmx1b/+ or Prrx1-Cre;R26Lmx1b/R) were obtained from crosses of Prrx1-Cre;R26R-LacZ/+ males and R26Lmx1b/Lmx1b females. Littermates without the Cre (R26Lmx1b/+ or R26Lmx1b/R-LacZ) were used as controls. The mice were maintained in a mixed background, predominantly C57Bl6 and CD-1. The embryos were stage-matched using combinations of overall size, facial morphology, and the degree of limb digitation. For experiments using sections, morphologies of the tongue, palate, and teeth were used as additional criteria for staging.
Skeletal staining, silver nitrate staining, β-galactosidase staining
Skeletal staining was performed as described (Wallin et al., 1994) using Alizarin red for bone and Alcian blue for cartilage. Frozen sections were prepared as described (Jeong et al., 2012), and silver nitrate staining for the bone was performed using Von Kossa Histology Stain Kit (Diagnostic Biosystems, KT 028) following the manufacturer’s instructions. β-galactosidase staining was performed as described (Jeong et al., 2004).
Micro-CT
P0 heads were scanned with Skyscan at 40 kV, 250 µA, and 7.47 µm resolution, and reconstructed using NRecon program.
RNA in situ hybridization
Whole mount RNA in situ hybridization was performed as described using a digoxigenin-labeled probe (Wilkinson and Nieto, 1993). Section RNA in situ hybridization was performed on frozen sections as described (Schaerenwiemers and Gerfinmoser, 1993). The templates for RNA probes for Acan, Bmp4, Dlx5, Runx2, Sp7, Sox9, and Wnt7b were obtained from other researchers, and further information is available upon request. The template for Lmx1b probe was PCR-amplified from a commercial full-length cDNA clone (Open Biosystems, ID 40129984), adding a T3 polymerase-binding site. Primer sequences are provided in S1 Table.
Immunofluorescence and immunohistochemistry
Immunofluorescence was performed on frozen sections as described (Jeong and McMahon, 2005). The primary antibodies were rabbit anti-Ki67 (Abcam, ab15580, 1:400), rabbit anti-cleaved caspase-3 (Cell Signaling Technology, #9661, 1:300), and rabbit anti-Sp7 (Abcam, ab209484, 1:1000). Alexa Fluor secondary antibodies (Invitrogen) were used for detection. 4',6-diamidino-2-phenylindole (DAPI) was used to stain nuclei. The numbers of cells were counted manually from two sections per embryo. Two-tailed Student’s t-test was used for comparison of cell proliferation rates.
Immunohistochemisty for p-Smad1/5/9 was performed using rabbit anti-phospho-Smad1(Ser463/465)/ Smad5(Ser463/465)/ Smad9(Ser465/467) antibody from Cell Signaling Technology (#13820, 1:800). TSA Biotin Kit (Perkin-Elmer, NEL700A001KT) was used to amplify the signal and 3,3’-Diaminobenzidine tablets (DAB) (Sigma-Aldrich, D4293) were used for detection.
Western blot
For western blot comparing EMM and SOM of wild type embryos, the tissue was dissected from 11~16 of E12.5 wild type CD-1 embryos, as illustrated in S1 Fig, to prepare one lysate sample for each region. Three samples of EMM and SOM each were used for the quantitative analysis described below. For western blot comparing EMM of control and Lmx1b LOF mutant embryos, EMM was dissected as in S1 Fig and three samples per genotype were analyzed, where each sample consisted of EMM from 4 embryos.
The tissue was homogenized and lysed in radioimmunoprecipitation assay (RIPA) buffer containing Halt phosphatase inhibitor (Thermo Scientific, 78420) and Complete mini protease inhibitor (Sigma-Aldrich, 11836153001). Protein concentration of the lysate was determined by Bradford assay (Bradford, 1976), and 2~20 ug of total protein was loaded to each lane of a gel, depending on the marker to be examined. The primary antibodies were rabbit anti-pSMAD1/5/9 (Cell signaling #13820, 1:1000), mouse anti-active β-catenin (ABC) clone 8E7 (Millipore 05-665, 1:2000), rabbit anti-pErk1/2 (p44/42 MAPK) (Cell signaling #4370, 1:2000), rabbit anti-p38 MAPK (Cell signaling #9211, 1:1000) and rabbit anti-β-actin (Cell signaling #8457, 1:1000). The secondary antibodies were goat anti-mouse IgG horseradish peroxidase (HRP) (Thermo Fisher A10551, 1:1000) and donkey anti-rabbit IgG-HRP (Novex A16023, 1:10,000). Following enhanced chemiluminescence reaction with SuperSignal West Dura Extended Duration Substrate (Thermo Fisher, 37071), the signals were detected by Gel Doc (BioRad). The intensities of the bands were quantified using Image Lab (BioRad), and they were normalized against the intensity of the β-actin band from each sample. Two-tailed Student’s t-test was used for statistical comparison.
Reverse transcription and real-time quantitative PCR (RT-qPCR)
EMM and/or SOM were dissected from E12.5 embryos as in S1 Fig. EMM from E11.5 embryos was dissected as in S1 Fig except that the ectoderm was not removed due to a technical difficulty. The tissue was homogenized using a pestle and QIAshredder (Qiagen), and total RNA was extracted using RNeasy Mini Kit (Qiagen). Reverse-transcription was performed using Transcriptor First Strand cDNA Synthesis Kit (Roche, 04379012001). qPCR was performed using Power SYBR Green PCR Master Mix (Applied Biosystems, 4367659), and a housekeeping gene Peptidylpropyl isomerase (Ppib) was used as an internal standard (Pachot et al., 2004). mRNA levels of all the genes were normalized against Ppib level within each sample, using a formula 2^(−ΔCt(Gene of interest − Ppib))×1000, and compared across different samples. For all the genes, >5 samples per genotype were analyzed and two-tailed Student’s t-test was used for a statistical comparison. The primers used for qPCR are listed in S1 Table. Most primer sequences were from PrimerBank (Spandidos et al., 2010).
Results
Key osteogenic signaling pathways are active in both EMM and SOM
To understand why EMM does not generate an ossification center while SOM does, we considered two possibilities: absence of osteogenic signals in the dorsal head, or presence of anti-osteogenic factors at this location.
The first possibility seemed unlikely based on the previously documented expression patterns of the genes encoding bone morphogenetic proteins (BMP), wingless-type MMTV integration site family proteins (WNT), and fibroblast growth factors (FGF), which are well known to promote osteogenic differentiation of mesenchyme cells (Long, 2011; Senarath-Yapa et al., 2013; Wu et al., 2016). The dorsal forebrain expresses multiple Bmp genes at mid-gestation, and Bmp4 continues to be highly expressed in the diencephalon roof plate (Furuta et al., 1997; Han et al., 2007) (Fig 2A,B). Similarly, several Wnt genes including Wnt1, Wnt3, Wnt3a, Wnt7b and Wnt8b are expressed in the dorsal brain (Harrison-Uy and Pleasure, 2012; Roelink and Nusse, 1991; Wilkinson et al., 1987), in the pallium of the telencephalon (Fig 2C) and/or dorsal midline of the diencephalon and the mesencephalon. Also, Fgf8 is expressed in the dorsal diencephalon (Crossley and Martin, 1995; Kataoka and Shimogori, 2008). The primary roles of these secreted proteins are to regulate the growth and patterning of the brain (Chizhikov and Millen, 2005; Harrison-Uy and Pleasure, 2012; Hoch et al., 2009; Kataoka and Shimogori, 2008; Martinez-Ferre and Martinez, 2012). However, because EMM is located close to the dorsal brain, we thought that the BMP, WNT, and FGF signals from the brain would also be readily available to EMM.
A) The head of an E13.5 wild type mouse embryo processed by whole mount RNA in situ hybridization. B,C) Coronal sections of an E13.5 head processed by RNA in situ hybridization. The arrows in A and B point to the strong expression of Bmp4 in the diencephalon roof plate (DR) of the forebrain. The arrow in C points to the expression of Wnt7b in the telencephalon pallium (TP) of the forebrain. D,E) Coronal sections of an E13.5 head processed by immunohistochemistry for phosphorylated (p)-SMAD1/5/9. Pa: parietal bone rudiment in SOM. The arrow in E points to p-SMAD1/5/9 detected in EMM adjacent to DR. The dotted line in E is the boundary between DR and EMM. F) Dissection of EMM and SOM from an E12.5 head. See S1 Fig for details. G) Western blot analysis of EMM and SOM at E12.5 for an active form of intracellular signal transducers. β-actin is shown as a loading control. H) Quantitative comparison of the western blot results from EMM and SOM. The band intensity of each marker was normalized against the β-actin band from the same sample. Data from triplicate samples are shown (diamonds), along with the average (horizontal bar) and the standard deviation (error bars). **: p<0.01. n.s.: not significantly different (p>0.05). Bars in A–C,F: 0.5 mm. Bars in D,E: 0.1 mm.
To confirm that the cells in EMM are indeed receiving the above signals, we examined the intracellular mediators of BMP, WNT, and FGF signaling. Binding of BMP to its receptors can elicit cellular responses through SMAD1/5/9 or p38 MAPK (= MAPK14, mitogen activated protein kinase 14) (Wu et al., 2016). Canonical WNT signaling leads to stabilization of β-catenin, which then controls the transcription of target genes (Nusse and Clevers, 2017; Yang, 2012). FGF was shown to use signal transduction pathways involving extracellular signal-regulated kinase 1 and 2 (ERK1/2, = MAPK1/MAPK3) or p38 MAPK in the context of osteogenesis (Ornitz and Marie, 2015; Rodriguez-Carballo et al., 2016). By immunohistochemistry, we found phosphorylated (p)-SMAD1/5/9 in both EMM and SOM (Fig 2D,E). In addition, we isolated EMM and SOM (Fig 2F, S1 Fig) and performed western blot, which detected activated forms of SMAD1/5/9, β-catenin, p38 MAPK, and ERK1/2 in both EMM and SOM (Fig 2G). A quantitative analysis revealed that, on average, EMM cells had less p-SMAD1/5/9 than SOM cells (Fig 2H). In contrast, more p-ERK1/2 was found in EMM than in SOM although the difference was not statistically significant (p=0.07).
