The next frontier in antibody-drug conjugates: challenges and opportunities in cancer and autoimmune therapy

ADC therapy for malignant tumors

As a novel precision therapeutic strategy, ADCs have demonstrated significant potential for application in various types of malignant tumors in recent years. By targeting tumor-associated antigens and delivering highly potent cytotoxic payloads, ADCs can achieve selective killing of tumor cells while minimizing toxic damage to normal tissues^{[22,32,33,50]}.

In hematological malignancies, ADCs precisely recognize and bind to differentiation antigens highly expressed on the surface of malignant cells, thereby achieving high-intensity killing at lower doses^[3,51,52].

In acute myeloid leukemia (AML), the target CD33 has garnered widespread attention. In fact, CD33 is not relevant only for hematologic malignancies. Typical ADC drugs such as Gemtuzumab Ozogamicin utilize an anti-CD33 monoclonal antibody to precisely deliver the DNA-damaging molecule calicheamicin into tumor cells. Following internalization, the payload is released in the acidic environment of lysosomes, thereby inducing double-strand DNA breaks and triggering apoptosis^[53-55].

For ALL, the primary target is the CD22 molecule. Inotuzumab ozogamicin (INO) is an ADC targeting CD22. After internalization into the lysosomes of tumor cells, calicheamicin is released and binds to DNA, leading to apoptosis. A Stage III clinical study (NCT01564784) has shown that INO exhibits an overall response rate (ORR) of 80.7% in patients with relapsed or refractory ALL, with 78.4% of these patients achieving minimal residual disease (MRD) negativity^[56,57].

For lymphoma, molecules such as CD19, CD22, CD30, and CD79b are commonly targeted, highly expressed molecules. Loncastuximab tesirine has demonstrated significant single-agent antitumor activity in patients with relapsed/refractory diffuse large B-cell lymphoma (DLBCL), with an ORR of 48.3% and a complete response (CR) rate of 24.1%. The Polatuzumab vedotin in combination with bendamustine and rituximab (pola-BR) regimen achieved a CR rate of 40.0%, significantly higher than the 17.5% in the control group, and significantly prolonged patients’ progression-free survival (PFS) and overall survival (OS)^{[33,51,58-62]}.

Compared to the relatively homogeneous target expression in hematological malignancies, solid tumor therapy faces challenges such as tumor heterogeneity, drug penetration, and uneven target expression. However, ADCs continue to show broad therapeutic promise in multiple cancer types^[22,63].

In breast cancer, ADC drugs have become a breakthrough in targeted therapy. trastuzumab emtansine (T-DM1) was the first approved HER2-targeting ADC, which precisely identifies HER2-positive tumor cells via its antibody component and delivers emtansine intracellularly. Trastuzumab deruxtecan (T-DXd), through improved linker design and a more potent topoisomerase I inhibitor, has demonstrated broader applicability, especially in patients with HER2-low breast cancer, by killing adjacent HER2-negative cells via the “bystander effect”. Sacituzumab govitecan has been approved for treating patients with refractory or metastatic triple-negative breast cancer (TNBC)^[64-68].

In lung cancer, ADC drugs have demonstrated significant therapeutic potential against various targets. In non-small cell lung cancer (NSCLC), HER2 and Trop-2 are important targets. T-DXd, as a HER2-targeting ADC, has shown notable antitumor activity. Sacituzumab govitecan, by targeting Trop-2, is particularly suitable for patients with relapsed or refractory disease. In small cell lung cancer (SCLC), approximately 80% of patients have high DLL3 expression, and Rovalpituzumab tesirine has demonstrated an ORR of 38% in patients with high DLL3 expression^[69-72].

In urothelial carcinoma, enfortumab vedotin is an ADC drug targeting Nectin-4, which has been approved for the treatment of refractory metastatic bladder cancer. Clinical trials have shown that enfortumab vedotin exhibits high response rates and significant survival benefits in bladder cancer patients who are platinum-resistant or have failed immunotherapy.

Overall, ADCs in cancer treatment, through the combination of precise targeting and potent payloads, not only extend patient survival time but also provide more options for patients with different molecular subtypes of tumors, serving as an important bridge connecting targeted therapy and chemotherapy^[73,74].

However, as ADCs are widely applied across various oncological indications, a series of structural limitations and mechanistic challenges have gradually been exposed in clinical practice. Especially when expanding to non-malignant indications such as autoimmune diseases, how to maintain the balance between efficacy and safety is becoming a central focus of research.

Major challenges and limitations of ADCs in therapy

Currently, despite the aforementioned advantages of ADCs, this therapeutic modality still faces considerable limitations, such as complex pharmacokinetics, numerous adverse reactions, and manufacturing and translational hurdles^[37,93,94]. Furthermore, when applied to the treatment of autoimmune diseases, ADCs also encounter a considerable number of additional challenges. Beyond the traditional adverse reactions and delivery obstacles, drug resistance, as a more profound limiting mechanism, is also becoming increasingly prominent in ADC therapy. Its mechanisms are complex and heterogeneous, and it has become a core obstacle to limited clinical efficacy. This will be systematically elaborated on in Section 5.

Antigen-related resistance

Sufficient target antigen expression on tumor cells is a prerequisite for ADC efficacy. Reduction in the quantity and alterations in the distribution of the target antigen are the most direct factors contributing to ADC resistance. Studies have found that after long-term exposure to T-DM1, HER2 protein levels significantly decrease in resistant cell lines, suggesting that continuous ADC pressure can lead to target antigen downregulation^[110,111].

