www.elsevier.com/locate/visres

Vision Research 47 (2007) 612–623

Microglial changes occur without neural cell deathin diabetic retinopathy

David Gaucher a,b,c, Jean-Armand Chiappore a,b, Michel Paques a,b,d, Manuel Simonutti a,b,Christian Boitard e, Jose A. Sahel a,b,d,f, Pascale Massin c, Serge Picaud a,b,d,*

a INSERM U-592, Hopital St. Antoine, Laboratoire de Physiopathologie Cellulaire et Moleculaire de la Retine, Batiment Kourilsky, 184 rue du

Faubourg Saint-Antoine, 755751 Paris Cedex 12, Franceb Universite Pierre et Marie Curie Paris-6, Paris, France

c Assistance Publique-Hopitaux de Paris, Hopital Lariboisiere, Service d’Ophtalmologie, Universite Paris 7, Paris, Franced Fondation Ophtalmologique Adolphe de Rothschild, Paris, France

e INSERM U-561, Hopital St. Vincent de Paul, Paris, Francef Centre Hospitalier National d’Ophtalmologie des Quinze-Vingts, Paris, France

Received 7 February 2006; received in revised form 29 July 2006

Abstract

Very early neuroglial changes have been observed to precede major vascular changes in the retina of diabetic patients and animalmodels. We investigated the sequence of these neuroglial changes further, in mice with alloxan-induced diabetes. Diabetes was inducedby a single injection of Alloxan into C57/Bl6 mice, which subsequently received daily insulin injections. Diabetic and control animalswere weighed and their blood glucose levels were determined weekly. Electroretinographic recordings and scanner laser ophthalmoscope(SLO) examinations were carried out 15 days, one month and three months after the onset of diabetes. Diabetes induction was confirmedby the presence of glucose in the urine, a tripling of blood glucose level, weight loss and an increase in glycated haemoglobin levels. Threemonths after diabetes onset, the electroretinogram b/a wave amplitude ratio was decreased at the highest light intensities and oscillatorypotentials were delayed. The retinal fundus and vessels remained unchanged. No cell apoptosis was detected in vertical and horizontalsections of the retina by TUNEL or immunocytochemistry for the active caspase 3. No increase in GFAP-immunostaining indicative of aglial reaction was observed in Muller glial cells. By contrast, changes in the morphology of microglial cells were observed, with markedshortening of the dendrites. Thus, the microglial reaction occurs very early in progression to diabetic retinopathy, at about the same timeas early electroretinographic modifications. The absence of apoptotic cells, contrasting with previous results in mice with streptozotocin-induced diabetes, is consistent with insulin neuroprotection.� 2006 Elsevier Ltd. All rights reserved.

Keywords: Experimental diabetes; Mouse; Retina; Electroretinography; Microglia; Apoptosis; Insulin

1. Introduction

Diabetic retinopathy (DR) is a common complication ofdiabetes that can lead to legally recognised blindness. Thiscondition has been extensively studied but its pathophysi-ology remains poorly understood. DR is classically consid-ered to be a microangiopathy because microaneurysms and

0042-6989/$ - see front matter � 2006 Elsevier Ltd. All rights reserved.doi:10.1016/j.visres.2006.11.017

* Corresponding author. Fax: +33 1 49 28 46 05.E-mail address: picaud@st-antoine.inserm.fr (S. Picaud).

the proliferation of new vessels are observed, together witha breakdown of the blood retinal barrier (BRB). The BRBis located at the junctions between endothelial cells in thecapillaries. Its normal function is regulating the movementof molecules into and out of the retina. BRB breakdown isroutinely identified in patients with DR in clinical practice,by fluorescein angiography, in which the abnormal leakageof fluorescein from the capillaries is observed (Davis,1992). These vascular changes may result from biochemicaldisorders, such as activation of the polyol and b-kinasepathways, and the production of glycated products and

D. Gaucher et al. / Vision Research 47 (2007) 612–623 613

reactive oxidative species due to an increase in serum glu-cose concentration (Ciulla, Amador, & Zinman, 2003).Retinal cell degeneration and blindness may result fromthese vascular changes and subsequent retinal ischaemia.

However, several recent studies have suggested that neu-rodegenerative events may precede the vascular changes.Indeed, changes in retinal function have been reported inhumans and animals before the detection of retinal vascu-lar lesion in the form of BRB rupture or microaneurysms.In diabetic patients, colour vision and contrast sensitivitymay deteriorate without detectable clinical symptoms ofDR (Ewing, Deary, Strachan, & Frier, 1998). Electroretin-ograms (ERG) show very early changes in oscillatorypotentials (OP) and changes are also observed on multifo-cal ERG (Bresnick & Palta, 1987; Fortune, Schneck, &Adams, 1999; Tzekov & Arden, 1999; Yoshida, Kojima,Ogasawara, & Ishiko, 1991). Histologically, neuronalapoptosis has been observed in ganglion cells before anychange could be detected in the retinal vascular system(Barber et al., 1998). This degeneration is associated withglial reactions in Muller cells, generally observed followingthe appearance of retinal lesions (Abu-El-Asrar, Dralands,Missotten, Al-Jadaan, & Geboes, 2004).

Proliferative diabetic retinopathy has rarely beendescribed in diabetic animal models, although neovascular-isation has been described in a few transgenic animals (Laiet al., 2005; Ruberte et al., 2004; Yamada, Yamada, Higu-chi, & Matsumura, 2005). Changes of various magnitudeshave been reported on ERG: increases in OP latencies,and decreases in OP amplitudes, a- and b-waves (Ewinget al., 1998; Hancock & Kraft, 2004; Li, Zemel, Miller, &Perlman, 2002; Lowitt, Malone, Salem, Kozak, & Orfalian,1993; Sakai, Tani, Shirasawa, Shirao, & Kawasaki, 1995).Neuronal apoptosis has been described in the ganglion celllayer in various diabetic animal models (Agardh, Bruun, &Agardh, 2001; Barber et al., 1998; Martin, Roon, Van Ells,Ganapathy, & Smith, 2004; Ning et al., 2004). Neuronaldegeneration has also been detected in other retinal layersin some models (Agardh et al., 2001; Ning et al., 2004).This degenerative process has been attributed to the glialreaction of Muller cells, as in diabetic humans (Agardhet al., 2001; Asnaghi, Gerhardinger, Hoehn, Adeboje, &Lorenzi, 2003; Li et al., 2002; Lieth et al., 1998; Rungger-Brandle, Dosso, & Leuenberger, 2000), or to microglial cellactivation in some cases (Barber et al., 2005; Rungger-Brandle et al., 2000; Zeng, Ng, & Ling, 2000). However,these neuroglial reactions are not observed in all animalmodels of DR (Agardh et al., 2001; Asnaghi et al., 2003;Barber et al., 2005, 1998; Li et al., 2002; Lieth et al.,1998; Martin et al., 2004; Ning et al., 2004; Rungger-Bran-dle et al., 2000). In particular, neuronal apoptosis was firstdescribed in rats with streptozotocin-induced diabetes(Barber et al., 1998), but is not systematically observed inmice with streptozotocin-induced diabetes (Asnaghi et al.,2003; Martin et al., 2004).

