Abstract

   Anti-HER2 antibodies are effective but often lead to resistance in
   patients with HER2+ breast cancer. Here, we report an epigenetic
   crosstalk with aberrant glycerophospholipid metabolism and inflammation
   as a key resistance mechanism of anti-HER2 therapies in HER2+ breast
   cancer. Histone reader ZMYND8 specifically confers resistance to cancer
   cells against trastuzumab and/or pertuzumab. Mechanistically, ZMYND8
   enhances cPLA2α expression in resistant tumor cells through inducing
   c-Myc. cPLA2α inactivates phosphatidylcholine-specific phospholipase C
   to inhibit phosphatidylcholine breakdown into diacylglycerol, which
   diminishes protein kinase C activity leading to interleukin-27
   secretion. Supplementation with interleukin-27 protein counteracts
   cPLA2α loss to reinforce trastuzumab resistance in HER2+ tumor cells
   and patient-derived organoids. Upregulation of ZMYND8, c-Myc, cPLA2α,
   and IL-27 is prevalent in HER2+ breast cancer patients following
   HER2-targeted therapies. Targeting c-Myc or cPLA2α effectively
   overcomes anti-HER2 therapy resistance in patient-derived xenografts.
   Collectively, this study uncovers a druggable signaling cascade that
   drives resistance to HER2-targeted therapies in HER2+ breast cancer.

   Subject terms: Breast cancer, Cancer therapeutic resistance,
   Glycolipids, Epigenetics
     __________________________________________________________________

   While HER2-targeted therapies such as trastuzumab can be effective in
   patients with breast cancer, resistance often develops. Here, the
   authors demonstrate that the histone reader ZYMND8 promotes
   glycerophospholipid metabolic reprogramming via c-Myc and cPLA2α to
   increase secretion of IL-27, mediating resistance to HER2-targeted
   antibodies in preclinical models of breast cancer.

Introduction

   Human epidermal growth factor receptor 2-positive (HER2+) breast
   cancer, accounting for ~15% of all breast cancer cases, is an
   aggressive subtype whose growth is reliant on overexpression (OE) of
   the HER2 receptor tyrosine kinase. The current standard of care for
   HER2+ breast cancer involves targeting HER2 with anti-HER2 monoclonal
   antibodies and/or tyrosine kinase inhibitors, which has remarkably
   improved outcomes for patients with early disease^[51]1,[52]2. The
   FDA-approved antibodies, including trastuzumab, pertuzumab, and
   ado-trastuzumab emtansine (T-DM1), bind HER2 protein on the surface of
   cancer cells to block HER2-mediated signaling pathways that control
   cancer cell proliferation or elicit cytotoxic effects in cancer
   cells^[53]3. Large clinical studies have demonstrated that the dual
   HER2 blockade with trastuzumab plus pertuzumab achieves a higher
   pathological complete response rate than either drug alone in the
   neoadjuvant setting and thus it is FDA approved as first-line therapy
   in combination with chemotherapy for patients with early and advanced
   HER2+ breast cancer^[54]4–[55]6. Despite the initial response to
   HER2-targeted therapies, many patients with metastatic HER2+ breast
   cancer often have therapy resistance and experience tumor recurrence,
   ultimately leading to death of the disease^[56]7. Previous studies have
   identified several trastuzumab resistance mechanisms, including genetic
   alterations in ERBB2 gene (encoding HER2 protein), hyperactivation of
   phosphoinositide 3-kinase (PI3K) signaling and alternate signaling
   pathways, and immune evasion^[57]3,[58]7. However, these alterations
   may not explain all de novo and acquired mechanisms of drug resistance
   to the double HER2 blockade with trastuzumab and pertuzumab. While
   HER2-targeted antibody drug conjugates, like trastuzumab deruxtecan
   (T-DXd) and T-DM1, are effective in patients, there are some
   limitations associated with their use, including therapy resistance,
   toxicities, limited impact on heterogeneous tumors, and high costs.
   Thus, understanding the mechanism of resistance to trastuzumab plus
   pertuzumab is crucial, because it will not only provide insights into
   diverse adaptive mechanisms and biomarkers, but also contribute to
   optimizing combination therapies and to the development of personalized
   therapies capable of preventing treatment resistance and inducing
   cures.

   Epigenetic alterations are often induced by cancer treatments and drive
   the therapy resistance. The histone reader ZMYND8 is a member of
   bromodomain family proteins and acts as an oncoprotein to promote
   growth of human cancers, including brain, colorectal, kidney, liver,
   and estrogen-positive/triple-negative breast cancers^[59]8–[60]12. It
   interacts with transcription factors and epigenetic modifiers to
   modulate the expression of genes whose protein products contribute to
   angiogenesis, antioxidant response, cell migration and invasion, DNA
   damage response, epithelial-mesenchymal transition, innate immune
   response, and metabolism in cancer cells^[61]8,[62]13–[63]17. A recent
   bioinformatics analysis reported that, amongst bromodomain family
   members, only ZMYND8 mRNA was upregulated in human HER2+ breast tumors
   resistant to trastuzumab^[64]18. However, the role of ZMYND8 in
   anti-HER2 therapy resistance remains unknown.

   Metabolic reprogramming is another hallmark of human cancers and
   frequently induced by oncogenes such as c-Myc^[65]19. HER2+ breast
   cancer cells have the elevated levels of glycerophospholipids including
   phosphatidylcholine (PC), phosphatidylethanolamine (PE), and
   triacylglycerol (TAG), compared with other subtypes of breast cancer
   cells^[66]20. PC is the most abundant glycerophospholipid that acts as
   a major structural component of cell membrane in eukaryotes. It is
   synthetized via the de novo Kennedy pathway and the remodeling pathway
   (Lands’ cycle) and broken down by a family of phospholipases leading to
   generation of the second messengers involved in inflammation, cell
   death, and immunity^[67]21. Cytosolic phospholipase A2 alpha (cPLA2α)
   is a well-characterized phospholipase isoform and activated by
   intracellular calcium and kinases^[68]22. cPLA2α preferentially
   catalyzes the hydrolysis of the sn-2 ester bond of glycerophospholipids
   to produce arachidonic acid (AA) and lysophospholipids^[69]22. Previous
   studies have shown upregulation of cPLA2α promoting breast cancer
   metastasis^[70]23, however, its precise mechanism remains largely
   unknown. Whether cPLA2α associates with elevated glycerophospholipids
   contributing to HER2-targeted therapy resistance has not been explored
   yet.

   Here, we show that ZMYND8 is more potently induced in HER2+ human
   breast cancers resistant to anti-HER2 therapies. ZMYND8 is necessary
   and sufficient to drive trastuzumab or trastuzumab plus pertuzumab
   resistance in preclinical models of HER2+ breast cancer. ZMYND8
   activates the c-Myc-cPLA2α axis to inhibit PC metabolic breakdown into
   diacylglycerol (DAG) in tumor cells, which causes inactivation of
   protein kinase C (PKC) to induce interleukin-27 (IL-27) expression,
   thereby mediating tumor resistance to the HER2 antibodies. Genetic or
   pharmacological inhibition of c-Myc or cPLA2α markedly re-sensitizes
   patient-derived xenografts (PDXs) and tumor organoids to trastuzumab or
   trastuzumab plus pertuzumab in vitro and in mice. Collectively, these
   findings define the ZMYND8-c-Myc-cPLA2α-IL-27 axis as the molecular
   mechanism of anti-HER2 therapy resistance and identify both c-Myc and
   cPLA2α as druggable targets capable of surmounting HER2+ breast tumor
   resistance to HER2-targeted therapies.

Results

ZMYND8 protein is upregulated in human HER2+ breast cancer resistant to HER2
antibodies plus chemotherapy

   To investigate the role of ZMYND8 in HER2+ breast cancer, we queried
   ZMYND8 mRNA in The Cancer Genome Atlas (TCGA) cohort and found
   increased ZMYND8 mRNA levels in HER2+ breast tumors compared with
   normal breast tissues (Supplementary Fig. [71]1a). To assess ZMYND8
   upregulation at the protein level, we performed immunohistochemistry
   (IHC) with anti-ZMYND8 antibody in human HER2+ tumors. ZMYND8 protein
   levels were significantly elevated in tumor cells (Supplementary
   Fig. [72]1b, c). Notably, breast cancer patients with lower tumoral
   ZMYND8 expression achieved a complete response to HER2 antibodies-based
   therapies, whereas those with higher expression had only a partial or
   no response (Fig. [73]1a and Supplementary Table [74]1). We further
   analyzed HER2+ breast tumors harvested from the same patients before
   and after HER2-targeted therapy, and found ZMYND8 protein upregulation
   in tumors resistant to trastuzumab, pertuzumab, and docetaxel
   (Fig. [75]1b, c). In line with human tumors, ZMYND8 protein was
   upregulated in trastuzumab-resistant HR6 cells as compared with
   isogenic trastuzumab-responsive BT474T cells (Fig. [76]1d and
   Supplementary Fig. [77]1d, e), both of which were isolated from HER2
   overexpressing BT474 xenografts in mice treated with or without
   trastuzumab, respectively^[78]24. Intriguingly, ZMYND8 mRNA was not
   induced in human HER2+ breast tumors after neoadjuvant treatment with
   trastuzumab plus lapatinib in the PAMELA clinical trial^[79]25
   (Supplementary Fig. [80]1f). Similar results were found in pre-treated
   HER2+ breast tumors from patients with pathologic complete response vs.
   those with residual disease as well as in BT474T vs. HR6 cells
   (Supplementary Fig. [81]1g, h). These results suggest that ZMYND8 is
   upregulated by HER2 targeted therapies at the protein level. To
   determine the specificity of ZMYND8 induction in HER2
   antibody-resistant tumor cells, we studied ZMYND8 expression in two
   neratinib-resistant cell lines, OVCAR8NR and 5637NR, and found no
   changes in ZMYND8 protein levels between resistant cells and their
   isogenic counterparts (Supplementary Fig. [82]1i). Similarly, the
   chemotherapy drug, docetaxel, failed to induce ZMYND8 expression in
   BT474T cells (Supplementary Fig. [83]1j). Kaplan–Meier survival
   analysis showed that ZMYND8 negatively correlated with overall survival
   of patients with HER2+ breast cancer (Supplementary Fig. [84]1k).
   Collectively, these results indicate that ZMYND8 is induced in human
   HER2+ breast cancer by the current standard of care with anti-HER2
   antibodies plus chemotherapy, which may predict a poorer therapeutic
   response.

Fig. 1. ZMYND8 drives trastuzumab resistance in HER2+ breast cancer in vitro.

