Abnormal cerebral hemodynamics and blood-brain barrier permeability detected with perfusion MRI in systemic lupus erythematosus patients

Objective


Introduction
Neuropsychiatric (NP) involvement includes several manifestations of the nervous system in systemic lupus erythematosus (SLE), estimated to occur in more than half of all SLE patients, with a prevalence varying between 12 and 95 % (Unterman et al., 2011). Neuropsychiatric SLE (NPSLE) patients often require more intense treatment with glucocorticoids and cytostatic drugs, and report increased fatigue and reduced quality of life compared to SLE patients without attributable NP events (non-NPSLE) (Hanly et al., 2004). The most prevalent central nervous system (CNS) manifestations associated with NPSLE are diffuse, including headaches, mood disorders and cognitive dysfunction, each having an estimated prevalence of at least 20 % (Unterman et al., 2011). Other common CNS manifestations have more focal character, such as cerebrovascular disease and seizures (Unterman et al., 2011). Postulated pathophysiologies contributing to NPSLE consist of mainly ischemic or autoimmune processes, including systemic and intrathecal autoantibody production, focal or general vascular pathology, as well as disruption of the blood-brain barrier (BBB) (Nikolopoulos et al., 2019;Stock et al., 2017;Ota et al., 2022). The imaging characteristics for each of these mechanisms differ, although features may overlap (Ota et al., 2022).
In light of these previous MRI perfusion and permeability findings (Borrelli et al., 2003;Papadaki et al., 2018;Papadaki et al., 2019;Zhuo et al., 2020;Jia et al., 2019;Gasparovic et al., 2010;Wang et al., 2012;Emmer et al., 2010;Zimny et al., 2014;Chi et al., 2019;Kamintsky et al., 2020), the present study was conducted to further evaluate perfusionbased measures with DSC-MRI in an SLE cohort, while also investigating potential discrepancies when using different attribution models for NPSLE. Our primary aim was to investigate whether SLE patients exhibit altered patterns of cerebral perfusion compared to HC, and NPSLE compared to non-NPSLE patients, in specific predefined anatomical brain regions. We also aimed at investigating whether BBB dysfunction and altered vascular properties were present in WMH compared to NAWM in SLE patients. We applied different perfusion and permeability metrics including cerebral blood flow (CBF), cerebral blood volume (CBV), mean transit time (MTT) and the BBB leakage parameter K 2 .

Participants
This cross-sectional study included 64 female SLE patients, consecutively recruited independently of NP events or disease activity from the tertiary rheumatology outpatient clinic at Skåne University Hospital, Lund, Sweden. Each patient fulfilled at least four of the Systemic Lupus International Collaborating Clinics (SLICC) classification criteria for SLE, including at least one clinical and one immunological (Petri et al., 2012). In order to reduce study group heterogeneity, inclusion criteria were female, age between 18 and 55 years (reducing the likelihood of age-related MRI abnormalities) and right-handedness (reducing potential differences in hemisphere dominance between subjects). Exclusion criteria were any contraindication to performing MRI such as claustrophobia, pacemaker, metal implant, pregnancy or contrast agent allergy. SLE disease activity was assessed using the Systemic Lupus Erythematosus Disease Activity Index 2000 (SLEDAI-2 K) (Gladman et al., 2002 Feb) and irreversible organ damage using the SLICC/American College of Rheumatology (ACR) Damage Index (SDI) (Gladman et al., 1996).
A rheumatologist and a neurologist clinically evaluated all SLE patients regarding ongoing or history of NP events attributed to SLE, which were then defined according to the 19 case definitions published by the ACR (The American College of Rheumatology, 1999), herein stated as the "NPACR model". A diagnosis of NPSLE required consensus between the two clinicians. The rheumatologist further classified the NPACR patients according to the attribution models SLICC A and SLICC B (Hanly et al., 2007;Hanly et al., 2008), including patients with attributable NP events within 6 months and 10 years prior to SLE diagnosis, respectively. The SLICC models excluded NP events based on contributing non-SLE factors in the NPACR model, while also excluding minor NP events commonly found among the general population: headache, anxiety, mild depression, mild cognitive dysfunction and polyneuropathy without electroneurography confirmation (Ainiala et al., 2001). Each patient had to fulfill the requirements of at least one NP manifestation according to the different attribution models to be included in the respective NPSLE group. Conversely, patients without attributable NP events were referred to as non-NPSLE.
