Inflammatory signature of cerebellar neurodegeneration during neonatal hyperbilirubinemia in Ugt1 ^-/- mouse model

Mice were housed and handled according to institutional guidelines, and experimental procedures approved by the ICGEB board, with full respect to the EU Directive 2010/63/EU for animal experimentation. Ugt1 ^-/- mice in the FVB/NJ background have been generated previously [17]. Homozygous mutant animals were obtained from heterozygous mating. Male and female pups were used for these studies. WT littermates were used as control. Animals used in this study were at least 99.8% FVB/NJ genetic background, obtained after more than ten backcrosses with wild-type FVB/NJ mice. Mice were kept in a temperature-controlled environment with 12/12-h light/dark cycle. They received a standard chow diet and water ad libitum.

Blood samples were collected at different time points in mutant and WT littermates by decapitation in EDTA-collecting tubes, at the moment of sacrificing the animals, as previously described [17]. Total bilirubin (TB) determination in plasma was performed using Direct and Total Bilirubin Reagent kit (BQ Kits, San Diego, CA) adapting the method to use minimal volumes (10 μl of plasma), as previously described [17].

Tissues for bilirubin content determination were collected and analysed as previously described [27].

Total RNA from mouse cerebellum was prepared using EuroGOLD Trifast (Euroclone, Milano, Italy). One microgram of total RNA was reverse-transcribed, as previously described [18]. Total cDNA (1 μL) was used to perform qPCR using the specific primers listed in Additional file 1. qPCR was performed using the iQ SYBR Green Supermix (Bio-Rad) and a C1000 Thermal Cycler CFX96 Real Time System (Bio-Rad). Expression of the gene of interest was normalized to the Gapdh housekeeping gene. Data were analysed using the ΔΔCt method. All primers are listed in Additional file 1.

Cerebella were dissected, homogenized in lysis solution buffer (150 mM NaCl, 1% NP-40, 0.5% DOC, 0.1% SDS, 50 mM Tris HCl pH 8, 2× protease inhibitors) and analysed by western blot as described previously [16]. Primary antibodies used were as follows: anti-calbindin 1:2000 (Synaptic Systems, Goettingen, Germany), anti-NeuN 1:2000 (Millipore, Temecula, CA), anti-Glial Fibrillary acidic protein (GFAP) 1:1000 (Sigma, St. Louis, MO), anti-Iba1 1:1000 (Wako, Neuss, Germany), anti-cleaved caspase3 1:500 (Cell Signalling), anti-caspase3 1:800 (Santa Cruz Biotechnology), anti-CHOP 1:1000 (Santa Cruz, sc-56107), anti-HO1 1:1000 (Enzo Life Sciences, ADI-OSA-110) and anti-LC3 1:1000 (Sigma, L7543). Antibody anti-tubulin mAb E7 (Developmental Studies Hybridoma Bank, Iowa City, IA) or anti-Actin (Sigma, A2066) was used for loading control determination. Representative images are shown in the figures. Each timepoint was run in a separate gel.

Nissl staining and immunofluorescence analysis of cerebellum samples were performed as previously described [17, 18]. The study was performed in a double-blind fashion: the genotype of the animals was unknown to the operator, while a different investigator analysed the data. Measurements were averaged for each animal.

Briefly, after blocking, specimens were incubated with the primary antibody for 2 h at RT in blocking solution with anti-calbindin 1:400 (Synaptic Systems, Goettingen, Germany), anti-NeuN 1:400 (Millipore, Temecula, CA), anti-glial fibrillary acidic protein, GFAP 1:200 (Sigma, St. Louis, MO), anti-iba1 1:200 (Wako, Neuss, Germany), anti-TNFα 1:100 (sc-1350, Santa Cruz), anti-NFKβ 1:100 (sc-8008, Santa Cruz), anti-MRC1 1:100 (CD206 clone MR5D3, MCA2235GA, BioRad), anti-CD68 1:100 (clone FA-11, BioRad), anti-CHOP 1:100 (GADD153, sc-56107, Santa Cruz), anti-HO1 1:200 (Enzo Life Sciences, ADI-OSA-110), anti-Fas 1:100 (CD95, NB120-13550, Novus Biologicals) and anti-HO1 1:200 (ADI-OSA-110, Enzo Life Sciences). Antigen retrieval with sodium citrate pH6 was performed for CHOP and HO1 immuno-stainings prior to blocking solution step. After 3× 5-min washes with blocking solution, specimens were incubated with secondary antibody (Alexa Fluor 488 or 568; Invitrogen Carlsbad, CA) for 2 h at RT. Nuclei were visualized by addition of Hoechst (10 μg/ml, Invitrogen) for 5 min after secondary antibody solution. Nissl-stained sections were mounted in Eukitt (Fluka, St. Louis, MO, USA) while immunofluorescence (IF) slides were mounted in Mowiol 4-88 (Sigma-Aldrich). Images were acquired on a Nikon Eclipse E-800 epi-fluorescent microscope with a charge-coupled device camera (DMX 1200 F; Nikon Amstelveen, The Netherlands). Digital images were collected using ACT-1 (Nikon) software.

