EBioMedicine 62 (2020) 103133 Contents lists available at ScienceDirect EBioMedicine journal homepage: www.elsevier.com/locate/ebiom Research paper Efficacy of adeno-associated virus gene therapy in a MNGIE murine model enhanced by chronic exposure to nucleosides a, Raquel Cabrera-Pe Ferran Vila-Julia reza, Yolanda Ca maraa, Miguel Molina-Berenguera, Silvia Lope-Piedrafita , Michio Hirano , Federico Mingozzie, Javier Torres-Torronterasa,*, b,c d Ramon Martía,* a Research Group on Neuromuscular and Mitochondrial Diseases, Vall d’Hebron Research Institute, Universitat Auto noma de Barcelona, and Biomedical Network Research Centre on Rare Diseases (CIBERER), Instituto de Salud Carlos III, Barcelona, Catalonia, Spain b Servei de Ressonancia Magnetica Nuclear, Universitat Auto noma de Barcelona, Cerdanyola del Valles, Catalonia, Spain c n Biomedica en Red en Bioingeniería, Biomateriales y Nanomedicina (CIBER-BBN), Cerdanyola del Valles, Catalonia, Spain Centro de Investigacio d Department of Neurology, H. Houston Merritt Neuromuscular Research Center, Columbia University Irving Medical Center, New York, NY, United States e Spark Therapeutics, Philadelphia, PA 19104, United States A R T I C L E I N F O A B S T R A C T Article History: Background: Preclinical studies have shown that gene therapy is a feasible approach to treat mitochondrial Received 17 September 2020 neurogastrointestinal encephalomyopathy (MNGIE). However, the genetic murine model of the disease Revised 5 November 2020 (Tymp/Upp1 double knockout, dKO) has a limited functional phenotype beyond the metabolic imbalances, Accepted 5 November 2020 and so the studies showing efficacy of gene therapy have relied almost exclusively on demonstrating correc- Available online xxx tion of the biochemical phenotype. Chronic oral administration of thymidine (dThd) and deoxyuridine (dUrd) to dKO mice deteriorates the phenotype of the animals, providing a better model to test therapy Keywords: approaches. MNGIE Gene therapy Methods: dKO mice were treated with both dThd and dUrd in drinking water from weaning until the end of Nucleosides the study. At 8 - 11 weeks of age, mice were treated with several doses of adeno-associated virus (AAV) Mitochondrial disease serotype 8 vector carrying the human TYMP coding sequence under the control of different liver-specific pro- Thymidine phosphorylase moters (TBG, AAT, or HLP). The biochemical profile and functional phenotype were studied over the life of the animals. Findings: Nucleoside exposure resulted in 30-fold higher plasma nucleoside levels in dKO mice compared with non-exposed wild type mice. AAV-treatment provided elevated TP activity in liver and lowered sys- temic nucleoside levels in exposed dKO mice. Exposed dKO mice had enlarged brain ventricles (assessed by magnetic resonance imaging) and motor impairment (rotarod test); both were prevented by AAV treatment. Among all promoters tested, AAT showed the best efficacy. Interpretation: Our results show that AAV-mediated gene therapy restores the biochemical homeostasis in the murine model of MNGIE and, for the first time, demonstrate that this treatment improves the functional phenotype. Funding: This work was funded in part by the Spanish Instituto de Salud Carlos III, and the Generalitat de Cat- alunya. The disclosed funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. © 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) 1. Introduction systemic accumulation of its substrates, the nucleosides thymidine (dThd) and deoxyuridine (dUrd) [2,3]. dThd and dUrd accumulation Mitochondrial neurogastrointestinal encephalomyopathy causes an imbalance in the mitochondrial deoxyribonucleoside tri- (MNGIE) is caused by mutations in the nuclear gene TYMP [1], which phosphate (dNTP) pool that interferes with mitochondrial DNA lead to a deficiency of thymidine phosphorylase (TP) activity and (mtDNA) replication, thereby inducing mtDNA depletion, multiple deletions, and point mutations [4-8]. The main clinical features of MNGIE are progressive gastrointesti- * Corresponding authors. nal dysmotility, cachexia, ptosis, ophthalmoplegia or ophthalmopare- E-mail addresses: javier.torres@vhir.org (J. Torres-Torronteras), ramon.marti@vhir. org (R. Martí). sis, demyelinating peripheral neuropathy, symmetrical distal https://doi.org/10.1016/j.ebiom.2020.103133 2352-3964/© 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) 2  et al. / EBioMedicine 62 (2020) 103133 F. Vila-Julia transplant-related complications, which are aggravated in MNGIE Research in context patients by the poor clinical condition when most are treated [13]. More recently, orthotopic liver transplantation was proposed as an Evidence before this study alternative option [14], and the reported cases indicate that success- Mitochondrial neurogastrointestinal encephalomyopathy ful engraftment results in a sustained nucleoside reduction [15-18]. (MNGIE) is a fatal metabolic disease caused by mutations in the However, liver transplantation is also an aggressive intervention, and nuclear gene TYMP that encodes thymidine phosphorylase. In it is constrained by a limited supply of donors. patients, thymidine phosphorylase dysfunction causes accumu- Over the last few years, several preclinical studies have proven the lation of the nucleosides thymidine and deoxyuridine, which effectiveness of lentiviral hematopoietic stem cell gene therapy or interferes with mitochondrial DNA replication. Hence, all ther- liver-targeted adeno-associated virus (AAV) vector gene therapy to apy approaches for MNGIE are addressed to reach permanent correct MNGIE biochemical disarrangements in a Tymp/Upp1 double reduction of these toxic metabolites. Over the last years, several knockout (dKO) mouse model of the disease [5,19-24]. However, as preclinical studies have demonstrated that gene therapy is a the dKO mouse model has a limited and elusive functional phenotype feasible strategy for this disorder. Transduction of hematopoi- except for its clear metabolic imbalances, these preclinical studies etic progenitors with the correct version of the gene using lenti- have relied almost exclusively on demonstrating correction of the viral vectors, or the liver, using an adeno-associated virus (AAV) biochemical imbalances caused by the lack of TP activity. It has been vector, resulted in effective and sustained restoration of nucleo- shown that chronic oral administration of dThd and dUrd to dKO side homeostasis in the Tymp/Upp1 double knock out murine mice significantly enhances the phenotype beyond biochemical model of this disorder. However, this animal model does not derangements and provides a better model to test therapy recapitulate the clinical manifestations of the disease, and all approaches [25]. Taking advantage of this improved model, we show evidence so far collected supporting gene therapy for MNGIE here that the liver-targeted AAV gene therapy not only corrects the relies almost exclusively on demonstrating correction of the biochemical imbalances of the nucleoside-stressed model, but also biochemical phenotype in mice. prevents other functional changes such as those related to its neuro- logical phenotype. Added value of this study 2. Methods Previous studies demonstrated that challenging the MNGIE mouse model with a chronic oral administration of the toxic 2.1. Vector construction, production, and titration nucleosides exacerbated the model leading to the manifesta- tion of neurological phenotype and thus providing a better tool The human coding sequence of TYMP (hcTYMP) was cloned in to test therapeutic intervention. In this study, we show that three different vectors (AAV-TBG, AAV-AAT, and AAV-HLP) and pack- AAV-mediated gene therapy prevents the enlargement of brain aged in AAV8 serotype as previously described [19]. AAV-TBG con- ventricles and the neuromotor dysfunction observed in the tained hcTYMP under regulation of the thyroxine binding globulin mice. Hence, we prove that liver-targeted gene therapy using (TBG) promoter in a single-stranded configuration. AAV-AAT con- an AAV vector prevents the manifestation of a neurological tained hcTYMP under regulation of the a 1-antitrypsin promoter phenotype beyond correcting the biochemical nucleoside and the apolipoprotein E hepatic control region (ApoE-AAT cassette) imbalance. This is the first time liver-targeted AAV-mediated in a single-stranded configuration. AAV-HLP contained hcTYMP gene therapy for MNGIE shows efficacy on clinical parameters under regulation of the hybrid liver-specific