Structure Determination of Lipid A with Multiple Glycosylation Sites by Tandem MS of Lithium-Adducted Negative Ions

Abstract

FLATⁿ is a tandem mass spectrometric technique that can be used to rapidly generate spectral information applicable for structural elucidation of lipids like lipid A from Gram-negative bacterial species from a single bacterial colony. In this study, we extend the scope and capability of FLATⁿ by tandem MS fragmentation of lithium-adducted molecular lipid A anions and fragments (FLAT^n-Li) that provides additional structural and diagnostic data from FLATⁿ samples allowing for the discrimination of terminal phosphate modifications in a variety of pathogenic and environmental species. Using FLAT^n-Li, we elucidated the lipid A structure from several bacterial species, including novel structures from arctic bacterioplankton of the Duganella and Massilia genera that favor 4-amino-4-deoxy-l-arabinopyranose (Ara4N) modification at the 1-phosphate position and that demonstrate double glycosylation with Ara4N at the 1 and 4′ phosphate positions simultaneously. The structures characterized in this work demonstrate that some environmental psychrophilic species make extensive use of this structural lipid A modification previously characterized as a pathogenic adaptation and the structural basis of resistance to cationic antimicrobial peptides. This observation extends the role of phosphate modification(s) in environmental species adaptation and suggests that Ara4N modification can functionally replace the positive charge of the phosphoethanolamine modification that is more typically found attached to the 1-phosphate position of modified lipid A.

Graphical Abstract

INTRODUCTION

Lipid A is an essential structural component of the bacterial cell, stringently required for survival of almost all known Gram-negative bacteria.¹ This complex glycolipid is the membrane portion of the medically relevant biomolecule lipopolysaccharide (LPS), also known as endotoxin, and is the pathogen-associated molecule primarily responsible for the toxicity of Gram-negative bacteria during infection and sepsis. The toxicity of lipid A is dependent on its structure, due to specific interactions with the MD-2/TLR-4 receptor complex. Therefore, it is important to determine the structure of lipid A to better understand the role of this molecule in disease and further elucidate the structure–activity relationship (SAR) of this toxin. Indeed, we and others have demonstrated that the chemical structure of lipid A varies among species,^2–4 and these differences in structure are relevant to function.^5–9 Moreover, the structural modification of lipid A is also related to antibiotic resistance mechanisms of Gram-negative bacteria.^10–12 For instance, resistance to cationic antimicrobial peptides (including colistin, or polymyxin E)¹³ results from modification of lipid A phosphate moieties by positively charged groups such as phosphoethanolamine (PEtN) and 4-amino-4-deoxy-l-arabinopyranose (hereafter simply aminoarabinose or Ara4N), which increase the positive charge character of the outer membrane components and disrupt the function of cationic antimicrobial peptides.¹ Therefore, accurate determination of lipid A structure can be a key metric to predict and understand bacterial antibiotic resistance, as well as the toxicity of the membrane.

Lithium (Li)-adducted lipid cations facilitate identification of lipids by providing additional fragment information during tandem mass spectrometry (MS²).^14–16 This strategy has been employed for structural analysis of phospholipids,¹⁴ triacylglycerols,¹⁵ and steroids.¹⁶ However, the formation of Li adducts with lipid A and its fragmentation and characterization in tandem MS have not been studied. Electrospray ionization (ESI)-based tandem mass spectrometry and/or low-resolution MALDI-TOF MS have been most commonly used for analysis of the lipid A structure.^17,18 However, time-consuming extraction and purification methods are essential prior to both of these lipid A structural analyses, due to contaminating biomaterials and salts. To resolve these issues, our group previously developed and demonstrated FLAT¹⁹ (Fast Lipid Analysis Technique) and FLATⁿ (FLAT with MSⁿ) for structural analysis of lipid A from primary biomass.²⁰ These methods use on-surface LPS derivatization without isolation or purification from the biological substrate, followed by matrix-assisted laser desorption ionization (MALDI) MSⁿ allowing rapid structural analysis in 2 h.²⁰ To expand the utility of FLATⁿ, we explored a negative-mode lithium-adducted form of the lipid A anion ([lipid A + Li – 2H]⁻) and conducted fragmentation using MALDI trapped ion-mobility time-of-flight (timsTOF) tandem MS. We call this new extended methodology FLAT^n-Li (FLAT with MSⁿ and lithium adducts), which allows for detection and structural characterization of labile phosphate modifications on lipid A, such as glycosylation with Ara4N, a modification which is frequently lost during solvent purification methods. We further demonstrate here that this technique resolves the exact chemical location of the Ara4N on either the 1 or 4′-phosphate group in these Gram-negative bacteria; this is a novel method that deciphers site-specific modifications of lipid A without chromatographic or chemical degradation techniques and with greater fidelity to the in vivo structure.

