Antitumor Activity and Biomarker Analysis for TROP2 Antibody–Drug Conjugate Datopotamab Deruxtecan in Patient-Derived Breast Cancer Xenograft Models Open Access

Trophoblast cell-surface antigen-2 (TROP2) is a transmembrane glycoprotein that functions in various signaling pathways associated with carcinogenesis and is overexpressed in many cancers, including breast cancer (1, 2). There are now several TROP2-targeted therapies in development, including TROP2-targeted antibody–drug conjugates (ADC), including sacituzumab govitecan (SG) and datopotamab deruxtecan (Dato-DXd, DS-1062a). SG was recently approved by the FDA for metastatic triple-negative breast cancer (TNBC), hormone receptor–positive (HR⁺) breast cancer, and urothelial cancer and is currently being assessed in other tumor types. Dato-DXd is currently being evaluated in several tumor types, including metastatic non–small cell lung cancer (NSCLC), HR⁺ breast cancer, and TNBC (3, 4). In the TROPION-PanTumor01 trial, the objective response rate (ORR) by blinded independent central review was 26.8% [95% confidence interval (CI), 14.2%–42.9%] and 31.8% (95% CI, 18.6%–47.6%) for patients with HR⁺/HER2⁻ breast cancer and TNBC, respectively (4). In the phase III trial TROPION-Breast01 (5), Dato-DXd significantly improved progression-free survival (PFS) compared with the investigator’s choice of chemotherapy (6.9 vs. 4.9 months; HR 0.63; 95% CI, 0.52–0.76; P < 0.0001) in 732 patients with inoperable or metastatic HR⁺ breast cancer who had received previous chemotherapy (5).

Most ADCs work by delivering cytotoxic payloads to tumor markers on the cancer cell surface, with internalization and release of the payload (6). The antitumor activity of ADCs such as mirvetuximab soravtansine targeting folate receptor α have been demonstrated to correlate with target expression (7). However, the role of TROP2 expression in antitumor activity of anti-TROP2 ADCs has been more controversial. Preclinically, higher TROP2 expression correlated with SG efficacy, although SG also had growth inhibitory effects in tumors with low to moderate TROP2 expression and deficiencies in homologous recombination repair (8). FDA approval of SG is independent of TROP2 expression levels for TNBC and HR⁺ breast cancer as well as bladder cancer, all diseases that frequently express TROP2. Efficacy of SG in TROPiCS-02 trial in HR⁺ breast was not dependent on TROP2 expression (9). In the ASCENT trial in TNBC, patients with higher TROP2 expression had a trend toward higher ORR and longer PFS and overall survival (OS) compared with those with lower expression, but SG led to improved oncologic outcomes compared with the investigator’s treatment choice even in the lower TROP2 expression cohort (10, 11). However, lack of TROP2 expression and a novel mutation (T256R) in TACSTD2 (the gene encoding TROP2) that impairs TROP2 membrane localization have been reported as mechanisms of intrinsic and acquired resistance to SG, respectively, suggesting a role of TROP2 expression and membrane localization for antitumor activity (12).

In prior work, Dato-DXd was shown to have potent antitumor activity in multiple cell lines of different tumor lineages and in NSCLC line–derived xenografts and patient-derived xenografts (PDX; 13). The antitumor activity of Dato-DXd was preclinically associated with expression of not only TROP2 but also Schlafen-11 (SLFN11), a biomarker of sensitivity to DNA topoisomerase I inhibitors (13–16). However, in TROPION-PanTumor01, objective responses were seen in patients with NSCLC treated with Dato-DXd, irrespective of TROP2 expression (3). In this study, tested the antitumor activity and pharmacodynamic (PD) effects of Dato-DXd in breast cancer PDXs (BCX) that have shown resistance to standard chemotherapy and are varied in TROP2 expression. We also functionally characterized the impact of TROP2 expression on Dato-DXd activity.

All animal experiments were approved by and performed in accordance with the Institutional Animal Care and Use Committee of MD Anderson Cancer Center. Patient provided informed consent for tissue acquisition for PDX development by image-guided biopsy or surgery. This study was approved by the Institutional Review Board at the University of Texas MD Anderson Cancer Center (lab07-0950) and conducted in accordance with Declaration of Helsinki and U.S. Common Rule. All patients provided written informed consent. PDX establishment and genomic analysis were previously described (17, 18). Tumor fragments were implanted into 6 to 8 weeks old female athymic nude mice. Briefly, a 0.3-cm incision was made, and an approximately 27-mm³ tumor fragment was inserted into the s.c. pocket. The skin was then closed with wound clips. The mice were treated when the tumor volume reached at least 200 mm³. Cell line xenografts were generated by injecting 5 × 10⁶ cells subcutaneously over the mammary fat pad of athymic nude mice. The cells were injected in a 1:1 mix (100 μL) of Matrigel and DME/F12.

Dato-DXd and isotype control IgG-DXd with equivalent drug (DXd)–antibody ratios (IgG-DXd) were obtained from Daiichi Sankyo. BCX models were treated q3w with Dato-DXd and IgG-DXd administrated by i.v. injection. Drugs were stored at −80°C in a dark freezer. The dilution buffer contained 10 mmol/L histidine buffer (pH 6.0), 9% sucrose, and 0.02% polysorbate 80. Olaparib (Chemietek) was resuspended in 5% DMSO and 30% cyclodextran and given daily by oral gavage.

