Activation of STING due to COPI-deficiency

COPA syndrome is caused by loss-of-function mutations in the COP-α subunit of coatomer protein complex I (COPI), which participates in retrograde vesicular trafficking of proteins from the Golgi to the endoplasmic reticulum (ER). Disease manifests early in life with arthritis, lung pathology, kidney dysfunction and systemic inflammation associated with NF-κB activation and type I interferon (IFNαβ) production. Here, we generated in vitro models for COPA syndrome and interrogated inflammatory signalling pathways via a range of biochemical and molecular biological techniques. Results were confirmed with cell lines in which mutant COPA was overexpressed and with COPA syndrome patient PBMCs. We identified Stimulator of Interferon Genes (STING), as a driver of inflammation in COPA syndrome. Furthermore, we found that genetic deletion of COPG1, another COPI subunit protein, induced NF-κB and type I IFN pathways similar to COPA-deficiency. Finally, we demonstrate that in vitro, inflammation due to COPA syndrome mutations was ameliorated by treatment with the small molecule STING inhibitor H-151. Therefore, inflammation induced by deletion of COPI subunits in general suggests a link between retrograde trafficking and STING regulation, and this innate immune sensor represents a novel therapeutic target in COPA syndrome.


Introduction
Coatomer subunit α (COPA) syndrome is a recently identified rare disorder, involving complex pathology with dysregulation of the innate and adaptive immune system (1).
The disease is caused by autosomal dominant mutations in the COPA gene, which encodes the α−subunit of the coatomer complex I (COPI). COPI mediates retrograde trafficking of cargo proteins from Golgi to endoplasmic reticulum (ER) and within cis-Golgi compartments (2). To date, 11 families with mutations in COPA have been identified worldwide (1, [3][4][5][6][7][8][9][10]. Clinically, patients present symptoms with different severity, including interstitial lung disease with or without pulmonary haemorrhage, inflammatory arthritis and immune-mediated kidney disease. Autoantibodies and increased levels of Th17 cells have been identified in the majority of patients (1).
Most striking is the disease onset during childhood or early adulthood and the incomplete penetrance, leaving some individuals unaffected, despite carrying the mutation (1). Interestingly, across all reported COPA syndrome patients, a total of 5 missense mutations within exon 8 and 9 of the COPA gene have been identified, which translates into a 14 amino acid stretch within the WD40 domain of the COPA (COPα) protein. Being highly conserved between species and implicated in proteinprotein interactions, COPA mutants were shown to have impaired binding efficiency to cargo proteins, therefore causing defective retrograde transport (1). As a consequence, ER stress, activation of the unfolded protein response (UPR) and nuclear factor kappa B (NF-κB) pathway activation were suggested pathomechanisms, which have already been linked to lung disease and autoimmunity in other studies (11)(12)(13). Furthermore, Volpi and colleagues described elevated transcription levels of type I interferons (IFNs) and interferon-stimulated genes (ISGs) in peripheral blood of COPA syndrome patients, suggesting a role of type I IFN signalling and a dysregulated innate immune response in disease pathogenesis (3). This finding is supported by two recent case reports showing therapeutic benefit of JAK1/2 inhibition in COPA syndrome patients (10,14).
However, the innate immune sensor as well as the molecular mechanisms underlying COPA syndrome pathogenesis remain unclear.
Due to the involvement of COPA in intracellular trafficking and its subcellular localization, we wondered about a role of COPA in STING (Stimulator of IFN gene, also known as MITA, MPYS and ERIS) (15)(16)(17)(18) pathway regulation. Inactive STING forms homodimers that localize to the ER membrane and has been identified as an adapter protein downstream of multiple intracellular nucleotide sensors including cyclic-GMP-AMP-synthase (cGAS) (19,20). Upon recognition of cytosolic double stranded DNA as a danger-associated molecular pattern (DAMP), cGAS synthesizes 2'3'-cGAMP, a second messenger detected by STING (21). Secondary messenger binding results in a conformational change in STING, which enables translocation to the Golgi apparatus. This is an essential step for downstream signalling to occur (22). At the Golgi compartment, transcription factors such as NF-κB and IFN regulatory factor 3 (IRF3) are activated via IκB kinase epsilon (IKKε) and TANKbinding-kinase 1 (TBK1) in a phosphorylation-dependent manner (23). Translocation of these transcription factors to the nucleus then results in gene expression of proinflammatory cytokines and type I and III IFNs (IFNαβ and IFNλ) (24,25).
Identifying the innate immune pathway activated in COPA syndrome will provide valuable insights in disease pathology leading to beneficial pharmacological intervention for this inflammatory condition.

Cell culture
All used cell lines were obtained from ATCC.

