Anti–N-methyl-D-aspartate receptor antibodies, cognitive dysfunction, and depression in systemic lupus erythematosus

Abstract

Objective

To assess the association of cognitive dysfunction and depression with serum antibodies to N-methyl-D-aspartate (NMDA) receptor (anti-NR2) and analyze clinical and neuroimaging correlates in patients with systemic lupus erythematosus (SLE).

Methods

Sixty patients underwent neurocognitive assessment, evaluation for depression with the Beck Depression Inventory II (BDI-II) and psychiatric interview (Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition [DSM-IV] criteria), brain magnetic resonance imaging, and proton magnetic resonance spectroscopy imaging (¹H-MRSI). Cognition was assessed in 5 domains: memory, attention/executive, visuospatial, motor, and psychomotor, and adjusted to each individual's best level of prior cognitive functioning estimated from the reading subtest of the Wide Range Achievement Test–3 (WRAT-3). Serum anti-NR2 antibodies were measured by enzyme-linked immunosorbent assay using a pentapeptide from the human NMDA receptor.

Results

Cognitive dysfunction was found in 28 of 60 patients (mild in 8, moderate in 20) before adjustment for WRAT-3 and in 35 of 60 patients (mild in 15, moderate in 11, and severe in 9) after adjustment for WRAT-3. The changes were most pronounced in the memory and visuospatial domains. There was no significant association between anti-NR2 antibody levels and cognition. On ¹H-MRSI, patients with moderate or severe cognitive dysfunction had significantly higher choline:creatine ratios in the dorsolateral prefrontal cortex and the white matter, compared with patients with mild or absent cognitive dysfunction. Anti-NR2 antibodies were significantly correlated with BDI scores; patients with BDI-II scores of ≥14 had higher serum levels of anti-NR2 antibodies (P = 0.005, 95% confidence interval 0.83, 4.31), and there was a trend toward higher anti-NR2 antibody levels among patients who fulfilled the DSM-IV criteria for major depression.

Conclusion

Serum anti-NR2 antibodies are associated with depressive mood but not with cognitive dysfunction in SLE at a given time point. Larger longitudinal studies are needed to address the possible association between anti-NR2 antibodies and depression in SLE.

Cognitive dysfunction and depression are common manifestations of neuropsychiatric systemic lupus erythematosus (SLE) (1, 2), with the reported prevalence of cognitive dysfunction ranging from 20% to 80% in previous studies (3–6). The pathogenetic mechanisms of both cognitive decline and depression in SLE remain elusive and are likely multifactorial. A variety of factors have been implicated, including antiphospholipid antibodies (7, 8), autoantibody-induced direct neuronal damage (9), medications (10, 11), and primary neurologic and psychiatric disorders affecting cognition (4). Some studies have shown that higher disease activity was predictive of subsequent cognitive dysfunction (12), whereas others showed no association between cognitive dysfunction and disease activity or use of corticosteroids (5, 10, 11, 13). In one study (14), clinically diagnosed depression associated with cognitive dysfunction was found to be an independent risk factor predicting poor test performance. However, cognitive decline and depression in SLE may often occur in the absence of any known risk factors and without a recognizable inflammatory response.

Cognitive function is usually assessed with a battery of neurocognitive tests. The reading subtest of the Wide Range Achievement Test–3 (WRAT-3) (15) is an instrument used in neuropsychology to estimate the best level of prior (premorbid) cognitive functioning in English first-language speakers (16, 17). To improve the precision of measuring cognitive decline in SLE, we used the reading subtest of the WRAT-3 to estimate the premorbid cognitive abilities in participants in the present study and compared their current performance on neurocognitive tests with their estimated level of prior cognitive functioning.

