Research ArticleCancer

Trastuzumab emtansine (T-DM1) renders HER2+ breast cancer highly susceptible to CTLA-4/PD-1 blockade

See allHide authors and affiliations

Science Translational Medicine  25 Nov 2015:
Vol. 7, Issue 315, pp. 315ra188
DOI: 10.1126/scitranslmed.aac4925

Targeted therapy with more punch

Overexpression of the human epidermal growth factor receptor 2 (HER2) is common in breast cancer, and it is associated with poor outcomes despite the availability of trastuzumab, an antibody against HER2, and other HER2-targeted agents. The reason for the poor outcomes is that many patients develop resistance to the targeted drugs. Müller et al. have now shown that this resistance can be overcome with trastuzumab emtansine, an antibody-drug conjugate that combines the HER2-targeting ability of trastuzumab with a cytotoxic drug, which the antibody delivers directly to the tumor. In addition to its cytotoxic effects, treatment with trastuzumab emtansine activated a strong antitumor immune response and effectively combined with immune checkpoint inhibitors, suggesting that it can be used in combination therapy.


Targeted drug delivery with antibody-drug conjugates such as the HER2-directed ado-trastuzumab emtansine (T-DM1) has emerged as a powerful strategy for cancer therapy. We show that T-DM1 is particularly effective in eliciting antitumor immunity in patients with early breast cancer (WSG-ADAPT trial) and in a HER2-expressing orthotopic tumor model. In the latter, despite primary resistance to immunotherapy, combined treatment with T-DM1 and anti–CTLA-4/PD-1 (cytotoxic T lymphocyte–associated protein-4/programmed cell death protein-1) was curative because it triggered innate and adaptive immunity. Tumor rejection was accompanied by massive T cell infiltration, TH1 (T helper 1) cell polarization, and, notably, a substantial increase in regulatory T cells. Depletion of regulatory T cells resulted in inflammation and tissue damage, implying their essential role in protecting the host during therapy. This study provides insights into the mechanisms of T-DM1’s therapeutic activity and a rationale for potential therapeutic combination strategies with immunotherapy.


About 15 to 20% of patients with breast cancer exhibit amplification and overexpression of the human epidermal growth factor receptor 2 (HER2) (1, 2). This genetic alteration promotes cancer cell proliferation and survival, resulting in increased tumor growth and poor clinical outcome in the absence of HER2-targeted therapy (1, 2). Trastuzumab is a HER2-specific monoclonal antibody (mAb), which represents a milestone in breast cancer therapy. Trastuzumab has improved the outcome of patients with early and advanced HER2-positive breast cancer, which has confirmed the value of HER2 as an effective therapeutic target (35). Several other HER2-targeted agents, including the dual HER1/HER2 kinase inhibitor lapatinib and the HER2/HER3 dimerization inhibitor pertuzumab, which are administered as part of combination treatment regimens with chemotherapy, endocrine treatment, and/or another HER2-blocking agent, have also demonstrated clinical efficacy (68). Nonetheless, de novo and acquired resistance to HER2 blockade eventually occurs in most, if not all, patients with advanced disease (9). Therefore, new therapeutic avenues are clearly needed not only in early but also in metastatic HER2-positive breast cancer, which to date remains incurable.

Ado-trastuzumab emtansine (T-DM1) is an antibody-drug conjugate (ADC), composed of trastuzumab and the cytotoxic moiety DM1, a maytansine derivative, covalently conjugated to the antibody via a stable thioether linker (10). By combining the target specificity of a mAb with the therapeutic potential of a powerful cytotoxic agent, ADCs offer selective ablation of cancer cells and markedly increase the therapeutic window (11). T-DM1 is a first-in-class drug, which was developed to improve the efficacy of trastuzumab. Upon binding to HER2, T-DM1 undergoes receptor-mediated internalization and subsequent proteolytic digestion, releasing the cytotoxic catabolite lysine-Nε-4-(N-maleimidomethyl)cyclohexane-1-carboxylate (MCC)–DM1 (12). In addition, T-DM1 blocks HER2-mediated signal transduction, mediates antibody-dependent cell-mediated cytotoxicity (ADCC), and inhibits shedding of the HER2 extracellular domain (13). The therapeutic activity of T-DM1 in trastuzumab-resistant HER2-positive breast cancer, both in preclinical cancer models and in prospective clinical trials (14, 15), along with its favorable safety profile, prompted numerous trials of this compound in several therapeutic indications (16).

Maytansine derivatives such as DM1 bind to a site on β-tubulin that is distinct from the vinca domain and are potent inhibitors of microtubule assembly (17, 18). Early clinical trials with systemically applied maytansine showed not only moderate therapeutic activity but also substantial toxicity (19). This disappointment resulted in the development of synthetic derivatives of maytansine such as DM1, which have 100- to 1000-fold higher cytotoxicity than clinically used anticancer drugs (20) and are used as the cytotoxic payload in ADCs. We recently reported that ansamitocin P3, a direct precursor in the synthesis of DM1, which has a direct cytotoxic effect on tumor cells, is capable of inducing the full spectrum of maturational changes in dendritic cells (DCs), resulting in robust activation of antitumor immunity (21). Our results demonstrate that Ansamitocin P3 facilitates antigen uptake and migration of tumor-resident DCs to the tumor-draining lymph nodes, thereby potentiating the tumor-specific T cell response in vivo. Notably, we observed similar immune-mediated mechanisms for other microtubule-depolymerizing agents, such as dolastatin-10 and its synthetic analog monomethyl auristatin E (MMAE) (22, 23). The latter is used as cytotoxic payload in the ADC brentuximab vedotin, which results in high and sustained clinical responses in patients with CD30+ lymphomas (24). Indeed, brentuximab vedotin induces activation of patient DCs, T cells, and B cells, reflecting augmentation of tumor-specific immunity (22, 25).

