Introduction
Immunotherapy with immune checkpoint inhibitors is usually considered in advanced metastatic breast cancer. The intention is to reduce the tumor burden by restoring a durable anti-tumor immune response. As breast cancer is not a highly immunogenic disease in general, treatment efficacy seems to depend on the molecular breast cancer subtype and the expression of PD-L1 [
1,
2].
Monotherapy against programmed cell death protein-1 (PD-1) showed only the modest tumor and durable response rates in breast cancer (4–25%) [
2]. In the need of novel strategies, in vivo studies provided convincing evidence that a dual blockade against PD-1 and the novel immune checkpoint receptor lymphocyte-associated gene-3 (LAG-3) [
3] can result in tumor reduction and increase of survival [
4,
5] by restoring CD8
+ T cell function [
6]. In human tumor tissue, LAG-3 is co-expressed with PD-1 on activated but exhausted CD8
+ T cells [
7]. Particularly, highly immunogenic tumors susceptible to immunotherapy like melanoma, microsatellite instable colorectal cancer, or triple-negative breast cancer carry PD-1
+/LAG-3
+ CD8
+ tumor-infiltrating T cells [
8].
In breast cancer, the predictive value of LAG-3 expression remains still unclear. There is some prognostic evidence associating high LAG-3 expression with improved overall survival (OS) but uncertain significance with respect to disease-free survival (DFS) [
9]. Most ongoing clinical trials are investigating anti-LAG-3 drugs in combination with a dual blockade against PD-1 in solid tumors in advanced disease with promising survival benefits and long duration of response rates for those who profited [
5].
Interestingly, most of these trials are considering neither the amount nor the distribution of tumor-infiltrating CD8
+ T cells. For immunotherapy to be active though, CD8
+ T cells must be present within the tumor bed—referred to as inflamed or clinically “hot” tumors [
10‐
13]. Moreover, while the advanced metastatic disease is targeted, PD-1 and LAG-3 expression within distant metastatic breast cancer remains understudied. Their assessment within different anatomical sites of distant metastasis might however be crucial given that the composition of the tumor immune microenvironment is heterogeneous and is critically influenced by organ-specific parenchymal cells [
14‐
16].
To address this knowledge gap, we assessed the CD8+ T cell immune phenotype as well as PD-1 and LAG-3 expression in primary tumors (PBTs) with intrapatient matched distant metastases (METs) in a retrospective cohort of 95 breast cancer patients by using immunohistochemistry on whole sections. Metastases had occurred at either brain, bone, liver, or soft tissue.
Discussion
By investigating a large intrapatient matched PBT distant MET breast cancer cohort [
2,
17,
25‐
27], we have shown that (i) PD-1
+/LAG-3
+ is strongly associated with a “hot” immune phenotype and (ii) differs between METs and PBTs. As described, LAG-3 expression was only observed in PD-1-positive cases [
7] with an overall low frequency [
23].
For this study, we translated the intratumoral spatial distribution pattern of CD8
+ T cells into three defined immune phenotypes, intended to reflect the clinical terms “hot” or “cold” tumors. We based our evaluation on the following tumor compartments: tumor center comprising the intratumoral/tumor epithelial and stromal compartment and the invasive margin compartment. While these compartments are arbitrary to a certain extent since T cells are thought to freely move through tissue and are as such not stuck to a particular compartment as implied by a snap-shot-like impression in FFPE tissue, these three compartments are well-established and proposed as such by consensus agreements [
22].
As published previously [
17], we evaluated the intratumoral compartment of the “tumor center” [
22] compartment based on the notion that a direct contact between CD8
+ T cells and tumor cells must occur for CD8
+ T cell-mediated cytotoxicity. Within this concept, also metastatic biopsies containing the “tumor center” fulfill the necessary criteria to identify immune phenotypes as the value of the invasive margin—often lacking in metastatic biopsy material—becomes limited. Systematic studies assigning the spatial distribution of CD8
+ T cells to a certain immune phenotype and, moreover, validating the biological significance of these immune phenotypes to the response to immune checkpoint inhibition are lacking up to date. There is thus no consensus classification of immune phenotypes yet. While we are convinced that our proposed immune phenotype classification may serve as a valuable surrogate marker also applicable in metastatic tissue, our suggested immune phenotypes certainly need to be validated in a cohort that includes responders and non-responders to immune checkpoint inhibition.
In contrast to the previous results, neither the immune phenotype nor PD-1/LAG-3 expression was associated to the molecular breast cancer subtype [
28] possibly due to the overall small number of patients in our cohort.
