Focusing on toxicity management: Challenges and strategies for HER2-targeted antibody–drug conjugates in breast cancer

Abstract

Antibody–drug conjugates (ADCs) targeting human epidermal growth factor receptor 2 (HER2) have revolutionized the treatment landscape of HER2-positive breast cancer, significantly improving patient survival. However, their growing clinical application has revealed a spectrum of serious adverse events (AEs) that can compromise quality of life, reduce treatment compliance, and, in some cases, lead to life-threatening outcomes or premature therapy discontinuation. This review provides a comprehensive overview of the toxicity profiles and underlying mechanisms of HER2-targeted ADCs, with a focus on representative agents such as trastuzumab emtansine (T-DM1) and trastuzumab deruxtecan (T-DXd). We examine the pathogenesis of common toxicities, including thrombocytopenia, interstitial lung disease, cardiotoxicity, and hepatotoxicity, and summarize clinical evidence for monitoring and intervention strategies. Emphasis is placed on the importance of early identification and standardized management to mitigate risk. Furthermore, we discuss emerging approaches to improve ADC safety through structural optimization, including advances in antibody engineering, linker design, and payload selection. This review aims to guide clinicians and researchers in improving the safety and clinical utility of HER2-targeted ADCs in breast cancer treatment.

Graphical abstract

Highlights

• •HER2-targeted ADCs improve survival in breast cancer but cause serious toxicities, requiring systematic management.
• •ADC toxicities arise from both on- and off-target mechanisms, offering insights for intervention strategies.
• •Next-generation ADCs use bispecific design and optimized linkers to enhance safety and therapeutic index.

1. Introduction

In recent years, antibody-drug conjugates (ADCs), as a next-generation targeted therapeutic strategy, have demonstrated significant breakthroughs in oncology. Human epidermal growth factor receptor 2 (HER2)-targeted ADCs such as trastuzumab emtansine (T-DM1) and trastuzumab deruxtecan (T-DXd) have demonstrated remarkable therapeutic potential in breast cancer, benefiting not only patients with HER2 overexpression but also those with HER2-low tumors [1,2]. However, the complex molecular structure of ADCs, while enabling their targeted efficacy, is also associated with a range of potentially severe adverse events, including thrombocytopenia, interstitial lung disease (ILD), gastrointestinal toxicity, cardiotoxicity, and ocular injury—some of which may reach grade ≥3 severity and threaten treatment continuity or patient safety [3,4]. These toxicities not only compromise treatment adherence and patient quality of life but may also limit the broader clinical application of ADCs. Thus, elucidating the molecular basis of these toxicities and developing early warning systems, intervention strategies, and optimized dosing regimens has become a critical research priority. This review focuses on the toxicity challenges of HER2-targeted ADCs in breast cancer therapy, providing a systematic overview of their mechanistic underpinnings, representative adverse events, and mitigation strategies. The goal is to inform future ADC development toward safer and more effective therapeutic applications.

2. Advances in ADC for HER2-positive breast cancer

ADCs are a novel class of targeted therapeutics composed of a monoclonal antibody, a cytotoxic payload, and a chemical linker. Their design is inspired by Paul Ehrlich's “magic bullet” concept [[5], [6], [7]]. ADCs function by binding the antibody component to tumor-specific surface antigens, followed by receptor-mediated endocytosis and intracellular trafficking through the endosome–lysosome pathway [8]. In the lysosomal environment, cleavable linkers are degraded by acidic pH or proteases, releasing the cytotoxic payload. In contrast, non-cleavable linkers require complete degradation of the antibody backbone to release the active drug [9].The released payload induces cell death—primarily through DNA binding or inhibition of microtubule polymerization, ultimately triggering apoptosis or other forms of cytotoxicity (Fig. 1) [10]. In the context of breast cancer, two HER2-targeted ADCs have been approved for clinical use: trastuzumab emtansine (T-DM1) and trastuzumab deruxtecan (T-DXd) [11]. Several other anti-HER2 ADCs are currently under clinical investigation. A summary of ongoing trials and agents is presented in Table 1.

Fig. 1.

Mechanism of action of antibody-drug conjugates (ADCs). An ADC consists of three key components: a monoclonal antibody, a cytotoxic payload, and a chemical linker. The antibody moiety binds to tumor-specific surface antigens, triggering receptor-mediated endocytosis and subsequent intracellular trafficking through the endosome-lysosome pathway. Within the lysosomal compartment, cleavable linkers undergo degradation via acidic pH or proteolytic enzymes, releasing the cytotoxic payload. The liberated payload induces cell death through DNA intercalation or microtubule disruption, ultimately triggering apoptosis or other programmed cell death mechanisms.

Table 1.

Characteristics of anti-HER2 ADCs.

ADCs	Antibody	Linker	Cytotoxic Payload	DAR	Adverse Events
T-DM1	Trastuzumab	SMCC	DM1	3.5	Hepatotoxicity, thrombocytopenia, fatigue [3]
T-DXd	Trastuzumab	Enzyme-cleavable tetrapeptide	DXd	8	Neutropenia, nausea, anemia [12]
ARX788	Humanized HER2 mAb	Non-cleavable	AS269	1.9-2.0	Ocular toxicity, myelosuppression, dermatologic effects [13]
RC48	Anti-HER2 mAb	Enzyme-cleavable	MMAE	3-4	Peripheral neuropathy, diarrhea, myelosuppression [14]
A166	Anti-HER2 mAb	Enzyme-cleavable	Duo-5	3-4	Keratopathy, hypophosphatemia, dry eye [15]
PF-06804103	Anti-HER2 mAb	Valine-citrulline (vc0101)	Aur0101	4	Alopecia, neuropathy, keratitis, arthralgia [16]
SYD985	Trastuzumab	Valine-citrulline cleavable linker	seco-DUBA	2.7	Conjunctivitis, keratitis, fatigue, respiratory disorders [17]
ZW49	Zanidatamab	Valine-citrulline (vc) cleavable linker	MMAE	2–4	Keratitis, alopecia, diarrhea [18]
ALT-P7	Trastuzumab biobetter (HM2)	Cleavable cysteine-containing peptide linker	DM4	2	Neurotoxicity, fatigue, nausea [19]
BL-M07D1	Trastuzumab	Cathepsin B-cleavable linker	Ed-04	8	Anemia, neutropenia, keratitis [20]
SHR-A1811	anti-HER2 mAb (Engineered)	Cleavable peptide linker	SHR9265 (DXd-like topo I inhibitor)		Neutropenia, nausea, anemia [21]

