Autophagy Is Involved in Mesenchymal Stem Cell Death in Coculture with Chondrocytes

Objective Cartilage formation is stimulated in mixtures of chondrocytes and human adipose–derived mesenchymal stromal cells (MSCs) both in vitro and in vivo. During coculture, human MSCs perish. The goal of this study is to elucidate the mechanism by which adipose tissue–derived MSC cell death occurs in the presence of chondrocytes. Methods Human primary chondrocytes were cocultured with human MSCs derived from 3 donors. The cells were cultured in monoculture or coculture (20% chondrocytes and 80% MSCs) in pellets (200,000 cells/pellet) for 7 days in chondrocyte proliferation media in hypoxia (2% O2). RNA sequencing was performed to assess for differences in gene expression between monocultures or coculture. Immune fluorescence assays were performed to determine the presence of caspase-3, LC3B, and P62. Results RNA sequencing revealed significant upregulation of >90 genes in the 3 cocultures when compared with monocultures. STRING analysis showed interconnections between >50 of these genes. Remarkably, 75% of these genes play a role in cell death pathways such as apoptosis and autophagy. Immunofluorescence shows a clear upregulation of the autophagic machinery with no substantial activation of the apoptotic pathway. Conclusion In cocultures of human MSCs with primary chondrocytes, autophagy is involved in the disappearance of MSCs. We propose that this sacrificial cell death may contribute to the trophic effects of MSCs on cartilage formation.


Introduction
There is a range of current treatment modalities for symp tomatic and focal cartilage defects. 1,2 These include bone marrow stimulation techniques like microfracture or auto logous chondrocyte implantation (ACI) 3 and variations thereof. 4 Unfortunately, microfracture generates fibrous cartilaginous scar tissue and therefore provides nonana tomic restoration of articular surface. ACI, and related tech niques, have demonstrated superior mid and longterm outcomes as compared with the simpler microfracture pro cesses. 5 However, culturing of chondrocytes in a 2dimen sional environment to obtain sufficient cells for implantation can lead to changes in the chondrocyte phenotype. 6,7 Furthermore, substantial numbers of chondrocytes are harvested from an otherwise intact articular area creating additional damage in the joint surface. 6,8 To reduce the number of chondrocytes (CHs) required for cell implantation, a combination of chondrocytes and mesen chymal stromal cells (MSCs) has been studied. 9 Wu et al. 10 demonstrated a beneficial effect on cartilage formation over the respective monocultures. MSCs increased chondrocyte proliferation and stimulated deposition of cartilage matrix. However, the trophic effect generated by MSCs is followed by a counter loop where the chondrocytes signal the MSCs to undergo cell death. This mechanism has been confirmed using a variety of MSC sources both in vitro and in vivo and in a clinical trial in which cartilage defects were implanted with a mixture of preoperatively isolated chondrocytes and allogenic bone marrow derived-stem cells. 1114 It was postulated that the mechanism behind the death of the MSCs is likely related to one of the deliberate "suicide" programs present within cells. 15 These suicide programs are usually induced intrinsically or extrinsically by external stimuli such as mechanical stress, oxidative processes, and drug treatments. The programmed suicide death has usually 2 main forms, apoptosis and autophagy. 1618 These 2 path ways are intricately interconnected and the upregulation of one leads to downregulation of the other. 19 Apoptosis is triggered by biochemical events that induce characteristic changes in the morphology of the cell (i.e., membrane blebbing and nuclear fragmentation). The apop totic process can be intrinsically activated by the release of cytochrome c from mitochondria or can be activated extrin sically by death receptors. Both pathways lead to the activa tion of a series of caspases which mediate the cell destruction. 16 Autophagy, is a catabolic process where the cell degrades cytoplasmic components important for survival, thereby leading to a so called "selfeating" phenomenon. This orga nized degradation and recycle activity uses vesicles, known as autophagosomes, which can contain organelles, proteins, and other components. These autophagosomes subse quently fuse with lysosomes that degrade both the cargo and the vesicles. 19 In light of these results, we decided to investigate which of the 2 molecular mechanisms was involved in the possible cell death of MSCs in cocultures. Pellets containing the combination of human chondrocytes and human MSCs were cultured for 1 week and analyzed for changes in gene and protein expression characteristic for the apoptotic or autophagy pathways. Our data show a clear prevalence in activation of the autophagic pathway. We hypothesize that this mechanism could be a selfsacrifice mechanism of the MSCs which could contribute to the trophic effect of these cells on chondrocytes.

