The CN-index of hnRNP-DL Predicts an Individual “Window of Treatment Success” in RA Patients

Background There is a need for biomarker to identify patients ‘at risk’ for rheumatoid arthritis (risk-RA) and to better predict the therapeutic response and in this study we tested the hypothesis that novel native and citrullinated heterogeneous nuclear ribonucleoprotein (hnRNP)-DL autoantibodies could be possible biomarkers. Methods Using Protein macroarray and ELISA, epitope recognition against hnRNP-DL was analysed in sera from different developed RA disease and diagnosed SLE patients. Toll-like receptor (TLR) 7/9 and myeloid differentiation primary response gene 88 (MyD88)-dependency were studied in sera from murine disease models. HnRNP-DL expression in cultivated cells and synovial tissue was analysed by indirect immunouorescence, immunoblot and immunohistochemistry. Responder RA patients who had received MTX or α-TNF inhibitor therapy (Enbrel®). A-C, α-DL were measured by ELISA in patient sera from the EIRA cohort treated with MTX (n = 192) with 161 EULAR Responder and 31 EULAR non-Responder among 6 months. Above these values ROC analyses were performed for detecting DAS28 therapy response. D-F, α-DL were measured by ELISA in patient sera from the Predict cohort treated with Enbrel® therapy with 6-month EULAR response data (n = 94, responder n = 63, non-Responder n = 31). Based on the signals, ROC analysis was performed for detecting DAS28 therapy response. scleroderma; Sjö: Sjögren´s syndrome; PsA: psoriasis arthritis; MB: ankylosing spondylitis; OA: Osteoarthritis; HAM/TSP: HTLV-1-associated myelopathy/tropical spastic paraparesis; MCTD: mixed connective tissue disease; SE: shared epitope; PC: parenchymal changes in the lung; MTX: methotrexate; Interleukin; TNF: tumor necrosis factor; TLR: Toll-like receptor; MyD88: myeloid differentiation primary response gene 88; SIGIRR/ Single IL-1-related receptor/ Toll/interleukin-1 receptor 8; SIGIRR agonist; SDS–PAGE: ATXN:


Abstract Background
There is a need for biomarker to identify patients 'at risk' for rheumatoid arthritis (risk-RA) and to better predict the therapeutic response and in this study we tested the hypothesis that novel native and citrullinated heterogeneous nuclear ribonucleoprotein (hnRNP)-DL autoantibodies could be possible biomarkers.

Methods
Using Protein macroarray and ELISA, epitope recognition against hnRNP-DL was analysed in sera from different developed RA disease and diagnosed SLE patients. Toll-like receptor (TLR) 7/9 and myeloid differentiation primary response gene 88 (MyD88)-dependency were studied in sera from murine disease models. HnRNP-DL expression in cultivated cells and synovial tissue was analysed by indirect immuno uorescence, immunoblot and immunohistochemistry.
Results hnRNP-DL was highly expressed in stress granules, citrullinated in the rheumatoid joint and targeted by autoantibodies either as native or citrullinated proteins in patient subsets with different developed RA disease.
Structural citrullination dependent epitopes (SCEs) of hnRNP-DL were detected in 58% of the SLE patients although 98% of these sera were α-CCP2-negative. To obtain a speci c citrullinated signal value, we subtracted the native antibody value from the citrullinated signal. This CN DL (Citrullinated-Native-hnRNP-DL)-index identi ed and the bioinformatic value was explored, as a new value for an "individual window of treatment success" in early RA and for the detection of RF-IgM/α-CCP2 seronegative RA patients (24-46%). Negative CN DL -index was found in SLE patients, risk-RA-and early RA-cohorts such as EIRA where the majority of these patients are DAS28responders to methotrexate (MTX) treatment (87%). High positive CN DL -values were associated with more severe RA, shared epitope and parenchymal changes in the lung. Speci cally, native α-hnRNP-DL is TLR7/9-dependent, associated with pain and ROC-analysis revealed an association to initial MTX or etanercept treatment response, especially in seronegative RA patients.

