The Prognostic Value of microRNAs Varies with Patient Race/Ethnicity and Stage of Colorectal Cancer

Abstract

Purpose

MicroRNAs (miRNAs) have potential prognostic value for colorectal cancers (CRCs); however, their value based on patient race/ethnicity and pathologic stage has not been determined. The goal was to ascertain the prognostic value of 5 miRNAs with increased expression in CRCs of African American (Black) and non-Hispanic Caucasian (White) patients.

Experimental Design

TaqMan^® qRT-PCR was used to quantify expression of miR-20a, miR-21, miR-106a, miR-181b, and miR-203 in paired normal and tumor CRC archival tissues collected from 106 Black and 239 White patients. The results were correlated with overall survival based on patient race/ethnicity and pathologic stage. Since decisions regarding adjuvant therapy are important for Stage III CRCs, and since miR-181b appeared to have prognostic value only for Stage III Black patients, we assessed its prognostic value in a separate cohort of Stage III CRCs of Blacks.

Results

All 5 miRNAs had higher expression in CRCs (>1.0-fold) than in corresponding normal tissues. High expression of miR-203 was associated with poor survival of Whites with Stage IV CRCs (HR=3.00, 95% CI=1.29–7.53), but in Blacks it was an indicator of poor survival of patients with Stage I and II CRCs (HR=5.63, 95% CI=1.03–30.64). Increased miR-21 expression correlated with poor prognosis for White Stage IV patients (HR=2.50, 95% CI=1.07–5.83). In both test and validation cohorts, high miR-181b expression correlated with poor survival of only Black patients with Stage III CRCs (HR=1.94, 95% CI=1.03–3.67).

Conclusion

These preliminary findings suggest that the prognostic value of miRNAs in CRCs varies with patient race/ethnicity and stage of disease.

Introduction

In the US, colorectal cancer (CRC) is the third most common malignancy and the second leading cause of cancer-related deaths among men and women (1). CRC is a heterogeneous disease, especially with respect to patient race/ethnicity, tumor stage, and genetic alterations that contribute to its progression. Identification of molecular determinants involved in the progression of CRCs will aid in evaluating patient prognosis and/or provide targets for cancer prevention and/or therapy. MicroRNAs (miRNAs) are implicated in the progression of and prognosis for CRCs (2–5).

MiRNAs are a family of small (20 to 24-nucleotide) single-stranded, non-coding RNAs that regulate gene expression at the post-transcriptional level (6). They are thought regulate protein production by binding to a complementary site in mRNA, preventing it from being translated or targeting it for destruction, or by transcriptional gene-silencing at the chromatin level (7–9). The number of known miRNAs in humans now exceeds 1,500 (10). Each miRNA may regulate hundreds of mRNAs (11).

About half of the human miRNAs are located at cancer-associated regions of the genome, suggesting that they are involved in tumorigenesis (12). Based on cancer-associated alterations in their expression, miRNAs may act as tumor suppressors or oncogenes (13) and are implicated in tumor progression and metastasis.(14) Many are dysregulated in human tumors (9). Microarray analyses have revealed specific down-regulation of let-7 expression in lung tumors, but not in breast or colon cancers (15), suggesting that the effects of miRNAs are organ-specific.

Due to deletion of the p53 tumor suppressor gene in colon cancer cell lines, several miRNAs are characterized as abnormal. Among these, miR-15b, miR-181b, miR-191, and miR-200c are overexpressed in CRCs compared to normal colorectal tissues, and survival analyses indicate that patients with higher miR-200c expression have shorter survival times compared to patients with lower expression (2). Accumulation of miR-143 and miR-145 is down-regulated in cells derived from CRCs (16). MiR-21 is expressed at high levels in colonic carcinoma cells, and high miR-21 expression is associated with poor survival and decreased therapeutic response (3). In CRCs, miR-133b and miR-145 are down-regulated, and miR-31, miR-96, miR-135b, and miR-183 are up-regulated (17).

Although previous studies have highlighted the potential value of miRNA expression profiles in the prognosis for CRC patients (2–5), their value with respect to race/ethnicity and tumor stage is not known. In the present investigation, we determined the expression levels of five miRNAs (miR-20a, miR-21, miR-106a, miR-181b, and miR-203), which were the most highly overexpressed miRNAs amongst a panel of 389 miRNAs assayed in CRCs (3). Furthermore, their prognostic value was analyzed in a consecutive, retrospective CRC patient cohort of 142 African Americans (Blacks) and 239 non-Hispanic Caucasians (Whites) with respect to race/ethnicity and tumor stage.

