ERBB2 gene reduction and alteration of DNA repair were observed after receipt of HER2-targeted therapies in patients with HER2+ BC

We analyzed paired pre- and post-treatment samples from 10 patients who underwent anti-HER2 treatment and chemotherapy for metastatic BC and experienced disease progression during treatment. Of these, 5 patients received T-DM1. HER2 expression was reduced in the post-treatment biopsy in 4 of the 10 patients, including 3 of the 5 patients who received T-DM1 therapy after trastuzumab and pertuzumab combination therapy (Fig. 1A). The reduction in HER2 expression in these patients was accompanied by the loss of ERBB2 amplification. At disease progression after T-DM1, 3 of 5 patients (60%) had lost ERBB2 amplification. In contrast, in patients treated with trastuzumab and pertuzumab alone, only 1 of the 5 (20%) had lost HER2 amplification.

Fig. 1

DNA repair pathways are activated after HER2-targetd drug treatment in patients with HER2+ BC. A Targeted WGS was performed on 10 paired patient samples after treatment with trastuzumab/pertuzumab or T-DM1. The table shows gene amplification or variation after treatment. B The gene list was analyzed using STRING software (version 11.0) to show similar categories by functions of genes. In the context of the STRING analysis, k-means clustering was applied to identify groups of genes with similar behavior. Each color indicates co-regulated gene modules related to specific canonical signaling pathways. C Targeted DNA sequencing was performed to identify gene alteration profiling from five pairs before and after T-DXd treatment and six after T-DXd treatment in BC patients tissue samples. Genomic DNA was collected from FFPE tissue samples. D Gene expression analysis in three pairs before and after T-DXd treatment in HER2+ BC patients tissue samples. Total RNA was collected from FFPE tissue slides, and an RNA-seq analysis was conducted

In patients with reduced HER2 expression, heterogeneity across specimen sites was observed. In three cases, HER2 expression was negative in some specimens, even in the pre-treatment biopsy—for example, HER2 was expressed in the primary BC tumor but not in the lymph nodes. In such cases, HER2 status was more likely to change after treatment. In contrast, in six cases, regardless of multiple biopsies, HER2 was consistently positive.

To explore the mechanism of HER2 loss and the development of resistance to anti-HER2 treatment, we analyzed genomic changes between the paired pre- and post-treatment samples, using the STRING database to identify protein–protein interactions. A pathway analysis showed multiple gene alterations after anti-HER2 treatment (Fig. 1B). In particular, genes related to the DNA repair pathway were amplified, including TOP2A, RAD21, RAD52, and MCL1. Of note, TOP2A, RAD21, and MCL1 were simultaneously upregulated in several cases. MCL1 amplification was observed in the three cases, including two cases with unchanged HER2 expression and one case with reduced expression. TOP2A amplification was observed in two cases with unchanged HER2 expression.

Genetic alterations and gene expression profiles revealed upregulation of DNA repair pathway in patients with HER2+ BC after receipt of T-DXd

The most common genetic alterations in human BC samples from SNUH were TP53 mutations (82.3%), ERBB2 amplifications (56.3%), CDK12 amplifications (43.8%), MYC amplifications (37.5%), and CDKN2A/2B copy number loss (37.5%) (Fig. 1C, Sup. Table 5, and Sup. Table 6). Some interesting features were noted when comparing the paired samples from the same patients (Sup. Table 5). While PT02, PT03, and PT04 had similar mutational profiles regardless of treatment history, we observed differences between pre- and post-T-DXd samples in PT01 and PT05. PT01 lost the CD274 splicing mutation and CDK12 truncation mutation during treatment. In contrast, PT01 gained the GNAS R201H hotspot activating mutation. PT05 harbored a MAP3K1 nonsense mutation and AURKA amplification, which were not detectable after T-DXd treatment. However, the TP53 R273C mutation emerged after T-DXd treatment, along with pathogenic mutations in ARID1A and FGFR4. In addition, copy number loss of JAK2/CD274/PDCD1LG2 was noted in PT05_Post as well as amplifications of CCND1/FGF19.

