Cross sign T2 hyperintensities in atrophic spinal cord of hereditary spastic paraplegia type 5

Background: Hereditary spastic paraplegias type 5 (SPG5) is an inherited neurodegenerative disease with 27-hydroxycholesterol abnormal accumulation. Imaging and pathologic manifestations remain poorly understood due to the rare incidence. This study reveals the MRI features of SPG5, and aims to investigate a promising imaging diagnostic biomarker for SPG5. Methods: We prospectively recruited SPG5 patients and matched healthy controls from Neurogenetic Diseases Centers of Fujian province in China, clinical and MRI data of whom were collected. Abnormalities of spinal cord and brain were characterized and quantied by conventional and quantitative MRI. Comparisons were conducted between MRI and cerebrospinal uid (CSF) bioindicators. Results: Seventeen SPG5 patients were enrolled (11 men, 6 women; age range, 13–49 years; median disease duration, 14 years). For the rst time, T2 hyperintensities with “+” form (cross sign) in atrophic spinal cord was found among all SPG5 patients. To grade severity of this sign, we set up a scoring scale (cross-sign scores) in cervical spinal cords. Unexpectedly, total cross-sign scores showed a strong positive correlation with disability scale scores (r = 0.687, P = 0.002) and disease duration (r = 0.520, P = 0.032). Although total spinal cord area was reduced (cervical levels: 12-27%; thoracic levels 41-60%), no correlation was found between spinal cord atrophy and disease severity. In CSF, a positive correlation was identied between 27-hydroxycholesterol and neurolament light (r = 0.468, P = 0.049), although 27-hydroxycholesterol and neurolament light were unrelated to disease severity. Conclusion: Cross sign spinal cord diagnostic biomarker


Background
Hereditary spastic paraplegias (HSP) are a large, genetically diverse group of inherited neurological disorders characterized by a length-dependent distal axonopathy of the corticospinal tract, resulting in lower limb spasticity and weakness (1). Hereditary spastic paraplegia type 5 (SPG5), a subtype of HSP, is caused by autosomal recessive loss-of-function mutations in CYP7B1, that encodes oxysterol-7ahydroxylase. This mutations leads to the accumulation of CYP7B1 substrates, such as 27hydroxycholesterol , in plasma and CSF (2,3).
Previous reports have indicated that hypercholesterolemia confers a higher risk for Alzheimer's disease, and evidences suggest that 27-OHC plays a pivotal role in Alzheimer's-associated neural injury (4)(5)(6). As an oxidized derivative of cholesterol, 27-OHC has cytotoxic and pro-apoptotic properties, which is increased approximately 0.5-fold (40-80 ng/mL) in patients with cognitive impairment and 10-fold (600-1300 ng/mL) in patients with SPG5 (3,7). The observation of extremely high levels of 27-OHC prompted us to hypothesize that an accumulation of neurotoxic oxysterols could cause abnormalities in the spinal cord and brain. In fact, T2 white matter hyperintensities of the brain have been reported in case studies of SPG5(8) , (9) . However, image manifestations of SPG5 are poorly understood owing to limited diagnostic tools for many patients.
Our center focuses on HSP research and have set up an HSP cohort to record natural disease history. To date, 34 SPG5 patients have been long-term followed, 28 of whom were previously reported (10). In this study, we applied conventional and quantitative MRI techniques to characterize and quantify signal and structural changes of spinal cord and brain among SPG5 patients. Subsequently, we analyzed the correlations between MRI-derived measurements with the clinical status and molecular biomarkers of these patients.

Methods
Standard protocol approvals, registrations, and patient consents The study protocol and informed consent procedures were approved by the institutional review board at This report provides the rst description of a T2 hyperintensity cross sign in the spinal cords of SPG5 patients, a form of hyperintensity resembling a "+" observed in T2* MEDIC, T2-PD, and T2 sequences. This sign corresponds with the symmetric hyperintensities of the anterior and dorsal columns with a preserved anterior horn (in contrast with a blurred dorsal horn) on the T2* MEDIC sequence ( Fig. 1).
Cross-sign scoring (CSS) scale was performed according to the following observations (outlined in T2*-MEDIC datasets of all patients and healthy controls were excluded due to technical limitations, so cross sign was assessed only in the C2 to C7 vertebral disc levels. A single axial T2*-MEDIC image of the most affected slice in every intervertebral disc level was identi ed for scoring. The two participating radiologists were instructed in the rating system for this study (C.D.R. who has 30 years of neuroradiology experience and Y.L. who has 10 years of neuroradiology experience); both were blinded to the clinical information. The intraclass correlation coe cients were also calculated. The summed scores of all intervertebral disc levels were recorded.

