The P aracoccus denitrificans NarK‐like nitrate and nitrite transporters—probing nitrate uptake and nitrate/nitrite exchange mechanisms

Summary Nitrate and nitrite transport across biological membranes is often facilitated by protein transporters that are members of the major facilitator superfamily. Paracoccus denitrificans contains an unusual arrangement whereby two of these transporters, NarK1 and NarK2, are fused into a single protein, NarK, which delivers nitrate to the respiratory nitrate reductase and transfers the product, nitrite, to the periplasm. Our complementation studies, using a mutant lacking the nitrate/proton symporter NasA from the assimilatory nitrate reductase pathway, support that NarK1 functions as a nitrate/proton symporter while NarK2 is a nitrate/nitrite antiporter. Through the same experimental system, we find that Escherichia coli NarK and NarU can complement deletions in both narK and nasA in P. denitrificans, suggesting that, while these proteins are most likely nitrate/nitrite antiporters, they can also act in the net uptake of nitrate. Finally, we argue that primary sequence analysis and structural modelling do not readily explain why NasA, NarK1 and NarK2, as well as other transporters from this protein family, have such different functions, ranging from net nitrate uptake to nitrate/nitrite exchange.


Introduction
The transport of nitrate into bacterial cells is important for two processes, assimilation (where nitrate is transported to the cytoplasm and is reduced to nitrite and then to ammonium) and respiration (where nitrate can act as a terminal electron acceptor). The active sites of the assimilatory and the respiratory nitrate reductase are cytoplasmic, thus requiring nitrate import. The negatively charged nitrate has to enter the cytoplasm against a membrane potential of 180 mV (negative in the cytoplasm) which, if transport was relying on a passive nitrate entry pore, would restrict cytoplasmic nitrate concentrations to a mere 0.001% of the external concentration. Both assimilation and respiration occur at extracellular nitrate concentrations in the low micromolar range (Parsonage et al., 1985). This means that, unless cytoplasmic nitrate reduction for either process can proceed at nanomolar concentrations of substrate, uptake of nitrate must be an active process, linked to an energy source or, at least, to an alternative way of compensating for the transport of negative charge into the cell.
In the case of assimilation, there is evidence that in some bacteria nitrate uptake is driven at the expense of ATP hydrolysis by an ABC-type transport protein (Ohashi et al., 2011) while in other organisms, this process involves a member of the major facilitator superfamily (MFS) (Ogawa et al., 1995;Gates et al., 2011). A subgroup of these MFS proteins, the NarK-like transporters, has been found to be implicated in nitrate uptake during respiration as well as during assimilation (Clegg et al., 2002;Sharma et al., 2006;Goddard et al., 2008). The different fates of nitrate transported into the cell for assimilation and respiration necessitate different properties of the transporters involved. During assimilation, nitrate is reduced to nitrite and then to ammonium in the cytoplasm, in which case nitrate import is the only transport step which is assumed to be required. On the other hand, during respiration with a periplasmic nitrite reductase, nitrite produced in the cytoplasm must be transported back to the periplasm for subsequent reductions to N 2 catalysed by periplasmic proteins. For these reasons, it has been proposed that there are two classes of MFS transporters involved in nitrate transport. Some NarK transporters, termed NarK1, exhibit high similarity to proteins found within operons encoding assimilatory nitrate reductases (Moir and Wood, 2001) and, based on biochemical or physiological evidence (Wood et al., 2002;Goddard et al., 2008), these proteins are predicted to be nitrate/proton (or conceivably sodium, but we argue later that this is less likely) symporters. Conversely, NarK2-like transporters are generally associated with respiratory processes and are believed to be nitrate/nitrite antiporters (Moir and Wood, 2001;Wood et al., 2002;Jia and Cole, 2005;Goddard et al., 2008).
Despite the proposed functions for NarK-like transporters outlined above, there is still significant uncertainty regarding their exact transport mechanisms. This uncertainty has recently been enhanced, rather than diminished, by the publication of the crystal structures of two proteins of this family from Escherichia coli, NarK and NarU. The genes encoding these proteins are associated with the structural genes for two isoforms of respiratory nitrate reductases found in E. coli and thus NarK and NarU might reasonably be assumed to have the same function, most likely nitrate/nitrite antiport. However, whereas Zheng et al. (2013) interpret their NarK structure in terms of this protein being a nitrate/nitrite antiporter and Fukuda et al. (2015) provide experimental support for antiporter function alongside their structure, Yan et al. (2013) chose to interpret their NarU structure in terms of a nitrate/cation (and probably not a proton) symporter. This conflicting evidence, alongside a continuing tendency to overlook the role of NarK protein-family members in nitrate assimilation, adds to the need for further experimental data which will contribute to pinpointing the exact function of each of these proteins.
To shed more light onto the function of NarK-like transporters, we focused on Paracoccus denitrificans, a model organism for studying denitrification, the first step of which involves the uptake of nitrate from the periplasm into the cytoplasm by the NarK transporter. In this organism, NarK is an unusual fusion of two NarK-like transporters, NarK1 and NarK2, and is essential for anaerobic growth when nitrate is the terminal electron acceptor (Goddard et al., 2008). The NarK1 and NarK2 activities show a strong interdependence and full-length NarK cannot mimic the individual activities of the distinct domains (Goddard et al., 2008). A schematic representation of the nitrate/nitrite transporters of P. denitrificans, along with some of the proteins interacting with nitrate or nitrite on either side of the inner membrane, is presented in Fig. 1 while a bioinformatics analysis of NarK1 and NarK2 in relation to other nitrate/nitrite transporters of the same family is shown in Fig. 2. In line with the classification mentioned above, P. denitrificans NarK1 and NarK2 should be respectively, a nitrate/proton symporter and a nitrate/nitrite antiporter (Wood et al., 2002;Goddard et al., 2008).
