Glucose metabolism during fasting is altered in experimental porphobilinogen deaminase deficiency

Abstract

Porphobilinogen deaminase (PBGD) haploinsufficiency (acute intermittent porphyria, AIP) is characterized by neurovisceral attacks when hepatic heme synthesis is activated by endogenous or environmental factors including fasting. While the molecular mechanisms underlying the nutritional regulation of hepatic heme synthesis have been described, glucose homeostasis during fasting is poorly understood in porphyria. Our study aimed to analyse glucose homeostasis and hepatic carbohydrate metabolism during fasting in PBGD-deficient mice. To determine the contribution of hepatic PBGD deficiency to carbohydrate metabolism, AIP mice injected with a PBGD-liver gene delivery vector were included. After a 14 h fasting period, serum and liver metabolomics analyses showed that wild-type mice stimulated hepatic glycogen degradation to maintain glucose homeostasis while AIP livers activated gluconeogenesis and ketogenesis due to their inability to use stored glycogen. The serum of fasted AIP mice showed increased concentrations of insulin and reduced glucagon levels. Specific over-expression of the PBGD protein in the liver tended to normalize circulating insulin and glucagon levels, stimulated hepatic glycogen catabolism and blocked ketone body production. Reduced glucose uptake was observed in the primary somatosensorial brain cortex of fasted AIP mice, which could be reversed by PBGD-liver gene delivery. In conclusion, AIP mice showed a different response to fasting as measured by altered carbohydrate metabolism in the liver and modified glucose consumption in the brain cortex. Glucose homeostasis in fasted AIP mice was efficiently normalized after restoration of PBGD gene expression in the liver.

Introduction

Acute intermittent porphyria (AIP, OMIM 176000) is an autosomal dominant metabolic disorder caused by a hepatic deficiency in the porphobilinogen deaminase enzyme (EC 4.3.1.8; PBGD) (1–5). The main clinical manifestation of AIP includes acute neuropsychiatric attacks when hepatic heme synthesis is activated by endogenous or exogenous precipitating factors, such as fasting, reproductive hormones, infection, drugs, alcohol or stress. These factors reduce the availability of the regulatory heme pool in the liver and this results in the up-regulation of the rate-limiting enzyme for the heme synthesis pathway, aminolevulinate synthase (ALAS1) (6,7). The acute attack is invariably accompanied by excess formation and renal excretion of the porphyrin precursors, aminolevulinic acid (ALA) and porphobilinogen (PBG), the metabolites prior to the deficient PBGD step. Intravenous administration of human hemin restores the regulatory heme pool in the liver and it is indicated in severe or prolonged attacks (1,2,6).

Early studies recognized that caloric deprivation precipitates acute attacks and showed that nutritional supplementation attenuates the severity of these attacks (8). The molecular mechanism of fasting in precipitating the crises of acute porphyria has been described in mice, with the description of the ability of peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1α) to activate the expression of hepatic ALAS1 (9). Indeed, PGC-1α is a coactivator of nuclear receptors and other transcription factors (10) involved in fatty acid oxidation, tricarboxylic acid (TCA) cycle flux, mitochondrial oxidative phosphorylation and gluconeogenesis (9–11). Fasting also induces hormonal and nutritional cues that play a role in maintaining whole-body glucose homeostasis and promoting energy balance (12).

While the molecular mechanisms underlying the nutritional regulation of hepatic heme synthesis are well known, glucose homeostasis during fasting are poorly understood in porphyria. In this study, body glucose consumption and hepatic carbohydrate metabolism were characterized in a mouse model of AIP (13) during the early stages of fasting. To evaluate the effect of PBGD deficiency on glucose homeostasis, we also included a cohort of AIP mice where the PBGD expression was restored specifically in the liver by the administration of a recombinant adeno-associated virus vectors of serotype 5 (rAAV5-PBGD) (14).

