Host-Derived Matrix Metalloproteinase-13 Activity Promotes Multiple Myeloma–Induced Osteolysis and Reduces Overall Survival

Multiple myeloma is an incurable plasma cell malignancy that colonizes the skeleton. Osteoclast-mediated bone destruction is a major hallmark of active disease. The resultant lesions cause great pain, contributing to patient morbidity and mortality. Interaction with the bone stroma can protect multiple myeloma cells from applied therapies, and the associated degradation of bone matrix generates growth factors that further drive the progression of the disease (1). Agents such as bisphosphonates and receptor activator of nuclear kappa B ligand (RANKL) inhibitors effectively reduce osteoclast numbers and protect against skeletal-related disease, but offer little benefit in regards to overall survival (2). These data suggest additional mechanisms facilitate myeloma–bone interaction and their identification could yield novel therapeutic targets for the treatment of the disease.

Mineralized bone matrix is primarily resorbed by osteoclasts. More than 90% of bone is comprised of mineralized type I collagen and therefore enzymes with collagenolytic activity are crucial for appropriate bone remodeling. Cathepsin K is an acidophilic collagenase suited to working in the low pH of the osteoclast lacunae (3–5). However, collagenolytic MMPs, namely MMP1, -2, -8, -13 and membrane-bound MMP-14, -15 are expressed in bone, and ablation of these MMPs in mice has demonstrated skeletal phenotypes, in particular, for MMP-14 (6–11). These phenotypes can be attributed to extracellular matrix (ECM) turnover and the processing of nonmatrix molecules such as growth factors and cytokines. Counterintuitively, mesenchymal stromal cells (MSC) and osteoblasts that are responsible for bone matrix deposition express the majority of these MMPs compared with osteoclasts (12–14). MMPs are also secreted by myeloma cells (15, 16). Interestingly, recent studies have reported a noncatalytic role for myeloma-derived MMP13 in disease progression (15), but the catalytic role for stroma-derived MMP13 remains unexplored. Here, our analyses of human myeloma biopsies demonstrate MMP13 expression in bone-lining osteoblasts and MSCs. This is consistent with literature showing widespread MMP13 expression in the skeleton by cells of MSC lineage, including chondrocytes, osteoblasts, and osteocytes (17–19). In addition, MMP13 appears not to be expressed by osteoclasts, although reports have localized MMP13 in peri-osteoclast cells and in the cement lines of the bone matrix (18, 20).

To date, no studies have examined the contribution of host-derived MMP13 on myeloma progression. This is, in large part, due to the paucity of myeloma models that allow for the ablation of host-derived genes of interest. The spontaneously arising murine myeloma cell line 5TGM1 is syngeneic to the KaLwRiJ substrain of C57BL/6. The cells do not grow in C57BL/6 mice, but we have previously reported that 5TGM1 engraftment in C57BL/6 recombinase activating gene-2 (RAG-2)-null mice is feasible (21, 22). We therefore generated novel RAG-2- and MMP13 double-null C57BL/6 mice that would allow us to address the contribution of host-derived MMP13 to the progression of multiple myeloma. Our findings show that host compartment MMP13 ablation significantly improved the overall survival of myeloma-bearing mice as a result of decreased bone resorption. Surprisingly, this was not due a deficiency in type I collagen turnover but rather differences in the bioavailability of factors, namely CXCL7, that promote osteoclast precursor recruitment and formation. To determine the translatability of MMP13 inhibition as a therapeutic approach, we also demonstrate the efficacy of a highly selective MMP13 inhibitor using an immunocompetent in vivo model. Taken together, our data identify, for the first time, a causal role for host-derived MMP13 catalytic activity in driving the progression of multiple myeloma.

Deidentified human patient specimens were collected through Moffitt Cancer Center's Institutional Review Board-approved Total Cancer Care protocol (MCC14690). All patients involved in this study provided written informed consent in accordance with recognized ethical guidelines as detailed in the Belmont Report. Animal experiments were performed under the University of South Florida-approved Institutional Animal Care and Use Committee (IACUC) protocol, IS0000309, IS0003489, and IS0005900. RAG-2/MMP13 double-null mice were generated by crossing RAG-2-null mice with MMP13–null mice, on a C57BL/6 background. Luciferase-labeled myeloma cells, 5TGM1-Luc (RRID:CVCL_VI66), and U266-Luc (RRID:CVCL_0566) were obtained from University of Texas, Health Science Center at San Antonio, San Antonio, TX (2012; ref. 23) and University of Virginia, VA (2014), respectively. MM1.S were obtained in 2015 from ATCC (catalog no. CRL-2974; RRID:CVCL_8792) and OPM2 was obtained in 2015 from Dr. Kenneth Shain (Moffitt Cancer Center, Tampa, FL; RRID:CVCL_1625). Cells were cultured in RPMI containing 10% FBS and 1% penicillin and streptomycin and used within 30 passages. Cells have recently tested negative for Mycoplasma by PCR in July 2020 (Bulldog Bio, catalog no. 25233), and were additionally authenticated against ATCC, DSMZ, or ExPASy short tandem repeat profiles. Compound 1 [(S)-17b in ref. 25], Compound 2 [(S)-17c in ref. 25] and Compound 4 (52 in ref. 24) were synthesized as described in the respective references (24, 25). For the majority of the studies, unless otherwise explicitly stated, we used Compound 1 as the MMP13 inhibitor and denoted it as MMP13i.

