A multi-iron enzyme installs copper-binding oxazolone/thioamide pairs on a nontypeable Haemophilus influenzae virulence factor

Significance

Nontypeable Haemophilus influenzae (NTHi) is a human pathogen that causes ear and lung infections. Unlike typeable H. influenzae, which is targeted by the Hib vaccine, NTHi remains a human health threat. We report that an essential virulence factor is posttranslationally modified by an enzyme containing a nonheme multi-iron cofactor. This iron enzyme belongs to an emerging family associated with natural product biosynthesis. Here, we show that the enzyme converts six cysteine residues on the virulence factor to oxazolone/thioamide pairs and that these modified residues bind copper. This study provides insight into the maturation of a virulence factor, suggests a role for copper at the host–pathogen interface during NTHi infection, and may lead to alternative strategies to combat NTHi infection.

Abstract

The multinuclear nonheme iron-dependent oxidases (MNIOs) are a rapidly growing family of enzymes involved in the biosynthesis of ribosomally synthesized, posttranslationally modified peptide natural products (RiPPs). Recently, a secreted virulence factor from nontypeable Haemophilus influenzae (NTHi) was found to be expressed from an operon, which we designate the hvf operon, that also encodes an MNIO. Here, we show by Mössbauer spectroscopy that the MNIO HvfB contains a triiron cofactor. We demonstrate that HvfB works together with HvfC [a RiPP recognition element (RRE)-containing partner protein] to perform six posttranslational modifications of cysteine residues on the virulence factor precursor peptide HvfA. Structural characterization by tandem mass spectrometry and NMR shows that these six cysteine residues are converted to oxazolone and thioamide pairs, similar to those found in the RiPP methanobactin. Like methanobactin, the mature virulence factor, which we name oxazolin, uses these modified residues to coordinate Cu(I) ions. Considering the necessity of oxazolin for host cell invasion by NTHi, these findings point to a key role for copper during NTHi infection. Furthermore, oxazolin and its biosynthetic pathway represent a potential therapeutic target for NTHi.

Ribosomally synthesized, posttranslationally modified peptide natural products (RiPPs) are a broad category of natural products with diverse biological roles. In RiPP biosynthesis, genome-encoded precursor peptides are posttranslationally modified by processing and tailoring enzymes, which install chemical moieties that confer potent activities (1–5). As such, classes of different RiPPs are often defined by specific posttranslational modifications (PTMs) or by the enzymes that confer them (6, 7). A recently established metalloenzyme family involved in RiPP biosynthesis is the multinuclear nonheme iron-dependent oxidase family (MNIO, Pfam PF05114), formerly called the domain of unknown function (DUF) 692 proteins. The first MNIO to be recognized was MbnB, which oxidizes two cysteine residues in its precursor peptide to oxazolone and thioamide pairs for the biosynthesis of the copper-scavenging natural product methanobactin (8–10). Since then, genome mining has led to the identification of additional MNIOs that perform diverse reactions, including heterocycle and macrocycle formation (11), carbon excision (12, 13), and amino acid cleavage (14, 15) (summarized in SI Appendix, Table S1).

The emerging view of MNIOs involves the formation of a heterodimer between the MNIO and its “partner” protein, which typically includes a winged helix-turn-helix domain known as a RiPP recognition element (RRE), responsible for binding of the precursor peptide (16). These partners include proteins that are homologous to DUF2063 and DUF1772 family proteins (8, 11, 15, 17). Alternatively, one MNIO, MovX, includes an additional domain resembling an RRE and does not need a partner (14). The MNIOs form a TIM barrel structure, and at the center lies a conserved metal-binding site that accommodates up to three iron ions (10, 17–19). The active cofactor of MbnB is a mixed-valent dinuclear Fe(II)/Fe(III) cluster (10); thus, questions remain regarding the function, if any, of the third iron ion. Regardless of the role of the third iron ion, other MNIOs have been reported to activate oxygen and modify substrates in the absence of additional reductant, implicating a mixed-valent cofactor in their reactions as well. Most MNIOs perform oxidations of cysteine residues in the precursor peptide (8, 11, 12) with the exception of two, MovX and ApyH, which modify the C-termini of their respective precursor peptides (14, 15).

Recently, Ahearn et al. reported a new virulence factor termed NTHI1441 from nontypeable Haemophilus influenzae 86-028NP (NTHi) (20). NTHi is the primary pathogen responsible for ear infections and exacerbations of chronic obstructive pulmonary disease (COPD), which affects 6.4% of American adults and is the fourth leading cause of death worldwide (21, 22). Knockout of NTHI1441 resulted in a 76% decrease in the ability of NTHi to invade human bronchial epithelial cells (20). Based on this observation, combined with the presence of surface-exposed NTHI1441 epitopes and detection of antibodies to NTHI1441 in the serum from individuals experiencing COPD exacerbations, NTHI1441 was posited to be a secreted virulence factor mediating host-cell entry and infection (20).

NTHI1441 is expressed as part of an operon that includes three other genes of unknown function (Fig. 1A). Notably, downstream of NTHI1441 are NTHI1443 and NTHI1444, which are predicted to encode an MNIO and an RRE-containing DUF2063-like protein, respectively (SI Appendix, Fig. S1). This observation led us to hypothesize that NTHI1441 represents a RiPP precursor peptide that is posttranslationally modified by the MNIO and its partner protein. NTHI1441, which bears no homology to methanobactins or any other known RiPPs, comprises 95 amino acids, with the first 24 residues predicted to be a signal peptide for secretion outside the cell. NTHI1441 contains eight cysteine residues, with six appearing in a C-terminal repeated AEGKCGEGKCG motif (Fig. 1B, R1 to R3). These cysteines are potential sites of PTM by NTHI1443.

Fig. 1.

The operon structure and amino acid sequence of NTHI1441. (A) The NTHI1441 operon is composed of four genes: NTHI1440, predicted to encode a DoxX protein; NTHI1441, annotated as a hypothetical protein; NTHI1443, an MNIO (DUF692) family protein; and NTHI1444, a protein with structural homology to DUF2063 proteins. (B) The sequence of NTHI1441 is composed of 95 amino acids, which we designate as the signal sequence (S, blue), the N-terminal sequence (N, purple), and the C-terminal repeating AEGKCGEGKCG motif (R1, R2, and R3, green). Cysteine residues are highlighted in red and labeled with their position numbers.

Herein, we report the characterization of NTHI1441, NTHI1443, and NTHI1444. Spectroscopic characterization of heterologously expressed NTHI1443 shows that it is isolated with a trinuclear iron cofactor. NTHI1444 forms a weak complex with NTHI1443 and readily binds the putative precursor peptide, suggesting a role for this partner protein in substrate recognition. Together with its partner protein, the MNIO carries out six four-electron oxidations of cysteine residues on NTHI1441 to produce six oxazolones and thioamides that coordinate copper. These combined data corroborate our assignment of this operon as a RiPP biosynthetic gene cluster (BGC), and thus we rename the operon the hvf BGC (for H. influenzae virulence factor) and the resultant RiPP oxazolin.

