CAFs secreted exosomes promote metastasis and chemotherapy resistance by enhancing cell stemness and epithelial-mesenchymal transition in colorectal cancer

CAFs and NFs were obtained from colorectal cancer tissues and corresponding normal colorectal mucosa in the study (Additional file 1: Figure S1A). CAFs were positive for alpha-smooth muscle actin (α-SMA), fibroblast activation protein (FAP), fibroblast specific protein 1 (FSP-1) and mesenchymal marker vimentin, while NFs weakly expressed these proteins (Additional file 1: Figure S1B-D). To explore the roles of CAFs during CRC progression, CRC cells were treated with conditioned medium of CAFs (CAFs-CM) or NFs (NFs-CM) prior to in vitro functional experiment. Compared to human CRC SW480, SW620 and LOVO cells treated with NFs-CM, cells treated with CAFs-CM showed increased ability of migration and invasion (Additional file 1: Figure S2A). Moreover, CRC cells treated with CAFs-CM showed increased 5-FU/L-OHP therapy resistance compared to those treated with NFs-CM (Additional file 1: Figure S2B-E, **P < 0.01).

Recent evidence demonstrated that exosomes secreted by a variety of cells were implicated in tumor metastasis and chemotherapy resistance [9, 10]. We speculated that CAFs might exert their effects on CRC cells by secreting exosomes into the culture medium in our experiment system. To verify this conjecture, we isolated exosomes from CAFs-CM and NFs-CM using differential ultracentrifugation (Additional file 1: Figure S3A). Transmission electron microscopy (TEM) revealed cup-shaped structures (Fig. 1a), and nanosight analysis showed a mean particle size of 50–100 nm diameter structures that are typical of exosomes (Additional file 1: Figure S3B). Moreover, exosome markers CD63, CD81, and TSG101 proteins were positively expressed in these vesicles (Fig. 1b, Additional file 1: Figure S3C). After that, we labeled CAFs-secreted exosomes (CAFs-exos) or NFs-secreted exosomes (NFs-exos) with fluorescent dye PKH67 and added them into CRC culture medium to track whether these exosomes could be internalized by CRC cells. As expected, green fluorescence signals were observed in NFs-exos and CAFs-exos treated SW480, SW620 and LOVO cells by laser scanning confocal microscope (LSCM), while no fluorescence signals were observed in PBS treated cells (Fig. 1c, Additional file 1: Figure S3D), suggesting the internalization of PKH67 labeled-exosomes by CRC cells. We further explored the internalization efficiency of CAFs-exos and NFs-exos by CRC cells. CRC cells were incubated with exosomes and the percentage of cells with fluorescence signals at different time points were used to evaluate CRC cells’ internalization efficiency of CAFs-exos and NFs-exos by LSCM. The uptake efficiency of CAFs-exos and NFs-exos by CRC cells increased in a time-dependent manner and more than 90% of SW480, SW620 and LOVO cells were positive for PKH67 fluorescence at 24 h (Additional file 1: Figure S3E). No significant difference was found between the internalization of NFs-exos and CAFs-exos by CRC cells (Additional file 1: Figure S3E, both P > 0.05).

In vitro and in vivo experiments were performed to evaluate the effect of NFs-CM, CAFs-CM, CAFs-exos or CAFs-CM depleted of exosomes treatment on CRC cells. Boyden chamber migration and matrigel invasion assay showed that both CAFs-CM and CAFs-exos significantly increased migration and invasion abilities of CRC cells compared to NFs-CM. However, CRC cells’ migration and invasion abilities were significantly decreased when cells were treated with CAFs-CM depleted of exosomes (Fig. 1d, Additional file 1: Figure S4A, **P < 0.01). The in vivo roles of CAFs-exos were further examined in mice models. CRC cells treated with CAFs-CM and CAFs-exos formed more and larger lung metastasis nodules compared with those treated with NFs-CM. Depletion of exosomes from CAFs-CM suppressed the formation of lung metastasis nodules (Fig. 1e, **P < 0.01).