NCBI dataset GSE47552 consists of RNA-sequencing data from CD138⁺ cells isolated from healthy donors and patients with monoclonal gammopathy of undetermined significance (MGUS), smoldering multiple myeloma (SMM), and symptomatic multiple myeloma (n = 99). MMP-1, MMP-8, and MMP13 expressions was extracted and analyzed using the web built-in GEO2R software per NCBI instructions. NCBI dataset GSE46053 consists of transcriptomic sequencing data from healthy donor- and myeloma patient–derived MSCs, which were or were not conditioned using human myeloma (MM1.S) conditioned media (n = 37). Myeloma-induced MMP-1, MMP-8, and MMP13 expressions were extracted and analyzed using the web built-in GEO2R software as per NCBI instructions. Multiple Myeloma Research Foundation (MMRF) IA14 is a record repository for the CoMMpass study, which tracks genomic status throughout myeloma disease progression in newly diagnosed treatment-naïve patients. Analyses of 770 enrolled individuals with RNA-sequencing data paired with their longitudinal clinical data were performed using built-in analytic tools online to compare gene expression with progression-free survival and overall survival.

Nonsequential FFPE, rehydrated tissue sections were rinsed with 1× TBST. Endogenous peroxidases were quenched using methanol peroxide. Antigen retrieval was performed using proteinase K (20 μg/mL) at 25°C for 10 minutes. Prior to staining, chamber slides were fixed with 4% paraformaldehyde. Cells and tissues were blocked at 25°C for 1 hour, and incubated overnight at 4°C in primary antibody diluted in blocking reagent: α-human/mouse MMP13 at 1:200 (Triple Point Biologics, catalog no. RP1-MMP13), α-human CD138 at 1:200 (BD Pharmingen, catalog no. 553712; RRID:AB_394998), and α-mouse IgG2b (Bethyl Laboratories, catalog no. A90–109A; RRID:AB_67157 at 1:200 and A90–109P; RRID:AB_67160 at 1:1,000 for immunofluorescence and Western blot analysis, respectively). Isotype controls were used to assess antibody specificity. Sections were then washed in 1× TBST, and incubated either with species-specific biotinylated or fluorescently labeled (Invitrogen) secondary antibodies at 1:1,000 in blocking buffer for 1 hour at 25°C for IHC and immunofluorescence (IF) staining, respectively. For IHC, biotinylated targets were visualized after 1× TBST washes using an avidin–biotin peroxidase complex and DAB. Sections were counterstained with hematoxylin, dehydrated, and mounted for brightfield microscopy. For IF studies, stained slides were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) and aqueously mounted for wide-field fluorescence microscopy.

MSCs were isolated from both male and female bone marrow flushes and fragmented bone chips harvested from neonatal RAG-2–null/wild-type and MMP13–null mice (26, 27). Tissues were digested in MSC medium [α-minimum essential medium (α-MEM) containing 15% FBS] and 1 mg/mL collagenase for 1 hour at 37°C with shaking at 150 rpm and were cultured in MSC medium for 3 days. Adherent populations were enriched by carefully removing nonadherent cells over time. MSCs were validated in osteogenic assays and expanded for further analysis.

Murine femoral and tibial bone marrow were plated in 10-cm petri dishes in 10 mL α-MEM medium containing 10% FBS to generate primary osteoclasts in vitro. Nonadherent population was replated in fresh dishes with media containing 25 ng/mL of M-CSF (PeproTech, catalog no. 315–02–10) to enrich for myeloid osteoclast precursor population. Myeloid progenitors were subsequently plated at 2–3 × 10⁵/well in 48- or 96-well cell culture plates and treated every 48 hours with osteoclastogenic medium: α-MEM containing 10% FBS, 25 ng/mL macrophage-colony stimulating factor (M-CSF), and 100 ng/mL RANKL (PeproTech, catalog no. 315–11–10) for 4 to 7 days. For treatment studies, MMP13i (Compound 1) was added to osteoclastogenic media (10 nmol/L–10 μmol/L) to assess its effect on differentiation. For MSC conditioned media studies on osteoclast formation, control and CXCL7-immunoprecipitated MSC conditioned media (24 hours in α-MEM with 15% FBS) was collected from wild-type and MMP13–null MSCs and used at 1:1 ratio with regular osteoclast media supplemented with RANKL, but without M-CSF. For treatment studies with recombinant CXCL7 protein, precursor cells were periodically fed with differentiation media containing full-length or MMP13–processed recombinant CXCL7 (Beta Lifesciences, catalog no. BL-0262PS) for up to 5 days. Differentiated cultures were fixed in 4% paraformaldehyde and stored at 4°C in 1× PBS.

Tartrate-resistant acid phosphatase (TRAcP) staining was performed on fixed in vitro differentiated osteoclast cocultures to detect multinucleated osteoclasts using Histostain-TRAcP Kit as per manufacturer's instructions (Invitrogen, catalog no. 85–0199). For in vitro resorption assays, bone marrow myeloid precursor cells were seeded at 6 × 10⁴ cells per 24-well in Osteo-Assay Surface plates (Corning, catalog no. 3988XX1) and cultured as above. Cells were further cultured for three days following osteoclast formation to permit resorption. Osteoclast numbers and resorptions were detected as per manufacturer's instructions thereafter.

MMP13 immunofluorescence in bone marrow stromal cells and osteoclast cultures was performed. After differentiation, cells were fixed in ice-cold methanol for 7 minutes, washed in 1× PBS prior to blocking for 30 minutes at 25°C using 3% milk in 1× PBS. The cells were then washed with 1× PBS, and incubated with primary antibody (α-mouse MMP13, 1:250 dilution in blocking solution; Triple Point Biologics, catalog no. RP1-MMP13) for 1 hour at 25°C. 1× PBS-washed cells were incubated with secondary antibody (Alexa-Fluor 488 α-mouse conjugated antibody diluted 1:1,000 in blocking solution; Invitrogen, catalog no. A32723; RRID:AB_2633275) for 30 minutes at 25°C. Cells were washed again with 1× PBS and mounted in aqueous mounting media containing phalloidin-488 to visualize actin (1:5,000; Invitrogen, catalog no. A12379) and viewed via wide-field fluorescence microscopy.

