Diffusion tensor imaging in neuropsychiatric systemic lupus erythematosus

Neuropsychiatric systemic lupus erythematosus (NPSLE) is a complex neurological disorder characterized by neuropsychological dysfunction [1]. Patients can present with either focal symptoms, consisting of stroke and/or transient ischemic attacks, or with central nonfocal symptoms of cognitive dysfunction, acute confusional state, seizures, or psychosis. NPSLE is likely initiated by inflammatory, thrombotic, and/or cardioembolic etiologies and further exacerbated by antibody, cytokine, and cytotoxin mediators [2, 3]. Neuropsychiatric manifestations, affecting some 75% of patients diagnosed with systemic lupus erythematosus (SLE), are associated with significantly increased morbidity and mortality [4]. Indeed, previous studies have demonstrated a wide range of brain abnormalities in NPSLE both during and after experiencing acute symptoms [5‐8].

Diffusion Tensor Imaging (DTI) offers increased resolution compared to conventional structural Magnetic Resonance Imaging (sMRI) regarding white matter microstructure by measurement of water diffusion through cellular compartments in vivo [9]. Compared to more isotropic movement of water in gray matter, water diffusion in white matter moves anisotropically, meaning that water diffuses preferentially along the length of the axon compared to perpendicular to the axon. This anisotropic diffusion of water appears to be due to the highly structured axonal membranes and their associated myelin sheaths [10]. By tracking the diffusion of water in the brain, the measure fractional anisotropy (FA) and mean diffusivity (MD) can be derived. Higher FA (and lower MD) suggests greater axonal coherence and myelination [9], increasing in a roughly linear manner that conforms to normal developmental brain processes [11, 12]. Measures of FA are usually considered to be overall measures of axonal integrity, reflecting either increased axonal caliber, increased myelin thickness, increased fiber coherence in a given direction, or some combination of these factors [13]. In contrast, MD, is a measure of the average molecular motion independent of the constraints of tissue boundaries, and is affected by cellular size and degradations in tissue integrity [14].

Relatively few studies have emerged showing water diffusivity changes in NPSLE. The first study compared 9 active to 10 chronic NPSLE patients using magnetization transfer imaging (MTI), a technique designed to compare bound to free protons in biologic tissue [15]. These researchers found that MTI values in active NPSLE patients differed significantly from chronic patients, who were similar to controls, interpreted to reflect the presence of inflammation in the active cohort. A second group used a region of interest (ROI) approach to assess a broad cohort of 34 patients diagnosed with SLE, using diffusion weighted imaging, a technique sensitive to the microscopic motion of water within extracellular space [16]. They found early diffusion changes in the frontal lobe, the genu of the corpus callosum, and the anterior internal capsule, even in the presence of normal magnetic resonance imaging (MRI) findings. Finally, a group compared 8 female NPSLE patients, with new onset of symptoms, to 20 healthy controls using diffusion tensor imaging (DTI). Using an ROI approach, they found that these patients differed from controls in a wide range of normal appearing gray and white matter regions including the insula, thalamus, parietal and frontal white matter, and corpus callosum [17].

Tract Based Spatial Statistics (TBSS) is an automated analysis methodology allowing for projection of individual subject diffusion tensor maps into a common space, allowing for localized statistical testing across the entire brain as opposed to ROI approaches [18]. TBSS is a relatively new analysis technique that has been applied to a broad range of neurological and psychiatric disorders to assess the microstructural integrity of white matter [18‐20]. We sought to use DTI and automated TBSS to assess WM abnormalities in a cohort of patients comprised of acute NPSLE, SLE without NPSLE, and normal controls. We hypothesized greater WM abnormalities in NPSLE compared to either SLE or controls.

