Abstract
The low overall survival rates of patients with breast cancer in sub-Saharan Africa (SSA) are driven by regionally differing tumor biology, advanced tumor stages at diagnosis, and limited access to therapy. However, it is not known whether regional differences in the composition of the tumor microenvironment (TME) exist and affect patients’ prognosis. In this international, multicentre cohort study, 1,237 formalin-fixed, paraffin-embedded breast cancer samples, including samples of the “African Breast Cancer-Disparities in Outcomes (ABC-DO) Study,” were analyzed. The immune cell phenotypes, their spatial distribution in the TME, and immune escape mechanisms of breast cancer samples from SSA and Germany (n = 117) were investigated using histomorphology, conventional and multiplex IHC, and RNA expression analysis. The data revealed no regional differences in the number of tumor-infiltrating lymphocytes (TIL) in the 1,237 SSA breast cancer samples, while the distribution of TILs in different breast cancer IHC subtypes showed regional diversity, particularly when compared with German samples. Higher TIL densities were associated with better survival in the SSA cohort (n = 400), but regional differences concerning the predictive value of TILs existed. High numbers of CD163+ macrophages and CD3+CD8+ T cells accompanied by reduced cytotoxicity, altered IL10 and IFNγ levels and downregulation of MHC class I components were predominantly detected in breast cancer samples from Western SSA. Features of nonimmunogenic breast cancer phenotypes were associated with reduced patient survival (n = 131). We therefore conclude that regional diversity in the distribution of breast cancer subtypes, TME composition, and immune escape mechanisms should be considered for therapy decisions in SSA and the design of personalized therapies.
Introduction
Cancer is an increasing public health concern in sub‐Saharan Africa (SSA), with an increase of 70% predicted by 2030, including an increase in female breast cancer (1–3). Because of diagnosis at an advanced stage, a weak health infrastructure, and limited treatment options, the survival rates of patients with breast cancer in SSA are particularly low (4–6). Furthermore, in the United States, patients with breast cancer of African-American ancestry have a poorer prognosis than their Caucasian counterparts, even after accounting for differences in access to health care (7). Despite limited information regarding patients with breast cancer from SSA, geographical diversity in breast cancer biology in this area has been suggested (8, 9). Because the SSA population is highly heterogeneous in terms of genetic variation, differences in immunologic features have been proposed as predictors of the overall survival (OS) of SSA patients (10, 11).
Breast cancer has been characterized as a heterogeneous disease comprising hormone receptor (HR)+Her2−, HR+Her2+, HR−Her2+, and triple-negative tumors that differ regarding their prognosis (12). Because of the heterogeneous distribution of breast cancer subtypes in SSA, regional differences have been suggested (8, 13–17). In addition, the quality of tumor processing is highly variable between countries, which may also negatively influence data interpretation (13).
The composition of the tumor microenvironment (TME) driven by intrinsic and extrinsic factors may also influence breast cancer patients’ prognosis in SSA (18–20). Intrinsic immune escape strategies for breast cancer include downregulation of MHC class I and upregulation of coinhibitory immune checkpoint molecules, such as programmed death ligand 1 (PD-L1), accompanied by reduced antitumoral immune responses and worse patient outcomes (20). In addition, the TME influences tumor progression, the response to therapy and outcomes (21). While an increased frequency of immune suppressive cells, a low frequency of CD8+ T lymphocytes and impaired cytotoxic immune responses are associated with reduced survival of patients with breast cancer, higher frequencies of tumor-infiltrating lymphocytes (TIL) are correlated with a favorable prognosis of breast cancer (22).
Currently, whether TIL counts in breast cancer tumors differ between ethnicities is the subject of controversial discussion. One study demonstrated only subtle differences within triple-negative tumors (23), whereas another study reported substantially higher frequencies of protumorigenic, suppressive immune cells in patients with breast cancer with African-American ancestry (24). So far, to the best of our knowledge, there are no published data focusing on the biological diversity of the TME of breast cancer from SSA. Therefore, the aim of this international, multicentre study was to decipher TME variations and immune escape strategies in 1,237 breast cancer patient samples from 10 countries in SSA and to compare them with European breast cancer samples and to analyze their impact on mortality.
Materials and Methods
Patients’ characteristics and study design
Formalin-fixed and paraffin-embedded (FFPE) tumor tissue blocks of 1,497 female patients with breast cancer from SSA were prospectively collected between 2006 and 2020 across 10 countries (Table 1). Samples were collected by local pathologists, the African Cancer Registry Network, and the International Agency for Research on Cancer. When possible, patient's clinical data and follow-up data (n = 400) were obtained. The patients from Southern Africa (SA) and Western Africa (WA) with available survival data were part of the African Breast Cancer-Disparities in Outcomes (ABC-DO) Study (6), and these survival data were truncated at 3 years. Because of the lack of detailed information on staging, all tumors >5 cm or with known metastasis were defined as advanced-stage tumors. On the basis of nonrepresentative tumor areas, 250 breast cancer samples were excluded, resulting in the inclusion of 1,237 breast cancer samples in the current study (Supplementary Fig. S1). Histopathological diagnoses of breast cancer were made according to the World Health Organization (WHO) classification of Breast Tumors 5th edition (25), including histopathological grading according Elston and Ellis. All samples were analyzed by conventional IHC using antibodies (Ab) directed against the estrogen receptor (ER), progesterone receptor (PR), Her2, and Ki-67 (Supplementary Table S1). As a control, 117 German (GER) breast cancer samples with similar proportions of IHC subtypes were collected at the Institute of Pathology of the Martin-Luther-University Halle-Wittenberg, Germany, 67 of 117 breast cancer samples are part of the PiA study (NCT01592825; ref. 26). This research weas performed in accordance with the Declaration of Helsinki. All data related to clinical samples were approved by local ethical institutions, as listed in Supplementary Table S2. If possible, informed written consent for the use of samples for research purposes was obtained from patients.
Characteristics of breast cancer samples analyzed.
