Abstract
MicroRNA (miRNA) profiling represents a promising source of cancer-related biomarkers. miRNA signatures are specific for each cancer type and subgroups of patients with diverse treatment sensitivity. Yet this miRNA potential has not been satisfactorily explored in rectal cancer (RC). The aim of the study was to identify the specific miRNA signature with clinical and therapeutic relevance for RC. Expressions of 2555 miRNA were examined in 20 pairs of rectal tumors and matched non-malignant tissues by 3D-Gene Toray microarray. Candidate miRNAs were validated in an independent cohort of 100 paired rectal tissues and in whole plasma and exosomes of 100 RC patients. To study the association of miRNA profile with therapeutic outcomes, plasma samples were taken repeatedly over a time period of 1 year reflecting thus patients’ treatment responses. Finally, the most prominent miRNAs were investigated in vitro for their involvement in cell growth. We identified RC-specific miRNA signature that distinguishes responders from non-responders to adjuvant chemotherapy. A predominant part of identified miRNAs was represented by the members of miR-17/92 cluster. Upregulation of miRNA-17, -18a, -18b, -19a, -19b, -20a, -20b and -106a in tumor was associated with higher risk of tumor relapse and their overexpression in RC cell lines stimulated cellular proliferation. Examination of these miRNAs in plasma exosomes showed that their levels differed between RC patients and healthy controls and correlated with patient’s treatment response. miRNAs from miR-17/92 cluster represent a non-invasive biomarker to predict posttreatment prognosis in RC patients.
Introduction
Late diagnosis accompanied with a lack of predictive markers are the main reasons for high cancer mortality. Despite advances in surgical techniques and improvements in neoadjuvant and adjuvant chemoradiotherapy regimens, patients with the same stage of particular cancer type may have different treatment susceptibility and long-term outcomes (1). Therefore, identification of new predictive markers of treatment response and prognostic markers of patient’s overall survival is inevitable for the advent of individualized treatment strategies.
MicroRNAs (miRNAs) regulate gene expression via posttranscriptional gene silencing by mRNA degradation or translational block. miRNAs aberrant expression and function may exert a severe effect on cellular integrity and contribute to tumorigenesis as they control fundamental biological processes such as apoptosis, cell proliferation, differentiation, angiogenesis and invasion (2). In this context, miRNAs are extensively studied in relation to cancer and their specific expression profiles were associated with different types of malignancies, including colorectal cancer (CRC) (3–6). The recent enthusiasm in elaborating miRNAs as a potential cancer-related biomarkers is driven by their unique features: they regulate expression of oncogenes and tumor suppressors, and more than 50% of miRNA genes are eventually located in the regions altered in human cancer by deletions or amplifications (7). Also, miRNAs are present in plasma and feces at detectable levels, and due to their small size and stem–loop structure, they are more stable than mRNAs. Finally, due to an imperfect sequence complementarity to its targets, a single miRNA can regulate the expression of more than hundreds mRNA transcripts simultaneously, having thus a potentially higher informative value in distinguishing different tissues and tumors than mRNAs do. Organization of miRNA genes in polycistronic clusters that are co-expressed and hypothesized to regulate functionally related genes is also of note (8).
During the last decade, there is rising evidence that aberrant miRNA expression has a functional role in the initiation and progression of CRC, although not considering possible distinctions between colon and rectum-located carcinomas (9,10). Due to many differences in embryogenesis, etiology, anatomy, genetics and treatment response between the colon and rectal tumors, CRC might no longer be regarded as a single etiopathogenic entity (11–13). Yet, there is only a limited number of studies that investigated rectal cancer (RC) alone, and a bulk of those few was specifically focused on miRNA profiling in relation to neoadjuvant treatment response in RC patients, as reviewed in (14). Studies reporting on miRNA expression patterns in respect to adjuvant therapy or long-term survival of patients, or those investigating miRNA levels not only in tumor but also in peripheral plasma, are strikingly rare.
