Several antibody–drug conjugates (ADC) showing strong clinical responses in solid tumors target high expression antigens (HER2, TROP2, Nectin-4, and folate receptor alpha/FRα). Highly expressed tumor antigens often have significant low-level expression in normal tissues, resulting in the potential for target-mediated drug disposition (TMDD) and increased clearance. However, ADCs often do not cross-react with normal tissue in animal models used to test efficacy (typically mice), and the impact of ADC binding to normal tissue antigens on tumor response remains unclear. An antibody that cross-reacts with human and murine FRα was generated and tested in an animal model where the antibody/ADC bind both human tumor FRα and mouse FRα in normal tissue. Previous work has demonstrated that a “carrier” dose of unconjugated antibody can improve the tumor penetration of ADCs with high expression target-antigens. A carrier dose was employed to study the impact on cross-reactive ADC clearance, distribution, and efficacy. Co-administration of unconjugated anti-FRα antibody with the ADC-improved efficacy, even in low expression models where co-administration normally lowers efficacy. By reducing target-antigen–mediated clearance in normal tissue, the co-administered antibody increased systemic exposure, improved tumor tissue penetration, reduced target-antigen–mediated uptake in normal tissue, and increased ADC efficacy. However, payload potency and tumor antigen saturation are also critical to efficacy, as shown with reduced efficacy using too high of a carrier dose. The judicious use of higher antibody doses, either through lower DAR or carrier doses, can improve the therapeutic window by increasing efficacy while lowering target-mediated toxicity in normal tissue.
Antibody–drug conjugates (ADC) have the potential to combine the high tumor selectivity of antibodies with the extraordinary potency of small-molecule drugs (1). With the recent approval of enfortumab vedotin (anti–nectin-4), trastuzumab deruxtecan (anti-HER2), and sacituzumab govitecan (anti–Trop-2) for solid tumors, there are currently nine FDA-approved ADCs, several more in late-stage clinical trials, and a large number in early clinical testing and preclinical development. Despite their promise, this class of anti-cancer agents has been hampered by a relatively narrow therapeutic index (2). Many approaches have been taken to increase this window, but several of these have resulted in shifts in the therapeutic window (i.e., increased efficacy with increased toxicity) rather than a true improvement/expansion of the window. For example, higher potency payloads can lower the efficacious dose, but they often cause a greater degree of toxicity resulting in a lower MTD. Likewise, the selection of “cleaner” targets with less normal-tissue expression can shift the dose-limiting toxicity from a target-mediated effect to non–target-mediated toxicity (3) without a major improvement in the MTD. In fact, nearly all ADCs have a toxicity profile determined by the payload at the MTD in first-in-human studies (4, 5). A key challenge is delivering this maximum dose of payload in a manner that reaches and kills as many cancer cells as possible while avoiding concentrations that are toxic in healthy tissues. More stable linkers and/or site-specific conjugation can often reduce non–target-antigen–mediated payload release and associated hematological toxicity (e.g., from deconjugation). However, the longer circulation of intact ADC can potentially increase toxicity from non–target-antigen–mediated internalization [e.g., micropinocytosis; refs. 6, 7, resulting in side effects, such as ocular toxicity (refs. 8, 9)] and/or other toxicities from a target-antigen–mediated mechanism (10). Higher potency payloads can increase in vitro cell killing, but this can be offset by increased toxicity necessitating lower doses. Newer approaches to expand the therapeutic window are greatly needed, particularly those that can increase cancer cell delivery/killing and reduce normal tissue uptake in the dose-limiting organ at a fixed payload dose.
Recent ADCs with clinical responsiveness have targeted antigens with significant normal tissue expression. Unfortunately, most antibodies have been developed in murine systems, and thus do not cross-react with mouse antigen. Although the role of cross reactivity and expression can be tested in cynomolgus monkeys for toxicity, the effect of normal tissue expression on efficacy cannot be gauged in these systems. To expand the therapeutic window and achieve clinical success, development of ADCs must account for the impact of normal tissue expression on both toxicity and efficacy, particularly for the type of high expression targets that are showing significant clinical relevance.
To examine the role of normal tissue expression, a cross-reactive antibody was generated in rabbit that binds both mouse folate receptor alpha (FRα) in normal tissues and the human FRα in tumor xenografts. The lead rabbit monoclonal antibody was chimerized into an IgG2a murine backbone and designated rmFR1–12. In this work, we look at the impact of a “carrier dose,” the co-administration of unconjugated rmFR1–12 with the rmFR1–12-ADC, on systemic clearance, local antibody distribution in the tumor, uptake of the ADC by normal tissues, and efficacy using this cross-reactive antibody. The effects of a carrier dose also depend on factors such as target expression and payload potency, so both high (4 million FRα/cell) and low (40,000 FRα/cell) antigen expression models were used along with a high potency DM4 payload (microtubule inhibitor) and very high potency DGN549 payload (DNA alkylator). Overall, cross-reactive ADC shows decreased efficacy due to increased plasma clearance as compared with a non–cross-reactive ADC, which can be improved by a carrier dose of unconjugated antibody by blocking target-mediated uptake in normal tissue.