In this study, we assessed BRB permeability and retinalcell activity in vivo, in mice with alloxan-induced diabetes,

a classical animal model of diabetes in which DR has notyet been investigated. Alloxan and STZ are both diabeto-genic agents with cytotoxic effects on pancreatic b-cellsresulting from the production of reactive oxygen species.Alloxan generates oxidative stress conditions, with asimultaneous massive increase in cytosolic calciumconcentration, whereas STZ induces DNA alkylation(Szkudelski, 2001). At the first sign of functional changesin Alloxan-induced mice, we looked for neuroglial reac-tions on vertical and horizontal retinal sections. Wereport here that changes in microglial cell morphologyare the first detectable cellular modifications occurringin mice with alloxan-induced diabetes, apparently preced-ing neuronal apoptosis and increases in blood barrierpermeability in this model.

2. Materials and methods

2.1. Animals and experimental diabetes induction

We used twenty C57/Bl6 male mice, which were handled according to theprinciples of the ARVO statement for the use of animals in ophthalmic andvision research. The mice were maintained on a standard diet, with watersupplied ad libitum, under a 12-h light/12-h dark cycle. Diabetes was inducedby a single injection of Alloxan (75 mg/kg) into the tail vein, as described byLarger, Becourt, Bach, and Boitard (1995). Animals were 10 weeks old andweighed around 25 g at the time of injection. Nine mice received injections,the others being used as controls. Diabetic mice received a daily subcutane-ous injection of three international units of insulin (Umuline Zinc�, LillyLaboratories, France), in the morning (9 a.m.). Insulin was administeredto keep the animals alive during the study period, because a few mice withalloxan-induced diabetes had previously been reared without treatment,but they had survived for no more than two weeks.

2.2. Glucose levels

Animals were weighed once per week. We checked for the presence ofglucose in the urine twice during the first week after injection (urine glu-cose test strips, Roche Diagnostics, France). Blood glucose levels weredetermined once per week for the three months of the study and justbefore the mice were killed. Blood samples were obtained from the lateralvein of the tail (blood glucose meter and OneTouch Ultra test strips, Life-scan, France).

2.3. Glycated haemoglobin determinations

Just before the animals were killed, blood samples from the right retro-orbital venous sinus were collected into calcium ethylenediaminotetra-ace-tic acid (EDTA). Glycated haemoglobin levels were determined automat-ically, by high-performance liquid chromatography, as described by Ouand Rognerud (1993). Briefly, we diluted 10 lL of blood in a hypotoniccell lysis solution (0.1% Brij, Sigma). Haemoglobins were then separatedon a porous silica column containing polyaspartic acid (Poly CatA, PolyLC, Houston, USA). Various ionic gradients were obtained by mixingtwo buffer solutions (20 mM bis-Tris, 2 nM KCN, pH 5.8 and 20 mMbis-Tris, 2 nM KCN, 75 mM trisodium citrate, pH 5.8). These gradientswere used to separate the haemoglobins, which were detected by spectros-copy at 418 nm.

2.4. Ocular examination, funduscopy and fluorescein angiographs

Fundus examinations and angiographs were performed on all mice, 15days, 1 month and 3 months after the pharmacological induction of diabe-tes. Cornea, eye lens and fundus were examined with a biomicroscope

614 D. Gaucher et al. / Vision Research 47 (2007) 612–623

(BQ, Haag-Streit, Germany) and a 90-dioptre lens (Volk, USA). Photo-graphs of the eye fundus and angiographs were obtained with a scanninglaser ophthalmoscope (SLO) (HRA, Heidelberg, Germany). Retinal angi-ography was carried out after the intraperitoneal injection of 25% fluores-cein in 0.9% NaCl. Mouse pupils were dilated by the ocular instillation oftropicamide eye drops (0.5%). Fundus images were obtained for each eyebefore, 1 and 5 min after the fluorescein injection.

2.5. Retinal electrophysiology

Electroretinograms (ERG) were obtained from all animals, 15 days,1 month and 3 months after diabetes induction. Animals were acclimatisedto the dark for 12 h before the recording. They were anaesthetised by intra-peritoneal injection (10 ll/g) of a mixture of ketamine (10%) and xylazine(7.5%) in 0.9% NaCl. Corneas were anaesthetised with oxybuprocaine chlor-hydrate (0.4%) and the pupil was dilated with tropicamide (0.5%). Each ani-mal was placed on a heating pad, and the eyelids were retracted to keep theeye open during the recording. A gold electrode was placed on the corneawith a drop of methylcellulose and the neutral and reference electrodes wereplaced on the tail and head of the animal, respectively. Light stimulation wasdelivered via a Ganzfeld light source, with flash intensities from 10�4 to25 cd s m�2. Responses were amplified, high- and low-pass-filtered(1–300 Hz) and digitised (Multiliner Vision, Toennies/Jaeger, Hoechberg,Germany). Oscillatory potentials (OP) were isolated for a light stimulus of3 cd s m�2 by filtering the responses with a high-pass filter at 100 Hz and alow-pass filter at 300 Hz. The amplitudes and latencies of the ERG a- andb-waves were measured at the maximum negative and positive peaks ofthe recordings with respect to the baseline before stimulation. As recom-mended by the ISCEV (International Society for Clinical Electrophysiologyof Vision), OP amplitudes were measured from the negative peak to the nextpositive peak whereas OP latencies were measured at the positive peaks. Wesummed OP amplitudes (Sampl) and latencies (Slat) for the first 5 OPbecause these parameters provide information about inner retinal cell func-tion (amacrine cells) and were found to be modified in other diabetic animalmodels (Hancock & Kraft, 2004; Hotta et al., 1997) and in diabetic patients(Bresnick & Palta, 1987; Yoshida et al., 1991).

