Background
Outcome for patients with human epidermal growth factor receptor 2 (HER2)-positive metastatic breast cancer (MBC) has dramatically improved since the introduction of trastuzumab. More recently, the addition of pertuzumab, ado-trastuzumab emtansine (TDM-1), trastuzumab-deruxtecan, tucatinib, neratinib, and margetuximab have further increased outcome in these patients [
1‐
6]. Follow-up of the CLEOPATRA study showed that a small group of patients experience long-term progression-free survival [
1]. We and others have previously shown in real-world cohorts that achieving radiological complete remission (rCR) is strongly associated with improved overall survival (OS) in patients with HER2-positive MBC [
7,
8]. Strikingly, survival ranges from a few months to several years and sometimes even decades. Insight into tumor characteristics and the immune microenvironment of primary and metastatic tumor samples from patients with MBC could provide insight into the underlying processes of such variable outcomes and potentially aid in personalization of treatment and ultimately long-term survival for more patients [
9].
Within the tumor immune microenvironment both protumor and antitumor cells play a role, such as protumor neutrophils promoting metastases or regulatory T-cells allowing tumor proliferation and on the other side CD8+ T-cells and natural killer (NK) cells that elicit favorable anti-tumor immune responses [
10]. The tumor immune microenvironment may be of particular importance in HER2-positive breast cancer as HER2 is a natural antigen and the response to HER2-targeted therapies is partly based on both the innate immune system via antibody-dependent cell cytotoxicity (ADCC) as the adaptive immune system, by means of NK-cell activation [
11,
12]. Preclinical studies have shown that activation of the immune system, in particular NK-cells, is necessary for trastuzumab efficacy [
13,
14]. Therefore the cellular composition of the tumor immune microenvironment may be associated with outcome in HER2-positive MBC, as has been shown for triple-negative MBC [
5].
In patients with HER2-positive MBC, the prognostic value of stromal tumor infiltrating lymphocytes (sTILs) showed conflicting results in retrospective analyses of several studies and a retrospective series of patients mainly evaluating sTILs in quantitative manner [
15‐
18]. In-depth characterization of immune cells may have key prognostic value and increase our understanding of the interaction between tumor and microenvironment [
19]. It may also provide clues for development of immune modulating agents that can be combined with anti-HER2 treatment. In patients with early breast cancer, multi-omics features and single-cell pathology data were highly correlated with outcome [
20,
21].
In this exploratory analysis we use a combination of next-generation RNA sequencing, multiplex immunofluorescence (mIF) multispectral analysis for spatial composition evaluation, and immunohistochemistry (IHC) to characterize the tumor immune microenvironment of 108 pre-treatment primary tumor samples and 39 samples of metastases of 124 patients with HER2-positive MBC, including 15 paired samples. Next, we evaluate if immune traits associate with rCR and survival in a real-world cohort of patients with long-term follow up.
Discussion
We characterized the tumor immune microenvironment in 147 samples of 124 real-world patients with HER2-positive MBC and in exploratory analysis evaluated whether the infiltration of specific immune cells was associated with rCR and with OS. In our study with long-term follow-up of more than 9 years, we have shown that the infiltration of B-cells, regulatory T-cells, and exhausted T-cells are associated with unfavorable outcome. Additionally, the extracellular matrix signature ECM1 (high vascularization in the ECM), is associated with more infiltration of sTILs and negatively associated with OS. Last, more PDL1-positive cells in metastases was associated with higher likelihood of rCR and trended toward better OS. Below we will discuss our most important findings on infiltration of specific immune cells and how these insights could serve the evaluation and development of novel immunomodulatory strategies.
In our cohort of patients with HER2-positive MBC, the presence of overall sTILs was not associated with OS nor with rCR, overall nor in ER-positive versus ER-negative subgroups. The lack of significant association of sTILs with outcome in the metastatic setting is in line with other retrospective cohort analyses and a post hoc analysis of the MA.31 study [
16,
18,
42]. However, post hoc analyses of the CLEOPATRA and PANACEA study showed a positive association of sTILs and OS [
15,
43]. In the CLEOPATRA study, all patients received dual HER2-blockade, which is shown to have a synergistic effect on increasing NK-cell migration [
15,
44]. The PANACEA study combined trastuzumab with PD1-inhibitor pembrolizumab and showed a better response rate in a subgroup of patients with metastatic samples that harbored at least 5% sTILs. This could suggest that patients with high sTILs in HER2-positive MBC might have unfavorable outcome with conventional anti-HER therapy but may benefit from a combination of anti-HER2 therapy with immune checkpoint inhibitors that can activate cytotoxic T-cells and enhance the anti-tumor immune response [
43].
