Locked Nucleic Acid In situ Hybridization Analysis of miR-21 Expression during Colorectal Cancer Development

Novel mechanisms of human gene regulation mediated by microRNAs (miRNA) have recently been established (1). MiRNAs control gene expression at the posttranscriptional level by targeting mRNAs for translational repression or mRNA degradation (1). Evidence is accumulating that many miRNAs are differentially regulated in normal development and cancers, and that deregulation of specific miRNAs may contribute to human cancer development and progression (1). To better understand the molecular mechanisms that regulate miRNA gene expression, our group has recently developed an algorithm to predict the promoters of human miRNA genes that are likely to be involved in evolutionarily conserved regulatory systems (2). Using this algorithm, we predicted 59 putative miRNA gene promoter regions, one of which was the miR-21 miRNA promoter. Because high miR-21 expression levels have been reported in various malignancies, including breast cancer (3–5), pancreatic cancer (3, 6, 7), cholangiocarcinoma (8), hepatocellular carcinoma (9), brain tumor (10), leukemia (11, 12), cervical cancer (13), ovarian cancer (14), colorectal cancer (3, 15), prostate cancer (3), lung cancer (3), and gastric cancer (3), we concentrated on miR-21 analysis. Our biochemical analyses confirmed that the miR-21 promoter predicted by the algorithm was correct (2). Transcription of primary miR-21 transcripts (pri-mir-21) was initiated 30 bp downstream of the promoter TATA box. Other evolutionally conserved regulatory elements present in the miR-21 promoter included the binding sites of activator protein (AP-1), Ets/PU.1, SP1, CCAAT/enhancer binding protein-α, nuclear factor I (NFI), serum response factor, p53, and signal transducer and activator of transcription 3. We further showed that phorbol 12-myristate 13-acetate activated miR-21 transcription through the several AP-1 and Ets/PU.1 binding sites in the miR-21 promoter, and also showed that miR-21 was induced during macrophage differentiation of HL-60 cells (16). Therefore, we hypothesized that increased miR-21 expression, which has been validated in many cancers, may reflect elevated tumor cell AP-1 activity. We also identified the negative transcriptional regulator nuclear factor I-B (NFIB) as a target for miR-21 regulation (16). Because the miR-21 promoter contains an NFIB binding site and NFIB efficiently suppresses AP-1–dependent miR-21 transactivation, we concluded that a double-negative feedback loop consisting of miR-21, NFIB, and the miR-21 promoter could self-reinforce miR-21 expression (16).

Several other miR-21 targets have also been suggested such as RECK (17), Sprouty2 (18), PTEN (8, 9), TPM1 (19), and PDCD4 (20–23). Tumor suppressor PDCD4 is reported to suppress protein synthesis (24, 25) through cytoplasmic binding of eIF4A and is also known to inhibit several transcription factors, including AP-1 (26, 27), SP1 (28), and β-catenin (26). Because DNA binding sites for some of these factors are present in miR-21 promoters, PDCD4 might also contribute to the down-regulation of miR-21 expression through double-negative feedback regulation. Indeed, it was recently reported that in Ras-induced cellular transformation of a rat thyroid cell line, PDCD4 reduces miR-21 activity, at least partly, by suppressing AP-1 activity (29). However, at this moment, there is even no clear proof that PDCD4 is a target of miR-21 in vivo because of the lack of efficient tools to detect miR-21 expression in a single-cell resolution.

As a first step toward understanding mechanisms regulating miR-21 expression in vivo, we examined miR-21 expression patterns during the cancer development. Colorectal tumors were chosen for the experimental analyses because the clinical and histologic features of colorectal cancers are relatively simple compared with other malignancies, and most of colorectal cancers are thought to develop from precancerous adenomas (30). For this purpose, we established a sensitive and stable in situ hybridization (ISH) method using formalin-fixed, paraffin-embedded (FFPE) tissues. Experimental results indicate that the frequency and extent of miR-21 expression increase during colorectal cancer progression from precancerous adenoma to advanced carcinoma.

Human cell lines. HEK293 (originated from embryonic kidney), HeLa (cervical carcinoma), MDA-MB435 (breast ductal carcinoma), PA-1 (embryonic carcinoma), NCC-IT (embryonic carcinoma), A427 (non small cell lung carcinoma), G401 (rhabdoid tumor), SW620 (colorectal adenocarcinoma), and AZ521 (gastric cancer) cells were maintained at 37°C in high-glucose DMEM supplemented with 10% FCS (Gibco/Invitrogen).

