Impaired STING Activation Due to a Variant in the E3 Ubiquitin Ligase AMFR in a Patient with Severe VZV Infection and Hemophagocytic Lymphohistiocytosis

Varicella zoster virus (VZV) is a neurotropic alphaherpesvirus exclusively infecting humans, in whom it causes two distinct pathologies: varicella (chickenpox) upon primary infection and herpes zoster (shingles) following reactivation [1]. VZV spreads via inhalation of infectious droplets and infects mononuclear cells in the tonsils, leading to viremia through infected T cells and dissemination to the skin to replicate and cause the characteristic vesicular eruptions known as varicella [1]. Importantly, VZV establishes latency in sensory neuronal ganglia, from which it can reactivate and spread to the skin as zoster or to the central nervous system (CNS), manifesting as meningitis, encephalitis, cerebellitis, or vasculopathy with stroke [2, 3]. About 20% of patients hospitalized with chickenpox experience neurological complications [4], suggesting a correlation between impaired systemic control of the infection and development of severe varicella with dissemination to the CNS. Knowledge on the pathogenesis of VZV infection and immunity remains incomplete, and although implementation of a VZV vaccine with the Oka strain in certain countries has resulted in reduced frequency of VZV cases, this infectious disease remains a worldwide health problem [5].

VZV is an enveloped DNA virus, which can infect epithelial cells, peripheral blood mononuclear cells (PBMCs), and neurons [1, 6]. The genome of VZV comprises 71 known open reading frames (ORF)s, of which the VZV latency-associated transcript (VLT) is the only viral gene known to be transcribed during latency, together with a VLT-ORF63 fusion transcript [7, 8]. Unrelated clinical VZV isolates exhibit only little variability in virulence [9], suggesting that defective host immunity is the major determinant of disease severity rather than differences between virus strains. The precise determinants of protective immunity toward VZV are incompletely understood, in part because VZV is a strictly human pathogen, and limited data are available from humanized mice studies [10]. Studies regarding inborn errors of immunity (IEI) that cause increased susceptibility to VZV are therefore of particular importance. From these studies, it is clear that cellular immunity mediated by natural killer (NK) cells and T cells plays a particularly important role during VZV infection, as individuals with IEI affecting NK and T cell function may present with severe and disseminated VZV infections [11‐13]. Moreover, genetic defects in the interferon-γ-receptor 1 (IFNGR1) and tyrosine kinase (TYK) 2 interfere with macrophage function and were found to predispose to disseminated infection with mycobacteria and VZV [14, 15]. Finally, genetic defects in the innate cytosolic DNA sensor RNA polymerase III (POL III) were identified and demonstrated to cause selectively increased susceptibility to VZV CNS infection and pneumonitis in otherwise healthy children and adults [16‐18].

The innate immune system utilizes pattern recognition receptors (PRRs) to detect pathogen-associated molecular patterns (PAMPs) in order to mount protective immune responses. This includes the production of cytokines and interferons (IFN)s, the latter exhibiting antiviral activity through their ability to induce IFN-stimulated genes (ISGs) [19]. Different classes of PRRs are involved in the recognition of virus infections, including membrane-associated Toll-like receptors (TLR)s, cytosolic RNA-sensing retinoic acid-inducible gene 1 (RIG-I)-like receptors, and finally cytosolic DNA sensors [19]. Within the group of DNA sensors, TLR9 detects unmethylated DNA, RNA polymerase III (POL III) recognizes AT-rich DNA, while absent in melanoma (AIM)2, gamma-interferon-inducible protein 16 (IFI16) and cyclic GMP-AMP synthase (cGAS) senses double-stranded (ds) DNA in a sequence-independent manner [20, 21]. Among these, cGAS has emerged as the main cytosolic DNA sensor, signaling through the adaptor molecule Stimulator of Interferon Genes (STING) to induce mainly type 1 IFN responses, but also to contribute to nuclear factor kappa B (NF-κB) activation [22].