Materials and Methods
Krogh cylinder model
Computational simulations to mechanistically predict intratumoral distribution of the ADC with and without a carrier dose was performed using a previously published Krogh Cylinder Model (11). Differential equations and ADC kinetic parameters have been detailed in Supplementary Information.
Rabbits were immunized with recombinant 300–19/Folr1 cells expressing murine FRα. Hybridomas were screened by flow cytometry with mouse and human FRα-positive and -negative cell lines for production of anti-FRα antibodies. The resulting lead antibody was chimerized to a mouse IgG2a (muIgG2a) backbone and designated rmFR1–12. Anti-human FRα ADCs, M-SPDB-DM4 and M-s-SPDB-DM4 have been characterized previously (12). DGN549 ADC of anti-human FRα antibody M9346A was prepared by the method previously described (13).
M109 (murine lung cancer) cells were cultured from ImmunoGen's cell bank for binding affinity studies without further characterization. KB (ATCC CCL-17–containing HeLa chromosomal markers, human carcinoma/papilloma cell line; refs. 12, 14, 15) and OV90 (16) human ovarian cancer cells were obtained from the ATCC. All lines were characterized by the vendor using routine DNA profiling; no further authentication was performed by the authors. M109 cells were cultured in RPMI-1640 supplemented with 10% FBS, KB cells in Eagle's MEM with 10%, and OV90 cells in a 1:1 mixture of MCDB-105 and Medium-199 with 15%. Cells were expanded and maintained as recommended in a humidified 37 °C, 5% CO2 incubator.
Binding of rmFR1–12 and its DM4 and DGN549 conjugates to M109 and KB cells was evaluated flow-cytometrically by indirect immunofluorescence. Briefly, 2 × 104 cells per well in a 96-well plate were incubated for 2 hours at 4°C with an anti-FRα antibody diluted to various concentrations in assay medium [RPMI-1640 supplemented with 2% (w/v) BSA (Sigma)]. The cells were then washed with cold assay medium, stained with fluorescein isothiocyanate–labeled goat anti-murine or anti-human immunoglobulin G (IgG) antibody for 40 minutes at 4°C in the dark, washed with cold assay medium, fixed in a solution of PBS, pH 7.4 (PBS) containing 1% formaldehyde, and analyzed using a FACSCalibur flow cytometer (BD Biosciences). Data were analyzed via GraphPad Prism sigmoidal dose–response (variable slope) function. The concentration of the antibody achieving half-maximal binding was taken as its apparent affinity (Kd).
Dilutions of conjugates in the appropriate culture medium were added to wells of 96-well flat-bottomed plates containing 1 × 103 cells per well. The plates were incubated at 37°C, 5% CO2 for 5 days. Cell viability was determined by the WST-8 assay (Donjindo Molecular Technologies, Inc.) in accordance with the manufacturer's protocol, and IC50 values were generated using a sigmoidal dose–response (variable slope) nonlinear regression curve fit (GraphPad Software Inc.).
In vivo studies
All procedures were carried out in accordance with the “Guide for the Care and Use of Laboratory Animals” (National Research Council). Eight to 12-week-old female CB-17 SCID mice were obtained from Charles River Laboratories and quarantined for 7 days before study initiation. Mice were inoculated subcutaneously with cells (1 × 107 cells per mouse) resuspended in serum-free culture media (KB cells) or 50% Matrigel in serum-free culture media (OV90 cells). When tumor volume reached approximately 100 mm3, mice were randomized (N = 6 per group) and received a single intravenous bolus injection of PBS (control), DM4 conjugate or DGN549 conjugate at the indicated doses (based on antibody concentration). For biodistribution studies, radiolabeled DM4 conjugate was administered at the indicated doses. Unconjugated mFR1–12 was used in co-administration studies at the indicated doses. Mice received a single intravenous bolus injection. Tumor dimensions were measured two or three times per week and volume was calculated as ½ (L × W × H), where L is the length, W is the width, and H is the height of the tumor. A partial tumor regression was defined as a reduction in tumor volume by 50% or greater. A complete tumor regression was scored when no palpable tumor could be detected. Termination criteria was set at tumor volumes exceeding 1,500 mm3 in volume, >20% body weight loss (BWL), or clinical observations of the mice being in distress. Statistical differences in tumors sizes was assessed using Prism following the nadir during regrowth. Antigen density in vitro was assessed via flow cytometry with indirect fluorescence using a FITC-labeled secondary antibody with phycoerythrin-labeled anti-FRα antibody using QuantiBRITE PE kit (Becton Dickinson), and in KB and OV90 tumor models was assessed by ex vivo immunohistochemical staining of formalin-fixed paraffin-embedded (FFPE) tumor tissues.
