Effects of high-fat diet and Apoe deficiency on retinal structure and function in mice

To investigate the effects of a high-fat diet (HFD) and apolipoprotein E (Apoe) deficiency on retinal structure and function in mice. Apoe KO mice and wild-type C57BL/6J mice were given a low-fat diet (LFD) or a HFD for 32 weeks. Blood glucose, serum lipids, body weight and visceral fat weight were evaluated. Retinal sterol quantification was carried out by isotope dilution gas chromatography-mass spectrometry. The cholesterol metabolism related genes SCAP-SREBP expressions were detected by qRT-PCR. Retinal function was recorded using an electroretinogram. The thickness of each layer of the retina was measured by optical coherence tomography. Fundus fluorescein angiography was performed to detect retinal vasculature changes. Immunohistochemical staining was used to determine the expression of NF-κB, TNF-α and VEGFR2 in the retina among HFD, HFD Apoe−/−, LFD Apoe−/− and WT mice retinas. HFD feeding caused the mice to gain weight and develop hypercholesterinemia, while Apoe−/− abnormalities also affected blood lipid metabolism. Both HFD and Apoe deficiency elevated retinal cholesterol, especially in the HFD Apoe−/− mice. No up-regulated expression of SCAP-SREBP was observed as a negative regulator. Impaired retinal functions, thinning retinas and abnormal retinal vasculature were observed in the peripheral retinas of the HFD and Apoe−/− mice compared with those in the normal chow group, particularly in the HFD Apoe−/− mice. Moreover, the expression of NF-κB in the retinas of the HFD and Apoe−/− mice was increased, together with upregulated TNF-α mRNA levels and TNF-α expression in the layer of retinal ganglion cells of the peripheral retina. At the same time, the expression level of VEGFR2 was elevated in the intervention groups, most notably in HFD Apoe−/− mice. HFD or Apoe gene deletion had certain adverse effects on retinal function and structure, which were far below the combined factors and induced harm to the retina. Furthermore, HFD caused retinal ischemia and hypoxia. Additionally, Apoe abnormality increased susceptibility to ischemia. These changes upregulated NF-κB expression in ganglion cells and activated downstream TNF-α. Simultaneously, they activated VEGFR2, accelerating angiogenesis and vascular permeability. All of the aforementioned outcomes initiated inflammatory responses to trigger ganglion cell apoptosis and aggravate retinal neovascularization.

Because the levels of esterified cholesterol were very low in mice retinas, unesterified (free) and total cholesterol were evaluated. The levels of retinal cholesterol (unesterified and total cholesterol) were increased compared to wild-type animals although they did not achieve statistical significance. In HFD Apoe−/− mice, the retinal cholesterol contents were much higher than other three groups.  Fig. 2A).
To explore more details, the mean thickness was then computed for each of the layers in eight regions, namely, the central superior (CS), central nasal (CN), central inferior (CI), central temporal (CT), paracentral superior (PS), paracentral nasal (PN), paracentral inferior (PI) and paracentral temporal (PT) regions ( Fig. 3C(f)). The circles used were 0.2 mm, 0.6 mm and 1.2 mm in diameter, respectively. The value in each region of the heat map represents the mean thickness of the retinal area (Fig. 5).
These data indicated that the HFD alone could significantly reduce each layer's thickness of the peripheral retina of the mice, especially for the composite layer (NFL + GCL + IPL). The hypercholesterolemia caused by Apoe gene knockout was not sufficient to change the thickness of the peripheral retina of the mice. However, double factors further reduced the thickness of the NFL + GCL + IPL in the peripheral retina. These results indicate that hypercholesterolemia caused by HFD or Apoe gene deletion was sufficient to cause blood vessel abnormalities in the peripheral retina. Capillary density and nonperfusion zones were more significantly increased in the HFD Apoe −/− mice than in the age-matched HFD C57BL/6J mice or LFD Apoe −/− mice.