2.6. Cell labelling

Animals were killed three months after diabetes induction. The left eyeof each animal was enucleated for the labelling of cells on vertical retinalsections, whereas the right eye was dissected for the preparation of a flat-mounted retina. Eye cups and retina were fixed overnight in 4% parafor-maldehyde (PAF) at 4 �C. A few flat-mounted retina (from one controland two diabetic mice) were embedded in orbithine carbamyl transferase(OCT) medium and cut into 40 lm-thick horizontal retinal sections. Entireretinas were stained for GFAP as described by Rungger-Brandle et al.(2000), by incubation with the polyclonal anti-GFAP antibody (2.5 days,4 �C), followed by washing and incubation with the secondary antibody(1.5 days, 4 �C). Retinae were incubated with antibodies in a permeabilis-ing solution containing 1% Triton X-100 in phosphate-buffered saline(PBS). Vertical retinal sections were cut from eye cups embedded inOCT medium after cryoprotection in various sucrose solutions. Verticalsections were permeabilised by incubation for 1 h at 37 �C with 0.5% Tri-ton X-100 in a solution containing 1% bovine serum albumin (BSA) and0.05% Tween 20 in PBS. Sections were incubated overnight at 4 �C withprimary antibodies diluted in permeabilising solution. The primary anti-bodies used were rabbit anti-GFAP (1/200, Dako, USA) and rabbitanti-phosphotyrosine (1/500, Zymed, USA) antibodies. Sections werewashed several times in PBS and then incubated for 1 h at room temper-ature with the secondary antibody in permeabilising solution supplement-ed with 4 0,6-diamidino-2-phenylindole (DAPI) for nuclear staining (1/500,Sigma). The secondary antibody used was an Alexa fluor 488- or 594-con-jugated goat anti-rabbit antibody (1/500, Molecular Probes, USA).

Neuronal apoptosis was investigated by terminal UTP nick-end label-ling (TUNEL) coupled with fluorescein (TUNEL-FITC detection kit;Roche Diagnostics, France). Vertical sections (5–8 per retina) and hori-

zontal sections (6 per eye) of retina were permeabilised by incubationfor 5 min with 0.1% Triton X-100 in PBS. They were then stained by incu-bation with 4 0,6-diamidino-2-phenylindole (DAPI) for 5 min, and incubat-ed with the kit solution for 1 h at 37 �C.

2.7. Cell quantification

All specimens were viewed with a fluorescence microscope (LeicaDM5000B, Germany) equipped with a digital camera (Coolsnap fx, Pho-tometrix, Roper Scientific, USA) linked to a computer running imageanalysis software (Metavue 6.2R4, Universal Imaging Corp, USA). Digi-tal images were acquired using identical settings in all cases. Microglialcells were counted and the length of their dendrites was determined ontwo different retinal sections including the optic nerve for each animal.Dendrites were measured with Metavue software.

2.8. Statistical analysis

Statistical comparisons were made by unpaired t-tests when possible,and by non-parametric Mann–Whitney U-tests otherwise. Statistical anal-ysis was carried out with Statview 5.0 (SAS Institute Inc., USA). Signifi-cance tests were performed with (p = 0.05) taken as the threshold forsignificance.

3. Results

3.1. Diabetes follow-up

Following the injection of alloxan, the induction of dia-betes was confirmed based on weight stagnation, polydip-sia and polyuria in mice (Fig. 1). The day after alloxaninjection, glucose was detected in the urine and blood glu-cose levels at least doubled in injected mice. The blood glu-cose levels ranged from 5.89 to 13.68 mM in controlanimals and from 19.68 to 33.36 mM in diabetic mice.Blood glucose levels remained three times higher in diabeticthan in control animals throughout the three-month studyperiod and the weight gain of these animals was reduced byapproximately 50% (Fig. 1). Three months after the diabe-tes induction, glycated haemoglobin levels were 2.72%(±0.06, n = 9, SEM) in control mice and 6.54% (±0.82,n = 5, SEM) in diabetic animals. Thus, despite the contin-uous insulin treatment, the diabetic mice displayed chronichyperglycaemia.

3.2. No vascular lesion

We evaluated the integrity of the blood retinal barrier,by examining diabetic and control animals by fluoresceinangiography with a scanning laser ophthalmoscope. Nohaemorrhage, microaneurism or proliferation of new ves-sels was detected. Fig. 2 shows fluorescein angiograms forcontrol and diabetic animals. We assessed the diffusion ofthis dye out of the capillary network by analysing angio-grams taken at different time points after fluorescein injec-tion, because fluorescein diffusion into the retinal tissueprovide evidence for the breakdown of the internal BRB(Rakoczy et al., 2003). However, no fluorescein diffusioninto the retinal tissue was detected and the morphologyand density of the retinal capillary network remained

Blood glucose (mM)

05

101520

2530

3540

0 1 2 3Months

diabetic mice

control mice

Weight (g)

20

22

24

26

28

30

32

34

0 1 2 3Months

diabetic mice

control mice

Fig. 1. Blood glucose concentration (A) and animal weight (B) in control mice (n = 9 at study ending) and mice with alloxan-induced diabetes (n = 5 at theend of the study) over a 3-month period. Each graph shows changes in these parameters for the control (squares) and diabetic (triangles) mice during the 3months of the experiment. The induction of diabetes was demonstrated by the increase in blood glucose concentration and the stabilisation of body weight.(Means and the standard errors of means (SEM) are indicated.)

D. Gaucher et al. / Vision Research 47 (2007) 612–623 615

unchanged in diabetic and control mice throughout theentire study (Fig. 2). These observations suggest thatthe BRB was not ruptured during the three months ofthe experiment in mice with alloxan-induced diabetes.

3.3. Functional changes

We investigated the possible effects of diabetes on neuro-nal cell activity in our model, by recording electroretino-grams at various time points after the induction ofdiabetes. We checked that the cornea and ocular lens wereclear before each recording, because cataracts are oftenobserved in diabetic animals. No statistically significantchange in ERG was detected one month after diabetesinduction. After three months, the amplitudes and latenciesof the ERG a- and b-waves remained unchanged, but a sig-nificant decrease in the b-wave/a-wave amplitude ratio (b/a) was observed for light intensities of 10 and 25 cd s m�2 (ttest, p = 0.014 and 0.018 respectively). The sum of OPlatencies was also higher in diabetic mice (n = 5, t test,p = 0.034) than in control mice (n = 9) (Fig. 3, Table 1).These observations suggest that three months of diabetesled to changes in neuronal cell activity.

3.4. Absence of neural apoptosis

In diabetic animal models, neuronal apoptosis has beenreported as early as one month after diabetes induction(Barber et al., 1998). We therefore investigated the distribu-tion of apoptotic cells in vertical and horizontal retinal sec-tions, using TUNEL and cleaved caspase 3 detection. Eventhree months after the induction of diabetes, no apoptoticcells were detected with either of the cell death markers.For example, Fig. 4A and B shows the absence of TUNELstaining on vertical retinal sections from control animalsand diabetic mice three months after diabetes induction.