We and others have previously demonstrated several structural modifications of the hydrophobic domain of lipid A in psychrophiles²¹ (cold-adapted organisms with optimal growth temperatures below 12 °C and capability often down to 0 °C or lower) which correlate with homeoviscous adaptation to cold growth,²² including the use of shorter and more unsaturated acyl residues^23,24 and widened substrate specificity of acyltransferases incorporating a range of acyl lengths in both the primary and secondary linkages.^25–27 We now show that Gram-negative psychrophiles also modify both the 1 and/or 4′ terminal phosphates of the lipid A backbone with Ara4N by characterizing the complete lipid A structure of the planktonic psychrophiles Duganella aceris (D. aceris), Massilia rubra (M. rubra), and Pseudomonas lactucae (P. lactucae) in comparison to the known Ara4N-modified structure of Proteus mirabilis (P. mirabilis).

Our work presented here also suggests an additional and perhaps more evolutionarily fundamental role of phosphate modification in the biological function of lipid A in free-living species. Lipid A is an essential structural component of the bacterial cell, stringently required for survival of almost all known Gram-negative bacteria.¹ As the main component of the outer leaflet of the outer membrane, lipid A is the scaffold for interaction with and adaptation to not just the host immune system but the broader environment throughout diverse environmental niches. Several organisms whose lipid A is characterized in this work are environmental psychrophiles, bacteria that have undergone evolutionary adaptation to diverse cold environments. They have an optimal growth temperature below 15 °C, and many can grow even at subzero temperatures.^28–30

EXPERIMENTAL SECTION

Materials.

Lithium chloride^16,31 (LiCl, ≥99.98%), methanol (MeOH, 98%), norharmane (NRM, 98%),^32,33 and chloroform (HPLC grade) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Certified endotoxin-free water was sourced from Gibco (Grand Island, NY, USA). ITO glass slides were obtained from Delta Technologies LTD (Loveland, CO, USA). All other chemicals were obtained from Sigma-Aldrich unless otherwise noted. Proteus mirabilis ATCC 7002 (P. mirabilis) was obtained from American Type Culture Collection (ATCC). Marine 2216 premixed media, Reasoners (R2) premixed media, and necessary chemical stocks were obtained from Fisher Scientific. For plate growth, 1.5% Difco Bacto-Agar was added to all media.

Bacteria Strains, Isolation, and Cultivation.

P. mirabilis isolates were cultivated at 37 °C in Lysogeny broth (LB, Becton-Dickenson, Hunt Valley, MD, USA) supplemented with 1.5% Bacto-agar. Duganella aceris (D. aceris) and Massilia rubra (M. rubra) were isolated from the Kuparuk River (North Slope Borough, AK) by culture on Reasoners agar at 4 °C. D. aceris was isolated from unfiltered surface water sterile-collected at 70.3300 N, 149.0073 W on June 11th, 2019 and M. rubra was isolated from unfiltered surface water sterile-collected at 68.6461 N, 149.4109 W on June 3rd, 2019. Pseudomonas lactucae (P. lactucae) was isolated from the Chesapeake Bay (Annapolis, MD) by culture on Marine agar at 4 °C from surface water sterile-collected at 38.9180 N, 76.4104 W on February 20th, 2020. The latter three cultures were colony-purified at least three times prior to Sanger 16S colony-based PCR sequencing (Azenta, NJ). Sequencing data were validated and assembled using Geneious Prime software (Biomatters, NZ), and BLAST to the GenBank nt database (ncbi.nlm.nih.gov) was used to identify the strains. For lipid A characterization by FLATⁿ, FLAT^n-Li, and/or GC-MS FAME, these environmental strains were grown on solid media as described above or in 500 mL of liquid media at 8 °C with 215 rpm rotary shaking until confluent.

