Comparative osteogenic performance of Chinese hamster ovary cell-derived versus Escherichia coli-derived recombinant human bone morphogenetic protein-2 Authors: Charles Matthews, Lyndsey Johnson, Whitney Holden, John von Benecke Published: 2026 ABSTRACT AbstractAimsRecombinant human bone morphogenetic protein-2 (rhBMP-2) is used clinically for bone regeneration. The Food and Drug Administration-approved clinical product (Infuse) is Chinese hamster ovary (CHO)-derived, but Escherichia coli (EC)-derived rhBMP-2 may offer manufacturing advantages. This study compared CHO- and EC-derived rhBMP-2 potency in vitro and in vivo.MethodsFour EC-derived rhBMP-2 preparations were screened using alkaline phosphatase (ALP) assays in C2C12 and W20-17 cells. The most potent (EC-MAX) was compared to CHO-rhBMP-2. A total of 15 male athymic rats received bilateral absorbable collagen sponge (ACS) implants loaded with 17 µg rhBMP-2 (CHO or EC-MAX) in muscle pouches for four weeks. Bone formation was assessed by radiography, micro-CT, and histology.ResultsEC-MAX showed 2.2-fold higher osteogenic activity in W20-17 cells and 1.3-fold in C2C12 cells versus CHO-rhBMP-2. In vivo, EC-MAX produced ~ ten-fold greater bone volume (88.5 mm³ (SD 12.8) vs 9.2 mm³ (SD 5.4)) and tissue volume (203.3 mm³ (SD 47.9) vs 15.0 mm³ (SD 8.8)). Histology confirmed more mature bone architecture in EC-MAX implants with adipocyte-filled marrow spaces.ConclusionPotency cannot be generalized across products. Inter-supplier potency variation exists among EC-derived rhBMP-2 products. EC-MAX demonstrated superior osteoinductive potential both in vitro and in vivo versus CHO. However, substantial variability highlights the need for product-specific characterization.Cite this article: Bone Joint Res 2026;15(6):686–695.Keywords: rhBMP-2, Osteogenesis, Escherichia coli, recombinant human bone morphogenetic protein-2, rat model, Micro-CT, Bone formation, radiography, Collagen, Alkaline phosphatase, bone regeneration, ossicles Article focus Does expression system (Chinese hamster ovary (CHO) vs Escherichia coli (EC)) influence recombinant human bone morphogenetic protein-2 (rhBMP-2) osteoinductive potency? Does expression system (Chinese hamster ovary (CHO) vs Escherichia coli (EC)) influence recombinant human bone morphogenetic protein-2 (rhBMP-2) osteoinductive potency? Is there inter-supplier variability among EC-derived rhBMP-2 products? Is there inter-supplier variability among EC-derived rhBMP-2 products? What structural differences might explain observed potency variation? What structural differences might explain observed potency variation? Key messages The most potent EC-derived rhBMP-2 demonstrated significantly greater osteoinductive responses in the models tested. The most potent EC-derived rhBMP-2 demonstrated significantly greater osteoinductive responses in the models tested. Substantial inter-supplier potency variation exists, reconciling conflicting literature reports. Substantial inter-supplier potency variation exists, reconciling conflicting literature reports. rhBMP-2 potency cannot be assumed equivalent based on protein mass; each product requires independent validation. rhBMP-2 potency cannot be assumed equivalent based on protein mass; each product requires independent validation. Strengths and limitations This study had a comprehensive methodology: 10 × 10 batch in vitro comparisons across two cell lines and analysis of structural characterization, including mass spectrometry, nuclear magnetic resonance (NMR), and circular dichroism (CD). The use of ASTM International (ASTM) standard assessment of ectopic bone formation provided the in vivo assessment. This study had a comprehensive methodology: 10 × 10 batch in vitro comparisons across two cell lines and analysis of structural characterization, including mass spectrometry, nuclear magnetic resonance (NMR), and circular dichroism (CD). The use of ASTM International (ASTM) standard assessment of ectopic bone formation provided the in vivo assessment. This was an in vivo study with direct head-to-head comparison using identical doses, carriers, and procedures. This was an in vivo study with direct head-to-head comparison using identical doses, carriers, and procedures. Ectopic model limitations include a non-orthotopic site, masked immunogenicity in athymic rats, and accelerated rodent bone metabolism. Ectopic model limitations include a non-orthotopic site, masked immunogenicity in athymic rats, and accelerated rodent bone metabolism. Introduction Bone morphogenetic protein-2 (BMP-2) drives bone repair and regeneration.1 Recombinant human BMP-2 (rhBMP-2) from Chinese hamster ovary (CHO) cells (Infuse; Medtronic, USA) has been used clinically for decades,2 but CHO production is resource-intensive.3,4Escherichia