Tertiary lymphoid structures are associated with lower axillary residual nodal burden in breast cancer patients after neoadjuvant chemotherapy Authors: Zhihao Wei, Konglan Lin, Wen Deng, Yuanle Chen, Yiwen Lu Published: 2025 ABSTRACT BackgroundTertiary lymphoid structures (TLSs) are associated with an improved chemotherapy response and favorable clinical outcomes in various tumor types, including breast cancer. The number of TLSs increases in breast cancer after neoadjuvant chemotherapy (NAC), potentially activating the immune landscape of the tumor microenvironment. However, the potential relationship between TLSs and clinicopathological characteristics in patients with breast cancer after receiving NAC is poorly understood.MethodsThis retrospective study included clinicopathological data from breast cancer patients treated at Sun Yat-Sen Memorial Hospital between 2012 and 2023. Hematoxylin‒eosin staining and immunohistochemical analysis were performed to evaluate TLSs and lymphatic vessels in two cohorts comprising 326 NAC-treated and 304 treatment-naive breast cancer patients, with validation conducted in a cohort of 23 paired biopsy specimens collected before and after receiving NAC. RNA fluorescence in situ hybridization (RNA-FISH) was employed to quantify lymphatic endothelial cell patterns. TLS gene expression scores were determined to validate the observed differences in TLS patterns in breast cancer patients.ResultsTLSs were more common in breast cancer samples after receiving NAC (p < 0.0001). Compared with breast cancer patients with a low density of TLSs (TLS-low) (n = 92) and with the absence of TLSs (TLS-absent) (n = 142), patients with a high density of TLSs (TLS-high) (n = 92) exhibited significantly lower stages of residual axillary disease (Bonferroni-adjusted p < 0.0167) after receiving NAC, but there were no clinical stage differences before receiving NAC (p > 0.05). Compared with patients with a low density of TLSs after receiving NAC, patients with a high density of TLSs exhibited a lower residual axillary nodal burden (p < 0.0001). Transcriptomic analysis validated that patients with NAC-treated breast cancer with an axillary pathological complete response (pCR) attained higher TLS gene scores than those with a non-pCR axilla did (p < 0.0001). Lymphatic vessel density and size were highly correlated with TLS density in the NAC cohort, and these results agree with those of a validation cohort (n = 23) that included a paired-sample set of breast cancer patients before and after receiving NAC. The RNA-FISH results revealed that increased transcript levels of lymphatic vessel-specific markers were significantly associated with a higher density of TLSs in breast cancer. Additionally, lymphatic-specific gene expression was positively correlated with TLS gene scores in two large breast cancer datasets.ConclusionsThese findings highlight the association between TLS density and the axillary residual nodal burden, and illustrate the potential relationship between lymphatic vessels and TLSs in NAC-treated breast cancer patients.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40001-025-03556-6. Introduction Cancer biology is characterized by the acquisition of core hallmarks, such as sustained proliferation and evasion of cell death, which are driven by genetic mutations. During tumor progression, cancer cells interact with cells within the tumor microenvironment (TME), often taking advantage of neighboring components to facilitate tumor expansion [1, 2]. Targeting tumor or key neighboring components within the TME represents a significant strategy in cancer treatment [3]. The treatment strategies of cancer have evolved over the past two centuries, currently including surgery, chemotherapy, radiotherapy, endocrine therapy, immunotherapy, and targeted therapy [4]. Surgical interventions may restore immune capacity for robust adaptive responses within the TME [5]. Chemotherapies and radiotherapies initially disrupt the TME through tumor cell killing, which may induce pro-survival and immunosuppressive responses in the remaining cells, yet simultaneously enhance local antitumor immune infiltration [1, 6]. Endocrine therapy exerts dual immunomodulatory effects on the TME [7]. Immunotherapies modify interactions between tumor cells and other cells within the TME, thereby activating durable immune responses that mediate long-term tumor eradication [6]. Targeted therapies enable precise modulation of the TME through the identification of specific biomarkers expressed by tumor cells or critical neighboring cell populations [6, 8]. The therapeutic efficacy of these treatment strategies depends on the extent of immune activation within the TME [1]. Tertiary lymphoid structures (TLSs) are ectopic lymphoid aggregates that develop at sites of chronic inflammation, including the TME [9]. TLSs are dominated by organized B and T cells, along with supportive stromal and immune components such as follicular dendritic cells (FDCs), fibroblasts, lymphatic vessels and blood vessels (e.g., high endothelial venules) [9, 10]. TLSs can initiate local antitumor immune responses within the TME by directly exposing tumor antigens to immune cells, thereby generating efficient and sustained immunity without relying mostly on activating lymph nodes [11]. TLSs play a crucial role in the antitumor immune response and are thought to be potential biomarkers associated with favorable clinical outcomes, including less lymph node metastasis [9, 12–15]. It has been suggested that cytotoxic lymphocytes from TLSs in tumors might infiltrate the periphery of draining lymph nodes through lymphatic vessels, and these cytotoxic lymphocytes may contribute to the prevention of lymph node metastasis [15]. Therefore, TLSs in malignant tumors have significant research and clinical application value. However, in untreated breast cancer and other immune-cold tumors, TLSs are detected only in a small number of patients [9, 11, 16], and the relationship between lymphatic vessels and TLSs remains unclear. Neoadjuvant chemotherapy (NAC) refers to systemic chemotherapy before surgery to downstage tumors, eradicate micrometastatic disease, reduce nodal disease, and create favorable conditions for subsequent surgery [17, 18]. NAC facilitates rapid assessment of treatment effectiveness and accelerates the development of treatments for early-stage breast cancer [19]. Mechanistically, NAC can promote the release of tumor-specific antigens and enhance the antitumor immune response by upregulating damage-associated molecular pattern (DAMP) molecules within the TME [20]. Studies have indicated that NAC potentially enhances the antitumor immune response, shifting it toward a state of immune activation [14, 21–24]. Our previous and recent studies have demonstrated that intratumoral TLSs increase after receiving NAC, particularly in patients with a better therapeutic response [14, 25, 26]. Retrospective studies have demonstrated that TLSs are associated with better prognosis in patients with cancers following NAC [26–28]. However, few studies have explored the association between TLSs and specific clinicopathological characteristics in these patients, such as the relationship between TLSs and axillary residual nodal burden. In this study, we retrospectively investigated TLSs in tumor tissues and their relationships with clinicopathological characteristics in patients with breast cancer after receiving NAC (NAC cohort) and no prior NAC treatment (treatment-naive cohort). We also analyzed the association between TLSs and the axillary residual nodal burden in the NAC cohort and large breast cancer datasets and described the potential relationship between lymphatic vessels and TLSs. Methods Patients and human tumor samples Patients with breast cancer in the neoadjuvant treatment cohort were identified for inclusion in this study if they had undergone surgical resection for locally advanced, non-distant metastatic breast cancer after receiving NAC at Sun Yat-Sen Memorial Hospital between 2012 and 2023. Patients at the same hospital during the same period were selected in the untreated control cohort if they had undergone surgical resection for breast cancer without prior systemic treatment. A total of 630 female breast cancer patients with invasive breast cancer were retrospectively enrolled from Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University (Guangzhou, China), including those who did not receive NAC (n = 304) and those who underwent NAC and subsequent surgery (n = 326) between 2012 and 2023. The inclusion criteria were as follows: (i) biopsy-proven invasive breast cancer; (ii) available baseline data. The exclusion criteria were as follows: (i) the presence of occult breast tumors; (ii) the presence of distant metastasis or another malignancy. Clinical and pathological data were collected for patients all included. The clinicopathological characteristics of patients with breast cancer who did (NAC cohort) and did not receive neoadjuvant therapy (treatment-naive cohort) prior to surgery were detailed in Table 1. To reduce potential spatial sampling biases, we also selected 23 patients who underwent core needle biopsy in the NAC cohort and analyzed their paired biopsy specimens and surgical resection specimens. Tissue blocks with only or predominantly nontumor tissues or in situ carcinoma tissues were excluded.Table 1Clinicopathological characteristics of the NAC cohort and treatment-naive cohort of patients with breast cancerCharacteristicNAC cohort (n = 326)Treatment-naive cohort (n = 304)p valueMedian age at surgery, y (range)45 (25–74)45 (24–71)0.9614Median body