The periosteal requirement and temporal dynamics of BMP2‐induced middle phalanx regeneration in the adult mouse

Abstract Regeneration of mammalian limbs is restricted to amputation of the distal digit tip, the terminal phalanx (P3). The adjacent skeletal element, the middle phalanx (P2), has emerged as a model system to investigate regenerative failure and as a site to test approaches aimed at enhancing regeneration. We report that exogenous application of bone morphogenetic protein 2 (BMP2) stimulates the formation of a transient cartilaginous callus distal to the amputation plane that mediates the regeneration of the amputated P2 bone. BMP2 initiates a significant regeneration response during the periosteal‐derived cartilaginous healing phase of P2 bone repair, yet fails to induce regeneration in the absence of periosteal tissue, or after boney callus formation. We provide evidence that a temporal component exists in the induced regeneration of P2 that we define as the “regeneration window.” In this window, cells are transiently responsive to BMP2 after the amputation injury. Simple re‐injury of the healed P2 stump acts to reinitiate endogenous bone repair, complete with periosteal chondrogenesis, thus reopening the “regeneration window” and thereby recreating a regeneration‐permissive environment that is responsive to exogenous BMP2 treatment.


INTRODUCTION
Salamanders display astonishing regenerative ability whereas mammals, such as mice and humans, appear resistant to regeneration, responding to amputation injury only by wound repair and scar formation. Closer analysis reveals that mice and humans do indeed have regenerative capabilities, but these are restricted to portions of the distal digit tip, the terminal phalanx (P3) (Borgens, 1982;Douglas, 1972;Illingworth, 1974). The regeneration response is amputationlevel-specific. Amputation levels proximal to the P3 nail matrix fail to elicit a significant regeneration response, resulting in bone truncation and soft tissue scar formation. In contrast, distal level amputations result in the restoration of lost structures (Agrawal, Kelly et al., 2011;Agrawal et al., 2010;Dawson et al., 2016;Fernando et al., 2011;Han, Yang, Lee, Allan, & Muneoka, 2008;Masaki via intramembranous ossification, and serves as a mammalian model for successful epimorphic regeneration (Fernando et al., 2011;Simkin, Sammarco, Dawson, Schanes et al., 2015;Simkin, Sammarco, Dawson, Tucker et al., 2015).
Amputation of the middle phalanx (P2) is regenerationincompetent and has emerged as a system to investigate regenerative failure and as a site to test approaches aimed at enhancing regeneration (Agrawal, Kelly et al., 2011;Agrawal et al., 2010;Dawson et al., 2016;Miura et al., 2015;Mu et al., 2013;Yu et al., 2012). Comparable to other long bones, P2 amputation results in bone truncation and skin healing with fibrotic scar formation.
Recent studies have shown that, despite the absence of a regenerative response, P2 amputation initiates a dynamic repair response similar to fracture healing and displays signs of soft tissue regeneration (Dawson et al., 2016). This finding is important since mammalian bones typically respond to fracture injury by successfully regenerating new bone to bridge the fracture (Einhorn, 2005;Shapiro, 2008). The P2 bone responds to amputation by forming a peripheral cartilaginous callus that is derived from the periosteum, osteoblast recruitment to form woven bone, and remodeling of the woven bone into a lamellar structure that resembles the original bone tissue (Dawson et al., 2016). This is a highly dynamic response that increases local stump bone volume only to remodel it back to the original stump morphology. We hypothesize that this repair response is analogous to a failed attempt at bone regeneration.
