Associations Between Diabetic Retinal Microvasculopathy and Neuronal Degeneration Assessed by Swept-Source OCT and OCT Angiography

Purpose: To provide clinical evidence of the associations between retinal neuronal degeneration and microvasculopathy in diabetic retinopathy (DR). Methods: This case-control study included 76 patients (76 eyes) with type 2 diabetes mellitus (DM), and refraction error between −3.0 and +3.0 D. The eyes were assigned into DM (without DR), non-proliferative DR (NPDR), and proliferative DR (PDR) groups. Age-, sex-, and refractive error-matched normal subjects were enrolled as controls. The mean retinal thickness (mRT), the relative mean thickness of the retinal nerve fiber layer (rmtRNFL, mtRNFL/mRT), ganglion cell layer (rmtGCL), ganglion cell complex (rmtGCC) layer, foveal avascular zone area (FAZa), FAZ perimeter (FAZp), FAZ circularity index (FAZ-CI), and vessel density (VD) in superficial capillary plexus (SCP) and deep capillary plexus (DCP) were assessed by swept-source optical coherence tomography (OCT) and OCT angiography (OCTA). Group comparison and Spearman's partial correlation coefficient analysis were applied to evaluate the correlation between these morphological parameters. Results: rmtRNFL, FAZa, and FAZp in SCP and DCP increased with the DR severity (prmtRNFL < 0.001; pFAZa, SCP = 0.001; pFAZa, DCP = 0.005; pFAZp, SCP < 0.001; pFAZp, DCP < 0.001). The rmtGCL, FAZ-CI in SCP and DCP, and VD in DCP decreased with the DR severity (prmtGCL = 0.002, pFAZ−CI, SCP = 0.002; pFAZ−CI, DCP < 0.001, pVD, DCP < 0.001). After controlling age, sex, duration of diabetes, and hypertension, the rmtRNFL, FAZa in SCP and DCP, and FAZp in SCP and DCP were correlated with the severity of DR (p < 0.05), while VD in SCP and DCP, FAZ-CI, and rmtGCL were negatively correlated with the severity of DR (p < 0.05). The rmtGCL was negatively correlated with the FAZa in SCP (r = −0.34, p = 0.002) and DCP (r = −0.23, p = 0.033), and FAZp in SCP (r = −0.37, p = 0.001) and DCP (r = −0.32, p = 0.003), but positively correlated with VD in SCP (r = 0.26, p = 0.016), VD in DCP (r = 0.28, p = 0.012), and FAZ-CI in DCP (r = 0.31, p = 0.006). Conclusions: rmtRNFL, FAZ-CI in SCP and DCP, and FAZp in SCP are strong predictors of the severity of DR. The ganglion cell body loss is highly correlated with increased FAZp and FAZa, decreased FAZ-CI, and reduced VD with the severity of DR.


INTRODUCTION
Diabetic retinopathy (DR) is among the most significant and disabling chronic complications of diabetes mellitus (1,2). Although cumulative evidence from basic science and clinical research suggested that DR is a neurovascular disease (3,4), the correlation between diabetic retinal microvasculopathy and neuronal degeneration still remains uncertain, and their synergistic effect is yet to be illustrated in humans.
In previous studies, retinal neuronal degeneration and microvascular dysfunction have been shown to promote the occurrence and development of DR (5)(6)(7)(8). We further confirmed that retinal ganglion cell is the most susceptible retinal neuronal cell type in response to hyperglycemia in a diabetic animal model (3). Clinical studies further confirmed that retinal neurons dysfunction occurred before retinal microvasculopathy using electroretinography (ERG) and visual field examinations (9)(10)(11). However, the underlying mechanisms and the correlation between diabetic retinal microvasculopathy and neuron degeneration are yet uncertain. In this study, advanced swept-source optical coherence tomography (SS-OCT) and OCT angiography (OCTA) were used to further investigate the correlation between the early and sensitive parameters of retinal microvasculopathy including the foveal avascular zone area (FAZa), FAZ perimeter (FAZp), FAZ circularity index (FAZ-CI), and retinal vessel density (VD) in superficial capillary plexus (SCP) and deep capillary plexus (DCP), and the parameters for evaluation of the early morphological changes of ganglion cells including the relative mean thickness of the retinal nerve fiber layer (rmtRNFL, mtRNFL/mRT), the relative mean thickness of the ganglion cell layer (rmtGCL), and the relative mean thickness of the ganglion cell complex layer (rmtGCC) in DR.