In vivo xenograft growth and tumor growth inhibition was analyzed as previously described by the NCI Patient Derived Xenograft Network Consensus Guidelines (19). The % tumor volume change per time point was calculated as a relative level of tumor growth change from baseline: $δt=Vt-V0V0×100$⁠, in which $Vt$ is the tumor volume at time t and $V0$ is the tumor volume at baseline. Event-free survival (EFS) was defined as the time interval from the initiation of study to the first event or to the end of the study period for tumors that did not double in volume. The time to event was determined using linear interpolation based on the following formula: for animals for which there is no tumor volume measurement at time t but which have flanking volume measurements at time $t0$ and $t1$ such that $t0<t<t1$⁠. Then we use linear interpolation to compute the measurement. That is, we compute $ΔVt=ΔV0+β(t-t0)$⁠, in which $β=(ΔVt1-ΔVt0)/(t1-t0)$⁠. Using the Kaplan–Meier survival estimates, an EFS tumor/control (T/C) value was defined by the ratio of the median time to tumor doubling/quadrupling of the treatment group and the median time to tumor doubling/quadrupling of the control group. If the treatment group did not reach a doubling of tumor volume from initial tumor volume, then EFS T/C was defined as greater than the ratio of the last day of the study for the treatment group divided by the median time to tumor doubling or quadrupling for the control group. Log-rank test was used to compare survival distribution of the two treatment groups. All statistical analyses were performed using R software version 4.0.3. A P value < 0.05 was regarded statistically significant.

Tumors were collected rapidly and fixed in 10% neutral buffered formalin for 24 hours and washed with ethanol. Tumors were then processed by the Research Histology Core Facility (MD Anderson Cancer Center). Formalin-fixed paraffin-embedded PDX tissues of 4 micron thickness on positively charged slides were used to perform IHC using autostainers at MD Anderson and Daiichi Sankyo RD Novare. See Supplementary Tables S1 and S2 for further information including Research Resource Identifiers (RRID). In brief, after deparaffinization by each dewax solution followed by heat-mediated epitope retrieval, sections were blocked with peroxidase blocking solution. After endogenous peroxidase blocking, the slides were incubated with antibodies against the following markers: TROP2, γ-histone H2A.X (Ser139; γH2AX), human IgG (hIgG), DXd, p-KAP-1 (Ser824), cleaved caspase 3 (CC3), Ki67, and SLFN11. The antibodies were detected using each detection kit with diaminobenzidine as chromogen. After primary antibody incubation, the slides were counterstained with hematoxylin, dehydrated, and coverslipped. After staining, the slides were digitalized in a Nanozoomer S360 scanner (Hamamatsu Photonics K.K.) or an Aperio AT2 scanner (Leica Biosystems) under 20× objective magnification and evaluated by pathologists. Analysis was conducted using the digital image analysis platform using HALO v3.0 or v3.1 (Indica Labs). The images were annotated excluding necrotic areas, and a pathologist-trained specific algorithm was applied. The positive signals for each marker were evaluated using membrane module for TROP2 and hIgG and cytonuclear module for γH2AX, DXd, p-KAP-1, CC3, Ki67, and SLFN11. The percentage staining positivity (0%–100%) and staining intensity (0 = no staining, 1+ = weak staining, 2+ = moderate staining, and 3+ = strong staining) were used to generate H-scores (0–300).

Tumors were flash-frozen in liquid nitrogen and stored at −80°C. DXd concentration in tumor was determined with a validated LC-MS/MS method as previously described (20). The lower limit of quantitation was 0.01 ng/mL.

PDXs were flash frozen in liquid nitrogen and stored at −80°C. Frozen tissues were placed in bead lysis tubes (Precellys) for homogenization. Protein extraction, total protein quantification, reverse phase protein arrays (RPPA), and antibody-binding quantification were performed by MD Anderson Functional Proteomics Core Facility as previously described (21).

Whole-exome sequencing was performed on BCX.006, BCX.011, BCX.010, BCX.017, BCX.022, BCX.024, and BCX.144A by the Cancer Genomic Laboratory at MD Anderson Cancer Center. Additionally, targeted exome sequencing (T200.1) was conducted on BCX.051, BCX.055, BCX.094, and BCX.100 by the same institute. The sequencing procedures and genes included in the T200 panel have been previously described (22). Fastq files were aligned to the hg19 reference genome using Burrows-Wheeler Aligner (RRID: SCR_010910), followed by the removal of duplicate reads with Picard (RRID: SCR_006525) Somatic single-nucleotide variants and small indels were identified using VarScan2 (RRID: SCR_006849). Copy number analysis was performed using a previously established algorithm (23). For copy number calls, amplifications were defined as having an estimated copy number greater than 4, and deletions as having an estimated copy number less than 1.

Frozen tumor fragments from 17 early-passage PDX models were lysed and homogenized in lysis buffer. Total RNA was then isolated using Norgen BIOTEK Total RNA Purification Plus Kit. Genomic RNA was quantified using PicoGreen (Invitrogen, RRID: SCR_008817), and its quality was assessed via the 2200 TapeStation (Agilent, RRID: SCR_013575). RNA from each sample was reverse transcribed into double-stranded cDNA using Ovation RNA-Seq System V2 Kit (Nugen). Libraries were prepared with KAPA kits, and gene expression profiles were captured using NimbleGen whole-exome V3 probes. FASTQ files were processed using STAR (RRID: SCR_004463) with a two-step alignment procedure, as well as TopHat (RRID: SCR_013035) in conjunction with Cufflinks (RRID: SCR_014597). Additional RNA quantification was performed using HTSeq (v0.11.0) with htseq-count (v2.0.2; RRID: SCR_011867). Gene counts were normalized and used to fit a model with the R package DESeq2 (RRID: SCR_015687). Using the log-normalized counts, we characterized the expression profiles of BRCAness (24) and homologous recombination deficiency (HRD) gene signatures (25). A predictor distinguishing BRCA-like from non–BRCA-like tumors was calculated by summing the weighted expression of BRCAness genes, following the method described by Konstantinopoulos and colleagues (24).

