Retinopathy with central oedema in an INS C94Y transgenic pig model of long-term diabetes
Diabetic retinopathy is a severe complication of diabetes mellitus that often leads to blindness. Because the pathophysiology of diabetic retinopathy is not fully understood and novel therapeutic interventions require testing, there is a need for reliable animal models that mimic all the complications of diabetic retinopathy. Pig eyes share important anatomical and physiological similarities with human eyes. Previous studies have demonstrated that INS C94Y transgenic pigs develop a stable diabetic phenotype and ocular alterations such as cataracts. The aim of this study was to conduct an in-depth analysis of pathological changes in retinas from INS C94Y pigs exposed to hyperglycaemia for more than 2 years, representing a chronic diabetic condition.
Eyes from six INS C94Ypigs and six age-matched control littermates were analysed via histology and immunohistochemistry. For histological analyses of retinal (layer) thickness, sections were stained with H&E or Mallory’s trichrome. For comparison of protein expression patterns and vessel courses, sections were stained with different antibodies in immunohistochemistry. Observed lesions were compared with reported pathologies in human diabetic retinopathy.
INS C94Ypigs developed several signs of diabetic retinopathy similar to those seen in humans, such as intraretinal microvascular abnormalities, symptoms of proliferative diabetic retinopathy and central retinal oedema in a region that is cone rich, like the human macula.
The INS C94Ypig is an interesting model for studying the pathophysiology of diabetic retinopathy and for testing novel therapeutic strategies.
KeywordsAnimal model Cataract Colour vision Diabetes Eye disease Macular oedema Pathophysiologic processes Retinopathy Vascular complications
Dolichos biflorus agglutinin
Lycopersicon esculentum lectin
Neurofilament heavy chain
Smooth muscle actin
Individuals with diabetes mellitus are showing an increasing incidence of diabetic retinopathy, a severe complication that often results in blindness . A recent population-based study reported that 13% of individuals newly diagnosed (screening-detected) with type 2 diabetes already showed signs of diabetic retinopathy . Visual impairment is primarily a sequelae to macular oedema, an exudative fluid accumulation in the macula that affects one in 15 individuals with diabetes . Currently, more than 20 million individuals worldwide suffer from this form of sight-threatening diabetic retinopathy and numbers are rising exponentially with the incidence of diabetes . To date, the pathophysiology of diabetic retinopathy is not fully understood, and the translational value of established animal models (e.g. for impairment of cone function in the macula) is limited by differences in retinal architecture and cellular composition in the central area [4, 5]. Nevertheless, rodent models have great value for studying pathophysiological matters such as vasogenic oedema formation in streptozotocin-induced diabetic rats, or the direct effects of drugs on retinal vascular function and thickness [6, 7].
Human vision depends heavily on the macula, an area with increased density of cones, cone bipolar cells and ganglion cells, providing high optical resolution . Many mammals possess a specialised retinal region comparable with the macula, with an increased density of cones, ganglion cells and bipolar cells, either in the form of a roughly circular area centralis or as a horizontal visual streak . In contrast to rodents, the cone density observed in the visual streak of the pig retina (20,000–40,000 cones/mm2)  is quite similar to that in the outer regions (non-foveal region) of the human macula (20,000–40,000 cones/mm2) .
In general, pig and human eyes are similar in their size, anatomy and physiological characteristics. As animals with high life expectancy, the use of pigs enables the study of the long-term pathophysiological consequences of chronic metabolic disturbances . The INS C94Y transgenic pig is a novel large-animal model of diabetes mellitus that is characterised by impaired insulin secretion, decreased beta cell mass and permanent hyperglycaemia . Cataracts develop within the first week of life in these INS C94Ypigs, indicating an ocular pathophysiological phenotype . Since this model could potentially be useful for testing novel therapeutic strategies counteracting diabetic retinopathy and offers the opportunity to investigate the effects of long-term hyperglycaemia on visual function and the development of retinal oedema, we aimed to characterise the ocular phenotype of long-term diabetic INS C94Ypigs, particularly in the retina. Observed changes were related to disease characteristics in humans in order to prove the translational quality of this novel model for diabetic retinopathy pathophysiology.
