Bacterial Production, Characterization and Protein Modeling of a Novel Monofuctional Isoform of FAD Synthase in Humans: An Emergency Protein?

FAD synthase (FADS, EC 2.7.7.2) is the last essential enzyme involved in the pathway of biosynthesis of Flavin cofactors starting from Riboflavin (Rf). Alternative splicing of the human FLAD1 gene generates different isoforms of the enzyme FAD synthase. Besides the well characterized isoform 1 and 2, other FADS isoforms with different catalytic domains have been detected, which are splice variants. We report the characterization of one of these novel isoforms, a 320 amino acid protein, consisting of the sole C-terminal 3′-phosphoadenosine 5′-phosphosulfate (PAPS) reductase domain (named FADS6). This isoform has been previously detected in Riboflavin-Responsive (RR-MADD) and Non-responsive Multiple Acyl-CoA Dehydrogenase Deficiency (MADD) patients with frameshift mutations of FLAD1 gene. To functionally characterize the hFADS6, it has been over-expressed in Escherichia coli and purified with a yield of 25 mg·L−1 of cell culture. The protein has a monomeric form, it binds FAD and is able to catalyze FAD synthesis (kcat about 2.8 min−1), as well as FAD pyrophosphorolysis in a strictly Mg2+-dependent manner. The synthesis of FAD is inhibited by HgCl2. The enzyme lacks the ability to hydrolyze FAD. It behaves similarly to PAPS. Combining threading and ab-initio strategy a 3D structural model for such isoform has been built. The relevance to human physio-pathology of this FADS isoform is discussed.


Introduction
Riboflavin, or vitamin B2, is the precursor of flavin mononucleotide (FMN) and flavin adenine dinucleotide (FAD), which are essential cofactors of hundreds of oxidoreductases (EC1) but also of other flavoproteins and flavoenzymes which are not oxidoreductases. These enzymes are involved in In the investigation accounted for in this report, we have over-produced in E. coli and purified with a yield sufficient to allow functional characterization and structure/function relationships studies the novel putative natural isoform starting at Met268 and named hFADS6. This isoform is able to perform both Mg 2+ -dependent FAD synthesis and the reverse reaction (i.e., FAD pyrophosphorolysis), but not FAD hydrolysis, the latter finding correlating with the lack of the molybdopterin binding domain in the novel variant ( Figure 1).
Molecules 2018, 23,116 3 of 16 In the investigation accounted for in this report, we have over-produced in E. coli and purified with a yield sufficient to allow functional characterization and structure/function relationships studies the novel putative natural isoform starting at Met268 and named hFADS6. This isoform is able to perform both Mg 2+ -dependent FAD synthesis and the reverse reaction (i.e., FAD pyrophosphorolysis), but not FAD hydrolysis, the latter finding correlating with the lack of the molybdopterin binding domain in the novel variant ( Figure 1).

Figure 1.
Schematic representation of the human FAD synthase protein isoforms. The protein length and the enzymatic activities of six human FAD synthase isoforms are reported. Molybdopterin binding-like domain, responsible of FAD hydrolysis, is indicated in dark grey (N-terminus); phosphoadenylyl sulfate reductase domain (PAPS), competent in FAD synthesis and in pyrophosphorolysis, is highlighted in light grey (C-terminus). hFADS1, hFADS2 and hFADS3 possess both catalytic domains; hFADS4 and hFADS5 present the sole molybdopterin binding domain; on the contrary, hFADS6 has the sole PAPS reductase domain. Numbers indicate amino acid positions in the resulting protein sequence of the FADS1 isoform.

Cloning, Expression and Purification of the hFADS6 Isoform
The cDNA coding for human FADS isoform 6 (hFADS-6) was cloned between EcoRI and XhoI restriction sites of the pH6EX3 expression vector, generating a fusion peptide with a N-terminal 6-His-tag, whose theoretical molecular mass is 38.266 kDa. To optimize the heterologous production of the protein, several E. coli strains were screened for transformation, i.e., BL21(DE3)pLysS, JM109, RosettaGami2(DE3)pLysS, Rosetta(DE3)pLysS. E. coli Rosetta(DE3)pLysS strain, that is supplemented with transfer RNA (tRNA)-specific for rare codons, emerged as the best choice in terms of production of the recombinant protein ( Figure 2). Molybdopterin binding-like domain, responsible of FAD hydrolysis, is indicated in dark grey (N-terminus); phosphoadenylyl sulfate reductase domain (PAPS), competent in FAD synthesis and in pyrophosphorolysis, is highlighted in light grey (C-terminus). hFADS1, hFADS2 and hFADS3 possess both catalytic domains; hFADS4 and hFADS5 present the sole molybdopterin binding domain; on the contrary, hFADS6 has the sole PAPS reductase domain. Numbers indicate amino acid positions in the resulting protein sequence of the FADS1 isoform.