Materials and Methods

Patients and Tissue Sample Collection

Included were 142 Black and 239 White CRC patients who had undergone surgical resection for “first primary” sporadic CRC between 1985 and 2004 at the University of Alabama at Birmingham (UAB). Thus, this was an ‘unselected’ patient population. Patients with multiple primaries within the colorectum, with multiple malignancies, or with a family or personal history of cancer (due to distinctive molecular pathways) were excluded. We excluded from the study population those patients who died within a week of their surgery; those with surgical margin-involvement, unspecified tumor location, multiple primaries within the colorectum, or multiple malignancies; and those patients with a family history of hereditary non-polyposis colorectal cancer (HNPCC), familial adenomatous polyposis (FAP), or personal histories of CRC. Since the information from the patient charts may not be reliable in identifying the familial vs. sporadic nature of CRCs, our cohort can be described as a ‘consecutive’ patient population. The intent of using patients from this time period was to maximize post-surgery follow-up. Formalin-fixed, paraffin-embedded (FFPE) tissue blocks of CRCs and their corresponding normal (benign colonic epithelial) surgical specimens were collected from UAB. Of 142 Blacks, 106 were included as an initial test cohort and 36 as a separate validation cohort. The sample size and power were estimated based on a previous study (3). Demographic, clinical, and pathologic information was collected from medical records, physician charts, and pathology and radiology reports. These data included age, race, gender, tumor location, tumor-node-metastasis (TNM) stage, tumor size, survival times, status, and relapse. Hematoxylin and eosin-stained slides were reviewed by two pathologists (CKS & WEG) together. Pathologic staging was performed according to the criteria of the American Joint Commission on Cancer (18). Histological grade was assessed according to World Health Organization criteria. Well and moderately differentiated tumors were pooled into a low-grade group and poor and undifferentiated tumors into a high-grade group (19). The Institutional Review Board of UAB approved this study.

Patient race/ethnicity and Follow-up Information

Information on patient race/ethnicity was obtained from their charts, and assignment was self-described or self-identified. We recognize, however, that there is some diversity in identification within any race/ethnic group.

Follow-up information was retrieved from the UAB Tumor Registry. Patients were followed either by the patient’s physician or by the UAB Tumor Registry until their death or the date of the last documented contact within the study time frame. The Tumor Registry ascertained outcome (mortality) information directly from patients (or living relatives) and from the physicians of the patients through telephone and mail contacts. This information was further validated against the state Death Registry. Demographic data, including patient age at diagnosis, gender, race/ethnicity, date of surgery, date of the last follow-up (if alive), date of recurrence (if any), and date of death, were collected. The Tumor Registry updated follow-up information every six months, and follow-up of our retrospective cohorts ended in September 2012. The median follow-up periods for Blacks and Whites were 19 years (range 7–29 years) and 15 years (range 7–30 years), respectively.

RNA Isolation and qRT-PCR

Total RNA was extracted from macro-dissected tumor and corresponding normal FFPE samples using TRIzol® Reagent (Invitrogen, CA). The quality of RNA was determined with a NanoDrop2000 spectrophotometer (Thermo Fisher Scientific, NE). qRT-PCR of miRNAs was performed using TaqMan® microRNA assays (Applied Biosystems, CA) and was conducted by two-step RT-PCR according to the manufacturer’s protocol. Briefly, cDNA synthesis was first accomplished using the TaqMan® MicroRNA Reverse Transcription Kit (Applied Biosystems, CA). The template consisted of 10 ng of total RNA, and miRNA-specific primers were used (provided in TaqMan® MicroRNA kits). This PCR reaction was in a 15-µL reaction mixture and was performed on the iQ™5 Real-Time PCR Detection System (BioRad, CA) for 30 min at 16°C, 30 min at 42°C, and 5 min at 85°C. The synthesized target cDNA was amplified using sequence-specific primers from the TaqMan® kits. This PCR reaction was accomplished in a 20-µL mixture (1.33 µL of template cDNA) and was performed for 10 min at 95°C, 15 sec at 95°C, and 1 min at 60°C for 40 cycles. Signals were collected at the end of each cycle.

