Structural and microvascular changes of the peripapillary retinal nerve fiber layer in Von Hippel–Lindau disease: an OCT and OCT angiography study

Von Hippel–Lindau (VHL) disease is an autosomal dominant genetic disease caused by VHL gene mutation. Retinal hemangioblastomas (RH) are vascularized tumors and represent the main ocular manifestation of the disease. Histopathologically, RH are composed of capillary vessels and stromal cells, the neoplastic population of the lesion. The origin of these stromal cells remains controversial, even if they are hypothesized to be glial cells. The aim of the present study was to investigate neuronal and microvascular changes of the peripapillary retinal nerve fiber layer, in which glial cells, neurons and capillaries (the radial peripapillary capillary plexus) interact. VHL patients with or without peripheral RH were enrolled and compared to healthy controls. Mean peripapillary retinal nerve fiber layer (pRNFL) thickness was measured by means of optical coherence tomography (OCT). The following vascular parameters of the radial peripapillary capillary plexus were quantified using OCT angiography: Vessel Area Density,Vessel Length Fraction, Vessel Diameter Index and Fractal Dimension. One hundred and nine eyes of 61 patients, and 56 eyes of 28 controls were consecutively studied. Mean pRNFL was significantly thinner in VHL eyes without RH versus eyes with RH and controls. Mean pRNFL thickness did not differ between VHL eyes with RH and controls. All OCTA vascular parameters were reduced in VHL eyes with or without RH versus controls, with significative difference for Vessel Diameter Index. The same OCTA parameters did not significantly differ between VHL eyes with or without RH. In VHL eyes without RH, pRNFL thinning may be the consequence of impaired perfusion of the radial peripapillary capillary plexus, while the increase of pRNFL thickness in VHL eyes with RH may depend on possible activation and proliferation of the other RNFL resident cells, the glial cells.

OCTA vascular parameters did not significantly differ between VHL eyes with or without RH. (Table 2).

Discussion
In the present study, the peripapillary retinal nerve fiber layer was thinner in VHL eyes without RH, both compared to controls and to eyes with RH. At OCTA analysis of the radial peripapillary capillary plexus, the Vessel Diameter Index, which refers to the diameter of the vessels 12 , was significantly reduced in VHL eyes compared to healthy controls, independently of the presence or absence of RH, showing a plexus characterized by analomalous thin capillaries. The radial peripapillary capillary plexus is a distinct retinal vascular network, located in the pRNFL, characterized by a specific radial distribution of capillaries, tightly coupling and suppling the nerve fibers 11,13,14 . Retinal blood flow is highly correlated to neural activity, in a strict neurovascular metabolic coupling 15 . Therefore, the reduced perfusion of the this plexus may be correlated to the reduction in thickness of the RNFL, as occurs in other retinal diseases like diabetes and glaucoma, whose neurodegenerative nature is well demonstrated 13,[16][17][18] . In a retina-specific knockout VHL animal model, excessive vessel reduction and regression have already been found 19 . Therefore, an impaired radial peripapillary capillary plexus, characterized by reduction in peripapillary perfusion, may hypothetically cause reduction of retinal nerve fibers. www.nature.com/scientificreports/ On the other hand, in VHL eyes with RH, showing the same radial peripapillary capillary plexus perfusion impairment, pRNFL was thicker than in eyes without RH. Therefore, radial peripapillary capillary plexus changes may be excluded in the pathophysiology of this relative thickening. The retinal nerve fiber layer is composed not only by nerve fibers and vessels, but also by glial cells, mostly Müller cells and astrocytes 20,21 (from 18 to 42%). Moreover, the thickness of pRNFL has been shown to be linearly correlated to the number of retinal astrocytes 20 . The role of astrocytes in VHL disease remains controversial, but histopathologic studies have already revealed trasformed and increased astrocytes in the stromal populations of hemangioblastomas, including the retinal ones 5,9,10,[22][23][24] . Under normal conditions, astrocytes maintain the homeostasis of the nervous tissue -controlling, protecting and supporting neuronal function-and regulate local blood flow 25 . Under pathological conditions, atrocytes react through a process called reactive astrogliosis, with hyperplasia/proliferation of the cells, increased number of their processes and larger cell body size 25,26 . In an experimental model, the deletion of VHL gene in retinal atrocytes induced not only increased astrocyte cell number in the RNFL, but also upregulation of retinal VEGF production, which is a key factor in the pathophisiology and growth of hemangioblastomas 27,28 . Therefore, the relative increase in pRNFL thickness we detected in eyes with RH and not in eyes without RH, represents in our view human in vivo clinical evidence of local activation of retinal astrocytes with consequent upregulation of retinal VEGF production, previoulsy demonstrated in animal models and in hystopathological studies.
In VHL patients, OCT and OCTA have been mainly used to study posterior pole RH or macular capillary changes related to RH [29][30][31][32] . To the best of our knowledge, this is the first OCT and OCTA study investigating the structural and perfusion characteristics of the peripapillary nerve fiber layer, in which glial cells, neurons and capillaries of the radial peripapillary capillary plexus interact.
Genotype-phenotype correlations were not investigated in the present research. This could represent a limit of the study, since we cannot exclude that the genotype of VHL germline mutation may differently influence the analyzed parameters, as higher prevalence of progression-related complications and preretinal fibrosis have already been correlated to it 33,34 . A further analysis of genotype versus OCT and OCTA parameters is ongoing.  www.nature.com/scientificreports/ In conclusion, in VHL eyes, pRNFL is thinner than in controls hypothetically as a consequence of reduced perfusion of the radial peripapillary capillary plexus, as shown in other retinal disorders. When peripheral RH are present, pRNFL thickness relatively increases compared to eyes without RH. This fact may be due to the proliferation of retinal astrocytes, as previously histologically reported. This observation may offer a holistic interpretation of the involvement of different retinal cells in VHL progressive ocular disease. A clinical application of our findings may be that, in the follow-up of VHL patients, the increase of pRNFL thickness, compared to baseline, may guide retinal examination for peripheral RH screening. Larger population and longer follow up are mandatory to confirm our observations.

