*Article* **Retinal Venular Tortuosity Jointly with Retinal Amyloid Burden Correlates with Verbal Memory Loss: A Pilot Study**

**Oana M. Dumitrascu 1,\* , Ryan Rosenberry <sup>2</sup> , Dale S. Sherman <sup>3</sup> , Maziyar M. Khansari <sup>4</sup> , Julia Sheyn <sup>5</sup> , Tania Torbati <sup>5</sup> , Ayesha Sherzai <sup>6</sup> , Dean Sherzai <sup>6</sup> , Kenneth O. Johnson <sup>7</sup> , Alan D. Czeszynski <sup>7</sup> , Steven Verdooner <sup>7</sup> , Keith L. Black <sup>5</sup> , Sally Frautschy <sup>8</sup> , Patrick D. Lyden <sup>9</sup> , Yonggang Shi <sup>4</sup> , Susan Cheng <sup>2</sup> , Yosef Koronyo <sup>5</sup> and Maya Koronyo-Hamaoui 5,10,\***


**Abstract:** Introduction: Retinal imaging is a non-invasive tool to study both retinal vasculature and neurodegeneration. In this exploratory retinal curcumin-fluorescence imaging (RFI) study, we sought to determine whether retinal vascular features combined with retinal amyloid burden correlate with the neurocognitive status. Methods: We used quantitative RFI in a cohort of patients with cognitive impairment to automatically compute retinal amyloid burden. Retinal blood vessels were segmented, and the vessel tortuosity index (VTI), inflection index, and branching angle were quantified. We assessed the correlations between retinal vascular and amyloid parameters, and cognitive domain Z-scores using linear regression models. Results: Thirty-four subjects were enrolled and twenty-nine (55% female, mean age 64 ± 6 years) were included in the combined retinal amyloid and vascular analysis. Eleven subjects had normal cognition and 18 had impaired cognition. Retinal VTI was discriminated among cognitive scores. The combined proximal mid-periphery amyloid count and venous VTI index exhibited significant differences between cognitively impaired and cognitively normal subjects (0.49 ± 1.1 vs. 0.91 ± 1.4, *p* = 0.006), and correlated with both the Wechsler Memory Scale-IV and SF-36 mental component score Z-scores (*p* < 0.05). Conclusion: This pilot study showed that retinal venular VTI combined with the proximal mid-periphery amyloid count could predict verbal memory loss. Future research is needed to finesse the clinical application of this retinal imaging-based technology.

**Keywords:** retinopathy; retinal vessels; retinal fluorescence imaging; amyloid; cognitive decline; Alzheimer's disease

**Citation:** Dumitrascu, O.M.; Rosenberry, R.; Sherman, D.S.; Khansari, M.M.; Sheyn, J.; Torbati, T.; Sherzai, A.; Sherzai, D.; Johnson, K.O.; Czeszynski, A.D.; et al. Retinal Venular Tortuosity Jointly with Retinal Amyloid Burden Correlates with Verbal Memory Loss: A Pilot Study. *Cells* **2021**, *10*, 2926. https:// doi.org/10.3390/cells10112926

Academic Editors: Maurice Ptito and Joseph Bouskila

Received: 29 June 2021 Accepted: 25 October 2021 Published: 28 October 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

#### **1. Introduction**

By 2025, the number of people aged 65 years and older with Alzheimer's dementia (AD) is projected to reach 7.1 million, which is almost a 22% increase from 2020 [1,2]. The contribution of vascular disease to cognitive performance is increasingly recognized, as the mechanisms linking vascular dysfunction and neurodegeneration are better characterized [3–7]. Recent reports implicate cerebral vascular pathology as an early and core contributor to the development of not only vascular dementia but also AD [7–11], a neurodegenerative condition and looming public health threat [2,12]. Considering the emerging vascular hypothesis [8,13–15], there is a critical need to incorporate both vascular and AD biomarkers [16–20] into predictive models to allow for early and sensitive detection of AD and mixed dementias. Yet, imaging of the skull-shielded brain poses various limitations for widespread screening in the clinical setting. The retina is a central nervous system organ that exhibits Aβ deposition and vascular changes [21–33] and is far more accessible for repeated and high-resolution imaging [34–42]. Dysfunctional pericytes in the blood-brain barrier (BBB) are significant contributors to the pathogenesis of vascular cognitive impairment, including cerebral small vessel and cerebral large vessel disease, as well as AD [28,43]. BBB pericyte injury is a predictor of apolipoprotein E (APOE) ε4-associated cognitive decline [4]. In contrast, BBB dysfunction mediates cerebral Aβ deposition, the retinal–blood barrier mirrors the BBB, and its disruption in the form of retinopathy was shown to predict cognitive decline [28,44–49]. Post-mortem retinal vessels derived from patients with mild cognitive impairment (MCI) and AD exhibited early and progressive pericyte loss as well as Aβ accumulation inside retinal pericytes, correlating with similar AD pathology in the brain [28]. Several studies demonstrated the linkage between retinal vascular fractal dimensions, caliber, and both tortuosity and cognitive deterioration [45–51]. The retinal arteriolar central reflex to vessel width ratio in digital retinal photographs was significantly higher in APOE ε4 allele carriers [48], hence the retina may allow for non-invasive monitoring of the effects of APOE ε4 on the cerebrovascular disease. Similarly, as targeting vascular risk factors is being considered in dementia prevention trials [52], retinal vascular assessments could offer a window for assessing the response to various interventions.

Recent work has highlighted the promising utility of retinal fluorescence imaging, an emerging technique capable of non-invasively imaging and quantifying the retinal amyloid, which is the pathological marker of AD [22,23,26,34,53–55]. Using this technique, our group previously identified a significant association between retinal amyloid count, especially in the proximal mid-periphery area, and the severity of cognitive impairment as well as hippocampal volumes [34,35]. As the same retinal imaging modality also allows for retinal vasculature analysis, we aimed to quantitatively examine both retinal vascular and retinal amyloid biomarkers in a cohort of subjects with cognitive decline. In this proof-of-concept exploratory study, we sought to examine the relationship between retinal microvascular features and retinal amyloid burden, with global and domain-specific cognitive scores.

#### **2. Materials and Methods**

#### *2.1. Participants*

This pilot study was approved by the Cedars-Sinai Institutional Review Board. All subjects older than 40 years of age presenting to our Neurology clinic with subjective cognitive decline and interest in undergoing retinal fluorescence imaging were included in this cohort. All subjects underwent a neurological examination, a standard battery of neuropsychological tests, and standard-of-care 3 Tesla non-contrast structural brain magnetic resonance imaging (MRI). No exclusion criteria were prespecified, except for a history of glaucoma, allergy to mydriatic eye drops, curcumin, or vitamin E. All subjects provided written informed consent prior to the commencement of the study.

#### *2.2. Retinal Imaging*

After ocular dilation, the retinal imaging was performed with a confocal scanning ophthalmoscope (RetiaTM, CenterVue SpA) that utilizes blue light for the excitation of curcumin emission to obtain fluorescent images of the retina, following a study design described in prior reports (Figure 1A) [34,35]. Curcumin has high affinity and specificity for the β-pleated sheets of Aβ, specifically for Aβ42, oligomers, and fibrils, which are linked to AD [56–61]. The researchers conducting the retinal image processing and quantifications were blinded to the patients' clinical characteristics. β β β

*Cells* **2021**, *10*, x FOR PEER REVIEW 3 of 16

**Figure 1.** Study timeline of brain and retinal imaging followed by sectoral amyloid and vascular analysis. Study design scheme illustrating that subjects underwent baseline brain imaging and neuropsychological evaluation, followed by retinal fluorescence imaging after 4 days of daily oral curcumin intake (**A**). Illustration of the region of interest in the right eye supero-temporal retinal quadrant and its three subregions, which were used for quantifying retinal amyloid counts (**B**). Illustration of the region of interest used for the retinal vascular analysis. The red circle indicates the center of the optic nerve-head and the smallest yellow circle shows the optic nerve-head area. The two larger circles indicate the region of interest for the vascular analysis, which were 1.5 and 4 times the diameter of the optic disc. The branching angle and tortuosity of vessels within the region of interest were calculated. Arteries and veins are outlined by red and blue lines, respectively (**C**). Graphs illustrating differences in total amyloid (**D**) and proximal amyloid counts (**E**) when stratified by cognitive status. Graphs illustrating the differences between arterial branching angle (**F**) and the venous tortuosity index (**G**) when stratified by CDR. \* *p* < 0.05; \*\* *p* < 0.01, by two-tailed unpaired student *t*-test or one-way ANOVA and Bonferroni's post-hoc test. Abbreviations: MRI, magnetic resonance imaging; PP, posterior pole; PMP, proximal mid-periphery; DMP, distal mid-periphery; ODD, optic disc diameter; CDR, Clinical Dementia Rating; and VTI, vessel tortuosity index.