Body glucose consumption was assessed using the radiotracer 2-[¹⁸F] fluoro-2-deoxy-d-glucose (¹⁸F-FDG). The rationale for its use is based on the principle that this glucose analogue is taken up in the cells and becomes trapped after phosphorylation; thus, the measured signal approximates uptake of glucose (15,16). AIP mice showed reduced ¹⁸F-FDG uptake in the brain cortex after 14 h of fasting which could be reversed by PBGD-liver gene delivery. Given that brain is one of the organs most responsive to changes in the availability of glucose during fasting, metabolomic analyses in serum and liver samples were subsequently performed to understand blood glucose homeostasis and hepatic carbohydrate metabolism during fasting. Glycogen breakdown was the main mechanism to maintain blood glucose homeostasis in wild-type mice while hepatic gluconeogenesis and ketogenesis become major energy sources in AIP mice after 14 h of fasting. Restitution of PBGD gene expression in the liver of AIP mice blocked ketogenesis and restored glycogen breakdown.

Results

Reduced ¹⁸F-FDG uptake in brain cortex of the AIP mouse was recovered upon restitution of PBGD gene expression in the liver

The ¹⁸F-FDG uptake in the whole body was visualized by means of positron emission tomography (PET) imaging in wild-type (n = 8) and AIP mice (n = 8) after 14 h of fasting (Supplementary Material, Fig. S1). To quantitatively evaluate the PET images, mice were sacrificed and the brain, heart, kidney, liver, lung and spleen were extracted to measure specific ¹⁸F-FDG retention in each organ (Supplementary Material, Table S1). A decrease in ¹⁸F-FDG level was significant (P < 0.05) in AIP mouse brain compared with wild-type animals, but no difference was observed in other organs.

Given that brain functions depend almost exclusively on the availability of glucose and different brain regions have different energy requirements, differences in the regional brain ¹⁸F-FDG uptake were evaluated with statistical parametric mapping (SPM) of 3D autoradiography. SPM analysis of cerebral 3D autoradiography highlighted reduced ¹⁸F-FDG uptake in the primary somatosensorial cortex of AIP mice (n = 6) when compared with wild-type animals (n = 6) (Fig. 1A and B) and to AIP mice upon PBGD gene delivery leading to liver-specific over-expression of the protein (n= 3) (Fig. 1C and D). Moreover, no cortical differences were detected between wild-type animals and AIP mice undergoing PBGD-liver gene delivery (image not shown). These data indicate that cortical ¹⁸F-FDG uptake was efficiently recovered after PBGD over-expression in the liver of AIP mice. To validate the SPM results and compare ¹⁸F-FDG uptake within the neocortex area, a separate volume of interest analysis was performed over 3D-autoradiographic volumes using a volume of interest map described by Ma and co-workers (17). The percentage of ¹⁸F-FDG uptake was lower in AIP mice (31 ± 2%) when compared with both wild-type (34 ± 1.6%; P = 0.04, Mann–Whitney's U-test) and AIP mice treated with PBGD-liver gene delivery (34 ± 1.8%; P = 0.04, Mann–Whitney's U-test).

Figure 1.

A voxel-based statistical analysis of 3D autoradiographic brain images that anatomically localizes regions with low ¹⁸F-FDG uptake between pairs of groups. Briefly, all autoradiographic slices from each brain were digitized and aligned overlaid on a canonical MRI scan to create a 3D volume stack. Then, all the statistical 3D volume-rendered images corresponding to the same group of animals (wild-type, AIP or AIP treated with PBGD-liver gene therapy) were averaged to create an image representing the mean value of ¹⁸F-FDG uptake in the different voxels. Finally, this average value for each voxel was statistically compared with the same voxel from the comparison group, giving a statistical image that anatomically localized significant changes between pair of comparisons. (A) Statistical 3D image showing in yellow brain areas with a significant (P < 0.01) decrease in the ¹⁸F-FDG uptake in AIP mice (n = 6) compared with the wild-type animals (n = 6). (B) Representative coronal slices corresponding to brain sections indicated in (A) (right) that anatomically localize over a canonical MRI scan the ¹⁸F-FDG uptake differences and quantify fold-change decreases in AIP brains according to the colour code scale flanking the 3D volumetric brain image. (C) 3D volume-rendered image and (D) representative coronal slices corresponding to brain sections indicated in (C), showing reduced ¹⁸F-FDG uptake areas when AIP mice (n = 6) were compared with those AIP animals treated with PBGD-liver gene therapy (n = 3). AIP mice showed reduced ¹⁸F-FDG uptake in the primary somatosensory cortex (yellow areas in B4-6 and D1-6) and also in the posterior parietal cortex and primary and secondary auditory and visual cortices (yellow areas in the middle and lower right side of D6-8). The posterior parietal cortex corresponds to a high-order associative cortex that receives information from primary and secondary somatosensory, auditory and visual cortices.