Animal model, sample preparation for histology and immunohistochemistry
The generation of and main pathophysiological findings in INS C94Ypigs have been described previously [11, 13]. For this study, we used six INS C94Y male and female pigs and six age-matched control littermates (ten pigs aged 24 months and two aged 40 months) generated at the Molecular Animal Breeding and Biotechnology, Gene Centre, LMU Munich, in their facility in Oberschleissheim, Germany. All animal experiments were performed according to the German Animal Welfare Act and were permitted by local authorities (ROB 55.2-1-54-2532-68-11). Animals were housed under controlled conditions, with a once-a-day feeding regimen and free access to water. Continuous glucose monitoring was performed using the Guardian REAL-Time System (Medtronic, Meerbusch, Germany) with a glucose sensor inserted subcutaneously behind the ear and transfer of collected data via a connected transmitter to an external computer. Blood glucose levels were determined once or twice daily using a glucometer. In addition, fructosamine levels were monitored as an adequate evaluation of medium-term glucose control in pigs [14, 15]. Fructosamine levels were determined in EDTA-plasma using an AU480 Autoanalyzer and an adapted reagent kit from Beckman-Coulter (Krefeld, Germany). At the time point of sampling of the eyes, mean fructosamine levels were 380.17 ± 41.50 μmol/l in control pigs and 684.83 ± 61.96 μmol/l in INS C94Ypigs (Student’s t test, p < 0.0001), confirming a chronic hyperglycaemic state in the INS C94Ypigs. Animals were euthanised after 24 or 40 months, respectively. Immediately after enucleation, eyes were dissected to obtain sections containing low and high ganglion cell density areas  of the porcine retina, as recently described . Briefly, a sagittal section was cut through the optic nerve, dividing the eye cup into temporal and nasal halves to obtain cross-sections representing central and peripheral retinas, characterised by 4000–5000 (central) and 500–1000 ganglion cells per mm2 (periphery), respectively . In pigs, this area of high cone density, called the visual streak, is a broad horizontal band situated around and above the optic nerve. The highest density, with 40,500 cones per mm2, can be found in a small area of the visual streak above the optic nerve in the nasal direction . This area has cone densities comparable with those in the human macula and is preferentially damaged in diabetic retinopathy [3, 10]. The temporal part was fixed in formalin, and the nasal part in Bouin’s solution. Dehydration, embedding, cutting and mounting on coated slides were performed as previously described .
Histology and immunohistochemistry, measurement of layer thickness and counting of blood vessel section profiles
Specimens from all six INS C94Ypigs and all six control pigs were used for all stainings. For histological analyses, sections were deparaffinised, rehydrated and stained with H&E or Mallory’s trichrome . For immunohistochemistry, heat antigen retrieval was performed . The following primary antibodies were used: monoclonal mouse anti-glutamine synthetase (GS; 1:1500; BD Biosciences, Heidelberg, Germany); mouse anti-smooth muscle actin (SMA; 1:50; Sigma-Aldrich, Taufkirchen, Germany); FITC-labelled lectins Dolichos biflorus agglutinin (DBA) and Lycopersicon esculentum lectin (LEL) (all 3 μg/ml; Vector, Burlingame, CA, USA); polyclonal rabbit anti-phosphorylated (p)IRS1 (Ser302-307; 1:500; Merck Millipore, Darmstadt, Germany); rabbit anti-aquaporin 4 (AQP4; 1:200; Alomone, Jerusalem, Israel); polyclonal rabbit anti-pIRS1 (Tyr895; 1:100; Merck Millipore); monoclonal rat anti-S-opsin (neat) , polyclonal rabbit anti-cone arrestin, (1:10,000)  and polyclonal rabbit anti-neurofilament heavy chain (NEFH; 1:500; Sigma-Aldrich). Fluorescence-labelled (488 or 568) secondary antibodies to mouse and rabbit IgG heavy and light chain (1:500; Invitrogen, Karlsruhe, Germany) and to rat IgG2c (1:200, Invitrogen, Karlsruhe, Germany) were used . Images were recorded using a Leica DMR (Leica, Wetzlar, Germany) or Axio Vision Imager M1 (Zeiss, Jena, Germany). All stainings included appropriate isotype controls. For quantification of retinal layer thickness, complete sections were scanned with a C9600-12 scanner and analysed using NDP.view version 2.4.26 (Hamamatsu, Herrsching am Ammersee, Germany). Retinal layer thickness was measured at different defined distances from the optic nerve. All measurements were performed independently by two researchers in separate experiments. The number of vessel section profiles was quantified over a length of 1 mm by counting LEL-positive areas.
Retinal layer thickness measurements (in μm) are presented as means ± SD. The Mann–Whitney U test was used to compare retinal layer thicknesses (complete and single layers). Student’s t test was used to compare vessel numbers and fructosamine levels between control and diabetic INS C94Ypigs. The significance threshold was p < 0.05 in all analyses.