It requires further investigation to determine whether the observed difference in the levels of p-SMAD1/5/9 or p-ERK1/2 contributes to the difference in the osteogenic abilities of EMM and SOM. Nonetheless, our result indicated that the lack of osteogenesis in EMM during normal development was not due to the complete absence of the inductive signals in this region. Therefore, we moved our focus to testing the other hypothesis, that there are anti-osteogenic factor(s) actively repressing ossification from EMM.
Lmx1b is expressed in the head mesenchyme in a pattern complementary to osteogenic genes at the beginning of calvarial bone development
To identify candidates for an anti-osteogenic factor in EMM, we searched the literature for genes whose inactivation led to excess ossification in the calvaria.
An earlier report showed that mouse mutants for Lmx1b, a member of LIM-homeobox family transcription factor genes, were born with a dysmorphic head in which all the sutures and fontanelles of the calvaria were replaced with a continuous sheet of bone (Chen et al., 1998a). While this phenotype suggested an anti-osteogenic function of Lmx1b, no details were reported on the cellular and molecular changes underlying the calvarial phenotype of Lmx1b−/− mutants. In fact, it remained unclear whether Lmx1b directly regulated calvarial development. Lmx1b is expressed in the brain, and Lmx1b−/− mutants exhibited various brain defects (Guo et al., 2007). The expression of secreted proteins was altered in their brain, and the midbrain and the hindbrain suffered from severe hypoplasia (Guo et al., 2007). Given the close interactions between the brain and the skull during development (Twigg and Wilkie, 2015), it was possible that the calvarial defects of Lmx1b−/− mutants were secondary to the perturbation in the signaling or mechanical influences from the brain.
To define the role of Lmx1b in calvarial development, first we examined its expression in the head mesenchyme from E9.5 to late stages of gestation by RNA in situ hybridization (Fig 3). Lmx1b was broadly expressed in the head mesenchyme and in the overlying surface ectoderm at E9.5 and E10.5 (Fig 3A–D). However, the expression at E9.5 was in a gradient declining toward the optic vesicle (Fig 3B), and by E10.5, a zone of Lmx1b-negative mesenchyme was established just above the eye (Fig 3D). At E12.5, the rudiments for the frontal bone and the parietal bone form within the head mesenchyme, expressing the early markers of osteogenic specification, distal-less homeobox 5 (Dlx5), runt related transcription factor 2 (Runx2), and Sp7 (also known as Osx) (Ishii et al., 2015; Karsenty, 2008; Long, 2011; Twigg and Wilkie, 2015). We found that these osteogenic genes were turned on in SOM where Lmx1b was absent, but not in EMM where Lmx1b was expressed (Fig 3E–M). We also performed reverse transcription followed by quantitative real time PCR (RT-qPCR) for Lmx1b, Dlx5, and Runx2 on EMM and SOM dissected from wild type embryos as in Fig 2F and S1 Fig. These results were consistent with RNA in situ hybridization data (Fig 3N), which validated the accuracy of our dissection. Examination of transverse sections showed that Lmx1b expression was absent or very low in the entire antero-posterior extent of SOM (Fig. 3O–Q).
A–M) Coronal sections of the head of wild type embryos processed by RNA in situ hybridization. Apical (Ap) – basal (Ba) axis is indicated in A. B and D are an enlargement of the boxed areas in A and C, respectively. H–M are an enlargement of the boxed areas in E–G. The bracket in D indicates Lmx1b-negative mesenchyme above the eye. The arrowhead in E points to the basal margin of Lmx1b expression in the mesenchyme. The arrow in H points to Lmx1b expression in EMM. The arrows in L and M indicate the frontal bone rudiment. N) EMM and SOM were isolated from E12.5 wild type embryos as in Fig 2F, and analyzed by reverse transcription (RT) followed by quantitative real-time PCR (qPCR) for gene expression. The expression levels were normalized using a house keeping gene, Ppib, as an internal standard. Data from six embryos are shown in each chart (diamonds), along with the average (horizontal bar) and the standard deviation (error bars). ***: p<0.001. O) A schematic of an E13.5 head showing the plane of the sections in P and Q. P,Q) Horizontal sections of the head from an E13.5 wild type embryo processed by RNA in situ hybridization. Anterior (An) – posterior (Po) axis is indicated in P. The brackets indicate SOM. R–T) Coronal sections of the head of an E14.5 wild type embryo processed by RNA in situ hybridization. The arrowhead in R points to the basal margin of Lmx1b expression in the mesenchyme. Br: brain, co: coronal suture, Ec: ectoderm, Ey: eye, Fr: frontal bone, HM: head mesenchyme, OV: optic vesicle, Pa: parietal bone. Bars in A,C: 0.1 mm, bars in E,P,R: 0.5 mm.
Later in development, as the calvarial bones grew apically, Lmx1b was rapidly down-regulated in both the area and the intensity of expression (Fig 3R–T). By E16.5, Lmx1b expression became undetectable in the calvaria by RNA in situ hybridization.
The above data suggested that Lmx1b could directly regulate calvarial development through its expression in the embryonic head mesenchyme, and that its role could be to restrict the areas of bone formation within the calvarial primordium.
Head mesenchyme-specific inactivation of Lmx1b leads to craniosynostosis affecting multiple sutures
To directly test whether Lmx1b expressed within the head mesenchyme is essential to calvarial development, we generated a tissue-specific Lmx1b knockout mutant using Prrx1-Cre (Logan et al., 2002) (Prrx1-Cre;Lmx1bfl/−, referred to as Lmx1b LOF). Prrx1-Cre is active broadly in the head mesenchyme from ~E9.5, but not in the brain or the surface ectoderm. While Prrx1-Cre acts in both SOM and EMM (Logan et al., 2002) (Fig 4A), it is less efficient in the anterior part of the head mesenchyme than in the posterior part (Logan et al., 2002). Accordingly, mapping of Prrx1-Cre domains at E18.5 showed that the anterior parts of the frontal bone and the interfrontal suture were labeled in a highly mosaic fashion. Other calvarial bones and sutures were almost completely labeled (S2 Fig).
A) A coronal section of the head of an E13.5 Prrx1-Cre;R26R-LacZ/+ embryo processed by β-galactosidase staining (blue) and counter-stained with nuclear fast red. B,C) The head of a control (Lmx1bfl/+) and a Prrx1-Cre;Lmx1bfl/− mutant (= Lmx1b LOF) at P0. Abnormal bulges in the mutant are indicated in C. D,E) Lateral views of the skull stained with Alizarin red for bone and Alcian blue for cartilage. The arrow in E points to a hole in the skull created by the protruding brain. F–I) Dorsal views of the calvaria stained with Alizarin red (F,G) or reconstructed from micro-CT scans (H,I). The arrows in G and I point to fused sutures. J,K) Coronal sections of the skull from micro-CT scans at the position of the dotted line in F. L–N) Coronal sections of the head processed by von Kossa staining (L,M) or β-galactosidase staining (N). M and N are adjacent sections from a Prrx1-Cre;Lmx1b;R26R-LacZ/+ mutant. The arrow in M points to the bone in place of the mid-sutural mesenchyme in the mutant. O,P) Superior views of E16.5 calvariae stained with Alizarin red. The arrow in P points to the heterotopic ossification center at the vertex. The arrowheads in P point to the basal part of the coronal sutures. Br: brain, co: coronal suture, de: dermis, Fr: frontal bone, if: interfrontal suture, IP: interparietal bone, la: lambdoidal suture, Pa: parietal bone, sa: sagittal suture. Bar: 1 mm.
As with Lmx1b−/− mutants, Lmx1b LOF newborns had a dysmorphic head (n=18), with bulges in the front and back of the head (Fig 4B,C). The skulls of 14 mutants were examined by skeletal staining or micro-computed tomography (micro-CT) at P0. All of them exhibited craniosynostosis affecting multiple sutures. The interfrontal suture and both coronal sutures were fused invariably, and at least one lambdoidal suture was fused in almost all of the mutants (13 out of 14 pups) (Fig 4F–K). Apparently, the reduction in the skull vault due to craniosynostosis caused the brain to protrude through the posterior part of the calvaria, making a big hole at the location of the sagittal suture and the posterior fontanelle (Fig 4C,E,G).
LMX1B is necessary to suppress osteogenic specification of EMM
To identify the primary changes responsible for the craniosynostosis in Lmx1b LOF mutants, we examined calvarial development at progressively earlier stages. The sections of Lmx1b LOF head at E18.5 showed the presence of bone within what should be the mid-sutural mesenchyme of the interfrontal suture (Fig 4L,M; n=4). In contrast, the overlying dermal mesenchyme did not undergo ossification despite the deletion of Lmx1b by Prrx1-Cre in this tissue as well (Fig 4M,N).
At E16.5, the frontal bones and the parietal bones from left and right were growing toward the dorsal midline of the head, but they were still at a distance from the midline in both the control and Lmx1b LOF mutant embryos (Fig 4O,P). However, the mutants had an additional ossification center at the vertex, which replaced the anterior fontanelle and bridged all four pieces of the surrounding endogenous bone (Fig 4P; n=4). This phenotype suggested de novo heterotopic ossification to be a major factor for the craniosynostosis of Lmx1b LOF mutants although overgrowth of the endogenous bones could have contributed as well. It was also evident from E16.5 skulls that the coronal sutures of the mutants were initially established on the basal side, where they arise from SOM (Deckelbaum et al., 2012), but lost on the apical side because of the ossification from the vertex (Fig 4P). Combined with the lack of Lmx1b expression in SOM (Fig 3P,Q), this result indicated that Lmx1b was not necessary for specification of coronal suture progenitors within SOM (see Discussion).
