Diminished apoptosis in hypoxic porcine retina explant cultures through hypothermia

Simulation of hypoxic processes in vitro can be achieved through cobalt chloride (CoCl2), which induces strong neurodegeneration. Hypoxia plays an important role in the progression of several retinal diseases. Thus, we investigated whether hypoxia can be reduced by hypothermia. Porcine retinal explants were cultivated for four and eight days and hypoxia was mimicked by adding 300 µM CoCl2 from day one to day three. Hypothermia treatment (30 °C) was applied simultaneously. Retinal ganglion, bipolar and amacrine cells, as well as microglia were evaluated via immunohistological and western blot analysis. Furthermore, quantitative real-time PCR was performed to analyze cellular stress and apoptosis. In addition, the expression of specific marker for the previously described cell types were investigated. A reduction of ROS and stress markers HSP70, iNOS, HIF-1α was achieved via hypothermia. In accordance, an inhibition of apoptotic proteins (caspase 3, caspase 8) and the cell cycle arrest gene p21 was found in hypothermia treated retinae. Furthermore, neurons of the inner retina were protected by hypothermia. In this study, we demonstrate that hypothermia lowers hypoxic processes and cellular stress. Additionally, hypothermia inhibits apoptosis and protects neurons. Hence, this seems to be a promising treatment for retinal neurodegeneration.

www.nature.com/scientificreports www.nature.com/scientificreports/ after each medium change. No differences were seen within the groups for each point in time, indicating that degenerative effects were not induced by the cultivation of retinae (Fig. 1C).
For further investigations of the effects of hypothermia on CoCl 2 , we performed hematoxylin & eosin staining of retinal cross-sections (Sup. Fig. S1A). As described previously 11 , CoCl 2 lead to a reduction of the retinal thickness. To evaluate whether hypothermia inhibited neurodegenerative effects of CoCl 2 on porcine retina, retinal thickness was measured. At both investigated points in time, the retinal thickness was reduced significantly through CoCl 2 (4 days: p = 0.01; 8 days: 0.03) in comparison to control retinae. For both points in time, four and eight days, hypothermia preserved retinal thickness, that CoCl 2 + 30 °C treated retinae were significantly thicker than CoCl 2 + 37 °C treated ones (4 days: p = 0.003; 8 days: p = 0.002) and no difference were seen between CoCl 2 + 30 °C retinae and control + 37 °C retinae (p > 0.6; Sup. Fig. S1B). These results indicate that hypothermia lowers oxidative stress induced by CoCl 2 , and preserved retinal thickness, which was reduced by CoCl 2 -treatment.
HSP70, a chaperon belonging to the heat shock protein family, is important for the correct folding process of proteins, and accumulates under stress conditions 19 . To evaluate the effect of CoCl 2 and hypothermia on cellular stress HSP70 mRNA expression was analyzed (Fig. 2F). At the early point in time, the HSP70 expression level was significantly higher in the CoCl 2 + 37 °C group (29.8 ± 5.7-fold; p = 0.0002) than in the control + 37 °C. CoCl 2 -stressed retinae treated with hypothermia (8.7 ± 7.4-fold; p = 0.04) still had an increased HSP70 expression compared to control ones, but interestingly, lowering the temperature significantly diminished the HSP70 expression in comparison to the CoCl 2 -stressed retinae at 37 °C (p = 0.0002). After eight days, the CoCl 2 + 37 °C group (8.6 ± 3.2-fold; p = 0.0002) presented a significantly increased mRNA expression level compared to the control + 37 °C group. Hypothermia treatment led to a significant reduction of HSP70 mRNA expression in the CoCl 2 -stressed hypothermia group (0.8 ± 0.1-fold; p = 0.0002) in comparison to the CoCl 2 + 37 °C group. Most importantly, these results prove the complete inhibition of cellular stress after hypothermia treatment since no differences were seen between the CoCl 2 + 30 °C group and the control + 37 °C group (p > 0.9; Fig. 2F). Additionally, we performed western blot