DM4 ADCs cellular processing
DM4 catabolites within normal tissue and FRα-positive cancer cells were quantified using tritiated DM4 ADC (17), wherein [3H] was incorporated in the methoxy group at the C20 position of the payload. Briefly, protein-free radioactivity (processed ADC) and protein-associated radioactivity (unprocessed ADC) were separated by acetone extraction, and the associated radioactivity measured via liquid scintillation counting, and the data were used to calculate %ID/g, percentage of processed ADC, and pmol processed ADC.
Mice were inoculated subcutaneously with 1 × 107 KB cells per mouse resuspended in serum-free culture media. Tumors grew to approximately 200 mm3 before the mice were injected intravenously with saline or 2.5 mg/kg Alexa488-labeled rmFR1–12 antibody alone or concurrently with either 10 or 25 mg/kg unlabeled rmFR1–12 antibody. Twenty-four and 48 hours after treatment, mice were injected with Tomato Lectin DyLight594 (Vector Laboratories) for visualization of active blood vessels. After 5 minutes, mice were euthanized, tumors were collected and fixed in 10% neutral-buffered formalin for paraffin embedding.
For staining, all chemicals were procured from Sigma-Aldrich, unless noted otherwise. FFPE tumors were cut for histology into 5-μm sections using a microstat. Tumor sections were deparaffinized and antigen-retrieved before imaging (Supplementary Methods). Immunofluorescence studies were performed by blocking tumor sections with 0.5%–2% BSA for 15 minutes, following by staining with 6 μg/mL AlexaFluor647-labeled chicken anti-mouse antibody (Thermo Fisher) and 15 μg/mL AlexaFluor555-labeled rat anti-mouse CD31 antibody in 0.5% BSA for 40 minutes at room temperature in the dark. Slides were washed for 5 minutes in PBS, followed by staining with Hoechst 33342 for 1 minute, then washed again for 3 minutes in PBS. Stained slides were imaged using a ×20 objective on an upright Olympus FV1200 confocal microscope in the 405 (Hoechst 33342), 488 (ADC), 542 (CD31), and 635 (chicken anti-mouse antibody) lasers. Images were analyzed using ImageJ.
Matching cellular delivery with payload potency results in highest efficacy
To first demonstrate the impact of receptor expression and payload potency on cell killing in vitro, a non–cross-reactive antibody to human FRα, the M antibody (M9346A), and cross-reactive antibody binding to both human and mouse antigen were conjugated to DM4 (“low” potency) or DGN549 (“high” potency) payloads and tested in a “high” expression and “low” FRα expression model (Supplementary Fig. S1; refs. 12, 18). Quotation marks emphasize that the potency is context dependent, denoting a relative comparison in this case. Likewise, expression levels require absolute quantification because target receptors per cell vary over many orders of magnitude (e.g., 105 receptors/cell is considered low for HER2 but high for HER3). KB cells express approximately 4 million folate receptors per cell in vitro and form xenografts with “high” homogeneous target expression, whereas OV90 cells express approximately 40,000 folate receptors per cell with “low” heterogeneous membrane expression. Low and high levels in this case are relative to the range of FRα expression observed in various xenograft models and patients with ovarian carcinoma.
The antibody M9436A (M antibody) does not bind mouse FRα and only targets the xenograft human cancer cells. In the high expression KB model, the more potent DGN549 payload, which must be administered at low antibody doses (0.29 mg/kg M-DGN549) due to the high potency/low tolerability, is likely not able to penetrate the tumor tissue far, reaching few cells (Fig. 1A), and resulting in modest efficacy (Fig. 1B). In contrast, cures were achieved in all animals treated with approximately 2.5 mg/kg of M-s-SPDB-DM4, because the higher dose can penetrate deeper in the tissue while still delivering lethal amount of payload to cells. In low expression models, ADC tumor penetration (regardless of payload used) is often fairly efficient. However, low receptor expression causes fewer total ADC molecules to internalize into tumor cells, limiting cellular payload accumulation. Treatment of OV90 with M-s-SPDB-DM4 likely results in subtherapeutic concentrations of the “low potency” DM4 that are insufficient to kill tumor cells. However, treatment with the high potency M-DGN549 ADC could result in complete regression of tumors from delivery of lethal concentration of DGN549 (Fig. 1C). In the low expression OV90 model, the ADC with the lower potency DM4 payload only achieved modest efficacy given at the same dose that achieved complete regression in the high expression KB tumors. The ADC with the more potent DGN549 showed limited activity at the 0.29 mg/kg dose against the high-expressing xenograft, whereas a 2-fold lower dose led to complete responses in the low-expressing OV90 model (Fig. 1B and D).
