MMA secreted from tumor cells activates fibroblasts in the tumor microenvironment
Aberrations in the enzymes downstream of methylmalonyl-CoA in the propionate metabolism pathway, namely methylmalonyl-CoA mutase (MUT), methylmalonyl-CoA epimerase (MCEE), methylmalonic aciduria type A protein (MMAA), or cob(I)yrinic acid a,c-diamide adenosyl-transferase (MMAB) result in pathogenic systemic MMA accumulation in methylmalonic acidemias (10–13), and drive cancer drug resistance and metastasis through increased MMA accumulation in vitro and in vivo(3)(Fig. 1a). We profiled the transcripts of these metabolic enzymes in individual cells obtained from resected human lung cancer primary tumors and metastases, and found that tumor cells with reduced expression of these genes were enriched in mesenchymal subpopulations (Fig. 1a, 1b). Given this, and our previous findings that metastatic inducers drive MMA production and pro-aggressive effects on tumors through dysregulation of propionate metabolism, we wondered if tumor-produced MMA might also act on other cell types in the TME(3). Fibroblasts comprise the major component of the TME, and in some solid tumors even outnumber malignant cells(14). We knocked down MUT in A549 lung carcinoma and A375 melanoma cells to simulate MMA accumulation by altered propionate metabolism during early steps of metastasis, and co-cultured these cells with MRC5 lung and BJ dermal fibroblasts, respectively (Fig. 1c-e). Five days of co-culture markedly increased CAF markers in the fibroblasts, suggesting that tumor-produced MMA is secreted and activates fibroblasts in the stroma (Fig. 1f). Conversely, blocking MMA production in A375 cells by knockdown of PCCA, a component of propionyl-CoA-carboxylase, repressed their ability to induce the activation and infiltration of fibroblasts in the tumor in vivo (Figure s1a-c). Notably, an RNA-sequencing dataset of 501 whole tumors from patient lung squamous cell carcinomas showed a correlation between low MUT, MCEE, MMAA and MMAB levels (indicating high MMA) and high expression of cancer-associated fibroblast markers ACTA1 (encoding for SMA) and FAP (Figure s1d), suggesting that human tumors with greater levels of MMA do indeed harbor a great proportion of inflammatory CAFs.
We have previously demonstrated that MMA in the serum is largely encapsulated in lipid vesicles, allowing for accelerated entry into cells at much lower concentrations compared to free MMA (15). When we isolated extracellular lipid vesicles from the conditioned media of MUT-knocked down tumor cells (EVsshMUT−A549), we found that they indeed carried more MMA compared to control vesicles (EVsshGFP−A549), and could induce CAF markers when used to treat fibroblasts (Figure s1e-f). Depletion of these vesicles from the conditioned media of MUT-knocked down cells abolished its ability to induce CAF markers in fibroblasts, confirming that tumor-produced MMA, like the MMA in the serum of elderly people, is delivered and acts on cells through extracellular vesicles (Figure s1g).
Treatment of MRC-5 and BJ fibroblasts with exogenous MMA reproduced the effect of co-culture with or extracellular vesicles from MUT-knockdown tumor cells on CAF marker expression in a dose-dependent manner (Fig. 1g). The ability of exogenous MMA to induce CAF markers in fibroblasts was similar to that of the conditioned media and lipid vesicles from MUT-knocked down tumor cells, as well as other known CAF inducers, including TGFβ (Figure s1h). Proliferation was not affected by 1mM MMA treatment, and mildly decreased under 5mM of MMA (Figure s1i). We confirmed that MMA activation of CAFs was not simply due to decreased pH or altered TCA cycle flux, as other acids from the propionate metabolism pathway were unable to reproduce the phenotype (Figure s1j). Intriguingly, MMA also induced CAF production of matrix metalloproteinases (Fig. 1f), which contribute to the extracellular matrix (ECM) remodeling that promotes intravasation of tumor cells into the bloodstream in early stages of metastasis (16).
