Heterogeneity and distribution characteristics of tertiary lymphoid structures predict prognostic outcome in esophageal squamous cell carcinoma

By screening the pathological information of esophageal cancer patients hospitalized at Henan Cancer Hospital (China) from January 2015 to December 2022, and combining follow-up data, and the ESCC cohort was analyzed for heterogeneity (n=117) and prognostic assessment (n=112). The case screening process and research design are shown in Figure 1. This study was conducted in accordance with the Declaration of Helsinki and received approval from the Ethics Committee of Henan Cancer Hospital (Ethics Approval No: 2021-KY-0092–001). Written informed consent was obtained from all participants involved in the research.

By consulting the medical records, we confirmed the clinical and pathological data of qualified cases (including the age of first diagnosis, gender, treatment plan and therapeutic effect, etc.) and made statistical analysis. The patients were mainly divided into two cohorts: surgery group and neoadjuvant treatment group. Surgery cohort refers to cases that do not receive neoadjuvant therapy. The neoadjuvant cohort refers to those who received chemotherapy drugs and immunological drugs before surgery.

The paraffin-embedded ESCC tissue specimens were cut into 5 μm thick continuous sections and baked in the oven for 2 h at 65°C. The tissue sections were successively transferred to xylene and alcohol for dewaxing to water; Hematoxylin solution was stained for 5 min and differentiated into 1% hydrochloric acid and alcohol. Water back blue after eosin staining; Transfer to different concentrations of alcohol (75%, 85%, 95%, 100%) and xylene in order to dehydrate transparent, neutral gum tablets. The results of HE staining were observed by optical microscope.

Two pathologists who were unaware of the patient’s clinical information and prognosis observed the HE sections with a microscope to assess whether TLS was present in the tumor, and counts the number of TLSs. The number of TLSs in each section was also recorded for peritumoral (5 mm area around tumor tissue) and incisal margin. Qupath software was used to scan HE slices, and TLSs (presence/quantity) was determined twice according to the preliminary screening results. For TLS positive cases, the diameter of the largest TLS and the tumor area and peri-tumor area were measured. Tumor area (T area), incisal margin area (N area), peritumoral area (P area). Here, TG, TA, and TT represent the presence of GC, Aggregates (AGG) and total TLS in the tumor region, respectively. PG, PA, and PT represent the presence of GC, AGG, and total TLS in the peritumoral area, respectively. NG, NA, and NT represent the presence of GC, AGG, and total TLS in the incisal margin area, respectively. We analyzed the TLS distribution and density of all tumor wax blocks (four-eight) in each case, and analyzed the spatial heterogeneity among different sections.

Tumor infiltrating lymphocytes (TILs) include intratumoral tumor infiltrating lymphocytes (iTIL) and stromal tumor infiltrating lymphocytes (sTIL) in the stroma that are present in the tumor’s cancer nest. The clinical information was unknown by two independent pathologists in advance. The tumor-stroma ratio (TSR) 50% was taken as the truncation value. TSR<50% and TSR≥50% were defined as the interstitial rich group and the interstitial sparse group, respectively, so as to compare the prognosis difference between the two groups.

Multiplex immunofluorescence staining was performed on formalin-fixed, FFPE tissue sections to assess TLS-related immune cell distribution in both tumor and peritumoral areas. Briefly, 4-μm-thick sections were deparaffinized, rehydrated, and subjected to antigen retrieval using citrate buffer (pH 6.0) in a pressure cooker. Endogenous peroxidase activity was blocked with 3% hydrogen peroxide, followed by incubation in 5% BSA blocking buffer. Primary antibodies were applied sequentially using tyramide signal amplification (TSA), including CD3 (FITC, 520 nm), CD8 (Texas Red, 620 nm), CD20 (Cy5, 690 nm), and pan-cytokeratin (TRITC, 570 nm), with DAPI used for nuclear counterstaining. Each antibody was followed by HRP-conjugated secondary antibody and TSA fluorophore, with microwave treatment for antibody stripping between rounds. Stained slides were scanned using a fluorescence microscope with appropriate filters, and images were analyzed using image processing software.

SPSS 25.0 software was used for statistical analysis. Continuous variables are presented as mean ± standard error (SE) or median (interquartile range, IQR), and categorical variables are expressed as frequency (percentage). Univariate and multivariate analyses were performed using Cox proportional hazards regression models, with hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) calculated. The data were compared between groups using the χ 2 test and paired t- test, as appropriate. Survival curves were generated by the Kaplan-Meier method and compared using log-rank tests. All statistical tests were two-sided, with p values <0.05 considered statistically significant.