A carrier dose can increase or decrease efficacy depending on target expression
Unconjugated antibody that is co-administered with the ADC will block some of the perivascular-binding sites, allowing deeper penetration into the tissue (at the cost of fewer ADC molecules delivered per cell). As the antibody dose is increased in the high expression KB model (Fig. 1B), the ADC is “diluted” and spreads out deeper into the tumor tissue. On the basis of the poor tissue penetration of M-DGN549 in the high expression model, a carrier dose is expected to improve tissue penetration and increase efficacy. According to predictive simulations (11), the addition of 2.5 mg/kg of unconjugated antibody to 0.29 mg/kg of M-DGN549 results in similar overall payload uptake in tumor as 0.29 mg/kg of M-DGN549-ADC but distributed more homogeneously throughout the tumor (Fig. 2B). This translates to improved efficacy in vivo as expected (Fig. 2C). A super-saturating dose of 10 mg/kg unconjugated antibody, however, reduces total payload uptake in the tumor and therefore efficacy. Once the tumor is saturated with antibody, competition of any additional unconjugated antibody with the ADC is antagonizing and can reduce efficacy. This phenomenon is apparent in the low expression model OV90 (Fig. 2D), where 0.29 mg/kg M-DGN549 is predicted to show efficient tissue penetration, consistent with previous observations (16), and high payload uptake (Fig. 2E), thereby demonstrating maximum efficacy (Fig. 2F). Both the 2.5 and 10 mg/kg carrier doses super-saturate the tumor and antagonize payload uptake and efficacy. Conceptually, efficacy is driven by the cellular payload delivery versus payload potency; the carrier dose changes the cellular payload delivery, which can be beneficial or detrimental depending on how well-matched the original ADC dose is to receptor expression and payload potency.
Cross-reactivity of antibody/ADC increases plasma clearance and reduces efficacy
To test the impact of normal tissue expression and binding, a rabbit anti-mouse FRα antibody was generated. The variable domain of the rabbit antibody was then grafted onto a murine framework to generate a chimera, denoted rmFR1–12 (Supplementary Fig. S2). Fortuitously, the chimera binds human and mouse folate receptor with similar affinity and does not compete with folate at even supraphysiological concentrations (Supplementary Fig. S3). When injected into CD-1 mice, the plasma clearance of rmFR1–12-s-SPDB-DM4 significantly increased resulting in a 3-fold reduction in AUC (Fig. 3A), presumably due to significant binding in normal tissue leading to target-mediated drug disposition (TMDD; ref. 19). To determine the impact of faster clearance on tumor uptake, the DM4 payload component was radiolabeled, and SCID mice with KB xenografts were administered 5 mg/kg of rmFR1–12-s-SPDB-DM4 or M-s-SPDB-DM4. Tumor uptake was significantly lower for the cross-reactive rmFR1–12-s-SPDB-DM4 compared with the non–cross-reactive M-s-SPDB-DM4, indicating the faster plasma clearance reduced tumor uptake (Fig. 3B). Next, tumor growth was measured to see whether the increased plasma clearance and lower tumor uptake had an impact on efficacy. At a 2.5 mg/kg dose, which resulted in complete cures for the non–cross-reactive M-s-SPDB-DM4 (Fig. 3C), efficacy of rmFR1–12-s-SPDB-DM4 was significantly reduced, presumably due to binding to normal tissues that increased clearance and lowered tumor uptake (Fig. 3D).
A carrier dose can improve tumor exposure and efficacy in low and high expression models
Because FRα is expressed in multiple tissues, even low-level expression can result in significant target-mediated uptake and clearance due to the large normal tissue “sink.” These effects are typically dose-dependent, and an increase in antibody dose can saturate the target-mediated clearance, resulting in linear kinetics. Although ADCs are usually administered at doses that are limited by payload toxicity, unconjugated antibodies are often well-tolerated even at very high doses. Increasing doses of the cross-reactive rmFR1–12 antibody resulted in a higher Cmax and slower clearance as expected (Fig. 4A), and increasing doses of the rmFR1–12 ADC showed both dose-dependent pharmacokinetics and efficacy response (Supplementary Fig. S4). By combining carrier doses with a constant 2.5 mg/kg dose of ADC, the Cmax of the ADC does not change. However, the unconjugated antibody blocks normal tissue receptors, slowing clearance of the ADC and increasing tumor exposure (Fig. 4B).
