IFNγ-induced Chemokines Are Required for CXCR3-mediated T-Cell Recruitment and Antitumor Efficacy of Anti-HER2/CD3 Bispecific Antibody

Therapies that direct T cells to tumors, including adoptive transfer of genetically engineered T cells and T-cell–dependent bispecific (TDB) antibodies, have been clinically validated for the treatment of B-cell leukemias as well as lymphomas, (1, 2) and have demonstrated promising activity for the treatment of multiple myeloma (3, 4). In addition to hematologic indications, several TDBs targeting solid tumors are currently in clinical trials. We have previously generated anti-HER2/CD3 TDB for the treatment HER2-positive cancers (5). Anti-HER2/CD3 TDB redirected T cells to eradicate HER2-expressing cancer cells and exhibited potent antitumor activity in preclinical mouse models of HER2-overexpressing tumors.

CD3-bispecific antibodies bridge T cells to tumor cells by binding the T-cell receptor (TCR) CD3 subunit and a tumor antigen expressed on cancer cells, thereby facilitating T-cell–mediated killing of target cells. Simultaneous binding of the bispecific antibody to T cells and tumor targets leads to rapid formation of an immunologic synapse by induction of TCR clustering and exclusion of a key negative regulator of TCR activity (CD45) from the synapse (6). Because these molecules mediate T-cell activation via binding to CD3 rather than TCR and peptide/MHC interactions, CD3-bispecific antibodies are able to redirect a polyclonal T-cell response to tumor cells independent from tumor-antigen specificity of the T cell. Several reports have demonstrated that treatment of preclinical tumor models with CD3-bispecific antibodies leads to an increase in tumor-infiltrating T cells (1, 7–10). The mechanism(s) underlying increased T-cell accumulation has yet to be described.

A clear correlation is emerging between response to immunotherapy and immunogenicity of the tumor. Moreover, analysis of tumor biopsies from clinical trials has revealed a correlation among clinical response to immunotherapy with immune cell profiles of tumors. A significant correlation has been observed between the tumor mutational burden (number of mutations per coding sequence in tumor DNA) and the objective response rate to anti–PD-1 therapy within multiple indications (11). Similarly, patients with mismatch repair deficiency, resulting in high number of somatic mutations, demonstrate exceptionally high response rates to anti–PD-1 treatment across various tumor indications (12). Response rates for PD1/PD-L1 monotherapy blockade range from 10% to 40% and responses to anti-CTLA treatment are rare (13–15). Therefore, there is an urgent medical need in oncology to extend clinical benefit of immunotherapies for the majority of patients that respond poorly by augmenting tumor inflammation. Because of their unique properties, CD3-bispecific antibodies could potentially increase tumor T-cell infiltration within patients and indications that are predicted to respond poorly to checkpoint inhibitors. Several clinical trials have demonstrated a positive association between tumor-infiltrating lymphocytes, immune activation, and CD8⁺ T cells in HER2-positive breast cancer (16–19). Inhibitors for immune checkpoints such as PD-1/PD-L1 are currently being tested in clinical trials for HER2-positive breast cancer. On the basis of the low frequency of mismatch repair deficiency and median mutation load detected in breast cancer (11, 20), this indication may not be optimal for checkpoint blockade.

In this study, we demonstrate that anti-HER2/CD3 TDB was highly efficacious in the treatment of mammary tumors with low baseline T-cell content. TDB treatment induced robust inflammation thereby converting tumors to an inflamed phenotype. We characterize the mechanisms of TDB-induced T-cell accumulation and demonstrate that it was essential for antitumor activity. On the basis of robust preclinical activity, anti-HER2/CD3 TDB demonstrates promise as a single agent. In addition, due to lack of activity for immunotherapies in poorly inflamed tumors, TDB strategies may be leveraged to augment tumor inflammation, thereby conferring clinical benefit to checkpoint inhibitors regardless of baseline immune status.

Bispecific IgG TDB assembly from half antibodies was performed as described previously (5, 21). Human IgG1 TDBs included N297G substitution to attenuate the Fc-mediated effector functions. Fc modification in murine IgG2A TDBs was L234A, L235A, P329G (LALA-PG). For muIgG2a anti-HER2/CD3 TDB (5), the “knob” arm is murine anti-HER2 4D5 (22) and the “hole” is chimeric anti-murine CD3 2C11 (23). In human IgG1 version of anti-HER2/CD3 TDB (4D5/40G5), the “knob” arm is human anti-HER2 4D5 (22) and the “hole” is human anti-human CD3, clone 40G5 (24).

