Slc2a10 knock-out mice deficient in ascorbic acid synthesis recapitulate aspects of arterial tortuosity syndrome and display mitochondrial respiration defects

Abstract

Arterial tortuosity syndrome (ATS) is a recessively inherited connective tissue disorder, mainly characterized by tortuosity and aneurysm formation of the major arteries. ATS is caused by loss-of-function mutations in SLC2A10, encoding the facilitative glucose transporter GLUT10. Former studies implicated GLUT10 in the transport of dehydroascorbic acid, the oxidized form of ascorbic acid (AA). Mouse models carrying homozygous Slc2a10 missense mutations did not recapitulate the human phenotype. Since mice, in contrast to humans, are able to intracellularly synthesize AA, we generated a novel ATS mouse model, deficient for Slc2a10 as well as Gulo, which encodes for L-gulonolactone oxidase, an enzyme catalyzing the final step in AA biosynthesis in mouse. Gulo;Slc2a10 double knock-out mice showed mild phenotypic anomalies, which were absent in single knock-out controls. While Gulo;Slc2a10 double knock-out mice did not fully phenocopy human ATS, histological and immunocytochemical analysis revealed compromised extracellular matrix formation. Transforming growth factor beta signaling remained unaltered, while mitochondrial function was compromised in smooth muscle cells derived from Gulo;Slc2a10 double knock-out mice. Altogether, our data add evidence that ATS is an ascorbate compartmentalization disorder, but additional factors underlying the observed phenotype in humans remain to be determined.

Introduction

Arterial tortuosity syndrome (ATS, MIM#208050) is a rare autosomal recessive connective tissue disorder, characterized by severe patterning defects in blood vessel development. The main features of the disease include elongation, tortuosity and stenosis of the large and middle-sized arteries, with an increased risk for aneurysm formation and ischemic events (1–3). Patients may present with additional connective tissue-related features, such as cutis laxa, diaphragmatic hernias, joint laxity, distinctive craniofacial malformations and a marfanoid habitus. Finally, corneal thinning causing pellucid corneas and keratoconus was recently described (2,4). The disease has shown to be variable in terms of severity, ranging from early mortality to mild manifestations in adulthood (1,2,5,6). Arterial histopathology shows disorganization and fragmentation of the elastic fibers (1,7–10). Cultured fibroblasts of ATS patients showed disorganization of the actin cytoskeleton and multiple extracellular matrix (ECM) components, including fibronectin, fibrillin, type 3 and type 5 collagen and decorin (11,12).

ATS is caused by loss-of-function mutations in the SLC2A10 gene, encoding the GLUT10 protein (3). GLUT10 belongs to the GLUT family of facilitative transporters, consisting of 14 transmembrane proteins, enabling the transport of monosaccharides, polyols and other small carbon compounds across the membranes of eukaryotic cells (13). GLUT members show specific functionalities, but some functional redundancy has been observed. Recent evidence identified GLUT10 as a transporter of dehydroascorbic acid (DHA), the oxidized form of ascorbic acid (AA), but its subcellular localization remains enigmatic and may change upon physiological states. GLUT10 has been shown to reside in mitochondria (14), the nuclear envelope (3) and/or the endoplasmic reticulum (ER) (15–17).

It remains to be solved how defective DHA transport results in a vasculopathy related to Loeys-Dietz (18) or Marfan syndrome (19,20), respectively, caused by perturbed transforming growth factor beta (TGFβ) pathway signaling or altered ECM proteins. In the absence of relevant vascular patient material, several ATS disease models have been developed. Slc2a10 knockdown in a morpholino-based zebrafish model showed abnormal vascular patterning and mitochondrial respiration (21). Two knock-in mouse models, either carrying a homozygous likely pathogenic G128E or S150F missense substitution in Slc2a10 (22,23), were previously phenotyped by two different research groups. While no ATS-related abnormalities could be discerned in either mutant by one of both groups at 3 months of age (23), the other group found elastic fiber proliferation in older homozygous G128E mutant mice only (22). It is unclear whether this is related to the elastic fiber fragmentation seen in the vascular wall of ATS patients. Recent follow-up studies in this mouse model indicated altered redox homeostasis and mitochondrial dysfunction as pathogenic mechanisms underlying ATS (24), similar to observations in the morpholino-based zebrafish knockdown model (21) and in ATS knockdown cell lines (rat A10, mouse 3T3-L1) (14). Altered redox homeostasis contributes to the pathogenesis in other aneurysm-related diseases (25).

A growing evidence that GLUT10 functions as an AA transporter has led to the hypothesis that the innate AA synthesizing capacity of mice could rescue the phenotype. Therefore, we characterized a novel mouse model for ATS, with gene disruptions of both Slc2a10, the causal gene for ATS, and Gulo, which encodes L-gulonolactone, the enzyme catalyzing the last step in AA biosynthesis in mouse.