   [85]Fig. 1
   [86]Open in a new tab

   a ZMYND8 IHC analysis in human HER2+ breast tumors with complete
   (n = 4) or partial/no response (n = 8) to HER2-targeted therapy. b, c
   ZMYND8 IHC analysis in human HER2+ breast tumors before and after
   trastuzumab+pertuzumab+docetaxel treatment. Representative IHC images
   (b) and quantification (c, n = 3). d Immunoblot analysis of ZMYND8 and
   HER2 proteins in BT474T and HR6 cells (n = 3). e Immunoblot analysis of
   ZMYND8 and HER2 proteins in parental and ZMYND8 KO HR6 cells (n = 3).
   f, g Clonogenic growth of parental and ZMYND8 KO HR6 cells treated with
   vehicle or trastuzumab for 6 days. Representative colony images (f) and
   quantification (g, n = 2). h Immunoblot analysis of ZMYND8 protein in
   SC, ZMYND8 KD1, and ZMYND8-rescued HR6 cells (n = 3). i, j Clonogenic
   growth of SC, ZMYND8 KD1, KD1 + EV, and KD1 + ZMYND8 HR6 cells treated
   with vehicle or trastuzumab (Tras.) for 6 days. Representative colony
   images (i) and quantification (j, n = 3). k Immunoblot analysis of
   ZMYND8 protein in EV- and ZMYND8-BT474T cells (n = 3). l, m Clonogenic
   growth of EV- and ZMYND8-BT474T cells treated with vehicle or
   trastuzumab for 6 days. Representative colony images (l) and
   quantification (m, n = 3). n Immunoblot analysis of ZMYND8 protein in
   EV- and ZMYND8-SK-BR-3 cells (n = 3). o, p Clonogenic growth of EV- and
   ZMYND8-SK-BR-3 cells treated with vehicle or trastuzumab for 6 days.
   Representative colony images (o) and quantification (p, n = 3). q, r
   Immunoblot analysis of ZMYND8 and HER2 proteins in parental and ZMYND8
   KO organoids (n = 2). s–v Growth of parental, ZMYND8 KO1 and KO2
   organoids-1 treated with vehicle or trastuzumab for 9 days.
   Representative organoid images (s, u) and organoid viability
   quantification (t, v, n = 3). Data from biological replicates represent
   mean ± SEM. P-value was determined by two-way ANOVA with Tukey’s test
   (j, m, p, t, and v), unpaired 2-tailed Student’s t-test (a), or paired
   2-tailed Student’s t-test (c). Scale bar, 50 μm (b) or 200 μm (s, u).
   Source data are provided as a Source Data file.

ZMYND8 induction confers HER2 antibody resistance in HER2+ breast cancer

   To determine whether ZMYND8 regulates HER2+ breast cancer progression,
   we deleted ZMYND8 in SK-BR-3 cells using CRISPR/Cas9, and found that
   knockout (KO) of ZMYND8 significantly inhibited colony growth and
   migration of SK-BR-3 cells (Supplementary Fig. [87]2a–e). Next, we took
   advantages of isogenic HR6 and BT474T cells to study whether ZMYND8
   drives trastuzumab resistance in HER2+ breast cancer. ZMYND8 KO
   significantly reduced HR6 colony growth without affecting HER2
   expression (Fig. [88]1e and Supplementary Fig. [89]2f, g). Next, we
   seeded ZMYND8 KO1 or KO2 HR6 cells 1.5-fold more than HR6 cells to
   achieve their comparable colony numbers among vehicle groups in
   clonogenic survival studies. Trastuzumab killed ZMYND8 KO1 or KO2 cells
   but not parental HR6 cells in a dose-dependent manner (Fig. [90]1f, g).
   Sensitization of HR6 cells to 0.25 μg/mL trastuzumab was also observed
   when ZMYND8 protein was knocked down with either of two independent
   short hairpin RNAs (shRNAs), which was counteracted by re-expression of
   ZMYND8 (Fig. [91]1h-j and Supplementary Fig. [92]2h–j), suggesting a
   specific role of ZMYND8 in trastuzumab resistance. Contrary to ZMYND8
   loss-of-function, ZMYND8 OE rendered both BT474T and SK-BR-3 cells
   resistant to trastuzumab treatment (Fig. [93]1k–p and Supplementary
   Fig. [94]2k, l). To validate the role of ZMYND8 in trastuzumab
   resistance observed in isogenic cell lines, we generated
   three-dimensional organoids from two HER2 therapy-resistant breast
   cancer PDX characterized by strong ZMYND8 expression (Fig. [95]1q, r).
   The organoids maintained HER2 antibody-resistant properties
   (Fig. [96]1s–v). ZMYND8 was depleted by its single guide RNAs (sgRNAs)
   without altering HER2 protein levels (Fig. [97]1q, r). ZMYND8 KO1 or
   KO2 significantly suppressed organoid growth and further reduced
   organoid viability following 20 μg/mL trastuzumab treatment, indicating
   that ZMYND8 promotes trastuzumab resistance in HER2+ tumor organoids
   (Fig. [98]1s–v).

   To determine the specificity of ZMYND8 in trastuzumab resistance, we
   treated BT474T cells overexpressing ZMYND8 or empty vector (EV) with
   the HER2 tyrosine inhibitor lapatinib or neratinib and found that
   ZMYND8 OE did not affect the response of BT474T cells to lapatinib or
   neratinib (Supplementary Fig. [99]3a–d). Similarly, ZMYND8 KO1 or KO2
   did not alter the sensitivity of BT474T cells, SK-BR-3 cells, and
   organoids to lapatinib, neratinib, or docetaxel (Supplementary
   Fig. [100]4a–r), although ZMYND8 KO significantly inhibited the growth
   of BT474T and SK-BR-3 cells (Supplementary Fig. [101]5a–f). Together,
   these findings uncover a specific role of ZMYND8 in driving trastuzumab
   resistance in HER2+ breast cancer.

   HR6 cells were also resistant to 20-50 μg/mL pertuzumab alone or
   0.2 μg/mL (lower dose) trastuzumab plus 20–50 μg/mL pertuzumab
   (Fig. [102]2a, b). ZMYND8 KO1 or KO2 re-sensitized HR6 cells to
   0.2 μg/mL trastuzumab plus 50 μg/mL pertuzumab (Fig. [103]2a, b),
   whereas ZMYND8 OE significantly increased BT474T cell resistance to
   0.25 μg/mL trastuzumab plus 50 μg/mL pertuzumab (Fig. [104]2c, d).
   Likewise, tumor organoid resistance to 2 μg/mL (lower dose) trastuzumab
   plus 50 μg/mL pertuzumab was also alleviated by ZMYND8 KO1 or KO2
   (Fig. [105]2e–h). Collectively, our data indicate that ZMYND8 is
   necessary and sufficient to drive resistance to trastuzumab or
   trastuzumab plus pertuzumab in HER2+ breast cancer cells and organoids.

Fig. 2. ZMYND8 drives trastuzumab plus pertuzumab resistance in HER2+ breast
cancer in vitro.

   [106]Fig. 2
   [107]Open in a new tab

   a, b Clonogenic growth of parental, ZMYND8 KO1 and KO2 HR6 cells
   treated with trastuzumab (Tras.), pertuzumab (Pert.), or
   trastuzumab+pertuzumab (Tras.+Pert.) for 6 days. Representative colony
   images (a) and quantification (b, n = 3). c, d Clonogenic growth of EV
   and ZMYND8 overexpressed BT474T cells treated with vehicle or
   trastuzumab+pertuzumab for 6 days. Representative colony images (c) and
   quantification (d, n = 3). e–h Growth of parental, ZMYND8 KO1 and KO2
   organoids treated with vehicle, trastuzumab, pertuzumab, or
   trastuzumab+pertuzumab for 9 days. Representative organoid images (e,
   g) and organoid viability quantification (f, h, n = 3). Data from
   biological replicates represent mean ± SEM. P-value was determined by
   two-way ANOVA with Tukey’s test (b, d, f, and h). Scale bar, 200 μm.
   Source data are provided as a Source Data file.

   To determine if ZMYND8 drives trastuzumab resistance in HER2+ breast
   tumors in vivo, we implanted HR6 and ZMYND8 KO1 or KO2 cells into the
   mammary fat pad of female NSG mice, respectively. Mice were
   administrated intraperitoneally with vehicle or 20 mg/kg trastuzumab
   twice per week for 2 weeks when the tumor volume reached ~100 mm^3.
   ZMYND8 KO1 or KO2 effectively inhibited HR6 tumor growth (Fig. [108]3a,
   b). Trastuzumab treatment, though unable to block HR6 tumor growth,
   synergized with ZMYND8 KO1 or KO2 to induce tumor regression in mice
   (Fig. [109]3a, b). In contrast, ZMYND8 OE significantly reduced
   sensitivity of BT474T tumors to trastuzumab at both high (15 mg/kg) and
   low (5 mg/kg) doses in mice (Fig. [110]3c–e). Taken together, our
   results indicate that ZMYND8 drives trastuzumab resistance in HER2+
   breast tumors in vivo.

Fig. 3. ZMYND8 drives trastuzumab resistance in HER2+ breast cancer in mice.

   [111]Fig. 3
   [112]Open in a new tab

   a Growth curve of parental, ZMYND8 KO1 and ZMYND8 KO2 HR6 tumors in
   mice (n = 5) treated with vehicle (Veh.) or trastuzumab (Tras.). b
   Immunoblot analysis of ZMYND8 in treated tumors. c, d Growth curve of
   parental and ZMYND8 overexpressed BT474T tumors in mice (n = 5) treated
   with vehicle or trastuzumab at the dose of 15 mg/kg (c) or 5 mg/kg (d).
   e Immunoblot analysis of ZMYND8 in treated tumors. Data from biological
   replicates represent mean ± SEM. P-value was determined by two-way
   ANOVA with Tukey’s test (a, c, and d). Source data are provided as a
   Source Data file.

c-Myc induction by ZMYND8 is responsible for HER2 antibody resistance in
HER2+ breast cancer

   To determine whether ZMYND8 downstream target genes mediate HER2
   antibody resistance, we analyzed ZMYND8 transcriptome in isogenic
   BT474T and HR6 cells by RNA-sequencing (RNA-seq). 2066 genes were
   differentially regulated in HR6 cells compared with BT474T cells
   (Supplementary Fig. [113]6a). However, ZMYND8 KO inhibited 176 genes
   but induced 400 genes in HR6 cells (FDR < 0.05, LogCPM > 0, |fold
   change | > 2, Supplementary Fig. [114]6a). Hallmark pathway analysis
   with gene set enrichment analysis (FDR < 0.1) revealed, amongst
   differentially expressed genes, only c-Myc target genes were inhibited
   by both ZMYND8 KO1 and KO2, although a couple of signaling pathways
   were enriched in HR6 cells (Supplementary Fig. [115]6b–f). c-Myc was
   the top gene strongly induced in HR6 cells compared with BT474T cells,
   which was controlled by ZMYND8 (Fig. [116]4a–c). c-Myc is known to be
   induced by BRD4^[117]26. We previously showed that ZMYND8 and BRD4
   cooperate to activate hypoxia response in breast cancer^[118]8. To
   determine whether ZMYND8 induces c-Myc through BRD4, we treated BT474T
   cells overexpressing ZMYND8 or EV with a BRD4 inhibitor JQ-1 (1 μM) or
   vehicle for 6 hr. JQ-1 treatment significantly diminished
   ZMYND8-induced c-Myc expression at both mRNA and protein levels in
   BT474T cells (Supplementary Fig. [119]7a, b). These results indicate
   that ZMYND8 induces c-Myc expression through BRD4 in HER2+ breast
   cancer.

