Autoantibody profiles and clinical association in Thai patients with autoimmune retinopathy

Autoimmune retinopathy (AIR) is a rare immune-mediated inflammation of the retina. The autoantibodies against retinal proteins and glycolytic enzymes were reported to be involved in the pathogenesis. This retrospective cohort study assessed the antiretinal autoantibody profiles and their association with clinical outcomes of AIR patients in Thailand. We included 44 patients, 75% were females, with the overall median age of onset of 48 (17–74, IQR 40–55.5) years. Common clinical presentations were nyctalopia (65.9%), blurred vision (52.3%), constricted visual field (43.2%), and nonrecordable electroretinography (65.9%). Underlying malignancy and autoimmune diseases were found in 2 and 12 female patients, respectively. We found 41 autoantibodies, with anti-α-enolase (65.9%) showing the highest prevalence, followed by anti-CAII (43.2%), anti-aldolase (40.9%), and anti-GAPDH (36.4%). Anti-aldolase was associated with male gender (P = 0.012, OR 7.11, 95% CI 1.54–32.91). Anti-CAII showed significant association with age of onset (P = 0.025, 95% CI − 17.28 to − 1.24), while anti-α-enolase (P = 0.002, OR 4.37, 95% CI 1.83–10.37) and anti-GAPDH (P = 0.001, OR 1.87, 95% CI 1.32–2.64) were significantly associated with nonrecordable electroretinography. Association between the antibody profiles and clinical outcomes may be used to direct and adjust the treatment plans and provide insights in the pathogenesis of AIR.


Methods
Study design and ethics. This retrospective cohort study assessed the secondary data of clinical presentation, history, and antiretinal autoantibody profiles from medical records of suspected AIR patients in Siriraj Hospital, Thailand. All procedures were performed in accordance with the relevant guidelines and regulations. The study was approved by the institutional review board (IRB) of the Faculty of Medicine Siriraj Hospital Mahidol University (approval number SI531/2020, dated June 24th, 2020). All the data that we analyzed and reported, including the patient history, ophthalmological examination, and antibody detection in patient sera, had been collected in the course of routine diagnostic analysis for suspected AIR patients. Therefore, the IRB had approved the exempt of informed consent from individual patients. Data collection. Due to the rarity of AIR, we reviewed medical records of all AIR patients attending the Ophthalmology clinic from November 2013 until December 2020. The suspicion of AIR cases began by ruling out other possible causes of visual abnormality and signs of severe inflammation. Additional data comprised clinical presentations, including decreased best corrected visual acuity (BCVA), abnormal color vision, nyctalopia, or reduced peripheral visual field (VF), personal or family history of autoimmune diseases and/or malignancies, abnormal electroretinography (ERG), optical coherence tomography (OCT), and fundus photography. Finally, the diagnosis of AIR cases was supported by the positive result of at least one antiretinal autoantibody.
Antiretinal autoantibody detection. As part of routine diagnosis for suspected AIR cases, patient sera were collected and sent to the Ocular Immunology Laboratory, Casey Eye Institute, Oregon Health & Science University, Portland, OR, USA. The reactivity was tested using immunoblotting. The first 20 patients were tested for autoantibodies against retinal and other unidentified proteins prior to the availability of the specific diagnostic panels. Afterwards, 22 patients were subjected for the autoimmune retinopathy panel (CAII, HSP27, retinal arrestin, tubulin, GAPDH, aldolase, α-enolase, and PKM2), while two patients with history of cancer were tested for the cancer-associated retinopathy panel (recoverin, CAII, HSP60, TULP1, Rab6, GAPDH, aldolase, α-enolase, and PKM2).
Statistical analysis. The demographic data and antibody profiles were presented descriptively. The association between categorical or ordinal variables were assessed using chi-square and Fisher exact test. The mean/ median between variables with continuous data was analyzed using independent t-test or Mann-Whitney U test. The hypothesis was two-tailed with P < 0.05 considered as statistically significant. All missing data were excluded from analysis. The statistical analysis was performed using PASW Statistics version 18.0.0.

