MMP13 and TIMP1 are functional markers for two different potential modes of action by mesenchymal stem/stromal cells when treating osteoarthritis

Mesenchymal stem cells (MSCs) have been investigated as a potential injectable therapy for the treatment of knee osteoarthritis, with some evidence of success in preliminary human trials. However, optimization and scale‐up of this therapeutic approach depends on the identification of functional markers that are linked to their mechanism of action. One possible mechanism is through their chondrogenic differentiation and direct role in neo‐cartilage synthesis. Alternatively, they could remain undifferentiated and act through the release of trophic factors that stimulate endogenous repair processes within the joint. Here, we show that extensive in vitro aging of bone marrow‐derived human MSCs leads to loss of chondrogenesis but no reduction in trophic repair, thereby separating out the two modes of action. By integrating transcriptomic and proteomic data using Ingenuity Pathway Analysis, we found that reduced chondrogenesis with passage is linked to downregulation of the FOXM1 signaling pathway while maintenance of trophic repair is linked to CXCL12. In an attempt at developing functional markers of MSC potency, we identified loss of mRNA expression for MMP13 as correlating with loss of chondrogenic potential of MSCs and continued secretion of high levels of TIMP1 protein as correlating with the maintenance of trophic repair capacity. Since an allogeneic injectable osteoar therapy would require extensive cell expansion in vitro, we conclude that early passage MMP13+, TIMP1‐secretinghigh MSCs should be used for autologous OA therapies designed to act through engraftment and chondrogenesis, while later passage MMP13−, TIMP1‐secretinghigh MSCs could be exploited for allogeneic OA therapies designed to act through trophic repair.


| INTRODUCTION
The concept of multipotent mesenchymal stem cells (MSCs) was established by Caplan in the early 1990s 1 following the seminal work of Friedenstein in the 1960s and 1970s, demonstrating the presence of osteogenic precursor cells in bone marrow. [2][3][4] Building on this early work, there have been many studies demonstrating the capacity of MSCs to differentiate in vitro with clear evidence for multipotent skeletal lineage differentiation, 5 including chondrogenesis, [6][7][8][9][10][11][12][13] osteogenesis, [14][15][16][17] and adipogenesis 9,18,19 as well as more limited evidence for pluripotent differentiation including endodermal and ectodermal pathways. 1,20 In one of these studies, we generated clonal populations of bone marrow MSCs and showed that individual cells retained the capacity for chondrogenesis, with varying degrees of potency. 9 More recently, evidence has grown that MSCs may support tissue repair through mechanisms that do not directly relate to their multipotential differentiation capacity. 21 Caplan has described MSCs as having "trophic" capacity by which, following implantation, they induce neighboring cells to secrete active molecules, for example, in the treatment of stroke, myocardial infarction, or in meniscal cartilage repair. 22 Trophic repair is most likely mediated through the production by MSCs of large amounts of growth factors and other mediators. 20,[22][23][24][25] We have previously developed a therapeutic strategy for meniscal cartilage repair based on the trophic properties of MSCs 26 that has shown some evidence of efficacy in preclinical and clinical trials. 27 A second mechanism contributing to trophic repair is the ability of MSCs to suppress immune responses by a range of mechanisms including downregulation of T cell proliferation. 9,24,[28][29][30][31] This important property of MSCs has been used clinically to support the engraftment of donated hematopoietic cells and to prevent graft vs host disease. 24,32 There have been several studies describing the loss of differentiation capacity with increasing passage of MSCs in vitro, [33][34][35][36] with other studies suggesting that in vivo aging also leads to a loss of differentiation capacity after ex vivo isolation of the aged cells. 37  Patient details can be seen in Table S1. Cells were suspended in stem cell expansion medium consisting of low glucose Dulbecco's Modified Eagles Medium (Sigma-Aldrich) supplemented with 10 vol%/vol% fetal bovine serum (FBS, Sigma-Aldrich), 1 vol%/vol% Glutamax (Gibco), and 1 vol%/vol% Penicillin/Streptomycin (Sigma-Aldrich). The serum batch was selected to promote the growth and differentiation of MSCs. 41 The medium was also supplemented with 10 ng/mL FGF-2 (Peprotech). This growth factor has been previously shown to enhance the MSC proliferation rate in vitro, 12,42 to retain MSCs as undifferentiated cells during proliferation 6,43 and to enhance chondrogenic differentiation when the FGF-2 expanded MSCs are subsequently exposed to differentiation conditions. 12,42 The cell suspension was separated from any bone in the sample by repeated washing with media. The cells were centrifuged at 500g for 5 minutes and the supernatant/fat removed. The resulting cell pellet was resuspended in medium, and then plated at a seeding density of between 1.5 × 10 5 and 2.0 × 10 5 nucleated cells per cm 2 . These flasks were incubated at 37 C in a humidified atmosphere of 5% CO 2 and 95% air. Four days were allowed before the first medium change and then the medium was changed every other day until adherent cells reached 90% confluence and were ready for passaging.