Total RNA was extracted from cells with TRIzol (Invitrogen) as per manufacturer's instructions. For MMP13 and Cytokine RNA expression, cDNA was generated by a standard reverse transcription reaction, and RT-PCR mixtures were generated using SYBR Green Reagent (Applied Biosystems, catalog no. 4309155) and reactions were performed and quantified using ABI-7900HT instrument and SDS 2.3 software under manufacturer's instructions (Applied Biosystems; RRID:SCR_018060). Refer to Supplementary Table S1 for RT-PCR primer information.

Standard cell lysis protocols were used to isolate total proteins from cell cultures. For immunoprecipitation experiments, wild-type and MMP13–null MSCs were washed with 1× PBS and serum-starved in α-MEM for 1 hour at 37°C prior to incubation with fresh α-MEM for 24 hours to generate conditioned media. Conditioned media were precleared with rProtein G with an hour-long rocking at 4°C and subject to washes and immunoprecipitation using additional rProtein G (15 μL/sample; Invitrogen) and α-mouse CXCL7 antibody at 1 μg/0.5 mg total protein constituted in 1× PBS to 1 mL final volumes (Abcam, catalog no. 231102; RRID:AB_949345). Immunoprecipitation was performed at 4°C overnight on rocker and the resulting media were sterilized using sterile filter and centrifugation for use either in in vitro cultures or protein blotting. Refer to Supplementary Materials and Methods for isolation of bone marrow supernatant.

For immunoblotting, 25 μg of total protein was electrophoresed and transferred to nitrocellulose. Of note, to blot for less abundant proteins, conditioned media samples were concentrated using Vivaspin-6 3000 MWCO spin filters as per manufacturer recommendations (Sartorius, catalog no. VS0691). Transferred blots were blocked for 1 hour at 25°C (1× TBST containing 5% nonfat dairy milk) prior to overnight incubation in primary antibodies: α-mouse MMP13 at 1:1,000 (Triple Point Biologics, catalog no. RP1-MMP13), and α-mouse CCL-2 (Thermo Fisher Scientific; catalog no. MA5–17040; RRID:AB_2538512), α-mouse CXCL7 (R&D Systems, catalog no. AF793; RRID:AB_355606), and α-mouse β-actin (Cell Signaling Technology, catalog no. 3400S) all at 1:1,000. Of note, α-mouse CXCL7 antibody binds to Lys40-Tyr113 and detects active protein. The next day, blots were washed extensively prior to detection with horseradish peroxidase–labeled secondary antibody and ECL using Odyssey Fc Imaging System (LI-COR; RRID:SCR_013430 and RRID:SCR_013715). ELISA was used for the quantification of carboxyterminal telopeptide of type I collagen (ICTP; AVIVA Systems Biology, catalog no. OKEH00680) and CXCL7 (RayBiotech, catalog no. ELM-TCK1–5) in ex vivo isolated specimens and culture media as per manufacturer's instructions.

For cytokine array analysis, 18-hour conditioned media from wild-type or MMP13–null MSCs treated with vehicle or MMP13i were collected in phenol-free α-MEM after an hour-long serum starvation. Cytokine blotting using C2000 Cytokine Array Kit was performed with conditioned media as per kit instructions (RayBiotech, catalog no. AAM-CYT-2000–4; RRID:AB_1547202) and detected using Odyssey Fc Imaging System (LI-COR; RRID:SCR_013430 and RRID:SCR_013715).

For proteomic analysis of MMP13–processed CXCL7, refer to Supplementary Materials and Methods.

For genetic ablation studies, 5TGM1-Luc cells (1 × 10⁶ in 100 μL 1× PBS) were tail vein injected into age-matched 6-week-old mice that were RAG-2–null or RAG-2-MMP13 double-null (n = 20). Multiple myeloma affects men and women equally; therefore, all in vivo studies included male and female mice to remove potential sex disparity as a confounding factor in our observations and analyses. Tumor burden was monitored using bioluminescence imaging with IVIS system. Quantitation was performed by secondary research personnel on deidentified and randomized data in a blinded fashion using the Living Image software (Perkin Elmer; RRID:SCR_018621). Murine whole blood was collected weekly by submandibular phlebotomy and serum levels of IgG2b determined by ELISA analysis according to manufacturer's instructions, also in randomized and blinded methods (Bethyl Laboratories, catalog no. A90–109P; RRID:AB_67160 and starter kit, catalog no. E101). Mice were euthanized upon reaching the clinical endpoint (hind limb paralysis and/or >10% reduction in body weight). Tumor-bearing tibias were excised and fixed in 10% buffered formalin overnight for further histologic analyses.

For pharmacologic ablation studies with MMP13i (Compound 1), age-matched 6-week-old immunocompetent KaLwRiJ mice (n = 30) were inoculated with 5TGM1-Luc by tail vein route (1 × 10⁶ in 100 μL 1× PBS). Both male and female mice were included to account for potential sex disparity as confounding factor. Tumors were allowed 7 days to seed prior to randomization into treatment groups and initiation of daily intraperitoneal injections using vehicle (1× PBS containing 10% DMSO and 10% Tween-80) or MMP13i at 20 mg/kg body weight (diluted at 5 mg/mL in vehicle). Tumor burden was monitored using bioluminescence imaging and quantitated with IVIS Living Image software in blinded methods as described previously (Perkin Elmer; RRID:SCR_018621). Mice were weighed and monitored for toxicity and well-being daily and euthanized upon reaching the clinical endpoint of hindlimb paralysis. Tumor-bearing tibias were excised and fixed for further histologic analyses. All in vivo studies were independently repeated.

