Transplanted bone marrow mononuclear cells and MSCs impart clinical benefit to children with osteogenesis imperfecta through different mechanisms

Bone marrow transplantation (BMT) is an established therapeutic modality for both malignant and nonmalignant disorders of hematopoietic stem cells. After wide recognition that bone marrow also contains progenitors of bone,^1-4  we postulated that BMT should be applicable to the treatment of osteopoietic as well as hematopoietic disorders.⁵  Nilsson et al demonstrated that transplantation of whole bone marrow leads to donor-derived osteopoiesis in mice,⁶  while Pereira et al showed that systemically infused murine mesenchymal stromal cells (MSCs), which are plastic adherent in vitro,⁷  engrafted in bone.³  We showed that BMT in children with osteogenesis imperfecta (OI), a genetic disorder of collagen type I, the major structural protein in bone, leads to donor-derived osteopoiesis and consequent improvement in the microscopic structure of bone⁵  and in the clinical manifestations of OI.⁸  Recently, BMT in a murine model of OI has corroborated our early human studies.⁹  Taken together, these data validate the functional competence of donor-derived osteopoietic cells, providing the necessary proof to move forward with the development of marrow cell-based treatments for disorders of bone.

Despite this progress, the cellular mechanism(s) by which BMT gives rise to robust osteopoietic activity remains unproven. Pereira et al reported that systemically infused murine MSCs engrafted in the bone of a murine model of OI, and generated a small but statistically significant increase in collagen,¹⁰  supporting the prevailing view that BMT-associated donor-derived osteopoiesis was attributable to the engraftment and differentiation of MSCs. Thus, we reasoned that a decrease in the rate of clinical improvement in our OI patients after BMT⁸  might be corrected with a boost of donor-derived MSCs, which in fact led to a second wave of accelerated growth velocity in all 5 evaluable patients.¹¹  This result suggested that MSCs isolated on the basis of their adherence to plastic may provide adequate therapy for patients with OI or other bone disorders. However, the issue is complicated by work showing that so-called nonadherent bone marrow cells (NABMCs) have measurable osteoprogenitor activity,^12-14  raising questions as to the developmental origin of the transplantable marrow osteoprogenitors that give rise to donor-derived osteopoiesis and hence to the marrow population most likely to yield clinical improvement in patients.

Here we show that NABMCs are significantly more robust transplantable osteoprogenitors than MSCs in mice, suggesting NABMC would be effective cell therapy for bone disorders. Translating this laboratory observation to a pilot clinical trial, T cell–depleted marrow mononuclear cells, comprising < 0.01% MSCs, engraft in bone after intravenous infusion and lead to a remarkable acceleration of growth in some OI patients, suggesting vigorous osteoprogenitor activity in humans as predicted by our animal model. Finally, we demonstrate that NABMCs produce their clinical activity by engrafting in bone, differentiating to osteoblasts, and contributing normal collagen to affected bone, whereas MSCs stimulate growth by secreting a soluble factor that indirectly stimulates growth-plate chondrocyte activity.

Bone marrow cells were harvested from C57BL/6 mice (The Jackson Laboratory) or H2K-GFP transgenic mice.¹⁵  Whole bone marrow cells were cultured at a density of 1 × 10⁶/cm² in α minimum essential medium (α-MEM) supplemented with 10% fetal bovine serum (FBS). NABMCs were collected at day 5 by washing with phosphate-buffered saline (PBS). MSCs were established from adherent expanding bone marrow cells. When the adherent cells reached near confluence, they were harvested by treating with 0.25% trypsin solution for 5 minutes at 37°C, and then gently scraping with a cell scraper. Cells were split 1:3 at each passage resulting in a seeding density of approximately 5000 cells/cm². At passage 5, we confirmed with flow cytometry that they lacked the expression of CD45 and other hematopoietic lineage markers (TER119, CD3, B220, CD11b, and Gr-1) whereas they expressed typical MSC markers (CD29, CD49e, CD90, CD105, and Sca-1; supplemental Figure 1, available on the Blood Web site; see the Supplemental Materials link at the top of the online article). We also demonstrated that they retained trilineage (osteogenic, adipogenic, and chondrogenic) differentiation potential (supplemental Figure 1). Finally, there was no evidence of transformation of the MSCs as determined by 2 approaches. First, there was no observable overgrowth of subpopulations of cells in culture which would indicate increased and uncontrolled proliferation of cells reflecting an escape from senescence. Second, MSCs (1 × 10⁶ cells/mouse) were transplanted intradermally onto the back of syngeneic mice and observed for 8 weeks without evidence of localized cell proliferation indicative of tumor formation (data not shown). Whole bone marrow, NABMCs, or MSCs were transplanted into irradiated C57BL/6 with several dosages or into oim/oim mice (B6 background) as previously described.¹⁶  All animal protocols were approved by the Institutional Animal Care and Use Committees of either St Jude Children's Research Hospital or The Children's Hospital of Philadelphia.

