Induction of Ab3 and Ab3′ Antibody Was Associated with Long-Term Survival after Anti-G_D2 Antibody Therapy of Stage 4 Neuroblastoma¹

MoAbs³ can effect tumor kill by activating complement-mediated plus cell-mediated antibody-dependent cytotoxicities. Antitumor antibodies belonging to subclasses such as mouse IgG2a, mouse IgG3, or mouse-human IgG1 chimeric have been used successfully in clinical trials (1). Although measurable antitumor effects were modest,occasional patients were noted to have unexpectedly prolonged survival. Often, HAMA response was observed when patients were treated with murine antibodies. Although some of the HAMA response was directed at the constant regions of the murine IgG, a portion of the response was specific for the variable regions bearing unique antigenic determinants called idiotopes. Anti-idiotypic (anti-id) antibodies can potentially induce a human anti-anti-idiotypic response (2). Those anti-ids that recognize idiotopes within the framework region of the immunizing monoclonal antibodies are termed Ab2α. Of interest are anti-ids (Ab2β) that recognize the antigen-binding site of the immunizing MoAb and mimic the original antigen. Because of this mimicry, they form part of a self-regulating network (3). Vaccination with anti-id has induced protective immunity against viral (4), bacterial (5), and parasitic(e.g., trypanosomiasis) infections (6). Anti-id vaccines that mimic carbohydrate or glycolipid antigens have shown immunological advantages over the natural antigen (7, 8):(a) unlike natural carbohydrate antigens, anti-id vaccines can stimulate cellular in addition to humoral immune response (9); and (b) anti-id vaccines may be more immunogenic than the non-protein antigens they mimic. Several anti-ids have been raised against MoAbs recognizing cell surface tumor targets. Immunization of animals with these anti-ids can generate antibodies(i.e., Ab3′) that recognize the original tumor antigen (10, 11, 12, 13, 14, 15, 16); in some reports, specific T-cell-mediated immunity was also found (13, 17, 18, 19, 20). The potential role of an anti-idiotypic network in colon cancer was first described in patients treated with monoclonal antibody CO-17-1A (21). In subsequent studies, Ab2 and Ab3 (8, 22, 23, 24, 25, 26) and T-cell immunity (13, 18, 20, 27) were demonstrated in responding patients.

Previously, we reported 34 patients with high-risk metastatic neuroblastoma treated with anti-G_D2 monoclonal antibody 3F8 at the time of minimal disease. Thirteen patients remained progression free 4–12 years after the initiation of 3F8 treatment. A self-limiting HAMA response was found to be associated with long-term survival (28). No late (after 4 years) relapse was seen. We postulate that an idiotype network might be responsible for the long-term tumor control. Natural antitumor antibodies have been described previously (29, 30), which correlated strongly with clinical outcome in patients with neuroblastoma. The antigen specificities include HUD (29) and a M_r 260,000 molecular weight cell membrane protein (31). HUD is a nuclear antigen found in 78% of human neuroblastomas, and anti-HUD antibody was present in a small subset of patients. Among patients with small cell lung cancer,those with anti-HUD antibodies typically present with autoimmune paraneoplastic encephalomyelitis and/or paraneoplastic sensory neuronopathy, and these patients had longer survival. If 3F8 enhances tumor antigen processing or presentation, one might expect an enhanced antibody response to HUD, which in turn will have a positive effect on survival. In this report, we measure serum Ab3 (anti-anti-idiotype),Ab3′ (anti-antigen), antimouse IgG3, and anti-HUD antibody responses and study their impacts on patient survival.

Thirty-four patients received 3F8 between 1987 and 1995 after treatment of their stage 4 neuroblastoma with chemotherapy and/or radiation(Table 1). Thirteen patients (group I)were either in early relapse, second or subsequent remission, or with stable microscopic disease; 21 patients (group II) were in first remission and were treated with N6 chemotherapy, of whom 17 received only N6, whereas 4 had some additional chemotherapy prior to N6 (28). Informed consent for 3F8 treatment was obtained in accordance with Institutional Review Board guidelines. All 13 patients in group I had failed at least one chemotherapy regimen, and 11 had progressive or persistent disease after myeloablative therapies. Their first relapses occurred at a median of 20 months from diagnosis (range,7–88 months). Two patients were outliers, with biopsy-proven skeletal relapse at 82 and 88 months from initial diagnosis. Eleven of these 13 patients achieved remission after reinduction therapy. 3F8 was administered at a median of 8.6 months (range, 1.4 to 18.6 months) from the time of relapse for these group I patients. For patients in first remission (group II), 3F8 treatment was initiated at a median of 9.2 months (range, 6.7–14.6 months) from diagnosis after they finished induction therapy. Because all 34 patients were treated with 3F8 at a time of near-complete remission and their long-term survivals were comparable, they were analyzed as one group in this report.

