Association of Lipoprotein and Thyroid Hormones With Cognitive Dysfunction in Patients With Systemic Lupus Erythematosus

Background: Neuropsychiatric systemic lupus erythematosus (NPSLE) occurs up to 75% of adult SLE patients, and is one of the major causes of death in SLE patients. Cognitive dysfunction is a typical clinical feature of NPSLE, which seriously affects the quality of life of patients. Dyslipidemia and thyroid disease, which were prevalent in SLE patients, both related to the neuropsychiatric disturbances, including signicant psychiatric and cognitive disturbance. This study aimed to investigate whether cognitive dysfunction in patients with systemic lupus erythematosus (SLE) was related to the expression of serum thyroid hormone and lipoproteins. Methods: A total of 121 patients with SLE and 65 healthy controls (HCs) at Nanjing Drum Tower Hospital completed a cognitive function test, then 81 SLE patients were divided into high cognition (n=33) group and low cognition group (n=48). The differences in clinical and laboratory tests and the correlations between IgG, IgM, albumin, T3, and T4 levels and cognitive function were analyzed. The enzyme-linked immunosorbent assay was used to determine the serum levels of 4 lipoproteins (APOE, APOA1, IGF-1, and IGFBP7) in 81 patients. Results: Patients with SLE were less educated with abnormal cognitive function compared with HCs. The levels of albumin, T3 (P < 0.05), and T4 decreased in low-cognition patients, while D-dimer, anti-dsDNA antibody, and IgM levels increased. The serum IgG and IgM levels showed a signicant negative correlation with partial cognitive function. The serum T3 and T4 levels positively correlated with cognition. The expression of APOE, APOA1, IGF-1, and IGFBP7 showed no difference between the high-and low-cognition groups. However, the serum APOE level negatively correlated with Line Orientation, APOA1 positively correlated with Coding, and IGFBP7 negatively correlated with Graphic copy (P < 0.05), and IGF-1 had no correlation with any cognitive functions. Conclusions: Thyroid hormones (T3 and T4) and lipoproteins (APOE, APOA1, and IGFBP7) were associated with cognitive dysfunction in SLE. Whether T3 and T4 can be used in clinical practice as the biomarkers of cognitive dysfunction in SLE needs further exploration.


Introduction
In this study, the correlation between cognitive function and the levels of serum thyroid hormones (T3 and T4) and lipoproteins (APOE, APOA1, and IGFBP7) in SLE were evaluated. This study aimed to discover the changes in the levels of serum biomarkers and proteins possibly related to the cognitive function in SLE, thus prompting diagnosis or predicting prognosis of SLE.

Patients and study design
A total of 121 patients with SLE who visited the Department of Rheumatology, Nanjing Drum Tower Hospital, Nanjing, China, from May 2019 to May 2020 were prospectively enrolled. All patients were diagnosed according to the SLE criteria of the American College of Rheumatology [32]. The disease activity of these patients was measured using the Systemic Lupus Erythematosus Disease Activity Index (SLEDAI) [33][34]. Patients who had other autoimmune diseases; had a history of familial hyperlipidemia and/or thyroid disease, diabetes mellitus, and/or other rheumatic diseases; and took lipid-lowering agents or thyroid medications were excluded. Active patients were classi ed into two groups: low-cognition group (n = 33; RBANS score:  and high-cognition group (n = 48; RBANS score: , based on the RBANS score. Patients who developed neuropsychiatric syndromes not attributable to SLE (electrolyte imbalances, infections, or medications) were excluded. Meanwhile, 65 age-and gender-matched healthy controls (HCs) were recruited from the Physical Examination Center of Nanjing Drum Tower Hospital. This study was approved by the ethics committee at The A liated Drum Tower Hospital of Nanjing University Medical School (ID: SC201700201) and undertaken according to the guidelines of the Declaration of Helsinki. At entry, patients completed a standardized medical history, laboratory tests, and analyses. All the detections were carried out at the clinical laboratory of Nanjing Drum Tower Hospital.
The demographic features of the patients and HC samples are shown in Table 1.