Participants
This study followed the Ethical Principles for Medical Research Involving Human Subjects of the Declaration of Helsinki and was approved by the Ethics Committee of Beijing Tongren Hospital, Capital Medical University. All the enrolled subjects signed an informed consent form before participation.

Sample Size Calculation
Combined with our pilot study results, which enrolled at least 15 subjects per group to show the statistical significance, Cochran's sample size formula was applied and the previous studies were referenced (27). Sample size was calculated at 95% confidence level with a margin of error of ±5%. We determined the minimum sample size to be 20 subjects in each group to detect the difference (0.83) between means of the OCT and OCTA parameters with a significance level of (alpha) 0.05.

Inclusion and Exclusion Criteria
The inclusion criteria were as follows: (1) participants aged 40-70 years with T2DM; (2) refractive errors between −3.00 D and +3.00 D. T2DM was defined according to the American Diabetes Association (ADA) guidelines of DM (28)(29)(30)(31)(32)(33), while DR was defined and classified by ophthalmologists according to the 2017 "Diabetic Retinopathy: A Position Statement of Diabetic Retinopathy" by the American Diabetes Association (34). Age-, sex-, and refractive error-matched normal subjects were enrolled as controls.
The exclusion criteria were as follows: (1) cysts detected in the neural retina by OCT; (2) macular edema secondary to any other retinal vascular diseases; (3) patients with other fundus diseases, such as age-related macular degeneration, uveitis, and retinal inherited diseases; (4) eyes with history of posterior surgery within 1 year; (5) eyes that underwent pan-retinal photocoagulation or focal/grid laser treatment; (6) eyes treated with intravitreal anti-vascular endothelial growth factor (VEGF) agent or steroid within one year; (7) patients with glaucoma or ocular hypertension; (8) fundus examination could not be performed due to ocular media opacities; (9) patients could not tolerate examinations due to severe systemic diseases.

Detection and Quantification of the Morphological Parameters by OCT
Imaging with DRI-Triton OCT: All subjects were examined by SS-OCT (DRI OCT1 Atlantis scanner, Topcon Co., Ltd., Tokyo, Japan). A 9 × 9 mm scanning range mode was selected to obtain B-scan images. The thickness of the retinal nerve fiber layer (RNFL), ganglion cell layer (GCL), and ganglion cell complex (GCC) in nine macular regions was measured automatically using TOPCON Advanced Boundary Segmentation (TABS) software. The nine regions were designed according to the Early Treatment Diabetic Retinopathy Study Classification (ETDRS) system. The ETDRS map was divided into nine parts: one 1 mm diameter circle at the center, outside four inner-quarter annuluses on the outside between the 1 and 3 mm diameter circles, and four outer-quarter annuluses between the 3 and 6 mm diameter circles (Figure 1). All images were measured independently by two ophthalmologists. Computer-aided manual correction of OCT segmentation was used for correcting thickness measurements in cases with errors.

