We analyzed 57 patients with RRMM enrolled in a phase IIb clinical trial (registered at www.clinicaltrials.gov) evaluating treatment with Isa-Len-Dex.8 These patients had a confirmed diagnosis of MM, had progressed or were refractory to the immediate prior therapy, and had received more than two lines of anti-MM treatment, including many of whom were Len exposed and refractory.8 This phase Ib dose-escalation study assessed the safety, efficacy, and pharmacokinetics of Isa given in two schedules (3, 5, or 10 mg/kg every other week (Q2W) or 10 or 20 mg/kg weekly (QW) for 4 weeks and then Q2W thereafter (QW/Q2W)), in combination with Len 25 mg (days 1–21) and Dex 40 mg (QW). Patients received 28-day treatment cycles. Twenty-four patients were treated in the dose-escalation phase (3 mg/kg Q2W, n=4; 5 mg/kg Q2W, n=3; 10 mg/kg Q2W, n=6; 10 mg/kg QW/Q2W, n=3; 20 mg/kg QW/Q2W, n=8). Three patients were replaced for dose-limiting toxicity evaluation due to grade 3 infusion-associated reactions that occurred during the first Isa infusion and led to treatment discontinuation (1 patient at 3 mg/kg Q2W and 2 patients at 20 mg/kg QW/Q2W). We included all 57 patients with RRMM that received at least one dose of Isa-Len-Dex therapy despite dosage level 3, 5, 10, or 20 mg/kg and treatment schedule. Baseline patient demographics, disposition, and disease characteristics were summarized previously.8

Isa mAb was provided by Sanofi. Anti-HLA-ABC-FITC (clone: REA230), anti-CD38-FITC-PE (clone: IB6), anti-CD3-FITC (clone: BW264/56), and anti-CD56-PE (clone: AF12-7H3) were purchased from Miltenyi Biotec Inc (Auburn, California, USA). EasySep Direct human NK cell isolation kit was purchased from STEMCELL Inc (Vancouver, British Columbia, Canada). Carboxyfluorescein succinimidyl ester (CFSE), fixable viability dye (FVD), fetal bovine serum (FBS), interleukin 6 (IL-6), IL-2, and Penicillin-Streptomycin antibiotics (10,000 U/mL) were purchased from Thermo Fisher Scientific (Waltham, Massachusetts, USA).

We obtained genomic DNA samples from >40 human myeloma cell lines from Keats lab (https://www.keatslab.org/myeloma-cell-lines). HLA typing was performed using a Luminex-based oligonucleotide probe hybridization method according to the manufacturer’s instructions (One Lambda, Canoga Park, California, USA). Based on the HLA class I types, 12 MM cell lines with specific HLA class I ligands (online supplemental table 1), such as those having homozygous HLA-C (C1C1 or C2C2) and with or without HLA-A3/11 or Bw4 ligands were identified for purchasing from DMSZ (https://www.dsmz.de/catalogues/catalogue-human-and-animal-cell-lines.html) or JCRB cell banks (http://cellbank.nibiohn.go.jp/english/). The MM cell lines were cultured in Roswell Park Memorial Institute 1640 (RPMI-1640) medium (Invitrogen, Carlsbad, California, USA) supplemented with 10% heat-inactivated FBS (Sigma, St. Louis, MO), IL-6, and antibiotics (100 units/mL penicillin, 100 mg/mL streptomycin) at 37°C in a humidified atmosphere with 5% CO₂. These MM cell lines were routinely tested using MycoAlert PLUS Mycoplasma Detection Kit from Lonza (Walkersville, MD) and STR profiling to ensure they were free from mycoplasma.

Genomic DNA from human PBMC and MM cell lines was extracted using the Qiagen DNA extraction kit (Qiagen, Valencia, California, USA). The HLA class I and KIR genotyping were performed using Luminex-based oligonucleotide probe hybridization methods according to the manufacturer’s instructions (One Lambda, Canoga Park, California, USA). Prediction of KIR haplogroups and centromeric and telomeric gene clusters were performed as described previously.29 Briefly, those having only the genes of group A haplotype KIRs (3DL3-2DL3-2DL1-2DP1-3DP1-2DL4-3DL1-2DS4-3DL2) were assigned as AA genotype carriers. The BB genotype refers to be lacking any of the four A haplotype-specific KIR genes (2DL1, 2DL3, 3DL1, and 2DS4). The AB genotype is heterozygous, carrying all A haplotype-specific KIR genes plus other KIR genes.

The single-nucleotide polymorphisms of FCGR3A-F158V were analyzed by a single-step multiplex allele-specific PCR and melting curve-based assay using modified primers sets.30 The primer sequences are: V allele-specific reverse primer: 5′-AGACACATTTTTACTCCCAAC-3′; F allele-specific reverse primer: 5′-GCGGGCAGGGCGGCGGGGGCGGGGCCGGTGATGTTCACAGTCTCTGATCACACATTTTTACTCCCAAA-3′; FCGR3A-specific common forward primer: 5′-TCCAAAAGCCACACTCAAAGTC-3′). The sequence and specificity of the primers were verified by direct sequencing of PCR amplicons from nine subjects (online supplemental figure 1). Only 50 patients were typed for FCGR3A-V158F dimorphism, and the quantity of DNA from the remaining 7 patients was not sufficient to perform this genotyping.

