Alpha-1-Antitrypsin Promoter Improves the Efficacy of an Adeno-Associated Virus Vector for the Treatment of Mitochondrial Neurogastrointestinal Encephalomyopathy

Abstract

Mitochondrial neurogastrointestinal encephalomyopathy (MNGIE) is a devastating disease caused by mutations in TYMP, which encodes thymidine phosphorylase (TP). In MNGIE patients, TP dysfunction results in systemic thymidine and deoxyuridine overload, which interferes with mitochondrial DNA replication. Preclinical studies have shown that gene therapy using a lentiviral vector targeted to hematopoietic stem cells or an adeno-associated virus (AAV) vector transcriptionally targeted to liver are feasible approaches to treat MNGIE. Here, we studied the effect of various promoters (thyroxine-binding globulin [TBG], phosphoglycerate kinase [PGK], hybrid liver-specific promoter [HLP], and alpha-1-antitrypsin [AAT]) and DNA configuration (single stranded or self complementary) on expression of the TYMP transgene in the AAV8 serotype in a murine model of MNGIE. All vectors restored liver TP activity and normalized nucleoside homeostasis in mice. However, the liver-specific promoters TBG, HLP, and AAT were more effective than the constitutive PGK promoter, and the self-complementary DNA configuration did not provide any therapeutic advantage over the single-stranded configuration. Among all constructs, only AAV-AAT was effective in all mice treated at the lowest dose (5 × 10¹⁰ vector genomes/kg). As use of the AAT promoter will likely minimize the dose needed to achieve clinical efficacy as compared to the other promoters tested, we propose using the AAT promoter in the vector eventually designed for clinical use.

Introduction

Mitochondrial neurogastrointestinal encephalomyopathy (MNGIE) is a rare autosomal recessive disease caused by mutations in the nuclear gene TYMP, which encodes thymidine phosphorylase (TP).¹ The main features of MNGIE are severe gastrointestinal dysmotility, cachexia, progressive external ophthalmoplegia, peripheral neuropathy, leukoencephalopathy, and signs of mitochondrial dysfunction in muscle biopsy. It is a devastating disorder, as around 80% of patients do not survive beyond 40 years. Complications derived from the gastrointestinal dysfunction and critical nutritional status are the most common causes of death.²

In MNGIE patients, TP dysfunction results in systemic accumulation of TP substrates, the nucleosides thymidine (dThd) and deoxyuridine (dUrd),³ which leads to a mitochondrial deoxyribonucleoside triphosphate (dNTP) imbalance that interferes with mitochondrial DNA (mtDNA) replication.^4,5 More specifically, increased dThd concentration results in expansion of thymidine triphosphate (dTTP) and secondary depletion of deoxycytidine triphosphate (dCTP) in mitochondria.⁶ mtDNA replication is affected, and muscle and other tissues show mtDNA depletion, as well as multiple deletions and somatic point mutations, which account for the mitochondrial dysfunction and clinical phenotype.^7–9 Several therapies have been proposed for MNGIE patients, with the common aim of achieving permanent clearance of the toxic systemic metabolites.^10–13 Nonetheless, only allogeneic hematopoietic stem cell transplantation and more recently, orthotopic liver transplantation have achieved sustained reductions in nucleoside concentrations to undetectable (or barely detectable) levels and an improvement in the clinical outcome.^14–18 However, both of these interventions are associated with considerable mortality rates and complications, which are particularly pronounced in MNGIE patients because of their poor medical condition at the time of treatment.^14,19 The difficulty of finding a compatible donor is an additional limitation of these treatments.

Based on this background, it seems evident that there is a need to develop safer therapeutic approaches for MNGIE, and to this end, gene therapy could be an option. We first explored the feasibility of gene therapy for this purpose in preclinical in vitro and in vivo models using a lentiviral vector targeting hematopoietic stem cells.²⁰ The results were consolidated in a follow-up study in lentivirus-treated animals²¹ that showed long-term transgene expression and sustained biochemical normalization. However, lentiviral vectors are associated with a risk of oncogenesis because of their integrative nature.²² To overcome this drawback, we later developed an alternative strategy based on use of an adeno-associated virus (AAV) vector transcriptionally targeted to the liver. Our previous results in a murine model of MNGIE using thyroxine-binding globulin (TBG) promoter (AAV2/8-TBG vector) demonstrated that AAV-mediated liver-targeted TYMP expression normalizes nucleoside metabolism²³ in a sustained long-term effect.²⁴

However, gene therapy has been associated with a dose-dependent immune response, and therefore, it is preferable to minimize the vector dose administered.^25–27 In this line, we have now focussed on investigating several alternative vectors having different promoters and DNA configurations. Specifically, we studied the effect of four promoters, the constitutive phosphoglycerate kinase (PGK) promoter and three liver-specific promoters—TBG promoter, hybrid liver-specific promoter (HLP), and alpha-1-antitrypsin (AAT) promoter—and two DNA configurations—single-stranded DNA and self-complementary DNA—on expression of the TYMP transgene when used in the AAV2/8 serotype in a murine model of the disease.

Materials and Methods

AAV vector construction, production, and titration

A schematic representation of the four vectors investigated in this study is shown in Figure 1. In all cases, the AAV2/8 serotype was used.

Figure 1.

