The m6A methylation regulators expression level in GC
We performed the analysis of differentially expressed genes in 35 normal tissues and 375 tumor tissues downloaded from the TCGA database. As shown in Figure 1A and 1B, 19 of 22 m6A methylation regulators were expressed in GC, among which, 17 m6A methylation regulators were significantly upregulated in GC tissues, including m6A “writer” (KIAA1429, METTL14, METTL3, METTL4, RBM15, RBM15B, WTAP) and “reader” (DGCR8, EIF3A, EIF3B, ELAVL1, SRSF2, YTHDC1, YTHDC2, YTHDF1, YTHDF2, YTHDF3).
The YTHDF2 expression level in baseline clinicopathological characteristics
To understand the relationship between 19 m6A methylation regulators and patient prognosis, we performed univariate Cox regression analysis for 19 regulators, respectively. We found that m6A “reader” YTHDF2 was obviously associated with the OS (Additional file 1). The results also indicated that YTHDF2 might perform as a protective factor role in GC (HR = 0.629, p = 0.027). Consequently, GC patients were divided into low and high expression groups according to the expression level of YTHDF2 in GC tissues (Figure 2A). Combining the baseline clinicopathological characteristics of the patients, we found that gender and subtype were both related to the expression of YTHDF2 (Table 1). Yet, the YTHDF2 expression level was significantly higher in females than in males, while no difference was found in MSS, MSS-L, and MSS-H (Figure 2B - I).
Table 1 Baseline clinicopathological characteristics
Variables
|
Total
(N=305)
|
YTHDF2 Low
(N=152)
|
YTHDF2 High
(N=153)
|
P value
|
Age, n (%)
|
|
|
|
0.122
|
≤65
|
139 (45.6)
|
76 (24.9)
|
63 (20.7)
|
|
>65
|
166 (54.4)
|
76 (24.9)
|
90 (29.5)
|
|
Gender, n (%)
|
|
|
|
0.049
|
female
|
113 (37.0)
|
48 (15.7)
|
65 (21.3)
|
|
male
|
192 (63.0)
|
104 (34.1)
|
88 (28.9)
|
|
Stage, n (%)
|
|
|
|
0.287
|
Stage I
|
39 (12.8)
|
22 (7.2)
|
17 (5.6)
|
|
Stage II
|
99 (32.5)
|
53 (17.4)
|
46 (15.1)
|
|
Stage III
|
135 (44.3)
|
59 (19.3)
|
76 (24.9)
|
|
Stage IV
|
32 (10.5)
|
18 (5.9)
|
14 (4.6)
|
|
Grade, n (%)
|
|
|
|
0.666
|
G1-2
|
112 (36.7)
|
54 (17.7)
|
58 (19.0)
|
|
G3-4
|
193 (63.3)
|
98 (32.1)
|
95 (31.1)
|
|
T, n (%)
|
|
|
|
0.219
|
T1-2
|
75 (24.6)
|
42 (13.8)
|
33 (10.8)
|
|
T3-4
|
230 (75.4)
|
110 (36.1)
|
120 (39.3)
|
|
M, n (%)
|
|
|
|
0.633
|
M0
|
285 (93.4)
|
141 (46.2)
|
144 (47.2)
|
|
M1
|
20 (6.6)
|
11 (3.6)
|
9 (3.0)
|
|
N, n (%)
|
|
|
|
0.098
|
N0
|
93 (30.5)
|
53 (17.4)
|
40 (13.1)
|
|
N+
|
212 (69.5)
|
99 (32.5)
|
113 (37.0)
|
|
Subtype, n (%)
|
|
|
|
<0.0001
|
MSS
|
204 (66.9)
|
119 (39.0)
|
85 (27.9)
|
|
MSI-L
|
48 (15.7)
|
17 (5.6)
|
31 (10.2)
|
|
MSI-H
|
53 (17.4)
|
16 (5.2)
|
37 (12.1)
|
|
Risk factors of patient survival
To explore the risk factors that can influence the prognosis of GC patients in baseline clinicopathological characteristics, univariate and multivariate Cox regression analyses were used for determining the risk factors for OS. As a result, YTHDF2 expression, patient age, and tumor stage were confirmed as significant predictive factors for OS (Table 2). Patients with higher YTHDF2 expression had significantly better OS. Compared with patients younger than 65 years old, those > 65 years old were at high risk of worse OS. In terms of stage, an increased risk of poor prognosis was detected in stage III and IV, as compared with stage I and II.