For the comparison group, perfusion MRI data from 20 HC (10 female, 10 male) collected from a previous study (Knutsson et al., 2014) were reanalyzed and compared with the data obtained from the SLE patients. As part of that study, a neurologist evaluated the HC subjects and their medical histories were reviewed, to exclude cognitive deficits and any medications that could potentially have altered the cerebral perfusion. Additionally, conventional MRI scans were performed to rule out visible pathology in the HC group.
Written informed consent was obtained from all subjects and the project was approved by the local ethics committee (the SLE study has ethical permits 2012/254, 2012/677 and 2014/778, and the HC study has permit 2008/703).
The HC cohort was examined on a 3 T MRI scanner (Achieva, Philips Medical Systems, Best, The Netherlands). The structural series included: axial T2-weighted FLAIR (axial plane resolution 0.43×0.43 mm 2 , slice thickness 5 mm, TE = 140 ms, TR = 12000 ms and TI = 2850 ms), and a T1-weighted Look-Locker sequence before and after contrast agent administration (axial plane resolution 1.7×1.7 mm 2 , slice thickness 6 mm, TE = 3.5 ms and TR = 8.0 ms). PWI was performed using a DSC-MRI gradient-echo EPI sequence with the following sequence parameters: axial plane resolution 1.72×1.72 mm 2 , slice thickness 5 mm, FOV 220×220 mm 2 , image matrix 128×128, TE = 29 ms, TR = 1243 ms, 20 slices, flip angle 60 • and an acquisition time of 1:30 min. A gadolinium contrast agent of 0.1 mmol/kg body weight (Dotarem®, Gothia Medical, Guerbet, Paris, France) was injected during the DSC-MRI imaging. In addition, a prebolus of 0.02 mmol/kg body weight gadolinium contrast agent (Dotarem®, Gothia Medical, Guerbet, France) preceded the DSC-MRI measurement (Knutsson et al., 2014) and was not further considered in the present study. Both bolus doses were administered at an injection rate of 5 ml/s, followed by a 20 ml saline flush at the same rate.

MRI perfusion analysis
The DSC-MRI images were processed using the software nordicICE (NordicNeuroLab, Bergen, Norway) and the perfusion processing was carried out as follows: The signal in each voxel was converted to transverse relaxation-rate change ΔR2*, and a linear relationship between ΔR2*(t) and the contrast agent concentration was assumed (Fisel et al., 1991). The CBV maps were calculated by dividing the area of the concentration time curve in each voxel with the area of the arterial input function (AIF). The CBF and MTT maps were calculated using the central volume theorem (CBF = CBV/MTT) and Zierler's area-to-height relationship after deconvolution of tissue concentration curves with the AIF using a singular value decomposition (SVD) deconvolution algorithm (Wu et al., 2003). The AIF was defined by taking the average of semiautomatically selected voxels within the branches of the middle cerebral artery (MCA) around the Sylvian fissure. The contrast agent leakage was corrected for using a linear fitting algorithm and K 2 was calculated using an estimation of the extravasation of the contrast agent by determining deviation from a 'non-leaky' reference tissue response curve (Boxerman et al., 2006).
