HDAC Inhibitor Panobinostat Engages Host Innate Immune Defenses to Promote the Tumoricidal Effects of Trastuzumab in HER2⁺ Tumors

Approximately 20% to 30% of breast cancers overexpress the HER2/neu oncogene–encoding HER2 (1). Active dimers of HER2 with other family members promote tumorigenesis by enhancing tumor cell growth and survival (2). Targeted blockade of HER2 signaling with the humanized antibody (Ab) trastuzumab (Herceptin) has substantially improved prognosis for HER2-positive breast cancer patients (3). By binding HER2, trastuzumab can induce cell-cycle arrest and/or inhibit DNA repair (3). It can also trigger immune-mediated tumor cell death through antibody-dependent cell-mediated cytotoxicity (ADCC) and may evoke therapeutically beneficial adaptive immune responses (4). However, despite these pleiotropic actions of trastuzumab, less than 35% of HER2⁺ breast cancers demonstrate initial response to trastuzumab and of those patients with metastatic disease who initially respond to anti-HER2-based chemotherapy, 70% relapse within a year of treatment initiation (5).

Histone deacetylase inhibitors (HDACi) are anticancer drugs that can mediate a diverse array of biological responses, including induction of apoptosis and cell-cycle arrest, inhibition of angiogenesis, and promotion of tumor cell differentiation (6). From an immunological perspective, HDACi can modulate the effector functions of activated immune cells and promote the efficacy of immunotherapeutic strategies in established solid tumors (7, 8). HDACi may engage anticancer immune responses by directly altering tumor cell immunogenicity, and have the potential to rescue the functional activity of exhausted CD8⁺ T cells (9, 10). By modulating the activity and accumulation of regulatory immune cells in tumors, HDACi may also influence the immunosuppressive nature of the tumor microenvironment (10).

HDACi have been shown to potentiate the antitumor effects of trastuzumab and lapatinib against trastuzumab-sensitive, HER2⁺ tumor cell lines in vitro (11–14). Such findings have established grounds for testing HDACi in combination with trastuzumab and chemotherapy in early-phase clinical trials with promising activity (15, 16). Here, we examine the combined anticancer activity and mechanisms of synergy between trastuzumab and the HDACi panobinostat in models of trastuzumab-sensitive and -refractory HER2⁺ cancer. We demonstrate the novel curative capacity of this anti–HER2-based combination therapy and its ability to engage host innate immune cells and effector molecules to evoke durable therapeutic responses against trastuzumab-refractory HER2⁺ tumors.

Six- to 8-week-old SCID mice were obtained from the Australian Resources Centre (Perth, Western Australia). NOD/SCID IL2Rγ^−/− (NSG) and BALB/c MMTV-Her2/neu transgenic mice were bred in-house and genotyped at the Peter MacCallum Cancer Centre (PMCC; ref. 17). All animal experiments were performed in accordance with institutional guidelines of the PMCC.

The AU565 and BT474 cell lines were purchased from the ATCC (2009; both lines were last authenticated in July 2015). BT474 tumor cells were maintained in RPMI1640, 10% FBS, 2 mmol/L l-glutamine, 0.1 mmol/L nonessential amino acids, 1 mmol/L sodium pyruvate, and 0.1 mmol/L HEPES. AU565 tumor cells were maintained in RPMI-1640, 10% FBS, 2 mmol/L l-glutamine and 0.1 mmol/L HEPES. Through multiple passages of the AU565 tumor line, we derived a variant (AU565^pv) that could be xenografted into NSG and SCID mice. MSCV-GFP/Bcl2 and MSCV-GFP/AKT (myristolated AKT (MyrAKT)) were transduced into the BT474 and/or AU565^pv cells as described previously (18–20). The rat HER2⁺ murine H2N100 and H2N113 mammary carcinoma cell lines (derived from MMTV Her2/neu transgenic mice; ref. 21) and YAC-1 cell line were cultured in RPMI1640, 10% FBS, 2 mmol/L l-glutamine and 0.1 mmol/L HEPES. All cell lines tested negative for Mycoplasma contamination by PCR.