Note: mAb = monoclonal antibody; FDA = U.S. Food and Drug Administration; SMCC = succinimidyl 4-(N-maleimidomethyl)cyclohexane-1-carboxylate.

2.1. Clinical advances of T-DM1 in HER2-Positive breast cancer

T-DM1 is composed of trastuzumab, the microtubule inhibitor DM1, and a stable thioether linker (SMCC). It delivers DM1 specifically to HER2-positive tumor cells, where it disrupts microtubule assembly, leading to cell cycle arrest and apoptosis [22,23]. In addition, the trastuzumab component retains its ability to mediate antibody-dependent cellular cytotoxicity (ADCC), thereby enhancing the antitumor immune response [24]. T-DM1 has demonstrated significant clinical efficacy in several pivotal trials. In the EMILIA trial, T-DM1 significantly outperformed lapatinib plus capecitabine as second-line therapy for advanced breast cancer, extending median progression-free survival (PFS) by 3.2 months (9.6 vs 6.4 months) and overall survival (OS) by 5.8 months (30.9 vs 25.1 months) [3]. For heavily pretreated and refractory patients, the TH3RESA trial showed that T-DM1 significantly improved both PFS (6.2 vs 3.3 months) and OS (22.7 vs 15.8 months) compared with physician's choice of therapy [25,26]. In early-stage breast cancer, the KATHERINE trial demonstrated that switching to T-DM1 in patients with residual invasive disease after neoadjuvant therapy significantly improved 3-year invasive disease-free survival (88.3% vs 77.0%) and reduced the risk of distant recurrence (10.5% vs 15.9%) [27]. Nonetheless, the clinical utility of T-DM1 is constrained by dose-limiting toxicities and intrinsic resistance in some patients, highlighting the need for further optimization strategies [26].

2.2. Clinical advances of T-DXd in HER2-Positive breast cancer

T-DXd consists of an anti-HER2 antibody, a cleavable tetrapeptide linker, and the topoisomerase I inhibitor DXd. Upon HER2-targeted internalization, DXd stabilizes the Top1–DNA cleavage complex, inducing DNA damage during replication and promoting tumor cell apoptosis [28]. A uniform drug-to-antibody ratio (DAR) and stable linker design reduce off-target toxicity. The membrane-permeable DXd enables a bystander effect, allowing diffusion into adjacent antigen-negative tumor cells and expanding therapeutic reach [[28], [29], [30]]. In the DESTINY-Breast01 trial in patients with advanced breast cancer, T-DXd achieved an objective response rate (ORR) of 60.9%, a disease control rate (DCR) of 97.3%, a median progression-free survival (PFS) of 16.4 months, and a median duration of response (DoR) of 14.8 months [4]. The DESTINY-Breast03 trial demonstrated that T-DXd significantly outperformed T-DM1, with a median PFS of 29.0 vs 7.2 months and OS of 52.6 vs 42.7 months. The 36-month PFS rate was also markedly improved (45.7% vs 12.4%) [12]. The TUXEDO-1 study reported an intracranial response rate of 78.6% and a median PFS of 21 months in the per-protocol population (PPP) of patients with brain metastases refractory to systemic therapy, including both newly diagnosed and previously treated lesions. [31]. The DESTINY-Breast04 trial evaluated T-DXd in HER2-low metastatic breast cancer, showing significant improvement in PFS (9.9 vs 5.1 months) and OS (23.4 vs 16.8 months) compared to treatment of physician's choice (TPC) [32]. Despite its clinical efficacy, interstitial lung disease (ILD), observed in approximately 10% to 14% of patients treated with T-DXd, remains a key safety concern. Therefore, baseline pulmonary evaluation and routine high-resolution CT (HRCT) monitoring are crucial for timely detection and intervention [32,33].

