Dependence of heme biosynthesis on iron-sulfur cluster biogenesis: human ALAD is an unrecognized iron-sulfur protein

Gang Liu National Institute of Child Health and Human Development Debangsu Sil Pennsylvania State University https://orcid.org/0000-0002-0718-3925 Wing-Hang Tong National Institute of Child Health and Human Development, NIH Nunziata Maio National Institute of Child Health and Human Development J. Martin Bollinger Jr. The Pennsylvania State University Carsten Krebs Pennsylvania State University Tracey Ann Rouault (  rouault@mail.nih.gov ) National Institute of Child Health and Human Development https://orcid.org/0000-0003-0062-0245


Introduction
In mammalian cells, substantial amounts of iron are consumed by heme biosynthesis and iron-sulfur cluster (ISC) biogenesis 1,2 . Heme is synthesized by eight sequential enzymatic steps 3 , of which the rst takes place in the mitochondrial matrix, where 5-aminolevulinate synthase (ALAS) catalyzes the condensation of succinyl-CoA with glycine to generate ALA ( Figure S1) 3,4 . Two ALAS genes are present in vertebrates, a ubiquitously expressed ALAS1 and an erythroid-speci c ALAS2 2,5 . The mRNA of ALAS2 has an iron responsive element (IRE) in its 5'-untranslated region (UTR) and is post-transcriptionally regulated by the iron regulatory proteins (IRP1 and IRP2) 6,7 . Since IRP1 loses its IRE-binding activity when it ligates an Fe-S cluster, ALAS2 protein levels are indirectly regulated by ISC biogenesis because apo-IRP1 binding represses ALAS2 translation 6,7 . ALAS1, which catalyzes the rate-limiting step of heme biosynthesis in non-erythroid cells 3 , is subject to negative feedback regulation by cellular heme content 8,9 . After its synthesis in mitochondria, ALA is exported to the cytosol, where ALA dehydratase (ALAD) catalyzes the second step of heme biosynthesis by condensing two ALA molecules into porphobilinogen ( Figure S1). ALAD is evolutionarily conserved and constitutes an important enzyme for chlorophyll synthesis, the corrin ring of vitamin B12, and other important tetrapyrroles 10,11 . After three additional enzymatic reactions in the cytosol and two in the mitochondria, the nal insertion of ferrous iron into protoporphyrin IX to generate heme occurs in the mitochondrial matrix and is catalyzed by ferrochelatase (FECH) ( Figure   S1) 4 . Human FECH is an Fe-S protein 12,13,14 , and its post-translational stability depends on coordination of a [Fe 2 S 2 ] cluster 12,15 .
Iron-sulfur clusters are ancient prosthetic groups with essential biological functions 16,17,18,19 . In the mitochondria of mammalian cells, ISCs are assembled de novo by a complex composed of NFS1, ISD11, ACP, the ISCU scaffold and frataxin 20 . After assembly, nascent clusters are transferred by an HSPA9/HSC20 chaperone/co-chaperone system directly to recipient proteins or through intermediate scaffolds 21 . Previous models have proposed that initial Fe-S synthesis occurs solely in the mitochondrial matrix 22 . However, the core mammalian ISC components have also been identi ed in the cytosol and nucleus, and accumulating evidence shows that the ISC biogenesis machineries likely operate independently to generate nascent clusters in several subcellular compartments of multicellular eukaryotes 23,24,25 . We previously reported that binding of HSC20 to a leucine-tyrosine-arginine (LYR) motif of succinate dehydrogenase complex subunit B (SDHB) was essential for ISC incorporation into SDHB 26 . The LYR motif was also identi ed in other HSC20-binding proteins 26,27 . Therefore, we speculated that analyzing protein sequences for the presence of the LYR motif and motifs with similar chemical properties (LYR-like motifs) could be used to discover candidate Fe-S proteins 26,28 .
