Role of Flavinylation in a Mild Variant of Multiple Acyl-CoA Dehydrogenation Deficiency

Mutations in the genes encoding the α-subunit and β-subunit of the mitochondrial electron transfer flavoprotein (ETF) and the electron transfer flavoprotein:ubiquinone oxidoreductase (ETF:QO) cause multiple acyl-CoA dehydrogenation deficiency (MADD), a disorder of fatty acid and amino acid metabolism. Point mutations in ETF, which may compromise folding, and/or activity, are associated with both mild and severe forms of MADD. Here we report the investigation on the conformational and stability properties of the disease-causing variant ETFβ-D128N, and our findings on the effect of flavinylation in modulating protein conformational stability and activity. A combination of biochemical and biophysical methods including circular dichroism, visible absorption, flavin, and tryptophan fluorescence emission allowed the analysis of structural changes and of the FAD moiety. The ETFβ-D128N variant retains the overall fold of the wild type, but under stress conditions its flavin becomes less tightly bound. Flavinylation is shown to improve the conformational stability and biological activity of a destabilized D128N variant protein. Moreover, the presence of flavin prevented proteolytic digestion by avoiding protein destabilization. A patient homozygous for the ETFβ-D128N mutation developed severe disease symptoms in association with a viral infection and fever. In agreement, our results suggest that heat inactivation of the mutant may be more relevant at temperatures above 37 °C. To mimic a situation of fever in vitro, the flavinylation status was tested at 39 °C. FAD exerts the effect of a pharmacological chaperone, improving ETF conformation, and yielding a more stable and active enzyme. Our results provide a structural and functional framework that could help to elucidate the role that an increased cellular FAD content obtained from riboflavin supplementation may play in the molecular pathogenesis of not only MADD, but genetic disorders of flavoproteins in general.

In agreement, our results suggest that heat inactivation of the mutant may be more relevant at temperatures above 37°C. To mimic a situation of fever in vitro, the flavinylation status was tested at 39°C. FAD exerts the effect of a pharmacological chaperone, improving ETF conformation, and yielding a more stable and active enzyme. Our results provide a structural and functional framework that could help to elucidate the role that an increased cellular FAD content obtained from riboflavin supplementation may play in the molecular pathogenesis of not only MADD, but genetic disorders of flavoproteins in general.
Multiple acyl-CoA dehydrogenation deficiency (MADD, 2 MIM 231680; also designated glutaric acidemia type II (GAII)) is caused by mutations in either of the genes encoding the two subunits of electron transfer flavoprotein (ETF) or the monomeric enzyme ETF:ubiquinone oxidoreductase (ETF:QO) (1,2). The flavin adenine dinucleotide (FAD) in ETF receives electrons from at least 12 different mitochondrial FAD-containing dehydrogenases, which are involved in fatty acid ␤-oxidation and amino acid metabolism (3). ETF is subsequently oxidized by the membrane-bound ETF:QO. This last component, which harbors both a FAD and an [4Fe-4S] cluster, mediates the transfer of reducing equivalents to the respiratory ubiquinone pool, leading to subsequent ATP production (4,5). As a result of deficient function of either ETF or ETF:QO, the acyl-CoA dehydrogenases are blocked because they cannot transfer the electrons gained in the dehydrogenation reactions, entailing accumulation of various acyl-esters in blood and urine, hence the term MADD (1). The clinical features of patients suffering from MADD are rather heterogeneous. It ranges from lethal cases with neonatal anomalies to mildly affected individuals, presenting in childhood or adulthood with hypoglycemic, encephalopathy, and/or myopathy (1,6). There is evidence that the severity of the clinical phenotype, to some extent, depends on the location and nature of mutations in the genes encoding ETF or ETF:QO, with null mutations severely affecting mRNA expression, processing and/or stability being associated with lethal disease and missense mutations leaving some residual ETF/ETF:QO enzyme activity being associated with milder clinical forms (6 -9). In patients with milder disease variants, symptoms are often