Given the probable role of ALA and PBG accumulation in the occurrence of neurological symptoms in AIP, the urinary excretion of these porphyrin precursors was quantified in AIP mice after 14 h of fasting. Caloric deprivation for this short period induced hepatic ALAS1 expression (data not shown). However, ALAS1 activation only resulted in a small increase in the excretion of heme precursor in AIP mice (ALA from 44 µg ALA/mg creat., range [15–78] to 52 µg ALA/mg creat., range [39–62]; and PBG from 19 µg PBG/mg creat., range [7–47] to 74 µg PBG/mg creat., [35–129]).

In order to understand the role of the hepatic PBGD deficiency in glucose homeostasis, we characterized blood glucose levels and metabolomics profiles in the serum and liver of AIP mice in response to fasting.

Increased serum ketone bodies in AIP mice subjected to 14 h of fasting

Metabolomic analysis of serum samples from wild-type mice before and after the 14-h fasting period showed an increase in circulating levels of pyruvate (Fig. 2A), lactate (Fig. 2B) and alanine (Fig. 2C). These metabolites are substrates for glucose production and their increased serum levels are the natural metabolic response of the body to maintain normoglycemia during fasting (18). Serum levels of glycerol fell (Fig. 2D) and ketone body concentration (Fig. 2E and F) were unchanged after 14 h of fasting in wild-type mice.

Figure 2.

Metabolomic changes in the serum of AIP mice after 14 h fasting period. The effects of fasting on serum concentration of (A) pyruvate, (B) lactate, (C) alanine, (D) glycerol, (E) hydroxybutyric acid and (F) acetone in wild-type animals and AIP mice treated or not with PBGD-liver gene therapy. Main gluconeogenic precursors are lactate, glycerol (which is a part of the triacylglycerol molecule), alanine and glutamine. Altogether, they account for over 90% of overall gluconeogenesis. The well-fed group corresponds to mice with free access to food. The fasting groups were animals subject to 14 h of fasting, from 8 p.m. to 10 a.m.

In contrast to wild-type mice, serum from fasted AIP mice showed high circulating levels of hydroxybutyrate (Fig. 2E) and acetone (Fig. 2F) when compared with well-fed AIP mice. Indeed, the serum of fasted AIP mice showed reduced levels of pyruvate (Fig. 2A), lactate (Fig. 2B) and alanine (Fig. 2C) when compared with fasted wild-type animals. These results suggest that AIP mice developed a different metabolic response to fasting when compared with wild-type mice.

Of interest, wild-type animals and AIP mice treated with PBGD-liver gene delivery presented similar circulating levels of these metabolites after 14 h fasting. These data indicate that the metabolic response to fasting in AIP mice returned to normal upon specific over-expression of the PBGD protein in the liver.

Hepatic gluconeogenesis and ketogenesis instead of glycogenolysis in 14 h fasted AIP mice

The liver of fasted wild-type showed reduced glycogen levels (Fig. 3A), indicating that glycogenolysis was the primary metabolic process activated in normal livers to maintain normal levels of hepatic glucose-6-phosphate (Fig. 3B) and normoglycemia (Fig. 4A) during fasting. Similar results were obtained in AIP mice treated with PBGD-liver gene delivery. In contrast, fasted livers from AIP mice showed unchanged glycogen levels (Fig. 3A) while intrahepatic levels of pyruvic acid (Fig. 3C) were markedly reduced when compared with both well-fed AIP animals and fasted wild-type mice. These data suggest that fasted AIP livers activate gluconeogenesis instead of glycogenolysis in order to maintain both hepatic (Fig. 3B) and blood (Fig. 4A) glucose levels within the normal range.