Thus far, there are two well-characterized nodes in the heme biosynthetic pathway of mammalian cells at which defects in the Fe-S biogenesis machinery can suppress heme synthesis. First, FECH of higher eukaryotes contains a [Fe 2 S 2 ] cluster that is proposed to stabilize the enzyme 12,13 . Second, ALAS2 is post-transcriptionally regulated in erythroid cells by the interconversion of IRP1 between a holo-form that functions as cytosolic aconitase and an apo-IRE-binding protein that represses ALAS2 expression 7 .
Therefore, Fe-S biogenesis defects can block heme synthesis by either repressing ALAS2 synthesis in erythroid cells or inactivating FECH. However, it remains unexplained how Fe-S biogenesis defects in yeast result in impaired heme production 29,30 , given that yeast FECH is not an Fe-S protein and yeast lack IRPs 12,14 . The fact that Fe-S biogenesis de ciency impaired heme biosynthesis in yeast drove us to search for unrecognized intersections between the Fe-S and heme pathways using bioinformatics to screen heme biosynthetic enzymes for the presence of LYR motifs. We found that ALAD contained a LYRlike motif (A 306 F 307 R 308 ). Our previous extensive studies have revealed that the LYR-like motif can be de ned as follows: position 1, a small amino acid with a more aliphatic character, including leucine, isoleucine, valine, alanine and a few others. The second position must be either Y or F (aromatic) and the third position must be either R or K (positively charged) 26,28 . Here, we report that ALAD needs a [Fe 4 S 4 ] cofactor for optimal enzymatic activity upon over-expression in human cells and in bacteria. We performed functional studies to elucidate the molecular mechanisms by which human ALAD acquires and coordinates the [Fe 4 S 4 ] cluster, and to demonstrate the biological signi cance of the [Fe 4 S 4 ] cluster.

Results
Human ALAD coordinates a previously unrecognized [Fe 4 S 4 ] cluster First, we explored whether there are as-yet-uncharacterized convergent points between the Fe-S and heme biosynthetic pathways. We rst simultaneously knocked down the mitochondrial and cytosolic ISCU isoforms, which constitute the primary Fe-S biogenesis scaffolds in mammalian cells. Over-expression of yeast FECH, a non-Fe-S enzyme, did not restore heme biosynthesis in ISCU-de cient cells [Figures S2A-S2C and Supplemental Discussion (SD)]. Then we overexpressed a cytosolic ISCU mutant that impairs Fe-S biogenesis in the cytosol of mammalian cells and found that heme biosynthesis was inhibited (detailed in the SD and Figures S2D-S2F). Taken collectively, these results implied that one or more cytosolic heme biosynthetic steps was dependent on Fe-S biogenesis.
Then we screened the amino acid (AA) sequences of cytosolic heme biosynthesis enzymes for conserved LYR or LYR-like motifs to search for a potential Fe-S protein. Sequence alignment showed that human ALAD contains six cysteines, C119, C122, C124, C132, C162 and C223, and a LYR-like motif (Ala306-Phe307-Arg308) that is conserved among eukaryotic ALAD orthologs ( Figure S3). Therefore, in terms of AA sequence and composition, ALAD appeared to represent a good candidate as an Fe-S protein.
As we have previously shown that de novo Fe-S biosynthesis occurs simultaneously in the mitochondrial and cytosolic compartments of mammalian cells 23 , we rst determined whether human ALAD interacted with cytosolic HSC20. When the cytosolic fraction of HeLa cells was immunoprecipitated with anti-HSC20 antibody, the eluate contained ALAD ( Figure 1A), suggesting that endogenous cytosolic HSC20 (C-HSC20) interacted with ALAD. Because HSC20 binds directly to LYR motifs, permitting transfer of nascent clusters from ISCU directly to recipient proteins 16,26 , we then determined whether the AFR motif of human ALAD might be important for its interaction with HSC20. We transfected HeLa cells with plasmids expressing FLAG/MYC-tagged wild-type ALAD (ALAD WT) or AFR to alanine mutant (Mut AFR-AAA). Lysates from these cells were immunoprecipitated with anti-FLAG antibody, and the eluates were analyzed by western blotting. As shown in Figure 1B, wild-type ALAD interacted speci cally with C-HSC20 but not with the lower-molecular-weight mitochondrial HSC20 (M-HSC20), as expected, given that ALAD is a cytosolic enzyme. No interaction between C-HSC20 and the AFR-AAA mutant was detected. Collectively, these data demonstrated that ALAD interacted with C-HSC20 and that the AFR motif of ALAD was important for the interaction.
