Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Immunoarchitectural patterns as potential prognostic factors for invasive ductal breast cancer

Immunoarchitectural patterns as potential prognostic factors for invasive ductal breast cancer www.nature.com/npjbcancer ARTICLE OPEN Immunoarchitectural patterns as potential prognostic factors for invasive ductal breast cancer 1,2,3 1,3 2,3 1 2 2 2 2 1✉ Xue Du , Zhe Zhou , Yun Shao , Kun Qian , Yongfang Wu , Jun Zhang , Miao Cui , Jingjing Wang , Shengqi Wang and Yanhong Tai Currently, tumor-infiltrating lymphocytes (TILs) in invasive breast cancers are assessed solely on the basis of their number, whereas their spatial distribution is rarely investigated. Therefore, we evaluated TILs in 579 patients with invasive breast cancer of no special type (IBC-NST) with a focus on their spatial distributions in tumor center (TC) and invasive margin (IM). We also assessed a new factor, namely para-tumor infiltrating lymphocytes (PILs) in the para-tumor lobular area (Para). Five immunoarchitectural patterns (IPs) were observed, which were significantly associated with clinicopathological features, especially molecular subtypes, histological grades, clinical stages, and programmed death-ligand 1 (PD-L1) expression. High-TIL density (IP1/2) correlated with favorable disease-free survival (DFS) in TNBC patients (p = 0.04), but opposite results were observed for luminal B subtype patients (both the lowest TIL and PIL densities (IP5) correlated with good DFS, p = 0.013). Luminal B patients with high TILs in the IM and low TILs in the TC (IP3) exhibited the worst DFS, whereas those with low TILs (similar to IP5) and high PILs (IP4) exhibited poor DFS. We also identified TIL subpopulations with significantly different IPs. Our findings suggest that IP can be a potential prognostic factor for tumor immunity in IBC. npj Breast Cancer (2022) 8:26 ; https://doi.org/10.1038/s41523-022-00389-y INTRODUCTION In this study, we investigated hematoxylin and eosin (H&E)- stained sections of 579 tissue samples from invasive breast cancer Breast cancers are clinically and molecularly heterogenous, with of no special type (IBC-NST) patients to define tumor immu- 5–10 intrinsic subtypes . Each subtype displays varied molecular 2,3 noarchitectural patterns (IPs) and TIL density. Therefore, compre- characteristics that form the basis for therapeutic resistance and hensive analysis of the identified IPs was performed with respect different therapeutic strategies . Immunotherapy and combined to lymphocyte density, location, immunophenotyping, and neoadjuvant chemotherapy are being aggressively developed, combined histopathological characteristics (such as the histologi- with anti- programmed death-ligand 1 (PD-L1) exhibiting strong cal grade, clinical stage, molecular type, and survival status) in immunomodulatory therapeutic potential against breast cancer . patients with IBC-NST. A pre-existing immunological response might enhance the 5–7 efficacy of conventional cytotoxic chemotherapy . However, despite accumulating evidence, the translation from basic tumor immunology to clinical practice remains problematic . PD-L1 and RESULTS tumor mutational burden (TMB)-based immunotherapeutic clinical Stratification of IBC-NSTs into five IPs trials have shown favorable results in a small subset of invasive We assessed 579 primary IBC-NST cases for tumor immunity and breast cancer (IBC) patients, mainly triple-negative breast cancer grouped them into five IPs, as indicated in the flowchart (Figs. 1 (TNBC) patients . Previous studies have shown that high count of and 2). IP2 (19/579, 3.28%) had the least number of cases, followed tumor-infiltrating lymphocytes (TILs) cannot constantly warrant a by IP1 (69/579, 11.92%), IP3 (110/579, 19.00%), IP4 (130/579, good outcome in all IBC patients. In luminal-HER2-negative patients, high TIL count is considered an adverse prognostic 22.45%), and IP5 (251/579, 43.35%) (Table 1). We displayed the factor for survival ; however, the TILs should be studied with a cross-referenced H&E and leukocyte common antigen (LCA) new perspective for a comprehensive understanding of the tumor stained sections, and TissueGnostics images of typical cases to microenvironment. highlight the distinct differences of five IPs in Fig. 3. A recent investigation into more reliable predictors revealed that immune contextures, such as TIL density and spatial Lobular involvement, cancerous embolus, and histological localization, are associated with clinicopathological characters grade and PD-L1 expression based on molecular subtypes, and were IP4 had significantly higher frequency of lobular involvement therefore considered appropriate immunotherapeutic candidates. (121/130, 93.08%) compared with that in the other four IPs. A However, the association of clinicopathological characters with para-tumor infiltrating lymphocytes (PILs) located in the para- similar trend was also observed with IP1 (47/69, 68.12%), which tumor lobular area (Para) remains uncertain. Therefore, the had high PIL counts similar to IP4, but unlike IP2 (5/19, 26.32%), quantitative molecular and spatio-morphological parameters of IP3 (46/110, 41.82%), and IP5 (98/251, 39.04%) (Fig. 4a, b, infiltrating lymphocytes interactions should be explored to Supplementary Table 1). Moreover, the cancerous embolus ratio was also significantly higher in IP4 (42/130, 32.31%) than that in improve the identification of predictive markers. 1 2 3 Beijing Institute of Radiation Medicine, Beijing, PR China. Department of Pathology, Fifth Medical Center of Chinese PLA General Hospital, Beijing, PR China. These authors contributed equally: Xue Du, Zhe Zhou, Yun Shao. email: sqwang@bmi.ac.cn; taiyanhong29@163.com Published in partnership with the Breast Cancer Research Foundation 1234567890():,; X. Du et al. Fig. 1 Outline of the traditional software-assisted assessment procedure used for analyzing TILs and PILs in IBC-NST samples. TILs tumor- infiltrating lymphocytes, PILs para-tumor infiltrating lymphocytes, IBC-NST invasive breast cancer of no special type. IP1 (8/69, 11.59%), IP3 (23/110, 20.91%), and IP5 (47/251, 18.73%) grade 1 cases were found in the IP1 and IP2 groups, and grade 3 (Fig. 4c, d, Supplementary Figure 2, Supplementary Table 1). cases (62.32% in IP1 and 52.63% in IP2) were more common than The histological-grade distribution differed significantly grade 2 cases (37.68% in IP1 and 47.37% in IP2). In contrast, in IP5, between each of the five IPs (χ² = 84.84, p < 0.001; Fig. 4e, f). No grade 1 (14.34%) and grade 2 (73.71%) cases accounted for npj Breast Cancer (2022) 26 Published in partnership with the Breast Cancer Research Foundation 1234567890():,; X. Du et al. one case each of luminal A subtype was found in IP1 and IP2. Conversely, patients with the TNBC or HER2 + subtype accounted for a greater proportion of IP1 (19/69, 27.54% and 14/69, 20.29%, respectively) compared to that of IP2 (each 3/19, 15.79%), IP3 (19/ 110, 17.27% and 10/110, 9.09%, respectively), IP4 (19/130, 14.62% and 10/130, 7.69%, respectively), and particularly IP5 (6/251, 2.39% and 10/251, 3.98%, respectively). Luminal B subtype and luminal- HER2 subtype cases were distributed evenly into the five IPs (Supplementary Table 1). Pearson’s chi-square test of molecular subtypes revealed significant differences between IPs (χ² = 88.097, p < 0.001; Fig. 5f, g), and IP5 contained significantly more mild subtype cases than other IPs, whereas aggressive subtype cases were significantly more common in IP1 than IP3–IP5. The IP2 group showed no significant difference in the molecular-subtype distribution with other IPs, except for IP5 (Supplementary Fig. 2). The distribution of IPs in the luminal B subtype differs significantly in three histological grades (χ = 18.494, p = 0.018). Pattern integration The clinicopathological characteristics of the 579 patients were summarized by heatmapping (Fig. 5h). Although the IP2 cases shared similar prognosis-related clinicopathological characteristics with the IP1 cases (such as a higher histological grade, more aggressive molecular subtype, and lower clinical stage), the differences between IP2 and IP3–5 were less significant than those between IP1 and IP3–5 (Supplementary Figure 2). As there Fig. 2 Flow chart of the process used to identify five different immunoarchitectural patterns. IP immunoarchitectural pattern. were only 19 IP2 cases (and even less when subdivided), the low degree of significance may partially reflect the small sample size. 88.05% of all cases, whereas grade 3 cases accounted for only We found that the curve for IP5 was unique, whereas the curve 11.95% of the cases (Supplementary Table 1). The grade tendencies of IP1 and IP2, as well as IP3 and IP4, were highly distribution was not different between IP3 and IP4, both of which similar (Supplementary Figure 3). These three characteristics are exhibited more grade 3 cases and fewer grade 1 cases compared the major factors that affect disease-free survival (DFS); thus, we to IP5. No significant differences were identified between the IP1 merged the IPs into three groups (IP1/2, IP3/4, and IP5) as and IP2, and IP2 and IP3 pairs (Supplementary Fig. 2). prognostic groups for subsequent DFS analysis (Fig. 6). Clinical stage, tumor size, and lymph node metastasis DFS analysis Tumor-size measurements were missing in one case each for IP4 In this study, Formalin-fixed, paraffin-embedded (FFPE) tissues and IP5; thus, these cases were omitted from subsequent analyses. were collected within 3 years to meet the quality requirements for Most patients (496) exhibited early stage (stage I, 269 cases or nucleic acid-isolation for genome-wide mRNA-expression analysis stage II, 227 cases) disease, and only 81 patients exhibited late- and exome sequencing. Given the short duration of the disease, stage disease (stage III, 74 cases or stage IV, 7 cases). A significant we only assessed the IPs as a prognostic factor for DFS, with a difference was found between patients exhibiting early- and late- median survival time of 91.63 months (IQR 69.40–113.86). No stage disease among the 5 IPs (χ² = 12.055, p = 0.017). The significant association was found with DFS among the five IPs. proportion of late-stage cases in IP1 (2.90%) was significantly When stratified by molecular subtypes, the three merged-IP lower than that in IP3 (19.09%), IP4 (12.40%), and IP5 (16.40%) (Fig. groups exhibited significant associations with DFS (log-rank = 4g, h, Supplementary Table 1). Only one late-stage case was found 3.054, p = 0.217; Breslow = 8.724, p = 0.013; Fig. 6a–e). In the in IP2 (1/19, 5.26%), although it was not significantly different from luminal B subtype, all three merged IPs (log-rank = 8.711, p = other IPs (Supplementary Fig. 2). 0.013; Fig. 6b) and all five individual IPs (log-rank = 10.121, p = Tumor sizes differed significantly between the 5 IPs (χ = 0.038; Fig. 6f) showed significant associations with DFS, where IP5 22.035, p < 0.001), with the lowest median tumor size in IP4 and was associated with a favorable outcome and IP3 was associated the highest in IP2 (Fig. 4i, Supplementary Fig. 2, Supplementary with the worst DFS. Patients in the IP4 group (distinguished from Table 1). The differences were analyzed in terms of the number of lymph node metastases, but no statistical significance was found those in the IP5 group by high PIL count) had the same poor DFS (χ = 3.994, p = 0.407. Fig. 4j, Supplementary Fig. 2). rate as those in the IP3 group. In contrast, among patients with the TNBC subtype, those in the merged IP1/2 group had significantly improved DFS than patients in the IP5 group (log-rank = 6.419, Molecular subtype p = 0.040; Fig. 6e). Similar trends were observed for patients with Immunohistochemistry (IHC)-staining of samples from 579 the luminal B-HER2 and HER2+ subtypes, but the association with patients revealed that the ER- and PR-expression levels gradually DFS was not significant for either subtype (Fig. 6c, d). Among increased from IP1 to IP5 (Fig. 5a, b, Supplementary Fig. 2), patients with the luminal A subtype, only two were included in the whereas the proliferation indicator Ki-67 revealed an opposite IP1 and IP2 groups, and the DFS curves of the merged-IP groups trend (Fig. 5c). HER2 + cases were significantly lower in IP5 than in overlapped without a significant association in terms of DFS (Fig. the other IPs (Fig. 5d, e). 