Rescue of bilirubin-induced neonatal lethality in a mouse model of Crigler-Najjar syndrome type I by AAV9-mediated gene transfer

Abstract

Crigler-Najjar type I (CNI) syndrome is a recessively inherited disorder characterized by severe unconjugated hyperbilirubinemia caused by uridine diphosphoglucuronosyltransferase 1A1 (UGT1A1) deficiency. The disease is lethal due to bilirubin-induced neurological damage unless phototherapy is applied from birth. However, treatment becomes less effective during growth, and liver transplantation is required. To investigate the pathophysiology of the disease and therapeutic approaches in mice, we generated a mouse model by introducing a premature stop codon in the UGT1a1 gene, which results in an inactive enzyme. Homozygous mutant mice developed severe jaundice soon after birth and died within 11 d, showing significant cerebellar alterations. To rescue neonatal lethality, newborns were injected with a single dose of adeno-associated viral vector 9 (AAV9) expressing the human UGT1A1. Gene therapy treatment completely rescued all AAV-treated mutant mice, accompanied by lower plasma bilirubin levels and normal brain histology and motor coordination. Our mouse model of CNI reproduces genetic and phenotypic features of the human disease. We have shown, for the first time, the full recovery of the lethal effects of neonatal hyperbilirubinemia. We believe that, besides gene-addition-based therapies, our mice could represent a very useful model to develop and test novel technologies based on gene correction by homologous recombination.—Bortolussi, G., Zentilin, L., Baj, G., Giraudi, P., Bellarosa, C., Giacca, M., Tiribelli, C., Muro, A. F. Rescue of bilirubin-induced neonatal lethality in a mouse model of Crigler-Najjar syndrome type I by AAV9-mediated gene transfer.

The Crigler-Najjar type I (CNI) syndrome is a rare autosomal recessive disease characterized by a severe congenital nonhemolytic hyperbilirubinemia. This pathological condition is caused by the complete absence of bilirubin glucuronidation activity in the liver, due to mutations in the uridine diphosphoglucuronosyltransferase 1a1 (UGT1A1) gene (1, 2). Bilirubin glucuronidation is essential to convert unconjugated bilirubin (UCB) to a water-soluble and nontoxic compound easily excretable in the bile, thus avoiding accumulation of toxic levels of UCB in the plasma.

Untreated babies with CNI rapidly develop high plasma levels of UCB (20–50 mg/dl), which result in serious neurological damage and ultimately lead to death. Before the introduction of phototherapy, CNI was fatal by the age of 1–2 yr due to kernicterus (3, 4). Currently, patients with CNI undergo 10–12 h of daily phototherapy treatment. Even though phototherapy is the treatment of choice during the first years of life due to its high efficiency and ease of application at home, phototherapy becomes less effective during puberty because of increased skin thickness and pigmentation and decreased surface/mass ratio. Thus, patients with CNI are under constant risk of developing permanent brain damage, and the only permanent cure at present is liver transplantation, which is not devoid of complications and lifelong associated risks (5).

The monogenic nature of CNI and the diffusible effect of the glucuronidation activity make gene therapy an attractive therapeutic approach. Indeed, during the past 2 decades, a wide spectrum of approaches involving both viral and nonviral gene transfers has been evaluated using a rat model of the disease (6).

In the current study, we generated a novel mouse model for CNI by targeting a nonsense point mutation into the Ugt1 gene. This mouse model resembles most major features of the human syndrome, such as neonatal hyperbilirubinemia and early lethality due to bilirubin-induced neurological damage. The cerebellar architecture in the knock-in mice was significantly affected, together with reductions in Purkinje cell number and dendritic arborization. The disease could be effectively prevented by an adeno-associated virus (AAV)-based gene therapy approach consisting of a single intraperitoneal injection of an AAV9 vector coding for the human UGT1A1 protein in neonatal mice. AAV treatment rescued the lethal phenotype, and mice reached adulthood without any evidence of neurological damage.

We believe that this new model may represent a very promising tool to study and develop therapies for CNI.

MATERIALS AND METHODS

Animals

Mice were housed and handled according to institutional guidelines, and experimental procedures approved by the International Centre for Genetic Engineering and Biotechnology (ICGEB) board. Animals used in this study were of ∼94% C57Bl/6 genetic background, obtained after 4 backcrosses with C57Bl/6 mice. Animals were kept in a temperature-controlled environment with a 12–12 h light-dark cycle. They received a standard chow diet and water ad libitum. For AAV gene transfer, mutant pups at postnatal day 2 (P2) were intraperitoneally injected with a single dose of AAV9/CMV-hUGT1A1 (∼1.7×10¹⁰ viral particles).

Generation of Ugt1^−/− mice

To generate the Ugt1 mutant mice, the entire Ugt1a1 locus, plus the promoter and the 3′ gene-flanking regions, was amplified by PCR using Pfu DNA polymerase system (Expand High FidelityPlus; Roche, Monza, Italy) and cloned into a pFrlt1 targeting vector. PCR was performed using genomic DNA derived from the c129Sv/J embroynic stem cells that were subsequently used to perform the electroporation. The total length of homology region was ∼8 kbp, divided into 3 fragments: the long arm of homology (5.1 kbp) included the Ugt1a1 gene promoter, exons 1 to 3 and intervening introns; the fragment containing exon 4 with its flanking introns (∼0.7 kbp); and the short arm (2.8 kbp) containing exon 5 and downstream sequences (see Fig. 1). Absence of mutations was confirmed by aligning the available genomic sequence from the gene bank (C57Bl/6) with that obtained from the sequencing of ≥2 independent clones (originated from different PCR reactions). To generate the single base deletion in exon 4, we used an in vitro site-directed mutagenesis system (Stratagene, Santa Clara, CA, USA) with specific synthetic oligonucleotides carrying the mutation (Ugt1a1mutDIR; Ugt1a1mutREV). Two loxP sites and 2 Frt sites were included in the targeting construct flanking exon 4 and the Neo cassette, respectively.

