Cytosolic d‐type CpG‐oligonucleotides induce a type I interferon response by activating the cGAS‐STING signaling pathway

Cytosolic DNA receptor cyclic GMP‐AMP (cGAMP) synthase (cGAS) has been shown to be critically involved in the detection of cytosolic, self‐ and non‐self‐DNA, initiating a type I IFN response through the adaptor protein Stimulator of Interferon Genes (STING) and interferon regulatory factor 3 (IRF3). Current studies propose that canonical binding of dsDNA by cGAS depends on DNA length, but not on base sequence. In contrast, activation of TLR9 is sequence dependent. It requires unmethylated CpG dinucleotides in microbial DNA, which is mimicked by synthetic oligodeoxynucleotides (ODN). Here, we provide evidence that d‐type ODN (D‐ODN), but not K‐type ODN (K‐ODN), bind to human cGAS and activate downstream signaling. Transfection of D‐ODN into a TLR9‐deficient, human monocytic cell line (THP‐1) induced phosphorylation of IRF3 and secretion of IFN. This response was absent in cells with CRISPR/Cas9‐mediated cGAS‐ or STING‐deficiency. Utilizing a protein pulldown approach, we further demonstrate direct binding of D‐ODN to cGAS. Induction of a type I IFN response by D‐ODN was confirmed in human primary monocytes and monocyte‐derived macrophages. These results are relevant to our understanding of self–nonself‐discrimination by cGAS and to the pharmacologic effects of ODN, which currently are investigated in clinical studies.


Introduction
The utilization of synthetic oligonucleotides is currently being studied in a broad range of diseases, including malignancies, allergies, and infectious diseases, and oligodeoxynucleotides (ODN), have been approved as vaccine adjuvants [1,2]. ODN contain unmethylated CpG motifs that are prevalent in bacterial and viral but not in vertebrate genomic DNA [3]. These CpG motifs are recognized by TLR9. Within human PBMC, functional TLR9 signaling in response to pathogens is restricted to B cells and plasmacytoid dendritic cells (PDC), while murine TLR9 is also expressed by monocytes and macrophages [4,5]. Distinct classes of synthetic ODN have been described, depending on their structural and immunostimulatory properties [6][7][8][9]: K-ODN (also referred to as CpG-B) trigger early secretion of IFN-β and production of proinflammatory cytokines, including IL-6 and TNF-α, by PDC mediated by the transcription factors NF-κB and interferon regulatory factor (IRF5) [10]. K-ODN also activate B cells to proliferate and secrete immunoglobulins [11,12]. D-ODN (also referred to as CpG-A) induce high levels of IFN-α secretion from PDC via an IRF7-dependent autocrine/paracrine feedback loop [7,13,14]. Structurally, K-ODN are linear, ss molecules containing multiple unmethylated CpG motifs. D-ODN form a central hairpin structure and have a poly-G tail that leads them to form G-tetrads [15].
While TLR9 binds CpG-containing DNA within endosomes, detection of cytosolic DNA is primarily mediated by the receptor cyclic GMP-AMP (cGAMP) synthase (cGAS). Upon binding of DNA, the protein cGAS induces GTP and ATP to form the second messenger cGAMP. cGAMP then binds stimulator of interferon genes (STING) [16][17][18][19], which triggers phosphorylation of IRF3 via TBK1. IRF3 acts as a transcriptional regulator of inflammatory genes and upregulates IFN-β [20]. Recent studies suggest a major role for the cGAS-STING pathway in both the immune defense against pathogens and in autoimmune diseases [21,22]. The minimal cGAS activation motif is a matter of current scientific debate. Several studies suggest that human cGAS preferentially recognizes long dsDNA, with a minimum length requirement of 40-45 base pairs. The cGAS activation by long dsDNA is sequence independent [23][24][25]. This concept was challenged by Herzner et al., who observed that short stretches of base-paired DNA within ssDNA stem-loop structures of HIV-1 DNA activate cGAS in a sequence-dependent manner [26].
Although TLR9 was shown to be critical for CpG-ODN-induced immune responses, there is much less known about additional effects elicited by cytosolic CpG-ODN. Better knowledge of potential TLR9-independent effects of CpG-ODN could lead to an improved rationale and indication of how to use various types of CpG-ODN in clinical settings.
In this study, we report that transfected D-ODN but not K-ODN initiate a type I IFN response by human monocytes and macrophages that lack functional TLR9. We further demonstrate direct binding of cGAS by D-ODN. Knockdown experiments confirmed that this immune response depends on cGAS-STING signaling and not on TLR9 sensing, suggesting that the cGAS-STING pathway plays a critical role in the cytosolic recognition of D-ODN.

