Cell and immunotherapy offer transformative potential for treating diseases like cancer and autoimmune disorders by modulating the immune system. The development of these therapies is resource-intensive, with the majority of drug candidates failing to progress beyond laboratory testing. While recent advances in machine learning have revolutionised areas such as protein engineering, applications in immunotherapy remain limited due to the scarcity of large-scale, standardised datasets and the complexity of cellular systems. In this work, we address these challenges by leveraging a high-throughput experimental platform to generate data suitable for fine-tuning protein language models. We demonstrate how models fine-tuned using a preference task show surprising correlations to biological assays, and how they can be leveraged for few-shot hit maturation in CARs. This proof-of-concept presents a novel pathway for applying ML to immunotherapy and could generalise to other therapeutic modalities.

supplementary table (excel)

Supplemental tables and figures

supplementary table (excel)

3103 Background: Engineered T cells expressing chimeric antigen receptors (CAR) are now delivering clinically relevant results in patients with advanced hematological malignancies. One critical area for future development is to modulate gene expression thereby endowing the engineered T cell with specific desired features that enhance anti-tumor activity. Methods: Short-hairpin RNA (shRNA) were cloned individually or multiplexed within micro-RNA scaffolds that enabled the co-expression of the individual shRNA with a CAR and a selectable marker all driven by a PolII promoter within a single retroviral vector. Primary human T cells transduced with the CAR-shRNA vectors were selected, expanded in vitro, subjected to negative selection to eliminate any remaining TCR+ cells and examined for target gene expression and functional activity. Results: A 500bp DNA fragment incorporating a shRNA-specific for CD3ζ cloned into a retroviral vectoreffectively knocked down expression of CD3ζ in transduced BCMA-specific CAR T cells. The consequent reduction of cell surface TCR expression resulted in minimal cytokine production upon TCR stimulation in vitro providing a potential allogeneic CAR T approach. These CAR T cells showed no demonstrable evidence of GvHD induction when infused in NSG mice yet maintained BCMA-specific CAR activity in KMS-11 and RPMI-8226 established myeloma models. Initial studies further confirmed that two shRNA could be expressed from a single retroviral vector to modulate the expression of multiple genes. Further engineering of the microRNA framework reduced the size of the transgene load to 394bp while enabling the expression of up to 4 shRNA within a single vector. shRNA specific for CD3ζ, beta-2-microglobulin, CD52 and diacylglycerol kinase alpha were engineered into the framework downstream of a CD19-CAR. Transduced Jurkat cells showed concurrent knockdown of the respective gene products at the mRNA and protein levels. Conclusions: A first-in-human clinical trial evaluating the first-generation single shRNA-vector in the context of a BCMA-targeting CAR as a non-gene edited approach to allogeneic CAR T cell therapy will be initiated in 2020. The proof of principle study here shows that multiple shRNAs are active within a single viral vector thereby avoiding the need for bespoke individual clinical reagents to target multiple genes. The multiplexed shRNA vector system is now in further development to explore whether this strategy can enhance the therapeutic potential of CAR T cells.

To fulfill a productive infection cycle the human immunodeficiency virus (HIV) relies on host-cell factors. Interference with these co-factors holds great promise in protecting cells against HIV infection. LEDGF/p75, encoded by the PSIP1 gene, is used by the integrase (IN) protein in the pre-integration complex of HIV to bind host-cell chromatin facilitating proviral integration. LEDGF/p75 depletion results in defective HIV replication. However, as part of its cellular function LEDGF/p75 tethers cellular proteins to the host-cell genome. We used site-specific editing of the PSIP1 locus using CRISPR/Cas to target the aspartic acid residue in position 366 and mutated it to asparagine (D366N) to disrupt the interaction with HIV IN but retain LEDGF/p75 cellular function. The resulting cell lines demonstrated successful disruption of the LEDGF/p75 HIV-IN interface without affecting interaction with cellular binding partners. In line with LEDGF/p75 depleted cells, D366N cells did not support HIV replication, in part due to decreased integration efficiency. In addition, we confirm the remaining integrated provirus is more silent. Taken together, these results support the potential of site-directed CRISPR/Cas9 mediated knock-in to render cells more resistant to HIV infection and provides an additional strategy to protect patient-derived T-cells against HIV-1 infection as part of cell-based therapy.

