Histone acetylation as a new mechanism for bilirubin-induced encephalopathy in the Gunn rat

Bilirubin neurotoxicity has been studied for decades and has been shown to affect various mechanisms via significant modulation of gene expression. This suggests that vital regulatory mechanisms of gene expression, such as epigenetic mechanisms, could play a role in bilirubin neurotoxicity. Histone acetylation has recently received attention in the CNS due to its role in gene modulation for numerous biological processes, such as synaptic plasticity, learning, memory, development and differentiation. Aberrant epigenetic regulation of gene expression in psychiatric and neurodegenerative disorders has also been described. In this work, we followed the levels of histone 3 lysine 14 acetylation (H3K14Ac) in the cerebellum (Cll) of the developing (2, 9, 17 days after the birth) and adult Gunn rat, the natural model for neonatal hyperbilirubinemia and kernicterus. We observed an age-specific alteration of the H3K14Ac in the hyperbilirubinemic animals. The GeneOntology analysis of the H3K14Ac linked chromatin revealed that almost 45% of H3K14Ac ChiP-Seq TSS-promoter genes were involved in CNS development including maturation and differentiation, morphogenesis, dendritogenesis, and migration. These data suggest that the hallmark Cll hypoplasia in the Gunn rat occurs also via epigenetically controlled mechanisms during the maturation of this brain structure, unraveling a novel aspect of the bilirubin-induced neurotoxicity.

. Total Serum Bilirubin (TSB), calculated free bilirubin (cBf) in the blood, cerebellar weight, and Western blot analysis of the level of histone 3 acetylation (H3K14Ac) P: post-natal age in days, Adult: more than 1-year-old. Black bars jj rats, White bars: ctrls. (A) TSB (µM); (B) cBf (nM), (C) Cll weight (mg/animal). Results are expressed as mean ± S.D. of 6-15 animals each group and post-natal age. One way ANOVA followed by Tukey post-test: ***p < 0.001. (D) H3K14Ac levels in the Cll of jj animals vs. ctrl. Results are as mean ± S.D. of 3-6 animals each group and post-natal age. Unpaired t-test with Welch correction, *p < 0.05 vs. age matched ctrl. described 14 . Differently from TSB, the calculated Bf (cBf, Fig. 1B) level in jj pups dropped during development (P2 Σ150 nM, P9 Σ120 nM, P17 Σ35 nM, ever significantly higher than in ctrl, one-way ANOVA: p ≤ 0.0001, followed by Tukey post-test, p ≤ 0.001), felling to the levels not statistically different from those in ctrl in the adult age (adult jj Σ7 nM; One way ANOVA, followed by Tukey post-test, p > 0.05).
Cll weight (Fig. 1C) was similar in jj and ctrl littermates up to P9, becoming significantly different at P17 (Σ30% weight loss vs. age-matched ctrl, one way ANOVA followed by Tukey post-test: p < 0.001), and increasing later on (Adult: Σ40%, one way ANOVA followed by Tukey post-test: p < 0.001).
Western blot analysis of global acetylation of histone H3K14. The follow the level of H3K14Ac in the developing cerebellum of jj and controls rats by Western blot, we used the 07-353 anti-H3K13Ac antibody. At P2, no significant difference was observed in the level of H3K14Ac in the Cll of jj animals compared to age-matched ctrl (Fig. 1D) (unpaired t-test with Welch correction, p = 0.2687). The level of H3K14Ac in jj was significantly increased (1.65 ± 0.54 fold, unpaired t-test with Welch correction, p < 0.0222) at P9 and significantly decreased at P17 (0.67 ± 0.18 fold, unpaired t-test with Welch correction, p < 0.0187). In the adults there was no difference in the level of H3K14Ac between jj and ctrl (unpaired t-test with Welch correction, p = 0.4508).

ChIP-Seq analysis.
To link the effect of hyperbilirubinemia on H3K14Ac with the genes controlled by this epigenetic mechanism, the 07-353 anti-H3K13Ac antibody used for Western blot analysis was also used to perform chromatin immunoprecipitation, followed by DNA sequencing (ChIP-Seq -full result available on GEO repository # GSE109145). After removal of duplicate DNA fragments and DNA fragments present in both jj and ctrl (physiological genes), 1884 unique DNA sequences were identified. Since variations in the level of histone acetylation in the promoter region positively correlate with gene transcription 9,15 , we focused on peaks identified by ChIP-Seq on the promoter regions (Table 1: 255 genes). As shown in Fig. 2, the functional annotation analysis of the corresponding genes [16][17][18] revealed an enrichment for genes involved in CNS development (Σ45%), metabolism & homeostasis (Σ31%), signalling (Σ13%), response to stimuli & communication (Σ5%), transport (Σ5%), and binding (Σ2%).
Morphological features of the Gunn rat Cll. Since our results strongly suggested an impact of bilirubin on the genetic program of CNS maturation, we systematically followed the histological development of the cerebellum of jj rats in the attempt to interpret the genetic results. No morphological alterations between jj and ctrl were obvious at P2 (Fig. 3A,B). In both jj and ctrl animals, Purkinje cells were organized in 3-5 layers, with a round/oval shape and a reticulated cytoplasm (Fig. 3B). At P9, in spite of a conserved architecture, signs of cellular sufferance/death, microgliosis, extracellular matrix abnormalities and edema were evident in jj pups. PCs in ctrl displayed a clear definition of the plasma-membrane, cytoplasm, and nuclear areas, and a round/drop shape, and were organized in 3/1 layers. On the contrary, in jj pups, PCs were largely present in 4/2 layers, with an undefined, irregular shape. At P17, microgliosis and signs of cellular sufferance were still present in jj rats. PCs in ctrl were well differentiated, with a drop shape, and almost completely organized in a single layer, diffusely in 2/1 layers and still presenting the altered morphology described at P9 in jj. In the adult animal, the effect of Cll hypoplasia was well appreciable, with a less developed structure characterized by large spaces between the folia (Fig. 3A). Microgliosis was reduced but still present. No PC's neurites were visible in jj rats, where PCs appeared atrophic and apoptotic (Fig. 3B).

