key: cord-0292782-99foqn5y
authors: Rojas, M.; Ramirez-Santana, C.; Acosta-Ampudia, Y.; M. Monsalve, D.; Rodriguez-Jimenez, M.; Zapata, E.; Naranjo-Pulido, A.; Suarez-Avellaneda, A.; J. Rios-Serna, L.; Prieto, C.; Zambrano-Romero, W.; Valero, M. A.; D. Mantilla, R.; Zhu, C.; Li, Q.-Z.; Toro-Gutierrez, C. E.; J. Tobon, G.; Anaya, J.-M.
title: Polyautoimmunity Clusters as a New Taxonomy of Autoimmune Diseases
date: 2021-08-23
journal: nan
DOI: 10.1101/2021.08.15.21262029
sha: 0c3d4db8c9e73015449756fe5fee37959e19146c
doc_id: 292782
cord_uid: 99foqn5y

Polyautoimmunity (PolyA) is an emerging concept that may help to develop a better classification of autoimmune diseases (ADs). Thus, we aimed to develop new taxonomy based on PolyA. Two-hundred and fifty-four consecutive patients were included with rheumatoid arthritis (RA, n:146), systemic lupus erythematosus (SLE, n:45), Sjogrens syndrome (SS, n:29), autoimmune thyroid disease (AITD, n:17) and systemic sclerosis (SSc, n:17). Clinical features, autoantigen array chip, lymphocytes immunophenotype and cytokine profile were assessed simultaneously. The coexistence of two or more ADs with classification criteria was termed Overt PolyA, whereas the presence of autoantibodies unrelated to the index AD, without criteria fulfillment, was named Latent PolyA. Combination of IgG autoantibodies yielded high accuracy for classification of ADs. In SLE, Histone H2A, Sm/RNP, ssDNA, and dsDNA IgG autoantibodies were the most predictive autoantibodies for this condition. Laminin, Ro/SSA (52 kDa), and U1-snRNP B/B for SS; Thyroglobulin for AITD; Ribo Phosphoprotein P1, and CENP-A for SSc. Interestingly, Thyroglobulin and U1-snRNP B/B' were mutual diagnostic biomarkers in SS and SSc. Latent PolyA showed in nearly 70% of patients, whereas overt PolyA was most common in AITD (82.4%) and SLE (40%). Cluster analysis based on autoantibodies yielded three clusters of which clusters 2 and 3 exhibited high frequency of latent and overt PolyA with distinctive clinical and immunological phenotypes. Combination of autoantibodies demonstrated high performance for classification of ADs. Patients with both latent and overt PolyA cluster together and exhibit differential clinical and immunological features. High prevalence of latent and overt PolyA advocates for routinary surveillance in clinical settings.

disclosed the highest rates of management with corticosteroids, disease-modifying antirheumatic drugs (DMARDs), and antimalarials, whereas patients with RA reported the highest rates of management with biologics.

Patients with RA presented low activity of disease, according to DAS-28. On the contrary, most patients with SLE showed moderate clinical reported activity (i.e., SLAQ score ≥ 3). Raynaud, telangiectasias, dysphagia, and sclerodactyly were most common in SSc. Arthritis and arthralgias were most frequently presented in RA, whereas malar rash was distinctive of SLE (Table 1) . Interestingly, some clinical manifestations were equally distributed across diseases, probably due to PolyA. Clinical inflammatory biomarkers (i.e., ESR, CRP and fibrinogen) and thyroid function did not differ among diseases. However, patients with SLE disclosed lower levels of total leukocytes, lymphocytes, and hemoglobin (Table 1) .

Initially, we evaluated the expression of autoantibodies compared with 38 healthy volunteers. After quality control filtering (i.e., SNR > 3), 111 IgG and 97 IgM autoantibodies were included in the final analysis. It was found that 25 IgG autoantibodies in SLE, 7 in SS, 2 in AITD, and 12 in SSc were over-expressed (i.e., Log2 fold change ≥ 1, and p-value FDR <0.05) (Supplementary Material). There were no over-expressed nor under-expressed IgG autoantibodies in patients with RA (Fig. 1A) and this finding was related to the lack of rheumatoid factor (RF) and citrullinated antigens included in the microarray. Patients with different index ADs shared several IgG autoantibodies (i.e., PolyA) (Fig. 1B) .

Patients with SLE showed over-expression of IgG autoantibodies against nuclear, thyroid, complement and collagen-associated antigens (Fig. 1C) . This was similar for SS, in which autoantibodies for ribonuclear proteins and thyroid antigens were over-expressed. Concerning AITD, only IgG autoantibodies against thyroid antigens were significantly over-expressed. Patients with SSc showed over-expression of autoantibodies against nuclear, cytoplasmatic and thyroid antigens (Fig. 1C) .