Each course of treatment consisted of 10 mg/m²3F8 i.v. daily over 90 min. The duration of treatment was 5 days per course for the initial nine patients in group I, and 10 days (2 consecutive weeks) for the rest of the patients. Patients were premedicated with i.v. antihistamine (usually diphenhydramine or hydroxyzine), followed by i.v. bolus of morphine sulfate or Dilaudid plus a continuous infusion of opioids (e.g., morphine sulfate at 0.1 mg/kg/h). MoAb 3F8, packaged in 2% human serum albumin,was diluted to a final volume of 70 ml with 5% human serum albumin in normal saline. Ten ml of 3F8 solution were infused over 30 min, 20 ml over the next 30 min, and the final 40 ml over the last 30 min. Morphine sulfate infusion was generally stopped 15 min before completion of 3F8 infusion. Additional doses of morphine sulfate(0.1 mg/kg i.v.) were used as needed for pain, diphenhydramine (1 mg/kg) for urticaria, and s.c. epinephrine 1:1000 (0.01 ml/kg; maximum,0.3 ml) for angioedema. After each course of 3F8 treatment, HAMA titer was in general measured every 2–4 weeks during the first 2 months and periodically afterward. In the absence of progressive disease, patients were retreated with 3F8 if their HAMA titer was <1000 units/ml. The median intervals between each of the four 3F8 courses were 2.1, 2.8,and 1.5 months, respectively. The median length of overall 3F8 treatment (for those who received more than one cycle) was 5.4 months.

Ninety-six-well, flat-bottomed polyvinyl microtiter plates (Dynex Technologies, Chantilly, VA) were coated with antigen ganglioside G_D2 (Advanced Immunochemical, Long Beach, CA) at 20 ng/well. The purity of G_D2 was >95%. Control wells for serum background subtraction were not coated with antigen. One hundred μl/well of 0.01% gelatin (Sigma Chemical Co., St. Louis,MO) in PBS was used as filler protein to saturate unbound sites for 1 h at room temperature. After washing with PBS (plates were washed three times with PBS between steps), patient sera (50 μl) in duplicates were serially diluted in 0.03% BSA in PBS and were allowed to react with G_D2-coated plates for 2.5 h at 37°C. A standard curve was constructed using serial dilutions of human-mouse chimeric 3F8. After being washed with PBS, a peroxidase-conjugated affinity purified goat-anti-human IgG (heavy and light chains) antibody (Jackson Immunoresearch, West Grove, PA) diluted(1:1000 to 0.8 μg/ml) in 0.5% BSA in PBS was added to the wells. After a 1-h incubation at 4°C and washing, a color reaction was performed at room temperature using hydrogen peroxide as substrate and o-phenylenediamine (Sigma) as chromogen. Upon stopping the reaction with 30 μl of 5 n sulfuric acid, the microtiter plates were analyzed at 490 nm using MRX microplate reader(Dynex Technologies). Based on the titration curves using human-mouse chimeric 3F8, anti-G_D2 IgG (Ab3′) titer was calculated in units/ml of binding activity (each unit equivalent to 10 ng of chimeric 3F8). Because there was reactivity with light chains, we could not rule out a low level of IgM in our measurements. Ab3′ after 3F8 was expressed both as a ratio of pre-3F8 levels and as increments of Ab3′ over pretreatment levels. Conclusions drawn from either calculation were similar. The day-to-day variability of these assays was ±20%. Experiments were generally repeated once with similar conclusions.

Six rat anti-3F8-ids (A1G4, idio2, C4E4, A2A6, C2H7, and C2D8; Ref. 32) were digested using immobilized ficin (Pierce,Rockford, IL) according to the manufacturer’s instructions. The size of the fragment was confirmed by SDS-PAGE. Retention of binding of the antibody F(ab′)₂ fragment to 3F8 was measured by ELISA. Completeness of the digestion was assayed using peroxidase-conjugated mouse antirat IgG secondary antibodies, specific for F(ab′)₂ and Fc (data not shown).