The seventeen NPSLE (Mean Age = 38.6 +/- 11.8; 94% Female; WRAT-Reading = 44.3 +/- 7.7) and sixteen SLE with no NPSLE (Mean Age = 37.4 +/- 13.2; 94% Female; WRAT-Reading = 44.4 +/- 8.5) were compared to the twenty healthy controls (Mean Age = 32.5 +/- 10.7; 90% Female; WRAT-Reading = 47.5 +/- 7.3) across major demographic variables. The three groups did not differ significantly from each other in terms of age (F = 1.44, p = .25), gender (Chi-Square = .28, Probability = .87), ethnicity (Fishers Exact p = .73), or premorbid cognitive functioning (F = .96, p = .39).

Next, we sought to compare the groups across DTI measures of FA and MD. SLE patients did not differ from control subjects on either FA or MD measures after controlling for age and sex (df = 32). When NPSLE subjects were compared to control subjects, numerous regions of significant FA and MD differences were observed (Figure 1a/b).

Significant regions of differences between acute NPSLE patients and controls, controlling for age and sex (df = 33), are presented in Table 3.

Finally, when comparing NPSLE to SLE patients, controlling for age and sex (df = 29), numerous regions of significant FA and MD differences were again observed. These are presented in Table 4.

There was one interesting regional similarity wherein NPSLE patients differed in terms of FA measures both from SLE patients and control subjects: the body of the corpus callosum [MNI (x, y, z) = (-8, 5, 25) (-12, 6, 27)]. Similarly NPSLE patients differed from both controls and SLE patients in MD measures obtained within the left arm of the forceps major [MNI (x, y, z) = (-23, -56, 21) (-23, -56, 22)], and the left anterior corona radiata [MNI (x, y, z) = (-24, 15, 19) (-23, 22, 20)].

Finally, we conducted post hoc analyses designed to determine whether the significant MD differences were being driven by the principal eigenvector (AD) or the average of the secondary eigenvectors (RD), which have been ascribed to predominantly axonal versus myelin processes respectively [25]. These post hoc analyses (Additional file 1: Supplemental Table S1) found significant AD and RD differences in all regions identified to differentiate NPSLE patients from SLE and controls including: the body of the corpus callosum [AD NPSLE > Controls MNI (x, y, z) = (-13, 15, 23); RD NPSLE > Controls MNI (x, y, z = (-13, 7, 27), RD NPSLE > SLE MNI (x, y, z) = (-14, 8, 28)], left arm of the forceps major [RD NPSLE > Controls MNI (x, y, z = (-23, -56, 21), RD NPSLE > SLE MNI (x, y, z) = (-23, -56, 22)], and left anterior corona radiata [AD NPSLE > SLE MNI (x, y, z) = (-22, 15, 25), RD NPSLE > Controls MNI (x, y, z = (-24, 15, 19).

The current results were obtained in a relatively young (less than 60 years old), large cohort of both acute NPSLE and SLE patients without NPSLE studied with voxel-based DTI techniques. Interestingly, there were no significant FA or MD differences observed between the sixteen SLE patients without NPSLE and the twenty matched controls, many of the SLE patients of whom had subcortical white matter and periventricular lesions. This could be due to the lower disease burden of this cohort, which did not reach a "threshold" detectable by our imaging and statistical methodology, or due to the non-overlapping lesion burden of the SLE cohort as compared to the NPSLE cohort. It will be important to determine in future studies, with larger samples, whether there are systematic effects of NPSLE that accumulate in the brain in such a way that they affect white matter in a systematic (as opposed to sporadic) manner, as this preliminary study suggests.

In contrast, when comparing the acute NPSLE patients to controls, we found numerous FA and MD changes reflecting diffuse white matter abnormalities. These abnormalities were also present when comparing acute NPSLE to SLE patients, suggesting that either the acute effects of the NPSLE disease, or its treatment, results in white matter changes discernable with conventional MRI techniques. Moreover, in post hoc analyses, we were not able to differentiate AD or RD changes as predominant in driving MD changes. Thus, we are unable to definitively say whether axonal or myelin processes drive the increase in MD; rather, the increased MD in the white matter regions appears to reflect increased overall diffusivity both along and perpendicular to the axon. Again, future longitudinal studies, at various stages in the acute symptomatology will be necessary to determine how disease treatment and symptom resolution are reflected in white matter changes reflected in FA and MD signal changes.