Background The Natural Killer Group 2D (NKG2D) receptor is a NK cell activating receptor that binds to eight different ligands (NKG2DL) commonly over-expressed in cancer, including MICA and MICB. The product candidate CYAD-01 are chimeric antigen receptor (CAR) T-cells encoding the full length human NKG2D fused to the intracellular domain of CD3ζ. Data from preclinical models have shown that CYAD-01 cells specifically target solid and hematological tumors. Encouraging preliminary results from the Phase I clinical trial THINK, assessing CYAD-01 safety, showed initial signals of objective clinical responses in patients with r/r AML and MDS. The clinical development of CAR T-cells has been limited by several challenges including achieving sufficient numbers of cells for clinical application. We have previously shown that NKG2D ligands are transiently expressed on activated T cells and that robust cell yields are generated through the addition of a blocking antibody and a PI3K inhibitor during cell manufacture. Here, we investigated the ability of an optimized short hairpin RNA (shRNA) technology to modulate NKG2DL expression on CYAD-01 cells and to determine if there is an increase in the anti-tumor activity of NKG2D-based CAR T-cells (termed CYAD-02). Methods Molecular and cellular analyses identified MICA and MICB as the key NKG2DL expressed on activated T-cells and highly likely to participate in driving fratricide. In silico analysis and in vitro screening allowed the identification of a single shRNA targeting the conserved regions of MICA and MICB, thus downregulating both MICA and MICB expression. The selected shRNA was incorporated in the NKG2D-based CAR vector, creating the next-generation NKG2D-based CAR T-cell candidate, CYAD-02. In addition, truncated versions of the NKG2D receptor were generated to explore the mechanisms of action of NKG2D receptor activity in vivo. The in vivo persistence and anti-tumor activity of CYAD-02 cells was evaluated in an aggressive preclinical model of AML. Results Injection of CAR T-cells bearing truncated forms of the NKG2D-CAR in immunosuppressed mice resulted in similar persistence to the control T-cells. In contrast, CYAD-01 cells had reduced persistence, suggesting that the recognition of the NKG2DL by the NKG2D receptor could contribute to this effect. Analysis of cell phenotype upon CAR T-cell activation showed that MICA and MICB were transiently expressed on T-cells during manufacturing. These results collectively suggested that downregulating MICA and MICB expression in CYAD-01 cells could be a mean to increase CAR T-cell persistence in vivo. Candidate shRNA were screened for efficient targeting of both MICA and MICB at the mRNA and protein level. T-cells transduced with a single vector encoding for the NKG2D-based CAR and the selected shRNA targeting MICA and MICB (CYAD-02) demonstrated 3-fold increased expansion during in vitro culture in the absence of the blocking antibody used to increase cell yield during manufacture. When injected into immunosuppressed mice, CYAD-02 cells generated with the Optimab process showed 10-fold higher engraftment one week after injection and potent anti-tumor activity resulting in 2.6-fold increase of mouse survival in an aggressive AML model. Conclusions By using a single vector encoding the NKG2D-based CAR next to a shRNA targeting MICA and MICB and combined with improved cell culture methods, CYAD-02, the next-generation of NKG2D-based CAR T-cells, demonstrated enhanced in vivo persistence and anti-tumor activity. Following FDA acceptance of the IND application, a Phase 1 dose-escalation trial evaluating the safety and clinical activity of CYAD-02 for the treatment of r/r AML and MDS is scheduled to start in early 2020. Disclosures Fontaine: Celyad: Employment. Demoulin:Celyad: Employment. Bornschein:Celyad: Employment. Raitano:Celyad: Employment. Machado:Horizon Discovery: Employment. Moore:Avvinity Therapeutics: Employment, Other: Relationship at the time the work was performed; Horizon Discovery: Employment, Equity Ownership, Other: Relationship at the time the work was performed; Centauri Therapeutics: Consultancy, Other: Current relationship. Sotiropoulou:Celyad: Employment. Gilham:Celyad: Employment.

