One of the core values of ADCs lies in chemically linking highly specific mAbs to potent cytotoxic drugs through linkers. This design enables the precise delivery of cytotoxic agents to tumour cells by leveraging the specificity of the antibody for target antigens while minimizing off-target effects on healthy tissues. ADCs can be used to overcome the low target specificity of traditional chemotherapies, thereby increasing therapeutic efficacy while minimizing side effects. In the DESTINY-CRC01 trial, CRC patients with HER2 expression (IHC 3+ or IHC 2+/ISH+) achieved a significant objective response rate (ORR) with T-DXd, whereas no objective responses were observed in patients with low HER2 expression levels (IHC 2+/IHC 1+/ISH−) (Yoshino et al., 2023). These results indicate that in addition to precise targeting by mAbs, ADCs can also align with target expression levels to support individualized treatment strategies.

Linker design enables responsive release triggered by tumour microenvironment (TME) conditions, such as acidic pH, enzyme activity, or reductive stress, which facilitates individualized control of release kinetics to adapt to tumour heterogeneity. Beyond classical cleavable linkers, researchers have developed smart responsive linkers that activate payload release by sensing specific TME features. For example, hypoxia-activated azobenzene linkers remain stable in normal tissues (O₂ > 10%) but are cleaved in hypoxic tumour microenvironments (O₂ < 1%), releasing monomethyl auristatin E (MMAE) and fully restoring ADC cytotoxicity (Xiao et al., 2024). In addition, differentiated release strategies using linker combinations have enabled temporally controlled release in dual-payload ADCs. For instance, KH815 employs a combination of irinotecan (a topoisomerase I inhibitor) and triptolide (an RNA polymerase II inhibitor). The former is rapidly released inside tumour cells via lysosomal enzyme-cleavable linkers, directly disrupting DNA replication, whereas the latter is triggered by pH-sensitive linkers in the TME to target stromal or vascular cells, achieving a “dual-hit” mode of tumour cell killing plus microenvironment remodelling (Mullard, 2025). Such innovations enable the release patterns of ADCs to be adjusted on the basis of individual tumour TME characteristics, laying the foundation for personalized therapy.

Traditional ADCs have primarily relied on microtubule inhibitors, topoisomerase inhibitors, or DNA-damaging agents. In CRC, patient stratification is driven mainly by biomarker-defined targets (e.g., HER2, GCC, CEACAM5) and tumour biology, whereas payload class modulates efficacy and safety within these target-positive populations—particularly in the setting of heterogeneous antigen expression and acquired resistance. Payload diversification has therefore become a core strategy for improving the therapeutic index and overcoming heterogeneity/resistance (Beck et al., 2017; Conilh et al., 2023). Table 2 summarizes representative ADC payload classes and clinical selection considerations in CRC. In addition to conventional payloads, non-traditional payloads—including immunostimulatory agents (e.g., TLR/STING agonists), RNA polymerase inhibitors (e.g., amatoxins), and apoptosis modulators (e.g., Bcl-xL inhibitors)—may help address resistance by engaging antitumour immunity or targeting tolerant/dormant tumour-cell states (Wang et al., 2025b; Grairi and Le Borgne, 2024). Dual-payload ADCs (e.g., KH815) that combine mechanistically distinct warheads are also being explored to deliver a “dual-hit” effect on core tumour vulnerabilities such as replication and transcription (Mullard, 2025). Overall, payload diversification complements target-based stratification and is driving ADCs towards greater precision and efficacy in CRC.

The pharmacokinetic properties and overall therapeutic index of an ADC are profoundly influenced by its drug-to-antibody ratio (DAR) (Dumontet et al., 2023). The DAR defines the average number of payload molecules conjugated per antibody, and achieving an optimal balance is essential to ensure sufficient payload delivery without compromising antibody integrity (Chau et al., 2019). Early ADCs developed using conventional nonsite-specific conjugation methods typically yielded heterogeneous mixtures. Fractions with high DARs (≥4) exhibited stronger in vitro cytotoxic activity but exhibited increased hydrophobicity, resulting in accelerated clearance, shortened systemic exposure, and a narrower therapeutic window. In contrast, fractions with low DARs carried fewer payloads, leading to inadequate antitumour activity (Fu et al., 2022; Chis et al., 2024). The advent of site-specific conjugation strategies—such as engineered cysteine conjugation, enzymatic peptide ligation, and glycan-based conjugation—has enabled the generation of ADCs with homogeneous and controlled DARs, which are typically set at 2, 4, or 8 (Khongorzul et al., 2020; Samantasinghar et al., 2023; Choi et al., 2024; Li M. et al., 2024). These approaches allow precise and stoichiometric attachment at defined positions, markedly improving the homogeneity, stability, and activity of ADCs. Modern ADCs further integrate site-specific conjugation with tumour microenvironment–responsive linkers to achieve optimized pharmacokinetic profiles, maintaining stability and tolerability in systemic circulation while ensuring efficient payload release in tumour tissues, thereby broadening the therapeutic window (Drago et al., 2021).

