Differential microRNA expression profiles associated with microsatellite status reveal possible epigenetic regulation of microsatellite instability in gastric adenocarcinoma.

BACKGROUND: Although microsatellite instability (MSI) is a powerful predictive biomarker for the efficacy of immunotherapy, the mechanism of MSI in sporadic gastrointestinal cancer is not fully understood. However, epigenetics, particularly microRNAs, has been suggested as one of the main regulators that contribute to the MSI formation.
METHODS: We used microRNA expression data of 386 gastric adenocarcinoma samples from The Cancer Genome Atlas (TCGA) database to identify differential microRNA expression profiles by different MSI status. We also obtained putative common target genes of the top differential microRNAs with miRanda online tools, and we analyzed these data by Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes pathway enrichment (KEGG).
RESULTS: We found that 56 and 67 gastric adenocarcinoma samples were positive for low and high MSI, respectively, and that a high MSI status was associated with age, sex and subregion (P=0.049, 0.014 and 0.007, respectively). In the 67 samples with a high MSI status, expression levels of 14 microRNAs were upregulated but five microRNAs were downregulated as assessed by the fold change (FC), compared with that of the 56 samples with a low MSI status (P<0.05, |FC| >2). Further analysis suggested that the expression of miR-210-3p, miR-582-3p, miR-30a-3p and miR-105-5p predicted a high MSI status (P=4.93×10-10, 5.63×10-10, 3.23×10-9 and 7.64×10-4, respectively). Regulation of the transcription pathways ranked the top of lists from both GO and KEGG analyses, and these microRNAs might regulate DNA damage-repair genes that were also associated with a high MSI status.
CONCLUSIONS: MiR-30a-3p and miR-105-5p are potential biomarkers for the MSI-H gastric adenocarcinoma, possibly by altering expression of DNA damage-repair genes. 2020 Annals of Translational Medicine. All rights reserved.

Introduction

More than one million patients suffered from gastric carcinoma (GCa) with an estimated 783,000 GCa-related deaths in 2018, making it the fifth most common and the third most deadly cancer worldwide (1). For advanced GCa, the palliative and systemic chemotherapies are the mainstay of treatments, and its median overall survival (OS) is only 10–12 months (2). Recently, the immune checkpoint inhibitors have been used to treat advanced GCa with a high-frequency microsatellite instability (MSI-H) or mismatch repair defects (dMMR) (3,4). The Food and Drug Administration (FDA) has granted an accelerated approval to pembrolizumab for pediatric and adult solid tumor patients with MSI-H or dMMR, and MSI-H has emerged as a key predictive biomarker for immunotherapy in GCa. Therefore, it is critical to identify the mechanism underlying the MSI-H formation in GCa.

Microsatellites are short tandem repeats of DNA, which are widely distributed in the eukaryotic genome, and they are mostly located in the non-coding regions of genes or near the telomere regions of chromosomes, likely caused by defects in mismatch repair (MMR) that plays important roles in maintaining genome stability. The gain or loss of tandem repeats resulting in the alteration of microsatellite length is called microsatellite instability (MSI) (5). It is generally considered that MSI arises from the impairment of MMR machinery and is associated with tumorigenesis (6), while dMMR originates from germline mutations in the MMR genes commonly seen in the Lynch syndrome (7), but the majority of sporadic MSI result from somatic mutational inactivation or epigenetic silencing of the MMR genes (8,9). Previous studies demonstrated that more than half of MSI-positive GCa manifested hypermethylation in the promoter of MLH1, a key member of the MMR genes, while another nearly 40% of MSI-positive GCa originated from unknown genetic or epigenetic alterations (10).

As an integral part of epigenetic regulators, non-coding RNAs, including microRNAs, play irreplaceable roles in RNA degradation and post-transcriptional regulation of gene expression. It has been shown that microRNAs are aberrantly expressed in various types of malignancies, functioning either as oncogenes or tumor suppressor genes (11,12). Therefore, whether microRNAs play a role in epigenetic regulation of MSI-H formation needs further exploration.

A previous study has explored the relationship between the expression of certain microRNAs and MSI-H in colorectal cancer with a small set of 39 samples (13), but the relationship between microRNAs and the MSI status in GCa has not been fully investigated yet. Therefore, additional studies on the relationship between microRNAs and microsatellite status in GCa may help elaborate the molecular mechanism underlying MSI formation and the efficacy of immunotherapy. Such a relationship also likely provides new biological markers for immunotherapy in GCa.

