Association Between a History of Nontyphoidal Salmonella and the Risk of Systemic Lupus Erythematosus: A Population-Based, Case-Control Study

Objective We investigated the correlation between nontyphoidal Salmonella (NTS) infection and systemic lupus erythematosus (SLE) risk. Methods This case-control study comprised 6,517 patients with newly diagnosed SLE between 2006 and 2013. Patients without SLE were randomly selected as the control group and were matched at a case-control ratio of 1:20 by age, sex, and index year. All study individuals were traced from the index date back to their NTS exposure, other relevant covariates, or to the beginning of year 2000. Conditional logistic regression analysis was used to analyze the risk of SLE with adjusted odds ratios (aORs) and 95% confidence intervals (CIs) between the NTS and control groups. Results The mean age was 37.8 years in the case and control groups. Females accounted for 85.5%. The aOR of having NTS infection were significantly increased in SLE relative to controls (aOR, 9.20; 95% CI, 4.51-18.78) in 1:20 sex-age matching analysis and (aOR, 7.47; 95% CI=2.08-26.82) in propensity score matching analysis. Subgroup analysis indicated that the SLE risk was high among those who dwelled in rural areas; had rheumatoid arthritis, multiple sclerosis, or Sjogren’s syndrome; and developed intensive and severe NTS infection during admission. Conclusions Exposure to NTS infection is associated with the development of subsequent SLE in Taiwanese individuals. Severe NTS infection and other autoimmune diseases such as rheumatoid arthritis, multiple sclerosis, or Sjogren’s syndrome also contributed to the risk of developing SLE.


INTRODUCTION
Systemic lupus erythematosus (SLE), whose global prevalence and incidence respectively range from 9-241 and 0.3-23.2 per 1,00,000 person-years (1), is among the top 20 leading causes of death in women aged 5-64 years (2). The multifactorial etiology of SLE, including genetic factors, immune dysfunction-related factors, environmental and hormonal factors, can be deemed as the result of exposome-epigenome-genome interactions (3). Genetic susceptibility alone, however, is insufficient to trigger the onset of SLE, with the evidence of only 24% concordance rates among monozygotic twins with SLE manifestation (4). This indicates that besides genetic factors, the environment also plays a crucial role in the pathophysiology of SLE (5). In addition, heritability, shared (familial) environmental-, and nonshared environmental factors were reported to account for SLE susceptibility in 43.9%, 25.8%, and 30.3% Taiwanese patients, respectively (6).
The major environmental factors influencing SLE development include exposure to ultraviolet radiation, particulate air pollution, trace elements, alcohol use, and infections (7). A correlation between persistent infections and the development of autoimmune diseases has been reported. Infections by viruses, especially Epstein-Barr virus, parvovirus B19, cytomegalovirus, varicella-zoster virus (8), and retrovirus, might play a pivotal pathogenetic role in SLE and are highly associated with SLE (9). Likewise, bacterial infections are also linked to SLE. Gut commensals, for instance, were reported to contribute to the immune pathogenesis of lupus nephritis (10). Additionally, a common anti-DNA idiotype (16/6Id) carried by antibodies against Mycobacterium tuberculosis accounts for the antinuclear autoantibody positivity in tuberculosis patients (11), which may lead to autoimmune phenomena (12).
The association between salmonellae infections and the development of autoimmune diseases (13)(14)(15)(16) has been reported recently. Salmonella enterica serovar Typhimurium and Enteritidis seem to act more aggressively than other bacterial infections in SLE patients. Other than causing localized gastroenteritis, salmonella infection in SLE patients can cause bacteremia and complications with high mortality rates (17,18).
Nontyphoidal Salmonellae (NTS) is a major cause of diarrhea and foodborne gastroenteritis worldwide, and the risk of invasive gastroenteritis is six-fold higher than that of other bacterial infections (19). The global burden of NTS gastroenteritis was estimated to be 94 million cases and 1,55,000 deaths yearly worldwide, with an especially high incidence of 4 cases per 100 persons and over 88,000 associated deaths in 2006 in east Asia (20).
Salmonella infection due to defective Toll-like receptor (TLR)/myeloid differentiation primary response 88 signaling results in disruption of the cellular plasma membrane in B cells, highlighting the importance of correlations between genotype and environmental factors (21).
We hypothesized that a past NTS infection increases the risk of subsequent SLE. As the epidemiological relationship between NTS and SLE has not been clarified, we conducted this original nationwide case-control study to address this important issue.

