These exploratory biomarker analyses of patients from the primary and exploratory cohorts of the DESTINY-Gastric01 trial identified relevant prognostic or predictive biomarkers in patients with HER2+ or HER2-low gastric and GEJ cancer treated with T-DXd. Patients with higher levels of HER2-associated biomarkers, including HER2 IHC positivity, HER2 ISH positivity, tumor HER2 mRNA levels, plasma gene apCN and/or serum HER2ECD, had numerically higher ORR compared with patients with lower HER2-associated biomarkers. For example, patients with plasma HER2 amplification in ctDNA had an ORR of 61% compared with 34% in patients with no amplification. These results suggest that T-DXd has activity even in patients with HER2-low tumors, which confirms previous findings from the HER2-low DESTINY-Gastric01 exploratory cohorts in which the confirmed ORR was 26.3% (95% CI 9.1–51.2%) and 9.5% (95% CI 1.2–30.4%) in the IHC 2+/ISH− and IHC 1+ cohort, respectively5. However, some discordance was observed between plasma HER2 amplification and HER2 IHC positivity in tumor tissue collected before T-DXd treatment in the current study (Extended Data Table 2). Although probably due to a technological limitation in detecting plasma HER2 amplification or tumor heterogeneity, it is important to note there is a possibility that location of metastasis is related to levels of ctDNA14. This may also play a role in discordance between tumor and plasma biopsy results. However, NPA was high, which may support the use of ctDNA to guide treatment if there are no available tissue samples. Further investigation is warranted to validate whether plasma HER2 amplification could replace tissue IHC/ISH in HER2 scoring.

In the current study, tumor HER2 mRNA level, plasma gene apCN and HER2ECD level were all correlated with HER2 status in tumor biopsy samples assessed with IHC/ISH. Although HER2 amplification and associated overexpression in gastric and GEJ cancer is poorly understood, our results might provide some insights into this relationship15. Alterations in other genes were found to overlap with HER2 amplification in our analysis, as determined by ctDNA analysis at baseline. Although bTMB did not appear to have a major effect on response to treatment, patients with alterations in key signal transduction genes (MET, EGFR, FGFR2 or PIK3 GoF) had numerically lower ORR. There were trends for worse outcomes in patients with MET, EGFR and FGFR2 amplification; however, interpretation is limited by the small number of patients with these alterations and the lack of a control arm in our study. It remains unknown whether these trends are related to resistance to T-DXd or are simply prognostic effects.

We found no significant relationship between PD-L1 RNA expression and T-DXd efficacy. However, given recent findings from the KEYNOTE-811 trial16, further investigations are warranted to explore the relationship between PD-L1 IHC, HER2 IHC and the efficacy of T-DXd. Our data suggested there might be higher ORR in the HER2-low gastric cancer cohort of patients with high levels of plasma HER2ECD compared with those with low HER2ECD levels. However, these results are not conclusive because of the small number of patients in the exploratory cohorts, and confirmation in additional HER2-low cohorts would be required for validation. This might be provided in the ongoing DESTINY-Gastric03 (NCT04379596) and EPOC2203 (jRCT2031230477) studies, which evaluated T-DXd combinations for HER2-low gastric cancer.

HER2 ctDNA amplification was detected in approximately half of the patients in this study (45% (37 of 82)) before T-DXd treatment and in 33% of patients (27 of 82) by EoT. This is consistent with previous observations for other HER2-targeting therapies for gastric cancer such as trastuzumab and lapatinib17,18.

It is unlikely that HER2 expression is lost on a per-cell basis. Instead, it is possible that the tumor was heterogeneous at baseline, with the ratio of HER2-expressing to -nonexpressing cells changing during treatment, which could indicate that non-HER2-amplified cells predominated in the tumor. For these patients, switching to non-HER2-targeted therapy may be an appropriate choice; however, further investigation in this subgroup of patients would be required to confirm this.

