Sjögren Syndrome in Systemic Lupus Erythematosus: A Subset Characterized by a Systemic Inflammatory State

Objective. An often-neglected subset of patients with systemic lupus erythematosus (SLE) is those with secondary Sjögren syndrome (SLE-sSS). Further, primary SS overlaps and can be difficult to delineate from SLE. To shed light on the SLE-sSS subset, we investigated a large and well-characterized SLE cohort, comparing patients with SLE-sSS and SLE patients without SS (SLE-nonsSS) and controls. Methods. We included 504 consecutive patients with SLE, fulfilling the 1982 revised American College of Rheumatology criteria, and 319 controls from the general population, matched for age and sex to the first 319 patients. SLE-sSS was defined according to the American-European Consensus Criteria (AECC). A thorough clinical examination, including subjective and objective quantifications of sicca symptoms, was performed in all participants. Autoantibodies and 20 selected cytokines were measured by luminex and multiplex analysis, respectively. Results. SLE-sSS, as defined by AECC, occurred in 23% of the patients with SLE. In comparison to SLE-nonsSS, the SLE-sSS group was older and more frequently female. Leukopenia and peripheral neuropathy were more frequent and nephritis less frequent. Circulating levels of 6/20 investigated proinflammatory cytokines [tumor necrosis factor-α, interleukin (IL) 6, monocyte chemoattractant protein 4, macrophage inflammatory protein 1β, IL-12/IL-23p40, and interferon γ–induced protein 10], total IgG, anti-SSA/Ro52, anti-SSA/Ro60, anti-SSB/La antibodies, and rheumatoid factor (IgM and IgA) were higher in the SLE-sSS group (p < 0.05 for all comparisons). Conclusion. The frequency of SLE-sSS increased with age and affected roughly one-quarter of all patients with SLE. Despite less internal organ involvement, a systemic inflammatory state with high levels of proinflammatory cytokines is present in the SLE-sSS subgroup. This is a novel observation that may affect future understanding and treatment of the SLE-sSS subset.

Systemic lupus erythematosus (SLE) is a chronic autoimmune systemic disease characterized by enhanced autoantibody production and formation of immune complexes. SLE is also a very heterogeneous condition involving many organ systems, and disease activity varies from persistently mild to life-threatening 1 . A state of systemic inflammation, often associated with complement consumption, enhanced activity in the type I interferon system, as well as high levels of proinflammatory cytokines [e.g., tumor necrosis factor-α (TNF-α), interleukin 6 (IL-6), IL-8, and interferon γ-induced protein (IP)-10], is common in SLE 2,3,4,5 . While the general prognosis of SLE has improved, mortality rates remain more than 2-fold higher than in the general population, and cardiovascular diseases constitute a growing share of mortality causes 6,7,8 .
It has become increasingly clear that different subsets of the SLE population exist. Already in 1959, Heaton suggested that Sjögren syndrome (SS) is a chronic and relatively benign form of SLE 9 . More recently, several studies have identified autoantibody clusters/immune phenotypes, which vary in clinical symptoms, biomarkers, and prognosis 10,11,12,13,14 . An SLE phenotype characterized by antibodies to SSA and SSB antigens (also referred to as Ro/La) consistently appeared in these studies, but this supposedly milder SLE subset has so far achieved limited scientific attention.
The diagnosis of SS is a clinical entity, based on dryness of eyes and mouth due to destructive inflammation in the exocrine glands, especially tear and salivary glands. SS can exist in isolation, as primary SS (pSS), or with other rheumatic diseases, referred to as secondary SS (sSS). A major difference according to the 2002 Revised American-European Consensus Criteria (AECC) is the classification where the serologic item (SSA/SSB antibodies) is included for pSS, but not for sSS 15 . In SLE these autoantibodies are common, usually stable over time, and they appear early, even several years before disease onset 16,17 .
The clinical SLE-sSS phenotype has been described as a mild version of SLE with dominance of skin and joint manifestations and with less severe internal organ involvement, and in particular, less nephritis 9,18,19 . Differences and similarities between pSS and SLE with SS (SLE-sSS) have been studied 20 , but to what extent the inflammatory pattern differs between SLE-sSS and SLE patients without SS (SLE-nonsSS) is not known, and this information may be important regarding treatment perspectives.