promoter, which is com- in the animal model of the disease, which constitutes an impor- posed of an abbreviated form of the ApoE-AAT cassette [26] in a self- tant step towards the translation of this therapy to patients. complementary configuration. For construction of AAV-HLP, the pAV-HLP-Luc plasmid was provided by Dr. Nathwani’s group (Uni- Implications of all the available evidence versity College London Cancer Institute, London, United Kingdom) At present, hematopoietic stem cell transplantation and liver and the luciferase gene was replaced by the hcTYMP sequence. All transplantation are the only available therapies for MNGIE vectors were produced and titrated as indicated elsewhere [19]. patients. However, both treatments are constrained by the availability of a suitable donor and involve complex medical 2.2. Animal procedures interventions associated to life-threatening risks. Gene therapy may constitute a safer alternative, and this study provides evi- All animal procedures were conducted in accordance with the dence that liver targeted AAV-mediated gene therapy is effec- rules established by the Generalitat de Catalunya for the Care and tive on the neurological phenotype in the enhanced animal Use of Laboratory Animals and in compliance with the ARRIVE guide- model of the disease. This will support conducting clinical stud- lines. The protocols were approved by the Ethics Committee for Ani- ies that will bring this potential treatment closer to patients. mal Experimentation of the Vall d’Hebron Research Institute (Permit Number: 14 06). A double knockout mouse for the genes Tymp and Upp1 previ- ously described was used for this study [5]. AAV transduction effi- weakness, and diffuse leukoencephalopathy on brain magnetic reso- ciency in mice is higher in males than females [27]. Thus, to avoid nance imaging (MRI). Disease onset is usually between the first and sex-related variation in the results, only male animals were used in fifth decades, in most cases before the age of 20 years, and the condi- this study. tion progressively degenerates, causing premature death of affected For chronic oral nucleoside exposure (“exposed” mice), the ani- patients usually during the third to fourth decade of life [9]. mals’ drinking water was supplemented with 16.6 g/L of dThd and Several therapeutic strategies have been proposed for MNGIE, all dUrd (68.6 mM dThd and 72.3 mM dUrd) from weaning up to the of them addressed to reducing the toxic nucleoside accumulation end of the study (22-month-old mice) [25]. Mice that were not [10]. Allogeneic hematopoietic stem cell transplantation was the first exposed to nucleoside administration are designated as “naïve” in treatment providing a permanent nucleoside reduction and clinical this study. improvement or stabilization in patients [11,12]. However, it is an To study pharmacokinetics (PK) following acute nucleoside aggressive therapy associated with a high mortality risk due to administration (peak plasma concentration, area under the curve  et al. / EBioMedicine 62 (2020) 103133 F. Vila-Julia 3 [AUC], time to reach plasma concentration in naïve wild type [WT] Real-Time PCR System (Applied Biosystems). hcTYMP DNA copy animals), a single dose was administered by oral gavage (10 mL/g number was quantified using the predesigned TaqMan MGB expres- body weight of 16.6 g/L of dThd and dUrd aqueous solution, ie, sion assay Hs00157317_m1 and referred to the murine single-copy 166 mg dThd + 166 mg dUrd/kg body weight). Blood samples were nuclear gene Ang1 using the predesigned TaqMan MGB gene expres- collected immediately before nucleoside administration, and at sion assay Mm00833184_s1 as previously described [19]. Quantifica- 30 min, 1 h, 1.5 h, 2 h, 3 h, 5 h, and 8 h after administration. In the PK tions were based on a standard curve prepared with different study, mice were deprived of access to drinking water during the first dilutions of plasmids containing hcTYMP DNA or a specific region of 3 h after gavage. the Ang1 gene. To detect the influence of possible metabolic adaptation on PK in chronically dThd/dUrd-treated mice (exposed group), acute dThd/ 2.5. Motor performance dUrd administration was studied under two different conditions: mice without chronic oral dThd/dUrd treatment (naïve group), and Motor coordination, balance and exercise tolerance were assessed mice already receiving chronic dThd/dUrd treatment (exposed in 25-week-old mice by testing their time to fall in a rotarod appara- group). In this latter case, dThd/dUrd was withdrawn from drinking tus (Ugo Basile Mouse Rota-Rod). The protocol included three train- water 24 h before acute administration for the PK study. ing sessions of 2 min each at a constant 4-rpm speed prior to the test. Therapeutic vectors were administered to 8- to 11-week-old dKO The test session was then started at a speed of 4 rpm for 15 s, fol- mice in a single intravenous injection (tail vein) of AAV-TBG at doses lowed by a constant acceleration rate of 1 rpm/s for 45 s. Three test (in vector genomes per kg, vg/kg) of 5 £ 1011, 1012, 2 £ 1012 and sessions were carried out with each mouse and the three time-to-fall 1013, and AAV-AAT or AAV-HLP, both at doses of 2 £ 1012 and 1013. values were recorded and averaged [25]. Blood samples from live animals were collected from the saphe- nous vein using EDTA capillaries (Microvette 200K3E, Sarstedt). At the end of the study (age 22 months), mice were euthanized by cervi- 2.6. Magnetic resonance imaging cal dislocation. Heart blood was immediately collected in EDTA- treated tubes, and tissues (liver, brain, small intestine and gastrocne- Eighteen months-old mice were anesthetized with 4% (v/v) and mius skeletal muscle) were excised and immediately frozen in liquid maintained with 1.5% (v/v) isoflurane in O2 at 1 L/min. Mice were nitrogen and stored at 80 °C. Before analysis, mouse tissues were placed in prone position on a water-jacketed bed to maintain body ground to a fine powder with a ceramic mortar and pestle without temperature (37 § 1 °C) and immobilized with bite-bar and ear-bars. defrosting in liquid nitrogen. Stable respiration rate (50 80 bpm) and rectal temperature were monitored (SA Instruments, Stony Brook, NT). 1 2.3. TP activity and nucleoside determination H-MRI studies were performed at the joint nuclear magnetic resonance facility of the Universitat Auto noma de Barcelona and Plasma dThd and dUrd concentrations were determined by HPLC- Centro de Investigacio  n Biome dica en Red—Bioingeniería, Biomater- UV, as previously described [22]. For TP activity, 100 200 mg of pul- iales y Nanomedicina (CIBER-BBN) (Cerdanyola del Valle s, Catalo- verized tissue was homogenized with a plastic pestle in 800 mL of nia), Unit 25 of NANBIOSIS. Experiments were conducted on a 7T lysis buffer. Homogenates were centrifuged at 20,000 x g for 30 min Bruker BioSpec 70/30USR scanner (Bruker BioSpin GmbH, Ettlingen, at 4 °C, the supernatant protein concentration was measured [28], Germany) equipped with a 72-mm inner diameter linear volume and TP activity was determined as described elsewhere [29]. coil as a transmitter and a dedicated mouse brain surface coil as a For dThd and dUrd quantification in tissue, 100 200 mg of pul- receiver. MRI data were acquired and processed on a Linux com- verized samples was homogenized with a plastic pestle in lysis buffer puter using Paravision 5.1 software (Bruker BioSpin GmbH, Karls- with addition of 1 mM internal standard (isotope-labeled dThd: ruhe, Germany). 13 C1015N2-dThd) and TP activity inhibitors to prevent in vitro degra- Low-resolution T2-weighted fast spin-echo images were initially dation of nucleosides in homogenates: 100 mM of 5‑chloro-6-[1-(2- obtained in axial, sagittal, and coronal planes for use as reference iminopyrrolidinyl) methyl] uracil hydrochloride (strong TP-specific scout images. The following imaging parameters were applied: TE inhibitor that does not inhibit uridine phosphorylase (UP), used in (effective echo time) 36 ms; TR (repetition time) 2 s; ETL (echo train liver homogenates containing high TP activity derived from AAV length) 8; FOV (field of view) 1.92 £ 1.92 cm2 ; MTX (matrix size) transduction) or 100 mM 6-amino-5-bromouracil (TP and UP inhibi- 128 £ 128; ST (slice thickness) 1 mm; gap (gap between slices) tor) and 100 mM tetrahydrouridine (cytidine deaminase inhibitor), 0.1 mm; NS (number of slices) 14 axial, 9 sagittal, 5 coronal; NA (num- used in small intestine homogenates containing high TP and UP activ- ber of averages) 1; and Expt (experiment time) 24 s per orientation. ity) [30]. Homogenates were centrifuged at 20,000 x g for 30 min at T2 maps were acquired using a multi-spin multi-echo sequence. 