Figure 3.

Metabolomic changes in the liver of AIP mice after a 14 h fasting period. The effects of fasting on hepatic concentration of (A) glycogen, (B) glucose-6-phosphate, (C) pyruvic acid, (D) hydroxybutyrate (E) acetates and (F) ATP/ADP ratio in wild-type animals and AIP mice treated or not with PBGD-liver gene therapy. Well-fed correspond to mice with free access to food. Fasting were animals subject to 14 h of fasting, from 8 p.m. to 10 a.m.

Figure 4.

Blood glucose, glucagon and insulin levels in mouse strains showing different hepatic PBGD activity. (A) Glycemia was determined in fasted AIP, T2, T1 and wild-type strains showing 33, 55, 65 and 100% of normal PBGD activity, respectively. (B) Glucagon and insulin levels were measured in the serum of AIP and wild-type mice after 14 h fasting. Two months before serum extraction, a cohort of AIP and wild-type mice were injected with 5 × 10¹² gc/kg (i.v.) of an AAV vector carrying PBGD gene or the luciferase gene marker (luc) driven by a liver-specific promoter (14). (C) Glucose tolerance tests were conducted in fasted AIP, T2, T1 and wild-type strains and (D) in fasted AIP mice treated or not with PBGD-liver gene therapy.

Given that hepatic glycogenolysis is regulated by glucagon (19) and insulin (20), circulating levels of these hormones were measured after 14 h of fasting. AIP mice showed high serum insulin and low serum glucagon levels (Fig. 4B) even in normoglycemia (Fig. 4A). This insulinemia seems to be associated with the hepatic PBGD deficiency in AIP mice because its circulating level partially tended to recover wild-type values upon restitution of PBGD expression in the liver (Fig. 4B). The effect of the viral vector was ruled out because the administration of the same vector carrying luciferase gene marker (rAAV5-luc) did not modify insulin levels in either AIP or in wild-type mice (Fig. 4B).

Hepatic PBGD deficiency was also associated with a delayed glucose tolerance test (GTT) in the three transgenic mouse strains generated by gene targeting (13). GTT was most delayed in the AIP strain (T1/T2 compound heterozygote), while T1 and T2 strains showed an intermediate delay between wild-type and AIP mice (Fig. 4C). In addition, PBGD over-expression in the liver of AIP mice by means of a PBGD gene delivery vector (Fig. 4D) resulted in a GTT similar to that obtained in wild-type animals (Fig. 4C). All these data support some degree of insulin resistance associated with hepatic PBGD deficiency.

Finally, the liver of fasted AIP mice showed increased levels of hydroxybutyrate (Fig. 3D) and acetate (Fig. 3E), suggesting ketosis activation as an alternative energy source to the observed inhibition of the glycogenolysis pathway. In this way, the energy state of the livers was measured by the intrahepatic ATP/ADP ratio (Fig. 3F). While fasting shifted the equilibrium to ATP synthesis in the liver of wild-type animals, the ATP/ADP ratio was unchanged in AIP mice. Of interest, restitution of PBGD expression in the liver of AIP mice blocked ketosis activation (Fig. 3D and E) and restored a normal ATP/ADP ratio (Fig. 3F) in response to fasting. These data were in accordance with the observed restoration of hepatic glycogenolysis (Fig. 3A).

Discussion

Metabolic differences in glucose homeostasis in AIP mice compared with wild-type were described after a 14 h fasting period. Caloric deprivation for this short period slightly induced the expression of hepatic ALAS1 resulting in a small increase in urinary PBG excretion, while ALA excretion remained unchanged. Indeed, all groups of mice showed normoglycemia and hepatic glucose-6-phosphate levels within the normal range, suggesting that the fasted livers accomplished a good glucose homeostasis. However, wild-type mice responded to fasting by activating hepatic glycogenolysis and increasing serum glucose precursors (lactate, pyruvate and alanine). In contrast, AIP mice were unable to activate glycogen breakdown and reacted to nutritional deprivation by stimulating hepatic gluconeogenesis and ketogenesis.