Fig. 4. c-Myc drives HER2 antibody resistance in HER2+ breast cancer in
vitro.

   [120]Fig. 4
   [121]Open in a new tab

   a mRNA levels of c-Myc and its target genes induced in HR6 cells
   compared with BT474T cells (n = 2). b, c RT-qPCR (b) and immunoblot (c)
   analyses of ZMYND8 and c-Myc in BT474T, HR6, ZMYND8 KO1 HR6, and ZMYND8
   KO2 HR6 cells (n = 3). d Immunoblot analysis of ZMYND8 and c-Myc in
   BT474T cells expressing EV or c-Myc (n = 3). e, f Clonogenic growth of
   EV and c-Myc overexpressed BT474T cells treated with vehicle or
   trastuzumab (Tras.) for 6 days. Representative colony images (e) and
   quantification (f, n = 3). g Immunoblot analysis of ZMYND8 and c-Myc in
   parental and c-Myc KO1 or KO2 HR6 cells (n = 3). h, i Clonogenic growth
   of parental and c-Myc KO1 or KO2 HR6 cells treated with vehicle or
   trastuzumab for 6 days. Representative colony images (h) and
   quantification (i, n = 3). j, k Clonogenic growth of HR6 cells treated
   with vehicle, MYCi361, trastuzumab, or MYCi361+trastuzumab for 6 days.
   Representative colony images (j) and quantification (k, n = 3). l–o
   Growth of tumor organoids treated with vehicle, trastuzumab, MYCi361,
   or trastuzumab+MYCi361 for 9 days. Representative organoid images (l,
   n) and organoid viability quantification (m, o, n = 3). p–s Growth of
   tumor organoids treated with vehicle, trastuzumab, MYCi361,
   trastuzumab+pertuzumab, trastuzumab+MYCi361, or
   trastuzumab+pertuzumab+MYCi361 for 9 days. Representative organoid
   images (p, r) and organoid viability quantification (q, s, n = 3). Data
   from biological replicates represent mean ± SEM. P-value was determined
   by bioinformatics with edgeR (a), one-way ANOVA with Dunnett’s test (b)
   or Tukey’s test (q, s), or two-way ANOVA with Tukey’s test (f, i, k, m,
   and o). Scale bar, 200 μm. Source data are provided as a Source Data
   file.

   c-Myc amplification has been reported in trastuzumab resistant HER2+
   breast tumors, but its role in trastuzumab resistance remains
   controversial^[122]27–[123]29. To determine whether c-Myc regulates
   trastuzumab resistance in HER2+ breast tumors, we overexpressed c-Myc
   in trastuzumab-responsive BT474T and SK-BR-3 cells and found that c-Myc
   OE rendered these cells resistant to trastuzumab (Fig. [124]4d-f and
   Supplementary Fig. [125]8a–c). We next depleted c-Myc in HR6 cells by
   either of two independent sgRNAs. In contrast to c-Myc
   gain-of-function, c-Myc KO1 or KO2 robustly re-sensitized HR6 cells to
   0.25 μg/mL trastuzumab (Fig. [126]4g-i). Similar results were observed
   when c-Myc KO1 or KO2 cells were treated with 0.2 μg/mL trastuzumab
   and/or 50 μg/mL pertuzumab (Supplementary Fig. [127]8d, e). In line
   with genetic loss-of-function studies, pharmacological inhibition of
   c-Myc with a small molecule inhibitor MYCi361 showed a synergistic
   growth inhibition of HR6 cells in a dose-dependent manner when combined
   with 0.25 μg/mL trastuzumab (Fig. [128]4j, k). Similarly, the combined
   treatment of 4 μM MYCi361 and 20 μg/mL trastuzumab robustly inhibited
   growth of tumor organoids, while their single treatment or combined
   treatment with lower doses of trastuzumab had no effect on organoid
   viability (Fig. [129]4l-s). MYCi361 treatment also re-sensitized tumor
   organoids to 5 μg/mL trastuzumab plus 50 μg/mL pertuzumab
   (Fig. [130]4p–s).

   To determine if ZMYND8 drives trastuzumab resistance through c-Myc, we
   restored c-Myc expression in ZMYND8 KO1 HR6 cells by lentiviral
   transduction (Fig. [131]5a) and then assessed whether c-Myc
   re-expression abolishes the effect of ZMYND8 KO on trastuzumab
   sensitization. Trastuzumab treatment effectively reduced ZMYND8 KO1
   colony growth, which was blocked by c-Myc expression (Fig. [132]5b, c).
   Given that ZMYND8 protein was induced by c-Myc (Fig. [133]4d and
   Supplementary Fig. [134]8a), we also restored ZMYND8 expression in
   c-Myc KO1 HR6 cells and found that ZMYND8 expression did not antagonize
   c-Myc KO to restore trastuzumab resistance (Fig. [135]5d–f). To
   determine the c-Myc rescue effect in vivo, we next implanted HR6,
   ZMYND8 KO1, and ZMYND8 KO1+c-Myc cells into the mammary fat pad of
   female NSG mice and mice were administrated with vehicle or 10 mg/kg
   trastuzumab once the tumor volume reached ~100 mm^3. As expected,
   ZMYND8 KO1 overcame HR6 tumor resistance to trastuzumab treatment.
   However, the expression of c-Myc restored growth of ZMYND8 KO1 tumors
   and partially blocked the antagonizing action of ZMYND8 KO1 on
   trastuzumab resistance in mice (Fig. [136]5g, h). Collectively, these
   findings indicate that c-Myc is a downstream effector of ZMYND8 driving
   trastuzumab resistance in HER2+ breast tumors.

Fig. 5. c-Myc is required for ZMYND8-induced trastuzumab resistance.

   [137]Fig. 5
   [138]Open in a new tab

   a Immunoblot analysis of c-Myc and ZMYND8 proteins in parental, ZMYND8
   KO1 and ZMYND8 KO1+c-Myc HR6 cells (n = 3). b, c Clonogenic growth of
   parental, ZMYND8 KO1 and ZMYND8 KO1+c-Myc HR6 cells treated with
   vehicle or trastuzumab for 6 days. Representative colony images (b) and
   quantification (c, n = 3). d Immunoblot analysis of ZMYND8 and c-Myc in
   parental, c-Myc KO1, and c-Myc KO1 + ZMYND8 HR6 cells (n = 2). e, f
   Clonogenic growth of parental, c-Myc KO1, and c-Myc KO1 + ZMYND8 HR6
   cells treated with vehicle or trastuzumab for 6 days. Representative
   colony images (e) and quantification (f, n = 2). g, h Growth of
   parental, ZMYND8 KO1 and ZMYND8 KO1+c-Myc HR6 tumors in mice treated
   with vehicle or trastuzumab. Tumor growth curve (g) and the percentage
   of tumor inhibition (h, n = 5). Data from biological replicates
   represent mean ± SEM. P-value was determined by one-way ANOVA with
   Dunnett’s test (h) or two-way ANOVA with Tukey’s test (c and g). Source
   data are provided as a Source Data file.

The ZMYND8-c-Myc axis alters cPLA2α-dependent glycerophospholipid metabolism
in resistant HER2+ breast cancer

   Recent studies suggest a role of metabolic reprogramming in trastuzumab
   resistance, although the underlying mechanism remains
   unknown^[139]30,[140]31. c-Myc is a master regulator of cancer cell
   metabolism and regulates therapy resistance^[141]19,[142]32. Thus, we
   hypothesized that ZMYND8 drives HER2 antibody resistance through
   c-Myc-induced metabolic alterations. To this end, we unbiasedly
   quantified the global metabolites in isogenic BT474T, HR6, and ZMYND8
   KO1 or KO2 HR6 cells by mass spectrometry. 72 metabolites were
   significantly altered in either of 4 groups (p < 0.05, Supplementary
   Fig. [143]9a). Amongst these metabolites, multiple PC species were
   significantly increased in HR6 cells (fold change > 1.3), which was
   reversed by ZMYND8 KO1 or KO2 (Supplementary Fig. [144]9a). Pathway
   enrichment analysis revealed that only phospholipid biosynthesis was
   significantly enriched in HR6 cells (Supplementary Fig. [145]9b). To
   validate these results, we next quantified phospholipids in a
   comprehensive way with a modified and specific mass spectrometry
   approach in isogenic BT474T and HR6 cells. Levels of multiple
   glycerophospholipid species including PC, phosphatidylserine (PS),
   phosphatidylglycerol (PG), and lysophosphatidylcholine (LPC) were
   significantly increased by more than 1.5-fold in resistant HR6 cells
   compared with BT474T cells (p < 0.001, Fig. [146]6a). In contrast, DAG
   and TAG species were significantly decreased in HR6 cells (p < 0.01,
   Fig. [147]6a). Notably, alterations of PC, PS, DAG, and TAG species
   were reversed by either ZMYND8 KO1 or KO2 or c-Myc KO1 or KO2
   (p < 0.05, Fig. [148]6a). These results indicate that both ZMYND8 and
   c-Myc are responsible for elevated PC and PS species and reduced DAG
   and TAG species in HER2 antibody resistant breast tumor cells.

Fig. 6. The ZMYND8-c-Myc axis alters glycerophospholipid metabolism in
resistant HER2+ breast cancer via induction of cPLA2α.

   [149]Fig. 6
   [150]Open in a new tab

   a Heatmap of glycerophospholipid species in BT474T (n = 10), HR6
   (n = 10), ZMYND8 KO1 HR6 (n = 5), ZMYND8 KO2 HR6 (n = 5), c-Myc KO1 HR6
   (n = 5), c-Myc KO2 HR6 (n = 5), cPLA2α KO1 HR6 (n = 5), and cPLA2α KO2
   HR6 cells (n = 5). b Heatmap of expression of metabolic genes involved
   in glycerophospholipids in BT474T, HR6, ZMYND8 KO1 HR6, and ZMYND8 KO2
   HR6 cells (n = 2). c RT-qPCR analysis of PLA2G3, PLA2G4A, PLAAT3, and
   PLA2G4F mRNAs in BT474T, HR6, ZMYND8 KO1 HR6, and ZMYND8 KO2 HR6 cells
   (n = 3). d Analysis of PLA2G4A mRNA expression in human HER2+ breast
   tumors before and after HER2-targeted therapy from the PAMELA clinical
   trial. e Immunoblot analysis of ZMYND8 and cPLA2α in BT474T, HR6,
   ZMYND8 KO1 HR6, and ZMYND8 KO2 HR6 cells (n = 3). f RT-qPCR analysis of
   PLA2G4A, ZMYND8 and MYC mRNAs in parental, c-Myc KO1, and c-Myc KO2 HR6
   cells (n = 3). g Immunoblot analysis of c-Myc and cPLA2α in parental,
   c-Myc KO1, and c-Myc KO2 HR6 cells (n = 3). h Immunoblot analysis of
   ZMYND8, c-Myc, and cPLA2α in parental, ZMYND8 KO1, and ZMYND8 KO1+c-Myc
   HR6 cells (n = 3). i Nucleotide sequence of the c-Myc binding element
   (in red) at the promoter of the PLA2G4A gene. j ChIP-qPCR assay with
   anti-c-Myc antibody or IgG in HR6 cells (n = 3). k Genome browser
   snapshot of ZMYND8 ChIP-seq in MDA-MB-231 cells ([151]GSE108833). rep,
   replicate. l Heatmap of glycerophospholipid species in tumor
   organoids-1 treated with vehicle or CDIBA for 6 days (n = 5). Data from
   biological replicates represent mean ± SEM. P-value was determined by
   one-way ANOVA with Dunnett’s test (c and f), paired 2-tailed Student’s
   t-test (d), or unpaired 2-tailed Student’s t-test (j). Source data are
   provided as a Source Data file.