Given this mechanism of immune engagement observed for the microtubule-depolymerizing cytotoxic moieties used in ADCs, we hypothesized that the high level of clinical activity observed with T-DM1 (13) relies, at least in part, on the induction of an efficient antitumor immune response. The latter would allow the combination of T-DM1 treatment with immunotherapeutic strategies such as antibodies blocking immune checkpoints, including cytotoxic T lymphocyte–associated protein-4 (CTLA-4) or programmed cell death protein-1 (PD-1), which should help overcome resistance and potentially result in long-term benefit. We analyzed paired tumor samples from breast cancer patients before and after short-term preoperative treatment with T-DM1 within the WSG-ADAPT (adjuvant dynamic marker-adjusted personalized therapy) trial (26), which revealed a substantial increase in the number and density of tumor-infiltrating T cells. To further substantiate these data and explore combinations of T-DM1 with immune checkpoint inhibition, we studied a HER2-expressing, trastuzumab-resistant, orthotopic and syngeneic breast cancer model, which has been extensively used during the preclinical development of T-DM1. The resistance of this model to trastuzumab, which forms the antibody backbone of T-DM1, allowed us to exclusively focus on the therapeutic and immunologic effects of the cytotoxic payload DM1, as well as on its interaction with immune checkpoint blockade.


Intratumoral recruitment of T cells upon T-DM1 therapy

Building on our findings that ansamitocin P3 (21) and its synthetic derivative DM1 (fig. S1) induce phenotypic and functional DC maturation, we hypothesized that T-DM1 induces lymphocytic infiltration into the tumor microenvironment of patients with HER2-positive breast cancer. Matched tumor biopsies (pretherapeutic versus on-treatment) from 28 treatment-naïve patients diagnosed with HER2-positive/estrogen receptor (ER)–positive breast cancer, who underwent a single treatment of T-DM1 monotherapy within a subtrial of the WSG-ADAPT protocol (NCT01745965) (26), were subjected to histomorphological evaluation. The WSG-ADAPT study is a large multicenter umbrella trial with breast cancer subtype–specific subprotocols; the ADAPT HER2+/ER+ subtrial assesses early therapy response to T-DM1 monotherapy, T-DM1 plus endocrine therapy, or trastuzumab plus endocrine therapy. A second core biopsy after initial diagnosis (“pretherapeutic”) is performed after 3 weeks of induction therapy (“T-DM1–mono”). Tumor-infiltrating lymphocytes (TILs) dispersed in the stroma between the carcinoma cells (stromal TILs) were assessed independently by two trained histopathologists (27, 28), according to current histomorphological consensus guidelines. At the time of diagnosis, stromal TILs were sparse. Accordingly, the lymphocyte-predominant phenotype (stromal TILs ≥40%) was rare in HER2-positive/ER-positive breast cancer [3 of 28 (10%) patients] before therapy, which is in line with previous findings (29). Stromal TILs increased in response to T-DM1 (P = 0.005, Wilcoxon matched pairs test) (Fig. 1A). On average, the percentage of stromal TILs was significantly higher in on-treatment biopsies (pretherapeutic: mean, 13.77 ± 2.7%; on therapy: mean, 20.84 ± 2.9%) (Fig. 1A), whereas the CD8/CD4 ratio did not change significantly (fig. S2). Histomorphological evaluation after 3 weeks of T-DM1 monotherapy revealed that 12 of 28 (43%), 8 of 28 (29%), 6 of 28 (21%), and 2 of 28 (7%) of treatment-naïve patients exhibited marked cytopathic changes (histological regression grade 1), extensive tumor destruction (regression grade 2), no histologically evident treatment effect in the tumor cell compartment (regression grade 0), and only residual carcinoma in situ or no tumor cells in the tumor bed (regression grade 3 and 4), respectively (30). Quantitative changes in stromal TILs appeared to be unrelated to the histological regression grade: increased presence of stromal TILs was noted in some patients with a regression grade 0. Conversely, a histiocytic resorption reaction with very few TILs was encountered in the residual tumor bed of a patient with grade 4 regression (Fig. 1B). Most cases, however, showed a subtle to marked (often regional) increase in TILs, including CD4+ and CD8+ T cells (Fig. 1, A and C).

Fig. 1. Development of TILs upon T-DM1 therapy in human and murine breast cancers.

(A) Percentage of stromal TILs in human breast cancer core needle biopsies before and after T-DM1 monotherapy (3 weeks). Unpaired data are presented on the left side, whereas on the right side, the data are presented as a trend diagram with matched pairs linked by lines. The histological regression grade is coded by color [pretherapeutic: mean, 13.77 (SEM, 2.76); T-DM1–mono: mean, 20.84 (SEM, 2.91)]. Data on the left were analyzed using the Wilcoxon test. Data on the right were analyzed by paired t test. (B) Complete histological regression (grade 4) after T-DM1 monotherapy. The residual tumor bed shows numerous histiocytes and very few lymphocytes. Scale bar, 50 μm. (C) Histological regression grade 1, with regional increase in TILs, including CD4+ and CD8+ T cells. Scale bar, 50 μm. (D) H&E (left panel) and anti-HER2 (right panel) immunohistochemical staining of an orthotopic Fo5 tumor. Scale bar, 40 μm. (E) Therapeutic response of Fo5 tumor–bearing mice upon treatment with one (1×) or two (2×) doses of T-DM1 (15 mg/kg). Mice were treated when tumors reached an average size of 80 mm3 and were euthanized when tumors reached 1200 mm3. (F and G) Representative cryosections from control and T-DM1–treated tumors (one dose) were stained for CD4 (F) and CD8 (G) 11 days after treatment [blue: 4′,6-diamidino-2-phenylindole (DAPI); red: HER2; green: CD4/8]. Scale bar, 100 μm. (H) Single-cell suspensions from whole-tumor digests were restimulated for 4 hours with phorbol 12-myristate 13-acetate (PMA)/ionomycin and analyzed for IFN-γ production using fluorescence-activated cell sorting (FACS). (I) Fo5 tumor–bearing mice were depleted of CD4 and CD8 T cells using depleting antibodies and treated as in (E). (E and I) Pooled data from at least three independent experiments (n ≥ 18).

To further delineate the immunological mode of action of T-DM1 in vivo, we used orthotopically grown tumors derived from transgenic mice that overexpress human HER2 under the control of the murine mammary tumor virus promoter (Fo5) as a model of human HER2–overexpressing breast cancer (31). Hematoxylin and eosin (H&E) and α-human-HER2 staining are shown in Fig. 1D, and tumor responses after one (1×) or two (2×) administrations of T-DM1, in terms of survival, are shown in Fig. 1E. In line with our observations in human primary breast cancers, we confirmed an increase in TILs, particularly T cells, upon T-DM1 therapy in the Fo5 breast cancer model (fig. S3 and Fig. 1, F and G). Accordingly, we found a strong increase in interferon-γ (IFN-γ) production in T-DM1–treated mice (Fig. 1H). To further characterize the importance of T cells for the therapeutic efficacy of T-DM1, we depleted CD4+ and CD8+ T cells with T cell–depleting antibodies, which severely reduced the overall survival time and rate of T-DM1–treated mice (Fig. 1I). This observation substantiates a crucial role for T cells in T-DM1–mediated immune activation and tumor rejection.