Copy number losses of many DNA repair–related genes, including BRCA2, RAD51B, ATM, and MRE11, were also observed in PT05_Post, although their exact roles would require additional study. Interestingly, when focusing on the ERBB2 copy numbers in the paired samples, we observed a trend of decreasing ERBB2 copy numbers in post-T-DXd samples compared to pre-T-DXd samples; however, cautions should be taken for this trend might have stemmed from the technical limitations of targeted sequencing or the discrepancy in tumor purity in each sample. Taken together, we found that the overall mutational landscapes of pre-T-DXd and post-T-DXd human samples are largely similar; however, certain discrepancies do exist that necessitate additional study in a larger cohort of T-DXd-treated population.

RNA-seq was conducted on the paired samples from PT01, PT02, and PT03. To elucidate the enriched cellular pathways in pre- and post-T-DXd treatment BC samples, we performed GSEA on three paired samples. Most importantly, MYC_TARGETS_V2 (normalized enrichment score [NES] = 2.144; FDR q-value < 0.001), MTORC1_SIGNALING (NES = 2.134; FDR q-value = 0.002), and DNA_REPAIR (NES = 1.625; FDR q-value = 0.002) were significantly enriched in post-T-DXd samples (Fig. 1D, Sup. Table 7).

HER2-directed ADC-resistant HER2+ BC cell lines show reduced ERBB2 gene and HER2 protein expression

To better understand the mechanisms of the molecular changes in patient samples after the development of resistance to HER2-directed ADC therapies, we established HER2-directed ADC–resistant HER2+ BC cell lines. First, we determined the ERBB2 gene copy number in five HER2+ BC cell lines: SKBR3, BT474, HCC1954, SUM190, and HCC1419. We confirmed that all tested HER2+ BC cell lines were ERBB2 gene amplified (ERBB2 gene copy number: > 10) compared to the triple-negative BC cell line MDA-MB-231 that was known to HER2-negative cell line (ERBB2 gene copy number: 1.01) (Fig. S1A). Next, we evaluated the sensitivity of HER2+ BC cell lines to HER2-ADC and observed a dose-dependent response to T-DM1 or T-DXd treatment in the 5-day short-term treatment condition (Fig. S1B).

On the basis of proliferation data, we selected SUM190 (which had the highest HER2 copy number) and HCC1954 (which had the lowest) (Fig. S1A). Before generating HER2-directed ADC-resistant cell lines, we first tested the antitumor effects of T-DM1 and T-DXd in xenograft models and confirmed that single-agent treatment led to tumor shrinkage in both SUM190 (-66.84% GI and -25.56% GI respectively, P < 0.001) and HCC1954 (-18.60% GI and -22.82% GI respectively, P < 0.05) xenografts compared to in vehicle control (Fig. S1C).

To generate HER2-directed ADC–resistant BC cell lines, we treated SUM190 and HCC1954 parent cells with T-DM1 or T-DXd at the 80% inhibitory concentration (IC80) for 3–5 days and then replaced the culture media with fresh complete media until cells recovered at a normal growth rate. This treatment/recovery cycle was repeated for about 6–12 months (Fig. 2A). We confirmed that no TDM1R and TDXdR HER2+ BC cell lines showed growth inhibition when treated with 2 µg/mL of T-DM1 or T-DXd while parent BC cell lines showed over 90% cell death with the same treatment (Fig. 2B; TDM1R and TDXdR indicate resistance to TDM1 and T-DXd, respectively). We also confirmed that both SUM190-TDXdR and HCC1954-TDXdR cell lines did not show growth inhibition when treated with the T-DXd payload, DXd(P < 0.0001, Fig. 2C) but did show growth inhibition in TDM1R-resistant cell lines (P < 0.01, Fig. 2C).

