Modeling, optimization, and comparable efficacy of T cell and hematopoietic stem cell gene editing for treating hyper‐IgM syndrome

Abstract Precise correction of the CD40LG gene in T cells and hematopoietic stem/progenitor cells (HSPC) holds promise for treating X‐linked hyper‐IgM Syndrome (HIGM1), but its actual therapeutic potential remains elusive. Here, we developed a one‐size‐fits‐all editing strategy for effective T‐cell correction, selection, and depletion and investigated the therapeutic potential of T‐cell and HSPC therapies in the HIGM1 mouse model. Edited patients’ derived CD4 T cells restored physiologically regulated CD40L expression and contact‐dependent B‐cell helper function. Adoptive transfer of wild‐type T cells into conditioned HIGM1 mice rescued antigen‐specific IgG responses and protected mice from a disease‐relevant pathogen. We then obtained ~ 25% CD40LG editing in long‐term repopulating human HSPC. Transplanting such proportion of wild‐type HSPC in HIGM1 mice rescued immune functions similarly to T‐cell therapy. Overall, our findings suggest that autologous edited T cells can provide immediate and substantial benefits to HIGM1 patients and position T‐cell ahead of HSPC gene therapy because of easier translation, lower safety concerns and potentially comparable clinical benefits.

Appendix Table S1. Off-target analysis of the selected CD40LG gRNA.

Nonlinear mixed-effects model analysis of longitudinal data in Fig 5B and 5C
Due to the type of longitudinal trajectories of data in Fig 5B and 5C, the comparison of the longitudinal trend of the two groups was performed with a nonlinear mixed-effects (NLME) regression with the following asymptotic model: where the time was considered as continuous with the origin shifted at day 1 (for a better interpretation of the parameters of the model), while the dependent variable Y was transformed in the square root scale to meet the assumption of normality of the residuals. In the model, the parameter represents the horizontal asymptote (i.e. the value of the plateau reached by √ ). The parameter 0 represents the value at day 1 and the parameter is the natural logarithm of the rate constant. For evaluating differences between the two groups, in the full model, all parameters were allowed to depend on the group variable: = 0 + 1 ( ), 0 = 0 + 1 ( ), = 0 + 1 ( ) where = 1, for NC group, and = 0 for CPA 300 group. For each dependent variable, the final model was obtained with a backward variable selection of fixed-effects covariates (where pvalues less than 0.05 were considered significant). Random effects were set on the asymptote to account for mice heterogeneity. When necessary few observations were excluded from the analysis since they were outliers for the full model. For evaluating differences of IgG concentration among groups overall and, within each group, between pre and post values, a full linear mixed-effects (LME) model was estimated, followed by a post-hoc analysis with the R package phia. In each post-hoc analysis, p-values were adjusted with Bonferroni's correction to account for multiple comparisons or testing, depending on the analysis. The full LME model included the following terms: time (pre vs post), group and an interaction term between time and group. When necessary, to meet the assumptions of the LME model, an adequate transformation of the dependent variable was used (the natural logarithmic transformation for data of first boost in Fig 5F and 6E and data of Fig 7C and the square root transformation for data of second boost in Fig 6H). Random effects of the LME model were set on the intercept term and they were defined either to account only for the variability among mice in case of a single experiment (data of Fig 6H), or as nested to account for the variability among experiments and among mice within each experiment (data of Fig 5F, 6E and 7C). Only the results of post-hoc analysis are reported.
Post-hoc analysis of IgG concentration of first boost in Fig 5F: Post-hoc analysis testing the overall difference among groups: Post-hoc analysis of IgG concentration of first boost in Fig 6H: Post-hoc analysis testing the overall difference among groups: Linear mixed-effects model analysis of data in Fig 6B and 6C For evaluating differences among groups at a fixed time-point accounting for data belonging to different experiments, a linear mixed-effects (LME) model with the only group term was estimated, followed by a post-hoc analysis comparing all pairs of groups with the R package phia.
In each post-hoc analysis, p-values were adjusted with Bonferroni's correction to account for multiple comparisons. To meet the assumptions of the model, an adequate transformation of the dependent variable was used in the corresponding LME model (for data in Fig  Post-hoc analysis of n. recipient CD3+ cells at day 3\4 of Fig 6B (