The efficacy of our labelling technique was checked bystaining retinal sections from an rd (retinal degeneration)mouse by TUNEL. As expected, in this animal presentingphotoreceptor degeneration, several TUNEL-positiveapoptotic cells were detected in the outer nuclear layer(Fig. 4C, arrows). TUNEL was also performed on horizon-tal sections, because only a few retinal ganglion cells arepresent in vertical sections and these cells have been report-ed to undergo degeneration early in progression to DR indiabetic patients and in rats with streptozotocin-induceddiabetes (Barber et al., 1998). No TUNEL-positive cellswere detected on horizontal sections containing large num-bers of retinal ganglion cells (data not shown). Despite theobserved change in retinal cell activity, no retinal cell apop-tosis was detected three months after the induction of dia-betes with alloxan.

3.5. Absence of macroglial reaction

As a glial reaction has been described in diabetic animalmodels and human patients (Agardh et al., 2001; Hancock& Kraft, 2004; Li et al., 2002; Lieth et al., 1998; Rungger-Brandle et al., 2000; Sakai et al., 1995), we labelled verticalretinal sections with an anti-GFAP antibody. In controlanimals, this antibody labelled astrocytes in the ganglioncell layer (Fig. 4D). Three months after the induction ofdiabetes, no abnormal glial reaction was observed in dia-betic mouse retina. A few Muller cells were stained in theinner retina (Fig. 4E, arrow), but such staining was alsoobserved in a few control retinas. We stained a retinal sec-tion from an rd mouse with the anti-GFAP antibody, toobtain information about the nature of reactive glial cells.In this mouse, the anti-GFAP antibody strongly labelledmany glial cell processes extending from the outer plexi-form layer to the inner limiting membrane (Fig. 4F, arrow).We investigated possible changes in morphology in these

Fig. 2. Absence of vascular changes in mice with alloxan-induced diabetes. Retinal angiographies showing the superficial vessels (A–C) and deep capillarynetworks (D–E) in a control animal (A and D) (n = 5) and in mice with alloxan-induced diabetes (n = 5) (B, C and E) at the end of the 3-month studyperiod. These images show normally perfused retinal vessels, with no new vessels and no increase in vascular permeability.

616 D. Gaucher et al. / Vision Research 47 (2007) 612–623

labelled astrocytes by staining the whole retina for GFAP.In diabetic animals, the morphology of the GFAP-positiveastrocytes was normal on flat-mounted retina (Fig. 4G andH). This examination demonstrated the absence of amacroglial reaction in mice with alloxan-induced diabetes,three months after diabetes induction.

3.6. Microglial changes

Microglial cells in the ganglion cell layer, the innerplexiform layer and the inner nuclear layer werevisualised by specific staining with an anti-phosphotyrosineantibody (Fig. 5). The specificity of immunolabelling was

ERG b/a wave amplitude ratio

0

2

4

6

8

10

100 300 1000 3000 10000 25000Intensity (mcd.s.m-2)

diabetic

control

Oscillatory potentials

0

200

400

600

800

Sampl Slat

µV

200

400

600

800

Sampl Slat

µV

0

2

4

6

8

10

100 300 1000 3000 10000 25000Intensity (mcd.s.m-2)

0

2

4

6

8

10

100 300 1000 3000 10000 25000Intensity (mcd.s.m-2)

* *

0

200

400

600

800

Sampl Slat

µV

*

msec

msec

msec

200

400

600

800

0

200

400

600

800

200

400

600

800

diabeticcontrol

Scotopic ERG, 3 months Oscillatory potentials, 3 months

60 ms

100 µV60 ms

100 µV

diabeticcontrol

15 days 15 days

1 month

3 months 3 months

1 month

0

0

0

B F

GC

D H

Fig. 3. Functional changes were recorded in mice with alloxan-induced diabetes, 15 days (A and E), 1 month (B and F) and 3 months (C, D, G and H)after the induction of diabetes. (A–C) The ratio of B and A waves on the electroretinogram (ERG) was significantly lower in diabetic animals than incontrol animals for the strongest stimulus, after 3 months of diabetes (C). (D) Representative recording of scotopic ERG responses to a 10 cd s/m2 lightstimulus (arrowhead), showing the decrease in b/a amplitude ratio in a diabetic mouse (dark line) 3 months after the onset of diabetes, and comparisonwith a control mouse (grey line). (E–G) Oscillatory potentials (OP) remained unchanged in terms of summed amplitudes (Sampl) at all time points whereasthe sum of their latencies (Slat) was significantly higher in diabetic than in control mice after three months (G). (H) Representative recording of OP to a3 cd s/m2 light stimulus (arrowhead), showing the longer OP latency in a diabetic mouse (dark line) 3 months after the onset of diabetes than in a contromouse (grey line). (Means and standard errors of the means (SEM) are indicated in the bar graphs, *significant difference, t test, p < 0.05).

D. Gaucher et al. / Vision Research 47 (2007) 612–623 617

l

Table 1Changes in electroretinogram measurements in diabetic mice

Time (months) a wave lV ± SD b wave lV ± SD Ratio b/a ±SD Sampl Ops lV ± SD Slat Ops ms ± SD

Stimulus (cd s m�2) 10 10 10 3 3

Control mice 0.5 182.1 ± 48.5 447.5 ± 139.2 2.4 ± 0.4 781.5 ± 250.2 159.6 ± 25.1Diabetic mice 0.5 216.7 ± 71 479.3 ± 217 2 ± 0.3 726.6 ± 230.5 174.7 ± 34.6

Control mice 1 228.4 ± 39.7 448.3 ± 137 1.9 ± 0.4 647.8 ± 193.8 174.4 ± 15.3Diabetic mice 1 189.9 ± 47.7 368.1 ± 65.9 2 ± 0.5 566.1 ± 79.7 183.7 ± 9.4

Control mice 3 163.6 ± 36.1 365 ± 58.6 2.2 ± 0.3 686.2 ± 139.4 171.9 ± 10.9Diabetic mice 3 207.7 ± 63.9 368 ± 99.1 1.8 ± 0.3* 611.4 ± 196.7 187.2 ± 12.6*

Mean amplitudes of a and b waves, mean b/a amplitude ratio and mean of the summed oscillatory potential (OP) amplitudes (Sampl) and latencies (Slat)in mouse retina, 15 days, 1 month and 3 months after diabetes induction.

