A Novel Affinity Engineered Anti-CD47 Antibody With Improved Therapeutic Index That Preserves Erythrocytes and Normal Immune Cells

Therapeutic blockade of the CD47/SIRPα axis by small molecules or monoclonal antibodies (mAbs) is a proven strategy to enhance macrophages-mediated anti-tumor activity. However, this strategy has been hampered by elevated on-target toxicities and rapid clearance due to the extensive CD47 expression on normal cells (“antigen sink”) such as red blood cells (RBCs). To address these hurdles, we report on the development of STI-6643, an affinity-engineered fully human anti-CD47 IgG4 antibody with negligible binding to normal cells. STI-6643 exhibited no hemagglutination activity on human RBCs at concentrations up to 300 µg/mL yet specifically blocked the CD47/SIPRα interaction. Of particular interest, STI-6643 preserved T cell functionality in vitro and showed significantly lower immune cell depletion in vivo in contrast to three previously published competitor reference anti-CD47 clones Hu5F9, AO-176 and 13H3. In cynomolgus monkeys, STI-6643 was well-tolerated at the highest dose tested (300 mg/kg/week) and provided favorable clinical safety margins. Finally, STI-6643 displayed comparable anti-tumor activity to the high-affinity reference clone Hu5F9 in a RAJI-Fluc xenograft tumor model as monotherapy or in combination with anti-CD20 (rituximab) or anti-CD38 (daratumumab) mAbs. These data suggest that STI-6643 possesses the characteristics of an effective therapeutic candidate given its potent anti-tumor activity and low toxicity profile.

d7 and d29 relative to first antibody infusion. For Pharmacokinetic analysis, blood was collected and processed to plasma at 0 min, 5 min, 15 min and 2, 8, 24, 72 and 168 h. Plasma STI-6643 levels were measured using a validated ELISA method with a dynamic range of 0.2 to 5 µg/mL. TK parameters were estimated using Phoenix WinNonlin v1.4 pharmacokinetic software (Certara,

USA)
For the high-dose study STI-6643 when given by IV bolus injection once weekly (on days 1, 8, 15, and 22) for a total of 4 doses to cynomolgus monkeys (n=10 per group; 5 males and 5 females) at a dose level of 0 and 300 mg/kg/dose (dose volume of 10 mL/kg for all groups). Animals underwent a 6-week recovery period before being necropsied on Day 71. Clinical pathology looking at hemoglobin concentration, RBC and lymphocyte counts were conducted between d7 and d28 relative to first antibody infusion and analyzed by the ADVIA 2120 Automated Hematology Analyzer. For coagulation and clinical chemistry blood samples were collected at -7 and 28 days relative to first antibody infusion and analyzed by the Sysmex CS-5100 Automated Coagulation Analyzer and Hitachi-7180 Automatic Clinical Analyzer, respectively. Pharmacokinetic (PK) analysis was conducted in male and female cynomolgus monkey plasma on days 1 and 15, following once weekly slow IV bolus injection of STI-6643 at dose levels of 30, 90 and 150 mg/kg/dose. Blood was collected (on days 1 and 15) and processed to plasma at 0 min (pre-dosing), 5 min, 15 min and 2, 8, 24, 72 and 168 h. Plasma STI-6643 levels were measured using a validated ELISA method with a dynamic range of 0.1 to 3.2 µg/mL. TK parameters were estimated using Phoenix WinNonlin v6.4 pharmacokinetic software (Certara, USA) using a noncompartmental approach consistent with the IV (slow bolus) route of administration. All parameters were generated from STI-6643 individual concentrations in plasma from Days 1 and 22. The primary toxicokinetic parameters Tmax, Cmax and AUC(0-168 h) were calculated using Watson LIMS 7.4.2 software with at least two plasma concentrations values above lower limit of quantification (LLOQ) and reported as Median (Min and Max), Mean ± SD and Mean ± SD, respectively. T1/2 was calculated using Phoenix WinNonlin v6.4 software and reported as Mean ± SD. SPSS Statistics 21 was used to conduct t-test analysis for gender differences. When both p value was <0.05 and exposure (Cmax and AUC(0-168 h)) ratio was outside a 0.5-2.0 range, data were considered statistically different between genders. Plasma concentration below the lower limit of quantitation (0.1 µg/mL) was considered as below the quantification limit (BQL). If the concentrations of more than half of the samples were below LLOQ at the same time point, data were excluded from the average calculation, reported as "NC."

Data and statistical analysis
Bioluminescence imaging data: Bioluminescence was quantified using the Living Image® software from PerkinElmer. The radiance (p/sec/cm 2 /sr) was set between 1.0E+05 and 5.0E+06 for each analysis and bioluminescence picture. Tumor growth data are given in total flux (p/s).
When data of at least two groups were compared over time (bioluminescence values and anti- Log-rank tests are used to assess statistical significance in survival data. All statistical tests were 2-sided, and results were considered statistically significant at P < 0.05. Outliers were identified and excluded using the ROUT method (at Q=1%) in the GraphPad Prism software.
For the dose-finding TK study: Statistical analysis was performed by using pooled males and females monkeys for each group: For data collected and/or reported in Provantis™ (in-life (clinical observations, body weights, food consumption); clinical pathology (clinical chemistry, coagulation, hematology); postmortem (organ weights, necropsy, histopathology)), Levene's test was used to assess the homogeneity of group variances parametric assumption at the 5% significance level. Datasets with at least 3 groups were compared using an overall 1-way ANOVA F-test or Kruskal-Wallis test (if parametric assumptions were not met) at the 5% significance level.
The above pairwise comparisons were conducted using a 2-sided Dunnett's or Dunn's test, respectively, if the overall test was significant. All significant pairwise comparisons were reported at the 1 and 5% significance levels. For data external to the Provantis™ system, the assumptions that permitted use of a parametric ANOVA were verified using the Shapiro-Wilkes test for normality of the data and Levene's test for homogeneity of variance, with p ≤ 0.001 level of significance required for either test to reject the assumptions. If both assumptions were fulfilled, a single-factor ANOVA was applied, with animal grouping as the factor, utilizing a p ≤ 0.05 level of significance. If the parametric ANOVA was significant at p ≤ 0.05, Dunnett's test was used to identify statistically significant differences between the control group and each test article dosed group at the 0.05 level of significance. If either of the parametric assumptions was not satisfied, the Kruskal-Wallis non-parametric ANOVA procedure was used to evaluate intergroup differences (p ≤ 0.05). The Dunn's multiple comparison test was applied if this ANOVA was significant, again utilizing a significance level of p ≤ 0.05.
For the high-dose TK study: Statistical analysis was performed by Provantis system. Comparison between test article-treated groups and the vehicle control group was performed using the following statistical methods: (1) Data within groups were evaluated for homogeneity of variance by Levene's test. For data whose variances were homogeneous (p>0.05), a one-way analysis of variance (ANOVA) was performed on the data; for nonhomogeneous data (p≤0.05), a logarithmic transformation was automatically applied to obtain log data, and a Levene's test was applied to the log data again. For log data whose variances were homogeneous (p>0.05), a one-way analysis of variance (ANOVA) was performed on the log data; for nonhomogeneous log data (p≤0.05), a rank transformation was applied on the log data to obtain rank data before Kruskal-Wallis test being