Cognitive dysfunction study
The cognitive dysfunction in SLE is characterized by de cits in attention, learning and recall, verbal and nonverbal uency, language, visuospatial skills, executive functions, and motor dexterity [35][36][37]. A total of 121 patients with NPSLE and 65 normal volunteers were invited to participate in this study. All participants provided additional medical information using an ad-hoc questionnaire and had a physical examination.
The RBANS consisted of ve indexes: immediate memory, visuospatial/constructional, attention, language, and delayed memory [38]. Stimuli were contained in a wire-bound, easel-type booklet, making the test easily portable and allowing for bedside administration. The total administration time was [20][21][22][23][24][25][26][27][28][29][30] min. The battery indexes were assessed using the following tests [38]: a. List Learning: The examiner read aloud a list of 10 words, which were semantically unrelated, early acquired, relatively high-imagery, and as phonetically unique as possible. The participants were asked to recall (without regard for order) as many of these items as possible. After four trials, the number of words correctly recalled on the fourth trial was used as a measure of verbal learning and memory; the total score was 40 points.
b. Immediate Story Memory: The examiner read a short story including 12 keywords aloud twice, and then the participants were asked to recall this story as possible. The number of keywords correctly recalled was used as a measure of immediate story memory to avoid complicated scoring rules; the total score was 24 points.
c. Figure Copy: The participants were asked to remember a geometric gure comprising 10 parts and then copy this gure in 4 min. Each part received a 2-point score (accuracy and placement); the total score was 20 points.
d. Line Orientation: Two lines were presented at an angle, and the participants needed to select the corresponding lines on a simultaneously presented array. The subject's task on each item was to identify the matching lines. One point was given for each correctly matched line; the total score was 20 points.
e. Picture Naming: The participants were shown 10 pictures and asked to name the objects of the pictures. Semantic cues were given if the object was obviously misperceived (e.g., ''umbrella'' for mushroom). Naming correctly (including the correct answer according to the prompts) was used as a measure of picture naming; the total score was 10 points.
f. Semantic Fluency: The participants should generate the words as much as possible for a given semantic category (e.g., fruits and vegetables) within 60 s. The semantic categories aimed to minimize retrieval demands and thereby more speci cally tap semantic stores rather than retrieval strategies. One point was given for each correct word; the total score was 20 points.
g. Digit Span: The Digit Span was analogous to digits forward on the Wechsler Adult Intelligence Scale (WAIS). There were two string of digits in each item, increasing from 2 to 9 digits. The second given length string was read when the rst string was failed. The failure of attempts of two same items was given 0 points; the total score was 16 points.
h. Coding: First, the participants were provided with a row of graphics, each of which corresponded to a number. Later, they were provided with some gures without numbers. They were asked to ll in the number under each gure provided for the second time as fast as possible, according to the corresponding relationship between the gure and the number provided for the rst time. Numbers rather than symbols were chosen for the response to avoid the possible detrimental effect of constructional apraxia on performance. The score was the total number of items completed in 90 s.
i. List Recall: In this test, the participants were asked to recall the words provided in the "List Learning" test as many as possible. One point was given for each correctly recalled words, and the total score was 16 points.
j. List Recognition: In this test, the examiner read 20 words aloud, including 10 words provided in the "List Learning" test. The participants were asked to answer whether the words were read before. One point was given for each correct recognition, and the total score was 20 points.
k. Delay Story Recall: In this test, the participants were asked to recall the story provided in the "Story Memory" test. The number of keywords correctly recalled was used as a measure of delay story memory to avoid complicated scoring rules. One point was given for each correctly recalled keywords, and the total score was 12 points.
l. Figure Recall: In this test, the participants were asked to recall and draw the gure provided in the " Figure  SCALING: For the purpose of the present study, a single reference sample was used to provide the scaling metrics. The norm was established based on the test results of 540 people aged 20-89 years. Raw scores were converted into scaled scores for all participants in the present study using the same scaling metric, regardless of age [38][39].

Measurement of serum autoantibodies
The levels of autoantibodies and lipoproteins of patients in the serum were measured within 1 month of the cognitive function testing. In this phase of the study, participants had a blood sample drawn. A venous puncture was done, and 2 mL of blood was collected in a serum separator tube. The serum from the patients was prepared by centrifugation at 3000 rpm for 10 min in a clinical centrifuge. The autoantibodies, including anti-ds-DNA antibody, anti-Sm antibody, anti-Ribonucleoprotein antibody (anti-RNP), anti-Sjögren's syndrome A antibody (anti-SSA), anti-Sjögren's syndrome B antibody (anti-SSB), anti-ribosomal-P antibody, and anti-β2-glycoprotein-I antibody, and immunoglobulins including Immunoglobulin G (IgG), Immunoglobulin M (IgG), Immunoglobulin A (IgA), and Immunoglobulin E (IgE), were analyzed.