Cell Culture and expansion
Human adipose tissue-derived MSCs were extracted from lipoaspirates obtained from consenting 2 male healthy donors (A211 and A283) and 1 female donor (A258) (Supplementary Table 1) as previously described. 2022 Chondrocytes were extracted from healthy looking cartilage of a donor undergo ing an amputation. The use of cells for this study were approved by the Mayo Clinic Institutional Review Board. 22 MSCs were cultured in standard medium (Gilco's advanced modified Eagle medium; MEM) supplemented with 1% penicillinstreptomycin, 1% GLUTAmax, 5% human platelet lysate PLT max, and 0.2% heparin. Chondrocytes were cultured in chondrocytes proliferation medium (Gilco's advance MEM, 10% fetal bovine serum [FBS], 1% penicillinstreptomycin, 1% GLUTAmax, 0.2 mM ascorbic acid 2phosphate, and 4 mM proline). Both cell types were cultured in normal oxygen conditions (21% oxygen). MSCs were used at passage 5 and primary chon drocytes at passage 4.

Pellet Coculture
Cell pellets were generated by seeding 200,000 cells per well in a 96well plate. Cells were cultured as monocultures or as cocultures with a ratio of 80% MSCs/20% chondro cytes. Both mono and cocultures were cultured in chondro genic proliferation media (same as above) under hypoxia (2% oxygen). Medium was changed every three days. Pellets were harvested at day 7 for RNAseq and immuno fluorescence analysis.

RNa extraction
Total RNA was isolated from the samples using the Direct zol RNA kit as instructed by the manufacturer (Zymo Research, Irvine, CA). For each condition, 4 to 5 cell pellets were pooled to obtain sufficient RNA yield for downstream analysis. Nanodrop (Thermo Fisher Scientific, Waltham, MA) was used to determine the purity and concentration of the RNA extracted.

RNa Sequencing and analysis
RNA sequencing and bioinformatic analysis was performed by the Mayo Clinic RNA sequencing and bioinformatic cores as described previously. 2225 RNA libraries were pre pared using the TruSeq RNA library preparation kit (Illumina, San Diego, CA) following the manufacturer's instructions. The polyA mRNA of each sample was purified from the total RNA using oligo dT magnetic beads. To mul tiplex sample loading on the flow cells, specific indexes were incorporated at the adaptor ligand using the TruSeq kit. The constructs were purified and enriched using 12 cycles of PCR. Agilent Bioanalyzer DNA 1000 chip and Qubit fluo rometry (Invitrogen, Carlsbad, CA) were used to control the quality and the concentration of the samples. Libraries were loaded onto flow cells at concentrations of 8 to 10 pM to generate cluster densities of 700,000/mm 2 26 and quantification of gene expression using the HTSeq software. 27 Normalized gene counts were also obtained from MAPRSeq where expression values for each gene were normalized to 1 mil lion reads and corrected for gene length (Reads Per Kilobase pair per Million mapped reads, RPKM). RNAseq data were deposited in the Gene Expression Omnibus of the National Center for Biotechnology Information (GSE142831).

identification of Coculture Regulated genes
To estimate the relative RNA contribution of the MSCs in the cocultures, we used the sex mismatch between the male donors (A211 and A283) and the female chondrocyte donor. The average level of expression of 7 unique male markers in the coculture was 57% of the levels in the respective monocultures ( Table 1). As expected, the Ymarkers were not expressed in the female chondrocytes and MSC (A258). We used these numbers to calculate the expected value of gene expression of a given gene using the following for mula for each donor pair and calculated the average of the 3 donors. Only genes with RPKM values above 0.3 in each of the samples were included in the analysis.
The expected value is valid under the assumption that gene expression in cocultures is the sum of gene expression in MSCs and chondrocytes and is not influenced by the inter action between both cell types. The real value was then compared with the expected value to determine the fold change (Fc) difference.
The majority of genes have an Fc of around 1 indicating that the observed gene expression is the sum of the expression in the MSCs and chondrocytes. Genes with an Fc > 2 cutoff were considered upregulated genes. Genes with an Fc < 0.5 were considered downregulated genes.

