The core concept of this study was to use the differential expression of miRNAs as a tool to further unravel the biology of human pancreatic cancer in the hope of uncovering new targets or vulnerabilities that might be exploited therapeutically; and/or finding new biomarkers or combinations of biomarkers that might contribute to better or earlier diagnosis [
25]. We believe that the study of miRNAs may be particularly fruitful since each miRNA targets multiple mRNAs, regulating the cellular levels of the corresponding proteins
The data presented here can be viewed in a number of different ways, simple lists of miRNAs DE between tumor and normal and between F1 and tumors (and in some cases both), and because of our previously published information on DE mRNAs and proteins, it is also possible to tentatively identify predicted protein targets of specific DE miRNAs.
Previous analyses of these proteomic [
19,
24] and mRNA [
20] or datasets on these clinical and F1 samples have looked at each dataset individually. mRNA analysis [
20] showed an enrichment of genes associated with proliferation, cell cycle, and mitotic processes. Label-free LC-MS-based proteomic analysis carried out on this sample set identified a number of membrane proteins in both comparisons (TvN, F1vT) that had the potential to be utilized either as biomarkers or ADC drug conjugate targets [
24].
The study of miRNAs in the development, progression, and prognosis of PDAC has intensified in the last decade. Interestingly there are few studies including transcriptomic and proteomic analysis as well as miRNA analysis on the same sample set. This limitation has meant that protein and gene targets of interesting miRNAs have predominantly been identified using predictive software and have not been validated in matched samples [
8,
25,
26,
27,
28,
29]. Mattie et al. [
30] performed miRNA analysis on matched tumor vs. PDX samples; however, this study was complicated by a high degree of homology with mouse miRNA sequences. Hanoun et al. [
31] initially published a study looking at the silencing of miR148a as a potential therapeutic target in PDAC. However, following further studies in which they looked at the role this miRNA had in the modulation of protein expression, they discovered that miR148a was responsible for only minimal protein regulation and were unable to identify protein targets in cell lines [
32]. This indicates the importance of looking at parallel analysis where possible, as it may help to reduce the number of false leads emerging from in silico studies [
33] miR148a was shown to be significantly downregulated in our patient tumor when compared with the normal-adjacent samples fold change of −8.15, it was, however, unchanged in the F1vT comparison. The miRWalk prediction software identified nine targets for this miRNA; however, when compared with the protein list, none of these were inversely expressed.
The investigation of miRNAs, with mRNA and proteomic investigations in parallel from the same samples, as we propose, may lead to the identification of proteins that could be druggable targets. Furthermore, RNA therapeutics are now a reality, and miRNA manipulation has the potential to become a useful addition to the therapeutic armamentarium against pancreatic cancer [
15,
34].
4.1. Tumor vs. Adjacent Normal Tissue Comparison
Examination of the dysregulation of miRNA in our dataset provided a short list of interesting miRNA; the top 10 up and down differentially expressed miRNAs in the (a) TvN comparison (
Table 1), (b) F1vT comparisons (
Table 2), and (c) miRNAs that are differentially expressed in the same direction in both comparisons (
Table 3). Of the 39 miRNAs within this priority list, 7 of them predicted an effect on multiple proteins, which were found to be inversely expressed in the corresponding protein dataset (
Table 4,
Table 5 and
Table 6). Several studies in gastric cancer [
35], triple-negative breast cancer [
36], and lung cancer [
37] have indicated roles for miR509. This miRNA, however, has not been implicated in the literature on pancreatic cancer. Our data indicated that miR509 was shown to be downregulated in the patient tumors when compared to the normal adjacent, with a fold change of −1.85 (
Supplementary Table S1). It was predicted to target a number of proteins in our list, including COTL1, LRPAP1, and SL4A2 (
Table 4). Predicted targets of the remaining miRNAs were shown to be differentially expressed in the TvN comparisons, although none of their predicted targets were found to be inversely expressed in the proteomic dataset, but alterations in the expression of miR708 [
38], miR210 [
39,
40], and miR214 [
41,
42] have all been studied in pancreatic cancer.