analyses to evaluate HSP70 protein levels (Fig. 2G,H). A significantly increased signal intensity of HSP70 was noted after four days in the CoCl 2 + 37 °C group (231.4 ± 30.8%) in comparison to the control + 37 °C group (100.0 ± 17.8%; p = 0.015). In contrast to the results of qPCR analyses regarding HSP70, the signal intensity of HSP70 was increased in the CoCl 2 + 30 °C group (230.5 ± 42.5%) in comparison to the control group (p = 0.044). Nevertheless, western blot analyses of HSP70 after eight days, were in accordance with those results seen in the qPCR. The addition of CoCl 2 , at 37 °C, led to a strongly increased signal intensity (271.0 ± 60.3%) in comparison to the control + 37 °C group (100.0 ± 19.9%; p = 0.005). Interestingly, hypothermia treatment normalized the signal intensity of HSP70 completely, resulting in no differences between Hypoxic cells were stained with anti-HIF-1α (red, arrowheads). DAPI was used to visualize cell nuclei (blue). (B) Statistical evaluation of HIF-1α cell counts showed, that CoCl 2 led to a strongly elevated number of HIF-1α + cells located in the GCL after four and eight days. Hypothermia inhibited hypoxic processes and lowered hypoxia in most of the cells located in the GCL. (C) In regard to the number of hypoxic cells in the whole retina, CoCl 2 again led to an increased hypoxia, whereas hypothermia alleviated hypoxic processes. (D) mRNA levels of HIF-1α were evaluated via qPCR. Analyses revealed an increased HIF-1α mRNA expression in the CoCl 2 + 37 °C group at both days compared to the control + 37 °C group. At day four, hypothermia led to a control-like HIF-1α expression. (E) mRNA levels of iNOS were analyzed with qPCR. The iNOS mRNA expression was significantly increased by CoCl 2 after four days. This effect was counteracted by hypothermia. At day eight, CoCl 2 had no effect on iNOS expression, whereas both hypothermia treated groups, showed a reduced iNOS mRNA expression in comparison to the control group. (F) qPCR analysis regarding HSP70. CoCl 2 strongly elevated the HSP70 mRNA expression level at days four and eight. At both www.nature.com/scientificreports www.nature.com/scientificreports/ the CoCl 2 + 30 °C (139.3 ± 29.1%) and the control + 37 °C group (p = 0.79; Fig. 2H). In summary, these findings indicate a total counteraction of cellular stress and an early prevention of hypoxic processes in CoCl 2 -stressed retinae after hypothermia treatment.
Neuroprotection via hypothermia. RGCs transfer the electrochemical information via their axons, which build the optic nerve, to the brain. Since RCGs are affected in glaucoma it is important to establish new therapeutic approaches that protect retinal neurons, most of all RGCs.
Since it is known that CoCl 2 leads to a degeneration of rod bipolar cells 11 , we additionally investigated the amount of PKCα + rod bipolar cells after four and eight days (Fig. 5D,E). Regarding the amount of PKCα + rod bipolar cells, the number of bipolar cells was not changed in any of the groups in comparison to the control + 37 °C group (100.0 ± 5.2% PKCα + cells; Fig. 5E) at day four. However, at day eight, a significant loss of bipolar cells was noted in the CoCl 2 + 37 °C group (65.7 ± 5.3% PKCα + cells; p = 0.006) in comparison to control + 37 °C retinae (100.0 ± 8.0% PKCα + cells). Even more, a total rescue of PKCα + cells was achieved by hypothermia (CoCl 2 + 30 °C: 94.1 ± 7.1% PKCα + cells; p = 0.029), resulting in no difference between control + 37 °C and hypothermia treated CoCl 2 + 30 °C retinae (p > 0.9; Fig. 5E).
These findings show, that the damaging effect of CoCl 2 on amacrine cells was not attenuated by hypothermia. While the total bipolar cell population was not affected neither by CoCl 2 nor by hypothermia, the late loss of rod-bipolar cells was totally counteracted after hypothermia.
Apoptotic mechanisms were reduced through hypothermia. To investigate underlying mechanisms that lead to protection of RGCs after hypothermia, we evaluated apoptosis. To this end, we analyzed the expression of several genes that are involved in apoptosis.