Figure 1.

Generation of Ugt1 knock-in mice. A) Targeted disruption of the mouse Ugt1 locus. Maps of the Ugt1a1 locus, the targeting vector, the targeted allele, and the FlpE-ed allele are shown. Exonic regions are represented as open boxes. Exon (EX) 4 mutated is flanked by LoxP sites (solid box flanked by shaded arrowheads), while PGK-Neomycin cassette is flanked by Frt sites (Neo, shaded box flanked by ovals). Probes used in the Southern blot analysis to screen targeted embryonic stem clones and mice (B probe and 3′ probe) are indicated as solid black bars. B indicates BglII restriction sites. Fragments generated by digestion of genomic DNA with BglII are indicated for the wild-type (WT) and targeted loci (4.9 and 5.1, respectively). B) Partial sequences of the WT and mutant alleles of exon 4 are shown, together with their translation products. Deleted guanine and the generated stop codon are indicated in the WT and mutant (MUT) sequences (bold underscored G and bold TGA and Stop, respectively). C) Southern blot analysis of genomic DNA from WT, heterozygous (HET), and mutants after Flp-mediated recombination (without the neomycin cassette), digested with BglII. B probe was used. MWM, molecular weight marker.

The full targeting vector was then electroporated into embroynic stem cells, and the resulting ∼200 G418-resistant clones were screened by Southern blot hybridization with the 3′ probe (a probe external to the homology region). We obtained 2 positive clones that were confirmed by cutting the genomic DNA with BglII and using the internal B probe. The B probe was obtained performing a PCR (B-probe dir and B-probe rev) and digesting the resulting product with EcoRV restriction enzyme to obtain a fragment of ∼1.1 kbp. To verify the absence of undesired genomic rearrangements in the locus and to confirm homologous recombination events, additional Southern blot analysis was performed, digesting the genomic DNA of the positive clones with a number of different restriction enzymes and hybridizing them with different probes (Neo probe and 3′ probe). Furthermore, in the positive clones, the region corresponding to the exon 4 fragment was sequenced to further confirm the presence of the introduced mutation in exon 4 and the presence of the loxP sites in the recombinant allele. Before blastocyst microinjection, karyotyping of the positive clones was performed to verify the correct chromosome complement.

Positive clones were injected into C57Bl/6 blastocysts to obtain chimeras. Male chimeras were mated with C57Bl/6 females to check for germline transmission (F1 generation). F1 offspring were analyzed by Southern blot for the transmission of the mutated exon 4 within the Ugt1 locus; genomic DNA from tail biopsies was digested with BglII, and the membrane was hybridized with the B probe. F1 heterozygous mice were mated with FlpE transgenic “deleter” mice [Tg(ACTFLPe)9205Dym, C57Bl/6 background] to remove the Neo cassette from the targeted locus and obtain the F2 generation. Southern blot using the Neo probe confirmed the absence of the Neo cassette. Heterozygous mice without the Neo cassette were backcrossed with C57Bl/6 wild-type mice for 2 generations to obtain an F4 generation (∼94% C57Bl/6 background). Heterozygous mice of the F4 generation were then mated to obtain homozygous Ugt1^−/− mutant mice. PCR genotyping was performed using the following primers: Ex4 Screen dir; Ugt1 9934 rev.

Phototherapy treatment

Newborn pups were exposed to blue fluorescent light (λ=450; 20 μW/cm²/nm; Philips TL 20W/52 lamps; Philips, Amsterdam, The Netherlands) for 12 h/d (synchronized with the light period of the light-dark cycle). Intensity of the lamps was monitored monthly with an Olympic Mark II Bili-Meter (Olympic Medical, Port Angeles, WA, USA).

Preparation of total RNA from the mouse liver and Northern blot analysis

Total RNA from liver was prepared as described previously (7). To perform Northern blot analysis, 20 μg of total liver RNA was denaturated and run on a 1.2% agarose formaldehyde gel and blotted onto a nylon membrane (Hybond-N; Amersham Biosciences, Uppsala, Sweden). The membrane was incubated with UltraHyb prehybridization solution (Ambion, Austin, TX, USA) and subsequently hybridized with a P³²-radiolabeled probe located in exon 5 of the Ugt1 gene (Northern dir; Northern rev). After being washed, the membrane was exposed overnight using a Cyclone phospho-screen (Packard Bioscience Co., Downers Grove, IL, USA) and the detection of the radioactive signal was done with Cyclone Storage phospho-imager (Packard Bioscience). The membrane was stripped and hybridized with a Gapdh probe. Densitometric analysis was performed using Quantity One 4 software (Bio-Rad Laboratories, Hemel Hempstead, UK).

Preparation of protein extracts and Western blot analysis

Liver and skeletal muscle tissues were surgically removed, homogenized in lysis buffer, and analyzed by Western blot as described previously (7). Primary antibodies used were as follows: anti-human UGT1 rabbit polyclonal antibody (Santa Cruz Biotechnology, Heidelberg, Germany) and anti-β-tubulin mAb E7 (Developmental Studies Hybridoma Bank, Iowa City, IA, USA).

UGT1a1 activity determination and HPLC-MS analysis

Microsomes from liver (5-d- and 5-mo-old mice) and skeletal muscle (from 5-mo-old mice) of each genotype were prepared as described previously (8). Protein concentration was determined by the bicinchoninic acid assay (9).