Cloning, Expression and Purification of the hFADS6 Isoform
The cDNA coding for human FADS isoform 6 (hFADS-6) was cloned between EcoRI and XhoI restriction sites of the pH6EX3 expression vector, generating a fusion peptide with a N-terminal 6-His-tag, whose theoretical molecular mass is 38.266 kDa. To optimize the heterologous production of the protein, several E. coli strains were screened for transformation, i.e., BL21(DE3)pLysS, JM109, RosettaGami2(DE3)pLysS, Rosetta(DE3)pLysS. E. coli Rosetta(DE3)pLysS strain, that is supplemented with transfer RNA (tRNA)-specific for rare codons, emerged as the best choice in terms of production of the recombinant protein ( Figure 2). Indeed, a band at about 36.0 kDa was present in the induced cell lysate after 2 h induction ( Figure 2A lane 2), which was absent in non-induced cell lysate and supernatant (Figure 2A lanes 1  and 3, respectively). Different growth conditions and Isopropyl β-D-1-thiogalactopyranoside (IPTG) concentrations were tested to improve expression; IPTG concentration from 0.1 mM to 1 mM did not influence protein expression (not shown). The cell lysates were separated in a soluble and insoluble cell fraction as described in Materials and Methods. The amount of expressed protein recovered in the soluble fraction increased with time of growth (Figure 2A lanes 4-6). However, after 4 h, which was the best condition ( Figure 2A, lane 5), protein degradation was also observed (Figure 2A lane 6), which was not prevented by lowering the temperature to 20 • C (Figure 2A, lane 7). A lower amount of protein was recovered in the cell pellet after 4 h induction ( Figure 2A, lane 8), accounting for about 30% of the protein in the soluble fraction. The protein was identified by immunoblot performed with both the in-house produced anti-FADSs antibody (Figure 2B lane 1 and 5 ) and the anti-His antibody as a control ( Figure 2C, lanes 1 and 5 ). In both cases a strong immunoreaction was observed in the induced cell lysate, which was virtually absent in the non-induced lysates. The faint immunoreaction at higher molecular mass might be due to aggregation of a very small fraction of the protein. To purify the protein of interest, the soluble fraction, corresponding to lane 5 of Figure 2 and to lane 1 of Figure 3, was applied on a Ni 2+ chelating column and eluted as described in Materials and Methods. The column was then washed with 50 mM imidazole ( Figure 3A lane 3); the elution of the 6His-hFADS6 was started with 100 mM imidazole ( Figure 3A lanes 4 and 5) and was completed by increasing the imidazole concentration to 250 mM. The purified protein resulted in an apparently homogenous band ( Figure 3A, lanes 6 and 7). The apparent molecular mass on SDS-PAGE was about 36 kDa, very close to the theoretical molecular mass of the 6His-hFADS6. The protein yield was about 25 mg·L −1 of cell culture.
positions in the resulting protein sequence of the FADS1 isoform.

Cloning, Expression and Purification of the hFADS6 Isoform
The cDNA coding for human FADS isoform 6 (hFADS-6) was cloned between EcoRI and XhoI restriction sites of the pH6EX3 expression vector, generating a fusion peptide with a N-terminal 6-His-tag, whose theoretical molecular mass is 38.266 kDa. To optimize the heterologous production of the protein, several E. coli strains were screened for transformation, i.e., BL21(DE3)pLysS, JM109, RosettaGami2(DE3)pLysS, Rosetta(DE3)pLysS. E. coli Rosetta(DE3)pLysS strain, that is supplemented with transfer RNA (tRNA)-specific for rare codons, emerged as the best choice in terms of production of the recombinant protein ( Figure 2). Spectrophotometric analysis of purified fraction 5 ( Figure 3B) as well as of fractions 4, 6 and 7 (not shown) showed a typical flavoprotein absorbance spectrum, with a main peak at 274 nm and two minor peaks at 350 and 450 nm ( Figure 3B) indicating that this novel purified isoform, as the previously characterized isoform 2, tightly binds flavins [25]. High performance liquid chromatography (HPLC) analysis of the supernatant obtained by acidic treatment of purified hFADS6 demonstrates that the bound cofactor is FAD ( Figure 3C). The molecular ratio of FAD/monomer as determined by optical properties as well as by HPLC (see Materials and Methods) gave a value of 0.60 ± 0.015.

Molecular Characterization of hFADS6
The aggregation state of hFADS6 and of hFADS2 was assessed by size exclusion chromatography and hydrodynamic radius measurement performed by tandem HPLC-GPC/LS. As shown in Figure 4A, hFADS6 migrated in HPLC-GPC as a single peak with an elution volume of 13 mL, much higher than what observed for hFADS2 (elution volume, 11 mL). The addition of 1 mM DTT to the elution buffer did not alter the chromatographic behavior of either protein, if not for improving the sharpness of individual peaks, likely because of the dissociation of minor amounts of disulfide linked aggregates. Online MALS (Multi-Angle Light Scattering) measurements performed on the peaks obtained in the presence of DTT gave molecular masses of 36.5 ± 5 kDa for hFADS6 and 111 ± 17 kDa for hFADS2, confirming that the hFADS6 and the hFADS2 had a monomeric and dimeric nature, respectively.
hFADS6 has a CD signal originating from alpha-helical regions in secondary structure, and its CD spectrum nearly overlaps that of hFADS2 ( Figure 4B), if not for minor differences in signal intensity, that could be related to the different length of the two isoforms. Within the limitations due to incomplete coverage of the far ultra-violet (UV) region and to the approximations inherent to these models, the alpha helical content in hFADS6 was estimated at 39%, comparable to 32% in hFADS2. The presence and thermal stability of secondary structure elements was assessed on either