It is well documented that bone morphogenetic protein (BMP) signaling is required for endogenous digit tip regeneration (Han, Yang, Farrington, & Muneoka, 2003;Yu, Han et al., 2010). Indeed, the local application of BMP2 induces level-specific regeneration of neonatal digit amputations Yu et al., 2012) as well as regeneration of neonatal and adult limb amputations (Ide, 2012;Masaki & Ide, 2007;Yu et al., 2012). For induced neonatal digit regeneration, a microcarrier bead is used for targeted application of BMP2 that stimulates the formation of an endochondral ossification center that organizes the patterned ossification response (Yu et al., 2012). The BMP2 response in limb amputations is less well characterized, in part because the amputated limb stump is large and an adequate BMP2 vehicle has not been identified. Urist originally identified BMP2 as an agent that induced ectopic bone formation when implanted in vivo (Reddi & Huggins, 1972;Urist, 1965;Wozney et al., 1988). More recently, BMPs have been shown to act redundantly during skeletal development, functioning early in mesenchymal cell condensation and later as a requirement for endochondral ossification Barna & Niswander, 2007). In adults, BMP2 is important in the early periosteal response to fracture Wang, Huang, Xue, & Zhang, 2011), and is required for cell differentiation during the many stages of fracture healing (Minear, Leucht, Miller, & Helms, 2010;Wang et al., 2011;Yu, Lieu et al., 2010). Together, these studies demonstrate that BMP signaling plays a critical role in both skeletal development and endogenous skeletal repair and regeneration, and can be used as a treatment to enhance skeletal regeneration. In this study, P2 level amputation of the adult mouse digit is induced to regenerate by treatment with a BMP2 slow release vehicle. The induced regeneration response was characterized in detail using quantitative micro-computed tomography ( CT), histology, and immunohistochemistry. BMP2 treatment induced endochondral ossification at the amputation site that completely restored the amputated P2 bone length, but not the joint or the distal P3 skeletal element. The regenerated bone integrates with the P2 stump and tendon repair was observed. Histologically, BMP2-induced regeneration is mediated by the formation of a distal chondrogenic callus that undergoes subsequent ossification. Distal callus formation requires the stump periosteum and the formation of the peripheral chondrogenic callus that forms during the amputation response (Dawson et al., 2016). The responsiveness to BMP2 is transient suggesting a dynamic wound environment that we have defined as the "regeneration window." The healing of long-term amputations results in a stump that is refractory to BMP2 treatment; however, re-injury of the P2 stump reinitiates the formation of the peripheral callus as well as the responsiveness to BMP2 treatment. These studies identify and characterize a mechanism of BMP2-induced regeneration following digit amputation in adult mice.

BMP2-induced P2 regeneration
Amputation of the middle phalanx (P2) is a model system to investigate regenerative failure as well as induced regeneration of the skeletal element and associated soft tissues Agrawal et al., 2010;Dawson et al., 2016;Miura et al., 2015;Mu et al., 2013;Yu et al., 2012). Digit amputation transects the P2 diaphysis, the dorsal ligament, the skin and associated structures including the dermis and hair follicles, and the ventral fibrocartilage and associated tendon (Fig. 1A, B). In neonatal mice, targeted BMP2 treatment using a microcarrier bead vehicle after wound closure induced a P2 segment-specific regenerative response (Yu et al., 2012). Using a similar approach but with a slow release vehicle, we implanted BMP2 (n = 5) or bovine serum albumin (BSA) (n = 5) containing XeroGel at the time of wound closure (9 days post amputation [DPA]) between the wound epidermis and stump bone. At the time of treatment, the amputated P2 stump bone displayed a lateral periosteal chondrogenic callus and the digit stump was capped distally by a wound epidermis (Dawson et al., 2016). Experimental (BMP2 XeroGel) and control (BSA XeroGel) treated digits were followed using CT imaging for 160 days and changes in phalangeal length and anatomy were quantified (Fig. 1C).
CT 3D rendered images at 1 day post implantation (DPI) show no gross anatomical differences between BMP2 and BSA treatment, with both displaying periosteal remodeling (Fig. 1C, arrowheads). By 7 DPI, the stump bones from both treatment groups show evidence of peripheral callus formation associated with the periosteum (Fig. 1C).
The peripheral callus typically forms following simple P2 amputation (Dawson et al., 2016). By 14 DPI, four of five BMP2-treated digits show the formation of a boney distal callus that is contiguous with the stump bone and extends the length of the P2 element (Fig. 1C). The distal regeneration of bone results in a significant increase in bone length compared to BSA control digits (Fig. 1D, P ≤ 0.05). The lengthening To investigate the mechanism of BMP2-induced regeneration, we carried out detailed histological and immunological analyses at 8 DPI.