The chemotherapy sensitivity of CRC cells to clinical drugs 5-FU/L-OHP was also detected in our experimental system. CAFs-CM treated CRC cells had higher abilities of survival and colony formation and lower cell apoptosis compared to NFs-CM-treated cells under 5-FU/L-OHP therapy (Fig. 1f-h, S4B, S4C, *P < 0.05, **P < 0.01). Furthermore, we isolated exosomes from CAFs-CM and found that CAFs-exos treatment could also increase survival and colony formation abilities and decreased cell apoptosis of CRC cells compared to those of NFs-CM-treated cells under 5-FU/L-OHP therapy (Fig. 1f-h, Additional file 1: Figure S4B, C, **P < 0.01). However, depletion of exosomes in CAFs-CM could abolish the protection of CAFs-CM and CAFs-exos treatment on CRC cells under 5-FU/L-OHP therapy (Fig. 1f-h, Additional file 1: Figure S4B, C, **P < 0.01).

Accumulating evidences indicate that cancer stem cells play important roles during metastasis and chemoresistance of many human tumors. In our study, we found that CAFs-exos treated SW480, SW620 and LOVO cells (named SW480^CAFs-exos, SW620^CAFs-exos and LOVO^CAFs-exos cells) formed larger spheres compared to cells treated with NFs-exos (named SW480^NFs-exos, SW620^NFs-exos and LOVO^NFs-exos cells) (Additional file 1: Figure S4D, **P < 0.01). Moreover, SW480^CAFs-exos, SW620^CAFs-exos and LOVO^CAFs-exos cells showed higher proportion of CRC^CD133+/CD44+ cells compared to SW480^NFs-exos, SW620^NFs-exos and LOVO^NFs-exos cells (Fig. 1i, Additional file 1: Figure S4E). Consistently, cancer stemness markers CD133, CD44, and OCT4 were obviously increased in SW480^CAFs-exos, SW620^CAFs-exos and LOVO^CAFs-exos cells by real-time PCR, western blot and immunofluorescence assays (Fig. 1j, k, Additional file 1: Figure S4F), indicating the enhancement of CRC cell stemness phenotypes by CAFs-exos. We also found that SW480^CAFs-exos, SW620^CAFs-exos and LOVO^CAFs-exos cells expressed higher mesenchymal markers (N-cadherin and Vimentin) and lower epithelial markers (E-cadherin) compared to SW480^NFs-exos, SW620^NFs-exos and LOVO^NFs-exos cells, suggesting the induction of epithelial to mesenchymal transition in CRC cells by CAFs-exos (Fig. 1l). These results reveal that CAFs secreted exosomes can promote metastasis and 5-FU/L-OHP resistance, possibly by enhancing stemness and EMT in CRC.

Exosomes are bilateral structures that could shuttle proteins and microRNAs (miRNAs) into adjacent cells within the microenvironment. To investigate how CAF-exos exert their effects on CRC cells, NFs-exos, CAFs-exos, SW480^NFs-exos, and SW480^CAFs-exos cells were sequenced using miRNA microarray assay (Fig. 2a). Among the differentially expressed miRNAs, the levels of miR-92a-3p, miR-181d, miR-221, miR-125b, miR-185, and miR-625 were significantly increased in CAFs-exos and SW480^CAFs-exos cells, which were further validated by real-time PCR (Fig. 2b). We focused on miR-92a-3p in the following experiments because its expression was highest in CAFs-exos and SW480^CAFs-exos cells and the underlying mechanisms of miR-92a-3p in the regulation of CRC aggressiveness and chemotherapy resistance remain to be characterized. The endogenous miR-92a-3p level was detected in normal colorectal mucosa cell line NCM460, seven human CRC cell lines, human CRC tissues and matched normal colorectal mucosa, CAFs, NFs, CAFs-exos and NFs-exos. Real-time PCR results showed that the level of miR-92a-3p was significantly higher in CAFs and CAF-exos compared to NFs, NFs-exos, CRC tissues, cell lines, and was lowest in NCM460 cells (Fig. 2c, **P < 0.01).

