A novel protective role for microRNA-3135b in Golgi apparatus fragmentation induced by chemotherapy via GOLPH3/AKT1/mTOR axis in colorectal cancer cells

Chemotherapy activates a novel cytoplasmic DNA damage response resulting in Golgi apparatus fragmentation and cancer cell survival. This mechanism is regulated by Golgi phosphoprotein-3 (GOLPH3)/Myo18A/F-actin axis. Analyzing the functions of miR-3135b, a small non-coding RNA with unknown functions, we found that its forced overexpression attenuates the Golgi apparatus fragmentation induced by chemotherapeutic drugs in colorectal cancer (CRC) cells. First, we found that miR-3135b is downregulated in CRC cell lines and clinical tumors. Bioinformatic predictions showed that miR-3135b could be regulating protein-encoding genes involved in cell survival, resistance to chemotherapy, and Golgi dynamics. In agreement, ectopic transfection of miR-3135b in HCT-15 cancer cells significantly inhibited cell proliferation, sensitized cells to 5-fluoruracil (5-FU), and promoted late apoptosis and necrosis. Also, miR-3135b overexpression impaired the cell cycle progression in HCT-15 and SW-480 cancer cells. Because GOLPH3, a gene involved in maintenance of Golgi structure, was predicted as a potential target of miR-3135b, we studied their functional relationships in response to DNA damage induced by chemotherapy. Immunofluorescence and cellular ultrastructure experiments using antibodies against TGN38 protein, a trans-Golgi network marker, showed that 5-FU and doxorubicin treatments result in an apoptosis-independent stacks dispersal of the Golgi ribbon structure in both HCT-15 and SW-480 cells. Remarkably, these cellular effects were dramatically hindered by transfection of miR-3135b mimics. In addition, our functional studies confirmed that miR-3135b binds to the 3′-UTR of GOLPH3 proto-oncogene, and also reduces the levels of p-AKT1 (Ser473) and p-mTOR (Ser2448) signaling transducers, which are key in cell survival and autophagy activation. Moreover, we found that after treatment with 5-FU, TGN38 factor coimmunolocalizes with beclin-1 autophagic protein in discrete structures associated with the fragmented Golgi, suggesting that the activation of pro-survival autophagy is linked to loss of Golgi integrity. These cellular effects in autophagy and Golgi dispersal were reversed by miR-3135b. In summary, we provided experimental evidence suggesting for the first time a novel role for miR-3135b in the protection of chemotherapy-induced Golgi fragmentation via GOLPH3/AKT1/mTOR axis and protective autophagy in colorectal cancer cells.


MicroRNA-3135b expression is repressed in colorectal cancer cells and tumors. To initiate
the study of miR-3135b, a non-coding RNA located in chromosome 6p21.32 with no previous involvement in human cancers, we analyzed its expression levels in three CRC cell lines and normal colon cells using stem loop qRT-PCR assays. Data showed that miR-3135b levels are significantly downregulated in HCT-15, SW-480 and Caco2 cancer cells in comparison with normal CRL-1790 colon cells (Fig. 1A). To further investigate the clinical relevance of miR-3135b, we also analyzed its expression during the early and late stages of carcinogenesis using previous published data obtained from a miRNAs profiling of 1893 normal/colorectal carcinoma-paired samples, and 290 adenoma tissues samples deposited in NCBI GEO database (GSE115513) 24 . Results showed a slight but significant downregulation of miR-3135b expression in CRC stages 1 and 3 of tumor progression relative to normal colon tissues, whereas no significant changes in miR-3135b levels were detected between normal tissues and CRC stages 2 and 4 (Fig. 1B). Also, we evaluated if miR-3135b expression could be regulated by chemotherapeutic drugs. Data evidenced that the low miR-3135b expression is even more suppressed after DNA damage induced by 5-FU and doxorubicin treatments in comparison to untreated controls in both HCT-15 and SW-480 cancer cells (Fig. 1C,D).

MicroRNA 3135b is predicted to regulate genes involved in Golgi apparatus integrity.