Because The Cancer Genome Atlas (TCGA) database provides a large number of microRNA sequencing dataset of GCa tissue samples (14), we evaluated differential expression of microRNAs in GCa with different microsatellite status by analyzing the available high-throughput microRNA data in the TCGA database. Furthermore, we used additional data from Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes pathway enrichment (KEGG) databases to identify the pathways that may be regulated by differentially expressed microRNAs, which may provide possible molecular mechanisms underlying the MSI-H formation.

Methods

Data acquisition

The raw microRNA sequencing data and clinical information were downloaded from the FireBrowse database (http://www.firebrowse.org/). The inclusion criteria of GCa tissue samples were as follows: (I) the samples with pathologically confirmed diagnosis of GCa; (II) the samples with both microRNA sequencing data and clinical information; and (III) the samples with microsatellite status information. As a result, a total of 386 GCa samples were included in the analysis. The relationship between the microsatellite status and clinical features of the samples were assessed by the Chi square test, and P<0.05 was considered statistically significant.

Analysis of differentially expressed microRNAs in GCa tissues by microsatellite status

We processed microRNA expression data by using R language packages (version 3.5.1) and analyzed the differentially expressed microRNAs in GCa tissues with a microsatellite status, i.e., MSI-H, MSI-low (MSI-L) and microsatellite stable (MSS), by the limma package in R. We calculated the fold changes (FC) of the expression levels of individual microRNAs, and the FCs in differentially expressed microRNAs with |FC| >1 and P<0.05 were considered statistically significant. Because MSI-H and MSS GCa had the most differentially expressed microRNAs (), we thus focused on MSI-H and MSS in the subsequent analyses. To identify more significantly differentially expressed microRNAs, we calculated the expression levels of microRNAs for both MSI-H and MSS GCa with |FC| >2 and P<0.05.

Table S1

Different expressed microRNAs in MSI-H and MSS gastric adenocarcinoma with P<0.05 and |fold change| >1