intermittent and only become evident during periods of illness and catabolic stress, indicating that in this group of patients, in whom residual ETF/ETF:QO enzyme activity allows modulation of the enzymatic phenotype, the disease severity does not depend only on the nature of the gene defect but also on cellular factors that may modulate the enzymatic phenotype (6). This potential for in vitro modulation of the enzymatic phenotype has been established for a large number of disease-causing mutations affecting flavin-containing mitochondrial acyl-CoA dehydrogenases (10). In these cases, missense mutations have been shown to impair folding to the native structure and/or destabilize the native folded structure, resulting in decreased enzyme levels, which are subject to modification by environmental conditions like temperature and availability of quality control chaperones and proteases (11)(12)(13). The cofactor FAD is another cellular factor that may modulate the enzymatic phenotype of disease-causing mutations in mitochondrial flavoproteins. It has been observed for acyl-CoA dehydrogenases and for ETF that levels of available FAD have a strong impact on folding and maintenance of the native structure (14 -17). Moreover, supplementation of certain MADD patients (3, 18 -22), and some patients with isolated deficiencies of acyl-CoA dehydrogenases (23)(24)(25), with FAD or its precursor riboflavin results in an increase of the enzymatic activities and improvement of clinical symptoms (as reviewed in Ref. 26), indicating that the enzymatic phenotype can be amenable to modulation by FAD in vivo also. Decreased dietary intake of riboflavin, and also certain physiological conditions like pregnancy (reviewed in Ref. 27), fasting (28), exercise (reviewed in Ref. 29), and infections (reviewed in Ref. 30)) may induce a depletion in FAD content, which will pose high demands on flavoproteins and perhaps especially on those with folding defects. This raises the interesting hypothesis that availability of FAD may be a predisposing factor in the cellular pathogenesis of genetic disorders of acyl-CoA dehydrogenations. However, the additional possibility of a ligand-induced folding effect exerted upon FAD binding to a mutant protein, which becomes destabilized upon an adverse physiological condition, has not yet been explored. Here we address this hypothesis by studying in vitro the effect of flavinylation on the folding, conformational quality and proteolytic susceptibility of a mild variant of human ETF (ETF␤-D128N). Studies were also performed under physiological heat stress conditions, to mimic the adverse physiological factors of a fever event.

EXPERIMENTAL PROCEDURES
Chemicals-All reagents were of the highest purity grade commercially available. Octanoyl-CoA, FAD, AMP, and urea were purchased from Sigma. Isopropyl-␤-D-thiogalactopyranoside was purchased from VWR International.
Gene Expression and Protein Purification-Escherichia coli JM109 cells from Promega transformed with ETF plasmids for the wild type (WT), and ETF␤-D128N variant were grown as described previously (6). Briefly, cells were grown in dYT medium (16 g of Bacto Tryptone, 10 g of Bacto Yeast extract and 5 g of NaCl) supplemented with 10 g⅐ml Ϫ1 kanamycin at 37°C or 30°C, up to an absorbance at 532 nm between 0.5 and 0.8, and then induced with 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside for 4 h. Cells were harvested by centrifugation, resuspended in 10 mM Hepes, 10% ethylene glycol, and 0.5 mM phenylmethylsulfonyl fluoride (Roth) (Buffer A) in presence of 0.1 mg⅐ml Ϫ1 FAD, DNase (PVL), and disrupted in a French press. The soluble extract was applied to a 10-ml Q-Sepharose fast flow (Amersham Biosciences) equilibrated in buffer A. The column was washed with five column volumes (Vc) of buffer A, and bound proteins were eluted by a linear gradient ranging from 0 -1 M NaCl, in buffer A. ETF eluted as a pure protein at ϳ250 mM salt, as confirmed by SDS/PAGE. The proteins were fast-frozen using liquid nitrogen and stored at Ϫ80°C. To ensure full occupancy of FAD sites, the protein was incubated with 2.5-fold molar excess FAD at 4°C overnight, and the free cofactor was removed with extensive washings by ultrafiltration/dilution. This procedure yielded FAD-to-protein ratios higher than 0.95 both for WT and ETF␤-D128N. Unless otherwise mentioned, all experiments were performed with the proteins containing full occupancy of FAD sites.