The rate-limiting step in the hepatic glycogen degradation is catalyzed by the glycogen phosphorylase isozyme (EC 2.4.1.1). Glucagon activates hepatic isozyme while insulin induces its dephosphorylation and concomitant inactivation (21). Given that circulating levels of glucagon in AIP mice are already low after 14 h of fasting, it is feasible to contemplate a delayed activation of the glycogenolysis caused by a lack of glucagon-activation of glycogen phosphorylase isozyme. Altered serum levels of insulin and glucagon appear to be associated with deficient hepatic PBGD activity in AIP mice because PBGD-liver gene delivery tended to normalize its circulating levels and restored hepatic glycogenolysis in response to fasting.

Hepatic PBGD deficiency was also associated with a delayed glucose peak in the GTT suggesting that AIP mice have some degree of peripheral insulin resistance (22). As a consequence, the liver of AIP mice could fail to respond to the normal actions of the insulin altering important insulin-dependent signal transduction. In order to determine whether the defect was associated with the expression of the insulin receptor, its kinase activation or the subsequent signal transduction in the hepatocyte, we compared the amount of phosphorylated and non-phosphorylated insulin receptor in the liver of wild-type and AIP mice treated or not with liver gene therapy. Preliminary results suggest that porphyric hepatocytes showed a fully functional insulin receptor (data not shown) and further investigation will focus on the transactivation signal through the insulin receptor in porphyric hepatocytes (23).

In human porphyria, different studies performed before the establishment of hemin treatment confirmed abnormal carbohydrate metabolism in porphyria patients during an acute attack (24,25). Indeed, Waxman and colleagues (26) described abnormal GTTs in nine patients during the acute phase of the disease. Four of these also showed excessive serum insulin release.

Fasted AIP mice also showed increased levels of ketone bodies in serum and liver in spite of normoglycemia. This phenomenon suggests the activation of hepatic ketogenesis as an alternative energy source. This fast-induced ketogenesis allows porphyric organs to dispose of two different energy sources, glucose and ketone bodies.

Given the relative lack of local carbohydrate stores in the brain (glycogen, mostly in astrocytes), its energy requirements have been traditionally accepted to be supplemented by ketone body oxidation (27,28). The brain adapts progressively to use ketone bodies as an energy source and also as a substrate for lipid synthesis during prolonged starvation, strenuous exercise or a low carbohydrate diet (29). Thus, the reduced ¹⁸F-FDG uptake observed in the AIP brain might suggest a progressive adaptation of cortical neurons to use ketone bodies in AIP mice. Unfortunately, this hypothesis remains speculative and we cannot establish whether reduced glucose uptake in the brain cortex responds to changes in glucose/energy requirements in the brain cortex or altered glucose transport into the neurons caused by peripheral insulin resistance (30). However, since reduced glucose uptake was only observed in a specific region of the cortex, our data suggest that the passage of glucose through the blood–brain barrier and its transport into the majority of brain cell types were not altered in AIP mice.

Elevated serum lactate levels in well-fed AIP mice treated or not with PBG-liver gene delivery could suggest an activation of muscle lactic fermentation caused by mitochondrial energy defects in the muscle of AIP mice. It is difficult to establish whether this finding is linked to PBGD deficiency. However, Herrick et al. (31) reported high blood lactate levels in AIP subjects after glucose loading, which may be indicative of mitochondrial energy failure. A recent report demonstrated a close relationship between heme synthesis pathway activation, cataplerosis of the TCA cycle and mitochondrial energy failure in the liver of these AIP mice (32). The authors suggest that heme synthesis activation, as mediated by phenobarbital administration, results in a marked production and excretion of heme precursors causing a massive withdrawal of hepatic succinyl-CoA. The TCA cycle is then unable to supply reduced cofactors for the mitochondrial respiratory chain.