ID	|Fold change|	FDR
MIMAT0000267	4.264228785	1.19E-09
MIMAT0000102	3.685301663	0.006835307
MIMAT0001080	3.556060866	3.33E-06
MIMAT0019814	2.958476637	7.36E-07
MIMAT0000264	2.629836721	9.58E-07
MIMAT0000097	2.2943782	1.57E-06
MIMAT0001536	2.260993124	5.81E-06
MIMAT0000682	2.253289919	1.75E-06
MIMAT0000064	2.17596124	8.47E-08
MIMAT0004797	2.175298156	5.12E-10
MIMAT0004550	2.146435041	1.07E-13
MIMAT0001620	2.110032652	1.95E-06
MIMAT0000318	2.099751905	6.27E-07
MIMAT0004514	2.065385362	1.99E-10
MIMAT0000728	2.055748198	0.018643679
MIMAT0000088	2.043643173	1.16E-08
MIMAT0004571	2.042890454	3.33E-06
MIMAT0000261	2.04123805	5.84E-05
MIMAT0005920	2.041077668	9.58E-07
MIMAT0003247	1.967459056	7.00E-12
MIMAT0004978	1.958067971	0.000202137
MIMAT0000416	1.9302329	0.006835307
MIMAT0000087	1.888299921	1.51E-09
MIMAT0004603	1.88693558	3.33E-06
MIMAT0000262	1.886753228	0.009523598
MIMAT0004569	1.886303163	4.28E-07
MIMAT0000226	1.863228556	0.034900381
MIMAT0000763	1.854717086	0.000168194
MIMAT0000259	1.854088832	0.000152194
MIMAT0000280	1.848938995	0.000436882
MIMAT0023712	1.83850692	2.87E-05
MIMAT0022727	1.817873973	0.000820333
MIMAT0004928	1.814319928	1.57E-06
MIMAT0004671	1.799173378	0.000918538
MIMAT0026476	1.797700431	0.017216632
MIMAT0000095	1.784378739	0.000144718
MIMAT0000432	1.76175037	0.001486309
MIMAT0004543	1.75685845	0.002155453
MIMAT0000731	1.743121407	2.82E-07
MIMAT0000646	1.736930564	3.72E-05
MIMAT0000441	1.736859396	0.011953815
MIMAT0000461	1.722576282	2.82E-07
MIMAT0004985	1.704920565	1.92E-06
MIMAT0002821	1.688341807	5.14E-08
MIMAT0000274	1.687004101	0.001231438
MIMAT0014990	1.684572501	1.95E-06
MIMAT0000098	1.676111094	0.000716002
MIMAT0003301	1.65950854	8.98E-05
MIMAT0000423	1.656296737	0.000505055
MIMAT0000460	1.651988139	0.00842623
MIMAT0000222	1.645849461	0.012964121
MIMAT0000434	1.63691306	0.007586202
MIMAT0004503	1.634758825	1.60E-09
MIMAT0004808	1.629740599	0.00046192
MIMAT0000275	1.627004618	0.000238957
MIMAT0002820	1.625678725	7.68E-07
MIMAT0004552	1.612372117	6.72E-06
MIMAT0003249	1.612048881	0.010417378
MIMAT0004584	1.609889855	0.001992171
MIMAT0000617	1.608859172	0.004202522
MIMAT0000458	1.597347527	9.68E-07
MIMAT0004701	1.587045803	0.000344391
MIMAT0001635	1.582951142	0.003614507
MIMAT0003321	1.578075864	2.14E-06
MIMAT0000732	1.57312827	0.000202137
MIMAT0000091	1.568779278	0.021909939
MIMAT0005951	1.568437217	0.001091738
MIMAT0003266	1.566948994	0.000858332
MIMAT0004598	1.556933144	0.009084325
MIMAT0000279	1.539353915	0.000913669
MIMAT0000100	1.535708318	3.72E-05
MIMAT0005593	1.535395154	0.01027584
MIMAT0019828	1.534879077	0.012964121
MIMAT0003322	1.528267532	0.000152194
MIMAT0004553	1.522662311	0.00069812
MIMAT0004494	1.520974269	0.000153516
MIMAT0003256	1.506753261	1.95E-06
MIMAT0000250	1.505492791	0.000182247
MIMAT0004657	1.50475595	0.006570749
MIMAT0019927	1.487788526	0.002742594
MIMAT0000066	1.485850578	0.000168194
MIMAT0003241	1.485388014	0.000297465
MIMAT0000440	1.462396455	0.004367022
MIMAT0004484	1.457547295	0.000211644
MIMAT0000070	1.457190008	0.020741236
MIMAT0000258	1.456145364	5.02E-06
MIMAT0004501	1.449910617	0.004769434
MIMAT0004558	1.447237294	0.000246507
MIMAT0019208	1.444843925	0.005494834
MIMAT0004491	1.442718703	0.001672409
MIMAT0003284	1.436431103	0.001878293
MIMAT0019731	1.436293111	0.014194363
MIMAT0003328	1.42749683	0.007706134
MIMAT0000425	1.421126413	0.004043419
MIMAT0000257	1.418223168	0.001649559
MIMAT0004946	1.416993761	0.045261694
MIMAT0004693	1.41325744	0.004769434
MIMAT0003294	1.411728246	0.007355318
MIMAT0004500	1.409612341	0.007706134
MIMAT0003214	1.404498847	0.00012945
MIMAT0000073	1.389784965	0.046751481
MIMAT0017992	1.388505386	0.009810861
MIMAT0000435	1.385016404	0.031244762
MIMAT0018090	1.378677265	0.011942527
MIMAT0004680	1.375526596	0.045473954
MIMAT0004567	1.373418522	0.0002005
MIMAT0004762	1.36682349	0.000324367
MIMAT0004496	1.365947461	0.011296214
MIMAT0002809	1.364319375	0.017142299
MIMAT0015020	1.362157667	0.001405775
MIMAT0019761	1.359728653	0.003082614
MIMAT0004658	1.358847641	5.44E-05
MIMAT0026738	1.355709859	0.01451574
MIMAT0004485	1.355149568	0.004570055
MIMAT0019940	1.353872523	0.016585451
MIMAT0004559	1.351495552	0.000531265
MIMAT0003298	1.344515815	0.030099922
MIMAT0004766	1.340227545	0.013711544
MIMAT0004511	1.334616141	0.007355318
MIMAT0009451	1.33018041	0.003874558
MIMAT0000227	1.329184638	0.015147391
MIMAT0004570	1.329108637	0.049617671
MIMAT0000761	1.322464512	0.04850475
MIMAT0003880	1.320928375	0.010417378
MIMAT0004489	1.319285996	0.006606604
MIMAT0004615	1.316873881	0.001146656
MIMAT0000263	1.307626663	0.047508536
MIMAT0022977	1.302419895	0.049617671
MIMAT0018968	1.300534423	0.012964121
MIMAT0000071	1.298361538	0.049435248
MIMAT0004562	1.297079857	0.002991261
MIMAT0003218	1.295222184	0.037362735
MIMAT0019957	1.295063209	5.57E-05
MIMAT0004481	1.28407505	0.015092373
MIMAT0003888	1.283997216	0.02830075
MIMAT0016847	1.278806814	0.029834138
MIMAT0000273	1.278561281	0.02097475
MIMAT0000443	1.276801539	0.016908865
MIMAT0004568	1.276610985	0.003938986
MIMAT0019926	1.274212267	0.024043734
MIMAT0027520	1.270893252	0.02599086
MIMAT0000686	1.26864232	0.023253041
MIMAT0004801	1.264646444	0.028512767
MIMAT0017993	1.259293649	0.008591579
MIMAT0026475	1.258239468	0.000108124
MIMAT0018187	1.254531779	0.004861908
MIMAT0019200	1.251418655	0.021063842
MIMAT0027587	1.248703564	0.011942527
MIMAT0000084	1.241918862	0.009981334
MIMAT0005948	1.235974877	0.005387006
MIMAT0000276	1.235392983	0.005452818
MIMAT0000082	1.233482791	0.004079498
MIMAT0004556	1.226711709	0.03676473
MIMAT0004482	1.222780366	0.024043734
MIMAT0004811	1.222701793	0.001520118
MIMAT0004560	1.221800524	0.010558205
MIMAT0004499	1.217676858	0.001986628
MIMAT0030020	1.213847919	0.014766417
MIMAT0004486	1.213506142	0.027241827
MIMAT0026765	1.212785085	0.002877419
MIMAT0003323	1.205382345	0.010834503
MIMAT0019918	1.19835748	0.009810861
MIMAT0000418	1.197773473	0.042862387
MIMAT0018936	1.18811845	0.039064377
MIMAT0019696	1.183583085	0.037227484
MIMAT0018360	1.181985012	0.045261694
MIMAT0015070	1.16721465	0.040369259
MIMAT0022710	1.15590751	0.02229369
MIMAT0027608	1.153721284	0.029834138
MIMAT0005936	1.151194014	0.028512767
MIMAT0022500	1.146345252	0.023666317
MIMAT0022483	1.142537625	0.024583611
MIMAT0019751	1.141954821	0.034214514
MIMAT0022280	1.122611782	0.013711544

