miR-196b-5p Regulates Colorectal Cancer Cell Migration and Metastases through Interaction with HOXB7 and GALNT5

Colorectal cancer is the third most common cancer in males and the second most common cancer in females. In 2015, about 69,090 men and 63,600 women were diagnosed with colorectal cancer in the United States (1). The introduction of new drugs including inhibitors of angiogenesis and EGFR signaling have improved survival time in metastatic colorectal cancer (2), but despite these improvements, metastatic colorectal cancer still remains an incurable disease in most patients. This has prompted research to uncover the molecular mechanisms, which drive metastatic colorectal cancer, to help identify biomarkers better able to predict clinical outcome, and to inform novel therapeutic strategies in the context of increasingly personalized cancer care (3–6).

The discovery of the miRNAs in human cancer has added new insights into understanding cancer biology and has opened the door for novel diagnosis and treatment tools in patients with cancer (7, 8). miRNAs are a class of small, endogenous, noncoding, single-stranded RNAs that play a crucial role as posttranscriptional regulators for their larger counterparts mRNA expression (9). These small RNA molecules are known important regulators of carcinogenesis and in general, one particular miRNA can act as a tumor suppressor or a tumor promoter (“oncomiR”), depending on the cellular and molecular context (10, 11). Besides their potential as diagnostic and prognostic biomarkers, one of the most promising themes in recent years has been the development of therapeutic strategies which target miRNA differentially expressed in cancer versus normal tissue (12). Several miRNAs have been previously identified relevant for colorectal cancer pathogenesis (13–18). In this study we focused on miR-196b-5p, a miRNA that has been involved in carcinogenesis of other epithelial types of cancer. In colorectal cancer, previously published data suggested that high miR-196b-5p levels are associated with response to chemoradiotherapy (19, 20). In colorectal cancer, the influence of miR-196-5p is more controversial, as one study including 126 patients with colorectal cancer reported that high miR-196-5p levels were associated with short disease and overall survival (21), whereas a second study comprising more than 400 patients from 3 independent cohorts suggested the opposite was true (22).

This study aims to clarify the biological role and clinical significance of miR-196b-5p in patients with colorectal cancer by quantitating expression in 2 large independent patient cohorts, linked to a database containing comprehensive patient demographic and outcome data. The impact of miR-196b-5p in vitro and in vivo on the hallmark processes of cancer progression was also analyzed to gain deeper insight into the colorectal cancer relevant biological mechanisms influenced by miR-196b-5p.

To explore the prognostic value of miR-196b-5p, we measured the expression levels by qRT-PCR in two independent cohorts of central European patients. The cohorts comprise randomly selected cancerous tissue of histologically confirmed patients with colorectal cancer between 30 and 90 years of age who were diagnosed between the years 2005 and 2012. The patients' tissue was derived from either the Institute of Pathology, Medical University of Graz, Austria (cohort 1, n = 110; ref. 13) or the Department of Comprehensive Cancer Care, Masaryk Memorial Cancer Institute, Czech Republic (cohort 2, n = 182). The patients' clinicopathologic data were retrieved from medical records at the same institutions. This study has been approved by the Institutional Review Boards (No. 23-545 ex 10/11, No. 25-324 ex 12/13, and 2015/1690/MOU). All cases were reviewed based on pathology reports and histologic slides for pTNM categories. Patients were treated by standard surgical procedures and received adjuvant treatment when appropriate (stage II with risk factors or stage III). If advanced disease at the date of diagnosis, the patients then received medical treatment according to a physician-choice following the recommendation of the last version of European Society of Medical Oncology (ESMO) guidelines (23). Dates of death were obtained from the medical history, central registry of the Austrian or Czech Bureau of Statistics, or by telephone calls to their relatives as previously reported. miRNA measurement by qRT-PCR from cancerous tissue was performed as follows: Two to 8 10-μm-thick tissue sections were used for microdissection to obtain areas with at least 70% tumor cell content. miRNAs were isolated using the miRNeasy FFPE Kit 50 (Qiagen) according to the manufacturer's instructions. cDNA was synthesized from 500 ng of total RNA using a miScript Reverse Transcription Kit (Qiagen). Quantification of miRNAs was performed using the miScript SYBR Green PCR Kit (Qiagen) and the specific miScript Primer Assays for miR-196b-5p (hsa-miR-196b-5p, MIMAT0001080, assay name: Hs_miR-196b_1) and RNU6b (assay name: Hs_RNU6-2_11; all miScript Primer Assays from Qiagen) according to the manufacturer's recommendations on a Light Cycler 480 real-time PCR device (Roche). Expression values were calculated using normalization to RNU6B [after the formula 2⁁-(target gene – RNU6B)], log2-transformed, and further used for statistical analysis.