Parameters . | Eastern SSA n = 444 (35.9%) . | Central SSA n = 99 (8.0%) . | Southern SSA n = 242 (19.6%) . | Western SSA n = 452 (36.5%) . | Caucasians n = 117 . |
---|---|---|---|---|---|
Countries (N) | Ethiopia (289)* | Congo (99) | Namibia (137)* | Ivory Coast (94) | Germany (117) |
Malawi (16) | South Africa | Mali (83) | |||
Mauritius (72) | (105)* | Nigeria (275)* | |||
Uganda (67)* | |||||
Woman of Black Ethnicity (%) | 100% | >95% | >95% | >95% | 0% |
Mean age (range) | 46.6 (16–97) | 49.1 (27–86) | 55.5 (25–97) | 47.6 (22–91) | 64.3 (35–85) |
Tumor size (cm) | |||||
<5 (% of known) | 173 (64.6%) | 7 (13.2%) | 100 (42.7%) | 72 (37.9%) | 71 (60.7%) |
>5 (% of known) | 95 (35.4%) | 46 (96.8%) | 134 (57.3%) | 118 (62.1%) | 46 (39.3%) |
Unknown | 176 | 46 | 8 | 190 | NA |
Simplified tumor stage | |||||
Early | 192 (43.2%) | 7 (7.0%) | 100 (41.2%) | 87 (19.2%) | 71 (60.7%) |
Advanced | 139 (31.1%) | 46 (46.5%) | 134 (55.4%) | 168 (37.2%) | 46 (39.3%) |
Unknown | 113 (25.5%) | 46 (46.5%) | 8 (3.3%) | 197 (43.6%) | 0 (0.0%) |
Available survival data | 139 (31.3%) | 0 (0.0%) | 192 (79.3%) | 69 (16.2%) | 0 (0.0%) |
Number of patients alive after 3 years follow-up | 80 (57.5%) | NA | 129 (67.2%) | 26 (35.6%) | NA |
Histology | |||||
NST | 365 (82.2%) | 89 (89.8%) | 194 (80.2%) | 390 (86.2%) | 87 (74.4%) |
non-NST | 79 (17.8%) | 10 (10.2%) | 48 (19.8%) | 62 (13.8%) | 30 (25.6%) |
Grading | |||||
G1 | 23 (5.2%) | 3 (3.0%) | 40 (16.5%) | 28 (6.2%) | 18 (15.4%) |
G2 | 182 (41.0%) | 38 (38.4%) | 101 (41.7%) | 188 (41.6%) | 64 (54.7%) |
G3 | 237 (53.4%) | 58 (58.6%) | 96 (41.8%) | 236 (52.2%) | 35 (29.9%) |
IHC subtype | |||||
Luminal A–like | 132 (29.7%) | 28 (28.3%) | 89 (36.8%) | 89 (19.7%) | 72 (61,5.0%) |
Luminal B–like | 163 (36.7%) | 37 (37.4%) | 94 (38.8%) | 113 (25.0%) | 21 (18,0%) |
Her2+ | 48 (10.8%) | 13 (13.1%) | 17 (7.0%) | 72 (15.9%) | 7 (6.0%) |
TNBC | 101 (22.7%) | 21 (21.2%) | 42 (17.4%) | 178 (39.4%) | 17 (14.5%) |
TILs | |||||
Low | 210 (47.2%) | 49 (49.5%) | 140 (57.9%) | 219 (48.5%) | 60 (51.3%) |
Intermediate | 161 (36.6%) | 41 (41.4%) | 73 (30.2%) | 182 (40.3%) | 47 (40.2%) |
High | 73 (16.3%) | 9 (9.1%) | 29 (12.0%) | 51 (11.3%) | 10 (8.5%) |
PD-L1 expression by TILs | |||||
Negative | 186 (41.9%) | 34 (34.3%) | 98 (40.5%) | 260 (57.5%) | 89 (76.2%) |
Positive | 176 (39.6%) | 63 (63.6%) | 73 (30.2%) | 169 (37.4%) | 28 (23.8%) |
Missing | 82 (18.5%) | 2 (2.1%) | 71 (29.3%) | 23 (5.1%) | 0 (0.0%) |
MHC class I expression | |||||
HC (range) | 137.3 (0–300) | 201.1 (30–300) | 148.9 (10–300) | 113.3 (0–300) | 175.5 (10–300) |
β2-m (range) | 166.5 (0–300) | 182.8 (0–300) | 193.0 (10–300) | 123.1 (0–300) | 186.6 (40–300) |
Parameters . | Eastern SSA n = 444 (35.9%) . | Central SSA n = 99 (8.0%) . | Southern SSA n = 242 (19.6%) . | Western SSA n = 452 (36.5%) . | Caucasians n = 117 . |
---|---|---|---|---|---|
Countries (N) | Ethiopia (289)* | Congo (99) | Namibia (137)* | Ivory Coast (94) | Germany (117) |
Malawi (16) | South Africa | Mali (83) | |||
Mauritius (72) | (105)* | Nigeria (275)* | |||
Uganda (67)* | |||||
Woman of Black Ethnicity (%) | 100% | >95% | >95% | >95% | 0% |
Mean age (range) | 46.6 (16–97) | 49.1 (27–86) | 55.5 (25–97) | 47.6 (22–91) | 64.3 (35–85) |
Tumor size (cm) | |||||
<5 (% of known) | 173 (64.6%) | 7 (13.2%) | 100 (42.7%) | 72 (37.9%) | 71 (60.7%) |
>5 (% of known) | 95 (35.4%) | 46 (96.8%) | 134 (57.3%) | 118 (62.1%) | 46 (39.3%) |
Unknown | 176 | 46 | 8 | 190 | NA |
Simplified tumor stage | |||||
Early | 192 (43.2%) | 7 (7.0%) | 100 (41.2%) | 87 (19.2%) | 71 (60.7%) |
Advanced | 139 (31.1%) | 46 (46.5%) | 134 (55.4%) | 168 (37.2%) | 46 (39.3%) |
Unknown | 113 (25.5%) | 46 (46.5%) | 8 (3.3%) | 197 (43.6%) | 0 (0.0%) |
Available survival data | 139 (31.3%) | 0 (0.0%) | 192 (79.3%) | 69 (16.2%) | 0 (0.0%) |
Number of patients alive after 3 years follow-up | 80 (57.5%) | NA | 129 (67.2%) | 26 (35.6%) | NA |
Histology | |||||
NST | 365 (82.2%) | 89 (89.8%) | 194 (80.2%) | 390 (86.2%) | 87 (74.4%) |
non-NST | 79 (17.8%) | 10 (10.2%) | 48 (19.8%) | 62 (13.8%) | 30 (25.6%) |
Grading | |||||
G1 | 23 (5.2%) | 3 (3.0%) | 40 (16.5%) | 28 (6.2%) | 18 (15.4%) |
G2 | 182 (41.0%) | 38 (38.4%) | 101 (41.7%) | 188 (41.6%) | 64 (54.7%) |
G3 | 237 (53.4%) | 58 (58.6%) | 96 (41.8%) | 236 (52.2%) | 35 (29.9%) |
IHC subtype | |||||
Luminal A–like | 132 (29.7%) | 28 (28.3%) | 89 (36.8%) | 89 (19.7%) | 72 (61,5.0%) |
Luminal B–like | 163 (36.7%) | 37 (37.4%) | 94 (38.8%) | 113 (25.0%) | 21 (18,0%) |