Locked nucleic acid–modified oligonucleotide probes. Locked nucleic acid (LNA)–modified oligonucleotide probes labeled with FITC at their 3′-ends were obtained from Molecular Biology of ThermoElectron GmbH. The sequences of miR-21 probe and the scramble control probe for a negative control were 5′-TLcAALcATLcAGLtCTLgATLaAGLcTA-3′ and 5′-CLaTTLaATLgTCLgGALcAALcTCLaAT-3′, respectively. La, Lt, Lc, and Lg were LNA monomers corresponding to the bases A, T, C, and G, respectively.

ISH with LNA-modified oligonucleotide probes. Five-micrometer-thin sections of FFPE tissues adhered to glass slides were deparaffinized in three consecutive xylene baths for 1 min each, followed by 1 min each in serial dilutions of ethanol (100%, 100%, 95%, 95%) and three changes of diethyl pyrocarbonate–treated water. Slides were then immersed in 0.3% H₂O₂ for 30 min at room temperature, washed thrice with diethyl pyrocarbonate–treated water, digested with 400 μg/mL proteinase K (Roche) at 37°C for 15 min, washed thrice with diethyl pyrocarbonate–treated water, submerged in 95% ethanol for 1 min, and air-dried completely. Slides were then hybridized in incubation chambers overnight at 37°C in an oven, using 0.2 μmol/L LNA–modified probes diluted with mRNA ISH solution (DAKO). After hybridization, slides were rinsed thrice in 0.5× SSC, washed for 30 min at 50°C in 0.5× SSC/0.1% Brij35 (Sigma), and rinsed twice in TBS. An anti-FITC horseradish peroxidase–conjugated antibody (DAKO, P5100) at 1:100 dilution in TBS/1% bovine serum albumin was applied to the slides for 60 min at room temperature, followed by three washes in TBS/0.1% Tween 20 (TBS-T). For amplification of antibody signals, FITC-conjugated phenol (fluorescyl-tyramide, DAKO, K1497) was applied to the slides for 30 min at room temperature, followed by three washes in TBS-T. Finally, an anti-FITC antibody conjugated to horseradish peroxidase (DAKO, K1497) was added to the slides for 30 min at room temperature, followed by three washes in TBS-T. The reaction products were visualized using a 50 mg/dL 3,3′-diaminobenzidine tetrahydrochloride solution containing 0.003% hydrogen peroxide.

Immunohistochemical staining. Deparaffinization, endogenous peroxidase inactivation, antigen retrieval of FFPE clinical tissues, and immunostaining with anti-PDCD4 (ab51495, Abcam) antibody were done as described previously (31). The immunostained sections were evaluated independently by two pathologists in conjunction with the H&E-stained sections from the same lesions.

Lentivirus vectors. For the polymerase II–driven (SV40 promoter) vector, a part of pri-miR-21 sequence (from −6 bp to +65 bp of the 5′-end of mature miR-21 sequence) was inserted between the BamHI and EcoRI sites of pLSP (22) to generate pLSP-pre-mir21_short. For the polymerase III–driven (U6 promoter) vector, the mouse U6 promoter and a part of the pri-miR-21 sequence (from −57 bp to +115 bp of the 5′-end of mature miR-21 sequence) were inserted into the BamHI and EcoRI sites of pLSP to generate pLSP-mU6-pre-mir21_long. Sequences of synthetic oligonucleotide pairs for shCre (targeting Cre-recombinase of P1phage) and shU3-12 (targeting part of MuLV-LTR) were described previously. They were inserted into pLPS as described above to generate pLPS-shCre and pLSP-shU3-12 (32), respectively, and used for the negative controls. Vesicular stomatitis virus-G pseudotyped lentiviral vectors were produced using the ViraPower Lentiviral Expression System (Invitrogen) according to the manufacturer's instructions.

Virus transduction and protein analysis. HEK293 cells were transduced with vesicular stomatitis virus-G pseudotyped vectors at multiplicity of infection of 1 to 2 and selected with puromycin for 1 wk. About 1 wk after the end of selection, cells were disrupted for total protein preparation. Western blotting analysis with anti-PDCD4 and anti-actin antibodies was done as described previously (22).

Colorectal tissue samples. Thirty-four endoscopically resected colorectal polyps and 39 surgically excised colorectal tumors were selected from a list of patients with colorectal lesions who underwent endoscopic or surgical operation in 2006. All tissue samples were banked at the Fujita Health University School of Medicine, Aichi, Japan. For the endoscopically resected samples, the patients were between 35 and 83 y old (mean age, 58.8 ± 12.0 y) and included 23 males and 11 females. For the surgically excised samples, the patients were between 41 and 83 y old (mean age, 65.7 ± 9.4 y) and included 23 males and 16 females. Among the 35 cases of surgically removed colorectal cancer, the clinical tumor stage distribution was stage I in 1 patient, stage II in 16 patients, stage III in 15 patients, and stage IV in 2 patients, according to the International Union Against Cancer classification. This study was approved by the institutional ethical review board for human investigation at Fujita Health University.