To determine whether CAFs-exos increase miR-92a-3p level in CRC cells, the expression of miR-92a-3p was determined in CRC cells incubated with CAFs-exos or NFs-exos. MiR-92a-3p level was significantly increased in SW480^CAFs-exos, SW620^CAFs-exos and LOVO^CAFs-exos cells, but not in SW480^NFs-exos, SW620^NFs-exos and LOVO^NFs-exos cells, suggesting CAFs-exos were associated with the increase of miR-92a-3p in CRC cells (Fig. 2d, **P < 0.01). To determine how miR-92a-3p was increased in CRC cells, CRC cells were incubated with CAFs-exos or NFs-exos. Real-time PCR assay showed that miR-92a-3p was significantly increased in CRC cells incubated with CAFs-exos for 24 h (Fig. 2e, **P < 0.01). Moreover, we detected the level of pre-miR-92a (precursor of miR-92a-3p) and found it was unchanged in CRC cells incubated with NFs-exos or CAFs-exos (Fig. 2f, P > 0.05), suggesting that the increase of miR-92a-3p in CRC cells was not the result of miRNA endogenous synthesis but more likely a direct transfer by CAFs-exos.

Further efforts were made to explore whether the increase of miR-92a-3p in CRC cells was caused by direct exosomal transfer from CAFs to CRC cells. CRC cells were firstly transfected with miR-92a-3p sponge or miR-NC prior to incubation with NFs-exos or CAFs-exos (Fig. 2g). MiR-92a-3p was significantly decreased in miR-92a-3p-sponge transfected cells. However, the level of miR-92a-3p in these cells was obviously increased after incubation with CAFs-exos rather than NFs-exos (Fig. 2g, **P < 0.01). We further co-cultured SW480 cells with control CAFs or CAFs that were transiently transfected with fluorescein amidite (FAM)-tagged miR-92a-3p for 24 h using a transwell co-culture system (Fig. 2h). Interestingly, green fluorescence signals of miR-92a-3p-FAM were observed in SW480 cells co-cultured with miR-92a-3p-FAM expressing CAFs, rather than in SW480 cells co-cultured with control CAFs (Fig. 2h). To further determine whether the transfer of miR-92a-3p is mediated by exosomes, SW480 cells were incubated with exosomes derived from the conditioned medium of miR-92a-3p-FAM expressing CAFs or control CAFs. Consistently, fluorescently labelled green signals were also observed in SW480 cells incubated with exosomes secreted by miR-92a-3p-FAM expressing CAFs, while no fluorescent signals were observed in SW480 cells incubated with exosomes secreted by control CAFs (Fig. 2i), confirming that miR-92a-3p was directly transferred from CAFs to CRC cells via exosomes.

We then set out to explore the role of CAFs-exosomal miR-92a-3p in the aggressiveness of CRC cells. Boyden chamber assay showed that the number of migrated and invaded SW480^CAFs-exos, SW620^CAFs-exos and LOVO^CAFs-exos cells was increased compared to SW480^NFs-exos, SW620^NFs-exos and LOVO^NFs-exos cells, while transfection of miR-92a-3p-sponge into CAFs-exos could significantly reduce the migration and invasion of SW480^CAFs-exos, SW620^CAFs-exos and LOVO^CAFs-exos cells. However, reintroduction of miR-92a-3p mimics into CRC cells could reverse miR-92a-3p-sponge mediated inhibition of cell migration and invasion (Fig. 3a, Additional file 1: Figure S5A, B, ** P < 0.01), suggesting CAFs-secreted exosomal miR-92a-3p could significantly enhance cell migration and invasion of CRC.

To further explore the effect of CAFs-secreted miR-92a-3p in vivo, SW620 cells were firstly injected in the flank of mice to form subcutaneous tumors. Tumors were then resected, minced, and transplanted to the mucosa of ileocecal junction. Macroscopy results showed that tumors were formed in the ileocecal junction and liver. Moreover, the number of liver metastases in mice injected with SW620^CAFs-exos cells was more than those injected with SW620^NFs-exos cells, while injection of SW620 cells treated with CAFs-exos transfected with miR-92a-3p-sponge could reduce the number of liver metastases compared to control (Fig. 3b, **P < 0.01).