Structural Analysis-The crystallographic structure of human ETF (PDB code: 1efv) was visualized using PyMOL (DeLano Scientific). Analysis of the molecular interactions, cofactor contacts, topological features, and generation of the model of the ETF␤-D128N was carried out using the WhatIF web server and the PDBsum data base (31).
Biochemical Methods and Activity Assays-Protein concentration was determined using the Bradford assay. Flavin content was determined using the molar extinction coefficient ⑀ 436 nm ϭ 13,400 M Ϫ1 ⅐cm Ϫ1 reported for FAD bound to ETF (5). The enzymatic activity of the purified proteins was measured monitoring DCPIP reduction, in a coupled assay in which recombinant human MCAD (0.13 M) and octanoyl-CoA (13 M), were employed, as described in Ref. 20. One unit of catalytic activity is defined as nmol of DCPIP reduced per minute, in the conditions used in the assay. All specific activities reported are based on total flavin content.
Spectroscopic Techniques-UV/visible spectra were recorded at room temperature in a Shimadzu UVPC-1601 spectrometer. Far-UV CD spectra were recorded on a Jasco J-715 spectropolarimeter with Peltier temperature control. A quartz polarized 1-mm path length quartz cuvette (Hellma) was used, and protein concentrations ranged from 0.1-0.2 mg⅐ml Ϫ1 . Fluorescence spectroscopy was performed using a Cary Eclipse instrument. For tryptophan emission studies, excitation was set at 280 nm, whereas for FAD emission, excitation was carried out at 436 nm. Unless otherwise noted, the slits for excitation and emission were set to 5 and 10 nm, respectively. Typically, ETF protein concentration in fluorescence studies was 1 M.
Flavin-depleted ETF-Apo-ETF was prepared using KBr solutions as in Ref. 32. Briefly, proteins were incubated in 3 M KBr, 10 mM Hepes pH 7.0, and 20% ethylene glycol solution for 1 h. Released FAD was washed by repeated ultrafiltration/dilution in the presence of 1 mM dithiothreitol and a 10-fold molar excess AMP. In the case of the D128N variant, we also used a milder procedure taking advantage of the spontaneous process of flavin release, which had already been described for ETF (33), to produce a FAD-depleted fraction: when the protein is stored at Ϫ20°C during 60 days, a decrease in FAD content and in the specific activity was observed, without a significant alteration of the protein folding, as determined by far-UV CD. Unbound FAD was removed by repeated ultrafiltration/dilution, yielding a sample with FAD content as low as 10%.
Dissociation Constant for FAD Binding-Apo-ETF (1-4 M) in the presence of 1 mM dithiothreitol, 10-fold molar excess of AMP, 10 mM Hepes pH 7.0, and 20% ethylene glycol was placed in a quartz cuvette at 15°C. After each addition of known amounts of FAD the mixture was allowed to equilibrate for 2 min, and the visible spectrum was measured. The spectra of free and bound FAD are clearly distinct, making possible to quantify the concentration of each species at a given point. The fraction of bound FAD (f bound ) was calculated using the proportionality where A i is absorbance at the wavelength i, and ⑀ i is extinction coefficient of free FAD at a wavelength. In this case, the wavelengths 465 nm (⑀ 1 ϭ 10190 M Ϫ1 cm Ϫ1 ) and 435 nm (⑀ 2 ϭ 10150 M Ϫ1 cm Ϫ1 ) were used to distinguish between bound and free FAD, respectively. The concentration of free FAD in the equilibrium was measured by the equation [FAD] free ϭ [FAD] total Ϫ [ETF] total ϫ f bound . The dissociation constant (K D ) was obtained by fitting the data of two independent experiments, using the equation, Kinetics of Flavin Release-Wild-type ETF and the ETF␤-D128N (1 M), both with full occupancy of FAD sites, were incubated at 39°C for up to 60 min in 10 mM Hepes, 20% ethylene glycol, pH 7.0. The kinetics of flavin release was monitored from the increase of the 530-nm emission peak arising from free FAD upon excitation at 436 nm. After 1 h of incubation, the samples were boiled for 5 min, to release the remaining FAD, and a spectrum was collected at 39°C. The fraction of bound FAD was determined at each point in respect to the point of 100% release of FAD.
