Panel of serum miRNAs as potential non-invasive biomarkers for pancreatic ductal adenocarcinoma

Early-stage diagnosis of pancreatic ductal adenocarcinoma (PDAC) is difficult due to non-specific symptoms. Circulating miRNAs in body fluids have been emerging as potential non-invasive biomarkers for diagnosis of many cancers. Thus, this study aimed to assess a panel of miRNAs for their ability to differentiate PDAC from chronic pancreatitis (CP), a benign inflammatory condition of the pancreas. Next-generation sequencing was performed to identify miRNAs present in 60 FFPE tissue samples (27 PDAC, 23 CP and 10 normal pancreatic tissues). Four up-regulated miRNAs (miR-215-5p, miR-122-5p, miR-192-5p, and miR-181a-2-3p) and four down-regulated miRNAs (miR-30b-5p, miR-216b-5p, miR-320b, and miR-214-5p) in PDAC compared to CP were selected based on next-generation sequencing results. The levels of these 8 differentially expressed miRNAs were measured by qRT-PCR in 125 serum samples (50 PDAC, 50 CP, and 25 healthy controls (HC)). The results showed significant upregulation of miR-215-5p, miR-122-5p, and miR-192-5p in PDAC serum samples. In contrast, levels of miR-30b-5p and miR-320b were significantly lower in PDAC as compared to CP and HC. ROC analysis showed that these 5 miRNAs can distinguish PDAC from both CP and HC. Hence, this panel can serve as a non-invasive biomarker for the early detection of PDAC.

Pancreatic ductal adenocarcinoma (PDAC) is the fourth leading cause of cancer-related deaths worldwide with a 5-year survival rate of only 9% 1 . Most of the times when the PDAC is diagnosed, it is already at an advanced stage where it cannot be resected. Hence, there is an urgent need to find biomarkers for the early detection of PDAC.
MicroRNAs (miRNAs) are a subset of small non-coding RNAs that regulate gene expression at the posttranscriptional level by binding to the 3ʹ untranslated region (UTR) of target mRNAs 2 . Altered expression of miRNAs has been shown to be involved in the regulation of crucial pathological processes in tumorigenesis, progression and metastasis [3][4][5] . Aberrant expression of miRNAs has been detected in a wide variety of cancers including breast, lung, colorectal, hepatocellular and even pancreatic cancer [6][7][8][9][10] . Several groups have shown that expression of miR-21, miR-155, miR-196a and miR-222 increases in pancreatic tumour tissue [11][12][13][14] . Recent studies have looked at the possibility of using these miRNAs as possible diagnostic and prognostic markers 15,16 . Several reports indicate that miRNA expression profiles could be useful in the diagnosis of specific cancer types 17,18 .
It has been well established that serum contains a large number of stable miRNAs derived from various tissues/organs. Highly expressed miRNAs in specific tissues may leak into circulation as biomarkers for tissue injuries, and recent studies have shown the utility of miRNAs as biomarkers for cancers like pancreatic, colorectal, lung, prostate cancer etc [19][20][21][22][23] . A few studies have profiled the plasma miRNAs in PDAC and even shown their clinical significance 19,24 .
A large, retrospective cohort study in which patients had chronic pancreatitis (CP) of at least five years duration before being diagnosed with PDAC found a 14-fold increased risk in these patients 25  www.nature.com/scientificreports/ risk factor for the development of PDAC; it seems a promising idea to screen the CP patients for the malignant transformation associated phenotypic changes. However, till date, no blood-based marker has been shown to be useful in the early detection of PDAC in this high-risk, asymptomatic population. Early diagnosis for PDAC requires markers with high sensitivity and specificity. The standard serum marker, sialylated Lewis blood group antigen CA19-9, is widely used, but its use is limited to monitoring responses to therapy and not as a diagnostic marker 26 .