Enzyme-linked immunosorbent assay
The expression of apolipoprotein E (APOE, RayBio, USA), apolipoprotein A1 (APOA1, R&D system, USA), insulin-like growth factor-1 (IGF-1, Fcmacs, China), and insulin-like growth factor binding protein 7 (IGFBP7, Arigobio, China) were detected using an enzyme-linked immunosorbent assay (ELISA) kit following the manufacturer's protocols. In short, a 96-well plate was coated with the antibody (100 μL/well) in duplicate and incubated at 4℃ overnight. The plate was then washed ve times with wash buffer and dried gently on a lint-free paper towel. After the sample was diluted, 100 μL of the detection samples was added to each well, and the plate was incubated for 2 h at room temperature. An equal volume (100 μL/well) of the substrate solution was added, incubated at room temperature for 30 min, and then protected from light. The reaction was stopped by adding 100 μL/well of 1M H2SO4, and the optical density was determined at a wavelength of 450 nm 5 min later [40]. The results were tted to the standard curve, and the detection ranges were as follows: APOE (1.5-400 ng/mL), APOA1 (6.3-200 ng/mL), IGF-1 (62.5-4000 pg/mL), and IGFBP7 (625-40000 pg/mL).

Statistical analysis
Data were expressed as means ± standard deviation. Differences between two groups were determined using the unpaired-sample Student t test if the variance was normally distributed, and the Mann-Whitney U test was used for non-normally distributed data. Differences among observed frequencies were tested using the chi-square test, while Pearson's correlation coe cient was used to calculate the correlation between variables [40]. The correlations were analyzed using Spearman's correlation test. Data were analyzed and visualized using SPSS16.0 software or GraphPad Prism 7.0 (GraphPad Software, CA, USA), and a two-tailed P value < 0.05 was considered statistically signi cant.

Clinical characteristics and cognitive function of patients with SLE and HCs
The demographic features of 121 patients with SLE and 65 HCs in the study are summarized in Table 1.
The mean age of patients and HCs was 33.88 ± 11.77 and 36.69 ± 13.88 years, respectively. The mean level of education was 11.95 ± 3.05 years in the patient group, which was signi cantly lower than that in the HC group (P < 0.01).  Figure Copy and Semantic Fluency showed no differences (Table 2). These tests corresponded to ve cognitive domains: immediate memory, visuospatial/construction, attention, language, and delayed memory. After categorizing the aforementioned cognitive function tests according to these ve cognitive domains, the difference between patients and normal volunteers was analyzed. The results showed that levels of immediate memory, visuospatial/construction, attention, and delayed memory were signi cantly lower in patients than in HCs, while language features showed no obvious difference (Fig. 1).  The Graph Pad Prism 7.0 software were used for statistical analysis.

Demographic features, clinical manifestations, and treatment of patients
Based on the RBANS score, 81 patients were divided into high-cognition group (n = 33) and low-cognition group (n = 48). As shown in Table 3, no statistically signi cant differences in age and sex were found between the two groups. The mean level of education was 10.77 ± 3.64 years in the low-cognition group, which was lower than that in the high-cognition group (P < 0.01). The levels of SLEDAI (13.91 ± 11.79 vs 9.38 ± 5.79) signi cantly increased in patients with low cognition compared with those in high cognition (all P < 0.05). Data are expressed as number (percentage). P values are based on Chi-square test for normally distributed variables. *P < 0.05, **P < 0.01, *** P < 0.005, vs. Low RBANS Score. The SPSS Statistics 16.0 software were used for statistical analysis.
Patients with low cognition had more frequent anemia (23.4% vs 3.03%) compared with patients with high cognition (P < 0.05). Meanwhile, the prevalence of rash, mucosal ulcers, hematuria, proteinuria, pyuria, arthritis, vasculitis, pleurisy, pericarditis, low complement, thrombocytopenia, leukopenia, lupus nephritis, and neurological disorder showed no difference between the two groups. In addition, the treatment with prednisone, hydroxychloroquine, cyclophosphamide, azathioprine, methotrexate, and cyclosporine was not different between the two groups (Table 3).