HER2, also known as ErbB2, is a member of the EGFR family (Marchiò et al., 2021); HER2 forms heterodimers with HER3 to regulate specific signalling pathways within the tumour microenvironment, including activation of the PI3K/AKT/mTOR pathway (Díaz-Serrano et al., 2018; Miricescu et al., 2020), stimulation of the MAPK/ERK cascade through RAS activation (Su, 2018), and inhibition of GSK3β to maintain Wnt/β-catenin signalling (Liu et al., 2018; Kim Y. et al., 2023). These signalling cascades collectively govern tumour proliferation, epithelial–mesenchymal transition, metabolic reprogramming, therapeutic resistance, and metastasis (Shimozaki et al., 2024). HER2 also contributes to immune modulation. It induces the expression of cytokines such as interleukin-6, which activates the JAK/STAT pathway, leading to the upregulation of programmed death-ligand 1 (PD-L1) expression and promoting cancer stem cell self-renewal (Zhang et al., 2016; Hassan and Seno, 2022). Concurrently, HER2 signalling results in the recruitment of immunosuppressive cells, including regulatory T cells (Tregs) and M2-polarized macrophages, thereby reshaping the immune microenvironment to facilitate tumour immune evasion (Jing et al., 2023; Jiang et al., 2024). HER2 is overexpressed in multiple solid tumours, including gastric cancer (Joshi and Badgwell, 2021), colorectal cancer (Suwaidan et al., 2022), and breast cancer (Oh and Bang, 2020). Importantly, HER2 has emerged as one of the most successful therapeutic targets across diverse malignancies (Arteaga et al., 2011; La Salvia et al., 2019; Xie et al., 2020).

HER3, also known as ErbB3, is another member of the EGFR family (Sierke et al., 1997) and is overexpressed in various solid tumours, including CRC (Campbell et al., 2010). HER3 can be phosphorylated through heterodimerization with other receptor tyrosine kinases, such as HER2(Olayioye et al., 2000; Campbell et al., 2010), thereby activating the PI3K/AKT and MAPK/ERK signalling pathways (Yarden and Sliwkowski, 2001; Baselga and Swain, 2009; Ocana et al., 2012; Rathore et al., 2025). Given its high expression in tumours and efficient internalization upon antibody binding, HER3 is considered a promising target for ADC development(Haikala and Jänne, 2021).

Human epidermal growth factor receptor (EGFR), also known as ErbB1/HER1, is a member of the ErbB family (Kumagai et al., 2021). In the absence of specific ligands, EGFR remains inactive (Yarden and Sliwkowski, 2001). The binding of ligands to the extracellular domain induces homodimerization or heterodimerization with other ErbB RTKs, triggering phosphorylation of the tyrosine kinase domain. This activates downstream signalling pathways, including the RAS/RAF/MAPK pathway (Hynes and Lane, 2005) and the PI3K/AKT pathway (Glaviano et al., 2023). Currently, anti-EGFR monoclonal antibodies such as cetuximab and panitumumab are effective only in a small subset of cancer patients (Ciardiello and Tortora, 2008; Bokemeyer et al., 2011; Douillard et al., 2014; Van Cutsem et al., 2015). Therefore, EGFR-targeted ADCs may represent a novel strategy to overcome these limitations.