In the TvN comparison, it would be expected that we would see the alteration in the expression of proteins and miRNAs associated with tumorigenesis and proliferation in the pathways associated with this transition. miR21 has been extensively studied in PDAC, with high expression levels being linked with shorter overall survival [
43,
44,
45], while miR331-5p has not been reported in any cancer-linked study. miR21-5p and miR331-5p were shown to target Dimethylarginine Dimethylaminohydrolase 1 (DDAH1). Reduced levels of DDAH1 have been shown to mediate cell invasion and metastasis in gastric cancer through the WNT signaling pathway. Additionally, downregulation was linked with poor prognosis [
46]. These miRNAs were both shown to be upregulated in the TvN comparison, and the expression of DDAH1 was shown to be downregulated in the membrane protein fraction of the same comparison. Our data suggest that both miR181c and miR181d target phosphatidylethanolamine binding protein 1 (PEPB1) with a reduction in expression in the TvN in both the membrane and cytoplasmic fraction. Loss of expression of PEBP1 (also known as raf kinase inhibitor protein (RKIP), an endogenous inhibitor of the MAPK pathway) has been shown to be linked with PDAC metastasis and poor overall survival [
47,
48].
miR210-3p was predicted to target endoplasmic reticulum protein 27 (ERP27), and the protein was shown to be inversely expressed in the membrane fraction (
Table 4). Transcriptomic analysis of the cancer genome atlas has identified ERP27 as a pivotal gene in the pathogenesis and progression of PDAC [
49]. Increased glutathione S-transferase M3 (GSTM3) expression has been linked with improved overall survival in PDAC [
50]. This protein was inversely regulated by miR143-5p, which was found to be upregulated in the tumor samples when compared with the normal-adjacent samples. Literature searches showed no previous studies linked with cancer and the expression of miR4742. Our proteomic data shows that miR4742 targets a number of proteins within the dataset, including superoxide dismutase 2 (SOD2), cysteine and glycine-rich protein 1 (CSRP1), and protein S100-A16 (S100A16). SOD2 is involved in oxidative phosphorylation and the regulation of ROS (reactive oxygen species), and upregulation in its expression has been linked with acquired resistance to ROS-inducing anticancer drugs and potential irradiation in pancreatic cancer cell lines [
51]. While little is known about CSRP1 in pancreatic cancer, S100A16 has been extensively studied in PDAC. Roles in metastasis [
52], epithelial–mesenchymal transition (EMT) [
53], and increased expression linked with poor prognosis have all been reported in the literature. In our protein data, we do see an increased expression in the membrane fraction in both the TvN and F1vT comparison of S100A16.
4.2. F1 Xenograft vs. Tumor Comparison
In the case of the F1vT comparison, our previous studies on mRNA and protein expression changes on the samples that were used here have highlighted the importance of proteins and mRNA that drive the important process of tumor engraftment in tumorigenesis. As described by Coleman et al. [
24], the selection pressure that is exerted on the tumor cells during the engraftment process may result in the altered expression of proteins that drive tumor growth. The upregulation of miR125a in the PDAC samples agrees with what is already shown in the literature [
54], and it has been shown that its expression can contribute to cell survival [
55]. One of its predicted targets is SGT1 homolog (SUGT1), a highly conserved protein generally found in the nucleus, involved in the kinetochore function and essential for the G1/S and G2M transition, which was shown to be increased in expression in the cytosolic fraction of the tumors in comparison to the tumor sample., The expression of this protein has been potentially linked with a poor prognosis in colorectal cancer [
56].
Keklikoglou et al., 2015, found that the expression of miR206 was downregulated in pancreatic tumors vs. non-malignant normal tissue [
57]. In our study, miR206 was not differentially expressed in the TvN comparison; however, it was significantly upregulated in the F1vT comparison with a fold change of 26.01. Wu et al. showed that the expression of miR-206 inhibited hepatocellular carcinoma cell migration and invasion while promoting apoptosis by targeting Peptidylprolyl Isomerase B (PPIB) [
58]. Another study evaluated the expression of 41 genes as a function of in vitro radioresistance in the NCI-60 cancer cell line panel and found that PPIB had the strongest direct correlation. They also showed that siRNA downregulation of PPIB leads to radio-sensitization of the cancer cells [
59]. PPIB expression was shown to be downregulated in both comparisons (TvN and F1vT) in our proteomic data.