4 °C, and protein concentration of the supernatant was measured T2-weighted images at 24 different TE values between 10 and [28]. The supernatant was deproteinized by ultrafiltration (10 kDa 240 ms were acquired, with TR 3 s, FOV 1.92 £ 1.92 cm2, MTX Amicon Ultra filters; Merck Millipore) at 14,000 x g for 30 min at 4 ° 128 £ 128, NS 8, ST 1 mm, and gap 0.1 mm, covering the whole brain C. The ultrafiltrate was later used to measure nucleosides by LC-MS/ excluding the olfactory bulb and cerebellum (Supp. Figure 6a), with MS (liquid chromatography coupled to tandem mass spectrometry). Expt 4 min 48 s. Components of the sample were resolved and detected as previously Voxel-wise T2 relaxation maps were calculated using Paravision described [23]. Serial dilutions of dThd and dUrd standards 5.1 software. Regions of interest corresponding to the total brain and (0.5 20 mM) in lysis buffer and 1 mM of internal standard were proc- excluding ventricular areas were manually outlined for each slice essed in parallel to generate calibration curves. dThd and dUrd con- and the mean T2 region values were calculated for each mouse. centrations in the samples were obtained from interpolation of the For ventricular volume measurement, brain volume was manually peak areas relative to the internal standard in the calibration curve. outlined in each slice. Within the selected brain region, ventricular pixels were automatically pre-defined on T2 maps as pixels having 2.4. Vector copy number determination T2 values above 60 ms. The analysis was reviewed and non-ventricu- lar regions were manually removed. The sum of ventricular pixels in DNA was isolated with phenol-chloroform from approximately all slices was then multiplied by the voxel volume 25 mg of frozen liver and dissolved in 10 mM Tris HCl, pH 8.0. Vector (0.15 £ 0.15 £ 1 mm3) to provide the total ventricular volume for genome copies per cell were quantified by qPCR on the 7900HT Fast each mouse. 4  et al. / EBioMedicine 62 (2020) 103133 F. Vila-Julia 2.7. Mitochondrial isolation and dNTP determination blood samples between 8 and 9 am (8- to 30-fold increase in exposed WT mice, and 13- to 26-fold increase in exposed dKO mice) (Fig. 1a Liver and brain mitochondria were isolated and dNTPs were and Supplementary Fig. 1). As mice were exposed to dThd and dUrd extracted as previously described [4,19]. For dNTP quantification, dry in drinking water ad libitum, it was not possible to determine the extracts were thawed and re-dissolved in 40 mL of 40 mM Tris HCl, time of the last intake before blood collection. This may account for pH 7.4. Mitochondrial dNTP content was determined using the previ- the wide, randomly distributed variation in plasma concentrations; ously described polymerase-based method [31] with some modifica- that is, time differences between the last water intake and sample tions [19]. collection. Thus, we decided to explore a more controlled set of parameters, 2.8. Mitochondrial DNA copy number analysis by studying nucleoside PK in 35 to 40 week-old mice after a single dThd/dUrd administration by oral gavage in naïve WT and dKO mice. Total DNA was isolated from approximately 25 mg of pulverized We also studied exposed WT and dKO mice to detect whether a pos- frozen tissue with the QIAamp DNA Mini Kit (Qiagen) and dissolved sible metabolic adaptation in response to chronic dThd and dUrd in 10 mM Tris HCl (pH 8). MtDNA copy number was quantified by exposure influenced PK. After the single oral dose, plasma concentra- qPCR on the 7900HT Fast Real Time PCR System (Applied Biosystems) tions rose above naïve dKO levels, reaching a peak value between 1 using a probe targeted to the mitochondrial gene mt-Rnr2 (encoding and 1.5 h after oral gavage (Fig. 1b and Supplementary Fig. 2a). There- the 16S rRNA) and referred to the nuclear gene Ang1, following the after, deoxyribonucleoside (dN) levels rapidly declined, even in dKO protocol detailed elsewhere [32]. mice, which lack the required catabolic enzymes, owing to urinary excretion (data not shown). Eight hours after oral administration, 2.9. Histochemical staining of small intestine dThd and dUrd plasma levels remained above normal plasma levels in all groups. Peak concentrations (Fig. 1c and Supplementary Fig. 2b) Longitudinal sections of jejune were analysed using the adapted and AUCs (Fig. 1d and Supplementary Fig. 2c) showed a clear ten- “Swiss roll” technique [33]. Jejune sections were fixed in 4% neutral dency to be higher in dKO mice than WT mice, in both the naïve and buffered formalin and embedded in paraffin for histological examina- exposed groups, respectively. We also found that AUC values were tion. Three-mm-thick slices were stained with Masson’s trichrome. higher in exposed dKO than in naïve dKO mice (Fig. 1d and Supple- Images were taken using a FSX100 fluorescence microscope and the mentary Fig. 2c). software FSX-BSW (Olympus, Tokyo, Japan). 3.2. AAV-TYMP treatment limits nucleoside stress in chronically dThd/ 2.10. Statistical analysis dUrd -exposed dKO mice Statistical analyses were performed using the GraphPad Prism 6 Exposed dKO mice were treated with three different liver-tar- program (GraphPad Software, Inc.). For multiple comparisons (multi- geted AAV vectors carrying hcTYMP (AAV-TBG, AAV-AAT, or AAV- ple comparisons with one reference group, and multiple compari- HLP) at different doses. Plasma dThd and dUrd monitoring over 35 sons of all groups compared with each other) the Kruskal-Wallis test weeks showed the following: although AAV treatment failed to followed by the Dunn's multiple comparisons test was used. When reduce circulating levels of these nucleosides at most time points, the the only interest was to compare two groups defined by only one frequency of randomly distributed samples in which nucleosides con- variable (e.g., exposed group vs. non-exposed group), we used the centrations decreased to levels within the range of naïve dKO mice Mann-Whitney U test. In any case, tests used are always indicated in (Supplementary Fig. 3) was higher than that of untreated exposed the figure legends. For statistical purposes, undetectable values were dKO mice (Fig. 1a). As the last nucleoside intake (in drinking water) considered as zero. Nucleoside PK parameters, including peak before blood sample collection might have been interfering with plasma concentration, AUC, and time to reach plasma WT concentra- detection of the effect of AAV treatment on dN concentration, we tion, were calculated from the graphical curves obtained by plotting analyzed the effect of AAV treatment on PK parameters in gavage- nucleoside plasma concentrations vs time, defined by connecting exposed dKO mice (Fig. 1b-e). Before gavage dosing, plasma dThd adjacent measurements with straight lines, and taking as baseline and dUrd concentrations in all AAV-treated mice but one were in the the mean plasma nucleoside concentration at time 0 h in naïve WT WT range or below (dThd, 1.9 3.5 mM; dUrd, 2.5 4.2 mM). It should mice. be noted that the group of exposed mice had been deprived of oral dThd and dUrd in drinking water 24 h before PK assessment. AAV 2.11. Role of funding source treatment reduced peak dThd concentrations to well below the peak concentrations observed in exposed dKO mice, regardless of the vec- The disclosed funders had no role in study design, data collection tor construct or dose used, although statistical significance was not and analysis, interpretation, decision to publish, or preparation of the reached for the two AAV-TBG lowest doses (Fig. 1c). AAV treatment manuscript. led to even more pronounced reductions in the AUC (Fig. 1d). In con- trast to what occurred in AAV-untreated mice, which did not reach 3. Results WT nucleoside levels after gavage nucleoside administration over 8 h of monitoring (Fig. 1b), all AAV-treated exposed dKO mice reached 3.1. Pharmacokinetics of dThd and dUrd in the murine model WT levels in less than 8 h (Fig. 1e). A dose-dependent effect in the time reduction to reach WT dThd concentration was observed with In agreement with previous reports [19-24], in the absence of oral all vectors; a similar, but less marked dose-dependent effect was nucleoside administration (naïve mice), circulating dThd and dUrd observed on peak dThd concentration and dThd AUC. Comparison concentrations were around 3-fold higher in dKO mice than