Predictive value of peripheral blood leukocytes-based methylation of Long non-coding RNA MALAT1 and H19 in the chemotherapy effect and prognosis of gastric cancer

Highlights

• •Methylation levels of MALAT1 and H19 were associated with the response to chemotherapy and prognosis of gastric cancer (GC).
• •Methylation levels of MALAT1 and H19 in peripheral blood leukocytes could be used to predict the chemotherapy effect and prognosis of GC.
• •MALAT1 expression was correlated to the microsatellite instability and tumor mutational burden.
• •MALAT1 and H19 expression were related to multiple immune checkpoint and immune pathways.

Abstract

Background

The predictive value of the methylation of Long non-coding RNA (lncRNA) metastasis-associated lung adenocarcinoma transcript 1 (MALAT1) and H19 promoters in peripheral blood leukocytes as a non-invasive biomarker for the chemotherapy effect and prognosis gastric cancer (GC) is unclear.

Methods

The DNA methylation of H19 and MALAT1 between chemotherapy-sensitive and non-sensitive groups and between groups with better and worse survival of GC was compared using regression analyses. Several predictive nomograms were constructed. The genetic alteration of MALAT1 and H19 and the association between gene expression and immune status in GC were also investigated using bioinformatics analysis.

Results

Higher genetic methylations in peripheral blood were noticed in GC groups with poorer survival. The constructed nomograms presented strong predictive values for the chemotherapy effect and 3-year survival of disease-free survival, progression-free survival, and overall survival, with the area under the curve as 0.838, 0.838, 0.912, and 0.925, respectively. Significant correlations between MALAT1 or H19 expression and marker genes of immune checkpoints and immune pathways were noticed. The high infiltration of macrophages in H19-high and low infiltration of CD8+ T cells in MALAT1-high groups were associated with worse survival of GC.

Conclusions

MALAT1 and H19 have the potential to predict the chemotherapy response and clinical outcomes of GC.

Graphical abstract

Introduction

As the fifth type of malignancy that is most seen globally, gastric cancer (GC) is also the third leading cause of cancer-related death worldwide, accounting for more than 1 million incident cases and more than 700,000 deaths each year [1]. Although there has been great progress in the diagnostic and treatment strategies, the specific pathogenesis of GC remains to be illuminated [2]. Since the clinical symptoms and signs of GC are not obvious and the unwillingness to take invasive gastroscopy examinations, early diagnosis, and treatment are facing significant challenges [3]. Therefore, over three-quarters of GC patients are diagnosed in the advanced stage with a 5-year survival rate of only 5 %−10 % [4]. Moreover, in the process of chemotherapy for GC, the development of drug resistance often leads to treatment failure, and drug resistance is considered one of the major obstacles that need to be handled [5]. Therefore, it is important to develop diagnostic and predictive methods that are less invasive to improve the early diagnosis and prognosis of GC.

Epigenetic abnormalities are heritable, disrupting the epigenetic landscapes associated with genetic changes usually occur in cancer [6]. DNA methylation is an important form of epigenetic modification and is found crucially participate in the carcinogenesis of many tumors, including GC [7]. Numerous recent studies have reported that the expression of Long non-coding RNAs (lncRNAs) is associated with the development of tumorigenesis and the treatment effect of GC [8,9]. Moreover, it has been found that lncRNAs play an important role in regulating tumorigenesis and the progress of many cancers by participating in the process of gene transcription mediated by DNA methylation [10]. As one of the most conserved lncRNAs, the regulation of metastasis-associated lung adenocarcinoma transcript 1(MALAT1) by DNA methylation has been found in many human tumors, including GC, prostate, and breast [11], [12], [13]. The lncRNA H19 is also one of the important lncRNAs located on chromosome 11p15.5, and its pro-tumorigenic and riboregulatory roles have been widely reported in multiple malignancies [14,15]. Different from the oncogenic role of MALAT1, H19 presented with both oncogenic and suppressor properties in diverse cancers and participates in nearly all stages of cancer progression, including migration, invasion, proliferation, and metastasis [16]. Numerous studies have found that the role of H19 in cancer formation and progression may be associated with its abnormal methylation [17,18]. In GC, during the process of cancer occurrence and metastasis, the overexpression of lncRNA H19 was reported, and such aberrant expression also led to poor clinical outcomes in GC [19]. Based on the evidence presented above, we hypothesize that the DNA methylations of these genes may play an important role in the occurrence and development of GC. Recently, multiple tissue-based studies have investigated the role of lncRNA MALAT1 and H19 in the carcinogenesis and prognosis of GC. However, whether the peripheral blood leukocyte-based methylation levels of these genes could act as potential biomarkers that are less invasive for predicting the carcinogenesis and development of GC is still unclear.

In our previous research, we discussed the potential value of MALAT1 and H19 from the peripheral blood leukocytes in the early diagnosis of GC [12]. We aimed to identify the predictive value of MALAT1 and H19 methylation in peripheral blood to predict the effect of chemotherapy and the prognosis of GC using a prospective follow-up study.

Materials and methods

Study participants

One hundred and fifty patients who were diagnosed by histology and pathology as GC from the First Affiliated Hospital of Anhui Medical University were included (median age: 62.79 years, range: 26.42–82.08, 114 males, 36 females, patients undergoing curative surgery (CS): 51, patients in locally advanced or metastatic stage (LAMS): 99). The detailed inclusion and exclusion criteria, the data and sample collection methods, as well as the DNA methylation detection technology have been described in detail in our previous study [12,20]. The examined methylation levels of each CpG sites of H19 and MALAT1 genes among GC and health controls were presented as heatmap in Figure S1-S5.

DNA methylation data extraction based on data from the MethSurv database

Using the online tool of the MethSurv database [21], we explored the association between genes’ DNA methylation levels in each CpG site in tissues and the prognosis of GC. The hazard ratio (HR) and confidence interval (CI) were provided.