A mixture of equal amounts of anti-idiotype F(ab′)₂ at 100 μg/ml was used to coat 96-well,flat-bottomed polyvinyl microtiter plates at 50 μl/well. Control wells without antigen were used for serum background subtraction. After incubation at 37°C for 1 h, the antigen was removed. One hundredμl of 0.01% gelatin (Sigma) in PBS were added as filler protein for 1 h at room temperature and then washed three times with PBS. Patient sera in duplicate serial dilutions were preincubated in diluent containing 50 μg/ml of 2E6 (a rat IgG1 specific for the Fc portion of 3F8) in 0.03% BSA in PBS for 1 h at 37°C to inhibit human antibodies directed against non-3F8-idiotypic determinants. Serum samples were then added at 50 μl/well to the microtiter plates. A standard curve was constructed using a human-mouse chimeric 3F8. After a 2.5-h incubation at 37°C and PBS washing, a peroxidase-conjugated goat antihuman IgG (H+L; Jackson Immunoresearch Laboratories) was added at 100 μl/well and allowed to incubate at 4°C for 1 h. Upon washing, color reaction was carried out. Absorbance of the wells were then read using an MRX microplate reader, and antibody titer,calculated in units/ml Ab3 after 3F8 treatment, was expressed as a ratio relative to pretreatment levels. When expressed as increments of Ab3 over pretreatment levels, similar conclusions were found.

Ten μg/ml of NS.7, a mouse IgG3 myeloma (American Type Culture Collection, Rockville, MD) was used to coat microtiter plates at 50μl/well. Patient sera were diluted in 0.03% BSA in PBS, and the anti-IgG3 titer was assayed using ELISA, as described previously (28). Recombinant HUD protein, kindly provided by Dr. Josef Dalmau (Department of Neurology, Memorial Sloan-Kettering Cancer Center, NY), was used as antigen in coating microtiter plates. A high-titer human anti-HUD serum (gift of Dr. J. Dalmau) was used for constructing the standard curves. Anti-HUD titer was calculated in units/ml based on the standard.

The end points for this analysis were PFS and overall survival. PFS was defined as the interval from first 3F8 treatment to progression, or last follow-up. Overall survival was defined as the interval from first 3F8 treatment to death, or last follow-up. The score test from the proportional hazards model was used to determine an association between the PFS or overall survival rates and Ab3 or Ab3′ titers (both titers being continuous value assay measurements). Patients were deemed evaluable if the length of their PFS or overall survival exceeded the time period when antibody titers were measured (i.e., 6 and 14 months, respectively). The test of significance was based on a permutation distribution of the score statistics. Permutation procedure was applied because of the small number of subjects in this study (33).

The ages of the 34 patients (20 males and 14 females) ranged from 1.1 to 7.5 years at diagnosis and from 3.2 to 15.2 years at the time of 3F8 treatment (Table 1). At diagnosis, distant metastases to bones were present in 29 patients, and 31 patients had bone marrow disease. Most patients had high-risk features: unfavorable histology (16 of 17),serum lactate dehydrogenase >1500 units/ml (9 of 24), serum ferritin>142 ng/ml (20 of 23), and MYCN >10 copies (7 of 25). Treatment with 3F8 was initiated 1.6–9.1 (median, 3.4) months after the last course of chemotherapy. Patients received a total of 5–40 days (50–400 mg/m²) of 3F8. Fourteen patients are alive, and 13 (1.8–7.4 years at diagnosis) are progression-free (53–143 months from the initiation of 3F8 treatment) without further systemic therapy.

Sera collected around 6 and 14 months after the initial 3F8 treatment were assayed for anti-G_D2 antibody (Ab3′)activity by ELISA. Median Ab3′ at 6 months was 40% above pretreatment levels (i.e., in half of the patient group, Ab3′ was ≥40%above pretreatment level; range, 40–1450%). When the median is used as a cutpoint, patients with Ab3′ above the median had a significantly better long-term PFS than those below the median (71%versus 21%, P = 0.02; Table 2; Fig. 1,A); overall survival was also significant (59% versus 25%, P = 0.04;Table 2; Fig. 1,B). Similar PFS and overall survival trends(P = 0.21 and 0.16, respectively) were found for the 14 months post 3F8 titers (Table 2). Increasing the Ab3′ threshold at either 6 or 14 months to 300% above pre-3F8 levels increased the separation between the high versus low Ab3 groups for both PFS and survival curves (data not shown).