in WT between the different vectors showed that AAV-AAT had the most mice (Fig. 1a and Supplementary Fig. 1). A marked increase in plasma pronounced effect on dThd PKs. dThd and dUrd concentrations above normal WT levels was observed The greatest effect was achieved with AAV-AAT at 1013 vg/kg, in the exposed groups. When the time of blood sample collection was which reduced plasma dThd concentration to <3.5 mM in a not considered, plasma dThd and dUrd concentrations showed wide mean of 2.2 h (range 1.8 2.6 h) after oral gavage. Mice treated with dispersion in the exposed groups (data not shown). This variability 1013 vg/kg of AAV-AAT also showed the lowest mean dThd peak was greatly reduced after changing the protocol and collecting all concentration (13.8 mM, range 5.5 16.9) and the lowest mean  et al. / EBioMedicine 62 (2020) 103133 F. Vila-Julia 5 Fig. 1. Effect of AAV treatment on dThd pharmacokinetics. a) Plasma dThd concentration in WT and dKO mice exposed to dThd and dUrd supplementation in drinking water start- ing from weaning (exposed group), or without exposure (naïve group). The monitored range encompasses 5 to 34 weeks of age. Each symbol and line represents a single mouse. Gray dotted and shadowed areas indicate the concentration range in naïve dKO and WT groups, respectively. b) Plasma dThd concentrations before (0 h) and after a single adminis- tration of dThd and dUrd by oral gavage (166 mg/kg each) in naïve and exposed WT mice, naïve and exposed dKO mice, and exposed dKO mice previously treated with different doses of AAV-TBG, AAV-AAT and AAV-HLP vectors. Mice were 35 to 45 weeks old at the time of the experiment. Each point represents mean§range. White and gray areas between dotted lines indicate the concentration range in naïve dKO and WT groups at time zero, respectively. c, d and e) PK results (mean+SD) obtained from experiments depicted at b. Sample sizes in panels b to e are: naïve WT, n = 8; naïve dKO, n = 6; exposed WT, n = 5; exposed dKO, n = 7; TBG-treated mice with 5 £ 1011 vg/kg n = 2, 1012 vg/kg n = 4, 2 £ 1012 vg/kg n = 5, and 1013 vg/kg n = 5; AAT-treated mice with 2 £ 1012 vg/kg n = 6 and 1013 vg/kg n = 5; HLP-treated mice with 2 £ 1012 vg/kg n = 6 and 1013 vg/kg n = 4. Asterisks indicate statistically significant differences (*p<0.05, **p<0.01; panels c and d, Kruskal-Wallis test followed by the Dunn's multiple comparisons test, all groups compared with dKO exposed; panel e, Kruskal-Wallis test followed by the Dunn's multiple comparisons test, all groups compared with each other). AUC, area under the curve. dThd AUC (23.8 mmoles dThd x hour/L, range 15.8 29.3). Similar AAV, with levels comparable to those of exposed WT mice in all effects of AAV treatment on PK were observed for dUrd (Supple- groups and doses except the lowest AAV-TBG dose). In plasma, mentary Fig. 2). reductions were especially prominent in mice treated with the AAV- Nucleoside-exposed mice were treated with dThd/dUrd in drink- AAT vector (and reached statistical significance for the dose 2 £ 1012 ing water from weaning until the end of the study (at 22 months vg/kg), whereas there was no effect in small intestine with any treat- old). At this point, mice were killed and dThd and dUrd levels were ment. The greatest reductions were obtained for the AAV-AAT vector analyzed in plasma and liver (AAV target organ), as well as brain and in plasma, liver, and brain. Similar results were observed for the dUrd small intestine (two affected organs in MNGIE patients). Chronic oral concentrations (Supplementary Fig. 4). nucleoside exposure resulted in increased dThd median levels in all tissues in WT mice (plasma 5-fold, liver 2-fold, brain 4-fold, and small 3.3. AAV-TYMP treatment induces increased TP activity in liver intestine 90-fold, as compared with median levels in naïve WT group), and showed a higher increase in dKO mice (plasma 20-fold, Chronic nucleoside exposure induced increased TP activity in liver liver 10-fold, brain 7.5-fold, and small intestine 275-fold, as com- tissue in WT mice (mean value 39 nmoles Thy/h/mg prot in naïve pared with median levels in naïve WT group) (Fig. 2). In addition, WT, and 112 in exposed WT), and in dKO mice (mean value 1.2 in AAV-treatment reduced dThd and dUrd levels in nucleoside-exposed naïve dKO, and 9.9 in exposed dKO) (Fig. 3a), but not in small intes- mice at the end of the study in liver (reductions compared to exposed tine, brain, or skeletal muscle (Supplementary Fig. 5). Treatment with dKO AAV-untreated, with levels comparable to those of WT naïve AAV vectors provided TP activity to liver of exposed mice in a dose- mice) and in brain (reductions compared to untreated exposed dKO dependent manner (Fig. 3b). AAV-AAT was the most effective vector 6  et al. / EBioMedicine 62 (2020) 103133 F. Vila-Julia Fig. 2. Effect of AAV treatment on dThd levels in 22-month-old mice. Plasma, liver, brain, and small intestine dThd concentration in WT (n = 7 9) and dKO (n = 8) naïve mice, WT (n = 4) and dKO (n = 10) exposed mice, and dKO exposed mice treated with different AAV doses (TBG-treated mice with 5 £ 1011 vg/kg n = 2, 1012 vg/kg n = 4, 2 £ 1012 vg/kg n = 6 and 1013 vg/kg n = 4; AAT-treated mice with 2 £ 1012 vg/kg n = 6 and 1013 vg/kg n = 4; HLP-treated mice with 2 £ 1012 vg/kg n = 6 and 1013 vg/kg n = 3). Horizontal lines indicate the median concentration for each group. Asterisks indicate statistically significant differences (*p<0.05, **p<0.01; Kruskal-Wallis test followed by the Dunn's multiple comparisons test, all groups compared with dKO exposed). in providing TP activity to liver, in agreement with its nucleoside 3.4. Histological analysis of small intestine and mtDNA copy number reduction. At the highest dose (1013 vg/kg), liver TP activity mean val- ues in exposed mice increased 40- to 75-fold compared to WT values: Histological analysis of the small intestine (Masson’s trichrome 2091§140 nmoles Thy/h/mg prot for AAV-TBG (mean§SD); 3031§ staining) did not reveal any detectable abnormalities induced by 509 for AAV-AAT and 1535§456 for AAV-HLP (Fig. 3b). nucleoside exposure in this tissue. Also, we could not find that nucle- Vector copy number (VCN) increased in liver in a dose-depen- oside exposure induced mtDNA depletion in brain and small intes- dent manner (Fig. 3c). In an attempt to compare the efficacy of tine. (Supplementary figure 6). the promoters carried by the three vectors, we normalized TP activity per VCN, and the results revealed that TP/VCN ratios were not constant for each vector, but declined as VCN increased 3.5. Effect of the AAV-TYMP treatment on neurological phenotype (Fig. 3d). This apparent “saturation of efficacy” makes it difficult to compare the efficacies of different promoters, but the fact that the Motor coordination, balance and exercise tolerance were deter- highest TP/VCN ratios were always observed with the AAV-AAT mined using the rotarod performance test in 25 week-old mice vector for all different VCN values (Fig. 3d), suggests that AAT was (Fig. 4). The performance in this test can be influenced by impaired the most effective promoter among those tested. The TP vs VCN function of the skeletal muscles and/or the central nervous system. distribution was adjusted to the nonlinear function: Multiple comparison analysis of all groups represented in Fig. 4 Y = intercept + slope x log(X), where Y is TP activity and X is VCN. (Kruskal-Wallis test, followed by the Dunn's multiple comparisons Semilogarithmic transformation of the function showed that the test) did not detect statistical differences for any group, probably due highest slope among the three vectors corresponded to AAV-AAT to the high number of groups. However, when we only analysed the (p<0.05, extra sum of squares F-test). effect of nucleoside exposition, we observed that exposed dKO mice Although all the promoters used here are liver-specific, AAV treat- had a reduced time to fall, as compared with WT naïve mice ment provided TP activity to small intestine in exposed dKO mice, as (p = 0.006, Mann-Whitney U test) (Fig. 4). In AAV-treated mice, well as skeletal muscle treated with the highest vector doses, and rotarod performance varied widely, but in most AAV-treated mice brain in a few individual mice. AAV-induced TP activities did not (69%; 27 out of 39) was within or above the naïve WT normal range reach levels above WT values in non-hepatic tissues except in a single (19 44 s), whereas only 33% (3 out of 9) of exposed dKO mice skeletal muscle sample (Supplementary Fig. 5). showed a