Peripheral blood samples were obtained from 11 selective donors with specific KIR+HLA compositions (online supplemental table 2). The NK cells were isolated by an immunomagnetic negative-selection method using the EasySep direct procedure according to the manufacturer’s instructions. The purity of NK cells following isolation was evaluated by flow cytometry, and those with >95% were used. These freshly isolated NK cells were stimulated overnight in IL-2 (100 U/mL) added RPMI-1640 medium supplemented with 10% heat-inactivated FBS.

To evaluate the impact of KIR receptors and cognate HLA class I ligand interactions on Isa-dependent NK cytolysis of MM cell line targets, a recently developed flow cytometry method using CFSE labeling and FVD staining was employed.31 Briefly, MM cell line targets were prelabeled with CFSE for 8 min and then divided into 10,000 cells/well in a 96-well culture plate. Target killing was initiated by the addition of following six different coincubation treatments overnight at 37°C: (1) no treatment as normalization control; (2) Len (10 µM); (3) Dex (10 µM); (4) Len and Dex (10 µM each); (5) Isa (10 µg/mL); (6) IL-2 activated NK cells (2.5:1 as the effector:target ratio); (7) IL-2 activated NK cells+Isa. The cells were harvested and stained with FVD for 20 min for the detection of dead cells. The cells were then washed with phosphate-buffered saline containing 2% FBS and then subjected to flow cytometry analysis using Fortessa x-20 (BD Biosciences, San Jose, California, USA). Data were collected, analyzed using Flowjo software. MM cell death rate was calculated by numbers of normalized CFSE⁺FVD⁺ cells/numbers of normalized CFSE⁺ cells. Len or Dex treatment for 72 hours with dosages ranging from 0 to 100 µM resulted in 0%–20% of death with all 10 MM cell lines except Dex-treated KAS61, which showed up to 40% cell death (online supplemental figures 2 and 3). Len and Dex (10 µM each) combined showed an additive effect on MM cell line death during 24 hours (online supplemental figure 3). To avoid the Len/Dex-mediated cytotoxicity background, we used untreated MM cell lines in all subsequent experiments to precisely examine the impact of KIR+HLA interactions on Isa-dependent NK cytolysis.

In the phase IIb study, progression-free survival (PFS) was defined as the time from the first dose of study medication to either disease progression or death, whichever came first. PFS was censored at the date of the last valid assessment before the cut-off date in the absence of either disease progression or death before the analysis cut-off date. The PFS estimates were calculated using the Kaplan-Meier method.32 All cytolysis experiments were conducted in triplicates and repeated at least twice. The mean percentage ±SD of each experimental condition was calculated from 3 to 6 replicate wells. A paired t-test was used to compare cell killing by different conditions. The coefficient of drug interaction (CDI) was used to calculate combination synergism, as previously described.33 The CDI is calculated as following: CDI = AB/(A×B). AB is the ratio of the combination groups to the control; A or B is the ratio of the single-drug group to the control. The CDI value <1, = 1, or >1 indicates that the combination is either synergistic, additive, or antagonistic, respectively. The statistical analysis was performed using GraphPad Prism V.6 (GraphPad Software, La Jolla, California, USA). Differences were considered to be significant at a p value below 0.05.

To examine whether KIR and its cognate HLA class I ligand combinations known to set NK cell functional threshold influence response to Isa-Len-Dex therapy, we determined the correlation between specific KIR+HLA combinations and median PFS time (MST). Nearly one-fourth of the patient cohort (24.6%; 14/57) carried the KIR3DL2+HLA-A3/11+ combination, and these patients had significantly prolonged PFS compared with the rest of the cohort lacking this combination (MST=19.35 vs 6.11 months, HR=0.43 (95% CI 0.22 to 0.88), p=0.02) (figure 2A). In contrast, patients with KIR2DL1+HLA-C2C2+ had significantly reduced PFS (MST=5.73 vs 13.33 months, HR=3.54 (95% CI 0.08 to 0.98), p=0.05) compared with the patients missing this combination (figure 2B). It is important to note that none of the 14 patients who carried the KIR3DL2+HLA-A3/11+ combination had a KIR2DL1+HLA-C2C2+ combination.

The KIR3DL2 is a framework gene and exists ubiquitously in all individuals, and thus the prolonged PFS is arguably ensured only by HLA-A3/11. To confirm the prolonged PFS is mediated by KIR3DL2+HLA-A3/11+ combination and not by HLA-A3/11 alone, we assessed the presence of other core HLA-A antigens that do not bind KIR3DL2 (eg, HLA-A1, A2, A24, A26, A29, and A68) and found no correlation between these core HLA-A antigens and PFS (online supplemental figure 4).

