CD38-bispecific antibody pretargeted radioimmunotherapy for multiple myeloma and other B-cell malignancies

CD38, a transmembrane glycoprotein with high surface density and uniform expression on multiple myeloma (MM) cells¹  and relatively low expression on normal myeloid and lymphoid cells,²  has proven a successful target for monoclonal antibody (mAb)-based immunotherapy in MM.^3,4  Very effective agents introduced in the past decade have made complete response (CR) to induction therapy possible in almost half of MM patients.^5,6  However, disease eradication remains elusive, creating conditions that strongly favor the persistence and evolution of therapy-resistant malignant plasma cell clones. As a result, the vast majority of the 120 000 people in the United States living with MM⁷  will ultimately die of progressive disease.^8,9  High-dose chemotherapy followed by autologous stem cell transplantation (ASCT) increases CR rates and prolongs disease-free survival,^10-14  and the number of ASCTs for MM increases annually, emphasizing the importance of ASCT in current treatment paradigms.¹³  Yet, disease almost invariably recurs even after ASCT.

Radioimmunotherapy (RIT) promises 2 critical improvements over other MM treatments: less-toxic conditioning for ASCT and the potential to eradicate disease. The radiosensitivity of malignant plasma cells outside of the bone marrow is well documented in clinical settings. Local recurrence of solitary extramedullary plasmacytomas occurs in <10% of cases after external-beam radiotherapy (RT) alone,¹⁵  and sustained local disease control and durable symptom relief has been reported for 98% of lesions receiving >10 Gy.¹⁶  Furthermore, this excellent efficacy of external-beam RT for extramedullary plasmacytomas occurs even in patients with poor-risk cytogenetics and active MM,¹⁷  suggesting that targeted RIT is agnostic to certain high-risk features.

RIT selectively delivers radiation to target cells at disseminated disease sites, facilitating escalation to radiation doses not achievable through external-beam RT. The efficacy of RIT is well established for several hematologic malignancies,^18-20  and has been successfully integrated into ASCT-conditioning regimens with a significant improvement in progression-free survival and overall survival among patients with non-Hodgkin lymphoma (NHL) and acute myeloid leukemia when targeting CD20 and CD45 antigens, respectively.^18,21-24  In contrast, few studies have examined RIT in MM^25-30  and none have explored pretargeted RIT (PRIT), a 2-step process shown in the clinic to be markedly superior to conventional, single-step RIT.^31-36  In conventional RIT, a targeting antibody (Ab) is directly labeled with a radioactive molecule. In 2-step PRIT, a nonradioactive targeting Ab is administered first and allowed to localize to tumor sites. The second step, administered after this “cold” Ab has maximally accumulated in the tumor, is a low-molecular-weight radioactive moiety with a high affinity for the Ab. The small size of the second reagent facilitates rapid tumor penetration, rapid capture and retention by the pretargeted Ab, and rapid clearance of unbound radioactive molecules from the blood. This 2-step approach greatly decreases radiation absorption by healthy tissues. The efficacy of PRIT can be further amplified by administering a “clearing agent” (CA) prior to the radioactive reagent. The CA accelerates clearance of any unbound Ab from the bloodstream, greatly reducing the chance of radioactive molecules attaching to unbound Ab.

Unlike many surface antigens, CD38 is stable on the cell surface,³⁷  and this trait combined with high density and uniform expression in MM and NHL make CD38 an excellent target for PRIT. We recently documented striking therapeutic efficacy of PRIT in MM xenograft models using anti-CD38-streptavidin (SA; OKT10-SA) and the β-emitter ⁹⁰Y.³⁷  Objective remissions were observed within 7 days in 100% of mice treated with 800 to 1200 µCi CD38-SA PRIT, including 100% complete remissions (no detectable tumor in treated mice compared with tumors in control mice that were 2982% ± 1002% of initial tumor volume) by day 23. Despite these dramatic results, concerns with current PRIT approaches persist, including immunogenicity of bacterially derived SA, and the presence of endogenous biotin, which might block binding of radiolabeled biotin.

