Analyzing the associations between tertiary lymphoid structures and postoperative prognosis, along with immunotherapy response in gastric cancer: findings from pooled cohort studies

Gastric cancer (GC) is a major cause of cancer-related mortality globally (Sung et al. 2021), with exceptionally high disease burden in East Asia (Etemadi 2017). GC is featured by high heterogeneity at histological, cellular, (epi)genomic, and proteomic levels, accompanied by distinct clinical outcomes (Wadhwa et al. 2013; Ge et al. 2018). Despite good response to treatment in early-stage GC, advanced GC is highly aggressive, with a median overall survival (OS) time of around 10 months (Smyth et al. 2020). In recent years, immunotherapy has revolutionized the treatment landscape of advanced GC and significantly extended patients’ survival (Li et al. 2021a). However, only a limited population benefited from immunotherapy, and a wide variation of response rates was reported, calling for better biomarkers for stratified treatment (Janjigian et al. 2021; Shitara et al. 2020).

Proteomic markers such as PD-L1 expression (Kim et al. 2018) and molecular markers such as microsatellite instability (MSI) (Kwon et al. 2021) have been established as potential biomarkers of prognosis and responses to immune checkpoint inhibitors (ICIs), whereas findings between different clinical trials were inconsistent. Recently, components in the tumor microenvironment (TME) of GC, such as the contents (Ren et al. 2021) and spatial location (Chen et al. 2022) of different cell types, and specialized structures like tertiary lymphoid structure (TLS) (Yu et al. 2022), have received increasing attention for they could affect prognosis and immunotherapy efficiency.

Composed of cellular aggregates in non-lymphoid organs under inflammatory conditions like infection and tumor, TLSs show analogical functional and structural features with lymph nodes (Schumacher and Thommen 2022). Mature TLS is characterized by the B-cell zone that involves the germinal center and is surrounded by the follicular helper T cells. The T-cell zone containing dendritic cells and high endothelial venules is also crucial to TLS (Sautès-Fridman et al. 2019). TLSs signify privileged regions for immune cell maturation and antigen presentation, serving as the crucial milieu for anti-tumor immunity. Emerging evidence indicated that TLS presence strongly correlated with higher immunoreactivity and better clinical outcomes of GC. For instance, Li and colleagues reported that TLS presence was indicative of favorable OS based on a cohort containing 63 GC cases (Li et al. 2020), and Yin et al. further showed that TLS was a promising predictor for longer survival of Epstein-Barr Virus (EBV)-associated GC (Yin et al. 2022). Moreover, Jiang et al. proposed that TLS positively correlated with superior response of ICIs based on a cohort containing 13 GC samples (Jiang et al. 2022). However, some studies were limited by small sample sizes or specific GC subtypes, thus may lack generalizability to some extent. Moreover, associations between TLS and prognosis and therapeutic sensitivity of GC are controversial, probably because samples from different ethnicities and different detecting approaches for TLS were applied.

Consequently, we conducted a meta-analysis to clarify the prognostic and predictive values of TLSs in GC. Simultaneously, we performed bioinformatic validation to capture the biological underpinnings by TLS-related gene signature in external cohorts. Our comprehensive analyses provided the latest evidence for the relationships between TLS and GC, probably conveying a powerful biomarker for clinical practice.

Through a comprehensive meta-analysis of 11 studies containing 4,224 GC cases, we pinpointed that TLS correlated with favorable prognosis and ICIs sensitivity of GC. Biological validation in the TCGA-STAD and PRJEB25780 cohorts further corroborated that TLS presence signified higher immunoactivity in TME.

The prognosis of GC is known to be affected by tumor- and host-correlated characteristics, such as age, pTNM stage, and histologic subtypes. Consequently, we first interrogated the relationships between TLS levels and clinicopathologic features. Results showed that higher TLS levels correlated with smaller tumor size and earlier T and N stages, consistent with the findings in breast cancer (BC) (Wang et al. 2022) and lung cancer (Rakaee et al. 2021). Interestingly, higher TLS levels were discovered in diffuse and mix than intestinal subtypes of GC. Since TLS is distinguished by B-cell enriched regions, a recent single-cell atlas also documented significantly higher B-cell infiltration as a salient feature of DGC (Kumar et al. 2022). Wang et al. also reported that TLS presence predicted higher tumor-infiltrating lymphocyte (TIL) levels of BC by meta-analysis (Wang et al. 2022). Nonetheless, we could not assess whether TLS levels correlate with higher infiltration of TILs in meta-analysis due to the nature of the data.

Systematical meta-analysis indicated that higher TLS levels strongly predicted favorable OS of postoperative GC, with HR of 0.36 and 0.55 of univariate and multivariate Cox analysis, respectively. Sensitivity analyses further validated the stability and robustness of our findings. Noteworthy, high heterogeneity was found between the included studies. Thus, subgroup analyses stratified by the TLS detecting methods, sample size, and median age of the included cases were carried out to interrogate potential heterogeneity. Strikingly, the prognostic value of TLS remained significant across subgroups and was more notable in younger than elder GC cases. The heterogeneity decreased as expected, whereas it remained modest. Moreover, meta-analysis implicated that GC patients with high TLS levels significantly benefited from anti-PD-1 inhibitors as later-line therapy than those with low TLS levels. However, considering the limited included cases and retrospective design of studies, the predictive effects of TLS on ICIs sensitivity need to be interpreted cautiously.