Analysis of the layer thickness was performed on Nissl-stained sections by measuring the layer depth (μm) as previously described [16]. PC density analyses were performed as previously described [16, 17].

The Prism package (GraphPad Software, La Jolla, CA) was used to analyse the data. Results are expressed as mean ± S.D. Values of p < 0.05 were considered statistically significant. Depending on the experiment, Student’s t test or two-way ANOVA, with Bonferroni’s post hoc comparison tests, were used, as indicated in the legends to the figures and text. Correlation analyses were done using the Pearson coefficient to assess the linearity between two variables and calculate two-tailed p value (95% of confidence interval).

We have previously reported that bilirubin induces apoptosis in the cerebellum of C57Bl/6 Ugt1 ^-/- mice at P4 [18], a strain carrying the same genetic mutation but showing a more severe phenotype [16]. Here, we used a less severe mouse strain (Ugt1^-/-/FVB/NJ) to study the apoptotic pathway in the cerebellum, which is the most vulnerable region to bilirubin toxicity in rodent animal models (Fig. 1a) [17, 28, 29]. No increase in caspase 3 cleavage was observed at P5 (Fig. 1b), but apoptosis occurred at P8, as demonstrated by a twofold increase of activated caspase 3 (Fig. 1b). These results are in line with our previous data showing that, despite some of the mice were rescued from mortality by PT starting at P8, the accumulated damage was not reversible [17].

Neurodegeneration is a well-known feature of bilirubin-induced neurological dysfunction (BIND). To have a deeper insight in the mechanisms leading to cerebellar hypoplasia, we first estimated by western blot (WB) analysis the levels of calbindin, a Purkinje cell (PC)-specific marker [30]. We observed that PCs were not affected at P5, while at P8 and P10, there was a significant reduction in calbindin levels in Ugt1 ^-/- cerebellar protein extracts (56 and 61% decrease, at P8 and P10, respectively; Fig. 1c). Similarly, IF analysis showed a reduction in PC number of 30 and 50% at P8 and P10, respectively, and the almost complete disappearance of their dendritic arborization (Fig. 1d).

Likewise, we assessed the degree of cerebellar damage in brain sections of Ugt1 ^-/- mice over the phases preceding death (Fig. 1e and Additional file 2A). The long-term exposure of cerebellar neurons to high levels of bilirubin resulted in an overall reduction in cerebellar cortex thickness of 20 and 32%, at P8 and P10, respectively (Fig. 1e). Layer-thickness quantification showed that the external germinal layer (EGL)- and molecular layer (ML)-thickness were significantly reduced at both P8 and P10, while the internal granular layer (IGL)-thickness was reduced only at P10 (Additional file 2A).

Next, we verified whether bilirubin affected mature cerebellar granule cells (GCs) by determining NeuN expression, a nuclear protein antigen whose expression in the cerebellum is restricted to differentiated granule neurons [31]. Western blot experiments showed that bilirubin had no evident effects on differentiated GCs in all time points analysed (Fig. 1f and Additional file 2B). IF analysis performed at P8 and P10 showed no obvious reduction in cell density of neurons located in the IGL (Fig. 1g and Additional file 2C, respectively), suggesting that differentiated GCs were not affected by chronic exposure to bilirubin.

Collectively, our data support the concept that vulnerability of different neuronal cell types in vivo is associated with their developmental stage.