6a). Both univariable and multivariable analyses showed that the Among the 136 luminal A subtype samples, most were distributed in IP5 (88/136, 64.71%), followed by IP4 (27/136, IP had no significant associations with DFS (Supplementary 19.85%), and IP3 (19/136, 13.97%) (Supplementary Table 1). Only Table 2). Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 26 X. Du et al. Table 1. Population clinicopathological characters. Characteristics No. Percentage Characteristics No. Percentage Age ≤35 32 5.53% TNM I 269 46.46% 35–50 218 37.65% II 227 39.21% ≥50 329 56.82% III 74 12.78% IV 7 1.21% Sex Female 578 99.83% Unknown 2 0.35% Male 1 0.17% ER status Negative 118 20.38% Postmenopausal not yet 312 53.89% Positive 461 79.62% Postmenopausal 260 44.91% Unknown 6 1.04% PR status Negative 170 29.36% Null (male) 1 0.17% Positive 409 70.64% Family genetic history not 560 96.72% HER2 status Negative 474 81.87% Family genetic history 15 2.59% Positive 105 18.13% Unknown 4 0.69% Ki67 ≤14% 151 26.08% Tumor size T1( ≤ 2) 379 65.46% å 14% 428 73.92% T2(2–5) 155 26.77% T3( ≥ 5) 10 1.73% Molecular type Luminal A 136 22.70% T4 33 5.70% Luminal B 272 45.41% Unknown 2 0.35% Luminal HER2 58 9.68% HER2+ 47 7.85% Lymho node N0 377 65.11% TNBC 66 11.02% N1 143 24.70% N2 38 6.56% Immuno- architectural Pattern 1 69 11.92% N3 21 3.63% 2 19 3.28% 3 110 19.00% Metastasis M0 572 98.79% 4 130 22.45% M1 7 1.21% 5 251 43.35% Spatial differences in the immune cell subpopulations subtype on PD-L1 expression in lymphocyte predominant breast cancer (LPBC) patients. PD-L1 was mainly expressed in immune We selected 77 typical cases of different IPs (IP1 = 28, IP2 = 19, cells (ICs) (103/208, 49.52%), including lymphocytes, macrophages, IP3 = 17, and IP4 = 13) to compare the subpopulations of TILs dendritic cells, plasma cells, and granulocytes, but was rarely located in stromal tumor center (TC) invasive margin (IM), and PILs in Para. The TIL subpopulations in the TCs and IMs of IP1/2 group expressed in tumor cells (13/208, 6.25%). Positive ICs were characterized by dark brown punctate or linear membrane were averaged to obtain total TIL-subpopulation densities to staining. The spatial distributions of positive ICs resembled those represent “hot tumors” group. Likewise, we combined each PIL subpopulation (CD4, CD8, and CD20) of IP1/4 group to represent of the IPs (Fig. 7g–k). “hot Para” group. We compared the ratio of T cells (CD4+ CD8+) PD-L1-expression levels were significantly higher in the TNBC to B cells (CD20+), and the ratio of cytotoxic T cells (CD8+)to subtype than in the HER2+ subtype (p = 0.0031, n = 170; Fig. 7l). T cells in these groups. We found that the proportion of T cells PD-L1 positivity was significantly associated with IPs (p < 0.001, (z = −3.455, p = 0.001) and cytotoxic T cells (z = −2.448, p = n = 208; Fig. 7m), where ICs in both the IP1 (84.06%) and IP2 0.014) in “hot tumor” group was significantly higher than that in (89.47%) groups showed the highest positive rates, followed by “hot Para” group. In contrast, the proportion of B cells was IP3 (31.71%) and IP4 (31.82%), while ICs in the IP5 group had the significantly higher in “hot Para” group than that in “hot tumor” lowest expression rate (2.86%) (Supplementary Table 3). The same group (z = −3.112, p = 0.002; Fig. 7a–c). tendency was also observed in the TNBC (p < 0.001, n = 66) and We then compared the ratio of T cells to B cells, and the ratio of HER2+ subtypes (p < 0.001, n = 104; Fig. 7n, o). In cohort 2, PD-L1 cytotoxic T cells to T cells for the following two pairs: IM of IP3 and expression in ER+/HER2− patients was high in both IP1 and IP2 IP1/2; IM of IP3 and Para of IP1/4. No significant differences were groups, with no significant difference (p = 0.31, n = 38; Fig. 7p). found (Fig. 7d–f). Survival analysis revealed that in all 208 cases, including 66 TNBC cases, a PD-L1-positive status was significantly associated The PD-L1-positive rate differed significantly between IPs and with a better prognosis (p = 0.018 and p = 0.020, respectively; Fig. correlated with DFS in all cases and TNBC subtypes 7q, r). No predictive significance was found for survival in patients with the HER2+ and ER+/HER2− subtypes (Fig. 7s, t). To eliminate From the 579 IBC-NSTs cases, 208 typical cases were selected for the influence of TIL count on DFS in TNBC, we subdivided the PD-L1 staining and were divided into two cohorts. Cohort 1 contained all enrolled patients with TNBC and HER2+ cases (with TNBC cases into those with high-TIL count (P1/2) and low-TIL all IPs). Cohort 2 contained all ER+/HER2− patients with IP1 and count (P3/4/5). No significant predictive value was found for either IP2 was used as a control to eliminate the effect of molecular group, based on PD-L1 expression (Fig. 7u, v). npj Breast Cancer (2022) 26 Published in partnership with the Breast Cancer Research Foundation X. Du et al. Fig. 3 Schematic diagram of immune cells’ distribution and morphology in five IP patterns of breast cancer tissues. H&E-stained sections, IHC sections, and TissueGnostics analysis of IP1 (a–c), IP2 (d–f), IP3 (g–i), IP4 (j–l), and IP5 (m–o). The IHC sections were stained for the LCA marker. The red lines in the IHC sections show the margins of the tumor areas. TissueGnostics analysis of the yellow frames in panels b, e, h, k, and n is shown in c, f, i, l, and o, respectively. The black and red lines shown for the TissueGnostics analysis represent the margins of the tumor areas and para-tumors, respectively. H&E hematoxylin and eosin, IHC immunohistochemistry, IP immune pattern, LCA leukocyte common antigen. Tumor-mutation signature and immune-response gene mutation rate in the IP5 group than in the other IP groups, expression including breast-cancer-specific oncodriver genes (TP53, ERBB2, We selected 40 cases from the five IPs to detect TMBs based on MAP3K4, BRCA1, ERBB4, and PIK3CA). By comparing the expression of immune-response genes whole-exome sequencing. No significant differences were between TCs and IMs in the IP1/2 (n = 6) and IP1/4 Para (n = 8) observed (Supplementary Figure 4). We then compared the groups, we identified a cluster of highly expressed genes in IP1/2 differences between oncodriver genes using the Catalog of Somatic Mutations in Cancer database (http://cancer.sanger.ac. TCs and IMs (Supplementary Fig. 4), including T cell activation- uk/census) and found that 44% (141/320) of the genes had a lower associated genes (CD8A, CXCR6, IL-12A, IL-8, and EBI3), the immune Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 26 X. Du et al. Fig. 4 Major clinical characteristics analysis in five immunoarchitectural patterns. Lobular involvement, cancerous embolus, tumor histological grade, clinical stage, tumor size, and number of metastatic lymph nodes in five immunoarchitectural patterns, in terms of distributions (a, c, e, g, i, j) and percentages (b, d, f, h). The correlations between patterns and categorical variables (including tumor grades, molecular subtypes, lobule cancerization, and vascular invasion) were analyzed by the χ² test (for trends) or Fisher’s exact test. Continuous variables (tumor size and number of metastatic lymph nodes) were assessed by the Kruskal–Wallis H tests under K independent-sample tests. npj Breast Cancer (2022) 26 Published in partnership with the Breast Cancer Research Foundation X. Du et al. Fig. 5 ER, PR, Ki-67 expression and HER2 status in five immunoarchitectural patterns and heatmap of all enrolled patients’ clinical characteristics. ER (a), PR (b), Ki-67 (c) expression and HER2 (d) status in five immunoarchitectural patterns The percentages of HER2-positive cases for each pattern (e). Distributions (f) and percentages (g) of molecular subtypes in each of the five IPs. h Heatmap of the clinical characteristics of all 579 patients included in this study. Molecular type, 1, luminal A subtype; 2, luminal B subtype; 3 represent luminal B HER2 subtype; 4 represent HER2 subtype; and 5 represent TNBC subtype. checkpoint gene ICOS (also known as CTLA4), and MHC II true. No cases with high-TC TIL density and low IM density were molecules, such as HLA-DRB1. Equivalently, high expression of found. Tumors with high-TIL densities in both areas fulfill the LPBC genes related to B-cell activation—including CD40LG and CD79B— criteria; thus, TCs provide sensitive and corroborative evidence to MHC molecules, CD1D, and B/T/Th1 cell activation genes— CCL21, confirm the LPBC status, thereby distinguishing IP1/2 from other IPs. In IP1 and IP2, the TIL density and subpopulation in TC and IM IL12B, and PTPRC—was observed in IP1/4 Para (Supplementary showed high accordance with each other. This homogeneity Fig. 4). allowed us to propose a hypothesis that TILs in TCs disperse from IMs, which are regions characterized by higher vessel density . DISCUSSION We then demonstrated that patients with IP1/2 had the best DFS In this study, we used a traditional software-assisted assessment in TNBC and better DFS than that for IP3 (high TIL density in IMs system for assessing IPs of IBC-NSTs on whole slide images (WSIs). only) in luminal B-type cases, further suggesting that high-TIL and This rigorous assessment system enabled us to establish the CD8+ T-cell densities reflect anti-tumor immunity and are 12,13 classification criteria for IPs by calculating the areas of the indicative of a good prognosis . IP1 and IP2 are demarcated lymphatic nuclei instead of those of whole cells and accurately based on higher PIL levels in IP1 compared with those in IP2. distinguished IM and TC regions. The finding that the density of TILs was higher in IMs than in At the morphological level, the densities and spatial distribu- TCs was also reported by Mi et al. , suggesting the distinct role of tions of TILs are heterogenous in different areas (TC, IM, and Para) IMs in the tumor immune architecture. They further discovered of the same block, but FFPE blocks of the same tumor are that the IM is multifaceted and may serve pro- and anti-tumor homogenous, as reported by Mani . In our study, specifically, TIL functions simultaneously with higher CD8+ expression and more count in TCs was similar to those in IMs; high-TC TIL count FOXP3+ cells. We deduced that those cases with higher CD8+ TILs indicated high-IM TIL count, but the reverse was not necessarily in IMs might show the same pattern as IP1/2 group, whereas the Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 26 X. Du et al. Fig. 6 Kaplan–Meier curves for disease-free survival for all patients according to the molecular types. a–e Log-rank (Mantel–Cox) χ² test = 3.054, p = 0.217; Breslow (generalized Wilcoxon) χ² square test = 8.724, p = 0.013; Tarone–Ware χ² test = 8.724 P = 0.049. Kaplan–Meier curves for luminal B subtype in all five patterns f. rest cases belonged to IP3. In IP3, TILs were mainly restricted to Most studies have focused on the immune microenvironment in the IM region and maintained at a medium level (>10%), and their the tumor area. In this study, we expanded the scope of subcellular populations were also altered, with a trend toward lymphocyte evaluation to the Para. As expected, corresponding reduced CD8/(CD4+ 8) ratios and increased CD20/(CD4+ 8) ratios, to “hot” immune tumors, we found another equally obvious “hot” though the alterations were not significant. A similar pattern of immune region localized in the IP1 and IP4 Para with high “altered immune tumors” was reported for colorectal cancer , frequency of lobular involvement. This distinction could be used based on the expression levels and spatial distributions of CD3 to stratify LPBC into IP1 and IP2, and “cold” immune tumors into and CD8. The density and subpopulation heterogeneity of TILs in IP4 and IP5. The fact that no significant difference in the frequency IP1/2 and IP3 indicate their multifaceted functions, which was of lobular involvement occurred between IP1 and IP4 indicates supported by their differential DFS in luminal B-type and that the immune status within a tumor area is not responsible for TNBC cases. driving lobular involvement. Histological grade influence on IP predictive ability in the The subpopulation of IC and gene mutations in different IPs. luminal B subtype. In this study, all five individual IPs of luminal B Previously, it was discovered that infiltrating ICs not only function and merged IPs of TNBC cases showed significant associations to control tumor growth and progression, but also help to create with DFS. Since previous reports confirmed that histological grade an immunosuppressive environment in which the tumor