Dysfunctional cortico–basal ganglia–thalamic circuit and altered hippocampal–amygdala activity on cognitive set-shifting in non-neuropsychiatric systemic lupus erythematosus




To explore sequential brain activities throughout cognitive set-shifting, which is critical to understanding the basic pathophysiology of cognitive dysfunction, in patients with new-onset systemic lupus erythematosus (SLE) without neuropsychiatric symptoms.


Fourteen patients with new-onset SLE but without neuropsychiatric symptoms and 14 healthy controls matched for age, sex, education level, and intelligence quotient with the patients performed a cognitive set-shifting task derived from the Wisconsin Card Sorting Test while they were undergoing event-related functional magnetic resonance imaging of the brain. Blood oxygen level–dependent signals were compared between different stages of cognitive set-shifting in the lupus patients and in the healthy subjects.


Lupus patients and healthy subjects demonstrated comparable cognitive function performance, but the cortico–basal ganglia–thalamic–cortical circuit and amygdala–hippocampus coupling, which were involved in response inhibition and active forgetting–learning dynamics, respectively, were demonstrated to be compromised in patients with SLE. Moreover, an increase in contralateral cerebellar–frontal activity was found to compensate for the compromised cortico–basal ganglia–thalamic–cortical circuit in lupus patients in order to maintain their cognitive test performance as comparable to that of the healthy subjects.


Our study revealed significant differences in the sequential brain signals during cognitive set-shifting between patients with SLE without neuropsychiatric symptoms and healthy subjects. The results prompt further in-depth investigation for the functional neural basis of cognitive dysfunction involving the aforementioned neural circuits and compensatory pathways in patients with SLE.

Systemic lupus erythematosus (SLE) is a chronic and multisystemic autoimmune disorder characterized by protean clinical manifestations. Neuropsychiatric SLE (NPSLE) is one of the most common manifestations of lupus, and it has hindered improvements in the survival of these patients over the last 5 decades (1). Of the various symptoms of NPSLE, cognitive impairment, which primarily affects attention, planning, reasoning, working memory, and executive function, has been frequently described (2). It is evident that the cognitive functioning of patients with NPSLE is inferior to that of lupus patients without neuropsychiatric symptoms and that of healthy individuals (3, 4), as they demonstrate poorer performance in a number of domains of cognitive function testing, in particular, attention, information processing, learning, memory, and executive function (3–5).

Real-time functional magnetic resonance imaging (fMRI) of the brain while subjects were performing cognitive function tasks revealed that brain activities, which were manifested by blood oxygen level–dependent (BOLD) signals, were significantly altered in patients with NPSLE as compared with healthy subjects (6, 7). However, since these fMRI studies involved lupus patients with neuropsychiatric manifestations, the findings failed to give insights as to whether the real-time brain activities in lupus patients without clinically overt neuropsychiatric symptoms were altered. Such knowledge is particularly relevant because it is currently believed that lupus patients, even those without clinically obvious neuropsychiatric symptoms, perform poorer cognitively on neuropsychiatric tests than do healthy individuals (4, 8, 9).

Until recently, we used brain fMRI to evaluate the event-related brain activities of lupus patients without neuropsychiatric symptoms (10). Using the Wisconsin Card Sorting Test (WCST) for assessing neural activities corresponding to response selection and feedback evaluation, which evaluate the performance of goal-directed tasks and strategic planning skills of executive function, respectively, lupus patients without neuropsychiatric symptoms demonstrated significantly increased brain activities in regions that enhanced event anticipation, attention, and working memory during RS to compensate for their poor strategic planning as a result of deactivated brain regions involved in motor planning, decision-making, sensory integration, error detection, and conflict processing (10). Very recently, a study performed by another group of investigators using a different functional imaging protocol demonstrated attenuated brain activities in the cerebellum, posterior cingulate, and the adjacent precuneus in the default mode network in lupus patients without neuropsychiatric manifestations (11). While both of these studies demonstrated altered brain signals in certain anatomic regions of the brain in lupus patients without neuropsychiatric symptoms, it is pathologically more relevant to map the functional abnormalities that occur in neural circuits by analyzing sequential activation of the brain while patients are going through various sequential stages of cognitive function testing.

