Methanobactin transport machinery

The Mbn operons and their component genes were originally identified via bioinformatics (17) (Fig. 1). We have now extended and updated this analysis with a focus on transport-related genes, resulting in a more complete model for Mbn transport. TBDTs are present in operons for four of five Mbn groups (17) (Fig. 1), consistent with the need for active transport of a large metal-bound natural product. However, the TBDTs present in Mbn operons are more heterogeneous than might be expected, given the high conservation of other Mbn-related genes. As noted previously (17), MbnT genes that neighbor genes encoding sigma and anti-sigma factor pairs also encode the N-terminal extensions necessary for signal transduction between TBDTs and anti-sigma factors (34–36). As additional Mbn operons have been identified, it has become increasingly clear that these fecIRA-like mbnIRT triads are present only in group I Mbn operons, suggesting that only those MbnTs play a role in signal transduction. A larger comparison of putative MbnTs with extant TBDTs in the Joint Genome Institute/Integrated Microbial Genomes & Microbiomes and UniProt databases indicates that the differences between MbnTs extend beyond the presence or absence of an N-terminal extension. MbnTs from group I Mbn operons (MbnT1s) are distinct in sequence and domain content from group II MbnTs (MbnT2s), and both are distinct from MbnTs in group III and IV (MbnT3s); all MbnTs appear to be distinct from previously categorized TBDT groups (SI Appendix, Fig. S3 A and B and SI Results and Discussion and Datasets S1–S3). Notably, many methanotrophs contain multiple genes encoding MbnTs, sometimes from multiple MbnT groups.

Although not described in the initial bioinformatics-based characterization of Mbn operons (17), there is some evidence that PBPs may also play a role in Mbn uptake. These proteins are periplasmic (in Gram-negative bacteria) or membrane-anchored (in Gram-positive bacteria) and are associated with ATP-binding cassette (ABC) importers (37, 38). The only Mbn operons containing PBPs are group IIA operons, which include the MbnE gene, but close homologs of the Methylocystis (Mc.) and the Ms. trichosporium OB3b mbnE genes can be found elsewhere in the genomes of other Mbn-producing methanotrophs. In some cases, multiple potential homologs are present. All these PBPs are members of solute binding protein (SBP) Family 5 (39), which comprises primarily oligopeptide-binding proteins; in a sequence similarity network of this family, MbnEs cluster separately from all characterized PBPs except a microcin C7 transporter, YejA (SI Appendix, Fig. S4A and Dataset S4). Potential homologs are observed in the genomes of most species producing group III–V Mbns, but they are also similar to YejA (40). The similarities between the two PBP subgroups (SI Appendix, Fig. S4B) complicate any definitive assignment without evidence of coregulation with Mbn operons or biochemical interactions with Mbns. However, it is notable that yejA encodes a microcin C7 transporter; microcin C7 is a ribosomally produced, posttranslationally modified natural product with a heptapeptide backbone and a final mass within the range observed for Mbns (40).

A previous set of gene regulation studies confirmed that the entirety of the mbn operon in Ms. trichosporium OB3b [mbn(E)IRTABCMNPH] (SI Appendix, Fig. S2) is coregulated in a Cu-dependent manner (32). Many of the genes in the operon lack close homologs elsewhere in the genome. However, there are several TBDTs that are closely related to mbnT as well as two other PBPs classified via bioinformatics as oligopeptide-binding proteins and possible mbnE homologs. In total, there are 45 TBDTs and 3 Family 5 PBPs in the Ms. trichosporium OB3b genome (41). Thus, to ensure that we were pursuing the candidates most likely to be related to Mbn transport, we monitored Cu-dependent gene regulation of the closest mbnE and mbnT homologs.

