The 12-chemokine TLS signature score was positively associated with an enhanced CIC and elevated immune cell infiltration.
A 12-chemokine RNA signature, originally identified as a metagene signature related to inflammation and tumor-localized ectopic TLS presence (referred to as the TLS signature), was comprehensively analyzed in TCGA pancancer datasets. The 12 genes encoding chemokines were commonly dysregulated in various cancer types compared to the respective normal controls (Fig. 1a). We delineated the IGSs of each CIC step within different subgroups when tumor samples were classified according to the signature levels (Fig. 1b). A strong correlation between the 12-chemokine TLS signature and all steps of the CIC was observed, including T cell immunity, dendritic cell (DC) enrichment, and expression of checkpoint molecules (Fig. 1b-c, Fig. S1-2). On the other hand, we observed different patterns of the impact of the TLS signature score on genomic correlates (Fig. 1d). In bladder urothelial carcinoma (BLCA), skin cutaneous melanoma (SKCM), colon adenocarcinoma (COAD), etc., the TLS signature score was positively correlated with TMB and TNB, while contradictory findings were observed in adrenocortical carcinoma (ACC) and thyroid carcinoma (THCA). We next analyzed bladder cancer and melanoma in depth, as a high TLS signature score may reflect a TNB-high/immune-inflamed subtype in these tumors, suggesting that the signature may be capable of playing a predictive/prognostic role for immunotherapy.
A high TLS signature score was a favorable prognostic factor in the TCGA BLCA dataset.
First, we explored the phenotypic differences between TLS signature score-high and TLS signature score-low patients in the TCGA BLCA dataset. A weak (r = 0.13) but statistically significant (p < 0.05) positive association was identified between the TLS signature score and TMB or TNB (Fig. 2a, b). When patients were subclassified as TLS signature score high or TLS signature score low (cutoff: top 25%), we observed a strong enrichment of infiltrating immune cells and expression of immune checkpoint molecules in the TLS signature score high subgroup (Fig. 2c and Fig. S3). Moreover, the TLS signature score-high subgroup was characterized by an interferon-γ (IFN-γ)-dominant phenotype (C2 subtype, Fig. 2d and Fig. S4), accompanied by elevated BCR/TCR richness and Shannon diversity (Fig. 2e) and increased immune signature expression (Fig. 2f). Overall, a high TLS signature score indicated enhancement of multiple factors of the immune cycle (Fig. 2g), and patients within the TLS signature score-high subgroup experienced a prolonged progression-free interval (PFI, p = 0.039) and increased overall survival (OS, p = 0.055) months (Fig. 2h, i).
The TLS signature score was associated with improved clinical outcome in the IMvigor210 cohort.
Next, we investigated how the TLS signature score may impact immunotherapeutic responses in a cohort of locally advanced or metastatic urothelial carcinoma (mUC) patients receiving atezolizumab treatment. In accordance with the data from TCGA BLCA datasets, the TLS signature score was positively associated with TNB and infiltrating immune cells (Fig. 3a, b). When the samples were subclassified into two subgroups (cutoff: top 25%), TLS signature score-high samples had higher PD-L1 staining on tumor cells (TCs) or on tumor-infiltrating immune cells (ICs) (Fig. 3c) than TLS signature score-low samples, and TLS signature score-high samples had enrichment of the inflamed phenotype (Fig. 3d, e) and elevated immune signatures such as cytolytic activity, IFN-γ signaling, and MHC expression (Fig. S5). Collectively, the expanded immune cycle in TLS signature score-high patients (Fig. 3f) may have contribute to the increased disease control rate (DCR), overall response rate (ORR) and prolonged OS (Fig. 3g, h). Specifically, in the IMvigor210 mUC cohort, we observed that the DCR and ORR of the TLS signature score-high subgroup were 59% and 32%, respectively, compared with only 39% and 20% in TLS signature score-low subgroup, respectively (Fig. 3g, p < 0.05 for all comparisons). More importantly, a significant difference in OS was also observed between the TLS signature score-high and the TLS signature score-low groups (Fig. 3h, HR: 0.64; OS: 15.9 vs 7.7 m; p = 0.0017).
A high TLS signature score correlates with an enhanced CIC and predicts the therapeutic response in melanoma patients receiving immunotherapy.
A previous study showed that TLS formation confers distinct T cell phenotypes in melanoma, and a 9-gene signature associated with TLSs was subsequently derived, which predicted clinical outcomes in patients receiving immunotherapy [9]. Here, we showed that the 12-chemokine TLS signature, which was originally derived from colorectal carcinoma (CRC), is also indicative of elevated TMB, TNB, immune infiltration, and checkpoint expression (Fig. 4a-c and Fig. S6). The median score was used as the cutoff, and TLS signature score-high melanoma samples showed an enhanced CIC phenotype (Fig. 4d). An improved PFI but no an improved OS duration were observed in the TLS signature score-high subgroup in the TCGA BLCA dataset (Fig. 4e, f). More intriguingly, we then validated the clinical utility of the TLS signature for predicting immunotherapy response by using two publicly available datasets (GSE91061 and GSE35640). As expected, the 12-chemokine TLS signature score was also a predictive biomarker for melanoma patients who received anti-PD1/PD-L1 immunotherapy (Fig. 4g) or MAGE-A3 treatment (Fig. 4h). In addition, a high TLS signature score indicated strong immune infiltration and immune responses in both datasets (Fig. S7-8). However, since TNB data were not available in these two melanoma immunotherapy datasets, we were unable to determine the associations between the TLS signature score and the IGSs of the immune cycle.