Assessment of chemotherapy effect

To examine the value of these genes in predicting the response of platinum and fluoropyrimidine on GC, we first used the largest publicly available pharmacogenomics database, the Genomics of Drug Sensitivity in Cancer (GDSC) (https://www.cancerrxgene.org/) and R package of “pRRophetic” to explore the difference of 5-Fluorouracil and Cisplatin response between expression-high and expression-low subgroups of H19 and MALAT1, based on the expression levels. Next, we aimed to investigate the value of genes’ DNA methylation in predicting the chemotherapy effect of GC, 50 of 97 GC patients with stage IV who received platinum plus fluorouracil (PPF) treatment at the first line were selected, and we used the Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST1.1) to examine their chemotherapy effects [22]. Treatment time was set as the starting point to observe the efficacy of chemotherapy, and the imaging studies were evaluated every two cycles. The chemotherapy response was categorized as progressive disease (PD), stable disease (SD), partial response (PR), and complete response (CR) according to RECIST1.1.

Assessment of prognosis of GC patients

The total enrolled population was separated into two groups: 51 GC patients at I-III stages and undergoing CS and 99 patients in LAMS (including 97 GC patients at stage IV and 2 GC patients who did not receive CS and were at stage III). The primary endpoint for the former groups was disease-free survival (DFS) and overall survival (OS), while progression-free survival (PFS) and OS were used to examine the latter groups. The definition of the survival outcomes was the following: DFS was defined from the date of CS to the date of the last follow-up (October 21, 2021) or recurrence, and PFS was defined as the time interval between the starting point of treatment and the date of disease progression or death of the disease, and OS was calculated from the date of first treatment or CS to the time of death of any cause.

Development of the nomograms in predicting the chemotherapy effect and survival of GC

Patients (50) treated with PPF were categorized into PD groups and nPD groups (including CR, PR, and SD) to evaluate the potential value of MALAT1 and H19 methylation in predicting the chemotherapy efficacy of GC patients. We then performed logistic binary and multinomial regression analyzes (adjusted for age, sex, smoke, and drink) to explore the associations of the genes’ methylation levels between nPD and PD groups. Similarly, to investigate the relationship between genes’ methylation and the prognosis of GC patients, univariate and multivariable (adjust for age, sex, smoke, and drink) Cox regression analyses were applied, including DFS, PFS, and OS. According to the logistics regression analysis, we constructed a nomogram [23] to predict the chemotherapy effect of GC patients who received PPF. Concordance index (C-index), Hosmer–Leeshawn test (P > 0.05), and calibration plots were used to test the nomograms’ calibration ability [24], while decision curve analysis (DCA) [25] and receiver operating curve (ROC) [26] were applied to examine the nomograms’ clinical values. Based on the Cox regression analysis, we combined the CpG sites with a higher area under the curve (AUC) (statistically significant) and relevant clinicopathological parameters, such as TNM stages, age, sex, smoke, and drink to construct the predictive nomograms for predicting the DFS, PFS, and OS of GC. Calibration plots, DCA curves, and ROC plots were generated, and the cut-off values, sensitivity (Sen), specificity (Spe), positive predictive value (PPV), and negative predictive value (NPV) were also calculated. Moreover, based on the Cox regression results, the risk score was obtained, and the patients were separated into the low and high-risk groups according to the median risk score. Notably, to avoid overfitting the selected variables, we also performed a least absolute shrinkage and selection operator (Lasso) regression model in OS [27]. In this study, significant threshold of P-value was set as less than 0.05.

Interaction analysis among CPG sites and between CPG sites and environmental factors on the chemotherapy effect and prognosis of GC

To explore the interaction between CpG sites’ methylation and environmental factors in the chemotherapy efficacy and prognosis of GC, the above-identified CpG sites were classified as hypermethylation(> median) and hypomethylation (≤ median) in terms of additive and multiplicative interaction analysis. The presence of additive interaction between two factors can be inferred when the 95 %CI for the relative excess risk (RERI) and attributable proportion (AP) on the additive scale do not include 0, and the 95 % CI for the synergy index (S) excludes 1 [28]. Therefore, in such cases, it is feasible to establish the association between the two factors and conduct further analysis to explore their mode of interaction and their impact on the outcome.

Alteration and immune analysis of MALAT1 and H19

We also utilized the cBioPortal database to investigate the associations between genetic alterations in MALAT1 or H19 and disease-specific survival (DSS) and OS in GC [29]. We then used the TIMER and CIBERSORT algorithms to evaluate the association between gene expression and infiltration score of different immune cells in the TIMER (http://timer.cistrome.org/) database [30]. We evaluated the association between MALAT1/H19 expression and the level of 8 typical immune-checkpoint related transcripts, including CD274, PDCD1LG2, TIGIT, CTLA4, HAVCR2, LAG3, PDCD1, and SIGLEC15. To further evaluate the role of MALAT1/H19 in predicting the immunotherapy response in GC, the Tumor Immune Dysfunction and Exclusion (TIDE) algorithm was used [31]. In addition, the clinical outcome of GC and immune infiltration among expression-high and expression-low groups were explored using the TIMER database.

Functional enrichment analysis

To explore the potential mechanism of lncRNA MALAT1 and H19 in the carcinogenesis of GC, we performed a Gene set enrichment analysis (GSEA) using the GSEA software (version 3.0). Based on the median value of MALAT1 and H19, the GC patients in TCGA cohort were separated into expression-high and low subgroups, respectively. Then, the different KEGG and HALLMARK functions between expression-high and low groups were examined.

Results

Effect of genes’ methylation level in tissues on the prognosis of GC

Utilizing the MethSurv tool, we then investigated the prognostic values of each CpG site of gene methylation in tissues in the clinical outcomes of GC, and five hypermethylated CpG sites of H19 were found to be associated with better OS of GC, while two hypermethylated CpG sites (H19-Body-N_Shore-cg06197492, MALAT1-TSS200-Island-cg12498916) was related to worse OS of GC (Table 1, Figure S6).