RTqPCR analysis of selected genes.
Due to the surprising percentage of enrichment for genes involved in CNS development, we decided to confirm and quantify the epigenetic control of a selected panel of genes, by assessing their expression by RTqPCR (selected genes are those in red in Fig. 2B, in which their biological functions based on the Gene Ontology analysis are indicated. RTqPCR results are in Fig. 4). Ptk2 (protein tyrosine kinase 2 beta, considered a key gene in neurite outgrowth and elongation, synapses formation, and actin reorganization 19 ), was significantly down-regulated in P2 jj pups (Σ2 fold vs. age-matched ctrl, unpaired t-test with Welch correction, p < 0.047), normalizing thereafter. Mag (myelin-associated glycoprotein), barely detectable immediately after birth, was highly expressed in ctrl and Σ2.5 fold down-regulated in jj pups at P9 (unpaired t-test with Welch correction, p < 0.0402), reversing to a Σ1.2 fold up-regulation at P17 (unpaired t-test with Welch correction, p < 0.0306). Icam1 (intracellular adhesion molecule 1, expressed mainly by the endothelial cells forming the blood-brain barrier, involved in cell adhesion, leucocytes 20 and monocytes extravasation 21 , and morphogenesis) was up-regulated 1.6 fold in P17 jj rats (unpaired t-test with Welch correction, p < 0.0416). Similarly, we observed a Σ2.2 fold increase (unpaired t-test with Welch correction, p < 0.0315) of Chmp1a (charged multi-vesicular body protein 1a, regulating the neural progenitor cell proliferation 22 ). In adult jj Cll, Col4a3 (collagenase 4a3, the major structural component of the basal membrane, involved in the extracellular matrix remodeling 23 , providing the functional compartmentalization of the brain by clustering of growth factors, neurotransmitters/ions receptors, as well contributing to migration and differentiation 24 ), Casp6 (caspase 6 -proliferation and morphogenesis - Fig. 2B), and Arghap4 (Rho GTPase-activating protein, inhibiting the cell motility and axon outgrowth via regulating the cytoskeleton dynamics 25 ) were upregulated Σ2.5fold (unpaired t-test with Welch correction, p < 0.00547), Σ1.9fold (unpaired t-test with Welch correction, p < 0.0287) and Σ1.6 fold (unpaired t-test with Welch correction, p < 0.0142) respectively. No modulation of Anxa2 (annexin2), Agrn (Agrin), and Tubb2b (Tubulin2b) was detected at any post-natal age in jj rats (data not shown). Il6 (intron region segment resulting from ChIP-Seq analysis) was also investigated. In ctrl animals the Il6 level rapidly decreases from P2 to P9, stabilizing thereafter. In jj pups, a significant down-regulation of Il6 was present immediately after birth compared to ctrl animals (Σ2.9fold, unpaired t-test with Welch correction, p < 0.0315), while a 1.65 fold up-regulation was noticed at P9 (unpaired t-test with Welch correction, p < 0.0248), normalizing later on.

