T cell Bispecific Antibodies in Node-Positive Breast Cancer: Novel Therapeutic Avenue for MHC class I Loss Variants

Abstract

Background

Tumor-infiltrating lymphocytes (TILs) represent a prognostic factor for survival in primary breast cancers (BC). Nonetheless, neoepitope load and TILs cytolytic activity are modest in BC, compromising the efficacy of immune-activating antibodies, which do not yet compete against immunogenic chemotherapy.

Patients and methods

We analyzed by functional flow cytometry the immune dynamics of primary and metastatic axillary nodes (mLN) in early breast cancers after exposure to T cell bispecific antibodies (TCB) bridging CD3ε and HER2 or CEACAM5, before and after chemotherapy. HLA class I loss was assessed by whole exome sequencing and immunohistochemistry. 100 primary BC, 64 surrounding “healthy tissue” (HT) and 24 mLN related-parameters were analyzed.

Results

HLA loss of heterozygosity was observed in early BC, at a clonal and subclonal level and was associated with regulatory T cells and Tim3 expression restraining the immuno-stimulatory effects of neoadjuvant chemotherapy. TCB bridging CD3ε and HER2 or CEACAM5 could bypass MHC class I loss, partially rescuing T cell functions in mLN.

Conclusion

TCB should be developed in BC to circumvent low MHC/peptide complexes.