Guanylyl cyclase C (GCC) is the receptor for diarrhoea-inducing heat-stable enterotoxins as well as the endogenous ligands guanosine and uridine (Waldman and Camilleri, 2018). In normal gastrointestinal tissue, GCC expression is restricted to the apical surface of epithelial cells at tight junctions, where it is isolated from systemic circulation (Carrithers et al., 1996). Upon ligand binding, GCC undergoes receptor-mediated endocytosis (Urbanski et al., 1995). In contrast, GCC is expressed in more than 95% of primary and metastatic colorectal cancers and in approximately 65% of oesophageal, gastric, and pancreatic tumours (Urbanski et al., 1995; Waldman et al., 1998; Buc et al., 2005). As such, the GCC has attracted interest as a potential ADC target.

Carcinoembryonic antigen-related cell adhesion molecule-5 (CEACAM5), also known as CEA or CD66e, is a well-established diagnostic marker in CRC (MacSween et al., 1972; Beauchemin and Arabzadeh, 2013). CEACAM5 is also highly expressed in multiple malignancies, such as pancreatic cancer, gallbladder cancer, non-small cell lung cancer, and bladder cancer, but its expression in normal tissues is limited (Eades-Perner et al., 1994; Beauchemin and Arabzadeh, 2013). As a serum biomarker, CEACAM5 demonstrated the highest sensitivity (93.7%) and specificity (96.1%) in a CRC cohort (Tiernan et al., 2013). Its tumour-specific overexpression makes CEACAM5 an ideal target for ADC-based therapies.

In addition to HER2, HER3, EGFR, GCC, and CEACAM5, several emerging targets, including CDH17, PD-L1, and LGR5, have recently been investigated in CRC. These targets are characterized by relatively specific expression profiles and are often closely associated with tumour initiation and progression, immune evasion, stem cell maintenance, and drug resistance. Although the number of ADCs developed against these targets remains limited, most agents still at the preclinical or early clinical stage have already demonstrated certain antitumour potential. To provide a more straightforward overview of these advances, representative information on the corresponding ADCs is summarized in Table 4.

Rational combinations of antibody–drug conjugates (ADCs) with conventional chemotherapy are increasingly explored to deepen responses and delay or overcome resistance, provided that the partner drug adds complementary antitumor pressure without duplicating the ADC’s dose-limiting toxicities (Pretelli et al., 2024; Wei et al., 2024). Because most ADCs require binding, internalization, and intracellular processing to release active payload, the sequence of administration can meaningfully affect efficacy: preclinical data suggest that sequential dosing can produce greater tumor cell damage than concurrent administration (Wei et al., 2024). Despite these potential benefits, the major barrier to clinical development remains tolerability, as ADC–chemotherapy combinations frequently exacerbate grade ≥3 adverse events, consistent with overlapping systemic toxicities driven by off-target/on-target off-tumor exposure to payload (Tan et al., 2025). This risk is further shaped by ADC design features such as linker stability and DAR, which influence systemic free payload exposure and correlate with severe toxicities, underscoring the need for careful partner selection and schedule/dose optimization in combination regimens (Markides et al., 2024; Tang S. C. et al., 2024; Tan et al., 2025).

In both preclinical and clinical studies, combining different forms of chemotherapy with ADCs has been widely recognized as an effective strategy to overcome resistance and achieve favourable therapeutic outcomes (Yardley, 2013). The synergistic interactions between chemotherapy and ADCs include targeting different stages of the cell cycle or modulating the expression of tumour cell surface antigens. The sequence of administration plays a critical role in treatment efficacy, which may be related to the efficiency of ADC internalization (Wahl et al., 2001). However, the major challenge limiting the clinical development of these combinations is the significant increase in toxicity, which is likely caused by overlapping toxicity from the off-target and off-tumour effects of ADC payloads (Martin et al., 2016; Cortés et al., 2020).

In preclinical models, RC48 combined with gemcitabine demonstrated synergistic antitumour activity both in vitro and in vivo. Further RNA-seq analyses indicated that the combination may have synergistic effects on CRC cells through the regulation of multiple signalling pathways, such as the p53, PI3K-AKT, and MAPK pathways and pathways involved in the cell cycle(Liu et al., 2024). Another study revealed that Oba01 (an ADC targeting DR5) combined with the CDK inhibitor abemaciclib significantly enhanced the in vivo inhibition of microsatellite stability (MSS) and microsatellite instability-high (MSI-H) CRC PDX growth without inducing toxicity or resistance (Zhou D. et al., 2025). In rectal cancer PDX models, AMT-562 initially showed promising efficacy, but the tumours quickly relapsed; the combination of AMT-562 with rabusertib resulted in more durable antitumour effects (Weng et al., 2023a). Collectively, these findings highlight the potential of combining ADCs with conventional chemotherapeutics as novel treatment strategies for advanced CRC, particularly the refractory MSS subtype. For patients with a heavy tumour burden and rapid progression, ADC–chemotherapy combinations can achieve rapid tumour control.