Spatial heterogeneity of the target antigen within tumors can also impair ADC efficacy. Spatial heterogeneity in tumor cells leads to significant differences in antigen expression across different tumor regions. Even with an overall determination of “HER2 high expression”, a substantial number of HER2-negative subpopulations may still exist, reducing ADC penetration depth and coverage. In the ZEPHIR study, PET imaging using radiolabeled antibodies to assess HER2 distribution revealed that in approximately 46% of patients, less than half of the tumor regions showed uptake of the tracer antibody^[112]. In such cases, the ADC and its bystander effect may be insufficient to completely eradicate antigen-negative target cells. As treatment progresses, target cells may develop new mutations under selective pressure, thereby affecting antigen expression and cell tolerance^[66-68,113].

Changes in the nature of the target antigen can also mediate resistance. “Antigen drift”, where the target antigen undergoes structural changes due to gene mutations, leading to difficulties in ADC recognition, is one such instance. In anti-HER2 therapy, tumor cells can express a truncated form of HER2 (p95HER2) that lacks the extracellular domain, reducing the efficacy of trastuzumab-based drugs^[114]. After treatment with the anti-CD22 ADC INO in ALL, 11% of relapsed patients developed acquired CD22 mutations, including forms such as protein truncation and epitope alterations^[115].

Impaired drug transport and efflux mechanisms

For an ADC to be effective, it must be internalized by tumor cells and release its payload intracellularly. The ADC cannot function if it is not efficiently internalized, if the payload is not effectively released, or if it is expelled from the cell before it can act.

Alterations in endocytic pathways are a significant resistance mechanism. Under normal conditions, most ADCs enter lysosomes via clathrin-mediated endocytosis. However, resistant cells can alter this pathway, leading to resistance^{[108,110,116]}. For instance, in vitro studies by Sung et al. found that T-DM1-resistant cells overexpress Caveolin-1 (CAV-1), causing HER2 receptors to be internalized primarily through the caveolae-mediated pathway. This results in reduced co-localization of the antibody-drug complex with lysosomes, thereby diminishing the cytotoxic effect of T-DM1^[117].

Lysosomal dysfunction itself can also affect the release of the cytotoxic payload. Decreased activity of proteolytic enzymes or insufficient acidification within lysosomes can lead to incomplete antibody degradation, preventing the full release of the payload. Ríos-Luci et al. reported that some T-DM1-resistant cells exhibit elevated lysosomal pH, leading to impaired processing of T-DM1 within the lysosome. This highlights the role of maintaining lysosomal acidity in combating resistance^[118].

For ADCs with non-cleavable linkers, payload release is more dependent on specific transport mechanisms. The specific metabolite Lysine-MCC-DM1, generated after the lysosomal degradation of T-DM1, requires the lysosomal membrane protein SLC46A3 for its export. Research by Tomabechi et al. found that effective SLC46A3 inhibitors, such as clarithromycin, erythromycin, and rifabutin, can significantly reduce T-DM1 cytotoxicity. This is consistent with the fact that downregulation of SLC46A3 expression in cancer cells is associated with T-DM1 resistance. Low SLC46A3 expression is correlated with the T-DM1 resistance phenotype and can serve as a predictive biomarker for its efficacy^[119].

Tumor cells can significantly reduce the cytoplasmic concentration of payloads by upregulating drug efflux pumps. ATP-binding cassette (ABC) transporters (e.g., P-glycoprotein/MDR1, BCRP/ABCG2) actively expel hydrophobic chemotherapeutic drugs from the cell, thereby reducing the intracellular concentration of the active drug. Commonly used payloads (including vinca alkaloid/taxane-type tubulin polymerization inhibitors and camptothecin-type topoisomerase inhibitors) are known substrates of ABC transporters^[41,120,121].

P-gp (MDR1) is one of the most extensively studied ABC transporters, and it can recognize and pump out auristatin- and maytansinoid-class payloads. The DM1 payload of T-DM1 can be effluxed by P-gp, and resistant cells often exhibit P-gp upregulation. Li and Wang et al. also found that resistant breast cancer clones showed cross-resistance to free DM1 and vinca alkaloids, accompanied by upregulation of the MDR1 gene, indicating the role of P-gp-mediated efflux^[122,123].

Overall, ABC transporter-mediated drug efflux is one of the common mechanisms of ADC resistance. When this mechanism is present, simply increasing the ADC dosage is often of little benefit, as the cells can rapidly pump out the toxin. To address this situation, researchers are exploring strategies such as using drugs that are less easily recognized by efflux pumps or co-administering efflux inhibitors to overcome MDR-mediated ADC resistance (see Section 6 for details).

Mechanisms of cell death pathway inhibition

Even if ADCs successfully deliver their toxic payload into the cell, tumor cells can still evade apoptosis by modulating damage response and cell death pathways. Dysfunction of apoptotic signaling pathways is a common mechanism of tumor resistance to various therapies, and ADC resistance is no exception. In cases of resistance to INO in acute leukemia, inactivating mutations in tumor suppressor genes such as TP53 and ATM are often detected^[115]. Furthermore, tumor cells can counteract the effects of the toxin by upregulating survival signals or anti-apoptotic proteins. For example, overexpression of anti-apoptotic proteins such as BCL-2 and BCL-XL can render tumor cells insensitive to apoptotic stimuli mediated by Gemtuzumab Ozogamicin or BV. In preclinical models, combining these ADCs with BCL-2 inhibitors can partially restore their killing effect, supporting the role of anti-apoptotic pathways in resistance^[124,125].

