The crystal structure of the heme d 1 biosynthesis-associated small c-type cytochrome NirC reveals mixed oligomeric states in crystallo

The crystal structure of the c-type cytochrome NirC from Pseudomonas aeruginosa has been determined and reveals the simultaneous presence of monomers and 3D domain-swapped dimers in the same asymmetric unit.


Introduction
Monoheme c-type cytochromes are a subfamily of c-type cytochrome proteins and are well known electron transporters in the respiratory chain of organisms in all domains of life. In bacteria, they are also involved in H 2 O 2 scavenging and act as an electron-entry point for nitrate, nitrite and nitric oxide reductases, which function as terminal oxidases in the respiratory chain under anaerobic conditions (denitrification; Bertini et al., 2006). The monoheme c-type cytochrome domain features three -helices that surround one c-type heme covalently attached to the cysteines of the characteristic sequence motif CXXCH via thioether bonds. The central iron cation of heme is usually coordinated by a histidine residue at its proximal site, whereas a histidine or, more rarely, a methionine ligates the other side in most cases. Many monoheme c-type cytochromes can form dimers and oligomers by exchanging the N-or C-terminal -helix, which offers attractive protein-engineering opportunities (Hirota et al., 2010;Hayashi et al., 2012;Nagao et al., 2015). We have recently embarked on a structural biology program to investigate the enzymes involved in the biosynthesis of heme d 1 , an essential cofactor of the cd 1 nitrite reductase NirS, which catalyses the reduction of nitrite to nitric oxide in, for example, the opportunistic pathogen Pseudomonas aeruginosa (Schobert & Jahn, 2010). The nir operon of P. aeruginosa, which encodes most of the enzymes involved in heme d 1 biosynthesis in addition to nirS itself, also contains open reading frames for two monoheme c-type cytochromes termed nirM and nirC ( Fig. 1; Arai et al., 1990;Kawasaki et al., 1995). Whereas NirM is absent in many NirS-producing bacteria and has been assigned to function as an electron transporter for the nitrite reductase (Hasegawa et al., 2001), orthologs of NirC are found in nearly all NirS producers, but their function is subject to ongoing discussions. It has been demonstrated that NirC from P. aeruginosa can act as an electron donor for the nitrite reductase in vitro (Hasegawa et al., 2001). However, the redox potential of NirC from Paracoccus pantothrophus has been found not to be optimal for the transport of electrons from the bc 1 complex to nitrite reductase (Hasegawa et al., 2001;Zajicek et al., 2009;). A double knockout of the two bettersuited electron transporters cytochrome c 550 and pseudoazurin rendered P. denitrificans unable to denitrify despite the presence of NirC (Pearson et al., 2003). Further, mutants of P. denitrificans and P. fluorescens that are impaired in their ability to express nirC are unable to produce active nitrite reductase owing to a lack of heme d 1 . Together, these findings implicate the role of NirC as lying in heme d 1 biosynthesis (Ye et al., 1992;Boer et al., 1994); however, this could not be confirmed in P. aeruginosa (Hasegawa et al., 2001). To further investigate NirC, we have determined the crystal structure of heterologously produced P. aeruginosa NirC. We find a unique crystal packing that contains monomeric NirC together with two different dimeric forms. To our knowledge, this is the first time that monomers and dimers of a monoheme c-type cytochrome have simultaneously been observed in the same asymmetric unit.