Fig. 2

Anti-HER2 antibody–drug conjugate (HER2-directed ADC)–resistant HER2+ BC cell line generation. A TDM1R and TDXdR cell lines generated by continuous treatment/recovery cycle with HER2-directed ADC. SUM190 (1 million) and HCC1954 (500,000) cells were added to the 100-mm culture dish. The next day, cells were treated with T-DM1 or T-DXd at the 80% inhibitory concentration (IC80) for 3–5 days and then replaced with fresh complete media until cells recovered at a normal growth rate. This treatment/recovery cycle was repeated for about 6–12 months. B Clonogenic assay. Parent TDM1R and TDXdR cell lines were treated with 2 µg/ml of T-DM1 or T-DXd for 14 days, and viability was measured by an SRB staining assay. Experiments were repeated three times independently. Data were collected from three biological replicates. C Antiproliferation effect of T-DXd payload and DXd in parent and TDM1R and TDXdR cell lines. Cells were treated with DXd for 14 days, and viability was measured by the SRB staining assay. Data were collected from three biological replicates. D T-DXd significantly reduces tumor growth in SUM190-TDM1R (n = 12 per group) and HCC1954-TDM1R (n = 9 per group) xenograft models. A multiple t-test comparison was used to compare tumor size between the control and treatment groups. E TDM1R and TDXdR cell lines showed reduced HER2 expression. The ImageJ program was used to measure intensity. Western blotting. F FACS analysis. TDM1R and TDXdR cell lines showed reduced cell-surface HER2 expression. Cells were maintained without drug for 7 days and collected to measure HER2 expression on the cell surface with anti-HER2-PE. Three biological replicates showed similar results. G Droplet digital PCR assay. CNV indicates copy number variation. TDM1R and TDXdR cell lines showed a reduced ERBB2 gene copy number. Each box shows mean with standard deviation; **, P < 0.01; ***, P < 0.001, ****, P < 0.0001, n.s. not significant. Data were collected from three biological replicates

To confirm the antitumor effect of T-DXd in TDM1R BC models, we conducted a xenograft assay and confirmed that T-DXd significantly reduced tumor growth in both SUM190-TDM1R (70.5% GI) and HCC1954-TDM1R (-85.9% GI) xenograft models (Fig. 2D, P< 0.01). Next, we determined the HER2 expression levels in TDM1R and TDXdR BC cell lines using Western blotting and fluorescence-activated cell sorting (FACS) analysis. Both T-DM1- and T-DXd-resistant HER2+ BC cell lines had reduced total HER2 protein and cell surface HER2 expression levels (Fig. 2E and F). To determine whether reduced HER2 protein expression in TDM1R and TDXdR BC cell lines was due to downregulation of gene expression levels, we performed a droplet digital PCR assay and observed reduction of ERBB2 gene copy numbers in all TDM1R and TDXdR BC cell lines (SUM190-TDM1R and SUM190-TDXdR: > 80% reduction, P < 0.0001, SUM190-TDM1R and SUM190-TDXdR: > 40% reduction, P < 0.001) compared to in parent cell lines (Fig. 2G).

Genetic alterations cause downregulation of HER2 gene expression

TDM1R and TDXdR BC cell lines showed reduced HER2 gene copy number and protein levels (Fig. 2). We validated the HER2 gene copy number data by fluorescence in situ hybridization analysis using the T-DM1R and TDXdR BC cell lines. As shown in Fig. 3A, parent SUM190 and HCC1954 cell lines showed HER2 gene amplification (HER2/CER17 ratio: 3.497 and 3.002, respectively), but SUM190-TDM1R (0.7004), SUM190-TDXdR (0.1352), HCC1954-TDM1R (0.6826), and HCC1954-TDXdR (1.851) cell lines showed less HER2 gene amplification in both metaphase and interphase than did parent cells (P < 0.0001). To elucidate how HER2-directed ADC reduces HER2 gene expression in HER2+ BC cell lines, we first assessed chromosome instability, which is closely related to cancer development, gain or loss of gene expression, and therapeutic resistance [33, 34]. We analyzed chromosomes in 35 metaphase cells from each parent cell line and TDM1R and TDXdR BC cell lines. We observed chromosomal aberrations in both parent and TDM1R and TDXdR BC cell lines; however, these aberrations were not increased in TDM1R and TDXdR BC cell lines compared to in parent cells (Fig. S2A). Interestingly, the SUM190-TDXdR cell line showed truncation of the HER2 gene-amplified region (Fig. S2B). On the basis of this observation, we analyzed the gene copy number ratios between the parent and TDM1R and TDXdR BC cell lines of the gene copy numbers of ERBB2, MIEN1, MIR4728, and PGAP3 near the HER2 gene location, using whole-genome sequencing data. In the SUM190-TDM1R and SUM190-TDXdR cell lines, the gene copy numbers of ERBB2, MIEN1, MIR4728, and PGAP3 were reduced by at least 50% compared to the SUM190 parent cell line. We observed similar reductions in copy numbers of these genes in the HCC1954 cell lines (Fig. 3B). Interestingly, KPL4, which is relatively resistant to T-DXd compared to SUM190 (> tenfold higher than IC50 of SUM190) and chronically exposed to T-DXd with 2 µg/ml (KPL4-TDXdR), retained HER2 gene expression, but the T-DM1-resistant KPL4 cell line showed reduced HER2 gene copy numbers (Fig. S2C). Altogether, these data suggest that chronic exposure to the HER2-ADCpayload altered HER2 gene expression in TDM1R and TDXdR HER2+ BC cell lines.

