Smad4 loss identifies aggressive subtype with immunotherapy and anti-HER-2 treatment resistance in gastric cancer

Abstract

Background

SMAD4 mutation and homozygous deletion represent the most prevalent genomic events driving aggressive biological behavior in gastric cancer (GC). However, clinical outcome and therapeutic response in GC patients with Smad4-loss remains obscure.

Methods

This study included 990 GC patients from four independent clinical centers including the Zhongshan Hospital (ZSHS) cohort, the Cancer Genomic Atlas (TCGA) cohort, the Samsung Medical Center (SMC) cohort and the Memorial Sloan Kettering Cancer Center (MSKCC) cohort.

Results

In ZSHS cohort, 60/454 GC patients harbored Smad4-loss are characterized by lower pN stage, well histology differentiation, lower EBV infection, null p53 staining and lower tumor proliferation. Smad4-loss GC patients exhibit miserable overall survival across ZSHS cohort and TCGA cohort. Moreover, Smad4-loss GC patients yield no impact on adjuvant chemotherapy, poor outcome upon anti-PD-1 immunotherapy or anti-HER-2 therapy. Interestingly, Smad4-loss GC show more well and intermediate differentiation and lower Ki67 staining. Furthermore, Smad4-loss GC exhibit tumor immunosuppressive contexture characterized with enriched CXCL13⁺CD8⁺T cells, reduced IFN-γ⁺ cells and GZMB⁺ cells infiltration.

Conclusions

Smad4 loss yields poor clinical outcome, immunotherapy and anti-HER-2 treatment resistance and tumor immunosuppressive contexture in GC patients. Our findings provide clues for further detailed biological investigation and aggressive clinical management in Smad4-loss GC patients.

Introduction

Gastric cancer (GC) is the fifth most common neoplasm and the fourth leading cause of cancer-related death worldwide [1]. Several genetic alterations have been identified in this lethal disease, including loss of the TP53, ARID1A, CDKN2A, SMAD4, CDH1, RNF43 and APC tumor suppressor genes and activation of the ERBB2, FGFR2, CCNE1, MYC, PIK3CA, KRAS and CTNNB1 oncogenes [2–4]. Although these genomic discoveries have provided crucial insights into the molecular events within multiple molecular subtypes and have spurred efforts to develop improved therapeutic agents targeting various driver pathways, the vast majority of potential driver genes have not been thoroughly analyzed in this cancer type.

SMAD4, also known as DPC4 (Deleted in pancreatic cancer locus 4) located on chromosome 18q21.1, encodes a critical cytoplasmic mediator in the transforming growth factor-β (TGF-β) signaling pathway and has been identified as a candidate tumor suppressor gene whose inactivation may play a role in pancreatic and possibly other human cancers [5, 6]. Somatic mutations or deletions in SMAD4 are present in up to 50% of pancreatic cancers and 30% of colorectal cancers [5, 7, 8]. The cardinal mutation in pancreatic carcinogenesis is KRAS activation occurring at the earliest stages of the disease, whereas SMAD4 mutations are thought to facilitate the progression of pancreatic cancer [9]. Recent researches of pancreatic cancer clinical samples have confirmed that loss of Smad4 protein correlates with worse clinical outcomes [10], and studies in mouse models have indicated that loss of Smad4 confers a metastatic phenotype [11]. Additionally, SMAD4 is also germline mutated in a subgroup of patients with familial juvenile polyposis, an autosomal dominant disorder characterized by a predisposition to hamartomatous polyps and gastrointestinal cancers, defining SMAD4 as a tumor suppressor gene in gastrointestinal carcinogenesis [12].

In GC, somatic mutations and homozygous deletions in SMAD4 have been found in 14.2% of patients, especially in the non-hypermutated CIN or MSS/TP53⁺ molecular subtype [2, 3]. Moreover, loss of Smad4 protein due to homozygous deletion or loss of heterozygosity at 18q21.1 and SMAD4 mutations is associated with worse clinical outcomes and more aggressive phenotypes in GC [13–15]. Mechanistically, SMAD4 inactivation disrupts the TGF-β signaling pathway, leading to impaired regulation of cell proliferation and apoptosis [16]. This disruption promotes epithelial-mesenchymal transition, increasing the invasive potential of GC cells [17]. Furthermore, SMAD4 loss has also been reported to alter the tumor microenvironment, fostering immune evasion by promoting the recruitment of immunosuppressive cells such as regulatory T cells and myeloid-derived suppressor cells [18, 19]. However, no studies have examined the clinical, histological and immunological characteristics of Smad4-loss GC patients as a separate clinical entity to develop effective individualized therapeutic measures for them.

