Structure-function correlations in Retinitis Pigmentosa patients with partially preserved vision: a voxel-based morphometry study

Retinitis Pigmentosa is a group of hereditary retinal dystrophy disorders associated with progressive peripheral visual field loss. The impact of this retinal loss in cortical gray matter volume has not been addressed before in Retinitis Pigmentosa patients with low vision. Voxel-based morphometry was applied to study whole brain gray matter volume changes in 27 Retinitis Pigmentosa patients with partially preserved vision and 38 age- and gender-matched normally sighted controls to determine whether peripheral visual loss can lead to changes in gray matter volume. We found significant reductions in gray matter volume that were restricted to the occipital cortex of patients. The anteromedial pattern of reduced gray matter volume in visual primary and association cortices was significantly correlated with the extent of the peripheral visual field deficit in this cohort. Moreover, this pattern was found to be associated with the extent of visual field loss. In summary, we found specific visual cortical gray matter loss in Retinitis Pigmentosa patients associated with their visual function profile. The spatial pattern of gray matter loss is consistent with disuse-driven neuronal atrophy which may have clinical implications for disease management, including prosthetic restoration strategies.

Retinitis Pigmentosa (RP) is an important cause of human visual disability characterized by early peripheral visual field defects (VFD). The non-syndromic form of RP has a worldwide prevalence of about 1/4000 individuals [1][2][3][4] . It is classified as an inherited retinal dystrophy disorder which is characterized by photoreceptor degeneration and subsequent visual loss 3 . In the typical form of RP, retinal degeneration affects rod photoreceptors in the mid-periphery, advancing towards the macula and fovea with cone photoreceptors affected only in later stages. Patients initially present with nyctalopia, followed by the progressive peripheral VFD, and eventually blindness. The age of onset is highly variable, ranging from childhood to mid-adulthood 1, 2 .
Voxel-based morphometry (VBM) is a well-established technique 5-7 that has been used to report local gray matter (GM) volume alterations in several disorders leading to VFD [8][9][10][11][12][13][14][15][16][17][18][19] . However, some of these studies have only focused on patients whose late VFD had already evolved to blindness and include multiple neuro-ophthalmological pathologies within the same study. Thus, different mechanisms of visual loss and putative brain adaptation were likely present [16][17][18] . VBM studies in patients with partially preserved vision due to VFD have mainly concerned glaucoma, also characterized by peripheral VFD, or age-related macular degeneration (AMD) characterized by central or pericentral VFD [8][9][10][11][12][13][14][15] . GM loss has been reported in the visual cortex of AMD and glaucoma patients in cortical representations matching the retinal lesion projection zones 8 . Several other studies have found cortical changes, mainly volumetric reduction, in patients with central retinal degeneration 10,11,19,20 . Furthermore, in a morphometric study involving patients with hereditary macular dystrophies with central scotomata, the authors found a correlation between scotomata size and GM volume, suggesting a direct association between the lack of visual input and GM atrophy 11 . Regarding peripheral VFD, the focus of morphometry studies has been mainly on glaucoma 8,[12][13][14][15]21 . Li et al. (2012) 12 had analyzed the extraoccipital cortex in progressive stages of primary open-angle glaucoma and did not find GM volumetric changes in the early stage group. On the contrary, in the advanced-late stage group, reduced GM volume was found in multiple cortical brain regions including the primary visual cortex 12 . Conversely, these authors found higher GM volume in regions neighboring the most damaged brain region, which may reflect the possibility of structural reorganization. Similar studies also reported volumetric decreases and increases in several brain regions in glaucoma 14,15 .
To the best of our knowledge, only one study has investigated the presence of visual cortical volumetric reduction due to peripheral VFD in disorders other than glaucoma. In that study, wider calcarine fissures were reported in nine patients with retinal pathologies, particularly in the anterior and middle portions, which is also consistent with peripheral VFD 22 . Nevertheless, only four patients presented with pigmentary degeneration of the retina and, since the authors reported only the calcarine fissure width, it remains unclear whether more widespread changes related to the visual loss were present.
The present work aims to determine whether progressive peripheral vision loss in RP patients might lead to structural cortical changes, not only in the visual cortex but also in the whole brain. We hypothesize that RP patients with low vision would have reduced GM volume, mainly in the visual cortex due to visual field constriction, as previously shown for other late-onset VFD pathologies 8,11,14 . Furthermore, considering the VFD similarities with glaucoma, neighboring regions with increased GM volume might also be present 14,15 . This question is relevant as GM alterations caused by peripheral VDF in a pathology other than glaucoma have not been previously addressed. For that purpose, a whole-brain VBM analysis was used to investigate the possible link between peripheral VDF and GM abnormalities in the currently largest cohort of low vision RP patients included in morphometric studies.

