Tumor-reactive monoclonal antibody (mAb) therapy is a targeted approach to cancer treatment. An increasing number of these mAbs have been clinically approved.1 However, in some patients, treatment with mAbs generates antibody responses against the therapeutic mAb.

With the use of murine, chimeric, or humanized/human mAbs, patients can generate human antimouse antibody, human antichimeric antibody (HACA), or human antihuman antibody (HAHA) responses against the therapeutic mAb.2–4 These antitherapeutic antibodies may potentially “neutralize” the therapeutic mAb via increased clearance and/or interference with the tumor-binding of the mAb. To minimize mAb immunogenicity, humanzed and fully human mAbs are now the main mAb class in clinical trials.5

A source of immunogenicity on humanized/human mAbs, distinct from the idiotype, is the antibody allotypes, the heritable allogeneic polymorphisms in the antibody constant regions.6 Most immunoglobulin (Ig)G1 mAbs (chimeric or fully human) in common clinical use express at least 1 of 4 well-described IgG1 allotypes (GM1, GM2, GM3, or GM17), and all express a common kappa light chain allotype (KM1, KM1,2, or KM3).6 In vivo exposure to an Ig with a non-self-allotype can induce antibodies against the foreign allotype.6

Dinutuximab (a chimeric anti-GD2 mAb, ch14.18), given together with interleukin 2 (IL2) and granulocyte-macrophage colony stimulating factor (GM-CSF), is the standard of care treatment for children with high-risk neuroblastoma.7 8 The SJGD2 phase I clinical trial conducted at St Jude Children’s Research Hospital (SJCRH) sought to test the efficacy of a variant anti-GD2 mAb, designated hu14.18K322A, for patients with recurrent/refractory neuroblastoma. While screening for HAHA in the phase I clinical trial of the hu14.18K322A,9 10 we found that 9 of 38 patients had antibodies against our therapeutic mAb (hu14.18K322A) prior to receiving any treatment, and thus demonstrated pretreatment antitherapeutic antibodies (PATA) recognizing the hu14.18K322A mAb. Here we will show that these PATA recognized a determinant(s) shared among other mAbs, distinct from the 14.18 idiotype. Furthermore, PATA+ patients showed improved outcome compared with PATA− patients after hu14.18K322A treatment.

PATA was detected in the pretreatment sera from 9 of 38 evaluable patients from the SJGD2 phase I study of hu14.18K322A (figure 1A). These nine PATA+ patients had no prior exposure to hu14.18K322A or any other mAb. The PATA reactivity was due to antibody reactivity against hu14.18K322A, as indicated by the IgM antibody found in patient serum (figure 1B), and this antitherapeutic reactivity remained detectable throughout the 1-month treatment course (figure 1C). This stable reactivity pattern for these nine PATA+ patients is distinct from the pattern for the PATA−, HAHA+ patients, who had HAHA detected in their serum after treatment began (figure 1D), or those patients that remained PATA− and HAHA− during therapy (figure 1D).

In prior anti-GD2 clinical trials, we noted a decrease in peak serum levels of anti-GD2 mAb in treatment courses after the detection of the HACA or HAHA reactivity.4 In contrast to our previous observations, the majority of PATA+ patients had peak anti-GD2 mAb concentrations that were higher than that detected for PATA− patients treated at comparable hu14.18K322A doses (p=0.078, figure 1E). Although not statistically significant, consistent with figure 1E, PATA+ patients had increased levels for both area under the curve and half-life, commensurate with decreased clearance-levels of Hu14.18K322A, compared with PATA− patients (online supplementary figure 1). Thus, unlike HACA/HAHA, acquired after mAb exposure, PATA do not neutralize detectible serum levels of the therapeutic mAb.

In a binding inhibition assay, PATA+ serum samples do not inhibit hu14.18K322A from binding to its binding partner (1A7 mAb), thus PATA+ samples show no anti-idiotypic activity (table 1). In contrast, serum samples from a PATA− patient who developed HAHA after hu14.18K322A exposure and a pool of serum samples from patients who developed antitherapeutic antibody after exposure to other 14.18-based anti-GD2 mAbs4 15 show strong inhibition of hu14.18K322A binding to the 1A7 mAb, demonstrating induction of anti-idiotypic antibody (table 1).