Conclusion
CN DL -index de nes patients is a possible biomarker for develop RA and the "window of treatment success" thereby potentially closing the sensitivity gap.
Autoantibodies against hnRNP-A2/B1 (RA33) occur in about 20-40% of RA, SLE and mixed connective tissue disease (MCTD) patients [16]. Autoantibodies to hnRNP-A1 can be found in RA, SLE and MCTD, but probably are cross-reacting α-hnRNP-A2/B1 antibodies [17]. Although hnRNP-A2/B1 is citrullinated in the rheumatoid joint, and it can be targeted either as a citrullinated and or native protein in distinct subsets of RA patients [18]. Previously, we have described autoantibodies directed to the TNFα regulatory protein hnRNP-D (AUF1) to occur in 33% of SLE, 20% of RA, and 17% of MCTD patients [19]. Although predominantly localized in the nucleus, hnRNPs are exported additionally into the cytosol, where they form new autoimmune target structures in stress granules, P-bodies or RNA transport particles [19][20][21].
The hnRNP-D-like protein (hnRNP-DL) protein, which is also known as JKTBP, is related to the autoantigen hnRNP-D/AUF1. Due to its binding properties and structural features [22], hnRNP-DL,-D and -AB-form the D-subgroup of hnRNPs. These proteins exhibit a modular structure and conserved residues, two adjacent RNA binding domains (RBD) followed by a glycine-rich C-terminal auxiliary domain. However, they are very distinct in each of the unique N-terminal regions [23,24].
HnRNP-DL acts as a transcription factor [25], participates in metabolism and biogenesis of mRNA [3], is able to shuttle between the nucleus and the cytoplasm and binds both to nuclear and cytoplasmic mRNAs [24], especially when containing AU-rich elements (AREs) as found within the 3'-UTR of many proto-oncogenes and cytokine mRNAs [26,27]. Up to now, three alternatively spliced hnRNP-DL transcript variants have been described, hnRNP-DL isoform 1-3, whereas proteins only were described for isoform 1 and 2 [23]. Splenocytes from pristane-primed rats restimulated with hnRNPs (-A1,-A2/B1 and -A3) induce a highly in ammatory and erosive arthritis in naïve recipient rats [6]. Furthermore, human TNFα-transgenic mice, which develop a massive erosive in ammatory polyarthritis, generate α-hnRNP-autoantibodies [28]. This supports the hypothesis of a pathogenic role of native hnRNPs in erosive arthritis and suggests that autoimmunity to nucleic acid-associated autoantigens has the potential to contribute to RA development [18]. HnRNPs may also induce pro-in ammatory cytokines, relevant for arthritis development in rats, which involve TLR7 and TLR9 but not TLR4[6].
For α-hnRNP-A2/B1, clinical associations have already been shown for RA severity, with antibodies against the citrullinated protein occurring more frequently in erosive RA and antibodies against the native protein in milder disease [18,29]. For citrullinated peptides it has already been shown that the formation of a delta value with the corresponding arginine peptide increased diagnostic sensitivity and indicated association to shared epitope (SE) [30].
In our study, the delta value of ELISA signals was evaluated as a possible biomarker to obtain a new clinical value, as the difference between the α-citrullinated and α-native protein value. hnRNPs were further investigated in the immunopathogenesis of RA, demonstrating the clinical relevance of autoantibodies, for predicting therapeutic success, early parenchymal changes in the lung, and SE in RA. For the rst time, structural epitopes resulting from the citrullination process were investigated.

Material And Methods
This paragraph is integrated in the Additional le 1.

Results
Protein macroarray screening identi es the hnRNP-DL protein as a novel autoantigen targeted in rheumatoid arthritis (RA).
Sera from 26 RA patients and 40 control subjects, including osteoarthritis (OA) patients (n = 20) and self-reported healthy blood donors (n = 20), were analysed on protein macroarrays [31]. The 20 most sensitive autoantigens only found in the RA group are listed in the Additional le 2.