Retinal layer thickness is substantially increased in INS C94Ypigs, especially in the nerve fibre/ganglion cell layer
INS C94Ypigs have intraretinal vascular abnormalities that are typical of diabetic retinopathy
Several intraretinal vessels of INS C94Ypigs were irregular and coiled (Fig. 1a–k: control pigs; Fig. 1b–l: diabetic INS C94Y pigs), with changes in endothelial cells and the inner membrane (Fig. 1g–j). Loop formation of capillaries in the retinas of INS C94Ypigs (Fig. 1k: control pig; Fig. 1l: diabetic INS C94Y pig) was detectable. In addition, extraretinal haemorrhage was visible in the gross pathology of the fundus of diabetic INS C94Ypigs, with erythrocytes in the vitreous (Fig. 1l, arrow). Disruption of the inner limiting membrane (Fig. 1l, arrow) and the nerve fibre/ganglion cell layer in INS C94Ypigs resulted in cotton-wool spots.
Findings in the eyes of diabetic INS C94Ypigs indicate changes in the intraretinal vessel course and the function of colour vision
Altered abundance and localisation of cell metabolism and water-clearance proteins in INS C94Ypig retinas
An increase in NEFH abundance indicated an enlargement of neurons/axons in the nerve fibre/ganglion cell layer (Fig. 3g, h), and not of the Müller glial endfeet (Fig. 3k, l). The main water channel of the retina, AQP4 (Fig. 3m, n), changed from a Müller glial endfeet pattern in control retinas (Fig. 3m) to an entire Müller glial cell accented expression pattern in diabetic INS C94Ypigs, with an accentuation of the endfeet at the inner limiting membrane (Fig. 3n, p).
Long-term diabetes in INS C94Ypigs is associated with signs of severe diabetic retinopathy
There is an urgent need for adequate animal models that mimic the important features of diabetic retinopathy. The importance of large-animal models is evolving, given their remarkable resemblance to the anatomical, physiological and pathological characteristics of many diseases in humans, such as diabetic retinopathy. The major sight-threatening complication of this disease is macular oedema . Animal models for investigating certain features of this condition exist [5, 6, 7, 22] but, for greater translational value to humans, more suitable animal models, featuring as many hallmark pathologies observed in the human disease condition as possible, are urgently needed. In this study, we characterised the ocular phenotype of an INS C94Ypig model of long-term diabetes mellitus (≥24 months). All INS C94Ypigs showed several typical ocular lesions of diabetes mellitus (see text box), such as cataracts and severe retinopathy. Cataracts developed in the very early stages of diabetes mellitus in these pigs  and progressed to mature cataracts in all cases analysed.
Interestingly, the retinas of INS C94Ypigs revealed significant changes in retinal architecture and the levels of functionally relevant proteins. There was a very prominent increase in retinal thickness in diabetic INS C94Ypigs (Fig. 1), which was particularly confined to the regions 2 and 3 mm from the optic nerve, where pigs have a region of similar cellular composition to the macula in humans . Closer examination of the affected retinal layers revealed that INS C94Ypigs developed retinal oedema with a spatial distribution similar to that seen in humans with macular oedema . An increase in the mean central thickness of 1 mm in individuals with diabetic retinopathy defines a centre involving diabetic macular oedema, and this central thickness is correlated with the decrease in visual acuity .
In addition, typical vascular abnormalities (i.e. intraretinal microvascular abnormalities) of human diabetic retinopathy  were seen in the retinas of diabetic INS C94Ypigs. Retinal vessel calibre changes are associated with complications in humans with diabetes , and there is a consistent relationship between wider retinal venular calibres, lower fractal dimensions and the progression to proliferative retinopathy .
Furthermore, INS C94Ypigs developed typical irregularities at the inner limiting membrane, resulting in disruption of this barrier with the accumulation of hard exudates and cotton-wool spots, which are characteristic features of proliferative diabetic retinopathy. This is the most common microvascular end-stage complication, with a 16-year incidence of 31% in type 1 diabetes  and 15% in type 2 diabetes (over 20 years) . The duration of diabetes is the strongest predictor for the development and progression of retinopathy, and models with longevity are therefore favourable . A further advantage of the INS C94Ypig model is that all major complications were detectable after 2 years, which is very rapid compared with other large-animal models such as large primates, where the development of these signs has required more than 15 years .