To the best of our knowledge, no previous study has used Next Generation Sequencing (NGS) to identify miRNAs expressed in CP. Hence, in the present study, both high-throughput NGS and quantitative reverse transcription PCR (qRT-PCR) assays were used to characterize the miRNA expression profile in FFPE tissue and serum from PDAC and CP patients. The goal was to identify a panel of serum miRNAs that could serve as a diagnostic biomarker for the early detection of PDAC in a high-risk group like CP patients.

Results
Selection of candidate miRNAs from the NGS data. NGS analysis was performed to identify the differentially expressed miRNAs in 27 PDAC, 23 CP, and 10 normal pancreatic tissue specimens from autopsy cases. Results from PDAC and CP groups were compared with the normal pancreas. A miRNA was considered differentially expressed when it had both fold change ≥ 2.0 or ≤ 0.5 and a p-value < 0.05. Based on these criteria, 219 miRNAs were identified to be differentially expressed between patients with PDAC and CP, including 120 up-regulated miRNAs and 99 down-regulated miRNAs. From the list of differentially expressed miRNAs, 8 miR-NAs were selected for further validation with qRT-PCR-four up-regulated miRNAs (miR-215-5p, miR-122-5p, miR-192-5p and miR-181a-2-3p) with the highest fold change between PDAC and CP, and the four most downregulated miRNAs (miR-30b-5p, miR-216b-5p, miR-320b, and miR-214-5p) in PDAC versus CP ( Table 1).

Levels of miRNAs in serum samples by qRT-PCR.
The levels of the eight selected miRNAs were measured in 125 serum samples (50 PDAC, 50 CP, and 25 HC) using qRT-PCR. The levels of these miRNAs were then compared among the three groups, including the comparisons between PDAC and CP, and between PDAC and HC. As shown in Fig. 2a-c, three miRNAs including miR-215-5p (P < 0.001), miR-122-5p (P < 0.01), and miR-192-5p (P < 0.001) were significantly up-regulated in the serum of PDAC patients compared to that of CP patients. Similarly in comparing the PDAC and HC samples, the same three candidate miRNAs (miR-215-5p, miR-122-5p and miR-192-5p) had significantly up-regulated levels (p < 0.0001 for each) in the serum of PDAC patients than in the HC (Fig. 2a-c). However, the serum levels of miR-181a-2-3p, exhibited no significant difference among the three studied groups (p = 0.45, 0.38 and 0.56; Fig. 2d) even though its tissue expression was much higher in PDAC (Fig. 1d).
As shown in Fig. 2a-c, the circulating levels of miR-215-5p, miR-122-5p and miR-192-5p in serum samples gradually increased from HC to CP to PDAC, suggesting their role in tumor progression. It is noteworthy that the difference in levels from HC to CP was not significant; however, between CP and PDAC, the enhancement in their levels became highly significant (Fig. 2a-c). Furthermore, the levels of miR-30b-5p in serum gradually decreased from HC to CP to PDAC, suggesting its role as a tumor suppressor (Fig. 2e). www.nature.com/scientificreports/ Further assessment of the expression levels of miR-215-5p, miR-122-5p and miR-192-5p from FFPE tissue samples ( Fig. 1a-c) showed an overall upregulation in PDAC tissues which correlated well with the higher levels in PDAC serum samples ( Fig. 2a-c). Similarly, the expression levels of miR-30b-5p and miR-320b from FFPE tissue samples ( Fig. 1e and g) showed an overall downregulation in FFPE PDAC tissue samples and also in PDAC serum samples ( Fig. 2e and g). The serum levels of these miRNAs replicated the expression level patterns seen in FFPE tissues suggesting that these miRNAs could serve as a non-invasive biomarker for the early detection of PDAC.
The tissue and serum levels of the miRNAs were also analyzed according to the stage of PDAC (Tables 2 and  3). The results showed that six of the miRNAs were significantly different between FFPE tissues of CP and early stages of PDAC (Table 2). However, miRNA levels were not different when comparing early stages of PDAC with advanced stages. Similarly, in serum samples of CP and early stages of PDAC, five of the miRNAs showed significantly different levels (Table 3). This suggests that these miRNAs may be relevant for early detection of PDAC.