The WCST, which is primarily regarded as a unitary system for evaluating executive function that recruits cognitive processes as a whole during task performance, has been increasingly used to identify the neural basis of different cognitive stages of executive function. Using tailored set-shifting tasks derived from the original WCST (12), the modified version of the WCST was able to demonstrate that the prefrontal cortex, hippocampus, and striatum were sequentially involved across cognitive set-shifting processes in a stage-specific manner (12–15). We therefore undertook this fMRI study using the modified WCST to explore brain activities across all stages of cognitive set-shifting processes in patients with new-onset SLE who did not have clinically overt neuropsychiatric symptoms. Our goal was to map the potential functional abnormalities in neural circuits by analyzing event-related sequential activation of the brain.


Study participants.

Patients with new-onset SLE who satisfied the American College of Rheumatology classification criteria for the disease (16) were recruited from the Lupus Clinic of the National University Hospital, Singapore. SLE disease activity was measured with the Systemic Lupus Erythematosus Disease Activity Index (SLEDAI) (17), and assays for serum levels of complement components 3 and 4 and anti–double-stranded (anti-dsDNA) antibodies were carried out in the hospital laboratory. Healthy subjects recruited from staff working at the National University of Singapore were individually matched with the lupus patients in terms of age, sex, education level, and intelligence quotient (IQ). Each participant's IQ was evaluated with the Wechsler Abbreviated Scale of Intelligence (WASI), which comprises verbal, performance, and the full IQ.

Participants were ineligible for this study if they were left-handed, were not proficient in English, had a history of neurologic disorders, had symptoms of anxiety and/or depression (Hospital Anxiety and Depression Scale score ≥8), (18) had a history of a psychiatric disorder or treatment with psychotropic medications, or were positive for antiphospholipid antibodies. Written informed consent was obtained from the study participants, and the study was approved by the ethics committee at our institution.

Wisconsin Card Sorting Test.

All participants were instructed to perform the computer-based modified WCST. Prior to the test, all participants underwent training to ensure satisfactory comprehension of the sorting criteria. During response selection (RS), 4 cards appeared along the top of a blue screen as a reference and remained unchanged as the test ensued. On each trial, a candidate card that appeared at the center bottom of the blue screen was to be matched according to 1 of the 3 rules (color, number, or shape) with 1 of the reference cards, as randomly generated by the program (Figure 1). Participants were given 4 seconds to respond, after which either a bar would appear under the selected reference card or else the words “too late” would appear on the blue screen, which signified trial termination, and the test would move on to the next trial. After the 4-second selection time, fixation display of a white cross would appear at the center of the blue screen. After another 5 seconds, feedback would appear on the screen as “Right” or “Wrong,” indicating a correct or incorrect card selection, respectively. The feedback stimulus would appear for 500 msec, during which feedback evaluation (FE) was allowed, and then the display would change to fixation until the inception of the next trial.

Figure 1.

Sample sequence of the modified Wisconsin Card Sorting Test, showing the course of cognitive set-shifting. RS = response selection; PF = positive feedback; NF = negative feedback. Color figure can be viewed in the online issue, which is available at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)1529-0131.

The identified rule remained unchanged for a random 3–5 successive correct feedbacks until another rule was randomly generated. Appearance of the first negative feedback (NF) after successive correct feedbacks signaled a “shift” event for the study subject to abandon previously prepotent rule and search for a new rule, subsequent negative feedback (second NF), if applicable, would serve as a “generate” event to identify the correct rule. The first positive feedback (PF) in a series of correct feedbacks was regarded as an “update” event to register the newly confirmed cognitive set in memory, while subsequent positive feedback (second PF) served as a “maintain” event to keep the prevailing rule until a signal was received to shift to another rule.