Results
Baseline characteristics of patients. There were 44 patients with suspected AIR (Table 1 and Supplementary Table 1), 33 females and 11 males, with the median age of onset and at diagnosis of 48 (17-74, IQR 40-55.5) and 58.5 (23-75, IQR 50.75-64) years, respectively. In the first visit, common clinical presentations were nyctalopia (65.9%), blurred vision (52.3%), and reduction in peripheral VF (43.2%). More than half of the patients presented with normal color vision (59.1%) and nonrecordable ERG (65.9%). Interestingly, we found 12 (all female) out of 44 patients with underlying autoimmune diseases (systemic lupus erythematosus (SLE) 50%, rheumatoid arthritis (RA) 33%, and others 17%) and two female patients with confirmed malignancy (CAR). Most patients (65.9%) reported neither related underlying conditions nor family history of autoimmune diseases. The median age of onset in patients with underlying autoimmune disease was significantly lower than those without (40 vs  www.nature.com/scientificreports/ had VF data only in one eye) had a constricted VF, with a mean reduction of 77.3%. Sixty-six (83.5%) of 79 evaluated eyes had a reduction of more than 50% from the normal value, 16.5% had a reduction of less than 50%. Central and paracentral scotoma were found in 3.8% and 27.8% of the evaluated eyes, respectively. Five patients (11.4%) had a history of receiving chloroquine or hydroxychloroquine for autoimmune conditions, for which data was available from four patients. The average duration of the treatment in these patients was 3.75 years. Three patients had the onset of AIR 2, 3, and 9 years after discontinuing chloroquine, while in one patient, the ophthalmological symptoms began during chloroquine treatment.
Follow up data. In this retrospective cohort study, we followed the patients ranging from 1 month to 12 years 3 months (median 1.8 years, IQR 1-3.9). The ophthalmological examination during the follow up was summarized in Table 2 and Supplementary Table 2 in the Supplement. The BCVA was decreased in 42.5% of evaluated eyes and remained stable in 27.6%. Interestingly, we found 29.9% of eyes with slight improvements in the BCVA. The patients with improved BCVA comprised of follow up time of less than or equal to one year (four patients) and more than one year (ten patients). We also observed the changes of the VF in 59 eyes with available follow up data, with periods between the examinations ranging from 0.7-6.8 (median 2.4, IQR 1.5-3.83) www.nature.com/scientificreports/ years. The peripheral VF was reduced in 55.9% of evaluated eyes, with an average reduction of 36.3% from the baseline, and stable in 15.3%. We noticed improvements of less than 50% from the baseline VF in 25.4% of eyes and more than 50% in 3.4%. The baseline OCT was evaluated in 43 out of 44 patients, however the follow up data was only available from 33 patients with follow up time ranging from 0.5 to 10 (median 2, IQR 1-5) years. Compared to the baseline, 57.6% of patients showed stable OCT findings (mostly in patients with follow up time less than four years), 33.3% had worsening conditions, and 9.1% displayed no pathology. In most patients, the disease progression advanced from the peripheral retina to the macula, except in two patients, where the central retina showed more severe pathology than the peripheral areas. The overall OCT observation showed that the loss of ellipsoid zones (EZ) occurred early. The foveal EZ were usually preserved until the late stage of the disease before they eventually vanished. We found significant association between the duration of disease (the time between onset until diagnosis) and conditions of EZ at baseline (P = 0.007). Despite the worsening of the retinal layers and reduction of peripheral VF, some patients maintained good BCVA due to the preservation of the EZ in the fovea. Retinal pigment epithelium (RPE) changes started with attenuation, followed by atrophic disappearance. The outer nuclear layer (ONL) became thinner to flat. In patients with longer follow up periods, the choroid also showed thinning due to atrophy of the inner layer. The fundus photography appearance in most of our patients displayed generalized RPE atrophy, scattered pigment clumps, and arteriolar attenuation, as shown in Fig. 1a,b. The example of OCT from one of the patients demonstrated RPE attenuation, shortening of EZ, and flattening of the ONL (Fig. 1c-f).