Then, we determined the levels of plasma and cerebellar bilirubin at different time points, ranging from P2 to P10 (Fig. 2). Mutant animals showed a time-dependent increase in plasma bilirubin levels, associated to the worsening of the phenotype. Importantly, there was a clear correlation between the levels of UCB in the cerebellum and those found in plasma (p < 0.001; Fig. 2f). These results demonstrate the tight association between the accumulated levels of UCB in the cerebellum and the increase in neurodegeneration and apoptosis.

Since neurotoxicity is normally associated with neuroinflammation [32, 33], we investigated whether bilirubin-induced cell death was accompanied by an increase of inflammatory markers in the cerebellum of the Ugt1 ^-/- mice.

First, we analysed astrocyte response to bilirubin, by determining glial fibrillary acidic protein (GFAP) levels, an astrocyte-specific marker. No astrocytosis was detected in the early time point of bilirubin exposure (P5, Fig. 3a). Conversely, GFAP levels were two and threefold increased at P8 and P10, respectively, as determined by WB analysis (P8 and P10, Fig. 3a).

IF analysis performed at P8 and P10 revealed that the glial scar was selectively localized in those cerebellar fissures where PC loss was more prominent (Fig. 3b), such as fissures IV to VIII (Additional file 3A, section of whole cerebellum at P10). To note, the time-dependent increase in GFAP protein levels was also evident at mRNA level (Additional file 3B). These finding were in good correlation with the time-dependent loss of PCs (Fig. 1d).

Next, we determined the cerebellar microglia response to bilirubin by the analysis of ionized calcium-binding adapter molecule 1 (Iba1) as cell-specific marker. We observed that bilirubin triggered an increase in Iba1 mRNA levels starting from P8, and reaching up to threefold increase at P10, as determined by real-time RT-PCR (Additional file 4A). However, at protein levels, we observed a twofold increase in Iba1 only at P10 (Fig. 3c). Furthermore, IF staining of cerebellar sections showed that Iba1-positive cells were located in those zones more affected by bilirubin (Fig. 3d and Additional file 4B).

Interestingly, among all the inflammation markers tested (Fig. 4a and Additional file 5), only TNFα was differentially expressed in mutant animals compared to WT littermates, reaching up to ~10-fold (two-way ANOVA, p < 0.05) in very severe conditions, such as P10 (Fig. 4a). Immunofluorescence analysis of cerebellar sections at P10 showed the presence of TNFα- and NFKβ-positive cells in cerebellar samples from mutant mice, which co-stained with microglia-positive cells (Fig. 4b–d).

These data underline the complexity of the inflammatory response triggered by bilirubin in vivo and indicate that neuroinflammation is an important player contributing to each stage of the damage in our experimental model. Overall, these data show that the inflammatory response in vivo increases over time as bilirubin accumulates in the cerebellum and parallel the onset of neurodegeneration.

To better characterize the microglia activation during prolonged bilirubin exposure, we performed real-time RT-PCR analysis of the molecular markers that discriminate the activation of microglia/macrophage states, such as those characterizing the pro-inflammatory M1 microglia (CD68 and CD86) and the anti-inflammatory M2 microglia (MRC1, Arg1 and MRC2).

We observed that the messenger RNA (mRNA) of the M1 pro-inflammatory microglia marker CD68 showed 2.0-, 2.5- and 4.0-fold increase at P5, P8 and P10, respectively, compared to WT samples at the same time points (Fig. 5a). Unexpectedly, when the mRNA expression of CD86, another component of M1 type microglia, was compared between Ugt1 ^-/- mice and WT littermates, no upregulation was observed.

By contrast, real-time RT-PCR analysis revealed an inverted trend for the M2 type, which is considered to compose the neuroprotective counterpart of microglia. MRC1 mRNA in Ugt1 ^-/- mice was upregulated at P5 and showed decreased expression over time (the same pattern was observed for Arg1, but differences were not statistically significant), while MRC2 was significantly activated only at P8, and mRNA levels decreased at P10, both in WT and Ugt1^-/- cerebella (Fig. 5a).

IF experiments reinforced the findings obtained by real-time PCR analysis showing MRC1-positive cells at P5 (Fig. 5b), and of CD68 at P10 (Fig. 5c). In particular, a portion of MRC1-positive cells co-stained with Iba1 at P5, confirming the activation of M2 microglia at the early stages of the disease. Likewise, we reported an important increase in CD68-positive cells at P10, providing evidence of pro-inflammatory M1 type of microglia activation at the later stages of the disease (Fig. 5c).

