Introduction

Iron is an essential nutrient for bacteria and fungi; however, acquisition under aqueous, aerobic conditions is challenging because of the insolubility of iron (III). Microorganisms overcome this by chelating iron (III) with secreted siderophores for extracellular sequestration and uptake across cellular membranes [1]. Siderophores are low molecular weight compounds that have reported affinities for iron (III) between 1029 and 1043 M-1 [2]. Chelation occurs through oxygen coordination with characteristic subunits (e.g., trihydroxamate; Figure 1). In addition to common cellular processes that require iron as a cofactor (e.g., TCA cycle, DNA replication), siderophore-mediated iron acquisition aids in the virulence of organisms such as Staphylococcus aureus and Pseudomonas aeruginosa [3, 4]. To date, >500 siderophores have been discovered and despite their similar function, the structures vary widely, playing a major role in recognition by membrane-bound proteins and diffusion away from cellular membranes [5, 6]. Trihydroxamate siderophores are synthesized by nonribosomal peptide synthetase-independent synthetases (NISs) and commonly contain N-hydroxy-N-succinyl-cadaverine blocks [7]. Recent work, however, has revealed the high propensity for variable subunit incorporation, such as a lipid tail, desaturation event, or acetylation, modulating the bioactivity of these compounds [8, 9].

Figure 1
figure 1

Structures of the trihydroxamate siderophores (a) ferrioxamine B (Fer B), and (b) ferrioxamine E1 (Fer E1)

Previous CID fragmentation studies have provided diagnostic, nominal/accurate mass information for iron (III) bound and apo (unbound) siderophores at the MS and MS2 level and include several common ferrioxamines such as B, G, D2, and E (Figure 1a, b) [1014]. The iron (III) unbound state is frequently more abundant in bacterial extracts and has therefore been more widely studied. However, recent MS-based studies discovered a suite of low abundance (~ng/L), iron (III) bound trihydroxamates, suggesting that additional iron bound siderophores remain to be analyzed [9]. In addition to fragmentation patterns similar to the ferrioxamines mentioned above, the iron (III) bound forms are readily identifiable because of the 54Fe/56Fe isotope ratio (6.4% Fe54/100% Fe56). For these reasons, the fragmentation patterns of these features, [M – 2H + Fe]+, are of great significance for compound discovery as low abundant species represent an unexplored source of novel molecules.

In this work, recurrent fragmentation pathways have been characterized based on accurate tandem mass spectrometry (MS2 and MS3; positive mode; nano-LC-ESI-CID-LTQ-Orbitrap) analysis of a linear trihydroxamate siderophore, ferrioxamine B (Fer B), and cyclic ferrioxamine E1 (Fer E1), which were chosen for their similar scaffolds yet diverse fragmentation pathways. Diagnostic fragmentation pathways were confirmed with accurate masses observed in MS2 studies and proposed at the MS3 level. We sought to characterize these ion chemistries for the future elucidation of unique subunit structures in previously uncharacterized, low abundant, iron-coordinated siderophores [9].

Materials and Methods

Desferrioxamine B and ferrioxamine E1 were purchased from Sigma-Aldrich (St. Louis, MO, USA) and diluted in MS-grade methanol (EMD) for analysis. Iron (III) chloride (Sigma-Aldrich) was added at 1 mg/mL to 0.5 mg/mL desferrioxamine B to produce chelated ferrioxamine B. Accurate mass data of the iron (III) bound form was acquired using nano-LC-ESI-linear ion trap quadrupole-Orbitrap (Thermo Scientific, Waltham, MA, USA, LTQ Orbitrap XL). Ions were fragmented with a normalized collision energy of 35.0, FT resolution 7500, and 30.0 ms activation time. The top three most abundant species were selected for MS2 and subsequent MS3 studies. Accurate mass spectra of precursor and fragment ions were acquired by the Orbitrap and spectra were then analyzed using Xcalibur processing software.

