Abstract
While previous studies have described associations between specific microRNAs and colorectal cancer (CRC) metastasis, our understanding of microRNA regulation of metastatic spread remains largely unexplored. To identify microRNAs critical for disease progression, we measured microRNA expression in primary CRC tumors and synchronous liver metastases in 19 cases using quantitative polymerase chain reaction (qPCR) arrays. We identified 16 microRNAs significantly differentially expressed between primary tumors and liver metastases that distinguish primary tumors and liver metastases by hierarchical clustering. Combinations of microRNAs expressed in the primary tumor and in the metastatic tumor are associated with survival, but these signatures have no microRNAs in common. We found that increased expression of miR-210 and miR-133b in liver metastases compared to primary tumors is associated with lower survival. We propose that evaluating the change in expression between primary and metastatic tumors in each patient may lead to improved biomarker development.
Colorectal cancer (CRC) is a leading cause of cancer-related mortality worldwide (1). Tumor recurrence remains a major challenge. Metastasis is the lethal form of the disease that needs to be prevented or treated. Although significant progress has been made characterizing the molecular features of primary tumors, we have a relatively poor understanding of metastases. Colorectal tumors often metastasize to the liver and the extent of liver infiltration is a major determinant of survival (2). Sequencing reveals that liver metastases may be genetically distinct from primary CRC tumors (3). However, tumors readily adapt to new microenvironments and many of these changes are reflected in changes in microRNA expression, as we have found in ovarian cancer (4).
MicroRNAs are emerging as powerful biomarkers due to their high tissue and environmental specificity (5). The microRNA environment of metastatic disease is relatively unexplored. Unfortunately, patients who present with metastatic colorectal adenocarcinoma have an almost uniform poor outcome and subsequently succumb to their disease. After lymph nodes, the liver is the next most frequent site of involvement.
Analogously to proteins, the set of microRNAs regulating metastasis are often not the same as those found to be differentially regulated relative to normal tissues (4). A number of microRNAs are reported to be associated with survival in CRC (6, 7). Most studies have focused on expression in primary tumors under the hypothesis that the nature of the primary tumor sets the path for future disease progression (8).
We examined the differential expression of microRNAs in primary CRC tumors with synchronous liver metastases, and analyzed microRNA expression in association with overall survival. Our goal was to gain new insights into metastasis, the form of the disease that is lethal to patients. One major problem in developing tumor biomarkers is the significant heterogeneity, both intra-tumorally and between patients, as well as the myriad of disease sub-types. We hypothesize that improved understanding of metastasis could lead to new biomarkers by evaluating the molecular changes occurring during metastasis. These changes may be more indicative of the type of late-stage disease to be treated.
Materials and Methods
Case selection. The Lifespan Hospital System pathology archives were searched for cases of colon adenocarcinoma that presented with histologically-confirmed liver metastases. Selection of patients and analysis was approved by the Rhode Island Hospital Institutional Review Board, approval #206011. Twenty-seven cases from Rhode Island Hospital and Miriam Hospital that presented with colon cancer and synchronous liver metastases between 1995 and 2012 were selected (Tables I and II). Length of survival was determined by retrospectively reviewing the Lifespan Health System electronic medical record. Last date of contact and patient status (mortality) were determined via tumor registry review of clinical records. Slides were reviewed by an attending pathologist (MBR and CE) for presence or absence of tumor. Tumor tissue with greater than 70% tumor cells in both primary tumor site and associated metastatic lesion was selected for analysis.
Laser capture microdissection and RNA extraction. Tissue sections of 10 μm were cut and mounted onto plain glass slides for manual microdissection or PEN membrane slides (Life Technologies, Carlsbad, CA, USA) for UV laser capture microdissection. The sections were de-paraffinized, stained, dehydrated through graded alcohols using the Paradise FFPE reagent System (Life Technologies) and subjected to manual microdissection or UV laser capture microdissection within two hours of deparaffinization. Care was taken to capture areas with similar proportions of cancer cells. Laser-captured cells were collected on LCM Macro CapSure caps (Life Technologies) using the Arcturus XT LCM instrument (Life Technologies). Total RNA (containing both mRNA and microRNA) was extracted using the RecoverAll Total Nucleic Acid Extraction Kit for formalin-fixed paraffin-embedded tissues, with DNAse incubation (Life Technologies).