Chronic HFD and/or Apoe deficiency impaired retinal function. Chronic HFD and Apoe deletion
were capable of inducing subtle changes in retinal structure, but it remains unknown whether these two factors could cause malfunction of the retina. Therefore, the retinal function in the mice was tested with ERG, and the impairment of retinal function was observed. Dark-and light-adapted ERGs were performed in the 32nd week.
The amplitudes of the a-wave and b-wave, which represent the functions of the photoreceptors and bipolar cells, respectively (Table 2), were significantly decreased in the HFD Apoe −/− mice, HFD-C57BL/6J mice and LFD Apoe −/− mice-especially the HFD Apoe −/− mice (Fig. 7A, B). Further, chronic HFD and Apoe deletion induced even more decreased amplitudes of the ERG a-and b-wave (Fig. 7C).
Considering the effect of intraocular pressure (IOP) on the mouse retina, we checked the IOP at the 6th week and 32nd week in each group with an Icare tonometry (Icare ic200, NC, USA). No significant changes were shown in IOP among all of the groups.  www.nature.com/scientificreports/ the RGCs activated the expression of downstream TNF-α. The TNF-α expression in the retinal NFL + GCL + IPL composite layer of the HFD C57BL/6J mice and the LFD Apoe −/− mice was increased, especially in the HFD Apoe −/− mice (Fig. 8C). Further, RT-PCR analysis confirmed that the high-fat diet and Apoe gene knockout induced TNF-α mRNA levels that were significantly increased compared to those in the normal control group (Fig. S1). The increase in inflammatory agents directly acted on RGCs to aggravate cell death.
In addition, as we mentioned, neovascularization in the retinal vasculature was noted among the different groups. It is known that retinal neovascularization is directly related to vascular endothelial growth factor (VEGF) expression, The latter caused angiogenesis and increased vascular permeability. We further examined the expression of VEGFR2 in the central and peripheral areas of the retina. With hematoxylin-eosin (H&E) staining, there were no differences in the expression of VEGFR2 in the HFD Apoe −/− mice, LFD Apoe −/− and HFD C57BL/6J mice www.nature.com/scientificreports/ compared to the normal control group in the central retina. Interestingly, in the peripheral area of the retina, the expression levels of VEGFR2 in the LFD Apoe −/− and HFD C57BL/6J mice (LFD Apoe −/− : 8.00 ± 0.76/0.004 mm 2 , n = 8; HFD C57BL/6J: 11.13 ± 1.13/0.004 mm 2 , n = 8) were significantly increased compared to those in the LFD C57BL/6 mice (6.00 ± 0.93/0.004 mm 2 , n = 8) (P1 = 0.001, P2 < 0.0001). In the HFD Apoe −/− mice, the increase in VEGFR2 expression was more significant (P4 < 0.0001) (Fig. 8D).
The results showed that both HFD and Apoe −/− increased the expression of NF-κB in the RGCs and activated the expression of downstream TNF-α, further promoting RGC apoptosis. No further changes in VEGFR2 expression in the central retina were induced by either a single factor or two factors. However, HFD or Apoe −/− increased the expression of VEGFR2 compared to the normal controls, and the expression of VEGFR2 increased more notably under the two factors together in the peripheral retina.

