Correlation between cortical lesions and cognitive impairment in multiple sclerosis

Abstract Objectives Gray matter (GM) damage is well known as a fundamental aspect of multiple sclerosis (MS). Above all, cortical lesions (CLs) burden, detectable at MRI with double inversion recovery (DIR) sequences, has been demonstrated to correlate with cognitive impairment (CI). The aim of this study was to investigate the role of CLs number in predicting CI in a cohort of patients with MS in a clinical practice setting. Materials and methods Thirty consecutive patients with MS presenting CLs (CL+) at high‐field (3.0 T) MRI 3D‐DIR sequences and an even group of MS patients without CLs (CL‐) as a control, were investigated with the Rao Brief Repeatable Battery of Neuropsychological Tests (BRB), Version A. Total and lobar CLs number were computed in CL+ patients. Results Among the sixty patients with MS enrolled, forty‐seven (78.3%) had a relapsing‐remitting course, while thirteen (21.7%) a progressive one, eleven secondary progressive, and two primary progressive. Compared to CL−, CL+ patients had a greater proportion of progressive forms (p = .03). The most affected region was the frontal lobe (73.3% of patients), followed by temporal and parietal ones (both 60.0%). Multivariate (logistic regression) analysis revealed a significant correlation between total CLs number and the presence of mild cognitive impairment defined as pathologic score in at least one BRB test (p = .04); it was also correlated with deficit at PASAT 3 (p = .05) and Stroop Test (p = .02). Conclusions We confirmed CLs number, evaluated with a technique quite commonly available in clinical practice, as a predictive factor of CI in patients with MS, in order to improve the diagnosis and management of CI and monitor potential neuroprotective effects of therapies.

Because conventional MRI techniques cannot easily reveal GM lesions, more specific sequences have been introduced, such as double inversion recovery (DIR), which selectively suppresses signal from WM and cerebrospinal fluid (CSF), improving the detection of CLs (Geurts et al., 2005;Turetschek et al., 1998).
International guidelines have been proposed to facilitate detection of CLs . Moreover, a three-dimensional version (3D-DIR) allows for a shorter acquisition time and a reduction in artefacts (Roosendaal et al., 2009). With the application of these techniques, several studies have demonstrated that CLs are a typical finding in all phenotypes and stages of MS, including clinically isolated syndrome (CIS) (Calabrese et al., 2007(Calabrese et al., , 2008Geurts et al., 2005) and radiologically isolated syndrome (Calabrese & Gallo, 2009;Giorgio et al., 2011), although they are more represented in the progressive forms and with longer disease duration (Calabrese et al., 2015;Kutzelnigg et al., 2005;Mike et al., 2011;Roosendaal et al., 2009).
Cognitive impairment is an important component of MSrelated disability at both the earlier and later stages of the disease, affecting from 40% to 70% of patients with MS. Several cognitive domains, such as long-term memory, attention and concentration, executive functioning, and information processing speed (IPS), may be involved (Amato, Zipoli, & Portaccio, 2006;Chiaravalloti & DeLuca, 2008;Jongen, Ter Horst, & Brands, 2012;Lovera & Kovner, 2012). In recent years, some MRI studies using DIR sequences have demonstrated a significant correlation between CLs burden, mainly measured as CLs volume, and cognitive dysfunction in patients with MS (Calabrese et al., 2007Damasceno et al., 2015;Mike et al., 2011;Nelson et al., 2011;Papadopoulou et al., 2013;Roosendaal et al., 2009).
Although there is also some evidence of a correlation between regional GM and WM damage measures and performances at neuropsychological tests (Morgen et al., 2006;Nocentini et al., 2014;Sperling et al., 2001), in our knowledge, there are only sporadic reports of a correlation between regional CLs number and impairment within specific cognitive domains (Roosendaal et al., 2009).
With this study, conducted with high-field 3.0 Tesla (T) MRI 3D-DIR sequence, a technique nowadays quite available in the clinical practice setting in Europe, we aimed to investigate the role of CLs load, measured as CLs number, in predicting CI in a cohort of patients affected by relapsing-remitting (RR) and progressive MS. We also analyzed the regional distribution of CLs in the various cerebral lobes and their relative contribution to CI.

