Human perivascular stem cells prevent bone graft resorption in osteoporotic contexts by inhibiting osteoclast formation

Abstract The vascular wall stores mesenchymal progenitor cells which are able to induce bone regeneration, via direct and paracrine mechanisms. Although much is known regarding perivascular cell regulation of osteoblasts, their regulation of osteoclasts, and by extension utility in states of high bone resorption, is not known. Here, human perivascular stem cells (PSCs) were used as a means to prevent autograft resorption in a gonadectomy‐induced osteoporotic spine fusion model. Furthermore, the paracrine regulation by PSCs of osteoclast formation was evaluated, using coculture, conditioned medium, and purified extracellular vesicles. Results showed that PSCs when mixed with autograft bone induce an increase in osteoblast:osteoclast ratio, promote bone matrix formation, and prevent bone graft resorption. The confluence of these factors resulted in high rates of fusion in an ovariectomized rat lumbar spine fusion model. Application of PSCs was superior across metrics to either the use of unpurified, culture‐defined adipose‐derived stromal cells or autograft bone alone. Under coculture conditions, PSCs negatively regulated osteoclast formation and did so via secreted, nonvesicular paracrine factors. Total RNA sequencing identified secreted factors overexpressed by PSCs which may explain their negative regulation of graft resorption. In summary, PSCs reduce osteoclast formation and prevent bone graft resorption in high turnover states such as gonadectomy‐induced osteoporosis.


| INTRODUCTION
As early as the 17th century, the modern technique of bone grafting has been used to repair or replace skeletal tissues. Despite the ability of bone grafts to take successfully, the primary failure of this methodology is via osteoclast mediated resorption, 1 which can be seen in diverse clinical scenarios from spine fusion 2 to cleft palate repair. 3 Bone grafts are particularly prone to resorption in states of high bone turnover, 4 such as chronic kidney disease, 5 rheumatoid arthritis, 6 corticosteroid-induced osteoporosis, and postmenopausal osteoporosis. 7 Indeed, the clinical entity of osteoporosis has a dramatic impact on numerous outcomes after spine fusion procedures, such as pseudoarthrosis, pedicular screws pull-out, rod breakage, cage subsidence, proximal junction kyphosis, adjacent fractures, and persistent pain. [8][9][10][11] Methods to augment bone grafts by minimizing chances of bone graft resorption would be clinically impactful, particular in the context of high bone turnover states.
Here, we describe an effective and clinically translatable approach to augment osteoporotic bone grafts with human perivascular progenitor cells. Briefly, PSCs when mixed with autograft bone induced an increase in osteoblast:osteoclast ratio, induced bone matrix formation, and prevented bone graft resorption. The confluence of these factors resulted in high rates of spine fusion in a gonadectomy-induced osteoporosis model.
Within in vitro experiments, we determined that PSCs negatively regulate osteoclast formation, and do so via nonvesicular paracrine factors.

| Human PSC purification from adipose tissue
PSCs were isolated from human subcutaneous adipose tissue via fluorescence-activated cell sorting (FACS) as previously reported. 24,29,34,35 Human lipoaspirates were obtained under IRB approval at Johns Hopkins University with a waiver of informed consent (protocol number IRB00119905). Fat tissue was stored for less than 48 hours at 4 C before processing. The stromal vascular fraction (SVF) of human lipoaspirate was obtained by enzymatic digestion. 35 The resulting SVF was further processed for cell sorting, using a mixture of the following

| Animals and conditions
Female 12-week-old athymic rats were used (strain code 316, Charles River Laboratories Inc, Wilmington, Massachusetts), which exhibit essentially normal bone repair. [36][37][38] Experimental procedures were consistent with ethical principles for animal research and were approved by Johns Hopkins University IACUC (protocol number RA19M268). Throughout the study, rats were single housed in an IVC system rack using polypropylene cages (37 cm × 25 cm × 24 cm), with 12/12 night/day cycles, 21 C (±2 C) and 50% (±20%) relative humidity. All rats had ad libitum access to complete rat food and filtered water. Animal allocation is described in Supplementary Tables S2 and S3.