MMTV.huHER2.FVB/n transgenic female mice have been described previously (25) and were maintained on a FVB/n background strain. The double transgenic mouse model expressing human HER2 and human CD3e (26) were maintained on a FVB/n and Balb/c mixed background. The animals were dosed and monitored according to the guidelines from the Institutional Animal Care and Use Committee at Genentech, Inc. Anti-HER2/CD3-TDB was administered intravenously as described in the figures. For CD4⁺ T-cell depletion, groups were treated with anti-CD4.GK.1.5TI.Rat.IgG2b (Genentech Inc.) at 500 μg per mouse, via intraperitoneal injection, every 4–5 days for a total of 3 doses. For FTY720 (Sigma SML0700) studies, FTY720 was administered at 0.5 mg/kg, three times a week, via intravenous injection, for one week.

A 1-cm incision was made in the skin just rostral to the third mammary fat pad on female FVB WT mice. A pocket for the tumor was made into the no. 2/3 mammary fat pad and a 2 × 2 mm MMTV-HER2-transgenic Founder #5 (Fo5) tumor section (27) was placed into the pocket. The skin was closed using wound clips. Wound clips were removed at 7–10 days postsurgery and mice were monitored for appearance of palpable tumors. When tumor volumes grew to an average of approximately 150 mm³, they were placed into treatment cohorts with an equivalent tumor volume average size. Mice were treated as indicated. Anti-IFNγ (clone XMG1.2) and anti-CXCR3 (clone CXCR3-173) antibodies (BioXcell) were diluted in PBS and administered to mice intraperitoneally at 500 μg per injection at 1 day prior and days 1, 6, 8, and 13 following anti-HER2/CD3 TDB treatment.

IHC was performed on 4-μm thick formalin-fixed, paraffin-embedded tissue sections mounted on glass slides. Primary antibodies against CD3, clone SP7 (RM-9107-S, Thermo Fisher Scientific), and HER2, clone 4B5 (790-2991, Ventana Medical Systems Inc.) were used. CD3 staining was performed on the DAKO autostainer, utilizing Target pH6 (Dako) antigen retrieval. Sections were incubated with 1:200 dilution of primary antibody followed by biotinylated goat anti-rabbit secondary (Jackson ImmunoResearch Laboratories), and detected with ABC-HRP (Vector Laboratories). Slides were visualized with DAB (Pierce), counterstained, dehydrated, and coverslipped for viewing. HER2 staining was carried out on the Ventana Discovery XT automated platform (Ventana Medical Systems) according to the manufacturer's recommendations.

Processing of MMTV-huHER2 tumors for flow cytometry was done using Human Tumor Dissociation Kit (Miltenyi Biotec) according to manufacturer's instruction. Briefly, tumors were cut into small pieces of 2–4 mm and transferred to gentleMACS C-tubes (Miltenyi Biotec) in 5 mL of fresh-made ice-cold RPMI medium with enzyme mix. C-tubes were placed onto gentleMACS Octo Dissociator with Heaters. GentleMACS program 37C_m_TDK_2 was used to dissociate tumors. Samples were filtered through a 70-μm cell strainer and spun down at 1,300 rpm for 5 minutes. Single-cell suspensions were prepared in FACS buffer (PBS, 2% FBS and 1 mmol/L EDTA) and stained against the indicated markers for flow cytometric analysis. Fo5 tumors were explanted, weighed, coarsely mechanically digested, and transferred into gentleMACS C-tubes (Miltenyi Biotec) containing digestion buffer containing DMEM-high glucose supplemented with 5% heat-inactivated FBS, HEPES (10 mmol/L), Collagenase-D (2 mg/m), DNase (40 U/mL), and trypsin inhibitor (1 mg/mL). Tumor samples were then dissociated in digestion buffer using the gentleMACS Dissociators (Mouse_tumor_3, Miltenyi Biotec) then incubated at 37°C for 45 minutes at 180 rpm rotation. Tumors were then mechanically dissociated into single suspension using gentleMACS Dissociators (Mouse_tumor-_1, Miltenyi Biotec). Digested tumors were passed through a 100-μm strainer and centrifuged at 1,450 rpm for 10 minutes at 4°C. Red blood cell lysis was performed with 5 mL of ACK lysis buffer and lysis was subsequently quenched with 10 mL of DMEM and centrifuged. Cells were then resuspended at 1 mL per 0.25 g of original tumor mass/mL of FACS buffer for further antibody staining.