Her2+ | 48 (10.8%) | 13 (13.1%) | 17 (7.0%) | 72 (15.9%) | 7 (6.0%) |
TNBC | 101 (22.7%) | 21 (21.2%) | 42 (17.4%) | 178 (39.4%) | 17 (14.5%) |
TILs | |||||
Low | 210 (47.2%) | 49 (49.5%) | 140 (57.9%) | 219 (48.5%) | 60 (51.3%) |
Intermediate | 161 (36.6%) | 41 (41.4%) | 73 (30.2%) | 182 (40.3%) | 47 (40.2%) |
High | 73 (16.3%) | 9 (9.1%) | 29 (12.0%) | 51 (11.3%) | 10 (8.5%) |
PD-L1 expression by TILs | |||||
Negative | 186 (41.9%) | 34 (34.3%) | 98 (40.5%) | 260 (57.5%) | 89 (76.2%) |
Positive | 176 (39.6%) | 63 (63.6%) | 73 (30.2%) | 169 (37.4%) | 28 (23.8%) |
Missing | 82 (18.5%) | 2 (2.1%) | 71 (29.3%) | 23 (5.1%) | 0 (0.0%) |
MHC class I expression | |||||
HC (range) | 137.3 (0–300) | 201.1 (30–300) | 148.9 (10–300) | 113.3 (0–300) | 175.5 (10–300) |
β2-m (range) | 166.5 (0–300) | 182.8 (0–300) | 193.0 (10–300) | 123.1 (0–300) | 186.6 (40–300) |
Note: Follow-up data were available from countries marked with an asterisk (*).
Abbreviations: β2-m, β2-microglobuline; HC, heavy chain; Her2+, human epidermal growth factor receptor 2 positive; NA, not applicable; PD-L1, programmed death ligand-1; TNBC, triple-negative BC.
Quality score
To quantify tissue quality, a quality score was developed using the following parameters: (i) general visual impression (from +1 for “bad” to +5 points for “very good”); (ii) staining of non-neoplastic breast glands for ER (−1 or +1 point); (iii) staining of proliferating cells (internal control via stained mitosis) using an anti-Ki-67 ab (−1 or +1 point); and (iv) whether the tissue detached during the staining procedure (−1 or +1 point). Samples with scores <2 points were excluded.
Histopathology and IHC
Breast cancer subtypes were classified according to the surrogate IHC subtype classification published in 2013 (12). TILs were analyzed on hematoxylin and eosin (H&E) slides as described elsewhere (27). IHC staining was performed on a Bond III automated immunostainer (Leica Biosystems Nussloch GmbH) using the Bond Polymer Refine Detection Kit (DS9800-CN). Antibodies against beta-2-microgloblin (β2-m), major histocompatibility complex (MHC) class I heavy chain (HC), programmed death receptor (PD)-1, PD-L1, perforin, granzyme B, phosphorylated signal transducer and activator of transcription (pSTAT)1, and tapasin (tpn; Supplementary Table S1) were applied as recommended by the manufacturer. PD-L1 testing was performed with the CAL10 antibody clone (Zytomed), which showed a good correlation with the widely used SP142 antibody clone (Ventana; ref. 28). Immunostaining was assessed by using semiquantitative H-scoring as described elsewhere (29).
RNA isolation and expression analysis
Prior to RNA isolation, the breast cancer tissues were microdissected from 505 samples. Microdissection was performed by a pathologist using a Zeiss axioscope 5 microscope (Zeiss) and tumor tissue was marked with a pencil to exclude preexisting breast parenchyma. For RNA isolation, two to four 10-μm-thick tissue slides were used. Deparaffinization of FFPE tissues was performed with 2 × xylene for 5 minutes, followed by incubation in 96% and 70% ethanol for 2 minutes each. Proteinase K digestion (Qiagen, catalog no.19131) was performed for up to 2 hours at room temperature followed by 15 minutes at 80°C. RNA was isolated with a Qiagen miRNeasy FFPE Kit (Qiagen, catalog no. 217504) according to the manufacturer's instructions. RNA expression analysis was performed using the NanoString nCounter XT Assay according to the Hybridization protocol for the nCounter XT CodeSet Gene Expression Assay (NanoString nCounter) and data analyzed with the nCounter Expression Data Analysis Guide (MAN- C0011-04 from 2017). From each sample, 400 ng of FFPE-derived RNA were used. All steps were performed as described in nCounter XT Assay User Manual (MAN-10023-11 from 2016). For the analysis, a custom codeset design was used and all respective genes are summerized in Supplementary Table S3. The expression levels were evaluated with the NanoStringNorm (https://github.com/sgrote/NanoStringNormalizeR/) R package.
Construction of tissue microarrays
FFPE tissues from 374 patients with intermediate to very high tissue quality and with sufficient material were included in tissue microarrays containing two 0.6-mm tissue cylinders of each donor block using a manual tissue arrayer (Beecher Instruments Inc.). Two tissue cores from the tumor centre and the invasive margin were selected and arrayed on a recipient paraffin block.