2.3. Advances in novel HER2-Targeted antibody-drug conjugates for breast cancer

Several next-generation HER2-directed ADCs have emerged with encouraging antitumor activity in breast cancer. Among them, ARX788 leverages non-natural amino acid incorporation for site-specific conjugation of an anti-HER2 monoclonal antibody with a tubulin inhibitor, AS269 [34]. In a phase I trial, ARX788 achieved an ORR of 65.5% and a DCR of 100% in patients with HER2-positive advanced breast cancer, with a median DoR of 14.4 months [35]. Combination regimens are currently under active investigation. In phase I/II trials, RC48 achieved ORRs of 42.9% in HER2-overexpressing and 33.3% in HER2-low breast cancer, with a median PFS of 5.7 and 5.1 months, respectively, and a manageable safety profile [14]. A166, an anti-HER2 ADC conjugated with the novel microtubule inhibitor Duo-5, demonstrated an objective response rate of 73.9% and a median progression-free survival of 12.3 months at the 4.8 mg/kg dose level in a phase I study, with efficacy appearing dose-dependent and toxicity remaining manageable [15]. PF-06804103 is an anti-HER2 ADC composed of a monoclonal antibody conjugated via a cleavable linker to the microtubule inhibitor Aur0101. In a phase I trial, it demonstrated an objective response rate of 71% in HER2-positive breast cancer and 52.4% in patients with HER2-low disease [16]. In addition, trastuzumab duocarmazine (SYD985) is an ADC constructed from trastuzumab linked via a cleavable linker to seco-DUBA. Its phase III TULIP trial demonstrated a significantly superior median PFS compared with physician's choice of treatment (7.0 vs. 4.9 months) [17]. Zanidatamab zovodotin (ZW49) is a bispecific ADC targeting HER2/HER3. Its phase I study showed a manageable safety profile in HER2-positive solid tumors, with an ORR of 13% in breast cancer patients [36]. In a phase I study (NCT03281824), ALT-P7 has shown preliminary signals of objective response in heavily pretreated HER2-positive metastatic breast cancer patients; however, detailed efficacy outcomes have not yet been fully reported in a peer-reviewed publication [19]. BL-M07D1 is an ADC based on a humanized trastuzumab conjugated to the topoisomerase I inhibitor Ed-04. Preclinical studies indicated its antitumor activity exceeds that of T-DM1 and T-DXd, and a phase I trial is currently underway in patients with metastatic breast cancer [37]. Recently, SHR-A1811 demonstrated significantly prolonged median PFS in the phase III HORIZON-Breast01 trial in previously treated HER2-positive advanced breast cancer (30.6 vs. 8.3 months; HR 0.22), with a manageable safety profile [21]. In addition to the agents discussed above, several other HER2-targeted ADCs are currently under clinical or preclinical development.

While structural optimizations of antibody, linker, and payload have expanded the clinical utility of HER2-targeted ADCs in both HER2-positive and HER2-low breast cancer, these advances have also introduced a distinct spectrum of toxicities that demand mechanistic insight and effective management strategies.

3. Mechanisms of ADC toxicity

ADCs exhibit unique toxicity profiles distinct from traditional chemotherapy or monoclonal antibodies, primarily due to variations in target antigen specificity, linker stability, and cytotoxic payload. These toxicities are generally classified as on-target or off-target (Fig. 2).

Fig. 2.

Toxicity mechanisms of antibody-drug conjugates (ADCs). ADC -related toxicities are categorized as on-target or off-target effects: (1) On-target toxicity: The antibody binds to target antigens expressed on normal cells, leading to payload release in healthy tissues. (2) Premature payload release: Systemic instability of linkers results in extracellular payload dissociation, causing diffuse tissue damage. (3) FcγR-mediated uptake: Nonspecific internalization by immune cells or FcγR-expressing tissues via antibody Fc domain interactions. (4) Non-specific endocytosis: Pinocytosis/macropinocytosis of intact ADCs or free payload by normal cells. (5) Bystander effect: Payload released from antigen-positive tumor cells diffuses into adjacent normal cells via passive transport or solute carriers.

3.1. On-target toxicity

Although ADCs are designed to selectively target tumor cells, their antibody component may also bind the same antigen expressed in normal tissues, leading to off-tumor payload release and associated toxicities. The incidence and severity of such events depend on the antigen's expression pattern in healthy tissues and the pharmacologic properties of the cytotoxic payload [38].

3.1.1. Hepatotoxicity

Drug-induced liver injury (DILI) refers to hepatic dysfunction or tissue damage caused by regulated pharmaceuticals, including chemically synthesized drugs, biologics, and traditional medicines [39]. As a HER2-targeted ADC, T-DM1 carries a boxed warning for hepatotoxicity from the U.S. FDA, necessitating regular monitoring of liver enzymes during therapy. T-DM1–associated hepatotoxicity typically manifests as elevated serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST). In the phase III EMILIA trial, AST and ALT elevations were reported in 22.4% and 16.9% of patients, respectively, with grade 3–4 events occurring in 4.3% and 2.9% of cases [3].

To elucidate the molecular basis of T-DM1–induced hepatotoxicity, Endo et al. established integrated in vitro and in vivo models, which demonstrated that hepatocytes internalize T-DM1 via low-level HER2 expression on their surface [[40], [41], [42]]. This uptake leads to intracellular release of the DM1 payload, disruption of the mitochondrial outer membrane, and activation of intrinsic apoptosis pathways [42]. Additionally, T-DM1 upregulates pro-inflammatory cytokines such as TNF-α in hepatic tissue, further exacerbating liver injury [42]. Similar target-dependent hepatotoxicity has also been observed with other ADCs, notably gemtuzumab ozogamicin (GO), which targets CD33—a protein expressed in hepatic sinusoidal cells. Due to severe liver toxicity, including hepatic veno-occlusive disease, GO was voluntarily withdrawn from the U.S. market in 2010 [43]. Beyond HER2-mediated uptake, T-DM1 also engages a HER2-independent mechanism involving direct binding of DM1 to cytoskeleton-associated protein 5 (CKAP5) on hepatocyte membrane. This interaction induces membrane destabilization, calcium influx, microtubule disassembly, and subsequent activation of calcium-dependent cell death pathways [41]. Of note, this CKAP5-mediated hepatotoxicity appears specific to maytansinoid-based payloads such as DM1 and DM4, as ADCs containing MMAE have not shown comparable hepatotoxic profiles in preclinical models (Fig. 3) [41].

Fig. 3.

