Smoc1 and Smoc2 regulate bone formation as downstream molecules of Runx2

Runx2 is an essential transcription factor for bone formation. Although osteocalcin, osteopontin, and bone sialoprotein are well-known Runx2-regulated bone-specific genes, the skeletal phenotypes of knockout (KO) mice for these genes are marginal compared with those of Runx2 KO mice. These inconsistencies suggest that unknown Runx2-regulated genes play important roles in bone formation. To address this, we attempted to identify the Runx2 targets by performing RNA-sequencing and found Smoc1 and Smoc2 upregulation by Runx2. Smoc1 or Smoc2 knockdown inhibited osteoblastogenesis. Smoc1 KO mice displayed no fibula formation, while Smoc2 KO mice had mild craniofacial phenotypes. Surprisingly, Smoc1 and Smoc2 double KO (DKO) mice manifested no skull, shortened tibiae, and no fibulae. Endochondral bone formation was also impaired at the late stage in the DKO mice. Collectively, these results suggest that Smoc1 and Smoc2 function as novel targets for Runx2, and play important roles in intramembranous and endochondral bone formation.

The work attempts to provide improved information on the functional role of SMOC1/2 proteins in skeletal development, using a combination of quantitative gene expression studies in limb-bud cells, in vitro assays to show effects on osteoblast differentiation, and some targeted mouse genetics to link these with bone development in vivo. It is a novel study, in as far as the authors reveal a genetic pathway that includes Bmp2 and Runx2 inducing expression of the paralogous Smoc genes: Smoc1 and Smoc2. They also give a phenotypic description of Smoc1;Smoc2 double knockout mice, which is also novel.
Specific comments: 1. No statistics are given for the microarray analysis (e.g. number of replicates, p-values, FDR etc.) It was not explained how this data was analysed. The data in Table 1 only shows eight genes and these are all very similar in their gene family/type, which is a somewhat surprising result. Are these the top genes or the only upregulated genes? It is essential to include a comprehensive list of the values for all genes included on the array somewhere in the manuscript. The GEO data was not accessible to this reviewer as the data is currently private and is scheduled to be released on Feb 20, 2020. No token was provided. 2. In the microarray analysis a FC of +1.46 and +1.26 for Smoc1 and Smoc2, respectively, seems very low compared to the data presented in Figure  8. Is there any evidence that Runx2 binds to the Smoc genomic loci, either at the promoter regions or cis-regulatory regions, and that it directly regulates their expression? In silico analyses could be used to identify any Runx2 motifs in these regions, and Chromatin immunoprecipitation analyses could confirm this experimentally. 9. What is the direct link of Smoc gene function to bone formation? Do they effect mesenchymal stem cell differentiation into osteoblasts or chondrocytes? It would be useful to address this robustly in either embryos, or more thoroughly in the cell culture systems used. 10. Can the authors add DN-Runx2 and BMP2 in combination to the Limb Bud cells to confirm the direct relationship between these and Smoc gene expression. Do they still see increases in Smoc expression in this context? 11. The In Situ hybridisation images in figure 2 are of low resolution and difficult to interpret. In addition, they are not novel as several studies have previously shown Smoc1 developmental expression at E11.5 (Rainger et al. 2011, Abouzid et al, 2011, and also Smoc2. To make better use of this data, the specific expression of both paralogues should be clearly annotated in specific tissue regions, and compared to show where they overlap or are individually specific. These in situs may also be somewhat over-exposed, as the limb bud expression does not clearly show the anterior-posterior specific expression of Smoc1 and its absence in the regions adjacent to the AER. Could the limbs be removed from the embryos to more clearly show the specific expression patterns and reduce some shadowing from the rest of the embryo body? 12. I was unsure of the relevance of Col2A1 in situ hybridisation in Figure 2. How does this relate to Smoc/Runx2 or Bmp2? A statement is required to provide this link more clearly. 13. There are no scale bars on the in situ images. The E11.5 embryos look older and I think these may not be accurately staged. 14. Quantitative analyses of the Alkaline phosphatase assays are required (Fig 3e-f). These images are currently insufficient to draw any meaningful conclusions from alone and the authors interpretation is therefore subjective. 15. There is inadequate information and phenotypic analysis of the mouse crosses. No metrics are given for the incidence of viable embryos for each genotype and penetrance rates of each phenotype are not given.  Figure 5 is not convicing. There is no quantitation of signal, and the specimens shown are of poor quality and/or of low resolution. What does it mean if there is less signal in a tissue that is smaller than its control? These are largely phenomena and do not give any insight into mechanisms of bone formation mediated by  Figure 1f is not labelled. Figure 1g requires an axis legend. 25. Sample sizes are not provided on an experiment-by experiment basis, which is important to interpret the data shown.