Sample Preparation.

Colonies of each bacterial strain were obtained using an inoculation loop then resuspended in deionized water (200 μL). The bacterial pellet was harvested by centrifugation (8000g, 5 min) and then resuspended in 10 μL, which was used for MS analysis. For FLATⁿ or FLAT^n-Li, 1 μL of the bacterial resuspension was analyzed as previously described.^19,34 Briefly, 1 μL of the prepared citrate buffer solution (0.2 M citric acid, 0.1 M trisodium citrate, pH 3.5) was deposited onto the sample spot on an ITO slide. The plate was incubated in a humidified, closed glass chamber for 30 min at 110 °C. After heating, the ITO slide was removed from the chamber, cooled, thoroughly washed with water using a pipet, and left to dry on the laboratory bench.

Lipid A fatty acid content was measured via gas chromatography coupled to flame ionization detection (GC-FID) after acid hydrolysis, methylation, and hexane extraction. For sample preparation, cleavage of the ester and amide bonds linking lipid A acyl chains to the glucosamine backbone was achieved by incubating hot phenol-water extracted LPS in the presence of 2 M methanolic HCl at 90 °C for 18 h, subsequently converting fatty acids into fatty acid methyl esters. These fatty acid methyl esters were extracted using hexane and analyzed via GC-FID. C₁₅ was used as an internal standard for quantitation. For sample analysis, GC-FID separation and detection was carried out on a Shimadzu GC-MS QP2010 Ultra twin-line machine (one to MS and one to FID), where both lines were equipped with a DB-5 ms (30 m × 0.25 mm i.d. × 0.25 μm film thickness, 5% phenyl-95% dimethylpolysiloxane) analytical column from Agilent. The GC instrument was equipped with a split/splitless inlet. Inert split/splitless liners were used. For analysis, 1 μL of sample was injected into the FID line inlet in split mode at an injection temperature of 250 °C. The following GC temperature program was used: 80 °C (hold 1 min), 25 °C/min ramp to 160 °C (hold 1 min), 10 °C/min ramp to 265 °C, 1 °C/min ramp to 270 °C (hold 1 min), and 10 °C/min ramp to 295 °C (hold 1 min). Ultrahigh purity helium (UHP300) was used as a carrier gas maintained at a constant pressure of 110.7 kPa translating to a linear velocity of 31.3 cm/s. The FID was maintained at a temperature of 325 °C with a makeup flow of 30 mL/min helium (UHP300), 40 mL/min compressed hydrogen, and 400 mL/min compressed air. For data evaluation, LabSolutions LC/GC version 5.99 (Shimadzu) was utilized.

Mass Spectrometry.

A Bruker Matrix-Assisted Laser Desorption/Ionization trapped ion mobility spectrometry Time-of-Flight Mass Spectrometry MALDI (tims TOF) MS was used for FLAT (direct biomass MS for visualization of lipid A primary molecular ions), FLATⁿ (direct biomass MS/MS for fragmentation and structural analysis of lipid A), and FLAT^n-Li (direct MS/MS of lithium-doped biomass for greater fragmentation and resolution of lipid A, particularly of glycosylated structures). This MALDI (tims TOF) MS is equipped with a dual ESI/MALDI source with a SmartBeam 3D 10 kHz frequency tripled Nd:YAG laser (355 nm); the system was operated in “qTOF” mode (TIMS deactivated) for these experiments. Ion transfer tuning was used with the following parameters: funnel 1 RF, 440.0 Vpp; funnel 2 RF, 490.0 Vpp; multipole RF, 490.0 Vpp; CID energy, 0.0 eV; and deflection delta, −60.0 V. Quadrupole was used with the following values for MS mode: ion energy, 4.0 eV. The m/z scan range for MS is set to 1000 to 3000. Collision cell activation of ions used the following values for MS mode: collision energy, 9.0 eV; collision RF, 3900.0 Vpp. In the MS/MS mode, the precursor ion was chosen by typing the targeted m/z value including two digits to the right of the decimal point. Typical isolation width and collision energy were set to 4–6 m/z and 100–110 eV, respectively. The m/z scan range for MS/MS is set to 200 to 3000. Focus Pre TOF used the following values for transfer time, 110.0 μs, and prepulse storage, 9.0 μs. Agilent ESI tune mix was used to perform calibration of the m/z scale. MALDI parameters in qTOF were optimized to maximize intensity by tuning ion optics, laser intensity, and laser focus. All spectra were collected at a 104 μm laser diameter with beam scan on using 800 laser shots per spot and 70 and 80% laser power, respectively. Both MS and MS/MS data were collected in negative ion mode. In all cases, 10 mg/mL of norharmane (NRM)^32,35 in 1:2 MeOH/CHCl₃ (v:v) was used for lipid A detection. For generation of the lithium adducted form of the lipid anion, 1 μL of LiCl (100 mM, deionized water) was added to 99 μL of NRM solution. NRM solution (1 μL) was deposited on the sample spot. All MALDI (timsTOF) MS and MS/MS data were visualized using mMass (Ver 5.5.0).³⁶ Peak picking was conducted in mMass using the following parameters: S/N threshold, 3.0; absolute intensity threshold, 1.0; relative intensity threshold, 1.0%; picking height, 50; apply baseline; and apply smoothing. Identification of all fragment ions was determined based on exact mass as predicted by Chemdraw Ultra (Ver10.0).

