Different patterns of cerebral perfusion in SLE patients with and without neuropsychiatric manifestations

Abstract To investigate brain perfusion patterns in systemic lupus erythematosus (SLE) patients with and without neuropsychiatric systemic lupus erythematosus (NPSLE and non‐NPSLE, respectively) and to identify biomarkers for the diagnosis of NPSLE using noninvasive three‐dimensional (3D) arterial spin labeling (ASL). Thirty‐one NPSLE and 24 non‐NPSLE patients and 32 age‐ and sex‐matched normal controls (NCs) were recruited. Three‐dimensional ASL‐MRI was applied to quantify cerebral perfusion. Whole brain, gray (GM) and white matter (WM), and voxel‐based analysis (VBA) were performed to explore perfusion characteristics. Correlation analysis was performed to find the relationship between the perfusion measures, lesion volumes, and clinical variables. Receiver operating characteristic (ROC) analysis and support vector machine (SVM) classification were applied to differentiate NPSLE patients from non‐NPSLE patients and healthy controls. Compared to NCs, NPSLE patients showed increased cerebral blood flow (CBF) within WM but decreased CBF within GM, while non‐NPSLE patients showed increased CBF within both GM and WM. Compared to non‐NPSLE patients, NPSLE patients showed significantly reduced CBF in the frontal gyrus, cerebellum, and corpus callosum. CBF within several brain regions such as cingulate and corpus callosum showed significant correlations with the Systemic Lupus Erythematosus Disease Activity Index (SLEDAI) and the Systemic Lupus International Collaborating Clinics (SLICC) damage index scores. ROC analysis showed moderate performance in distinguishing NPSLE from non‐NPSLE patients with AUCs > 0.7, while SVM analysis demonstrated that CBF within the corpus callosum achieved an accuracy of 83.6% in distinguishing NPSLE from non‐NPSLE patients. Different brain perfusion patterns were observed between NPSLE and non‐NPSLE patients. CBF measured by noninvasive 3D ASL could be a useful biomarker for the diagnosis and disease monitoring of NPSLE and non‐NPSLE patients.


| INTRODUCTION
Neuropsychiatric systemic lupus erythematosus (NPSLE) is a severe and life-threatening type of SLE. It has a wide range of clinical neurological and psychiatric manifestations (e.g., stroke, epilepsy, cognitive deficits, psychosis, and mood disorders) that decrease quality of life (Ercan et al., 2015;Rhiannon, 2008). Complex underlying pathological mechanisms are reported in SLE, particularly in NPSLE, including neuroinflammation, demyelination, neuron injury, and vasculopathy . It is crucial to differentiate NPSLE from SLE without neurologic or psychiatric symptoms (named non-NPSLE; . However, the discrimination of NPSLE from non-NPSLE is difficult in a clinical setting, requiring stable and reliable biological and radiographical markers (Castellino, Govoni, Giacuzzo, & Trotta, 2008;Cervera et al., 2006;Sarbu, Bargalló, & Cervera, 2015).
Perfusion imaging has been used to assess the microvasculature changes in SLE patients. The cerebral hypoperfusion of frontal, temporal, and partial areas has been observed in NPSLE patients by positron emission tomography (PET) and single-photon emission computed tomography (SPECT) studies, which could provide early biomarkers for SLE (Kodama et al., 1995;Sahebari, Rezaieyazdi, Khodashahi, Abbasi, & Ayatollahi, 2018). Recently, MR perfusion imaging studies with dynamic susceptibility contrast (DSC) reported that the cerebral blood volume (CBV), and cerebral blood flow (CBF) of certain specific brain regions (e.g., posterior cingulate gyrus) were highly related to NPSLE (Gasparovic et al., 2010;Papadaki et al., 2018;Wang et al., 2012;Zimny et al., 2014). Compared to normal controls (NC), both hyperperfusion and hypoperfusion, were illustrated within gray (GM) and white matter (WM) in SLE patients (Gasparovic et al., 2010;Papadaki et al., 2018;Wang et al., 2012;Zimny et al., 2014). Previous studies have demonstrated hypoperfusion within GM and WM in NPSLE compared to non-NPSLE (Papadaki et al., 2018;Wang et al., 2012). Accumulated evidence indicates that abnormal perfusion (CBF or CBV) is not specific to NPSLE because perfusion is also altered in non-NPSLE patients. Discordant results were shown in a few studies (Emmer et al., 2010;Waterloo et al., 2001). Emmer et al. (2010) used ROIs (regions of interest), including subcortical GM and central WM, to investigate the perfusion alterations in NPSLE but did not find any significant difference among groups (NPSLE patients, non-NPSLE patients and NCs).
In clinical practice, the attribution of neuropsychiatric manifestations to SLE mainly depends on expert physician judgment. Neither clinical measurements nor conventional imaging has adequate diagnostic precision. The abovementioned PET and DSC-MRI perfusion imaging techniques have high sensitivity (80-100%) but low specificity (<67%) in diagnosing NPSLE . PET, SPECT, or DSC-MRI are not suitable for every SLE patient due to irradiation or contrast injection.
Arterial spin labeling (ASL) imaging, which is a contrast agent-free perfusion imaging technique, has been validated as being useful for the evaluation of brain perfusion related to various neurological and psychiatric diseases, including Alzheimer's disease, Parkinson's disease, brain tumor, stroke, seizures, and schizophrenia (Haller et al., 2016;Le et al., 2014;Oura et al., 2018;Schneider et al., 2019). In particular, the development of three-dimensional (3D) ASL-MRI with GRASE (a technique combining gradient rapid echo and spin echo) or FSE (fast spin echo) acquisition dramatically improves the perfusion image quality with a high signal-to-noise ratio and spatial-resolution (Alsop et al., 2015;Uetani et al., 2018). However, no previous investigation has focused on the ability of ASL to diagnose and evaluate brain perfusion in NPSLE patients.
Therefore, the purpose of this study was to investigate the underlying perfusion pattern of NPSLE and non-NPSLE using noninvasive 3D ASL-MRI and to evaluate its clinical significance.  (Hanly et al., 2010;Singh et al., 2006). The inclusion criteria were as follows: (a) Table 1 and Table S1). According to the above inclusion criteria, exclusion criteria and completeness of clinical information, 9 NPSLE patients were excluded due to incomplete scanning, history of hemicrania or incomplete disease history; 11 non-NPSLE patients were excluded due to incomplete scanning, history of hemicrania, or unclear NP history; and 4 NCs were excluded due to incomplete scanning or poor image quality. In total, 31 NPSLE, 24 non-NPSLE, and 32 NCs were included in the final analysis.
All subjects were informed of the purpose of the collection of their information, and this study was approved by the Human and Animal Ethics Committee of Peking Union Medical College Hospital.