can is a significant independent factor for DFS in luminal but not TNBC thrive . IC subpopulations in the tumor area, combined with patients , it is reasonable to verify whether the predictive ability density and location, could predict the survival of patients with of IPs is attributed to the histological grade. Using the chi-square colorectal cancer more accurately than the classical test, we found that the IP distribution in the luminal B subtype tumor–node–metastasis (TNM) system , where a high CD8/CD3- differs significantly in three histological grades. The DFS differ- density ratio correlated with a good prognosis . However, ICs in ences by the Kaplan–Meier survival were no longer evident when IM are multifaceted , and may exhibit pro- and anti-tumor stratified by grade. These findings indicated that the DFS functions simultaneously based on the different levels of CD8 and predictive ability of IPs in the luminal B subtype may have a FoxP3 . Furthermore, B cells and plasma cells can also adopt close association with histological grade. While in the TNBC either effector or regulatory phenotypes, and hence, exhibit cohort, the chi-square test failed due to the existence of many positive or negative anti-tumor associations depending on the 19–21 variables smaller than 5, which confirmed the deduction that the contextual factors . In this study, statistical analysis showed skewed distribution of TNBC in histological grades could be a that T cells and CD8-positive cytotoxic T cells were significantly major factor contributing to its low predictive ability . more abundant in tumor areas of IP1 and IP2 than in Paras, further npj Breast Cancer (2022) 26 Published in partnership with the Breast Cancer Research Foundation X. Du et al. Fig. 7 Different immune cell subpopulations in tumor-associated areas and para-tumor areas for all five IPs identified in this study. The ratio of T lymphocytes (CD4− and CD8-positive) to B lymph cells (CD20-positive) a, B lymph cells (CD20-positive) to T lymph cells (CD4− and CD8-positive) b, and cytotoxic T cells (CD8-positive) to T lymph cells c among IP1/2 TCs and IMs and IP1/4 Paras. The ratio of T lymph cells to B lymph cells d, B lymph cells to T lymph cells e, and cytotoxic T cells (CD8-positive) f to T lymph cells among IP1/2 IMs, IP3 IMs, and IP1/4 Paras. PD-L1-positive immune cells (SP142, brown DAB staining) observed with all five patterns g–k. PD-L1 expression in immune cells among all five patterns for all patients m, all patients with TNBC or HER2+ tumors l, all patients with TNBC n, all HER2+ patients o, and all luminal HER2− patients p. Kaplan–Meier curves for disease-free survival according to the median expression level of PD-L1 in all patients q, all patients with TNBC r, all HER2+ patients s, all ER+/HER2− patients t, all patients with TNBC in the IP1 and IP2 groups u, and all patients with TNBC in the IP3, IP4, and IP5 TNBC groups v. IP Immunoarchitectural pattern, TC tumor center, IM invasive margin, Para para-tumor area, TNBC triple- negative breast cancer. suggesting that T cells (especially cytotoxic T cells) exhibit anti- ERBB2/4, MAP3K4, BRCA1, and PIK3CA. Further cases are needed to tumor functions in tumor areas. In contrast, high density of B cells verify whether the heterogeneous immunogenicity of IBC is in the Para might promote tumor progression, considering the attributable to driver gene mutation signatures, as suggested in para-tumor immune microenvironment mentioned above. prior studies . Genome-wide mRNA-expression analysis in ICs revealed a cluster IPs can be used in preliminary screening for PD-L1 expression. of immune-related genes that were differentially expressed in the PD-L1 expression in the five IPs showed very similar trends in the tumor area and para-tumor area, which deserves further explora- following three groups: all selected cases, the TNBC subtype, and tion in terms of the function and clinical significance. The median the HER2+ subtype. These highly repetitive expression trends CD20/(CD4+ CD8) and CD8/(CD4+ CD8) ratios in IP3 IMs were suggest that PD-L1 expression coincides well with the IP- between those of IP1/2 IMs and IP1/4 Paras, which might indicate classification scheme. Therefore, we speculate that PD-L1 expres- a compromised status between the host and tumor. sion was closely related to the density of TILs, a phenomenon 23–25 IP5 was characterized by very low or no IC infiltration in both reported by other researchers , rather than the tumor the TC and Para, with similar weak immune reactions or immune molecular subtypes. Through survival analysis, we found that ignorance immune type described by Camus and colleagues . PD-L1 expression was significantly associated with a better DFS for Patients in IP5 mainly presented a lower histological grade (88%) all 208 cases and in patients with TNBC. Surprisingly, the HER2+ and a luminal molecular type (86%). We speculate that the subtype cases, with a similar trend in PD-L1 positivity and different determining factor for different immune reactions within all five IPs in TNBC, did not show the same predictive value in terms of IPs was the tumor antigen load. However, patients in IP1 and IP2 PD-L1 expression. Considering that the PD-L1 expression level in did not show a higher TMB than IP5. We identified 320 oncodriver TNBC is significantly higher than that in the HER2+ subtype, we gene mutations in 40 patients. Among these, nearly half (44%) speculate that the PD-L1-expression level is a better predictor of exhibited a markedly lower mutation rate in IP5, including TP53, DFS than the PD-L1-positivity status (1% cut-off). Therefore, it is Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 26 X. Du et al. clinically more meaningful to provide exact PD-L1-positive IC Evaluating key TIL parameters and determining IPs percentages than to simply provide a positivity status . Moreover, TILs are defined as mononucleated lymphoid cells, which include lymphocytes and plasma cells. The overall assessment of stromal TILs owing to the close relationship and similar DFS predictive value (sTILs), rather than hotspots of intense infiltration, was conducted using between PD-L1 expression and IPs, TNBC cases that intersected H&E-stained sections to eliminate interior heterogeneity. To elucidate the with IP1/2 were mostly PD-L1 positive, whereas cases with spatial distribution of TILs, the tumor areas were divided into the TC and IM complete TIL deficiency in IP5 could be considered PD-L1 as per the guidelines of the International Immuno-Oncology Biomarker 19,32 negative. Therefore, predicting the PD-L1-expression status using Working Group (Fig. 1). We used a traditional software (StrataQuest IP classification based on H&E staining is feasible and can be 6.0.213, TissueGnostics, Vienna, Austria)-assisted assessment system to considered for preliminary screening of patients with IBC-NST. It is precisely evaluate the ratio of TILs or PILs to the stromal area in TCs, IMs, and Para. Lobular involvement defined as invasive tumor cells expand into recommended to report both TILs and PD-L1 as a combined 27 the lobules in Para, which was commonly seen in BC. immune-oncological biomarker in daily practice . The procedural steps are outlined in Fig. 1. Five consecutive According to the conception of tumor immunoediting, IP of IBC complementary steps of software-assisted manual assessment were represents the final immune manifestation of ductal carcinoma performed: 1. The boundaries of tumor area (red) and Para (yellow) were in situ (DCIS) progression. By comparing of pure DCIS, DCIS mixed manually delineated by pathologists. Then the confounder regions (gray) synchronous IBC, and IBC, the immune cell subsets and spatial that were to be excluded from these areas were annotated, including 28,29 necrosis, tertiary lymphoid structures, intermixed normal tissue; 2. The ratio distribution were found highly variable . Further studies need of stromal area to tumor area or Para was manually assessed by two to focus on certain key immunomodulatory switches, such as pathologists to eliminate bias from the variability of stromal to tumor area CD103 and CTLA4, in specific IP during the progression from ratio between different cases. 3. The IM area was automatically mapped by DCIS to IBC, to trace the formation of the IPs and select the more the software, and thus, IM and TC were segmented; 4. The numbers of TILs progressive DCIS for early immunotherapy. (red) and PILs (red) in TC, IM, and Para were automatically labeled and In summary, we established five morphological IPs by studying counted. 5. The ratio of TILs or PILs nucleus area to the stromal area in TC, the immune architecture of IBC-NSTs in terms of density, spatial IM, and Para was calculated. By calculating the ratio of nucleus to the whole lymphocyte area, a distribution, subpopulation of lymphocytes, TMB, and immune- threshold value of 20% ratio of the lymphatic nucleus area to the stromal associated gene expression profiles. The clinical significance of the area in the TC was used as a cut-off for the quantity scores of consecutive different IPs is as follows: (1) high-TIL counts in the TC represented sTILs, in order to identify LPBC (50% stromal TILs) samples. A cut-off ratio of the most robust indicator for LPBC, with a high proportion of CD8 10% lymphatic nucleus area to stromal area in the IM was used to + T cells, indicating favorable DFS in TNBC; (2) high PIL counts distinguish samples with intermediate and low immunoscores. We further included PILs, based on a cut-off ratio of 20% lymphatic nucleus area to with a high proportion of CD20+ B cells indicated poor DFS similar stromal area to investigate the influence of tumor immunity on the Para. to TILs mainly in the IM in luminal B subtype IBC; (3) PD-L1 Finally, five different types of IPs were determined, as shown in Fig. 2. positivity significantly correlated with the IC counts and could be predicted by IPs; and (4) IPs of IBC-NST might be a potential IHC prognosis factor, especially in the TNBC and luminal B subtypes. FFPE tissue specimens were sectioned at 4 μm thickness. Antibodies There are several limitations to our investigation. First, this study against CD4 (clone number UMAB64, ZSbio, Wuxi, China), CD8 (clone was a retrospective evaluation, so it is not clear whether the IP number SP16, ZSbio), CD20 (clone number L26, ZSbio), and PD-L1 (clone system may help predict the therapeutic responses. Second, the number SP142, Ventana Medical Systems, Oro Valley, AZ) were used for short clinical follow-up period for these patients further restricted IHC, which was performed using a Benchmark Ultra System, the OptiView the predictive value in terms of survival, especially for luminal DAB IHC Detection Kit (Ventana Medical Systems). A rabbit monoclonal tumors, where late recurrences are common. Third, future studies negative immunoglobulin (Ventana Medical Systems) was used as negative should explore the genetic variation underling tumor IPs to screen control, and FFPE tonsil tissue stained with PD-L1 was used as positive control. All negative and positive controls were used for each batch of for novel markers that can serve as potential immunotherapeutic samples. The PD-L1 staining scores of TILs (cut-off value: 1% staining of any targets in breast cancer. Despite these limitations, our study intensity) were evaluated by two board-certified pathologists according to provides unique insights into the tumor microenvironment 25,34 the IM passion 130 Trial criteria . The positivities of CD markers were architecture and its potential clinical prospects. assessed based on a digital-pathology computational workflow (Strata- 14,33 Quest 6.0.213 software for WSIs . The densities of CD4-, CD8-, and CD20-positive lymphocytes were evaluated through counting the numbers METHODS of nuclei encompassed by DAB-stained cytoplasm and membrane (Supplementary Fig. 1). A threshold was established to ensure the accurate Cohort enrollment labeling of positive cells for each region. The output parameters are The study cohort consisted of 579 consecutively archived IBC-NST samples presented as the number of positive lymphocytes/mm (for TCs, IMs, that were surgically excised at the time of diagnosis between 08/2015 and and Para). 