Typical quantitative evaluations of sTILs do not capture different composite immune cell populations that may specifically influence the pro-tumor anti-tumor equilibrium in the tumor microenvironment [
10]. For instance, Tregs are a key regulator of the T-cell response. We found a negative association with outcome for Tregs. This finding is in line with several meta-analyses that showed a poor prognosis with high infiltration of Tregs in early breast cancer [
45]. It has also been shown that infiltration of Tregs correlates with poor prognostic factors such as ER-negativity, HER2-positivity, lymph node metastasis and high histological grade in early breast cancer [
46]. Studies evaluating infiltration of regulatory T-cells in MBC are sparse but confirm a negative association with survival [
18]. On a positive note, a recent study in ER-positive metastatic breast cancer showed effective reduction in activated Tregs in the microenvironment after treatment with tamoxifen, pembrolizumab and vorinostat (a histone deacetylase inhibitor), presumably reflecting a remodeling of the tumor microenvironment toward anti-tumor immunity [
47].
Recently Sobral-Leite and colleagues showed that infiltration of Tregs is associated with downstream activation of the PI3K pathway [
48], a known resistance mechanism in HER2-positive and ER-positive breast cancer. This finding could explain the strong negative association with survival we found in this HER2-positive MBC cohort and provides rationale for evaluating therapies that combine HER2-targeted, PI3K inhibition and Treg-targeted agents. Drugs targeting the PI3K pathway are combined with anti-PD(L)1-targeting drugs in three phase 1 studies for patients with solid tumors with an overactivated PI3K pathway (NCT03673787, NCT03257722, NCT04317105).
Another recent study in breast cancer indicated that the number of Tregs is closely correlated with that of (IL10+) Bregs in TIL aggregates in marginal regions of tumors [
49]. Similarly to Tregs, Bregs are negative regulators of anti-tumor immune response and associated with progression of several cancers, including breast cancer [
49]. Moreover, B-cells can induce transformation of CD4+ T-cells to Tregs via TGF-β and IL10 and direct contact via the PD1-PDL1 axis in mice models [
50,
51]. We did not characterize B-cell subpopulations, therefore we do not know whether Bregs or active B-cells determine the negative association with OS and rCR. Nevertheless, we report a consistent negative impact on outcomes with the infiltration of B-cells. In our cohort less than 1% of patients had a tertiary lymphoid structure (TLS), the prognostic value of TLS was therefore not included in our analyses. We did analyze the spatial colocalization of T-cells and B-cells, which was not associated with outcome (Additional file
4: Results Table). To the best of our knowledge, no study reported a specific role of B-cells in MBC. In early breast cancer, however, infiltration of B-cells is associated with high pathological complete response rates in in several studies [
52,
53]. This observation might indicate plasticity of B-cells and a more immune suppressive role of B-cells in metastatic cancer, which deserves further study and possibly evaluation of targeting Bregs in MBC [
54].
We also found a negative association between the
exhausted T-cell signature in primary tumors and outcome. Exhaustion can be seen as a self-preserving transient state of T-cells, induced under chronic stimulation of antigens. Exhausted T-cells can also be recognized by high expression of CTLA-4, LAG-3, PD-1, and TIM-3 [
55]. Therefore presence of exhausted T-cells is associated with benefit from PD1- or PDL1-targeted therapy. In our cohort, none of the patients received such therapy, probably explaining the unfavorable association of the exhausted T-cell signature with outcome. PD1 and PDL1 expression in the primary tumor was not associated with outcome in our cohort. Strikingly, we noticed a higher chance of achieving rCR in samples when PD1-positive T-cells were close to PDL1-positive stromal cells or tumor cells in metastatic samples. In general, PDL1 expression is associated with decreased survival in early breast cancer [
56]. However, inhibition of PD1/PDL1, as well as other checkpoints expressed by exhausted T-cells, shows high efficacy in reversing the exhaustion and stimulating anti-tumor activity of T-cells [
55]. Among patients with HER2-positive MBC, the PANACEA study showed an objective response in 7 of 46 (15%) of the patients with PDL1 expression. In patients without expression of PDL1 no responses were seen [
43]. The KATE2 study demonstrated no clinically significant PFS benefit with the addition of atezolizumab (anti-PDL1) to TDM1 (versus placebo + TDM1) in patients not selected for PDL1 expression [
57]. Similarly, no significant clinical activity was seen in heavily pre-treated HER2-positive MBC who received durvalumab (anti-PDL1) and trastuzumab in a phase 1 study [
58]. Several phase 2 and 3 studies are recruiting patients with HER-positive MBC and will evaluate a combination of trastuzumab with or without pertuzumab with an anti-PD(L)1 inhibitor, summarized by Griguolo [
59] and Costa [
60]. To the best of our knowledge, no studies are yet evaluating a combination of HER2-targeted therapy and anti-lymphocyte-activation gene 3 (LAG3) or anti-T cell immunoglobulin and mucin domain-containing protein (TIM3) (and anti-PD(L)1), which could also be an appealing strategy to overcome T-cell exhaustion.
Novel immunomodulatory strategies for patients with HER2-positive MBC might focus on NK-cells. In our study, presence of NK-cells in the primary tumor was not associated with outcome, only NK-cells in metastases showed a trend toward more rCR (OR > 10,
P = 0.054, Additional file
4). This could suggest that there were not enough NK-cells for a meaningful benefit or NK-cells are inhibited by other infiltrating cells [
12]. NK-cells can be inhibited via stimulation of the NK inhibitory receptor (NK group 2 member A [NKG2A]) in the context of MHC class I [
11]. Monalizumab, which targets NKG2A, has shown benefit in phase 1 and 2 studies in gynecological malignancies [
61] and head and neck cancer [
11], respectively. Monalizumab in combination with trastuzumab is now being evaluated in the phase 2 MIMOSA study in patients with HER2-positive MBC (NCT04307329).
