Argininosuccinic aciduria fosters neuronal nitrosative stress reversed by Asl gene transfer

Argininosuccinate lyase (ASL) belongs to the liver-based urea cycle detoxifying ammonia, and the citrulline-nitric oxide cycle synthesising nitric oxide (NO). ASL-deficient patients present argininosuccinic aciduria characterised by hyperammonaemia and a multi-organ disease with neurocognitive impairment. Current therapeutic guidelines aim to control ammonaemia without considering the systemic NO imbalance. Here, we observed a neuronal disease with oxidative/nitrosative stress in ASL-deficient mouse brains. A single systemic injection of gene therapy mediated by an adeno-associated viral vector serotype 8 (AAV8) in adult or neonatal mice demonstrated the long-term correction of the urea cycle and the citrulline-NO cycle in the brain, respectively. The neuronal disease persisted if ammonaemia only was normalised but was dramatically reduced after correction of both ammonaemia and neuronal ASL activity. This was correlated with behavioural improvement and a decrease of the cortical cell death rate. Thus, the cerebral disease in argininosuccinic aciduria involves neuronal oxidative/nitrosative stress not mediated by hyperammonaemia, which is reversed by AAV gene transfer targeting the brain and the liver, acting on two different metabolic pathways via a single vector delivered systemically. This approach provides new hope for hepatocerebral metabolic diseases.


MAIN TEXT
Adeno-associated virus (AAV) vector mediated gene therapy has achieved promising results in recent clinical trials in liver 1 and neurodegenerative 2 inherited diseases, and led to the market approval of the first gene therapy product in the western world to treat lipoprotein lipase deficiency 3 . These successes underpin the current interest in this technology as illustrated by the rapidly expanding number of gene therapy based clinical trials 4 . Among various AAV capsid variants, AAV8 has demonstrated its efficacy in liver transduction in preclinical 5 and clinical studies 6 . This serotype also efficiently transduces other tissues including the central nervous system after systemic injection in neonatal mice 7 .
As with many liver inherited metabolic diseases, urea cycle defects exhibit a high rate of mortality and neurological morbidity in infancy despite conventional treatment 8 . Successful correction of the urea cycle via AAV-mediated gene therapy has been reported in mouse models of ornithine transcarbamylase deficiency 9,10 , argininosuccinate synthetase deficiency 11,12 , and arginase deficiency 13 . Argininosuccinic aciduria (ASA; OMIM 207900) is the second most common urea cycle defect with a prevalence of 1/218,000 live births 14 . In addition, ASA is an inherited condition proven to cause systemic nitric oxide (NO) deficiency 15 as the disease is caused by mutations in argininosuccinate lyase (ASL), an enzyme involved in two metabolic pathways: i) the liver based urea cycle which detoxifies ammonia, a highly neurotoxic compound generated by protein catabolism and ii) the citrulline-NO cycle, present in most organs, producing NO from L-arginine via nitric oxide synthase (NOS) (Supplementary Fig. 1) 16 . Patients may exhibit an early-onset phenotype with hyperammonaemic coma in the first 28 days of life, or a late-onset phenotype with either acute hyperammonaemia or a chronic phenotype with neurocognitive impairment and progressive liver disease 17 . Compared to other urea cycle defects, ASA patients present with an unusually high rate of neurological and systemic complications 17 contrasting with a lower rate of hyperammonaemic episodes. Various pathophysiological mechanisms have been hypothesised to account for this paradox, including impaired NO metabolism 18 . A hypomorphic Asl Neo/Neo mouse model shows impairment of both urea and citrulline-NO cycles and reproduces the clinical phenotype with impaired growth, multi-organ disease, hyperammonaemia and early death 15 . Common biomarkers of ASA include increased ammonaemia, citrullinaemia, plasma argininosuccinic acid, orotic aciduria and reduced argininaemia 18 .
In this study, we have characterised the neuropathophysiology of the disease studying the brain of the hyperammonaemic Asl Neo/Neo mouse and have used a systemic AAV-mediated gene therapy approach as a proof-of-concept study to rescue survival and protect the ASLdeficient brain from both hyperammonaemia and cerebral impaired NO metabolism. To achieve this, we designed a single-stranded AAV8 vector carrying the murine Asl (mAsl) gene under transcriptional control of a ubiquitous promoter, the short version of the elongation factor 1 α (EFS) promoter. The vector was administered systemically to adult and neonatal Asl Neo/Neo mouse cohorts.