MMA-treated fibroblasts secrete EVs to promote pro-aggressive reprogramming in tumor cells
To determine if the secretome of MMA-activated CAFs might direct tumor cell behavior, we cultured tumor cells with conditioned media from vehicle- or MMA-treated fibroblasts (CMveh−MRC5/BJ and CMMMA−MRC5/BJ) and observed a marked increase in markers of EMT (Fig. 2a). Additionally, co-injection of A549 tumor cells with MMA-treated MRC5 fibroblasts into mice significantly increased the ability of tumor cells to metastasize, indicating that one or more secreted factors from MMA-activated CAFs promotes a pro-metastatic phenotype in cancer cells (Fig. 2b).
Next, we aimed to identify the components of the conditioned media secreted by MMA-treated fibroblasts that was driving the EMT phenotype in tumor cells. EVs are loaded with signaling molecules and genetic material, and function as essential signaling mediators in the tumor microenvironment (17, 18). Considering that MMA is delivered from tumor cells to fibroblast in EVs, we looked to see whether the fibroblast messengers reciprocally driving EMT in tumor cells were also contained in EVs. From MRC-5 lung and BJ dermal fibroblasts, we isolated EVs from the conditioned media after vehicle or MMA treatment (EVsveh−MRC5 and EVsveh−BJ, or EVsMMA−MRC5 and EVsMMA−BJ, respectively) (Fig. 2c). We did not observe a significant difference in the number or size of EVs secreted by MMA-treated fibroblasts (EVsMMA−MRC5/BJ) compared to those secreted by vehicle-treated fibroblasts (EVsveh−MRC5/BJ) (Figure S2a-b). Survey of extracellular vesicle marker proteins confirmed the purity of these EVs (Figure S2c). To determine if the CAF-secreted factor driving EMT in tumor cells was being delivered through these structures, we then added EVsveh−MRC5/BJ or EVsMMA−MRC5/BJ to their tissue-matched A549 or A375 tumor cells (Fig. 2c). Upon treatment of tumor cells with EVsMMA−MRC5, we once again observed an increase in EMT markers (Fig. 2d). In contrast, the supernatant from the CMMMA−MRC5/BJ after isolation of the EVs lost its ability to induce this effect (Figure S3a). Intriguingly, when A549 tumor cells treated with isolated EVsMMA−MRC5 were then cultured in normal media, they converted back to an epithelial phenotype after five days, highlighting the plasticity of EMT (Figure S3b). Importantly, when EVsMMA−MRC5-treated tumor cells were released from EVsMMA−MRC5 treatment, but subsequently co-cultured in the presence of untreated fibroblasts, the tumor cells maintained their aggressive phenotype (Figure S3b). This underscores the importance of a positive feedback loop between the tumor and stroma, wherein fibroblast activation drives tumor cell aggression, which reciprocally drives more fibroblast activation, ultimately leading to metastatic progression. A375 and A549 tumor cells treated with EVsMMA−MRC5/BJ also exhibited increased resistance to chemotherapeutic and targeted therapy drugs, and displayed increased colony formation in soft agar compared to tumor cells treated with EVsveh−MRC5/BJ (Fig. 2e and 2f). Additionally, tumor cells treated with EVsMMA−MRC5/BJ exhibited increased invasion and migration ability in transwell assays, and formed more metastases following a subcutaneous primary tumor implantation in vivo (Fig. 2g, 2h). Intriguingly, despite having significantly higher metastases formation, tumor cells treated with EVsMMA−MRC5 did not form significantly larger primary tumors (Fig. 2h). This indicates that the EVs isolated from MMA-activated fibroblasts specifically drive an aggressive, metastatic phenotype in tumor cells, rather than increased cell proliferation.