Data Source
This study was conducted using data from Taiwan's National Health Insurance Research Database (NHIRD), which consists of original claim data and registration files for almost 99% of Taiwan's population. Registry for Catastrophic Illness Patients (RFCIP) belongs to a subset of NHIRD that lists diseases requiring long-term care. Patients in RFCIP are exempted from medical expenses as to protect vulnerable beneficiaries. Therefore, the reliability and accuracy of SLE diagnosis received by the enrolled patients were assured. We also employed another subset of NHIRD, the Longitudinal Health Insurance Research Database (LHIRD), in this study. The LHIRD comprises all the original claims data of 1,000,000 people randomly sampled from the year 1997 to 2013 registry of the NHIRD. The positive predictive value of NHIRD claims data can be up to 84.6% with a strict definition of study group selection. In Taiwan, hospital-licensed medical records technicians validated the coding before claiming the reimbursements, and the National Health Insurance Administration authority verified the audit. The Institutional Review Board of Taichung Veterans General Hospital (TCVGH CE14149B-1) approved this study.

Study Population
As shown in Figure 1, we identified patients who had been diagnosed with SLE by the International Classification of Diseases, Ninth Revision (ICD-9) codes (710.0) in the RFCIP during the period 1997-2013 (n=16,487). The index date was defined as the first date of SLE diagnosis during 2006-2013. Individuals diagnosed with SLE before 2006 were excluded. The control group was randomly selected from LHIRD during 2006-2013, which contained outpatients and inpatients data and was matched at a ratio of 1:20 by age, sex, and index year. Totally, 6,517 patients were included in the SLE group, and 130,340 non-SLE individuals served as the control group. We further performed propensity score matching (PSM) at a 1:2 ratio to minimize the potential confounding effects of sex, age, and selected comorbidities on the incidence of SLE.

Statistical Analysis
We measured the balance of baseline characteristics in the agematched, sex-matched, and propensity score-matched populations by the absolute standardized difference (ASD), whose value if <0.1 was considered a small difference. Student's t-test was used to compare variables between cases and controls for continuous variables, and chi-square test was used for comparing categorical variables. We used multivariable conditional logistic regression models to estimate the association between NTS and SLE. To further control for potential confounding factors, all models were adjusted for comorbidities and medications that might be associated with SLE. We used the following four models to estimate the effect of NTS on the risk of SLE: (1) model 1, NTS exposure alone, (2) model 2, NTS exposure + demographic variables, (3) model 3, NTS exposure + demographic variables + medical utilization and comorbidities at baseline in age-matched and sex-matched populations, (4) model 4, a conditional logistic model with NTS exposure alone in propensity score-matched population. Furthermore, we performed subgroup analyses by sex, age, urbanization, insurance amount, cumulative cost of NTS-related visits, and number of NTS visits. To improve the reliability of NTS ICD coding, we enrolled only patients using antibiotics at admission for further ascertainment. Conditional logistic regression of adjusted odds ratios (aORs) with 95% confidence intervals (CIs) was conducted. All data analyses were performed using SAS ® (version 9.4; SAS Institute, Inc., Cary, NC, USA). Statistical significance level was set at P-value <0.05 in the two-tailed test. Table 1 demonstrates the baseline characteristics of 136,857 and 19,401 patients in 1:20 age-sex matching and in 1:2 PSM analyses, respectively. The mean age of patients in both SLE and non-SLE groups was 37.8 years. There was an obvious female   Analysis of the interval between the first NTS visit and the index date revealed that an interval of 3.53-6.08 years was associated with a higher SLE risk, with an aOR of 1.68 (95% CI=1.11-2.56) in model 1 and 1.69 (95% CI=1.11-2.57) in model 2. In addition, NTS patients who were admitted with antibiotics use had a higher risk of SLE, with an aOR of 5.71 (95% CI=3. .80) in model 1; 5.71 (95% CI=3.32-9.80) in model 2; 5.00 (95% CI=2.88-8.70) in model 3; 3.78 (95% CI=1.68-8.48) in model 4.