Multispectral imaging, spatial distribution analysis, and immunoscoring
The frequency and localization of immune cell subpopulations and cancer cells from 374 samples was determined by multispectral imaging (MSI) and is representatively shown in Fig. 1. The staining procedure was performed as described recently (30). The Ab panel included CD3, CD8, FOXP3, CD20, CD163, and panCK (Supplementary Table S1). For imaging, the PerkinElmer Vectra Polaris platform was employed. Cell segmentation and phenotyping were performed using the inForm software (PerkinElmer Inc., Version 2.4). The frequency and spatial distribution of cell populations were analyzed using an R script (https://github.com/akoyabio).
TILs and their quantitative distribution within the TME of breast cancer samples. Representative H&E staining of FFPE breast cancer samples with low (A), intermediate (int.; B), and high (C) infiltration of TILs. Scale bars, 20 μm. In total, TILs were analyzed in 1,237 samples. MSI was performed on 374 samples with representatively shown pictures with low (D), intermediate (E), and high (F) infiltration of TILs. MSI allows visualization of immune cell subpopulations including CD3+CD8− T cells (yellow), CD3+CD8+ T cells (red), CD3+FOXP3+ T cells (turquoise), CD163+ macrophages (orange), and CD20+ B cells (green) as well as pan-cytokeratin+ cancer cells (gray). G, Box plots (line represents the mean) indicating that TILs were found to be higher in luminal-B–like (LuB), HER2-positive (Her2+) and TNBC subtypes, compared with luminal-A–like (LuA; P < 0.0001, multivariate analysis). H, No differences in the quantity of TILs in different regions in SSA were detected (Eastern Africa, EA; Southern Africa, SA; Central Africa, CA; Western Africa, WA), while the TIL frequency was tendentially lower in GER breast cancer samples. I, Kaplan–Meier curve: the amounts of TILs indicates a prognostic impact with poorer survival in breast cancer with lower numbers of TILs (P = 0.001, log-rank Test). J, Pearson correlation map with association levels: a higher numbers of TILs are linked with increased frequencies of all immune cell subpopulations analyzed. Pearson correlation coefficients are represented by different colors defined in the scale bar on the right side of the correlation map. The immune cell subpopulations analyzed by MSI revealed higher numbers of CD3+ T cells (K), CD3+CD8+ T cells and CD163+ macrophages (M2; L) in WA. No relevant difference in the frequency of CD20+ B cells was shown. However, in GER samples, the frequency of regulatory T cells was slightly higher when compared with those in all regions of SSA. Except for the log-rank tests, all P values were estimated in multivariate analyses and are adjusted for differences in age, grading, quality score, stage, IHC subtype and region of origin as confounding factors. All P values are shown in the graphs.
TILs and their quantitative distribution within the TME of breast cancer samples. Representative H&E staining of FFPE breast cancer samples with low (A), intermediate (int.; B), and high (C) infiltration of TILs. Scale bars, 20 μm. In total, TILs were analyzed in 1,237 samples. MSI was performed on 374 samples with representatively shown pictures with low (D), intermediate (E), and high (F) infiltration of TILs. MSI allows visualization of immune cell subpopulations including CD3+CD8− T cells (yellow), CD3+CD8+ T cells (red), CD3+FOXP3+ T cells (turquoise), CD163+ macrophages (orange), and CD20+ B cells (green) as well as pan-cytokeratin+ cancer cells (gray). G, Box plots (line represents the mean) indicating that TILs were found to be higher in luminal-B–like (LuB), HER2-positive (Her2+) and TNBC subtypes, compared with luminal-A–like (LuA; P < 0.0001, multivariate analysis). H, No differences in the quantity of TILs in different regions in SSA were detected (Eastern Africa, EA; Southern Africa, SA; Central Africa, CA; Western Africa, WA), while the TIL frequency was tendentially lower in GER breast cancer samples. I, Kaplan–Meier curve: the amounts of TILs indicates a prognostic impact with poorer survival in breast cancer with lower numbers of TILs (P = 0.001, log-rank Test). J, Pearson correlation map with association levels: a higher numbers of TILs are linked with increased frequencies of all immune cell subpopulations analyzed. Pearson correlation coefficients are represented by different colors defined in the scale bar on the right side of the correlation map. The immune cell subpopulations analyzed by MSI revealed higher numbers of CD3+ T cells (K), CD3+CD8+ T cells and CD163+ macrophages (M2; L) in WA. No relevant difference in the frequency of CD20+ B cells was shown. However, in GER samples, the frequency of regulatory T cells was slightly higher when compared with those in all regions of SSA. Except for the log-rank tests, all P values were estimated in multivariate analyses and are adjusted for differences in age, grading, quality score, stage, IHC subtype and region of origin as confounding factors. All P values are shown in the graphs.
Statistical analysis
The Mann–Whitney U test was employed to compare clinical data. Variables originating from gene expression analyses were assessed by multivariate regression using generalized linear models (Binomial, Poisson or Gamma families, depending on the response variable) to estimate ORs and confidence intervals (CI) while adjusting for the following confounding factors: age (continuous variable, by year), grading (binary variable, grade 3 vs. 1 or 2), quality score (discrete quantitative variable, 1 to 10), stage (binary variable, advanced vs. early), IHC subtype (categorical variable), and region of origin (categorical variable). Higher OR denoted concordance between higher gene expression values and increasing variable values (age, grading, quality, and stage) or as compared with a reference category (“luminal A–like” in the case of IHC subtype and “East” in the case of Region). Missing age entries were imputed using the median age of patients of the country of origin. Otherwise, patients with missing information in any other variable were excluded from regression analyses. Multivariate regressions were performed using the statsmodels library for Python (www.statsmodels.org). Linear correlations between variables were estimated as Pearson correlations. Survival analyses were performed on 400 patients from SSA (recruited between 2005 and 2017, maximum follow-up time of 36 months) using Kaplan–Meier estimators and with differences calculated with log-rank tests (unadjusted) or Cox proportional hazards models (adjusted for confounding factors). Cox proportional hazards models were implemented and plotted alongside the corresponding Kaplan–Meier curves using the Lifelines library for Python (https://lifelines.readthedocs.io/). Except when stated otherwise, all statistical analyses were performed using IBM SPSS Statistics or GraphPad Prism v9.
Data availability
The data generated in this study are available in the article and its Supplementary Data or upon request from the corresponding author.