Here, we report a comprehensive characterization of RC tissues using miRNA expression profiling as a potential predictive and prognostic marker. We screened 20 pairs of rectal tumors and paired non-malignant tissues for expression levels of 2555 miRNA. Tumor-specific miRNA expression profiles were further verified in an independent cohort of 100 paired RC tissues and compared with profiles of plasma and exosomes, repeatedly taken in a time period of 1 year. This unique set of samples enabled us to correlate tissue and plasma/exosome miRNA expression profiles and investigate them in relation to patient’s prognosis. The most promising miRNAs were examined for their involvement in cellular processes such as proliferation and migration in in vitro experiments using RC cell lines.
Materials and methods
Design of the study
The design of the study is depicted in Figure 1. Briefly, large-scale miRNA screening was performed on a ‘discovery cohort’ of 20 pairs of rectal tumor and adjacent mucosa samples to search for miRNAs deregulated in malignant tissue. Tumors were distributed among all four tumor–node–metastasis (TNM) stages (each TNM stage was represented by five tumors), and none of them was exposed to neoadjuvant treatment before the sampling (Supplementary Table 1, available at Carcinogenesis Online). Verification of candidate genes identified in the screen was conducted on a ‘rectal validation cohort’ of 100 pairs of rectal tumors and non-malignant tissues. This cohort included both untreated patients and those receiving neoadjuvant therapy (Supplementary Table 2, available at Carcinogenesis Online). Findings on rectal tumors were further compared with colon tumors to identify miRNAs that are specific for RC. ‘Colon validation cohort’ consisted of 102 paired tissue samples (Supplementary Table 3, available at Carcinogenesis Online). All the tissue samples were obtained during surgical resection of the tumor. Panel of miRNAs, which expression profile was associated with RC patients’ treatment response, was further analyzed in blood plasma samples. ‘Plasma cohort’ involved 100 RC patients and 31 healthy individuals (Supplementary Table 4, available at Carcinogenesis Online). Plasma samples were collected from patients in 6 month intervals, reflecting condition at the diagnosis (T0), at the termination of adjuvant therapy (T1) and 6 months later (T2), described in our previous work in details (15). Finally, miRNAs related with treatment response in RC were studied by in vitro assays for their effect on proliferation and migration of RC cells.
An overview of the experimental design. A scheme summarizes the number of miRNAs analyzed in different sample cohorts.
Study patients and collection of biological specimens
All patients included in the study were recruited at the General University Hospital at the Charles University and Thomayer Hospital in Prague, or Teaching Hospital and Medical School of Charles University in Pilsen, Czech Republic. Ethics committees of all hospitals approved the study, and all patients signed informed consent. Patients diagnosed with either RC or colon cancer were sampled for tumor tissue, adjacent non-malignant tissue and/or peripheral blood. Tissue samples were deep-frozen in liquid nitrogen immediately after resection, and further stored at −80°C, and so were plasma samples isolated from ethylenediaminetetraacetic acid (EDTA)-treated whole fresh blood by centrifugation at 500g for 10 min. Clinical data of all patients are represented by tumor location, TNM stage at diagnosis, neoadjuvant and adjuvant treatment regimens, treatment response by means of local recurrence of tumor following the adjuvant therapy and 5 year survival.
Isolation and quality control of RNA from human samples
Deep-frozen tissue samples were homogenized using MagNA Lyser Green Beads in the MagNA Lyser Instrument (F. Hoffmann-La Roche). Total RNA including miRNA was isolated using mirVana™ miRNA Isolation Kit without miRNA enrichment (Thermo Fisher Scientific). Quality and concentration of miRNA were analyzed by Agilent Bioanalyzer 2100 (Agilent Technologies), using Small RNA Analysis Kit (Agilent Technologies). Plasma/Serum Circulating and Exosomal RNA Purification Kit (Norgen Biotek) was used for isolation of exosomal content and cell-free miRNAs. Exosomal miRNA was isolated by ExoQuick™ Exosome Precipitation Solution (System Biosciences). Qubit 3.0 Fluorometer and Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific) were used to control for quality of miRNAs.
miRNA screening with 3D-Gene microarray
Total RNA of 400 ng per sample was mixed with miRNA spike (Cat. no. TRT-XR304; Toray) and further labeled, hybridized and washed according to the instruction manual from Toray (H-M-R miRNA protocol 4-Plex V E2 1 EG) using Toray miRNA labeling kit (Cat. no. TRT-XE211) and Toray Human miRNA Oligo chip 4plex (Cat. No TRT-XR518). The intensity of each miRNA was analyzed with the 3D-Gene Scanner 3000 (Toray) with auto gain, auto focus and auto analysis settings.