Here, we describe the production and characterization of a new anti-CD38-bispecific protein that entirely removes biotin binding and SA from PRIT, while maintaining high CD38-binding affinity and high avidity for the radiolabeled second step. We also directly compare therapeutic efficacy of the new CD38-bispecific with the CD38-SA approach. Several PRIT technologies have been developed to facilitate avidity of the radiolabeled second-step reagent for the pretargeted tumor-bound Ab. These include the SA-biotin and bispecific Ab approaches used here, complementary hybridization (Watson-Crick pairing) of phosphorodiamidate morpholino DNA oligomers,^38,39  and cyclooctene-modified Abs binding to radiolabeled tetrazine ligands.^40,41  Although prior studies show that all PRIT methods are superior to single-step RIT, no head-to-head comparisons have examined which PRIT method is most promising for clinical development of CD38 targeting. We present a comparative analysis of the biodistribution and therapeutic efficacy of the 2 most popular PRIT strategies. Our findings demonstrate that efficacy, reduced immunogenicity, and absence of interference from endogenous biotin all make anti-CD38-bispecific PRIT an excellent candidate for clinical translation.

Methods detailing the construction of the 028-Fc-C825 bispecific (anti-CD38 × anti-⁹⁰Y-labeled 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid [Y-DOTA]) fusion gene, isolation of CHO cell clones stably expressing the 028-Fc-C825–bispecific Ab, purification of the bispecific Ab, and synthesis of the DOTAY-dextran (DYD) CA are described in supplemental Methods (available on the Blood Web site). Construction of the control, 2H7-Fc-C825 (anti-CD20 × anti-Y-DOTA) bispecific was described previously.⁴² 

Conjugates of the OKT10 anti-CD38 mAb and SA were synthesized, purified, and characterized as previously published.^43,44  A synthetic, dendrimeric CA containing 16 N-acetylgalactosamine residues and a single biotin residue per molecule (NAGB) was obtained from Aletheon Pharmaceuticals (Seattle, WA) for use with the OKT10-SA conjugate.

⁹⁰Y (PerkinElmer, Seattle, WA) labeling of 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (DOTA)-biotin was performed using 12 mg/mL DOTA-biotin, 500 mmol/L ammonium acetate (pH 5.3), and ⁹⁰Y heated for 60 minutes at 84°C. After cooling to room temperature, 100 mmol/L diethylenetriaminepentaacetic acid (DTPA) was added, and labeling was efficiency determined using avidin-agarose beads.⁴³ 

The human MM cell lines H929, U266B1, and RPMI 8226 were obtained from ATCC (Manassas, VA). These lines and the CD38⁺, CD20⁻ human Burkitt lymphoma (BL) cell line Namalwa were authenticated by DNA profiling (ATCC kit 135-XV), tested for mycoplasma, and maintained in log-phase growth at >95% viability (trypan-blue exclusion) in RPMI 1640 media supplemented with 10% fetal bovine serum (FBS), 50 U/mL penicillin G, and 50 μg/mL streptomycin sulfate, for no more than 6 weeks after thawing.

Log-phase growth H929 cells (0.5 × 10⁶ per group) were harvested and washed once with 1 mL of Hank balanced salt solution (HBSS)–2% FBS (HBSS) buffer. For CD38-blocking groups, cells were resuspended in 40 µL of HBSS buffer containing 40 µg of daratumumab (Janssen R&D, Raritan, NJ) and incubated at 4°C for 30 minutes. Then, 2 µg of bispecific fusion protein and 1 µg of Y-DOTA–biotin were added to all groups and cells incubated at 4°C for 30 minutes, followed by 2 washes. Recovered cells were resuspended in 40 µL of HBSS buffer containing 2 µL of phycoerythrin-labeled SA (AnaSpec Inc, Fremont, CA), incubated at 4°C for 30 minutes, washed 3 times, resuspended in 500 µL of phosphate-buffered saline (PBS) buffer containing 1% formaldehyde and analyzed on a Guava Easycyte mini cytometer.