Proton magnetic resonance spectroscopy imaging (¹H-MRSI) is an application of magnetic resonance spectroscopy that allows noninvasive measurement of brain neurometabolites, such as N-acetylaspartate (NAA), creatine (Cr), and choline (Cho). The NAA concentration is considered to be a measure of neuronal density and neuronal integrity in the adult central nervous system (CNS) (18); change in choline is associated with membrane breakdown, possibly related to inflammation (19), demyelination (20), or ischemia (21), and creatine concentration is thought to reflect intracellular energy stores (18). It is commonly accepted that the ratios of these neurometabolites provide more reliable estimates than the absolute values of the single metabolites. Several studies have examined neurometabolic changes in the brains of patients with neuropsychiatric SLE, and have shown reductions in the NAA:Cr ratio (22) and NAA:Cho ratio (23) and increases in the Cho:Cr ratio (21, 24, 25). In the present study, we performed ¹H-MRSI and measured concentrations of NAA, choline, and creatine in the brains of SLE patients.

Anti-NR2 antibodies are a subset of pathogenic murine anti–double-stranded DNA (anti-dsDNA) antibodies that cross-react with a consensus peptide sequence of the extracellular, ligand-binding domain of mouse and human N-methyl-D-aspartate (NMDA) receptors NR2a and NR2b (26). Under physiologic conditions, NR2 receptors are expressed on the neurons throughout the hippocampus and cortex (27) and bind the neurotransmitter glutamate; they have been implicated in fundamental mechanisms of synaptic plasticity underlying learning and memory (28, 29). In mice, introduction of anti-NR2 antibodies into the CNS either by direct intracranial injection or through experimentally induced breakdown in the blood–brain barrier caused excitatory neuronal cell loss and led to impaired memory, learning (30), and fear response (31). Other preclinical studies have shown that NMDA receptor antagonists had antidepressant-like activity in animal models (32). Furthermore, NMDA receptor abnormalities have been observed in patients with major depression and in suicide victims (33, 34). These data suggest that anti-NR2 antibodies may be associated with various manifestations of neuropsychiatric lupus, including cognitive dysfunction and depression.

The primary objective of the present study was to test whether serum anti-NR2 antibodies are associated with cognitive dysfunction in SLE. The secondary objectives were to analyze the association of clinical, serologic, and neuroimaging features of SLE with cognitive dysfunction. The analysis of association between anti-NR2 antibodies and depression was performed after analysis of the primary objective. Exploratory analyses were performed to address the association between anti-NR2 antibodies and concentrations of neurometabolites measured by ¹H-MRSI.

PATIENTS AND METHODS

Patients.

Sixty consecutive patients were enrolled in this cross-sectional study between December 2002 and June 2004. All patients were ≥18 years of age and fulfilled the 1997 updated American College of Rheumatology criteria for the diagnosis of SLE (35, 36). Patients were screened and recruited from the rheumatology clinic at the National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS), National Institutes of Health and from local tertiary care centers and private rheumatology practices. Patients were ineligible for the study if they had a history of neurologic disease leading to brain damage (including head injury resulting in loss of consciousness, strokes, seizures, or toxic exposure), if they had had a clinically documented transient ischemic attack within 6 months of the screening visit, or if they were currently being treated with anticonvulsant agents. Patients for whom English was not the first language were also excluded since the neurocognitive tests used have been validated only for English as first language speakers. Patients were excluded from the neuroimaging studies if they had any of the usual contraindications for magnetic resonance imaging (MRI). The study was approved by the Institutional Review Board of the NIAMS. All study participants provided written informed consent prior to institution of the study procedures.

Data collection.

Each participant was evaluated during 1 or 2 study visits, scheduled no longer than 2 weeks apart. In addition to a complete history, physical examination, and routine laboratory testing, patients were evaluated for depression with a self-administered questionnaire, the Beck Depression Inventory, Second Edition (BDI-II) (37), and psychiatric interview performed by a psychiatrist who was blinded with regard to the anti-NR2 antibody levels and results of the neurocognitive tests, neurocognitive assessment, and MRI of the brain. Patients were considered clinically depressed if they met Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) (38) criteria for the diagnosis of major depression on the day of study evaluation. Current SLE disease activity was measured by SLE Disease Activity Index (SLEDAI) (39, 40). Information on current and past medication regimens was obtained during the interview and from patients' medical records.

Laboratory analysis.