Mouse peripheral blood was collected in tubes coated with EDTA (BD Biosciences). Red blood cells were lysed using ACK lysis buffer (Thermo Fisher Scientific) twice. Remaining cells were resuspended in FACS buffer. Small aliquots were taken for cell counting using beads and flow cytometry. The rest of the samples were stained with the indicated markers for flow cytometry analysis.

All antibodies for flow cytometry cell staining were purchased from BD Biosciences or BioLegend unless indicated otherwise. Ki67 clone SolA15 was from Thermo Fisher Scientific and FoxP3 Transcription Factor Staining Buffer Kit was purchased from Invitrogen.

The live/dead fixable near-infrared dye (Invitrogen) was used to exclude dead cells. The FOXP3 buffer kit (eBioscience) was used for FOXP3 and IFNγ staining. Intracellular fixation and permeabilization buffers from eBioscience were used for cytokine staining. For IFNγ intracellular staining, the BioLegend cell activation cocktail with Brefeldin A was used for 4 hours to restimulate T cells prior staining.

Mouse blood was collected in serum separator tubes (BD Biosciences). The tubes with samples were centrifuged at 10,000 rpm for 10 minutes. The sera were transferred into a clean 96 V-bottom plate and stored in −80°C freezer until they were ready for analysis.

Mouse serum or tumor tissue lysates were analyzed using the mouse cytokine group I 23-Plex Bio-Plex Pro (Bio-Rad) or Premix Panel I 32-Plex (Millipore). Briefly, lysate samples were normalized to a total protein concentration of 20 mg/mL, using tissue lysate buffer. Normalized samples were further diluted in 1:4 in Bio-Rad Sample Diluent with 0.5% BSA (SB) before analysis. Assays were performed according to Bio-Plex Pro Cytokine, Chemokine, and Growth Factor Assay Instruction Manual. Standards were made by resuspending lyophilized cytokine standard with 500 μL of complete media. Dilutions were made to create 10 standard points with a dilution factor of 2.27. Fifty microliters of coupled beads were added into a V-bottom 96-well plate. The beads were washed twice with the wash buffer before adding the 50 μL of blank, standards, controls, and unknown samples. The plate was allowed to incubate at room temperature for 30 minutes on a shaker at 850 revolutions per minute (rpm). After the 30-minute incubation, the plate was washed three times with wash buffer. Twenty-five microliters of detection antibodies were added to the wells and incubated for another 30 minutes at room temperature on a shaker at 850 rpm. After 30 minutes, the plate was washed again three times with wash buffer. After the wash, 25 μL of streptavidin-PE were added and incubated for 10 minutes at room temperature. After 10 minutes, the plate was washed three times then was resuspended in 125 μL of assay buffer. The plate was read on the Bio-Plex 200 system and analyzed with the Bio-Plex Manager Software (Bio-Rad Laboratories). CXCL9 and CXCL11 were measured using standard ELISA (R&D Systems) according to manufacturer's instruction. Serum and tumor from Fo5 tumor–bearing mice were harvested 1 day following treatment. Tumors were frozen and pulverized into a fine powder while maintaining their frozen state with mortar and pestle. Approximately 0.2 mg of tumor powder was placed into cell lysis buffer (Cell Signaling Technology) and homogenized using Lysing Matrix D (MPBio). Supernatants were clarified by centrifugation and measured for protein content using the BCA assay (Pierce). Supernatant total protein content was normalized between samples by dilution. Serum and tumor supernatant were subjected to cytokine analysis using an immunology multiplex assay (EMD Millipore). For Fo5 tumor supernatants, cytokine content was normalized to total protein content within inputted samples.

Splenic and lymph node T cells from female FVB mice were magnetically separated. Isolated T cells were incubated with 2.0 μmol/L CellTrace Violet (Molecular Probes C34557) for 20 minutes in PBS at 37°C. A total of 2 × 10⁷ T cells in saline were injected to Fo5 tumor–bearing female FVB mice via tail vein 1 day prior to treatment. Mice were treated intravenously with 12.5 μg of anti-HER2 TDB or control TDB (anti-gD/CD3) diluted in PBS 4 days prior to tumor harvest.