High-throughput quantitative PCR of human samples
All RNA samples underwent reverse transcription (RT) with TaqMan MicroRNA Reverse Transcription Kit (Thermo Fisher Scientific) using PCR cycler C1000 Touch™ thermal cycler (Bio-Rad). All samples were preamplified using a pool of TaqMan MicroRNA Assays (Thermo Fisher Scientific) and C1000 Touch™ thermal cycler (Bio-Rad). All assays were validated for a multiplex RT, preamplification and for their quantitative PCR (qPCR) performance. Successfully validated probes are characterized in Supplementary Table 5, available at Carcinogenesis Online. qPCR was performed using the high-throughput platform BioMark™ HD System (Fluidigm) using 48.48 or 96.96 Dynamic Array™ IFC for Gene Expression.
Cell culture and transfection
Human rectal carcinoma cells SW1463 were purchased from Sigma–Aldrich (Cat. no. 90112713). The authentication of the cells has been provided by using the short tandem repeat profiling method (Generi Biotech Ltd, Czech Republic). Last short tandem repeat analysis was carried out 6 months before the start of the experiments. Cells were cultured in RPMI-1640 media (Sigma–Aldrich), supplemented with 10% fetal bovine serum (Gibco), 1 mM l-glutamine (Biosera), 1 mM sodium pyruvate and 1 mM penicillin–streptomycin (Biosera). Cells, maintained at 37°C, in 20% O2 and 5% CO2 environment, were transfected in 6-well plate format with 2.5 pmol of MISSION miRNA mimics (hsa-miR-18a, has-miR-18b, hsa-miR-19a, hsa-miR-19b; Sigma–Aldrich) or with miRNA mimics negative control with no homology to human genome (HMC0003; Sigma–Aldrich) using Lipofectamine® RNAiMAX 2000 (Invitrogen™). Efficiency of transfection was analyzed by qPCR measuring expression levels of transfected miRNAs compared with negative control.
Viability and proliferation assays
Cells were plated for colony formation assay onto 6-well plates in a density of 500 cells per well and cultured in normal RPMI media. After 12 days, colonies were fixed with 3% formaldehyde, stained with 1% crystal violet and colonies were manually counted. For proliferation assay, cells were plated onto 96-well plates in a density of 3 × 104 cells per well. Metabolic activity of the cells was measured 24 h after plating by adding water-soluble tetrazolium-1 solution into the media as recommended by the manufacturer (Roche). Absorbance at 450 and 690 nm was measured on BioTek ELx808 absorbance microplate reader (BioTek).
Migration assay
Cell migration was assayed for by using Transwell permeable supports 8.0 µm (Corning). Cells in a density of 1 × 104 were seeded on the top of a Transwell supports in 24-well plate format and cultured in RPMI media supplemented with 0.5% fetal bovine serum. Cells were allowed to migrate for 48 h through the membrane into the RPMI media containing 20% fetal bovine serum. Migrated cells were fixed with 3% formaldehyde, stained with 1% crystal violet and counted in four random fields under ×200 magnification.
Preprocessing of qPCR data
Fluidigm Real-Time PCR Analysis software (Fluidigm) and GenEx qPCR data analysis software version 6 (MultiD) were used for preprocessing of qPCR data. Missing data from 3D-Gene microarray were replaced by value of background signal, non-malignant tissue was subtracted from tumor tissue, and log scale was applied to all data. Statistical difference from 0 was tested by one sample test. Data from BioMark™ HD System were treated as follows: baseline correction of the amplification curves was set as linear derivative, Cq threshold was set identically for each array (each assay individually), data from multiple arrays were normalized with interplate calibrators. Technical replicates were tested for outliers with Grubbs test, outliers were removed and replicates were averaged. Cq values that were out of limit of quantification were removed. Missing data were replaced with the highest Cq value +2. Expression levels in tissues were normalized to expression levels of miRNA 186, selected by the NormFinder algorithm. Expression levels in plasma samples were normalized with an averaged expression of all analyzed miRNAs. Data were further transformed into relative quantities, and log scale transformation was used to ensure normal data distribution.
