Concomitant dysregulation of microRNAs miR-151-3p and miR-126 correlates with improved survival in resected cholangiocarcinoma

Abstract

Background

MicroRNAs (miRNAs) are small non-coding genes which become dysregulated in cancer and may predict survival. The role of miRNAs in outcomes in cholangiocarcinoma (CC) has not been reported.

Methods

RNA was extracted from 32 resected CCs along with adjacent uninvolved bile duct epithelium. A total of 43 miRNAs were quantified using NanoString™. Clinicopathologic characteristics and outcomes were captured and compared. Overall survival curves were created using the Kaplan–Meier method; factors, including miRNA expression, were compared by log-rank, chi-squared or Cox regression analyses.

Results

Absolute expression of each miRNA was compared with overall survival after excluding perioperative deaths (n= 3). One upregulated (miR-151-3p; P= 0.003) and one downregulated (miR-126; P= 0.023) miRNA in resected CC relative to adjacent normal bile duct epithelium correlated with survival on univariate analysis. Clinical factors and these miRNAs were compared. Dysregulated miR-151-3p and miR-126, respectively, were the only factors that correlated with improved overall survival [41.5 months vs. 12.3 months (P= 0.002) and 21.9 months vs. 15.1 months (P= 0.02), respectively]. In eight patients, both miRNAs were dysregulated. In the remainder, only one or neither showed dysregulation. Concomitant dysregulation correlated with the best overall survival (58.7 months vs. 15.1 months; P < 0.000; n= 8); clinicopathologic factors in these groups were otherwise similar.

Conclusions

In resected CC, the concomitant dysregulation of both miR-151-3p and miR-126 was the factor related to the greatest improvement in overall survival. Further analysis of the targets of these miRNAs may yield potential therapeutic targets or prognostic biomarkers.

Introduction

Cholangiocarcinoma (CC) is a primary hepatic malignancy originating from bile duct epithelium. It is the second most common primary hepatic neoplasia and, although rare, accounts for approximately 3% of all gastrointestinal malignancies.¹ Over 3500 cases will be diagnosed in the USA this year alone. Further, the worldwide incidence of CC and associated mortality rates are increasing.² Hepatobiliary cancer accounts for 13% of the 7.6 million cancer-related deaths worldwide each year; of these, CC accounts for 10–20%.³ Because of its poor prognosis, incidence and prevalence are essentially the same.

Risk factors for CC include chronic inflammation, primary sclerosing cholangitis (PSC), infestation with liver flukes, congenital disorders (e.g. choledochal cysts), hepatolithiasis, viral hepatitis, and lifestyle choices (i.e. ethanol use, obesity, smoking). Most patients with CC have no identifiable risk factors. Effective screening tools for CC are lacking and the disease remains difficult to diagnose. In patients identified as high risk (i.e. patients with chronic viral hepatitis and PSC), there are no reliable screening methods. Even histopathology can be deficient in making a diagnosis after resection or biopsy.

The only definitive curative therapy is surgical resection in those patients without distant or locally advanced disease. Despite aggressive resection, overall survival is poor and 5-year survival rates range from 20% to 35%.^4–9 In patients with advanced disease, survival remains limited.¹⁰ In patients with advanced disease, the standard of care includes gemcitabine-based chemotherapy or enrolment in a clinical trial. Growing understanding of the molecular and cellular aetiology of this disease is enabling the development of novel targeted therapies. However, little is known about the pathogenesis of CC.

MicroRNAs (miRNAs) are small (20–22 nucleotides), non-coding RNA fragments that have critical functions in various biological processes. To date, over 1000 miRNAs have been reported and shown to play a role in cell proliferation, apoptosis and differentiation. They are linked to oncogenesis through their function as oncogenes or tumour suppressors.¹¹

The present authors have noted anecdotally that a subset of patients undergoing resection for CC appear to do better than others for unclear reasons. Therefore, this group hypothesized that key miRNAs can be utilized to characterize a subset of patients with resectable CC in whom overall survival will be improved and that these miRNAs will thus be prognostic. Further, these miRNAs modulate protein expression involved in tumour growth and may serve as potential targets for future therapies.