At 8 DPI, we observe a large mass of cells that formed a cartilaginous callus distal to the amputation plane associated with the BMP2 Xero-Gel vehicle ( Fig. 2A, B). The cartilaginous distal callus is contiguous with, and appears to be an expansion of, the peripheral stump callus

Periosteum removal inhibits BMP2-induced P2 regeneration
The periosteum is required for P2 peripheral callus formation (Dawson et al., 2016). To investigate the role of periosteal tissue in induced regeneration, BMP2 (n = 4) or BSA (n = 4) XeroGel was implanted into To assess for cartilaginous growth in response to BMP2 or BSA treatment in periosteum-removed digits, we harvested treated digits at 8 DPI and performed detailed histological and immunological analysis. BMP2-or BSA-treated digits show no cartilaginous or boney growth adjacent to the periosteum or distal to the amputation plane, yet exhibit boney growth within the endosteal/marrow space (Fig. 3C, F). Accordingly, immunostaining for ACAN revealed no positive chondrogenic signal in serial sections of BMP2-or BSA-treated digits ( Fig. 3D, G). Analogous to the histological findings, immunostaining using the osteoblast marker Osterix (OSX) showed immunopositive cells within the endosteal/marrow region of the BMP2-or BSA-treated digits, yet we did not observe positive signal adjacent to the external bone surface or distal to the amputation plane (Fig. 3E, H). Our previous cell lineage studies showed that the endosteum/marrow space of P2 contributed osteoblasts in response to injury, congruent with our findings here (Dawson et al., 2016). Therefore, it appears that P2 endosteal/marrow derived osteoblasts are not the target of BMP2 treatment and furthermore cannot compensate for the periosteal tissue in induced regeneration. Taken together, these data provide evidence that the periosteal tissue is required for BMP2-induced P2 regeneration.

Regeneration window
BMP2-induced regeneration of the neonate P2 digit is an established model of induced mammalian regeneration Yu et al., 2012). Using this neonatal model, we explored whether the potential to respond to BMP2 changes as the amputation wound matures, thus inquiring if the efficacy of induced regeneration is modulated at different stages of wound healing. BMP2 treatment of the neonatal digit at 4 DPA, the approximate time of wound closure, restores the P2 skeletal length (Yu et al., 2012). BMP2-induced regeneration of the neonate digit is attenuated with treatment prior to wound closure or treatment at later stages of wound healing ( wound maturation, and thus identifies a "regeneration window" with respect to BMP2 responsiveness during the wound healing process. To investigate the regeneration window in the adult digit amputation model, digits were amputated at the P2 level and allowed to heal for 24 days, over 2 weeks post wound closure. By 24 DPA, the P2 cartilaginous peripheral callus had been completely replaced with woven bone (Fig. 4B). Immunostaining for ACAN confirmed the absence of chondrocytes at 24 DPA (Fig. 4C), and immunostaining for OSX revealed positive signal within the central marrow region and in the woven bone of the peripheral callus (Fig. 4D, arrowheads). At 24 DPA, BMP2 or BSA XeroGel (n = 4) was implanted into the distal digit tip between the stump bone and the apical epidermis. BMP2-induced regeneration was quantified using CT scans over the next 4 weeks.
CT renderings of 24 DPA BMP2-or BSA-treated digits show that, by 1 DPI, both treatment groups had previously undergone peripheral callus formation and ossification (Fig. 4E). At 14, 21, and 28 DPI, time points typically associated with a robust BMP2-induced regeneration response, no distal elongation of bone in response to BMP2 treatment was observed. Rather, continued remodeling of the peripheral callus in BMP2-and BSA-treated digits was apparent (Fig. 4E). Statistical analysis confirms no significant change in bone length from 1 DPI to 28 DPI and no significant differences between the two treatment groups (Fig. 4F, P > 0.05). 24 DPA BMP2-treated and BSA control digits were analyzed at 8 DPI to assess for cartilaginous distal callus formation (Fig. 4G, H). 8 DPI 24 DPA BMP2-treated digits displayed some small clusters of peripheral chondrocytes (arrowhead); however, we did not observe the cartilaginous distal callus that typifies induced regeneration (Fig. 4G). There was no indication of cartilaginous growth in 8 DPI 24 DPA BSA control digits (Fig. 4H). These findings point to a regeneration window of BMP2 responsiveness during amputation healing of the adult digit and, coupled with the periosteal requirement for induced regeneration, suggest that this window may be linked to dynamic changes associated with the chondrogenic response of the injured periosteal tissue.
Working under the general hypothesis that active chondrogenesis is a target for exogenous BMP2 we carried out re-injury studies to determine if a healed amputation wound (BMP2 unresponsive) could be stimulated to become BMP2 responsive. In other words, can the regeneration window be reopened? P2 digits were amputated and allowed F I G U R E 4 P2 wound maturation attenuates BMP2-induced regeneration. (A) Neonate P2 skeletal length is restored after BMP2 treatment at 4 DPA, the approximate time of wound closure. BMP2 treatment prior to or after wound closure attenuates the regeneration response (ANOVA, ±SEM, *P or ♦P ≤ 0.05, **P ≤ 0.01, ***P or ♦♦♦P ≤ 0.001, ****P or ♦♦♦♦P ≤ 0.0001; asterisks denote significance in BMP2-treated digits compared to BMP2 treatment at 4 DPA; diamonds denote significance between BMP2 and BSA treatment). as well as new cartilaginous growth associated with the peripheral callus that formed external to the initial periosteal response (Fig. 5B, callus outlined). ACAN immunostaining is localized to the margin of the new peripheral callus, but not distal to the amputation plane; thus the re-injury healing response recapitulates the cartilaginous peripheral response of the original amputation injury (Fig. 5C).