Statistical analysis
Statistical analysis was performed by GenEx qPCR data analysis software version 6 (MultiD), by SAS 9.3 software (SAS Institute) and R version 3.4.0. Analyses were conducted separately for different sources of tissues, i.e. rectum or colon. Mixed general linear model was used for comparisons. Groups were compared using contrasts. The significance level was set up as alpha of 0.05. P values were corrected for multiple testing by Bonferroni correction.
Results
Large-scale miRNA screening identified multiple miRNAs deregulated in RC
In the screening of 2555 miRNAs executed in the ‘discovery cohort’ of 20 rectal tumors matched with their non-malignant mucosa, we identified 71 miRNAs that were deregulated in tumor tissue. The levels of miRNAs were not affected by any treatment of the patients applied before the tissue sampling, as only neoadjuvant therapy-naive patients were subjected to the screen. Identified miRNAs were deregulated in tumor tissue irrespectively of its TNM characteristics. Overall, 57 of 71 miRNAs were downregulated, whereas 14 were upregulated in tumor (Supplementary Table 6, available at Carcinogenesis Online). In the next step, we verified expression profiles of miRNAs identified in the screen in a larger set of rectal samples. First, 71 TaqMan MicroRNA Assays were comprehensively tested for their performance in high-throughput BioMark HD System platform. Twenty-two probes passed all selection criteria based on the minimum information for publication of quantitative real-time PCR experiments (MIQE) guidelines (Supplementary Table 5, available at Carcinogenesis Online) (16). Those 22 miRNAs were then tested in ‘rectal validation cohort’, consisting of 100 matched tumor and non-malignant tissues from RC patients. Tissues from both, neoadjuvantly treated and untreated, patients were included in this analysis. We confirmed that 13 miRNAs were significantly deregulated in tumor tissue, 10 being upregulated and 3 downregulated (Figure 2A). These differences between tumor and normal tissue were observed only in samples from neoadjuvant therapy-naive patients, whereas they were not detectable in samples from patients receiving chemoradiotherapy before the tumor resection (Figure 2B).
miRNAs with significantly different expression profiles between tumor tissue (T) and matched non-malignant tissue (N). (A) Panel of miRNAs deregulated in rectal tumors of 69 neoadjuvantly untreated patients. (B) The differences in miRNA levels between tumor and non-malignant tissue are not detectable in 32 patients treated with neoadjuvant therapy before the sampling. (C) Panel of miRNAs deregulated in 102 colon tumors. miRNAs common for both rectal and colon carcinomas are shown in black bars. Bars represent mean fold change of tumor tissue (T) relative to non-malignant tissue (N). Error bars represent 95% confidence interval.
These data show the presence of deregulation of miRNA expression in rectal tumor tissue compared with non-malignant tissue and suppression of these differences by neoadjuvant therapy.
Rectal and colon carcinomas have similar miRNA profiles
We further tested whether miRNAs deregulated in rectal tumors are specific for this type of cancer. For this purpose, we analyzed all 22 miRNAs selected in the screen and passing all the selection criteria for qPCR performance in the ‘colon validation cohort’ comprising 102 tumors and non-malignant tissues from patients having colon cancer. Most of the analyzed miRNAs were commonly deregulated in both cancer types, and only miRNA 744 was observed to be upregulated in RC but not in colon cancer (Figure 2A). On the other hand, some miRNAs that were selected in the screen but not confirmed to be deregulated in rectal tumors appeared to be deregulated in colon tumors only. Namely, miR-133a, -133b, -1, -143 and -28 were downregulated whereas miR-103 was upregulated in colon tumor tissue (Figure 2C).
In summary, our data suggest that tumors localized in colon or rectum have similar miRNA profiles with exceptions of few miRNAs that are specific for either of two specimens.