Autoantibodies tested included antinuclear antibodies, anti-dsDNA, anticardiolipin antibodies (aCL) (IgM and IgG isotypes), anti–β₂-glycoprotein I antibodies (anti-β₂GPI) (IgA, IgM, and IgG isotypes), anti–ribosomal P antibodies (anti-P), a panel of antineuronal nuclear antibodies (ANNA-1 and ANNA-2), and IgG anti-NR2 antibodies. Tests to determine the serum levels of anti-P, anti-β₂GPI, ANNA-1, and ANNA-2 were performed at Mayo Medical Laboratory (Rochester, MN). Anti-NR2 antibody levels were measured by enzyme-linked immunosorbent assay (ELISA) using the 283–287–pentapeptide consensus sequence of the human NMDA receptor NR2a and NR2b subunits, according to a previously described technique (26). Since the optical density (OD) values in the controls differed slightly on each ELISA plate, we used the Z scores of the anti-NR2 concentrations to define the level of the antibodies. Anti-NR2 antibody levels were compared with those of healthy controls and categorized as absent (Z score ≤1), low (Z score between 1 and 2), or high (Z score ≥2).

Neuropsychological assessment and ascertainment of cognitive function.

Cognitive function was assessed with the following neurocognitive tests: the California Verbal Learning Test (41), the Rey-Osterrieth Complex Figure Test (copy and recall) (42), the Stroop Color, Word, and Color-Word Tests (43), the Digit Symbol-Coding Subtest of the Wechsler Adult Intelligence Scale (44), the Trail Making Test (part B) (45), the Controlled Oral Word Association Test (45), and the Grooved Pegboard Test (45) for both the dominant and nondominant hands. The tests were grouped into 5 domains to evaluate 5 aspects of cognition: memory, attention/executive function, visuospatial skills, motor function, and psychomotor speed (Figure 1). The reading subtest of the WRAT-3 was administered to each patient to estimate premorbid cognitive abilities and to assess the decline in cognitive function by comparing the estimated premorbid and current cognitive status. Cognitive function was assessed using 2 parallel analyses: unadjusted (based on the current performance on the neurocognitive tests) and adjusted (current performance adjusted to the premorbid level of cognition, estimated based on results of the reading subtest of the WRAT-3).

Figure 1. Neurocognitive assessment. Neurocognitive tests were grouped to assess different cognitive domains (memory, attention/executive, visuospatial, motor, and psychomotor). CVLT = California Verbal Learning Test.

Unadjusted analysis.

For each patient, the raw scores from each test were compared with published norms (age-, sex-, and education level–corrected, as appropriate) and transformed into Z scores to express the deviation from the normal mean. Mean domain Z scores were defined as the average of the Z scores from the tests comprising each domain. To express cognitive function as a composite score, the Z score for each domain was transformed into a Domain Cognitive Dysfunction Score as shown in Table 1, with higher values representing more impairment in a particular domain. The sum of all Domain Cognitive Dysfunction Scores across the 5 domains resulted in the Global Cognitive Dysfunction Score, which was transformed into a Global Cognitive Dysfunction Category as described below.

Table 1. Scoring and categorization of cognitive dysfunction*

* The composite score is constructed from the bottom to the top of the table.
Global Cognitive Category	Defined from Global Cognitive Dysfunction Score (GCDS)
Absent	GCDS 0–1
Mild	GCDS 2–3
Moderate	GCDS 4–5
Severe	GCDS ≥6
Global Cognitive Dysfunction Score	Sum of Domain Dysfunction Scores over 5 domains
Domain Cognitive Dysfunction Score (DCDS)	If Z score is ≥−1 SD, then DCDS is 0; if Z score is <−1 SD but ≥−2 SD, then DCDS is 1; if Z score is <−2 SD, then DCDS is 2.
Domain Z scores	Average of the Z scores in the tests comprising each domain
Test Z scores	Compared with age- and sex-matched published normal values
Test raw scores	Obtained from performance on the neurocognitive testing

Adjusted analysis.

To determine the degree of individual cognitive decline, we estimated each participant's premorbid cognitive abilities from performance on the reading subtest of the WRAT-3. The domain Z scores were then adjusted for each patient's premorbid cognitive function by subtracting the WRAT-3 Z score from the domain Z score (46). The adjusted Domain Cognitive Dysfunction Scores and Global Cognitive Dysfunction Scores were derived from the adjusted domain Z scores through the same transformations as in the unadjusted analysis. The patient's Global Cognitive Dysfunction Category was then classified as absent, mild, moderate, or severe, based on the Global Cognitive Dysfunction Scores in both adjusted and unadjusted analyses (Table 1).