IL-6 in fibroblast-secreted EVs mediates tumor cell metastatic signaling
As we did not see a change in the number and size of EVs induced by MMA treatment, we speculated that the potent tumor cell response observed after MMA treatment could be due to differentially loaded EV cargo. To identify the active factor in EVs from MMA-treated fibroblasts driving metastatic progression, we performed proteomic analysis on EVsveh−MRC5 and EVsMMA−MRC5. One of the most significantly upregulated secreted proteins in EVsMMA−MRC5 compared to EVsveh−MRC5 was IL-6, a pro-inflammatory cytokine that has been implicated in promoting EMT and metastasis (Fig. 3a, Figure s4a) (19–21). We also observed that genes driving IL-6/JAK/STAT3 pathway activity were enriched in more mesenchymal cells characterized by downregulation of key genes restricting MMA production from human lung cancer tumor and metastasis tissue samples (Figure S4b). Indeed, both IL-6/JAK/STAT3 signaling, measured by JAK2 and STAT3 phosphorylation, and TGFβ signaling, measured by phosphorylation of SMAD proteins, were activated in A549 cells upon treatment with EVsMMA-MRC5 (Fig. 3b). Notably, while EVsMMA−MRC5 increased Y705 phosphorylation of STAT3, which is the main regulator of cytokine-induced JAK/STAT3 signaling, it did not affect phosphorylation at S727 (Fig. 3b), suggesting a specificity in EVsMMA−MRC5 mediated downstream signaling. To determine the necessity of these signaling cascades for the ability of EVsMMA−MRC5 to drive EMT, we blocked their activation in A549 tumor cells using the TGFβR or STAT-3 inhibitors, SB431542 and cryptotanshinone, respectively (Fig. 3c). Inhibition of these pathways effectively blocked EMT induction by EVsMMA−MRC5 in A549 tumor cells, re-sensitized cells to drug treatment, and suppressed the increase in invasion and migration (Fig. 3d-f). Similarly, knockdown of IL6R in tumor cells suppressed both IL-6/JAK/STAT3 and TGFβ signaling, and suppressed EVsMMA−MRC5-induced EMT marker expression, drug resistance, and invasion and migration, suggesting that IL-6R activation functions upstream of TGFβ pathway signaling in this context (Fig. 3g-j, Figure S4c). Additionally, treating A549 lung tumor cells with tocilizumab, an IL-6R antibody and inhibitor, replicated the effect of IL-6R knockdown, effectively blocking IL-6/JAK/STAT3 and TGFβ signaling and suppressing the induction of EMT and drug resistance by EVsMMA−MRC5 (Figure S4d-f). Finally, we knocked down IL6 in MRC-5 fibroblasts before treating them with MMA and isolated their secreted EVs. While IL-6 knockdown in fibroblasts did not have any effect on the ability of MMA to induce CAF marker expression in fibroblasts, it effectively suppressed the ability of EVsMMA−MRC5 to induce IL6/JAK/STAT3 and TGFβ signaling in tumor cells, and was sufficient to abolish the EMT-inducing effect of EVsMMA−MRC5 and their ability to boost drug resistance, invasion and migration (Figure s5).
MMA activates fibroblasts through ROS activated NF-κB and TGFβ signaling
Next, we set out to characterize the mechanism by which MMA treatment of fibroblasts led to activation of the CAF phenotype and IL-6 loading into and secretion from extracellular vesicles. We performed RNA-seq on MRC-5 fibroblasts treated with vehicle or MMA, and a pathway enrichment analysis of the RNA-seq data showed an upregulation of genes in the NF-κB and TGFβ signaling pathways in MMA-treated fibroblasts (Fig. 4a). Crosstalk between these two pathways has been described previously, wherein TGFβ signaling leads to the sequential phosphorylation of TAK1, IKK, and NF-κB (Fig. 4b)(22). We confirmed that these pathways are activated in MRC-5 fibroblasts upon MMA treatment, or treatment by EVs derived from MMA-producing tumor cells (EVsshMUT−A549) (Fig. 4c, s6a). Using time course analysis, we noted that p65 phosphorylation occurred later than SMAD3 and TAK1 phosphorylation (Fig. 4c). Pharmacological inhibition of TGFβR using SB43152, but not of TAK1 and IKK using Takinib and IKK16, respectively, effectively suppressed the induction of CAF markers by MMA, suggesting that the MMA-induced CAF phenotype is largely regulated by TGFβ separately from NF-κB signaling (Fig. 4d-e). Similarly, genetic knockdown of TGFBR1, but not CHUK1 (encoding for IKK1), negated the ability of MMA to induce CAF markers (Figure s6b-c). Interestingly, pharmacological inhibition of TGFβR, TAK1 and IKK were all individually able to reduce IL-6 loading into EVsMMA−MRC5, indicating that MMA-induced IL-6 secretion through extracellular vesicles is mediated by NF-κB downstream of TGFβ-TAK1-IKK activation, and we saw the same effect with genetic knockdown of TGBR1 and CHUK1 (Fig. 4f, s6d-e). In line with this and our earlier findings demonstrating the necessity of IL-6, all three inhibitors abrogated the ability of EVsMMA−MRC5 to induce IL-6/JAK/STAT3 and TGFβ signaling in A549 tumor cells, along with EMT (Fig. 4g-h, Figure s6f-g). Additionally, all three inhibitors were able to suppress the ability of EVsMMA−MRC5 to increase drug resistance in tumor cells, although this effect was small using SB43152 or TAKinib (Fig. 4i).