lengthier time to fall.  et al. / EBioMedicine 62 (2020) 103133 F. Vila-Julia 7 Fig. 3. TP activity in liver after AAV treatment. Liver TP activity in (a) WT (n = 8) and dKO (n = 8) naïve mice and in untreated WT (n = 4) and dKO (n = 10) exposed mice. (b) Liver TP activity in untreated WT and dKO exposed mice and in dKO exposed mice treated with different AAVs doses (TBG-treated mice with 5 £ 1011 vg/kg n = 2, 1012 vg/kg n = 4, 2 £ 1012 vg/kg n = 6 and 1013 vg/kg n = 4; AAT-treated mice with 2 £ 1012 vg/kg n = 6 and 1013 vg/kg n = 4; HLP-treated mice with 2 £ 1012 vg/kg n = 6 and 1013 vg/kg n = 3). (c) Liver vector copy number (VCN) in AAV-treated dKO exposed mice. TBG-treated mice with 5 £ 1011 vg/kg n = 1, 1012 vg/kg n = 4, 2 £ 1012 vg/kg n = 5 and 1013 vg/kg n = 4; AAT-treated mice with 2 £ 1012 vg/kg n = 6 and 1013 vg/kg n = 4; HLP-treated mice with: 2 £ 1012 vg/kg n = 5 and 1013 vg/kg n = 3. Horizontal lines indicate the median for each group. (d) Non-linear regression of liver TP activity (data from panel b) vs. liver VCN (data from panel c) in AAV-treated mice (Y = intercept + slope x log(X)). Note that the low VCN region is magnified to facilitate observation of the differences between the three vectors. Asterisks indicate statistically significant differences (*p<0.05, **p<0.01; for panel a, Mann-Whitney U test; for panel b, Kruskal-Wallis test followed by the Dunn's multiple comparisons test, all groups compared with each other, two different analyses for each dose). All MNGIE patients have diffuse leukoencephalopathy on brain extreme values; r = 0.358, p = 0.005, n = 59, including all values) and MRI [9], and chronic oral exposure to dThd and dUrd enhances leu- with rotarod test performance (Spearman r = 0.348, p = 0.0068, koencephalopathy in the dKO murine model [25]. We performed n = 59, excluding two ventricular volume extreme values). Ventricu- brain MRI studies in 84-week-old naïve and exposed WT and dKO lar volumes did not significantly correlate with plasma or liver dThd mice. Larger T2-weighted hyperintense areas (indicative of cerebro- levels. spinal fluid) were found in exposed dKO mice than in naïve dKO, naïve WT, or exposed WT mice (Fig. 5a and Supplementary Fig. 7). In 3.6. AAV-TYMP treatment normalized the mitochondrial dNTP balance contrast to what was reported in the former study of nucleoside exposed dKO mice, we did not find leukoencephalopathy in our In MNGIE, nucleoside overload leads to an unbalanced dNTP mito- experimental setting, probably because of different magnetic reso- chondrial pool, resulting in mtDNA instability [4]. We quantified nance protocols [25]. Hyperintensity was mainly located in the third dNTP levels in mitochondrial extracts from liver and brain in 22- and lateral ventricles, suggesting ventricular enlargement. The vol- month-old mice (Fig 6). dTTP content was significantly increased in ume of the third and lateral ventricles was quantified using T2 maps, brain mitochondria of exposed dKO mice. The same trend was and it was significantly greater in exposed dKO mice than in naïve observed in liver mitochondria, but statistical significance was not WT mice whereas it was unchanged in both naïve dKO and exposed reached, probably owing to the wide dispersion of the values (liver WT mice (Fig. 5b). AAV-AAT and AAV-HLP treatment prevented ven- dTTP range: 5.3 54.7 pmoles/mg protein). The dTTP expansion in tricular enlargement in exposed dKO mice: most animals treated liver was largely prevented in exposed dKO mice treated with AAVs, with these vectors had normal ventricular volumes, with the excep- as most values of AAV treated mice returned to dTTP values with the tion of one mouse treated with 1013 vg/kg AAV-AAT that developed naïve WT range, although only the group treated with AAV-TBG at hydrocephalus. In contrast, AAV-TBG failed to prevent ventricular 1013 vg/kg reached statistical significance (Fig. 6a). AAV treatment enlargement in most cases (Fig. 5b). also limited dTTP expansion in brain, but the effect was only partial Ventricular volumes at 84 weeks of age significantly correlated as dNTP in AAV-treated animals did not reach WT levels (and reached with brain dThd levels at the end of the study (94 weeks) in all groups statistical significance only in the group treated with AAV-AAT at (Spearman r = 0.324, p = 0.014, n = 57, excluding 2 ventricular volume 2 £ 1012 vg/kg) (Fig. 6b). The observations regarding mitochondrial 8  et al. / EBioMedicine 62 (2020) 103133 F. Vila-Julia because gastrointestinal dysmotility is the most debilitating symp- tom in MNGIE patients [9]. However, it does show some new func- tional and radiological alterations (neuromotor dysfunction and ventricular alterations in brain) that are absent or much less promi- nent in the non-nucleoside exposed murine model. We used this enhanced preclinical model to expand on our previous studies dem- onstrating the feasibility and efficacy of AAV-based liver-targeted gene therapy for MNGIE [19,21,23]. Systemic dThd and dUrd levels in naïve dKO mice were 2- to 4- fold higher than levels in naïve WT mice, as was originally reported [5] and consistently observed here, which contrasts with the much more pronounced accumulations (more than 50-fold) observed in MNGIE patients [35,36]. In agreement with the study where the stressed model was established [25], we found that oral administra- tion of dThd and dUrd to dKO mice exacerbates systemic accumula- tion of both dNs. Furthermore, the highest dThd and dUrd accumulations were seen in small intestine, likely because of the Fig. 4. Motor function. Motor function assessed by rotarod test in 25-week-old WT and presence of ingested dNs, before or during intestinal absorption. In dKO naïve and exposed mice, and dKO exposed mice treated with different AAVs contrast, we found detectable dThd and dUrd levels in all tissues ana- doses. Horizontal lines show median time to fall in each group. Gray areas indicate the lyzed from naïve WT mice, and exposed dKO mice had dN levels only range of values in the naïve WT group. No statistical differences found, Kruskal-Wallis 20-fold above WT plasma values, and 10-fold above liver and brain test followed by the Dunn's multiple comparisons test, all groups compared with dKO exposed). WT values. These discrepancies likely result from the use of TP activ- ity inhibitors during sample processing for dN determination, which prevented in vitro dN degradation in WT samples and reduced the dCTP levels in liver conversely mirrored those of dTTP; that is, differences between naïve WT and exposed dKO concentrations. As exposed dKO showed dCTP depletion as compared with WT levels, was expected, the excess of dThd and dUrd resulted in enhanced which was largely prevented by AAV treatment, reaching statistical mitochondrial dNTP imbalances (dTTP increase and dCTP depletion) significance for all AAV-AAT treated groups, and for the group treated in the two tissues analyzed (liver and brain). with AAV- HLP at 2 £ 1012 vg/kg (Fig. 6c). No statistical differences We found an unreported effect of dN exposure on mice: an were detected in mitochondrial dCTP levels in brain (Fig. 6d). As to increase in TP activity in liver, but not in skeletal muscle, small intes- the purine deoxynucleotides, mitochondrial dGTP levels were tine, or brain. A similar effect (TYMP upregulation) has been reported expanded in liver (an increase that was not prevented by AAV treat- in cultured human cells in response to dThd and nucleoside analogues ment), while dATP showed a tendency to be depleted in the same [37]. Exposure to dN doubled liver TP activity in WT mice and, inter- organ (statistically non-significant), but the effect was prevented in estingly, also increased activity in dKO animals. Upregulation of Tymp most AAV treated groups. In contrast, mitochondrial dATP and dGTP or other genes encoding enzymes with TP activity (such as Upp1 and levels were barely affected or not affected at all in brain (Supplemen- Upp2) may account for this increase in WT mice. Additionally, dThd tary Fig. 8). excess could be inhibiting the ubiquitin-mediated TP degradation as it has been recently reported in human differentiating reticulocyte cell 4. Discussion cultures [38]. In contrast, Tymp and Upp1 genes are molecularly inacti- vated in the dKO model [5,39]; therefore, dN-induced TP activity in Over the last few years, gene therapy has been proposed to reduce these animals should be mediated by upregulation of another the relevant risks associated with the currently available