Our study demonstrates that high oxygen conditions lead to a modification of markers for autophagic cell death in the developing brain. Moreover, we showed that a single rhEpo-treatment neutralizes these modifications. In previous studies, our group demonstrated a marked reduction of inflammatory mediators, oxidative stress, apoptotic cell death and pro-apoptotic factors in immature rodent brains exposed to hyperoxia after rhEpo treatment [31,37,38].

As demonstrated by several groups, oxidative stress culminating in ROS generation is believed to be essential to trigger autophagy by various mechanisms [19,39]. Once autophagy is activated the formation of phagophores (precursors of autophagosomes) is initiated by Beclin-1 [40]. Here, we show that hyperoxia modulates Beclin-1 expression in the immature brain resulting in an up-regulation after 12 h of hyperoxia. In line with this finding, previous reports revealed that oxidative stress-induced autophagy coincides with an increase in the level of Beclin-1 in cortical neurons, glioma U251 and neuronal PC12 cells [41–43]. The elongation of phagophores requires two ubiquitin-like conjugation systems; the Atg12-Atg5-Atg16L and the phosphatidylethanolamine-LC3 system [4,18]. Some Atg genes, including LC3, were found to be up-regulated during oxidative stress-induced autophagy induction, suggesting a possible function for ROS in regulation of Atg protein expression [44]. Noteworthy, in our model of oxygen toxicity we demonstrated an increased expression of Atg3, Atg5, Atg12, LC3A-II and LC3B-II after 12 h of oxygen exposure suggesting that hyperoxia induced oxidative stress in the immature brain might trigger these acute changes in autophagy activity.

Studies using mutant mice with targeted deletion of Atg5 or Atg7 specifically in the brain suggested that autophagy is constitutively active and essential for neuronal survival [13–15]. However, these studies analyzed the function of autophagy activity proteins in the healthy adult brain and used a constitutive knock out model neglecting the dynamic changes of autophagic processes and their relevance in different pathological conditions. In contrast, we were interested in the regulation of autophagy activity in the immature brain and applied different levels of oxidative stress by varying the duration of hyperoxia. This might have led to an altered tissue environment with modified inflammatory mediators [38]. These different experimental setups might explain the discrepancy between hyperoxia induced brain damage and enhanced autophagy activity. Interestingly the autophagic response seems to depend on the duration of oxygen exposure, since we observed a marked down-regulation of all autophagy related proteins analyzed after 24 h.

In previous studies we demonstrated an increase in apoptotic cell death present after 6 h of hyperoxia and evident up to 72 h [29,30]. The dynamic changes of apoptotic processes combined with the regulation of autophagic activity shown here support the hypothesis of an interaction between these two mechanisms. This indicates an induction of autophagy by hyperoxia-triggered acute ROS formation with enhanced activation of apoptosis as previously shown [28]. However, with sustained oxygen exposure, triggering both extrinsic and intrinsic apoptotic cascades [29,32], autophagy might be inhibited by apoptosis [28]. In this regard it has been shown that caspase-2 deficiency leads to increased autophagy and prolonged neuronal survival whereas mitochondrial oxidative stress induced apoptosis is inhibited [45]. Of note, we have recently demonstrated that the caspase-2 initiated intrinsic apoptotic pathway plays a major role in hyperoxia-induced cell death and might be a potential explanation for the observed increased apoptosis [32] as well as reduced autophagy after longer durations of hyperoxia. This might be highly relevant for the resulting damage to the developing brain and long term functional outcome [46], since the injury cascade is in full progress within this critical time window and autophagy is necessary to degrade and recycle damaged cellular material [6,19,20].

Interestingly rhEpo reverses these hyperoxia mediated effects. However, whether these rhEpo triggered modulations are primary or secondary effects remains to be clarified. One possible mechanism would be amelioration of ROS-formation by rhEpo [36,37] indirectly inhibiting oxygen induced autophagy within the early stage (12 h). When apoptosis is predominant at 24 h it might induce anti-apoptotic signaling thereby enabling autophagy. Another explanation could be a direct modulating process on autophagy by stabilization of the outer mitochondrial membrane mediated by Bcl-2 stabilization. This would lead to impaired permeabilization of the outer mitochondrial membrane or directly via an interaction of Bcl-2 with Beclin-1 [36,47].

Although we indicate that autophagy is involved in hyperoxia mediated brain damage, we cannot exclude that our results are partly associated with apoptosis, since it has been reported that e.g. Beclin-1 may have dual roles, in apoptosis and autophagy [41,48]. Knockout mice deficient in Beclin-1 show both, neuro-degeneration and lysosomal abnormalities [49]. Since the autophagic process is dependent on the interaction of Beclin-1 and the anti-apoptotic protein B-cell lymphoma 2 (Bcl-2), the Beclin-1-Bcl-2 complex functions as a regulator of autophagy depending on the tissue environment and the pathological conditions [50]. Accordingly, we recently demonstrated that hyperoxia triggers a decrease in Bcl-2 expression as a part of the intrinsic apoptotic pathway in the developing brain [32]. This finding might explain the observed modulation of Beclin-1 expression counteracted by rhEpo, giving similar results in a model of traumatic brain injury [36]. Furthermore, it has been shown that Beclin-1 is a substrate of caspase-3, which is a key player in the intrinsic apoptotic pathway. The cleavage products of Beclin-1 reduce autophagy and promoted apoptosis [51].