Breast cancer cell lines BCX.010CL and BCX.011CL were generated from the PDXs BCX.010 and BCX.011 in our laboratory. Following in vitro expansion, the cell lines were validated to match the respective PDXs by short tandem repeat analysis as well as confirmed negative by Mycoplasma testing. The cell lines were grown in vitro for less than 10 passages. For TROP2 overexpression, BCX.010CL and BCX.011CL cells were transduced with control or TROP2-expressing lentivirus which were packaged in HEK-293T cells (RRID: CVCL_0063) by transfection of lentiviral TROP2 vector pLV[Exp]-Puro-TRE > hTACSTD2 (VectorBuilder, VB231019-1506zkr) or transfected with control plasmid or plasmid expressing TROP2-DDK pCMV6-entry-TROP2 (OriGene, RC202519). The transduced or transfected cells were selected by puromycin or G-418, respectively. TROP2 antibody was purchased from Abcam. Antibodies against cleaved PARP, cleaved caspase 3 (cCasp3), H2AX, γH2AX (Ser139), KAP1, p-KAP1 (Ser824), CHK1, and p-CHK1 (Ser345) were purchased from Cell Signaling Technology. Anti–β-actin antibody was purchased from Sigma-Aldrich (RRID: SCR_008988). Goat anti-rabbit Alexa Fluor-680 (RRID: AB_2535758) and goat anti-mouse Dylight-800 (RRID: AB_10703265) secondary antibodies were purchased from Life Technologies and Rockland Immunochemicals, respectively.

Two pairs of cell lines (BCX.010CL-control and BCX.010CL-TROP2 and BCX.011CL-control and BCX.011CL-TROP2) were seeded in 96-well plates at densities of about 0.6 × 10⁴ cells/100 μL per well in triplicates for each treatment dose. After adhering overnight, 100 μL of serially diluted solutions of IgG-DXd, Dato-DXd, DXd, and vehicle control were added. Cells were incubated at 37°C for 72 hours. Cells were then fixed with 50% trichloroacetic acid followed by staining with 0.4% sulforhodamine B solution. Optical densities (OD) of each well were read at 490 nm by plate reader Synergy 4 (BioTek). The IC₅₀ value was determined using CalcuSyn software (Biosoft; RRID: SCR_020251).

Cells were seeded in 6-well plates at a density of 1,000 cells per well in triplicates for each treatment group. The following day, cells were treated with vehicle control, IgG-DXd, and Dato-DXd at 1,000 ng/mL. Culture medium was changed with fresh drugs twice a week, and cells were cultured for 8 days. Cell colonies were then fixed in 10% formalin and stained with 0.05% crystal violet in 25% methanol. Following staining, colonies were imaged, and total colony area was quantitated using NIH ImageJ v.1.48 software [ImageJ (RRID: SCR_003070), U. S. NIH].

Cells were seeded at a density of 3 to 4 × 10⁵ cells per well into 6-cm plates. The following day, cells were treated with vehicle control, IgG-DXd, and Dato-DXd at 1,000 ng/mL. After 96 hours, cells in the supernatant and trypsinized-attached cells were combined and then washed with PBS. Cells were stained with annexin V-FITC and propidium iodide using eBioscience Annexin V-FITC Apoptosis Detection Kit (BMS500FI-300; RRID: AB_2575598). The labeled cells were analyzed by flow cytometry. The percentage of total apoptotic cells were summed by annexin V (A+−) and annexin V (A++) cells which represent late and early apoptotic cells, respectively.

Following treatments, cells were washed with cold PBS and lysed in 1× Laemmli buffer. The protein concentrations in the cell lysates were measured using Pierce BCA Protein Assay Kit (Thermo Fisher Scientific; (RRID: SCR_008452). The same amount protein for each sample (20–50 μg/lane) was loaded to SDS-PAGE gel, followed by transferring proteins to a 0.2-μm nitrocellulose membrane (Bio-Rad Laboratories, RRID: SCR_008426)). Membranes were blocked with blocking buffer Blocker Casein in PBS (Thermo Fisher Scientific) at room temperature for 1 hour, followed by immunoblotting with the primary antibodies in 5% BSA TBST buffer at room temperature overnight. After washing with TBST buffer, the immunoblotting membrane was then probed with the secondary antibodies with fluorescence conjugation. The immunoblots were visualized, and the immunoblotting signal intensity quantitated using the Odyssey IR imaging system (Li-Cor Biosciences).

IgG-DXd and Dato-DXd were labeled with Alexa fluorescent dye using Alexa Fluor 488 Conjugation Lightening Kit (Abcam, RRID: ab236553), following the manufacturer’s protocol. Final ADC–Alexa 488 concentration was 1 μg/μL. To observe ADC binding to the cell membrane, cells were seeded on cover slides in 6-well plates overnight and then fixed with 4% paraformaldehyde, followed by blocking with 3% BSA/PBS at room temperature for 1 hour. Cells were then incubated with ADC–Alexa 488 at 2 μg/mL at room temperature for 2 hours, followed by PBS wash twice. For internalization assay, ADC– Alexa 488 was added into the culture medium in the wells at 2 μg/mL for various times, from 10 minutes to 72 hours. Cells were fixed with 4% paraformaldehyde and washed with PBS. The cover slides were mounted onto glass slides with DAPI (4′,6-diamidino-2-phenylindole) counterstaining. Fluorescence on cell surface and in cytoplasm was monitored using an Olympus BX51 fluorescent microscope and imaged using Nuance imaging system.

RNA sequencing (RNA-seq) data are available at Gene Expression Omnibus with accession ID GSE235812. The raw sequencing data for this manuscript have been successfully submitted to Sequence Read Archive with accession number PRJNA1169861. The data generated in this study are available upon request from the corresponding author.