Table 1.

Effect of H19 and MALAT1 methylation level in tissues on prognosis of GC.

CpG	HR	CI	P-value
H19-TSS1500-S_Shore-cg00237904	1.334	(0.937;1.897)	0.110
H19-TSS1500-Island-cg01539474	0.741	(0.518;1.059)	0.100
H19-TSS1500-S_Shore-cg01895612	1.129	(0.819;1.557)	0.459
H19-TSS1500-Island-cg01977486	0.651	(0.46;0.922)	0.016
H19-TSS1500-S_Shore-cg02657360	1.093	(0.759;1.573)	0.632
H19-TSS1500-Island-cg02694715	1.366	(0.96;1.944)	0.083
H19-TSS1500-S_Shore-cg02886509	0.819	(0.573;1.171)	0.274
H19-TSS1500-S_Shore-cg03996735	0.759	(0.528;1.091)	0.136
H19-TSS1500-Island-cg04088212	1.382	(0.969;1.969)	0.074
H19-Body-N_Shore-cg04647234	0.696	(0.49;0.987)	0.042
H19-TSS1500-S_Shore-cg04817190	0.777	(0.543;1.114)	0.170
H19-TSS1500-Island-cg04975775	1.069	(0.774;1.477)	0.685
H19-Body-N_Shore-cg06197492	1.394	(1.007;1.931)	0.045
H19-TSS1500-S_Shore-cg06749854	0.839	(0.584;1.205)	0.342
H19-TSS1500-S_Shore-cg06765785	1.182	(0.814;1.716)	0.378
H19-TSS1500-Island-cg10154633	0.723	(0.506;1.033)	0.075
H19-Body-N_Shore-cg10602543	0.535	(0.381;0.749)	<0.001
H19-Body-N_Shore-cg11492040	1.212	(0.833;1.764)	0.315
H19-Body-N_Shore-cg11716026	0.583	(0.379;0.895)	0.014
H19-TSS1500-Island-cg11735853	0.831	(0.578;1.195)	0.319
H19-TSS1500-Island-cg13581483	0.833	(0.583;1.191)	0.316
H19-Body-N_Shore-cg15394860	0.868	(0.629;1.196)	0.386
H19-TSS1500-Island-cg15886040	1.223	(0.885;1.692)	0.223
H19-Body-Island-cg15963714	0.663	(0.479;0.916)	0.013
H19-TSS1500-Island-cg16303279	0.733	(0.514;1.047)	0.088
H19-TSS1500-S_Shore-cg16675558	0.785	(0.548;1.126)	0.188
H19-TSS1500-Island-cg17769238	0.762	(0.536;1.083)	0.130
H19-TSS1500-S_Shore-cg18104242	0.84	(0.584;1.207)	0.345
H19-TSS1500-S_Shore-cg18362496	0.711	(0.501;1.009)	0.056
H19-TSS1500-S_Shore-cg18454954	0.76	(0.514;1.124)	0.169
H19-Body-N_Shore-cg22172494	0.782	(0.567;1.079)	0.134
H19-TSS1500-S_Shore-cg23476401	0.859	(0.599;1.23)	0.406
H19-Body-N_Shore-cg23977670	0.721	(0.508;1.023)	0.067
H19-TSS1500-Island-cg24409677	1.323	(0.922;1.896)	0.128
H19-TSS1500-S_Shore-cg24510613	0.783	(0.546;1.122)	0.183
H19-TSS1500-S_Shore-cg24605090	0.779	(0.545;1.113)	0.170
H19-TSS1500-S_Shore-cg25281616	0.84	(0.587;1.203)	0.342
H19-TSS1500-S_Shore-cg25579157	0.744	(0.523;1.058)	0.099
H19-TSS1500-S_Shore-cg25821896	0.792	(0.555;1.128)	0.196
H19-TSS1500-Island-cg25838645	0.767	(0.538;1.094)	0.143
H19-Body-Island-cg25852472	0.916	(0.627;1.339)	0.652
H19-TSS1500-Island-cg26469586	0.703	(0.493;1.002)	0.051
H19-Body-Island-cg26808784	0.701	(0.494;0.995)	0.047
H19-Body-Island-cg26857192	0.869	(0.601;1.256)	0.455
H19-TSS1500-S_Shore-cg27300742	0.73	(0.512;1.042)	0.083
MALAT1-Body-Island-cg01185801	1.356	(0.963;1.91)	0.081
MALAT1-TSS200-Island-cg02943285	1.185	(0.808;1.738)	0.385
MALAT1-Body-S_Shelf-cg04073608	1.215	(0.88;1.677)	0.237
MALAT1-TSS200-Island-cg04868132	1.114	(0.808;1.535)	0.511
MALAT1-Body-S_Shore-cg04977124	0.866	(0.59;1.27)	0.461
MALAT1-Body-S_Shore-cg05491695	1.256	(0.908;1.738)	0.168
MALAT1-Body-S_Shore-cg07799005	1.163	(0.786;1.72)	0.450
MALAT1-TSS1500-N_Shore-cg10631284	1.429	(0.936;2.183)	0.098
MALAT1-TSS200-Island-cg12498916	1.359	(0.984;1.877)	0.063
MALAT1-Body-S_Shelf-cg14690315	0.869	(0.614;1.23)	0.429
MALAT1-TSS200-Island-cg15574972	1.311	(0.922;1.865)	0.131
MALAT1-Body-S_Shore-cg17153055	1.256	(0.827;1.907)	0.286
MALAT1-Body-S_Shore-cg18501142	1.297	(0.912;1.845)	0.147
MALAT1-TSS1500-N_Shore-cg19878733	1.286	(0.931;1.778)	0.127
MALAT1-TSS200-Island-cg20503416	0.805	(0.54;1.201)	0.289
MALAT1-Body-S_Shore-cg22311458	0.744	(0.52;1.064)	0.105
MALAT1-TSS1500-N_Shore-cg23566411	0.812	(0.554;1.19)	0.286
MALAT1-Body-S_Shelf-cg26489875	1.273	(0.898;1.805)	0.176

GC: Gastric Cancer; HR: Hazard ratio; CI: confidence interval.