Comparison of clinical and laboratory characteristics of patients
The study investigated the laboratory parameters of patients enrolled in this study ( 14 μg/dL; P < 0.05). The levels of anti-dsDNA antibody (P < 0.05) and IgM (P < 0.01) and the number of B cells (P < 0.05) were higher and the D-dimer levels were considerably higher in patients with low cognition (P < 0.001). The incidence of other clinical indexes, including globulin, 24-h urine protein (24-h UP), C3, C4, C-reactive protein, and creatinine (Cr), displayed no difference between the two groups.
In addition, the difference in the proportion of various blood cells in the peripheral blood between the two groups was detected. The results showed that the number of B cells was higher in the low-cognition group (0.15 ± 0.16 % vs 0.10 ± 0.13 %; P < 0.05), but no difference was found in the numbers of erythrocytes, WBCs, CD3 + cells, CD3 + CD4 + cells, CD3 + CD8 + cells, and natural killer cells between the groups ( Table 4). As shown in Table 4, the serum levels of anti-dsDNA antibodies (P < 0.05) and IgM (P < 0.01) were both higher in patients with low cognition. However, no statistically signi cant differences in positive rates of anti-Sm antibodies, anti-RNP antibodies, anti-SSA antibodies, anti-SSB antibodies, anti-Rib P antibodies, and anti-β 2 GPI antibodies, as well as the levels of IgG, IgA, and IgE, were found between the two groups.

Correlation analysis between cognitive function and IgG, IgM dsDNA, and albumin levels
The characteristics of SLE are high levels of autoantibodies and damages to multiple organs [41]. The serum anti-dsDNA antibodies are considered a diagnostic marker and one of the classi cation criteria for SLE [42][43][44]. Moreover, abnormal immunoglobulin levels were observed in many of the patients. The results showed that the IgG level was inversely associated with neuropsychological tests of immediate memory (List Learning: r = -0.344, P < 0.01; Immediate Story Recall: r = -0.264, P < 0.05; Fig. 2A Fig. 2K). Nevertheless, no correlation was found with other cognitive functions (Supplementary Fig. 2). However, dsDNA did not have a signi cant correlation with any cognitive function ( Supplementary Fig. 3).
The cerebrospinal uid (CSF)/serum quotient of albumin, known as quotient albumin (Q albumin), is widely accepted as a biomarker for estimating blood-brain barrier (BBB) function [45]; BBB plays a critical role in the pathogenesis of NPSLE [46]. The potential relationship between the serum albumin levels in SLE and cognitive dysfunction was analyzed. The results showed that the albumin levels correlated positively with List Learning, Immediate Story Recall, and Figure Copy (all P < 0.05; Fig. 2L, M, and 2N), as well as List Recognition Test and Delayed Story Recall (both P < 0.01; Fig. 2O and 2P), but had no relationship with other cognitions (Supplementary Fig. 4).

Correlation analysis between cognitive function and C3, C4, and lipoprotein levels
Studies indicated that a signi cant higher prevalence of thyroid autoantibodies was observed in patients with SLE compared with HCs [47][48]. Many studies tried to associate thyroid abnormalities with clinical ndings of SLE, but with no uni ed conclusion [49][50][51]. This study explored the relationship between the serum T3 and T4 levels and cognitive functions in patients with SLE. A regression analysis revealed that the serum T3 levels positively correlated with Immediate Memory, including List Learning (r = 0.293, P < 0.05; Fig. 3A) and Immediate Story Recall (r = 0.269, P < 0.05; Fig. 3B). Moreover, Figure Copy  The results indicated that T3 and T4 levels positively correlated with most cognitive functions. Increasing evidence shows that thyroid hormones (mainly T3 and T4) were involved in all aspects of lipid metabolism [52][53][54]. The relationship between thyroid hormones and lipid metabolism is evident in patients with thyroid dysfunction [55][56]. Moreover, lipid metabolism was markedly altered in patients with SLE [57]. The serum of 81 patients was collected for ELISA, and the expression of four lipoproteins (APOE, APOA1, IGF-1, and IGFBP7) was analyzed to explore any correlation between lipid metabolism and cognitive dysfunction in patients with SLE. The results showed no signi cant difference in the expression of these four proteins between the two groups ( Fig. 4A, 4B, 4C, and 4D). Then, the correlations between the levels of these four lipoproteins and cognitive function were analyzed. The serum APOE levels signi cantly negatively correlated with Line Orientation (r = -0.206, P <0.05; Fig. 4E), and IGFBP7 and Figure Recall (r = -0.223, P < 0.05; Fig. 4F), whereas APOA1 showed a positive correlation with Coding (r = -0.207, P < 0.05; Fig. 4G). Apart from this, no other correlations were found.