| Patients
We enrolled a group of thirty consecutive patients with MS diagnosis according to revised 2010 McDonald criteria, referred to MS Centre of Parma between October 2014 and January 2015 and presenting CLs (CL+) at a brain MRI scan with DIR sequences. An even group of consecutive MS patients without CLs (CL−) was enrolled as a control. All individuals were investigated for MS-related CI with specific neuropsychological tests.
Depression was evaluated by the Beck Depression Inventory (BDI) (Hautzinger, 1991), with regard to the well-known influence of mood disorders on cognitive performance. Patients with severe depression (BDI > 31) were excluded from the study. The study protocol was approved by the "Ethics Committee Province of Parma," and all subjects gave written informed consent prior to participation.

| Image acquisition
All MR images were obtained with a 3.0 T MR scanner (Discovery MR750; GE Healthcare, Milwaukee, WI) equipped with an 8-channel phased-array head coil. We performed 3D fast spin echo (Cube) DIR T2-weighted pulse sequence in order to obtain optimal GM-WM contrast for optimal visualization of cortical MS lesions. This technique is based on 3D-FSE-XETA (eXtended Echo Train Acquisition), an advancement of 3D-FSE in which refocusing flip angle modulation and 2D acceleration enable volumetric coverage with high in-and through-plane resolution within clinical scan times of 3-6 min. Cube DIR sequences were obtained using two adiabatic inversion recovery RF pulses to null the signal from CSF and WM. The TI values for nulling the CSF were set at 3,000-5,000 ms (TIcsf) and WM (TIwm) at 750-800 ms. The Cube DIR sequence parameters were as follows: TE/TR 85/6,900 ms, receiver bandwidth 62.50 kHz, FOV 24 × 24 cm, imaging matrix 256 × 256, section thickness 1.6 mm, 116 sections, NEX 2.0 with parallel imaging acceleration factor 2.0. Intrinsic low SNR was exceeded increasing NEX.  Raw score at each test was corrected considering also age, gender, and education, according to normative Italian data (Amato et al., 2006). Tests with scores 2 SDs below the mean normative values were considered pathological. Patients with at least one altered test were considered affected by mild cognitive impairment (MCI), while patients with at least two altered tests were considered affected by definite cognitive impairment (DCI).

| Statistical analysis
Chi-square test, Student's t test, and median test were used to compare demographic and clinical features between the two groups of patients, with and without CLs at DIR sequences, for categorical and continuous variables, respectively.
To investigate the relationship between MRI and neuropsychological variables, univariate correlations between continuous variables were assessed using the Pearson correlation coefficient.
A multivariate analysis (logistic regression) was performed to determine the relative contribution of demographic (age and gender), clinical (EDSS, disease duration and phenotype), and MRI (total, frontal, temporal, and parietal CLs number and total T2/FLAIR lesions number) variables in predicting the presence of DCI, MCI, and pathologic performance in single BRB tests. We also applied the same multivariate analysis in a subgroup including only patients affected by the RR phenotype.
All statistical analyses were performed using SPSS version 22 (IBM Corp.) statistical package.
In more detail, with regard to the regional pattern of distribution of CLs in CL+ patients, we found that the most affected area was the frontal lobe (22 patients, 73.3%), with a mean burden per patient of 3.8 ± 3.85 lesions (range 1-15), followed by temporal (18 patients, 60.0%), with a mean burden of 3.2 ± 2.90 lesions (range 1-10) per patient, and parietal lobe (18 patients, 60.0%) with a mean burden of 1.7 ± 1.32 lesions (range 1-6) per patient. The occipital lobe was spared in all patients but one.
In Figure 1, there are some examples of DIR CLs images in our population.