mass index at surgery, kg/m2 (range)22.9 (15.8–38.5)22.8 (15.4–33.3)0.6917Tumor subtype TNBC28 (8.6%)15 (4.9%)0.8074 HER2 + 132 (40.5%)82 (26.9%) HR + /HER2-166 (50.9%)207 (68.0%)Clinical T stage cT119 (5.8%)23 (7.6%)0.2913 cT2217 (66.5%)213 (70.1%) cT365 (19.9%)54 (17.8%) cT425 (7.7%)14 (4.6%)Clinical N stage cN093 (28.5%)90 (29.6%)0.9639 cN1195 (59.8%)181 (59.5%) cN219 (5.8%)15 (4.9%) cN319 (5.8%)18 (5.9%)LVI No265 (81.3%)229 (75.3%)0.7361 Yes61 (18.7%)75 (24.6%)Tumor histological type Ductal287 (88.0%)274 (90.1%)0.3568 Lobular21 (6.4%)7 (2.3%) Othera18 (5.5%)23 (7.5%)Pathological tumor statusb ypT0/ypTis53 (16.3%)- pT1/ypT1180 (55.2%)164 (53.9%) pT2/ypT287 (26.7%)135 (44.4%) pT3/ypT36 (1.8%)5 (1.6%)Pathological nodal statusc pN0/ypN0186 (57.1%)201 (66.1%) pN1/ypN184 (25.8%)65 (21.4%) pN2/ypN240 (12.3%)22 (7.2%) pN3/ypN316 (4.9%)16 (5.3%)Pathological AJCC Staged -/Stage y0, yis47 (14.4%)– Stage I/Stage yI99 (30.4%)118 (38.8%) Stage II/Stage yII123 (37.7%)147 (48.4%) Stage III/Stage yIII57 (17.5%)39 (12.8%)Respond to NACe Responder220 (67.5%)– Non-Responder106 (32.5%)–AJCC American Joint Committee on Cancer, HER2 human epidermal growth factor receptor 2, HR hormone receptor, LVI lymphovascular invasion, NAC neoadjuvant chemotherapy, pCR pathologic complete response, TNBC triple-negative breast cancer, TLS tertiary lymphoid structureaIncludes all tumors that were neither ductal nor lobularbRefers to tumor size on surgical specimens. Pathological tumor status after receiving NAC treatment refers to the largest contiguous focus of residual invasive tumors in surgical specimens after NAC. Pathological T staging following neoadjuvant therapy is denoted with a yp prefix to T classification following the AJCC 8th editioncPathological nodal stage refers to the number of lymph nodes affected by metastases. Pathological nodal stage after receiving NAC treatment refers to the number of lymph nodes affected by metastases of residual invasive tumors after NAC. Pathological N staging following neoadjuvant therapy is denoted with a yp prefix to N classification following the AJCC 8th editiondCategories indicate AJCC 8th edition anatomic stageseA responder after NAC is defined as having a partial response (PR) or a complete response (CR); a non-responder after NAC is defined as having no PR or CR Table 1Clinicopathological characteristics of the NAC cohort and treatment-naive cohort of patients with breast cancerCharacteristicNAC cohort (n = 326)Treatment-naive cohort (n = 304)p valueMedian age at surgery, y (range)45 (25–74)45 (24–71)0.9614Median body mass index at surgery, kg/m2 (range)22.9 (15.8–38.5)22.8 (15.4–33.3)0.6917Tumor subtype TNBC28 (8.6%)15 (4.9%)0.8074 HER2 + 132 (40.5%)82 (26.9%) HR + /HER2-166 (50.9%)207 (68.0%)Clinical T stage cT119 (5.8%)23 (7.6%)0.2913 cT2217 (66.5%)213 (70.1%) cT365 (19.9%)54 (17.8%) cT425 (7.7%)14 (4.6%)Clinical N stage cN093 (28.5%)90 (29.6%)0.9639 cN1195 (59.8%)181 (59.5%) cN219 (5.8%)15 (4.9%) cN319 (5.8%)18 (5.9%)LVI No265 (81.3%)229 (75.3%)0.7361 Yes61 (18.7%)75 (24.6%)Tumor histological type Ductal287 (88.0%)274 (90.1%)0.3568 Lobular21 (6.4%)7 (2.3%) Othera18 (5.5%)23 (7.5%)Pathological tumor statusb ypT0/ypTis53 (16.3%)- pT1/ypT1180 (55.2%)164 (53.9%) pT2/ypT287 (26.7%)135 (44.4%) pT3/ypT36 (1.8%)5 (1.6%)Pathological nodal statusc pN0/ypN0186 (57.1%)201 (66.1%) pN1/ypN184 (25.8%)65 (21.4%) pN2/ypN240 (12.3%)22 (7.2%) pN3/ypN316 (4.9%)16 (5.3%)Pathological AJCC Staged -/Stage y0, yis47 (14.4%)– Stage I/Stage yI99 (30.4%)118 (38.8%) Stage II/Stage yII123 (37.7%)147 (48.4%) Stage III/Stage yIII57 (17.5%)39 (12.8%)Respond to NACe Responder220 (67.5%)– Non-Responder106 (32.5%)–AJCC American Joint Committee on Cancer, HER2 human epidermal growth factor receptor 2, HR hormone receptor, LVI lymphovascular invasion, NAC neoadjuvant chemotherapy, pCR pathologic complete response, TNBC triple-negative breast cancer, TLS tertiary lymphoid structureaIncludes all tumors that were neither ductal nor lobularbRefers to tumor size on surgical specimens. Pathological tumor status after receiving NAC treatment refers to the largest contiguous focus of residual invasive tumors in surgical specimens after NAC. Pathological T staging following neoadjuvant therapy is denoted with a yp prefix to T classification following the AJCC 8th editioncPathological nodal stage refers to the number of lymph nodes affected by metastases. Pathological nodal stage after receiving NAC treatment refers to the number of lymph nodes affected by metastases of residual invasive tumors after NAC. Pathological N staging following neoadjuvant therapy is denoted with a yp prefix to N classification following the AJCC 8th editiondCategories indicate AJCC 8th edition anatomic stageseA responder after NAC is defined as having a partial response (PR) or a complete response (CR); a non-responder after NAC is defined as having no PR or CR Clinicopathological characteristics of the NAC cohort and treatment-naive cohort of patients with breast cancer AJCC American Joint Committee on Cancer, HER2 human epidermal growth factor receptor 2, HR hormone receptor, LVI lymphovascular invasion, NAC neoadjuvant chemotherapy, pCR pathologic complete response, TNBC triple-negative breast cancer, TLS tertiary lymphoid structure aIncludes all tumors that were neither ductal nor lobular bRefers to tumor size on surgical specimens. Pathological tumor status after receiving NAC treatment refers to the largest contiguous focus of residual invasive tumors in surgical specimens after NAC. Pathological T staging following neoadjuvant therapy is denoted with a yp prefix to T classification following the AJCC 8th edition cPathological nodal stage refers to the number of lymph nodes affected by metastases. Pathological nodal stage after receiving NAC treatment refers to the number of lymph nodes affected by metastases of residual invasive tumors after NAC. Pathological N staging following neoadjuvant therapy is denoted with a yp prefix to N classification following the AJCC 8th edition dCategories indicate AJCC 8th edition anatomic stages eA responder after NAC is defined as having a partial response (PR) or a complete response (CR); a non-responder after NAC is defined as having no PR or CR The NAC regimens were as follows: doxorubicin (60 mg/m2) plus cyclophosphamide (600 mg/m2) every three weeks for four cycles, followed by paclitaxel (80 mg/m2) for 12 weeks or docetaxel (75 mg/m2) and cyclophosphamide (600 mg/m2) every three weeks for four to six cycles. Human epidermal growth factor receptor 2 (HER2)-positive patients were treated with concomitant trastuzumab (initial loading dose of 4 mg/kg and subsequent doses of 2 mg/kg/week). Therapeutic effects were evaluated according to the standard Response Evaluation Criteria in Solid Tumors. Pathological complete response (pCR) was defined as the absence of residual invasive carcinoma or ductal carcinoma in situ (DCIS) in any excised breast or regional lymph node tissue detected by microscopy. The occurrence of pCR after receiving NAC was defined as the absence of residual invasive carcinoma in the breast and regional lymph nodes (ypT0/Tis, N0); breast-only or node-only pCR was defined as the absence of any residual invasive carcinoma or DCIS limited to the breast or axilla detected by microscopy. There was no anatomic stage group assigned if there was a pCR to neoadjuvant therapy. Complete response (CR) was defined as the disappearance of all target lesions and any pathological lymph nodes (whether target or non-target) must have a reduction in short axis to < 10 mm. Partial response (PR) was defined as at least a 30% reduction in the sum of the longest diameter of target lesions detected by clinical examination and imaging studies. A responder was defined as having a PR or a CR; a non-responder was defined as having no PR or CR. Hormone receptor (HR) and HER2 status conditions were assessed using immunohistochemical (IHC) staining or fluorescence in situ hybridization (FISH) on paraffin-embedded tumor slides, following the guidelines of the American Society of Clinical Oncology/College of American Pathologists. Enrolled patients had confirmed HER2-positive (3 + by immunohistochemistry and/or positive by FISH) invasive breast cancer, following the 2007 American Society of Clinical Oncology/College of American Pathologists guidelines. HR positivity was determined on the basis of the presence of ≥ 1% tumor cells expressing estrogen and/or progesterone receptors. All the samples were obtained from patients who provided informed consent, and all the relevant procedures were approved by the Internal Review and Ethics Boards of Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University (approval number: SYSKY-2025-182-01). TLS identification based on hematoxylin‒eosin (H&E) images All cases were blind-reviewed by two operators, ZW and KL, for the presence of TLSs according to H&E staining of tumor sections. The analyzed samples were core needle biopsies (n = 23) and surgical specimens (n = 630). Furthermore, a blinded assessment was independently performed on these samples by a third operator (YC). Only lymphoid aggregates that consisted of at least 50 lymphocytes were considered as TLSs[29]. The TLS density was determined by calculating the number of TLSs per mm2 of breast tumor sections. Assessment of the inter-rater agreement of the pathological assessment The reproducibility of assessing the TLS status of breast tumor samples with the pathology method was tested by training the third operator, YC, for TLS screening on a set of 150 samples. The third operator, YC, subsequently analyzed 630 tumor samples blindly. The Cohen’s