08/2018. All enrolled patients provided written informed consent to use these samples for translational research, as approved by the Ethics DNA isolation and whole-exome sequencing Commission of the General Hospital of China PLA (approval number ky- We selected samples from 40 patients with different molecular subtypes 2020-1-4) and the study was compliant with the ‘Guidance of the Ministry and matched FFPE and blood samples to detect germline mutations. of Science and Technology (MOST) for the review and Approval of human Genomic DNA was extracted using a QIAamp DNA FFPE Tissue Kit or a Genetic Resources’, China. For TIL quantification, all specimens were DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany). Whole-exome reviewed for pathomorphology on H&E-stained sections to select the most sequencing libraries were constructed using an NEBNext Ultra DNA Library representative tumor block. Biomarker expression levels and other Prep Kit (Illumina, San Diego, CA) and according to the manufacturer’s clinicopathological features were extracted from the archives. Histological protocol. The average whole-exome sequencing depth was 100× for FFPE grades were scored according to the Nottingham semi-qualitative scoring tumors samples, and 60× for their normal tissue counterparts. system . Tumor molecular subtyping (expression of the associated parameters, including ER, PR, HER2, and Ki-67) and TNM staging were RNA isolation and Illumina immune-response gene panel evaluated based on World Health Organization’s classification of breast sequencing tumors (5th edition) . All H&E-stained and IHC-stained sections were scanned using a KFBIO Scans cope high-resolution scanner at 40× Macrodissections were performed on whole H&E-stained sections to magnification (Konfoong Biotech, Ningbo, China), and all parameters collect TILs from tumor areas and PILs from Para. Total RNA was extracted assessed in this study were examined using WSIs. using a miRNeasy FFPE Kit (Qiagen, Hilden, Germany) following the npj Breast Cancer (2022) 26 Published in partnership with the Breast Cancer Research Foundation X. Du et al. manufacturer’s instructions. RNA-sequencing libraries were constructed 11. Stamatelos, S. K., Bhargava, A., Kim, E., Popel, A. S. & Pathak, A. P. Tumor using the AmpliSeq Immune Response Panel (Illumina, San Diego, CA) ensemble-based modeling and visualization of emergent angiogenic hetero- according to the manufacturer’s protocol. Then, the quality of the library geneity in breast cancer. Sci. Rep. 9, 5276 (2019). (2 × 150 base-pair, paired-end reads) was checked, and it was sequenced 12. Camus, M. et al. Coordination of intratumoral immune reaction and human using a HiSeq 2500 System (Illumina, San Diego, CA). colorectal cancer recurrence. Cancer Res. 69, 2685–2693 (2009). 13. Angell, H. & Galon, J. From the immune contexture to the Immunoscore: the role of prognostic and predictive immune markers in cancer. Curr. Opin. Immunol. 25, Statistical analysis 261–267 (2013). Tumor sizes, metastatic lymph nodes, and the scores for CD4, CD8, CD20, 14. Mi, H. et al. Digital pathology analysis quantifies spatial heterogeneity of CD3, PD-L1, ER, PR, and Ki-67 staining were treated as continuous variables. They CD4, CD8, CD20, and FoxP3 immune markers in triple-negative breast cancer. were tested for normality and homogeneity of variance and were found Front Physiol. 11, 583333 (2020). not normally distributed and not homogenous. Non-parametric tests 15. Galon, J. & Bruni, D. Approaches to treat immune hot, altered and cold tumours (Mann-Whitney U test of Two-Independent Samples; Kruskal-Wallis H test with combination immunotherapies. Nat. Rev. Drug Disco. 18, 197–218 (2019). of K Independent samples) were conducted to assess the difference 16. Liang, Q., Ma, D., Gao, R. F. & Yu, K. D. Effect of Ki-67 expression levels and between groups. Some of these parameters were missing, and therefore, histological grade on breast cancer early relapse in patients with different omitted in subsequent analysis. We assessed the correlations between Immunohistochemical-based Subtypes. Sci. Rep. 10, 7648 (2020). patterns and categorical variables (clinical stage, histological grade, 17. Schreiber, R. D., Old, L. J. & Smyth, M. J. Cancer immunoediting: integrating HER2 status, molecular type, lobular involvement, and cancerous embolus) immunity’s roles in cancer suppression and promotion. Science 331, 1565–1570 by performing χ² tests for trends or Fisher’s exact tests. Survival endpoints (2011). were evaluated as DFS. Survival curves were prepared based on 18. Mlecnik, B. et al. Histopathologic-based prognostic factors of colorectal cancers Kaplan–Meier estimates and compared using the log-rank test. The Cox are associated with the state of the local immune reaction. J. Clin. Oncol. 29, proportional hazards model was used to perform univariable and 610–618 (2011). multivariable analyses to further identify the variables independently 19. Hendry, S. et al. Assessing tumor-infiltrating lymphocytes in solid tumors: a associated with DFS. Statistical analyses were performed using SPSS 23 practical review for pathologists and proposal for a standardized method from (IBM Corporation, Somers, NY). All statistical tests were 2-sided, and P the International Immunooncology Biomarkers Working Group: Part 1: assessing values < 0.05 were considered as statistically significant. the host immune response, TILs in invasive breast carcinoma and ductal carci- noma in situ, metastatic tumor deposits and areas for further research. Adv. Anat. Pathol. 24, 235–251 (2017). Reporting summary 20. Tan, T. T. & Coussens, L. M. Humoral immunity, inflammation and cancer. Opin. Further information on research design is available in the Nature Research Immunol. 19, 209–216 (2007). Reporting Summary linked to this article. 21. Shah, S. et al. Increased rejection of primary tumors in mice lacking B cells: inhibition of anti-tumor CTL and TH1 cytokine responses by B cells. Int J. Cancer 117, 574–586 (2005). DATA AVAILABILITY 22. He, Y., Jiang, Z., Chen, C. & Wang, X. Classification of triple-negative breast can- DNA-seq and RNA-seq datasets were uploaded on National center for biotechnology cers based on Immunogenomic profiling. J. Exp. Clin. Cancer Res. 37, 327 (2018). information (NCBI) under the Accession Number of PRJNA786713. The datasets 23. Van Berckelaer, C. et al. Infiltrating stromal immune cells in inflammatory breast generated and/or analyzed during the current study are available from the cancer are associated with an improved outcome and increased PD-L1 expres- corresponding author on reasonable request. sion. Breast Cancer Res 21, 28 (2019). 24. Buisseret, L. et al. Tumor-infiltrating lymphocyte composition, organization and PD-1/ PD-L1 expression are linked in breast cancer. Oncoimmunology 6, e1257452 Received: 17 May 2021; Accepted: 21 January 2022; (2017). 25. Hoda, R. S. et al. Clinical and pathologic features associated with PD-L1 (SP142) expression in stromal tumor-infiltrating immune cells of triple-negative breast carcinoma. Mod. Pathol. 33, 2221–2232 (2020). 26. Erber, R. & Hartmann, A. Understanding PD-L1 testing in breast cancer: a practical REFERENCES approach. Breast care 15, 481–490 (2020). 1. Curtis, C. et al. The genomic and transcriptomic architecture of 2,000 breast 27. Gonzalez-Ericsson, P. I. et al. The path to a better biomarker: application of a risk tumours reveals novel subgroups. Nature 486, 346–352 (2012). management framework for the implementation of PD-L1 and TILs as immuno- 2. Poudel, P., Nyamundanda, G., Patil, Y., Cheang, M. C. U. & Sadanandam, A. Het- oncology biomarkers in breast cancer clinical trials and daily practice. J. Pathol. erocellular gene signatures reveal luminal-A breast cancer heterogeneity and 250, 667–684 (2020). differential therapeutic responses. NPJ Breast Cancer 5, 21 (2019). 28. Gil Del Alcazar, C. R. et al. Immune escape in breast cancer during in situ to 3. Lehmann, B. D. et al. Identification of human triple-negative breast cancer sub- invasive carcinoma transition. Cancer Disco. 7, 1098–1115 (2017). types and preclinical models for selection of targeted therapies. J. Clin. Investig. 29. Toss, M. S. et al. The prognostic significance of immune microenvironment in 121, 2750–2767 (2011). breast ductal carcinoma in situ. Brit J. Cancer 122, 1496–1506 (2020). 4. Steenbrugge, J. et al. Anti-inflammatory signaling by mammary tumor cells 30. Savas, P. et al. Single-cell profiling of breast cancer T cells reveals a tissue-resident mediates prometastatic macrophage polarization in an innovative intraductal memory subset associated with improved prognosis. Nat. Med. 24, 986–993 mouse model for triple-negative breast cancer. J. Exp. Clin. Cancer Res. 37, 191 (2018). (2018). 31. Allison, K. H. WHO Classification of Tumours 5th Edition Breast Tumours. Lyon: 5. Denkert, C. et al. Tumour-infiltrating lymphocytes and prognosis in different IARC (2019). subtypes of breast cancer: a pooled analysis of 3771 patients treated with 32. Salgado, R. et al. The evaluation of tumor-infiltrating lymphocytes (TILs) in breast neoadjuvant therapy. Lancet Oncol. 19,40–50 (2018). cancer: recommendations by an International TILs Working Group 2014. Ann. 6. Yu, X. et al. Prognostic and predictive value of tumor-infiltrating lymphocytes in Oncol. 26, 259–271 (2015). breast cancer: a systematic review and meta-analysis. Clin. Transl. Oncol. 18, 33. Miller, T. J. et al. The prognostic and predictive value of SOX2(+) cell densities in 497–506 (2016). patients treated for colorectal cancer. Cancers 12, (2020). 7. Denkert, C. et al. Tumor-associated lymphocytes as an independent predictor of 34. Aktas, B. Y., Taban, H. & Aksoy, S. Atezolizumab and Nab-paclitaxel in advanced response to neoadjuvant chemotherapy in breast cancer. J. Clin. Oncol. 28, triple-negative breast cancer. N. Engl. J. Med. 380, 985–986 (2019). 105–113 (2010). 8. Marra, A., Viale, G. & Curigliano, G. Recent advances in triple negative breast cancer: the immunotherapy era. BMC Med. 17, 90 (2019). 9. Kim, J. Y., Kim, W. G., Kwon, C. H. & Park, D. Y. Differences in immune contextures among different molecular subtypes of gastric cancer and their prognostic ACKNOWLEDGEMENTS impact. Gastric Cancer 22, 1164–1175 (2019). We thank Xiaochang Zhang (Beijing Institute of Radiation Medicine) for his help in 10. Mani, N. L. et al. Quantitative assessment of the spatial heterogeneity of tumor- generating images with R software for this research. This work was supported by the infiltrating lymphocytes in breast cancer. Breast Cancer Res. 18, 78 (2016). National Natural Science Foundation of China (grant number 81972485). Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 26 X. Du et al. AUTHOR CONTRIBUTIONS Reprints and permission information is available at http://www.nature.com/ reprints Xue Du, Zhe Zhou, and Yun Shao carried out the studies, statistical analysis, and drafted the manuscript. These three authors contributed equally to this work and are Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims considered as “co-first author”. Kun Qian participated in sample collection and in published maps and institutional affiliations. patient follow-up. Yingfang Wu, Jun Zhang, Miao Cui, Jingjing Wang, and Yun Shao reviewed the H&E-stained sections and performed the TissueGnostics image analysis. Xue Du and Zhe Zhou were involved in next-generation sequencing (NGS) and NGS data analysis. Yanhong Tai and Shengqi Wang designed the studies and critically Open Access This article is licensed under a Creative Commons revised the manuscript for important intellectual content. Tai and Wang are Attribution 4.0 International License, which permits use, sharing, considered as corresponding authors. All authors read and approved the final adaptation, distribution and reproduction in any medium or format, as long as you give manuscript. appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless COMPETING INTERESTS indicated otherwise in a credit line to the material. If material is not included in the The authors declare no competing interests. article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. ADDITIONAL INFORMATION org/licenses/by/4.0/. Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41523-022-00389-y. © The Author(s) 2022 Correspondence and requests for materials should be addressed to Shengqi Wang or Yanhong Tai. npj Breast Cancer (2022) 26 Published in partnership with the Breast Cancer Research Foundation http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png npj Breast Cancer Springer Journals

Immunoarchitectural patterns as potential prognostic factors for invasive ductal breast cancer

Loading next page...
