From hematoma to lamellar bone — the cellular cascade of secondary fracture repair and how recombinant BMP-2/BMP-7 accelerate osteogenesis
The instant a bone fractures, the injury is not merely mechanical — it is vascular. Cortical, periosteal, and medullary blood vessels tear, and within minutes a hematoma fills the fracture gap and surrounding soft tissue. Far from being inert debris, this clot is the biological seed of the entire repair cascade: a fibrin-rich scaffold saturated with growth factors that converts a sterile injury into a coordinated regenerative program.
Within the fracture hematoma, activated platelets degranulate their alpha granules, releasing platelet-derived growth factor (PDGF-BB), transforming growth factor-beta (TGF-β1), and vascular endothelial growth factor (VEGF) into the clot matrix. This growth-factor cocktail is the earliest chemotactic signal of fracture repair — it draws circulating and resident inflammatory cells to the injury site within hours.
Neutrophils arrive first (peak at 24 hours), clearing necrotic debris and releasing reactive oxygen species and proteases that further remodel the clot. Macrophages follow (peak day 2–3) and become the dominant orchestrators of the early inflammatory phase: they polarize from a pro-inflammatory M1 phenotype (secreting IL-1β, IL-6, TNF-α) toward a pro-regenerative M2 phenotype (secreting IL-10, TGF-β) over the first week, a transition now recognized as a rate-limiting checkpoint for healing — macrophage depletion studies in rodents delay callus formation by 2–3 fold.
IL-6 in particular acts systemically as well as locally: it stimulates hepatic acute-phase protein synthesis and, at the fracture site, promotes angiogenesis and osteoclast precursor recruitment. TNF-α, in modest and transient amounts, is required for normal healing (TNF-receptor knockout mice show delayed union), but chronic or excessive TNF-α — as seen in smokers, diabetics, or NSAID-treated patients — suppresses MSC osteogenic differentiation and is a recognized contributor to delayed union.
By 48–72 hours, mesenchymal stem cells (MSCs) — recruited from the cambium layer of the periosteum, the endosteum, bone marrow stroma, and circulating pericyte populations — have homed to the fracture site along SDF-1/CXCR4 chemokine gradients. These MSCs are multipotent: their downstream fate (chondrocyte vs. osteoblast) is dictated largely by local oxygen tension and mechanical strain, setting the stage for the divergent healing paths of Stage 2.
Approximately 5–10% of long-bone fractures progress to delayed union or nonunion, and a poorly regulated or prolonged inflammatory phase — driven by smoking, NSAID use, diabetes, or open/high-energy trauma — is among the most consistently identified modifiable risk factors, which is why many trauma protocols now restrict NSAID use in the first 2 weeks after fracture.
As inflammation subsides, the fracture site transitions into an anabolic phase. Recruited MSCs differentiate along a chondrogenic path, producing a fibrocartilaginous "soft callus" that mechanically splints the fracture — not by mineral rigidity, but by the compressive resilience of cartilage matrix, exactly like the tissue found in growth plates and joint surfaces.
Chondrogenic commitment of MSCs is governed by the transcription factor SOX9, activated synergistically by TGF-β/BMP signaling and by the hypoxic microenvironment at the fracture gap core. Oxygen tension there falls below 5% — too low to support osteoblastic bone formation (which requires a well-vascularized niche) but well tolerated by chondrocytes, which rely on anaerobic glycolysis. This hypoxia is sensed through HIF-1α stabilization, which directly upregulates SOX9 and downstream cartilage matrix genes (Col2a1, aggrecan).
The resulting fibrocartilaginous callus is mechanically distinct from bone: it tolerates the substantial interfragmentary strain still present at this stage (secondary bone healing, unlike the near-zero-strain environment required for primary/direct healing under rigid fixation, tolerates strains up to 10%). This is the biological logic behind the Perren "strain theory" of fracture healing — tissue differentiation follows a strain-dependent hierarchy: granulation tissue tolerates >100% strain, cartilage ~10–15%, and lamellar bone only ~2%.