Discussion
As we known, HFD animals have been used as models to study type 2 diabeticmellitus (T2DM), metabolic syndrome, and obesity for many years 2,11,12 . In recent years, the effect of HFD on the retinal structure and function has become a popular topic 13 . In the human body, the APOE gene is located on chromosome 19 14 . APOE is a Table 2. Amplitude changes of a-wave and b-wave under different stimulus intensity under dark adaptation and light adaptation.   www.nature.com/scientificreports/ lipid-related protein that is present in both the serum and CNS 15 . As a ligand for low-density lipoprotein receptors, it is involved in chylomicron and very low-density lipoprotein clearance, and it maintains the body lipid metabolism balance. This protein not only plays an important role in lipid metabolism but also participates in other important biological functions, such as immune regulation and neurological pathway regulation (neuron repair and remodeling). Therefore, in the population, APOE gene abnormalities could lead to cardiovascular 16 , cerebrovascular and nervous system-related diseases 17 . For these reasons, this gene has become a new target for basic research and clinical treatment 18 . Apoe-deficient (Apoe −/− ) mice spontaneously developed hypercholesterolemia, atherosclerosis and retinopathy 6 . The effect of cholesterol abnormalities caused by Apoe gene deletion on the retina has also been a research focus. In our study, we used the HFD-fed mice and the Apoe gene knockout mice to observe the effects of diet and genes on the structure and function of the retina. First of all, compared with wild-type mice fed a normal diet, the body weight and the amount of VAT in the HFD C57BL/6J and HFD Apoe −/− mice were significantly increased. In the HFD Apoe−/− mice, serum lipid and systemic glucose levels were highest among the four groups. Although the cholesterol levels of the HFD C57BL/6 and LFD Apoe −/− mice were higher than in the control group, they did not achieve the level of the HFD Apoe −/− mice, consistent with previous studies by Plump and Zhang 19,20 . Interestingly, due to gene defects, the blood lipids and blood glucose of the LFD Apoe −/− mice were higher than in the control group, while the body weights were similar to those in the control group. These results indicated that hyperglycemia or hyperlipemia did not necessarily go hand in hand with obesity. Then, we measured the retinal lipid profile. We found that the retinal sterol profile was elevated, especially in HFD-ApoE−/− mice. As a negative regulator, no significant up-regulation of SCAP-SREBP expression was observed. These results were consistent with Nicole El-Darzi's study 21 .
Next, we further demonstrated that a 32-week feeding period of the high fat diet to Apoe-deficient mice and wild-type C57BL/6 mice resulted in dramatic reduction in the a-and b-wave amplitude of ERG compared to that in regular mice, indicating that HFD impacted the function of the retina. These results were consistent with the results of most studies 22,23 . Some other studies showed that only decreased amplitude of the a-and b-waves prolonged the latency of ERG as seen in the HFD Apoe −/− mice, whereas the retinal light response of the HFD C57BL/6J and LFD Apoe −/− mice was similar to that of the normal control group (LFD C57BL/6J) 22 . Controversially, in the present study, there was still apparently decreased a-and b-wave's amplitude of ERG in the LFD Apoe −/− mice compared with that of the normal control group, although the reduced amplitude of ERG was not as significant as in the HFD Apoe −/− mice. This outcome suggested that hypercholesterolemia might play a major role in damage to neuroretinal cells and impairment of retinal function among these risk factors.
Third, to distinguish different thicknesses between different regions, we examined each layer thickness of the central and peripheral retina using SD-OCT 24 . We found that, for the central retina, diet and gene deficiency have little effect on the thickness of each retinal layer. This result agreed with the study by Ritland JS 25 , perhaps because of the rich blood supply, good nutrients and lack of aggregation of inflammatory cytokines in the central area. This hypothesis was confirmed in our subsequent findings on fundus fluorescence angiography (FFA), which showed fewer changes occurring in blood vessels (vascular area, percentage of vascular retina, number of total branch points and nonperfused area) in the HFD or Apoe −/− group than in the normal control group in the central retina. In the peripheral retina, the thickness of the NFL + GCL + IPL composite layer in the HFD and Apoe −/− mice was significantly thinner, and that in the HFD Apoe −/− mice was much thinner. Since the retinal vessels are mainly located in the NFL layer, FFA tests were more sensitive in detecting the vessel distribution and blood perfusion in this specific structure. These OCT changes were also supported by the FFA results. We found that the percentage of vascular retina and number of total branch points in the HFD C57BL/6J, LFD Apoe −/− and HFD Apoe −/− mice were all increased. These findings might be related to the microvessels supplying the peripheral retina, rendering the peripheral retina more prone to ischemia and hypoxia. It could not excrete metabolites in a timely manner and received less nutrition in this state. Because of rod cells mainly located in the peripheral retina, night blindness can also occur when photoreceptors dysfunction.