To investigate whether human ALAD coordinates an ISC in human cells, we quanti ed 55 Fe incorporation into C-terminally FLAG/MYC-tagged POLD1 (POLD1), a known [Fe 4 S 4 ] protein that acquires its cluster in the cytosol 31 , C-terminally FLAG/MYC-tagged ALAD WT and C-terminally FLAG/MYC-tagged Mut AFR-AAA in HEK293 cells transfected with either a pool of non-targeting siRNAs (NT) or with si-RNAs directed against human ISCU (si-ISCU). We found that ALAD and POLD1 bound comparable amounts of 55 Fe in control HEK293 cells transfected with non-targeting siRNAs ( Figure 1C and 1D). Mut AFR-AAA, which did not interact with C-HSC20 ( Figure 1B), bound signi cantly less 55 Fe than ALAD WT in NT HEK293 cells ( Figure 1C, P<0.001). Compared with control cells, ALAD and POLD1 bound signi cantly lower amounts of 55 Fe in HEK293 cells with defective ISC assembly machinery, achieved by silencing of human ISCU ( Figure 1C and 1D, P<0.001). Taken together, these results demonstrated that ALAD binds iron, presumably in the form of an ISC, in human cells. To determine whether human ALAD can indeed coordinate an ISC in human cells, we then overexpressed C-terminally FLAG/MYC-tagged ALAD WT and Mut AFR-AAA in Expi293 cells and puri ed the recombinant proteins anaerobically. Anaerobically puri ed recombinant ALAD migrated as a single band on SDS-PAGE ( Figure 1E) and exhibited a shoulder at ~420 nm in its UV-vis spectrum ( Figure 1F), suggesting the presence of [Fe 4 S 4 ] cluster(s) 32, 33, 34 .
We next expressed N-terminally MBP-tagged human ALAD (MBP-ALAD) in growing bacteria along with the A. vinelandii isc operon, which encodes the basic Fe-S biogenesis proteins and can assist iron-sulfur cluster assembly in bacteria, and puri ed the expressed protein anaerobically 35,36 . The anaerobically puri ed MBP-ALAD was brown ( Figure 1G), migrated as a single band on SDS-PAGE ( Figure 1H), and also exhibited a shoulder at ~420 nm in its UV-vis spectrum ( Figure 1I). 57 Fe-enriched MBP-ALAD sample was analyzed by Mössbauer and EPR spectroscopies. The 4.2-K Mössbauer spectrum collected in a 0.53 mT magnetic eld applied parallel to the direction of γ radiation ( Figure 1J, Figure S4) 17,37 . Taken together, these results demonstrate that each human ALAD monomer harbors one [Fe 4 S 4 ] cluster when it is expressed along with the isc operon and puri ed anaerobically.

To validate the presence of [Fe 4 S 4 ] cluster(s) and determine the stoichiometry of the [Fe 4 S 4 ] cluster(s), a
[Fe 4 S 4 ]-ALAD has signi cantly higher enzymatic activity than zinc-ALAD ALAD was previously identi ed as a zinc enzyme, and structures with zinc in the active site have been studied 38,39 . Therefore, to determine whether the [Fe 4 S 4 ] cluster of human ALAD is functionally signi cant, we compared the enzymatic activities of aerobically puri ed wild-type apo-ALAD (aero-apo-ALAD WT, see the Methods section for details), aerobically puri ed wild-type ALAD (aero-ALAD WT, aerobically puri ed wild-type ALAD prior to treating with EDTA, ), aerobically puri ed wild-type zinc bound-ALAD (aero-Zn-ALAD WT), and the anaerobically puri ed wild-type ALAD. As demonstrated by the ICP-MS analysis, aero-apo-ALAD WT did not contain iron or zinc, while aero-ALAD WT contained no zinc and a small amount of iron ( Figure 2A). The enzymatic activity of aero-Zn-ALAD WT was approximately six times and two times higher than those of aero-apo-ALAD WT and aero-ALAD WT, respectively (Figure 2A and 2B, P<0.01). These results suggested that the enzymatic activity of apo-ALAD can be partly restored by zinc reconstitution. Wild-type ALAD expressed along with isc operon and puri ed anaerobically (ALAD WT) did not contain zinc (Figure 2A). Its enzymatic activity was approximately three times higher than that of aero-Zn-ALAD WT ( Figure 2B the third bar versus the second bar, P<0.01). These results demonstrate that [Fe 4 S 4 ]-ALAD has signi cantly higher enzymatic activity than zinc-ALAD.