Recent reports have described the occurrence of hemophagocytic lymphohistiocytosis (HLH) during VZV infection in the setting of IEI [23]. HLH is a rare but life-threatening clinical syndrome characterized by uncontrolled activation of immune cells and severe systemic inflammation [24]. Familial (or primary) HLH (fHLH) is a genetic disorder most often caused by autosomal recessive loss of function mutations in genes that disrupt the cytotoxic activity of NK and CD8 + T cells, including PRF1, UNC13D, STX11, STXBP2, RAB27A, LYST, AP3B1, SH2D1A, and BIRC4 [25]. While autosomal recessive primary HLH presents early in childhood, hypomorphic mutations in HLH genes are generally associated with fHLH onset in adulthood [26]. Sporadic (or secondary) HLH develops without a known underlying genetic cause but rather presents in association with malignancies, autoimmune or autoinflammatory diseases, or is triggered by infections [27, 28]. Interestingly, both familial and sporadic HLH might present with an infectious trigger, and herpesviruses are the most common causes of virus-induced HLH, with Epstein-Barr-virus (EBV) and cytomegalovirus (CMV) accounting for more than 50% of virus-associated HLH [27]. Similarly, VZV can trigger HLH upon primary infection or reactivation, although this is mostly reported in patients with underlying immunosuppression or fHLH [23, 29‐31].

Here, we describe a patient with severe clinical manifestations of VZV infection and HLH, whom we studied in great detail in order to understand the underlying genetic and immunological pathogenesis. We hypothesize that severe VZV disease in some otherwise healthy individuals may be caused by monogenic IEI, not necessarily displaying complete clinical penetrance. Based on the current knowledge on VZV infection pathogenesis, innate and adaptive immune circuits, and HLH-associated molecules, we searched whole exome sequencing (WES) data for rare gene variants that may underlie susceptibility to severe VZV infection and/or HLH. In this patient, we identified a monoallelic variant in the gene encoding the ubiquitin ligase autocrine motility factor receptor (AMFR) which regulates cGAS-STING signaling through STING poly-ubiquitination. This finding was functionally validated in the by showing reduced STING activation and IFN responses as well as increased VZV replication in patient PBMCs and fibroblasts from the patient compared to controls. We propose that defects in AMFR, which impair cGAS-STING-mediated sensing of foreign DNA, confer enhanced susceptibility to severe disseminated VZV infection and risk of secondary hyperinflammation.

The major finding of the present study is the description of a novel predisposition to severe viral infection involving abnormal AMFR-mediated polyubiquitination and decreased activation of STING in a child with a cellular phenotype showing impaired VZV-induced type I IFN responses and increased VZV replication and a clinical presentation with severe VZV infection and HLH. To our knowledge, this is the first report of a defect in STING signaling in humans causing increased susceptibility to viral infection.

Following the initial discovery of STING [37] and the subsequent discovery of cGAS [38], the cGAS-STING pathway is now appreciated as the major driver of the antiviral IFN response. Other DNA sensors, including RNA POL III, have been identified by many independent groups, but their respective roles in innate immunity are not fully clarified and may be dependent on cell type and cellular context [16, 39, 40]. cGAS recognizes foreign and host-derived dsDNA in a sequence-independent manner and catalyzes the formation of 2′3′-cGAMP that serves as a second messenger to the ER-bound adaptor protein STING [41, 42], which traffics to the Golgi, where it recruits TBK1, thus leading to phosphorylation of STING on serine 366, and this in turn leads to recruitment of IRF3 [43]. This positions IRF3 for phosphorylation by TBK1 allowing IRF3 dimerization, nuclear translocation, and induction of IFN-β gene expression [44]. The role of ubiquitination in STING biology is well documented and includes multiple types of ubiquitin linkage [45]. For instance, K63 ubiquitination by TRIM56, TRIM32, and Mul1 is important for the activation of signaling [46‐48] and is targeted by HSV1 to facilitate CNS infection [49]. Moreover, K48 ubiquitination by RNF5 and RNF90 has been reported to promote STING degradation and downregulate signaling [50, 51]. Finally, K27-linked ubiquitination of STING by AMFR promotes STING ER-to-Golgi trafficking [35], which is a rate-limiting step in STING signaling [52] and promotes activation of the pathway and antiviral defense [35].