Glucuronidation assay was performed as described previously (8) with minor modifications. Briefly, assays were performed in triplicate after 1 h of enzymatic reaction using the following conditions: 10 mM MgCl₂, 50 mM Tris-HCl (pH 7.5), 10 μg/μl phosphatidycholine, 15 μM bilirubin (bilirubin was previously dissolved in DMSO at a concentration of 0.33 μg/μl), 1 mM uridine diphosphate-glucuronic acid, and 0.5 μg/μl microsomal proteins (previously incubated for 1 h with digitonin at a concentration of 0.35 mg/mg of microsomes; ref. 10) in a total volume of 100 μl. Reactions were stopped with 100 μl of methanol with 0.02% of butylated hydroxytoluene. Samples were centrifuged at 10,000 g for 10 min at 4°C, and supernatants were collected for HPLC-MS analysis (LC-MS) as described previously (8) and adapted to our instrumentation.

Supernatant was transferred into a conical vial for injection into the LC-MS system. The HPLC used was a Surveyor Thermo Finnigan system with pump plus autosampler and a diode array detector (Thermo Finnegan, San Jose, CA, USA). Bilirubin and its glucurono-conjugated species were injected and separated on a 4.6-mm inner diameter × 100-mm Luna C18 column (3-μm particle size; Phenomenex, Torrance, CA, USA) with a security guard cartridge with the same stationary phase (3-mm inner diameter × 4 mm; Phenomenex). The mobile phase A was 1 mM ammonium formate in water, and the mobile phase B was 1 mM ammonium formate in methanol. Separation was achieved using a linear gradient of 70% B to 95% B in 2.5 min at a flow rate of 0.9 ml/min. After 0.5 min, the column was reequilibrated to initial conditions over 3 min, stopping the runs at 10 min. The absorbance of the eluted pigments was monitored at 455 nm with 195 nm as a reference wavelength.

Mass spectrometry characterization and detection of bilirubin and the mono- and dibilirubin glucuronide conjugates formed were performed using a LCQ Deca XP Plus model (Thermo Finnigan), utilizing a standard electrospray ionization (ESI) source operated in positive mode and with an ion trap detector.

Bilirubin determination in plasma

Blood was obtained from P2 and P5 mice by decapitation, and from adult mice by cardiac puncture, in EDTA-collecting tubes, and kept in the dark. Total plasma bilirubin levels were determined as described previously (11), adapting the method to use minimal volumes. Standard curves were performed by dilution of a stock solution of bilirubin (20 mg/dl) in BSA (4 g/dl). Purified bilirubin was previously dissolved in DMSO (3 mg/ml) and then added to BSA in PBS. Two commercial bilirubin reference standards (Bio-Rad Laboratories, Irvine, CA, USA) were included in each set of analysis as quality control. Absorbance values at 600 nm were obtained by using a multiplate reader (LD 400C Luminescence Detector; Beckman Coulter, Milan, Italy).

Tissue histology and immunoflourescence

Total brains from all genotypes were extracted and fixed with 4% paraformaldehyde in PBS overnight a 4°C. After cryoprotection in 20% sucrose in PBS and 0.02% sodium azide, specimens were frozen in optimal cutting temperature compound (Sigma-Aldrich, St. Louis, MO, USA), and sagittal sections were obtained in a cryostat. For Nissl staining, slides (14- or 30-μm-thick sections) were briefly submerged in cresyl violet solution (5 mg/ml cresyl violet acetate in 0.3% acetic acid; Sigma-Aldrich), dehydrated, and mounted.

For immunofluorescence, slices (30 μm thick) were blocked for 2 h at room temperature in 2.5% BSA in PBS plus 0.1% Triton X-100. After blocking, specimens were incubated overnight with the primary antibody mAB anti-calbindin (1:250; Synaptic Systems, Göttingen, Germany) and, after 3× washes in PBS plus 0.05% TX-100, for 3 h with secondary antibodies conjugated with Alexa Fluor 568 (Invitrogen, Carlsbad, CA, USA). Nuclei were visualized by addition of Hoechst (10 μg/ml; Invitrogen) for 5 min after the secondary antibody solution.

Nissl-stained slides were mounted using Eukitt (Fluka, St. Louis, MO, USA) while immunostained 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 1200F; Nikon, Amstelveen, The Netherlands). Digital images were collected using ACT-1 (Nikon) software and analyzed using ImageJ (U.S. National Institutes of Health, Bethesda, MD, USA).

Analysis of the layer thickness was performed on Nissl-stained sections (in 30 μm sections) by measuring the layer depth (μm) in lobules IV, V, and VI (n=3 animals/genotype; 5 sections/animal). Purkinje cell number was calculated by counting calbindin-positive cells in vermis sections along the entire cerebellum perimeter (5 slices/animal; 3 animals/genotype). Total distance and number of counts were recorded for each slide. Cell numbers were then expressed as linear density (cells/mm). The genotype of the animals was unknown to the operator. A second operator analyzed the data, and measurements were averaged for each animal. The results are expressed as means ± sd for each genotype.

Rotarod analysis

The ability to maintain balance on a rotating cylinder was measured with an accelerating rotarod apparatus, as described previously (12). Before the test, mice were trained with 3 trials at constant speed (slow, moderate and high: 2.8, 5.5, and 8.0 rpm, respectively). After the training session, mice were placed on the rotating rod with a linear increase in rotation speed from 2.5 to 48 rpm over a 5-min period. Mice that fell within 15 s of starting the trial were given a second chance. Mice were subjected to 3 trials of accelerating rotarod test for 2 consecutive days, and the latency to fall was recorded. The test was repeated each month. The genotype of the animals was unknown to the operator. To avoid minor differences between the tracks and time of the test, both the tracks used and the order of the mice were recorded. Mice changed tracks and order daily. All the experiments were done in the same time of the day.