The human colorectal cancer cell lines HRT18, RKO, DLD1, SW480, and HCT116 were purchased from ATCC. The colorectal cancer cell lines were authenticated at the Cell bank of the Core Facility of the Medical University of Graz (Graz, Austria), by performing a short tandem repeat (STR) profiling analysis (Kit: Promega, PowerPlex 16HS System; catalog no. DC2101, last date of testing: March 3, 2016). Mycoplasma testing was performed using the Venor GeM Mycoplasma Detection Kit (Minerva Biolabs). Cells were cultured as described in detail in the Supplementary Materials section.

For functional studies, we transfected the colorectal cancer cell lines either with miR-196b-5p mimics (10 nmol/L), inhibitor (50 nmol/L), respectively), or GALNT5 and HOXB7 siRNAs (20 nmol/L) and for confirmation of results obtained from transient approaches and studying in vivo effects, we generated stable miR-196b-5p–overexpressing or inhibitor-expressing cell lines by lentiviral transduction, as it is described in all details in the Supplementary Materials section.

For relative mRNA and miRNA detection, absolute miRNA, and protein quantification, we applied standard methods of quantitative PCR, digital PCR (dPCR), and Western blot analysis as described in detail in the Supplementary Materials. Primer sequences are listed in Supplementary Table S1.

To investigate whether miR-196b-5p expression has an influence on cellular growth and chemosensitivity to common drugs, we applied the WST-1 proliferation assay as described in the Supplementary Materials section. We used 3 independent assays to monitor cell migration and invasion (xCELLigence Real Time Assay, Transwell and Scratch Assay) of colorectal cancer cells after transient or stable miR-196b-5p overexpression and inhibition. These assays and the in vivo experiments are described in every detail in the Supplementary Materials section.

To detect the most differentially expressed genes upon forced miR-196b-5p–overexpressing colorectal cancer cells, a microarray analysis was performed as described in the Supplementary Materials part. To identify putative mRNA targets, we used several prediction algorithms, qRT-PCR, Western blot analysis, and luciferase assay as specified in the Supplementary Materials.

All statistical analyses were performed using SPSS version 20 software (SPSS Inc.) or MedCalc software (version 13.1.2.0). In the Supplementary Material, we are listing the Recommendations for Tumor Marker Prognostic Studies (REMARK criteria) in relation to our reported results here (24). Unpaired or paired Student t test, Fisher exact test, χ² test, Spearman correlation, Mann–Whitney test, and Kruskal–Wallis test were applied where appropriate to analyze the association between miR-196b-5p expression and clinicopathologic parameters. Data of gene expression were log2-transformed. Overall survival was defined as the time from date of diagnosis to the date of death by any cause, and it was assessed using the Kaplan–Meier method. For survival analysis, we separated patients into low/high groups using cut-off value that separates the patients calculated according to the P value in log-rank test between the 25th and 75th of the cohort as was published previously (15). Univariate and multivariate Cox proportional hazards models including age, gender, tumor grade, tumor stage (according to the AJCC/UICC 2010 TNM classification), and miRNA expression levels. To test the proportional hazard assumption in Cox models, Schoenfeld residuals test was used. The reported results included HRs and 95% confidence intervals (CI). The log-rank test was performed to compare the survival curves of individual groups. A 2-sided P < 0.05 was considered statistically significant.