Mechanisms of T-DM1-induced hepatotoxicity via HER2-and CKAP5-dependent pathways. (1) HER2-dependent pathway: Following binding to HER2 on tumor cell surfaces, the T-DM1/HER2 complex undergoes internalization and lysosomal degradation, releasing DM1 to induce apoptosis. This represents the dominant hepatotoxicity mechanism in HER2-positive tumors. (2) CKAP5-dependent pathway: DM1 specifically binds to CKAP5 on hepatocyte membranes, causing membrane damage and Ca²⁺ influx. Elevated cytosolic Ca²⁺ disrupts microtubule networks, ultimately triggering apoptotic cell death.

In summary, T-DM1 induces hepatotoxicity via both HER2-mediated uptake and HER2-independent interaction with CKAP5, constituting a multifactorial model of hepatocellular injury. These findings deepen our understanding of ADC-associated toxicity and lay the groundwork for mechanism-guided, personalized hepatotoxicity management strategies.

3.1.2. Cardiotoxicity

Cardiotoxicity is a recognized safety concern in HER2-targeted therapies and has drawn increasing clinical attention with the advent of HER2-targeted ADCs. In the EMILIA study, 1.7% of patients treated with T-DM1 experienced a ≥10% reduction in LVEF, with no cardiac-related deaths reported [3]. In the DESTINY-Breast03 trial, LVEF decline was observed in 4.3% of patients receiving T-DXd and 1.5% of those receiving T-DM1, with no grade ≥3 heart failure events reported in either group [12].

The HER2 receptor plays a pivotal role in cardiomyocyte development, homeostasis, and repair, primarily through activation of downstream signaling pathways such as PI3K/Akt and Ras/MAPK, which regulate cellular proliferation, survival, and energy metabolism. HER2-targeted ADCs may disrupt receptor dimerization and downstream signaling, leading to metabolic dysregulation, sarcomeric disarray, and impaired regenerative capacity in cardiomyocytes [44]. At the subcellular level, these effects are associated with mitochondrial dysfunction, including reduced activity of respiratory chain complexes, decreased ATP production, and accumulation of reactive oxygen species (ROS). Collectively, these changes result in reversible myocardial contractile dysfunction and ventricular remodeling [45]. The risk of cardiotoxicity is significantly increased when HER2-ADCs are administered in combination with anthracyclines due to a synergistic “dual-hit” effect. While anthracyclines induce primary cardiomyocyte injury via oxidative stress, HER2-targeted therapies impair cardiac repair by inhibiting the HER2/neuregulin-1 (NRG1) signaling axis. Together, these mechanisms markedly elevate the incidence of adverse cardiac events [46].

3.1.3. Ocular toxicity

Ocular toxicity is a notable adverse effect associated with several HER2-targeted ADCs. A166 has been linked to a high incidence of eye-related events, with corneal disorders reported in 84% and blurred vision in 74% of patients. Grade ≥3 ocular toxicities occurred in 16%, primarily due to severe keratopathy (31%) [15]. Mild-to-moderate ocular events have also been reported in clinical trials of T-DM1 and T-DXd, including dry eye in 4% and 11% of patients, respectively, and conjunctivitis in 4% of those treated with T-DM1 [47]. In the ACE-Breast-02 study, ARX-788 demonstrated a considerable rate of ocular toxicity as well, with dry eye (55.0%) and blurred vision (35.0%) being the most common events. Notably, grade ≥3 dry eye and blurred vision were reported in 9.1% and 12.3% of patients, respectively [13].

HER2 is expressed in corneal epithelial cells and contributes to limbal stem cell proliferation and epithelial wound healing. HER2-targeted ADCs may interfere with these physiological processes, compromising epithelial homeostasis and regeneration, and thereby contributing to ocular surface toxicities such as keratopathy [48]. In addition, ADCs may enter ocular cells through non-specific uptake mechanisms such as macropinocytosis, leading to intracellular accumulation of cytotoxic payloads and exacerbation of local toxicity. A study by Zhao et al. demonstrated that inhibiting macropinocytosis-mediated uptake in ocular tissues reduced ADC-induced toxicity, suggesting a potential approach for the development of safer ADC formulations [49].

3.2. Off-target toxicity

Off-target toxicity is a major dose-limiting concern in ADC therapy and primarily arises from multiple mechanisms, including: (1) premature cleavage of unstable linkers in circulation, leading to systemic release of cytotoxic payloads; (2) Fc gamma receptor (FcγR)-mediated nonspecific uptake by non-target immune or stromal cells; and (3) the bystander effect, wherein membrane-permeable payloads diffuse into adjacent antigen-negative cells, causing collateral damage [50].

Linker stability is a key determinant of ADC pharmacology, directly influencing payload specificity and off-target toxicity. First, Cleavable linkers (e.g., hydrazone or peptide-based) may be prematurely degraded by plasma hydrolases or nonspecific enzymes during systemic circulation, resulting in early payload release before reaching tumor sites [51]. These free payloads—especially lipophilic small molecules—can passively diffuse into normal cells, disrupting vital processes such as DNA replication or microtubule assembly, and inducing systemic toxicities like myelosuppression and hepatotoxicity [52]. Premature release also compromises tumor targeting, increasing payload exposure in healthy tissues and aggravating toxicity [50]. Second, Fcγ receptor (FcγR)-mediated nonspecific uptake. The Fc region of ADCs can bind to FcγRs on immune cells, mediating ADCC and phagocytosis [53]. In ADC therapy, this interaction may also result in unintended internalization by non-target immune or stromal cells, leading to cytotoxin release and off-target toxicity [54]. To reduce this risk, Fc domains may be engineered to diminish FcγR affinity [55], and Fc-free formats such as Fab-based ADCs are being explored for enhanced specificity [56]. Third, the bystander effect. Membrane-permeable payloads released in antigen-positive cells may diffuse into neighboring antigen-negative cells [57]. This expands therapeutic reach in tumors with heterogeneous expression, but also increases the risk of collateral toxicity in healthy tissues. ADCs such as T-DM1 and ARX-788, which release maytansinoid payloads (e.g., DM1 or MMAF), can cause unintended injury to bone marrow, liver, and lung through passive diffusion and microtubule or DNA damage [50].