Fig. 3

HER2-directed ADC-resistant HER2+ BC cell lines showed reduced ERBB2 gene amplification. A Fluorescence in situ hybridization analysis. The red color indicates the amplification of ERBB2 gene, and the green color indicates the centromere on chromosome 17 (CEP17). A total of 25 individual cells were evaluated for ERBB2 gene amplification by measuring the HER2/CEP17 ratio from each cell line. Each box shows mean ± s.d.; ****, P < 0.0001. Three biological replicated experiments showed similar results. B Whole-genome sequencing data analysis. ERBB2, MIEN1, MIR4728, and PGAP3 gene copy numbers were reduced on chromosome 17 in TDM1R and TDXdR cell lines. C Transcriptome analysis of ERBB2 gene. All ERBB2 probes were pulled out from Affymetrix Clariom D Human microarray data and clustered by differential expression. Expression indicates log2. Data were collected from three biological replicates. D Alternative splicing was increased in ERBB2, MIEN1, MIR4728, and PGAP3 genes on chromosome 17 compared to in parent cells. Transcriptome Analysis Console (TAC, Affymetrix, Inc) software used for the Affymetrix Clariom D Human microarray database to compare differential gene splicing between HER2-ADC-resistant cells and their parent cells. Data were collected from three biological replicates

Alternative splicing is also a well-known mechanism for regulating gene expression [35]. To measure the alternative splicing of HER2 mRNA in TDM1R and TDXdR BC cell lines, we first checked the binding intensity of all HER2 gene probes using the Clariom D human microarray chip. Gene expression data indicated that 93% (54 of 58) of HER2 gene probes showed reduced HER2 gene expression in both TDM1R and TDXdR BC cell lines compared to in parent cell lines (Fig. 3C), and we confirmed that the alternative splicing index of ERBB2, MIEN1, MIR4728, and PGAP3 genes was significantly increased in both types of TDM1R and TDXdR BC cell lines, resulting in mRNA reduction (Fig. 3D). Taken together, our data indicated that ADC-mediated genetic alterations such as deletion, reduction of gene copy number, and alternative splicing cause downregulation of HER2 gene expression in HER2+ BC cell lines.

There are four major potential mechanisms of resistance against ADC: 1) reduced ADC uptake due to the reduction of the target molecule, 2) efflux of the payload, 3) epigenetic modification, and 4) bypass of the payload’s antitumor effect by activation of signaling pathways [36, 37]. The mechanism of target protein reduction appears to be at play in our observation of reduced HER2 expression in TDM1R and TDXdR cell lines. To confirm whether reduced HER2 protein level is the reason for T-DXd resistance, we overexpressed HER2 in TDXdR BC cell lines and measured the antiproliferation effect of T-DXd. Overexpression of HER2 did not induce the antiproliferation effect of T-DXd (Fig. S3A). To show that reduced HER2 expression in TDM1R and TDXdR cell lines is sufficient for ADC uptake and therapeutic efficacy, we evaluated the antitumor effect of T-DXd in SUM190-TDM1R and HCC1954-TDM1R xenograft models, which have reduced HER2 expression (Fig. 2D and Fig. S8C). T-DXd significantly reduced tumor growth in both SUM190-TDM1R and HCC1954-TDM1R xenograft models (Fig. 2D, P< 0.01). These data indicated that reduced HER2 expression is not a major cause of resistance to HER2-ADC, and reduced HER2 expression in TDM1R and TDXdR BC cell lines is still sufficient to induce tumor growth inhibition by HER2-directed ADC endocytosis.