Overall, METs turned “cold” suggesting reduced immunogenicity of METs in general [
25]. Interestingly, METs of “cold” PBTs always remained “cold” at their matched metastatic site, while “hot” PBTs either stayed “hot” or turned “cold” in their corresponding METs. These observations were independent of the breast cancer molecular subtype. While the underlying mechanisms remain unclear, our findings imply a tumor intrinsic immunogenicity and may explain the low response rates to immunotherapy in metastatic breast cancer [
2,
5].
From a clinical translational point of view, these results strongly favor the spatial assessment of CD8
+ T cells together with PD-1/LAG-3 within metastatic tissue if immune modulatory therapy is considered. In case of synchronous metastases at different anatomical locations, biopsy material of either all metastatic sites or one of clinically greatest importance, such as brain metastases, should be discussed. The value of a combined assessment of PD-L1 and tumor-infiltrating lymphocytes was recently proposed as a more comprehensive immuno-oncological biomarker in breast cancer [
29]. Whether our suggested evaluation of immune phenotypes together with PD-1/LAG-3 within metastatic tissue may serve as an even more comprehensive immuno-oncological biomarker needs further validation in larger and prospective cohorts.
Among the distant metastatic sites, brain and soft tissue METs displayed more prevalently an inflamed but exhausted immune phenotype. To distinguish site-specific immune changes [
14] from the molecular breast cancer subtype, we included additional brain metastases. Again PD-1
+/LAG-3
+ expression correlated to “hot” brain METs regardless of the molecular breast cancer subtype supporting our previous notion of a tumor intrinsic immunogenicity.
In our cohort, an inflamed PD-1
−/LAG-3
− immune phenotype in the PBT was associated with an improved DFS implying a negative DFS prognostic impact of PD-1/LAG-3 expression. While this certainly needs to be confirmed in a larger cohort, this adverse prognostic significance of PD-1/LAG-3 expression was not unexpected given their inhibitory effects on the immune response in general [
6]. Nevertheless, these observations were in contrast to a recent publication describing improved DFS in PD-1- and LAG-3-positive primary breast cancers [
23].
In recent reports, LAG-3 expression is associated with different DFS prognostic outcomes which may be due to a small number of LAG-3/PD-1 positive cases, heterogenous methods employed, and different LAG-3-positive cutoffs [
9]. While Burugu et al. [
23] used tissue microarrays (TMAs) and focused on intratumoral lymphocytes, the paper by Bottai et al. [
28] described data using whole slides and stromal lymphocytes. This illustrates the inconsistencies and limitations between methods and the tumor compartment evaluated. TMAs are usually constructed using only small tumor cores taken in regions with high tumor content and not selected based on the presence of abundant immune infiltration. TMAs may as such neither reflect the intratumoral heterogeneity nor give the complete picture of the presence of LAG-3-positive cells as the whole tissue sections we used. Furthermore, due to our interest in LAG-3 and PD-1 expression in metastatic tissue, our cohort is biased for patients with advanced metastatic disease thus differing from an average breast cancer cohort.
Our study fills an important knowledge gap in metastatic breast cancer in two main regards: (i) the immune phenotype and PD-1/LAG-3 expression within metastatic breast cancer are significantly different from the primary tumor and among anatomical metastatic sites and (ii) PD-1
+/LAG-3
+ expression is strongly associated with a “hot” immune phenotype. Taken together, not the primary tumor but metastatic breast cancer should be analyzed for the immune phenotype and PD-1/LAG-3 expression to reveal the metastasis-associated immune pathology. This dual evaluation in metastatic sites may eventually improve the stratification of advanced breast cancer patients for immunotherapy given that CD8
+ T cells must be present within the tumor bed for an effective immunotherapy response [
10].
Conclusions
In summary, LAG-3 was exclusively observed in PD-1+ cases with an overall low frequency. PD-1+/LAG-3+ expression was associated with a “hot” immune phenotype both in PBTs and METs regardless of the breast cancer molecular subtype. Disease-free survival was significantly improved in inflamed but PD-1−/LAG-3− PBTs. In our cohort, METs of “cold” PBTs always remained “cold” at their matched metastatic site. In contrast, “hot” PBTs either remained “hot” or turned “cold” in their corresponding METs. Among the anatomical sites of metastases, brain and soft tissue metastases were more commonly inflamed with signs of exhaustion.
Our study emphasizes the careful assessment of the immune phenotype and PD-1/LAG-3 expression in metastatic breast cancer tissue to overcome intrapatient tumor heterogeneity. Furthermore, analysis of metastatic breast cancer tissue may improve the stratification of advanced breast cancer patients for a dual anti-PD-1/anti-LAG-3 immunotherapy.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.