Results

Generation of the Slc2a10;Gulo double knock-out mouse model

We developed a novel mouse model for ATS in a mixed 129/SvEv-C57BL/6 genetic background by crossing constitutive knock-out lines for Gulo and Slc2a10, both harboring a selection cassette replacing multiple exons (Supplementary Material, Fig. S1A). To take into account any possible effect of a variable genetic background on the acquired data, all studies were performed on littermates from the F2 generation (Supplementary Material, Fig. S1B). RT-qPCR-based expression analysis, using primers targeting the sequence replaced by the selection cassette, confirmed the absence of wild-type (WT) Slc2a10 RNA in the skin and aorta obtained from Slc2a10^−/− animals (DKO and Slc2a10 KO) (Fig. 1). Double knock-out mice (DKO—Gulo^{tm1mae/tm1mae};Slc2a10^−/−) were born with expected Mendelian frequency and had a normal lifespan to at least 9 months of age, similar to their control littermates (Gulo KO—Gulo^{tm1mae/tm1mae};Slc2a10^+/+, Slc2a10 KO—Gulo^+/+;Slc2a10^−/− and WT—Gulo^+/+;Slc2a10^+/+). DKO mice developed normally and showed no obvious gross anomalies, skeletal dysmorphisms or skin abnormalities. However, the weight of female DKO mice was slightly reduced, at each measurement time point (Supplementary Material, Fig. S2).

Figure 1

Slc2a10 expression levels in skin and aortic tissue. RT-qPCR analysis identifies a near absence of Slc2a10 WT mRNA in mice homozygous for Slc2a10 mutations (Slc2a10 KO and DKO), compared to mice harboring the Slc2a10 WT allele (Gulo KO and WT). Displayed average values represent the relative expression data in the skin or aorta for three male and three female mice. Error bars represent the SD, ^*P ≤ 0.05; ^**P ≤ 0.01; ^***P ≤ 0.001.

Slc2a10;Gulo DKO mice do not completely copy the human ATS phenotype

We carried out serial ultrasound measurements at multiple locations in the cardiovascular system of 10 male and 10 female mice of each selected genotype (DKO, Slc2a10 KO, Gulo KO, WT) at 6 weeks and 3, 6 and 9 months of age. Diameters of the ascending aorta, distal ascending aorta, aortic arch and descending aorta showed a subtle but statistically significant decrease in DKO mice compared to their WT littermates, for a given age and sex (Fig. 2). More distal measurements (carotid arteries) did not reveal any significant diameter alterations (Supplementary Material, Fig. S3). Fractional shortening and ejection fraction were comparable between the selected genotypes at 6 weeks and 3 months of age (Supplementary Material, Fig. S4). We did not detect aortic or mitral valve insufficiency using pulsed Doppler. Detailed microscopic analysis of the aortic tree, eye arterioles and the circle of Willis, obtained from polymer replicas of the cardiovascular system at 9 months of age, did not indicate any omnipresent structural anomalies such as tortuosity, abnormal implantation of the aortic side branches or aneurysms. However, one out of three analyzed female DKO mice had a local stenosis in the vertebral artery (Fig. 3; Supplementary Material, Fig. S5). At 9 months, elastic fiber staining of aortic tissue showed mild age-related changes, such as sporadic elastic fiber fragmentation or accumulation, in the tunica media of the aortic wall across all selected genotypes.

Figure 2

Serial ultrasound measurements obtained in WT, Gulo KO, Slc2a10 KO and DKO mice at 6 weeks and 3, 6 and 9 months of age. Shown data were acquired at the level of the aortic root, proximal ascending aorta, distal ascending aorta and aortic arch. Left and right column images represent data obtained in male (M) and female (F) animals, respectively. Shown P-values were obtained using a linear mixed model with covariance pattern modeling, without interactions. As a result, P-values should be interpreted when making the comparison for a given age and sex. Error bars represent 95% confidence intervals. ^*P ≤ 0.05; ^**P ≤ 0.01.

Figure 3

Vascular corrosion casts of WT, Gulo KO, Slc2a10 KO and DKO female mice. Representative images of the aortic arch and its side branches, the circle of Willis and the eye arterioles are shown. A local narrowing in the vertebral artery in the circle of Willis could be observed in one DKO mouse (white-framed enlargement). Scale bars: 2 mm.

Picrosirius red (PR) polarization staining on the skin tissue showed no major alterations in collagen deposition, but patches of increased collagen deposition could be discerned in female DKO mice (Fig. 4; Supplementary Material, Fig. S6).