   To determine the mechanism underlying ZMYND8/c-Myc-induced
   glycerophospholipid alterations, we analyzed the expression of
   metabolic genes involved in PC and PS metabolism in our RNA-seq
   dataset. The PC biosynthesis gene LPCAT1 and many phospholipases
   including PLA2G2A, PLA2G2F, PLA2G3, PLA2G4A, PLA2G4F, PLA2G6, PLB1,
   PLBD1, PLCB1, PLCD3, PLCD4, and PLCE1 were significantly dysregulated
   in HR6 cells compared with BT474T cells, with PLA2G3 and PLA2G4A
   dependent on ZMYND8 (Fig. [152]6b). RT-qPCR validated the specific
   induction of PLA2G4A mRNA only by ZMYND8 in HR6 cells (Fig. [153]6c).
   Consistently, PLA2G4A mRNA was robustly increased by 2.2-fold in human
   HER2+ breast tumors after neoadjuvant treatment in the PAMELA clinical
   trial^[154]25 (n = 91, Fig. [155]6d). PLA2G4A encodes cPLA2α protein, a
   member of the cytosolic calcium-dependent PLA2 group IV family that
   catalyzes the hydrolysis of glycerophospholipids^[156]22. We further
   found that ZMYND8 was required for cPLA2α protein induction in HR6
   cells (Fig. [157]6e). cPLA2α mRNA and protein expression were also
   induced by c-Myc in HR6 cells (Fig. [158]6f, g). Notably, re-expression
   of c-Myc rescued cPLA2α protein levels in ZMYND8 KO1 HR6 cells
   (Fig. [159]6h), suggesting that ZMYND8 induces cPLA2α expression
   through c-Myc. To determine if PLA2G4A is a direct c-Myc target gene,
   we analyzed the PLA2G4A promoter and identified a putative c-Myc
   binding E-box element 5’-CACGTG-3’ (Fig. [160]6i). Chromatin
   immunoprecipitation (ChIP)-qPCR assay showed that c-Myc bound the E-box
   element at the PLA2G4A promoter in HR6 cells (Fig. [161]6j). In
   contrast, there was no direct binding of ZMYND8 to the PLA2G4A promoter
   (Fig. [162]6k). In line with these results, mRNA expression correlation
   analysis revealed a positive correlation between MYC and PLA2G4A in
   HER2+ tumors resistant to neoadjuvant double HER2 blockade in the
   PAMELA clinical trial (Supplementary Fig. [163]10a). However, ZMYND8
   mRNA did not correlate with MYC or PLA2G4A mRNA (Supplementary
   Fig. [164]10b, c), which may be due to the fact that ZMYND8 protein
   levels, rather than mRNA levels, were induced in resistant tumors
   (Fig. [165]1a and Supplementary Fig. [166]1f). Together, these findings
   indicate that the ZMYND8-c-Myc axis induces cPLA2α expression in breast
   cancer cells resistant to HER2 antibodies.

   To determine if cPLA2α is responsible for ZMYND8/c-Myc-induced
   glycerophospholipid alterations, we quantified glycerophospholipids in
   isogenic BT474T, HR6, and cPLA2α KO1 or KO2 HR6 cells by mass
   spectrometry. Loss of cPLA2α prevented increases in PC and PS species
   but did not alter LPC and PG species in HR6 cells (p < 0.01,
   Fig. [167]6a). Reduction of DAG and TAG species was also reversed by
   cPLA2α KO in HR6 cells (p < 0.05, Fig. [168]6a). Consistently,
   pharmacological inhibition of cPLA2α with CDIBA (20 μM) significantly
   reduced PC species but increased TAG species in tumor organoids
   (p < 0.01, Fig. [169]6l). Together, these results indicate that the
   ZMYND8-c-Myc-cPLA2α axis reprograms PC, DAG, and TAG species in breast
   cancer resistant to HER2 antibodies.

   Next, we examined whether cPLA2α drives HER2 antibody resistance. We
   depleted cPLA2α by either of two independent sgRNAs in HR6 cells
   (Fig. [170]7a). cPLA2α KO1 or KO2 re-sensitized HR6 cells to 0.25 μg/mL
   trastuzumab (Fig. [171]7b, c). Similar results were found in HR6 cells
   treated with 0.2 μg/mL trastuzumab plus 50 μg/mL pertuzumab
   (Fig. [172]7d, e). To determine whether the catalytic activity of
   cPLA2α is required for trastuzumab resistance, we re-introduced
   wildtype (WT) cPLA2α or its catalytically inactive mutant S505A into
   cPLA2α KO1 HR6 cells (Fig. [173]7f). Expression of WT cPLA2α in cPLA2α
   KO1 HR6 cells fully restored trastuzumab resistance, whereas S505A
   mutant failed to do so (Fig. [174]7g, h). Consistently, CDIBA
   dose-dependently reduced HR6 colony growth when combined with
   0.25 μg/mL trastuzumab, though its single treatment had a marginal
   effect (Fig. [175]7i, j). Likewise, resistant organoids were
   re-sensitized to 20 μM CDIBA and 20 μg/mL trastuzumab combo, but not to
   CDIBA alone or 5 μM CDIBA and 20 μg/mL trastuzumab combo
   (Fig. [176]7k–n). Adding 20 μM CDIBA to 5 μg/mL trastuzumab plus
   50 μg/mL pertuzumab combo also effectively reduced viability of
   resistant organoids (Fig. [177]7o–r). In vivo xenograft studies further
   showed that cPLA2α KO1 or KO2 inhibited HR6 tumor growth and robustly
   overcame trastuzumab resistance in NSG mice (Fig. [178]7s–u).
   Collectively, these results indicate that cPLA2α induced by the
   ZMYND8-c-Myc axis drives HER2 antibody resistance in vitro and in mice.

Fig. 7. cPLA2α drives HER2 antibody resistance in HER2+ breast cancer.

   [179]Fig. 7
   [180]Open in a new tab

   a–c Clonogenic growth of parental, cPLA2α KO1, and cPLA2α KO2 HR6 cells
   treated with vehicle or trastuzumab (Tras.) for 6 days. Immunoblot
   analysis (a, n = 3), representative colony images (b), and
   quantification (c, n = 3). d, e Clonogenic growth of parental, cPLA2α
   KO1, and cPLA2α KO2 HR6 cells treated with vehicle, trastuzumab,
   pertuzumab (Pert.), or trastuzumab+pertuzumab for 6 days.
   Representative colony images (d) and quantification (e, n = 3). f–h
   Clonogenic growth of parental, cPLA2α KO1, cPLA2α KO1 + cPLA2α-WT and
   cPLA2α KO1 + S505A mutant HR6 cells treated with vehicle or trastuzumab
   for 6 days. Immunoblot analysis (f, n = 3), representative colony
   images (g), and quantification (h, n = 3). i, j Clonogenic growth of
   HR6 cells treated with vehicle, trastuzumab, CDIBA, or
   trastuzumab+CDIBA for 6 days. Representative colony images (i) and
   quantification (j, n = 2). k–n Growth of tumor organoids treated with
   vehicle, trastuzumab, CDIBA, or trastuzumab+CDIBA for 9 days.
   Representative organoid images (k, m) and organoid viability
   quantification (l, n, n = 3). o–r Growth of tumor organoids treated
   with vehicle, trastuzumab, CDIBA, trastuzumab+pertuzumab,
   trastuzumab+CDIBA, or trastuzumab+pertuzumab+CDIBA for 9 days.
   Representative organoid images (o, q) and organoid viability
   quantification (p, r, n = 3). s–u Growth of parental, cPLA2α KO1 and
   cPLA2α KO2 HR6 tumors in mice treated with vehicle or trastuzumab.
   Tumor growth curve (s), tumor weight (t), and the percentage of tumor
   inhibition (u, n = 5). Data from biological replicates represent
   mean ± SEM. P-value was determined by one-way ANOVA with Tukey’s test
   (p and r) or Dunnett’s test (u), or two-way ANOVA with Tukey’s test (c,
   e, h, l, n, s, and t). Scale bar, 200 μm. Source data are provided as a
   Source Data file.

IL-27 induction by the ZMYND8-c-Myc-cPLA2α axis drives anti-HER2 therapy
resistance in HER2+ breast cancer

   Next, we assessed whether cPLA2α induces the expression of
   proinflammatory genes leading to ZMYND8/c-Myc-mediated HER2 antibody
   resistance. Proinflammatory genes dysregulated in our RNA-seq dataset
   were chosen for RT-qPCR validation with a goal of identifying
   proinflammatory genes induced by the ZMYND8-c-Myc-cPLA2α axis in HR6
   cells. Amongst 11 proinflammatory genes, only IL27 was induced by
   ZMYND8 in HR6 cells compared with BT474T cells (Fig. [181]8a), whereas
   other cytokines or chemokines were repressed or not regulated by ZMYND8
   (Fig. [182]8a), consistent with our previous findings^[183]13. IL-27
   protein secretion was attenuated by genetic loss of ZMYND8 or c-Myc or
   pharmacological inhibition of c-Myc in HR6 cells and tumor organoids
   (Fig. [184]8b–d). In line with this, IL27 mRNA was increased by 2-fold
   in HER2+ breast tumors after neoadjuvant double HER2 blockade in the
   PAMELA clinical trial^[185]25 (Fig. [186]8e). Next, we studied whether
   cPLA2α controls IL-27 secretion in resistant tumor cells. KO of cPLA2α
   reduced IL-27 mRNA and protein levels in HR6 cells, which was rescued
   by cPLA2α WT but not S505A mutant (Fig. [187]8f, g), suggesting that
   the catalytic activity of cPLA2α is required for IL-27 protein
   secretion. Pharmacological inhibition of cPLA2α also dose-dependently
   reduced IL-27 protein secretion in resistant tumor organoids
   (Fig. [188]8h). These results indicate that the ZMYND8-c-Myc-cPLA2α
   axis induces IL-27 secretion in resistant HER2+ breast tumors, which
   requires the catalytic activity of cPLA2α.