kappa coefficient (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\upkappa }$$\end{document}κ) test was conducted to measure interrater reliability for the presence of TLSs[29, 30]. The inter-rater agreement rate (Cohen’s \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\upkappa }$$\end{document}κ score) was 0.834 for the presence of TLSs (95% CI [0.811, 0.857]; p < 0.0001) (Supplementary Table 1). Immunohistochemistry Formalin-fixed paraffin-embedded (FFPE) breast tumor tissue sections (thickness 4 μm) were deparaffinized in xylene and rehydrated in a graded ethanol series. Next, the deparaffinized slides were placed in boiling EDTA to repair the antigen for 3 min. Subsequently, the samples were incubated in 3% hydrogen peroxide for 10 min to block endogenous peroxidase activity and then were washed with phosphate-buffered saline (PBS) three times. Afterward, all the dried sections were incubated in 10% bull serum albumin for 30 min, followed by incubation overnight at 4 °C with rabbit anti–human PDPN (Cat# ZM-0465) at a 1:100 dilution. The biotinylated secondary antibodies were labeled using streptomycin avidin-peroxidase solution (Dako) at a 1:500 dilution. Finally, the samples were stained using a Simple DAB Stain Kit (Dako) according to the manufacturer’s instructions, and the sections were counterstained with hematoxylin. The sections from each sample (processed with non-immunized serum instead of the primary antibody) were employed as negative controls. Quantification of podoplanin-positive (PDPN+) lymph vessels To quantify the parameters of the lymph vessels, five primary breast tumor sections from each patient were stained with the rabbit anti–human PDPN (Cat# ZM-0465) and screened by two experienced operators, ZW and KL, to select the maximum staining area (hot spot) under × 100 magnification (Olympus BX63) [31, 32]. Next, the mean value of PDPN+ lymphatic vessels within five hot spots for each tumor section was computed at × 200 magnification. Lymphatic vessel density (LVD) was defined as the number of lymphatic vessels within five × 200 fields. Afterward, five consecutive fields within the hot spots under × 200 magnification were captured using a digital image capture device (Olympus BX63) for the calculation of the lymphatic vessel area (LVA) and lymphatic vessel perimeter (LVP). The LVA was depicted as the mean lumen area of the lymphatic vessels within the five × 200 fields, and the LVP was depicted as the mean perimeter of the lymphatic vessels within the five × 200 fields. A semiquantitative automated counting approach was used to validate the manual results above. Briefly, 326 breast tumor samples were obtained from the NAC cohort. IHC images of entire clinical tissue sections were captured and deposited using the side scanner Axio Scan.Z1 (Zeiss). The scan of the slices was fully automated. We used the image analysis software APtime (version 2.0.0–3596) to quantify the density, area, and perimeter of the lymphatic vessels in the whole-slide images (WSIs). Cell detection was based on nuclear staining (hematoxylin) with a traditional nuclear segmentation type, and the nuclear contrast threshold was set at 0.5 with an intensity of 0.015 and an aggressiveness of 0.6. The nuclear size ranged from 5 to 150 (µm2), the minimum roundness of segmentation was set at 0.05, and the cytoplasmic radius was set at 1.5 µm. For the analysis, the PDPN+ lymphatic endothelial cells (LECs) were defined by the DAB channel and histological characteristics of the vessels. The filter size was set at 1.66 µm, with a threshold of 0.39. Subsequently, the software automatically calculated the following parameters of the lymphatic vessels in the WSIs. The LVD was defined as the number of lymphatic vessels within the whole slide area [33]. The LVA was depicted as the mean lumen area of the lymphatic vessels within the whole slide area, and the LVP was depicted as the mean perimeter of the lymphatic vessels within the whole slide area [33]. To analyze the consistency between manual counting and automated counting of lymphatic vessels, we randomly selected 50 IHC-stained WSIs from all available histopathological images. A third experienced operator, YC, who was blinded to the results of automatic TLS segmentation, manually identified the lymphatic vessels on these selected WSIs and determined the lymphatic vessel count and size in the tumor region for each WSI. Manual calculations were conducted using an Olympus BX63 digital image capture device as mentioned above. The intraclass correlation coefficients (ICCs) and Bland‒Altman plots were used to assess the agreement between manual counting based on IHC-stained WSIs and automated lymphatic vessel calculation. The Bland‒Altman plot revealed favorable consistency between the automatic lymphatic vessel counting and the manual lymphatic vessel counting (LVD: ICC = 0.9248, 95% CI 0.8669–0.9574, p < 0.0001; LVA: ICC = 0.9984, 95% CI 0.9972–0.9991, p < 0.0001; LVP: ICC = 0.9291, 95% CI 0.8778–0.9594, p < 0.0001) (Supplementary Fig. 1). RNA fluorescence in situ hybridization analysis RNA fluorescence in situ hybridization (RNA-FISH) was performed on freshly cut 4 µm FFPE tissue sections from the IHC cohort using a Fluorescence in Situ Hybridization Kit for RNA (Beyotime) according to the manufacturer’s instructions. In short, tissue sections were baked for 1 h at 65 °C and then deparaffinized and rehydrated in a graded ethanol series. The slides were then digested with 20 µg/mL proteinase K in prewarmed PBS for 10–20 min at room temperature (RT), subsequently fixed in 4% paraformaldehyde for 10 min and washed for 2 × 5 min in PBS-T (1 × PBS with 0.1% Triton X-100) at RT before proceeding with RNA-FISH. The fluorescence-labeled detection probes PDPN and LYVE1 (target sequences in Supplementary Table 2; Shanghai GenePharma Co., Ltd., Shanghai, China) were mixed at a concentration of 1 μM in hybridization buffer (10% dextran sulfate/2 × saline-sodium citrate tween (SSCT) (0.3 M NaCl, 0.03 M sodium citrate, 0.1% Tween-20) /50% formamide/4% polyvinylsulfonic acid) to a final volume of 25 μL. After spotting the probes, the slides were covered with a coverslip, sealed with rubber cement, and denatured for 2.5 min at 92 °C before being transferred to a 37 °C humidified chamber overnight. The next day, the slides were washed with 2 × SSCT for 15 min at 60 °C, 15 min at RT, and 5 min at RT in 0.2 × saline-sodium citrate (SSC). The sections were counterstained and mounted with Antifade Mounting Medium with DAPI (Beyotime). Images were generated using the side scanner Axio Scan. Z1 (Zeiss). All the samples were scanned at 40 × magnification. All the images were automatically preprocessed and evaluated using the in situ hybridization module of the HALO® image analysis software (Indica Labs). The fluorescence intensity was quantified using Image-Pro Plus 6 software. Briefly, three fields of view per image were randomly selected for analysis. The area and integrated optical density (IOD) were measured using the ‘Count and Measure Objects’ function. The mean fluorescence intensity of PDPN or LYVE1 for each field was calculated as the IOD/area, and the final reported fluorescence intensity per section was derived from the average of the three fields. The relative fluorescence intensity of each marker was calculated by dividing the mean fluorescence intensity per section by the mean fluorescence intensity of the TLS-absent sections. Validation of dataset retrieval and preprocessing Thirty-nine TLS signature genes were obtained from Sautès-Fridman et al. [9]. Datasets were selected if they satisfied the following criteria: [1] patients with breast cancer with transcriptome sequencing results; [2] sample size greater than 30. The exclusion criterion was healthy individuals with transcriptome sequencing results. Transcriptional matrix (FPKM) and corresponding clinical information for 1231 breast cancer patients were downloaded from The Cancer Genome Atlas (TCGA) (https://portal.gdc.cancer.gov/). Additionally, data of 935 breast cancer patients before receiving NAC (pre-NAC)-treatment were obtained from Wolf et al. [34] and the National Center for Biotechnology Information Gene Expression Omnibus (GEO) database under accession code GSE194040, data of 30 pre-NAC-treatment breast cancer patients were sourced from Park et al. [21] and the GEO database under accession code GSE123845, data of 120 pre-NAC-treatment breast cancer patients were derived from Desmedt et al.[35] and the GEO database under accession code GSE16446, data of 162 pre-NAC-treatment breast cancer patients were acquired from Sammut et al.[36] and the GEO databases under accession code GSE25066, GSE32603, GSE41998, and data of 121 pre-NAC-treatment breast cancer patients were obtained from Axelrod et al.[37]. Next, we transferred the fragments per kilobase million (FPKM) values into transcripts per kilobase million (TPM) values. We used a log2-transformed TPM matrix of counts, which was normalized by subtracting the average log2 expression of all genes of each transcriptomic dataset. TLS gene signatures were extracted from each transcriptomic dataset. TLS gene scores were calculated using the R package GSVA with the thirty-nine-gene TLS signatures proposed by Sautès-Fridman et al.