Since the transition from M2 to M1 phenotype is generally associated with inflammation-induced pathologies, taken together these data indicate that exposure to bilirubin triggers the activation of pro-inflammatory microglia with a parallel decrease in the neuroprotective one.

It has been shown that bilirubin is involved in endoplasmic reticulum (ER) stress response activation in in vitro cell cultures [34‐36]. Thus, we investigated whether ER stress is a key event in the onset of bilirubin toxicity in vivo.

Concomitantly to the activation of inflammation, we observed that the bilirubin insult also triggered the activation of different ER stress markers during the early phase of hyperbilirubinemia, as determined by real-time RT-PCR analysis (Fig. 6a and Additional file 6A). mRNA levels of ATF3, CD95/Fas and CHOP increased as aggravating stages of bilirubin toxicity moved forward. Since CHOP is a key effector in ER stress response, we performed IF analysis of cerebellar sections at P5 and at P10. IF analysis showed that CHOP signal was increased in mutant mice compared to WT littermates in all cerebellar layers at P5 (external germinal layer, Purkinje cell layer and internal granular layer, Fig. 6b).

As neurodegeneration moved forward, we observed that Fas/CD95 and CHOP signal concentrated in the PCL (Fig. 6c, d). In particular, at P10, CHOP was mainly localized in the nucleus of severely neurodegenerated PCs, but not in granule cell neurons (Fig. 6d and Additional file 6D). Moreover, we observed that CHOP-positive Purkinje cells were surrounded by Iba1-positive cells (Fig. 6d, e and Additional file 6E) indicating that microglia cells were actively engulfing degenerated neurons.

Finally, we focused our attention on oxidative stress, which is considered a major process triggered by bilirubin in vivo [18]. By real-time RT-PCR, different markers of oxidative stress were analysed to assess their activation status (Fig. 7a and Additional file 7A-B). We could not detect any variation in mRNA levels of the analysed cytochrome P450 family 1 subfamily A member 2 (Cyp1a1), and cytochrome P450 family 2 subfamily A member 5 (Cyp2a5) (Additional file 7B), nor in Nrf2, an important transcription factor involved in cellular response to oxidative stress (Fig. 7a). On the contrary, both the mRNA and protein expression levels of HO1 (heme oxygenase 1), a key enzyme in heme catabolism, were increased in RNA and protein extracts from Ugt1 ^-/- cerebella, in a time-dependent manner, starting from P8 and reaching more than fourfold increase at P10 at protein level (Fig. 7b). IF analysis showed an important increase of HO1-positive cells in the Purkinje cell area (P5 and P10, Additional file 7C and Fig. 7c, respectively). All HO1-positive cells were also Iba1 positive, but not Calbindin or NeuN positive, suggesting that some microglia cells experienced oxidative stress as a consequence of their phagocytic activity on degenerated PCs.

Because of the close connection between oxidative stress, neuro-apoptosis, inflammation and autophagy, we investigated whether the latter was activated in vivo in response to bilirubin.

To determine whether autophagy is triggered by hyperbilirubinemia, we performed a WB analysis and quantified the levels of the LC3-II form, as a marker of autophagy. We observed the increase the LC3-II form only in cerebellar total extracts of Ugt1 ^-/- animals at P10 (Fig. 8), corresponding to the most severe time point analysed.

These data showed, for the first time, that bilirubin triggers autophagy in vivo in a mouse model of severe neonatal hyperbilirubinemia.

Although mildly elevated bilirubin concentrations are considered beneficial due to its antioxidant properties, high levels of bilirubin lead to considerable oxidative stress [37, 38]. This is particularly evident in the central nervous system when ROS production overwhelms antioxidant defences [39], resulting in neurodegeneration.

Typically, the redox-sensitive pathway is activated at the onset of brain injury as a protection to overcome oxidative stress. This was well demonstrated in tissue-culture studies showing that UCB-induced ROS excess impairs glutathione homeostasis [35, 40] and mitochondrial metabolism resulting in a cellular energetic crisis, swelling, cytochrome c release [41] and calcium perturbation [42]. All these events activate the mitochondrial apoptotic pathway [43] leading to increased oxidative stress and cellular redox status imbalance [44, 45]. Likewise, others and we showed that oxidative stress is a key feature of neurotoxicity animal models of systemic hyperbilirubinemia [18, 26, 46]. However, oxidative stress response to bilirubin did not elicit the induction of Cyps in our mouse model (Additional file 7B), differently to what observed in the Gunn rat [47], an animal model of the disease presenting a milder phenotype.