To determine whether this increased exposure had an impact on efficacy, increasing doses of the unconjugated rmFR1–12 antibody were co-administered with rmFR1–12-DGN549 and tumor growth was monitored. Notably, the unconjugated antibody has no detectable efficacy in these tumor models by itself (Supplementary Fig. S4). The addition of 5 and 10 mg/kg of the cross-reactive rmFR1–12 antibody to 0.59 mg/kg of the DGN549-ADC resulted in a significant increase in efficacy in the high-expressing KB model (Fig. 5A). The increased efficacy at 10 mg/kg is notable, because the same 10 mg/kg carrier dose decreased efficacy in the non–cross-reactive system (Fig. 2B). To image the impact of dose on tissue distribution, 2.5 mg/kg of fluorescently labeled rmFR1–12 was injected intravenously into mice bearing high expression KB tumors with or without 10 mg/kg of non-fluorescent rmFR1–12. At 48 hours after injection, the 2.5 mg/kg dose was heterogeneously distributed whereas the addition of 10 mg/kg carrier dose resulted in efficient tissue penetration and homogeneous targeting (Fig. 5B). Interestingly, the carrier doses improved efficacy in the low expression OV90 model at all doses tested (up to 10 mg/kg, Fig. 5C). This is in direct contrast with the non–cross-reactive ADC where any carrier dose decreased efficacy (Fig. 2C). Therefore, the impact of a carrier dose on TMDD and increasing tumor uptake can be equally important as tissue penetration and cellular delivery effects.
Optimal carrier dose depends on both systemic (Plasma) and local (Tumor) effects
Administration of carrier antibody improved the efficacy of the high potency rmFR1–12-DGN549 in high- and low-expressing xenograft models. Next, we tested whether a carrier dose also had an impact on the activity of the lower potency rmFR1–12-s-SPDB-DM4 that can be administered at higher doses (e.g., 2.5 mg/kg of ADC). The cross-reactive rmFR1–12-s-SPDB-DM4 at 2.5 mg/kg had reduced efficacy in the high expression KB model (Fig. 3D) and resulted in heterogeneous distribution (Fig. 5B). The addition of 2.5 mg/kg of rmFR1–12 antibody to 2.5 mg/kg of ADC improved efficacy, but a higher carrier antibody dose (10 mg/kg) resulted in the greatest efficacy with 4/6 complete responses (Fig. 6A).
Despite the improvement noted from the carrier dose for both payloads and in the high and low expression models, it is anticipated that too large of a carrier dose will decrease efficacy by blocking too much cellular uptake (similar to Fig. 2C and F). To test this hypothesis, a 25 mg/kg dose of rmFR1–12 antibody was co-administered with 2.5 mg/kg of rmFR1–12-s-SPDB-DM4. This indeed lowered efficacy close to levels seen without any carrier dose. Tumor uptake of ADC payload measured using radiolabeled DM4 (Fig. 6B) was found to increase with the addition of a 2.5 or 10 mg/kg carrier dose of antibody from slower plasma clearance. However, a 25 mg/kg carrier dose reduced the tumor uptake to less than that of the ADC-only dose (although not significantly different). It is interesting to note that total payload uptake was similar for both the 2.5 and 10 mg/kg carrier dose (Fig. 6B), but the 10 mg/kg carrier dose was much more effective (Fig. 6A). This is consistent with the 10 mg/kg carrier dose improving efficacy by two mechanisms: increasing the total tumor uptake of ADC (relative to no carrier dose) by blocking normal tissue uptake and improving tissue penetration (relative to the 2.5 mg/kg carrier dose, Fig. 5B; Supplementary Fig. S6).
Potential impact of carrier doses on toxicity
ADCs often suffer from a narrow therapeutic window due to dose-limiting toxicity, which can be either target-mediated or non–target-mediated depending on the particular target, antibody, and payload among other factors. Even with cross-reactive antibodies, mice are not ideal models for tissue-specific uptake given differences in expression levels relative to humans, but the impact of a carrier dose on biodistribution can highlight the expected effect in other species (e.g., cynomolgus monkeys, humans). The biodistribution of rmFR1–12-s-SPDB-DM4 (radiolabeled DM4) was measured as a function of the carrier dose in both liver and kidney. The liver uptake was largely independent of the carrier dose (Fig. 6C), indicating primarily non–target-mediated uptake. However, folate receptor is expressed at a high level in the kidney. The carrier dose significantly decreased radiolabeled DM4 uptake in the kidney proportional to the carrier dose, with 25 mg/kg reducing uptake almost 5-fold (Fig. 6D).