MMTV-huHER2 transgeneic (TG) mice are immunocompetent mice that express the human HER2 transgene under the control of the murine mammary tumor virus (MMTV) promoter. Expression of HER2 within mammary epithelium leads to spontaneous tumor development (25). We have previously shown that anti-HER2/CD3 TDB treatment resulted in tumor regression in this and other mouse models (5). To better understand the mechanism of antitumor activity, we investigated anti-HER2/CD3 TDB-induced immune responses using the HER2-MMTV model. T-cell dynamics were analyzed using anti-CD3 IHC to detect tumor-infiltrating T cells. CD3 staining was sparse in control-treated tumors (Fig. 1A). In sharp contrast, anti-HER2/CD3 TDB-treated tumors demonstrated an intense CD3 signal indicative of robust TDB-mediated accumulation of intra-tumoral T cells. Anti-HER2 staining confirmed comparable, high (IHC 3+) HER2 expression on tumor cells between control and anti-HER2/CD3 TDB-treated tumors. MMTV-huHER2 tumors treated with anti-HER2/CD3 showed markedly increased CD3⁺ lymphocytes both within the tumor and at the tumor–normal border compared with vehicle-treated tumors. Tumors from anti-HER2/CD3-treated mice showed no changes in tumor architecture, stromal composition/degree of fibrosis, acute inflammation or necrosis by hematoxylin and eosin compared with vehicle-treated mice. As no significant changes in the tumor and tumor microenvironment were histologically observable, further IHC or other studies would be necessary to characterize this fully.

To confirm IHC findings and further clarify TDB-induced alterations in immune contexture, we analyzed tumors by flow cytometry. The most notable immunologic effect of anti-HER2/CD3 TDB treatment (6 days) was an increase of tumor-infiltrating CD45⁺ lymphocytes and CD8⁺ T cells (Fig. 1B). Nearly all of the tumor-infiltrating CD8⁺ T cells were effector or effector memory subtype 6 days after anti-HER2/CD3 TDB treatment (Supplementary Fig. S1).

TDB treatment induced a significant increase in IFNγ⁺ CD8⁺ T cells (Fig. 1B), indicating that anti-HER2/CD3 TDB not only enriched for intratumoral CD8⁺ T cells, but also enhanced CD8⁺ T-cell activation. Increased PD-1⁺ expression on CD8⁺ T cells (Fig. 1B), an inhibitory receptor that is upregulated in response to T-cell activation, further supports that anti-HER2/CD3 TDB increased the population of activated T cells in tumors. PD-1 upregulation by anti-HER2/CD3 TDB treatment is consistent with our previous in vitro findings (5) in human peripheral blood mononuclear cells (PBMC).

Although TDBs can bind and activate CD4⁺ T cells, anti-HER2/CD3 TDB did not substantially impact the frequency of tumor-infiltrating CD4⁺ T cells (Fig. 1B; Supplementary Fig. S1). Almost all CD4⁺ T cells were PD-1⁺ suggesting that anti-HER2/CD3 TDB treatment engaged and activated CD4⁺ cells in tumors. Anti-HER2/CD3 TDB treatment clearly impacted CD4⁺ T-cell subtypes by increasing the fraction of T regulatory (Treg) cells, characterized as CD25⁺FoxP3⁺ CD4⁺ T cells (Fig. 1B). Hypothetically, increased Treg frequency can be due to the conversion of CD4⁺ T cells to a Treg phenotype or caused by an influx of Treg cells from the periphery. Increased Treg proportions, coupled with elevated PD-1 expression, may illustrate potential immunosuppressive feedback mechanisms downstream of immune cell activation by anti-HER2/CD3 TDB.

To confirm these key findings of anti–HER2/CD3-TDB-induced immune responses, we utilized an orthotopic, syngeneic mammary tumor transplant model (Fo5; ref. 27). In agreement with our observations in transgenic MMTV-huHER2 model (Fig. 1), anti-HER2/CD3 TDB administration inhibited tumor growth and induced a similar immune response in the Fo5 breast cancer model (Supplementary Fig. S2).

In summary, these results demonstrate that anti-HER2/CD3 TDB treatment robustly increased activated T cells in HER2-positive mammary tumors that are poorly infiltrated at steady state. Taken together, treatment with anti-HER2/CD3 TDB converted noninflamed mammary tumors to an immune-inflamed phenotype.

Although CD8⁺ T cells exhibited higher in vitro potency, both CD4⁺ and CD8⁺ T cells, can be activated by CD3-bispecific antibodies in vitro (6, 28), as well as, in vivo (Fig. 1B) and both subtypes are capable of killing tumor cells in vitro. The contrasting responses of CD8⁺ and CD4⁺ T cells to anti-HER2/CD3 TDB in MMTV-huHER2 tumors suggested that CD8⁺ T cells are likely critical mediators of TDB-induced efficacy in vivo. To assess the contribution of CD4⁺ T cells to the antitumor response of anti-HER2/CD3 TDB therapy, we depleted CD4⁺ T cells from MMTV-huHER2 TG mice prior to treatment (Fig. 2A–C). Anti-HER2/CD3 TDB treatment achieved comparable antitumor effects irrespective of CD4⁺ T-cell presence, indicating that CD4⁺ T cells were dispensable for antitumor efficacy of anti-HER2/CD3 TDB treatment (Fig. 2D and E).