Long-bones were scanned at 6-μm increments across 1,000-μm thickness in the metaphysis 500 μm from the growth plate using μ35 instrument for high-resolution μCT analysis (Scanco; RRID:SCR_017119). Trabecular bone volume to total volume ratio (BV:TV) was determined from reconstructed images using manufacturer's software (Scanco; RRID:SCR_017119). Following reconstruction, 3D models of woven bone were built and analyzed by a blinded researcher using consistent thresholding on deidentified bone scan data to assess bone quality. For histomorphometry, bones were subsequently decalcified (changes of 14% EDTA pH 7.4 every two days for 3 weeks) and nonsequential formalin-fixed paraffin-embedded (FFPE) tissue sections were stained with hematoxylin and eosin (H&E). ImageJ (RRID:SCR_003070) was used for trabecular bone measurements, with the area of analysis beginning 500 μm from the growth plate, and extending for 1,000 μm toward the diaphysis. Osteoclast TRAcP staining was performed to as described previously (Invitrogen, catalog no. 85–0199) on FFPE sections, and the osteoclast data were manually calculated from multiple 20× fields of view using brightfield microscopy.

Quantified data are represented as mean with SEM when applicable. Statistical analyses were performed by the Moffitt Biostatistics Core when scaling in vivo studies to ensure robustness, power, and detectable HRs given 5% type I error by a two-sided log-rank test. For statistical analyses of any two treatment groups, Student t test was applied. For statistical analyses of three groups or more, one-way ANOVA was performed. Differences were considered significant if P < 0.05 and noted with asterisks (n.s., not significant).

MMP13 is highly expressed in the cancer–bone microenvironment (28–30). Using publicly available datasets, we initially examined MMP13 expression levels in isolated CD138⁺ bone marrow plasma cells derived from healthy control individuals and patients with varying stages of multiple myeloma (n = 99; GSE:47552). Analyses suggest that, although MMP13 was detected, there does not appear to be a difference at the mRNA level between control and patients diagnosed with MGUS, SMM, or symptomatic multiple myeloma (Fig. 1A). Interestingly, analysis of the MMRF CoMMpass cohort dataset (IA14) showed no MMP13 expression in CD138⁺ myeloma cells in 56% of patients (433 of 770). Of the patients that did express MMP13, albeit at very low levels, myeloma-derived MMP13 did not correlate with progression-free survival but did correlate with overall survival (Supplementary Fig. S1A and S1B).

We performed similar analyses on additional collagenases (MMP-1, -8) expressed by CD138⁺ cells and found MMP-1 is expressed at significantly higher levels in active multiple myeloma but observed no differences in MMP-8 expression (Supplementary Fig. S2A and S2B). We and others have shown that MMP expression is often induced in the surrounding stroma (31). Cancer-derived stimuli including interleukins and PTHrP induce MMP13 expression in bone stroma (19, 29, 32). In keeping with this observation, we identified that MMP-1, -8, and -13 expression was significantly enhanced in primary bone marrow–derived cells incubated with human multiple myeloma cells (MM.1S; Fig. 1B; Supplementary Fig. S2C and S2D; GSE:46053, n = 37; refs. 17, 33). Of these MMPs, MMP13 expression is largely confined to the skeletal tissues making it an attractive therapeutic target as opposed to systemically expressed MMPs. We therefore focused on identifying the cellular sources of MMP13 in human myeloma biopsies (Fig. 1C) and consistently observed positivity in the stromal compartment, specifically in cuboidal bone-lining cells compared with CD138-stained myeloma cells (Fig. 1D). Interestingly, and in keeping with previous reports (18, 34), we detected MMP13 positivity in the bone matrix itself within bone cement lines (Fig. 1C and D).

Given the high level of expression of MMP13 by bone-lining cells in human myeloma specimens and its potential role in processing type I collagen, we hypothesized that stromal MMP13 contributes to myeloma disease progression. Addressing the contribution of stromal genes to multiple myeloma progression in vivo has been challenging, but we addressed this issue by generating RAG-2/MMP13 double-null animals that are receptive to engraftment with the murine multiple myeloma cell line 5TGM1 (22). In keeping with our findings in human tissues, IHC analysis of wild-type and MMP13–null (MMP13^−/−) tissues demonstrated that MMP13 expression was largely confined to bone-lining cells and the cement lines of the bone matrix, while, as expected, MMP13 expression was not detected in MMP13–null animals (Supplementary Fig. S3A; ref. 6). RT-PCR and immunofluorescence consistently demonstrated the presence of MMP13 in wild-type MSCs but not osteoclasts (Supplementary Fig. S3B and S3C). We also observed that 5TGM1 conditioned media (RPMI) significantly increased MMP13 mRNA expression in wild-type MSCs compared with controls (RPMI media alone; Supplementary Fig. S3D) that was in keeping with human analyses (Fig. 1B). In contrast, RT-PCR revealed variable MMP13 expression across mouse and human myeloma cell lines (Supplementary Fig. S4).