Each scanning session consisted of 5 runs, and each run lasted for 8 minutes. Figure 1 shows a sample sequence from the modified WCST. During the performance of the WCST and concomitant imaging, the sequence of the “shift,” “generate,” “update,” and “maintain” events was kept constant throughout the course of the WCST, while the predefined time interval of the imaging sequence was entirely independent of the WCST event sequence.

Image acquisition.

Functional imaging was performed with a Siemens Symphony 1.5T MRI scanner. A blipped gradient-echo echo-planar imaging sequence was applied for functional imaging, using the following parameters: repetition time 3,000 msec, flip angle 90°, field of view 192 × 192 mm, and pixel matrix 64 × 64. Each run contained 156 whole-brain acquisitions in a plane parallel to the line between the anterior and posterior commissures on the sagittal scout images, using the following parameters: 32 oblique axial slices of 3 mm thickness, 0.3-mm gap between slices, with descending interleaved slice acquisition. T1-weighted anatomic reference images were acquired using a magnetization-prepared rapid gradient-echo sequence, as follows: pixel matrix 256 × 256, field of view 206 × 206 mm, 80 slices of 2 mm thickness, and in the coronal plane.

Image processing and statistical analysis of the BOLD signal.

Image processing and analyses were performed using the software Brain Voyager QX (version 2.1; Brain Innovations). Preprocessing steps, including slice scan-time correction, motion correction, spatial smoothing (8-mm full-width half-maximum) and linear trend removal, were performed prior to statistical mapping of brain activation during the fMRI paradigm. Functional images were subsequently registered to the T1-weighted images, and the realigned images were transformed into Talairach space.

Beta-score maps were computed in the Talairach space using a random-effects general linear model. This model was constructed with separate regressors relative to a fixation baseline, convolved with a canonical hemodynamic response function peaking at 6 seconds after initiation of card stimulus or feedback, for RS and FE. Jittering of the fixation intervals between the FE and the subsequent RS facilitated event deconvolution (15). A region of interest (ROI) was computed by analyzing the effect contrasts of the 2 categorical task factors, class (positive/negative) and order (first/subsequent), for both the RS and the FE. Significant clusters and peak voxels that survived a false discovery rate (a statistical method for correcting for multiple comparisons) of q < 0.05 were reported (19). The cluster threshold was calculated with the Cluster-Level Statistical Threshold Estimator tool in the Brain Voyager QX software package.

Statistical analysis.

Results are expressed as the mean ± SD except where indicated otherwise. Data from healthy controls and SLE patients were compared by Mann-Whitney U test for continuous variables and by chi-square test for categorical variables. P values (2-tailed) less than 0.05 were considered significant. All statistical analyses were performed with the PASW Statistics software (version 18; SPSS).


Characteristics of the study participants and performance on the WCST.

Fourteen patients with new-onset SLE and 14 healthy controls matched to the patients in terms of age, sex, education level, and IQ were recruited. Upon recruitment, the majority of the SLE patients had active lupus, as indicated by low serum levels of C3 and C4 and a high serum level of anti-dsDNA. Two patients had SLEDAI scores ≤4, indicating low levels of disease activity, while the other 12 patients had SLEDAI scores >4. With regard to performance on the WCST, the SLE patients and controls were comparable in terms of the number of rules identified, the number of errors per rule, and the reaction time. Table 1 summarizes the demographic features, clinical data, and WCST performance data of the study participants.

Table 1. Demographic and clinical characteristics and WCST performance of the study participants*
 SLE patients (n = 14)Healthy controls (n = 14)
  • *

    Except where indicated otherwise, values are the mean ± SD or the mean ± SD (range). WCST = Wisconsin Card Sorting Test; SLE = systemic lupus erythematosus; IQ = intelligence quotient; anti-dsDNA = anti–double-stranded DNA; SLEDAI = Systemic Lupus Erythematosus Disease Activity Index.