[9]. In summary, the TLS signatures included the genes CCL2, CCL3, CCL4, CCL5, CCL8, CCL18, CCL19, CCL21, CXCL9, CXCL10, CXCL11, CXCL13, CD200, FBLN7, ICOS, SGPP2, SH2D1A, TIGIT, PDCD1, CD4, CCR5, CXCR3, CSF2, IGSF6, IL2RA, CD38, CD40, CD5, MS4A1, SDC1, GFI1, IL1R1, IL1R2, IL10, CCL20, IRF4, TRAF6, STAT5A and TNFRSF17. The signature score was calculated as the mean gene expression by the R package using the GSVA method. Analysis and visualization were performed using the R package ggplot2. Statistical analysis Statistical analyses were performed using R version 4.4.2 and the associated packages and GraphPad Prism (V.8.0.1; GraphPad Software). The figure legends, text, or methods provide information regarding the statistical details and methods. Data are presented as either the mean or median of continuous values and were compared between two groups using Student’s t-test or Mann‒Whitney U test. For paired sample comparisons, paired t-tests were used. Unless otherwise specified, analysis for significance was performed by one-way or two-way analysis of variance (ANOVA) when more than two groups were compared. All the statistical tests described in this work were two-sided. Categorical variables in the tables were compared using the χ2 test or Fisher’s exact test. p-values < 0.05 were regarded as statistically significant, which were defined as *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001. p > 0.05 was denoted as ns (not significant). Bonferroni adjustment was used to adjust for the multiple testing among three groups of TLS types in Table 2 (Bonferroni-adjusted p < 0.0167 were regarded as statistically significant. Bonferroni-adjusted p > 0.0167 was denoted as ns (not significant)).Table 2NAC patients’ demographics and clinicopathological characteristics according to the density of TLSsCharacteristicTLS-high group (n = 92)TLS-low group (n = 92)TLS-absent group (n = 142)p valueAge in years, median (range), y46 (28–65)44.5 (27–67)44.5 (25–74)0.4293Body mass index, median(range), kg/m223 (17.2–30.8)22.5 (16–30)22.9 (15.8–38.5)0.5831Tumor subtype TNBC13 (14.1%)3 (3.3%)12 (8.5%)0.1034 HER2 + 36 (39.1%)42 (45.7%)54 (38.0%) HR + /HER2-43 (46.7%)47 (51.1%)76 (53.5%)Clinical T stage cT17 (7.6%)5 (5.4%)7 (4.9%)0.8361 cT259 (64.1%)64 (69.6%)94 (66.2%) cT318 (19.6%)19 (20.7%)28 (19.7%) cT48 (8.7%)4 (4.3%)13 (9.2%)Clinical N stage cN036 (39.1%)18 (19.6%)39 (27.5%)0.1334 cN148 (52.2%)62 (67.4%)85 (59.9%) cN25 (5.4%)5 (5.4%)9 (6.3%) cN33 (3.3%)7 (7.6%)9 (6.3%)LVI No80 (87.0%)73 (79.3%)112 (78.9%)0.2630 Yes12 (13.0%)19 (20.7%)30 (21.1%)Tumor histological type Ductal77 (83.6%)87 (94.5%)123 (86.6%)0.5980 Lobular6 (6.5%)1 (1.1%)14 (9.8%) Othera9 (9.7%)4 (4.3%)5 (3.5%)Pathological tumor status after NACb ypT0/Tis24 (26.1%)8 (8.7%)21 (14.8%)0.0036 ypT145 (48.9%)61 (66.3%)74 (52.1%) ypT223 (25%)19 (20.7%)45 (31.7%) ypT3–4 (4.3%)2 (1.4%)Pathological nodal status after NACc ypN083 (90.2%)44 (47.8%)59 (41.5%) < 0.0001 ypN19 (9.8%)27 (29.3%)48 (33.8%) ypN2–16 (17.4%)24 (16.9%) ypN3–5 (5.4%)11 (7.7%)yAJCC Staged -/Stage y0, yis24 (26.1%)5 (5.4%)18 (12.7%) < 0.0001 Stage yI38 (41.3%)33 (35.9%)28 (19.7%) Stage yII30 (32.6%)33 (35.9%)60 (42.3%) Stage yIII–21 (22.8%)36 (25.4%)Respond to NACe Responder64 (69.6%)59 (64.1%)97 (68.3%)0.6925 Non-Responder28 (30.4%)33 (35.9%)45 (31.7%)AJCC American Joint Committee on Cancer, HER2 human epidermal growth factor receptor 2, HR hormone receptor, LVI lymphovascular invasion, NAC neoadjuvant chemotherapy, pCR pathologic complete response, TNBC triple-negative breast cancer, TLS tertiary lymphoid structureaIncludes all tumors that were neither ductal nor lobularbPathological tumor status after receiving NAC treatment refers to the largest contiguous focus of residual invasive tumors in surgical specimens after NAC. Pathological T staging following neoadjuvant therapy is denoted with a yp prefix to T classification following the AJCC 8th editioncPathologic nodal stage after receiving NAC treatment refers to the number of lymph nodes affected by metastases of residual invasive tumors after NAC. Pathological N staging following neoadjuvant therapy is denoted with a yp prefix to N classification following the AJCC 8th editiondCategories indicate AJCC 8th edition anatomic stageseA responder after NAC is defined as having a partial response (PR) or a complete response (CR); a non-responder after NAC is defined as having no PR or CR Table 2NAC patients’ demographics and clinicopathological characteristics according to the density of TLSsCharacteristicTLS-high group (n = 92)TLS-low group (n = 92)TLS-absent group (n = 142)p valueAge in years, median (range), y46 (28–65)44.5 (27–67)44.5 (25–74)0.4293Body mass index, median(range), kg/m223 (17.2–30.8)22.5 (16–30)22.9 (15.8–38.5)0.5831Tumor subtype TNBC13 (14.1%)3 (3.3%)12 (8.5%)0.1034 HER2 + 36 (39.1%)42 (45.7%)54 (38.0%) HR + /HER2-43 (46.7%)47 (51.1%)76 (53.5%)Clinical T stage cT17 (7.6%)5 (5.4%)7 (4.9%)0.8361 cT259 (64.1%)64 (69.6%)94 (66.2%) cT318 (19.6%)19 (20.7%)28 (19.7%) cT48 (8.7%)4 (4.3%)13 (9.2%)Clinical N stage cN036 (39.1%)18 (19.6%)39 (27.5%)0.1334 cN148 (52.2%)62 (67.4%)85 (59.9%) cN25 (5.4%)5 (5.4%)9 (6.3%) cN33 (3.3%)7 (7.6%)9 (6.3%)LVI No80 (87.0%)73 (79.3%)112 (78.9%)0.2630 Yes12 (13.0%)19 (20.7%)30 (21.1%)Tumor histological type Ductal77 (83.6%)87 (94.5%)123 (86.6%)0.5980 Lobular6 (6.5%)1 (1.1%)14 (9.8%) Othera9 (9.7%)4 (4.3%)5 (3.5%)Pathological tumor status after NACb ypT0/Tis24 (26.1%)8 (8.7%)21 (14.8%)0.0036 ypT145 (48.9%)61 (66.3%)74 (52.1%) ypT223 (25%)19 (20.7%)45 (31.7%) ypT3–4 (4.3%)2 (1.4%)Pathological nodal status after NACc ypN083 (90.2%)44 (47.8%)59 (41.5%) < 0.0001 ypN19 (9.8%)27 (29.3%)48 (33.8%) ypN2–16 (17.4%)24 (16.9%) ypN3–5 (5.4%)11 (7.7%)yAJCC Staged -/Stage y0, yis24 (26.1%)5 (5.4%)18 (12.7%) < 0.0001 Stage yI38 (41.3%)33 (35.9%)28 (19.7%) Stage yII30 (32.6%)33 (35.9%)60 (42.3%) Stage yIII–21 (22.8%)36 (25.4%)Respond to NACe Responder64 (69.6%)59 (64.1%)97 (68.3%)0.6925 Non-Responder28 (30.4%)33 (35.9%)45 (31.7%)AJCC American Joint Committee on Cancer, HER2 human epidermal growth factor receptor 2, HR hormone receptor, LVI lymphovascular invasion, NAC neoadjuvant chemotherapy, pCR pathologic complete response, TNBC triple-negative breast cancer, TLS tertiary lymphoid structureaIncludes all tumors that were neither ductal nor lobularbPathological tumor status after receiving NAC treatment refers to the largest contiguous focus of residual invasive tumors in surgical specimens after NAC. Pathological T staging following neoadjuvant therapy is denoted with a yp prefix to T classification following the AJCC 8th editioncPathologic nodal stage after receiving NAC treatment refers to the number of lymph nodes affected by metastases of residual invasive tumors after NAC. Pathological N staging following neoadjuvant therapy is denoted with a yp prefix to N classification following the AJCC 8th editiondCategories indicate AJCC 8th edition anatomic stageseA responder after NAC is defined as having a partial response (PR) or a complete response (CR); a non-responder after NAC is defined as having no PR or CR NAC patients’ demographics and clinicopathological characteristics according to the density of TLSs AJCC American Joint Committee on Cancer, HER2 human epidermal growth factor receptor 2, HR hormone receptor, LVI lymphovascular invasion, NAC neoadjuvant chemotherapy, pCR pathologic complete response, TNBC triple-negative breast cancer, TLS tertiary lymphoid structure aIncludes all tumors that were neither ductal nor lobular bPathological tumor status after receiving NAC treatment refers to the largest contiguous focus of residual invasive tumors in surgical specimens after NAC. Pathological T staging following neoadjuvant therapy is denoted with a yp prefix to T classification following the AJCC 8th edition cPathologic nodal stage after receiving NAC treatment refers to the number of lymph nodes affected by metastases of residual invasive tumors after NAC. Pathological N staging following neoadjuvant therapy is denoted with a yp prefix to N classification following the AJCC 8th edition dCategories indicate AJCC 8th edition anatomic stages eA responder after NAC is defined as having a partial response (PR) or a complete response (CR); a non-responder after NAC is defined as having no PR or CR Results TLSs are more common in breast cancer samples after receiving NAC To better understand the clinical significance of TLSs in patients with breast cancer treated with NAC, we identified a treatment cohort (n = 326) that had undergone surgical resection of breast cancer after receiving NAC between 2012 and 2023. Given that TLSs can occur in treatment-naive breast cancer patients [14], we also identified a control cohort (n = 304) of patients who had undergone surgical resection for breast cancer without prior NAC and whose tumors served as a surrogate for the pre-NAC breast cancer microenvironment (Table 1). This treatment-naive cohort was similar to the NAC cohort in terms of age, body mass index (BMI), tumor molecular subtype, clinical T/N stage, lymphovascular invasion (LVI), and tumor histological type. To evaluate the presence of TLSs, we performed H&E staining on tumor tissue sections from all enrolled breast cancer patients (n = 630) (refer to Methods). In NAC-treated breast tumors, we observed many TLSs in the tumor stroma with almost no residual viable tumor. However, in the treatment-naive cohort, the predominant area was residual viable tumors lacking TLSs (Fig. 1a). Next, we analyzed the presence of TLSs in the two cohorts (Fig. 1b). We found that TLSs were present in 184 patients in the NAC cohort (n = 326) (56.4%) and 22 patients in the treatment-naive cohort (n = 304) (7.2%) (p < 0.0001) (Fig. 1c). Among all the cohorts (n = 630) in the combined NAC-treated and treatment-naive group, patients with TLSs exhibited a significantly lower proportion of positive lymph nodes compared to those without TLSs (Fig. 1d). These results raised the possibility that TLSs could emerge as the front line of