RC miRNA profile associates with patient’s treatment response
Identified RC miRNA signatures were further analyzed with clinical characteristics of patients that reflect the invasiveness of the disease (TNM) and patient’s adjuvant treatment response (recurrence of the tumor after adjuvant therapy and 5 year survival of the patients since diagnosis). In this analysis, only 69 neoadjuvantly-untreated patients were considered. We have identified one miRNA that was in association with progression of tumor, i.e. TNM at diagnosis. As shown in Figure 3A, expression levels of miR-21 were significantly higher in tumors diagnosed with TNM stage IV (n = 15) than in TNM stage I–III classified tumors (n = 50). Moreover, multiple miRNAs (Figure 3B) were observed to be differently expressed in rectal tumors that have positively responded to adjuvant therapy (i.e. no recidive at the end of the treatment administration, n = 50) than those that did not respond well and exhibited relapse (i.e. local recidive of the tumor at the end of the treatment, n = 7). This miRNA panel seems to be specific for RC as none of identified miRNAs showed any association with treatment response or survival of patients with colon cancer (data not shown). Interestingly, majority of treatment-associated miRNAs identified by us were not random but organized in miR-17/92 cluster (Figure 3C).
miRNAs associated with tumor progression and treatment response. (A) The expression level of miRNA 21 is associated with disseminated rectal cancer, being more upregulated in tumors classified with TNM stage IV (n = 15) than in tumors with TNM stage I–III (n = 50). (B) Panel of miRNAs associated with recidive of rectal carcinoma following adjuvant therapy. Group of patients with no recidive (n = 50) was compared with the group of patients with the local recurrence of tumor after adjuvant therapy (n = 7). Majority of identified miRNAs belong to the miR-17/92 cluster. Bars represent mean fold change of tumor tissue (T) relative to non-malignant tissue (N). Error bars represent 95% confidence interval. (C) Members of the miR-17/92 cluster and its two paralogues, miR-106a/363 and -106b/25. miRNAs identified in this study are highlighted in bold. Distribution of miRNAs into their particular families is defined by different colors: members of miR-17 family in orange, miR-18 family in turquoise, miR-19 family in green and miR-92 family in purple.
Exosomal profile of miR-17/92 cluster reflects posttreatment regression of RC
Next, we focused on the miRNAs belonging to the miR-17/92 cluster, as many of those were found to be tumor deregulated and simultaneously associated with sensitivity to adjuvant chemotherapy. We attempted to examine those miRNAs (miR-17, -18a/b, -19a/b, -20a/b and -106a) in whole plasma and exosomes, and to study their expression profile in relation to patient’s treatment outcome. The ‘plasma sample cohort’ involved 100 RC patients sampled for peripheral blood at the diagnosis (T0), at the termination of adjuvant therapy (T1) and a year from the diagnosis at the regular examination (T2). Only patients with no signs of recidive were analyzed at T2. The actual numbers of analyzed samples in the patient group were 52 for exosomes and 88 for whole plasma at T0, 35 and 38 at T1 and 22 and 27 at T2. Patient’s exosomal and plasma profiles were compared with those of 31 healthy controls. All analyzed miRNAs were detectable in both whole plasma and exosomes. miRNA levels from cell-free plasma did not distinguish patients having cancer from healthy controls by any of analyzed miRNAs. Their expression levels were not different between individual patient’s samplings from T0 to T2. On the contrary, exosome miRNA levels were deviated from reference levels (i.e. those detected in healthy individuals) only at the presence of the tumor (T0) and these differences were diminished a year after diagnosis when tumor was no longer present/active (T2) (Figure 4A–H).
Plasma levels of miR-17/92 cluster of miRNAs differently expressed between rectal cancer patients and healthy controls. (A–H) Levels of miRNAs measured in plasma exosomes but not in whole plasma differ between patients having rectal cancer (T0) and 31 healthy controls. Patients were sampled at three time points approximately 6 months apart starting at the diagnosis (T0), at the end of adjuvant therapy (T1) and during the posttreatment examination (T2). Patients’ exosomal miRNA levels altered over this time interval and were no longer different from healthy controls at T2 sampling, reflecting thus a disease remission. Only patients with no signs of recidive were analyzed at T2. Number of patients analyzed at each time point was as follows, T0: exosomes n = 50, whole plasma n = 88; T1: exosomes n = 35, whole plasma n = 38; T2: exosomes n = 22, whole plasma n = 27. Box plots represent median of log relative values and interquartile range. *P < 0.05, ***P < 0.001.