The six highly conserved cysteines, Lysine 252, and the LYR motif of human ALAD are important for the enzymatic activity of human ALAD that contains a [Fe 4 S 4 ] cluster in vitro To identify the AAs important for the coordination/acquisition of the [Fe 4 S 4 ], we mutagenized each of the six highly conserved cysteines, and Lys252, which has been shown to form a Schiff base with the carbonyl of one of the two ALA substrates in the enzymatic active site pocket ( Figures S5 and S6).
Additionally, we mutagenized the LYR-like motif of human ALAD to generate eight MBP-ALAD mutants (C119A, C122A, C124A, C132A, C162A, C223A, K252M, and Mut AFR-AAA). We expressed these mutants along with the bacterial isc operon and puri ed the over-expressed proteins anaerobically. To investigate whether the mutants coordinated a [Fe 4 S 4 ] cluster, we plotted their molar extinction coe cients based on their UV-vis spectra. As shown in Figure 2C, none of UV-vis spectra of the mutants exhibited the absorption at ~420 nm, which is a hallmark feature of [Fe 4  To determine whether the N-terminal MBP tag hindered the formation of a homo-octamer of ALAD subunits that constitute the previously de ned active form of ALAD 40 , and to assess how [Fe 4 S 4 ] cluster coordination affected ALAD quaternary structure, the anaerobically puri ed wild-type and mutant ALAD proteins were subjected to native gel electrophoresis. As shown in Figure 2D, most of the ALAD WT monomers formed ~640 kD homo-octamers, while all the ALAD mutants and the aero-apo-ALAD WT mostly assembled into ~480 kD homo-hexamers, a previously identi ed less active multimeric enzyme 40 .
These results suggested that MBP-ALAD can form homo-octamers in vitro and that [Fe 4 S 4 ] cluster coordination correlated with formation of higher proportions of ALAD homo-octamers.
The [Fe 4 S 4 ] cluster of human ALAD is crucial for its enzymatic activity and heme biosynthesis in HepG2 cells Next, we investigated the biological signi cance of the [Fe 4 S 4 ] cluster of ALAD in human HepG2 cells. We co-transfected HepG2 cells with siRNAs to knockdown (KD) the expression of endogenous ALAD, and with a plasmid directing expression of either recombinant wild type ALAD (the ALAD WT group) or the ALAD mutant in which the AFR motif had been replaced by alanines (Mut AFR-AAA group). Effective KD of endogenous ALAD was achieved in HepG2 cells ( Figure 3A). Recombinant ALAD WT or the Mut AFR-AAA mutant were well expressed with undetectable background expression of endogenous ALAD ( Figures  3A and 3B, P values <0.01). We used ALAS1 protein levels and the activities of the heme-dependent enzymes cytochrome P450 1A2 (CYP1A2) and catalase, as readouts of the integrity of the heme biosynthetic pathway 41,42,43 . As shown in Figures 3A, 3D and 3E, there were no signi cant differences in the protein level of ALAS1 and catalase and CYP1A2 activities between the ALAD WT group and untransfected HepG2 cells. In cells expressing ALAD-Mut AFR-AAA, however, we observed signi cantly decreased ALAD, catalase and CYP1A2 activities in cell lysates as compared with untransfected HepG2 cells and with cells expressing ALAD WT ( Figures 3A and 3C-3E, P values <0.01). These results demonstrated that the AFR-AAA mutant was non-functional and exhibited signi cantly lower enzymatic activity than wild-type ALAD. Therefore, overexpression of ALAD-Mut AFR-AAA did not restore heme biosynthesis that was suppressed by siRNA-induced ALAD de ciency. Given that the AFR motif of human ALAD is required for [Fe 4 S 4 ] cluster acquisition, these results demonstrate that the [Fe 4 S 4 ] of human ALAD is important for its enzymatic activity and heme biosynthesis in HepG2 cells.