| Image preprocessing
ASL image preprocessing was carried out using the ASAP (Automatic Software for ASL Processing) toolbox (Mato, García-Polo, O'Daly, Hernández-Tamames, & Zelaya, 2016). First, the T1w structural and ASL images were reoriented to the anterior commissure-posterior commissure (AC-PC) plane, and the image origin was set to the AC. Second, structural images were segmented into GM, WM, and CSF. Thus, the segmentation of GM and WM probability maps were obtained, as well as the brain mask for the brain subtraction of ASL images. Third, the structural images were co-registered to ASL images and downsampled into the ASL space. Fourth, partial volume correction (PVC) was performed using Asllani estimates in ASL image space, which could estimate the CBF for GM and WM independently (Mato et al., 2016). Then, the subtracted and PVC ASL images within the brain mask were normalized into the MNI space (voxel values were interpolated and upsampled into 2 mm × 2 mm × 2 mm) based on the co-registration and segmentation nonlinear transformation information. Finally, the corrected CBF images of GM and WM were smoothed with a 3D Gaussian kernel (full-width at half maximum, FWHM = 6 mm × 6 mm × 6 mm).
For whole brain (WB) level comparisons, the mean CBF within GM, WM, and WB (GM + WM) were calculated. A CBF ratio of GM/WM was obtained by dividing the mean CBF within GM by the mean CBF within WM. For voxel level comparisons, voxel-based analysis (VBA) was performed to find the altered CBF of the local brain regions.

| Statistical analysis
Statistical analyses were carried out using IBM SPSS statistics Version 25 and the DPABI toolbox (a toolbox for Data Processing & Analysis of Brain Imaging. Release V3.1; Yan, Wang, Zuo, & Zang, 2016). For measurement data, parametric ANOVA or nonparametric Kruskal-Wallis H analysis was carried out and followed by corresponding post hoc two-sample tests (Tukey's test for parametric analysis and the Mann-Whitney U test for nonparametric analysis). For categorical data, Fisher's exact test was carried out. For ASL images, voxel-based statistical analysis was performed using ANOVA with age as a covariate, followed by a pairwise two-sample t-test with 5,000 permutations to define the t-value threshold. An uncorrected p < .01 and cluster size >30 voxels were considered to be a significant difference.  3 | RESULTS

| WM lesions
Twenty-three NPSLE and six non-NPSLE patients showed WM lesions on FLAIR images. There was a significant difference in the incidence of WM lesions between NPSLE and non-NPSLE patients (p < .001, refer to Table S1), but there was no significant difference in WM lesion volumes (p = .469). As the WM lesion distribution map ( Figure S1) shows, the WM lesions were mainly distributed within the periventricular and deep WM.

| Perfusion abnormalities
The WB level results are listed in

| Relationship between perfusion measures, clinical variables, and WM lesion volume
Abnormal CBF within several brain regions (e.g., cingulate, corpus callosum, and bilateral frontal gyrus) showed weak correlations with clinical information including SLEDAI and SLICC scores and showed no significant correlation with disease duration, total cumulative duration of steroid use, or WM lesion volume when age was used as a covariate. The details of this analysis are shown in Table 2.
We found no significant association between CBF measures and other clinical information (e.g., renal disorder) or laboratory parameters (e.g., anti-dsDNA) related to the manifestation of SLE, vascular risk factors and immunological indicators.