Maps of CBF, CBV and MTT were analyzed using the in-house MATLAB R2016b (RRID:SCR_001622) software Eval GUI (developed by Markus Nilsson at Lund University, Sweden). Based on earlier findings and previously investigated areas in SLE (Borrelli et al., 2003;Papadaki et al., 2018;Papadaki et al., 2019;Zhuo et al., 2020;Jia et al., 2019;Gasparovic et al., 2010;Wang et al., 2012;Emmer et al., 2010;Zimny et al., 2014;Chi et al., 2019;Kamintsky et al., 2020), 26 regions of interest (ROIs) were placed in pons, insulae, hypothalami, nuclei caudatus, putamina, posterior thalami, corpus callosum, cingulate gyrus, frontal and parietal gray matter (GM), and frontal, parietal, temporal and occipital white matter (WM). The manually drawn ROIs were 40 pixels in size (volume 0.59 ml), except for the insula ROIs which were 20 pixels in size (volume 0.30 ml). The ROIs were manually placed by a junior colleague (TS) with guidance from a senior neuroradiologist with 30 years of experience (PCS). The FLAIR or T1-weighted images were used as anatomical reference and any signal alterations such as WMHs were avoided. Corresponding color-coded CBF maps were used to avoid ROI placement directly in the intravascular space (example shown in Fig. 1). The researchers were blinded to clinical patient information during these steps.
The values for CBF, CBV and MTT obtained from the ROIs were normalized for each subject, by dividing with the respective value of the middle cerebellar peduncle. This generated relative perfusion-based values (denoted rCBF, rCBV and rMTT), which were summarized as means with standard deviation for each ROI and compared between SLE and HC, as well as between NPSLE, non-NPSLE and HC (using the SLICC A, SLICC B and NPACR model). The differences in rCBF, rCBV and rMTT were then presented as the mean increase or decrease in percent in SLE patients relative to the values obtained in the HC group (% SLEvsHC = ((SLE/HC) − 1) × 100). The same calculations were used for the NPSLE patients relative to the non-NPSLE patients (% NPSLEvsnon-NPSLE = ((NPSLE/non-NPSLE) − 1) × 100), the NPSLE patients relative to the HC group (% NPSLEvsHC = ((NPSLE/HC) − 1) × 100) and the non-NPSLE patients relative to the HC group (% non-NPSLEvsHC = ((non-NPSLE/HC) − 1) × 100).

White matter hyperintensities
The WMH regions for each SLE patient were manually delineated in the axial FLAIR images, using the Eval GUI software tool, by two trained researchers (TS and TR) and evaluated by an experienced neuroradiologist (PCS). WMHs smaller than 0.015 ml were removed to reduce the impact of partial volume effects (PVEs) and false positives. To ensure that the segmented WMH maps overlapped correctly with the PWI images, the following preprocessing steps were applied: The FLAIR images were co-registered to PWI-space using the Linear Image Registration Tool (FLIRT) (Jenkinson and Smith, 2001;Jenkinson et al., 2002) in the FMRIB Software Library (FSL) software 5.0.10 (RRID:SCR_002823) (Smith, 2002) using sinc interpolation. The WMH maps were then transferred to PWI-space using the transformation matrices obtained from the above-mentioned FLIRT co-registration, using Near-estNeighbor interpolation.
ROIs were manually placed in WMHs larger than or equal to 3 pixels in PWI-space (volume ≥0.044 ml) (example shown in Fig. 2). Two ROIs of 40 pixels each were placed in NAWM of the centra semiovale (volume 1.18 ml), as this area was located in the vicinity of the majority of the WMHs. Relative perfusion-based values (rCBF, rCBV and rMTT), normalized to the respective value of the middle cerebellar peduncle, and absolute values for K 2 were calculated for the WMHs and NAWM in the SLE cohort. The differences in rCBF, rCBV, rMTT and K 2 were then presented as the mean increase or decrease in percent in the WMHs relative to the values obtained in NAWM (% WMHvsNAWM = ((WMH/ NAWM) − 1) × 100). Descriptive values regarding WMH size (ml/ WMH), WMH volume (ml/patient) and WMH burden (total number of measurable WMHs for each subject) were calculated using the ROIs in PWI-space. During all the above steps clinical patient information was blinded.