To distinguish of MSI-H subtypes from MSS using microRNAs expression profiles

We also used microRNA expression data to distinguish MSI-H from MSS subtype by a stepwise logistic regression analysis, and P<0.05 was considered statistically significant. We then constructed the receiver operating characteristic (ROC) curves to illustrate prediction accuracy of the models containing each of the microRNAs, respectively. We also used ROCs from the models that included all of the microRNAs with P<0.05 using the pROC package of R.

Prediction of genes targeted by MSI-H-related microRNAs and mapping of the target signaling pathway genes

We divided the MSI-H-related microRNAs into two groups of either upregulated or downregulated expression levels and compared their MSS. We selected the top six upregulated and three downregulated microRNAs for the two groups, respectively, by using more stringent criteria (P<0.01 and |FC| >2.175). The genes targeted by upregulated and downregulated microRNAs in GCa with MSI-H were predicted, respectively, according to miRanda (http://www.microrna.org/microrna/home.do) online analytic tools, and the putative genes with a short variable region (SVR) score less than −0.5 were included for further analysis. We further explored the signaling pathways and processes of the predicted genes by using the Annotation, Visualization and Integrated Discovery (DAVID) database (v6.8, https://david.ncifcrf.gov/summary.jsp). Finally, we performed GO and KEGG pathway enrichment analyses for the target genes with P<0.05 and gene counts ≥3 sets as the cut-off criteria for the comparisons.

Statistical analysis

The expression levels of microRNAs in GCa tissues were analyzed and compared by the unpaired t-test. The statistical analyses were performed by using the IBM SPSS statistics software program version 20.0 (IBM Corp., NY, USA) and R language (version 3.5.1). P values were two-sided with a significance level of 0.05.

Results

Different clinicopathological traits of GCa with different MSI status

In the present study, we included the data for 386 GCa samples from the TCGA database, and the number of the samples with MSS, MSI-L and MSI-H was 263, 56 and 67, respectively. Their general clinical traits are presented in . The associations between the MSI status and detailed clinical traits, including age at diagnosis, sex, family history, helicobacter pylori infection, gastric subregion, histologic type, histologic grade and TNM pathological stage are presented in . We found that the MSI status was significantly associated with age at diagnosis (P=0.049), sex (P=0.014) and gastric subregion (P=0.007). Overall, the proportion of MSI-H positive tumors increased as age increased, while the proportion of MSS tumors decreased as age increased, but no obvious trend was seen for MSI-L tumors. Specifically, 33 of 129 (25.6%) patients with age >70 years had MSI-H positive tumors, 23 of 128 (18.0%) patients with age of 61–70 years had MSI-H positive tumors, and 11 of 129 (8.5%) patients with age ≤61 years had MSI-H positive tumors; female GCa patients (25.2%) were more likely to develop MSI-H tumors than male GCa patients (13.3%); and the MSI-H was more likely found in distal (37.1%) and body (35.5%) GCa than in proximal (13.7%) and junction (11.4%) GCa (). No differences were observed for other patients’ traits ().