Schematic representation of the four adeno-associated virus (AAV) vectors studied. In all cases, vector serotype corresponded to AAV2/8. ITR, inverted terminal repeats; TBG, thyroxine-binding globulin promoter; hcTYMP DNA, human thymidine phosphorylase gene coding sequence; WPRE, woodchuck hepatitis virus post-transcriptionally regulatory element; BGH, bovine growth hormone polyadenylation signal; TK1 intron, stuffer DNA corresponding to the human thymidine kinase 1 (TK1) intron 3 sequence; PGK, phosphoglycerate kinase promoter; HLP, hybrid liver-specific promoter; AAT, α-1-antitrypsin promoter; HBB2 intron, human haemoglobin subunit β-2 intron sequence; HBB2, bovine haemoglobin subunit β-2 polyadenylation signal; scAAV, self-complementary adeno-associated vector.

The AAV-TBG vector was generated at the AAV Vector Core of the Telethon Institute of Genetics and Medicine (Pozzuoli, Italy), as previously detailed.²³ Briefly, the coding sequence of the human TYMP gene (hcTYMP) was PCR amplified, cloned into the pCR2.1 TOPO vector (Invitrogen, Carlsbad, CA), sequence verified, and finally cloned into the single EagI restriction site of the AAV2/8-TBG vector. The vector was produced by transfection of 293 cells and purified by CsCl gradient.

For generation of the AAV-PGK vector, the intron 3 sequence of TK1 was PCR amplified, subcloned into the TOPO-TA vector (Invitrogen), and finally cloned between the HindIII and EcoRV restriction sites in the pSMD2 plasmid. The PGK promoter and hcTYMP sequences were obtained from the previously described p305-TP lentiviral vector²⁰ and subcloned between the XhoI and XbaI restriction sites of the pcDNA3.1(+) plasmid. Finally, the PGK-hcTYMP fragment was cloned between the NotI and PmeI restriction sites of the pSMD2 plasmid to obtain the AAV-PGK vector. After sequence verification, the vector was produced by the Vectors Production Unit of the Animal Biotechnology and Gene Therapy Centre (Autonomous University of Barcelona, Barcelona, Catalonia), by triple transfection of 293 cells. The vector was purified by iodixanol gradient following previous polyethylene glycol precipitation.

For construction of the self-complementary AAV-HLP (scAAV-HLP) vector, the pAV-HLP-Luc plasmid was provided by Dr. Natwani's group (University College London Cancer Institute, London, United Kingdom). The Luciferase gene was replaced by the hcTYMP sequence, which was cloned into the XbaI restriction site. Then, the HLP-hcTYMP fragment was purified and cloned into the SnaBI restriction site of the custom-synthesized scAAV-pA plasmid (GeneArt Platform, Life Technologies). The final scAAV-HLP vector was sequence-verified and produced by the Vectors Production Unit following the protocol described for the AAV-PGK vector.

Finally, for construction of the AAV-AAT vector, the hcTYMP sequence was obtained from the scAAV-HLP vector and cloned between the XbaI and XhoI restriction sites of the pSMD2-ApoE-hAAT-UGT1A1-2.1 plasmid. After sequence verification, production was carried out following a protocol based on triple transfection of 293 cells and CsCl gradient purification.

The vectors produced were titrated in our laboratory following a previously described method based on double-stranded DNA quantification using the Quant-iT PicoGreen dsDNA Assay Kit (ThermoFisher Scientific).²⁸ The AAV titer was calculated as the difference between the vector genomes value from lysed samples and the vector genome values for non-lysed samples, in order to quantify encapsidated DNA only. In the case of the AAV-TBG vector (previously tested and reported)²³ vector doses were recalculated according to the titers obtained by the picogreen method.

Animal procedures

All animal procedures were performed in accordance with protocols approved by our institutional review board. Male Tymp^−/−/Upp1^−/− double knockout (KO) mice, 8–12 weeks old, were treated with a single intravenous tail injection of the different vectors at different doses: 5 × 10¹⁰, 2 × 10¹¹ and 5 × 10¹¹ vg/kg for the TBG vector; 5 × 10¹⁰, 2 × 10¹¹, 5 × 10¹¹, 10¹² and 2 × 10¹² vg/kg for the AAT vector and 2 × 10¹¹, 5 × 10¹¹, 10¹² and 2 × 10¹² vg/kg for the PGK and HLP vectors. Age-matched untreated double KO and wild type (wt) mice were used as controls. Saphenous vein blood samples were collected weekly during the first month and every 2–4 weeks up to 28 weeks after treatment using EDTA capillaries (Microvette 200K3E, Sarstedt) to assess plasma nucleoside concentration and alanine aminotransferase (ALT) activity. Body weight was monitored and general health status assessed by qualitative survey (e.g., movement, coat appearance, behavior) throughout the study. Mice were killed by cervical dislocation 34 weeks after vector administration. At this point, intracardiac blood was extracted for dThd and dUrd determination and tissue samples were collected, immediately frozen in liquid nitrogen, and stored at −80°C for different purposes.