On the other hand, resistance and adaptation targeting the site of action of the ADC’s payload can also lead to resistance. For example, the DM1 payload of T-DM1 targets microtubules, and long-term administration may lead tumor cells to develop resistance by modulating microtubule-associated molecules. This includes mutations in β-tubulin genes, alterations in isotype expression profiles, or upregulation of microtubule-stabilizing proteins, thereby attenuating DM1’s interference with microtubule dynamics. For topoisomerase I inhibitor payloads, tumor cells can counteract toxicity by enhancing their DNA damage repair capabilities. For instance, with DXd, the active metabolite of DS-8201, studies have found loss-of-function mutations in the DNA repair-related gene SLX4 (an assembly scaffold protein for endonucleases) in tumors. These mutations cause cells to respond differently to DNA damage induced by topoisomerase inhibition, thereby avoiding cell death^[126].

Thus, when cellular signal transduction and damage response to the toxin are altered, the lethal effect of ADCs is greatly diminished. These mechanisms enable resistant cells to bypass the execution of apoptotic programs in response to ADC-induced damage, thereby evading death. In the future, combining ADCs with pro-apoptotic drugs or drugs targeting repair pathways may potentially restore the sensitivity of such resistant cells to ADCs.

Specificity of ADC resistance mechanisms in autoimmune diseases

Although ADCs face resistance challenges in both autoimmune diseases and cancer treatment, the pathological characteristics and therapeutic needs of autoimmune diseases dictate that their resistance mechanisms possess significant uniqueness beyond the traditional ADC resistance observed in oncology.

The immunogenicity of ADCs may be a prominent and unique challenge in their therapeutic application. Research by Pizano-Martinez et al. shows that the incidence of anti-drug antibodies (ADAs) induced by biologics in patients with autoimmune diseases can be as high as 40%^[127]. This difference stems from the pre-existing overactivated state of the immune system in autoimmune disease patients, making them more prone to generating immune responses against exogenous proteins. Carrasco-Triguero noted that while ADA generation does not necessarily imply a reduction in ADC titer or therapeutic efficacy, there are indeed cases where the corresponding ADC is undetectable in patients with elevated ADAs^[128].

Cross-reactivity is another immunogenicity challenge. Carrasco-Triguero’s research indicates that although patients generally develop ADAs only against the mAb, they might also simultaneously produce antibodies against the antibody, linker, and payload, increasing the risk of immunogenicity. In autoimmune diseases, such multi-component immune responses may be more complex, as the patient’s immune system has already mounted abnormal responses to multiple self-antigens and might more readily recognize and attack different components of the ADC^[128].

Another difficulty is the highly heterogeneous and dynamically changing nature of pathogenic immune cell populations in autoimmune diseases. With the advancement of single-cell transcriptomic technologies, an increasing number of studies have noted the heterogeneity of pathogenic immune cell expression in autoimmune diseases. As Zhao et al. pointed out, these newly discovered cell subtypes may suggest new therapeutic targets^[129]. However, this also places higher demands on the design of novel ADCs; if these rare cell subpopulations cannot be targeted, the desired therapeutic outcomes may not be achieved, thereby affecting the ADC’s therapeutic coverage and durability of treatment.

In summary, ADC resistance mechanisms in autoimmune diseases exhibit significant uniqueness, primarily manifested in increased immunogenicity risks and dynamic changes in immune cell subpopulations. These unique mechanisms necessitate the adoption of strategies different from those for oncology ADCs when developing ADCs for autoimmune diseases, including more stable molecular designs, lower immunogenicity, multi-target coverage, and personalized treatment regimens. Only by fully recognizing and addressing these unique challenges can ADCs exert their potential advantages in the treatment of autoimmune diseases.

Molecular-level strategy innovations

Innovation at the molecular design level is the most direct and effective approach to overcoming ADC resistance. Through meticulous refinement of the three core components of ADCs - the antibody, linker, and payload - the drug’s anti-resistance capabilities can be fundamentally enhanced. This subsection will primarily discuss how to bypass transporter-mediated resistance mechanisms through optimized payload and linker design, as well as how to utilize multi-target and dual-payload strategies to broaden the killing spectrum and reduce the risk of resistance.

The initial focus is on the optimization of ADC components, namely the antibody, linker, and payload. The choice of antibody directly influences the targeting specificity and tumor penetration of an ADC. In recent years, small-format antibody fragments [such as Fab fragments, single chain variable fragments (scFv), and designed ankyrin repeat proteins (DARPins)] have demonstrated unique advantages in novel ADC designs^[22,130,131]. Furthermore, modifications to the Fc region can also alter functions such as the ADCC activity of ADCs. For example, the S239D/I332E mutations in the Fc region have been demonstrated to significantly improve binding to FcγRIIIa, thereby enhancing ADCC^[132].