To investigate the functions of miR-3135b, we searched for its potential gene targets and associated signaling pathways by performing a bioinformatic analysis using TargetScan and miRmap softwares. Results revealed that a number of genes involved in Golgi structure and functions, such as GOLPH3, MYO18A, SET1, ST6GAL1 and ST6GAL2, contain potential miR-3135b binding sites at their 3′-UTRs, suggesting that they could be regulated at the posttranscriptional level (Fig. 2). Moreover, several protein-encoding genes involved in apoptosis, cell  (24) deposited in NCBI GEO database (GSE115513). Expression analysis was carried out using the Limma package in R (https :// www.bioco nduct or.org/packa ges/relea se/bioc/html/limma .html) which employs gene-wise linear models and empirical Bayes method. Multiple testing correction by considering an FDR < 0.05 as statistically significant was used. Horizontal lines in the middle of box plot denote the median. Lines extending from the boxes indicate variability outside the upper and lower quartiles. NS, Non-significant. (C-D) qRT-PCR assays for miR-3135b expression in (C) HCT-15 and (D) SW-480 colorectal cancer cell lines treated with doxorubicin and 5-FU compared to baseline levels. Data were normalized with the endogenous small-nucleolar RNU44. Statistical significance was determined by Student's t-test and p < 0.05 were considered statistically significant. MicroRNA-3135b impairs proliferation, cell cycle progression and promotes late apoptosis and necrosis. Next, we wondered if miR-3135b could regulate cell proliferation, and apoptosis, and sensitizes CRC cells to 5-FU therapy. Results showed that proliferation of HCT-15 cells transfected with miR-3135b mimic is significantly decreased in comparison with non-transfected and scramble transfected control cells at 12, 24 and 36 h (Fig. 3A). Next, we calculated the half-maximal inhibitory concentrations (IC 50 ) for 5-FU in HCT-15 cells using MTT assays, and performed combined treatments using miR-3135b mimic and 5-FU. Data evidenced that restoration of miR-3135b slightly but significantly sensitizes HCT-15 cells to 5-FU treatment in comparison with 5-FU monotherapy after 24 h and 36 h (Fig. 3A). To evaluate the effects of miR-3135b on early and late apoptosis, we performed double staining of cells with annexin V-FITC (AN) and propidium iodide (PI). The proportion of cells that exhibit AN−/PI−, AN+/PI−, AN−/PI+ and AN+/PI+ staining representing normal, early apoptotic, necrotic and late apoptotic cell populations, respectively, was determined using fluorescence activated cell sorting. Data showed a significant increase in the percentage of apoptotic cells after intervention with miR-3135b mimic in comparison to mock and scramble transfected controls (Fig. 3B,C). Moreover, a dramatic increment of cells in late apoptosis and necrosis after miR-3135b plus 5-FU combined treatments was found (Fig. 3B, C). Next, we evaluated the effects miR-3135b and combined treatments in cell cycle progression of CRC cell lines. Data indicated that transfection of miR-3135b, alone or in presence of 5-FU results in a significant increase of the percentage of cells in G1 phase and a diminution of cells in S1 phase relative to untreated control, suggesting that inhibition of cell proliferation could be due, at least in part, to cell cycle arrest (Fig. 4A,B). This effect was more pronounced in HCT-15 cell line in comparison to SW-480 cancer cells.
