Human glutaredoxin-1 can transfer copper to isolated metal binding domains of the P1B-type ATPase, ATP7B

Intracellular copper (Cu) in eukaryotic organisms is regulated by homeostatic systems, which rely on the activities of soluble metallochaperones that participate in Cu exchange through highly tuned protein-protein interactions. Recently, the human enzyme glutaredoxin-1 (hGrx1) has been shown to possess Cu metallochaperone activity. The aim of this study was to ascertain whether hGrx1 can act in Cu delivery to the metal binding domains (MBDs) of the P1B-type ATPase ATP7B and to determine the thermodynamic factors that underpin this activity. hGrx1 can transfer Cu to the metallochaperone Atox1 and to the MBDs 5-6 of ATP7B (WLN5-6). This exchange is irreversible. In a mixture of the three proteins, Cu is delivered to the WLN5-6 preferentially, despite the presence of Atox1. This preferential Cu exchange appears to be driven by both the thermodynamics of the interactions between the proteins pairs and of the proteins with Cu(I). Crucially, protein-protein interactions between hGrx1, Atox1 and WLN5-6 were detected by NMR spectroscopy both in the presence and absence of Cu at a common interface. This study augments the possible activities of hGrx1 in intracellular Cu homeostasis and suggests a potential redundancy in this system, where hGrx1 has the potential to act under cellular conditions where the activity of Atox1 in Cu regulation is attenuated.

Recombinant hGrx1, Atox1 and WLN5-6 proteins were overexpressed and purified (Fig. S1). All purified proteins, including Atox1 and WLN5-6, in addition to unlabeled and/or 15 N-and 13 C/ 15 N-labelled hGrx1 were confirmed to be isolated as apo-proteins as determined by inductively coupled plasma mass spectrometry (ICP-MS) and a colorimetric assay with Bcs. Purified apo-proteins were loaded with Cu(I) as previously described 10 and after size exclusion chromatography (SEC) to remove excess Cu(I), were analyzed for Cu(I) content. As expected, the Atox1 and hGrx1 proteins were found to bind Cu(I) with a Cu(I):protein stoichiometry of 1:1 while WLN5-6, which has two MBDs, bound Cu(I) with a stoichiometry of 2:1 (Table S1). The apo-and Cu(I)-bound proteins prepared in this way were used for both the copper-exchange and nuclear magnetic resonance (NMR) experiments.
To determine the binding affinities of the individual proteins for Cu(I), purified apo-proteins at various concentrations were added individually to a reaction mixture containing the probe complex [Cu I Bcs 2 ] 3− . These experiments yielded dissociation constants (K DCu(I) ) for the hGrx1, Atox1, and WLN5-6 proteins of 10 −15.8 , 10 −17.5 and 10 −17.8 M at pH 7.0, respectively (Table S2). These determinations agree closely with those previously reported 10,44 (Table S2).

The hGrx1, Atox1 and WLN5-6 proteins participate in Cu(I)-transfer with apo-protein partners.
To establish whether the hGrx1 and WLN5-6 proteins could interact to facilitate Cu exchange, we performed protein co-incubation assays followed by SEC-ICP-MS. The individual apo-and Cu(I)-proteins were incubated together before separation by SEC. The column eluent was simultaneously monitored for the presence of protein (A 280 ) and for Cu (ICP-MS) ( Fig. 1A-C). Co-incubation of the Cu(I)-hGrx1 and apo-WLN5-6 proteins yielded a SEC profile with two major protein peaks where the first peak (WLN5-6), co-eluted with Cu while the second peak (hGrx1) was essentially Cu-free (Fig. 1D), indicating Cu-transfer from Cu(I)-hGrx1 to apo-WLN5-6 to yield apo-hGrx1 and Cu(I)-WLN5-6. The reverse experiment (apo-hGrx1 with Cu(I)-WLN5-6) did not result in Cu exchange, indicating the former transfer was irreversible (Fig. 1E). In addition, through separation of the proteins by anion exchange and colorimetric analyses of the Cu(I) content of the eluents, we were able to reproduce the results of previous studies that showed irreversible Cu exchange between Cu(I)-hGrx1 and apo-Atox1 and no exchange between Cu(I)-Atox1 and apo-WLN5-6 or Cu(I)-WLN5-6 and apo-Atox1 (Figs. S2, S3) 10,21 .