Statistical analysis
All statistical analyses were carried out using the Python package SciPy 1.9.1 (RRID:SCR_008058) (Virtanen et al., 2020 Mar). Each perfusion comparison was performed separately, as described below, and the same approach was used to assess differences in age between the groups. A Shapiro-Wilk test (p < 0.05) was used to check for normal distribution. Differences between SLE and HC were analyzed using a two-sided t-test or a Mann-Whitney U test. Comparisons between NPSLE, non-NPSLE and HC applied the SLICC A, SLICC B and NPACR models separately, using a one-way ANOVA with post-hoc Tukey HSD test or Kruskal-Wallis test with pairwise Mann-Whitney U tests. Bonferroni correction was applied to each of the following perfusion analyses separately to account for multiple comparisons, setting the statistical significance to p < 0.00064 (calculated 0.05/(26×3), as each group comparison had 26 ROIs with 3 perfusion parameters (rCBF, rCBV, rMTT) each): SLE vs HC, SLICC A vs non-SLICC A vs HC, SLICC B vs non-SLICC B vs HC, NPACR vs non-NPACR vs HC. An additional analysis with only females in the HC cohort was also performed, and results are presented in the supplementary material.
WMHs in SLE were compared to NAWM with a two-sided t-test or a Mann-Whitney U test for rCBF, rCBV, rMTT and K 2 . These four perfusion-based parameters were analyzed (in terms of WMH vs NAWM) for each of the following groups separately: SLE, SLICC A, SLICC B, NPACR, non-SLICC A, non-SLICC B and non-NPACR. The NPSLE and non-NPSLE WMH groups were also compared using the same statistical tests regarding age, WMH burden, WMH size and WMH volume (using the SLICC A, SLICC B and NPACR models separately).

Clinical characteristics
An overview of the clinical characteristics of the 64 SLE patients is presented in Table 1. The median SLEDAI-2 K score was 2 (range 0-18), with 16 patients (25 %) having SLEDAI-2 K ≥4, and the median SDI score was 0 (range 0-5), with 20 patients (31 %) having SDI ≥1. The majority of the patients were treated with immunosuppressive medications such as hydroxychloroquine (80 %), prednisolone (78 %) or other disease-modifying antirheumatic drugs (DMARDs) (59 %). Due to severe hydrocephalus in one HC, the ROIs for perfusion measurements could not be placed correctly in this subject. Therefore, this HC had to be excluded, leaving 19 HCs (10 female, median age 31 years, range 25-74 years) to be included in the final analysis. There were no significant age differences between SLE and HC (p = 0.23) or between NPSLE, non-NPSLE and HC, regardless of attribution model (SLICC A: p = 0.18, SLICC B: p = 0.46 and NPACR: p = 0.35).  The frequency of NPSLE was 20%, 30%, and 59% when applying the SLICC A, SLICC B and NPACR models, respectively ( Table 2). The most frequent NP manifestations according to the NPACR model were cognitive dysfunction (36 %), headache (31 %), autonomic neuropathy (16 %), depression (14 %) and anxiety (14 %). Autonomic and cranial neuropathy were most frequent when further applying the SLICC A and SLICC B models. Assessment of disease duration, SLEDAI-2 K and SDI in the NPSLE and non-NPSLE patients revealed no significant differences between the groups (data not shown).

Regional perfusion differences
The mean differences between the SLE patients and the HC group are presented in Fig. 3 and Supplementary Table 1. On average, 2.5 % increase in rCBF (range -23.2 -28.3 %), 1.0 % decrease in rCBV (range − 39.8 -61.5 %) and 7.3 % decrease in rMTT (range − 28.3 -19.1 %) was found. After Bonferroni correction, the SLE cohort displayed significantly decreased rCBF in the pons, decreased rCBV in the bilateral putamina and posterior thalami, and decreased rMTT in the right anterior insula and bilaterally in the hypothalami, putamina and right posterior thalamus, compared to HC. Increased rCBF was present in the posterior corpus callosum and increased rCBV in the anterior corpus callosum, compared to HC. To investigate possible sex-related differences, we repeated the above-mentioned analysis with only female HC. Bilateral putamina in rCBV and the right anterior insula in rMTT showed a significant decrease in the regional perfusion after correction for multiple comparisons (Supplementary Table 2).