MicroRNA-3135b attenuates Golgi disruption in response to therapy. The induction of DNA double strand breaks (DSBs) by chemotherapeutic drugs has been shown to activate a cytoplasmic DNA damage response, i.e. the GOLPH3/MYO18A/F-actin pathway, to promote Golgi apparatus dispersal and cancer cells survival 5 . Our bioinformatics analyses identified the GOLPH3 gene as a potential target of miR-3135b   www.nature.com/scientificreports/ ( Fig. 2). GOLPH3 is a proto-oncogene involved in the maintenance of Golgi architecture and vesicular trafficking system, as well as in the activation of AKT and mTOR transducers involved in eukaryotic cells survival. Therefore, we decided to evaluate the potential role of miR-3135b in the regulation of GOLPH3 and Golgi apparatus structure in response to chemotherapy treatments. First, we examined the levels of phosphorylated H2A.X (Ser139) histone, an early DNA damage marker, in HCT-15 cells exposed to 5-FU. As expected, immunoblotting results confirmed a significant increase in phospho-H2A.X (Ser139) levels after 5-FU intervention in comparison to non-treated cells, indicative of the presence of DNA DSBs and DNA damage (Fig. 5A). Next, we established the model of Golgi fragmentation using treatments with 5-FU and doxorubicin compounds and tracked the morphologic alterations in the organelle structure using antibodies against TGN38, a trans-Golgi network protein marker, followed by immunofluorescence and confocal microscopy assays in CRC cells. Data showed that TGN38 signal is located at the trans-Golgi network in ring-shaped structures around the nucleus in control HCT-15 cells (Fig. 5B). Quantitative analysis of TGN38-immunostained cells indicated that both 5-FU and doxorubicin drugs significantly alter the Golgi apparatus structure inducing the stacks dispersal of the perinuclear ribbon ( Fig. 5B, middle panels). These morphologic changes were accompanied by a significant increase in the relative Golgi area of HCT-15 cells treated with 5-FU and doxorubicin relative to non-treated control (Fig. 5B,D). Interestingly, the transfection of miR-3135b mimics in combination with the drugs dramatically hindered the Golgi fragmentation induced by 5-FU or doxorubicin alone and significantly reduced the Golgi area close to untreated control (Fig. 5B, D). These findings were confirmed in a second CRC cell line, SW-480. Likewise, the TGN38 signal was found at the trans-Golgi network surrounding the nucleus in SW-480 cells (Fig. 5C), in a very similar stain pattern as found in HCT-15 cells (Fig. 5B). The fragmentation of the Golgi structure induced by 5-FU and doxorubicin was also significantly impaired after transfection of miR-3135b mimic in comparison with non-transfected control in SW-480 cells (Fig. 5C,E). These data demonstrate that drug treatments induced significant alterations of the Golgi ribbon ultrastructure depicted as stacks dispersal, and highlight the protective effect of miR-3135b on Golgi integrity in two different CRC cell lines. To further confirm the immunofluorescence findings, a cellular ultrastructure analysis of Golgi apparatus using transmission electronic microscopy (TEM) was performed in the same conditions as previously described. TEM results confirmed the presence of significant changes in the organelle ultrastructure associated with stacks dispersal of the Golgi ribbon structure after 5-FU treatments, and corroborated the protective role of miR-3135b on Golgi integrity in HCT-15 cells (Fig. 5F).

Golgi dispersal occurs independently of apoptosis in CRC cells. To rule out the possibility that
Golgi fragmentation could be due to apoptosis activation, we immunodetected the caspase-3 protein in HCT-15 cells using confocal microscopy. As expected, after treatment with 5-FU and doxorubicin, a subpopulation . Moreover, most non-apoptotic cells exhibited Golgi fragmentation uniformly dispersed throughout the cytoplasm, which was accompanied with an increase in relative Golgi area (Fig. 6A,B). These data strongly suggest that the organelle dispersion was predominantly independent of apoptosis as previously reported in cervical cancer HeLa cells 5 . Next, we also discarded the possibility that expression of TGN38 protein used as trans-Golgi network marker could be targeted by miR-3135b. For this, we performed Western blot assays to detect TGN38 levels in controls and transfected cells with miR-3135b mimics. After densitometric analysis of immunodetected bands, we did not find any significant changes in the expression of the TGN38 protein following the transfection of miR-3135b mimics at 30 and 60 nM concentrations (Fig. 6C).