qRT-PCR results confirm that, although mbnE is down-regulated by Cu in concert with the Mbn operon, the two most closely related PBPs, a hypothetical protein (MettrDRAFT_1622) and yejA (MettrDRAFT_3640) (SI Appendix, Fig. S2), are not (SI Appendix, Fig. S5 and Table S1 and Dataset S5). The TBDT results are more complicated. Two mbnIRTPH operons are present elsewhere in the genome in close proximity to one another (SI Appendix, Fig. S2). Both operons (containing the mbnT1 homologs MettrDRAFT_1229 and MettrDRAFT_1241) are Cu-regulated, but down-regulation levels differ (particularly notably for the nonregulatory genes mbnPH, which are 5- to 10-fold down-regulated, compared with the 25- to 50-fold down-regulation of their mbn operon counterparts) (Dataset S5). Differences in regulation patterns after Cu addition are also observed, particularly in the operon containing the mbnT1 gene MettrDRAFT_1241: unlike the regulatory genes in the main mbn operon or the other mbnIRTPH operon, which respond to Cu addition swiftly, the mbnIR homologs in this operon are not down-regulated completely within the first 15 min, but continue decreasing at the 5-h and 24-h time points. These regulatory differences are statistically significant. Whether analyzed via hierarchical clustering or calculation of gene-to-gene Spearman correlation values, expression patterns of genes within the mbn operon correlate significantly more closely to other genes within that operon than they do to genes outside of that operon (SI Appendix, Fig. S5 and Table S1 and Dataset S5). These differing responses to Cu suggest that, even if the MbnT1s encoded by these operons are capable of binding Mbn, the systems are not fully interchangeable with the native mbn operon. Further work will be necessary to identify the extent of regulatory crosstalk between the three systems and native substrate-triggered regulation and/or uptake. However, these gene-regulation results provide support for the preliminary assignment of the originally identified mbnE and mbnT genes (MettrDRAFT_3413 and MettrDRAFT_3421) as the PBP and TBDT most likely to function in import of the native Mbn. A possible function of the close MbnT1 homologs could be uptake of nonnative CuMbns produced by other methanotrophs in the environment.

We took a two-pronged approach to definitively assign MbnT’s function: loss-of-function experiments using a gene disruption mutant of Ms. trichosporium OB3b and gain-of-function experiments using a heterologous host. For the first approach, we used marker exchange mutagenesis to prepare a strain of Ms. trichosporium OB3b in which mbnT was replaced with a gene conferring gentamicin resistance (SI Appendix, Fig. S6). Both WT and ΔmbnT cells were incubated with ⁶⁵CuMbn (final concentration of 5 μM) for 2 h followed by washing and inductively coupled plasma mass spectrometry (ICP-MS) analysis. Natural Cu has two stable isotopes, ⁶³Cu and ⁶⁵Cu, in a 2.3:1 ratio. By adding ⁶⁵CuMbn to cells, we can monitor the levels of ⁶⁵Cu directly and correlate the values to the amount of ⁶⁵CuMbn in the cells. Our results show a clear difference in the levels of ⁶⁵Cu in WT Ms. trichosporium OB3b and ΔmbnT Ms. trichosporium OB3b (Fig. 2A), with the ΔmbnT strain exhibiting much lower levels of ⁶⁵Cu. This result supports the experiments reported recently by Gu and coworkers (29).

We then expressed MbnT in E. coli (SI Appendix, Fig. S7) and tested for uptake of CuMbn. E. coli was previously shown to be incapable of CuMbn uptake (24). In this experiment, MbnT expression was induced in E. coli cells that were then incubated for 3 h before ⁶⁵CuMbn addition (final concentration of 5 μM), followed by incubation for 30 min, washing, and ICP-MS analysis. The data reveal a clear difference in the ⁶⁵Cu content of cells expressing MbnT and cells containing only the empty vector (Fig. 2B). These combined data provide strong in vivo evidence for the identification of MbnT as the CuMbn importer.