Antitumor activity of immune checkpoint inhibitors and T-DM1 therapy

We analyzed the expression of immune checkpoint receptors on T cells and their cognate ligands on tumor-associated macrophages (TAMs). A markedly increased expression of CTLA-4 was noted on both CD4+ and CD8+ T cells upon T-DM1 treatment. Although PD-1 expression was only slightly up-regulated, its ligand PD-L1 showed a strong up-regulation on TAMs (Fig. 2, A and B). On the basis of these findings, our previously reported synergistic interaction of ansamitocin P3 with immune checkpoint inhibition (21), the clinical successes of CTLA-4 (32) and PD-1/PD-L1 (33, 34) blocking antibodies, and the very potent therapeutic interaction of α-CTLA-4 and α-PD-1 (35), we assessed a combination therapy consisting of T-DM1 and CTLA-4– and PD-1–blocking antibodies. The combination of CTLA-4– and PD-1–blocking antibodies alone was completely ineffective; however, combined application with T-DM1 resulted in strong antitumor efficacy, both with one and two administered doses of T-DM1, the latter resulting in almost 100% of complete cures (Fig. 2C). No visible side effects were observed in the treated animals. As depicted in Fig. 2D, the mice that remained tumor-free after combination treatment were protected from tumor rechallenge, whereas all control mice rapidly developed tumors, indicating a strong, T cell–centered, immunological memory formation in the treated mice.

Fig. 2. Expression of CTLA-4, PD-1, and PD-L1 and combination therapy of T-DM1 with α-CTLA-4/PD-1.

(A and B) Single-cell suspensions from Fo5 tumors were stained for TAM markers as well as for PD-1, PD-L1 (A), and CTLA-4 (B). (C) Therapeutic response of Fo5 tumor–bearing mice upon treatment with one (1×) or two (2×) doses of T-DM1 (15 mg/kg) alone, α-CTLA-4/PD-1 alone, or their combination (T-DM1 + α-CTLA-4/PD-1). (D) Mice that remained tumor-free after combination therapy and control mice were rechallenged in the contralateral mammary gland using cell suspensions from whole Fo5 tumor digests. (C) Pooled data from at least four independent experiments (n ≥ 23). (D) Data from pooled, tumor-free mice, which remained tumor-free after combination treatment (n = 18), and control mice (n = 14) from three independent experiments.

Patterns of immune cell infiltrates and chemokines/cytokines in treated mice with breast cancer

We next assessed changes in intratumoral immune cell subsets in the different treatment arms. Figure 3A depicts the breakdown of CD45+ immune cells along with the CD45 cell fraction. Expansion of CD45+ cells occurred with T-DM1 treatment and particularly with the combination therapy. Figure 3 (B to D) and fig. S4 show a detailed comparison of T cell receptor (TCR) γ/δ T cells, B cells, natural killer T (NKT) cells, NK cells, and CD11b+/CD11c+ myeloid cells. Whereas B cells decreased only after combination treatment (fig. S4B), CD11b+/CD11c+ myeloid cells strongly decreased as compared to all CD45+ cells upon T-DM1 monotherapy and combination treatment (Fig. 3B). In contrast, a relative increase in TCRγ/δ T cells, NKT cells, and NK cells was observed after T-DM1 treatment (fig. S4, A, C, and D). Similarly, we noted an increase in the frequency of EOMES-positive NK cells, which exert potent antitumor effector functions (36). TAMs lost the M2-specific expression of arginase and gained expression of PD-L1 upon T-DM1 treatment (Fig. 3, C and D), which was most pronounced for the combination therapy. The latter finding is indicative of a strong, lymphocyte-centered, immune activation associated with increased IFN-γ production, which has been shown to mediate PD-L1 up-regulation (37). With trastuzumab alone and in combination with CTLA-4/PD-1–blocking antibodies, we noticed no therapeutic benefit and no prominent expansion of the CD45+ cells in the tumor stroma (fig. S5). This finding strongly suggests that the immune modulatory properties of the cytotoxic payload DM1 are mainly responsible for the pronounced immune infiltration. We used FACS and immunofluorescence to analyze T cell numbers and observed a strong increase in CD4+ and CD8+ T cells in mice treated with T-DM1, which was even more pronounced for the combination therapy (Fig. 4, A to C). We also conducted Luminex analysis to measure cytokine, chemokine, and growth factor concentrations (Fig. 4D and fig. S6). Both T helper 1 (TH1) and TH2 cytokines substantially increased, which further supports the immune-promoting role of T-DM1 and its interaction with CTLA-4/PD-1–blocking antibodies. Our data also provide an explanation for the increased T cell and NK cell numbers owing to the increased amount of proinflammatory chemoattractants such as MIG/CXCL9, MIP-1α/CCL3, MIP-1β/CCL4, and RANTES in tumors from mice treated with T-DM1 and the combination. Furthermore, we noticed a drop in tumor-promoting chemokines such as KC/CXCL1, macrophage colony-stimulating factor (M-CSF), and, most notably, vascular endothelial growth factor (VEGF), whereas basic fibroblast growth factor concentration increased.

Fig. 3. Patterns of intratumoral immune cell subsets after T-DM1 therapy.

(A) Summary of CD45+ immune cell subsets and CD45 cells as determined by FACS. Subsets are depicted as percentage of all acquired live events; n ≥ 40 mice per group from eight pooled, independent ex???periments. Large diagram: CD45 (dark gray), myeloid/DC (red), lymphoid/other (dark blue); smaller diagram: CD4 (conventional) (green), CD4 (Treg) (violet), CD8 (dark blue), NK cells (orange), NKT cells (light blue), TCRγ/δ T cells (yellow), and B cells (light gray). (B) CD11b/c-positive cells as percentage of all CD45+ cells. (C) Arginase expression in TAMs. (D) PD-L1 expression on TAMs. (B to F) Pooled data from at least five independent experiments (n ≥ 21).

Fig. 4. Characterization of intratumoral T cells after T-DM1 therapy.

