Clinical, pathological and functional characterization of riboflavin-responsive neuropathy

Riboflavin metabolites are critical components of the mitochondrial electron transport chain. Manole et al. describe the genetics, phenotypes and neuropathology of a large patient series with riboflavin-responsive neuropathy. Using in-vitro and in-vivo models, they reveal mitochondrial dysfunction in the disorder, and validate riboflavin esters as a potential therapeutic strategy.


Introduction
Brown-Vialetto-Van Laere (BVVL) syndrome is an autosomal recessive neurological disorder first described by Brown in 1894 and later by Vialetto and Van Laere (Brown, 1894;Vialetto, 1936;Van Laere, 1966). Affected patients mostly present with neuropathy, bilateral sensorineural deafness, bulbar palsy and respiratory compromise. Other cranial nerve palsies, optic atrophy, upper and lower motor neuron involvement and ataxia can occur particularly as disease progresses, mimicking conditions such as amyotrophic lateral sclerosis (ALS), Madras motor neuron disease and Nathalie syndrome (Anand et al., 2012;Manole et al., 2014). Deafness is the most common sign of this condition, with most affected individuals exhibiting hearing loss during the disease course. The time between the onset of deafness and the development of other manifestations varies but is usually in early childhood (Manole and Houlden, 2015).
Previous work has revealed strong links between mutations in two genes (SLC52A2 and SLC52A3) and BVVL syndrome, both of which encode riboflavin transporters (Green et al., 2010;Johnson et al., 2012;Foley et al., 2014). The role of another riboflavin transporter-encoding gene, SLC52A1, in BVVL syndrome pathogenicity is still uncertain, as it was found to be defective in only one case (Ho et al., 2011). SLC52A2 and SLC52A3 mutations include missense, nonsense, frame-shift, and splice-site alterations, but uniformly result in loss-of-function through reduced riboflavin transporter expression and/or riboflavin uptake (Foley et al., 2014;Udhayabanu et al., 2016). Riboflavin (7,8-dimethyl-10-ribityl-isoalloxazine; vitamin B 2 ) is a water-soluble compound and acts as a precursor for flavin mononucleotide (FMN) and flavin adenine dinucleotide (FAD). Both FMN and FAD function in biological redox reactions such as in the mitochondrial electron transport chain (ETC) (Powers, 2003). Since riboflavin cannot be synthesized by mammals de novo, riboflavin transporters are indispensable for normal cellular metabolism, suggesting that reduced intracellular riboflavin is a critical pathological mediator of BVVL syndrome. Indeed, although infants with early-onset riboflavin transporter deficiency rapidly become ventilator-dependent and usually die in the first year of life, treatment with high-dose riboflavin supplementation partially ameliorates the progression of this neurodegenerative condition, particularly if initiated soon after the onset of symptoms (Foley et al., 2014). In a recent paper by Rizzo et al. (2017), motor neurons from induced pluripotent stem cells (iPSCs) were derived from two patients with BVVL syndrome carrying either SLC52A2 or SLC52A3 mutations. They found a reduction in axon elongation coupled with perturbations in neurofilament composition and reduced autophagic/mitophagic flux. However, despite these findings the pathophysiological mechanisms underlying motor neuron degeneration in BVVL syndrome are still unclear.
Motor neurons are thought to be uniquely susceptible to impaired energy metabolism because of their high metabolic rate and axonal length (Cozzolino and Carri, 2012). Furthermore, mitochondrial perturbations causing alteration of the ETC and increased oxidative stress are thought to be involved in the pathomechanisms of neurodegeneration in some motor neuron diseases such as ALS or spinal muscular atrophy (Johnson et al., 2010;Cozzolino and Carri, 2012; Bartolome et al., 2013). Given the critical role of riboflavin in the generation of substrates used for the ETC, we hypothesized that reduced riboflavin transport results in impaired ETC, which may in turn contribute to neurodegeneration.
Here, we expand the clinic-genetic spectrum of riboflavin transporter genes and then perform a series of experiments to identify the underlying effects of loss of riboflavin transporter function on neuronal integrity and mitochondrial function. We review clinical case histories and undertake pathological evaluation of brain and spinal cord of two patients with confirmed SLC52A3 mutations who presented either in infancy or in later childhood. Finally, we investigate the in vitro cellular effects of SLC52A2 mutations on metabolism and ETC function and also the in vivo consequences of loss of the SLC52A2/SLC52A3 homologue in Drosophila melanogaster, and test whether these can be mitigated by supplementation with a riboflavin derivative.

Study subjects
Patients were enrolled and informed consent from the patient and/or their parental guardian was obtained. DNA was collected from a total of 132 suspected cases (probands and their relatives) presenting with cranial neuropathies and sensorimotor neuropathy with or without respiratory insufficiency. Patients' DNA samples were collected at medical centres in England (including from patients originating from Pakistan, India, Saudi Arabia, Kuwait, Iran and Turkey) as well as from medical centres in Wales, Scotland, Northern Ireland, Ireland, France, Belgium, the Netherlands, Greece, Malta, Russia, Lebanon, Iceland, Australia and the USA, following the announcement of an on-going molecular study at the UCL Institute of Neurology, University College London (National Hospital for Neurology and Neurosurgery, Queen Square, London) of patients presenting with this phenotype. This study was ethically approved by the UCL/University College London Hospital Joint Research Office (99/N103), and written informed consent to perform a skin biopsy and fibroblasts was obtained as appropriate.