Finally, we examined whether zinc addition affected ALAD enzymatic activity by adding zinc to the medium of mammalian cells over-expressing either WT ALAD or the Mut AFR-AAA (100 µM ZnCl 2 for 48 hours). As shown in Figure 3A, levels of human metallothionein MT2A, whose expression is regulated by zinc 44,45 , were greatly increased, suggesting that levels of cellular zinc indeed increased. However, there were no signi cant changes in either the protein levels of ALAS1 or the activities of ALAD, catalase or CYP1A2 after adding ZnCl 2 to the medium ( Figures 3A and 3C-3E). These results indicated that increasing intracellular zinc concentrations did not increase ALAD enzymatic activity in HepG2 cells overexpressing either wild-type ALAD or the AFR-AAA mutant. Notably, FECH protein levels were not affected by zinc addition, indicating that the coordination of [Fe 2 S 2 ] cluster by human FECH was also unaffected by increased intracellular zinc levels ( Figure 3A).

Discussion
In this study, we found that human ALAD requires a [Fe 4 S 4 ] cluster to acquire its full enzymatic activity both in vitro and in HepG2 cells. More importantly, the [Fe 4 S 4 ] cluster of ALAD is crucial for heme biosynthesis, unveiling a node of regulation of heme biosynthesis by ISC biogenesis that operates at the second step of the heme biosynthetic pathway.
To determine whether one or more steps of the cytosolic reactions of heme biosynthesis depended on Fe-S cluster biogenesis, we overexpressed a form of cytosolic ISCU that lacked a mitochondrial targeting signal and contained the mutation D46A, which prevents the ISCU scaffold from transferring its Fe-S cluster to recipient proteins in mammalian cells. Our results demonstrated that two heme-dependent enzymes, catalase and cytochrome P450 oxidase, lost activity when the D46A ISCU construct was expressed to suppress cytosolic Fe-S biogenesis. We then used bioinformatics to evaluate whether the AA sequences of any of the cytosolic heme biosynthesis enzymes contained LYR-like motifs and potential cluster-ligating cysteines, which we have shown may represent features of candidate Fe-S proteins 28 . Only human ALAD was found to contain an LYR-like motif, consisting of Ala306-Phe307-Arg308. ALAD had been previously characterized as a zinc protein, as ALAD puri ed from beef liver contained much larger amounts of zinc than iron 46 . Also, the low enzymatic activity of metal-free ALAD could be partially restored by zinc addition 46,47 . In later studies, it was demonstrated that ALAD octamers required four zinc ions for enzymatic activity 38 . However, all these studies were performed under aerobic conditions that can interfere with identi cation of iron-sulfur clusters. In fact, many Fe-S proteins have been previously mischaracterized as zinc proteins, such as CPSF30 48 , human CISD2 49 , and yeast POL3 31 . Therefore, we decided to explore whether ALAD is an Fe-S protein. According to our results, human ALAD can coordinate an ISC in human cells. In addition, each human ALAD monomer harbored one [Fe 4 S 4 ] cluster when it was expressed along with the isc operon and puri ed anaerobically.