Values of methylation levels in predicting the chemotherapy response of GC

Using the online database, we first investigated the relationship between the expression of H19 and MLAT1 and the IC50 of 5-Fluorouracil and Cisplatin. As shown in Fig. 1A-D, we found that GC patients in H19 and MALAT1-low group have a relatively lower 5-Fluorouracil IC50, and patients with higher MALAT1 expression also have a higher Cisplatin IC-50, suggesting that GC patients with a relatively lower expression on H19 and MALAT1 are more sensitive to the treatment of 5-Fluorouracil and Cisplatin. We hypothesize that the DNA methylation level of these two LncRNAs could also predict the chemotherapy effect of platinum and fluoropyrimidine in GC. Among 50 GC patients with stage IV and undergoing CS, 48 % (n = 24) patients experienced disease progression or death. The comparisons of the two genes’ methylation levels in peripheral blood between the PD and nPD groups were displayed in Table S1. Differences were observed for H19b5, HA9d3, H19d4, MALAT1a1, and MALAT1b9 sites (adjusted P-value<0.05 for all), and the differences in the methylation levels between PD and nPD groups of these islands and sites were shown in Fig. 1E, F. After combining the identified three CpG sites (H19b5, HA9d3, and MALAT1b9) and the relevant clinical parameters, including age, sex, smoke, and drink, a predictive nomogram model was constructed (Fig. 1G). The C-index of the generated nomogram was 0.838, with a Sen of 0.792, Spe of 0.846, PPV of 0.826, and NPV of 0.815, suggesting a greater predictive value than the above CpG sites alone (the AUC of H19b5, H19d3, MALAT1b9 alone were 0.687, 0.639, and 0.643, respectively) (Table S5, Fig. 1H). The calibration curve analysis and the test of Hosmer–Lemeshow (P = 0.899) suggested that the built model could predict the chemotherapy response of GC with an accurate predictive capacity (Fig. 1I). The wide range of the threshold probabilities of the DCA curve further corroborated the clinical value of the nomogram in predicting the chemotherapy effect of GC (Fig. 1J).

Fig. 1.

Predictive model for predicting the chemotherapy of GC (A-D) Chemotherapy response prediction of 5-Fluorouracil and Cisplatin for H19 and MALAT1 in GC. The box in the box plot represents the median and the upper and lower quartiles, while the horizontal lines extending from the box to the top and bottom of the plot depict the maximum and minimum values after removing outliers. (E) Forest plots of the CpG region/sites with the P-value of multinational regression analysis less than 0.05. (F) Differences of the methylation levels between PD and nPD groups of the identified CpG region/sites. (G) Logistic regression-based nomogram model for predicting the chemotherapy of GC. (H) Receiver operating characteristic curve of the CpG sites alone and nomogram. (I) The calibration curve and the Hosmer–Lemeshow (H-L) test of the nomogram. (J) Decision curve analysis of the nomogram. GC: gastric cancer, PD: Progressive Disease, nPD: including CR (Complete response), PR (Partial Response), and SD (stable disease).

Values of methylation levels in predicting the clinical outcomes of GC

To better investigate the potential value of MALAT1 and H19 in predicting the clinical outcomes of GC, multiple variables were discussed that could reflect the survival outcome, including DFS, PFS, and OS. As of the date of the last follow-up (October 21, 2021), 16 patients experienced recurrence (31.4 % of 51 GC patients who were at I-III stages and undergoing CS), 95 patients experienced disease progression or death (96 % of 99 patients in LAMS), and 106 patients were dead (70.7 % of all GC patients). The median follow-up of DFS, PFS, and OS was 1146 (95 %CI, 1118–1173) days, 1088 (95 %CI, 1038–1137) days, and 1146 (95 %CI, 1115–1176) days, respectively. The comparisons between categories in the methylation of MALAT1 or H19 for DFS, PFS, and OS were presented in Table S2-S4. The CpG regions and sites that were statistically significant are shown in Table 2 and Fig. 3A (P-value of multivariate Cox regression<0.05 was presented). Specifically, we found that compared with nPD group, H19c13 was hypermethylated in the PD group, with HR as 2.562 (95 %CI, 1.326–4.952, P = 0.005). After adjusting for age, sex, smoke, and drink, we found that the H19c14 site was hypomethylated instead, with HR as 0.398 (95 %CI, 0.165–0.96, P = 0.04). Combining H19c13, and other clinical variables, including sex, smoke, and drink, we built a pragmatic nomogram model to predict the DFS of GC, and a robust c-index of 0.841 was noticed (Fig. 2A). Fig. 2B, C suggested that the nomograms have great predictive ability and clinical capacity. The constructed nomogram also has great values in predicting the 3-year survival of DFS in GC, with an AUC of 0.838, Sen of 0.824, Spe of 0.788, PPV of 0.660, and NPV of 0.899 (Table S5, Fig. 2D).

Table 2.

Association between methylation levels of H19 or MALAT1 promoters in peripheral blood and survival status among GC patients (P < 0.05).