Abbreviations

ATP

Adenosine Triphosphate

BC

Breast Cancer

CAFs

Cancer-Associated Fibroblasts

CAR T cells

Chimeric Antigen Receptor T cells

CEACAM5

Carcinoembryonic Antigen-Related Cell Adhesion Molecule 5

CEACAM5-T

CD3-CEACAM5 bispecific antibody

CGH

Comparative genomic hybridisation

CRT

Calreticulin

CTL

Cytotoxic T Lymphocytes

Ctrl

Control

CXCL10

C-X-C motif chemokine 10

DC

Dendritic Cells

DCIS

Ductal Carcinoma

EBC

Early Breast Cancer

EpCAM

Epithelial Cell Adhesion Molecule

ER

Estrogen Receptor

ERBB2

Erb-B2 Receptor Tyrosine Kinase 2

FAP

Fibroblast Activation Protein

Foxp3

Forkhead box protein P3

FPR1

Formyl Peptide Receptor 1

GrzB

Granzyme B

HER2

Human Epidermal growth factor Receptor 2

HER2-T

CD3-HER2 bispecific antibody

HER2-UNT

untargeted HER2-UNT Antibody

HLA

Human Leukocyte Antigen

HLA LOH

HLA loss of heterozygosity

HMGB1

High Mobility Group B1

HR

Hormone Receptor

HT

Healthy Tissues

ICD

Immunogenic Cell Death

IDC

invasive DC

IFN

Interferon

mAB

Monoclonal Antibodies

MAPK

Mitogen-Activated Protein Kinase

MHC

Major Histocompatibility Complex

mLN

Metastatic Lymph Nodes

NFATC1

Nuclear Factor of Activated T-Cells 1

NF-κB1

Nuclear Factor-Kappa B1

NK

Natural Killer

NR

Resistance

PD-1

Programmed cell Death-1

PD-L1

Programmed Death-Ligand 1

R

Responsiveness

RNASeq

RNA Sequencing

SEM

Standard Error of the Mean

SLN

Sentinel Lymph Node

TCB

T cell bispecific antibodies

TCGA

The Cancer Genome Atlas

TCR

T Cell Receptor

TILs

Tumor-Infiltrating Lymphocytes

Tim3

T-cell Immunoglobulin and Mucin-domain-3

TLR4

Toll-Like Receptor

TNBC

Triple Negative Breast Cancer

Introduction

Like colon carcinoma, breast cancers (BC) have been one of the first anatomopathological entities to be described as “immune infiltrated” but paradoxically turned out to be relatively “immunoresistant”. First, distinct molecular subtypes of BC are intrinsically immunogenic in that some immunological metrics (1) harbor independent prognostic and predictive values for overall or progression-free survival during therapies. Pathologic analysis can reveal not only the numbers of TILs, the presence of follicular CD4⁺ T cells but also the pattern of infiltration, including tertiary lymphoid organs, relevant mainly in triple negative breast cancer (TNBC) and human epidermal growth factor receptor 2 (HER2)⁺ BC (2). TILs density is also a predictor of pathological complete response following neoadjuvant chemotherapy (3). Secondly, conventional therapies against BC play a major role in activating the immune system by inducing immunogenic cell death (ICD) (4,5). Third, despite several tumor-associated antigens described as immunogenic in BC vaccination trials, they were not protective against disease progression (6)(7). Hence, in contrast to other malignancies, there have been few clinical trials using programmed death 1(PD-1) or - ligand-1 (PD-L1) monoclonal antibodies (mAb) which achieves objective response rates in the 12-19% range when administered alone (8), or combined with taxanes in first-line metastatic TNBC (9). Hence, the lack of a clinically relevant immuno-dominant T cell receptor (TCR) repertoire capable of expanding post-chemotherapy remains a challenging issue. The current strategies to circumvent this issue are chimeric antigen receptor redirected T cells and T cell bispecific antibodies (TCB). TCB redirect T cells to tumor cells, by connecting a T cell via CD3ε□to a tumor-specific antigen on the surface e.g. HER2 or CEACAM5 (10)(11). This induces the formation of a transient cytolytic synapse between the cytotoxic T cell and the tumor cell resulting in T-cell activation, proliferation, and serial lysis of tumor cells. TCB-mediated T-cell activation does not rely on the presence of MHC class I molecules and tumor-specific peptide antigens (10). We have previously described novel heterodimeric 2+1 T cell bispecific antibodies based on a human IgG1 backbone using an Fc-region without residual FcγR binding (12)(13). Here, we show that i) immunogenic chemotherapy failed to switch on effector T cells in metastatic lymph nodes (mLN) of BC where Tim3 and Treg dominated the phenotype, ii) high prevalence of HLA class I loss of heterozygosity in tumor cells reaching axillary LN in paired specimens could explain TILs inactivation, iii) HER2-TCB or CEACAM5-TCB reinstate TILs effector functions in mLN, circumventing MHC class I loss variants.

Materials and methods

Patients and cohorts

Study approval

Institutional review board approval was granted by the University Paris Saclay and Gustave Roussy/Kremlin Bicêtre for the prospective cohort which was previously described (SAFIR01 (13)). The human study protocols were in accordance with the Declaration of Helsinki principles, and all patients provided informed consent before enrollment in the study.

Prospective cohort of stage I/II BC patients. Tumor infiltrating cells were analyzed prior to chemotherapy (in N=38 patients) or after neoadjuvant chemotherapy (in N=97 patients). A total of 100 primary BC lesions (table S1 and S2), 64 paired “Healthy Tissue” (HT) and 24 dissociated metastatic lymph nodes were analyzed after surgery at diagnosis or post-chemotherapy.

Experimental Design

Details feature in Supplemental materials. Briefly, BC lumps, mLN and paired surrounding healthy tissues were dissociated and subjected to ex vivo assays as previously described (14). The ex vivo assay consisted in stimulating the immune cells from 14 primary BC and 21 dissociated mLN with mAbs (T cell bispecific antibodies (TCB) HER2-CD3, CEACAM5-CD3 or controls), anti-FAP-41BBL, and combinations) or cytokines (rIL-2) (Fig. S2) for 2-6 days in 48-well plates as described in the Fig. S2, followed by flow cytometric analysis of T and NK cells and cytokine monitoring in ELISA of culture supernatants. We arbitrarily defined “responding” lesions, those exhibiting > 1.5-fold change over two different controls (medium and untarget-anti-CD3) in 3 independent biological readouts (out of 6 readouts, GrzB or Ki67 positivity on CD4⁺, CD8⁺ T cells or NK cells).