We assessed TROP2 cell membrane expression in 52 BCXs by IHC. TROP2 was expressed (H-score > 10) in the majority of both HR⁻ and HR⁺ models (Fig. 1A and B). However, five models had no expression, four had an H-score of 1 to 9, and 20 had a H-score of 10 to 99. Notably, of the eight models with no or very low expression (H-score < 10), five were metaplastic/spindle cell TNBC.

For 27 BCXs, we also assessed the membrane expression of TROP2 in the matched patient tumors. We found a high correlation between TROP2 membrane expression in the patient tumors and the matched PDXs (r = 0.7; P < 0.0001; Fig. 1C and D). However, the expression of TROP2 was statistically lower in the PDXs versus patients (P = 0.04; Supplementary Fig. S1).

Given the ambiguity about the role of TROP2 expression on the antitumor activity of TROP2 ADCs, we treated 11 PDXs exhibiting a range of TROP2 membrane expression (H-score from 0 to 300; Supplementary Fig. S2). We included high-TROP2–expressing models (Fig. 2A) but also enriched for low-TROP2–expressing models (Fig. 2B). Notably, 10 of the 11 PDX models were generated from residual breast tumors after neoadjuvant therapy (Supplementary Table S3). We tested two dose levels of Dato-DXd and IgG-DXd (10 and 1 mg/kg every 21 days). Dato-DXd and IgG-DXd had a dose-dependent activity.

The 11 PDXs varied in sensitivity to Dato-DXd (Supplementary Tables S4 and S5). Figure 3A summarizes the antitumor activity of 10 mg/kg Dato-DXd and the equivalent dose of IgG-DXd by showing both absolute change from baseline (top) as well as growth inhibition relative to the untreated controls (T/C ratio; bottom). Four models (36.4%) had a partial response (>30% decrease from baseline) when treated with 10 mg/kg Dato-DXd (Fig. 3A). Eight models (72.7%) had a T/C <0.4 when treated with 10 mg/kg Dato-DXd; a T/C ratio less than 0.40 equates to treatment resulting in tumor volume reaching <40% of the volume of untreated tumors at day 21. Seven models (63.6%) had prolonged EFS-2 (time to doubling) with 10 mg/kg Dato-DXd (Supplementary Table S5).

All five models with high TROP2 expression (here defined as H-score ≥200) demonstrated antitumor activity with Dato-DXd, and four of these models regressed when treated with Dato-DXd. The antitumor activity in the models with low or no TROP2 was more variable, with no model demonstrating sustained regression by day 21 with Dato-DXd. High-dose Dato-DXd caused growth inhibition in three TROP2-low models: BCX.051 (H-score 96; T/C TV: 0.31), BCX.010 (H-score 1; T/C TV; 0.15), and transient growth inhibition in BCX.144A (H-score 0; T/C TV:0.25). Notably, BCX.010 and BCX.144A also demonstrated similar growth inhibition with IgG-DXd, and interestingly, both BCX.010 and BCX.144A had high nuclear expression of SLFN11 (Fig. 3A).

IgG-DXd also had antitumor activity to a varying extent across the models. Two of the BCXs that regressed with 10 mg/kg Dato-DXd (BCX.022 and BCX.024) also had tumor regression with 10 mg/kg IgG-DXd. However, at equivalent 1 mg/kg doses of Dato-DXd and IgG-DXd, Dato-DXd had greater activity in these models. Four models had greater sensitivity to 10 mg/kg Dato-DXd compared with the equivalent dose of IgG-DXd. In total, at either 1 or 10 mg/kg, Dato-DXd had improved efficacy over IgG-DXd in five of five TROP2-high models (H-score > 200) and one of six TROP2-low models (H-score < 100). The TROP2-low model that was more sensitive to Dato-DXd than IgG-DXd had a TROP2 H-score of 96.

The models underwent genomic sequencing by whole-exome or targeted exome sequencing, The alterations observed in actionable genes in these models are shown in Fig. 3A and listed in Supplementary Tables S6 and S7. Prior studies have suggested that TOPO1 inhibitors may be more active in the context of BRCAness (26). Our study was not designed to address this question specifically, as we did not enrich for models with DNA damage response (DDR) alterations. None of the models tested had germline pathogenic alterations in BRCA1 or BRCA2. One of the Dato-DXd responsive BCXs had an ATM deletion. Supplementary Figure S3 and Supplementary Tables S8 and S9 demonstrate alterations in an expanded panel of DDR-related genes. The Dato-DXd responders and nonresponders did not differ in their expression of a reported BRCAness signature (Supplementary Figs. S4 and S5; ref. 24). Of the 230 genes in the HRD signature, 222 genes were found in our RNA-seq expression data (25). The heatmap of the expression of the 222 HRD signature genes for the PDXs are shown in Supplementary Fig. S6. Although the four sensitive models clustered more closely, this separation was not statistically significant. We had previously reported the PARP inhibitor sensitivity of several of these models by testing antitumor activity of talazoparib (18). Interestingly, the most PARP inhibitor sensitive models BCX.022 and BCX.024 were also models that we found were sensitive to not only Dato-DXd but also the control ADC, suggesting there indeed may be overlap in PARP and TOPO1 inhibitor sensitivity (Supplementary Fig. S7).

As SLFN11 expression has been associated with TOPO1 sensitivity, we assessed the expression of SLFN11 as well as TROP2 in the models by IHC (Fig. 3A; Supplementary Table S10; ref. 14). Interestingly, several models such as BCX.010, BCX.011, and BCX.144A that were TROP2-negative did not respond to Dato-DXd despite demonstrating nuclear SLFN11 expression (Figs. 2 and 3).