Results
Demographic and visual data.  Table 2), which may lead to an overestimation of the retinal and the RNFL thickness. The mean deficit of the visual field was significantly higher for the RP group when compared to the control group in both eyes [right eye U = 0.50, p = 7.274 × 10 −11 and left eye U = 4.00, p = 1.746 × 10 −10 ; 4 RP patients could not perform the examination with both eyes and 1 RP patient with the left eye]. Patients' visual field diameter ranged from 5 to 48 degree in the right eye and 10 to 48 degrees in the left eye, as assessed with static perimetry. It was not possible to estimate the visual field extent of 4 RP patients, who could not perform the exam with both eyes, and from the left eye of 1 RP patient. Additionally, no statistical differences were found between the right and left eye within the RP and control groups for any of the visual parameters studied: visual acuity [RP Z = −0.11, p = 0.911; controls Z = −0.57, p = 0.566], retinal thickness [RP t (25)   VBM data. Figure 1 shows the statistical parametric map of reduced GM volume in RP patients when compared to controls. Reduced GM volume was found in both hemispheres and involved the following brain regions: right and left calcarine, right and left lingual, right and left cuneus, and right occipital superior gyrus ( Table 3). The analysis with the contrast RP > Controls yielded no statistically significant results (i.e. no regions with significant increased GM volume was found in RP patients). Figure 2 shows the statistical parametric maps of statistically significant correlations between cortical GM volume and visual field loss (visual field diameter and mean deficit variables) in RP patients. The same regions showed GM loss related with visual function deterioration. Correlation analysis between GM volume and preserved visual field diameter showed significant positive associations in the following brain regions: right and left cuneus, right and left calcarine, right and left lingual, right and left precuneus. Correlation analysis between visual field mean deficit and GM volume showed negative associations in the following brain regions: right and left calcarine and right and left lingual, and in right and left cuneus and right occipital superior, respectively (Table 4). No other visual parameters (visual acuity, retinal thickness, or RNFL thickness) or disease duration showed statistically significant correlations with GM volume.

Discussion
Our results showed that Retinitis Pigmentosa (RP) patients with low (partially preserved) vision have significantly reduced gray matter (GM) volume when compared to healthy controls in several components of the primary and association visual cortices. Moreover, visual cortical atrophy in RP patients was associated with the magnitude of peripheral visual degeneration. To our knowledge, this is the first whole-brain morphometric study conducted in this pathology.
In our study, GM loss was found bilaterally in calcarine sulci, lingual gyri, cuneus, and also in the right occipital superior gyrus. Patterns of loss in anteromedial occipital cortex were correlated with the extent of field loss   Given that the visual cortex is organized into retinotopic maps, it is relevant to discuss whether our findings are consistent with the topology of visual loss. In the retina, each visual field representation has receptive fields at nearby locations in the image. Thus, these topological maps preserve the spatial arrangement of the retinal image especially in the primary visual cortex (V1), where adjacent stimuli in the visual field are represented in adjacent positions in the visual cortex. If progressive VFD occur over the years, the corresponding part of the visual cortex gradually loses its input. Thus, continued visual input loss due to VFD can result in structural changes at the visual cortex level, particularly in V1 [24][25][26] . In this way, the typical peripheral VFD of RP corresponds to a partial loss of input to the more anterior and middle visual cortex, which is consistent with our findings. Accordingly, our data (group comparisons and correlation analyses with visual function parameters) corroborates this hypothesis since the observed GM reductions were present in these regions of the occipital cortex, while the more posterior regions were spared (Fig. 1). Moreover, this spatial pattern of reduced GM volume is consistent with previous VBM studies focusing in peripheral VFD 8,22 and, at the same time, mirror the GM volume reductions observed in the case of central VFD. In our study, brain areas showing GM volume reduction included the primary visual cortex but also extended towards the association cortices. These results emphasize that even in low vision patients, widespread GM atrophy is present in several regions of the primary and secondary visual cortices and beyond. Possible explanations for the observed GM reduction in the visual cortex are disuse-driven neuronal atrophy and/ or transneuronal degeneration 23 . The structural alterations of afferent visual pathways in RP have been investigated by measuring retinal nerve fiber layer (RNFL) and retinal thickness, using optical coherence tomography (OCT). Retinal thinning in RP is consensual across studies [27][28][29] , and our findings of reduced retinal thickness as measured by OCT (Table 1) support this notion. On the other hand, RNFL findings have been contradictory in RP. It has been reported to be increased, decreased, or maintained within normal limits in several studies [29][30][31] . In the current work, the average RNFL thickness of RP patients is similar to the control group (Table 1), which suggests axonal integrity of the optic nerve. Optic pathway trans-synaptic neurodegeneration (Wallerian degeneration) is well established in patients with glaucoma-induced retinal ganglion cells (RGC) damage 9, 13-15, 23, 32 . In macular degeneration, volumetric loss of the visual pathway has also been observed 13,32,33 . In this case, the   adverse effect of photoreceptor degeneration (first-order neurons) on RGC (third-order neurons) was proposed as a possible explanation in that disease 19 . Nevertheless, a diffusion tensor imaging study of the optic nerve in RP revealed pathologic structural alterations of the optic nerve 34 , while another found decreased integrity of the optic radiation in RP patients 35 . Further studies are necessary to assess a better understanding of the role of RGC and the visual pathway alterations in RP.