3.2.1. Thrombocytopenia

T-DM1–induced thrombocytopenia represents a major dose-limiting toxicity observed across multiple clinical trials. It is characterized by a decrease in peripheral platelet counts due to impaired megakaryocyte differentiation and maturation in the bone marrow [58]. In the EMILIA trial, grade 3–4 thrombocytopenia occurred in 12.9% of patients receiving T-DM1, compared with 0.2% in the control group. The overall incidence was 28.0% versus 2.5%, respectively [3]. Similarly, in the TH3RESA study, grade 3–4 thrombocytopenia was observed in 6.0% of patients treated with T-DM1 and 3.0% of controls, with total incidence rates of 21.0% and 4.0%, respectively [26].

Uppal et al. employed a human hematopoietic stem/progenitor cell model and demonstrated that T-DM1–induced thrombocytopenia is primarily mediated by FcγRIIa-dependent internalization [59]. Through its Fc domain, T-DM1 binds FcγRIIa on megakaryocyte surfaces, facilitating receptor-mediated uptake. The released DM1 payload disrupts microtubule assembly and cytoskeletal dynamics, impairing megakaryocyte maturation and reducing platelet production [59]. However, blocking FcγRIIa did not fully prevent T-DM1 internalization, suggesting the presence of alternative uptake pathways. To investigate this, Zhao et al. studied AGS-16C3F, an ADC targeting ENPP3 (absent in megakaryocytes) with an MMAF payload, using the same in vitro model [60]. Despite lacking ENPP3 expression, AGS-16C3F was internalized and disrupted megakaryocyte differentiation, implicating macropinocytosis as a non–target-dependent uptake mechanism. These findings complement those of Uppal et al. and suggest that megakaryocyte sensitivity to ADCs may be generally driven by intracellular payload release rather than target-specific uptake. Together, these studies highlight that ADCs may enter megakaryocytes via both FcγRIIa-mediated and nonspecific pathways, such as macropinocytosis.

3.2.2. Interstitial lung disease

ILD is one of the most serious and potentially fatal adverse events associated with T-DXd therapy [12]. In the DESTINY-Breast01 trial, the overall incidence of ILD was 13.6%, with Grade 5 events (fatal) reported in 2.2% of patients [4]. The DESTINY-Breast03 trial showed a higher ILD incidence of 16.7% in the T-DXd group, significantly greater than the 3.4% observed with T-DM1; Grade ≥3 ILD occurred in 0.8% of patients [12]. Similarly, the DESTINY-Breast04 trial reported an ILD incidence of 12.1%, with Grade 3–4 events in 1.3% and Grade 5 in 0.8% of patients [32].

Recent studies have revealed that T-DXd–induced ILD represents an off-target toxicity mediated by Fc domain interaction with Fcγ receptor III (FcγRIII) on alveolar macrophages (AMs), independent of antigen recognition [54,61]. Upon Fc–FcγRIII binding, T-DXd is internalized by AMs, leading to their apoptosis and polarization toward a proinflammatory phenotype marked by high SPP1 expression. These activated AMs release proinflammatory cytokines and chemokines, disrupting local immune homeostasis and inducing interstitial inflammation and lung injury [61] Notably, this process originates in the perivascular lung niche and occurs independently of HER2 targeting, indicating a macrophage-driven, non–antigen-dependent toxic pathway. Based on this mechanism, Wei et al. proposed an “ADC + A″ strategy: using parental antibodies or human IgG1 monoclonals to preoccupy FcγRIII on AMs, thereby reducing T-DXd uptake, dampening inflammatory signaling, and mitigating ILD risk [61].

3.2.3. Gastrointestinal toxicity

Gastrointestinal (GI) toxicities are among the most prevalent off-target adverse effects of ADC therapy. In the DESTINY-Breast04 trial, T-DXd significantly increased the incidence of nausea (73.0%), vomiting (34.0%), and diarrhea (22.4%) compared to the control group (23.8%, 9.9%, and 18.0%, respectively), with Grade ≥3 events reaching 4.6%, 1.3%, and 1.1%, respectively [32]. T-DXd delivers DXd, a topoisomerase I inhibitor, which disrupts the proliferation of gastrointestinal mucosal epithelial cells, inducing apoptosis and mucosal injury [62]. In addition, DXd may activate enterochromaffin cells to release serotonin (5-hydroxytryptamine, 5-HT), stimulating peripheral vagal afferents and the central emetic center, and causing nausea and vomiting [63].Moreover, the bystander effect of T-DXd allows the cytotoxic payload to diffuse into adjacent normal intestinal cells, amplifying tissue injury and compromising mucosal integrity. Its prolonged plasma half-life may sustain GI toxicity [63]. Finally, HER2 contributes to epithelial barrier maintenance and mucosal repair in the gut; inhibition of HER2 signaling by ADCs may impair mucosal regeneration, thereby worsening GI toxicity [64].

4. Mitigation strategies

4.1. Clinical management strategies

Based on the previously discussed toxicity mechanisms of HER2-targeted ADCs and clinical trial evidence, this section systematically outlines standard management strategies for key AEs. Table 2 compiles evidence-based management recommendations for representative ADCs, including both approved and investigational agents, to serve as a practical reference for clinical decision-making. All management measures and incidence rates are derived from pivotal clinical trials and established expert consensus.

Table 2.

Adverse events and management strategies of HER2-Targeted antibody-drug conjugates in HER2-Positive breast cancer.