Molecular Characterization of hFADS6
The aggregation state of hFADS6 and of hFADS2 was assessed by size exclusion chromatography and hydrodynamic radius measurement performed by tandem HPLC-GPC/LS. As shown in Figure 4A, hFADS6 migrated in HPLC-GPC as a single peak with an elution volume of 13 mL, much higher than what observed for hFADS2 (elution volume, 11 mL). The addition of 1 mM DTT to the elution buffer did not alter the chromatographic behavior of either protein, if not for improving the sharpness of individual peaks, likely because of the dissociation of minor amounts of disulfide linked aggregates. Online MALS (Multi-Angle Light Scattering) measurements performed on the peaks obtained in the presence of DTT gave molecular masses of 36.5 ± 5 kDa for hFADS6 and 111 ± 17 kDa for hFADS2, confirming that the hFADS6 and the hFADS2 had a monomeric and dimeric nature, respectively.
hFADS6 has a CD signal originating from alpha-helical regions in secondary structure, and its CD spectrum nearly overlaps that of hFADS2 ( Figure 4B), if not for minor differences in signal intensity, that could be related to the different length of the two isoforms. Within the limitations due to incomplete coverage of the far ultra-violet (UV) region and to the approximations inherent to these models, the alpha helical content in hFADS6 was estimated at 39%, comparable to 32% in hFADS2. The presence and thermal stability of secondary structure elements was assessed on either protein through temperature ramp experiments carried out by progressively heating protein solutions from 20 • C to 95 • C at 1 • C·min −1 . In these experiments, both proteins gave a well-defined transition between the structured and the denatured state, confirming that even the shorter protein construct in hFADS6 has a defined structure. Tm values for the transition to denatured state in hFADS6 was 51.6 • C, slightly lower than the Tm measured for hFADS2 (53.4 • C). Whether the lower thermal stability of hFADS6 with respect to hFADS2 depends on the absence of structure-stabilizing features in the shorter isoform or on its monomeric state remains to be ascertained.
Molecules 2018, 23, 116 6 of 16 protein through temperature ramp experiments carried out by progressively heating protein solutions from 20 °C to 95 °C at 1 °C·min −1 . In these experiments, both proteins gave a well-defined transition between the structured and the denatured state, confirming that even the shorter protein construct in hFADS6 has a defined structure. Tm values for the transition to denatured state in hFADS6 was 51.6 °C, slightly lower than the Tm measured for hFADS2 (53.4 °C). Whether the lower thermal stability of hFADS6 with respect to hFADS2 depends on the absence of structure-stabilizing features in the shorter isoform or on its monomeric state remains to be ascertained.

Functional Characterization of hFADS6 Recombinant Protein
To ascertain if the over-expressed recombinant hFADS6 protein was able to synthesize FAD, the purified protein was incubated with the substrates, in the absence or presence of Mg 2+ , which is Both FADS isoforms show a cooperative folding/unfolding transition. Data were processed to calculate the midpoint temperature (T m ) for the native/denatured transition (53.4 • C for hFADS2 vs. 51.6 • C for hFADS6).

Functional Characterization of hFADS6 Recombinant Protein
To ascertain if the over-expressed recombinant hFADS6 protein was able to synthesize FAD, the purified protein was incubated with the substrates, in the absence or presence of Mg 2+ , which is specifically required by other human FADS isoforms [21,22]. The FAD synthesis reaction ( Figure 5A) was started by the addition of 0.13 µM purified hFADS6 to ATP (100 µM) and FMN (2 µM). An evident fluorescence decrease (λ excitation at 450 nm, λ emission at 520 nm) was observed, indicating the conversion of the highly fluorescent FMN into the less fluorescent FAD. The initial rate of FAD synthesis in the cuvette, calculated as described in Materials and Methods, was about 0.49 nmol·min −1 (corresponding to 2.3 min −1 in this preparation, 2.7 ± 0.13 min −1 as the mean of three different protein preparations.) As expected for FADSs, in the absence of Mg 2+ , nearly no fluorescence decrease was observed [21,22].
The reverse reaction, i.e., pyrophosphorolysis, was also tested, as well as the ability to cleave FAD by hydrolysis, as previously reported for the bi-functional isoform 2 [28]. As expected on the basis of the lack of the MPTb domain, isoform 6 of the enzyme was able to catalyze the pyrophosphorolysis, but not the hydrolysis ( Figure 6A); not even the presence of Co 2+ could promote the ability of hFADS6 to catalyze hydrolysis, at a difference from the bi-functional variant. A control was also made by using the -SH reagent mersalyl. Mersalyl induces FAD hydrolysis in the hFADS2 isoform, but was found inactive in this regard with the hFADS6 variant. The kinetics of the reverse reaction was measured as a function of NaPPi concentration ( Figure 6B). A KmPPI of 0.042 ± 0.006 mM was calculated by interpolating the data in the Michaelis-Menten equation. The KmFAD could not be measured due to instrumental limitations that do not allow measurements of Km values lower than 0.1 µM as in the case hFADS6, similarly to hFADS2 [25]. A relevant feature of hFADS2 is the presence of redox reactive cysteines, some of which implicated in the catalytic activity [26]. To directly evaluate if some Cys residues of hFADS6 play a role in enzyme function, the effect of HgCl 2 was tested. Addition of HgCl 2 inhibited the FAD synthesis reaction, in agreement with previous data described for the truncated form of FADS missing the N-terminal domain [26]. This inhibitory effect corroborates the proposal that free Cys residues located in the PAPS domain participate in the catalytic cycle of the hFADSs.
Kinetic analyses of FAD synthesis as function of MgCl 2 ( Figure 5B), FMN ( Figure 5C) or adenosine triphosphate (ATP) ( Figure 5D) concentration were performed. A hyperbolic dependence was found for FAD synthetic rate on Mg 2+ concentration, with 50% of maximum activity (Mg 2+ 50 ) at 0.15 mM ± 0.02 mM, in line with hFADS2 values [25]. Data fitting according to the Michaelis-Menten equation using the Grafit program (see Material and methods) gave apparent values of: K m FMN 0.13 ± 0.01 µM and K m ATP 6.91 ± 0.51 µM (data from 3 independent experiments). Overlapping results were obtained by interpolating the same data in the Lineweaver-Burk equation using the Grafit program (not shown). These experiments gave k cat values of 2.8 and 3.0 min −1 , respectively. In three different experiments a mean value of 2.9 ± 0.05 min −1 was evaluated.
The reverse reaction, i.e., pyrophosphorolysis, was also tested, as well as the ability to cleave FAD by hydrolysis, as previously reported for the bi-functional isoform 2 [28]. As expected on the basis of the lack of the MPTb domain, isoform 6 of the enzyme was able to catalyze the pyrophosphorolysis, but not the hydrolysis ( Figure 6A); not even the presence of Co 2+ could promote the ability of hFADS6 to catalyze hydrolysis, at a difference from the bi-functional variant. A control was also made by using the -SH reagent mersalyl. Mersalyl induces FAD hydrolysis in the hFADS2 isoform, but was found inactive in this regard with the hFADS6 variant. The kinetics of the reverse reaction was measured as a function of NaPPi concentration ( Figure 6B). A K m PPI of 0.042 ± 0.006 mM was calculated by interpolating the data in the Michaelis-Menten equation. The K m FAD could not be measured due to instrumental limitations that do not allow measurements of K m values lower than 0.1 µM as in the case hFADS6, similarly to hFADS2 [25].