In the surgery cohort, the median overall survival (OS) and disease-free survival (DFS) were 33 months and 15 months, respectively (n=112). Tumor staging analysis revealed that most patients presented with advanced disease: T3 (51.8%) was the most common, followed by T2 (36.6%), T1 (9.8%), and T4 (1.8%). In terms of lymph node involvement, 48.2% were classified as N0, 33.9% as N1, and 17.9% as N2 (Table 1). In the neoadjuvant cohort, the number of cases with Tumor regression grade(TRG)classification 0, 1, 2, 3 were 9, 6, 11, 3, respectively. More than two-thirds of patients were men (72.3% in the surgery cohort and 75.8% in the validation neoadjuvant cohort). The median age was 66 years (range: 50–79 years) in the surgery cohort and 63 years (range: 47–74 years) in the neoadjuvant cohort. Univariate analysis identified several factors significantly associated with overall survival (OS), including T stage (p = 0.027), N stage (p = 0.002), tumor grade (TG, p = 0.026), tumor regression (TR, p = 0.008), necrosis ratio (NR, p = 0.043), tumor-stroma ratio (TSR, p = 0.006), intratumoral tumor-infiltrating lymphocytes (iTIL, p < 0.001), and stromal TILs (sTIL, p < 0.001). However, in the multivariate analysis, T stage lost its statistical significance (p = 0.537), suggesting it may be confounded by other variables. Independent predictors of OS included N stage, TG, TR, necrosis area (NA), necrosis grade (NG), necrosis type (NT), TSR, iTIL, and sTIL (Table 1). Additionally, the analysis results for DFS showed that in the univariate analysis, T stage (p = 0.027), PA (p = 0.024), PG (p = 0.047), and PT (p = 0.006) were all significantly associated with DFS. In the multivariate analysis, only PA (p = 0.044) was identified as an independent predictive factor (Table 2). The analysis results indicate that the factors influencing OS and DFS are not entirely the same.

These results highlight that while several pathological and immunological features are associated with survival outcomes, the factors independently predicting OS and DFS differ. OS appears to be more strongly influenced by tumor biology and immune microenvironment, whereas DFS is more closely linked to features of perineural invasion.

The spatial distribution of tertiary lymphoid structures (TLSs) in esophageal cancer varied across tumor (T), peritumoral (P), and nodal (N) regions. As illustrated in Figure 2A, TLS were observed in distinct spatial compartments, including intra-tumoral, peri-tumoral margin, and extra-tumoral (stromal) regions. Representative histological features of TLSs, including GCs and AGGs, are shown in Figure 2A. In the surgery cohort, varying numbers of GC and AGG were observed in the T, P, N regions. The HE morphology of TLS is illustrated in Figure 2B. Within the tumor region, 106 patients (90.6%) were TLS-positive, among which 24 (20.5%) and 105 (89.7%) exhibited GC and AGG, respectively (Figure 2C). In the peritumoral region, 116 cases (99.1%) were TLS-positive, with 55 cases (47.0%) showing GC structures (Figure 2D). At the tumor margin, the numbers of GC-positive, AGG-positive, and TLS-positive cases were 55 (47.0%), 86 (73.5%), and 102 (87.2%), respectively (Figure 2E). TLS distribution was also assessed before and after neoadjuvant treatment, with results summarized in Figures 2F–J. In the T regions before and after treatment, GC structures were identified in 3 cases each, and AGGs in 11 and 13 cases, respectively. In the P region, GCs and AGGs were found in 4 and 15 cases, respectively. In the N region, 4 patients had GCs, and 8 had AGGs.

These results demonstrate that AGGs are consistently more frequent than GCs across all spatial regions and treatment phases. Furthermore, TLSs remain prevalent even after neoadjuvant therapy, indicating a potentially persistent immune response within the tumor microenvironment.

TLSs, including GCs and AGGs, demonstrated notable spatial heterogeneity across T, P, and N regions. Figures 3A–C show representative HE-stained sections from 117 ESCC patients, illustrating the detection of GC, AGG, and TLS. Assessment of different paraffin blocks from the same tumor revealed uneven distribution patterns.