We then investigated the interaction of MbnT with CuMbn in vitro. MbnT was expressed in E. coli, solubilized in n-dodecyl-N,N-dimethylamine-N-oxide and purified via Ni-affinity chromatography (SI Appendix, Fig. S8). Size-exclusion chromatography suggests that the purified protein exists in multiple oligomeric states. Binding of Ms. trichosporium OB3b CuMbn to MbnT was assessed using surface plasmon resonance (SPR) (42). SPR is a detergent-free means to characterize ligand binding that has been used previously to characterize TBDTs (43). Purified MbnT was immobilized on a sensor chip via amine coupling, and solutions of CuMbn were applied. Binding of CuMbn to the immobilized MbnT elicits a response that is not observed with the application of buffer solutions (Fig. 3A). The response curves were fit with an equation describing the approach to equilibrium with pseudo-first-order kinetics (44); the pseudo-first-order rate constants, k_obs, obtained in these traces directly correlate to the concentration of CuMbn. A plot of k_obs versus the concentration of CuMbn can be fit to a line (Fig. 3B) where the slope is the rate constant for association (k_on) and the intercept is the rate constant for dissociation (k_off) (44). An equilibrium constant for dissociation, K_d, of 6.22 μM, was obtained. The determined K_d value is higher than those reported for other TBDTs binding to their cognate substrates (45–48). However, most of those studies were performed in vivo using native TBDTs whereas MbnT was isolated from a heterologous host, and in vitro binding experiments were conducted in the absence of membranes or detergents.

The specificity of Ms. trichosporium OB3b MbnT for its cognate CuMbn was investigated via SPR experiments in which cobalamin (a proxy for a large, metal-containing compound), ferric pyoverdine (a proxy for a peptide-derived, metal-containing natural product), and ferric citrate (a small, metal-containing compound) were applied to immobilized MbnT. None of these molecules gave an appreciable SPR response (SI Appendix, Fig. S9A), indicating that they do not bind to MbnT. To further ensure that observed interactions result from specific recognition of CuMbn by MbnT, we immobilized myoglobin (a protein that should not bind CuMbn or Cu) to the SPR sensor chip and applied CuMbn. No SPR response was observed for any concentration of CuMbn (SI Appendix, Fig. S9B). To probe the specificity of a given MbnT for its cognate CuMbn, we also cloned, expressed, and purified the MbnT from Mc. rosea SV97. Although Mc. rosea SV97 CuMbn has two nitrogen-containing heterocycles and two thioamide groups like Ms. trichosporium OB3b CuMbn, there is no N-terminal transamination or disulfide formation, the first oxazolone is replaced by a pyrazinedione, and a sulfonated threonine is present (SI Appendix, Fig. S1B) (14, 16). Interestingly, the addition of Ms. trichosporium OB3b CuMbn to Mc. rosea SV97 MbnT does elicit an SPR response (Fig. 3C), yielding a K_d value of 17.6 μM (Fig. 3D). These data suggest a complex mechanism for CuMbn recognition and binding, wherein MbnT binds only structurally similar substrates, and individual MbnTs may bind cognate CuMbns with higher affinity than nonnative CuMbns. Similar trends are observed for the ferric pyoverdine receptor FpvA from Pseudomonas aeruginosa ATCC 15692, which, unlike other Pseudomonas TBDTs, binds noncognate pyoverdines, albeit with different affinities (28).

To investigate whether MbnE might play a role in binding CuMbn in the periplasm, perhaps for delivery to periplasmic enzymes or, alternatively, to an inner membrane ABC transporter (37), we cloned, expressed, and purified Ms. trichosporium OB3b MbnE. Most Ms. trichosporium OB3b MbnE is expressed in inclusion bodies, but a small amount remained soluble and was purified via affinity chromatography. Size-exclusion chromatography revealed the protein to be a single oligomer (SI Appendix, Fig. S10), and circular dichroism spectroscopy suggests that it is mostly composed of α-helices and unordered loops (SI Appendix, Fig. S11).