Fig. 8. cPLA2α induces IL-27 secretion leading to HER2 antibody resistance in
HER2+ breast cancer.

   [189]Fig. 8
   [190]Open in a new tab

   a RT-qPCR analysis of indicated mRNAs in BT474T, HR6, ZMYND8 KO1 HR6,
   and ZMYND8 KO2 HR6 cells (n = 3). n.d, not detected. b, c ELISA
   analysis of IL-27 protein secretion from indicated cells and organoids
   (n = 3). d ELISA analysis of IL-27 protein secretion from tumor
   organoids treated with vehicle or MYCi361 (n = 3). e Analysis of IL27
   mRNA expression in human HER2+ breast tumors before and after
   HER2-targeted therapy from the PAMELA clinical trial. f IL27 mRNA
   levels in parental, cPLA2α KO1 and KO2 HR6 cells (n = 3). g ELISA
   analysis of IL-27 protein secretion from parental, cPLA2α KO1, cPLA2α
   KO1 plus cPLA2α-WT or S505A HR6 cells (n = 3). h ELISA analysis of
   IL-27 protein secretion from tumor organoids treated with vehicle or
   CDIBA (n = 3). i–l Clonogenic growth of BT474 (i, j) and SK-BR-3 (k, l)
   cells pre-treated with IL-27 protein for 30 min followed by
   co-treatment with vehicle or trastuzumab (Tras.) for 6 days.
   Representative colony images (i, k) and quantification (j, l, n = 3).
   m, n Clonogenic growth of BT474 cells pre-treated with conditional
   media from HR6 cells or ZMYND8 KO counterparts for 30 min followed by
   co-treatment with vehicle or trastuzumab for 6 days. Representative
   colony images (m) and quantification (n, n = 3). o, p Clonogenic growth
   of cPLA2α KO1 and KO2 HR6 cells pre-treated with IL-27 protein for
   30 min followed by co-treatment with vehicle or trastuzumab for 6 days.
   Representative colony images (o) and quantification (p, n = 2). q–t
   Growth of tumor organoids pre-treated with IL-27 protein for 30 min
   followed by co-treatment with vehicle or trastuzumab+CDIBA for 9 days.
   Representative organoid images (q, s) and organoid viability
   quantification (r, t, n = 3). Data from biological replicates represent
   mean ± SEM. P-value was determined by one-way ANOVA with Dunnett’s test
   (a–d, f–h, j, l, n, r, and t), or paired 2-tailed Student’s t-test (e).
   Scale bar, 200 μm. Source data are provided as a Source Data file.

   To determine a role of IL-27 in trastuzumab resistance, we pre-treated
   trastuzumab-sensitive BT474 and SK-BR-3 cells with IL-27 protein,
   followed by 0.25 or 2.5 μg/mL trastuzumab treatment. Clonogenic assay
   showed that IL-27 at the dose of 50 ng/mL and 100 ng/mL fully induced
   resistance of both BT474 and SK-BR-3 cells to trastuzumab
   (Fig. [191]8i–l). Consistently, a partial resistance to trastuzumab was
   observed in BT474 cells incubated with conditional media from HR6 cells
   but not ZMYND8 KO1 or KO2 counterparts (Fig. [192]8m, n). We next
   studied whether IL-27 secretion contributes to
   ZMYND8/c-Myc/cPLA2α-mediated HER2 antibody resistance. To this end, we
   supplemented IL-27 protein in cultured cPLA2α KO1 or KO2 HR6 cells
   before treatment of 0.25 μg/mL trastuzumab or vehicle. Supplementation
   with IL-27 protein dose-dependently reversed sensitivity of cPLA2α KO1
   or KO2 HR6 cells to trastuzumab (Fig. [193]8o, p). Similarly, IL-27
   protein dose-dependently counteracted cPLA2α inhibitor CDIBA to
   increase trastuzumab resistance in tumor organoids (Fig. [194]8q-t).
   Next, we compared the effect of IL-27, ZMYND8, c-Myc, and cPLA2α on
   HER2 antibody resistance. To this end, we generated IL-27 KO1 and KO2
   HR6 cells (Supplementary Fig. [195]11a, b), and these cells, along with
   HR6 cells and ZMYND8, c-Myc, and cPLA2α KO counterparts, were treated
   with 0.25 μg/mL trastuzumab or vehicle. Individual KO of ZMYND8, c-Myc,
   cPLA2α, and IL-27 achieved a comparable sensitivity of HR6 cells to
   trastuzumab (Supplementary Fig. [196]11c, d). Similar results were
   observed when these cells were treated with 0.2 μg/mL trastuzumab plus
   50 μg/mL pertuzumab (Supplementary Fig. [197]11e, f). These results
   support that ZMYND8, c-Myc, cPLA2α, and IL-27 work together within the
   same signaling axis to drive resistance to HER2 antibodies.
   Collectively, these findings indicate that IL-27 induced by the
   ZMYND8-c-Myc-cPLA2α axis drives HER2 antibody resistance in breast
   cancer.

cPLA2α induces IL-27 secretion through blocking DAG-induced activation of PKC

   We next studied if the metabolite downstream of cPLA2α controls IL-27
   secretion in drug resistant HER2+ breast tumors. AA (20-200 μM) failed
   to influence IL27 levels in HR6 cells (Fig. [198]9a). Interestingly,
   supplementation with a cell membrane-permeable DAG analog,
   1,2-Dioctanoyl-sn-glycerol (DOG), dose-dependently inhibited IL27 mRNA
   expression in HR6 cells (Fig. [199]9b). DAG is generated from several
   precursors including glycerophospholipids, TAG, and
   phosphatidylinositol 4,5-bisphosphate by phospholipase C and D,
   sphingomyelin synthases, PA phosphohydrolases, and lipases^[200]33.
   However, none of these known DAG metabolic enzymes were altered and/or
   dysregulated by ZMYND8 in HR6 cells (Supplementary Fig. [201]6a),
   suggesting that these enzymes may not be involved in DAG reduction in
   resistant tumor cells. Phosphatidylcholine-specific phospholipase C
   (PC-PLC) is known to catalyze the hydrolysis of PC to DAG, although its
   coding gene has not been identified yet^[202]34. Treatment of a
   specific PC-PLC inhibitor D609 (20 or 50 μM) increased IL27 mRNA in
   cPLA2α KO1 but not parental HR6 cells (Fig. [203]9c), suggesting that
   cPLA2α may competitively inhibit PC-PLC to block the hydrolysis of PC
   to DAG, leading to IL-27 secretion. We further showed that DOG
   treatment dose-dependently induced PKC phosphorylation (Fig. [204]9d),
   whereas AA (20-200 μM) failed to do so (Fig. [205]9e). In line with
   increased DAG, PKC phosphorylation was remarkably induced by cPLA2α KO1
   or KO2 in HR6 cells (Fig. [206]9f), which was reversed by cPLA2α WT but
   not S505A mutant (Fig. [207]9g). Treatment of CDIBA similarly increased
   PKC phosphorylation in HR6 cells in a dose-dependent manner
   (Fig. [208]9h). Activation of PKC was blocked by D609 treatment (20 or
   50 μM) in cPLA2α KO1 HR6 cells (Fig. [209]9i), further supporting a
   signaling pathway that cPLA2α inactivates PC-PLC to decrease DAG and
   subsequent PKC inactivation in resistant HER2+ breast tumor cells. To
   determine whether the activity of PKC is required for IL-27 inhibition,
   we treated cPLA2α KO1 HR6 cells with a PKC inhibitor Go6983. Treatment
   of Go6983 dose-dependently abolished PKC phosphorylation and caused
   IL27 derepression in cPLA2α KO1 HR6 cells (Fig. [210]9j, k).
   Consistently, trastuzumab resistance was restored in cPLA2α KO1 HR6
   cells following D609 or Go6983 treatment (Fig. [211]9l–o). In line with
   IL-27 inhibition, DOG treatment dose-dependently sensitized HR6 cells
   to 0.25 μg/ml trastuzumab (Fig. [212]9p, q), but AA failed to do so
   (Fig. [213]9r, s). Together, these results indicate that reduction of
   DAG by cPLA2α inactivates PKC to induce IL-27 secretion in resistant
   HER2+ breast tumor cells.

Fig. 9. cPLA2α induces IL-27 secretion through blocking DAG-induced
activation of PKC.

   [214]Fig. 9
   [215]Open in a new tab

   a, b IL27 mRNA levels in HR6 cells treated with AA (a) or DOG (b) for
   3 days (n = 3). c IL27 mRNA levels in parental and cPLA2α KO1 HR6 cells
   treated with D609 for 3 days (n = 3). d, e Immunoblot analysis of PKC
   phosphorylation (P-PKC) in HR6 cells treated with vehicle, DOG (d,
   n = 3), or AA (e, n = 2) for 3 days. cPLA2α KO1 HR6 cells as a positive
   control. f, g Immunoblot analysis of indicated protein in parental,
   cPLA2α KO1 or KO2, and rescued HR6 cells (n = 3). h–j Immunoblot
   analysis of P-PKC in indicated cells treated with vehicle, CDIBA (h,
   n = 2), D609 (i, n = 2), or Go6983 (j, n = 2) for 3 days. k IL27 mRNA
   levels in parental and cPLA2α KO1 HR6 cells treated with vehicle or
   Go6983 for 3 days (n = 9). l–s Clonogenic growth of parental and cPLA2α
   KO1 HR6 cells pre-treated with vehicle, D609 (l, m, n = 3), Go6983 (n,
   o, n = 3), DOG (p, q, n = 3), or AA (r, s, n = 2) for 30 min followed
   by co-treatment with vehicle (Veh.) or trastuzumab (Tras.) for 6 days.
   Representative colony images (l, n, p, and r) and quantification (m, o,
   q, and s). Data from biological replicates represent mean ± SEM.
   P-value was determined by one-way ANOVA with Dunnett’s test (b), or
   two-way ANOVA with Tukey’s test (c, k, m, o, and q). Source data are
   provided as a Source Data file.