Gene/Regions/Sites		Methylation Levela		Univariate Cox Regression		Multivariable Cox Regression
		nPD/Alive	PD/Dead	HR (95 %CI)	P-value	HR (95 %CI)	P-value
DFS	H19c13	97.7(97.12,98.18)	98.25(97.67,98.87)	2.562(1.326,4.952)	0.005	2.472(1.266,4.824)	0.008
	H19c14	98.39(98,98.68)	98.27(97.8,98.46)	0.516(0.225,1.183)	0.118	0.398(0.165,0.96)	0.04
PFS	H19C	95.67(95.41,95.93)	95.84(95.44,96.13)	1.489(0.98,2.263)	0.062	1.565(1.019,2.404)	0.041
	MALAT1C	67.47(63.6,71.76)	72.43(69.29,74.51)	1.051(1.007,1.097)	0.023	1.052(1.006,1.1)	0.027
	H19a2	43.41(42.82,45.59)	44.22(43.12,45.74)	0.937(0.893,0.983)	0.008	0.929(0.882,0.978)	0.005
	H19b2	96.52(96.43,98.35)	97.48(97.03,97.86)	1.288(0.979,1.693)	0.07	1.356(1.016,1.81)	0.038
	H19b12	96.9(96.55,97.67)	97.22(96.78,97.51)	1.296(0.955,1.76)	0.096	1.482(1.061,2.072)	0.021
	H19b14	93.06(92.07,95.22)	95.47(94.8,96.22)	1.403(1.169,1.684)	2.77E-04	1.461(1.206,1.769)	1.06E-04
	H19c2	85.67(83.8,88.47)	86.92(85.12,88.92)	1.114(1.026,1.209)	0.01	1.128(1.037,1.228)	0.005
	H19c4	94.43(92.32,94.57)	95.85(95.08,96.59)	1.307(1.102,1.55)	0.002	1.338(1.121,1.598)	0.001
	H19c7	97.59(95.32,97.7)	97.94(97.44,98.4)	1.386(1.082,1.775)	0.01	1.36(1.049,1.763)	0.02
	H19c17	97.98(96.91,99.66)	97.7(97.28,98.09)	0.78(0.605,1.005)	0.054	0.762(0.586,0.992)	0.044
	H19c19	98.09(96.95,99.67)	97.25(96.69,97.56)	0.806(0.644,1.01)	0.061	0.791(0.626,0.999)	0.049
	MALAT1b9	1.63(0.97,1.69)	0.76(0.52,0.92)	0.488(0.273,0.87)	0.015	0.474(0.258,0.871)	0.016
	MALAT1c2	70.87(69.17,75.34)	73.95(71.24,76.39)	1.052(1.007,1.1)	0.023	1.052(1.005,1.102)	0.031
	MALAT1c3	61.61(55.89,66.55)	67.38(63.04,70.97)	1.04(1.007,1.074)	0.018	1.039(1.005,1.075)	0.024
	MALAT1c5	69.8(64.1,75.62)	75.55(72.39,78.4)	1.059(1.017,1.103)	0.006	1.058(1.014,1.104)	0.009
	MALAT1c6	75.84(73.93,78.48)	79.81(77.28,82.5)	1.057(1.004,1.112)	0.033	1.06(1.005,1.117)	0.031
OS	MALAT1	8.76(8.63,8.95)	8.93(8.71,9.12)	2.312(1.193,4.481)	0.013	2.432(1.23,4.809)	0.011
	MALAT1A	20.77(20.48,21.12)	21.22(20.6,21.73)	1.397(1.076,1.814)	0.012	1.425(1.089,1.864)	0.01
	H19b7	91.95(90.61,92.91)	92.64(91.64,93.37)	1.108(0.988,1.243)	0.079	1.143(1.011,1.291)	0.032
	H19b9	94.51(93.17,95.49)	94.6(94.01,95.6)	1.151(1.01,1.311)	0.034	1.18(1.031,1.351)	0.016
	H19b10	97.61(97.13,97.92)	97.33(96.76,97.73)	0.724(0.549,0.955)	0.022	0.711(0.538,0.94)	0.017
	H19b14	95.23(94.14,95.77)	95.39(94.78,96.1)	1.301(1.083,1.564)	0.005	1.322(1.099,1.59)	0.003
	H19b15	97.35(96.89,97.85)	97.79(97.4,98.05)	1.5(1.094,2.056)	0.012	1.517(1.104,2.084)	0.01
	H19c2	86.54(84.67,87.65)	86.78(84.95,88.92)	1.076(1.009,1.147)	0.025	1.087(1.015,1.164)	0.017
	H19c3	95.54(94.66,96.25)	95.99(95.13,96.69)	1.178(1.031,1.345)	0.016	1.198(1.043,1.376)	0.01
	H19c4	95.27(94.39,95.91)	95.8(95.01,96.58)	1.232(1.077,1.409)	0.002	1.254(1.09,1.443)	0.002
	H19c7	97.78(96.99,98.34)	98(97.47,98.42)	1.347(1.088,1.666)	0.006	1.358(1.093,1.687)	0.006
	H19c13	97.74(97.3,98.28)	98.05(97.64,98.65)	1.311(1.024,1.677)	0.032	1.32(1.028,1.695)	0.029
	MALAT1a4	11.85(10.94,13.09)	12.48(11.72,13.74)	1.209(1.064,1.374)	0.004	1.222(1.069,1.397)	0.003
	MALAT1b2	0.49(0.35,0.67)	0.56(0.41,0.74)	1.774(1.054,2.986)	0.031	1.759(1.037,2.983)	0.036

^a: Methylation level is expressed as a percentage. Data was expressed as median (P25, P75). PD: Progressive Disease, nPD: including CR (Complete response), PR (Partial Response), and SD (stable disease).

Fig. 3.