The effect of CAFs-exosomal miR-92a-3p on CRC cells response to 5-FU/L-OHP therapy was also determined in vitro and in vivo. SW480^CAFs-exos, SW620^CAFs-exos and LOVO^CAFs-exos cells showed higher survival and colony formation abilities and lower cell apoptosis compared to cells treated with NFs-exos. Suppressing miR-92a-3p in CAFs-exos could significantly inhibit survival and colony formation and induce cell apoptosis of SW480^CAFs-exos, SW620^CAFs-exos and LOVO^CAFs-exos cells. Re-introduction of miR-92a-3p mimics into CRC cells reversed these effects (Fig. 3c, d, Additional file 1: Figure S5C, D, *P < 0.05, **P < 0.01). In vivo experiment showed that the size of subcutaneous tumors derived from SW480 cells treated with PBS was the largest, while nude mice injected with SW480 cells under 5-FU/L-OHP therapy formed the smallest tumors. Importantly, tumors derived from SW480^{CAFs-exos-miR-92a-3p-sponge} cells were smaller than CAFs-exos groups and larger than NFs-exos group under 5-FU/L-OHP therapy (Fig. 3e, Additional file 1: Figure S5E, **P < 0.01,). Moreover, tumors derived from SW480^CAFs-exos cells yielded decreased apoptosis compared to those of SW480^NFs-exos group. Inhibition of miR-92a-3p in CAFs-exos could induce cell apoptosis (Fig. 3f, *P < 0.05, ** P < 0.01).

In addition, SW480^CAFs-exos, SW620^CAFs-exos and LOVO^CAFs-exos cells formed more and larger spheres compared to those formed by SW480^NFs-exos, SW620^NFs-exos and LOVO^NFs-exos cells. Transfection of CAFs-exos with miR-92a-3p-sponge significantly reduced the number and size of spheres, while re-introduction of miR-92a-3p mimics in CRC cells could increase spheres formation (Fig. 3g, Additional file 1: Figure S6A, **P < 0.01). Consistently, SW480^CAFs-exos, SW620^CAFs-exos and LOVO^CAFs-exos cells showed higher proportion of CRC^CD133+/CD44+ cells compared to SW480^NFs-exos, SW620^NFs-exos and LOVO^NFs-exos cells. Suppressing miR-92a-3p in CAFs-exos decreased the proportion of CRC^CD133+/CD44+ cells while reintroduction of miR-92a-3p increased CRC^CD133+/CD44+ cell proportion (Fig. 3h, Additional file 1: Figure S6B, *** P < 0.001). Moreover, cell stemness markers CD133, CD44 and OCT4 were obviously increased in SW480^CAFs-exos, SW620^CAFs-exos and LOVO^CAFs-exos cells (Fig. 3i, j, S6C, **P < 0.01, ***P < 0.001). Transfection of CAFs-exos with miR-92a-3p-sponge could decrease CD133, CD44 and OCT4, while re-introduction of miR-92a-3p mimics in CRC cells could increase these proteins in CRC cells (Fig. 3i, j, Additional file 1: Figure S6C). Moreover, SW480^CAFs-exos, SW620^CAFs-exos and LOVO^CAFs-exos cells expressed lower epithelial marker E-cadherin but higher mesenchymal markers N-cadherin and vimentin compared to SW480^NFs-exos, SW620^NFs-exos and LOVO^NFs-exos cells. However, suppressing miR-92a-3p in CAFs-exos increased E-cadherin and decreased N-cadherin and vimentin in CRC cells. Reintroduction of miR-92a-3p in CRC cells could largely enhance the N-cadherin and vimentin and inhibit E-cadherin expression in CRC cells (Fig. 3k). These above data demonstrate that CAFs promote metastasis and 5-FU/L-OHP resistance and enhance stemness and EMT in CRC by transferring exosomal miR-92a-3p to CRC cells.