The proportion of cases with complete positivity (all four blocks positive) was low in the T region (GC: 0%; AGG: 50.0%; TLS: 50.0%), but relatively higher in the P and N regions (e.g., TLS: 76.5% in P) (Figures 3D-F). Figures 3G–I further illustrate the degree of spatial consistency, defined as either all positive or all negative across blocks. GC distribution was completely consistent in 79.5% of T region samples. AGG and TLS were most stable in the P region, with consistency rates of 75.2% and 78.6%, respectively. These findings highlight the region-specific variability of TLSs and underscore the need for multi-site sampling to accurately assess the immune microenvironment in ESCC.

In addition to analyzing TLSs, we extended our investigation to TILs to provide a more comprehensive view of immune infiltration in ESCC. TLSs offer localized information, while TILs reveal the broader spatial distribution and intensity of immune cell presence across tumor tissues. Figure 4A demonstrates HE staining images showing the different proportions of iTIL, sTIL, and TSR in tumor tissues. And in the ESCC surgery cohort, iTIL mostly account for 1%-10% (Figures 4B). sTIL are divided into four fractional regions: 0-20%, 21-40%, 41-60%, and 61-80%, with case numbers of 33 (28.2%), 47 (40.2%), 26 (22.2%), and 11 (9.4%), respectively (Figures 4C). TSR was ≥50% in 113 patients (96.6%), while only 4 patients (3.4%) had a TSR <50% (Figures 4D). These findings suggest a dominant stromal component in ESCC and indicate that iTIL infiltration is generally modest, whereas sTIL presence is more heterogeneous.

In addition to TLSs, the spatial distribution and stability of TILs were evaluated. Figures 5A–C display HE staining images showing patterns of iTIL, sTIL, and TSR across multiple blocks from the surgery cohort (n = 117). Most patients exhibited a TSR ≥ 50% (88.9%) and low iTIL levels (1-10%) in 92.3% of cases. In contrast, sTIL expression showed greater variability among different tumor sections.

Consistency analysis (Figure 5D) revealed high spatial stability for TSR (91.5%) and iTIL (88.9%), while sTIL demonstrated significantly lower consistency (40.2%), indicating a more heterogeneous stromal immune distribution. Further insights were gained from the neoadjuvant cohort. By comparing pre-treatment biopsy samples with post-treatment surgical tissues, spatial heterogeneity was again observed. Among 15 cases with marked tumor regression (TRG 0–1), TLSs remained detectable post-treatment in 46.7% of T regions, 80.0% of P regions, and 53.3% of N regions. The consistent presence of TLSs across multiple tissue blocks in these cases (Figures 5E–F) reinforces the dynamic and spatially variable nature of the immune response following therapy.

The presence and spatial distribution of TLSs within tumor tissues exhibit distinct prognostic implications in ESCC. KM survival analysis revealed that high presence of TG was significantly associated with longer OS in ESCC patients (log-rank p < 0.05) (Figures 6A, B). In contrast, the presence of TLSs in the peritumoral area showed no significant correlation with OS. However, high presence of perineural area (PA), grade (PG), and type (PT) was significantly associated with prolonged DFS (Figures 6C–H). These findings indicate that spatially resolved TLS evaluation provides valuable prognostic information beyond traditional pathological parameters. In addition, multiplex immunofluorescence staining results were consistent with the IHC findings. In the tumor region, TLS structures were characterized by aggregated CD3⁺ T cells and scattered CD20⁺ B cells. CD8⁺ cytotoxic T cells were frequently found within the CD3⁺ T-cell zones (Figure 6I). In the peritumoral region, TLS were segregated CD3⁺ T-cell and CD20⁺ B-cell zones, suggesting the presence of more mature TLS, including germinal center-like structures (Figure 6I). These findings confirm the presence and spatial organization of TLS-associated immune cells observed by IHC.

Building on conventional pathological diagnostics, the evaluation of TLS related indicators can provide more diagnostic and therapeutic references for the prognosis assessment of ESCC. In the future, by integrating multi-omics data (such as genomics, proteomics, and metabolomics) and predictive models using artificial intelligence, the diagnostic accuracy and treatment outcomes for ESCC can be further enhanced. This comprehensive analytical approach enables doctors to more precisely understand the biological behavior of tumors, offering more personalized and accurate treatment plans for patients (Figure 7).

The original contributions presented in the study are included in the article/. Further inquiries can be directed to the corresponding author/s. 1