Apoptosis can be induced in two different ways 20 . The extrinsic pathway is activated when a ligand binds to a specific "death" receptor. This leads to the activation of caspases, like caspase 8. Therefore, the early extrinsic apoptosis pathway was evaluated via caspase 8 (casp. 8) expression (Fig. 6B). At day four, CoCl 2 + 37 °C stressed retinae (4.9 ± 2.1-fold; p = 0.002) showed a significant higher mRNA expression of caspase 8 than control retinae. This effect was lowered by hypothermia treatment (CoCl 2 + 30 °C: 3.45 ± 1.69-fold; p = 0.314), but an mRNA increased expression was still observable in comparison to the control + 37 °C group (p = 0.084; Fig. 6B). After eight days, the caspase 8 mRNA expression was significantly increased in the CoCl 2 + 37 °C retinae (2.6 ± 1.0-fold expression; p = 0.004). Once again, a reduction of temperature prevented the caspase 8 overexpression in the CoCl 2 + 30 °C group (1.9 ± 0.6-fold; p = 0.229) in comparison to the CoCl 2 + 37 °C group. Protective effects of hypothermia were seen once again, since no alterations were detectable comparing the expression in both CoCl 2 stressed groups (p = 0.244; Fig. 6B).
The other way to induce apoptosis is the intrinsic pathway, in which stress signals lead to a secretion of cytochrome c from the mitochondria into the cytoplasm 20 . This alters the activation state of several pro-or anti-apoptotic proteins, which then leads to apoptosis. The intrinsic apoptosis pathway was analyzed via the Bax/Bcl-2 ratio (Fig. 6C). After four days, no changes were noted within all groups (Fig. 6C). At day eight, no difference was observable between hypothermia treated CoCl 2 + 30 °C retinae (1.14 ± 0.31-fold; p > 0.9) and control + 37 °C ones. In contrast, a significant reduction of the Bax/Bcl-2 ratio, comparing CoCl 2 + 30 °C and CoCl 2 + 37 °C groups (p = 0.048), was observable (Fig. 6C).
As our results show, hypothermia counteracted the cell cycle arrest of retinal cells triggered by CoCl 2 . Both, the expression of caspase 8 and the amount of apoptotic RGCs were strongly reduced after hypothermia treatment, which indicates that hypothermia lowered the apoptotic mechanisms.

Rescue of microglia and macroglia.
To evaluate microglia, CD11b mRNA expression, a gene which encodes for microglia receptors, was analyzed (Fig. 7A). A 3.3-fold reduction was noted in the CoCl 2 + 37 °C group in comparison to the control + 37 °C group (p = 0.098) after four days. In accordance, protective effects of hypothermia were seen, since no alterations were observable between the control + 37 °C and CoCl 2 + 30 °C group. After eight days, CoCl 2 again induced a 5-fold reduction of CD11b mRNA expression in the CoCl 2 + 37 °C group (p = 0.001). Hypothermia completely protected the CD11b mRNA expression in the CoCl 2 + 30 °C group. Hence, no differences were notable compared to the control + 37 °C (p = 0.489; Fig. 7A).
Investigations of macroglial response revealed a significantly reduced GFAP mRNA expression in both CoCl 2 groups at day 4 (37 °C: 0.19 ± 0.20-fold p = 0.001; 30 °C: 0.40 ± 0.30-fold; p = 0.039; Sup. Fig. S2A). At day eight, no differences were observed within all investigated groups (Sup. Fig. S2A). Western blot analyses showed no changes in GFAP signal intensities in any of the groups neither at four nor at eight days (Sup. Fig. S2B,C). The evaluation of GFAP immunoreactivity revealed the same results as seen in qPCR analyzes (Sup. Fig. S2D). Significant lower GFAP signals were seen in the CoCl 2 + 37 °C groups in both points in time (4 days S2E).

Discussion
The aim of this study was to investigate possible neuroprotective effects of hypothermia (30 °C) on CoCl 2 -stressed cultured porcine retina explants. We demonstrated that hypoxic damage, such as oxidative stress, due to CoCl 2 on retinal cells was diminished through hypothermia. Furthermore, hypothermia had inhibiting effects on the apoptosis and led to an enhanced cell survival.