| Cell passaging and calculation of population doublings and doubling time
At the end of each passage, the MSCs were harvested using 0.25% trypsin-EDTA (Sigma-Aldrich), pooled, counted, and then divided into different centrifuge tubes for reseeding and further growth, for immediate use in measurement of % integration of meniscal cartilage, for storage in liquid nitrogen for subsequent use in differentiation protocols as well as for genomic and proteomic analysis. The cells for each

Significance statement
This study has shown that mesenchymal stem cell (MSC) chondrogenesis is a transient property of these cells, which is lost as they age in vitro, whereas trophic repair potency is maintained until the MSCs cease to grow. These findings are significant, as they highlight the importance of defining the intended mode of action when preparing MSCs for injection into osteoarthritic joints. Results show that the development of injectable MSC therapies for osteoarthritis must take into account the transient nature of chondrogenic potency relative to their sustained trophic potency with increasing passage, and specific strategies should be adopted to exploit one or other of these mechanisms of action. patient were passaged continuously without freezing, until growth arrest, defined as no detectable increase in cell number between passages (see Table S1). At each passage, the total number of harvested MSCs was determined. The first cell harvest after seeding of fresh bone marrow was taken as passage 0. The number of cells reseeded at the start of passage 1 was used as the baseline for calculation of the first population doubling (PD) value at the end of passage 1.
Downstream analyses of the MSCs were undertaken from passage 1 onward.
The number of PDs was calculated using the following formula:

| Biochemical analysis
Cartilage constructs were freeze-dried and weighed at the end of the 35-day tissue engineering period. The extracellular matrix was fully solubilized by overnight digestion with 2 mg/mL bovine pancreatic trypsin (Sigma-Aldrich) which was then boiled for 15 minutes to inhibit the action of the enzyme. 44 In order to obtain the dry weight of extracellular matrix in the construct, remaining undigested scaffold material was freeze-dried, weighed, and subtracted from the original dry weight. The amounts of proteoglycan in the digests was measured as sulfated glycosaminoglycan (GAG) using a dimethylmethylene blue (Sigma-Aldrich) colorimetric assay. 45

| Osteogenesis and adipogenesis
Whole MSC populations or MSC clones were grown in monolayer until 50% to 70% confluent prior to osteogenic differentiation or 100% confluent prior to adipogenic differentiation. In both cases, control cells were then cultured in minimum essential medium (α-MEM;  Figure S1) and increasing number of lipid droplets following adipogenic differentiation (see Figure S3).

| Assembling and culture of constructs
Sandwich constructs of two ovine meniscal cartilage disks interposed with a seeded scaffold were assembled as previously described 27 using skin clips and cultured in vitro in ultra-low attachment 6-well plates in expansion medium with 10 ng/mL FGF-2 for 7 days followed by culture in an integration medium consisting of high glucose DMEM containing 10 vol %/vol% FBS, 1 vol%/vol% Glutamax, 1 vol%/vol% penicillin/streptomycin, insulin, and 80 μM ascorbic acid 2-phosphate for 33 days. The medium was replenished twice every week. The constructs were incubated at 37 C on a rotating platform throughout the culture period. At the end of culture, the constructs were prepared for histological analysis by fixation in 10 vol%/vol% neutral buffered formalin (Sigma-Aldrich).

| Histomorphometric analysis
Histomorphometry was carried out using a method that we developed and characterized in previous studies. 26,27,46 Fixed constructs were dehydrated and paraffin embedded. Samples were cut into 4 μm sections and stained with H&E for the study of morphological details. All histological sections were scanned using a Leica Aperio slide scanner and histomorphometric analysis was performed under blind conditions, using ImageScope software (Leica). Two perpendicular sections, one at the edge and another one at the center of each construct, were used.
For each section, the entire length of the implant/meniscus interface was measured, as well as the length of any areas of integration at the interface. The repair index was then determined as:     47 Gene level count data were generated from the Bowtie2 alignments using htseq-count version 0.9.0. The R library DESeq2 was used to produce rlog transformed count data.