Nucleoside and TP activity determinations

Plasma dThd and dUrd concentrations were analyzed by high-performance liquid chromatography with ultraviolet absorption detection, as previously described.²⁰ For TP activity and tissue nucleoside determinations, a portion of frozen sample was homogenized in TP activity lysis buffer (50 mM Tris–HCl, pH 7.2; 10 mL/L Triton X-100; 2 mM phenylmethylsulfonyl fluoride; 0.2 mL/L 2-mercaptoethanol), using a Potter S homogenizer. Homogenates were centrifuged at 20,000 g for 30 min at 4^°C, and supernatants were separated into two aliquots. One aliquot was used for protein determination²⁹ and TP activity³⁰ determination, as reported. The other aliquot was frozen until use for measurement of nucleoside concentration by liquid chromatography coupled to tandem mass spectrometry.²³

ALT activity determination

To test whether the treatment induced hepatocellular toxicity, plasma ALT activity was monitored by periodic determinations (every 4 weeks) over the post-treatment period using a spectrophotometric method. ALT-mediated transamination of alanine to α-ketoglutarate was coupled to reduction of the product pyruvate via lactate dehydrogenase, with nicotinamide adenine dinucleotide (NADH) consumption spectrophotometrically monitored at λ = 340 nm. Fifty microliters of plasma was incubated for 10 min at 37°C in 720 μL of 0.1 M Tris-HCl (pH 7.5) containing 44 mM L-alanine, 0.2 mM NADH, and 1 U/mL of lactate dehydrogenase. Absorbance at 340 nm was monitored for 10 min (blank), and the ALT reaction was then triggered by the addition of 80 μL of 130 mM α-ketoglutarate to follow the slope for an additional 10 min. ALT activity was expressed as micromoles of pyruvate formed per minute and liter of plasma (IU/L) based on the molar extinction coefficient of NADH at 340 nm [6300 L/(mol·cm)].

Western blot

Protein extracts were obtained by homogenization of approximately 20 mg of tissue in TP activity lysis buffer. In all cases, 20 μg of total protein was electrophoresed on 10% polyacrylamide gel under denaturing conditions, and then transferred to polyvinylidene difluoride membranes using the Bio-Rad semi-dry system. Transferred proteins were blotted with a rabbit polyclonal TP antibody (ab69120, Abcam), a mouse monoclonal β-actin antibody (A5316, Sigma-Aldrich), or a mouse monoclonal GAPDH antibody (GA1R, Ambion). Peroxidase-conjugated goat anti-rabbit and rabbit anti-mouse immunoglobulins (Dako) were used as secondary antibodies. Finally, membranes were treated with the Immobilon Western chemiluminescent kit (Merck Millipore), and bands were visualized in an ODYSSEY Fc (Li-COR) detector and quantified using Image J software (U.S. National Institutes of Health, Bethesda, MD).

Histological analysis

Immunofluorescent histological analysis of human TP was performed in 10 μm liver cryosections fixed with acetone:methanol (1:1 v/v). Sections were then blocked with bovine serum albumin 2% (w/v) in Tris buffered saline buffer (50 mmol/L Tris–HCl, pH 7.6; 125 mmol/L NaCl) for 30 minutes and incubated with 10 μg/mL anti-TP rabbit polyclonal antibody (Abcam) overnight at 4°C. After washing, sections were stained with AlexaFluor 488 conjugated goat anti-rabbit immunoglobulin G (Invitrogen). Nuclear staining was performed with 1 μg/mL Hoechst 33342.

Liver mitochondrial dNTP measurement

Liver mitochondria were isolated as previously described.⁶ A volume (0.1–0.3 mL) of suspension with isolated mitochondria containing 0.5 mg protein was treated with trichloroacetic acid (final concentration 0.5 M) and centrifuged at 20,000 g for 5 min at 4°C. Supernatants were neutralized with 1.5 volumes of 0.5 M tri-N-octylamine in Freon (1,1,2-trichlorotrifluoroethane) and centrifuged for 10 min at 10,000 g at 4°C. Half the aqueous upper phase was recovered, dried under speed vacuum centrifugation at room temperature, and stored at −80°C until measurement. For dNTP quantification, dry dNTP extracts were thawed and dissolved in 40 μL of 40 mM Tris-HCl (pH 7.4) (final equivalent protein concentration, 12.5 μg/μL). For dATP and dTTP determination, nondiluted and 1:3 diluted samples were used. For dGTP and dCTP measurement, 1:5 and 1:10 dilutions were required. Duplicates of each sample were assessed. dNTP content was determined using the previously described polymerase-based method⁶ with minor modifications. Briefly, 10 μL of reaction mixture containing 5 μL of dNTP extract, 40 mM Tris-HCl (pH 7.4), 10 mM MgCl₂, 5 mM dithiothreitol, 0.25 μM oligoprimer, 0.75 μM [8-³H] dATP, 12–21 Ci/mmol (or [methyl-³H] dTTP for the dATP assay), and 0.025 U/μL of Sequencing Taq DNA Polymerase (Bioron) was incubated for 1 hour at 48°C. Reaction mixtures with 40 mM Tris-HCl (pH 7.4) dNTP standards were processed in triplicate in parallel. After incubation, 9 μL of the mixture was spotted on a 96-position DEAE filtermat (Perkin Elmer) and left to air dry. Filters were washed 6 times with 5% Na₂HPO₄, once with water, and once with 70% ethanol, and left to dry again. The retained radioactivity was determined by scintillation counting in a 1450 MicroBeta TriLux detector (Perkin Elmer) after incubating the dry filtermats with MetiLex solid melt-on scintillator (Perkin Elmer) at 80°C. dNTP amounts were calculated from interpolation on calibration curves. dTTP concentrations below 0.8 pmol/mg protein were undetectable and considered as zero. dATP could not be determined because values always fell below the lower limit of quantification.