First of all, to explore the clinical significance of miR-196b-5p in patients with colorectal cancer, we measured miR-196b-5p expression by qRT-PCR in 2 independent cohorts and further analyzed its role as a potential prognostic factor. Patient characteristics for both cohorts are summarized in Supplementary Table S2. In cohort 1, we found a significantly lower level of miR-196b-5p expression in stage III and stage IV patients (P < 0.05; Supplementary Fig. S1A), whereas no difference was found for tumor grade (Supplementary Fig. S1B). In cohort 2, we confirmed a significant lower expression for miR-196b-5p in stage III and IV patients (P < 0.05; Supplementary Fig. S1C), and detected lower expression values in high-grade (G3/G4) tumors (P < 0.05; Supplementary Fig. S1D). For evaluating whether the levels of miR-196b-5p are associated with clinicopathologic parameters and survival of patients with colorectal cancer, we divided the cohorts into 2 groups (low and high miR-196b-5p expression) according to the criteria mentioned in the Material and Methods section (15). In cohort 1, low levels of miR-196b-5p were significantly associated with stage IV colorectal cancer at date of diagnosis (i.e., metastatic disease, P < 0.05, χ² test), whereas no association was observed for gender, age, or tumor grading (data not shown). In cohort 2, low levels of miR-196-5p were associated with high (G3 + G4) tumor grade and also with stage IV colorectal cancer (P < 0.05, χ² test). Figure 1 shows the Kaplan–Meier curves for 5-year overall survival and reveals that low miR-196b-5p expression is significantly associated with poor prognosis in both independent cohorts (P < 0.05, log-rank test). Univariate Cox proportional analysis identified high tumor grade (G3 vs. G1 + G2, in cohort 1), advanced tumor stage (stage IV vs. stage II–III), and low expression of miR-196b-5p as poor prognosticators for OS (all P values <0.05), whereas age and gender were not significantly associated with OS in both cohorts (Supplementary Table S3). To test whether the prognostic value of low miR-196b-5p expression prevailed after adjustment of other risk factors, multivariate analyses were performed using a Cox proportional hazard model. Multivariate analyses including age, gender, tumor grade, tumor stage, and miR-196b-5p expression, demonstrated that low miR-196b-5p expression was an independent predictor for poor survival in cohort 1 (HR 0.50; 95% CI, 0.29–0.84; P = 0.01; Supplementary Table S4).

Restricting our survival analyses on localized tumor stages only (stage II/III) and using the same cut off-value defined above, we found again a statistically significant poor survival for patients with low miR-196b-5p expression (P < 0.05 for both cohorts, log-rank test; Supplementary Fig. S2A and S2B). In cohort 1, we could also find similar and significant results in stage IV metastatic patients (Supplementary Fig. S2C). Finally, checking data from level 3 Illumina miRNASeq (IlluminaSequencingTechnology:Genome Analyzer) of the Cancer Genome Atlas (TCGA) dataset showed that low miR-196b-5p expression was associated with poor disease-free survival in univariate (HR 0.67; 95% CI, 0.48–0.93; P = 0.018348067) as well as multivariate analysis after adjusting for stage (HR 0.6912; 95% CI, 0.49–0.96; P = 0.02843).

Given the findings that low levels of miR-196b-5p were significantly associated with metastatic disease and poor patient outcome in both cohorts, we preceded our study to characterize underlying cellular and molecular mechanisms of this observation. At first, we measured miR-196b-5p expression in 8 different colorectal cancer cell lines using relative qRT-PCR and validated our findings by absolute digital PCR quantification in an independent laboratory. A strong and significant correlation for cell lines (Spearman correlation R = 0.881; P = 0.004; Supplementary Fig. S3A and S3B) and 10 selected patient samples (Spearman correlation R = 0.794; P = 0.006; Supplementary Fig. S3C and S3D) was found. In addition, we divided the cell lines into a miR-196b-5p low/high category as well as the patient samples and compared the expression values from the cell lines to patient sample to show miR-196b-5p levels in the cell line model system in relation to tissue samples (Supplementary Fig. S4A and S4B). To explore the biological role of miR-196b-5p, we started to perform gain (miR-196b-5p mimic) and loss (miR-196b-5p inhibitor) of function experiments by transiently manipulating miR-196b-5p levels in 3 independent colorectal cancer cell lines (HCT116, RKO, and HRT18; Supplementary Fig. S4C and S4D). To confirm the success of transfection of miR-196b-5p mimics, we validated decrease of known target genes HOXA9 and GATA6 in 2 independent colorectal cancer cell lines (Supplementary Fig. S4E and S4F; refs. 25, 26).