Demographic and clinical characteristics  
 Age, years39.38 ± 13.934.07 ± 14.4
 Sex, no. female/male12/212/2
 Education level, years14.43 ± 3.413.43 ± 3.4
 Full IQ96.23 ± 15.6100.43 ± 11.0
 C3, mg/dl (normal 85–185)70.46 ± 25.0 (28–103)
 C4, mg/dl (normal 10–50)15.42 ± 10.9 (3–38)
 Anti-dsDNA, IU (normal <20)132.50 ± 111.3 (1–250)
 SLEDAI9.93 ± 5.7
 Prednisolone, mg/day15.32 ± 18.4
WCST performance  
 No. of rules identified19.54 ± 3.519.50 ± 3.0
 No. of errors per rule3.04 ± 2.52.51 ± 0.7
 Reaction time, msec1,679.01 ± 411.31,808.51 ± 228.8

Signals on fMRI during response selection after negative feedback versus after positive feedback.

Following the rule-shift signal, brain regions that yielded greater activation during RS after the first NF as compared with the RS after the first PF (RS after first NF > RS after first PF) implied involvement in response inhibition of the previously prepotent rule and the need to search for a new rule. In healthy control subjects, significantly activated BOLD signals were found in both middle frontal gyri, the left medial frontal gyrus, the left inferior parietal lobule, the right angular gyrus, the right middle occipital gyrus, the right middle temporal gyrus, the left fusiform gyrus, the right claustrum, the right globus pallidus (GP), and both thalami. The GP and thalami, being the critical components of the cortico–basal ganglia–thalamic–cortical circuit, were involved in response inhibition and change in behavioral set (20, 21).

In patients with SLE, besides brain activation in various regions within the frontal and parietal lobes, the declive of the cerebellar vermis was activated while signals in the right GP and thalami were concomitantly absent. However, activated BOLD signals in the right parahippocampal gyrus and left posterior cingulate were elicited during the condition contrast of the RS after the first NF < RS after the first PF, which was absent in the healthy controls, implying that SLE patients required additional activity in these 2 regions to boost reconfiguration of the response strategy for adapting to a new rule. The brain activation profiles of healthy controls and SLE during RS after the first NF versus after the first PF are shown in Table 2 and Figure 2.

Table 2. Region of interest during RS and FE in healthy controls and patients with SLE*
 Brain regionTalairachHemisphereBA
  • *

    RS = response selection; FE = feedback evaluation; SLE = systemic lupus erythematosus; BA = Brodmann area; NF = negative feedback; PF = positive feedback.

Healthy controls      
 RS after first NF > RS after first PF1 Middle frontal gyrus412530R9
  (response inhibition)2 Middle frontal gyrus−34506L10
 3 Medial frontal gyrus−11945L8
 4 Angular gyrus32−5639R39
 5 Inferior parietal lobule−37−5642L40
 6 Middle occipital gyrus30−83−3R18
 7 Middle temporal gyrus50−320R21
 8 Fusiform gyrus−37−56−9L37
 9 Globus Pallidus14−23R
 10 Thalamus12−79R
 11 Thalamus−10−89L
 12 Claustrum29190R
 RS after first NF < RS after first PFNo significant voxels     
 RS after second NF vs. RS after second PFNo significant voxels     
 Second NF > second PF13 Precentral gyrus−43130L6
  (generation of new rule identity)14 Middle frontal gyrus29493R10
 15 Middle frontal gyrus−37503L10
 16 Medial frontal gyrus−11945L8
 17 Inferior parietal lobule35−5642R7
 18 Inferior parietal lobule−43−4442L40
 Second NF < second PF19 Precentral gyrus32−2957R4
  (maintenance of established rule)20 Anterior cingulate−743−6L32
 21 Posterior cingulate5−5018R30
 22 Cingulate gyrus−10−2039L24
 23 Precuneus14−4451R7
 24 Inferior parietal lobule−49−2627L40
 25 Postcentral gyrus56−2321R40
 26 Middle temporal gyrus−52−2−9L21
 27 Middle temporal gyrus−46−599L39
 28 Hippocampus29−29−9R
 29 Hippocampus−27−29−9L
 30 Amygdala−19−11−13L
 First NF vs. first PFNo significant voxels     
 First NF vs. second NFNo significant voxels     
 First PF vs. second PFNo significant voxels     
Patients with SLE      
 RS after first NF > RS after first PF31 Superior frontal gyrus−315515L10
  (response inhibition)32 Middle frontal gyrus321051R6
 33 Middle frontal gyrus29493R10
 34 Middle frontal gyrus−492527L46
 35 Superior parietal lobule−31−6851L7
 36 Inferior parietal lobule44−5645R40
 37 Inferior parietal lobule−37−5042L40
 38 Inferior parietal lobule−55−4748L40
 39 Thalamus−16−86L
 40 Claustrum29223R
 41 Declive of cerebellar vermis35−62−21R
 42 Declive of cerebellar vermis5−74−18R
 RS after first NF < RS after first PF43 Posterior cingulate−7−4721L30
  (reconfiguration of RS)44 Parahippocampal gyrus23−14−15R28
 RS after second NF vs. RS after second PFNo significant voxels     
 Second NF > second PF45 Superior frontal gyrus294912R10
  (generation of new rule identity)46 Middle frontal gyrus413430R9
 47 Middle frontal gyrus−31586L10
 48 Middle frontal gyrus−462524L46
 49 Cingulate gyrus−42833L32
 50 Superior parietal gyrus−37−5954L7
 51 Inferior parietal gyrus41−5345R40
 52 Claustrum29190R
 Second NF < second PF53 Precentral gyrus50−26R6
  (maintenance of established rule)54 Paracentral lobule−19−3854L5
 55 Anterior cingulate−731−9L32
 56 Hippocampus29−26−12R
 First NF vs. first PFNo significant voxels     
 First NF vs. second NFNo significant voxels     
 First PF vs. second PFNo significant voxels     
Figure 2.