In order to distinguish to what extent apoptosis and autophagy contribute to hyperoxia mediated cell death further studies are required. Nevertheless, our data suggest that the simultaneous regulation of a broad range of autophagy activity proteins after oxygen exposure play an important role in the pathology of neonatal brain damage.

Twelve or 24 h after hyperoxia initiation on P6, Wistar rat pups (n = 8 per group) were sacrificed with chloral hydrate (1 mg/kg, IP.) and were transcardially perfused with normal saline solution. After decapitation the olfactory bulb and cerebellum were removed, brain hemispheres were snap-frozen in liquid nitrogen and stored at −80 °C until further analysis (quantitative real-time polymerase chain reaction (PCR) and Western blotting).

Total cellular RNA was isolated from snap-frozen tissue by acidic phenol/chloroform extraction and DNase I treated (Roche Diagnostics, Mannheim, Germany); 2 μg of RNA was reverse transcribed with Moloney murine leukemia virus reverse transcriptase (Promega, Madison, WI, USA) in 25 μL of reaction mixture. The resulting cDNA (1 μL) was amplified by real-time PCR. The PCR product of Atg3, Atg5, Atg12 and Beclin-1 was quantified in real-time, using a dye-labeled fluorogenic reporter oligonucleotide probe and primers (Table 1). All probes were labeled at their 5′ ends with the reporter dye 6-carboxy-fluoresceine (FAM), at their 3′ ends with the quencher dye 6-carboxytetramethylrhodamine (TAMRA) and were purchased from BioTez (Berlin, Germany). Hypoxanthineguanine phosphoribosyl-transferase (HPRT) was used as internal standard. Real-time PCR and detection were performed in triplicate and repeated four times for each sample using a total reactive volume of 13 μL which contained 6.5 μL of 2× TaqMan Universal PCR Master Mix (Applied Biosystems, Foster City, CA, USA), 2.5 μL of 1.25 μM oligonucleotide mix and 0.5 μL (0.5 μM) of probe. The PCR amplification was performed in 96-well optical reaction plates for 40 cycles with each cycle at 94 °C for 15 s and 60 °C for 1 min. Each plate included at least three “no template controls”. The expressions of Atg3, Atg5, Atg12, Beclin-1 and HPRT were analyzed with the real-time PCR ABI Prism 7500 sequence detection system (Applied Biosystems) according to the 2^−ΔΔC_T method [52].

Snap-frozen tissue was homogenized in RIPA (radioimmunoprecipitation assay) buffer (1% NP40, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM EDTA, 1 mM EGTA, 1 mM Na₃VO₄, 20 mM NaF, 0.5 mM DTT, 1 mM PMSF and protease inhibitor cocktail in PBS pH 7.4). The homogenate was centrifuged at 1050 g (4 °C) for 10 min, and the microsomal fraction was subsequently centrifuged at 17,000 g (4 °C) for 20 min. After collecting the supernatant, protein concentrations were determined using the bicinchoninic acid kit (Interchim, Montluçon, France). Protein extracts (25 μg per sample) were denaturated in Laemmli sample loading buffer at 95 °C, separated by 10% or 15% sodium dodecyl sulfate polyacrylamide gel electrophoresis and electro-transferred in transfer buffer to a nitrocellulose membrane (0.2 μm pore, Protran; Schleicher & Schüll, Dassel, Germany). Nonspecific protein binding was prevented by treating the membrane with 5% non-fat dry milk in Tris-buffered saline/0.1% Tween 20 for 2 h at room temperature. Equal loading and transfer of proteins was confirmed by staining the membranes with Ponceau S solution (Fluka, Buchs, Switzerland). The membranes were incubated overnight at 4 °C with rabbit monoclonal anti-Atg5 (55 kDa; 1:1.000; Cell Signaling Technology, Danvers, MA, USA), rabbit monoclonal anti-Atg12 (55 kDa; 1:1.000; Cell Signaling Technology), rabbit monoclonal anti-Beclin-1 (60 kDa; 1:1.000; Cell Signaling Technology), rabbit monoclonal anti-LC3A (14 kDa; 1:500; Cell Signaling Technology) or rabbit monoclonal anti-LC3B (14 kDa; 1:500; Cell Signaling Technology), respectively. Unfortunately we did not get any specific Atg3 signal for three commercial Atg3 antibodies tested. Secondary incubations were performed with horseradish peroxidase-linked anti-rabbit (1:2.000; Amersham Biosciences, Bucks, United Kingdom) antibody. Positive signals were visualized using enhanced chemiluminescence (ECL; Amersham Biosciences) and quantified using a ChemiDoc^™ XRS+ system and the software Quantity One^® (Bio-Rad, Munich, Germany). Membranes were stripped, then washed, blocked and re-probed overnight at 4 °C with mouse anti-β-actin monoclonal antibody (42 kDa; 1:10.000; Sigma-Aldrich, Taufkirchen, Germany).

Data were analyzed using IBM SPSS Statistics (IBM, Frankfurt, Germany) and presented as mean + standard error (SEM). Group effects were assessed by analysis of variance (ANOVA) followed by post-hoc independent sample t-test multiple comparison. p-values are presented after Bonferroni correction. Adjusted p-values of <0.05 were considered statistically significant.