Interestingly, in our mouse model, the increase in ROS production resulted from the presence of activated microglia cells in the region where Purkinje cells degenerate. These cells, which are positive for HO-1, are involved in the removal of death cells and are associated with inflammation. We observed that HO-1 expression levels increment over time in the damaged tissue, with HO1-positive microglia cells found as early as P5. This observation is in agreement Silva and colleagues showing that an excess of activated microglia produces high levels of nitric oxide, which in turn mediates microglia-induced neuronal cell death [48].

The endoplasmic reticulum (ER) is very sensitive to perturbations of energy levels, intracellular calcium and redox state. Important modifications of its balance result in a condition named ER stress, characterized by an excess of unfolded proteins, and the induction of genes involved in cell survival and protein degradation, aiming to restore the normal state. However, when ER stress is sustained for long periods of time, cells are unable to overcome the accumulation of unfolded proteins and activate the apoptotic pathway.

We report that mRNA levels of key ER stress markers, such as ATF3 and Fas, were significantly induced in response to high bilirubin levels. However, no changes in transcriptional levels were observed in other genes related to unfolded protein response (UPR) and ER stress (GPR78, GRP94, P58 ^IPK , FKBP11 and DR5) at any time point analysed (Additional file 6A). At the early stages of the disease, CHOP signal was increased in cerebellar sections of mutant mice, with more intense signal in the Purkinje cell layer. At later stages, CHOP signal was localized in degenerating Purkinje cells, surrounded by microglia Iba1-positive cells, suggesting their involvement in removing cellular debris from death Purkinje cells. These results are in agreement with previous findings performed in tissue culture cell lines [20, 35, 36], showing that CHOP and ATF3 are key mediators in ER-stress bilirubin-induced response.

Interestingly, the increased expression of ATF3, Fas/CD95 and CHOP genes indicates that cells sense the overwhelming ER stress and activate the PERK-mediated pathway [49, 50]. Since bilirubin levels continuously increase over time [17], cells are not able to revert ER stress and activate the apoptotic pathway, a hypothesis supported by the increased levels of cleaved caspase 3 and TUNEL-positive cells observed in later stages of the disease [17].

Altogether, these observations provide evidence that ER stress is involved in the onset of bilirubin-induced neuro-apoptosis of Purkinje cells in vivo.

The neurodegenerative phenotype in the cerebellum of the mice used in this study is progressively enhanced after sustained high levels of bilirubin and correlates well with the reduced cerebellar development (Fig. 1e). Neurodegeneration is associated with reduced arborization of PC and progressive cell loss (Fig. 1d). In contrast, mature granule cells were not affected by bilirubin toxicity (Fig. 1f). These results are in line with previous observations performed by our lab in the Ugt1 ^-/- mouse model, in both C57Bl/6 and FVB/NJ genetic backgrounds [16‐18]. In particular, we have previously demonstrated that PCs showed a neurodegenerative phenotype as determined by FluoroJade positivity, TUNEL assay and Sholl analysis [17, 18]. These data strengthen the concept that immature neurons are more sensitive to insults compared to more differentiated ones [51] and that bilirubin affects survival of neuronal precursors by impairing their neurogenesis, arborization and synaptic connectivity [17, 52].

In the nervous system, degenerating cells can stimulate inflammation. Although most neurodegenerative diseases involve inflammation, also ER stress is a critical mechanism that can regulate inflammation and is linked through several intersecting cellular pathways, particularly those mediated by ROS [50].

We showed that NFKβ, a well-known inflammatory mediator, responds to high bilirubin levels by the upregulation of its mRNA in a whole tissue extract and by its increase in microglia cells which migrate to the site of tissue injury, together with the increase of TNFα in microglia and astrocytes. To this respect, NFKβ plays a dual role because it exerts both neurotoxic and neuroprotective functions depending on the kinetics and the expression in the injured tissue. In fact, NFKβ regulates the defence response to stress stimuli (inflammation and oxidative stress) and cellular fate (apoptosis, proliferation and differentiation). However, prolonged inflammation leads to the activation of genes, mediated by NFKβ that boosts oxidative stress, inflammation and apoptosis [50]. This is in line with our results showing the exacerbation of neurotoxicity during the more severe conditions observed at P10 (increase in HO1, caspase 3, astrocytosis and gliosis). Similarly, NFKβ is activated in response to pathological condition like trauma, ischemia or different types of neurodegenerative disease [39].