PCR and Sanger sequencing
Primer sequences, PCR and Sanger sequencing conditions for SLC52A1, SLC52A2 and SLC52A3 were as in Foley et al. (2014). Segregation of pathogenic variants was also assessed. Where one heterozygous mutation was identified, deletions in the other allele were investigated by array CGH (comparative genomic hybridization) but no deletions or insertions were identified. Mutation positions are based on NCBI reference sequences for complementary DNA. SLC52A2 mutation positions are based on sequences NM_024531.4 for the nucleotide sequence and NP_078807.1 for the protein sequence. SLC52A3 mutation positions are based on sequences NM_033409.3 for the nucleotide sequence, and NP_212134.3 for the protein sequence.

SLC52A2 patient fibroblasts
For further details of the generation of human fibroblast cultures and cell culture, determination of flavin status in fibroblasts, and assessment of ETC complex I, II, and citrate synthase activities in fibroblasts see the Supplementary material.

Drosophila studies
For details of semi-quantitative RT-PCR of whole flies, measurement of mitochondrial membrane potential, isolation of Drosophila mitochondria and measurement of mitochondrial respiration, and immunohistochemistry of the Drosophila larval neuromuscular junction, see Supplementary material.

Stocks and culture conditions
Fly strains were obtained from the Bloomington Drosophila Stock Center (Indiana, USA) and Vienna Drosophila Resource Center (Austria). All transgenic insertions used in this study were outcrossed at least five times into an isogenic (iso31) background. These were: cg11576 UAS-RNAi (VDRC 7578, Dietzl et al., 2007;and HMC04813, Perkins et al., 2015), and daughterless-GAL4 (Bloomington stock 55850). Flies were reared at 25 C on standard fly food consisting of corn meal, yeast, sucrose, glucose, wheat germ, soya flour, nipagin and propionic acid. For experiments with compound supplementation, 0.1 mg/ml riboflavin (Sigma) or 0.1 mg/ml riboflavin-5'-lauric acid monoester (RLAM) was added to the food. Flies were kept under 12-h light: 12-h dark cycles. , ETC complex I, II, II/III  redox activities and citrate synthase assay Riboflavin, FMN and FAD content, and ETC complex I, II redox activity and citrate synthase activities were measured in flies ($10/genotype) as described above for patient fibroblasts. Prior to analysis all samples were subjected to three cycles of freeze/thawing to lyse membranes. A Lowry assay was used to normalize for protein concentration (Frolund et al., 1995). Complex II/III activity was determined at 30 C using the method of King (1967), which followed the succinate-dependent antimycin A sensitive reduction of cytochrome c at 550 nm.

Electron microscopy of fly brains
Flies were decapitated and the heads were perfused with 3% glutaraldehyde in 0.1 M sodium cacodylate buffer and 5 mM CaCl 2 overnight. Brains were then treated with 1% osmium tetroxide for 3 h at 4 C and embedded in Araldite Õ CY212 resin. Thin sections were stained with toluidine blue and the brain visualized with the light microscope. Ultrathin sections (70 mm) were stained with lead citrate and uranyl acetate and digital images taken on a Philips CM10 TEM with MegaView G3 digital imaging system with Radius Software (Olympus). Mitochondrial morphology was measured blind to genotype.

Larval behaviour
Larval locomotion was tested by placing individual third instar larvae in the centre of petri dishes (8.5 cm diameter, 1.4 cm height) coated with 10 ml of 4% agar. On average, 30 larvae were tested per strain. The number of grid squares (1 cm) entered per min by the larva was measured per each genotype.

Adult behaviour
Adult flies were kept as groups of males and females in a 12 h:12 h light-dark cycle at 25 C for 1 day prior to testing. Single virgin females ($1 day old) were loaded into glass tubes (with 2% agar and 4% sucrose food) and monitored using the Drosophila Activity Monitoring System (Trikinetics) in 12 h:12 h light-dark conditions at 25 C with an approximate intensity of 700-1000 lx during the light condition. For experiments involving supplementation, riboflavin or RLAM was added to the above food. Fly activities were deduced from the number of times flies broke beams of infrared light passing through the middle of the tube. Locomotor activity was recorded in 30-min bins and an analysis was performed on the second day after loading. Data were pooled from at least two independent experiments. The relative locomotor activities per 30-min bin for individual flies were averaged for each genotype and also the average locomotor activity per day was calculated. Locomotion graphs were generated using GraphPad Prism 6 and Microsoft Excel.

Life span
Adult female flies were collected from eclosion and transferred to fresh food tubes (10 flies/tube) with or without RLAM supplementation. Each day, death events were scored and viable flies were transferred to fresh tubes. Survival proportions were plotted as percentage of live flies against days. Approximately 100 flies were tested for each genotype.

Statistics
Statistics were performed using GraphPad Prism 6. The significance between the variables was shown based on the P-value obtained (ns indicates P 4 0.05, *P 5 0.05, **P 5 0.005, ***P 5 0.0005). Data are presented as box plots illustrating 80% of the data distribution, together with the median and 10th, 25th, 75th and 90th percentiles unless otherwise stated.