Cellular growth was not significantly influenced in the HCT116 and HRT18 cell line, whereas the RKO cells showed a slight but significant (P = 0.0028) decrease of growth after 96 hours of ectopic miR-196b-5p overexpression (Supplementary Fig. S5A–S5C). On the basis of previous reports about a possible association of miR-196b-5p expression and response to chemotherapy, we next explored the influence of miR-196b-5p expression levels on drug sensitivity for the 3 most commonly used colorectal cancer drugs (i.e., 5-fluorouracil, irinotecan, and oxaliplatin; refs. 19, 20). Testing several concentrations of these drugs, we could not detect any significant differences or trends with regard to chemotherapy sensitivity in low or high miR-196b-5p–expressing cells (Supplementary Figs. S6, S8A, and S8C). As miR-196b-5p expression levels were correlated with patient survival and metastatic stage IV disease, we next analyzed the migration behavior. For a screening approach, we applied a real-time cell analysis migration measurement method (xCELLigence system), which enabled us to measure the migratory process at multiple time points over 48 hours. After treatment with the miR-196b-5p inhibitor (i.e., low miR-196b-5p levels), all 3 cell lines showed the earliest start and highest rate of migration in this assay system, whereas high miR-196b-5p expression levels (i.e., miR-196b-5p mimic) led to the lowest migration rates in colorectal cancer cells (Supplementary Fig. S9A–S9C). To substantiate these findings in independent migration assays, we performed a scratch assay and a transwell assay for all 3 cell lines. We confirmed that low levels (i.e., inhibitor) of miR-196b-5p are associated with increased migration, whereas high levels of miR-196b-5p lead to decreased migration of colorectal cancer cells (Fig. 2). In addition to these 3 cell lines, we included DLD1 (with the highest miR-196b-5p levels) and SW480 (with very low expression levels) to perform a migration assay after inhibition (DLD1) or forced expression (SW480) of miR-196b-5p. Using the scratch wound healing assay, miR-196b-5p overexpression led again to significantly decreased migration in SW480 cells, whereas inhibition of miR-196b-5p resulted in significantly increased migration in DLD1 cells (Supplementary Fig. S10A and S10B). In addition, we found that low levels of miR-196b-5p increase HCT116 cancer cell invasion in 2 independent invasion assays (Supplementary Fig. S11A–S11C). To further confirm the results of transient transfections under different conditions, we generated 3 cell lines with stable miR-196b-5p overexpression or inhibitor expression (Supplementary Fig. S12A and S12B). Although we could successfully obtain a green fluorescent signal in all the transfected cells (indicating a successfully stable transfection), the miR-196b-5p levels could not be significantly decreased by introduction of the inhibitor construct. Notably, this is in line with the mode of action of miRNA inhibitors, which after binding to the miRNA lead to a loss of function rather than to reduce the expression levels (27). Consequently, we confirmed the same effects on cancer cell migration in all stably manipulated cell lines using again 2 independent migration assays (Fig. 3A–E).