Evaluation of GFP⁺ cells to assess donor engraftment was performed as previously described.¹⁶  Briefly, deparaffinized sections were incubated with rabbit anti–green fluorescent protein (GFP) antibody (1:250; Invitrogen) followed by color development with NOVARed (Vector Laboratories). The osteoblasts were identified as the large, cuboidal cells with eccentrically placed nuclei along the endosteal surface as previously described.^16,17  GFP⁺ osteoblasts were enumerated by visual examination of stained slides by two investigators, who were blinded to the experimental conditions. Osteopoietic cell engraftment (chimerism) is the percentage of observed bone cells that express GFP (the marker of donor origin). For each animal, 3 histologic sections (200 cells/section) were evaluated; the mean of the 3 sections was accepted as the osteopoietic cell engraftment of that animal. Immunohistochemical staining of proα2(I) was performed with a rabbit anti–mouse polyclonal anitserum (1:100) specific for the proα2(I) polypeptide, generated by us (C.L.P.),¹⁸  with a goat anti–rabbit biotinylated secondary antibody (1:200; Vector Laboratories). Horseradish peroxidase was visualized with NOVARed (Vector Laboratories).

All photomicrographs were obtained with a digital AxioCam HRc camera (Carl Zeiss MicroImaging LLC) attached to a Zeiss AxioImager.A1 microscope (Carl Zeiss MicroImaging LLC) via a 20×/0.5 NA dry-objective lens at room temperature using AxioVision 4.5 software (Carl Zeiss MicroImaging LLC). All photomicrographs were cropped using Photoshop CS4 (Adobe Systems) and labeled using Illustrator CS4 (Adobe Systems).

Children with severe osteogenesis imperfecta (type III) were enrolled in a clinical trial approved by the Institutional Review Board of St Jude Children's Research Hospital, the Food and Drug Administration, and the National Marrow Donor Program. All subjects were previously enrolled in a clinical trial of BMT as treatment for OI.^5,8  Bone marrow was harvested from the original marrow donors according to standard clinical practice and then depleted of CD3⁺ cells with the CliniMACS device (Miltenyi Biotec) using the negative selection program according to the manufacturer's instructions. The freshly harvested/processed marrow cell product was infused into the children without a chemotherapy preparatory regimen.

Primary mouse epiphyseal chondrocytes were isolated as previously described with minor modifications.¹⁹  Briefly, femoral and humeral heads from neonatal C57BL/6 mice (2-5 days old) were collected using a stereoscopic dissecting microscope and first digested with 0.25% trypsin solution for 20 minutes at 37°C followed by digestion with 86.5 U/mL collagenase type I (Worthington Biochemical Corporation) at 37°C overnight (∼ 16 hours). The freshly isolated primary chondrocytes were directly used for the chondrocyte proliferation assay. As a control to verify that the cells retained characteristics of chondrocytes after isolation as well as after expansion in the proliferation assay, a population of freshly isolated chondrocytes were cultured at a density of 15 000 cells/cm² in Dulbecco modified Eagle medium containing 10% FBS for 10 days. The expanded population of chondrocytes was characterized by Alcian blue staining. The freshly isolated and expanded cells expressed the chondrocyte-specific genes, collagen type II, SRY-box (Sox) 5 and 6, collagen type IX, aggrecan, as determined by RT-PCR analysis (supplemental Figure 1).

