Antibody–Drug Conjugates (ADCs) for Personalized Treatment of Solid Tumors: A Review

Chemotherapy with cytotoxic compounds has been the mainstay of systemic cancer treatment for half a century [1, 2]. Over many years, “optimal” combinations of different cytotoxic agents having different cell-killing mechanisms were developed to maximize the antitumor activity [2]. However, these treatments are generally only partially effective in solid tumors, in part because the maximal achievable doses are limited by systemic toxicity, and as a result long-term remissions are rarely seen in patients with metastatic disease. Medicinal chemists and oncologists have long sought ways to increase the delivery of cytotoxic chemicals to cancer cells for increased efficacy, while minimizing exposure of normal healthy tissue [3]. The invention of monoclonal antibodies [4] offered the possibility of exploiting their specific binding properties as a mechanism for the selective delivery of a cytotoxic agent to cancer cells upon chemical conjugation of a cytotoxic effector to a tumor-binding antibody to create an antibody–drug conjugate (ADC). This article is a review based on previously conducted studies and does not involve any new studies of human or animal subjects performed by either of the authors.

All three component parts of an ADC, the antibody, the cytotoxic agent, and the linker that joins them, are critical elements in its design. The antibody moiety should be specific for a cell surface target molecule that is selectively expressed on cancer cells, or overexpressed on cancer cells relative to normal cells [5]. The payload of an ADC must be highly cytotoxic so that it can kill tumor cells at the intracellular concentrations achievable following distribution of the ADC into solid tumor tissue, and because only a limited number of payloads can be linked to an antibody molecule (typically an average of 3–4 payloads per antibody) without severely compromising its biophysical and pharmacokinetic properties [5‐7]. Indeed, the breakthrough in ADC research that eventually led to the creation of approved and marketed products came with the realization that the cytotoxic agents suitable for ADC approaches need to have potency in the picomolar range to be able to be delivered in sufficient quantity to enough cancer cells to effect a therapeutic benefit [5‐7]. The cytotoxic compounds used in approved and marketed ADCs are derivatives of calicheamicin, a class of highly cytotoxic enediyne antibiotics which kill cells by causing DNA double-strand breaks [5, 8], and derivatives of the potent antimitotic microtubule-disrupting agents, dolastatin 10 (auristatins) [6, 9] and maytansine [7, 10, 11], the chemical structures of which are shown in Fig. 1a. The two tubulin-acting agents were evaluated in phase I and phase II clinical trials and showed minimal activity at their recommended phase II doses of 2.0 mg/m² (maytansine) and 0.4 mg/m² (dolastatin 10) when each was given once every 3 weeks [12‐15]. Gastrointestinal and central neurologic toxicities were dose-limiting for maytansine [12, 13], while myelosuppression was dose-limiting for dolastatin 10 [14, 15].

The third vital component of an ADC is the linker that forms a chemical connection between the payload and the antibody. The linker should be sufficiently stable in circulation to allow the payload to remain attached to the antibody while in circulation as it distributes into tissues (including solid tumor tissue), yet should allow efficient release of an active cell-killing agent once the ADC is taken up into the cancer cells [5]. Linkers can be characterized as either cleavable, where a chemical bond (or bonds) between the payload and the attachment site on the antibody (usually an amino acid) can be cleaved intracellularly [16‐18], or as non-cleavable, where the final active metabolite released within the cell includes the payload and all elements of the linker still attached to an amino acid residue of the antibody, typically a lysine or a cysteine residue, following complete proteolytic degradation of the ADC within the lysosome of the cell [19‐21].

The design goal for an ADC is to harness the potent tumor cell-killing action of the payload to an antibody, while retaining the favorable in vivo pharmacokinetic and biodistribution properties of the immunoglobulin, as well as any intrinsic biologic or immunologic activity it may have [5‐7, 21, 22]. Much of the selection of the optimal antibody, the ideal linker–payload chemistry, and the optimal number of payload molecules linked per antibody molecule, are determined empirically, with a focus on maximizing the therapeutic index of the ADC [5, 22]. A summary of the chemistry, biochemistry, and preclinical evaluation leading to the selection of the ADC design that became ado-trastuzumab emtansine (T-DM1), the first ADC to receive full approval for a solid tumor indication (vide infra), has been reviewed elsewhere [7], and serves as an exemplar for the work involved in ADC optimization prior to declaring a candidate for clinical evaluation.