Table 1

Clinicopathological characteristics of gastric adenocarcinoma cases in the TCGA database

Traits	No. of cases (%)
All subjects	386 (100.0)
Age at diagnosis
<50	32(8.3)
51–60	97 (25.1)
61–70	128 (33.2)
71–80	103 (26.7)
>80	22 (5.7)
NA	4 (1.0)
Sex
Female	131 (33.9)
Male	255 (66.1)
Microsatellite status
MSS	263 (68.1)
MSI-L	56 (14.5)
MSI-H	67 (17.4)
Gastric subregion
Antrum/distal	143 (37.1)
Cardia/proximal	53 (13.7)
Fundus/body	137 (35.5)
Gastroesophageal junction	44 (11.4)
NA	9 (2.3)
Family history
Yes	18 (4.6)
No	315 (81.6)
NA	53 (13.8)
HP infection
Yes	19 (5.1)
No	162 (41.7)
NA	205 (53.2)
Stage
Stage I	50 (13.0)
Stage II	123 (31.9)
Stage III	174 (45.1)
Stage IV	31 (8.0)
NA	8 (2.1)

TCGA, The Cancer Genome Atlas; MSS, microsatellite stable; MSI-L, microsatellite instability low; MSI-H, microsatellite instability high; NA, not available; HP, Helicobacter pylori.

Table 2

Differences in the frequencies of MSI status by clinicopathological features in gastric adenocarcinoma cases in TCGA database

Variables	MSI-H (n=67)	MSI-L (n=56)	MSS (n=263)	P (group)	P (subgroup)
Mean of age ± SD (years)	69.08±9.54	64.51±10.83	64.23±10.71	0.004*
Neoplasm subdivision (%)				0.002*
Antrum/distal	37 (56.9)	20 (37.0)	86 (33.3)		1.000
Cardia/proximal	3 (4.6)	10 (18.5)	40 (15.5)		0.007#
Fundus/body	24 (36.9)	18 (33.3)	95 (36.8)		0.202
Gastroesophageal junction	1 (1.5)	6 (11.1)	37 (14.3)		0.002#
Sex
Male	34 (50.7)	39 (69.6)	182 (69.2)	0.014*
Female	33 (25.2)	17 (13.0)	81 (61.8)
Histological type (%)				0.965
STAD, signet ring type	2 (3.0)	0 (0.0)	9 (3.4)		1.000
STAD, diffuse type	11 (16.4)	9 (16.4)	47 (17.9)		0.512
STAD, NOS	21 (31.3)	22 (40.0)	89 (33.8)		0.399
SIAD, mucinous type	4 (6.0)	2 (3.6)	15 (5.7)		0.829
SIAD, NOS	14 (20.9)	12 (21.8)	44 (16.7)		0.396
SIAD, papillary type	2 (3.0)	1 (1.8)	5 (1.9)		0.580
SIAD, tubular type	13 (19.4)	9 (16.4)	54 (20.5)		0.755
Neoplasm histologic grade (%)				0.717
G1	2 (3.0)	1 (1.8)	4 (1.6)		1.000
G2	20 (30.3)	20 (35.7)	99 (38.8)		0.579
G3	44 (66.7)	35 (62.5)	152 (59.6)		0.848
Pathologic T stage (%)				0.168
T1	6 (9.0)	3 (5.4)	12 (4.6)		1.000
T2	13 (19.4)	15 (26.8)	49 (18.6)		0.503
T3	23 (34.3)	25 (44.6)	132 (50.2)		0.164
T4	25 (37.3)	13 (23.2)	70 (26.6)		0.724
Pathologic N stage (%)				0.030*
N0	30 (45.5)	19 (33.9)	70 (27.2)		1.000
N1	17 (25.8)	14 (25.0)	69 (26.8)		0.255
N2	8 (12.1)	16 (28.6)	54 (21.0)		0.033#
N3	11 (16.7)	7 (12.5)	64 (24.9)		0.017#
Pathologic M stage (%)				0.142
M1	1 (1.5)	2 (3.8)	19 (7.6)		–
M0	64(98.4)	51(96.2)	231(92.4)		–
Pathologic TNM stage (%)				0.153
Stage I	14 (20.9)	8 (14.8)	28 (10.9)		1.000
Stage II	25 (37.3)	20 (37.0)	78 (30.4)		0.535
Stage III	25 (37.3)	23 (42.6)	126 (49.0)		0.053
Stage IV	3 (4.5)	3 (5.6)	25 (9.7)		0.064