#### *2.3. Retinal Amyloid Quantification*

The set of retinal images were processed using an automated retinal fluorescence measurement software system (NeuroVision Imaging, Inc., Sacramento, CA, USA). A combination of algorithms, including background correction, followed by characterization of the corrected retina using a mixture model, were used to identify pixels that were abnormally bright. Specifically, the primary factor of the variation in pixel intensity is illumination variability across the entire field of view (e.g., edges of the image become dark). This variability is addressed by estimating the background level and correcting it. The secondary factor in pixel variability is the structure to which it depends on. Vessels appear dark or hypofluorescent, while the amyloid appears identically as bright or

hyperfluorescent. The background correction produces all vessels at a more consistent pixel value. In a similar manner, this occurs for the retina and amyloid spots. Choosing the appropriate threshold is possible using the mixture model, which characterizes hypofluorescent, isofluorescent, and hyperfluorescent pixels appropriately. A common region of interest (ROI) in the supero-temporal quadrant was applied with a field of view of 50 degrees, positioned on the image center using fovea and optic nerve-head centers as reference points to correct for eye rotation, with a zone around the fovea and optic nerve-head masked, as previously reported [35]. The ROI was further divided into three subregions: posterior pole, proximal mid-periphery, and distal mid-periphery (Figure 1B). Retinal amyloid count was quantified in the target ROI and three specified subregions.

#### *2.4. Retinal Vascular Quantification*

From the same retinal fundus images, an ROI was defined within a circumpapillary region centered on the optic nerve-head (ONH) and extending between 1.5 and 4 ONH radii (Figure 1C) [62]. Before the analysis, retinal images were visually inspected to ensure vessels were visible and that there was no reflectivity that could influence the result. The major vessels were detected after intensity normalization to minimize the effect of other influencing factors. Retinal vessels within the ROI were segmented using the Frangi vesselness filter to generate a binary image [63]. The vessels were classified into arteries and veins by a human observer based on the facts that retinal arteries are brighter in color and thinner in width compared to veins [64]. For each vessel segment on the binary image, vessel endpoints were selected, and distance transformation was used to extract the vessel centerline. The extracted centerlines were smoothed using a cubic spline with a regularization parameter of 3 × 10−<sup>5</sup> . For each centerline, several geometric features, including the vessel tortuosity index (VTI), vessel inflection index, and branching angle, were non-automatically quantified. The VTI was calculated for each centerline based on a combination of local and global centerline geometric variables, as explained previously, that can detect alterations in the retinal vessels' curvature with pixel-level accuracy [65]. Equation (1) shows the formula for the VTI.

$$\text{VTI} = 0.1 \times \left( \text{SD}\_{\text{\Theta}} \text{N.M.} \frac{\text{L}\_{\text{A}}}{\text{L}\_{\text{C}}} \right) \tag{1}$$

where SDθ the is standard deviation of the angle difference between lines tangent to each centerline pixel and a reference axis (i.e., x-axis), and M is the average ratio of the centerline length to its chord length between pairs of inflection points, including centerline endpoints. N is number of critical points where the first derivative of the centerline vanishes, while L<sup>A</sup> and Lc are the length of the vessel and its chord length, respectively. The VTI is shown to provide good correspondence with human perception of tortuosity and is invariant to rigid transformations. Similar to other measures of tortuosity, VTI is unitless. Its minimum value is zero, while it has no theoretical maximum as it can increase with the twistedness of a vessel. The vessel inflexion index was determined based on a number of inflection points along the vessels. Mathematically, these were pixels where the second derivative of the centerline vanishes. The vessel inflexion index represents local changes in the tortuosity of vessels and was found to be robust for ranking the tortuosity of vessels with similar lengths [66]. The branching angle of the vessels was calculated interactively using the open-source tool GIMP 2.8.

#### *2.5. Cognitive Evaluation*

All participants underwent a standard battery of neuropsychometric testing performed by a licensed neuropsychologist (DS). Standard neuropsychological testing included the Montreal Cognitive Assessment (MOCA), global Clinical Dementia Rating (CDR), as well as general cognitive (ACS-test of Premorbid Functioning) and specific cognitive domain assessments: attention and concentration (Wechsler Adult Intelligence Scale (WAIS)-IV); verbal memory (California Verbal Learning Test (CVLT) II, Wechsler Memory

Scale (WMS)-IV, and Logical Memory II); non-verbal memory (Rey Complex Figure Test and Recall (RCFT) 30 min, and Brief Visuo-Spatial Memory Test Revised (BVMT-R) Delayed Recall); language (Fluency-Letter (FAS) and Fluency-category (animals)); visuo-spatial ability (Rey Complex Figure Test and Recognition Trial (RCFT) Copy); speed of information processing (Trails A and B); and symptom validity and functional status (SF-36 Physical Component Score (PCS) and Mental Component Score (MCS)). We also evaluated the subject's emotional status using the Beck Depression Inventory II, Geriatric Depression Scale, and Profile of Mood State/Total Mood Disturbance.

#### *2.6. Statistical Analysis*

Descriptive statistics were calculated for patient demographics and clinical characteristics. Unless otherwise stated, data are expressed as mean ± standard deviation. Subjects were partitioned into three groups according to the Clinical Dementia Rating (CDR) (0.5, questionable impairment; 1, mild cognitive impairment; and 2, moderate cognitive impairment) [67] and dichotomized using MOCA, which demonstrates excellent sensitivity and specificity for both mild cognitive impairment (MCI) and AD. Using the cutoff score of <26, the MOCA has excellent sensitivity for MCI (90%) and AD (100%), as well as for the specificity for normal controls (87%). Positive (PPA) and negative predictive accuracy (NPA) were also reported to be excellent with a PPA of 89% and NPA of 91% for MCI, and a PPA of 89% and NPA of 100% for AD [68]. The subjects were also partitioned into groups according to the neuropsychometric diagnosis (normal cognition versus impaired cognition).

To produce combined indices of retinal vascular and amyloid measures, each variable was first inspected for normality; any non-normal variables were then log-transformed to produce a normal distribution. Each normalized variable was then standardized to a mean of 0 and a standard deviation set equal to 1. While higher amyloid count was associated with worse cognitive function, higher venous vascular tortuosity index (VTI) values were associated with better cognitive function. To account for this inverted scale, the standardized values of venous VTI were multiplied by −1. Standardized variables were then summed to produce exploratory, combined index measures of retinal amyloid and retinal vascular features.

Differences in continuous variables between levels of CDR were assessed through one-way analysis of variance (ANOVA), with Bonferroni's post-hoc test for the correction of multiple comparisons. Differences in the continuous variables between diagnostic scores were assessed using Student's *t*-test. Linear regression was performed to assess the relationship between retinal vascular and retinal amyloid measures, as well as to assess the relationship between combined retinal vascular and amyloid counts, and cognitive parameters. All statistical analyses were performed using STATA v15.1 (StataCorp, College Station, TX, USA) with an a priori significance level of 0.05.

#### **3. Results**

Our study included a total of 34 subjects that presented to our Neurology clinic with cognitive concerns. Out of those 34, 29 had retinal images of sufficient quality to undergo both retinal amyloid and vascular analysis; their demographics and preexisting conditions are shown in Table 1. Mean MOCA was 26 (range of 4–32) and median MOCA was 27. Eleven subjects had a CDR of 0.5, 15 had a CDR of 1, and 3 had a CDR of 2. Regarding the formal neuropsychometric cognitive evaluation, 11 (37.93%) patients had normal cognition and 18 (62.06%) had impaired cognition (six with amnestic MCI, nine with multidomain MCI, two probable AD cases, and one with possible fronto-temporal lobar degeneration).

Linear regression analyses revealed that the venous branching angle correlated with the distal mid-periphery amyloid count (*p* = 0.03) and the arterial inflexion index correlated with the posterior pole amyloid count (*p* = 0.02). There were no associations between retinal vascular parameters and amyloid count in the proximal mid-periphery (Table S1).