Targeting c-Myc or cPLA2α overcomes tumor resistance to HER2 antibodies in
mice

   We next assessed the therapeutic potential of MYCi361 and CDIBA on
   combating HER2 antibody resistance in vivo. To this end, HR6 cells were
   orthotopically implanted into the mammary fat pad of female NSG mice
   and mice were administrated intraperitoneally with trastuzumab
   (20 mg/kg), MYCi361 (50 mg/kg), or both when the tumor volume reached
   about 100 mm^3. MYCi361 treatment synergistically inhibited HR6 tumor
   growth when combined with trastuzumab, while its individual treatment
   had a modest inhibitory effect (Fig. [216]10a and Supplementary
   Fig. [217]12a, b). Combined treatment of MYCi361 and trastuzumab
   dramatically increased the levels of a cell death marker cleaved
   caspase-3 (CC3) but not a cell proliferation marker Ki67 in HR6 tumors
   (Fig. [218]10b–e). Likewise, 2 mg/kg CDIBA and 20 mg/kg trastuzumab
   combo also effectively increased CC3 levels and killed HR6 tumors in
   mice (Fig. [219]10f-h and Supplementary Fig. [220]12c, d). Notably,
   treatment of 50 mg/kg MYCi361 or 20 mg/kg CDIBA enhanced the response
   of human HER2+ PDX tumors to trastuzumab in mice (Fig. [221]10i–n and
   Supplementary Fig. [222]12e–h). MYCi361, CDIBA, trastuzumab, or their
   combos had little effect on mouse body weight, suggesting that mice are
   well tolerated with these treatments (Supplementary Fig. [223]12i–l).
   Lastly, we found that treatment of 50 mg/kg MYCi361 or 20 mg/kg CDIBA
   robustly increased CC3 levels and abrogated HER2+ PDX tumor resistance
   to 30 mg/kg trastuzumab plus 30 mg/kg pertuzumab in mice without
   obvious toxicities (Fig. [224]10o–q and Supplementary Fig. [225]12m–o).
   Both inhibitors significantly extended survival of tumor-bearing mice
   (Fig. [226]10r). Together, these findings indicate that targeting c-Myc
   or cPLA2α can effectively overcome anti-HER2 therapy resistance in the
   clinically relevant breast cancer mouse models.

Fig. 10. Targeting c-Myc or cPLA2α overcomes HER2 antibody resistance in
mice.

   [227]Fig. 10
   [228]Open in a new tab

   a Growth curve of HR6 tumors in mice treated with vehicle, trastuzumab
   (Tras.), MYCi361 (MYCi), or trastuzumab+MYCi361 (n = 5). b, c
   Representative Ki67 IHC image (b) and quantification (c) in tumors
   (n = 5). d, e Representative CC3 IHC image (d) and quantification (e)
   in tumors (n = 5). f Growth curve of HR6 tumors in mice treated with
   vehicle, trastuzumab, CDIBA, or trastuzumab+CDIBA (n = 5). g, h
   Representative CC3 IHC image (g) and quantification (h) in tumors
   (n = 5). i Growth curve of BCM-4169 tumors in mice treated with
   vehicle, trastuzumab, MYCi361, or trastuzumab+MYCi361 (n = 4). j, k
   Representative CC3 IHC image (j) and quantification (k) in tumors
   (n = 4). l Growth curve of BCM-4169 tumors in mice treated with
   vehicle, trastuzumab, CDIBA, or trastuzumab+CDIBA (n = 5). m, n
   Representative CC3 IHC image (m) and quantification (n) in tumors
   (n = 5). o Growth curve of BCM-4169 tumors in mice treated with
   vehicle, trastuzumab+pertuzumab (Pert.), trastuzumab+pertuzumab+MYCi361
   or CDIBA (n = 5). p, q Representative CC3 IHC image (p) and
   quantification (q) in tumors (n = 5). r Kaplan–Meier survival curve of
   BCM-4169 tumor-bearing mice treated with vehicle,
   trastuzumab+pertuzumab, trastuzumab+pertuzumab+MYCi361 or CDIBA. s A
   working model of the ZMYND8-c-Myc-cPLA2α-IL-27 axis-induced anti-HER2
   therapy resistance in HER2+ breast cancer. Data from biological
   replicates represent mean ± SEM. P-value was determined by one-way
   ANOVA with Dunnett’s test (q) or Tukey’s test (e, h, k, and n), two-way
   ANOVA with Tukey’s test (a, f, i, l, and o), or log-rank test (r).
   Scale bar, 100 μm. Source data are provided as a Source Data file.

Discussion

   While several heterogenous trastuzumab resistance mechanisms have been
   previously proposed, the significant scientific barriers persist in
   designing the successful combination therapies capable of combating
   anti-HER2 therapy resistance in patients with breast cancer. In this
   study, we identified a signaling pathway driving HER2 antibody
   resistance in HER2+ breast cancer, which yields ZMYND8, c-Myc, cPLA2α,
   and IL-27 as druggable targets and biomarkers specifically for breast
   cancer resistant to trastuzumab and pertuzumab but not HER2 tyrosine
   kinase inhibitors (Fig. [229]10s). Importantly, our findings provide a
   proof-of-concept of c-Myc and cPLA2α inhibitor therapeutics as the
   rational strategies that can overcome resistance to anti-HER2 therapies
   (Fig. [230]10s). Therefore, this work has paved the way for a profound
   impact on both fundamental and translational aspects of human HER2+
   breast cancer.

   We showed that ZMYND8 is a key epigenetic regulator that initiates
   signal transduction of anti-HER2 therapy resistance by enhancing c-Myc
   transcription (Fig. [231]10s). ZMYND8’s transcriptional coactivator
   function requires BRD4^[232]8, which is known to control c-Myc
   transcription^[233]35. Interestingly, ZMYND8 is induced by HER2
   targeted therapies at the protein but not mRNA levels in resistant
   HER2+ tumors. c-Myc controls ZMYND8 protein upregulation but the
   underlying mechanism requires further investigation. Nevertheless,
   these findings indicate a mutual regulation between ZMYND8 and c-Myc.
   c-Myc amplification has been detected in trastuzumab resistant human
   HER2+ breast tumors^[234]27. Our functional studies showed that both
   ZMYND8 and c-Myc are sufficient and necessary for acquired resistance
   to HER2 antibodies, highlighting the pivotal role of epigenetic
   dysregulation in anti-HER2 therapy resistance in breast cancer.
   Importantly, our present studies have expanded our understanding of
   ZMYND8 and c-Myc to the field of glycerophospholipid research. While
   ZMYND8 is known to regulate multiple signaling pathways in cancer
   cells^[235]8,[236]13–[237]17, the ZMYND8-c-Myc axis is specific for
   HER2 antibody resistant breast cancer. Several intrinsic trastuzumab
   resistance mechanisms, including activation of PI3K/AKT and HER2
   amplification, have been reported^[238]3,[239]7. We found that ZMYND8
   loss does not alter HER2 protein levels. However, whether ZMYND8
   regulates additional intrinsic resistance mechanisms to HER2-targeted
   therapy needs to be explored in the future.

   Our findings here provide the functional evidence of aberrant
   glycerophospholipid metabolism promoting anti-HER2 therapy resistance
   in breast cancer, which is supported by recent studies showing elevated
   lipid droplets in trastuzumab resistant HER2+ breast cancer
   cells^[240]31. Glycerophospholipids are previously reported to be
   elevated in HER2+ breast tumor cells^[241]20, suggesting that increased
   glycerophospholipids prime breast tumor cells to develop an intrinsic
   resistance to anti-HER2 treatment. Previous studies demonstrated that
   HER2 OE increases fatty acid biosynthesis through stimulating the
   expression of fatty acid synthase^[242]36. However, HER2 expression is
   not altered in ZMYND8/c-Myc-overexpressed breast tumor cells. Notably,
   we uncover that cPLA2α is upregulated in HER2+ breast tumors from
   patients who do not respond to the dual HER2 blockade in the PAMELA
   trial and acts as a key regulator of aberrant glycerophospholipid
   metabolism and anti-HER2 therapy resistance in HER2+ breast cancer.
   cPLA2α is a primary phospholipase A2 that catalyzes the hydrolysis of
   glycerophospholipids to produce AA and lysophospholipids and also has
   phospholipase A1 activity under a certain condition^[243]37. Our
   studies suggest that cPLA2α may inactivate a currently unidentified
   PC-PLC, which requires its phospholipase A2 activity. As a result,
   activation of cPLA2α leads to a decrease in DAG production in resistant
   HER2+ breast tumor cells. These findings reveal a previously
   unappreciated function of cPLA2α in glycerophospholipid metabolism.
   Previous evidence has indicated that cPLA2α activation significantly
   correlates with HER2 overexpression^[244]38, and that PIK3CA mutation
   is associated with increased cPLA2α activity^[245]39. These findings
   suggest that cPLA2α may be also involved in resistance to HER2-targeted
   therapies in tumors with HER2 amplification and/or PIK3CA mutation.
   Thus, our study may have important therapeutic implications for
   resistant tumors driven by HER2 amplification and/or PIK3CA mutation.

   cPLA2α has been shown to promote metastasis of cancers, including
   triple-negative breast cancer and melanoma^[246]23,[247]40. AA and
   lysophospholipids are primarily responsible for cPLA2α-mediated
   physiological and pathological processes^[248]22. Unexpectedly, we
   found that, rather than AA, DAG negatively promotes resistance to HER2
   antibodies in breast cancer through inhibiting inflammatory cytokine,
   providing insights into multifaceted regulation of inflammation by DAG.
   Recent mass spectrometry studies have revealed increased DAG species in
   breast tumor tissues compared to adjacent normal breast tissues,
   although their levels remain unknown in the HER2+ breast tumor
   subtype^[249]41. Given DAG’s ability to enhance cytotoxic T
   cell-mediated antitumor immunity^[250]42, our studies suggest that
   increasing DAG by cPLA2α inhibitor may yield a more robust impact on
   breast tumor sensitivity to HER2 antibodies in immunocompetent models,
   compared to what we have observed in immunodeficient mouse models. PKC
   plays a primary role in DAG-triggered signaling^[251]42. The PKC family
   consists of 10 isoforms with distinct functions^[252]43. Therefore, it
   is of high importance to identify the PKC isoform that is responsible
   for DAG-induced breast tumor sensitivity to anti-HER2 therapies.

   HER2 expression is known to induce inflammation in breast cancer, which
   acts as a double-edged sword in a context-dependent manner. We showed
   that IL-27 plays an autocrine role in ZMYND8-c-Myc-cPLA2α-induced
   anti-HER2 antibody resistance in breast cancer (Fig. [253]10s). IL-27
   belongs to the IL-6/IL-12 family that regulates inflammation and immune
   response^[254]44. Interestingly, we found that IL-27 is suppressed by
   PKC activation. While PKCα and PKCθ have been shown to induce IL-2,
   IL-12, and IL-27 in immune cells^[255]45,[256]46, PKCζ inhibits IL-6
   and IL-12 expression and tumorigenesis by reducing histone
   acetylation^[257]47. Given that PKCα promotes trastuzumab resistance
   through activation of the ERK pathway in HER2+ breast cancer^[258]48,
   our studies exclude a possible role of PKCα in IL-27 regulation.
   Previous studies demonstrate that IL-12 potentiates the antitumor
   effect of trastuzumab^[259]49. In contrast, IL-6 has been reported to
   induce trastuzumab resistance in HER2+ breast tumors^[260]50. These
   previous findings, along with ours, suggest a complex role of the
   IL-6/IL-12 family cytokines in anti-HER2 therapy resistance in breast
   cancer, potentially influenced by their upstream regulators and/or
   downstream effectors. IL-27 plays a critical context-dependent role in
   immune cells^[261]51, which are enriched in the microenvironment of
   HER2+ breast tumors^[262]52. It is of high interest to comprehensively
   study the autocrine and/or paracrine effect of IL-27 on anti-HER2
   therapy resistance in the preclinical animal models with proficient
   immune systems. IL-27 antibody has been shown to be well tolerated with
   a promising antitumor activity in patients with solid tumors in a phase
   I clinical trial ([263]NCT04374877). Our findings suggest that
   neutralizing IL-27 with its antibody may be another effective strategy
   that can potentiate the current HER2-targeted therapies in breast
   cancer patients.