Because we observed robust expression of MMP13 in the bone stroma, we next determined whether host-derived MMP13 contributes to myeloma progression in vivo. In three independent studies, wild-type or MMP13–null animals (n = 10/group) were inoculated with luciferase-expressing 5TGM1 cells. Tumor burden was monitored weekly by bioluminescence imaging. Log-scale analysis showed no difference between the groups in regards to tumor growth rate (Fig. 2A and B). Measurement of weekly serum IgG2b concentrations confirmed bioluminescence data (Fig. 2C). Surprisingly, despite no apparent difference in tumor growth rates, overall survival in the MMP13–null multiple myeloma-bearing mice was significantly higher than that of the wild-type group with median survival times of 43 and 39 days, respectively (P = 0.0011; Fig. 2D). Of note, for the 5TGM1 model, this increase in overall survival is in keeping with reports for approved myeloma therapies such as melphalan, bortezomib, and bisphosphonates (35, 36). IHC analysis for IgG2b in ex vivo myeloma-bearing bone tissue confirmed that the multiple myeloma burden in the wild-type and MMP13–null groups was similar (Fig. 2E).

MMP13–null mice have hypertrophic growth plates and delays in endochondral ossification during skeletal development (6, 37), while adult mice have increased trabecular bone volume (6, 38). Using high-resolution μCT, we confirmed increased trabecular bone: total volume (BV:TV) in age-matched (12-week-old) tumor-naïve MMP13–null mice compared with tumor-naïve wild-type controls (Fig. 2F). The age of these control mice was chosen to correspond with the approximate age at which the myeloma-bearing mice reached their clinical endpoint. To determine multiple myeloma BV:TV differences between the groups, we used the ratios obtained from tumor-naïve mice as a means of normalization (Tumor BV:TV). Using this approach, we observed that bone loss was significantly reduced in the MMP13–null mice compared with their wild-type counterparts (Fig. 2G). We also noted significant differences in trabecular bone (Tr.) parameters such as spacing, patterning factor, thickness, and number in MMP13–null myeloma-bearing animals compared with wild type, results that are consistent with reduced bone loss in the MMP13–null myeloma-bearing animals (Fig. 2H–K). Analysis of BV:TV in nonsequential sections from tumor-bearing tibia confirmed μCT results (Fig. 3A and B). To identify the potential causes for reduced bone volume in the MMP13–null mice, we examined MMP13’s ability to process type I collagen. MMP processing of type I collagen into cross-linked ICTP is distinguishable from cathepsin K activity, which generates N- and C-terminal peptides (NTX, CTX; refs. 12, 39, 40). Using an ICTP-specific ELISA, we observed no differences between wild-type versus MMP13–null bone marrow supernatants derived from multiple myeloma–bearing animals (Supplementary Fig. S5). We also examined and found no differences in numbers of osteoclasts/mm of tumor–bone interface or size between the wild-type and MMP13–null groups despite osteoclasts' key role in bone resorption (Fig. 3C–E). Because these analyses were done at study endpoint, temporal differences between the groups may not have been observable. We therefore repeated the study and harvested tumor-bearing tibias at day 21 postinoculation (Supplementary Fig. S6A and S6B). Again, we observed no differences in osteoclast number between the groups (Supplementary Fig. S6C).

Given there was no difference in osteoclast size in vivo at either study mid- or endpoint, we hypothesized that temporal differences in osteoclast formation and/or activity may be responsible for the slower resorption in the MMP13–null myeloma-bearing animals. We therefore conducted in vitro studies to test this hypothesis. Because osteoclasts do not express MMP13 (Supplementary Fig. S3C), we performed osteoclastogenic assays over 7 days using whole bone marrow–derived cocultures from wild-type and MMP13–null mice. Our in vitro data demonstrate significantly more but smaller multinucleated osteoclasts in the MMP13–null cultures (Fig. 4A–C). Furthermore, when normalizing to osteoclast numbers, we also observed MMP13–null osteoclasts were less resorptive than wild-type controls (Fig. 4D and E). Our in vitro data on osteoclast numbers at day 7 contrasted previous reports showing fewer numbers of osteoclasts at day 5 (18, 28). We therefore more closely examined temporal osteoclast formation rates and found that wild-type osteoclast formation peaks one day earlier than MMP13–null osteoclasts, and that relative osteoclast numbers between groups vary significantly depending on assay duration (Fig. 4F and G). Taken together, these results demonstrate that stromal MMP13 is critical for efficient formation of osteoclasts and their activity.