RESULT AND DISCUSSION

Determination of Structure of Ara4N Modified Lipid A from Proteus mirabilis.

Proteus mirabilis (P. mirabilis) is one of the main pathogens responsible for urinary tract infections (UTIs).³⁷ Given that the structure of Ara4N modified lipid A from P. mirabilis is known,³⁸ it served as a model for and validation of FLAT^n-Li using alkali-metal adducts of the lipid A molecular anion and tandem MS fragments. Figure 1 shows FLAT^n-Li (Figure 1A) and FLATⁿ (Figure 1B) mass spectra, the composition of assigned ions (Figure 1C), and the chemical structure of lipid A (Figure 1D). We observe that the addition of lithium to the sample provides more intense fragment ions and 10 times more visible fragment features (298 compared with 29 in Figure 1A and B) than in FLATⁿ MS/MS of the same sample. For example, observe the difference in intensity and feature number between the red and black spectra in the dashed box in Figure 1A,B, highlighting the spectral region expanded in the inset. The FLAT^n-Li spectrum includes both adducted and nonadducted peaks, as well as additional fragmentation due to a neutral loss of sugars, phosphates, and/or acyl units as well as cross-ring fragments of the backbone sugars. The Li-adduct tandem MS FLAT^n-Li method is particularly effective for fragmentation of Ara4N-modified lipid A (for example, compare the intensities of 1961.31 to 1955.30 in Figure 1) and determination of the location of terminal phosphate moieties on the 1 and/or 4′ positions. In P. mirabilis, the ion at m/z 849.49 in Figure 1A shows that Ara4N is attached to the 4′-phosphate group, matching previous literature.³⁸ This ion originates from the dissociation of the ester bond cleavage of the acyl chain from the 3′-position with cross-ring fragmentation (red dashed line in Figure 1D) from the reducing end of glucosamine (3′α + ^0,4A₂), indicating that the modification is located at the 4′ position. In this way, cross-ring fragmentation in FLAT^n-Li provides critical information about the site of Ara4N modification and can assign modification sites on the phosphate groups in lipid A using a comparison of anomeric, ester, and cross-ring cleavage products consistent with known CID fragmentation chemistry and analysis.³⁹ In addition to the expected 4′ phosphate glycosylation discussed above, the increased sensitivity of FLAT^n-Li also reveals a weak Y₁ fragment signal at m/z 841.48 (Figure S1). Y₁ is the β(1→6) glycosidic bond dissociation fragment that does not contain the nonreducing sugar of the molecule; this fragment at 841.48 demonstrates low-level incorporation of Ara4N at the 1-phosphate position instead of the 4′ position, made possible by the unusual bifunctional glycosyltransferase ArnT, which can transfer Ara4N to either phosphate of the lipid A molecule,^40,41 and the absence of competing phosphate modifications. In addition to enhanced fragmentation efficiency of phosphate and glycosidic linkages, we also observed a more neutral loss of the ester-linked fatty acids in the lithium-adduct FLAT^n-Li spectra, including fatty acid loss from the sugar linkage (3α and 3′α fragments) and from the acyl-oxyacyl linkage (2′ε and 3′ε fragments). For instance, the base peak in Figure 1A at m/z 1733.10 originates from a loss of a secondary acyl-oxyacyl fatty acid (2′ε or 3′ε fragmentation). The enhanced fragmentation of FLAT^n-Li in Figure 1A provides more robust structural information for these lipids compared to that of FLATⁿ of the same lipid A in Figure 1B, the nonadducted spectrum where the intense base peak at m/z 1824.24 suppresses the signal of other ions. Ions originating from multiple neutral losses are also observed in the FLAT^n-Li spectrum of Figure 1A, such as a loss of 3α and 3′α with a loss of Ara4N or Ara4N-phosphate at m/z 1131.65 and 1045.67, respectively, and cross-ring cleavage ions with an additional loss of fatty acid or sugar moieties is also detected at m/z 724.44, 722.42, and 718.43.