Binding of Ms. trichosporium OB3b CuMbn to its cognate purified MbnE was then tested via SPR. The application of CuMbn elicits an SPR response with k_obs directly proportional to the concentration of CuMbn (Fig. 4A), up to 30 μM. At 60 μM CuMbn, the k_obs diverges from linearity (Fig. 4B), perhaps due to inactivation that reduces the protein’s ability to bind substrate. The plot of k_obs vs. [CuMbn] fits to a line that yields k_on of 1.4 × 10⁻³ μM⁻¹⋅s⁻¹, k_off of 8.5 × 10⁻³ s⁻¹, and a K_d value of 6.07 μM. This value is within the range that has been observed for binding of oligopeptides to PBPs (K_d values of 0.3 nM–50 μM) (49, 50). The specificity of MbnE for CuMbn was investigated by adding cobalamin, ferric pyoverdine, and CuSO₄ to the immobilized protein. In these experiments, we observed rapid association followed by an even faster dissociation (SI Appendix, Fig. S12A), indicating that these molecules do not bind to an appreciable extent. Therefore, MbnE, like MbnT, is specific for CuMbn and does not bind soluble Cu, siderophores, or metal chelates.

To investigate whether MbnE recognizes and binds noncognate CuMbns, we cloned, expressed, and purified MbnEs from Mc. hirsuta CSC1 and Mc. parvus OBBP. The Mc. hirsuta CSC1 mbnE gene is found within the Mbn operon (SI Appendix, Fig. S2 and Dataset S6); there are two close nonoperon mbnE homologs in Mc. parvus OBBP, but the one used in this study, O5ODRAFT_02185, is more similar to group II mbnEs (SI Appendix, Fig. S2). Both MbnEs are monomeric (SI Appendix, Fig. S13), and the Mc. hirsuta CSC1 MbnE comprises primarily α-helices and unordered loops (SI Appendix, Fig. S14). Binding of cognate CuMbn to the Mc. hirsuta CSC1 MbnE was demonstrated by SPR experiments in which semipurified CuMbn (spent medium from the Cu-starved cultures supplemented with CuSO₄ and purified on a Diaion HP-20 column) was applied to the immobilized protein. SPR responses were observed (Fig. 4C), and the rates and amplitudes of the response curves were directly proportional to the amount of semipurified CuMbn applied. Using the absorbance at 322 nm and the reported extinction coefficient at that wavelength (14), we estimate the amounts of applied CuMbn to be 5, 15, and 30 μM each. A plot of the k_obs vs. these concentrations of CuMbn can be fit to a line with a k_on of 1.9 × 10⁻³ μM⁻¹⋅s⁻¹ and k_off of 2.3 × 10⁻² s⁻¹ determined, yielding a K_d of 12.1 μM (Fig. 4D). Like Ms. trichosporium OB3b MbnE, Mc. hirsuta CSC1 MbnE does not bind cobalamin, ferric pyoverdine, and CuSO₄, but surprisingly, it also does not bind the semipurified CuMbn from Ms. trichosporium OB3b (SI Appendix, Fig. S12B). These data suggest that MbnEs may be less promiscuous than MbnTs in their ability to recognize and bind noncognate CuMbns.