The first ADC therapeutics to reach the market were developed in hematologic malignancies, where target antigens are generally well-characterized, lineage-specific cell surface molecules that are uniformly expressed and thought to be more accessible to antibodies circulating in plasma, relative to those present on solid tumors [22]. Such antigens are usually highly restricted in their distribution, meaning that healthy non-hematopoietic tissue and pluripotent hematopoietic stem cells are not targeted by the ADC. The first ADC to receive marketing approval from the US Food and Drug Administration (FDA) was gemtuzumab ozogamicin (Mylotarg^®), an ADC created by conjugation of calicheamicin to an anti-CD33 antibody via an acid–labile linkage [23, 24]. CD33 is a marker of normal and malignant cells of the myeloid lineage. Based on a single arm phase II trial, the ADC received accelerated approval in 2000 as a single agent (dosing 9 mg/m² on days 1 and 15) for treating patients with acute myeloid leukemia (AML) in first relapse, and who are at least 60 years old. The overall response rate (ORR) was 26% in these patients, including a subset having incomplete platelet recovery [23, 24], while side effects included delayed hematopoietic recovery and hepatic veno-occlusive disease [24, 25]. Approval was conditioned on execution of randomized clinical trials to confirm patient benefit [24]. In 2010, the ADC was withdrawn from the US market (although it is still marketed in Japan), following an unsuccessful confirmatory phase III trial which compared the effect of adding a single dose of 6 mg/m² to standard remission induction therapy in patients younger than 60 years [26]. Recently, the results of several other randomized studies, adding gemtuzumab ozogamicin to various induction regimens, utilizing either a single 3 mg/m² dose or fractionated dosing (3 mg/m² on days 1, 4 and 7), to reduce the incidence of hepatic veno-occlusive disease, have suggested improved overall survival (OS) in AML patients whose disease has favorable and intermediate cytogenetic characteristics [27], and have renewed interest in this ADC, and in CD33 as a target [28].

The second ADC to receive marketing approval was brentuximab vedotin (Adcetris^®), made by conjugation of the potent auristatin tubulin agent, MMAE, to an anti-CD30 antibody via a cleavable valine-citrulline (vc) dipeptide linker [6]. The CD30 antigen, a marker for activated lymphocytes, is highly expressed on the Reed–Sternberg cells of Hodgkin lymphoma (HL) as well as on the malignant cells of anaplastic large cell lymphoma (ALCL). Early phase clinical trials of the “naked” anti-CD30 antibody showed no responses in HL, and, in ALCL patients, only 2 of 41 (5%) had a complete response (CR) [29]. However, arming the antibody with vcMMAE created a highly active compound that received accelerated approval from FDA in August, 2011, for treatment of patients with relapsed HL or systemic ALCL [6]. In the pivotal single arm phase II trials, brentuximab vedotin showed a 75% ORR in patients with relapsed or refractory HL, with a median duration of response of 20.5 months [30], and an 86% ORR (57% CR) in patients with systemic ALCL [31]. The principal dose-limiting toxicity (DLT) was neutropenia, while the main toxicity upon repeated administration was neuropathy [30, 31]. Several phase III trials are in progress to confirm clinical benefit of brentuximab vedotin in randomized studies, both as a single agent and in combination with approved agents (http://​www.​clinicaltrials.​gov) [6].

A second ADC using the calicheamicin payload, inotuzumab ozogamicin, which targets the B cell antigen CD22 has completed a phase III trial (INO-VATE ALL) in relapsed or refractory acute lymphoblastic leukemia [32‐34]. The results show that treatment with the ADC produced a higher rate of CR (80.7%) compared to chemotherapy (29.4%), with 23% of the patients alive at 2 years in the ADC-treatment group compared to 10% in the chemotherapy comparator group [34]. Like gemtuzumab ozogamicin, the most concerning non-hematologic toxicities in patients treated with inotuzumab ozogamicin were adverse events in liver, especially veno-occlusive disease [34].

