GPT-4 as a biomedical simulator.

BACKGROUND: Computational simulation of biological processes can be a valuable tool for accelerating biomedical research, but usually requires extensive domain knowledge and manual adaptation. Large language models (LLMs) such as GPT-4 have proven surprisingly successful for a wide range of tasks. This study provides proof-of-concept for the use of GPT-4 as a versatile simulator of biological systems. METHODS: We introduce SimulateGPT, a proof-of-concept for knowledge-driven simulation across levels of biological organization through structured prompting of GPT-4. We benchmarked our approach against direct GPT-4 inference in blinded qualitative evaluations by domain experts in four scenarios and in two quantitative scenarios with experimental ground truth. The qualitative scenarios included mouse experiments with known outcomes and treatment decision support in sepsis. The quantitative scenarios included prediction of gene essentiality in cancer cells and progression-free survival in cancer patients. RESULTS: In qualitative experiments, biomedical scientists rated SimulateGPT's predictions favorably over direct GPT-4 inference. In quantitative experiments, SimulateGPT substantially improved classification accuracy for predicting the essentiality of individual genes and increased correlation coefficients and precision in the regression task of predicting progression-free survival. CONCLUSION: This proof-of-concept study suggests that LLMs may enable a new class of biomedical simulators. Such text-based simulations appear well suited for modeling and understanding complex living systems that are difficult to describe with physics-based first-principles simulations, but for which extensive knowledge is available as written text. Finally, we propose several directions for further development of LLM-based biomedical simulators, including augmentation through web search retrieval, integrated mathematical modeling, and fine-tuning on experimental data.

Removal of innate immune barriers allows efficient transduction of quiescent human hematopoietic stem cells.

Quiescent human hematopoietic stem cells (HSC) are ideal targets for gene therapy applications due to their preserved stemness and repopulation capacities; however, they have not been exploited extensively because of their resistance to genetic manipulation. We report here the development of a lentiviral transduction protocol that overcomes this resistance in long-term repopulating quiescent HSC, allowing their efficient genetic manipulation. Mechanistically, lentiviral vector transduction of quiescent HSC was found to be restricted at the level of vector entry and by limited pyrimidine pools. These restrictions were overcome by the combined addition of cyclosporin H (CsH) and deoxynucleosides (dNs) during lentiviral vector transduction. Clinically relevant transduction levels were paired with higher polyclonal engraftment of long-term repopulating HSC as compared with standard ex vivo cultured controls. These findings identify the cell-intrinsic barriers that restrict the transduction of quiescent HSC and provide a means to overcome them, paving the way for the genetic engineering of unstimulated HSC.

Oncogene-induced maladaptive activation of trained immunity in the pathogenesis and treatment of Erdheim-Chester disease.

Trained immunity (TI) is a proinflammatory program induced in monocyte/macrophages upon sensing of specific pathogens and is characterized by immunometabolic and epigenetic changes that enhance cytokine production. Maladaptive activation of TI (ie, in the absence of infection) may result in detrimental inflammation and development of disease; however, the exact role and extent of inappropriate activation of TI in the pathogenesis of human diseases is undetermined. In this study, we uncovered the oncogene-induced, maladaptive induction of TI in the pathogenesis of a human inflammatory myeloid neoplasm (Erdheim-Chester disease, [ECD]), characterized by the BRAFV600E oncogenic mutation in monocyte/macrophages and excess cytokine production. Mechanistically, myeloid cells expressing BRAFV600E exhibit all molecular features of TI: activation of the AKT/mammalian target of rapamycin signaling axis; increased glycolysis, glutaminolysis, and cholesterol synthesis; epigenetic changes on promoters of genes encoding cytokines; and enhanced cytokine production leading to hyperinflammatory responses. In patients with ECD, effective therapeutic strategies combat this maladaptive TI phenotype; in addition, pharmacologic inhibition of immunometabolic changes underlying TI (ie, glycolysis) effectively dampens cytokine production by myeloid cells. This study revealed the deleterious potential of inappropriate activation of TI in the pathogenesis of human inflammatory myeloid neoplasms and the opportunity for inhibition of TI in conditions characterized by maladaptive myeloid-driven inflammation.

Antiviral immunity and nucleic acid sensing in haematopoietic stem cell gene engineering.

The low gene manipulation efficiency of human hematopoietic stem and progenitor cells (HSPC) remains a major hurdle for sustainable and broad clinical application of innovative therapies for a wide range of disorders. Given that all current and emerging gene transfer and editing technologies are bound to expose HSPC to exogenous nucleic acids and most often also to viral vectors, we reason that host antiviral factors and nucleic acid sensors play a pivotal role in the efficacy of HSPC genetic manipulation. Here, we review recent progress in our understanding of vector-host interactions and innate immunity in HSPC upon gene engineering and discuss how dissecting this crosstalk can guide the development of more stealth and efficient gene therapy approaches in the future.