We identi ed α-hnRNPDL with second highest intensity score. HnRNP-A2/B1 and hnRNP-D (AUF1), have already been described as autoantibody targets in RA [11,19]. Structure of hnRNP-DL and sequence alignment with hnRNP-D is shown in Additional le 1: supplementary Fig. 1. One of two different hnRNP-DL clones, expressing the protein fragment from amino acid 81 to 420, revealed autoantibody reactivity in 20% of RA sera (Additional le 2). This hnRNP-DL fragment was termed hnRNP-DL mir (major immunogenic region). Isoform hnRNP-DL2 (amino acid 120-420) could not be detected by RA sera.
Autoantibodies against native and citrullinated hnRNP-DL are predominantly present in sera of systemic lupus erythematosus (SLE) and RA patients.
To verify the results from protein macroarray screening, hnRNP-DL mir was expressed in E. coli BL21(DE3)pLysS, puri ed and tested for reactivity in ELISA as native (DL) and citrullinated protein version (cit-DL), using 1010 sera obtained from Risk-RA cohort (n = 71), from early RA cohorts (LURA n = 106; EIRA n = 404), from an established RA cohort (Predict n = 127), control cohorts of other autoimmune diseases (n = 216) and from healthy controls (n = 86).The majority of α-DL was found in sera of patients with SLE (34%) and RA (6-21%) and in patients with psoriasis arthritis (15%), patients with MS (5%) and scleroderma (5%) as well as healthy controls (2%) (  Table 1). Although α-cit-DL signals of seronegative patients were lower than those of seropositive patients, they were still signi cantly higher than in other diseases in EIRA and Predict cohort (Additional le 1: supplementary Table 2). Noticeable 58% of the SLE patients in comparison to other diseases were α-cit-DL positive (α-DL 18%), although 98% of the tested SLE sera were α-CCP2-negative. We determined the difference between the ELISA signals, to get a value that describes the relationship between α-cit-DL and α-DL. This value we named CN DL   Anti-DL autoantibodies were detectable in early RA. Therefore, we investigated aCCP2-positive healthy subjects with musculoskeletal symptoms, classi ed as Risk-RA cohort, differentiating between subjects developing arthritis during follow-up and those remaining healthy without arthritis diagnosis. Further we analysed α-DL autoantibody association with certain risk factors for RA. We plotted respectively α-cit-DL, α-DL and the CN DLindex in the LURA cohort with the parenchymal changes in the lung and in the EIRA cohort with the genetic risk factor shared epitope.
In the Risk-RA cohort α-cit-DL and CN DL -Index were signi cantly elevated in progressors ( Fig. 2A), in the LURA cohort in patients with parenchymal lung changes ( Fig. 2B) and in the EIRA in patients with shared epitope, particularly in those carrying two copies (Fig. 2C). No signi cant differences were found for α-DL antibodies.  Mann-Whitney U test was performed for analysing signi cance of indicated groups (*p < 0.05, **p < 0.01, ***p < 0.001 ****p < 0.0001). OD, optical density; nm, nano meter; ns, not signi cant; PC, parenchymal changes in lung; SE, shared epitope High α-DL autoantibody levels found in 6-month EULAR responders for MTX or Enbrel® treatment We examined our biomarkers (α-cit-DL, α-DL and CN DL -index) with therapy data of the EIRA and Predict cohort.
Because α-DL correlated negatively to the CN DL -Index in the seronegative group (Additional le 1: supplementary   Fig. 3), we analysed MTX-treated EIRA patients with negative CN DL -index. 87% of these patients were responders.