3D Model of hFADS6
Structural models of hFADS6 apo-protein were computed with three different programs, using homology and/or threading procedures. The C-terminus part of hFADS6 (PAPS domain) is homologous to already experimentally characterized proteins, with identity/similarity levels adequate for successfully producing homology models with an accurate architecture: for instance, query primary structure (aa 109-320) has a 36% similarity with Saccharomyces cerevisiae YDL045c FAD synthetase. Accordingly, the output from the three different modeling approaches was very . The FAD cleavage reaction (i.e., pyrophosporolysis) was started by the addition of the 6-His-hFADS6 protein (0.13 µM) and was measured by the initial rate of fluorescence increase (λ excitation = 450 nm, λ emission = 520 nm). FAD hydrolysis was measured under the same conditions, if not for omitting NaPPi and MgCl 2 , and in the presence of either 1 mM CoCl 2 (black line) or 1 mM CoCl 2 plus 0.5 mM Mersalyl (dark grey); (B) NaPPi concentration dependence. FAD cleavage (i.e., pyrophosporolysis) rate, catalyzed by purified 6-His-hFADS6 (0.13 µM), was measured fluorimetrically at 37 • C in 2 mL of 50 mM Tris/HCl pH 7.5, in the presence of 5 mM MgCl 2 , 0.5 µM FAD, and of the given NaPPi concentrations.

3D Model of hFADS6
Structural models of hFADS6 apo-protein were computed with three different programs, using homology and/or threading procedures. The C-terminus part of hFADS6 (PAPS domain) is homologous to already experimentally characterized proteins, with identity/similarity levels adequate for successfully producing homology models with an accurate architecture: for instance, query primary structure (aa 109-320) has a 36% similarity with Saccharomyces cerevisiae YDL045c FAD synthetase. Accordingly, the output from the three different modeling approaches was very similar when predicting by homology modeling the organization of the C-terminus part of hFADS6. On the contrary, the produced models largely diverged in the arrangement of the N-terminus part, for which no homologous templates can be identified. For this reason, Swiss Model, which is a software exclusively based on homology modeling, failed in predicting 3D structure of the hFADS6 N-terminus. Robetta or I-TASSER, which also implement ab initio modeling, produced dissimilar 3D structures for this part. The output provided by Robetta was selected as final apo-hFADS6 model, for its better geometrical features. FAD ligand was transferred, after superposition with the crystallographic structure of the Candida glabrata FMN Adenylyltransferase (PDB ID: 3g6k) for generating the holo-hFADS6 form (Figure 7, PAPS domain in white, N-terminus part in red). In more general terms, the predicted model for hFADS6 had an alpha-helical content of 37% (117 residues out of 320), close to that calculated from CD spectra (39%). The content of beta-strands cannot be determined by the limited coverage of our CD spectra, but it was estimated at 10% in the model (32 residues out of 320).

Discussion
The FADS6 isoform was very recently discovered, by transcriptome analysis, in patients harboring frameshift mutations of FLAD1 gene triggering RR-MADD. This FADS6 isoform can be anyway produced from the altered gene since it starts from Met268 whose codon is downstream the frameshift ( [30], and Figure 1). However, it was not known if this isoform could display some functions related to FAD production. We have demonstrated in this work that this shorter isoform is able to synthesize FAD. Therefore, its importance, even though expressed in small amount, resides in human pathophysiology, that is, to synthesize sufficient amount of FAD, allowing survival of patients affected by disruptive mutations of the FLAD1 gene. hFADS6 harbors additional 61 N-terminus amino acid residues with respect to the artificial PAPS mutant. Hence, it could then be considered the natural counterpart of the artificial protein previously produced by recombinant DNA strategies [29]. Therefore, it was very important to ascertain the capacity of the novel hFADS6 isoform to synthesized FAD. FADS6 displayed FAD synthesizing activity, but not FAD hydrolase activity, which resides in the N-terminal domain of the full length FADS1 or 2 [28]. Moreover, FADS6 showed some similar features to the other previously characterized longer FADS isoforms,