immunofluorescence Staining
Pellets were harvested for immunofluorescent staining as previously described. 28 Cellpellets were washed with phos phatebuffered saline (PBS) and fixed with 10% formalin for 15 minutes. Samples were then embedded in cryomatrix (Thermo Fisher) and cut into 10μm sections with a cryo tome (Shandon). Sections were permeabilized with 0.5% Triton X100 in PBS for 10 minutes at room temperature followed by animal serum treatment (5%, 1 hour, room tem perature) to block nonspecific binding. Sections were incu bated overnight at 4°C in a humidified chamber with antibodies against LC3B (1:500 dilution, MAB85582, R&D System) and SQSTM1/p62 (1:200 dilution, ab56416, Abcam). Subsequently, slides were washed with 0.1% Tween 20 in PBS and incubated with Alexa Fluorconjugated secondary antibodies (Alexa 568 or Alexa 488, Abcam) for 50 minutes at room temperature in a humidified chamber. Nuclei were counterstained with 4,6diamidino2phenylin dole (DAPI, Molecular Probes) and images were taken with a fluorescence microscope (Nikon Eclipse E400).

transcriptome Changes in MSCs and Chondrocyte Cocultures
The relative contribution of the MSCs in gene expression in the cocultures dropped from 80% (based on seeding ratio) a the values of the monoculture are compared with the coculture and an average of the 7 genes was obtained. Based on these data, we assumed that 57% of the rPKM in the cocultures was derived from the mesenchymal stromal cells and 43% was derived from chondrocytes.
to 57% after 1 week in coculture ( Table 1). This confirms the disappearance of MSCs from cocultures and is in agree ment with previous observations. 9 Cluster analysis of RNAseq data from monocultured MSCs, chondrocytes, and cocultures at day 7 shows cluster ing of the distinct conditions in their respective groups (Fig.  1b). There is visible variation present in heatmap patterns when comparing the 3 MSC donors, indicative of interdo nor variation. 29 Moreover, the up and downregulated genes in the hCHs and MSCs in the monoculture differ greatly from the coculture suggesting an interaction between the 2 cells. Based on the similarity in gene expression patterns between the 3 cocultures, it is likely that common pathways are influenced by the interaction between the MSCs and the chondrocytes.
A total of 362 genes met the inclusion criteria of >2fold upregulation and P < 0.05 in at least one of the cocultures compared with the expected value assuming no cellular interaction between the MSCs and chondrocytes. In total, 137 genes were downregulated. The number of upregulated genes was nearly 3 times higher compared with the down regulated genes, which may indicate a predominance of pathway activation when cells are placed in coculture. A Venn diagram was created to identify common up or downregulated genes in each of the cocultures. Cocultures of chondrocytes with donors A258 and A283 represented a closer pattern of gene up and downregulation as compared with the coculture involving donor A211, which further highlights the presence of interdonor variability (Fig. 1c). In total, 92 genes were more than 2fold upregulated in all 3 cocultures and 74 were more than 2fold downregulated (Supplementary Figure 1).

Upregulation of autophagic and apoptotic Pathways in MSCs and Chondrocyte Cocultures
Among the 92 upregulated genes, 86 were identified (Supplementary Table 6) by the STRING software and 51 were found having known interactions (medium confidence 0.4). Two main clusters were obtained: one consisting of a series of histones (HIST1H2AG, HIST1H2BG, HIST3H2A, HIST1H1E) which may indicate an effect on cell prolifera tion in line with previous observations. 10 The second, a larger cluster of 51, consisted of a series of genes mainly present in cell death processes like autophagy and apoptosis (37 out of 51 genes 16,30 ) (Fig. 2). The genes identified encode surface (RARRES3, 31 TNFRSF10A, 32 PIK3R3, 33 TRAF1, 34 Table 2). In general, most of the GOterms obtained from the ClueGO analysis represented terms like cell stress or death pathways ( Fig. 2 and Supplementary Table 3). Moreover, 52.17% of the terms GOpathways detected were related to ubiquitinspecific processing proteases. Ubiquitination rep resents a fundamental process in the autophagic machinery. 38 These data suggest that the coculturing of MSCs and chon drocytes may induce autophagic cell death.