In our present study, subjective and objective symptoms of sSS, defined according to AECC 15 , from a large and well-characterized cohort of consecutive SLE patients and matched controls are presented. To our knowledge, no previous study has investigated patients with SLE and matched population controls for both subjective and objective symptoms of SS and associated SSA/SSB autoantibodies. The primary outcomes were occurrence, and clinical and immunological characteristics of the SLE-sSS subgroup. As secondary outcomes, we performed stratified analyses based on anti-SSA/SSB profiles.  24 .

MATERIALS AND METHODS
All participants were evaluated for the individual items of SS according to the AECC 15 . Subjective ocular and oral symptoms were recorded as defined by the validated questionnaire. Objective measurements were Schirmer's test and whole unstimulated salivary flow (WUSF) over 15 min 25 . For controls, these objective measurements were only performed if subjective symptoms were present. Data on fulfillment of each item were collected. The amount of tears and saliva were recorded. If a salivary gland biopsy had been performed earlier, results were retrieved from the patient's records. Patients who had not had a salivary gland biopsy, when this was needed to determine whether they fulfilled the sSS criteria, were regarded as not having sSS. The attending rheumatologist assessed these patients as not having sSS, and hence that a salivary gland biopsy was not necessary.
Laboratory measurements. Fasting blood samples were drawn at inclusion. Laboratory tests were performed at the Swedish Board for Accreditation and Conformity Assessment (SWEDAC; www.swedac.se)-accredited Clinical Chemistry and Immunology Laboratories at the Karolinska University Hospital. Routine laboratory tests and analyses of complement factors and immunoglobulins were performed on fresh samples according to clinical routine.
IgA, IgG, and IgM rheumatoid factor (RF) were measured with a Phadia2500 instrument at Phadia Thermofisher Uppsala. Cutoffs for RF isotypes were determined as > 95% specificity compared with 100 blood donors for IgA and IgM RF and for 285 population controls for IgG RF.
Mesoscale Discovery (MSD) multiplex analysis of cytokines EDTA-plasma samples were analyzed on the MSD V-PLEX Human Cytokine 30-plex kit (K15054D; MSD) according to the manufacturer's instructions, as previously reported 4 . In short, the plasma samples were thawed at room temperature and diluted twice for proinflammatory and Personal non-commercial use only. The Journal of Rheumatology Copyright © 2020. All rights reserved. cytokine analyses, and 4 times for chemokine analysis in sample diluents. The diluted samples were incubated on the MSD plates for 2 h at room temperature with shaking. Plates were washed and incubated an additional 2 h with detection antibodies. After washing, 2× Read buffer T was added and the plates were analyzed in a Sector Imager 6000. Calibrator and plasma samples were analyzed in duplicates. Using the MSD Workbench software, the responses of the calibrator concentrations were plotted as log signal unit on the vertical (y) axis versus log concentration on the horizontal (x) axis. A weighted 4-parameter logistic fit equation was used for curve fitting and back calculation of plasma sample concentrations.
Statistical analysis. Patient and control characteristics are presented as mean ± SD, median (interquartile range), or percentages, depending on data type and distribution.
Groups were compared with the Student t test, Wilcoxon rank-sum test, or chi-square tests as appropriate. We used nonparametric tests when log transformation of continuous variables did not give a nearly normal distribution. Adjustment for age was performed by multiple logistic regression.
Calculations were performed using JMP software (SAS Institute). A 2-sided p value < 0.05 was considered statistically significant.
The local Ethics Committee at Karolinska Institutet approved the study (Dnr 03-556, Dnr 2017/1570-32). All study subjects gave written informed consent to participate in the study.

