Cross-regulation of [2Fe–2S] cluster synthesis by ferredoxin-2 and frataxin

Protein purification

Homo sapiens ISCU2 (Q9H1K1-1), (NIA)₂ (Q9Y697, Q9HD34 and O14561), FXN (Q16595), FDXR (P22570) and wild-type and mutant FDX2 (Q6P4F2) proteins were prepared as previously described¹⁸. BL21 Star (DE3) cells were transformed with the corresponding plasmids: pCDFDuet-His₆-Thr-NFS1 (site 1), ISD11 (site 2) with pET21(+)-ACP for (NIA)₂, pET28a(+)-His_6-Thr-ISCU2 for ISCU2, pET28a(+)-His_6-Thr-FDX2 for wild-type and mutant FDX2 constructs, pET28a(+)-His₆-Thr-FXN for FXN, and pCDFDuet-His₆-FDXR with pET28a(+)-GroEL for FDXR, where His₆ corresponds to a 6× histidine tag and Thr corresponds to a thrombin cleavage site with sequence SQDLVPRGS. Cells were cultured in LB with appropriate antibiotics at 37 °C at 180 rpm. Cell culture transformed with wild-type and mutant FDX2 were supplemented with FeSO₄ to a final concentration of 10 µM, and FDXR growth was supplemented with riboflavin to a final concentration of 3.75 µM. Protein overexpression was induced by IPTG to a final concentration of 1 mM. (NIA)₂ and FDXR were induced when the optical density at 600 nm (OD₆₀₀) reached 0.6, and incubated overnight at 25 °C (FDXR) or 18 °C ((NIA)₂). ISCU2, FXN and FDX2 were induced when the OD₆₀₀ reached 0.7 and incubated for 3.5 h at 30 °C. Cells were collected by centrifugation at 4,500 rpm for 20 min at 4 °C with a Beckman Coulter JLA-10.500 rotor, and resuspended in His buffer A (20 mM Tris-HCl, 250 mM NaCl and 5 mM imidazole, pH 8.0) with the addition of protease inhibitor (cOmplete Protease Inhibitor Cocktail). Cells were lysed by sonication and cell debris was removed by centrifugation at 45,000 rpm for 45 min at 4 °C with a Beckman Coulter Type 70 Ti rotor. Supernatant was loaded to a Cytiva 5-ml HisTrap HP pre-equilibrated with His buffer A and eluted over a linear gradient of 0–100% His buffer B (20 mM Tris-HCl, 250 mM NaCl and 500 mM imidazole, pH 8.0). Fractions were pooled and concentrated by ultracentrifugation using Amicon Ultra-15 Centrifugal Filter Units with a molecular-mass cut-off of 10 kDa (ISCU2, FXN, FDX2 and FDXR) or 50 kDa for (NIA)₂. Proteins (ISCU2, FXN, FDX2 and (NIA)₂) were buffer-exchanged into SEC buffer (20 mM Tris-HCl and 250 mM NaCl, pH 8.0) using a NAP-5 desalting column, and FDXR was buffer-exchanged into ion-exchange buffer A (20 mM Tris-HCl and 25 mM NaCl, pH 8.0). ISCU2, FXN and FDX2 were treated with recombinant thrombin protease to remove the His₆-tag. ISCU2 was incubated with 50 molar eq. DTT and 100 eq. DTPA for 60 min at 20 °C. (NIA)₂ was incubated with 50 eq. DTT, 2 eq. PLP and 4 eq. TCEP for 60 min at 20 °C. Proteins were subjected to size-exclusion chromatography using HiLoad 16/600 columns packed with Superdex 75 pg (ISCU2, FXN and FDX2) or 200 pg for (NIA)₂ pre-equilibrated with SEC buffer (FXN and FDX2) or SEC buffer + 10 mM DTT (ISCU2 and (NIA)₂). Proteins were eluted and fractions were pooled. FDXR was loaded onto a Cytiva 5-ml HiTrap Q-Sepharose column and eluted with a linear gradient of 0–100% ion-exchange buffer B (20 mM Tris-HCl, 500 mM NaCl, pH 8.0). Proteins were concentrated to 0.5 ml using the corresponding Amicon Ultra-15 Centrifugal Filter Unit and buffer-exchanged under anaerobic conditions (< 2 ppm O₂) using a NAP-5 column into either degassed Tris buffer (20 mM Tris-HCl and 100 mM NaCl, pH 8.0) or sodium phosphate buffer (50 mM Na₂HPO₄ and 150 mM NaCl, pH 8.0) sparged with argon for 60 min. Protein concentrations were determined by UV-visible spectroscopy at 280 nm in urea buffer (100 mM Tris-HCl and 8 M urea, pH 7.5) using absorption coefficients of 52,260 M⁻¹ cm⁻¹, 9,970 M⁻¹ cm⁻¹, 26,930 M⁻¹ cm⁻¹ and 44,920 M⁻¹ cm⁻¹ for (NIA)₂, ISCU2, FXN and FDXR, respectively. The FDX2 protein concentration was determined by UV-visible spectroscopy at 456 nm in Tris buffer using an absorption coefficient of 10,000 M⁻¹ cm⁻¹ based on the concentration of the [2Fe–2S] cluster as determined using the ferrozine method. Final preparations were aliquoted, flash-frozen in liquid N₂ and stored in liquid N₂.