Levels of miR2467 were shown to be significantly increased in the F1 samples in comparison to the original patient tumor. This miRNA was predicted to target a large number of proteins, and when examined in our protein dataset, a large number of proteins showed inverse expression. There are limited studies in the literature on this miRNA, with one indicating a downregulation of the miRNA in colorectal tumor samples compared to tumor-adjacent normal tissue [
60]. We observed a decrease in the expression of Aflatoxin B1 aldehyde reductase member 2 (AKR7A2) in both of our comparisons, not reported in the proteomic analysis by Cui et al. [
61]. Furthermore, this protein was found to be targeted by two miRNAs in our analysis (miR2467 and miR1290). The latter has been suggested as a biomarker in PDAC [
62]. Wei et al. [
63] and Tavaano et al. [
64] both examined the expression of miR1290, with and without CA19-9 in serum and plasma, respectively, and noted that while the expression was found to be higher in PDAC patients, it alone was insufficient to select patients at risk of developing PDAC. Both studies indicated that combining the expression of miR1290 with Ca19-9 was more effective. Whole miRNome and proteome profiling revealed miR1290 as a novel hypoxia-associated microRNA, which was highly abundant in hypoxic extracellular vesicles (EVs) and also exhibited a signature consisting of six proteins including AKR7A2, which was significantly associated with a poor prognosis for melanoma patients [
65].
miR615 overexpression has been shown to inhibit cell proliferation, migration, and invasion in vitro [
66]. miR615 was increased in the F1 samples compared to the original patient tumors, and a number of predicted target proteins were found to be inversely expressed in the matching transition in proteomic data. These include several enzymes involved in the reduction and oxidation of aldehydes (AKR7A2, aldehyde dehydrogenase 1 family member B1 (ALDH1B1). Aldehyde dehydrogenase (ALDH2)) and aldehyde dehydrogenases are often highly expressed in cancer stem cells. ALDH1B1 may have a role in cell proliferation [
67]. It has also been suggested as a potential biomarker of colon cancer [
68]. It was another predicted target of miR615 found in our proteomic dataset to be downregulated in the F1vT comparison as well as in the cytosolic fraction of the TvN comparison. Calnexin (CANX), a member of the calnexin family of molecular chaperones, is a calcium-binding, endoplasmic reticulum (ER)-associated protein that interacts transiently with newly synthesized N-linked glycoproteins, facilitating protein folding and assembly. It may also play a central role in the quality control of protein folding by retaining incorrectly folded protein subunits within the ER for degradation. This protein has the potential as a biomarker in breast [
69] and lung patients [
70]. Several of the other miR615-3p targets identified in our analysis showed consistent changes across both comparisons and are involved in processing membrane protein targets to ER (SECG1A1, SPC53, TM95F2).
The miR-378 family has been associated with numerous cancers, including: osteosarcoma [
71], cervical [
72], and colon [
73]. MiR-378e has not yet been extensively studied for an oncogenic/oncosuppressive role. Wang et al. [
73] have reported an association between miR-378e and colorectal cancer, as the survival rate of patients in the high-expression group was significantly higher than that in the low-expression group. In our dataset, it was shown to be differentially upregulated in the F1 samples vs. the patient samples. miR409 expression, which was downregulated in the F1 tumors vs. patient tumor samples, has been linked with PDAC cell proliferation, invasion, and migration [
74] proteins. Predicted to be targeted by miR409, Catenin Delta 1 (CTNND1) along with Catenin alpha 1 (CTNNA1) and Catenin Beta 1 (CTNNB1) expression was shown to indicate a poor prognosis for pancreatic cancer patients [
75]. In the F1 samples, this protein was shown to be differentially upregulated versus the patient samples (
Table 5).
4.3. miRNA Altered in Both Comparisons
Across the two transitions, from normal to the tumor and then tumor to F1 models, there was a small cohort of miRNAs that were shown to be DE, namely miR4534, miR4743, miR222, miR3154, and miR6831. miR-6831-5p had been previously identified as being differentially expressed in PDAC vs. normal control serum samples, as shown by Aita et al. [
76]. KEGG pathway analysis showed enrichment for the pancreatic cancer term. Dysregulation of miR6831-5p has been associated with gastric [
77] and colorectal cancer [
78]. Additionally, PLD3, which we found to be downregulated in both the cytoplasmic and membrane protein fraction of the F1 tumor samples compared to the patient material, has been shown to be a potential marker of senescence [
79]. Limited studies have been completed on miR4534; however, one study indicated the potential roles of miR4534 in prostate cancer through the regulation of PTEN [
80]. miR3154 has been investigated in leukemia, with its expression shown to be higher at initial diagnosis compared to complete remission [
81]. miR222 has been widely studied in pancreatic cancer. Elevated levels of these miRNAs have been associated with poor survival [
82,
83,
84,
85].
Of course, a caveat with any research of this kind using tissue extracts is the fact that the samples consist of a heterogenous mixture of cell types and connective tissue, so we cannot, for example, be certain that an alteration observed in the tumor samples reflects an alteration in expression in the tumor cells (for definitive attribution some form of in situ hybridization would be needed for miRNAs and mRNAs, and immunohistochemistry in the case of proteins).