RESULTS
Pathophysiology of the brain disease in ASA 6 ASL deficiency causes a systemic NO deficiency due to the loss of a protein complex that facilitates channeling of exogenous L-arginine to NOS 15 . To explore the effect on cerebral NO metabolism, various surrogate biomarkers were investigated. NO concentrations from wild type (WT) and Asl Neo/Neo mice were evaluated by measurement of nitrite (NO 2 -) and nitrate (NO 3 -) ions, downstream metabolites of NO 19 , and were found to be significantly increased in Asl Neo/Neo mice in brain homogenates (Fig. 1A). Similarly, cyclic guanosine monophosphate (cGMP), .a signalling pathway physiologically upregulated by NO generated by coupled NOS 20 , when measured in brain homogenates, was also found to be increased in Asl Neo/Neo mice (Fig. 1B). Low tissue L-arginine is a consequence of ASL deficiency downstream the metabolic block and can cause NOS uncoupling 21 , which leads to the production of reactive oxygen species including superoxide ion (O 2 -) or peroxynitrite (ONOO -) with the latter nitrating specific tyrosine residues and generating nitrotyrosine, a marker of oxidative/nitrosative stress 22 . This process can modify the protein structure and function, altering enzymatic activity or triggering immune response 23 . The detoxification of peroxynitrite by reduced glutathione (GSH) can generate nitrite via the reaction ONOO -+ 2GSH → NO 2 -+ GSSG + H 2 O 24 . Analysis of glutathione concentrations in in brain homogenates of Asl Neo/Neo mice showed that they were reduced although this did not reach statistical significance (Fig. 1C). In the cortex of WT and Asl Neo/Neo mice, immunostaining against nitrotyrosine was increased in Asl Neo/Neo mice in cells identified morphologically as neurons (Fig. 1D). Staining of glial fibrillary acidic protein (GFAP) and CD68, markers of astrocytic and microglial activation, respectively, did not show any difference ( Fig. 1E and   1F, respectively). Immunohistochemistry against neuronal NOS (nNOS or NOS1), inducible NOS (iNOS or NOS2) and endothelial NOS (eNOS or NOS3) showed an increased staining of all three enzymes in Asl Neo/Neo mice in cells morphologically identified as neurons (nNOS and iNOS) and endothelial cells (eNOS) (Fig. 1G, 1H and 1I, respectively). The brain morphology did not differ between WT and Asl Neo/Neo mice ( Supplementary Fig. 2).
Collectively these data suggest that a neuronal oxidative/nitrosative stress plays a role in the neuropathology of ASA. However hyperammonaemia per se can cause brain toxicity through oxidative stress 25 . To investigate whether neuronal oxidative/nitrosative stress is a primary mechanism involved in the phenotype of patients with ASA or is secondary to hyperammonaemia, we designed a gene therapy approach to normalise ammonaemia and conditionally target neuronal ASL activity.

AAV8.EFS.GFP vector targets liver and cerebral neurons
In order to extend survival and ameliorate the brain phenotype, we designed a vector that was not only able to transduce the liver to correct the defective urea cycle but also the brain, especially neurons. Neonatal C57Bl/6 mice received an intravenous injection of a single stranded AAV8.EFS.GFP vector (3.4x10 11 vector genomes/pup) and were culled at 5 weeks of life alongside uninjected control littermates. Fluorescence microscopy revealed green fluorescent protein (GFP) expression in the liver and the brain ( Fig. 2A). Anti-GFP immunostaining confirmed a high rate of hepatocyte transduction across the hepatic lobule ( Fig. 2B). Anti-GFP ELISA showed the liver as the main peripheral organ transduced ( Supplementary Fig. 3A, 3B). Anti-GFP brain immunostaining showed a clear pattern of neuronal transduction (Fig. 2C, 2D) throughout, most prominently in the cortex and decreasing rostro-caudally ( Supplementary Fig. 3C). Survival was improved significantly in both adult-and neonatally-injected groups. Sustained growth improvement was observed in adult-injected mice (Fig. 3C) with a peak of growth velocity in the 2 weeks following the injection of gene therapy ( Supplementary Fig. 4B). In neonatally-injected mice, a significant improvement of growth was transiently observed until day 30 (Fig. 3D) consistent with a growth speed similar to WT animals until day 15 ( Supplementary Fig. 4B). Later in follow-up, no significant difference of weight was observed between the surviving untreated and neonatally-treated Asl Neo/Neo mice (Fig. 3E).
A specific fur pattern with sparse, brittle hair called trichorrhexis nodosa was observed in untreated Asl Neo/Neo mice, mimicking symptoms observed in ASA patients 26   Urinary orotic acid levels were increased significantly in Asl Neo/Neo mice at 10 weeks of age compared with WT mice. Orotic acid concentration was normalised in two adult-injected mice at 10 weeks however it did not reach statistical significance neither in the adult-nor the neonatally-treated groups ( Supplementary Fig. 6D). Plasma alanine amino transferase (ALT) levels were normalised in both adult-and neonatally-injected mice ( Supplementary Fig. 6E).
Haematoxylin and eosin (H&E) staining of liver samples showed vacuolated cytoplasm in untreated Asl Neo/Neo mice; cytoplasmic glycogen deposits were identified by periodic acid Schiff (PAS) staining. This feature was markedly improved following adult, but not neonatal injections ( Supplementary Fig. 7).