Direct study of nitrate transport is challenging, as suitable isotopes of nitrogen are rare and have extremely short half-lives, especially if they were to be used alongside bacterial growth rates. Therefore, in this study we used complementation experiments to elucidate the transport mechanisms of NarK1 and NarK2. This is a classical approach which we have used before, for example to demonstrate that both NarK1 and NarK2 can, but to different extents, complement the deletion of the entire narK gene (Goddard et al., 2008). Here, we explored whether NarK and its individual domains can complement for another NarK-like protein, NasA, in P. denitrificans. NasA clusters with NarK1 in phylogenetic analysis (Moir and Wood, 2001) (Fig. 2B), but is located within the nas operon. It is proposed to be a nitrate/proton symporter (Gates et al., 2011) which supplies nitrate to the cytoplasmic assimilatory nitrate reductase NasC (Fig. 1). The resulting nitrite is further reduced in the cytoplasm (by NasB, Fig. 1) so there is no requirement for nitrite export, in contrast to the dissimilatory Nar system.
Given the lack of certainty regarding the exact mechanisms of NarK-like transporters, in addition to our complementation experiments we performed a number of residue-specific studies in order to identify aminoacids with a central role in the transport process. Firstly, we investigated the role of the linker region between NarK1 and NarK2 (Wood et al., 2002;Goddard et al., 2008); several homologues of a fused NarK1-NarK2 protein have been identified and the length and primary amino acid sequence of this linker region is highly variable (Goddard et al., 2008). It is, therefore, not clear whether the linker is just a tether between the two domains or it is involved in a regulatory interaction between NarK1 and NarK2 (Goddard et al., 2008). Secondly, we focused on the proline residues within the transmembrane domains of both NarK1 and NarK2; prolines are key in such environments [reviewed in (Cordes et al., 2002)] and are likely to be involved in the catalytic activity or the structural flexibility of NarK-like proteins. Thirdly, we examined by structural modelling whether there are amino acids within or around the transport pores of NarK1 and NasA which could be reversibly protonated, a putative feature that would support the proposal that these proteins are nitrate/proton symporters.
By combining complementation experiments with sitedirected mutagenesis, we were able to add to the evidence for the proposed functions of P. denitrificans NarK1 and NarK2 and, in line with these functions, to examine the importance of conserved amino acid residues or specific regions of the NarK fusion. In addition, through cross-complementation studies, we compared the functions of NarK1 and NarK2 with the functions of two nitrate/nitrite transporters found in E. coli, NarK and NarU, as well as with another P. denitrificans transporter, NasH (Fig. 1), and thus, we obtained insight into the range of functionalities and mechanisms encountered in bacterial nitrate/nitrite transport proteins.

Results
Complementation of P. denitrificans DnasA P. denitrificans NasA supplies nitrate to the cytoplasmic NasC nitrate reductase. The resultant nitrite is further converted by the NasB nitrite reductase to ammonium in the cytoplasm of the cell (Fig. 1). Therefore, there is no requirement for NasA to transport nitrite out of the cell and naturally it has been assumed to be a nitrate/proton symporter (Gates et al., 2011). Aerobic growth with nitrate as the sole nitrogen source is abolished by the loss of nasA (Gates et al., 2011), so the nitrate/proton transport deficient background provided by this strain can be used as a test platform for complementation analyses. Both NarK1 (predicted nitrate/proton symporter) and NarK2 (predicted nitrate/nitrite antiporter) can import nitrate but NarK2 should obstruct the action of NasB by exporting nitrite to the periplasm. This means that the extent of complementation, and thus aerobic growth of a nasA deletion strain with nitrate as a sole nitrogen source, provides an excellent whole-cell system to assay for the role of any NarK-like transporter of interest.
Complementation of DnasA with plasmid-borne nasA proved more troublesome than previous complementations of narK mutants using a comparable procedure. Initially, we sought to express nasA by choosing a translational start that corresponded to that of the very closely related NarK1 protein. Following failure to observe any complementation, nasA starting at an ATG (position 2129) to the 5' side of our original choice was investigated for complementation and was successful Fig. 1. Schematic representation of the nitrate/nitrite transporters of P. denitrificans, discussed in this study, along with some of the proteins interacting with nitrate or nitrite on either side of the inner membrane. In P. denitrificans, NarK1 and NarK2 are a single protein fusion called NarK. Proteins involved in the assimilation pathway are shown in white while proteins involved in respiration are shown in black. Nitrate respiration occurs under anaerobic conditions whatever the nitrogen source, whereas assimilatory proteins are expressed aerobically or anaerobically only when preferred nitrogen sources, such as ammonium, are absent. Uncertainties about the roles of NasA and NarK1/ NarK2 are discussed in this study. Nar, and not Nap, is the default enzyme for respiration. (Fig. 3A). This result adds to the previous evidence (Gates et al., 2011) that the nasA::kan mutation is nonpolar.
NarK1 successfully complemented aerobic growth of the nasA::kan mutant which showed near-WT doubling time and growth yield, albeit with a longer growth lag (Fig. 3A). This was comparable to the behaviour seen during complementation with plasmid-borne nasA. By contrast, plasmid-borne narK2 was inferior at restoring growth, both in terms of rate and maximum cell density reached (note that Fig. 3A is in semi-log scale), as was the full-length narK construct (Fig. 3A). This suggests that NasA and NarK1 function similarly, most probably as nitrate/proton symporters. On the other hand, the A. Alignment of P. denitrificans NasA (Pd_NasA), NarK1 (Pd_NarK1) and NarK2 (Pd_NarK2), E. coli NarK (Ec_NarK) and NarU (Ec_NarU), and A. nidulans CRNA* (An_CRNA* lacks residues 221-310 of the WT protein sequence; the extended predicted intracellular loop of WT An_CRNA (residues 221-310) was removed because it compromises the alignment) was performed using Multalin (Corpet, 1988). An_CRNA* is included to give a comparison with a eukaryotic NarK-like protein. Residues which are conserved in all proteins are shown in red whereas residues which are conserved in half or more of the proteins are in blue. B. Phylogenetic tree of the same nitrate/ nitrite transporter as shown in (A). In this case, the WT sequence of An_CRNA was used. The tree was constructed using Multalin (Corpet, 1988). The scale bar represents 10 point accepted mutations (PAM).
inability of NarK2 to replace NasA is consistent with its role as a nitrate/nitrite antiporter whose function results in the export of nitrite generated from the reduction of nitrate from the cell rather than its retention for cytoplasmic assimilation. Note that these differences in complementation are not due to expression levels of these proteins as we found that C-terminally hexahistidinetagged NarK, NarK1 and NarK2 were produced at similar levels (Fig. 4A). It should also be recalled that expression levels of NarK2 have been shown to be sufficient to support respiration in a narK deletion, whereas NarK1 alone was less effective (Goddard et al., 2008).