The recent study by Bright Asare-Bediako revealed that the HFD mouse is a useful model for examining the effects of prediabetes and hypercholesterolemia on the retina 26 . Interestingly, the severity of retinopathy depended on HFD feeding duration 12 . It was considered that retinal neovascularization occurs in mice after 7 months HFD feeding 27 . Bright Asare-Bediako et al. 26 found that C57BL/6 mice fed with HFD showed a reduction in a-wave and b-wave amplitudes at 6 months and increased retinal nerve infarcts and vascular leakage, as well as reduced vascular density at 12 months. This evidence supported our findings. The HFD-induced changes appeared to occur slower than those observed in T2DM models but were consistent with other retinopathy models, showing neural damage prior to vascular changes. This result falled in line with our findings 26 . We confirmed that a 32-week HFD (8 months) could induce both retinal vascularization and neurologic changes.
Finally, the underlying mechanism for vascularization and neurologic changes were detected. A large aggregation of studies has focused on underlying mechanisms caused by HFD. Mykkänen et al. believed that 12-week HFD feeding could induce differential products of a variety of stress-related genes in the retina, such as nitric oxide synthase and hydroxynonenal, leading to retinal degeneration 28 . The other study suggested that changes in the diabetic retinal fundus caused by HFD were due to decreased glucose tolerance and enhanced insulin resistance 29 . These pathological changes were related to TLR4-dependent macrophages or microglia activation and perivascular oxidative stress marker upregulation 11 . Chang et al. observed that 3-6 months of HFD feeding impaired retinal function through calcium signaling abnormalities and microglial activation. The reduction in ERG components in early diabetic retinopathy reflected the reduction in neuronal activity in retinal light responses, which might have occurred due to decreased calcium signaling transduction in neurons based on changes in PI3K-Akt signaling in L-type voltage-gated calcium channels and plasma membrane Ca 2+ -ATPase pathway regulation ability 30  www.nature.com/scientificreports/ APOE contains four subtypes, and type 4 (Apoe4) works on blood vessels, which might be associated with VEGF expression downregulation 31 . Hypercholesterolemia caused by Apoe gene knockout partially or completely blocked the retinal vessels and upregulated the expression of VEGF, which resulted in dramatic retinal neovascularization 23 . All of the aforementioned findings indicated that APOE has a certain protective effect on retinal vessels. Chrysostomou et al. also suggested that the outward growth of glia-related synapses seems to be mediated by Apoe partially 32 . Apoe deletion led to a decrease in the number of activated glial cells and a reduction in the outward growth of RGC nerve processes, hinting that Apoe had a protective effect on retinal nerve cells 33,34 . Drouet et al. 35 pointed out that lipid-bound Apoe, not lipid-free Apoe, can protect neurons from apoptosis. They believed that Apoe was necessary for the prevention of RGC apoptosis. This protective effect of Apoe called for lipids, but no cholesterol was required. Their research showed that the cholesterol-fed Apoe gene knockout mice had retinal dysfunction (decreased ERG), while little or no calretinin immunoreactivity (neuronspecific calcium binding protein) was found in the layers of INL and GCL in the cholesterol-fed Apoe −/− mice, which indicated that the neural retinal cells of the cholesterol-fed Apoe −/− mice were undergoing cell death. It was believed to be related to the upregulation of the Bax immune response 22 . In addition, compared to the HFD Apoe −/− mice fed cholesterol for 25 weeks, the retinal abnormalities in the cholesterol-fed Apoe −/− mice for 35 weeks were more obvious, suggesting that the retinal degeneration in the Apoe −/− mice is supposed to be associated with the cholesterol feeding duration 22 . Furthermore, different macronutrient compositions of high-fat diet should be considered as key factors which may lead to the retina impairment in varying degrees. Therefore, it was reasonable to assume that the retinal dysfunction was due to the combination of several pathological conditions (e.g., hypercholesterolemia, atherosclerosis, hypertension and Apoe deficiency). Conversely, the opposite view also exists. Gregory et al. suggested that the increased loss of ganglion cells in the retina was not affected by the Apoe gene 25 .