Discussion
The FADS6 isoform was very recently discovered, by transcriptome analysis, in patients harboring frameshift mutations of FLAD1 gene triggering RR-MADD. This FADS6 isoform can be anyway produced from the altered gene since it starts from Met268 whose codon is downstream the frameshift ( [30], and Figure 1). However, it was not known if this isoform could display some functions related to FAD production. We have demonstrated in this work that this shorter isoform is able to synthesize FAD. Therefore, its importance, even though expressed in small amount, resides in human pathophysiology, that is, to synthesize sufficient amount of FAD, allowing survival of patients affected by disruptive mutations of the FLAD1 gene. hFADS6 harbors additional 61 N-terminus amino acid residues with respect to the artificial PAPS mutant. Hence, it could then be considered the natural counterpart of the artificial protein previously produced by recombinant DNA strategies [29]. Therefore, it was very important to ascertain the capacity of the novel hFADS6 isoform to synthesized FAD. FADS6 displayed FAD synthesizing activity, but not FAD hydrolase activity, which resides in the N-terminal domain of the full length FADS1 or 2 [28]. Moreover, FADS6 showed some similar features to the other previously characterized longer FADS isoforms, such as Mg 2+ dependence and strong inhibition by Hg 2+ , confirming that some of the Cys residues highlighted in the model (Figure 7) play crucial roles in this PAPS domain [26]. The FAD synthesizing activity is about 70% with respect to that measured for hFADS2 [25]; however, the catalytic efficiency (k cat /K m ) results higher than that of hFADS2 due to the lower Km values of hFADS6 for both FMN and ATP compared to those of hFADS2. The data are in agreement with those found for PAPS [29]. Concerning the Mg 2+ binding, we expect a similar behavior as in C. glabrata FMN Adenylyltransferase, i.e., Mg 2+ makes a bridge between bound ATP and the protein binding site. In fact, hFADS6 harbors two residues, D141 and D225, which are homologous to those responsible for Mg 2+ binding in C. glabrata enzyme (D66 and D168) [34].
From a structural standpoint, it is noteworthy that the short and monomeric hFADS6 appears to be appropriately folded, with a thermal stability of its folding remarkably close to that of the much larger (and dimeric) hFADS2. The presence of DTT has effect neither on the aggregation state nor on the hydrodynamic radius of the two proteins, confirming the absence of structurally relevant disulfide bonds in the structure.
In agreement with the high sequence identity between the two primary structures, the models for the 3D structure of the C-terminus part of hFADS6 and of the PAPS mutant previously produced through recombinant DNA are similar to one another (Figure 7, white structure), except for the N-terminal 61 amino acids of the hFADS6 (Figure 7, red structure). Besides the overall close similarity of the fold, even the positions of the Cys residues in the two structures mostly overlap [29]. Thanks to the availability of a number of templates with similarity with the target higher than 30%, the prediction of the general architecture of this part of hFADS6 can be considered accurate. Conversely, lack of adequate templates forces to rely on threading and ab initio modeling for the N-terminus part of hFADS6, accepting a lower confidence of the outcome. Indeed, the results provided by Robetta and I-TASSER for the N-terminus of the hFADS6 diverge for their overall orientation with respect to the PAPS domain. In neither case, however, the access to the catalytic site is hampered by the N-terminus part of the protein. Similarity of the 3D arrangement of these two isoforms in the C-terminus part is the structural basis for a conserved catalytic activity. Indeed, docking of FAD on hFADS6 ( Figure 7A) led to similar results with respect to the PAPS [29].
Despite the hFADS6 synthase function overlaps that of hFADS2, clear structural differences were observed between the two proteins. Indeed, differently from the dimeric structure of hFADS2, hFADS6 had an unequivocal monomeric form. Since the major difference in hFADS6 with respect to hFADS2 is the lack of the molybdopterin binding domain, this difference could be the cause of the absence of the dimeric structure. In other words, in hFADS2, the molybdopterin binding domain should play some role in the interaction between the two monomers. Indeed, previous studies based on site-directed mutagenesis addressed the relationships between the molybdopterin binding and the PAPS domains [26]. Further conditioning of the structure determined by the redox state (SH/S-S balance) of the protein is likely to occur; a crucial role for some Cys residues in the PAPS domain of the hFADS2 was assessed by site-directed mutagenesis. This issue is under investigation for hFADS6 in comparison with the hFADS2 isoform.
The first evidence of the expression of hFADS6 isoform in human tissues was provided in 2016 [30]. However, a clear correlation on expression level in healthy and pathological conditions is still lacking. Another interesting question concerns the putative localization of this novel isoform. Since this hFADS6 lacks the mitochondrial targeting peptide [27] a sole cytosolic localization is expected. This point is particularly relevant for cellular bioenergetics. Indeed, the relative activity of mitochondrial vs. cytosolic isoforms of FADS might regulate the transport directionality of the mitochondrial FAD transporter [35]. Therefore, under the conditions in which frameshift mutations of FLAD1 gene alter the relationships among splice variants, this novel emergency protein becomes relevant and we expect that flavoprotein biogenesis in mitochondria is ensured by FAD import. Experiments are on the way to directly prove these assumptions in the frame of measuring the relative abundance of the natural hFADS6 in different tissues.
In conclusion, FADS6 represents a previously neglected FADS isoform whose role was highlighted by the discovery of FADS gene natural mutations causative of RR-MADD [30]. Very importantly, this isoform may represent an emergency protein, which can be produced also in patients from an altered gene. In this respect, this protein represents a target for therapy intervention in patients harboring FADS defects, especially for those in which Rf therapy is unsuccessful.

Materials
All chemicals and monoclonal anti-polyhistidine-peroxidase antibody (A7058) were from Sigma-Aldrich (Saint Louis, MO, USA), if not otherwise specified. E. coli Rosetta(DE3)pLysS strain was purchased from Novagen (Madison, WI, USA). Restriction endonucleases and other cloning reagents were purchased from Fermentas (Burlington, ON, Canada); MegaMan Human Transcriptome Library was from Agilent Technologies (Santa Clara, CA, USA). Anti-hFADSs antibody (patent number WO2009107158 A1) was in-house produced.