It has been described that cobalt, like hypoxia, triggers the stabilization of the α-subunit of hypoxia-inducible factor (HIF-1) by preventing its degradation 21 . Increased levels of HIF-1α activate the expression of certain genes, like iNOS and heat shock proteins (HSPs) 22,23 . Furthermore, cobalt leads to DNA fragmentation, caspase activation, and to ROS-production through the uncoupling of mitochondrial respiration 24,25 . Toxic effects of cobalt include a loss of mitochondrial membrane potential, the inhibition of the proteasome degradation, resulting in cell death 22,26 . Nevertheless, the treatment of CoCl 2 can simulate a disease process and can cause symptoms very similar to those of hypoxia 22 .
Due to the fact, that CoCl 2 stabilizes HIF-1α, HIF-1α mRNA expression was investigated to verify that hypoxia was successfully induced in retinae stressed with CoCl 2 . In our study, CoCl 2 -stressed hypoxic retinae presented a higher HIF-1α mRNA expression as well as a higher number of hypoxic cells, showing that CoCl 2 successfully induced hypoxia. The stabilization of HIF-1α results in a higher expression of genes that encode several proteins, like heat shock proteins (HSPs) and inducible nitric oxide synthase (iNOS), which are essential to manage hypoxic stress 27 . This effect was successfully seen in CoCl 2 -stressed retinae, in which the mRNA expression of HSP70 was strongly elevated 22 . HSPs are chaperons which are important for the defense against cellular stress. They prevent misfolding and protein aggregations. Especially HSP70 is required for the transcriptional activity as well as for the accumulation and function of HIF-1α 28 . Our results show that hypothermia led to a strong reduction not only of HIF-1α but also of HSP70, pointing out that HIF-1α and HSP70 are strongly linked. The same indirect effect of CoCl 2 on HSP70 was also described in other studies, strengthening our suggestions 29,30 . Nevertheless, our results do not clarify, whether hypothermia first inhibits HIF-1α accumulation which than leads to the reduction of HSP70, or whether hypothermia prevents the HSP70 expression which results in a reduced HIF-1α amount.
Divalent metal ions, such as cobalt, induce a disturbance of the mitochondrial respiration chain and stimulate the rupture of the outer cell membrane, resulting in ROS production and oxidative stress 6,8,22 . Our results suggest that CoCl 2 not only mimics hypoxia through the stabilization of HIF-1α, but also leads to a strongly elevated level of ROS, indicating that it triggers oxidative stress. In accordance with other publications, our results show that hypoxia and oxidative stress trigger apoptotic mechanisms 24,25 , which then led to the significant loss of RGCs in the retinae. These mechanisms were completely counteracted by hypothermia, which probably first blocked the interaction of CoCl 2 and HIF-1α, then HSP70 and iNOS were strongly reduced, leading to decreased oxidative stress and alleviated apoptosis mechanisms.
The prominent loss of RGCs was associated with an increased p21 mRNA expression, a gene which induces cell cycle arrest, and caspase 8, a hallmark for extrinsic apoptosis 20 . An overexpression of HIF-1α can induce apoptotic processes in different ways: the interaction of HIF-1α and p53, a tumor suppressor gene, leads to the activation of apoptotic mechanisms 2 . In addition, HIF-1α activates p21 and lowers the cell viability 31 .