To assess chondrocyte proliferation, 2 × 10⁴ freshly isolated chondrocytes were inoculated into each well of a 24-well plate (lower well) followed by inoculation of 2 × 10⁴ MSCs or chondrocytes onto a cell-culture insert (BD Biosciences) which became the upper well in a transwell system. The cells in the lower and upper wells were maintained separately overnight in α-MEM culture medium supplemented with 10% FBS. After approximately 16 hours (overnight), the lower and upper wells were assembled into a single transwell system. After 3 days of cell culture, the relative cell number of chondrocytes was measured using an MTT Cell Proliferation assay (Invitrogen) according to the manufacturer's instructions.

Indirect and/or systemic effect of MSC transplantation on chondrocyte proliferation was evaluated as follows. Freshly isolated mouse chondrocytes were seeded in 96-well plates at 5 × 10³ cells/well (∼ 15 000 cells/cm²) and maintained in DMEM with 10% FBS overnight. Cells were then treated with 1% mouse serum samples collected from MSC (1 × 10⁶) injected or PBS injected control mice for up to 6 days. The time to serum collection and test serum concentration were previously optimized to reveal chondrocyte proliferation in this bioassay (supplemental Figure 2). Relative cell numbers were measured at days 0, 2, 4, and 6 by using the CyQUANT cell proliferation assay kit (Molecular Probes). Fluorescence was measured with a Synergy HT microplate reader (BIO-TEK Instruments Inc) with excitation at 485 nm and emission detection at 530 nm.

Tibias, taken from C57BL/6 mice 7 days after either MSC (1 × 10⁶) or PBS injection were fixed in 4% paraformaldehyde, decalcified in 10% EDTA (pH 7.4) for 7 days at 4°C, and embedded in paraffin. Serial 5-μm-thick sections (40 per tibia) were mounted and every 10th section (4 per tibia) was evaluated using the proliferating cell nuclear antigen (PCNA) staining kit (Invitrogen) with 0.5% methyl green counterstaining. The PCNA-positive cells were enumerated using AxioVision 4.5 software (Carl Zeiss MicroImaging LLC).

Longitudinal body growth of oim/oim mice was evaluated by measuring length between L1and L6 lumbar vertebrae. Oim/oim mice were injected with either 1 × 10⁶ MSC or PBS via tail vein at 4 weeks of age. Four weeks after the injection, the body weight was measured and the lumbar spine was assessed by first obtaining radiographs with a piXarray100 (Bioptics) imager and then the distance between L1 and L6 was measured using AxioVision 4.5 software (Carl Zeiss MicroImaging LLC).

Statistical analyses were performed using the Student t test for comparison of 2 variables and 1-way ANOVA for comparison of multiple variables, unless otherwise indicated. All statistical analyses were run in Prism 4 (GraphPad Software) and a P value of < .05 was considered statistically significant.

We initially predicted that MSCs constitute the principal marrow osteoprogenitors and that osteopoietic engraftment was greater after BMT, compared with infusions of ex vivo–expanded MSCs because of the use of a radiation/chemotherapy preparative regimen, as MSCs may home to irradiated or otherwise damaged tissues.²⁰  To test this hypothesis, we transplanted 1 × 10⁶ GFP⁺ MSCs with 1 × 10⁵ Sca-1⁻ marrow cells into mice (n = 10 per group) after total body irradiation (TBI) of 1125 cGy or 550 cGy or no TBI; the donor-derived osteopoietic cell (osteoblast and osteocyte) engraftment was evaluated by determining the number of GFP-expressing osteopoietic cells as a percentage of the total number of osteopoietic cells in histologic sections. At 3 weeks posttransplantation, all 3 groups of mice demonstrated a similar low level of donor-derived (GFP⁺) osteopoietic cell engraftment (mean ± SEM, 1.55% ± 0.16% vs 1.64% ± 0.13% vs 1.26% ± 0.15%, respectively, P = .19; Figure 1A), suggesting that the conditioning regimen alone does not underlie the substantial osteopoietic engraftment after BMT.