To screen the downstream targets of miR-92a-3p, we utilized five bioinformatics and online prediction database (DIANAmT, miRanda, miRDB, miRWalk, Targetscan) and screened FBXW7 and MOAP1 as potential downstream targets of miR-92a-3p with highest predictive values. The TargetScan algorithm showed the potential complementarity sequence in the 3’UTR of FBXW7 and MOAP1 to the seed sequence of miR-92a-3p (Fig. 4a). To substantiate the site-specific repression of miR-92a-3p on FBXW7 and MOAP1, we constructed mutated FBXW7 3′-UTR and MOAP1 3′-UTR luciferase reporter. Dual-luciferase activity assay showed that the luciferase activity of FBXW7 or MOAP1 with wild type 3’UTR was significantly suppressed in miR-92a-3p expressing HEK293A, SW480, SW620 and LOVO cells (Fig. 4b, **P < 0.01). In contrast, the luciferase activity of FBXW7 or MOAP1 with 3’UTR mutation was not changed (Fig. 4c, P > 0.05). Moreover, FBXW7 and MOAP1 were significantly decreased in miR-92a-3p expressing CRC cells compared to Mock cells (Fig. 4d, e, ** P < 0.01). In addition, both FBXW7 and MOAP1 proteins were suppressed in SW480^CAFs-exos, SW620^CAFs-exos and LOVO^CAFs-exos cells compared to SW480^NFs-exos, SW620^NFs-exos and LOVO^NFs-exos cells while re-introduction of FBXW7 and MOAP1 in CRC cells could increase their levels (Fig. 4f). These results indicate that FBXW7 and MOAP1 are downstream targets of miR-92a-3p in CRC cells.

Having validated FBXW7 and MOAP1 as downstream targets of miR-92a-3p, we further explored their roles in miR-92a-3p-mediated regulation of CRC. SW480, SW620 and LOVO cells were transfected with FBXW7 and MOAP1 plasmids separately. Boyden chamber assay showed that re-introduction of FBXW7 could inhibit miR-92a-3p mediated promotion of migration and invasion in CRC cells (Fig. 4g, h, Additional file 1: Figure S7A, B, ** P < 0.01). In addition, re-expression of FBXW7 and MOAP1 significantly reduced colonies formation and survival abilities of CRC cells under 5-FU/L-OHP treatment (Fig. 4i, k, Additional file 1: Figure S7C, D, * P < 0.05, ** P < 0.01). Overexpression of FBXW7 inhibited sphere formation, CD133, CD44, OCT4, cyclinD1, c-myc, N-cadherin and increased E-cadherin in miR-92a-3p expressing CRC cells, leading to inhibition of cell stemness, proliferation and EMT phenotypes (Fig. 4l-n, Additional file 1: Figure S7E, F, ** P < 0.01). Re-introduction of MOAP1 induced the expression of apoptotic proteins BAX and cleaved caspase 3 in miR-92a-3p expressing CRC cells (Fig. 4o), thus promoting cancer cell apoptosis.

To further explore the effect of FBXW7 and MOAP1 on metastasis and 5-FU/L-OHP resistance of CRC, we constructed SW620/Mock, SW620/miR-92a-3p, SW620/miR-92a-3p/FBXW7, SW620/miR-92a-3p/MOAP1 cells by lentivirus infection. Tail vein injection and subcapsular injection of the spleen assay showed that metastatic nodules formed in lung and liver by SW620/miR-92a-3p cells were significantly increased compared to those formed by SW620/Mock cells, while the formation of metastatic nodule by SW620/miR-92a-3p/FBXW7 cells was obviously decreased compared to SW620/miR-92a-3p cells (Fig. 5a, b, ** P < 0.01).