Consistent with the skeletal phenotype, we found that osteogenic genes Runx2 and Sp7 were ectopically expressed in the vertex mesenchyme of Lmx1b LOF mutants at E14.5 (Fig 5A–D; n=4) (the terms ‘heterotopic’ and ‘ectopic’ are used in this paper following the recommendations of (Sarnat, 1995): an anatomical structure at an abnormal location within the correct organ is called heterotopic, whereas gene expression at an abnormal location is called ectopic even if it is within the correct organ). Separately from the ectopic Sp7 at the vertex (Fig 5E,F), Sp7 associated with the SOM-derived bone rudiments also appeared to extend more apically in Lmx1b LOF mutants than in controls (Fig 5G,H; n=4). However, at this point, we do not know whether this extension of Sp7 is from SOM cells that have migrated farther than normal, or from the mutant EMM cells that have been induced to undergo osteogenesis, possibly by the signals from the osteogenic front.
A–D) Coronal sections of E14.5 heads processed by RNA in situ hybridization. The apical end of the head is shown. The arrows in B and D point to the ectopic expression of Runx2 and Sp7 at the vertex. E–H) Coronal sections of E13.5 heads processed by immunofluorescence for Sp7. E and F are from the boxed areas in G and H, respectively. The arrow in F points to the ectopic Sp7 at the vertex. The arrow in H points to the apparent apical extension of Sp7 from SOM of the mutant. I,J) RT-qPCR comparison of gene expression in EMM of control and Lmx1b LOF mutant embryos. The expression level of each gene was normalized to the internal standard Ppib. Results from individual samples (5 to 6 per genotype) are presented as diamonds, along with the average (horizontal bar) and the standard deviation (error bars). *: p<0.05, **: p<0.01, ***: p<0.001, n.s.: not significantly different (p>0.05). K) Western blot from EMM of control and Lmx1b LOF mutant embryos. L) Quantitative comparison of the western blot results. 3 lysate samples per genotype were used for the western blot, where each lysate sample was made from EMM of four embryos. M,N) Coronal sections of E13.5 heads processed by immunofluorescence for Ki67 (red). The nuclei were stained with DAPI (blue). O) Ki67 cells and total cells were counted from EMM demarcated in H and I, and the percentage of Ki67 cells was calculated. Results from three embryos per genotype are presented along with the average and the standard deviation. Bar: 0.1 mm.
Through RT-qPCR, the up-regulation of Dlx5 and Runx2 was detectable in the mutant EMM at E11.5–E12.5, around the same time when osteogenesis normally begins in SOM (Fig 5I,J). These results indicated that at least some of the EMM cells of Lmx1b LOF mutants were incorrectly specified into the osteogenic fate. Furthermore, the early onset of the phenotype was in line with the fact that the region-specific expression of Lmx1b was established in the head mesenchyme as early as E10.5 (Fig 3). Together, our results supported the idea that Lmx1b is required to pattern the head mesenchyme preceding the actual calvarial bone formation.
To better understand the changes in the mutant EMM, we examined additional genes that were known to be expressed in the head mesenchyme during normal development. Engrailed1 (En1) encodes a homeobox transcription factor, and it is expressed in SOM from ~E11 (Deckelbaum et al., 2006). We did not detect up-regulation of En1 in Lmx1b LOF mutants (Fig 5J). This suggested that although the mutant EMM had gained osteogenic competence, it did not fully take on the SOM identity.
Msh homeobox genes Msx1 and Msx2 are expressed in both EMM and SOM, and they promote calvarial bone formation from SOM while repressing it in EMM (Han et al., 2007; Roybal et al., 2010; Wilkie et al., 2000). We found that Msx2, but not Msx1, was significantly down-regulated in EMM of Lmx1b LOF mutants (Fig 5J). Although this change is consistent with the possibility that Lmx1b inhibits osteogenesis in EMM at least in part through Msx2, it cannot account for the calvarial phenotype of Lmx1b LOF mutants on its own. Mutation of Msx genes led to heterotopic ossification only when all four copies of Msx1;Msx2 were deleted (Roybal et al., 2010). Twist1, encoding a basic helix-loop-helix transcription factor, is expressed throughout the head mesenchyme just like Msx1 and Msx2 (Rice et al., 2000). Twist1 is required for osteogenic specification of SOM, but inhibits later steps of osteoblast differentiation (Bialek et al., 2004; Goodnough et al., 2012; Goodnough et al., 2016). Its expression was unaffected in Lmx1b LOF mutants (Fig 5J).
We also examined the activities of major osteogenic signaling pathways, namely, BMP-SMAD pathway, canonical WNT pathway, p38 MAPK pathway, and FGF-ERK pathway. We did not find a consistent difference between the control EMM and Lmx1b LOF mutant EMM that could explain the mutant phenotype (Fig 5K,L). Similarly, we did not detect a significant difference in the expression of Axin2 and Gli1 (Fig 5J), which are transcriptional read-outs of the canonical WNT pathway and Hedgehog pathway, respectively (Jho et al., 2002; McMahon et al., 2003). These data supported the notion that was introduced in Fig 2, that the osteogenic signaling activities are not the absolute limiting factor preventing osteogenesis in EMM. For example, even though the level of p-SMAD1/5/9 in wild type EMM was lower than in wild type SOM (Fig 2), this level was sufficient for osteogenesis in Lmx1b LOF mutant EMM (Fig 5L). In addition, based on our results, it is unlikely that Lmx1b regulates osteogenesis by modulating the activation of the above four signaling pathways. Rather, we posit that LMX1B functions independently of these signals to counter their pro-osteogenic effect.
We investigated the effect of the loss of LMX1B on proliferation and apoptosis of EMM cells, using Ki67 and cleaved caspase-3 as markers, respectively (Fernandes-Alnemri et al., 1994; Scholzen and Gerdes, 2000); we did not find a significant difference between controls and Lmx1b LOF mutants (Fig 5M–O, S3 Fig). Combined, our analysis of Lmx1b LOF mutants revealed that Lmx1b was necessary to inhibit osteogenic specification of EMM.
LMX1B is sufficient to inhibit osteogenic specification of SOM
To obtain further evidence that LMX1B has an anti-osteogenic activity in the developing calvaria, we performed a gain-of-function (GOF) experiment of overexpressing Lmx1b in the head mesenchyme. We used a mouse line in which Lmx1b coding sequence was inserted into ROSA26 locus following a floxed stop cassette (R26Lmx1b) (Li et al., 2010). In embryos with both R26Lmx1b and Prrx1-Cre (Prrx1-Cre;R26Lmx1b/+, referred to as Lmx1b GOF), LMX1B was constitutively expressed in the head mesenchyme including SOM, which normally does not express Lmx1b.
At birth, all of the Lmx1b GOF mutants had severely diminished calvarial bone (Fig 6A–D; n=17). The parietal bone was affected more than the frontal bone, which was consistent with the difference in Prrx1-Cre activity between the two parts of the calvaria (S2 Fig). Based on the labeling by a Cre reporter, Lmx1b-overexpressing cells populated the normal location of the parietal bone at P0, but they were present as cartilage and a layer of loose mesenchyme instead of bone (Fig 6E–G; n=2). The residual parietal bone in the mutants was associated with a small number of cells in which Prrx1-Cre was inactive (Fig 6H,I; n=2)
A–D) Lateral views (A,B) and dorsal views (C,D) of the skull of a control and Prrx1-Cre;R26Lmx1b/+ mutant (= Lmx1b GOF) newborns. The skulls were stained with Alizarin red and Alcian blue. E–I) Coronal sections of P0 heads along the dotted line in A. F,G and H,I are pairs of adjacent sections from a Prrx1-Cre;R26Lmx1b/R-LacZ mutant. E, F, and H were stained for bone (brown), and G and I were stained for β-galactosidase activity (blue). In F and G, arrows and arrowheads indicate a layer of loose mesenchyme and cartilage, respectively, in the parietal region of the mutant head. In I, the open arrowhead points to the cells within the mutant calvaria in which Prrx1-Cre was inactive. de: dermis, Fr: frontal bone, IP: interparietal bone, Pa: parietal bone. Bars in A and C: 1 mm, bars in E and H: 0.2 mm.
Examination of osteogenic genes indicated that the initial specification of SOM cells into the osteoblast fate was disrupted in Lmx1b GOF mutants (Fig 7; n=3). At E12.5, Dlx5 and Runx2 were expressed in the parietal bone primordium of control embryos, but they were very weakly expressed at the equivalent position within SOM of the mutants (Fig 7A–D). Because Msx2 was down-regulated in EMM of Lmx1b LOF mutants (Fig 5F), we also examined it in Lmx1b GOF mutants. Compared with the control embryos, Lmx1b GOF mutants had decreased expression of Msx2 in the parietal bone primordium but increased expression in the mesenchyme immediately apical to it, indicating that the regulatory relationship between Lmx1b and Msx2 is complex (Fig 7E,F; n=3). By E14.5, there was no expression of osteogenic genes Dlx5, Runx2, and Sp7 in the presumptive parietal bone region of the mutants (Fig 7G–L; n=2). As shown at P0, Lmx1b-overexpressing cells were properly occupying the parietal region at E13.5, but they failed to make a skeletogenic condensation (Fig 7Q,R; n=3).
A–P) Coronal sections of the heads through the parietal bone rudiment (Pa) processed by RNA in situ hybridization. In E and F, arrows and arrowheads indicate down-regulation and up-regulation, respectively, of Msx2 in the Lmx1b GOF mutant. Q,R) Coronal sections of the head of Prrx1-Cre;R26R-LacZ/+ (control) and Prrx1-Cre;R26Lmx1b/R-LacZ (Lmx1b GOF) embryos processed by β-galactosidase staining (blue). The arrows point to the absence of mesenchyme condensation in the mutant. S,T) Coronal sections of E13.5 heads processed by immunofluorescence for Ki67 (red). The nuclei were stained with DAPI (blue). U) Ki67 cells and total cells were counted from the parietal bone area demarcated in S and T, and the percentage of Ki67 cells was calculated. Results from three embryos per genotype are presented as diamonds, along with the average (horizontal bar) and the standard deviation (error bars). n.s.: not significantly different (p>0.05). Bar: 0.2 mm.