Trastuzumab (Herceptin, Roche) was obtained through the PMCC Pharmacy. Panobinostat was provided by Novartis Pharmaceutical Inc. (Shanghai, China). For in vitro studies, panobinostat was dissolved in dimethyl sulfoxide (Calbiochem) at a 10 mmol/L stock solution. For in vivo studies, panobinostat was dissolved by sonication in 5% dextrose to make a 1 mg/mL stock solution. Depletion antibodies to asialoGM1 [natural killer (NK) cells; ≥90% depletion efficiency; Wako Pure Chemical] and CD8β (53.5.8; ≥85% depletion efficiency; BioXCell) were prepared as described previously (22). Blocking or neutralizing antibodies to CD16/CD32 Fc receptors (2.4G2), IFNγ (XMG1.2), CXCR3 (CD183) and CD11b were prepared as described previously (antibodies from BioXCell; ref. 22).

Tumor cells (10³) were treated with 0–10 μg/mL trastuzumab or 10 μg/mL CIg for 5 days. For the last 16 hours, 0.5 μCi tritiated thymidine was added. Levels of tritiated thymidine incorporation were measured using a β-counter (PerkinElmer). Six-replicate wells/group analyzed. Cell death assays: 10⁴ tumors cells were treated with: (i) 1–200 nmol/L panobinostat or corresponding DMSO concentrations for 24–48 hours or (ii) 1 μg/mL trastuzumab or control human IgG1 (Enzo Life Sciences) for 96 hours before addition of 20 nmol/L panobinostat or DMSO for 24 hours. Vehicle- and drug-treated cells were stained with Annexin-V-FITC (BD Pharmingen)/propidium iodine (PI) and analyzed on a LSRII analyzer (BD Biosciences). Two replicate wells/group were analyzed. A 4-hour ⁵¹Cr release assay (as described; ref. 22) was used to assess the effects of panobinostat on NK-cell cytotoxicity and tumor cell sensitivity to NK-cell killing. IL2–activated mouse NK cells were: (i) cultured in 10 nmol/L panobinostat or DMSO for 20 hours before being cocultured with ⁵¹Cr-labeled target cells or (ii) cultured with ⁵¹Cr-labeled targets that had been pretreated for 20 hours with 5–50 nmol/L panobinostat.

Immunoblots were performed on whole-cell lysates (19). Tumor cell cultures were treated as outlined above in the presence of 10 μmol/L qVD (InSolution Q-VD-OPh; Calbiochem). Protein/sample was resolved on 10% SDS-PAGE gels. Antibodies against human HER2 (tyr1248; Cell Signaling Technology), AKT (Cell Signaling Technology), pAKT (Ser473; 193H12; Cell Signaling Technology), Bcl-2 (Cell Signaling Technology), and β-Actin (Sigma-Aldrich) were used as probes that were detected using horseradish peroxidase–conjugated secondary Abs (DAKO) and enhanced chemiluminescence (ECL, Amersham; Lumilight, Roche).

BT474 (3 × 10⁶) or AU565^pv (1.5 × 10⁶) tumor cells were orthotopically injected into the fourth mammary fat pad of SCID or NSG mice. 0.72 mg estrogen pellets (Innovative Research of America) were implanted subcutaneously at the time of BT474 tumor inoculation. In some experiments, tumor-free NSG mice were reconstituted with 1.5 × 10⁶ mouse splenic NK cells, twice enriched (90%–95% purity) using an NK Cell Isolation Kit (StemCell Technologies). Seven days after NK-cell reconstitution, AU565^pv tumor cells were orthotopically injected. A total of 5 × 10⁵ H2N100 or H2N113 tumor cells were injected subcutaneously into MMTV-HER2/neu transgenic mice.

Tumor growth was monitored every 2 to 3 days and represented as tumor area. Drug treatment commenced when tumors reached 25 mm². Xenograft models: Mice were treated intraperitoneally with vehicle (PBS + 5% glucose water), panobinostat (15 mg/kg on days 1–5 and 8–12), trastuzumab (10 mg/kg on days 1, 4, 8, 11), or panobinostat plus trastuzumab.

α-rHER2 Abs (7.16.4; 100 μg on days 1, 4, 8, 11) were administered with panobinostat (15 mg/kg) or doxorubicin. Doxorubicin was injected intravenously (2 mg/kg) or intratumoral (i.t.; 50 μL of a 2 mmol/L stock) on days 0 and 7. Isotype control immunogobulins (CIg, 100–300 μg; Bio X cell), depleting (α-asialoGM1, 100 μg; α-CD8β, 100μg), neutralizing (α-CD11b, 300 μg; α-IFNγ, 200 μg) and blocking (α-CD16/CD32 [2.4G2], 100 μg; α-CXCR3, 150 μg) antibodies were administered intraperitoneally on days –1, 0, 3, 6, 9, 12, 15, 18 relative to therapy.