Chemical Stability Studies-ETF␤-D128N (16 M) depleted of flavin by milder procedure (see above), was incubated overnight with 10 mM Hepes, 10% ethylene glycol pH 7.8 at 4°C, in the absence and in the presence of a 2.5-fold excess of FAD (40 M). Samples were diluted in different urea solutions, in 2 mM Hepes pH 7.8, equilibrated at room temperature for 15 min, after which the Trp emission spectra were recorded. Accurate concentration of the urea stock solutions was confirmed by refractive index measurements (34), and the pH was verified before the experiment. Fluorescence emission data in the 300 -400-nm interval was analyzed determining the average emission wavelength (AEW), which takes into account both variations in the emission intensity and position (35). The denaturation curves were determined plotting the AEW as a function of urea concentration. Data were fitted to a sigmoid curve allowing the determination of an apparent C m (denaturant concentration at the curve midpoint). The biological activity of the samples incubated overnight was also determined using the assay described above (20).
Limited Proteolysis with Trypsin-Wild type and ETF␤-D128N with full occupancy of FAD sites were incubated overnight in 10 mM Hepes, 10% ethylene glycol pH 7.8 at 4°C, in the presence of a 2.5-fold molar excess of FAD. Control samples without added FAD were also prepared. These fractions were used to test the susceptibility toward trypsin digestion. Samples were incubated with trypsin (bovine pancreas trypsin; PVL) at 35°C in 0.1 M Tris/HCl pH 8.5, at a 10-fold excess over the protease. As a control, identical samples without trypsin were submitted to the same procedure. Aliquots with 0.05 nmol of protein were sampled at different time points up to 2 h, and the reaction was stopped by adding SDS-PAGE loading buffer (2% SDS and 5% ␤-mercaptoethanol). As an internal standard for the quantity of loaded protein in the gel, bovine serum albumin (5 M) final concentration was also added to loading buffer solution. The products of the proteolysis reaction were analyzed by 12% SDS/PAGE, which were stained with Coomassie Blue. Protein was quantified densitometrically (Bio-Rad Chemidoc XRS), and the percentage of undigested protein was calculated in respect to the total amount of protein at time 0. Reported values refer to the sum of intensities of ␣ and ␤ bands.
Thermal Stress Studies-ETF␤-D128N and wild-type ETF (12 M), in 10 mM Hepes, 20% ethylene glycol pH 7.0, were incubated (100 l) at different temperatures (33-44°C) for 1 h, and the residual activity was determined in relation to the activity at time 0 (duplicate measurements were performed for each temperature, and two independent protein batches were assayed). To study the effect of FAD on the loss of activity, ETF␤-D128N (16 M) was incubated at 39°C up to 90 min, in the presence of a 2.5-fold excess of FAD (40 M). Control samples without added FAD were also prepared and treated in the same conditions. At ϳ10-min intervals, 3.5 l of the samples were collected to determine the residual activity as indicated above. For the measurements using fluorescence spectroscopy, spectra were collected at 2-min intervals, up to 1 h. Trp emission was recorded from 300 -400 nm with excitation at 280 nm. Unless otherwise noted, ETF concentration used was 1 M.