It was believed that HFD and Apoe deficiency-induced retinopathy is a type of preproliferative diabetic retinopathy, in which the death of perithelial cells and endothelial cells leads to retinal ischemia 36 . However, the reasons for apoptosis in ganglion cells remain poorly understood. We proposed the following question: As a continuation of the CNS, the retina is rich in blood vessels and nerves. Does an HFD result in increased inflammation response, while Apoe gene deletion leads to the increased sensitivity of retinal nerve cells to ischemic injury? To further explore the causes of the thinning of the NFL + GCL + IPL composite layer and increased vascular area, nonperfused area and hemangiomatous area in HFD and Apoe −/− mice, we conducted the following steps.
First, we detected the number of apoptotic RGCs in the peripheral and central retina. RGCs were stained with Hoechst 33258, and the neurons containing fragmented or reduced nuclei were regarded as apoptotic cells. It was demonstrated that the percentage of apoptotic RGCs in hypercholesterolemic mice induced by HFD and Apoe −/− out of the total number of neurons was increased in the peripheral retina compared to the central retina. Second, as we know, NF-κB is an important nuclear transcription factor. It participates in the body's inflammatory reaction and immune response and can regulate apoptosis and the stress response 37 , so we examined NF-κB expression in the NFL + GCL + IPL composite layer and found that it was significantly higher in the HFD and Apoe −/− mice than in the normal control group. Puig et al. showed that HFD can increase TNF-α and thus activate microglia and macrophages in the mouse brain 38 . To confirm the role of TNF-α in these conditions, we investigated the expression of the corresponding inflammatory factor, TNF-α, regulated by NF-κB, which was also increased in this particular layer. Then, we confirmed that the increased expression of TNF-α in the retina is directly related to the upregulated expression of TNF-α mRNA. The inflammatory responses initiated through NF-κB and TNF-α resulted in the apoptosis of RGCs. In the other hand, we further detected VEGFR2 levels, which caused angiogenesis and increased vascular permeability. The results showed that the expression of VEGFR2 was higher in the HFD and Apoe −/− mice than in the control group, in agreement with Sheng's findings 8 . He indicated that CNS neurons of the Apoe −/− mice were more sensitive to ischemic injury than wild-type neurons under a normal diet 8 . Liu's study 39 suggested that HIF-1α plays an essential role in systemic responses to hypoxia and targets VEGF. Therefore, it might be a choice to evaluate HIF-1α expression. It is worth noting that a single factor (such as HFD or Apoe −/− ) caused retinal injury far less than that caused by combined factors (HFD + gene abnormality). In summary, inflammatory response is considered as a contributor to chronic retinopathy. Herein, We prove the hypothesis: these changes appears to be the inflammatory responses initiated www.nature.com/scientificreports/ through NF-κB and TNF-α, resulted in apoptosis of RGCs. Also, increased VERFR2 levels caused angiogenesis and increased vascular permeability. Due to the increasing global population and improved living conditions, the number of people with HFD and hypercholesterolemia is growing remarkably, and related systemic diseases have become a major social issue of global concern. This study showed that a high-fat diet could lead to prediabetic retinopathy, related to the duration of HFD feeding. For the population with APOE abnormalities, a reasonable dietary structure can delay or alleviate the retinal structural and functional changes. However, because of genetic predispositions, for those people without dietary control, retinal dysfunctions rapidly progress and are difficult to control, which can severely impact quality of life.