options to enzyme, likely UP2 (encoded by Upp2). In fact, naïve dKO mice have treat MNGIE (allogeneic stem cell transplantation and orthotopic some detectable TP activity in liver that must be due to UP2. liver transplantation) [11-13,15-18,34]. Since the first proof-of- All MNGIE patients present diffuse leukoencephalopathy on brain concept preclinical studies demonstrating the feasibility of using len- MRI, but this abnormality is barely detectable in naïve dKO mice, and tiviral[22] and AAV vectors [23], several studies have shown that is only seen when they are very old [5]. The first study of the nucleo- these strategies provide maintained effectiveness at long term in the side-exposed dKO model demonstrated that chronic exposure to preclinical model [19-21,24]. The main endpoint used to draw con- dThd and dUrd led to increased brain MRI FLAIR intensity signal clusions was the ability of the vectors to achieve genetic correction in around the third and lateral ventricles [25]. In our study, we failed to target cells of the murine dKO model, thereby providing TP activity detect MRI differences between naïve dKO and naïve WT mice, show- and normalization of systemic dThd and dUrd levels. ing how elusive this feature is in the unexposed mouse model. None- However, the dKO mouse is a limited model due to the absence of theless, and in agreement with the previous report [25], dKO mice a clinically relevant phenotype, which consists of a lack of TP activity (but not WT mice) did develop the MRI phenotype under dN expo- (except for some residual activity in tissues expressing Upp2), moder- sure, although we found enlarged ventricle volumes rather than the ate dThd and dUrd systemic accumulation, and a modest mitochon- diffuse parenchymal leukoencephalopathy that is typically observed drial dNTP imbalance. Any additional traits beyond the biochemical in MNGIE patients. Combined analysis of T2 maps and high-resolu- consequences of the absence of TP have only been glimpsed in very tion T2-weighted images revealed that the periventricular signal old animals and, even so, as elusive and barely detectable signs [5]. hyperintensity corresponds to enlarged ventricular volume, with no The model can be enhanced by providing the animals with dThd and fluid transfer to the surrounding white matter, in contrast to what is dUrd in drinking water, which exacerbates the phenotype and high- observed in other hydrocephalic models with severe ventricular lights the importance of dThd and dUrd overload in the pathophysiol- enlargement [40]. Exposure to dNs also succeeded in the develop- ogy of MNGIE [25]. This enhanced model does not recapitulate the ment of a motor function phenotype. At 25 weeks of age, exposed main clinical features of the disease (which manifest mainly as severe dKO mice showed a shorter time to fall in the rotarod test, which is gastrointestinal dysmotility, myopathy, peripheral neuropathy and indicative of motor coordination impairment in this model, in agree- brain diffuse leukoencephalopathy). Specifically, the lack of gastroin- ment with previous findings [25]. Furthermore, rotarod time to fall at testinal phenotype is an important limitation of this animal model, 25 weeks of age inversely correlated with MRI ventricular volume  et al. / EBioMedicine 62 (2020) 103133 F. Vila-Julia 9 Fig. 5. Brain MRI analysis. (a) Coronal T2-weighted and T2 map images of 1-mm sections (region corresponding to slice 6 in Supp Figure S7) showing ventricular enlargement in dKO exposed mice (red arrows). (b) Ventricular volume in 80 to 84-weeks old WT (n = 9) and dKO (n = 5) naïve mice, non-treated WT (n = 5) and dKO (n = 12) exposed mice, and dKO exposed mice treated with different AAVs doses (TBG-treated mice with: 5 £ 1011 vg/kg n = 2, 1012 vg/kg n = 3, 2 £ 1012 vg/kg n = 5 and 1013 vg/kg n = 5; AAT-treated mice with: 2 £ 1012 vg/kg n = 6 and 1013 vg/kg n = 4; HLP-treated mice with: 2 £ 1012 vg/kg n = 6 and 1013 vg/kg n = 3). Ventricular volume was quantified in T2 maps as the sum of the ventricular volumes calculated in all sections analyzed in Supp Figure S6 (see materials and methods for calculations). Gray areas indicate the range of values in the naïve WT group. Asterisks indicate statistically significant differences (*p<0.05; Kruskal-Wallis test followed by the Dunn's multiple comparisons test, all groups compared with dKO exposed). Ven- tricular volume values above the mean plus 3 SD of all data were considered outliers and excluded from the analysis. Two outliers were identified in dKO exposed AVV-treated with TBG 1012 vg/kg, and TBG 2 £ 1012 vg/kg, one in each group. assessed at 84 weeks of age, which concurs with previous studies hcTYMP expression [21,23]. The need to optimize therapeutic gene associating hydrocephalus with motor function defects [41]. expression and minimize the vector dose to transfer therapy to As was discussed above, most effects of dN exposure seen in our MNGIE patients prompted us to explore new vector constructs [19]. study are consistent with previously reported findings [25], but we The ApoE-AAT enhancer-promoter cassette has been used in several failed to replicate some other important features. After careful histo- AAV-based preclinical and clinical studies [42-45]. The self-comple- logical analysis (Masson’s trichrome staining), we found no abnor- mentary genome configuration resulted in faster and stronger thera- malities in small intestine of exposed mice, and no mtDNA depletion peutic gene expression in these studies, and it has also been was detected in brain or in small intestine. successfully used in patients [46-48]. We found that use of the AAV- In previous preclinical studies, we used the TBG promoter in AAV AAT vector to carry hcTYMP in the dKO model resulted in greater gene therapy for MNGIE and proved that it is effective for driving efficacy as compared to other hepatic promoters or a self- 10  et al. / EBioMedicine 62 (2020) 103133 F. Vila-Julia Fig. 6. Mitochondrial dNTPs. Mitochondrial thymidine triphosphate (dTTP) and deoxycytidine triphosphate (dCTP) content in liver (a and c) and brain (b and d) of WT (n = 9) and dKO (n = 8) naïve mice, non-treated WT (n = 4) and dKO (n = 10) exposed mice and dKO exposed mice treated with different AAVs doses (TBG-treated mice with: 5 £ 1011 vg/kg n = 2, 1012 vg/kg n = 4, 2 £ 1012 vg/kg n = 6 and 1013 vg/kg n = 4; AAT-treated mice with: 2 £ 1012 vg/kg n = 6 and 1013 vg/kg n = 4; HLP-treated mice with: 2 £ 1012 vg/kg n = 6 and 1013 vg/kg n = 3). Horizontal lines indicate the median for each group. Gray areas indicate the range of values in the naïve WT group. Asterisks indicate statistically significant differ- ences (*p<0.05, **p<0.01; Kruskal-Wallis test followed by the Dunn's multiple comparisons test, all groups compared with dKO exposed). complementary configuration [19]. Now, taking advantage of the Previously, we observed that treatment with AAV-TBG, AAV-HLP enhanced nucleoside-exposed animal model to expand the study, we and AAV-AAT results in sporadic cases of increased TP activity in found that AAV-AAT provides an effect beyond the previously non-hepatic tissues [19], and this was confirmed here in an indepen- reported biochemical normalization. AAV-AAT performed better than dent set of experiments. In our previous report, we found that this other liver-targeted vectors in restoring biochemical homeostasis “illegitimate activity” seems not to be caused by ectopic expression, (nucleoside reduction, liver TP activity, dNTP normalization), and it as transgene DNA or mRNA could not be consistently found in sam- also improved the brain MRI abnormalities manifested in the ples with TP activity. Hence, this activity might respond to exosomal exposed mouse model. export from liver [49]. In any case, this extra-hepatic effect was irrel- Our attempts to determine the effect of different vectors on dN evant in quantitative terms in both the previous and the current homeostasis revealed an unexpected limitation of the model. As mice study. received dNs in their drinking water ad libitum, the variability in dN dNTP imbalances were ameliorated in liver as a consequence of levels seen over the day was likely determined by the last time the the systemic dN reduction. The effect in brain, however, was only mouse drank water. To circumvent this confounding factor, we partial. The fact that liver, but not brain, is the target of the vectors decided to determine PK parameters (peak concentration, AUC, and tested here may account for the modest effect in brain. In