Because we observed no differences in type I collagen turnover between myeloma-bearing wild-type and MMP13–null mice, but did note defects in osteoclast formation rates and function, we next examined whether MMP13 ablation altered production/bioavailability of factors that could influence osteoclast behavior. To this end, we focused on MSCs because we found this population to be a major source of MMP13. We generated conditioned media from wild-type and MMP13–null MSCs and performed cytokine array analysis that identified differences in factors potentially responsible for regulating osteoclast behavior (Fig. 5A). As a control, we included a novel selective inhibitor of MMP13 catalytic activity (MMP13i; ref. 25). Incubation of wild-type MSCs with MMP13i mirrored decreases noted with MMP13–null MSCs. Importantly, these results indicated MMP13 catalytic, rather than noncatalytic activity regulated MSC cytokine/growth factor bioavailability (Supplementary Fig. S7A and S7B). RT-PCR analysis was used to validate cytokine array data at the transcript level (Fig. 5B). Using a candidate approach, we focused on CXCL7 because, in MMP13–null MSCs, it had elevated mRNA levels but reduced active protein levels in the conditioned media (Fig. 5A and B). CXCL7 is proteolytically activated from the latent proplatelet basic protein (PPBP) by MMP-3, and cathepsin G (41–43). Interestingly, we also observed that MMP-3 expression was increased at the transcriptional and protein level in MMP13–null MSCs, but this increase was not sufficient to compensate for MMP13 loss and rescue bioavailability of active CXCL7 (Fig. 5A and B). Western blot analysis corroborates CXCL7 cytokine array results in both MSC lysate and conditioned media (Fig. 5C). Notably, treatment of wild-type MSCs with MMP13i reduced the amount of secreted CXCL7 to that noted with the MMP13–null MSCs, further supporting MMP13 control of CXCL7 levels (Fig. 5D). Proteolytically activated CXCL7 promotes osteoclastogenesis but has not been described as an MMP13 substrate (44, 45). Immunoblotting demonstrated processing of recombinant human CXCL7 full-length peptide upon incubation with recombinant human active MMP13 (Fig. 5E; Supplementary Fig. S8A). Human and mouse CXCL7 share 79.3% homology and therefore, this was verified with murine MMP13 and CXCL7 (Supplementary Fig. S8B). Mass spectrometry identified novel MMP13 direct cleavage sites in both human and murine full-length CXCL7 propeptides (Supplementary Fig. S8A and S8B). Specifically, human CXCL7 was predominantly processed at Glu²⁷, which is one amino acid away from the canonical activating cleavage site observed with Cathepsin G (44, 45). Importantly, MMP13–processed CXCL7 significantly increased osteoclast formation in MMP13–null cultures (Supplementary Fig. S8C). We also noted that the addition of exogenous recombinant MMP13 alone partially rescued CXCL7 levels in MMP13–null whole bone marrow cocultures (Supplementary Fig. S9A and S9B). To further test the importance of MSC-derived CXCL7 in osteoclast formation, we immunodepleted CXCL7 from wild-type MSC conditioned media and observed significantly reduced wild-type osteoclast numbers that were similar to those induced by MMP13–null MSC conditioned media (Fig. 5F and G). These results indicate that MMP13 regulation of CXCL7 bioavailability from MSCs is important for enhancing osteoclast formation.

Recent reports demonstrated that myeloma-derived MMP13 can contribute to osteoclast fusogenesis in a noncatalytic manner. Our data thus far indicate that bone stroma–derived MMP13 catalytic activity is important for regulating the bioavailability of important osteoclastogenic factors such as CXCL7. To further explore whether MMP13 activity was necessary for osteoclast formation, we tested a series of MMP13–selective inhibitors with nanomolar range IC₅₀s (Compounds 1, 2, and 4). In previous studies, these compounds were shown to selectively target the catalytic domain of MMP13 (Supplementary Fig. S10A and S10B; refs. 24, 25, 46). Of these reagents, we observed that Compound 1 potently suppressed wild-type osteoclast formation in vitro at concentrations <100 nmol/L compared with the other inhibitors (Fig. 6A). The remainder studies were therefore conducted with Compound 1, designated as MMP13i. MMP13 inhibition by MMP13i compromised the viability of myeloma cell lines (MM1.S, 5TGM1, OPM2, U266) albeit at concentrations >1 μmol/L (Fig. 6B). Interestingly, MMP13i did not affect the viability of MSCs or CD11b-isolated monocytes at concentrations <5 μmol/L (Supplementary Fig. S10C).

To determine the in vivo efficacy of the MMP13i, 6-week-old immunocompetent syngeneic KaLwRiJ mice (n = 30) were inoculated with 5TGM1-Luc cells to establish skeletal lesions. After 7 days, mice were randomized into vehicle control and MMP13i groups. Bioluminescence imaging results demonstrated a delay in myeloma growth over time (Fig. 6C). Similar to our genetic studies, we also observed a significant increase in overall survival in the MMP13i–treated group compared with vehicle control (median survival times 35 vs. 42 days, respectively; Fig. 6D). We noted that MMP13i treatment resulted in a trend of decreased osteoclast numbers (Supplementary Fig. S11A and S11B) with no difference in BV:TV ratios between the vehicle control and MMP13i treatment group in tumor-bearing mice. MMP13i treatment of tumor-naïve mice did, however, demonstrate an increase in BV:TV compared with controls (Fig. 6E). Collectively, based on these data, our working model is that that multiple myeloma induces MMP13 in the surrounding bone stroma and that MMP13 processing of factors such as CXCL7 enhances osteoclast formation rates, leading to increased myeloma-induced bone disease. We propose that inhibition of MMP13 activity can block this mechanism and thereby significantly enhance overall survival (Fig. 6F).

Multiple myeloma induces systemic skeletal lesions that greatly impact patient quality of life. The vicious cycle of myeloma–bone interaction increases bioavailability of cytokines and growth factors that enhance tumor growth and contribute to therapy resistance (1, 47). Mechanisms governing reciprocal interactions between the myeloma and surrounding bone microenvironment are therefore of potential therapeutic importance. Here, we have shown that MMP13 ablation from the host stroma significantly extended overall survival of myeloma-bearing mice. Interestingly, this effect appears to be due to MSC-derived MMP13 regulating the availability of multiple cytokines including CXCL7 that control osteoclast formation and activity rather than collagen turnover. Importantly, our studies show that this effect depends on MMP13 catalytic activity because an MMP13–selective inhibitor could significantly extend overall survival in multiple myeloma–bearing mice.