Detection of miR-21 expression levels by ISH with LNA-modified oligonucleotide probes. To screen for miR-21–positive and miR-21–negative control cultures for ISH, levels of primary transcript of miR-21 (pri-miR-21) and mature miR-21 were determined (Supplementary Fig. S1A and B). Expression level patterns of pri-miR-21 and mature miR-21 were very similar, suggesting that there were no significant rate-limiting steps in the processing of miR-21 production among these cell lines. Among the nine cell lines tested, we used malignant HeLa and MDA-MB435 cells as positive controls and nonmalignant HEK293 cells as a negative control.

A LNA-modified probe was designed in which every third DNA nucleotide was substituted with a corresponding LNA monomer, which was subsequently used for ISH with LNA-modified oligonucleotide probes (LNA-ISH). With fluorescent microscopic observation, the FITC-labeled probe clearly stained the cytoplasm of HeLa and MDA-MB435 cells, whereas HEK293 cell cytoplasmic staining was much weaker (Fig. 1A). Because mature miRNAs are present in the cytoplasm (33), the probe most likely detected the mature miR-21 signals. To apply this LNA-ISH to clinical FFPE samples, we also attempted to detect miR-21 by combining LNA-ISH with the biotin-free tyramide signal amplification system. Using this method, 3,3′-diaminobenzidine staining of miR-21 was visible in the cytoplasm (brown; Fig. 1A).

To confirm the specificity of miR-21 detection and more precisely quantify miR-21 expression levels, we produced two lentiviral vectors carrying a portion of pri-miR-21 (including the entire pre-miR-21) driven by the SV40 promoter (polymerase II–driven pLSP-pre-mir21_short) or the mouse U6 promoter (polymerase III–driven pLSP-mU6-pre-mir21_long; Supplementary Fig. S2). HEK293 cells were transduced with pLSP-pre-mir21_short, pLSP-mU6-pre-mir21_long, and pLSP-shGFP (control), respectively, and stable transductants were selected for further analyses. Expression of miR-21 in these established cell lines was analyzed (Supplementary Fig. S3), and LNA-ISH for miR-21 was done. As shown in Fig. 1B, HEK293 cells transduced with miR-21–expressing vectors exhibited a clear cytoplasmic staining when compared with control cells, and pLSP-mU6-pre-mir21_long–transduced cells showed denser staining than pLSP-pre-mir21_short–transduced cells. From these results, we concluded that our LNA-ISH could detect miR-21 in a semiquantitative manner.

High miR-21 expression levels were observed not only in cancer cells but also in cancer-associated fibroblasts from colorectal FFPE tissues. We first performed the LNA-ISH for miR-21 using the surgically excised advanced colorectal cancer tissues. In the 34 slides examined, precancerous adenomatous lesions were found in five cases: three were separate polyps and two were adenomatous masses adjacent to malignant adenocarcinoma. Including these five lesions, we performed the LNA-ISH for miR-21 on 39 lesions (Supplementary Table S1). The expression of miR-21 in adenocarcinoma was much higher than the expression in normal mucosa (Fig. 2; Table 1). In the surgical samples, however, increased miR-21 expression was barely detectable in precancerous adenomas. Unexpectedly, up-regulation of miR-21 was observed not only in malignant cells but also in the stromal fibroblasts adjacent to the tumor (Fig. 2B; Supplementary Fig. S4A; Supplementary Table S1). Overexpression of miR-21 was never observed in fibroblasts far from the tumor mass (Fig. 2B). When an equivalent LNA probe but in which the miR-21 oligonucleotide sequence has been scrambled was used instead, this control probe showed no significant staining in tumor regions, normal tissue, and the stromal fibroblasts adjacent to the tumor (Supplementary Fig. S5B and D), showing clear contrast to the staining by the miR-21 probe in the sequential FFPE colorectal tissue sections (Supplementary Fig. S5A and C). All these results suggest that nonmalignant stromal fibroblasts adjacent to tumors might induce miR-21 expression due to factors secreted from the nearby tumors.