Comparisons between NPSLE and non-NPSLE patients revealed a general trend towards higher rCBF (mean increase 13.3 %) and rCBV (mean increase 13.9 %) using the SLICC A model, and a trend towards lower rMTT using the NPACR model (mean decrease 7.3 %) (Supplementary Table 3). No statistically significant perfusion differences were found in any of the ROIs, regardless of attribution model, after correction for multiple comparisons. When comparing NPSLE and non-NPSLE patients to HC, significant perfusion differences were found in the same areas for both NPSLE and non-NPSLE, using the SLICC A and SLICC B models. After correction for multiple comparisons, decreased rCBF was found in the pons, decreased rCBV in the right putamen and right posterior thalamus, and decreased rMTT in the right putamen and bilaterally in the hypothalami, compared to HC. When using the NPACR model significant results were found in the same areas, and, in addition, a significantly decreased rMTT in the right posterior thalamus and right anterior insula, compared to HC. The results are presented in Supplementary Table 4.

Perfusion and contrast leakage differences in WMHs compared to NAWM
Measurable WMHs were present in 31 SLE patients. Of these, 9, 12, and 16 patients were NPSLE patients when using the SLICC A, SLICC B, and NPACR attribution models, respectively. In general, the SLE patients had a WMH burden of 2 (range 1-25), WMH size of 0.074 ml/WMH (range 0.044-0.222 ml/WMH) and WMH volume of 0.192 ml/patient (range 0.044-3.062 ml/patient). Descriptive data regarding the WMHs are presented in Table 3. No significant differences in age or the WMH metrics were found between the NPSLE and non-NPSLE patients.
Significantly higher rCBF, rCBV, rMTT and K 2 were present in WMHs   compared to NAWM in the SLE patients. The same significant pattern was found in all non-NPSLE groups regardless of the attribution model. NPSLE patients in the SLICC A and SLICC B models showed a significant increase in regional perfusion in WMHs compared to NAWM regarding rCBF, and in the NPACR model regarding rCBV. Higher K 2 was also found in all groups in WMHs compared to NAWM, and significantly among the non-NPSLE patients. The results in the SLE, NPSLE and non-NPSLE patients are presented in Table 3.

Discussion
This explorative study, using DSC-MRI, demonstrated abnormal cerebral perfusion in several areas of the brain in SLE patients compared to HC. In particular, we found a decrease in rMTT to be the most prevalent finding and it differed significantly in the right anterior insula, hypothalami, putamina as well as the right posterior thalamus. Previous findings of hypoperfusion in the thalami (Papadaki et al., 2018;Oda et al., 2005), insulae (Zhuo et al., 2020;Giovacchini et al., 2010) and basal ganglia (Papadaki et al., 2018;Papadaki et al., 2019;Zhang et al., 2005 Dec 15;Shen et al., 1999) suggest, in agreement with our results, that these areas could be affected in SLE patients. In addition to perfusion alterations, as measured with DSC-MRI, single-photon emission computed tomography (SPECT) and ASL, these areas further show potential involvement in SLE patients with altered microstructure on diffusion tensor imaging (DTI) as well as GM atrophy on conventional MRI (Hughes et al., 2007;Cagnoli et al., 2012). In contrast to a previous ASL study (Zhuo et al., 2020), we found decreased rCBF in the pons in SLE patients, which could reflect or predispose the presence of T2 hyperintensity lesions in the region (Abreu et al., 2005), and increased rCBF and rCBV in the corpus callosum. Our findings of decreased rCBV in the putamina and right posterior thalamus also contradicts some previous reports suggesting that SLE patients could have regional and even overall hyperperfusion compared to HC (Gasparovic et al., 2010;Wang et al., 2012). Together, this emphasizes that perfusion differences found with DSC-MRI and other imaging techniques in SLE patients are still inconclusively explored. However, our findings are expected to further the general knowledge in the field rather than contributing to individualized clinical practice at this point. The results confirm that SLE patients differ from HC also regarding perfusion metrics, and contribute to a better understanding of NPSLE pathogenetic mechanisms and how they are reflected in the CNS. Perfusion MRI measurements may contribute to individual assessments of SLE patients in the NPSLE context.