All mice were sacrificed when tumors reached 100–150 mm². Mice cleared of tumor were monitored out to days 80–100 posttumor clearance.

Analysis of tumor-infiltrating leukocytes was conducted on days 1, 2, 6, 9 posttherapy initiation. Dissected tumors were mechanically and enzymatically dissociated in collagenase IV (Worthington Biochemical Corporation). Cell suspensions were stained for the pan-leukocyte marker CD45.2 (eBioscience). Tumor-associated CD49b⁺ NK cells (DX5) were characterized for the differentiation markers CD27 (LG.7F9) and CD11b (M1/70), activation receptor CD69 (H1.2F3) and CD16/CD32 Fc-receptors (93). Myeloid cell populations were staining for Ly-6C (HK1.4), CD11b and Ly-6G (all antibodies from eBioscience). Cell viability determined using DAPI (4′,6-diamido-2-phenlindole; Invitrogen). Cells were analyzed on a LSRII analyzer (BD Biosciences). All tumor samples were analyzed using the same viable cell (DAPI⁻) collection gate cutoff to eliminate any biases relating to therapy-induced differences in tumor size/group.

Mouse splenic NK cells were enriched using a NK cell isolation kit (80%–85% purity; StemCell Technologies) and cultured for 5-days in 1,000 U/mL IL2. Human NK cells were isolated from peripheral blood of healthy donors using a MACS NK cell isolation kit and autoMACs separator (MACS; Miltenyi Biotec) and cultured for 1 to 3 days in 25 U/mL IL2. NK cells were maintained in RPMI-1640 supplemented, 10% FBS, 2 mmol/L l-glutamine, 10 mmol/L HEPES, 1 mmol/L sodium pyruvate, 100 μmol/L nonessential amino acids, 50 μmol/L 2-ME. NK cells were treated with 2.5–20 nmol/L panobinostat for 20 hours before being stained with antibodies to mouse CD49b, CD27, CD11b, CD69, and CD16/CD32 (BD Bioscience) or human CD56, CD16 and CD69 (BD Biosciences) as described above and analyzed on a LSRII analyzer.

RNA was extracted from whole AU565^pv tumors, harvested from SCID mice 2 days posttreatment initiation with vehicle, panobinostat, trastuzumab or panobinostat plus trastuzumab (as outlined above) using an RNeasy Plus Mini Kit (Qiagen). RNA was sequenced on the Illumina HighSeq 2000. Sequenced data were processed using Seqliner v0.2.0 (seqliner.org) RNA-seq analysis pipeline. Short read data were first aligned using TopHat/bowtie2 (23) to the human reference genome GRCh37.p12. All unmapped reads were subsequently aligned to the mouse reference genome GRCm38/mm10. Read counting was performed with HTSeq-Count from HTSeq (24) using transcript annotation from Degust Version 0.20 (Victorian Bioinformatics Consortium, Monash University and VLSCI's Life Sciences Computation Centre). The Voom-LIMMA workflow was used to normalize data and compute differential gene expression (25). All processed and raw RNAseq data files were deposited into the Gene Expression Omnibus/Sequencing Read Archive.

Statistical differences between groups were analyzed with an unpaired Student t test (in vitro experiments) or a Mann–Whitney U or Log-rank (Mantel–Cox) test (in vivo experiments) using GraphPad Prism (GraphPad Software).

The combinatorial effects of panobinostat and trastuzumab were examined in vitro against the trastuzumab-sensitive HER2⁺ breast tumor line BT474. Trastuzumab suppressed the proliferation of BT474 tumor cells in a dose-dependent manner (Fig. 1A) and increased their sensitivity to panobinostat-induced cell death as determined by Annexin-V/PI staining (Fig. 1B).

