Introduction
Diabetic retinopathy remains a leading cause of new-onset blindness in adults [
1], but its pathogenesis is not fully understood. It is known that ‘residual risks’ exist even after controlling for hyperglycaemia [
2]. Among these, plasma lipoproteins have been implicated, as supported by numerous clinical and epidemiological studies; see reviews [
3‐
5] and some key studies [
6‐
14]. Most of these studies focus on standard measures of circulating lipids and lipoproteins, and in general, they show associations between ‘dyslipidaemia’, for example elevated plasma LDL-cholesterol or apolipoprotein B (apoB), and the severity of diabetic retinopathy. Some studies have gone further, addressing qualitative (and not just quantitative) changes in plasma lipoproteins, including the extent of LDL modification by oxidation [
15]. We have previously reported associations between diabetic retinopathy severity and average lipoprotein particle diameters and the distribution of ‘subclasses’ of HDL, LDL, and VLDL in the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications (DCCT/EDIC) cohort [
10]. Also in that cohort, we found that circulating immune complexes containing oxidised LDL (ox-LDL) predicted severe diabetic retinopathy many years later [
13]. Overall, however, associations between plasma lipoproteins and diabetic retinopathy, although statistically significant in population studies, are too weak to define individual risk or prognosis, as recently underlined in two detailed reports [
14,
15].
We have hypothesised that despite these rather unimpressive associations between plasma lipoproteins and diabetic retinopathy, lipoproteins actually play a central role in diabetic retinopathy progression, but one that is ‘hidden’, occurring in tissue, and only operative after extravasation and modification. In the retina, in contrast to arteries during atherogenesis, leakage from plasma to tissue only occurs if the blood-retinal barriers (BRBs) are compromised. Put simply, plasma lipoprotein levels may be largely irrelevant to the retina if BRBs are intact, and their effects following BRB breakdown may depend more on the extent of leakage than on their plasma concentrations. In support of this concept, we have observed that intra-retinal deposits of apoB-100 (the main protein component of LDL and VLDL), ox-LDL and ox-LDL immune complexes are entirely absent in healthy human retinas, but present in diabetic retinas even before the onset of clinical retinopathy, and thereafter in greater amounts commensurate with diabetic retinopathy severity [
16,
17]. Furthermore, we have consistently observed toxicity of LDL (modified in vitro to simulate stresses found in vivo), towards cultured human retinal cells, including capillary endothelial cells, pericytes, retinal pigment epithelial (RPE) cells, and Müller cells, promoting oxidative and endoplasmic reticulum (ER) stresses, inflammation and death [
17‐
24]; ox-LDL is also injurious to neurons [
25]. Based on these considerations, we hypothesise that after extravasation into the retinal tissue, lipoproteins become modified by glycation and oxidation and, thus, become cytotoxic. In this way, episodes of initially transient, localised vascular leakage may lead to vicious cycles of chronic pathology.
Rodent models are commonly used in diabetic retinopathy research, but they have limitations. Rodents do not develop the advanced, proliferative retinopathy that afflicts humans [
26]. Oxygen-induced retinopathy models in new-born rodents exhibit neovascularisation, but the animals are not a model of diabetes. Regarding a possible role of extravasated lipoproteins in the propagation of diabetic retinopathy, a critical difference between rodents and humans lies in their lipoprotein metabolism: rodents have substantially lower LDL levels than humans and cholesterol is transported primarily in HDL [
27]. Therefore, rodent models of diabetes fail to replicate the chronic exposure to extravasated, modified LDL that we have demonstrated in human diabetic retinopathy. In the present study, using intravitreal injection of human LDL (with or without modification), we assessed the effects on retinas in a mouse model of diabetes and in non-diabetic mice.
Discussion
We evaluated, for the first time, the effects of normal and modified LDL given by intravitreal injection on the diabetic retina. We showed that in the presence, but not in the absence, of short duration diabetes (i.e., 12 weeks), mouse retinas became highly sensitive to the injurious insult of HOG-LDL, but that normal LDL had little effect. Specifically, HOG-LDL altered the retinal architecture and impaired retinal function as measured by ERG, vascular leakage, VEGF expression, inflammation, ER stress, and propensity to apoptosis; effects that replicate some features of human clinical diabetic retinopathy [
36]. Thus, these data provide in vivo proof-of-concept that lipoproteins, when extravasated and modified, have important and dire consequences in the propagation of diabetic retinopathy, belying the relatively weak associations between circulating levels and diabetic retinopathy.
Rodents have low endogenous LDL levels, and we considered chronic intravascular infusion of either human or mouse LDL (with or without modification) to be unsatisfactory for both practical and immunologic reasons [
37]. Our goal was to simulate the accumulation of intra-retinal modified LDL and to understand its contribution to the pathogenesis of diabetic retinopathy. We chose the intravitreal route to access the retina in a manner analogous to therapeutic approaches in humans. It should be mentioned that, to our knowledge, the endogenous vitreous levels of N-LDL or ox-LDL in diabetic humans have not been documented in the literature, either in the presence or absence of retinopathy. It has been reported that non-diabetic human and bovine aqueous humour contains only HDL, but not LDL [
38]. However, in view of our findings of progressive extravasation and modification of LDL in the diabetic retina [
16,
17], both N-LDL and modified LDL may exist in the vitreous of diabetic patients.