Clinical and genetic analysis of patients with BVVL syndrome
We screened 132 patients with phenotypes suggestive of BVVL syndrome and identified 20 patients (15%) with riboflavin transporter mutations. Genetic and clinical details of the 20 individuals carrying mutations in SLC52A2 or SLC52A3 are summarized in Table 1. Thirteen of the probands were males. Age of symptom onset was available for 15 patients (mean: 8.2 years; range: 7 months-21 years). First symptoms usually occurred during childhood (11 patients) and less frequently during teenage years (four patients) or adulthood (one patient), and were mostly secondary to cranial nerve involvement (Table 1). Only one patient presented with symptoms not related to cranial nerve involvement (limb weakness). Sensorineural hearing loss was both a common presenting symptom (eight patients) and a common clinical feature during follow-up (17 patients). Other frequent clinical features include optic atrophy (14 patients), weakness of facial and bulbar muscles (13 patients) and sensorimotor peripheral axonal neuropathy (16 patients). Limb weakness was more severe in the upper than in the lower limbs in nine patients. Ten patients developed some degree of respiratory involvement with three requiring assisted ventilation, and 10 patients had dysphagia and/or chewing difficulty, six of them requiring nasogastric tube or gastrostomy feeding.
Of the 22 mutations identified, eight were found in the SLC52A2 locus and 14 in SLC52A3, of which five SLC52A2 and nine SLC52A3 variants were novel (Table  1). In contrast, no SLC52A1 mutations were observed. None of the variants were found in the homozygous state in the ExAC database of over 100 000 controls, suggesting pathogenicity (Table 1). Consistent with this hypothesis, all except one mutation (p.Arg212Cys in SLC52A3) were predicted to be at least probably damaging by SIFT and PolyPhen-2 algorithms (scores ranging between 0.55 and 1). The mutations reside in transmembrane helices and in the intracellular and extracellular loops. Although three homozygous and seven compound heterozygous mutations were identified, five mutations (all in SLC52A3) were identified on only one allele. These heterozygous individuals did not differ substantially in phenotype including age of presentation from the rest of the cohort of mutation-positive cases.
There was no correlation between the nature of pathogenic variant and phenotype severity, although in the case of Patient AM2, nonsense mutations on both alleles resulted in a truncated protein, and this genotype correlated with rapid progression of symptoms and death at 2 years of age.