Materials and methods

Institutional review board approval was obtained for the retrieval of data from the archival files of the Department of Pathology at this institution. A total of 135 patients presenting with CC between 1993 and 2007 were subsequently identified. Of these, 90 underwent surgical exploration. Of the 69 who underwent resection with curative intent, tissue samples for 32 patients were of adequate volume for RNA extraction. Three 2-mm cores of tumour and matched adjacent benign bile duct epithelium were punched from paraffin blocks. Deparaffinization using 100% xylene at 50 °C for 3 min was followed by RNA extraction using the RecoverAll™ Total Nucleic Acid Isolation Kit (Ambion®, Life Technologies, Corp., Grand Island, NY, USA) according to the manufacturer's guidelines. Calculation of the miRNA using the NanoString nCounter® Analysis System (NanoString Technologies, Inc., Seattle, WA, USA) was performed according to the manufacturer's protocol. This method is described elsewhere.¹² Selected miRNAs were validated by reverse transcription polymerase chain reaction (RT-PCR) using standard protocols and are reported elsewhere.¹³

Statistical analysis

NanoString nCounter® provides a quantitative assessment of miRNA expression. Dysregulated miRNA expression was defined as aberrant expression compared with that in normal adjacent bile duct epithelium. The expression of each dysregulated miRNA was dichotomized as ‘high’ or ‘low’ relative to the mean tumour expression of each miRNA. These data were then correlated with clinical outcomes.

The Kaplan–Meier method was used to construct overall and progression-free survival curves. Time-to-event variables were calculated from the time of resection. Log-rank analysis was used to compare groups. Perioperative deaths were considered as deaths that occurred within 60 days of resection or during the index admission and were excluded from all survival and recurrence analyses. Progression was determined by the date when either intra- or extrahepatic progression of disease was noted radiographically. Categorical variables were analysed using Pearson's chi-square or Fisher's exact test as appropriate. Student's t-test was used to analyse continuous variables. All tests were examined at a two-tailed significance level of 0.05. Cox regression multivariate analysis was used to compare factors associated with survival. All factors were included and analysed in a forward stepwise fashion. Logistic regression multivariate analysis was used to compare associations with the differential expression of miRNAs of interest. All factors were included and analysed in a forward stepwise fashion; however, because data for carbohydrate antigen 19-9 (CA 19-9) were missing for 12 patients, this factor was excluded from the final analysis.

Results

A total of 32 patients with CC resected with curative intent for whom adequate pathologic specimens were available for miRNA analysis were identified. Patients were, on average, in their seventh decade (Table 1). The majority were female and had extrahepatic tumours, eight of which were hilar. Almost half of the patients had elevated CA 19-9 and/or obstructive jaundice preoperatively (Table 1).

Table 1.

Demographic data for patients undergoing resection for cholangiocarcinoma (n= 32)

Characteristics	Data
Age, years, median (range)	66 (47–85)
Gender, male, n	12
Tumour location, n	Intrahepatic, 11
	Hilar, 8
	Distal, 13
Elevated CA 19-9, n	15
CA 19-9, mL, median (range)	51.3 (<15–9000)
Elevated bilirubin, n	13
Adjuvant chemotherapy, n	10/29

CA 19-9, carbohydrate antigen 19-9.

The majority of the patients (n= 20, 63%) were classified as having stage I or II disease (American Joint Commission on Cancer, Cancer Staging Manual, 7th edn).¹⁴ Thus, only about one third of patients underwent adjuvant chemotherapy (Table 1). Two thirds of the tumours were moderately differentiated and showed perineural invasion (PNI) (Table 2). Only one third of patients demonstrated lymphovascular invasion (LVI) or nodal disease. The majority of the patients had negative resection margins (Table 2).

Table 2.

Perioperative and pathologic factors in patients undergoing resection for cholangiocarcinoma (n= 32)

Factors	Patients, n
AJCC cancer stage14
Stage I	9
Stage II	11
Stage III	7
Stage IV	5
Grade 2 disease	20
Nodal disease present	9
Negative margins	18
Fibrosis	16
Perineural invasion	21
Lymphovascular invasion	13
Complications	7
Mortality	3

AJCC, American Joint Commission on Cancer.