Materials and methods
Animals. Male mice (C57BL/6J, wild-type and Apoe −/− ) were used in this study. All Apoe −/− mice have the same genetic background as C57BL/6J mice. All of the experimental mice were purchased from Beijing Vital River Laboratories Co., Ltd. [SCXK (JING) 2016-0006, Beijing, China] and fed in the laboratory animal room of Tianjin University of Science and Technology (temperature 22 °C, relative humidity 40-60%) with a light:dark cycle of 12:12 h. Starting from the 6th week, a HFD (36.72% fat, 13.19% proteins, 50.09% carbohydrates) and a low-fat diet (LFD, control, 7.16% fat, 15.90% proteins, 81.07% carbohydrates) were given to two groups of Apoe −/− mice 40 . Two groups of wild-type mice also received the HFD and LFD (control groups). During the experiment, body weight and food intake were monitored, and body weight was measured weekly. After 32 weeks of feeding, the tail veins of the mice were used to examine blood glucose, mouse epididymis tissues were collected, and the serum lipids of the mice were measured to evaluate obesity (Fig. 1A) 41 . All of the animal procedures in this study were approved by the Institutional Animal Care and Use Committee of Nankai University and complied with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research.

Electroretinography (ERG).
Retinal function was evaluated with ERG following a previously described procedure. Recording was conducted in a dark room after at least 24 h of adaptation. Thirty-eight-week-old mice were anesthetized by intraperitoneal injection of 5% chloral hydrate (purchased from the General Hospital of Tianjin Medical University, 0.01 mL/g body weight) and were placed on a heating pad to keep the body temperature at 37 °C. Proparacaine hydrochloride eye drops were applied for ocular surface anesthesia. The pupils were dilated with tropicamide 0.5% half an hour before recording. A drop of 1% carbomer gel was used to keep the eye moist to ensure electrical contact and to protect the cornea. Subcutaneous needle electrodes were placed at the back and the tail as reference and ground electrodes, respectively. A loop wire, as an active electrode, was gently positioned on the center of the cornea to record signals. All of the procedures were performed under dim red light. Full-field ERGs were recorded with a visual electrophysiology system (RETI-port/scan 21, Roland Consult Co., Ltd., Brandenburg, Germany). A series of stimulus intensities (0.01, 3.0 and 10.0 cd s/m 2 ) was applied for dark-adapted ERGs. Light-adapted ERGs were recorded to stimuli of 3.0-10.0 cd s/m 2 ) superimposed on the background light after light adaptation. The amplitude of the a-wave and b-wave was measured from the baseline to the first trough and from the first trough to the next peak and were analyzed using ERG View software (Roland Consult Co., Ltd., Brandenburg, Germany).

Fundus photography and fundus fluorescein angiography (FFA).
Digital color fundus photographs were obtained using a MICRON IV comprehensive system for rodent retinal imaging (Phoenix Research Labs, Pleasanton, CA, USA) after pupil dilation. Anesthesia and pupil dilation were induced as described above. For FFA, the mice were injected intraperitoneally with 10% sodium florescence dye at a dose of 0.01 mL per 5-6 g body weight, and fundus images were obtained using MICRON IV. FFA images include the central and peripheral regions. The central region is defined as three times the diameter of the optic disc, and the rest of the retina is considered the peripheral region (Fig. 3A). Retinal vessel density was calculated using Angio Tool software (National Institutes of Health National Cancer Institute, Gaithersburg, MD), including the vascular area, percentage of the vascular retinal area, total number of branch points and retinal filling area (as opposed to the nonperfused area) 10 .
Optical coherence tomography (OCT). Anesthesia and pupil dilation were induced as described above.