Results
General epidemiologic characteristics
Regional variations in clinical, pathologic, and immunologic parameters of patients with breast cancer are shown in Table 1. Among women with tumors of known stage, early-stage tumors were the least common in samples from Central Africa (CA; 13%) but higher in those from WA (38%), SA (43%), and Eastern Africa (EA; 58%). Regional differences in breast cancer subtypes exceeded 10% absolute differences for luminal A–like tumors (ranging from 19% of tumors in WA to 37% in SA), luminal B–like (25% of WA tumors compared with 37%–38% in all other regions) and triple-negative breast cancer (TNBC; almost 40% in WA vs. 17%–23% in other regions).
Breast cancer subtype–specific regional differences in the frequency of TILs
The prevalence of TILs was assessed in 1,237 breast cancer samples. Univariate analysis showed low (1%–9.9%) or intermediate (10%–39.9%) infiltration of TILs in most breast cancer samples, while only a few cases demonstrated high TIL infiltration (>40%), as shown in Fig. 1A–C. An increased frequency of TILs was found in luminal B–like (adjusted OR: 1.42, 95% CI: 1.23–1.64), Her2+ (OR: 1.48, 95% CI: 1.23–1.79), and TNBC (OR: 1.25, 95% CI: 1.07–1.46) when compared with the luminal A–like subtype (Fig. 1G; Supplementary Fig. S2A). In multivariable analysis, neither regional differences in the TIL frequency within SSA (Fig. 1H) nor an association with the tissue quality, tumor stage, menopausal status, or age was detected (Supplementary Fig. S2A), but higher grading and the luminal B–like, Her2+, and TNBC subtypes were associated with higher TIL densities. However, after adjustment for age and clinical features, no relevant differences in TIL frequencies could be demonstrated in comparison with GER samples (n = 117). Of note, the distribution of TILs within the respective IHC subtypes showed regional diversity, with only GER samples showing the highest TIL densities in TNBC (Supplementary Fig. S2B). Low TILs correlated with a worse survival probability independently of other clinicopathological parameters [n = 400; Fig. 1I, HR = 2.953 (95% CI: 1.624–5.371), P = 0.005]. As shown in Supplementary Fig. S2C, TILs are a prognostic marker particularly in ER− tumors. Moreover, the prognostic impact of TILs was the highest in EA, while in SA and WA the significance was lower (Supplementary Fig. S2D).
Correlation of the TIL frequency with the immune cell composition of breast cancer
The density of TILs was positively correlated with CD2, CD3, CD4, CD8, CD20, and FOXP3 expression (Fig. 1J, P < 0.0001), while CD56 expression was relatively lower. Using multiplex IHC, the frequency and localization of CD3+, CD3+CD8−, CD3+CD8+, CD3+FOXP3+ T cells, CD20+ B cells, and CD163+ M2 macrophages in relation to each other as well as to panCK+ breast cancer cells were analyzed (Fig. 1D–F). A positive correlation was found between the number of TILs and the frequency of CD3+, CD3+CD8−, CD3+CD8+, CD3+FOXP3+ T cells and CD20+ B cells (P < 0.0001), while no association between TILs and CD163+ M2 macrophages was detected (P = 0.242; Fig. 1D–F).
Regional differences in the immune cell repertoire of breast cancer
Regarding immune cell composition, WA breast cancer samples had a higher mean frequency of CD3+ T cells (particularly CD3+CD8+) than EA and GER breast cancer samples (OR: 1.93, 65% CI: 1.26–2.94; Fig. 1K; Supplementary Fig. S3). In all intrinsic breast cancer subtypes, the proportion of CD3+CD8+ T cells was the highest in WA within SSA, but comparable with GER samples. The percentage of CD20+ B cells and CD3+FOXP3+ T cells in breast cancer samples did not differ between SSA regions, whereas the frequency of CD163+ M2 macrophages was higher in samples from WA (OR: 1.92, 95% CI: 1.42–2.61; Fig. 1M; Supplementary Fig. S3C). The link between TME and tumor-specific features upon adjustment for confounders is given in Supplementary Fig. S2A and S3A–S3C.
Altered spatial distribution of immune cell subpopulations in breast cancer of distinct regions
The frequency of the different immune cell subpopulations was on average higher at the invasive margin than in the tumor centre. Increased numbers of CD163+ macrophages were found in the tumor centre of WA samples (P = 0.001). Spatial distribution analysis revealed a shorter mean distance of 74.8 μm between CD3+CD8− and CD3+CD8+ T cells in WA samples compared with a mean distance of 155.1 μm in breast cancer samples from all other regions (Fig. 2A, P < 0.0001). In addition, a closer proximity of CD163+ macrophages to CD3+ T-cell subsets was found in breast cancer samples from WA, with an average of 86.5 μm compared with 128.6 μm in samples from other regions (Fig. 2B, P = 0.005). Regarding the minimal distance between tumor cells and CD3+CD8+ T cells, closer contact was found in WA samples, with an average of 9.4 μm compared with 14.2 μm in samples from other regions (Fig. 2C, P = 0.025). The minimal distance between tumor cells and CD3+CD8+ T cells was closer in tumors positive for PD-L1 (Fig. 2D, P = 0.059).
Spatial immune cell distribution in synopsis with PD-L1 expression in breast cancer of SSA. Box plots (line represents the mean) showing that regardless of PD-L1 expression, a closer proximity of CD3+CD8− and CD3+CD8+ T cells is evident for Western SSA (WA) samples (A). A closer spatial proximity of CD3+ T cells and CD163+ macrophages exist in WA (B). Finally, a closer spatial proximity of CD3+CD8+ T cells and tumor cells was present in WA samples as well (C). Depending on the PD-L1 status, minor differences in the spatial proximity of CD3+CD8+ T cells and tumor cells could be shown, with a slightly higher proximity in PD-L1–positive samples (D). Representative IHC staining of breast cancer samples without (E) and with low (F) and high (G) PD-L1 expression. Scale bars, 50 μm. The prevalence of both breast cancer samples positive for PD-L1, analyzed with the immune cell score (H) and the tumor proportion score (I), was higher in TNBC. However, only minor differences were seen in the respective regions within SSA (J), while PD-L1 expression was significantly lower in GER samples. Kaplan–Meier curves show better OS of patients with PD-L1 breast cancer in both Her2+ (K) and TNBC (L) subtypes. Except for the log-rank tests, all P values were estimated in multivariate analyses and are adjusted for differences in age, grading, quality score, stage, IHC subtype, and region of origin as confounding factors.