Results

HvfB Houses a Triiron Cofactor.

HvfA (NTHI1441), HvfB (NTHI1443), and HvfC (NTHI1444) can be heterologously expressed with 6xHis-tags in Escherichia coli NiCo21(DE3) and purified to homogeneity by Ni-affinity and size-exclusion chromatography (SEC) (SI Appendix, Fig. S2). Since all the iron ligands found in the other MNIOs are conserved in HvfB (SI Appendix, Fig. S1), HvfB was expressed in TB media supplemented with 150 µM ferrous ammonium sulfate (SI Appendix, Table S2). The purified protein is pale yellow, shows absorbance features consistent with a nonheme iron cofactor (Fig. 2A), and copurifies with 2.2 to 2.6 molar equivalents of Fe. The 4.2-K Mössbauer spectra of ⁵⁷Fe-enriched HvfB were measured under varying applied magnetic field (Fig. 2B). The spectra are best simulated using a four-site model with three unique high-spin (S = 5/2) ferric sites (84% of the total absorption signal) that belong to a magnetically coupled iron cluster having a total electron spin of S = 5/2 with Mössbauer parameters listed in Table 1. These parameters closely resemble those from the triiron cofactor observed in MbnBC (8). Specifically, one of the three ferric sites exhibits an ⁵⁷Fe hyperfine tensor with positive principal components, which indicates that the local electron spin of this site is antiparallel to the total spin of the cluster. In addition, the remaining two ferric sites exhibit ⁵⁷Fe hyperfine tensors with negative principal components, suggesting that the local spins of these two sites are parallel to the total spin. The fourth site constitutes a diamagnetic (S = 0) species representing 16% of the total absorption signal. Its Mössbauer parameters are consistent with high-spin ferric centers; thus, we assign it to a diferric cluster. Upon treatment with dithionite, the magnetic Mössbauer spectrum collapses into a single quadrupole doublet with parameters consistent with a high-spin (S = 2) ferrous species (Fig. 2C and Table 1). These spectroscopic features showing a mixture of di- and triiron species in HvfB are similar to what is observed for MbnB (8) but with a larger fraction of triiron cluster present in HvfB.

Fig. 2.

Spectroscopic characterization of the HvfB iron cofactor. (A) The UV–visible spectrum of 100 µM HvfB shows an absorbance at ~350 nm in the as-isolated protein (red), which decreases upon reduction with ~1 mol. equiv. sodium dithionite (blue). (B) The 4.2-K field-dependent Mössbauer spectra of 1 mM as-isolated HvfB (black vertical bars) are best fit with a four-site model (gray lines for the total simulations) corresponding to a diferric cluster (black dashed lines) and a triferric species with three distinct iron sites (red, blue, and green dashed lines), with parameters listed in Table 1. (C) Dithionite-reduced HvfB (1 mM) produces a low-temperature Mössbauer spectrum composed of a high-spin ferrous doublet (black vertical bars for the experimental data and the gray line for the spectral simulation) with parameters listed in Table 1.

Table 1.

Parameters used to simulate the 4.2-K Mössbauer spectra of HvfB as-isolated and after treatment with dithionite

Sample	Site	δ (mm/s)	ΔEQ (mm/s)	η†	Ax/gnβn, Ay/gnβn, Az/gnβn (T)	% Iron	Assignment
As isolated*	1	0.55	−0.57	3	+15.9, +15.9, +15.9	28	Triferric
	2	0.50	0.59	0	−17.6, −19.4, −19.4	28
	3	0.46	−0.40	−3	−18.2, −18.2, −19.5	28
	4	0.51	0.68	0.8	0, 0, 0	16	Diferric
Dithionite-treated‡	1	1.26	2.87	-	-	100	Ferrous

^*The current Mössbauer simulations use an S = 5/2 spin Hamiltonian to simulate the triiron cofactor with the individual iron sites listed in the table as Sites 1 to 3. The spin Hamiltonian parameters used for the current simulation are D = 0.3 cm⁻¹, E/D = 0.2, g = [2, 2, 2].

^†The asymmetric parameter of quadrupole splitting, η, normally is defined with a value range between 0 and 1 if the three principal components of the EFG tensor are chosen in the following order: |V_xx| ≤ |V_yy| ≤ |V_zz|. Here, η = 3 or −3 indicates that the largest component of the EFG tensor is oriented along y or x defined by the zero filed splitting tensor.

^‡The overall spin state and the sign of ΔE_Q are not determined for the dithionite-treated HvfB.

Complex Formation between Operon Proteins.

Some of the MNIOs characterized to date must be coexpressed with their partner DUF2063- or DUF1772-family proteins for solubility and stability (8, 12), but we were able to express and purify HvfB and HvfC individually. Combining the two proteins, however, did not show formation of a heterodimer by SEC. Only upon coexpression of the two proteins is a significant amount of heterodimer observed, although with additional peaks corresponding to the individual proteins still present (Fig. 3A). Complex formation between HvfB and HvfC was corroborated by coexpression and purification of the two proteins with only HvfB 6xHis-tagged. Purification by Ni-affinity chromatography yields both proteins (SI Appendix, Fig. S2), suggesting that HvfC binds HvfB on the column. We then used SEC to examine the ability of each protein to interact with the putative substrate, HvfA. When combined, HvfC, but not HvfB, showed an additional peak indicative of an HvfA–HvfC complex (Fig. 3B), consistent with the proposed role of other partner proteins in substrate recognition (17). Combining or coexpressing HvfA with HvfB and HvfC did not produce convincing evidence for a ternary complex by SEC, suggesting that the complex may be transient or unstable in vitro. Regardless, the formation of a ternary complex is implied by the need for all three proteins for modification activity, described below.

Fig. 3.

Complex formation between operon proteins. Proteins were separated by SEC using a Superdex 200 Increase 10/300 GL column, monitored by the absorbance at 280 nm. (A) HvfB and HvfC only form a complex upon coexpression. (B) HvfA forms a complex with HvfC, but not with HvfB. Gray dashed lines denote distinct proteins or complexes.

Detection of a PTM by Intact Mass Spectrometry.