BT474 tumor cells were grown orthotopically in the mammary fat pad of NSG mice that lack T, B, and NK cells and have a dysfunctional myeloid compartment (26). Single-agent panobinostat or trastuzumab therapy suppressed tumor growth and prolonged survival of tumor-bearing mice by 20 to 30 days relative to the vehicle treated group (Fig. 1C). Complete tumor clearance was achieved in mice treated with the combination of panobinostat and trastuzumab (Fig. 1C). Similar responses to panobinostat and trastuzumab treatment, alone and in combination were observed in SCID mice (with a competent innate immune system) orthotopically transplanted with BT474 tumors (Fig. 1D). SCID mice depleted of NK cells could also support the curative activity of the combination therapy (Fig. 1E). The combined ability of panobinostat and trastuzumab to induce antitumor responses, independent of innate and adaptive immune cells suggests that direct induction of apoptosis is sufficient to mediate the curative effects of the combination therapy in trastuzumab-sensitive HER2⁺ tumors.

The combined therapeutic activity of panobinostat and trastuzumab was also examined in MMTV-HER2/neu transgenic mice bearing subcutaneously implanted H2N tumors. These tumor lines express oncogenic rat HER2 (rHER2) and are sensitive to the cytostatic and cytotoxic effects of anti-HER2 therapy (21). In vitro treatment of H2N100 tumor cells with the anti-rHER2 mAb, 7.16.4 increased their sensitivity to the apoptotic effects of panobinostat (Fig. 2A). Cotreatment of MMTV-HER2/neu transgenic mice with panobinostat and 7.16.4 was significantly more effective at controlling H2N100 tumor growth than either single agent alone (Fig. 2B). Depletion of CD8⁺ T or NK cells did not compromise the antitumor effects of the combination therapy against H2N100 tumors (Fig. 2C); reinforcing the dominant contribution of tumor-intrinsic antiproliferative and apoptotic mechanisms to the combined tumor growth inhibitory effects of panobinostat and anti-HER2 therapy in HER2⁺ tumors that are sensitive to the cytostatic effects of anti-HER2 therapy.

We next examined the combined therapeutic activity of panobinostat and trastuzumab in HER2⁺ xenografts that are inherently resistant to the cytostatic effects of trastuzumab. The HER2⁺ AU565^pv tumor line (confirmed to express similar levels of HER2 as the BT474 tumor line and was resistant to the antiproliferative effects of trastuzumab in vitro (Supplementary Fig. S1 and Fig. 3A) was modestly sensitive to the apoptotic effects of panobinostat. Addition of trastuzumab did not increase panobinostat-induced cell death (Fig. 3B).

In NSG mice bearing established AU565^pv tumors, trastuzumab treatment did not inhibit tumor growth and the administration of panobinostat, alone or in combination with trastuzumab, resulted in only a modest delay in tumor growth (Fig. 3C). In SCID mice a strikingly different antitumor effect of trastuzumab and panobinostat was observed (Fig. 3D). As single agents, both were more effective at slowing AU565^pv tumor growth compared with that seen in NSG mice, but when used in combination, trastuzumab and panobinostat evoked complete regression of the AU565^pv tumors in all mice (Fig. 3D). These data suggest that cotreatment with trastuzumab and panobinostat can overcome tumor intrinsic resistance to trastuzumab by engaging host innate immune responses present in SCID but not NSG mice.

We next sought to derive a trastuzumab-resistant BT474 tumor line. Since oncogene-driven activation of the AKT pathway can confer resistance to trastuzumab (27), we assessed the impact of trastuzumab and panobinostat, alone and in combination on the expression of activated (pAKT) and total AKT protein levels in the BT474 tumors cells. As single agents, both panobinostat and trastuzumab reduced pAKT expression levels (Fig. 4A). Complete ablation of phosphorylated AKT was observed in BT474 cells cotreated with panobinostat and trastuzumab (Fig. 4A). In contrast, no effect of panobinostat and/or trastuzumab on pAKT was observed in the trastuzumab-resistant AU565^pv tumor cells (Fig. 4A). Based on these findings, the BT474 tumor line was transduced to express constitutively active myristoylated AKT (MyrAKT), which rendered it insensitive to the cytostatic effects of trastuzumab (Fig. 4B), blocked the dephosphorylation of AKT following exposure to trastuzumab, panobinostat and the combination of both agents (Fig. 4C) and suppressed the combined apoptotic effects of panobinostat and trastuzumab (compare Figs. 1B and 4D).