We then compared the enzymatic activity of [Fe 4 S 4 ]-ALAD with that of zinc-ALAD and found that [Fe 4 S 4 ]-
ALAD had signi cantly higher enzymatic activity than zinc-ALAD. ALAD catalyzes the asymmetric condensation of two ALA molecules to form porphobilinogen. The ALA molecule that becomes the propionyl side chain and pyrrole nitrogen-containing part of porphobilinogen is named P-side ALA, whereas that which contributes the acetyl side chain and amino nitrogen-containing part of porphobilinogen is known as the A-side ALA. According to previous studies, the rst step of this enzymatic reaction is the formation of P-side Schiff base of ALA with the nitrogen of lysine 252, followed by A-side ALA binding to the other side of enzymatic active site pocket 47,50,51,52,53 . It was proposed that in zinc-ALAD, the zinc functioned as a Lewis acid to bind A-side ALA 51 . Based on the resemblance of the ALAD reaction to those catalyzed by aconitase 54 , dihydroxyacid dehydratase 55 , and quinolinate synthase 56,58 , we speculate that the [Fe 4 S 4 ] cluster of ALAD is coordinated by only three Cys ligands, and the non-Cys-coordinated iron may function as a Lewis acid and may be coordinated by the ALA substrate, thereby enabling dehydration to take place ( Figure S5) 54,57 . As a [Fe 4 S 4 ] cluster occupies a larger space than Zn(II), we further speculate that the [Fe 4 S 4 ] cluster of ALAD forces the A-side ALA to move closer to the P-side ALA than when ALA is bound to Zn(II), thereby optimizing proximity of the two ALA molecules and facilitating the condensation reaction.
Most Fe-S proteins coordinate their clusters through cysteines 17,58 . To identify the cysteines important for [Fe 4 S 4 ] cluster acquisition and/or coordination, we mutagenized the six highly conserved cysteines of human ALAD to alanine residues (A). According to previous studies, lysines 210 (K210) and 263 (K263) in yeast ALAD, which correspond to K199 and K252 in human ALAD, form Schiff bases with two molecules of ALA to facilitate their condensation 59,60 . Therefore, we also mutagenized K252 in human ALAD to methionine to investigate the impact of a K252M mutant on [Fe 4 S 4 ] cluster coordination and ALAD enzymatic activity. To examine the biological signi cance of the AFR motif, we substituted it with three alanines (Mut AFR-AAA). According to our results, all the AA residues investigated are important for [Fe 4 S 4 ] cluster coordination and the enzymatic activity of human ALAD in vitro. Our ndings are consistent with those of a previous study in which a combination of C122A, C124A and C132A greatly reduced activity of human ALAD that ligates zinc 38 . In both the zinc-containing form of the enzyme which can be generated in vitro, and the Fe-S-containing form detected after overexpression in bacteria, cysteines 122, 124 and 132 appear to be required for acquisition of enzymatic activity. Consistent with these ndings, a Cys132Arg mutation was found to be responsible for ALAD porphyria in a 14-year-old boy 61 . Furthermore, we speculate that the structure of ALAD with a [Fe 4 S 4 ] cluster may resemble the crystal structure of Zn-ALAD ( Figure S6) 39 Figure S6) in positions where they have been proposed to form a disul de bond that stabilizes ALAD tertiary structure 38 . Therefore, mutagenesis of the cysteine residues that ligate zinc in the available crystal structure may directly affect ISC ligation, whereas mutagenesis of C119 and C162 can impair cluster ligation as a result of misfolding of ALAD. Note that the K252M mutant completely lost its enzymatic activity. We speculate that this may be caused by loss of binding of one of the two ALA molecules that are condensed to form porphobilinogen in the "active site pocket" as previously de ned 59,60 , but Fe-S ligation was notably absent as well for unclear reasons.
The quaternary structure of human ALAD can switch between a homo-hexameric and a homo-octameric con guration 40 . Wild-type ALAD monomers tend to form a homo-octamer with full enzymatic activity through interactions between their N-terminal domains 40,63 . In the presence of ALAD mutations or of "morphlocks", small molecules that stabilize speci c quaternary structures, ALAD can assemble into a homo-hexameric con guration with lower enzymatic activity than the homo-octameric con guration 63,64 . In this study, we found that [Fe 4 S 4 ] cluster coordination correlated with increased proportions of ALAD homo-octamers, which may be another reason why [Fe 4 S 4 ]-ALAD has high enzymatic activity.