Transient transfection
The pCMV6-Entry-COPA-myc-DDK plasmid (Origene) was used as a template to

Immunoprecipitation of mCitrine-STING for mass spectrometry
Immunoprecipitation (IP) experiments were performed similarly as described (27).
Briefly, 3x10 6 HEK293T cells were seeded in 10cm dishes. The following morning cells were transiently transfected with 10µg of either pEF-BOS EV control or pEF-

Cytokine measurements
To measure protein levels of CXCL10 and IFNβ, THP-1 cells were seeded at 1. Using the GraphPad Prism8 software, statistical comparison was made either by paired ratio t-test or one-way ANOVA as stated in the figure legends (p-value < 0.05 considered statistically significant).

Generation of cellular models for COPA syndrome.
Aiming to study the protein function of COPA in the context of innate immune signalling we used CRISPR/Cas9 genome editing to delete COPA in human monocytic THP-1 and epithelial HeLa cell lines. This approach was based on the assumption that the reduced target protein binding efficiency of loss of function mutations identified in COPA syndrome patients can be mimicked by reduction of wildtype COPA protein levels. We generated three COPA deficient THP-1 pooled cell lines using different single guide (sg) RNAs targeting different exons of the COPA gene. In the monocytic human THP-1 cell line, reduction of COPA protein levels coincided with spontaneous phosphorylation of the transcription factor signal transducer and activator of transcription 1 (STAT1), which is downstream of type I and III IFN signalling (Fig. 1A). Complete deletion of COPA could not be achieved because it is essential for cell survival (34). We proceeded to use the THP-1 cell line with sgRNA 1 (termed COPA deficient ) as it demonstrated the greatest reduction in COPA levels and concomitant increase in STAT1 phosphorylation (pSTAT1).
An elevation of proinflammatory cytokine gene expression levels (IL1B, IL6, IL12, IL4, IL23) has previously been described in B cell lines derived from COPA syndrome patients and a type I IFN signature was reported in patient PBMCs (1, 3).
In order to independently confirm these findings in a different cell line, sgRNA1 was used to generate COPA deficient HeLa cells (Fig. 1D). Similarly, the reduction of COPA protein levels resulted in increased transcription of proinflammatory cytokines and type I IFN (Fig. 1E), as well as elevated baseline phosphorylation of TBK1 (pTBK1), a signalling molecule downstream of several pattern recognition receptors (Fig. 1D). Therefore, we have successfully generated COPA deficient THP-1 and HeLa cell lines that model inflammatory manifestations observed in COPA syndrome as a useful tool for future studies to investigate the molecular mechanisms underlying this disease.
Overactive STING pathway drives inflammatory signalling in COPA syndrome model cell lines.
To identify the innate immune sensor that is driving the inflammatory response in our in vitro models of COPA syndrome, we genetically deleted several candidate innate immune receptors in COPA deficient THP-1 cells (Suppl. Fig. 1). Using this approach, we excluded the involvement of inflammasome sensor NLRP3 (Suppl. Fig. 1A), cytoplasmic RNA sensor PKR (Suppl. Fig. 1B) and RNA-sensors RIG-I and MDA-5 by deletion of shared adaptor protein mitochondrial antiviral-signalling protein (MAVS) (Suppl. Fig. 1C). Furthermore, deletion of Unc93 homolog B1 (UNC93B1), an adaptor protein essential for endosomal TLRs and cell surface TLR5 stability and signalling (35), was not able to ameliorate the inflammatory phenotype in COPA deficient THP-1 cells (Suppl. Fig. 1D).
Another candidate immune sensor is STING, which functions as adapter protein that triggers inflammation downstream of multiple cytoplasmic nucleic acid sensors including cGAS. Inactive STING localizes on the ER membrane and requires anterograde trafficking to translocate to Golgi compartments upon activation (36).
Although the anterograde transport route is mediated by coatomer complex II (COPII) vesicles which COPA is not a part of, all cellular secretory pathways form a network within the cell and are tightly regulated. Therefore, we hypothesized that reduced functionality of retrograde transport might potentially interfere with STING trafficking and prolong signalling. In line with this, we identified COPA as a potential interaction partner for STING in an unbiased mass spectrometry-proteomics experiment, where we analysed protein lysates following pulldown of overexpressed mCitrine (mCit)-tagged human STING in HEK293T cells ( Fig. 2A). This finding was also independently confirmed by Keskitalo and colleagues (37).
CRISPR/Cas9-mediated genetic deletion of STING ameliorated the spontaneous pSTAT1 signal (Fig. 2B) as well as inflammatory gene expression in COPA deficient THP-1 cells (Fig. 2C). Interestingly, TNF transcription levels are not significantly reduced upon STING deletion and are likely the result of NF-κB pathway activation following ER stress, as previously suggested (1). As an independent confirmation for the role of STING, genetic deletion in COPA deficient HeLa cells ameliorated type I IFNmediated inflammatory baseline signalling as assessed by qRT-PCR and immunoblot analysis for pTBK1 (Suppl. Fig. 2). In the context of potential pharmaceutical intervention, treatment of COPA deficient THP-1 cells with STING inhibitor H-151 was able to ablate baseline pSTAT1 (Fig. 2D) and type I IFNmediated gene transcription (Fig. 2E), thus indicating a possible targeted treatment for COPA syndrome patients.
Given the requirement for STING's ER-to-Golgi translocation to activate downstream signalling molecules, we performed immunofluorescence (IF) studies and aimed to investigate whether defective retrograde transport via COPA-deficiency results in STING accumulation at the Golgi, which therefore mediates increased signalling. As positive control, parental HeLa cells were stimulated with HT-DNA, a double stranded DNA molecule to activate cGAS, which results in distinct STING translocation that can be observed as puncta formation and indicates STING activation (Fig. 2F). In COPA deficient HeLa cells without further stimulation, STING did not form clear puncta like the positive control (Fig. 2F). However, considering that loss of COPA is associated with Golgi dispersal (38), localization patterns of activated STING might be significantly aberrant. Overall these results suggest that spontaneous STING activation is driving the inflammatory response in COPA syndrome model THP-1 and HeLa cell lines, although this may involve smaller signalling complexes at dispersed Golgi fragments.