Notably, IKK inhibition had a greater effect than both TGFβR inhibition or TAK1 inhibition in reducing IL-6 loading into EVsMMA−MRC5, which also corresponded with a greater effect in suppressing the potency of EVsMMA−MRC5 for promoting EMT and drug resistance in A549 tumor cells (Fig. 4f-i). This suggested that the NF-κB activation downstream of TGFβR signaling was supplemented by a certain level of NF-κB activation independent of TGFβR signaling, together producing the full effect of IL-6 loading into EVsMMA−MRC5 and the full potency of EVsMMA−MRC5 to induce EMT and increase drug resistance in tumor cells.
As increased generation of reactive oxygen species (ROS) has been established to trigger both NF-κB and TGFβ signaling(23, 24), we conjectured that ROS activation of NF-κB both independently and through TGFβR-TAK1-IKK-NF-κB signaling may be at the apex of the MMA signal that induces the CAF phenotype and function. Additionally, pathway enrichment analysis of RNA-seq data showed that the oxidative stress response was upregulated in MMA-treated MRC-5 fibroblasts (Fig. 4a). Indeed, MMA treatment, as well as treatment by EVs from MMA-producing tumor cells (EVsshMUT−A549) of tumor cells increased ROS with peak levels at 6 hours, corresponding to the peak in TGFβ and NF-κB signaling, while also increasing malondialdehyde (MDA), a marker of oxidative stress, over several days (Fig. 4c, Fig. 5a, Figure s7a-c). While ROS induction by MMA was similar to that observed by other ROS inducers, including rotenone, TTFA, and hydrogen peroxide, these other inducers were unable to drive the same level of CAF activation in the fibroblasts (Figure s7d-e). This suggests that MMA may increase ROS through a specific mechanism, or that MMA activates other processes that work with ROS to induce activation of fibroblasts.
Treatment of these fibroblasts with the antioxidants N-acetyl-cysteine (NAC) or SkQ1 effectively inhibited MMA induction of NF-κB and TGFβ signaling, along with the MMA-induced increase in CAF markers and increased IL-6 loading into EVsMMA−MRC5 (Fig. 5b-e). When EVsMMA−MRC5 were collected from fibroblasts that were co-treated with antioxidants, they were no longer able to activate IL6/JAK/STAT3 or TGFβ signaling in A549 tumor cells (Fig. 5f). Consistently, antioxidant treatment of fibroblasts suppressed the ability of EVsMMA−MRC5 to induce the EMT phenotype and increase drug resistance in A549 tumor cells, and reversed the ability of these tumor cells to form metastases in vivo (Fig. 5g-i). Together, our data illustrates a mechanism wherein exposure of fibroblasts to MMA generates ROS and induces oxidative stress, which activates NF-κB and TGFβ signaling. Canonical TGFβ signaling regulates CAF marker expression, while NF-κB signaling, which is activated by ROS both independently of and downstream of TGFβ signaling through TAK1 and IKK, regulates IL-6 association and secretion with vesicles. In tumor cells, IL-6 enriched EVsMMA−MRC5 activates IL-6/JAK/STAT3 and TGFβ signaling, promoting EMT and the acquisition of pro-aggressive traits (Fig. 6).