Targeted therapies—including monoclonal antibodies, tyrosine kinase inhibitors, and antiangiogenic agents—have proven clinical safety and efficacy in tumours with specific mutations, overexpression, or amplification. However, resistance and clonal heterogeneity narrow the therapeutic window of targeted monotherapies, leading to the emergence of ADC–targeted therapy combinations. The goal is to achieve more potent inhibition of oncogene-dependent signalling pathways, increase the availability of surface antigens, and sensitize tumours with low antigen expression. Synergistic mechanisms include improving intratumoural drug delivery via antiangiogenic effects (Ponte et al., 2016; Bordeau et al., 2021), modulating tumour antigen expression (La Monica et al., 2017; Haikala et al., 2022), overcoming intratumoural heterogeneity and resistance (Saatci et al., 2018), and inducing synthetic lethality.

Combining ADCs with targeted agents—such as anti-angiogenic therapies—may enhance antitumour activity in CRC and warrants clinical evaluation. As a monotherapy, AMT-562 showed limited efficacy in CRC, but when it was combined with bevacizumab or cetuximab, it demonstrated significantly increased antitumour activity (Weng et al., 2023a). Currently, clinical trials of RC48 in combination with bevacizumab or tislelizumab are underway in patients with HER2-positive advanced CRC, aiming to further evaluate the efficacy and safety of the combination treatment in a second-line setting (NCT05785325, NCT05493683). The PROCEADE-CRC-01 trial is currently being conducted to evaluate efficacy of Precemtabart tocentecan combined with bevacizumab ± capecitabine or bevacizumab + 5-fluorouracil(Kopetz et al., 2025a).

The combination of ADCs with immune checkpoint inhibitors (ICIs) is based on a dual mechanism of “localized high-potency cytotoxicity + immune activation.” This combination treatment may potentially overcome the limitations of previous immunotherapies in CRC while enabling more refined individualized treatment. In tumours, ADCs deliver highly potent cytotoxins in a targeted manner, leading to tumour cell death and the release of tumour antigens and damage-associated molecular patterns. This promotes dendritic cell–mediated antigen presentation and T-cell activation. Moreover, ICIs (e.g., anti-PD-1/PD-L1) promote adaptive immune suppression, amplifying the ADC-induced primary immune response (Villacampa et al., 2024).

ADC + ICI strategies should be guided by multidimensional biomarkers—including molecular target expression, MSI status, and tumour-infiltrating immune phenotypes—to achieve true personalized treatment. CRC is highly heterogeneous both molecularly and immunologically, comprising MSI-H (“immune-hot”) and MSS (“immune-cold”) subtypes, different oncogenic drivers (HER2, c-MET, KRAS/BRAF, etc.), and variable antigen expression. Since the vast majority of metastatic CRCs are MSS and unresponsive to ICIs alone, ADCs may act as “tumour heaters” by inducing antigen release and recruiting effector immune cells, thereby substantially expanding the proportion of MSS patients who could benefit from ICIs (Yu et al., 2024). In syngeneic mouse tumour models, RC48 and anti–PD-1 monotherapy inhibited tumour growth. Moreover, their combination further enhanced efficacy without significant body weight loss, indicating that RC48 can sensitize tumours to immunotherapy independent of microsatellite status (Wu X. et al., 2023). Similarly, 84-EBET (a CEACAM6-targeting ADC) combined with an anti–PD-1 antibody showed marked synergy (Kogai et al., 2025).

To enhance clinical relevance, toxicity evaluation for ADCs in colorectal cancer should be reported in a standardized manner (CTCAE grading), with explicit documentation of grade ≥3 events, serious adverse events, and the frequencies of dose interruption, reduction, and discontinuation. Cross-trial comparisons should be interpreted cautiously because eligibility criteria, prior lines of therapy, baseline organ function, and supportive-care practices vary substantially across studies; these factors can inflate or attenuate apparent toxicity rates in heavily pretreated mCRC populations.