All PATA+ patients had PATA reactivity directed against the Fc component of hu14.18K322A (figure 2A), but not the Fab component. Serum from these PATA+ patients also recognized the Fc end of another mAb, rituximab. Rituximab (human anti-CD20) has the GM1,17 allotype, which is a different allotype than hu14.18K322A (GM3 allotype; online supplementary figure 2).6 Sera from eight of nine PATA+ patients bound to intact rituximab and the Fc fragment, but not the Fab fragment, of rituximab (figure 2A).

Figure 2B demonstrates that all nine PATA+ patients had significantly higher reactivity than the PATA− patients against a mouse IgG2a isotype control mAb and 14G2a (anti-GD2 mAb; mouse IgG2a Fc region). In contrast, the reactivity of sera from all PATA+ patients was not >0.7 OD cut-off against mouse IgG1 mAb isotype control or against panitumumab (anti-EGFR mAb; human IgG2 Fc region); there was no significant difference between PATA+ versus PATA− patients to recognize mouse IgG1 mAb, but there was a slight elevation in reactivity of PATA+ to panitumumab versus to PATA− patients.

Of the 38 patients enrolled in the phase I study, only 4 remained progression-free at 2.5 years after starting treatment (table 2). All 4 of these patients are in the PATA+ group of 9 patients; 0 of 29 PATA− patients were non-progressors (p=0.002; table 2). As prior studies of anti-GD2-based mAb agents suggest more likely benefit in the absence of bulky disease,16 the disease burden was evaluated qualitatively (online supplementary tables 1 and 2). The distribution of patients with high disease burden (classes 3 and 4) versus low disease burden (classes 1 and 2) was similar in PATA+ patients (5 with high burden and 4 with low burden) and PATA− patients (18 with low burden and 11 with high burden) (online supplementary tables 1 and 2). At 2.5 years after starting treatment, only the 4 PATA+ patients with low disease burden (all were class 2) had no disease progression without further therapy; all 34 other patients (including the 5 PATA+ patients with class 4 high disease burden and the 29 PATA− patients) showed progression (p<0.001; table 2).

To our knowledge, this is the first study describing an association of the presence of PATA with the improved clinical outcome of a tumor-reactive mAb trial. Although the specific epitope(s) are yet to be determined, we can rule out certain well-defined sources of mAb immunogenicity. PATA reactivity in this study cannot be accounted for by focusing only on the hu14.18K322A point mutation, as we detect PATA directed against rituximab, which does not contain this mutation. PATA is not directed against the 14.18 idiotype, as PATA did not diminish hu14.18K322A binding to 1A7 (table 1).

Human IgG1-based mAbs have well-characterized, serologically defined allotypes (four on the heavy chain and three on the kappa light chain) (online supplementary figure 2).6 Most IgG1 mAbs in clinical practice express a kappa light chain with a common allotype (KM3). Of the PATA+ patients, six of nine had evaluable DNA for genotyping; all six were found to have the KM3 allele (data not shown); they would be unlikely to make an antibody to KM3 as it would not be foreign. Further, the PATA reactivity binds primarily to the Fc region, rather than Fab, of both hu14.18K322A and rituximab (figure 2). Among the allotypes on the Fc fragment of IgG1, only one allotype (GM1) is expressed on rituximab (online supplementary figure 2). Of the six PATA+ patients (genotyped by pyrosequencing for GM allotypes), two of them were positive for GM1 (data not shown). GM1, GM2, and GM3 are not expressed on the Fc region of hu14.18K322A. Therefore, if sera from PATA+ patients recognize a region that is expressed on both hu14.18K322A and rituximab, then it must be a different allotype from the well-described allotypes found on human IgG1 (online supplementary figure 2).

None of the PATA+ patients strongly recognize the human IgG2 isotype, panitumumab, or a murine IgG1 isotype antibody. These data suggest that the epitope(s) recognized on hu14.18K322A by serum samples from PATA+ patients are not present on panitumumab or the murine IgG1 mAb tested. They also suggest that PATA reactivity against hu14.18K322A (produced in the rat myeloma cell line, YB2/0, with decreased fucosylation activity) is not directed against murine/rodent glycosylation patterning as both rituximab and panitumumab (recognized vs not recognized by PATA+) are produced in the CHO cell line, which has lower levels of glycosylation.17