A transient accumulation of nitrite was observed in strains expressing full-length NarK (Fig. 3C). This is similar to observations made during anaerobic growth (Goddard et al., 2008) and is consistent with the ability of these transporters to act as nitrate/nitrite antiporters. The subsequent drop in extracellular nitrite can be probably attributed to the nitrite import activity of NasH ( Fig.  1), encoded within the nas operon of P. denitrificans and predicted to transport nitrite bi-directionally across the cytoplasmic membrane (see below). Complementation with plasmid-borne nasA also resulted in nitrite accumulation ( Fig. 3C), which would be contrary to it performing a role solely in nitrate uptake. However, we interpret this transient accumulation as a result of higher nitrate uptake by the complemented strain compared to the WT strain: the plasmids expressing NasA and NarK are based on the same expression vector and therefore, we expect relatively similar expression levels of the two proteins. The amounts of NarK produced from the plasmid are clearly sufficient to support respiration (Wood et al., 2002;Goddard et al., 2008) and as much lower flux through NasA is needed for assimilation there could be more NasA expressed than required in the NasAcomplemented strain. In this case, NasH would be exporting the excess nitrite produced by the assimilatory nitrate reductase (NasC) acting on the imported excess nitrate, thus ensuring that cytoplasmic nitrite concentration does not rise to toxic levels inside the cell and causing a transient nitrite accumulation. It should be noted that although the expression level of NasA does not look equivalent to that of NarK according to Fig. 4B, one needs to take into account that the reaction of different proteins with a specific antibody can vary, especially for integral membrane proteins where their affinity tags are P. denitrificans (A) DnasA or (B) DnasA DnasH were grown aerobically in the presence of nitrate as a sole nitrogen source, while harboring pEG276-derived plasmids expressing full-length P. denitrificans NarK (᭛), P. denitrificans NarK1 (D), P. denitrificans NarK2 (᭺), P. denitrificans NasA (1) or empty vector (*). WT Pd1222 is indicated by (w). Extracellular nitrite accumulation was also determined in the same strains ((C) for DnasA and (D) for DnasA DnasH). Three biological replicates were performed for each experiment and representative data is shown. likely to be inaccessible; anti-polyhistidine antibodies are known to often cause such problems. Fig. 4B shows that both proteins were efficiently produced and complementation assays [ (Wood et al., 2002;Goddard et al., 2008) for NarK and Fig. 3A for NasA] along with the transient nitrite accumulation (Fig. 3C) further supports that they are expressed at equivalent levels.

Complementation of P. denitrificans DnasA DnasH
The P. denitrificans nitrite transporter NasH is homologous to NirC, a protein which is thought to transport nitrite bi-directionally across the cytoplasmic membrane during the growth of E. coli with nitrate as the terminal electron acceptor (Jia et al., 2009) (Fig. 1). Both NasH and E. coli NirC are distinct from the MFS transporter superfamily; in fact Salmonella typhimurium NirC has been shown to adopt a pentameric structure similar to FocA (Wang et al., 2009;Waight et al., 2010;Lu et al., 2012). Evidence from disruption of nasH in P. denitrificans is in line with a role in nitrite transport (Gates et al., 2011); as with disruption of nasA, a DnasA DnasH strain is unable to grow aerobically with nitrate as the sole nitrogen source (Gates et al., 2011). We note that Rycovska et al. (2012) argue that NirC is a nitrite/proton antiporter but in this manuscript we adhere to the view of Lu et al. (2012) who report that this protein is a nitrite channel.
Expression of NarK1 from a plasmid was able to restore near-WT growth cell density to the DnasA DnasH strain, but with a longer lag phase than for the WT (Fig. 3B). However, unlike the single nasA::kan mutant, neither NarK2 nor full-length NarK when expressed from a plasmid were able to restore the growth in the double deletion mutant (Fig.  3B). Nitrite accumulated in the extracellular growth medium in the DnasA DnasH mutant complemented with NarK1 expressed from a plasmid but this time the accumulation was not transient (compare Fig. 3C and D). It has previously been demonstrated that NarK1 is capable of low-level nitrite export along with nitrate import (Goddard et al., 2008). Thus, this result provides further evidence that NasH can contribute to uptake of nitrite released during nitrate assimilation. The latter is supported by the unsuccessful complementation of the DnasA DnasH mutant with full-length NarK or NarK2. In this case, nitrite accumulated to the level of 120 lM for complementation with NarK and 50 lM for complementation with NarK2 ( Fig. 3D), even though there was no significant growth, which is consistent with NasH importing nitrite back into the cell and also with NarK2 being a nitrate/nitrite antiporter.
The DnasA DnasH mutant could not be complemented by nasA, which can be explained by NasA's function solely as a nitrate importer and NasH's function as a bidirectional transporter of nitrite; NasH, when present, carries out nitrite export (along with nitrite uptake, see above) but if it is absent, nitrite can build up to a toxic level and cell growth is poor. This also agrees with the transient nitrite accumulation observed during the complementation of the DnasA mutant by plasmid-borne nasA, where, NasH performs a nitrite export function when the nitrate import from NasA is overwhelming (Fig. 3C).