Vector and mtDNA copy number and mRNA quantification

DNA was isolated from frozen liver samples by phenol–chloroform extraction, dissolved in 10 mM Tris-HCl (pH 8) and quantified by spectrophotometry (NanoDrop Spectrophotometer, Thermo Scientific). Detection and quantification of vector and mtDNA copy number was performed by quantitative real-time PCR (qPCR) using an ABI PRISM 7900HT sequence detector (Applied Biosystems). For hcTYMP DNA copy number quantification, we used the predesigned TaqMan MGB expression assay Hs00157317_m1 (Applied Biosystems) referred to the murine single-copy nuclear gene Ang1 predesigned TaqMan MGB gene expression assay Mm00833184_s1 (Applied Biosystems). Similarly, mtDNA content was measured using an MGB 16S custom-designed TaqMan probe (FAM-5′ AAGTCCTACGTGATCTGAGGT 3′-MGB) and primers (16S Forward: AATGGTTCGTTTGTTCAACGATT and 16S Reverse: AGAAACCGACCTGGATTGCTC) and referred to the same murine single-copy nuclear gene Ang1. In all cases, quantifications were obtained by interpolation on standard curves of several dilutions of plasmids containing hcTYMP DNA or a specific region of the Ang1 or 16S genes.

For mRNA quantification, total RNA was extracted from 20–30 mg of ground tissue using the RNeasy Mini Kit (Qiagen) and quantified by spectrophotometry (NanoDrop Spectrophotometer, Thermo Scientific). One μg of isolated RNA was treated with DNAse I, Amplification Grade (Invitrogen). Reverse transcription was performed on DNA-free RNA using the High-capacity cDNA Reverse Transcription Kit (Applied Biosystems). The resulting cDNA was used for qPCR analysis using the probe indicated above for hcTYMP and referred to Ppia mRNA (TaqMan Gene Expression Assay Mm02342430_g1).

Statistical analysis

Statistical analysis was performed with the GraphPad Prism 6 software (GraphPad Software, Inc.). Nonparametric tests were used in most comparisons because of the limited number of cases per group (n < 10). For comparisons between groups with a larger number of cases, parametric tests were used when data were compatible with a normal distribution. The tests used in each particular case are indicated in the figure legends. For statistical purposes, undetectable values were considered as zero.

Results

AAV-AAT was the most effective vector among all tested

Our previous work showed that a single intravenous injection of single-stranded AAV8 vector carrying the human TYMP coding sequence (hcTYMP), transcriptionally targeted to liver under control of the thyroxine-binding globulin promoter (AAV-TBG), reduces the systemic nucleoside overload in a mouse model of MNGIE.^23,24 Here, we compared the efficacy of this AAV-TBG vector with 3 alternative AAV8 vectors containing the same hcTYMP sequence under the control of different promoters and with different DNA configurations: AAV-PGK, with the constitutive phosphoglycerate kinase promoter and single-stranded DNA; scAAV-HLP, with the previously described hybrid liver-specific promoter³¹ and self-complementary DNA; and AAV-AAT, with the α-1-antitrypsin liver-specific promoter and apolipoprotein E hepatocyte control region enhancer³² and single- stranded DNA (Fig. 1). The HLP and AAT promoters are similar, consisting of a core enhancer from the human apolipoprotein hepatic control region, upstream of the α-1-antitrypsin gene promoter. However, in the case of HLP, a modified reduced sequence was used to overcome the size limitation of self-complementary vectors.

All results obtained here with AAV-PGK, scAAV-HLP, and AAV-AAT were compared with those obtained previously using the AAV-TBG vector in the same mouse model.²³ In order to accurately compare the results, the reported AAV-TBG doses were recalculated and corrected here according to the titers obtained with a picogreen dsDNA quantification method.²⁸ This technique, used for all vectors in this study, differs from the methods used for vector quantification in the original report (dot-blot and qPCR).²³ The vector genome re-assessment with the new method revealed that the titers had been significantly overestimated (4 × ) in the original report due to high amounts of nonencapsidated DNA. For this reason, the doses of AAV-TBG vector quoted here are 4-fold lower than those originally reported.²³ In all cases, we treated 8- to 12-week-old double KO Tymp^−/−/Upp1^−/− male mice⁴ with a single intravenous tail injection of the described vectors at several doses, ranging from 5 × 10¹⁰ to 2 × 10¹² vg/kg.