Figure 1.

MS/MS spectra of lipid A from P. mirabilis. Spectra were collected and analyzed as described in the text. This figure shows (a) a FLAT^n-Li mass spectrum from the Li-adduct precursor ion at m/z 1961.31 (lipid A + Ara4N), with spectral inset enhancing the 700–850 Da range, (b) a FLATⁿ spectrum of the precursor ion at m/z 1955.30 to illustrate the contrast in features generated by the presence of lithium, (c) a table of theoretical and observed exact masses of assigned ions, and (d) the structure of P. mirabilis lipid A as determined from these data. The lithium adducted fragments are labeled with a hashtag (#) mark in the spectrum and the table of ions. In order to differentiate 1- and 4′-phosphage group modification, we labeled the m/z value with asterisk (*) and asterisk-prime (*′) marks, respectively. The m/z values in the table were obtained from abstraction of the neutral loss of the denoted letters from precursor ions.

De Novo Structure Elucidation of Lipid A from Massilia rubra.

In the section above, we characterized P. mirabilis lipid A to verify and validate the capability of FLAT^n-Li for Ara4N structural identification. In this section, we now characterize lipid A molecules with unknown structures, demonstrating the power and applicability of the technique for de novo structure prediction and systematic analysis of Ara4N-modified lipid A in psychrophilic bacteria. The novel lipid A structure presented in Figure 2 is that of M. rubra^29,42 isolated from the waters of the Kuparuk River in North Slope Borough, Alaska. The spectrum shown in Figure 2A represents the three most abundant deprotonated ion species of lipid A ([Lipid A – H]⁻, [Lipid A + Ara4N – H]⁻, and [Lipid A + 2Ara4N – H]⁻) in M. rubra. The ion at m/z 1389.79 represents a bisphosphorylated lipid A structure with free phosphate at both the 1-and 4′-phosphate groups, similar to classical Pseudomonas structures in that it shows 3-O-deacylation and 2-OH modification of the 2-N acyl-oxyacyl unit.⁴³ Addition of a single Ara4N to this structure was observed by the molecular ion at m/z 1520.85 and doubly modified with Ara4N at the 1 and 4′ positions and was detected by the ion at m/z 1651.91—a molecular species that has not previously been observed in wild-type lipid A but is again made possible in the absence of other phosphate modifications by the bifunctional nature of ArnT. FLATⁿ of the ion at m/z 1389.79 (spectrum in Figure 2B and MS/MS fragment table in Figure 2C) provides additional structural insight into the base lipid A structure; the neutral loss of phosphoric acid (−97.97 Da) at m/z 1291.82 demonstrates the presence of labile phosphate, suggested at the anomeric 1-position (B₂ fragmentation) according to the previous literature.⁴⁴ Moreover, the loss of acyl residues from this fragment provides valuable structural information, such as B₂ + 3′α at m/z 1103.68, suggesting the presence of hydroxydecanoic acid (chemical formula, C₁₀H₂₀O₃; molecular mass, 188.14) or B₂ + 3′α + 2ε at m/z 887.50, which indicates the presence of hydroxylauric acid (chemical formula, C₁₂H₂₄O₃; molecular mass, 216.17). Cross-ring fragmentation (A₂) with a loss of acyl chain shows that the N-2′ acylamide is an acyl-oxyacyl (two-acyl residue) unit. For instance, ^0,2A₂ + 3′α at m/z 694.36 shows the N-2′ acylamide has one hydroxydecanoic acid and lauric acid. Furthermore, the intensity ratio between two cross-ring fragments with a loss of acyl chain, ^0,2A₂ + 3′α and ^0,4A₂ + 3′α at m/z 634.34, indicates that this lipid A is penta-acylated, a pattern previously observed in the spectroscopy of penta-acylated lipid A in Pseudomonas aeruginosa.⁵ Last, the glycosidic bond dissociation, the Y₁ ion at m/z 626.33, elucidates that the lipid A has two fatty acid chains on the reducing-end sugar of the glucosamine backbone, again as in Pseudomonas. To confirm the acyl composition of lipid A, we conducted fatty acid methyl ester (FAME) analysis via gas chromatography coupled to flame ionization detection (GC-FID) and determined acyl chain composition (Figure S2). The results match with the assigned structure by FLATⁿ; the 3-OH fatty acids in direct linkage to the sugars are 3-OH decanoyl residues, and the acyl-oxyacyl residues are dodecanoate and 2-OH dodecanoate in equal abundance (Figure 2D).