Production, purification, and characterization of rAAV vectors

The AAV-hUGT1a1 vector used in this study is based on AAV type 2 backbone in which the inserted human UGT1a1 cDNA is under the transcriptional control of the strong, constitutive cytomegalovirus (CMV) immediate early promoter. Infectious vector stocks were prepared by the AAV Vector Unit at ICGEB Trieste (http://www.icgeb.org/avu-core-facility.html), as described previously (13) with few modifications. In brief, infectious recombinant AAV vector particles were generated in HEK293 cells by a cross-packaging approach, whereby the vector genome was packaged into AAV capsid serotype 9 (14, 15). Viral stocks were obtained by CsCl gradient centrifugation; rAAV titer, determined by measuring the copy number of viral genomes in pooled, dialyzed gradient fractions, as described previously (16), was in the range of 1 × 10¹¹ to 1 × 10¹² genome copies/ml.

PCR and real-time PCR to detect viral genomic particles (VGPs) in tissues

Total DNA from tissues was extracted using the Wizard SV Genomic DNA Purification System (Promega, Madison, WI, USA) according to the manufacturer's instructions. The vector genome copy number was quantified by real-time PCR using Taqman probe technology (Applied Biosystems, Carlsbad, CA, USA) and primers specific for the CMV promoter sequence. Real-time PCR was performed using the following primers: pZac dir; pZac rev.

RT-PCR to detect hUGT1A1 expression in AAV-transduced tissues

Total RNA from liver and skeletal muscle was extracted as described previously. Total RNA (1 μg) was reverse transcribed with M-MLV reverse transcriptase (Invitrogen) according to manufacturer's instructions. Total cDNA (1 μl) was used to perform PCR using specific primers able to amplify the human version of the UGT1A1 mRNA and not the endogenous mouse version (hUGT1A1: RT-pZac-1241dir, RT-hUGT1A1–529 rev; Gapdh primers: RT-mGapdh dir, RT-mGapdh rev).

Statistics

Results are expressed as means ± sd. The Prism package (GraphPad Software, La Jolla, CA, USA) was used to analyze the data. Values of P < 0.05 were considered statistically significant.

List of primers

Primers used were as follows: Ugt1a1mutDIR: gtgaccctgaatgtccttaaatgactgctgatgatttg, Ugt1a1mutREV: caaatcatcagcagtcatttaaggacattcagggtcac; B-probe dir: tggcatccttttctgacttacc, B-probe rev: ttggaggatgtcagaggatttc; Ex4 Screen dir: tcaccagagtaggcatctcatc, Ugt1 9934 rev: gctgtaagacaatcttctcc; Northern dir: gctacaaggagaacatcatgcg, Northern rev: aaaggaacacctccagagacc; pZac2 dir: gtgtccactcccagttcaat, pZac2 rev: gtggtttgtccaaactcatc; pZac dir: tgggcggtaggcgtgta, pZac rev: gatctgacggttcactaaacgag; RT-pZac-1241dir: cagtgcttctgacacaacagtct, RT-hUGT1A1–529 rev: gcaagaagaatacagtgggcag; and RT-mGapdh dir: gcatggactgtggtcatgag, RT-mGapdh rev: ccatcaccatcttccaggag.

RESULTS

Generation and characterization of the Ugt1 knock-in mice

To generate the Ugt1 knock-in mice, a single base deletion was introduced by gene targeting in exon 4 (Fig. 1A). This mutation, similar to the one present in the hyperbilirubinemic Gunn rat model (1), leads to a frameshift generating a stop codon immediately downstream (Fig. 1B). The PGK-Neo cassette was removed from the Ugt1 locus by crossing heterozygous mice with a transgenic strain expressing the FlpE recombinase.

Gene expression was determined by Northern and Western blot analyses of total liver RNA and microsomal protein preparations, respectively. Northern blot analysis showed that Ugt1 mRNA levels of mutant mice at P5 were 22% of wild-type control values, while the mRNA levels of the heterozygous pups were intermediate (58%) between the wild-type and homozygous mutant levels (Fig. 2A). Regardless of the presence of Ugt1 mRNA in mutant mice, no band was detected by Western blot analysis (Fig. 2B), suggesting that the truncated protein, lacking the transmembrane domain, was rapidly degraded. Moreover, enzymatic assays using UCB as a substrate demonstrated that liver microsomes from P5 mutant mice had no Ugt1a1 glucuronidation activity (Fig. 2C), confirming observations in the Western blot analysis.

Figure 2.

Characterization of Ugt1 knock-in mice. A) Left panel: Northern blot analysis of total liver RNA extracts (20 μg) from WT, heterozygous, and homozygous mutant mice using a probe in the exon 5 of Ugt1 mRNA. Gapdh probe was used as a loading control on the same membrane, after stripping of the Ugt1 probe. Right panel: densitometric quantification of the bands. B) Left panel: Western blot analysis of total liver protein extracts (20 μg) from mice of all genotypes using an anti-UGT1 antibody. Anti-β-tubulin mouse antibody was used as loading control. Right panel: densitometric quantification of the bands. C) HPLC ultraviolet-chromatograms show the elution profile of the Ugt1a1 activity-products after incubation of bilirubin with liver microsomes from WT, heterozygous, and homozygous mutant mice (each graph represents the corresponding genotype). Peaks corresponding to unconjugated and monoconjugated bilirubin are indicated.

Absence of Ugt1a1 enzyme causes hyperbilirubinemia and early neonatal lethality

Mutant mice developed jaundice as early as 36 h after birth, evidenced by the yellowish-orange color of the skin (Fig. 3A). Within a few days after birth, jaundice progressed to a severe hyperbilirubinemic condition, reaching UCB levels ∼42 times higher than in control littermates at P5 (wild type: 0.29±0.17 mg/dl, hetero: 0.39±0.20 mg/dl, and mutant: 12.22±1.49 mg/dl; Fig. 3B and Table 1). This was due to the lack of bilirubin-glucuronidation activity in Ugt1 mutant mice (Fig. 2C), leading to high levels of UCB in plasma and tissues.

Figure 3.