Discussion
In the present study, the cognitive function was lower in patients with SLE compared with HCs. Lower education; higher SLEDAI score; lower albumin, T3, and T4 levels; and higher D-dimer, anti-dsDNA antibodies, and IgM levels were found in patients with low cognition compared with those with high cognition. Besides, cognitive function was closely associated with the IgG, IgM, albumin, T3, and T4 levels in patients. Apart from this, four lipoproteins (APOE, APOA1, IGF-1, and IGFBP7) were detected by ELISA. Although these four lipoproteins showed no difference between the two groups, APOE, APOA1, and IGFBP7 expression correlated with cognition. An important nding of the present study was that the abnormal levels of serum T3, T4, and lipoproteins might be associated with cognitive dysfunction in SLE.
NPSLE is a severe complication of SLE, which results in severe neurodegenerative changes and threatens life [58]. NPSLE is characterized by a variety of neurological manifestations, making the diagnosis of NPSLE a formidable challenge for rheumatologists [59][60]. Patients with NPSLE had multiple NP events, of which 91.2% affected the CNS, including cognitive dysfunction, headache, mood disorder, seizures, anxiety, and psychoses. The most common clinical manifestations were cognitive dysfunction, which occurred in 42.1% of patients [61]. In addition, this study showed that nervous system involvement was the initial presentation of SLE, while more than 50% of NPSLE occurred within the rst 5 years after the onset of SLE [61]. This nding indicated that patients with SLE were likely to undergo cognitive dysfunction before the diagnosis of NPSLE. Therefore, 121 patients with SLE and 65 HCs were recruited for the cognitive assessment using RBANS. The results showed that the cognitive functions of patients with SLE, including immediate memory, visuospatial/constructional, attention, and delayed memory, were signi cantly lower than the cognitive functions of HCs before the diagnosis of NPSLE. Patients with SLE had a worse response in the lingual gyrus compared with the HCs; this region was associated with visual attention, visual encoding/processing, and working memory [62][63]. This opinion might explain why patients with SLE performed worse on attention, memory task, and visuospatial/constructional memory.
SLE is a prototypic autoimmune disease caused by the loss of B cell tolerance and subsequent recognition of self-antigens and becoming autoreactive [64][65]. In this study, the proportion of B cells obviously increased in patients. Also, the B cell hyper-reactivity could induce an accumulation of autoreactive plasma cells [66], which secreted a variety of autoantibodies, including anti-dsDNA, anti-Sm, anti-Ro, anti-SSA, anti-SSB, anti-Rib-P, and anti-β2-GPI, to mediate the occurrence of SLE [67]. The levels of anti-dsDNA and IgM increased in the low-cognition group compared with the high-cognition group, but anti-Sm, anti-RNP, anti-SSA, anti-SSB, anti-Rib-P, anti-β2-GPI, IgG, IgA, and IgE levels all showed no difference. The serum levels of anti-dsDNA antibodies uctuated with disease activity, whereas others were not altered (anti-Sm, anti-Ro, anti-La, anti-cardiolipin) [68] owing to the different origins of antibodysecreting cells, which partially explained the results. However, in a cohort of 100 randomly selected patients, the higher positive rate of IgG anti-cardiolipin antibodies was found in patients with moderate/severe cognitive impairment, but no difference was found in the incidence of other autoantibodies [67]. Besides, in the present study, the plasma albumin level, as a biomarker of disease activity in SLE [69], decreased in the low-cognition group. The results indicated that the changes in serum IgM, dsDNA, and albumin levels might be associated with cognitive dysfunction in patients with SLE. As expected, IgM showed negative correlations with cognitive function, albumin correlated positively with cognitive function, but dsDNA showed no correlation.