Flow cytometric analyses and cytokine/chemokine monitoring

TILs from 77 primary BC and 15 mLN were stained with fluorochrome-coupled mAbs (detailed in table S3), incubated for 20 min at 4°C and washed with PBSX1. Cell samples were acquired on a Cyan ADP 9-color (Beckman Coulter) and BD FACS Canto II flow-cytometers with single-stained antibody-capturing beads used for compensation (Compbeads, BD Biosciences). Data were analyzed with Kaluza 1.3 (Beckman). Supernatants from cultured cells were monitored using commercial ELISA (BD Biosciences).

Staining of HER2 and HLA class I

Immunohistochemistry was performed using HLA-I, mouse anti-human HLA A, B and C mAb (MBL, clone EMR8-5, 0,5μg/ml, 1:750 dilution, on 3μm-thick sections of formalin-fixed, paraffin-embedded breast cancer tissue sections on a Leica Bond RX with their Polymer Refine Detection System. Also refer to Supplementary materials.

HLA class I genes scoring

A total of 695 metastatic BC from clinical trial SAFIR01 (13) and SAFIR02 from Gustave Roussy (France) and 1080 primary BC from TCGA and METABRIC were included in the analysis. Segments with copy number variants were detected by Comparative genomic hybridisation (CGH) array and the raw data were processed using a R package rCGH (15). Also refer to supplemental materials.

LOHHLA (Loss Of Heterozygosity in Human Leukocyte Antigen) methodology and algorithm

Copy number inference of HLA alleles was performed using LOHHLA (15) on whole exome sequencing data. For further details refer to Supplemental materials.

Statistics

Data analyses and graphical representations were performed with Prism 7 (GraphPad, San Diego, CA, USA) and R version 3.4. The nonparametric Mann-Whitney test was used for comparison between primary BC versus HT or mLN. Correlations between different parameters were assessed by nonparametric Spearman test. p-values are two-sided and 95% confidence intervals for the statistic of interest are reported.

Results

Metastatic lymph nodes differ from primary lesions in the dynamics of immune composition in BC

We investigated the composition of immune infiltrates in early breast cancers (EBC) amenable to surgical resection, either before chemotherapy (“adjuvant” setting) or after conventional anthracycline/taxane-based chemotherapy (“neoadjuvant” setting). Patients baseline characteristics in the adjuvant and neoadjuvant setting are detailed in Table S1 and Table S2, respectively. Flow cytometric analyses of freshly dissociated primary tumors at diagnosis were compared with surrounding “healthy” tissues (HT), revealing several cancer-associated immune hallmarks. Cancerous lesions were enriched in leukocytes (Fig. S1a) in an immunosuppressive contexture. Regulatory T cells defined by CD25^highCD127^low/CD4⁺ T cells (Fig. S1b, left panel) were accompanied with conventional T cells overexpressing PD-1/PD-L1 (Fig. S1c, left and middle panels) while non-T and non-NK cells (CD3^-CD56^-, mostly myeloid and B cells (14)) expressed higher levels of PD-L1 (Fig. S1c, right panel) in tumors. The immune profiles of hormone receptor (HR)⁺, HER2⁺, TNBC did not significantly differ in primary or metastatic lymph nodes (mLN) except for CD25^highCD127^low CD4⁺ Treg cells that preferentially accumulated in HER2⁺ and TNBC (Fig. S1b right panel, Table S4).

We next examined the immune profiles of tumors subjected to neoadjuvant chemotherapy, mostly represented by locally advanced BC. Surprisingly, the immune microenvironment of mLN appeared different from their primary lesion. First, the proportion of CD8⁺ TILs markedly increased with chemotherapy in primary lesions with a significant rise of the CD8/CD4 ratio but less so in mLN (Fig. 1a left panel, Fig. 1b). Second, the three immune checkpoints (Treg, PD-1, Tim3), hallmarks of T cell activation, were decreased in mLN CD4⁺ TILs compared to primary tumors prior to chemotherapy (Fig. 1c). Third, chemotherapy failed to induce degranulation of CTLs in mLN (Fig. 1a right panel) but did induce Tim3 expression (superior to that observed in primary lesions) and Treg in CD4⁺ TILs of mLN (Fig. 1c). Altogether, the CD8/CD4 ratio was largely in favor of CD4⁺ T cells despite neoadjuvant chemotherapy in mLN (Fig. 1b). Interestingly, the live CD45^- fraction (Fig. 1d) was positively associated with a favorable and higher CD8/CD4 ratio in mLN in the neoadjuvant setting (Fig. 1e).