addition, time to reach WT values) after a single gavage administration, instead the anatomic complexity of the brain, constituted by many cell types of single dN values in one blood sample, as more reliable biochemical distributed among different functional areas, may mask local effects, endpoints to study the effect of the vectors on dN homeostasis in the as our method for mitochondrial dNTP determination is performed model. A vector dose-response was observed in most cases, and com- on total brain mitochondria. Interestingly, dCTP depletion could not parison between the three vectors showed that AAV-AAT was the be detected in brain mitochondria, which correlates with the absence most efficient to reduce plasma dN levels, which is consistent with of mtDNA depletion in this organ, in agreement with previous experi- their better comparative ability to provide TP activity to liver. As was mental data showing that the cause of mtDNA depletion in MNGIE is expected, all vectors led to substantial dose-dependent TP activity in limited availability of dCTP [4]. liver, and the analysis of TP activity/vector copy number ratios indi- Of note, this is the first time that AAV-mediated gene therapy has cated that the AAT promoter was the most efficient. All these obser- shown efficacy on disease-related features in MNGIE beyond normal- vations confirm and reinforce our previous results on the efficacy of ization of the biochemical imbalances. Nucleoside-exposed mice different vectors [19]. treated with AAV-HLP and AAV-AAT vectors recovered normal  et al. / EBioMedicine 62 (2020) 103133 F. Vila-Julia 11 ventricular volumes on MRI, while this effect was only partial (in program 2018FI_B_01115 to FV). This work was funded in part by the some mice) and not statistically significant with the AAV-TBG vector. Spanish Instituto de Salud Carlos III (Grants PI15/00465, PMP15/ When all mice were taken as a whole, ventricular volumes signifi- 00025 and PI18/01574 to R.M., co-funded with E.R.D.F.). The dis- cantly correlated with dThd levels in brain, reinforcing the notion closed funders had no role in study design, data collection and analy- that normalization of the biochemical imbalances is directly related sis, decision to publish, or preparation of the manuscript. with improvements in this neurological trait. However, the results did not allow us to draw clear conclusions on the effect of therapy on Data sharing statement motor function, as assessed with the rotarod test. Time to fall in the rotarod test was shorter in nucleoside-exposed dKO mice, which can The data that support the findings of this study are available from be due to alterations in the central nervous system and/or the skele- the corresponding authors, RM and JT, upon reasonable request. tal muscle. This reduction seemed to be prevented in most groups of AAV-treated mice. However, if this observation suggests a positive Supplementary materials effect of the treatment, it is uncertain why was it not observed at the highest vector doses. The reasons for this apparent paradox remain Supplementary material associated with this article can be found, to be explored. in the online version, at doi:10.1016/j.ebiom.2020.103133. The results of this study show that the AAV-AAT vector performed the best among those tested, efficiently providing TP activity to the liver and restoring dN and dNTP homeostasis. In addition, functional References improvements on the phenotype beyond normalization of biochemi- [1] Nishino I, Spinazzola A, Hirano M. Thymidine phosphorylase gene mutations in cal markers were proven for AAV-mediated MNGIE gene therapy in MNGIE, a human mitochondrial disorder. Science 1999;283(5402):689–92. the nucleoside-exposed dKO mouse model. AAV transduction rates [2] Marti R, Nishigaki Y, Hirano M. Elevated plasma deoxyuridine in patients with are lower in humans than in mice [50,51]; therefore, using more thymidine phosphorylase deficiency. Biochem Biophys Res Commun 2003;303 (1):14–8. potent promoters, within a reasonable range of potency, will be nec- [3] Spinazzola A, Marti R, Nishino I, Andreu AL, Naini A, Tadesse S, et al. Altered thy- essary to make AAV treatment possible in humans at feasible vector midine metabolism due to defects of thymidine phosphorylase. J Biol Chem doses. This will have an impact on both biosafety and the cost of 2002;277(6):4128–33. [4] Gonzalez-Vioque E, Torres-Torronteras J, Andreu AL, Marti R. Limited dCTP treatment, and contribute to facilitating implementation of this ther- availability accounts for mitochondrial DNA depletion in mitochondrial apy in patients, which constitutes a significant advance towards the neurogastrointestinal encephalomyopathy (MNGIE). PLoS Genet 2011;7(3): treatment of this devastating disorder with a much less invasive ther- e1002035. [5] Lopez LC, Akman HO, Garcia-Cazorla A, Dorado B, Marti R, Nishino I, et al. Unbal- apy than those currently available. anced deoxynucleotide pools cause mitochondrial DNA instability in thymidine phosphorylase-deficient mice. Hum Mol Genet 2009;18(4):714–22. Contributors [6] Nishigaki Y, Marti R, Copeland WC, Hirano M. Site-specific somatic mitochondrial DNA point mutations in patients with thymidine phosphorylase deficiency. J Clin Invest 2003;111(12):1913–21. F. Vila-Julia, J. Torres-Torronteras and R. Martí designed the study [7] Nishigaki Y, Marti R, Hirano M. ND5 is a hot-spot for multiple atypical mitochon- and wrote the manuscript; F. Vila-Julia , R. Cabrera-Pe rez, Y. Ca mara, drial DNA deletions in mitochondrial neurogastrointestinal encephalomyopathy. M. Molina-Berenguer, S. Lope-Piedrafita, and J. Torres-Torronteras Hum Mol Genet 2004;13(1):91–101. [8] Pontarin G, Ferraro P, Valentino ML, Hirano M, Reichard P, Bianchi V. Mitochon- performed the experiments and collected and analysed the data; F. drial DNA depletion and thymidine phosphate pool dynamics in a cellular model Vila-Julia, S. Lope-Piedrafita, J. Torres-Torronteras and R. Martí partic- of mitochondrial neurogastrointestinal encephalomyopathy. J Biol Chem ipated in data interpretation; M. Hirano generated and provided the 2006;281(32):22720–8. [9] Garone C, Tadesse S, Hirano M. Clinical and genetic spectrum of mitochondrial mouse model; F. Mingozzi produced and provided the AAV-AAT vec- neurogastrointestinal encephalomyopathy. Brain 2011;134(Pt 11):3326–32. tor; all the participating authors contributed to the critical revision of [10] Cabrera-Pe rez R, Torres-Torronteras J, Vila-Julia  F, Ortega FJ, Ca mara Y, Barqui- the manuscript. All the authors read and approved the final version nero J, et al. Prospective therapeutic approaches in mitochondrial neurogastroin- testinal encephalomyopathy (MNGIE). Expert Opin Orphan Drugs 2015;3 of the manuscript. (10):1167–82. [11] Halter J, Schupbach WM, Casali C, Elhasid R, Fay K, Hammans S, et al. Allogeneic Declaration of Competing Interests hematopoietic SCT as treatment option for patients with mitochondrial neurogas- trointestinal encephalomyopathy (MNGIE): a consensus conference proposal for a standardized approach. Bone Marrow Transpl 2010;46(3):330–7. RM, FV and MM report grants from the Instituto de Salud Carlos III [12] Hirano M, Marti R, Casali C, Tadesse S, Uldrick T, Fine B, et al. Allogeneic stem cell during the conduct of the study; JT reports grants from Departament transplantation corrects biochemical derangements in MNGIE. Neurology de Salut, Generalitat de Catalunya (PERIS program), during the conduct 2006;67(8):1458–60. [13] Halter JP, Michael W, Schupbach M, Mandel H, Casali C, Orchard K, et al. Alloge- of the study; RM, MH, JT and YC report grants and non-financial sup- neic haematopoietic stem cell transplantation for mitochondrial neurogastroin- port from Modis Therapeutics, personal fees and other from Modis testinal encephalomyopathy. Brain 2015;138(Pt 10):2847–58. Therapeutics, outside the submitted work; in addition, RM and MH [14] Boschetti E, D'Alessandro R, Bianco F, Carelli V, Cenacchi G, Pinna AD, et al. Liver as a source for thymidine phosphorylase replacement in mitochondrial neurogas- have a patent “Deoxynucleoside therapy for diseases caused by unbal- trointestinal encephalomyopathy. PLoS ONE 2014;9(5):e96692. anced nucleotide pools including mitochondrial DNA depletion syn- [15] D'Angelo R, Boschetti E, Amore G, Costa R, Pugliese A, Caporali L, et al. Liver trans- dromes” (PCT/US16/038110) with royalties paid to Modis plantation in mitochondrial neurogastrointestinal encephalomyopathy (MNGIE): clinical long-term follow-up and