Long-term improvement of the NO metabolism in the liver
Liver NO levels assessed by nitrite/nitrate levels were reduced in untreated Asl Neo/Neo mice.
Reduced liver glutathione was decreased in untreated Asl Neo/Neo mice (Supplementary Fig.   8B).

Long-term correction of the cerebral NO metabolism in neonatally-but not adult-treated Asl Neo/Neo mice
Cortical ASL enzyme activity in untreated Asl Neo/Neo mice was 14.1% of WT activity. In mice injected as adults, this activity was unchanged (16.2% of WT activity) but increased dramatically in mice injected neonatally with 64.8% of WT activity being evident at time of culling ( Fig. 5A).
To assess the effect of the improved ASL activity on the NO metabolism in brains of neonatally-treated Asl Neo/Neo mice, we measured nitrite/nitrate levels. Compared to WT brains, nitrite/nitrate levels were increased in untreated Asl Neo/Neo mice and in adult-injected mice by 3.4 and 2.5 times, respectively, whereas in neonatally-injected Asl Neo/Neo mice the levels were not significantly different from WT mice (Fig. 5B). To examine if this decrease in nitrite/nitrate levels in neonatally-treated mice was correlated with a modification of the oxidative/nitrosative stress, we quantified cortical nitrotyrosine staining. There was no significant difference between neonatally-injected mice and WT mice. In contrast, adult injected mice and untreated Asl Neo/Neo mice showed a significant increase in the percentage of immunoreactivity (Fig. 5C, 5D).

Impact of gene therapy on behaviour and cell death in the brain
Behavioural tests were performed to assess open field exploration. At 3 months of age, there was a significant reduction in the walking distance measured in the untreated Asl Neo/Neo mice, whereas an improvement was seen in both adult-and neonatally-injected groups (Fig. 6A).
Performance with an accelerating rotarod at the same age was significantly reduced in untreated Asl Neo/Neo mice but not significantly different from WT in both adult-and neonatallyinjected groups (Fig. 6B).
Cell death was assessed by TUNEL staining and was found to be significantly increased in the cortex of untreated Asl Neo/Neo mice compared to WT. Cell death was reduced in adultinjected and neonatally-injected mice compared to untreated Asl Neo/Neo mice (Fig. 6C, 6D).