**Table 1.** Demographics and medical history of subjects in the combined retinal vascular and retinal amyloid analysis.

The analysis of retinal vascular and amyloid measures according to strata of cognitive function showed that the retinal PMP amyloid count and total amyloid count were significantly higher in the cognitively impaired compared to normal cognition participants (PMP: 144 ± 52 vs. 85 ± 32, *p* = 0.0012; total: 343 ± 90 vs. 247 ± 82, *p* = 0.04; Figure 1D,E and Table 2). There was no significant difference in the venous branching angle (*p* = 0.98) or arterial VTI (*p* = 0.53) across levels of CDR, whereas the arterial branching angle reached near significance (*p* = 0.066; Figure 1F). Venous VTI was significantly different across levels of CDR (mean ± SD of venous VTI values across increasing CDR categories: 0.13 ± 0.02, 0.13 ± 0.02, and 0.09 ± 0.02; *p* = 0.026; Figure 1G). Given these group differences and because of the independence of retinal vascular and retinal amyloid measures, the following combined amyloid-vascular indexes were calculated as exploratory variables: proximal mid-periphery amyloid count-venous VTI, total amyloid count-venous VTI, proximal mid-periphery amyloid count-arterial branching angle, and total amyloid count-arterial branching angle. One-way ANOVA revealed significant group differences in the VTI indices when compared according to the CDR level (Figure 2A–D). The combined proximal mid-periphery amyloid-venous VTI index was the only combined index measure exhibiting significant group differences when the cognitively impaired were compared to the cognitively normal subjects (0.49 ± 1.1 vs. −0.91 ± 1.4, *p* = 0.006; Figure 2E and Table 2).

**Table 2.** Vascular and amyloid parameters stratified by the cognitive status.



**Figure 2.** Combined retinal amyloid and vascular parameters in patients stratified by cognitive scores. Graphs illustrating the differences in the combined proximal mid-periphery amyloid-arterial branching angle index (**A**), total amyloid-arterial branching angle index (**B**), proximal mid-periphery amyloid-venous tortuosity index (**C**), and total amyloid-venous tortuosity index (**D**) when stratified by CDR score. Graphs illustrating the differences between the combined proximal mid periphery amyloid-venous tortuosity index (**E**) and total amyloid-venous tortuosity index (**F**) when stratified by the cognitive status. Bar graphs show the mean and deviation (\* *p* < 0.05 and \*\* *p* < 0.01 by two-tailed paired Student's *t*-test). Abbreviations: VTI, vessel tortuosity index; ABA, arterial branching angle; PMP, proximal mid-periphery; and CDR, Clinical Dementia Rating.

− −

− −

−

−

− −

−

−

−

We performed regression analyses to evaluate the correlations between retinal vascular geometric parameters and retinal amyloid counts with cognitive domain Z-scores. We found that the venous branching angle correlated with the WAIS-IV-digit span Z-score (Beta −0.045 (SE 0.015), *p* = 008). The total amyloid count correlated with the SF-36-MCS Z-score (Beta −0.004 (SE 0.002), *p* = 0.046), whereas the proximal mid-periphery amyloid count correlated with two verbal memory measures, namely CVLT-II Long Delay (Beta −0.009 (SE 0.003), *p* = 0.027) and WMS-IV LM-II (Beta −0.007 (SE 0.003), *p* = 0.028). The distal mid-periphery amyloid count correlated with non-verbal memory, RCFT Delayed Recall (Beta −0.01 (SE 0.005), *p* = 0.04), and SF-36-MCS (Beta −0.014 (SE 0.004), *p* = 0.004; Table 3).

**Table 3.** Retinal vascular and amyloid parameter predictors of cognitive domain measures.


Abbreviations: VTI, vessel tortuosity index; PMP, proximal mid-periphery; DMP, distal mid-periphery; WAIS, Wechsler Adult Intelligence Scale; CVLT, California Verbal Learning Test; WMS LM-II, Wechsler Memory Scale Logical Memory II; RCFT, Rey Complex Figure Test and Recall; MCS, Mental Component Score; and Std. Err, standard error.

> The combined proximal mid-periphery amyloid-venous VTI index correlated with both verbal memory performance Z-scores (WMS-IV LM-II (Beta −0.537 (SE 0.138), *p* = 0.001) and CVLT-II Long Delay (Beta −0.370 (SE 0.176), *p* = 0.046)), as well as with the mental component of the cognitive-related quality-of-life score (SF-36-MCS (Beta −0.338 (SE 0.153), *p* = 0.039); Figure 3C,D). The combined total amyloid-venous VTI index correlated with WMS-IV LM-II (Beta −0.440 (SE 0.132), *p* = 0.003) and SF-36-MCS (Beta −0.302 (SE 0.141), *p* = 0.045; Figure 3A,B and Table 3).

*Cells* **2021**, *10*, x FOR PEER REVIEW 10 of 16

**Figure 3.** Retinal amyloid count combined with retinal venous VTI correlated with verbal memory and cognitive-related quality-of-life measures. Graphs illustrating the correlations between the combined total amyloid-venous tortuosity index and verbal memory (**A**) and cognitive-related quality-of-life Z-scores (**B**), and the correlations between the combined proximal mid-periphery amyloid-venous tortuosity index and verbal memory (**C**) and cognitive-related quality-of-life Z-scores (**D**). Abbreviations: WMS-IV LM II, Wechsler Memory Scale IV Logical Memory II; WMS-IV, Wechsler Memory Scale IV; SF-36 MCS, SF-36 Mental Component Score; PMP, proximal mid-periphery; and VTI, vessel tortuosity index.

#### **4. Discussion**

The main findings from this exploratory investigation of retinal fluorescence imaging are that retinal vascular features do not significantly correlate with retinal amyloid deposition in the proximal mid-periphery area; proximal mid-periphery retinal amyloid count correlates with verbal memory; and the combination of the retinal amyloid and venous tortuosity index into standardized index scores can provide a more comprehensive indicator of cognitive performance.

Microvascular damage is increasingly recognized as a critical initiator of vascular cognitive impairment and AD pathology [5,9,69]. Vascular dysregulation, leading to cerebral amyloid accumulation, and the link between cerebrovascular disease and dementia are

explained by several mechanisms [4,5,70]. Pericyte loss and deficient vascular plateletderived growth factor receptor-β signaling were identified in both the retinal and cerebral vasculature in subjects with MCI and AD [4,28]. Prior reports demonstrated that retinal vasculature may be used as a biomarker of early or preclinical dementia [71], and retinal microvascular abnormalities in MCI and dementia have been demonstrated using various retinal vasculature imaging modalities (e.g., retinal fundus photography [48,72,73], optical coherence tomography angiography [74,75], high-frequency flicker-light stimulation [51,76], and the retinal function imager [74]). Conversely, den Haan et al. [77] showed that retinal vascular measures did not differ between patients with AD and control participants, and venular tortuosity was smaller in subjects with greater white matter disease burden. Previous investigations have also shown a strong relationship between retinal vasculopathy and brain amyloid deposition [30]. Sharafi et al. [30] evaluated the relationship between retinal vascular statuses (vessel diameter and both tortuosity and spatial-spectral texture measures) using hyperspectral retinal imaging and CNS amyloid status (assessed with (18) F-florbetaben positron-emission tomography). They found that retinal venules of amyloid-positive subjects showed a higher mean tortuosity compared with the amyloid-negative subjects. This study suggested that the inclusion of metrics related to retinal vasculature and the surrounding tissue-related texture could improve the discrimination performance of the cerebral amyloid status [30].

As both retinal amyloid accumulation and retinal vascular pathology [28,34–37] are reported in patients with MCI, we explored the interplay between retinal vascular geometric measures and retinal amyloid burden using retinal fluorescence imaging. Prior studies showed that the retinal proximal mid-periphery area may be the target of amyloid quantification to reflect cerebral AD pathology, as it correlates with cognitive performance and hippocampal volume [35,37]. In this pilot cohort, we found that retinal vascular features correlated with amyloid deposition in the posterior pole and retinal distal mid-periphery area, but not with the proximal mid-periphery area. A possible explanation is that this investigation measured physical features of the retinal vasculature (e.g., branching angle and tortuosity index) and did not assess functional endpoints. Additionally, our quantitative vascular analysis could not target the smaller retinal blood vessels. The mechanisms driving vascular remodeling and amyloid deposition may occur at different rates, leading to the appearance of these clinical signs at different stages in disease progression. The investigation of subjects with mainly mild cognitive impairment in our cohort may explain why venous VTI was lower in subjects with worse cognition and why the other arterial or venous vascular parameters did not show any significant differences across cognitive strata. This hypothesis is supported by the lack of association between retinal vascular features and most of the neuropsychometric cognitive scores in our cohort. Conversely, retinal vascular measures did not correlate with any cognitive measures except for attention and concentration. The total and proximal mid-periphery retinal amyloid count correlated with verbal memory measures, while the distal mid-periphery amyloid count correlated with non-verbal memory measures. Interestingly, in this cohort with early cognitive disorders, subjects with higher amyloid counts and worse cognition levels had lower retinal venular tortuosity. The only combined index that discriminated between individuals with impaired cognition and normal cognition was the proximal mid-periphery amyloid count-venular VTI. This combined index was also associated with verbal memory and the 'mental component' summary of psychological functioning (SF-36 Mental Component Score). This latter finding reflects the association with cognitive-related psychological and emotional functioning. This appears to represent an exclusive contribution, as physical functioning status, as demonstrated by the SF-36 Physical Component Score, was not associated with amyloid or vascular retinal markers.