Composite general linear model of the blood oxygen level–dependent activation pattern seen on functional magnetic resonance imaging in healthy control subjects and patients with new-onset systemic lupus erythematosus (SLE) during the condition contrasts of response selection (RS) after the first negative feedback (NF) > RS after the first positive feedback (PF) as well as the second NF > the second PF. Red/yellow areas indicate activation; green/blue areas indicate deactivation. Images are numbered according to the brain region numbering shown in Table 2.

Signals on fMRI during negative feedback versus positive feedback.

Regions that showed greater activation at the second NF as compared with the second PF (second NF > second PF) reflected involvement in generating the identity of a new rule instead of keeping the prevailing one. In healthy controls, both middle frontal gyri, the left medial frontal gyrus, the precentral gyrus, and both inferior parietal lobules were activated during this condition. In lupus patients, the right superior frontal gyrus, the bilateral middle frontal gyri, the right inferior parietal lobule, the left superior parietal lobule, the left cingulate gyrus, and the right claustrum were activated instead.

Brain regions that were activated in the second PF as compared with the second NF (second PF > second NF) signified their contribution to maintain the established rule rather than to generate the identity of a new rule. During this condition, healthy controls had involvement of their right precentral gyrus, right precuneus, left inferior parietal lobule, left middle temporal gyrus and both hippocampi, left amygdala, right posterior cingulate, left anterior cingulate, and left cingulate gyrus (limbic system). In lupus patients, limbic involvement in rule maintenance occurred only in the right hippocampus and left anterior cingulate. Regions of brain activation in healthy controls and SLE for second NF versus second PF are described in Table 2 and Figure 2.


We found that the cortico–basal ganglia–thalamic–cortical circuit, which is involved in response inhibition, was dysfunctional in lupus patients, even if they did not manifest clinically overt neuropsychiatric symptoms. To compensate for the dysfunction of the circuit in the lupus patients, the contralateral cerebellar and frontal areas of the brain were activated. Additionally, lupus patients showed altered activities at the amygdala and hippocampus while they tried to maintain prepotent rules during cognitive set-shifting.