Relative expression values were calculated using the comparative C_T method and by normalizing with miRNA RNU6B as an endogenous control. Fold change = 2^−ΔΔC_T, where ΔΔC_T = ΔC_T (C_TmiRNA − C_{T RNU6B) tumor} − ΔC_T (C_TmiRNA − C_{T RNU6B})_normal. The experiments were performed in triplicate, and the investigators were blinded to clinical data and survival outcome information until completion of all assays.

Statistical Analyses and Validation

Sample Size and Power Calculations

The sample size and power analysis were estimated based on a prior study of miRNAs in CRCs (3). In that analysis, increased expression of miR-21 was a poor prognostic indicator of survival, and 26 of 72 patients (36%) were positive for its overexpression. The hazard ratio was 2.5 (p=0.01). Thus, we proposed to evaluate 96 samples. A minimum of 96 samples in each racial group will provide enough power to detect a hazard ratio ≥ 2.2. Therefore, the sample size (n=96) was sufficient to identify a statistically significant prognostic value for miRNA expression.

Statistical Analysis

Chi-square tests were used to assess the univariate associations of baseline characteristics with miRNA expression for Black and White patients, separately. The baseline characteristics included demographic variables (age and gender) and pathologic variables (tumor location, size, grade, nodal status, distant metastasis, and stage). Cluster analyses, performed to define miRNA expression cutoff points, were based on Euclidean distances, which are geometric distances in multidimensional space. This method was chosen to establish cut-off points based on numerical technique. The cutoff of each miRNA was calculated for Blacks, Whites, and combined patients separately, but remained constant within each race throughout the study. Descriptive statistics were used to describe the basic features of the fold change of each miRNA (Table 2). The type-I error rate of each test was controlled at <0.05. All analyses were performed (by BZ & SB) with SAS application version 9.2 (SAS Institute Inc., Cary, NC).

Table 2.

Expression of miRNAs in CRCs

miRNA	Combined (N=345) FCa,b (Q1-Q3)c	Blacks (N=106) FCa,b (Q1-Q3)c	Whites (N=239) FCa,b (Q1-Q3)c
miR-20a	2.05 (0.72–7.31)	4.14 (1.51–27.3)	1.60 (0.61–4.78)
miR-21	2.10 (0.92–5.38)	3.79 (0.85–11.2)	2.03 (1.00–4.10)
miR-106a	2.00 (0.61–8.14)	4.33 (1.06–47.5)	1.58 (0.58–4.58)
miR-181b	1.33 (0.59–3.97)	3.97 (0.68–22.8)	1.06 (0.58–2.34)
miR-203	2.10 (0.40–9.06)	5.98 (0.69–71.5)	1.49 (0.38–5.70)

N, total number of cases; FC, fold change.

^aThreshold cycle (C_T) is the unit of measurement in qRT-PCR to measure relative gene expression. Difference in C_T (ΔΔ C_T) = Tumor change in Ct minus paired non-tumor change in C_T.

^bFold change calculated from ΔΔCt, where fold change = 2^{-(median ΔΔ CT)}. Median fold change values are represented.

^cQ1 and Q3 for inter-quartile range of fold change values.

Deaths due to CRC were the outcomes (events) of interest. Survival analysis was used to model the relationship between the time to death due to CRC and miRNA expression. Those patients who died of any cause other than CRC and those who were alive at the end of the study were considered to be censored. A log-rank test and Kaplan-Meier survival curves were used to compare patient survival in the high and low expression groups for each miRNA. Multivariate Cox regression survival models were built by including all five miRNAs and the known confounding covariates (age, sex, tumor location, tumor stage, tumor size, and tumor grade). To detect the relationship for patients with different tumor stages, survival analyses were also performed for patient groups with each tumor Stage (I+II, III, and IV) separately. The final Cox model was obtained based on stepwise selection criteria. All the above analyses were conducted for Black, White, and combined patient populations separately. The final Cox model was obtained based on stepwise selection criteria. All the above analyses were conducted for Black, White, and combined patient populations separately, and Benjamini-Hochberg corrections were used to adjust for multiple comparisons testing errors (20). This method is an accepted approach for adjusting for multiple comparisons, and similar corrections have been used in other studies of cancer biomarkers (21, 22).