Clinically actionable toxicities can be broadly grouped into hematologic suppression, gastrointestinal toxicity, peripheral neuropathy, ocular events, hepatotoxicity, and pneumonitis/interstitial lung disease (ILD) (Mahalingaiah et al., 2019; Hackshaw et al., 2020; Powell et al., 2020). These patterns are influenced by payload class (Colombo and Rich, 2022; Zhu et al., 2023), linker stability (Donaghy, 2016), and, in some cases, target expression in normal tissues (Hu et al., 2023). A pragmatic summary of common adverse events, plausible drivers, recommended monitoring, and risk-mitigation strategies for CRC practice is provided in Table 5.

Across CRC-directed ADC programs, microtubule-disrupting payloads are most commonly associated with peripheral neuropathy and ocular surface events, often in a cumulative-exposure pattern (Powles et al., 2021). In contrast, Topoisomerase I inhibitor–based payloads more frequently produce myelosuppression and gastrointestinal toxicity and have also been linked to pneumonitis/ILD in some settings, underscoring the need for early recognition and protocolized intervention (Bardia et al., 2021). DNA-damaging payloads may carry broader systemic toxicity and a narrower therapeutic window, making dose optimization and patient selection particularly critical. Because mCRC patients frequently have baseline gastrointestinal symptoms, prior neurotoxic chemotherapy exposure, and substantial hepatic tumour burden, risk stratification and proactive supportive care should be integrated into trial design and routine practice to preserve treatment continuity and maximize benefit.

Beyond the payload-class patterns and the clinically oriented risk-mitigation framework summarized above (Table 5), ADCs can also induce unexpected or disabling toxicities that are not readily predicted by antigen specificity alone. The mechanisms remain incompletely understood and may involve extra-tumoural antigen expression, nonspecific uptake by normal tissues, and/or systemic enzymatic cleavage resulting in exposure to deconjugated species and free payload. Historically, this phenomenon was highlighted by the LeY-targeting conjugate BR96–doxorubicin, which produced prominent gastrointestinal toxicity rather than the “expected” hematologic or cardiac toxicities typically associated with doxorubicin. (Saleh et al., 2000). Cardiotoxicity has also been observed with trastuzumab-based platforms, including T-DM1 and T-DXd, consistent with potential on-target/off-tumour liability of the antibody component, whereas such toxicity appears less common with other DXd-based programs (Hu et al., 2023).

Importantly, interstitial lung disease (ILD)/pneumonitis has been reported across multiple ADCs with variable incidence and severity and may occur with patterns that are not strictly dependent on antigen specificity or payload type, underscoring the need for early recognition and protocolized intervention (Hackshaw et al., 2020; Kumagai et al., 2020; Tarantino et al., 2021). In addition to between-drug variability, clinically meaningful heterogeneity exists among patients receiving the same ADC, as baseline organ function, comorbidities, and inter-individual PK/PD differences can modify both toxicity phenotype and severity(Powell et al., 2022; Tang M. et al., 2024); therefore, eligibility criteria, monitoring intensity, and dose-modification algorithms should be tailored accordingly. When ADCs are combined with chemotherapy or other systemic agents, these risks may be amplified through overlap or unexpected synergy, further reinforcing the value of predefined stopping rules and organ-specific management pathways.

The most common mechanism of ADC resistance is the downregulation of target antigen expression, which parallels the principles of resistance observed with monoclonal antibodies. With increasing drug exposure, target antigens may undergo mutations, copy number alterations, or structural modifications, thereby reducing the likelihood of antibody–antigen binding. Preclinical studies have demonstrated that when used to generate xenograft tumours responsive to high-dose T-DM1, JIMT1 cells resistant to trastuzumab acquire secondary resistance following cyclic T-DM1 treatment, accompanied by reduced HER2 expression (Loganzo et al., 2015; Loganzo et al., 2016). Interestingly, the bystander effect mediated by cleavable linkers can partially overcome the reduced response associated with lower antigen expression (Sung et al., 2016). In addition to decreased antigen levels, dimerization of the target antigen with another cell surface receptor can also contribute to ADC resistance. For example, NRG-1β, a ligand that induces HER2/HER3 heterodimerization, inhibits T-DM1-mediated cytotoxicity in HER2-amplified breast cancer cells. This resistance can be overcome by combination with pertuzumab, which blocks HER2/HER3 dimerization and downstream signalling (Phillips et al., 2014).