While experiencing the maximum cognitive demand for withholding the previous prepotent rule and the response inhibition throughout the cognitive set-shifting processes, healthy subjects showed increased brain activation in the right middle frontal gyrus, the right GP, and both thalami in a synchronous pattern across the cognitive set-shifting processes (Figure 3A). The dorsolateral prefrontal cortex is involved in working memory, in the switching of cognitive sets, and in the inhibitory control of the prepotent response (20, 22). Since the right middle frontal gyrus belongs to the dorsolateral prefrontal cortex, it implies that it participates in conditions that require functional working memory and switching of cognitive sets. Furthermore, it was previously proposed that response suppression required the functional integrity of the prefrontal cortex, basal ganglia, and thalami. Functionally, these 3 regions form the cortico–basal ganglia–thalamic–cortical circuit that was believed to coordinate the function of the cortical regions involved (21–25).

Figure 3.

Brain signal changes across the cognitive set-shifting process. A, Cortico–basal ganglia loop activity in healthy control subjects. B, Left thalamus activity in healthy controls and patients with new-onset systemic lupus erythematosus (SLE). C, Cerebellar–contralateral frontal activity in SLE patients. The Talairach x, y, and z coordinates for the right declive of the cerebellar vermis are shown in the key to C to distinguish the two graphs. Red/yellow areas indicate activation; green/blue areas indicate deactivation. Numbered images at the bottom correspond to the numbered β score graphs at the top. Values are the mean and images are the composite of 14 healthy control subjects and 14 SLE patients. NF = negative feedback; RS = response selection; PF = positive feedback. Color figure can be viewed in the online issue, which is available at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)1529-0131.

Studies have identified the indirect pathway that connected these regions, showing that the cortex, striatum, external segment of the GP, subthalamic nucleus, internal segment of the GP, thalami, and cortex were involved sequentially in implementing response suppression and in changing the behavioral set (26). However, we did not observe brain activation at the right GP and right thalamus in patients with SLE when the cognitive demand for inhibitory response control was dominant, implying that information flow from the striatum to the thalamus was compromised, which resulted in the deficit of response control in SLE patients under this circumstance. Although the left thalamus was still activated in the lupus patients, the magnitude of the activation signal in the same region was lower than that in the healthy controls (Figure 3B).

Besides its role in the response inhibition circuit, a number of lesion studies have shown that the GP is one of the important neural components that mediate executive cognitive function (27, 28). For example, patients with advanced Parkinson's disease were unable to shift attention after undergoing bilateral pallidotomy (29). Similarly, disturbance of metabolism in the thalamus was demonstrated in children with Sturge-Weber syndrome (30), of which cognitive dysfunction is one of the most common symptoms. Therefore, the absence of activity in the right GP and right thalamus in our lupus patients suggests that they had impaired response inhibition and reduced capability for behavioral change secondary to dysfunction of the cortico–basal ganglia–thalamic–cortical circuit.

Notably, the absence of brain activation at the right GP and right thalamus as mentioned above was coupled with an increase in brain activity at the declive of the right cerebellar vermis in lupus patients. A few anatomic, physiologic, and neuroimaging studies demonstrated that cerebellar damage led to impairment of executive function, spatial cognition, language, and personality change (31). On the other hand, overreliance on cerebellar activities was demonstrated in patients with schizophrenia, alcoholism, and cocaine abuse during evaluation of working memory, verbal working memory, and executive control, respectively (32–35). In our study, the increased activity in the right cerebellum was coupled with activation of the left prefrontal cortex (left superior frontal gyrus and middle frontal gyrus) in lupus patients. This activation pattern is reminiscent of increased brain activation in the left frontal–right cerebellar circuit in alcoholics during verbal memory testing (34), which is consistent with the observations that cerebellar activation often occurs in conjunction with contralateral frontal lobe activation (35). In our study, analysis of brain activities at the declive of the cerebellar vermis and the left middle and superior frontal gyri revealed a rather similar activation pattern during the course of cognitive set-shifting (Figure 3C), further implying the conjugated role of the right cerebellum and left frontal regions during different stages of cognitive set-shifting.