In general, metastatic samples had significantly lower infiltration of sTILs, especially CD8+ T-cells compared to the primary tumor. Additionally, spatial analyses from the mIF data showed less colocalization of tumor cells and immune cells in the metastases compared to primary tumors. These results are in line with several studies, that have concordantly shown that metastases harbor fewer sTILs compared to their corresponding primary tumors [
43,
62,
63]. On top of that, the site of metastases influences the number of sTILs, with lung metastases harboring more sTILs than bone or liver metastases [
18,
43]. It should also be noted that sTILs are dynamic and subject to host factors and changes in the tumor as well as treatment.
We used the Morisita-Horn index to evaluate colocalization of immune cells and tumor cells and between different immune cells. Rather than evaluating the nearest neighbor, the Morisita-Horn index evaluates segregation of cells within a neighborhood [
39]. We therefore believe this index provides a better representation of the tumor microenvironment. To strengthen our findings, we also used the Spatial Proximity Score ([SPS]; Additional file
2: Fig. S3B) to evaluate impact of nearest neighbor within the tumor immune microenvironment (Additional file
4: Results Table). The associations with outcome of colocalization of Tregs to other T-cells, PD1-positive T-cells to PDL1-positive stroma and tumor was similar using the SPS or the Morisita-Horn index. The Ecoscore, which classifies the tumor microenvironment as more pro-tumor or more anti-tumor (Additional file
2: Fig. S3C) [
64], showed no association with outcome in our cohort (Additional file
4: Results Table).
To the best of our knowledge, our study is the largest real-world cohort of patients with HER2-positive MBC in which the tumor immune microenvironment is evaluated in detail. Another strength of our study is the high correlation of the results over the different technologies that we used to evaluate the tumor immune microenvironment. Nevertheless, using a real-world cohort comes with some limitations. First, mIF and IHC biomarkers were assessed using TMAs which may overestimate the expression of biomarkers compared to whole slides, although the TMA results and gene expression analyses largely corresponded [
65]. Second, availability and evaluability of tissue created some limitation for thorough evaluation, including a selection of relatively easier sites for biopsies, excluding bone biopsies, which are more common in ER-positive tumors, relatively few (
n = 26) matched primary tumors and metastatic samples, and some metastatic samples (
n = 7, 18%) were taken after treatment, which could have induced a bias in less sTILs and maybe proportionally less samples of patients who achieved rCR. Third, not all patients with metachronous MBC (i.e., metastases developed after treatment for the primary tumor), received trastuzumab as neoadjuvant and/or adjuvant therapy and none of the patients received pertuzumab which is now the recommended first line of treatment for HER2-positive MBC. In addition, 23 patients with oligometastases received local treatment for their metastases as well as systemic therapy. The number of patients in subsets is small, which limits our ability to draw definite conclusions in HER2-positive/ER-positive and HER2-positive/ER-negative subgroups. We could not perform separate analyses in the synchronous subgroup, the subgroup with oligometastases or according to different treatment schedules. Despite these limitations, we provide insight in the tumor immune microenvironment that potentially can help designing treatment combinations for patients with HER2-positive MBC with a unique dataset containing long median follow-up of nine years. In addition, we contribute to the scientific community a resource collection of deep, multi-platform, multi-technology immune composition data for clinically well-annotated HER2-positive MBC. Future studies could elaborate on the results showed in this paper, for instance, in depth analysis of intracellular signaling that influences the function of immune cells could further help in designing treatment strategies.
In conclusion, we present insight into the tumor immune microenvironment and its association with outcome in a real-world cohort of patients with HER2-positive MBC. In our cohort, the infiltration of B-cells and regulatory T-cells in the primary tumor microenvironment are associated with unfavorable OS and lower probability of achieving radiological complete remission. These findings provide insight and rationale to further explore a combination of HER2-targeted therapy with targeted immune-modulating therapy to improve durable responses in more patients.
Acknowledgments
We would like to acknowledge the NKI-AVL Core Facility Molecular Pathology & Biobanking (CFMPB) for supplying NKI-AVL Biobank material and/or lab support, the Genomics Core Facility and the Tumor Registry from the Netherlands Cancer Institute, SlideScore, UCSF, Breast Cancer Research Foundation, Breast Cancer Research – Atwater Trust, Give Breast Cancer the Boot, the Helen Diller Family Comprehensive Cancer Center (HDFCCC) Biorepository and Tissue Core facility of University of California, San Francisco, Agendia N.V., the nationwide network and registry of histo- and cytopathology in the Netherlands (PALGA Foundation), Akoya Biotechnology, PerkinElmer, and BioRender. Special thanks go out to Annegien Broeks, Jan Hudecek, Ron Kerkhoven, Wim Brugman, Iris de Rink, Arno Velds, and Jennifer Bolen.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.