DISCUSSION
In this nationwide, population-based case-control study, we found that the risk of new-onset SLE was 9.20-fold and 7.47fold higher in patients with a history of NTS than in matched controls by 1:20 sex-age matching and PSM analyses,  Table 2, our data implied that SLE was prone to rise among NTS-infected patients who had Sjogren's syndrome, Rheumatoid arthritis, Multiple sclerosis, and who lived in rural area rather than suburban and urban area. In Table 3, among ambulatory care or admission patients, the more intensive NTS visit in one NTS infection episode, the higher of NTS cumulative cost gave rise to higher odds ratio of future SLE, which largely occurred among 3 to 6 years after NTS infection. Among admission only patients, the more critical NTS infection is (who needed hospitalization and with both IV and oral antibiotics use), the higher odds of future SLE will be. Despite study illustrated that Salmonellosis of long duration was characterized by hyperactivity of autoimmune reactions (28), there was no advanced description on the duration from severe Salmonellosis to the onset of autoimmune disease. Environmental exposures other than infections might also depend on the level of urbanization. First, the smoking behavior of Taiwanese individuals was estimated to be 71.63% and was largely distributed (47.94%) in suburban area (29), which correlated with our observation. Second, dwelling in areas in which the content of particulate matter (2.5 μm) was over 35 μg/ m³ was associated with 15% higher long-term mortality risk according to the air quality guideline (30); such areas included many suburban and rural areas in Taiwan in 2013 (31) and are linked to many autoimmune diseases (7). Third, heavy metals were largely spatially distributed in the soil of rural townships during 1982-1986 in Taiwan (32). Major pollution events (33) in the past, such as aluminum slag, electric arc furnace dust pollution, and rice cadmium toxicity, largely in rural areas possibly trigger the onset of SLE or aggravate the condition. Last, low-income households that can afford only low insurance amounts are mainly located in rural areas (34), and they accounted for the higher prevalence of SLE in this study.
Other autoimmune diseases, such as Sjogren's syndrome, Rheumatoid arthritis, Multiple sclerosis, were prone to develop high risk of SLE in NTS infected patients in our study. Changes in symbiosis and dysbiosis of microbiome could flare autoimmune diseases (35). In patients with Sjogren's syndrome, the changes of oral microbiome in glandular and mucosal tissues brought about activated plasma cell and CD4+ cells that ruined the secretory acini of salivary glands. Similarly, increased levels of IFN-g and IL-17 caused by infiltration of CD4+ cells and dendritic cells in the conjunctiva causing ocular disease (36). On the other hand, in patients with Rheumatoid arthritis, dental, saliva, or gut microbiome was distinguished from healthy controls (37). Its HLA-DRB1 molecules, trans-membrane glycoproteins of macrophages and dendritic cells that present antigens to T cells, appear to forge the microbiome signature and build an inflammatory environment in animal study (38). Furthermore, gut microbiome shown to be altered in patients with relapsing Multiple sclerosis, causing changes of the metabolic pathways or the Gut-associated lymphoid tissues (GALTs) and aggravation of inflammatory demyelination in central nervous system (39).
Among environmental factors in SLE pathogenesis, microbiome has gained much attention in recent years. The mechanism by which NTS infection increases the risk of SLE remains unclear and disputed. Yet, we propose that long-term NTS stimulates abundant proinflammatory cytokines and high oxidative stress through innate and adaptive immune response, leading to the onset of SLE.

Interaction Between Innate and Adaptive Immune Systems
After reaching the lower intestine, salmonella bacteria translocate across Peyer's patches where they are engulfed by phagocytic cells. Salmonella are recognized by patternrecognition receptors, triggering cascades toward the induction of interleukin (IL)-1b, IL-18, and IL-23; cell pyroptosis and inefficient clearance of phagocytized bacteria (40). Infections by bacteria such as Escherichia coli and Salmonella are wellknown triggers of pyroptosis (41), which may play a pathogenic role in releasing host nuclear autoantigens in SLE (18).