AMFR/g78 is an E3 protein ubiquitin ligase embedded in the ER membrane, important for the degradation of misfolded proteins in the ER-associated degradation response (ERAD). Upon microbial DNA challenge, AMFR catalyzes K27-linked polyubiquitination of STING, and this ubiquitin chain creates an anchoring platform for recruiting and activating TBK1, resulting in IRF3 phosphorylation and IFN-β induction [35]. Moreover, AMFR is also involved in regulating cholesterol biosynthesis which may impact ER membrane fluidity and STING trafficking from ER to Golgi, thereby influencing IFN activation and ISG responses [53]. Indeed, previous research has shown that STING signaling is linked to cholesterol metabolic pathways and cellular cholesterol content. Specifically, reprogramming of lipid metabolism leading to altered cholesterol synthesis via the mevalonate pathway can influence the threshold for type-I IFN production through STING activation [54]. The missense variant causing an amino acid substitution from arginine to cysteine is localized at the well-preserved position 594 within the G2BR domain, which is responsible for binding to the E2 ubiquitin-conjugating enzyme UBE2G, thus facilitating interaction between the RING domain and E2 enzyme and subsequently ubiquitin transfer. Deletion of the G2BR domain was reported to abolish the ERAD function of AMFR [55], and a pathogenic variant in this domain would therefore be expected to cause disturbed AMFR-mediated functions, most notably defective ubiquitination and activation of STING [35]. Indeed, we found that the cGAS-STING pathway is impaired in patient PBMCs stimulated by 2′3-cGAMP or dsDNA when measuring pSTING, pTBK-1, and ISG15, as well as decreased response to HSV1 despite the integrity of TLR3 and TLR9 signaling pathways in the patient. We were able to attribute this directly to decreased STING ubiquitination in cells expressing AMFR R549C as compared to cells expressing AMFR WT. Moreover, we demonstrate that the AMFR R549C variant interferes with AMFR WT in a partially dominant-negative manner with dose–response kinetics, although we cannot formally exclude a contribution from haploinsufficiency.

The effect of the AMFR variant may be through interaction with endogenous WT AMFR expressed from the other allele, thus eventually resulting in limiting amounts of AMFR for full STING activation. Alternatively, mutant AMFR may bind to STING and by competition cause dysregulation/insufficient activation of a fraction of cellular STING available for activation. At the molecular level, structural biology may provide some hypotheses. The R595C mutation in the patient AMFR variant is located in the E2 ubiquitin-conjugating enzyme binding domain responsible for interaction with the E2 ligase UBE2G2/UPC7 [56] and is part of a stretch of positively charged residues on positions 594–596. Therefore, we speculate that the R595C mutation impairs the recruitment of the E2 ubiquitin ligase to AMFR. Since AMFR is part of a larger protein complex including, for instance, INSIGR1, some likely mechanisms for the dominant negative effect include (i) more than one AMFR molecule is involved in each AMFR-containing complex, and the patient variant, therefore, leads to impaired recruitment of the E2 ligase to the complexes containing WT:mutant and mutant:mutant AMFR; (ii) there is only one AMFR molecule per complex, but other proteins in the complex are rate-limiting, and complexes containing mutant AMFR, therefore, compete for these proteins, hence exerting dominant negative activity toward complexes containing WT AMFR. Altogether, elucidation of the exact nature of the mechanism of the dominant negative activity required further investigation.

Finally, we establish the causal relationship between genotype and phenotype by demonstrating the reconstitution of the cellular phenotype (cGAS-STING signaling) in patient PBMCs complemented with WT AMFR. Whether the AMFR-STING defect is relatively specific to VZV and HSV1, or alternatively involves other DNA viruses remains to be resolved, but our finding of normal responses to SeV may indicate a defect mainly related to infection by DNA viruses. Collectively, the precise role of cGAS STING-mediated DNA sensing and IFN induction in humans remains to be determined.

The existence of defective innate DNA sensing predisposing to VZV infection, possibly through impaired signaling in response to the presence of viral cytosolic DNA, is similar to the previously reported IEI affecting another cytosolic DNA sensor RNA polymerase III, which has been described in children and adults with severe VZV CNS encephalitis, vasculopathy, and recurrent VZV meningoencephalitis [16‐18]. Therefore, defective DNA sensing may be a common theme in VZV predisposition and might act in a partially cell-type-dependent manner. In the case of POL III deficiency, with CNS involvement constituting a prominent feature, we hypothesized that a particularly important role of POL III in recognizing the AT-rich VZV genome may be at play in certain cell types, such as neuronal cells, in which cGAS expression is low/absent [16, 17]. On the other hand, one study identified an important role of STING in mounting type I and type III IFN responses to VZV in human dermal fibroblasts and HaCaT keratinocytes with potential implications for varicella pathogenesis and suggesting an important role of cGAS-STING in the skin [57]. Recently, the cGAS-STING DNA sensing pathway was demonstrated to be required for IFN induction and VZV restriction during VZV infection in THP1 cells, and the VZV protein ORF9 was reported to antagonize cGAS-STING signaling and IFN production [58]. The observation that POL III dominates as a VZV sensor in mononuclear cells might in fact explain why we do not observe significantly decreased IFN induction in PBMCs in response to VZV and only could measure a modest increase in VZV replication in this cell type from the patient. The pattern of inheritance appears to be autosomal dominant with incomplete penetrance since the mother and two younger siblings also carry the AMFR R594C variant and so far have not experienced severe VZV infection. However, PBMCs from the mother of the patient also exhibit decreased phosphorylation of IRF3 (although less pronounced than in the patient), as well as significantly reduced IFNα/β responses to VZV infection together with increased VZV replication. The data suggest a milder cellular phenotype in the mother. However, since fibroblasts from the mother were not available, we were unable to make a precise comparison of antiviral responses in different cell types between patient, mother, and healthy controls. Indeed, fibroblast which produces more vivid antiviral responses than PBMCs may more precisely reflect the physiological situation in vivo. Moreover, it also has to be kept in mind that VZV infection models in primary human cells combined with the difficulty of measuring intracellular signaling pathways by a slowly replicating virus (i.e., not a potent agonist added at one given time) that can only be used in relatively low titers can be variable and is unlikely to give a completely straight biochemical picture in vitro regarding cellular responses. Incomplete penetrance is a frequent observation in IEI and may be due to influences from other genetic variants, differences in allele expression (for monoallelic variants), differences in viral load, or other immunomodulatory external factors. In order to hopefully avoid future severe illness in the siblings, they have been vaccinated against VZV by the live attenuated VZV vaccine.