* Significant difference, t test, p < 0.05.

Fig. 4. Absence of macroglial reaction and apoptosis in animals with alloxan-induced diabetes, after 3 months. (A–C) No TUNEL-positive apoptotic cellswere detected in the diabetic (B) and control (A) mice, whereas such cells were detected in rd mice (C). (D–F) GFAP expression in retinal sections ofdiabetic mice (E) was no stronger than that in control animals (D), whereas stronger staining was observed with sections from rd mice (F). (G and H)GFAP-positive astrocytes displayed no morphological alterations on flat-mounted retinas from control animals (G) and diabetic mice (H). Scale barsrepresent 10 lm. GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; ONL, outer nuclear layer.

618 D. Gaucher et al. / Vision Research 47 (2007) 612–623

demonstrated on a vertical section of retina from a mousestrain in which the gene encoding the chemokine receptorCX3CR1 was replaced by the gene encoding enhancedgreen fluorescent protein (EGFP) (Jung et al., 2000). Inthese animals, brain and retinal microglial cells, which

would normally express CX3CR1, become EGFP-positive.These EGFP-positive retinal microglial cells wereimmunolabelled with the anti-phosphotyrosine antibody(Fig. 5A–C). In vertical retinal sections, microglial cellnuclei and processes were clearly identified and quantified

Fig. 5. Microglial changes in mice with alloxan-induced diabetes. (A–C) Microglial cell immunolabelling with the anti-phosphotyrosine antibody in mice.A microglial cell expressing green fluorescent protein (GFP) (A) is immunolabelled with the anti-phosphotyrosine antibody (B) in a CX3CR1-deficientmouse, as indicated on the merged image (C). (D, E, G and H) Retinal sections stained with the anti-phosphotyrosine antibody showing immunopositivemicroglial cells in control mice (D and G) and in mice with alloxan-induced diabetes, after 3 months (E and H). Microglial cells appeared to have largercell soma and shorter cell processes in diabetic animals (E and H) than in control mice (D and G). In (D and E), the nuclear dye DAPI was used to detectthe various nuclear layers. (F) Quantification of phosphotyrosine-immunopositive microglial cells in different retinal layers of control animals and micewith alloxan-induced diabetes, after 3 months. Microglial cell nuclei were quantified in the GCL, IPL and INL. No statistically significant differences in thenumber of microglial cells were found between controls and diabetic animals (p > 0.05) (B). (I) Quantification of the cumulative length of microglial celldendrites (arrows in D and E), measured on complete retinal sections in control (G) and diabetic animals (H). Dendrites were significantly shorter in thediabetic mice (*p = 0.01). GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer. Scale bars represent 10 lm (A–H) and 25 lm (Gand H). (Means and the standard errors of means (SEM) are indicated in the graph.)

D. Gaucher et al. / Vision Research 47 (2007) 612–623 619

620 D. Gaucher et al. / Vision Research 47 (2007) 612–623

in the inner nuclear layer (INL), inner plexiform layer(IPL) and ganglion cell layer (GCL). Diabetic and controlanimals had similar numbers of phosphotyrosine-positivemicroglial cells (Fig. 5F), but the cell bodies of these cellsseemed to be larger in diabetic animals, with thicker prox-imal processes and shorter dendrites (Fig. 5D–E). Thecumulative length of microglial dendrites was determinedon complete sections including the optic nerve (Fig. 5Gand H), and was found to differ significantly between dia-betic and control animals (Fig. 5I, t test, p = 0.01). Asmicroglial cells are known to display a change in morphol-ogy, including dendrite contraction, in the early stages ofactivation, these observations suggest that microglial cellswere activated three months after the onset of diabetes inmice with alloxan-induced diabetes.

4. Discussion

This study on mice with alloxan-induced diabetesshowed that changes to oscillatory potentials and ERGb/a amplitude wave ratio first became detectable threemonths after the onset of diabetes. No change in blood ves-sel permeability was detected in vivo by fluorescein angiog-raphy carried out with a scanning laser ophthalmoscope.No ganglion cell apoptosis or macroglial reaction wasdetected, but changes in microglial cells were observed,with a decrease in the length of microglial cell processes,which is known to occur early in microglial activation.Insulin treatment was insufficient to regulate blood glucoselevels, but may have protected against neuronal apoptosis.

4.1. The sequence of events in the retina following the

induction of diabetes

In diabetic patients, OPs provide the first sign of retinalalteration, as their amplitude decreases and their latencyincreases even at a pre-retinopathy stage characterised byan absence of detectable vascular changes on funduscopy(Shirao & Kawasaki, 1998). Other ERG changes have alsobeen observed, including a change in ERG pattern (Caputoet al., 1990), scotopic and photopic b-wave alterations(Holopigian, Seiple, Lorenzo, & Carr, 1992) and a specificreduction of S-cone ERG (Yamamoto, Kamiyama, Nitta,Yamada, & Hayasaka, 1996). Changes in OP latenciesand amplitudes have also been described in various animalmodels (Li et al., 2002; Sakai et al., 1995; Segawa, Hirata,Fujimori, & Okada, 1988). In rats with STZ-induced diabe-tes, these OP changes are detected as early as two weeksafter diabetes induction (Li et al., 2002; Sakai et al.,1995). Other changes in ERG responses, such decreasesin b-wave magnitude, have also been observed in these ani-mal models (Li et al., 2002). In our mice with alloxan-in-duced diabetes, no significant changes were detected inERG a- and b-waves, but changes in the b/a wave ampli-tude ratio were statistically significant three months afterthe onset of diabetes. Indeed, this ratio could be used tonormalise ERG measurements between mice. However,

the variability of this measure in control animals at the var-ious time points studied (Table 1) limits the value of the dif-ference observed at three months. By contrast, OP latenciessteadily increased over the first three months, providingevidence for the importance of the statistically significantincrease in OP latency reported at 3 months. Similarincreases in OP latency have been reported for streptozoto-cin-treated rats after five weeks of diabetes (Li et al., 2002;Sakai et al., 1995) and in diabetic patients with and withoutretinopathy (Bresnick & Palta, 1987; Yoshida et al., 1991).The presence of such increases in OP latency in diabeticpatients without retinopathy and in our alloxan-treatedmice is consistent with the notion that this change is a veryearly marker of diabetes-induced retinal changes. TheseERG changes are consistent with the suggested changesin the GABA-ergic system or with dysfunction of the innerretinal vasculature (Hanitzsch, Kuppers, & Flade, 2004;Ishikawa, Ishiguro, & Tamai, 1996a; Ishikawa, Ishiguro,& Tamai, 1996b; Shirao & Kawasaki, 1998; Wachtmeister,1998).