We also investigated the interactions between Mc. parvus OBBP MbnE and CuMbn by SPR. There are two distinct Mbn operons and MbnEs in Mc. parvus OBBP (SI Appendix, Fig. S2). The first operon is a group I Mbn operon, containing a 16-amino-acid-long precursor core peptide. The second is a group IIb operon that has a shorter, eight-residue core peptide. There is at least one Cu-binding compound that can be isolated from the spent medium of Cu-starved Mc. parvus OBBP cells via a Diaion HP-20 column (SI Appendix, Fig. S15): a low molecular weight species that is likely to be the group IIb CuMbn. We obtained Diaion- and HPLC-purified (SI Appendix, Fig. S15A) samples of this low-molecular-weight compound in the presence and absence of Cu. It has an Mbn-like, Cu-responsive, UV-visible spectrum (SI Appendix, Fig. S15B), and a Cu-bound form of this species appears after Cu addition (SI Appendix, Fig. S15 C and D). A second higher-molecular-weight species was observed in some samples (SI Appendix, Fig. S15C), but low yields and apparent instability render it difficult to ascertain if this is the group I Mbn. Application of fractions containing the HPLC-purified Cu-binding compound to immobilized MbnE results in an SPR response, and the response is dependent on the amount of the samples added (Fig. 4E). Application of the semipurified CuMbn from Ms. trichosporium OB3b did not give an appreciable response, nor did cobalamin, ferric pyoverdine, and CuSO₄ (SI Appendix, Fig. S12C). The fact that a response is also observed with the low-molecular-weight Cu-binding compound (Fig. 4F) strongly suggests that this compound binds specifically to MbnE. Thus, we have demonstrated specific interactions between three different MbnEs and their cognate CuMbns and no apparent cross-recognition by MbnE of CuMbns from different species.

As a first step toward defining the molecular basis for CuMbn recognition by MbnE, we crystallized and determined the 1.9-Å resolution structure of Mc. parvus OBBP MbnE (SI Appendix, Table S2) in the absence of CuMbn. This apo structure (residues 35–609) reveals two major domains (Fig. 5A). The N-terminal domain includes residues 35–313, and the second domain includes residues 314–575. The C-terminal residues 576–609 form a β-strand and a loop that are part of the N-terminal domain. The domain–domain interface includes several extended loops that likely rearrange during the binding of substrate. There is a cavity at the domain–domain interface that measures ∼15 × 30 Å (Fig. 5B) with solvent access from one side only. To determine whether this cavity is large enough to accommodate CuMbn or CuMbn-like molecules, we used the GRAMMX server to generate docking models with the crystal structures of the structurally distinct CuMbns from Ms. trichosporium OB3b and Mc. sp. M. These models indicate that the interdomain cavity is indeed of sufficient size to bind either CuMbn (Fig. 5 C and D).

A Dali search (51) with the apo Mc. parvus OBBP MbnE structure revealed several PBPs and SBPs structurally similar to Mc. parvus OBBP MbnE. Of particular interest is the structure of the SBP AppA from Bacillus subtilis (PBD 1XOC, rmsd 1.151 Å for 122 Cα atoms). AppA was copurified and crystallized with a nonapeptide (52). Surface representations of both structures reveal that MbnE is in an open conformation, whereas AppA is in a closed conformation with its nonapeptide substrate inaccessible to solvent (SI Appendix, Fig. S16). There are two major structural differences between the proteins: a loop in the N-terminal domain of the Mc. parvus MbnE is ordered into an α-helix in AppA, and an α-helix in the C-terminal domain is shifted significantly toward the cavity in AppA (Fig. 5E). These regions of AppA contact the bound nonapeptide and shield it from solvent, and this structural comparison provides significant insight into how MbnEs might bind CuMbn.

Since the discovery of Mbns over a decade ago, these natural products have been postulated to play a central role in methanotroph Cu acquisition (19, 53, 54). Although uptake of intact CuMbn by methanotrophs has been demonstrated (24), the detailed molecular mechanisms of its recognition, import, and export have not been elucidated. Here we show that both the MbnT TBDT and the MbnE PBP encoded by Mbn operons are involved in Mbn transport. Disruption of the gene encoding MbnT in Ms. trichosporium OB3b impairs Mbn import, and expression of MbnT in E. coli confers the ability to import CuMbn. In addition, purified MbnT and MbnE interact with their cognate CuMbns with similar affinities. Notably, MbnT can recognize noncognate CuMbns whereas MbnE is not similarly promiscuous. The crystal structure of apo MbnE provides further insight into how MbnEs might bind their cognate CuMbns. These studies provide a framework for understanding not only Cu acquistion by methanotrophs, but also the transport of Mbn and related molecules by a range of bacteria.