pathogenic implications. J Neurol 2020. Therapeutics; MH reports grants and other support from Entrada Ther- [16] D'Angelo R, Rinaldi R, Pironi L, Dotti MT, Pinna AD, Boschetti E, et al. Liver trans- apeutics, grants from Muscular Dystrophy Association and grants from plant reverses biochemical imbalance in mitochondrial neurogastrointestinal NIH, outside the submitted work; FM is an employee of Spark Thera- encephalomyopathy. Mitochondrion 2017;34:101–2. [17] De Giorgio R, Pironi L, Rinaldi R, Boschetti E, Caporali L, Capristo M, et al. Liver peutics; SL reports that the Nuclear Magnetic Resonance facility where transplantation for mitochondrial neurogastrointestinal encephalomyopathy. she works charged the VHIR a fee for the MRI services; and RM, YC, JT Ann Neurol 2016;80(3):448–55. and RC have a patent “Treatment of mitochondrial diseases” (PCT/ [18] Kripps K, Nakayuenyongsuk W, Shayota BJ, Berquist W, Gomez-Ospina N, Esqui- vel CO, et al. Successful liver transplantation in mitochondrial neurogastrointesti- EP2016/062636) with royalties paid to Modis Therapeutics. nal encephalomyopathy (MNGIE). Mol Genet Metab 2020;130(1):58–64. [19] Cabrera-Pe rez R, Vila-Julia  F, Hirano M, Mingozzi F, Torres-Torronteras J, Martí R. Acknowledgments Alpha-1-antitrypsin promoter improves the efficacy of an adeno-associated virus vector for the treatment of mitochondrial neurogastrointestinal encephalomyop- athy. Hum Gene Ther 2019;30(8):985–98. JT and FV were funded by fellowships granted by the Generalitat [20] Torres-Torronteras J, Cabrera-Pe rez R, Barba I, Costa C, de Luna N, Andreu AL, et al. de Catalunya (PERIS program, SLT002/16/00370 to JT and FI-AGAUR Long-term restoration of thymidine phosphorylase function and nucleoside 12  et al. / EBioMedicine 62 (2020) 103133 F. Vila-Julia homeostasis using hematopoietic gene therapy in a murine model of mitochon- [36] Valentino ML, Marti R, Tadesse S, Lopez LC, Manes JL, Lyzak J, et al. Thymidine and drial neurogastrointestinal encephalomyopathy. Hum Gene Ther 2016;27(9): deoxyuridine accumulate in tissues of patients with mitochondrial neurogas- 656–67. trointestinal encephalomyopathy (MNGIE). FEBS Lett 2007;581(18):3410–4. [21] Torres-Torronteras J, Cabrera-Pe rez R, Vila-Julia  F, Viscomi C, Ca mara Y, Hirano M, [37] Tsuneyoshi K, Haraguchi M, Hongye Z, Gotanda T, Tachiwada T, Sumizawa T, et al. et al. Long-term sustained effect of liver-targeted adeno-associated virus gene Induction of thymidine phosphorylase expression by AZT contributes to enhance- therapy for mitochondrial neurogastrointestinal encephalomyopathy. Hum Gene ment of 50 -DFUR cytotoxicity. Cancer Lett 2006;244(2):239–46. Ther 2018;29(6):708–18. [38] Meinders M, Shoemark D, Dobbe JGG, Streekstra GJ, Frayne J, Toye AM. Expression [22] Torres-Torronteras J, Gomez A, Eixarch H, Palenzuela L, Pizzorno G, Hirano M, and retention of thymidine phosphorylase in cultured reticulocytes as a novel et al. Hematopoietic gene therapy restores thymidine phosphorylase activity in a treatment for MNGIE. Mol Ther Methods Clin Dev 2020;17:822–30. cell culture and a murine model of MNGIE. Gene Ther 2011;18(8):795–806. [39] Cao D, Leffert JJ, McCabe J, Kim B, Pizzorno G. Abnormalities in uridine homeo- [23] Torres-Torronteras J, Viscomi C, Cabrera-Perez R, Camara Y, Di Meo I, Barquinero J, static regulation and pyrimidine nucleotide metabolism as a consequence of et al. Gene therapy using a liver-targeted AAV vector restores nucleoside and the deletion of the uridine phosphorylase gene. J Biol Chem 2005;280(22): nucleotide homeostasis in a murine model of MNGIE. Mol Ther 2014;22(5):901–7. 21169–75. [24] Yadak R, Cabrera-Pe rez R, Torres-Torronteras J, Bugiani M, Haeck JC, Huston MW, [40] Mandell JG, Neuberger T, Drapaca CS, Webb AG, Schiff SJ. The dynamics of brain et al. Preclinical efficacy and safety evaluation of hematopoietic stem cell gene and cerebrospinal fluid growth in normal versus hydrocephalic mice. J Neurosurg therapy in a mouse model of MNGIE. Mol Ther Methods Clin Dev 2018;8:152–65. Pediatr 2010;6(1):1–10. [25] Garcia-Diaz B, Garone C, Barca E, Mojahed H, Gutierrez P, Pizzorno G, et al. Deoxy- [41] Lee MJ, Chang CP, Lee YH, Wu YC, Tseng HW, Tung YY, et al. Longitudinal evalua- nucleoside stress exacerbates the phenotype of a mouse model of mitochondrial tion of an N-ethyl-N-nitrosourea-created murine model with normal pressure neurogastrointestinal encephalopathy. Brain 2014;137(Pt 5):1337–49. hydrocephalus. PLoS ONE 2009;4(11):e7868. [26] McIntosh J, Lenting PJ, Rosales C, Lee D, Rabbanian S, Raj D, et al. Therapeutic lev- [42] George LA, Sullivan SK, Giermasz A, Rasko JEJ, Samelson-Jones BJ, Ducore J, et al. els of FVIII following a single peripheral vein administration of rAAV vector Hemophilia B gene therapy with a high-specific-activity factor IX variant. N Engl J encoding a novel human factor VIII variant. Blood 2013;121(17):3335–44. Med 2017;377(23):2215–27. [27] Davidoff AM, Ng CY, Zhou J, Spence Y, Nathwani AC. Sex significantly influences [43] Manno CS, Pierce GF, Arruda VR, Glader B, Ragni M, Rasko JJ, et al. Successful transduction of murine liver by recombinant adeno-associated viral vectors transduction of liver in hemophilia by AAV-Factor IX and limitations imposed by through an androgen-dependent pathway. Blood 2003;102(2):480–8. the host immune response. Nat Med 2006;12(3):342–7. [28] Bradford MM. A rapid and sensitive method for the quantitation of microgram [44] Miao CH, Ohashi K, Patijn GA, Meuse L, Ye X, Thompson AR, et al. Inclusion of the quantities of protein utilizing the principle of protein-dye binding. Anal Biochem hepatic locus control region, an intron, and untranslated region increases and sta- 1976;72:248–54. bilizes hepatic factor IX gene expression in vivo but not in vitro. Mol Ther 2000;1 [29] Martí R, Lo pez LC, Hirano M. Assessment of thymidine phosphorylase function: (6):522–32. measurement of plasma thymidine (and deoxyuridine) and thymidine phosphor- [45] Nayak S, Herzog RW. Progress and prospects: immune responses to viral vectors. ylase activity. Methods Mol Biol (Clifton, NJ) 2012;837:121–33. Gene Ther 2010;17(3):295–304. [30] Perez-Perez MJ, Priego EM, Hernandez AI, Camarasa MJ, Balzarini J, Liekens S. [46] Nathwani AC, Gray JT, Ng CY, Zhou J, Spence Y, Waddington SN, et al. Self-comple- Thymidine phosphorylase inhibitors: recent developments and potential thera- mentary adeno-associated virus vectors containing a novel liver-specific human peutic applications. Mini Rev Med Chem 2005;5(12):1113–23. factor IX expression cassette enable highly efficient transduction of murine and [31] Sherman PA, Fyfe JA. Enzymatic assay for deoxyribonucleoside triphosphates nonhuman primate liver. Blood 2006;107(7):2653–61. using synthetic oligonucleotides as template primers. Anal Biochem 1989;180 [47] Nathwani AC, Reiss UM, Tuddenham EG, Rosales C, Chowdary P, McIntosh J, et al. (2):222–6. Long-term safety and efficacy of factor IX gene therapy in hemophilia B. N Engl J [32] Blazquez-Bermejo C, Molina-Granada D, Vila-Julia  F, Jime nez-Heis D, Zhou X, Med 2014;371(21):1994–2004. Torres-Torronteras J, et al. Age-related metabolic changes limit efficacy of deoxy- [48] Wang Z, Ma HI, Li J, Sun L, Zhang J, Xiao X. Rapid and highly efficient transduction nucleoside-based therapy in thymidine kinase 2-deficient mice. EBioMedicine by double-stranded adeno-associated virus vectors in vitro and in vivo. Gene 2019;46:342–55. Ther 2003;10(26):2105–11. [33] Moolenbeek C, Ruitenberg EJ. The "Swiss roll": a simple technique for histological [49] Simpson RJ, Kalra H, Mathivanan S. ExoCarta as a resource for exosomal research. studies of the rodent intestine. Lab. Anim. 1981;15(1):57–9. J Extracell Vesicles 2012;1. [34] Hirano M, Casali C, Tadesse S, Stanzani M, Savage D. Sustained biochemical and [50] Lisowski L, Dane AP, Chu K, Zhang Y, Cunningham SC, Wilson EM, et al. Selection clinical improvements two years postallogeneic stem cell transplantation in a and evaluation of clinically relevant AAV variants in a xenograft liver model. patient with MNGIE. Neurology 2008;70:A406–A7. Nature 2014;506(7488):382–6. [35] Marti R, Spinazzola A, Tadesse S, Nishino I, Nishigaki Y, Hirano M. Definitive diag- [51] Vercauteren K, Hoffman BE, Zolotukhin I, Keeler GD, Xiao JW, Basner-Tschakarjan nosis of mitochondrial neurogastrointestinal encephalomyopathy by biochemical E, et al. Superior in vivo transduction of human hepatocytes using engineered assays. Clin Chem 2004;50(1):120–4. AAV3 capsid. Mol Ther 2016;24(6):1042–9.