In diabetic patients and animals, retinal cell apoptosis inthe various retinal layers has been described as an earlyevent, preceding vascular alterations (Asnaghi et al.,2003; Barber et al., 1998; Martin et al., 2004; Ning et al.,2004; Park et al., 2003). In animal models, this cell death,detected by both TUNEL and immunolabelling for activecaspase-3, has generally been detected in the ganglion celllayer one month after the induction of diabetes, but itmay also occur earlier, after as little as two weeks (Ninget al., 2004). Retinal cell death has also been observed inthe photoreceptor layer and the inner nuclear layer, butduring a later phase of the disease (Ning et al., 2004; Parket al., 2003). In mice with STZ-induced diabetes, Asnaghireported no increase in the number of apoptotic cells after10 and 24 weeks of diabetes, whereas Martin reported suchretinal apoptosis at 2, 6 and 12 weeks (Asnaghi et al., 2003;Martin et al., 2004). In mice with alloxan-induced diabetes,no cell death was detected in any cell layer after threemonths of diabetes, even if horizontal sections were usedso that large numbers of ganglion cells could be studiedat a glance. These results indicate that the early ERGchanges are not a consequence of retinal cell death.

Several studies have reported that ERG modificationsand ganglion cell apoptosis are followed by a glial reactionin diabetic patients and animal models (Agardh et al., 2001;Hancock & Kraft, 2004; Li et al., 2002; Lieth et al., 1998;Rungger-Brandle et al., 2000; Sakai et al., 1995). In ratswith STZ-induced diabetes, the glial reaction appearedafter 6–15 weeks of diabetes (Li et al., 2002; Lieth et al.,1998). In mice with alloxan-induced diabetes, ERG modifi-cations appeared only after 3 months of diabetes and noglial reaction, as assessed by GFAP labelling, could bedetected. Further studies should determine whether this gli-al reaction and ganglion cell death occur even later in thismodel.

Microglial cells are also activated in rats with STZ-in-duced diabetes (Rungger-Brandle et al., 2000; Zeng et al.,

D. Gaucher et al. / Vision Research 47 (2007) 612–623 621

2000) and in Ins2 Atika mice (Barber et al., 2005). In rats,this activation is demonstrated by the hypertrophy ofmicroglial cells after 1 month of diabetes, the number ofcells increased after 4 months these cells migrated intothe outer nuclear layer after 14 to 16 months (Zeng et al.,2000). Rungger-Brandle et al. (2000) detected an increasein the number of microglial cells as early as 1 month afterdiabetes induction, before the increase in GFAP expres-sion. Barber found swollen and retracted microglial cellprocesses in Ins2 Atika mice (Barber et al., 2005), asobserved in our mice with alloxan-induced diabetes. Noincrease in the number of retinal microglial cells was detect-ed in mice with alloxan-induced diabetes, but we didobserve a change in the morphology of these cells, asdescribed above. These changes in the morphology ofmicroglial cells are known to be associated with themicroglial ‘‘alert’’ activation stage (Raivich et al., 1999).Our findings suggest that microglial cells may be activatedin the diabetic mouse retina before ganglion cell death andthe subsequent macroglial reaction. These observations areconsistent with findings for Ins2 Atika mice showingmicroglial activation but no increase in GFAP expressionby Muller cells (Barber et al., 2005). Microglial cell activa-tion was correlated with the ERG changes observed in ourstudy, and may trigger the ganglion cell death reported inother diabetic animal models. This sequence of events issupported by the neuroprotection conferred by the anti-in-flammatory drug, minocycline, in animals with streptozo-tocin-induced diabetes (Krady et al., 2005). The sequenceof retinal modifications early in DR can therefore bededuced from the combined results of all these studies:microglial activation may initially be associated with OPchanges and ERG b/a amplitude ratio reduction; thesemodifications are followed by ganglion cell apoptosis andsubsequent macroglial reaction.

4.2. Differences between diabetic animal models

There seem to be discrepancies in the sequence of cellu-lar events in the various diabetic animal models described.For example, the permeability of the BRB increases as ear-ly as three to ten days after the induction of diabetes in ratsby streptozotocin injection (Diani et al., 1987). In Ins2 Ati-ka mice, the BRB was broken down only after 12 weeks ofdiabetes, and in our mice with alloxan-induced diabetes,even this period was not long enough. These differencesmay result from differences in the techniques used to studythe BRB, or from differences in the animal strains and spe-cies (as supported by the findings of Zhang et al., 2005) orin the origin of the diabetes, with the BRB seeming to bemuch more resistant in diabetic mice than in diabetic rats.

Another main difference between these animal modelsconcerns the induction of neuronal apoptosis, principallyin the ganglion cell layer. Ganglion cell apoptosis has beenreported to occur as early as 15 days after the onset of dia-betes in mice with STZ-induced diabetes (Martin et al.,2004) and at 1 month in KKAY spontaneous diabetic

mice, which also have a triglyceride metabolism disorder(Ning et al., 2004). In the Ins2 atika mice, apoptotic cellswere detected as four weeks after the onset of diabetesand the number of cell bodies in the ganglion cell layerwas found to be smaller than in the control after 22 weeksof diabetes (Barber et al., 2005). By contrast, no ganglioncell apoptosis was detected in our model of mice with allox-an-induced diabetes, even after three months. Alloxan andSTZ both trigger pancreatic b-cell necrosis and diabetes:alloxan does so by creating an oxidative stress whereasSTZ induces DNA alkylation (Szkudelski, 2001). No dataare available concerning the relative toxicities of these twodrugs to the mouse retina. However, the reported STZ-in-duced ganglion cell death is unlikely to be due to a directtoxic effect on the retina, for two reasons. Firstly, Asnaghiet al. (2003) reported an absence of glial reactions and gan-glion cell apoptosis in mice with STZ-induced diabetes,three months after the induction of diabetes (Asnaghiet al., 2003). Secondly, insulin injection can reduce thenumber of apoptotic ganglion cells in rats with STZ-in-duced diabetes (Barber et al., 1998).