Anti-HER2/CD3 TDB treatment has previously been shown to induce T-cell proliferation and expansion in vitro (5), which may account for increased intratumoral T-cell numbers. However, increased CD8⁺ T-cell accumulation may also be mediated by active recruitment to tumors. To test the contribution of T-cell recruitment in vivo, we used FTY720 (Fingolimod), a sphingosine 1-phosphate (S1P) receptor agonist that inhibits S1P-dependent lymphocyte egress from secondary lymphoid organs and blocks trafficking of lymphocytes between blood and tissue (29). Mice were pretreated with FTY720 prior to anti-HER2/CD3 TDB treatment, which resulted in the expected reduction of peripheral blood CD45⁺ lymphocytes, including CD8⁺ T cells (Fig. 3A). Combination treatment with anti-HER2/CD3 TDB and FTY720 abolished anti-HER2/CD3 TDB-mediated accumulation of CD45⁺ and CD8⁺ T cells within tumors, indicating that the recruitment of T cells from the periphery was the major mechanism of T-cell accumulation in MMTV-huHER2 tumors in response to anti-HER2/CD3 TDB administration (Fig. 3B).

To determine the contribution of CD8⁺ T-cell proliferation in response to anti-HER2/CD3 therapy, CellTrace Violet (CTV)-labeled splenic and lymph node T cells from naïve FVB female mice were adoptively transferred into Fo5 tumor–bearing mice 1 day prior to treatment. Labeled CD8⁺ T cells were detected in tumors treated with control TDB, suggesting a low level of passive baseline trafficking of circulating CD8⁺ T cells into tumors (Fig. 3C). Anti-HER2/CD3 TDB treatment for 4 days resulted in a 4-fold increase in labeled CD8⁺ T cells (Fig. 3C).

The percentage of CD8⁺ T cells with diluted CTV, indicative of cell division, increased significantly in anti-HER2/CD3 TDB-treated tumors (from 15% to 73%; Fig. 3D). No evidence of CD8⁺ T-cell proliferation was observed in spleen or lymph node suggesting that anti-HER2/CD3 TDB treatment impacted CD8⁺ T-cell proliferation selectively within tumors (Supplementary Fig. S3A). In addition, anti-HER2/CD3 TDB treatment significantly increased the percentage of Ki67-positive CTV⁺ CD8⁺ T cells within tumor (Fig. 3E). The correlation between diluted CTV and Ki67 supports the proliferative status of these cells (Supplementary Fig. S3B).

Together, these data indicated that two mechanisms, TDB-induced CD8⁺ T-cell recruitment from the periphery and proliferation, contributed to the accumulation of tumor-infiltrating CD8⁺ T cells.

Anti-HER2/CD3 TDB treatment converted noninflamed tumors to an inflamed phenotype. We hypothesized that T-cell triggering by anti-HER2/CD3 TDB treatment may lead to cytokine and chemokine signaling that actively recruited T cells to tumor. Rapid cytokine release in peripheral blood has been widely documented following CD3-bispecific antibody and CAR-T therapies (30, 31). Consistent with these findings, systemic induction of multiple cytokines and chemokines was detected in serum 2 hours after anti-HER2/CD3 TDB treatment. Notably, all cytokine levels in periphery returned to baseline levels within 24 hours after treatment (Supplementary Fig. S4).

Next, we analyzed TDB-induced cytokine levels locally within tumors. As expected, substantial elevation of T-cell cytokines, including IFNγ, IL6, IL2, TNFα, and IL4, were detected in tumors 2 hours after dosing (Fig. 4A). In addition, several cytokines and chemokines not typically derived from T cells were highly induced, such as IL1β, GM-CSF, CXCL1, CCL2, CCL3, CCL4, and CCL11 (Fig. 4B), suggesting that initial T-cell triggering resulted in a dynamic inflammatory response potentially involving several cell types, including myeloid cells and stromal cells, in addition to T cells. These results indicated that anti-HER2/CD3 TDB-treated tumors acquire an inflammatory phenotype, which may influence immune cell trafficking and function in tumors.

The marked increase of CD8⁺ T cells in tumors following anti-HER2/CD3 TDB treatment (Fig. 1) implied that active T-cell–specific chemotaxis promoted recruitment of CD8⁺ T cells into tumors. CXCL-9, -10, and -11 are well-documented IFNγ-regulated T-cell chemoattractants that function via binding to CXCR3 expressed on T cells (32). The tissue gradient of CXCR3 cognate ligands is essential for the recruitment of CXCR3-expressing T cells to the site of inflammation.