Impact of the ETF␤-D128N
Mutation-ETF is a heterodimer composed of 30 kDa (ETF␣) and 28 kDa (ETF␤) subunits, containing one structural AMP, and one catalytic FAD group (2). The MADD ETF␤-D128N mutation (6,36), which is here investigated, occurs in a ␤-turn within a very conserved region of the ␤-subunit, that comprises the segment 126-AIDDD-130 (Fig. 1). These residues have contacts with the AMP and are involved in intersubunit interactions with residues from the ␣-subunit, via hydrogen bonds and non-bonded contacts. In particular, Asp-128 is directly involved in a network of interactions: it is salt-bridged to the nearby Lys-11 (2.96 Å) from the ␤-subunit N terminus; and it contacts with several residues from the ␣-subunit (Ile-148, Tyr-149, Asn-152). Also, ETF␤-D130 is hydrogen-bonded to ETF␣-Q265, a residue that contacts with the FAD ring. Therefore, although Asp-128 is not directly involved in an interaction with the flavin, it is located in a relatively sensitive region of the protein, in respect to dimer contacts and flavin binding. Indeed, this is noted by the decreased catalytic activity of purified ETF␤-D128N variant (400 units⅐mg Ϫ1 ), which is only ϳ30% of that of the wild-type protein. Also, the temperature deactivation profile of the ETF␤-D128N variant is much more pronounced than that of the wild-type protein (Fig. 2). For the ETF␤-D128N variant, a rapid drop of the activity is observed if the protein is incubated for 60 min at temperatures above 37°C, whereas at 37°C and below, more than 85% of the initial activity is retained.
Folding and Conformation of ETF␤-D128N-To define the structural characteristics of the ETF␤-D128N variant protein, we have investigated the folding and conformational properties of the purified protein (Fig. 3). Circular dichroism (CD) analysis on the far-UV region showed that the spectrum of the mutant protein, is dominated by minima at 222 and 209 nm, which are typical of a well-folded ␣/␤ protein (Fig. 3A). In fact, the two ETF subunits, although they share very little sequence identity, have the same structural topology, consisting of a three-layered ␣␤␣ sandwich Rossman fold architecture. A comparison between the far-UV CD spectra of the mutant and wild-type ETF shows that the mutation has no impact on the overall fold of the protein. More subtle effects on the tertiary structure were monitored using fluorescence spectroscopy. The protein contains two tryptophan residues, one in each subunit (ETF␣-W199 and ETF␤-W144), which are relatively accessible to the solvent, making them particularly sensitive conformational probes. The Trp emission maximum of ETF␤-D128N is 10-nm red-shifted in respect to that of wild-type protein (supplemental Fig. S1A). As a lower emission maximum denotes a more solvent-shielded tryptophan, this shift in the emission from ϳ320 to ϳ330 nm indicates that the aromatic moieties in ETF␤-D128N are more easily accessible to water molecules. This result suggests looser tertiary contacts in ETF␤-D128N, in comparison to wild-type ETF.
The FAD moiety in ETF␤-D128N was analyzed by visible absorption spectroscopy. The two bands typical of flavin, with maxima at 436 and 373 nm, were observed (Fig. 3B). The latter, is ϳ3-nm blue-shifted in respect to the band observed in wildtype ETF, which has a maximum at 376 nm. Additionally, the intensity of the FAD fluorescence is a reporter of the status of the flavin moiety, as the FAD emission is quenched when it is bound to the protein (37). The FAD, which is in an extended conformation according to the crystal structure, has a similar emission spectrum in both wild-type and mutant proteins, denoting a characteristic weak band centered at ϳ500 nm.
Cofactor Dissociation in ETF-To verify if the ETF␤-D128N mutation induces a perturbation of the flavin binding site, the dissociation constants for FAD binding to wild-type ETF and ETF␤-D128N mutant were measured. When substoichiometric quantities of FAD were added to wild-type ETF apoprotein, the characteristic ETF visible spectrum

FIGURE 3. Far-UV CD (A) and visible absorption (B) spectra of ETF␤-D128N (-) and wild-type ETF (⅐⅐⅐⅐⅐⅐).
The protein concentrations used were 1.7 M for the far-UV CD spectroscopy and 17 M for the visible absorption analysis. In both cases, the buffer used was 2 mM Hepes, pH 7.8, and the FAD:protein ratio was higher than 0.95. Spectra were recorded at room temperature. FEBRUARY 13, 2009 • VOLUME 284 • NUMBER 7 JOURNAL OF BIOLOGICAL CHEMISTRY 4225 became immediately evident (Fig. 4A). In the case of ETF␤-D128N, an increased aggregation of the apoprotein was noted during the experiment, as evidenced from the increase of absorbance at lower wavelengths due to scattering (Fig. 4B). Nevertheless, the absorption spectrum characteristic of ETF was still observed.