Figure 1. Regulatory T cells and Tim3 expression in BC metastatic lymph nodes (mLN) despite immunogenic chemotherapy.

a. Relative percentages of the whole CD8⁺ cell fraction among live CD3⁺ cells and of the granzyme B expressing cells in the gate of CD8⁺ T cells. b. Ratio of CD8/CD4 tumor infiltrating T lymphocytes (TILs) in all 4 settings. c. Expression of distinct immune checkpoints on CD4⁺ T cells from primary tumor BC or mLN. d. Fraction of CD45^- cells remaining alive after dissociation. e. Spearman correlations between CD45^- cells and the ratio of CD8/CD4 in TILs in primary tumor versus mLN in the adjuvant and neodjuvant settings. The number of specimen is annotated. Unpaired Mann-Whitney t’test: *p<0.05, **p<0.01, ***p<0.001.

Immunogenic chemotherapy could switch on effector T cells in primary lesions but failed to do so in metastatic lymph nodes where Tim3 and Treg dominated the scenario.

Immune effects of IgG-based T-cell bispecific antibodies (TCB) in BC

To reinstate resident T cell functions in mLN, we utilized IgG-based T-cell bispecific antibodies (TCB) on freshly dissociated BC (in the previously described in sitro assays (14), Fig. S2). The bioactivity of the HER2-CD3ε TCB (HER2-T) was compared to medium alone and untargeted control Ab (Ctrl-T) with rIL-2 (as a positive control) in 3-16 lesions. The results of the functional assays describing the dynamics of the immune profile at Day 1 of co-cultures are summarized in the heat-map presented in Fig. 2a and Supplemental Fig. 3a for tumors handled in adjuvant and neoadjuvant settings, respectively. Briefly, HER2-T triggered degranulation of CD8⁺ and NK TILs (Fig. 2b-c, Fig. S3b-c), their proliferation (Fig. 2d, Fig. S3d) as well as IFNγ□cytokine release in T cells (Fig. 2e, Fig. S3e), mainly in mLN in the adjuvant setting. However, HER2-T did not increase TNFα □production by T cells nor the inflammatory contexture of BC (Fig. 2a, Fig. S3a). Of note, primary tumor-residing T cells already exposed and responding to chemotherapy failed to further react to HER2-T in most patients, and these reagents also failed to ameliorate the low degranulation and proliferative capacity (Fig. S3b-d) as well as IFNγ□cytokine release in tumor T cells (Fig. S3e). Importantly, HER2-T was not efficient at reinstating mLN TILs functions after neoadjuvant chemotherapy, regardless of TCB dose (Fig. S4).

Figure 2. Metastatic lymph nodes respond to HER2 targeted-T (HER2-T) cell bispecific antibodies.

a. Heatmap depicting the biological effects of HER2-T in mLN and primary tumors over controls in adjuvant settings performed at Day 1 (grey color: ND). b-e. T (b, d, e) and NK (c) functions investigated prior to chemotherapy using flow cytometry after intracellular staining, gating on live CD45⁺ cell fraction of dissociated primary BC or mLN. Relative percentages of granzyme B expressing CD8⁺ T and NK cells (b, c) at Day 1 or Day 6. d. Relative percentages of proliferative (% Ki67⁺) CD8⁺ T or NK cells (at Day 6). e. IFNγ□expressing CD8⁺ or CD4⁺ T cells tested at Day 1. Each dot represents one specimen and lines represent means of all dots. The number of specimen is annotated for each conditions: medium, versus isotype control mAb [Ctrl-T (isotype control for the HER2-T), HER2-T or recombinant human IL-2 (rIL-2). ND: Not determined. Unpaired Mann-Whitney t’test: *p<0.05, **p<0.01, or indicated p-value for trends. Also refer to Fig. S3-4 for comprehensive analyses of immune parameters in neoadjuvant settings and dose effects.