Significant upregulation of intratumoral CXCL-9 and CXCL-11 was detected 2 hours after TDB treatment (Fig. 5A). Consistent with these findings in MMTV-HER2 TG model, CXCL9 and CXCL10 were significantly elevated within the tumors 24 hours after anti-HER2/CD3 TDB treatment in the Fo5 transplant model (Supplementary Fig. S5).

Next, the expression of the chemokine receptor CXCR3 was determined on T cells. Approximately 15%–50% of peripheral blood CD8⁺ T cells were positive for CXCR3 in nontreated tumor-bearing MMTV-huHER2 mice (Fig. 5B). In contrast, practically all CD4⁺ T cells were negative for the CXCR3 expression (Fig. 5B). The majority of CXCR3⁺CD8⁺ T cells displayed a T central memory (Tcm) cell subtype (Tcm, CD8⁺CD62L⁺CD44⁺). These data indicated that a large number of circulating CD8⁺ Tcm cells likely responded to CXCL-9, -10, and -11 and home to the chemoattractant source in the tumor site. The divergence in expression of CXCR3 between CD4⁺ and CD8⁺ T cells may explain the preferential accumulation of CD8⁺ T cells in tumors after anti-HER2/CD3 TDB treatment.

The expression of CXCR3 can be regulated by TCR activation and cytokines (33). Treatment with anti-HER2/CD3 TDB increased the fraction of circulating CXCR3⁺CD8⁺ T cells from appoximately 30% to 80% by day 7 and frequencies were maintained until day 21 (Fig. 5C). Increased CXCR3⁺CD8⁺ T cells within the blood coincided with increased percentages of CD8⁺ Tcm cells (Fig. 5C). These changes can be the result of CXCR3 and/or CD62L upregulation on CD8⁺ T cells, although we cannot exclude the possibility of mobilization and redistribution of CXCR3⁺ cells or/and CD8⁺ Tcm from other lymphoid organs.

The change of CXCR3 on CD4⁺ T cells after anti-HER2/CD3 TDB treatment was less prominent than that on CD8⁺ T cells. The percentage of CXCR3⁺CD4⁺ T cells in circulation increased from approximately 5% to 20% 1 day after TDB treatment, with a minor increase of CD4⁺ Tcm cells (Fig. 5D). In summary, these results support the induction of active chemokine-mediated T-cell recruitment by anti-HER2/CD3 TDB treatment through CXCR3 and its cognate chemokine ligands.

The induction of CXCR3 ligands within tumors and CXCR3 on T cells after anti-HER2/CD3 TDB treatment suggested involvement of this pathway in T-cell recruitment. To elucidate whether CXCR3 signaling is required for CD8⁺ T-cell accumulation within tumors, mice bearing Fo5 tumors were treated with an anti-CXCR3 antibody that abolishes binding of CXCL-10 and CXCL-11 to CXCR3 (ref. 34; Fig. 6A). Cotreatment with anti-HER2/CD3 TDB and anti-CXCR3 blunted anti-HER2/CD3 TDB-induced accumulation of intratumoral CD8⁺ T cells (Fig. 6C), demonstrating that the CXCR3 pathway is a key contributor to TDB-induced CD8⁺ T-cell recruitment.

CXCR3 ligands are induced in many cell types in response to IFNγ (33). Because IFNγ is detected both within tumors and serum following anti-HER2/CD3 TDB treatment (Fig. 4), we hypothesized that IFNγ secretion resulting from TDB-induced T-cell activation drives CXCR3-dependent recruitment of CD8⁺ T cells to the tumor. To test this, Fo5 tumor–bearing mice were treated with an anti-IFNγ neutralizing antibody (Fig. 6A). IFNγ blockade together with anti-HER2/CD3 TDB treatment reduced serum CXCL-9 and CXCL-10 to baseline levels in 4 of 5 mice (Fig. 6B). In addition, IFNγ blockade diminished anti-HER2/CD3 TDB-mediated CD8⁺ T-cell accumulation within tumors (Fig. 6C). These results demonstrate that both induction of T-cell chemokines and accumulation of intratumoral CD8⁺ T cells by anti-HER2/CD3 TDB were dependent on IFNγ.