The fact that cartilage was found in P0 Lmx1b GOF mutants at a position comparable to the parietal bone of control pups (Fig 6E–G) prompted us to consider whether the Lmx1b-overexpressing cells had undergone a fate change from osteogenic to chondrogenic. Such change was shown in mouse mutants lacking canonical WNT signaling in the head mesenchyme, which also suffered from loss of calvarial bone (Day et al., 2005; Goodnough et al., 2012; Hill et al., 2005; Tran et al., 2010). Therefore, we examined the early markers of chondrocyte specification and differentiation, SRY-box 9 (Sox9) and Aggrecan (Acan) (de Crombrugghe et al., 2001). The expression of Sox9 and Acan was restricted to the basal domain of the head mesenchyme in both control and Lmx1b GOF mutant embryos, and they were not expressed in the presumptive parietal bone region of the mutants where the osteogenic markers were lost (Fig 7M–P; n=2). This result argued against the idea that the calvarial bone deficiency in Lmx1b GOF mutants was driven by a direct bone-to-cartilage fate conversion of mesenchymal progenitors. Overexpression of Lmx1b was not sufficient to specify the chondrogenic fate, though LMX1B clearly did not inhibit cells from becoming a part of the cartilage (Fig 6G). Rather, we believe that the undifferentiated mesenchyme between the cartilage and the dermis (Fig 6F,G) was what had become of the parietal bone progenitors upon overexpression of Lmx1b, and that the presence of the cartilage was a secondary consequence to the absence of the bone. During normal development, cartilage appears in the parietal region first, but it is degraded later as the parietal bone grows over it. This enzymatic removal of the cartilage is dependent on the signals from the bone (Holmbeck et al., 1999; Zhou et al., 2009).
Similar to Lmx1b LOF mutants, Lmx1b GOF mutants did not show a significant difference in proliferation or apoptosis of SOM cells from control embryos (Fig 7S–U, S3 Fig). Therefore, we conclude that LMX1B specifically inhibited osteogenic specification of SOM cells.
Anti-osteogenic function of Lmx1b is tissue-specific
Our results from both LOF and GOF experiments have established that LMX1B has an anti-osteogenic function in the developing calvaria. This raised a question whether it was true in other parts of the body, specifically in the limbs, where Lmx1b is strongly expressed during normal development and regulates dorso-ventral patterning (Chen et al., 1998a; Cygan et al., 1997; Riddle et al., 1995; Vogel et al., 1995). Because Prrx1-Cre has robust activity in limb buds (Logan et al., 2002), we examined the limbs of Lmx1b LOF and GOF mutants for any evidence of abnormal ossification.
The limb skeleton of Lmx1b LOF mutants showed a patterning defect such as loss of the ulna and the patella, which was reported previously (Fig 8A–D) (Chen et al., 1998b). However, we did not find evidence of heterotopic or excess ossification either from whole mount Alizarin red staining (Fig 8A–D; n=3) or von Kossa staining of the sections (Fig 8E,F; n=4). Similarly, the limb skeleton of Lmx1b GOF mutants was mildly abnormal in their shape (Fig 8G–J) (Li et al., 2010), but the level of ossification appeared grossly comparable to the control limbs based on Alizarin red staining (Fig 8G–J; n=4) and von Kossa staining (Fig 8K,L; n=2). These results indicated that the role of Lmx1b as an essential anti-osteogenic factor is specific to calvarial development.
A–D,G–J) Forelimbs (FL) and hind limbs (HL) stained with Alizarin red and Alcian blue. The arrows point to the morphological abnormalities in the mutant limbs. E,F,K,L) Sections through the humerus (Hum) processed by von Kossa staining. Neither Lmx1b LOF nor Lmx1b GOF mutants displayed an obvious ossification phenotype in the limbs. All four limbs from each animal were sectioned and examined, and the humerus is shown as a representative result. Bar: 1 mm.
Key osteogenic signaling pathways are active in both EMM and SOM
To understand why EMM does not generate an ossification center while SOM does, we considered two possibilities: absence of osteogenic signals in the dorsal head, or presence of anti-osteogenic factors at this location.
The first possibility seemed unlikely based on the previously documented expression patterns of the genes encoding bone morphogenetic proteins (BMP), wingless-type MMTV integration site family proteins (WNT), and fibroblast growth factors (FGF), which are well known to promote osteogenic differentiation of mesenchyme cells (Long, 2011; Senarath-Yapa et al., 2013; Wu et al., 2016). The dorsal forebrain expresses multiple Bmp genes at mid-gestation, and Bmp4 continues to be highly expressed in the diencephalon roof plate (Furuta et al., 1997; Han et al., 2007) (Fig 2A,B). Similarly, several Wnt genes including Wnt1, Wnt3, Wnt3a, Wnt7b and Wnt8b are expressed in the dorsal brain (Harrison-Uy and Pleasure, 2012; Roelink and Nusse, 1991; Wilkinson et al., 1987), in the pallium of the telencephalon (Fig 2C) and/or dorsal midline of the diencephalon and the mesencephalon. Also, Fgf8 is expressed in the dorsal diencephalon (Crossley and Martin, 1995; Kataoka and Shimogori, 2008). The primary roles of these secreted proteins are to regulate the growth and patterning of the brain (Chizhikov and Millen, 2005; Harrison-Uy and Pleasure, 2012; Hoch et al., 2009; Kataoka and Shimogori, 2008; Martinez-Ferre and Martinez, 2012). However, because EMM is located close to the dorsal brain, we thought that the BMP, WNT, and FGF signals from the brain would also be readily available to EMM.
A) The head of an E13.5 wild type mouse embryo processed by whole mount RNA in situ hybridization. B,C) Coronal sections of an E13.5 head processed by RNA in situ hybridization. The arrows in A and B point to the strong expression of Bmp4 in the diencephalon roof plate (DR) of the forebrain. The arrow in C points to the expression of Wnt7b in the telencephalon pallium (TP) of the forebrain. D,E) Coronal sections of an E13.5 head processed by immunohistochemistry for phosphorylated (p)-SMAD1/5/9. Pa: parietal bone rudiment in SOM. The arrow in E points to p-SMAD1/5/9 detected in EMM adjacent to DR. The dotted line in E is the boundary between DR and EMM. F) Dissection of EMM and SOM from an E12.5 head. See S1 Fig for details. G) Western blot analysis of EMM and SOM at E12.5 for an active form of intracellular signal transducers. β-actin is shown as a loading control. H) Quantitative comparison of the western blot results from EMM and SOM. The band intensity of each marker was normalized against the β-actin band from the same sample. Data from triplicate samples are shown (diamonds), along with the average (horizontal bar) and the standard deviation (error bars). **: p<0.01. n.s.: not significantly different (p>0.05). Bars in A–C,F: 0.5 mm. Bars in D,E: 0.1 mm.
To confirm that the cells in EMM are indeed receiving the above signals, we examined the intracellular mediators of BMP, WNT, and FGF signaling. Binding of BMP to its receptors can elicit cellular responses through SMAD1/5/9 or p38 MAPK (= MAPK14, mitogen activated protein kinase 14) (Wu et al., 2016). Canonical WNT signaling leads to stabilization of β-catenin, which then controls the transcription of target genes (Nusse and Clevers, 2017; Yang, 2012). FGF was shown to use signal transduction pathways involving extracellular signal-regulated kinase 1 and 2 (ERK1/2, = MAPK1/MAPK3) or p38 MAPK in the context of osteogenesis (Ornitz and Marie, 2015; Rodriguez-Carballo et al., 2016). By immunohistochemistry, we found phosphorylated (p)-SMAD1/5/9 in both EMM and SOM (Fig 2D,E). In addition, we isolated EMM and SOM (Fig 2F, S1 Fig) and performed western blot, which detected activated forms of SMAD1/5/9, β-catenin, p38 MAPK, and ERK1/2 in both EMM and SOM (Fig 2G). A quantitative analysis revealed that, on average, EMM cells had less p-SMAD1/5/9 than SOM cells (Fig 2H). In contrast, more p-ERK1/2 was found in EMM than in SOM although the difference was not statistically significant (p=0.07).
It requires further investigation to determine whether the observed difference in the levels of p-SMAD1/5/9 or p-ERK1/2 contributes to the difference in the osteogenic abilities of EMM and SOM. Nonetheless, our result indicated that the lack of osteogenesis in EMM during normal development was not due to the complete absence of the inductive signals in this region. Therefore, we moved our focus to testing the other hypothesis, that there are anti-osteogenic factor(s) actively repressing ossification from EMM.
Lmx1b is expressed in the head mesenchyme in a pattern complementary to osteogenic genes at the beginning of calvarial bone development
To identify candidates for an anti-osteogenic factor in EMM, we searched the literature for genes whose inactivation led to excess ossification in the calvaria.
An earlier report showed that mouse mutants for Lmx1b, a member of LIM-homeobox family transcription factor genes, were born with a dysmorphic head in which all the sutures and fontanelles of the calvaria were replaced with a continuous sheet of bone (Chen et al., 1998a). While this phenotype suggested an anti-osteogenic function of Lmx1b, no details were reported on the cellular and molecular changes underlying the calvarial phenotype of Lmx1b−/− mutants. In fact, it remained unclear whether Lmx1b directly regulated calvarial development. Lmx1b is expressed in the brain, and Lmx1b−/− mutants exhibited various brain defects (Guo et al., 2007). The expression of secreted proteins was altered in their brain, and the midbrain and the hindbrain suffered from severe hypoplasia (Guo et al., 2007). Given the close interactions between the brain and the skull during development (Twigg and Wilkie, 2015), it was possible that the calvarial defects of Lmx1b−/− mutants were secondary to the perturbation in the signaling or mechanical influences from the brain.
To define the role of Lmx1b in calvarial development, first we examined its expression in the head mesenchyme from E9.5 to late stages of gestation by RNA in situ hybridization (Fig 3). Lmx1b was broadly expressed in the head mesenchyme and in the overlying surface ectoderm at E9.5 and E10.5 (Fig 3A–D). However, the expression at E9.5 was in a gradient declining toward the optic vesicle (Fig 3B), and by E10.5, a zone of Lmx1b-negative mesenchyme was established just above the eye (Fig 3D). At E12.5, the rudiments for the frontal bone and the parietal bone form within the head mesenchyme, expressing the early markers of osteogenic specification, distal-less homeobox 5 (Dlx5), runt related transcription factor 2 (Runx2), and Sp7 (also known as Osx) (Ishii et al., 2015; Karsenty, 2008; Long, 2011; Twigg and Wilkie, 2015). We found that these osteogenic genes were turned on in SOM where Lmx1b was absent, but not in EMM where Lmx1b was expressed (Fig 3E–M). We also performed reverse transcription followed by quantitative real time PCR (RT-qPCR) for Lmx1b, Dlx5, and Runx2 on EMM and SOM dissected from wild type embryos as in Fig 2F and S1 Fig. These results were consistent with RNA in situ hybridization data (Fig 3N), which validated the accuracy of our dissection. Examination of transverse sections showed that Lmx1b expression was absent or very low in the entire antero-posterior extent of SOM (Fig. 3O–Q).