This demonstrates that Cu(I)-hGrx1 is able to transfer Cu to both apo-protein partners Atox1 and WLN5-6. However, Cu was not observed to transfer from Cu(I)-Atox1 and Cu(I)-WLN5-6 to apo-hGrx1. This observation correlates with the significantly weaker Cu binding affinity of hGrx1 compared with the Atox1 and WLN5-6 proteins (Table S2) 1 . Crucially, the ability of Cu(I)-hGrx1 to transfer Cu to WLN5-6 has not been reported previously and augments the possible roles for hGrx1 in intracellular Cu homeostasis beyond Cu exchange with Atox1 10 .
The failure of the Atox1 and WLN5-6 proteins to participate in Cu-exchange in both directions is consistent with previously published studies and has been attributed to the inability of the proteins to interact productively for Cu transfer 21 . However, the observation of Cu transfer from Cu(I)-Atox1 to individual MDBs of ATP7B appears to vary depending on the experimental conditions 21,32,34,54 . Specifically, the WLN5 and WLN6 domains have been shown to receive Cu from Cu(I)-Atox1, both when present as individual domains in solution 54 and as part of a multidomain (WLN1-6) protein fragment 34 . Transfer of Cu from Cu(I)-hGrx1 to WLN5-6 occurs despite the presence of Atox1. Our observations of pairwise Cu(I)-transfer between the hGrx1 and the apo-proteins Atox1 and WLN5-6 led us to extend the study to a mixture of all three of these proteins. Cu(I)-hGrx1 was incubated with a mixture of the apo-proteins Atox1 and WLN5-6 (at a molar ratio of 1:1:1), followed by re-separation by SEC-ICP-MS. Remarkably, we observed via SEC-ICP-MS that the majority of the Cu co-eluted with the WLN5-6 protein, rather than the hGrx1 and Atox1 proteins which, despite optimization of the conditions of the experiment, co-eluted as a single peak (Fig. 1F). The Cu(I) distribution was estimated (through the Cu(I) contents of the corresponding eluents, that is, the relative areas under the peaks in the elution profile, Fig. 1F) at 9:1 WLN5-6:hGrx1/Atox1. This was confirmed by separation of the proteins by anion exchange and colorimetric analyses (Fig. S2, where the hGrx1 and Atox1 proteins were resolved). These analyses indicate that on incubation, Cu transferred from Cu(I)-hGrx1 to the WLN5-6 protein preferentially (yielding Cu(I)-WLN5-6), despite the presence of equimolar Atox1 in the mixture. This result agrees with our observation that both the Atox1 and WLN5-6 apo-proteins can receive Cu from Cu(I)-hGrx1 (Figs. 1D, S2, S3). Importantly, since Cu was not observed to transfer from Cu(I)-Atox1 to apo-hGrx1 or between Atox1 and WLN5-6, the predominance of Cu(I)-WLN5-6 as a product of this experiment, indicates direct, preferential Cu-transfer from Cu(I)-hGrx1 to apo-WLN5-6, rather than equilibration of Cu between the proteins in the mixture. That is, the distribution of Cu between the WLN5-6 and Atox1/Grx1 peaks cannot be accounted for by consideration of the Cu(I)-binding affinities of the Atox1 and WLN5-6 proteins alone, which indicate approximate two-fold tighter binding of Cu(I) to WLN5-6 versus Atox1 (Table S2). We therefore hypothesized that the Cu(I)-binding affinities of the individual proteins are not the only factors that determine Cu delivery from Cu(I)-hGrx1 when multiple protein partners are present and that the affinities of the protein-protein interactions between these proteins contribute to Cu transfer. We therefore sought to examine more closely the protein-protein interactions between the Cu(I)-hGrx1 and the Atox1 and WLN5-6 proteins.