Seven patients (22%) had perioperative complications, including multi-organ system failure (n= 1), sepsis (n= 2), anastomotic leak or bleeding (n= 3) and bronchobiliary fistula (n= 1) (Table 2). Three patients (9%) died perioperatively, one of whom was the patient with multi-organ failure. The other two died at >30 days after discharge of unknown causes. Median overall survival was 17.8 months. Seven patients were lost from follow-up. Of the remaining 22 patients, 16 (73%) patients experienced disease recurrence at a median of 17.0 months (range: 6.6–89.0 months). In the majority of patients (n= 15, 68%), recurrence was either intrahepatic or in the porta hepatis. Only one patient, who has been followed for 89 months, survived to the completion of this study.

Expression of 43 miRNAs in the study specimens was found to be dysregulated in comparison with that in adjacent normal bile duct epithelium. Of these, 17 were upregulated and 26 were downregulated. When these data were correlated with outcomes, dysregulations of two miRNAs were found to be associated with improvements in overall survival from the time of resection. High expression of miR-151-3p was associated with improved median survival of 41.5 months compared with 12.3 months (P= 0.004). Low expression of miR-126 was associated with improved median survival of 21.9 months compared with 15.1 months (P= 0.033). A stronger association with overall survival emerged when both miRNAs were dysregulated (58.7 months vs. 15.1 months; P < 0.000) (Fig. 1). These miRNAs did not predict recurrence or time to recurrence (P= 0.71).

Figure 1.

High expression of miR-151-3p and low expression of miR-126 predict overall survival after resection for cholangiocarcinoma (n= 9)

When other clinicopathologic factors were analysed individually for association with overall survival, the only factors found to be prognostic were dysregulations of miR-126 and miR-151-3p (Table 3). When all clinicopathologic factors were analysed in a multivariate stepwise fashion, only differential expression of both miRNAs correlated with an improvement in overall survival (P= 0.04, hazard ratio 0.201, 95% confidence interval 0.043–0.928). The absence of LVI correlated with the differential expression of both miRNAs when clinicopathologic factors were compared (Table 4). When all factors were analysed in a forward stepwise fashion, no factors appeared to have an independent association; however, the trend towards an association with the absence of LVI remained.

Table 3.

Clinicopathologic factors prognostic of overall survival

Factors	Univariate analysis	Multivariate analysis
High miR-151-3p expression	0.004	0.944
Low miR-126 expression	0.033	0.434
Dysregulation in both miRNAs	0.000	0.040
Tumour size	0.674	0.613
Nodal metastasis	0.122	0.359
Margins	0.368	0.217
Tumour location	0.311	0.791
Grade	0.324	0.213
Fibrosis	0.767	0.229
Elevated bilirubin	0.546	0.690
Elevated CA 19-9	0.875	0.909
Perineural invasion	0.771	0.648
Lymphovascular invasion	0.548	0.132
Adjuvant chemotherapy	1.000	0.756

CA 19-9, carbohydrate antigen 19-9.

Table 4.

Clinicopathologic factors correlated with concordant dysregulation of both miR-151-3p and miR-126 (n= 29)

Factors	Univariate analysis	Multivariate analysis
Age <70 years	0.681	0.412
Gender	1.000	0.769
Tumour size	0.116	0.441
Nodal metastasis	1.000	0.891
Margins	0.704	0.516
Tumour location	0.129	0.288
Grade	0.359	0.365
Fibrosis	1.000	0.822
Elevated bilirubin	0.856	0.916
Elevated CA 19-9	0.608	a
Perineural invasion	0.938	0.412
Lymphovascular invasion absent	0.050	0.053

^aElevation of CA 19-9 was excluded from the analysis because these data were missing for 12 patients.

CA 19-9, carbohydrate antigen 19-9.

Discussion

It has become clear that miRNAs are important in normal physiologic processes and even more important in cancer. Often intronic, these non-coding fragments of RNA are capable of affecting the post-transcription regulation of up to hundreds of genes. They are highly conserved among species and their small size makes them remarkably resistant to degradation. Thus, it is possible to utilize small tissue samples from tumour types that are inherently difficult to study. The present study utilized the latest NanoString™ technology to identify a set of dysregulated miRNAs and, for the first time, identified two miRNAs, miR-151-3p and miR-126, with prognostic potential. These two miRNAs, particularly when concordantly dysregulated, are the only factors prognostic of survival in this sample of patients.

