Microsatellite instability and Epstein-Barr virus combined with PD-L1 could serve as a potential strategy for predicting the prognosis and efficacy of postoperative chemotherapy in gastric cancer

Introduction

Gastric cancer (GC) is still an important global health problem (Lyons et al., 2019) and ranks fifth in terms of incidence but third in terms of mortality, with over 1,000,000 new cases and an estimated 769,000 deaths worldwide in 2020 (Sung et al., 2021). Notably, more than 40% of the worldwide GC cases and deaths occur in China. Based on the latest data from the National Central Cancer Registry of China (NCCRC) in 2015, GC was the second most prevalent cancer and third leading cause of cancer-related deaths in China (Wang et al., 2019). At present, GC patients are still mainly treated with complete tumor resection, and postoperative adjuvant radiotherapy and chemotherapy or neoadjuvant chemotherapy can improve the survival time of patients with advanced GC (Tan, 2019). Surgical resection combined with chemotherapy remains the first-line treatment for advanced GC in China. Even though the GC survival rate has increased due to the improved systemic management over recent decades, the prognosis of GC patients remains poor in China, with a disappointing low 5-year relative survival rate of 35.9% (Yang et al., 2018). More importantly, a variable response to treatment has been observed based on the one-size-fits-all approach to treatment, showing that GC is a heterogeneous disease (Digklia & Wagner, 2016). Thus, there is an urgent need for novel biomarkers to precisely predict the prognosis of GCs and to achieve personalized treatment.

The Cancer Genome Atlas (TCGA) has recently proposed four comprehensive molecular subtypes for GC based on gene expression profile studies (Cancer Genome Atlas Research Network, 2014), which provide new prognostic biomarkers and potential therapeutic implications, particularly the Epstein-Barr virus (EBV)-positive subtype and microsatellite instability (MSI) subtype (Kawazoe et al., 2017). Recent studies have suggested that patients with different molecular subtypes of GC may have different prognoses, and some subtypes, such as MSI, may affect the efficacy of chemotherapy for GC (Smyth et al., 2017). However, in this regard, molecular typing alone may not be sufficient, and more markers should be considered for prognostic evaluation and choice of treatment. Previous studies indicate that the EBV and MSI subtypes may show complex reactions in the immune system, which exhibit high levels of tumor infiltrating lymphocytes (Chang et al., 2018). It indicates that the immune microenvironment of the tumor and tumor infiltration may be different in GC patients with different molecular types. PD-L1 has emerged as the most promising biomarker, which is associated with the cancer immune microenvironment (Jin et al., 2017). PD-L1 expression in tumor or tumor infiltrating immune cells are used as an adjunct diagnostic criteria for the application of PD-1/PD-L1 pathway inhibitors (Muro et al., 2016), but the prognostic value of PD-L1 is controversial in GC patients (Liu et al., 2020). In recent years, higher PD-L1 expression has been found to be closely associated with EBV-positive and MSI GCs (Lee et al., 2017; Derks et al., 2016). However, it is unclear whether the value of PD-L1 expression in predicting GC prognosis is contingent upon EBV and MSI statuses (Choi et al., 2020). Overall, given the complex immune interactions that occur within the tumor microenvironment and available evidence, the addition of PD-L1 expression based on stratification by EBV and MSI statuses would provide better prognostic insight and elucidate the therapeutic implications in GC, whereas the available reports regarding this pattern are insufficient.

In this study, we focused on the molecular subtypes stratified by EBV and MSI statuses according to the TCGA study and analyzed their prognostic value. Furthermore, we examined the level of PD-L1 expression and explored its prognostic value, as well as its role in response to postoperative chemotherapy among different molecular subtypes for GC patients.