Figure 4

Histological analysis of WT, Gulo KO, Slc2a10 KO and DKO female mice. Picrosirius red (PR) polarization staining (upper row) for collagen reveals the presence of patches with accumulated collagen deposition (white arrowheads) in the skin of DKO mice. Verhoeff-Van Gieson (VVG) elastic fiber staining (middle row) shows mild elastic fiber anomalies in the aortic wall of all studied mice (black arrowheads).

Transmission electron microscopy (TEM) analysis of female skin samples showed a normal elastic fiber architecture (Supplementary Material, Fig. S7). Collagen fibril diameters were significantly reduced in DKO compared to WT mice (Fig. 5).

Figure 5

Electron microscopic quantification of collagen diameters in WT and DKO skin samples. DKO mice show decreased collagen fiber diameters, compared to WT mice (measurements on EM micrographs of two animals per genotype). TS: transverse section, LtS: longitudinal section. Scale bar upper row: 1.0 μm, lower row: 0.3 μm. Error bars represent 95% confidence intervals. ^***P ≤ 0.001.

In vitro analysis of Slc2a10;Gulo DKO vascular smooth muscle cells shows decreased ECM deposition

We examined the ECM deposition in vascular smooth muscle cells (VSMCs) of the DKO and WT genotypes (Fig. 6). In contrast to WT VSMCs, DKO VSMCs did not show a base network of recognizable and organized fibronectin fibers stretching between the cells but presented with a reduced intensity of fibronectin staining with a more fuzzy appearance (Fig. 6A; Supplementary Material, Fig. S8). A similar observation was made for fibrillin-1 where stained fibers appeared thinner in DKO VSMCs (Fig. 6B; Supplementary Material, Fig. S8). WT VSMCs showed clear staining of tropoelastin that appeared in a fibrillar structure (Fig. 6C), next to the presence of a strong fiber network of both fibulin-4 (Fig. 6D) and fibulin-5 (Fig. 6E). In DKO VSMCs, a reduced staining of fibulin-4 and fibulin-5 (Fig. 6D and E; Supplementary Material, Fig. S8) and an overlapping network of tropoelastin and fibulins (Fig. 6C; Supplementary Material, Fig. S8) could be observed. We further assessed LTBP-4, which was present in clear fibrillar structures (Fig. 6F) in both WT and DKO VSMCs but appeared less intense in DKO VSMCs, organized in a more densely packed fibril network (Fig. 6F; Supplementary Material, Fig. S8). Taken together, these data show that DKO VSMCs displayed decreased deposition of the various components of the ECM compared to WT VSMCs.

Figure 6

Immunofluorescence staining for fibronectin (A), fibulin-1 (B), tropo-elastin (C), fibulin-4 (D), fibulin-5 (E) and LTPB-4 (F) shows deposition of the ECM components after 7 days of culture. DKO VSMCs show less distinct ECM structures compared to WT VSMCs. Scale bars represent 100 μm.

Slc2a10;Gulo DKO VSMCs do not show altered activation of canonical TGFβ signaling

We investigated canonical TGFβ signaling activation in VSMCs derived from WT and DKO mouse aortas. Under an unstimulated serum-deprived condition, DKO VSMCs showed similar levels of Smad2 phosphorylation compared to WT VSMCs (Fig. 7A and B). Following TGFβ stimulation, both WT and DKO VSMCs demonstrated a strong and comparable increase in phosphorylation of Smad2 after 15 min (Fig. 7A). Quantification of phosphorylated Smad2 to total Smad2 ratio revealed a further increase to its maximum at 60 min of stimulation after which it decreased (Fig. 7A and B). In conclusion, our data do not support aberrations in the activation of the canonical TGFβ signaling pathway in DKO VSMCs compared to control VSMCs.

Figure 7

DKO VSMCs do not show altered activation of the canonical TGFβ signaling pathway. DKO VSMCs do not show increased phosphorylation of Smad2 without TGFβ stimulation and do not differ from WT VSMCs in basal conditions. Stimulation with TGFβ led to increased phosphorylation of Smad2 in both WT and DKO VSMCs. There was no difference in pSmad2/Smad2 ratio between the two genotypes. The data show two independent cell lines from two experiments. The data points represent average values of an independent cell line per experiment. Results are depicted as mean ± SD.

Slc2a10;Gulo DKO VSMCs show a reduced maximum oxygen consumption rate

Since GLUT10 impairment has previously been associated with mitochondrial dysfunction in a Glut10 mouse model with missense mutations (24) and a slc2a10 zebrafish knockdown model (21), we analyzed the oxygen consumption rate (OCR) of DKO compared to WT VSMCs as a marker for mitochondrial function (Fig. 8A). Phase I represents the basal and unstimulated OCR. Addition of oligomycin in phase II inhibits the ATP synthase and reduces the OCR. In phase III, treatment with fluoro-carbonyl cyanide phenylhydrazone (FCCP) uncouples oxygen consumption from ATP production and raises OCR to its maximum. Addition of antimycin A in phase IV inhibits complex III of the mitochondria and blocks mitochondrial respiration. Figure 8B and C contains the quantification of the mean value of phases I and III per genotype. Basal OCR did not differ between both genotypes (Fig. 8A and B). However, when VSMCs were treated with FCCP to reach maximum respiration, DKO VSMCs displayed decreased maximum respiration rates compared to WT VSMCs (Fig. 8A and C).