A–M) Coronal sections of the head of wild type embryos processed by RNA in situ hybridization. Apical (Ap) – basal (Ba) axis is indicated in A. B and D are an enlargement of the boxed areas in A and C, respectively. H–M are an enlargement of the boxed areas in E–G. The bracket in D indicates Lmx1b-negative mesenchyme above the eye. The arrowhead in E points to the basal margin of Lmx1b expression in the mesenchyme. The arrow in H points to Lmx1b expression in EMM. The arrows in L and M indicate the frontal bone rudiment. N) EMM and SOM were isolated from E12.5 wild type embryos as in Fig 2F, and analyzed by reverse transcription (RT) followed by quantitative real-time PCR (qPCR) for gene expression. The expression levels were normalized using a house keeping gene, Ppib, as an internal standard. Data from six embryos are shown in each chart (diamonds), along with the average (horizontal bar) and the standard deviation (error bars). ***: p<0.001. O) A schematic of an E13.5 head showing the plane of the sections in P and Q. P,Q) Horizontal sections of the head from an E13.5 wild type embryo processed by RNA in situ hybridization. Anterior (An) – posterior (Po) axis is indicated in P. The brackets indicate SOM. R–T) Coronal sections of the head of an E14.5 wild type embryo processed by RNA in situ hybridization. The arrowhead in R points to the basal margin of Lmx1b expression in the mesenchyme. Br: brain, co: coronal suture, Ec: ectoderm, Ey: eye, Fr: frontal bone, HM: head mesenchyme, OV: optic vesicle, Pa: parietal bone. Bars in A,C: 0.1 mm, bars in E,P,R: 0.5 mm.
Later in development, as the calvarial bones grew apically, Lmx1b was rapidly down-regulated in both the area and the intensity of expression (Fig 3R–T). By E16.5, Lmx1b expression became undetectable in the calvaria by RNA in situ hybridization.
The above data suggested that Lmx1b could directly regulate calvarial development through its expression in the embryonic head mesenchyme, and that its role could be to restrict the areas of bone formation within the calvarial primordium.
Head mesenchyme-specific inactivation of Lmx1b leads to craniosynostosis affecting multiple sutures
To directly test whether Lmx1b expressed within the head mesenchyme is essential to calvarial development, we generated a tissue-specific Lmx1b knockout mutant using Prrx1-Cre (Logan et al., 2002) (Prrx1-Cre;Lmx1bfl/−, referred to as Lmx1b LOF). Prrx1-Cre is active broadly in the head mesenchyme from ~E9.5, but not in the brain or the surface ectoderm. While Prrx1-Cre acts in both SOM and EMM (Logan et al., 2002) (Fig 4A), it is less efficient in the anterior part of the head mesenchyme than in the posterior part (Logan et al., 2002). Accordingly, mapping of Prrx1-Cre domains at E18.5 showed that the anterior parts of the frontal bone and the interfrontal suture were labeled in a highly mosaic fashion. Other calvarial bones and sutures were almost completely labeled (S2 Fig).
A) A coronal section of the head of an E13.5 Prrx1-Cre;R26R-LacZ/+ embryo processed by β-galactosidase staining (blue) and counter-stained with nuclear fast red. B,C) The head of a control (Lmx1bfl/+) and a Prrx1-Cre;Lmx1bfl/− mutant (= Lmx1b LOF) at P0. Abnormal bulges in the mutant are indicated in C. D,E) Lateral views of the skull stained with Alizarin red for bone and Alcian blue for cartilage. The arrow in E points to a hole in the skull created by the protruding brain. F–I) Dorsal views of the calvaria stained with Alizarin red (F,G) or reconstructed from micro-CT scans (H,I). The arrows in G and I point to fused sutures. J,K) Coronal sections of the skull from micro-CT scans at the position of the dotted line in F. L–N) Coronal sections of the head processed by von Kossa staining (L,M) or β-galactosidase staining (N). M and N are adjacent sections from a Prrx1-Cre;Lmx1b;R26R-LacZ/+ mutant. The arrow in M points to the bone in place of the mid-sutural mesenchyme in the mutant. O,P) Superior views of E16.5 calvariae stained with Alizarin red. The arrow in P points to the heterotopic ossification center at the vertex. The arrowheads in P point to the basal part of the coronal sutures. Br: brain, co: coronal suture, de: dermis, Fr: frontal bone, if: interfrontal suture, IP: interparietal bone, la: lambdoidal suture, Pa: parietal bone, sa: sagittal suture. Bar: 1 mm.
As with Lmx1b−/− mutants, Lmx1b LOF newborns had a dysmorphic head (n=18), with bulges in the front and back of the head (Fig 4B,C). The skulls of 14 mutants were examined by skeletal staining or micro-computed tomography (micro-CT) at P0. All of them exhibited craniosynostosis affecting multiple sutures. The interfrontal suture and both coronal sutures were fused invariably, and at least one lambdoidal suture was fused in almost all of the mutants (13 out of 14 pups) (Fig 4F–K). Apparently, the reduction in the skull vault due to craniosynostosis caused the brain to protrude through the posterior part of the calvaria, making a big hole at the location of the sagittal suture and the posterior fontanelle (Fig 4C,E,G).
LMX1B is necessary to suppress osteogenic specification of EMM
To identify the primary changes responsible for the craniosynostosis in Lmx1b LOF mutants, we examined calvarial development at progressively earlier stages. The sections of Lmx1b LOF head at E18.5 showed the presence of bone within what should be the mid-sutural mesenchyme of the interfrontal suture (Fig 4L,M; n=4). In contrast, the overlying dermal mesenchyme did not undergo ossification despite the deletion of Lmx1b by Prrx1-Cre in this tissue as well (Fig 4M,N).
At E16.5, the frontal bones and the parietal bones from left and right were growing toward the dorsal midline of the head, but they were still at a distance from the midline in both the control and Lmx1b LOF mutant embryos (Fig 4O,P). However, the mutants had an additional ossification center at the vertex, which replaced the anterior fontanelle and bridged all four pieces of the surrounding endogenous bone (Fig 4P; n=4). This phenotype suggested de novo heterotopic ossification to be a major factor for the craniosynostosis of Lmx1b LOF mutants although overgrowth of the endogenous bones could have contributed as well. It was also evident from E16.5 skulls that the coronal sutures of the mutants were initially established on the basal side, where they arise from SOM (Deckelbaum et al., 2012), but lost on the apical side because of the ossification from the vertex (Fig 4P). Combined with the lack of Lmx1b expression in SOM (Fig 3P,Q), this result indicated that Lmx1b was not necessary for specification of coronal suture progenitors within SOM (see Discussion).
Consistent with the skeletal phenotype, we found that osteogenic genes Runx2 and Sp7 were ectopically expressed in the vertex mesenchyme of Lmx1b LOF mutants at E14.5 (Fig 5A–D; n=4) (the terms ‘heterotopic’ and ‘ectopic’ are used in this paper following the recommendations of (Sarnat, 1995): an anatomical structure at an abnormal location within the correct organ is called heterotopic, whereas gene expression at an abnormal location is called ectopic even if it is within the correct organ). Separately from the ectopic Sp7 at the vertex (Fig 5E,F), Sp7 associated with the SOM-derived bone rudiments also appeared to extend more apically in Lmx1b LOF mutants than in controls (Fig 5G,H; n=4). However, at this point, we do not know whether this extension of Sp7 is from SOM cells that have migrated farther than normal, or from the mutant EMM cells that have been induced to undergo osteogenesis, possibly by the signals from the osteogenic front.
A–D) Coronal sections of E14.5 heads processed by RNA in situ hybridization. The apical end of the head is shown. The arrows in B and D point to the ectopic expression of Runx2 and Sp7 at the vertex. E–H) Coronal sections of E13.5 heads processed by immunofluorescence for Sp7. E and F are from the boxed areas in G and H, respectively. The arrow in F points to the ectopic Sp7 at the vertex. The arrow in H points to the apparent apical extension of Sp7 from SOM of the mutant. I,J) RT-qPCR comparison of gene expression in EMM of control and Lmx1b LOF mutant embryos. The expression level of each gene was normalized to the internal standard Ppib. Results from individual samples (5 to 6 per genotype) are presented as diamonds, along with the average (horizontal bar) and the standard deviation (error bars). *: p<0.05, **: p<0.01, ***: p<0.001, n.s.: not significantly different (p>0.05). K) Western blot from EMM of control and Lmx1b LOF mutant embryos. L) Quantitative comparison of the western blot results. 3 lysate samples per genotype were used for the western blot, where each lysate sample was made from EMM of four embryos. M,N) Coronal sections of E13.5 heads processed by immunofluorescence for Ki67 (red). The nuclei were stained with DAPI (blue). O) Ki67 cells and total cells were counted from EMM demarcated in H and I, and the percentage of Ki67 cells was calculated. Results from three embryos per genotype are presented along with the average and the standard deviation. Bar: 0.1 mm.
Through RT-qPCR, the up-regulation of Dlx5 and Runx2 was detectable in the mutant EMM at E11.5–E12.5, around the same time when osteogenesis normally begins in SOM (Fig 5I,J). These results indicated that at least some of the EMM cells of Lmx1b LOF mutants were incorrectly specified into the osteogenic fate. Furthermore, the early onset of the phenotype was in line with the fact that the region-specific expression of Lmx1b was established in the head mesenchyme as early as E10.5 (Fig 3). Together, our results supported the idea that Lmx1b is required to pattern the head mesenchyme preceding the actual calvarial bone formation.
