Characterization of phthiocerol and phthiodiolone dimycocerosate esters of M. tuberculosis by multiple-stage linear ion-trap MS[S]

Both phthiocerol/phthiodiolone dimycocerosate (PDIM) and phenolic glycolipids are abundant virulent lipids in the cell wall of various pathogenic mycobacteria, which can synthesize a wide range of complex high-molecular-mass lipids. In this article, we describe linear ion-trap MSn mass spectrometric approach for structural study of PDIMs, which were desorbed as the [M + Li]+ and [M + NH4]+ ions by ESI. We also applied charge-switch strategy to convert the mycocerosic acid substituents to their N-(4-aminomethylphenyl) pyridinium (AMPP) derivatives and analyzed them as M + ions, following alkaline hydrolysis of the PDIM to release mycocerosic acids. The structural information from MSn on the [M + Li]+ and [M + NH4]+ molecular species and on the M + ions of the mycocerosic acid-AMPP derivative affords realization of the complex structures of PDIMs in Mycobacterium tuberculosis biofilm, differentiation of phthiocerol and phthiodiolone lipid families and complete structure identification, including the phthiocerol and phthiodiolone backbones, and the mycocerosic acid substituents, including the locations of their multiple methyl side chains, can be achieved.


Alkaline hydrolysis and preparation of N -(4aminomethylphenyl) pyridiniumderivative with reagent
Alkaline hydrolysis to yield myroceranoic acids from PDIM was carried out following the protocol as previously described ( 34 ) with modifi cation. To the tube containing dried PDIM, 400 l methanol, 200 l diethylether, and 200 l tetrabutylammonium hydroxide (40 wt% solution in water) were added. The tube was capped and heated at 100°C overnight and cooled to room temperature, and 50 l HCl (37%) was added. Following addition of 1 ml water and 1 ml chloroform, vortexing for 1 min, and centrifuging at 1,200 g for 2 min, the organic layer was transferred to another tube and was washed twice with 1 ml water. The fi nal organic layer containing FFAs was dried under nitrogen, and N -(4-aminomethylphenyl) pyridinium (AMPP) derivative was made with the AMP+ Mass Spectrometry Kit, according to the manufacturer's instruction. Briefl y, the dried sample was resuspended in 20 l ice-cold acetonitrile/N,N-dimethylformamide PDIM ( Fig. 1 ) and PGL respectively consist of a longchain 3-methoxy, 4-methyl, 9,11-dihydroxy glycol (phthiocerol) and a p -glycosylated phenylglycol (glycosyl phenolphthiocerol) backbone diesterifi ed with di-, tri-, and tetra-methyl-branched long-chain mycocerosic (mycoceranic) acids ( 10,19,20 ). Dependent on the species, the long-chain diol backbone ranges in size from C 25 to C 36 , and the mycocerosic acid chain ranges from C 23 or C 24 to C 32 ( 3,4,10,21,22 ). Other variants with 3-keto or 3-hydroxy diols also exist. The phthiocerol family consists of phthiocerol A, phthiocerol B, phthiodiolone A, and phthiotriol A, of which phthiocerol A and phthiodiolone A are the most commonly present. In M. marinum , a major analogous PGL family consisting of the long-chain ␤ -diol backbone modifi ed with a phenolic group at the terminal, to which a 3-O -methylrhamnose is ␤ -O -linked to the phenol ring, is also present ( 9,23 ).
MS has played important roles in the elucidation of the structures of PDIMs and PGLs. For example, Minnikin et al. ( 10,(24)(25)(26) and Daffe et al. ( 9,(27)(28)(29) used GC/MS analysis together with NMR spectroscopy for complete characterization of DIMs, following extraction steps, chromato graphic separations, and chemical reactions. Recently, MALDI-TOF ( 16,29,30 ) and ESI-Fourier transform ion cyclotron resonance ( 31 ) MS have also been applied for profi ling PDIM and PGLs. However, the ESI/ MS/MS method useful for direct structural identifi cation of these complex lipids has not been established. Here, we describe MS n linear ion trap (LIT) with high-resolution MS toward characterization of the structures of PDIMs, which were desorbed as the [M + Alk] + (Alk = Li, Na, NH 4 ) ions by ESI. This mass spectrometric approach affords realization of the structures of this lipid family isolated from the biofi lm formed by M. tuberculosis , including the identities of the mycocerosic acid side chains and the phthiocerol backbone.

Bacterial strains and growth conditions
M. tuberculosis Erdman was cultured at 37°C in Middlebrook 7H9 or Middlebrook 7H10 agar plates supplemented with 60 l/l oleic acid, 5 g/l BSA, 2 g/l dextrose, 0.003 g/l catalase, 0.5% glycerol, and 0.05% Tween 80 (broth) or in Sauton's liquid medium unless otherwise indicated. Bacterial biofi lms were inoculated with stationary phase planktonic cultures into Sauton's media at a 1:100 dilution. Culture vessels were closed tightly to restrict oxygen for 3 weeks and then vented, as described by Ojha ( 33 ).