Figure 8

DKO VSMCs show a reduced maximum OCR. In basal conditions, DKO and WT VSMCs do not differ in OCR, while treatment with FCCP results in a decreased maximum OCR of DKO VSMCs. Data represents two independent cell lines per genotype measured in three technical replicates. The data points represent average values of an independent cell line per experiment. Results are depicted as mean ± SD. ^**P ≤ 0.01.

Discussion

Arterial tortuosity may occur in both innate and acquired vasculopathies, such as persistent hypertension and neoangiogenesis, and may associate with an increased vascular risk, including aneurysm formation and blood vessel rupture. Monogenic conditions may be helpful in delineating the underlying pathophysiology of altered vascular patterning. In this study, we developed and characterized a novel mouse model for ATS, since the previously reported mouse models for ATS failed to recapitulate the major hallmarks of the human disease (22,23). Our model addressed two concerns regarding the previous models. Firstly, we developed an Slc2a10 knock-out, which abolishes GLUT10 function while the previously reported missense mutations were only predicted to be pathogenic (22,23). Secondly, we impaired AA synthesis in our model by disrupting the Gulo gene, aiming to mimic the physiological conditions in human more reliably. Indeed, Gulo encodes L-gulonolactone oxidase, an ER membrane-bound enzyme that catalyzes the last step in the vitamin C biosynthesis pathway (26). It became mutated in many species, such as teleost fishes, guinea pigs and anthropoid primates, including humans, making these species unable to synthesize vitamin C and rely on its dietary uptake (27). Vitamin C plays an essential role as an antioxidant and in collagen synthesis. Since recent evidence indicates that GLUT10 function may include transport of DHA into subcellular compartments (14,17,24), it could be hypothesized that murine developmental processes do not suffer from Slc2a10 mutations and localized vitamin C hypovitaminosis, which would explain the absence of a relevant phenotype in Slc2a10 mutant mice (16,23).

Nevertheless, in this study, we only observed mild cardiovascular manifestations in the DKO ATS mouse model. Ultrasound-guided characterization revealed a subtle stenosis from the proximal ascending aorta to the descending aorta in DKO mice but only statistically significant when compared with their WT littermates. Moreover, the reduced weight in the DKO mice further questions the biological relevance of these results.

Histopathology of ATS patient vascular tissue invariably shows substantial fragmentation of the elastic laminae in the tunica media (1), while conversely the previously described Glut10^G128E mouse model mainly presented with increased arterial wall thickness, most prominent at the age of 15 months (22,24). Our model only reveals mild elastic fiber anomalies in all genotypes that are likely age-related changes. Nevertheless, dermal collagen histology revealed patches of increased collagen deposition in female DKO mice, which has also been observed in skin tissue of ATS patients (1) and might contribute to their ‘doughy’ skin characteristics. Ultrastructural analysis of skin tissue of DKO mice additionally revealed collagen fiber diameter alterations in DKO mice, but did not reveal major defective elastin deposition, which contrasts with TEM observations in ATS patients (1).

The observed mild phenotypic abnormalities seem to be more pronounced in female mice, suggesting a sex discrepancy, even for nulliparous mice. Influences of sex in connective tissue homeostasis and disease have been previously recognized. However, most studies indicate that male mouse models for connective tissue disorders show more severe phenotypes than female mice (28–31), which contrasts with our study.

Despite the absence of a major clinical or histological phenotype, we did identify defective ECM deposition in cultured VSMCs of DKO mice. AA is a cofactor for cross-linking enzymes such as lysyl oxidase and lysyl oxidase-like enzymes. Hence, it is not surprising that elastin and collagen assembly are primarily affected. Nevertheless, our observations also indicate reduced deposition of the early-stage fibril-forming networks including the fibronectin and fibrillin-1 microfibrillar scaffolds. This reflects an earlier observation that in the absence of AA, not only deposition of collagen but also of fibronectin is greatly reduced (32). Indeed, collagen I fibers interact preferentially with relaxed (but in a lesser extent to stretched) fibronectin fibers (32), and collagen 1 increases fibronectin deposition (33). Therefore, ECM assembly cannot be seen as a consequential process but rather as an interactive and growing mechano-sensing network of components. Similarly, fibulin-4 and fibulin-5 deposition was clearly reduced, while LTBP4 deposition was only slightly affected. Though further experimentation is necessary, we hypothesize that latent TGFβ-binding proteins might be upregulated due to their primary role in elastogenesis (34). Our results are in line with findings by Zoppi et al. (12), who previously showed ECM disarray in skin fibroblasts of ATS patients. In addition, a recent study revealed an abnormal microfibrillar scaffold and incomplete elastic core assembly in skin samples of ATS patients (1) that likely reflect defective ECM assembly at multiple levels, as shown in our model. Since defective ECM assembly is often associated with vascular phenotypes (35,36), this might indicate that mice may rely on other mechanisms that overcome the deleterious defects of abnormal ECM assembly.