All four BCXs that had objective responses (tumor regression of at least 30%) to Dato-DXd had high expression of TROP2 by IHC. To statistically compare TROP2 expression, we stratified the 11 BCXs based on response to Dato-DXd into responders (BCX.017, BCX.024, BCX.022, and BCX.055) and nonresponders (BCX.006, BCX.011, BCX.094, BCX.100, BCX.051, BCX.144A, and BCX.010). The models that responded to Dato-DXd had significantly higher TROP2 protein expression by IHC (P = 0.02; Fig. 3B) and RPPA (P = 0.03; Fig. 3C). In addition, TROP2 (TACSTD2) RNA expression by RNA-seq was higher in the responders versus nonresponders (P = 0.04; Fig. 3D).

We also analyzed the relative antitumor activity of Dato-DXd compared with IgG-DXd in relationship to TROP2 H-score. We found a strong negative correlation between TROP2 H-score and the ratio of tumor volume in the Dato-DXd–treated BCXs to the tumor volume of IgG-DXd–treated BCXs at day 21 or the day of last measurements before maximum size (r = −0.87; P = 0.0004; Fig. 3E). In addition, improvement in EFS-2 correlated with TROP2 H-score (r = 0.93; P = 0.0068; Fig. 3F) for models with a significant number of events. Taken together, these data suggest that TROP2 expression correlates with increased antitumor activity of Dato-DXd compared with IgG-DXd.

Upon comparison of the functional proteomic profile of responders and nonresponders by RPPA, there was only one statistically significant difference on functional proteomics when controlled for multiple testing at a FDR of 0.5; the sensitive models had enrichment for expression of MET transcriptional regulator MACC1. Therapeutic targets, such as FAK and PARP, were enriched in responding models, whereas resistant tumors were enriched for mesenchymal features, including lower E-cadherin and higher caveolin and AXL; however, these differences did not reach statistical significance when controlled for multiple testing (Supplementary Table S11).

We sought to validate the role of TROP2 expression on Dato-DXd antitumor activity by generating isogenic cell lines that differed by TROP2 expression. We generated breast cancer cell lines BCX.010CL and BCX.011CL from the two Dato-DXd–resistant, TROP2-negative BCXs: BCX.010 and BCX.011. We then generated stable cell lines by transducing control versus or TROP2-expressing virus. Western blotting confirmed overexpression of TROP2 in the BCX.010CL-TROP2 and BCX.011CL-TROP2 cell lines compared with control cell lines BCX.010CL-control and BCX.010CL-control (Fig. 4A). The protein levels of exogenously expressed TROP2 reached TROP2 levels similar to levels observed in TROP2-positive cell line MDA-MB-468 cells. IHC of cell pellets showed strong positive staining in TROP2-expressing cell lines (Supplementary Fig. S8). We then assessed the efficacy of Dato-DXd and IgG-DXd in the matched cell lines by cell viability assay. The results showed that after 72 hours of treatment, IgG-DXd had very limited activity (IC₅₀ > 100 μg/mL) in all four cell lines. Similarly, Dato-DXd had limited activity (IC_50s > 100 μg/mL) in TROP2-negative control cell lines. In contrast, Dato-DXd had significantly enhanced antitumor activity (>10,000 folds) in the TROP2-overexpressing cell lines BCX.010CL-TROP2 and BCX.011CL-TROP2 versus control cell lines and IgG-DXd (Fig. 4B and C). To assess apoptosis, we performed annexin V staining assay. Dato-DXd significantly increased the percentage of apoptotic cells in both BCX.010CL and BCX.011CL compared with controls (Fig. 4D and E). We also assessed Dato-DXd effect on cell colony formation. The result showed that overexpression of TROP2 significantly sensitized the cells to Dato-DXd–induced inhibition of colony formation (Fig. 4F and G). The effects of Dato-DXd on DNA damage signaling and apoptosis in TROP2 cell lines was verified by immunoblotting. Dato-DXd increased apoptosis markers cleaved PARP and CC3 levels in in both BCX.010CL-TROP2 and BCX.011CL-TROP2 cells (Fig. 4H). Dato-DXd also induced DNA damage markers γH2AX, pKAP1, pCHK1, and pCHK2 in TROP2-expressing cell lines compared with control cell lines (Fig. 4H). In BCX.010CL-TROP2 cells, Dato-DXd not only induced cell apoptosis but also increased cell necrosis detected by flow cytometry (Supplementary Fig. S9A and S9B). Immunofluorescence study showed that Dato-DXd–Alexa 488 attached to the cell surface and internalization in the TROP2-overexpressing cell lines but not the matched control cell lines (Supplementary Fig. S10).

Next, we tested the antitumor activity of Dato-DXd and the control ADC against the BCX.011 isogenic pair in vivo. If untreated, these cell line xenografts uniformly maintained the respective expression status of TROP2, as shown by IHC (BCX.011CL-control median H-score = 0.3; BCX.011CL-TROP2 median H-score = 195.3; Fig. 4I). At 10 mg/kg, Dato-DXd caused regression of BCX.011CL-TROP2 xenografts but had almost no effect BCX.011CL-control xenografts (Fig. 4J and K; Supplementary Tables S12 and S13). After 28 days of treatment, we collected the available tumors and performed immunoblotting of tissue lysates of these tumors. Whereas all BCX.011CL-control tumors regardless of treatment did not express TROP2, BCX.011CL-TROP2 xenografts did express TROP2 with the exception of a single tumor that regrew after first regressing with 10 mg/kg Dato-DXd treatment (Supplementary Fig. S11A and S11B).