Macromolecule production
The gene encoding NirC was amplified by polymerase chain reaction from genomic DNA extracted from P. aeruginosa PA14, omitting the signal peptide for periplasmic export as determined by SignalP5.0 (Almagro Armenteros et al., 2019). The fragment was cloned between the NotI/KpnI restriction sites of a modified pCOLADuet plasmid (pVP008-ompA), yielding an expression vector that produces NirC with an N-terminal OmpA signal peptide for periplasmic export in Escherichia coli followed by a Tobacco etch virus (TEV) protease-cleavable StrepII-tag. Residue Glu71 was mutated to alanine by QuikChange Mutagenesis for the reasons outlined below. The success of cloning and mutagenesis was verified by the Eurofins Genomics sequencing service.
E. coli C43(DE3) cells (Miroux & Walker, 1996) were cotransformed with the NirC expression plasmid and with pEC86, a plasmid containing the cytochrome c maturation system of E. coli, which is necessary for the covalent attachment of heme to a suitable polypetide chain (Arslan et al., 1998). LB medium (8 Â 1 l) supplemented with 34 mg l À1 chloramphenicol, 30 mg l À1 kanamycin, 200 mM -aminolevulinic acid and 12.5 mM iron(II) sulfate was inoculated with 10 ml overnight culture and cultivated at 37 C with mild shaking until an optical density at 600 nm of between 0.6 and 0.8 was reached. The cultures were then induced with 1 mM isopropyl -d-1-thiogalactopyranoside and further incubated at 20 C for 20 h. The cells were harvested by centrifugation and the pellets were stored at À20 C until needed.
The cell pellets were thawed, resuspended in lysis buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl) supplemented with protease inhibitors (cOmplete mini EDTA-free) and lysed by sonication. The lysate was centrifuged for 30 min at 36 000g followed by an additional centrifugation of the supernatant for 60 min at 100 000g to remove cell debris.
The cleared cell lysate was loaded onto a StrepTactin HC column (IBA, Gö ttingen, Germany) equilibrated with protein buffer (10 mM Tris-HCl pH 8.0, 150 mM NaCl) and connected to an Ä KTApurifier FPLC system (GE Healthcare, Boston, USA). The column was washed with five column volumes and subsequently eluted with three column volumes of protein buffer containing 5 mM d-desthiobiotin.  Depiction of the nir operon, indicating the function of each gene product in nitrite reduction and heme d 1 biosynthesis. A grey background highlights the periplasmic location of the gene product. elution fractions and UV-Vis absorption at 410 nm confirmed the presence of heme. Fractions containing no impurities as judged by SDS-PAGE analysis were pooled and digested with TEV protease at 4 C overnight with simultaneous dialysis against protein buffer. To remove undigested protein, the resulting solution was again loaded onto the StrepTactin column and the flowthrough was collected. As a final purification step, NirC was subjected to size-exclusion chromatography using a HiLoad 16/60 Superdex 75 column (GE Healthcare, Boston, USA). The pure protein was concentrated to 40 mg ml À1 with Vivaspin 20 (molecular-mass cutoff 3000) ultrafiltration units (Sartorius AG, Gö ttingen, Germany) and the concentration was determined spectrophotometrically using a NanoDrop 2000 (Thermo Fisher Scientific, Waltham, USA) and a calculated extinction coefficient of " 280 = 11 mM À1 cm À1 . Macromolecule-production information is summarized in Table 1.

Crystallization
Crystallization experiments were set up using a HoneyBee 961 pipetting robot (Digilab Genomic Solutions, Hopkinton, USA) and were monitored with a Rock Imager 1000 automated microscope (Formulatrix, Bedford, USA). Optimization screens were prepared utilizing a Formulator pipetting robot (Formulatrix, Bedford, USA). Experiments to crystallize wild-type NirC never yielded well diffracting crystals. Therefore, a mutation (E71A) suggested by the surfaceentropy reduction server (Goldschmidt et al., 2007) was introduced, leading to crystal growth in a few conditions of The Cryos Suite sparse-matrix screen (Qiagen, Hilden, Germany). The respective precipitants were further optimized by varying the concentration of precipitant and salts as well as the pH. Crystals usually began to appear after one week and took three more weeks to grow to full size (50-100 mm; Supplementary Fig. S1). The crystals were harvested and directly flash-cooled in liquid nitrogen without the addition of cryoprotectants. Crystallization information is summarized in Table 2.

Data collection and processing
Data were collected on beamline P11 of the PETRA III synchrotron, DESY, Hamburg, Germany (Burkhardt et al., 2016). The anomalous data set was processed using DIALS (Winter et al., 2018), POINTLESS (Evans, 2011) and AIMLESS (Evans & Murshudov, 2013) from the CCP4 suite . Despite the low completeness in the highest resolution shell, the data set was sufficient to obtain initial phases via SAD phasing. The data set used for final refinement was processed using autoPROC (Vonrhein et al., 2011) executing XDS (Kabsch, 2010), POINTLESS (Evans, 2011), AIMLESS (Evans & Murshudov, 2013) and STAR-ANISO (Tickle et al., 2018). Anisotropic data truncation was applied assuming a local I/(I) of 1.2 to include useable data beyond the spherical resolution of 2.83 Å . The extent of the anisotropy is visualized in Supplementary Fig. S5. Datacollection and processing statistics are summarized in Table 3.

