A comprehension of the genes connected to TLS molecular subtyping

BRCA patients were subtyped into two clusters, C1 and C2, using PAM and nine TLS-specific genes, with C1 comprising 470 cases and C2 comprising 614 cases, in the TCGA dataset. C1 exhibited significantly higher expression of TLS-specific genes and notable differences included: (1) enhanced ImmPort pathway activities such as antigen presentation, and TCR and BCR signaling; (2) increased immune infiltration of T cells, B cells, and DCs; and (3) greater expression of immune checkpoints such as PD-1, PD-L1, and CTLA-4 (Fig. 1A). UMAP analysis clearly separated the clusters (Figure S1A) and showed prolonged overall survival in C1 (Figure S1B). Expression of TLS-specific genes was notably higher in C1 (Figure S1C), and enrichment analysis of upregulated DEGs in C1 identified activation of immunological pathways related to cytokines, chemokines, and interferons (Figure S1D). Six ML algorithms for classification, including Pamr, RF, SVM, LassoLR, XGBoost, and Boruta, identified ten intersecting genes critical for TLS subtyping (Fig. 1B).

Fig. 1

Identification of TLS-associated tumor suppressor LGALS2

(A) Heatmap showed the distribution of nine TLS-specific genes, ImmPort-based immune-related pathways, TIMER-based immune cells, and immune checkpoints in two TLS-related clusters. (B) Venn plot showed the intersected genes among the DEGs between two TLS-related clusters optimized by six ML algorithms for classification. (C) Heatmap showed the relationships between gene modules and TLS phenotype. (D) Correlation plot showed the brown module’s correlation between module membership and gene significance for TLS. (E) PPIRWR procedure for determining the TLS-associated gene ranking. (F) Petal chart showed the intersected genes that were significantly positively associated with each of the nine immunotherapeutic signatures from the top 10% of the TLS-associated gene ranking according to PPIRWR. (G) Venn plot showed the three intersected best genes in the context of TLS through molecular subtyping, WGCNA, and PPIRWR. (H) The univariate Cox regression analysis on the three intersected best TLS-related genes. (I) Venn plot showed the intersected gene LGALS2 was the final emerging gene among the three genes with associations with TLS by three ML algorithms for survival. (J) Circos plot showed the KEGG pathways correlated with LGALS2 and the distribution of genes enriched in pathways in human chromosomes. (K) GSEA-based GO immunological pathways for LGALS2. (L) Radar plot showed the correlation between LGALS2 and cancer immunogram traits. (M) Bar plot showed the correlation between LGALS2 and ImmPort-based immunological pathways

WGCNA discerning the TLS-attached gene module

Using WGCNA and the ssGSEA algorithm based on nine TLS-specific genes, ten gene modules were discovered, with the brown gene module being identified as having the highest correlation (R = 0.69) with TLS, as shown in Fig. 1C. The brown module exhibited a high correlation (R = 0.82) between module membership and gene significance for TLS (Fig. 1D). This study also linked the brown module to PROGENy-based oncological pathways such as JAK-STAT, NF-κB, TNF-α, and TRAIL, showing strong positive correlations with TLS (Figure S2A). Gene characteristics within the brown module were further detailed, including expression levels and interactions (Figure S2B).

Topology-based filtering of genes dependent on TLS by PPIRWR

The PPIRWR procedure for determining the TLS-associated gene ranking is shown in Fig. 1E. Using the gene ranking vector, GSEA assessed KEGG oncological (Figure S3A) and immunological (Figure S3B) pathways, highlighting enrichment in pathways such as BRCA, P53/TGF-β, and PD-L1/PD-1 checkpoints. A significant correlation was found among nine immunotherapeutic signatures, including CYT, IFNγIS, AyersExpIS, GEP, RohIS, DavoliIS, chemokineIS, RIR, and ImmuneScore, demonstrating their reliability for predicting immunotherapeutic responses, as shown in Figure S3C. From the top 10% of PPIRWR-ranked genes, 242 were selected as the final PPIRWR TLS-associated genes, that positively correlated with each immunotherapeutic signature (Fig. 1F).

TLS-associated tumor suppressor LGALS2 identified

CLEC10A, CD79B, and LGALS2 were identified as significant in the context of TLS through molecular subtyping, WGCNA, and PPIRWR (Fig. 1G). Univariate Cox regression analysis confirmed their significance (Fig. 1H). Then three ML algorithms for survival were applied. LassoCox optimized LGALS2 alone (Figure S4A), while CoxBoost and RSF evaluated all three genes (Figures S4B and S4C), solidifying LGALS2 as a significant protective prognostic gene (Fig. 1I and S4D). High LGALS2 expression correlated with prolonged survival in several independent datasets including METABRIC, GSE10309, and GSE96058 in GEO (Fig. 2A), with declining expression at advanced tumor stages indicating its tumor-suppressing nature (Figure S4E).