Female FoxN1Nu athymic nude mice (Envigo, Hayward, CA) were maintained under standard protocols approved by the Fred Hutchinson Cancer Research Center (FHCRC) Institutional Animal Care and Use Committee (IACUC). CD38⁺ tumor cells (10⁷) were injected subcutaneously in the right flank 7 to 11 days prior to experiments to produce 50- to 80-mm³ tumor xenografts. To attenuate tumor rejection due to natural killer (NK) cell activity, mice were injected intraperitoneally with anti-asialoGM1 Ab (Wako, Richmond, VA) 1 day prior to tumor implantation, 5 days later, and weekly thereafter. All mice were placed on a biotin-deficient diet (TestDiet, Richmond, IN) 7 days prior to PRIT studies.

Groups of 3 to 5 mice bearing H929 flank tumors were injected via the tail vein (IV) with 2.8 nmol (210-420 μg) of 028-Fc-C825 (CD38 bispecific) or 2H7-Fc-C825 (CD20 control bispecific). Optimal Ab dosing was determined in prior experiments.^37,42  Twenty-three hours later, mice were injected with 5 µg of DYD CA, followed 1 hour later by 1.2 nM (2 μg) DOTA-biotin labeled with 20 to 40 μCi (0.74-1.48 MBq) of ⁹⁰Y. For blood clearance studies, retro-orbital blood sampling was performed at serial time points up to 24 hours after ⁹⁰Y injection. For biodistribution studies, blood samples, tumors, and body organs were harvested at 6 to 120 hours. ⁹⁰Y dose in each tissue sample was counted on a γ counter and the percentage of injected dose per gram (percent injected dose [ID] per gram) was calculated.⁴⁵ 

Mean absorbed doses to organs and tissues were calculated from the activity-over-time curves generated in biodistribution experiments. The calculations estimate organ self-dose plus cross-organ dose by accounting for organ mass, specific absorbed energy fraction, the emission spectrum of the radionuclides, and the β-particle absorbed fractions for small organs.⁴⁵  The results were expressed as radiation absorbed dose (gray) per unit administered activity (per millicurie).

Therapeutic efficacy of CD38 bispecific and CD38-SA PRIT were assessed in groups of 8 to 10 mice (sample size determined by power analysis) bearing flank H929 or Namalwa xenografts. Mice were randomized into groups with equivalent mean tumor volumes, and treatments were administered as in “Blood clearance and biodistribution studies,” with the addition of a CD38-SA (OKT10-SA) Ab treatment group that received 50 μg of NAGB as CA. All groups received second-step DOTA-biotin labeled at 600, 1000, or 1200 μCi of ⁹⁰Y (22.2, 37, or 44.4 MBq). Tumor size and body weight were measured 3 times a week following treatment. Mice were killed when they experienced excessive weight loss, hind-limb paralysis, or exceeded tumor volume limits per IACUC requirements.

In murine xenograft studies, treatment effects on tumor growth rate were determined by first calculating tumor growth rate for each mouse as area under the curve of tumor volume over time, standardized for the number of days the mouse was alive. Treatment effects on standardized tumor area under the curve were then determined using analysis of variance. Treatment effects on mouse survival were determined by log-rank comparisons of Kaplan-Meier survival functions. Analyses were performed using JMP 12.2.0 (SAS Institute, Cary, NC) and GraphPad Prism 7 (GraphPad Software, La Jolla, CA).