| Osteoporosis induction and assessment of bone mass
Animals were ovariectomized through a dorsal bilateral approach to induce osteoporosis. 39,40 To perform ovariectomy (OVX), animals were anesthetized with inhaled isoflurane (3% induction, 2% maintenance) delivered with combined oxygen and nitrous oxide (1:2 ratio)

Significance statement
Perivascular progenitor cells exert positive regulatory effects on osteoblasts to heal bones, yet their potential role in osteoclast regulation is not known. It is observed that human perivascular progenitor cells reduce osteoclast formation, thereby preventing bone graft resorption and yielding better outcomes in a preclinical xenograft model. In the future, perivascular stem cells could be used to augment bone grafts, serving as a pro-anabolic, antiosteoclastic stimulus for better outcomes in orthopaedics.
along with subdermal injection of sustained-release buprenorphine (1.2 mg/kg SC, q72h) and enrofloxacin (5 mg/kg). A 10-mm longitudinal skin incision was made at the costovertebral area bilaterally. The peritoneal cavity was explored, and bilateral ovaries excised. Peritoneum and skin were then closed with 4-0 resorbable sutures (Ethicon Inc, Somerville, New Jersey). Dual-energy x-ray absorptiometry (DXA) based assessment for bone mineral density (BMD) was performed to confirm bone mass loss postoperatively, performed every 4 weeks after OVX, using a UltraFocus Faxitron equipment (Faxitron Bioptics, Tucson, Arizona). BMD was measured considering a lumbar spine region of interest (ROI) encompassing the L3 to L6 vertebral bodies. 41 In addition, body weight was recorded every 4 weeks.

| Osteoporotic bone graft preparation and cell supplementation
A finely minced osteoporotic bone graft was prepared from corticocancellous pelvic and lumbar spinous apophysis bones derived from syngeneic female animals 12 weeks after OVX using a dental bone morcelizer device (G.S. online store, Seattle, Washington) (Figure 1A). 39 A total amount of 0.60 g of bone graft was used per spinal fusion (0.30 g per side) according to previous published studies. 42 In general, one donor provided the bone graft needed for three surgical procedures. Implants were prepared using 0.5 × 10 6 total PSCs or adipose-derived stromal cells (ASCs) per animal (0.25 × 10 6 per fusion spine), adapted from our previous study. 21 Briefly, each bone graft (0.30 g) was placed into an individual well of a 24-well plate and evenly dispersed across the well. Next, 0.25 × 10 6 PSCs in 500 μL DMEM were seeded onto the bone graft and incubated at 37 C for 1 to 2 hours. Cell-bone graft interaction was characterized in vitro focusing on adhesion kinetics and cell viability. 43

| Spinal fusion procedure
Twelve weeks after OVX, posterolateral lumbar spinal fusion was performed. Anesthesia and analgesia were the same as described above.
Spinal fusion surgeries were performed as previously described. [44][45][46] Briefly, a 3-cm midline incision was carried out in order to perform a L4/L5 surgical access. The transverse processes of L4 and L5 were bilaterally exposed by blunt muscle splitting technique. High-resolu-

| Assessment of spinal fusion by manual palpation
At 8 weeks postimplantation, the lumbar spine specimens were taken out en bloc. Manual palpation was performed to evaluate the reduction of motion between the lumbar spines of rats, as previously described. 47 The samples were tested by three blinded observers and scored on a scale of 1 to 5 by flexing or extending the specimens. The scoring criteria were as follows: 1 indicates motion between vertebrae, with no bone mass formation; 2 indicates motion with a unilateral bony mass; 3 indicates motion with bilateral bony masses; 4 indicates no motion between vertebrae, with moderate bilateral bone masses bridging transverse processes; and 5 indicates no motion, with abundant bilateral bone.

