Background
Tumor infiltrating lymphocytes (TILs) have been associated with improved response to chemotherapy and better overall survival in breast cancer [
1‐
4], and conversely, there is evidence that suppressive immune cells facilitate tumor evasion of the host immune system [
5]. More specifically, the presence of CD8+ T cells is associated with improved long-term survival in both HER2+ and triple-negative breast cancers [
6], and there is evidence that chemotherapeutic agents promote CD8+ T cell-mediated immunogenic tumor cell killing [
7]. Evidence suggests that tumor-specific factors, such as hormone receptor (HR) status and molecular subtype, are key associations with immune infiltration in breast tumors [
6]. However, these factors do not fully describe the observed variation in the tumor immune microenvironment of breast cancers and there is growing evidence that patient-level modifiable factors may also play a role.
Recent studies show that body mass index (BMI) is significantly correlated with systemic inflammation [
8] and may influence inflammation within the breast microenvironment. BMI has been shown to be positively correlated with breast tumor and normal tissue inflammation in patients with ER-negative breast cancer [
9]. Interestingly, overweight or obese patients may have improved responses to immunotherapy in other cancers [
10], which could indicate a distinct tumor immune microenvironment.
There is also evidence that the inflammatory potential of an individual’s diet may influence breast cancer risk and outcomes, based on analyses of the Women’s Health Initiative cohort [
11,
12]. These studies utilized a literature-derived nutrient-based dietary inflammatory index to assess the inflammatory potential of the diet. A related inflammatory index, the empirical dietary inflammatory pattern (EDIP), is a food-based score of dietary inflammatory potential [
13,
14] that has been shown to be associated with the immune microenvironment in colorectal cancer. Specifically, pro-inflammatory diet (higher EDIP score) was associated with colorectal cancers low or no intratumoral periglandular reaction (fewer/no immune cells), but not with cancers that had intermediate or high peritumoral lymphocytic reaction [
15].
Based on these data, we hypothesized that patient factors including adiposity (BMI change from age 18), physical activity, and dietary inflammatory potential (EDIP score) may play an important role in the breast cancer tumor immune microenvironment. In the current study, we aimed to develop immune cell-specific expression signatures based on immunohistochemistry (IHC) data, and then, we evaluated associations of BMI change since age 18, physical activity, and EDIP score with the expression signatures.
Discussion
There is a growing understanding of the critical role of the breast tumor immune microenvironment in treatment response and outcomes in breast cancer, yet modifiable patient factors that impact immune infiltration in breast cancer remain poorly understood. The Nurses’ Health Study offers a large-scale, unique cohort to evaluate interactions between patient lifestyle exposures and the immune microenvironment. Specifically, this study is unique in its integration of multiple immune cell-specific immunohistochemistry (CD8, CD4, CD20, and CD163) with published immune gene expression signatures, including the IRIS gene sets, which we recently demonstrated can effectively interrogate the breast cancer tumor immune microenvironment in a large clinical trial [
33]. Using paired IHC and gene expression data from over 250 unique patient tumors for each marker, we successfully derived improved expression-based predictors of immune cell subset infiltration and then evaluated the association with patient factors: BMI change from age 18, physical activity, and EDIP.
Higher BMI change since age 18 demonstrated the strongest association between immune cell-specific expression scores and patient factors, specifically GSAct, CD4 T cell score, and CD163 macrophage score. BMI change in adulthood has been associated with breast cancer risk [
34]. The role of obesity in the tumor immune microenvironment is of great interest, driven initially through evidence in melanoma that obesity was associated with longer survival in patients receiving immunotherapy [
10]. Mechanistic work suggested that, paradoxically, obesity results in increased immune aging, tumor progression and PD-1-mediated T cell dysfunction yet increased efficacy of PD-1/PD-L1 blockade in murine models and patients with cancer [
35]. Several recent papers investigated the association of TILs and BMI. In one, high stromal TILs were associated with increased event-free survival in lean (hazard ratio [HR] = 0.22, 95% CI = 0.08–0.62;
P = 0.004) but not in heavier patients [
36]. A second showed that TIL density was significantly lower in obese than in normal weight and overweight patients [
37]. Importantly, neither of these studies evaluated specific immune cell subsets as TILs encompass diverse pro- and anti-tumor immune cells.
While most FDA-approved immunotherapies target CD8+ cytotoxic T cells, the role of CD4+ helper T cells [
38,
39] and CD163+ macrophages [
40‐
42] is less well defined but increasingly seen as key players in the breast cancer microenvironment. It is established that T‐lymphocyte populations change with obesity (reviewed in [
43]). In our model, for each 10 unit (kg/m
2) increase in BMI—roughly equivalent to an individual going from BMI 20 (normal weight) to BMI 30 (obese), the percentage of cells positive for CD4 and CD163 increased 1.6% and 1.4%, respectively. It remains unclear whether this percent change could explain differences in response to therapy. In obesity, there is evidence that interferon gamma-producing pro-inflammatory CD4‐positive Th1 cells are increased, whereas anti‐inflammatory CD4‐positive Th2 and Treg cells are reduced [
43]. Intriguingly, we have previously shown in the Nurses’ Health Study that higher BMI was associated with increased expression of genes associated with IFN alpha and gamma response in ER- tumor and ER- tumor-adjacent tissues [
9]. In addition, in multivariable models of CD8, CD4, CD163, and GeparSixto scores, both HER2 and TNBC status significantly contributed to each model reinforcing the importance of investigating TME metrics within specific breast cancer subtypes.