Intact mass analysis of HvfA by liquid chromatography–electrospray ionization time-of-flight mass spectrometry (LC–ESI-TOF MS) showed two chromatographic peaks with masses 10,273.5 and 7,973.8 Da (Fig. 4 A–C, Table 2, and SI Appendix, Fig. S3). The first 24 amino acids of HvfA are predicted by SignalP (23) to encode a secretion signal with a signal peptidase I cleavage site between residues 24 and 25. The mass of the smaller peptide is consistent with the cleaved peptide (calculated mass = 7,981.90 Da), suggesting cleavage by the E. coli peptidase during heterologous expression. The relative amounts of cleaved and uncleaved peptide varied across preparations. The detected masses of the intact and cleaved peptide differ from the calculated masses by −8 Da. Considering that HvfA contains eight cysteine residues, we reasoned that the cysteine residues may be in disulfide bonds. Treatment with the reductant DTT indeed restores the detected masses to the theoretical masses (SI Appendix, Fig. S4, Table 2).

Fig. 4.

Intact protein mass spectrometry of HvfA expressed alone and coexpressed with HvfBC. (A) Representative extracted ion chromatograms showing the two species detected: Peak 1 (blue) and Peak 2 (green), extracted using m/z values shown in SI Appendix, Fig. S3. Panels (B) and (C) show that the deconvoluted masses of Peak 1 and Peak 2, respectively, from the mass spectra are shown in SI Appendix, Fig. S3. (D) The UV–Visible spectra of 50 µM HvfA_mod (red) and HvfA expressed alone (black) or coexpressed with HvfB (light blue) or HvfC (dark blue) individually. Deconvoluted masses from intact protein MS of HvfA_mod are shown for Peak 1 (E) and Peak 2 (F). The corresponding chromatograms and mass spectra are shown in SI Appendix, Fig. S5.

Table 2.

Intact masses of HvfA expressed alone (unmodified) or coexpressed with HvfB and HvfC (modified), representing the whole peptide (Peak 1) and the peptide with the signal sequence cleaved (Peak 2), with and without DTT treatment

		Peak 1		Peak 2
		Mass (Da)	Δm (Da)	Mass (Da)	Δm (Da)
Expressed alone (HvfAunmod)	As isolated	10273.5 ± 0.2	−8.2	7973.7 ± 0.3	−8.2
	+ DTT	10280.8 ± 0.3	−0.9	7981.1 ± 0.5	−0.8
Coexpressed (HvfAmod)	As isolated	10255.6 ± 0.4	−26.1	7954.3 ± 0.3	−27.6
	+ DTT	10257.1 ± 0.5	−24.6	7957.6 ± 0.3	−24.3

Mass differences (Δm) are relative to the theoretical masses (10,281.7 and 7981.9 Da for Peaks 1 and 2, respectively). The reported error represents the SD of at least three independent measurements.

Having characterized the unmodified peptide, we then investigated whether HvfA is modified by HvfB and HvfC. We adopted a heterologous coexpression approach, in which 6xHis-tagged HvfA was coexpressed with HvfB, HvfC, or both, followed by isolation of HvfA by Ni-affinity and SEC. HvfA coexpressed with HvfB and HvfC showed new absorbance features at 258 and 302 nm that were not observed when expressed alone or with either HvfB or HvfC individually (Fig. 4D). We will refer to this preparation of the peptide as HvfA_mod to differentiate it from HvfA_unmod (expressed alone). LC–ESI-TOF MS of HvfA_mod showed two peaks (SI Appendix, Fig. S5), again corresponding to the whole peptide and that with the signal peptide cleaved. Within each of these chromatographic peaks, there was a mixture of the unmodified peptide and another species with masses 10,255.6 and 7954.3 Da (Δm ≈ −26 Da from the calculated masses, Fig. 4 E and F), representing one or multiple PTMs. LC–MS of HvfA coexpressed with HvfB and HvfC individually yielded HvfA_unmod, corroborating the need for both proteins to achieve modification (SI Appendix, Fig. S6). Treatment of HvfA_mod with DTT showed a mass change of ~+2 Da, suggesting that only one disulfide bond remains in the modified peptide (Table 2 and SI Appendix, Fig. S7). This leaves −24 Da attributable to PTM(s).

Localization of the PTM to C-terminal Cysteine Residues.

To localize the PTM, multiple different truncated forms of HvfA, depicted in Table 3, were coexpressed with HvfB and HvfC and isolated. Examination of the 302 nm feature shows that each truncated peptide retains the PTM (albeit to different extents) with the exception of ΔR1-3, which lacks all three C-terminal repeats (SI Appendix, Fig. S8). The intact mass of ΔR1-3 was identical whether expressed alone or with HvfB and HvfC and is 2 Da less than the calculated mass (Table 3 and SI Appendix, Figs. S12 and S15). This difference disappears upon treatment with DTT (Table 3 and SI Appendix, Fig. S16 and Table S3), suggesting that the disulfide bond observed for HvfA_mod forms between the two N-terminal cysteine residues, C38 and C43, which are not in a conserved motif (Fig. 1B). Complementary truncations of the N-terminus (ΔS and ΔSN) retained the ~−24 Da mass difference associated with the PTM, localizing the PTM to the repeating AEGKCGEGKCG sequences (Table 3 and SI Appendix, Figs. S9 and S10). Truncation of two of the three repeats (ΔR2-3) decreased the mass difference to −10 Da, of which −2 Da can be attributed to the disulfide bond and −8 Da to the PTM (SI Appendix, Figs. S11 and S16). Individual Cys-to-Ala mutants (SI Appendix, Fig. S17 and Table S4) of residues in the repeating motif showed Δm = −22 Da, which is +4 Da relative to the wild-type (WT) protein. Therefore, the PTMs installed on HvfA_mod by HvfBC can be assigned as six −4 Da modifications. Interestingly, mutation of the cysteine residues involved in the N-terminal disulfide bond (C38A and C43A) produced a variety of lesser-modified products (SI Appendix, Table S4 and Fig. S18), which will be discussed below.

Table 3.

Comparison of the WT and truncated HvfA variants

HvfA variant		302 nm peak	Theoretical mass (Da)	Detected mass (Da)	Δm (Da)	Δm (reduced, Da)
WT		✓	10281.7	10255.6 ± 0.4	−26.1	−24.3
ΔS		✓	8113.1	8086.7 ± 0.3	−26.4	−24.6
ΔSN		✓	6101.9	6077.4 ± 0.2	−24.6	−23.6
ΔR2-3		✓	7443.5	7432.8 ± 0.3	−10.4	−8.7
ΔR1-3		✗	5394.1	5391.7 ± 0.3	−2.4	−0.4
ΔSR2-3		✓	5274.8	5270.0 ± 0.3	−10.6	−8.5
ΔSNR2-3		✓	3263.7	3255.6 ± 0.1	−8.1	−8.3

The cartoons depicting the truncations show the signal peptide in blue, the following 20 N-terminal residues in purple, the AEGKCGEGKCG motifs in green, and the intermediate sequences between the repeats as black lines. Portions of the peptide that have been truncated are shown as gray dashed lines. The presence of the 302 nm chromophore was determined from the spectra shown in SI Appendix, Fig. S8. The intact MS data are shown in SI Appendix, Figs. S9–S14 and for the DTT-treated (reduced) peptides in SI Appendix, Fig. S16 and Table S4. Δm represents the mass differences between the theoretical mass (unmodified) and the detected modified peptide. Δm (reduced) represents the mass differences from the theoretical mass and the detected mass of the DTT-treated, modified peptide.