Detection and Quantification of the OCTA Parameters
Optical coherence tomography angiography images were obtained using a 6 × 6 mm scanning mode in the macular area. Individual images in SCP, DCP, outer retina, and choriocapillaris plexus were generated by Topcon IMAGEnet R 6 software with the computer-aided manual correction in cases with errors (35). FAZa and FAZp in SCP and DCP were measured by Image J software (version 1.48, available online at http://rsb.info.nih.gov/ ij/, National Institutes of Health, Bethesda, MD, USA). Foveal avascular zone circularity index (FAZ-CI) (36) is an indicator and novel biomarker representing disruption of the parafoveal capillary network. The shape of FAZ is closer to a regular circular shape when FAZ-CI is close to 1. The formula of the FAZ-CI is: OCTA images were binarized with Ostu threshold method using Image J software (37). The VD in superficial and deep capillary networks was measured by calculating the proportion of vessel area with blood flow over the total area measured (38) (Figure 2).
To avoid individual differences and other confounding factors, we used the mean value (the average of ETDRS nine subfields) and relative thickness to describe the morphological parameters of retinal ganglion cells determined by SS-OCT. The relative mean thickness of RNFL (rmtRNFL), GCL (rmtGCL), and GCC (rmtGCC) was defined as the mean thickness of RNFL, GCL, or GCC/mean retinal thickness, respectively.

Statistical Analysis
Statistical analysis was performed using SPSS software (SPSS, Inc. 24.0, Chicago, IL, USA). Age of patients, the duration of diabetes, the FAZa, rmtRNFL, rmtGCL, and rmtGCC in each group were described by means ± standard deviation (mean ± SD) or median (interquartile range). The comparisons between the groups were analyzed by one-way analysis of variance (ANOVA) or Kruskal-Wallis test. Spearman's partial correlation analysis was used to analyze the correlation between the parameters of retinal degeneration and microvasculopathy or between various morphological parameters with the severity of DR after controlling the age, gender, duration of diabetes, and hypertension. A value of P < 0.05 was considered statistically significant.
No statistically significant difference was found in the gender and age among the four groups (p gender = 0.633, p age = 0.086). No significant difference was observed in the duration of DM in the DM, NPDR, and PDR groups (p = 0.171). Also, no significant difference was noted in the fasting blood glucose in the DM, NPDR, and PDR groups (p = 0.658). The glycosylated hemoglobin levels in the DM, NPDR, and PDR groups did not differ significantly (p = 0.464). Furthermore, no significant difference was detected in the duration of hypertension among the four groups (χ 2 = 2.20, p = 0.333) ( Table 1). FIGURE 1 | Measurement and analysis of the thickness of RNFL, GCL, and GCC layers. The ETDRS map (nine subfields) was utilized to calculate the average thickness of the RNFL, GCL, and GCC layers. RNFL, retinal nerve fiber layer; GCL, ganglion cell layer; GCC, ganglion cell complex layer; ETDRS, the Early Treatment Diabetic Retinopathy Study.

Correlations of the Mean Macular Thickness With Diabetic Retinopathy Severity
According to the general trend, the mean macular thickness increased with DR severity. However, there was no statistically significant difference observed between the normal (277. 60

Relative Mean GCL Thickness Decreased With Diabetic Retinopathy Severity
The rmtGCL thickness (GCL/RT) layer was 0.27 ± 0.01 in the DM group, 0.27 ± 0.02 in the NPDR group, 0.25 ± 0.02 in the PDR group, and 0.26 ± 0.02 in the normal control group, respectively. The rmtGCL thickness was significantly lower in the PDR group than that in the normal control, DM, and NPDR FIGURE 2 | Representative OCTA images illustrate the quantification methods of VD, FAZa, FAZp, and FAZ-CI in SCP and DCP. OCTA images were obtained using a 6 × 6 mm scanning mode in the macular area. Individual images in SCP, DCP, outer retina, and choriocapillaris plexus were generated by Topcon IMAGEnet ® 6 software with the computer-aided manual correction in cases with errors in eyes with normal control (healthy subjects), DM (no DR), NPDR, and PDR. The FAZa and perimeter (yellow circle) were quantified by Image J software. Ostu method was applied to calculate the VD using Image J software. OCTA, optical coherence tomography angiography; VD, vessel density; FAZ, foveal avascular zone; FAZ-CI, FAZ circularity index; SCP, superficial capillary plexus; DCP, deep capillary plexus; DM, diabetes mellitus; DR, diabetic retinopathy; NPDR, non-proliferative diabetic retinopathy; PDR, proliferative diabetic retinopathy.  Table 2).  Table 2).