In this study, we included 990 GC patients with matched genomic, pathological and clinical annotations from four independent clinical centers and stratified them into Smad4 loss and Smad4 non-loss subgroups to interrogate the impact of Smad4 loss on clinical outcome, therapeutic response and tumor immune contexture. These findings could provide clues for further detailed biological investigation and aggressive clinical management in Smad4-loss GC patients.

Patients and Methods

Study cohorts

To assess the predictive value of Smad4 status, this study included 990 patients from four independent cohorts: the Zhongshan Hospital (ZSHS) cohort (n = 454), the Cancer Genomic Atlas (TCGA) cohort (n = 433), the Samsung Medical Center (SMC) cohort (n = 43) and the Memorial Sloan Kettering Cancer Center (MSKCC) cohort (n = 60).

Patient Inclusion

The ZSHS cohort includes 454 GC patients approved by the Clinical Research Ethics Committee of Zhongshan Hospital, Fudan University (Shanghai, China; approval number: Y2015-054), whose clinical and pathological features were presented in Table 1. All data were gathered retrospectively, and the survival periods were defined as months after surgery. Patients who received adjuvant chemotherapy (ACT) were defined as patients with tumor-node-metastasis (TNM) II/III advanced tumors treated with at least one cycle of 5- fluorouracil- based ACT, according to Chinese Society of Clinical Oncology and National Comprehensive Cancer Network guidelines for GC and patients’ wills.

Table 1.

Clinical characteristics of the sample set according to Smad4 status in ZSHS cohort (n = 454).