 
/lp/springer-journals/immunoarchitectural-patterns-as-potential-prognostic-factors-for-HAl5DOmAK0
Publisher
Springer Journals
Copyright
Copyright © The Author(s) 2022
eISSN
2374-4677
DOI
10.1038/s41523-022-00389-y
Publisher site
See Article on Publisher Site

Abstract

www.nature.com/npjbcancer ARTICLE OPEN Immunoarchitectural patterns as potential prognostic factors for invasive ductal breast cancer 1,2,3 1,3 2,3 1 2 2 2 2 1✉ Xue Du , Zhe Zhou , Yun Shao , Kun Qian , Yongfang Wu , Jun Zhang , Miao Cui , Jingjing Wang , Shengqi Wang and Yanhong Tai Currently, tumor-infiltrating lymphocytes (TILs) in invasive breast cancers are assessed solely on the basis of their number, whereas their spatial distribution is rarely investigated. Therefore, we evaluated TILs in 579 patients with invasive breast cancer of no special type (IBC-NST) with a focus on their spatial distributions in tumor center (TC) and invasive margin (IM). We also assessed a new factor, namely para-tumor infiltrating lymphocytes (PILs) in the para-tumor lobular area (Para). Five immunoarchitectural patterns (IPs) were observed, which were significantly associated with clinicopathological features, especially molecular subtypes, histological grades, clinical stages, and programmed death-ligand 1 (PD-L1) expression. High-TIL density (IP1/2) correlated with favorable disease-free survival (DFS) in TNBC patients (p = 0.04), but opposite results were observed for luminal B subtype patients (both the lowest TIL and PIL densities (IP5) correlated with good DFS, p = 0.013). Luminal B patients with high TILs in the IM and low TILs in the TC (IP3) exhibited the worst DFS, whereas those with low TILs (similar to IP5) and high PILs (IP4) exhibited poor DFS. We also identified TIL subpopulations with significantly different IPs. Our findings suggest that IP can be a potential prognostic factor for tumor immunity in IBC. npj Breast Cancer (2022) 8:26 ; https://doi.org/10.1038/s41523-022-00389-y INTRODUCTION In this study, we investigated hematoxylin and eosin (H&E)- stained sections of 579 tissue samples from invasive breast cancer Breast cancers are clinically and molecularly heterogenous, with of no special type (IBC-NST) patients to define tumor immu- 5–10 intrinsic subtypes . Each subtype displays varied molecular 2,3 noarchitectural patterns (IPs) and TIL density. Therefore, compre- characteristics that form the basis for therapeutic resistance and hensive analysis of the identified IPs was performed with respect different therapeutic strategies . Immunotherapy and combined to lymphocyte density, location, immunophenotyping, and neoadjuvant chemotherapy are being aggressively developed, combined histopathological characteristics (such as the histologi- with anti- programmed death-ligand 1 (PD-L1) exhibiting strong cal grade, clinical stage, molecular type, and survival status) in immunomodulatory therapeutic potential against breast cancer . patients with IBC-NST. A pre-existing immunological response might enhance the 5–7 efficacy of conventional cytotoxic chemotherapy . However, despite accumulating evidence, the translation from basic tumor immunology to clinical practice remains problematic . PD-L1 and RESULTS tumor mutational burden (TMB)-based immunotherapeutic clinical Stratification of IBC-NSTs into five IPs trials have shown favorable results in a small subset of invasive We assessed 579 primary IBC-NST cases for tumor immunity and breast cancer (IBC) patients, mainly triple-negative breast cancer grouped them into five IPs, as indicated in the flowchart (Figs. 1 (TNBC) patients . Previous studies have shown that high count of and 2). IP2 (19/579, 3.28%) had the least number of cases, followed tumor-infiltrating lymphocytes (TILs) cannot constantly warrant a by IP1 (69/579, 11.92%), IP3 (110/579, 19.00%), IP4 (130/579, good outcome in all IBC patients. In luminal-HER2-negative patients, high TIL count is considered an adverse prognostic 22.45%), and IP5 (251/579, 43.35%) (Table 1). We displayed the factor for survival ; however, the TILs should be studied with a cross-referenced H&E and leukocyte common antigen (LCA) new perspective for a comprehensive understanding of the tumor stained sections, and TissueGnostics images of typical cases to microenvironment. highlight the distinct differences of five IPs in Fig. 3. A recent investigation into more reliable predictors revealed that immune contextures, such as TIL density and spatial Lobular involvement, cancerous embolus, and histological localization, are associated with clinicopathological characters grade and PD-L1 expression based on molecular subtypes, and were IP4 had significantly higher frequency of lobular involvement therefore considered appropriate immunotherapeutic candidates. (121/130, 93.08%) compared with that in the other four IPs. A However, the association of clinicopathological characters with para-tumor infiltrating lymphocytes (PILs) located in the para- similar trend was also observed with IP1 (47/69, 68.12%), which tumor lobular area (Para) remains uncertain. Therefore, the had high PIL counts similar to IP4, but unlike IP2 (5/19, 26.32%), quantitative molecular and spatio-morphological parameters of IP3 (46/110, 41.82%), and IP5 (98/251, 39.04%) (Fig. 4a, b, infiltrating lymphocytes interactions should be explored to Supplementary Table 1). Moreover, the cancerous embolus ratio was also significantly higher in IP4 (42/130, 32.31%) than that in improve the identification of predictive markers. 1 2 3 Beijing Institute of Radiation Medicine, Beijing, PR China. Department of Pathology, Fifth Medical Center of Chinese PLA General Hospital, Beijing, PR China. These authors contributed equally: Xue Du, Zhe Zhou, Yun Shao. email: sqwang@bmi.ac.cn; taiyanhong29@163.com Published in partnership with the Breast Cancer Research Foundation 1234567890():,; X. Du et al. Fig. 1 Outline of the traditional software-assisted assessment procedure used for analyzing TILs and PILs in IBC-NST samples. TILs tumor- infiltrating lymphocytes, PILs para-tumor infiltrating lymphocytes, IBC-NST invasive breast cancer of no special type. IP1 (8/69, 11.59%), IP3 (23/110, 20.91%), and IP5 (47/251, 18.73%) grade 1 cases were found in the IP1 and IP2 groups, and grade 3 (Fig. 4c, d, Supplementary Figure 2, Supplementary Table 1). cases (62.32% in IP1 and 52.63% in IP2) were more common than The histological-grade distribution differed significantly grade 2 cases (37.68% in IP1 and 47.37% in IP2). In contrast, in IP5, between each of the five IPs (χ² = 84.84, p < 0.001; Fig. 4e, f). No grade 1 (14.34%) and grade 2 (73.71%) cases accounted for npj Breast Cancer (2022) 26 Published in partnership with the Breast Cancer Research Foundation 1234567890():,; X. Du et al. one case each of luminal A subtype was found in IP1 and IP2. Conversely, patients with the TNBC or HER2 + subtype accounted for a greater proportion of IP1 (19/69, 27.54% and 14/69, 20.29%, respectively) compared to that of IP2 (each 3/19, 15.79%), IP3 (19/ 110, 17.27% and 10/110, 9.09%, respectively), IP4 (19/130, 14.62% and 10/130, 7.69%, respectively), and particularly IP5 (6/251, 2.39% and 10/251, 3.98%, respectively). Luminal B subtype and luminal- HER2 subtype cases were distributed evenly into the five IPs (Supplementary Table 1). Pearson’s chi-square test of molecular subtypes revealed significant differences between IPs (χ² = 88.097, p < 0.001; Fig. 5f, g), and IP5 contained significantly more mild subtype cases than other IPs, whereas aggressive subtype cases were significantly more common in IP1 than IP3–IP5. The IP2 group showed no significant difference in the molecular-subtype distribution with other IPs, except for IP5 (Supplementary Fig. 2). The distribution of IPs in the luminal B subtype differs significantly in three histological grades (χ = 18.494, p = 0.018). Pattern integration The clinicopathological characteristics of the 579 patients were summarized by heatmapping (Fig. 5h). Although the IP2 cases shared similar prognosis-related clinicopathological characteristics with the IP1 cases (such as a higher histological grade, more aggressive molecular subtype, and lower clinical stage), the differences between IP2 and IP3–5 were less significant than those between IP1 and IP3–5 (Supplementary Figure 2). As there Fig. 2 Flow chart of the process used to identify five different immunoarchitectural patterns. IP immunoarchitectural pattern. were only 19 IP2 cases (and even less when subdivided), the low degree of significance may partially reflect the small sample size. 88.05% of all cases, whereas grade 3 cases accounted for only We found that the curve for IP5 was unique, whereas the curve 11.95% of the cases (Supplementary Table 1). The grade tendencies of IP1 and IP2, as well as IP3 and IP4, were highly distribution was not different between IP3 and IP4, both of which similar (Supplementary Figure 3). These three characteristics are exhibited more grade 3 cases and fewer grade 1 cases compared the major factors that affect disease-free survival (DFS); thus, we to IP5. No significant differences were identified between the IP1 merged the IPs into three groups (IP1/2, IP3/4, and IP5) as and IP2, and IP2 and IP3 pairs (Supplementary Fig. 2). prognostic groups for subsequent DFS analysis (Fig. 6). Clinical stage, tumor size, and lymph node metastasis DFS analysis Tumor-size measurements were missing in one case each for IP4 In this study, Formalin-fixed, paraffin-embedded (FFPE) tissues and IP5; thus, these cases were omitted from subsequent analyses. were collected within 3 years to meet the quality requirements for Most patients (496) exhibited early stage (stage I, 269 cases or nucleic acid-isolation for genome-wide mRNA-expression analysis stage II, 227 cases) disease, and only 81 patients exhibited late- and exome sequencing. Given the short duration of the disease, stage disease (stage III, 74 cases or stage IV, 7 cases). A significant we only assessed the IPs as a prognostic factor for DFS, with a difference was found between patients exhibiting early- and late- median survival time of 91.63 months (IQR 69.40–113.86). No stage disease among the 5 IPs (χ² = 12.055, p = 0.017). The significant association was found with DFS among the five IPs. proportion of late-stage cases in IP1 (2.90%) was significantly When stratified by molecular subtypes, the three merged-IP lower than that in IP3 (19.09%), IP4 (12.40%), and IP5 (16.40%) (Fig. groups exhibited significant associations with DFS (log-rank = 4g, h, Supplementary Table 1). Only one late-stage case was found 3.054, p = 0.217; Breslow = 8.724, p = 0.013; Fig. 6a–e). In the in IP2 (1/19, 5.26%), although it was not significantly different from luminal B subtype, all three merged IPs (log-rank = 8.711, p = other IPs (Supplementary Fig. 2). 0.013; Fig. 6b) and all five individual IPs (log-rank = 10.121, p = Tumor sizes differed significantly between the 5 IPs (χ = 0.038; Fig. 6f) showed significant associations with DFS, where IP5 22.035, p < 0.001), with the lowest median tumor size in IP4 and was associated with a favorable outcome and IP3 was associated the highest in IP2 (Fig. 4i, Supplementary Fig. 2, Supplementary with the worst DFS. Patients in the IP4 group (distinguished from Table 1). The differences were analyzed in terms of the number of lymph node metastases, but no statistical significance was found those in the IP5 group by high PIL count) had the same poor DFS (χ = 3.994, p = 0.407. Fig. 4j, Supplementary Fig. 2). rate as those in the IP3 group. In contrast, among patients with the TNBC subtype, those in the merged IP1/2 group had significantly improved DFS than patients in the IP5 group (log-rank = 6.419, Molecular subtype p = 0.040; Fig. 6e). Similar trends were observed for patients with Immunohistochemistry (IHC)-staining of samples from 579 the luminal B-HER2 and HER2+ subtypes, but the association with patients revealed that the ER- and PR-expression levels gradually DFS was not significant for either subtype (Fig. 6c, d). Among increased from IP1 to IP5 (Fig. 5a, b, Supplementary Fig. 2), patients with the luminal A subtype, only two were included in the whereas the proliferation indicator Ki-67 revealed an opposite IP1 and IP2 groups, and the DFS curves of the merged-IP groups trend (Fig. 5c). HER2 + cases were significantly lower in IP5 than in overlapped without a significant association in terms of DFS (Fig. the other IPs (Fig. 5d, e). 