(A and B) Tumor-infiltrating CD4 (A) and CD8 (B) T cells are displayed as percentage of total live events (left), percentage of all CD45+ cells (middle), and cells/cm3 tumor tissue (right). (C) Representative tumor sections treated as indicated were stained for CD4 (top panel) and CD8 (bottom panel) (blue: DAPI; red: HER2; green: CD4 or CD8, respectively). Scale bar, 100 μm. (D) Tumors from mice treated as in (C) were analyzed for the indicated cytokines/chemokines. (A and B) Pooled data from at least eight independent experiments (n ≥ 36). (D) Pooled data from two independent experiments (n = 8). Data are depicted as box and whisker plots with whiskers showing min to max.

Activation and proliferation of CD4+ and CD8+ T cells infiltrating murine breast tumors

We next performed a functional characterization of tumor-infiltrating T cells in control and treated mice (Fig. 5, A and B). For T-DM1 and particularly for the combination, we observed a marked increase in the expression of inhibitory receptors such as CTLA-4, PD-1, and TIM-3 (T cell immunoglobulin and mucin-domain containing-3), consistent with T cell activation. The proliferation marker Ki-67 was highly expressed in CD8+ TILs in the combination group. Along the same lines, we also observed a strong increase in IFN-γ and granzyme B production. However, granzyme B production was not observed in CD4+ T cells. Consistent with a potent immune activation and antitumorigenic activity of intratumoral T cells, we found a strong TH1 immune deviation as evidenced by expression of the TH1-specific transcription factor Tbet in CD4+ T cells. We also analyzed granzyme B and IFN-γ production in mice treated with trastuzumab and trastuzumab plus α-CTLA-4/PD-1 but did not observe any significant immune-modulatory effects (fig. S7).

Fig. 5. Functional characterization of tumor-infiltrating T cells.

(A) Tumor-infiltrating CD4 and CD8 T cells were analyzed for the indicated parameters. (B) The parameters analyzed in (A) were quantitated and are displayed as percentage of total CD4+ cells or CD8+ T cells, as indicated. (B) Pooled data from at least five independent experiments (n ≥ 21).

Regulatory T cells and their critical role in preventing autoimmune inflammation in treated mice with breast cancer

The ratio of CD8+/regulatory T cells (Tregs) has been proposed as a useful predictive index for the success of cancer immunotherapy, reflecting that depletion or inactivation of CD4+ Tregs may represent a potentially effective strategy to enhance antitumor immunity (38, 39). However, most experimental data supporting this concept have been obtained from subcutaneous, cell line–based cancer models. Using the orthotopic breast cancer model, we found an increase in Tregs [CD45+CD11b/cTer119CD4+FOXP3+CD25+(Helios+)], which was most pronounced with the combination treatment at the tumor site (Fig. 6A and fig. S8A). Whereras the CD8+/CD4+ ratio remained essentially unchanged, the CD8+/Treg ratio declined for the combination group with T-DM1 (fig. S8B), in striking contrast to the other treatment groups (fig. S8). Notably, the Treg frequency was also increased in the periphery (fig. S8C). The increased Treg frequency at the tumor site was not observed with trastuzumab (fig. S9).

Fig. 6. Quantitation and characterization of tumor-infiltrating CD4 T cells.

(A) From left to right, tumor-infiltrating CD4 T cells are displayed as percentage of all CD45+ cells, FOXP3+ CD4+ T cells as percentage of total CD4+ T cells, and FOXP3+/CD25+ CD4+ T cells as percentage of total CD4+ T cells. Pooled data from at least six independent experiments (n ≥ 32). (B) Tumor-infiltrating CD4 T cells were analyzed for the indicated parameters. (C) eFluor-labeled CD8+ T cells (effector) were cocultured with Tregs (suppressors) in the presence of plate-bound anti-CD3/CD28 to determine the capacity of the Tregs to suppress CD8+ T cell proliferation. (D) Survival rates of mice bearing Fo5 tumors (as in Figs. 1 and 2) that were depleted of CD4, CD8, or CD4 and CD8 T cells and treated with T-DM1 (two doses) and α-CTLA-4/PD-1 (combo) or left untreated as indicated [pooled data from three independent experiments are shown (n ≥ 16)]. (E and F) Autoimmune phenotype in mice receiving combination therapy and the indicated T cell–depleting antibodies. (E) Images of mice from the indicated treatment arms. Both top images in (F) were taken at the same magnification. Scale bar, 40 μm. The left image in (F) shows a normal ear of a wild-type mouse. On the right in (F), the epidermis is massively hyperplastic (white arrow) and shows intracorneal pustule formation and marked intercellular edema due to an epidermotropic lymphocytic infiltrate (*). We observed scattered apoptotic keratinocytes and sparse lymphocytic infiltrate in the edematous dermis (open arrow). The lower left image in (F) shows a normal parotid gland of a wild-type mouse. Scale bar, 40 μm. On the lower right, the parotid gland contains multiple lymphocytic foci associated with intercalated ducts (white arrow). The lymphocytes invade the epithelium of the intercalated duct and form lymphoepithelial lesions within the acini (*). Some acinar cells are apoptotic (open arrow).

Tregs from tumors treated with combination therapy expressed high amounts of CTLA-4, Ki-67, and Tbet but did not produce increased concentrations of interleukin-17 (IL-17) and displayed only a moderate increase in IFN-γ production (Fig. 6B). We noted no difference in the suppressive capacity between Tregs from control and combination-treated animals (Fig. 6C). Next, we separately depleted either CD4+ or CD8+ T cells in mice receiving the combination treatment (Fig. 6D). Depletion of CD8+ T cells markedly decreased the therapeutic efficacy of the combination. Furthermore, depletion of CD4+ T cells (which include Tregs) diminished the therapeutic efficacy, although considerably less compared to CD8+ T cell depletion. This indicates that CD4+ T cells contribute to the antitumor effects of the combination therapy and that the elevated Treg frequencies, despite maintaining their suppressive capacity, do not interfere with tumor rejection in the combination. In addition, tumor-free mice had developed a memory response toward Fo5 tumors, and 18 of 18 mice were protected in a rechallenge experiment, whereas all control mice developed breast tumors (Fig. 2D).

Tregs were critical for the prevention of adverse events, and their depletion resulted in severe autoimmune inflammatory symptoms in more than half of the mice receiving combination therapy. Two tumor-free mice in this group were excluded from the analysis in Fig. 6D; one mouse was euthanized because of severe autoimmune symptoms and the other died. Histologic changes in the skin of these mice revealed features of eczema and graft-versus-host disease, and the appearance of their salivary glands resembled Sjögren’s syndrome (Fig. 6, E and F).