Characteristic	Total (n = 454)	Smad4 non-loss (n = 394)	Smad4 loss (n = 60)	P-value
Age, years				0.071
Median (range)	59.5 (29–90)	59.0 (29–90)	61.5 (39–83)
Sex, n (%)				0.099
Male	321	284 (72.1)	37 (61.7)
Female	133	110 (27.9)	23 (38.3)
Primary tumor location, n (%)				0.741
Proximal	111	94 (23.9)	17 (28.3)
Middle	68	59 (15.0)	9 (15.0)
Distal	275	241 (61.1)	34 (56.7)
pT stagea, n (%)				0.888
T1	76	64 (16.6)	12 (20.3)
T2	61	54 (14.0)	7 (11.9)
T3	88	77 (20.0)	11 (18.6)
T4	219	190 (49.4)	29 (49.2)
pN stage, n (%)				0.014
N0	173	142 (36.0)	31 (51.7)
N1	49	40 (10.2)	9 (15.0)
N2	86	82 (20.8)	4 (6.7)
N3	146	130 (33.0)	16 (26.6)
pM stage, n (%)				0.420
M0	445	387 (98.2)	58 (96.7)
M1	9	7 (1.8)	2 (3.3)
TNM stageb, n, (%)				0.366
I	101	84 (21.8)	17 (28.8)
II	102	87 (22.6)	15 (25.4)
III	232	207 (53.8)	25 (42.4)
IV	9	7 (1.8)	2 (3.4)
Differentiation, n (%)				0.019
Well	23	16 (4.1)	7 (11.7)
Intermediate	87	71 (18.0)	16 (26.7)
Poor	343	306 (77.7)	37 (61.6)
Undifferentiation	1	1 (0.2)	0 (0.0)
WHO histologic type, n (%)				0.608
Adenocarcinoma*	382	330 (83.8)	52 (86.7)
Mucinous adenocarcinoma	29	27 (6.8)	2 (3.3)
Poorly cohesive carcinoma	39	33 (8.4)	6 (10.0)
Neuroendocrine carcinoma	3	3 (0.8)	0 (0.0)
Undifferentiated carcinoma	1	1 (0.2)	0 (0.0)
Lauren classification, n (%)				0.331
Intestinal	282	245 (62.2)	37 (61.7)
Diffuse	111	93 (23.6)	18 (30.0)
Mixed	61	56 (14.2)	5 (8.3)
TCGA subtype, n (%)				0.062
EBV	38	37 (9.4)	1 (1.7)
MSI	86	77 (19.5)	9 (15.0)
GS	87	70 (17.8)	17 (28.3)
CIN	243	210 (53.3)	33 (55.0)
ACRG subtype, n (%)				0.307
MSI	86	77 (19.5)	9 (15.0)
MSS/EMT	89	72 (18.3)	17 (28.3)
MSS/TP53+	97	86 (21.8)	11 (18.3)
MSS/TP53-	182	159 (40.4)	23 (38.4)
EBER status, n (%)				0.044
Positive	38	37 (9.4)	1 (1.7)
Negative	416	357 (90.6)	59 (98.3)
MMR status, n (%)				0.403
Proficient	368	317 (80.5)	51 (85.0)
Deficient	86	77 (19.5)	9 (15.0)
E-cadherin status, n (%)				0.067
Positive	365	322 (81.7)	43 (71.7)
Negative	89	72 (18.3)	17 (28.3)
p53 status, n (%)				< 0.001
WT	158	139 (35.3)	19 (31.7)
Null	96	71 (18.0)	25 (41.7)
OE	200	184 (46.7)	16 (26.6)
ARID1A statusc, n (%)				0.155
Non-loss	254	223 (73.1)	31 (63.3)
Loss	100	82 (26.9)	18 (36.7)
Ki67				< 0.001
Median (range)	25.0 (0–90)	25.0 (0–90)	15.0 (0–45)
PD-L1 CPSd, n (%)				0.646
≥ 1	292	249 (73.9)	43 (76.8)
< 1	101	88 (26.1)	13 (23.2)
Her-2 IHC, n (%)				0.455
0	364	313 (79.4)	51 (85.0)
1+	31	27 (6.9)	4 (6.7)
2+	30	26 (6.6)	4 (6.7)
3+	29	28 (7.1)	1 (1.6)
Chemotherapye, n (%)				0.165
Not applied	188	158 (41.3)	30 (50.8)
Applied	254	225 (58.7)	29 (49.2)
Relapse, n (%)				0.794
No	235	203 (51.5)	32 (53.3)
Yes	219	191 (48.5)	28 (46.7)
Vital status, n (%)				0.172
Alive	234	208 (52.8)	26 (43.3)
Deceased	220	186 (47.2)	34 (56.7)
Overall survival, months				0.004
Median (range)	65.0 (2–79)	Not reached (2–79)	23.0 (2–75)

Statistically significant p < 0.05 values are in bold.

Patients with unknown data were excluded from Chi-square or Fisher’s exact test. pT stage, pathological assessment of the primary tumor; pN stage, pathological assessment of the regional lymph nodes. pM stage, pathological assessment of the metastasis. Adenocarcinoma* Adenocarcinoma, not otherwise specified, TNM Tumor Node Metastasis, TCGA the Cancer Genome Atlas, EBV Epstein-Barr Virus, MSI Microsatellite Instable, GS Genome Stable, CIN Chromosome Instable, ACRG the Asian Cancer Research Group, OE Overexpression, WT Wild Type, CPS Combined Positive Score. ^aPatients with unknown pT stage (n = 10), ^bPatients with unknown TNM stage (n = 10), ^cPatients with unknown ARID1A status (n = 100), ^dPatients with unknown PD-L1 status (n = 61), ^ePatients with unknown chemotherapy status (n = 12).

Public data sets

For TCGA cohort, 433 GC patients with genomic data and clinical information downloaded from http://www.cbioportal.org were included, and a total of 45 patients were excluded due to lack of SMAD4 mutation or copy number variation information or survival information. The SMC cohort was from a single-center, phase 2 trial in which 61 patients with metastatic gastric cancer (mGC) who were treated with pembrolizumab monotherapy (ClinicalTrials.gov, NCT#02589496), 18 patients were excluded due to incomplete clinical information or unknown transcriptomic data, and the clinical and pathological features of SMC cohort were shown in Supplementary Table 1. The MSKCC cohort was comprised of 305 patients with stage IV esophagogastric adenocarcinoma, of which 60 patients treated with anti-HER2 therapy with therapeutic response data were included and patient characteristics of the MSKCC cohort were retrieved from https://www.cbioportal.org/study/summary?id=egc_msk_2017.