Plasma nucleoside concentration was monitored over 34 weeks following treatment to investigate treatment efficacy with the different vectors. The results for plasma dThd are shown in Fig. 2, and similar results were obtained for dUrd. The lowest dose of all vectors that reduced dThd to wt levels at some point during the monitoring period was 5 × 10¹¹ vg/kg (Fig. 2 and Supplementary Fig. S1). At lower doses, AAV-PGK failed to reduce dThd to wt levels at any point over the period monitored. The most effective vector in terms of reducing plasma nucleoside concentration was AAV-AAT, which, at the lowest dose tested (5 × 10¹⁰ vg/kg), brought dThd down to wt levels at 1 week post-treatment in all mice, and maintained the concentration at wt levels or below over the entire period monitored. scAAV-HLP, which had a similar promoter in a self-complementary configuration, achieved the same nucleoside reduction at 4 weeks after treatment with a dose of 2 × 10¹¹ vg/kg, suggesting that the self-complementary configuration does not accelerate TYMP transgene expression. At the time mice were killed (34 weeks after treatment administration), plasma dThd was at wt level or below in 65% of animals treated with AAV-PGK, 83% of those treated with AAV-TBG, 94% of those treated with scAAV-HLP, and 97% of those treated with AAV-AAT (Supplementary Fig. S1). These results indicate that in comparison with the previously described AAV-TBG vector, a more potent and faster effect on plasma nucleoside reduction is obtained with the AAV-AAT vector.

Figure 2.

Plasma thymidine (dThd) concentration. Plasma dThd concentration in the various groups of mice during the monitoring period. The number of animals (n) in each treatment group is 8–10 except for the previously described TBG vector: (n = 5-6), wt (n = 11), and KO (n = 13); where wt and KO correspond to age-matched, untreated wild-type and double knock-out Tymp^−/−/Upp1^−/− mice, respectively. The yellow area corresponds to the wt concentration range. Different dose groups are separated by vertical dotted lines. Horizontal short black lines indicate the median. At the last time point (24 weeks after vector administration), dThd levels were significantly lower than those of KO (p < 0.01) in all groups except for the 5 × 10¹⁰ TBG and 2 × 10¹¹ PGK groups (Mann–Whitney test).

To assess whether the vectors restored nucleoside homeostasis in tissue, we determined dThd and dUrd concentrations in liver and other commonly affected tissues in MNGIE patients (brain, skeletal muscle, and small intestine) at completion of the monitoring time (Figs. 3d and 4b). Dose-dependent nucleoside reductions to wt levels were observed in liver, brain, and skeletal muscle, with varying efficacy, depending on the vector. As was observed in blood, the most pronounced effect was obtained with AAV-AAT. In small intestine, significant dThd and dUrd reductions only occurred with scAAV-HLP and AAV-AAT (p < 0.01 for all doses, Mann-Whitney test, except for the lowest scAAV-HLP dose, 2 × 10¹¹ vg/kg). In agreement with the plasma findings, liver-specific promoters were more effective than the constitutive PGK promoter in tissues.

Figure 3.

Biochemical and molecular results in liver 34 weeks after treatment. (a) AAV vector genome copy number. Human hcTYMP DNA copy number was assessed by qPCR and referred to the murine single-copy gene Ang1. Results for all groups were significantly higher from those of KO (p < 0.01, Mann-Whitney test). (b) Thymidine phosphorylase (TP) activity in liver extracts. Results for all groups except 5 × 10¹⁰ TBG and 2 × 10¹¹ PGK were significantly higher from those of KO (p < 0.05, Mann–Whitney test) (c) TP activities normalized per vector genome copy number. Results were significantly higher in animals treated with the AAT vector when comparing with PGK and HLP groups (p < 0.01, Mann-Whitney test). However, significant differences were not detected between AAT and TBG groups at none of the tested doses. (d) Liver dThd concentration. Yellow area corresponds to the age-matched wt concentration range. Results for all groups except 5 × 10¹⁰ TBG and 2 × 10¹¹ PGK were significantly lower than those of KO (p < 0.001, Mann–Whitney test). In all panels, n = 8–11 in each group except TBG (n = 5–6), wt (n = 8–16), and KO (n = 7–17). Different dose groups are separated by vertical dotted lines. Horizontal short black lines indicate the median.

Figure 4.

TP activity and dThd concentration in different tissues. (a) TP activity measured in brain, gastrocnemius muscle, and small intestine 34 weeks after treatment. Using the Mann–Whitney test, the wt, HLP, and AAT groups showed significant differences relative to KO in brain (p < 0.01). In gastrocnemius and small intestine, all groups treated with TBG, HLP, and AAT at doses higher than 5 × 10¹¹ vg/kg showed significant differences compared with KO (p < 0.05); the PGK groups showed no differences relative to KO in any case. (b) dThd concentration in brain, gastrocnemius, and small intestine at the end of the treatment. The yellow area corresponds to the age-matched wt concentration range. Excluding animals treated with 5 × 10¹⁰ TBG and 2 × 10¹¹ PGK, all groups showed significant differences relative to KO in both brain and gastrocnemius (p < 0.01, Mann–Whitney test). In the case of the small intestine, only results from the HLP (except lowest dose) and AAT groups differed significantly from those of KO (p < 0.05, Mann–Whitney test). Different dose groups are separated by vertical dotted lines. In both panels, N = 8–11 for each group except for TBG (n = 5–6), wt (n = 15–16), and KO (n = 14–17), and horizontal short black lines indicate the median.

Transduction and TYMP expression were virtually restricted to liver with all vectors tested

Eight months after treatment, hcTYMP DNA copy number was assessed by quantitative real-time PCR in DNA extracted from liver, brain, skeletal muscle, and small intestine, but was only detectable in liver (Fig. 3a), thus indicating poor transduction efficiency in the other tissues. TYMP expression was studied by determining TP enzyme activity (Figs. 3b and 4a) and TP protein levels using Western blot (Supplementary Fig. S2) and immunostaining (Supplementary Fig. S3).

