Identification of DElncRNAs, DEmiRNAs and DEmRNAs in pancreatic cancer
Microarray expression data of pancreatic cancer associated lncRNAs, miRNAs, and mRNAs were collected from GEO: This consisted of a total of 3 lncRNA expression datasets, including 122 pancreatic tumor tissue and 27 adjacent non-tumor tissue, 3 miRNA expression datasets, including 170 pancreatic tumor tissue and 38 non-tumor tissue, and 3 mRNA expression datasets, including 195 pancreatic tumor tissue and 32 non-tumor tissues (Table 1).
Table 1. Selected GEO dataset
Microarray
|
GEO accession
|
Number of tumour samples
|
Number of adjacent non-tumor samples
|
LncRNA microarray
|
GSE55643
|
45
|
8
|
|
GSE56560
|
28
|
7
|
|
GSE60979
|
49
|
12
|
|
Total
|
122
|
27
|
miRNA microarray
|
GSE24279
|
136
|
22
|
|
GSE32678
|
25
|
7
|
|
GSE41369
|
9
|
9
|
|
Total
|
170
|
38
|
mRNA microarray
|
GSE56560
|
28
|
7
|
|
GSE60979
|
49
|
12
|
|
GSE62165
|
118
|
13
|
|
Total
|
195
|
32
|
Using thresholds of |Log2FC|>1 in at least two datasets and p-value<0.05 in three datasets, the results show that 23 DElncRNAs (Table S1), including 15 upregulated DElncRNAs and 8 downregulated DElncRNAs, 40 DEmiRNAs(Table S2), including 31 upregulated DEmiRNAs and 9 downregulated DEmiRNAs, 1613 DEmRNAs (Table S3), including 1069 upregulated DEmRNAs and 544 downregulated DEmRNAs, were significantly abnormal expressed in pancreatic cancer. The top 5 obvious upregulated and downregulated DElncRNAs, DEmiRNAs and DEmRNAs were respectively shown in Table 2, 3, 4.
Table 2. Top 5 significantly up-regulated and top 5 downregulated DEmiRNAs in three GEO dataset.
|
|
Pancreatic cancer (n = 170), normal pancreas ( n = 38)
|
miRNA name
|
Up
/Down
|
GSE24279
|
GSE32678
|
GSE41369
|
P value
|
Log2FC
|
P value
|
Log2FC
|
P value
|
Log2FC
|
hsa-miR-217
|
Down
|
3.56E-17
|
-5.42539
|
|
|
0.00152
|
-3.57232
|
hsa-miR-216a
|
Down
|
2.99E-11
|
-4.74284
|
|
|
0.000781
|
-2.63622
|
hsa-miR-130b
|
Down
|
5.91E-21
|
-3.91727
|
|
|
0.000984
|
-2.61235
|
hsa-miR-141
|
Down
|
0.0000481
|
-2.35824
|
0.002175
|
2.114211
|
0.0481
|
-2.12812
|
hsa-miR-200c
|
Down
|
5.16E-09
|
-1.85622
|
0.001694
|
2.216186
|
0.0499
|
-1.68123
|
hsa-miR-145
|
Up
|
6.82E-12
|
1.748881
|
|
|
7.79E-06
|
3.244704
|
hsa-miR-150
|
Up
|
8.67E-15
|
3.328354
|
|
|
0.000944
|
3.28386
|
hsa-miR-1825
|
Up
|
0.025
|
4.068449
|
0.010976
|
1.225039
|
|
|
hsa-miR-135b
|
Up
|
|
|
2.14E-05
|
3.442542
|
4.63E-08
|
2.819987
|
hsa-miR-21
|
Up
|
|
|
8.13E-05
|
3.418418
|
0.000019
|
2.992536