Neuropathological analysis of patients with BVVL syndrome
To characterize in detail the neuropathological symptoms of BVVL syndrome, we undertook a comprehensive pathological examination of two patients carrying compound heterozygous SLC52A3 mutations (Patients AM2 and AM4; Table 1). These patients represent two ends of the spectrum of severity of BVVL syndrome.
Patient AM2 had a normal birth at term. His motor skills were mildly delayed and he never acquired the ability to roll over completely front to back. He achieved the ability to sit with minimal support at age 7 months. From about 8 months of age he began to exhibit more clear signs of the condition such as ptosis and neck weakness. He was admitted at the age of 9 months for investigations but no diagnosis was made at that time. His condition quickly progressed to include respiratory muscle weakness, and ventilator dependence at the age of 1 year. He also developed severe weakness in his shoulder girdle areas and proximal upper limbs. Weakness subsequently developed in forearm muscles, distal lower limb muscles and thighs, trunk and face, and progressed to the point that he could only weakly move his eyelids and had very limited sideways movements of his eyes. Nerve conduction studies and EMG showed a severe neuropathy. He died at 2 years of age of respiratory failure.
The cerebral cortex, hippocampus and cerebral white matter were unremarkable. The deep grey nuclei were not available for assessment. The neuronal density in the substantia nigra was normal for the patient's age, but the third and fourth cranial nerve nuclei showed severe neuronal loss, gliosis and microglial activation ( Supplementary  Fig. 1). In the pons there were two symmetrical sharply demarcated lesions surrounding both fifth cranial nerves ( Fig. 1A-D). In these lesions we observed prominent neovascularization, dense infiltration of macrophages and widespread myelin loss, with relative preservation of axons. The fifth cranial nerve was vacuolated and its nucleus showed severe neuronal loss and gliosis ( Supplementary Fig. 1). In the medial lemniscus at the level of the midbrain and pons there was prominent gliosis, microglial activation and vacuolation of the neuropil, which was also seen in the central tegmental tract. In the medulla, the ninth, 10th and 12th cranial nerve nuclei showed severe neuronal depletion and gliosis with pale corresponding nerve tracts. The neuronal loss in the eighth cranial nerve was moderate ( Supplementary Fig. 1). The medial lemniscus, spinocerebellar tract and medullary reticular formation were all gliotic. Inferior olivary nuclei, in particular the dorsal and ventral parts, showed severe neuronal loss and gliosis. In the cerebellum, there was no significant cortical atrophy [ Supplementary Fig. 2B(i)], but the cerebellar white matter showed widespread vacuolation.
In the middle cerebellar peduncle there was a lesion with similar morphology to those seen in the pons and medulla.
In the spinal cord, there was moderate neuronal loss in the anterior horns and moderately severe atrophy of spinothalamic tracts, spinocerebellar tracts, gracile fasciculus and nucleus in the medulla, and to a lesser extent cuneate fasciculus and nucleus in the medulla ( Supplementary Fig.  1). There was severe fibre loss and macrophage activation in the anterior spinal nerve roots, whilst the poster spinal nerve roots were densely populated by myelinated fibres with little macrophage activation (Fig. 1E-H). Patient AM4 also had a normal birth at term. He had hearing problems from the age of 8 years and was diagnosed with sensorineural hearing loss at the age of 11. He developed optic atrophy and difficulty walking at the age of 16 years. At age 17 years he presented with dysarthria and subsequently developed swallowing difficulties. EMG showed widespread denervation and nerve conduction studies were consistent with an axonal motor neuropathy. He died at 19 years of respiratory insufficiency.
The cerebral neocortex, hippocampus, amygdala, caudate nucleus putamen, globus pallidus, thalamus and cerebral deep white matter showed no apparent abnormality. Mild gliosis was evident in the dorsal part of the optic tract. In the midbrain, the substantia nigra was densely populated by lightly pigmented neurons in keeping with the patient's age. The corticospinal tracts were unremarkable in the cerebral peduncles of the midbrain. The third cranial nerve nucleus was not available for the assessment and the fourth cranial nerve nuclei showed a mild degree of neuronal loss and accompanying mild gliosis ( Supplementary  Fig. 3). In the pons there was prominent neuronal loss in the loci coerulei with free pigment deposits in the neuropil and gliosis (Supplementary Fig. 3). The neuronal density in the pontine basal nuclei and the fibre density in the transverse and corticospinal tracts were unremarkable. The superior cerebellar peduncles showed mild gliosis and vacuolation. The fifth and seventh cranial nerve nuclei showed severe neuronal loss ( Supplementary Fig. 3 and Fig. 2A and C). In the sixth cranial nerve nucleus the gliosis was mild and the neuronal density was relatively well-preserved ( Fig. 2 and Supplementary Fig. 3). In the medulla (Fig. 2D-F and Supplementary Fig. 3), the nuclei of the eighth cranial nerve showed severe neuronal loss and prominent gliosis and the nuclei of the 10th and 12th cranial nerves and nucleus ambiguus showed moderately severe neuronal loss and gliosis. The nerve tracts of the available fifth, 10th and 12th cranial nerves ( Fig. 2A, E and F, insets) showed reduced density of the myelinated fibres. The gracile nucleus showed severe, while the cuneate nucleus showed mild neuronal loss and gliosis ( Supplementary  Fig. 3). In the cerebellar cortex there was widespread Purkinje cell loss and Bergmann gliosis and some degree of neuronal loss in the granular cell layer [ Fig. 2G and Supplementary Fig. 2A(i)].
The cerebellar nuclei (fastigii, globosus and emboliformis) showed severe neuronal loss and gliosis bilaterally, while the dentate nuclei were mildly gliotic with less conspicuous neuronal loss (Fig. 2H-I). There was a severe pallor, gliosis and prominent microglial/macrophage activation in the medial lemniscus (Fig. 2J), spinothalamic tracts and inferior cerebellar peduncles (Fig. 2K). The corticospinal tracts in the pyramids (Fig. 2L) were densely populated by myelinated fibres and the inferior olivary nuclei showed only mild patchy gliosis ( Supplementary Fig. 3). In the ventral aspect of the upper spinal cord at the junction with the lower medulla there were bilateral symmetric lesions involving the anterior horns, lateral reticular nucleus, supraspinal nucleus, spinothalamic tracts and medial longitudinal fasciculus (Fig.  3A-E). In these lesions there was neovascularization and dense infiltration of macrophages with a near complete loss of myelin, while neurons and axons were relatively preserved ( Fig. 3A-E). Below these lesions the cervical cord showed a severe atrophy of the neurons in the anterior horns and to a lesser extent posterior horns ( Supplementary Fig. 3). There was a severe, slightly asymmetrical atrophy of the uncrossed anterior corticospinal tracts, whilst the lateral corticospinal tracts were densely populated by myelinated fibres with only mild vacuolation [ Fig. 2L(i) and Supplementary Fig. 3]. Severe symmetrical atrophy with prominent pallor, macrophage infiltration and vacuolation of the anterior and posterior spinocerebellar tracts [ Fig. 2K(i)], spinothalamic tracts and gracile and to a lesser extent cuneate fasciculi ( Supplementary Fig. 3) was also observed. While the posterior nerve roots were densely populated by myelinated fibres with only mild macrophage activation, in the anterior nerve roots there was moderately severe loss of myelinated fibres and prominent infiltration of macrophages ( Fig. 3F-J).
Amyloid-b, hyper phosphorylated tau, -synuclein, TDP-43, p62 or ubiquitin positive inclusions were not observed in Patients AM2 and AM4. In both cases the inflammatory reaction was restricted to microglial activation and macrophage infiltrates with no significant lymphocytic inflammation.