Complementation of P. denitrificans DnarK
If the entire narK gene is deleted in P. denitrificans, the resultant deletion strain cannot grow anaerobically using nitrate as respiratory electron acceptor (Wood et al., 2002;Goddard et al., 2008). Anaerobic growth of the A. NarK full length, NarK1 or NarK2 and (B) NasA were expressed from equivalent constructs with a C-terminal hexahistidine-tag. Extracts from P. denitrificans DnasA strains expressing the relevant protein were analysed by SDS-PAGE, followed by Western blot analysis using an anti-pentahistidine antibody (Qiagen). NarK1, NarK2 and NasA migrate faster than predicted based on their molecular weight; this is common for integral membrane proteins and it is consistent with previous studies (Goddard et al., 2008). The protein band detected in all lanes of (A), migrating below the 28 kDa protein standard, is due to non-specific binding of the antibody. The same expression pattern was observed in a DnasA DnasH strain.
narK mutant was previously shown to be restored by complementation with full-length NarK, NarK1 or NarK2 (Goddard et al., 2008) although NarK1 was the least effective. In this study we found that NasA failed to complement for the absence of NarK; when NasA was expressed from a plasmid bearing a C-terminal hexahistidine tag, immunoblotting using an anti-pentahistidine antibody indicated that it was efficiently expressed (Fig.  4B). In addition, as we have mentioned above, it is likely that NasA levels when this protein is expressed from a plasmid are higher than in the WT strain (see transient nitrite accumulation in Fig. 3C). Thus, the failure of NasA to complement a narK deletion once more indicates that NasA is solely a nitrate import transporter. NasH is not expressed during anaerobic growth in the presence of ammonium and therefore, nitrite export to the periplasm, the location of the respiratory nitrite reductase (Fig. 1), is absent and NasA is unable to rectify this, thereby not supporting denitrification.
Complementation of P. denitrificans DnasA by E. coli NarK and NarU NarK and NarU from E. coli are believed to be nitrate/ nitrite antiporters (Clegg et al., 2002;Jia and Cole, 2005;Clegg et al., 2006;Jia et al., 2009;Zheng et al., 2013). A recent crystal structure of NarU was solved with and without its substrate (nitrate) (Yan et al., 2013) and, based on structural and primary sequence information, it has been proposed that the transport mechanism of NarU deviates from the canonical 'rocker-switch' model of substrate translocation. On the other hand, the crystal structure of NarK from E. coli, also solved with and without its substrate (Zheng et al., 2013), suggested a 'rocker-switch' model for nitrate/nitrite exchange.
From our previous results, we know that the E. coli NarK and NarU can complement a deletion in P. denitrificans NarK (Goddard et al., 2008) but we would predict that these proteins would not be able to complement a deletion in nasA. Nevertheless, we assessed the ability of E. coli NarK and NarU to complement a DnasA strain during aerobic growth in the presence of nitrate as the sole nitrogen source. Both of these transporters were able to complement the nasA deletion, restoring WT growth kinetics and maximal cell density (Fig. 5). This demonstrates that E. coli NarK and NarU can facilitate the net uptake of nitrate; nitrate/nitrite exchange should be detrimental to growth under these conditions.
Targeted mutagenesis studies on P. denitrificans NarK To investigate aspects of the mechanistic basis for nitrogen-oxyanion transport by NarK1 and NarK2, we undertook a series or targeted mutagenesis studies. We investigated the ability of several NarK (and NarK1 or NarK2 domains) variants in restoring growth of a P. denitrificans DnarK strain under anaerobic growth conditions, when nitrate was used as the sole terminal electron acceptor. The strain used also lacks napD (Goddard et al., 2008), resulting in the absence of the periplasmic nitrate reductase NapAB (Fig. 1). This was necessary in order to avoid complications which may have otherwise arisen from the reduction of nitrate to nitrite without the requirement for transmembrane transport of nitrate.
The NarK1-NarK2 linker region. Currently, 'fused' NarK1-NarK2 transporters have been identified in more than a dozen organisms but only NarK from Paracoccus pantotrophus (Wood et al., 2002) and P. denitrificans (Goddard et al., 2008) have been experimentally characterized. Fused NarK transporters show a good degree of similarity, with the largest regions of variability at the extreme N-and C-termini and the 'linker' region between the predicted helices 12 and 13 (effectively the point at which the C-terminus of NarK1 is fused to the Nterminus of NarK2, Fig. 6). A functional interaction has been demonstrated between NarK1 and NarK2 of P. denitrificans (Goddard et al., 2008) and it could be that the linker region tethers the two subunits in close proximity to enable this interaction to occur. The length of this linker region varies between organisms and sequence alignment of more than twenty such domains, revealed the presence of only two completely conserved residues, a leucine and a proline (L425 and P426 in P. denitrificans NarK). We carried out mutagenesis of these residues to determine if they played any essential role in activity. NarK L425A , NarK P426A , NarK L425A,P426A and NarK DL425DP426 (lacking both L425 and P426) mutants were all able to restore anaerobic growth of the DnarK strain with nitrate as the terminal electron acceptor. The growth rates and maximal cell densities achieved by strains expressing these mutants were indistinguishable from those of the comparable strain expressing WT NarK (not shown). Therefore, the linker region most likely functions solely to tether the two domains. Tethering could promote heterodimerisation between NarK1 and NarK2 over homodimerisation, something that in itself would be important for the function of the NarK fusion, so it is not entirely surprising that the conserved residues of the linker play no essential role in the transport processes catalysed by the two domains.
Mutagenesis of proline residues in NarK1 and NarK2. The functional interaction between the NarK1 and NarK2 domains in the NarK fusion leads to complex growth phenotypes when a single domain (narK1 or narK2) in the full-length gene is inactivated (Goddard et al., 2008). It is, therefore, not straightforward to determine the effects of mutations in one subunit in the presence of the other fused subunit. Individually, narK1 and narK2 are capable of supporting growth in a DnarK strain (Goddard et al., 2008), allowing us to investigate the effect of mutations within the two domains without such complications.

MFS
proteins typically contain a twelvetransmembrane (TM) helix core, composed of two sixhelix sub-domains surrounding a central ligand-binding cavity and most of them are thought to function via a 'rocker switch' model in which a single transport site alternately faces either the inside or the outside of the cell (Abramson et al., 2004;Law et al., 2007). It is likely that this re-orientation of the transport site involves significant conformational changes, during which the relative positions of helices significantly differ. Proline residues usually have a key role in dynamic processes involving transmembrane a-helices where they occur with a relatively high frequency compared to their soluble counterparts [examples reviewed in (Cordes et al., 2002)]; due to the steric hindrance caused by their ring structure and as their backbone nitrogen is unavailable for normal hydrogen bonding they introduce a bend in a-helices which contributes to the their flexibility.
To date, only a single essential proline residue has been reported in a homologue of NarK1; the P50S mutation in NarK2, a NarK1-like transporter from Mycobacterium tuberculosis, abolished signalling (Giffin et al., 2012). Mutation of P113 in E. coli NarU, a homologue of P. denitrificans NarK2, to cysteine or leucine was also detrimental to its function but mutation to alanine was not (Jia et al., 2009). In this study, we performed mutagenesis on proline residues within and close to the predicted ends of TM helices in NarK1 and NarK2 (Fig. 6). Although conserved residues are preferable targets for mutagenesis, we also investigated less well-conserved residues as it is largely unknown if all NarK1-or NarK2like transporters operate by the same mechanism or even if they transport the same substrates.
NarK1 contains seven such prolines (Fig. 6); four are highly conserved within NarK1 homologues (P58, P157, P302 and P375) and three are more poorly conserved (P240, P303 and P325). We mutated each of these residues to alanine and tested the ability of each variant to denitrificans NarK demonstrating the location of important residues. The 24 transmembrane helices and connecting loops of NarK, as predicted by the TMHMM program, are illustrated and the relative positions of residues mutated in this study are marked. White circles represent non-essential amino acids and black circles represent residues that abolish the growth of a DnarK strain complemented with the appropriate plasmid; mutations in the essential arginines have been described previously (Goddard et al., 2008). M442, the start codon of NarK2, is indicated with a grey circle. restore growth of a DnarK strain. Only the P302A mutation rendered NarK1 inactive. NarK1-expressing cells are characterized by an increased lag phase before exponential growth compared to the WT NarK [approximately 16 h compared to 6 h (Goddard et al., 2008)]. As such the NarK1 P302A -expressing strain was incubated for 72 h to ensure that the absence of growth was not simply due to a vastly increased lag phase. No growth above the level of a DnarK strain containing the empty expression vector was observed at this time. Expression of the NarK1 P302A mutant at an equivalent level to the WT NarK1 was confirmed via Western blot analysis of a strain expressing a C-terminally hexahistidine-tagged protein (NarK1 P302A -His; Fig. 7).
NarK2 contains eight prolines within its transmembrane helices (Fig. 6, NarK2 starts at the methionine preceding helix 13 in this figure). Of these, six are highly conserved within NarK2-like proteins (P512, P546, P617, P652, P729 and P847) and the other two are more poorly conserved (P474 and P709). It should be noted that the residue numbers correspond to the fulllength NarK protein of which NarK2 comprises the Cterminal half, with the first residue of NarK2 being the endogenous M442 (Fig. 6). The ability of each proline to alanine mutant to restore growth of a DnarK strain was tested and the NarK2 P512A , NarK2 P729A and NarK2 P847A mutants were unable to restore anaerobic growth to a DnarK strain. Expression of each of these transporters was confirmed to be at equivalent levels to the WT via Western blot analysis as above (Fig. 7).

Discussion
Function of bacterial NarK-like transporters P. denitrificans NarK1 is involved in providing nitrate for respiration, but its sequence is more related to proteins like NasA (Fig. 1) from the assimilatory pathway (Fig. 2). The role of NasA in nitrate import is indicated by the fact that in P. denitrificans its absence compromises growth on nitrate but not on nitrite (Gates et al., 2011) and by the presence of several NasA analogues with a clear role in nitrate uptake in eukaryotes (Zhou et al., 2000;Krapp et al., 2014). Therefore, it is plausible that the function of NarK1 is to catalyse net nitrate uptake, a step that in principle would be anyway needed for the activation of NarK2 which is widely assumed to require cytoplasmic nitrite before it can import nitrate. In this context, our finding that NarK1 can complement a nasA deletion (Fig. 3A) provides key evidence for its role as a nitrate/proton symporter. This is further supported by the mutagenesis data presented in previous studies (Wood et al., 2002;Goddard et al., 2008;Gates et al., 2011). The failure of NasA to complement a DnarK mutant, while NarK1 can do so (Goddard et al., 2008), indicates that NarK1 must be able to also catalyse nitrite export to some extent. By contrast, NarK2 cannot complement a deletion in nasA (Fig. 3A) and nitrite accumulates in the extracellular medium when NarK2 is expressed in a DnasA DnasH strain (Fig. 3D). This strongly supports a model where NarK2 is a nitrate/nitrite antiporter.

Function-to-structure relationships of bacterial NarK-like transporters
The proposal that NasA and NarK1 are members of a wider family of net nitrate uptake proteins while NarK2 is an example of a nitrate/nitrite antiporter, appears rational until a structural analysis of this family of proteins is performed. We have built a homology model of P. denitrificans NasA and NarK1 based on the published E. coli NarK structure (Zheng et al., 2013) (Fig. 8A) and we found that both NasA and NarK1 overlay extremely well with E. coli NarK (PDB 4U4V), with root mean square deviations (RMSDs) of 2.97 Å over 368 residues and 2.72 Å over 376 residues respectively. The same was observed when a homology model for NarK2 was built based on the published E. coli NarU (Yan et al., 2013) (Fig. 8B) as well as the E. coli NarK structure (Zheng et al., 2013) (not shown). This is not entirely unexpected as the P. denitrificans proteins present on average 50% primary sequence identity to the E. coli transporters. However, based on these homology models it seems that these five proteins (P. denitrificans Fig. 7. Analysis of the expression of NarK1 and NarK2 proline variants. For each of the NarK1 or NarK2 proline mutations which affected the function of these domains, the proline variant was expressed from a construct with a C-terminal hexahistidine-tag. Extracts from P. denitrificans DnarK strains expressing the relevant variant were analysed by SDS-PAGE followed by Western blot analysis using an anti-pentahistidine antibody (Qiagen). NarK1 and NarK2 migrate faster than predicted based on their molecular weight; this is common for integral membrane proteins and it is consistent with previous studies (Goddard et al., 2008). The protein band detected in all lanes, migrating below the 28 kDa protein standard, is due to non-specific binding of the antibody. NarK1, NarK2 and NasA and E. coli NarK and NarU), which are predicted to have very different functions, have practically identical structures. This raises two possibilities. One possibility is that subtle sequence and structural differences between NasA, NarK1 and similar proteins in organisms that can only assimilate nitrate on one hand, and NarK2 and similar proteins in organisms that can respire on nitrate on the other, account for the proposed functional differences. A second, much more surprising possibility, which has to be considered, is that A. Overlay of the homology models of P. denitrificans NarK1 and NasA on the crystal structure of E. coli NarK. The crystal structure of E. coli NarK (PDB: 4U4V) is shown in cartoon and is coloured from the N-terminus (blue) to the C-terminus (red) while the homology models of NarK1 and NasA are shown in grey and magenta respectively. The main residues involved in nitrate binding are strictly conserved in NarK1 and NasA and overlay extremely well onto the equivalent side chains of E. coli NarK (active-site inset). B. Overlay of the homology model of P. denitrificans NarK2 on the crystal structure of E. coli NarU. The crystal structure of E. coli NarU (PDB: 4IU8) is shown in cartoon and is coloured from the N-terminus (blue) to the C-terminus (red) while the homology model of NarK2 is shown in grey. The main residues involved in nitrate binding are strictly conserved in NarK2 and overlay extremely well onto the equivalent side chains of E. coli NarU (active-site inset). all these proteins function as nitrate/nitrite antiporters although the idea that NasA, an assimilatory protein, performs nitrate/nitrite exchange is of course counterintuitive.