Materials & Methods

Results

Molecular subgroups classified by MSI and EBV statuses

In the present study, we defined EBV-positive and MSI subgroups separately according to TCGA classification (Cancer Genome Atlas Research Network, 2014). Patients with MSI status were assigned to the MSI subgroup (n = 52, 23.2%), patients with EBV infection were assigned to the EBV⁺ subgroup (n = 11, 4.9%), and patients without MSI and EBV infection were assigned to the MSS/EBV⁻ subgroup (n = 161, 71.9%). It is worth noting that 2 patients were both positive for MSI and EBV infection, which were not classified in the subsequent subgroup analysis.

MSS/EBV⁻ subgroup was associated with poor prognosis in GC

Log-rank test investigated that MSS/EBV⁻ GCs showed the worst OS compared to MSI and EBV⁺ GCs (P = 0.016, Fig. 2A). The associations between other clinical characteristics and GC survival are shown in Table S2. Results of multivariate Cox regression also showed that patients with MSS/EBV⁻ appeared to have the worst outcomes (HR = 1.547, 95% CI [0.965–2.481], P = 0.077, Table 2) compared to those with MSI. When MSI and EBV⁺ cases were combined as a reference group (non-MSS/EBV⁻ subgroup), multivariate analysis revealed that GCs with MSS/EBV⁻had poorer OS (HR = 1.610, 95% CI [1.046–2.479], P = 0.031, Table S3). Also, PD-L1 expression tended to correlate with improved survival, but multivariate analysis showed marginal significance (HR = 0.684, 95% CI [0.458–1.021], P = 0.063, Table 2).

Table 2:

Multivariate analyses of risk factors affecting overall survival (OS) in 224 gastric cancer patients.

Characteristics		HR	95% CI	P
Molecular subtypes	MSI (n = 52)	1.00		0.077
	EBV+ (n = 11)	0.755	0.275–2.076
	MSS/EBV− (n = 161)	1.547	0.965–2.481
PD-L1	Positive	1.00		0.063
	Negative	0.684	0.458–1.021
TNM stage	I/II	1.00		0.009
	III	2.117	1.208–3.709
Vascular invasion	Negative	1.00		0.020
	Positive	2.216	1.135–4.330
CTx	No	1.00		<0.001
	Yes	0.477	0.331–0.686

DOI: 10.7717/peerj.11481/table-2

Notes:

MSI

Microsatellite instability

MSS

Microsatellite stable

EBV

Epstein-Barr virus

PD-L1

Programmed cell death ligand 1

CTx

Chemotherapy

HR

Hazard ratio

CI

Confidence interval

95% CI and P values were calculated with multivariate Cox regression, adjusting for the variables that P < 0.100 from the univariate analysis, such as PD-L1, WHO classification, histological grade, vascular invasion, neural invasion, postoperative chemotherapy and TNM stage.

Discussion

Recently, more evidence has been emerged about the potential prognostic and therapeutic implications of molecular subtypes, particularly through the stratification according to EBV and MSI status. According to The Cancer Genome Atlas (TCGA) study and recent clinical trials, EBV and MSI GC subgroups exhibited high expression of PD-L1 and could benefit from therapy with PD-1/PD-L1 antibodies. However, very few studies employed combinatorial analysis of MSI, EBV-infection and PD-L1 to evaluate the prognosis and chemotherapy efficacy of gastric cancer. As postoperative chemotherapy is still the first-line treatment for GC patients, so we try to combine conventional chemotherapy with molecular subtyping and PD-L1 expression, which may help to select the patients who would benefit the most from chemotherapy and provide evidence for individualized treatment of patients with GC. In the present study, we observed that GC patients with MSS/EBV⁻ were associated with the worst prognosis compared with those with MSI and EBV⁺. Beyond that, we found that PD-L1-positive expression can predict better OS in the MSS/EBV⁻ subgroup. At the same time, MSS/EBV⁻ GCs can also benefit more significantly from chemotherapy after surgery. Nevertheless, MSS/EBV⁻ GCs with PD-L1⁺ cannot benefit from postoperative chemotherapy.