Modern linker design thus may emphasize not only stability in the bloodstream but also specific drug release in the target organ microenvironment. As mentioned earlier, cleavable linkers can take advantage of differences between the intracellular and extracellular environment, ensuring that payload release mainly occurs within target cells. While linkers responsive to a single stimulus (e.g., pH or a specific protease) have already shown clear benefits in ADC design, multifunctional “intellective” linkers that combine multiple stimuli offer even higher precision in drug release. Some linkers integrate acid responsiveness with protease sensitivity, enabling them to be “partially activated” under low-pH conditions and subsequently cleaved by tumor-associated enzymes. For example, hydrazone bonds respond to acidic pH, making them active in the mildly acidic tumor environment^[133]. Meanwhile, Cathepsin B, which is often overexpressed in tumor cells, specifically recognizes and cleaves linkers containing enzyme-sensitive sequences (such as Val-Cit). Such linkers enable a two-step release mechanism: partial activation in the acidic microenvironment, followed by complete payload release upon Cathepsin B cleavage^[133-135].

Additionally, the development of site-specific conjugation technologies for linkers has made it possible for each antibody molecule to carry a uniform number of payloads (resulting in a homogeneous drug-to-antibody ratio, DAR), enabling positional control, maximally reducing product heterogeneity, and preserving key epitope-binding regions^[94]. Of course, the limitations of such strategies lie in higher demands on antibody engineering and conjugation chemistry, and correspondingly increased production costs. These innovations are expected to further enhance the efficacy of ADCs against resistant tumors while maintaining safety, leading to more refined ADC molecular designs.

As previously mentioned, resistance to many ADCs stems from the overexpression of ABC transporters (such as P-gp) by tumor cells. To address this mechanism, Kovtun et al. designed an ADC in which the traditional hydrophobic SMCC linker was replaced with a hydrophilic PEG4-Mal linker. While carrying the same DM1 toxin, this hydrophilic linker-ADC exhibited higher cytotoxic activity in MDR1-overexpressing tumor models, successfully bypassing P-gp-mediated resistance^[41]. In fact, such masking groups can not only enhance efficacy but also reduce non-specific interactions of ADCs with normal tissues, mitigating side effects such as aggregate formation, rapid clearance, and potential high immunogenicity caused by highly hydrophobic payloads.

To enhance the killing spectrum, another key strategy involves engineering bispecific and dual-payload ADCs. These next-generation designs represent a critical direction for addressing resistance driven by antigen loss or signaling pathway compensation. Although these new types of ADCs are currently in early clinical or preclinical research stages, the results presented so far have been largely encouraging.

Bispecific or multi-specific ADCs, by simultaneously recognizing two antigens, improve tumor binding affinity and internalization efficiency, and can circumvent resistance caused by antigen loss associated with single-target approaches. To date, over 100 different design formats for bispecific ADCs (bsADCs) have been reported in the literature^[136]. Andreev et al. constructed a bispecific ADC that bridges HER2 and the prolactin receptor (PRLR), which can co-target PRLR on the surface of HER2-positive tumor cells, leading to significantly enhanced internalization and killing effects of the ADC in tumors co-expressing HER2 and PRLR^[137]. Furthermore, combining two non-tumor-specific but complementary targets can also reduce off-target toxicity. For example, Luo et al. developed a dual-target ADC for EpCAM and CLDN3, which requires high co-expression of both antigens on tumors for effective killing, exhibiting minimal toxicity to normal cells that only highly express EpCAM but have low CLDN3 levels, thereby alleviating the toxicity limitations of single-target EpCAM ADCs^[138]. In summary, dual-target ADCs, through a “dual recognition” mechanism, enhance tumor specificity and internalization efficiency, and can, to some extent, overcome resistance caused by antigen heterogeneity or downregulation.

On a parallel track, dual-payload or multi-payload ADCs conjugate two cytotoxic drugs with complementary mechanisms of action onto the same antibody, aiming to achieve synergistic killing and reduce the probability of resistance. Yamazaki et al. constructed an anti-HER2 dual-payload ADC carrying both MMAE and MMAF toxins, which achieved complete remission in breast cancer xenograft models with low HER2 expression and resistance to traditional T-DM1, significantly outperforming the efficacy of either single-payload ADC^[139]. Mechanistically, the more potent MMAF effectively kills resistant cells, while the conjugated MMAE exerts a bystander effect to eliminate nearby antigen-negative tumor cells, thereby enhancing overall efficacy. Kumar et al. also synthesized a hetero-trifunctional linker, enabling the simultaneous conjugation of the microtubule inhibitor MMAE and the DNA intercalator PBD dimer onto a single antibody, achieving more efficient killing^[140].

Dual-payload ADCs hold the promise of broadening the antitumor spectrum and reducing the chances of tumor cells developing tolerance to a single drug, but their development also faces challenges such as precise control of the DAR and pharmacokinetic balancing. Currently, the world’s first dual-payload ADC to enter clinical trials is KH815 (an anti-TROP2 ADC) developed by China’s Kanghong Pharma (NCT06885645): this drug links two types of toxins, a topoisomerase I inhibitor and an RNA polymerase II inhibitor, to enhance antitumor efficacy by inducing DNA double-strand breaks on one hand and inhibiting RNA synthesis on the other. Interestingly, KH815 was also observed in preclinical studies to downregulate the expression of resistance-related proteins such as P-gp and HSP70 in tumor cells, thereby increasing sensitivity to chemotherapy. Table 5 lists a summary of dual-target and dual-payload ADCs in clinical trials.