Imaging procedures.

All MRI examinations were performed using a 3T magnet (General Electric Medical Systems, Milwaukee, WI). Three-dimensional gradient-echo T1-weighted images, proton-density, and T2-weighted scans were acquired. The T1-weighted scans were repeated after intravenous administration of 0.1 mmole/kg Magnevist (Berlex, Wayne, NJ). All films were read by an independent neuroradiologist who was blinded with regard to the results of the neurocognitive testing.

¹H-MRSI was performed with a multislice imaging technique (4 slices, 7.5mm cubic voxel dimensions, spin-echo slice selection, repetition time 2,300 msec, echo time 280 msec, water suppression, lipid signal from the scalp suppressed with outer volume saturation pulses, nominal voxel resolution 0.42 ml). Regions of interest (ROIs) were drawn on the left and right dorsolateral prefrontal cortex and the white matter of the centrum semiovale, on structural MRI scans corresponding to the MRSI slices (47, 48). Metabolite signals were reported as ratios of the area under the peaks for NAA, creatine (plus phosphocreatine), and choline averaged over the voxels in the ROIs. Quality control procedures were undertaken to eliminate voxels with obvious artifact.

Statistical analysis.

Global Cognitive Dysfunction Scores were compared in patients grouped by antibody level and depression status. The binary outcome variable for the antibody testing was serum anti-NR2 antibody status, defined as present versus absent or low/absent versus high. The binary outcome variable for depression status was clinical depression present versus absent. The results were verified through analysis of the domain Z scores and single-test Z scores. Descriptive statistics were computed for all study variables. When the continuous variables met assumptions for parametric distribution, the unpaired t-test was used for comparison between groups. Variables not meeting distribution assumptions were analyzed using the nonparametric Wilcoxon rank sum test. Two sided P values less than 0.05 were considered significant. One-way analysis of variance (ANOVA) was used to compare concentrations of anti-NR2 antibodies in patients with different degrees of cognitive impairment. A correlation matrix was constructed and multivariable regression analysis performed to explore any association between Global Cognitive Dysfunction Score and anti-NR2 antibody level, age, education level, current prednisone dosage on the day of neurocognitive evaluation, cumulative steroid dose over lifetime, SLEDAI score, or depression status. Logistic regression analysis was performed to account for confounding effects of disease activity and duration, steroid treatment, age, and education level on the association between depressive mood (defined as a BDI-II score of ≥14) and anti-NR2 antibody positivity. StatView version 5.0.1. and SAS E-Guide 3.0 software (SAS, Cary, NC) was used for all analyses.

RESULTS

Demographic characteristics of the patients.

The mean age of the patients enrolled in the study was 41 years (range 19–67). Approximately three-fourths of the participants were female, 53% were white, and 27% were African American. Importantly, the mean level of education was 14.9 years, indicating a relatively high level of expected cognitive abilities in this cohort. Eighty percent of the patients had received high-dose steroids in the past for severe lupus manifestations, but at the time of study evaluation the majority of patients were receiving low-dose or no corticosteroids because their disease was in remission (Table 2).

Table 2. Demographic characteristics of the study participants*

* Except where indicated otherwise, values are the number (%). Anti-dsDNA = anti–double-standed DNA; SLEDAI = Systemic Lupus Erythematosus Disease Activity Index.
Age, mean ± SD years	41 ± 13
Female/male	47 (78)/13 (22)
Ethnicity
White	32 (53)
African American	16 (27)
Asian	4 (7)
Hispanic	8 (13)
Education, mean ± SD years	14.9 ± 3.06
Disease duration, mean ± SD years	13.8 ± 10.2
Anti-dsDNA positive	33 (55)
SLEDAI ≤4	56 (93)
History of organ system involvement
Renal	41 (68)
Hematologic	26 (43)
Musculoskeletal	43 (72)
Mucocutaneous	30 (50)
Current daily prednisone dosage, median (range)	2.5 (0–30)
Prednisone dosage subgroup
0 mg/day	22 (37)
<3 mg/day	16 (26)
3–<10 mg/day	12 (20)
10–<20 mg/day	7 (12)
≥20 mg/day	3 (5)
Cumulative prednisone dose
<10 gm	12 (20)
10–20 gm	12 (20)
>20 gm	36 (60)
Concurrent immunosuppressive therapy
Mycophenolate mofetil	5 (8)
Hydroxychloroquine	39 (65)
Cyclophosphamide	5 (8)