As to the possible link between this variant causing increased VZV replication and severe VZV infection to the development of HLH, this remains insufficiently understood. However, some previous observations in the literature may suggest a possible pathophysiological mechanism. Among cases of secondary HLH, viral infection is the most common trigger [59], and more recently, it has been suggested that monoallelic variants in known HLH genes, or less penetrant other gene variants associated with IEI, may underlie HLH [26, 60]. This may be the case for other IEI affecting innate immunity as exemplified by a previous study demonstrating disseminated VZV and HLH in a patient with GATA2 defect [23]. In this context, enhanced viral replication and an increased load of PAMPs may increase the risk of developing a hyperinflammatory state. Further, we cannot exclude that dysregulated STING activation as described in our patient may lead to hyperinflammation during a high viral burden. We suggest that the variant may not have a very strong direct disease-causing effect but rather lowers the threshold for virus-induced inflammation and HLH. The link between STING-signaling and reduced threshold for HLH is further underpinned by a recent study implicating the well-established HLH-disposing gene UNC13D in STING regulation [61]. Collectively, these complex cellular interactions between pathogen/PAMP and the immune system that may trigger a cytokine storm and HLH cannot be sufficiently mimicked and studied in an in vitro system. Based on the present report, we would encourage others with the relevant expertise to investigate this matter in a physiological context, which would however require modified animals, since VZV is strictly a human pathogen.

Previously, Goldbach-Mansky and associates described a vascular and pulmonary syndrome in patients with gain-of-function variants in the STING-encoding gene STING1 (formerly known as TMEM173) and suggested the name STING-associated vasculopathy with onset in infancy (SAVI) for this autoinflammatory interferonopathy [62]. The authors identified three different mutations in exon 5 of STING1 in these six patients and observed constitutive STAT1 phosphorylation and increased constitutive and inducible type I IFN expression in patient cells [62]. Since then, several other publications have described additional patients and extended the clinical phenotype [63, 64]. Moreover, other IEIs also affecting IFN production, such as STAT2 deficiency, translate into partially overlapping clinical manifestations together with the interferonopathy signature [65, 66]. However, despite the description of numerous IEIs affecting pathogen and PAMP receptors and their downstream IFN-inducing signaling pathways causing susceptibility to severe viral infections, defects in the cGAS-STING pathway in humans have not been previously reported. The clinical and cellular phenotype described here is significantly different from the one ascribed to SAVI, and notably, we did not observe increased IFN nor ISG levels in vitro under the given conditions examining patient cells harvested after acute illness. However, we cannot entirely exclude that elevated ISGs may have been present in the circulation of the patient during the acute episode, and thus we cannot rule out a degree of autoinflammation/interferonopathy caused by the AMFR variant and possibly associated to the development of VZV-triggered HLH in the patient.

In conclusion, we suggest a novel genetic etiology of severe VZV disease in childhood, also representing the first IEI related to a defect in the STING-cGAS pathway. The present work contributes information on the question of the role of DNA sensors in host defense in human immunology. Together with our previous work on POL III, the available data suggest that POL III and cGAS-STING may each play non-redundant roles in VZV immunity in humans, depending on the cell type and tissue involved. Collectively, the precise role of cGAS STING-mediated DNA sensing and IFN induction in humans remains to be further studied. Identification of human IEIs involving these molecules remains a powerful tool for establishing the contributions of relevant immune signaling pathways in humans and for gaining valuable insights into how dysregulated cGAS-STING signaling may lead to human disease, ranging from autoinflammatory interferonopathy to severe viral infection.