We then considered that osteopoietic engraftment after BMT may be derived mainly from marrow cells in the NABMCs. Thus, we transplanted 2 × 10⁶ GFP⁺ NABMCs or a similar dose of GFP⁺ MSCs, both with 1 × 10⁵ Sca-1⁻ marrow cells, into lethally irradiated (1125 cGy) mice (n = 10 per group). At 3 weeks after transplantation, the mice transplanted with NABMCs demonstrated a mean (± SEM) of 13.3% ± 1.3% donor-derived osteopoietic cell engraftment, compared with only 1.7% ± 0.2% in mice transplanted with MSCs (P < .001; Figure 1B). These results showed that in our murine model, NABMCs are more robust osteoprogenitors than MSCs suggesting that they may provide superior marrow cell therapy for patients with bone disorders.

We next asked whether our observations in a murine model of osteopoietic engraftment after BMT would translate to the clinical setting. We therefore enrolled 5 children with type III OI, who had previously undergone BMT,^5,8  in a trial of boost infusions of marrow mononuclear cells, which are commonly used clinical grafts. Bone marrow, freshly harvested from the original marrow donors, was depleted of CD3⁺ cells with the use of a CliniMACS device. The resultant marrow grafts contained a median of 2.26 × 10⁸ total nucleated cells per kilogram of recipient weight (range, 1.36-5.50 × 10⁸), 6.96 × 10⁶ CD34⁺ cells/kg (range, 3.69-7.48 × 10⁶), and 6.0 × 10³ CD3⁺ cells/kg (range, 0.0-43.0 × 10³). MSCs in the graft were enumerated by 2-color flow cytometry for D7FIB and CD45 expression,²¹  and by determining the adherent cell yield of in vitro culture of an aliquot of the freshly harvested marrow and the processed marrow cell product. The median MSC dose was 1.13 × 10³ cells/kg (range, 0.68-2.75 × 10³), representing an approximate 20-fold reduction of MSCs in the processed marrow graft compared with freshly harvested marrow because of removal by the precolumn filter and/or possibly by nonspecific adherence to the magnetic beads, and a 1000- to 2000-fold lower dose than is typically used in studies of MSC-based cell therapy.²²  The measured cell constituency of the grafts, then, revealed < 0.01% MSCs (> 99.99% NABMCs) among the total nucleated cells. As anticipated, there were no adverse events associated with the cell infusions.

A male patient, who had a female marrow donor, underwent orthopedic surgery of his arm (for clinical indications), at 2 months after the cell infusion. Fluorescence in situ hybridization for X and Y chromosomes of culture-expanded osteoblasts obtained from a bone sample revealed 0.8% donor osteoblasts (supplemental Figure 3A). Flow cytometric analysis of the cultured osteoblasts demonstrated < 0.05% hematopoietic cell contamination (supplemental Figure 3B), indicating that the observed female cells represent osteopoietic differentiation of donor cells. Given that the osteopoietic graft does not seem to be long-lived²³  and more than 5 years have elapsed since he underwent BMT, we postulate that the source of these donor-derived osteopoietic cells was the marrow mononuclear cell boost received on this clinical trial.

The median growth velocity measured from the 5 patients more than 6 months before the marrow cell infusion was 0.14 cm/month (range, 0.1-0.38 cm/month; Figure 2), which was more than 4 standard deviations below mean for age- and sex-matched healthy children²⁴  consistent with the recognized growth deficiency of severely affected OI children.²⁵  Three of the 5 patients accelerated their growth velocity to a median of 1.0 cm/month, (range, 0.97-1.1 cm/month) over the first 3 months after infusion, which is > 2 standard deviations above the mean (supplemental Table 2). The 2 patients who lacked a response grew at a median of only 0.085 cm/month (range, 0.07-0.10 cm/month). There was no difference in the pretreatment growth velocities of responders versus nonresponders. The beneficial effects of NABMC transplantation were transient, with growth velocities returning to pretreatment levels by 6 months.