Mutations in COPA drive STING-dependent inflammation.
In order to validate our findings in patient samples, we analysed PBMCs from a COPA syndrome patient carrying the c.698G>A (p.R233H) mutation with clinical presentation of severe polyarticular arthritis and lung disease (3). At the time of sample collection, the patient was treated with Prednisone and Rituximab. Flow cytometry analysis revealed elevated levels of pTBK1, particularly in CD14-expressing monocytes (Suppl. Fig. 3). Treatment with STING inhibitor H-151 was able to ameliorate baseline pTBK1 as shown in representative histograms and quantified as fold change of pTBK1 levels after H-151 treatment (measured by mean fluorescence intensity, (MFI)), thereby confirming basal STING activation (Fig. 3A,   B).
To further study inflammation in the context of COPA syndrome we generated overexpression plasmids encoding previously published loss-of-function COPA mutations E241K and R233H by mutagenesis PCR. We overexpressed these constructs in HEK293T cells which do not express detectable levels of endogenous STING (21,39) and no basal activation of pTBK1 or pIRF3 was observed (Fig. 3C).
Therefore, we co-expressed mCit-tagged STING together with the myc-tagged COPA mutant plasmids and could indeed detect increased levels of pIRF3 (Fig. 3C).
Importantly, these experiments suggest that inflammation driven by COPA mutations in HEK239T cells only occurs in presence of STING, indicating the dysregulation of STING in COPA syndrome.

Deficiency in COPI-mediated retrograde transport activates STING signalling
Besides COPA (subunit α), 6 other subunit proteins COP β, β', δ, ε, γ, ξ as well as small GTPase Arf1 are essential for functional retrograde transport between Golgi and ER and within cis-Golgi compartments (Fig. 4A). The opposite transport direction between ER and Golgi (anterograde transport) is mediated by COPII complex subunits SEC13, SEC31, SEC23, SEC24 and small GTPase Sar1 (Fig.   4A). Here, we sought to determine whether activation of the STING signalling pathway is directly linked to loss of COPA function, or whether disruption of intracellular trafficking routes, both anterograde and retrograde, results in inflammation via this pathway. Therefore, we randomly selected COPI-subunit COPG1 and COPII-subunit SEC13 and used the CRISPR/Cas9 technology to delete these subunits in THP-1 cells.
Interestingly, only deficiency in COPI complex proteins COPA and COPG1 but not the COPII protein SEC13 results in spontaneous phosphorylation of STAT1 and inflammatory gene transcription (Fig. 4B,D). In order to determine whether the inflammatory phenotype can be ameliorated by inhibition of STING signalling, THP-1 Immunoblot as well as qRT-PCR analysis show amelioration of pSTAT1 (Fig. 4C) and reduced transcription of proinflammatory genes and ISGs after inhibitor treatment (Fig. 4D). These findings indicate that indeed, the STING pathway is activated upon deletion of different COPI subunit proteins, however deletion of the COPII subunit protein SEC13 does not result in inflammation.
Collectively, our results demonstrate spontaneous activation of the STING signalling pathway in COPA syndrome. Furthermore, it appears disruption of COPI trafficking in general drives activation of STING signalling. Thus, mutations in other COPI proteins may underlie as yet undescribed disorders with immune/autoinflammatory pathology.