Methods

Study approval

   Animal experiments were approved by the Animal Care and Use Committee
   at UT Southwestern Medical Center with the protocol number 2017-102199.
   The deidentified human tumor tissues, PDX, and organoids were approved
   by the Institutional Review Board at UT Southwestern Medical Center
   with written informed consent (STU082018-004).

Plasmid constructs

   Human ZMYND8, c-Myc, cPLA2α, and IL-27 sgRNAs listed in Supplementary
   Table [264]2 were cloned into lentiCRISPR-V2 vector as described
   previously^[265]8. Human ZMYND8 and scrambled control (SC) shRNAs were
   cloned into pLKO.1 vector as described previously^[266]8. Human ZMYND8,
   c-Myc and cPLA2α cDNAs were cloned into pLVX-N-FLAG vector.
   Catalytically inactive S505 A mutant was generated by site-directed PCR
   mutagenesis. All plasmids were verified by DNA sanger sequencing.

Cell culture

   Parental BT474 cells (from ATCC, HTB-20), isogenic BT474T and HR6 cells
   (from C.L.A), 5637 and 5637NR cells (from C.L.A), OVCAR8 and OVCAR8NR
   cells (from C.L.A), and SK-BR-3 cells (ATCC, HTB-30) were cultured in
   RPMI1640 or McCoy’s 5A with 10% heat-inactivated fetal bovine serum
   (FBS, Sigma) at 37 °C in a 5% CO[2]/95% air incubator. OVCAR8NR and
   5637NR cells were maintained with 1000 nM and 600 nM neratinib,
   respectively. Trastuzumab (10 μg/mL) was added to HR6 cells for
   maintenance of resistance. ZMYND8, c-Myc, cPLA2α, and IL-27 KO or KD
   cells were freshly generated by lentiviral transduction of sgRNAs or
   shRNAs in the presence of 10 μg/mL polybrene (Sigma) and selected with
   puromycin (1 μg/mL). ZMYND8 or c-Myc overexpressed BT474T and SK-BR-3
   cells were freshly generated by lentiviral transduction of ZMYND8 or
   c-Myc cDNA and selected with puromycin. ZMYND8-rescued HR6 cells were
   freshly generated by lentiviral transduction of ZMYND8 cDNA in ZMYND8
   KD or c-Myc KO cells. c-Myc-rescued HR6 cells were freshly generated by
   lentiviral transduction of c-Myc cDNA in ZMYND8 KO cells. cPLA2α-WT and
   S505A mutant-rescued HR6 cells were freshly generated by lentiviral
   transduction of respective WT and S505A cPLA2α cDNA in cPLA2α KO cells.
   Cells within three passages were used for experiments. All cell lines
   were mycoplasma-free.

Clonogenic assay

   HR6 (1.0 × 10^4 cells/well), ZMYND8 KO (1.5 × 10^4 cells/well), ZMYND8
   KD (1.5 × 10^4 cells/well), ZMYND8 KD + ZMYND8 (1.0 × 10^4 cells/well),
   c-Myc KO (1.5 × 10^4 cells/well), ZMYND8 KO+c-Myc (1.5 × 10^4
   cells/well), cPLA2α KO (1.0 × 10^4 cells/well) cells, BT474T
   (0.8-1.0 x 10^4 cells/well), BT474T-c-Myc OE (0.8 × 10^4 cells/well),
   BT474T-ZMYND8 OE (0.8 × 10^4 cells/well), BT474T-ZMYND8 KO
   (1.0-2.0 × 10^4 cells/well), SK-BR-3 (0.5 × 10^4 cells/well),
   SK-BR-3-c-Myc OE (0.5 × 10^4 cells/well), SK-BR-3-ZMYND8 OE (0.5 × 10^4
   cells/well), SK-BR-3-ZMYND8 KO (0.5-1.5 × 10^4 cells/well), and BT474
   (1.0 × 10^4 cells/well) cells were seeded onto a 96-well plate and
   cultured for 24 h and then treated with indicated drugs for 6 days.
   Colonies were fixed with 100% methanol for 15 min and stained with 0.1%
   crystal violet (MilliporeSigma). After staining, the colony area or
   crystal violet intensity was quantified^[267]8. Conditional media were
   generated from HR6 cells (3.0 × 10^5 cells/well) and ZMYND8 KO
   counterparts (5.0 × 10^5 cells/well) cultured for 3 days. 200 μL
   conditional media were added to BT474 cells 30 min before treatment
   with 0.25 μg/mL trastuzumab, and replaced every 3 days.

Cell migration assay

   Cell migration assay was performed in 24-well plate with an 8 μm-pore
   Boyden chamber. 1 × 10^5 cells were seeded in serum-free medium in the
   upper chamber and allowed to migrate for 36 h toward the bottom chamber
   containing medium with 10% FBS. The migratory cells were fixed with
   100% methanol, stained with 0.1% crystal violet, and imaged using a
   Zeiss Axio Observer Z1 fluorescence microscope. The number of migratory
   cells was counted from three independent microscope fields per sample.

Metabolomics assay

   1 × 10^6 cells were cultured with fresh medium with dialyzed 10% FBS
   (MilliporeSigma) for 2 h and then rinsed once with 3 mL of ice-cold
   normal saline solution. Cells were lyzed with 500 μL of pre-cooled 80%
   (vol/vol) methanol in water and incubated for 5 min on ice. After
   scraping from the plate, the cell lysate/methanol mixture was
   transferred to a 1 mL Eppendorf tube. The tubes were frozen in liquid
   nitrogen and repeated for three freeze-thaw cycles between liquid
   nitrogen and 37 °C water. After the third thaw, the tubes were vortexed
   for 1 min, and then centrifuged at ~20,000 g for 15 min at 4 ^oC. The
   supernatant containing metabolites was carefully transferred into a new
   tube. Protein was quantified in the supernatant using the Pierce BCA
   kit. The supernatant equivalent to 10 μg of protein was transferred to
   a new Eppendorf tube and dried using a SpeedVac. The dried samples were
   resuspended in 100 μL acetonitrile/water 80:20 (vol/vol), vortexed
   rigorously for 1 min, and centrifuged at ~20,000 g for 15 min at 4 ^oC.
   After collecting the supernatant, metabolites were quantified by LC-MS.
   The heatmap was generated by the statistical analysis module on the
   MetaboAnalyst website.

Lipid profiling assay

   250,000 cells were lysed at room temperature for lipid extraction using
   a modified methyl-tert-butyl ether method in a fresh 16 × 100 mm glass
   tube with a PTFE-lined cap (Fisher Scientific, Pittsburg, PA)^[268]53.
   After vortexing and centrifugation at 2,671 × g for 5 min, the organic
   phase was collected, spiked with 20 μL of a 1:5 diluted Splash
   Lipidomix standard solution (Avanti, Alabaster, AL), and dried under
   N[2] air flow. The samples were resuspended in hexane and transferred
   to 1.5 mL amber GC vials. Lipids were analyzed by LC-MS/MS using a
   SCIEX QTRAP 6500^+ (SCIEX, Framingham, MA) equipped with a Shimadzu
   LC-30AD (Shimadzu, Columbia, MD) high-performance liquid chromatography
   system and a 150 × 2.1 mm, 5 μm Supelco Ascentis silica column
   (Supelco, Bellefonte, PA). Samples were injected at a flow rate of
   0.3 mL/min at 2.5% solvent B (methyl tert-butyl ether) and 97.5%
   Solvent A (hexane). Solvent B was increased to 5% over 3 min and then
   to 60% over 6 min. Solvent B was decreased to 0% during 30 s while
   Solvent C [90:10 (v/v) isopropanol-water] was set at 20% and increased
   to 40% during the following 11 min. Solvent C is increased to 44% over
   6 min and then to 60% over 50 sec. The system was held at 60% solvent C
   for 1 min prior to re-equilibration at 2.5% of solvent B for 5 min at a
   1.2 mL/min flow rate. Solvent D [95:5 (v/v) acetonitrile-water with
   10 mM ammonium acetate] was infused post-column at 0.03 mL/min. Column
   oven temperature was 25 °C. Data was acquired in positive and negative
   ionization mode using multiple reaction monitoring and analyzed using
   MultiQuant software (SCIEX). The identified lipid species were
   normalized to its corresponding internal standard and protein
   concentration.

Immunoblot assay

   Cells were lyzed with NETN lysis buffer (150 mM NaCl, 1 mM EDTA, 10 mM
   Tris-HCl, pH 8.0, 0.5% IGEPAL CA-630, and protease inhibitor cocktail),
   followed by sonication for 15 s. Cell lysis was resolved by sodium
   dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to
   nitrocellulose membrane. The following primary antibodies were used:
   anti-ZMYND8 antibody (1:1000, A302-089A, Bethyl Laboratories),
   anti-c-Myc antibody (1:500, sc-40, Santa Cruz Biotechnology),
   anti-cPLA2α antibody (1:500, 2832S, Cell Signaling Technology),
   anti-HER2 antibody (1:1000, sc-33684, Santa Cruz Biotechnology),
   anti-FLAG (1:5000, F9291, Sigma), anti-P-PKC (1:500, 9371T, Cell
   Signaling Technology), and anti-β-actin (1:5000, 81115-1-RR,
   Proteintech). Proteins were visualized by chemiluminescence with ECL
   prime (GE Healthcare).

RT-qPCR assay

   Total RNA was isolated from the cultured cells or tumor organoids with
   Trizol and treated with DNase I (ThermoFisher). First-strand cDNA was
   synthesized using the iScript cDNA synthesis kit (Bio-Rad). Real-time
   qPCR was performed^[269]13 and data were normalized to RPL13A mRNA.
   qPCR primers used in this study are shown in Supplementary
   Table [270]3.

RNA-seq assay

   Total RNA was isolated from the cultured cells using the RNeasy mini
   kit (Qiagen) and treated with DNase (Qiagen). mRNA was used for library
   preparation with the KAPA mRNA HyperPrep Kit (Roche) and sequenced on
   the Illumina NextSeq 500. Bioinformatics analysis was conducted using
   the Galaxy RNA-seq workflow. Fastq files were subjected to quality
   check using FastQC (Galaxy version 0.74 + galaxy0) and then trimmed
   using Trimmomatic (Galaxy version 0.39 + galaxy2). Trimmed fastq files
   were mapped to hg38 using HISAT2 (Galaxy version 2.2.1 + galaxy1).
   Duplicates were removed using MarkDuplicates (Galaxy version 3.1.1.0).
   Read counts were generated using featureCounts (Galaxy version
   2.0.3 + galaxy2) and the differential expression analysis was performed
   using edgeR (Galaxy version 3.36.0 + galaxy4). The differential
   expression genes with a false discovery rate (FDR) <0.05, logCPM > 0,
   and |fold change | > 2 as cutoff criteria were selected for the pathway
   analysis using GSEA (v4.3.3). The heatmap and bubble plots were
   generated with SRplot^[271]54.