We performed comprehensive mechanistic studies employing in vitro cell-killing experiments to confirm the role of ADCC in inter-individual variations of efficacy to Isa treatment for MM. We used primary NK cells with distinct KIR receptor repertoire (online supplemental table 2) and MM cell lines expressing minimal HLA class I ligands (online supplemental table 1). The NK cells alone can kill MM cell lines, but the degree of killing differs substantially among effector (NK cells) and target (MM cell lines) combinations (online supplemental figure 5A,B). Such variation in NK cytolysis is presumably due to the “missing ligand” phenomenon, where the absence of HLA class I ligand from MM cell targets to which the effector NK cells were initially licensed by their cognate HLA class I ligands. Supplementing Isa in NK+MM reactions enhanced the NK cytolysis into several folds via Isa-dependent NK cytolysis mechanism (online supplemental figure 5A, C and D).

We performed in-depth mechanistic studies using in vitro cell-killing experiments to confirm the opposing effects of KIR3DL2+HLA-A3/11+ and KIR2DL1+HLA-C2C2+ combinations on PFS are arbitrated by Isa-dependent NK cytolysis. The MM cell lines that are positive or negative for HLA-A3/11 or HLA-C2C2 ligands were subjected to killing by NK cells of 11 donors with different KIR+HLA constellations in the presence of Isa mAb. Consistent with clinical survival results, MM cells expressing HLA-A3/11 (ANBL6 expresses HLA-A3, and KMS18 expresses HLA-A11) showed a significantly improved Isa-dependent NK cytolysis compared with those MM cells (SKMM2 and JK6L) that were negative for HLA-A3/11 (p<0.002, figure 2C). The MM cells expressing HLA-C2C2 (Karpas620 and ARD) were resistant to Isa-dependent NK cytolysis compared with those MM cells (AMO1, KAS61, KMS20, and XG2) that were negative for HLA-C2C2 (p<0.0001, figure 2D). These effects were independent of the varied KIR receptor repertoire of NK cells used in these ADCC assays.

Although the difference was not statistically significant, patients with KIR2DL3+HLA-C1C1+ combination showed a tendency toward an improved PFS compared with patients missing this combination (MST=9.69 vs 6.11 months, HR=0.81 (95% CI 0.42 to 1.57), p=0.55) (online supplemental figure 6A). No difference was observed in PFS with patients carrying KIR3DL1+HLA-Bw4+ combination compared with patients missing this combination (online supplemental figure 6B). The strength of KIR3DL1-mediated inhibition differs from HLA-Bw4 subtypes.34 35 The patients with strongly inhibiting ligand HLA-Bw4^I80 and KIR3DL1 combinations showed a trend towards improved PFS compared with patients carrying weakly inhibiting KIR3DL1+HLA-Bw4^T80+ combination (MST=11.99 vs 5.98 months, HR=1.81 (95% CI 0.71 to 4.60), p=0.4); however, this difference was not significant (online supplemental figure 6C). No difference was observed in PFS in patients having KIR3DL1+HLA-Aw4 (ie, HLA-A23, A24, A25, and A32) (online supplemental figure 6B) or in patients having or missing KIR2DL2+HLA-C1C1+ combinations (online supplemental figure 6D). Similarly, the activating KIR genes do not contribute to the PFS (online supplemental figure 7).

To determine if dimorphic residue 158 of FcRIIIa/CD16 influences the therapeutic outcome of Isa mAb by modulating Isa-dependent NK cytolysis, we compared the PFS between patients having high-affinity or low-affinity FCGR3A genotype. The distribution of FCGR3A-158 dimorphism in the patient cohort was 30% F/F, 28% V/F, and 7% V/V. Although not statistically significant, patients with high-affinity FCGR3A-158VV/VF genotype showed a trend of improved PFS compared with those carrying low-binding FF genotype (MST=13.93 vs 9.35 months, HR=0.55 (95% CI 0.24 to 1.27), p=0.37) (figure 3A). Next, we evaluated the functional implications of FCGR3A-158 dimorphism in Isa-dependent NK cell-mediated cytotoxicity. The presence of Isa in the in vitro killing assay increased NK cell cytolysis of MM cell line targets (figure 3B). The Isa-mediated NK cytolysis increased significantly with the presence of FCGR3A-158VV by over a factor of 3 compared with the NK cells that are homozygous for phenylalanine (p=0.0008) (figure 3B).

We also investigated the effects of other KIR genotypes and the combined effects of multiple favorable KIR+HLA combinations (online supplemental table 3). Patients carrying high-affinity FCGR3A-158V allele and KIR3DL2+HLA-A3/11+ combination showed the highest MST (MST >32 vs 6.11 months, HR=0.35 (95% CI 0.16 to 0.74), p=0.007) (figure 4A). Similarly, patients carrying KIR3DL2+HLA-A3/11+ and KIR2DL3+HLA-C1C1+ combinations showed improved PFS (MST=19.35 vs 7.39 months, HR=0.45 (95% CI 0.21 to 0.96), p=0.04) (figure 4B). Of interest, 15.8% of patients carrying centromeric BB motif or centromeric AA motif with HLA-C2C2+ ligand showed worse PFS (online supplemental table 3).