Reactivation of a developmental Bmp2 signaling center is required for therapeutic control of the murine periosteal niche

Reviewer #2:

[…] This is an important and comprehensive study from a group who has contributed much and in an original manner to this area of biomedical research. For the most part, the data are strong and convincing and do strengthen the argument that BMP2 plays a rather unique and non-redundant role in establishing the phenotypic properties and repair capacity of periosteum. As the authors stress, future identification of means to stimulate periosteal BMP2 action or expression could lead to new therapies to strengthen or repair bone. There are only a few and relatively minor concerns.

A key conclusion reached is that "although many BMPs are expressed in bone, periosteal BMP signaling and bone formation require only Bmp2 in the Prx1-Cre lineage". It would be important to know whether Bmp2 ablation may have resulted in changes in expression of other Bmps, particularly in periosteum.

New data in Figures 10A-B show that Bmp7 is also expressed in the periosteum and therefore does not compensate for loss of Bmp2. This is fully consistent with the in vitro data showing that recombinant BMP7 does not compensate for loss of Bmp2 in primary osteoblast cultures. Expression of Bmp4 in periosteal cells was at the lower limit of detection by QPCR and therefore also cannot compensate for lack of Bmp2.

Also, given that Prx1-Cre does not target only the periosteal progenitors but the entire mesenchymal cell population within the limbs, it would be important for the authors to discuss possible changes in tissue interactions within the mutant limbs influencing outcomes.

Based on our Bmp2-lacZ data, we do not see evidence of Bmp2 expression that would influence other tissue interactions within the mutant limbs and that it is likely that if BMP signaling is required for those interactions, other BMPs present in the limb would provide the signal. While it is somewhat surprising that BMP2 would have such a highly specific role in limb skeletogenesis, the role is fully consistent with the Bmp2-lacZ localization pattern. As genetic ablation of Bmp4 or Bmp7 do not affect periosteal apposition but have other ligand-specific effects on the limb skeleton (2008 Tsuji, 2010 Tsuji, and 2006 Bandyopadhyay), we believe ligand localization is a major determinant of this phenotype.

As the authors correctly underline, it is quite interesting that Dkk1 haploinsufficiency and treatment with iPTH or SOST antibody did not appear to rescue periosteal growth and fracture repair defects in the Bmp2;Prx1-Cre mice. In fact, the latter treatments increased the frequency of spontaneous fractures in mutant mice. The authors mention, but do not deal with, this striking outcome, and it would be important to not only face it, but also try to account for it.

We observed increased frequency of spontaneous fractures only with iPTH treatment, as shown in Figures 2-T. It is well established that iPTH activates bone resorption alongside bone formation, where the net effect in wildtype mice is increased bone mass. In the absence of BMP2, activation of bone formation response is severely blunted and thus the altered balance shifts towards increased bone fragility.

The ChIP-sequencing data in Figure 6 are somewhat perplexing as they were generated using whole E10.5 mouse limb buds and immortalized E13.5 limb bud cells in vitro. Both of these represent early embryonic and highly heterogenous cell populations, and it is hard to know to what extent they speak (or can speak) about the function of selected genes (Axin2, Lef1 and Grhl3) specifically in periosteal cells. The fact that Bmp2 ablation does not affect embryonic long bone development and that bone formation has not started by E13.5 also question the significance of the data.

Question 1) Heterogeneity

- For E10.5, the ChIP-seq analysis automatically enriches spefically for cells that are actively engaged in canonical Wnt signaling.

- For E13.5, the immortalized limb bud cells we used are a clonal line of osteochondral-progenitors, not a heterogeneous population (Rosen, 1994).

Question 2) Time points were selected for analysis

- E13.5 is when Osx1+ osteoblasts and the bone collar (primitive periosteum) first appear (Rodda and McMahon, 2006). These cells persist for the first 30 days of life and directly mediate periosteal bone growth during early post-natal life (Maes, 2010).

Question 3) Function of selected genes

- Axin2 and Lef1 are only used as positive controls to demonstrate that the Bcl9-pulldown method enriches for previously established canonical Wnt target genes.

- We propose that Grhl3 acts as a co-factor with Smads for induction of Sp7. For this to be true, Grhl3 would need to be induced prior to induction of Sp7, and this hypothesis is confirmed in this figure.

Reviewer #3:

This manuscript from Salazar and colleagues presents information of potential interest to many in the field. While the concepts being addressed are of interest, there are numerous points that should be clarified that are listed below.

1) Please show the bone phenotypes (Figure 1) for Prx1-Bmp2 heterozygotes if they are available. Please show a schematic diagram or a reference on Bmp2-Lacz allele (Figure 2A).