| DXA and microcomputed tomography (μCT) assessments of bone graft sites
Cell augmented bone grafts were assessed using a combination of DXA and μCT. First, the BMD of bone graft sites was prospectively analyzed every 4 weeks with DXA using regions of interest between the transverse processes of the L4 and L5 vertebrae. Second, general morphological description and morphometric analysis were performed using ex vivo microCT using a Skyscan 1275 scanner (Bruker-MicroCT, Kontich, Belgium) with the following settings: 55 kV, 181 μA, 1 mm aluminum filter in 180 , six frames per 0.3 with a 20-μm voxel size. Images were reconstructed using NRecon. DataViewer software was used to realign the images and quantitative parameters were assessed using Skyscan CTan software (SkyScan, Kontich, Belgium) as previously published. 21,45 Briefly, polygonal ROIs were outlined including the minced bone graft and the newly formed bone matrix between the L4 and L5 transverse processes. Preexisting bone structures were excluded. A threshold value range of 61 to 255 was used. After global thresholding was carried out, a three-dimensional (3D) data analysis including bone volume, bone volume/tissue volume, bone surface, trabecular thickness, trabecular spacing, and trabecular number was performed.

| Histology and immunohistochemistry
Specimens were fixed in 4% paraformaldehyde for 48 hours, decalcified with 14% ethylenediaminetetraacetic acid for up to 2 months and embedded in the optimum cutting temperature compound. Sections were prepared at 10-μm thickness with a Cryofilm type 3c microtome (SECTION-LAB, Hiroshima, Japan). Sections were stained with hematoxylin and eosin (H&E) and Goldner's trichrome. Tartrate-resistant acid phosphatase (TRAP) and alkaline phosphatase (ALP) were used as functional enzymatic methods using previously published methods. 48 Histomorphometric parameters related to osteoclasts and osteoblasts were measured using the OsteoMeasure system (OsteoMetrics, Atlanta, Georgia) 49

| Osteoclastogenesis assays
In order to assay the paracrine effects of PSCs on osteoclast formation, bone marrow monocyte/macrophage lineage cells (BMMs) were first harvested from 8 to 10 weeks old wild-type male mice by flushing the marrow space of femora and tibiae. The flushed bone marrow cells were cultured in α-minimum essential medium (α-MEM) containing 10% FBS, 1% penicillin/streptomycin, and 20 ng/mL macrophage colony-stimulating factor (M-CSF) (R&D Systems, Minneapolis, Minnesota) for 3 days. BMMs were then incubated in 48-well plates (5 × 10 4 cells per well) with 10 ng/mL M-CSF and 50 ng/mL RANKL (R&D) for 5 days. Next, TRAP staining was performed on cultured osteoclasts using a commercially available kit (Sigma-Aldrich) prior fixation with PFA 4% for 10 minutes. After osteoclast differentiation, TRAP-positive multinucleated cells containing more than three nuclei were identified as osteoclasts. ImageJ was used to quantify total osteoclasts number and area per field of view (×4). 51  was collected and either applied during osteoclast differentiation (at a concentration of 2% or 5% CM), or further processed for EV purification, as previously published. 29 Briefly, EVs were isolated by serial centrifugation at 300g for 10 minutes, 2000g for 30 minutes, 10 000g for 30 minutes, and 120 000g for 4 hours at 4 C. The resulted EV pellet was resuspended in PBS. EV isolates were validated as previously published 29 and in accordance with guidelines set forth by the International Society for Extracellular Vesicles (ISEV) 52 using a combination of size distribution evaluation using nanoparticle tracking analysis (Nanosight), visualization of EVs with transmission electron microscopy, and western blot to confirm enrichment in tetraspanin molecules but without cellular contaminants (CD9, CD63, CD81, calnexin). 29 PSC-EV were applied to BMMs during osteoclast differentiation at concentrations based on our prior reports in other cell types. 29

| Transcriptomics
The RNA content of PSC-EVs and parent PSCs was detected by total RNA sequencing as previously described. 29 Briefly, total RNA

| Statistical analysis
Statistical analysis was performed using an appropriate analysis of variance to analyze more than two groups, followed by a post hoc Tukey's test. A Fisher's exact test was used to analyse categorical variables such as fusion score analysis. The statistical software, GraphPad Prism 8.1 Version (GraphPad Software, San Diego, California) was used for all statistical analyses. *P < .05 and **P < .01 were considered significant.