In several aneurysm syndromes, dysregulation of the canonical TGFβ signaling pathway is at play in aneurysm formation (37,38). However, while we observed ECM abnormalities, DKO VSMCs do not show altered phosphorylation of Smad2 upon TGFβ stimulation over time as well as under basal, unstimulated, conditions. An initial report indicated a prolonged signal increase for downstream readouts of TGFβ signaling in arterial tissue of one ATS patient (3). This finding remained unconfirmed on both fibroblast cultures (12) and skin or artery tissue (1) in an extended number of patients. A role for noncanonical αvβ3 integrin-mediated TGFβ signaling in the ATS pathomechanisms has been suggested (12). Our results add to the uncertainty regarding the role of TGFβ signaling in the pathomechanisms underlying many elastic fiber diseases, but cannot account for a possible contribution of spatiotemporal regulation of TGFβ signaling.

A number of studies have pointed towards a key role for oxidative stress in ATS pathogenesis. For instance, GLUT10 has been identified as a mitochondrial DHA transporter, and it was shown that increased mitochondrial targeting of GLUT10 and associated increased mitochondrial DHA uptake was triggered by stress and aging conditions (14,24). In a morpholino slc2a10 knockdown zebrafish model, altered OCR and differential gene expression analysis showed involvement in oxidative signaling (21). In the Glut10^G128E mouse model, Syu et al. (24) similarly identified a decreased maximum OCR in early passage Glut10^G128E VSMCs. In this study, we compared basal and maximal OCR in VSMCs of DKO and control mice and identified a normal basal OCR in DKO VSMCs. However, following treatment with FCCP, a decreased maximum OCR could be observed in DKO VSMCs, compared to WT. Since basal OCR was comparable in DKO and WT VSMCs, the need for a stressor to dysregulate mitochondrial function in DKO VSMCs could explain the reduced maximum OCR after FCCP treatment. However, it has been suggested that an aberrant maximal OCR only corresponds to the initial symptoms of mitochondrial dysfunction caused by SLC2A10 mutations, due to reactive oxygen species (ROS) that trigger arterial wall remodeling. In its turn, this leads to additional ROS production, and disease progression promotes further decline in mitochondrial function. It is also plausible that mitochondrial dysfunction is a more ubiquitous phenomenon associated with aneurysmal disease. Van der Pluijm et al. investigated mitochondrial function in fibulin-4^R/R and Tgfbr-1^M318R/+ mouse VSMCs and found decreased mitochondrial respiration in these mouse models. In addition, skin fibroblasts of aneurysmal patients with FBN1, TGFBR2 or SMAD3 mutations revealed mitochondrial dysfunction as well (25).

Finally, due to the mild nature and low abundance of relevant phenotypic alterations, it could be argued that this novel ATS DKO mouse model does not appear to be a suitable disease model to study the pathomechanisms underlying ATS (22,23). While the lack of a severe phenotype could be attributed to the genetic background of the used mouse strains, possibly hindering phenotypic penetrance (39), the underlying cause could also be biological robustness. This is a phenomenon by which the functionality of the mutated gene is maintained by increased expression of related genes (40,41), as was demonstrated earlier in a number of other mouse disease models (42,43). One mechanism of biological robustness is transcriptional adaptation (44,45). This mechanism is triggered by mRNA degradation products and is believed to act via sequence similarity (46,47). Since Slc2a10 is a member of a large family of facilitative glucose transporter proteins harboring a high sequence similarity, this mechanism is plausible (48). It could also be argued that an extensive dose (60 mg/kg body weight) of ascorbic may have masked the development of a relevant phenotype, since this dose exceeds the daily recommended dose in human (1 mg/kg body weight) extensively. However, species differences might complicate the direct comparison or extrapolation of the optimal daily dose between mouse and human. A daily AA dose of 1 mg/kg body weight corresponds in human to plasma AA levels of 25 μm. To reach comparable levels in mouse plasma, a significantly higher dose (80 mg/kg body weight) needs to be administered (49).

Nevertheless, the current DKO mouse model could be of importance in revealing redundant gene networks, alternative pathways or modifier genes, possibly providing clues for future therapies. Moreover, since isolated VSMCs of DKO mice mimicked cellular disturbances found in cell cultures of ATS patients, these VSMCs could potentially be used for further identification of the molecular mechanisms that underlie ATS, while analyses at the histological level may reveal adaptive mechanisms that contribute to the phenotypic rescue.