Generation of a novel Drosophila model of BVVL syndrome
Mitochondrial dysfunction has long been documented in neurodegenerative diseases (Palomo and Manfredi, 2015). For example, SOD1 mutations in familial ALS have been shown to lead to abnormalities in mitochondrial morphology, both in biopsies and post-mortem tissues of human patients (Sasaki and Iwata, 2007;Sasaki, 2010) and in cellular and mouse models of the disease (Magrane et al., 2009(Magrane et al., , 2014Vinsant et al., 2013;Palomo and Manfredi, 2015). However, whether mitochondrial function is perturbed by BVVL syndrome-linked mutations is unclear. We hypothesized that lower levels of intracellular riboflavin as a result of mutated riboflavin transporters would lead to reduced levels of FMN and FAD, which in turn would lead to impairments at the level of the ETC complex I and complex II. We sought to test this hypothesis in vivo.
Previous work has used a knockout of the mouse SLC52A3 orthologue to model BVVL syndrome (Intoh et al., 2016;Yoshimatsu et al., 2016). However, SLC52A3 knockout mice die within 48 h of birth, precluding phenotypic analysis at later developmental stages. To circumvent this issue we turned to the fruit fly, Drosophila. Comparative BLAST analysis revealed a single Drosophila homologue of SLC52A3/hRFVT3, the previously uncharacterized gene cg11576 (E-value: 2.74 Â 10 À74 ; next closest homologue, E-value: 2.95). As described below, we name the cg11576 gene drift (Drosophila riboflavin transporter). The DRIFT amino acid sequence exhibits 36.9% identity and 53.1% similarity with hRFVT3, and the L1 loop and GXXXG motifs characteristic of riboflavin transporters are also conserved ( Supplementary Fig. 4) (Russ and Engelman, 2000;Zhang et al., 2010). The L1 loop is a region of the protein shown to recognize riboflavin through both hydrogen bonds and van der Waals interactions (Zhang et al., 2010), while the GXXXG motif is required for dimerization (Russ and Engelman, 2000). DRIFT is also highly homologous to the hRFVT3 paralogues hRFVT1 and hRFVT2 ( Supplementary Fig. 4). Furthermore, 10/19 and 11/25 residues in hRFVT2 and hRFVT3, respectively, that are altered by BVVL syndrome-linked mutations are either identical or functionally similar in DRIFT (Supplementary Fig. 4 and 5). Mammalian riboflavin transporters exhibit wide domains of expression, including the nervous system, intestine, testes, and placenta (Yonezawa and Inui, 2013). Similarly, drift is transcribed in several adult Drosophila tissues, including the head, gut, abdomen and thorax The posterior nerve roots are densely populated with myelinated fibres with (J 1 ) minimal macrophage activity. (F 2 -I 2 ) In the anterior nerve roots there is a moderately severe loss of myelinated fibres and (J 2 ) prominent macrophage activation. Scale bars = 2.5 mm in A-E and F-J; 80 mm in A 1 -E 1, F 1 -J 1 and F 2 -J 2 .
( Supplementary Fig. 6A). This broad expression pattern is in agreement with RNAseq data from the Drosophila ModEncode Project (Graveley et al., 2011).
To examine the biochemical and phenotypic consequences of loss of DRIFT, we disrupted drift expression using either P-element insertions or transgenic RNA interference (RNAi). Insertion of a MiMIC transposon (cg11576 MI04904-GFSTF.2 ) within intron 1 of drift resulted in viable heterozygote adults but homozygote lethality prior to the L3 larval stage. Furthermore, global knockdown of drift using two independent transgenic RNAi lines (7578 and HMC04813) driven via a strong global promoter (actin-Gal4) also resulted in either complete or highly penetrant lethality prior to the adult stage. Thus, similarly to SLC52A3 (Yonezawa and Inui, 2013;Intoh et al., 2016;Yoshimatsu et al., 2016), and consistent with a fundamental requirement for riboflavin transport in metazoans (Biswas et al., 2013), DRIFT is likely to be essential in Drosophila.
Using a distinct global driver (daughterless-Gal4; da-Gal4) in concert with the 7578 RNAi line, we achieved robust drift knockdown as determined by semi-quantitative RT-PCR ( Supplementary Fig. 6C). However, this knockdown did not result in early lethality (suggesting weaker RNAi expression by da-Gal4 relative to actin-Gal4), facilitating analysis of drift knockdown flies at later developmental stages. Using whole-body tissue from drift knockdown adults and associated control lines (heterozygotes for da-Gal4 or the drift RNAi transgene), we found that drift knockdown resulted in a substantial reduction of in vivo riboflavin levels as well as the riboflavin metabolites FMN and FAD (Fig. 4A-C). These results, combined with the high homology of DRIFT to hRFVT1-3, strongly suggest that DRIFT is a bona fide riboflavin transporter.
We next asked whether drift knockdown resulted in reduced mitochondrial activity. Indeed, ETC complex I activity was significantly reduced by drift knockdown (Fig. 4D), while ETC complex II and II-III activity exhibit a trend towards lower levels, albeit non-significant ( Fig. 4E and F). Thus, ETC complex I activity appears particularly sensitive to DRIFT expression.
We also tested whether similar reductions in riboflavin metabolism and ETC activity could be observed in BVVL syndrome patient tissue. Using fibroblasts derived from BVVL syndrome patients with riboflavin transporter mutations and healthy age-matched controls (Supplementary material), we found a significant reduction in the intracellular levels of FMN and FAD in patient fibroblasts when grown in low extracellular riboflavin conditions (Supplementary Fig. 7A and B). Levels of intracellular riboflavin in patient fibroblasts frequently fell below the threshold of detection under these conditions, but not in control fibroblasts (data not shown), consistent with defective riboflavin transport. Furthermore, we observed a significant reduction in ETC complex I and complex II activity in patient fibroblasts compared to controls (Supplementary Fig. 7C and D). Taken together, the above in vivo and in vitro data suggest a conserved role for riboflavin transporters in regulating ETC activity, particularly ETC complex I. drift knockdown lowers mitochondrial membrane potential and affects respiratory chain activity Given the reduced ETC complex I activity in drift knockdown flies, we investigated whether loss of drift affected mitochondrial membrane potential (ÁÉ m ), an indicator of mitochondrial health and function. Mitochondrial membrane potential was assessed in the brains of 1-day-old flies using tetramethylrhodamine, methyl ester (TMRM) fluorescence and imaging with confocal microscopy. This revealed a significant reduction in mitochondrial membrane potential in drift knockdown brains compared to control ( Supplementary Fig. 8A).
We next isolated mitochondria from drift knockdown and control flies and assessed respiratory chain activity by measuring oxygen consumption using a Clark-type oxygen electrode. The steps in respiration were compared using substrates and inhibitors specific to individual respiratory complexes. We found that drift knockdown leads to significantly increased oxygen consumption in state 2 (ADP independent) respiration, when using substrates for complex I (5 mM pyruvate and 5 mM malate) but not substrates for complex II (5 mM succinate in the presence of 5 mM rotenone; Supplementary Fig. 8B). The respiratory control ratio (RCR) is the ratio of state 3 respiration (ADP stimulated) to state 4 respiration (no ADP present), and is considered an indication of the degree of coupling of mitochondrial respiratory chain activity to oxidative phosphorylation (Chance and Williams, 1955). RCR was significantly lower in drift knockdown mitochondria compared to control flies ( Supplementary Fig. 8C). Furthermore, the efficiency of oxidative phosphorylation (measured as ratio ADP/O) was also significantly lower in drift knockdown mitochondria ( Supplementary Fig. 8D). Collectively, these results suggest that loss of drift activity reduces mitochondrial membrane potential as a result of mitochondrial uncoupling.