The 028-Fc-C825–bispecific Ab fusion protein, designed to recognize the CD38 surface antigen and the yttrium-DOTA ligand, was constructed by fusing DNA fragments encoding scFv of the 028 anti-CD38 human Ab and the ds-scFv of the affinity-enhanced C825 Ab to both sides of a human immunoglobulin G1 Fc containing a N-linked glycosylation (NLG) linker (Figure 1A). CHO-DG44 cells were transfected with the fusion construct DNA and high-expressing clones were selected using methotrexate. The monomeric bispecific protein spontaneously dimerized to form a 160-kDa molecule, which was purified from culture supernatants by affinity chromatography and characterized by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) (Figure 1B); enzyme-linked immunosorbent assay showed that the bispecific captured the Y-DOTA ligand in a concentration-dependent fashion (Figure 1C). Bifunctional flow cytometry analysis demonstrated that the bispecific bound only to cells expressing CD38, that binding was blocked by unmodified anti-CD38 mAb binding the same epitope, and that the bispecific captured the Y-DOTA ligand (Figure 1D-E). The anti-CD38 mAb OKT10 did not block cell binding of the CD38 bispecific, indicating that the binding epitopes did not overlap (supplemental Figure 1).

Pharmacokinetics (PK) analysis of the CD38-bispecific protein required our standard, 2-step PRIT protocol because direct radioiodination of the bispecific protein impaired binding. For PRIT, mice were injected first with the unlabeled protein, followed 23 hours later with CA, then 1 hour later with ⁹⁰Y-DOTA–biotin. Blood samples were taken at 5 minutes, 30 minutes, and 1, 4, and 24 hours after the ⁹⁰Y injection. Figure 2 demonstrates that without CA, blood clearance of ⁹⁰Y was 72.8% ID per gram after 30 minutes and 86.8% after 24 hours. With 5 µg of CA, clearance was 97.8% ID per gram at 30 minutes and 99.3% after 24 hours. We additionally tested increasing doses of CA; all doses produced virtually identical results (eg, clearance using 32 µg of CA was 98.2% ID per gram at 30 minutes and 99.8% after 24 hours; data not shown). We therefore used the 5-µg dose for further experiments.

Biodistributions of radioactivity in blood, tumors (H929 xenografts), and normal organs were compared between CD38-bispecific PRIT and control (CD20-targeted) bispecific PRIT (Figure 3A). Tumor-bearing mice were injected at −24 hours with unlabeled protein, at −1 hour with CA, then at 0 hour with ⁹⁰Y-DOTA–biotin. Assessed 24 hours after ⁹⁰Y-DOTA–biotin injection, tumor-to-normal tissue ratios of absorbed radiation were 16:1 for blood, 14:1 for lung, 12:1 for liver, and 19:1 for kidney for the CD38-bispecific group (Figure 3A). For the control-bispecific group, ratios were <2:1 for the same tissues (Figure 3A). We also evaluated radioactivity biodistributions over time, taking tissues 6, 24, 48, and 120 hours after ⁹⁰Y-DOTA–biotin injections. These studies confirmed the tumor specificity of CD38-bispecific PRIT, and further demonstrated high retention of radiation in tumors over time (Figure 3B). Control-bispecific PRIT showed no tumor targeting or retention (Figure 3C).

We estimated radiation-absorbed doses to tumors, whole body, and 10 normal tissues from time-activity curves generated in CD38-bispecific biodistribution experiments, using a dosimetry method that calculates both organ self-dose absorbed fractions and β-particle cross-organ dose contributions, per unit of administered activity (Figures 3B and 4).⁴⁵  Absorbed radiation dose to tumor was 43.8 Gy/mCi, and tumor-to-normal organ ratios of absorbed dose were 51:1 for the femur, 23:1 for the whole body, 15.4:1 for the lung, 15:1 for the kidneys, 10:1 for the blood, and 7.4:1 for the liver (Figure 4).