We did not observe a change in multidrug-resistant genes such as MDR1 and ABCG2 on the microarray analysis, but the epigenetic modulator EGR1 and carrier protein gene SLC6A14 were significantly elevated in TDM1R and TDXdR BC cell lines. To validate whether EGR1 or SLC6A14 is involved in HER2-directed ADC resistance, we knocked them down using RNAi in DXd-resistant BC cell lines and performed a proliferation assay with T-DXd. Silencing of EGR1 or SLC6A14 did not increase the efficacy of T-DXd in SUM190-TDXdR and HCC1954-TDXdR cell lines (Fig. S3B). These data indicated that HER2-directed ADC resistance is not caused by a reduction in intracellular DXd payload level or HER2 expression.

DNA damage response pathway can be targeted to enhance the antitumor effect of T-DXd in TDM1R and TDXdR HER2+ BC cell lines

We confirmed the presence of ERBB2 gene alterations in patients with HER2+ BC after anti-HER2 therapy and in TDM1R and TDXdR HER2+ BC cell lines. Both payloads, DM1 and DXd, are well known to induce DNA damage response pathways due to increased genotoxic stress by inhibition of DNA replication, transcription, recombination, and chromatin remodeling. Hence, we hypothesized that the DNA damage response pathway can be targeted to overcome the resistance to HER2-directed ADCs. We analyzed microarray data from TDM1R and TDXdR BC cell lines using Transcriptome Analysis Console Software to identify canonical pathways and found that DNA damage response was activated in T-DM1R and TDXdR BC cell lines (Fig. S4A). To confirm these findings, we also performed a gene set enrichment analysis and found that a DNA repair, G2M checkpoint, and mitotic spindle pathways were activated in the TDXdR cell lines (Fig. 4A and Sup. Table 8), which are known to regulate mitosis. In TDM1R BC cell lines, we only observed an activated DNA damage response pathway in SUM190-TDM1R cell line (Sup. Table 9). A further signaling network analysis showed that DNA repair pathway genes—such as non-homologous end joining, mismatch repair, nucleotide excision repair, base excision repair, and homologous recombination repair pathway–related genes—were expressed in TDM1R and TDXdR BC cell lines (Fig. S4B-E). We further determined the expression level of DNA repair pathway proteins using a reverse-phase protein array database and observed up-regulation of ATR, pATR, ATM, Chk1, Chk2, Rad50, and Rad51 in HER2-ADC-resistant cell lines compared to its parent cells (Fig. S4F-H). These data support our hypothesis that DNA damage response pathway inhibition could enhance the efficacy of HER2-directed ADC in HER2-directed ADC–resistant HER2+ BC.

Fig. 4

A gene expression analysis and synthetic lethal kinome library high-throughput RNAi screening revealed that the DNA repair pathway is a target for enhancing the efficacy of T-DXd in TDM1R and TDXdR cell lines. A Functional gene-set enrichment analysis using Affymetrix Clariom D Human Transcriptome array data. B Illustration of synthetic lethal kinome library high-throughput RNAi screening. C and D STRING interaction analysis of the top 50 target genes from kinome library high-throughput RNAi screening. K-means clustering was applied to identify groups of genes with similar behavior. Each color indicated co-regulated gene modules related to a specific canonical signaling pathway. SUM190-TDM1R (C), SUM190-TDXdR (D). E Bliss independence dose–response assay. Cells were treated with T-DXd and elimusertib for 5 days, and viability was measured using SRB staining. The data shown are representative of three independent experiments with similar results—the table indicates viability, and the Bliss synergy score was evaluated and visualized using Synergyfinderplus software (right, www.synergyfinderplus.org). F Clonogenic assay. Cells were treated with T-DXd and/or elimusertib for 14 days, and cell viability was measured by SRB staining. Data are presented as mean ± standard deviation. Two-tailed unpaired Student’s t-test. Experiments were repeated in triplicate. G. Western blotting. Cells were treated with T-DXd (1 µg/ml) and/or elimusertib (100 nM) for 48 h, and whole-cell lysates were collected for immunoblotting. Protein expression was normalized with actin level in control cells from each TDM1R and TDXdR cell line using ImageJ software. The data shown are representative of three independent experiments with similar results