Cognitive dysfunction and individual cognitive decline.

Based on current performance on the neuropsychological tests, patients' mean domain Z scores were in the low average to mildly impaired range (Z score of −0.6 or less) across 3 of the 5 domains (Figure 2A). The lowest scores were in the memory and visuospatial domains. When the WRAT-3 (group mean Z score 0.15) was administered to each patient, it became evident that in approximately half of the patients, the expected level of functioning was in the high average range or better compared with the general population (46.7% scored at or above a Z score of 0.6), and their current performance across all 5 cognitive domains represented a substantial deviation from their estimated premorbid level of cognition. When patients were categorized into Global Cognitive Dysfunction Categories (Figure 2B), unadjusted analysis revealed that 28 of 60 (47%) had global cognitive dysfunction (moderate in 8, mild in 20). Incorporation of WRAT-3 results into the scoring resulted in a higher proportion of patients showing global cognitive dysfunction (35 of 60 [58%]), with more severe impairment (severe in 9, moderate in 11, mild in 15).

Figure 2. A, Cognitive performance compared with premorbid cognitive abilities. Values are the mean and SEM of current performance in different domains (memory, attention/executive, visuospatial, motor, and psychomotor) compared with the expected level of cognition assessed with the reading subtest of the Wide Range Achievement Test 3 (WRAT) (mean ± SEM 0.15 ± 0.13). B, Distribution of patients by Global Cognitive Dysfunction Category in the unadjusted analysis (open bars) (cognitive dysfunction absent in 53%, mild in 33%, moderate in 13%) and the analysis adjusted for premorbid cognitive ability (solid bars) (cognitive dysfunction absent in 42%, mild in 25%, moderate in 18%, severe in 15%).

Analysis of association between anti-NR2 antibodies and cognitive dysfunction.

Anti-NR2 antibodies were measured in all patients, and levels were categorized as absent (n = 40), low (n = 4), or high (n = 16). When we compared adjusted Global Cognitive Dysfunction Scores between the subgroup of patients who did and the subgroup who did not have serum anti-NR2 antibodies, we found no difference in cognitive abilities between the 2 groups (mean ± SD Global Cognitive Dysfunction Score 2.62 ± 2.46 and 2.65 ± 2.4, respectively; P = 0.97). To explore the possibility of association between only high-level antibodies and cognitive dysfunction, we compared patients with high levels of anti-NR2 antibodies and those who had either low-level or absent anti-NR2. The Global Cognitive Dysfunction Scores in patients with high-level anti-NR2 antibodies did not differ from scores in the remainder of the patients (mean ± SD 2.6 ± 2.5 and 2.8 ± 2.3, respectively; P = 0.82). Analyses with Global Cognitive Dysfunction Score unadjusted for premorbid cognitive abilities revealed similar results. Moreover, we found no association with anti-NR2 antibodies when patients' performances across single tests and domains were compared.

We also compared the mean concentration of anti-NR2 antibodies in patients classified into the 4 Global Cognitive Dysfunction Categories. No association was found between higher levels of anti-NR2 antibodies and more pronounced dysfunction, in either the adjusted analysis (Figure 3) or the unadjusted analysis (data not shown) (F = 0.8, P = 0.9 by ANOVA).

Figure 3. Cognitive impairment and levels of antibodies to N-methyl-D-aspartate receptor (anti-NR2). In the adjusted analysis, no association was found between degree of cognitive impairment and level of anti-NR2 antibodies (F = 0.8, P = 0.9 by analysis of variance). Values are the mean and SEM Z scores for anti-NR2 antibody levels (0.653 ± 0.544, 0.871 ± 0.807, 0.316 ± 0.676, and 0.653 ± 1.045 in patients in the absent, mild, moderate, and severe cognitive dysfunction groups, respectively).