| Quantitative PCR
These were filtered to remove genes with less than 1 average count.
Statistical analyses were performed in R version 3.4.4 and graphical representations were done using the R package ggplot2.
Normalized proteomics and transcriptomics data were integrated and preliminary exploratory analyses revealed a relevant heterogeneity between patients. In order to discriminate between changes related to patient heterogeneity and changes related to passage, differentially expressed variables over passage were calculated with a two-way analysis of variance (ANOVA) to account for patient variability as confounder. This followed multiple testing corrections using the These were translated into integers 0 to 1 to undertake the calculation. showing no sign of growth arrest even at passage 30 ( Figure 1A and Table S1). Figure 1B  MSCs that is lost with cell expansion in vitro.

| Trophic repair by MSCs is maintained with increasing passage
We used meniscal cartilage integration and suppression of T-cell proliferation as assays relating to different aspects of trophic repair. Transcriptomics and proteomics data were processed according to data standards and integrated. The integrated data were further analyzed by Ingenuity Pathway Analysis to identify key genes and proteins that change over passage and map them to biological functions and predict possible upstream regulators (see Figures 6 and 7). B, PCA score plot of first two principal components of all genes and proteins. There is a segregation of patients (see blue and orange ellipses) that can also be linked to changes in osteogenic capabilities (see Figure S2), while changes related to passage are capture mainly in PC2. C, Venn diagram of significant variables after two-way analysis of variance revealing 338 genes and proteins that vary significantly and independently of the patient group variation. D, PCA of the 338 significant variables over passage. Shown the score plot of the first two principal components (capturing approximately 70% of variance). As expected, there is a marked segregation over passage. Heatmap of scaled data from the 338 significant proteins and genes that are represented in rows and clustered via Ward hierarchical clustering. Columns represent biological samples, each column is a patient sample and they are ordered by passage with red being passage 1, green passage 5, turquoise passage 10, and purple passage 15. PCA, principal component analysis related genes and proteins. The methodological approach to transcriptomic and proteomic analysis is illustrated in Figure 5A. Data structure was appraised via PCA. A score plot of the first two principal components is shown in Figure 5B. The changes captured over passage are accounted for in PC2 while PC1 captures considerable variation between patients which can be stratified in two groups (shown with blue and orange ellipses). These two groups of patients also show differences in osteogenic capabilities (see above). With the aim of identifying genes and proteins changing solely over passage, independently of this patient variability, we performed a two-way ANOVA for each variable, as described in the Section 2. Figure 5C shows the significant variables of these tests. Only 338 genes and proteins are significant uniquely over passage, independently of patient to patient variation or any related iteration. These 338 genes and proteins were used to calculate PCA, of which the first two principal components are shown in Figure 5D together with a heatmap of the 338 variables For each of these 338 significant proteins and genes, we calculated the fold change with respect to passage 1 and analyzed these data with Ingenuity Pathway Analysis (IPA) (QIAGEN, Inc., https:// www.qiagenbioinformatics.com/products/ingenuity-pathwayanalysis). IPA's core analysis, overlaid with the global molecular network within the software resulted in the identification of a number of canonical pathways, functions, and upstream regulators found to be significantly over-represented within this list and therefore linked to the loss of multipotential differentiation capacity of the cells. The largest and most significant change was in the cell-cycle master regulator FOXM1 gene pathway, with clear evidence from transcriptomic data for downregulation of the FOXM1 gene itself ( Figure 6A) and with six out of seven of its downstream effectors also predicted by IPA to be downregulated ( Figure 6B). Other upstream regulators predicted to be deactivated were the prostaglandin receptor PTGER2, and members of the Vascular endothelial growth factor family, while upstream regulators that were also predicted to be positively activated over passage included the proliferation regulator NUPR and the cytoskeleton regulator MYOC. Quantitative data from our transcriptomic analyses and the associated IPA predictions of changes in the downstream effectors of these regulators are shown in Figure S5; however, none of the F I G U R E 6 Ingenuity Pathway Analysis of integrated transcriptomics and proteomics data. A, Ingenuity pathway analysis identified FOXM1 and four other master regulators genes (see Figure S5) that vary with increasing passage number. FOXM1 data from transcriptomics analysis show inhibition of gene expression with increasing passage number. Each bar is the mean ± SEM for results using MSCs from each of the four patients. B, Ingenuity pathway analysis of the integrated transcriptomics and proteomics data set predicts inhibition of five out of six of the identified downstream regulators in the FOXM1 canonical pathway. C, Ingenuity pathway analysis identified multiple genes and proteins that mapped to the search terms "Cell Movement," "Cell Migration," or "Wound Healing." Most of these identified genes and proteins did not vary significantly over passage as determined by analysis of variance. The most highly expressed genes and proteins are listed for each of the search terms. D, CXCL12 data from transcriptomics analysis show continuous gene expression with increasing passage number. Each bar is the mean ± SEM for results using MSCs from each of the four patients. E, CXCL12 data from proteomics analysis show continuous protein secretion with increasing passage number. Each bar is the mean ± SEM for results using MSCs from each of the four patients. MSCs, mesenchymal stem cells changes and predicted downstream effects were as clear-cut as for FOXM1.