| Cognitive impairment
We found DCI, determined by failure in ≥2 BRB tests with scores 2 SD below the mean value score, in 11 patients (18.3%), while MCI, defined by failure in ≥1 BRB test, was present in 24 patients (40.0%).
Analyzing the frequency of CI in the two groups of patients CL+ and CL−, we found a greater proportion of patients with both DCI and MCI in the CL+ group, even though they did not reach the statis-  Table 3.
The multivariate analysis conducted in the RRMS subgroup (n = 47) demonstrated a significant correlation between total CLs number and SDMT (p = .006). Even total T2/FLAIR lesions number was correlated only with SDMT (p = .001). As for the entire population, even this subgroup analysis revealed only age as independent predictor of DCI (p = .007); age was also correlated with SDMT (p < .001).

| D ISCUSS I ON
In our study, CLs were mainly located in the frontal lobe (48.3%), followed by the temporal (33.1%) and the parietal (18.0%) ones; we found only a single occipital lesion (0.6%) in one patient. Moreover, CLs were more common in the progressive forms (p = .03). These results are consistent with the literature. In fact, the development of CLs has been reported more frequently in the temporal and frontal lobes, while the occipital lobe is less affected (Calabrese, Favaretto, Martini, & Gallo, 2013 In the interpretation of these results, we have to consider several limitations presented by our study. First, the size of our sample was probably too limited and the "matching" of the two groups too incomplete to allow definitive conclusions. It is likely, in fact, that the lack of a statistically significant correlation between CLs and CI is mainly due to this reason. We had also to consider the lack of a control group well matched to the patients. Secondly, the MRI technique we employed, that is 3D-DIR, is far from perfect in detect-  (Seewann et al., 2012). In addition to that, we have to consider that subpial CLs, which are also associated with physical and cognitive dysfunction in patients with MS, are poorly detected by conventional MRI and even DIR . However, other comparative neuropathological and imaging studies demonstrated that the number of CLs detected at MRI correlates well with their total number revealed by histology, concluding that "the tip of the iceberg detected by MRI and its bulk differ only in size." (Calabrese et al., 2013;Geurts et al., 2012;Seewann et al., 2011) Then, there are some consensus recommendations for MS CLs scoring using DIR MRI , while other more accurate techniques, such as "Phase Sensitive Inversion Recovery" (PSIR) or ultra-high-field MRI lack of standardization and are less available in clinical practice.
In addition to the previously reported limitations, we measured only CLs number but not volume, even though several studies reported CLs volume as a more accurate predicting factor of CI.
However, volumetric measures are usually limited to research setting, while CLs count can be easily evaluated. Then, the regional CLs distribution we adopted (i.e., cerebral lobes) is quite rough, as it does not consider functional pathways. Finally, we have to consider that CLs burden is only one factor of the complex biological interplay between WM disease and GM disease which likely sustains CI. A lot of previous MRI studies reported correlation between CI and various neuroimaging markers, including WM lesion load and CLs, whole-brain, cerebellar, deep GM, callosal and cortical atrophy, and normal appearing WM damage (Daams et al., 2016).
In spite of these limitations, our study confirmed the correlation between total CLs number and low performances in several cognitive tests which explore memory, attention, and IPS, as shown in previous studies Mike et al., 2011;Nelson et al., 2011;Roosendaal et al., 2009). Moreover, we found this correlation also with CLs number detected in single cerebral lobes, with the exception of the occipital one, where we found only one lesion. In our knowledge, there are only sporadic reports in the literature of a direct correlation between regional CLs number and deficit in specific neuropsychological tests, such as one study describing the association between hippocampal lesion number and visuospatial memory (Roosendaal et al., 2009). Besides, some previous studies described the relationship between regional GM atrophy and cognition (Morgen et al., 2006;Nocentini et al., 2014); another study, instead, reported a correlation between frontal and parietal WM lesions volume and impairment on neuropsychological tests evaluating sustained attention, processing speed, and verbal memory (Sperling et al., 2001). Also, our data reported an association between single BRB tests and CLs number detected in various cerebral lobes, according to the aforementioned studies. This lack of anatomic specificity could be explained by the involvement of different cognitive functions associated with several cortical areas during the performance of a single test (Nocentini et al., 2014

ACK N OWLED G M ENTS
None.