In conclusion, we developed a novel mouse model for ATS, DKO for Gulo and Slc2a10. While our model does not phenocopy human ATS, it did reveal alterations at the cellular level including disturbed ECM assembly and altered cellular respiration. Our model confirms a role for ascorbate synthesis and compartmentalization in the disease mechanisms leading to ATS and may be helpful to identify the mechanisms underlying the phenotypic rescue in mice.

Materials and Methods

Animals

Heterozygous knock-out mice for Slc2a10, in which a selection cassette replaces a sequence ranging from exon 2 to the beginning of exon 5, were purchased from Taconic Biosciences (Germantown, New York, USA, cat. nr. TF2438) (Supplementary Material, Fig. S1A). The mutation resides in a mixed 129/SvEv-C57BL/6 genetic background. Heterozygous knock-out mice for Gulo, in which a selection cassette replaces exons 3 and 4, were purchased from MMRRC (RRID MMRRC_000015-UCD) (Supplementary Material, Fig. S1A). This mouse line, residing in a mixed 129/SvEv-C57BL/6 genetic background as well, was previously described by Maeda et al. (49). The two lines were crossed according to a breeding scheme depicted in Supplementary Material, Fig. S1B. To allow for valid data generation and to take into account any potential influence of genetic background, all comparisons were made between littermates from the F2 generation. Male and female mice of the following genotypes were studied: double knock-out (DKO—Gulo^{tm1mae/tm1mae};Slc2a10^−/−), single knock-out (Gulo KO—Gulo^{tm1mae/tm1mae};Slc2a10^+/+ and Slc2a10 KO—Gulo^+/+;Slc2a10^−/−) and wild type (WT—Gulo^+/+;Slc2a10^+/+). All procedures were conducted in compliance with the European Parliament Directive 2010/63/EU and with the approval of the Ghent University ethical committee on animal experiments (Permit Number: ECD 13/17). All animals were fed ad libitum with a standard mouse breeding feed, supplemented with 300 mg/kg AA (ssniff® Spezialdiäten). The administered AA dose is based on an average food intake of 4 g/day, for a mouse with an average weight of 20 g. This dose translates to a daily dose of 60 mg/kg body weight, which is comparable to the average administered AA dose of the first report on this model (110 mg/kg), by Maeda et al. (49). Mice were maintained in a fully controlled animal facility (12:12 h light/dark cycle at ±22°C).

Genotyping

Seven-day-old mice were toe-clipped, after which genomic DNA was prepared from the tissue using the KAPA Express Extract DNA Extraction Kit (Sigma-Aldrich, St. Louis, Missouri, USA, cat. nr. KK7103). For both Gulo and Slc2a10, the DNA was PCR-amplified using specific primers detecting the WT and mutant alleles. The genotype was determined based on PCR product presence and band size (Supplementary Material, Table S1).

Quantitative real-time PCR

Skin and aortic tissues from three male and three female mice of each selected genotype were harvested at the age of 9 months and submerged in RNAlater™ Stabilization Solution (Thermo Fisher Scientific, Waltham, Massachusetts, USA, cat. nr. AM7020) prior to the RNA extraction procedure, carried out with the RNeasy® Mini kit (Qiagen, Hilden, Germany, cat. nr. 74106). Next, cDNA synthesis was performed using the iScript cDNA synthesis kit (Bio-Rad, Hercules, Californa, USA, cat. nr. 1708891). RT-qPCR reactions were carried out in quadruple on an LC480 machine (Roche, Basel, Switzerland). The reaction mix consisted of cDNA, LightCycler® 1536 DNA Probes Master (Roche, Basel, Switzerland, cat. nr. 05502381001), LightCycler® 480 High Resolution Melting Dye (Roche, Basel, Switzerland, cat. nr. 04909640001) and specific primers amplifying the Slc2a10 transcript or expressed repeat elements (Supplementary Material, Table S2) (50), to which all reactions were normalized. Data analysis was carried out with qBase+ (Biogazelle). Values plotted correspond to the means of three male and three female biological replicates. Error bars shown represent 95% confidence intervals. Data were analyzed using Welch’s ANOVA, followed by a Games-Howell post hoc test.