MicroRNA expression analysis. RNA was prepared following the manufacturer's instructions. Briefly, 5 ng of RNA was reverse-transcribed using the Taqman MicroRNA Reverse Transcription Kit and the Megaplex RT primer Human Pool A (Life Technologies). The reverse-transcribed cDNA was then pre-amplified for 12 cycles using Taqman PreAmp Master Mix and the Megaplex PreAmp primers, Human Pool A. 377 microRNAs were assayed on a Taqman Human microRNA Array card A using an ABI HT7900 (Life Technologies). MicroRNAs were considered to be significantly expressed if more than 10 cases had Ct values less than 32. ΔΔCt was calculated relative to U6 snRNA. Samples were prepared in two batches and batch correction was applied to the U6 corrected ΔCt values in each sample using Partek® software version 6.6 (Partek Inc., St. Louis, MO, USA). Significance was determined by fold changes greater than 2 and p-values <0.05 by a paired t-test followed by multiple hypothesis correction (9). Statistical analysis was performed in R (www.r-project.org/). Hierarchical clustering was performed in Gene-E (Broad Institute, Cambridge, MA, USA) using Pearson correlation as the distance metric.
Survival analysis. For genes expressed at higher levels in metastases compared to primary tumors, one point was assigned when the expression was higher than the median. For microRNAs expressed at a lower level in metastases compared to primary tumors, one point was assigned when the expression was lower than the median. The sum of these points constitutes the score. When the score is less than half the number of microRNAs, the case is assigned to the low risk group. When the score is greater than half the number of microRNAs, the case is assigned to the high risk group. Kaplan–Meier analysis in R was used to test if the classification was associated with survival.
Summary of patient demographics and tumor characteristics.
Results
A total of 19 out of 27 CRC cases from the Lifespan Hospital System archives were selected for microRNA expression analysis on low-density qPCR arrays (II). Portions of formalin-fixed paraffin-embedded tissue sections were enriched for cancer cells by laser capture microdissection. Isolated RNA was reverse-transcribed and amplified to interrogate 377 microRNAs on Taqman low-density array cards.
Sixteen microRNAs were found to have significant differential expression between the primary and metastatic tumors at p<0.05 using a paired t-test, with average fold changes greater than 2 (Figure 1). Four microRNAs were found to have FDR-corrected p-values of less than 0.05 (Table III). Five microRNAs were down-regulated, and 11 microRNAs up-regulated in liver metastases. The four most differentially expressed microRNAs, excluding miR-122, discriminate primary from metastatic tumors in 18 out of the 19 cases screened with qPCR array cards (Figure 1B).
Patient information for each case.
To extend the study, we tested expression of five microRNAs with uncorrected p<0.05, q<0.15 and fold change more than 2 using assays targeting one microRNA, plus U6 snRNA, in an additional eight samples (Figure 2). miR-523 and miR-548a-3p were not detected in the extended set with individual primers and were removed from further analysis.
The most significantly differentially expressed microRNA was found to be the liver-specific microRNA, miR-122. miR-122 constitutes ~70% of all microRNA content in liver (10). Higher miR-122 expression in metastases indicates that the Taqman qPCR approach detects the small amounts of residual liver cells remaining after enrichment. A related study also observed high expression of miR-122 in liver metastases, even from a mixture of 95% tumor and 5% liver tissue (7). The large fold changes reflect the absence of miR-122 expression in primary tumors compared to its significant expression in metastases.
The second ranked microRNA, miR-885-5p, was also very significantly differently expressed in primary and metastatic tumors, with >1,000-fold increased expression in liver metastases compared to primary tumors (Figure 1). Up-regulation of miR-885-5p may be an indicator of liver pathology (2, 11).
We tested the association between microRNA expression and overall survival using a scoring approach similar to that developed by Kang et al. and ourselves (3, 4, 12, 13) (Figure 2). We used the Cox proportional hazards model to determine if higher or lower expression is associated with higher risk. One point was assigned if the microRNA expression was higher than the mean if the hazard ratio is greater than one, and vice versa for hazard ratios less than one. Out of the 16 most differentially expressed microRNAs, only miR-133b and miR-210 fit the Cox model with p<0.2. Patients with tumors with elevated expression levels of both miR-210 and miR-133b in metastatic disease had a significantly shorter overall survival by Kaplan-Meier analysis (Figure 2A). Interestingly, higher miR-210 expression was observed in most metastases, but the majority of metastases had lower miR-133b expression (Figure 3). Combining the 19 cases measured on array cards with the measurement of an additional eight tumor pairs measured by individual qPCR assays targeting only the microRNAs of interest and U6 snRNA for normalization improved separation of the high- and low-risk groups in a total of 27 patients (Figure 2). These observations suggest that when comparing the relative expression between primary and metastatic tumors, a more reliable measurement is obtained that can be independently measured and may be less dependent on the variation of expression among patients, or in another cohort of patients.