|
Table 3.Top 5 significantly upregulated and downregulated DEmRNAs in three GEO dataset (logFC>3)
|
|
Pancreatic cancer (n =195), normal pancreas (n =32)
|
mRNA name
|
Up/
down
|
GSE56560
|
GSE60980
|
GSE62165
|
P value
|
Log2FC
|
P value
|
Log2FC
|
P value
|
Log2FC
|
ALB
|
down
|
6.2E-09
|
-5.929303
|
0.000022
|
-4.49572445
|
1.08E-14
|
-5.3
|
PNLIPRP1
|
down
|
0.000208
|
-4.93908
|
0.00312
|
-3.19119699
|
1.53E-15
|
-6.15
|
SERPINI2
|
down
|
0.00000774
|
-4.562986
|
0.000383
|
-4.02319809
|
1.94E-11
|
-6.47
|
CELA2A
|
down
|
0.00123
|
-4.271562
|
0.0086
|
-3.82716481
|
0.000000262
|
-6.02
|
AQP8
|
down
|
0.000000552
|
-3.68139
|
0.00000159
|
-4.06761851
|
9.79E-15
|
-6.17
|
MMP11
|
up
|
0.0000763
|
1.952182
|
2.59E-11
|
3.88137784
|
2.29E-27
|
4.86
|
GPRC5A
|
up
|
1.01E-12
|
3.512951
|
8.64E-10
|
2.2962543
|
3.55E-26
|
4.32
|
SLC6A14
|
up
|
0.00000065
|
3.930053
|
0.00000325
|
2.99596191
|
2.08E-14
|
4.65
|
POSTN
|
up
|
5.24E-09
|
4.378546
|
0.00000458
|
3.02612901
|
|
|
CEACAM6
|
up
|
1.26E-11
|
4.549112
|
5.42E-08
|
3.9202177
|
4.25E-13
|
3.96
|
Table 4.Top 5 significantly upregulated and downregulated DElncRNAs in three GEO dataset.
|
|
Pancreatic cancer (n = 122), normal pancreas (n =27)
|
LncRNA name
|
Up/
down
|
GSE55643
|
GSE60980
|
GSE56560
|
P value
|
Log2FC
|
P value
|
Log2FC
|
P value
|
Log2FC
|
LINC00671
|
Down
|
|
|
4.35E-06
|
-1.20411
|
3.02E-15
|
-2.15519
|
LINC00029
|
Down
|
|
|
2E-07
|
-1.73827
|
5.79E-07
|
-1.03315
|
FAM167A-AS1
|
Down
|
|
|
0.000514
|
-1.37194
|
1.43E-10
|
-1.25994
|
LINC00339
|
Down
|
0.00262
|
-1.50678
|
5.93E-06
|
-1.55368
|
1.15E-08
|
-1.88799
|
AKR7A2P1
|
Down
|
8.94E-05
|
-1.30069
|
4.25E-05
|
-1.87539
|
|
|
KRT19P2
|
Up
|
0.0285
|
1.279934
|
2.06E-09
|
2.987603
|
|
|
KRT16P6
|
Up
|
0.00594
|
2.026907
|
2.95E-07
|
2.754911
|
|
|
KRT16P2
|
Up
|
0.000937
|
2.834359
|
3.51E-08
|
3.462639
|
|
|
AFAP1-AS1
|
Up
|
0.000946
|
3.660372
|
0.000138
|
1.438305
|
0.0103
|
1.495363
|
HOXA11-AS
|
Up
|
0.00086
|
3.700575
|
0.000532
|
1.507758
|
|
|
Prediction of DEmiRNA-targeted DEmRNAs by miRWalk
We used the miRwalk database to predict the target gene on the above 40 DEmiRNA. A total of 2600 miRNA-mRNA relationship pairs were contained 34 DEmiRNAs and 1798 target mRNAs. By intersecting with DEmRNAs, a total of 83 miRNA-mRNA relationship pairs, including 23 DEmiRNAs and 69 DEmRNAs (Table S4), that 19 over-expressed DEmiRNAs downregulated 34 DEmRNAs (Figure 1A), while 4 under-expressed DEmiRNAs upregulated 35 DEmRNAs (Figure 1C).