Cloning of cDNA Coding for hFADS6
The 963 bp cDNA coding for the hFADS6 isoform [30] was amplified from the Mega Man Human Transcriptome Library with the forward and reverse primers 5 -CCGGAATTCTATGAAG GGACTATTCCAAAACCCAG-3 and 5 -CCGCTCGAGCTACCCCTGCTGTCCTGGGAAGGGG-3 , containing the EcoRI and XhoI sites, respectively. The amplified cDNA was then cloned in the EcoRI/XhoI sites of the pH6EX3 expression vector. The resulting recombinant plasmid, defined as pH6EX3-hFADS6, encodes a 6His-tagged fusion protein corresponding to hFADS6 carrying the extra N-terminal sequence MSPIHHHHHHLVPRGSEASNS.

Expression of Recombinant hFADS6 Protein in E. coli
The expression host E. coli Rosetta(DE3)pLysS strain was transformed with the pH6EX3-hFADS6 construct by calcium chloride treatment. Selection of positive colonies was performed on LB-agar plates containing 34 µg·mL −1 chloramphenicol and 100 µg·mL −1 ampicillin. E. coli Rosetta(DE3)pLysS cells carrying the recombinant plasmid were inoculated in 10 mL of LB medium (1% Bacto Peptone, 0.5% Bacto Yeast extract, 1% NaCl, pH 7.0) supplemented with 100 µg·mL −1 ampicillin and 34 µg·mL −1 chloramphenicol, and cultured overnight at 37 • C with rotary shaking (160 rpm). A 5 mL aliquot of the cell culture was transferred to 0.5 L of fresh LB medium supplemented with 100 µg·mL −1 ampicillin and 34 µg·mL −1 chloramphenicol and grown at 37 • C to A 600 equal to 0.6-0.8. Then, 0.4 mM IPTG was added to induce the expression of the recombinant protein, 6-His-hFADS6. Growth was continued for 2 to 8 h at 28 • C, or overnight at 20 • C. Bacteria were harvested by centrifugation at 3000×g for 10 min at 4 • C and the pellets stored at −20 • C. The bacterial pellet (about 3 g wet weight) was thawed on ice for 15 min and resuspended in 30 mL of start buffer (500 mM NaCl, 40 mM Hepes/Na, pH 7.4) supplemented with Protease Inhibitor Cocktail (P8849, Sigma-Aldrich, 1 mL/20 g of cells wet weight) and 0.5 mM Phenylmethylsulfonyl fluoride (PMSF). Cells were disrupted by mild sonication at 4 • C (one cycle of 10 min and one cycle of 5 min with 1 s Pulse ON and 1 s Pulse OFF, at 40 W) using a VCX130 PB sonifier (Branson, Danbury, CT, USA). The soluble and the insoluble cell fractions were separated by centrifugation of the cell lysate at 20,000×g for 30 min at 4 • C. The pellet (inclusion bodies and cell debris), containing the insoluble over-expressed proteins, was re-suspended in 15 mL of start buffer, aliquoted and used for SDS-PAGE analysis. The supernatant, containing the soluble over-expressed 6-His-hFADS6, was used for SDS-PAGE analysis, FADS activity assay, and further protein purification (see below).

Purification of Recombinant 6-His-hFADS6
A 40 mL aliquot of the soluble cell fraction, obtained as outlined above, was applied onto a Chelating Sepharose Fast Flow column (3.5 mL packed resin), previously treated with 200 mM NiSO 4 according to the producer's protocol and equilibrated with the start buffer. The column was first washed with 35 mL start buffer, then eluted with a step gradient of 50, 100, 250, and 500 mmol·L −1 imidazole in the same buffer. Prior to storing or further processing, fractions containing the purified recombinant protein were desalted by gel filtration on a PD10-column in 40 mM Hepes/Na, 5 mM β-mercaptoethanol, pH 7.4. These protein samples were stable for at least 30 days at 4 • C.

Protein Concentration and FAD/Protein Monomer Ratio Measurements
Protein concentration was measured with the method of Bradford, using BSA as standard [36]. In an alternative procedure, protein concentration of the purified 6-His-hFADS6 was estimated by absorbance spectra, which were recorded on an Ultrospec 3100 pro spectrophotometer (Amersham Biosciences, Piscataway, NJ, USA), essentially as in [25]. To this aim, the contribution of the bound FAD had to be subtracted from the A 280 readings. Because A 280 for both free FAD and hFADS6-bound FAD is 1.7-fold A 450 , the A 280 , actually due to the apo-protein, may be calculated from the Equation: A 280 apo-enzyme = A 280 − (A 450 × 1.7) The protein concentration was then estimated by using ε 280 (47.705 mM −11 ·cm −1 , 1.247 mg/mL −1 ·cm −1 ), as calculated from the protein sequence by using the ExpasyProtParam tool (Swiss Institute of Bioinformatics, Lausanne, Switzerland). Measurements made by either the spectrophotometric or the Bradford method differed by no more than 7%. The FAD/protein monomer ratio (given as percent flavinylation, Fl%) could be estimated from the absorbance spectrum by considering:

FAD Content as Assayed by HPLC Analysis
The purified 6-His-hFADS6 (15.4 µg in 100 µL, 4 µM) was treated with 10% perchloric acid to remove tightly but not covalently bound molecules. The supernatant was neutralized and aliquots were analyzed for flavin content by HPLC, as previously reported [25,37]. Standard aqueous of FAD, FMN and riboflavin were used to identify the protein-bound flavin cofactor.