Simultaneously, VEGF released by hypertrophying chondrocytes and macrophages drives angiogenic sprouting from the periosteal and medullary vasculature into the periphery of the callus. This invading capillary network does not yet penetrate the cartilage core — it forms a vascular envelope around it, setting up the spatial template for the endochondral ossification front that will sweep centripetally inward during Stage 3. Callus size at this stage correlates inversely with fixation rigidity: fractures treated with flexible intramedullary nailing or casting (secondary healing) form large, biomechanically robust external calluses, while rigidly plated fractures form minimal callus and heal by direct osteonal remodeling instead.
The soft cartilaginous callus is a temporary scaffold, not a permanent tissue. Over weeks 3 through 6, it is systematically replaced by woven bone through endochondral ossification — the same fundamental process that lengthens the fetal skeleton and the long-bone growth plate, here redeployed for repair and driven by a local surge of BMP-2 and BMP-7 signaling.
Chondrocytes at the center of the soft callus undergo a striking phenotypic switch: they stop proliferating, enlarge roughly fivefold (hypertrophy), and begin expressing type X collagen and alkaline phosphatase — enzymes and matrix proteins that promote calcification of the surrounding cartilage matrix. Hypertrophic chondrocytes also secrete VEGF at far higher levels than proliferative chondrocytes, actively recruiting the vascular invasion that will remodel them.
As the matrix calcifies, hypertrophic chondrocytes undergo programmed cell death, leaving behind a honeycomb of calcified cartilage "cores." Invading capillary buds — carrying osteoblast precursors, osteoclast precursors, and perivascular MSCs — infiltrate this scaffold. Osteoclast-like chondroclasts resorb the calcified cartilage while osteoblasts deposit osteoid directly onto the residual calcified cartilage cores, producing woven bone: a disorganized, high-turnover bone matrix with randomly oriented collagen fibrils that is mechanically inferior to lamellar bone but forms rapidly.
This entire process is orchestrated by a local surge of bone morphogenetic proteins, chiefly BMP-2 and BMP-7 (also called osteogenic protein-1, OP-1), secreted by hypertrophic chondrocytes, osteoblasts, and platelets. BMPs signal through heterodimeric type I/II serine-threonine kinase receptors (BMPR-IA, BMPR-IB, BMPR-II), phosphorylating SMAD1/5/8, which complexes with SMAD4 and translocates to the nucleus to activate RUNX2 — the master transcription factor for osteoblast differentiation. RUNX2 in turn drives expression of osteocalcin, osteopontin, and bone sialoprotein, the structural and regulatory proteins of the woven bone matrix. This endogenous BMP surge is precisely what recombinant BMP-2/BMP-7 therapy (Stage 4) seeks to amplify or substitute for when natural healing is insufficient.
When the natural BMP surge is insufficient — in open fractures with soft-tissue and periosteal loss, segmental bone defects, or established nonunions — recombinant human BMP-2 (rhBMP-2, marketed as InductOs in the EU and Infuse in the US) or BMP-7/OP-1 can be delivered directly to the fracture site on an absorbable collagen sponge (ACS) carrier, supraphysiologically driving osteoinduction. This is one of the few growth-factor therapies with a genuine, if narrow, place in orthopedic clinical practice.
rhBMP-2 is produced recombinantly in Chinese hamster ovary (CHO) cells and lyophilized onto a bovine type I collagen sponge, which is rehydrated intraoperatively and packed into or around the fracture site. The BESTT trial (Governale et al., and subsequent multicenter RCTs) evaluating rhBMP-2 for open tibial shaft fractures treated with intramedullary nailing found a significant reduction in secondary surgical interventions (hardware exchange, bone grafting, delayed wound closure) compared with standard care, particularly for Gustilo-Anderson type IIIA/IIIB open fractures. BMP-7/OP-1 has similarly been used, chiefly for tibial nonunions as an alternative to autologous iliac crest bone graft, under a humanitarian device exemption.