Fig. 2

Immunological features of LGALS2

(A) The univariate Cox regression analysis on LGALS2 in four independent BRCA datasets. (B) Heatmap showed the correlation between LGALS2 and infiltration of immune cells according to ESTIMATE, MCPcounter, Pornpimol-ssGSEA, and TIMER. (C) Box plot showed the levels of the nine immunotherapeutic signatures in two LGALS2-stratified groups. (D) UMAP plot showed the distribution of 17 microenvironment cell types. (E) UMAP plot showed the expression pattern of LGALS2 in 17 microenvironment cell types. (F) Violin plot showed the expression pattern of LGALS2 in 17 microenvironment cell types. (G) Correlation plot showed the correlation between LGALS2 and CD86 in DCs. (H) The reconstruction of the pseudo-time trajectory of DCs. (I) The expression pattern of LGALS2 in the pseudo-time trajectory of DCs. (J) The correlation between LGALS2 and the pseudo-time of DCs. (K) The correlation between LGALS2 and the differentiation potential of DCs. (L) The cell communication pattern of LGALS2 + and LGALS2- DCs in terms of the MHC-I signaling pathway. (M) Heatmap showed the correlation between LGALS2 and GO immunological pathways

LGALS2 biological functional annotation

KEGG enrichment analysis strongly associated many important immunological pathways, including cytokine and chemokine signaling, T and NK cell activities, and checkpoints, with LGALS2 (Fig. 1J). GSEA-GO highlighted the involvement of LGALS2 in enhancing immunological processes related to T cells, B cells, DC, and IFNγ signaling (Fig. 1K and S5A). LGALS2-associated DEGs showed close ties to adaptive immune response and leukocyte activation based on Metascape (Figure S5B). High LGALS2 expression in BRCA patients also upregulated immunogram traits (Fig. 1L). The correlation of LGALS2 with ImmPort pathways was remarkable (Fig. 1M). The distribution patterns of Fges in five major categories and 12 minor categories suggested that LGALS2 significantly enhances immunity (Figures S5C and S5D).

Immunological features of LGALS2

Analysis of the role of LGALS2 in the BRCA cancer immune cycle revealed that it enhances several anti-tumor immune steps (Figure S6A). High LGALS2 expression in BRCA patients correlates with enriched immune cells such as B cells, various T cells, and activated DCs, indicating its key role in the immune microenvironment according to ESTIMATE, MCPcounter, Pornpimol-ssGSEA, and TIMER (Fig. 2B). Moreover, its interaction with immunomodulators (Figure S6B) and alignment with immunotherapeutic signatures (Fig. 2C) underscores LGALS2’s potential as an effective immunotherapeutic biomarker.

Positions of LGALS2 in the microenvironment at the scRNA-seq and stRNA-seq level

From Fig. 2D, the 17 primary cell types were classified in the scRNA-seq of BRCA. LGALS2’s expression patterns, as shown in Fig. 2E, showed high levels in myeloid cells, especially DCs which had the highest expression (Fig. 2F). This led to LGALS2 being identified as a DC marker, which was also confirmed by the scRNA-seq data of HCC (Figure S7A) and NSCLC (Figure S7B). The stRNA-seq of BRCA revealed that DC enrichment scores align with the expression patterns and spatial distribution of LGALS2 (Figure S7C and S7D). It was also confirmed by its strong association with the DC marker CD86 in DCs (Fig. 2G). We explored whether LGALS2 serves as a mature DC marker, noting increased LGALS2 activity over pseudo-time (Fig. 2H-J). CytoTRACE analysis showed a negative correlation between LGALS2 and DC differentiation potential, establishing its effectiveness as a mature DC marker (Fig. 2K). Moreover, mature DCs with LGALS2 showed enhanced interactions with CD8 + T cells compared to their immature counterparts (Fig. 2L). Functional analysis via GO and AUCell revealed that DCs with high LGALS2 levels resemble mature DCs, involved in key immune processes related to DCs (Fig. 2M).

Co-localization of LGALS2 and TLS-associated DCs

TLS distribution in BRCA samples was assessed using HE staining to identify lymphoid structures. Subsequent IHC staining targeted CD20 for B cells and CD3 for T cells in serial sections, mapping their proximity within TLSs. Additional IHC staining for CD86 and LGALS2 on consecutive sections highlighted the colocalization of LGALS2 with TLS-associated DCs, suggesting LGALS2’s role in DC-mediated immune responses in TLSs. Interestingly, LGALS2 showed low expression in tumor cells surrounded with TLSs, consistent with the findings in the scRNA-seq analysis (Figure S8).

In vivo validation of LGALS2

In vivo validation was performed to elucidate the immunoregulatory roles of LGALS2 in BRCA. The mRNA and protein expression of LGALS2 was significantly reduced in primary mouse DCs in the sh-LGALS2 group (Figures S9A and S9B). The tumor weight (Figures S9C and S9D) and volume (Figures S9C and S9E) were significantly increased in the C57BL/6 mice with an injection of LGALS2-suppressed DCs. Besides, the proportion of CD3 + CD4+ (Figure S9F), CD3 + CD8+ (Figure S9G), and CD8 + IFNγ+ (Figure S9H) was significantly reduced in the sh-LGALS2 group (Figure S9J), while CD8 + PD-1+ (Figure S9I) T cells was significantly increased in the sh-LGALS2 group (Figure S9J).