For in vivo confirmation of this potentially prometastatic phenotype, we used stable miR-196b-5p inhibitor (silencing) expressing HCT116 cells and their respective control and labeled them with a luciferase-expressing vector for in vivo imaging purposes. As shown in Fig. 3F, mice intrasplenic injected with miR-196b-5p silenced cells showed more frequently [6 mice (75%) vs. 2 mice (28.5%) and more intense (mean photons emission: 317814 vs. 59191; P = 0.09; Fig. 3F; Supplementary Fig. S13A] signals. Regarding the gold standard of pathologic examination, 7 out of 8 mice (87.5%) in the group of miR-196b-5p inhibitor showed intra-abdominal metastases in the whole abdomen, whereas only 2 out of 7 (28.5%) mice in the control group were positive for this feature (P < 0.05; Fig. 3G and H). Further macroscopic and histologic examination of liver tissue revealed that 7 of 8 mice (87.5%) in the group of miR196b-5p inhibitor versus 2 out of 7 mice in the control cell group had detectable metastases (P = 0.041; Fig. 3I; Supplementary Fig. S13B and S13C). The median number of metastases per liver was 2.5 in the group of miR-196b-5p inhibitor versus 0 in the control group (P = 0.04; Supplementary Fig. S13D). Finally, we used human specific primers to detect human gene expression of GAPDH and the putative target genes HOXB7 and GALNT5, and could confirm the findings of histologic examination [i.e., human gene expression was detectable in 7 of 8 (87.5%) livers of miR-196b-5p–silenced cells versus 2 out of 7 (28.5%) in control livers, P < 0.05]. Thus, the in vivo data support our in vitro findings, that colorectal cancer cells with low miR-196b-5p expression showed a higher degree of metastases formation in vivo.

In search for possible molecular mechanisms that trigger this metastatic behavior, we next measured changes of expression levels of classical epithelial–mesenchymal transition (EMT) markers in miR-196b-5p–overexpressing and silenced cells, but could not detect any significant differences (Supplementary Fig. S14A and S14B). To detect potential molecular miRNA::mRNA interactions in an unbiased approach, we performed a microarray-based whole transcriptome profiling analysis in 3 independent biological replicates comparing the HCT116 miR-196b-5p stably overexpressing cells against control cells. On the basis of the microarray results, showing 38 significantly upregulated and 41 significantly downregulated genes (see also the heatmap in Fig. 4; Supplementary Table S5), we performed an in silico target prediction approach for these mRNAs and selected potentially interacting mRNA for further analysis (see detailed description in the Supplementary Methods section and Supplementary Table S6). To confirm the reliability of the microarray results, we further selected potential interactors with previously published data about involvement in carcinogenesis (i.e., GALNT5, HOXB7, MUC13, DKK1, CPA4, KRTAP2-3, and KRTAP1). We could confirm the downregulation of all these genes (100% concordance) from the microarray by quantitative RT-PCR (Supplementary Fig. S15A). To confirm these 7 downregulated candidates in an independent cell line model, we measured all 7 genes in stable miR-196b-5p overexpression HRT18 cells. In this independent cell line model, we could validate HOXB7 and GALNT5 as significantly downregulated (Supplementary Fig. S15B). To further explore whether these 2 genes are influenced by miR-196b-5p in a different experimental setting, we transiently transfected HCT116 and HRT18 cells with miR-196b-5p. Both genes were significantly downregulated on mRNA (Fig. 5A and B) and protein levels (Fig. 5C) upon ectopically miR-196b-5p overexpression. In contrast, introducing transiently miR-196b-5p inhibitor to HCT116 and HRT18 cells significantly increased the mRNA and protein expression of GALNT5 and HOXB7 (Supplementary Fig. S15C–S15E). To test whether the HCT116 cells used for the in vivo experiment (with stable miR-196b-5p silencing) have upregulation of GALNT5 and HOXB7 target genes, we performed mRNA and protein detection in comparison with the control cells. Both target genes were significantly upregulated on mRNA and protein level in miR-196b-5p–silenced cells (Supplementary Fig. S15F and S15G). In an attempt to confirm this upregulation of target genes in metastases of our mouse study, a qRT-PCR analysis using human-specific primers for detecting human RNA from liver metastases indicated human gene expression in 7 of 8 (87.5%) mice of miR-196b-5p–silenced cells versus 2 out of 7 (28.5%) mice in control cells (P < 0.05). Overall, we observed higher expression levels of the target genes HOXB7 and GALNT5 in miR-196b-5p–silenced liver metastases, but the number of affected livers in the control group (n = 2) did not allow meaningful statistical conclusions (Supplementary Fig. S16A and S16B). A direct interaction of miR-196b-5p and GALNT5 and HOXB7 was further substantiated by cloning the predicted wild-type and mutated 3′-UTR-binding site (Fig. 5D) in a luciferase-reporter plasmid. Reduced luciferase activity after miR-196b-5p cotransfection in HEK cells was observed in the GALNT5 and HOXB7 wild-type sequences, whereas the mutated binding site restored the normal luciferase activity (Fig. 5E and F). To further prove the interaction of miR-196b-5p with the putative binding site, we inhibited the endogenous miR-196b-5p expression in cells with high miR196b-5p expression (HCT116-stable miR-196b-5p overexpression) and cotransfected the wild-type 3′UTR-luciferase constructs of HOXB7 or GALNT5. The inhibition of miR-196b-5p resulted in increased luciferase activity for both genes, supporting our hypothesis (Supplementary Fig. S16C and S16D). Finally, we performed correlation analysis between miR-196b-5p and GALNT5/HOXB7 in tissue samples using publicly available data of TCGA. A significantly negative correlation between miR196b-5p and both target genes was found (P < 0.05, Spearman correlation; Supplementary Fig. S17).To test the significance of HOXB7 and GALNT5 in clinical samples, we analyzed the TCGA dataset and found that high HOXB7 expression is associated with poor survival (Supplementary Fig. S18). Using GEO microarray datasets (GSE12945 and GSE14333; refs. 28, 29), we identified a poor prognostic value of high HOXB7 and GALNT5 expression levels in 2 additional datasets (Supplementary Figs. S19 and S20).

