3.1 Decreased GRAP2 expression in LUAD
The TIMER database was used to assess the transcriptional level of GRAP2 in different tumors and normal tissues. Results showed that GRAP2 transcriptional level was significantly lower in tumor tissues in matched normal tissues, such as bladder urothelial carcinoma, breast invasive carcinoma, colon adenocarcinoma, LUAD, lung squamous cell carcinoma (LUSC), prostate adenocarcinoma, rectum adenocarcinoma. However, higher transcriptional level was observed in tumors such as esophageal carcinoma and kidney renal clear cell carcinoma (Figure 1A).
We further used the TCGA database to examine the transcriptional level of GRAP2 in tumor tissues compared with adjacent normal tissues in LUAD. The results showed that the transcriptional level of GRAP2 in tumor tissues was significantly lower than that in unmatched adjacent normal tissues (P<0.01) (Figure 1B). These results were verified in matched tumor tissues and adjacent normal tissues (Figure 1C). The similar results were found in LUSC (Supplementary Figure 1A and 1B).
Table 1 summarized the correlation between GRAP2 expression and clinical characteristics in LUAD. The results showed that the low expression of GRAP2 is associated with late T stage, N stage, pathologic stage, and worse primary therapy outcome (Figure 1D-1G and Table 1). The protein expression level of GRAP2 was further investigated by immunohistochemical staining, we found that the GRAP2 protein level was obviously decreased in LUAD tissues compared with adjacent normal tissues (Figures 2A and 2B). Finally, we investigated GRAP2 expression in LUAD cell lines, data showed that GRAP2 mRNA level in three LUAD cell lines (A549, H1975 and H1299) was significantly down-regulated compared with that in a normal lung epithelial cell line (BEAS -2B) (Figure 2C).
Table 1
Correlation between GRAP2 expression and the clinicopathological features of the LUAD cases from TCGA.
Characteristic
|
Low expression of GRAP2
|
High expression of GRAP2
|
p
|
n
|
267
|
268
|
|
T stage, n (%)
|
|
|
< 0.001
|
T1
|
65 (12.2%)
|
110 (20.7%)
|
|
T2
|
160 (30.1%)
|
129 (24.2%)
|
|
T3
|
28 (5.3%)
|
21 (3.9%)
|
|
T4
|
14 (2.6%)
|
5 (0.9%)
|
|
N stage, n (%)
|
|
|
0.002
|
N0
|
158 (30.4%)
|
190 (36.6%)
|
|
N1
|
53 (10.2%)
|
42 (8.1%)
|
|
N2
|
50 (9.6%)
|
24 (4.6%)
|
|
N3
|
1 (0.2%)
|
1 (0.2%)
|
|
M stage, n (%)
|
|
|
0.330
|
M0
|
187 (48.4%)
|
174 (45.1%)
|
|
M1
|
16 (4.1%)
|
9 (2.3%)
|
|
Pathologic stage, n (%)
|
|
|
0.002
|
Stage I
|
128 (24.3%)
|
166 (31.5%)
|
|
Stage II
|
65 (12.3%)
|
58 (11%)
|
|
Stage III
|
55 (10.4%)
|
29 (5.5%)
|
|
Stage IV
|
16 (3%)
|
10 (1.9%)
|
|
Primary therapy outcome, n (%)
|
|
|
0.008
|
PD
|
47 (10.5%)
|
24 (5.4%)
|
|
SD
|
19 (4.3%)
|
18 (4%)
|
|
PR
|
2 (0.4%)
|
4 (0.9%)
|
|
CR
|
149 (33.4%)
|
183 (41%)
|
|
Gender, n (%)
|
|
|
0.154
|
Female
|
134 (25%)
|
152 (28.4%)
|
|
Male
|
133 (24.9%)
|
116 (21.7%)
|
|
Age, n (%)
|
|
|
0.723
|
<=65
|
131 (25.4%)
|
124 (24%)
|
|
>65
|
129 (25%)
|
132 (25.6%)
|
|
Smoker, n (%)
|
|
|
0.443
|
No
|
34 (6.5%)
|
41 (7.9%)
|
|
Yes
|
227 (43.6%)
|
219 (42%)
|
|
Age, meidan (IQR)
|
65 (58, 72)
|
66.5 (59, 72)
|
0.591
|
Table2. Correlation analysis between GRAP2 expression and immune markers of immune cells from TIMER.