Contrarily to some studies in glaucoma 12, 14, 15 , we did not observe GM hypertrophy in RP patients when compared to controls. Our results of reduced GM volume restricted to the occipital region in the absence of GM increases suggests that, notwithstanding the VFD similarities, the neuropathological features may differ between the two conditions. In fact, the primary retinal damage differs between glaucoma and RP (i.e. RGC versus photoreceptors, respectively) which may lead to a distinct impact in the disease progression. However, we cannot rule out the possibility that the conservative whole-brain analysis used in this study might have missed subtle increases in GM volume.

Figure 2.
Correlations between cortical gray matter volume and visual field loss in the Retinitis Pigmentosa patients. The same altered regions (Fig. 1) present a correlation between cortical gray matter volume and visual field extent in the Retinitis Pigmentosa patients (p < 0.001 with a cluster correction of 150 voxels). In the upper part, it is represented the direct (positive) correlation between gray matter volume and visual field diameter (in degrees). A smaller visual field diameter corresponds to reduced gray matter volume. In the lower part, it is displayed the inverse (negative) correlation between mean deficit (in dB) and gray matter volume. High mean deficit values correspond to reduced gray matter volumes. The color bar indicates the range of t-values with white/yellow representing more significant differences (higher t-values), orange indicating less significant differences (middle range t-values), and red indicating nonsignificant differences (lower t-values). In the correlation analysis for the RP group, GM volume was directly (positively) correlated with visual field diameter in an occipital cluster extending to the posterior region of the medial parietal cortex, and was inversely (negatively) correlated with the visual field mean deficit in two occipital clusters ( Table 4). The cortical regions which correlate with visual parameters remarkably overlap with the regions of reduced GM volume observed in the between-group comparisons. No other visual parameters (visual acuity, retinal thickness, and RNFL thickness) or disease duration showed significant correlations with GM volume. One previous study showed that this measure of disease severity did not correlate with GM reductions caused by central VFD 11 . However, this was expected because the age of symptoms onset was reported by the patients which makes this parameter highly subjective and an imprecise measure of disease severity 1, 2, 36 . Thus, RP's visual field loss, here characterized by the parameters visual field diameter and mean deficit, seems to be the only measure related to the GM volume atrophy. In other words, our data suggest that GM atrophy in the visual cortex of RP patients with partially preserved vision is likely driven by the extent of the visual field loss. In a recent functional magnetic resonance imaging (MRI) work from our group, which included retinotopic assessment of half of the RP patients from the current study, with more preserved visual function, V1 functional alterations were also dependent on the extent of visual field loss and not related to disease duration 36 . Our findings are in line with studies in other hereditary retinal dystrophies, which also reported correlations between the visual field loss and GM atrophy 11,20 .