Finally, to test the hypothesis whether miR-196b-5p exerts its effect on cancer cell migration by influencing HOXB7 and GALNT5 expression, we first performed siRNA-mediated knockdown experiments for both genes (Supplementary Fig. S21A and S21B) to prove the phenocopy of the miR-196b-5p phenotype. Reducing the expression levels of each of the mRNAs led to a reduced migration of cancer cells (Fig. 6A and B), and this effect was even more pronounced when both genes were knocked-down in parallel (Fig. 6C). Next, we explored whether expression of previously published downstream effector genes of HOXB7, including CCND1 and CDKN1B (30, 31) are indeed influenced by overexpression of miR-196b-5p in colorectal cancer cell lines. For both effector genes, we observed differential expression of mRNAs after forced miR-196b-5p expression, which is in line with previously published studies (Supplementary Fig. S21C and S21D). Finally, we did a rescue experiment to show that knockdown of GALNT5 and HOXB7 reduces the migration in cells with transient inhibition of miR-196b-5p. As shown in Fig. 6, siRNA-mediated knockdown of GALNT5 and HOXB7 expression in HCT116 cells with transient miR-196b-5p inhibitor expression (those cells exhibiting the promigratory/prometastatic phenotype above) resulted in a significant reduction of cellular migration in Transwell assay (Fig. 6D).

In this study, we sought to identify an association between miR-196b-5p expression and clinical outcome in colorectal cancer patients. miR-196b-5p has been implicated in several tumor scenarios including upregulation in acute lymphoblastic leukemia (32) and downregulation in glioblastoma (33), cervical cancer (34), and B-cell lineage lymphoma (35). These discordant observations highlight the fact that miR-196b-5p may function in a context specific way as either an oncomiR or a tumor suppressor gene, as has been described for many other miRNAs in an excellent recently published review (11).

In respect to colorectal cancer, there is conflicting data about the biological role and hence the potential prognostic utility of miR-196b-5p, which this study has attempted to resolve. By analyzing 292 colorectal cancer specimens from patients in 2 independent cohorts, we confirmed that low miR-196b-5p expression is significantly associated with poor prognosis. Our findings are in line with the data from Boisen and colleagues, who presented a large 3 cohort study and demonstrated that higher miR-196b-5p expression levels were independently associated with improved outcome in colorectal cancer whether or not bevacizumab therapy had been utilized (22). In addition, our results are underlined by a recently published study of Fantini and colleagues who found out that miR-196b-5p is significantly upregulated in pretreatment surgical resection colorectal cancer samples and that the upregulation of miR-196b correlates with less severe clinicopathologic characteristics (25).