Results

MicroRNA Expression Profiles of CRCs

The characteristics of Black (test and validation) and White patient cohorts are given in Table 1. Expression levels of five miRNAs in the combined population, Blacks, and Whites are given separately in Table 2. We recently reported that all five miRNAs are stable in FFPE tissues stored over long periods of time (23). For the combined population, all miRNAs had higher expression in CRCs compared to normal tissues (>1.0-fold; range, 1.33 to 2.10). All five had higher expression in CRCs from Blacks (range, 3.97 to 5.98) as compared to CRCs from Whites (1.06 to 2.03) (Table 2).

Table 1.

Clinicopathological and molecular features of CRC patients in study

Variable	Blacks		Whites
	Test cohort	Stage III Validation cohort
	N=106 (%)	N=36 (%)	N=239 (%)
Age group (years)
< 65	52 (49)	13 (36)	117 (49)
≥ 65	54 (51)	23 (64)	122 (51)
Gender
Female	71 (67)	18 (50)	104 (44)
Male	35 (33)	18 (50)	135 (56)
Tumor location
Proximal colon	28 (26)	19 (53)	114 (48)
Distal colon	69 (65)	12 (33)	69 (29)
Rectum	9 (9)	5 (14)	56 (23)
Depth of tumor invasion
pT0	0 (0)	0 (0)	0 (0)
pT1	4 (4)	0 (0)	8 (3)
pT2	11(10)	2 (6)	29 (12)
pT3	73 (69)	27 (75)	165 (69)
pT4	18 (17)	7 (19)	37 (15)
aNodal status
N0	57 (54)	0 (0)	132 (55)
N1–3	49 (46)	36 (100)	97 (41)
bDistant metastasis
M0	89 (85)	36 (100)	199 (83)
M1	16 (15)	0 (0)	40 (17)
Tumor stage
I	9 (9)		29 (12)
II	46 (43)		100 (42)
III	35 (33)	36 (100)	71 (30)
IV	16 (15)		39 (16)
Tumor grade
Low	86 (81)	31 (86)	185 (78)
High	20 (19)	5 (14)	53 (22)
cTumor size (cm)
≤ 5	66 (62)	18 (50)	152 (64)
> 5	40 (38)	16 (44)	87 (36)

N, total number of cases; %, percentage of N;

^aInformation missing for 10 White cases.

^bInformation missing for 1 Black case (test).

^cInformation missing for 2 Black cases (validation).

Survival Analyses

Data on both univariate Kaplan-Meier and multivariate Cox regression survival analyses of the combined patient population as well as patient groups categorized according to race/ethnicity, tumor stage, and miRNA status are provided in Table 3 and Table 4. Univariate survival curve figures are provided only for test and validation data on miR-181b to demonstrate its independent prognostic value in Blacks with Stage III CRCs.

Table 3.

Univariate survival analyses based on miRNA status

Race	miRNA	a,bLog rank p-value	Total (N)	Low (N)	High (N)
Combined
All Stages	miR-20a	0.020	344	233	111
	miR-21	0.223	344	232	112
	miR-106a	0.005	344	231	113
	miR-181b	0.004	344	234	110
	miR-203	0.002	344	233	111
Stage I+II	miR-20a	0.236	184	136	48
	miR-21	0.396	184	125	59
	miR-106a	0.236	184	136	48
	miR-181b	0.303	184	138	46
	miR-203	0.360	184	136	48
Stage III	miR-20a	0.877	105	68	37
	miR-21	0.567	105	69	36
	miR-106a	0.158	105	63	42
	miR-181b	0.125	105	64	41
	miR-203	0.045	105	63	42
Stage IV	miR-20a	0.965	55	29	26
	miR-21	0.155	55	38	17
	miR-106a	0.585	55	32	23
	miR-181b	0.722	55	32	23
	miR-203	0.920	55	34	21
Blacks
All Stages	miR-20a	0.065	106	72	34
	miR-21	0.149	106	72	34
	miR-106a	0.049	106	72	34
	miR-181b	0.003	106	72	34
	miR-203	0.005	106	71	35
Stage I+II	miR-20a	0.368	55	36	19
	miR-21	0.287	55	38	17
	miR-106a	0.471	55	38	17
	miR-181b	0.189	55	38	17
	miR-203	0.021	55	39	16
Stage III	miR-20a	0.017	35	26	9
	miR-21	0.331	35	24	11
	miR-106a	0.039	35	25	10
	miR-181b	0.008	35	24	11
	miR-203	0.071	35	22	13
Stage IV	miR-20a	0.879	16	10	6
	miR-21	0.891	16	10	6
	miR-106a	0.766	16	9	7
	miR-181b	0.891	16	10	6
	miR-203	0.879	16	10	6
Whites
	miR-20a	0.978	238	161	77
	miR-21	0.680	238	161	77
	miR-106a	0.415	238	160	78
	miR-181b	0.123	238	162	76
	miR-203	0.696	238	163	75
Stage I+II	miR-20a	0.873	129	94	35
	miR-21	0.399	129	93	36
	miR-106a	0.486	129	96	33
	miR-181b	0.371	129	99	30
	miR-203	0.408	129	97	32
Stage III	miR-20a	0.513	70	42	28
	miR-21	0.089	70	42	28
	miR-106a	0.919	70	38	32
	miR-181b	0.509	70	42	28
	miR-203	0.811	70	39	31
Stage IV	miR-20a	0.482	39	25	14
	miR-21	0.023	39	26	13
	miR-106a	0.184	39	26	13
	miR-181b	0.752	39	21	18
	miR-203	0.029	39	27	12