Prediction of DElncRNA-DEmiRNA coaction by LncBase 2.0
We mapped the above mentioned 40 DEmiRNAs into the LncBase2.0 and searched for the DEmiRNA-targeted lncRNAs. By intersecting with DElncRNAs, There were 71 interactions between 32 DEmiRNAs (hsa-let-7d, hsa-let-7i, hsa-10a, hsa-125a-5p, hsa-125b, hsa-130b) and 12 DElncRNAs(RUNX1-IT1, HOXA11-AS, UNC5B-AS1, FAM167A-AS1, LINC01869, LINC00339, LINC00671, CMAHP, LINC00029, AFAP1-AS1, CYTOR, MIR4697HG). In detail, 7 over-expressed DElncRNAs downregulated 8 DEmiRNAs, while 5 under-expressed DElncRNAs upregulated 24 DEmiRNAs (Figure 2, Table S5).
Construction of ceRNA interaction network
We established a ceRNA regulatory network with pre-treated data (Figure 3, Table S6). In the network, there are 11 DElncRNAs, 19 DEmiRNAs, and 66 DEmRNAs. At length, 5 under-expressed DElncRNA upregulate 15 DEmiRNAs, while 15 over-expressed DEmiRNAs downregulate 31 DEmRNAs; 6 over-expressed DElncRNAs downregulate 4 DEmiRNAs, while 4 down-expressed DEmiRNAs over-regulate 35 DEmRNAs.
Furthermore, we topologically calculated the connectivity among each gene to elucidate its significance in the ceRNA interaction network (Table 5, Table S7). They were considered to be the pivotal genes with degree>5 among the lncRNAs and miRNAs, including LINC00029, FAM167A-AS1 , LINC00671; hsa-miR-130b; hsa-miR-141; hsa-miR-200c; hsa-miR-23a; hsa-miR-217; hsa-miR-125a-5p; hsa-miR-24; hsa-miR-27a; hsa-miR-199a-5p; hsa-miR-34a; hsa-miR-34c-5p. Since hsa-miR-130b was the most connected degree with 20 in the ceRNA regulatory network, we inferred that it might have a strong impact on the pancreatic cancer process. The lncRNA-miRNA-mRNA network also revealed that DElncRNAs indirectly acted with DEmRNAs.
Table 5. Topology parameters of 14 hub gene (degree>5) in the ceRNA network
Gene
|
Type
|
Betweenness
|
Closeness
|
Degree
|
LINC00029
|
lncRNA
|
678.87489
|
26.48333
|
10
|
FAM167A-AS1
|
lncRNA
|
782.85275
|
25.5
|
9
|
LINC00671
|
lncRNA
|
641.19704
|
25.9
|
8
|
hsa-miR-130b
|
miRNA
|
1261.33333
|
27.31667
|
20
|
hsa-miR-141
|
miRNA
|
784.66667
|
20.91667
|
10
|
hsa-miR-200c
|
miRNA
|
635.66667
|
19.45
|
10
|
hsa-miR-23a
|
miRNA
|
528.35005
|
24.23333
|
9
|
hsa-miR-217
|
miRNA
|
576.33333
|
19.45
|
8
|
hsa-miR-125a-5p
|
miRNA
|
426.70879
|
21.85238
|
7
|
hsa-miR-24
|
miRNA
|
486.77879
|
22.51667
|
7
|
hsa-miR-27a
|
miRNA
|
371.21975
|
22.9
|
7
|
hsa-miR-199a-5p
|
miRNA
|
528.86061
|
21.05
|
6
|
hsa-miR-34a
|
miRNA
|
227.63663
|
22.23333
|
6
|
hsa-miR-34c-5p
|
miRNA
|
175.35741
|
22.23333
|
6
|
Function enrichment analysis of DEmRNAs in ceRNA network
Using the Enrichr, we analyze the GO and KEGG analysis of 66 DEmRNAs based on the constructed ceRNA interaction network.