Analysis of association between other antibodies and cognitive dysfunction and between clinical characteristics and cognitive dysfunction.

We found no association between cognitive dysfunction and anti-dsDNA, aCL, anti-P, or anti-β₂GPI antibodies (data not shown). All patients were negative for ANNA-1 and ANNA-2 antibodies. In a correlation analysis, only education level showed inverse and mild correlation with cognitive dysfunction, in both adjusted and unadjusted analyses (r [Spearman's correlation] = 0.4, P = 0.0015 in the unadjusted analysis); statistical significance remained after accounting for SLE disease activity, depression, steroid therapy, and age. In the regression analysis, we found no association between cognitive dysfunction and disease activity, depression, prednisone treatment, or age.

Anti-NR2 antibodies and depression.

Eleven of sixty patients had depressive symptoms (BDI-II score ≥14). Six of these 11 patients had high levels of anti-NR2 antibodies, as compared with 10 of 48 patients without depressive symptoms. Higher mean levels of serum anti-NR2 antibodies were observed in patients with depressive symptoms than in those without (mean ± SD Z score 2.75 ± 3.0 versus 0.2 ± 2.5) (P = 0.005, 95% confidence interval [95% CI] for the difference 0.83, 4.31) (Figure 4A), and patients with high levels of anti-NR2 antibodies were found to have higher BDI-II scores compared with patients with low-level or absent anti-NR2 antibodies (P = 0.006, 95% CI for the difference 1.9, 10.8). In logistic regression analysis, anti-NR2 concentration remained significantly associated with depressed mood after controlling for SLEDAI, disease duration, age, education level, and steroid treatment (P = 0.004). Moreover, BDI-II scores correlated with anti-NR2 antibody Z scores (r = 0.4, 95% CI 0.16, 0.59), whereas none of the other variables did.

Figure 4. Depression and levels of antibodies to N-methyl-D-aspartate receptor (anti-NR2). A, Mean and SD Z scores for anti-NR2 antibody levels in patients with depressive mood (defined as a score of ≥14 on the Beck Depression Inventory, Second Edition [BDI-II]) (n = 11) and those without depressive mood (n = 48) (mean ± SD 2.75 ± 3.0 and 0.2 ± 2.5, respectively) (P = 0.005, 95% confidence interval [95% CI] for the difference 0.83, 4.31, by t-test). One patient was omitted from the analysis because of incomplete data on the BDI-II. B, Mean and SD Z scores for anti-NR2 antibody levels in patients defined as having clinical depression (Depressed; n = 10) and those not defined as having clinical depression (Non-depressed; n = 50) (mean ± SD 1.827 ± 3.0 and 0.4 ± 2.7, respectively) (P = 0.14, 95%CI for the difference −3.3, 0.47).

Ten patients meeting DSM-IV criteria for major depression on the day of study evaluation were considered clinically depressed. These patients had higher mean levels of anti-NR2 antibodies compared with the 50 patients not meeting the DSM-IV criteria, but the difference did not reach statistical significance (mean ± SD Z score 1.827 ± 3.0 versus 0.4 ± 2.7) (P = 0.14, 95% CI −3.3, 0.47) (Figure 4B).

Brain imaging and cognitive dysfunction.

Fifty-three study participants underwent MRI of the brain. Seven patients could not complete the MRI study: 4 withdrew because of anxiety, and 3 did not fit into the scanner. Of the 53 patients, 6 were found to have 2 or more nonspecific white matter lesions and/or silent lacunar infarctions. Since multiple nonspecific punctate lesions are frequently found in SLE patients and likely represent ischemic changes (49, 50), we compared cognitive performance of patients with 2 or more punctate lesions and/or silent lacunar infarctions on the MRI (n = 6) with the rest of the study participants (n = 46). One patient with incidentally detected meningioma was excluded from the analysis.