Two or more retinal vascular abnormalities were associated in a dose-response manner with an increased risk of disabling dementia in a prior study [49]. It is possible that combined amyloid–vascular indexes are better discriminators of cognitive function, with the potential for use as outcome measures in AD and mixed dementia trials. Our study is limited by a small sample size, heterogeneity, and the absence of genetic and CSF or brain amyloid biomarkers. Similarly, our patients were not evaluated for all ocular conditions other than a history of glaucoma. Due to the limited sample size, we could not adjust for the presence of traditional vascular risk factors or the presence of retinopathy, which are known contributors to retinal vascular geometric changes. Given the heterogeneity in the sample size across study groups, further confirmation of these preliminary results will be necessary in the future for specific groups of early AD and vascular and mixed dementias.

Our findings underscore the potential value of the exploratory amyloid-vascular indexes presented herein. Future investigations are warranted to explore the clinical utility of retinal fluorescence imaging in concert with combined amyloid-vascular index measures. More comprehensive cohort studies including a larger sample size and a greater range of disease severity among the participants could help to elucidate the stage at which retinal amyloid and/or retinal venular versus arterial impairments begin to develop in cognitive disorders. Given the cost and technical requirements of gold-standard methods for assessing cerebral amyloid deposition and vascular pathology, further validation of these retinal imaging methods could potentially yield greater accessibility to testing, thus facilitating more extensive clinical trials as well as improving the detection of early dementia.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/cells10112926/s1, Table S1: Retinal vascular parameter correlation with retinal amyloid measures.

**Author Contributions:** Conceptualization, O.M.D., R.R., S.F., P.D.L., S.C., Y.K. and M.K.-H.; methodology, O.M.D., R.R., M.M.K., A.S., D.S., D.S.S., Y.K., P.D.L., S.C. and M.K.-H.; software, R.R., J.S., T.T., K.O.J., A.D.C., S.V., K.L.B., S.F. and M.K.-H.; validation, D.S.S, J.S., T.T., S.C.; formal analysis, R.R., M.M.K., Y.S., S.C. and M.K.-H.; investigation, D.S.S., A.S., D.S., K.L.B., P.D.L. and M.K.-H.; resources, K.L.B., P.D.L. and M.K.-H.; data curation, O.M.D., R.R., J.S., T.T., Y.K., S.C. and M.K.-H.; writing—original draft preparation, O.M.D.; writing—review and editing, O.M.D., R.R., D.S.S., M.K.-H., K.O.J., A.D.C., S.V., K.L.B., S.F., Y.K., S.F., P.D.L., Y.S., S.C. and M.K.-H.; visualization, S.C., P.D.L. and M.K.-H.; supervision, P.D.L., Y.K., M.K.-H. and K.L.B.; project administration, O.M.D. and M.K.-H.; funding acquisition, K.L.B. and M.K.-H. All authors have read and agreed to the published version of the manuscript.

**Funding:** We received support from a National Institute on Aging award (AG044897, Koronyo-Hamaoui, PI) and from the Saban, Gordon, and Marciano Private Foundations (Koronyo-Hamaoui, PI). The funders had no role in the design or conduct of this research.

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki and was approved by the Institutional Review Board (or Ethics Committee) of Cedars-Sinai Medical Center (protocol code 00052349, approved in 2018).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** Data available upon request due to restrictions, e.g., privacy or ethical restrictions.

**Acknowledgments:** We thank Mia Oviatt for editing the manuscript. The authors dedicate the manuscript to the memory of Salomon Moni Hamaoui and Lillian Jones Black, who died of Alzheimer's disease.

**Conflicts of Interest:** Black, Verdooner, Koronyo, and Koronyo-Hamaoui are founding members of NeuroVision Imaging Inc., 1395 Garden Highway, Suite 250, Sacramento, CA 95833, USA. Dr. Frautschy is co-inventor of the US patent US9192644B2 for a curcumin formulation. Johnson, Czeszynski, and Verdooner are currently employed by NeuroVision Imaging Inc. The remaining authors declare that the research study was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

#### **References**


### *Article* **Retinal Protection from LED-Backlit Screen Lights by Short Wavelength Absorption Filters**

**Celia Sanchez-Ramos <sup>1</sup> , Cristina Bonnin-Arias <sup>1</sup> , Vanesa Blázquez-Sánchez <sup>1</sup> , Victoria Aguirre-Vilacoro <sup>1</sup> , Teresa Cobo <sup>2</sup> , Olivia García-Suarez <sup>3</sup> , María Jesús Perez-Carrasco <sup>4</sup> , Cristina Alvarez-Peregrina 5,\* and José A. Vega 3,6**


**Abstract:** (1) Background: Ocular exposure to intense light or long-time exposure to low-intensity short-wavelength lights may cause eye injury. Excessive levels of blue light induce photochemical damage to the retinal pigment and degeneration of photoreceptors of the outer segments. Currently, people spend a lot of time watching LED screens that emit high proportions of blue light. This study aims to assess the effects of light emitted by LED tablet screens on pigmented rat retinas with and without optical filters. (2) Methods: Commercially available tablets were used for exposure experiments on three groups of rats. One was exposed to tablet screens, the other was exposed to the tablet screens with a selective filter and the other was a control group. Structure, gene expression (including life/death, extracellular matrix degradation, growth factors, and oxidative stress related genes), and immunohistochemistry in the retina were compared among groups. (3) Results: There was a reduction of the thickness of the external nuclear layer and changes in the genes involved in cell survival and death, extracellular matrix turnover, growth factors, inflammation, and oxidative stress, leading decrease in cell density and retinal damage in the first group. Modulation of gene changes was observed when the LED light of screens was modified with an optical filter. (4) Conclusions: The use of short-wavelength selective filters on the screens contribute to reduce LED light-induced damage in the rat retina.

**Keywords:** retinal light injury; LED screen; optical filter; retinal protection

### **1. Introduction**

As early as 1966 it was suggested that exposure to low-intensity short-wavelength light for a long time can induce retinal damage [1], with the action spectrum (400–440 nm) of blue light the most dangerous [2,3], able to trigger or exacerbate macular and retinal damage [4,5]. Both human and animal studies suggest that excessive levels of blue light induces immediate photochemical damage to the retinal pigment epithelial cells (RPE), photoreceptors, and ganglion cells [2,3,6–10]. Thus, the phototoxicity of blue light may contribute to the progression and severity of age-related macular degeneration (AMD) and vision loss.

Nowadays, light-emitting diodes (LEDs) are gradually becoming the majority of the domestic light sources, replacing conventional ones [2,11,12]. The most commonly used

**Citation:** Sanchez-Ramos, C.; Bonnin-Arias, C.; Blázquez-Sánchez, V.; Aguirre-Vilacoro, V.; Cobo, T.; García-Suarez, O.; Perez-Carrasco, M.J.; Alvarez-Peregrina, C.; Vega, J.A. Retinal Protection from LED-Backlit Screen Lights by Short Wavelength Absorption Filters. *Cells* **2021**, *10*, 3248. https://doi.org/10.3390/ cells10113248

Academic Editors: Maurice Ptito and Joseph Bouskila

Received: 29 September 2021 Accepted: 16 November 2021 Published: 19 November 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

method to produce white LED lights is the use of blue diodes covered with a phosphor compound; thus, one of the biggest risks of LEDs is its emission spectrum, since it contains a large blue light component [2,12,13]. Additionally, there is time-related degradation of phosphorus which leads to a progressively increasing release of short wavelengths (blue light) [9]. Since LEDs do not directly emit ultraviolet (UV) and infrared (IR) wavelengths, the blue light risk is the main ones to focus on when considering LEDs and LED systems [12].