DRIFT is required for normal mitochondrial morphology
Given the alterations in mitochondrial function resulting from reduced DRIFT levels, we also assessed whether mitochondrial morphology was regulated by DRIFT. To do so, we used electron microscopy to enable ultrastructural examination of mitochondria within the neuropil regions (axons, dendrites and synaptic domains) of the adult nervous system in drift knockdown flies and an associated control. Strikingly, we frequently observed grossly abnormal mitochondria containing swollen vacuoles in drift knockdown neurons, coupled with fragmented mitochondrial cristae and a concomitant reduction in the area of mitochondrial cristae ( Fig. 4G and H). Grouping mitochondria from control and drift knockdown neurons in three categories (exhibiting no, minor, or major mitochondrial swelling and cristae fragmentation) (Fig. 4I), we found that morphologically normal mitochondria were depleted in drift knockdown neurons, whereas the proportion of mitochondria exhibiting gross structural defects was greatly increased (Fig. 4J). Thus, we have identified a profound neuronal endophenotype associated with reduced riboflavin transporter expression: vacuolar mitochondria with fragmented cristae.

drift knockdown results in reduced locomotion and lifespan in Drosophila
The viability resulting from drift knockdown via da-Gal4 allowed us to assess whether riboflavin transporter knockdown impacts post-embryonic organismal phenotypes in a manner consistent with BVVL syndrome pathology (Foley et al., 2014;Manole et al., 2014). To test for locomotor defects, we used the Drosophila activity monitor (DAM) system to perform automated recordings of adult locomotion, measured as the number of infra-red beam breaks across a 24 h day/night cycle. Under 12 h light: 12 h dark conditions, control 1-2 day old adult female Drosophila exhibited peaks of activity at dawn and dusk, and relative quiescence during the afternoon and night (Fig. 5A and B). In contrast, peak activity and total beam breaks in drift knockdown adults were substantially reduced ( Fig. 5C  and D). Furthermore, drift knockdown resulted in greatly reduced lifespan, with 99% mortality within 4 days posteclosion (Fig. 5E). As BVVL syndrome is an early-onset disorder, we also tested for locomotor defects at earlier stages of the Drosophila lifecycle. Indeed, we found that drift knockdown third instar larvae also exhibit significantly reduced locomotion, as measured by the number of grid crosses per minute on an agar plate (Fig. 5F). These Drosophila phenotypes mimic motor dysfunction and early mortality observed in patients with BVVL syndrome, suggesting a conserved link between riboflavin transporter dysfunction, locomotor strength and lifespan.
Recently, reductions in axonal length and transcriptional signatures of altered cytoskeletal composition were identified in BVVL syndrome iPSC-derived motor neurons (Rizzo et al., 2017). To test for altered axonal development and signatures of cytoskeletal dysfunction in vivo, we examined the neuromuscular junction of drift knockdown third instar larvae. Interestingly, despite the clear reduction in locomotion in drift knockdown larvae (Fig. 5F), motor neuron development appeared unperturbed. Motor neuron axons correctly innervated the muscle surface and there was no reduction in the number of synaptic boutons per motor neuron following drift knockdown ( Supplementary Fig. 9A and B). Mutations in cytoskeletal components such as Ankyrin, Spectrin, HTS/Adducin and Futsch (a microtubule-binding protein) cause well-defined alterations in the morphology of larval motor neurons. These include synaptic retraction (disassembly of active zones and degradation of inter-bouton axonal regions, leaving small isolated boutons) (Pielage et al., 2005(Pielage et al., , 2008(Pielage et al., , 2011, aberrant filopodia-like protrusions (Pielage et al., 2011), or a reduction in synaptic growth (Roos et al., 2000). As we did not observe similar phenotypes at the neuromuscular junction of drift knockdown larvae, we infer that cytoskeletal organisation is grossly normal in this background, and therefore that locomotor defects can initially occur in the absence of altered synaptic development and/or neurodegeneration following drift knockdown.