Spatial immune cell distribution in synopsis with PD-L1 expression in breast cancer of SSA. Box plots (line represents the mean) showing that regardless of PD-L1 expression, a closer proximity of CD3+CD8− and CD3+CD8+ T cells is evident for Western SSA (WA) samples (A). A closer spatial proximity of CD3+ T cells and CD163+ macrophages exist in WA (B). Finally, a closer spatial proximity of CD3+CD8+ T cells and tumor cells was present in WA samples as well (C). Depending on the PD-L1 status, minor differences in the spatial proximity of CD3+CD8+ T cells and tumor cells could be shown, with a slightly higher proximity in PD-L1–positive samples (D). Representative IHC staining of breast cancer samples without (E) and with low (F) and high (G) PD-L1 expression. Scale bars, 50 μm. The prevalence of both breast cancer samples positive for PD-L1, analyzed with the immune cell score (H) and the tumor proportion score (I), was higher in TNBC. However, only minor differences were seen in the respective regions within SSA (J), while PD-L1 expression was significantly lower in GER samples. Kaplan–Meier curves show better OS of patients with PD-L1 breast cancer in both Her2+ (K) and TNBC (L) subtypes. Except for the log-rank tests, all P values were estimated in multivariate analyses and are adjusted for differences in age, grading, quality score, stage, IHC subtype, and region of origin as confounding factors.
Breast cancer subtype–specific PD-L1 expression and its clinical relevance
The immune inhibitory molecule PD-L1, known to be frequently overexpressed in cancer (31) was assessed in 1,059 breast cancer samples (Fig. 2E–G). PD-L1 staining of tumor cells was found in 28.3% of the breast cancer samples. In all, 45.4% of breast cancer samples expressed PD-L1 on TILs, and less frequently on tumor cells (Fig. 2H and I), which both positively correlated with the number of TILs (P < 0.0001) and the frequency of CD3+ T cells (P = 0.004). In general, a higher proportion of PD-L1+ samples was demonstrated in TNBC (OR: 2.96, 95% CI: 1.86–4.71) than in luminal A–like IHC subtypes (Fig. 2H), which was only slightly increased in Her2+ (OR: 1.53, 95% CI: 0.87–4.71) and luminal B–like (OR: 1.25, 95% CI: 0.79–1.97). Furthermore, slightly lower expression of PD-L1 was observed in WA samples (Supplementary Fig. S4A) as well as in GER samples. PD-L1 expression had no impact on breast cancer patients’ outcomes in the total cohort but was correlated with improved survival in Her2+ and TNBC subtypes (Fig. 2K and L, HR: 1.90 and 1.19, respectively).
Low regional differences in the function of immune cells
Because RNA expression analyses revealed low or undetectable expression of IFNγ, IL2, IL7, IL10, and IL15 in samples of low tissue quality, only breast cancer samples of intermediate to very high quality were evaluated for cytokine expression (n = 505). The expression of IL2, IL7, and IL15 was very low in all breast cancer samples analyzed, whereas higher IL10 and IFNγ mRNA expression levels were detected in advanced-stage tumors (Fig. 3A and B), with the lowest expression of IL10 in samples from SA and slightly higher expression in samples from WA (Fig. 3A). The IFNγ expression levels were highly variable in WA samples, with a 3-fold higher mean value compared with breast cancer samples from EA (OR: 1.98, 95% CI: 1.03–3.81; Fig. 3B). Higher IFNγ expression was positively correlated with the TIL density (P < 0.001, Fig. 3C). The expression of the cytotoxic markers granzyme B and perforin was low, ranging between 0% and 10%, with the lowest mean frequencies of perforin in samples from WA (Supplementary Fig. S4B and S4C).
Cytokine expression and its influence on immune cell infiltration. The expression of IL10 (A) and IFNγ (B) is visualized with forest plots. Only samples with detectable RNA of the respective genes were considered. The different sample sizes correspond to the number of specimens for which data were available for all variables examined in each test. Pearson correlation map (C) showing the connection of IFNγ expression and TILs and immune cell subpopulations. Pearson correlation coefficient is displayed by different colours defined in the scale bar on the right-hand side of the figure.
Cytokine expression and its influence on immune cell infiltration. The expression of IL10 (A) and IFNγ (B) is visualized with forest plots. Only samples with detectable RNA of the respective genes were considered. The different sample sizes correspond to the number of specimens for which data were available for all variables examined in each test. Pearson correlation map (C) showing the connection of IFNγ expression and TILs and immune cell subpopulations. Pearson correlation coefficient is displayed by different colours defined in the scale bar on the right-hand side of the figure.
Regional differences in the frequency of MHC class I–mediated immune escape in breast cancer
Because IFNγ is a regulator of MHC class I components (32), the expression of the MHC class I HC, β2-m, and the antigen processing and presentation machinery (APM) component tpn was determined (Fig. 4A–F). Lower expression of these proteins was detected in breast cancer samples from WA as compared with those from SA (OR: 2.06, 95% CI: 1.02–4.16), CA (OR: 7.72, 95% CI: 1.98–30.18), and Germany (OR: 1.00, 95% CI: 1.0–1.0, Fig. 4H–J; Supplementary Fig. S5A). No major differences in MHC class I HC and β2-m expression were detected regarding the breast cancer subtypes (HC: P = 0.292, β2-m: P = 0.190; Fig. 4G), age, grading or clinical stage (Supplementary Fig. S5A and S5B). Furthermore, breast cancer samples with lower expression of the MHC class I HC (P = 0.006) showed poorer patient survival rates independently of other clinicopathological parameters (n = 131; Fig. 4K; Supplementary Fig. S6). Because basal MHC class I expression depends on functional IFNγ signaling (33), breast cancer samples were also analyzed for pSTAT1 (Fig. 4L–N), which was positively correlated with IFNG RNA expression (P = 0.006), with the highest expression in breast cancer samples from CA and SA.