T-DM1 has greatly influenced the development of ADCs for solid malignancies, owing to its clinical activity even in patients refractory to other HER2-directed therapies and its favorable safety profile. The distinct in vitro and in vivo efficacy has been attributed mainly to disruption of microtubules, resulting in cell cycle arrest at the G2-M interface and tumor cell apoptosis, through intracellular release of DM1 (40). The contribution of antitumor immunity to T-DM1’s therapeutic mode of action has not been considered so far. Intrigued by our recent data showing that the precursor of DM1 ansamitocin P3 reinstates immunosurveillance by driving a distinct DC maturation program (21, 22), we used a transplantable, immunocompetent, and orthotopic breast cancer model derived from a transgenic mouse mammary tumor virus (MMTV)–human HER2–driven murine tumor model (31). Notably, because large established tumors from this model do not respond to single-agent trastuzumab therapy but continue to express high amounts of HER2 (41), this mouse model represents an opportunity to evaluate the in vivo efficacy of therapeutic strategies in trastuzumab-resistant breast cancer. The latter still represents a clinical challenge because a fraction of treated patients experience primary resistance to trastuzumab, and about 70% of initial responders acquire secondary resistance (42). Hence, battling trastuzumab resistance requires more effective, targeted therapies and/or combination treatments. Taking advantage of this model, we aimed at specifically dissecting the immune-modulatory capacities of DM1 and, subsequently, its interaction with immunotherapy. We demonstrate that T-DM1 increases TILs in human primary breast cancers and induces infiltration by effector T cells in murine breast tumors. In the latter, these T cells are essential for its full therapeutic activity. Although primary resistance to immune checkpoint-blocking antibodies occurred, combining T-DM1 treatment with blockade of the PD-1/CTLA-4 inhibitory pathway resulted in complete cures and greatly enhanced T cell responses, including complete tumor rejection and memory formation in our model. Unexpectedly, the therapeutic benefit was accompanied by a substantial increase in functionally suppressive Tregs, which were essential to protect mice from therapy-induced, immune-mediated autoimmune disease during tumor rejection.

The interplay between tumor cells and immune cells is gaining increasing recognition as a major determinant of breast cancer progression and treatment (43). Although recent data from large cohorts have indicated an association between the presence of lymphocytic infiltration and improved prognosis (4446), substantial heterogeneity is observed (45, 47). Particularly in the triple-negative subtype, increased immune infiltrate predicts both response to chemotherapy and improved survival (48). Moreover, for HER2+ disease, Perez and colleagues (49) were recently able to demonstrate that gene expression patterns related to immune function are strongly linked to clinical outcome in trastuzumab-treated patients. The role of FOXP3-positive Tregs cells involved in the maintenance of immunological self-tolerance and suppression of antitumor immune responses (50) remains controversial in breast cancer but may confer a worse prognosis with an increased risk of relapse (51, 52).

The field of cancer immunotherapy is rapidly expanding with the demonstration of improved overall survival and durable responses mediated by PD-1 and PD-L1 inhibitors (33, 53, 54). This therapeutic activity is noted across multiple cancer types and even in patients refractory to standard treatments. Although some patients respond remarkably well, innate resistance represents a hurdle for the clinical success of these therapies. Exciting perspectives include combining immunotherapy with chemotherapy or targeted therapy (5557). From a mechanistic point of view, these therapies may act on tumor cells by destroying them and/or modulating their propensity to elicit antitumor immune responses, or alternatively by stimulating immune effector cells, either directly or by relieving immunosuppressive mechanisms in the tumor microenvironment. Additionally, the immunologic profile of the tumor may be substantially altered by neoadjuvant chemotherapy. It has been demonstrated particularly in breast cancer that this approach may lead to accumulation of effector T cells and/or depletion of potentially immunosuppressive cells such as Tregs, which may correlate with an increased likelihood of pathological responses and, ultimately, better prognosis (46, 58). Notably, and consistent with our findings, it was recently reported that intratumoral infiltration with FOXP3+ Tregs and CD8+ effector T cells upon neoadjuvant chemotherapy was associated with pathological complete responses (59). Although one could speculate that the simultaneous presence of FOXP3+ Tregs and effector T cells may cause the suppression of the latter, our data argue for an important host-protective role of Tregs during the induction of potent, TH1-polarized antitumor immune responses, which are characterized by Tbet and IFN-γ up-regulation in T cells. In line with this, Tregs in the combination group, while maintaining their suppressive capacity, acquired a TH1-polarized phenotype, indicative of a strong TH1-polarized immune response. This finding has been previously described for infectious diseases, where it allows Tregs, upon peripheral differentiation and specialization, to maintain normal immune homeostasis during TH1 responses in vivo (6062). Differential depletion of CD4+ (which contain all Tregs) and CD8+ T cells revealed not only that conventional CD4+ T cells substantially contributed to the antitumor immunity induced by combination therapy but also that depletion of CD4+ Tregs led to severe autoimmune phenotypes in more than half of the mice receiving combination therapy (63, 64). Several recent studies, including ours, have described a partial tumor-specific reduction of Tregs. The latter is mediated by multiple mechanisms, including the depleting activity of the anti–CTLA-4 antibody clone 9D9 [immunoglobulin G2b (IgG2b)], which has been used throughout this study (21, 23, 65, 66). However, Treg depletion may be affected by several factors, such as mouse strain/substrain, tumor type, and model. In contrast to the subcutaneous B16 and MC-38 (C57BL/6) tumor models, Treg depletion could not be reproduced in the Balb/c-based CT-26 tumor model using the original 9D9 clone (65, 66). The Fo5 tumor model used in this study is based on orthotopic implantation of tumor pieces, derived from established tumors, whereas most of the published work analyzes anti–CTLA-4–mediated Treg depletion in subcutaneous tumor cell line–based models. Highlighting the contribution of tumor-specific factors, recent studies by Simpson et al. (66) and Romano et al. (67) indicate that only a specific subset of tumor resident macrophages in melanoma patients was able to mediate CTLA-4–dependent ADCC of Tregs. Accordingly, the B16 melanoma model seems to be among the tumor models most permissive to anti–CTLA-4 therapy and therapy-induced Treg depletion (66). Only limited information is available for other tumor types and particularly orthotopic tumor models, such as the breast cancer model used in our study. These findings will therefore require further in-depth validation in future studies. Note that the overall response rates of tremelimumab (human IgG2, no ADCC) and ipilimumab (IgG1, ADCC) were comparable in clinical phase 3 trials, although tremelimumab seems to solely mediate an increase in effector T cells with no or a weak increase in Tregs, whereas ipilimumab seems to also reduce Tregs in clinical responders (65, 67, 68). It can therefore be speculated that an ADCC-optimized anti–CTLA-4 antibody could be even more efficacious in combination with T-DM1, if toxicity remains manageable.