Discussion
Cll hypoplasia is a hallmark of hyperbilirubinemia in rodents [26][27][28][29] , and cerebellar involvement with morphological and behavioral abnormalities has also been reported in severely hyperbilirubinemic neonates [30][31][32] . Inflammation and oxidative stress are considered the major mechanisms of bilirubin neurotoxicity, whereas the impact of hyperbilirubinaemia on CNS development has been only marginally envisaged, and evaluated mostly by in vitro experiments 33,34 . Unexpectedly, the known inflammatory or oxidant effectors of bilirubin neurotoxicity have been not identified in our data (ChIP-Seq, followed by Gene Ontology analysis), revealing that 45% of genes displaying a Histone 3 lysine 14 acetylation are related to CNS development. Indeed, only 3 genes among all the 255 identified TSS-Promoter sequences have been previously reported in the literature for their association with hyperbilirubinemia, namely myelin 28,31,32,34 , tubulin 35 , and Icam1 36 .
The down-regulation of Mag has been reported in in vitro studies, in agreement with the defective myelination observed both in bilirubin neurotoxicity models 28,34 and neonates 32 . Mag down-regulation is also a known consequence of bilirubin-induced perturbation of the oligodendrocytes maturation. A possible additional link between what has been previously described and the present results is the fact that histone acetylation is a known mechanism controlling oligodendrocyte differentiation and myelin production, both in physiological CNS development and in repair processes after demyelination 6,10 .
Our data are in agreement with the literature also in relation to Il6, whose intron sequence was identified by ChIP-Seq analysis. Il6 is a well-known effector of bilirubin neurotoxicity and possibly linked with the reported defective myelination. In fact, apart from the possible inflammatory activity, Il6 is involved in oligodendrogenesis 37,38 , a process active up to P45 in rodents and 2 years in humans 39 , and reactivated in pathological conditions. During reactivation, injured neurons and oligodendrocytes may reactivate myelin synthesis by overexpressing Il6 and its receptor (Il6r/CD126), restoring normal behavior in injured animals 10,40 . Both Mag and Il6 present a fluctuating behavior, being significantly down-regulated in the early post-natal life, and reverting thereafter to the level of age-matched controls (Fig. 4). Notably, in our work, IL6 modulation (P9) precedes Mag increase (P17), supporting the inductor role of Il6 in myelination described in the literature 10,40 . The fluctuating expression of Il6 and Mag (firstly up-, then down regulated), is present also for H3K14Ac levels, increasing at P9, and reverting under the level of age-matched controls at P17, and normalizing in the adult age.
The regulation of the other genes is more difficult to be analyzed since they are very new in the bilirubin field and no data are provided by literature. While we still have to confirm the role of the various genes identified in this study through methods such as gene silencing in vitro, our work suggests that the epigenetic impairment of neurodevelopmental processes in hyperbilirubinemia may be a relevant mechanism of bilirubin neurotoxicity. It is worth mentioning that Chmp1a, Arghap4, Casp6, Ptk2, Col4a3 are genes involved in key steps of brain development as proliferation, migration, morphogenesis, neurite outgrowth and elongation, synaptogenesis, extracellular matrix formation and compartmentalization, as well the pathological axonal degeneration and apoptosis observed 19,22,25,41,42 in jj rats. By adding epigenetic dysregulation to the list of the mechanisms related to bilirubin-induced neuronal damage, we can confirm and expand the concept of a widespread toxic effect of the pigment on the CNS 43 , improving our understanding of the cellular and molecular mechanisms of bilirubin induced damage to CNS.

Materials and Methods
Animals. Gunn rats (Hds Blue:Gunn-UDPGT j , P2, 9, 17; P ± 1 day. Adult = more than 1 year old) were obtained from the SPF animal facility of CBM S.c.a.r.l. (AREA Science Park, Basovizza). Ages were selected based on previous evidence 26,44 . Animals were housed in a temperature-controlled environment (22 ± 2 °C), on a 12 hours light/dark schedule, and ad-libitum access to food and water. The study was approved by the animal care and use committee of the CBM Scarl and the competent Italian Ministry. All procedures were performed according to the Italian Law (decree 87-848) and European Community directive (86-606-ECC). Maximal effort to minimize the number of the animals used and their sufferance was done. TSB, cBf and Cerebellum weight quantification. Serum and Cll were collected as previously described 26,45 . In brief, blood samples were collected during the sacrifice (decapitation under urethane anaesthesia 1.0-1.2 g/kg IP) and centrifuged at 2000 rpm, 20 min RT. Total serum bilirubin (TSB) was quantified by the diazo reaction, as previously described 26 . Free bilirubin was calculated (cBf) by applying the formula and the albumin-bilirubin dissociation constants for Gunn pups detailed in literature 14 . Cerebellum was dissected immediately after the sacrifice, and the weight recorded by a precision balance.
Western blot analysis of the levels of H3K14Ac. Western blot was performed as previously described 44,45 . In brief, Cll were mechanically homogenized by glass-glass Dounce (in 0. 25    Protein tyrosine kinase 2 beta; Il6: Interleukin 6. P: post-natal age in days, Adult: more than 1-year-old. White bars: ctrl; Black bars: jj. Results are expressed as mean ± S.D. of 6 animals each genotype/age. Unpaired t-test with Welch correction, *p < 0.05; **p < 0.05; ***p < 0.005 vs. age-matched controls. Statistics. The statistical analysis was performed by GraphPad InStat for Windows (GraphPad Software, Inc, La Jolla, CA, USA). The ANOVA test, followed by Tukey-Kramer multiple comparison tests, was used to analise TSB, cBf, and Cll weight during the development. The unpaired two-tailed Student's t-test, based on unequal variance, was applied to evaluate the difference between jj and controls at the same age (Western blot, RTqPCR). All data are expressed as mean ± S.D. of multiple biological repetition. A p-value lower than 0.05 was considered statistically significant.

Data Availability
ChIP-Seq -full result available on GEO repository # GSE109145.  Table 2. Primers specification.