In considering the second possibility, one needs to take into account that for this to happen the following steps would have to take place in P. denitrificans: nitrate would have be imported into the cytoplasm for assimilation and reduced to nitrite, nitrite would have to be subsequently exported back to the periplasm in exchange for nitrate via NasA, and exported nitrite would have to then be transported back into the cytoplasm by NasH for reduction to ammonium for assimilation purposes (Fig. 1). However, absence of NasH is more detrimental to bacterial growth in the presence of nitrite than nitrate (Gates et al., 2011) which would not be the case if NasA and NasH acted sequentially to exchange periplasmic nitrate for cytoplasmic nitrite and then to take up nitrite. The complementation experiments presented in this work and previously (Goddard et al., 2008), which clearly show different functions for NarK1, NarK2 and NasA, further support the notion that these proteins cannot all be nitrate/nitrite antiporters. For example, complementation by NarK2 is much superior in restoring nitrate respiration than nitrate assimilation, consistent with its dominant antiporter role and inconsistent with export of nitrite, followed by reuptake by NasH under assimilatory conditions. Another indication in favour of the different functions of NarK-like proteins comes from organisms like Bradyrhizobium diazoefficiens USDA 110 that lack analogues of NasH and the Nar-type respiratory nitrate reductase and only have a NarK-like protein which functions to supply nitrate to an assimilatory nitrate reductase (Cabrera et al., 2016). This Bradyrhizobium NarK supplies nitrate for assimilation, so if after its reduction nitrite were to be exported in exchange for incoming nitrate there would be no obvious candidate for nitrite uptake. We can, therefore, conclude that despite their structural similarities (as predicted by our homology models, Fig. 8), there is compelling biological evidence in this study and in the literature supporting the different functions of NarK-like transporters.
A family of fungal nitrate transporters provides additional evidence in favour of some NarK-like proteins functioning in nitrate/proton symport rather than nitrate/ nitrite exchange. The sequence of a nitrate transporter from Aspergillus nidulans, usually termed CRNA, aligns very well with the proteins of the bacterial NarK family, especially when its large cytoplasmic loop (residues 221-310) is omitted ( Fig. 2A). In line with the sequence similarity, the sequence of CRNA can also be easily threaded onto the crystal structure of the E. coli NarU (Yan et al., 2013). The fact that CRNA and bacterial NarK-like proteins are so similar reinforces the hypothesis that very subtle differences must underpin the diversity of functions observed for these transporters. Notably, CRNA clusters phylogenetically with the group of NarK proteins that are proposed to be nitrate/ proton symporters (like P. denitrificans NarK1 and NasA, Fig. 2B) and in fact electrophysiological experiments have confirmed its role in the symport of two protons with one nitrate anion (Zhou et al., 2000). The same function has been reported for other related fungal proteins, for example a transporter from Tuber borchii (Montanini et al., 2006). The phylogenetic analysis, together with the experimental evidence for the function of these fungal analogues, align with the function of NasA as a nitrate/proton symporter rather than a nitrate/ nitrite antiporter. Quite how NasA performs this function is still puzzling as its transmembrane parts do not contain any of the residues which are known to catalyse proton transport (for example, aspartate, glutamate, histidine or lysine) unlike the plant NRT1.1 nitrate/proton symporter (Parker and Newstead, 2014a;Sun et al., 2014), an MFS protein where a histidine residue is found to facilitate proton transport across the membrane.
Finally, the possibility that any members of the NarK protein family catalyse nitrate/alkali metal-cation symport [proposed by Yan et al. (2013) for E. coli NarU] seems, at this point, remote as there is only one example of such a process catalysed by an MFS protein. This sole example is the melibiose/sodium or proton symporter for which the structure shows that it has an unusual constellation of charged residues capable of binding a sodium cation or a proton adjacent to the melibiose binding site (Ethayathulla et al., 2014). Such features are not present in known NarK protein-family structures (see active sites in Fig. 8) or structures of any other, currently known, MFS proteins.
Function of E. coli NarK and NarU E. coli NarK and NarU have been thought to be nitrate/ nitrite antiporters and phylogenetically they are close to NarK2 (Fig. 2B). In fact they are similar enough to NarK2 (sequence identity 50%) that the homology model of NarK2 built on the E.coli NarU structure (Yan et al., 2013) (Fig. 8B) overlays onto the crystal structure (PDB: 4IU8) (Zheng et al., 2013) with an RMSD of only 4.1 Å over 360 residues. The main nitrate-binding site displays an almost identical arrangement of side chains in the NarK2 model (inset, Fig. 8), suggesting a very similar function and interaction with nitrate ligands in NarK2 and E. coli NarU (but also E. coli NarK whose structure is very similar to that of NarU). However, unlike P. denitrificans NarK2, both E. coli NarK and NarU fully complemented a nasA deletion, which strongly suggests that they are capable of facilitating the net uptake of nitrate despite their proposed nitrate/nitrite antiporter function.