ChIP-qPCR assay

   Cells were fixed in 10 mL of culture media containing 1% formaldehyde
   for 10 min at room temperature and quenched by addition of 1 mL of 2 M
   glycine. After washing with TBSE buffer (20 mM Tris-HCl, pH 7.5, 150 mM
   NaCl, 1 mM EDTA), cells were lysed in lysis buffer (10 mM Tris-HCl, pH
   8.0, 100 mM NaCl, 10 mM EDTA, 0.25% Triton X-100, and protease
   inhibitor cocktail) and centrifuged. The nuclei pellet was collected
   and lysed in lysis buffer (50 mM HEPES-KOH, pH 7.5, 150 mM NaCl, 1 mM
   EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 1% SDS, and protease
   inhibitor cocktail) to extract chromatin. The chromatin DNA was
   fragmented by sonication and then centrifuged at 21,000 g for 30 min at
   4 °C. The supernatant was diluted in ChIP immunoprecipitation buffer
   (50 mM HEPES-KOH, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1%
   sodium deoxycholate, 0.1% SDS, and protease inhibitor cocktail) and
   then subjected to immunoprecipitation overnight in the presence of
   Dynabeads (ThermoFisher) with anti-c-Myc antibody (sc-40, Santa Cruz
   Biotechnology) or normal mouse IgG (sc-2025, Santa Cruz Biotechnology)
   at 4 °C. Beads were washed, eluted, and reverse-crosslinked at 65 °C
   for 4 h, followed by treatment with proteinase K at 45 °C for 1 h. The
   precipitated DNA was purified with MinElute PCR purification kit
   (Qiagen) and quantified by real-time qPCR assay. The primers used for
   ChIP-qPCR are listed in Supplementary Table [272]2. Fold enrichment was
   calculated based on Ct as 2^−Δ(ΔCt), where ΔCt = Ct[IP] - Ct[Input] and
   Δ(ΔCt) = ΔCt[antibody] - ΔCt[IgG.]

IL-27 ELISA assay

   HR6 cells (3.0 × 10^5 cells/well) as well as ZMYND8 KO (5 × 10^5
   cells/well), c-Myc KO (5 × 10^5 cells/well), cPLA2α KO (3.0 × 10^5
   cells/well), and IL-27 KO (3.0 × 10^5 cells/well) counterparts were
   seeded onto a 6-well plate with 1 mL of culture medium. Organoids
   (5.0 × 10^4 cells/well) were seeded onto a 48-well plate (pre-coated
   with 50 μL basement membrane extract) with 0.2 mL of culture medium and
   transduced with ZMYND8 sgRNA lentivirus or treated with MYCi361 (2 μM
   or 4 μM) or CDIBA (5 μM or 20 μM). Four days after culture, media were
   collected and subjected to IL-27 protein measurement with a human IL-27
   ELISA kit (ThermoFisher) according to the manufacturer’s instructions.
   The concentration of IL-27 protein was normalized to HR6 cell number or
   organoid viability.

Animal studies

   Female mice were used in our study because we studied breast cancer in
   women. 1 × 10^6 HR6 or BT474T cells suspended in PBS/Matrigel (1:1,
   Corning) were implanted into the second left mammary fat pad of
   6-8 week-old female NSG mice. Human HER2+ PDX tumors (8 mm^3) were
   implanted into the fourth left mammary fat pad of 6-8-week-old female
   NSG mice. Mice were subcutaneously administrated with 17β-estradiol (1
   μmol per mouse) daily before the tumor volume reached to 100 mm^3, and
   then were intraperitoneally administrated with trastuzumab (5-30 mg/kg,
   twice per week for 2 weeks), trastuzumab plus pertuzumab (30 mg/kg,
   twice per week for 2 weeks), MYCi361 (50 mg/kg/day, 2 days on/2 days
   off for 4 rounds), or CDIBA (2–20 mg/kg, daily for 2 weeks). Tumor
   volume was measured as described previously^[273]8. About 2 weeks after
   drug treatment, tumors were harvested, imaged, and weighted. Mouse
   survival studies were terminated upon reaching a tumor length of 2 cm.
   The maximum tumor burden permitted by the Animal Care and Use Committee
   at UT Southwestern Medical Center is 2 cm in length. In this study, the
   length of all tumors did not exceed 2 cm. Mice were housed in groups of
   4–5/cage, with temperatures ranging from 65-75 °F, humidity between
   40-60%, and a 12 h light/12 h dark cycle.

IHC assay

   IHC assay was performed on a Dako Autostainer Link 48 system^[274]8.
   The slides were baked, deparaffinized, and hydrated, followed by
   antigen retrieval in a Dako PT Link. The tissues were incubated with a
   peroxidase block and then a following primary antibody to detect the
   target antigen. The anti-ZMYND8 antibody (1:1500, A302-089A) was
   purchased from Bethyl Laboratories. The anti-Ki67 antibody (1:10000,
   7309-1-AP) was purchased from Proteintech. The anti-CC3 antibody (1:
   1500, 9661S) was purchased from Cell Signaling Technology. The EnVision
   FLEX visualization system (Agilent) was used to visualize the staining.

Patient-derived tumor organoids

   Human metastatic HER2+ PDX tumors (BCM-4169 and HCI-032) were harvested
   from female NSG mice when the tumor volume reached to 500-600 mm^3. PDX
   tumors were washed twice with 10 mL of DMEM/F-12 medium supplemented
   with 1x GlutaMax, 10 mM HEPES, and 50 U/mL penicillin/50 μg/mL
   streptomycin/100 μg/mL neomycin and minced into small fragments. Tumor
   fragments were digested for 2 h in 10 mL of dissociation media
   (DMEM/F-12, 300 U/mL collagenase, 100 U/mL hyaluronidase, 10 ng/mL EGF,
   1 mg/mL insulin) with constant shaking at 275 rpm at 37 ^oC, filtered
   through a 100-μm strainer, and centrifuged at 200  g for 5 min. The
   cell pellet was resuspended in 10 mL of red blood cell lysis buffer for
   5 min at 37 ^oC and centrifuged at 200  g for 5 min. Cell dissociation
   was performed again by incubating cell pellets for 5 min with 3 mL
   pre-warmed trypsin in a 37 ^oC incubator, and terminated by 6 mL of
   neutralization solution (2% FBS in PBS). After centrifugation at 200  g
   for 5 min, tumor cells were counted and suspended with PBS containing
   0.5% BSA. Mouse Cell Depletion Cocktail (Miltenyi Biotec, 130-104-694)
   was used to eliminate potential mouse cells from human tumor cells.
   Human tumor cells were resuspended in cold basement membrane extract
   and cell suspension was seeded to 24-well plate and solidified for
   20 min at 37 ^oC. Tumor organoids-1 from BCM-4169 were cultured in
   DMEM/F-12 medium supplemented with 1/10 volume R-Spondin 1 media
   collected from R-Spondin1-expressing 293T cells, 5 nM NRG1, 5 ng/mL
   FGF7, 20 ng/mL FGF10, 5 ng/mL EGF, 100 ng/mL Noggin, 500 nM A83-01,
   5 mM Y-27632, 500 nM SB202190, 1x B27 supplement, 1.25 mM
   N-Acetylcysteine, 5 mM Nicotinamide, 1x GlutaMax, 10 mM HEPES, and
   50 U/mL penicillin/50 μg/mL streptomycin/100 μg/mL neomycin. Tumor
   organoids-2 from HCI-032 were cultured in DMEM/F-12 medium supplemented
   with 5% FBS, 1 μg/mL hydrocortisone, 10 ng/mL EGF, 10 μM Y-27632,
   100 ng/mL FGF2, 1 mM N-Acetylcysteine, 10 nM NRG1, 10 mM HEPES, 1 x
   Glutamax, and 50 U/mL penicillin/50 μg/mL streptomycin/100 μg/mL
   neomycin. Organoids were cultured in a 5% CO[2]/95% air incubator at 37
   ^oC, and passaged every 2-3 weeks. Established tumor organoids were
   trypsinized and resuspended in organoid media containing 5% basement
   membrane extract but lacking EGF and NRG1. 20,000 cells/well were
   seeded onto a basement membrane extract (50%)-coated 96-well plate and
   treated with trastuzumab (2–20 μg/mL), pertuzumab (50 μg/mL), MYCi361
   (2–4 μM), CDIBA (5–20 μM), and/or IL-27 (50–100 ng/mL). ZMYND8 sgRNA
   lentivirus were transduced into cells 1 day prior to drug treatment.
   Seven to ten days later, the organoid viability was measured with MTS
   assay. Organoids were photographed using a Zeiss Axio Observer Z1
   fluorescence microscope.

Human tumor studies

   Human deidentified HER2+ breast tumors and their adjacent normal breast
   tissues were obtained from surgical breast cancer pathologists Y.F. and
   Y.P. for IHC studies. Staining was scored with H-score and verified by
   Yong Wang and Y.F^[275]8. The clinical responses of tumors to
   treatments were provided by Y.F. and Y.P.. For ZMYND8 IHC studies in
   paired tumors, two HER2+ tumors (Cases 1 and 2) were collected from
   patients after the completion of 4 cycles of trastuzumab, pertuzumab,
   and docetaxel. The tumor in Case 3 was a recurrent metastatic brain
   tumor collected from the patient 3 years after completing 4 cycles of
   treatment with trastuzumab, pertuzumab, and docetaxel. The biopsies
   collected from the same patient prior to treatments were used as
   controls. Gene expression data were retrieved from the PAMELA
   trial^[276]25, the Alliance trial (dbGaP Study Accession:
   phs001291.v1.p1)^[277]55, and TCGA downloaded from the UCSC Cancer
   Browser^[278]56.

Statistics

   Statistical analysis was performed using GraphPad Prism (Version
   10.4.1) by 2-tailed Student’s t-test between 2 groups, and 1- or 2-way
   ANOVA with multiple testing correction within multiple groups.
   Kaplan–Meier survival curve was analyzed by log-rank test. The number
   of biological replicates is shown in figures or figure legends. Each
   biological replicate in in vitro cell experiments includes two-three
   technical replicates. Data are expressed as mean ± SEM. A p < 0.05 was
   considered significant.

Supplementary information

   [279]Supplementary Information^ (6.9MB, pdf)
   [280]Transparent Peer Review file^ (16.7MB, pdf)

Source data

   [281]Source Data^ (36.2MB, xlsx)

Acknowledgements