Confirmation that hGrx1 binds Cu(I) via two surface-exposed Cys residues. To confirm the location of the Cu(I) binding site in hGrx1 (between residues C23 and C26) 10 , we produced 15 N and 13 C/ 15 N-labeled apo-hGrx1 and Cu(I)-hGrx1 for NMR studies. Apo-hGrx1 gave a high quality 15 N-1 H heteronuclear single quantum coherence (HSQC) spectrum with sharp and well dispersed peaks consistent with a well-folded and (2020) 10:4157 | https://doi.org/10.1038/s41598-020-60953-z www.nature.com/scientificreports www.nature.com/scientificreports/ monomeric protein. Out of 115 peaks expected in the 15 N-1 H-HSQC spectrum from the 106-residue construct of apo-hGrx1 (excluding non-native residues from cleavage at the N-terminus), ~108 peaks were observed. Using standard triple resonance experiments, near complete HN-N, Cα and Cβ assignments of the apo-hGrx1 spectrum were made (96%, 95% and 95%, respectively), however peaks could not be assigned to residues T22, A50, T51, N52 and H53, suggesting this region undergoes conformational exchange (Fig. S4). In addition, Y25 and C26 presented as relatively weak peaks in most spectra. Conformational exchange for residues located at or near the C23-XX-C26 active site, was also reported in the NMR structural analysis of reduced hGrx1 45 . One peak in the apo-hGrx1 15 N-1 H-HSQC spectrum (124.8/7.132 ppm) was not assigned as no corresponding signals could be observed in any of the triple resonance experiments. The assigned chemical shifts have been deposited into Biological Magnetic Resonance Bank (BMRB; accession number 27650).
A 15 N-1 H-HSQC spectrum of Cu(I)-hGrx1 was recorded and a comparison of the apo-hGrx1 and Cu(I)-hGrx1 spectra revealed that a subset of peaks (F18, I19, K20, C23, Y25, C26, I48, A67, T69, S84 and D85) showed significant chemical shift positional or signal intensity changes ( Fig. S5A-C). These residues cluster in regions that encircle residues C23 and C26 ( Fig. S5A-C), confirming these two cysteine residues are directly involved in the coordination of Cu(I). For example, the peak assigned to residue C23 showed one of the most significant positional changes, while the weak but observable peaks assigned to residues Y25 and C26 in the apo-hGrx1 spectrum disappeared completely in the Cu(I)-hGrx1 spectrum ( Fig. S5A-C). In addition, the peak assigned to residue T69, located ~8Å from C23, shifted significantly and changed in its intensity. This residue has been demonstrated to be sensitive to the redox status and the structure of the active site including being displaced in structures minimized with the C23 thiolate group instead of a thiol 45,55 . Overall, this spectral comparison locates the Cu(I) binding site between residues C23 and C26, which agrees with previous mutagenesis and Cu binding studies of hGrx1 10 . At higher concentrations of partner proteins, the observed spectral changes continued to progress, such that spectra recorded at partner:Cu(I)-hGrx1 ratios of 3.8:1 and 7.1:1 (partner:Cu(I)-hGrx1; for Atox1 and WLN5-6, respectively) showed additional spectral changes in comparison to the spectrum of apo-hGrx1. These observations may be ascribed to interactions between newly formed apo-hGrx1 and the partner proteins. That is, the formation of weak and transient apo-hGrx1-Cu(I)-Atox1 and apo-hGrx1-Cu(I)-WLN5-6 and/or apo-hGrx1-apo-Atox1 and apo-hGrx1-apo-WLN5-6 complexes. This indicates that apo-hGrx1-protein partner complexes persisted following Cu transfer. The spectral changes for these titrations, in terms of elucidating the molecular details and thermodynamics of these transient protein-protein complexes were challenging to decipher as they resulted from multiple hGrx1 species (apo-hGrx1, Cu(I)-hGrx1 and Cu(I)-hGrx1-partner, apo-hGrx1-Cu(I)-partner and apo-hGrx1-apo-partner protein complexes) that were present at varying concentrations throughout the titrations. Therefore, we conducted NMR titration studies with apo-15 N-hGrx1 and the apo-proteins Atox1 and WLN5-6 in order to simplify the analyses and to directly measure the affinities of the protein-protein interactions.