Cytosolic D-ODN induces IRF3 phosphorylation and type I IFN production by TLR9-deficient THP-1 cells
Primary human monocytes do not respond to CpG ODN stimulation of TLR9 [4,27,28] and can even suppress TLR9-induced IFN-secretion by PDC [29,30]. Unexpectedly, we observed dosedependent transcription and secretion of IFN-β by naïve THP-1 cells transfected with D-ODN ( Fig. 1A and B) at levels comparable to cytosolic herring testes DNA (htDNA, a dsDNA molecule known to activate cGAS). We also compared transfected D-ODN to the transfection reagent itself: transfected D-ODN induced significantly higher levels of IFN-β than lipofectamine alone (Supporting information Fig. S1A and B). Significant levels of IFN-β were also induced by p(I:C) and p(dA:dT) which trigger type I IFN responses through TLR3 and RIG-I-like receptors [31][32][33]. To examine whether transfected D-ODN induce downstream mediators of type I IFNs, we also investigated levels of CXCL10 [34]. Again, we found significant secretion of CXCL10 in response to D-ODN (Fig. 1C). Interestingly, low levels of IL-6 secretion were also induced by D-ODN transfection (Fig. 1D).
TLR9-dependent induction of type I IFN by CpG-ODN depends on the activation of IRF5 and IRF7 [3], whereas IRF3 mediates the IFN response to cytosolic dsDNA [20,36,37]. We, therefore, evaluated whether D-ODN activated IRF3 in THP-1 cells. Results showed that cells transfected with D-ODN phosphorylated IRF3 as did the positive controls dsDNA and p(dA:dT) [20,32,38,39] (Fig. 1F, quantification in Supporting information Fig. S1D). In summary, these findings suggest the involvement of cytosolic receptors rather than endosomal TLR9 in the induction of type I IFN by THP-1 transfected with D-ODN.

Transfection of D-ODN induces STING oligomerization
To further define the signaling pathway triggered by cytosolic D-ODN upstream of IRF3, we evaluated the activation of the cGAS-STING pathway. Cytosolic DNA induces synthesis of cGAMP by cGAS, which acts as second messenger, promoting STING activation and oligomerization into a supramolecular complex [40]. HEK cells stably transduced with an mCherry-tagged STING construct (henceforth "HEK Sting") do not express cGAS. Sting oligomerization within HEK Sting cells can be stimulated if cocultured with cGAMP producing cells (cGAMP is transferred between cells through gap junctions) [40][41][42]. HEK cells stably expressing high levels of cGAS (henceforth "HEK cGAS high ") activated HEK Sting cells in the same culture [41,42]. STING oligomerization by HEK cells with low cGAS expression (henceforth "HEK cGAS low ") was significantly increased by D-ODN transfection ( Fig. 2A, quantification in Fig. 2B), but not by K-ODN transfection (Supporting information Fig. S2). Together, these data indicate that cytosolic D-ODN induce synthesis of cGAMP and activate STING, consistent with D-ODN directly activating cGAS.

IFN-β-induction by D-ODN requires cGAS and STING
Since HEK cGAS low cells transfected with D-ODN triggered STING oligomerization in bystander cells, we examined whether cGAS itself was required for sensing transfected D-ODN. WT, cGAS KO, STING KO, and cGAS/STING double-knockout (DKO) THP-1 cells were transfected with D-ODN. Consistent with previous results (

D-ODN but not K-ODN binds to and activates cGAS
The ability of nontransfected D-ODN to induce type I IFN production is sequence dependent. The critical hexameric CpG motif lies at the apex of a stem-loop structure composed of three selfcomplimentary phosphodiester base pairs [6]. This class of ODN also carries a 3'-prime poly-G-tail composed of phosphorothioate (PO) nucleotides. In contrast, K-ODN are ss, linear 12-30 mers with up to three CpG-motifs composed completely of phosphodiester (PS) nucleotides and lacking either a palindromic motif or poly-G tail [7].
We set out to determine whether cGAS/STING activation depends on the class of the transfected CpG ODN. In contrast to htDNA and D-ODN, transfection with K-ODN did not induce significant IFN-β mRNA expression by THP-1 cells (Fig. 4A, left panel). K-ODN similarly failed to induce secretion of IFNβ, CXCL10 (Fig. 4A, middle and right panel) or IL-6 (data not shown). These findings verify that THP-1 cells do not respond to TLR9 agonists [7].
G10 is a D-ODN used in clinical studies targeting allergic diseases and solid tumors [43][44][45][46]. G10 is composed of three central CpG motifs in a 30 mer and contains both a 3' and 5' poly-G-tails on a PO backbone. To investigate whether D-ODN lacking a PS backbone can activate cGAS, we transfected WT and KO cells with G10 and compared the effect to D-ODNs with backbones including PS nucleotides. Significant secretion of IFN-β and CXCL10 was elicited by all of these constructs in the WT but not STING KO or cGAS KO cells (Fig. 4). Stimulation with p(dA:dT) or LPS induced CXCL10 and IL-6 secretion (respectively) in all cell lines, confirming their viability and responsiveness ( Fig. 4B to D). Following transfection of cGAMP, we detected significant secretion of IFN-β and CXCL10 in WT and cGAS KO but not in STING KO cells.
To further corroborate these results, we also studied IRF3 phosphorylation upstream of IFN secretion. Phosphorylation of IRF3 was observed after transfection of both D-ODN and G10 but not K-ODN (Fig. 4E). cGAMP and htDNA were used as positive controls.
Utilizing a protein pulldown approach, we investigated whether these ODN interacted directly with cGAS and whether this interaction was ODN class dependent. THP-1 lysates were exposed to biotinylated DNA constructs which were then captured by streptavidin beads. Biotinylated htDNA, D-ODN, and G10 all pulled down cGAS protein while K-ODN did not ( Fig. 4F and Supporting information Fig. S4). As expected, biotinylated D-ODN and htDNA did not bind cGAS in cGAS-KO cells. This set of experiments shows that cytosolic D-but not K-class ODN binds to and activate the canonical cGAS/STING signaling pathway.

Transfected D-ODN induces type I IFN in primary human monocytes and macrophages
THP-1 is a monocyte-like cell line originally derived from a patient with acute monocytic leukemia. To determine whether primary monocytes from healthy humans are capable of sensing cytosolic D-ODN, we isolated CD14+ cells from healthy volunteers. When transfected with D-but not K-ODN, these cells responded with IFN-β and IL-6 secretion (Fig. 5A). In monocytes, n = 5 for p(I:C); n = 4 for D-ODN 0.03 μg/mL; n = 3 for htDNA 3 μg/mL. (D) n = 9 for naïve control; n = 8 for D-ODN 1 μg/mL; n = 7 for D-ODN 0.1 μg/mL and 0.3 μg/mL; n = 5 for htDNA 1 μg/mL, p(I:C) and p(dA:dT); n = 4 for D-ODN 0.03 μg/mL and htDNA 3 μg/mL p(I:C); n = 3 for D-ODN 3 μg/mL. (E) Naïve THP-1 cells were analyzed for TLR9-expression and (F) for phosphorylation of IRF3 after 24 h of transfection by the indicated ligands (western blot). CAL-1 cells and HEK293 cells with and without TLR9 were used as controls (in E). One representative blot from two (E) or three (F) independent experiments is shown. Statistical significance was determined by paired (A) and unpaired (B-D) student's t-test. *p < 0.05 versus negative control. CXCL10 secretion was significantly higher after transfection of D-ODN compared to K-ODN. cGAMP and htDNA served as positive controls in these experiments. Similarly, IFN-β, CXCL10, and IL-6 secretion was significantly detected in human macrophages derived from CD14+ selected monocytes transfected with D-ODN. (Fig. 5B). Of note, transfection reagent alone or nontransfected D-ODN had no effects on cytokine production. Transfection of all DNA-constructs exerted some cytotoxic effects (Supporting information Fig. S5), as measured by activity of lactate dehydrogenase (LDH) in the supernatant. These effects, however, did not correlate with the capability of DNA-constructs to induce IFN-β.
Consistent with these findings, we observed expression of cGAS, but not TLR9 in human monocyte-derived macrophages by using immunoblot (Fig. 5C and D).
We conclude that the ability to sense cytosolic D-ODN is not confined to immortalized cell lines (HEK293, THP-1).

Discussion
Previous studies showed that D-ODN-driven IFN production was TLR9 dependent (cells lacking TLR9 did not respond). This manuscript suggests that D-ODN can signal via two independent mechanisms of action. One is TLR9 dependent and the other is cGAS/STING dependent.
This work describes a novel signaling pathway through which D-class ODN induces type I IFN secretion. TLR9-negative THP-1 cells transfected with D-ODN were stimulated to initiate robust IFN-β and CXCL10 secretion. Importantly, K-ODN did not activate these cells suggesting that the CpG sequence alone is insufficient to elicit this response. We further demonstrated STING oligomerization in response to transfected D-ODN in TLR9-deficient HEK293 cells. Utilizing cGAS/STING-KO cells, we proved that the cGAS/STING-pathway is required to detect cytosolic D-ODN. Finally, we demonstrate that D-but not K-ODN directly binds to cGAS. These findings were validated in primary human monocytes, increasing their clinical relevance, and represent the first description of CpG-ODN-mediated activation of the cGAS/STINGsignaling pathway.
Current research shows that cGAS binds to the sugar phosphate backbone of dsDNA [47,48]. Indeed, we previously demonstrated that inhibition of cGAS activity by binding the 24-bp ODN A151 is partially mediated by its PS backbone [42]. Consistent with this, we could not detect cGAS activation by K-ODN on a PS backbone (Fig. 4). By comparison, the D-ODNs that activated cGAS both contained stretches of PO nucleotide DNA (Fig. 3  and 4).
Recent studies demonstrated that the backbone structure and length but not sequence of dsDNA determine cGAS:dsDNA interaction and downstream signaling [21,[23][24][25][47][48][49][50]. Jakobsen et al. showed that cytosolic HIV-1 derived ssDNA shorter than 40-bp is also capable of inducing type I IFN [51]. This stimulatory capacity depended on stretches of dsDNA within the stem-loop structure and was mediated by IFI16 [51,52]. Similarly, Herzner et al. demonstrated cGAS-signaling in response to HIV-1-derived short ssDNA [26]. The stimulatory capacity of such short ssDNA was found to be sequence dependent: the minimal cGAS recognition motif (termed "G3-YSD") forms a stem-loop structure due to its palindromic DNA sequence, with a base-paired stem of  12-20 bp of dsDNA flanked by two or three unpaired Gs at either side. Of note, Herzner et al. concluded that intermolecular quadruplex interactions-which are facilitated by longer G-rich sequences-were not involved in G3-YSD recognition. In our study, we used predefined ODN which had been designed for optimal stimulation of TLR9 in primary human PBMC [7,10,53]. Interestingly, all D-ODN contained a palindromic sequence, forming stem-loop structures with a stem of 3-4 bp of dsDNA and 2-9 Gs at either end. In principle, this is consistent with the minimal cGAS activation motif described by Herzner et al., extending their finding that stretches of 12-bp dsDNA are sufficient for cGAS activation to even shorter stretches of dsDNA [26].
In summary, we demonstrate a completely novel mechanism of ODN-sensing by the innate immune system in the cytosol. Our work is the first to establish that cytosolic D-ODN, but not K-ODN, are capable of inducing a robust cGAS/STING-dependent IFN response. Our study also confirms that cGAS is capable of sensing short stretches of dsDNA within ssDNA stem-loop structures. Future studies should be directed toward dissecting the precise requirements for D-ODN sensing by cGAS. Based on our findings, we suggest that activation of cGAS by cytosolic D-ODN depends on a central palindromic sequence, resulting in a stem-loop structure with stretches of dsDNA, flanked by at least two Gs at either side. We cannot rule out the possibility that the ODN backbone also contributes to cGAS activity. Future studies need to investigate whether cGAS-stimulation depends on higher, quaternary structures mediated by longer poly-G tails and Hogsteen effects and whether D-ODN on PS backbones are capable of cGAS signaling.
Some earlier studies also documented TLR9-independent immunological effects of ODN [13,[54][55][56][57]. Based on our data, experimental results obtained with transfected CpG-ODNs [13,56,57] must be interpreted carefully, meticulously distinguishing between TLR9-and cGAS-dependent effects. Future studies need to clarify whether cGAS-stimulation by ODN can be exploited clinically such as for vaccination or cancer therapy. Carefully selected and transfected ODN could potentially combine the activation of B cells, PDC, and macrophages. In the case of tolerogenic ODN, it needs to be elucidated whether cGAS stimulation needs to be avoided. In addition, future studies need to investigate whether highly TLR9-or cGAS-selective ODN can be synthesized, depending on optimal sequence and structure.

Materials and methods
All cell culture and transfection reagents were tested negative for endotoxin (<0.1 EU/mL) by LAL assay.

Transfection of cGAMP, htDNA, and ODN
Endotoxin-free (lack of response in HEK-TLR4 cells) cGAMP (cyclic [G(2',5')pA(3',5')p]) was purchased from Invivogen (San Diego, CA, USA). dsDNA isolated from herring testis (htDNA) was purchased from Sigma-Aldrich (Saint Louis, MO, USA). For some studies, an equimolar mixture of three ODN sequences was used ("D-ODN" and "K-ODN"), as previous studies demonstrated that such mixtures more consistently stimulated cells from multiple donors than did individual ODN. Sequences and backbone structure can be found in Table 1. Endotoxin-free (<0.1 EU/mL at 1 mg/mL) D-ODN (equimolar mixture of D19, D29, and D35) and K-ODN (equimolar mixture of K3, K23, and K123) were synthesized at the Core Facility of the Center for Biologics Evaluation and Research facility, Food and Drug Administration (Bethesda, MD, USA) and kindly provided by D. M. Klinman [6]. A. M. Krieg kindly provided G10.
THP-1 cells with targeted mutations were kindly provided by T. Zillinger. To generate targeted mutations, THP-1 cells were coelectroporated with a gRNA-and a Cas9-expression plasmid for cGAS KO and for STING KO as described earlier [42]. The gRNA target sequences used were GGCCGCCCGTC-CGCGCAACT(GGG) for cGAS KO and CTAGCCCCCAAAGGGT-CACC(AGG) for STING KO (PAM sequence in parenthesis). Sequential KO procedures were performed to create DKO cells.
Capital letters depict bases with phosphorothioate backbone, lowercase letters depict a phospodiester backbone.
All studies using primary cells were performed after written approval from the ethics committee of the medical faculty of the university of Bonn ("Ethik-Kommission der Medizinischen Fakultät," "Rheinische Friedrich-Wilhelms Universität Bonn") and after obtaining written informed consent from the donors. Investigations were conducted according to the principles expressed in the Declaration of Helsinki. Primary human monocytes were isolated from peripheral blood by CD14+ selection, following the manufacturer's protocol (Miltenyi Biotec, Bergisch Gladbach, Germany) and cultured in supplemented RPMI 1640 (including L-glutamine, 10% heat-inactivated FCS, and 1% sodium pyruvate).

Detection of cytokines by ELISA
Commercially available ELISA kits for IFN-β, IL6, and CXCL10 were purchased from R&D Systems (Minneapolis, MN, USA). All ELISA assays were performed according to the manufacturer's instructions.

qRT-PCR
Total RNA was extracted from cells using TRIzol reagent (Thermo Fisher Scientific), as specified by the manufacturer; cDNA was synthesized with the High-Capacity cDNA Reverse Transcription kit (Applied Biosystems, Darmstadt, Germany) according to the manufacturer's instructions. Gene expression levels (normalized to 18s) were analyzed using the ViiA 7 Real-Time PCR system (Applied Biosystems). All reagents and probes used in these studies were purchased from Applied Biosystems. The following Taq-Man assays were used: IFN-B (Hs00985639 m1), 18s (Hs0287368 g1).
Relative amount protein was quantified using ImageJ software (NIH, USA). The amount of protein was measured as the grey mean value within a specified region of interest. Background measurements were subtracted. Finally, the amount of protein was expressed in relation to the corresponding loading control, corrected for background.

Fluorescence microscopy
Coculture experiments of cGAS high and cGAS low HEK293T cells with STING-expressing HEK293T cells were performed as previously published [42]. Briefly, 1.5 × 10 4 cGAS high or cGAS low cells were cocultured with 1.5 × 10 4 STING reporter cells in a 96-well format and stimulated as described above. After 24 h, images were collected using an Olympus IX81 microscope with 20× magnification. STING complexes were automatically counted using the ZEN imaging software (Zeiss, Jena, Germany).

Cytotoxicity assay
Cytotoxic effects of DNA transfection were measured utilizing a commercially available assay (Promega, USA). The colorimetric assay indirectly quantifies the activity of LDH in the supernatant by measuring the production of formazan. Cells treated with lysis buffer were used as positive control.

Statistical analysis
Statistical analysis was performed using GraphPad Prism 5 (GraphPad Software, La Jolla, CA, USA). Data are expressed as mean ± SEM with p < 0.05 considered statistically significant, as determined by two-sided student's t-test and indicated by "*" within the figures, unless stated otherwise.