anti-tumor immune responses after receiving NAC and were associated with better clinical prognosis.Fig. 1Assessment of the presence of TLSs in primary breast tumors. a Representative FFPE breast tumors stained with H&E. (upper) A representative treatment-naive breast tumor stained with H&E showing the absence of TLSs. The dotted line shows the boundary between the residual viable tumor and the stroma. High-magnification images of representative tumor bed and tumor stroma are shown on the right; (lower) A representative NAC-treated breast tumor stained with H&E showing the presence of TLSs. High-magnification images of representative TLSs delineated by the dotted line are shown on the right. Scale bars, 1 mm for the low-magnification image and 200 μm for the high-magnification image. b Representative images of H&E staining for treatment-naive and NAC specimens. The TLSs are delineated with black lines, highlighting their proximity to tumor cells. c The percentage of breast cancer patients with and without TLSs in the treatment-naive and NAC cohort. Comparisons were performed using the χ2 test. ****p < 0.0001. d Comparison of the percentage of positive lymph nodes to the total number of dissected local lymph nodes in the combined NAC-treated and treatment-naive groups with respect to the presence of TLSs. In the box plots, the centerline represents the median, and the box limits represent the lower and upper quartiles. The upper whisker extends to the minimal, maximum, or third quartile plus the 1.5 × interquartile range (IQR). The lower whisker extends to the maximal, minimum, or first quartile minus 1.5 × IQR. Comparisons were performed using the Mann–Whitney U test. ****p < 0.0001. LN lymph node, NAC neoadjuvant chemotherapy, TLS tertiary lymphoid structure Assessment of the presence of TLSs in primary breast tumors. a Representative FFPE breast tumors stained with H&E. (upper) A representative treatment-naive breast tumor stained with H&E showing the absence of TLSs. The dotted line shows the boundary between the residual viable tumor and the stroma. High-magnification images of representative tumor bed and tumor stroma are shown on the right; (lower) A representative NAC-treated breast tumor stained with H&E showing the presence of TLSs. High-magnification images of representative TLSs delineated by the dotted line are shown on the right. Scale bars, 1 mm for the low-magnification image and 200 μm for the high-magnification image. b Representative images of H&E staining for treatment-naive and NAC specimens. The TLSs are delineated with black lines, highlighting their proximity to tumor cells. c The percentage of breast cancer patients with and without TLSs in the treatment-naive and NAC cohort. Comparisons were performed using the χ2 test. ****p < 0.0001. d Comparison of the percentage of positive lymph nodes to the total number of dissected local lymph nodes in the combined NAC-treated and treatment-naive groups with respect to the presence of TLSs. In the box plots, the centerline represents the median, and the box limits represent the lower and upper quartiles. The upper whisker extends to the minimal, maximum, or third quartile plus the 1.5 × interquartile range (IQR). The lower whisker extends to the maximal, minimum, or first quartile minus 1.5 × IQR. Comparisons were performed using the Mann–Whitney U test. ****p < 0.0001. LN lymph node, NAC neoadjuvant chemotherapy, TLS tertiary lymphoid structure TLSs correlate with lower axillary residual nodal burden in the NAC cohort To evaluate the abundance of TLSs, we classified the sections into three types: with the absence of TLSs (TLS-absent) (n = 142), with a low density (TLS-low) (n = 92) or a high density of TLSs (TLS-high) (n = 92) using the median total TLS density (0.005894/mm2) as the cut-off value as described previously [38–40] (Fig. 2a). We explored the associations between the density of TLSs and clinicopathological characteristics in the NAC cohort (Table 2). We found that there were no differences among the clinical stages or other clinical characteristics before receiving NAC among NAC patients with a high density of TLSs (TLS-high) (n = 92), with a low density (TLS-low) (n = 92) or with the absence of TLSs (TLS-absent) (n = 142) (all p > 0.05). Pathological tumor status (ypT) after receiving NAC exhibited differences among the three groups (p = 0.0036). However, there were differences only in ypT between the TLS-high group and the TLS-low group (Bonferroni-adjusted p < 0.0167). Additionally, the TLS-high, TLS-low and TLS-absent groups of breast cancer patients revealed differences in pathological nodal status (ypN) after receiving NAC (p < 0.0001). Interestingly, the TLS-high group showed a significantly lower ypN than the TLS-low and TLS-absent group (all Bonferroni-adjusted p < 0.0167). Also, the yAJCC stages of residual invasive breast cancer were different among these three groups after NAC treatment (p < 0.0001). The TLS-high group exhibited notably lower yAJCC stages of residual invasive breast cancer compared to the TLS-absent group and TLS-low group (all Bonferroni-adjusted p < 0.0167). No differences were observed in ypN and pathological yAJCC stages between the TLS-absent group and TLS-low group (all Bonferroni-adjusted p > 0.0167). Taken together, these results revealed a close relationship between the density of TLSs and breast cancer pathological yAJCC stages, especially the residual axillary nodal burden, in patients with breast cancer after receiving NAC.Fig. 2A higher density of TLSs is correlated with a lower residual nodal burden in the NAC cohort. a Representative images of the absence of TLSs (TLS-absent), a low density (TLS-low) or high density of TLSs (TLS-high) sections from patients with breast cancer after receiving NAC, as evaluated by H&E staining. The TLSs are delineated with black lines. Scale bar, 500 μm. b–d Boxplot showing the TLS density stratified by the breast cancer subtype (b), tumor stage after receiving NAC (yStage) (c), pathological complete response (pCR) and axillary pCR (d) after receiving NAC in TLS-high and TLS-low groups. e Boxplot showing the TLS density stratified by axillary pathological status after receiving NAC (ypN) in TLS-high and TLS-low groups. f Boxplot showing TLS gene scores in large breast cancer datasets stratified by pCR after receiving NAC. g Boxplot showing TLS gene scores in large database, including breast cancer patients stratified by lymph node metastasis-positive and lymph node metastasis-negative status in the treatment-naive cohort (left), and patients stratified by axillary pCR and non-pCR status in the NAC cohort (right). In the box plots, the centerline represents the median, and the box limits represent the lower and upper quartiles. The upper whisker extends to the minimal, maximum, or third quartile plus the 1.5 × interquartile range (IQR). The lower whisker extends to the maximal, minimum, or first quartile minus 1.5 × IQR. Comparisons were performed using the Mann–Whitney U test. ns not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. HER2 human epidermal growth factor receptor 2, HR hormone receptor, NAC neoadjuvant chemotherapy, pCR pathological complete response, TLS tertiary lymphoid structure, TNBC triple-negative breast cancer, ypN pathological nodal status after receiving NAC, yStage tumor stage after receiving NAC; LM lymph node metastasis A higher density of TLSs is correlated with a lower residual nodal burden in the NAC cohort. a Representative images of the absence of TLSs (TLS-absent), a low density (TLS-low) or high density of TLSs (TLS-high) sections from patients with breast cancer after receiving NAC, as evaluated by H&E staining. The TLSs are delineated with black lines. Scale bar, 500 μm. b–d Boxplot showing the TLS density stratified by the breast cancer subtype (b), tumor stage after receiving NAC (yStage) (c), pathological complete response (pCR) and axillary pCR (d) after receiving NAC in TLS-high and TLS-low groups. e Boxplot showing the TLS density stratified by axillary pathological status after receiving NAC (ypN) in TLS-high and TLS-low groups. f Boxplot showing TLS gene scores in large breast cancer datasets stratified by pCR after receiving NAC. g Boxplot showing TLS gene scores in large database, including breast cancer patients stratified by lymph node metastasis-positive and lymph node metastasis-negative status in the treatment-naive cohort (left), and patients stratified by axillary pCR and non-pCR status in the NAC cohort (right). In the box plots, the centerline represents the median, and the box limits represent the lower and upper quartiles. The upper whisker extends to the minimal, maximum, or third quartile plus the 1.5 × interquartile range (IQR). The lower whisker extends to the maximal, minimum, or first quartile minus 1.5 × IQR. Comparisons were performed using the Mann–Whitney U test. ns not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. HER2 human epidermal growth factor receptor 2, HR hormone receptor, NAC neoadjuvant chemotherapy, pCR pathological complete response, TLS tertiary lymphoid structure, TNBC triple-negative breast cancer, ypN pathological nodal status after receiving NAC, yStage tumor stage after receiving NAC; LM lymph node metastasis Next, we aim to figure out the relationship between the specific number of TLS density and clinicopathological characteristics in TLS-high and TLS-low groups, particularly the residual axillary nodal burden. We observed a greater density of TLSs in triple-negative breast cancer (TNBC) after receiving NAC when other subtypes had a relatively low TLS density (p < 0.05) (Fig. 