*, P value was less than 0.05; #, the subgroup contributed the difference within the groups. TCGA, The Cancer Genome Atlas; GA, gastric adenocarcinoma; MSS, microsatellite stable; MSI-L, microsatellite instability low; MSI-H, microsatellite instability high; NA, not available; SD, standard deviation; GE, gastroesophageal; STAD, stomach adenocarcinoma; NOS, not other specified; SIAD, stomach intestinal adenocarcinoma.

Figure 1

The proportion of GCa patients with different MSI status grouped by (A) age, (B) sex, and (C) gastric subregion.

MicroRNA expression profiles by MSI status

To explore the differences in the frequencies of MSI-H, MSI-L and MSS in the microRNA expression profiles, all the differentially expressed microRNAs (defined as P<0.05 with |FC| >1) among these three groups were assessed and compared with each other (). We found that MSI-L and MSS tumors had similar microRNA expression profiles, but MSI-H tumors had the most different expression profiles in comparison with MSS ().

Table S2

Different expressed microRNAs in MSI-H and MSI-L gastric adenocarcinoma with P<0.05 and |fold change| >1

ID	|Fold change|	FDR
MIMAT0000267	2.468875448	0.010219
MIMAT0019814	2.300115342	2.11E-02
MIMAT0000763	2.112081476	0.009511
MIMAT0000682	2.111826398	0.000405
MIMAT0000318	1.910117734	0.000405
MIMAT0001536	1.880922659	0.013484
MIMAT0005920	1.823869638	0.010219
MIMAT0001620	1.790420257	0.002833
MIMAT0003247	1.713692493	0.009511
MIMAT0000088	1.697596859	0.034798
MIMAT0004558	1.68404046	4.05E-04
MIMAT0004514	1.666236563	0.012608
MIMAT0004571	1.658890595	0.023278
MIMAT0004701	1.64181437	0.013933
MIMAT0003328	1.614420354	0.040287
MIMAT0000646	1.583097336	0.029688
MIMAT0000257	1.571716635	0.009511
MIMAT0004550	1.556305761	0.024059
MIMAT0002809	1.555288061	0.009511
MIMAT0000458	1.549880272	1.26E-02
MIMAT0000100	1.546333728	0.01603
MIMAT0014990	1.533451709	0.021695
MIMAT0002821	1.522236606	0.007251
MIMAT0000731	1.507560293	0.040287
MIMAT0002820	1.498727252	0.034798
MIMAT0000066	1.471321879	0.049187
MIMAT0003321	1.469350506	0.034169
MIMAT0004559	1.424003999	0.009511
MIMAT0004503	1.401724895	3.99E-02
MIMAT0017993	1.385063909	0.007251
MIMAT0005948	1.32884243	0.044563

Table S3

Different expressed microRNAs in MSI-L and MSS gastric adenocarcinoma with P<0.05 and |fold change| >1

ID	|Fold change|	FDR
MIMAT0002830	17.76279758	0.026251
MIMAT0027459	2.815585575	1.32E-05
MIMAT0014998	1.821807692	0.026251
MIMAT0015050	1.635801689	0.0118073
MIMAT0019958	1.423560534	0.0421419
MIMAT0000084	1.302644305	0.0303131
MIMAT0000078	1.292059736	0.0118073

Figure 2

The number of commonly expressed microRNAs and differentially expressed microRNAs between MSI-H vs. MSI-L, MSI-H vs. MSS, and MSI-L vs. MSS. “Total miRNAs” means all the microRNAs investigated in the present study.

To analyze the association between microRNA expression and MSI, we further analyzed the difference in microRNA expression between the MSI-H and MSS groups. By using a more stringent criterion (P<0.05 and |FC| >2), we found that a total of 19 differentially expressed microRNAs were identified between MSI-H and MSS samples, of which 14 were upregulated and five were downregulated in MSI-H samples, compared with those in MSS samples (). The Volcano plot is presented to show microRNA expression levels with P<0.05 and |FC| >2 ().