Figure 3. Metastatic lymph nodes respond to CEACAM5 targeted-T cell bispecific antibodies (CEACAM5-T).

T and NK cell functions in primary BC and mLN assessed after dissociation and incubation with CEACAM5-T at increasing dosing, prior to adjuvant chemotherapy monitored by flow cytometry after intracellular staining. a. Heat-maps depicting the biological effects of CEACAM5-T in mLN and primary tumors over controls in adjuvant settings performed at Day 1 and Day 6 (grey color: ND). b-c. Relative percentages of granzyme B, Ki67 or IFN□expressing CD8⁺ T cells (at Day 1 or Day 6) (b) or NK cells (c). Each dot represents one specimen and lines represent means of all dots. The number of specimen is annotated for each condition: medium, versus isotype control mAb [Ctrl-T (isotype control for the CEACAM5-T), CEACAM5-T or recombinant human IL-2 (rIL-2), as outlined. d. Pooled data from all 10 specimen segregated according to values >10% positivity, in control (medium, Ctrl-T) versus TCB exposed wells. Chi-square test and Unpaired Mann-Whitney t’test: *p<0.05, **p<0.01.

Next, we investigated the CEACAM5-CD3ε T cell bispecific antibody (CEACAM5-T) in about 10 lesions at increasing doses in mLN only in adjuvant settings. Its format incorporates bivalent binding to CEACAM5, a head-to-tail fusion of CEACAM5- and CD3ε-binding Fab domains as well as an engineered Fc region which completely abolished binding to FcγRs and C1q (16). As observed using the HER2-T, the CEACAM5-T induced T and NK cell effector functions and entry into the cell cycle (Ki67), even at low dosing of TCB (Fig. 3). Interestingly, in 3/5 cases tested, CEACAM5-T turned HER2-T-resistant lesions into positive ones (Fig. S5a-b).

In as much as i) Tim3 and PD-1 were expressed at high levels on CD4⁺ TILs from mLN (Fig. 1), ii) Tim3 and 4-1BBL (CD137L) were upregulated by HER2-T (Fig. S5c-d), iii) Fibroblast Activation Protein (FAP) was expressed on stromal cells in BC (Fig. S5e), we undertook to admix HER2-T with other bispecific antibodies (anti-FAP-4-1BBL and controls) (listed in Fig. S2) to boost T cell functions. In some cases (5 cases out of 12 tested), lesions resistant to HER2-T could respond to anti-FAP-4-1BBL antibodies for proliferation or GrzB and Th1 cytokine release (Fig. S5f-h).

We next analyzed which factors could predict responsiveness in in sitro assays to TCB. HER2 and CEACAM5 expression levels by tumor cells allowed to some extent to contrast responding (R) from non-responding (NR) lesions (Fig. S6a-c). Naïve tumors unexposed to chemotherapy, highly metastatic lesions (>3N+) or abundant CD45^- resident cells in mLN and high proportion of CD3⁺ at diagnosis in mLN favored the R phenotype in the adjuvant setting (Fig. S6d-f). mLN NK cells contrasted R from NR in neoadjuvant cases (Fig. S6g).

Altogether, redirecting polyclonal T cells to metastatic tumor cells, by connecting them via CD3 to a tumor-specific antigen on the surface of the LN tumor cell was efficient at triggering effector T (and indirectly NK) cell functions prior to chemotherapy in BC.