2b), suggesting that the high TLS density might be due to more intense immune responses after receiving NAC. We also observed a negative correlation between the TLS density and the yAJCC stages of breast cancer patients after receiving NAC (all p < 0.01) (Fig. 2c). In addition, we observed that NAC-treated patients with a pCR exhibited a greater density of TLSs (p < 0.001) (Fig. 2d, left). These results might explain the survival advantage of these patients. Furthermore, TLSs were associated with axillary pCR (ypN0) and non-pCR (p < 0.0001) (Fig. 2d, right). Notably, TLS density was significantly higher in patients with no lymph node metastasis following NAC (all p < 0.05) (Fig. 2e). These results indicated the relationship between the greater density of TLSs and better axillary node prognosis in breast cancer patients after receiving NAC. Subsequently, we used the TLS gene scores to quantify the TLS patterns and predict the real situation of specific individuals because of the heterogeneity among treatment-naive and NAC-treated breast cancer patients. First, we used mRNA expression matrixes from pre-NAC-treated patients in large datasets to analyze the relationship between TLS gene scores and pCR (refer to Methods). We found that patients who achieved pCR after receiving NAC attained higher TLS gene scores compared with non-pCR patients (p < 0.0001) (Fig. 2f). We subsequently investigated the TLS gene scores of patients grouped by pathological axillary nodal status. In the large datasets including the breast cancer mRNA expression matrix after receiving NAC (refer to Methods), patients with an axillary residual nodal burden had lower TLS gene scores (p < 0.0001) (Fig. 2g). However, no differences were observed in patients receiving no prior treatment (treatment-naive) (p > 0.05) (Fig. 2g). Collectively, these results revealed the distribution patterns and prognostic implications of TLSs in NAC-treated breast cancer patients, indicating that compared with those with the absence or lower density of TLSs, patients with a greater density of TLSs in breast cancer after receiving NAC had a lower residual nodal burden. Association between lymphatic vessels and TLSs in the NAC cohort Infiltrating tumor cells can pass through the lymphatic vessels and reach the lymph nodes with lymph fluid [41]. Tumoral lymphatics are considered to undergo dynamic changes that can facilitate tumor metastasis [42]. To understand the changes in lymphatic vessels in breast cancer following NAC, we examined the number and size of lymphatic vessels in the interstitial tissue of primary tumor sites in both NAC and treatment-naive breast cancer patients. We performed lymphatic vessel (PDPN+) IHC staining on breast tumors from both cohorts. We used the hot spot (refer to Methods) to calculate the LVD, LVA, and LVP of the lymphatic vessels in at least five hot spots. We subsequently used semiquantitative automated counting to validate our manual results (refer to Methods). The Bland‒Altman plot revealed favorable consistency between our automatic and manual lymphatic vessel counting based on IHC-stained WSIs (LVD: ICC = 0.9248, 95% CI 0.8669–0.9574, p < 0.0001; LVA: ICC = 0.9984, 95% CI 0.9972–0.9991, p < 0.0001; LVP: ICC = 0.9291, 95% CI 0.8778–0.9594, p < 0.0001) (Supplementary Fig. 1a-c). Next, we analyzed the parameters of the lymphatic vessels on the basis of the results of the IHC staining for PDPN (Fig. 3a). We found that the NAC cohort tended to have significantly larger lymphatic vessels (p < 0.0001), while no differences were found in the LVD (p > 0.05) (Fig. 3b–d). Further correlation analysis revealed that in NAC-treated breast tumor tissues, higher TLS density correlated with a higher LVD (r = 0.6381, p < 0.0001) and a smaller size (LVA, r = − 0.6532, p < 0.0001; LVP, r = − 0.2135, p < 0.0001) (Fig. 3e–g), whereas the correlation was not statistically significant in treatment-naive tissues (LVD, r = − 0.0749, p = 0.7405; LVA, r = − 0.3807, p < 0.0805; LVP, r = 0.0525, p < 0.8165) (Supplementary Fig. 2). Next, we selected 23 patients in the NAC cohort with prior needle biopsies and obtained their paired pre- and post-NAC biopsy samples. We performed hematoxylin‒eosin (H&E) staining and IHC analysis to investigate TLSs and lymphatic vessels in this paired-sample set of female patients before and after receiving NAC (Fig. 4a). We found that the number of TLSs significantly increased after receiving NAC (p < 0.0001) (Fig. 4b). Additionally, a significantly larger lymphatic vessel density, area and perimeter were detected in the breast tumor tissues of patients after receiving NAC (all p < 0.0001) (Fig. 4c). These results validated our observations in the NAC and treatment-naive cohorts, indicating that NAC could induce the presence and abundance of TLSs and the changes in lymphatic vessels.Fig. 3Association of lymphatic vessels and TLSs in treatment-naive patients and NAC patients. a Representative images of immunohistochemical (IHC) staining for podoplanin-positive (PDPN+) lymphatic vessels (arrow) in breast cancer samples from treatment-naive patients and NAC-treated patients. Scale bar, 100 μm. b–d Comparison of the lymphatic vessel density (LVD) (b), lymphatic vessel area (LVA) (c), and lymphatic vessel perimeter (LVP) (d) between treatment-naive patients (n = 304) and NAC-treated patients (n = 326). In the box plots, the centerline represents the median, and the box limits represent the lower and upper quartiles. The upper whisker extends to the minimal, maximum, or third quartile plus the 1.5 × interquartile range (IQR). The lower whisker extends to the maximal, minimum, or first quartile minus 1.5 × IQR. Comparisons were made via the Mann–Whitney U test. ns not significant, ****p < 0.0001. e–g Analysis of the correlation between the LVD (e), LVA (f), LVP (g) and TLS density in breast cancer tissues and TLS density from patients after receiving NAC (n = 184). The parameters were quantified using an Olympus BX63 microscope and the image analysis software APtime (version 2.0.0–3596). p-values are indicated. Statistical significance was determined via Pearson’s correlation test. NAC neoadjuvant chemotherapy, TLS tertiary lymphoid structureFig. 4Comparison of TLSs and lymphatic vessels in patients before and after receiving NAC. a Representative images of H&E and immunohistochemical (IHC) staining for TLSs (delineated with black lines) and podoplanin-positive (PDPN+) lymphatic vessels (arrow) in breast cancer samples from patients before (pre-NAC) (above) and after receiving NAC (post-NAC) (below). Scale bars, 1 mm for the low-magnification image, 200 μm for the middle-magnification image (left and middle), and 50 μm for the high-magnification image (right). b Comparison of TLS density in breast cancer patients before (pre-NAC) (n = 23) and after receiving NAC (post-NAC) (n = 23). c Comparison of the lymphatic vessel density (LVD), lymphatic vessel area (LVA), and lymphatic vessel perimeter (LVP) in breast cancer patients before (pre-NAC) (n = 23) and after receiving NAC (post-NAC) (n = 23). The parameters were quantified using an Olympus BX63 microscope and the image analysis software APtime (version 2.0.0–3596). Comparisons were performed using a paired-samples t-test. ns not significant, ****p < 0.0001. NAC neoadjuvant chemotherapy, TLS tertiary lymphoid structure Association of lymphatic vessels and TLSs in treatment-naive patients and NAC patients. a Representative images of immunohistochemical (IHC) staining for podoplanin-positive (PDPN+) lymphatic vessels (arrow) in breast cancer samples from treatment-naive patients and NAC-treated patients. Scale bar, 100 μm. b–d Comparison of the lymphatic vessel density (LVD) (b), lymphatic vessel area (LVA) (c), and lymphatic vessel perimeter (LVP) (d) between treatment-naive patients (n = 304) and NAC-treated patients (n = 326). In the box plots, the centerline represents the median, and the box limits represent the lower and upper quartiles. The upper whisker extends to the minimal, maximum, or third quartile plus the 1.5 × interquartile range (IQR). The lower whisker extends to the maximal, minimum, or first quartile minus 1.5 × IQR. Comparisons were made via the Mann–Whitney U test. ns not significant, ****p < 0.0001. e–g Analysis of the correlation between the LVD (e), LVA (f), LVP (g) and TLS density in breast cancer tissues and TLS density from patients after receiving NAC (n = 184). The parameters were quantified using an Olympus BX63 microscope and the image analysis software APtime (version 2.0.0–3596). p-values are indicated. Statistical significance was determined via Pearson’s correlation test. NAC neoadjuvant chemotherapy, TLS tertiary lymphoid structure Comparison of TLSs and lymphatic vessels in patients before and after receiving NAC. a Representative images of H&E and immunohistochemical (IHC) staining for TLSs (delineated with black lines) and podoplanin-positive (PDPN+) lymphatic vessels (arrow) in breast cancer samples from patients before (pre-NAC) (above) and after receiving NAC (post-NAC) (below). Scale bars, 1 mm for the low-magnification image, 200 μm for the middle-magnification image (left and middle), and 50 μm for the high-magnification image (right). b Comparison of TLS density in breast cancer patients before (pre-NAC) (n = 23) and after receiving NAC (post-NAC) (n = 23). c Comparison of the lymphatic vessel density (LVD), lymphatic vessel area (LVA), and lymphatic vessel perimeter (LVP) in breast cancer patients before (pre-NAC) (n = 23) and after receiving NAC (post-NAC) (n = 23). The parameters were quantified using an Olympus BX63 microscope and the image analysis