Our findings underline the immune-promoting properties of T-DM1 therapy. We report the striking observation that a therapeutic maneuver with T-DM1 can transform large established HER2-positive tumors, which are resistant to CTLA-4/PD-1 blockade, to tumors which are highly vulnerable to immune attack. Ultimately, this resulted in complete cures in our preclinical model, highlighted by skewing of treated tumors toward a T cell–inflamed phenotype with tumor-infiltrating T cells, a broadly increased chemokine profile, and a type I IFN signature (69). Mechanistically, the therapeutic interaction with immune checkpoint inhibition is explained by increasing the number and density of T cells along with the activation of targets for T cell–directed therapies upon immune activation. Simultaneously, T-DM1 reduces the number of tumor cells, thereby minimizing the chronic exposure of T cells to tumor-derived antigens, which may render them nonfunctional (70). We therefore propose that these therapies function in a nonredundant but complementary manner: the ADC with its cytotoxic payload reinstates immunosurveillance by driving a distinct DC maturation program and, subsequently, T cell infiltration, whereas CTLA-4/PD-1–blocking therapy potentially reinvigorates exhausted T cells. Additional studies will be required to elucidate the extent to which this specific therapeutic situation can be compared with other studies using CTLA-4–permissive tumor models. In addition, because HER2 is a self-antigen and its sequence is typically unmodified in human cancer, immune tolerance may differentially modulate the immune response after T-DM1 treatment in our preclinical model. In addition, we provide clear evidence that Tregs protect the host from potent immune responses induced by the combination therapy. Our data therefore reveal a mechanism of action for T-DM1 and have direct implications for clinical efforts combining ADCs such as T-DM1 with checkpoint inhibition.


Study design

The objective of the in vivo efficacy studies was to evaluate the therapeutic and immunological activity of T-DM1 alone and in combination with CTLA-4/PD1–blocking antibodies. These treatments were evaluated in the human HER2–expressing Fo5 tumor model. Sample sizes (n = 6 per group for survival analysis and 4 to 6 mice per group for FACS analysis) were determined on the basis of homogeneity and consistency of tumor growth characteristics in the selected model and were sufficient to evaluate treatment efficacy. Data from all performed survival experiments were pooled. Mice were euthanized (considered “dead”) when tumors reached a volume of 1200 mm3. Animals were randomly assigned to groups on the basis of tumor size. All animals were included in the survival analysis.

The objective of the patient study was to compare TILs from matched tumor biopsies (pretherapeutic versus on-treatment) from 28 treatment-naïve patients diagnosed with HER2-positive/ER-positive breast cancer who underwent a single treatment of T-DM1 monotherapy within a subtrial of the WSG-ADAPT protocol (NCT01745965). After initial diagnosis (pretherapeutic), a second core biopsy was performed after 3 weeks of induction therapy (T-DM1–mono). TILs dispersed in the stroma between the carcinoma cells (stromal TILs) were assessed independently by two trained histopathologists, according to current histomorphological consensus guidelines.


FVB mice were obtained from Janvier Labs or bred in the animal facility of the Department of Biomedicine, University of Basel, Switzerland. All animals were housed under specific pathogen–free conditions and in accordance with Swiss federal regulations. Perpendicular tumor diameters were measured with calipers, and tumor volume was calculated according to the following formula: Tumor volume (mm3) = d2*D/2, where D and d are the longest and shortest diameters of the tumor (in millimeters), respectively.

In vivo tumor models and treatments

Female FVB mice (8 to 10 weeks old) were orthotopically transplanted with pieces of Fo5 (MMTV–human HER2) breast tumors (1 to 2 mm in diameter) into one mammary gland (40). Once tumors reached an average volume of 80 mm3 (day 0), mice were treated with T-DM1 (15 mg/kg) and/or anti-mouse PD-1 (10 mg/kg) (RMP1-14; rat IgG2a; Bio X Cell) and anti-mouse CTLA-4 (10 mg/kg) (9D9; mouse IgG2b; Bio X Cell) as indicated in the figures and legends. Tumor volume was measured three times per week as described above. T-DM1 was given once (day 0) or twice (day 0 and 14). For survival experiments, anti–CTLA-4/PD-1 was given as monotherapy on days 0, 2, 4, 7, and 10, for 1× T-DM1 on days 7, 9, 11, 14, and 17, or for 2× T-DM1 on days 21, 23, 25, 28, and 31. For FACS analysis, T-DM1 was given once on day 0 and anti–CTLA-4/PD-1 either as monotherapy on days 0, 2, 4, and 7 or in combination with T-DM1 (one dose) on days 2, 4, 6, and 9. Trastuzumab was administered on the same schedule as T-DM1. Tumors were analyzed by FACS on day 11. For T cell depletion, anti-CD4 (GK1.5; rat IgG2b, Bio X Cell) and anti-CD8 (53-6.72; rat IgG2a, Bio X Cell) depleting antibodies (10 mg/kg) were given at day −2 and day 0 and then weekly until the end of the experiment or until autoimmune symptoms occurred. Mice were euthanized when tumors reached a volume of 1200 mm3. All mice included in the figures (survival analysis) reached this tumor size. All surviving mice reflecting the tail of the curve remained tumor-free until completion of the experiments and euthanization. Cohort sizes were typically six mice per group for survival analysis and four to six mice per group for FACS analysis. Independent data sets were used for the survival analyses in Figs. 2 and 6.