Methods
This was a cross-sectional study and informed consent was obtained from each subject; data collection was compliant with the tenets of the Declaration of Helsinki. The approval from the Ethics Committee for Clinical Practice of the Azienda Ospedaliera di Padova for the study was obtained (Prot.34971/AOP/2018). VHL patients, referred by the Familial Tumor Unit of the Veneto Institute of Oncology (IOV-IRCCS), underwent scheduled eye examination including: best-corrected visual acuity (BCVA) measurement using standard Early Treatment Diabetic Retinopathy Study (ETDRS) charts, anterior segment evaluation at slit lamp, intraocular pression measurement, fundus examination by means of indirect ophthalmoscopy and 90-diopter-lens biomicroscopy. Patients who presented congenital or acquired anterior segment diseases; epiretinal macular membrane; posterior pole and optic nerve -treated or untreated RH; glaucoma; refractive errors ≥ 6 diopters; history of ocular inflammation; concomitant presence of other retinal vasculopathies were not included in the research. The posterior pole and optic nerve RH were excluded because they may interfere with the correct visualization and quantification of RNFL and vascular parameters. Moreover, patients with pituitary stalk hemangioblastomas, which may compromise optic nerve pathway, were also excluded 35 . An age-matched healthy group was recruited for comparison.
OCT and OCT angiography. After pupil dilatation (obtained with 1% tropicamide eye-drops solution), enrolled eyes performed OCT and OCTA using Spectralis HRA + OCTA (Heidelberg Engineering, Heidelberg, Germany) in a semi-dark room, late in the morning and with the in-built eye-tracker always active to obtain high quality scans and correct foveal centration of the scans. The OCT scan protocol included a peripapillary 3.5 mm ring N-RNFL (N-site/axonal menu) scan, centered onto the optic nerve head, to automatically provide the peripapillary retinal nerve fiber layer (pRNFL) thickness value. pRNFL thickness was expressed both as mean pRNFL (pRNFL) and as sectorial pRNFL (Temporal, T-pRNFL; Temporal Superior, TS-pRNFL; Nasal Superior, NS-pRNFL; Nasal, N-pRNFL; Nasal Inferior, NI-pRNFL; Temporal Inferior, TI-pRNFL; Papillo-Macular Bundle, PMB).
An OCTA scan pattern of 10° × 10° (3.0 × 3.0 mm; 512 B-scans separated by 6 micron) centered onto the optic nerve head was acquired. The inbuilt software automatically generated the enface OCTA image of the radial peripapillary capillary plexus, which is the radial capillary monolayer network encircling the optic disc 11,13 . As previously described 29 in order to guarantee high quality OCTA images, only images with a signal strength more than 30 in "Q score" (on a scale of 0 to 40 for Spectralis, Heidelbeerg), were analysed 36,37 . Correct foveal centration of the scans was obtained with the inbuilt eye tracking system. A skilled technician checked each image after acquisition to detect any motion artifacts or segmentation errors, and eventually repeated examination. Open-source available ImageJ software (National Institutes of Health, Bethesda, MD, USA) allowed quantitative analysis of the OCTA en face images, obtaining quantitative parameters of the radial peripapillary capillary plexus: Vessel Area Density (VAD), Vessel Length Fraction (VLF), Vessel Diameter Index (VDI) and Fractal Dimension (FD), as previously described 12,38 . OCTA en face images were automatically converted into a binary image. VAD, which provides the estimate of real vessel density in the examined area, taking into consideration both vessel length and vessel diameter, was obtained dividing the number of black pixels in the binary image by the total number of image pixels. The binarized image was then "skeletonized" deleting the outer boundary of the binarized image so that each vessel segment had a single pixel width. By dividing the number of vessel pixels in the skeletonized image by the total number of image pixels VLF was obtained, quantifing the vessel length regardless of the vessel diameters. VDI parameter, sensitive to vascular dilation in the OCT-A images as it presents the vessel size information regardless of the vessel length, was obtained by processing both the binary and the skeletonized images to calculate the average vessel caliber. FD, representing complexity of images, was calculated both over the skeletonized images (FDsk) using a box counting technique and over binary images (FDbin) 12,38 . Statistical analysis. All variables were summarized according to the conventional methods of descriptive statistics: mean and standard deviation for quantitative variables; absolute and relative (percentage) frequencies for qualitative variables. Normal distribution of parameters was checked by Shapiro-Wilk's test.
Gender distribution and age of enrolled patients and controls were compared using chi-square test and t-Student test for independent samples, respectively. Mean pRNFL and sectorial pRNFL thickness values were analyzed with one-way ANOVA model. When applicable, statistical models were adjusted for replication of measures in both eyes of patients and controls.
A mixed effects ANOVA model was applied to compare OCTA parameters between groups, with repeated measures (both eyes), and adjusted for age. Statistical significance was set at p < 0.05 level. All the analyses were performed by SAS® 9.4 statistical software (SAS Institute, Cary, NC, USA).

Data availability
The authors have full access to all the data and take responsibility for the integrity of the data and the accuracy of the data analysis as well as the decision to submit to publication.