Recent studies assessing LED effects on the retina using ambient LED lighting have demonstrated LED phototoxicity [2,3,9,11,12,14,15]. To our knowledge, no data are available regarding backlit LED screens. People spend increasing amounts of time viewing backlit LED screens, such as personal computers, tablets, smartphones, etc., at short distances. Thus, this raised a range of public concerns about their potential risks as retinal hazards [16–18]. In 2017, Lin et al. investigated the effects of periodic exposure to the blue LED in rats showing marked retinal damage [19] characterized by an accumulation of macrophages (identified because express ionized calcium binding adapter molecule-1), drusen-like materials around the outer segments of the photoreceptors, and finally degeneration of the photoreceptors [7]. Additionally, LED induce strong damage in rat retinal pigment epithelium (RPE) characterized by the breakdown of the blood–retinal barrier and the induction of necrotic cell death [20]. Furthermore, Balb/c mice exposed to LED light show a reduction of outer nuclear layer (ONL), increase in TUNEL-positive apoptotic nuclei, changes in several differentially expressed genes, and downregulation of ubiquitin and autophagy function [21]. At the basis of those changes is the oxidative stress induced by LED [22] which affects mitochondria and triggers mitochondrial death signaling pathways [17,23], as well as accumulation of damaged lysosomes and subsequent lysosomal cell death [24].

On the other hand, the deleterious effects of LED lights on the retina can be prevented partially by blue-light-blocking films [25], or more effectively by lenses with a brown or gray tint [26].

Previous studies from our research group have demonstrated that exposure to blue light reduces the number of retinal cells, upregulates genes related to cell death, downregulates genes involved in cell survival, increases the activity of genes involved in extracellular matrix degradation, and alters the expression of growth factor that participates in cell maintenance and survival. These effects can be prevented or reversed by filtering blue light [27,28]. Therefore, this study aims to evaluate the effects of light emitted by back-lit LED tablet screens on the retina of pigmented rats and determine whether or not they can be modified by partially filtering out the emitted short-wavelengths of the visible spectrum. Based on the multiple pathways that can participate in LED-induced retinal damage, and our previous experience and studies, the following genes were analyzed: genes involved in cell survival and death (Bcl-2; Bcl-XL; Bax; Bak; Bcl-XS, Caspases-3 and 9), genes related to the extracellular matrix (ECM) turnover (MMPs-2 and 9, ADAMTS-12 and 14, TIMPs-1 and 2), genes related to growth factors (BDNF-Trk-B system, VEGF-VEGFr-2 system), and inflammation (TNF-α, SODs-1 and 2). Furthermore, immunohistochemistry was used to study determined the protein product of the genes showing greater variation after exposure to LED. The results could support the effects of short-wavelengths emitted by back-lit LED screens on gene regulation and assess the efficacy of the filter in removing excessive blue light radiation, thus potentially providing a retinal photoprotective effect.

#### **2. Materials and Methods**

#### *2.1. Animals and Rearing Conditions*

Male Lister-Hooded rats obtained from Harlan Laboratories Models, S.L. (Barcelona, Spain), were housed at the bioterium of the Medical School of the Universidad Complutense de Madrid (UCM, Madrid, Spain). The animals were kept in a dark environment for 14 days to remove the effect of light exposure from their previous breeding location, with access to food and water ad libitum. The use of rats complied with the Statement for the Use of

Animals in Ophthalmic and Vision Research (ARVO 2013). Animals were treated humanely and with regard to the alleviation of suffering. This study has the approval of the Animal Experimentation Committee of the UCM and the Department of Health of the Comunidad de Madrid, Spain (Reference PROEX 310/15).

Rats were divided into three groups, with 12 animals per group. Group 1 was exposed to the light emitted by the LED-backlit tablet screens; group 2 was exposed to the light emitted by the LED-backlit tablet screens with a selective short-wavelength absorption filter adhered to the screen; the control group was unexposed to LED-backlit tablet screens. Animals were housed in clear Makrolon Polycarbonate cages, 58 × 38 × 18 cm in size, with two animals per cage.

#### *2.2. Light Source and Exposure*

Light emitted by blank LED-backlit tablet screens, with size 231 × 147.2 × 8.7 mm (Cristal IPS multitouch 8.9", 1920 × 1200 px Full HD), set at full brightness was used. The selective short-wavelength absorption filter used for group 2 exposure conditions was Reticare ® Intensive (Tecnología Sostenible y Responsable S.L., Madrid, Spain). Before starting the study, the filter was adhered to each tablet screen, following the manufacturer's instructions.

The photo exposure process was designed aiming at simulating the conditions of use of touch screens by children. The screens were placed surrounding the cages, at a distance between 4.72 and 5.90 inches of the animal's eyes, leaving the upper and lower areas without screens. As for the devices used for photo exposure, tablets were used instead of smartphones for two reasons: firstly, because the time of use of tablets, in general, is longer than smartphones; and secondly, because the tablets emit with less intensity (roughly half) than mobile phones.

Tablet screens were set at 10 cm from each of the four cage walls. For groups 1 and 2, each cage contained six screens (without or with filter, respectively), one screen in each of the two short walls, and two screens in each of the two long walls. No screens were attached to cage ceilings. Figure 1 shows the characteristic of the light LED screen emission and the illuminance (lux) measured inside the study cages, as well as the transmittance curve of selective short-wavelength absorption filter (Reticare ® Intensive). LED-backlit screen tablet light emission (with and without filter) was measured with an Ocean Optics USB2000 + Spectrometer.


Animals of groups 1 and 2 were exposed to 8-hr dark/16-hr LED-backlit screen tablet light cycles for three months. All animals, including a control group, were sacrificed at the end of the exposure period.

β

−

#### *2.3. Tissue Collection, RNA Extraction, and qPCR*

The animals were sacrificed with an overdose of pentobarbital sodium (200 mg/kg, ip) and eyes were removed. The left eyes of each animal were frozen at −80 ◦C and maintained until quantitative polymerase chain reaction (qPCR) analysis. Right eyes were fixed in 4% formaldehyde for 24 h, then washed in tap water, and routinely processed for paraffin inclusion, and used for structural and immunohistochemical studies.

Total RNA was extracted from the whole retina using a commercial kit (Trizol Reagent, Invitrogen, Carlsbad, CA, USA), following the manufacturer's instructions. After precipitation and cold ethanol washing, RNA was dried and dissolved in an appropriate volume of Tris–EDTA buffer (10 mM Tris–HCl pH 8.0 and 1mM EDTA-Na2). Each sample was treated with 1U of DNase I for 1-h at 37 ◦C to digest genomic DNA. RNA was precipitated, washed, and dissolved again in the same buffer. RNA solution was quantified at 260 nm (Biomate 3, Thermo Electron Corporation, Waltham, MA, USA) and its purity was assessed by the ratio of 260/280 nm readings. We used the High Capacity cDNA Archive kit (Applied Biosystems, Foster City, CA, USA), random hexamers, and 10 µg of total RNA to make cDNA following the manufacturer's instructions. Then, 1 µg of cDNA was used to detect the expression of the genes analyzed in this qPCR study.

Quantitative PCR was performed using 1 µg of the cDNA, and the primers used to detect the investigated genes, and the β-actin were designed based upon the published mRNA sequences for Rattus norvegicus (Table 1; GenBank accession numbers are included).


**Table 1.** Genes sequences analyzed in this study.

Homemade TaqMan probes were labeled at the 50 with 60 FAM fluorochromes for all investigated sequences, and VIC fluorochrome for β-actin, while the 3′ ends were labeled with the Minor Groove Binder (MGB) quencher. The PCR reactions were performed using the TaqMan Universal PCR Master Mix (Applied Biosystems, Foster City, CA, USA) using 5 pmol of each primer and 9 pmol of both target and β-actin probe. The assays were performed in triplicate in independent experiments using a 7500 PCR real-time system (Applied Biosystems), and quantification was calculated using the 2-∆∆Ct algorithm against β-actin and expressed as the n-fold difference compared to an arbitrary calibrator, chosen as a higher value than ∆∆Ct.

Average values obtained in control group animals were compared to those exposed to LED-backlit tablet screen light, without and with filter (group 1 and 2, respectively), and the results were expressed as fold difference compared to control (relative expression).

#### *2.4. Structural and Immunohistochemical Studies*

The structure of the retinal in all animal groups evaluated by staining representative formol fixed, paraffin embedded sections (10 per eye and animal) with hematoxylin & eosin. The sections were then scanned by an SCN400F scanner (Leica Biosystems™, Newcastle, UK), and the scans were computerized using SlidePath Gateway LAN software (Leica, Leica Biosystems™). Then, the whole thickness of the retina was measured, evaluating in 10 points the distance from the inner to the other surface of the retina.