Moreover, the size of subcutaneous tumors formed by SW480/miR-92a-3p was significantly larger than that formed by SW480/Mock cells (Fig. 5c, Additional file 1: Figure S8A, B, ** P < 0.01) under 5-FU/L-OHP therapy. However, re-introduction of FBXW7 into SW480/miR-92a-3p cells could significantly decrease SW480/miR-92a-3p cells derived subcutaneous tumors (Fig. 5c, Additional file 1: Figure S8A, ** P < 0.01). In addition, SW480/miR-92a-3p/FBXW7 cells derived tumors showed lower ki67 proliferation index, increased TUNEL positive apoptotic cells, and higher level of caspase3 apoptotic proteins by IHC and TUNEL assay (Fig. 5d-f, Additional file 1: Figure S8C, E, G, ** P < 0.01). Furthermore, we found consistent in vivo results in tumors formed by SW480/Mock, SW480/miR-92a-3p, and SW480/miR-92a-3p/MOAP1 expressing cells (Fig. 5g-j, Additional file 1: Figure S8B, D, F, H, ** P < 0.01). Thus, these results showed that FBXW7 and MOAP1 could attenuate miR-92a-3p mediated metastasis and chemotherapy resistance of CRC cells in vivo.

To determine the potential mechanisms underlying the role of FBXW7 in abrogating the tumor-promoting effects of miR-92a-3p, we investigated the Wnt/β-catenin pathway. The introduction of exogenous miR-92a-3p significantly increased β-catenin expression in the nuclear fractions in SW480, SW620 and LOVO cells, whereas re-introduction of FBXW7 in miR-92a-3p expressing cells decreased the expression of β-catenin in the nuclear fractions (Fig. 6a, b). Moreover, overexpression of miR-92a-3p increased CD133, CD44, OCT4, N-cadherin, MMP7, MMP9 and decreased E-cadherin in CRC cells. In contrast, overexpression of FBXW7 in miR-92a-3p expressing cells showed the opposite effect (Fig. 6a). Furthermore, we found that FBXW7 inhibited β-catenin in the nucleus fractions by promoting β-catenin ubiquitination and degradation in CRC cells (Fig. 6b-e). We also investigated the mechanism of MOAP1 in attenuating the effects of miR-92a-3p on CRC chemotherapy resistance. Overexpression of miR-92a-3p suppressed MOAP1 as well as Bax, cytochrome c, caspase9, caspase3 in CRC cells under 5-FU/L-OHP therapy (Fig. 6f). Re-expression of MOAP1 in miR-92a-3p expressing cells showed less expression of apoptotic proteins and cytochrome c release to the cytoplasm (Fig. 6f, g).

To investigate the clinical values of miR-92a-3p for CRC, we detected its expression in 40 cases of CRC tissues and corresponding normal mucosa using real-time PCR. The expression of miR-92a-3p was higher in CRC tissues compared to normal mucosa. Moreover, miR-92a-3p was higher in CRC with metastasis compared to CRC without metastasis (Fig. 7a, *** P < 0.001). The expressions of FBXW7 and MOAP1 were lower in CRC tissues than in normal mucosa (Fig. 7b, c). Correlation analysis showed that miR-92a-3p was negatively correlated with FBXW7 and MOAP1 in CRC with (Fig. 7d) or without metastasis (Fig. 7e). The above results further validate that FBXW7 and MOAP1 are both downstream target genes of miR-92a-3p.

To ascertain whether exosomal miR-92a-3p is associated with CRC metastasis, we detected the level of exosomal miR-92a-3p in the serum in healthy persons, CRC patients with metastasis and those without metastasis. Results showed that exosomal miR-92a-3p level was highest in CRC patients with metastasis, gradually decreased in CRC patients without metastasis, and lowest in the healthy people (Fig. 7f). Moreover, we detected the level of exosomal miR-92a-3p in the serum of 18 cases of 5-FU/L-OHP sensitive and resistant CRC patients. Results showed that exosomal miR-92a-3p was significantly higher in 5-FU/L-OHP resistant CRC patients compared to that in 5-FU/L-OHP sensitive CRC patients (Fig. 7g). These results strongly suggest that exosomal miR-92a-3p may be a predictor of CRC metastasis and chemotherapy resistance.