e15040 Background: The majority of CAR-T cell therapy studies use the patients’ own autologous cells which carry many potential restrictions related to manufacturing such as quantity and quality of cells. Sourcing cells from an allogeneic healthy donor avoids many of these manufacturing issues. However, the recognition of recipient HLA by the T Cell Receptor of the donor T cells drives a graft versus host (GvH) response leading to normal tissue damage thereby limiting the potential of allogenic T cell therapy. Blunting this response is critical for the successful development of a safe allogeneic CAR-T cell therapy. Hypothesis: Inhibition of TCR activity using shRNAs targeting the CD3 complex can reduce or eliminate TCR driven GvH activity. Objective: To identify candidate shRNA that can knockdown CD3 and achieve inhibition of TCR signaling in primary human T cells. Methods: Peripheral blood derived T cells from healthy donors were transduced with the appropriate vector then subjected to puromycin selection for further 48 hours after transduction. On day 8, cells were finally harvested and characterized. Results: There was a systematic decrease of CD3e or CD3z expression at the RNA level in cells transduced respectively with shRNA against CD3e or CD3z chains of the TCR complex in the Jurkat T cell line as well as in human primary T cells This also correlated with a systematic decrease of CD3e and TCRa/b expression at the protein level in cells transduced with either shRNA against CD3e or CD3z Most interestingly shRNA targeting of the CD3 complex drove reduced TCR signaling as determined by reduced levels of cytokine release upon mitogenic stimulation Conclusions: Targeting of the CD3 complex with shRNA reduced the level of CD3 expression in a Jurkat T cell line as well as in primary human T cells. This resulted in decreased in vitro responses of the engineered T cells during mitogenic stimulation, most likely due to impaired TCR assembly and cell surface expression. These studies support the concept of exploiting shRNA technology for allogenic CAR-T cell therapy that using a single vector, non-gene editing approach.

Graft-versus-host disease (GvHD) assessment has been shown to be a challenge for healthcare professionals, leading to the development of the eGVHD App (www.uzleuven.be/egvhd). In this study, we formally evaluated the accuracy of using the App compared to traditional assessment methods to assess GvHD. Our national multicenter randomized controlled trial involved seven Belgian transplantation centers and 78 healthcare professionals selected using a 2-stage convenience sampling approach between January and April 2017. Using a 1:1 randomization stratified by profession, healthcare professionals were assigned to use either the App (APP) or their usual GvHD assessment aids (No APP) to assess the diagnosis and severity score of 10 expert-validated clinical vignettes. Our main outcome measure was the difference in accuracy for GvHD severity scoring between both groups. The odds of being correct were 6.14 (95% CI: 2.83-13.34) and 6.29 (95% CI: 4.32-9.15) times higher in favor of the APP group for diagnosis and scoring, respectively (P<0.001). Appassisted GvHD severity scoring was significantly superior for both acute and chronic GvHD, with an Odds Ratio of 17.89 and 4.34 respectively (P<0.001) and showed a significantly increased inter-observer agreement compared to standard practice. Despite a mean increase of 24 minutes (95% CI: 20.45-26.97) in the time needed to score the whole GvHD test package in the APP group (P<0.001), usability feedback was positive. The eGVHD App shows superior GvHD assessment accuracy compared to standard practice and has the potential to improve the quality of outcome data registration in allogeneic stem cell transplantation.

The NUP214-ABL1 fusion is a constitutively activated tyrosine kinase that is significantly associated with overexpression of the TLX1 and TLX3 transcription factors in T cell acute lymphoblastic leukemia (T-ALL). Here we show that NUP214-ABL1 cooperates with TLX1 in driving T-ALL development using a transgenic mouse model and human T-ALL cells. Using integrated ChIP-sequencing, ATAC-sequencing, and RNA-sequencing data, we demonstrate that TLX1 and STAT5, the downstream effector of NUP214-ABL1, co-bind poised enhancer regions, and cooperatively activate the expression of key proto-oncogenes such as MYC and BCL2. Inhibition of STAT5, downregulation of TLX1 or MYC, or interference with enhancer function through BET-inhibitor treatment leads to reduction of target gene expression and induction of leukemia cell death.

Mammalian genomes encode a plethora of long non-coding RNA (lncRNA). These transcripts are thought to regulate gene expression, influencing biological processes from development to pathology. Results from the few lncRNA that have been studied in the context of the immune system have highlighted potentially critical functions as network regulators. Here we explored the nature of the lncRNA transcriptome in regulatory T cells (Tregs), a subset of CD4+ T cells required to establish and maintain immunological self-tolerance. The identified Treg lncRNA transcriptome showed distinct differences from that of non-regulatory CD4+ T cells, with evidence of direct shaping of the lncRNA transcriptome by Foxp3, the master transcription factor driving the distinct mRNA profile of Tregs. Treg lncRNA changes were disproportionally reversed in the absence of Foxp3, with an enrichment for colocalisation with Foxp3 DNA binding sites, indicating a direct coordination of transcription by Foxp3 independent of the mRNA coordination function. We further identified a novel lncRNA Flatr, as a member of the core Treg lncRNA transcriptome. Flatr expression anticipates Foxp3 expression during in vitro Treg conversion, and Flatr-deficient mice show a mild delay in in vitro and peripheral Treg induction. These results implicate Flatr as part of the upstream cascade leading to Treg conversion, and may provide clues as to the nature of this process.