The association between thyroid disease and SLE has been reported for more than 50 years [70]. Both hypothyroidism and thyroid nodules are found more frequently in patients with SLE than in the general population [71]. The rate of thyroid cancer is twice as prevalent in patients with SLE as in those without SLE [72]. In recent years, the impact of serum thyroxine levels on cognitive function has attracted widespread attention gradually, but no consistent conclusions have been drawn. A study on mild cognitive impairment and dementia showed that patients with relatively high T3 levels showed impairment in memory and visuospatial and executive functions [73]. However, patients with acute coronary syndrome having low T3 levels had a poorer health-related quality of life (including general health, social functioning, and role emotional) than those with normal levels at a 1-year follow-up [74]. A study on subjective cognitive decline showed that higher T3 levels were associated with better verbal memory performance (immediate and delayed recall tasks) in APOE ε4 carriers [75]. The results showed that T3 and T4 levels decreased in patients with low cognition, and a signi cant positive correlation was found between T3 and T4 levels and cognitive function, which was consistent with the aforementioned view and supported that T3 and T4 had protective cognitive functions.
Dyslipidemia is one of the major risk factors for SLE, leading to a high prevalence of premature atherosclerosis and coronary artery disease in patients with SLE [76][77]. In addition, impaired renal function is also associated with dyslipidemia [78][79]. Thyroid hormones in uence all aspects of lipid metabolism; especially T3 induces LDL receptor gene expression, enhancing LDL clearance [80]. A report revealed that low free T3 was an independent risk factor for dyslipidemia in patients with SLE [81]. The present study did not show the difference in the expression of serum lipoproteins (APOE, APOA1, IGF-1, and IGFBP7) between the two groups, but three of the aforementioned parameters showed a correlation with the cognitive function: APOE negatively correlated with Line Orientation, APOA1 positively correlated with Coding, and IGFBP7 negatively correlated with Figure Recall. Four years ago, an interesting result was reported: APOE knock-out (KO) mice had synaptic loss and cognitive dysfunction. Although these mice had synaptic loss and dysfunction similar to that in APOE KO mice, they did not have the learning and memory impairment observed in APOE KO mice; also, the memory de cit in the APOE KO mice was speci c to female mice [82]. These results indicated that the cognitive impairment in mice was speci cally caused by abnormal peripheral blood lipids but not by the expression of APOE in the brain. In AD, accelerated cognitive decline and abnormal internal environment, structure, and function of the brain were also found in APOE ε4 carriers [83]. Leon et al. compiled "The Role of APOE in Cerebrovascular Dysfunction" and pointed out multiple mechanisms of APOE involvement in cognitive dysfunction [84], which could be used to explain the results of the present study. ApoA1 is responsible for transporting cholesterol to the liver. It is a critical component in the formation of HDL. The overexpression of ApoA1 might effectively inhibit the age-related decline in memory and learning ability [85][86]. Moreover, Kawano et al. found that the levels of ApoA1 were strikingly lower in a group of late-onset nonfamilial AD [87].
These studies and the results of the present study both proved a positive correlation between APOA1 and cognitive functions. IGF-1 is an essential neurotrophic factor produced both peripherally and in the brain; adequate levels of serum IGF-1 may be necessary for normal cognitive functioning [88][89]. The serum IGF-1 levels were signi cantly elevated in patients with SLE and inversely correlated with age [90]. Shankar et al. supported that both high and low levels of IGF-1 might induce poor cognitive function and that optimum levels of IGF-1 might be associated with better cognitive function [91]. Therefore, this explained why the correlation between IGF and cognitive function was not observed in patients with high and low cognitive function.

Conclusion
In conclusion, the ndings indicated that T3 and T4 levels frequently decreased in patients with low cognition, and the cognitive function was associated with T3, T4, APOE, APOA1, and IGFBP7 levels in patients with SLE. The results suggested that T3 and T4 could be used in clinical practice as biomarkers for cognitive dysfunction in SLE. Availability of data and materials The datasets generated during and analyzed during the current study are not publicly available due to respect participants' rights to privacy and to protect their identity, but are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.