6a). Both univariable and multivariable analyses showed that the Among the 136 luminal A subtype samples, most were distributed in IP5 (88/136, 64.71%), followed by IP4 (27/136, IP had no significant associations with DFS (Supplementary 19.85%), and IP3 (19/136, 13.97%) (Supplementary Table 1). Only Table 2). Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 26 X. Du et al. Table 1. Population clinicopathological characters. Characteristics No. Percentage Characteristics No. Percentage Age ≤35 32 5.53% TNM I 269 46.46% 35–50 218 37.65% II 227 39.21% ≥50 329 56.82% III 74 12.78% IV 7 1.21% Sex Female 578 99.83% Unknown 2 0.35% Male 1 0.17% ER status Negative 118 20.38% Postmenopausal not yet 312 53.89% Positive 461 79.62% Postmenopausal 260 44.91% Unknown 6 1.04% PR status Negative 170 29.36% Null (male) 1 0.17% Positive 409 70.64% Family genetic history not 560 96.72% HER2 status Negative 474 81.87% Family genetic history 15 2.59% Positive 105 18.13% Unknown 4 0.69% Ki67 ≤14% 151 26.08% Tumor size T1( ≤ 2) 379 65.46% å 14% 428 73.92% T2(2–5) 155 26.77% T3( ≥ 5) 10 1.73% Molecular type Luminal A 136 22.70% T4 33 5.70% Luminal B 272 45.41% Unknown 2 0.35% Luminal HER2 58 9.68% HER2+ 47 7.85% Lymho node N0 377 65.11% TNBC 66 11.02% N1 143 24.70% N2 38 6.56% Immuno- architectural Pattern 1 69 11.92% N3 21 3.63% 2 19 3.28% 3 110 19.00% Metastasis M0 572 98.79% 4 130 22.45% M1 7 1.21% 5 251 43.35% Spatial differences in the immune cell subpopulations subtype on PD-L1 expression in lymphocyte predominant breast cancer (LPBC) patients. PD-L1 was mainly expressed in immune We selected 77 typical cases of different IPs (IP1 = 28, IP2 = 19, cells (ICs) (103/208, 49.52%), including lymphocytes, macrophages, IP3 = 17, and IP4 = 13) to compare the subpopulations of TILs dendritic cells, plasma cells, and granulocytes, but was rarely located in stromal tumor center (TC) invasive margin (IM), and PILs in Para. The TIL subpopulations in the TCs and IMs of IP1/2 group expressed in tumor cells (13/208, 6.25%). Positive ICs were characterized by dark brown punctate or linear membrane were averaged to obtain total TIL-subpopulation densities to staining. The spatial distributions of positive ICs resembled those represent “hot tumors” group. Likewise, we combined each PIL subpopulation (CD4, CD8, and CD20) of IP1/4 group to represent of the IPs (Fig. 7g–k). “hot Para” group. We compared the ratio of T cells (CD4+ CD8+) PD-L1-expression levels were significantly higher in the TNBC to B cells (CD20+), and the ratio of cytotoxic T cells (CD8+)to subtype than in the HER2+ subtype (p = 0.0031, n = 170; Fig. 7l). T cells in these groups. We found that the proportion of T cells PD-L1 positivity was significantly associated with IPs (p < 0.001, (z = −3.455, p = 0.001) and cytotoxic T cells (z = −2.448, p = n = 208; Fig. 7m), where ICs in both the IP1 (84.06%) and IP2 0.014) in “hot tumor” group was significantly higher than that in (89.47%) groups showed the highest positive rates, followed by “hot Para” group. In contrast, the proportion of B cells was IP3 (31.71%) and IP4 (31.82%), while ICs in the IP5 group had the significantly higher in “hot Para” group than that in “hot tumor” lowest expression rate (2.86%) (Supplementary Table 3). The same group (z = −3.112, p = 0.002; Fig. 7a–c). tendency was also observed in the TNBC (p < 0.001, n = 66) and We then compared the ratio of T cells to B cells, and the ratio of HER2+ subtypes (p < 0.001, n = 104; Fig. 7n, o). In cohort 2, PD-L1 cytotoxic T cells to T cells for the following two pairs: IM of IP3 and expression in ER+/HER2− patients was high in both IP1 and IP2 IP1/2; IM of IP3 and Para of IP1/4. No significant differences were groups, with no significant difference (p = 0.31, n = 38; Fig. 7p). found (Fig. 7d–f). Survival analysis revealed that in all 208 cases, including 66 TNBC cases, a PD-L1-positive status was significantly associated The PD-L1-positive rate differed significantly between IPs and with a better prognosis (p = 0.018 and p = 0.020, respectively; Fig. correlated with DFS in all cases and TNBC subtypes 7q, r). No predictive significance was found for survival in patients with the HER2+ and ER+/HER2− subtypes (Fig. 7s, t). To eliminate From the 579 IBC-NSTs cases, 208 typical cases were selected for the influence of TIL count on DFS in TNBC, we subdivided the PD-L1 staining and were divided into two cohorts. Cohort 1 contained all enrolled patients with TNBC and HER2+ cases (with TNBC cases into those with high-TIL count (P1/2) and low-TIL all IPs). Cohort 2 contained all ER+/HER2− patients with IP1 and count (P3/4/5). No significant predictive value was found for either IP2 was used as a control to eliminate the effect of molecular group, based on PD-L1 expression (Fig. 7u, v). npj Breast Cancer (2022) 26 Published in partnership with the Breast Cancer Research Foundation X. Du et al. Fig. 3 Schematic diagram of immune cells’ distribution and morphology in five IP patterns of breast cancer tissues. H&E-stained sections, IHC sections, and TissueGnostics analysis of IP1 (a–c), IP2 (d–f), IP3 (g–i), IP4 (j–l), and IP5 (m–o). The IHC sections were stained for the LCA marker. The red lines in the IHC sections show the margins of the tumor areas. TissueGnostics analysis of the yellow frames in panels b, e, h, k, and n is shown in c, f, i, l, and o, respectively. The black and red lines shown for the TissueGnostics analysis represent the margins of the tumor areas and para-tumors, respectively. H&E hematoxylin and eosin, IHC immunohistochemistry, IP immune pattern, LCA leukocyte common antigen. Tumor-mutation signature and immune-response gene mutation rate in the IP5 group than in the other IP groups, expression including breast-cancer-specific oncodriver genes (TP53, ERBB2, We selected 40 cases from the five IPs to detect TMBs based on MAP3K4, BRCA1, ERBB4, and PIK3CA). By comparing the expression of immune-response genes whole-exome sequencing. No significant differences were between TCs and IMs in the IP1/2 (n = 6) and IP1/4 Para (n = 8) observed (Supplementary Figure 4). We then compared the groups, we identified a cluster of highly expressed genes in IP1/2 differences between oncodriver genes using the Catalog of Somatic Mutations in Cancer database (http://cancer.sanger.ac. TCs and IMs (Supplementary Fig. 4), including T cell activation- uk/census) and found that 44% (141/320) of the genes had a lower associated genes (CD8A, CXCR6, IL-12A, IL-8, and EBI3), the immune Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 26 X. Du et al. Fig. 4 Major clinical characteristics analysis in five immunoarchitectural patterns. Lobular involvement, cancerous embolus, tumor histological grade, clinical stage, tumor size, and number of metastatic lymph nodes in five immunoarchitectural patterns, in terms of distributions (a, c, e, g, i, j) and percentages (b, d, f, h). The correlations between patterns and categorical variables (including tumor grades, molecular subtypes, lobule cancerization, and vascular invasion) were analyzed by the χ² test (for trends) or Fisher’s exact test. Continuous variables (tumor size and number of metastatic lymph nodes) were assessed by the Kruskal–Wallis H tests under K independent-sample tests. npj Breast Cancer (2022) 26 Published in partnership with the Breast Cancer Research Foundation X. Du et al. Fig. 5 ER, PR, Ki-67 expression and HER2 status in five immunoarchitectural patterns and heatmap of all enrolled patients’ clinical characteristics. ER (a), PR (b), Ki-67 (c) expression and HER2 (d) status in five immunoarchitectural patterns The percentages of HER2-positive cases for each pattern (e). Distributions (f) and percentages (g) of molecular subtypes in each of the five IPs. h Heatmap of the clinical characteristics of all 579 patients included in this study. Molecular type, 1, luminal A subtype; 2, luminal B subtype; 3 represent luminal B HER2 subtype; 4 represent HER2 subtype; and 5 represent TNBC subtype. checkpoint gene ICOS (also known as CTLA4), and MHC II true. No cases with high-TC TIL density and low IM density were molecules, such as HLA-DRB1. Equivalently, high expression of found. Tumors with high-TIL densities in both areas fulfill the LPBC genes related to B-cell activation—including CD40LG and CD79B— criteria; thus, TCs provide sensitive and corroborative evidence to MHC molecules, CD1D, and B/T/Th1 cell activation genes— CCL21, confirm the LPBC status, thereby distinguishing IP1/2 from other IPs. In IP1 and IP2, the TIL density and subpopulation in TC and IM IL12B, and PTPRC—was observed in IP1/4 Para (Supplementary showed high accordance with each other. This homogeneity Fig. 4). allowed us to propose a hypothesis that TILs in TCs disperse from IMs, which are regions characterized by higher vessel density . DISCUSSION We then demonstrated that patients with IP1/2 had the best DFS In this study, we used a traditional software-assisted assessment in TNBC and better DFS than that for IP3 (high TIL density in IMs system for assessing IPs of IBC-NSTs on whole slide images (WSIs). only) in luminal B-type cases, further suggesting that high-TIL and This rigorous assessment system enabled us to establish the CD8+ T-cell densities reflect anti-tumor immunity and are 12,13 classification criteria for IPs by calculating the areas of the indicative of a good prognosis . IP1 and IP2 are demarcated lymphatic nuclei instead of those of whole cells and accurately based on higher PIL levels in IP1 compared with those in IP2. distinguished IM and TC regions. The finding that the density of TILs was higher in IMs than in At the morphological level, the densities and spatial distribu- TCs was also reported by Mi et al. , suggesting the distinct role of tions of TILs are heterogenous in different areas (TC, IM, and Para) IMs in the tumor immune architecture. They further discovered of the same block, but FFPE blocks of the same tumor are that the IM is multifaceted and may serve pro- and anti-tumor homogenous, as reported by Mani . In our study, specifically, TIL functions simultaneously with higher CD8+ expression and more count in TCs was similar to those in IMs; high-TC TIL count FOXP3+ cells. We deduced that those cases with higher CD8+ TILs indicated high-IM TIL count, but the reverse was not necessarily in IMs might show the same pattern as IP1/2 group, whereas the Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 26 X. Du et al. Fig. 6 Kaplan–Meier curves for disease-free survival for all patients according to the molecular types. a–e Log-rank (Mantel–Cox) χ² test = 3.054, p = 0.217; Breslow (generalized Wilcoxon) χ² square test = 8.724, p = 0.013; Tarone–Ware χ² test = 8.724 P = 0.049. Kaplan–Meier curves for luminal B subtype in all five patterns f. rest cases belonged to IP3. In IP3, TILs were mainly restricted to Most studies have focused on the immune microenvironment in the IM region and maintained at a medium level (>10%), and their the tumor area. In this study, we expanded the scope of subcellular populations were also altered, with a trend toward lymphocyte evaluation to the Para. As expected, corresponding reduced CD8/(CD4+ 8) ratios and increased CD20/(CD4+ 8) ratios, to “hot” immune tumors, we found another equally obvious “hot” though the alterations were not significant. A similar pattern of immune region localized in the IP1 and IP4 Para with high “altered immune tumors” was reported for colorectal cancer , frequency of lobular involvement. This distinction could be used based on the expression levels and spatial distributions of CD3 to stratify LPBC into IP1 and IP2, and “cold” immune tumors into and CD8. The density and subpopulation heterogeneity of TILs in IP4 and IP5. The fact that no significant difference in the frequency IP1/2 and IP3 indicate their multifaceted functions, which was of lobular involvement occurred between IP1 and IP4 indicates supported by their differential DFS in luminal B-type and that the immune status within a tumor area is not responsible for TNBC cases. driving lobular involvement. Histological grade influence on IP predictive ability in the The subpopulation of IC and gene mutations in different IPs. luminal B subtype. In this study, all five individual IPs of luminal B Previously, it was discovered that infiltrating ICs not only function and merged IPs of TNBC cases showed significant associations to control tumor growth and progression, but also help to create with DFS. Since previous reports confirmed that histological