| Regulators of cell migration and wound healing can be linked to the trophic properties of MSCs
As shown in Figure 4, the trophic repair capacity of MSCs remains unchanged with increasing passage. With the aim of further investigating this phenomenon, we hypothesized that (a) the proteins and genes involved in regulating this function would not present a significant change in abundance/expression over passage and (b) functions related to cell movement, cell migration, and wound healing are likely to be mechanistically involved in trophic repair. To investigate this hypothesis, we used the IPA database to identify the proteins and genes involved within the three terms listed above. Then we mapped those lists to our data, extracted overlapping variables, and appraised their significance over passage. The large majority of the proteins and genes mapping to these search-terms were found to be unchanged over passage and expressed at a consistent level across all four passages of our MSC cultures ( Figure 6B), supporting the hypothesis that genes and proteins expected to be necessary for trophic repair continue to be expressed in aging cells, when multipotent differentiation capacity has been lost but trophic repair capacity remains high. We considered CXCL12 (also called stromal cell-derived factor 1) to be of particular importance at it is associated with all three of our search terms ( Figure 6C) and was expressed consistently highly at both gene and protein level ( Figure 6D,E).

| Marker genes and proteins
The IPA analysis outlined above demonstrated downregulation of the FOXM1 canonical pathway with increasing passage/loss of multipotent differentiation and continuous expression of CXCL12 and other cell migration and wound healing genes and proteins with increasing passage. However, we consider it necessary also to identify genes and proteins that may not be part of canonical pathways or gene/protein families, but that can be used as specific markers of F I G U R E 7 Gene and protein markers of the in vitro MSC aging process. The MMP13 gene was selected as a marker of early passage cells which is lost with aging of MSCs in vitro while secretion of the TIMP-1 protein was selected as a marker of MSCs that is independent of in vitro aging. A, MMP13 data from transcriptomics analysis show decreasing gene expression with increasing passage number. Each bar is the mean ± SEM for results using MSCs from each of the four patients. B, MMP13 data from quantitative polymerase chain reaction analysis shows decreasing gene expression with increasing passage number. Each bar is the mean ± SEM for results using MSCs from each of the four patients. C, TIMP-1 data from proteomics analysis show continuous protein secretion with increasing passage number. Each bar is the mean ± SEM for results using MSCs from each of the four patients. D, TIMP-1 data from enzyme linked immunosorbent assay analysis show continuous protein secretion with increasing passage number. Each bar is the mean ± SEM for results using MSCs from each of the four patients. E, enzyme linked immunosorbent assay analysis of TIMP-1 secreted by MSC/collagen scaffold constructs shows continuous protein secretion with increasing passage number for fresh constructs but reduced secretion from constructs that have been freeze-thawed under conditions that reduce their viability. For each cell source, the result is shown for one fresh compared with one frozen construct. MSCs, mesenchymal stem cells cellular aging in vitro. Such markers would aid in comparison of studies of cells from one laboratory to another or in determining the functionality of an MSC population being used for therapeutic purposes.
We therefore analyzed the gene array and protein data to identify candidate markers.
Within the significant variable genes identified by transcriptomic analysis, there was a significant decrease with increasing passage in the gene for matrix metalloproteinase 13 (MMP13; Collagenase 3; Figure 7A and Table S2) and a significant increase with increasing passage in the gene for Insulin-like growth factor binding protein 5 (IGFbinding protein 5; Table S2). There was no significant change in other MMP genes (Table S2) or IGFBP genes (Table S3), nor in any of the genes of the transforming growth factor family (Table S4) Within the proteomic data set, we identified those proteins expressed at highest abundance at all passages (Table S5). The most abundant of these proteins was metalloproteinase inhibitor 1 (tissue inhibitor of metalloproteinase 1 [TIMP1]; Table S5 and Figure 7C). We went on to validate these results using an ELISA kit assay to deter- consistent with previous studies of its biological function. It is a protooncogene that is a key master regulator in the survival of cancer stem cells. 48,49 It has also been shown to be highly expressed in multipotent and pluripotent stem cells and to be critical to the maintenance of stem cell potency 50,51 and to the induction of pluripotency through reprogramming. 