Coculture treatment induces autophagic Flux
To further distinguish whether the MSCs disappear from coculture by apoptosis or autophagy, the expression of pro apoptotic, antiapoptotic, apoptotic, and autophagic mark ers was assessed (Fig. 3a-d).
First, we looked at 4 proapoptotic markers (BAD, BAX, BIM, BID) (Fig. 3a). These markers belong to the BCL2 celldeathregulatorfamily and they initiate and/or mediate the activation of apoptosis. Here, by comparing the mono culture with the coculture, the variability among the MSCs donors is clear. Some of the genes are overexpressed in mono or coculture depending on the donor. However, the trends between the 2 conditions do not present any statisti cal difference, which suggests inactivation of the apoptotic pathway (Supplementary Table 4). This is furthermore sup ported by the gene expression of four of the main caspase pathway regulators (CASP3, CASP5, CASP8, CASP9) 16 that did not change or were less expressed in cocultures (Fig. 3c). Moreover, immunofluorescence images were obtained to determine the level of caspase3 (Fig. 4b).
Caspase3 is the final protein of the apoptotic cascade cycle. 39 During apoptosis, the cells present disrupted nuclei comprising high level of caspase3 (Fig. 4a, left). However, in both mono and coculture, this characteristic is barely present indicating absence of apoptotic activation.
We further checked 2 of the main antiapoptotic markers (BCL2 and BCL2L1). Although both did not reach a statis tical significance (0.11 BCL2 and 0.15 BCL2L1), both pre sented, for each of the 3 donors, an increase in expression in coculture ( Fig. 3b and Supplementary Table 4). This sug gests an activation of antiapoptotic processes, which may cause the inactivation of the caspase cycle. Furthermore, as regulated necrosis can be triggered by binding of TNFα and FAS ligand we looked at overexpres sion of RIPK1, RIPK3, and MLKL. However, none of these genes were significantly upregulated in the cocultured com pared with the monoculture (Supplementary Fig. 4) exclud ing a role for necrosis in the disappearance of the MSCs.
We next determined the effects of autophagic markers in the cocultures (Fig. 4). First, we looked at multiple markers highly present during autophagy activation (Fig. 3d). Among the 8 individual markers, 7 were statistically upregulated in the coculture compared with the monoculture, indicating acti vation of the autophagic machinery at the gene level. We fur ther looked at the protein level using immunofluorescence. Here, 2 wellrecognized markers (LC3B and P62) were used. LC3B is conjugated to the autophagosome during autophago some formation. P62/SQSTM1 protein interacts with both LC3BII and ubiquitin protein and is degraded in autopha golysosomes. 40 In inactivated autophagy, these 2 markers can be singularly present ( Supplementary Fig. 3 A211) or can be present together but not colocalized ( Supplementary Fig. 3  A258). On the contrary, in active autophagy both markers are present and colocalize (Fig. 4a), which indicates creation of the autophagosome. Even if the behavior of the 3 donors is different, it is clear that the level of combined LC3B and P62 are higher in the coculture compared with the monoculture for all the donors, indicating upregulation of the autophagic machinery. 41 Taken together, these results suggest that MSCs exhibit enhanced autophagic flux.