In Predict cohort (Enbrel®-treatment) no CN DL -index/response association were found since all patients had equally high positive CN DL -index and none of them negative values. ROC analysis of α-cit-DL or CN DL -index showed no speci c response cutoff. But with α-DL we identi ed 23% of the EIRA patients as MTX responder and in the seronegative group 25% (90% speci city). Among the established RA cohort (Predict) α-DL reached 13% sensitivity and even 25% within the seronegative group for the detection of Enbrel® response (100% speci city;  High baseline titer against α-hnRNP-DL mir (DL) is rather present in 6-month EULAR Responder RA patients who had received MTX or α-TNF inhibitor therapy (Enbrel®). A-C, α-DL were measured by ELISA in patient sera from the EIRA cohort treated with MTX (n = 192) with 161 EULAR Responder and 31 EULAR non-Responder among 6 months. Above these values ROC analyses were performed for detecting DAS28 therapy response. D-F, α-DL were measured by ELISA in patient sera from the Predict cohort treated with Enbrel® therapy with 6-month EULAR response data (n = 94, responder n = 63, non-Responder n = 31). Based on the signals, ROC analysis was performed for detecting DAS28 therapy response. OD, optical density; nm, nano meter; vs., versus; RA, rheumatoid arthritis; MTX, Methotrexat; seropos., rheumatoid factor IgM and/or α-CCP2 positive patients; seroneg., rheumatoid factor IgM and α-CCP2 negative patients.
Anti-cit-DL and α-DL increase the serodiagnostic sensitivity in early RA.
All RA cohorts were analysed to determine diagnostic sensitivities of α-cit-DL and α-DL, in RF IgM/ α-CCP2seropositive and -negative patients.
The calculated cutoff versus healthy controls (96% speci city), identi ed 80% of the subjects in the Risk-RA cohort, which are exclusively aCCP2-positive. In the LURA/ EIRA cohort 32/73% of the seronegative patients were identi ed. In the Predict cohort all patients could be identi ed with one of our biomarkers and α-DL response was on average the lowest (6%). In SLE 84% in total were detected (α-DL: 34%; α-cit-DL: 80 %). In other autoimmune diseases about half (48%) of the patients were detected in total with our biomarker set.
Using the cutoff versus other diseases (96% speci city), we detected 51% of the seronegative established RA patients and 8-17% of the early RA patients (Table 1). A nity-puri ed α-DL autoantibodies from RA patient sera were used for localization of hnRNP-DL in HeLa-and HEp-2 cells. Sparing the nucleoli in interphase cells, staining with the α-DL autoantibodies showed a nucleoplasmic staining with large speckles (Fig. 4A-a-b). However, the nucleoplasmic staining produced by α-hnRNP-D and α-hnRNP-A2/B1 antibodies was more homogeneous (Fig. 4A-e-f) and stained as well as α-DL autoantibodies discrete cytoplasmic foci when cells were stressed by arsenite (Fig. 4A-c, -e-f). Notably, the colocalization experiment showed α-DL antibodies stained a subset of cytoplasmic stress granules (Fig. 4A-c), independent of size and localization. hnRNP-D could be detected in nearly all granules ( Fig. 4A-g, yellow), like the controls Ataxin2 and RCK/p54 (Fig. 4A-d/h).
Since previous studies demonstrated hnRNP-A2/B1 and hnRNP-D to be highly expressed in synovial tissue of RA patients and arthritic mice [19,28,32,33], we analysed the expression of hnRNP-DL in the human joint. Speci c rabbit antibodies recognizing hnRNP-DL 1 and − 2 expression were tested by immunohistochemistry in synovial tissue of RA and OA patients and from healthy controls (Fig. 4B). These analyses revealed hnRNP-DL to be highly expressed in RA tissue. Nuclear and cytoplasmic expression was seen in cells of RA synovial tissue, in contrast to the exclusive nuclear staining observed in OA and normal tissue (Fig. 4B, arrows).
We further investigated the expression of hnRNP-DL under in ammatory conditions in IL1α-and TNFα-stimulated HepG2-, as well as in IL6-stimulated HeLa cells by immunoblotting (Fig. 4C). TNFα and particularly IL1α upregulate, whereas IL6 downregulates the expression of hnRNP-DL and furthermore induces its degradation.
We further detected citrullinated proteins of the same molecular weight as of hnRNP-DL (Fig. 4D) in the synovial tissue. Anti-DL in animal models of RA and SLE with association to TLR7/9 and MyD88 -supports reference to clinical pain Anti-citDL/α-DL autoantibodies, in baseline samples, are associated with pain VAS after 6 months of various treatments of EIRA patients; Additional le 1: supplementary Table 3-5).