Description
|
Gene markers
|
None
|
Purity
|
Cor
|
P
|
Cor
|
P
|
CD8+ T cell
|
CD8A
|
0.727
|
***
|
0.674
|
***
|
|
CD8B
|
0.646
|
***
|
0.593
|
***
|
T cell (general)
|
CD3D
|
0.728
|
***
|
0.654
|
***
|
|
CD3E
|
0.81
|
***
|
0.768
|
***
|
|
CD2
|
0.817
|
***
|
0.773
|
***
|
B cell
|
CD19
|
0.58
|
***
|
0.482
|
***
|
|
CD79A
|
0.51
|
***
|
0.403
|
***
|
Monocyte
|
CD86
|
0.569
|
***
|
0.467
|
***
|
|
CD115(CSF1R)
|
0.545
|
***
|
0.451
|
***
|
Neutrophils
|
CD66b(CEACAM8)
|
0.229
|
***
|
0.229
|
***
|
|
CD11b(ITGAM)
|
0.473
|
***
|
0.382
|
***
|
|
CCR7
|
0.669
|
***
|
0.578
|
***
|
Natural killer cell
|
KIR2DL1
|
0.285
|
***
|
0.238
|
***
|
|
KIR2DL3
|
0.363
|
***
|
0.303
|
***
|
|
KIR2DL4
|
0.317
|
***
|
0.253
|
***
|
|
KIR3DL1
|
0.292
|
***
|
0.236
|
***
|
|
KIR3DL2
|
0.393
|
***
|
0.335
|
***
|
|
KIR3DL3
|
0.147
|
***
|
0.13
|
**
|
|
KIR2DS4
|
0.323
|
***
|
0.269
|
***
|
Dendritic cell
|
HLA-DPB1
|
0.616
|
***
|
0.544
|
***
|
|
HLA-DQB1
|
0.431
|
***
|
0.329
|
***
|
|
HLA-DRA
|
0.568
|
***
|
0.483
|
***
|
|
HLA-DPA1
|
0.594
|
***
|
0.525
|
***
|
|
BDCA-1(CD1C)
|
0.376
|
***
|
0.29
|
***
|
|
BDCA-4(NRP1)
|
0.204
|
***
|
0.169
|
***
|
|
CD11c(ITGAX)
|
0.558
|
***
|
0.469
|
***
|
Th1
|
T-bet(TBX21)
|
0.73
|
***
|
0.677
|
***
|
|
STAT4
|
0.624
|
***
|
0.544
|
***
|
|
STAT1
|
0.478
|
***
|
0.405
|
***
|
|
IFN-γ(IFNG)
|
0.56
|
***
|
0.499
|
***
|
|
TFN-α(TFN)
|
0.42
|
***
|
0.296
|
***
|
Th2
|
GATA3
|
0.492
|
***
|
0.393
|
***
|
|
STAT6
|
0.2
|
***
|
0.236
|
***
|
|
STAT5A
|
0.658
|
***
|
0.592
|
***
|
|
IL13
|
0.28
|
***
|
0.219
|
***
|
Tfh
|
BCL6
|
0.066
|
***
|
0.071
|
|
|
IL21
|
0.408
|
***
|
0.374
|
***
|
Th17
|
STAT3
|
0.113
|
*
|
0.137
|
**
|
|
IL17A
|
0.33
|
***
|
0.271
|
***
|
Treg
|
FOXP3
|
0.652
|
***
|
0.562
|
***
|
|
CCR8
|
0.653
|
***
|
0.569
|
***
|
|
STAT5B
|
0.469
|
***
|
0.478
|
***
|
|
TGFβ(TGFB1)
|
0.38
|
***
|
0.291
|
***
|
Effector T-cell
|
CX3CR1
|
0.381
|
***
|
0.324
|
***
|
|
FGFBP2
|
0.327
|
***
|
0.278
|
***
|
|
FCGR3A
|
0.483
|
***
|
0.394
|
***
|
Effector memory T-cell
|
DUSP4
|
-0.023
|
|
-0.032
|
|
|
GZMK
|
0.737
|
***
|
0.673
|
***
|
|
GZMA
|
0.681
|
***
|
0.613
|
***
|
Resident memory T-cell
|
CD69
|
0.683
|
***
|
0.617
|
***
|
|
CXCR6
|
0.78
|
***
|
0.729
|
***
|
|
MYADM
|
0.181
|
***
|
0.079
|
|
General memory T-cell
|
CCR7
|
0.669
|
***
|
0.578
|
***
|
|
SELL
|
0.621
|
***
|
0.527
|
***
|
|
IL7R
|
0.633
|
***
|
0.537
|
***
|
Exhausted T-cell
|
HAVCR2
|
0.559
|
***
|
0.456
|
***
|
|
LAG3
|
0.576
|
***
|
0.509
|
***
|
|
CXCL13
|
0.092
|
*
|
-0.016
|
|
|
LAYN
|
0.281
|
***
|
0.14
|
**
|
3.2 Low GRAP2 expression is an independent prognostic factor for overall survival in LUAD
We investigated whether GRAP2 expression correlates with prognosis in patients with LUAD. We divided the LUAD patients from the TCGA database into high (top 50% samples) and low (50% remaining samples) cohorts according to GRAP2 expression level. LUAD patients with higher expression of GRAP2 expression exhibited good overall survival (OS) (HR=0.61, P=0.001) according to the Kaplan-Meier survival analysis (Figure 3A). However, the correlation between GRAP2 expression and the OS in LUSC was not significant (HR=0.95, P=0.727) (Supplementary Figure 1C). Therefore, we will only analyze the role of GRAP2 in LUAD in the follow-up. Subgroup analysis showed that high GRAP2 expression was significantly associated with longer OS in LUAD under the following features: T2 stage (HR = 0.66, P = 0.035), N0&N1 stage (HR = 0.67, P = 0.018), M0 stage (HR = 0.54, P = 0.001), Pathological stage III (HR = 0.55, P = 0.048), Primary therapy outcome, PD&SD (HR = 0.47, P = 0.008), Residual tumor R0 (HR = 0.54, P = 0.001), Smoker ( HR = 0.61, P = 0.004), Male patients (HR=0.63, P=0.034), Female patients (HR=0.59, P=0.012), Age >65 years (HR=0.57, P=0.007) and Age <=65 years (HR=0.62, P=0.003) (Figure 3B-3L). Cox analysis was also used to explore the correlation between GRAP2 expression and OS. High GRAP2 expression was significantly associated with longer OS (Univariate Cox: hazard ratio HR = 602, 95% CI = 0.448–0.808, P <0.001; Multivariate Cox: hazard ratio HR = 602, 95% CI = 0.448–0.808, P <0.001) (Figure 4). Finally, we used an independent external GEO data set GSE37745 to verify our results, the results were consistent with those above, LUAD patients with higher expression of GRAP2 showed higher OS compared with patients with lower expression of GRAP2 (Supplementary Figure 2). These data indicate that GRAP2 is a tumor suppressor gene and can be used as an independent prognostic factor for OS in LUAD.
3.3 GRAP2 is associated with immune response in LUAD
To examine the biological function of GRAP2 in LUAD, we used the GEPIA database to detect the co-expression pattern of GRAP2 in LUAD. The first 25 genes (top 25) are positively correlated with GRAP2 expression, and the last 25 genes (bottom 25) are negatively correlated with GRAP2 expression in the heat map (Figure 5A). We used the “Link Interpreter” module of the Linked Omics website to evaluated functional enrichment in the genes co-expressed with GRAP2 (top 600) using GO and KEGG annotations, and found that these genes were enriched in immune response processes, such as Th17 cell differentiation, T cell activation, initial immune deficiency, cytokine receptor activation and so on (Figure 5B). We also performed GO and KEGG annotations analysis in LUSC. The data showed that enriched pathways in the genes co-expressed with GRAP2 (top 600) were closely associated with the immune response in LUSC, but there are few overlapping enrichment items between LUAD and LUSC (Supplementary Figure 3).
To gain further insight into the underlying mechanisms that GRAP2 regulates LUAD prognosis. The survival-related and down-regulated genes in LUAD were screened using GEPIA database. We crossed the 600 genes that co-expressed with GRAP2 with 731 survival-related and down-regulated genes in LUAD, we identified a gene-set containing 91 genes at the intersection (Figure 5C). We performed functional enrichment in these 91 genes using GO and KEGG annotations, several biological processes appeared as particularly enriched, including external side of plasma membrane, specific granule membrane, MHC protein complex, T cell activation and lymphocyte differentiation (Figure 5D).
We further used protein-protein interaction (PPI) and correlation analysis to explore the interactions between these 91 proteins. We found a higher enriched protein–protein interaction network among these proteins than random proteins (Figure 6A). Furthermore, most of these proteins in the network have a strong positive correlation with each other (Figure 6B). Therefore, these established genes co-expressed with GRAP2 particularly related to immune response, which may be the molecular mechanisms that GRAP2 affecting the prognosis of patients with LUAD, and can be used as multi-gene biomarkers to predict the survival in LUAD.