MicroRNA-3135b targets the GOLPH3/AKT/mTOR axis. To gain insights into the molecular mechanism underlying the protective role of miR-3135b on Golgi dispersal induced by drugs, we investigated the expression levels of GOLPH3, p-AKT1 (Ser473) and p-mTOR (Ser2448) proteins by Western blot assays using non-treated, mock, and scramble transfected HCT-15 cells, as well as cells transfected with miR-3135b mimic alone or in combination with 5-FU. Results showed that the expression of GOLPH3, and p-AKT1 (Ser473) and p-mTOR (Ser2448) proteins does not present significant variations among non-treated, mock and scramble controls (Fig. 7A,B). In contrast, the p-AKT1 (Ser473) and p-mTOR (Ser2448) levels were significantly repressed after miR-3135b mimic transfection at 30 nM and 60 nM, whereas GOLPH3 expression was inhibited only www.nature.com/scientificreports/ at 60 mM (Fig. 7A,B). Treatment with 5-FU resulted in a significant increase of GOLPH3, and a reduction of p-AKT1 (Ser473) and p-mTOR (Ser2448). No significant changes in α-tubulin expression used as control were found in treated and control cells. In light of these data, it was reliable to propose that miR-3135b may regulate Golgi dispersal caused by chemotherapy through direct binding to its cognate targets. Bioinformatic analysis predicted that 3′-UTR of GOLPH3 gene contains a potential miR-3135b binding site (Fig. 7C). Therefore, to define if miR-3135b can exert a posttranscriptional repression of GOLPH3 gene, we performed luciferase reporter assays. A DNA fragment corresponding to 3′-UTR of GOLPH3 was cloned downstream of the luciferase-coding region of pmiR-LUC vector (Fig. 7C). In addition, a mutated version of the predicted miR-3135b binding site was included as control. Thereafter, wild-type and mutated pmiR-LUC-GOLPH3 and pmiR-LUC control constructs were transfected into HCT-15 cells and luciferase activity was analyzed after 24 h. Data showed that co-transfection of recombinant plasmids and miR-3135b mimics result in a significantly reduction of luciferase activity in comparison to controls (Fig. 7D). In contrast, when the mutated sequence of GOLPH3 3′-UTR was assayed, no significant changes in luciferase activity were observed suggesting that miR-3135b binding was specific.
Golgi dispersal is associated to autophagy activation. Recent studies showed that endo-membranous organelles such as endoplasmic reticulum and the Golgi apparatus are essential for the assemblage and production of autophagic-vesicles during the early stages of autophagy activation 10 . In addition, autophagy triggering depends on a functional mTOR pathway. As we observed that dispersal of Golgi in cancer cells treated with 5-FU and doxorubicin was associated with activation of mTOR signaling (and hindered by miR-3135b), we hypothesized that a pro-survival autophagic pathway could be in turn activated in response to chemotherapy, which may result in recycling of Golgi fragments allowing cell survival of cancer cells. To test this hypothesis, we performed immunofluorescence assays using specific antibodies against TGN38 and beclin-1 autophagy protein in HCT-15 and SW-480 cells treated with 5-FU. Data showed that beclin-1 signal is very weak in both non-treated HCT-15 and SW-480 control cells. Remarkably, after treatment with 5-FU, the beclin-1 signal was www.nature.com/scientificreports/ significantly increased and located around the nucleus in a pattern overlapping with TGN38 and associated to fragmented Golgi, suggesting that ongoing autophagy could be occurring in SW-480 and HCT-15 cells (Fig. 8).
The co-immunolocalization of both proteins was better observed in HCT-15 than in SW-480 cells (Fig. 8B). Moreover, the restoration of miR-3135b in 5-FU treated cells led to a significant reduction in Golgi fragmentation and the consequent loss of beclin-1 signal, suggesting the inhibition of autophagy which was accompanied with the recovery of Golgi structure integrity in HCT-15 and SW-480 cells (Fig. 8A-D).