Type I Interferon (IFN-I)
The autoantigens generated from uncleared apoptotic cells and neutrophil extracellular traps may induce the proliferation and differentiation of autoreactive B cells to produce autoantibodies, which then participate in the formation of immune complexes and tissue injuries. Both autoantigens and immune complexes promote the production of IFN-I by activating plasmacytoid dendritic cell signaling via amyloid/DNA complexes (42,43). IFN-I hampers the clearance of apoptotic cells by macrophages and induces B cells to produce autoantibodies (44). In addition, the immune complexes containing antinuclear autoantibodies could provoke plasmacytoid dendritic cells to produce more IFN-I, leading to SLE (45,46). Therefore, IFN-I has been recognized as the central pathogenic cytokine in SLE onset (47).

IL-17, IL-18, and IFN-g
During NTS infection, there is a massive release of proinflammatory cytokines including IL-23 and IL-18 (40). Antigen-presenting cells produce IL-23, which induces and maintains the activation of Th17 cells (48), and induces various innate and adaptive immune response cells, including innate lymphoid cells (49) and Th17 cells (50) to release the proinflammatory cytokines IL-17 and IL-22. IL-17, in conjunction with B cell activating factor, promotes the activation and proliferation of B cells along with their antibody production and class switching (51,52). Lupus patients have been reported to have increased serum levels of IL-17 (53) and IL-17-producing cells in the peripheral blood (54,55). IL-18 is vital in stimulating Th1 cells to release IFN-g (56), which plays a critical role in macrophage activation for the control of persistent salmonella infection (57)(58)(59). At the same time, the elevation of IL-17 and IL-18 in SLE patients may trigger the inflammatory process, and an increase in the IL-18/IL-4 ratio and mRNA levels of IFN-g/IL-4 (60) tilts the cytokine profile toward inflammatory cascades (53).
In real world practice, we prescribed intravenous antibiotics to patients with severe NTS infection and transited to oral antibiotics when discharged in Taiwan (61,62). The optimal antibiotic treatment duration for severe NTS gastrointestinal infection was 3-7 days and for NTS bacteremia ranges from 2 weeks (in healthy individuals) to 6 weeks (in immunocompromised patients) (63). Therefore, the data in "IV + Oral antibiotics use ≥ 3 days during visit" group (main model) was more persuasive than that in other groups. Despite low sample size gave rise to low statistical power, the high aOR (aOR=7.47, 95% CI= 2.08-26.82) in "IV + Oral antibiotics use ≥ 3 days during visit" subgroup using PSM strongly implied a correlation between NTS infection and future SLE development.
The advantages of using NHIRD in research include the availability of a large sample size, population-based data, and long-term comprehensive follow-up data. Our study is representative of the general population, and the possible measurable confounding factors of SLE were adjusted through matching with age, sex, demographic variables, and comorbidities through PSM. In addition, we strengthen the validity of the coding of NTS-associated diseases by filtering the NTS patients into those for whom only admission records data were available and those who were administered antibiotics to treat NTS infection. Therefore, the diagnosis of symptomatic NTS infection in our study population can be considered to be reliable.
Several limitations should be considered when interpreting our findings. First, epidemiologic evaluation of evident NTS infection was challenging as many infections were not clinically recognized. The use of retrospective ICD-9-based methods to select study groups of NTS infection might have led to selection bias. Second, NHIRD lacked information on smoking status, a well-known confounder of SLE and the most common risk factor for COPD development (64). Therefore, we managed to apply COPD prevalence as a proxy for smoking status (29). Third, non-differential misclassification bias on the diagnosis of NTS, a dichotomous variable, may exist. However, such bias always drives the result toward the null, and attenuates real effect estimates (65). Fourth, this was a single-country evaluation. Thus, our findings may not be applicable to non-Asian ethnic groups. Finally, the relation between symptomatic NTS and SLE was only demonstrated based on epidemiological evidence.

CONCLUSION
Exposure to NTS infection may be associated with an increased risk SLE and the risk increased with the individual severity of NTS infection. More epidemiologic and in vivo investigations are warranted to examine whether causality present between NTS infection and its trigger on SLE.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding authors.

ETHICS STATEMENT
The Institutional Review Board of Taichung Veterans General Hospital (TCVGH CE14149B-1) approved this study. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.