Deparaffinized and rehydrated sections were processed for detection SYN using the EnVision antibody complex detection kit (DakoCytomation, Copenhagen, Denmark), following supplier's instructions. Briefly, the endogenous peroxidase activity and non-specific binding were blocked, and the sections were then incubated overnight at 4 ◦C with the primary antibodies included in the Table 2. Subsequently, sections were rinsed incubated with Dako EnVision System labeled polymer-HR anti-mouse IgG (DakoCytomation) for 30 min at room temperature. Finally, sections were washed, and immunoreaction visualized using 3-3′ -diaminobenzidine as chromogen. To ascertain structural details some sections were counterstained with Mayer's hematoxylin, dehydrated, and mounted with Entellan® (Merk, Dramstadt, Germany).


**Table 2.** Antibodies used in the study.

<sup>1</sup> Watham, MA, USA; <sup>2</sup> Abcam: Cambridge, UK; <sup>3</sup> Seattle, WA, USA; <sup>4</sup> Santa Cruz, CA, USA; <sup>5</sup> Dallas, TX, USA; <sup>6</sup> Waltman, MA, USA.

The variations in the intensity of immunostaining developed for each investigated antibody was evaluated in five sections per animal separated by 100 µm. Each section was scanned with an SCN400F scanner (Leica Biosystems™) and annotated using SlidePath Gateway LAN software (Leica Biosystems™). A 1 mm<sup>2</sup> grid was then applied randomly onto 2 × 500 µm enlarged images in 4 non-overlapping fields (4 mm<sup>2</sup> per section; 20 mm<sup>2</sup> per subject), and the free nerve endings and sensory corpuscles within the grid were counted by two independent observers. The results are expressed as absolute numbers for the densities of the sensory corpuscles and free nerve endings per cm<sup>2</sup> . The intensity of immunostaining developed with the different retinal layers was evaluated in arbitrary units of grey levels ranging from 1 (black) to 256 (white) using an image analysis system (MIP System, Servicio de Análisis de Imágenes, Universidad de Oviedo, Spain). The intensities were therefore divided into four groups (64 units of grey level each), referred in the text and tables as strong (1–64, ++++), high (65–128, +++), intermediate (129–192), and low (193 to level of the background in control sections, +). Intensities of grey higher than those of background of the corresponding control sections were considered unreactive. Although we attempted to process all samples identically, there may have been variations in the intensity of the final immunoreaction due to technical aspects, i.e., differences in penetration of the fixative. On the other hand, for the immunohistochemical study, the retinal segment and orientation of the sections were not taken into account, and therefore the thickness of the retina may appear in the images with greater thickness in the experimental animals than in the controls.

#### *2.5. Statistical Analysis*

Significant differences among the three groups were assessed with the Kruskal–Wallis H test, and *p*-values < 0.05 were considered statistically significant (marked in the figures as \* *p* < 0.05, \*\* *p* < 0.01).

#### **3. Results**

The study was carried out with 36 male Lister-Hooded rats obtained from Harlan Laboratories Models, S.L. (Barcelona, Spain), housed at the bioterium of the Medical School of the Universidad Complutense de Madrid (UCM, Madrid, Spain), with food and water ad libitum, kept in a dark environment for 14 days to remove the effect of light exposure from their previous breeding location. Rats were divided into 3 groups, with 12 animals per group. Group 1 was exposed to the light emitted by the LED-backlit tablet screens; Group 2 was exposed to the light emitted by the LED-backlit tablet screens with a selective short-wavelength absorption filter adhered to the screen; the control group was unexposed to LED-backlit tablet screens.

#### *3.1. Structural Study*

The retinal structure was studied in the three groups of animals. In those exposed to LED-backlit tablet screen light, there was a significant decrease in the whole retinal thickness (reduction of 23.82 ± 6.21%), apparently due to a reduction in the number of cells in both the outer and inner nuclear layers. The whole thickness of the retina was measured by evaluating in 10 points the distance from the inner to the other surface of the retina. These structural changes were almost reversed by the selective short-wavelength absorption filter (Figure 2).

#### *3.2. Gene Expression Study*

#### 3.2.1. Life/Death Cell-Related Genes

Different genes related to cell survival and cell death were analyzed. The eyes of the animals exposed to LED-backlit tablet screen light without filter (group 1) showed a downregulation of the genes encoding anti-apoptotic proteins, especially Bcl-2 (−13.2-fold), and to a lesser extent Bcl-XL (−3.5-fold), as Figure 3a shows. Selective short-wavelength absorption almost eliminated these responses. Regarding genes directly or indirectly

involved in cell death, exposition to the LED-backlit tablet screen light without filter resulted in a moderate up-regulation of Bax and Bak (5- and 3.3-fold, respectively) and a marked up-regulation of Bcl-XS (16.3 folds), as Figure 3b shows. Similarly, in this group (group 1), there was an up-regulation of both caspase-3 (6-fold) and caspase-9 (11.8-fold) genes. This scenario changes after the use of the selective short-wavelength absorption filter. In fact, in this experimental situation, Bak and caspase-3 were undetectable, and the expression of Bax, Bcl-XL, and caspase-9 remained up-regulated for the controls, but the expression was significantly reduced for the group of animals exposed to LED-backlit screens, as Figure 3c shows.

**Figure 2.** Retinal structure from control rats, rats exposed to LED Screen (LED-S), and rats exposed to LED screen with a protective filter (LED-S-P).

#### 3.2.2. Extracellular Matrix Degradation: MMP/TIMPs System and ADAMTS Genes

− − Figure 4 shows that exposure to LED-backlit tablet screen light without filter (group 1) entails an increase in MMP-2 and MMP-9 mRNA expression, suggesting an increase in the extracellular matrix (ECM) turnover mediated by these proteases. As for changes in TIMPs expression, exposure to light emitted by LED-backlit tablet screens caused a decrease of both TIMP-1 and TIMP-2. The use of the selective short-wavelength absorption filter almost reversed the decrease, although in levels below control group values, as Figure 4c shows. Regarding ADAMTS-12, no differences were found between the control group versus either group 1 or group 2. On the other hand, exposure to LED-backlit tablet screen light without filter (group 1) increased the levels of ADAMTS-14 expression, whereas exposure effects were reverted with the filter (group 2), as Figure 4b shows.

#### 3.2.3. BDNF/TrkB and VEGF/VEGFR-2 System, and TNF-α

Figure 5a shows how exposure to LED-backlit tablet screen light did not modify expression levels of BDNF either with or without the selective short-wavelength absorption filter, while TrkB gene expression increased significantly. It is important to highlight that, with the filter, TrkB gene expression persisted at a high level, although below the level observed in group 1 (without filter).

**Figure 3.** Gene expression comparisons between group 1 vs. control and group 2 vs. control. (**a**) Anti-apoptotic gene expression, (**b**) pro-apoptotic gene expression, and (**c**) gene expression of apoptotic-related enzymes. \* *p* < 0.05; \*\* *p* < 0.01.

α

**Figure 5.** Variation of the expression in both experimental vs. control groups, of (**a**) TrKB; (**b**) VEHG and VEGFr2; and (**c**) and TNF-α. \* *p* < 0.05; \*\* *p* < 0.01.

α. \*

Figure 5b shows that after exposure to LED-backlit tablet screen light (group 1), levels of VEGF mRNA increase moderately, while levels of its receptor increased by nearly 14-fold. These effects were highly attenuated using the selective short-wavelength absorption filter (group 2): VEGF levels were practically identical to controls, and VEGFr2 were lower, although above controls.

Finally, Figure 5c shows that TNF-α mRNA expression levels increased up to 14-fold in group 1 rats (without filter) but were like controls in group 2 (with filter). α mRNA expression levels increase

#### 3.2.4. Oxidative Stress

Figure 6 shows how the expression of superoxide dismutase-1 and -2 were strongly up-regulated after exposure to LED-backlit tablet screen light (group 1). These effects are highly attenuated, but increased for the controls, using the selective short-wavelength absorption filter.

**Figure 6.** Variation in the expression of SOD1 and SOD2 in both experimental vs. control groups. \* *p* < 0.05; \*\* *p* < 0.01.

Table 3 summarizes the results of the variations in the expression of the genes analyzed, as a complement of the figures.

