[Artificial intelligence and secure use of health data in the KI-FDZ project: anonymization, synthetization, and secure processing of real-world data]. / Künstliche Intelligenz und sichere Gesundheitsdatennutzung im Projekt KI-FDZ: Anonymisierung, Synthetisierung und sichere Verarbeitung für Real-World-Daten.

The increasing digitization of the healthcare system is leading to a growing volume of health data. Leveraging this data beyond its initial collection purpose for secondary use can provide valuable insights into diagnostics, treatment processes, and the quality of care. The Health Data Lab (HDL) will provide infrastructure for this purpose. Both the protection of patient privacy and optimal analytical capabilities are of central importance in this context, and artificial intelligence (AI) provides two opportunities. First, it enables the analysis of large volumes of data with flexible models, which means that hidden correlations and patterns can be discovered. Second, synthetic - that is, artificial - data generated by AI can protect privacy.This paper describes the KI-FDZ project, which aims to investigate innovative technologies that can support the secure provision of health data for secondary research purposes. A multi-layered approach is investigated in which data-level measures can be combined in different ways with processing in secure environments. To this end, anonymization and synthetization methods, among others, are evaluated based on two concrete application examples. Moreover, it is examined how the creation of machine learning pipelines and the execution of AI algorithms can be supported in secure processing environments. Preliminary results indicate that this approach can achieve a high level of protection while maintaining data validity. The approach investigated in the project can be an important building block in the secure secondary use of health data.

A conserved isoleucine in the binding pocket of RIG-I controls immune tolerance to mitochondrial RNA.

RIG-I is a cytosolic receptor of viral RNA essential for the immune response to numerous RNA viruses. Accordingly, RIG-I must sensitively detect viral RNA yet tolerate abundant self-RNA species. The basic binding cleft and an aromatic amino acid of the RIG-I C-terminal domain(CTD) mediate high-affinity recognition of 5'triphosphorylated and 5'base-paired RNA(dsRNA). Here, we found that, while 5'unmodified hydroxyl(OH)-dsRNA demonstrated residual activation potential, 5'-monophosphate(5'p)-termini, present on most cellular RNAs, prevented RIG-I activation. Determination of CTD/dsRNA co-crystal structures and mutant activation studies revealed that the evolutionarily conserved I875 within the CTD sterically inhibits 5'p-dsRNA binding. RIG-I(I875A) was activated by both synthetic 5'p-dsRNA and endogenous long dsRNA within the polyA-rich fraction of total cellular RNA. RIG-I(I875A) specifically interacted with long, polyA-bearing, mitochondrial(mt) RNA, and depletion of mtRNA from total RNA abolished its activation. Altogether, our study demonstrates that avoidance of 5'p-RNA recognition is crucial to prevent mtRNA-triggered RIG-I-mediated autoinflammation.

Inhibition of cellular RNA methyltransferase abrogates influenza virus capping and replication.

Orthomyxo- and bunyaviruses steal the 5' cap portion of host RNAs to prime their own transcription in a process called "cap snatching." We report that RNA modification of the cap portion by host 2'-O-ribose methyltransferase 1 (MTr1) is essential for the initiation of influenza A and B virus replication, but not for other cap-snatching viruses. We identified with in silico compound screening and functional analysis a derivative of a natural product from Streptomyces, called trifluoromethyl-tubercidin (TFMT), that inhibits MTr1 through interaction at its S-adenosyl-l-methionine binding pocket to restrict influenza virus replication. Mechanistically, TFMT impairs the association of host cap RNAs with the viral polymerase basic protein 2 subunit in human lung explants and in vivo in mice. TFMT acts synergistically with approved anti-influenza drugs.

G-rich DNA-induced stress response blocks type-I-IFN but not CXCL10 secretion in monocytes.

Excessive inflammation can cause damage to host cells and tissues. Thus, the secretion of inflammatory cytokines is tightly regulated at transcriptional, post-transcriptional and post-translational levels and influenced by cellular stress responses, such as endoplasmic reticulum (ER) stress or apoptosis. Here, we describe a novel type of post-transcriptional regulation of the type-I-IFN response that was induced in monocytes by cytosolic transfection of a short immunomodulatory DNA (imDNA), a G-tetrad forming CpG-free derivative of the TLR9 agonist ODN2216. When co-transfected with cytosolic nucleic acid stimuli (DNA or 3P-dsRNA), imDNA induced caspase-3 activation, translational shutdown and upregulation of stress-induced genes. This stress response inhibited the type-I-IFN induction at the translational level. By contrast, the induction of most type-I-IFN-associated chemokines, including Chemokine (C-X-C Motif) Ligand (CXCL)10 was not affected, suggesting a differential translational regulation of chemokines and type-I-IFN. Pan-caspase inhibitors could restore IFN-ß secretion, yet, strikingly, caspase inhibition did not restore global translation but instead induced a compensatory increase in the transcription of IFN-ß but not CXCL10. Altogether, our data provide evidence for a differential regulation of cytokine release at both transcriptional and post-transcriptional levels which suppresses type-I-IFN induction yet allows for CXCL10 secretion during imDNA-induced cellular stress.

Sequence-specific activation of the DNA sensor cGAS by Y-form DNA structures as found in primary HIV-1 cDNA.

Cytosolic DNA that emerges during infection with a retrovirus or DNA virus triggers antiviral type I interferon responses. So far, only double-stranded DNA (dsDNA) over 40 base pairs (bp) in length has been considered immunostimulatory. Here we found that unpaired DNA nucleotides flanking short base-paired DNA stretches, as in stem-loop structures of single-stranded DNA (ssDNA) derived from human immunodeficiency virus type 1 (HIV-1), activated the type I interferon-inducing DNA sensor cGAS in a sequence-dependent manner. DNA structures containing unpaired guanosines flanking short (12- to 20-bp) dsDNA (Y-form DNA) were highly stimulatory and specifically enhanced the enzymatic activity of cGAS. Furthermore, we found that primary HIV-1 reverse transcripts represented the predominant viral cytosolic DNA species during early infection of macrophages and that these ssDNAs were highly immunostimulatory. Collectively, our study identifies unpaired guanosines in Y-form DNA as a highly active, minimal cGAS recognition motif that enables detection of HIV-1 ssDNA.

A Conserved Histidine in the RNA Sensor RIG-I Controls Immune Tolerance to N1-2'O-Methylated Self RNA.

The cytosolic helicase retinoic acid-inducible gene-I (RIG-I) initiates immune responses to most RNA viruses by detecting viral 5'-triphosphorylated RNA (pppRNA). Although endogenous mRNA is also 5'-triphosphorylated, backbone modifications and the 5'-ppp-linked methylguanosine ((m7)G) cap prevent immunorecognition. Here we show that the methylation status of endogenous capped mRNA at the 5'-terminal nucleotide (N1) was crucial to prevent RIG-I activation. Moreover, we identified a single conserved amino acid (H830) in the RIG-I RNA binding pocket as the mediator of steric exclusion of N1-2'O-methylated RNA. H830A alteration (RIG-I(H830A)) restored binding of N1-2'O-methylated pppRNA. Consequently, endogenous mRNA activated the RIG-I(H830A) mutant but not wild-type RIG-I. Similarly, knockdown of the endogenous N1-2'O-methyltransferase led to considerable RIG-I stimulation in the absence of exogenous stimuli. Studies involving yellow-fever-virus-encoded 2'O-methyltransferase and RIG-I(H830A) revealed that viruses exploit this mechanism to escape RIG-I. Our data reveal a new role for cap N1-2'O-methylation in RIG-I tolerance of self-RNA.