Interestingly, INS C94Ypigs revealed an altered abundance of proteins important for colour vision. Impaired colour vision is a common observation in individuals with diabetic retinopathy, starting in early, uncomplicated stages and increasing with macular oedema, in which two-thirds of individuals have colour discrimination abnormalities . Some unique features of the INS C94Ypig model, such as alterations in cones (Fig. 2k–p; text box), are replicated in other existing animal models, such as marmosets  and rats [6, 7].
Disruption of the inner limiting membrane in INS C94Ypigs points to changes in Müller glial cell function. This was further substantiated by changes in proteins with important functions for Müller glial cell metabolism (e.g. GS) and the major retinal water channel protein AQP4 . The development of retinal oedema is associated with Müller glial cell dysfunction in osmotic glial swelling , and the finding of similar alterations in diabetic INS C94Ypigs therefore points to a similar pathophysiology. Müller cells are particularly important for retinal protective responses under stressful conditions . AQP4 has been described in Müller cell (endfoot) membranes of pigs  and is involved in retinal volume regulation . The observed alterations in GS and AQP4 levels (Fig. 3) in diabetic INS C94Ypigs merit future investigation to clarify the potential role of Müller glial cells in oedema formation. Because of the similar function of pig and human Müller cells (e.g. both possess Ca2+-activated K+ channels), which are not ubiquitously present in all species , the pig is a good model in which to study these alterations.
This study represents a unique possibility to analyse the ocular consequences of long-term exposure (>2 years) to hyperglycaemia. The pig model replicates several hallmarks of diabetic retinopathy observed in humans (text box) that, to our knowledge, have not previously been found combined in one animal model. Marmosets (Callithrix jacchus) maintained on a 30% galactose-rich diet for 2 years have been reported to develop characteristic retinal vascular lesions with macular oedema and morphological characteristics, including acellular capillaries and pericyte loss, vessel tortuosity and capillary basement membrane thickness . The findings from that study indicated that hyperhexosaemia triggered retinal vascular changes similar to those in human diabetic retinopathy; however, changes in cones or Müller glial cells were not described . In the INS C94Ypig model, longevity and physiological (e.g. Müller glial cell function) and anatomical similarities (a central, cone-rich area) with the human retina enabled specialised studies of these aspects.
Clearly, one limitation of this study was the monitoring of only one time point, presumably representing late-stage retinopathy, because we sampled at the endpoint of a long-term study designed to generate the first INS C94Ypig biobank . Therefore, we have no information on how rapidly pathophysiological features develop in the pig, and comparison with the disease dynamics of other models is limited. In future studies, investigation of the dynamics of diabetic retinopathy features should include an in vivo imaging technique such as optical coherence tomography, which is already standard when monitoring rodent models for diabetic retinopathy [6, 7]. However, this has yet to be established in the pig and the very early, severe cataract formation involving the entire lens with opacity at 4.5 months requires surgical removal of the lens prior to in vivo imaging . If the changes develop after only months then this would be a drawback for the model, because it would make it impractical for several applications. The next step is to analyse earlier time points and, as cataract development starts as early as day 8 after birth , we hypothesise that other pathologies will also appear at an early point.
In summary, we confirmed our hypothesis by showing that the retinal changes occurring in the INS C94Ypig make it an interesting and valuable model for further studies of diabetic retinopathy and its pathophysiology, and of proliferative diabetic retinopathy, intraretinal microvascular abnormalities or macular oedema risk stratification, and novel treatment strategies.
The authors thank S. Nüske and A. Scholz (Teaching and Experimental Farm, LMU Munich) for providing pig control tissues. We thank R. Degroote (Institute of Animal Physiology, LMU Munich) for critically revising the manuscript and R. Wanke (Institute of Animal Pathology, LMU Munich) for critical discussions.
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.
This work was supported by the von Behring-Röntgen Foundation (63-0004); the Foundation for the Support of Science in Ophthalmology Marburg e.V. (to CAD); the Federal Ministry of Education and Research (16EX1024A); and the German Centre for Diabetes Research (DZD) (to EW).
Duality of interest
The authors declare that there is no duality of interest associated with this manuscript.
CAD conceived and designed the experiments. KJHK, BA, SH, SMH, PBU, KL and CAD performed the experiments. KJHK, BA, SH, SMH, PBU, KL, WS, EW and CAD analysed the data. SR, AB and EW developed and characterised the INS C94Ypig model. CAD wrote the manuscript. All authors critically read and provided comments on the manuscript. All authors approved the final version to be published. CAD is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.