Several possible mechanisms could explain diffusion differences affecting acute NPSLE patients. The first includes immune-mediated vascular or neuronal injury and subsequent neuronal and metabolic dysfunction resulting in edematous processes that increase water content in WM regions of the brain [26]. The second relates to therapy, whereby the introduction of corticosteroids, or other immunosuppression drugs (i.e. cyclophosphamide), and/or disease modifying antirheumatic drugs, can affect water content in the brain parenchyma. The third mechanism may be related to platelet or fibrin macro- or microembolism from Libman-Sacks endocarditis or anticardiolipin antibodies causing multiple areas of macroscopic or microscopic ischemia, infarctions, and microhemorrhages with surrounding edema [2]. Decreased FA, in the same subject, is harder to interpret, being potentially related to demyelination, axonal loss, ischemia, and/or inflammation [27]. However, metabolic changes, observed with proton magnetic resonance spectroscopy, have been found in acute NPSLE, suggesting demyelination, reactive brain inflammation, ischemia or infarction as likely mechanisms leading to decreased FA values [8].

The regions where lower FA and higher MD were found in acute NPSLE are not where patients generally accumulated lesions related to large infarcts or chronic ischemic damage, although diffuse microscopic ischemic injury not detected on MRI is common on histopathologic studies of NPSLE. It is of note that none of these regions were identified on radiological scan to be regions of old infarct in NPSLE (Additional file 2: Supplemental Images S2). Thus, these results likely reflect acute injury effects of NPSLE as opposed to cumulative disease processes. One previous study of eight acute NPSLE patients (6 treated with steroids), found decreased FA in the thalamus, corpus callosum and parietal and frontal WM [17]. A second study of thirty-four patients with SLE, found diffusion abnormalities limited to the corpus callosum when patients were compared to controls [16]. The current results provide new evidence that FA and MD may have diagnostic use in NPSLE by demonstrating, for the first time, regional brain specificity, and by distinguishing NPSLE from SLE patients, further indicating the potential diagnostic specificity of this technique for patients in the acute stage of this difficult disease.

Strengths of the current study include: 1) the relatively large patient cohort, 2) the relatively youth (less than 60 years old) of the patient cohort compared to previous studies, 3) whole brain as compared to ROI analyses, 4) ease and reproducibility of TBSS methodology, and 5) the lack of differences between SLE patients without NPSLE and controls strengthen the specificity of our findings. A limitation of the study is the lack of repeated measures of the acute NPSLE patients as they progress through the acute phase of their disease. This would help to establish whether therapy or disease characteristics predominated over time in determining DTI abnormalities. However, these extremely ill patients are difficult to study repetitively in the acute setting, and these data would have likely increased the differences between acute NPSLE as compared to SLE patients or healthy controls. Future studies will determine if patients can be subcategorized into more tractable groups amenable to sensitive neuroimaging studies.

These results suggest that great care is needed when selecting NPSLE patients to participate in neuroimaging studies. Patients with SLE but no NPSLE appear to have different diffusion characteristics than those with acute symptoms (e.g., seizure, transient ischemic attack, acute confusion). These differences, in turn, appear to affect significantly the diffusion parameters in multiple white matter regions throughout the brain. As water diffusivity is critically important to the interpretation of numerous imaging paradigms, including functional magnetic resonance imaging (fMRI), perfusion weighted imaging (PWI), and magnetic resonance spectroscopy (MRS), the current data would suggest that these patients be treated as a different cohort with respect to imaging analyses.

We found significant FA and MD changes in the white matter of acute patients diagnosed with NPSLE. SLE patients were not significantly different from control subjects on FA or MD measures. These results reflect an automated, replicable, and sensitive assay of white matter abnormalities in the acute phase of neuropsychiatric lupus, requiring less than 10 minutes of imaging on a standard MRI system.