These data suggest that miRNA expression profile in plasma exosomes rather than in whole plasma reflects the miRNA expression profile of the tumor tissue and a course of the disease.
Levels of miRNA of family 18 and 19 influence cell proliferation and migration
Because both miRNAs of family 18 (18a and 18b) and family 19 (19a and 19b) were commonly associated with treatment response in this study while they were not previously related to pathogenesis of RC, we investigated the impact of their expression levels on proliferation and migration of RC cells. To this end, we have overexpressed miR-18a/b and -19a/b in the SW1463 RC cells and measured their ability to proliferate, migrate and form single-cell colonies. The significant increase in expression levels of all analyzed miRNAs following the miRNA mimics transfection was confirmed by qPCR (Figure 5A). We observed a significant increase in proliferation activity of SW1463 cells overexpressing miR-18b, as measured by water-soluble tetrazolium-1 assay (Figure 5B) and by colony formation assay (Figure 5C). These results suggest that high levels of miR-18b in tumor tissue, as observed by us in human samples, might provide a proliferation advantage to tumor cells. In addition, we observed a decrease in migration of SW1463 cells when either miR-18a or -19a were overexpressed (Figure 5D).
Expression levels of miRNAs from miR-17/92 cluster influence proliferation and migration of rectal cancer cells. (A) The efficiency of SW1463 cells transfection by miRNA mimics represented as a fold change in expression levels of particular miRNAs compared with miRNA mimics negative control (NC). (B) The effect of miR-18a/b and -19a/b overexpression on the proliferation capacity, (C) colony-forming activity, and (D) migration activity of SW1463 miRNA mimics transfected cells. Results are expressed as a percentage of NC that was set to 100% and represent mean and standard error of the mean of at least two independent experiments. *P ˂ 0.05.
Taken together, these data show that miR-18 and -19, members of miR-17/92 family cluster, modulate proliferation and migration capacity of RC cells.
Discussion
miRNAs represent master regulators of gene expression and, in the context of cancer, they play role of oncogenes or tumor suppressors (17). Their deviated expression profiles are cancer-type specific. Despite the differences between colon cancer and RC in treatment strategies and patient’s prognosis, CRC is usually studied as one entity. To reflect the possible molecular differences between these two specimens, we aimed to identify RC-deregulated miRNAs in confrontation with colon cancer and in relation to patient’s therapeutic outcomes.
By assaying for 2555 miRNAs in 20 tumor samples, we identified a panel of 71 miRNAs deregulated in RC. Majority of them were underexpressed in tumors (57 versus 14), suggesting that miRNAs are mostly downregulated during rectal carcinogenesis. Global downregulation of miRNA expression seems to be common feature in cancer (18,19). This cancer-related miRNA suppression can be explained by the defective miRNA biogenesis, mutations in miRNA coding regions or epigenetic modifications to their promoters. As a result, miRNA inhibitory regulation of genes involved in cell differentiation and apoptosis is disturbed, which affects their ability to prevent tumor development (17). Up to date, several studies aimed at investigation of miRNA expression in RC (20–27). Majority of them, however, suffered from small sample size and/or limited set of analyzed miRNAs, not applying an initial unbiased miRNA array screen to identify true candidates. Moreover, different methods of tissue storage were exploited. These discrepancies most probably underlie the lack of consistent observations across the studies and it is rather difficult to interpret our data in the context of existing literature. An exception is represented by the study of Gaedcke et al., in which 57 deep-frozen RC tumors and matched non-malignant tissues were analyzed (28). The authors, unlike us, used miRCURY LNA microarray to screen for expression of 2090 miRNAs, but the design of the study is mostly comparable with ours. Authors identified 20 up- and 29 downregulated