In our study, fasted AIP mice showed a low ATP/ADP ratio when compared with wild-type and AIP mice treated with PBGD-liver gene delivery. These data suggest that fasting could have an impact on energetic metabolism in porphyric livers. The slight increase in the excretion of porphyin precursors observed in our fasted AIP mice ruled out cataplerosis of the TCA cycle, and mitochondrial energetic failure related to succinyl-CoA availability during fasting, as recently reported in Homedan et al. (32). However, we cannot exclude reduced flux of the TCA cycle and impaired mitochondrial energy activation caused by withdrawal of pyruvate and acetyl-CoA which were redirected towards gluconeogenesis and ketogenesis (Fig. 5).

Figure 5.

Proposed model of glucose homeostasis in fasted AIP mice. Some degree of insulin resistance reduces glucose uptake and stimulates fat mobilization in fasted AIP mice. In the liver, fatty acids must be broken down into acetyl-CoA in order to obtain energy. Acetyl-CoA is not being recycled through the TCA cycle because the TCA intermediates (mainly oxaloacetate) have been depleted to feed the gluconeogenesis pathway. The resulting accumulation of acetyl-CoA activates ketogenesis. Ketone bodies are transported from the liver to other tissues, where they can be reconverted to acetyl-CoA to produce energy, via the TCA cycle. AIP mice showed glucose hypometabolism in the primary somatosensorial cortex which is reversed with PBGD-liver gene therapy (see Fig. 1). These findings raise the interesting possibility that activated hepatic ketogenesis might affect the activity of ketone-sensitive neurons.

In conclusion, porphyric livers respond to a short fasting period by activating gluconeogenesis and ketogenesis. Restitution of PBGD gene expression in the liver tended to normalize circulating insulin and glucagon levels, blocked ketogenesis and restored glycogenolysis. In the brain cortex, AIP mice showed reduced ¹⁸F-FDG uptake that could reflect neuron adaptation to ketone bodies as an energy source complementary to glucose.

Materials and Methods

Animals

Three- to five-month-old wild-type, homozygous T1 (C57BL/6-^{pbgdtm1(neo)Uam}, heterozygous T2 (C57BL/6-^{pbgdtm2(neo)Uam}) and compound heterozygous T1/T2(AIP) mice strains (13) showing 100, 65, 55 and 33% of hepatic PBGD activity, respectively, were used in this study. Only AIP mice replicate the drug-precipitated biochemical abnormalities of acute porphyria as experienced by humans and develop neuropathological features resembling those of patients with the inherited disease (13,14). The AIP strain was used in this study. Additionally, cohorts of homozygous T1 and heterozygous T2 mice were also included to determine serum glucose in the GTT analysis. The rAAV5 were generated and purified as previously described (14). The therapeutic PBGD gene (rAAV5-PBGD) or the luciferase gene marker (rAAV5-luc) is driven by a liver-specific promoter. Two months before analysis, a cohort of AIP mice received a single intravenous injection of 5 × 10¹² gc/kg of a rAAV5-PBGD vector (group AIP-GT) resulting in the restoration of normal PBGD expression in the liver (14). Experimental protocols were performed according to European Council Guidelines and approved by the local Animal Ethics Committee.

GTT

For GTT analysis, mice were fasted for 14 h (from 8 p.m. to 10 a.m.) and then injected with d-glucose (Hospira, IL, USA) (2 g per kg body weight, i.p.). Blood was collected from tail veins at 0, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 75, 90 and 105 min and glucose levels were measured with a glucose meter (Bayer). Glucagon (Sigma-Aldrich, MO, USA) and insulin (Merck Millipore, Germany) levels were measured by ELISA kits in serum blood samples taken in different cohorts of fasted mice.