Cerebellar activities, coupled with brain activation in the contralateral frontal areas, have been suggested to compensate for articulatory and inhibitory controls in patients with alcoholism and cocaine abuse, respectively, in order to maintain normal cognitive function (34, 35). Taken together with another recent finding of the cerebellum–basal ganglia interconnections in an anatomic study performed in primates (36), we postulated that in patients with SLE, the increased right cerebellar activities and its coupled left prefrontal activities may serve to compensate for the dysfunction of inhibitory control secondary to the compromised cortico–basal ganglia–thalamic–cortical circuit in order to maintain the performance of the set-shifting tasks. Consistent with the findings of other functional imaging studies, the compensatory activations that aided the dysfunctional cortico–basal ganglia–thalamic circuit found in our study illustrate the adaptability and plasticity of the brain to recruiting neural networks to serve those functions that have been damaged. For example, in order to maintain performance levels, generalized brain activations were demonstrated in subjects with age-related degeneration in motor performance of the hands (37). Relevant to SLE, a few recent studies demonstrated generalized brain activation in lupus patients during the performance of neuropsychological tasks (6, 7, 10, 38). Importantly, compensatory BOLD signal intensity decreased in lupus patients with disease duration of >10 years, suggesting that compensatory activations may decrease when irreversible neuron damage has set in (38).

When being challenged to adapt to a new rule during cognitive-set shifting, the right parahippocampal gyrus and left posterior cingulate were activated in SLE patients in order to reconfigure their response strategy and to adapt to a new rule, whereas among the healthy controls, no significant brain activities in these two regions were observed. The altered activities of the posterior cingulate and parahippocampus were consistent with two recently published studies in SLE patients without NPSLE and in patients with schizophrenia (11, 39), which suggested that there was attenuation of the intrinsic default mode network, episodic memory problem, and related cognitive deficits. The default mode network is regarded as a resting state of brain function and is involved in the planning of future events and ongoing information processing.

The condition contrast second NF > second PF indicated brain regions where involvement of new rule generation dominated the maintenance of an established rule. In healthy controls, both hippocampi were deactivated during this condition contrast, indicating their involvement in maintaining an identified rule. Most brain regions that contributed more to rule generation than to rule maintenance in SLE patients and healthy subjects overlapped, suggesting that brain activities underlying this cognitive process remained intact in lupus patients. Further analysis revealed that activities of both hippocampi initially decreased during the “generate” event, when the identity of the new rule was unknown. Subsequently, during the “update” and “maintain” events, while the new rule was identified and maintained in memory, both hippocampi were activated. This finding is consistent with the putative role of the hippocampus in set-shifting, in which decreased hippocampal activity at the generation event would facilitate active forgetting of the previous rule and a shift to a new cognitive set, while the subsequent increased hippocampal activity at the update and maintenance events would enhance learning of the new rule to guide ensuing trials (15). Interestingly, the left amygdala appeared to follow a synchronous activity pattern with both hippocampi across all stages of set-shifting (Figure 4A).

Figure 4.

Brain signal changes in the hippocampus and amygdala across the cognitive set-shifting process. A, Hippocampal and amygdala activity in healthy control subjects. B, Right hippocampal activity in healthy controls and patients with new-onset systemic lupus erythematosus (SLE). Red/yellow areas indicate activation; green/blue areas indicate deactivation. Numbered images at the bottom correspond to the numbered β score graphs at the top. Values are the mean and images are the composite of 14 healthy control subjects and 14 SLE patients. NF = negative feedback; RS = response selection; PF = positive feedback. Color figure can be viewed in the online issue, which is available at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)1529-0131.

The neural activity underlying FE has been proposed as a result of the response to reward or punishment of the feedback stimulus (14). Hippocampal activities appear to be reinforced and maintained by the amygdala through the generation of rewards (40), suggesting the coupled role of the amygdala and the hippocampus in set-shifting processes observed in our study. However, involvement of the left hippocampus and left amygdala in rule maintenance was absent in our SLE patients, implying a compromise of the active forgetting–learning dynamics in set-shifting despite the fact that the right hippocampus was still required to maintain the established rule in a pattern similar to that in healthy controls (Figure 4B). As in patients with SLE, altered amygdala and hippocampal activities within the default mode network have been demonstrated in patients with depression and Alzheimer's disease (41, 42). These findings may signify a potential common pathologic basis of the disrupted forgetting–learning dynamics that involves the amygdala and hippocampi in patients with SLE, depression, and dementia.