To test if healed and re-injured amputations can be induced to regenerate, the 24 DPA re-injured P2 bone stump was treated with BMP2 (n = 8) or BSA (n = 5) XeroGel 9 days after re-injury. Implants were placed between the newly formed wound epidermis and the reinjured bone stump. Digit regeneration was quantified using CT for 5 weeks. CT 3D renderings at 1 DPI indicate that the BMP2-treated and BSA control digits have undergone prior periosteal bone remodeling (Fig. 5D). Statistical analysis of bone length changes indicates a To determine if the re-injury regeneration response was mediated by a distal cartilaginous callus, we harvested samples at 10 days post BMP2 or BSA treatment for histological and immunological analysis.
At 10 DPI, a distal callus (outlined) has formed associated with the BMP2 source and shows evidence of chondrogenesis (Fig. 5F, inset).
Immunostaining for ACAN shows chondrocytes in the distal callus but largely associated with the boundaries of the callus and not present in the central region of the distal callus (Fig. 5G). ACAN positive cells are also observed directly associated with the BMP2 XeroGel implant (Fig. 5G, arrowheads). In control BSA-XeroGel-treated digits ACAN immunopositive chondrocytes were not observed distal to the amputation plane (Fig. 5J). While chondrocytes were not observed in the central region of the distal callus, this region stained positive for the osteoblast marker OSX (Fig. 5H) indicating that the central region of  (Fig. 5H, K). The spatial organization of chondrocytes and osteoblasts within the distal callus was distinct from BMP2-induced regenerates from an initial amputation which raises the possibility that the response to re-injury may not be identical. Nevertheless, these studies support the conclusion that the regeneration window of previously healed amputation injuries has the potential to be "reopened" by re-injury, thus reinitiating the repair response, and the cells of the wound are able to mount a regenerative response to BMP2 treatment.

DISCUSSION
The mouse digit is a unique model system to investigate regeneration; not only is it proof of concept that mammals can indeed regenerate certain structures, such as after distal P3 amputation, it is also a system in which to assess regenerative failure and induced regeneration of P2.
By focusing on the healing response following P2 level amputations, a dynamic bone and soft tissue repair response was characterized (Dawson et al., 2016). The periosteum of the amputated bone reacts in a manner similar to fracture healing by forming a transient proliferative chondrogenic callus that builds new bone peripherally only to have the new bone remodeled back to the original stump morphology. In the current study we have found that targeted treatment with BMP2 can modify this periosteal response to create a chondrogenic distal callus that facilitates the regeneration of new bone to restore the amputated P2 bone. While imperfect, the regenerated new bone is anatomically integrated with the stump bone, possesses a bone marrow that is contiguous with the stump marrow and, after remodeling, displays anatomical features of the amputated P2 bone. These structural similarities are suggestive that the BMP2-induced regenerated bone is appropriately patterned, which is consistent with previous studies on BMP2-induced regeneration in neonates (Yu et al., 2012). Further, we find evidence that some of the surrounding soft tissue, i.e. tendon fibers, undergo repair in conjunction with the newly forming bone.
These findings support the general conclusion that targeted growth factor treatment to modify/extend intrinsic repair mechanisms can be effective for enhancing regeneration following amputation. Stimulating skeletal regeneration by distraction osteogenesis also involves enhancing intrinsic repair mechanisms (Ilizarov, 1989); however, distraction osteogenesis induces the regeneration of new bone by direct ossification (Dhaliwal, Kunchur, & Farhadieh, 2016;Hvid, Horn, Huhnstock, & Steen, 2016) and so it is distinct from BMP2-induced regeneration which stimulates formation of new bone de novo by endochondral ossification.
BMP2 stimulates segment-specific P2 regeneration by inducing the formation of a distal chondrogenic callus that is contiguous with the peripheral callus that forms in response to simple amputation.