Mucosa-associated lymphoid tissue 1 (Malt1) regulates immune cell function by mediating the activation of nuclear factor κB (NF-κB) signaling through both its adaptor and proteolytic function. Malt1 is also a target of its own protease activity and this self-cleavage further contributes to NF-κB activity. Until now, the functional distinction between Malt1 self-cleavage and its general protease function in regulating NF-κB signaling and immune activation remained unclear. Here we demonstrate, using a new mouse model, the importance of Malt1 self-cleavage in regulating expression of NF-κB target genes and subsequent T cell activation. Significantly, we further establish that Treg homeostasis is critically linked to Malt1 function via a Treg intrinsic and extrinsic mechanism. TCR-mediated Malt1 proteolytic activity and self-cleavage was found to drive Il2 expression in conventional CD4+ T cells, thereby regulating Il2 availability for Treg homeostasis. Remarkably, the loss of Malt1-mediated self-cleavage alone was sufficient to cause a significant Treg deficit resulting in increased anti-tumor immune reactivity without associated autoimmunity complications. These results establish for the first time that inhibition of MALT1 proteolytic activity could be a viable therapeutic strategy to augment anti-tumor immunity.

The Janus kinase 3 (JAK3) tyrosine kinase is mutated in 10% to 16% of T-cell acute lymphoblastic leukemia (T-ALL) cases. JAK3 mutants induce constitutive JAK/STAT signaling and cause leukemia when expressed in the bone marrow cells of mice. Surprisingly, we observed that one third of JAK3-mutant T-ALL cases harbor 2 JAK3 mutations, some of which are monoallelic and others that are biallelic. Our data suggest that wild-type JAK3 competes with mutant JAK3 (M511I) for binding to the common γ chain and thereby suppresses its oncogenic potential. We demonstrate that JAK3 (M511I) can increase its limited oncogenic potential through the acquisition of an additional mutation in the mutant JAK3 allele. These double JAK3 mutants show increased STAT5 activation and increased potential to transform primary mouse pro-T cells to interleukin-7-independent growth and were not affected by wild-type JAK3 expression. These data extend our insight into the oncogenic properties of JAK3 mutations and provide an explanation of why progression of JAK3-mutant T-ALL cases can be associated with the accumulation of additional JAK3 mutations.

Background: Hemophagocytic lymphohistiocytosis (HLH) is a severe inflammatory condition driven by excessive CD8(+) T-cell activation. HLH occurs as both acquired and familial hemophagocytic lymphohistiocytosis (FHL) forms. In both conditions, a sterile or infectious trigger is required for disease initiation, which then becomes self-sustaining and life-threatening. Recent studies have attributed the key distal event to excessive IFN-gamma production; however, the proximal events driving immune dysregulation have remained undefined.Objective: We sought to investigate the role of regulatory T (Treg) cells in the pathophysiology of experimental FHL.Methods: Because mutation in perforin is a common cause of FHL, we used an experimental FHL mouse model in which disease in perforin-deficient mice is triggered by lymphocytic choriomeningitis virus (LCMV). We assessed Treg and CD8(+) T-cell homeostasis and activation during the changing systemic conditions in the mice. In addition, human blood samples were collected and analyzed during the HLH episode.Results: We found no primary Treg cell defects in perforin-deficient mice. However, Treg cell numbers collapsed after LCMV inoculation. The collapse of Treg cell numbers in LCMV-triggered perforin-deficient, but not wild-type, mice was accompanied by the combination of lower IL-2 secretion by conventional CD4(+) T cells, increased IL-2 consumption by activated CD8(+) T cells, and secretion of competitive soluble CD25. Moreover low Treg cell numbers were observed in untreated patients experiencing HLH flares.Conclusion: These results demonstrate that excessive CD8(+) T-cell activation rewires the IL-2 homeostatic network away from Treg cell maintenance and toward feed-forward inflammation. These results also provide a potential mechanistic pathway for the progression of infectious inflammation to persistent inflammation in patients with HLH.