¹H-NMR-based metabolomics on serum and liver samples

¹H-NMR spectra in either serum or liver samples were recorded at 310 K on a Bruker Avance III 600 spectrometer operating at a proton frequency of 600.20 MHz using a 5-mm CPTCI triple resonance (¹H, ¹³C and ³¹P). For serum samples, 100 μl were mixed with 300 μl of phosphate buffer (0.75 mm Na₂HPO₄ adjusted to pH 7.4, and 20% D₂O to provide the field frequency lock). The final solution was transferred to a 5-mm NMR tube and Carr-Purcell-Meiboom-Gill sequencing (CPMG, spin-spin T2 relaxation filter) was performed with a total time filter of 410 ms to attenuate the signals of serum macro-molecules to a residual level. For each sample, 128 transients were collected into 32 K data points using a spectral width of 12 kHz with a relaxation delay of 2 s and an acquisition time of 1.36 s. A line-broadening function of 0.3 Hz was applied to all spectra prior to Fourier transformation.

In the case of liver samples, ∼50 mg of tissue were removed, flash frozen and mechanically homogenized in 2 ml of a cold mixture (CH₃CN:H₂O 1:1 v/v, T= 0°C, 5 min). Each homogenate was centrifuged at 5000 g for 15 min at 4°C. Supernatants (hydrophilic metabolites) were lyophilized overnight. For NMR measurements, the hydrophilic extracts were reconstituted in 600 μl D₂O containing 0.67 mm trisilylpropionic acid and transferred into 5-mm NMR tubes. The lipophilic extracts were subsequently extracted in 700 μl of a solution CDCl₃/CD₃OD (2:1) containing 1.18 mm tetramethylsilane and then vortexed, homogenized for 20 min, centrifuged for 15 min at 6000 g at room temperature and transferred into 5-mm NMR tubes. The 1D Nuclear Overhauser Effect Spectroscopy with a spoil gradient was used to record 1D ¹H-NMR spectra. A total of 256 transients were collected across a 12 kHz spectral width at 300 K into 64k data points, and exponential line broadening of 0.3 Hz was applied before Fourier transformation. A recycling delay time of 8 s was applied between scans to ensure correct quantification. The frequency spectra were phased, baseline corrected and then calibrated to either trisilylpropionic acid or tetramethylsilane (δ = 0.0 ppm), using TopSpin software (version 2.1; Bruker, Germany). Liver metabolite identification and quantification were performed as previously described (33).

Ex vivo autoradiography

Given that spatial resolution remains one of the most critical technical limitations in the assessment of the mouse brain by PET imaging, specific regional brain ¹⁸F-FDG uptake was evaluated with ex vivo autoradiography. This is a technique with a spatial resolution approximately one order of magnitude better than even the best small animal PET devices (100–200 μm versus 1–2 mm, respectively) (34). Brain metabolic activity was assessed in six wild-type and six AIP male mice. An additional group of three male AIP mice was injected with rAAV5-PBGD 3 months before the autoradiography procedure. After a fasting period of 14 h, animals were i.p. injected with ¹⁸F-FDG (18.4 ± 0.1 MBq; InstitutoTecnológico PET, Madrid, Spain) dissolved in 0.2 ml of 0.9% NaCl solution. Before the ¹⁸F-FDG injection, glucose blood levels were measured with test strips (OneTouch UltraEasy glucometer, LifeScan, USA). After a conscious uptake period of 45 min, the animals were sacrificed and their brains quickly removed and immediately frozen by immersion in cold isopentane on dry ice. Serial coronal sections (40 µm thick) from the olfactory bulb to the cerebellum were rapidly obtained at −20°C by using a cryostat (Leica CM1850, Nussloch, Germany). One of every three slices was collected yielding ∼80 sections per animal, at a 120 µm pitch. The brain slices were thaw-mounted onto Superfrost Plus microscope glass slides (Menzel-Glaser, Braunschweig, Germany). After collection, the slices were heat-dried at 37°C on a hot plate and then exposed together with ¹⁸F-FDG self-made standards to an autoradiographic film (Agfa Curix RP2 Plus, Mortsel, Belgium) for 1 h. Then, the films were manually developed and left to dry in a warm air stream. All the brain slices were digitally captured at a resolution of 1728 × 1296 pixels (Leica DC300F, Nussloch, Germany).