DISCUSSION
This study provides new insight into the pathophysiology of the brain disease in ASA and presents proof-of-concept of hepatocerebral phenotypic correction of the hypomorph Asl Neo/Neo mouse model via systemic AAV-mediated gene therapy.
Compared to other urea cycle defects, the neurological disease in ASA is a paradox between low rates of hyperammonaemic decompensation and high rates of neurological complications, including neurocognitive delay, abnormal neuroimaging, epilepsy and ataxia 18 . The Asl Neo/Neo mouse model was used to study its unreported neurological phenotype and investigate neuropathophysiology. NO plays a complex and ambiguous role in the brain, involved in both inflammation-related neurotoxicity and cGMP-mediated neuroprotection 27,28 . This is consistent with our observation of increase of both neuronal nitrotyrosine staining and nitrite/nitrate levels. In the brain, hyperammonaemia increases nNOS and iNOS-mediated NO synthesis via an increase in extracellular glutamate and oxidative stress 25 . In our study, correction of hyperammonaemia only did not modify the oxidative/nitrosative stress, suggestive of an independent brain-specific causative mechanism. A measurable benefit in cell death in the cortex is observed when neuronal ASL activity is restored. Neurons are more vulnerable to oxidative stress than astrocytes in vitro, as they cannot overexpress γ-glutamyl transpeptidase to replenish their intracellular glutathione content 30 . Therefore, they rely on the paracrine glutathione supply from astrocytes when exposed to reactive nitrogen species and oxidative stress 30 . This might explain the neuronal staining observed for nitrotyrosine and the efficacy of neuronal-targeted gene therapy. This nitrosative stress caused by peroxynitrite, generating nitrotyrosine, has been implicated previously in the pathophysiology of various neurodegenerative diseases: Parkinson's disease 31 , Alzheimer's disease 32,33 and amyotrophic lateral sclerosis 34 . In urea cycle defects, a neuronal disease caused by oxidative stress as a consequence of low tissue arginine has been hypothesised as playing a role in the brain pathophysiology 35 . Although the precise biochemical mechanisms regulating NO metabolism in different cerebral cell types in ASA remain elusive, NOS coupling and uncoupling appear to co-exist in the brain accounting for both the physiological NO-cGMP pathway and nitrosative/oxidative stress, respectively (Fig. 7) as observed in Alzheimer's disease 27 . No evidence of neuroinflammation was observed in astrocytes and microglial cells, when GFAP and CD68 were assessed by immunohistochemistry suggesting that these cell types are not primarily involved. Thus neuronal oxidative/nitrosative stress seems to play a key-role in the ASA brain disease.
In murine animal models, AAV8 is known for its ability to widely transduce the brain after intracranial administration 36,37 . After systemic injection, most of the organs and especially the liver are successfully targeted 38 but the neurotropism is influenced by the age of infusion and the dose of vector administered. For instance, successful brain transduction with AAV8 and a CMV promoter after systemic delivery has been reported previously until day 14 of life but not later (1.5x10 11 vg/mouse) 39 . However brain transduction was barely detectable in adult mice after a similar experiment (1x10 11 vg/mouse) 38 . Increasing the dose of vector improved brain transduction in adult mice. Indeed intravenous injection in adult mice with AAV8 and EF1α promoter showed mild brain cell transduction at 1x10 11 to 2x10 12 vg/mouse 40,41 , but widespread neuronal and astrocytic transduction at 7.2x10 12 vg/mouse (approximately 2.9x10 14 vg/kg) 42 . The transient ability for AAV vectors to cross an immature blood brain barrier in the neonatal mouse brain is not well understood and could be due to immaturity or receptor-mediated transcytosis 39,43 . The age at injection does not only allow an increased  11,13 . In this study, an increase from 14.5% to 18% of WT liver ASL activity was observed 9 months after neonatal injection. This provided a persistent correction of ammonaemia but was not sufficient to normalise other biochemical parameters of the disease (plasma amino acids, orotic aciduria). In ASA, the increased urine secretion of the argininosuccinic acid that removes two nitrogen moieties may explain reduced tendency to develop hyperammonaemic episodes compared to proximal urea cycle defects 16 . AAVmediated correction of other models of urea cycle disorders has shown that a small (approximately 3%) improvement in liver enzyme levels and ureagenesis can restore survival and improve ammonia levels 55,56 . Controlling orotic aciduria in the Spf ash mouse model of ornithine transcarbamylase deficiency however required 5 times more vector compared to that necessary to normalise ammonaemia 9,57 . In that respect, our study provides a hierarchization in the biomarkers significance according to the ASL residual activity in ASA. Plasma amino acids and urine orotic acid required a liver ASL activity of >18% for normalization whereas ammonaemia was seen to normalize when ASL activity was only 14.5-18% of WT activity.
However these figures might be biased by the persistence of the non-integrating transgene delivered by AAV vector. As reported previously in shRNA-induced hyperammonaemic Spf ash mice 9 , the AAV-encoded enzymatic activity required to normalize ammonaemia might be higher than the endogenous residual activity required in a non-hyperammonaemic subject.
Indeed the reduced transgenic expression from non-integrated episomes compared to endogenous chromosomal alleles has been suspected recently from results of a liver-directed clinical trial 58 . The fur phenotype observed in ASL-and argininosuccinate synthase-deficient mice is likely to be caused by hypoargininaemia as arginine represents up to 10% of hair composition 17 . The long-term phenotypic improvement of the fur in adult-injected mice is consistent with the improved plasma arginine levels.
A small but significant increase of the nitrite/nitrate levels was observed in the livers of adultinjected mice thereby suggesting that restoration of ASL activity had a positive effect on the function of both urea and citrulline-NO cycles. A decrease in reduced glutathione in untreated Asl Neo/Neo mice reflects increased cellular levels of oxidative stress. Whilst systemic NO deficiency 15 and increased oxidative stress 21 have been described previously, it was shown that long-term correction of neuronal ASL achieved a marked decrease of the cortical oxidative/nitrosative stress, independent of improvement in ammonia levels.
ASA patients are at high risk of developing neurological complications even if hyperammonaemic episodes do not occur 18 . Our data provide further evidence for a neuronal disease with oxidative/nitrosative stress independent of ammonaemia, and illustrates the pathophysiological importance of disturbed NO metabolism in the ASA brain. Any therapy aiming to preserve the neurological status of ASA patients needs to protect the brain from two potential insults, hyperammonaemia and disturbed cerebral NO metabolism. Similar to liver transplantation 59 , a gene therapy approach targeting only hepatocytes will cure the urea cycle defect but will not correct the symptoms related to ASL deficiency in extra-hepatic tissues, especially the brain 21 . This study provides proof-of-concept for phenotypic correction of the Asl Neo/Neo mouse model using AAV technology. These promising results raise the possibility of combining two sequential systemic injections: i) a first early (neonatal) injection of a gene therapy vector that would transiently restore the urea cycle in the liver and will transduce neurons to modify the long-term natural course of the neuronal disease, and ii) a second injection in infancy or adulthood targeting the liver for long-term correction of the urea cycle.
The potential for humoral immune response generated by the first AAV injection will need to be considered for the second injection 60 . It is possible however, similarly to what is reported for neonatal rodents 61 , that the immaturity of the immune system in humans at the time of neonatal injection might prevent the induction or diminish the magnitude of humoral response against the AAV capsid 62 . An alternative AAV capsid, which does not cross-react with neutralising antibodies developed, might be a valid option 63,64 . Several inherited metabolic diseases with hepatocerebral phenotype might benefit from a similar dual targeting approach such as mitochondrial diseases caused by nuclear genetic defects (e.g. POLG1, MPV17, DGUOK genes) and some lysosomal storage disorders (e.g. neuronopathic Gaucher disease, mucopolysaccharidosis type I and II). Depending on the pathophysiology of the disease, specific brain cell-types can be selectively targeted in modifying either promoter and/or age at injection 36,37 .   Analysis of ASL enzyme activity. Liver and brain samples were snap-frozen in dry ice at time of collection after perfusion of the animal with PBS to remove residual blood in tissues.