The first ADC to receive marketing approval for treatment of a prevalent solid tumor was ado-trastuzumab emtansine (Kadcyla^®; T-DM1), made by conjugation of the sulfhydryl group of the maytansinoid DM1 to lysine amino groups of the anti-human epidermal growth factor receptor 2 (HER2) antibody, trastuzumab, via reaction with the bifunctional non-cleavable linker, succinimidyl-4-(N-maleimidomethyl)cyclohexane-1-carboxylate (SMCC) [7, 35, 36]. Figure 1b shows a molecule of T-DM1 (representative of a typical ADC), illustrating the chemical structure of the linker–payload moiety that is attached to lysine residues of the antibody (left panel), and a representation of a molecular model of one antibody molecule (molecular mass 150 kDa) conjugated to four molecules of the maytansinoid DM1 (right panel; molecular weight of linker-payload ~1 kDa). Since there are about 90 lysine amino groups in trastuzumab, and the conjugation reaction with SMCC is carefully controlled to yield a conjugate with an average of 3.5 DM1 molecules per antibody molecule [7], the ADC compound is heterogeneous with respect to loading of the payload and its distribution across the potential sites of reaction on the antibody [37, 38]. Mass spectroscopy shows that about 70–80% of the antibody molecules are conjugated to between 2 and 5 molecules of DM1, while only about 3% of the antibody remain unconjugated [37, 39, 40]. Peptide mapping methodology shows that up to 70 individual lysine residues are partially modified by linker–payload to an average level of about 4% with a range from 25% to <1% [40]. While the heterogeneity of an ADC presents a challenge for process and analytical development and process control, the challenge can be met by utilizing appropriate development strategies and modern techniques of protein and chemical analysis [7, 37‐40].

Phase I and phase II clinical trials in patients with HER2-positive metastatic breast cancer (mBC) established a recommended phase II dose/schedule of 3.6 mg/kg given by intravenous infusion once every 21 days [41‐44]. The principal DLT was reversible thrombocytopenia [41, 45], with reversible low-grade increases in hepatic transaminases also observed [41]. Pharmacokinetics were non-linear in the dose range 0.3–4.8 mg/kg as anticipated from preclinical studies and prior clinical experience with trastuzumab, given the normal tissue expression of HER2 [46]. The half-life was approximately 4 days at the 3.6 mg/kg dose level [47]. Plasma levels of free payload were very low [41, 47]. An early signal of antitumor activity (5 confirmed partial responses from 24 enrolled patients; ORR 20.8%) in the phase I study was confirmed in two phase II trials in patients (~110 patients each) who were heavily pre-treated for their HER2-positive mBC, with ORRs of 25.9% [43], and 34.5% [44].

These early clinical studies led to a pivotal phase III trial (“EMILIA”) where patients with mBC who had progressed following treatment with a taxane plus trastuzumab were randomized to receive T-DM1 or the approved second-line treatment of lapatinib (a tyrosine kinase inhibitor which also targets HER2) plus capecitabine [48]. The ORR (43.6% vs. 30.8%; p < 0.001), progression-free survival (PFS; 9.6 vs. 6.4 months; hazard ratio 0.65; p < 0.001), and OS (30.6 vs. 25.1 months; hazard ratio 0.68; p < 0.001), all significantly favored the T-DM1 arm versus the comparator arm of the study [48]. Furthermore, the incidence of adverse events of grade ≥3 was less in the T-DM1 arm (40.8%, with thrombocytopenia and transaminitis most common, consistent with the phase I/II experience) relative to the comparator arm (57.0%, with diarrhea, hand–foot syndrome and vomiting most common). The rate of cardiac adverse events, a concern with HER2-targeted therapy, was low (<2%) in both arms. The “EMILIA” trial established the safety and effectiveness of T-DM1, and, in February, 2013, the FDA granted full approval as a therapy for treating patients with HER2-positive late-stage mBC previously treated with a taxane and trastuzumab. To date, T-DM1 is the only ADC to have received full approval from FDA based on a randomized phase III study in any indication.