Histology and immunohistochemistry (IHC)

Haematoxylin and eosin (H&E) staining, and immunohistochemistry were performed on formalin-fixed paraffin-embedded surgical specimens of GC using standard protocols. The detailed information of IHC antibodies is described in Supplementary Table 2.

Digital IHC images were taken using Image Pro plus 6.0 (Media Cybernetics Inc.) under high-power magnification (HPF, ×200 magnification). IHC staining results were independently scored under 3 randomized fields by two pathologists who were blind to clinical information. Any discrepancies exceeding 5% in scores were re-evaluated to reach a consensus.

Evaluation of Smad4 status

For ZSHS cohort lacking genomic information, as previously reported [13], the results of immunostaining were considered to be positive if ≥ 10% of the neoplastic cells were stained. Positive immunostaining of Smad4 in tumor cells was regarded as normal Smad4 expression (Smad4 non-loss GC), while negative immunostaining Smad4 expression (<10% of the stained neoplastic cells) was defined as deficient Smad4 expression (Smad4-loss GC). For TCGA and MSKCC cohorts, SMAD4-loss GC was defined as patients with SMAD4 mutation (missense mutation, nonsense mutation, frame shift deletion, frame shift Insert, splice site) or copy number homozygous deletion tumors. SMAD4 loss (mutation or homozygous deletion) was shown significantly associated with loss of nuclear Smad4 protein expression [20, 21]. In addition, Smad4 protein expression correlates with SMAD4 mRNA expression [22]. For SMC cohort, 16.3% of patients exhibiting significantly low SMAD4 mRNA expression was defined as SMAD4-loss. SMAD4 loss (mutation or homozygous deletion) was significantly associated with decreased SMAD4 mRNA expression (Supplementary Table 3).

Statistical analysis

Kaplan-Meier analysis was conducted for survival analyses. Pearson’s chi-squared test and Fisher’s exact test were applied to evaluate the association between categorical variables. Multivariable logistic regression and Cox regression analysis were performed to evaluate associations with outcomes. The Mann-Whitney U test was used to assess the difference in a continuous variable between groups. The statistical analysis was two-tailed and P < 0.05 was considered statistically significant. All analyses were conducted by IBM SPSS Statistics v20.0 and R 4.1.2 software.

Results

Clinicopathologic characteristics of GC patients with Smad4 loss

The ZSHS cohort included 454 GC patients, of which 60 participants (13.2%) harbored Smad4 loss. The clinical and demographic characteristics according to Smad4 status are summarized in Table 1. There was no statistically significant difference between the Smad4-loss and the Smad4-non-loss subsets in terms of age, sex, TNM stage, WHO histologic type, Lauren classification, TCGA subtype, ACRG subtype, chemotherapy administration, or tumor relapse. However, compared with Smad4-non-loss patients, Smad4-loss patients were more likely to exhibit stage N0 or N1 (66.7% vs. 46.2%, P = 0.014), well and intermediate histology differentiation (38.4% vs. 22.1%, P = 0.019), EBER negative (98.3% vs. 90.6%, P = 0.044), p53 null (41.7% vs. 18.0%, P < 0.001), E-cadherin negative (28.3% vs. 18.3%, P = 0.067) and low levels of proliferative marker Ki67 staining (15.0% vs. 25.0%, P < 0.001) status.