Significantly, GO BP analysis performed that candidate genes targeted by downregulated DEmiRNAs were participated in positive regulation of transcription by RNA polymerase II, positive regulation of transcription, DNA-templated (Figure 4A). CC analysis performed that these candidate genes were significantly concentrated in mitotic spindle, cell-cell junction and adherens junction (Figure 4C). For MF items, they took part in E-box binding, disordered domain specific binding and phosphatidylinositol-3,4,5-trisphosphate binding (Figure 4E). While for GO BP analysis demonstrated that candidate genes targeted by upregulated DE-miRNAs were significantly involved in the pyrimidine nucleobase catabolic process, pyrimidine nucleobase metabolic process and regulation of cytokine production (Figure 4B). CC analysis performed that these genes were significantly enriched inspanning components of the plasma membrane, PRC1 complex and endoplasmic reticulum membrane (Figure 4D). At last, MF analysis for these genes showed that they were major enriched in pyridoxal phosphate binding, potassium ion transmembrane transporter activity, and prostaglandin E receptor activity (Figure 4F).
Further, KEGG pathway analysis was revealed for 66 DEmRNAs based on the constructed ceRNA network. Candidate genes targeted by downregulated DEmiRNAs were significantly taking participated in bacterial invasion of epithelial cells, yersinia infection, microRNAs in cancer, transcriptional misregulation in cancer, adherens junction, melanoma, non-small cell lung cancer, Endocytosis, small cell lung cancer and prostate cancer (Table 6). The KEGG pathways for candidate genes targeted by upregulated DE-miRNAs were shown in Table 7.
Table 6. The top 10 enriched KEGG pathways for the intersection of target genes of downregulated DE-miRNAs and upregulated DEGs.
Term
|
count
|
Gene
|
p-value
|
Bacterial invasion of epithelial cells
|
3
|
CLTC;FN1;MET
|
3.29E-04
|
Yersinia infection
|
3
|
TBK1;WIPF1;FN1
|
0.001753
|
MicroRNAs in cancer
|
4
|
BMPR2;ZEB1;E2F3;MET
|
0.00203
|
Transcriptional misregulation in cancer
|
3
|
ZEB1;PPARG;MET
|
0.004548
|
Adherens junction
|
2
|
LMO7;MET
|
0.006854
|
Melanoma
|
2
|
E2F3;MET
|
0.007042
|
Non-small cell lung cancer
|
2
|
E2F3;MET
|
0.007042
|
Endocytosis
|
3
|
WIPF1;CLTC;ARAP2
|
0.00961
|
Small cell lung cancer
|
2
|
FN1;E2F3
|
0.011284
|
Prostate cancer
|
2
|
ZEB1;E2F3
|
0.012483
|
Table 7. The top 10 enriched KEGG pathways for the intersection of target genes of upregulated DE-miRNAs and downregulated DEGs
Term
|
count
|
Gene
|
p-value
|
beta-Alanine metabolism
|
2
|
ALDH6A1;ABAT
|
9.84E-04
|
Propanoate metabolism
|
2
|
ALDH6A1;ABAT
|
0.001265
|
Valine, leucine and isoleucine degradation
|
2
|
ALDH6A1;ABAT
|
0.002509
|
Phosphonate and phosphinate metabolism
|
1
|
PCYT1B
|
0.009265
|
Selenocompound metabolism
|
1
|
CTH
|
0.026036
|
Mannose type O-glycan biosynthesis
|
1
|
B3GALNT2
|
0.035067
|
Proximal tubule bicarbonate reclamation
|
1
|
SLC4A4
|
0.035067
|
Renin-angiotensin system
|
1
|
ANPEP
|
0.035067
|
Maturity onset diabetes of the young
|
1
|
NR5A2
|
0.039553
|
Butanoate metabolism
|
1
|
ABAT
|
0.042532
|
Validation of DElncRNA and DEmRNAs in the ceRNA Network
TCGA and GTEx datasets were analyzed by using GEPIA to demonstrate that aberrant expression of DElncRNAs and DEmRNAs were significantly over-expressed between pancreatic cancer and normal pancreas tissues (Figure 5) based on optimal cutoff (|Log2FC| >1, p<0.01).