RESULTS
The 504 patients and 319 matched controls were well matched for age, but the female percentage was slightly lower among the patients (86.3% vs 92.2%; p = 0.01). The patients with SLE were 33.9 ± 15.5 years at diagnosis, and their disease duration was 12.1 ± 12.3 years. Basic characteristics are presented in Table 1. Patients with SLE-sSS versus SLE-nonsSS. SLE-sSS was present in 23.2% of the patients with SLE. Patients with SLE-sSS were older at inclusion (54.6 ± 13.6 vs 43.4 ± 14.7 yrs; p < 0.0001) and at SLE onset (40.4 ± 15.6 vs 31.9 ± 14.9 yrs; p < 0.0001). The percentage of patients with SLE-sSS increased with age, as demonstrated in Figure 1. The frequency of females was higher in the SLE-sSS group compared to SLE-nonsSS (95.7% vs 83.4%; p = 0.0007; Table 1).

DISCUSSION
According to the revised AECC 15 , SLE-sSS occurred in 23% of patients in this large and well-defined cohort of consecutive SLE patients. It is a novel observation that, despite less internal organ involvement, higher levels of proinflammatory cytokines were present in the SLE-sSS group than in the SLE-nonsSS. We also confirm that older age, female sex, and SSA/SSB antibodies, not included in the AECC definition of sSS, were positively associated with SLE-sSS. However, notably a large minority (39%) of SLE-sSS patients were negative for all 3 SSA/SSB antibodies. To our knowledge, this is the first study to investigate whether systemic inflammation, as measured by cytokine levels, differs between SLE-sSS and SLE-nonsSS. Although SLE-sSS is often considered a less severe form of SLE, the levels of several proinflammatory cytokines -TNF-α, IL-6, MCP-4, MIP-1β, IL-12/IL23p40, and IP-10 -were higher in SLE-sSS than in SLE-nonsSS patients. As previously reported, the investigated cytokines were also upregulated in SLE versus controls, and they were positively associated with SLE disease activity 4 . TNF-and IL-6 have also been reported as high in pSS 27 . Interestingly, an IP-10 antagonist ameliorated the progression of autoimmune sialoadenitis in MRL/lpr mice 28 . Further, hypergammaglobulinemia, a well-known feature of SS 29 , was in our study consequently more common among SLE-sSS patients, although the levels of IgG were usually below 20 mg/ml. We previously reported that low total IgM levels were associated with an SSA/SSB-positive profile in SLE 29 , but using the AECC, IgM levels were similar in SLE-sSS and SLE-nonsSS subgroups.
The occurrence of SLE-sSS in our study (23%) is higher than that reported by most previous studies (6-14%) 18,20,30,31,32 . An important reason for this discrepancy is likely attributable to our meticulous investigation procedures. In contrast to most studies, we measured tear and saliva production in all patients with SLE, regardless of whether subjective symptoms were present. It was a general finding that objectively reduced production of tear and saliva is more common than subjectively reported sicca symptoms. The high frequency of SLE-sSS is likely also affected by ethnicity. Our study population is mostly of European white origin and Baer, et al previously reported that in whites, as compared to other American ethnicities, frequencies of SLE-sSS were higher at 18% versus 14% 18 .
The frequencies of SLE-sSS increased with age. Patients with SLE-sSS were on average 9 years older than SLE-nonsSS, both at disease onset and at inclusion. Similar age differences were also reported previously 18,20,30,33,34 . The difference in age at SLE onset did however not remain after controlling for present age. Comparable findings were reported by Baer, et al, who investigated a large multiethnic SLE cohort 18  To further clarify whether sicca symptoms correlated with age, we stratified SLE-sSS, SLE-nonsSS, and controls by age. The results demonstrated that the prevalence of both subjective and objective sicca symptoms increase with age also in the SLE-nonsSS group and in controls. Thus, in some of the older patients with SLE-sSS, the AECC criteria may diagnose patients with sSS who have more age-related than immunological aberrations. Taken together, the present and previous studies demonstrate that sSS is an age-related complication among patients with SLE 18,19 .