MHC class I loss in tumor-draining LN metastases

Three arguments point to a key role of tumor cells in triggering T cell reactivity in BC: i) the CD8/CD4 ratio of the mLN was positively correlated with the proportion of CD45^- cells (supposedly live residual tumor cells) (Fig. 1e), ii) the number of N+ and the proportion of CD45^- cells correlated with the efficacy of TCB (Fig. S6d-e), iii) chemotherapy did not pave the way to a better efficacy of TCB, suggesting that elimination of tumor cells jeopardized the cross-linking of the target antigen with the CD3ε (Fig. 2). Moreover, chemotherapy appeared to better target the primary lesions than the mLN by effectively reprogramming the TME while rendering mLN T cells more immunosuppressed (Tim3, Treg) (Fig. 1). These changes may account for the better efficacy of TCB in the mLN in adjuvant settings. Hence, we hypothesized that the density of the relevant MHC class I/ peptide complexes for the fitness of the T cell repertoire residing in mLN may be below the threshold of T cell reactivity. Immunohistochemistry staining of the specimen revealed a marked downregulation of MHC class I in 47 mLN compared to 52 primary BC, including 45 paired mLN/primary tumors (Fig. 4a-b left panel), across all molecular subtypes, including luminal cases for which TILs are not prognostic markers (Fig. 4c). Of note, about 1/3 of primary tumors harbored low/no MHC class I expression at diagnosis. Chemotherapy restored the MHC class I expression in distinct cases in mLN (Fig. 4b, right panel). Hence, other cases of HLA loss could be due to genetic deletions. Comparative genomic hybridization focusing on chromosome 6 and 15 where the loci of heavy chains of MHC class I molecules and the ß2 microglobulin (B2M) reside respectively, revealed genetic defects in B2M, more in metastatic (SAFIR01) than non-metastatic cohorts (TCGA) (Fig. 5a), and mostly in triple negative breast cancers, as assessed in SAFIR01 as well as in two distinct publicly available cohorts of primary BC (TCGA, METABRIC) (Fig. 5b).

Figure 4. Diagnosis of MHC class I deficiency in early BC by immunohistochemistry.

a. Representative micrograph pictures of IHC stainings using anti-HLA ABC mAb on paired primary and mLN BC at different magnitudes. b-c. H-score of MHC class I expression using anti-HLA ABC antibody in primary and mLN from BC, segregating the BC cases according to adjuvant versus neoadjuvant settings and the molecular subtyping. The paired cases are linked with a line. Paired Wilcoxon statistics: *p<0.05, **p<0.01, ***p<0.001.

Figure 5. Subclonal HLA LOH in early BC.

a-b. Loss of B2M in metastatic BC. CGH array of N=695 metastatic lesions (SAFIR01) compared with TCGA and METABRIC cohorts (N=1080), analyzing chromosome 6 and 15 associated-MHC loci. c. Percentage of BC patients (N=8) harboring (orange) or not (grey) HLA LOH as measured by HLA LOH and allele-specific characterization of clonal (red) and subclonal (purple) status. d. Number of BC patients with matched primary tumor and lymph node metastasis exhibiting no HLA LOH (grey), HLA LOH in both primary tumor and lymph nodes (green), HLA LOH in primary tumor only (blue) and HLA LOH in metastatic lymph nodes only (magenta). e. Percentage of regions harboring HLA LOH for central, peripheral and lymph node metastases. Analysis was performed for the 5 patients identified with HLA LOH. f. Representative example of the Log Ratio calculated in HLA LOH from a peripheral and a metastatic lymph node sample of a BC patient with HLA LOH. Open circles represent mismatch positions between the two HLA alleles. g. HLA LOH status and association with response to chemotherapy.

To further investigate somatic loss of MHC class I molecules in primary tumor and mLN, we sequenced 9 tumor regions per patient (3 primary central, 3 primary peripheral, and 3 mLN) and a matched normal sample for 8 patients with luminal breast cancer according to a computational tool recently developed (17). We identified loss of heterozygosity of class I HLA alleles (HLA-LOH) in 5 out of 8 patients (62%; Fig. 5c). HLA LOH was subclonal in all five patients (Fig. 5d) with only HLA-A allele being subject to clonal LOH in one patient (Fig. 5c). Interestingly, we found a trend for enrichment for HLA loss in mLN as compared to central regions of the primary tumors (Fig. 5e-f, Table S5; two-sided p-value of Fisher’s exact test 0.06).