We studied the efficacy of CD38-bispecific PRIT in 2 xenograft models, and found that optimized PRIT dosing cured at least 75% of mice in every experiment. Athymic mice (n = 8-10 per group) bearing H929 MM or Namalwa NHL xenografts were injected first with cold Ab, then 23 hours later with CA, and then in 1 more hour with ⁹⁰Y-DOTA–biotin. Previous PRIT studies by our group^37,42  indicated that 1200 µCi ⁹⁰Y provides the most favorable therapeutic ratio, and we first examined efficacy of the CD38-bispecific Ab at 1200 µCi. In the H929 model, treatment with CD38-bispecific PRIT dramatically reduced tumor growth (Figure 5A) and increased survival (Figure 5B) relative to treatment with bispecific control PRIT (1200 µCi) and untreated control groups (P ≤ .0002, CD38 bispecific vs either control). In total, 19 of 20 mice in the 2 control groups died of tumor progression within 30 days (1 untreated mouse showed spontaneous tumor regression), whereas in the CD38-bispecific group 8 mice sustained complete remissions through the end of the study (150 days) and 2 died of tumor progression, both after day 50. In a second CD38⁺ tumor xenograft model (Namalwa), we added head-to-head comparisons of the CD38-bispecific Ab vs the CD38-SA (OKT10-SA) Ab, each labeled with 1200 µCi ⁹⁰Y. In this experiment, both CD38 treatments resulted in 100% CRs by day 21, followed by 1 subsequent relapse in the CD38-SA group and zero relapses in the CD38-bispecific group (Figure 6). Thus, at this dose, the 2 CD38 treatments reduced tumor growth and increased survival with high and equivalent efficacy (P < .0001 for either CD38 treatment vs either control, P = .48 for CD38 bispecific vs CD38-SA). To further characterize the dose-response relationship of CD38 PRIT, we also evaluated 600 and 1000 µCi ⁹⁰Y pretargeted to xenograft tumors. These reduced-dose experiments compared CD38-bispecific PRIT vs CD38-SA PRIT, and both anti-CD38 systems reduced tumor growth (Figure 7A) while improving survival (Figure 7B) in a marginally dose-dependent manner (P < .062, PRIT 600 µCi vs PRIT 1000 µCi). Importantly, at these reduced doses, the CD38-bispecific Ab strongly outperformed CD38-SA, resulting in long-term survival of 61% vs 22%, respectively, with 600 µCi, and 78% vs 47%, respectively, with 1000 µCi (P < .004, CD38 bispecific vs CD38-SA). In contrast, all controls, including mice treated with cold (no radiolabel) CD38 bispecific, cold CD38-SA, or control bispecific at 1000 µCi, died of tumor progression by day 15 (Figure 7, P < .0001, any CD38 PRIT group vs any control group).

PRIT using all Abs was well tolerated, with minimal weight loss and recovery to starting weight within 14 days of treatment (supplemental Figure 2). Over all studies (Figures 5-7), 2 of 18 mice treated with CD38 bispecific at 1200 μCi and 2 of 18 treated with CD38 bispecific at 1000 μCi died due to low body weight before day 18. All other deaths in all groups resulted from tumor progression. Anecdotal data suggest that early weight loss and mortality may be ameliorated as follows. Between our final 2 studies, we implemented husbandry changes to reduce external radiation exposure from cage mates and bedding; mortality before the changes was 2 of 8 mice and mortality after was 0 of 10 mice for 1000 μCi CD38-bispecific treatments (groups combined in Figure 7).

Combinations of immunomodulatory- and proteasome inhibitor–based therapies frequently induce remission in MM patients, but relapse is nearly inevitable and the need for development of potentially curative treatments remains critical. Our results demonstrate 75% to 88% cure rates using PRIT in 2 CD38-expressing murine xenograft tumor models. These results are consistent with the steep dose-response relationship between radiation and hematologic malignancies. In MM, external-beam radiation can cure isolated plasmacytomas and provide sustained local disease control in 98% of lesions receiving >10 Gy¹⁶ ; in follicular NHL, another generally incurable B-cell malignancy, external-beam radiation can also eradicate disease that is limited to a single site of involvement.^46,47 