Sera at 6 months and 14 months after the initial 3F8 treatment were also assayed for Ab3 by ELISA. The median Ab3 titer at 6 months was 1.2(i.e., a 20% increment above pretreatment levels). Patients with Ab3 above the median had better PFS (64% versus 29%, P = 0.06; Fig. 2,A; Table 2) and survival(59% versus 25%, P = 0.03; Fig. 2,D; Table 2). At 14 months, the median Ab3 was 1.5, and the correlation of Ab3 and outcome was also significant for both PFS (90%versus 36%, P = <0.01) and overall survival (65% versus 19%, P = 0.04). As one might expect, stepwise increase in the cutpoint of Ab3 in the statistical analysis (e.g., increasing the Ab3 threshold from 1.2 to 1.9 in Fig. 2) widened the separation in the PFS and survival curves between the Ab3-high and Ab3-low groups. Using the permutation test based on score statistics from the Cox proportional hazard model, the Ab3 level was a strong predictor of survival,especially if measured at 14 months from first 3F8 treatment (Table 3). On the other hand, Ab3′ was statistically less significant.

Serum antimouse IgG3 activity in the 6- and 14-month sera were measured using an IgG3 myeloma NS.7 that does not bear the 3F8 idiotype. Using median titer as the threshold, the PFS and survival curves were not significantly different between the high- and low-titer groups (Table 2). The 6-month sera were then assayed for general antitumor activity using HUD as the target antigen. Previous studies have shown that patients with neuroblastoma or small cell lung cancer developed anti-HUD antibodies that appeared to correlate with survival (29, 34) Again, the PFS and survival durations between the high- and low-titer groups were not significantly different (Fig. 3; Table 2).

Our data suggest that Ab3 (anti-anti-idiotypic) and Ab3′(anti-G_D2) responses were induced in patients after treatment with anti-G_D2 Ab1 antibody 3F8. More importantly, Ab3 and Ab3′ titers correlated significantly with PFS and survival. This is consistent with the idiotype network hypothesis,where elevations of Ab3 and Ab3′ titers may have a causal relationship with long-term tumor control. We further demonstrated that the rise in Ab3 and Ab3′ antibody titers was not merely part of a global immune recovery, typically seen in patients after cessation of chemotherapy. Instead, a shift of the immune repertoire toward G_D2 and the 3F8-idiotype was necessary for prolonged survival. Thus, immune responses in survivors directed at irrelevant epitopes, such as anti-mouse IgG3 or anti-HUD antibody, did not correlate with clinical outcome. If the idiotype network was indeed important for long-term tumor control, one would expect some dose-response relationship, i.e., higher Ab3 or Ab3′ titer correlated with better outcome. Such relationships were observed in our analyses (e.g., Fig. 2). Although Ab3 titers were measured within the first 14 months in this study, such titers appeared to persist for years in the few patients we studied (data not shown). Because these patients have remained in remission from 4 to 12 years since, one might hypothesize that such titers may be responsible for long-term tumor control. Previously, we observed that only the pattern II response (i.e., transient low HAMA/Ab2 titer) was associated with superior clinical outcome (28). We interpret these findings to mean that high-titer Ab2 was not inducive to idiotype network formation. Furthermore, the initiation of the network appeared relatively early after antibody treatment. Because the number of cycles of 3F8 was also associated with good outcome (28), we hypothesize that a potential strategy in anti-G_D2 immunotherapy might be to continue to measure Ab3′ and Ab3 responses after each subsequent cycle of 3F8 until the ratio of post/pre levels above a threshold level is reached.