The outcome of this clinical trial was especially noteworthy since unlike growth after BMT^5,8  or MSC infusions,¹¹  only a subset of children responded to this cell therapy. To identify the factors that may have contributed to the observed effects on growth velocity, we attempted to correlate responsiveness to the NABMC graft with age, sex, time from BMT, total nucleated cell dose, CD34⁺ cell dose, MSC dose and hematopoietic chimerism (supplemental Figure 4). Surprisingly, the only factor that could be related to a positive response to NABMC transplantation was mixed hematopoietic chimerism as the 2 nonresponders were complete (100%) and the 3 responders were mixed (< 90%) donor chimeras (supplemental Figure 4). Because engraftment of transplantable osteoprogenitors is saturable,¹⁶  we reasoned that this result might reflect the extent of occupancy of stem cell niches by donor osteoprogenitors.

To test this notion, we transplanted 2 groups (n = 10 each) of lethally irradiated FVB/N mice (H-2k^q) with a CD3⁺ cell-depleted bone marrow graft consisting of 2 × 10⁵ cells from a GFP-transgenic C57BL/6 donor (H2-k^b) and 1.8 × 10⁶ cells from either an FVB/N donor or a C57BL/6 donor. The resultant 2 cohorts represent either 10% donor chimeras or complete donor chimeras. At 3 weeks after BMT, 5 mice in each group were infused with a marrow boost of 5 × 10⁶ CD3⁺ cell-depleted bone marrow cells from a GFP-transgenic C57BL/6 mouse. After 5 additional weeks (8 weeks from BMT), all mice were analyzed for donor-derived osteopoietic cell engraftment. As expected, the nonboosted control groups demonstrated similar levels of engraftment (mean ± SEM, 4.8% ± 1.2% vs 5.9% ± 1.4%, P = .57; supplemental Figure 5). The 2 boosted groups also showed similar engraftment (10.2% ± 2.1% vs 10.9% ± 0.4%, P = .75; supplemental Figure 5), indicating that the degree of hematopoietic chimerism, and by inference, the degree of donor-HSC chimerism, is not a contributing factor in whether or not a boost of donor cells engrafts in bone. Importantly, these data also indicate that post-BMT infusion of marrow cell boosts without a conditioning regimen in mice, which was the clinical protocol, leads to approximately 5% donor-derived osteopoietic cell engraftment, which appears to be clinically significant in patients (Figure 2), but less than the 15%-20% donor-derived osteopoietic engraftment observed with marrow cell infusions after a radiation conditioning in mice.¹² 

While all children accelerated their growth after BMT,⁵  some children responded slowly without measureable acceleration in the first 3 months, while others rapidly accelerated their growth. Interestingly, the slow responders after BMT were nonresponders after marrow cell boost, while the BMT rapid responders all responded to the boost as well (Table 1). Given that the specific collagen mutation affects the severity of the OI phenotype by influencing the intermolecular interactions of collagen^26,27  and each of our patients is affected by a different mutation,^5,8  our data suggest the collagen mutation may be a principal determinant of the response to NABMC therapy. In contrast, all children respond to MSC therapy (supplemental Table 1),¹¹  suggesting that these 2 cell populations exert their growth effects by different mechanisms.

Noting that NABMCs are more robust osteoprogenitors than MSCs, we predicted that they likely exert their biologic effect in OI by engrafting in the bone microenvironment and differentiating to osteoblasts. To test this idea, we transplanted 2 × 10⁶ NABMCs from a GFP-transgenic donor into homozygous oim/oim mice, a model of OI in which the animals lack expression of the proα2 protein subunit of collagen I.²⁸  Immunohistochemical staining for GFP of femur sections obtained 4 weeks after transplantation revealed 15.2% ± 1.9% donor-derived osteopoietic cell engraftment (n = 5 mice). Immunohistochemical staining using proα2-specfic antiserum unambiguously demonstrated a contribution of proα2 protein to the bone matrix, which must have been produced by the donor-derived osteopoietic cells (Figure 3 bottom left panel). Although not strictly quantitative, proα2 protein was not definitively detected in bone sections with modest levels of donor-derived osteopoiesis suggesting a low, but clearly measurable, level of proα2 in the bone matrix of the mice with the greater levels of donor osteopoietic cell engraftment (Figure 3). By contrast, mice (n = 5) transplanted with GFP-transgenic MSCs demonstrated only 0.9% ± 0.2% donor-derived osteopoietic cell engraftment, and there was no evidence of proα2 protein expression (Figure 3 bottom right panel). These findings suggest that the beneficial effects of transplanted NABMCs, but not MSCs, are mediated by the contribution, to the bone matrix, of normal collagen produced by osteoblasts that differentiated from donor NABMCs, consistent with our view that the response to NABMC therapy is collagen-dependent.