Animals. The
Protein extraction was performed on ice. Samples were homogenised in lysis buffer (50 mM Tris, 150 mM NaCl, 1% Triton adjusted to pH 7.5) and centrifuged at 16,000 g for 20 min at 4°C. Protein quantification of the supernatant was performed using the Pierce TM BCA protein assay kit (ThermoFisher Scientific, Rockford, IL, USA) according to manufacturer's instructions.
Liver ASL activity was measured in duplicate samples. 20µg protein was added to a buffer solution of 20 mM Tris, 1 mM argininosuccinic acid, 0.02 nM L-citrulline-d7 and incubated for 2 hours at 37°C. Brain ASL activity was also measured in duplicate samples. 80µg protein was added to a buffer solution of 20 mM Tris, 30 µM argininosuccinic acid, 0.02 nM Lcitrulline-d7 and incubated for 2 hours at 37°C. A 4:1 volume of methanol was added to stop the reaction, and centrifuged at 9,500 g for 2 minutes. The supernatant was analysed by the LC-MS/MS method described above. ASL activity was calculated by subtracting the amount of argininosuccinic acid post-incubation from that pre-incubation.
Nitrite and nitrate measurement. Measurement of nitrite and nitrate levels were performed using a modified Griess reaction protocol 69 . Samples were collected carefully in order to minimise the risk of nitrite and nitrate contamination. All glassware and plastic ware were cleaned with double-distilled water (ddH 2 O). Animals were anaesthetised and perfused on ice.
Organs and brain were collected on ice in less than 3 minutes and snap frozen on dry ice. Glutathione analysis. Reduced glutathione was measured as described previously 70