The phenotype of Bmp2; Prx1-Cre heterozygotes has been previously shown in Tsuji et al., 2006.

A schematic showing the structure of the Bmp2-lacZ allele has been provided in new Figure 2—figure supplement 1A-C.

2) Although LacZ staining and GFP fluorescence (Figure 2) show Bmp2 expression and Bmp2 activation nicely, sometimes the signal could be too weak to detect due to the limitation of the techniques. Such as there is a significant Bmp2 activation (GFP, Figure 1U) in the adult bone marrow, but there is no Bmp2 expression (LacZ, Figure 1H and I) at the same place. So, a qPCR with different bone fractions by serial collagenase digestion of cortical bone would be a good confirmation on the location of Bmp2 in different stages.

We thank the reviewer for the comments and apologize for not being clear about how the signaling reporter works. The BRE-gfp reporter is not specific for BMP2 signaling but rather provides a net readout of all BMP signaling. The observation that GFP expression is maintained in the marrow compartment of Bmp2 cKOs is consistent with the fact that many BMPs besides Bmp2 have been shown previously by several groups to be expressed in the bone marrow where they contribute to hematopoiesis as well as skeletal biology. By the same rationale, the observation that GFP persists in the marrow of Bmp2 cKOs while being attenuated in the periosteum, strongly supports our model that other BMPs do not compensate for BMP2 in the periosteum. We propose that BMP2 is the main BMP family member with activity in the periosteum, whereas other BMPs, such as Bmp4 and Bmp7 compensate in other compartments (2008 Tsuji, 2010 Tsuji, and 2006 Bandyopadhyay).

3) No Col1a1 expression in WT/cKO osteocytes within the cortical bone (Figure 3F)?

In a lineage-tracing experiment with Col1a1-Cre, osteocytes would be indeed be “marked” as descendants of a population that at some point expressed bona fide Col1a1 mRNA. However, the primary function of bone-embedded osteocytes is no longer matrix-production but mechanical sensing. As part of this change in cell function, osteocytes would not necessarily be expected to be actively transcribing and depositing new Col1a1 in the surrounding bone as unabated production of collagen would eventually eliminate the lacunar space they occupy. The observation that osteocytes are not transcribing Col1a1 mRNA, regardless of Bmp2 expression, is therefore the expected outcome.

4) Bmp2-Prx1 cKO appears to not affect iPTH-induced trabecular bone gain (Figure G and N). The author claims that Bmp2 is a mediator of iPTH for periosteal bone formation. However, there was no significant change in periosteal expansion in either WT or cKO mice (Figure 4H). It is also not clear why there was a significant increase in periosteal bone formation in the cKO mice as well (Figure 4J and L).

Periosteal BFR and MAR are direct readouts of periosteal function while uCT total area is indirect. Increased total area requires a longer amount of time to be measurable compared to increased BFR/MAR since it is a cumulative effect over time of bone forming activity.

5) Figure 4M is a little confusing. According to Figure 2H, Bmp2 is expressed in adult bone osteocytes (not periosteal cells). However, iPTH only induces ID3 in only periosteal cells but not osteocytes themselves. Is this suggesting that osteocytes do not respond to Bmp2 although they express Bmp2? The reviewer is also confused by the cKO bone morphology in Figure 4M, which shows massive woven bones and relative stronger (compared with the WT) ID3 expression in osteocytes.

Figure 2H shows Bmp2-lacZ expression in osteocytes of untreated juvenile mice of 2 weeks of age. These data cannot be directly compared to histology in 4M which shows the cortical bone of a 1 month old animal that was treated for the past 2 weeks with iPTH.

The authors thank the reviewer for pointing out the highly porous nature of the PTH-treated cKO cortical bone in Figure 4M. We have used the lamellar structures as guides to distinguish true cortical bone with osteocytes from the intra-cortical pockets of heterogeneous endosteum and marrow. This greatly clarifies that Id3+ nuclei can be found in the periosteum and marrow of PTH-treated WT mice, but not the periosteum of PTH-treated cKO mice. Importantly, osteocytes in cortical bone of PTH-treated mice are Id3-, regardless of Bmp2 genotype. The data across the paper are therefore consistent in that osteocytes are negative for Id1, Id3, and Col1a1, as would be expected for cells that should not be actively engaged in BMP signaling and are not making bone matrix.

6) Figure 4Q-T do not have error bars or statistical analysis done.

Data in Figure 4Q represent an absolute number (% of total animals scored with a binary outcome). Error bars and statistical analysis are therefore not applicable here.