Despite their ability to induce marked growth acceleration, ex vivo–expanded MSCs do not seem to engraft appreciably in bone in children with OI.¹¹  Similarly, intravenously infused MSCs do not engraft readily in murine bone (Figure 1 and Dominici et al¹² ) or generate measurable collagen deposits in our oim/oim murine model (Figure 3). Because linear growth in long bones occurs through the proliferation of chondrocytes in the growth plate,²⁹  we first considered that MSCs may be engrafting specifically at this site. MSCs derived from the GFP-transgenic C57BL/6 mouse were injected intravenously into wild-type mice. Four weeks after the injection, immunohistochemical staining for GFP was performed on the bone sections; however, there were no GFP-positive chondrocytes in the growth plate, indicating donor cells did not engraft in the growth plate after the injection (Figure 4A). In the absence of demonstrable engraftment, we reasoned that MSCs may be stimulating growth through the release of soluble mediators that act directly on chondrocytes.³⁰  However, when freshly isolated murine chondrocytes are cocultured with MSCs, chondrocytes, or medium alone (control) in a transwell system to prevent cell-to-cell contact, their growth profiles were essentially the same (Figure 4B).

We next considered that the MSCs may activate a biochemical pathway, leading to chondrocyte growth stimulation. To pursue this notion, we infused 1 × 10⁶ MSCs or phosphate-buffered saline (PBS) into mice (n = 5 per group) and collected serum from the recipients on day 2 or 7 after cell infusion. Over a 6-day interval, chondrocytes in tissue culture proliferate only minimally when maintained in standard chondrocyte medium without supplementation, or with the addition of fresh MSC medium or MSC-conditioned medium (Figure 4C). When supplemented with 1% (v/v) serum obtained from mice at 2 or 7 days after PBS infusion, or 2 days after MSC infusion, chondrocyte proliferation was greater than in medium without supplementation; however, serum from MSC-infused mice did not generate a greater response than control serum. Strikingly, supplementation with serum taken from mice at 7 days after MSC infusion stimulated significantly greater proliferation of chondrocytes (P < .001), indicating the activity of a soluble mediator in 7-day, but not 2-day, serum (Figure 4C).

These findings suggest that MSCs induce the production of a soluble mediator from a second tissue that activates the proliferation of chondrocytes in the growth plate, but does not address whether MSCs secrete a mediator which acts at a distant second tissue, or direct cell-to-cell contact is required. To distinguish between these possibilities, we infused 300 μL of MSC-conditioned medium (4 days of culture with confluent MSCs in a 15-cm dish with 10 mL of medium), a similar volume of fresh MSC medium, or 1 × 10⁶ MSCs into mice (n = 9 per group). After 7 days, serum obtained from mice infused with conditioned medium was found to stimulate chondrocyte proliferation in vitro more readily than serum from control animals (P = .015) and similar to serum from mice infused with MSCs (P = NS; Figure 4D), indicating the activity of MSCs is because of a soluble mediator released from the cells which may act at a remote site.

To establish the physiologic significance of this MSC effect, we assessed the expression of PCNA in the intact growth plate in response to MSC treatment. The femoral growth plate of mice (n = 12/ group) killed 7 days after MSC infusion revealed 65.4% ± 1.4% PCNA-expressing cells compared with 52.5% ± 2.9% in control animals infused with PBS (mean ± SEM, P < .001; Figure 5A). The observed activity was most prominent in the proliferating zone of the growth plate in agreement with findings during physiologic growth.²⁹  Our data indicate that MSCs stimulate the secretion of a soluble mediator from a second tissue that leads to growth-plate cell proliferation.