Discussion
Early diagnosis of PDAC presents a challenge due to the lack of sensitive and specific biomarkers; hence, there is a need for new early diagnostic tools. Recent studies by Jin et al. 27 and others 19,26,28,29 have demonstrated that miRNAs are stable in serum/plasma and hence, can be readily detected by various assays such as NGS technology, miRNA microarray, and qRT-PCR. More importantly, the unique serum/plasma miRNA expression profiles for pancreatic cancer may serve as fingerprints for their detection 19,26,28,29 . A serum miRNA-based biomarker   www.nature.com/scientificreports/ associated with pancreatic cancer could possibly allow detection of the tumors without invasive procedures such as biopsy and surgery. In this study, the expression levels of tissue miRNAs in PDAC, CP and normal pancreas were determined by NGS. Eight differentially expressed miRNAs were selected and their levels were quantified in serum of PDAC, CP patients and healthy controls by qRT-PCR. The results showed that 5 serum miRNAs; three up-regulated (mir-215-5p, mir-122-5p, and mir-192-5p) and two down-regulated (mir-30b-5p and mir-320b) have strong potential as a biomarker for the early detection of PDAC.
In the current study, a significant increase was observed in the expression of miR-215-5p, miR-122-5p and miR-192-5p in PDAC tissues compared to both CP tissue and normal pancreatic tissue using two different techniques (NGS and qRT-PCR). This suggests their involvement in tumour progression; however, there are conflicting reports in the literature about the role of these miRNAs in various cancers. The expression of miR-215-5p was reported to be up-regulated in gastric cancer and gliomas 30,31 ; with higher miR-215-5p expression associated with higher-grade of glioma 32 . But reports in other cancer tissues suggest that it serves as a tumor suppressor 33 . Similarly, a lot of studies have looked at the expression of miR-122-5p in cancer tissues with most of them showing a tumor suppressor role for miR-122-5p. Most studies in pancreatic cancer and hepatocellular carcinoma show that miR-122-5p is underexpressed in tumor tissue 15,24,[34][35][36][37] . However, studies in renal cell carcinoma and colorectal cancer show overexpression of this miRNA in the tumor tissues 38,39 . Likewise, controversial results have been reported about the alteration of miR-192-5p in PDAC tissue specimens. While Flammang et al. 40 and Botla et al. 41 reported lower expression of miR-192-5p in PDAC tissue; others have reported increased expression in pancreatic cancer tissue 24,42,43 . In fact, ectopic expression of miR-192-5p associates with enhanced cell proliferation and migration, reduced apoptosis and promotes cell cycle progression in pancreatic cancer 43 .
Interestingly, studies measuring the circulatory levels of all these three miRNAs show higher levels in cancer patients compared to controls, which is in concordance with data in the present study. At the blood level, it has been shown that serum miR-215-5p has a relatively high value as a promising biomarker in diagnosing hepatocellular carcinoma 44 and osteosarcoma 45 . In keeping with our findings, different studies identified the up-regulation of miR-122-5p in plasma or whole blood of patients with PDAC 24,46,47 . Mazza et al. found that though plasma miR-122-5p levels had low specificity as a diagnostic marker with no significant difference between levels in CP and PDAC patients but it could serve as an independent negative prognostic factor as it significantly associated with tumor stage and metastasis 47 . Studies in other cancer types like hepatocellular carcinoma and colorectal cancer also show increased serum levels of miR-122-5p 38,48 . Levels of miR-192-5p have also been reported to be higher not only in serum or plasma of pancreatic cancer patients but also in the serum of esophageal squamous cell carcinoma (ESCC) patients 24,42,49 .