Flavinylation Role in Acyl-CoA Dehydrogenation Deficiencies
Because the spectrum of free and bound FAD are distinct, it is possible to quantify the concentration of each species at a given point (as described under "Experimental Procedures"). By fitting the data with a one-site binding equation (Fig. 4, A and B, inset) the dissociation constant of K D ϭ 0.04 Ϯ 0.01 M and K D ϭ 0.04 Ϯ 0.02 M was obtained for the wildtype protein and ETF␤-D128N, respectively. These low values are close to the detection limit that can be measured in these conditions, and should correspond to an upper limit.
Considering that the effect of the mutation on the specific activity is more pronounced above 37°C (Fig.  2), the kinetics of flavin dissociation at 39°C was investigated for the ETF␤-D128N in comparison to wild-type ETF (Fig. 5). Given that the intensity of the emission spectrum of free FAD is higher than that of bound FAD (supplemental Fig. S1B), the release of flavin was determined monitoring the increase in flavin fluorescence emission, similarly to a previously described methodology (32). The point of 100% flavin release was measured after boiling the protein for 5 min. To rule out that the increase in intensity was due to conversion of FAD into FMN (10-fold more fluorescent), commercial FAD was treated in the same conditions, and no significant change in intensity was observed. Replicate assays were made, allowing to determine an apparent rate of release k off of 0.38 Ϯ 0.06 h Ϫ1 for ETF␤-D128N and 0.13 Ϯ 0.02 h Ϫ1 for WT (n ϭ 3). The value obtained for the wild-type ETF is in excellent agreement with the reported value in the literature (0.12 h Ϫ1 (32)). After incubation at 39°C, the original fluorescence spectrum could not be reverted by incubation at 4°C for 24 h. Moreover, the release of flavin correlates with a decrease in specific activity (see below, Fig.  8A), which could not be rescued by incubation with excess FAD. The results showed that under these conditions, the ETF␤-D128N variant releases FAD irreversibly at a 3-fold higher rate than wildtype ETF, denoting a weaker interaction of the cofactor with the protein at 39°C or a less stable apoprotein.
Flavinylation Improves the Conformational Stability of the apo D128N Variant-To test if flavin insertion or its absence has an effect on the conformational stability and overall folding of the protein, we analyzed the effect of flavinylation on the ETF␤-D128N variant. For this purpose, we determined urea denaturation curves, monitoring tertiary structure and packing changes in the protein from tryptophan-fluorescence emission.
The effect of FAD was tested using as a starting point a fraction of apo ETF␤-D128N, which was depleted in flavin (Ͻ0.1 mol FAD/mol protein) but that nevertheless retained, at least partially, the secondary structure and overall folding, as inferred from far-UV CD (see "Experimental Procedures"). In agreement with the lack of FAD, the enzymatic activity of this fraction was residual. The urea denaturation profile of this fraction yielded a cooperative transition, with an apparent midpoint denaturant transition (C m ) of 1.4 M (Fig. 6A). However, incuba- tion with FAD resulted in a dramatic increase of the C m up to 2.3 M, and improved the cooperativity of the transition as reflected by the steeper slope (Fig. 6A). These effects, which are indicative of a substantial increase in the conformational stability of ETF, correlate with an insertion of FAD in the protein, as denoted from the 14-fold increase in the enzymatic activity upon incubation (from 12 to 166 units⅐mg Ϫ1 ). Nevertheless this increase was less than the initial activity, when a ratio of 0.95 FAD:protein was obtained, meaning that apoprotein could not be totally rescued. This suggests that a heterogeneous solution of apoprotein was obtained, in agreement with some degree of irreversible aggregation seen before for the apoprotein of ETF␤-D128N mutant in the reconstitution experiments. Inspection of the tryptophan emission spectrum at the different conditions, allowed us to evaluate the changes in the protein packing and tertiary contacts, as a result of FAD binding. The presence of flavin induced a more compact conformation as seen by the tryptophan emission shift from 337 nm in apoprotein to 333 nm in the reconstituted form, which denotes a more solvent-shielded Trp moiety (38) (Fig. 6B).