The MTT value can be described using the central volume theorem, stating that the MTT is given by the ratio CBV/CBF. Therefore, either an abnormal relative decrease in rCBV or an increase in rCBF (or both these effects) would result in a decreased rMTT. Interestingly, we found several regions that followed this pattern leading to a decreased rMTT, Fig. 3. Visualization of the perfusion differences in percentage between SLE and HC for rCBF (top), rCBV (middle) and rMTT (bottom). Only ROIs with a significant perfusion difference i.e., increases of more than 15 % and a decrease of more than − 15 %, are shown. An overview of the exact percentages for each ROI can be found in Supplementary Table 1. ROI = Region of interest, SLE = Systemic lupus erythematosus, HC = Healthy Controls, rCBF = Relative cerebral blood flow, rCBV = Relative cerebral blood volume, rMTT = Mean transit time. some of which were significant after correction for multiple comparisons, as presented in Supplementary Table 1. While a reduction in rMTT is not often reported in the literature, it has been hypothesized that a shortened rMTT in acute stroke could be due to errors attributed to the SVD deconvolution when there is poor collateral circulation and restricted diffusion or blood flow (Doucet et al., 2016;Murayama et al., 2017;Dababneh et al., 2011). However, the importance and interpretation of these findings in the context of SLE remains to be further evaluated, as to the best of our knowledge, no previous DSC-MRI research has found any rMTT differences between SLE patients and HC (Papadaki et al., 2018;Wang et al., 2012;Emmer et al., 2010).
Correct diagnosis of NPSLE is still a clinical challenge, and difficulties may arise when determining whether previous and ongoing NP symptoms are primarily caused by SLE, secondary to unrelated causes (such as infections or medications) or due to comorbidities (Bortoluzzi et al., 2018). Thus, different more or less stringent attribution models have been developed, and to our knowledge, this is the first work to use the more stringent SLICC A and SLICC B models alongside the NPACR model to evaluate cerebral perfusion in NPSLE and non-NPSLE patients. We found that none of the investigated metrics revealed any significant differences, in concordance with some previous DSC-MRI studies that also were unable to distinguish the patients (Papadaki et al., 2019;Wang et al., 2012;Emmer et al., 2010;Zimny et al., 2014). This could be an expression of the variability in perfusion-based values among the patient groups, and the heterogeneity of the clinical phenotypes, including the possibility of subclinical NP involvement among patients labeled as non-NPSLE within the current definitions (Wang et al., 2012;Bortoluzzi et al., 2018 Mar). Additionally, as a consequence of the consecutive recruitment of patients, NP manifestations in the NPSLE patient groups were in general not acute onset nor necessarily active despite ongoing symptoms. We emphasize that other study designs are needed to specifically evaluate SLE patients with acute onset or ongoing NP involvement related to cerebral perfusion.
Interestingly, we demonstrated significant alterations in the perfusion-based metrics for both NPSLE patients and non-NPSLE patients when compared with HC. These results are in agreement with the findings in the general SLE group, and suggest that SLE patients, independently of NP involvement, might present with altered cerebral perfusion compared to HC. A decrease in rCBV has previously been reported in the thalami and lenticular nuclei in NPSLE patients compared to HC (Papadaki et al., 2018), which is in line with our findings of reduced rCBV in the same areas, and particularly in the' right putamen and right posterior thalamus after correcting for multiple comparisons. Even in non-NPSLE patients, these regions, which are part of the corticobasal ganglia-thalamic-cortical circuit, have shown decreased signaling on blood oxygenation level-dependent (BOLD) functional MRI (fMRI), which has been hypothesized to cause impairment of both response control and behavioral change (Ren et al., 2012). Furthermore, a previous study, using quantitative susceptibility mapping (QSM), demonstrated iron accumulation in the putamina in NPSLE patients compared to both non-NPSLE patients and HC (Ogasawara et al., 2016). These findings, potentially caused by underlying neuroinflammation, are suggested to further underestimate the perfusion values in this area (Yamada et al., 2002). Although a later study was unable to replicate these changes (Bulk et al., 2021), a possible iron-induced susceptibility effect should, still, be considered. While no obvious calcifications were seen in the basal ganglia in our SLE cohort, the inconclusive results from previous studies concerning iron accumulation (Ogasawara et al., 2016;Bulk et al., 2021) warrant the inclusion of QSM in future SLE studies.