It is worth noting that dual-target and dual-payload designs do not necessarily guarantee higher efficacy. Researchers have developed two anti-HER2 ADCs loaded with MMAE + SG3457 and MMAF + PNU-15968, respectively; unfortunately, neither demonstrated superior efficacy in in vitro experiments. This suggests that there are more interactions between dual-payload/dual-target ADCs and tumor cells that warrant further exploration, and the generation of their combined therapeutic effects depends on in-depth molecular interaction studies and basic/clinical trials^[140,143].

In addition to the classic optimization designs mentioned above, there have also been recent reports on other optimization strategies such as Probody-drug conjugates, Degradative-antibody conjugates, and Immune receptor agonists. However, they are still in very early preclinical research stages and will not be elaborated upon here^[94].

Innovations in therapeutic strategies

The challenge of resistance to ADC monotherapy is prompting a shift in clinical practice toward combination treatment strategies. Synergy between different drugs can not only enhance efficacy but, more importantly, prevent and overcome the development of resistance through multi-mechanistic parallel attacks. This subsection will systematically elaborate on the principles, clinical practice, and prospects of strategies combining ADCs with different types of drugs, such as immune checkpoint inhibitors (ICIs), small molecule targeted drugs, traditional chemotherapy, and emerging cell therapies.

ADCs can be combined with ICIs to exert synergistic antitumor immunity. ADC-mediated tumor cell apoptosis not only directly kills tumor cells but also releases tumor-associated antigens and danger signals, inducing immunogenic cell death (ICD) and activating dendritic cells, thereby enhancing T cell infiltration and tumor-specific immune responses. However, under chronic antigen stimulation, infiltrating T cells often exhibit an exhausted state. Combining ADCs with ICIs can fully leverage the tumor antigen release and immune activation effects triggered by ADCs, with ICIs releasing the “brakes” on T cells for tumor clearance, forming a mechanistic complement between the two^{[13,38,144,145]}. T-DXd; DS-8201a, as a novel ADC, has demonstrated significant antitumor activity and potential to overcome resistance when combined with immunotherapy. Currently, multiple clinical trials evaluating the combination of T-DXd with ICIs are underway. Early evidence has shown that the combination of T-DXd and ICIs can elicit antitumor effects exceeding those of chemotherapy or ICI monotherapy in various cancer types. In the BEGONIA Phase Ib/II trial (NCT03742102), durvalumab (a PD-L1 inhibitor) plus T-DXd achieved an ORR of 66.7% in patients with metastatic triple-negative breast cancer (mTNBC), compared to 58.3% in the control group of durvalumab plus paclitaxel, suggesting a quantifiable improvement in efficacy with the addition of T-DXd. In the urothelial cancer cohort of the DS8201-A-U105 Phase Ib study (NCT0352357), nivolumab (a PD-1 inhibitor) plus T-DXd achieved an ORR of 36.7%, including 4 CR and 7 partial responses, with a median duration of response of 13.1 months, demonstrating a depth and durability of response difficult to achieve with traditional regimens. Mechanistically, T-DXd enhances tumor antigen presentation and CD8⁺ T cell infiltration through ADCC and the “bystander effect”, providing an ample effector cell basis for ICIs to lift immune suppression, thereby creating synergistic amplification. These phenomena reveal that ADCs can “turn” “cold tumors”, which are traditionally insensitive to immunotherapy, into “hot tumors” rich in T cell infiltration, thereby enabling patients who have partial responses or resistance to ADC monotherapy to achieve more comprehensive and durable tumor control^[146].

While this combination therapy has not yet been shown to produce novel, uncontrollable toxicities, the potential for overlapping toxicities remains a key consideration, as both ADCs and ICIs can independently cause immune-related adverse events (irAEs)^[147]. Optimizing the administration sequence and dosage, and identifying the most synergistic yet safely manageable combinations, will be key to the successful application of this strategy in the future.

Combining ADCs with small-molecule targeted drugs or chemotherapy offers another promising strategy to overcome resistance. Because resistant tumors often exhibit complex mechanisms such as poly-clonality and signaling pathway compensation, which are difficult to control comprehensively with a single agent. Combining ADCs with other antitumor drugs can leverage differences in their mechanisms of action to achieve better synergistic effects, a strategy widely considered effective for overcoming resistance^[148].

In terms of molecular targeting, for ADC-resistant cells exhibiting cell cycle dysregulation or signaling abnormalities, combining ADCs with small molecule drugs that inhibit the corresponding pathways can restore ADC sensitivity. In HER2-positive breast cancer, Cyclin E overexpression can lead to T-DM1 resistance^[149]. Preclinical studies by Witkiewicz et al. found that the addition of CDK4/6 inhibitors can arrest excessive tumor cell proliferation and restore T-DM1-induced mitotic arrest and killing effects^[150]. Similarly, as previously mentioned, overexpression of anti-apoptotic proteins like BCL-2 and BCL-XL can render cells insensitive to ADCs. In vivo studies by Zoeller et al. further confirmed that neutralizing BCL-2/BCL-xL can enhance T-DM1’s killing of resistant tumors and prolong survival. This indicates that for resistant cells reliant on anti-apoptotic pathways or cell cycle escape, combining ADCs with corresponding small molecule inhibitors (such as BCL-2 inhibitors, PLK1 inhibitors, etc.) is a feasible approach^[151].