Overall, ETF deflavinylation, which results in a structure with poorer tertiary interactions and decreased biological activity despite the maintenance of an ␣/␤ fold, can be rescued to some extent by reincorporation of FAD in the protein.
Flavinylation Modulates the Proteolytic Susceptibility of ETF␤-D128N-To test if the conformational stabilization afforded by flavinylation could contribute to minimize the propensity of the mutant ETF␤-D128N to be degraded, the resistance toward proteolysis was investigated. The assay used consisted of monitoring the extent of limited proteolysis by trypsin in the presence of FAD and in its absence. The rationale for this approach is that a destabilized conformation will have a higher number of cleavage sites accessible to digestion, as a result of a higher flexibility of the polypeptide chain (39). Therefore, this approach allows us to evaluate the propensity for in vivo degradation. Trypsin digestion was carried out at 35°C for up to 2 h. Samples drawn every 30 min were analyzed by SDS-PAGE (Fig. 7). The protein samples without trypsin were heated for 2 h. In the absence of FAD, the two subunits start to be substantially degraded after 60 min of incubation with trypsin (Fig. 7). After 120 min, the protein amount decreases to 70% (supplemental Fig. S2B). Interestingly, the presence of FAD could preserve the native conformation at 85% (supplemental Fig. S2B). Control experiments in the absence of trypsin and on wild-type ETF were also carried out (supplemental Fig. S2A). These results suggest that flavinylation, and the presence of FAD, induce a protease-resistant conformation in ETF. Moreover, the presence of FAD induces changes in the conformation of the ETF␤-D128N variant, which restore its proteolytic resistance to levels of the wild-type protein (supplemental Fig. S2B).
External FAD Preserves Folding and Enzymatic Activity during Thermal Stress-We have designed an in vitro experiment aimed at testing a possible effect of dietary riboflavin supplementation for a mild mutation like ETF␤-D128N during thermal stress. To analyze the relevance of an increased FAD level during a fever episode, the effect of incubating ETF␤-D128N during 60 min at 39°C, in the absence and in the presence of a 2.5-fold excess of FAD, was investigated (Fig. 8). Activity measurements during the thermal perturbation showed that the presence of FAD prevents loss of activity that goes down to 50% of the initial value (225 units⅐mg Ϫ1 ), in comparison to control  samples (Fig. 8A). In a parallel experiment, Trp fluorescence emission showed that the presence of FAD lowers the shift in the maximum wavelength, indicating that it contributes to maintain the aromatic residues in a more solvent-shielded environment. In contrast, in the absence of FAD, the increased Trp emission is indicative of poorer and less compact tertiary structure (Fig. 8B). In conclusion, the presence of FAD during thermal stress of ETF␤-D128N prevents the loss of tertiary structure and protein compactness, as well as the loss of enzymatic activity.

DISCUSSION
In numerous diseases involving missense mutations in mitochondrial flavoproteins, symptoms are in some cases improved by dietary riboflavin supplementation (3,(21)(22)(23)26). Perhaps because of the many reports on successful treatment of a variant group of MADD patients, the "riboflavin-responsive MADD patients" (3, 21, 22, 40 -43), milder forms of MADD in general are treated with riboflavin, but without knowing the molecular mechanism of this treatment. MADD is therefore a good candidate to study FAD sensitive conformational instability of flavoproteins due to disease-causing missense mutations. We chose the ETF␤-D128N variant protein for the following reasons: it was identified in homozygous form in a patient with MADD, who had no obvious symptoms until the age of 2 years when disease symptoms were precipitated in connection with virus infection and fever. The patient died after 2 days. A sibling from whom no genetic material was available, had died suddenly and unexpectedly at the age of 6 months (36). Subsequent analysis in cultured fibroblasts from the patient showed residual activity and presence of reduced levels of both ETF subunits (6). Upon expression of the mutant protein in an E. coli system, decreased levels of the mutant protein compared with wild type were observed. This could be partially rescued by growing the bacterial cells at lower temperature, but not by chaperone co-overexpression. Furthermore, the ETF␤-D128N mutant protein displayed reduced thermal stability (6). Based on these observations, it was hypothesized that the mutant protein displays decreased conformational stability, which is sensitive to environmental conditions like temperature, and likely to the availability of FAD cofactor.