A phase III trial (“TH3RESA”) in patients with progressive HER2-positive advanced mBC who had received at least two HER2-directed regimens in the advanced setting, including trastuzumab and lapatinib, and previous taxane therapy in any setting, also favored T-DM1 over the comparator arm (physician’s choice), with improved PFS (6.2 vs. 3.3 months; hazard ratio 0.528; p < 0.001) [49], and median OS of 22.7 versus 15.8 months for the comparator arm (p = 0.0007) [50]. Of note, patients enrolled in TH3RESA had had a median of 4 prior regimens for treating their advanced disease, with 28–30% of patients receiving >5 such prior treatments [49]. Again, improved efficacy was achieved together with a reduced incidence of grade ≥3 adverse events (32% for T-DM1 vs. 43% for the comparator) [49, 50].

A phase III trial of T-DM1 (with and without pertuzumab) versus trastuzumab plus taxane in the first-line treatment of HER2-positive mBC (“MARIANNE”) showed that T-DM1 met the pre-defined non-inferiority criteria, with similar ORR (60–70%) and PFS (~14 months) for all treatment arms [51]. Neither experimental arm showed superior PFS to the trastuzumab plus taxane arm, and addition of pertuzumab to T-DM1 did not improve PFS despite preclinical data that showed synergistic activity for this combination [52]. The duration of response for patients responding to treatment with T-DM1 was ~21 months, while that of patients responding to treatment with trastuzumab plus taxane was 12.5 months. T-DM1 was better tolerated, with patients having improved (prolonged) health-related quality of life [51].

The efficacy results of the MARIANNE trial were disappointing in view of the initial promise of T-DM1 in early phase I and phase II trials [53, 54]. However, the median PFS seen with T-DM1 treatment in MARIANNE (14.1 months; n = 367 patients) was similar to that observed in the prior small phase II trial (14.2 months; n = 67 patients). A major difference between the phase II and phase III studies is in the median PFS of the control trastuzumab plus taxane arms, reported to be 9.2 months (n = 70 patients) and 13.7 months (n = 365 patients), respectively [51, 53]. As noted by Perez et al. [51], the latter value is consistent with results of other phase III trials, for example, the CLEOPATRA study of pertuzumab plus trastuzumab plus docetaxel (PFS of comparator arm = 12.4 months, n = 406 patients) [55]. In the small phase II study [53], more patients in the control arm were first diagnosed at an earlier stage of disease and had received prior (neo)adjuvant therapy with a taxane (40.0%) versus the T-DM1 arm (32.8%), while in the control arm of MARIANNE, 32.9% of patients had received (neo)adjuvant taxane therapy [51, 56]. One might speculate that more cases of prior (neo)adjuvant therapy in the control arm of the small phase II study [53, 56], might account for its relatively poor PFS (9.2 months) compared with larger studies (12.4 and 13.7 months in phase III studies cited above).

What can we learn from the preclinical and clinical development of T-DM1 with respect to the relationship between target expression on the cancer cells of a solid tumor, and clinical benefit with the ADC? With the caveat that any such learning may be unique to the biology of the HER2 antigen in mBC, there are some observations that may apply generally.

In a phase II trial, mBC patients (n = 112) were enrolled on the basis of having had prior treatment with trastuzumab, from which it was inferred that they were HER2-positive [43]. However, the HER2 status of their disease was retrospectively reassessed on archival primary tumor specimens at a central laboratory (specimens available in 95 cases). While the ORR for the entire population was 25.9%, it was higher (33.8%) in patients confirmed as having HER2-overexpressing mBC (n = 24), and only 4.8% in patients whose disease was reassessed as HER2-normal (n = 21). There were similar trends in a second phase II study in patients (n = 110) who must have had at least two prior HER2-directed regimens [44]; the ORR was 34.5% in the overall population, while in those patients whose HER2 status was confirmed at a central laboratory (in 80 of 95 available specimens), the ORR was 41.3%. The implication of these observations is that, if the phase I trial (ORR 20.8%) had been conducted in all comers with mBC, instead of with patients likely to have HER2-overexpression, the ORR could well have been zero because only about 20% of all mBC patients overexpress HER2, while the majority of cases express normal levels of HER2 (“negative” in the jargon of the test). Thus, patient selection, albeit indirect, may have been a key factor in engendering enthusiasm for continued development of the molecule.