For a more detailed analysis of HvfA, HvfA_mod and HvfA_unmod were subjected to capillary electrophoresis top-down–MS/MS (CE–TD MS/MS). The peptide with the signal sequence cleaved was more readily detected by this technique than the intact peptide (Fig. 5 and SI Appendix, Fig. S19). Using the extracted ion electropherograms (EIEs) at m/z = 664 to 666, two peaks can be resolved for HvfA_unmod (Fig. 5A). Treatment with DTT collapses these into a single peak (SI Appendix, Fig. S20). Because CE is sensitive to different proteoforms, the peaks likely correspond to different arrangements of disulfide linkages among the eight cysteines of the unmodified peptide. The MS/MS spectra localize 1 Da mass losses to Cys-containing fragments (Fig. 5 B and C and SI Appendix, Fig. S20), which disappear upon reduction by DTT (SI Appendix, Figs. S20 and S21). For HvfA_mod, CE afforded the resolution of several peaks apparent in the EIEs (Fig. 5 D–F and SI Appendix, Fig. S19). The masses of these peaks differ from each other by −2 and −4 Da and thus correspond to peptides with different numbers of PTMs (SI Appendix, Table S5). [The −2 Da difference is the net change due to +2 Da reduction of a disulfide bond and a single −4 Da modification, whereas the peaks that differ by −4 Da correspond to a second PTM after one PTM has already occurred. These differences all become −4 Da each in the DTT-treated peptide (SI Appendix, Figs. S21 and S22 and Table S6).] Similar results were obtained for the full-length peptide with the signal sequence intact (SI Appendix, Figs. S23 and S24 and Tables S7 and S8). Thus, TD MS/MS corroborates the presence of −4 Da PTMs on each of the six cysteine residues in the three AEGKCGEGKCG motifs in modified HvfA with the two N-terminal cysteine residues involved in a disulfide bond.

Fig. 5.

CE–TD MS/MS analysis of HvfA lacking the signal peptide. (A) Extracted ion electropherogram (EIE) (m/z 663-666) of HvfA_unmod. Peaks a′ and a have the same masses but different charge state distributions, potentially resulting from heterogeneity in disulfide bond formation. (B) Depiction of the unmodified peptide. This depiction of the disulfide bonds between neighboring cysteines is arbitrary; the connectivity is not known. (C) Fragmentation pattern of peak a of HvfA_unmod. Eight cysteine residues were identified with H loss (gray boxes) due to disulfide bonds. (D) EIE (m/z 663 to 666) of HvfA_mod. (E) Annotation of Cys modifications and mass shifts among separated peaks (a to g). Depictions of peaks d to f containing three to five modifications are abbreviated with an ellipsis (...). The precise locations of the PTMs in the partially modified peptides (peaks b to f) cannot be determined and are depicted left-to-right for simplicity. (F) Fragmentation pattern of peak g of HvfA_mod. The first two cysteines (gray boxes) show H loss due to disulfide bond formation. The other six cysteines (orange boxes) show 4 Da losses. In both (C) and (F), Cys residues of the AEGKCGEGKCG motif are denoted by an asterisk. Raw MS1 and MS/MS data are shown in SI Appendix, Fig. S19, and average masses for each peak are shown in SI Appendix, Table S6.

NMR Structural Elucidation of the HvfA PTM.

For structural characterization of the PTM, modified and unmodified [U-¹³C,¹⁵N]-HvfA-ΔSR2-3, which contains one AEGKCGEGKCG motif, was subjected to a suite of standard double- and triple-resonance NMR experiments that enabled almost complete (92%) backbone assignment of the peptide. Notably, the 2D ¹H-¹⁵N heteronuclear single quantum coherence (HSQC) and HNCACB spectra show the disappearance of C53 and C58 amide protons in the modified peptide (Fig. 6 A and B and SI Appendix, Fig. S25), suggesting that these cysteine amides are part of a proton-deficient modification. By contrast, the Cβ shift values of the two cysteines in the N-terminal portion of the peptide corroborate that these residues (C38, C43) are oxidized in a disulfide bond (Fig. 6B). Comparison of the HvfA_mod and HvfA_unmod ¹H-¹⁵N HSQC spectra reveal dramatic chemical shift perturbations (CSPs) in a subset of residues near the modified cysteine residues (Fig. 6A and SI Appendix, Fig. S25–S27). The residues C-terminal to the modified cysteines, G54 and G59, show the largest magnitude CSPs of ~1 (¹H δ) and ~12 ppm (¹⁵N δ) for both residues (Fig. 6A and SI Appendix, Fig. S27); the dramatic deshielding of these resonances indicates a reduction in electronegativity. The two lysines N-terminal of C53 and C58, K52 and K57, show significantly shielded Cα CSPs of ~10 ppm upon modification (Fig. 6C), suggesting that they are adjacent to an aromatic ring.

Fig. 6.

NMR characterization of the HvfA-ΔSR2-3 PTM. (A) ¹H-¹⁵N HSQC overlay for HvfA_mod (red) and HvfA_unmod (black) with largest CSPs (G54 and G59) noted. (B) Overlay of HNCACB spectra for Cys residues (C38 and C43, in disulfide bonds; C53 and C58, sites of modification). Cβ peaks are shown in black and red for HvfA_unmod and HvfA_mod, respectively. The average Cβ position for oxidized or reduced Cys, as described by ref. 24, is denoted by a red or blue box, respectively. (C) HNCACB strips for representative Lys residues (K47, distant; K52 and K57, proximal to modification), with HvfA_unmod shown in black and HvfA_mod in red. The red box is indicative of average Lys Cα shift values ~52 ppm, per ref. 24. (D) HvfA_mod HNCO with G54 and G59-coupled carbons at ~165 ppm. Inset: Magnified G54/G59 peaks with ¹J_CC of 90.99 Hz. (E) Depiction of the unmodified Lys-Cys-Gly (greens) and the modified peptide with an oxazolone and thioamide pair (blues). Detected chemical shift values are labeled. R represents the lysine sidechain. (F) G54/G59 carbon 1D traces (solid lines) showing the change in coupling because of shifting decoupling π-pulse offset (red to black). A coupled/decoupled "three peak" model fitting by Bayesian parameter estimation yields model predictions (dots). (G) Theoretical decoupling profiles for varying pulse offsets (solid lines) with model-fitted spin inversion levels shown as dots plotted on the decoupling profiles. The average position of the inversion levels on the decoupling profiles shows that the chemical shift of coupled carbon is ~122 ppm (dashed line).