Assays of [2Fe–2S] cluster assembly

Fe–S cluster assembly reactions were performed under anaerobic conditions (< 2 ppm O₂) in a glove box. Kinetic assays were routinely composed of 20 µM apo-ISCU2 incubated with 20 µM ferrous ammonium sulfate (Fe-(NH₄)₂(SO₄)₂), 2 µM of the (NFS1–ISD11–ACP) complex, 1 µM FDXR and 40 µM NADPH. Titrations were performed with 0–50 molar equivalents of FXN (corresponding to 0–100 µM, where one equivalent is relative to the concentration of the (NFS1–ISD11–ACP) complex), and 0–10 eq. of either wild-type or mutant FDX2 (0–20 µM). The reaction mix was transferred into a 384-well plate and incubated at 25 °C in a Tecan Spark microplate reader. Biosynthesis of [2Fe–2S] was initiated by the injection of 30 µM -cysteine to a final volume of 100 µl and kinetics were measured at 456 nm at 25 °C. The data were collected using SparkControl Magellan 3.0. The kinetic rates of [2Fe–2S] were calculated on the basis of the slope at the start of the curve corresponding to the maximum rate.

ARBS assays

Assays were routinely performed in sodium phosphate buffer (50 mM Na₂HPO₄ and 150 mM NaCl, pH 8.0) at 25 °C with a concentration of ISCU2 and equimolar ferrous ammonium sulfate and/or the (NFS1–ISD11–ACP) complex at 7.5 µM. Concentrations of FDX2 and FXN ranged from 0–10 eq. (0–75 µM) as indicated within the text. For studies following persulfide formation on NFS1, reactions were initiated with 7.5 µM l-cysteine. Assays following persulfide transfer from NFS1 to ISCU2 were initiated with 15 µM l-cysteine. Persulfide reduction assays were initiated with 7.5 µM l-cysteine followed by the addition of a preincubated mixture of FDX2, FDXR and NADPH to final concentrations of 7.5 µM, 3.75 µM and 40 µM, respectively. Reactions were sampled at respective timepoints by mixing 15 µl of the reaction mix with 5 µl of a stop mix making up a fivefold molar excess of mal-dPEG relative to the total thiol concentration (including l-cysteine and total cysteine residues), a 2.5-fold molar excess of EDTA relative to the total iron concentration and a final concentration of 1% SDS. To monitor the persulfidation state of ISCU2 in the presence of FDX2 and prevent overlap of alkylated ISCU2 and FDX2 on gel, which both contain four cysteine residues and have similar molecular weights of 14.4 kDa and 14.3 kDa, respectively, we identified an optimal concentration of SDS in which the solvent-accessible cysteine residues of ISCU2 were available for alkylation, whereas the cysteine residues of FDX2 ligating the [2Fe–2S] cofactor remained partially buried and shielded from alkylation (Supplementary Fig. 1). A final concentration of 0.08% SDS was found to be optimal for this detection. After 30 min of reaction with the stop mix, leading to full alkylation of ISCU2 and/or NFS1, 10 µl of reducing loading dye (60 mM Tris-HCl, 25% glycerol, 2% SDS, 700 mM 2-mercaptoethanol and 0.1% bromophenol blue) was added. Reaction aliquots were analysed by SDS–PAGE on 8% (NFS1) and 14% (ISCU2) acrylamide/bis-acrylamide 19:1 gels. The gels were imaged using an Odyssey Clx scanner (Li-COR) and the data were collected and analysed using Image Studio 5.2.