The literature on the E. coli NarK and NarU function is conflicting. Fukuda et al. (2015) show evidence that NarK should be a nitrate/nitrite antiporter whereas Yan et al. (2013) propose a cation symport function for NarU. Cole and colleagues have argued in the past that E. coli NarU might catalyse nitrate/proton symport under conditions of low intracellular nitrite, as part of the uptake process, and nitrite/proton symport as the export process (Jia et al., 2009). Such nitrite export was argued to be followed by nitrite import by NirC, with subsequent cytoplasmic reduction to ammonium (Jia et al., 2009). Provided that the number of protons is equal for the two processes, this is energetically indistinguishable from simple nitrate/nitrite exchange and may well explain the complementation of the nasA mutant seen in this work. Of course, in order for NarU to catalyse net nitrate uptake, it would need to reorient towards the periplasmic surface with an empty binding site, whereas to support respiration, it would have to reorient with nitrite bound in its active site. This would be unorthodox; usually only certain ligand states of transporters can reorient in order to avoid futile cycles of activity [although variable specificity has been observed, for example in the proton-coupled oligopeptide (POT) transporters where the number of protons transported varies depending on the size of the peptide (Parker et al., 2014b)]. Nitrate/proton symport would also require a protonatable residue near the active site to facilitate nitrate uptake but no obvious candidates (for example, aspartate, glutamate, histidine or lysine) can be found in E. coli NarK and NarU. Of course taking into account the importance of the two conserved arginines for the function of all of these transport proteins (Yan et al., 2013;Zheng et al., 2013) (Fig. 6) one could argue that an obvious question is why are two positive charges needed to bind the monovalent nitrate or nitrite. Given that there is precedent in certain enzymes (Guillen Schlippe and Hedstrom, 2005) for arginine residues to have sufficiently perturbed pK a values to act as bases, an alternative role for one or both of these conserved arginines is also worth considering. However, for E. coli NarK binding of the nitrate between the two positively charged arginines has been found to be integral for the occurrence of the necessary conformational changes during the transport cycle (Fukuda et al., 2015) and its recent crystal structure has revealed a positively charged substrate-translocation pathway (Zheng et al., 2013). These observations mean that a proton symport mechanism is highly unlikely. Therefore, most evidence points towards a nitrate/nitrite antiporter function of E. coli NarK and NarU, but more detailed experiments would be needed to explain their apparent ability to facilitate net nitrate uptake in P. denitrificans shown in this work (Fig. 5).
Bacterial NarK-like transporters compared to mitochondrial transport proteins An explanation for the wide variety of functionalities among the NarK-like proteins could come from the work of (Robinson et al., 2008). In this study, it was proposed that a family of mitochondrial transport proteins adopt a range of transport mechanisms which are regulated by the strength of salt-bridge networks forming between residues which are positioned at the water-membrane interfaces of the transporter on either side of the mitochondrial membrane. These networks can have a key role in the transport process as substrate binding provides the necessary energy to overcome them and allows reorientation of the transporter within the membrane. Strong networks on both sides of the membrane are indicative of strict equimolar exchange of substrates, with energy provision required in both directions and provided by substrate binding. A weak network on one side and a strong network on the opposite side favours symport from the side containing the weak network. Finally, weak networks on either side of the membrane allow a transporter to possess both symport and antiport mechanisms depending on the concentrations of the substrate/s in each cellular compartment.
Based on the above model, P. denitrificans NarK1, a net nitrate importer, should have a high affinity for nitrate (Goddard et al., 2008) when open to the periplasm but a low affinity for nitrite when open to the cytoplasm. In this study and elsewhere (Goddard et al., 2008), it has been observed that NarK1 facilitates lowlevel nitrite export, along with nitrate import. A strong salt bridge on the cytoplasmic side would favour reorientation of NarK1 to the conformation in which the periplasmic binding site would open even in the absence of nitrite binding, which agrees with the overall net nitrate importer function of NarK1. On the other hand, strong networks on both sides of the membrane would promote nitrate/nitrite exchange for NarK2. NarK and NarU from E. coli could have weak networks on both sides, allowing both uptake of nitrate and export of nitrite, even when this process is not equimolar, due to spontaneous reorientation of the transporters. This provides a possible explanation for these transporters being able to complement an essentially symport process (aerobic nitrate assimilation) when they are normally associated with an antiport mechanism. It is likely that in the dissimilatory nitrate reductase pathway, nitrite is at a relatively high cytoplasmic concentration which would support nitrate/nitrite antiport. However, as nitrite is reduced to ammonia in the assimilatory pathway, its concentration might be lower and spontaneous reorientation of the transporter could allow net nitrate uptake in the absence of nitrite export. While this model provides some explanation for the functions of bacterial NarK-like transporters, further work would be required in order to identify conserved residues involved in the salt bridges above or other features that might allow switching of transporters from antiport to symport functions. Such features might include specific protein-protein interactions with respiratory or assimilatory nitrate reductases, analogous to the protein-protein interactions within the denitrification system reported by Borrero-de Acuna et al. (2016), or a posttranslational modification of the NarK-like proteins themselves; the control of a plant nitrate/proton symporter by phosphorylation illustrates the latter possibility (Sun et al., 2014).