JAK/STAT signaling and MHC class I pathway component expression in SSA. Representative IHC stainings of breast cancer samples with MHC class I HC (A and D), β2-m (B and E), and the APM component tpn (C and F) are shown with no or weak and strong staining patterns, respectively. A lower expression of HC (H), β2-m (I), and tpn (J) was found in WA, while no differences can be demonstrated in the respective intrinsic subtypes (G). K, The prognostic influence of MHC class I expression is representatively shown for HC. Kaplan–Meier curve illustrates poor survival in patients with low HC expression in SSA (P = 0.006, log-rank test). Activation of IFNγ signaling via phosphorylated pSTAT1 is shown with representative low, intermediate, and high nuclear expression analyzed by IHC (L and M). All P values were estimated in multivariate analyses and are adjusted for differences in age, grading, quality score, stage, IHC subtype, and region of origin as confounding factors.
JAK/STAT signaling and MHC class I pathway component expression in SSA. Representative IHC stainings of breast cancer samples with MHC class I HC (A and D), β2-m (B and E), and the APM component tpn (C and F) are shown with no or weak and strong staining patterns, respectively. A lower expression of HC (H), β2-m (I), and tpn (J) was found in WA, while no differences can be demonstrated in the respective intrinsic subtypes (G). K, The prognostic influence of MHC class I expression is representatively shown for HC. Kaplan–Meier curve illustrates poor survival in patients with low HC expression in SSA (P = 0.006, log-rank test). Activation of IFNγ signaling via phosphorylated pSTAT1 is shown with representative low, intermediate, and high nuclear expression analyzed by IHC (L and M). All P values were estimated in multivariate analyses and are adjusted for differences in age, grading, quality score, stage, IHC subtype, and region of origin as confounding factors.
Discussion
Genetic, immunologic, environmental, and socioeconomic factors are known to influence breast cancer pathophysiology and disease progression (12, 21, 34). So far, regional differences in breast cancer biology have been proposed to be due to the ethnicity-dependent distribution of intrinsic breast cancer subtypes (9, 15, 16, 34–36). Studies on breast cancer in women of African ancestry, mainly performed in the United States, suggest more aggressive breast cancers with a higher proportion of TNBC linked with a poor outcome when compared with patients with breast cancer of European ancestry (10, 11, 23, 24, 37, 38). It is noteworthy that African Americans do not represent the regional diversity in SSA, because the vast majority of this population has its ancestry in Western, Central-Western, and South-Western SSA (39, 40). However, a study comparing Tanzanian, African American, and European American patients with breast cancer also demonstrated higher proportions of more aggressive intrinsic breast cancer subtypes in EA (41). Although The Cancer Genome Atlas database has provided information about the genomic landscape of breast cancer, including TNBC, these databases mainly comprise breast cancer specimens from women of European ancestry and only a low number of breast cancer cases of African ancestry (42). Recent results based on the Nigerian Breast Cancer Study of TNBC using whole-exome sequencing demonstrated twice as many TP53 mutations, a higher frequency of GATA3 mutations and a lower frequency of PIK3CA mutations when compared with Caucasians (9, 43–45). In addition, aberrant regulation of DNA damage repair genes was shown in breast cancer of women of African ancestry (38), which can lead to the activation of the G2–M checkpoint (46). Next to regional differences in the mutational landscape in SSA, an altered TME composition was previously reported in U.S. patients with breast cancer depending on their ethicity, with only subtle differences in the amounts of TILs, but increased frequencies in immune suppressive cell subpopulations, such as regulatory T cells (Treg) and M2 macrophages, have been suggested (24, 47). However, to the best of our knowledge, no data exist regarding regional TME diversity and immune escape mechanisms in SSA.
In this study, regional TME differences and their impact on the survival of patients with breast cancer from 10 SSA countries were analyzed by determination of the number of TILs as well as the composition and function of the immune cell repertoire. Because tissue quality and its variation across centres is a potential biasing factor, a scoring system was established to quantify tissue quality to include it in the modeling and thus prevent false-negative results.
While the TIL counts in breast cancer were comparable within the SSA regions, their mean values were higher in all SSA regions when compared with breast cancer samples from GER. This is in line with studies from the United States showing a higher TIL frequency in African Americans than in European Americans (48). In contrast, analysis of the immune cell repertoire demonstrated regional diversity, with higher numbers of CD3+ T cells and CD163+ M2 macrophages in WA. These results are in line with data on breast cancer in African American women, reporting higher levels of these immune cell subsets (23, 24), in which the majority of patients probably have WA ancestry (39, 40). Moreover, an ancestry-associated gene expression profile of TNBC in women of African descent with significantly higher levels of CD3+ T cells in African Americans and patients from Ghana when compared with European Americans and Ethiopian patients was demonstrated (49). This study confirmed previous reports that increased numbers of TILs are a favorable prognostic factor in breast cancer (27, 50). However, TILs showed a more pronounced prognostic value in ER− breast cancer in the current study like in previous studies as well (51), with differences in their prognostic significance between SSA regions. A higher predictive value of TIL frequencies was shown in EA, whereas analysis of TIL frequencies in WA showed no significant influence on survival. Findings from two prospective clinical trials in the United States also revealed that African-American patients had higher proportions of high TILs and no influence of increased TILs on disease-free-survival (52, 53).