During image acquisition, the mouse cornea was always kept moist with the application of carbomer gel to avoid corneal opacity. Image-guided spectral-domain OCT images 2 μm in resolution were obtained using InSight software (Phoenix Research Labs, Pleasanton, CA). The upper, upper-middle, middle, middle-lower and lower areas of the retina were scanned as follows. The scanning lines were positioned 3 PD and 1.5 PD from the optic disc separately. When measuring the central area, a full-length retinal scan was performed across the middle of the disc (Fig. 3C a-e). The retina can be divided into eight regions for the measurement of retinal thickness of each layer: the central superior (CS), central nasal (CN), central inferior (CI), central temporal (CT), paracentral superior (PS), paracentral nasal (PN), paracentral inferior (PI) and paracentral temporal (PT) regions 13 . The sizes of the optic disc, central retina and peripheral retina were 0.2 mm, 0.6 mm and 1.2 mm, respectively ( Fig. 3C(f)) 24 .
Immunohistochemistry and immunofluorescence staining. Immunohistochemistry was performed as follows. Eyeballs were enucleated and fixed with formalin (with 5% acetic acid) overnight. After placement in 0.1 M PBS, the eyes were subsequently placed in a gradient of ethyl alcohol and xylene for dehydration before embedding in liquid paraffin. Tissue sections 4 μm in thickness were cut using a cryostat (Microtome Cryostat Microm HM525; Thermo Fisher Scientific, Walldorf, Germany). After deparaffinization and rehydra- www.nature.com/scientificreports/ tion, the sections were blocked with 10% goat serum for 30 min. Then, the sections were incubated with primary antibody overnight at 4 °C. After triple washing with PBS for 5 min, the slides were incubated with secondary antibody for 1 h. Then, a series of hematoxylin staining and mounting was performed. For immunofluorescence staining, tissue sections on slides were fixed with 10% goat serum and then incubated with a single primary antibody for single immunofluorescence staining for 1 h. The slides were washed three times with 1 × PBS before incubation with DAPI counterstaining for 30 min. To detect apoptosis, RGCs were stained with Hoechst 33258 (Thermo Fisher Scientific, Waltham, MA, USA) and Ultra Cruz (Santa Cruz Biotechnology, Dallas, TX, USA) mounting medium. The slides were washed three times with 1 × PBS before incubation with a respective secondary antibody for single immunofluorescence staining or two different respective secondary antibodies for double immunofluorescence staining: anti-NF-kB p65 antibody (1:100)  Quantification of retinal sterols. Mice were anesthetized via intraperitoneal injection of 80 mg/kg ketamine and 15 mg/kg xylazine and sacrificed by cervical dislocation. After eyeballs were enucleate, the retina was carefully taken out. Samples of the washed retina of the same genotype were combined and homogenized. Sterols were quantified by isotope dilution gas chromatography-mass spectrometry (GC-MS), as described(Tianjin University of Science and Technology). Unesterified cholesterol and total cholesterol were measured by GC-MS with deuterated sterol analogs as internal standards 42 . Reverse transcription, quantitative real-time polymerase chain reaction. Total RNA was isolated from the retina using TRIzol reagent (Invitrogen, CA, USA) according to the manufacturer's instructions. Quantitative real-time PCR (qPCR) was performed on an Mx3000P (Agilent Technologies) using gene-specific primers for GAPDH, TNF-α, SCAP and SREBP-1 (Table 3) and Power SYBR Green PCR Master Mix (Applied Biosystems). GAPDH was used as an internal loading control to normalize all of the PCR products. The band intensities of the amplified DNAs were compared after visualization on a UV transilluminator. Data analysis. All of the data are expressed as the mean ± standard error of the mean (SEM). Figures and statistical analysis were performed using GraphPad Prism software (GraphPad Prism 5, GraphPad Prism Software, Inc., San Diego, CA, USA). One-way ANOVA followed by Bonferroni's correction was used when multiple groups were compared. P < 0.05 was considered statistically significant.