Smad4 loss yields poor prognosis in GC patients

The ZSHS cohort (n = 454) and TCGA cohort (n = 433) were analyzed to assess the impact of Smad4 loss on prognosis in GC patients. In the ZSHS cohort, the median overall survival was 23.0 months among 60 Smad4-loss GC patients and not reached among 394 Smad4-non-loss GC patients, respectively (log-rank P = 0.032, Fig. 1a left). Moreover, the median overall survival was 21.0 months among 30 Smad4-loss GC patients without adjuvant treatment and not reached among 158 Smad4-non-loss GC patients without adjuvant treatment, respectively (log-rank P = 0.030, Fig. 1a right). Similar poor prognosis was found in SMAD4-loss GC patients for both all and adjuvant treatment- naïve patients in the TCGA cohort (log-rank P = 0.028; log-rank P = 0.004, respectively, Fig. 1b). Multivariate analyses with the cox proportional hazards model showed that Smad4 loss remained an independent prognostic variable after adjustment for classic prognostic variables including age, sex, TNM stage and differentiation status (ZSHS cohort: all patients, HR, 1.59; 95% CI, 1.09–2.31; P = 0.015, Fig. 1c; adjuvant treatment-naïve patients, HR, 1.91; 95% CI, 1.02–3.57; P = 0.043, Supplementary Table 4; TCGA cohort: all patients, HR, 1.92; 95% CI, 1.24–2.99; P = 0.004, Fig. 1d; adjuvant treatment-naïve patients, HR, 1.78; 95% CI, 1.03–3.08; P = 0.038, Supplementary Table 5), confirming Smad4-loss GC patients as an aggressive subtype with poor prognosis, which was consistent with previous studies [13, 14].

Fig. 1. Poor prognosis in Smad4-loss gastric cancer patients.

a Survival curves comparing overall survival between Smad4-loss and Smad4 non-loss GC patients in ZSHS cohort. b Survival curves comparing overall survival between SMAD4-loss and SMAD4 non-loss GC patients in TCGA cohort. c, d Multivariate Cox regression analyses of the correlation between Smad4/SMAD4 status, clinicopathologic factors and OS in (c) ZSHS cohort and (d) TCGA cohort. P-value was obtained using the log-rank test. P < 0.05 was considered statistical significance. ZSHS Zhongshan Hospital, TCGA The Cancer Genome Atlas.

Smad4 loss confers anti-PD-1 immunotherapy and anti-HER-2 treatment resistance in GC patients

GC patients who received ACT treatment from the ZSHS cohort (n = 236) and TCGA cohort (n = 143) were analyzed to assess the impact of Smad4 loss on ACT treatment benefit. However, compared to Smad4-non-loss GC patients, no benefit from ACT was observed among Smad4-loss GC patients in the ZSHS and TCGA cohorts (Fig. 2a, b; Supplementary Tables 6, 7). Furthermore, GC patients upon pembrolizumab treatment from the SMC cohort were analyzed to evaluate the impact of SMAD4 loss on anti-PD-1 immunotherapy response. None of seven SMAD4-loss GC patients from the SMC cohort elicited response to pembrolizumab (Overall Response Rate (ORR), 0% for SMAD4-loss versus 30.6% (11/36) for non-loss; P = 0.163, Fig. 2c). Moreover, 28.6% of SMAD4-loss GC patients achieved disease control rate (DCR), which was lower than 63.9% in SMAD4-non-loss GC patients (P = 0.083, Supplementary Table 1). However, SMAD4-loss yielded no immunotherapy response after adjusting for PD-L1 and MSI status (odds ratio [OR], 0.32; 95% CI, 0.04–2.41; P = 0.266, Supplementary Table 8).

Fig. 2. Unfavorable response upon anti-PD-1 immunotherapy or anti-HER-2 therapies in SMAD4-loss gastric cancer patients.

a, b Kaplan-Meier analyses of overall survival among patients with TNM II/III stage diseases treated with ACT in Smad4 loss and non-loss group in (a) ZSHS cohort and (b) TCGA cohort. P-values were obtained using the log-rank test. c Proportion of patients responding to PD-1 inhibitor immunotherapy in SMC cohort: complete response (CR)/partial response (PR) and stable disease (SD)/progressive disease (PD): 30.6%/69.4% in the SMAD4 non-loss group and 0.0%/100.0% in SMAD4 loss group (P = 0.163, Fisher’s exact test). d Multivariable logistic regression of disease-related feature for anti-PD-1 response after adjusting for PD-L1 and MSI status. e Proportion of patients responding to anti-HER-2 therapy in MSKCC cohort: CR/PR and SD/PD: 69.6%/30.4% in SMAD4 non-loss group and 42.9%/57.1% in SMAD4 loss group (P = 0.069, χ² test). f Multivariable logistic regression of disease-related feature for anti-HER-2 response.