coli (EC) expression may reduce costs,3,4 although preclinical studies report inconsistent EC-derived rhBMP-2 performance compared with CHO forms.5-9 CHO-derived rhBMP-2 is glycosylated, whereas EC-derived rhBMP-2 is non-glycosylated, as E. coli lacks N-linked glycosylation machinery.3,7,10-12 This may influence solubility, release kinetics, and potency.8 EC-derived rhBMP-2 expressed as inclusion bodies requires denaturation/refolding to restore activity;13,14 inefficiencies reduce specific activity. We compared CHO- and EC-derived rhBMP-2 directly, both in vitro and in vivo. This study illuminates how expression system differences influence rhBMP-2 potency, with implications for alternative rhBMP-2 therapeutic development and regulatory evaluation. Methods In vitro relative potency assays We evaluated rhBMP-2 from four EC-derived commercial sources - Cowellmedi (South Korea), CG Bio (South Korea), R&D Systems (USA), and Locate Bio (UK). Locate Bio material was provided under a confidentiality agreement; further details are available upon request. All materials were purchased on commercially available terms, with no financial support from any manufacturer. Each EC-derived sample was screened for osteogenic activity using cell-based assays. The most active preparation, designated EC-MAX, was selected for comparison with CHO-derived rhBMP-2 (Infuse). To minimize the variability associated with primary mesenchymal stem cells,15-18 we used immortalized cell lines: C2C12 myoblasts (ATCC CRL-1772) and W20-17 stromal cells (ATCC CRL-2623), both widely used for BMP-2 bioactivity assessment. Following ASTM F2131-21, the standard test method for measuring BMP osteoinductive activity, alkaline phosphatase (ALP) induction was used as a surrogate marker of osteogenic differentiation. Cells were plated in 96-well plates and exposed to serial dilutions of each rhBMP-2 preparation at 4.3 fold in the W20-17 assay and 2.2 fold in the C2C12 assay. The plates were exposed for either 48 hours (C2C12) or 24 hours (W20-17), as summarized in Table I. Table I.Alkaline phosphatase (ALP) assay cell types, treatment duration, and serial dilution ranges.PhaseCell typeDurationSerial dilution range*Initial screeningC2C12 mouse myoblasts (ATCC CRL-1772)48 hrs2.88 µg/ml to 0.07 µg/mlW-20-17 (ATCC CRL-2623) mouse stromal cells24 hrs1.369 µg/ml to 0.01 ng/ml10-lot × 10-lot testingW-20-17 (ATCC CRL-2623) mouse stromal cells24 hrs1.369 µg/ml to 0.01 ng/mlC2C12 mouse myoblasts (ATCC CRL-1772)48 hrs0.72 µg/ml to 0.001 µg/ml*Serial dilutions at 4.3-fold in the W-20-17 assay and 2.2-fold for the C2C12 assay. Alkaline phosphatase (ALP) assay cell types, treatment duration, and serial dilution ranges. Serial dilutions at 4.3-fold in the W-20-17 assay and 2.2-fold for the C2C12 assay. After treatment, cell lysates were collected, and ALP activity was quantified using a colorimetric assay based on p-nitrophenyl phosphate hydrolysis (Abcam ab83369, UK). Optical density at 450 nm was measured on a Tecan Infinite 200 PRO plate reader (Switzerland). Dose-response curves (semi-log) were generated in GraphPad PRISM v10.5.0 (GraphPad Software, USA), and non-linear regression was used to calculate the half-maximal effective concentration (EC50), with lower EC50 values indicating higher rhBMP-2 potency. Following EC-MAX identification, we performed a 10 × 10 batch comparison. Ten independent batches of EC-MAX were tested against ten batches of CHO-rhBMP-2 (Infuse) and the World Health Organization (WHO) international rhBMP-2 standard National Institute for Biological Standards and Control (NIBSC) 93/574. Both W20-17 and C2C12 cell lines were used with ALP dose-response assays run in triplicate for each batch. Relative potency was calculated by normalizing to an internal standard (single Infuse lot in C2C12 assay) or to the NIBSC standard (W20-17 assay). Outliers were identified and removed using robust regression and outlier removal (ROUT) method (Q = 1%) in GraphPad Prism. In vitro structural and physicochemical comparability assessment To determine structural and physicochemical properties between EC-MAX and CHO BMP-2, a structured comparability framework was applied as shown in Table II. Table II.Testing for structural and physiochemical properties between the most potent Escherichia coli and Chinese hamster ovary preparations. The results are shown in the Supplementary Material.Framework componentProperty comparatorBackbone identitySequenceChain-length distributionHigher-order structureDisulfide patternCircular dichroismNuclear magnetic resonancePost-translational modifications and variantsDeamidationOxidationGlycosylationPyroglutaminationMisfolded disulfidesQuantity and extinctionConcentrationε280 Testing for structural and physiochemical properties