Discussion
In this study, we have studied the mechanism involved in the progressive cell death of the MSCs in coculture with chondrocytes.
We used the sex mismatch between the female chondro cyte and male MSC donors to estimate the relative contribu tion of the MSCs to the gene expression in the cocultures. The expression of Ychromosomespecific RNAs proved stable across donors. By assuming that the expression of Yspecific markers is not influenced by the coculture condi tions, a notion which is supported by previous observations, it is possible to estimate the contribution of the male MSC donors to global RPKM in coculture with female chondro cytes. Using this approach, we concluded that the relative contribution of MSCs to the RPKM in the cocultures dropped from 80% to 57% after 1 week of culture. This supports previous observations where after 4 weeks of culture, MSCs have almost completely disappeared from cocultures with chondrocytes due to cell death irrespective of the origin of the stem cells. 9 Signs of increased cell death were first noted at day 7 and increased at day 14. 10 We rea soned that the 7day time point marked the beginning of the disappearance of MSCs from cocultures and selected this time point for an RNAseq analysis. At this time point, the transcriptome of the cocultures was substantially and statis tically different from the respective monocultures indicating the presence of nonadditive interactions between the 2 cell populations. 362 genes were upregulated of 2fold in at least 1 of the cocultures. Of these genes, 92 were consistently upregulated in each of the 3 cocultures. The considerable interdonor variability is in line with previous studies. 42 Interestingly, the donor A258 (female) and A283 (male), presents higher overlap in terms of both up and downregu lated genes in coculture suggesting that the trophic role of MSCs is a generic, sexindependent property of these cells as noted before. 43 We do realize that the method for selecting genes specifically regulated by the interaction between chondrocytes and MSCs has limitations. For example, genes that are inversely regulated in MSCs and chondrocytes may be missed. The genes identified in this study thus represent a snapshot of genes that are regulated in cocultures.
STRING analysis identified 2 interconnected networks of which the main included more than half of the total upregulated genes. This network describes genes involved in apoptotic or autophagic processes and comprised pro teins involved in subsequent steps of these pathways start ing from membrane receptor proteins (TNFRSF10A, PIK3R3, TRAF1) up to fundamental enzymatic compo nents (OAS3). Most of the genes detected are upregulated in one (RARRES3), or both processes (DIRAS3, ISG15, BMP2, PHLDA2). Other genes are involved in the inhibi tion of one (HBEGFfor apoptosis) or both pathways (ATF3). Moreover, genes such as BIRC3 and NCF2 can upregulate one process while inhibiting the other. Interestingly, these results differ from those of Wu et al. 44 where a microarray analysis identified clusters related to intracellular cell cycle regulators, extracellular matrix pro duction and secreted growth factors (FGF1 and BMP2). This variability could be related to the time points consid ered. Wu et al. 44 performed the analysis at the second day of culture, whereas our study focused on a later time point (day 7). At this earlier time point, a dominant effect on cell proliferation was found whereas the disappearance of MSCs from the cocultures was first noted at day 7. 9,10 Also, in our study, we show the upregulation of many proliferation markers, however, this upregulation was modest and conse quently did not meet the cutoff of 2fold used (Supplementary Table 7). No upregulation of cartilage matrix genes was noted, which could be explained by the early time point considered. Interestingly, a decrease in COL10A1 and COL3A1 mRNA expression was visible indicating a reduc tion in the hypertrophic activity ( Supplementary Fig. 5).
Like in the study of Wu et al., 44 we have found upregula tion of BMP2. This suggests that BMP2 plays a fundamen tal and prolonged role in the coculture effect. This contrasts FGF1, which in this study was not found upregulated. Combined this data may suggest that FGF1 functions as trigger in an initial phase in chondrocyte proliferation, whereas BMP2 is required for longer period of time and may sustain cartilage matrix formation. 4548 Alternatively, the differences could be explained by the use of adipose MSCs versus bone marrow MSCs in the study by Wu et al. 44 This seems, however, unlikely given the consistency in the trophic effect of MSCs from a variety of sources in cocul ture with primary chondrocytes. 9 Moreover, the use of a more physiologically relevant hypoxic environment, used in this study, compared with the normoxic environment in the study by Wu et al., 44 may also have contributed to the differences in gene expression. Apoptosis can be initiated by an extrinsic or, an intrinsic process both ending with the cleavage of the procaspase3. 16 Here, the RNA expression data suggested the possible acti vation of the extrinsic pathway by membrane receptor proteins (TRAF1, TNFRSF10A) rather than the activation of the intrinsic pathway by genes like DIABLO, HTRA2, AIFM1, ENDOG, and CAD (Supplementary Table 5). Since we did not found an in increase or difference in the active form (caspase3 within the nuclei and nuclei debris) at the protein level, we concluded that the cell death via the canon ical apoptotic pathway is likely not driving MSCs death.