Many ADCs currently in late-stage clinical trials or recently FDA approved, such as enfortumab vedotin (anti–nectin-4), sacituzumab govitecan (anti–Trop-2), trastuzumab deruxtecan (anti-HER2), and mirvetuximab soravtansine (anti-FRα), target tumors expressing high receptor levels (20–23). Target antigens overexpressed on tumor cells are often accompanied by lower but still significant expression in normal tissue. Because antibodies are frequently generated using mouse systems, cross-reactivity between mouse and human targets is rare. This can create a disconnect between estimates of therapeutic doses, typically measured in mouse models, and tolerable doses based on toxicity, typically measured in non-human primates. The impact of normal tissue uptake and increased clearance seen in primate models is not recapitulated in the mouse models used for efficacy, where the ADC only binds to cells in the tumor. However, TMDD and clearance is a widespread and important phenomenon that must be taken into account during drug development (19, 24, 25). In this work, we generated a cross-reactive murine antibody and ADC to test the impact of antibody dosing, target expression, and payload potency on efficacy in an FRα-targeted system where normal tissue expression is relevant.
One method to overcoming the sink for ADCs caused by expression of receptors in normal tissues is to increase the dose. Once the normal tissue receptors are saturated, the pharmacokinetics become linear and clearance is reduced. Indeed, several ADCs recently approved or in late-stage clinical trials are administered at high doses, exceeding that of the first ADC for solid tumors (ado-trastuzumab emtansine). Co-administration of unconjugated antibody can also improve tissue penetration and efficacy as demonstrated by adding trastuzumab to the clinical dose of ado-trastuzumab emtansine (26). However, adding too much unconjugated antibody (at a fixed ADC dose) could block cellular uptake to an extent that results in a decrease in efficacy. The appropriate amount of antibody to maximize efficacy depends on several factors, most prominently the receptor expression (the major driver of cell delivery) and payload potency (the major driver of cell death). To first examine how increased antibody doses impact efficacy without the confounding effects of normal tissue binding, a non–cross-reactive anti-folate antibody was tested with two different payloads in a high and low expression model.
In a “clean” system, where the ADC only binds to human FRα in the xenograft, the highest efficacy was seen when the cellular delivery matches the potency as we and others have reported (2, 27, 28). DM4-conjugates are typically tolerated at several mg/kg doses in the clinic, whereas DNA-alkylating conjugates are often tolerated at <0.5 mg/kg. Treatment with the DM4-ADC resulted in complete responses in the high-expressing tumor model at the 2.5 mg/kg antibody dose due to high cellular uptake rate (Fig. 1A). In contrast, the lower antibody dose of the DGN549-ADC gave poor responses despite the high potency due to the lower antibody tissue penetration (Fig. 1B). However, in a low-expressing model, the opposite was true. The DM4-conjugate was “potency-limited” due to cellular accumulation of DM4 being below the concentration required for cell death, but the high potency DGN549 gave complete responses (Fig. 1C and D).
Comparing the DGN549-ADC treated tumors in the high and low expression system, the efficacy of this ADC actually improves with lower expression despite the ADC being more potent on the high-expressing cells in vitro (Supplementary Fig. S1). This is similar to the result seen in a prostate cancer model with an ADC targeting PSMA (29). Together with the different sensitivities between clinical patient tumors, this is likely a major confounding factor for why a clear correlation between target expression and efficacy has not been found. Because lower expression can result in increased ADC tissue penetration even at low doses, using high potency payloads can lead to higher efficacy by allowing better penetration of lethal concentrations of the payload into the tissue.
To investigate the effects of a carrier antibody dose in a more representative scenario where the antibody and ADC bind receptors expressed in normal tissue, a cross-reactive antibody was developed. A rabbit antibody that binds with comparable affinity with mouse and human folate receptor was grafted onto a murine framework for these studies. When administered at the same dose as the non–cross-reactive version, plasma clearance increased resulting in a 3-fold drop in AUC and significantly lower tumor uptake. The efficacy of the ADC was appreciably compromised compared with the “clean” system with no normal tissue binding. By co-administering increasing doses of the antibody with the ADC (both cross-reacting), the plasma concentration of antibody/ADC increases, presumably due to saturation of the normal tissue, resulting in linear kinetics. Although Cmax of the ADC (given at a fixed dose) is not affected, the AUC increased due to lowered target-mediated clearance. Note that these are SCID mice, so the reduced clearance is attributed to blocking folate receptor rather than Fc-receptors (30). However, the conclusions are the same in either case. The improvement in systemic exposure with the addition of unconjugated antibody resulted in increased efficacy for all carrier doses tested with the high potency DGN549-ADC and high expression model. Interestingly, even in the low expression model, where the expectation is that a carrier dose would reduce cellular delivery too much (and lower efficacy; ref. 31), the improvement in tumor uptake was enough to overcome this limitation, resulting in higher efficacy with the carrier dose (Fig. 5). Therefore, even targets with expression as low as 40,000 receptors per cell can benefit from carrier doses depending on the potency of the payload and normal tissue receptor levels.