The presence of SSA/SSB autoantibodies and their positive association to sicca symptoms is well recognized. Earlier studies reported that these antibodies occur together 12,16 . We observed that the prevalence of oral and ocular dryness was  more pronounced in patients who are positive for all 3 antibodies (SSA/Ro52, SSA/Ro60, and SSB/La). Our study thus supports the presence of an autoantibody cluster related to the SLE-sSS subset 10,11,29 . Although the presence of the SSA/SSB antibodies was associated with sicca symptoms, a considerable fraction of patients with SLE-sSS (39%) were negative for all 3 SSA/SSB antibodies. We also report higher frequencies of IgM and IgA RF in patients with SLE-sSS. Thus, the presence of sicca symptoms is not exclusively related to SSA/SSB antibodies, but also to increasing age, RF, and possibly to other unidentified factors.
Overall organ damage was more severe in the SLE-sSS group, but this difference did not remain after age adjustment. Regarding disease activity, we performed 2 validated indices: SLEDAI 23 , a qualitative (presence vs absence) recording of symptoms and laboratory aberrations; and SLAM 22 , which grades symptom severity and includes subjective symptoms such as fatigue, headache, and arthralgia. SLEDAI measurements did not differ, whereas with SLAM, higher disease activity scores were observed in the SLE-sSS than in the SLE-nonsSS subgroup. This observation could possibly be explained by the inflammatory state, which may cause subjective and general symptoms such as muscle and joint pain, headache and fatigue, which are included in the SLAM but not in the SLEDAI index.
Consistent with earlier studies 18,30,34 , the female predominance was more pronounced among SLE-sSS patients as compared to SLE-nonsSS (96% vs 84%). The very high percentage of female patients with SLE-sSS (96%) is similar to reports in pSS 30 . The average age of disease onset for SLE-sSS was older than for SLE-nonsSS (40.4 ± 15.6 vs 33.9 ± 14.9 yrs) but still considerably younger than the average age of onset for pSS, which is 55 years 35,36 .
In clinical practice, it is often difficult to delineate pSS from SLE-sSS. Organ manifestations commonly reported in pSS are fever, lymphadenopathy, parotid gland enlargement, Raynaud phenomenon, interstitial lung disease, peripheral neuropathy, and vasculitis 37,38,39 . All these clinical features, except parotid gland enlargement, were investigated in our present study, but only peripheral neuropathy differed and was more frequent in SLE-sSS than in SLE-nonsSS, demonstrating that the majority of these manifestations are shared between SLE and pSS. Of SLE manifestations, leukopenia was more common in the SLE-sSS group and there was also a positive trend for photosensitivity, both of which have been associated with the type I interferon signature. Nephritis was, in accordance with previous studies, less frequent in the SLE-sSS group, and anti-dsDNA positivity showed a similar trend 18,20,30,34 . We conclude that patients with SLE-sSS and patients with pSS have many similarities and it is not surprising that among the patients with SLE-sSS, according to AECC, 21% also fulfilled the criteria for pSS, if we disregard the fact that they are diagnosed with SLE (a prerequisite for being evaluated for pSS).
The strengths of this study are the objective measurements of sicca symptoms in all patients with SLE, according to the AECC criteria and recommended practice in Sweden, in a well-characterized consecutively collected cohort of patients with SLE. However, if the investigating rheumatologist did not consider sSS to be present, we did not refer  only Schirmer's test and WUSF are used 40 , and we have considered the risk for misclassification, but the rheumatologists assessed these patients clinically as SLE-nonsSS. To further illustrate this issue, we excluded the 132 patients in whom a biopsy could have made a difference, but this did not essentially change the characteristic differences between SLE-sSS and SLE-nonsSS. Excluding these patients would also have changed the study's cross-sectional and consecutive design. Finally, we did not measure saliva and tear production in controls without sicca symptom. Because positive objective measures were more common than subjective complaints, we may have underestimated the number of controls with impaired saliva/tear production. Our investigations of the SLE-sSS subset demonstrate that it affects roughly one-quarter of patients with SLE, and the frequency increases with age. Autoantibodies SSA/SSB occur in the majority, but a large minority (39%) were SSA/SSB-negative. SLE-sSS shares many features with pSS such as a striking female predominance, older age at onset, and neuropathy. It is a novel observation, with possible therapeutic implications, that an inflammatory state occurs in SLE-sSS with higher levels of proinflammatory cytokines than in SLE-nonsSS.