Even within the HER2-positive population, there is likely a range of HER2 expression levels above the minimum level that saturates the read-out of the approved immunohistochemistry test. In confirmed HER2-positive patients, whose levels of HER2 mRNA were assessed by quantitative reverse transcriptase/polymerase chain reaction (qRT-PCR) methodology in a phase II study, those with ≥ median values had an ORR of 36% with a median PFS not reached, while those with < median HER2 mRNA levels had an ORR of 28% with a median PFS of 4.2 months [43]. While the numbers are small, similar trends were observed in other phase II trials [44, 54], including first-line treatment settings [53, 56].

Collectively, these observations in clinical studies with T-DM1 are consistent with a simple hypothesis for driving success for ADCs in solid tumors; that is, that the amount of payload delivered to a cancer cell in a tumor is a function of how much ADC can be given to the patient (i.e., the dose and schedule, within bounds dictated by tolerability), and how much ADC the cell can receive (antigen density). The higher the cell surface density of the target antigen, the more ADC can be taken up and metabolized by the cell, to release the active cytotoxic agent. Furthermore, in the case of HER2, it has been amply demonstrated preclinically that the higher the antigen density on a tumor xenograft, the greater the amount of trastuzumab that is retained within the tumor mass relative to a non-binding control antibody [57]. Thus, for an ADC, personalization of therapy can mean selecting only those patients for treatment whose cancers express the target antigen above a threshold level of antigen per cell necessary for anti-tumor activity. However, this criterion is by no means sufficient to predict anti-tumor activity, and further work needs to be done to establish more biomarkers for response to ADC therapeutics beyond simply the presence of the target antigen, for example, markers for sensitivity to the payload, or for aspects of target biology that may affect internalization and intracellular trafficking (may influence rate and extent of payload release from the ADC) [19, 20, 36].

There are >60 ADCs currently in clinical evaluation, the majority of which (72%) utilize a version of the two classes of potent tubulin-binding microtubule-disrupting agents found in the two currently approved ADCs described above [5, 58, 59]. Most ADCs in development are targeting solid tumors and are in early clinical development, a testimony to the relatively recent surge in enthusiasm for the ADC approach following the approvals of brentuximab vedotin and ado-trastuzumab emtansine. Such enthusiasm must be tempered by the discontinuation of development of several compounds, mainly for reasons of insufficient activity at the maximum doses that can be tolerated upon repeat administration [58, 59]. Some of the reasons for the relatively narrow therapeutic window ascertained from the clinical experience in developing ADCs have been discussed elsewhere [58, 59]. However, several ADCs have advanced into pivotal studies in both solid tumor indications and in hematologic cancers (listed in Table 1). The remainder of this review will focus on the five ADCs currently in pivotal studies for treating solid tumors.

The common theme that emerges from the clinical experience obtained during development of the one marketed ADC and the five more now in pivotal clinical trials for the treatment of solid tumors is the importance of patient selection for the level of target expression on tumor cells. The development of T-DM1 was greatly aided by the fact that trastuzumab itself was already a marketed therapeutic agent for which there was already a marketed diagnostic test to define those breast cancer patients (~20% of the total) whose cancer overexpressed HER2 and who were thus most likely to benefit from treatment with T-DM1 [7]. In the case of the five ADCs in pivotal trials for mesothelioma, FRα-overexpressing platinum-resistant ovarian cancer, gpNMB-positive TNBC, EGFR-amplified glioblastoma and DLL3-expressing SCLC, the relationship between anti-tumor activity and the level of target expression in the tumor required investigation during clinical evaluation of phase I and phase II trials. All five ADCs utilize a diagnostic test (generally an immunohistochemistry test) in their respective pivotal clinical trials in order to select the target-positive patient population most likely to benefit from treatment with the ADC, by virtue of the expression of the antibody target above a pre-defined threshold level identified during early clinical development.