The impact of a carrier dose depends on a variety of factors, including the dose administered, so adding a carrier dose does not always improve distribution and/or efficacy. This is apparent when increasing the dose after the tumor is already saturated; any increase in a carrier dose will not improve tissue penetration since all accessible antigen within the tumor is already targeted. A more subtle but important effect occurs at lower doses, particularly in high expression models. To allow deeper tissue penetration, the carrier dose must block a sufficient amount of antigen on perivascular cells. In essence, the antibody must “fill up” the binding sites on the first cell layer to reach the next. Therefore, adding a small carrier dose at the low end of the dose range may not improve penetration, and this may contribute to the lack of improvement with only 1 mg/kg carrier dose in Fig. 5A.
Carrier doses improved the efficacy of the lower potency DM4-ADC in the high expression model when using the cross-reactive antibody. Despite the improvement from the carrier doses, there still exists a limit beyond which increased antibody doses start to lower efficacy. In this system, 2.5 mg/kg of ADC administered with a 10-fold higher antibody dose (25 mg/kg) super-saturated the tumor, lowered uptake, and reduced efficacy to levels seen with no carrier dose. Taken together, minimal efficacy is seen with the ADC alone; addition of the 2.5 mg/kg carrier dose increases the tumor uptake (systemic effect, Fig. 6B) but only marginally increases efficacy; addition of a 10 mg/kg carrier dose provides the same improvement in tumor uptake (systemic effect) but also improves tissue distribution (local effect) to result in a strong response. Note that before saturation, no decrease in tumor uptake of the ADC was seen, similar to Kadcyla (27) and consistent with theoretical predictions (32). Beyond this dose (25 mg/kg carrier dose in the cross-reactive system), there is no improvement in tissue penetration, and increased carrier doses simply reduce the tumor and cellular uptake of the payload, lowering efficacy. This is one of several (33, 34) considerations necessary when optimizing DAR and conjugation chemistry. The optimal results can differ depending on whether a saturating dose or subsaturating dose is used (19, 35).
Mouse models are not ideal for measuring toxicity, and the doses used here did not significantly impact mouse body weight (Supplementary Figs. S8 and S9). However, the biodistribution of the payload in this system provides some potential insight on the impact of carrier doses. The non–target-mediated uptake in the liver is the same at all carrier doses whereas the target-mediated kidney uptake is dramatically reduced. This observation matches computational models (27) and experimental data (e.g., antibody pre-dosing with anti-TENB2 ADCs reduces intestinal uptake while maintaining tumor efficacy; refs. 36, 37), and could help lower target-mediated toxicity.
Antibody distribution in clinical tumors is highly heterogeneous (38). Clinical tumors often show different characteristics than cell line xenografts, including greater stromal tissue and lower vascularity. The higher vascularity and permeability of mouse models could mask the importance of drug tissue penetration in human tumors that may be even lower due to reduced vascularity. Computational simulations of clinical tumors, supported by clinical observations, can help predict the impact of dosing in patients to guide drug design. Increasing the antibody dose of ADCs (e.g., by reducing the DAR) has the potential to improve human tumor tissue penetration and enhance tumor uptake, and these results could explain some clinical data. The anti-MUC16 ADC sofituzumab vedotin (DAR 3.5) showed a 17% response rate when administered at 2.4 mg/kg every 3 weeks (39). When the antibody was modified to a site-specific DAR 2 ADC, it could be dosed at 5.2 mg/kg every 3 weeks, generating a 45% response rate (40).