To investigate whether this mechanism may be operational in OI, underlying the growth observed in patients with OI, we infused MSCs (10⁶ cell/mouse) or PBS (controls) into 4-week-old oim/oim mice (n = 7/group). Computer-assisted measurement of the overall length of the lumbar vertebrae 4 weeks after infusion (n = 4/group) showed that the treated mice had a significantly greater length than controls (1.79 ± 0.02 cm vs 1.70 ± 0.02 cm, mean ± SEM, P < .001; Figure 5B). Moreover, the ratio of the body weight 4 weeks after infusion to the preinfusion weight was assessed in both groups to determine the general growth-promoting effect of MSCs. The treated group had a 54.1% ± 3.8% increase in body weight, compared with only 34.9% ± 3.3% among controls (mean ± SEM, P = .003; Figure 5C) over the 4 weeks of observation.

Recognizing that NABMC therapy in children with OI stimulates growth throughout the skeleton, not solely at sites of donor cell engraftment, we considered whether these cells may also be associated with the release of a soluble growth-promoting mediator. To examine this idea, we infused 2 × 10⁶ NABMCs into mice and serum was collected 7 days after the infusion. The serum was assessed in our chondrocyte proliferation in parallel with serum from mice injected with PBS or MSCs. The assay demonstrated that the serum from NABMC-infused mice stimulated chondrocyte proliferation significantly greater than the serum from PBS-infused mice; however, it was significantly less than the serum from MSC-infused mice (Figure 6), suggesting a potential supplementary mechanism of NABMC-stimulated growth.

In our original study of BMT as therapy for children with severe OI,⁵  it was potentially questionable whether the observed clinical benefits were produced exclusively by the infused bone marrow cells or perhaps partly by the conditioning regimen or other peritransplantation drugs. Here, bone marrow MNCs infused in the absence of radiochemotherapy conditioning or post–cell-infusion medicines were shown to induce striking accelerations of growth in a subset of children with OI, unambiguously demonstrating the therapeutic capacity of this cell population in OI.

Our working model of the cellular basis for this therapeutic activity predicts that transplanted donor cells with osteopoietic potential engraft in bone and differentiate to osteoblasts and osteocytes, which then contribute normal collagen to the OI bone matrix. The importance of this mechanism is underscored by the notion that the defect in OI quality of bone is fundamental to the pathogenesis of OI.³¹  Indeed, the severity of the autosomal-dominant OI phenotype is determined by the specific collagen amino acid substitution as well as the relative abundance of normal and mutated collagen.^26,27,32  When donor-derived cells engraft in bone and contribute normal collagen to the bone matrix, the ratio of normal to mutated collagen increases. Histomorphometric analysis of OI-affected bone after BMT suggests that this ostensible increase of normal collagen is associated with an improvement in the quality of bone.⁵  Thus, we postulate that the response to therapy with NABMC should be, in part, dependent on the specific collagen mutation, just as the clinical phenotype is, in part, dependent on the mutation. This theory is compatible with our data in which we observed both responders and nonresponders among the 5 children with different collagen mutations (Figure 2). Direct support for this idea is provided by biochemical studies of collagen I derived from patient 1 which suggested that this mutation would render the child poorly responsive to increasing the relative amount of normal collagen in his OI bone matrix.³³  Consistent with this prediction and our theory that NABMC responsiveness is collagen mutation-dependent, this patient was a nonresponder in this NAMBC trial and a slow responder after BMT (Table 1), but he vigorously responded, as did all of the patients, to MSC therapy (supplemental Table 1).