The HNCO spectrum of HvfA_mod-ΔSR2-3 reveals significant perturbation in the carbonyls N-terminal to G54 and G59 (Fig. 6 D and E), with resonances occurring at ~165 ppm. These chemical shifts are significantly upfield from the typical peptide carbonyl resonance at 174 ± 3.5 ppm (24), consistent with the replacement of the carbonyl oxygen with a less electronegative atom such as sulfur. The ~165 ppm signals are split with a strong ¹J_CC coupling of about 90 Hz; the expected ¹J_C′Cα coupling for an unmodified carbonyl is approximately 55 Hz. To further probe the atom connectivity of the modification, modified HNCO spectra were recorded with decoupling pulses arrayed over a series of offset frequencies (Fig. 6F). These pulses are usually used to decouple Cα from CO, but in this case, they were used to decouple the ¹J_CC heteronuclear coupling of the carbons at ~166 ppm. As the offset frequency is arrayed over values ranging from 50 to 90 ppm, the skirt of the decoupling pulse covers a range of 80 to 160 ppm (Fig. 6G). During the series, the coupled doublet initially present (low offset values of the pulse, Fig. 6F, dark shading) collapses until a fully decoupled singlet appears (high offset values, Fig. 6F, red). The level of collapse at each offset frequency was modeled with continuous values between 1 (fully coupled) and −1 (fully decoupled). These decoupling values were then plotted onto the skirt profile of the decoupling pulse to estimate a chemical shift of ~122 ppm for the coupled carbon. Together with the −4 Da mass shift and absorbance features, these data suggest that the PTMs are oxazolone/thioamide pairs as observed previously in methanobactin and methanobactin-like products (Fig. 6E) (14, 25). The glycine shift differences likely result from the replacement of the carbonyl oxygen with a thio-enol and the shielding observed for the lysines is consistent with the formation of an oxidized aromatic ring.

Copper Binding to Modified HvfA.

Considering the role of the oxazolone/thioamide pairs of methanobactin in copper binding, we investigated the metal binding properties of HvfA_mod. As isolated, HvfA_mod did not copurify with any metals (SI Appendix, Table S9). Addition of Cu(I) to the peptide under anaerobic conditions showed a shift in the absorbance maximum from 302 to 321 nm, indicative of the interaction of the oxazolone/thioamide with copper (Fig. 7A) (26). In comparison, addition of Cu(I) to HvfA_unmod did not alter its absorption spectrum (SI Appendix, Fig. S28). Addition of Cu(II) to HvfA_mod resulted in degradation of the peptide, evidenced by disappearance of the 302 nm peak (SI Appendix, Fig. S29) and the inability to detect the intact peptide by MS. When treated with 3 mol. equiv. of Cu(I) and desalted, both forms of HvfA retain copper (2.7 eq. for HvfA_mod, 2.2 eq. for HvfA_unmod, SI Appendix, Table S9). Native MS of both peptides revealed multiply Cu-bound forms (Fig. 7B and SI Appendix, Fig. S30–S32 and Table S10), with two to three copper ions per peptide bound to HvfA_unmod and six copper ions bound to the most abundant HvfA_mod species. To assess the possibility that HvfA_unmod binds copper weakly and/or nonspecifically, we performed competition assays with the high-affinity copper chelator bicinchoninic acid (BCA) [Cu(I) log β₂ = 17 (27, 28)]. No significant competition between BCA and HvfA_unmod was observed, indicating that BCA binds Cu(I) with higher affinity. By contrast, at least six equivalents of Cu(I) were needed to saturate BCA in the presence of HvfA_mod (Fig. 7C). The heterogeneous nature of HvfA_mod precludes quantification of the Cu(I) binding affinity, but the extent of competition with BCA suggests that HvfA_mod has multiple high-affinity Cu(I)-binding sites that are not present in HvfA_unmod.

Fig. 7.

Binding of Cu(I) by HvfA. (A) Absorption spectra of the titration of [Cu(CH₃CN)₄]PF₆ into 50 µM HvfA_mod. Inset: Single-wavelength data at 302 nm (black) and 340 nm (red). (B) Comparison of native MS of Cu(I)-bound HvfA_unmod (Top) and HvfA_mod (Bottom). Mass spectra are shown in SI Appendix, Figs. S30–S32. (C) Titration of Cu(I) into BCA (blue), a mixture of BCA and HvfA_unmod (black), and a mixture of BCA and HvfA_mod (red), as monitored by the absorbance of the Cu[BCA]₂ complex at 562 nm. BCA and HvfA concentrations were both 50 µM.

Discussion

HvfA (NTHI1441) was recently determined to be a virulence factor important in the invasion of bronchial epithelial cells by the human pathogen NTHi (20). Our data show that HvfA is a RiPP precursor peptide that is posttranslationally modified by the MNIO HvfB with help from its partner protein HvfC, which is involved in substrate recognition. Spectroscopic interrogation of the iron cofactor of HvfB reveals that, as isolated, the majority of the protein contains a triiron cofactor with a minor population of diiron cluster. To date, the only MNIO to have its active iron cofactor thoroughly investigated is MbnB, which activates O₂ from the mixed-valent Fe(II)/Fe(III) state (10, 29). The role of the third iron-binding site in HvfB and other MNIOs remains unclear. Nonetheless, these results demonstrate that the formation of the oxazolone/thioamide pair by a multinuclear iron cofactor is not unique to methanobactin and may be a common PTM among MNIO-associated RiPPs.

HvfA has an N-terminal signal peptide that confers secretion by the Sec translocon and is cleaved by signal peptidase I, which is present in both E. coli and NTHi (30, 31). As such, HvfA as purified from heterologous expression in E. coli is a mixture of signal-cleaved and intact peptide, and both forms show a significant extent of PTM. Further, HvfA with the signal peptide removed (ΔS) also contains six PTMs, suggesting that the signal peptide is not necessary for recognition by the HvfBC complex. Mutation of the two residues involved in a disulfide bond in the N-terminal region of the peptide (C38 and C43), however, produces detectable partially modified products (SI Appendix, Fig. S18), suggesting that this disulfide bond may be important for recognition by HvfBC. Yet, truncating the whole N-terminal region of the peptide (ΔSN) and further truncations of the C-terminal repeats (SI Appendix, Figs. S8–S13 and Table 3) still yield modified peptide. Thus, it appears that while an intact N-terminal region is optimal, perhaps for recognition by HvfC, HvfB can still modify even very small fragments of HvfA. The function of the fourth protein encoded by the operon, a membrane protein annotated as a DoxX homolog that we term HvfX, is yet unknown. DoxX from Mycobacterium tuberculosis was reported to form a membrane-associated complex with superoxide dismutase SodA and thiol-oxidoreductase SseA, which function to alleviate oxidative stress (32). Thus, we do not expect DoxX to have a biosynthetic function.