Absence of Ugt1a1 enzyme causes jaundice and early neonatal lethality. A) As early as 36 h after birth, mutant mice (red arrowheads) develop hyperbilirubinemia and become jaundiced; this is easily seen by the orange-yellowish coloration of the skin in P2 mutant mice. B) Total plasma bilirubin levels (mg/dl) of 5-d-old mice. Values represent means ± sd of WT (n=4); heterozygous (n=9) and homozygous mutant (n=7) plasma samples. Error bars = sd. ***P ≤ 0.001. Bottom panel: appearance of representative plasma samples derived from each genotype. C) Kaplan-Meier survival curves of mutant mice and their WT and heterozygous littermates (P≤0.001, WT vs. MUT).

Table 1.

Total plasma bilirubin levels

Treatment	WT	HET	MUT
P2	0.41 ± 0.21 (9)	0.81 ± 0.41 (12)	10.31 ± 0.44* (3)
P2 blue light	0.69 ± 0.65 (3)	0.48 ± 0.17 (5)	3.65 ± 1.17* (5)
P5	0.29 ± 0.17 (4)	0.39 ± 0.20 (9)	12.22 ± 1.49@ (7)
P15	0.17 ± 0.07 (4)	0.29 ± 0.18 (4)	–
P15 + AAV9	–	–	1.39 ± 0.13§ (2)
P150	0.18 ± 0.02 (2)	0.25 ± 0.17 (2)	–
P150 + AAV9	–	–	4.76 ± 1.40& (5)

Values represent means ± sd. Values in parentheses indicate number of samples analyzed.

^*P ≤ 0.001 vs. P2WT; ^#P ≤ 0.001 vs. P2MUT;

^@P ≤ 0.001 vs. P5WT;

^§P ≤ 0.001 vs. P5MUT;

^&P ≤ 0.001 vs. P5MUT.

Kaplan-Meier survival curve of 20 mutant pups showed that 50% of mice died within 5 d after birth (Fig. 3C) with no survivors after P11. UCB progressively accumulated in the brain of mutant mice, producing an intense and diffuse yellow staining (Fig. 4A). Moreover, neonatal mutant mice (P3–P6) displayed many of the clinical symptoms of kernicterus, including severe motor impairment characterized by a marked improper posture of the rear limbs, poor and slow movement, and decreased feeding (Supplemental Video S1). In mice that reached the second week of life (P7–P11), movement disorders such as athetosis, dystonia, ataxia, and stimulus-evoked hyperactivity resembling seizures were observed (data not shown).

Figure 4.

UCB induces neurological damage in the cerebellum of Ugt1 mutant mice. A) Comparison between WT and mutant mouse brains. As a result of severe hyperbilirubinemia, UCB accumulates in the brain of mutant mice, staining the entire tissue. B) Nissl staining of 5-d-old WT and mutant cerebellum slices. Misshapen fissures are indicated: I, IV, VIa, VIb, and IXb. Boxed areas indicate fields shown in C. C) High-magnification image of cerebellar layers from P5 WT and mutant littermates. EGL, external germinal layer; ML, molecular layer; IGL, internal granular layer. D) Layer depth of P5 WT and mutant cerebella (n=3/genotype). Scale bars = 1 mm (B); 100 μm (C). Error bars = sd. **P < 0.01; ***P < 0.001.

Bilirubin induces neurological damage in the cerebellum of Ugt1 mutant mice

To characterize the bilirubin-induced neurological damage in more detail, we performed histological analysis of the entire brain of P5 wild-type and mutant mice using Nissl staining. Gross histological analysis showed no major changes except for the cerebellum (Fig. 4B and Supplemental Fig. S1A). We observed that cerebella of mutant mice were less developed compared with those of wild-type littermates. Indeed, the cerebellar fissures were misshaped and, in particular, fissures in lobules I, IV, VIa, and IXb were almost absent. Furthermore, closer examination revealed that mutant mice had an abnormal stratification of the cerebellar layers (Fig. 4C). We measured cerebellum layer thickness of wild-type and mutant Nissl-stained sagittal sections, and we observed that mutant mice had a thinner external germinal layer (19.3±2.9 μm; 61% reduction, P<0.001) compared with control littermates (49.6±5.7 μm; Fig. 4D). We also observed a significant increase in the depth of the internal granular layer in mutant cerebella (wild type: 77.1±7.4 μm, mut: 96.7±4.8 μm; P≤0.01).

Fluorescent immunohistochemistry analysis of cerebellum using anti-calbindin antibody revealed that Purkinje cells of P5 mutant mice were severely affected and that their number was significantly reduced (37% reduction; P<0.001), compared with wild-type littermates (Fig. 5A, B). A detailed observation of histological preparations showed that, in mutant mice, the Purkinje cell layer architecture was very disorganized, and the dendritic arbor was underdeveloped (Fig. 5C, right panel). On the contrary, wild-type Purkinje cells settled into an ordered monolayer with an extensively arborized dendritic tree (Fig. 5C, left panel). Moreover, terminal deoxynucleotidyl transferase-mediated deoxyuridine triphosphate nick end-labeling (TUNEL) analysis of cerebellum highlighted the presence of TUNEL positive cells in P5 mutant mice (Fig. 6) suggesting the presence of apoptotic and/or necrotic cells. No other brain regions of mutant brain (hippocampus, cortex, and basal ganglia) were positive for the assay (Supplemental Fig. S1C, D), confirming Nissl histological results (Supplemental Fig. S1A, B).

Figure 5.

Purkinje cells are affected by high levels of UCB. A) Fluorescent immunohistochemistry of P5 WT and mutant cerebella using Hoechst nucleus staining (blue) and anti-calbindin antibody (red). Left panels: low-magnification images of total cerebellum. Middle and right panels: high-magnification images of fields 1 and 2 (boxed areas in left panels). B) Quantification of Purkinje cell number (cells/mm) in WT and mutant cerebella (n=3/genotype). Error bars = sd. ***P < 0.001. C) High-magnification image of Purkinje cells in P5 WT and mutant cerebella. Scale bars = 1 mm (A, left panels); 400 μm (A, right panels); 100 μm (C).