Correlation Between Microvasculopathy and Neuronal Degeneration Among Different Groups
After controlling the age, sex, duration of diabetes, and hypertension, Spearman's partial coefficient correlation analysis showed that the rmtRNFL (r = 0.52, p < 0.001); FAZa in SCP

DISCUSSION
In this study, we found that increased rmtRNFL, FAZa, and FAZp, decreased rmtGCL, VD, and FAZ-CI are correlated with the severity of DR. rmtRNFL, FAZ-CI in SCP and DCP, and FAZp in SCP are stronger predictors of the severity of DR than the FAZa. The ganglion cell body loss (rmtGCL) is highly correlated with increased FAZa and perimeter, decreased FAZ-CI, and reduced VD with the severity of DR. These results indicated that diabetic microvasculopathy is highly correlated with neuronal degeneration, and neuron protection is an essential strategy to improve prevention and management in DR (3).
Neurovascular unit (NVU) is a new concept in neuroscience introduced by the Stroke Progress Review Group (39). The NVU is a complex structural and functional unit that regulates regional cerebral blood flow (CBF) and the delivery of nutrients (40). The dysfunction of NVU has been implicated in several diseases, such as stroke and Alzheimer's disease (41). Based on the concept of brain NVU and the close correlation between the retina and the brain in all aspects of function and anatomy, the concept of retinal neuronal vascular unit (RNVU) was proposed (42,43). Neurons and endothelial cells are the critical components of RNVU. The interactions between neurons and microvasculature in DR have gained increasing attention in recent years. In this study, we found that the ganglion cell body loss is highly correlated with the retinal vasculature changes (increased FAZp, decreased FAZ-CI, and reduced VD) with the severity of DR, which is consistent with previous reports (44,45).
A ganglion cell is mainly composed of a cell body (soma), a single axon, and dendrites. The electrical impulses transmitted by neurons mainly depend on axon transport, the process by which nerve cells transfer substances from the cell body to the axon tip. This ATP-dependent process plays an important role in maintaining the physiological activities of neurons and can occur in two delivery directions: anterograde (from the cell body to the axon tips) and retrograde (from the axon tips back to the cell body) transport. The microtubule cytoskeleton and motor proteins are critical for long-range axon transport.
The axons of the GCL constitute RNFL. Increasing evidence speculated that mitochondrial dysfunction and axoplasmic flow stasis in optic neuropathy (44,45) lead to axonal swelling and contribute to most of the neurodegenerative diseases including DR (46)(47)(48)(49). Eyes with early Leber Hereditary Optic Neuropathy (E-LHON) showed a thicker RNFL, but a thinner RNFL was seen in eyes with atrophic LHON (A-LHON) in comparison with the control group, thereby suggesting that the changes in RNFL thickness represent the different course of the disease (50). In a diabetic rat model, expression level of the synaptic proteins including vesicular glutamate transporter-1, syntaxin-1, and synaptotagmin-1 was decreased in the GCL, but the level of dynein motor proteins increased in the GCL after 8 weeks of diabetes, indicating that axonal transport contributes to early signs of neural dysfunction in the diabetic rat retinas (51). Furthermore, the degree of increased thickness of RNFL is closely correlated with visual evoked potential (VEP)    amplitude loss and degree of optic atrophy, indicating that during the progression of DR, a thinner RNFL is a sign of optic atrophy following the swelling stage of axons of ganglion cells (52), and the degree of thickened RNFL may be an indicator of optic atrophy for patients with DR. Furthermore, in a primate model of human non-arteritic anterior ischemic optic neuropathy, RNFL swelling is confirmed as a strong and independent predictor of optic nerve atrophy, which agrees with our results (52). In this study, we have shown that the relative thickness of RNFL increased with DR severity concomitant with the decreased thickness of the GCL layer. This phenomenon suggested that the swelling of the GCL axon preceded the GCL cell body loss under hyperglycemia, thereby reflecting the unique hypoglycemia-mediated pathological damage of GCL. The pathological evidence, either in animal models or humans, is consistent with the findings of the current study (53)(54)(55)(56)(57). Furthermore, FAZp, FAZ-CI, and VD are good sensors of retinal ischemia and have been implicated in