Since identifying PATA in 9 of 38 patients in this trial, we considered why we had not observed PATA in past trials of anti-GD2 mAb using our bridge ELISA assays. In previous studies, our biotinylation kit recommended 20 molar excess of the biotinylation reagent. Recently, we optimized this to instead use 5–10 molar excess for our assays, resulting in adequate biotin-labeling while retaining mAb function/specificity. We confirmed that excess biotinylation interferes with allotype binding by reanalyzing samples from a COG phase III trial, ANBL0032,7 for which we had found that 0 of 126 patients evaluated demonstrated PATA in our previously used bridge ELISA assay. Using our optimized biotinylation protocol, we randomly selected 40 of these 126 patients and reassayed their pretreatment samples, and found that 3 of 40 patients (8%) demonstrated a PATA response (using the same 0.7 OD “cut-off”; data not shown). Thus, reduced biotinylation of hu14.18K322A allows for more sensitive detection of PATA than our prior assays, likely due to the excessive biotinylation causing blocking of the binding sites recognized by antiallotype mAbs.

Separately, we questioned whether the patient population we studied here was more likely to have PATA than other populations of cancer patients. In these patients with relapsed or refractory neuroblastoma, virtually all patients would have had combined, myelosuppressive chemotherapy, requiring multiple transfusions of blood products (particularly platelets) resulting in exposure to allogeneic blood products (plasma and Ig). We screened for PATA in a hu14.18K322A trial of newly diagnosed pediatric neuroblastoma patients, who had not yet had any therapeutic interventions or exposure to blood products (Trial NB2012 of SJCRH; NCT01857934).18 Of the first 32 patients enrolled in the study, none were PATA+ in their pretreatment samples; the incidence of PATA positivity in these two St Jude studies of hu14.18K322A (9/38 vs 0/32) was significantly different (p=0.003). Thus, prior exposure to allogeneic blood products may be inducing antibody responses to allogeneic epitopes shared between the therapeutic hu14.18K322A and the Igs of allogeneic blood product donors, that are seen as “foreign” by the immune systems of the PATA+ patients.

Prior studies of tumor-reactive mAbs have looked for associations of disease outcome and the detection of antitherapeutic antibodies and focused primarily on the importance of serum containing antibody to the therapeutic mAb that developed after known exposure to the therapeutic mAb. Correlations with favorable outcomes have been demonstrated,19–21 and others found outcome correlations with “anti-anti-idiotype” antibody formation.22 Yet, such correlations may be confounded as patients who form antitreatment antibodies may have stronger immune systems.23 Our findings suggest that antitherapeutic antibodies found in the serum of patients being treated with tumor-reactive mAbs may have different in vivo effects, depending on what component the antitherapeutic antibodies recognize on the tumor-reactive mAb. Our findings suggest that PATA may act in conjunction with the anti-GD2 mAb therapy, as PATA-associated benefit required hu14.18K322A treatment.

It will be helpful to define the molecular epitopes on therapeutic mAbs that are recognized by PATA. Sequencing the IgG constant regions for PATA+ versus PATA− patients may help test the hypothesis that PATA+ patients have subtle differences in their IgG Fc region compared with the Fc sequence from most therapeutic mAb, making the therapeutic mAb appear immunologically foreign to the PATA+ patients. If PATA+ patients have a better outcome when treated with tumor-reactive mAbs, then determining the mechanisms of PATA will be critical. The presence of PATA may increase the circulation time of the therapeutic mAb by extending its half-life, as figure 1E and online supplementary figure 1 suggest. Alternatively, PATA “cross-linking” of tumor-bound mAb on the tumor may enhance tumor killing via: (a) inducing tumor cell apoptosis24; (b) increasing mAb avidity to the antigen target; or (c) increasing the Fc density at the cell surface, thereby augmenting the antibody-dependent cell-mediated cytotoxicity or complement-dependent cytotoxicity potential (online supplementary figure 3).25

Several tumor-reactive mAbs show clear clinical efficacy. Identification of newer mAb targets on tumors will likely generate a clinical armamentarium of therapeutic mAbs, allowing for recognition of virtually any cancer type. Although some patients respond to tumor-reactive mAb-based treatments, many still do not. To increase the efficacy of mAb therapeutics, their mechanisms of action and interface with the immune system must be rigorously studied. The findings presented here support the possibility that tumor-reactive mAbs may interact with endogenous non-neutralizing serum antibodies in a manner augmenting antitumor effects. Further clarification of this phenomenon and the underlying mechanisms might provide leads for the development of more effective mAbs and mAb-based treatment regimens.