Regarding chemotherapy, combining ADCs with traditional chemotherapeutic drugs also shows complementary advantages: The Phase II TEAL study demonstrated that T-DM1 combined with lapatinib and albumin-bound paclitaxel significantly increased the pathological CR rate in patients, exhibiting a stronger tumor-shrinking effect compared to traditional “dual antibody + chemotherapy” regimens^[152]. For instance, early studies (BP22572/NCT00934856) found that combining T-DM1 with the classic taxane drug docetaxel could also enhance efficacy, but with more severe adverse reactions. This suggests that in designing ADC + traditional chemotherapy combination regimens, careful consideration must be given to balancing dosage and treatment duration to avoid severe adverse reactions. Overall, rational drug pairing and optimization of administration sequence based on resistance mechanisms are expected to maximize the efficacy of combination therapies and reduce the incidence of resistance^[153].

Furthermore, the integration of ADCs with other emerging therapies is also an area of exploration. For example, linking ADC therapy with adoptive cell immunotherapies such as CAR-T cells is thought to potentially overcome extreme cases of resistance. CAR-T cell therapy has achieved breakthrough success in hematological malignancies but is often limited in solid tumors due to antigen heterogeneity and an immunosuppressive microenvironment. When tumor cells “escape” ADC attack due to antigen negativity, CAR-T cells targeting different antigens can be considered to eliminate remaining cancer cells, achieving complementarity.

Kim et al. further validated the feasibility of this synergistic strategy by establishing CEACAM5-expressing NSCLC models. When the anti-CEACAM5 ADC SAR408701 showed limited efficacy or resistance, CAR-T cells targeting the same antigen could still effectively induce cytotoxic responses and significantly inhibit tumor growth in vivo^[154]. This indicates that CAR-T therapy can serve not only as a supplementary treatment for ADC-resistant patients but also, in early combination, intervene proactively to reduce the risk of relapse or ADC resistance. Particularly noteworthy is that CAR-T cells have a higher dependency on antigen expression levels, whereas ADCs may, to some extent, rely on the “bystander effect” to kill low-expressing or even antigen-negative tumor cells, thus exhibiting a high degree of mechanistic complementarity. Zhao’s review also proposed from a systemic perspective that this “ADC + CAR-T” dual-track immune attack strategy can not only cover a broader antigen spectrum but also address complex resistance and immune escape mechanisms in the tumor microenvironment. ADCs are responsible for delivering an initial, highly effective cytotoxic strike, while CAR-T cells, relying on immune memory and self-amplification capabilities, achieve sustained clearance, thereby forming a synergistic system of acute strike and long-term immune surveillance^[155].

Overall, data on the concurrent or sequential application of ADCs and CAR-T cells are still very limited. Considering the complex preparation and management processes required for CAR-T cell therapy, the feasibility of combining ADCs with it will face more difficulties in clinical research^[156]. However, it can be speculated that the combined application of multiple therapeutic strategies will be key to overcoming ADC tumor resistance in the future; ADCs, acting as “missiles”, can certainly be combined with immune cells or other drugs to achieve better outcomes. In summary, at the therapeutic strategy level, both the combination with classic drugs and the integration of novel immune/cell therapies have greatly broadened the application boundaries of ADCs, providing a richer arsenal for tackling the problem of resistance. We have reason to expect that after rigorous clinical trial screening, several effective ADC combination regimens will enter the standard of care, helping more patients overcome the challenges posed by drug resistance.

Unique strategies to address ADC resistance in autoimmune diseases

The application of ADC therapy in autoimmune diseases faces a series of unique challenges distinct from those in oncology, which can lead to reduced actual efficacy or “resistance” phenomena. These include enhanced drug immunogenicity, heterogeneity in target expression at lesion sites, and the safety requirements of long-term medication. Therefore, innovative strategies targeting these specific mechanisms are needed to improve the efficacy and tolerability of ADCs in autoimmune diseases. Several potential key strategies are summarized below.

Reducing the Immunogenicity Risk of ADCs: Patients with autoimmune diseases often require long-term, repeated administration. Even if an ADC formulation has low reported immunogenicity in cancer treatment, the accumulation of minor immune responses during long-term use can lead to the production of ADAs, thereby diminishing efficacy. Consequently, ADC design must aim to minimize immunogenicity. Regarding the requirement to reduce cross-reactivity and immunogenicity risks in autoimmune diseases, current ADC modifications are still focused on the antibody, linker, and payload components. For example, selecting low immunogenicity “small molecule” drugs rather than bacterial/plant toxins, and optimizing linkers to avoid the exposure of new antigenic epitopes, are considered^[103]. PROTAC payloads, acting as catalytic degraders, can specifically degrade pathogenic proteins at very low doses; their small molecule structure inherently possesses very low immunogenicity, thereby reducing immune system recognition and antibody generation^[157]. Similarly, antibody-siRNA conjugates deliver RNAi components specifically to lesion cells, silencing pathogenic genes intracellularly and avoiding the off-target damage to non-target cells caused by traditional cytotoxic drugs (chemically modified siRNAs can also inhibit innate immune activation)^[158]. Furthermore, some studies have attempted to reduce immune system recognition by engineering antibody frameworks to remove T cell epitopes or by using human proteases/enzymes as toxic effectors. Wang et al. developed a low-immunogenicity anti-CXCR4 antibody based on a human antibody framework and successfully conjugated the immunosuppressive drug Dasatinib for T cell-specific delivery^[159]. These low-immunogenicity design strategies are expected to maintain ADC activity during chronic administration while maximally reducing the production of ADAs.