Table 3

Differentially expressed microRNAs between MSI-H and MSS gastric adenocarcinoma in the TCGA database

ID	Accession number	Fold Change	FDR
miR-210-3p	MIMAT0000267	4.264228785	1.19E-09
miR-196b-5p	MIMAT0001080	3.556060866	3.33E-06
miR-203b-3p	MIMAT0019814	2.958476597	7.36E-07
miR-203a-3p	MIMAT0000264	2.629836721	9.58E-07
miR-429	MIMAT0001536	2.260993124	5.81E-06
miR-200a-3p	MIMAT0000682	2.253289919	1.75E-06
miR-582-3p	MIMAT0004797	2.175298156	5.12E-10
miR-200a-5p	MIMAT0001620	2.110032652	1.95E-06
miR-200b-3p	MIMAT0000318	2.099751905	6.27E-07
miR-29b-1-5p	MIMAT0004514	2.065385362	1.99E-10
miR-375-3p	MIMAT0000728	2.055748198	0.01864
miR-200b-5p	MIMAT0004571	2.042890454	3.33E-06
miR-183-5p	MIMAT0000261	2.04123805	5.84E-05
miR-1266-5p	MIMAT0005920	2.041077668	9.58E-07
miR-30a-3p	MIMAT0000088	-2.04364317	1.16E-08
miR-30c-2-3p	MIMAT0004550	-2.14643504	1.07E-13
has-let-7c-5p	MIMAT0000064	-2.17596124	8.47E-08
miR-99a-5p	MIMAT0000097	-2.2943782	1.57E-06
miR-105-5p	MIMAT0000102	-3.68530166	0.00684

TCGA, The Cancer Genome Atlas; GA, gastric adenocarcinoma; MSS, microsatellite stable; MSI-L, microsatellite instability low; MSI-H, microsatellite instability high; FDR, false discovery rate.

Figure 3

Volcano plot of differentially expressed microRNAs between MSI-H and MSS GCa samples. The red dots represent upregulated microRNAs with a P value <0.05 and |FC| >2, and the blue dots represent downregulated microRNAs with a P value <0.05 and |FC| <2.

MicroRNAs that predicted the MSI-H status

By the microRNA expression profiles from the TCGA database, we found that four microRNAs (miR-210-3p, miR-582-3p, miR-30a-3p and miR-105-5p) could accurately distinguish the MSI-H tumors from the MSS tumors (P=4.93×10−10, 5.63×10−10, 3.23×10−9 and 7.64×10−4, respectively). To further validate the accuracy of the prediction models, ROCs of the miR-210-3p, miR-582-3p and miR-30a-3p were constructed, and the area under the curve (AUC) was 0.784, 0.757 and 0.738 for these three microRNAs, respectively, and the increase in these AUCs was statistically significant (P<0.01 for all), while the ROC of miR-105-5p could not be performed due to the missing expression data of some samples. When the three microRNAs were combined, the AUC of the combined prediction model increased to 0.886 (P=0.0004), indicating that the MSI-H subtype could be accurately distinguished from the MSS subtype by this combined prediction model ().

Figure 4

ROC curves showed that three microRNAs could accurately distinguish the MSI-H status from MSS alone or combined together. A: miR-210-3p; B: miR-30a-5p; C: miR-582-3p; D: miR-210-3p, miR-582-3p and miR-30a-5p combined together.

Biological signaling pathway enrichment for MSI-H related microRNAs

According to the cut-off criteria (P<0.01 and |FC| >2.175), we considered the top six microRNAs of the 14 upregulated microRNAs and the top three microRNAs of the five downregulated microRNAs as the MSI-H-related microRNAs. By using the miRanda online analysis tools, we identified a total of 171 genes of upregulated microRNAs and 119 genes of downregulated microRNAs. Then, we performed an enrichment analysis to elucidate biological functions of these target genes. We found that the GO biological process (BP) terms were mainly enriched in the regulation of transcription (DNA templated); positive regulation of transcription (DNA templated); positive regulation of transcription from RNA polymerase II promoter, and negative regulation of transcription from polymerase II promoter (). In addition, the KEGG pathways were significantly enriched in those for transcription mis-regulation in cancer ().

Figure 5

The significantly enriched GO biological processes and KEGG pathways of putative genes targeted by the selected microRNAs. (A) GO biological processes; (B) KEGG pathways.

Discussion

Current choice of therapies for the advanced GCa are limited, and the prognosis is still relatively poor. For GCa patients with MSI-H or dMMR, however, recent therapeutic regimes of using PD-1/PD-L1 inhibitors alone or a combination with chemotherapy have achieved a remarkable progress (15-17). Based on the findings from the present study, 17.1% of the GCa patients had MSI-H tumors (18), which means nearly 1/6 of the GCa patients may benefit from the PD-1/PD-L1 mono-antibody therapy.