In NSG mice bearing established BT474-MyrAKT tumors, trastuzumab demonstrated minimal to no effect on tumor growth (Fig. 4E). The anticancer effects of panobinostat were reduced, as was the combined curative activity of panobinostat and trastuzumab (compare Fig. 1C and 4E). These data indicate that constitutive expression of activated AKT was sufficient to protect BT474 tumors from rejection by the combination therapy in NSG mice. Remarkably however in SCID mice, complete regression of established BT474-MyrAKT tumors was evident in all mice cotreated with panobinostat and trastuzumab (Fig. 4E).

We next sought to examine the contribution of innate effector cells to the curative effects of the combination therapy in SCID mice bearing established trastuzumab resistant HER2⁺ tumors. The therapeutic activity of panobinostat and trastuzumab, alone and in combination, was examined in AU565^pv tumor-bearing SCID mice treated with an α-CD11b–targeted antibody that can disrupt the migratory activity of innate immune cells such as NK cells, macrophages and/or neutrophils, all of which express Fc-receptors and can facilitate the cytotoxic actions of trastuzumab (28). Compared with the control immunoglobulin (CIg)–treated mice, α-CD11b treatment completely abrogated the capacity of the combination treatment to eradicate established AU565^pv tumors (Fig. 5A). Blockade of the Fc-receptors CD16 (FcγRIII) and CD32 (FcγRII), which are expressed on NK cells (FcγRIII only) and most myeloid cells (29), also reduced the combined curative activity of panobinostat and trastuzumab in AU565^pv tumor-bearing SCID mice (Fig. 5B), suggesting that ADCC contributed to the antitumor effects of panobinostat and trastuzumab. Given the reported capacity of NK cells to support the ADCC activity of trastuzumab (30), we examined the combinatorial effects of panobinostat and trastuzumab in AU565^pv tumor-bearing SCID mice depleted of NK cells. NK cell depletion abolished the curative activity of the combination therapy (Fig. 5C). We also demonstrated that NK cell reconstitution of NSG mice could significantly enhanced the combined therapeutic effects of panobinostat and trastuzumab in AU565^pv tumors (Fig. 5D). These studies demonstrate that NK cells are necessary and sufficient to invoke antitumor responses to HER2⁺ trastuzumab-resistant cancers following cotreatment with panobinostat and trastuzumab.

Analysis of the immune compartment of AU565^pv tumors harvested from in vivo therapy experiments revealed that cotreatment with panobinostat and trastuzumab evoked a ≥2-fold increase in tumor-associated CD49b⁺ NK cell frequency, relative to the vehicle and single-agent treatments (Fig. 6A and B). A large proportion of these AU565^pv-associated NK cells comprised of CD27^highCD11b^high and CD27^lowCD11b^high cells (Fig. 6A), which are the most potent NK effector populations (31). The observed increase in tumor-associated NK cell frequency following cotreatment with panobinostat and trastuzumab was evident within one day of treatment initiation, before any detectable therapeutic responses (Fig. 6C). No therapy-induced increases in CD49b⁺ NK cell frequency were observed in the spleens of these mice (Supplementary Fig. S2A and S2B). The administration of antibodies to CD11b, which we previously demonstrated abrogated the curative activity of panobinostat and trastuzumab in AU565^pv tumor-bearing SCID mice (Fig. 5A), also selectively restricted therapy-induced NK cell infiltration into the tumors (Fig. 6D); highlighting an important mechanistic link between tumor-associated NK cell frequency and the combined curative activity of panobinostat and trastuzumab.