Figure 6.

UCB induces cell death in the cerebellum of Ugt1 mutant mice. Comparison between P5 WT (left panels) and mutant mouse cerebella (right panels) by TUNEL analysis of cerebellum and counterstaining with methyl green. Arrows indicate apoptotic and/or necrotic cells. Boxed areas indicate fields shownin bottom panels. Scale bars = 1 mm (top panels); 200 μm (bottom panels).

AAV9 gene therapy rescues neonatal lethality in Ugt1 knock-in mice

To rescue the lethal phenotype of mutant mice, we moved toward a gene therapy approach based on AAV gene transfer. Due to early death of mutant mice, the gene therapy treatment of pups was combined with phototherapy during the first week of life to avoid undesired brain damage before full expression of the therapeutic gene. Newborn pups were exposed to blue light (λ=450 nm, 12 h/d), and mutants were injected i.p. at P2 with a single dose of AAV9 vector containing hUGT1A1 under the control of CMV promoter (∼1.7×10¹⁰ VGPs). Pups were maintained under phototherapy for another 6 d and then removed from blue light treatment and kept under normal light conditions. We observed that P2 mutant mice exposed to phototherapy from birth had significantly lower plasma bilirubin levels (3.65±1.17 mg/dl; n=5) as compared with mutant mice not treated with blue light (10.31±0.44 mg/dl; n=3; P≤0.001; Table 1). Although the treatment alone improved the survival of mutant mice, it was not sufficient to extend their life span to >20 d, with a 50% survival of 18 d (P≤0.001; Fig. 7A).

Figure 7.

hUGT1A1-AAV9 gene therapy treatment rescues neonatal lethality in Ugt1 knock-in mice. A) Kaplan-Meier survival curves of WT/heterozygous mice (dashed line), untreated mutant mice (solid line with circles), mutant mice treated only with phototherapy (solid line with triangles), and mutant mice treated with hUGT1A1-AAV9 (solid line). Median survival: MUT, 5 d; MUT + phototherapy, 18 d; P ≤ 0.001, MUT vs. MUT + phototherapy. B) Time course of motor coordination performance on rotarod of WT and AAV-treated mutant mice (WT: n=9; mut+AAV9: n=8). Repeated-measures ANOVA test, nonsignificant.

In contrast, we observed that all mutant mice treated with AAV9 CMV-hUGT1A1 survived (Fig. 7A). AAV-treated mutant mice reached adulthood and were indistinguishable from their control littermates. To determine more complex neurological functions, such as motor coordination and balance abilities, we tested their performance in the accelerating rotarod test. After the training session (see Materials and Methods) mice were subjected to 3 trials/session of accelerating rotarod test for 2 consecutive days, and the latency to fall was recorded. The test was repeated 1×/mo (up to 5 mo of age) until the animals were killed. The results obtained indicated that mutant mice treated with AAV had no obvious motor coordination impairment compared with wild-type control littermates at any time point (Fig. 7B).

Bilirubin levels of AAV-treated mutant mice were significantly reduced at 15 d after birth (Table 1). Those values were slightly increased at 5 mo postinjection but were well below the levels of untreated younger mice.

AAV9-hUGT1A1 protein is expressed in skeletal muscle of AAV-treated mutant mice

Mice were killed 5 mo after hUGT1A1-AAV9 treatment, and tissue samples from liver, skeletal muscle, heart, intestine, kidney, and diaphragm were collected. PCR analysis of total genomic DNA showed the presence of viral genomic DNA in liver, skeletal muscle, heart, and diaphragm (Fig. 8A). We focused our analysis in liver and skeletal muscle and quantified by real-time PCR the number of VGPs. We observed that at 5 mo postinjection, skeletal muscle of AAV-treated mice contained a higher level of absolute VGPs than liver (mean 33,700±30,971 and 10,343±4867 VGP/30 ng total genomic DNA, respectively; Table 2).

Figure 8.

hUGT1A1 expression in AAV9-treated mice. A) Tissue distribution of VGPs in AAV9-treated mutant mice at 5 mo postinjection. PCR of 30 ng total genomic DNA from liver (L), heart (H), intestine (I), kidney (K), diaphragm (D), and skeletal muscle (SM) of AAV-treated mutant mice at 5 mo postinjection. B) RT-PCR of total liver and skeletal muscle RNA from WT, heterozygous, and AAV-treated mutant mice at 15 d and 5 mo postinjection. Mouse Gapdh was used as endogenous control. Plasmid containing the hUGT1A1 cDNA was used as a positive control (+). C) Western blot analysis of total liver and skeletal muscle protein extracts (50 μg) from WT, heterozygous, and AAV-treated mutant mice at 15 d and 5 mo postinjection, using an anti-UGT1A antibody. Anti-β-tubulin mouse antibody was used as loading control. Mutant mouse total protein extract (M) was used as a negative control; total liver extract from WT mouse was used as a positive control. D) HPLC UV-chromatograms showing the elution profile of the UGT1A1 activity products after incubation of bilirubin with liver and skeletal muscle microsomes from WT, heterozygous, and AAV-treated mutant mice (each graph represents corresponding genotype). Peaks corresponding to unconjugated and monoconjugated bilirubin are indicated.

Table 2.

VGPs in AAV9-treated mutant mice

Tissue	VGPs
	15 d	150 d
Liver	14,833 ± 5179 (3.0 ± 1.0)	1852 ± 851 (0.4 ± 0.2)
Skeletal muscle	33,700 ± 30971 (6.7 ± 6.2)	10,343 ± 4867 (2.1 ± 1.0)

Values represent means ± sd. VGPs in 30 ng of total genomic DNA was quantified by real-time PCR in liver and skeletal muscle samples of 15- and 150-d-old AAV9-treated mutant mice. Values in parentheses indicate number of VGP copies per diploid genome. Three samples per genotype and tissue were analyzed.