Addressing Heterogeneity in Target Antigen Expression: As previously mentioned, the pathogenesis of autoimmune diseases is complex, involving multiple immune cell subsets and molecular pathways, and the expression of pathogenic targets is highly heterogeneous among different patients or disease stages. Single-target ADCs can hardly cover all pathogenic cell populations, and residual pathogenic cells can lead to disease persistence or relapse. One strategy to address this issue is to refine multi-target combinations and adaptive designs. On one hand, BsADCs can be developed and explored in autoimmune diseases. Although BsADCs are currently studied mainly in oncology, this strategy is equally attractive for combating heterogeneous target cells in autoimmune diseases. On the other hand, personalized multi-target regimens are also an important direction, involving the selection of different ADC combinations or modularly switching targets based on the patient’s pathogenic pathway. In fact, ADCs targeting different immune cells have already shown correspondingly excellent effects^[160]. Preclinical studies have shown that an anti-E-selectin Dex (Dexa-AbhEsel) ADC, an anti-TNF-α glucocorticoid ADC, and an anti-CD163 Dex (Cymac-001) ADC, used to deliver dexamethasone as an anti-inflammatory payload in vitro and in vivo respectively, have all achieved positive preclinical results in eliminating dysfunctional immune cells and pro-inflammatory cytokines^[49,88,161]. The aforementioned development of single-cell transcriptomics and other techniques has identified more rare subpopulations of heterogeneous pathogenic immune cells. While this poses greater challenges for ADC design, it also represents a new direction; new ADC designs may leverage characteristic markers of these rare cell populations obtained from single-cell transcriptomics to achieve broader disease coverage.

Although efforts by the academic community to explore ADC resistance in autoimmune diseases are noticeable overall, we must still recognize that related research remains considerably insufficient. Clinical data capable of supporting evidence-based medicine-level conclusions are extremely scarce, and current explorations face significant difficulties in clinical translation. The keys for the future lie in: establishing unified immunogenicity assessment and follow-up systems; validating safety and efficacy persistence under long-term repeated administration; accurately delineating pathogenic cell lineages using single-cell omics and spatial multi-omics to provide an operational framework for stratified medicine with multi-target or programmable ADCs; and truly translating cutting-edge technologies such as PROTAC-ADCs and siRNA-ADCs into clinical practice and real-world cohorts to form an interdisciplinary closed-loop validation. Only through multi-center, longitudinal, large-sample clinical trials, as well as collaborative iteration between industry and academia, can these conceptual strategies be truly transformed into replicable, affordable, and durably effective therapeutic regimens.

In summary, to address the multidimensional challenges of ADC resistance, scientific research and clinical practice are concurrently advancing innovations in molecular optimization and strategy integration. Whether it is the emergence of new-generation ADC molecules or the exploration of “ADC+X” combination regimens, these efforts are continuously expanding the frontiers of ADC therapy. In the field of solid tumors, multi-target, multi-payload, and multi-modal combinations will help crack the code of resistance, whereas in the realm of autoimmune diseases, the exploration and application of customized solutions addressing immunogenicity and target complexity are expected to promote the precise clinical implementation of ADCs in non-malignant diseases.

AI-driven ADC molecular design

In the realm of monoclonal antibodies, AI technology is thoroughly transforming antibody design and optimization processes. This transformation is driven by AI’s capacity to learn the fundamental rules of antigen binding, to non-linearly optimize interdependent design parameters, and to generate novel antibodies de novo. Protein structure prediction tools like AlphaFold can accurately predict the three-dimensional structures of antibody-antigen complexes, guiding the optimization of antibody affinity and specificity^[162]. Zheng et al. reported that antibody engineering guided by a deep generative model based on LSTM (Long Short-Term Memory) for structure prediction could increase affinity by up to 1800-fold, while AI-based final antibody generation and screening time could be reduced by 60%, concurrently minimizing off-target binding^[163]. The latest version, AlphaFold 3, has further improved prediction accuracy, especially in forecasting protein-protein and protein-small molecule interactions. It can precisely identify key residues at binding interfaces, predict multiple potential epitopes on an antigen to support bispecific ADC design, and assess cross-reactivity risks^[164,165].

Concerning drug disposition and stability, molecular dynamics simulations combined with machine learning can predict the stability of different ADCs in blood circulation. A platform-based quantitative systems pharmacology (QSP) model developed by Scheuher et al. can predict the in vivo half-life of ADCs by considering factors such as pH, enzyme activity, and serum protein binding^[166]. However, this method currently cannot provide direct guidance for the design of linkers and payloads.

Overall, while AlphaFold 3 builds upon the powerful protein structure prediction capabilities of AlphaFold 2 and further expands its application scope to accurately predict the interaction structures of complex molecules - including the binding modes of proteins with other proteins, ligands, RNA, and DNA - its application in prediction and development within the ADC field is currently limited to monoclonal antibody discovery. The simulation of more complex drug architectures like ADCs, which are composed of an antibody, a linker, and a drug, still awaits further research and exploration.