Few IgM autoantibodies were over-expressed in ADs (Supplementary Appendix) ( Fig. 2A) and sharing of IgM autoantibodies was less likely (Fig. 2B ). IgM autoantibodies against nuclear antigens in AITD and SLE were over-expressed. Interestingly, patients with SSc showed under-expression of autoantibodies for nuclear and myelinic antigens, but overexpression of liver associated autoantibodies (i.e., LC1) (Fig. 2C ). There were not underexpressed nor over-expressed IgM autoantibodies in patients with RA and SS.

Next, we focused on the evaluation of over-expressed IgG autoantibodies (i.e., Log2 fold change ≥ 1, and p-value FDR <0.05) as hallmarks for the identification of index AD. To avoid the bias of patients with overt PolyA, this analysis only those patients without such condition. Multivariate logistic regression yielded that IgG autoantibodies against nuclear antigens disclosed the best performance in SLE (Table 2 ). In AITD, IgG against Tg was the only associated autoantibody. Tg was associated with classification of SS and SSc. In addition, autoantibodies against ribonucleoproteins and centromere were also associated with classification of SS and SSc, respectively (Fig. 3A) .

Next, given the thresholds of the IgG autoantibodies obtained (Table 2) , we evaluated their positivity in all included diseases to estimate the frequency of latent PolyA. Since Tg and U1−snRNP B/B' IgG autoantibodies were shared in several conditions, only those estimated thresholds for AITD and SS were used, respectively.

Positivity for included autoantibodies in all patients is shown in Figure 3B . Although IgG autoantibodies obtained by multivariate analysis exhibited high frequency in specific index conditions, patients showed positivity for other autoantibodies (i.e., PolyA). The overall frequency of these autoantibodies was low in RA, whereas SLE, SS, and SSc showed higher positivity rates (Fig. 3B ).

Then, we looked for those autoantibodies that were not associated with overt ADs to estimate the occurrence of latent PolyA in each condition (Fig. 3C ). Latent SLE was the most common in patients with RA and SS. On the other hand, latency for SSc was the most common in patients with AITD and SLE, whereas latency for AITD was the most common in SSc. This analysis yielded that more than 70% of patients presented at least 1 type of latency for SLE, SS, AITD, or SS (Fig. 3D ).

Based on classification criteria, the frequency of overt PolyA was estimated. AITD was the most common cause of overt PolyA in SS and RA (Fig. 3E) . Conversely, RA was the most common cause of overt PolyA in AITD. In patients with SLE and SSc, overt PolyA was predominantly defined by SS. In contrast to latent PolyA, overt PolyA was less frequent, and AITD presented the highest rates (Fig. 3F ).

naïve CD4+ T cells were decreased (Fig. 4H ). On the other hand, Cluster 3, showed high levels of IL-8, activated CD4+ and CD8+ T cells (Fig.4H ).

In this proof-of-concept study, it is confirmed that PolyA allows a new taxonomy of ADs. Clusters of PolyA were well differentiated and characterized by a unique immune signature (i.e., cytokine response and cellular subphenotypes) (Fig. 4) . Moreover, latent PolyA was more frequent than overt PolyA (2, 3, 5) .

Several autoantibodies share different specificities across ADs. For example, anti-SSA/Ro and anti-SSB/La are considered the two most classic autoantibodies in SS (6) , and nearly 63% of patients show positivity to anti-SSA/Ro (21) . However, this autoantibody was also associated with the development of SLE (22). Anti-SSA/Ro in the presence of anti-SSB/La tends to identify patients with SS. It was found that 29 out of 35 patients with both anti-Ro/SSA and anti-La/SSB had SS, whereas of 53 with only anti-Ro/SSA, 23 had SS, 25 had SLE, and 13 had another disease (23) . This suggests that the combination of some autoantibodies in the diagnostic approach of ADs may improve the sensitivity and specificity of these tests (3) .

The microarray analysis allowed the identification of mixtures of autoantibodies that helped to develop high accuracy models for classification of ADs. Thus, confirming the usefulness of antibody combination in the diagnosis of disease. Tg and U1−snRNP B/B' autoantibodies were shared as diagnostic biomarkers for AITD, SS and SSc. This may suggest that some autoantibodies yielded similar specificities across diseases, or a phenomenon of latent PolyA, in which these patients may develop overt PolyA in the future. Usefulness of the models obtained deserves further attention and confirmation in larger studies.