As mentioned before, a HIF-1α overexpression was observed in CoCl 2 -stressed retinae at both points in time. In the early point in time, hypothermia totally stabilized the expression of HIF-1α to control level. This effect was accompanied by a total reduction of cellular stress, namely control-like HSP70 and iNOS expression-as well as ROS-levels, causing strongly lowered extrinsic apoptosis. After eight days, qPCR analyzes regarding HIF-1α plexiform layer; INL = inner nuclear layer; qPCR = quantitative real-time PCR; IHC = immunohistochemistry. Values are mean ± SEM, A-C: n = 6-7/group; E: n = 10/group. Statistical differences to control + 37 °C group are marked with * and differences to CoCl 2 + 37 °C group with #. # ,*p < 0.05; ## ,**p < 0.01; ### ,***p < 0.001. Scale bar = 20 µm. (2019) 9:4898 | https://doi.org/10.1038/s41598-019-41113-4 www.nature.com/scientificreports www.nature.com/scientificreports/ Figure 7. Hypothermia protected microglia. (A) Relative mRNA expression of Cd11b, a gene that is expressed by microglia, was analyzed via qPCR. At day four, a slightly decreased expression of CD11b was noted in the CoCl 2 + 37 °C retinae, which was counteracted via hypothermia. After eight days, a significant reduction of CD11b was observed in the CoCl 2 + 37 °C group. The damaging effect of CoCl 2 was again counteracted by hypothermia in the CoCl 2 + 30 °C group. (B) Representative pictures of microglia stained with anti-Iba1 (red) after four and eight days are shown. Anti-Fc γ -R was used as an activity marker for microglia (green). Fc γ -R + and Iba1 + cells were counted as active microglia. Cell nuclei were visualized with DAPI (blue). (C) The significant loss of microglia due to CoCl 2 was counteracted through hypothermia at both points in time. (D) CoCl 2 led to a significantly reduced number of active microglia. Hypothermia treatment rescued active microglia at four days, but not at eight days. Abbreviations: GCL = ganglion cell layer; IPL = inner plexiform layer, INL = inner nuclear layer; qPCR = quantitative real-time PCR; IHC = immunohistochemistry. Values are mean ± SEM. A: n = 6-7/ group; C, E: n = 10/group. Statistical differences to control + 37 °C group are marked with * and differences to CoCl 2 + 37 °C group with #. # p < 0.05; ### ,***p < 0.001. Scale bar = 20 µm.
www.nature.com/scientificreports www.nature.com/scientificreports/ indicate that the effect of hypothermia was lower than after four days. Anyway, the reduced number of HIF-1α + cells in the retina show that after eight days there was still a positive effect through hypothermia.
In accordance, after eight days, p21 expression was diminished in the hypothermia treated CoCl 2 -stressed retinae and caspase 8 expression was lower than in CoCl 2 retinae. Hence, hypothermia led to an early inhibition of hypoxic processes and reduced the apoptosis. In addition, hypothermia inhibited caspase 3 at the earlier, but not at the later point in time. It is known that hypothermia protects cells through a diminished apoptosis rate by inhibiting the lactate dehydrogenase 32 and by reducing the caspase 3 expression 33 . Moreover, it was shown for retinal explants of mice, that hypothermia treatment during the retinal dissection led to a strongly decreased apoptosis rate in the retina 34 . Caspase 8 is activated at an early stage of extrinsic apoptosis, whereas caspase 3 is cleaved in later stages of intrinsic apoptosis 35 . A possible explanation would be that hypothermia rather inhibits the extrinsic than the intrinsic pathway. Furthermore, other upstream proteins, besides caspase 8, might be the reason for the lacking inhibition of caspase 3 after hypothermia at the later point in time.
Bax is a pro-apoptotic protein, which induces the intrinsic apoptosis by opening mitochondrial pores and supporting the secretion of cytochrome c into the cytoplasm. Bcl-2, on the other hand, is an anti-apoptotic protein that inhibits Bax and prevents the secretion of cytochrome c 36 . In our study, the Bax/Bcl-2 ratio was slightly increased in CoCl 2 -stressed retinae after eight days, but not in hypothermia treated CoCl 2 -stressed ones. Hence, hypothermia seems to inhibit the apoptosis via increasing the Bcl-2 expression or decreasing the Bax expression. Several studies revealed that CoCl 2 -induced hypoxia leads to apoptosis. Kuehn et al. demonstrated that CoCl 2 -treated porcine retinae showed an increased Bax expression level after four days of cultivation 11 . Chang et al. observed that hypoxia activates the intrinsic apoptosis via Bax and caspase 3, whereas Lee et al. reported that CoCl 2 leads to apoptosis by activating both pathways simultaneously 37,38 .
In previous studies a loss of calretinin + amacrine and PKCα + bipolar cells was noted through CoCl 2 6,11 . This is the first study that shows a time dependent damage due to CoCl 2 on calretinin + amacrine cells, parvalbumin (PVALB) expressing displaced amacrine as well as PKCα + bipolar cells.