AI in non-invasive prediction of receptor expression status

The therapeutic efficacy of ADCs is highly dependent on the expression level of the target antigen, and traditional detection methods require invasive biopsies to obtain tissue samples. This approach faces significant limitations in the assessment of neoadjuvant therapy or for metastatic lesions where tissue samples are difficult to obtain. Deep learning methods applied to digital imaging offer new possibilities for non-invasively predicting receptor expression status.

Digital pathology combined with deep learning has demonstrated immense potential in predicting receptor expression status. Results from the Phase I HEROHE challenge showed that HER2 status in breast cancer can be predicted solely from routine hematoxylin and eosin (H&E) stained slides, circumventing the need for special immunohistochemistry (IHC) techniques^[167]. A deep learning model developed by Howard et al., integrating clinical features and digital pathology, achieved an area under the receiver operating characteristic curve (AUC) of 0.78-0.89 in external validation cohorts, significantly outperforming traditional clinical nomograms^[168]. Research by Zhao et al. indicated that although some difficulties remain in assessing recurrence risk, molecular features of TNBC, including PIK3CA mutations, BRCA2 mutations, and PD-L1 expression, can now be predicted with considerable accuracy from whole slide images (AUC 95%CI > 0.65). This is highly valuable for guiding targeted ADC therapy and combination treatments^[169]. This suggests considerable theoretical feasibility for such predictive applications in the context of ADC therapy, although research in this specific area remains nascent.

Deep learning-based prediction in medical imaging is also of significant importance. A deep learning model developed by Szep et al., based on whole-tumor apparent diffusion coefficient texture analysis, can predict the hormone status of breast cancer^[170]. Although the term “ADC” in this specific study refers to a radiological parameter (Apparent Diffusion Coefficient) and not an ADC, the methodology it presents - using AI to analyze medical imaging for predicting tumor characteristics - is highly relevant for future ADC drug delivery and efficacy prediction. Radiomics methods developed by Chaunzwa et al. can predict the histological type and receptor expression of NSCLC from routine CT images^[171]. Overall, such research holds considerable translational value in the ADC development field. However, these models are currently limited to specific diseases, and generalization to pan-cancer applications remains challenging. For the field of autoimmune diseases, model transferability is even more difficult, highlighting the challenges faced in using non-invasive prediction of receptor expression status in ADC development.

Moreover, translating these technologies into clinical practice also faces significant challenges. Harmon et al. pointed out that the generalizability of deep learning models across images acquired from different institutions and equipment is a key limitation^[172]. Rai and Neri et al. emphasized that the “black box” nature of these models, i.e., their lack of interpretability, limits their clinical acceptance, necessitating explainable AI (XAI) techniques to enhance trust^[173,174].

In conclusion, while deep learning techniques for non-invasive prediction of receptor expression status hold broad prospects, their clinical translation still needs to overcome key challenges such as generalizability, interpretability, and prospective validation. With further technological advancements and in-depth clinical validation, these methods may eventually become important pillars of ADC precision therapy.

Redefining the future of ADCs: AI, disease heterogeneity, and design innovation

Reflecting on the developmental trajectory of ADCs - from the materialization of the “magic bullet” concept to their widespread application in solid tumor therapy and tentative expansion into autoimmune diseases - we have witnessed the rapid evolution of this platform technology. However, with increasing depth in clinical and molecular understanding, a more fundamental reflection has emerged: Are we still employing 20th-century design logic to construct therapeutic systems for 21st-century complex diseases?

The traditional concept of ADCs is based on a linear “target-deliver-kill” logic, which is highly effective under conditions of high target expression, internalizable antigens, and acute treatment objectives. However, when ADCs are extended to chronic diseases characterized by high heterogeneity in target expression and requiring finely tuned effects (such as autoimmune diseases), this approach reveals its limitations.

Future ADCs may need to transcend the singular “antibody + toxin” structural paradigm and be re-envisioned as molecular delivery systems with adaptive regulatory capabilities. AI would no longer be merely a tool for molecular screening but would be embedded within the design logic, assisting in the construction of “intelligent ADCs” capable of real-time sensing and dynamic response: selectively releasing immunomodulatory factors in pro-inflammatory environments; maintaining a low-dose state during the repair phase; or even dynamically adjusting the payload release window based on the spatiotemporal expression of individual biomarkers. This concept not only addresses the discussions in this paper regarding “payload reconciliation” and “target heterogeneity” but also signifies the potential for a shift from structural-level optimization to system-level design.

As previously stated, autoimmune diseases do not always require the “elimination” of target cells but rather the “re-establishment of immune balance”. Therefore, future ADCs should perhaps be viewed more as “regulatory systems”, and their evaluation criteria should shift from toxicity intensity and killing efficiency to the ability to reconstruct immune homeostasis. This implies a fundamental shift in perspective: ADCs are no longer tools of confrontation but rather interveners for balance and repair.

After moving beyond the “structural engineering” phase of ADCs, we need to enter a phase of innovation in design logic. ADCs with true future potential will likely be AI-assisted adaptive systems possessing “dynamic decision-making capabilities”. Operating within this conceptual framework may be essential to overcoming the deep-seated obstacles currently faced in heterogeneous diseases, chronic treatment, and immune balance regulation, thereby advancing toward precision medicine in its truest sense.