Nonlinear mixed-effects model analysis of data in Fig 6D
The nonlinear relationship between n. recipient CD3+ cells at day 3\4 and n. donor CD4+ cells at day 18\24 was analyzed with a nonlinear mixed-effects (NLME) model analysis by using an asymptotic model. The standard NLME model with horizontal right asymptote, was reparametrized in the following way to enhance the interpretation of the results with respect to the aim of the analysis: where the X was n. recipient CD3+ cells at day 3\4, while the dependent variable Y (n. donor CD4+ cells at day 18\24) was transformed in the square root scale to meet the assumption of normality of the residuals. In the model, the parameter 0 represents the value at X=0 and the parameter is the natural logarithm of the rate constant. The parameter delta represents the difference between R0 and the value of the horizontal asymptote (i.e. the value of the plateau reached by √ for higher values of X). Thus, a positive value of the delta parameter denotes that a decrease of X corresponds to an increase of Y. Random effects were set on the delta parameter to account for heterogeneity among groups and among experiments. One observation was excluded from the analysis since it was an outlier for the full model. Linear mixed-effects model analysis for longitudinal data in Fig 6F, 6G, 1H, 2F, EV4A, EV1J When the nonlinear trajectories over time could not be modeled with known nonlinear mixedeffects models or data were not adequate for applying this kind of models, for evaluating differences among groups over time, a full linear mixed-effects (LME) model was estimated, which included the following terms: time (treated as categorical variable), group and an interaction term between time and group. In case of two groups, results of the comparisons at each time-points were retrieved directly from the estimated LME model. In case of more than three groups, a posthoc analysis with the R package phia was performed for testing differences between all pairs of groups at each time-point. In case of data in Fig 2F, also a post-hoc analysis for testing the overall differences among groups was performed. In each post-hoc analysis, p-values were adjusted with Bonferroni's correction to account for both multiple testing and comparisons. In this case, only the results of post-hoc analysis are reported. When necessary, to meet the assumptions of the model, an adequate transformation of the dependent variable was used in the corresponding LME model (the square root transformation for data in Fig 6F, the cubic root transformation for data in Fig 2F, the natural logarithmic transformation for RFI data in Fig 1H, RFI data in Fig EV1J, the log(x+0.01) transformation for data in Fig 6G due to the presence of zeros) and, eventually, few observations were excluded from the analysis since they were outliers for model. Random effects of the LME model were set on the intercept term and they were defined either to account only for the variability among mice (data of Fig 6F, 6G, 2F and EV4A), or as nested to account for the variability among donors and among samples with the same donor (data of Fig 1H and EV1J). Fig EV4A (

Linear mixed-effects model analysis of longitudinal data in Fig 2G and 4E
Due to the linear longitudinal trajectories of data in Fig 2G and 4E, the comparison of the longitudinal trend among groups was performed with a linear mixed-effects (LME) with time treated as a continuous variable. In order to test eventual differences among the groups, firstly a full model was estimated including the following terms: time, group and an interaction term between time and group (in case of the data of Fig 4E, the origin of the time was shifted at week 6 for a better interpretation of the parameters of the model). Then, the final model was obtained with a backward variable selection of fixed-effects covariates (where p-values less than 0.05 were considered significant). Random effects were set on the intercept to account for mice heterogeneity. In case of the analysis of the data in Fig 4E,

Nonlinear mixed-effects model analysis of longitudinal data in Fig 4D
Due to the type of longitudinal trajectories of data in Fig 4D, the comparison of the longitudinal trend among the groups was performed with a nonlinear mixed-effects (NLME) model regression with the following asymptotic model: where the time was considered as continuous with the origin shifted at week 6 (for a better interpretation of the parameters of the model). In the model, the parameter represents the horizontal asymptote (i.e. the value of the plateau reached by the dependent variable). The parameter 0 represents the value at week 6 and the parameter is the natural logarithm of the rate constant. For evaluating differences among groups, in the full model, all parameters were allowed to depend on the group variables: = 0 + 1 ( + 56)+ 2 ( + 56), 0 = 0 + 1 ( + 56) + 2 ( + 56), = 0 + 1 ( + 56) + 2 ( + 56) where + 56 = 1, for NGFR+GSE56 group, and + 56 = 0 otherwise, + 56 = 1, for NR+GSE56 group, and + 56 = 0 otherwise. The final model was obtained with a backward variable selection of fixed-effects covariates (where p-values less than 0.05 were considered significant). Random effects were set on the asymptote to account for mice heterogeneity.
Final NLME model: