The GAR/RGG motif defines a family of nuclear alarmins

The nucleus is the target of autoantibodies in many diseases, which suggests intrinsic nuclear adjuvants that confer its high autoimmunogenicity. Nucleolin (NCL) is one abundant nucleolar autoantigen in systemic lupus erythematosus (SLE) patients and, in lupus-prone mice, it elicits autoantibodies early. With purified NCL, we observed that it was a potent alarmin that activated monocytes, macrophages and dendritic cells and it was a ligand for TLR2 and TLR4. NCL released by necrotic cells also exhibited alarmin activity. The NCL alarmin activity resides in its glycine/arginine-rich (GAR/RGG) motif and can be displayed by synthetic GAR/RGG peptides. Two more GAR/RGG-containing nucleolar proteins, fibrillarin (FBRL) and GAR1, were also confirmed to be novel alarmins. Therefore, the GAR/RGG alarmin motif predicts a family of nucleolar alarmins. The apparent prevalence of nucleolar alarmins suggests their positive contribution to tissue homeostasis by inducing self-limiting tissue inflammation with autoimmunity only occurring when surveillance is broken down.


Introduction
Despite of the central tolerance mechanisms, polyreactive B cells remain common in the naive repertoire as potential origins of pathogenic autoantibodies 1,2 . Nuclear antigens are frequently targeted by these autoantibodies 3 . In systemic lupus erythematosus (SLE) patients, pathogenic IgG autoantibodies originate from polyreactive B cells and, in mice, polyreactive B cells induce autoreactive germinal centers to activate autoreactive B cells of broader specificity 4,5 . Aberrant monospecific B-cell somatic hypermutation also gives rise to autoreactive B cells 6 .
Among patients with antinuclear autoantibodies (ANA), 10-15% develop autoantibodies that react predominantly with the nucleolus [7][8][9] . Nucleolar proteins mostly function in ribosome biogenesis and their high autoimmunogenicity are not explained 10,11 . Some ribonucleoproteins (RNPs) contain pro-inflammatory RNA components, which can confer immunogenicity to the protein components 12,13 . The nucleoli contain abundant RNPs but pro-inflammatory nucleolar RNA or proteins have not been reported 11 . Nucleolin (NCL) is an abundant nucleolar protein with multiple RNA-binding motifs 10,14,15 . It is not a known RNP component but it is a major autoantigen in a subgroup of SLE patients 16 . In lupus-prone mice, NCL-specific autoantibodies appeared early before other autoantibodies, suggesting an early pathogenic role 17 .
Besides the nucleolus, NCL is also a nucleocytoplasmic shuttling protein and participates in nuclear import and export 18,19 . It localizes on the cell surface 20,21 , and could be secreted 22 . Cell surface NCL interacts with diverse proteins 20,21 , such as P-and L-selectins 22,23 , and the respiratory syncytial virus 24 , and can mediate protein endocytosis 20,21 . NCL is targeted to the cell surface through cytoskeleton or small vesicles 22,25 . How cell surface NCL is related to autoimmunity is unclear.
Necrotic cells release nuclear alarmins that can cause tissue inflammation and also release autoantigens that activate B cells [26][27][28] . Interestingly, in necrotic cells, the highly autoimmunogenic nucleoli are the dominant targets of complement protein C1q 29 . C1q binds to the nucleolar region where NCL localizes 29 . Binding of C1q to nucleoli activates C1q-associated proteases, which effectively degrade nucleolar proteins including NCL 29,30 . This adds C1q and its associated proteases to the intracellular and extracellular surveillance mechanisms, which function to limit necrotic tissue inflammation and autoimmunity [31][32][33][34] . The importance of this C1q-mediated surveillance is reflected by deficiency of C1q or C1qassociated proteases causing ANA production and severe SLE pathogenesis 35 . With these observations, we asked whether, like high mobility group box 1 (HMGB1) 36 , NCL possesses intrinsic alarmin activity that confers nucleolar immunogenicity.
NCL, FBRL, and GAR1 sequences were cloned in the pcDNA3.1 plasmid with C-terminal HA (Figs. 3A and 5), to transfect the human embryonic kidney 293T cells (ATCC) 38 . After culturing for 48 h in Dulbecco's modified Eagle medium containing 10% (v/v) heat-inactivated fetal bovine serum, L-glutamine (2 mM), and penicillin/streptomycin (100 units/ml) in the presence of 5% CO 2 , cells were homogenized to isolate cytoplasm and TxNE 37 , which were combined to incubate overnight with anti-HA-Agarose (0.3 ml). Beads were washed with the wash buffer (30 ml), and eluted using 3.5 M MgCl 2 , collecting 10 × 0.3 ml fractions. Proteins were dialyzed in PBS and measured using the Bradford reagent (Bio-Rad, Hercules, CA). Endotoxin was monitored using an LAL Endotoxin Assay (GenScript, Piscataway, NJ) and contamination by RNA and other TLR ligands were excluded based on RNase and trypsin digestion ( Supplementary Fig. S1).
Sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and western blotting Protein samples were boiled in the presence of dithiothreitol (10 mM) and, after separation on 12.5% (w/v) SDS-PAGE gels, electro-blotted on PVDF membranes. Blots were blocked for 1 h with 5% (w/v) non-fat milk in TBS-T (50 mM Tris, pH 7.4, 150 mM NaCl and 0.1% (v/v) Tween 20), incubated overnight with antibodies, and after washing incubated for 1 h with HRP-conjugated secondary antibodies. Blots were developed using the Pierce SuperSignal West Pico chemiluminescent substrate (ThermoFisher Scientific).

Blood cell isolation and culturing
Buffy coats were obtained from the Singapore Health Sciences Authority Blood Transfusion Services from healthy blood donors with institutional ethics approval. Peripheral blood mononuclear cells (PBMC) were isolated using Ficoll-Paque (GE Healthcare). PBMC (1 × 10 7 /ml) were re-suspended in RPMI medium containing 5% (v/v) bovine calf serum (HyClone) and cultured for 1 h to harvest adherent monocytes, which were re-suspended at 1.5 × 10 6 cells/ml. Monocytes were cultured with macrophage colony-stimulating factor (20 ng/ml) to generate macrophages, and dendritic cells (DC) were cultured using granulocyte-macrophage colony-stimulating factor (20 ng/ml) and IL-4 (40 ng/ml) 38 .

NF-κB luciferase assay
TLR activation was measured using an NFκB-based luciferase assay (Promega). Firefly luciferase was expressed under the NF-κB gene promoter (p5xNFκB-Luc, Stratagene, San Diego, CA) and Renilla luciferase was expressed under the CMV promoter (pRL-CMV, Promega) 39 . 293T cells were transfected in 24-well plates with these luciferase vectors and co-transfected with TLR and co-receptor vectors 39 , each at 0.1 μg/well using TurboFect (ThermoFisher Scientific). After 24 h, cells were harvested and stimulated in 96-well plates with purified proteins, synthetic peptides, or microbial TLR ligands. NFκBmediated firefly luciferase activity was measured and normalized to Renilla luciferase activity in each well and expressed as relative NFκB activation.

TLR2-binding assay
Ninety-six-well plates were coated with purified proteins (1.0 μg/well) and blocked for 1 h with PBS containing 1% (w/v) bovine serum albumin (PBS-BSA). TLR2 was diluted (0.375 to 6 μg/ml) to incubate with the plates overnight at 4°C and bound TLR2 was detected by incubating for 1 h with a mouse anti-His antibody (Sigma-Aldrich) and then 30 min with an HRP-conjugated secondary antibody (DAKO, Glostrup, Denmark). TLR2 was also coated (2 μg/ml) to incubate with NCL, NCL-HA, or the NCL peptides. Bound NCL was detected using a rabbit anti-NCL antibody and bound NCL-HA and its mutants were detected using a mouse anti-HA antibody (1 μg/ml), followed by HRP-conjugated secondary antibodies. Plate-coated TLR2 was also pre-incubated with mouse TLR2 or TLR4-blocking antibodies (5 μg/ml) before incubation with NCL or NCL-HA, and bound NCL or NCL-HA was detected using a rabbit anti-NCL antibody. Coated TLR2 was incubated with biotin-Ahxtagged peptides and the bound peptides were detected with streptavidin-HRP. Plates were all developed using the 3, 3′, 5, 5′-Tetramethylbenzidine (TMB) substrate solution (ThermoFisher Scientific).

Statistical analysis
All experiments were performed in triplicates. Data were representative of three independent experiments and presented as mean ± SD. Statistics was performed by oneway ANOVA or student t-test to compare two groups using the Prism software (GraphPad Prism 7). p-values were indicated by *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. p < 0.05 was considered statistically significant.

NCL activates leukocytes
Purified NCL was coated to stimulate PBMC, using HMGB1 as a control (Fig. 1A) 36,40 . TxNE was also applied to immobilized non-immune mouse IgG1 (Ms IgG1) and equivalent elution was used as a control. The protein-free 10 th fractions eluted from the affinity resins (E10) were also combined as a control. Plate-coated NCL and HMGB1 induced TNFα and IL-1β from PBMC (Fig. 1B, C), and from monocytes (Fig. 1D, E). The E10 fraction and IgG1 elution showed no cytokine induction (Fig. 1B, E). NCL also activated macrophages and DC (Fig. 1F, G). It consistently induces more cytokines than HMGB1, which is a ligand for TLR2, TLR4, and TLR5 40,41 . Although NCL binds to RNA, RNase digestion of plate-coated NCL showed no impairment in PBMC activation (Supplementary Fig. S1B), but trypsin digestion abolished NCL activation of PBMC ( Supplementary Fig. S1C). Trypsindigested TLR ligands, i.e., LPS, LTA and poly I:C, showed no reduction in PBMC activation.
NCL and HMGB1 were compared with LPS in TNFα and IL-1β induction ( Fig. 1H-J). Both NCL and HMGB1 induced an early IL-1β surge and a late TNFα surge (Fig. 1H, I). TNFα production was partially blocked by an IL-1β-specific antibody, suggesting autocrine IL-1β stimulation of monocytes ( Supplementary Fig. S2). LPS did not induced the type of early IL-1β or late TNFα surge, which were observed with NCL and HMGB1 stimulation (Fig. 1J). The similar cytokine induction by NCL and HMGB1 suggests their activation of similar receptors which, for HMGB1, are TLRs.

TLR2 binds to the NCL GAR/RGG motif
To determine whether TLR2 binds to the NCL GAR/ RGG motif, purified NCL, NCL-HA and NCL(649)-HA were coated to incubate with TLR2. TLR2 bound to NCL and NCL-HA in dose-dependent and saturable manners but it lacked binding to NCL(649)-HA (Fig. 3C). When TLR2 was coated to incubate with NCL-HA and its mutants, NCL(649)-HA and NCL(522)-HA, NCL-HA but not the two mutants bound to TLR2 (Fig. 3D). If coated TLR2 was pre-blocked with an anti-TLR2 antibody (Fig. 2E), it no longer bound to NCL or NCL-HA (Fig. 3E). The TLR4-blocking antibody showed no inhibition (Fig. 3E), suggesting that the NCL GAR/RGG motif binds to monocyte surface TLR2 to cause its activation.
Other NCL-HA mutants were also coated and TLR2 bound to the alarmin-active NCL(698)-HA but not the alarmin-inactive NCL-HA mutants (Fig. 3A, F). NCL (670)-HA contained a residual GAR/RGG region but it only showed weak TLR2 binding (Fig. 3A, F). Therefore, TLR2 only binds to the GAR/RGG motif on NCL. We then asked whether synthetic GAR/RGG peptides would also bind to and activate TLR2.
both proteins (Supplementary Fig. S1C). Collectively, our data suggest that the GAR/RGG motif is an alarmin motif, which may predict more nuclear alarmins 44 .
Our preliminary data showed NCL, FBRL and HMGB1 release by UV-induced necrotic cells ( Supplementary Fig. S6A) 29 . NCL purified from the culture was similar to TxNE-derived NCL in monocyte activation ( Supplementary Fig. S6B), suggesting a common pathway by which these nuclear alarmins signal tissue injuries. A FBRL sequence showing a GAR/RGG and a GAR/RG motif. Three peptides were synthesized as indicated and overlapping residues were highlighted (orange font). The GAR/RGG motif was deleted to generate the FBRL(Δ8-64)-HA mutant. B FBRL-HA and FBRL(Δ8-64)-HA were purified and plate-coated (40 μg/ml) to stimulate PBMC for 24 h and TNFα production was measured by ELISA. C PBMC were stimulated for 24 h with the FBRL peptides (10, 50, or 200 μg/ml) and TNFα production was determined by ELISA. D GAR1 sequence showing two GAR/RGG motifs. E GAR1-HA (40 μg/ml) was purified and coated to stimulate PBMC and TNFα production was determined by ELISA. F Comparison of FBRL, GAR1, and NCL in alarmin activities. The purified proteins were plate-coated (40 μg/ml) to stimulate PBMC for 24 h. TNFα production was determined by ELISA. Cell controls, no stimulation. Triplicate experiments were performed to obtain data as mean ± SD. Data was analyzed by one-way ANOVA *p < 0.05, ****p < 0.0001, n.s., not significant.
Here, we report that NCL activates TLR2 and TLR4 ( Supplementary Fig. S4). These surface TLRs generally lack access to endosomal alarmins.
While apoptotic cells maintain plasma membrane integrity and conceal nuclear antigens and alarmins, these intracellular components are variably exposed by necrotic cells 28,31 . We previously showed that the nucleolus was progressively exposed by necrotic cells 29 . We show here that NCL is released by UV-induced necrotic cells and the released NCL can activate monocytes ( Supplementary Fig. S6). Like NCL, FBRL and HMGB1 were also released by these UV-irradiated necrotic cells (Supplementary Fig. S6).
Cell surface NCL presents a challenge to the above proposition as it is not hidden from cell surface TLRs and autoantibodies in patients. It is unclear whether cell surface NCL exposes its GAR/RGG motif in sufficient density to activate TLRs on neighboring cells. Soluble NCL activated monocytes much more weakly than platecoated, multivalent NCL ( Supplementary Fig. S5). Surface NCL may only activate TLRs when it is highly expressed on live cells but necrotic cell nucleoli naturally express abundant NCL 29 . Surface NCL could surge transiently, e.g., during serum-induced cell growth 22 , which could help recruit immune or stem cells 49,50 . Surface NCL undergoes extensive modifications, including phosphorylation and glycosylation 25,51 , which may also modify its alarmin activity.
The discovery of increasing nuclear proteins with alarmin activities suggests hierarchical signals associated with tissue injury that orchestrate multiple homeostatic and defense responses 52,53 . One physiological benefit could be the promotion of CD8 T cell immunity when DC capture necrotic cells caused by viral infections or malignancy [54][55][56] . However, autoimmunity can occur when necrotic cells breach immune surveillance 27,57 . For example, necrotic cells release HMGB1-associated nucleosomes that elicit ANA and cause lupus-like diseases 58 . During SLE disease flares, nuclear antigens surge in the blood, suggesting excessive necrotic cell death 59,60 . On the contrary, HMGB1 also dampens TLRmediated macrophage activation to reduce the necrosis of these tissue scavengers 61 . Tissue inflammation also recruits complement, which is a self-limiting humoral scavenging system for effective tissue cleansing 34 . Through inducing tissue inflammation, alarmins can recruit diverse cell type to sites of tissue injury 62 , including stem cells, which are required for tissue repair 49,50 .

Ethics statement
The contents of this manuscript are the authors' own work, which have not been previously published elsewhere. Buffy coat samples were collected from donors with informed consent. This study was approved by the National University of Singapore Institutional Review Board (Reference code: H-19-031E).

Conflict of interest
A related patent application (no. 10202006656Q) has been submitted by S.W. and J.L.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.