The frequency of latent and overt PolyA matches with prior studies in which different causes of PolyA were described (1) (2) (3) . In this study, we found combinations of diverse ADs conforming the spectrum of PolyA. Latent PolyA was most common than overt PolyA. This may suggest that most patients with an index condition present autoantibody positivity for other ADs, and thus inferring that primary or secondary labels of ADs are inaccurate. In this line, patients with any ADs should be tested for other types of PolyA. This may have implications for follow-up and treatment (i.e., primary prevention). As mentioned, positivity for autoantibodies predate the appearance of overt ADs (6) , and in SSc, it has been suggested that PolyA may influence deleterious outcomes such as pulmonary fibrosis and mortality (24, 25) . As a corollary PolyA should be considered in all studies dealing with ADs, including epidemiological, genetic, and clinical trials.

The possible shortcomings of our study must be acknowledged. The main objective of this cross-sectional study was to estimate the frequency of latent and overt PolyA in outpatient clinics. In this line, this study reflects the frequency of ADs in our settings, being RA the most frequent, and supporting the unequal final sample size per group. In addition, patients with RA did not show over-expressed autoantibodies. The lack of RF and citrullinated antigens included in the autoantigen array chip could be associated with these results.

Further studies implementing these antigens are warranted. Frequency latencies were estimated only for those ADs included. Thresholds and specificities for other ADs were not estimated (e.g., gastrointestinal or endocrinological ADs). It is necessary to test this microarray in other ADs, to assess thresholds of positivity, and to improve the clinical efficiency of this technique. It is likely that tests for positivity for other ADs would have expanded PolyA, and latent PolyA would have reached higher rates.

A cross-sectional study was conducted from December 1st, 2018, to November 30th, 2019, in three tertiary specialized rheumatology centers; two were in Bogota, Colombia, including the Dermatology and Rheumatology Foundation (FUNINDERMA), and the Center for Autoimmune Diseases Research (CREA). The remaining center was the "Centro de Referencia en Osteoporosis, Reumatología & Dermatología", in Cali, Colombia.

Two-hundred and eighty-one consecutive patients attending the outpatient clinic were assessed. Only those patients with the following index conditions were considered: RA, SLE, SS, AITD, and SSc. The patients fulfilled either the 1987 American College of Rheumatology (ACR) classification criteria for RA (26) , the 1997 ACR criteria for SLE (27) , the 2013 ACR/European League Against Rheumatism (EULAR) classification criteria for SSc (28) , or the revised American-European Consensus Group for SS (29) . For AITD, patients with autoimmune hypothyroidism (AH) were classified as follows: 1) confirmed AH (i.e., thyroid dysfunction, thyroid stimulating hormone (TSH) >4.1 μ IU/mL or levothyroxine treatment, and the presence of anti-thyroperoxidase (TPO) or antithyroglobulin (Tg) antibodies), 2) euthyroid patients with positive anti-TPO or anti-Tg antibodies, 3) non-autoimmune hypothyroidism (thyroid dysfunction and absence of anti-TPO or anti-Tg antibodies) (30).

Then, foreign patients (n: 2), patients with prior history of neoplasia (n: 4), and unfulfillment of classification criteria (n: 21) were excluded. A final sample size of 254 patients was included in the analyses as follows: RA (n: 146), SLE (n: 45), SS (n: 29), AITD (n: 17) and SSc (n: 17). In addition, a group of 38 healthy volunteers (i.e., subjects without overt autoimmunity nor familial autoimmunity) were included as a control group.

This study was done in compliance with the Act 008430/1993 of the Ministry of Health of the Republic of Colombia, which classified it as minimal-risk research. All the patients were asked for their consent and were informed about the Colombian data protection law (1581 of 2012). The institutional review board of the Universidad del Rosario approved the study design.

The patients' demographic and cumulative clinical data were simultaneously obtained by standardized report form, physical examination, and chart review. Data included age, age at onset of disease, familial autoimmunity and familial autoimmune disease, and treatment on inclusion. All patients were evaluated for rheumatological or associated autoimmune clinical manifestations. These variables are described in Table 1 .

In addition, if any patient after inclusion presented positivity for autoantibodies by ELISA and Immunoblot not related to the index condition, clinicians evaluated the subject once again, and performed clinical tests to confirm classification criteria for overt PolyA. These included Schirmer and unstimulated saliva flow rate test for SS, and TSH for AITD. Patients with positivity for anti-phospholipid antibodies were tested within 12 weeks to confirm the classification criteria.