Identification of microRNAs differentially expressed between primary and metastatic tumors. A: Bar graph of average fold changes of the 16 most significantly differentially expressed microRNAs. B: Hierarchical clustering of the four most differentially expressed microRNAs, except for miR-122, discriminates primary from metastatic tumors.
Higher expression of miR-210 and miR-133b in metastases compared to primary tumors identifies high-risk patients. A: Kaplan–Meier analysis of 19 cases from the qPCR arrays shows association of the change of expression between primary and metastatic tumors for miR-210 and miR-133. Adding eight cases measured by qPCR specifically targeting miR-210 and miR-133 for a total of 27 cases improved association with survival, as indicated by the clearer separation of the high- and low-risk groups comparing the set of 19 cases. Red dotted lines indicate the high-risk group and the green line indicates the low-risk group. B: The change of expression of miR-210 and miR-133b is directly correlated with overall survival.
miR-133b is often found to be a tumor suppressor (4, 14) and was expressed at low levels in most liver metastases (Figure 1). However, the high-risk patients exhibited higher miR-133b expression compared to the low-risk patients, whom often had no detectable miR-133b expression in the primary tumor (data not shown). It is possible that metastases with miR-133b expression are more lethal because these more aggressive tumors compensate for higher miR-133b expression through other mechanisms, such as adaptation to hypoxia through miR-210. The inherent expression of miR-210 and miR-133b in either primary or metastatic tumors is not associated with survival (Figure 4).
Validation of microRNA expression by Taqman qPCR of individual microRNAs. The difference in expression between primary tumors and liver metastases is plotted for the 19 cases from the qPCR arrays and the eight cases tested with individual assays.
Within this dataset, microRNAs can be identified whose inherent expression in either the primary tumor or liver metastasis is associated with survival (Figure 4). However, the set of microRNAs in the primary tumor and that in the liver metastases associated with survival are completely different (Figure 4). These observations suggest that microRNA expression in each tumor type is significantly different from the other, likely reflecting adaptive changes to the local microenvironment for each tumor. These differences may be critical for treatment decisions as the metastases are the form of the disease that needs to be targeted by therapy.
Discussion
The goal of the present study was to investigate the expression of microRNAs of hepatic metastatic colon adenocarcinoma relative to that of the primary tumor and test for correlation with survival. We hypothesize that metastases are a more aggressive stage of the disease and that molecular changes in metastases reflect the type of advanced disease that might better predict patient outcomes. Our work at identifying microRNAs and mRNAs in ovarian cancer supports this hypothesis (4, 5, 10, 13). To test this hypothesis in CRC, we measured microRNA expression and evaluated its association with overall survival. As far as we are aware, our study is the first to show that increased miR-210 in association with increased miR-133b expression in liver metastases compared to primary tumors correlates with overall survival.
Recent findings from another study identify microRNAs differentially expressed between CRC primary tumors and liver metastases using microarrays (7). They also identified miR-10b and miR-210, but not miR-133b, as being differentially expressed between primary tumors and liver metastases. They might not have detected miR-133b because they used microrarrays, which are not as sensitive as qPCR assays. However, the overall list of microRNAs is significantly different (Table IV) This could be because of the modest size of the two studies or because of the different technologies. A recent report suggests that even from the same RNA samples, different technologies result in significantly different observations (15), suggesting that extensive validation from multiple assays is necessary to test microRNA expression. Our study also differs in methodology, as we performed microdissection to enrich for cancer cells, and to reduce liver cell contamination. Both studies identified high miR-122 expression, as this microRNA is extremely highly expressed in liver cells such that even if just 5% of the cell population are liver cells, high expression levels are observed. Because no significant expression of miR-122 was observed in the primary tumors, any expression in the metastases leads to large expression differences.