Only a small proportion of MSI-H GCa arises from germline mutations of the MMR genes (19). It is known that microRNAs play important roles in epigenetic regulation and that among the sporadic GCa, MSI-H is associated with epigenetic regulation, but the mechanism of MSI-H formation remains ambiguous (10,20,21). Previous studies have revealed that some microRNAs had a consistent expression pattern in both tumor tissues and circulatory plasma, serving as important predictive biomarkers for various types of malignant tumors (22,23). Therefore, the present study focused on the relationship between microRNA expression profiles and the MSI status in GCa, aiming at revealing the mechanism underlying the MSI-H formation.

Firstly, we found that the MSI-H status in 386 GCa patients was correlated with some clinicopathological features, e.g., the MSI-H status increased as age increased, with a higher frequency in female patients and patients with distal GCa located in the pylorus or body of stomach. These findings are consistent with those described in a review of other previously published results from fewer tumor samples (24).

Secondly, the present study also suggests that the microRNA expression profiles of MSS, MSI-L and MSI-H showed a trend change in GCa tumor samples. Although the difference between MSS and MSI-L was rather small, the difference between MSI-H and MSI-L was relatively remarkable and associated with aging. These trends indicate that it is a continuous change from MSS to MSI-H, consistent with the dividing method of MSI in colon cancer (25). Furthermore, we found that both MSI-H and MSS were significantly associated with microRNA expression levels.

MiR-210, which ranks the top of the most significantly differentially expressed microRNAs, has been reported to impair the functions of DNA damage-repair genes, possibly causing DNA replication errors (26,27), which may lead to the MSI formation (28). As for miR-196b, there are a few reports on the role of miR-196b in GCa. For example, a couple of studies have suggested that miR-196b promotes the metastasis and invasion of GCa cells (29,30). Other studies had shown that the high expression of miR-196b significantly impaired DNA damage-repair functions (31). Hence, we speculate that the high expression levels of miR-196b in the MSI-H-related GCa may affect the stability of the genome through the impairment of DNA damage-repair functions.

Studies have revealed that miR-203 also inhibits invasion and metastasis of GCa cells. For example, one study found that the expression of miR-203 was negatively correlated with expression of ataxia-telangiectasia mutated (ATM) protein (32), while another study demonstrated that the ATM gene was highly mutated and that the expression of the ATM protein was downregulated in MSI-H-related GCa tissues (33). Since ATM plays a critical role in DNA damage-induced signaling and initiation of cell cycle checkpoint signaling, it is reasonable to assume that miR-203 may contribute to MSI-H by targeting the ATM gene.

miR-429 and miR-200a, as the members of the miR-200 family, were significantly upregulated in MSI-H GCa tissues than in the MSS subtype. One study demonstrated that expression levels of the miR-200 family increased substantially in GCa tumor tissues, compared with that of normal tissues, indicating that the miR-200 family may play an important role in promoting GCa cell growth (34).

The miR-105, miR-99a and hsa-let-7c were the three microRNAs downregulated the most in MSI-H GCa, compared with the MSS subtype. Few studies reported the roles of miR-105 and has-let-7c in GCa. One study reported, however, that the miR-99 family of microRNAs could regulate DNA damage response by targeting SNF2H (35), while other studies showed that overexpression of the miR-99 family in prostatic cancer cells could inhibit the expression of SNF2H and reduce DNA damage-repair rate and overall repair efficiency (36,37), although the role of miR-99 in GCa has not been reported yet.

To further explore the functions of the above-mentioned nine microRNAs, we searched for the predicted target genes of these microRNAs and analyzed their related pathways and GO annotations by using bioinformatics online tools. We found that these nine microRNAs could regulate a variety of genes in several key signaling pathways, including regulation of transcription (DNA templated), positive regulation of transcription from RNA polymerase II promoter, positive regulation of transcription (DNA templated) and negative regulation of transcription from polymerase II promoter. It has been suggested that abnormal signaling pathways, such as the KRAS signaling pathway and the base-excision repair pathway, may contribute to the formation of MSI-H in gastrointestinal and endometrial cancers (38-40). Therefore, we assume that other DNA damage repair pathways may also play important roles in the formation of MSI-H, in addition to the impairment of the MMR pathway; however, further investigations are needed to test this hypothesis and unravel the underlying molecular mechanisms.

Conclusions

In the present study, we identified nine significantly differentially expressed microRNAs in GCa tumor tissues, and the results suggested that the pathways related to DNA damage-repair functions, other than MMR, were associated with MSI formation in GCa. Because of limited sample size and the limitations in bioinformatics analysis, further rigorous laboratory experiments in molecular and functional investigations are needed to substantiate these results.

The article’s supplementary files as