The role of CXCR3 and IFNγ pathways in antitumor efficacy was evaluated using the Fo5 tumor model. Anti-HER2/CD3 TDB combination treatment with either anti-CXCR3 or anti-IFNγ blunted efficacy. Single-agent treatment with anti-CXCR3 or anti-IFNγ did not impact tumor growth (Fig. 6D). In summary, our results indicate that anti-HER2/CD3 TDB treatment induced IFNγ that subsequently upregulated CXCR3 ligands in the tumor that facilitated intratumoral CD8⁺ T-cell accumulation. Finally, these results indicate that T-cell recruitment is essential for anti-HER2/CD3 TDB antitumor activity.

In general, breast cancers are not well-infiltrated by immune cells relative to indications, such as melanoma or non–small cell lung cancer (35). The lymphocyte-predominant phenotype (stromal TILs ≥40%) constitutes only 10% of the HER2-positive/estrogen receptor (ER)-positive breast cancer subpopulation (36). We used a genetically engineered murine model harboring human HER2-expressing mammary tumors to evaluate the therapeutic activity of anti-HER2/CD3 TDB and delineate immune cell fate in mice. The female mice spontaneously develop HER2-overexpressing tumors in an immunocompetent host. The resulting tumors represent a bona fide “immune desert” phenotype (37), demonstrating deficiency of both cellular immune content and soluble proinflammatory molecules and thus recapitulate a major proportion of human breast cancers.

In this study, we demonstrated that anti-HER2/CD3 TDB therapy increased T-cell infiltration, thereby converting tumors to an inflamed phenotype. A three-pronged mechanism accounted for the TDB-induced increase in T cells detected in mammary tumors: (i) treatment-independent T-cell homing to tumors, (ii) intratumoral CD8⁺ T-cell proliferation in response to anti-HER2/CD3 treatment and, (iii) active T-cell recruitment by anti-HER2/CD3 TDB-induced chemokine signaling that was the dominant mechanism for T-cell accumulation and is required for therapeutic activity.

Response to anti-HER2/CD3 TDB is triggered by initial proinflammatory signaling characterized by rapid induction of proinflammatory cytokines (e.g., IFNγ, IL1β, TNFα, and IL2) and chemokines (e.g., CXCL1, CCL2, CCL3, CCL4, and CCL11) within tumors. The definitive sequence of local inflammatory dynamics in tumors is not fully understood, but the cascade is likely necessary for triggering subsequent anti-HER2/CD3 TDB-mediated responses. The cytokine release caused by anti-HER2/CD3 TDB can be a double-edged sword in patients. While the local cytokine release in the tumor is essential for effective cancer immunity, systemic cytokine release in circulation could result in toxicities. Cytokine release syndrome (CRS) is a complex set of adverse effects associated with systemic cytokine elevation triggered by T-cell–redirecting therapies. Systemic cytokine release is the most common cause of toxicity in clinical use of T-cell–redirecting therapies and can be life-threatening (38–40). Multiple strategies have been developed to manage and mitigate systemic cytokine release and resulting CRS. Optimal clinical dosing strategy would minimize the release of toxic systemic cytokines, while simultaneously allowing for effective local immune response to facilitate T-cell recruitment and eradication of tumors.

Regulatory T (Treg) cells are vital for the preservation of immune tolerance and prevention of exacerbated immune responses. We detected a significant increase in CD25⁺Foxp3⁺CD4⁺ Treg cells in tumors treated with anti-HER2/CD3 TDB. Studies with iTreg suggest that high-affinity TCR signaling with suboptimal costimulation favors induction of Foxp3 and generation or iTreg cells (41). Therefore, it is possible that strong CD4⁺ T-cell activation induced by anti-HER2/CD3 TDB could lead to differentiation of CD4+ T cells to iTregs. It remains to be determined in other preclinical systems or in patients whether Treg cells may interfere with the activity of anti-HER2/CD3 TDB by suppressing T effector cells. It was recently shown that Tregs did not limit antitumor combination efficacy of trastuzumab emtansine (T-DM1) and anti-CTLA-4/PD-1 therapy, but instead served a host-protective role (36). Because Treg cells inhibited CD8⁺ T-cell proliferation ex vivo, depletion of Tregs in vivo reduced antitumor efficacy following combination therapy and the majority of animals died from autoimmune inflammatory syndrome. The role of Tregs in cancer therapy is complex and warrants further investigation in the context CD3-bispecific antibodies.