Cell lines, culture, and DC activation

The immature mouse DC cell line SP37A3 (provided by Merck KGaA) was cultured in Iscove’s modified Dulbecco’s medium (IMDM) (Sigma) supplemented with 10% heat-inactivated and endotoxin-tested fetal bovine serum (PAA), sodium pyruvate (Gibco), penicillin/streptomycin/l-glutamine mix (Gibco), minimum essential medium (MEM) nonessential amino acids (Sigma), ciprofloxacin (Bayer), and 0.05 mM 2-mercaptoethanol (Gibco). IMDM complete medium was supplemented with recombinant mouse granulocyte-macrophage colony-stimulating factor (GM-CSF; 20 ng/ml) and recombinant mouse M-CSF (20 ng/ml) (both from PeproTech). SP37A3 DCs were plated in 96-well plates (8 × 104 cells per well) and activated with cytotoxic compounds at the indicated concentrations or lipopolysaccharide (500 ng/ml) as positive control (Sigma; Escherichia coli 026:B6). After 24 hours of incubation, the DC phenotype was assessed by flow cytometry. Dead cells were identified by SYTOX Green (Invitrogen) staining and excluded from the analysis. Cytokines (IL-1β, IL-6, and IL-12p40) in supernatants of DC cultures were detected by standard sandwich enzyme-linked immunosorbent assay (ELISA) procedures using commercially available kits (BD/eBioscience). Ansamitocin P3 was provided by the National Cancer Institute (NCI, Bethesda, MD). DM1 was purchased from Concortis Biosystems.

Immunofluorescence staining of Fo5 tumors

Tumors were embedded in Tissue-Tek optimal cutting temperature compound (OCT) using standard protocols and stored at −70°C. The organs were cut with a cryotome into 5-μm-thick sections. The sections were air-dried for 15 min and fixed for 5 min in ice-cold acetone. Then, the slides were dried for 5 min and stored at −70°C until further use. For immunofluorescence staining, the sections were retrieved from the −70°C freezer and incubated for 5 min at room temperature (RT). Slides were washed for 5 min in 1× phosphate-buffered saline (PBS) to remove the OCT and then air-dried for 5 min and circled with a DAKO Pen. The slides were then washed for 5 min in 1× PBS. Eighty microliters of blocking solution [20% bovine serum albumin (BSA) in PBS] was added to each section, and the sections were incubated for 2 hours at RT in a dark and humid incubation chamber. Subsequently, the slides were washed three times for 5 min in 1× PBS. The slides were air-dried for 3 min, and the primary antibodies (α-CD4, α-CD8, and α-HER2 diluted in 3% BSA in 1× PBS) were added and incubated at 4°C overnight in a dark and humid incubation chamber. The slides were subsequently washed three times for 5 min in 1× PBS and air-dried for 3 min, and the secondary antibodies [Alexa Fluor 488, phycoerythrin (PE)] were added for 2 hours at RT in a dark and humid incubation chamber. Next, the slides were washed three times for 5 min in 1× PBS and embedded using DAKO mounting medium. Slides were stored in the dark (4°C) until further use. We used the following antibodies: α-CD4 from Bio-Rad (MCA4635GA), α-CD8 from BD Biosciences (558733), α-HER2 from Cell Signaling (2165S), Alexa Fluor 488 goat α-rat IgG from Life Technology (A11006), and PE goat α-rabbit IgG (H+L) from Invitrogen (A10542).

Tumor digestion

On day 11 after treatment initiation, mice were euthanized, and tumors were mechanically dissociated and digested with accutase (PAA), collagenase IV (Worthington), hyaluronidase (Sigma), and deoxyribonuclease type IV (Sigma). Single-cell suspensions were prepared and stained against the indicated markers for flow cytometric analysis.

mAbs and flow cytometry

The antibodies used were α-arginase (BD) and the corresponding isotype control (mouse IgG1) and anti-mouse IgG1–PE (Southern Biotech), α-CD3–PE (BD), α-CD3–PE–Cy7 (BD), α-CD4–BV421 (BD), α-CD4–PE–Cy7 (BD), α-CD8–BV605 (BD), α-CD11b–FITC (BioLegend), α-CD11c–FITC (BioLegend), α-CD11c–APC (BD), α-CD19–PE (BD), CD19–PE–Cy7 (BD), α-CD25–BV421 (BD), α-CD25–Alexa Fluor 647 (BD), α-CD45–PerCP (BD), α-CTLA-4–PE (BD), α-EOMES–PE (eBioscience), α-F4/80–BV421 (BioLegend), α-FOXP3–PE (eBioscience), α-FOXP3–Alexa Fluor 647 (eBioscience), α-granzyme B–Alexa Fluor 647 (BD), α-IFN-γ–APC (eBioscience), α-Ki-67–BV421 (BD), α-Ly-6C–BV605 (BD), α-Ly-6G–PE–Cy7 (BD), α-NK1.1–BV421 (BD), α-PD-1–BV421 (BD), α-PD-1–APC (BD), α-PD-L1–BV421 (BD), α-PD-L1–PE (BD), α-PD-L1–APC (BD), α-Tbet–Alexa Fluor 647 (BD), α-TCRγ/δ–PE (BioLegend), α-TCRγ/δ–APC (BioLegend), and α-TIM-3–PE (eBioscience). The live/dead fixable near-infrared dye (Invitrogen) was used to exclude dead cells. The FOXP3 buffer kit from eBioscience was used for CTLA-4, FOXP3, EOMES, Ki-67, and Tbet staining. Intracellular fixation and permeabilization buffers from eBioscience were used for arginase and cytokine staining. The eBioscience cell stimulation cocktail (PMA/ionomycin), containing protein secretion blockers, was used for 4 hours to restimulate T cells.

Regulatory T cell suppression assay

Cells were stained in magnetic cell sorting (MACS) buffer with α-CD4–PE and α-CD25–APC for Tregs and α-CD8–Alexa Fluor 488 for effector T cells. The staining was carried out for 45 min on ice in the dark. Then, the cells were washed twice with MACS buffer and sorted with a BD Influx cell sorter for CD4+/CD25+ Tregs. CD8+ effector cells were sorted from naïve FVB spleens. Before cell sorting, a 96-well U-bottomed cell culture plate was coated with anti-CD3 and anti-CD28 (10 μg/ml) and incubated overnight at 4°C or for 2 hours at 37°C in PBS. Sorted effector cells were stained with eFluor 670 according to the manufacturer’s instructions. Cells were counted, resuspended in an appropriate volume of medium, and incubated at 37°C, 5% CO2 for 3 to 4 days. Subsequently, the cells were stained using α-CD4–PE, α-CD25–APC, and α-CD8–Alexa Fluor 488 and analyzed by FACS.