Combined population includes all stages and all races/ethnic groups;

^aLog-rank, p-value estimated from Kaplan-Meier univariate analysis

^bAdjustment for multiple comparisons has been made for the P values using the Benjamini-Hochberg method.

Table 4.

Cox multivariate regression analysis to evaluate the independent prognostic value of miRNAs in CRCs

Combined				Blacks				Whites
Prognostic variables	Indicator of poor prognosis	aHazard ratio (95% CI)	cP-value	Prognostic variables	Indicator of poor prognosis	aHazard ratio (95% CI)	cP-value	Prognostic variables	Indicator of poor prognosis	aHazard ratio (95% CI)	cP-value
All stages (N=344)				All stages (N=106)				All stages (N=238)
miRNA				miRNA				Tumor stage
miR-203	High	1.49 (1.03–2.17)	0.036	miR-181b	High	3.55 (1.50–8.41)	0.004	II vs. I	II	3.55 (1.08–11.63)	0.038
Tumor stage				Tumor stage				III vs. I	III	6.22 (1.90–20.37)	0.002
II vs. I	II	1.60 (0.75–3.40)	0.224	II vs. I	II	0.80 (0.27–2.40)	0.701	IV vs. I	IV	26.43 (7.91–88.32)	<.0001
III vs. I	III	2.97 (1.41–6.28)	0.004	III vs. I	III	2.38 (0.83–6.79)	0.107
IV vs. I	IV	10.11 (4.66–21.93)	<.0001	IV vs. I	IV	4.19 (1.31–13.41)	0.017	Tumor grade
Tumor grade				Tumor location				High vs. low	High	1.54 (0.99–2.38)	0.057
High vs. low	High	1.60 (1.13–2.28)	0.009	Proximal colon vs. Rectum	Proximal colon	1.60 (1.13–2.28)	0.061
				Distal colon vs. Rectum	Distal colon	1.43 (0.33–6.21)	0.644
				Tumor grade
				High vs. low	High	2.21 (1.14–4.29)	0.019
Stage I+II (N=184)				Stage I+II (N=55)				Stage I+II (N=129)
Tumor grade				miRNA				Tumor grade
High vs. low	High	1.68 (0.86–3.28)	0.132	miR-203	High	5.63 (1.03–30.64)	0.046	High vs. low	High	1.52 (0.60–3.83)	0.382
				Tumor grade
				High vs. low	High	3.67 (1.30–10.37)	0.015
Stage III (N=105)				Stage III (N=35)				Stage III (N=70)
miRNA				miRNA				Tumor grade
miR-203	High	1.90 (1.05–3.42)	0.034	miR-181b	High	7.94 (1.60–39.30)	0.012	High vs. low	High	1.40 (0.63–3.12)	0.416
Tumor grade				Tumor grade
High vs. low	High	1.27 (0.66–2.42)	0.480	High vs. low	High	0.28 (0.03–2.42)	0.249
Stage IV (N=55)				bStage IV (N=16)				Stage IV (N=39)
miRNA								miRNA
miR-21	High	3.25 (1.37–7.72)	0.008					miR-21	High	2.50 (1.07–5.83)	0.034
miR-203	High	2.19 (1.10–4.33)	0.026					miR-203	High	3.01 (1.30–6.98)	0.011
Tumor grade								Tumor grade
High vs. low	High	3.76 (1.85–7.63)	0.0003					High vs. low	High	3.12 (1.29–7.53)	0.012

^aAdjusted for miRNA expression levels, age, gender, TNM tumor stage, tumor location within the colorectum, tumor grade, tumor size, and race (for analyses of the Combined patient population); CI, confidence interval.