Discussion
During the last decades, the search for genes and signaling pathways responsible for resistance to chemotherapy in human cancers has been exhaustive. However, it was not expected that a cytoplasmic DNA damage response mechanism operates in response to cytotoxic therapies in cancer cells, changing our assumption that only the nucleus orchestrates the DNA damage response. The effect of DNA damage on Golgi integrity has been largely neglected in past studies as it was thought as a consequence of apoptosis generated by extensive DNA damage. However, a report of Farber-Katz et al., demonstrated that Golgi fragmentation occurs after DNA damage caused by chemotherapeutic drugs as a mechanisms based on the activation of DNA-PK/GOLPH3/MYO18a pathway to allow cancer cells survival, which is largely unrelated to apoptosis 5 . A very limited number of reports have studied the dispersion of Golgi apparatus as a novel survival mechanism involved in the response to DNA damage in cancer cells [ 5,16 , this report]. Moreover, the involvement of microRNAs functioning as either oncogenes or tumor suppressors in this phenomenon is largely unknown 25 . Inspired by these facts, here we investigated the functional relationships between miR-3135b and the Golgi apparatus dispersal in response to therapy with DNA-damaging drugs. Our data showed that miR-3135b is downregulated in colorectal cancer cells and clinical tumors suggesting that it could function as tumor suppressor gene (Fig. 1). A previous study using a genome-wide profiling of www.nature.com/scientificreports/ microRNAs evidenced that miR-3135b is repressed in stool samples of patients with colorectal adenocarcinoma 26 , but no functional analysis of this particular microRNA was performed. On the other hand, several studies indicated that microRNAs might regulate GOLPH3 expression [27][28][29][30] . Here, we found that overexpression of miR-3135b inhibits cell proliferation, and also has a synergistic effect with 5-FU treatment in activating apoptosis, perhaps through repression of GOLPH3, which is consistent with previous studies that have shown that the inhibition of GOLPH3 by miR-126 promotes apoptosis and inhibits cell proliferation in esophageal squamous cell carcinoma 27 . Likewise, miR-134 suppressed cell proliferation by direct binding to the 3′-UTR of GOLPH3 in gastric cancer cells 28 . Cell cycle assays showed that miR-3135b+5-FU treatment increases the percentage of cells in G1 phase and decreases the ratio of S phase cells in both cell lines in comparison with control cells. Here, we also evaluated the functional relationships between miR-3135b and AKT1/mTOR signaling which is active in almost all cancers. These pathways have been largely associated to cell survival, resistance to chemotherapy and metastasis in diverse types of malignancies 31,32 . For instance, it was reported that miR-10a significantly inhibits cell proliferation and migration, and promotes apoptosis of MCF-7 breast cancer cells, which was attributed to suppression of phosphorylation levels of AKT and mTOR 33 . Our data showed that overexpression of miR-3135b in HCT-15 cells results in reduction of p-AKT1 (Ser473) and p-mTOR (Ser2448) levels, which may explain, in part, the inhibition of cell proliferation. Interestingly, and consistent with AKT1/mTOR pathways activation, other reports showed that the dispersion of the Golgi apparatus via GOLPH3 has the ability to enhance the hypersialylation of RTK receptors by promoting the transduction of the PI3K/AKT/mTOR signaling 34,35 .
On the other hand, it was shown that damage to DNA with doxorubicin and camptothecin triggers the dispersion of the Golgi through the cytoplasm and activates the protein kinase of DNA damage and DNA-PK GOLPH3/MYO18A axis 5 . We also found that exposure to 5-FU induces fragmentation of Golgi apparatus, which was hindered by miR-3135b in HCT-15, and SW-480 colorectal cancer cells. Thus, by simulating a cellular environment of DNA damage using 5-FU, we found novel findings on the regulation of Golgi dispersal by miR-3135b, maybe by targeting GOLPH3, adding a piece to the puzzle. These findings are in agreement with data showing that inhibition of GOLPH3 with an interfering RNA also prevents dispersion of the Golgi apparatus in response to doxorubicin 5,30,36 .
Interestingly, we found a potential functional link between Golgi dispersal, autophagy activation and miR3135b. It is well known that endoplasmic reticulum and the Golgi apparatus are pivotal for the assemblage of autophagic-vesicles 10 . Our data showed that the autophagy protein beclin-1 is located at vesicles associated with the fragmented Golgi, suggesting that ongoing autophagy could be occurring in response to drug treatment to allow cell survival (Fig. 8). We found that cells with excessive Golgi dispersion exhibit strong staining for beclin-1; conversely, cells that maintained the structure of the Golgi apparatus presented a low beclin-1 signal, which could indicate that the activation of autophagy has a direct correlation with the loss of the Golgi structure and mTOR activation in cancer cells. A similar mechanism on the dependence of Golgi ribbon structure, and autophagy activation via mTOR has also been documented in previous studies 37 .