WMHs seen on T2-weighted FLAIR images have been proposed to be caused by a variety of different factors in SLE patients (Ota et al., 2022;Luyendijk et al., 2011;Shaharir et al., 2018), and our secondary aim included investigating the vascular properties of these hyperintensities with DSC-MRI. We found significantly increased values for all perfusionbased measurements (rCBF, rCBV and rMTT) in WMHs compared to NAWM in the SLE cohort, indicating hyperperfusion and a relatively higher increase in rCBV compared to rCBF, as required for an observation of increased rMTT. These results are in agreement with previous studies in healthy individuals showing a higher rMTT in WMHs compared to NAWM (Dewey et al., 2021;Marstrand et al., 2002), but are in contrast to reports of reduced rCBF (Marstrand et al., 2002;Brickman et al., 2009;Promjunyakul et al., 2015) and rCBV (Sachdev et al., 2004) in the tissue surrounding WMHs, also found for SLE patients (Gasparovic et al., 2010). While these prior findings could be related to small vessel disease or ischemia, which are possible factors contributing to WMH load (Luyendijk et al., 2011;Shaharir et al., 2018;Magro Checa et al., 2013 Jun), few studies have specifically investigated perfusion-based metrics in WMHs in SLE. However, hyperintensity lesions in the context of multiple sclerosis (MS) have been reported to show increased CBF and CBV compared to NAWM, related to contrast enhancement as Table 3 Descriptive WMH data for the SLE group and the SLICC A, SLICC B and NPACR groups. The differences in rCBF, rCBV, rMTT and K 2 are presented as the mean increase in percent in WMHs relative to NAWM. Values shown in bold were statistically significant (p < 0.05). an expression of inflammatory activity (Lapointe et al., 2018;Yin et al., 2018). Although WMHs in SLE often have different characteristics on conventional MRI compared to those seen in MS, SLE patients could also present with demyelinating disease as an NP event, complicating the diagnosis (Magro Checa et al., 2013 Jun). Given the non-specific nature of WMHs in SLE (Ota et al., 2022;Jennings et al., 2004;Luyendijk et al., 2011), our findings raise the possibility of inflammation with vasodilation rather than ischemia, highlighting the importance of further research to provide nuance and additional knowledge.
In addition, we also found K 2 to be increased in WMHs compared to NAWM in the SLE cohort, which reflects the level of contrast agent leakage into the extravascular extracellular space and would be an indicator of a compromised BBB. This study is, to our knowledge, the first to investigate the BBB using K 2 specifically in WMHs for SLE patients. Previous research has used DCE-MRI and found the permeability weighted volume transfer constant (K trans ) to be increased in SLE compared to HC, while utilizing an explorative method of ROI placement in certain brain regions (Chi et al., 2019;Kamintsky et al., 2020). Since we measured K 2 in absolute values without normalization, we decided it was most suitable to compare differences between WMHs and NAWM within the SLE group and not compared to HC. Our results support the hypothesis of BBB dysfunction as part of the vascular pathology and potential WMH formation in SLE patients (Stock et al., 2017). This increase in BBB permeability is also consistent with DCE-MRI research in healthy individuals (Dewey et al., 2021), MS patients with active lesions (Yin et al., 2018) and cerebral small vessel disease (Huisa et al., 2015;Wong et al., 2019), where similar results have been found indicating a disruption of the BBB in different WMHs compared to NAWM. Our and others' findings suggest that K 2 should be of value, especially when used in tandem with K trans and measurements of cerebral perfusion, to investigate the integrity of the BBB related to vasodilation, and possibly affected by neuroinflammation in SLE patients (Nikolopoulos et al., 2019;Stock et al., 2017).