7) What does no Imin change in Bmp2 cKO mice mean in Figure 5I?

Imin is the minimum value of moment of inertia for the bone, which occurs relative to the cross-sectional orientation with the least resistance to bending. Wildtype mice treated with sclerostin antibody exhibit increased Imin while cKO mice exhibit no change. This estimated material property is based on diaphyseal (cortical) morphology obtained by microCT, and predicts that wildtype bones, but not cKO bones are more resistant to bending (and fracture) after treatment.

8) It is convincing that Bmp2 cKO inhibits anti-SOST-induced bone expansion (Figure 5G-J). Since there is no bone phenotype when Bmp2 is deleted in osteoblasts (Sp7-cre or Col1-cre) and BV/TV is normal in Prx1-Bmp2 cKO mice (Fig4G and N), it appears that Bmp2 is important for periosteum, but not for trabecular bone formation. What is the hypothesis that BV/TV cKO is not increased by anti-SOST (Figure 5F) other than development versus treatment?

In Figure 10C, in vitro experiments reveal that BMSC cells exhibit a “bell-shaped curve” response to ectopically activated WNT signaling, where low concentrations of recombinant Wnt3a activate matrix calcification but higher amounts become inhibitory. This inhibitory WNT threshold is lowered in the absence of Bmp2, as BMSC lacking Bmp2 are able to mineralize under normal conditions but not when exposed to recombinant Wnt3a. In other words, the inhibitory WNT threshold is reduced when the net amount of BMP signaling is reduced. in vivo, we hypothesize that endosteal osteoblasts are able to function in an unchallenged environment due to compensation by other BMPs (probably BMP4/7), but that the reduced net strength of pan-BMP signaling has made them more sensitive to Wnt-inhibition. In a future study, this interesting possibility could be measured by monitoring bone mass in Bmp2; Prx1-Cre mice treated with decreasing doses of SOST-ab.

9) Please provide an additional and higher powered picture for Figure 5O. It seems that all Prx1+ cells are BRE/GFP+ while all Prx1- cells are BRE/GFP-.

A new supplementary figure with separate images for GFP, DAPI, and RFP channels very clearly shows that the vast majority of cells derived from the Prx1-Cre; TdTomato lineage are not actively engaged in BMP signaling (GFP+).

10) Also, please provide an additional picture without DAPI/blue fluorescence for Figure 5N to show GFP in the periosteum.

A new figure panel has been added to Figure 5P.

11) Compare the result of Bcl9 ChIP-seq in limbs with publicly available Tcf4/β-catenin Chipseq data in other tissues/cell lines in terms of pathway enrichments. Is there a reason to choose Bcl9 antibody over Tcf4 antibody?

TCF/LEF proteins are constitutively bound to chromatin. In a WNT-OFF state, they function as transcriptional repressors of their bound target genes. In a WNT-ON state, additional co-factors (such as β-catenin, pygopus, and Bcl9) are recruited to genome-bound TCF, thereby converting TCFs from transcriptional repressors to transcriptional activators. TCF pulldown is therefore not able to discriminate between genes that are being repressed or induced, and thus a co-activator must be used to understand the subset of target genes that are being activated at any given time.

We used Bcl9 pulldown to selectively enrich for loci occupied by transcriptionally active β-catenin (by definition, only in WNT-ON cells). In support of our scientific rationale, Bcl9 mutant mice exhibit defects in distal limb and long bone development, demonstrating at the genetic level that β-catenin transcription in the developing limb is fully dependent on the Bcl9-β-catenin interaction (2018 Cantu Gen Dev).

12) Bmp2 is proposed to be a direct downstream target of canonical Wnt signaling (Figure 6E) in limb buds. Does Wnt3a treatment directly induce Bmp2 mRNA in limb bud cells?

Yes, a new Figure 11—figure supplement 1A shows that Bmp2 mRNA is upregulated by recombinant Wnt3a.

13) Are the authors proposing that Bmp2-Sp7-Col1a1 cascade is only present in the periosteum, but not in cortical bone or trabecular bone?

We propose that a general BMP-dependent cascade acts upstream of the Sp7/Col1a1. In the periosteum, our data provide evidence that this BMP-dependent cascade is activated primarily if not entirely by BMP2. Our study does not address which BMPs compensate for loss of BMP2 at non-periosteal sites. However, it reasonable to hypothesize that the major compensatory BMP at non-periosteal sites is BMP4, since mice with compound loss of Bmp2 and Bmp4 in Prx1-Cre lineage fail to make an appendicular skeleton (2006 Bandyopadhyay) while mice with loss of only Bmp2 manifest a periosteal-specific phenotype.