The maturation degree of TLS is postulated to affect its clinical significance by recent work. For instance, mature TLS with GC predicted significantly longer survival than total TLSs, whereas such prognostic effect attenuated when the formation of GC was damaged (He et al. 2023; Ling et al. 2022). Meanwhile, the prognostic value of TLS varied by its location in the tumor tissue. Intratumor TLS seemingly demonstrated a stronger prognostic effect than peritumor TLS, whereas findings were inconsistent between different cancer types (Sofopoulos et al. 2019; Li et al. 2021b). There is evidence that TLS levels remarkably varied from early to advanced stage of cancer (Sautès-Fridman et al. 2019). Additionally, studies showed that TLS levels significantly attenuate in the metastatic sites compared to the primary tumors, and can even be absent (Lee et al. 2019). And patients who hold TLS both in the primary tumor and metastatic site exhibit superior prognosis (Cipponi et al. 2012). However, we could not perform comprehensive subgroup analyses on the maturation degree or spatial location of TLS due to the nature of data. Therefore, future studies characterizing TLS with compositional, spatial, and functional details are imperative and encouraged.

Considering the unique puzzles in identifying and quantifying TLSs via traditional methods like H&E staining and IHC, we further validated the findings from meta-analysis and discovered the underlying biological underpinnings via TLS-relevant gene signature, which has been verified in multiple cancer types (Cabrita et al. 2020; Hou et al. 2022; Feng et al. 2021). Strikingly, higher TLS levels were found in DGC than in other subtypes, in congruence with the meta-analysis. Despite without significance, a trend of better prognosis was also observed in DGC individuals with higher TLS levels. Moreover, TLS level was proven to be an independent prognostic factor of intestinal GC.

The TME landscapes concerning TLS levels were subsequently depicted. Higher immune infiltration of major immune effector cells, such as T and B lymphocytes and natural killer cells, were observed in the high TLS-level group than in low ones, indicative of an “immune-hot” TME. Future studies integrating TLS levels and immune infiltration features within TME may offer a more comprehensive and robust prognosticator. Interestingly, higher mast cell (MC) infiltrates were found in the TLS-low group. Evidence showed that MCs could stimulate regulatory T cells to facilitate GC progression (Lv et al. 2023). Studies also reported that inhibition of the degranulation of MC attenuated the development of GC, signifying a potential target (Gunjigake et al. 2021). We also parsed the relationships between the TLS levels and the diversity and richness of the immune repertoire, which represents the strength and breadth of immune responses and acts a paramount role in anti-tumor immunity (Jiang et al. 2019). Higher diversity and richness of TCR and BCR were discovered in the TLS-high group, representing higher antigen presentation function.

Higher TLS levels also correlated with upregulated immune checkpoint genes expression. This may also partly explain why ICIs boost strong antitumor immunity in cancers with enriched TLSs (Petitprez et al. 2020). Intriguingly, ICIs could also instigate TLS formation. For instance, Sarah et al. reported the accumulation of TLS-correlated B cells in responders after neoadjuvant ICIs of melanoma (Helmink et al. 2020). Moreover, the genetic portraits significantly differed between different TLS-level groups. Elevated mutational frequencies of several genes that correlated with immune infiltration were observed in the TLS-high group. For example, a higher mutation rate of ARID1A, a tumor suppressor gene that is relevant to the MSI feature of cancers (Mullen et al. 2021), was found in the TLS-high group. ARID1A-mutated GC held higher TMB and PDL1 levels and favored higher immune cell infiltrates (Li et al. 2019). In brief, high TLS levels represent high immunogenicity and immunoactivity, possibly driving benefits from immunotherapy.

Eventually, we interrogated whether TLS levels could predict ICIs response in the PRJEB25780 cohort, in which GC patients received later-line pembrolizumab monotherapy. Intriguingly, more responders were identified in the TLS-high group than the low ones. Moderate accuracy in predicting benefit from ICIs of TLS-level was presented (AUC > 0.75), higher than the MSI and EBV-status, which are established biomarkers indicative of immunotherapy response (Bai et al. 2022; Yu et al. 2022).

The present work firstly and comprehensively offered substantial evidence for the clinical significance of TLS in GC by meta-analysis and biological validation. Meanwhile, several limitations should be noted. First, different scoring approaches and thresholds in evaluating high/low TLS levels were utilized in different studies. However, we could not perform corresponding subgroup analyses due to unavailable data, probably leading to bias. Second, the pooled sample size for discovering and validating the predictive effect of TLS on ICIs response was limited, thus may lack robustness. Third, high heterogeneity among studies caused potential publication bias in the meta-analysis. Additionally, all the included researches were retrospectively investigated and may risk intrinsic structural biases. Moreover, findings concerning the biological underpinnings of TLS were still at the speculative and analytic stage based on gene signature, without in vivo and in vitro functional validation.

Future studies should focus on establishing a common standard for identifying and quantifying TLS and future prospectively validating it in randomized trials for better clinical applications. Second, pinpointing dynamic changes of cellular components and location within TLS during immunotherapy leads to better comprehending its biological implications. Moreover, inducing the formation of TLS, like by intratumoral injection of vital cytokines such as CXCL13 (Delvecchio et al. 2021), administration of engineered cells (GeurtsvanKessel et al. 2009) and tumor vaccines (Zhang et al. 2021), may provide a neoteric perspective for synergistic immunotherapeutic method.

In brief, higher TLS levels positively correlate with higher immunogenicity and immunoactivity in TME, demonstrating potential in predicting postoperative prognosis and immunotherapy response of GC. Future studies with prospective design are needed to validate the clinical significance of TLS, individually or jointly with other markers, across different cancer types.