Our observation of the detrimental effect of high-titer HAMA/Ab2 has potential implications in the mechanism of the idiotype network. In contrast to patients with no HAMA/Ab2 response or those with persistently high titer, patients with self-limited HAMA/Ab2 response had significantly better PFS. HAMA formation may reflect an intact immune surveillance system, indirectly responsible for tumor control. However, general measures of global immune function do not correlate with tumor control among patients with neuroblastoma (35). More importantly, patients with the highest persistent HAMA response were generally not able to benefit from 3F8 treatment compared with those with the self-limited HAMA titer. In our analysis, antimouse and anti-HUD antibody titers (as a general measure of immune function) did not correlate with outcome. The alternative explanation invokes the idiotypic network of Jerne (36) that takes into account this paradoxical dose relationship between high levels of HAMA and failure to mount a substantial Ab3′ or Ab3 response. Because G_D2 is a self-antigen, one can hypothesize that significant rises in Ab3 or Ab3′ would not be possible unless suppressor pathways are removed and naive T or B cells are allowed to repopulate. After intensive chemotherapy that eliminates a large part of the lymphoid system, exposure to tumor-selective Ab1s that induce unique Ab2s may bias the recovering repertoire toward the specific antigen network. One can speculate that a weak HAMA/Ab2 response may be an indirect measure of a relatively “vacant” immune system. A high HAMA/Ab2 titer, on the other hand, may reflect a healthier but saturated lymphoid system with less chance for antigens to create a repertoire bias. The potential for biasing the immune system toward specific antigens has been well documented in murine models (37) and human disease states (38, 39). If true, one would expect such idiotype network to be successful only after intensive immuno-/myelosuppressive therapy. We have reported previously the association of Ab3 with prolonged survival among patients after autologous bone marrow transplantation (40). In other words, such network will not be easily induced if fully immunocompetent patients are treated with Ab1. An alternative explanation may lie in the quality of the HAMA response. In some murine models, only IgM anti-ids can induce a significant Ab3 response, whereas IgG anti-id was suppressive (41). It was postulated that in a normal antibody response, dominant B-cell clones(Ab2) would preempt Ab3 response against themselves by early switching from IgM to IgG secretion, before the immunogenic IgM Ab had time to activate anti-idiotypic B cells. After treatment with 3F8, two patterns of HAMA were noted (28). The pattern II HAMA response (low titer and transient) might represent an IgM anti-idiotype response,whereas the high and persistent pattern III HAMA might represent an IgG anti-idiotype response. This model could also explain why in most clinical studies of monoclonal antibodies, such an anti-idiotype network was not observed. Only in patients with heavy prior chemotherapy would T cells be significantly depleted such that class switch (IgM to IgG) would not occur. One can speculate that a protective idiotype network may be possible in patients receiving MoAbs if they were optimally timed with chemotherapy. Alternatively, immune modulators to prevent Ab2 class switch may improve the idiotype network. The suppressive effect of IgG anti-id was consistent with our findings in animals models, where the optimal dose range for anti-G_D2 antibody induction was unexpectedly low,⁴ suggesting that anti-ids can be tolerogenic when administered at conventional/high doses. Clearly, this could be detrimental if the host antitumor response contributed to tumor control.

It is noteworthy that the clinical benefit from Ab1 treatment correlated with antibody titer of Ab3 or Ab3′. As one might expect, the higher the titer at 6 or 14 months resulted in better outcome. For G_D2, this is equivalent to breaking self-tolerance by tipping the idiotype network balance toward Ab3′ and possibly higher orders of network antibodies. More importantly, this network of antibodies must be sufficiently self-regulated so that clinical signs or symptoms of autoimmunity would not appear, even when tumor control was successful. It is of interest that the timing of the Ab3′/Ab3 peak coincided with a declining or undetectable HAMA/Ab2 titer. A trivial explanation was the interference by Ab2 in Ab3′ or Ab3 assays, such that Ab3′ or Ab3 always mirrored the rise and the fall of Ab2. However, there were patients who had undetectable or declining HAMA/Ab2, as well as undetectable Ab3′ or Ab3 response. We like to interpret this to mean that the disappearance of Ab2, although conducive, is not sufficient for an Ab3′/Ab3 response. It is possible that strategies to break tolerance at the peak of the Ab2 response may facilitate Ab3′/Ab3 formation and render clinical use of anti-G_D2 antibodies more effective.

The association of anti-idiotypic response and improved outcome has been described in other tumor systems, e.g., 17-1A for colon cancer (25, 26, 42, 43, 44) and Ca125 for ovarian cancer (45, 46, 47, 48), although the paradoxical dose relationships of high HAMA/Ab2 and outcome have not been described previously. It is also possible that Ab1 was bound to free antigen and became complexed with HAMA, thereby allowing the antigen to be presented more efficiently to the immune system. Clearly, for these complexes to form,significant amounts of free antigens had to be available in circulation. Our patients were in complete or near-complete remission and generally had no detectable circulating G_D2antigens. Alternatively, the idiotype network could be responsible for the induction of anti-G_D2 or antitumor immune response (a subset of the Ab3), which may play a key role in long-term tumor control. In our study, the correlation of Ab3′ with outcome was generally not as good as Ab3, especially by 14 months after 3F8 treatment. It is possible that Ab3 represents an indirect measure of an antitumor repertoire broader than anti-G_D2 (Ab3′)alone, thus providing a better association with clinical outcome. Alternatively, because G_D2 is expressed on tumors and some normal tissues, the Ab3′ level is dependent on the tumor load(“acting as a sink”) and antibody avidity (which affects clearance), whereas Ab3 (most of which is not tumor-binding) is free to circulate and thereby easier to quantify. Furthermore, in our Ab3 assay, a panel of six anti-idiotypic specificities were used. It is possible that only certain idiotopes (as defined by anti-idiotypic antibodies) will be correlated with patient outcome, thereby providing a useful roadmap in our choice of anti-idiotypic vaccines. Further analysis will be needed to identify these protective epitope(s).