Inflammatory diet was positively associated with GSAct in bivariable analyses with a trend in multivariable analyses, suggesting that a more pro-inflammatory diet is associated with higher immunity. This differs from our hypothesis, which was based on data in colorectal cancer, where inflammatory diet was associated with a higher risk of developing colorectal cancer
only among tumors that had low tumor infiltrating lymphocytes [
15]. It is likely that the distinct settings, for example, direct exposure of colonic mucosa to dietary elements versus no exposure in breast tissue, may influence these effects. To our knowledge, this is the first study to examine the relationship between inflammatory diet and quantity of TILs and immune subsets in breast tumors.
We also did not find a correlation between physical activity and immune cell-specific expression scores, specifically not GSAct, CD8, or CD4, which had either bivariable or multivariable association with other patient factors. Physical activity has been hypothesized to be associated with biomarkers of inflammation in breast cancer survivors [
44]. In animal breast cancer models, effect of physical activity on the amount of TILs is conflicting [
45,
46]. In humans, exercise was not found to affect levels of circulating T cells in patients receiving chemotherapy for breast cancer [
47]. To our knowledge, this is the first study exploring the effect of physical activity on multiple immune cell subset infiltration in breast tumor tissue and further delineating activity versus inactivity and more granular evaluation of types of physical activity may provide additional insights. While physical activity was not associated with immune cell infiltration in this study, overall, patients with breast cancer have been shown to benefit from exercise during and after cancer-directed therapy, and this study should not be used to justify a sedentary lifestyle for these patients [
48,
49].
In this cohort of over 600 tumor–normal pairs and an independent validation cohort (TCGA), both GSAct and
CD8A expression metrics show modest—but consistent—correlation between breast cancer and tumor-adjacent normal breast tissue. This suggests that the adjacent normal breast may reflect an altered immune microenvironment in the context of breast cancer. In a smaller cohort, inflammation expression was elevated in adjacent normal tissue relative to reduction mammoplasty tissues [
50]. In the NHS, elevated inflammatory expression in adjacent normal breast tissues was associated with higher BMI and alcohol consumption specifically in ER- tumors [
9,
22]. It is possible that tumor-adjacent normal inflammatory gene expression is a “bystander” effect in response to the tumor. Additional work on the immune microenvironment of tumor-adjacent breast is warranted to understand if adjacent normal immune infiltration is associated with breast carcinogenesis, immune infiltration of established tumors, or therapy response.
This study does have limitations. Only a subset of all the subjects enrolled in NHS/NHSII were included in the TMA due to limited tumor tissue availability. However, the characteristics of participants included in the TMA were very similar to those of all the eligible cases, including BMI, physical activity, EDIP score, and other breast cancer risk factors (e.g., first-degree family history and parity). Adiposity has differential associations with breast cancer risk based on menopausal status, and the models were derived primarily in NHSII subjects and applied primarily in NHS subjects; importantly, menopausal status and NHS cohort were covariates included in multivariable models. We acknowledge that TILs in breast cancer are typically characterized using a standard approach based on the International TILs Working Group [
51]; however, given the computer-based quantification approach and diversity around stromal versus epithelial correlation across immune markers, we used total positive cells. In addition, the correlation between novel expression signatures and infiltrating immune cell number was modest, though significantly outperformed multiple established immune cell-specific signatures. We hypothesize that this reflects the fact that bulk transcriptome signatures may not be an optimal way to represent discrete infiltrating immune cells numbers and supports further work on single cell-based technologies such as single cell RNA sequencing and highly multiplexed immunofluorescence profiling.
Future studies of the association of patient factors, specifically BMI and EDIP, and patient outcomes are warranted. It would be interesting to investigate the effect of modification of BMI on patient outcomes with different systemic treatments. Associations between BMI, physical activity, and diet with other immune cell subsets should also be investigated in order to understand the complex relationships between immune cell infiltration and modifiable patient factors. Further, in future studies, the association between immune infiltration signatures and patient prognosis will be assessed, but this requires careful analysis beyond the scope of the present study. In addition, to improve gene expression signatures of specific immune subsets we plan to utilize more complex modeling, such as machine learning; however, a strength of our linear model-based score approach is the ability to identify and quantify contribution of individual components.
Acknowledgements
We would like to thank the participants and staff of the NHS/NHSII cohorts for their valuable contributions as well as the following state cancer registries for their help: AL, AZ, AR, CA, CO, CT, DE, FL, GA, ID, IL, IN, IA, KY, LA, ME, MD, MA, MI, NE, NH, NJ, NY, NC, ND, OH, OK, OR, PA, RI, SC, TN, TX, VA, WA, WY. The authors would like to acknowledge Catherine Carson, Celia Garr, Katherine Weber, and Kathy Hauck for clinical support. The authors assume full responsibility for analyses and interpretation of these data.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.