While further mechanistic evaluation is required to explain the BOLD signal changes, especially those involved in hippocampus–amygdala coupling, and the absence of BOLD signals in the hippocampus and amygdala in our patients with SLE, the well-investigated anti–N-methyl-D-aspartate (NMDA) receptor antibodies and anti–ribosomal P antibodies may be potentially contributory. The NMDA receptors are mainly expressed in the neurons of the hippocampi and amygdala. Under physiologic conditions, the neurotransmitter glutamate activates the receptors and mediates learning and memory by manipulating synaptic plasticity (43). In animal models, antagonizing the NR2 subunit of the NMDA receptor (anti-NR2 antibody) led to impaired memory and learning (43). Because anti-NR2 is present in the serum and cerebrospinal fluid of NPSLE patients (43), a potential pathogenetic role of anti-NR2 antibody in cognitive dysfunction has been advocated (44). Indeed, anti-NR2 purified from lupus patients was able to induce cognitive impairment, memory deficits, and hippocampal neurotoxicity in BALB/c mice (45, 46). It was recently shown that anti-NR2 antibody exerted its cytotoxic effects on neurons by increasing intracellular calcium levels through inhibition of the binding capacity of zinc at the zinc-binding site of the NMDA receptor, resulting in reduced cell viability (47).

As for the anti–ribosomal P antibodies, a clinical study in the 1980s demonstrated that the serum titer of anti–ribosomal P antibodies was selectively raised in lupus patients with active psychosis, although not all subsequent studies were able to replicate the associations between the antibodies and psychosis as well as depression in patients with SLE (48). Nevertheless, induction of autoimmune depression in C3H/HeJ mice by intracerebroventricular injection of affinity-purified anti–ribosomal P antibodies revealed that the antibodies bound specifically to the pyramidal cell layer and dentate gyrus of the hippocampus and other areas of the limbic system that could be significantly counteracted by antiidiotypic antibodies to anti–ribosomal P antibodies (49).

This study has a few limitations. First, the sample size is relatively small. However, sample sizes of this magnitude have consistently demonstrated sufficient sensitivity of BOLD signals to achieve statistical significance (14, 15). Second, patients with SLE were given prednisolone treatment at the time of the fMRI scan, and the possible effects of medication on BOLD signals may confound the interpretation of our results. Since the majority of patients underwent fMRI scanning within 2 months of the diagnosis of SLE, with a mean ± SD exposure to initial immunosuppressive therapy of 36.86 ± 35.5 days (range 8–143 days) prior to the scan, the effect of treatment on BOLD signals should be minimal.

In summary, our findings demonstrate that patients with SLE, even without clinically overt neuropsychiatric symptoms, had abnormal sequential brain activities involving the basal ganglia, hippocampi, and amygdala, which indicated a compromised cortico–basal ganglia–thalamic–cortical circuit and hippocampus–amygdala coupling in comparison with healthy subjects. These results translate into a potential compromise in response inhibition and the active forgetting–learning dynamics in lupus patients. Moreover, patients with SLE demonstrated overreliance on the cerebellar–contralateral frontal conjunction activities to compensate for the dysfunction of the compromised cortico–basal ganglia–thalamo–cortical circuit. While scientifically, our results highlight a potential pathologic brain network, which should prompt further investigation into the functional neuropathophysiology of cognitive dysfunction in patients with SLE, the practical clinical implication is that physicians who treat lupus patients may need to be aware of their potential reluctance to change their existing knowledge and adapt to new information, such as in situations where pharmacologic treatment and lifestyle need to be modified for the optimization of disease control and the prevention of organ damage.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Mak had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Ren, Ho, Mak.

Acquisition of data. Ren, Ho, Mak.

Analysis and interpretation of data. Ren, Mak.


We would like to thank Dr. Steven Graham for professional assistance regarding data acquisition, control recruitment, teaching, and initial analyses.