This study is not without limitations. The perfusion values could be affected by PVEs due to the DSC-MRI spatial resolution, which might include surrounding tissue in the ROIs. However, in order to minimize these effects, we placed each ROI carefully and with guidance of the higher spatial resolution of the FLAIR and MPRAGE images. Additional consideration was given to the WMHs, where hyperintensities smaller than 3 pixels (volume <0.044 ml) were excluded for this reason. Normalizing the perfusion values to a reference region reduces PVE related errors in the quantification (Jafari-Khouzani et al., 2015) but introduces the assumption that a certain brain region is unaffected by SLE. The middle cerebellar peduncle has previously been used with the motivation of being less often affected (Papadaki et al., 2018;Papadaki et al., 2019), and we wanted to use a similar approach. However, the overall lack of a common denominator in the literature (Gasparovic et al., 2010;Wang et al., 2012;Emmer et al., 2010;Zimny et al., 2014) could influence the comparability to other DSC-MRI studies in SLE. Another study limitation is the use of different scanners and imaging protocols for the SLE and HC cohorts. While the 90 • flip angle might increase the T1 effects in the acquired DSC images for the SLE cohort, the use of a longer TR, as well as normalization of the perfusion-based values, reduces the sensitivity to any dissimilarities related to different settings in the image acquisition (Jafari-Khouzani et al., 2015;Semmineh et al., 2018). Furthermore, quantification of K 2 can be hampered by arrival time delay and an additional correction method could be used in future studies (Leigh et al., 2012).
Finally, the conclusions may be hampered by the sample sizes of the study groups, although our overall SLE cohort is larger than most previous DSC-MRI studies (Borrelli et al., 2003;Gasparovic et al., 2010;Wang et al., 2012;Emmer et al., 2010;Zimny et al., 2014). However, larger sample sizes would allow for the analysis of additional subgroups, for example, evaluation of the perfusion based on specific NP manifestations and with different SLE disease activity in general. This could potentially gain further insight into the vascular pathophysiology behind the different phenotypes of NP involvement in SLE. In our cohort the patients had no or low ongoing SLE activity, and any immunosuppressive treatment may have influenced the results, in particular, by potentially reducing pathological findings. A larger sample size would also allow the inclusion of all sexes. Although SLE is a disease most commonly seen in females (Unterman et al., 2011), a generalization of our results should be avoided. Even if a small majority of the HC were female, there was indeed a difference in male-female composition between the SLE and HC groups. Besides previously documented differences in perfusion values between men and women using DSC-MRI (Shin et al., 2007), when removing the male HC in our analysis a similar pattern in all perfusion metrics was obtained in the all-female control group compared to the whole HC group. The general direction of increases and decreases in the regional perfusion ROIs compared to SLE remained consistent. However, significances in the perfusion metrics were reduced, mostly due to a decreased statistical power.

Conclusion
This study used DSC-MRI to demonstrate significant alterations in cerebral perfusion in female SLE patients with and without NP involvement, regardless of NPSLE attribution model, compared to HC. WMHs showed significantly increased rCBF, rCBV, rMTT and K 2 compared to NAWM in the SLE cohort. These findings provide additional insight into the role of BBB dysfunction and the vascular properties of WMHs in female SLE patients. Despite the disease being most frequent in females, generalization of our conclusions to SLE patients overall should be avoided. Several brain regions could be of interest in SLE, and further prospective well-regulated studies, evaluating cerebral perfusion and its clinical impact on the disease, need to be conducted.
The funding sources had no involvement in the study design, in the collection, analysis and interpretation of data, in the writing of the report and in the decision to submit the article for publication.