Figure 2.

FLAT and FLATⁿ spectra of lipid A from M. rubra. Spectra were collected and analyzed as described in the text. This figure shows (a) a FLAT mass spectrum of M. rubra lipid A showing molecular structures with no Ara4N (m/z 1389.79), one Ara4N (m/z 1520.85), and two Ara4N (m/z 1651.91) and (b) a FLATⁿ mass spectrum of the precursor ion at m/z 1389.79 to generate the full set of assigned ions (theoretical and actual masses tabulated in c) and elucidate the base structure of M. rubra lipid A (d). The structure of this lipid A is further characterized in Figure 3.

To conclusively establish the location of the Ara4N modifications in a level of characterization shown in Figure 1 for P. mirabilis (and that previously required NMR⁴⁵ in the absence of established structural information), we performed FLAT^n-Li of the Li adduct on the Ara4N-modified forms of Massilia lipid A using both the singly modified lipid A anion at m/z 1526.86 (Figure S3) and the novel doubly modified lipid A structure at m/z 1657.92 (Figure 3A). These spectra demonstrate that Ara4N modification is present at either the 1- or 4′-phosphate groups in the singly modified form (Figures 3A and S3A). In the expanded section of the mass spectrum from m/z 750 to 770 in Figure 3A, ^0,4A₂ + 3′α and Y₁ ions were observed at m/z 765.39 and 757.39, respectively. These ions are diagnostic for the presence of the Ara4N modification on either the 4′-phosphate or 1-phosphate group, respectively. The ions are also observed in singly modified lipid A shown in Figure S3A, showing that in Massilia the Ara4N modification can occur on both 1 and 4′-phosphate group even in singly modified lipid A, however, not in equal abundance. In the case of doubly modified lipid A by Ara4N, the intensity ratio of diagnostic ions at m/z 765.39 and 757.39 is 1.91 (showing a 2:1 asymmetry in the rate of fragmentation when both are present), but singly modified lipid A has 0.44, indicating that Ara4N modification selectively occurs at the 1-phosphate position in singly modified lipid A in Massilia rubra. If the single modification of lipid A by Ara4N occurred simultaneously with a 1:1 ratio, the intensity ratio of the diagnostic ion would be the same as in doubly modified lipid A in Figure 3A; however, the ratio roughly decreases 4-fold from 1.91 to 0.44 in the case of singly modified lipid A (Figure S3A), indicating that Ara4N is more than 4 times more abundant on the 1 position phosphate than the 4′ position in the singly modified structure. This is in contrast to the established lipid A structure shown earlier in this paper (P. mirabilis in Figures 1 and S1), which heavily favors the 4′ position as is common in lipid A structural literature due to frequent competition between ArnT and EptA.⁴⁶ However, the genomes of these particular strains of Duganella (GenBank biosample SAMN32961148) and Massilia (GenBank biosample SAMN32961147) lack EptA, and the presence of doubly glycosylated lipid A lacking any phosphoethanolamine (pEtN) incorporation in Massilia suggests that the organism is functionally replacing the common pEtN modification with additional Ara4N.