In vivo imaging

Serial ultrasound imaging was performed using a dedicated ultrasound apparatus (Vevo 2100, FUJIFILM VisualSonics, Toronto, Canada), equipped with a high-frequency linear array transducer (MS 550D, frequency 22–55 MHz). Imaging was carried out on mice of 6 weeks and 3, 6 and 9 months old. Ten male and 10 female mice of the selected genotypes were studied. The mice were anesthetized with 1.5% isoflurane mixed with 0.5 l/min 100% O₂. Body temperature was maintained at 37°C, by placing the mice on a heating pad. Artery and cardiac measurements were carried out in concordance with previously established guidelines (51). In short, the diameter measurements of the aorta at the level of the aortic root, proximal ascending aorta, distal ascending aorta, aortic arch and descending aorta were performed, next to diameter measurements of the carotid arteries and cardiac assessment. Three cardiac cycles for each animal were analyzed (VevoLAB 1.7.0, FUJIFILM VisualSonics, Toronto, Canada). For each measurement location, a linear mixed model with covariance pattern modeling was fitted to account for correlated responses (diameters) within the same subject. Genotype, age and sex were considered as categorical explanatory variables. Starting from a saturated mean model with unstructured covariance matrix and using residual maximum likelihood (REML), the covariance model was simplified by comparison with simpler structures through Akaike’s information criterion. The following covariance matrices that were as parsimonious as possible were selected: unstructured (aortic root), autoregressive (proximal ascending aorta), Toeplitz (distal ascending aorta) and compound symmetry (all other measurement locations). Afterwards, the fixed part of the model was simplified by testing simpler models using maximum likelihood. All final models included the main effects of genotype, age and sex without interaction terms and were refitted with REML. Values plotted correspond to the means of ±10 biological replicates. Error bars shown represent 95% confidence intervals.

Vascular corrosion casting

The vascular corrosion casting technique, based on the injection of a polymer to capture the 3D structure of the vasculature, was performed on three male and three female mice of each selected genotype at the age of 9 months as previously described (23). Briefly, 2–3 ml Batson’s solution (Polysciences, Batson’s #17 corrosion kit) was retrogradely injected in the abdominal aorta through a 26G catheter. After completion of the polymerization reaction, mouse bodies were macerated overnight in a 25% KOH solution. The resulting casts were cleaned, evaluated and photographed using a dissecting microscope, equipped with a 5 megapixel camera (Leica, Wetzlar, Germany).

Histology

From the three male and three female mice of each selected genotype, skin and aorta (9-month-old animals) were collected for histological analysis. Samples were formalin-fixed and paraffin-embedded, after which 5-μm-thick paraffin sections were made. Sections were subjected to the standard Verhoeff-Van Gieson and PR histological staining procedures, prior to visualization on a Zeiss Axio Observer Z1 microscope.

Transmission electron microscopy

Sample fixation was carried out in a 4% formaldehyde (EM grade), 2.5% glutaraldehyde (EM grade) and 0.1 M cacodylate buffer solution. Samples were placed in a vacuum oven for 30 min and left rotating for 3 h at room temperature. This solution was later replaced with fresh fixative, and samples were left rotating over night at 4°C. After washing, samples were post-fixed in 1% OsO₄ with K₃Fe(CN)6 in 0.1 M Na-cacodylate buffer, pH 7.2. After washing in double-distilled H₂O, samples were subsequently dehydrated through a graded ethanol series, including bulk staining with 2% uranyl acetate at the 50% ethanol step, followed by embedding in Spurr’s resin. To select the area of interest on the block and in order to have an overview of the phenotype, semi-thin sections were first cut at 0.5 μm and stained with toluidine blue. Ultrathin sections of a gold interference color were cut using an ultramicrotome (Leica EM UC6), followed by post-staining in a Leica EM AC20 for 40 min in uranyl acetate at 20°C and for 10 min in lead stain at 20°C. Sections were collected on formvar-coated copper slot grids. Grids were viewed with a JEM 1400plus transmission electron microscope (JEOL, Tokyo, Japan) operating at 80 kV. Collagen diameter measurements were carried out with Fiji (52). Error bars shown represent 95% confidence intervals. Data were analyzed with a two-tailed t-test.

VSMC isolation and cell culture

Mice (at an age of 25 days) were euthanized (CO₂) and autopsied according to standard protocols. Primary VSMCs from the thoracic aorta were isolated according to the collagenase digestion method of Proudfoot and Shanahan (53). Each cell line was derived from a single aorta. Primary VSMCs were cultured on gelatinized dishes in SmBM medium supplemented with the SmGM-2 kit (Lonza, Basel, Switzerland, cat. nr. CC-3182). Unless otherwise specified, two cell lines per genotype were assessed.