FIDA

Stock solutions of labelled FXN and FDX2, referred to as FXN_ALC and FDX2_ALC, were prepared by conjugation with ALC 480 using the corresponding Fidabio Protein Labelling Kit (Fida Biosystems). For labelling reactions, FXN and FDX2 were first buffer-exchanged into sodium phosphate buffer (50 mM Na₂HPO₄ and 150 mM NaCl, pH 8.0) using a NAP-5 column. Then, 75 µl of protein sample (FXN or FDX2) at 100 µM was incubated with 7.5 µl 1 M sodium bicarbonate. A 4 mg ml⁻¹ reactive dye stock solution was prepared by mixing 100 µg ALC 480 with 25 µl DMSO. Then, 7.5 µl of the dye stock solution was incubated at 21 °C for 30 min with the protein sample, protected from light, to induce labelling using a fivefold molar ratio of dye to protein. Excess dye was removed and proteins were transferred to Tris buffer (20 mM Tris-HCl and 100 mM NaCl, pH 8.0) using a NAP-5 column. The concentration of FXN_ALC was measured by UV-visible spectroscopy at 280 nm using an absorption coefficient of 26,930 M⁻¹ cm⁻¹. Because the ALC 480 dye interferes at wavelengths 400–540 nm, the concentration of FDX2_ALC was measured by UV-visible spectroscopy at 320 nm using an absorption coefficient of 20,450 M⁻¹ cm⁻¹ as determined by the ferrozine method. Labelled proteins were aliquoted, flash-frozen in liquid N₂ and stored in liquid N₂. As a control to ensure that the label was not interfering with protein function, we performed [2Fe–2S] cluster assembly kinetics, and observed that the efficiencies of both FXN_ALC and FDX2_ALC were comparable with those of the unlabelled proteins (Extended Data Fig. 2b).

Binding curves of FXN_ALC or FDX2_ALC with the (NIAU)₂ complex were generated using a premix method in which the indicator sample (the labelled protein) is preincubated with the analyte (the binding partner (NIAU)₂) before analysis by FIDA. For the FXN_ALC binding curve, data points are gathered by incubating a fixed concentration of 20 nM FXN_ALC with a series of (NIAU) concentrations spanning 0–16 µM (corresponding to 0–8 µM of the dimerized (NIAU)₂ complex) in a final volume of 20 µl. The data points of FDX2_ALC with the (NIAU)₂ complex is prepared by incubating a fixed concentration of 20 nM FDX2_ALC with a titration of 0–128 µM of the (NIAU) complex (corresponding to 0–64 µM of the dimerized (NIAU)₂ complex) in a final volume of 20 µl. Protein samples were prepared in Tris buffer (20 mM Tris-HCl and 100 mM NaCl, pH 8.0) and samples were incubated for a minimum of 10 min to ensure binding equilibrium is reached before assay.