software APtime (version 2.0.0–3596). Comparisons were performed using a paired-samples t-test. ns not significant, ****p < 0.0001. NAC neoadjuvant chemotherapy, TLS tertiary lymphoid structure Next, we explored the relationship between TLSs and lymphatic vessels. We first conducted RNA-FISH analysis to explore the expression of LEC-specific genes, including PDPN and LYVE1, in the TLS-absent, TLS-low and TLS-high groups (Fig. 5a). The results revealed that a greater density of TLSs in breast tumor tissue sections was significantly associated with increased transcript levels of PDPN and LYVE1 (all p < 0.05) (Fig. 5b). Next, we analyzed the correlation between LEC-specific markers (PROX1, PDPN, and LYVE1) and TLS gene scores in treatment-naive and NAC-treated breast tumor tissues using large datasets (refer to Methods). We found that TLS gene scores were not significantly correlated with the expression levels of LEC-specific markers in breast cancer tissues from the Desmedt database (Fig. 6a), whereas TLS gene scores were positively correlated with the expression levels of PROX1 (R = 0.196, p = 0.012) and PDPN (R = 0.219, p = 0.005) in the Sammut database (Fig. 6b), along with the expression levels of PROX1 (R = 0.233, p < 0.0001), PDPN (R = 0.315, p < 0.0001) and LYVE1 (R = 0.190, p < 0.0001) in the TCGA breast cancer dataset (Fig. 6c). This analysis revealed that LEC-specific gene expression may be associated with TLS gene scores in breast cancer, supporting the hypothesis that lymphatic remodeling might facilitate TLS formation and maturation, as discussed in the context of lymphatic vessel-mediated immune cell trafficking. Overall, we concluded that TLS density was associated with the density and size of lymphatic vessels, particularly in NAC-treated tissues. Additionally, NAC might affect the correlation between lymphatic vessels and TLS patterns. These data led us to hypothesize that lymphatic vessels in breast cancer might play critical roles in TLS-induced immune responses.Fig. 5Quantitative analysis of RNA fluorescence in situ hybridization (RNA-FISH) in lymphatic vessels in the immunohistochemical (IHC) cohort. a Representative H&E images of TLSs in breast cancer samples from the TLS-absent, TLS-low, and TLS-high groups (above) with RNA-FISH images of tissue sections (below). The TLSs are delineated with black lines in H&E images. RNA-FISH was performed using probes for PDPN mRNA (red) and LYVE1 mRNA (green), with DAPI (blue) counterstaining. Scale bars, 200 μm for H&E images and low-magnification RNA-FISH images, 20 μm for high-magnification RNA-FISH images. b Quantitative analysis of RNA-FISH signals across the TLS-absent, TLS-low, and TLS-high groups. The data are representative of three independent experiments, with n = 3 replicates each. The data are presented as the mean values ± SEMs. The relative fluorescence intensity was quantified using Image-Pro Plus 6 software. Comparisons were performed using one-way ANOVA. **p < 0.01, ***p < 0.001, ****p < 0.0001. MFI mean fluorescence intensity, TLS tertiary lymphoid structureFig. 6Association of specific gene expressions in lymphatic vessels and TLSs in the large datasets. a–c Pearson correlation between TLS gene scores and the expression of lymphatic-specific markers, including PROX1, PDPN, and LYVE1, in patients with breast cancer from large breast cancer databases. The correlations between TLS gene scores and the expression of lymphatic-specific markers in patients with breast cancer before receiving NAC in the Desmedt dataset (n = 120) are shown in (a). The correlations between TLS gene scores and the expression of lymphatic-specific markers in patients with breast cancer before receiving NAC in the Sammut dataset (n = 162) are shown in (b). The correlations between TLS gene scores and the expression of lymphatic-specific markers in breast cancer patients from the TCGA breast cancer dataset (n = 1231) are shown in (c). p-values are indicated (a–c). Statistical significance was determined via Pearson’s correlation test (a–c) Quantitative analysis of RNA fluorescence in situ hybridization (RNA-FISH) in lymphatic vessels in the immunohistochemical (IHC) cohort. a Representative H&E images of TLSs in breast cancer samples from the TLS-absent, TLS-low, and TLS-high groups (above) with RNA-FISH images of tissue sections (below). The TLSs are delineated with black lines in H&E images. RNA-FISH was performed using probes for PDPN mRNA (red) and LYVE1 mRNA (green), with DAPI (blue) counterstaining. Scale bars, 200 μm for H&E images and low-magnification RNA-FISH images, 20 μm for high-magnification RNA-FISH images. b Quantitative analysis of RNA-FISH signals across the TLS-absent, TLS-low, and TLS-high groups. The data are representative of three independent experiments, with n = 3 replicates each. The data are presented as the mean values ± SEMs. The relative fluorescence intensity was quantified using Image-Pro Plus 6 software. Comparisons were performed using one-way ANOVA. **p < 0.01, ***p < 0.001, ****p < 0.0001. MFI mean fluorescence intensity, TLS tertiary lymphoid structure Association of specific gene expressions in lymphatic vessels and TLSs in the large datasets. a–c Pearson correlation between TLS gene scores and the expression of lymphatic-specific markers, including PROX1, PDPN, and LYVE1, in patients with breast cancer from large breast cancer databases. The correlations between TLS gene scores and the expression of lymphatic-specific markers in patients with breast cancer before receiving NAC in the Desmedt dataset (n = 120) are shown in (a). The correlations between TLS gene scores and the expression of lymphatic-specific markers in patients with breast cancer before receiving NAC in the Sammut dataset (n = 162) are shown in (b). The correlations between TLS gene scores and the expression of lymphatic-specific markers in breast cancer patients from the TCGA breast cancer dataset (n = 1231) are shown in (c). p-values are indicated (a–c). Statistical significance was determined via Pearson’s correlation test (a–c) Discussion In recent years, increasing evidence has emphasized that TLSs can serve as potential biomarkers associated with favorable clinical outcomes in patients with tumors [9, 12, 13]. Our previous study and others have demonstrated that TLSs emerge in the inflammatory microenvironments after receiving NAC [14, 25, 26]. Chemotherapy reshapes the local immune infiltrate, shifting it from its baseline state to a treatment-induced inflammatory profile, which may lead to behavioral shifts in cells and the expression of genes that change the inflammatory environment [14, 21–24]. In breast cancer, the accumulation of ICOSL+ B cells in TLSs is associated with NAC effectiveness and improved patient prognosis [14]. Similarly, patients with TLS-present tumors, including those tumors that are resistant to standard treatments, have been reported to respond better to NAC compared with those with TLS-absent tumors exhibiting an immunosuppressive microenvironment [43]. The absence of these structures in prechemotherapy samples even indicated that TLSs could be induced during chemotherapy treatment [26]. By analyzing the clinicopathological information of breast cancer patients, we described the associations between TLSs and the clinicopathological characteristics of NAC-treated patients. Furthermore, the TLS gene scores facilitated the validation of the specific patterns and gene clusters of TLSs in patients with breast cancer treated with NAC. Additionally, we demonstrated the association between lymphatic vessels and TLSs through our two cohorts and large breast cancer datasets. Our study initially captured the presence of TLSs in breast cancer patients after receiving NAC and no prior treatments. We outlined the proportions of TLS-present and TLS-absent patients in the NAC cohort and treatment-naive cohort. TLSs were detected in 184 out of 326 patients in the NAC cohort (56.4%) and 22 out of 304 patients in the treatment-naive cohort (7.2%). Additionally, paired-sample specimens from patients with breast cancer before and after receiving NAC confirmed the potential effect of NAC on TLS formation. These results are consistent with the results of our previous study [14] and the findings reported by Kuwabara et al. [44] and Yang et al.