Up-regulation of miR-21 was frequently observed in precancerous adenomas but never in nontumorigenic polyps. We next analyzed miR-21 expression levels in endoscopically resected colorectal polyps that were <15 mm in diameter. Among the 34 samples examined, 28 cases were tumorigenic and 6 cases were nontumorigenic (Table 2; Supplementary Table S2). The ISH signal intensities of these endoscopically resected adenomas were far stronger than those of the five surgically excised adenomas (Supplementary Table S1; Table 1). Importantly, the RNA preservation of endoscopically resected tissues was much better than that of surgical excised tissues. This difference could be due to a longer time before immersion in formalin for the surgical samples (1-3 hours) compared with that for endoscopic ones (1-10 minutes) and also to the the much longer formalin fixation time for surgical tissues (24-96 hours) compared with the fixation time for endoscopic ones (<6 hours).

Expression of miR-21 in nontumorigenic lesions was not elevated above levels detected in normal colorectal mucosa (Fig. 3A). On the contrary, increased miR-21 expression was frequently observed in both precancerous adenomas and adenocarcinomas (Fig. 3B and C). As shown in Table 2 and Fig. 3, miR-21 expression was obviously higher in malignant adenocarcinoma than in precancerous adenoma. From these results, we concluded that elevation of miR-21 expression accompanies colorectal tumor development from precancerous adenoma to advanced carcinoma. We also expect that early detection of miR-21 up-regulation may have potential clinical application as a new biomarker for colorectal tumorigenesis.

MiR-21 and PDCD4 expression show mutually exclusive patterns in the areas around colorectal cancer tissues. It has been reported that PDCD4 is a target of miR-21 mostly by transiently introducing miR-21 RNA or antisense oligonucleotides for miR-21 exogenously (20–23). Recently, we have also shown that PDCD4 protein is up-regulated when cells were transduced with lentivirus vectors expressing newly developed decoy RNAs that specifically inhibit miR-21 function (22). We here tested whether HEK293 cells, which show marginal miR-21 expression (Fig. 1), reduce endogenous PDCD4 expression by stable expression of exogenous miR-21. By comparing PDCD4 protein levels in cells transduced with miR-21 expression lentivirus vector (pLSP-mU6-pre-mir21_long) and those with control vectors (or untransduced cells) by Western blotting, we have observed clear reduction in the steady-state levels of PDCD4 by exogenous miR-21 expression (Fig. 2C).

We next performed ISH for miR-21 and immunostaining of PDCD4 using sequential sections obtained from surgically resected colorectal cancer tissues. As shown in Fig. 4A, expression of PDCD4 was high in normal tissues (right) but was nondetectable in colorectal cancer cells. Areas with abundant miR-21 expression wrapped over the areas with low PDCD4 expression. Even when endoscopically resected samples were used, mutually exclusive expression patterns were frequently observed between miR-21 and PDCD4 in both malignant adenocarcinoma (Figs. 4B; Supplementary Fig. S6A) and precancerous adenoma (Figs. 4C; Supplementary Fig. S6B). These results support the hypothesis that PDCD4 is an in vivo target of miR-21 and further suggest that an early increase in miR-21 expression during colorectal tumorigenesis results in a decrease in PDCD4 expression.

MiR-21 expression and cancer development. MiR-21 expression has been reported to be one of the best hit miRNA in many profiling experiments designed to detect up-regulated miRNA in human cancer including colorectal carcinoma (3). Our ISH analysis on the colorectal carcinomas clearly detected high level expression of miR-21 in most of them (Fig. 2; Table 1), and importantly, we further showed that this miR-21 increase can be frequently detected from the adenoma stage in the section of endoscopic mucosal resection (Fig. 3; Table 2). Even in this early stage, we observed that PDCD4, a target of miR-21, was concomitantly reduced in miR-21-up-regulated regions (Fig. 4C). These results indicate the importance of up-regulation of miR-21 and down-regulation of PDCD4 as diagnostic biomarkers of colorectal carcinogenesis.

Recently, extensive analysis on miRNA profiles that are associated with prognosis and therapeutic outcome in colon adenocarcinoma was reported (15). Importantly, more advanced tumors expressed higher levels of miR-21 using either microarray data from the test cohort or the quantitative reverse transcription-PCR data from the validation cohort. By analysis on pooled cohorts, they further showed that high miR-21 expression is associated with a poor prognosis in either stage II or stage III colon cancer patients, indicating its potential as a prognostic biomarker (15).