To characterize the mechanisms by which panobinostat and trastuzumab could collaboratively promote tumor-associated NK cell infiltration, RNA sequencing was performed on whole AU565^pv tumors harvested from SCID mice 2 days posttreatment with panobinostat and trastuzumab, alone or in combination. In the trastuzumab-treated AU565^pv tumors, no significant changes in immune associated murine gene expression were detected relative to the control-treated tumors (Fig. 7A; GEO accession GSE81380). Panobinostat evoked differential changes in mouse interferon, chemokine and immune response gene signatures that were quantitatively and qualitatively amplified by the concomitant administration of trastuzumab (Fig. 7A, Table 1; GEO accession GSE81380). Most enriched of the gene signatures within the panobinostat + trastuzumab treated AU565^pv tumors was the interferon-γ response pathway, which correlated with an increase in expression of the IFNγ response motif STTTCRNTTT_VSIRF_Q6 and STAT1 (GSEA, v2.2.2; Table 1 and Supplementary Fig. S3A). A concomitant increase in mouse CXCL9, CXCL10 and CXCL11 chemokine expression was also detected (Supplementary Fig. S3B). These chemokines are ligands for the CXCR3 receptor and are induced by IFNγ–mediated signaling (32). Neutralization of IFNγ significantly abrogated the capacity of the combination therapy to increase NK cell frequency within AU565^pv tumors (Fig. 7B). A similar outcome was observed with the concomitant blockade of CXCR3 and Fc-receptors (Fig. 7B). Administration of the CXCR3 or Fc-receptor blocking antibody alone did not affect the ability of the combination therapy to induce a 2-fold increase in NK cell frequency (Fig. 7B). Consistent with these data, IFNγ neutralization or the concomitant blockade of CXCR3 and Fc-receptor signaling in AU565^pv tumor-bearing SCID mice significantly inhibited the combined anticancer effects of panobinostat and trastuzumab (Fig. 7C and D). Blockade of CXCR3 signaling alone did not significantly reduce the tumor growth inhibitory effects of the combination therapy (Fig. 7D). A similar outcome was observed with the blockade of CD16/CD32 Fc-receptor signaling in AU565^pv tumor bearing SCID mice treated with both panobinostat and trastuzumab, however the long-term curative activity of combination therapy was reduced (Fig. 5B). These data suggest that panobinostat can promote tumor-associated NK cell recruitment and the ADCC activity of trastuzumab by inducing CXCR3-reactive ligand expression locally within the tumor.

Phenotypically, panobinostat treatment promoted expression of the activation marker CD69 and restricted the modulation of Fc-receptor expression on AU565^pv-tumor associated NK cells (Supplementary Fig. S4); a phenomenon also observed in IL2-activated mouse and human NK cells treated with panobinostat in vitro (Supplementary Fig. S5A and S5B), and supported by the RNA-sequencing analysis of the panobinostat ± trastuzumab treated AU565^pv tumors (Supplementary Fig. S3B).

Examination of the in vitro cytotoxic activity of IL2-activated mouse NK cells cultured in the presence or absence of panobinostat, revealed that panobinostat could induce a modest improvement in the killing activity of mouse NK cells against treatment naive AU565^pv and YAC-1 tumor lines relative to the vehicle-treated NK cells (Supplementary Fig. S5C). Conversely, panobinostat treatment of AU565^pv tumor cells did not augment their sensitivity to NK cell killing (Supplementary Fig. S5D). Collective these data suggest that the dominant mechanisms by which panobinostat can enhance trastuzumab-mediated tumor clearance is by promoting tumor-associated NK cell infiltration and effector function.

The therapeutic benefit of combining HER2-targeted therapies and HDACi for the treatment of HER2⁺ cancers has to date largely been examined in vitro (11–14). Here, we focused on examining the combined anticancer activity of trastuzumab and panobinostat in models of established HER2⁺ cancer. Our in vivo-based findings highlight the potent ability of panobinostat to augment both the cytostatic and cytotoxic anticancer effects of trastuzumab. In trastuzumab-sensitive HER2⁺ tumors, tumor intrinsic mechanisms of cooperation between panobinostat and trastuzumab resulted in the shut-down of AKT signaling, which directly correlated with the curative activity of the combination therapy. The combined ability of panobinostat and trastuzumab to harness host immune effector mechanisms to eradicate HER2⁺ xenografts was revealed in the setting of trastuzumab-refractory disease. ADCC and IFNγ–mediated, CXCR3-dependent increases in tumor-associated NK cells underpinned the curative activity of the combination therapy in AU565^pv xenografts. This is the first study to demonstrate the immune-modulatory potential of this anti–HER2-based combination strategy, and its capacity to eradicate trastuzumab-refractory HER2⁺ cancer.