We identified potential kinase targets whose inhibition could overcome HER2-directed ADC resistance or enhance the antitumor efficacy of HER2-directed ADC in TDM1R and TDXdR BC cell lines. We focused on T-DXd for the synthetic lethal screening because recent clinical trial data indicated that T-DXd is associated with improved overall survival and progression-free survival compared to T-DM1 in patients with HER2+ metastatic BC [38]; T-DXd is also now a standard second-line therapy in these patients. We performed a non-biased high-throughput RNAi screening in SUM190-TDM1R and SUM190-TDXdR cells using kinome library pooled siRNA, consisting of 2,127 siRNAs targeting 709 kinase genes (Fig. 4B). Using the Sensitivity Score analysis (described in supplemental information), we selected top 50 target kinases from each of the SUM190-TDM1R and SUM190-TDXdR cell lines (Sup. Table 10). A STRING interactome pathway analysis identified DNA repair pathway–related genes as potential targets for combination therapy with T-DXd in both TDM1R and TDXdR BC cell lines. Ataxia-telangiesctasia, mutated (ATM), ataxia telangiectasia and Rad3-related (ATR), and CDK7 were identified in the SUM190-TDM1R cell line (Fig. 4C), and ATR was identified in SUM190-TDXdR cell line (Fig. 4D). Specifically, ATR was overlapped in both cell lines. Taken together, the patient data analysis and synthetic lethal screening data indicated that the DNA repair pathway is a potential target whose inhibition could enhance the efficacy of T-DXd in TDM1R and TDXdR BC cell lines. To verify the RNAi screening result, we tested two additional RNAi targeting ATR and evaluated its synergistic antiproliferation effect with T-DXd in HER2 ADC-resistant HER2+ BC cell lines. We observed that two ATR RNAi significantly inhibited growth and showed a combination antiproliferation effect with T-DXd in tested all HER2+ BC cell lines (P < 0.05, Fig. S5A and S5B).

ATR inhibitor elimusertib enhanced the antitumor efficacy of T-DXd in TDM1R and TDXdR HER2+ BC xenograft models

Patients’ gene alteration and gene expression data (Fig. 1), microarray data in TDM1R and TDXdR cell lines (Fig. 4A and Fig. S4A), and kinome RNAi screening results (Fig. 4C and D) indicated that the PI3K, cell cycle and DNA repair pathways are potential targets for combination therapy with T-DXd in TDM1R and TDXdR HER2+ BC. To validate which target can enhance the efficacy of T-DXd in TDM1R and TDXdR cell lines, we selected specific kinase inhibitors against PI3K (alpelisib), CDK4/6 (abemaciclib), Wee1 (AZD1775), Aurora kinase A (TAS-119), ATR (elimusertib), and PARP inhibitor (olaparib) for proliferation assays. We limited drugs that were only FDA-approved or had been used for clinical trials. As single agents, abemaciclib and elimusertib showed significant growth inhibition in all tested TDM1R and TDXdR cell lines (> 60% GI), but alpelisib, AZD1775, TAS-119, and olaparib did not (Fig. S5C and S5D). In combination with T-DXd, we found that only the ATR inhibitor elimusertib showed an enhanced antiproliferation effect in both T-DM1R and TDXdR BC cell lines (Fig. S5C and S5D). To evaluate the combination effect of T-DXd and elimusertib, we conducted a Bliss independence dose–response surface model [39, 40] under the 5-day short-term treatment condition. Compared to T-DXd or elimusertib monotherapy, combination therapy significantly inhibited antiproliferation in all tested TDM1R and TDXdR cell lines (Bliss synergy score: 11.34 in SUM190; 10.54 in SUM190-TDM1R; 10.14 in SUM190-TDXdR; 17.88 in HCC1954; 5.04 in HCC1954-TDM1R; 6.68 in HCC1954-TDXdR) (Fig. 4E, Sup. Figure 5E and 5F). Additionally, we tested small molecules Gartisertib (ATR inhibitor) and AZD1390 (ATM inhibitor) to confirm the target specificity of the DNA repair pathway. Similar to elimusertib and T-DXd combination data, we observed an enhanced antiproliferation effect in combination treatment in HER2 ADC-resistant SUM190 and HCC1954 cell lines (Fig. S5G and S5H). These data indicated that the DNA repair pathway, particularly ATR, is a potential target for combination with T-DXd in TDM1R and TDXdR HER2+ BC.