An esterified riboflavin derivative rescues drift knockdown phenotypes
Since BVVL syndrome pathology can be partially ameliorated by riboflavin treatment, we asked whether locomotor defects in drift knockdown flies could be rescued by supplementing Drosophila culture medium with riboflavin (0.1 mg/ml). However, we found no enhancement of locomotor activity following riboflavin supplementation ( Supplementary Fig. 10A). Riboflavin is a water-soluble vitamin that is easily excreted, leading to low bioavailability and short half-life. Furthermore, since riboflavin transporter expression is very low in drift knockdown flies ( Supplementary Fig. 6C), riboflavin in the Drosophila haemolymph may fail to be transported into relevant cell types. We sought to circumvent these issues using an esterified derivative of riboflavin (riboflavin-5'-lauric acid monoester; RLAM; 0.1 mg/ml) that could act as a prodrug, likely diffuse into the intracellular space independently of riboflavin transporter function and be cleaved by esterases to release active riboflavin (Fig. 6A). Strikingly, food supplementation with RLAM resulted in heightened locomotion ( Fig. 6B-D) and an extension of lifespan ( Fig.  6E) in drift knockdown flies. This rescue correlated with substantially increased complex I activity ($3-fold) following RLAM supplementation (Fig. 6F). In contrast to drift knockdown flies, RLAM did not induce a significant foldchange in locomotor activity in control adults, suggesting that RLAM is non-toxic at the concentration and time scale of supplementation tested ( Supplementary Fig. 10B).
The rescue of adult drift knockdown by RLAM is clearly partial. For example, activity in drift knockdown flies fed RLAM was still significantly lower than control flies fed RLAM (beams breaks/24 h: da 4 + controls = 1457 AE 58, n = 47; da 4 drift RNAi = 415 AE 27, n = 53; P 5 0.0001, Mann-Whitney U-test). We speculated that this partial rescue might be due to the fact that during pupariation, a period of substantial morphological growth and remodelling in the lifecycle of holometabolous insects, pupae will not be exposed to RLAM. Reduced intracellular riboflavin concentrations during pupariation may in turn impact critical aspects of development, resulting in adult flies that can only be partially rescued by RLAM supplementation. If true, we would expect to observe a more robust rescue with RLAM during earlier phases of development, when larval Drosophila are continually exposed to RLAM in the food. Indeed, in drift knockdown L3 larvae fed RLAM, we observed a rescue of larval locomotor activity to wild-type levels (Fig. 6G). Similarly to adult flies, riboflavin supplementation had no effect on the activity of drift knockdown L3 larvae (Supplementary Fig. 10C). Thus, we hypothesize that RLAM may represent a more efficient treatment method for BVVL syndrome patients, since-in contrast to riboflavin-cellular uptake of RLAM may still robustly occur in the absence of functional riboflavin transporters.