LNA-ISH combined with biotin-free tyramide signal amplification system is a useful technique for determination of miRNA expression levels in FFPE tissues. The LNA-modified oligonucleotide would be one of the most sensitive probes currently available for miRNA detection (34, 35). Nevertheless, it is quite difficult to detect miRNAs by ISH, especially in FFPE clinical tissues. It has been reported that the LNA-ISH technique could detect some miRNA species in FFPE samples (36). However, using LNA-modified probes alone or LNA-ISH combined with a universal immunoenzyme polymer method (Histofine Simple Stain MAX PO_MULTI purchased from Nichirei), we were not able to detect miR-21 expression in FFPE samples (data not shown). In our previous study (31), the biotin-free tyramide signal amplification system was used to detect the nuclear protein Brm, which is difficult to detect by immunohistologic methods (37). In the present study, we applied this method to LNA-ISH and were able to sensitively detect miR-21 expression. Because detection of nonspecific signals is not infrequent when using tyramide amplification, we confirmed that staining by the scramble control probe (Supplementary Fig. S5B) as well as staining unrelated to the LNA/DNA probe (Supplementary Fig. S4B) was only rarely detected.

High-level expression of miR-21 in cancer-associated fibroblasts may be induced by secreting factors originating from cancer cells. It is very interesting that the stromal fibroblasts around tumors frequently express miR-21 at high levels. In most cases from the present study, these cancer-associated fibroblasts showed strong miR-21 expression compared with distant normal fibroblasts and normal epithelial cells (Supplementary Table S1). In some cases, the fibroblast staining intensities were even more intense than those of adjacent malignant cells (Supplementary Fig. S4A). Therefore, we hypothesized that this is a non–cell-autonomous phenomenon and that cytokines secreted from the adjacent malignant tumors might contribute to miR-21 induction. In this regard, it has been reported that interleukin-6 levels are elevated in the cancer-associated fibroblasts around a colon cancer, in serum, and in tumor tissues from colorectal cancer patients (38, 39). It is also noteworthy that interleukin-6 induces the transcription of miR-21 in multiple myeloma cells through mediation of signal transducer and activator of transcription-3 activity (12). Importantly, signal transducer and activator of transcription-3 binding sites are present just upstream of the transcriptional start site in the miR-21 promoter. Therefore, interleukin-6 is a candidate for miR-21 induction in these cancer-associated fibroblasts.

MiR-21 RNA and PDCD4 protein expression patterns were mutually exclusive in colorectal epithelial cells. PDCD4 was highly expressed in normal colorectal epithelium, but PDCD4 expression was often reduced in precancerous colorectal regions (Fig. 4). This observation is consistent with a recent report that normal mucosa showed strong nuclear PDCD4, which was significantly reduced in adenomas (40). Our ISH analyses further indicated that cells with reduced PDCD4 expression frequently had elevated miR-21 expression, which was nondetectable in normal colorectal epithelium (Fig. 4). In progressive colorectal cancers, almost all cells expressed miR-21, whereas PDCD4 was almost undetectable. In summary of our ISH analysis, expression of miR-21 RNA and PDCD4 protein showed mutually exclusive patterns in colorectal epithelial cells. These observations support that PDCD4 is a good target of miR-21 in vivo. Because PDCD4 is a potent tumor suppresser, miR-21 may perform oncogenic functions, at least in part, through down-regulation of PDCD4.

We have previously shown that in human cell culture systems, a double-negative feedback loop operates between miR-21 and its target protein, NFIB, through the miR-21 promoter, which has the binding site of this negative transcriptional regulator (16). This means that the miR-21 gene integrates a system that self-reinforces its own expression. Whereas in adult rats NFIB mRNA is highly expressed (41), we are currently not able to perform specific immunohistochemical staining due to the absence of a specific anti-NFIB antibody that is applicable to FFPE clinical samples and is non–cross-reactive with other NFI family members. Therefore, direct evidence of NFIB involvement in vivo remains to be established. Using rat cell culture system, Dr. Verde's group very recently indicated that PDCD4 suppressed miR-21 promoter activity, at least in part, by inhibiting AP-1 activity (29). Therefore, these double-negative feedback loops operating through the miR-21 promoter may contribute to the self-reinforced expression of miR-21. This feedback could further lead to the mutually exclusive expression patterns observed between miR-21 RNA and PDCD4 protein in colorectal cancer. It also remains to be determined which transcriptional factors that are normally inhibited by PDCD4 are crucial for miR-21 gene induction in colorectal cancer. The molecular mechanisms involved in functional suppression of these transcription factors by PDCD4 could significantly advance our understanding of cancer progression. Therefore, elucidation of the molecular mechanisms involved in the entire regulatory network formed by miR-21 is important for understanding the initial stages of colorectal carcinogenesis.

No potential conflicts of interest were disclosed.