Approximately 20% of GC patients exhibit human epidermal growth factor receptor 2 (HER-2) overexpression or ERBB2 amplification [2], indicating anti-HER-2 therapy as the standard treatment in this patient subset. GC patients upon anti-HER-2 therapy from the MSKCC cohort were analyzed to determine the impact of Smad4 loss on anti-HER-2 therapy response. Notably, CR/PR rates were 6 of 14 (42.9%) for SMAD4-loss GC patients versus 32 of 46 (69.6%) for SMAD4-non-loss GC patients upon anti-HER-2 therapy from the MSKCC cohort (P = 0.069, Fig. 2e). Moreover, SMAD4 loss remained an independent predictor of anti-HER-2 therapy response (OR, 0.298; 95% CI, 0.076 to 1.167; P = 0.082, Fig. 2f, Supplementary Table 9) after adjustment for HER-2 status, RTK, RAS/RAF and PI3K/AKT alteration, which has been recognized as anti-HER2 treatment predictors previously [23, 24].

Smad4 loss elicits benign differentiation and proliferation behaviors in GC patients

Considering the aggressive clinical outcomes in Smad4-loss GC patients, we further explored fundamental histological features of the Smad4-loss and Smad4-non-loss subtypes of GC patients. Surprisingly, Smad4 loss was more common in well- or intermediate-differentiated GC patients compared to Smad4-non-loss tumors (38.4% vs. 22.1%, P = 0.019, Fig. 3a). Moreover, a lower proliferative Ki67 level was observed in Smad4-loss GC patients as compared with Smad4-non-loss tumors (median Ki67 15.0% vs. 25.0%, P < 0.001, Fig. 3b). The histopathologic findings for two representative samples including the well-differentiated tumor (case #127) which exhibited Smad4 loss and 10% proliferation Ki67 level, and the poorly-differentiated tumor (case #52) which exhibited Smad4 non-loss and 30% proliferation Ki67 level, are shown in Fig. 3c. These findings indicate Smad4 loss elicits benign differentiation and proliferation behaviors in GC patients.

Fig. 3. Well differentiation and low proliferation features of Smad4-loss gastric cancer.

a The proportion of GC patients with different degrees of differentiation in terms of Smad4 non-loss or Smad4 loss in ZSHS cohort. b Comparison of Ki67 level (mean ± SD) in tumor tissues of Smad4 non-loss patients (n = 394) and Smad4-loss patients (n = 60). c Two representative cases illustrating well differentiation and low proliferation status in the Smad4-loss sample (upper), while poor differentiation and high proliferation status in the Smad4 non-loss sample (down), confirmed by IHC. Scale bar, 100 μm.

GC patients from the TCGA cohort were analyzed to further assess the relationship between SMAD4 status and tumor genomic events. Increased fraction of genome altered (FGA) was found in SMAD4-loss versus SMAD4-non-loss tumors (P = 0.003, Supplementary Fig. 1A), while TMB showed no differences in both subsets (P = 0.89, Supplementary Fig. 1B). Further genetic evaluations on reported functional CNA genes correlated with OS revealed an increase in GATA6 amplification and 6q21.1 amplification (harboring HSP90AB1, NFKBIE, PTK7, TFEB, VEGFA) in SMAD4-loss tumors (Supplementary Fig. 1C, 1D), indicating SMAD4-loss tumors were prone to genomically unstable and acquisition of aggressive clinical outcomes despite benign cellular behaviors.