Adverse Event	Anti-HER2 ADC	Any Grade	≥3 Grade	Management Strategy	References
Thrombocytopenia	T-DM1	28.0%	12.9%	Grade 1–2: Continue at the same dose. Grade 3: Withhold until recovery to ≤ Grade 1, then resume at the same dose. Grade 4: Withhold until recovery to ≤ Grade 1, then resume at a reduced dose level. Platelet transfusion is required for patients with platelet count <10 × 109/L or bleeding with platelet count <50 × 109/L.	[3,13,15,32,65]
	T-DXd	23.7%	5.1%
	ARX788	29.5%	2.7%
Anemia	T-DM1	10.4%	2.7%	Grade 1–2: Supportive care with iron supplementation. Grade 3–4: Withhold treatment and consider transfusion until recovery to ≥ Grade 2, then resume at a reduced dose level.	[3,13,17,32,66,67]
	T-DXd	33.2%	8.1%
	RC48	15.4%	0.7%
	ARX788	11.4%	0.9%
	A166	23.5%	2.5%
	SYD985	14.2%	2.4%
Neutropenia	T-DM1	5.9%	2.0%	Grade 1–2: Continue dose with G-CSF support. Grade 3: Withhold treatment, administer G-CSF until recovery to ≥ Grade 2, then resume at the same dose. Grade 4: Withhold until recovery to ≥ Grade 2, then resume at a reduced dose level.	[13,17,32,66,68]
	T-DXd	33.2%	13.7%
	RC48	47.1%	17.6%
	SYD985	10.8%	4.9%
	ARX788	11.4%	0.9%

Interstitial Lung Disease (ILD)	T-DXd	12.1%	2.1%	Follow the "5S Strategy": Screening: Carefully select patients and establish appropriate monitoring based on initial risk assessment. Scan: Perform regular high-resolution computed tomography (HRCT) scans every 12 weeks. Synergy: Educate both the clinical team and patients about ILD risk and management strategies. Suspend: Immediately suspend T-DXd upon clinical suspicion of ILD, regardless of grade. Steroid: Initiate corticosteroid treatment promptly upon confirmed diagnosis of ILD.	[12,13,32,47,68]
	ARX788	32.7%	5.9%
Hepatotoxicity (ALT/AST Elevation)	T-DM1(AST)	22.4%	4.3%	Grade 1–2: Continue at the same dose. Grade 3: Withhold until recovery to ≤ Grade 2, then resume at a reduced dose level. Grade 4: Permanently discontinue.	[3,13,66,69]
	T-DM1(ALT)	16.9%	2.9%
	ARX788(AST)	70.9%	1.4%
	ARX788(ALT)	59.5%	2.3%
	RC48(AST)	71.3%	4.4%
	RC48(ALT)	61.8%	1.5%
Cardiotoxicity	T-DM1	1.7%	0.0%	For asymptomatic or mild (LVEF 40-49%) to moderate (LVEF <50%) dysfunction: Initiate ACEI/ARB therapy, continue ADC treatment, and increase LVEF monitoring frequency. For LVEF decline ≥10% from baseline or LVEF <40%: Withhold treatment and initiate heart failure therapy. May consider resuming treatment (potentially at a reduced dose) only after LVEF improves to ≥40% and symptoms resolve, with close cardiac monitoring. Permanently discontinue if LVEF does not improve or declines further.	[3,12,68,70]
	T-DXd	4.3%	0.0%

Adverse Event	Anti-HER2 ADC	Any Grade	≥3 Grade	Management Strategy	References
Ocular Toxicity	ARX788	74.5%	19.1%	Grade 1–2: Artificial tears and ophthalmology consultation. Grade 3: Withhold treatment. Grade 4: Permanently discontinue.	[13,15,17,71]
	A166	84.0%	30.9%
	SYD985	78.1%	21.2%
Peripheral Neuropathy	RC48	52.2%	5.9%	Grade 1–2: Supportive care (e.g., mecobalamin). Grade 3–4: Withhold until improvement to ≤ Grade 2, then resume at a reduced dose.	[15,17,66,68]
	SYD985	11.5%	0.0%
	A166	53.1%	6.2%
Nausea	T-DXd	73.0%	4.6%	Prophylaxis: Standard antiemetic regimen (dexamethasone + 5-HT3 receptor antagonist). For high-risk patients, add an NK1 receptor antagonist. Treatment: For breakthrough nausea/vomiting (NV), consider adding olanzapine. If Grade ≥3 NV persists for >7 days, reduce the ADC dose by one level.	[13,17,32,47,66,68]
	RC48	39.0%	1.5%
	ARX788	13.2%	0.5%
	SYD985	25.3%	1.0%
Vomiting	T-DXd	34.0%	1.3%
	T-DM1	19.0%	0.8%
	RC48	15.4%	0.0%
	ARX788	9.1%	0.0%
	SYD985	12.5%	0.3%

Adverse Event	Anti-HER2 ADC	Any Grade	≥3 Grade	Management Strategy	References
Diarrhea	T-DXd	22.4%	1.1%	Grade 1–2: Loperamide and lifestyle/dietary modifications. Grade 3–4: Immediately withhold treatment; manage with loperamide, consider adding octreotide, and provide intravenous hydration. Hospitalization may be required for severe cases.	[13,17,32,47]
	T-DM1	23.3%	1.6%
	ARX788	5.0%	0.5%
	SYD985	20.8%	1.0%