^bDue to the small number of available cases, multivariate analyses was not completed for Stage IV Black CRC patients.

^cAdjustment for multiple comparisons has been made for the P values using the Benjamini-Hochberg method.

Prognostic significance of miR-20a and miR-106a

Univariate survival analyses demonstrated that high expression of miR-20a was associated with shorter overall survival (OS) in the combined patient population when all stages were considered (Log-rank, p=0.02) and especially for Stage III Black CRC patients (log-rank, p=0.017) (Table 3). Similarly, high expression of miR-106a was associated with shorter OS for the combined population; for Blacks when all stages were considered; and for Blacks with Stage III CRCs (Log-rank, p=0.005, p=0.049, and p=0.039, respectively) (Table 3). In our multivariate analyses, however, these two miRNAs were not established as independent markers (Table 4).

Prognostic significance of miR-21

Increased expression of miR-21 correlated with shorter OS of only Stage IV patients (HR, 3.25; 95% CI, 1.37–7.72; p=0.008) (Table 4), especially for White patients with Stage IV CRCs (Log-rank, p=0.023, Table 3; and HR, 2.50; 95% CI, 1.07–5.83; p=0.034, Table 4).

Prognostic significance of miR-181b

Although high expression of miR-181b was associated with shorter OS of the combined patient population (log-rank, p=0.004, Table 3), it was an independent prognostic marker only for Black patients (Log-rank, p=0.003, Table 3; HR, 3.55; 95% CI, 1.50–8.41, Table 4), especially for those with Stage III CRCs (Test-set, Log-rank, p=0.008, Table 3 and Figure 1a; HR, 7.94; 95% CI, 1.60–39.30, Table 4).

Figure 1.

Kaplan-Meier univariate survival analysis based on miR-181b expression levels in the test (1A) and validation (1B) cohorts of Black Stage III CRC patients.

Since miR-181b appeared to have a prognostic value for Stage III, especially for Black patients, both in univariate and multivariate analyses, and studies demonstrating that markers are useful in determining prognosis of minority populations are rare, we validated the prognostic value of miR-181b in a separate cohort of 36 Stage III CRCs of Blacks. Moreover, traditional pathological features coupled with novel molecular markers could identify aggressive phenotypes within Stage III tumors and aid in identifying high risk patients to maximize the benefits of adjuvant therapy. The demographic and tumor characteristics for this validation cohort are given in Table 1. Due to non-availability of follow-up information, one case in the validation cohort was excluded from survival analyses.

Similar to the findings of the test cohort of Stage III Black patients, univariate analysis demonstrated that high expression of miR-181b was associated with short patient survival in the validation cohort (Log-rank, p=0.005) (Figure 1b). Since the findings of multivariate Cox regression analysis of the validation cohort were similar to findings for the test cohort (HR, 2.75; 95% CI, 1.17–6.48) (data not shown), and to increase our sample size in order to increase statistical power, the test and validation cohorts were pooled for multivariate analysis. Multivariate analyses demonstrated that Blacks with Stage III CRCs with high expression of miR-181b were 1.94 times more likely to die of CRC than patients in the combined population who had low miR-181b expression (HR, 1.94; 95% CI, 1.03–3.67) (Table 5).

Table 5.

Cox multivariate regression analysis to evaluate the independent prognostic value of miR-181b in the combined test and validation cohorts of Black Stage III CRCs (N=70)

Prognostic variables	Indicator of poor prognosis	aHazard ratio 95% (CI)	bP-value
miRNA
miR-181b	High expression	1.94 (1.03–3.67)	0.043
Tumor grade
High vs. low	High	1.89 (0.91–3.91)	0.092

^aAdjusted for miRNA expression levels, age, gender, TNM stage, tumor location within the colorectum, tumor grade, and tumor size; CI, confidence interval.