grade an immunosuppressive environment in which the tumor can is a significant independent factor for DFS in luminal but not TNBC thrive . IC subpopulations in the tumor area, combined with patients , it is reasonable to verify whether the predictive ability density and location, could predict the survival of patients with of IPs is attributed to the histological grade. Using the chi-square colorectal cancer more accurately than the classical test, we found that the IP distribution in the luminal B subtype tumor–node–metastasis (TNM) system , where a high CD8/CD3- differs significantly in three histological grades. The DFS differ- density ratio correlated with a good prognosis . However, ICs in ences by the Kaplan–Meier survival were no longer evident when IM are multifaceted , and may exhibit pro- and anti-tumor stratified by grade. These findings indicated that the DFS functions simultaneously based on the different levels of CD8 and predictive ability of IPs in the luminal B subtype may have a FoxP3 . Furthermore, B cells and plasma cells can also adopt close association with histological grade. While in the TNBC either effector or regulatory phenotypes, and hence, exhibit cohort, the chi-square test failed due to the existence of many positive or negative anti-tumor associations depending on the 19–21 variables smaller than 5, which confirmed the deduction that the contextual factors . In this study, statistical analysis showed skewed distribution of TNBC in histological grades could be a that T cells and CD8-positive cytotoxic T cells were significantly major factor contributing to its low predictive ability . more abundant in tumor areas of IP1 and IP2 than in Paras, further npj Breast Cancer (2022) 26 Published in partnership with the Breast Cancer Research Foundation X. Du et al. Fig. 7 Different immune cell subpopulations in tumor-associated areas and para-tumor areas for all five IPs identified in this study. The ratio of T lymphocytes (CD4− and CD8-positive) to B lymph cells (CD20-positive) a, B lymph cells (CD20-positive) to T lymph cells (CD4− and CD8-positive) b, and cytotoxic T cells (CD8-positive) to T lymph cells c among IP1/2 TCs and IMs and IP1/4 Paras. The ratio of T lymph cells to B lymph cells d, B lymph cells to T lymph cells e, and cytotoxic T cells (CD8-positive) f to T lymph cells among IP1/2 IMs, IP3 IMs, and IP1/4 Paras. PD-L1-positive immune cells (SP142, brown DAB staining) observed with all five patterns g–k. PD-L1 expression in immune cells among all five patterns for all patients m, all patients with TNBC or HER2+ tumors l, all patients with TNBC n, all HER2+ patients o, and all luminal HER2− patients p. Kaplan–Meier curves for disease-free survival according to the median expression level of PD-L1 in all patients q, all patients with TNBC r, all HER2+ patients s, all ER+/HER2− patients t, all patients with TNBC in the IP1 and IP2 groups u, and all patients with TNBC in the IP3, IP4, and IP5 TNBC groups v. IP Immunoarchitectural pattern, TC tumor center, IM invasive margin, Para para-tumor area, TNBC triple- negative breast cancer. suggesting that T cells (especially cytotoxic T cells) exhibit anti- ERBB2/4, MAP3K4, BRCA1, and PIK3CA. Further cases are needed to tumor functions in tumor areas. In contrast, high density of B cells verify whether the heterogeneous immunogenicity of IBC is in the Para might promote tumor progression, considering the attributable to driver gene mutation signatures, as suggested in para-tumor immune microenvironment mentioned above. prior studies . Genome-wide mRNA-expression analysis in ICs revealed a cluster IPs can be used in preliminary screening for PD-L1 expression. of immune-related genes that were differentially expressed in the PD-L1 expression in the five IPs showed very similar trends in the tumor area and para-tumor area, which deserves further explora- following three groups: all selected cases, the TNBC subtype, and tion in terms of the function and clinical significance. The median the HER2+ subtype. These highly repetitive expression trends CD20/(CD4+ CD8) and CD8/(CD4+ CD8) ratios in IP3 IMs were suggest that PD-L1 expression coincides well with the IP- between those of IP1/2 IMs and IP1/4 Paras, which might indicate classification scheme. Therefore, we speculate that PD-L1 expres- a compromised status between the host and tumor. sion was closely related to the density of TILs, a phenomenon 23–25 IP5 was characterized by very low or no IC infiltration in both reported by other researchers , rather than the tumor the TC and Para, with similar weak immune reactions or immune molecular subtypes. Through survival analysis, we found that ignorance immune type described by Camus and colleagues . PD-L1 expression was significantly associated with a better DFS for Patients in IP5 mainly presented a lower histological grade (88%) all 208 cases and in patients with TNBC. Surprisingly, the HER2+ and a luminal molecular type (86%). We speculate that the subtype cases, with a similar trend in PD-L1 positivity and different determining factor for different immune reactions within all five IPs in TNBC, did not show the same predictive value in terms of IPs was the tumor antigen load. However, patients in IP1 and IP2 PD-L1 expression. Considering that the PD-L1 expression level in did not show a higher TMB than IP5. We identified 320 oncodriver TNBC is significantly higher than that in the HER2+ subtype, we gene mutations in 40 patients. Among these, nearly half (44%) speculate that the PD-L1-expression level is a better predictor of exhibited a markedly lower mutation rate in IP5, including TP53, DFS than the PD-L1-positivity status (1% cut-off). Therefore, it is Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 26 X. Du et al. clinically more meaningful to provide exact PD-L1-positive IC Evaluating key TIL parameters and determining IPs percentages than to simply provide a positivity status . Moreover, TILs are defined as mononucleated lymphoid cells, which include lymphocytes and plasma cells. The overall assessment of stromal TILs owing to the close relationship and similar DFS predictive value (sTILs), rather than hotspots of intense infiltration, was conducted using between PD-L1 expression and IPs, TNBC cases that intersected H&E-stained sections to eliminate interior heterogeneity. To elucidate the with IP1/2 were mostly PD-L1 positive, whereas cases with spatial distribution of TILs, the tumor areas were divided into the TC and IM complete TIL deficiency in IP5 could be considered PD-L1 as per the guidelines of the International Immuno-Oncology Biomarker 19,32 negative. Therefore, predicting the PD-L1-expression status using Working Group (Fig. 1). We used a traditional software (StrataQuest IP classification based on H&E staining is feasible and can be 6.0.213, TissueGnostics, Vienna, Austria)-assisted assessment system to considered for preliminary screening of patients with IBC-NST. It is precisely evaluate the ratio of TILs or PILs to the stromal area in TCs, IMs, and Para. Lobular involvement defined as invasive tumor cells expand into recommended to report both TILs and PD-L1 as a combined 27 the lobules in Para, which was commonly seen in BC. immune-oncological biomarker in daily practice . The procedural steps are outlined in Fig. 1. Five consecutive According to the conception of tumor immunoediting, IP of IBC complementary steps of software-assisted manual assessment were represents the final immune manifestation of ductal carcinoma performed: 1. The boundaries of tumor area (red) and Para (yellow) were in situ (DCIS) progression. By comparing of pure DCIS, DCIS mixed manually delineated by pathologists. Then the confounder regions (gray) synchronous IBC, and IBC, the immune cell subsets and spatial that were to be excluded from these areas were annotated, including 28,29 necrosis, tertiary lymphoid structures, intermixed normal tissue; 2. The ratio distribution were found highly variable . Further studies need of stromal area to tumor area or Para was manually assessed by two to focus on certain key immunomodulatory switches, such as pathologists to eliminate bias from the variability of stromal to tumor area CD103 and CTLA4, in specific IP during the progression from ratio between different cases. 3. The IM area was automatically mapped by DCIS to IBC, to trace the formation of the IPs and select the more the software, and thus, IM and TC were segmented; 4. The numbers of TILs progressive DCIS for early immunotherapy. (red) and PILs (red) in TC, IM, and Para were automatically labeled and In summary, we established five morphological IPs by studying counted. 5. The ratio of TILs or PILs nucleus area to the stromal area in TC, the immune architecture of IBC-NSTs in terms of density, spatial IM, and Para was calculated. By calculating the ratio of nucleus to the whole lymphocyte area, a distribution, subpopulation of lymphocytes, TMB, and immune- threshold value of 20% ratio of the lymphatic nucleus area to the stromal associated gene expression profiles. The clinical significance of the area in the TC was used as a cut-off for the quantity scores of consecutive different IPs is as follows: (1) high-TIL counts in the TC represented sTILs, in order to identify LPBC (50% stromal TILs) samples. A cut-off ratio of the most robust indicator for LPBC, with a high proportion of CD8 10% lymphatic nucleus area to stromal area in the IM was used to + T cells, indicating favorable DFS in TNBC; (2) high PIL counts distinguish samples with intermediate and low immunoscores. We further included PILs, based on a cut-off ratio of 20% lymphatic nucleus area to with a high proportion of CD20+ B cells indicated poor DFS similar stromal area to investigate the influence of tumor immunity on the Para. to TILs mainly in the IM in luminal B subtype IBC; (3) PD-L1 Finally, five different types of IPs were determined, as shown in Fig. 2. positivity significantly correlated with the IC counts and could be predicted by IPs; and (4) IPs of IBC-NST might be a potential IHC prognosis factor, especially in the TNBC and luminal B subtypes. FFPE tissue specimens were sectioned at 4 μm thickness. Antibodies There are several limitations to our investigation. First, this study against CD4 (clone number UMAB64, ZSbio, Wuxi, China), CD8 (clone was a retrospective evaluation, so it is not clear whether the IP number SP16, ZSbio), CD20 (clone number L26, ZSbio), and PD-L1 (clone system may help predict the therapeutic responses. Second, the number SP142, Ventana Medical Systems, Oro Valley, AZ) were used for short clinical follow-up period for these patients further restricted IHC, which was performed using a Benchmark Ultra System, the OptiView the predictive value in terms of survival, especially for luminal DAB IHC Detection Kit (Ventana Medical Systems). A rabbit monoclonal tumors, where late recurrences are common. Third, future studies negative immunoglobulin (Ventana Medical Systems) was used as negative should explore the genetic variation underling tumor IPs to screen control, and FFPE tonsil tissue stained with PD-L1 was used as positive control. All negative and positive controls were used for each batch of for novel markers that can serve as potential immunotherapeutic samples. The PD-L1 staining scores of TILs (cut-off value: 1% staining of any targets in breast cancer. Despite these limitations, our study intensity) were evaluated by two board-certified pathologists according to provides unique insights into the tumor microenvironment 25,34 the IM passion 130 Trial criteria . The positivities of CD markers were architecture and its potential clinical prospects. assessed based on a digital-pathology computational workflow (Strata- 14,33 Quest 6.0.213 software for WSIs . The densities of CD4-, CD8-, and CD20-positive lymphocytes were evaluated through counting the numbers METHODS of nuclei encompassed by DAB-stained cytoplasm and membrane (Supplementary Fig. 1). A threshold was established to ensure the accurate Cohort enrollment labeling of positive cells for each region. The output parameters are The study cohort consisted of 579 consecutively archived IBC-NST samples presented as the number of positive lymphocytes/mm (for TCs, IMs, that were surgically excised at the time of diagnosis between 08/2015 and and Para). 