4.2. Antigen selection and antibody modification

The therapeutic index of ADCs is driven foremost by antigen selection, because antigen expression and trafficking dictate targeting precision and payload delivery. An ideal target typically meets four criteria: (1) tumor selectivity—high expression in tumor and low/absent expression in normal tissues; (2) an accessible and stable extracellular epitope to minimize shedding and nonspecific binding; (3) efficient internalization to ensure uptake of the ADC–antigen complex; and (4) stable expression under treatment pressure to limit downregulation or mutation [72]. In practice, few antigens satisfy all four, so antibody engineering is used to reclaim selectivity and widen the safety margin.Bispecific ADCs (bsADCs). By engaging two tumor-associated antigens simultaneously, bsADCs increase functional selectivity through co-expression (“AND-logic”) and/or avidity effects, thereby improving tumor targeting while sparing single-positive normal tissues. They can also mitigate resistance arising from downregulation of a single antigen [73]. For example, ZW49 (HER2 × HER3) improved responses in HER2-positive tumors with HER3 co-expression [74]. To curb EGFR-related skin toxicity, an EGFR × c-Met–MMAE bsADC preserved potent killing of double-positive tumor cells yet showed reduced binding and toxicity to normal cells, outperforming mono-EGFR ADCs [75,76]. These data support bsADCs as a structure-level strategy to raise selectivity without sacrificing potency; practically, success depends on antigen co-expression patterns, epitope geometry, and maintaining internalization competence.

Probody drug conjugates (PDCs). PDCs apply a mask–unmask design: a masking peptide is fused to the paratope via a tumor protease–cleavable linker, suppressing binding in normal tissues and restoring target engagement within protease-rich tumor microenvironments [77]. This conformational gating retains tumor delivery while reducing on-target toxicity in healthy tissues. Key considerations include interpatient and intratumoral protease heterogeneity (risk of under-activation) and preserving pharmacokinetics after unmasking. Together, rational target selection plus bsADC/PDC engineering provides complementary routes to improve selectivity, sustain internalization, and ultimately expand the ADC therapeutic window while controlling normal-tissue exposure.

4.3. Payload and linker optimization

One of the primary mechanisms of off-target toxicity in ADC therapy is the premature release of cytotoxic payloads in non-target tissues, largely due to suboptimal control over the timing and location of payload release. An ideal linker should remain highly stable in plasma to minimize systemic toxicity, while being efficiently cleaved within tumor cells or their microenvironment to ensure precise payload release. Thus, modulating the stability and cleavage kinetics of linkers can significantly reduce off-target toxicity and enhance the therapeutic index of ADCs.

Site-specific antibody-drug conjugation represents a key advancement in next-generation ADC design, enabling precise control over payload attachment sites on the antibody, thereby enhancing molecular homogeneity, stability, and therapeutic index [78]. The THIOMAB platform developed by Junutula's team employs site-directed cysteine mutagenesis (Ala114Cys) in the antibody CH1 domain to introduce reactive thiol groups (-SH) at predetermined positions, thereby enabling precise control of drug conjugation sites [79]. In contrast to conventional ADCs relying on stochastic lysine conjugation, this technology enables site-specific drug coupling at predetermined positions on each antibody molecule, yielding a homogeneous drug-to-antibody ratio (DAR = 2). The site-specific conjugation approach eliminates off-target toxicity associated with heterogeneous drug distribution in conventional lysine-conjugated ADCs, while the engineered cysteine-mediated drug-antibody linkage demonstrates superior stability. Experimental data demonstrated that the anti-MUC16 TDC, despite having only half the DAR of conventional ADCs, maintained comparable in vivo antitumor efficacy while exhibiting significantly improved maximum tolerated doses (MTDs) in both rat and cynomolgus monkey models. Notably, substantial reductions in hematological and hepatic toxicities were observed [79]. Consistent with these findings, toxicity evaluation of the anti-HER2 TDC in cynomolgus monkeys similarly demonstrated a substantially improved tolerance dose compared to conventional lysine-conjugated ADC molecules [80]. These results robustly demonstrate the therapeutic advantages of site-specific conjugation technology in improving the therapeutic window of ADCs. Currently, several breast cancer-targeted candidates employing this technology have advanced to clinical validation, including ladiratuzumab vedotin (NCT01969643), a LIV-1-targeting ADC utilizing cysteine-engineered conjugation, and ARX788 (NCT04829604), a HER2-directed ADC incorporating non-natural amino acid-based site-specific conjugation.

Moreover, Simmons et al. demonstrated that incorporating short-chain polyethylene glycol (PEG) segments into linkers enhances ADC hydrophilicity, reduces nonspecific interactions with non-target tissues, slows plasma clearance, and improves pharmacokinetics and therapeutic selectivity [81]. This strategy has proven clinically valuable in two FDA-approved ADCs—loncastuximab tesirine and sacituzumab govitecan—both of which utilize PEG modifications to broaden their therapeutic window and achieve effective yet tolerable outcomes.Ongoing research also focuses on novel payloads, including cytotoxins with unique mechanisms of action, multifunctional or immunomodulatory conjugates, and non-cytotoxic functional payloads. These strategies may fundamentally mitigate payload-associated toxicity and broaden the therapeutic scope and safety margins of ADCs [82]. In summary, optimization of linker chemistry, implementation of site-specific conjugation, hydrophilic modifications, and the development of novel low-toxicity payloads are collectively advancing ADC development toward higher therapeutic indices and reduced systemic toxicity.

4.4. Predictive tools for toxicity assessment

Given the structural complexity and mechanistic diversity of ADCs, traditional animal models incompletely predict human toxicities. To address this gap, machine learning (ML) has been proposed to integrate ADC chemical features (antibody sequence, linker chemistry, payload toxicophores) with patient multi-omics and clinical toxicology datasets to build high-accuracy adverse-event prediction models [83]. Such models can streamline toxicity assessment, reduce reliance on animal studies, and accelerate early development. Importantly, explainable ML (e.g., SHAP, LIME) can highlight contributory features and plausible mechanisms, thereby informing structure–linker–payload optimization and supporting individualized dosing and risk stratification. Early risk identification may also lower trial attrition and development costs [83].