For further research of the ceRNAs, we analyzed the association of the ceRNA network with clinical features in the TCGA database, such as neoplasm histologic grade and pathologic stage. The results showed that 1 lncRNA was associated with the pathologic stage, 2 miRNA were associated with years of birth, 4 miRNAs were associated with radiation therapy samples, 6 miRNAs were associated with pathologic M stage, and 18 mRNAs were connected to the clinical pathologic stage (P < 0.05; Figure 6-7, Table S8).
Relationship between key genes and overall survival
We conducted a survival analysis on each DElncRNA, DEmiRNA and DEmRNA in the ceRNA network to discover the potential prognosis RNA biomarker in pancreatic cancer. 1 DElncRNAs, 3 DEmiRNAs, and 19 DEmRNAs were tightly connected with overall survival, which met the cutoffs value (Figure 8, Table S9). Of these, 5 DEmRNAs appeared to exhibit protective functions, as pancreatic cancer patients with the higher expression level of these mRNAs were likely to get a longer prognosis. On the contrary, the remaining 1 DElncRNAs and 12 DEmRNAs were considered as risky factors because their expression was inversely connected to the prognosis of pancreatic cancer patients.
Positive correlation among the expression levels in ceRNA network
According to ceRNA theory, lncRNAs can positively influence mRNAs by competing with miRNAs for their binding sites. So we conducted a regression analysis between DElncRNAs and DEmRNAs targeted by miR-130b (Table S10), which was the most connected gene in the ceRNA network with a connection of degree 20. Intense positive associations were found between DElncRNAs and DEmRNAs, which were targeted by miR-130b in GSE60980 with an interaction score>0.4. It revealed the first three correlation coefficients of the interactions under the regulation of miR-130b, in which CYTOR interacted with FRMD6 and TCF4, while HOXA11-AS interacted with MET(Figure 9).
Clinicopathologic Significance of miR-130b in Pancreatic Cancer Patients
miR-130b was tested by qRT-PCR in the 11 pairs of PC tissues and matched NC tissues. Compared to the matched noncancerous pancreatic tissues, 11 PC tissues showed lower miR-130b expression with the median fold change of 1.23 (P<0.05). Moreover, Kaplan–Meier curves revealed that increased miR-130b was connective with tightly longer survival time in pancreatic cancer patients (p = 0.014) (Figure 10B).
MiR-130b Inhibited Cell Proliferation in vitro
To assess the function of miR-130b on PC cell proliferation activity, the proliferation rate of miR-130b overexpressing cells was assessed every 24h for 72h. miR-130b was successfully overexpressed in ASPC-1 cells and MIA PaCa-2 cells. In general, miR-130b enhanced the proliferative capacity of PC cells, as shown by the serial measurements of proliferation rates (Figure 11). These results proved that miR-130b could inhibit pancreatic cancer cell proliferation.
MiR-130b Suppressed Migration of Pancreatic Cancer Cell
To discover how miR-130b affected on pancreatic cancer cells, we treated MIA PaCa-2 and ASPC-1 cells with miR-130b mimics and control mimics in 50 nM. With overexpression of miR-130b, the proliferation of MIA PaCa-2 and ASPC-1 cells were significantly suppressed by 32.60401% and 23.79411%, respectively (P<0.01, Figure 12).
In addition, we used transwell to analyse the function of miR-130b in the pancreatic cancer cell migration. We discovered that the count of migrating ASPC-1 cells with miR-130b overexpression (2285 cells/high power magnification field (HP)) was significantly less than matched control groups(3615 cells/HP; Figure. 13A). Similar results were also found in MIA PaCa-2 cells (Figure 13B). It further revealed that the miR-130b mediated the migration of pancreatic cancer cells.