| Validation of low bone mass after OVX
In order to best mimic bone autografting in osteoporotic conditions, syngeneic athymic female rats were used to prepare the donor bone graft (Figure 1). Corticocancellous bone was harvested from rats 12 weeks after OVX, and bone graft was standardly prepared and applied in a posterolateral spinal fusion procedure to syngeneic animals that were also 12 weeks post-OVX ( Figure 1A). Body weight and lumbar BMD were monitored every 4 weeks after OVX ( Figure 1B-E).
Both donor and recipient ovariectomized animals had a similar starting weight and growth curve ( Figure 1B,C, mean starting weight: 183.2 ± 23.0 and 177.8 ± 24.4 g for donor and recipient rats, respectively). Likewise, mean lumbar BMD was similar between donor and recipient rats at the study outset ( Figure 1D,E, 267.7 ± 42.2 and 262.6 ± 56.3 mg/cm 2 , respectively). As expected, both donor and recipient rats demonstrated a significant decline in lumbar BMD after OVX (mean reduction in lumbar BMD of 45.7 ± 40.0 and 52.5 ± 47.7 mg/cm 2 among donor and recipient animals, respectively).
These findings confirmed the overall similarity of donor and recipient athymic rats, and the achievement of a low bone mass state in all groups after gonadectomy.

| Validation of bone graft supplementation with human PSCs
Our approach was to improve osteoporotic bone graft outcomes by the supplementation of human PSCs. As our prior studies have used polymers 25 or demineralized bone matrices 21,24,53 for cell delivery, we first attempted to optimize the seeding and viability of PSCs when placed on rat corticocancellous bone grafts. PSCs were isolated from fresh human lipoaspirates as previoulsy described, 22 These data suggested that a 2 hours incubation period would be of benefit for PSC-bone graft attachment, but without impairing cell viability.
We next validated that PSCs survive when added to bone graft in the early postoperative period after spinal fusion. PKH-labeled PSCs were seeded on osteoporotic bone grafts, and implanted in our lumbar spine fusion model. After 2 weeks, the bone graft site showed numerous PKH + PSCs ( Figure 2J). Having validated PSC viability, attachment, and survival on corticocancellous bone grafts, we next sought to examine their efficacy in preventing graft resorption.  Figure 3M). Here, bone grafts without cell supplementation showed a 43% reduction in BMD (**P < .01). Likewise, ASC-augmented bone grafts showed a 31% reduction in BMD (*P < .05). In contrast, PSC-augmented bone grafts showed no significant reduction in BMD.

| Human PSCs prevent bone graft resorption and improve posterolateral spine fusion
At the study endpoint, spine fusion was assessed using a previously validated grading system, by manually applying flexion/extension forces to assess intervertebral motion at the L4-L5 level ( Figure 3N). Results showed that PSC-augmented bone grafts yield a fusion rate of 66.7%. This is compared to a fusion rate of 12.5% among rats treated with bone grafts alone, and 25% among rats treated with ASC-augmented bone grafts. In summary, PSC-treated bone grafts showed minimal resorption, which was associated with a significantly higher incidence of spine fusion.