Structure solution and refinement
The structure was solved by SAD phasing with anomalous differences arising from the heme Fe atoms using the CRANK2 phasing pipeline ( Table 1 Macromolecule-production information.
Source organism P. aeruginosa Cloning primer †

GCCTGCTCAGCGCAGACGACGCCGG CCGGCGTCGTCTGCGCTGAGCAGGC
Cloning vector pVP008-ompA modified from pCOLADuet Expression vector pVP008-ompA-NirC-E71A Expression host E. coli C43(DE3) (Miroux & Walker, 1996) Complete amino-acid sequence of the recombinant protein ‡ MKKTAIAIAVALAGFATVAQAMASWSHPQFEKVDENLYFQGGGRDEHPDARRQAQLRHLLLQDCGSCHGLRLTG GLGPALTPEALRGKPRESLVATVLMGRPQTPMPPWAGLLSADDAGWLVDRLIEGEIAP † The NotI/KpnI restriction sites are underlined. ‡ The TEV protease cleavage site is underlined, the StrepII-tag is in bold, the OmpA leader peptide is in italics and the E71A mutation is in bold and underlined. Final refinement was performed against the higher resolution native data set using phenix.refine (Afonine et al., 2012) while lifting the applied NCS restraints, adding water molecules and riding H atoms, and applying TLS refinement. Depictions of the structure were made with PyMOL (version 1.8; Schrö dinger). Refinement statistics are summarized in Table 4.

Size-exclusion chromatography coupled to a multi-angle laser light detector
To assess the molecular mass of NirC, the isolated and concentrated protein solution (40 mg ml À1 ) was subjected to SEC-MALS experiments utilizing an Agilent Technologies 1260 Infinity II HPLC system (Santa Clara, USA) equipped with a Wyatt Optilab rEX diffraction-index detector and a mini-DAWN TREOS II multi-angle laser light-scattering detector (Wyatt, Santa Barbara, USA). Separation was achieved with a Superdex 75 Increase 10/300 GL column (GE Healthcare, Boston, USA) with protein buffer as the eluent. To calculate the molecular mass, a dn/dc of 0.195 ml g À1 was assumed.

Structural bioinformatics
Functional orthologs of NirC were identified by extracting orthologs from the OMA database (Altenhoff et al., 2018) followed by manually filtering for the presence of the genes encoding the nitrite reductase NirS and the heme d 1biosynthesis enzyme NirF in their genetic neighbourhood. After aligning the functional orthologs with Clustal Omega (Ashkenazy et al., 2016), sequence conservation was mapped onto the NirC monomer structure utilizing the ConSurf server ( Fig. 3 and Supplementary Fig. S2; Sievers et al., 2011). Structural homologs were identified with the DALI server (Holm & Laakso, 2016).