In this study, all available paraffin-embedded blocks containing tissue from patients who had undergone resection with curative intent and were recorded in the study institution's database were retrieved. Given the scarcity of availability and the inherent difficulty of extracting RNA from paraffin-embedded tissue, only 32 samples were found to include adequate RNA for analysis. Nonetheless, a set of 43 miRNAs that showed significant dysregulation in tumour tissue relative to adjacent normal ductal epithelium was identified. Of these, dysregulations of miR-151-3p and miR-126 were prognostic of overall survival, although they did not correlate with other typical clinicopathologic characteristics. The association with overall survival was particularly striking when both miRNAs were dysregulated and appeared to be an independent predictor of outcome. This novel expression profile and its relation to survival may aid in further identifying key miRNAs in the development of CC and elucidate the heterogeneous population that develops CC.

Similarly, other investigators have distinguished potential clinically important miRNAs in CC. One investigation demonstrated a cluster of 38 miRNAs distinguishing CC from normal tissue and demonstrated that the exogenous expression of miR-320 or miR-204 could negatively regulate Mcl-1 or Bcl-2 expression and facilitate chemotherapeutic drug-triggered apoptosis.¹⁵ Mcl-1 protein expression also appeared to be regulated by miR-29 in malignant cholangiocytes.¹⁶ Additionally, in cholangiocytes, miR-21 has been shown to be overexpressed in CC and may be oncogenic by inhibiting TIMP3, a potential tumour suppressor gene in the biliary tree,¹⁷ or modulating gemcitabine-induced apoptosis.¹⁸ One group investigated the expression of miRNAs in hepatocellular carcinoma (HCC) and CC related to clinical factors.¹⁹ They found several miRNAs to be associated with tumour stage, prognosis and cirrhosis in HCC, but did not find the same associations in CC.¹⁹ Finally, miR-9 identified in bile may be a diagnostic marker of CC.²⁰

Certainly, these miRNAs may play a protective role in the development of CC as this subset of patients does significantly better than others undergoing resection. The present authors are unaware of any studies investigating this potential. By contrast, one study of the effects of miR-151 in HCC has shown that miR-151-5p significantly increases migration and invasion by directly targeting a metastasis suppressor.²¹ This study did not find the same effect with miR-151-3p.²¹

Several authors have demonstrated the association between miR-126 and carcinogenesis.^22–26 Many of these studies have shown that a downregulation of miR-126 induced carcinogenesis and that miR-126 is a tumour suppressor itself.^22–26 These findings contrast with those of the present study. Low expression of miR-126 in the tumour samples in the present series correlated with improved median survival and may be protective in terms of overall survival. In concordance with the present findings, one investigation demonstrated that expression of miR-126 was lower in gastric cancer than in normal surrounding tissue, but this was not correlated with outcomes.²⁷ Additionally, miR-126 has been shown to play a role in inflammation and, more specifically, angiogenesis during inflammation. This role may also influence angiogenesis in tumour cells.²⁸ Just as there appears to be variation in the level of expression of miR-126 and its role in oncogenesis, there is also variation in the proposed downstream targets. It is this heterogeneity in the expression and action of miRNAs that warrants further investigation into their utility as potential cancer-specific therapeutic targets.

The correlation between an absence of LVI and the presence of miRNAs may be of additional prognostic importance. Many authors have shown that in tumours of the pancreatic and biliary tree, the presence of LVI portends a poor prognosis.^29,30 Although LVI was not prognostic in the present cohort of patients, the absence of LVI was found to correlate with the miRNAs associated with improved prognosis.

Given the retrospective nature of the present study and its use of archival tissue, its findings should be interpreted with caution, although they do offer a platform for future research. Although the current findings of the impact of dysregulation in these two miRNAs on overall survival are striking, the present sample size and the fact that selection was dependent on the harvesting of adequate RNA introduce the potential for unavoidable bias. Therefore, the present authors plan to validate these findings prospectively. Once validated, these miRNAs may be useful in stratifying patients for clinical trials and in determining which patients might benefit from adjuvant therapies. Further, the present authors are investigating target proteins of these miRNAs in immunohistochemistry to identify potential prognostic biomarkers that may also be useful in these contexts.

Conflicts of interest

None declared.