Our results support the hypothesis that atrophy and degeneration of the visual cortex can occur in patients with low vision even before blindness is settled. These findings may have important clinical implications in disease management because neurodegenerative pathophysiology may be early present in the course of the disease. Current investigational treatment modalities, such as gene replacement therapy, retinal or stem cell transplantation, pharmacologic neurotrophic factors, and neuroprosthetic devices should be implemented early in the course of the disease to avoid the progression to visual cortical atrophy 1, 2, 37 . Moreover, the success of retinal functional restoration highly depends on the integrity of the visual pathways to correctly lead the input to the visual cortex 38 . MRI has been proposed as a tool to evaluate to what extent an intervention may be effective 39 since patients with severe cortical structural atrophy may not benefit from retinal input restoration 40 . Also, MRI studies allow brain structural and functional assessment after retinal implants, enabling pre-and post-intervention disease management 41 . Our results further support this possibility for RP patients with low vision.
In summary, we showed striate and extrastriate visual cortical GM reduction in RP patients with low vision, associated with the extent of visual field loss.

Methods
Subjects. Twenty-eight Retinitis Pigmentosa (RP) patients were recruited from the Ophthalmology Department at the University of Coimbra Hospital, Portugal. Forty-three age-and gender-matched controls were recruited from the volunteers' database of the Institute for Biomedical Imaging and Life Sciences (IBILI) of the Faculty of Medicine, University of Coimbra, Portugal. All participants underwent an ophthalmologic examination (visual acuity assessment, automated perimetry, and optical coherence tomography [OCT]) and a structural magnetic resonance imaging (MRI) acquisition.
Patients with other neuro-ophthalmological or organic brain disorder were excluded from the study. One patient could not perform the MRI scan due to claustrophobia, being consequently excluded. Twenty-seven RP patients (12 females, 15 males; age range 20-66 years) were finally included ( Table 2).
None of the controls participants had a history of neurological or organic brain disorders. Exclusion criteria were: alterations in the ophthalmologic examination (automated perimetry and OCT), eye disorders known a priori, and alterations in the MRI structural images. Controls were required to have normal or corrected to normal vision (visual acuity ranged from 0.80 to 1.33 decimals [−0.10 to 0.12 logMAR]). Four control subjects were excluded due to alterations on the ophthalmological examination and another participant was excluded due to excessive movement in the MRI acquisition. Thirty-eight healthy controls (18 females, 20 males; age range 22-74 years) were included.
The study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Commission of Faculty of Medicine of University of Coimbra, Portugal. Written informed consent was obtained from all participants.
Ophthalmic examination. All participants underwent an ophthalmological examination including visual acuity assessment, static visual field evaluation with automated perimetry, and macular and retinal nerve fiber layer (RNFL) thickness measurements with OCT. All assessments were performed in each eye separately. Ophthalmological examinations were evaluated by an experienced ophthalmologist (ES) to rule out eye disorders in the control group.  Visual acuity was determined for each participant using a decimal chart (VA-Snellen chart) and the values were converted to the logarithm of Minimum Angle of Resolution (logMAR) scale before performing the statistical analysis. The right eye and left eye were evaluated separately with the same chart. When patients' visual acuity was so low that they could not read any of the largest letters or even count fingers, visual capacity was assessed with hand motion visual or light perception testing. The values for visual acuity in these patients (subjects 3 and 18 in Table 2) were considered to be 0.05 decimals (−1.30 logMAR).

Number of voxels
Static visual fields were evaluated using a MonCv3 multifunction perimeter (Metrovision, France) with a standardized program (except for 4 patients, who could not perform the exam for both eyes -subjects 3, 16, 18, and 24 in Tables 2 and 1 patient, who was unable to perform the exam using the left eye -subject 6 in Table 2). These RP patients could not perform the examination due to low visual acuity (inferior or equal to 0.33 decimals/−0.48 logMAR). A suprathreshold strategy (4 dB above the theoretical luminance threshold) was used for the 79 points tested in the central 24 degrees (radius) using a rapid thresholding strategy. When subjects could not execute the test due to a narrow visual field, a central 12 degrees (radius) assessment was performed. Testing was always carried out with appropriate refractive correction. Fixation was monitored in real time with an integrated camera by the operator. The static visual fields of 1 patient (subject 13 in Table 2) were evaluated with Humphrey Field Analyzer (Carl Zeiss Meditec, Germany) for the 10 central degrees due to unavailability of the MonCv3 multifunction perimeter on the day of examination. Visual field extent was then estimated for each eye of each patient in angles of visual degree (diameter). The maximum diameter of visual field was approximately estimated by defining the absolute scotoma as the black region (lower than 10 dB of sensitivity) obtained in the static perimetry sensitivity map (Fig. 3). Visual field mean deficit (in dB) was also obtained.