Severity of symptoms was assessed by either disease activity score 28 (DAS-28) for RA (31), systemic lupus activity questionnaire (SLAQ) for SLE (32), scleroderma skin patientreported outcome (SSPRO) for SSc (33, 34) , or EULAR SS patient reported index (ESSPRI) for SS (34) (35) (36) . In addition, most of the patients were systematically evaluated for erythrocyte sedimentation rate (ESR), C reactive protein (CRP), fibrinogen, and blood count. All data were collected in an electronic and secure database as described elsewhere (34) .

Samples were analyzed by an autoantigen array chip containing 128 antigens and controls at the Microarray and Immune Phenotyping Core Facility, UT Southwestern Medical Center. Briefly, the autoantigens and control proteins are printed in duplicates onto nitrocellulose film slides. Serum samples were pretreated with DNAse-I and diluted in phosphate buffered saline with Tween (PBST) for autoantibody profiling. The diluted serum samples were incubated with the autoantigen arrays, and autoantibodies binding with antigens on arrays were measured with cy3-conjugated anti-human IgG (Jackson ImmunoResearch Laboratories) and cy5-conjugated anti-human IgM (Jackson ImmunoResearch Laboratories), using a Genepix 4200A scanner (Molecular Device). The resulting images were analyzed with Genepix Pro 7.0 software (Molecular Devices). The median of the signal intensity for each spot was calculated and subtracted from the local background around the spot, and data obtained from duplicated spots were averaged. The background subtracted signal intensity of each antigen was normalized to the average intensity of the human IgG or IgM controls, which were spotted on the array as internal controls. Finally, the normalized fluorescence intensity (NFI) was generated as a quantitative measurement of the binding capacity of each antibody with the corresponding autoantigen. Signal-to-noise ratio (SNR) is another quantitative measurement of the true signal above background noise. SNR values equal to or greater than 3 were considered significantly higher than background, and therefore as true signals. The autoantibody which has the SNR value of less than 3 in more than 90% of the samples was considered negative and excluded from further analysis.

Serum of patients was collected in fasting state and spite of the treatment status. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted August 23, 2021. ;  Bioscience) as reported elsewhere (5) . For a detailed analysis of the cell phenotype, peripheral blood mononuclear cells were stained with fluorescent antibodies. A minimum of 100,000 lymphocytes per sample were acquired on a FACSCanto II™ flow cytometer (BD Biosciences™). Twenty-eight cell subsets (Supplementary Appendix,) were analyzed with FlowJo software version 9 (BD Biosciences™) as reported elsewhere (37) .

Univariate descriptive statistics were performed. Categorical variables were analyzed using frequencies, and quantitative continuous variables were expressed as the median and interquartile range (IQR). The Kruskal-Wallis, or Fisher's exact tests were used based on the results. Bonferroni correction was used for multiple testing in clinical manifestations.

Initially, we aimed to evaluate the over-expressed autoantibodies compared with healthy controls. First, data from autoantigen array was standardized by a robust linear model as previously described (38, 39) . The p-value was determined by unpaired t-test with a Benjamini and Bonferroni-Hochberg False Discovery Rate post-hoc correction (FDR). For all autoantibodies, the only selected were those that fulfilled: 1) p-value with a Benjamini-Hochberg FDR <0.05, and 2) Log2 fold change ≥ 1.

A logistic regression model was fitted to estimate the effect size of significant autoantibodies on ADs classification (i.e., Log2 fold change ≥ 1, and FDR <0.05). The dependent variable of the logistic model was the natural log of the odds of the index AD. The independent variables of the logistic model were selected through a backward selection procedure as previously described (40) .

From the selected autoantibodies, thresholds for positivity were obtained by maximizing the sum of sensitivity and specificity functions comparing it with healthy volunteers. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the curve (AUC) were estimated for each threshold (https://github.com/thie1e/cutpointr).

Next, we aimed to develop a new ADs classification based on those autoantibodies selected in the previous step. We used the mixed-cluster methodology proposed by Lebart et al. (41) First, a principal component analysis of the data was conducted. Next, the number of clusters by a hierarchical cluster analysis was determined, and finally, a consolidation step by k-means clustering was performed.