Therefore, we wanted to study the production of α-DL autoantibodies in the context of TLR and MyD88-knock-out mice, known to be involved in pain pathway [34,35]. Because hnRNP-DL is highly conserved in human and mouse Page 11/22 (similarity 98,5%[36]; Additional le 1: supplementary Fig. 1), we analysed α-DL in sera of mouse models of RA and SLE (Table 2).
Interestingly, in the interleukin-1 receptor antagonist-de cient (IL-1Ra −/− )-mouse arthritis model we found high signals of α-DL autoantibodies in all mice tested.
We analysed sera from TLR7-, TLR9-and TLR7/TLR9-double de cient lupus-prone MRL/lpr-mice. This investigation revealed that α-DL autoantibodies were TLR7/-9 dependent and only completely absent in the double de cient mice, while they remained detectable in about 50% of the single TLR7-or TLR9-knockout MRL/lpr-mice. MyD88 plays a central role in TLR-pathway [40]. We tested MyD88-de cient mice, which did not produce α-DL autoantibodies except two mice with very low titer. Further we tested knock-out mice of Toll interleukin-1 receptor 8 (TIR8, SIGIRR, IL1R8), a negative regulator of TLR-IL1-receptor family signaling. Genetic inactivation of this protein, which is associated with severe autoimmunity and high autoantibody production [41], increased prevalence of α-DL autoantibodies by 50%, with a three times higher mean level of ELISA signal intensity ( Table 2).

Discussion
RA antibody systems are remarkably diverse, characterized by the presence of those against native proteins as well as those containing posttranslational modi cations (PTMs) [42,43]. While current models of RA have embraced PTMs as core principles of pathogenesis [42,44], α-native protein antibodies are not adequately explained by the PTM-centric paradigm of autoantigen selection. The direct α-citrullinated protein-antibody response may depend on the presence of permissive factors, i.e. a genetic predisposition, as has been shown for α-cit-DL with its shared epitope (SE)-dependency and the continued production of modi ed antigen. Chronic bacterial infection, such caused by Aggregatibacter actinomycetemcomitans [45] or Porphyromonas gingivalis, which can citrullinate hnRNPs [46] or smoking [47,48], leading to overexpression of hnRNPs, as shown by our results with overexpression and citrullination of hnRNP-DL in RA joint.
In early RA, a serodiagnostic gap of 50-60% [49][50][51][52] left by using RF IgM/ α-CCP2 assays. This is of particular importance as patients considered to be autoantibody negative may erroneously not be diagnosed as having RA due to inappropriate therapeutic measures. In recent years, novel biomarkers have been described with sensitivities between 16-67% in α-CCP-negative RA cohorts [53]. However, the clinical utility of these biomarkers is questionable because diagnostic speci cities are largely unknown and will have to be shown in further studies.
It has already been published that it is important to study the citrullinated signal adjusted from the unmodi ed protein/peptide signal, to obtain the speci c signal, which is added or reduced by the modi cation. It has been shown that these autoantibodies occur speci cally in RA, but without clinical associations such as therapy response [30,46]. Therefore, we have introduced and tested a new biomarker CN DL -index which measures the difference of α-cit-DL and α-DL ELISA levels, covering both antibodies against citrullinated epitopes and structural citrullinated epitopes (SCEs). Negative CN DL -index was detectable at an early timepoint of arthritis and even before arthritis starts.
Moreover, RA patients with such negative CN DL -index tended to respond positively to MTX/Enbrel® therapy. As RA progresses, the CN DL -index became increasingly positive and was associated with SE, parenchymal changes in lung and lower the response to MTX therapy.