Since BMP2-induced regeneration is inhibited when the periosteum is surgically removed, the evidence suggests that periosteal cells of the stump play a role in the induced regeneration response. The periosteum is critical for callus formation in fracture repair and amputation healing (Colnot, 2009;Dawson et al., 2016); thus the formation of the peripheral chondrogenic callus is a prerequisite for the induced regeneration response. Periosteal chondrogenesis has been linked to the upregulation of BMP2 at the fracture site , periosteal-derived BMP2 is required for chondrogenic induction after bone injury (Wang et al., 2011), and periosteal cells form cartilage in response to BMP2, while BMP2 functions to induce osteogenesis (Minear et al., 2010) or to have no effect on endosteal cells (Yu, Lieu et al., 2010). These findings support the conclusion that BMP2 is a key endogenous factor controlling the periosteal response to injury and that treatment with BMP2 enhances an endogenous response by extending the period of BMP2 signaling.
In neonatal digits, BMP2-induced P2 regeneration does not involve callus formation, but induces the formation of a distal endochondral ossification center (Yu et al., 2012) that is similar to the distal callus observed in the current study in that they both mediate endochondral ossification. In addition to induced proliferation, BMP2 enhances cell recruitment to the wound environment in the neonate digit model by inducing SDF-1 production which recruits CXCR4 positive cells to the amputation wound . SDF-1 /CXCR4 signaling has been implicated in cell recruitment in the regenerating zebrafish fin (Dufourcq & Vriz, 2006), mouse digit tip regeneration , fracture repair (Kitaori et al., 2009), and BMP2-induced ectopic bone formation (Otsuru, Tamai, Yamazaki, Yoshikawa, & Kaneda, 2008). In BMP2-treated adult P2 amputations, the expression of CXCR4 by cells of the induced distal callus is suggestive that a similar mechanism acts to recruit cells to the distal digit stump (Fig. S1). These findings support a model in which BMP2 also functions indirectly to recruit cells that form the distal callus and, by doing so, provides a distal signal that organizes the induced regeneration response.
With this in mind, regenerative failure can result from a defect in the absence of regeneration-competent cells and/or one or all of the stepwise interactions. In any regeneration-incompetent injury model, induced regeneration indicates that responding cells have the potential to mount a regeneration response, yet fail to do so under normal conditions. Thus, the success of P2 regeneration indicates that regenerative failure stems from a restriction in BMP2 signaling and is not linked to an inherent inability of cells to respond to a regenerationpermissive wound environment. The results from a growing number of studies including matrix implantation (Agrawal, Kelly et al., 2011), histolytic degradation (Agrawal et al., 2012;Sammarco et al., 2015) and cell transplantation between regeneration-incompetent and regeneration-competent tissues (Wu et al., 2013) all support the general conclusion that P2 regenerative failure is causally linked to defects in the wound environment and is not limited by the availability of responsive cells. We also provide evidence that the wound environment is dynamic with respect to BMP2 responsiveness, transitioning from a responsive phase to a non-responsive phase that reflects a regeneration window during the healing response. A similar transitioning of the wound environment has been observed associated with BMP2 treatment in segmental bone defects (Hussein et al., 2012). The dynamics of a regeneration window are expected to be specific for the inducing agent and probably modify the responsiveness of wound cells to other regeneration-inducing treatments. The important conclusion is that the wound site is dynamically changing and therefore predicted to be differentially responsive to regeneration-inducing agents.
While this conclusion seems daunting with respect to understanding regenerative competency in mammals, the demonstration that a regeneration-permissive wound environment can be restored by re-injury provides encouragement that cells can regain regenerative capabilities long after an unsuccessful repair response. This indicates that a failed regenerative response does not modify the regenerative potential of cells at the wound site; they are able to initiate a healing response that can transition into a period of responsiveness to a regeneration-inducing agent such as BMP2. This conclusion has important clinical implications for potentially reactivating regenerative potential in amputated limbs after stump healing is completed.