^bAdjustment for multiple comparisons has been made for the P values using the Benjamini-Hochberg method.

Prognostic significance of miR-203

High expression of miR-203 in CRCs correlated with short survival in different groups of patients (Table 3 and Table 4). Increased miR-203 expression had an independent prognostic value for all CRC patients (combined population, Log-rank, p=0.002; HR, 1.90; 95% CI, 1.05–3.42), but especially for Blacks with early stage (I+II) CRCs (Log-rank, p=0.021; HR, 5.63; 95% CI, 1.03–30.64) and for Whites with Stage IV disease (Log-rank, p=0.029; HR, 3.01; 95% CI, 1.30–6.98).

Discussion

A panel of miRNAs that are over-expressed in CRCs (3) was analyzed to determine their prognostic value for Black and White CRC patients. The results show an increase in the expression of all the miRNAs in CRCs, which is consistent with previous reports (2, 3, 24, 25). Although the prognostic value of these miRNAs has been investigated in various cancers, including CRCs (2, 3, 24, 25), none of the earlier studies dealt with race/ethnicity and tumor stage. The current investigation demonstrates that the prognostic value of different miRNAs varies with tumor stage and patient race/ethnicity. The results show a greater fold increase for all five miRNAs in CRCs of Blacks than for their White counterparts.

In normal prostate tissues, miR-301, miR-219, miR-26a, miR-1b-1, and miR-30c-1 are expressed three times higher in Blacks than in Whites (26). Similarly, prostate cancer cell lines generated from Blacks have higher expression of miR-26a compared to lines derived from Whites of similar stage and pathological grade (27). These results suggest that racial differences exist in the expression levels of some miRNAs. Plausible mechanisms for aberrant expression of miRNAs include the alteration of miRNA copy numbers, epigenetic modification of miRNAs and/or miRNA processing proteins, and single-nucleotide polymorphisms (SNPs) in miRNA genes (28–31). Since, SNPs are specific to race/ethnicity (32), associations between SNPs in miRNAs or in their target genes may contribute to distinct phenotypic features in different race/ethnic groups. Also, dysregulations of specific miRNAs are associated with stage of the disease and survival in several malignancies (3, 33–36). Furthermore, miRNA expression profiles can reflect specific stages in tumor progression (37); and several miRNAs, referred to as “metastamirs,” are involved in tumor metastasis, even though some of these may not have obvious roles in tumorigenesis (38). These reports support our findings related to race- and stage-specific miRNAs in the prognosis of CRC patients.

In the present investigation, there was no substantial correlation between increased miR-21 levels and poor patient prognosis when all stages were considered, as previously reported (3, 39, 40). However, once the population was stratified by patient race/ethnicity and tumor stage, high expression of miR-21 was associated with poor prognosis of White patients with Stage IV CRCs. In the current study, miRNAs were isolated from macro-dissected tumor tissues, whereas prior studies evaluated miRNAs from whole CRC tissues (3, 27, 39). Highly expressed in the stroma of CRCs, miR-21 is associated with shorter disease-free survival (41). Moreover, there is a differential pattern of expression of miRNAs during CRC progression (42). These factors may have contributed to the inconsistency between our findings and those of previous studies. Similar to our findings, the high expression of miR-21 in advanced stages of CRCs correlates with distant metastases and shorter patient survival (39, 40, 43). Furthermore, highly expressed miR-21 is more common in Stage IV CRCs than in Stage II and Stage III CRCs (44). These reports support our finding of miR-21 as an independent prognostic indicator of patients with Stage IV CRCs.

Increased miR-181b was identified and validated as an independent marker of poor prognosis for Black patients with Stage III CRCs. There are high levels of miR-181b in CRCs (2) and in sessile serrated adenomas, know aggressive lesions, relative to hyperplastic polyps (45). Even though there is high miR-181b expression in CRCs, it was considered to have no prognostic value (46). Our findings, however, suggest that, for Stage III Black patients, miR-181b is a prognostic marker that could aid in identifying high-risk patients who would benefit from adjuvant therapy. MiR-181b acts as an epigenetic switch to inhibit the tumor suppressor CYLD, thereby increasing NF-κB activity and maintaining the transformed state of tumor cells (47). In silico analyses have predicted several target genes of miR-181b that are involved in cell cycle regulation, cell signaling, and chemosensitivity (2). SNPs in the miR-181b binding sites of the targets may be a reason for increased expression of miR-181b in CRCs. For example, in one of the proposed targets of miR-181b, GATA6, which has an oncogenic function in gastrointestinal cancers (48), three SNPs have been identified (49). Since SNPs are race/ethnic-specific and have prognostic value in CRCs (50), there may be an association between miR-181b and SNPs in its targets that contributes to CRC carcinogenesis in Black patients. However, the underlying mechanisms for the prognostic value of miR-181b in Black patients with Stage III CRCs need to be explored.