The absence of ganglion cell apoptosis in our study andin that by Asnaghi et al. (2003) may result from neuropro-tective effects of insulin. In both these studies, the diabeticmice received regular insulin injections, whereas, in otherstudies reporting ganglion cell apoptosis, the diabetic micereceived no insulin (Asnaghi et al., 2003; Barber et al.,1998; Martin et al., 2004; Ning et al., 2004; Park et al.,2003). Insulin is classically administered to normalise bloodglucose levels, to prevent complications due to hyperglyca-emia. In rats with STZ-induced diabetes, insulin injectionscan restore normal glycaemia and limit the number ofapoptotic ganglion cells (Barber et al., 1998). In our study,the ganglion cell survival cannot be attributed to insulin-mediated control over glycaemia, because blood glucoselevels were not normalised by insulin injections duringthe study period. Insulin was administered to prevent therapid death of untreated animals. Similarly, insulin hasbeen shown to mediate the neuroprotection of ganglioncells in rats with STZ-induced diabetes, despite incompletecontrol of hyperglycaemia (Asnaghi et al., 2003). Insulinneuroprotection has also been reported in cultured R28 ret-inal neurons, following serum deprivation (Wu et al.,2004). The injection of insulin into mice with alloxan- orSTZ-induced diabetes may, thus, also provide neuroprotec-tion for ganglion cells.

5. Conclusion

In summary, we report here the absence of neuronalapoptosis three months after the induction of diabetes byalloxan injection in mice. By contrast, the microglial chang-es and electroretinographic alterations reported in othermodels of DR were observed. In future studies, we willinvestigate whether this early inflammatory response cantrigger ganglion cell death and the macroglial reaction inanimals with DR.

622 D. Gaucher et al. / Vision Research 47 (2007) 612–623

Acknowledgments

We thank Dr. Nathalie Mario (Hopital St. Antoine,Paris, France) for the Glycated haemoglobin determina-tions. This work was supported by INSERM, UniversityPierre and Marie Curie (Paris VI), Assistance Publique-Hopitaux de Paris (AP-HP), Federation des Aveugles etHandicapes Visuels de France, RETINA-France, Associa-tion francaise contre les Myopathies (AFM) and the Euro-pean Economic Community (EVI-GENORET-512036).DG received a fellowship from Federation des Aveugleset Handicapes Visuels de France.

References

Abu-El-Asrar, A. M., Dralands, L., Missotten, L., Al-Jadaan, I. A., &Geboes, K. (2004). Expression of apoptosis markers in the retinas ofhuman subjects with diabetes. Investigative Ophthalmology and Visual

Science, 45(8), 2760–2766.Agardh, E., Bruun, A., & Agardh, C. D. (2001). Retinal glial cell

immunoreactivity and neuronal cell changes in rats with STZ-induceddiabetes. Current Eye Research, 23(4), 276–284.

Asnaghi, V., Gerhardinger, C., Hoehn, T., Adeboje, A., & Lorenzi, M.(2003). A role for the polyol pathway in the early neuroretinalapoptosis and glial changes induced by diabetes in the rat. Diabetes,

52(2), 506–511.Barber, A. J., Antonetti, D. A., Kern, T. S., Reiter, C. E., Soans, R. S.,

Krady, J. K., et al. (2005). The Ins2Akita mouse as a model of earlyretinal complications in diabetes. Investigative Ophthalmology and

Visual Science, 46(6), 2210–2218.Barber, A. J., Lieth, E., Khin, S. A., Antonetti, D. A., Buchanan, A. G., &

Gardner, T. W. (1998). Neural apoptosis in the retina duringexperimental and human diabetes. Early onset and effect of insulin.Journal of Clinical Investigation, 102(4), 783–791.

Bresnick, G. H., & Palta, M. (1987). Temporal aspects of the electroret-inogram in diabetic retinopathy. Archives of Ophthalmology, 105(5),660–664.

Caputo, S., Di Leo, M. A., Falsini, B., Ghirlanda, G., Porciatti, V.,Minella, A., et al. (1990). Evidence for early impairment of macularfunction with pattern ERG in type I diabetic patients. Diabetes Care,

13(4), 412–418.Ciulla, T. A., Amador, A. G., & Zinman, B. (2003). Diabetic retinopathy

and diabetic macular edema: pathophysiology, screening, and noveltherapies. Diabetes Care, 26(9), 2653–2664.

Davis, M. D. (1992). Diabetic retinopathy. A clinical overview. Diabetes

Care, 15(12), 1844–1874.Diani, A. R., Sawada, G. A., Zhang, N. Y., Wyse, B. M., Connell, C. L.,

Vidmar, T. J., et al. (1987). The KKAy mouse: a model for the rapiddevelopment of glomerular capillary basement membrane thickening.Blood Vessels, 24(6), 297–303.

Ewing, F. M., Deary, I. J., Strachan, M. W., & Frier, B. M. (1998). Seeingbeyond retinopathy in diabetes: electrophysiological and psychophys-ical abnormalities and alterations in vision. Endocrine Reviews, 19(4),462–476.

Fortune, B., Schneck, M. E., & Adams, A. J. (1999). Multifocalelectroretinogram delays reveal local retinal dysfunction in earlydiabetic retinopathy. Investigative Ophthalmology and Visual Science,

40(11), 2638–2651.Hancock, H. A., & Kraft, T. W. (2004). Oscillatory potential analysis and

ERGs of normal and diabetic rats. Investigative Ophthalmology and

Visual Science, 45(3), 1002–1008.Hanitzsch, R., Kuppers, L., & Flade, A. (2004). The effect of GABA and

the GABA-uptake-blocker NO-711 on the b-wave of the ERG and theresponses of horizontal cells to light. Graefe’s Archives for Clinical and

Experimental Ophthalmology, 7, 7.

Holopigian, K., Seiple, W., Lorenzo, M., & Carr, R. (1992). A comparisonof photopic and scotopic electroretinographic changes in early diabeticretinopathy. Investigative Ophthalmology and Visual Science, 33(10),2773–2780.

Hotta, N., Nakamura, J., Sakakibara, F., Hamada, Y., Hara, T., Mori,K., et al. (1997). Electroretinogram in sucrose-fed diabetic rats treatedwith an aldose reductase inhibitor or an anticoagulant. American

Journal of Physiology, 273(5 Pt 1), E965–E971.Ishikawa, A., Ishiguro, S., & Tamai, M. (1996a). Accumulation of

gamma-aminobutyric acid in diabetic rat retinal Muller cells evidencedby electron microscopic immunocytochemistry. Current Eye Research,

15(9), 958–964.Ishikawa, A., Ishiguro, S., & Tamai, M. (1996b). Changes in GABA

metabolism in streptozotocin-induced diabetic rat retinas. Current Eye

Research, 15(1), 63–71.Jung, S., Aliberti, J., Graemmel, P., Sunshine, M. J., Kreutzberg, G. W.,

Sher, A., et al. (2000). Analysis of fractalkine receptor CX(3)CR1function by targeted deletion and green fluorescent protein reportergene insertion. Molecular and Cellular Biology, 20(11), 4106–4114.