The same interaction surface facilitates Cu transfer from hGrx1 to Atox1 and WLN5-6. Apo-15 N-hGrx1 was titrated with apo-proteins Atox1 or WLN5-6 to final molar ratios of 2.0:1 (partner:hGrx1) (Figs. S8A-C, S9A-C). In these titrations, variations were observed from the same set of peaks that showed changes at the highest partner:hGrx1 ratios in the Cu(I)-hGrx1 titrations. These included peaks assigned to residues at and surrounding the C23-XX-C26 Cu binding site, which map to a common region on the surface of hGrx1. This mapped region is consistent for both the Atox1 and WLN5-6 titrations and indicates that hGrx1 interacts with both protein partners using the same interaction surface (Fig. 3A,B). Calculation of the electrostatic surface potential of the hGrx1 structure shows that this interaction surface is predominantly positively charged (Fig. 3C). This aligns with previous studies of metallochaperone P 1B -ATPase MBD interactions, which have reported that the complexes are mediated by complementary surface electrostatics 13,52,56 . The observed peak changes were fitted to a 1:1 binding model, which yielded K D s of 14 ± 6 μM and 7 ± 4 μM for the hGrx1-Atox1 and hGrx1-WLN5-6 interactions, respectively (Fig. S10A,B). These values are comparable to those determined for the protein-protein interactions between electron transfer partners 57 . Both metal and electron exchange require specific, yet transient protein-protein interactions. Given the same hGrx1 surface is involved in these protein-protein interactions, the difference in observed K D s for the hGrx1-Atox1 and hGrx1-WLN5-6 pairs is presumably due to differences in the surface structures of the Atox1 and WLN5-6 proteins (Fig. S11) 21,41,58,59 . Crucially, these observations indicate that protein-protein interactions between hGrx1 and the Atox1 and WLN5-6 proteins occur in the absence of Cu and that the affinities of these interactions are different between distinct pairs of protein partners.

Conclusion
In this study, we established that hGrx1 binds Cu in a 1:1 stoichiometry at a C23-XX-C26 site and that hGrx1 can deliver Cu to the apo-proteins Atox1 and WLN5-6 in solution. Importantly, Cu transfer from hGrx1 to WLN5-6 occurs preferentially in the presence of Atox1 and the interactions of hGrx1 with the Atox1 and WLN5-6 are not Cu dependent. The Cu-dependence of protein interactions that mediate intracellular Cu shuttling has been a controversial area of research. A number of studies have repeatedly stated that the interactions between proteins involved in Cu exchange are metal dependent. For example, examinations via NMR of Cu(I) delivery from Cu(I)-Atox1 to the multidomain protein ATP7B MBDs 1-6 34 , the determination of the structures of the Atx1-Ccc2a 52 and complexes between Atox1 and the MBD1 of ATP7A from yeast and human, respectively 53 , all described the observed intermolecular adducts as metal-mediated. This is despite the fact that extensive intermolecular interfaces between the proteins in the complexes have been defined 52,53 , that the surface electrostatics of the proteins have been shown to be crucial for complex formation 53,56 and that only minimal structural differences have been observed between apo-and Cu(I)-forms of these proteins and domains 37,60 . However, NMR experiments have also demonstrated interactions between MBDs 4-6 of ATP7B and Atox1 in the absence of Cu 27 . The results reported here give further support to the proposal that these interactions occur independently of the presence of Cu.