Our study has found, for the combined population that high expression of miR-203 is an independent marker of poor prognosis of CRCs. Further, miR-203 is an independent marker of poor prognosis for Blacks with Stage I or II disease, but, in Whites, it is a marker for Stage IV disease. In CRC patients <40 years of age, there is increased expression of miR-203 that correlates with aggressive tumor behavior (25). Elevated expression of miR-203 is a predictor of poor prognosis for pancreatic adenocarcinoma (51, 52) and for breast cancer (53). In contrast, there is decreased expression of miR-203 in lung cancer cells, and it inhibits proliferation and invasion (54). These findings suggest that the role of miR-203 varies with the type and stage of cancer and patient race.

Although the current study did not find an independent prognostic value for miR-20a or mi106a, in univariate analyses, their increased expression correlated with poor survival in the combined population. A meta-analysis of 20 studies on miRNA expression levels in CRCs found up-regulation of miR-20a in more than one study (55). High levels of miR-20a in gastric and gastrointestinal cancers correlate with reduced survival (56, 57). MiR-106a highly is expressed in metastatic CRC cell lines (58), CRCs tissues (3, 59) and stool samples of CRC patients (60). With respect to CRC prognosis, conversely to what we found, low expression of miR-106a was associated with decreased survival in a Spanish CRC population (61).

We acknowledge that there are limitations of our study. To begin, this study was conducted with a sample set collected at a single medical center, thus, it is not a population-based study. The samples used were remnants of diagnostic tissues that were processed and archived for more than two decades. Thus, there may be biases related to tissue collection and processing (62). For the discovery and validation of biomarkers, reference sample sets that are collected following standardized protocols are ideal. Further, the validation studies should ideally be performed on a sample cohort collected from a different institution. However, due to limited resources and the exploratory nature of this study, both our test and validation cohorts were collected from the same institution. Nevertheless, our future studies will focus on validating these findings in reference CRC samples collected from Blacks and Whites of different geographical regions of the United States. Lastly, although we used self-identified race to categorize patients into Blacks and Whites, we recognize that there is some diversity in identification within any race or ethnic group.

The present investigation is the first to evaluate the prognostic value of miR-20a, miR-21, miR-106a, miR-181b, and miR-203 in CRCs based on patient race/ethnicity and tumor stage. These miRNAs were expressed greater than 2-fold higher in primary CRCs of Black patients than in their White counterparts. Further, miR-21 is an independent prognostic marker for Stage IV CRCs of Whites; miR-181b is an independent prognostic marker for Stage III CRCs of Blacks; and miR-203 is an independent prognostic marker of Blacks with early-stage CRCs and for Whites with late-stage CRCs. Although these findings need to be validated in prospective studies, the results warrant that race/ethnicity of patients and stage of the disease should be considered in assessing the clinical utility of miRNAs in CRCs.

TRANSLATIONAL RELEVANCE.

The clinical utility of miRNAs, especially for assessing patient prognoses and predicting the efficacy of therapy, is promising. Since most prior studies were conducted in White patient populations, the value of miRNAs in CRCs based on race/ethnicity has not been assessed. We have demonstrated that miRNAs have distinct prognostic value. Increased expression of miR-21 and miR-203 were associated with poor survival of Whites with Stage IV CRCs, whereas miR-203 was an independent indicator for Blacks with early stage CRCs. In contrast, miR-181b was an independent prognostic marker only for Stage III CRCs of Blacks. These findings suggest that, for evaluation of the clinical utility of miRNAs in relation to CRCs, patient race/ethnicity and tumor stage should be considered. Furthermore, these findings have clinical implications in identifying aggressive phenotypes of CRCs and in identifying high-risk patients, thus maximizing the benefits of adjuvant therapy.