Citrullination is a hydrolytic reaction, the target protein mobility in SDS-PAGE will shift, yielding a noncharged citrulline amino acid and neutral urea through the hydrolysis of the strongly basic positively charged side chain of arginine by water. This charge shift affects protein structure, protein-protein interactions, and hydrogen bond formation, and it may cause protein denaturation [54,55] This study suggests an alternative model to the PTM centric model in which the antigen is initially targeted independent of citrullination, but may be depend on a structural change induced by cryptic PTM that causes the autoantibody binding. Demonstrated by 58% detection of the SLE patients were α-cit-DL positive (α-DL 18%) versus other diseases, although 98% of the tested SLE sera tested α-CCP2-negative. Citrullination leads to formation of a new SCE, whose recognition is independent of directly targeting the citrulline site. This new form of α-SCE-autoantibodies may explain the shift from an initial native autoantibody response against PTMs. DNA, RNA and TLR7/9 activation are required to generate α-hnRNPspeci c B cells and this complex induced RA in a pristane-induced arthritis model of RA [56]. Interestingly MyD88 de ciency leads to reduction of pain [34], which may explain the correlation of α-DL with pain-VAS after six months in the EIRA cohort. Autoantibodies against DL did not correlate to RF IgM or α-CCP2 or SE. These antibodies can be used speci cally in the seronegative group to predict the therapeutic outcome and pain level after six months of treatment.
The α-DL autoantibody level disappeared in the course of RA, inversely the α-cit-DL autoantibody level increased, independently from the therapeutic regime. Therefore, future therapies should use tolerance induction with the native RA autoantigens in 'high-risk' individuals to inhibit the epitope spreading to PTM autoantigens, as we have found the α-DL autoantibodies in major models of SLE/RA.
Native antigens as part of stress granules are used in existing models of experimental arthritis to induce arthritis, but not the citrullinated antigens[56] SE and speci c exogenous factors are missing in the studied animal models of RA and SLE, explaining the lack of ACPAs and SCE autoantibodies. Anti-native protein antibodies may represent markers for the detection of risk people in the earliest pre disease of RA, preceding the development of the ACPA response, predicting a mild disease. For α-hnRNP-A2/B1 autoantibodies has already been published that it is also associated with less erosive disease, exclusively in early RA [18,29]. Recently, several more reactivities against native proteins in RA have been published [43].
Therefore, it is important to measure other hnRNP-autoantibodies and in combination in future studies to evaluate them for personalized medicine.

Conclusions
We suggest that hnRNP-autoantibodies produced by patients with systemic autoimmune diseases target mRNA decay complexes which are part of stress granules. We hypothesize that increased formation and structural modi cation via bacterial or human enzymes of such protein complexes (e.g., in in ammatory processes with overexpression of IL1α and/or TNFα). This may lead to pathogenic autoimmune reaction against a structurally changed native hnRNP-DL introducing the SCE epitopes before PTM speci c targets increase of time. To summarize, introducing CN-index biomarker measuring the speci c anti-citrulline signal in autoantigens will help to facilitate objectively early RA treatment decisions, not measurable by commercial ACPA assays. Declarations Thanks to Prof. Gerhard Krönke from Erlangen, which provided additional sera of risk-RA.
We wish to thank EIRA study participants and the EIRA study group for their contributions; Professor Lars Klareskog for establishing the EIRA study: and for support and scienti c input; and Sarah Ohrendorf for comments reading the manusnscript. The study has been performed in compliance with the Declaration of Helsinki, with informed consent obtained from all study participants, and ethical approval granted at the Regional Ethical Review Board in Berlin, Erlangen, Vienna, and Stockholm.

Consent for publication
Not applicable Competing interests Distribution of ELISA signals of α-hnRNP-DLmir autoantibodies. reactivities were predominantly found in SLE and RA.

Figure 2
Anti-citrullinated hnRNP-DLmir autoantibodies are detectable even before the onset and in early status of disease. A-C, Anti-citrullinated hnRNP-DLmir (cit-DL), α-hnRNP-DLmir (DL) and ∆ OD between cit-DL and DL (ΔDL) were measured by ELISA. A, In Risk-patients of arthritis the OD-levels of citDL and ΔDL before onset are signi cantly speci c in the patient group where the arthritis has already been diagnosed compared to the group without diagnosis (n=71; non Arthritis n=34/Arthritis n=37; MannWhitneyU; citDL mediannon Arthritis=0. 19