The two oxazolone/thioamide pairs of methanobactin coordinate Cu(I) via the nitrogen atoms of the oxazolones and the thiolate sulfurs in a four-coordinate, distorted tetrahedral geometry (28). By analogy, mature HvfA, which we name oxazolin, contains six oxazolone/thioamide pairs and thus would be expected to bind three copper ions. Surprisingly, both HvfA_mod and HvfA_unmod bind multiple Cu(I) ions. Together, the native MS and BCA competition experiments suggest that HvfA_unmod can bind ~2 to 3 copper ions while HvfA_mod can bind ~6 to 7 copper ions. However, CE-MS shows that HvfA_mod is a mixture of peptides containing 0 to 6 PTMs, and integration of the electropherogram peaks (Fig. 5D) demonstrates that only ~35% of the peptide is fully modified, for an average ~2.4 modifications per peptide. Thus, an average Cu(I) binding stoichiometry by the HvfA_mod of 1 to 2, rather than 3, might be expected. The changes in the 302 nm chromophore level off with ~2 equivalents of copper (Fig. 7A), suggesting that as copper is added to HvfA_mod, it first binds to the oxazolone and thioamide moieties, with additional equivalents binding elsewhere, likely at low-affinity sites present in partially modified forms of HvfA. While it is difficult to parse the exact stoichiometry due to the mixture of species in HvfA_mod, the data are overall consistent with oxazolin binding 1 Cu(I) per pair of PTMs (Fig. 8), like methanobactin.

Fig. 8.

Model structure of copper-bound oxazolin. The bracketed portion is repeated three times. The variable residues between the repeated motifs are omitted and represented as ellipses (...).

The functional role of oxazolin is not yet known. One possibility is that it serves as a chalkophore, or copper-chelating metallophore, to scavenge copper for NTHi, like methanobactin does for methanotrophic bacteria (9, 33–35). As an essential nutrient, iron is sequestered by host cells as a defense against pathogens, a phenomenon known as nutritional immunity, and is thus scavenged by siderophores and iron acquisition systems by invading pathogens like NTHi (36–38). It may be that oxazolin is secreted by NTHi to acquire nutritional copper from the host and/or environment. An alternative hypothesis is that oxazolin serves as protection against toxic levels of copper. During infection, hosts stave off bacterial infection by accumulating levels of copper toxic to the bacterial pathogen (39, 40). Related to this notion, a recent preprint reports that an MNIO-produced natural product from Caulobacter vibrioides termed bufferin binds copper and is up-regulated under copper stress (41). Considering its importance for human cell invasion (20), oxazolin may be necessary to protect against high copper levels within the host cell. In support of this theory, mouse models of NTHi infection link up-regulated copper export and virulence, implying NTHi may experience host-induced copper toxicity (42). Another study has linked expression of the hvf operon to oxidative stress (43), which can be caused by copper. Investigation of the hvf operon regulation in response to copper and further in vivo studies will be required to establish the function of oxazolin. Notably, searches of the oxazolin precursor peptide HvfA against genomic databases identified >3,000 sequences, showing that the operon is conserved in a range of bacteria, including pathogens such as Pseudomonas aeruginosa and Neisseria gonorrhoeae (SI Appendix, Table S11). The identification of oxazolin and its biosynthetic pathway presented here could thus lead to alternative strategies for preventing and treating infections by NTHi as well as other bacteria.

Materials and Methods

Construct Design.

The genes encoding NTHI1441 (HvfA), NTHI1443 (HvfB), and NTHI1444 (HvfC) were codon-optimized and synthesized by GenScript. The optimized gene sequences are listed in SI Appendix, Table S12 with added restriction sites underlined. Details for the subcloning of pET21a(+)-hvfA, pRSFDuet-1-hvfB, pRSFDuet-1-hvfB, pRSFDuet-1-hvfBC, and pET28a(+)-TEV-hvfC are in the SI Appendix.

Protein Expression.

Briefly, expression of all proteins was carried out in E. coli NiCo21(DE3) cells, which were grown at 37 °C in Terrific Broth to mid-log phase, induced with IPTG, and cooled to 16 to 18 °C for overnight expression. More detailed methods are available in the SI Appendix.

Protein Purification.

Ni-affinity chromatography was used to isolate 6xHis-tagged HvfA expressed from pET21a(+)-hvfA. Cell paste was resuspended in phosphate-buffered saline (PBS) at pH 7.8 (~10 mL per g cell paste) containing 10 mM imidazole, 1 mM PMSF, 10 µg/mL DNase I, and cOmplete EDTA-free Protease Inhibitor Cocktail (one tablet per 100 mL, necessary to prevent peptide degradation), which was stirred on ice. Cells were lysed using a Branson sonicator at 4 °C, and the lysate was centrifuged at 100,000×g for 30 min. The cleared lysate was loaded onto a Ni-NTA column at room temperature by gravity and washed with PBS pH 7.8 with 25 mM imidazole. Protein was eluted with PBS pH 7.8 with 300 mM imidazole while collecting fractions. Fractions containing the protein of interest were identified by SDS-PAGE and pooled. The protein was concentrated using an Amicon 10 K MWCO centrifugal filter and further purified by SEC using a Superdex 200 10/300 GL column (GE HealthCare) in PBS pH 7.8 at 4 °C on an ÄKTA Pure FPLC. 2 mL of protein were injected onto the column at a flow rate of 1 mL/min, and the eluate was collected in 4 mL fractions. Fractions containing HvfA were again identified by SDS-PAGE, pooled, and concentrated, and aliquots were flash-frozen and stored at −80 °C until further use. When HvfA was coexpressed with HvfB, HvfC, or both, SEC was used to separate HvfA from the other proteins. Fractions containing only HvfA were saved, while fractions containing a mixture of proteins were discarded.

Ni-affinity chromatography was used to purify HvfB and HvfC expressed from pRSFDuet-1-hvfB and pRSFDuet-1-hvfBC, leveraging the 6xHis tag of HvfB, and from pET28a(+)-TEV-hvfC, which produces 6xHis-tagged HvfC. These purifications were carried out in an identical manner to HvfA with the inclusion of 5% glycerol in all the purification buffers for HvfB. When HvfB and HvfC were coexpressed, both proteins were obtained by Ni-affinity chromatography, as shown by SDS-PAGE. SEC was used to isolate the protein complex as indicated by a chromatographic peak with absorbance at 280 and 330 nm (the latter of which comes from HvfB) and by SDS-PAGE.