Smad4 loss is associated with a CXCL13⁺CD8⁺T-cell-enriched immune contexture in GC patients

Since Smad4 loss elicits benign differentiation and proliferation behaviors but aggressive clinical outcomes, we performed immune cell infiltration analysis on GC patients from the ZSHS cohort to further determine the impact of Smad4 loss on the tumor immune contexture (Supplementary Table 10). Interestingly, intratumoral CD8⁺T cells exhibited lower infiltration in Smad4-loss tumors compared with non-loss tumors (P = 0.062, Fig. 4a). Moreover, CXCL13⁺CD8⁺T cell subset, identified as immunosuppressive CD8⁺ T subset in our previous study [25], showed higher infiltration in Smad4-loss GC patients (P = 0.001, Fig. 4b). In addition, Smad4-loss GC patients demonstrated reduced infiltration of IFN-γ⁺ and GZMB⁺ cells (P =  0.049, Fig. 4c, P = 0.045, Fig. 4d, respectively). The histopathologic findings for two representative samples including the Smad4 non-loss tumor (case #189) exhibited high levels of CD8⁺T cell infiltration and low levels of CXCL13⁺CD8⁺T cell infiltration, while the Smad4-loss tumor (case #187) exhibited low levels of CD8⁺T cell infiltration and high levels of CXCL13⁺CD8⁺T cell infiltration, were shown in Fig. 4e, demonstrating Smad4 loss shapes CXCL13⁺CD8⁺T-enriched immune contexture in GC patients.

Fig. 4. CXCL13⁺CD8⁺ T cell-enriched immunosuppressive microenvironment of Smad4-loss gastric cancer.

a, d Comparison of CD8⁺T (a), CXCL13⁺CD8⁺T (b), IFN-γ⁺ (c) and GZMB⁺ (d) immune cell infiltration (normalized score) between Smad4 non-loss (n = 394) and Smad4-loss (n = 60) GC patients. Data were analyzed by Mann-Whitney U test. e Top, representative IHC images of CD8⁺T and CXCL13⁺CD8⁺T cell from Smad4 non-loss patients. Bottom, representative IHC images of CD8⁺ T and CXCL13⁺CD8⁺ T cell from Smad4-loss patients. Scale bar, 100 μm.

Discussion

In this study, we identified Smad4-loss GC patients as an aggressive subtype with poor prognosis, which was consistent with previous findings [13, 14]. Moreover, increasing evidence marks a groundbreaking change in the treatment of GC with the widespread application of immunotherapy and targeted therapies [26]. Our present results confirmed SMAD4-loss GC patients exhibit immunotherapy and anti-HER-2 treatment resistance, providing clues for further aggressive clinical management and potential novel treatment exploration in Smad4-loss GC patients. Furthermore, we confirmed Smad4-loss GC patients elicits benign differentiation and proliferation behaviors but CXCL13⁺CD8⁺T-enriched immune contexture, indicating tumor immunosuppressive contexture as potential crucial mechanism underlying aggressive clinical outcomes in Smad4-loss GC patients.

Previous studies revealed that genetic alterations have a dual impact on both the biological behavior of tumors and the immune microenvironment. On one hand, they influence the biological behavior of tumor cells, such as proliferation, growth, invasion, and metastasis. On the other hand, genetic alterations can also modify the immune microenvironment of tumors by changing the expression of surface antigens or secreted factors in tumor cells, thereby impacting the recognition and cytotoxicity of immune cells [27, 28]. Interestingly, we have found that these Smad4-loss GC patients actually show higher differentiated and lower proliferative biological behavior characteristics, which seem contradictory to the aggressive clinical outcomes. Furthermore, we found that CXCL13⁺CD8⁺ T cells, which has been proven to be associated with poor clinical outcomes in our previous study [25], were significantly enriched in Smad4-loss GC. Smad4 loss have been shown to impair CD8⁺T cell infiltration and CD8⁺ cytotoxic T cells activity by suppressing dendritic cell differentiation and promoting granulocytic myeloid-derived suppressor cell (G-MDSC) accumulation through CXCL1, leading to immune evasion in GC [19]. CXCL13⁺CD8⁺T cells were confirmed to be a highly exhausted subtype of intratumoral CD8⁺T cells, exhibiting impaired immune function and exhausted markers [29]. Thus, CXCL1 from Smad4-loss GC cells and G-MDSC accumulation may contributed to the enrichment of CXCL13⁺CD8⁺T cells. This subsequently caused decreased expression of the immune effector factors including IFN-γ and GZMB in Smad4-loss GC. Consequently, the suppressed anti-tumor immune response might account for the aggressive clinical outcomes of Smad4-loss GC, despite the highly differentiated and low proliferative biological behavior characteristics.