between the most potent Escherichia coli and Chinese hamster ovary preparations. The results are shown in the Supplementary Material. In vivo ectopic bone formation study We used ASTM F2529-13 (2021) as an accepted method for in vivo osteoinductive potential evaluation. This athymic nude rat model evaluates bone induction without immune response interference to human proteins. Sample size was determined from a previous rat study showing adequate power with n = 7 to 8 per group.19 A total of 17 male Hsd:RH-Foxn1rnu rats (~9 weeks old, 280 to 310 g; Envigo; Inotiv, USA) were obtained; two died during anaesthesia induction prior to incision, resulting in 15 treated animals (seven rats per group received bilateral implants of one type; one rat received one implant of each type). All procedures were approved by the institutional animal care and use committee and French Ministry for Education and Research (APAFIS#1437). Rats were acclimatized for one week, housed in individually ventilated cages (two per cage) at controlled temperature and humidity with a 12-hour light/dark cycle, given food and water ad libitum. Animals were monitored daily for general health, with clinical signs and body weight assessed twice weekly. Randomization was used to allocate the animals to each group. Implants consisted of ACS loaded with rhBMP-2. Collagen sponges (Infuse kit, bovine type I collagen, Lots: 7510100/M111701AAP and 7510050/M11806AAF) were ~0.175 cm³ (~1.25 × 0.7 × 0.2 cm) for consistency. Prior to surgery, each sponge was aseptically soaked with 100 µl of assigned rhBMP-2 solution (170 µg/ml), delivering 17 µg rhBMP-2 per implant, with ~30-minute soak time. The 17 µg dose selected was within the range based on established literature demonstrating reliable ectopic bone induction in rat.8,20 Infuse BMP-2 served as CHO-derived rhBMP-2, and EC-MAX as the E. coli-derived comparator. Surgeries were performed under general inhalation anaesthesia (isoflurane in air, 5% induction, 2% to 3% maintenance, 0.5 L/min flow). Pre- and postoperative analgesia was provided via subcutaneous buprenorphine (0.05 mg/kg, ≥ 30 minutes before incision and six to eight hours post-dose). For each implant, a longitudinal skin incision was made on the lateral thigh, parallel to the femur. Using blunt dissection, a pocket was created in the intramuscular plane between biceps femoris and gluteus superficialis muscles. A single rhBMP-2-loaded collagen sponge was inserted into each muscle pocket. Muscle fascia and subcutaneous layers were closed with sutures; skin was closed with surgical staples. All animals recovered uneventfully post implantation, with no complications, wound issues, or abnormal behaviours observed. Two assessors blinded to treatment groups (CM, JvB) evaluated mineralized tissue formation over time using digital radiographs (Faxitron MX-20; Hologic, USA) at Day 0 (post-implantation) and Day 28 (pre-euthanasia). At Day 28, radiographs were qualitatively examined for new bone formation. After four weeks, rats were euthanized, and implant sites with surrounding muscle and newly formed bone were harvested en bloc for microCT analysis. MicroCT scanning was performed using a Bruker SkyScan 1076 scanner (Bruker, USA). Scans were acquired at 70 kV with 0.6° rotation step and ~9 µm voxel size. One frame per rotation position was averaged. Raw projection data were reconstructed into cross-sectional images using NRecon software (Bruker), yielding 2D binary cross-sections of mineralized tissue. The 3D datasets were reoriented along the induced bone axis using DataViewer software (Extron, USA). Regions of interest corresponding to each implant area were defined manually. Threshold values for mineral tissue were 20 to 255. Within each region, standard morphometric parameters were quantified including bone volume (BV), tissue volume (TV), bone volume fraction (BV/TV%), and derived ratios (BS/BV, BS/TV). All measurements were calculated using CTAn image analysis software (Bruker). Implant site tissues were fixed in 10% neutral-buffered formalin for 14 to 28 hours (room temperature) and decalcified in 15% ethylene diamine tetraacetic acid (EDTA)/phosphate-buffered saline (PBS) for 40 days with gentle agitation. Decalcification was monitored by radiography (Faxitron MX20, Edimex). Samples were rinsed, dehydrated, and cleared using an automated processor (ASP6025; Leica, Germany) through graded ethanol, xylene, and molten paraffin. Tissues were embedded (Histostar; Thermo Fisher Scientific, USA) and sectioned at 4 μm (Histocore Autocut, Leica). Sections were floated on a 48°C water bath, mounted on Superfrost Plus slides, and dried overnight at 60°C. For staining, sections were deparaffinized in xylene and rehydrated through ethanol. For haematoxylin-eosin-saffron (HES) staining, sections were stained with haematoxylin, blued, stained with eosin, stained with saffron, dehydrated through ethanol, and cleared in xylene. For Masson’s trichrome staining, sections were stained with Weigert’s iron haematoxylin, treated with acid/70% ethanol, stained with fuchsin ponceau, treated with phosphomolybdic acid, stained with aniline blue, treated with 1% acetic acid, and then dried. Slides were coverslipped (Histocore Spectra CV mounting medium) and digitized at 20× using a Nanozoomer S60 (Hamamatsu, Japan). Images were exported as .ndpi and reviewed in NDP.view2 (v2.9.29) by a single blinded assessor (CM). Statistical analysis Quantitative data are reported as mean (SD). For the in vitro ALP assays, optical density (OD) values were normalized by subtracting the mean OD of media-only control wells. Semi-log dose–response curves were generated in GraphPad Prism v10.5.0 (GraphPad Software, USA), and EC50 values were calculated by non-linear regression. Relative potency was calculated from EC50 values by normalization to an internal Infuse reference lot for the C2C12 assay and to the WHO international rhBMP-2 standard, NIBSC 93/574, for the W20-17 assay. Outliers were identified and removed using the robust regression and outlier removal method, with Q = 1%. Comparisons of in vitro relative potency between EC-MAX and CHO-rhBMP-2 were performed in GraphPad Prism v10.5.0 using independent-samples two-sided t-tests assuming equal variances. For quantitative micro-CT outcomes, measurements of zero were retained in the analysis where no mineralized bone was detected at an implant site. Tissue volume, bone volume, tissue surface, bone surface, and intersection surface were compared between groups using independent-samples t-tests, whereas bone volume fraction, bone surface-to-volume ratio, and bone surface density were compared using Mann-Whitney U tests. Values of p < 0.05 were considered statistically significant. Results In vitro relative potency assays In initial screening of four EC-derived rhBMP-2 candidates, Locate Bio consistently showed the highest bioactivity (data not shown). This variant was labelled EC-MAX and used for extensive testing against CHO-derived rhBMP-2. In the subsequent comparison, ten independent lots of EC-MAX were evaluated against ten lots of CHO-rhBMP-2 in ALP assays using W20-17 and C2C12 cells. In the W20-17 assay, two low-potency CHO outliers were identified using predefined criteria (GraphPad Prism Robust Regression and Outlier Removal, Q = 1%) and excluded. EC-MAX demonstrated markedly greater potency in vitro versus CHO product. In W20-17 stromal cells, EC50 values for EC-MAX batches ranged from 0.0137 to 0.0967 µg/ml, whereas CHO-rhBMP-2 batches required higher concentrations (EC50 range 0.0443 to 0.2550 µg/ml) to achieve the same effect. EC-MAX’s EC50 in W20-17 cells was roughly half that of CHO (0.039 vs 0.087 µg/ml), indicating ~ 2.2 fold higher potency. Relative to the WHO/NIBSC rhBMP-2 international standard, defined as 100% relative potency, CHO-rhBMP-2 showed 93% relative ALP-inducing activity, whereas EC-MAX showed 232% relative ALP-inducing activity (p < 0.001). EC50 and relative potencies are shown in Supplementary Figure S1(a) and S1(b) respectively. In the C2C12 myoblast assay, a similar trend was observed. One CHO outlier (low potency) was identified (GraphPad Prism Rout method, Q = 1%) and excluded. EC-MAX batches showed EC50 values of 0.1630 to 0.4307 µg/ml, consistently lower than CHO-rhBMP-2 (0.2593 to 0.4617 µg/ml). Mean EC50 in C2C12 for EC-MAX was 0.266 µg/ml (SD 0.067) versus 0.330 µg/ml (SD 0.042) for CHO. EC-MAX exhibited roughly 1.3× higher potency. When normalized to the internal Infuse reference lot, defined as 100% relative potency, CHO-rhBMP-2 showed 97.4% ± 8.6% relative ALP-inducing activity, whereas EC-MAX showed 131.8% ± 24.5% as relative ALP-inducing activity (p < 0.001). EC50 and relative potencies are shown in Supplementary Figure S2(a) and S2(b), respectively. Overall, EC-MAX was more potent than CHO-rhBMP-2 in stimulating osteogenesis in both cell assays. In vitro structural and physicochemical comparability assessment EC-MAX and CHO-derived rhBMP-2 share complete sequence identity in the C-terminal 105-amino acid functional domain, including all conserved cysteines and receptor-binding interface as confirmed by peptide mapping analysis (Supplementary Figures S11 to S13) and sequence coverage analysis (Supplementary Figures S12 to S14). The key difference lies in N-terminal length: EC-MAX is a single 108-amino acid form (MW 12,131 Da, non-glycosylated), while CHO-derived material comprises two forms – a 131-residue form (MW 14,887 Da) and a 114-residue form (MW 12,904 Da) – present in 72:28 ratio. CHO-derived material is fully