Thirty-two patients (72.7%) received immunosuppressive medications, predominantly prednisolone (68.2%) and azathioprine (31.8%). The duration from the onset of the eye symptoms to the initial drug treatment ranged from 3 months to 30 years (median 7 years). We found that treatment was significantly associated with follow up variables, such as changes in VF (P = 0.009), OCT (P = 0.009), and the changes of EZ (P = 0.002) using chi-square test. In the group receiving treatment, 17 out of 32 (53%) patients showed stable BCVA, and 22 out of 31 (71%) patients displayed stable conditions of the EZ.
Autoantibody profile. Out of 44 patients with positive antiretinal autoantibodies, we observed 41 different autoantibodies, with average number of antibodies found in each patient 4 ± 2 (mean ± SD, range 1-11). In the present study, the autoantibodies were classified into three groups based on their target proteins: antibody against glycolytic enzymes, retinal proteins, and unidentified retina proteins. Autoantibodies against CAII, GAPDH, aldolase, and α-enolase were tested in all 44 patients. Antibodies toward arrestin, tubulin, HSP27, and Table 2. Summary of patient follow up. BCVA best corrected visual acuity, OCT optical coherence tomography, VF visual field, N/A not applicable. a One patient had a prosthetic eye. b VF was compared to the baseline. c Cyclosporine, leflunamide, dexamethasone, intravenous immunoglobulins (IVIG), rituximab.  Fig. 2a,b, the antibody against α-enolase had the highest prevalence (65.9%) in the glycolytic enzyme-specific autoantibody panel, followed by antibodies against aldolase (40.9%), GAPDH (36.4%), and PKM2 (21.4%). In the panel of retinal proteins, anti-CAII showed the highest prevalence (43.2%), followed by antibodies against arrestin (27.3%) and tubulin (26.2%). The remaining antibodies in this panel, including antibodies against recoverin, HSP27, HSP60, CRMP, TULP1, and Rab6, contributed to a prevalence of less than 15% each. Except for anti-60 kDa (15%), all antibodies against the unidentified retinal proteins had a prevalence of no more than 10%. Ten out of 12 patients with underlying autoimmune disease and both CAR patients had at least one autoantibody against the glycolytic enzymes. In patients without underlying autoimmune disease and malignancy, 27 out of 30 patients had at least one anti-glycolytic enzyme autoantibody. We found no significant difference between the two groups (P = 0.67, odds ratio 0.67) using Fisher exact test. Moreover, two CAR patients showed autoantibodies against recoverin, HSP60, and Rab6.    www.nature.com/scientificreports/ Out of 41 autoantibodies, 13 autoantibodies against the retinal proteins and glycolytic enzymes were analyzed further for the association with demographic (gender, age of onset, age at diagnosis, duration from onset to diagnosis), baseline (color vision, ERG, condition of the EZ, and occurrence of nyctalopia) and follow up variables (changes in BCVA, VF, OCT, and EZ). Aldolase was significantly associated with male gender in a Fisher exact test (P = 0.012, odds ratio 7.11, 95% CI 1.54-32.91). CAII showed significant association with age of onset (P = 0.025, 95% CI −17.28 to − 1.24) and PKM2 with age at diagnosis (P = 0.033, 95% CI 0.77-17.34) using independent t-test. We found significant association between disease duration from onset to diagnosis and anti-recoverin (P = 0.024), anti-CAII (P < 001), and anti-GAPDH (P = 0.045) by Mann-Whitney U test. Autoantibodies against the glycolytic enzymes, GAPDH (P = 0.001, odds ratio 1.87, 95% CI 1.32-2.64) and α-enolase (P = 0.002, odds ratio 4.37, 95% CI 1.83-10.37), were significantly associated with ERG findings using Fisher exact test. HSP27 was associated with BCVA in both eyes (right eye P = 0.001, 95% CI 0.26-0.88, left eye P < 0.001, 95%CI 0.38-1.01), while α-enolase, and PKM2 were associated with BCVA of the right eye at the last visit (P = 0.029, 95% CI − 0.97 to − 0.05, and P = 0.013, 95% CI 0.11-0.92, respectively) by independent t-test.