Although our mice are systemically null for MMP13, the studies herein have focused primarily on MSC- and osteoblast-derived MMP13 given our immunolocalization data in human and mouse multiple myeloma samples. Consistent with previous reports, this lineage is a major source of MMP13 with noted expression in chondrocytes and osteocytes (38). Underscoring the importance of MMP13 in osteoblast biology, tissue-specific knockout of MMP13 using type Iα collagen promoter-driven CRE recombinase recapitulates the developmental bone phenotypes noted in systemic MMP13–null mice (38). Importantly, MMP2- and -14–null mice also exhibit bone phenotypes, but they fail to compensate for the loss of MMP13, indicating a distinct role for this protease in bone remodeling. Cancer cells also induce MMP13 in MSCs and osteoblasts (Fig. 1B; Supplementary Fig. S3D; refs. 17, 28). To date, however, few studies have been able to examine whether stromal genes of interest such as MMP13 impact myeloma progression, particularly in vivo. This is largely due to the limited availability of genetically engineered models that are receptive to myeloma engraftment (48). Here, using RAG-2/MMP13 double-null mice, our data show that stromal MMP13 contributes to myeloma-induced bone destruction by regulating the bioavailability of nonmatrix molecules such as CXCL7 that impact osteoclast recruitment, formation and function as opposed to processing type I collagen. It is also important to note that several MMP-deficient mice carry a caspase-11 passenger mutation, including the MMP13–null mice used in this study (49). However, the in vitro and in vivo data obtained with our MMP13–selective inhibitor support, in large part, those obtained with the RAG-2/MMP13–null mice indicating negligible, if any, caspase-11 contribution. The similarities between the MMP13–null studies in immunocompromised mice and MMP13 inhibitor studies in immunocompetent KaLwRiJ mice also suggest immune cells, such as T cells, may not be involved in the effects mediated by host MMP13 (50, 51).

Ex vivo, our analysis of MSC conditioned media revealed the regulation of several nonmatrix molecules that have reported roles in controlling osteoclast biology including CXCL7 (44, 45). Expressed as an inactive propeptide (PPBP), CXCL7 undergoes successive rounds of proteolytic cleavage to eventually yield the active 7.6-kDa peptide (52). Here, we report for the first time that MMP13 regulates CXCL7 bioactivation and availability and is critical for MSC-induced osteoclast formation. Our mass spectrometry data demonstrated MMP13–directed CXCL7 cleavage sites resembling that of canonical activation sites by Cathepsin G processing (Supplementary Fig. S7A and S7B). We also noted that MMP13 could process CXCL7, at other distinct sites, albeit that these cleavage events were noted less frequently than the processing at Glu²⁷. These findings warrant further investigation into smaller CXCL7 fragments and how posttranslational modification by MMP13 potentially regulates cellular behavior in the tumor–bone microenvironment. Previously, we have shown this to be the case of parathyroid hormone related protein (PTHrP), a potent hormone involved in controlling cancer-induced osteolysis (32). While we acknowledge the currently unknown functional nature of the shorter fragments, our osteoclastogenesis treatment assays nonetheless support MMP13–processed CXCL7 retains osteoclastogenic activity (Supplementary Fig. S8).

While not examined here, it is also possible that the MMP13–CXCL7 interaction could influence the behavior of other cell types in the myeloma–bone microenvironment. For example, CXCL7 has been identified as a potent mediator of neutrophil chemoattraction and activation in various pathologies, in addition to noted roles in osteoclastogenesis (41, 44). Furthermore, we do not discount that myeloma- or MSC-derived MMP13 can regulate the bioavailability of CXCL7 expressed by other cell types/platelets in the bone marrow microenvironment. While MMP-3 and cathepsin G have been shown to participate in proteolytic activation of PPBP, we posit MMP13 is a key regulator of CXCL7 expression and activity given the low levels of CXCL7 detected in the conditioned media derived from MMP13–null MSCs and wild-type MSCs treated with MMP13i (52) and our data showing MMP13 can directly process CXCL7 to enhance its osteoclastogenic activity. Of note, MMP-3 expression is increased in MMP13–null MSCs (Fig. 5A) and MMP-3 has been reported to activate latent MMP13 (12). However, the increased levels of MMP-3 expression are insufficient to rescue protein levels of active CXCL7 supporting a role for MMP13 in the process. While we noted the dysregulation of other factors in the MMP13–null MSC conditioned media that can impact osteoclast biology, immune depletion of CXCL7 from conditioned media significantly inhibited the process.

Collectively, our study demonstrates a role for the catalytic activity of stromal MMP13 in multiple myeloma progression and overall survival. This is evidenced by our data examining CXCL7 processing and our in vitro and in vivo MMP13i data. In addition, the detected differences in nonmatrix factors in MMP13–null MSC conditioned media reinforce a role for MMP13 catalytic activity. This position is further supported by data showing that when wild-type MSCs are treated with a MMP13–selective inhibitor, the profile and levels of downregulated growth factors mirror those observed with the MMP13–null MSCs (Supplementary Fig. S7A). In agreement, published studies have also demonstrated MMP13 catalytic regulation of the MC3T3 osteoblast degradome (17, 53).

As stated, in the context of skeletal malignancies, MMP13 can also be derived from cancer cells (15, 54, 55). Indeed, our own studies show expression of MMP13 in CD138⁺ cells, and our analysis of the MMRF dataset also correlated MMP13 status with overall survival (Figs. 1D and 5B and C; Supplementary Fig. S1B). Interestingly, we noted that CD138⁺ myeloma cell MMP13 expression did not correlate with disease staging in publicly available datasets. Previous reports have shown that MMP13 derived from multiple myeloma cells (5TGM1) can contribute to osteoclast fusogenesis and that this effect is independent of the catalytic activity of MMP13 (15). This noncatalytic role for MMPs in osteoclast fusion has also been reported for MMP-14 (56). In our experiments, 5TGM1-derived MMP13 did not compensate for the loss of MMP13 in the bone stroma compartment, suggesting spatial localization may be important. Our results with MMP13–selective inhibitors further support a catalytic role for host-derived MMP13 in contributing to disease progression. The selectivity of the inhibitor was supported in experiments showing that the addition of the MMP13i to MMP13–null MSCs had no further effect on either downregulated or upregulated cytokines or growth factors (Supplementary Fig. S7B). MMP inhibitors as a modality to treat cancer were not successful in the clinical setting because of their largely broad-spectrum nature and dose-limiting side effects (57). However, the relatively deep catalytic pocket of MMP13 and restricted tissue expression of the protease to the tumor–bone microenvironment make it an ideal candidate for selective inhibition (58–61). Here, using a MMP13–selective inhibitor (IC₅₀ of 2.7 nmol/L), we demonstrate inhibition of MMP13 activity in vivo recapitulated the effects observed in our MMP13–null studies in regard to significantly improving overall survival in myeloma-bearing mice. Importantly, both genetic and pharmacologic ablation of MMP13 yielded comparable improved overall survival in mice as do standard-of-care treatments.