Figure 3.

MS/MS spectra of lipid A from M. rubra. Spectra were collected and analyzed as described in the text. This figure shows (a) a FLAT^n-Li mass spectrum of the doubly Ara4N-modified M. rubra lipid A precursor at m/z 1657.92 and (b) a FLATⁿ mass spectrum of the lithium-free precursor of the same structure at m/z 1651.92 to generate the full set of assigned ions (theoretical and actual masses tabulated in c) and elucidate the base structure of M. rubra lipid A with two glycosyl modifications (d).

The loss of phosphate-Ara4N is the base peak in Figure 3A at m/z 1440.89. In contrast, in the FLATⁿ spectrum in Figure 3B, the lost single and double Ara4N are the most intense ions at m/z 1520.85 and 1389.79, respectively. These ions suppress other fragments, making the enhanced characterization via FLAT^n-Li necessary, as described in the Determination of Structure of Ara4N Modified Lipid A from Proteus mirabilis section above. The loss of fatty acid in the position 3′-O is commonly observed with the cleavage of Ara4N or the cross-ring, such as ions at m/z 1121.69, 765.39, 694.36, and 634.34.

To further explore the environmental modification of lipid A with 1-P-Ara4N, we also characterized the novel lipid A structure of another arctic environmental species, D. aceris, which is closely related to Massilia. The spectral data for this lipid A are shown in Supplemental Figures S4–S7 and show that lipid A of D. aceris is nearly identical in structure to that described above in M. rubra, the only constitutional difference being a lack of hydroxylation of the secondary fatty acid in the 2-position acyl-oxyacyl unit (compare Figure 2 with Figure S4 and Figure 3 with Figure S6). FAME analysis by GC-FID (Figure S7) further supports this lack of hydroxylation because of the absence of 2-OH dodecanoate from the FAME profile. Like M. rubra, this lipid A also shows single modification with Ara4N with a preference for 1-P-Ara4N modification in the single modification (Figure S5); for instance, the diagnostic ions of ^0,4A₂ + 3′α and Y₁ ions observed at m/z 765.39 and 741.39 reveal Ara4N modification at the 4′-phosphate and 1-phosphage group, respectively. The ions are also observed in both doubly and singly modified lipid A (Figures S5A and S6A). In the case of lipid A with two Ara4N residues, the intensity ratio of diagnostic ions at m/z 765.39 and 741.39 is 2.62, but singly modified lipid A has 0.99, indicating that, like in M. rubra, Ara4N modification selectively occurs at the 1-phosphate position in singly modified lipid A in Duganella (Figure S5A), as well as Massilia. We also observed a nearly identical, doubly aminoarabinosylated lipid A structure in a third closely related psychrophilic species, Janthinobacterium svalbardensis (data not shown). This organism was isolated not from the arctic but from the winter Chesapeake Bay, showing that the expanded use of Ara4N modification is not restricted to arctic psychrophiles but is present in temperate-region psychrophiles as well.

Determination of Structure of Ara4N-Modified Lipid A from Pseudomonas lactucae.

Using the variety of spectral techniques employed for Massilia and Duganella, we also observe complex Ara4N modification of lipid A from a psychrophilic strain of the P. lactucae isolated from the winter waters of the Chesapeake Bay. The FLAT spectra of P. lactucae (Figure S8) and FLATⁿ (Figure S9) demonstrated lipid A ions typical of pseudomonads in general, with a single 2-OH unit and either no glycosylation (m/z 1615.99) or a single Ara4N glycosylation (m/z 1747.05), and are in congruence with many other characterized lipid A structures of the Pseudomonas genus.^44,47,48 In previous structural work on this species, however, the exact modification site by Ara4N was not conclusively determined. FLAT^n-Li shows that in this P. lactucae strain, the Ara4N modification can be localized to either the 1 or 4′ position in singly modified lipid A (Figure 4), similar to that of the singly modified ones in D. aceris and M. rubra discussed above. Unlike M. rubra and D. aceris, however, P. lactucae has only single Ara4N modification; there is no peak corresponding to the addition of a second Ara4N residue (predicted to be a mass of 1878.12 Da) in the FLAT spectrum shown in Figure S8.