The rationale for using panobinostat and trastuzumab in combination for the treatment of HER2⁺ cancer has been driven by in vitro observations demonstrating that HDACi can directly disrupt the HER2-signaling pathway by repressing the ERBB2 oncogene and/or inactivating PI3K/AKT signaling in tumor cells (33–35). By regulating the chaperone activity of HSP90, following inhibition of HDAC6, panobinostat can alter tumor cell sensitivity to the growth-inhibitory effects of HER2-targeted therapies by promoting the release and degradation of HSP90 client proteins, including HER2 (35, 36) and the subsequent up-regulation of cell-cycle regulatory proteins p21 and p27 (37). In line with this, loss of HER2 was observed in BT474 and BT474-MyrAKT tumor cells cotreated with panobinostat and trastuzumab (Supplementary Fig. S6). Such effects of HDACi and trastuzumab cotreatment on the HER2 signaling pathway have previously been examined in vitro (14, 35). In turn, HER2-targeted therapies have been reported to reprogram tumor cell sensitivity to cytotoxic therapy, including HDACi (11, 13) and/or down-regulate tumor cell expression of pro-survival proteins such as Bcl-2 (38). Our in vivo analyses of the combined molecular and therapeutic effects of trastuzumab and panobinostat treatment on trastuzumab-sensitive tumors reinforce the collaborative impact of these two anticancer agents on HER2 signaling. Notably, therapy-induced attenuation of pAKT proved to be a reliable biomarker of response to the curative effects of panobinostat and anti-HER2 therapy in NSG mice. Indeed, the expression of constitutively active AKT (MyrAKT) in BT474 tumor cells was sufficient to render them insensitive to the antiproliferative effects of trastuzumab and rejection by the combination therapy in severely immunocompromised recipient mice. These data, together with the demonstrated efficacy of the combination therapy in immune compromised mice bearing BT474 or H2N100 tumors suggest that tumor-intrinsic mechanisms underpin the collaborative interplay of panobinostat and trastuzumab in HER2⁺ tumors that are inherently sensitive to the cytostatic effects of anti-HER2 therapy. Interestingly doxorubicin failed to augment the efficacy of anti-HER2 therapy when administered intravenously or intratumorally in H2N tumor-bearing syngeneic mice; highlighting the superior therapeutic effects of panobinostat, in the combination setting, over conventional chemotherapy (Supplementary Fig. S7).

The combined potency of the tumor-intrinsic effects of panobinostat and trastuzumab in established BT474 and H2N tumors were likely dominant over any potential immune modulatory effects of the combination therapy. The combined ability of panobinostat and trastuzumab to harness host immune effector mechanisms and clear established disease was revealed in SCID mice bearing trastuzumab-resistant AU565^pv and BT474-MyrAKT tumors. Therapy-induced recruitment of NK cells into established AU565^pv tumors and promotion of trastuzumab-mediated ADCC was crucial to the curative success of the combination therapy. In line with this, in breast cancer patients, complete and partial therapeutic responses to trastuzumab were documented to correlate with higher in situ infiltration of leukocytes and NK/ADCC functions (30, 39). Interestingly in SCID mice, we also demonstrated that Bcl-2 overexpressing AU565^pv tumors were susceptible to rejection by the combination therapy (Supplementary Fig. S8), suggesting that the ADCC mechanisms promoted by the combination therapy could bypass key tumor-intrinsic survival pathways that can confer resistance to other conventional breast cancer therapies (40, 41). However, although we have identified an important role for NK cells in supporting the antitumor activity of the combination therapy, we cannot discount the contribution of other Fc-receptor–positive effector cells. Indeed, macrophages express both FcγRII and FcγRIII receptors (29) and can promote NK cell activity via the secretion of pro-inflammatory cytokines and/or direct cell-to-cell contact (28, 42). Interestingly, RNA sequencing of the AU565^pv tumors revealed elevated expression levels of the activating Fc receptors FcγRI and FcγRIV (Data not shown; GEO accession GSE81380), the expression of which is associated with monocyte-derived dendritic cells and monocytes/macrophages, respectively (29). Notably, trastuzumab is a humanized IgG1 antibody, the likes of which have been shown to bind all four mouse FcγRs. Human IgG1 antibodies possess similar reactivity against both human and mouse innate effector cells (including NK cells, macrophages and polymorphonuclear leukocytes) and induce potent ADCC in both human and mouse in vitro assay systems (43). Compared with human IgG2 and IgG3, IgG1 has proven to be the most potent human isotype in mouse models of cancer, capable of evoking a response profile similar to that of mouse IgG2a, the most potent IgG in mice (43).