To better understand the changes in the mutant EMM, we examined additional genes that were known to be expressed in the head mesenchyme during normal development. Engrailed1 (En1) encodes a homeobox transcription factor, and it is expressed in SOM from ~E11 (Deckelbaum et al., 2006). We did not detect up-regulation of En1 in Lmx1b LOF mutants (Fig 5J). This suggested that although the mutant EMM had gained osteogenic competence, it did not fully take on the SOM identity.
Msh homeobox genes Msx1 and Msx2 are expressed in both EMM and SOM, and they promote calvarial bone formation from SOM while repressing it in EMM (Han et al., 2007; Roybal et al., 2010; Wilkie et al., 2000). We found that Msx2, but not Msx1, was significantly down-regulated in EMM of Lmx1b LOF mutants (Fig 5J). Although this change is consistent with the possibility that Lmx1b inhibits osteogenesis in EMM at least in part through Msx2, it cannot account for the calvarial phenotype of Lmx1b LOF mutants on its own. Mutation of Msx genes led to heterotopic ossification only when all four copies of Msx1;Msx2 were deleted (Roybal et al., 2010). Twist1, encoding a basic helix-loop-helix transcription factor, is expressed throughout the head mesenchyme just like Msx1 and Msx2 (Rice et al., 2000). Twist1 is required for osteogenic specification of SOM, but inhibits later steps of osteoblast differentiation (Bialek et al., 2004; Goodnough et al., 2012; Goodnough et al., 2016). Its expression was unaffected in Lmx1b LOF mutants (Fig 5J).
We also examined the activities of major osteogenic signaling pathways, namely, BMP-SMAD pathway, canonical WNT pathway, p38 MAPK pathway, and FGF-ERK pathway. We did not find a consistent difference between the control EMM and Lmx1b LOF mutant EMM that could explain the mutant phenotype (Fig 5K,L). Similarly, we did not detect a significant difference in the expression of Axin2 and Gli1 (Fig 5J), which are transcriptional read-outs of the canonical WNT pathway and Hedgehog pathway, respectively (Jho et al., 2002; McMahon et al., 2003). These data supported the notion that was introduced in Fig 2, that the osteogenic signaling activities are not the absolute limiting factor preventing osteogenesis in EMM. For example, even though the level of p-SMAD1/5/9 in wild type EMM was lower than in wild type SOM (Fig 2), this level was sufficient for osteogenesis in Lmx1b LOF mutant EMM (Fig 5L). In addition, based on our results, it is unlikely that Lmx1b regulates osteogenesis by modulating the activation of the above four signaling pathways. Rather, we posit that LMX1B functions independently of these signals to counter their pro-osteogenic effect.
We investigated the effect of the loss of LMX1B on proliferation and apoptosis of EMM cells, using Ki67 and cleaved caspase-3 as markers, respectively (Fernandes-Alnemri et al., 1994; Scholzen and Gerdes, 2000); we did not find a significant difference between controls and Lmx1b LOF mutants (Fig 5M–O, S3 Fig). Combined, our analysis of Lmx1b LOF mutants revealed that Lmx1b was necessary to inhibit osteogenic specification of EMM.
LMX1B is sufficient to inhibit osteogenic specification of SOM
To obtain further evidence that LMX1B has an anti-osteogenic activity in the developing calvaria, we performed a gain-of-function (GOF) experiment of overexpressing Lmx1b in the head mesenchyme. We used a mouse line in which Lmx1b coding sequence was inserted into ROSA26 locus following a floxed stop cassette (R26Lmx1b) (Li et al., 2010). In embryos with both R26Lmx1b and Prrx1-Cre (Prrx1-Cre;R26Lmx1b/+, referred to as Lmx1b GOF), LMX1B was constitutively expressed in the head mesenchyme including SOM, which normally does not express Lmx1b.
At birth, all of the Lmx1b GOF mutants had severely diminished calvarial bone (Fig 6A–D; n=17). The parietal bone was affected more than the frontal bone, which was consistent with the difference in Prrx1-Cre activity between the two parts of the calvaria (S2 Fig). Based on the labeling by a Cre reporter, Lmx1b-overexpressing cells populated the normal location of the parietal bone at P0, but they were present as cartilage and a layer of loose mesenchyme instead of bone (Fig 6E–G; n=2). The residual parietal bone in the mutants was associated with a small number of cells in which Prrx1-Cre was inactive (Fig 6H,I; n=2)
A–D) Lateral views (A,B) and dorsal views (C,D) of the skull of a control and Prrx1-Cre;R26Lmx1b/+ mutant (= Lmx1b GOF) newborns. The skulls were stained with Alizarin red and Alcian blue. E–I) Coronal sections of P0 heads along the dotted line in A. F,G and H,I are pairs of adjacent sections from a Prrx1-Cre;R26Lmx1b/R-LacZ mutant. E, F, and H were stained for bone (brown), and G and I were stained for β-galactosidase activity (blue). In F and G, arrows and arrowheads indicate a layer of loose mesenchyme and cartilage, respectively, in the parietal region of the mutant head. In I, the open arrowhead points to the cells within the mutant calvaria in which Prrx1-Cre was inactive. de: dermis, Fr: frontal bone, IP: interparietal bone, Pa: parietal bone. Bars in A and C: 1 mm, bars in E and H: 0.2 mm.
Examination of osteogenic genes indicated that the initial specification of SOM cells into the osteoblast fate was disrupted in Lmx1b GOF mutants (Fig 7; n=3). At E12.5, Dlx5 and Runx2 were expressed in the parietal bone primordium of control embryos, but they were very weakly expressed at the equivalent position within SOM of the mutants (Fig 7A–D). Because Msx2 was down-regulated in EMM of Lmx1b LOF mutants (Fig 5F), we also examined it in Lmx1b GOF mutants. Compared with the control embryos, Lmx1b GOF mutants had decreased expression of Msx2 in the parietal bone primordium but increased expression in the mesenchyme immediately apical to it, indicating that the regulatory relationship between Lmx1b and Msx2 is complex (Fig 7E,F; n=3). By E14.5, there was no expression of osteogenic genes Dlx5, Runx2, and Sp7 in the presumptive parietal bone region of the mutants (Fig 7G–L; n=2). As shown at P0, Lmx1b-overexpressing cells were properly occupying the parietal region at E13.5, but they failed to make a skeletogenic condensation (Fig 7Q,R; n=3).
A–P) Coronal sections of the heads through the parietal bone rudiment (Pa) processed by RNA in situ hybridization. In E and F, arrows and arrowheads indicate down-regulation and up-regulation, respectively, of Msx2 in the Lmx1b GOF mutant. Q,R) Coronal sections of the head of Prrx1-Cre;R26R-LacZ/+ (control) and Prrx1-Cre;R26Lmx1b/R-LacZ (Lmx1b GOF) embryos processed by β-galactosidase staining (blue). The arrows point to the absence of mesenchyme condensation in the mutant. S,T) Coronal sections of E13.5 heads processed by immunofluorescence for Ki67 (red). The nuclei were stained with DAPI (blue). U) Ki67 cells and total cells were counted from the parietal bone area demarcated in S and T, and the percentage of Ki67 cells was calculated. Results from three embryos per genotype are presented as diamonds, along with the average (horizontal bar) and the standard deviation (error bars). n.s.: not significantly different (p>0.05). Bar: 0.2 mm.
The fact that cartilage was found in P0 Lmx1b GOF mutants at a position comparable to the parietal bone of control pups (Fig 6E–G) prompted us to consider whether the Lmx1b-overexpressing cells had undergone a fate change from osteogenic to chondrogenic. Such change was shown in mouse mutants lacking canonical WNT signaling in the head mesenchyme, which also suffered from loss of calvarial bone (Day et al., 2005; Goodnough et al., 2012; Hill et al., 2005; Tran et al., 2010). Therefore, we examined the early markers of chondrocyte specification and differentiation, SRY-box 9 (Sox9) and Aggrecan (Acan) (de Crombrugghe et al., 2001). The expression of Sox9 and Acan was restricted to the basal domain of the head mesenchyme in both control and Lmx1b GOF mutant embryos, and they were not expressed in the presumptive parietal bone region of the mutants where the osteogenic markers were lost (Fig 7M–P; n=2). This result argued against the idea that the calvarial bone deficiency in Lmx1b GOF mutants was driven by a direct bone-to-cartilage fate conversion of mesenchymal progenitors. Overexpression of Lmx1b was not sufficient to specify the chondrogenic fate, though LMX1B clearly did not inhibit cells from becoming a part of the cartilage (Fig 6G). Rather, we believe that the undifferentiated mesenchyme between the cartilage and the dermis (Fig 6F,G) was what had become of the parietal bone progenitors upon overexpression of Lmx1b, and that the presence of the cartilage was a secondary consequence to the absence of the bone. During normal development, cartilage appears in the parietal region first, but it is degraded later as the parietal bone grows over it. This enzymatic removal of the cartilage is dependent on the signals from the bone (Holmbeck et al., 1999; Zhou et al., 2009).
Similar to Lmx1b LOF mutants, Lmx1b GOF mutants did not show a significant difference in proliferation or apoptosis of SOM cells from control embryos (Fig 7S–U, S3 Fig). Therefore, we conclude that LMX1B specifically inhibited osteogenic specification of SOM cells.
Anti-osteogenic function of Lmx1b is tissue-specific
Our results from both LOF and GOF experiments have established that LMX1B has an anti-osteogenic function in the developing calvaria. This raised a question whether it was true in other parts of the body, specifically in the limbs, where Lmx1b is strongly expressed during normal development and regulates dorso-ventral patterning (Chen et al., 1998a; Cygan et al., 1997; Riddle et al., 1995; Vogel et al., 1995). Because Prrx1-Cre has robust activity in limb buds (Logan et al., 2002), we examined the limbs of Lmx1b LOF and GOF mutants for any evidence of abnormal ossification.