Metal Quantification.

Iron content was measured by the ferrozine assay (44), and other metals were measured by inductively coupled plasma mass spectrometry (ICP–MS). Detailed methods are available in the SI Appendix.

Site-Directed Mutagenesis and Truncations of HvfA.

Cysteine-to-alanine mutants of HvfA were generated by whole-plasmid amplification using semioverlapping primers containing the desired mutation (SI Appendix, Table S13). Truncations of HvfA were generated by PCR amplifying the whole plasmid and gene, excluding the portion of the gene sequence to be truncated, and the product was ligated via Gibson assembly. More detailed methods are available in the SI Appendix.

Complex Formation by SEC.

SEC experiments were carried out using an ÄKTA Pure FPLC and a Superdex 200 Increase 10/300 GL (Cytiva) column at 4 °C with a flow rate of 0.4 mL/min. Then, 500 µL aliquots of 300 µM HvfB, HvfC, or both (from coexpression and combination of individually expressed proteins), with and without the addition of 5 mol. equiv. of HvfA, were injected onto the column, and the absorbances at 280 and 330 nm were monitored. The approximate molecular weights of each protein and protein complex were estimated by comparison of the retention volumes to those of protein standards.

UV–Visible Spectroscopy.

UV–visible spectroscopy was performed on a Cary 3500 Compact Peltier UV–visible spectrophotometer at concentrations of 50 µM HvfA and 100 µM HvfB. All spectra were recorded at room temperature. To measure spectra of reduced HvfB, HvfB was rendered anaerobic by purging a sealed vial with N₂ gas for 30 min while gently stirring. Following transfer into a Coy anaerobic chamber, 0.5 µM methyl viologen dichloride and 2.5 mol. equiv. of sodium dithionite were added to the sample, each from anaerobic stock solutions. The protein was then placed in a sealed vial and removed from the anaerobic chamber to measure the UV–visible spectrum. Full reduction of the protein was confirmed by disappearance of the 315 nm chromophore of reduced dithionite.

Mössbauer Spectroscopy.

Preparation of ⁵⁷Fe-labeled protein is described in the SI Appendix. Mössbauer spectroscopy was performed with two spectrometers using Janis Research SuperVaritemp dewars (Wilmington, MA), which allow studies in applied magnetic fields up to 8 T at temperatures ranging from 1.5 to 200 K. The reported isomer shifts are relative to iron metal at 298 K. Spectral simulations were performed using the WMOSS software package (SEE Co, Edina, MN) and SpinCount software with the following S = 5/2 spin Hamiltonian equations, and all spectral figures were prepared using SpinCount software (45).

$$\begin{array}{c}\mathcal{H}=D\left({\stackrel{\rightharpoonup }{S}}_z^2-\frac{S(S+1)}{3}\right)+E\left({\stackrel{\rightharpoonup }{S}}_x^2-{\stackrel{\rightharpoonup }{S}}_y^2\right)+\beta \stackrel{\rightharpoonup }{S}\bullet \stackrel{\sim }{g}\bullet \stackrel{\rightharpoonup }{B}\hfill \\ +{\displaystyle \sum _i}\stackrel{\rightharpoonup }{S}\bullet \stackrel{\sim }{A}\bullet \stackrel{\rightharpoonup }{I_l}-{\displaystyle \sum _i}{g}_N{\beta }_N\stackrel{\rightharpoonup }{B}\bullet {\stackrel{\rightharpoonup }{I_l}}_i+{\mathcal{H}}_Q\hfill \end{array}$$

$${\mathcal{H}}_Q=\sum _i\frac{eQ{V}_{\mathit{zz}}}{4}\left[{\stackrel{\rightharpoonup }{I_{z,i}}}^2-\frac{I_i\left(I_i+1\right)}{3}+\frac{\eta }{3}\left({\stackrel{\rightharpoonup }{I_{x,i}}}^2-{\stackrel{\rightharpoonup }{I_{y,i}}}^2\right)\right].$$

Intact Protein Mass Spectrometry.

Intact mass spectra were measured using an Agilent 6230 LC–TOF operated in positive ion ESI mode equipped with an Agilent Series 1200 HPLC binary pump, an Agilent 1200 series autoinjector, and a 10 cm C18 column maintained at 35 °C. Samples contained approximately 50 µM HvfA in PBS containing 10 mM DTT if specified, and sample injections of 3 µL were used. Mobile phases were water with 0.1% formic acid (buffer A) and acetonitrile with 0.1% formic acid (buffer B). Separation was achieved using the following method: 100% buffer A isocratic for 0.5 min with a flow rate of 0.4 mL/min, gradient of 0 to 100% buffer B from 0.5 to 5.0 min, 100% buffer B isocratic until 7.25 min, gradient from 0 to 100% buffer A from 7.25 to 7.5 min and the flow rate was increased to 0.5 mL/min, 100% buffer A until 10.0 min. A blank (10 µL injection) was run in between each sample to prevent carryover. Data were analyzed using Agilent MassHunter BioConfirm 10.0. Chromatographic peaks were selected at the highest points to extract the mass spectrum. The mass was deconvoluted by the maximum entropy algorithm using the m/z range of 500 to 2,000 with an expected mass range of 2,000 to 20,000 Da and a mass step of 0.5 Da. Samples were measured in triplicate, and the reported mass values represent the average of at least three replicates.

Capillary Electrophoresis–Tandem Top-Down Tandem Mass Spectrometry (CE–TD MS/MS).

A CESI 8000 Plus CE system (Sciex) was coupled to an Orbitrap Exploris 480 spectrometer (Thermo Fisher Scientific) via an in-house-constructed electrokinetically pumped sheath-flow CE-MS nanospray interface for CE-MS/MS. The glass emitter (orifice size: 20 to 30 µm) on the interface was filled with sheath buffer (0.2% formic acid and 10% methanol) to generate electrospray at 2.2 to 2.5 kV. A 1-m linear polyacrylamide (LPA)-coated capillary (30 µm I.D., 360 µm O.D., 85 cm length) was used for CE separation. The inlet of the capillary was installed in the CE system and the outlet was inserted into the interface emitter. For CE, the capillary was filled with 10 nL (5 psi for 15 s) of HvfA sample (in PBS pH 7.8, containing 10 mM DTT if specified), followed by inserting the capillary inlet into a background electrolyte containing 5% acetic acid. Afterward, 30 kV of voltage was applied to the capillary for separation.