Patients with 2 or more lesions or silent cerebrovascular accidents demonstrated poorer performance on the Grooved Pegboard Test in both the dominant and nondominant hands (dominant hand P = 0.002, nondominant hand P = 0.046), resulting in scores reflecting impaired function in the motor domain (mean ± SD Domain Cognitive Dysfunction Score 1.167 ± 0.567 in patients with lesions and/or silent small cerebrovascular accidents, 0.370 ± 0.283 in patients with normal MRI results; P = 0.002).

Forty-two study participants underwent proton MRS of the brain. Neurometabolite concentration ratios were compared in patients with mild or absent cognitive dysfunction (n = 30) and patients with moderate or severe cognitive dysfunction (n = 12). Compared with patients in whom cognitive dysfunction was mild or absent, patients with moderate or severe cognitive dysfunction had higher Cho:Cr ratios in the left dorsolateral prefrontal cortex (mean ± SD 1.123 ± 0.119 versus 1.024 ± 0.079) (P = 0.003, 95% CI 0.035, 0.162), in the right dorsolateral prefrontal cortex (0.130 ± 0.137 versus 1.024 ± 0.017) (P = 0.006, 95% CI 0.033, 0.179), and in the white matter (1.440 ± 0.147 versus 1.322 ± 0.147) (P = 0.02, 95% CI 0.017, 0.22). There was no association between presence of depression or anti-NR2 antibodies and neurometabolite concentrations in the dorsolateral prefrontal cortex or white matter.

DISCUSSION

The primary goal of this study was to assess the association between serum antibodies to the NMDA receptor and cognitive decline in a cohort of lupus patients with no concurrent or prior history of CNS disease resulting in known brain damage that could explain cognitive dysfunction. We also evaluated an association between anti-NR2 antibodies and depression.

In this cross-sectional study, the association between serum anti-NR2 antibody levels and depression was suggested by several observations. Patients with BDI-II scores of ≥14 had significantly higher anti-NR2 levels than patients with lower scores, and the correlation between BDI scores and anti-NR2 levels remained significant after adjusting for disease activity assessed by SLEDAI, disease duration, age, education level, and steroid treatment.

Moreover, patients who met DSM-IV criteria for major depression on the day of evaluation had higher levels of anti-NR2 antibodies compared with patients who did not meet these criteria. Although the latter difference did not reach statistical significance, the trend observed in this study supports the notion of a possible association. The lack of statistical significance could be due to a variety of factors. First, our sample size was chosen based on the assumption of association between cognitive dysfunction and anti-NR2 antibody levels and not to demonstrate the difference in anti-NR2 levels between depressed and nondepressed patients. In addition, we used a very conservative approach to define clinical depression, by strictly applying DSM-IV criteria. Thus, even patients who had been diagnosed as having depression or were treated with antidepressant medications at the time of study were classified for the analysis as clinically depressed only if they met DSM-IV criteria for major depression on the day of study evaluation. Applying these strict criteria may underestimate the degree of association between anti-NR2 antibodies and depression, but avoids ascertainment or recall bias.

Our findings are consistent with those of a recently published independent study (51) that also showed a positive correlation between anti-NR2 antibodies and BDI scores in patients with SLE. A possible role of anti-NR2 antibodies in depression is supported by previous observations that intravenous administration of low doses of the NMDA receptor antagonist ketamine resulted in rapid improvement of depressive symptoms and in reduction in Hamilton Depression Rating Scale scores by ∼15 points (52, 53). Similarly, it has been found that memantine, another NMDA receptor antagonist, had antidepressive effects in patients with dementia (54). Of interest, many conventional antidepressive agents (monoamine oxidase inhibitors, tricyclic antidepressants, and selective serotonin reuptake inhibitors) were observed to induce reductions in NMDA messenger RNA levels (32). Since anti-NR2 antibodies can cause overstimulation of neuronal cells, mimicking other NMDA agonists, they may contribute to the development of depression in SLE.

We assessed cognitive dysfunction with a standard battery of neurocognitive tests, which we grouped to reflect dysfunction in different cognitive domains. We created, a priori, a system in which domain Z scores were transformed into the Domain Cognitive Dysfunction Scores and further into the composite score, the Global Cognitive Dysfunction Score. This approach enabled us to compare patients with severe abnormalities in 1 or 2 domains with patients with mild or moderate changes in several domains. Although this approach was not validated previously, our results have remained consistent across the different levels, supporting its validity.