[45], indicating that TLSs more commonly exist in breast cancer samples after receiving NAC. NAC, originally designed to debulk tumors and eradicate micrometastatic disease, has recently been recognized to remodel the TME [14, 17–19, 21, 22]. Previous studies have suggested that NAC can promote TLS development and maturation in tumors [25, 26]. This phenomenon may be due to the effect of chemotherapy on the TME in breast cancer. Evidence indicates that chemotherapy can provide long-term benefits when the disease remains controlled after treatment ends and is even cured, which depends on a lasting anticancer immune response [1]. Additionally, NAC can activate the antitumor immune response of immune cells in the breast TME [14, 21–24]. Thus, the high rate of TLS positivity may be due to these potential effects of NAC on breast cancer. The axillary status after receiving NAC has been consistently reported as a robust prognostic factor for patient survival independent of the primary tumor response and for determining further axillary treatment, including adjuvant therapy, which is important for patients with residual cancer burden (RCB)[46–48]. While more than half of patients with negative axillary lymph nodes achieve axillary pCR after neoadjuvant therapy, nodal pCR occurs in only approximately 40% of clinically node-positive patients [49]. Studies have shown that an increased tumor-infiltrating lymphocyte (TIL) concentration is associated with no axillary nodal burden after receiving NAC [50, 51]. TLSs are characterized by high-density and organized TIL infiltration [16]. It remains largely unknown whether the residual axillary nodal burden is correlated with TLS abundance following NAC in patients with breast cancer. In our study, we found that TLSs correlated with lower residual nodal disease after receiving NAC. Patients with a higher TLS density or gene expression showed reduced nodal involvement, which is consistent with findings in other cancers [15, 27, 52, 53]. Incorporating TLS assessment alongside nodal status may increase prognostic accuracy and help identify patients who may benefit from de-escalated axillary treatment or tailored adjuvant therapies. Additionally, we evaluated lymphatic vessels in the primary tumor sites in the two cohorts. Tumor-associated lymphatic vessels are considered to undergo dynamic changes that can facilitate tumor metastasis, including lymphangiogenesis and lymphatic enlargement [42]. Tumor-draining sentinel lymph nodes show enhanced lymphangiogenesis even before cancer metastasis, forming a permissive niche for metastatic survival [41]. Moreover, tumor progression and the response to therapy frequently involve modulation of the immune system or stromal components by nonclassical tumor regulators [54, 55], such as substances derived from lymphatic vessels. Tumor cells may rely on chemokine gradients as guidance cues and enter lymphatic vessels through endothelial junctions or induced discontinuities. Targeting lymphatic vessels has significant potential for the formation of TLSs through TME-induced antitumor immunity, thereby improving clinical treatment response rates in patients [56]. For example, local anesthetics can serve as microenvironment-targeting agents in modulating immune response, stress signaling, and cellular interactions to suppress tumor progression [57]. However, the crosstalk between the inflammatory TME and the local lymphatic vasculature and its impact on immune regulation remain poorly understood. NAC may enhance lymphatic vessel involvement in antitumor immunity. Some researchers have reported that chemotherapy may modify the lymphatic drainage patterns through the shrinkage and fibrosis of lymphatic vessels, as well as by obstructing lymphatic channels with cellular material or tumor emboli [58, 59]. However, in our current study, lymphatic vessels under NAC led to enlarged lymphatic lumens without changing their density, which is consistent with the findings of preclinical studies [60–62]. This enlargement may be driven by chemotherapy-induced activation of the VEGF-C/VEGFR3 pathway, injury-mediated lymphangiogenesis, or adaptive drainage of increased angiogenic flow [63–65]. The greater lumen area of the lymphatic vessels possibly reflects the need to drain the increased blood supply associated with angiogenesis [60]. However, drugs such as paclitaxel can also mildly suppress lymphangiogenesis via LEC autophagy [66]. Additionally, VEGF-C secreted by tumor cells and proinflammatory cytokines from immune cells may dilate vessels and reduce their pumping capacity [67]. Studies have also shown the potential relationship between lymphatic vessels and TLSs in inflammatory diseases and tumors [68–70]. Lymphatic vessels that could favor emigration of educated lymphocytes to draining lymph nodes are found in the vicinity of TLSs, leading to efficient trafficking of tumor-derived antigens and immune cells, which are essential for TLS formation [68]. Thus, lymphatic vessels can magnify the local recruitment and activation of immune cells and the anti-tumor immune responses. However, the relevance of lymphatic vessels and TLS abundance has not been studied in breast cancer after receiving NAC. Our analysis demonstrated that patients with a high density of TLSs after receiving NAC had significantly denser and smaller lymphatic vessels. Such correlations were not found in the treatment-naive cohort. This might be due to the dynamic interaction between TLSs and lymphatic vessels in the inflammatory microenvironment induced by NAC. Inflammatory microenvironments and tissue remodeling can mediate immune modulation. Helicobacter pylori infection creates an inflammatory microenvironment that not only regulates inflammatory pathways but also induces remodeling of immune cells by enhancing gastric cancer cell migration [71, 72]. During inflammation, newly formed TLSs can regulate dynamic lymphatic changes in different phases through the activity of various cytokines [73]. In mice with impaired pulmonary lymphatic flow, TLSs form in proximity to dilated lymphatic vessels and ultimately cause chronic lung inflammation, indicating a close relationship between lymphatic vessels and TLSs in the inflammatory microenvironment [74]. Besides, it has been reported that the formation of TLSs involves the generation of lymphatic vessels in Hashimoto’s thyroiditis [75]. Considering the role of lymphatic vessels in lymph node metastasis, along with the effect of NAC on lymphatic vessels and the close relationship between lymphatic vessels and TLSs, further investigation of the influence of NAC-altered lymphatics on TLS formation is warranted to reveal new strategies for improving immune responses and treatment outcomes. The limitations of this study include a lack of TLS sites, maturity analyses and a dissection of the immune cells in TLSs. Additionally, the retrospective nature of this study introduces inherent biases and limitations. Despite these limitations, our findings demonstrate a correlation between TLS density and residual axillary nodal burden, and a relationship between TLS density and the abundance and size of lymphatic vessels. Mechanistic studies are warranted to dissect the interaction between TLSs and lymph nodes with residual diseases as well as lymphatic vessels. NAC can enhance immune responses by remodeling the TME [14, 17–19, 21, 22]. Multiple strategies have emerged to improve therapeutic outcomes in tumor chemotherapy through comprehensive analysis of the TME following treatment. For instance, several high-throughput approaches, such as CRISPR screening, have been employed to identify genetic determinants underlying therapeutic resistance within the TME[76]. Integrating histological assessment with molecular profiling of the TME and artificial intelligence-based omics approaches may facilitate more precise evaluation of TLSs and support the development of personalized NAC regimens [77, 78]. Future precision medicine strategies, including targeted therapy, stem cell therapy, and chemodynamic therapy targeting the TME and modulating molecular mechanisms, hold promise for reversing the resistance to NAC and guiding individualized therapy in patients with breast cancer. Conclusions Our study highlighted that a higher density of TLSs in patients with breast cancer was associated with a lower residual axillary nodal burden and that the density and size of lymphatic vessels were correlated with TLS density in patients who underwent NAC. Moreover, we demonstrated the relationship between lymphatic vessel markers and TLS patterns on the basis of TLS gene scores among large datasets, which may increase confidence in the response to NAC and TLS-induced treatment strategies in breast cancer patients with axillary residual nodal burden. Supplementary Information Supplementary material 1. Bland‒Altman plot for agreement between manual lymphatic vessel counting and automated lymphatic vessel counting (n=50). (a-c) Manual counting of lymphatic vessel density (LVD) (a), lymphatic vessel area (LVA) (b), and lymphatic vessel perimeter (LVP) (c) on the basis of IHC-stained whole-slide images (WSIs). The red solid line of each plot shows the mean of the difference between manual lymphatic vessel counting and automated lymphatic vessel counting. The blue dashed lines show the upper/lower bounds of the 95% limits of agreement (95% LoA). Abbreviations: CI, confidence interval; ICC, intraclass correlation coefficient.Supplementary material 2. Analysis of the correlation between the lymphatic vessel density (LVD), lymphatic vessel area (LVA), lymphatic vessel perimeter (LVP) and tertiary lymphoid structure (TLS) density in breast cancer tissues from treatment-naive patients (n = 22).Supplementary material 3. Assessment of the inter-rater agreement rate between ZW and KL for the TLS screening.Supplementary material 4. Target sequences of RNA probes. Supplementary material 1. Bland‒Altman plot for agreement between manual lymphatic vessel counting and automated lymphatic vessel counting (n=50). (a-c) Manual counting of lymphatic vessel density (LVD) (a), lymphatic vessel area (LVA) (b), and lymphatic vessel perimeter (LVP) (c) on the basis of IHC-stained whole-slide images (WSIs). The red solid line of each plot shows the mean of the difference between manual lymphatic vessel counting and automated lymphatic vessel counting. The blue dashed lines show the upper/lower bounds of the 95% limits of agreement (95% LoA). Abbreviations: CI, confidence interval; ICC, intraclass correlation coefficient. Supplementary material 2. Analysis of the correlation between the lymphatic vessel density (LVD), lymphatic vessel area (LVA), lymphatic vessel perimeter (LVP) and tertiary lymphoid structure (TLS) density in breast cancer tissues from treatment-naive patients (n = 22). Supplementary material 3. Assessment of the inter-rater agreement rate between ZW and KL for the TLS screening. Supplementary material 4. Target sequences of RNA probes.