Creation of a 3D volume of autoradiography and voxel-based analysis

A 3D brain image reconstruction of autoradiographic data was conducted as previously published (34) with slight modifications. The autoradiographic brain images were preprocessed with the public domain ImageJ v 1.37 software (developed by Wayne Rasband at the National Institutes of Health, Bethesda, MD, USA) and available on the Internet at http://rsbweb.nih.gov/ij/download.html (accessed January 29, 2016). Briefly, first, a rectangular bounding box (820 × 620 pixels) was defined around each section. The image grey scale was then inverted and the background subtracted. A volume of interest was drawn as an isocontour over the autoradiographic images to generate a mask of cerebral areas. Then, serial adjacent autoradiographic coronal sections from each single brain were aligned and stacked using PMOD software (version 3.0; PMOD Technologies Ltd, Adliswil, Switzerland). The resultant 3D volumes were spatially normalized to a standard space using a magnetic resonance imaging (MRI) template of the mouse brain (17). To this end, an automatic registration between each single MRI scan and the three-dimensional brain volumes obtained from autoradiographic sections (3D autoradiography) was performed. Then a transformation matrix, previously calculated for MRI normalization, was applied to the autoradiographic 3D volume of the same animal (PMOD).

Statistical analysis

In order to localize significant differences in brain ¹⁸F-FDG uptake, a voxel-based analysis was performed using SPM software (SPM8, Wellcome Trust Centre for Neuroimaging, Institute of Neurology, University College, London, UK), that creates statistical images that localize significant changes on a whole-brain basis by applying a general linear model to each voxel. For SPM analyses, 3D autoradiographic volumes were smoothed with a Gaussian filter with an FWHM of 10 mm. Differences between groups were detected using a one-way ANOVA design using glucose levels as a nuisance covariate. In order to correct for intersubject differences in overall ¹⁸F-FDG uptake, images were normalized by proportional scaling to a mean value of 50. Parameters were estimated and contrasts were derived to create statistical parametric maps of T values, with a statistical threshold of P < 0.01 and an extent threshold of 30 contiguous voxels. Activated clusters were anatomically located using an MRI image and a volume of interest map (17). A statistical 3D volume-rendered image was formed with the mean value of ¹⁸F-FDG uptake of six wild-type mice in the different voxels composing the brain and was statistically compared with the average value in the corresponding voxels areas of the six AIP brains (Fig. 1). A statistical 3D volume-rendered image corresponding to a statistical comparison between AIP mice receiving or not PBGD-liver gene therapy is shown in Fig. 1C.

The non-parametric Mann–Whitney's U-test was used for comparisons between well-fed versus fasted animals in each group of mice. In the case of comparisons between the three groups (wild-type, AIP treated or not with PBGD-liver gene delivery), data were log transformed prior to ANOVA analysis to equalize variances and pairwise comparisons were made using Bonferroni's multiple comparison tests. Results are expressed as medians with ranges. Statistical analyses were carried out with GraphPad Prism 5.0 (GraphPad Software, Inc, La Jolla, CA, USA).

Funding

This work was supported in part by grants from Fundación de Investigación Médica Aplicada (CIMA—University of Navarra); Spanish Fundación Mutua Madrileña to A.F., M.B., A.S. and R.E.deS.; the European VII Framework Program—Project AIPGENE (grant FP7-Health-2010-261506 to A.F., A.S. and R.E.deS.) and Spanish Institute of Health Carlos III (FIS) co-financed by European FEDER funds (grant numbers PI061475, PI09/02639 and PI12/02785 to A.F., I.S-M, A.S. and R.E.deS.).

Acknowledgements

We are grateful to José Luis Lanciego, María Javier Ramirez, Xavier Correig, Carmen Unzu, Jesús Prieto and Matías Avila for scientific discussion and support. The authors also gratefully acknowledge Margarita Ecay and Izaskun Bilbao for their excellent work in the animal preparation and acquisition of the PET studies and Rubén Fernández de la Rosa for his assistance in the performance of autoradiographic images.

Conflict of Interest statement. The authors have declared that no competing interests exist. The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.

Author notes

†

M.C. and I.S.-M. contributed equally to this work.