Assessing the Impact of Cyclosporin A on Lentiviral Transduction and Preservation of Human Hematopoietic Stem Cells in Clinically Relevant Ex Vivo Gene Therapy Settings.

Improving hematopoietic stem and progenitor cell (HSPC) permissiveness to lentiviral vector (LV) transduction without compromising their biological properties remains critical for broad-range implementation of gene therapy as a treatment option for several inherited diseases. This study demonstrates that the use of one-hit ex vivo LV transduction protocols based on either cyclosporin A (CsA) or rapamycin enable as efficient gene transfer as the current two-hit clinical standard into bone marrow-derived CD34+ cells while better preserving their engraftment capacity in vivo. CsA was additive with another enhancer of transduction, prostaglandin E2, suggesting that tailored enhancer combinations may be applied to overcome multiple blocks to transduction simultaneously in HSPC. Interestingly, besides enhancing LV transduction, CsA also significantly reduced HSPC proliferation, preserving the quiescent G0 fraction and the more primitive multipotent progenitors, thereby yielding the highest engraftment levels in vivo. Importantly, no alterations in the vector integration profiles could be detected between CsA and control transduced HSPC. Overall, the present findings contribute to the development of more efficient and sustainable LV gene therapy protocols, underscoring the benefits of scaling down required vector doses, as well as shortening the HSPC ex vivo culture time.

Cyclosporine H Overcomes Innate Immune Restrictions to Improve Lentiviral Transduction and Gene Editing In Human Hematopoietic Stem Cells.

Innate immune factors may restrict hematopoietic stem cell (HSC) genetic engineering and contribute to broad individual variability in gene therapy outcomes. Here, we show that HSCs harbor an early, constitutively active innate immune block to lentiviral transduction that can be efficiently overcome by cyclosporine H (CsH). CsH potently enhances gene transfer and editing in human long-term repopulating HSCs by inhibiting interferon-induced transmembrane protein 3 (IFITM3), which potently restricts VSV glycoprotein-mediated vector entry. Importantly, individual variability in endogenous IFITM3 levels correlated with permissiveness of HSCs to lentiviral transduction, suggesting that CsH treatment will be useful for improving ex vivo gene therapy and standardizing HSC transduction across patients. Overall, our work unravels the involvement of innate pathogen recognition molecules in immune blocks to gene correction in primary human HSCs and highlights how these roadblocks can be overcome to develop innovative cell and gene therapies.

Lentiviral vectors escape innate sensing but trigger p53 in human hematopoietic stem and progenitor cells.

Clinical application of lentiviral vector (LV)-based hematopoietic stem and progenitor cells (HSPC) gene therapy is rapidly becoming a reality. Nevertheless, LV-mediated signaling and its potential functional consequences on HSPC biology remain poorly understood. We unravel here a remarkably limited impact of LV on the HSPC transcriptional landscape. LV escaped innate immune sensing that instead led to robust IFN responses upon transduction with a gamma-retroviral vector. However, reverse-transcribed LV DNA did trigger p53 signaling, activated also by non-integrating Adeno-associated vector, ultimately leading to lower cell recovery ex vivo and engraftment in vivo These effects were more pronounced in the short-term repopulating cells while long-term HSC frequencies remained unaffected. Blocking LV-induced signaling partially rescued both apoptosis and engraftment, highlighting a novel strategy to further dampen the impact of ex vivo gene transfer on HSPC. Overall, our results shed light on viral vector sensing in HSPC and provide critical insight for the development of more stealth gene therapy strategies.

Cyclosporin a and rapamycin relieve distinct lentiviral restriction blocks in hematopoietic stem and progenitor cells.

Improving hematopoietic stem and progenitor cell (HSPC) permissiveness to HIV-derived lentiviral vectors (LVs) remains a challenge for the field of gene therapy as high vector doses and prolonged ex vivo culture are still required to achieve clinically relevant transduction levels. We report here that Cyclosporin A (CsA) and Rapamycin (Rapa) significantly improve LV gene transfer in human and murine HSPC. Both compounds increased LV but not gammaretroviral transduction and acted independently of calcineurin and autophagy. Improved gene transfer was achieved across all CD34(+) subpopulations, including in long-term SCID repopulating cells. Effects of CsA were specific of HSPC and opposite to its known impact on HIV replication. Mutating the Cyclophilin A binding pocket of the viral capsid (CA) further improved transduction in combination with CsA. Tracking of the LV genome fate revealed that CsA relieves a CA-dependent early block and increases integration, while Rapa acts early in LV infection independently of the viral CA. In agreement, only Rapa was able to improve transduction by an integrase-defective LV harboring wild-type CA. Overall, our findings pave the way for more efficient and sustainable LV gene therapy in human HSPCs and shed light on the multiple innate barriers specifically hampering LV transduction in these cells.