Reply to Reviewer 1
We greatly thank the reviewer for reviewing our manuscript and providing us with constructive and important comments on our study. Reply: We thank the reviewer for these important comments. To address these points, we mated 3. Fig.1a Fig 12a). We also found an association of Hoxc11 with Runx2 by performing co-immunoprecipitation experiments expressions, although the identification of these is not our aim in the present study. We greatly appreciate the important suggestion that improved our study.
Minor comments: There are many typographical or editing errors.
We are very sorry for our errors and thank the reviewer for the careful reading of the original manuscript.
1. There is no title or legend explaining Table 1.
Reply: We replaced Table 1 of the microarray screening data with RNA-sequence data that were newly obtained to confirm our original data ( Supplementary Fig. 1, Supplementary Tables 1 and 2), and added the legend in the revised manuscript.

Information or reference explaining DN-Runx2 would be required.
Reply: The dominant-negative Runx2 (DN-Runx2) used in this study consisted of the amino acids 2-247 of Runx2 tagged with Flag epitope. This construct lacks the transcription activation domain at the C-terminal region. We added this information in the revised manuscript. Fig. 1g. is missed.

Label of X-axis in
Reply: We added a label to the X-axis in the revised manuscript.

Fig. 3C. Please change Smoc1 and Smoc2 in the X-axis to shSmoc1 and shSmoc2, respectively.
Reply: We corrected the labels of the X-axis to shSmoc1 and shSmoc2 in Fig. 3c. Reply: We are sorry for giving this misleading description. We made the correction as pointed out by the reviewer to avoid any misunderstanding by the readers.

Reply to Reviewer 2
This is an important report revealing the new molecular network regulating bone formation via We greatly thank the reviewer for critically reviewing our manuscript and providing us with constructive and important comments. We are also very pleased that the reviewer evaluated our findings as novel and providing improved information on the functional role of Smoc1/2 on skeletal development. Reply: We thank the reviewer for providing us with these important comments. We deeply apologize that the reviewer was unable to access the GEO data. In this study, we performed microarray analysis solely for the purpose of initial screening for genes that are upregulated by Bmp2, which regulates Runx2 expression and function. Therefore, the initial microarray was performed as a single screening experiment in the original manuscript. To confirm the original data and analyze it, we additionally performed RNA-sequence analyses with duplicate samples. Therefore, we replaced the original microarray data with the new RNA-sequence analyses data ( Supplementary Fig. 1

, Supplementary
Tables 1 and 2). In response to the reviewer's suggestion that a comprehensive list should be provided, we added this list in the revised manuscript (Supplementary Tables 1 and 2). We deposited the data in the NCBI database and will forward the token to the editorial office for the reviewer to access the original data.  (Fig. 1a,   b). To confirm the data, we extensively repeated RT-qPCR analyses as shown in the new Figure 1 for accurate quantitative evaluation. Additionally, we assume that the differences of the data between microarray analysis and RT-qPCR in the original manuscript were due to methodological differences between the two assays. Overall, we could confirm our results, and appreciate the reviewer's excellent comments.  Fig. 9g) and the data of the quantified length ( Supplementary Fig. 9h, i) in the revised manuscript. We greatly appreciate the reviewer's suggestions that have strengthened and improved our study. 15 below, we added whole skeletal appearance of all genotypes in Supplementary Fig. 11 in the revised manuscript. We also incorporated the data of the occurrence of phenotypes in all genotypes (Supplementary Table 3). We thank the reviewer so much for encouraging us to determine the phenotype of the double mutant mice more precisely and adequately. Reply: According to the reviewer's important and constructive comments, we performed whole-mount in situ hybridization using Runx2 KO embryos as shown in Figure 2. Smoc1 expression was observed in the skull, clavicle, and long bone at the E12.5 stage, whereas the expression in the limb bud was very weak at least at this stage (Fig. 2, Supplementary Fig. 5). We found that calvarial Smoc1 expression in Runx2 KO mice was moderately decreased compared with WT mice (Fig. 2a). Strong

Remarkably
Smoc2 expression was observed in the skull of WT mice (Fig. 2d), whereas very weak signals were detected in other tissues (Fig. 2e, f). Calvarial Smoc2 expression in Runx2 KO mice was clearly decreased compared with WT mice (Fig. 2d). These results support the data that

Can the authors add DN-Runx2 and BMP2 in combination to the Limb Bud cells to confirm the direct relationship between these and Smoc gene expression. Do they still see increases in Smoc expression in this context?
Reply: According to the reviewer's suggestion, we extensively investigated the effect of DN-Runx2 and this Bmp2-induced upregulation was inhibited by DN-Runx2. Considering the results of the induction by Runx2 (Fig. 1a, b, Supplementary Table 2) and the ChIP assay ( Supplementary Fig. 4 we showed the limbs, calvariae, and vertebrae from the embryo bodies. We replaced the original data with the new data in Fig. 2 in the revised manuscript.

I was unsure of the relevance of Col2A1 in situ hybridisation in Figure 2. How does this relate to
Smoc/Runx2 or Bmp2? A statement is required to provide this link more clearly.
Reply: We included the results of Col2a1 expression as a positive control for chondrogenesis and a positive control of in situ hybridization experiments. We totally agree with the reviewer's opinion. To avoid any misunderstanding by the readers, we removed the data.
13. There are no scale bars on the in situ images. The E11.5 embryos look older and I think these may not be accurately staged.
Reply: Because the magnification of all the panels were the same in the original figure (Fig. 2), we did not add scale bars on all panels. We are very sorry for confusing the presentation. As described above (#12), we replaced the data and incorporated the scale bars in all the panels.

13
14. Quantitative analyses of the Alkaline phosphatase assays are required (Fig 3e-f). These images are currently insufficient to draw any meaningful conclusions from alone and the authors interpretation is therefore subjective.
Reply: According to the reviewer's comment, we additionally performed the experiments in which we quantitatively determined the alkaline phosphatase activity using p-nitrophosphate as a substrate.
Alkaline phosphatase activity in osteoblasts infected with shSmoc1 or/and shSmoc2 was significantly decreased compared to that of the shGFP control. We incorporated Fig. 3f in the revised manuscript.
We thank the reviewer for this important comment.

Labelling of anatomy is lacking in Fig4.
Reply: We apologize for our careless errors, and labelled the anatomy in the revised figure. We thank the reviewer for such careful reading of the manuscript.  (Fig. 2).

The phenotypes revealed in
In addition, abnormal skull formation was observed at the earlier stages of E12.5 and E13.5 ( Supplementary Fig. 10). While we agree with the reviewer's comment, we do believe that our in vitro and in vivo data collectively indicate the association of Runx2 and Bmp2 with the role of Smoc1 and Smoc2 shown in Fig. 4, and that the differences in the patterning effect would be partly involved in the phenotype. Figure 5 is not convicing. There is no quantitation of signal, and the specimens shown  (Fig. 5d). The quantitative analyses strongly support our histological analyses. We added the new RT-qPCR data in Fig. 5 in the revised manuscript.

The data in
Although the Col10a1 expression level was similar in WT and Smoc1 and Smoc2 DKO mice (Fig. 4d), the Col10a1 expression pattern was quite different between them (Fig. 4c). It is also evident that endochondral bone formation was delayed at the late stage in the DKO mice. We incorporated these data in the revised manuscript ( Supplementary Fig. 5). We greatly thank the reviewer for the excellent comments.

Can the authors explain why the levels of Smoc1 mRNA is lover in Runx2 treated cells compared to Bmp2-treated cells? This seems surprising if the central tenet of this study is that Runx2 mediates
Smoc gene expression.
Reply: To respond to the reviewer's questions, we extensively repeated the RT-qPCR experiments, including more groups, and performed RNA-sequence analyses ( Fig. 1 and Supplementary Tables 1 and 2). We confirmed that Runx2 reproducibly induced Smoc1 and Smoc2 expressions; however, the effects of Runx2 on these expressions were different from those of Bmp2. One possible explanation is that we cannot accurately compare the effects of the transcription factor (Runx2) with those of the ligand (Bmp2). Another reason would be Runx2-independent pathways, as described above. Together with the results of the ChIP assays ( Supplementary Fig. 4) and whole-mount in situ hybridization experiments using Runx2 KO mice (Fig. 2), we are confident of our major proposal that Smoc1 and Smoc2 are direct targets of Runx2. We thank the reviewer for these comments that strengthened and improved our study.

Are Osterix and Osteocalcin also upregulated in this system? Can the authors therefore directly link Smoc function to bone formation?
Reply: We determined the expressions of Osterix and osteocalcin, both of which are well-known downstream molecules of Runx2, as positive controls to confirm the effects of Bmp2 treatment on the lumb bud cells. As expected, these genes were significantly upregulated in the Bmp2-treated group.
Osterix is an important transcription factor for bone formation, and Osterix KO mice lack bone formation. By contrast, osteocalcin is a well-known marker for osteoblasts. Because osteocalcin KO mice did not display impaired bone formation in a previous study (Ducy et al, Nature, 1996), osteocalcin does not appear essential for bone formation. Recently, Komori et al. (Int J Mol Sci, 2020) and We appreciate the reviewer's comment.
24. Figure 1f is not labelled. Figure 1g requires an axis legend.
Reply: We apologize for our careless errors. We replaced the original figure with new data and added the label to the axis in the revised manuscript.

Sample sizes are not provided on an experiment-by experiment basis, which is important to
interpret the data shown.
Reply: We are very sorry for not providing the sample numbers. We now describe the sample size for each experiment in the figure legends.
The authors have performed a tremendous amount of additional work, and the study is greatly improved. I have some remaining comments I feel would be important to address: 1) I welcome the RNAseq data, however the analysis is insufficiently described and not enough data has been presented. Please provide PCA analysis and more information on read quality counts. What were the criteria for inclusion/rejection of reads. Also, the supplemental legend to figure 1 refers to a LFC of >2, whereas the figure and main text use LFC>1. There are no metrics for adjusted p values, or what criteria were used to define down-regulated gene expression. This part of the study required significant supporting information to be included.
2) The discussion ends rather negatively ( To Reviewer 1's points: The reviewer's comments has been sufficiently addressed. They now show the data requested for all of the genotypes and at the developmental stage requested. The data is well presented in Fig.S11.

Reply to Reviewer 3
We greatly thank the reviewer for reading our revised manuscript and providing valuable comments. We are pleased that the reviewer considered our revised manuscript to be greatly improved. Furthermore, we greatly appreciate the reviewer's effort to evaluate our changes in response to the comments of Reviewer 1. Reply: We thank the reviewer for these important comments. Accordingly, we carefully checked the quality of our RNA-Seq data. We conducted PCA analysis of the data and confirmed the similarities within each group. We included the PCA analysis data in Supplementary Fig. 1a; our findings show that variance values clearly differ among Venus, Bmp2, and Runx2 groups. We also included the quality check data for each group following Fastqc analysis, as shown below. If the reviewer considers it necessary to incorporate these data in the manuscript, we are willing to include them as supplementary data.
Regarding the description of LFC, we apologize for confusion. In the revised manuscript, we

Reply to Reviewer 1's point
The reviewer's comments has been sufficiently addressed. They now show the data requested for all of the genotypes and at the developmental stage requested. The data is well presented in  They should also include a table indicating the penetrance of the phenotypes for each genotype, and at least show how many animals were observed with the phenotypes described. I could only find this for the calvaria phenotype but not the craniofacial defect.
Reply: Because the bone calcification of several mutant mice was impaired at P0, the micro CT photographs of the mice were not particularly clear; we included them in this letter to allow assessment by the reviewer. We quantified both nasal bone to eye distance and length of mandibular bone in the mice as shown below; however, the differences were not statistically significant. If the reviewer feels that it is appropriate to incorporate these data in the manuscript,