#### *3.3. Immunohistochemistry*

− − − Using immunohistochemistry, we analyzed the detection in retinal sections of the protein products of the genes whose expression were most affected by the experimental conditions to which the two groups of animals. The increases or decreases in the intensity of immunostaining paralleled that observed in the gene expression. The immunolabelling for Bcl-2 was greatly reduced in animals exposed to LEDs and partially recovered when light was filtered (Figure 7), whereas no notable variations were observed for Bcl-X (Figure 7). The detection of caspase-3 showed an increase in immunostaining in the group subjected to LED light and distribution and intensity similar to the controls in animals subjected to filtered light (Figure 7).


**Table 3.** Changes in the gene expression in the two established experimental groups in relation to the controls.

Regarding the two metalloproteases investigated, MMP2 did not show positive immunoreaction in the control animals, while showing a slight immunopositivity in the animals subjected to LEDs that was reduced in the group of filtered LED light (Figure 8). On the other hand, the immunoreaction for MMP9 was very similar in the three groups (Figure 8).

The immunoreactivity for TrkB, the high affinity receptor for the neurotrophin BDNF, was increased in animals exposed to LED, especially in the photoreceptors, but also in the inner nuclear layer and ganglionic cells layer, with levels similar to those of the animal controls when filtering LEDs (Figure 9).

The immunoreactivity for SOD1 increased in the photoreceptors layer of LED exposed rats and returned to similar levels of the control after LED filtering. However, SOD2 immunoreactivity increased in the photoreceptors, inner nuclear and ganglionic cell layers in LED-exposed animals, and the pattern of expression was not reverted by filtering (Figure 10).

**Figure 7.** Immunohistochemical detection of Bcl-2 (**upper panel**), Bcl-X (**central panel**) and Caspase-3 (**lower panel**) in the three groups of rats investigated. LED: animals exposed to LED-backlit screen for 3 months; LED + F: animals exposed to filtered LED-backlit screen for 3 months. GCL: ganglionic cells layer; INL: inner nuclear layer; ONL: outer nuclear layer.

**Figure 8.** Immunohistochemical detection of MMP2 (**upper panel**), and MMP9 (**lower panel**) in the three groups of rats investigated. LED: animals exposed to LED-backlit screen for 3 months; LED + F: animals exposed to filtered LED-backlit screen for 3 months. GCL: ganglionic cells layer; INL: inner nuclear layer; ONL: outer nuclear layer.

**Figure 9.** Immunohistochemical detection of TrkB (**upper panel**), and VEGFR2 (**lower panel**) in the three groups of rats investigated. LED: animals exposed to LED-backlit screen for 3 months; LED + F: animals exposed to filtered LED-backlit screen for 3 months. GCL: ganglionic cells layer; INL: inner nuclear layer; ONL: outer nuclear layer.

**Figure 10.** Immunohistochemical detection of SOD1 (**upper panel**), and SOD2 (**lower panel**) in the three groups of rats investigated. LED: animals exposed to LED-backlit screen for 3 months; LED + F: animals exposed to filtered LED-backlit screen for 3 months. GCL: ganglionic cells layer; INL: inner nuclear layer; ONL: outer nuclear layer.

The results of the quantitative study performed on immunohistochemical sections are in Table 4.

−/+

−/+ − − −

− −/+ −

− − −

−/+ −/+

− −/+

476


**Table 4.** Variations in the intensity of immunostaining in the control and the two experimental groups.

#### **4. Discussion**

The present study was designed to investigate the effects of the exposure to LEDbacklit tablet screen light on the structure, gene expression, and protein localization of the retina of a rat model. Moreover, we have analyzed whether or not those effects could be reversed partial or using a selective short-wavelength absorption filter. The genes and proteins investigated were related to cell survival and cell death, as well as some proteases, and protease-blockers involved in the turnover of the ECM, growth factors previously known to be affected by light exposure, one involved in inflammation, and others related to oxidative stress. This approach is necessary because of the multiple factors involved in phototoxicity [15]. Because the results of the gene expression and immunohistochemistry were in parallel, they are discussed together.

In 2001, Dawson showed that blue LEDs (460 nm) and argon lasers (458 nm) induced retinal damage with corneal irradiances of 10 J/cm<sup>2</sup> [29]. Ueda et al. also observed detrimental effects in the macula with blue LEDs (465 nm) [30]; besides, Shang et al. showed that the spectral range distribution of blue-white LEDs contains a significant fraction of short-wavelengths that induced irreversible retinal neuronal cell death in rats; they exposed Sprague-Dawley rats to white and blue LED light (750 lux) for 28 days, finding an increase in free radical production in the LED-exposed group [9]. Finally, Krigel et al. used albino Wistar and pigmented Long Evans rats exposed for 1–28 days to 500–6000 lux LEDs (cold white, blue, and green) [15]. The authors found that the blue

component of the white-LED reduced photoreceptor layer thickness and induced retinal toxicity. In previous in vitro studies, Chamorro et al. exposed human retina cells to LED lighting in 12-h dark/12-h light cycles affected RPE cell growth and induced cellular stress, increasing levels of reactive oxygen species, DNA damage, and apoptotic cells [3]. Additionally, when a selective blue-light absorption filter was added, phototoxic damage in human RPE cells was reduced, providing a retinal photoprotector effect. However, few studies have investigated long-term and low luminance light-induced retinal phototoxicity and, to our knowledge, all have been carried out using different ambient LED lights. In contrast to previous studies, we investigated the induced retinopathy in a rat model due with long-term, low-intensity exposure to commercially available LED-backlit tablet light. Thus, the main objective of the current study was to assess the long-term effects on the retina of pigmented rats by light emitted by a backlit commercially-available digital device LED screen (tablet), with and without a short-wavelength selective absorption filter, in 16-h light/8-h dark cycles.

Although albino rats are commonly used for studies about retinal damage [9], in our study we have used pigmented rats. Compared to albino strains, the age of the animals at the onset of the light exposure, and not ocular pigmentation, is the most important factor in regulating the severity of light-induced retinal damage [31]. However, following the Shang et al. protocol, the conditions of the animals analyzed included a 14-day pre-study housing period in a dark environment to clear the light exposure effect prior to breeding conditions. Experimental animal retinas were baseline evaluated after this pre-study washout period. The experimental conditions of the present study differ substantially and significantly in various aspects from that of Shang et al. study [9]. Firstly, in the light source used: those authors used single-wavelength blue LEDs (460 ± 10 nm) and custom-made PC white LEDs at domestic lighting levels, while we have used commercially available LED-backlit tablet screens, to replicate real-life usage. Secondly, the exposure conditions: in our study, back-lit LED tablet screens exposed to rats with and without a selective shortwavelength absorption filter, aimed to show that, by the use of the filter, the transmission reduction of the most energetic LED light emitted band (blue light) resulted in reduced cell damage, as shown in studies conducted by Sparrow et al. and Nagai et al., who used yellow optic filtered intraocular lenses [32,33]. Thirdly, the mean luminance used by Shang et al. was 750 lux, the usual domestic luminance level, while in our study mean luminance was 140 lux, which corresponds to the light emitted by a commercially available LED-backlit tablet screen-blank screen set at full brightness. Finally, in the exposure time: whereas Shang et al. used light exposure times of 3, 9, and 28 days under 12-h dark/12-h light cycles, we extended the study to 3 months under more aggressive cycles: 8-h dark/16-h light.

To our knowledge, this is the first study regarding the effects of LED-backlit screens on the mammalian retina, and it is relevant because the time we spend watching LEDbacklit screens of domestic devices (as personal computers, tablets, smartphones) at short distances has increased progressively in recent years, and this also includes infants and young people.

The retinal structure in the group of animals exposed to long periods of LED-backlit tablets light was altered, showing a significant decreases in the number of cells in the outer nuclear layer. No further significant changes were observed neither in the inner nuclear layer nor in the ganglion cell layer. These results agree with those obtained by Shang et al. in 2014 [9]. Lin et al. in 2017 reported marked retinal damage by regular exposure to a blue light-emitting diode in rats [19]. However, interestingly, these structural changes were prevented by filtering light specifically, as previously reported by different authors, including our research group.

The reduction in the number of retinal cells due to blue light exposure might be due to multiple factors, such as apoptosis of photoreceptors and RPE reported earlier [2,3,9,11,12,14,15]. Our results on the life/death-related genes further argue for the causes of cell loss after back-lit LED screens exposure. We observed that long periods of light exposure increase the expression of some genes related to cell death, such as Bax, Bak, and Bcl-XS as well as caspase 3 and caspase 9; conversely, the genes related to cell survival like Bcl-2 and Bcl-xl were down-regulated. Supporting our results, Lin et al. also observed increased expression of Bax and caspase-3, decreased expression of inhibited Bcl-2 and Bcl-xL, and inhibition of Bcl-2/Bax association in the RPE after regular exposure to blue light-emitting diode [19].