We next characterized somatic MHC class I loss in the diagnostic core biopsy in a cohort of 154 patients treated with neo-adjuvant standard of care chemotherapy. 18% of cases had evidence of LOH of at least one MHC class I allele prior to commencing chemotherapy, and tumors with presence of HLA LOH in the core biopsies were much less likely to have pathological complete response (OR 0.27; CI: 0.07-0.89, p=0.04, multiple logistic regression model factoring ER and HER2 status, Fig. 5g). In order to assess HLA abundance in the post-treatment samples, 24 cases were chosen based on their HLA LOH status at diagnosis (6 had evidence of HLA LOH) and presence of viable tumor tissue in both the primary breast tumor and metastatic lymph nodes. Following H-scoring, one sample was excluded as there were too few tumor cells in the breast tumor. In the 23 paired samples, there was a statistically significant decrease in H-score in the draining lymph nodes compared to breast tumor (p=0.04, Wilcoxon rank sum test), 12 samples showed a significant decrease while the other 11 samples showed no significant change (Fig.4b, neoadjuvant cases, p=0.004, Wilcoxon rank sum test in the paired analysis).

This observation showing the high frequency of genetic or non-genetic deficiencies in MHC class I expression in mLN further supports our hypothesis that metastatic lymph nodes are characterized by a lower density of MHC class I/peptide complexes than the primary BC tumors.

Discussion

The sentinel lymph node (SLN) is thought to be an important lymphoid organ for protecting against metastasis by eliciting anti-tumor immunity. However, our study unraveled, i) the profound immunosuppression residing in tumor-draining LN despite neoadjuvant chemotherapy supposed to reinstate local immuno-surveillance, ii) the capacity of TCB targeting CEACAM5 or HER2 to restore TILs effector functions despite low target antigen expression due to HLA-LOH.

The immunosuppressive microenvironment is a mainstay in BC, already detectable during the transition between in situ ductal carcinoma (DCIS) towards invasive BC (IDC). The CD8/Foxp3 ratio could be a hallmark of risk assessment for local recurrence and invasion in DCIS (18)(19). Our data show that neoadjuvant chemotherapy increased the proportion of Treg and Tim3 expression on CD4⁺ TILs in mLN while the CD8/CD4 ratio did not get ameliorated, in sharp contrast to what was observed in primary tumors.

The more immunosuppressive tumor environment of mLN (compared with the primary site) might be explained by the frequent loss of MHC class I molecules of the metastatic clones. Therefore, this study highlights the lack of effectiveness of conventional therapies to counteract natural immunosuppression mediated by BC progression to the first LN.

Importantly, bispecific T-cell engagers could restore T and NK cell functions in mLN (GrzB, Ki67, and IFNγ release in T cells), which are altered in IDC. Indeed, TILs gene products involved in TCR and cytokine receptor (IL-2RA and IL-10RA) signaling were upregulated in DCIS and normal breast compared with IDC (19). Moreover, DCIS CTLs harbored high clonal diversity in accordance with their high proliferative capacity and cytolytic pattern (19), features that are progressively lost in IDC. TCB tended to increase IL-2, IL-10, and CXCL10 (not shown) but did not increase inflammatory cytokines (Fig. 2,Fig 3,Fig. S3) that are known to compromise TCR signaling by accelerating CD3ζ□degradation (20).