Results and discussion
3.1. The simultaneous presence of NirC monomers and dimers in the asymmetric unit Approximately 2 mg of pure heme-bound NirC per litre of E. coli culture could reliably be obtained using the procedure described above. Size-exclusion chromatography as well as multi-angle light scattering suggested that the concentrated protein solution used in crystallization trials only contains monomers with a molecular mass of 11.5 kDa AE 0.9% (10.5 kDa expected; Fig. 2). Because the wild-type protein failed to crystallize, we attempted to reduce the surface entropy by mutating Glu71 to alanine, finally yielding cube-   shaped crystals that diffracted anisotropically and possessed a very large orthorhombic unit cell ( Supplementary Fig. S1). To obtain the initial phases, we made use of the presence of anomalous differences in diffraction data collected at the Fe K edge, leading to the positioning of 11 Fe atoms. Assuming that NirC contains one Fe atom bound to the covalently attached heme, this suggests a solvent content of 58%. A combination of automated model building and manual adjustments led to a final model with R work = 21.1% and R free = 24.7% that indeed consisted of 11 chains in the asymmetric unit. Interestingly, despite the use of monomeric NirC for crystallization, ten of these chains form dimers by exchanging their N-terminal -helices, as recognizable by different traces in the 2F o À F c density for the 'hinge loop' (Gly25-Gly33; Fig. 5). The dimers fall into two different types, as outlined below. Surprisingly, the 11th chain is monomeric, i.e. the crystal form obtained here contains both dimers and monomers of NirC at the same time. Analysis of the crystal packing reveals that the surfaceentropy-reducing mutation E71A enables packing interactions that are not possible with the glutamate side chain. An example is shown in Supplementary Fig. S6. Importantly, the mutation is not part of the dimer interfaces or the hinge loop,  suggesting that it does not cause the simultaneous presence of different NirC oligomeric states observed here. The formation of dimers from monomeric c-type cytochromes is a well known phenomenon (Hirota, 2019). It has been shown that the folding of horse cytochrome c starts with hydrophobic interactions between the N-and C-terminal helices, yielding 3D domain-swapped oligomers if the helices belong to different chains (Parui et al., 2013). Therefore, denaturation and subsequent refolding at high protein concentration is often utilized to produce such dimers (Hayashi et al., 2012;Nagao et al., 2015). Apparently, the high NirC concentration and the presence of precipitants in the crystallization experiments performed here are sufficient to allow an equilibrium between dimers and monomers.
Whereas 3D domain swapping is often observed in crystal structures (reviewed in Liu & Eisenberg, 2002), the simultaneous presence of different homooligomeric states of one protein in the same asymmetric unit is extremely rare and we are aware of only two other unambiguous examples, namely the FeSII protein from Azotobacter vinelandii (PDB entry 5frt; B. V. Kabasakal, C. A. R. Cotton, L. Lieber & J. W. Murray, unpublished work) and an engineered tenascin protein (PDB entry 2rbl; Hu et al., 2007). In both cases the asymmetric unit was found to contain monomers and 3D domain-swapped dimers, similar to the crystal structure of NirC reported here. In addition, there may be less obvious instances that do not involve 3D domain swapping, such as a non-oligomerizing variant of the auxin response factor from Arabidopsis thaliana (PDB entry 4nj7). This protein crystal-lized with 16 chains in the asymmetric unit, which may be regarded as dimers and pentamers (Korasick et al., 2014), although it also seems possible to explain the packing by assuming only monomers.
3.2. NirC shows high similarity to other monoheme c-type cytochrome domains encoded by the nir operon NirC displays a typical monoheme c-type cytochrome fold containing three -helices (1, 3 and 4) that surround a thioether-bound heme c moiety (Fig. 3). The iron cation is coordinated by His24 and Met62, which is consistent with mutagenesis studies performed using the orthologue from P. pantotrophus (Zajicek et al., 2009). The heme-binding crevice consists of a predominantly hydrophobic pocket established by Leu32, Pro34, Leu36, Thr52, Val53, Thr60, Pro61 and Trp65, which are also highly conserved in NirC proteins from other species. In P. aeruginosa NirC, the propionate groups of heme form hydrogen bonds to Lys44 and Arg57, two residues that are replaced by hydrophobic amino acids in some species, probably reflecting the fact that the heme propionate groups are solvent-exposed and do not require stabilizing interactions with the protein to be accommodated. Sequence alignment reveals that in NirC the hemeattachment motif (CXXCH) is always preceded by a conserved aspartate (Asp19), the side chain of which is solvent-exposed, and followed by a conserved GGLG motif residing on a loop connecting helices 2 and 3 ( Supplementary Fig. S3). Together with residues in the neighbourhood of the iron-ligating The NirC monomer shown as a cartoon with residues at a maximum distance of 4 Å from the covalently attached heme presented as sticks. Colours are based on sequence conservation as defined by ConSurf (Ashkenazy et al., 2016). The sequence alignment used to determine conservation is shown in Supplementary Fig. S3.
Met62, these motifs create a conserved surface patch, which may be required to form an interaction site for other proteins such as, for example, the heme d 1 -biosynthesis protein NirF, the function of which is not clear but may lie in an oxidoreduction within the last steps of the pathway (Klü nemann et al., 2020). Alternative hypotheses suggest involvement as an electron transporter of the nitrite reductase NirS (Hasegawa et al., 2001). A role in heme d 1 biosynthesis itself, on the other hand, seems to be supported by the finding that similarity searches with DALI (Holm & Laakso, 2016) identify the cytochrome c domains of the nir-operon proteins nitrite reductase (NirS) and the heme d 1 -biosynthesis enzyme NirN as most similar to NirC, while NirM and cytochrome c 552 , the known electron transporters that interact with NirS in nitrite reduction, differ to a larger extent ( Fig. 4 and Supplementary Table S1). In NirN and NirS, the cytochrome c domains are covalently attached to a -propeller domain and function as electron shuttles in the reactions catalysed by these enzymes. This suggests that NirC may perform a similar function with NirF, which is a standalone -propeller protein with high similarity to thepropeller domains of NirN and NirS. Interestingly, superposition of NirC and NirF onto the respective domains of NirS leads to a complex in which the conserved surface patch of NirC is oriented towards the propeller of NirF, and Leu32 of the conserved GGLG motif is buried in a hydrophobic pocket ( Supplementary Fig. S4). This may further corroborate the involvement of NirC in heme d 1 biosynthesis.