The limb skeleton of Lmx1b LOF mutants showed a patterning defect such as loss of the ulna and the patella, which was reported previously (Fig 8A–D) (Chen et al., 1998b). However, we did not find evidence of heterotopic or excess ossification either from whole mount Alizarin red staining (Fig 8A–D; n=3) or von Kossa staining of the sections (Fig 8E,F; n=4). Similarly, the limb skeleton of Lmx1b GOF mutants was mildly abnormal in their shape (Fig 8G–J) (Li et al., 2010), but the level of ossification appeared grossly comparable to the control limbs based on Alizarin red staining (Fig 8G–J; n=4) and von Kossa staining (Fig 8K,L; n=2). These results indicated that the role of Lmx1b as an essential anti-osteogenic factor is specific to calvarial development.
A–D,G–J) Forelimbs (FL) and hind limbs (HL) stained with Alizarin red and Alcian blue. The arrows point to the morphological abnormalities in the mutant limbs. E,F,K,L) Sections through the humerus (Hum) processed by von Kossa staining. Neither Lmx1b LOF nor Lmx1b GOF mutants displayed an obvious ossification phenotype in the limbs. All four limbs from each animal were sectioned and examined, and the humerus is shown as a representative result. Bar: 1 mm.
Discussion
The findings presented here provide multiple novel insights into the molecular genetic regulation of early stages of calvarial development, which is still poorly understood. We demonstrated that Lmx1b is a crucial regulator of early calvarial patterning. Its activity is required within the head mesenchyme to prevent ossification from the top of the head. This allows the vertex to be occupied by sutures and fontanelles at birth, whose positions are an important determinant of the overall growth and morphogenesis of the mammalian skull.
Patterning of the head mesenchyme
Head mesenchyme cells originate from the mesoderm and the cranial neural crest, and position themselves between the brain and the surface ectoderm at mid-gestation in mice (Deckelbaum et al., 2012; Jiang et al., 2002; Yoshida et al., 2008). The mesenchyme layer on the dorsal and lateral sides of the brain gives rise to the meninges, calvaria, and the dermis, in a deep-to-superficial order. The progenitor cells for the calvaria are further divided into those that will become the transient cartilage, bone, or the sutures/fontanelles. Currently, it is unclear when and how the embryonic head mesenchyme cells are specified into these distinct fates. Based on studies of other body parts such as the limb, patterning of the head mesenchyme most likely involves multiple levels of genetic regulation operating both sequentially and concomitantly. Until now, the only molecular mechanism that has been investigated in detail in this process is for the patterning along the deep-superficial axis of the mesenchyme layer. Here, the canonical WNT signaling and TWIST transcription factors play a crucial role in promoting development of the dermis and the bone in the superficial and intermediate levels by inhibiting chondrogenesis (Day et al., 2005; Goodnough et al., 2012; Goodnough et al., 2016; Goodnough et al., 2014; Hill et al., 2005; Tran et al., 2010).
Our current study defines another important step in the head mesenchyme patterning, namely, specification of EMM and SOM along the basal-apical axis. The fact that EMM and SOM follow distinct developmental paths has been noted for a decade (Roybal et al., 2010; Yoshida et al., 2008). However, it had remained unclear whether there was any intrinsic difference between the two areas of the mesenchyme. Development of the calvarial ossification centers from SOM, but not from EMM, could have been simply due to their different locations, in other words, the cues from the environment. Now, we provide the first set of evidence to challenge this notion. We have shown that major osteogenic signals such as WNT, BMP, and FGF are active in both locations. Also, we demonstrated that there was an intrinsic difference between EMM and SOM prior to the onset of osteogenesis, namely, the SOM-specific absence of Lmx1b expression. Finally, the fact that a head mesenchyme-specific manipulation of gene expression was sufficient for EMM to generate an ossification center indicates that the difference in osteogenic potential of EMM and SOM cannot be solely attributed to the surrounding tissues.
Little is known about the molecular basis of the regionalization of the head mesenchyme into EMM and SOM. To our knowledge, Lmx1b is the first gene to be identified that is expressed in EMM but absent from SOM. Conversely, En1 is the only gene with SOM-specific expression before bone development (Deckelbaum et al., 2006). So far, the studies on early calvarial development have focused on factors operating in SOM, mainly on the signaling pathways and transcription factors that are required for bone formation (Ishii et al., 2015; Rice, 2008; Twigg and Wilkie, 2015). As a result, a non-osteogenic fate of the head mesenchyme has been considered more or less a default outcome for the areas that have been left over after the locations of the bones are determined. However, our data on Lmx1b introduce a new, slightly different perspective (Fig 9). We found that Lmx1b was expressed in broad areas of the head mesenchyme from E9.5, but was already excluded from the prospective SOM by E10.5, which is earlier than the onset of the expression of En1 (~E11) or osteogenic genes (~E12) in SOM (Deckelbaum et al., 2012; Deckelbaum et al., 2006). Therefore, we propose that EMM’s identity as non-osteogenic mesenchyme is specified first via expression of Lmx1b. Subsequently, an Lmx1b-negative zone of mesenchyme is established above the eye through an unknown mechanism, and this zone becomes the osteo-competent SOM because of the absence of Lmx1b (Fig 9).
Lmx1b is broadly expressed in the head mesenchyme from mid-gestation stages, but its expression begins to be excluded from the mesenchyme above the eye around E10.5. This Lmx1b-negative domain becomes the osteogenic SOM, and gives rise to the frontal bone and the parietal bone from ~E12. The Lmx1b-positive mesenchyme on the apical side becomes non-osteogenic EMM, unable to undergo osteogenic specification despite active osteogenic signaling.
Anti-osteogenic function of Lmx1b during calvarial development
Through LOF and GOF experiments, we have established that Lmx1b inhibits osteogenic specification in the head mesenchyme primordium of the calvaria. This is the first evidence that active repression of osteogenesis by a dedicated anti-osteogenic factor is essential to early calvarial patterning. Previously, transcription factor genes Msx1, Msx2, Twist1, Twist2, and Foxc1 were shown to have an anti-osteogenic function during calvarial development (Bialek et al., 2004; Roybal et al., 2010; Sun et al., 2013). However, they also have well-established pro-osteogenic roles in the calvaria depending on the precise timing and location, and all of them are expressed in both EMM and SOM (Goodnough et al., 2012; Goodnough et al., 2016; Han et al., 2007; Huang et al., 2014; Ishii et al., 2003; Li et al., 1995; Liu et al., 1995; Rice et al., 2000; Rice et al., 2003; Roybal et al., 2010; Wilkie et al., 2000). As such, it is difficult to assess the contribution of their anti-osteogenic activity to the regional specification of the head mesenchyme. In contrast, there is no ambivalence in the inhibitory effect of Lmx1b on osteogenesis, and Lmx1b is expressed in EMM but not in SOM. These features are consistent with Lmx1b playing a crucial role in separating the osteogenic and non-osteogenic parts of the calvarial primordium.
At the same time, Lmx1b should not be considered a simple binary switch that determines the fate choice between soft tissue and bone on its own. During normal development, a part of SOM remains unossified and forms a coronal suture even though it does not express Lmx1b. Furthermore, in Lmx1b LOF mutants, the heterotopic ossification center appeared only from a specific antero-posterior position in the dorsal calvaria, even though Lmx1b was deleted from a much larger area (Fig 4). Similarly, the dermal mesenchyme lacking Lmx1b was spared from ossification. These results suggest that the absence of Lmx1b is only a permissive condition for osteogenic specification, not an instructive one. The mesenchyme can be assigned a non-osteogenic fate in the absence of Lmx1b if other requirements for the osteogenic specification are not met. Accordingly, we speculate that the anti-osteogenic function of LMX1B is indispensable specifically at the vertex because of the highly pro-osteogenic condition at this location, due to the abundant production of secreted osteogenic signals from the dorsal brain.
Lmx1b and craniosynostosis
The calvarial phenotype of Lmx1b LOF mutants stands out in its severity from those of other mouse models of craniosynostosis, which typically show fusion of 1–2 sutures or fusion of multiple sutures but only in the postnatal life (Holmes, 2012). Nonetheless, there are potential similarities between Lmx1b LOF mutants and other mutants in the cellular mechanisms leading to craniosynostosis. Previously identified mechanisms for synostosis include 1) enhanced differentiation and/or proliferation of osteoblasts at the osteogenic front of the suture (Yu et al., 2005; Zhou et al., 2000), 2) increased apoptosis in the suture (Chen et al., 2003), 3) a mis-specification of suture mesenchyme cells to become osteoblasts (Holmes and Basilico, 2012; Holmes et al., 2009), 4) loss of boundary integrity between osteogenic and non-osteogenic mesenchyme (Merrill et al., 2006; Ting et al., 2009; Yen et al., 2010), 5) loss of suture stem cell population (Zhao et al., 2015), and 6) ectopic chondrogenesis leading to endochondral ossification of the suture (He and Soriano, 2017; Maruyama et al., 2010). Craniosynostosis in Lmx1b LOF mutants can be best explained by 3). There was a mis-specification of EMM cells into osteoblasts, resulting in heterotopic ossification at the location of the prospective posterior interfrontal suture and the anterior fontanelle.
There are many reported cases of mutations in human LMX1B gene because its haplo-insufficiency leads to nail-patella syndrome (Chen et al., 1998b; Dreyer et al., 1998). This syndrome is characterized by hypoplasia of nails and kneecaps, renal failure, and glaucoma (Sweeney et al., 2003). However, craniosynostosis is not a general feature of these patients. So far, there is a report of only one family that carries a mutation in LMX1B and has craniosynostosis (Wilkie et al., 2010). The mutation is mis-sense, but its effect on LMX1B protein function has not been elucidated. One potential explanation for the absence of craniosynostosis in most patients with LMX1B mutation is that a defect more severe than haplo-insufficiency, such as a dominant-negative mutation, may be necessary to cause craniosynostosis.
In summary, we have demonstrated that Lmx1b plays an essential anti-osteogenic function to regulate early calvarial patterning. Given the unique role of Lmx1b in calvarial development, future investigation into its downstream genetic pathways will yield valuable information that can help understand human calvarial defects.