Protein extraction

Protein extraction buffer consisted of the following components: Tissue Extraction Reagent I (Invitrogen, cat. no. FNN0071) and Protease Inhibitor Mix [Halt Protease and Phosphatase Inhibitor Cocktail, EDTA-free (100×) (Thermo Scientific, cat. no. 78441)]. Protein concentration was determined with Pierce BCA Protein Assay Kit (Thermo Scientific, cat. no. 23225). Mixer Mill MM301 (Retsch) was used for mechanical disruption of frozen tumors. Ten milliliters of Tissue Extraction Reagent I with 1× Halt Protease and Phosphatase Inhibitor Cocktail was added per 1 g of tumor tissue. Tumors were homogenized for 2 min using the Mixer Mill with grinding balls and left for 30 min on ice with shaking every 5 min. Samples were centrifuged at 20,000g for 15 min at 4°C to pellet tissue debris. Supernatant was stored at −80°C.

Luminex and ELISA assays

Tumor tissue lysates were analyzed using the mouse cytokine group I (M60-009RDPD) and group II (MD0-00000EL) Bio-Plex Pro kits (Bio-Rad). Briefly, lysate samples were normalized to a total protein concentration of 9.2 mg/ml, using tissue lysate buffer. Normalized samples were further diluted to 1.2 mg/ml in Bio-Rad Sample Diluent with 0.5% BSA (SB) before analysis. Standards were reconstituted in 13% tissue extraction reagent and 87% SB to match the buffer composition of the analyzed samples. Reconstitution volumes and pooling of standards followed the Bio-Rad guidelines for cross-panel plexing for broad-range cytokine standard curves with an additional two low-concentration dilutions for a total of 10 standard dilutions plus two blank wells (reconstitution buffer). Standards (triplicate) and samples (duplicate) were run at 100 μl per well and incubated for 1.5 hours. Wash steps were performed using a Tecan HydroSpeed with the recommended magnetic bead wash protocol (MAG_96) using PBS with 0.05% Tween 20. All other steps and conditions followed the Bio-Rad manual for the kits. Samples were then analyzed on a Bio-Plex MAGPIX (Luminex/Bio-Rad) following the manufacturer’s guidelines. Absolute cytokine concentrations were determined using 5-PL curve fits in the Bio-Plex Manager Software (Bio-Rad). IFN-γ was measured using standard ELISA (eBioscience).

Patient samples and TILs

A total of 56 breast needle core biopsies (CNBs) corresponding to 28 patients diagnosed with HER2-positive and ER/PR-positive early breast cancer were retrieved from the central reference pathology tissue archive of the WSG-ADAPT trial (26, 71). All patients were treatment-naïve and underwent T-DM1 monotherapy for a total of 12 weeks before surgery. A pretherapeutic CNB and a second CNB obtained after 3 weeks of T-DM1 monotherapy were evaluated for the percentage of stromal TILs as recently recommended by an international working group (27). In brief, the percentage of stromal TILs was estimated over the entire histological section in steps of 5% by trained histopathologists. For cases with very few lymphocytes (covering <5% of the stromal area), TILs were estimated in steps of 1%. Patches with lymphocytic lobulitis and inflamed carcinoma in situ were excluded (27). The definitive percentage of stromal TILs was calculated as the mean of the percentages determined by two observers, and results were handled as metric data. The lymphocyte-predominant breast cancer type was defined as stromal TILs ≥40% (27). The Bland-Altman method and the Wilcoxon matched pairs test were used to assess changes in the quantity of stromal TILs before versus on-treatment. The histological regression grade was determined as described by Sinn et al. (30). Representative cases were subjected to immunohistochemical staining for CD4, CD8, and granzyme B using a Benchmark Ultra automated stainer (Ventana) and the antibodies SP35 (Zytomed Systems), C8/144B (DakoCytomation), and 760-4283 (Cell Marque). Representative scanned full sections of CNBs can be viewed by virtual microscopy at Please contact the corresponding author for the database password.


If not stated otherwise in the figure legend, samples were analyzed with GraphPad Prism software using Dunn’s multiple comparison test. Scatter dot plots are depicted as means with SEM. Whiskers indicate minimum and maximum.


Fig. S1. Effect of ansamitocin P3 and DM1 on the maturation of DCs.

Fig. S2. CD8/CD4 ratios in human breast tumors before and after T-DM1 monotherapy.

Fig. S3. H&E staining of treated Fo5 tumors.

Fig. S4. Patterns of intratumoral immune cell subsets after T-DM1 therapy.

Fig. S5. Tumor growth and tumor-infiltrating immune cell subsets in mice receiving trastuzumab instead of T-DM1.

Fig. S6. Luminex-based cytokine and growth factor measurement in T-DM1– and α-CTLA-4/PD-1–treated tumors.

Fig. S7. T cell function in mice receiving trastuzumab instead of T-DM1.

Fig. S8. Helios expression, CD8/CD4 and CD8/Treg ratios, and splenic Tregs.

Fig. S9. Helios expression and Tregs in mice receiving trastuzumab instead of T-DM1.


Acknowledgments: We thank M. Buchi and P. Herzig for excellent technical assistance with the mouse experiments and tumor digestions/stainings as well as tumor lysate preparations, respectively. A. Sutter (Merck KGaA) provided the SP37A3 cell line. Ansamitocin P3 was provided by the Developmental Therapeutics Program of the NCI/NIH. Funding: This work was supported by the Wilhelm Sander-Stiftung für Krebsforschung, a Swiss Group for Clinical Cancer Research (SAKK)/Amgen research grant, the Swiss National Science Foundation, the Krebsliga beider Basel, the Sassella-Stiftung, the Huggenberger-Bischoff Stiftung zur Krebsforschung, the Freiwillige Akademische Gesellschaft Basel, and the Whitaker International Scholars Program. The WSG trial was sponsored by Roche. Author contributions: P.M. and A.Z. developed the hypothesis, designed the experiments, conceived and supervised the project, coordinated the study, and wrote the manuscript as well as acquired funding. P.M., M.K., T.K., K.M., K.G., S.S., and M.C. developed the methodology, performed and analyzed experiments, and contributed to data interpretation. P.M designed all figures. D.S.T., H.K., N.H., M.v.B.-B., contributed to experimental design and revision of the manuscript. U.N., O.G., and S.R. contributed to the experimental design. Competing interests: N.H. has served on advisory boards and as a consultant for Roche/Genentech. U.N. has served on the advisory board for T-DM1. H.K. has participated in advisory boards for Roche, AstraZeneca, and Novartis. All other authors declare that they have no competing interests. Data and materials availability: The Fo5 tumor model was provided by Genentech under a material transfer agreement.

Stay Connected to Science Translational Medicine

Navigate This Article