glycosylated with high-mannose structures at Asn⁵⁶ (Supplementary Tables S1 and S2). Intact mass analysis revealed EC-MAX is essentially 100% homodimer with no detectable monomer, whereas CHO material contains ~ 4% inactive monomer (Supplementary Figures S3 to S11), directly reducing effective CHO dose. Disulfide bridge analysis confirmed expected bridges (Cys²-Cys⁶, Cys³-Cys⁷) but revealed that EC-MAX has higher levels of misfolded disulfide species, likely from bacterial refolding challenges (Supplementary Figures S15 and S16). Post-translational modifications differed quantitatively: EC-MAX showed higher deamidation (9.8% vs 4.3% at Asn²⁻⁶ region) and selective N-terminal methionine oxidation (3.0%), while 90% of CHO 1-114 form exhibits N-terminal pyroglutamination (Supplementary Tables S5 and S6, Figures S17 and S18). Higher-order structure: circular dichroism analysis revealed EC-MAX has lower β-sheet content (22 to 25%) versus CHO (27 to 29%), indicating subtly reduced structural order (Supplementary Figures S19-S26). Nuclear magnetic resonance spectroscopy showed that both materials exhibit peak broadening characteristic of BMP family proteins,21 but EC-MAX displays enhanced conformational exchange, manifested as greater line broadening and absence of upfield-shifted peaks visible in CHO spectra (Supplementary Figures S27 and S28). This finding indicates increased intermolecular dynamics, likely from absence of glycosylation-mediated steric shielding. This does not represent stable aggregation, as intact mass analysis confirmed both materials are predominantly dimeric. The functional significance of these structural differences is discussed below. In vivo ectopic bone formation study All 15 treated rats remained healthy over the four-week implantation period, with no adverse reactions to implanted rhBMP-2. By Day 28, new mineralized tissue was evident radiologically in both CHO and EC-MAX groups. An ARRIVE checklist is included in the Supplementary Material to show that the ARRIVE guidelines were adhered to in this study. Radiological assessment found a stark contrast in bone extent formed by EC-MAX versus CHO-rhBMP-2. In the CHO group, radiographs at 28 days typically showed isolated opaque nodules in thigh muscle. In contrast, EC-MAX radiographs revealed much larger radiopacity areas. Radiopaque masses in EC-MAX implants frequently bridged to adjacent femoral cortex. See Supplementary Figure S29 for representative radiological changes between day 0 and day 28. Micro-CT assessment Upon examination, visible morphological differences in new bone were found between the groups. High-resolution micro-CT imaging provided qualitative insight into structural differences. In the CHO group, high-density areas were localized to the implant periphery, with difficulty identifying clear trabecular structure, and lower-density areas centred within the implant. In the EC-MAX group, higher-density material was more uniformly distributed across the implant with apparent open trabecular structure. Representative images are shown in Figures 1 to 6. Micro-CT reconstruction of a Chinese hamster ovary-derived recombinant human bone morphogenetic protein-2 ectopic ossicle at four weeks. The mineralized volume is concentrated at the periphery, forming an incomplete cortical-like shell, with limited internal trabecular organization. Central regions show low mineral density, consistent with early-stage mineral deposition. Scale bar: 1 mm. Micro-CT reconstruction of the most active preparation of Escherichia coli (EC-MAX) recombinant human bone morphogenetic protein-2 ossicle at four weeks. The construct demonstrates robust mineralized tissue deposition. The thick peripheral mineralized shell and complex, well-developed internal trabeculae are indicative of structural organization at this timepoint. Scale bar: 1 mm. Micro-CT reconstruction of a Chinese hamster ovary-derived recombinant human bone morphogenetic protein-2 ossicle at four weeks. The construct demonstrates relatively sparse internal mineralization with a thin, irregular outer rim. Trabecular development is limited, reflecting early maturation. Scale bar: 1 mm. Micro-CT reconstruction of a Chinese hamster ovary-derived recombinant human bone morphogenetic protein-2 ossicle at four weeks. Peripheral mineralization is present but remains discontinuous, with a heterogeneous internal structure and minimal trabecular connectivity. This morphology is characteristic of less mature ectopic ossicles at this timepoint. Scale bar: 1 mm. Micro-CT reconstruction of the most active preparation of Escherichia coli (EC-MAX) recombinant human bone morphogenetic protein-2 ectopic ossicle at four weeks. The ossicle shows extensive, well-connected internal mineralization in a trabecular