Amacrine cells are located in the inner nuclear layer (INL) and play an important role for the modulation of signals to the RGCs 39 . It is known that they are vulnerable to glaucomatous damage [40][41][42] . There is a strong connection between amacrine cells and RGCs via gap-junctions, which leads to a secondary loss of amacrine cells after glaucoma-induced RGC loss 42 . In good accordance, we observed a delayed loss of amacrine cells at eight days which was induced by CoCl 2 and possibly strengthened by the early loss of RGCs. Hypothermia did not protect amacrine cells. We assume that the RGC loss due to CoCl 2 led to severe changes in the surrounding tissue, where dendrites of amacrine cells are located and therefore the protection of amacrine cells cannot be achieved.
Bipolar cells are also located in the INL and are connected to rods or cones. We used PKCα to label rod bipolar cells. A time-dependent loss of rod bipolar cells was observed. This was counteracted by hypothermia treatment. This later death is in accordance with affected bipolar cells in different rat glaucoma models 41,43 . Since hypothermia had a rescue effect on bipolar cells, but not on amacrine cells, we assume that the degeneration process of both cell types is different. To confirm this, further studies are necessary.
Besides HSP70, also the expression of inducible nitric oxide synthase (iNOS) depends on the activity of HIF-1α. CoCl 2 has degenerative effects on microglia through increasing the apoptosis rate and diminishing the cell viability via cell arrest 11,44 . Besides neurons, microglia were also protected by hypothermia. Interestingly, the gene encoding for the enzyme iNOS, which is mainly produced by microglia, was increased by hypoxic stress without a microglia response. HIF-1α might be involved in this pathway, since it is known that HIF-1α increases the iNOS expression by binding the transcription promotor 45,46 . Therefore, in our model iNOS regulation seems to be independent from a microglia response.
GFAP expression, is a hallmark for gliosis in retinal diseases or injuries. Macroglial signals in the porcine retina are stronger than in rodent retinae 47,48 . However, we detected a reduced GFAP expression in qPCR and immunohistochemical analyses, while western blot analyses revealed no changes in GFAP signal intensities at any of the investigated points in time. Those results indicate that CoCl 2 seems to have toxic effects not only on microglia but also on macroglia. However, our study and other studies reveal that in porcine degeneration models gliosis is not occurring during cultivation [48][49][50] . The preparation of retinal explants seems to induce a macroglial response itself. Therefore, it is difficult to detect further changes. Furthermore, in our ex vivo model, retinae are cultivated separate from the optic nerve. Thus, the immigration of astrocytes as a result of macrogliosis is not possible.

Conclusion
Hypoxic processes play a crucial role in several retinal diseases. Cobalt chloride (CoCl 2 ) is known to mimic hypoxic processes in vitro by stabilizing the transcription factor HIF-1α and leading to oxidative stress through a disruption of mitochondrial respiration. These effects were observed in the present study. This hypoxic damage due to CoCl 2 was associated with oxidative stress leading to increased apoptosis rates in CoCl 2 -stressed retinae. We demonstrated that hypothermia completely counteracted these mechanisms by probably disturbing the interaction of CoCl 2 and HIF-1α. This led to strongly reduced HSP70 and iNOS synthesis, alleviating oxidative stress and preventing apoptosis. Consequently, most RGCs and bipolar cells were rescued, while amacrine and horizontal cells were not protected.
In conclusion, we demonstrated that our CoCl 2 -induced hypoxia is a suitable model system to test potential therapies and that hypothermia could be a possible additional treatment for retinal diseases.