| Human PSCs improve μCT metrics of bone grafts
Cell augmented bone grafts were next assessed using high-resolution μCT imaging at the study endpoint (Figure 4). Three-dimen-

| Human PSCs uncouple osteoblast and osteoclast formation
Histological analyses were next performed on spine fusion sites, which further confirmed morphologic differences associated with F I G U R E 6 Legend on next page. PSC-augmented bone grafting. Goldner's trichrome staining within the control group showed residual bone graft material in a hypocellular fibrous background, with minimal newyl formed woven bone ( Figure 5A,B). In contrast, in PSC-treated groups, incorporation of the bone graft material was more evident, including newly formed woven bone and continuity between bone graft and the native transverse processes. Histochemical staining for ALP showed more intense areas of osteoblastic activity among bone-lining cells within PSC-augmented bone graft sites ( Figure 5C,D). Likewise, increased detection of the osteoblast specific marker OCN was found among PSC-augmented bone grafts ( Figure 5E-H). Osteoclast activity was assessed on TRAP stained sections ( Figure 5I was detected between groups ( Figure 5L).

| Human PSCs inhibit osteoclast formation by nonvesicular paracrine means
The paracrine effects of PSCs on osteoblast formation have been previously documented by our group. 30  Recently, we observed that EVs derived from PSCs retain many of the biologic features of their parent cells, including proosteoblastogenic effects. 29 PSC-derived EVs (PSC-EVs) were next examined for potential direct inhibition of osteoclast formation. As in our prior report, PSC-EVs were derived by ultracentrifugation, and their yield and purity confirmed 29 prior to application, and in accordance with ISEV criteria. 52 In contrast to coculture or CM assays, PSC-EV treatment led to no observable effect on osteoclastogenesis ( Figure 6G-I).
Results thus far suggested that OC inhibitory factors are present within parent PSCs but not PSC-derived EVs. To begin to investigate this phenomenon, a previously derived total RNA sequencing dataset was further interrogated (Figure 6J-M). 29 Here, three separate human PSC isolates were examined in comparison to their respective purified EVs. Transcripts were normalized by fragments per kilobasepair per million mapped (FPKM), and those with Log2 EVs and exosomes are of increasing therapeutic interest in relation to mesenchymal stem/progenitor cells and regenerative medicine. We recently observed that PSC-derived EVs, including exosomes and microvesicles, directly stimulate skeletal progenitor cells. 29 Also observed in this paper, PSC-EVs showed some modest differences in terms of recipient cell type, 29 62 Another interesting finding is enrichment for ApoE expression among PSCs. This gene encodes a protein with a key role in lipid metabolism that can also inhibit osteoclasts through downregulation of c-Fos, NFATc1, and nuclear factor-kappa B signaling. 63 Counterbalancing this, PSC also demonstrated expression of CXCL12, a member of CXC chemokine family, which is a known stimulatory cofactor for osteoclast development and function. 64 There are several limitations to our results. First, our observations regarding a negative regulatory effect on bone graft resorption were found in a postmenopausal (postgonadectomy) osteoporosis model. It will be interesting to determine if different scenarios of low bone mass, such as senile osteoporosis, or steroid-induced osteoporosis would yield a similar outcome with PSC-augmented bone grafting. Conversely, it is not clear if the same beneficial effect would occur in young animals. This is an important consideration, as bone grafting procedures are common in the pediatric population. Second, we specifically analyzed how PSCs induce paracrine effects on osteoclast formation in vitro. However, the identity of the exact nonvesicular secreted factors that negatively regulate OCs, whether one or multiple, has not been precisely defined. We anticipate that further study examining PSC CM using a combination of proteomics and neutralizing antibodies would represent a logical next step in this line of inquiry. Finally, it is not clear if PSCs only influence early OC formation, or if PSCs have negative regulatory effects on OC activity as well.

| CONCLUSION
In summary, PSCs reduce osteoclast formation via nonvesicular paracrine mechanisms and prevent bone graft resorption in high turnover states such as gonadectomy-induced osteoporosis. These data solidify the pleiotropic paracrine effects of PSCs on skeletal cells, and suggest the utility of such cells for cell-augmented bone grafting to prevent surgical failure.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author.