After identification of those autoantibody-based subgroups, immunological characteristics were evaluated for each group. Cytokine concentrations were analyzed after log transformation. Linear regression models were fitted to estimate the differences in cytokines and lymphocyte populations among clusters. All models were adjusted by age and sex. Post-hoc comparison of means was based on both adjusted Bonferroni p-values and Fisher's protected least significant differences procedure using t statistics based on Satterwhaite's approximation. The significance level of the study was set to 0.05. Statistical analyses were done using R software version 4.0.2. All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted August 23, 2021. ;

Table S1. Cellular markers used for immunophenotyping of lymphocytes by flow cytometry. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Under-expressed: 0 Over-expressed: 2

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Tregs  Effector CD8+ T cells  Th2  Naïve B cells  Th17  Th17/Th1 cells  Activated CD8+ T cells  Non-class memory B cells  Activated CD4+ T cells  Plasmablasts  Naïve CD4+ T cells  Th9  IL-8  Naïve CD8+ T cells  RANTES  CD19+ CD20-B cells  Central memory CD8+ T cells  Memory B cells  Transitional B (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted August 23, 2021. ; https://doi.org/10.1101/2021.08.15.21262029 doi: medRxiv preprint

Secondary Autoimmune Diseases No Longer Exist

Prevalence of latent and overt polyautoimmunity in autoimmune thyroid disease: A systematic review and meta-analysis

Cluster analysis of autoimmune rheumatic diseases based on autoantibodies. New insights for polyautoimmunity

Personalized medicine. Closing the gap between knowledge and clinical practice

Cytokine and autoantibody clusters interaction in systemic lupus erythematosus

Development of autoantibodies precedes clinical manifestations of autoimmune diseases: A comprehensive review

Diagnostic laboratory tests for systemic autoimmune rheumatic diseases: unmet needs towards harmonization

Specific autoantibodies precede the symptoms of rheumatoid arthritis: A study of serial measurements in blood donors

Antibodies of IgG, IgA and IgM isotypes against cyclic citrullinated peptide precede the development of rheumatoid arthritis

Antibodies against cyclic citrullinated peptide and IgA rheumatoid factor predict the development of rheumatoid arthritis

Development of Autoantibodies before the Clinical Onset of Systemic Lupus Erythematosus

Significance of Prediagnostic Thyroid Antibodies in Women with Autoimmune Thyroid Disease

Natural history of the transition from euthyroidism to overt autoimmune hypo-or hyperthyroidism: a prospective study

Latent autoimmune thyroid disease

Development of connective tissue disease in patients presenting with Raynaud's phenomenon: a six year follow up with emphasis on the predictive value of antinuclear antibodies as detected by immunoblotting

Prognostic significance of anticentromere antibodies and anti-topoisomerase I antibodies in Raynaud's disease. A prospective study

Autoantibodies Present Before Symptom Onset in Primary Sjögren Syndrome

Latent autoimmunity across diseasespecific boundaries in at-risk first-degree relatives of SLE and RA patients

The autoimmune tautology. A summary of evidence

The kaleidoscope of autoimmunity: multiple autoimmune syndromes and familial autoimmunity

IgG antibodies from patients with primary Sjögren's syndrome and systemic lupus erythematosus recognize different epitopes in 60-kD SSA/Ro protein

Autoantibodies predate the onset of systemic lupus erythematosus in northern Sweden

Anti-La (SS-B): a diagnostic criterion for Sjögren's syndrome?

Scleroderma-polymyositis overlap syndrome versus idiopathic polymyositis and systemic sclerosis: a descriptive study on clinical features and myopathology

Patients with Systemic Sclerosis/polymyositis Overlap Have a Worse Survival Rate Than Patients Without It

American College of Rheumatology/European League against Rheumatism criteria for classification of rheumatoid arthritis in the Nurses

Classification of Systemic Lupus Erythematosus: Systemic Lupus International Collaborating Clinics Versus American College of Rheumatology Criteria. A Comparative Study of 2,055 Patients From a Real-Life

Classification Criteria for Systemic Sclerosis: An American College of Rheumatology/European League Against Rheumatism Collaborative Initiative

Classification criteria of Sjögren's syndrome

Resilience in women with autoimmune rheumatic diseases

Validation of EULAR primary Sjögren's syndrome disease activity (ESSDAI) and patient indexes (ESSPRI)

Validación y adaptación al castellano del Índice Reportado por Pacientes con Síndrome de Sjögren del EULAR (ESSPRI-EULAR Sjögren's Syndrome Patient Reported Index), Reumatol

COVID-19 convalescent plasma composition and immunological effects in severe patients

Autoantigen Microarray for Highthroughput Autoantibody Profiling in Systemic Lupus Erythematosus

Robust-Linear-Model Normalization To Reduce Technical Variability in Functional Protein Microarrays

Protein array autoantibody profiles to determine diagnostic markers for neuropsychiatric systemic lupus erythematosus

The authors would like to thank Yhojan Rodriguez, María Higuera,