In contrast to the conflicting literature regarding these upregulated microRNAs, the available reports about the role of miR-30b-5p and miR-320b in cancer are more consistent. The lower expression of both these micro-RNAs in the present study indicated that they act as tumor suppressors in PDAC. Indeed, miR-30b-5p is found to be down-regulated in numerous human cancers, including PDAC 50 . Qin et al. analyzed the expression of miR-30b-5p in 90 patients of HCC and found that expression of miR-30b-5p was down-regulated in HCC tissue, thus inferring that miR-30b-5p acts as a tumor suppressor 51 . Moreover, another study showed that the expression level of miR-30b-5p in 32 ESCC tissues was significantly lower than that in adjacent normal tissues 52 . Liu et al. also found that miR-30b-5p functions as a tumor suppressor in renal cell carcinoma by targeting cell proliferation, metastasis and epithelial-to-mesenchymal transition 53 . As a member of the miR320 cluster, miRNA-320b has also been identified to play a suppressive role in various tumors, including colorectal cancer 54 , glioma 55 Table 4. Diagnostic performance of the five selected serum miRNAs in discriminating PDAC patients from chronic pancreatitis. *AUC, area under the curve; CI, confidence interval.  www.nature.com/scientificreports/ and nasopharyngeal carcinoma 56 . Tadano et al. found that miR-320 family, which inhibits cell proliferation, is frequently down-regulated in colorectal adenoma and submucosal invasive carcinoma tissues 57 . A recent study showed that transfection of pancreatic cancer cell lines with miR-320b mimics decreased their proliferation and invasion ability indicating that miR-320b has a tumor-suppressive role 58 .
Since, the present study revealed that the expression pattern of miR-215-5p, mir-122-5p and miR-192-5p is gradually increasing from HC to CP to PDAC (Fig. 2) while the levels of miR-30b-5p in serum gradually decreased from HC to CP to PDAC, hence, monitoring their serum levels at the stage of CP could be an effective marker for the early detection of PDAC.
Another interesting observation from this study is that pattern of serum miRNA levels does not necessarily correspond to their expression pattern in the tissue. In the present context, expression of miR-181a-2-3p, miR-216b-5p and miR-214-5p were significantly different in PDAC tissue compared to CP tissue by both NGS and qRT-PCR but there was no significant difference in their serum levels between the two groups of patients. This observation is not unique to this study as many other reports show little or no concordance between tissue and circulatory levels of the same microRNA 24,40 . This indicates that due care should be given to determine the miRNA levels in both tissue and serum as miRNAs showing concordance would serve as better biomarkers.

Conclusion
In conclusion, this study identified 5 serum miRNAs (miR-215-5p, miR-122-5p, miR-192-5p, miR-30b-5p and miR-320b) that could differentiate PDAC cases from CP and healthy controls with a high degree of accuracy. Thus, these identified miRNAs could potentially function as early non-invasive diagnostic biomarkers of PDAC. Further studies are needed to validate these findings with larger sample size. Future investigations could also explore the molecular mechanism employed by these miRNAs in the progression of pancreatic ductal adenocarcinoma.

Materials and methods
Ethics statement. This study was approved by the Institutional Ethics Committee, All India Institute of Medical Sciences, New Delhi, India. Written informed consent was obtained from participants for the use of their tissues and blood samples in this study. All methods were performed in accordance with the relevant guidelines and regulations. RNA isolation from FFPE tissue blocks and serum samples. One FFPE block was selected from each patient for miRNA analysis. From each of these blocks, three 10-µm sections were cut and taken on fresh slides; the tumour area was marked and micro-dissected after deparaffinization. The tissue from 3 slides (of the same block) was scraped into a 2 ml tube and mixed with the lysis buffer provided in the miRCURY RNA Isolation Kit (Tissue) from Exiqon (Exiqon, USA). The rest of the protocol was according to the instructions pro- www.nature.com/scientificreports/ vided in the kit. The integrity of extracted RNA samples was analyzed using Agilent Bioanalyzer 2100 (Agilent Technologies, CA, USA). Total RNA was extracted and purified from 500 μl of serum using the miRCURY RNA isolation kit -Biofluids (Exiqon, Denmark) following the manufacturer's instructions. Finally, the RNA was eluted in 50 μl volume.