Table 2 Cox regression analysis of prognostic factors for overall survival in gastric cancer (GC)
|
Univariate analysis
|
Multivariate analysis
|
HRa (95% CI b)
|
P
|
HR (95% CI)
|
P value
|
YTHDF2 expression
|
|
|
|
|
Low
|
Reference
|
|
Reference
|
|
High
|
0.63 (0.42-0.95)
|
0.027
|
0.59 (0.39-0.89)
|
0.012
|
Age
|
|
|
|
|
≤65
|
Reference
|
|
Reference
|
|
>65
|
1.63 (1.08-2.46)
|
0.019
|
1.89 (1.24-2.88)
|
0.003
|
Gender
|
|
|
|
|
female
|
Reference
|
|
|
|
male
|
1.38 (0.89-2.12)
|
0.147
|
|
|
Stage
|
|
|
|
|
Stage I
Stage II
Stage III
Stage IV
|
Reference
1.78 (0.80-3.95)
2.15(1.01-4.56)
3.70 (1.59-8.57)
|
0.156
0.047
0.002
|
Reference
1.82 (0.82-4.03)
2.32 (1.09-4.92)
4.23 (1.81-9.88)
|
0.138
0.029
<0.001
|
Grade
|
|
|
|
|
G1-2
|
Reference
|
|
|
|
G3-4
|
1.29 (0.86-1.95)
|
0.225
|
|
|
T
|
|
|
|
|
T1-2
|
Reference
|
|
|
|
T3-4
|
1.51 (0.93-2.46)
|
0.096
|
|
|
M
|
|
|
|
|
M0
|
Reference
|
|
|
|
M1
|
1.65 (0.80-3.40)
|
0.177
|
|
|
N
|
|
|
|
|
N0
|
Reference
|
|
|
|
N+
|
1.43 (0.90-2.29)
|
0.130
|
|
|
Subtype
|
|
|
|
|
MSSc
|
Reference
|
|
|
|
MSI-Ld
|
1.40 (0.79-2.46)
|
0.246
|
|
|
MSI-He
|
0.73 (0.42-1.28)
|
0.278
|
|
|
aHR: hazard ratio; b95% CI: 95% confidence intervals; cMSS: microsatellite stability; dMSI-L: microsatellite instability-low; eMSI-H: microsatellite instability-high
YTHDF2 expression and patient survival
We further performed Cox regression analyses for each clinicopathological subgroup of GC patients (Figure 3), which revealed that patients younger than 65 years old, male patients, patients with T3-4, M0, N0, and MSS subtype had a better prognosis.
To ensure the accuracy of the risk factors obtained by our previous analysis, and to exclude the interference of other factors, we used propensity score matching (PSM) to match patients with low and high expression of YTHDF2 on a 1:1 basis. The probability of the YTHDF2 expression level was used as a propensity score, and the nearest neighbor optimal matching algorithm was used to find the best match for each patient. Successful matching was considered as the standard difference (SD) less than 0.2 (Table 3).
Table 3 Propensity score matching (PSM) in gastric cancer (GC) patients based on YTHDF2 expression
Variables
|
Before matching
|
SDa
|
After matching
|
SD
|
Low (n=152)
|
High (n=153)
|
Low (n=106)
|
High (n=106)
|
Age
|
|
|
0.178
|
|
|
0.095
|
≤65
|
76 (24.9)
|
63 (20.7)
|
|
49 (23.1)
|
44 (20.8)
|
|
>65
|
76 (24.9)
|
90 (29.5)
|
|
57 (26.9)
|
62 (29.2)
|
|
Gender
|
|
|
0.227
|
|
|
0.04
|
female
|
48 (15.7)
|
65 (21.3)
|
|
33 (15.6)
|
35 (16.5)
|
|
male
|
104 (34.1)
|
88 (28.9)
|
|
73 (34.4)
|
71 (33.5)
|
|
Stage
|
|
|
0.224
|
|
|
0.124
|
Stage I
|
22 (7.2)
|
17 (5.6)
|
|
11 (5.2)
|
14 (6.6)
|
|
Stage II
|
53 (17.4)
|
46 (15.1)
|
|
35 (16.5)
|
34 (16.0)
|
|
Stage III
|
59 (19.3)
|
76 (24.9)
|
|
48 (22.6)
|
49 (23.1)
|
|
Stage IV
|
18 (5.9)
|
14 (4.6)
|
|
12 (5.7)
|
9 (4.2)
|
|
Grade
|
|
|
0.049
|
|
|
<0.001
|
G1-2
|
54 (17.7)
|
58 (19.0)
|
|
38 (17.9)
|
38 (17.9)
|
|
G3-4
|
98 (32.1)
|
95 (31.1)
|
|
68 (32.1)