RIT offers a delivery model designed to parlay the unique antitumor potency of radiation demonstrated in localized disease to a broader population of patients with multifocal disease by sparing healthy tissues while delivering a targeted radiation payload directly to malignant cells. Single-step RIT has demonstrated some promise in MM,^25-30  yet clinical applications have been very limited. Similar to CD20 single-step RIT, enthusiasm may be limited by low tumor-to-normal ratios (eg, <2.4:1 for kidney, lung, and liver in NHL^43,48 ). PRIT methods can greatly improve these therapeutic ratios. In MM models, we previously demonstrated that CD38-SA PRIT provides tumor-to-normal tissue dosimetry ratios of 6:1 for kidney, lung, and liver (Figure 5 in Green et al³⁷), and here show that CD38-bispecific PRIT provides ratios of 15:1 for kidney and lung and 7:1 for liver (Figure 4). To our knowledge, we have performed the only studies of PRIT in MM, and only 1 other group has used CD38 targeting for RIT (single-step).³⁰ 

Recent clinical successes with unmodified anti-CD38 mAb therapy in MM have motivated ongoing clinical studies of this therapy in B-cell NHL. High-density CD38 expression is a common feature of malignant B cells and the predictable growth kinetics associated with NHL cell lines have led to their frequent use in research models for the development of CD38-targeted MM therapy.³  Despite advances in the management of NHL, nearly 30% of these patients die of progressive disease within 5 years of diagnosis.⁴⁹  As in MM, treatment-refractory NHL typically retains sensitivity to radiation,^50,51  making CD38 PRIT a potentially effective treatment of such patients as well. Beyond tumors that share a common B-cell lineage, CD38 is expressed in most NK/T-cell lymphomas,⁵²  where 50% of patients die within 5 years⁵³  and overexpression of CD38 predicts poor outcomes for NK/T-cell lymphoma patients,⁵²  presenting another potential translational application for CD38 PRIT. We respectfully suggest that a broad range of potential indications for CD38 PRIT increases the probability of successful translation into a commercially viable radioimmunotherapeutic. To further evaluate CD38 as a target for RIT, we are conducting a clinical trial using a CD38 mAb directly conjugated to the α-emitter ²¹¹At (single-step RIT). We are also developing bispecific fusion protein constructs for α-emitter delivery that will facilitate head-to-head comparisons of α- and β-emitter PRIT.

Although the preclinical efficacy of PRIT appears clear, concerns have been raised regarding clinical translation. SA is a bacterial protein, and PRIT using SA might result in immunogenicity and thus limit the ability to administer multiple cycles of therapy, although 2 factors may mitigate this concern. First, the immunocompromised status of many patients with hematological malignancies limits immunogenic responses, and, second, PRIT is designed to be efficacious following a single dose, as demonstrated in both this manuscript and our previous studies. To reduce these concerns, several methods of modifying SA have been developed,^54-57  but approaches that eliminate SA are desirable. To achieve this, we developed the CD38-bispecific fusion protein, which exploits the same principles as the SA-biotin system, but replaces SA with a humanized yttrium-DOTA capturing C825 disulfide-stabilized scFv (Figure 1A). A bispecific construct harboring the same anti-yttrium–DOTA (C825) in an immunocompetent murine model (J.J.O., S.O., D.J.G., A.L.K., O.W.P., Bispecific antibody targeting CD45 and 90Y-DOTA in preclinical murine models of acute myeloid leukemia, unpublished observations) has demonstrated no evidence of toxicity. Reduced immunogenicity may allow for repeat dosing of CD38-bispecific PRIT, and fractionation may offer the opportunity to improve the therapeutic ratio in clinical settings. Our data here, however, suggest that PRIT is effective and well tolerated as a single-dose therapeutic.