There are also several regulatory considerations regarding co-administration of unconjugated antibodies with ADCs. Separate approval of an antibody and ADC is unlikely for several reasons, particularly for an antibody that is not expected to have efficacy on its own. However, current ADC formulations already have a distribution of different DARs with some unconjugated antibody present. In fact, gemtuzumab ozogamicin (Mylotarg) consists of a 1:1 mixture of unconjugated antibody and ADC (41, 42). Although this formulation was selected to avoid ADC aggregation (43), the evident role of tissue penetration may require a re-evaluation of the prevailing view that unconjugated antibody is always undesirable (1). Rather, the optimal amount of unconjugated antibody (if any) needs to be selected on the basis of target expression/internalization, payload potency, etc. Alternatively, the DAR could be reduced to the point of maximum efficacy. Typically, the DAR has been reduced only to the point where DAR-dependent clearance is minimized (33). However, depending on the target and payload, a further reduction could result in an improved therapeutic window, and this approach is under investigation (44, 45). Likewise, higher DARs and associated technologies may be needed for lower expression and/or lower potency payloads for appropriate “matching.” From an intratumoral payload delivery and antibody dosing perspective, adding unconjugated antibody to an ADC is similar to lowering the DAR. However, from the perspective of systemic distribution, effects such as DAR-dependent deconjugation and DAR-dependent clearance (33, 46, 47) can result in disparities from the variable DAR distribution despite the same average DAR; these must be taken into account when investigating a carrier dose versus lowering the DAR.
Finally, it is important to point out that both DM4 and DGN549 exhibit bystander killing, where the released payload can diffuse out of the targeted cell and kill adjacent cells (48–50). Tissue penetration is still an important consideration even with bystander-capable payloads; however, as evidenced experimentally in the current study (e.g., Fig. 1) and previous examples (27, 31). Improved tissue penetration of a single-domain anti-PSMA antibody also showed improved efficacy over larger constructs (29) when using the DGN549 payload. A computational analysis by Khera and colleagues (11) indicates that these results are due to the higher efficiency of direct cell killing relative to bystander killing.
In summary, increased antibody dosing with ADCs can improve both local delivery (better tissue penetration) and systemic delivery (reducing target-mediated clearance) resulting in increased efficacy. Importantly, blocking target-mediated uptake in normal tissues can also reduce toxicity in these tissues. Combined with the potential for improvements in other mechanisms of action, such as antibody signaling receptor blockade and/or Fc-effector functions, this can genuinely expand the therapeutic window rather than simply shift the bar. The utility of this technique is a function of many factors, including receptor expression in the target tissue, normal tissue expression, target internalization rate, and payload potency. Therefore, a judicious, quantitative pharmacology approach is necessary to clearly define the potential benefits of clinical application. These complex molecules require equally sophisticated simulations to identify optimal parameters for clinical efficacy. As demonstrated by current clinical ADCs and promising agents in the pipeline, these factors can be taken into account to gain both the targeting specificity of antibodies and the potency of small molecules for effective therapies.
R. Laleau reports other from ImmunoGen, Inc. (reports employment with ImmunoGen at the time of the research for the article) during the conduct of the study. B. Matin reports personal fees from Deciphera Pharmaceuticals outside the submitted work. R. Gregory reports other from ImmunoGen (reports employment) outside the submitted work. G.M. Thurber reports grants from NIGMS, as well as non-financial support from ImmunoGen (provided samples for imaging) during the conduct of the study; as well as non-financial support from Synaffix (conjugation services), grants from Crescendo, personal fees; as well as grants and non-financial support from Immunomedics; as well as personal fees from Seattle Genetics and Advanced Proteome Therapeutics; grants, personal fees, and non-financial support from Takeda; and personal fees from Genentech and Mersana Therapeutics outside the submitted work. No disclosures were reported by the other authors.
The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
J.F. Ponte: Conceptualization, formal analysis, supervision, investigation, methodology, writing-original draft, writing-review and editing. L. Lanieri: Conceptualization, resources, formal analysis, validation, investigation, methodology, writing-review and editing. E. Khera: Software, formal analysis, investigation, visualization, methodology, writing-original draft, writing-review and editing. R. Laleau: Data curation, investigation, methodology. O. Ab: Resources, data curation, investigation, methodology, writing-review and editing. C. Espelin: Supervision, investigation, project administration. N. Kohli: Conceptualization, resources, supervision, investigation, methodology. B. Matin: Software, investigation. Y. Setiady: Supervision, project administration. M.L. Miller: Conceptualization, investigation. T.A. Keating: Data curation, supervision, investigation. R. Chari: Supervision, project administration, writing-review and editing. J. Pinkas: Supervision, funding acquisition, project administration. R. Gregory: Resources, supervision, funding acquisition, project administration. G.M. Thurber: Conceptualization, supervision, funding acquisition, writing-original draft, project administration, writing-review and editing.
The financial support was provided by ImmunoGen, NIH R35 GM128819 (to G.M. Thurber), and the National Cancer Institute of the National Institutes of Health under Award number P30CA046592 by the use of the following Cancer Center Shared Resource(s): Histology.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.