|
68 (32.1)
|
|
T
|
|
|
0.141
|
|
|
0.089
|
T1-2
|
42 (13.8)
|
33 (10.8)
|
|
23 (10.8)
|
27 (12.7)
|
|
T3-4
|
110 (36.1)
|
120 (39.3)
|
|
83 (39.2)
|
79 (37.3)
|
|
M
|
|
|
0.055
|
|
|
0.185
|
M0
|
141 (46.2)
|
144 (47.2)
|
|
96 (45.3)
|
101 (47.6)
|
|
M1
|
11 (3.6)
|
9 (3.0)
|
|
10 (4.7)
|
5 (2.4)
|
|
N
|
|
|
0.19
|
|
|
0.063
|
N0
|
53 (17.4)
|
40 (13.1)
|
|
29 (13.7)
|
32 (15.1)
|
|
N+
|
99 (32.5)
|
113 (37.0)
|
|
77 (36.3)
|
74 (34.9)
|
|
Subtype
|
|
|
0.502
|
|
|
0.12
|
MSSb
|
119 (39.0)
|
85 (27.9)
|
|
75 (35.4)
|
74 (34.9)
|
|
MSI-Lc
|
17 (5.6)
|
31 (10.2)
|
|
15 (7.1)
|
12 (5.7)
|
|
MSI-Hd
|
16 (5.2)
|
37 (12.1)
|
|
16 (7.5)
|
20 (9.4)
|
|
aSD: Standard Deviation; bMSS: microsatellite stability; cMSI-L: microsatellite instability-low; dMSI-H: microsatellite instability-high
Cox regression analysis was used to analyze the prognostic factors of patients with PSM. The high expression level of YTHDF2 and stage IV were independent risk factors affecting the prognosis of patients (Table 4). Moreover, survival analysis showed that patients with high expression of YTHDF2 had significantly better OS, and patients with stage IV had a worse OS than those with other stages (Figure 4A and 4B).
Table 4 Cox regression analysis of prognostic factors for overall survival (OS) after matching
|
HRa (95% CIb)
|
P
|
YTHDF2 expression
|
|
|
Low
|
Reference
|
|
High
|
0.58 (0.37-0.91)
|
0.018
|
Age
|
|
|
≤65
|
Reference
|
|
>65
|
1.37 (0.86-2.16)
|
0.184
|
Gender
|
|
|
female
|
Reference
|
|
male
|
1.79 (1.05-3.06)
|
0.034
|
Stage
|
|
|
Stage I
Stage II
Stage III
Stage IV
|
Reference
1.32 (0.53-3.30)
2.15 (0.90-5.10)
3.15 (1.16-8.55)
|
0.548
0.084
0.024
|
Grade
|
|
|
G1-2
|
Reference
|
|
G3-4
|
1.68 (1.04-2.71)
|
0.034
|
T
|
|
|
T1-2
|
Reference
|
|
T3-4
|
1.92 (1.05-3.50)
|
0.033
|
M
|
|
|
M0
|
Reference
|
|
M1
|
1.29 (0.56-2.98)
|
0.548
|
N
|
|
|
N0
|
Reference
|
|
N+
|
1.34 (0.79-2.28)
|
0.272
|
Subtype
|
|
|
MSSc
|
Reference
|
|
MSI-Ld
|
1.64 (0.87-3.08)
|
0.124
|
MSI-He
|
0.73 (0.39-1.37)
|
0.322
|
aHR: hazard ratio; b95% CI: 95% confidence intervals; cMSS: microsatellite stability; dMSI-L: microsatellite instability-low; eMSI-H: microsatellite instability-high.
The potential role of YTHDF2 in GC
To explore the potential role of YTHDF2 in GC, we identified 3066 differential expression genes (DEGs) [|log2(Fold Change)| > 1 and p <0.05] between the low and high expression level of YTHDF2 (Figure 4C and 4D). The Gene Ontology (GO) analyses showed the DEGs were mainly enriched in cancer- and methylation-associated biological progress, cellular component and molecular function, such as histone modification, methylation, methyltransferase complex, histone methyltransferase complex, methyltransferase activity, and p53 binding (Figure 5A and 5B). The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways enrichment analyses were primarily enriched in several cancer-associated pathways, such as cell cycle, RNA degradation, mismatch repair, viral carcinogenesis, apoptosis, and p53 signaling pathway.