Hemophagocytic lymphohistiocytosis (HLH) is a severe inflammatory condition driven by excessive CD8+ T cell activation. HLH occurs in both acquired and familial (FHL) forms, with mutations in Perforin being a common cause of FHL. In both conditions a sterile or infectious trigger is required for disease initiation, which then becomes self-sustaining and life-threatening. Recent advances using experimental FHL triggered by lymphocytic choriomeningitis virus (LCMV) in Perforin-deficient mice have attributed the key distal event to be excessive IFNγ production, however the proximal events driving immune dysregulation have remained undefined. We investigated the role of regulatory T cells (Tregs) in the pathophysiology of experimental FHL. While we found no primary Treg defects in Perforin-deficient mice, Treg numbers collapsed following LCMV inoculation. The collapse of Treg numbers in LCMV-triggered Perforin-deficient, but not wildtype, mice was driven by the combination of lower IL-2 secretion by conventional CD4+ T cells, increased IL-2 consumption by activated CD8+ T cells and secretion of competitive sCD25 (IL-2 receptor). Together, these data demonstrate that excessive CD8+ T cell activation rewires the IL-2 homeostatic network away from Treg maintenance and towards feed-forward inflammation. In addition, reduced Treg number may contribute to the massive inflammation found in FHL. Disclosures No relevant conflicts of interest to declare.

ObjectiveTo identify the underlying genetic defect in a 16‐year‐old girl with severe early‐onset and refractory systemic lupus erythematosus (SLE), IgA deficiency, and mild lower limb spasticity without neuroradiologic manifestations.MethodsWhole‐exome sequencing and extensive immunologic analysis were performed on samples from the index patient.ResultsWe identified a de novo p.R779H IFIH1 gain‐of‐function mutation in a patient with severe early‐onset SLE, selective IgA deficiency, and mild lower limb spasticity. The same mutation in IFIH1 was recently identified in patients with Aicardi‐Goutières syndrome, a rare neuroimmunologic disorder associated with elevated levels of type I interferon (IFN). IFN induced with helicase C domain 1 functions as an intracellular innate immune receptor that senses viral nucleic acids and leads to the induction of type I IFN and proinflammatory cytokines. Despite systemic immunosuppressive treatment, disease activity persisted in the patient and was associated with elevated serum levels of IFNα and up‐regulation of IFIH1 itself.ConclusionThis finding adds a new genetic causation for Mendelian lupus and greatly extends the disease spectrum associated with mutations in IFIH1 (ranging from inflammatory encephalopathy to prototypic systemic autoimmune disease). This marked phenotypic heterogeneity, despite an identical mutation, demonstrates the importance of modifying factors in type I IFN–dependent pathologies caused by mutations in IFIH1.

Gain-of-function mutations in IFIH1 were identified in Aicardi-Goutieres syndrome (AGS), a rare neuroimmunological disorder associated with elevated levels of type I interferon and characterized by leucoencephalopathy, brain atrophy and intracranial calcifications leading to profound intellectual disability, spasticity and dystonia. IFIH1 functions as an intracellular innate immune receptor that senses viral nucleic acids and leads to the induction of type I interferon and proinflammatory cytokines.

NOD2 is an intracellular sensor that contributes to immune defense and inflammation. Here we investigated whether NOD2 mediates its effects through control of microRNAs (miRNAs). miR-29 expression was upregulated in human dendritic cells (DCs) in response to NOD2 signals, and miR-29 regulated the expression of multiple immune mediators. In particular, miR-29 downregulated interleukin-23 (IL-23) by targeting IL-12p40 directly and IL-23p19 indirectly, likely via reduction of ATF2. DSS-induced colitis was worse in miR-29-deficient mice and was associated with elevated IL-23 and T helper 17 signature cytokines in the intestinal mucosa. Crohn’s disease (CD) patient DCs expressing NOD2 polymorphisms failed to induce miR-29 upon pattern recognition receptor stimulation and showed enhanced release of IL-12p40 on exposure to adherent invasive E. coli. Therefore, we suggest that loss of miR-29-mediated immunoregulation in CD DCs might contribute to elevated IL-23 in this disease.