Next, we examined the combination effect of T-DXd and elimusertib on long-term treatment conditions using a clonogenic assay. Compared to T-DXd or elimusertib monotherapy, combination therapy significantly inhibited antiproliferation in all tested TDM1R and TDXdR cell lines (P < 0.0001 in SUM190-TDM1R; P < 0.01 in SUM190-TDXdR; P < 0.001 in HCC1954-TDM1R; P < 0.01 in HCC1954-TDXdR) (Fig. 4F).

We next determined whether the enhanced antiproliferation effect of the combination treatment was a result of apoptosis induction by ATR inhibition. In the SUM190-TDM1R and SUM190-DXdR cell lines, the combination of T-DXd and elimusertib induced DNA damage marker pH2AX and cleaved PARP expression compared to single-agent treatment with T-DXd or elimusertib (Fig. 4G, left panel, and Fig. S6A). In HCC1954-TDM1R cells, elimusertib single-agent treatment significantly induced pH2AX. The combination of T-DXd and elimusertib did not induce an increase in the apoptosis marker cleaved PARP but did result in reduced full-length PARP expression compared to single-agent T-DXd and elimusertib. We speculate that the apoptosis pathway activated before the 48-h time point. In HCC1954-TDXdR cells, we observed enhanced cleaved PARP expression upon combination treatment, but pH2AX expression was not increased by the combination treatment compared to single-agent elimusertib (Fig. 4G, right panel, and Fig. S6A). Unlike the reduction of HER2 expression by T-DXd treatment, pHER2 expression was significantly increased by T-DXd in all tested cell lines (Fig. 4G).

We determined whether elevated pHER expression affects its downstream molecules. We first analyzed a reverse-phase protein array database and observed downregulation of HER2 and upregulation of pHER2 expression with T-DXd treatment; however, we did not observe significantly increased expression of key downstream molecules, such as pAKT, pmTOR, pS6K, p70S6K, pMEK1, and pMAPK (pERK) (Fig. S6B). Further, we validated the reverse-phase protein array data using Western blotting and confirmed that T-DXd-mediated pHER2 induction does not induce its downstream molecules, pAKT and pERK (Fig. S6C and S6D).

Our in vitro data demonstrated that ATR inhibition enhances the efficacy of T-DXd in HER2-directed, ADC-resistant HER2+ BC cells. Next, we examined the synergistic antitumor effect of T-DXd and elimusertib in SUM190-TDM1R, SUM190-TDXdR, HCC1954-TDM1R, and HCC1954-TDXdR xenograft models. The doses were 10 mg/kg for T-DXd [7] and 10 mg/kg for elimusertib [41]. Compared to vehicle control, single-agent T-DXd treatment showed significant tumor shrinkage in both SUM190-TDM1R (74% shrinkage, P < 0.0001) and HCC1954-TDM1R (75% shrinkage, P < 0.0001) xenograft models until days 35–40, when recurrence occurred (Fig. 5A and B). Elimusertib single-agent treatment did not have an antitumor effect but showed a synergistic antitumor effect with T-DXd, and this combination showed more sustained tumor shrinkage than did T-DXd single-agent treatment in both the SUM190-TDM1R (P < 0.0086) and HCC1954-TDM1R (P < 0.0383) models (Fig. 5A and B). After the completion of combination treatment, some mice showed no residual tumors (3 of 13 mice with SUM190-TDM1R and 4 of 12 mice with HCC1954-TDM1R). In the SUM190-TDXdR xenograft model, single-agent T-DXd treatment did not show tumor shrinkage but rather continual tumor growth. As in the T-DM1-resistant model, single-agent elimusertib treatment did not show tumor growth inhibition (TGI); however, elimusertib combined with T-DXd showed a significant TGI effect compared to single-agent T-DXd in the SUM190-TDXdR model (Fig. 5C, 57% TGI, P < 0.0305). In the HCC1954-TDXdR xenograft model, we did not observe significant TGI upon combination treatment with T-DXd and elimusertib compared to single-agent T-DXd (Fig. 5D, 33% TGI, P < 0.5795). However, when we performed a paired comparison analysis, only three mice bearing HCC1954-TDXdR xenografts treated with the combination showed tumor progression; the remaining five mice showed tumor shrinkage or growth inhibition in the combination treatment group.