Spinal cord cross-sectional area measurement
After completion of the MRI scans, the 3D T2 images of the cervical and thoracic spine areas were simultaneously stitched to ensure full presentation of each cervicothoracic spinal cord (C1-T9) acquired by Compose software on a SIEMENS workstation (Fig. 3A). Spinal cord segmentation and cross-sectional area measurements were performed using Spinal Cord Toolbox, Version 4.01 (https://sourceforge.net/projects/spinalcordtoolbox/). Morphological spinal cord metrics were extracted at each vertebral level (from C1 to T9), including spinal cord cross-sectional area and sectional diameter, for further analyses.

Conventional brain MRI evaluation
Abnormal signals were assessed in T2, T1, FLAIR, and SWI routine sequences. Some neuroradiological signs, which have been previously reported in speci c subtypes of HSP (11), were also analyzed in the current SPG5 patients.
Visual rating of cerebral atrophy Visual rating of 3D T1 images from each participant was performed by two experienced neurologists who were trained for consistency in scale evaluation. Images were rated in native space with RadiAnt DICOM Viewer software. Six regions were rated based on existing scales and a simpli ed version of the overall rating was used with a detailed evaluation protocol, as previously described (12).
Quantitative brain MRI evaluation Voxel-based morphometry (VBM) was performed to assess alterations in brain volume and grey matter volume (GMV) of patients with SPG5 compared to healthy controls.  (T2*-MEDIC, T2-PD, T2). This sign manifested as a symmetric hyperintense signal of the anterior and dorsal columns with a preserved anterior horn (the dorsal horn appears blurred). In contrast, none of the healthy controls exhibited such a speci c sign (Fig. 1).
Cross sign scoring was performed by two experienced radiologists using the observational criteria outlined in Fig. 2A-B. Inter-scorer reliability assessment of reliability between radiologist scorers showed intraclass correlation coe cients from 0.9 to 0.95 at all levels (P < 0.001), thus demonstrating high reproducibility between observers. Cross signs were more pronounced at the C2 and C7 levels among all the explored cervical segments ( Fig. 2A) Total cross sign scores correlated with the disease severity. Spearman coe cients demonstrated positive correlations between total cross sign scores and SPRS (r = 0.687, P = 0.002) as well as disease duration (r = 0.520, P = 0.032). However, total cross sign scores did not correlate with either CSF 27-OHC or CSF NFL concentrations.

Spinal cord atrophy quanti cation
Since the proposed processing pipeline was fully automated, automatic registration into the Spinal Cord Toolbox template was successful in most cases (Fig. 3A). Manual correction was needed for the spinal cord segmentation of only 2 subjects (1 patient and 1 healthy control).
Spinal cord areas were signi cantly smaller in patients than in controls at all the evaluated cervical and thoracic levels (Fig. 3B-D). Moreover, the relative reduction in area was more pronounced at the thoracic levels (cervical levels: 12-27%; thoracic levels 41-60%), especially in T4 (35 vs 22 mm 2 , P < 0.001). At the T4 level, the relative reduction of antero-posterior diameters was greater than the reduction in transverse diameters (24% vs 18%, P = 0.101). The area under the ROC for the T4 spinal cord area was 0.976 (P < 0.001), and the cut-off value was 29 mm 2 (sensitivity 88% and speci city 94%).
In no instance did the total spinal cord area or spinal cord area at any of the examined levels correlate with disease duration, clinical scores, CSF 27-OHC, or CSF NFL (P > 0.05). Conventional MRI ndings in the brain No signal abnormalities were observed in the brain T2, FLAIR, or SWI sequences of any SPG5 patients or healthy controls. None of the speci c signs that have been previously reported for subtypes of HSP appeared in any of these SPG5 patients, including "Ear-of-the-lynx" sign, enlarged ventricles, brain white matter T2 hyperintensities, and bilateral T2 hypo-signal of the globus pallidus (Fig. 4).
Brain atrophy measurements Based on MRI visual rating scales, there were no differences in cerebral atrophy assessed with scores for each region or total scores from SPG5 patients compared to healthy controls (Fig. 5A).