For competition between FXN_ALC and either wild-type or mutant FDX2 constructs for the (NIAU)₂ complex, a fixed concentration of FXN_ALC (20 nM) and (NIAU) (2 µM) was titrated by unlabelled FDX2 spanning 0–64 µM. For measurements with reduced FDX2, deoxygenated solutions of the analyte, indicator and buffer were prepared under anaerobic conditions in a glove box. Then, 10 molar eq. of dithionite were added to each FDX2 sample to reduce and keep it reduced throughout the experiment. The full reduction of FDX2 was checked by UV-visible spectroscopy at 456 nm. The analyte, indicator and buffer were loaded into capped vials sealed with tape and placed into a septum sealed tube before analysis. We checked by UV-visible spectroscopy that FDX2 remained reduced under the conditions and duration of FIDA analysis after removal of the tape on top of the cap while keeping tape around the seal. For the competition between FDX2_ALC and FXN for the (NIAU)₂ complex, a fixed concentration of FDX2_ALC (20 nM) and (NIAU) (16 µM) was titrated by 0–256 µM unlabelled FXN.

Binding experiments were performed using a Fida 1 instrument (Fida Biosystems) using laser-induced fluorescence detection with an excitation wavelength of 480 nm. The data were collected using Fida 1 instrument software (v.2.42). A permanently coated capillary (outer diameter 375 µm, inner diameter 75 µm, length to detector 84 cm; Fida Biosystems) was routinely used with constant temperature control of 25 °C. The capillary was equilibrated with the analyte at 3,500 mbar for 30 s. After this, a small plug of the premixed indicator sample was injected at 50 mbar for 10 s. Finally, the indicator was mobilized by injecting the analyte at 400 mbar for 180 s and fluorescence was measured at the detector window. Competition binding curves were performed using a premix method with the same mobilization protocol.

The data points were analysed using Fida analysis software (v.2.32) with the single-species fit to determine the corresponding hydrodynamic radius. Binding curves were then fit using GraphPad Prism (v.8.0.2) to a single-site binding model⁴³ (FXN_ALC with (NIAU)₂ or FDX2_ALC with (NIAU)₂) or a competitive binding model (FXN_ALC–(NIAU)₂ or FDX2_ALC–(NIAU)₂ titrated with unlabelled FDX2 or FXN) (Supplementary Equation 1).

ITC

ITC measurements were performed at 25 °C with a MicroCal PEAQ-ITC system (Malvern Panalytical). Zn–ISCU2 was prepared by incubating apo-ISCU2 with 5 eq. DTT and 1.5 eq. ZnSO₄ for 60 min and then exchanged to remove excess DTT and ZnSO₄ using a NAP-5 column. All protein samples used in ITC experiments were buffer-exchanged into the same preparation of Tris buffer (20 mM Tris-HCl and 100 mM NaCl, pH 8.0) using a NAP-5 column.

The 0.24-ml sample cell was loaded with 20 µM of the (NIA)₂ or (NIAU–Zn)₂ complex (corresponding to 40 µM of NFS1). Aliquots (2 µl) of the partner proteins (apo-ISCU2, Zn-loaded ISCU2, FDX2 or FXN) were placed in the 40-µl syringe at 400 µM and injected into the sample cell every 180 s under a stirring speed of 500 rpm using the MicroCal PEAQ-ITC Control Software (v.1.41). For the interaction of FDX2 or FXN with ISCU2–Zn, the sample cell was loaded with 40 µM ISCU2–Zn and the syringe was loaded with 400 µM FDX2 or FXN. For the interaction of FXN with FDX2, the sample cell was loaded with 40 µM FXN and the syringe was loaded with 400 µM FDX2. For experiments measuring the interactions of the mutant FDX2 constructs with (NIA)₂, the cell was loaded with 10 µM of the (NIA)₂ complex, and the syringe with 200 µM of the corresponding FDX2 mutant.