The exposition to LED-backlit screens strongly up-regulates MMP-2 and MMP-9, to a lesser extent ADAMTS-14, and was without effect on ADAMTS-12, and down-regulates both TIMP-1 and TIMP-2. These changes could have deleterious effects on the retina due to the extracellular matrix (ECM). Increased MMP expression could induce a faster ECM turnover to elude matrix deposit genesis [34]. Our experiments indicate upregulation in MMP-2 and MMP-9 expression, especially in rats exposed to the unfiltered LED tablet screen, whereas Sanchez-Ramos et al. in 2010 did not find any change in MMP-2 expression in other animal model trials [27]. Plantner et al. detected an increase in this MMP [35]. As for MMP-9 expression, our results are consistent with those obtained by Papp et al. and by Sanchez-Ramos [27,36]. In general, these results cannot support the hypothesis that the formation of deposits that give rise to retinal drusen is due to a decrease in pigmentary epithelium MMPs [37], as there are two possible interpretations. On one hand, long-term exposure to light causes an increase in the expression of some MMPs and ADAMTS-14 concurrently by a decrease of TIMP-1 and TIMP-2. These facts could affect the retina due to ECM damage. On the other, the increase in MMPs and ADMATS-14 expression could be related to increased ECM turnover to avoid the appearance of deposits that causes drusen [27].

As for the analysis of the results of BDNF and its receptor TrkB, exposure to light in our study conditions, whether with or without filter, had no effects on BDNF gene expression, which is consistent with findings by Wen et al., who reported that continuous exposure to white light had no significant effects on BDNF expression [38]. Asai et al. also obtained similar results by Western blot densitometry [39]. On the contrary, TrkB receptor expression levels were affected by LED-backlit screen tablet light in rat retinas, both with and without the filter. However, compared to filtered conditions, gene change was greater when exposure was without the filter. Given TrkB essentially has a survival and protector role, the increased levels of TrkB could be related to retinal protection against phototoxicity [40]. Interestingly, maximum TrkB increases were obtained in rats exposed to light in unfiltered conditions, with a higher proportion of short wavelengths, the most deleterious for the retina [41]. Therefore, an increase in TrkB expression, after LED-light exposure, could be considered as a neuroprotection response. On the other hand, photoreceptor integrity and survival could be irreversibly compromised by the short-term stress of the RPE. Exposure to intense visible light activates the VEGF signal and, thus, the breakdown of the outer blood–retinal barrier. The resulting permeability increase seems to induce photoreceptor apoptosis due to light exposure. The breakdown and successive apoptosis effects can be prevented by inhibiting VEGF [42]. Results obtained in our study show an increase in VEGF expression in the retinas of rats exposed to LED-backlit tablet screen light. However, results obtained with the absorbing filter, which partially filters short wavelengths, do not show VEGF expression differences compared to control group values. However, it is important to point out that the toxic outcome of VEGF on photoreceptors is not a direct effect, but secondary to VEGF-induced RPE permeability. It seems plausible that after exposure to high levels of light, the breakdown of the epithelium integrity alters the metabolite exchange between retina and choroid, which, in turn, may further affect photoreceptors [42].

We also found after up-regulation of TNFα, SOD1, and SOD2 after exposition to LED tablet screen light. As far as we know, the regulation of the expression of TNFα by light has not been previously reported; nevertheless, this could be in agreement with the inflammatory mechanism that underlines the age-related macular degeneration induced or exacerbated by blue light [43]. Finally, the increase in SOD1 and SOD2, two enzymes intimately related to oxidative stress after light exposure, suggests a contribution to the whole retinal damage. Shang et al. observed free radical production in the retina after LED-exposition [9]. Similarly, Jaadane et al. and Nakamura et al. indicated that oxidative stress was partially involved in blue LED light-induced retinal damage [2,7].

Several studies have speculated on the damaging effects that short wavelength radiation can cause on the eye and the possible protective effect of optical filters that selectively absorb this light. In 2004, Sparrow et al. already suggested that a blue light partial absorption filter reduces approximately 80% and 78% in the death of the RPE exposed to blue (430 nm) and white light ( 5400 ◦K) [32]. Other studies, such as those published by Yanagi et al. in 2006 and by Hui et al. in 2009, demonstrated that the blue light partial absorption filter protects the RPE from the damage produced by short wave visible radiation, increasing cell viability by 42% and 79.5%, respectively [44,45]. On the other hand, in 2011, Zhou et al. observed that the viability of cells exposed to 430 ± 20 nm was reduced by 40% [46]. To evaluate possible artificial protection mechanisms, they interposed optical filters with different levels of a pigment that absorbs blue light and verified that the protection provided by the filter was a function of its absorbance. In 2013, Chamorro et al. published a study whose results showed that the absorption of blue light decreases apoptotic cell death by 50–89% and inhibits DNA damage by 57–81% [3].

Although all these studies have results consistent with ours, the research carried out by Chamorro et al. stands out, since, in their experiment, they used LED light sources, just like us. However, our study presents two important differences with theirs when analyzing the results. First, their light exposure system consisted of an experimental device with direct emission LEDs, while our lighting system consisted of LED-backlit screens, currently commercialized, which emit white light diffusely. It should be noted that, under normal conditions, a person can spend long hours watching screens and not specific sources of LEDs, which gives our study greater similarities with real experiences. Second, Chamorro et al. performed an in vitro experiment exposing an RPE cell culture to LED light. However, our study of animal experimentation has allowed us to expose the eyes to the LED radiation, keeping intact the ocular optical systems of protection, such as, for example, the filtering of light provided by the cornea, the lens, and the intraocular media. In addition, in our study, the retinal tissue analyzed was contained in the eye of a living animal during exposure to light, therefore, it kept intact the regeneration mechanisms of photopigments and cells that the eye presents under normal conditions of life during the exposure period. This gives our results greater consistency with the effects that blue light and selective absorbance filters for this radiation can produce on living retinal tissue.

In this study, we assessed genes involved in retinal damage in the classic model of pigmented rats. Although extrapolation of the results obtained in rats to humans is questionable, it is useful to gain knowledge on the type of detrimental effects that the new LED-backlit screens can generate. Additionally, it is important to consider that results obtained in in-vitro studies with human retinal cells are similar to those obtained in our study. Along this same line, Nagai et al. have shown that blue-light filtering intraocular lenses that absorb a percentage of short-wavelengths that reach the human retina, decrease the incidence of retinal degeneration [33].

Although further investigations are still needed, there is evidence about how LED light damage depends on the exposure time and the wavelength, and involves different pathways related to oxidative stress, inflammation, ECM degradation, regulation of apoptotic genes, and growth factors. As LED-backlit screens are watched at a short distance, we stress the importance of controlling excessive exposure to LED light currently imposed by the widespread use of digital devices such as computers, laptops, tablets, smartphones, etc. The use of short-wavelength selective filters on screens may contribute to reducing potentially irreversible damage to the human retina, by partially cutting off the excessive and most energetic blue-light emissions.

**Author Contributions:** Conceptualization, C.S.-R., V.A.-V. and J.A.V.; Data curation, T.C. and O.G.-S.; Formal analysis, V.B.-S., T.C. and O.G.-S.; Investigation, C.B.-A., T.C., O.G.-S. and C.A.-P.; Methodology, C.B.-A. and M.J.P.-C.; Project administration, C.S.-R. and J.A.V.; Resources, T.C. and M.J.P.-C.; Software, V.B.-S. and O.G.-S.; Supervision, C.S.-R., C.B.-A., V.A.-V. and J.A.V.; Validation, C.S.-R. and C.A.-P.; Visualization, M.J.P.-C.; Writing–original draft, C.A.-P.; Writing–review & editing, C.S.-R., C.B.-A., V.B.-S., V.A.-V., T.C., M.J.P.-C., O.G.-S. and J.A.V. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** This study has the approval of the Animal Experimentation Committee of the UCM and of the Department of Health of the Comunidad de Madrid, Spain (Reference PROEX 310/15).

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data that support the findings of this study are available from corresponding author upon reasonable request.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Cells* Editorial Office E-mail: cells@mdpi.com www.mdpi.com/journal/cells

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18

www.mdpi.com ISBN 978-3-0365-2652-2