Immunotoxin Activity Against B-Lineage Leukemia and Lymphoma Is Exposure Time Dependent: Implications for Trial Design

Objective: The Pseudomonas exotoxin (PE)-based anti-CD22 recombinant immunotoxin (RIT) Moxetumomab pasudotox (HA22) has a response rate of 86% in hairy cell leukemia (HCL) (Kreitman et al 2012), but only about 30% of children with multiply relapsed ALL respond to HA22 treatment (Wayne et al AACR 2014). In contrast, HA22 is active vs primary blasts from 24/35 (68%) children with ALL in vitro (Mussai et al 2010). The apparent discrepancy between clinical response rates and in vitro cytotoxicity against ALL might be explained by the short serum half-life observed in clinical trials (approximately 1 hour). We hypothesized that blasts need to be exposed to RIT for longer in order to die. We attempted to improve activity against B-cell malignancies with a modified immunotoxin, LMB-11, which is missing domain II, has an Fab instead of a dsFv, and has 7 mutations in domain III. These modifications have been associated with pre-clinical improvements including reduced side effects, prolonged half-life, and diminished immunogenicity. LMB-11 achieved sustained CR in a Burkitt lymphoma animal model (Bera et al 2014). In the current study, we used LMB-11 and HA22 to evaluate activity against ALL and to try to identify methods to maximize in vitro cytotoxicity with the goal to increase clinical response rates in future clinical trials.

Methods: Cytotoxicity was assessed with growth inhibition assays (WST-8). ALL cells were incubated with RIT for various times, washed, re-plated for a total of 72 h, stained for apoptosis by Annexin V-PE/7-AAD, and cell death analyzed by flow cytometry. NSG mice were inoculated with 5 million ALL cells IV on day 1 and treated with RIT from day 8 either IV or by continuous infusion with ALZET osmotic pumps implanted into the peritoneal cavity. Bone marrow (BM) samples were analyzed by flow cytometry for ALL infiltration and apoptosis.

Results:In vitro cytotoxicity (IC50) of cell lines varied 210-fold (0.02 to 4.6 ng/ml). The time that cells had to be exposed to RIT to kill >90% of the cells varied greatly from <30 minutes to >4 days. We observed similar exposure time variability on 8 primary ALL samples. Despite this variability, for 57% (8/14) of all samples tested, the two toxins HA22 and LMB-11 were similar in the time to induce cell death. For 38% (5/14) of samples, HA22 was up to 10-fold more active and killed more rapidly than LMB-11. One example was KOPN-8, which was killed after less than 30 minutes exposure to HA22, but needed more than 9 hours exposure to LMB-11 for similar effects. The difference in exposure time dependent killing between the two toxins was related to intracellular RIT-processing. Using two newly developed ALL xenograft models with KOPN-8 or HAL-01, we found that response in vivo correlated directly with the time needed to kill the cells in vitro. Because RITs have a short serum half-life in mice (<30 minutes) but killing of KOPN-8 or HAL-01 cells needed 9 h or 24 h exposure to LMB-11, we replaced bolus dosing with small doses given at frequent intervals or with continuous infusion. The change in dosing strongly improved response of both KOPN-8 and HAL-01 xenografts (table 1).

Conclusion: The previously unrecognized high variability of the time a CD22-targeting immunotoxin requires to kill ALL blasts provides a potential explanation for differences between in vitro cytotoxicity and the response rates observed in clinical trials. Our data indicate that ALL patients might show improved responses to anti-CD22 immunotoxins if treated with continuous infusion rather than the currently used bolus injection.