Prognostic model for predicting the OS of GC (A) Forest plots of the CpG region/sites with the P-value of Multivariable Cox regression analysis less than 0.05. (B) Cross-validation for tuning parameter screening in the LASSO regression model. (C) Coefficient profiles in the LASSO regression model. (D) Multivariable Cox regression-based nomogram for predicting the OS of GC. (E) The calibration curves of the nomogram for OS of GC. (F) Decision curve analysis of the nomogram for OS of GC. (G) Time-dependent ROC analysis of the nomogram for predicting 1-year, 2-year, and 3-years of OS. (H) Risk score analysis of a risk-score model based on based on the Cox regression results of OS. GC: gastric cancer, OS: overall survival, Lasso: least absolute shrinkage and selection operator.

Fig. 2.

Prognostic model for predicting the DFS and PFS of GC. (A, F) Multivariable Cox regression-based nomograms for predicting the DFS and PFS of GC. (B, G) The calibration curves of the nomograms for DFS and PFS of GC. (C, H) Decision curve analysis of the nomograms for DFS and PFS of GC. (D, I) Time-dependent ROC analysis of the nomograms for predicting 1-year, 2-year, and 3-years of DFS and PFS. (E, J) Risk score analysis of a risk-score model based on based on the Cox regression results of DFS and PFS of GC. DR: disease recurrence; nDR: disease did not recurrence, GC: gastric cancer, DFS: Disease-free survival, PFS: progression-free survival.

As for the potential value of the methylation in predicting the PFS of GC, similar analyses were conducted. As shown in Table 2 and Fig. 2A, 10 CpG sites and the H19C island were hypermethylated in the PD or dead group, compared to the nPD or alive group (HR>1, P-value<0.05 for all in the multivariate Cox analysis). While the hypomethylation was noticed for H19a2, H19c17, H19c19, and MALAT1b9 sites (HR <1, P-value<0.05 for all). After the ROC analysis, 3 CpG sites were selected for the construction of the nomogram model (H19b14, H19c4, and MALAT1b9). Combining these sites with age, we generated a predictive nomogram to predict the PFS of GC, and a c-index of 0.636 was noticed (Fig. 2F). Using the calibration curve and DCA, we also corroborated the clinical capacity of the model, and the results were shown in Fig. 3G, H. Based on the time-dependent ROC analysis, we observed good applicability in predicting the 3-year survival of PFS in GC, with an AUC of 0.963, Sen of 0.926, Spe of 1, PPV of 1, and NPV of 0.364 (Table S5, Fig. 2I).

Regarding OS, we found 11 CpG sites, MALAT1A island and MALAT1 gene were hypermethylated in the dead group, while the H19b10 site was hypomethylated instead (Table 2 and Fig. 3A). We applied a Lasso regression analysis to avoid the potential overfitting of the selected variables based on the multivariable Cox regression analysis in OS. As shown in Fig. 3B, C, 10 CpG sites were identified as the predictive variables with the highest values (MALAT1b2, H19c13, H19b15, H19c7, MALAT1a4, H19c4, H19b14, H19b7, H19b9, and H19c3). The above sites were selected for further ROC analysis, and the three sites with the highest AUC were MALAT1a4, H19b15, and H19c4. Combining these CpG sites and TNM stages, lymphatic metastasis, and viscera metastasis, we constructed a predictive nomogram for predicting the OS of GC, with the c-index of predicting the 3-year survival as 0.755 (Fig. 3D). The calibration curve and DCA analyses also indicate that the prediction of OS among GC was consistent with the prognosis (Fig. 3E, F).

Besides, each GC patient's risk score was calculated based on the previously identified variables. Based on the median value of the risk score, the patients can be divided into low- and high-risk groups, and we observed that patients with a higher risk score tend to have a worse chemotherapy response and clinical outcomes (Fig. 2E, J, and Fig. 3H).

Interaction between CPG sites and between CPG sites and environmental factors on the chemotherapy effect and prognosis of GC

As shown in Table S6 and S7, no additive interaction exists among the CpG sites and environmental factors on the effect of chemotherapy as well as the OS of GC, while multiplicative interaction was noticed between the methylation of MALAT1a4 site and age on OS of GC (P-value=0.006, Table S8).

Genetic characteristics, immunologic status, and enrichment analysis of MALAT1 and H19 in GC

Using the bioinformatics analysis, we also explored the association between the genetic alteration of these genes in tissue and the prognosis of GC. As shown in Fig. 4A, among 412 patients with GC from the TCGA database, the alteration frequency of MALAT1 and H19 was 7 % and 9 %, respectively. Additionally, the gene expression of MALAT1/H19 in various alteration types was evaluated and compared. As presented in Fig. 4B, we found only genetic alterations of MALAT1 changed the gene expression (P = 0004). However, the genetic alteration was not related to the survival of GC (P > 0.05 for all) (Figure 4C&D).

Fig. 4.

Genetic and immunologic landscape of MALAT1 and H19 in GC (A) Genetic alterations of MALAT1 and H19 in 412 GC patients from TCGA database. (B) Comparisons of MALAT1 or H19 expression between various mutation types. Associations between gene expression of MALAT1 or H19 and the marker genes of immune checkpoints and immune pathways in GC. (C&D) Association between the alteration of H19 or MALAT1 and the Overall Survival and Disease-Free Survival of GC. (E) Associations between gene expression of MALAT1 or H19 and the level of immune cell infiltration calculated by different algorithms, the immune checkpoint monocles, and TIDE score in GC. (F&G) The box plot visualizes significantly different immune cells between expression-high and expression-low subtypes. (H) Kaplan-Meier curves for the corresponding immune infiltrates and MALAT1/H19 expression in GC. GC: gastric cancer.