Krady, J. K., Basu, A., Allen, C. M., Xu, Y., Lanoue, K. F., Gardner, T.W., et al. (2005). Minocycline reduces proinflammatory cytokineexpression, microglial activation, and caspase-3 activation in a rodentmodel of diabetic retinopathy. Diabetes, 54(5), 1559–1565.

Lai, C. M., Dunlop, S. A., May, L. A., Gorbatov, M., Brankov, M., Shen,W. Y., et al. (2005). Generation of transgenic mice with mild andsevere retinal neovascularisation. British Journal of Ophthalmology,

89(7), 911–916.Larger, E., Becourt, C., Bach, J. F., & Boitard, C. (1995). Pancreatic islet

beta cells drive T cell-immune responses in the nonobese diabeticmouse model. Journal of Experimental Medicine, 181(5), 1635–1642.

Lieth, E., Barber, A. J., Xu, B., Dice, C., Ratz, M. J., Tanase, D., et al.(1998). Glial reactivity and impaired glutamate metabolism in short-term experimental diabetic retinopathy. Penn State Retina ResearchGroup. Diabetes, 47(5), 815–820.

Li, Q., Zemel, E., Miller, B., & Perlman, I. (2002). Early retinal damage inexperimental diabetes: electroretinographical and morphologicalobservations. Experimental Eye Research, 74(5), 615–625.

Lowitt, S., Malone, J. I., Salem, A., Kozak, W. M., & Orfalian, Z. (1993).Acetyl-L-carnitine corrects electroretinographic deficits in experimentaldiabetes. Diabetes, 42(8), 1115–1118.

Martin, P. M., Roon, P., Van Ells, T. K., Ganapathy, V., & Smith, S. B.(2004). Death of retinal neurons in streptozotocin-induced diabeticmice. Investigative Ophthalmology and Visual Science, 45(9),3330–3336.

Ning, X., Baoyu, Q., Yuzhen, L., Shuli, S., Reed, E., & Li, Q. Q. (2004).Neuro-optic cell apoptosis and microangiopathy in KKAY mouseretina. International Journal of Molecular Medicine, 13(1), 87–92.

Ou, C. N., & Rognerud, C. L. (1993). Rapid analysis of hemoglobinvariants by cation-exchange HPLC. Clinical Chemistry, 39(5),820–824.

Park, S. H., Park, J. W., Park, S. J., Kim, K. Y., Chung, J. W., Chun, M.H., et al. (2003). Apoptotic death of photoreceptors in the streptozo-tocin-induced diabetic rat retina. Diabetologia, 46(9), 1260–1268, Epub2003 Jul 1231.

Raivich, G., Bohatschek, M., Kloss, C. U., Werner, A., Jones, L. L., &Kreutzberg, G. W. (1999). Neuroglial activation repertoire in theinjured brain: graded response, molecular mechanisms and cues tophysiological function. Brain Research. Brain Research Reviews, 30(1),77–105.

Rakoczy, P. E., Brankov, M., Fonceca, A., Zaknich, T., Rae, B. C., & Lai,C. M. (2003). Enhanced recombinant adeno-associated virus-mediatedvascular endothelial growth factor expression in the adult mouseretina: a potential model for diabetic retinopathy. Diabetes, 52(3),857–863.

Ruberte, J., Ayuso, E., Navarro, M., Carretero, A., Nacher, V., Haurigot,V., et al. (2004). Increased ocular levels of IGF-1 in transgenic micelead to diabetes-like eye disease. Journal of Clinical Investigation,

113(8), 1149–1157.

D. Gaucher et al. / Vision Research 47 (2007) 612–623 623

Rungger-Brandle, E., Dosso, A. A., & Leuenberger, P. M. (2000). Glialreactivity, an early feature of diabetic retinopathy. Investigative

Ophthalmology and Visual Science, 41(7), 1971–1980.Sakai, H., Tani, Y., Shirasawa, E., Shirao, Y., & Kawasaki, K. (1995).

Development of electroretinographic alterations in streptozotocin-induced diabetes in rats. Ophthalmic Research, 27(1), 57–63.

Segawa, M., Hirata, Y., Fujimori, S., & Okada, K. (1988). Thedevelopment of electroretinogram abnormalities and the possible roleof polyol pathway activity in diabetic hyperglycemia and galactosemia.Metabolism, 37(5), 454–460.

Shirao, Y., & Kawasaki, K. (1998). Electrical responses from diabeticretina. Progress in Retinal Eye Research, 17(1), 59–76.

Szkudelski, T. (2001). The mechanism of alloxan and streptozotocinaction in B cells of the rat pancreas. Physiological Research, 50(6),537–546.

Tzekov, R., & Arden, G. B. (1999). The electroretinogram in diabeticretinopathy. Survey of Ophthalmology, 44(1), 53–60.

Wachtmeister, L. (1998). Oscillatory potentials in the retina: what do theyreveal. Progress in Retinal Eye Research, 17(4), 485–521.

Wu, X., Reiter, C. E., Antonetti, D. A., Kimball, S. R., Jefferson,L. S., & Gardner, T. W. (2004). Insulin promotes rat retinal

neuronal cell survival in a p70S6K-dependent manner. Journal

of Biological Chemistry, 279(10), 9167–9175, Epub 2003 Dec9165.

Yamada, H., Yamada, E., Higuchi, A., & Matsumura, M. (2005). Retinalneovascularisation without ischaemia in the spontaneously diabeticTorii rat. Diabetologia, 48(8), 1663–1668.

Yamamoto, S., Kamiyama, M., Nitta, K., Yamada, T., &Hayasaka, S. (1996). Selective reduction of the S cone electro-retinogram in diabetes. British Journal of Ophthalmology, 80(11),973–975.

Yoshida, A., Kojima, M., Ogasawara, H., & Ishiko, S. (1991). Oscillatorypotentials and permeability of the blood-retinal barrier in noninsulin-dependent diabetic patients without retinopathy. Ophthalmology,

98(8), 1266–1271.Zeng, X. X., Ng, Y. K., & Ling, E. A. (2000). Neuronal and microglial

response in the retina of streptozotocin-induced diabetic rats. Visual

Neuroscience, 17(3), 463–471.Zhang, S. X., Ma, J. X., Sima, J., Chen, Y., Hu, M. S., Ottlecz, A., et al.

(2005). Genetic difference in susceptibility to the blood–retina barrierbreakdown in diabetes and oxygen-induced retinopathy. American

Journal of Pathology, 166(1), 313–321.