| ROC analysis to differentiate NPSLE from non-NPSLE
Differentiating non-NPSLE and NPSLE is a challenge in clinical practice. In this work, the ratio of GM/WM was used to distinguish non-  Table 3 and Figure S2. The GM/WM ratio has a moderate ability to distinguish NPSLE from non-NPSLE, with an AUC of 0.737 and a cut-off value of 1.96 (sensitivity = 83.3%, specificity = 64.5%, and Youden index = 0.478).
We also investigated the ability of regional brain measurements to differentiate NPSLE from non-NPSLE. ROC analysis showed that all the statistically significant brain regions, such as the right cerebellum posterior lobe, bilateral superior and middle frontal gyrus and corpus callosum, showed a moderate performance (AUC > 0.7) for differentiating non-NPSLE from NPSLE. The details of this analysis are shown in Table 3 and Figure 2.

| Classification by SVM
The classification results are listed in Our data demonstrated an increase in GM perfusion in non-NPSLE patients and a decrease in NPSLE patients, implying the dynamic progression of perfusion alterations. Cohen et al. (2017) proposed that vascular injury may occur in all patients with SLE, but primary NP manifestations occurred only after exceeding a certain threshold of ischemic injury, which is consistent with our findings.
CBF significantly increased in GM in the temporal gyrus, cingulate gyrus, and cerebellum regions and in widespread WM in non-NPSLE patients compared with NCs. We hypothesized that this result may be caused by a compensatory mechanism in response to ischemia or injury, which was consistent with previous fMRI studies showing hyperactivities and hyperconnectivity within certain specific networks (e.g., the sensorimotor network) (Mikdashi, 2016;Niu et al., 2018;Nystedt et al., 2018;Papadaki et al., 2014;Wu et al., 2018;Zhang et al., 2017). However, in NPSLE patients, along with widespread vasculopathy and thrombosis, the perfusion of these brain regions was significantly decreased, which means a large reduction in compensatory capacity when the damage exceeded the threshold. In this work, 18 NPSLE patients showed obvious cerebral atrophy (only four non-NPSLE patients showed cerebral atrophy), which could also account for GM hypoperfusion. Non-NPSLE and NPSLE might share a similar vessel compensatory mechanism within WM, which presented in higher perfusion in distributed areas within WM both in non-NPSLE and NPSLE patients. This finding is consistent with a previous DSC study showing elevated CBF and CBV within normal-appearing WM in SLE patients compared to that in NCs (Gasparovic et al., 2010). The hyperpefusion within WM might also be attributed to potential vascular dysfunction caused by evaluated blood pressure in SLE patients (Gasparovic, Qualls, Greene, Sibbitt, & Roldan, 2012).
Histopathological studies revealed that nonspecific focal vasculopathy appeared in non-NPSLE, while diffuse vasculopathy and microthrombi were commonly found in NPSLE, which is the possible pathological basis for different patterns of perfusion abnormalities (Cohen et al., 2017;Sarbu, Sarbu, Bargallo, & Cervera, 2018;Sibbitt et al., 2010). Because of its vasculopathy nature, NPSLE is considered a subtype of inflammatory and immunologically mediated small vessel diseases. The corpus callosum is a key WM structure connecting hemisphere projection fibers. This territory is prone to be affected by  (Costallat et al., 2018;Lee et al., 2015;Wang et al., 2012). This perfusion pattern in the corpus callosum needs to be validated in a large sample with a longitudinal design and F I G U R E 2 ROC analysis of CBF extracted from statistically significant brain regions to differentiate NPSLE and non-NPSLE. The red point indicates the cut-off value has the potential to be an imaging biomarker for defining SLE patients vulnerable to progress to NPSLE.
Some previous works showed conflicting results demonstrating the hypoperfusion of GM and WM in both NPSLE and non-NPSLE patients (Papadaki et al., 2018;Wang et al., 2012). This discordant information may be due to variations in the sample sizes, MR scanners or analysis methods of these studies (Emmer et al., 2010;Wang et al., 2012;Zimny et al., 2014). For example, previous works performed CBF normalization with the assumption of stable perfusion within some brain regions (such as the cerebellum). However, we found that the perfusion within the cerebellum showed a significant difference between NPSLE and non-NPSLE patients, indicating that CBF in the cerebellum was altered by SLE disease. Therefore, we did not normalize cerebral CBF with CBF in the cerebellum. The alterations in cerebellum CBF and their relationship with cerebral perfusion should be studied in future studies.

CONFLICT OF INTEREST
None declared.

AUTHOR CONTRIBUTIONS
The project was designed, conceived and planned by Y.L., X.Z., and H.L. The data acquisition was performed by L.S. The data were processed and analyzed by Z.Z. The article was written and edited by Z.Z. and L.S. The article was revised by Y.D., J.H., X.Q., and S.H. All authors approved the final version of the manuscript.

PATIENT CONSENT FOR PUBLICATION
Obtained.

DATA AVAILABILITY STATEMENT
The data that support the findings in this study are available from the corresponding author upon request.