As was expected, a clear correlation was observed in liver between the vector genome copy number per cell and the administered dose in all cases: r_(AAV-TBG) = 0.9548, r_(AAV-PGK) = 0.6598, r_(scAAV-HLP) = 0.7880, r_(AAV-AAT) = 0.9596 (p < 0.0001 for all vectors, Pearson correlation test). The copy number values obtained here were higher than those we previously reported,^23,24 probably due to differences in the calibration curve slopes or other experimental conditions in different settings. The results showed wide interindividual variation at each dose; sampling bias could account, at least in part, for such variability, since the distribution of the vector in liver is not homogeneous, especially for low doses (Supplementary Fig. S3). Nonetheless, the copy number of all vectors per cell was within the same range for each dose (with a slight, non-significant trend towards higher copy number for AAV-PGK and lower copy number for AAV-TBG).

TYMP expression was dependent on the promoter used rather than the vector genome copy number in liver. In treated animals, liver TP activity increased in a dose-dependent manner (Fig. 3b), and Western blot and immunohistochemical analyses results were consistent with the enzyme activities (Supplementary Figs. S2a and S3). The immunostaining images revealed a patchy distribution of the expression for low doses, while a more homogeneous distribution is seen for higher doses, for all vectors. At doses of 2 × 10¹¹ vg/kg and higher, AAV-TBG and AAV-AAT showed similar efficacy in providing TP activity to the liver, which reached values 60-fold higher than wt levels at the highest vector doses. The efficacy of scAAV-HLP was slightly lower, and AAV-PGK, despite being the vector with the highest copy number, was clearly less effective: AAV-PGK restored TP activity to values similar to those in wt mice only at the highest dose (2 × 10¹² vg/kg). The results obtained at the lowest dose (5 × 10¹⁰ vg/kg) confirmed that AAV-AAT was more effective than AAV-TBG (p < 0.005, Mann–Whitney test): 7 of 8 animals treated with AAV-AAT showed above-normal liver TP activity, whereas only 2 of 6 animals treated with the AAV-TBG vector reached TP activities within the wt range.

When TP activity was normalized by vector copy number, the distribution of the ratios revealed that AAT and TBG are stronger promoters than HLP and PGK, but there was no evidence that AAT is stronger than TBG (Fig. 3c). In any case, blood dThd levels at the end of the study were, at the lowest vector dose, lower for the AAV-AAT vector compared with the AAV-TBG vector, similarly to what was observed in the in vivo monitoring (Fig. 1). Despite the high TP activity provided by the different vectors in liver tissue, especially the AAV-AAT vector, only small to moderate transient elevations of circulating alanine aminotransferase (ALT) above levels of untreated animals were detected in a few mice, indicating that the treatment was not associated with hepatotoxicity (Supplementary Fig. S4).

Of note, the product of the transgene was detected in tissues other than liver (despite undetectable hcTYMP DNA) in some mice treated with vectors carrying liver-specific promoters. This contrasts with the lack of effect of the constitutive PGK promoter, which did not lead to TP activity above the levels in untreated mice in non-hepatic tissues (Fig. 4a and Supplementary Fig. S2b).

The treatment restores mitochondrial dNTP homeostasis in liver

In MNGIE, nucleoside overload causes a dNTP imbalance that ultimately interferes with proper mtDNA replication. Untreated double KO mice showed expanded dTTP levels in liver mitochondria, as compared to levels in age-matched wt counterparts (p < 0.0001, Student's t-test). After treatment, dTTP increase was prevented, paralleling the dThd reductions observed in liver (Fig. 5). Although dCTP levels were not significantly low in untreated double KO mice, treatment with all vectors induced a slight but significant increase in liver dCTP, except at the lowest dose (p < 0.05, Mann–Whitney test). Finally, the slight changes observed in the dGTP pool are unlikely to be physiologically relevant, as dGTP remained within the normal range in most treated animals with all doses tested.

Figure 5.

Mitochondrial deoxyribonucleoside triphosphate pools in liver. Thymidine triphosphate (dTTP) was significantly higher in the KO group than in wt mice (p < 0.0001, Student's t-test). Treatment significantly decreased dTTP levels in all groups (p < 0.01, Mann–Whitney's test) except for TBG at the 5 × 10¹⁰ vg/kg dose. The treatment slightly but significantly increased dCTP levels in all groups except for the TBG, HLP, and AAT promoters at the lowest dose (p < 0.05, Mann–Whitney test). dGTP was slightly but significantly lower in all groups treated with the AAT promoter (except the 2 × 10¹¹ vg/kg dose, p < 0.05, Mann–Whitney test), but was increased in mice treated with the TBG promoter at 5 × 10¹¹ vg/kg (p = 0.03, Mann–Whitney test). In all panels, n = 8–11 for each group except for TBG (n = 5), wt (n = 12–13), and KO (n = 16). Different dose groups are separated by vertical dotted lines. Yellow area corresponds to the wt concentration range, and horizontal short black lines indicate the median.