We are aware that the present report has limitations that should be addressed in future works: (1) this is a basic research study which must be further confirmed by the analysis of miR-3135b and GOLPH3/AKT1/mTOR axis www.nature.com/scientificreports/ molecules in biopsies from CRC patients in order to evaluate their clinical value; (2) it is necessary to assess the protective role of miR-3135b in Golgi dispersal in response to other drugs and radiotherapy in a larger number of cell lines from diverse types of cancer; and (3) to characterize additional miR-3135b gene targets involved in the cytoplasmic DNA damage response. In conclusion, our results demonstrate that the GOLPH3/AKT/mTOR axis is negatively regulated by miR-3135b resulting in apoptosis activation and a decreased proliferative capacity and cell cycle arresting of HCT-15 and SW-480 cancer cells (Fig. 9). We also hypothesize that a pro-survival autophagic pathway could be activated which may result in recycling of Golgi fragments allowing cell survival of CRC cells in response to drugs treatments (Fig. 9). Our data also provide new insights into the role of microRNAs in the mechanisms of cytoplasmic DNA damage response, and suggest that manipulation of miR-3135b levels could be a potential therapeutic tool in colorectal cancer.  (2) Part of the previously described mechanism established that after DNA damage, GOLPH3 increases its affinity for myosin MYO18b resulting in tensile forces with F-actin cytoskeleton inducing Golgi fragmentation to confers cell survival and resistance to DNA-damaging agents 5,6 . (3) Here, we showed that ectopic overexpression of miR-3135b inhibits the Golgi dispersal, cell survival, cell cycle and sensitizes cells to 5-FU treatment, maybe by targeting GOLPH3 and AKT1/mTOR signaling. (4) In addition, 5-FU activates the autophagy associated to Golgi fragmentation which is hindered by miR-3135b mimics. Nonconfirmed miR-3135b potential targets participating in the mechanism are indicated by discontinuous red lines.

Validation of miR-3135b expression in an independent cohort.
Clinical and microRNAs microarray expression data from 1513 colorectal cancer patients were obtained from the Gene Expression Omnibus accession GSE115513 (PMID: 26740022) previously reported 24 . The differential expression analysis was carried out with the limma package (PMID: 25605792) using R environment, considering FDR < 0.05 as statistically significant.
RNA isolation. Total RNA was extracted from HCT-15, SW-480, Caco2 and CRL-1790 cells using TRIzol ® (Invitrogen) according to the manufacturer's instructions. RNA integrity was assessed by electrophoresis in 1% TAE agarose gel and bands were visualized with GelRed ® staining. www.nature.com/scientificreports/ assembled with vectashield/DAPI solution and documented using a confocal microscope. For measurement of Golgi area, images of HCT-15 and SW-480 cells stained with Golgi marker TGN38 antibodies were quantified by demarcation of the Golgi area with the polygon tool and calculation of its area using the ImageJ program. Luciferase reporter gene assays. The 3′-UTR sequence (wild-type and mutant) of the GOLPH3 gene was cloned downstream of the luciferase gene in the p-miR-report vector (Ambion) and recombinant plasmids were verified through automatic sequencing. The recombinant wild type and mutant pmiR-GOLPH3 plasmids and the p-miR-report control plasmid were transfected into HCT-15 cells using lipofectamine 2000 (Invitrogen). Then, cells were mixed with 1 × lysis buffer and transfered to a new tube. Finally, 20 µl of cell lysate were mixed with 100 µl of luciferase assay reagent and the light produced was measured with a Fluoroskan Ascent™ Microplate Fluorometer (Thermo Scientific).

Statistical analysis.
To compare means of more than two variables, one-way analysis of variance (ANOVA) followed by post-hoc test (Tukey) were used and p value lower than 0.05 was considered statistically significant. Experiments were performed three times by triplicate and results were represented as mean ± S.D. For miR-3135b expression assays in cells treated with drugs, we performed Student's t test to compare the expression of miR-3135 between two variables and p < 0.05 was considered statistically significant. Experiments were performed three times by triplicate and results were represented as mean ± S.D. The miR-3135b expression analysis in the previously published data sets from microarrays miRNAs profiling of 1893 normal/colorectal carcinomapaired samples, and 290 adenoma tissues samples deposited in NCBI GEO database (GSE115513) 24 , was carried out using the Limma package in R (www.bioco nduct or.org/packa ges/relea se/bioc/html/limma .html) employing gene-wise linear models and empirical Bayes methods. We have taken in account the multiple testing correction by considering an FDR < 0.05 as statistically significant.