A previous trial of T-DXd identified several variants in nonsmall cell lung cancer, including exon 20 insertions and variants at position 310 (ref. 19). In the current study, although we did not observe exon 20 insertions, there were variants at position 310. In addition, 11.0% (12 of 109) of patients had HER2 GoF variants and all patients with HER2 GoF variants had HER2 IHC 3+ or IHC 2+/ISH+ status. Patients with HER2 IHC 3+ and HER2 mutation had a higher ORR (87.5%, seven of eight) than those with HER2 IHC 3+ in the overall population (58.2%, 53 of 91), suggesting that patients with HER2 mutation may be more sensitive to T-DXd, even in those with HER2 IHC 3+ status. Analysis of ctDNA carried out at EoT identified three cases of acquired variants in the TOP1 gene (E709G, L429R and D533G) (Extended Data Fig. 2). The TOP1 gene encodes DNA topoisomerase I, which is the direct target of deruxtecan. Patients with triple-negative metastatic breast cancer reported variants in TOP1 with resistance to sacituzumab govitecan20. Sacituzumab govitecan is an ADC comprising an anti-Trop2 conjugated to another DNA TOP 1 inhibitor20. Of the three variants identified, D533G was previously found to be resistant to DNA topoisomerase I inhibition by camptothecin21. Structural analysis indicated that the D533 residue is in direct contact with camptothecin when it forms a complex with DNA and topoisomerase 1 (ref. 22). The D533G mutation may also provide resistance to deruxtecan, which is a derivative of camptothecin, although this theory needs to be confirmed by in vitro experiments. The E709G and L429R mutants have not previously been reported and neither the effect of these two variants on DNA topoisomerase I enzymatic activity, nor on its inhibition by camptothecin or deruxtecan, is known23. Although the presence of these acquired TOP1 mutations suggests that direct mutations of TOP1 may contribute to T-DXd resistance through provision of deruxtecan resistance, this appears to be a relatively rare event. In patients with acquired deruxtecan resistance but with maintained HER2 positivity, additional therapy with a HER2-targeting ADC incorporating a different payload might overcome such resistance.

Our study identified several genes with differential expression in T-DXd responders compared with nonresponders. HER2 and several other genes in the chromosome 17q12-21 region were expressed at higher levels in T-DXd responders compared with nonresponders. The expression level of those genes was associated with HER2 apCN, and increased expression levels of the genes on chromosome 17q12-21 is probably due to overall amplification of this region. Ultimately, the association of these genes with patient outcomes may be a secondary effect of the genomic amplification of this region and HER2 overexpression.

Results from studies in other cancer types have suggested that identification of specific expression patterns may provide prognostic benefit and allow for tailored therapeutic regimens24. The relationship between the timing of tumor sample collection used for assessment of HER2 IHC/ISH status and patient outcomes revealed that clinically meaningful ORR was observed regardless of the timing of sample collection in our study (that is, before or after/during the first Tmab treatment). HER2-targeted treatment can reduce the number of HER2+ cells (decreased HER2 expression following treatment with trastuzumab or other HER2-targeted agents has been observed in 16–32% of patients)10,11,12,25 and, consequently, confirmation of HER2+ following trastuzumab therapy is important before initiating another HER2-directed therapy. However, because the current analysis was from a third- or later-line study, there was a period of non-HER2-targeting therapy between the last dose of Tmab and initiation of T-DXd treatment. It is possible that HER2+ cells could regrow during this period of non-HER2-targeting therapy11. As the OS results suggest, HER2 status at baseline might be predictive of the efficacy of T-DXd. Therefore, although a fresh biopsy sample is useful for second-line therapy, it may have lower utility for third- or later-line therapy.

Limitations of these biomarker analyses include the fact that they were either exploratory or post hoc analyses with a small sample size conducted in the absence of external validation and data from a control arm. However, one strength of these exploratory analyses is that the data are from a robustly designed clinical trial of T-DXd compared with chemotherapy in patients with HER2+ advanced gastric cancer; the analyses presented are from patients in the T-DXd arm of the primary cohort as well as from those in the two exploratory cohorts, all of whom received T-DXd4. The biomarkers identified in this analysis are being investigated and validated in additional studies, including DESTINY-Gastric02 (ref. 6), DESTINY-Gastric03 (NCT04379596), DESTINY-Gastric04 (NCT04704934) and EPOC2203 (jRCT2031230477).