52 CXCL12 (also known as SDF-1) was found to be associated with all three of our search terms related to trophic repair and its gene and protein levels were maintained even at very late passage numbers, indicating a potential role for CXCL12 in MSC-mediated trophic repair. This observation is consistent with previous studies which have demonstrated its critical role in MSC-mediated induction of spinal cord repair 53 and myocardial repair, 54 as well as enhancing nerve cell survival in vitro 55 and mediating trophism between endothelial cells and tumor cells. 56 We have identified FOXM1 and CXCL12 through combined proteomic and genomic analysis, but while this in silico approach is a powerful tool, any mechanistic involvement of these regulators must await experimental confirmation.
Previous studies have clearly shown that multipotency of MSCs declines with both in vitro and in vivo aging. [33][34][35][36][37][38][39] Our observation here that chondrogenesis declines, osteogenesis tends to decline, but adipogenesis is retained at higher passage number is in agreement with the work of Yang et al. 34 Muraglia et al 36 26,46 Importantly, we found that MSCs that have been stimulated with transforming growth factor-ß to undergo chondrogenic differentiation are much less potent at promoting meniscal repair than the undifferentiated MSCs. 26 We went on to describe the use of undifferentiated MSCs seeded on a collagen scaffold to repair meniscal injury in a sheep preclinical model and in a first in human trial. 27 In the current study, we used our in vitro semiquantitative meniscal cartilage integration assay 26 as a model for trophic repair and made the very surprising observation that MSCs that have been cultured for up to 30 passages retain the same capacity for trophic repair as very early passage cells. These functional data were supported by our analysis of genes and proteins involved with cell movement and migration and with wound healing, showing that unlike those linked to differentiation, there was no significant change in their expression with increasing passage.
Another aspect of MSC trophism is the immunoregulatory effects of MSCs. Although a range of mechanisms are involved, it is clear that inhibition of T-cell proliferation is a critical component of their suppressive activity. 9,24,28-31 Human T cells can be strongly stimulated to proliferate using a combination of anti-CD3 and anti-CD28 antibodies. 58 Their rate of proliferation can be monitored by covalently labeling intracellular molecules with a fluorescent dye that is then diluted out by 50% with each cell division, tracked by FACS. 59 Adding MSCs into cultures of labeled, stimulated T cells suppresses the lymphocyte proliferation, so prolonging the accumulation of the dye. 9 In the current study, we have used this method to measure the immunoregulatory effects of MSCs at early and late passage and found no loss of potency with increasing passage. For human MSCs, the mechanism of T-cell suppression has been described as involving indoleamine 2,3dioxygenase-mediated tryptophan degradation. 60 Taken together, these results demonstrate that, as with trophic repair, immunoregulation is a fundamental property of MSCs that is not lost with increasing passage.
Previous studies have proposed that the loss of multipotent differentiation capacity of MSCs with increasing passage is related to progression of the cells to a senescent end-state. [33][34][35][36] However, the results described here demonstrate that even after extensive passaging in vitro, there is no apparent loss of trophic repair. The term "Mesenchymal Stem Cells" was coined by Caplan, 1 who went on to describe MSCs as "An injury Drugstore" 23 and more recently advocated a change in their name to "Medicinal Signaling Cells". 61 Other studies have questioned the definition of MSCs as stem cells because of the lack of rigorous confirmatory biological evidence, 20,25,31,62,63 with all of these studies calling for more experimental data before reaching a conclusion on the nomenclature. Others have been more forthright in concluding that MSCs are not stem cells and have called for an immediate change in nomenclature, in order to avoid the overhyped marketing of MSCs as "miracle cures." [64][65][66] Prockop emphasized that the essence of a stem cell should not be determined by its status at a single point in time. 25 In this study, we have investigated the in vitro differentiation and trophic behavior of MSCs across many passages over time and in this way, have reached the conclusion that the property of multipotency is relatively transient whereas the trophic effects of these cells is apparently permanent all the way through to the time of growth arrest.

| CONCLUSION
These studies demonstrate that the development of injectable MSC therapies for OA must take into account the transient nature of chondrogenic potency relative to their sustained trophic potency with increasing passage and specific strategies should be adopted to exploit one or other of these mechanisms of action.