network. A thickened peripheral shell is evident, suggesting more advanced mineral maturation. Scale bar: 1 mm. Micro-CT reconstruction of the most active preparation of Escherichia coli recombinant human bone morphogenetic protein-2 ossicle at four weeks. A dense and highly interconnected trabecular structure occupies the majority of the ossicle, with substantial peripheral mineralization. The high degree of internal complexity indicates ossicle maturation. Scale bar: 1 mm. Quantitative micro-CT analysis Quantitative micro-CT analysis revealed statistically significant differences in seven of eight morphometric parameters (Table III and Supplementary Figure S30), with EC-MAX exceeding CHO in all metrics except bone volume fraction (BV/TV%). Table III.Data from quantitative micro-CT analysis, presented as mean (SD) (n = 15 per group).Expression systemTV, mm3*BV, mm3*BV/TV, %†TS, mm2*BS, mm2*Intersection surface, mm2*BS/BV ratio, /mm†BS/TV density, /mm†CHO15.02 (8.80)9.22 (5.36)58.53 (18.57)45.86 (22.52)109.85 (75.84)2.65 (1.62)10.74 (5.06)6.45 (2.28)EC-MAX203.30 (47.88)88.52 (12.78)45.21 (10.40)298.67 (45.18)1,304.58 (359.33)15.02 (10.57)14.71 (3.18)6.39 (0.66)Statistical significancep < 0.001p < 0.001p < 0.001p < 0.001p < 0.001p < 0.001p = 0.003Not significant*Independent-samples t-test.†Mann-Whitney U test.BS, bone surface; BS/BV, bone surface to volume ratio; BV, bone volume; BV/TV, percent bone volume; CHO, Chinese hamster ovary; EC-Max, most active preparation of Escherichia coli; TS, tissue surface; TV, tissue volume. Data from quantitative micro-CT analysis, presented as mean (SD) (n = 15 per group). Independent-samples t-test. Mann-Whitney U test. BS, bone surface; BS/BV, bone surface to volume ratio; BV, bone volume; BV/TV, percent bone volume; CHO, Chinese hamster ovary; EC-Max, most active preparation of Escherichia coli; TS, tissue surface; TV, tissue volume. CHO implants demonstrated higher mean bone volume fraction (BV/TV: 58.53% (SD 18.57%)) versus EC-MAX (45.21% (SD 10.40%), p < 0.001, Mann-Whitney U test), despite producing less total bone. This reflects drastically different tissue volumes: CHO implants contracted to only 15 mm³ with densely packed bone, whereas EC-MAX implants expanded to 203 mm³ with mature trabecular architecture. BV/TV values in both groups fall within the normal rat cancellous bone range (40% to 61%).19,22,23 Bone surface-to-volume ratio (bone surface area divided by bone volume) was significantly higher in EC-MAX. EC-MAX bones averaged ~ 14.7 mm² of bone surface per mm³ of bone volume versus ~ 10.7 mm²/mm³ in CHO (p = 0.003, Mann-Whitney U test), indicating more surface area for a given bone volume, characteristic of more porous structure. In contrast, bone surface density (bone surface area normalized to total tissue volume) was not significantly different between groups, suggesting groups were comparable when scaling bone surface to entire implant size. There was no correlation between soak time and BV (Supplementary Figure S31). Histological assessment Histological assessment found essentially complete ACS carrier resorption in both groups, indicating successful cellular infiltration and tissue remodelling. Where present, it was characterized by loose, irregular, swirled collagen bundles without organized lamellar or trabecular structure, no lacunae and no mineralization front. Both CHO and EC-MAX contained trabecular bone areas with interconnected trabeculae, osteocyte lacunae embedded in matrix, and osteoblast-lined surfaces. Marrow-like spaces between trabeculae indicated osteogenesis. Histology corroborated the micro-CT findings, which demonstrated greater bone volume in the EC-MAX samples. Histologically, CHO implants showed extensive fibrovascular stroma and immature osteoid, features consistent with early-stage ectopic bone formation. In contrast, EC-MAX implants contained well-developed marrow adipocytes, indicating progression toward fatty (yellow) marrow maturation. Adipocyte-filled marrow spaces are characteristic of physiologically mature trabecular bone that has transitioned away from the hypercellular, osteogenic phase. CHO-derived ossicles exhibited compact trabeculae with narrow marrow spaces and minimal adipocytes – typical of earlier, more cellular marrow – whereas EC-MAX ossicles demonstrated wider trabecular spacing and abundant intratrabecular adipocytes, consistent with more advanced marrow maturation. Representative images are shown in Figures 7 to 10. Chinese hamster ovary (CHO)-derived recombinant human bone morphogenetic protein-2 (rhBMP-2) ectopic ossicle at four weeks (haematoxylin and eosin stain). A peripheral rim of woven trabecular bone (T) encloses a prominent fibrovascular core (FV), consistent with active early-stage ossicle maturation. Intertrabecular