The dose-response relationship is markedly non-linear and clinically important: physiologic BMP concentrations at a healing fracture are in the low tens of ng/mL, whereas the concentration delivered locally by an ACS-BMP-2 implant is on the order of 1–2 mg/mL — a supraphysiologic excess of three to five orders of magnitude, necessary because of rapid diffusion, proteolytic degradation, and the comparatively low binding affinity/retention of BMP-2 on the collagen carrier (much of the loaded dose diffuses away within days).
This supraphysiologic dosing carries a well-documented risk: heterotopic ossification (ectopic bone formation in surrounding soft tissue), reported in a meaningful minority of spine and some orthopedic applications, along with dose-dependent local swelling/seroma and, in spine fusion literature, associations with increased short-term cancer risk that led the FDA to issue a public health notification in 2008 and substantially curtailed off-label spine use. For fracture indications the risk-benefit profile is more favorable, but the cost (~$5,000 per treatment kit versus a fraction of that for autograft) and these safety signals mean rhBMP-2/7 use is now reserved for cases — open fractures with major bone loss, recalcitrant nonunion, patients for whom autograft harvest is undesirable — where the biology genuinely cannot proceed without pharmacologic amplification.
The Medtronic-sponsored 2011 Yale Open Data Access (YODA) independent reanalysis of rhBMP-2 spine trials found substantially higher adverse-event rates (including possible cancer signal) than originally published — a landmark case study in reanalysis of industry-funded device trials that reshaped FDA post-market surveillance requirements for orthobiologics.
Bridging callus is not the end of fracture repair — it is a mechanically crude intermediate that must be sculpted into definitive, load-adapted bone. Over months to years, coupled osteoclast resorption and osteoblast formation progressively replace disorganized woven bone with lamellar bone organized along the principal mechanical stress trajectories of the healed limb, ultimately restoring near-normal strength and, in growing bone, near-normal external contour.
Remodeling proceeds through discrete basic multicellular units (BMUs) — coordinated teams of osteoclasts and osteoblasts that travel through bone as a "cutting cone" (resorption front) trailed by a "closing cone" (formation front), converting woven bone and residual calcified cartilage into cylindrical osteons of lamellar bone oriented along the local dominant loading axis. Each BMU cycle — activation, resorption, reversal, formation — takes roughly 4 to 6 months in adult cortical bone.
The coupling between resorption and formation is governed principally by the RANKL/RANK/OPG axis. Osteoblasts and osteocytes express RANKL (receptor activator of NF-κB ligand), which binds RANK on osteoclast precursors to drive their differentiation and activation; osteoprotegerin (OPG), a soluble decoy receptor also secreted by osteoblasts, competitively inhibits this interaction. The RANKL:OPG ratio is the master rheostat of local bone turnover — mechanical loading, estrogen, and vitamin D all act substantially by shifting this ratio. As osteoclasts resorb bone, growth factors embedded in the matrix (TGF-β, IGF-1, BMPs) are released and recruit the next wave of osteoblasts, coupling resorption to formation and preventing a net loss of bone mass under normal conditions.
Mechanical strain sensed by osteocytes — the "mechanostat" of Harold Frost's theory — governs where remodeling reinforces bone (Wolff's Law): osteocytes in low-strain regions secrete sclerostin, which inhibits osteoblast Wnt/β-catenin signaling and permits net resorption, while osteocytes under higher physiological strain suppress sclerostin, favoring net formation. Over the remodeling phase, this feedback progressively reshapes the bulky external callus — no longer mechanically necessary once the fracture is bridged — down toward the original cortical diameter, while internally reorganizing trabecular and osteonal architecture along the limb's restored load-bearing axis. By 1–2 years post-fracture, mechanical strength in an uncomplicated healed long bone typically approaches 90–100% of the pre-injury intact bone, though the site may remain radiographically distinguishable (denser cortex, subtle contour change) indefinitely.
Osteoporosis drugs illustrate the RANKL/OPG axis pharmacologically in reverse: denosumab is a monoclonal antibody that mimics OPG by binding RANKL directly, suppressing osteoclast activity — the same pathway that couples resorption to formation during fracture remodeling is now a first-line therapeutic target for preventing fragility fractures in the first place.