In a mixture of the Cu(I)-hGrx1, apo-Atox1, and apo-WLN5-6 proteins, Cu was preferentially delivered to the WLN5-6 protein, despite the fact that the apo-Atox1 and apo-WLN5-6 proteins showed only a two-fold difference in their respective K DCu(I) values. This result can be reconciled by both the thermodynamics of the interactions between the proteins pairs and of the proteins with Cu(I). That is, the hGrx1 protein interacts with a higher affinity with the WLN5-6 protein than with the Atox1 and the binding affinity of WLN5-6 for Cu(I) is also higher. In fact, a calculation taking into account the respective values of K DCu(I) for each protein, the K D s for the hGrx1-Atox1 and hGrx1-WLN5-6 interactions and the fact that the WLN5-6 protein binds two equivalents of Cu, predicts a Cu distribution for this experiment of approximately 0.6%, 87.6%, 11.8% for the hGrx1, WLN5-6 and Atox1 proteins respectively, which correlates with our analysis. Interestingly, both proteins (Atox1 and WLN5-6) interact at the same interface with hGrx1, which is proximate to the Cu(I)-binding site. The predicted consequence of the different affinities of these proteins for hGrx1 is competition for binding at this common site and therefore competition for Cu exchange. Importantly, the outcome of this 'competition' in the cell, would not only be determined by the relative affinities of each protein for Cu(I) and for each other, but also the abundances of each protein 61 .
Crucially, we show here that hGrx1 is able to transfer Cu to the WLN5-6 protein, even in the presence of Atox1. To date, Atox1 has been the sole metallochaperone protein proposed to act in Cu delivery to the ATP7A and ATP7B proteins. The fact that hGrx1 can act in this way is significant for intracellular Cu homeostasis. This finding suggests a potential redundancy in this system and a role for hGrx1 under cellular conditions where the activity of Atox1 in Cu regulation is attenuated. For example, a number of lines of evidence suggest that both the cellular ratios of apo-Atox1/Cu(I)-Atox1 and reduced/oxidized Atox1 influence metabolic Cu flux and specifically the activity and trafficking of the Cu ATPases 37,38,62 . There is also evidence to suggest that Atox1 is not absolutely required for Cu delivery to the Cu(I)-ATPases 59,63 and that other Cu carriers may supplement Atox1 function. For example, ATOX1 knockout does not completely abrogate ATP7A function, suggesting that ATP7A may obtain Cu from alternative metal donor(s) 59,63 . The data presented in this study suggests that hGrx1 could assume such a role.

Methods
Protein overexpression and purification. The DNA sequence encoding hGrx1 was amplified via PCR and subcloned into a pGEX-6P-1 glutathione S-transferase fusion vector with an intervening PreScission Protease site for cleavage of the GST tag. The DNA sequences encoding Atox1 and the WLN5-6 protein fragment (encoding residues 486-633 of ATP7B) were amplified via PCR and subcloned into the pTEM-11 and pET-24d vectors, respectively. The pGEX-6p-1-hGrx1, pTEM-11-Atox1 and pET-24d-WLN5-6 plasmids were individually transformed into Escherichia coli strain BL21 Codon Plus (DE3). Cultures were grown in Luria Broth (LB) at 37 °C www.nature.com/scientificreports www.nature.com/scientificreports/ to an optical density OD 600 of approximately 0.8, induced with isopropyl β-D-1-thiogalactopyranoside (IPTG, 0.5 mM) and harvested after 16 h with shaking at 25 °C. The hGrx1 isotope labeled samples ( 15 N and/or 13 C) were grown in LB at 37 °C to an OD 600 of 0.6-0.8, harvested and washed twice in M9 salts before transfer to labeled M9 media and induction with IPTG (1.0 mM). Cells were harvested by centrifugation after 16 h with shaking at 25 °C 64 .
The purifications of the Atox1 and WLN5-6 proteins were conducted according to previously reported protocols with minor modifications 21,44  Cu-exchange. Protein samples (2 μg) of Cu(I)-loaded, apo-proteins and protein mixtures at 1:1 molar ratios were applied to a Bio-SEC. 3 column (3 mm particle size; 150 Å pore structure; 4.6 mm i.d., Agilent Technologies) in 200 mM ammonium nitrate (pH 7.6−7.8, adjusted with ammonium hydroxide) with 10 μg/L cesium (Cs) and antimony (Sb) as internal standards. Chromatographic separations of the proteins were monitored by measuring the UV absorbance of the eluent at 280 nm (indicating the presence of protein) and inductively coupled plasma mass spectrometry (ICP-MS) (Agilent Technologies 7700 x ICP-MS) 68 was used to monitor the presence of Cu.