Broadly speaking, regeneration is divided into categories designed to organize and contrast the various regenerative and reparative responses that occur in nature. Epimorphic regeneration, the designation within which amphibian limb and mammalian distal digit tip regeneration belong, is mediated by a proliferative blastema, a transient undifferentiated structure that mediates the restoration of the amputated structure (Carlson, 2007). The digit blastema forms by recruitment of cells from multiple tissue types to create a proliferative heterogeneous population that redifferentiates in a lineage-specific manner (Fernando et al., 2011;Lee et al., 2013;Lehoczky et al., 2011;Rinkevich et al., 2011;Simkin, Sammarco, Dawson, Schanes et al., 2015;Wu et al., 2013). Conversely, skeletal regeneration in response to fracture is defined as a tissue repair response not mediated by blastema formation (Carlson, 2007). Yet, like the blastema, the chondrogenic callus that forms in fracture healing and digit amputation is proliferative, displays similar recruitment characteristics, and is macrophage dependent (Miura et al., 2015, Simkin et al., 2017. What distinguishes the callus from a blastema is the differentiative state and relative homogeneity of composite cells (chondrogenic), and the fact that fracture healing is a nerve-independent process (Miura et al., 2015) which may be linked to the lack of cellular heterogeneity. The homogeneity of cells that make up the callus is strikingly similar to the cellular composition of the deer antler blastema that mediates their annual regeneration (Kierdorf, Kierdorf, & Szuwart, 2007). Based on the results presented here, the chondrogenic callus displays blastema characteristics and possesses considerable regenerative potential when properly stimulated. The distinction between a blastema and a callus is perhaps best considered in light of the idea that the blastema mediates intersegmental regeneration (i.e. the formation of segments) and the callus mediates intrasegmental regeneration (i.e. the formation of structures within a segment) (Satoh, Cummings, Bryant, & Gardiner, 2010). This delineation helps us to understand the relationship between epimorphic and tissue-specific regenerative responses.

Animals and surgery
Adult 8-week-old female C57Bl/6 mice were purchased from Charles River (Wilmington, MA) or bred in house at the Texas Institute for Genomic Medicine. Mice were anesthetized with isoflurane for all surgical procedures. Digits were amputated at the level of the second phalangeal element (P2; Fig. 1A, B) as described (Dawson et al., 2016).
Digits used in the P2 re-injury study were initially amputated at the P2 level and were allowed to heal for 24 days. At 24 DPA the apex of the healed digit stump skin was removed and the stump bone was re-injured by removing the distal cap of healed bone that encloses the bone marrow (Dawson et al., 2016). After re-injury, the stump wound was allowed to heal conservatively. For periosteum removal studies, the digits were amputated at P2 and the periosteum was mechanically removed as previously described (Dawson et al., 2016).
The distal amputation wound was closed with Dermabond (Ethicon, Somerville, NJ). Sham control digits were treated identically, but the periosteum was not removed. All animal use and techniques were compliant with the standard operating procedures of the Institutional Animal Care and Use Committees at Tulane University and the College of Veterinary Medicine and Biomedical Sciences at Texas A&M University.
At the time of treatment, a slow release implant containing either BMP2 (0.5 g/ L) or BSA (0.1%) was inserted into a surgically created pocket between the wound epidermis and the P2 bone stump, and the wound was closed using Dermabond. The slow release vehicle used was a sol−gel silica-based porous glass (XeroGel, Entellus Medical, Plymouth, MN). XeroGel is a hemostatic packing that uses a unique blend of polyethylene glycol and chitosan to separate tissues and prevent adhesions between mucosal surfaces, and functions as a slow release vehicle for embedded growth factors for up to 63 days (Agrawal & Sinha, 2016;Nicoll, Radin, Santos, Tuan, & Ducheyne, 1997;Santos, Radin, & Ducheyne, 1999). BMP2 or BSA containing XeroGel was prepared as described by Santos et al. (1999). Protein containing XeroGel solution was aliquoted into 1.0 L drops, air-dried at room temperature, and stored at −20 • C until use.

Histology and immunohistochemistry
Tissue processing has been previously described (Dawson et al., 2016).

Micro-computed tomography ( CT) scans, length, and image processing
P2 digits were scanned at weekly intervals beginning at 1 day post BMP2 or BSA implant (DPI) for 5 total weeks, with some treatment groups receiving an end-point scan at 160 DPI. Scans were performed using the vivaCT 40 (SCANCO Medical, Wayne, PA) as previously described (Dawson et al., 2016;Fernando et al., 2011). P2 digits were scanned at a voxel size of 10.5 m and energy of 45 kVp; 1000 projections per 180 o were captured at 380 msec using continuous rotation.
CT images were saved as a series of dicom files, and the dicom sequences were uploaded to ImageJ. Using the BoneJ (Doube et al., 2010) (Version 1.2.1) Optimized Threshold Plugin for ImageJ, images were segmented and converted to 3D renderings, and length measurements were quantified as previously described (Dawson et al., 2016).