Finally, we explored the biological signi cance of the [Fe 4 S 4 ] cluster of ALAD in human HepG2 cells. We performed functional assays to investigate whether overexpression of either wild-type human ALAD or the AFR-AAA mutant could rescue the heme biosynthetic defect of ALAD-knockdown cells. According to our results, the [Fe 4 S 4 ] of human ALAD is important for its enzymatic activity and heme biosynthesis in HepG2 cells. Interestingly, we also found that increased intracellular zinc concentrations, achieved by adding ZnCl 2 into the cell culture medium, did not increase ALAD enzymatic activity in wild-type ALAD-or AFR-AAA mutant-overexpressing HepG2 cells, suggesting that mammalian cells lack mechanisms that speci cally deliver zinc to ALAD in growing cells.
In conclusion, our ndings have several important biological implications. In iron-de cient non-erythroid cells, reduced ISC availability will decrease the enzymatic activity of ALAD, diminishing the potential accumulation of pyrroles and porphyrins synthesized downstream of ALAD that could otherwise occur in conjunction with reduced holo-FECH levels and/or increased ALAS1 protein levels. Porphyrin intermediates are toxic, and their accumulation causes several types of human porphyria 65,66,67 .
Therefore, the regulation of ALAD activity by ISC biogenesis may represent a new mechanism preventing iron-or ISC-de cient non-erythroid cells from building up harmful heme metabolic intermediates. In addition, given that Fe-S clusters are sensitive to oxidative stress and that ALAD was shown to be a direct target of oxidative stress 17, 68 , we speculate that oxidative damage of the [Fe 4 S 4 ] of human ALAD may be implicated in pathological conditions associated with both increased oxidative stress and anemia, such as chronic renal failure 69 . Furthermore, since we con rmed the biological signi cance of a LYR or LYR-like motif, our study validates our prediction that enzymes unrelated to the mitochondrial respiratory chain depend on Fe-S delivery from a co-chaperone/chaperone pathway that is critical for Fe-S acquisition 26 . Our work may encourage researchers to identify other novel Fe-S proteins based on presence of LYR motifs in proteins that contain cysteine ligands for Fe-S coordination in an adjacent peptide or protein complex.

Methods
All the data in this study are available upon request.

Cell culturing and transfection
Human cervical carcinoma HeLa cells, HEK293 cells, and hepatocellular carcinoma HepG2 cells were obtained from ATCC. Expi293 cells were purchased from ThermoFisher Scienti c. The stably-transfected HEK293T cell lines were established in our previous study 23  Immunoprecipitations of recombinant FLAG/MYC-tagged POLD1 (POLD1), wild-type ALAD (ALAD WT), or ALAD mutant of the AFR motif (Mut AFR-AAA) were performed on cytosolic fractions from HEK293 or Expi293 cells in a nitrogen recirculated glove box operated at <0.2 ppm O 2 48h after transfection with plasmids encoding POLD1, ALAD WT, or Mut AFR-AAA. Anti-FLAG immunoprecipitations were performed using M2-FLAG beads (Sigma). Washed FLAG M2 beads were added to the lysate and incubated for 2 h at room temperature under mild agitation. After the beads were recovered, recombinant proteins were competitively eluted with 100 μg/ml 3xFLAG peptide (Sigma).
UV-vis spectra were acquired using a NanoDrop spectrophotometer (ThermoFisher) using a column buffer supplemented with 100 μg/ml 3xFLAG peptide as the blank. 55 Fe incorporation assay HEK293 cells were grown in the presence of 1 μM 55Fe-Tf and transfected twice at 48h interval with si-RNAs targeting ISCU or with non-targeting si-RNAs. At the time of the second transfection with si-RNAs, cells were co-transfected with FLAG/Myc-tagged ALAD WT or Mut AFR-AAA, or with FLAG/Myc-tagged POLD1, which was included in the experiment as a positive control. Cytosolic extracts were prepared in the glove box after the second transfection and subjected to immunoprecipitation with anti-FLAG M2 agarose beads.