Methods
preparation of retinal explants. Porcine eyes were obtained from the local abattoir and retinae were prepared within three hours from enucleation. The preparation of retinal explants was performed as described previously 11,49,51 . Briefly, the eyeball was opened to separate anterior parts of the eye from the eye cup. The eye cup was incised four times to produce a cloverleaf-like shape. Next, one retinal explant sample per leaf was punched out www.nature.com/scientificreports www.nature.com/scientificreports/ in the central part of the retinal quadrant using a dermal punch (Ø = 6 mm, Pfm medical AG). Remaining retinal pigment epithelium was removed by washing retinal explants in Neurobasal-A medium (Life Technologies). Finally, retinal samples were placed on a Millicell culture insert (Millipore) with the GCL facing up and cultured in Neurobasal-A medium (Life Technologies) supplemented with 0.8 mM L-glutamine (Life Technologies), 2% B27 (Life Technologies), 1% N2 (Life Technologies) and 2% penicillin/streptomycin (Sigma-Aldrich), for four and eight days. Medium was exchanged completely at days zero, one, two and three. Additionally, half of the medium volume was replaced at days five and seven (Fig. 1A). The substance most commonly used to simulate a hypoxic environment is CoCl 2 8,9,52,53 , hence this was used in the current study. Hypothermia treatment and hypoxia induction via 300 µM CoCl 2 (Sigma-Aldrich) were performed simultaneously and took 48 h in total (Fig. 1A). Control groups were cultivated without the stressor and with or without additional hypothermia treatment, so that four groups were compared: control + 37 °C, control + 30 °C, CoCl 2 + 37 °C and CoCl 2 + 30 °C.

Measurement of reactive oxygen species (Ros)-level and pH-value.
For the measurement of ROS-level in cultured retinae, the non-lytic protocol of ROS-GLO ™ H 2 O 2 assay (Promega) was performed.
Retinae were cultured as described before and the measurement of ROS-level was performed according to the manufacture's protocol. For more detailed description please see supplementary part. The measurement of the pH-values was performed with LAQUAtwin B-712. Calibration was performed according to manufactural instructions. After the medium exchange, 200 µl medium was used for the measurement.
Quantitative real-time pCR (qpCR). The used primer for qPCR analyses are given in Supplementary   Table 1. qPCR analyses were performed as described previously 11,49,51 for n = 6/group at day four and n = 7/ group at days four and eight. All target genes were normalized against housekeeping genes encoding Histone H3 and β-Actin (Sup. Table 1). C t -Values of both housekeeping genes were not affected. The mean of all samples for β-actin was 18,84 ± 0.8 cycles and for histone H3 20.22 ± 0.9 cycles. Samples having C t -values 2 cycles higher or lower than the mean were excluded. Geometric mean of the C t -Values of both genes were calculated and used as reference. For the relative quantification the Δ-ΔC t -algorithm, with efficiency corrected calculation model, based on one sample was used 54 . All groups were compared to control + 37 °C or CoCl 2 + 37 °C groups.

Histological analysis of retinal cells. For immunohistochemical analyses retinal cross-sections
(n = 9-10/group/point in time) and flatmounts of retinae (n = 3/group) were prepared as described previously 11,49,51 . To identify different cell types and proteins specific primary antibodies and matched secondary antibodies were used (Sup. Table 2). For all stainings, 4′,6 diamidino-2-phenylindole (DAPI) was used to visualize the cell nuclei. Cross-sections and flatmounts were blocked with blocking-buffer containing 10-20% donkey or goat serum and 0.1-0.2% TritonX in PBS. Six cross-sections were stained per retina and cells were counted in 24 masked and defined image sections using ImageJ software. Regarding the evaluation of flatmounts, 9 images, including 4 peripheral and 5 central parts of the retina, were counted. For GFAP, the area was measured using an established protocol and an ImageJ macro 40,41 . For more information please see supplementary part.
Western blot. Western blot analyses were performed (n = 4/group/point in time) as described previously 11,49,51 . To this end, the primary antibodies (Sup. Table 2), diluted in the blocking solution, were incubated over night at 4 °C. After the washing steps, the secondary antibodies (Sup. Table 2) were applied for 60 min. Protein bands were recorded at 700 and 800 nm and evaluated with the Odyssey infrared imager system 2.1 (LI-COR Bioscience). HSP70, (70 kDa), β-III-tubulin (55 kDa) and GFAP (55 kDa) signal intensities were normalized to β-actin (42 kDa) signal intensities. statistical analyses. Groups were compared by one-way ANOVA, followed by Dunnett's post-hoc test (Statistica V 12; Statsoft). Results are presented as mean ± SEM. A p-value < 0.05 was considered as statistically significant. The level of significance was set to *p < 0.05, **p < 0.01, ***p < 0.001. Statistical differences compared to the control + 37 °C group are shown with *, differences compared to the CoCl 2 + 37 °C group are shown with # .