Autophagy is a cellular degradation pathway that is essential for survival. 49 However, if overexpressed, it could lead to neurodegeneration, cardiomyopathies, abnormali ties of skeletal development, and death as shown in mice studies. 18,50 LC3II is as a quantitative marker of autophagy required for the formation of the autophagosome and its expression is proportional to the amount of autophago somes in the cell. The P62/SQSTM1 protein serves as a link between LC3 and ubiquitinated substrates. 51 P62/SQSTM1 and P62bound polyubiquitinated proteins become incorpo rated into the completed autophagosome. From the data obtained, LC3II and P62/SQSTM1 were highly expressed, both at the gene and protein level, in the cocultures com pared to the monoculture conditions suggesting activation of the autophagic flux. Taken together, our data shows a strong association between the activation of autophagy and the disappearance of MSCs from cocultures with chondro cytes suggesting that the MSCs in coculture preferentially die by autophagy rather than by apoptosis. Formal proof of this hypothesis would require further studies for example by knocking down genes involved in the autophagic and apoptotic pathways.
This conclusion differs from the previous study per formed by Wu et al. 10 where high levels of TUNEL positiv ity staining were detected in the pellets. TUNEL staining detects the DNA breaks formed when DNA fragmentation occurs in the last phase of apoptosis. Normally, this kind of apoptosis is considered as caspasedependent procedure. However, cell death can proceed in caspaseindependent apoptotic pathway, in which TUNEL positive staining is also observed. 52 The mitochondria play a central role in both caspasedependent and caspaseindependent death pathways. 53 It has been recognized that mitochondria can release factors involved in caspaseindependent cell death, including apoptosisinducing factor (AIF) and endonucle ase G (EndoG). 5456 In fact, AIF is believed as a key media tor of poly ADPribose (PAR) polymerase (PARP) induced caspaseindependent cell death. 57 Indeed, it has been reported that autophagy is a cytosolic event that controls caspaseindependent macrophage cell death through RARP mediated pathway. 58 Moreover, autophagy activated by DNA damage can kill the cells through the autophagy regu lators in the absence of apoptosis. 5961 However, the major concern with these examples is that they represent a very artificial situation. Regardless, these data may explain the difference between these 2 studies.
It remains unclear which role MSCs cell death by autophagy plays in the coculture. We hypothesize that the autophagic extracellular vesicles generated may have an additional trophic effect. During autophagy, cells release a variety of signals, including extracellular vesicles, which after uptake by neighboring cells, induce cellular responses over short and/or longrange distances. 62 Indeed, research ers proposed the concept of "altruistic cell suicide" based on the observation that dying cells could induce prolifera tion of neighboring cells. 63 Based on our analysis, it is con ceivable that the activation of autophagy in MSCs likely initiates this "altruistic cell death" process in coculture with chondrocytes. We propose a sequential mechanism where growth factors (FGF1 and BMP2) are released by the MSCs and subsequently are further enhanced by the increased secretion of extracellular vesicles, and their uptake by neighboring chondrocytes. This mechanism can explain how MSCs stimulate cartilage formation while simultane ously disappearing both in vitro and in vivo.
Additional studies should focus on the mechanism behind the initiation of autophagy and genetic interference studies using, for example, knock down approaches, the release of extracellular vesicles, and their uptake. Particularly interest ing are the studies aimed at analyzing the content of the autophagic vesicles. Activation of autophagy in MSCs might be an efficient way to increase the formation of trophic extra cellular vesicles. This may help in optimizing intraarticular injection strategies based on MSCderived extracellular ves icles rather than MSCs themselves. 6466 The use of MSC derived extracellular vesicles rather than the cells themselves may avoid possible longterm phenotype changes of incorpo rated cells and attenuate many of the safety concerns related to the use of living cells.

Conclusion
In summary, here we provide evidence that MSCs in cocul ture with primary chondrocytes preferentially die by autophagy. We postulate that this altruistic cell death results in the formation of extracellular vesicles. These extracellu lar vesicles are an additional mechanism by which the MSCs stimulate chondrocyte proliferation and cartilage matrix formation in pellet cocultures.