Fig. 5

Combination treatment with T-DXd and elimusertib enhanced the antitumor effect compared with monotherapy in TDM1R and TDXdR HER2+ BC in vitro and in vivo. A-D Xenograft assay using SUM190-TDM1R (A), HCC1954-TDM1R (B), SUM190-TDXdR (C), and HCC1954-TDXdR (D). Cells were injected into the mammary fat pad of nude mice, and treatments were started when tumors were an average of 200—250 mm3. T-DXd (10 mg/kg) was administered one time on Day 0 via tail-vein injection. Elimusertib (10 mg/kg) was administered via oral gavage twice a day (6-h intervals) for 3 consecutive days per week. Data are presented as mean ± standard deviation. Left, tumor growth and tumor weight (endpoint) measurements. Table shows multiple t-tests between T-DXd and combination on each measurement date. Right, IHC images of expression levels of HER2, pH2AX, pATR, and Ki-67 in xenograft tumor tissues. Multiple t-test comparison tests were used for tumor growth. Table shows t-tests between T-DXd and a combination of elimusertib and T-DXd on each measurement date. A two-tailed unpaired Student’s t-test was used for tumor weight comparison. Scale bars = 200 µm. IHC intensity was evaluated using the ImageJ program. Each box shows the mean with standard deviation. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001. The data shown are representative of three tumor samples per group with similar results

We analyzed the expression levels of HER2, Ki-67, pATR, and pH2AX in xenograft tissues by IHC staining as observed in the Western blotting data (Fig. 4G), the expression levels of HER2, pATR, and Ki-67 were reduced in tumors from SUM190-TDM1R and HCC1954-TDM1R xenograft models treated with T-DXd combined with elimusertib compared to single-agent treatment with T-DXd or elimusertib (Fig. 5A and B). In tumors from SUM190-TDXdR and HCC1954-TDXdR models, we observed only the inhibition of proliferation marker Ki-67 by T-DXd combined with elimusertib (Fig. 5C and D). We speculate that a one-time injection of T-DXd is not sufficient to enhance tumor cell death on long-term follow-up because both SUM190-TDXdR and HCC1954-TDXdR xenograft models showed recurrence in all groups. There was no body weight loss in the mice treated with T-DXd, elimusertib, or the combination during the treatment period (Fig. S7), supporting the safety of the tested T-DXd and elimusertib treatment dose and schedule. Taken together, these results suggest that the ATR inhibitor elimusertib enhances the antitumor effect of T-DXd via the DNA damage pathway in both T-DM1- and T-DXd-resistant HER2+ BC. The IHC results demonstrated strong HER2 expression in all TDM1R and TDXdR BC xenograft models. The drug treatment started between 14 to 21 days after cell inoculation into mice. To confirm whether HER2 expression was restored during tumor engraftment in mice, we stopped T-DM1 and T-DXd treatment in all cell lines for 1 month and measured total and cell surface HER2 expression levels by Western blotting and FACS analysis, respectively (Fig. S8A and S8B), as well as IHC staining in SUM190 and HCC1954 parent tissue samples (Fig. S8C). We confirmed that HER2 levels were continually downregulated in TDM1R and TDXdR BC in cell lines and xenografts. These data indicated that the enhanced HER2 signals were caused by detection conditions without parent tissue samples.