Liver and skeletal muscle RNA samples from AAV-treated mutant mice were analyzed by semiquantitative RT-PCR to determine hUGT1A1 mRNA expression at 15 d and 5 mo after injection. We observed that the relative levels of hUGT1A1 mRNA in skeletal muscle had comparable expression levels at both 15 d and 5 mo postinjection (Fig. 8B, bottom panel). On the contrary, we could not detect hUGT1A1 mRNA in the liver of AAV-treated mutant mice at any of the analyzed time-points (Fig. 8B, top panel). Western blot analysis, using a polyclonal antibody that recognizes both mouse and human UGT1 families of proteins, confirmed the data obtained by RT-PCR, showing that skeletal muscle expressed hUGT1A1 at comparable levels at all time points analyzed, while no detectable band was present in liver protein extracts of hUGT1A1-AAV9-treated mice even at 15 d after injection (Fig. 8C). As expected, glucuronidation activity assays of wild-type and heterozygous controls showed enzymatic activity only in liver microsomes and not in skeletal muscle (Fig. 8D). Analysis of hUGT1A1-AAV9-treated mutant mice showed enzymatic activity only in skeletal muscle, while the liver had no detectable glucuronidation activity, confirming RT-PCR and Western blot analysis described above.

P5 mutant mice had a colorless gallbladder due to the absence of glucuronidated bilirubin, while gallbladders of their wild-type or heterozygous littermates had a yellowish color (Fig. 9A, top panels). Liver histology of hUGT1A1-AAV-treated mutant mice was normal (Fig. 9B), and their gallbladder showed a yellowish color, most probably as the result of bilirubin glucuronidation in skeletal muscle. (Fig. 9A, bottom panels).

Figure 9.

Normal liver histology in hUGT1A1-AAV9-treated adult mice. A) Gallbladders of P5 WT and P5 untreated mutant mice (top panels) and adult gallbladders of WT and AAV-treated mutant mice (bottom panels). B) Trichromic staining of liver sections from WT and hUGT1A1-AAV9 treated mutant mice. Scale bar = 100 μm.

AAV9 gene therapy prevents bilirubin-induced neurological damage in the cerebellum of mutant mice

Since mutant pups showed important neurological abnormalities due to bilirubin neurotoxicity, we determined whether gene therapy was enough to prevent that damage. To this aim, we performed a detailed histological analysis of the same mutant mice analyzed above, which were killed at 5 mo after injection of the hUGT1A1 AAV9 vector. Histological analysis of the cerebellum showed that mutant mice treated with AAV9 CMV-hUGT1A1 had cerebellar architecture and layer stratification indistinguishable from that of wild-type littermates (Fig. 10). Moreover, Purkinje cell count and dendritic arbor were comparable in wild-type and AAV-treated mutant cerebellar sections (Fig. 11), supporting the functional data obtained with the rotarod test, and confirming the successful protection from bilirubin-induced neurological damage.

Figure 10.

AAV9 gene therapy prevents bilirubin-induced neurological damage in the cerebellum of treated mutant mice. A) Representative Nissl staining of 5-mo-old WT and AAV-treated mutant cerebellum slices. Boxed areas indicate fields shown in B. B) High-magnification image of cerebellar layers from WT and AAV-treated mutant mice. C) Layer depth of 5-mo-old WT and mutant + AAV9 cerebella (n=3/genotype). Error bars = sd. Scale bars = 1 mm (A); 100 μm (B).

Figure 11.

hUGT1A1-AAV9 gene therapy prevents bilirubin-induced neurological damage in Purkinje cells of AAV-treated mutant mice. A) Fluorescent immunohistochemistry of 5-mo-old WT and AAV-treated mutant cerebella using Hoechst nucleus staining (blue) and anti-calbindin antibody (red). Left panels: low-magnification images of total cerebellum. Middle and right panels: high-magnification images of fields 1 and 2 (boxed areas in left panels). B) Quantification of Purkinje cell number (cells/mm) in WT and AAV-treated mutant cerebella (n=3/genotype). Error bars = sd. C) High-magnification images of Purkinje cells in WT and AAV9-treated mutant cerebella. Scale bars = 1 mm (A, left panels); 400 μm (A, right panels); 50 μm (C).

DISCUSSION

CNI is a rare genetic disorder with no permanent cure aside from liver transplantation. We describe here a novel mouse model of the CNI resembling the human pathology from both genetic and pathological points of view, and a therapeutic approach, based on AAV gene therapy vectors, rescuing the drastic effects of the disease.

We generated a mouse model carrying a 1-base deletion in Ugt1 exon 4 that produces an in-frame stop codon immediately downstream, generating a truncated and inactive form of the protein (17). This mutation, located in a common exon of the Ugt1 cluster, is identical to that present in Gunn rats (18) and similar to many of those found in patients with CNI patients (1, 2). This resemblance was fully accomplished after the deletion of the positive selection marker present inside the Ugt1 gene by means of the Flp recombinase, avoiding any type of secondary effects in the Ugt1 cluster or nearby genes due to the presence of the NeoR cassette in the locus, as already observed in other targeted mutations (19, 20). As in untreated patients with CNI, the lack of UGT1A1 activity invariably resulted in death from bilirubin neurotoxicity (3). The observation that full rescue is achieved by adding back only the UGT1A1 gene strongly suggests that the mortality of mutant pups is directly caused by bilirubin toxicity and is not related to other metabolic defects linked to the knockout of the other genes of the Ugt1 cluster.