08/2018. All enrolled patients provided written informed consent to use these samples for translational research, as approved by the Ethics DNA isolation and whole-exome sequencing Commission of the General Hospital of China PLA (approval number ky- We selected samples from 40 patients with different molecular subtypes 2020-1-4) and the study was compliant with the ‘Guidance of the Ministry and matched FFPE and blood samples to detect germline mutations. of Science and Technology (MOST) for the review and Approval of human Genomic DNA was extracted using a QIAamp DNA FFPE Tissue Kit or a Genetic Resources’, China. For TIL quantification, all specimens were DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany). Whole-exome reviewed for pathomorphology on H&E-stained sections to select the most sequencing libraries were constructed using an NEBNext Ultra DNA Library representative tumor block. Biomarker expression levels and other Prep Kit (Illumina, San Diego, CA) and according to the manufacturer’s clinicopathological features were extracted from the archives. Histological protocol. The average whole-exome sequencing depth was 100× for FFPE grades were scored according to the Nottingham semi-qualitative scoring tumors samples, and 60× for their normal tissue counterparts. system . Tumor molecular subtyping (expression of the associated parameters, including ER, PR, HER2, and Ki-67) and TNM staging were RNA isolation and Illumina immune-response gene panel evaluated based on World Health Organization’s classification of breast sequencing tumors (5th edition) . All H&E-stained and IHC-stained sections were scanned using a KFBIO Scans cope high-resolution scanner at 40× Macrodissections were performed on whole H&E-stained sections to magnification (Konfoong Biotech, Ningbo, China), and all parameters collect TILs from tumor areas and PILs from Para. Total RNA was extracted assessed in this study were examined using WSIs. using a miRNeasy FFPE Kit (Qiagen, Hilden, Germany) following the npj Breast Cancer (2022) 26 Published in partnership with the Breast Cancer Research Foundation X. Du et al. manufacturer’s instructions. RNA-sequencing libraries were constructed 11. Stamatelos, S. K., Bhargava, A., Kim, E., Popel, A. S. & Pathak, A. P. Tumor using the AmpliSeq Immune Response Panel (Illumina, San Diego, CA) ensemble-based modeling and visualization of emergent angiogenic hetero- according to the manufacturer’s protocol. Then, the quality of the library geneity in breast cancer. Sci. Rep. 9, 5276 (2019). (2 × 150 base-pair, paired-end reads) was checked, and it was sequenced 12. Camus, M. et al. Coordination of intratumoral immune reaction and human using a HiSeq 2500 System (Illumina, San Diego, CA). colorectal cancer recurrence. Cancer Res. 69, 2685–2693 (2009). 13. Angell, H. & Galon, J. From the immune contexture to the Immunoscore: the role of prognostic and predictive immune markers in cancer. Curr. Opin. Immunol. 25, Statistical analysis 261–267 (2013). Tumor sizes, metastatic lymph nodes, and the scores for CD4, CD8, CD20, 14. Mi, H. et al. Digital pathology analysis quantifies spatial heterogeneity of CD3, PD-L1, ER, PR, and Ki-67 staining were treated as continuous variables. They CD4, CD8, CD20, and FoxP3 immune markers in triple-negative breast cancer. were tested for normality and homogeneity of variance and were found Front Physiol. 11, 583333 (2020). not normally distributed and not homogenous. Non-parametric tests 15. Galon, J. & Bruni, D. Approaches to treat immune hot, altered and cold tumours (Mann-Whitney U test of Two-Independent Samples; Kruskal-Wallis H test with combination immunotherapies. Nat. Rev. Drug Disco. 18, 197–218 (2019). of K Independent samples) were conducted to assess the difference 16. Liang, Q., Ma, D., Gao, R. F. & Yu, K. D. Effect of Ki-67 expression levels and between groups. Some of these parameters were missing, and therefore, histological grade on breast cancer early relapse in patients with different omitted in subsequent analysis. We assessed the correlations between Immunohistochemical-based Subtypes. Sci. Rep. 10, 7648 (2020). patterns and categorical variables (clinical stage, histological grade, 17. Schreiber, R. D., Old, L. J. & Smyth, M. J. Cancer immunoediting: integrating HER2 status, molecular type, lobular involvement, and cancerous embolus) immunity’s roles in cancer suppression and promotion. Science 331, 1565–1570 by performing χ² tests for trends or Fisher’s exact tests. Survival endpoints (2011). were evaluated as DFS. Survival curves were prepared based on 18. Mlecnik, B. et al. Histopathologic-based prognostic factors of colorectal cancers Kaplan–Meier estimates and compared using the log-rank test. The Cox are associated with the state of the local immune reaction. J. Clin. Oncol. 29, proportional hazards model was used to perform univariable and 610–618 (2011). multivariable analyses to further identify the variables independently 19. Hendry, S. et al. Assessing tumor-infiltrating lymphocytes in solid tumors: a associated with DFS. Statistical analyses were performed using SPSS 23 practical review for pathologists and proposal for a standardized method from (IBM Corporation, Somers, NY). All statistical tests were 2-sided, and P the International Immunooncology Biomarkers Working Group: Part 1: assessing values < 0.05 were considered as statistically significant. the host immune response, TILs in invasive breast carcinoma and ductal carci- noma in situ, metastatic tumor deposits and areas for further research. Adv. Anat. Pathol. 24, 235–251 (2017). Reporting summary 20. Tan, T. T. & Coussens, L. M. Humoral immunity, inflammation and cancer. Opin. Further information on research design is available in the Nature Research Immunol. 19, 209–216 (2007). Reporting Summary linked to this article. 21. Shah, S. et al. Increased rejection of primary tumors in mice lacking B cells: inhibition of anti-tumor CTL and TH1 cytokine responses by B cells. Int J. Cancer 117, 574–586 (2005). DATA AVAILABILITY 22. He, Y., Jiang, Z., Chen, C. & Wang, X. Classification of triple-negative breast can- DNA-seq and RNA-seq datasets were uploaded on National center for biotechnology cers based on Immunogenomic profiling. J. Exp. Clin. Cancer Res. 37, 327 (2018). information (NCBI) under the Accession Number of PRJNA786713. The datasets 23. Van Berckelaer, C. et al. Infiltrating stromal immune cells in inflammatory breast generated and/or analyzed during the current study are available from the cancer are associated with an improved outcome and increased PD-L1 expres- corresponding author on reasonable request. sion. Breast Cancer Res 21, 28 (2019). 24. Buisseret, L. et al. Tumor-infiltrating lymphocyte composition, organization and PD-1/ PD-L1 expression are linked in breast cancer. Oncoimmunology 6, e1257452 Received: 17 May 2021; Accepted: 21 January 2022; (2017). 25. Hoda, R. S. et al. Clinical and pathologic features associated with PD-L1 (SP142) expression in stromal tumor-infiltrating immune cells of triple-negative breast carcinoma. Mod. Pathol. 33, 2221–2232 (2020). 26. Erber, R. & Hartmann, A. Understanding PD-L1 testing in breast cancer: a practical REFERENCES approach. Breast care 15, 481–490 (2020). 1. Curtis, C. et al. The genomic and transcriptomic architecture of 2,000 breast 27. Gonzalez-Ericsson, P. I. et al. The path to a better biomarker: application of a risk tumours reveals novel subgroups. Nature 486, 346–352 (2012). management framework for the implementation of PD-L1 and TILs as immuno- 2. Poudel, P., Nyamundanda, G., Patil, Y., Cheang, M. C. U. & Sadanandam, A. Het- oncology biomarkers in breast cancer clinical trials and daily practice. J. Pathol. erocellular gene signatures reveal luminal-A breast cancer heterogeneity and 250, 667–684 (2020). differential therapeutic responses. NPJ Breast Cancer 5, 21 (2019). 28. Gil Del Alcazar, C. R. et al. Immune escape in breast cancer during in situ to 3. Lehmann, B. D. et al. Identification of human triple-negative breast cancer sub- invasive carcinoma transition. Cancer Disco. 7, 1098–1115 (2017). types and preclinical models for selection of targeted therapies. J. Clin. Investig. 29. Toss, M. S. et al. The prognostic significance of immune microenvironment in 121, 2750–2767 (2011). breast ductal carcinoma in situ. Brit J. Cancer 122, 1496–1506 (2020). 4. Steenbrugge, J. et al. Anti-inflammatory signaling by mammary tumor cells 30. Savas, P. et al. Single-cell profiling of breast cancer T cells reveals a tissue-resident mediates prometastatic macrophage polarization in an innovative intraductal memory subset associated with improved prognosis. Nat. Med. 24, 986–993 mouse model for triple-negative breast cancer. J. Exp. Clin. Cancer Res. 37, 191 (2018). (2018). 31. Allison, K. H. WHO Classification of Tumours 5th Edition Breast Tumours. Lyon: 5. Denkert, C. et al. Tumour-infiltrating lymphocytes and prognosis in different IARC (2019). subtypes of breast cancer: a pooled analysis of 3771 patients treated with 32. Salgado, R. et al. The evaluation of tumor-infiltrating lymphocytes (TILs) in breast neoadjuvant therapy. Lancet Oncol. 19,40–50 (2018). cancer: recommendations by an International TILs Working Group 2014. Ann. 6. Yu, X. et al. Prognostic and predictive value of tumor-infiltrating lymphocytes in Oncol. 26, 259–271 (2015). breast cancer: a systematic review and meta-analysis. Clin. Transl. Oncol. 18, 33. Miller, T. J. et al. The prognostic and predictive value of SOX2(+) cell densities in 497–506 (2016). patients treated for colorectal cancer. Cancers 12, (2020). 7. Denkert, C. et al. Tumor-associated lymphocytes as an independent predictor of 34. Aktas, B. Y., Taban, H. & Aksoy, S. Atezolizumab and Nab-paclitaxel in advanced response to neoadjuvant chemotherapy in breast cancer. J. Clin. Oncol. 28, triple-negative breast cancer. N. Engl. J. Med. 380, 985–986 (2019). 105–113 (2010). 8. Marra, A., Viale, G. & Curigliano, G. Recent advances in triple negative breast cancer: the immunotherapy era. BMC Med. 17, 90 (2019). 9. Kim, J. Y., Kim, W. G., Kwon, C. H. & Park, D. Y. Differences in immune contextures among different molecular subtypes of gastric cancer and their prognostic ACKNOWLEDGEMENTS impact. Gastric Cancer 22, 1164–1175 (2019). We thank Xiaochang Zhang (Beijing Institute of Radiation Medicine) for his help in 10. Mani, N. L. et al. Quantitative assessment of the spatial heterogeneity of tumor- generating images with R software for this research. This work was supported by the infiltrating lymphocytes in breast cancer. Breast Cancer Res. 18, 78 (2016). National Natural Science Foundation of China (grant number 81972485). Published in partnership with the Breast Cancer Research Foundation npj Breast Cancer (2022) 26 X. Du et al. AUTHOR CONTRIBUTIONS Reprints and permission information is available at http://www.nature.com/ reprints Xue Du, Zhe Zhou, and Yun Shao carried out the studies, statistical analysis, and drafted the manuscript. These three authors contributed equally to this work and are Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims considered as “co-first author”. Kun Qian participated in sample collection and in published maps and institutional affiliations. patient follow-up. Yingfang Wu, Jun Zhang, Miao Cui, Jingjing Wang, and Yun Shao reviewed the H&E-stained sections and performed the TissueGnostics image analysis. Xue Du and Zhe Zhou were involved in next-generation sequencing (NGS) and NGS data analysis. Yanhong Tai and Shengqi Wang designed the studies and critically Open Access This article is licensed under a Creative Commons revised the manuscript for important intellectual content. Tai and Wang are Attribution 4.0 International License, which permits use, sharing, considered as corresponding authors. All authors read and approved the final adaptation, distribution and reproduction in any medium or format, as long as you give manuscript. appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless COMPETING INTERESTS indicated otherwise in a credit line to the material. If material is not included in the The authors declare no competing interests. article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. ADDITIONAL INFORMATION org/licenses/by/4.0/. Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41523-022-00389-y. © The Author(s) 2022 Correspondence and requests for materials should be addressed to Shengqi Wang or Yanhong Tai. npj Breast Cancer (2022) 26 Published in partnership with the Breast Cancer Research Foundation

Journal

npj Breast CancerSpringer Journals

Published: Feb 28, 2022

References