Another key inflammatory factor is TNFα, activated in the more severe phases of the pathology in the Ugt1 ^-/- pups. This factor is released by activated microglia and astrocytes in the CNS and mediates the amplification of neuroinflammation events and the initiation of the apoptotic cascade. As in the cases of NFKβ, TNFα expression is increased in many neurodegenerative conditions [53].

Our findings support previous data by the laboratory of Brites and colleagues, showing that exposure of astrocytes and microglia primary cultures to bilirubin triggers the release of the pro-inflammatory cytokines TNFα and NFKβ [48, 54, 55]. However, in our model, increased transcriptional levels of TNFα and NFKβ occur after elevated and prolonged levels of tissue bilirubin, while in in vitro cultures these events occur within hours upon bilirubin induction. In addition, other well-known pro-inflammatory mediators are unchanged, such as IL1β, IL6 and INFγ [56]. The discrepancy between in vitro and in vivo data might be related to intensity of the insult, interplay between different cell types and developmental status of the brain, conditions not fully recapitulated by in vitro primary cultures.

Activated microglia and astrocytes are the main effectors of the innate immune response in the CNS and are strongly activated in response to neurodegeneration, releasing neuroinflammatory factors. We showed that hyperbilirubinemia triggers the activation of both astrocytes and microglia in the cerebellum of Ugt1 ^-/- mice.

The activation of both astrocytes and microglia by bilirubin has been reported both in vitro [57] and in vivo. In particular, in vivo studies of Gunn rat cerebellum indicated that the increase in GFAP content is a common trait of hyperbilirubinemia [58]. Moreover, the involvement of glia and the role of TLR2 were shown in the humanized knockout mouse model of hyperbilirubinemia [26, 59], although the time course of the disease onset was not investigated. The block of inflammatory activation resulted in death of hyperbilirubinemic TLR2 KO mice, underlined the beneficial effects of inflammatory response in the context of bilirubin toxicity [26]. However, despite the important inflammatory response observed in our Ugt1 ^-/- animals, we could not detect TLR2 activation in response to bilirubin (Additional file 5D). In fact, because of the early neonatal lethality of our mouse model, we investigated the developmental stages within P5 and P10, while TLR2 activation was reported at later stages, namely P14 [26]. We speculate that the difference between the two models could be related to the extent and duration of the bilirubin insult, intrinsic differences between the two models, or a combination of both.

In addition, we investigated the interplay between inflammatory M1 and anti-inflammatory M2 microglia in the context of bilirubin toxicity. While M2 microglia was induced in at the early stages of the disease, decreasing over time as the severity of the phenotype increases, the trend of M1 microglia activation was the opposite, being progressively activated at later stages, as determined by Iba1 and CD68 M1 microglia markers. A similar trend of activation/deactivation of M1 and M2 microglia was also observed in other neurodegenerative pathologies such as spinal cord injury [60] and traumatic brain injury [61, 62].

Increasing evidences showed that autophagy is a crucial defence mechanism against neurodegenerative diseases and its deregulation is a key event that decides cell fate [63, 64].

Our results showed that autophagy is activated in response to bilirubin in the cerebellum of the Ugt1 ^-/- mice at P10 (Fig. 8). These observations are in line with experiments performed in neuroblastoma and endothelial cell lines exposed to high bilirubin levels [35, 65]. In both cases, the autophagy pathway was activated at very high levels of bilirubin (Bf = 100–140 nM) and at later stages (24–72 h) when cell death was evident.

Interestingly, autophagy is linked to ER stress through Fas/CD95, which is strongly induced in the cerebellum of Ugt1 ^-/- mice upon bilirubin exposure (Fig. 6). In fact, Fas/CD95 triggers cell death by the induction of caspase-mediated apoptosis but also stimulates autophagy through MAPK pathway in HeLa cells [66]. Moreover, CHOP is required to increase transcriptional levels of genes responsible for the correct formation and function of the autophagosome [67].

Taken together, these results indicate that autophagy is the latest pro-survival mechanism activated by cells to overcome bilirubin cellular stress prior to death.