Immunofluorescence of ECM

ECM protein production by VSMCs was determined by immunofluorescence. VSMCs were seeded at 50 000 cells/well in eight-well removable chamber slides and grown for 7 days to allow ECM deposition. VSMCs were fixed with an ice-cold 70:30 methanol/acetone mixture for 5 min and washed with PBS. Coverslips were blocked for 1 h with PBS supplemented with 10% normal goat serum (Agilent, Santa Clara, California, USA, cat. nr. X0907). Primary antibodies (Supplementary Material, Table S3) were incubated overnight at 4°C in PBS. Coverslips were washed three times with PBS for 5 min each prior to incubation with the secondary antibody for 1.5 h at room temperature (Molecular Probes, anti-rabbit Alexa Fluor 594, 1:1000). Coverslips were washed and mounted to glass slides with Vectashield supplemented with DAPI (Vector Laboratories, Peterborough, UK, cat. nr. H-1200) and sealed with nail polish. Images were recorded on a wide-field epifluorescent microscope (Axio Imager D2, Zeiss). Quantification of the immunofluorescent signal was performed by calculating the corrected total cell fluorescence (CTCF) of the ECM components corrected for the number of nuclei. The CTCF was determined by setting a color threshold to select the fibers in the image with Fiji image-analyzing software (52) and determining the integrated density of this area (intensity of the fluorescence). This measurement was corrected for the background fluorescence and the total area of the fibers and results in the CTCF. The CTCF was then divided by the number of nuclei that were present in the measured image. Data were corrected for outliers with the Grubbs’ test for outliers (54). Statistical analysis was performed with a nonparametric Mann–Whitney test. Significance was tested two-tailed. Results are expressed as mean ± SD.

TGFβ stimulation

VSMCs were seeded in six-well plates to reach confluence and were allowed to attach for 24 h. The following day, medium was changed to SmGM-2 medium without fetal calf serum, and VSMCs were serum deprived for 24 h prior to TGFβ stimulation. Protein samples were collected after 0 min, 15 min, 30 min, 1 h and 4 h of stimulation with TGFβ1 (BioVision, Milpitas, California, USA, cat. nr. 4342-5). Cells were scraped in PBS supplemented with protease inhibitor cocktail (Roche, Basel, Switzerland, cat. nr. 11836145001, 1:100) and phosphatase inhibitor cocktail (Sigma-Aldrich, St. Louis, Missouri, USA, cat. nr. P0044, 1:100) and lysed in an equal volume of 2× Laemmli buffer (4% SDS, 20% glycerol, 120 mm Tris pH 6.8) supplemented with protease inhibitor cocktail and phosphatase inhibitor. Lysates were first cleared from large DNA by passing through a 25G needle and then heated to 65°C for 10 min. Protein concentrations were measured with the Lowry protein assay as previously described (55). Equal amounts of protein were separated for size by SDS-PAGE and then transferred to a PVDF membrane (1 h, 100 V, Immobilon) and blocked with either 3% milk in PBS supplemented with 0.1% Tween-20 (1 h, room temperature). The primary antibody was incubated for 45 min at room temperature or overnight at 4°C for phosphorylated Smad2 (see Supplementary Material, Table S3 for primary antibodies). The membranes were washed five times with 0.1% Tween-20 in PBS and then incubated with horseradish peroxidase-conjugated secondary antibodies (Jackson ImmunoResearch, Ely, UK, 1:2000) for 1 h at room temperature. Bound secondary antibodies were detected with an Amersham Imager 600 (GE Healthcare Life Sciences, Chicago, Illinois, USA) using chemiluminescence. Band intensity was quantified using Fiji image-analyzing software (52). Data were corrected for outliers with the Grubbs’ test for outliers (54). Statistical analysis was performed with a nonparametric Mann–Whitney test. Significance was tested two-tailed. Results are expressed as mean ± SD.

Mitochondrial respiration

OCR was measured using an XF-24 Extracellular Flux Analyzer (Seahorse, Agilent, Santa Clara, California, USA). Respiration was measured in XF assay media (non-buffered DMEM), in basal conditions and in response to 1 μm oligomycin (ATP synthase inhibitor), 1 μm FCCP(uncoupler) and 1 μm antimycin A (complex III inhibitor). Smooth muscle cells were seeded at a density of 30 000 cells/well and analyzed after 24 h. Optimal cell densities were determined experimentally to ensure a proportional response to FCCP with cell number. For these experiments 6–8 wells were measured per time point (56,57). Data were corrected for outliers with the Grubbs’ test for outliers (54). Statistical analysis was performed with a nonparametric Mann–Whitney test. Significance was tested two-tailed. Results are expressed as mean ± SD.

Acknowledgements

B.C. is a senior clinical investigator of the Research Foundation – Flanders (FWO). M.R. was supported by the Research Foundation – Flanders (FWO) as a postdoctoral fellow. Ghent University Hospital is a member of the European Reference Networks for vascular and skin disorders (VASCERN and ERN-Skin).

Conflict of Interest statement. None declared.

Funding

Ghent University (Methusalem BOFMET2015000401 to A.D.P.); Research Foundation – Flanders (FWO) (FWOOPR2013025301); ‘Lijf and Leven’ grant [2014, ‘GAMMA’ (Genexpressie analyse ter detectie van de moleculaire mechanismen van aneurysmavorming)].