Among the 226 GC patients in our study, 54 (23.9%) patients were defined as having MSI by IHC, which was similar to that reported in TCGA cohort (22%) (Cancer Genome Atlas Research Network. 2014) and in ACRG Korean cohort (23%) (Angell et al., 2019). The incidence of EBV-associated GC ranged from 4–18% among different regions (Pikuła et al., 2020), our result (4.9%) is identical to the data of a recent Korean study (25/569, 4.4%) (Cho, Kang & Kim, 2019) and Taiwan study (65/1248, 5.2%) (Huang et al., 2019) that used the same ISH method. A review of PD-L1 expression in gastric cancer (Vrána et al., 2018) showed that the percentage of PD-L1⁺ in the TCs ranges from 9% to 49.1% (most of the trials were performed in Asian populations), similar to our results (41.2% in TCs). Previous studies have demonstrated a mutually exclusive relationship between GCs that were EBV-positive and deemed to have MSI (Choi et al., 2020). However, two cases representing both MSI and EBV-positive were observed in a recent study (2/287, 0.70%) (Pereira et al., 2018) and in our study (2/226, 0.88%). Previous studies have demonstrated that MSI and EBV⁺ subtypes were separately associated with better prognoses (Sohn et al., 2017), and our result is in line with the results from Pereira et al. (2018) and De Rosa et al. (2018) that suggested MSS/EBV⁻ GCs showed the worst OS compared with MSI and EBV⁺ GCs. The potential mechanism underlying better survival of these two GC subtypes is that intense inflammatory infiltration observed in EBV-positive and MSI GCs may play an important role in antitumor effects. CD8⁺ cytotoxic T cells were found to be more prevalent in these two subgroups based on the stimulation of many neo-antigens (existing in MSI and EBV⁺ tumors), which can promote the elimination of TCs (Choi et al., 2020).

As a standard regimen, postoperative chemotherapy has been recommended for patients with stage II or III GC. Nevertheless, its application to all stage II or III GC patients is unnecessary and may even be detrimental to some patients (Jiang et al., 2017). Among these, GCs with MSI may experience a detrimental effect from chemotherapy (especially for 5-FU based treatment), and this finding has been supported by several studies (Kim et al., 2015). Our study yielded a similar result, namely no significant benefit from postoperative chemotherapy (5-FU-based) was observed for MSI and EBV⁺ GCs. In this regard, cancer immunotherapy might offer a promising new option and has been demonstrated effective in MSI and EBV⁺ GCs. A recent phase II trial described the high overall response rates in MSI-H and EBV-positive GCs with pembrolizumab treatment in metastatic GCs (ORR: 85.7% for MSI-H and 100.0% for EBV-positive) (Kim et al., 2018). A possible explanation is that both MSI and EBV⁺ GCs are characterized by elevated mutation rates and high expression of PD-L1 and PD-L2, which can be used as possible targets for immunotherapy (Sohn et al., 2017), especially PD-1/PD-L1 directed therapy (Le et al., 2017). On the other hand, our results indicated that MSS/EBV⁻ GCs, which was associated with a poorer prognosis, showed improved OS with the administration of chemotherapy after surgery.