Overexpression and anaerobic puri cation of maltose-binding protein (MBP)-tagged human ALAD proteins BL21(DE3) competent bacterial cells (NEB) were co-transformed with pDB1282, a plasmid directing expression of Azotobacter vinelandii isc operon, and pMALc5x-MBP-ALAD plasmids, which directs expression of N-terminally MBP-tagged wild-type or mutant human ALAD proteins. The co-transformants were then cultured, induced for protein expression and lysed anaerobically. The N-terminally MBP-tagged ALAD proteins were then puri ed by a nity chromatography. See SD for details.
Ultraviolet-visible (UV-vis) spectroscopy, inductively coupled plasma mass spectrometry (ICP-MS), amino acid analysis (AAA), electron paramagnetic resonance (EPR) spectroscopy and Mössbauer spectroscopy Uv-vis spectra were acquired using a NanoDrop spectrophotometer (ThermoFisher) with column buffer supplemented with 10 mM maltose used as the blank. For ICP-MS, 25 µL sample was digested overnight and iron and zinc concentrations were determined with an Agilent 7900 ICP-MS system. AAA was performed by Alphalyse to precisely quantify the puri ed ALAD proteins. To verify the presence of an ironsulfur cluster, the anaerobically puri ed MBP-ALAD (250 µM) was subjected to both EPR and Mössbauer analyses. See SD for details.
Preparation of apo-ALAD and zinc-bound ALAD Aerobically puri ed human ALAD was treated with 10 mM EDTA for 1 h and passed through a G-25 column (GE Healthcare) to yield metal-free ALAD (apo-ALAD). For zinc reconstitution, apo-ALAD was incubated with 5 mM dithiothreitol and 20 equivalents of ZnCl 2 at room temperature for 2 h and subsequently passed through a G-25 column.
Western blotting, co-immunoprecipitation and native gel electrophoresis Organelle and cytosolic fractions were lysed with M-PER ® Protein Extraction Reagent (ThermoFisher). For western blotting, 20-40 µg protein sample was used. Mouse monoclonal anti-FLAG antibody was from Origene, anti-ISCU rabbit polyclonal serum was raised against a synthetic peptide fragment, as previously reported 25 ; anti α-TUB antibody was from Sigma. All the other primary antibodies for western blotting were from Abcam.
Co-immunoprecipitation (Co-IP) was carried out using the Pierce Co-Immunoprecipitation Kit (ThermoFisher). See SD for details.
For native gel electrophoresis, anaerobically puri ed ALAD proteins were separated, and the bands were visualized using the NativePAGE™ Bis-Tris Gel System (ThermoFisher) according to the manufacturer's instructions.
ALAD, cytochrome P450 (CYP), and catalase activity assays ALAD activity was determined according to a published protocol 70 . The CYP and catalase activities were analyzed by the CYP1A2 Activity Assay Kit (Abcam) and Catalase Activity Assay Kit (Abcam), respectively.

Statistical analysis
Where applicable, three independent replicates were performed, and results are presented as mean ± standard deviation (S.D.). Two-sided Student's t-test was used to determine the P value. P<0.05 and P<0.01 were considered signi cant (* or a ) and very signi cant (** or aa ), respectively. Statistical analyses of 55 Fe labeling experiments were performed with GraphPad Prism 7 using two-way analysis of variance  Human ALAD expressed in both human cells and bacteria (along with the isc operon) coordinates a [Fe4S4] cluster (A) Immunoprecipitation (IP) of cytosolic fractions of HeLa cells using anti-HSC20 antibody shows that ALAD co-precipitates with HSC20. The input was 5% of the cytosolic fraction lysate, and the remaining 95% of the lysate was subjected to IP with anti-HSC20 antibody. The eluates were then subjected to Western blotting with anti-HSC20 and anti-ALAD antibodies. (B) IP of whole cell lysates of