3.4 The expression level of GRAP2 correlate with the immune infiltration in LUAD
Since the hub genes were enriched in immune response-related pathways, the association between the immune score and GRAP2 expression was further explored. We divided the cases into two cohorts according to the expression level of GRAP2 and estimated the immune score by the ESTIMATE database. The immune score in cohort with high GRAP2 expression was significantly higher than in low GRAP2 expression cohort (Figure 7A).
Researches show that the survival time of several cancer patients is determined by the numbers and activity status of TIICs [6]. Therefore, we used the TIMER database to explore the correlation between GRAP2 expression and immune infiltration in LUAD. The results showed that numerous immune cells (except γδ T cells and Th2 cells) showed higher immune infiltration level in the cohort with high GRAP2 expression (Figure 7B). Furthermore, GRAP2 expression level was negatively correlated with tumor purity, however, it was positively correlated with infiltration levels of B cells (r=0.579, P<0.001), CD8+ T cells (r= 0.512, P<0.001), CD4+ T cells (r= 0.562, P <0.001), macrophages (r =0.252, PP<0.001), neutrophils (r=0.484, P<0.001) and dendritic cells (r =0.536, P<0.001) (Figure 7C). The correlation between GRAP2 expression and immune cell survival in LUAD was also examined. The data showed that cohort with high GRAP2 expression had a higher cumulative survival time in B cells (P=0) and dendritic cells (P=0.048), but not in CD8+ T cells, CD4+ T cells, neutrophils and macrophages (Figure 7D). These data suggest that GRAP2 plays a specific role in immune infiltration in LUAD.
3.5 GRAP2 is positively correlated with various immune markers
In order to further explore the relationship between GRAP2 and immune response, TIMER database was used to investigate the correlations between GRAP2 expression and diverse immune markers in LUAD. These immune markers can be used to characterize immune cells, including T cell (general), CD8+ T cell, Th1, Th2, Follicular helper T cell, Th17, Treg, Effector T-cell, Effector memory T-cell, Resident memory T-cell, General memory T-cell, Exhausted T-cell, B cell, Monocyte, Neutrophils, Natural killer cell and Dendritic cells. In clinical cancer biopsies, tumor purity is an important parameter to reflect the level of immune infiltration [23]. Our data showed that GRAP2 expression was positively correlated with most immune markers in multiple types of immune cells, 58 of 59 immune markers presenting a positive correlation with GRAP2 expression in LUAD (Table 2). These data further support that GRAP2 expression is significantly correlated with immune infiltration.
3.6 GRAP2 is positively correlated with chemokines/ chemokines receptors and MHC molecules
Chemokines and chemokine receptors play important role in the infiltration of immune cells towards the tumor [24]. To investigate the potential mechanism that GRAP2 regulating immune infiltration in LUAD, TISIDB database was used to examine the correlation between the expression level of GRAP2 and chemokines/ chemokine receptors. The heat map showed that GRAP2 expression was positively correlated with various chemokines and chemokine receptors in multiple tumors (Figure 8A, 8C). We also comprehensively explored the correlation between GRAP2 expression and chemokine/chemokine receptors using scatter plots. Data showed that GRAP2 expression was positively correlated with multiple chemokines, such as CCL4, CCL5, CCL18, CCL19, CXCL9, CXCL10, CXCL11, CXCL13 and XCL2 (Figure 8B). GRAP2 expression level was also positively associated with multiple chemokine receptors, such as CCR2, CCR4, CCR5, CCR6, CCR7, CCR8, CXCR3, CXCR4, and CXCR6 (Figure 8D).
A growing body of research has shown that downregulation of major histocompatibility complex class-I and -II (MHC-I and MHC-II) promote tumor immune suppression, metastatic progression and a poor prognosis in multiple tumors [25, 26]. Therefore, TISIDB database was used to explore the correlation between GRAP2 expression and MHC molecules. The heat map results showed that the MHC molecules in numerous tumors were significantly positively correlated with the expression of GRAP2 (Figure 9A). Scatter plots showed that the MHC molecules were positively correlated with the expression of GRAP2, including B2M, HLA-B, HLA-C, HLA-DMA, HLA-DMB, HLA-DOA, HLA-DOB, HLA-DPA1, HLA-DPB1, HLA- DOA1, HLA-DOA2, HLA-DRA, HLA-DRB1, HLA-E, HLA-F and TAP1 (Figure 9B-9Q). These data suggest that GRAP2 participating widely in modulating multiple immune-related molecules in LUAD to affect immune infiltration and immune response in TME.