Although there have been no studies to directly clarify why Smad4-loss GC exhibits highly differentiated and low proliferative biological behavior characteristics, some indirect evidence might make it reasonable. As a key lineage-associated transcription factor, GATA6 plays a crucial role in maintaining proper intestinal epithelial structure and regulating intestinal epithelial differentiation [30]. Furthermore, GATA6 was proven to indicate better degree of tumor cell differentiation as well as inhibited cell proliferation in digestive system malignancies like pancreatic cancer, liver cancer and so on [31–33]. In our study, we found that GATA6 was significantly amplified in SMAD4-loss GC. And the presence of intestinal differentiation-like characteristics in GC might indicate a well-differentiated status [34–36]. Thus, we speculated that GATA6 amplification might be associated with the specific biological behavior characteristics in SMAD4-loss GC. Moreover, GATA6 was reported to contribute to trastuzumab resistance in GC, which might be the possible mechanism for unfavorable response upon trastuzumab treatment in SMAD4-loss GC [37].

Smad4 is a central mediator of TGF-β signaling pathways, and its loss in GC leads to the dysfunction of the TGF-β signaling pathway, which is crucial for gastrointestinal carcinogenesis and progression [38, 39]. Besides, the increased TGF-β in tumor microenvironment is the most potent immunosuppressor via interference with the generation of tumor-specific cytotoxic T lymphocytes [40, 41]. Of particular interest, TGF-β functions as a potent tumor suppressor in pre-malignant cells. Loss of Smad4 in tumor cells not only eliminates the tumor-suppressive role of the TGF-β pathway, but also allows TGF-β to create an immune-suppressive tumor microenvironment that promotes tumor progression.

However, this study has potential limitations. Firstly, the retrospective nature of the analysis inherently carries the risk of selection bias and potential inaccuracies in data reporting. Additionally, we used SMAD4 mutation and copy number deletion to approximate Smad4 loss in the cohorts without Smad4 protein expression data. Although we have confirmed the correlation between SMAD4 mRNA expression and genetic alteration, a possible mismatch between Smad4 protein expression and genetic status might still affect the accuracy of our results. Lastly, considering the small number of patients in the SMC cohort and MSKCC cohort, the result should be interpreted cautiously and further studies are needed to validate our findings.

In summary, this is the first study to characterize the clinical and molecular portraits of Smad4-loss GC patients. Our data suggest that Smad4 loss identifies a subgroup of GC with aggressive behavior, and the identification of Smad4 loss as a predictive biomarker for immunotherapy or anti-HER-2 therapy resistance could have important implications for personalized treatment strategies and patient management in the future. Further research is needed to validate these findings in larger patient cohorts and to elucidate the underlying mechanisms driving treatment resistance in Smad4-loss GC.

Author contributions

MS, YG and JW for the acquisition of data, interpretation of data, statistical analysis, and drafting of the manuscript; ZZ, ZL, FS, CL and HH for technical and material support; RL, HL and JX for study concept and design, analysis and interpretation of data, drafting of the manuscript, obtained funding, and study supervision. All authors read and approved the final manuscript.

Funding

This study was funded by grants from National Natural Science Foundation of China (82103313, 82203201, 82272786, 82303966, 82373417), China Postdoctoral Science Foundation (2023M742327), Shanghai Municipal Natural Science Foundation (23ZR1409900), Shanghai Sailing Program (21YF1407600) and Natural Science Foundation of Fujian Province (2023J05294). All these study sponsors have no roles in the study design, in the collection, analysis and interpretation of data.

Data availability

Data and materials generated that are relevant to the results are included in this article. Other data are available from the corresponding author Prof. Xu upon reasonable request.

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

The study was approved by the Clinical Research Ethics Committee of Zhongshan Hospital, Fudan University, with the approval number Y2015-054. Written informed consent was obtained from each patient included and this study was performed in accordance with the Declaration of Helsinki.

Supplementary information

The online version contains supplementary material available at 10.1038/s41416-025-03002-8.