A distinct set of microRNAs expressed in primary tumors and liver metastases is associated with survival. All expressed microRNAs were tested for association with overall survival. Red dotted lines indicate the high-risk group and the green line indicates the low-risk group. Higher expression of the listed microRNAs in primary tumor (A) or liver metastases (B) is associated with poor survival, except for the lower expression of miR-9 in metastases, which is associated with high risk in the Kaplan–Meier plots.
miR-133b is reported to be a tumor suppressor in some contexts (16-18). Low expression of miR-133b in primary CRC tumors has been associated with poor survival (6) and is consistent with our findings that increased expression in liver metastases indicates higher risk. This may suggest a different function for miR-133b in the liver environment. Because miR-133b is among the more abundant microRNAs in the liver (19), we tested if miR-133b is correlated with miR-122 expression but found no correlation (data not shown, see Figure 3), suggesting that miR-133b indeed originates from cancer cells. In most of the liver metastases examined, miR-133b had lower expression than that of the primary tumor (Figure 3).
miR-210 is induced by hypoxia-inducible factor 1 and is up-regulated in most solid tumors (20, 21). miR-210 increases the metastatic potential of hepatocellular carcinoma via down-regulation of vacuole membrane protein 1 (22). The increase in miR-210 expression in liver metastases suggests that there may be more hypoxia in the metastases, or during the establishment of liver tumors. The presence of miR-210 is consistent with the presence of more drug-resistant cancer cells, which could populate metastases.
List of measured microRNAs with significant expression. RQ is the average RQ (Relative Quantitation to U6 snRNA).
Other microRNAs such as miR-21 have been reported to be associated with survival (23). However, many of these cases present before extensive metastasis has occurred. Our study examined metastases and thus all the cases had advanced disease, which confers inherently shorter survival than a cohort that includes cases with earlier phases of disease. Thus, the prognostic combination of miR-210 and miR-133b identifies the high-risk patients that do not respond well to current treatment.
Comparison between data of Pizzini et al. (7) and our list of identified microRNAs. microRNAs in bold are common between the two lists.
Measuring the difference in RNA expression between tumor types may be a powerful approach to predict patient outcomes as it indicates how the cancer cells adapt during disease progression in each patient. Insight into the evolving nature of tumors may provide more precise predictions and insights into patient outcomes as the phenotypes of primary tumors change significantly during metastasis (4, 13, 24). Comparing primary and metastatic tumors provides insight into how the disease is progressing in the patient, which can guide treatment decisions. These observations support the hypothesis that improved prognosis from molecular markers may benefit from the inherent normalization of evaluating the differences between primary and metastatic tumors compared to expression measurements in just one tumor per patient.
In conclusion, this is the largest qPCR study examining microRNAs in pairs of matched primary and metastatic CRC tumors. Our study supports the concept that the change in expression between primary tumors and liver metastases could be subject to less inherent variation among patients for prognostic predictions and be more indicative of the disease progression in each patient. A possible complication is the potential heterogeneity among different metastases in each patient, along with intra-tumoral heterogeneity, which affects all types of biomarker assays. Further research is warranted to investigate the heterogeneity of liver metastases. We nominate the candidate microRNAs, miR-210 and miR-133b, as potential prognostic markers. We emphasize that these candidates will require replication in independent CRC datasets as well as functional testing. Validated microRNA markers could provide new insights into disease progression and provide indicators for each patient of how their disease is progressing and how the metastases would most likely respond to treatment.
Acknowledgements
The Molecular Pathology Core of the COBRE Center for Cancer Research Development is funded by the National Institute of General Medical Sciences of the National Institutes of Health under Award Number P20GM103421.
Footnotes
-
This article is freely accessible online.
-
Competing Interests
The Authors have declared that no competing interests exist in regard to this study.
- Received April 11, 2014.
- Revision received June 6, 2014.
- Accepted June 9, 2014.
- Copyright© 2014 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved
References
In this issue
Jump to section
Related Articles
Cited By...
- Towards a consensus microRNA signature of primary and metastatic colorectal cancer
- Tissue micro-RNAs associated with colorectal cancer prognosis: a systematic review
- MicroRNAs associated with small bowel neuroendocrine tumours and their metastases
- Effect of Xenotransplantation Site on MicroRNA Expression of Human Colon Cancer Stem Cells