A second concern for SA-PRIT is the potential for endogenous biotin in patient blood and tissues to occupy and block SA-binding sites, preventing subsequent binding of the second-step radio-DOTA-biotin reagents. The bispecific approach obviates this concern, as binding of the C825 portion of the bispecific to the DOTA portion of the second-step reagent precludes any possible interference from endogenous biotin.

Our data suggest that bispecific PRIT has superior efficacy when compared with CD38-SA PRIT (Figure 7). The anti-CD38 region of the bispecific protein binds a different epitope of CD38 than the OKT10-SA protein, as demonstrated by the absence of OKT10 mAb blocking of the bispecific CD38 in cell-binding assays (supplemental Figure 1). The binding of separate epitopes raises the possibility that the first-step CD38 targeting constructs have differing therapeutic effects, but we demonstrated that neither construct has antitumor efficacy when administered without radiolabeling (Figure 7). Alternatively, the therapeutic differences may be a consequence of differing in vivo–binding efficiencies. The bispecific and SA treatments were designed such that all mice received an equivalent injected protein dose carrying an identical radiolabel dose, but subsequent tumor binding might differ for 2 reasons. First, it is feasible that CD38-SA binding, and thus PRIT efficacy, was reduced by very low-level blocking by residual endogenous biotin in mouse tissues, despite institution of a biotin-free diet. Second, differences in Ab-ligand interaction may be a function of each construct’s avidity for its respective epitope, resulting in kinetics that may favor the bispecific.

In conclusion, we have characterized a new CD38-bispecific targeting protein for PRIT of MM and NHL. The bispecific Ab is relatively easy to produce, exhibits excellent blood clearance using an inexpensive CA, and shows excellent tumor-to-normal organ ratios of absorbed activity reflected by favorable dosimetry in murine models. Moreover, the CD38 bispecific rapidly reduced disease burden, and at optimal doses ultimately cured 75% to 80% of mice in each of 2 xenograft models with minimal toxicity. CD38-SA was equally effective at the highest radiation dose, but the bispecific’s superiority over a range of doses, combined with its reduced risk of immunogenicity and lack of endogenous biotin interference, make the CD38 bispecific a prime candidate for clinical translation.

This work was supported by grants from the National Institutes of Health, National Cancer Institute K08CA151682 (D.J.G.), R01CA205248 (D.J.G), K24CA184039 (A.K.G.), K01CA188151 (J.J.O.), K23CA154874 (B.G.T.), P01CA044991 (O.W.P.), R01CA076287 (O.W.P), R01CA136639 (O.W.P.), and R01CA154897 (O.W.P.), and by Multiple Myeloma Opportunities for Research and Education (MMORE), the Brotherton Family Fund, and the David and Patricia Giuliani Family Foundation.

Contribution: D.J.G. and O.W.P. designed research, analyzed data, and wrote the paper; S.O. designed and performed research, analyzed data, and wrote the paper; Y.L. performed research, contributed vital new reagents, and analyzed data; M.L.C. performed research and analyzed data; A.L.K., M.N., and M.D.H. performed research; D.K.H., D.S.W., K.D.O., and K.D.W. contributed vital new reagents.; and D.R.F., A.K.G., T.A.G., J.J.O., and B.G.T. analyzed data.

Conflict-of-interest disclosure: A.K.G. consults for, and receives research support paid to the institution from, Gilead and Janssen. B.G.T. and O.W.P. receive Roche/Genentech and Mustang Bio research support paid to the institution and receive royalties from patents held in common with Mustang Bio. K.D.O. and K.S.W. hold US patent 8 648 176 (Engineered Proteins with High Affinity for DOTA Chelates). O.W.P. consults for Bayer, sits on the data safety monitoring board for Bristol-Myers Squibb and Roche, and has ownership interest in Emergent Biosolutions and PhaseRx. The remaining authors declare no competing financial interests.

Correspondence: Damian J. Green, Clinical Research Division, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave N., MS: D3-190, Seattle, WA 98109; e-mail: dgreen@fredhutch.org.