Discussion
The data of autoantibodies in AIR patients from different populations is needed to validate the diagnostic criteria for this disease. In this study, we explored the autoantibody profile of 44 Thai patients with suspected AIR. With a female predominance and a median age of onset of 48 years, we found two CAR patients and 12 patients with underlying autoimmune diseases. Together with the clinical presentations and ophthalmological examination, we found 41 autoantibodies against retinal proteins, which were classified into glycolytic enzymes, retinal proteins, and other unidentified proteins.
The late age of onset is one of the distinctive characteristics differentiating AIR from other inherited retinal diseases 11,14 . The median age of onset of AIR patients was reported in the fifth or sixth decades of life 12 . Patients with family history of autoimmune disease tend to have earlier age at diagnosis 16,17 . In our study, the overall median age of onset was 48 years. The median age of onset of patients with underlying autoimmune disease was significantly lower than those without underlying disease (40, IQR 37-51.25 vs 50, IQR 45-60.5 years). Interestingly, we found three patients with very young age of onset (17-20 years), one of which had underlying autoimmune disease, without any history of inherited retinal diseases. Previous studies also reported occasional patients with a young age of onset (11-20 years) [16][17][18] . This should be a point of consideration for the clinicians to include AIR as a differential diagnosis of retinal diseases in young patients. For CAR patients, the age at diagnosis (62-79 years) are older than those with npAIR (36-55 years) 16,18 . Our CAR patients had the age at diagnosis of 58 and 63 years, which is comparable to the previous reports.
Underlying autoimmune diseases or family history of autoimmune disorders have the tendency to affect female 19 and are often associated with npAIR 3,16 . The AIR patients in this cohort were predominantly female (33 in 44, 75%). Twelve out of 44 patients presented with underlying autoimmune disease, all of which were females. The most common underlying autoimmune disease in this study was SLE and RA, which were reported to primarily affect females 20 . SLE-and RA-related AIR has been reported occasionally in case reports 17,21,22 . Besides, hypothyroidism was also reported as the most common npAIR-related underlying autoimmune disease 17 .
In our study, the autoantibody against α-enolase, an important and ubiquitously expressed glycolytic enzyme, showed the highest prevalence, followed by anti-CAII autoantibody. Previous studies in AIR patients in the US and China reported similar trends of autoantibody profiles 2,23 . A study described the sensitivity and specificity of α-enolase as 92% and 31%, respectively, with a positive and negative predictive value of 43% and 87%, respectively 2 . The high prevalence of anti-α-enolase autoantibody might be induced by the molecular mimicry toward microbial and host enolases 24,25 . Another plausible explanation is the response against tumors since the highly mitotic cancer cells produce energy by aerobic glycolysis to promote tumor growth and metastasis [26][27][28] , which triggers the immune response to produce antibodies against the glycolytic enzymes 29,30 . The high rate of aerobic glycolysis occurring in the retinal cells, particularly the photoreceptors, renders the retina to be more sensitive to anti-glycolytic enzyme autoantibodies 31 . Anti-α-enolase autoantibody can penetrate the retinal tissue, reach its targets in the GCL and INL, and induce apoptotic cell death 32 . The autoantibody against CAII was also reported to induce target cell apoptosis by deranging the cells' metabolism through the inhibition of this protein's catalytic activities 33,34 . Anti-CAII is thus often associated with more severe clinical outcomes in AIR patients 34 , although we did not find significant association between anti-CAII autoantibody and the disease outcomes in this study.
In this female predominant disease, we found association between the third most common autoantibody in our patients, anti-aldolase and male gender. Risk estimate showed that anti-aldolase was seven times more likely to be found in male, but the significance of this finding requires further investigation.