Library construction and next-generation sequencing. Small RNA libraries were constructed using the Illumina Sample Preparation Kit, following the TruSeq Small RNA Sample Preparation Guide (Illumina, CA, USA) 59 . Approximately 500 ng of total RNA was used. The 3′ adaptors were ligated to the specific 3′OH group of miRNAs followed by 5′ adaptor ligation. The ligated products were reverse transcribed (Superscript III, Invitrogen, Whitefield, Bangalore) and amplified by PCR (17 cycles). The amplified libraries were size selected in the range of 140 bp-160 bp followed by overnight gel elution and salt precipitation 60 . Finally, the sequencing library was quantified by qRT-PCR using the Kapa Library Quantification Kit (Kapa Biosystems, Wilmington, MA, USA) 61 . The qPCR quantified library was subjected to sequencing on an Illumina sequencer (NextSeq500, Illumina) for 75 bp single-end chemistry.
Bioinformatics analysis of miRNA sequencing data. The adapter sequences and low-quality sequences were removed prior to data analysis 62 . Sequences > = 16 bp and < = 35 bp length were considered for further analysis. All the sequences were aligned to the Homo sapiens genome using bowtie-1.1.1. Aligned reads were extracted and checked for ncRNA (rRNA,tRNA,snRNA and snoRNA) contamination. The unaligned reads to ncRNAs were clustered based on the 100% coverage and similarity to generate the read count. Clustered reads were used for Known miRNA prediction.
Further, homology search of these miRNAs was done against all matured Homo sapiens miRNA sequences retrieved from miRbase-21 database using NCBI-blast-2.2.30 + . Known miRNAs and the read usage statistics reports were generated. DGE (Differential gene expression) analysis was carried out using DESeq tool. A miRNA was considered differentially expressed when the miRNA had both a fold change of ≥ 2 or ≤ 0.5 and a P-value < 0.05. Only differential miRNAs with high abundance reads were selected for further investigation.
Validation of miRNA expression by real-time PCR. The expression levels of the eight candidate miR-NAs (miR-215-5p, miR-122-5p, miR-192-5p, miR-181a-2-3p, miR-30b-5p, miR-216b-5p, miR-320b, and miR-214-5p) selected from the NGS data were measured by qRT-PCR. 6 μl of total RNA was converted to cDNA in a 20 μl reaction mixture using the miRCURY LNA Universal cDNA Synthesis Kit (Exiqon, Denmark) following the manufacturer's instructions. Quantification by real-time PCR was performed on Stratagene Mx3005P thermal cycler (Agilent Technologies, USA), using 4 μl of 1:40 diluted cDNA per tube along with the Exilent SYBR miRNA assay kits (Exiqon, Denmark), and LNA based specific primer assays (Qiagen, Germany). MxPro qRT-PCR software was used for the generation of the data. The miRNA expression was assayed in triplicate. Normalization of the miRNA expression levels was done using U6 as a reference control for tissue miRNAs and UniSp6 as an exogenous control for serum miRNAs. Results were analyzed by checking the relative expression of miRNAs by the 2 −ΔCT method. Statistical analysis. Statistical analysis was performed using STATA 14.0 statistical software (Texas, USA) and graphs were generated using GraphPad Prism 8.0. One-way ANOVA test was used to evaluate the differential expression of serum miRNA levels between PDAC, CP patients and healthy individuals. The significance of tissue miRNA expression levels was determined by Mann-Whitney test. The area under the ROC curves (AUC) generated by STATA 14.0 was used to evaluate the performance of the selected miRNAs in discriminating the PDAC from CP and HC. Sensitivity and specificity were calculated at an optimal cutoff point to determine the accuracy of differentially expressed miRNAs. A p-value < 0.05 was considered to be statistically significant.