For >60 yr, a rat model of the disease (the Gunn rat; ref. 21) has been intensively used both to study the disease and to develop possible therapeutic approaches (22). Homozygous mutant Gunn rats are unable to conjugate bilirubin and experience lifelong hyperbilirubinemia. However, this condition in Gunn rats leads to a milder phenotype compared with humans affected by CNI. In fact, Gunn rats survive and reach adulthood without any treatment, and they may only present cerebellar hypoplasia, hearing impairment, and minor behavioral defects (22), which disappear after a short period of phototherapy treatment (23). Despite these important differences with the human disease, this animal model has been highly informative to better understand the biology of CNI and validate different therapeutic approaches. Phototherapy substantially lowered plasma bilirubin in our mutant mice, but it was insufficient to extend life span beyond 3 wk, underscoring the higher severity of the disease in mice. Notwithstanding the presence of an identical mutation, the phenotype observed in our mice was much more severe than that of Gunn rats. Most likely, differences in the severity of the disease have to be attributed to species-specific gene variations, since it is well known that genetic background dramatically influences penetrance of targeted mutations (24).

A mouse knockout strain with a complete disruption of the Ugt1 gene cluster by the insertion of the Neo cassette has recently been reported (8) showing, similarly to our results, that homozygous mutant mice die within 2 wk after birth. Neither the neurological characterization of the mutant strain nor a therapeutic rescue of the phenotype was performed in that strain (8, 25).

As observed in patients with CNI, the lack of Ugt1a1 activity in mice invariably resulted in death from bilirubin encephalopathy, a condition also known as kernicterus (26). In humans affected by kernicterus, a selective yellow staining of discrete regions of the brain is observed (27), while in Ugt1 mutant mice, the accumulation of UCB into the brain was evident as a diffuse intense yellow-orange staining of the tissue. Surprisingly, rough histological analysis of the brain showed no major abnormalities in tissue architecture, except for those described in the cerebellum. A more detailed analysis of the brain showed that the cerebellum was particularly affected. Mutant mice at P5 showed cerebellar hypoplasia and Purkinje cell abnormalities similar to those observed in Gunn rats (28–30). Because of its postnatal development, the cerebellum is particularly vulnerable to both developmental and environmental insults, such as UCB toxicity (31). Hence, in our model, the observed cerebellar abnormalities are another important readout to monitor the effectiveness of the therapeutic protocols. In fact, to prove the efficacy of the gene therapy approach, we performed histological and functional analyses, which were lacking in all gene therapy studies performed so far in Gunn rats (6).

The gene therapy approach is the more promising therapy that may overcome liver transplantation, which is to date the only available definitive cure for CNI. Due to the absence of pathogenicity, mild immune response, and long duration of viral genome persistence, AAV vectors appear particularly appropriate for this purpose. Moreover, neonatal gene transfer is potentially advantageous for preventing early manifestations of genetic disease and for providing efficient gene expression at relatively low doses. In addition, since the neonatal immune system is still immature, gene therapy during this period may induce tolerance to the transgene products (32), especially in mice, where the activation of the endogenous Ugt1a1 gene also occurs at P0 (Supplemental Fig. S2).

Seppen et al. (33) performed a pivotal study in adult Gunn rats in which the researchers compared different AAV serotypes (1, 2, 6, and 8) for the correction of Ugt1a1 deficiency. They demonstrated that AAV1 (CMV-UGT1A1) is the most efficient in lowering the serum bilirubin levels (64% reduction) after 1 yr postinjection, followed by AAV8, AAV 6, and AAV 2. Moreover Flageul et al. (34) reported that injecting neonatal Gunn rats with a single dose of AAV8 CMV-UGT1A1 resulted in a transient normalization of the hyperbilirubinemia after gene delivery. In this respect, the AAV9 serotype was reported to systemically transduce various tissues in adult mice, such as liver, heart, muscle, and pancreas, in a dose-dependant manner (14). Here we combined a neonatal gene transfer approach using a single intraperitoneal injection of CMV-hUGT1A1 with AAV9 serotype. The results showed that the gene therapy protocol was successful and that mutant mice survived and reached adulthood without any sign of neurological damage, as demonstrated by behavioral and brain histological analysis. These results are very promising, especially since we showed that the neonatal expression of the transgene leads to survival of all mutant pups, with an important reduction in plasma bilirubin levels and normal cerebellar development and function. Plasma bilirubin levels at 5 mo after injection were higher than those of wild-type controls, but at levels compatible with life, and far from the risk of developing brain damage in adults.

Analysis of viral persistence and therapeutic gene expression revealed that, after i.p. injection, vector DNA could be found in several organs but was mainly expressed in the skeletal muscle, thus clearly indicating that either most of the viral DNA in the other tissues did not reach the nucleus in a transcription-proficient double-stranded form, or expression was shut off (as is known to occur for CMV promoter-driven transgenes in the liver; ref. 35). Alternatively, liver VGPs may have been diluted due to hepatocyte duplication during growth.

Successful transient UGT1A1 expression in extrahepatic tissues and reduction of bilirubin levels were already obtained in Gunn rats (36). Although skeletal muscle is not the natural organ for bilirubin glucuronidation, nonhepatic conjugated bilirubin is expected to clear through the liver and be excreted in bile, as reported in Gunn rats (37) and observed as a yellow color of the gallbladder in our AAV-treated mutant mice. Even if the expression in muscle is apparently effective, this is not the natural pathway of bilirubin detoxification, and further studies may be necessary to exclude unexpected and undesired secondary effects in the long term.

To summarize, we have presented here a new mouse model of CNI, closely mimicking genetic and phenotypic features of the human disease, and shown for the first time, the full recovery from the lethal effects of neonatal hyperbilirubinemia. We showed that expression of the protein in the muscle is able to correct the drastic effects of the disease. This strategy seems easier than the transduction of the liver, paving the way toward a more effective therapy for the CNI syndrome. In addition, given that our mutant mice have a single base deletion as often observed in patients, they could represent a useful model to set up and test novel gene therapy approaches, such as in vivo gene correction by homologous recombination (38), stem cells, or induced pluripotent stem cells from patients.