ADCs that have shown promising activity in solid tumors in early phase clinical trials, but that have not yet advanced into pivotal trials, also show most benefit in patients whose tumors express the highest levels of target antigen; for example, lifatuzumab vedotin which targets NaPi2b [86], and SAR566658 which targets the CA6 antigen of mucin-1 [87]. It is likely that a prototype diagnostic test will be a necessary accompaniment to virtually all ADCs in development for targeting solid tumors. In the future, it may be that non-clinical studies in patient-derived tumor xenograft models will better identify the threshold levels of target expression for favorable anti-tumor activity, and accelerate the development of appropriate calibration of a diagnostic test to measure such antigen levels, prior to initiation of clinical trials [88].

While the development pathway for three of the five ADCs in pivotal studies for treating solid tumors (Table 1) is to evaluate them as monotherapy versus a chemotherapy comparator arm—similar to the development of T-DM1 [48]—all three compounds are also being studied in combination with a variety of chemotherapeutic and biologic agents, as appropriate for the particular disease (http://​www.​clinicaltrials.​gov). The relatively benign side-effect profiles for many ADCs (relative to cytotoxic chemotherapy) suggest that they may be well suited to combine with other agents with the goal of further improvement in treatment outcomes for cancer patients, ultimately in first-line treatment settings. For example, mirvetuximab soravtansine is being evaluated in combination with either carboplatin, pegylated liposomal doxorubicin, or bevacizumab, in a multi-arm trial called FORWARD II (NCT02606305) [89].

In contrast, the initial strategy for the development of ABT-414 is in combination with chemo-radiation in the first-line treatment of glioblastoma, a disease for which the current standard of care has a poor outcome [90]. The development strategy for rovalpituzumab tesirine, on the other hand, is to seek an initial accelerated approval based on response rate in a single arm monotherapy trial in third-line or later treatment for SCLC, a setting for which there are no good treatment options available [91].

The recent findings of Zippelius et al. [92‐94] suggest that ADCs, especially those made with the potent microtubule-disrupting agents as the payload, may be combined with immune checkpoint inhibitors such as anti-PD1 antibodies (pembrolizumab, nivolumab) or anti-PD-L1 antibodies (atezolizumab) for enhanced and sustained anti-tumor effect. Not only do such ADCs induce immunogenic cell death, but the ADC-mediated tumor accumulation of a potent microtubule agent appears to activate intra-tumor dendritic cells, inducing uptake of antigens and the migration of antigen-loaded dendritic cells to lymph nodes where they can trigger activation of T-cells which may be directed towards tumor-associated antigens [92‐95]. Pre-clinically, Müller et al. have demonstrated that T-DM1 renders HER2-positive breast cancer highly susceptible to checkpoint blockade [94]. This research has stimulated interest in clinical evaluation of this potential, with ongoing clinical trials of T-DM1 in combination with pembrolizumab (anti-PD-1) and with atezolizumab (anti-PD-L1) (http://​www.​clinicaltrials.​gov), the recent addition of a combination of mirvetuximab soravtansine with pembrolizumab as an arm of the FORWARD II clinical study of this maytansinoid-ADC in combination regimens [89], and with the addition of an anti-PD1-containing arm to a combination trial of GV in treatment of melanoma (NCT02302339).

In 2013, ado-tratuzumab emtansine (T-DM1) became the first ADC approved for the treatment of a prevalent solid tumor (HER2-positive breast cancer). This success [7], after decades of disappointment in developing immunoconjugates as therapeutic agents [96], has sparked a flurry of research and development with over 80 ADCs being taken into clinical evaluation. There are now 5 more ADCs in pivotal clinical development for treating solid tumors (Table 1). The level of target expression is a key parameter in predicting the likelihood of patient benefit for these ADCs, and the co-development of a diagnostic test to assess the level of cell target antigen is important for selecting the patient population appropriate for receiving treatment. There is clearly much yet to learn about the optimal application of ADCs in the treatment of cancers, especially in establishing the best combination modalities for treating different cancers, including the role of the emerging immune–oncology therapies. Nonetheless, the opportunity for improved cancer therapy incorporating ADCs into treatment paradigms offer exciting possibilities for better outcomes for cancer patients.