miRNAs. We did not observe major overlap with these data; however, some of the miRNAs were identified in both studies and they all showed the same tendency of deregulation. In agreement, miR-17, -18a and -21 were upregulated, whereas miR-195, -378, -143, -145 and -1 were found to be commonly downregulated in both studies. In the follow-up validation of the screen-identified miRNAs, we confirmed 13 of them being significantly deregulated in a set of 100 rectal tumors, including miRNAs found by Gaedcke et al. (miR-17, -18a, -21, -195 and -378i). However, it needs to be pointed out that we were able to analyze only 22 of 71 screen-identified miRNAs in the validation study, because only those 22 have met all the criteria for optimal qPCR performance based on MIQE guidelines (16). This fact might have artificially restricted the actual number of successfully validated miRNAs. Notably, our list of confirmed upregulated miRNAs involved eight members of miR-17/92 cluster. miR-17/92 cluster is responsible for expression of six individual miRNAs, but together with its two paralog clusters miR-106a/363 and -106b/25, it involves 15 miRNAs from four families (family miR-17, -18, -19 and -20) (29). Interestingly, in contrary to general downregulation of majority of cancer-related miRNAs, miR-17/92 cluster levels are always elevated in malignant tissue, such as those derived from breast, lung, pancreas, prostate, stomach and colon (30,31). The increased expression might be attributed to tumor amplification of the cluster coding locus 13q31–32 that is frequently gained also in CRC (32). The broad existence of deregulation is perhaps not surprising in the context of cellular processes regulated by the cluster. These involve cell proliferation [by targeting i.e. transforming growth factor beta (TGF-β), SMAD and others], cell cycle regulation (i.e. p21, E2F1) or cell death (i.e. PTEN, BCL2) (29).
A preoperative neoadjuvant chemoradiotherapy is applied in locally advanced stages of RC. To avoid any depreciation of the expression data by exposure to neoadjuvant therapeutics, screen was performed on therapy-naive samples only. On the contrary, validation sample set involved both untreated patients and patients treated with neoadjuvant therapy to explore the effect of the treatment on miRNA expression. This strategy appeared to be accurate as the differences in expression profile between tumor and non-malignant tissues could only be observed in untreated patients. Svoboda et al. examined six rectal tumors before and 2 weeks after neoadjuvant chemoradiotherapy and they reported upregulation of miR-125b and -137 as a result of the treatment. They hypothesized that miRNA levels tend to change to normal levels after efficient tumor destruction as both miRNAs are known to be downregulated in CRC (23). The effect of treatment on suppressing the differences between tumor and non-malignant tissues may have two possible reasons: (i) therapy-responding tumors exhibit change of miRNA expression toward normal expression levels typical for adjacent non-tumor tissue; (ii) a ratio of malignant to stromal cells in the analyzed tumor samples of treated patients was shifted toward the latter leading thus to a loss of molecular profile attributable to tumorous tissue. Isolation of particular cell types from analyzed sample using laser microdissection together with single-cell qPCR might help to address these issues.
CRC is usually studied as a single disease, not distinguishing the colon and rectal tumor origin. By comparing the expression profiles of rectum and colon tumors, we indeed observed their very similar expression patterns, although small differences were noticed. Six miRNAs were specific for colon cancer, whereas miRNA 744 was upregulated exclusively in RC. On the other hand, miR-17, -18a/b, -20a/b, -21 and -106a were shown by us and by others to be commonly upregulated in all CRC (10,33,34). Apart from those, we identified two additional members of large miR-17/92 cluster, namely miR-19a and -19b. Our data, in the context of existing literature, highlight the fact that upregulation of miR-17/92 cluster is an important event in CRC development irrespectively on the location of the adenocarcinoma.