MSI revealed a higher proximity of CD3+CD8+ T cells and cancer cells in WA samples, with a mean distance of 9.4 μm compared with a distance of 14.2 μm in the other regions, suggesting a stronger interaction between tumor and immune cells in WA samples. The prognostic impact of the spatial distribution of immune to tumor cells is confirmed with a radius of 10 μm, while others implicate 30 μm as biologically relevant (54–56). Although a positive prognostic value is generally attributed to a high proportion of CD3+CD8+ T cells (57), the highest mean frequency of this immune cell subpopulation in SSA was detectable in WA breast cancer cases despite their unfavorable outcome. Interestingly, the frequency of CD3+CD8+ T cells in WA was comparable with the mean frequency of samples from GER. The unfavorable outcome in WA breast cancer might be explained by T-cell exhaustion or dysfunction, known to be more pronounced in African Americans (58). Functional status of CD8+ T cells in this study demonstrated lower expression of cytotoxic markers, but the T-cell exhaustion marker PD-1 was not altered when compared with that of GER breast cancer samples. In addition, high levels of IL10 and reduced IL2, IL7, and IL15 are known to be involved in impaired T-cell proliferation (59), which might be in line with the highest frequency of immunosuppressive M2-type CD163+ macrophages, which secrete IL10 (60), found among WA breast cancer samples. In addition, IFNγ produced by T and natural killer cells (43) exhibited higher expression levels in WA breast cancer samples. This was positively correlated with the expression of PD-L1 by TILs and tumor cells and MHC class I antigens by tumor cells known to be upregulated by IFNγ (61). A higher PD-L1 surface expression in TILs, which routinely has to be analyzed prior to immunotherapy in breast cancer (62), was further associated with improved survival in the Her2+ breast cancer subtype and TNBC. It is also noteworthy that minor regional differences, including slightly lower PD-L1 expression in TILs, were detected in breast cancer samples from WA, whereas a significant downregulation of MHC class I APM components was demonstrated in WA compared with that of other SSA regions. In particular, the MHC class I HC and tpn showed lower expression levels in breast cancer samples from WA when compared with all other regions in SSA and with GER breast cancer samples, which reflect a more pronounced MHC class I–mediated immune escape of breast cancer in the WA region (18, 19, 63–67).
Limitations
It is noteworthy that SSA regions are intrinsically variable in terms of genetic, environmental, lifestyle, and socioeconomic parameters and other factors, which were not available for our cohort. Furthermore, several indicators suggest that the samples included for each region were not unbiased, for example, among patients with a known tumor stage, the higher proportion of early-stage tumors in EA than in SA would not be expected at the population level. Therefore, adjustment for stage was included in all multivariable analysis. Moreover, quality issues with the available tissues are a significant problem in SSA countries. In addition, due to lack of adequate material in our study, it was not possible to perform genetic analysis to assess regional differences. Therefore, prospective studies are urgently needed to overcome these limitations and confirm the results obtained.
Conclusion
In addition to a higher proportion of ER− and biologically more aggressive tumors in WA (13) when compared with other SSA regions as well as European patients, the TME composition and expression of immune modulator molecules in WA show an unfavorable pattern. This possibly contributes to the particularly low survival rates of patients with breast cancer in WA. Thus, immune profiling of breast cancer in SSA underlines the clinical relevance of the TME and immune escape strategies linked to breast cancer patient outcomes. Considerable diversity is seen regarding these features between SSA regions. However, beyond the biological relevance shown in this study, further investigations are urgently required to identify the underlying immune regulatory mechanisms in patients with breast cancer of WA origin. This study highlights the relevance and importance of multicentre studies to analyze regional diversities in more detail, which might shape the design of clinical trials and therapeutic options.
Authors' Disclosures
T. Abebe reports grants from Susan G. Komen during the conduct of the study. No disclosures were reported by the other authors.
Authors' Contributions
M. Bauer: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing–original draft. M. Vetter: Writing–original draft, project administration, writing–review and editing. K. Stückrath: Data curation. M. Yohannes: Sample collection, clinical data collection. Z. Desalegn: sample collection. T. Yalew: Data curation, sample collection. Y. Bekuretsion: Data curation, sample collection. T.W. Kenea: Data curation, sample collection. M. Joffe: Conceptualization, writing–review and editing. E.J. Van Den Berg: Data curation, sample collection. J.I. Nikulu: Data curation, sample collection. K. Bakarou: Data curation, sample collection. S.S. Manraj: Data curation, sample collection. O.J. Ogunbiyi: Data curation, writing–review and editing, sample collection. I.-O. Ekanem: Data curation, sample collection. F. Igbinoba: Data curation, sample collection. M. Diomande: Data curation, sample collection. C. Adebamowo: Data curation, writing–review and editing. C.P. Dzamalala: Data curation, sample collection. A.A. Anele: Data curation, writing–review and editing, sample collection. A. Zietsman: Data curation, sample collection. M. Galukande: Data curation, sample collection. M. Foerster: Data curation, writing–review and editing, sample collection. I. dos-Santos-Silva: Data curation, writing–review and editing. B. Liu: Resources, data curation. P. Santos: Software, formal analysis, visualization, writing–review and editing. A. Jemal: Supervision, writing–review and editing. T. Abebe: Writing–review and editing. C. Wickenhauser: Resources, methodology, writing–review and editing. B. Seliger: Conceptualization, supervision, funding acquisition, writing–original draft, writing–review and editing. V. McCormack: Resources, writing–original draft, project administration, writing–review and editing. E.J. Kantelhardt: Conceptualization, resources, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.
Acknowledgments
This work was supported by grants from the Else-Kröner-Foundation (E.J. Kantelhardt, 2018_HA31SP), Susan G. Komen-Foundation (E.J. Kantelhardt, GTDR16378013, VM, IIR13264158), German Cancer Aid (B. Seliger, Integrate-TN, 70113450), the German Research Council (B. Seliger, Se581/33-1), German Academic Exchange Service (E.J. Kantelhardt, ID57216764), Ministry for Economic Cooperation and Development, and the Else-Kröner-Fresenius Foundation (E.J. Kantelhardt, ID81256434), Hoffmann-La Roche Ltd (E.J. Kantelhardt, 27.5.2014), and the NCI (V. McCormack, R01CA244559).
We want to thank all patients who provided tumor samples and pathology staff at the sites in SSA for providing the tissue material. We thank Maria Heise for excellent secretarial help. We thank Christine Fathke (MD) and Linda Dießel (MD) for support in rare histomorphological diagnoses and Sarah Voigtländer and Andreas Wilfer for excellent laboratory organization. Where authors are identified as personnel of the International Agency for Research on Cancer or WHO, the authors alone are responsible for the views expressed in this article, which do not necessarily represent the decisions, policies, or views of the International Agency for Research on Cancer or WHO.
The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.
Note: Supplementary data for this article are available at Cancer Immunology Research Online (http://cancerimmunolres.aacrjournals.org/).
References
Supplementary data
Supplementary Table 1
Supplementary Table 2
Supplementary Table 3