As an indirect argument supporting the relevance of T cells in BC immuno-surveillance, the emergence of antigen (MHC class I or proteins of the antigen processing machinery) loss variants has been described in HER2 positive BC, precluding proper recognition of HLA-A2-associated tumor epitopes by cytotoxic T cells (21). The proto-oncogene HER2 downregulates MHC class I expression through a RAS/Mitogen-activated protein kinase (MAPK)-dependent pathway (22). Likewise, chemotherapy and/or trastuzumab tend to eliminate HER2 overexpressing cells that are more sensitive to cytotoxic agents, paving the way to T cell responses, and the emergence of MHC class I loss variants. Several studies reported contradictory correlations between MHC class I loss and prognosis in BC (23)(24). MHC class I and II loss correlated with low MxA expression and fewer tertiary lymphoid structures and TILs in TNBC (25).

In advanced lung carcinomas, subclones harboring HLA-LOH were associated with a significantly elevated non-synonymous mutation/neoantigen burden and higher PD-L1 expression compared to subclones descended from the same ancestral tumor cell without HLA LOH (17). Using the same computational tool, we discovered a high prevalence (5/8 tumors) of HLA LOH in BC, suggesting HLA loss may be a pervasive mechanism of immune evasion in this cancer type as well, but at very early stages compared with lung tumors. Intriguingly, we found evidence for enrichment for HLA LOH in the lymph node metastasis (9/15 tumor samples) compared to central regions (3/15 tumor samples), suggesting HLA LOH may facilitate immune escape in the context of lymph node metastasis.

A considerable advantage of BiTE® or TCB molecules is that they directly engage any cytotoxic T cell and thus do not require the presence of preformed MHC class/peptide complexes, nor a high density of target antigens to mediate a full-blown T cell activation (10). Our study was biased towards probing novel therapeutic molecules onto available BC tissue samples that pathologists could provide after a comprehensive routine scoring of grade and tumor aggressiveness. Nonetheless, this bias enabled us to speculate that TCB could represent a reasonable alternative option to conventional chemotherapy (26). A related CEACAM5-T- cell bispecific (CEA-TCB, RO6958688, RG7802) antibody for the treatment of CEACAM5-expressing solid tumors currently in phase I clinical trials (NCT02324257), which led to encouraging results in advanced colorectal cancers. The high-avidity binding to CEACAM5 conferred by the antibody’s design, together with the bivalent binding mode to tumor antigen, translates into an excellent therapeutic index, optimal efficacy with low toxicity profile (27).

Theoretically, TCB could eliminate MHC class I loss variants selected by chemotherapy, since our neoadjuvant trials indicated that HLA class I loss raised from 18% to 50% after 6 cycles. Indeed, TCB only improved CTL functions in mLN, and not in primary lesions (Fig. 2–3), because the latter ones are already partially triggered by immunogenic chemotherapy (Fig. S3). However, mLN could hardly be reactivated by TCB pre-exposed to chemotherapy, although chemotherapy did not ignite T cells in this case. Therefore, it may be more effective to reactivate preexisting tissue resident memory T cells (TRM) (28) using TCB upfront prior to surgery and/or chemotherapy to facilitate tumor killing and cross-presentation of novel tumor antigens for the creation of a new TCR repertoire, hence preventing additional MHC class I loss variants generated post-chemotherapy.

In fact, our data appear to indicate that the initial representativeness in CD8⁺ T cells together with high (>70%) invasion by CD45^- cells correlate with the TCB-mediated capacity to trigger proliferative and cytotoxic functions of effector cells and cytokine release in the TME, bypassing the HLA LOH associated with the immunosuppressive milieu. These findings warrant pilot trials in window-of-opportunity settings before surgery and subsequent adjuvant chemotherapy to demonstrate the capacity of TCB to increase the CD8/CD4 ratio, GrzB on CD8⁺ T and NK cells and to decrease Ki67 in EpCAM⁺ tumor cells, not only in the primary tumor but most importantly in the metastatic LN.

Key message.

In this original article, the authors showed that HLA loss of heterozygosity is observed in early Breast Cancer, and is associated with Treg cells and Tim3 expression restraining the immuno-stimulatory effects of chemotherapy. T cell bispecific antibodies (TCB) bridging CD3ε and HER2 or CEACAM5 could bypass HLA loss, partially restoring T cell functions in axillary lymph nodes.