spaces contain emerging marrow-like tissue (M) with scattered adipocytes. Overall, the construct demonstrates robust osteoinduction but retention of a hypercellular, fibrovascular centre typical of the early remodelling phase. Scale bar: 500 μm. The most potent preparation of Escherichia coli (EC-MAX) recombinant human bone morphogenetic protein-2 (rhBMP-2) ectopic bone at four weeks (haematoxylin and eosin stain). The ossicle displays uniformly distributed trabecular bone (T) enclosing extensive adipogenic marrow-like tissue (M), with only minimal residual fibrovascular stroma. The trabecular network appears mature and spatially consistent. Scale bar: 2.5 mm. Chinese hamster ovary (CHO)-derived recombinant human bone morphogenetic protein-2 (rhBMP-2) ectopic bone formation at four weeks (haematoxylin and eosin stain). The ossicle comprises a continuous shell of immature trabecular bone (T) surrounding a fibrovascular stroma (FV). Marrow-like regions (M) are present, but remain limited in distribution, representing early transition toward marrow maturation. The predominance of fibrovascular tissue centrally is characteristic of earlier-stage CHO-induced ectopic ossicles. Scale bar: 500 μm. The most potent preparation of Escherichia coli (EC-MAX) recombinant human bone morphogenetic protein-2 ectopic ossicle at four weeks (haematoxylin and eosin stain). A well-developed peripheral trabecular rim (T) surrounds extensive marrow-like tissue (M) with abundant adipocyte differentiation, consistent with marrow maturation. A smaller central fibrovascular region (FV) persists. Scale bar: 2.5 mm. Discussion Our results demonstrate that EC-MAX surpasses CHO-derived rhBMP-2 in both in vitro assays and in vivo bone formation, producing approximately ten-fold greater bone volume with more mature trabecular architecture. These findings, combined with structural characterization, suggest multiple mechanistic hypotheses warranting direct experimental testing. The most striking structural difference between these products is glycosylation status. Crystal structures show that CHO-BMP-2 carries high-mannose N-glycans at Asn⁵⁶, and based on structural analysis of the BMP-2:BMPR complex, we hypothesize that these glycan structures, if positioned near the binding interface, could create steric interference during receptor engagement, reducing binding affinity.11,12,24 This could be tested through: direct receptor binding assays (SPR, ITC) comparing glycosylated vs non-glycosylated BMP-2; site-directed mutagenesis removing N56 glycosylation; and computational modelling of glycan-receptor interactions. If confirmed, this mechanism suggests the ten-fold in vivo potency difference reflects primarily reduced receptor binding efficiency rather than altered pharmacokinetics. Glycosylation may also influence clearance kinetics. High-mannose glycan structures are recognized by mannose-binding lectin (MBL) and CD206 macrophage mannose receptor, facilitating protein sequestration and accelerated clearance in therapeutic antibodies.25-27 We hypothesize CHO-BMP-2 may undergo faster tissue clearance through these innate recognition pathways, even in athymic rats retaining functional innate immunity. This could be tested through: radiolabelled protein pharmacokinetic studies measuring tissue retention and clearance; comparative studies in CD206-knockout mice; and serum half-life measurements. This mechanism would manifest as reduced area-under-curve exposure despite equal initial dosing. Beyond glycosylation, conformational dynamics may contribute to the observed potency differences. NMR spectroscopy revealed EC-MAX exhibits enhanced conformational exchange and peak broadening compared to CHO-BMP-2, consistent with increased intermolecular dynamics.21 Studies demonstrate reversible BMP-2 oligomerization maintains bioactivity,28,29 and our mass spectrometry confirms that both materials remain predominantly dimeric. We hypothesize that increased conformational flexibility at the dimer interface facilitates induced-fit receptor binding through greater conformational sampling. This could be tested through: receptor binding kinetics measuring association/dissociation rates; hydrogen-deuterium exchange mass spectrometry (HDX-MS) mapping conformational dynamics; and molecular dynamics simulations. Enhanced conformational flexibility may also help to explain why the increased misfolded disulfide content observed in EC-MAX did not preclude greater functional bioactivity in the ALP assays and the ectopic bone formation model. Finally, differences in bioactive dimer content may contribute modestly to the potency gap. Mass spectrometry showed EC-MAX is ~ 100% dimeric while CHO-BMP-2 contains ~ 4% monomeric protein incapable of receptor engagement. This 