Circular Dichroism Measurements
Circular dichroism (CD) spectra were recorded at 20 • C in 0.1 cm path length cells on 0.2 mg·mL −1 protein solutions in 40 mM Hepes/Na, 150 mM NaCl, 1 mM DTT, pH 7.4 by means of a J810 spectropolarimeter (Jasco Europe SrL, Cremella, Italy), and were analyzed by means of the J800 software (Jasco Europe SrL, Cremella, Italy). Similar conditions were used for temperature ramp experiments. Temperature was increased from 20 to 95 • C at a heating rate of 1 • C·min −1 by means of a computer-controlled Peltier-driven cell holder, while continuously monitoring the loss of CD signal at 220 nm. CD spectra were normalized in terms of Mean Residual Ellipticity ([ Molecules 2018, 23,116 FAD had to be subtracted from the A280 readings. Because A280 for both free FAD and hFADS6-FAD is 1.7-fold A450, the A280, actually due to the apo-protein, may be calculated from the Equ A280 apo-enzyme = A280 − (A450 × 1.7) The protein concentration was then estimated by using ε280 (47.705 mM −11 ·cm −1 , mg/mL −1 ·cm −1 ), as calculated from the protein sequence by using the ExpasyProtParam tool Institute of Bioinformatics, Lausanne, Switzerland). Measurements made by eithe spectrophotometric or the Bradford method differed by no more than 7%. The FAD/p monomer ratio (given as percent flavinylation, Fl%) could be estimated from the abso spectrum by considering: where 0.23 is the ε450 (FAD)/ε280 (apo-enzyme) ratio.

FAD Content as Assayed by HPLC Analysis
The purified 6-His-hFADS6 (15.4 µ g in 100 µ L, 4 µ M) was treated with 10% perchloric remove tightly but not covalently bound molecules. The supernatant was neutralized and a were analyzed for flavin content by HPLC, as previously reported [25,37]. Standard aque FAD, FMN and riboflavin were used to identify the protein-bound flavin cofactor.

Circular Dichroism Measurements
Circular dichroism (CD) spectra were recorded at 20 °C in 0.1 cm path length cells mg·mL −1 protein solutions in 40 mM Hepes/Na, 150 mM NaCl, 1 mM DTT, pH 7.4 by means o spectropolarimeter (Jasco Europe SrL, Cremella, Italy), and were analyzed by means of th software (Jasco Europe SrL, Cremella, Italy). Similar conditions were used for temperature experiments. Temperature was increased from 20 to 95 °C at a heating rate of 1 °C·min −1 by m a computer-controlled Peltier-driven cell holder, while continuously monitoring the loss signal at 220 nm. CD spectra were normalized in terms of Mean Residual Ellipticity ([Ө]MRE and the alpha-helical content was estimated from the [Ө]MRE at 222 nm according to Chen and [39].

Molecular Parameters
The molecular size of the purified 6-His-hFADS6 was assessed by HPLC-GPC, u Superose 12 column (10/300 GL, GE Healthcare, Milan Italy), fitted to a HPLC system (mo equipped with a tunable dual wavelength detector (Waters 2487, Milford, MA, USA) and a ref index detector (Optilab ® T-rEX, Wyatt, Santa Barbara, CA, USA). Columns were run at 0.5 mL in 40 mM Hepes/Na, 150 mM NaCl, pH 7.4, containing 1 mM DTT when indicated, and cali with suitable protein standards that encompassed the molecular size predicted for hFADS hydrodynamic properties and the molar mass of the purified 6-His-hFADS6 were assessed by ] MRE , [38]), and the alpha-helical content was estimated from the [ A280 apo-enzyme = A280 − (A450 × 1.7) The protein concentration was then estimated by using ε280 (47.705 mM −11 ·cm −1 , 1.247 mg/mL −1 ·cm −1 ), as calculated from the protein sequence by using the ExpasyProtParam tool (Swiss Institute of Bioinformatics, Lausanne, Switzerland). Measurements made by either the spectrophotometric or the Bradford method differed by no more than 7%. The FAD/protein monomer ratio (given as percent flavinylation, Fl%) could be estimated from the absorbance spectrum by considering: Fl% = [(A450/A280apo-enzyme)/0.23] × 100 where 0.23 is the ε450 (FAD)/ε280 (apo-enzyme) ratio.

FAD Content as Assayed by HPLC Analysis
The purified 6-His-hFADS6 (15.4 µ g in 100 µ L, 4 µ M) was treated with 10% perchloric acid to remove tightly but not covalently bound molecules. The supernatant was neutralized and aliquots were analyzed for flavin content by HPLC, as previously reported [25,37]. Standard aqueous of FAD, FMN and riboflavin were used to identify the protein-bound flavin cofactor.

Circular Dichroism Measurements
Circular dichroism (CD) spectra were recorded at 20 °C in 0.1 cm path length cells on 0.2 mg·mL −1 protein solutions in 40 mM Hepes/Na, 150 mM NaCl, 1 mM DTT, pH 7.4 by means of a J810 spectropolarimeter (Jasco Europe SrL, Cremella, Italy), and were analyzed by means of the J800 software (Jasco Europe SrL, Cremella, Italy). Similar conditions were used for temperature ramp experiments. Temperature was increased from 20 to 95 °C at a heating rate of 1 °C·min −1 by means of a computer-controlled Peltier-driven cell holder, while continuously monitoring the loss of CD signal at 220 nm. CD spectra were normalized in terms of Mean Residual Ellipticity ([Ө]MRE, [38]), and the alpha-helical content was estimated from the [Ө]MRE at 222 nm according to Chen and Yang [39].