Discussion
Since the core clinical imaging features of SPG5 has not been described previously, we undertook the present study not only to provide a detailed characterization of the structural signature of SPG5, but also to investigate what speci c patterns of spinal cord and brain damage correlate with clinical and pathogenic manifestations. To accomplish these goals, we enrolled a relatively large set of patients who underwent systematic clinical assessment combined with multimodal MRI and CSF markers evaluations. As a result, we have described the neuroimaging ndings from both clinical and molecular pathologic viewpoints. First, the signal abnormality at cervical levels was reliably re ected by conventional spinal cord MRI among the SPG5 patients, and "+" T2 hyperintensities (cross sign) in the atrophy spinal cord appeared in all 17 patients. Second, we set up cross sign scoring to grade the severity of this sign, and subsequently found that total cross sign scores bore a strongly positive correlation with the extent of disease degeneration, as assessed by standardized disability scales and disease duration. Third, as expected, each total spinal cord area had decreased signi cantly at each site investigated, in comparison with those of healthy controls, especially in the T4 level. However, no correlation was evident between spinal cord area and either disability or pathogenesis-associated molecular biomarkers. Fourth, no abnormality was present in conventional brain MRI or quantitative brain MRI-derived data, with the exception of a mildly reduced gray matter on the thalami.
T2 hyperintensities in spinal cord MRI are commonly associated with a large variety of causes (in ammation, infections, neoplasms, vascular, and spondylotic diseases), but it is rarely caused by HSP, except as reported for hereditary spastic paraplegias type 2 (SPG2), due to PLP1 gene mutations (13).
In this study, the included SPG5 patients were retained for long-term follow-up and, no other disease was identi ed. The cross sign spinal cord T2 hyperintensities, rst de ned in this report, appeared in all 17 of our patients on three different sequences (T2, T2-PD, T2*-MEDIC), and differed from the T2 hyperintensity morphology in patients with SPG2. Thus, this spinal cord cross sign is a speci c imaging feature for SPG5 and can therefore serve as a potentially useful diagnostic biomarker for SPG5, both in guidance for genetic testing or as a reference for interpreting genetic ndings.
Upon further investigation, we found that the spinal cord of SPG5 patients was atrophied but without correlation to the extent of disability or duration of disease, as reported for other subtypes of HSP (14). The lack of correlation may be partially explained by slow progression of SPG5 and low sensitivity of the imaging biomarker (spinal cord area). In contrast, the total cross sign score may provide a more sensitive imaging marker than other features of the spinal cord for quantifying the degree of neurodegeneration. As such, the cross-sign scores can be a valuable and informative imaging biomarker for monitoring disease process as well as the therapeutic response of SPG5 patients.
The accumulation of 27-OHC may be not only a biomarker but also a key factor in driving tissue damage in patients with SPG5. NFL is a protein component of the cytoskeleton of myelinated axons, and, as such, constitutes a putative biomarker to re ect axonal injury (15)(16)(17). Further, the concentration of CSF 27-OHC, which had a positive correlation with CSF NFL concentration in our study, could potentially indicate an association of neurotoxic 27-OHC with axonal injury. However, no signi cant correlation was identi ed between the concentrations of 27-OHC or NFL in the CSF with the progressive degeneration shown in clinical load (SPRS, disease duration) or in the worsening imaging indicators (total cross sign scores, spinal cord area). Therefore, although CSF 27-OHC and CSF NFL may be suitable markers for monitoring disease activities, they are not useful guides to the overall progression of SPG5.
Surprisingly, we failed to identify obvious cerebral atrophy in SPG5 through two advanced imaging techniques (visual rating of cerebral atrophy and quantitative brain measurements), as reported in Alzheimer's disease (18). However, emerging evidence has revealed common pathological mechanisms in neuronal injury from 27-OHC in familial Alzheimer's disease and SPG5 (6,(19)(20)(21). In contrast, spinal cord morphometry indicated more substantial atrophy in SPG5 patients. Moreover, speci c T2 hyperintensities visible on spinal cord MRI support the likelihood that severe spinal cord damage from 27-OHC might be implicated in SPG5 pathogenesis.
SPG5 (CYP7B1 gene mutation) and cerebrotendinous xanthomatosis (CTX, CYP27A1 gene mutation) are neurological diseases with mutations in genes that participate in cholesterol metabolism (22)(23)(24). Rare cases of genetically and biochemically con rmed 'spinal CTX' have been reported (25)(26)(27)(28)(29)(30). Spinal CTX differs from the classical cases with a relatively benign course, whose clinical presentations are dominated by spinal symptoms (spastic paraplegia, alteration to deep sensation, and urinary involvement), and paucity or absence of the classical neurological and systemic symptoms. These clinical features are almost indistinguishable from SPG5. In spinal CTX patients, spinal MRI studies revealed longitudinally extensive posterior and lateral column white matter abnormalities on T2 images, which strongly resemble our observations in SPG5 patients (29) , (30) . Together, the mutations associated with cholesterol metabolism and similar clinical and spinal cord MRI features suggest the likelihood of a similar pathogenic mechanisms for spinal cerebrotendinous xanthomatosis and SPG5.

Conclusion
In this work, we provide the rst evidence to our knowledge that T2-associated sequences show "+" signal hyperintensities in the atrophied spinal cords of SPG5 patients, and that this "cross sign" is associated with disease-related degeneration. This imaging sign can reinforce the validity of diagnoses, facilitate assessment of disease progression, and assist in monitoring of therapeutic effects. More importantly, the identi cation of new imaging features in SPG5 provides fresh insight into the pathogenic mechanism for oxidized derivatives of cholesterol.