Patterning of the head mesenchyme
Head mesenchyme cells originate from the mesoderm and the cranial neural crest, and position themselves between the brain and the surface ectoderm at mid-gestation in mice (Deckelbaum et al., 2012; Jiang et al., 2002; Yoshida et al., 2008). The mesenchyme layer on the dorsal and lateral sides of the brain gives rise to the meninges, calvaria, and the dermis, in a deep-to-superficial order. The progenitor cells for the calvaria are further divided into those that will become the transient cartilage, bone, or the sutures/fontanelles. Currently, it is unclear when and how the embryonic head mesenchyme cells are specified into these distinct fates. Based on studies of other body parts such as the limb, patterning of the head mesenchyme most likely involves multiple levels of genetic regulation operating both sequentially and concomitantly. Until now, the only molecular mechanism that has been investigated in detail in this process is for the patterning along the deep-superficial axis of the mesenchyme layer. Here, the canonical WNT signaling and TWIST transcription factors play a crucial role in promoting development of the dermis and the bone in the superficial and intermediate levels by inhibiting chondrogenesis (Day et al., 2005; Goodnough et al., 2012; Goodnough et al., 2016; Goodnough et al., 2014; Hill et al., 2005; Tran et al., 2010).
Our current study defines another important step in the head mesenchyme patterning, namely, specification of EMM and SOM along the basal-apical axis. The fact that EMM and SOM follow distinct developmental paths has been noted for a decade (Roybal et al., 2010; Yoshida et al., 2008). However, it had remained unclear whether there was any intrinsic difference between the two areas of the mesenchyme. Development of the calvarial ossification centers from SOM, but not from EMM, could have been simply due to their different locations, in other words, the cues from the environment. Now, we provide the first set of evidence to challenge this notion. We have shown that major osteogenic signals such as WNT, BMP, and FGF are active in both locations. Also, we demonstrated that there was an intrinsic difference between EMM and SOM prior to the onset of osteogenesis, namely, the SOM-specific absence of Lmx1b expression. Finally, the fact that a head mesenchyme-specific manipulation of gene expression was sufficient for EMM to generate an ossification center indicates that the difference in osteogenic potential of EMM and SOM cannot be solely attributed to the surrounding tissues.
Little is known about the molecular basis of the regionalization of the head mesenchyme into EMM and SOM. To our knowledge, Lmx1b is the first gene to be identified that is expressed in EMM but absent from SOM. Conversely, En1 is the only gene with SOM-specific expression before bone development (Deckelbaum et al., 2006). So far, the studies on early calvarial development have focused on factors operating in SOM, mainly on the signaling pathways and transcription factors that are required for bone formation (Ishii et al., 2015; Rice, 2008; Twigg and Wilkie, 2015). As a result, a non-osteogenic fate of the head mesenchyme has been considered more or less a default outcome for the areas that have been left over after the locations of the bones are determined. However, our data on Lmx1b introduce a new, slightly different perspective (Fig 9). We found that Lmx1b was expressed in broad areas of the head mesenchyme from E9.5, but was already excluded from the prospective SOM by E10.5, which is earlier than the onset of the expression of En1 (~E11) or osteogenic genes (~E12) in SOM (Deckelbaum et al., 2012; Deckelbaum et al., 2006). Therefore, we propose that EMM’s identity as non-osteogenic mesenchyme is specified first via expression of Lmx1b. Subsequently, an Lmx1b-negative zone of mesenchyme is established above the eye through an unknown mechanism, and this zone becomes the osteo-competent SOM because of the absence of Lmx1b (Fig 9).
Lmx1b is broadly expressed in the head mesenchyme from mid-gestation stages, but its expression begins to be excluded from the mesenchyme above the eye around E10.5. This Lmx1b-negative domain becomes the osteogenic SOM, and gives rise to the frontal bone and the parietal bone from ~E12. The Lmx1b-positive mesenchyme on the apical side becomes non-osteogenic EMM, unable to undergo osteogenic specification despite active osteogenic signaling.
Anti-osteogenic function of Lmx1b during calvarial development
Through LOF and GOF experiments, we have established that Lmx1b inhibits osteogenic specification in the head mesenchyme primordium of the calvaria. This is the first evidence that active repression of osteogenesis by a dedicated anti-osteogenic factor is essential to early calvarial patterning. Previously, transcription factor genes Msx1, Msx2, Twist1, Twist2, and Foxc1 were shown to have an anti-osteogenic function during calvarial development (Bialek et al., 2004; Roybal et al., 2010; Sun et al., 2013). However, they also have well-established pro-osteogenic roles in the calvaria depending on the precise timing and location, and all of them are expressed in both EMM and SOM (Goodnough et al., 2012; Goodnough et al., 2016; Han et al., 2007; Huang et al., 2014; Ishii et al., 2003; Li et al., 1995; Liu et al., 1995; Rice et al., 2000; Rice et al., 2003; Roybal et al., 2010; Wilkie et al., 2000). As such, it is difficult to assess the contribution of their anti-osteogenic activity to the regional specification of the head mesenchyme. In contrast, there is no ambivalence in the inhibitory effect of Lmx1b on osteogenesis, and Lmx1b is expressed in EMM but not in SOM. These features are consistent with Lmx1b playing a crucial role in separating the osteogenic and non-osteogenic parts of the calvarial primordium.
At the same time, Lmx1b should not be considered a simple binary switch that determines the fate choice between soft tissue and bone on its own. During normal development, a part of SOM remains unossified and forms a coronal suture even though it does not express Lmx1b. Furthermore, in Lmx1b LOF mutants, the heterotopic ossification center appeared only from a specific antero-posterior position in the dorsal calvaria, even though Lmx1b was deleted from a much larger area (Fig 4). Similarly, the dermal mesenchyme lacking Lmx1b was spared from ossification. These results suggest that the absence of Lmx1b is only a permissive condition for osteogenic specification, not an instructive one. The mesenchyme can be assigned a non-osteogenic fate in the absence of Lmx1b if other requirements for the osteogenic specification are not met. Accordingly, we speculate that the anti-osteogenic function of LMX1B is indispensable specifically at the vertex because of the highly pro-osteogenic condition at this location, due to the abundant production of secreted osteogenic signals from the dorsal brain.
Lmx1b and craniosynostosis
The calvarial phenotype of Lmx1b LOF mutants stands out in its severity from those of other mouse models of craniosynostosis, which typically show fusion of 1–2 sutures or fusion of multiple sutures but only in the postnatal life (Holmes, 2012). Nonetheless, there are potential similarities between Lmx1b LOF mutants and other mutants in the cellular mechanisms leading to craniosynostosis. Previously identified mechanisms for synostosis include 1) enhanced differentiation and/or proliferation of osteoblasts at the osteogenic front of the suture (Yu et al., 2005; Zhou et al., 2000), 2) increased apoptosis in the suture (Chen et al., 2003), 3) a mis-specification of suture mesenchyme cells to become osteoblasts (Holmes and Basilico, 2012; Holmes et al., 2009), 4) loss of boundary integrity between osteogenic and non-osteogenic mesenchyme (Merrill et al., 2006; Ting et al., 2009; Yen et al., 2010), 5) loss of suture stem cell population (Zhao et al., 2015), and 6) ectopic chondrogenesis leading to endochondral ossification of the suture (He and Soriano, 2017; Maruyama et al., 2010). Craniosynostosis in Lmx1b LOF mutants can be best explained by 3). There was a mis-specification of EMM cells into osteoblasts, resulting in heterotopic ossification at the location of the prospective posterior interfrontal suture and the anterior fontanelle.
There are many reported cases of mutations in human LMX1B gene because its haplo-insufficiency leads to nail-patella syndrome (Chen et al., 1998b; Dreyer et al., 1998). This syndrome is characterized by hypoplasia of nails and kneecaps, renal failure, and glaucoma (Sweeney et al., 2003). However, craniosynostosis is not a general feature of these patients. So far, there is a report of only one family that carries a mutation in LMX1B and has craniosynostosis (Wilkie et al., 2010). The mutation is mis-sense, but its effect on LMX1B protein function has not been elucidated. One potential explanation for the absence of craniosynostosis in most patients with LMX1B mutation is that a defect more severe than haplo-insufficiency, such as a dominant-negative mutation, may be necessary to cause craniosynostosis.
In summary, we have demonstrated that Lmx1b plays an essential anti-osteogenic function to regulate early calvarial patterning. Given the unique role of Lmx1b in calvarial development, future investigation into its downstream genetic pathways will yield valuable information that can help understand human calvarial defects.
Supplementary Material
1
2
3
4
1
2
3
4
Acknowledgments
We thank Dr. Jean-Pierre Saint-Jeannet and his laboratory members for helpful discussions and sharing equipment. We acknowledge Dr. Shoshana Yakar and Ripa Chowdhury of NYU College of Dentistry Micro-CT core for their help with the analysis. This work was supported by National Institutes of Health (R03 DE023617 and R01 DE026798 to J.J.)
Abstract
The calvaria (upper part of the skull) is made of plates of bone and fibrous joints (sutures and fontanelles), and the proper balance and organization of these components are crucial to normal development of the calvaria. In a mouse embryo, the calvaria develops from a layer of head mesenchyme that surrounds the brain from shortly after mid-gestation. The mesenchyme just above the eye (supra-orbital mesenchyme, SOM) generates ossification centers for the bones, which then grow toward the apex gradually. In contrast, the mesenchyme apical to SOM (early migrating mesenchyme, EMM), including the area at the vertex, does not generate an ossification center. As a result, the dorsal midline of the head is occupied by sutures and fontanelles at birth. To date, the molecular basis for this regional difference in developmental programs is unknown.
The current study provides vital insights into the genetic regulation of calvarial patterning. First, we showed that osteogenic signals were active in both EMM and SOM during normal development, which suggested the presence of an anti-osteogenic factor in EMM to counter the effect of these signals. Subsequently, we identified Lmx1b as an anti-osteogenic gene that was expressed in EMM but not in SOM. Furthermore, head mesenchyme-specific deletion of Lmx1b resulted in heterotopic ossification from EMM at the vertex, and craniosynostosis affecting multiple sutures. Conversely, forced expression of Lmx1b in SOM was sufficient to inhibit osteogenic specification. Therefore, we conclude that Lmx1b plays a key role as an anti-osteogenic factor in patterning the head mesenchyme into areas with different osteogenic competence. In turn, this patterning event is crucial to generating the proper organization of the bones and soft tissue joints of the calvaria.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.