Frequency Domain Cirrus OCT (software version 5.1.1.6, Carl Zeiss Meditec AG, USA) was performed to measure retinal thickness and retinal nerve fiber layer (RNFL) thickness in all participants, except in 1 patient due to the inaccessibility of the device (subject 23 in Table 2). A macular scan with a resolution of 512 × 128 and an optic disc scan with a resolution of 200 × 200 were acquired. Age of symptoms onset was self-reported by patients to determine the disease duration (Table 2).

MRI acquisition.
High-resolution brain anatomical images were acquired on a 3 T scanner (Magneton TrioTim, Siemens AG, Germany), using a 12-channel birdcage head coil. One 6-min T 1 -weighted magnetization-prepared rapid acquisition with gradient echo ( Statistical analysis of demographic and visual data. The statistical analysis was performed with IBM SPSS Statistics (version 23.0, IBM Corporation, USA). Continuous variables were tested for normality, separated by group, using the Kolmogorov-Smirnov test. Differences between the RP and control groups for non-normally distributed variables were tested using the Mann-Whitney U test (age, visual acuity, and visual field's mean deficit), while normally distributed variables were compared using the independent-samples t-test (retinal thickness and RNFL thickness). Gender and handedness comparisons between groups were performed using the Pearson's chi-squared test (χ 2 ). Comparisons between the left and right eye within each group were performed using the nonparametric Wilcoxon signed-rank test or the parametric paired-samples t-test. Correlations between disease duration and visual field diameter were investigated within the RP group using the non-parametric Spearman test, given the non-normality distribution of the corresponding variables. All p values below 0.05 were considered statistically significant.
VBM data analysis. Image processing. T 1 -weighted images were processed using the SPM12 version (Statistical Parametric Mapping, http://www.fil.ion.ucl.ac.uk/), compiled with the MATLAB version R2014a (The MathWorks, Inc., USA). Firstly, every scan was aligned to the anterior commissure manually. Then, the unified segmentation algorithm 42 was used to normalize, segment, and modulate the images. All T 1 -weighted images were registered to the Montreal Neurological Institute (MNI) coordinate system by registering the MRI images to the ICBM 152 template. Normalized images were segmented into gray matter (GM), white matter, and cerebrospinal fluid. GM images were modulated to correct for GM volume changes due to nonlinear registration. Additionally, the images were smoothed with a Gaussian kernel with 8 mm of full width at half maximum (FWMH) to ensure the normality of the data. Finally, voxel wise statistical analyses were applied to these processed GM images.
Significant clusters were overlaid in a T1 image of a random healthy subject registered to the MNI space for visualization purposes. This image can be found in the SPM folder. Anatomical labeling was performed in accordance with the work of Tzourio-Mazoyer et al. 43 using the xjview8 toolbox.
Statistical analysis. GM volume alterations were assessed by comparing RP patients with controls, using the standard general linear model (GLM) implementation in SPM12 for independent two-samples t-tests. Statistical parametric maps of two different contrasts (Controls > RP and RP > Controls) were created by applying a significant threshold of p < 0.01 with FDR correction for multiple comparisons and a cluster size of 20 voxels. Even though participants were age-and gender-matched (see the Results section), age and gender were included in the model as nuisance covariates. The total intracranial volume (TIV), known to be an important confound in VBM studies 44 , was also included as a nuisance variable. Nonetheless, no significant differences were found between groups for the TIV [independent samples t-test t (64) = −6.12, p = 0.543]. The TIV was calculated in MATLAB using the automated method as previously reported 45 .
To explore possible correlations between GM volume and patient's visual characteristics (visual field diameter, visual field mean deficit, visual acuity, retinal thickness, and RNFL thickness) as well as disease duration, post-hoc whole brain SPM multiple regression analyses were run. Since no statistical differences were found between the left and right eyes for all visual parameters within the RP group (see the Results section), left-and right-eye clinical parameters were averaged for the subsequent analysis to reduce the number of statistical comparisons. In this way, we excluded the 5 RP patients with missing data from both eyes when running the correlation analyses (subjects 3, 16, 18, 23, and 24 in Table 2). Again, age, gender, and TIV were added as nuisance variables to the regression model, and statistical parametric maps of direct (positive) or inverse (negative) correlation with visual field diameter, visual field mean deficit, visual acuity, retinal thickness, RNFL thickness or disease duration were computed using a p-value threshold (<0.001) with a cluster correction of 150 voxels.