Two of our patients had concurrent breast cancer and matched with the CAR criteria. The eye symptoms began two and seven years, respectively, before the diagnosis of cancer. Several types of cancers, including breast cancer, are reported to be associated with retinopathy 35 . The latency, or period from cancer diagnosis to onset of retinopathy or vice versa, was variable. A cohort study by Adamus et al. found that the time from cancer diagnosis to the onset of retinopathy was dependent on the cancer type and kinetics, immune response to cancer, and the cancer-induced autoimmune response toward the retina 12 . Nevertheless, the early diagnosis of CAR and appropriate management might improve the visual prognosis in these patients.
Both CAR patients in this study presented with autoantibodies against recoverin, HSP60, and Rab6. Reported to be related to various types of cancers, anti-Rab6 autoantibody could be found in both npAIR and pAIR patients 2,9 . The autoantibody against Rab6 is not detected in the normal population, thus making it a potentially specific biomarker for AIR, although the prevalence is low 10 . On the contrary, anti-HSP60 represents a natural antibody developed toward common pathogen invasions 36  www.nature.com/scientificreports/ through the molecular mimicry 37,38 . Autoantibody against HSP60 has also been related to several autoimmune diseases 37,39 , various types of cancer 2 , and even the normal population, although the concentration among these groups significantly differs 23 . Autoantibody targeting recoverin, a photoreceptor-specific protein, has been postulated to be a good biomarker for AIR due to its absence in the normal population 2,6,9 . Moreover, it is expressed in several tumors, making it a sensitive and specific autoantibody for CAR 8 . In our study, anti-recoverin was found in three out of 22 tested patients, two of which were CAR patients and one with family history of malignancy.
Previous reports suggested a worse prognosis in patients with positive anti-α-enolase antibodies 40 . We found significant association between anti-GAPDH, anti-α-enolase and nonrecordable ERG, which represents the absence of electrical response in the retina indicating the severe dysfunction. Thirteen patients with positive anti-GAPDH who were tested for ERG showed nonrecordable waves. Of the 25 patients with positive anti-αenolase and had the ERG reports, 23 of them showed nonrecordable ERG. Thus, our data not only confirmed the significance of anti-α-enolase antibody, but also provided additional information regarding autoantibody against GAPDH as a possible predictor for disease severity.
Anti-HSP27 autoantibody has been reported to cause symmetric, slow progression in CAR patients 9 . In our study, anti-HSP27 showed significant association with the BCVA at the last visit. All three patients with positive anti-HSP27 had good BCVA in both eyes at the time of diagnosis and the BCVA was stable or slightly decreased at the last follow up. However, the duration of follow up in these patients was less than 2 years. Thus, a higher number of patients is required to confirm the clinical correlation.
The primary therapy for AIR is a single or combination use of immunomodulatory agents, including corticosteroids, anti-metabolites, and T cell inhibitors 41 . A positive response to AIR treatment, as characterized by stable vision, improvement of VF and b-wave amplitude in ERG, and stable or even improvements of the EZ 16,42 , usually takes as long as one year or more before results could be observed 8,41 . In our study, 32 out of 44 patients received immunomodulators, where we found significant association between the treatment and the clinical outcomes of VF, OCT and condition of the EZ. Stable or improved BCVA and stable condition of EZ in one or both eyes were observed in 53% and 71% of patients receiving treatments, respectively. This positive response is in accordance with a previous study reporting 70% response rate after treatment with prednisone, cyclosporine, and azathioprine in a pool of CAR and npAIR patients 16 . We noticed that the therapeutic response in patients with underlying autoimmune disease was lower than those without (56% vs 77%). This finding corresponds to a study by Ferreyra et al. reporting the lesser response in patients with history of autoimmune diseases, suggesting that these patients might require stronger immunomodulatory treatments 16 .
Altogether, our study provided the correlation of autoantibody profiles and clinical presentations of AIR patients in Thailand. Since this study was not intended to identify antiretinal autoantibodies as the biomarkers for AIR, the autoantibody profile in the healthy population was beyond the scope of this study. In the future, we plan to assess the prevalence of autoantibody in healthy individuals and patients with other retinal dystrophies compared to AIR patients in the same population. This information could support the significance of these autoantibodies leading to rapid and proper management of AIR.

Data availability
The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.