The global conditions of the mass spectrometer were set to ion transfer tube temperature of 320 °C, in-source fragmentation energy of 15 V, RF lens of 60%, and intact protein mode turned on at low-pressure mode. A full MS scan was acquired at low resolution of 120,000 (at m/z of 200), m/z range of 450 to 2,000, normalized AGC target of 1,000%, and a microscan setting of 1. The MS/MS data were collected in data-dependent acquisition (DDA) mode. The top three most intense precursors (charge states of 5 to 60, minimal intensity of 10,000, and “auto” maximum injection time) in full MS spectra were selected for fragmentation (isolation window of 3 m/z and HCD energy of 25%). Other MS/MS settings include a resolution of 60,000 (at m/z 200), m/z range of 300 to 2,000, a microscan setting of 3, and normalized AGC target of 1,000%, with an auto maximum injection time. Dynamic exclusion was applied with a duration of 60 s, with exclusion of isotopes enabled.

For intact mass analysis, the electropherograms were analyzed by Thermo Xcalibur Qual Browser. Deconvoluted mass spectra were generated by averaging across the peak of interest in the electropherogram and processing with the Xtract algorithm. MS/MS data were imported into tdValidator for identification of protein sequences and modifications.

Sample Preparation for NMR.

For spectral simplification, the shortened HvfA-ΔSR2-3 peptide was chosen for structural elucidation by NMR. (Expression of the shorter peptide ΔSNR2-3 is low, and it could not be prepared in quantities needed for NMR.) Growth conditions in ¹³C/¹⁵N-labeled M9 media are in SI. To prepare the samples, the purified protein was concentrated (2.5 mM modified peptide, 5 mM unmodified), as quantified by the BCA assay, and exchanged into 100 mM PBS at pH 7.0. To each sample, D₂O was added at a final concentration of 5%. The samples were then transferred into 5 mm NMR tubes for data collection. Samples were stored at 4 °C in the NMR tubes between experiments.

NMR Data Collection.

All NMR data were collected on a Bruker NEO 600 MHz instrument equipped with a Z-gradient-enabled ¹H, ¹⁹F, ¹³C, ¹⁵N QCI-F cryoprobe. For both modified and unmodified peptide samples, various triple resonance experiments, including ¹H-¹⁵N HSQCs, HNCOs, and HNCACBs, were collected at 10 °C. The ¹H-¹⁵N HSQC spectra (hsqcfpf3gpphwg) were acquired at a size of 2,048 × 256 points using double quadrature detection (DQD) and States-TPPI in the F2 and F1 dimensions, respectively, with a total of eight scans per FID for HvfA_mod and four scans for HvfA_unmod. The HvfA_mod spectrum has a spectral width of 16.03 ppm and 18.00 ppm, giving FID resolutions of 9.39 Hz and 8.55 Hz for the F2 and F1 dimensions, respectively. The HvfA_unmod spectrum has a spectral width of 13.67 ppm and 18.00 ppm, giving FID resolutions of 8.00 Hz and 8.55 Hz for the F2 and F1 dimensions. The HNCO spectra (hncogp3d) were acquired at a size of 2,048 × 48 × 80 points using DQD, Echo-Antiecho, and States-TPPI acquisition modes for the F3, F2, and F1 dimensions, respectively, with a total of 4 scans per FID for both peptides. All HNCO spectra have a spectral width of 13.67 ppm, 18.00 ppm, and 14 ppm in the F3, F2, and F1 dimensions giving FID resolutions of 8.00 Hz, 45.59 Hz, and 52.80 Hz. The HNCACB spectra (hncacbgp3d) and HN(CO)CACB (hncocacbgp3d) were collected at a size of 2,048 × 48 × 128 points using DQD, Echo-Antiecho, and States-TPPI acquisition modes for the F3, F2, and F1 dimensions, respectively, with a total of eight scans per FID for both peptides. Both spectra have a spectral width of 13.67 ppm, 18.00 ppm, and 60 ppm in the F3, F2, and F1 dimensions giving FID resolutions of 8.00 Hz, 45.59 Hz, and 141.40 Hz, respectively. Methods describing spectral processing and assignment for the above experiments as well as the modified HNCO data acquisition, modeling, fitting, and extraction of the coupled ¹³C frequency are available in the SI Appendix.

Copper Titrations.

Titration of HvfA with copper was performed on a Cary 3500 Compact Peltier UV–visible spectrophotometer for aerobic experiments with Cu(II) and an Agilent 8453 within a Coy vinyl anaerobic chamber for anaerobic experiments using Cu(I). For Cu(II) experiments, a solution of 5 mM CuCl₂ was prepared in water and titrated into a solution of 50 µM HvfA_mod in 0.5 to 1 µL injections and thoroughly mixed, and the sample was measured on the UV–visible spectrophotometer after each addition. For Cu(I) experiments, a solution of 2 to 5 mM tetrakis(acetonitrile)copper(I) hexafluorophosphate ([Cu(CH₃CN)₄]PF₆) was dissolved in degassed acetonitrile in an anaerobic chamber. This solution was prepared fresh before each experiment. HvfA_mod was degassed, brought into the anaerobic chamber, and diluted into anaerobic PBS pH 7.8 to a final concentration of 50 µM. The titration was carried out in the same manner as the Cu(II) titrations except in the anaerobic chamber.

Preparation of the HvfA–Cu(I) Complex.

Cu(I)-bound modified and unmodified HvfA samples were prepared by adding 3 mol. equiv. of [Cu(CH₃CN)₄]PF₆ to anaerobic HvfA_mod (prepared as described above) in an anaerobic chamber. The solution was then desalted using a PD-10 column pre-equilibrated with the desired anaerobic buffer. Samples were then used for ICP–MS or native MS [in which case samples were desalted into 200 mM ammonium acetate (NH₄Ac) pH 7.2].

Native Mass Spectrometry.

Samples of Cu(I)-bound HvfA_mod and HvfA_unmod in 200 mM NH₄Ac were directly infused into an Orbitrap Exploris 480 mass spectrometer using a Sciex CESI 8000 system with 1 psi pressure for a flow rate of 55 nL/min. The glass emitter was filled with 10 mM NH₄Ac for generating native electrospray at 2.2 kV. For native electrospray MS, the in-source CID was 0 eV; intact protein mode was applied in standard-pressure mode (HCD gas pressure: 1.0); the extended trapping level was set to 1.0. Other source conditions were the same as for denatured MS. Full mass spectra were collected at a resolution of 120,000 (at m/z of 200), m/z range of 400 to 2,000, normalized AGC target of 200%, and a microscan setting of 1. The mass spectrum was deconvoluted in the range of m/z 700 to 2,000 in Thermo Xcalibur Qual Browser using the Xtract algorithm.

BLAST Search for Other Oxazolin BGCs.

Searching the Joint Genome Institute Integrated Microbial Genomes and Microbes database (46) for oxazolin homologs is described in SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.