several clinical studies to predict disease progression (58)(59)(60). Increased FAZp, FAZ-CI, and VD have been found to be an earlier sign of diabetic retinal vasculopathy (60)(61)(62) and are highly correlated with the severity of DR (12,62). In comparison with FAZa, which significantly varies in the population, FAZ-CI is a better measurement index to assess FAZ circularity with a value range from 0 to 1 (13). FAZ-CI has been found to be tightly correlated with visual acuity in patients with DR (13,14). We also found that enlargement of FAZp in SCP and DCP was correlated with DR severity as well as the duration of diabetes, which is in parallel to the previously published data (15)(16)(17). In previous studies, SCP and DCP vessel density was demonstrated to be decreased in DM (18) and highly correlated with DR progression (19), which are consistent with our results. We further demonstrated that microvascular injury is associated with GCL cell body loss in DR, providing the evidence for the first time that there is an interaction between retinal microvasculopathy and neuronal degeneration. Interestingly, we also found that in comparison with RNGL, GCL has a closer relationship with the vasculature ischemic parameters (FAZp, FAZ-CI, VD), indicating that the ganglion cell body is a good sensor for retinal ischemia.
The rapid revolution of OCT and OCTA has provided a powerful weapon for evaluating the morphological and structural changes in the various layers of the retina (20)(21)(22). Compared to traditional OCT, SS-OCT provides longer and deeper detection ability, ultra-high image acquisition speed, and a more powerful function of eliminating motion artifacts, thereby providing a distinct retinal microstructure (20,(23)(24)(25)(26). OCTA has the advantage of investigating retinal vasculature non-invasively layer-by-layer (Figure 2). Automatic quantitative OCTA metrics have provided valuable information for disease progression beyond that could be obtained from fundus fluorescein angiography (FFA) (63).
In this study, to avoid the individual differences, the intereye correlation, and the other confounding factors, we used the following: (1) average value of ETDRS nine subfields to describe the thickness of the retina, RNFL, GCL, and GCC (64,65); (2) all the values were automatically provided by Topton triton SS-OCT TABS software; (3) the relative mean thickness of RNFL, GCL, and GCC was defined as the mean thickness of RNFL, GCL, or GCC/mean retinal thickness. Gabriele et al. described a relative method in which RNFL thickness profiles relative to the distance from the disc center were computed for quadrants and clock hours (66). In our pilot study, a linear trend was established among the four groups, which was similar between the relative mean ETDRS macular thickness and absolute thickness; however, data distribution was based on relative thickness and indicated the differences among the groups.
Although the current study successfully showed the correlation between diabetic retinal vasculopathy and neuronal degeneration, these findings are limited due to the casecontrol design, and the causal relationship between neuronal degeneration and microvasculopathy cannot be determined, hence future validation using longitudinal data is warranted. Furthermore, the small sample size could not eliminate some confounding factors in the analysis, and so well-designed and large-scale prospective clinical trials are also needed to support the conclusions in the study. Morphological (OCT/OCTA) and functional evaluations (ERG) of neurons in a cohort study could also be considered to further determine the pathological changes in other neuronal cell types under hyperglycemia.
In summary, we found that there was a close correlation between diabetic retinal microvasculopathy and neuronal degeneration with respect to the occurrence and severity of DR in humans. The current results also hinted that neuroprotection is also critical for the prevention and managing of DR. Further, well-designed, large cohort studies are warranted to validate the findings of this study.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/supplementary materials, further inquiries can be directed to the corresponding author/s.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Beijing Tongren Hospital, Capital Medical University. The patients/participants provided their written informed consent to participate in this study.

AUTHOR CONTRIBUTIONS
XZ contributed to the conception and design of the study and revision of the manuscript. BQ and LZ organized the database