Given the finding that low expression of miR-196b-5p is significantly associated with poor patient outcome, we performed a series of experiments to elucidate the cellular and molecular mechanisms underlying this clinical observation. Cancer cell growth and chemosensitivity were not consistently effected in colorectal cancer cell lines in which miR-196b-5p expression had been manipulated, suggesting these hallmarks of cancer are not significantly impacted (19, 20).

Low miR-196b-5p expression was however associated with metastatic stage IV disease in both of our cohorts suggesting miR-196b-5p may be associated with metastatic rather than local tumor proliferation pathways. Hence, using 3 different colorectal cancer cell lines, and 2 methods of miR-196b-5p manipulation (transient and stable transfection), in a variety of complementary in vitro and in vivo disease progression assays we were able to identify a consistent phenotype whereby increased colorectal cancer migration/metastases appears to be associated with reduced cellular miR-196b-5p expression.

To elucidate a potential molecular link between miR-196b-5p expression and cancer cell migration, we performed an unbiased whole transcriptome analysis using colorectal cancer cells in which miR-196b-5p was stably overexpressed. On the basis of the assumption that genes directly targeted by miR-196b-5p are suppressed in miR-196b-5p–overexpressing cells, we identified several promising mRNA candidates. As we were primarily interested in direct interaction partners, we performed an in silico target prediction analysis of potential complementary sequence seeds and a complementary literature search to further identify miR-196b-5p targets with a putative influence on cancer cell migration. The process in its entirety identified 2 candidate genes, HOXB7 and GALNT5 that have been previously implicated in gastric and cervical cancer cell migration and/or invasion (34, 36, 37). In colorectal cancer, a study by Liao and colleagues reported that HOXB7 is significantly correlated with invasive and aggressive characteristics and poor survival of patients. Liao and colleagues also revealed that HOXB7 overexpression promotes proliferation and tumorigenic growth of human colorectal cancer cells, both in vitro and in vivo (30). These data prompted us to focus on these 2 genes and we demonstrated that forced miR-196b-5p expression decrease GALNT5 and HOXB7 expression on mRNA and protein levels. As a previous study by How and colleagues (34) has already identified direct interaction of miR-196b-5p and HOXB7 in cervical cancer, we validate the identical interaction site of the HOXB7 3′UTR (34). However, a role of GALNT5 in colorectal cancer has never been reported before, so we further focused our study on the regulation of GALNT5.

In general, the family of N-acetylgalactosaminyltransferases (GALLNTs) are responsible for glycosylation of O-glycans linked with N-acetylgalactosamine (O-GalNAcs). A study by Zhang and colleagues provides evidence of O-glycosylation influencing the composition of the extracellular matrix (ECM) and suggests a role for abundant protein modification in disease states where matrix composition and cell adhesion are altered (38). O-GalNAc glycosylation of ECM proteins is important for secretion and plays a role in modulating ECM composition and affecting cell adhesion, growth, and embryogenesis. O-GalNAc glycans have been implicated in numerous molecular pathways with key roles in tumor formation, cell invasion, and cancer progression (39). Some members of the GalNTs enzyme family have already been associated with several types of human cancer. A study by Peng and colleagues revealed that knockdown of GALNT7 markedly inhibits cervical cancer cell proliferation, migration, and invasion (40). For GALNT5 the only published study so far comes from He and colleagues, who demonstrated that low intratumoral GALNT5 expression is a significant negative prognostic predictor for patients with gastric cancer (37).

Finally, to test the hypothesis that these genes, HOXB7 and GALNT5, are factors which mediate the effect of miR-196b-5p, we performed migration assays following gene knockdown and rescue experiments in miR-196b-5p–silenced cells for both genes. Reduced levels of HOXB7 and GALNT5 phenocopied the effects of miR-196-5p and importantly, this antimigratory effect became yet more pronounced when both genes were knocked-down in parallel. These findings also highlight the fact that targeting multiple mRNAs in parallel may amplify the biological impact in future the potential therapeutic benefits of miRNAs.