Discussion
In this study, we have generated COPA syndrome model cell lines that recapitulate the IFN signature observed in PBMCs of COPA syndrome patient cells (3). We show that the type I IFN signature can be ablated by genetic deletion and pharmacological inhibition of STING. Interestingly, the clinical phenotype of COPA syndrome partially overlaps with symptoms reported in STING-associated vasculopathy with onset in infancy (SAVI), including severe systemic inflammation, recurrent fevers, interstitial lung disease, early onset in life and a predominant constitutive IFN gene activation (40). For SAVI, multiple case reports identified gain of function mutations in STING, causing spontaneous dimerization and therefore constitutive STING pathway activation (37,41,42). However, STING activation in COPA syndrome is less pronounced, as activated (phosphorylated) STING could not be detected in model cells lines and patient PBMCs (data not shown). Furthermore, we were not able to detect STING translocation to Golgi compartments in COPA deficient HeLa cells, which is thought to be a prerequisite for STING signalling. However, COPA deficient cells undergo partial Golgi fragmentation (38) similarly to cells treated with Brefeldin-A, a small GTPase inhibitor that prevents initiation of COPI and II vesicle formation and disrupts secretory trafficking (43). Therefore, STING signalling might occur off smaller Golgi fragments that could be below the level of detection, rather than from an intact Golgi complex, however this requires further investigation.
In our study we show that the STING pathway is activated upon deficiency of 2 different COPI-subunit proteins COPA and COPG1, but not COPII-subunit protein SEC13. Therefore, one could speculate that there may be other interferonopathies associated with defective retrograde transport caused by loss of COPI subunits other than COPA, that have not yet been described. Indeed, loss of function mutations in COPB2 (COPβ') or COPD (COPδ, encoded by ARCN1 gene) are linked to diseases associated with skeletal developmental defects and microcephalus formation (44,45). Given the manifestation of these conditions, an underlying IFN signature may be present but undocumented. However numerous other factors would be contributing and determine whether a particular COPI-deficiency results in more pronounced developmental or inflammatory pathology. This becomes evident in COPA syndrome, which presents with incomplete penetrance leaving some mutation carriers unaffected. Co-factors contributing to disease onset in individuals with COPA mutation have not yet been identified but could be related to ER stress and the threshold for inflammatory signalling via cGAS/STING.
The finding that retrograde trafficking regulates STING is novel and the underlying mechanisms are not yet fully understood. Notable is that STING does not encode a dilysine motif (KKxx, KxKxx), which has been shown to be required for direct binding of COPA`s WD40 repeat domain to protein cargo (46). Therefore, retrograde transport of STING would require an adaptor protein. This hypothesis is supported by prepublication data from two laboratories, demonstrating that COPA binding to STING is dependent on Surfeit protein 4 (SURF4) (47,48). This also agrees with our proteomic analysis, where SURF4 and COPA were immunoprecipitated with STING after overexpression in HEK293T cells (Figure 2A). Since STING signalling occurs from the ERGIC/Golgi compartments, it is conceivable that COPI-mediated retrograde transport may play an essential role to terminate signalling. However, whether activated STING is packaged into COPI vesicles and partially recycled to ER membranes or completely degraded via autophagy pathways remains to be determined. We were not able to visualize activated STING at the Golgi, likely due to the observed Golgi fragmentation following loss of COPA. The mechanism by which COPA deficiency leads to increased duration of activated STING signalling at the Golgi would also necessitate some initial signal for STING translocation from the ER.
Whether homeostatic STING cycles between ER and Golgi or tonic cGAS activation is required to initially drive STING translocation in COPA syndrome, is currently under investigation.
Overall, this study shows that the type I IFN signature described in COPA syndrome patients is caused by a spontaneous activation of the STING pathway. This finding is particularly exciting in the context of pharmacological intervention. So far, COPA syndrome patients have been treated with various combinations of immunosuppressive drugs including DMARDs (disease-modifying anti-rheumatic drugs), NSAIDs (nonsteroidal anti-inflammatory drugs) and biologics, which were only able to partially control the disease (1, [3][4][5][6][7][8][9][10]. Recently published case reports show treatment of two COPA syndrome patients with Janus kinase (JAK) 1/2 inhibitors Ruxolitinib or Barcitinib (10,14). Although the IFN signature and joint inflammation were reduced successfully, patient lung function was not significantly improved (10,14). Therefore, our study provides hope that targeting the STING pathway directly, for example with an inhibitor such as H-151, may be more specific and thus proves a greater beneficial approach to control type I IFN-mediated symptoms in COPA syndrome.