As for the association between MALAT1/H19 expression and the immunological characteristics in GC, we found the expression of MALAT1 and H19 was significantly correlated to 29 and 26 of the 60 immune checkpoints, respectively (Figure S7A, Table S9). In comparison, MALAT1 and H19 expression were also significant correlated with 71 and 73 of the 148 immune pathway-associated genes, respectively. (Figure S7B, Table S10). The immune cells’ infiltration differences between gene expression-high and low groups of H19 and MALAT1 were further examined in this research. As shown in Fig. 4E-G and Table S11, H19 expression was negatively correlated with the infiltration of B cells, CD4+ T cells, and positively associated with the level of macrophage infiltration. While MALAT1 expression was significantly related to the infiltrating CD8+ T cells, neutrophil, macrophage, and myeloid dendritic cells. Moreover, we noticed that the transcription levels of MALAT1 and H19 were negatively associated with the expression of several immune checkpoint monocles. Besides, a substantial positive correlation between H19expression and the TIDE score in GC was observed (Fig. 4E, Table S11). Notably, we found that the high infiltration of macrophage cells in H19 and MALAT1-high groups and the low infiltration of CD8+ T cells in MALAT1-high groups were associated with worse survival of GC. While the infiltration of other immune cells, such as the myeloid dendritic cells, B cells, and neutrophil was not related to the prognosis of GC (Fig. 4H).

In further study, we explored the differences in KEGG and HALLMARK functions between MALAT1/H19 high and low subgroups. The statistical enriched pathways were shown in Figure S8 and Table S12–15. Noticeable, we found the common KEGG and HALLMARK pathway was the process related to fatty acid metabolism, suggesting its potential role in the carcinogenesis and development of GC.

Discussion

Based on the complexity of the molecular biology of GC, the biology of the occurrence and prognosis of GC varies by histology and site. Cancer tissue is a commonly used specimen in GC-related research; however, tissue sampling faces difficulties and obstacles due to the high cost and poor clinical compliance in patients. Several traditional tumor biomarkers that are less invasive have been used to detect GC, such as carbohydrate antigen 19–9 and carcinoembryonic antigen; however, the diagnostic value is not strong enough [32]. Peripheral blood-based DNA methylation studies have emerged as a promising approach for early detection and diagnosis of cancer, including GC, and have become a hot research topic in recent years [7,12,33,34]. Using peripheral blood DNA methylation abnormalities as cancer-related biomarkers enables convenient, rapid, and minimally invasive sampling, which can increase compliance and facilitate population-based studies. In addition, peripheral blood leukocyte DNA offers advantages such as requiring smaller specimen volumes, easy accessibility, low cost, high quality, and stable detection results. Therefore, investigating the methylation pattern of H19 and MALAT1 in peripheral blood leukocytes can provide a noninvasive and convenient approach to studying their potential roles in GC development and progression

The higher expression of H19 or MALAT1 was associated with worse survival of GC, and the gene expressions were negatively related to their methylation levels [12]. Therefore, we hypothesis that the methylation level of H19 or MALAT1 may have an opposite prognostic value in GC. In this work, we observed that the worse survival of GC might be associated with the high infiltration of Macrophages in the H19-high group and with the low infiltration of CD8+ T cells in the MALAT-high group. Utilizing bioinformatics analysis, we noticed that the hypermethylation of H19 was linked to better survival of GC, while the hypermethylation of MALAT1 was related to worse OS of GC instead. The controversial function of DNA methylation of these genes might be due to the different interaction between DNA methylation and immune system, as the methylation status of genes was related to the immune cell differentiation and immune function [35]. In this research, the genes’ hypermethylation in peripheral blood leukocytes was related to a worse survival of GC. As the methylation status is a dynamic process, and its pattern is characterized by tissue specificity, the methylation status may be different in different tissue types and periods, so it is reasonable that the prognostic role of DNA methylation in GC between tissues and peripheral blood leukocytes is different. However, further studies should be implemented to validate such a hypothesis. Overall, we observed differential methylation of MALAT1 or H19 between groups with good and poor chemotherapy effects. Moreover, the prognostic nomogram models constructed by combining the CpG sites and relevant clinical parameters demonstrated robust AUC values and meaningful sensitivity and specificity. These findings suggest the potential value of MLAT1 and H19 methylation as non-invasive biomarkers for predicting chemotherapy response and prognosis in GC.

The potential value of DNA methylation in monitoring the chemotherapy effect has been discussed by previous studies [36], and the changes in DNA methylation caused by first-line chemotherapy were found to correlate to the survival outcomes [37]. For GC, chemotherapy remains the primary strategy of palliative therapy for patients with advanced or metastatic disease, and platinum plus fluoropyrimidine is the globally accepted first-line option for GC chemotherapy [38,39]. Thus, it is necessary to identify a practical clinical model to predict the chemotherapy effect of GC patients who undergo PPF treatment. The regulation of MALAT1 in response to chemotherapy has been reported in various cancers [40]. Of notice, Gong et al. reported that the role of MALAT1 in response to platinum-based chemotherapy is similar to that in lung cancer [41]. In comparison, the role of H19 in response to chemotherapy of cancers has not been widely studied. In the current research, we first identified the potential value of genes’ expression in predicting the chemotherapy response of PPF treatment on GC. Then, we further explored whether the genes’ DNA methylation in peripheral blood leukocytes could predict the chemotherapy sensitivity of PPF in GC. Of interest, we identified several CpG regions and sites of MALAT1 and H19, and the conducted predictive nomogram model has significant clinical value in predicting the response to chemotherapy of GC, which would help identify more effective treatment options for GC patients.