It should be mentioned that although the double KO mouse is a good biochemical model that recapitulates the biochemical imbalances observed in patients, it does not reproduce other molecular and clinical features, such as mtDNA depletion.⁴ We did not find mtDNA depletion in liver and, consistently, no effect of the treatment on this parameter was detected (Supplementary Fig. S5).

Discussion

Several experimental approaches developed in the past few years have indicated that gene therapy is feasible for treating MNGIE through use of an integrative lentiviral vector targeting hematopoietic progenitors^20,21,33 or an AAV8 vector transcriptionally targeted to liver.^23,24 These strategies were developed to provide an alternative to the currently available treatment options, allogeneic hematopoietic stem cell transplantation^14–16 and orthotopic liver transplantation,^17,18 which are associated with significant risks and difficulties, such as the need to find compatible donors.

Here, we attempted to improve the gene therapy strategy based on AAV vectors. We focused on these vectors because they show very low rates of integration into the host genome (relative to lentiviral vectors), and because similar AAV-based strategies have yielded good results in clinical studies.^34–36 As AAV-mediated TYMP expression has proven effective for clearing systemic accumulation of dThd and dUrd nucleosides causing MNGIE,^23,24 the next effort in this line would be to minimize the vector dose required for safety reasons. AAV administration can lead to the development of a detrimental immune response associated with short-lived transgene expression, as reported in some preclinical and clinical studies.^{25,35,37–39} An effective strategy to mitigate AAV vector immunogenicity in clinical practice has been reduction of the therapeutic dose via expression cassette optimization.³⁴ Additionally, smaller doses would lower the risk of vector insertion in the host genome and genotoxicity.⁴⁰ Dose reduction responds to economic considerations as well, as AAV manufacturing costs are high.

It is important to note that the predicted transduction efficiency of AAV in human liver is expected to be lower than that in mice,^41,42 which further supports the need to improve the biochemical efficacy of the vector. This difference between mice and humans should also be considered in relation to potential safety issues, because it suggests that the biochemical overcorrection observed in the treated murine model (dThd and dTTP levels below normal values) is not expected to occur in humans, as is discussed below.

Our results indicate that AAV-AAT is the most effective vector among those tested in this study, significantly improving the results of the previously reported AAV-TBG vector.^23,24 Use of the AAT promoter led to the highest TP activity in liver of treated mice and the fastest and most marked reductions in dThd and dUrd levels in blood and tissues. One week after treatment administration, AAV-AAT was the only vector that achieved normalized blood nucleoside levels at the lowest dose tested (5 × 10¹⁰ vg/kg), a considerable improvement in efficacy over that of the AAV-TBG vector.²³ However, when TP activity results are normalized by the vector genome copy number, the results support that AAT is more effective than PGK and HLP as a promoter, but there is no evidence that AAT is stronger than TBG. It should be noted that the wide dispersion of the TP activity/copy number ratios make these values very imprecise as estimates of the promoter strengths. In addition, for AAT, TBG, and HLP promoters, these ratios decrease at higher doses (which correspond to higher vector copy numbers), suggesting that the efficiency of the transgene expression is lower for higher vector copy numbers. Other groups have reported similar observations, which could be related with the saturation of the expression machinery or with epigenetic factors such as repeat-induced gene silencing.^31,36 Overall, the results do not allow us to conclude that the improved results observed for the AAV-AAT vector versus the AAV-TBG vector are due to a better efficacy of the AAT promoter.

Interestingly, the scAAV-HLP vector, which also acts under control of the alpha-1-antitrypsin promoter, did not provide better results than the single-stranded AAV-AAT vector, suggesting that the self-complementary configuration did not offer any advantages for transgene expression or function. This observation should be interpreted taking into account that the HLP promoter is a modified (smaller) version of the AAT promoter³¹ due to the size restrictions of the self-complementary vector. Also due to size limitations, the scAAV-HLP vector does not contain intron sequences, which are reported to contribute to mRNA protection and stability.⁴³ The self-complementary configuration favors faster expression of the transgene because scAAV vectors do not depend on cell replication machinery for their expression and are reported to be more stable and persistent within the host cell,⁴⁴ but the positive effect depends on the specific transgene used in each case.⁴⁵ In addition, systemic administration of scAAV vectors is known to enhance both the innate and adaptive immune response in a dose-dependent manner.^37,46 In agreement with this concept, we found that the dose-dependency of the effect of the scAAV-HLP vector was weaker than that of AAV-AAT, especially when TP activity in liver was examined. Taken together, these findings discourage use of the self-complementary configuration for a proposed AAV vector for MNGIE treatment.