NirC forms two different types of dimers
The dimers of NirC found here form by exchanging the N-terminal helix (Pro4-His24), which results in heme being coordinated by His24 and Met62 from two different chains, similar to what has been observed for dimers of cytochrome c 551 (NirM) from P. aeruginosa and cytochrome c 552 from Hydrogenobacter thermophilus (HT) (Hayashi et al., 2012;Nagao et al., 2015). Notably, dimerization does not have a major impact on the structure of the heme-binding site, indicating that monomers and dimers could display similar biochemical properties, as has already been found for other N-terminally domain-swapped cytochromes Hayashi et al., 2012).
Closer inspection reveals that the five dimers present in the asymmetric unit fall into two groups that are distinguishable by heme iron distance, internal symmetry and the relative position of the hinge loop (25-GLRLTGGLG-33; Fig. 5). In the three 'far dimers' the average iron-iron distance is 28.5 AE 0.2 Å , whereas it is 24.9 AE 0.1 Å in the two 'close dimers'. Interestingly, the two dimeric forms of NirC display different amounts of internal symmetry, as can be quantified by the 'GloA-score'. Naturally occurring homodimers are usually highly symmetric, with GloA-scores under 0.4 Å , while scores above 3 Å are indicative of substantial asymmetry (Swapna et  Comparison of the structure of monomeric NirC with the cytochrome c domains of NirS and NirN as well as with NirM from P. aeruginosa (PA) and cytochrome c 552 from Marinobacter hydrocarbonoclasticus (MH) after superposition of the covalently attached heme. All structures are depicted as cartoons and the covalently attached heme is shown as a ball-and-stick model. NirC is shown in blue and the compared structure is in white.
al., 2012). In the dimers of NirC observed here, the GloAscore is 0.9 AE 0.2 Å for the far dimers, compared with 3.05 AE 0.01 Å for the close dimers; this is indicative of substantial asymmetry in the latter, which is readily discernible after the superposition of one protomer of a dimer onto the other (Supplementary Fig. S2). The basis for the formation of different dimers in NirC seems to be rooted in the hinge loop, which with nine residues is much longer and allows more flexibility than the three residues found in P. aeruginosa cytochrome c 551 or HT cytochrome c 552 (Hayashi et al., 2012;Nagao et al., 2015). The importance of the length and hence the flexibility of the hinge loop is reflected in the observation that its elongation in HT cytochrome c 552 induced different Nand C-terminally domain-swapped dimers (Ren et al., 2015). In NirC, the increased flexibility leads to antiparallel -sheetlike interactions of Leu26, Arg27 and Leu28 in the far dimers and no interactions, but with elevated B factors of the hinge loop in the close dimers.
The relevance of dimer formation in NirC and the associated different levels of asymmetry are not clear at present. Previous work with eukaryotic monoheme c-type cytochromes suggested that 3D-domain-swapped oligomers are involved in apoptosis (Junedi et al., 2014), which seems to be in line with the finding that several other 3D-domain-swapped proteins have a regulatory function. Examples include bovine RNase, where only the dimer displayed toxicity towards tumour cells (Di Donato et al., 1995), and the regulatory domain of -isopropylmalate synthase from Mycobacterium tuberculosis, which forms via 3D domain swapping (Koon et al., 2004). Because dimer formation of NirC is clearly linked to the hinge loop and because this loop is highly conserved amongst NirC homologues from different species, it is possible that dimers of NirC also have a regulatory role. However, the fact that we did not detect dimers even in highly concentrated solutions, together with the observation that crystallization took relatively long and required the introduction of a surface-entropy- Depiction of the two different oligomerization states found in the crystal structure of NirC. On the left, the NirC monomer and both dimer conformations are shown as described in Fig. 1 after the superposition of one protomer. The backbone of NirC is shown as a cartoon with the covalently attached heme as a ball-and-stick model. On the right, the residues of the hinge loop are depicted as stick models with a 2F o À F c map at a level of 1, coloured according to the associated chain and individually orientated to allow an unobstructed view. reduction mutant that enabled crystal formation through new crystal contacts, currently argues for a crystallization artefact that points towards structural malleability of these proteins rather than a physiological importance of NirC dimers.
In summary, our work provides new insight into NirC, a monoheme c-type cytochrome that is likely to be involved in the biosynthesis of heme d 1 , an essential cofactor in nitrite reduction by denitrifying bacteria such as the opportunistic pathogen P. aeruginosa. While the simultaneous observation of NirC monomers and dimers in the same asymmetric unit of the crystal form obtained here may reiterate typical properties of this protein family, the mapping of conserved sequence motifs onto the structure of the monomer may offer new avenues to decipher the function of NirC.

Related literature
The following references are cited in the supporting information for this article: Brown et al. (1999), Klü nemann et al.