Base and prime editors (BEs and PEs) may provide more precise genetic engineering than nuclease-based approaches because they bypass the dependence on DNA double-strand breaks. However, little is known about their cellular responses and genotoxicity. Here, we compared state-of-the-art BEs and PEs and Cas9 in human hematopoietic stem and progenitor cells with respect to editing efficiency, cytotoxicity, transcriptomic changes and on-target and genome-wide genotoxicity. BEs and PEs induced detrimental transcriptional responses that reduced editing efficiency and hematopoietic repopulation in xenotransplants and also generated DNA double-strand breaks and genotoxic byproducts, including deletions and translocations, at a lower frequency than Cas9. These effects were strongest for cytidine BEs due to suboptimal inhibition of base excision repair and were mitigated by tailoring delivery timing and editor expression through optimized mRNA design. However, BEs altered the mutational landscape of hematopoietic stem and progenitor cells across the genome by increasing the load and relative proportions of nucleotide variants. These findings raise concerns about the genotoxicity of BEs and PEs and warrant further investigation in view of their clinical application.

Targeted gene editing in hematopoietic stem cells (HSCs) is a promising treatment for several diseases. However, the limited efficiency of homology-directed repair (HDR) in HSCs and the unknown impact of the procedure on clonal composition and dynamics of transplantation have hampered clinical translation. Here, we apply a barcoding strategy to clonal tracking of edited cells (BAR-Seq) and show that editing activates p53, which substantially shrinks the HSC clonal repertoire in hematochimeric mice, although engrafted edited clones preserve multilineage and self-renewing capacity. Transient p53 inhibition restored polyclonal graft composition. We increased HDR efficiency by forcing cell-cycle progression and upregulating components of the HDR machinery through transient expression of the adenovirus 5 E4orf6/7 protein, which recruits the cell-cycle controller E2F on its target genes. Combined E4orf6/7 expression and p53 inhibition resulted in HDR editing efficiencies of up to 50% in the long-term human graft, without perturbing repopulation and self-renewal of edited HSCs. This enhanced protocol should broaden applicability of HSC gene editing and pave its way to clinical translation.

Ex vivo gene editing in T cells and hematopoietic stem/progenitor cells (HSPCs) holds promise for treating diseases. Gene editing encompasses the delivery of a programmable editor RNA or ribonucleoprotein, often achieved ex vivo via electroporation, and when aiming for homology-driven correction of a DNA template, often provided by viral vectors together with a nuclease editor. Although HSPCs activate a robust p53-dependent DNA damage response upon nuclease-based editing, the responses triggered in T cells remain poorly characterized. Here, we performed comprehensive multiomics analyses and found that electroporation is the main culprit of cytotoxicity in T cells, causing death and cell cycle delay, perturbing metabolism, and inducing an inflammatory response. Nuclease RNA delivery using lipid nanoparticles (LNPs) nearly abolished cell death and ameliorated cell growth, improving tolerance to the procedure and yielding a higher number of edited cells compared with using electroporation. Transient transcriptomic changes upon LNP treatment were mostly caused by cellular loading with exogenous cholesterol, whose potentially detrimental impact could be overcome by limiting exposure. Notably, LNP-based HSPC editing dampened p53 pathway induction and supported higher clonogenic activity and similar or higher reconstitution by long-term repopulating HSPCs compared with electroporation, reaching comparable editing efficiencies. Overall, LNPs may allow efficient and harmless ex vivo gene editing in hematopoietic cells for the treatment of human diseases.

Glioblastoma multiforme (GBM) is the most common and lethal brain tumor characterized by a strongly immunosuppressive tumor microenvironment (TME) that represents a barrier also for the development of effective immunotherapies. The possibility to revert this hostile TME by immunoactivating cytokines is hampered by the severe toxicity associated with their systemic administration. Here, we exploited a lentiviral vector–based platform to engineer hematopoietic stem cells ex vivo with the aim of releasing, via their tumor-infiltrating monocyte/macrophage progeny, interferon-α (IFN-α) or interleukin-12 (IL-12) at the tumor site with spatial and temporal selectivity. Taking advantage of a syngeneic GBM mouse model, we showed that inducible release of IFN-α within the TME achieved robust tumor inhibition up to eradication and outperformed systemic treatment with the recombinant protein in terms of efficacy, tolerability, and specificity. Single-cell RNA sequencing of the tumor immune infiltrate revealed reprogramming of the immune microenvironment toward a proinflammatory and antitumoral state associated with loss of a macrophage subpopulation shown to be associated with poor prognosis in human GBM. The spatial and temporal control of IL-12 release was critical to overcome an otherwise lethal hematopoietic toxicity while allowing to fully exploit its antitumor activity. Overall, our findings demonstrate a potential therapeutic approach for GBM and set the bases for a recently launched first-in-human clinical trial in patients with GBM.

Immune escape represents a major driver of acute myeloid leukemia (AML) reemergence after allogeneic hematopoietic cell transplantation (allo-HCT), with up to 40% of relapses prompted by nongenomic loss of HLA class II expression in leukemia cells. By integrative analysis of gene expression, DNA methylation, and chromatin accessibility in paired diagnosis/relapse primary samples and in the respective patient-derived xenografts (PDX), we identify the polycomb repressive complex 2 (PRC2) as a key epigenetic driver of this immune escape modality. We report that loss of expression of HLA class II molecules is accompanied by a PRC2-dependent reduction in chromatin accessibility. Pharmacologic inhibition of PRC2 subunits rescues HLA class II expression in AML relapses in vitro and in vivo, with consequent recovery of leukemia recognition by CD4+ T cells. Our results uncover a novel link between epigenetics and leukemia immune escape, which may rapidly translate into innovative strategies to cure or prevent AML posttransplantation relapse. Significance: Loss of HLA class II expression represents a frequent mechanism of leukemia posttransplantation relapse. Here we identify PRC2 as the main epigenetic driver of this immune escape modality and show that its chemical inhibition can reinstate a proficient graft-versus-leukemia effect, providing an innovative rationale for personalized epigenetic immunotherapies. See related commentary by Köhler and Zeiser, p. 1410. This article is highlighted in the In This Issue feature, p. 1397

One of the most challenging tasks when adopting Bayesian networks (BNs) is the one of learning their structure from data. This task is complicated by the huge search space of possible solutions and by the fact that the problem is NP‐hard. Hence, a full enumeration of all the possible solutions is not always feasible and approximations are often required. However, to the best of our knowledge, a quantitative analysis of the performance and characteristics of the different heuristics to solve this problem has never been done before. For this reason, in this work, we provide a detailed comparison of many different state‐of‐the‐art methods for structural learning on simulated data considering both BNs with discrete and continuous variables and with different rates of noise in the data. In particular, we investigate the performance of different widespread scores and algorithmic approaches proposed for the inference and the statistical pitfalls within them.