In this study, we also created isogenic TROP2 cell line models with plasmid transfection which expresses TROP2 with a DDK tag at C′ terminal (Supplementary Fig. S12). Overexpression of TROP2 from both systems confers the cells high sensitivity of Dato-DXd treatment in vitro and in vivo (Fig. 4; Supplementary Fig. S12; Supplementary Tables S12 and S13). Interestingly, we found that whereas TROP2 expression from lentivirus is steady along with the passages (Supplementary Fig. S13 and S13A), the expression levels of TROP2-DDK from the plasmid cell line decreased in the cells with multiple passages detected by immunoblotting (Supplementary Fig. S13B and S13C). However, the xenograft derived from the plasmid cell line in which TROP2 was gradually decreased was still able to respond to Dato-DXd with lower TROP2 levels, suggesting that even a modest increase in TROP2 expression can enhance Dato-DXd activity (Supplementary Fig. S12D and S12E).

To identify early PD markers of response for Dato-DXd, we chose a set of TROP2-expressing BCXs based on sensitivity to Dato-DXd. For IHC profiling, we chose a Dato-DXd–resistant model, BCX.006, as well as two Dato-DXd–sensitive models, BCX.055 and BCX.017; BCX.055 and BCX.017 are both more sensitive to Dato-DXd compared with the IgG-DXd. We treated the models with 10 mg/kg Dato-DXd or 10 mg/kg IgG-DXd and collected replicate tumors 24 and 72 hours after treatment.

We assessed expression of TROP2 and γH2AX, as well as the distribution of hIgG and DXd (Supplementary Fig. S14). As expected TROP2 expression was high in BCX.055 and BCX.017 compared with BCX.006; there was no difference in TROP2 expression after 72 hours of Dato-DXd treatment (Fig. 5A). Dato-DXd treatment increased γH2AX expression significantly in both the Dato-DXd–responsive models, BCX.055 and BCX.017, versus both untreated tumor and IgG-DXd tumors. However, there was no change in γH2AX in the Dato-DXd–resistant model BCX.006 (Fig. 5B). There was a smaller but statistically significant increase in γH2AX with the IgG-DXd at 72 hours in the BCX.055 model.

At both 24 and 72 hours, the percentage of hIgG-positive cells was significantly higher with Dato-DXd treatment compared with IgG-DXd treatment in BCX.055 and BCX.017 but not in BCX.006 (Fig. 5C). However, there was a slight increase in hIgG-positive cells in BCX.006 treated with both IgG-DXd and Dato-DXd at 24 hours. In addition, the hIgG IHC staining following Dato-DXd was significantly higher overall in BCX.017 and BCX.055 versus BCX.006 at both timepoints. In BCX.055 and BCX.017, IHC demonstrated significantly higher percentage of DXd-positive cells compared with IgG-DXd at both 24 and 72 hours (Fig. 5D). At 72 hours, BCX.006 demonstrated higher levels of DXd-positive cells in both IgG-DXd– and Dato-DXd–treated tumors. DXd IHC showed a higher level of DXd-positive cells in BCX.017 and BCX.055 Dato-DXd–treated tumors compared with BCX.006 Dato-DXd–treated tumors. LC-MS/MS measurement of DXd concentrations in treated tumors also revealed statistically higher concentrations of DXd in Dato-DXd–treated BCX.055 and BCX.017 compared with both IgG-DXd–treated tumors and Dato-DXd–treated BCX.006 xenografts (Fig. 5E). We also assessed pKap1, Ki-67, and CC3 by IHC (Supplementary Fig. S15). pKap1 increased in both responder models at 72 hours but not in the nonresponder model. There was a modest CC3 increase at 72 hours in all three models.

To assess PD changes more broadly, we analyzed functional proteomics of the above models as well as an additional nonresponder model, BCX.094, using RPPA on PDX tumors treated with Dato-DXd at 24 or 72 hours At an FDR of 0.1, there were 43 Dato-DXd–induced changes in protein expression in responding models and 32 Dato-DXd–induced changes in the nonresponding model (Fig. 5F; Supplementary Tables S14 and S15). A total of 16 changes were commonly induced in both responders and nonresponders by Dato-DXd (Fig. 5G; marked red on Supplementary Tables S14 and S15). Consistent with the IHC results, the sensitive models demonstrated an increase in γH2AX as well as pCHK1, cleaved caspase 7, as well as p53 (Supplementary Table S14). In Dato-DXd–sensitive models, we also noted a significant increase in cyclin E1 expression, a finding we have previously reported with trastuzumab deruxtecan and DXd treatment in HER2-positive cell lines (27). In addition, although the BCXs were treated in immune-deficient nu/nu mice, the sensitive models demonstrated changes on immune-related markers (PD-L1, IL-6, lipocalin 2, and STING). The resistant BCXs had significant upregulation of cyclin B1/CDK1 signaling, as shown by an increase in cyclin B1, pCDK1 T14, PLK1 and Aurora B (Supplementary Table S15).

As we demonstrated that Dato-DXd activated DDR signaling, we hypothesized that the efficacy of Dato-DXd can be enhanced by combining it with PARP inhibitors. We selected four TROP2-expressing BCXs to test the combination of Dato-DXd and olaparib: a model with high Dato-DXd sensitivity (BCX.022), two models with moderate Dato-DXd sensitivity (BCX.100 and BCX.051), and a model with no sensitivity to Dato-DXd (BCX.006). Models were treated with 1 or 10 mg/kg and 3 or 10 mg/kg Dato-DXd or equivalent doses of IgG-DXd in combination with olaparib (100 mg/kg/day).

The high–Dato-DXd–sensitive model, BCX.022, regressed with 3 mg/kg Dato-DXd alone and in combination with olaparib (Fig. 6A; Supplementary Tables S16 and S17). However, the combination of 1 mg/kg Dato-DXd led to significantly more durable response compared with equivalent single agents (1 mg/kg Dato-DXd and olaparib vs. 1 mg/kg Dato-DXd EFS-2; P = 0.003). For the two models moderately sensitive to Dato-DXd, the combination of olaparib and Dato-DXd had significantly better antitumor activity compared with single agents at the higher dose of Dato-DXd [BCX.051 (Dato-DXd 3 mg/kg and olaparib vs. Dato-DXd alone EFS-4; P = 0.031) and BCX.100 (Dato-DXd 10 mg/kg and olaparib vs. Dato-DXd alone EFS-4; P = 0.015; Fig. 6B–C; Supplementary Tables S16 and S17)]. However, in the Dato-DXd nonresponder (BCX.006) model, neither the single agents nor the combination had antitumor activity (Fig. 6D).

TROP2 ADC Dato-DXd has emerged as a well-tolerated and effective therapeutic option for breast cancer (4). However, TROP2 is expressed in many tumor types, thus there is a great need to identify determinants of antitumor activity for Dato-DXd. Using a cohort of BCXs enriched for low-TROP2–expressing tumors, as well as isogenic cell lines without TROP2 expression, we have demonstrated that TROP2 expression is associated with response to Dato-DXd but that tumors may also vary in their sensitivity to IgG-DXd, suggesting potential variations in payload sensitivity or other mediators of sensitivity.

The importance of target expression has been well documented with some ADCs such as folate receptor α–targeted ADC mirvetuxumab (10). In contrast, there has been more controversy about the role of TROP2 expression for TROP2 ADCs. In the ASCENT trial, patients with TNBC treated with SG had an ORR of 44%, 38%, and 22% for high (H-score >200–300), medium (H-score 100–200), and low TROP2 expression (H-score 0 to <100; ref. 10), respectively. In total, 1 (14%) of 7 patients with an H-score of 0, and 3 (15%) of 20 patients with an H-score under 50 had a response (11). Upon final analysis, there was significant benefit for SG in patients with the two highest quartiles of TROP2 expression (220–275 and 275–300), The interaction between treatment and continuous TROP2 expression for OS was marginally significant (range, P = 0.04–0.05), and the interaction between treatment and TROP2 H-score expression by quartiles for OS was trending toward significance (P = 0.054). Taken together, the data demonstrated that the antitumor activity of SG may be greater in patients with higher TROP2 expression; however, in all TROP2 expression category cohorts, the efficacy of SG was better than the investigator’s choice of chemotherapy. In the TROPiCS-02 trial, when SG was compared with investigator-choice treatment in patients with HR⁺ breast cancer, improvement of PFS with SG treatment was observed in both TROP2 H-score <100 patients (median PFS 5.3 vs. 4.0 months) and higher-expressing patients (median PFS 6.4 vs. 4.1 months, HR 0.60). An overall survival benefit was seen with SG in both groups, further raising doubt about the role of TROP2 expression in the mechanism of action of SG (9).

Given the potency of deruxtecan as a payload, there has been debate about the role of target expression. In the TROPION PanTumor01 lung cancer cohort, TROP2 ADC Dato-DXd had activity across lung cancers with differing TROP2 expression levels (3). Similarly, objective responses were observed with HER3 ADC patritumab deruxtecan in EGFR-mutant lung cancer across a variety of expression levels (28). Notably, not only has T-DXd shown activity in HER2-low tumors in the DESTINY Breast-04 trial but also in HER2 IHC 0 patients in the DAISY trial, leading to debate as to whether HER2 expression is truly needed for efficacy (27, 28). However, T-DXd activity in the HER2-low setting is known to be more limited in some tumor types such as colorectal cancer (29). DESTINY PanTumor02 demonstrated that T-DXd had robust activity across a variety of HER2-expressing tumors, with an overall ORR of 37.1% in all patients enrolled with HER2 2+ or 3+ expression and with ORRs of 61.3% and 27.3% for patients who had central confirmation of HER2 3+ and 2+ expressions, respectively (30). T-DXd received accelerated FDA approval for unresectable or metastatic HER2 3+ solid tumors. Notably T-DXd was active in tumor agnostic fashion in tumors with 3+ expression, whereas the activity did vary across histologies with 2+ expression. It is unclear whether this due to possible differences in payload sensitivity, prior treatment history, or other factors.

In our study, we enriched our PDX cohort and demonstrated that four of five models with high TROP2 expression had durable responses and that introducing TROP2 into nonsensitive models introduces sensitivity, confirming the role of TROP2 for membrane binding and internalization. However, we also demonstrated that some non-TROP2–expressing models had growth inhibition often similar to the extent seen with IgG-ADC. We also have seen differences in sensitivity to IgG-ADC, with a few Dato-DXd–sensitive models also showing exquisite sensitivity to IgG-DXd, which may be due to differential sensitivity to DXd or other yet unknown factors. Our model selection did not allow us to test whether there is differential sensitivity in models with HRD. Further study is needed into the determinants of response to deruxtecan and Dato-DXd. Future studies can especially pursue TROP2-low tumors to evaluate strategies such as increasing TROP2 expression and/or enhancing payload sensitivity (31).

With increasing importance of ADCs in our armamentarium, there is a need to identify rational combinations to enhance antitumor activity and overcome intrinsic and acquired resistance. Our work was done in immune-deficient mice, but upregulation of STING signaling and PD-L1 expression may provide rationale for immunotherapy combinations. This finding supports the substantial clinical activity seen in the BEGONIA trial in which Dato-DXd was combined with durvalumab in first-line metastatic breast cancer (32).

Our in vivo PD data confirm the expected changes in DDR markers. The combination experiments demonstrated that PARP inhibitor combination enhanced in vivo antitumor activity in three of four PDX models. Thus, the combination of Dato-DXd with PARP inhibitors is a high-priority combination. However, the combination of sacitizumab govitecan with PARP inhibitors rucaparib and talazoparib demonstrated overlapping toxicity with significant cytopenia, with the latter combination ultimately being pursued with a modified schedule (33, 34). Dato-DXd has less hemotologic toxicity; its combination with selective PARP inhibitor saruparib is ongoing. In addition, the PD data on RPPA also highlighted multiple other changes, such as upregulation of cyclin E in Dato-DXd–sensitive models and PLK1 and Aurora B in resistant models. These and other changes may represent other potential combinatorial targets.

In summary, our study demonstrated that Dato-DXd is active in breast cancer PDXs predominantly generated from tumors resistant to standard chemotherapy. Our study was enriched for TROP2-low models and thus was able to demonstrate enhanced responses in TROP2-high PDXs, a finding that was also confirmed with isogenic paired cell lines varying in TROP2 expression. These findings confirm there is indeed an important contribution of TROP2 expression to Dato-DXd activity. However, there may be a growth-inhibitory effect also for low-TROP2–expressing tumors, and further study is needed to identify determinants of DXd sensitivity and hypothesis-driven combinations.

F. Meric-Bernstam reports grants and personal fees from AstraZeneca and Daiichi Sankyo during the conduct of the study as well as personal fees from Becton Dickinson, Calibr (a division of Scripps Research), Dava Oncology, Debiopharm, EcoR1 Capital, eFFECTOR Therapeutics, Elevation Oncology, Exelixis, GT Aperion, Incyte, Jazz Pharmaceuticals, LegoChem Biosciences, Lengo Therapeutics, Menarini Group, Molecular Templates, Protai Bio, Ribometrix, Roche, Tallac Therapeutics, Tempus, Zymeworks, Biovica, Cybrexa, FogPharma, Guardant Health, Harbinger Health, Karyopharm Therapeutics, LOXO-Oncology, Mersana Therapeutics, OnCusp Therapeutics, Sanofi Pharmaceuticals, Seagen, Theratechnologies, and Zentalis Pharmaceuticals; grants from Jazz Pharmaceuticals, Zymeworks, Aileron Therapeutics, Inc., Bayer Healthcare Pharmaceutical, Calithera Biosciences Inc., Curis Inc., CytomX Therapeutics Inc., Debiopharm International, eFFECTOR Therapeutics, Genentech Inc., Guardant Health Inc., Klus Pharma, Takeda Pharmaceutical, Novartis, Puma Biotechnology Inc., and Taiho Pharmaceutical Co.; personal fees and other support from Dava Oncology; and other support from European Organization for Research and Treatment of Cancer (EORTC), European Society for Medical Oncology (ESMO), and Cholangiocarcinoma Foundation outside the submitted work. In addition, F. Meric-Bernstam has a patent for WO2023060283A2 pending. K.W. Evans reports grants from the NCI and other support from Daiichi Sanchyo during the conduct of the study. T. Maejima reports personal fees and nonfinancial support from Daiichi Sankyo, Inc./Daiichi Sankyo, Co., Ltd. during the conduct of the study and outside the submitted work. T. Karibe reports personal fees from Daiichi Sankyo Co., Ltd. during the conduct of the study and outside the submitted work. L.A. Byers reports personal fees from AstraZeneca and Daiichi Sankyo outside the submitted work. D. Okajima reports personal fees from Daiichi Sankyo Co., Ltd. during the conduct of the study and outside the submitted work; in addition, D. Okajima has a patent for “Anti-TROP2 antibody–drug conjugate” issued to Daiichi Sankyo Co., Ltd. S. Damodaran reports grants from Guardant Health, Taiho Pharmaceuticals, Novartis, Serono, Sermonix, AstraZeneca, and Medilink outside the submitted work. No disclosures were reported by the other authors.

F. Meric-Bernstam: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. E. Yuca: Data curation, formal analysis, investigation, visualization, writing–review and editing. K.W. Evans: Data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. M. Zhao: Data curation, formal analysis, supervision, investigation, visualization, methodology, writing–original draft, writing–review and editing. T. Maejima: Data curation, formal analysis, investigation, visualization, methodology, writing–review and editing. T. Karibe: Data curation, formal analysis, investigation, visualization, writing–review and editing. M.G. Raso: Data curation, formal analysis, supervision, investigation, visualization, methodology, writing–review and editing. X. Tang: Data curation, investigation, visualization, writing–review and editing. X. Zheng: Data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. Y.Q. Rizvi: Data curation, investigation, visualization, writing–review and editing. A. Akcakanat: Investigation, methodology, writing–review and editing. S.S. Scott: Investigation, writing–review and editing. B. Wang: Data curation, investigation, writing–review and editing. L.A. Byers: Investigation, visualization, methodology, writing–review and editing. D. Tripathy: Investigation, writing–review and editing. D. Okajima: Conceptualization, resources, funding acquisition, validation, investigation, methodology, project administration, writing–review and editing. S. Damodaran: Validation, investigation, writing–review and editing.

We would like to acknowledge Susanna Brisendine for help in manuscript preparation and submission. This work was a funded collaboration between Daiichi Sankyo Co., Ltd. and MD Anderson Cancer Center. This work was also funded in part by a METAvivor Translational Research Award, the PDX Development and Trial Center U54 #CA224065, NIH T32 National Cancer CA009599, The MD Anderson Cancer Center Sheikh Khalifa Bin Zayed Al Nahyan Institute for Personalized Cancer, MD Anderson Breast Cancer Moonshot Program, NIH Clinical Translational Science Award 1UL1TR003167, the Nellie B. Connally Breast Cancer Research Endowment, the Barr funds, and the MD Anderson Cancer Center Support Grant (P30 CA016672).