Molecular Parameters
The molecular size of the purified 6-His-hFADS6 was assessed by HPLC-GPC, using a Superose 12 column (10/300 GL, GE Healthcare, Milan Italy), fitted to a HPLC system (mod. 515) equipped with a tunable dual wavelength detector (Waters 2487, Milford, MA, USA) and a refractive index detector (Optilab ® T-rEX, Wyatt, Santa Barbara, CA, USA). Columns were run at 0.5 mL·min −1 in 40 mM Hepes/Na, 150 mM NaCl, pH 7.4, containing 1 mM DTT when indicated, and calibrated with suitable protein standards that encompassed the molecular size predicted for hFADS2. The ] MRE at 222 nm according to Chen and Yang [39].

Molecular Parameters
The molecular size of the purified 6-His-hFADS6 was assessed by HPLC-GPC, using a Superose 12 column (10/300 GL, GE Healthcare, Milan Italy), fitted to a HPLC system (mod. 515) equipped with a tunable dual wavelength detector (Waters 2487, Milford, MA, USA) and a refractive index detector (Optilab ® T-rEX, Wyatt, Santa Barbara, CA, USA). Columns were run at 0.5 mL·min −1 in 40 mM Hepes/Na, 150 mM NaCl, pH 7.4, containing 1 mM DTT when indicated, and calibrated with suitable protein standards that encompassed the molecular size predicted for hFADS2. The hydrodynamic properties and the molar mass of the purified 6-His-hFADS6 were assessed by Multi Angle Light Scattering (MALS), using a Wyatt Dawn ® Heleos ® system connected in series to the aforementioned HPLC-GPC chromatographic system. Molar mass at different elution times was calculated by means of the Astra software (Wyatt, Santa Barbara, CA, USA), using a dn/dc value of 0.185.

Measurements of FAD Synthesis and FAD Pyrophosphorolysis Rate
The rate of FAD synthesis and FAD cleavage were measured as in [25], by exploiting the different fluorescence properties of FAD with respect to FMN. Fluorescence time courses (λ excitation at 450 nm; λ emission at 520 nm) were followed at 37 • C in a FP-3800 spectrofluorometer (Jasco, Easton, MD, USA). In each experiment, FAD and FMN fluorescence were calibrated by using standard solutions. The rate of FAD synthesis, expressed as nmol FAD min −1 ·(mg protein) −1 , was calculated from the rate of fluorescence decrease, measured as the tangent to the initial part of the experimental curve, as described in detail in [25]. For activity measurements, purified protein fractions (5-10 µg, 0.13-0.26 nmol protein as monomer, unless otherwise indicated) were incubated at 37 • C, in 50 mM Tris/HCl, pH 7.5, containing 5 mM MgCl 2 , 2 µM FMN, 100 µM ATP, and additional reagents as appropriate.
The rate of FAD pyrophosphorolysis was measured in the presence of 1 mM NaPPi (sodium pyrophosphate) and 0.5 µM FAD as in [25]. The rate of FAD cleavage was expressed as nmol FAD min −1 ·(mg protein) −1 , and was calculated from the rate of fluorescence increase, measured as the tangent to the initial part of the experimental curve. Under the same experimental conditions, but in the absence of NaPPi, the possible occurrence of FAD hydrolysis was investigated essentially as in [28,40], in the presence of 1 mM CoCl 2 .
To fit the experimental data and to obtain estimates of the kinetic parameters, the Grafit software was used (v. 5.0.13, 2006, by R.J. Leatherbarrow, Erithacus Software, Wilmington House, UK).

Electrophoretic Analysis
Proteins were separated by SDS-PAGE on 12% T polyacrylamide gels, according to Laemmli [41]. Quantitative evaluation of Coomassie Blue-stained protein bands was carried out using the Chemidoc imaging system (Bio-Rad, Hercules, CA, USA) and the Quantity One software (Bio-Rad, Hercules, CA, USA), as described previously [42].

Immunoblotting
SDS-PAGE-separated proteins were electro-transferred onto a nitrocellulose membrane using a trans-blot semidry electrophoretic transfer cell (Eppendorf, Hamburg, D). The immobilized proteins were incubated 1 h either with a 2500-fold dilution of an in-house produced polyclonal antiserum against hFADSs or with a 12,000-fold dilution of a peroxidase-conjugated anti-polyhistidine antibody. The bound antibodies were visualized with the aid of a secondary anti-rabbit immunoglobulin antibody, conjugated with alkaline phosphatase (1:5000 dilution).

3D Modeling of hFADS6
A three-dimensional model was obtained by submitting the target sequence to three homology and/or ab initio servers: Swiss Model (https://swissmodel.expasy.org), I-TASSER (https://zhanglab. ccmb.med.umich.edu/I-TASSER/) and Robetta (http://robetta.bakerlab.org) and by comparing their outputs. Swiss Model could only provide results for the C-terminal domain, whereas the threading/ab initio combined procedures implemented by the other two approaches produced almost complete models. Of the latter two, the model built by Robetta scored better in terms of protein geometry (Ramachandran plot, bond lengths, angles, dihedrals, rotamers, clashes, and contact energies), and was thus selected. FAD was transferred to it from one of the templates used by I-TASSER (namely, 3g6k); the resulting complex was then energy minimized with the Amber10:EHT forcefield and the reaction field electrostatics, by the Molecular Operating Environment v. 2016.0803.