By stimulating the local release of IFNγ and CXCR3-reactive chemokines CXCL9, CXCL10 and CXCL11, panobinostat could enhance the cytotoxic functions of trastuzumab by promoting tumor-associated NK cell recruitment. IFNγ is well documented to regulate NK cell frequency and activity within tumors, as well as polarize macrophages toward a tumor-protective M1-state (44). NK cells, particularly the CD27^highCD11b^high and CD27^lowCD11b^high subpopulations have also been reported to be responsive to CXCR3 ligands (31, 32, 45). By evoking tumor and/or stromal cell death, panobinostat in conjunction with the ADCC functions of trastuzumab may have stimulated a proinflammatory response, leading to the local production of IFNγ and in turn CXCL9, CXCL10, and CXCL11. Panobinostat may have also facilitated the local release of IFNγ by altering the activity of tumor-associated Fc-receptor–positive innate immune cells, through direct epigenetic effects or by promoting the cross communication between these cells (10). Trastuzumab-mediated engagement of Fc-receptors on innate immune cells may have further facilitated the local release of IFNγ (46) as well as stabilized immune cell interactions with the HER2⁺ tumor cells, a phenomenon that was shown to enhance NKG2D-mediated NK cell cytotoxicity in tumor-bearing syngeneic mice (47). Notably, although the neutralization of IFNγ was sufficient to abrogate the combined effects of panobinostat and trastuzumab on tumor-associate NK cell frequency and AU565^pv tumor clearance, the concomitant blockade of both CXCR3 and Fc-receptor signaling was required to achieve a similar outcome; highlighting the collaborative interplay of panobinostat and trastuzumab in evoking the eradication of HER2⁺ tumors that are refractory to the cytostatic effects of trastuzumab.

Given the pleiotropic antitumor responses elicited by HDACi, as well as the manageable toxicity profile for these agents (48, 49) we predict that panobinostat could greatly complement standard chemotherapy care and enhance the treatment of HER2⁺ breast tumors when combined with trastuzumab. Ultimately, the proven collaborative efficacy of panobinostat and trastuzumab in established HER2⁺ breast tumors, whether molecularly sensitive or resistant to trastuzumab therapy, provides strong rationale for pursuing the clinical development of this anti-HER2-based combination strategy.

R.W. Johnstone reports receiving a commercial research grant from Novartis and AstraZeneca, and has received speakers bureau honoraria from Novartis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Medon, J.A. Trapani, P.W. Atadja, M.A. Henderson, R.W. Johnstone, N.M. Haynes

Development of methodology: M. Medon, N.M. Haynes

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Medon, E. Vidacs, M.R. Jenkins, K.M. Ramsbottom, M.J. Smyth, M.A. Henderson, N.M. Haynes

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Medon, S.J Vervoort, J. Li, M.R. Jenkins, P.K Darcy, P.W. Atadja, M.A. Henderson, N.M. Haynes

Writing, review, and/or revision of the manuscript: M. Medon, J. Li, J.A. Trapani, M.J. Smyth, P.K Darcy, P.W. Atadja, M.A. Henderson, R.W. Johnstone, N.M. Haynes

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Medon, E. Vidacs, J. Li, M.A. Henderson, N.M. Haynes

Study supervision: P.W. Atadja, M.A. Henderson, R.W. Johnstone, N.M. Haynes

Other (performed animal work and procedures on the animals): E. Vidacs,

The authors thank Associate Prof. B. Chua for her intellectual input into the project, S.J. Haikerwal for technical assistance with the tissue analysis work, Dr. P. Beavis for his assistance with the preparation of the tumor samples for RNA sequencing, Dr. P. Neeson for provision of anti-human antibodies for the study of NK ligand expression on human breast cancer lines, Dr. C. Kearney for the provision of IL2-activated mouse NK cells, Dr. R. Pearson for provision of the MK-2206 inhibitor and K. Papastratos, S. Brown, and E. McGuire for technical assistance with the animal husbandry work.

This work was financially supported by Victorian Breast Cancer Research Consortium (R.W. Johnstone, M. Medon, and N.M. Haynes); National Health and Medical Research Council of Australia (NHMRC) and Victorian Cancer Agency (R.W Johnstone); Cancer Council of Victoria (N.M. Haynes and R.W. Johnstone).

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.