In the past years, miRNAs were tested as biomarkers of early cancer onset or treatment response, with promising results. Therefore, we evaluated miRNA profile of rectal tumors with clinical characteristics of patients and their treatment responses. miR-21 was overexpressed to a higher extent in advanced tumors with metastases (TNM stage IV). This observation in RC seems to be in agreement with several studies showing that high miR-21 levels are present in CRC with more advanced TNM and associated with worse therapeutic outcome and patient’s prognosis (35–38). The negative correlation of miR-21 expression with mortality has also been reported for RC alone (33,39). More interestingly, all eight members of miR-17/92 cluster identified in our study as being upregulated in RC were also significantly associated with the response of patients to adjuvant treatment. Levels of miR-17, -18a/b, -19a/b, -20a/b and -106a were highly upregulated in tumors exhibiting posttreatment relapse compared with those at the remission. Apart from miR-17/92 cluster, also expressions of miR-133a/b, -1, -143 and -195 were associated with treatment response, but in an opposite direction. None of the deregulated miRNAs correlated with treatment response of colon tumors, yet we identified a panel of miRNAs with predictive value specific for RC. To investigate whether this miRNA panel might be used as a non-invasive biomarker, miR-17/92 cluster expression profile was analyzed in peripheral blood. Cell-free plasma and plasma exosomes of RC patients were examined at the diagnosis and, to be able to associate miRNA profiles with treatment outcomes, 6 and 12 months later, spanning thus the period of chemotherapy and recovery. Exosomal, but not whole plasma levels of all miRNAs differed between healthy controls and patients at the diagnosis and were no longer altered in patients a year since diagnosis. Bearing in mind that the last sampling involved only patients at remission, we concluded that miRNA expression profile in plasma exosomes rather than in whole plasma reflects the course of the disease. Exosomes are cell-derived vesicles, present in body fluids that play an important role in intercellular communication. They are actively secreted by tumor cells to influence the surrounding microenvironment. This was previously documented for miR-17/92 cluster of miRNAs released from leukemia cells to have an impact on recipient cells (40,41). It has also been demonstrated that profile of miRNA is different between cell-free and exosomal components of blood. Exosomes offer better protection of miRNAs from degradation and, therefore, are miRNA enriched compared with cell-free plasma. As such, they provide a consistent source of miRNA for disease biomarker detection (42). For that reason, exosomal miRNAs have recently been tested as biomarkers of cancers, including CRC, with promising results (43,44). Here, we show for the first time that both tumor and exosomal levels of miR-17/92 cluster members associate with RC recurrence. To further explain our observation at a cellular level, we overexpressed miRs-18a/b and -19a/b in SW1463 RC cells and followed their proliferation and migration rates. In this context, it should be noted that posttransfection expression levels of miR-18b were much higher than those for other three studied miRNAs, and so that the positive observation might have been enhanced by this fact. Nevertheless, miR-18b upregulation related with the growth advantage of RC cells can contribute to the understanding of why is miR-18b and perhaps other members of miR-17/92 cluster so commonly overexpressed in CRC. In support to this notion, miR-19a was recently shown to promote proliferation and also migration in colon cancer cells (45). In our study, elevated levels of miR-18a or -19a had an inhibiting effect on RC cell migration. However, this seems to be also in contrary to what we observed in human samples. High expression levels were associated with tumor relapse proposing thus rather positive effect on migratory activity of cancer cells. What can mechanistically underlie miR-17/92 regulation of the cellular proliferation? One of the plausible explanations might rise from the knowledge that miR-17/92 inhibits translation of E2F1-3 transcription factors and they can, in turn, induce the expression of miR-17/92 cluster (46). The feedback system between miR-17/92 cluster and E2F provides a mechanism to keep regular cell-cycle progression under physiological conditions. However, overexpression of miR-17/92 in cancer disrupts the feedback loop to promote cell proliferation.
In conclusion, our study provides evidence that miRNA expression profiling is an informative molecular approach for selection of groups of patients with different sensitivities to chemotherapy. To the best of our knowledge, this is the first report of miRNA profile, detectable in tumor and in plasma exosomes, that predicts long-term therapeutic outcomes in RC patients. Identified list of miRNAs, if validated in an independent and perhaps larger study group, might be of substantial significance for RC treatment management.
Funding
The Czech Science Foundation (GACR 15-08239S) to J.S.; the Grant Agency of the Ministry of Health of the Czech Republic (AZV 15-26535A, 17-30920A) to V.V.; Ministry of Education, Youth and Sport of the Czech Republic (MEYS) (LQ1604 NPU II) and European Regional Development Fund and MEYS (CZ.1.05/1.1.00/02.0109 BIOCEV) to V.K.; project ‘Centrum of clinical and experimental liver surgery’ (UNCE/MED/006) to V.L. provided by the Charles University and National Sustainability Program I (NPU I) (LO1503), Ministry of Education, Youth and Sport of the Czech Republic.
Acknowledgements
The authors are thankful to all volunteers who contributed their biological material to the study and to all hospital employees who participated in samples collection.
Conflict of Interest Statement: The authors declare no conflict of interest.