MSCs have also been shown to be effective cell therapy for children with OI.¹¹  This observation was puzzling as convincing data to demonstrate a preferential localization (eg, homing) of intravenously infused MSCs into bone are lacking. Moreover, we were unable to show that MSCs contribute collagen to the bone matrix (Figure 3), a result that stands in marked contrast to early concepts predicting that MSCs differentiate to osteoblasts.^34,35  To address this issue, we conducted experiments showing that MSCs stimulate bone growth by secreting a soluble mediator(s) that initiates a biochemical pathway, involving 2 (and possibly more) intermediaries, ultimately resulting in growth-plate chondrocyte proliferation leading to bone elongation. Because the putative biochemical pathway targets chondrocytes, which do not substantially express collagen type I and therefore are unaffected by the OI mutation, the mechanism of action should be collagen mutation independent, a prediction supported by our clinical data.¹¹ 

In addition to osteoblast differentiation and collagen production, NABMC appear to also release a soluble mediator into the serum of NABMC-infused mice that can promote chondrocyte proliferation in our in vitro assay. While this result could be because of MSC contamination of the NABMC, we deem this explanation unlikely given the significant chondrocyte proliferation and extremely low potential MSC contamination within the NAMBC fraction. The soluble mediator apparent in the NAMBC may be an intermediary in the biochemical pathway stimulated by the MSC-associated factor, which leads to chondrocyte proliferation. This idea is especially intriguing because it invokes marrow hematopoietic cells as cellular components of the MSC-associated growth-promoting pathway. Whether this soluble factor contributes to the effects of NABMC in children with OI (Figure 2), in addition to the engraftment and differentiation of primitive cells, is currently unknown.

If NABMC-based therapy is truly mutation-dependent, why did all OI children who underwent BMT show a clinical benefit? We favor the notion that the potent transplantable osteoprogenitor residing within the NABMC population engrafts at discrete saturable niches in the bone marrow space.²³  Our current data suggest that a small fraction of these niches are available for engraftment under homeostatic conditions, as has been reported for the hematopoietic stem cell niche,³⁶  but marrow ablation may make substantially more niches available, permitting a larger number of osteoprogenitors to engraft, consistent with the outcome of OI children after BMT.^5,8  Thus, high levels of NABMC engraftment can generate clinical benefits even in the setting of unfavorable collagen mutations.

In summary, we have shown that both NABMC and MSCs are clinically effective agents for cell therapy of bone but function by entirely distinct mechanisms, raising important questions as to their optimal use in patients. The relatively short duration of maximal clinical effect with either graft remains a challenge. Repeated infusions of MSCs may be the most valuable strategy to stimulate bone growth, especially in light of the inability of growth hormone to uniformly benefit severely affected children.³⁷  Ultimately, the ideal therapeutic strategy may be to develop the capacity to durably engraft donor-derived osteopoietic cells from within the NABMC population to improve the quality of bone with repeated infusions of MSCs used as adjunct growth-promoting therapy.

The authors thank the Human Applications Laboratory at St Jude Children's Research Hospital for processing of the clinical bone marrow cell boosts, and John Gilbert for editorial review of this manuscript.

This work was supported in part by grants from the National Institutes of Health (R01 HL077643), St Jude Children's Research Hospital Cancer Center Support (CORE; grant P30 CA21765), American Lebanse Syrian Associated Charities, Center for Childhood Cancer Research, and Center for Molecular and Cellular Therapeutics at the Research Institute of The Children's Hospital of Philadelphia.

National Institutes of Health

Contribution: S.O. designed, performed, and analyzed research, and assisted with preparation of the manuscript; P.L.G. designed and performed clinical research; K.S., R.J., R.M., T.J.H., and E.V. performed research; C.L.P. designed and analyzed research; M.D. designed, performed, and analyzed research; M.I. conceived, designed, and analyzed research; and E.M.H. oversaw the entire project, conceived, designed, and analyzed laboratory and clinical research, and prepared the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

The current affiliation for P.L.G. is Division of Human Genetics, Growth & Development, Penn State Hershey Children's Hospital, Hershey, PA. The current affiliation for R.J. is Section of Neurology, St Christopher's Hospital for Children, Philadelphia, PA.

Correspondence: Edwin M. Horwitz, MD, PhD, The Children's Hospital of Philadelphia, Colket Translational Research Bldg 3010, 3501 Civic Center Blvd, Philadelphia, PA 19104; e-mail: horwitze@e-mail.chop.edu.