A complementary approach is wearable biosensing (WBS) for real-time, noninvasive monitoring of toxicity. WBS enables continuous tracking of vital signs and biomarkers in sweat, saliva, or interstitial fluid, supplying actionable data for dose adjustment and therapeutic-window management [84]. As a precedent, an ongoing CAR-T study (NCT05123001) uses an ECG- and photoplethysmography-based wearable system to surveil inflammatory and neurotoxicity signals and to trigger early clinical intervention. This framework provides a feasible template for ADC toxicity monitoring and early warning in routine care.

4.5. Dose optimization

Dose optimization strategies. Many ADC toxicities are exposure–dependent, making dosing refinement key to improving the therapeutic index [85], Four main approaches include:

First, Model-informed dose caps — Pharmacokinetic/pharmacodynamic (PK/PD) modeling defines safe exposure limits based on body surface area (BSA), body weight, and organ function, avoiding overexposure in patients with high BSA [86].

Second, Stage-tailored duration — Advanced disease is treated until progression/toxicity; adjuvant settings use fixed cycles (e.g., T-DM1 × 14). Extending duration (e.g., polatuzumab vedotin, 8 vs. 6 cycles) increased severe neuropathy by >50% [87]. Thus, duration limits should be tailored to disease stage and toxicity risk.

Frequency adjustment — Fractionated dosing lowers C_max, reducing peak-related toxicity, as in Mylotarg's reapproval after hepatotoxicity reduction [88].

Finally, Response-adapted dosing. Early response can guide dose de-escalation without compromising benefit. With inotuzumab ozogamicin, patients achieving CR/CRi after cycle 1 were dose-reduced from 1.8 to 1.6 mg/m², individualizing exposure while maintaining efficacy [89].

5. Summary

The use of antibody–drug conjugates (ADCs) has significantly improved outcomes for patients with HER2-positive breast cancer, yet associated adverse effects—particularly thrombocytopenia, interstitial lung disease, and cardiotoxicity—remain critical factors limiting clinical application and affecting quality of life. A comprehensive understanding of these toxicities, combined with standardized monitoring, early intervention, and multidisciplinary management, is essential to minimize treatment interruptions and maximize clinical benefits. Looking ahead, optimizing antigen selection, linker and payload design, and incorporating innovations such as bispecific and PDCs may enhance efficacy while further reducing toxicity risks, ultimately providing safer and more controllable treatment options for patients with HER2-positive breast cancer.

List of abbreviations

Abbreviation	English Full Name	Abbreviation	English Full Name
ADC	Antibody-Drug Conjugate	FDA	U.S. Food and Drug Administration
ADCC	Antibody-Dependent Cellular Cytotoxicity	GI	Gastrointestinal
AE	Adverse Event	HER2	Human Epidermal Growth Factor Receptor 2
ALT	Alanine Aminotransferase	HRCT	High-Resolution Computed Tomography
AM	Alveolar Macrophage	IgG	Immunoglobulin G
AST	Aspartate Aminotransferase	ILD	Interstitial Lung Disease
bsADC	Bispecific Antibody-Drug Conjugate	LVEF	Left Ventricular Ejection Fraction
BSA	Body Surface Area	mAb	Monoclonal Antibody
CKAP5	Cytoskeleton-Associated Protein 5	ML	Machine Learning
CR	Complete Response	MMAE	Monomethyl Auristatin E
CRi	Complete Response with Incomplete Hematologic Recovery	MTD	Maximum Tolerated Dose
CT	Computed Tomography	NRG1	Neuregulin-1
DAR	Drug-to-Antibody Ratio	ORR	Objective Response Rate
DCR	Disease Control Rate	OS	Overall Survival
DILI	Drug-Induced Liver Injury	PDC	Probody Drug Conjugate
DM1	Derivative of Maytansine 1	PEG	Polyethylene Glycol
DoR	Duration of Response	PFS	Progression-Free Survival
DXd	Deruxtecan payload	PK/PD	Pharmacokinetic/Pharmacodynamic
ECG	Electrocardiogram	ROS	Reactive Oxygen Species
EGFR	Epidermal Growth Factor Receptor	SMCC	Succinimidyl-4-(N-maleimidomethyl)cyclohexane-1-carboxylate
FcγR	Fc Gamma Receptor

CRediT authorship contribution statement

Xiaoyu Liu: Writing – original draft, Validation, Methodology, Data curation. Saige Yin: Writing – review & editing, Writing – original draft, Investigation, Funding acquisition. Xiyin Li: Writing – review & editing, Writing – original draft, Investigation, Funding acquisition, Data curation, Conceptualization. Jianyun Nie: Writing – review & editing, Writing – original draft, Investigation, Funding acquisition, Conceptualization.

Funding

This work was supported by the First-Class Discipline Team of Kunming Medical University (Kunming Medical University Breast Cancer Precision & Translational Medicine Research Team 2024XKTDYS08), National Natural Science Foundation of China (81960479 and 82360555), Biomedical Projects of Yunnan Key Science and Technology Program (202302 AA310046), Program of Yunnan Fundamental Research Project (202301AU070218), Project of Yunnan Applied Basic Research Project-Kunming Medical University Union Foundation (202301AY07001-165), Beijing Science and Technology Innovation Medical Development Foundation (KC2021-JF-0167-22), Youth Talent Basic Research Project of Yunnan Provincial Department of Education (2024J0243) and Yunnan Province Caiyun Postdoctoral Support Program Innovation Project.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.