CXCR3 is a receptor for T-cell specific chemokines CXCL-9, -10 and -11 (32). CXCR3 is expressed on Th1 CD4⁺ T, CD8⁺ T, NK, and NKT cells (34). All sensory components of this chemotactic signaling are regulated by IFNγ (33). The IFNγ-CXCR3 axis is a primary pathway that regulates T-cell migration to focal sites (42). Multiple studies have established the amplification loop concept where IFNγ-CXCR3-mediated recruitment of T cells and inflammation drives further recruitment and inflammation (43, 44). We speculate that in our model system, anti-HER2/CD3 TDB triggers rare T cells in the tumor to initiate this amplification loop. Activated T cells engaging tumor cells secrete IFNγ and other T-cell cytokines, thereby stimulating cytokine production by other cell types and initiating an inflammatory cascade. IFNγ sequentially initiates secretion of CXCL-9, -10, and -11 in tumors summoning more CD8⁺ T cells from the periphery to the tumor site. It is unclear which cell types produce these CXCR3 ligands and warrants further investigation. Infiltrating polyclonal CD8⁺ T cells are likely further activated by TDB treatment and subsequent tumor cell engagement that propels the amplification wheel of the inflammatory response. TDB treatment also induced local production of IL2 within the tumor, which may have a role in supporting CD8⁺ T-cell proliferation and cell expansion within tumors. In addition to T cells, NK cells and NKT cells express the chemokine receptor CXCR3 (34, 45). NK cells are innate effector cells with a critical role in cancer immunity and can mediate activity of therapeutic antibodies. NK cells also can be recruited to the site of inflammation in a CXCR3-dependent manner and may contribute to the antitumor activity of TDBs. Further studies are required to demonstrate how anti-HER2/CD3 TDB treatment affects NK-cell dynamics.

Although CXCL-9, -10, and -11 induce CXCR3 signaling, their distinct temporal and spatial patterns of expression, as well as, their differential binding affinities can orchestrate specificity with respect to the timing and location of the immune response. CXCL11 has the highest affinity for CXCR3, followed by CXCL10 and then CXCL9 (46). The concentration of CXCL11 is therefore lower compared with CXCL9 and CXCL10 (47). In vivo studies have demonstrated redundancy among these three chemokines, whereas other models have defined a single dominant CXCR3 ligand (48, 49). All three ligands were stimulated by anti-HER2/CD3 TDB across tumor models used herein. Interestingly, a CXCR3 blocking antibody (clone CXCR3-173) used in this study has previously been reported to lack the ability to inhibit CXCL9 binding to CXCR3, due to a nonoverlapping binding site (34). Inhibition of CXCL-10 and -11 binding significantly inhibited T-cell accumulation following anti-HER2/CD3 TDB treatment; however, other chemokines, for example, CXCL9 or CCL5, likely contributed to the modest accumulation of T-cell infiltration in the presence of anti-CXCR3 treatment. The specific contribution of these individual chemokines is not completely elucidated and requires additional investigation with the appropriate tools.

In summary, our results demonstrate that anti-HER2/CD3 TDB effectively induced CD8⁺ T-cell infiltration into HER2-positive breast tumors, thereby converting their immune contexture from an immune desert to an inflamed subtype. The main mechanism that contributed to acute antitumor activity of anti-HER2/CD3 TDB was IFNγ-induced CXCR3-mediated T-cell recruitment. The data presented were generated using limited in vivo model systems; therefore, although it provides a hypothesis and insight into the potential mechanism of action of the anti-HER2/CD3 TDB, the clinical relevance and translation of these findings are still unclear.

Long-term effects of TDB treatment on tumor immunity are largely unclear. It is tempting to speculate that TDB treatment could potentially expand tumor-specific T cells and result in release of neoantigens from cancer cells, resulting in epitope spreading. More clinical studies are required to evaluate whether TDB treatment can promote durable long-term antitumor immune responses in patients.

All authors have ownership interests (including patents) at Genentech. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J. Li, R. Ybarra, M.R. Junttila, K.B. Walsh, T.T. Junttila

Development of methodology: J. Li, R. Ybarra, P.D. Almeida, K.B. Walsh

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Li, R. Ybarra, J. Mak, A. Herault, P.D. Almeida, K. Totpal, M.R. Junttila, K.B. Walsh, A. Arrazate

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Li, R. Ybarra, A. Herault, P.D. Almeida, J. Ziai, M.R. Junttila, K.B. Walsh, T.T. Junttila

Writing, review, and/or revision of the manuscript: J. Li, R. Ybarra, A. Herault, P.D. Almeida, M.R. Junttila, K.B. Walsh, T.T. Junttila

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Li, R. Ybarra, K.B. Walsh

Study supervision: J. Li, R. Ybarra, A. Arrazate, K.Totpal, M.R. Junttila, K.B. Walsh, T.T. Junttila

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.