In addition to increased effectiveness, selection of a specific enhancer/promoter sequence has an influence on the incidence of malignancies after AAV gene delivery. In a recent study, AAT and TBG promoters both induced AAV integration in the Rian locus in mice, but use of the AAT enhancer/promoter did not lead to overexpression of local genes, unlike the TBG promoter.⁴⁰ It is worth mentioning that TYMP can act as a suicide gene in TYMP-expressing tumors when capecitabine is used, as the TP enzyme is needed to activate this prodrug; this would be an additional safety mechanism for treated patients.⁴⁷

Use of the PGK promoter did not succeed in providing TP activity to organs other than liver. The main reason why we tested a constitutive promoter was to achieve transgene expression in many different tissues to evaluate dose reductions. PGK was expected to induce TYMP expression in tissues where AAV8 is efficiently transduced. AAV8 serotype has been reported to transduce tissues such as heart, skeletal muscle, and brain, although less efficiently in these tissues than in liver.⁴⁸ Therefore, we expected the AAV-PGK vector to induce transgene expression in tissues other than liver, thereby contributing to systemic nucleoside clearance. However, the AAV-PGK vector did not provide TP activity to nonhepatic tissues, which is consistent with the failure to detect hcTYMP DNA in these tissues. This result agrees with the findings of a study in which genome copies in liver were ≥3 log greater than the values in other tissues after AAV8-mediated transduction.⁴⁸ Furthermore, the performance of AAV-PGK in liver was the poorest of all the vectors tested, in agreement with the fact that AAT and TBG are liver-specific promoters and more effective in this organ than the constitutive PGK promoter.

Interestingly, slight or moderate elevations of TP activity (and TP protein detected by Western blot) were observed in brain, skeletal muscle, and small intestine of some mice treated with vectors having liver-specific promoters, especially AAV-AAT at high vector doses. Since hcTYMP DNA copy number was undetectable in these tissues, we tested different DNA extraction methods in an attempt to maximize the yield of extraction for episomal DNA. Nonetheless, we were only able to detect hcTYMP DNA in few of these TP-positive samples by either qPCR or conventional PCR (data not shown). Similar negative results were obtained when we measured TYMP mRNA by qPCR in non-hepatic tissues with the highest TP activities in AAV-AAT treated mice, (i.e., RNA levels were undetectable in most samples and barely detectable [negligible] in few isolated cases [data not shown]). As increased TP activity must be derived from expression of the transgene in effectively transduced cells, we have to assume that the methods we used to detect TP protein and activity are more sensitive than those used to detect hcTYMP DNA. Illegitimate expression of such promoters has been reported; endogenous α-1-antitrypsin transcription is most abundant in liver cells, but it also occurs in a variety of other tissues, such as kidney, gastrointestinal tract (stomach and small intestine), pancreas, brain, adrenal glands, testes, blood monocytes, and tissue-specific macrophages.^49–51 Alternatively, the presence of TP protein in these nonhepatic tissues could be derived from exosomal export from liver, either as mRNA or protein. In fact, TP protein has been found in exosomes derived from nasopharyngeal carcinoma cells, platelets and thymus.⁵² Our observations also support that the liver-specific TBG and AAT promoters we used would be more powerful in these tissues than the PGK constitutive promoter.

In MNGIE, dThd accumulation causes a dNTP imbalance, which is the direct biochemical interference mechanism impairing mtDNA replication.^5,6,53 The results from the present study confirm that systemic dThd reduction lowers the mitochondrial dTTP content, as was observed in our previous studies.^21,24 The high TP activity achieved in liver, well above the normal values in mice, led to systemic dThd levels even below normal values, and the resulting mitochondrial dTTP levels in liver were also excessively reduced. The same effect has been reported in a long-term study of the AAV-TBG vector.²⁴ This observation may raise some concerns because of the risk of malignancies or other undesired effects on nuclear or mitochondrial DNA. There were no cases of hepatocarcinoma in the present study (mice killed at age 10 months) and our previous long-term AAV-TBG study did not show an increased incidence of liver tumors in animals treated with either dose and killed at 21 months of age.²⁴ Other factors support the notion that the risk of excessive reductions in humans is lower than that in mice. First, normal dThd and dUrd levels in humans are below the lower limit of detection of commonly used methods for nucleoside assessment (0.05 μmol/L in our study¹²). This means that levels reached in mice after treatment, although they are below normal murine values, are still above normal human values. Second, while nucleoside levels show a 20- to 500-fold increase in MNGIE patients,⁵⁴ the murine disease model exhibits moderate increases (only around 4-fold). Therefore, it is expected that the quantitative effect of treatment on nucleoside levels would be less pronounced in patients than in mice. Third, TP activities in mice treated with high vector doses were well above normal murine levels (25–75 nmol/h/mg prot) but were within the normal or only slightly higher than normal TP activity range in humans (1800–2000 nmol/h/mg prot³⁰). Finally, AAV transduction is reported to be around 20-fold higher in murine liver than in human hepatocytes,^41,42 which implies that the doses tested in this study are likely to promote lower TP activities in liver of MNGIE patients than those observed in mice. Taken together, these considerations point out relevant differences in nucleoside metabolism and AAV treatment response between mice and humans and suggest that the dThd excessive reduction and dTTP depletion observed in mice will not occur in humans.

In this study, all vectors tested restored TP activity in liver and reduced systemic nucleoside overload, but the construct including single-stranded DNA and the TYMP transgene under control of the AAT promoter was the most effective. These results further support the feasibility of using an AAV vector transcriptionally targeted to the liver as therapy for MNGIE, and provide evidence that the previously tested vector^23,24 can be improved if the transgene is under control of the AAT promoter. This strategy should be considered for clinical use in MNGIE, a severe disorder with limited treatment options.

Author Disclosure

No competing financial interest exist.

Supplementary Material

Supplementary Figure S1

Supplementary Figure S2

Supplementary Figure S3

Supplementary Figure S4

Supplementary Figure S5