An important question concerns possible immune responses to human LDL when introduced to the mouse. We consider such effects do not explain the present findings, first because retinal injury was induced by HOG-LDL but not by N-LDL, and second because HOG-LDL only induced retinopathy in diabetic but not in non-diabetic animals. Ocular immune privilege [
39] might have a role. However, it should be noted that modified LDL, with its altered immunogenicity, does trigger immune responses in humans, resulting in immune complexes which are implicated in vascular disorders. In the DCCT/EDIC cohort, we found that higher plasma levels of ox-LDL immune complexes predicted risk of diabetic retinopathy in type 1 diabetes [
13]. In patients with type 2 diabetes, higher circulating levels of autoantibodies against malondialdehyde-modified apoB-100 were associated with diabetic retinopathy [
11]. It is estimated that 95% of ox-LDL in circulation may be immune-complexed [
40]. Recently, we reported the presence of ox-LDL immune complexes in human diabetic retinas, and in cell culture; these exhibited greater toxicity than non-complexed ox-LDL towards pericytes [
17]. Thus, immune responses to ox-LDL are involved in diabetic retinopathy, but are unlikely to explain the present findings. In the present study, HOG-LDL induced inflammatory cell infiltration in both diabetic and non-diabetic animals, but the response was much more dramatic in the diabetic animals, and, therefore, the effect cannot be attributed to immunologic response to a lipoprotein from a different species.
One intriguing finding from this study is that diabetes appears to ‘prime’ the retina to the injurious effects of HOG-LDL. The dramatic response to HOG-LDL suggests that subclinical changes had already occurred, even though the retina appeared to be normal. The existence of subclinical injury in diabetes agrees with clinical observations; retinas remain apparently normal for over 5 years [
41], after which the development and rate of progression of diabetic retinopathy vary widely among individuals [
42]. A seminal study by Engerman and Kern [
43] showed that in diabetic dogs, poor glycaemic control caused subclinical damage, the progression of which could not be prevented by subsequent good glycaemic control. One plausible explanation for our data is that diabetes (characterised by hyperglycaemia) upregulates CD36 in retinal cells [
44], which mediates the toxicity of modified LDL [
17]: this is consistent with the observed synergistic effects of glucose and lipids in macrophages via the CD36 receptor in atherosclerotic lesions [
45]. Thus the data support a potentially important role of extravasated lipoproteins, in addition to hyperglycaemia, in the propagation of diabetic retinopathy.
Loss of BRB integrity is a critical feature of early diabetic retinopathy [
36]. In this study, intravitreal administration of HOG-LDL elicited extensive BRB breakdown in diabetic, but not in non-diabetic mice. Retinal leakage of fluorescein markers appeared scattered and patchy, as is seen in clinical diabetic retinopathy. There was also evidence of injury to the RPE layer, suggesting that both inner and outer BRBs were compromised. These effects were seen even though the LDL-related stress ‘approached’ the retinal vessels from the outside, not from the plasma. They were accompanied by progressively elevated VEGF in the retina over the two week timeframe studied, which likely contributed to the higher vascular permeability. Considering the short duration of this study, the present model simulates only the pre-proliferative changes of diabetic retinopathy. It would be of interest to see whether this increase in VEGF is sustained in the longer term, after repetitive or continuing exposure to HOG-LDL, and if this might induce neovascularisation (as is seen in proliferative diabetic retinopathy of humans) in the rodent model. Taking into account these in vivo data and our earlier observations in human retinal vascular cells [
17‐
24], there is now strong evidence that lipoprotein leakage and modification may cause further loss of BRB integrity, and thus constitute an important positive feedback loop propagating retinal injury.
We also studied some additional biochemical markers in mouse retina that reflect known features of diabetic retinopathy, including glial activation, ER stress and pro-apoptosis [
46]. HOG-LDL enhanced these processes in diabetic vs non-diabetic mice. Interestingly, retinal ER stress and BAX increased progressively during the two weeks of observation. Thus, the data are in agreement with our earlier findings in cultured human retinal cells exposed to HOG-LDL [
17,
20,
22‐
24], and with studies in the retinas of diabetic patients [
16,
23,
24].
Potential limitations include: (1) intravitreal injection may not accurately mimic the continuous LDL extravasation around retinal vessels; (2) only a single LDL dose was studied, and future studies should define an optimal, pathophysiologically relevant dose range; (3) the variables, including duration of diabetes before LDL challenge and that after injection, require optimisation; and (4) the mechanisms underlying the potentiating effects of HOG-LDL when combined with diabetes require investigation.
In conclusion, our study shows that modified LDL, if present in the retina at early stages of diabetes, significantly accelerates the progression of retinal injury and recapitulates many features of human diabetic retinopathy. It supports the hypothesis that, in addition to hyperglycaemia, and despite rather weak associations between plasma lipoproteins and human diabetic retinopathy, LDL (and presumably other lipoproteins) plays an important but ‘hidden’ role in diabetic retinopathy. Improved means to detect early vascular leakage and ‘glycoxidative/lipoxidative’ damage at the tissue and molecular levels may have value in diabetic retinopathy management. The work also provides feasibility to investigate the relative contribution of modified lipids and hyperglycaemia to diabetic retinopathy using an animal model that employs a simple intravitreal injection approach.