The aberrant methylation of lncRNA genes is crucially involved in the prognosis cancers, including GC [42]. As one of the important lncRNAs, the abnormal MALAT1 expression is related to the worse metastasis and survival of multiple cancers [13]. As the first identified lncRNA that was associated with cancer, the abnormal expression of lncRNA H19 was associated with the worse survival of numerous cancers [43]. Intriguingly, the aberrant methylation level of H19 was also associated with cancer formation [12], and its hypermethylation in tissue was related to the progression of liver cancer [17]. However, the above findings were based on the transcriptional levels of genes. Whether the two genes’ DNA methylation level could predict the GC's prognosis remains to be clarified. The commonly used classification of GC is the TNM stage, which can codify the tumor's anatomical extent and help create treatment decisions and predict clinical outcomes [44]. However, the TNM stage grouping-based model for survival prediction is built not for individual cancer patients but the patient population. More individualized predictive models with higher clinical guiding significance for the prognosis of GC are needed. In our study, we found the pragmatic nomogram that combined the CpG sites with higher predictive value and the TNM stage could intuitively and accurately predict the long-term prognosis of GC. Similarly, higher methylation of MALAT1 or H19 was also observed in groups with worse DFS and PFS, and the generated nomograms that combined the CpG sites and relevant clinical parameters presented great clinical applicability. Furthermore, there is a multiplicative interaction between the methylation of MALAT1 and age on the prognosis of GC, which could facilitate a better understanding of the mechanisms that affect the prognosis of GC.

We also found low genetic mutation frequencies among GC patients based on data from the TCGA database. Although the genetic alterations of MALAT1 changed the gene expression; however, such alteration was not correlated to the survival outcomes of GC, which was consistent with previous studies that the genes involved in GC-related pathways were less frequently associated with genetic alterations than with DNA methylation [45]. Moreover, the immune cell infiltration has been identified as one of the important hallmarks of cancer [46], and the immunity deficiency could contribute to tumorigenesis, progression, and worse clinical outcomes. Mounting evidence implies that the immune checkpoint mechanism has a significant impact on restraining T cell-mediated anti-tumor immune response within the tumor microenvironment, as per the findings of prior research [47]. Since the DNA methylation is crucially related to the presence of immune cell infiltration in breast cancer has been identified [48], we hypothesize that the two genes’ methylation in peripheral blood may interact with the infiltrating of immune cells during the prognosis of GC; however, further experiments are needed to discuss such hypothesis. Herein, we observed significantly associations between MALAT1/H19 expressions and the infiltration levels of multiple immune cells, as well as the levels of immune checkpoint monocles in GC. Moreover, a robust positively correlation between the TIDE score and H19 expression was observed. The above findings suggesting that MALAT1 and H19 could also serve as a potential target for immunotherapy of GC; however, more studies in vivo and in vitro are required to corroborate these finding. It has been reported that the effect of H19 on GC is mediated by the direct upregulation of ISM1 and the indirect suppression of CALN1 expression through miR-675 [18]. In the present study, we observed that pathways related to immunodeficiency, drug metabolism, and fatty acid metabolism were upregulated in H19-low patients, whereas apoptosis, G2M checkpoint, and lipid metabolism, including fatty acid metabolism, were enriched in MALAT1-low patients. These findings indicate a potential interaction between these pathways and MALAT1/H19 in the development of GC. However, further functional experiments are needed to establish this association and elucidate the underlying mechanisms of H19 and MALAT1 in GC.

Several shortcomings should be recognized. Firstly, the genes’ expression levels and their immune status in the peripheral blood of the recruited participates had not been detected. Thus, the direct analysis of these parameters is impossible. Secondly, the methylation levels of H19 and MALAT1 at different time points during the treatment have not yet been evaluated. Therefore, it remains unclear whether the treatment influences or is able to change the methylation status of these genes. Thirdly, the sample size was not large enough, and the sample size has not been estimated. Therefore, future multicenter prospective follow-up studies with larger samples are needed to collaborate with the above results and test the two genes’ methylation value in peripheral blood leukocytes in predicting the chemotherapy and clinical outcomes of GC and functional experiments are required to further elucidate the underlying mechanisms.

Conclusions

We found significant differential peripheral blood leukocytes-based methylation of MALAT1 and H19 between groups with worse chemotherapy effect and prognosis of GC, and the nomograms that contain the CpG sites and relevant clinical parameters showed great clinical applicability, suggesting that the methylation of these genes have the value to predict the chemotherapy and prognosis of GC.

Author contributions

FW conceived the study idea. DH collected the data. XL, LW, YW, TZ, ZY, NM, YL, FZ, and FW contributed to the analysis of the data. FW and DH wrote the initial draft with all authors providing critical feedback and edits to subsequent revisions. All authors approved the final draft of the manuscript. All authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. FW is the guarantor. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Natural Science Foundation for the Higher Education Institutions of Anhui Province of China (2022AH051145), the Research Fund of Anhui Institute of translational medicine (2022zhyx-C35), the National Natural Science Foundation of China (81602115), the Foundation of Supporting Program for the Excellent Young Faculties in Universities of Anhui Province in China (gxyq2019012) and the Outstanding Youth from the First Affiliated Hospital of Anhui Medical University.

Competing interests

The authors have no relevant financial or non-financial interests to disclose.

Ethical approval

Our study was approved by The Ethics Committee of Anhui Medical University. All patients provided written informed consent prior to enrollment in the study.

Data availability

All datasets involved in this study can be viewed in The Cancer Genome Atlas (TCGA), the Genomics of Drug Sensitivity in Cancer (GDSC) (https://www.cancerrxgene.org/), the Molecular Signature Database (MSigDB) (https://www.gsea-msigdb.org/gsea/msigdb/), cBioPortal (https://www.cbioportal.org/), and TIMER (http://timer.cistrome.org/) or data availability part of the corresponding articles. All the other data supporting the findings of this study are available within the article and its Supplementary Information Files, or from the corresponding authors upon reasonable request.

CRediT authorship contribution statement

Fang Wang: Conceptualization, Writing – original draft. Dingtao Hu: Data curation, Writing – original draft. Xiaoqi Lou: Formal analysis. Linlin Wang: Formal analysis. Yuhua Wang: Formal analysis. Tingyu Zhang: Formal analysis. Ziye Yan: Formal analysis. Nana Meng: Formal analysis. Yu Lei: Formal analysis. Yanfeng Zou: Formal analysis.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.