Residue-specific studies of NarK-like transporters
Although it has been demonstrated that NarK from P. denitrificans can transport nitrate and nitrite (Goddard et al., 2008), details of the molecular mechanisms underlying this process are incomplete. Fukuda et al. (2015) have performed extensive mutagenesis studies on E. coli NarK; however, they have not examined the role of proline residues which are known to be important for integral membrane proteins; prolines introduce kinks in transmembrane helices which tend to destabilize their packing and to facilitate conformational transitions between outward-facing, occluded and inward-facing states of the transporter (Senes et al., 2004) leading to a net transport of substrate across the membrane. One proline residue (P302), located on the periplasmic end of TM9, was found to be essential for NarK1 activity. In general, it has been demonstrated that in other MFS transporters, helices 1, 2, 4, 5, 7, 8, 10 and 11 form the central transport pore (Hirai et al., 2002;Abramson et al., 2003;Huang et al., 2003) and it is thought that most, if not all, MFS proteins have an overall conserved structure (Abramson et al., 2004). It is possible that the position of P302 on the periplasmic face of the transporter assists in the relative movement of the helices during the transport process. This would be in line with the transport model proposed for LacY where P123 and P327 cause kinks in transmembrane helices, providing the flexibility required for the hydrophilic cavity to open and close as appropriate (Abramson et al., 2003). In contrast to NarK2 from M. tuberculosis (a NarK1-like protein) (Giffin et al., 2012), the proline in TM2 (P50 in M. tuberculosis, P58 in P. denitrificans) was not found to be essential, indicating potential subtle differences in structure or transport mechanism of these proteins. NasA from Bacillus subtilis complemented an M. tuberculosis NarK2 deletion, consistent with a conclusion that the latter is, most likely, not a nitrate/nitrite antiporter (Giffin et al., 2012).
Three proline residues were found to be essential for the activity of NarK2. As with NarK1, one of these (P847) is located at the periplasmic end of a helix (TM11) which is predicted to form part of the transport pore. It is could be that this residue orientates the helix in the correct conformation for transport or imparts the flexibility required for reorientation of the helices during the transport process. The other two essential proline residues are located in TM2 (P512) and TM8 (P729) and are conserved within a large number of NarK2-like transporters. Interestingly, the essential arginine residues, R520 and R736, which can bind the substrate(s), are located in the same helices as these proline residues and in general movements of TM8 are widely implicated in the functions of the MFS protein family (Fukuda et al., 2015).

Conclusion
Our complementation experiments in nasA and nasH mutants provided experimental evidence supporting that in P. denitrificans NarK1 is a nitrate/proton symporter and NarK2 is a nitrate/nitrite antiporter but also that NasA is a nitrate/proton symporter and NasH is involved in bidirectional nitrite transport. As E. coli NarK and NarU were able to complement both narK and nasA deletions, we concluded that these proteins can operate in a number of transport modes, something which is counterintuitive based on their predicted structures and function as nitrate/nitrite antiporters. It, therefore, appears that the NarK-like proteins, despite their structural similarity, display a striking variety of activities from proteins that perform net nitrate uptake with a low background level of nitrite export (NasA) through to a protein with a predominantly nitrate/nitrite exchange activity (NarK2). The dominant activity of the latter would explain why steady state nitrate respiration in P. denitrificans is not impaired by the presence of the protonophore FCCP, which abolishes the membrane potential (Alefounder et al., 1981), whereas nitrate proton/symport activities of other transporters (for example NarK1) would explain why initiation of nitrate respiration under some conditions is retarded by the same compound (Boogerd et al., 1983).

DNA manipulations
DNA manipulations were performed by standard methods. PCRs were carried out using KOD DNA polymerase (from Thermococcus kodakaraensis) according to the supplier's instructions (Novagen) and all constructs generated were confirmed to be correct by sequencing. All oligonucleotides used in this study were synthesized by Sigma-Genosys and are shown in Table 1. The nasA gene was synthesized by GenScript and cloned into the pUC57 vector. In this construct, the ATG at position 2129 relative to that annotated at NCBI was used as a start codon because expression from a construct with the annotated start codon failed. This gene was then subcloned using EcoRI/HindIII into pEG276 for expression in P. denitrificans.

Construction of narK mutants
Inverse PCR was used to generate mutations in the narK, narK1 and narK2 genes; oligonucleotide combinations detailed in Table 1 were used. In each case, the template for these reactions was the appropriate pEG276-based clone (Table 2). Hexahistidine-tagged versions of the mutant ORFs, when needed, were generated from the appropriate clone pEG276-based clone and also encoded a C-terminal hexahistidine tag. The mutants generated are detailed in Table 2.
None of the plasmids used in this study contained any sequences that would render them under the control of anaerobic/aerobic regulatory mechanisms.

Bacterial strains, plasmids and growth conditions
Bacterial strains and plasmids used in this study are detailed in Table 2. P. denitrificans strains were grown in Luria-Bertani (LB) medium or in a defined mineral-salts medium [MM; (Robertson and Kuenen, 1983)] supplemented with 20 mM succinate as a carbon and energy source. Aerobic growth of P. denitrificans was conducted in the presence of 20 mM sodium nitrate, as described previously (Gates et al., 2011). Five millileter of growth medium in 50 ml universals was incubated in a rotary shaker at 250 r.p.m. and 37˚C. For anaerobic cultures, 300 ml of growth medium supplemented with 20 mM sodium nitrate in 300 ml bottles, were inoculated with 1% v/v of an aerobic culture, grown overnight in LB medium and cell density was determined at OD 600 . Cultures were incubated stationary at 37˚C. E. coli strains were grown aerobically in 5 ml LB medium as described for Paracoccus strains. When needed, strains were supplemented with the appropriate antibiotics at the following final concentrations: spectinomycin at 50 lg ml 21 , gentamycin at 20 lg ml 21 and carbenicillin at 100 lg ml 21 . For growth on solid media, liquid growth medium supplemented with 1.5% bacteriological agar was used. Plasmids were transferred from E. coli SM10 cells (Simon et al., 1983) to Paracoccus strains by conjugative mating as described previously (Moir and Ferguson, 1994).

Analysis of extracellular nitrite
Cells were harvested from anaerobic culture via centrifugation at 14,000 g for 1 min. Nitrite concentration in the medium was estimated colourimetrically as described previously (Nicholas and Nason, 1957).
Preparation of P. denitrificans extracts P. denitrificans strains were grown aerobically to an OD 600 of 1 before harvesting at 6,000 g for 10 min. Pellets were resuspended in BugBuster (Novagen) at 0.2 g dry pellet ml 21 and incubated at room temperature with gentle agitation for 30 mins. Extracts were subsequently used for Western blot analysis. providing the funds to sequence the genome of Paracoccus denitrificans PD1222. The authors declare no conflict of interest.