With our analysis we have here shown that the ETF␤-D128N mutation decreases significantly the specific activity and introduces some plasticity on the tertiary structure, but does not negatively impact the overall protein fold. The temperature inactivation profile of ETF␤-D128N shows that it becomes more unstable at temperatures above 37°C. In agreement, FAD binds very tightly in both ETF␤-D128N and wild type protein at 15°C, but during heat stress (39°C) the flavin is released 3-fold faster in the mutant, with activity loss. This in vitro result is in line with the in vivo situation where the symptoms emerged associated to an episode of stress.
We then moved to analyze if flavinylation would have an effect on ETF stability, and found that saturation of the FAD binding site significantly improved the stability and conformation of the apoprotein, and rescued destabilized forms, as inferred from urea unfolding studies. This suggests that under a stress condition or a decreased level of FAD, protein becomes deflavinylated and therefore destabilized. External FAD increases the proteolytic stability, suggesting that an excess of FAD in the patient cell may increase the lifetime and availability of an active form of the protein. Analysis of this effect in the context of a physiological heat stress condition, which mimicked a fever episode (incubation at 39°C for 1 h), clearly showed that exogenous FAD prevented conformational destabilization and activity loss. Studies of tissue samples or cultured fibroblasts from patients with MADD have shown an activity increase upon supplementation with riboflavin or FAD (26). This increase in activity may result from a direct saturation of the active site, or from an improvement of the in vivo maturation and assembly of the protein involved. It has been shown that FAD plays an important role during in vitro folding of mitochondrial medium-chain acyl-CoA dehydrogenase (MCAD), nucleating the formation of a competent oligomeric conformation during Hsp60-assisted folding (15). Here we postulate another possibility, which may occur concomitantly to the previous one. The hypothesis is based on the fact that externally added riboflavin leads to an increase in the FAD content, as shown in yeast mitochondria (44) and human mitochondria (42,43). In this scenario, increased availability of the FAD ligand would promote its binding, which induces conformational changes which propagate to the overall structure. This may occur by a structure nucleation effect, by which certain motions in the molecule are restrained, reducing its breathing. In the cell context, this effect also leads to a reduction of the sites, which are available to proteolytic degradation by cellular proteases.
As the ETF␤-D128N mutation is not located in the FAD binding domain, our observations provide a structural and molecular rationale to understand the basis of one of the many possible effects of riboflavin supplementation in patients with milder forms of genetic deficiencies of acyl-CoA dehydrogenation, where the mutation sites are distributed all over the structure, suggesting that destabilization of local interaction may lead to long-distance conformational changes that may affect FAD binding. In these patients with mutations that do not totally impair protein folding, but result in a less stable or inactive conformation, the increase in FAD concentration resulting from riboflavin supplementation may enhance the conformational quality of the affected protein, increasing the cellular availability and life time of a biologically active molecule, by a mechanism identical to the one we report in this work. This effect could be particularly relevant under certain pathophysiological conditions like fever but also fasting, infections, pregnancy, and low dietary intake of riboflavin that may cause low cellular FAD content and thereby poses high demands on folding defective flavoproteins that are up-regulated under less favorable folding conditions. We have experimentally shown that flavinylation maintained the properties of the ETF mutant variant after a heat stress condition, which simulated a fever event, a condition which is known to aggravate the symptoms of patients suffering from mild fatty acid ␤-oxidation disorders (6). These observations establish a proof of principle, which can be generalized and tested in many other genetic defects in mitochondrial flavoproteins.