Because this was a cross-sectional study of a highly educated cohort of patients, we were concerned that subtle declines in cognition might be masked if only current performance were measured. Including the reading subtest of the WRAT-3, an instrument that estimates the best level of prior cognitive abilities, allowed us to identify changes in cognition for each individual that would have not been recognized through comparison of patients' performance with published norms only. The major limitation of this and other similar tests is that the overall premorbid level of functioning may not be accurately estimated from a single “spared” domain (55). However, this test has been used as an estimate measure of premorbid cognition in other patient populations (46). Our analysis showed that adjusting for premorbid intelligence improves the sensitivity of neurocognitive testing in patients with SLE. Since the ability to estimate premorbid functioning may be more limited at the higher end of intellectual ability (56), our results may underestimate the actual decline in cognition in this cohort.

A recent study by Omdal et al (51) of anti-NR2 antibodies and cognitive performance in SLE demonstrated a modest but statistically significant correlation between anti-NR2 OD values and performance on the Visual Paired Associates test assessing immediate visual memory, the Grooved Pegboard test in the dominant hand, and BDI scores in patients with SLE. Whereas our results support the finding of association between BDI scores and anti-NR2 antibody levels, we did not see any relationship between the antibody concentrations and cognitive impairment. This could be related to methodologic differences in cognitive testing, antibody measurements, or selection of study participants.

Another possibility is that levels of anti-NR2 may fluctuate over time and may not always be detected in serum at a particular time point. Moreover, increased permeability of the blood–brain barrier, resulting in accelerated access of anti-NR2 antibodies to neuronal tissue, may have a transient but cumulative effect. Recent studies have demonstrated the necessity of disruption of the blood–brain barrier for development of anti-NR2 antibody–related cognitive dysfunction in lupus-prone mice (30). It is possible that a dose-dependent mechanism may take place, in which exposure of neuronal tissue to lower levels of anti-NR2 antibodies may lead to manifestations of depression, whereas exposure to higher levels or more prolonged exposure causes neuronal death, resulting in measurable cognitive impairment. Whether this is the case in patients with lupus remains to be tested in longitudinal studies assessing changes in cognition, levels of anti-NR2 antibodies, and the integrity of the blood–brain barrier over time.

In the regression analysis only education level was inversely associated with current test performance and overall cognitive decline. In contrast to the findings of other studies, we did not observe an association between cognitive dysfunction and aCL levels (7, 8), likely because we excluded patients with a history of cerebral thromboses and might therefore have selected against patients with aCL. Similar to other studies (4, 5, 10), we found no association between cognitive dysfunction and SLE disease activity, depression, steroid therapy, or age.

We demonstrated an increased Cho:Cr ratio on ¹H-MRSI in lupus patients with moderate or severe cognitive dysfunction. An elevated Cho:Cr ratio in the brain was previously observed to inversely correlate with neurocognitive performance in patients with SLE (23), adrenoleukodystrophy, or Williams syndrome (18). It is possible that neuronal loss, increase in inflammatory cellular density, postischemic inflammation, and presence of gliosis in certain brain areas may contribute to the development of cognitive dysfunction. Our results suggest that an increased Cho:Cr ratio in the cortex may represent a subclinical change occurring in the brain of cognitively impaired SLE patients that is not seen on conventional MRI, but can be captured by ¹H-MRSI. Therefore, ¹H-MRSI may play a role in future understanding of molecular mechanisms of impaired cognition in SLE.

Better understanding of the mechanisms of cognitive and psychiatric disorders in SLE will help to identify new therapeutic modalities that can prevent the progression of depression and cognitive decline and improve quality of life in these patients. Further research in this area is rapidly advancing, with many important questions to be addressed in future studies.

Acknowledgements

We thank Ana Gelabert, CRNP, Karen Bove Bettis, MRI technician, Jean Marie Bechtle, CRNP, and Geneva Jacobs, IRTA for their invaluable assistance with the study procedures.