Discussion
We expanded the spectrum of genetic defects in riboflavin transporters identifying SLC52A2 and SLC52A3 mutations in 15% of cases. It was previously noted that the distinct phenotype of upper limb and axial weakness, hearing loss and optic atrophy could be attributed only to patients harbouring SLC52A2 mutations (Foley et al., 2014). However, in our cohort of 20 patients, mutations in both SLC52A2 and SLC52A3 resulted in similar phenotypes. SLC52A3 mutations were more frequent than mutations in SLC52A2 in our cohort. This is in agreement with previous reports where BVVL syndrome-linked mutations were predominantly found in SLC52A3 (Bosch et al., 2011; Manole and Houlden, 2015). The high proportion of patients negative for mutations in the known riboflavin transporters indicates that novel genetic causes are yet to be found.
The neuropathology of BVVL syndrome has not been fully characterized (Sathasivam, 2008;Malafronte et al., 2013). In keeping with the spectrum of clinical findings, severe depletion of neurons in multiple cranial nerve nuclei, anterior horns of the spinal cord, cerebellar nuclei, Purkinje cells in the cerebellum, and degeneration of optic pathway, solitary tract, and spinocerebellar and pyramidal tracts was seen along with an axonal neuropathy in the sural nerve (Foley et al., 2014). Here we provide detailed neuropathological assessment of the brains and spinal cords in two patients with genetically confirmed mutations in the SLC52A3 gene who fell at the two extremes of age of presentation of BVVL syndrome. There was variably and severe neuronal loss and gliosis in the brain stem cranial nerve nuclei and anterior horns of the spinal cord with accompanying nerve root atrophy, which reflects the spectrum of clinical symptomatology. While no correlation between the mutation type and severity of clinical phenotypes was evident in the case series presented here, it is possible that different mutations directly influence the length of the disease and degree of atrophy of specific brain structures. For example, the rapid deterioration and demise of Patient AM2 due to formation of truncated protein originating from nonsense mutations on both alleles may explain less prominent atrophy of cerebellar Purkinje cells when compared with Patient AM4, the atrophy of which may require longer duration of the disease (Supplementary Fig. 2). In both patients there were symmetrical lesions in the brain stem characterized by prominent neovascularization, dense infiltration of macrophages, loss of myelin and relative preservation of neurons and axons. Although the Complex activity is expressed as a ratio to citrate synthase (CS) activity. Data were generated in at least three independent experiments; n = 10 for each genotype. (G) The number of grid breaks per min of wandering third instar larvae with and without RLAM supplementation; n = 20 for each genotype. **P 5 0.005, ***P 5 0.0005, ns = P 4 0.05, Mann-Whitney U-test (B and F) or Kruskal-Wallis test with Dunn's post hoc test (G).
anatomical distribution of the symmetrical lesions differed in both cases (Supplementary Fig. 1 and 3), the morphology of the lesions was identical in both cases and was similar to the pathology seen in mitochondrial encephalopathies (Tanji et al., 2001;Filosto et al., 2007). To the best of our knowledge such lesions have not been documented in any of the previous published cases of BVVL clinical syndrome and link with the mitochondrial abnormalities in Drosophila and patient fibroblasts.
Recent work has sought to model the effects of riboflavin transporter mutations in iPSC-derived motor neurons from patients with BVVL syndrome (Rizzo et al., 2017) or SLC52A3 knockout mice (Intoh et al., 2016;Yoshimatsu et al., 2016). Motor neurons can still be successfully differentiated from embryonic stem cells derived from mice lacking SLC52A3 (Intoh et al., 2016), suggesting that RFVT3 (the SLC52A3 gene product) is not required for motor neuron development in culture. In contrast, Rizzo et al., (2017) recently identified reduced axonal growth and transcriptional alterations in cytoskeletal factors, including neurofilament genes, in iPSCS-derived motor neurons from SLC52A3-BVVL syndrome patient cells. Our results are inconsistent with the premise that cytoskeletal defects occur in motor neurons of Drosophila larvae with reduced DRIFT, and suggest that DRIFT is not required for development of larval motor neurons, similar to findings for the SLC52A3 knockout mouse (Intoh et al., 2016). We note that the Drosophila genome lacks neurofilament genes, preventing us from assessing the potential relevance of this pathway to neuronal dysfunction in BVVL syndrome model Drosophila. However, the fact that locomotor defects and reduced lifespan still occur in drift knockdown flies in the absence of neurofilaments suggest that additional pathways are likely to contribute to neuronal dysfunction following loss of riboflavin transporter in flies, and potentially mammals. Based on our findings in both patient fibroblasts and Drosophila, we suggest that mitochondrial dysfunction may represent one such pathway that impacts neuronal signalling and viability in patients with BVVL syndrome.
Although fibroblasts are non-neural cells and not particularly vulnerable as a tissue, metabolic and mitochondrial abnormalities are commonly studied in this cell type (Distelmaier et al., 2009). We identified clear deficiencies in the activities of ETC complex I and complex II in patient-derived fibroblasts relative to fibroblasts from healthy controls. Interestingly, previous results have only shown evidence of marginally decreased ETC complex I activity in muscle cells from some SLC52A2 patients sensitive to reduced riboflavin uptake (Foley et al., 2014). Mitochondrial activity in different tissues and cell types may thus be differentially sensitive to reduced riboflavin uptake. Biochemical analysis of Drosophila drift knockdown tissue is consistent with data derived from patient fibroblasts, showing diminished levels of riboflavin, FMN and FAD, and reduced complex I activity when expression of the single Drosophila riboflavin transporter is perturbed. This in turn led to reduced mitochondrial membrane potential and abnormal mitochondrial respiratory chain activity.
Strikingly, we also observed profound abnormalities in mitochondrial ultrastructure in Drosophila neurons following drift knockdown. In contrast to the Drosophila genome, which encodes a single riboflavin transporter, the human genome contains three riboflavin transporter paralogs that exhibit partially overlapping expression domains, including within the brain (Yonezawa and Inui, 2013; www.human.brain-map.org). Given the potential for functional redundancy between riboflavin transporters co-expressed in the same neuron, we hypothesize that similar alterations in mitochondrial function and morphology in patients with BVVL syndrome are most likely to occur in neurons in which a single mutated riboflavin transporter is expressed. It will be intriguing to document neuronal subtypes with such unique riboflavin transporter signatures and to assess whether these are preferentially sensitive to corresponding riboflavin transporter mutations. At the whole-organismal level, there was impairment in locomotion at both the larval and adult stages of drift knockdown flies, reminiscent of the limb weakness and movement impairment of the BVVL syndrome patients. Moreover, drift knockdown flies had severely reduced life span, similar to untreated patients (Bosch et al., 2011). It is interesting to note that these lifespan and locomotor defects are rescued by ingestion of an esterified derivative of riboflavin, since this provides further evidence that the phenotypic signatures of drift knockdown flies are linked to riboflavin deficiency and consequent downstream metabolic defects. RLAM fulfils the criteria for being a treatment targeting energy metabolism, a fundamental mitochondrial process, in addition to circumventing the disease-related protein, features generally believed to be of great promise (Lin and Beal, 2006;Xuan et al., 2013;Foley et al., 2014).
It was previously thought that defective mitochondria occurs secondary to primary disease mechanism, but current research suggests that mitochondrial dysfunction may play a role in both onset and development of certain neurodegenerative diseases such as Alzheimer's disease, Parkinson's disease, Huntington's disease and ALS, where decreases in one or more mitochondrial complexes and corresponding oxidative stress have been reported in cell or animal models (Menzies et al., 2002;Hoglinger et al., 2003;Fukui and Moraes, 2007;McInnes, 2013;Yamada et al., 2014;Golpich et al., 2017). Our results from in vitro and in vivo models of a childhood neuropathy are in line with these findings, highlight a potential contribution of mitochondrial dysfunction to BVVL syndrome pathology, and suggest future therapeutic strategies based on esterified riboflavin.