Results and Discussion

Linear, Ferrioxamine B, and Cyclized Ferrioxamine E1: MS2

At the MS2 level, all three major product ions of Fer B were observed with bound iron ([M –2H + Fe]+) and detected with <6 ppm accuracy (SI Table 1, SI Figures 1 and 2) [11, 12]. Previously characterized diagnostic fragments were observed and include (m/z o , observed value; m/z t, theoretical value): m/z o = 597.2486 (–NH3), 496.1639 (hydroxamate cleavage between N7 and C8), and 414.1581 (amide cleavage between C11 and N12) [14]. Additional fragments were observed at lower abundance and smaller mass/charge ratios than the major cleavage events. Examples include the following (formulas match at <9 ppm): m/z t = 397.1295, m/z o = 397.1313 (C16H27N3O5Fe+), m/z t = 314.0560, m/z o = 314.0583 (C11H18N2O5Fe+), and m/z t = 270.0661, m/z o = 270.0678 (C10H18N2O3Fe+). It is interesting to note that while most fragments result from cleavage at two positions, with loss of a monohydroxyl unit being typical, one fragment, m/z = 314.0560, retains a hydroxamate –OH as a (bis) Fe (III)-dihydroxamate species (SI Figure 3). This fragment has been previously observed, predominantly at low pH (<2) and at nominal mass accuracy [14, 15]. Fragments such as these can be expected to differ within suites of siderophores or novel compound structures.

The top three major MS2 fragments of the cyclized siderophore, Fer E1 ([M – 2H + Fe]+), were detected with <2 ppm accuracy (SI Table 1, SI Figures 4 and 5). Previously characterized, diagnostic fragments were observed and include: m/z o = 636.2561 (–H2O), 619.2297 (N(1)-C(2), C(33)-N(1), (N(7)-O)-Fe, and N(7)-O cleavage), and 537.2237 (C(6)-N(7), C(11)-N(12) and (N(7)-O)-Fe cleavage) (SI Figure 5) [11, 12]. Another species of interest, at m/z t = 437.1244, resulted from cleavage at three positions: (1) between the α-methylene group (C2) and N(1), (2) the amide at C(22)-N(23), and (3) (N(29)-O)-Fe, accounting for the neutral loss of 217.1432 Da (-C9H19N3O3). This is especially informative because the linear Fer B cannot fragment in an analogous way because of the lack of a third amide and additional succinate equivalent as are found in Fer E1. Minor fragments observed include the following: m/z t = 471.1775, m/z o = 471.1773 (C18H33N5O6Fe+), and m/z t = 295.0745, m/z o = 295.0740 (C12H10N2O3Fe+) (SI Figure 4). As with Fer B, these ions map smaller regions of the Fer E1 structure and the masses of homologous fragments are expected to deviate for other compounds in this class.

Linear, Ferrioxamine B: MS3

Prior to this report, MS3 studies with accurate mass analysis have, to the best of our knowledge, not been performed on ferrioxamine B or E1. MS3 was pursued for access to additional diagnostic fragments of Fer B. The three most abundant ions in the Fer B MS2 spectrum (m/z = 597, 496, 414) were chosen for MS3 analysis (±0.5 Da mass selection window) (SI Figure 1). The detected MS3 fragments, proposed neutral losses, and ppm deviations are summarized in SI Tables 2 and 3. At the MS3 level, dominant fragments were formed as a result of water (–H2O; fragment a, Figure 2a), carbon monoxide (–CO; fragment i, Figure 2b), or carbon monoxide/water loss (–CO + –H2O; fragment o, Figure 2c) from the molecular ion. Carbon monoxide loss was common for fragments containing terminal (–CO) species from previous MS2 fragmentation. Additional neutral loss species can be found in SI Table 2. Despite the previously described structural variations observed between trihydroxamate containing siderophores, heteroatom composition at amide, oxime, and hydroxamate linkages remain consistent between the molecular suites. Because of this, the fragments resulting from cleavage of these specific bond types were of particular interest as they can direct future structural studies based on m/z deviations. Fragments arising from these bond breakage events are localized to the following regions: amide (e, t) oxime (b-h, l-n, r-t), and hydroxamate (c, g, k, n, q, s) (Figure 2, SI Tables 2 and 3).