However, it is important to note differences between the genetic and pharmacologic approaches taken in this study. For example, we did not observe a protective effect on myeloma-induced bone destruction with the MMP13i. We suspect that this is due to the short duration (between 31 and 54 days) of treatment compared with genetically null mice where MMP13 is absent from birth. Therefore, over a longer period, this protective effect of inhibiting MMP13 on cancer-induced bone disease would be manifested more. Supporting this position is the fact that treatment of wild-type tumor-naïve mice for the same period of time significantly increases bone volume (Fig. 6E). While we observed a decrease in osteoclast numbers in MMP13i–treated animals, this reduction did not reach statistical significance. Pharmacokinetic/pharmacodynamic studies have not been performed with the MMP13i and so it is possible that higher doses or improving inhibitor half-life in vivo could more potently protect against osteoclastogenesis and myeloma-associated bone disease. Importantly, we do not rule out that the noncatalytic function of MMP13 in osteoclast fusion may also explain why a more dramatic effect on osteoclast numbers between the MMP13 inhibitor and vehicle control groups was not noted. Nevertheless, given the increase in overall survival noted in the MMP13i cohort, we posit that MMP13 catalytic activity plays an important role in the progression of the disease. In support of targeting MMP13 activity therapeutically for skeletal malignancies, previous studies with MMP13 inhibitors such as 5-(4-phenoxy)-5-(2-methoxyethyl)-pyrimidine-2,4,6 (1H,3H,5H)-trione (IC₅₀ = 0.57 nmol/L) demonstrated efficacy in a bone-metastatic breast cancer model by mitigating metastatic tumor burden and tumor-induced bone disease (62, 63). Ongoing studies by our group will leverage MMP13–specific fluorescent triple-helical peptides (fTHP) for analysis of MMP13 catalytic activity in vivo or ex vivo and should help delineate between the potentially catalytic and noncatalytic functions of this enzyme (62, 64).

In conclusion, we have shown that MMP13 contributes to the overall survival of multiple myeloma–bearing animals. Our data demonstrate that this is due to slower rates of bone turnover that is not related to the ability of MMP13 to process type I collagen, but rather the regulation of bioavailable CXCL7 that in turn is critical for osteoclast formation and function. We further demonstrate the efficacy of a novel highly selective MMP13 inhibitor for the treatment of the disease using an immunocompetent preclinical animal model. In conclusion, we have shown that stromal MMP13 activity contributes to multiple myeloma reduced overall survival and that MMP13 inhibition is a tractable and valid target for the treatment of this currently incurable disease.

G. Shay reports grants from Multiple Myeloma Research Foundation outside the submitted work. J. Choi reports a patent for US Patent Apps. US20200181095A1, June 11, 2020 issued. R. Fuerst reports other support from FWF-Austrian Science Fund during the conduct of the study; in addition, R. Fuerst has a patent for WO2018226837A1 Selective Matrix Metalloproteinase-13 Inhibitors pending. G.B. Fields reports nonfinancial support from MMP Biopharma outside the submitted work; in addition, G.B. Fields has a patent for Compounds and Methods for Inhibition of Multiple Myeloma pending. C.C. Lynch reports grants from NCI during the conduct of the study. No disclosures were reported by the other authors.

C.H. Lo: Data curation, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. G. Shay: Data curation, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. J.J. McGuire: Data curation, investigation, methodology, writing–review and editing. T. Li: Data curation, validation, writing–review and editing. K.H. Shain: Conceptualization, resources, writing–review and editing. J.Y. Choi: Conceptualization, resources, writing–review and editing. R. Fuerst: Conceptualization, resources, writing–review and editing. W.R. Roush: Conceptualization, writing–review and editing. A.M. Knapinska: Conceptualization, resources, writing–review and editing. G.B. Fields: Conceptualization, resources, writing–review and editing. C.C. Lynch: Conceptualization, resources, formal analysis, supervision, funding acquisition, methodology, writing–original draft, project administration, writing–review and editing.

MMP13-null mice on a C57BL/6 background were kindly provided by Dr. Stephen M. Krane, MGH (Boston, MA). Luciferase-labeled myeloma cells (5TGM1-Luc) were obtained from Dr. Toshiyuki Yoneda via University of Texas, Health Science Center at San Antonio, San Antonio, TX. Luciferase-labeled U266 cells (U266-Luc) were obtained from Dr. Steven Grant at the University of Virginia (Charlottesville, VA). MMP13 inhibitor (MMP13i) is an invention of G.B. Fields, W.R. Roush, J.Y. Choi, and R. Fuerst (US Patent Appl. US20200181095A1, June 11, 2020). These studies were supported in part by R01-CA239214-01 (to C.C. Lynch) and R21-CA191981-01A1 (to C.C. Lynch).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.