Beyond aberrant antigen expression, tumour cells may also develop resistance directly to the cytotoxic payload. In the treatment of non-Hodgkin lymphoma, clinical efficacy was increased by replacing auristatin-based payloads with anthracycline-based payloads (Yu et al., 2015). In addition to payload type, the conjugation site and the average DAR are critical determinants of ADC performance (Yoder et al., 2019; Bai et al., 2020). Preclinical data indicate that while increasing the DAR enhances drug load, it also accelerates clearance, potentially reducing ADC efficacy (Sun et al., 2017).

ADCs are internalized into cancer cells primarily via endocytosis, which can occur through multiple pathways, including clathrin-mediated endocytosis, caveolae-mediated endocytosis, and clathrin–caveolin-independent endocytosis (Kalim et al., 2017). In a developed T-DM1-resistant model (N87-TM), upregulation of caveolin-1 (Cav-1), polymerase I, and transcriptional regulators involved in vesicle formation and endocytosis combined with reduced lysosomal colocalization impaired drug internalization and trafficking, resulting in T-DM1 insensitivity (Sung et al., 2018). In HER2-positive breast cancer, endophilin A2 promotes HER2 endocytosis and increases tumour cell sensitivity to trastuzumab. SH3GL1 knockdown in tumour cells decreases endocytosis and significantly reduces T-DM1-mediated cytotoxicity (Baldassarre et al., 2017).

Following receptor-mediated endocytosis, ADCs are trafficked to lysosomes, where their cytotoxic payloads are released through chemical or enzymatic cleavage. Most lysosomal hydrolases require an acidic environment for optimal activity, a condition maintained by vacuolar ATPase (Mellman, 1989). This mechanism has been demonstrated in the T-DM1-resistant cell line N87-KR, which exhibited binding and internalization kinetics similar to those of parental N87 cells but showed markedly reduced lysosomal acidification due to impaired vacuolar ATPase activity. This defect diminished T-DM1 catabolism and conferred resistance (Wang et al., 2017). Beyond pH regulation, specific transporters also play critical roles, particularly for ADCs with noncleavable linkers, as their metabolites require dedicated transport proteins to reach the cytoplasm. RNA sequencing studies have identified SLC46A3 as a key transporter that mediates the cytoplasmic translocation of maytansine-based catabolites. The silencing of SLC46A3 led to the lysosomal accumulation of metabolites and ultimately drug resistance (Hamblett et al., 2015).

The overexpression of ATP-binding cassette transporters plays a pivotal role in chemotherapy resistance and has been well documented across multiple cancer types. These efflux pumps facilitate the active extrusion of drugs from tumour cells, thereby conferring an multidrug resistance phenotype (Leonard et al., 2003; Yu et al., 2013; Dong et al., 2024). Since many ADC payloads, such as maytansinoids, are substrates of ATP-binding cassette transporters, their overexpression can directly impair ADC efficacy (Kovtun et al., 2010; Cianfriglia, 2013). For instance, studies in T-DM1-resistant cells demonstrated that, despite preserving HER2 overexpression, upregulation of ABCC2 and ABCG2 promoted DM1 efflux, leading to acquired resistance. Importantly, inhibition of these transporters was shown to restore T-DM1 sensitivity (Takegawa et al., 2017).

The activation of alternative signalling pathways is another key mechanism through which tumours acquire resistance to ADCs. A prototypical example is the PI3K/AKT/mTOR pathway, which regulates cell survival, growth, and metabolism. Aberrant activation of this pathway can reduce ADC sensitivity and attenuate payload efficacy. Clinically, a diminished therapeutic response to trastuzumab has been observed in patients harbouring PIK3CA mutations or with PTEN loss (Baselga et al., 2016). Mechanistic studies have indicated that PTEN deficiency or PIK3CA hyperactivation reduces trastuzumab sensitivity through constitutive activation of PI3K/AKT signalling (Berns et al., 2007). In addition, activation of the Wnt/β-catenin pathway has been implicated in ADC resistance. Overexpression of Wnt3 promotes β-catenin accumulation, thereby promoting tumour cell proliferation and invasion while simultaneously conferring resistance to trastuzumab (Wu et al., 2012).