Figure 2
figure 2

Fer B, top three MS2 fragments chosen for MS3 fragmentation and their resulting major and minor ions (theoretical m/z shown). Ions resulting from multiple fragmentation events are denoted with (*). Fragments not shown include: a, o, and p. Molecular formulas for a and o are provided in SI Table 2

Cyclized Ferrioxamine E1: MS3

The three most abundant ions in the Fer E1 MS2 spectra (m/z = 636, 619, 537) were chosen for MS3 analysis (±0.5 Da mass selection window) (Figure 3, SI Figure 4). All detected MS3 fragments, proposed neutral losses, and ppm deviations are summarized in SI Tables 2 and 3. Previous reports have indicated that the cyclized trihydroxamate, Fer E1, undergoes a unique combination of cleavage events compared with structurally similar, linear trihydroxamates such as Fer B [11, 12]. The expected ions from these cleavage events were observed at the MS2 (m/z = 636, 619, 537) and MS3 (m/z = 618, 591, 520, 437) level in this work. Similar to Fer B, high molecular weight and abundant fragments of Fer E1 were the result of water (–H2O; u) or carbon monoxide loss (–CO; fragment z, SI Table 2, Figure 3b) from the molecular ion. As was seen with Fer B, loss of water and carbon monoxide together also yielded major species (SI Table 2, fragments v and gg). However, unlike the linear molecule, Fer E1 underwent dehydration or decarbonylation concurrent with cleavage of several other bonds (e.g., C–C, C–N; SI Table 2, Figure 3b, c fragments aa and ii). Fragment ii (m/z = 420.1202) and related species hh (m/z = 437.1226) are of particular interest because they are yielded from the cleavage of amide-3 (between C(33)-N(1)), which is present only in Fer E1 as it is the site of cyclization.

Figure 3
figure 3

Fer E1, top three MS2 fragments chosen for MS3 fragmentation and their resulting major and minor ions (theoretical m/z shown). Ions resulting from multiple fragmentation events are denoted with (*). Fragments not shown include: u, v, x, y, cc, ff, gg, and kk. Molecular formulas for u, v, and gg are provided in SI Table 2

In comparison to the linear siderophore, Fer E1 underwent bond breakage between amide (w, dd, hh, ii) and C–N bonds (w, aa, ee, hh, ii) more frequently, whereas fragmentation between C–C bonds (bb) and oxime units [(N)-(O-Fe) or (N-O)-Fe, aa, bb] were less common than in Fer B. This was curious given the noncyclic structure of the MS2 fragments of Fer E1, suggesting that MS3 data can be correlated back to the parent geometry for the cyclized trihydroxamates. It is postulated that the cyclic, Fer E1, retains more oxime interactions because of the proximity of the Fe (III) to these units in the cyclic geometry [15]. Our data suggest that the result of these iron–oxygen interactions is less oxime breakage (aa, bb) and a variable fragmentation compared with the more flexible, linear Fer B.

Conclusion

We have annotated the first reported MS3 profiles for ferrioxamine B and E1 for all major fragments and neutral losses with <10 ppm mass accuracy. Owing to the growing evidence that suites of trihydroxamate siderophores are present at low concentrations (<ng/L) in the iron (III) bound ([M – 2H + Fe]+) state in microbial samples, the bound form and resulting tandem MS fragmentation patterns are of great significance for compound discovery. We demonstrate that cyclic and linear siderophore structures influence fragmentation patterns and are dependent on the chelation of the oxime units to iron (III). This insight will assist in future structure elucidation efforts with iron bound trihydroxamate siderophores.