For the analyses of protein mixtures that contained the hGrx1 and Atox1 proteins (which could not be resolved by SEC), separation of the proteins by anion exchange was carried out. Protein samples were applied to a mono Q 5/50 GL column (GE Healthcare) pre-equilibrated with binding buffer (20 mM Tris-MES, pH 8.0) and bound proteins were eluted by the application of a linear NaCl gradient (0.0-1.0 M NaCl, 20 mM Tris-MES, pH 8.0). The Cu contents of the eluted fractions were determined colorimetrically with Bcs 10 . NMR spectroscopy. 15 N-1 H heteronuclear single quantum coherence (HSQC) spectra were recorded at 4 °C using a Bruker Avance III 600 MHz NMR spectrometer equipped with a triple-resonance TCI cryogenic probehead. All stock protein solutions were dialyzed against MES/Na buffer (10 mM MES/Na, 50 mM NaCl, 1 mM TCEP, pH 6.0) overnight prior to titration studies. 15 N-1 H-HSQC spectra of 15 N-apo and 15 N-Cu(I)-hGrx1 (350-400 μl at 320-350 μM) were recorded after sequential additions of unlabeled apo forms of the proteins Atox1 or WLN5-6. For the titrations into 15 N-apo-hGrx1, the apo-proteins Atox1 or WLN5-6 (0.46 and 0.5 mM, respectively) were titrated to final molar ratios of 2.0:1 (partner:apo hGrx1). For titrations into 15 N-Cu(I)-hGrx1, final molar ratios of 3.8:1 and 7.1:1 (partner:hGrx1) were obtained upon sequential additions of the apo-proteins Atox1 or WLN5-6 (0.5 and 1.5 mM, respectively). Higher molar ratios lead to the observation of some protein precipitation and deterioration in the quality of the spectra recorded. D 2 O was added to each sample to a final concentration of 5% (v/v). Spectra were processed with Topspin 3.5 (Bruker Biospin) and analyzed using SPARKY (T. D. Goddard and D. G. Kneller, University of California, San Francisco). Spectral changes (either in intensity or peak position) were plotted against varying concentrations of the titrants. For peak positional changes, combined chemical shift perturbations, Δ chemical shift (ppm) were calculated based on the equation Δδ ppm = [(Δδ HN ) 2 + (0.154 × Δδ N ) 2 ] 1/2 where Δδ HN and Δδ N represent the chemical shift variations in the proton and nitrogen dimensions, respectively. Comparing apo and Cu(I)-hGrx1 spectra, peaks that were not overlapped and with positions changing by more than one standard deviation (SD) over the mean or those with intensity www.nature.com/scientificreports www.nature.com/scientificreports/ decreasing by >75% or increasing by >100% were deemed to be significantly affected. For titration with partner proteins, peaks with positions changed by more than one standard deviation (SD) over the mean or those with intensity changes >50% at molar ratios beyond 1:1 were deemed to be significantly affected. Resonances corresponding to a number of residues were initially evaluated. The peaks corresponding to Cys26 and Thr69 showed the clearest and most significant changes amongst those analyzed and therefore these were used for the affinity calculations. Binding affinities were determined from non-linear-least-square curve fitting to a 1:1 binding model based on intensity changes using Origin2016.
NMR assignment of apo-hGrx1. hGrx1 assignments were unavailable from the previously determined apo-hGrx1 structure (PDB code (1JHB) 45 ), so triple resonance experiments were recorded on a purified 13 C/ 15 N-labeled hGrx1 sample (0.5 mM) to allow the assignment of the 15 N-1 H-HSQC spectrum. Spectra were analyzed with Sparky (T. D. Goddard and D. G. Kneller, University of California, San Francisco). Backbone 15 N, H N , 13 C and 13 C resonances were obtained from HNCACB, CBCA(CO)NH, HCC(CO)NH and CC(CO) NH experiments using standard methods. 15 N and 13 C chemical shifts were referenced indirectly using 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS) according to their magnetogyric ratios.