Our study is not without limitations. As with all retrospective patient cohorts, they are prone to selection bias and heterogeneity in terms of treatment schedule and modalities. Although we used several experimental model systems, we cannot rule out that genes other than HOXB7 and GALNT5 might influence the cellular phenotype regulated by this miRNA. In our study, we did not perform in vivo knockdown experiments against HOXB7 and GALNT5 to diminish the prometastatic phenotype, but these experiments might be interesting future directions. In addition, further miR-196b-5p targeting strategies (with encapsulated mimics or aptamers) should be performed in appropriate preclinical models before testing this strategy in human patients with colorectal cancer. Nevertheless, using independent cohorts and several experimental models and techniques, we sought to minimize such effects and limitations.

In conclusion, the results of our study establish miR-196b-5p as a prognostic factor in colorectal cancer. miR-196b-5p promotes colorectal cancer metastases, at least in part, through the regulation of HOXB7 and GALNT5. Further prospective studies and clinical trials are warranted to proof the role of miR-196b-5p as a prognostic marker and furthermore evaluate if miR-196b-5p mimics carry therapeutic value in patients with colorectal cancer.

No potential conflicts of interest were disclosed.

The sponsor of the study had no role in study design, data collection, data analysis, data interpretation, writing of the report or the decision to submit this manuscript for publication.

Conception and design: V. Stiegelbauer, M. Karbiener, J. Haybaeck, O. Slaby, M. Pichler

Development of methodology: V. Stiegelbauer, M. Resel, E. Heitzer, M. Pichler

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): V. Stiegelbauer, P. Vychytilova-Faltejskova, A. Reicher, M. Resel, A. Wulf-Goldenberg, H. Stoeger, J. Haybaeck, M. Svoboda, G. Hoefler, O. Slaby, M. Pichler

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V. Stiegelbauer, M. Karbiener, A.-M. Pehserl, A. Reicher, M. Resel, E. Heitzer, C. Ivan, H. Ling, A. Deutsch, J.B. Adiprasito, J. Haybaeck, G. Hoefler, O. Slaby, A. Gerger, M. Pichler

Writing, review, and/or revision of the manuscript: V. Stiegelbauer, M. Karbiener, A.-M. Pehserl, M. Bullock, H. Ling, J.B. Adiprasito, H. Stoeger, J. Haybaeck, M. Svoboda, M. Stotz, G.A. Calin, A. Gerger, M. Pichler

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V. Stiegelbauer, A. Reicher, M. Resel, M. Bullock, H. Stoeger, M. Stotz

Study supervision: A. Deutsch, A. Gerger, M. Pichler

M. Pichler research was supported by an Erwin-Schroedinger Scholarship of the Austrian Science Funds (project no. J3389-B23) by a research grant of the Verein für Krebskranke of the Medical University of Graz, by the fund of the Oesterreichische National bank no. 14869 (to M. Pichler), and by a research grant of the Styrian Krebshilfe. This study was supported in part by funds of the Oesterreichische National bank (Anniversary Fund, project no. 15400 to A. Gerger). Furthermore, research activities from M. Stotz are supported by a START grant from the Medical University of Graz and a Hans und Blanca Moser Foundation for Early Stage Cancer Researcher. This work was supported by grant IGA NT/13860-4/2012 of the Czech Ministry of Health, by the project “CEITEC” (Central European Institute of Technology (CZ.1.05/1.1.00/02.0068) and the project BBMRI CZ (LM2010004). G.A. Calin is The Alan M. Gewirtz Leukemia & Lymphoma Society Scholar. Work in G.A. Calin's laboratory is supported in part by the NIH/NCI grants 1UH2TR00943-01 and 1 R01 CA182905-01, the UT MD Anderson Cancer Center SPORE in Melanoma grant from NCI (P50 CA093459), Aim at Melanoma Foundation and the Miriam and Jim Mulva research funds, the UT MD Anderson Cancer Center Brain SPORE (2P50CA127001), a Developmental Research award from Leukemia SPORE, a CLL Moonshot Flagship project, a 2015 Knowledge GAP MDACC grant, an Owens Foundation grant, and the Estate of C.G. Johnson Jr.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.