4% difference reduces effective CHO dose proportionally but cannot explain the ten-fold in vivo potency gap, suggesting that additional mechanisms contribute. These findings help reconcile conflicting reports of EC-versus-CHO rhBMP-2 performance in the literature by demonstrating substantial inter-supplier variability.5-9 The ectopic bone model provides mechanistic insights but has clinical translation limitations: non-orthotopic site, athymic rats masking immunogenicity, and accelerated rat bone metabolism versus humans.30 Furthermore, while a full dose response curve was evaluated in vitro, only a single dose of BMP-2 was evaluated in vivo, which limits clinical interpretation, as the observed relative performance may differ at higher or lower doses. These findings underscore that rhBMP-2 potency cannot be assumed equivalent based on protein mass. Each manufacturing process requires independent potency validation, and products cannot be assumed bioequivalent without comparative testing. This study extends reports of variable osteogenic potency between E. coli- and CHO-derived rhBMP-2 products.5-9 Systematic screening revealed considerable inter-supplier potency differences. At least one E. coli-derived product (EC-MAX), when manufactured with appropriate quality control, can exceed the osteoinductive potency of CHO-derived rhBMP-2 in preclinical models. Further studies should assess safety, long-term outcomes, optimal delivery matrices, immunological considerations, and cost-effectiveness. From a regulatory perspective, rhBMP-2 potency cannot be assumed equivalent across products based on protein mass alone. Each manufacturing process, particularly E. coli-derived proteins requiring in vitro refolding, requires independent potency validation. Products cannot be assumed bioequivalent without rigorous comparative testing. Regulatory approval should be contingent on product-specific efficacy and safety demonstration, not extrapolation from approved CHO-derived products. Author contributions C. Matthews: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Supervision, Validation, Writing – review & editing L. Johnson: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Writing – review & editing W. Holden: Data curation, Formal analysis, Investigation, Validation J. von Benecke: Data curation, Formal analysis, Project administration, Validation, Visualization, Writing – original draft, Writing – review & editing Funding statement The author(s) disclose receipt of the following financial or material support for the research, authorship, and/or publication of this article: Locate Bio provided salaries, materials, and other study-related costs. The research was conducted as part of Locate Bio’s internal research and development program. ICMJE COI statement All authors are employees of Locate Bio, a company actively involved in the development of rhBMP-2 products. This commercial interest represents a potential conflict of interest. Locate Bio funded the research, including the subcontracted experimental procedures performed by Atlantic Bone Screen and the authors may therefore have a financial interest in the outcome of the study. The data and conclusions presented in this article were generated as part of Locate Bio’s internal research and development activities. Data sharing The data that support the findings of this study are available from the corresponding author upon reasonable request. Acknowledgements The authors would like to thank Atlantic Bone Screen for performing the in vivo experimental work. Ethical review statement All procedures were approved by the institutional animal care and use committee and the French Ministry for Education and Research (APAFIS#1437). Open access funding Open access publication was funded by Locate Bio Ltd as part of its internal rhBMP-2 research and development activities. No external grant or funding award number applies. Supplementary material Supporting in vitro potency results, structural and physicochemical comparability analyses, and additional in vivo imaging/micro-CT data for recombinant human bone morphogenetic protein-2 derived from Chinese hamster ovary and Escherichia coli. © 2026 Matthews et al. This article is distributed under the terms of the Creative Commons Attribution No Derivatives (CC BY-ND 4.0) licence (https://creativecommons.org/licenses/by-nd/4.0/), which permits the reuse of the work for any purpose, including commercially, provided the original author and source are credited; however, it cannot be distributed to others in any adapted form. Data Availability The data that support the findings of this study are available from the corresponding author upon reasonable request. References Associated Data Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request.