Figure 4.

MS/MS Li-adduct spectra of lipid A from P. lactucae. Spectra were collected and analyzed as described in the text. This figure shows (a) a FLAT^n-Li mass spectrum of the singly Ara4N-modified M. rubra lipid A precursor at m/z 1753.05 and (b) a FLATⁿ mass spectrum of the lithium-free precursor of the same structure at m/z 1747.05 to generate the full set of assigned ions (theoretical and actual masses tabulated in c) and elucidate the base structure of P. lactucae lipid A with a single glycosyl modification of variable location (d). Hashed bonds (---) in structure indicate the nonstoichiometric presence of Ara4N at either the 1 or 4′ position, but not at both simultaneously.

The purpose of Ara4N modification of lipid A phosphates in these environmental psychrophiles is not yet clear, as these species are known to be nonpathogenic not just to humans but also to animals and plants. It is possible that it serves as a defense against competitive antimicrobials produced by other bacteria within the planktonic microenvironment, similar to its function in pathogenesis. However, it is also possible that this modification plays a role in homeoviscous adaptation to cold conditions by disrupting intermolecular cationic salt bridges between the phosphates of physically adjacent lipid A molecules in the membrane. This could have the effect of lowering the intermolecular attractive forces in the outer leaflet, resulting in lower phase-transition temperatures and greater membrane fluidity. It has been shown that the cold growth temperature (below 26 °C) increases Ara4N modification in lipid A of mesophilic pathogenic bacteria such as Yersinia pestis,^49,50 which provides intriguing support for this biophysical hypothesis.

CONCLUSIONS

In this work, we describe a new analytical method (FLAT^n-Li) for detailed lipid A structural analysis from primary biomass via intentional cationic adduct creation in the negative mode. This method does not require any instrumental modification or chromatographic techniques and represents a major improvement in de novo structural analysis of lipid A and placement of lipid A phosphate modifications over that of classical literature techniques. Since FLAT^n-Li employs interaction between carbohydrate residues in lipid A and alkali metal ions in the gas phase, it does not require any modification of our previous methods FLAT and FLATⁿ, which are designed for on-surface derivatization of LPS from Gram-negative bacteria and avoid the labor- and material-intensive organic extraction and molecular purification processes required by conventional analytical methods.

We have validated this technique using the well-characterized lipid A structure of P. mirabilis, tested the analytical approach by characterizing the novel glycosylated structure of lipid A from a psychrophilic member of a well-characterized genus (Pseudomonas lactucae), and implemented the technique to establish the de novo structure of a novel lipid A in the psychrophilic Burkholderiaceae M. rubra and D. aceris. In these latter two structures, the FLAT^n-Li technique was able to conclusively identify unusual lipid A modifications in which Ara4N residues are covalently attached preferentially to the 1-position phosphate (in contrast with the typical preference for 4′ phosphate aminoarabinosylation seen in lipid A of many pathogenic bacteria—including P. mirabilis as characterized here). We also show here that these species can incorporate two Ara4N residues into a single lipid A structure by glycosylating both the 1 and 4′ phosphate positions of lipid A simultaneously. Double modification with phosphoethanolamine (2 × PEtN), or with one PEtN and one Ara4N, has previously been seen in other species such as Salmonella,^51,52 but dual glycosylation with aminoarabinose is a novel finding.

We are currently conducting additional studies of the lipid A structure in many other psychrophilic and mesophilic environmental bacteria to explore these modifications in more detail; these additional structural studies will expand upon our understanding of the environmental structure–function relationship of lipopolysaccharide and may shed more evolutionary light on the presence, temperature dependence, and purpose of aminoarabinosylation and other structural modifications of lipid A in free-living bacteria.

Supplementary Material

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jasms.3c00014.

Additional MS (FLAT) and MS/MS (FLATⁿ and FLAT^n-Li) spectra, including expansions of the spectra and new spectra for clarification of P. mirabilis, P. lactucae, and M. rubra lipid A structure; MS and MS/MS spectra of an additional species, Duganella aceris, which is closely related to M. rubra and shows similar structural details related to the de novo lipid A structure; and also FAME GC/MS to confirm the fatty acid identities shown in the structures of the de novo structures of M. rubra and D. aceris (PDF)