In our study, we observed that both EBV-positive and MSI GCs showed a relatively higher proportion of PD-L1⁺ expression than MSS/EBV⁻ GCs. This result is consistent with reports in previous studies (Angell et al., 2019), which can be explained by observed amplification of the PD-L1 gene in EBV-positive GCs by TCGA study (Cancer Genome Atlas Research Network. 2014) and adaptive immune resistance in the tumor microenvironment. Intratumor lymphocytes can release interferon-γ to induce the upregulation of PD-L1 in TCs (Pereira et al., 2018). However, the positive expression rate of PD-L1 in MSS/EBV⁻ TCs varies greatly in different studies though there is few relevant studies, which could be attributed to the various method of PD-L1 expression determination and cut-off value of calling positive status used among different studies. In the study of De Rosa et al. (2018), only 3% PD-L1⁺ cases were found in MSS/EBV⁻ subgroup and a case was considered positive if membrane immunostaining was present in at least 5% of tumor cells, independently of the intensity. Choi et al. (2020) found 14% PD-L1-positive cases in MSS/EBV⁻ when considering both staining intensity and percentage of stained tumor cells, and any membranous staining was regarded as positive expression. In the present study, a higher proportion of PD-L1⁺ cases in MSS/EBV⁻ group (35%) are reported when scoring PD-L1 expression on the basis of both the intensity and percentage of PD-L1-positive TCs and considering both cell membrane and cytoplasm staining. We combined molecular subtypes with PD-L1 expression and found that PD-L1⁺ can predict a better OS in MSS/EBV⁻ subgroup but not in the other two subgroups. In contrast to our results, De Rosa et al. (2018) reported that PD-L1⁺ cannot predict the prognosis for GC patients with MSS/EBV⁻; however, the reliability of their result is doubtful due to a very limited number of cases with PD-L1⁺ (n = 2) in the MSS/EBV⁻ subgroup (n = 68). PD-L1 expression in TCs might be attributed to the response to endogenous host immunity (Kim et al., 2016). For the MSS/EBV⁻ subtype with a lower immune infiltration, the expression of PD-L1 might suggest the activation of the immune response in the tumor microenvironment, indicating a better prognosis.

Although MSS/EBV⁻ GCs can benefit from chemotherapy, it is still unclear whether there are new markers, especially PD-L1, that could subgroup them and better guide the selection of chemotherapy or immunotherapy. In the current study, chemotherapy could improve the prognosis for PD-L1-negative MSS/EBV⁻ GCs but not for PD-L1-positive MSS/EBV⁻ GCs. One possible explanation is that the survival benefit of PD-L1⁺ in this subtype could be attenuated by the immunosuppression caused by chemotherapy toxicity (Choi et al., 2019). Thus, for MSS/EBV⁻ GCs with PD-L1⁻, with an obvious benefit from chemotherapy, chemotherapy after surgery should be recommend for GCs of this subgroup. Even though MSS/EBV⁻ GCs with PD-L1⁺ exhibit better OS, we cannot give up evaluating the potential use of immune checkpoint inhibitors targeted at PD-L1 for this distinct subgroup of GCs considering the low efficacy of chemotherapy.

Therefore, we suggest that stratification by MSI and EBV statuses combined with PD-L1 expression should be used as a potential strategy to predict prognosis and identify GC patients who are most likely to respond to chemotherapy after surgery. This predictive strategy has the potential to guide patient subtype-oriented therapeutic decisions and avoid treating a certain proportion of resectable GC patients with unnecessary treatment.

Some limitations in the present study should be addressed. Firstly, only 13 patients were identified as EBV⁺ due to the low incidence of EBV-associated GC among Chinese people, and this could lead to a low level of statistical efficacy in EBV subgroup analysis. Further study with more patients is needed to verify our results. Secondly, the median follow-up time (39.7 months) was relatively short in our study. Clinical outcomes were not observed in 42% patients, and longer follow-up was needed. With the development of clinical practice, there have been PD-L1 test kit (PD-L1 IHC 22C3 pharmDx) approved by FDA as a companion diagnostic assay for the use of pembrolizumab in GC clinical treatment (Fashoyin-Aje et al., 2019). However, due to our earlier research, we used PD-L1 antibodies for scientific research, and the consistency between them needs further study.

Conclusions

In summary, we observed that MSS/EBV⁻ GCs were associated with the worst OS compared to that of GC patients with MSI and EBV⁺ subtypes. PD-L1 expression was associated with the MSI and EBV⁺ subtypes. MSS/EBV⁻ GCs could also gain a survival benefit from postoperative chemotherapy. On the other hand, PD-L1 positivity could predict a better prognosis for MSS/EBV⁻ GCs, whereas MSS/EBV⁻ GCs with PD-L1⁺ cannot gain survival benefit from chemotherapy. Thus, molecular typing plus PD-L1 expression could improve prognostic prediction for GC patients and help choose more effective treatments for specific subsets of GCs.

Supplemental Information