Raw thermograms were integrated using NITPIC (v.2.1.0)^44,45. The integrated ITC data were analysed using SEDPHAT (v.15.2b)⁴⁶. With respect to SEDPHAT nomenclature, interactions of (NIA)₂–Zn–ISCU2, (NIA)₂–apo-ISCU2 or (NIAU)₂–FXN were fitted using the model ‘A + B + B <-> {AB} + B <-> ABB; with two symmetric sites, macroscopic K’, where A corresponds to (NIA)₂ or (NIAU)₂, and B corresponds to apo-ISCU2, Zn–ISCU2 or FXN. The interaction between (NIA)₂ or (NIAU)₂ with FDX2 was modelled using ‘A + B + B <-> AB + B <-> BA + B <-> BAB; with 2 non-symmetric sites, microscopic K’, where A corresponds to (NIA)₂ or (NIAU)₂, and B corresponds to FDX2. The 68.3% error intervals determined during parameter fitting within SEDPHAT were calculated as previously described⁴⁷. ITC figures were produced using GUSSI (v.2.1.0)⁴⁸.

AlphaFold

We used our in-house implementation of ColabFold 1.3⁴⁹, which incorporates AlphaFold 2.2⁵⁰, to generate models for the ISC complex composed of NFS1, ISD11, ACP, ISCU2 and FDX2, with the corresponding Uniprot IDs Q9Y697, Q9HD34, O14561 (69–156), Q9H1K1 (35–167) and Q6P4F2 (56-186), respectively. For the construction of multimeric models, we used the multimer version of AlphaFold 2.2⁵¹ on a local workstation equipped with an RTXA6000. Amino acid sequences were used to query the UniRef30 database (June 2021) and the ColabFold environmental database (August 2021) using the MMseqs2 webserver (v.2)⁵². The resulting alignments were then used as input for AlphaFold, using ColabFold’s scripts, to generate five models per input. Each model generation involved six cycles and final refinements ranked according to local confidence score using predicted local distance difference test (pLDDT) and predicted aligned error (PAE).

Drosophila lines and culture methods

The fh-GAAs and daGS lines were previously described^39,53. The w¹¹¹⁸ control and UAS-Fdx1 RNAi (v104499) lines were obtained from the Vienna Drosophila Resource Center (VDRC) and the daGAL4 line (PGAL4-da.G32-2) from the Bloomington Stock Center. The composition of the fly food, methods for RU486 treatments and lifespan analysis were as previously described³⁹. For survival experiments, male flies of the same genotype were affected randomly to various doses of RU-486 treatments. Dead flies were counted concurrently with the transfer of flies to fresh medium containing the specified RU486 concentration, blinding was not applicable. 

Quantification of transcripts by quantitative PCR with reverse transcription (RT–qPCR)

Total RNA was extracted from 20 third instar male larvae or adults and treated with dsDNase (Thermo Fisher Scientific) before cDNA synthesis using SuperScript III Reverse Transcriptase (Thermo Fisher Scientific). Quantitative PCR was performed with the qPCR Mix (Promega) on a LightCycler 480 (Roche). The ribosomal gene rp49 was used as an internal reference for normalization. The primers used for amplifications were 5′-CCGCTTCAAGGGACAGTATCT-3′ and 5′-CACGTTGTGCACCAGGAACTT-3’ for rp49 (CG7939), 5′-GTCCACGGATGAAATAGTGAACA-3′ and 5′-GGGCCAAGTACAGAACATTGTC-3 or 5′-CTACACACCCCACAATGCTCT-3′ and 5′-TCACTATTTCATCCGTGGACTTG-3 for Fdx1, 5′-GGCCCTGAAAACGAAAGATGT-3′ and 5′-GCCCCGAAACCATCCAGAT-3 for Fdx2 and 5′-ACACCCTGGACGCACTGT-3′ and 5′-GTTGATCACATAGGTGCCGTG-3 for fh. Ct values were obtained using the second derivative method. Quantifications were made on four to six independent biological samples with four technical replicates for each sample. The data were collected and analysed using LightCycler 480 Software release 1.5.0. Statistical analysis were performed with Mann–Whitney tests or one-way ANOVA tests followed by post-hoc Tukey analysis for multiple comparisons.

Software

Unless otherwise stated, the data were analysed and plotted using GraphPad Prism 8.0.2 and Excel 2016. Protein structures were visualized using PyMOL 3.0. Figure layouts were created using PowerPoint 2016.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.