Bacterial lipid extraction
Bacterial biofi lms were harvested at the indicated time, pelleted, and boiled for 30 min. Samples were then extracted twice Ϫ (X = Cl, HCO 2 ) were also observed in the negative-ion mode; however, CID MS n on these latter ions failed to provide suffi cient information for structural identifi cation and will not be discussed further.
To gain insight into the fragmentation processes, we fi rst performed MS n on the [M + Li] + and [M + Na] + adduct ions of C 32:0 /C 32:0 -mC 35:0 PDIM standard to explore their utilities toward structural determination of PDIMs. As shown in Fig In contrast, the MS 2 spectrum of the [M+ NH 4 ] + ion at m/z 1,497.5 ( Fig. 2D ) is dominated by the ion at m/z 1,000 arising from losses of the 3,5,7,9-tetramethyloctaeicosanoic acid and NH 3 , similar to that observed for the [M+ NH 4 ] + ion of triacylglycerol (TAG) ( 36 ). The spectrum also contained the ion at m/z 968, arising from further loss of CH 3 OH, along with the ions at m/z 519 from additional loss of the 32:0-FA substituent, and at m/z 487 (519 -CH 3 OH) arising from further loss of the methoxy group. The losses of the FA substituents and NH 3 , and methanol were supported by high-resolution MS (data not shown). The observation of the ions at m/z 968 (1,000 -CH 3 OH) and at 487 (519 -CH 3 OH) corresponding to loss of CH 3 OH residue, supports the notion that the phthiocerol possesses a methoxy side chain, and the molecule belongs to the 3-methoxy 4-methyl PDIM family.
It appeared that the 32:0-FA substituent at C11 was fi rst cleaved from the phthiocerol backbone via a charge-remote fragmentation process to yield a lithiated ion of m/z 1,006.1 ( Fig. 2A ), which possesses a double bond at C-11 ( Scheme 1 ). This is followed by elimination of the remaining 32:0-FA at C9 to form the lithiated ion of m/z 525 that consists of a conjugated double bond at C9 and C11. This assumption is based on the fi ndings that the MS 4 spectrum of the ion of m/z 525 (1,486 → 1,006 → 525; Fig. 2E ) contained the major ion at m/z 493, arising from loss of the methoxy group (loss as CH 3 OH) attached to the phthiocerol chain, along with ions at m/z 231 arising from allylic cleavage of the C13-C14 bond and at m/z 161 arising from cleavages that may eliminate an alkene, acetylene, and H 2 ( Scheme 1 ).
Similarly, the 32:0-FA residue at C11 was also preferentially cleaved ( Fig. 2D ), resulting in the elimination of NH 3 and 32:0-FA residues to form the ion of m/z 1,000 in (4:1, v/v), and 20 l of ice-cold 1 M 3-(dimethylaminopropyl) ethyl carbodiimide hydrochloride in water was added. The vial was briefl y mixed on a vortex mixer and placed on ice. To the sample tube, 10 l of 5 mM N -hydroxybenzotriazole and 30 l of 15 mM AMPP (in distilled acetonitrile) were added. After the solution was heated at 65°C for 30 min and cooled to room temperature, 1 ml of water and 1 ml of n-butanol were added. The fi nal solution containing FA-AMPP derivative was vortexed for 1 min and centrifuged for 3 min at 1,200 g . The organic layer was transferred to another vial, dried under a stream of nitrogen, and stored at Ϫ 20°C until use.

MS
Both high-resolution ( R = 100,000 at m/z 400) higher energy collision activation dissociation and low-energy collision-induced dissociation (CID) MS/MS experiments were conducted on a Thermo Scientifi c LTQ Orbitrap Velos mass spectrometer with Xcalibur operating system. Lipid extracts in chloroform-methanol (2:1) were infused (1.5 l/min) to the ESI source, where the skimmer was set at ground potential, the electrospray needle was set at 4.0 kV, and temperature of the heated capillary was 300°C. The automatic gain control of the ion trap was set to 5 × 10 4 , with a maximum injection time of 50 ms. Helium was used as the buffer and collision gas at a pressure of 1 × 10 Ϫ 3 mbar (0.75 mTorr).
The MS n experiments were carried out with an optimized relative collision energy ranging from 30% to 45%, an activation q value at 0.25, and the activation time at 10 ms to leave a minimal residual abundance of precursor ion ( ‫ف‬ 20%). The mass selection window for the precursor ions was set at 1 Da wide to admit the monoisotopic ion to the ion trap for CID for unit resolution detection in the ion trap or high-resolution accurate mass detection in the Orbitrap mass analyzer. Mass spectra were accumulated in the profi le mode, typically for 2-10 min for MS n spectra (n = 2,3,4).

Nomenclature
For simplicity, the terms phthiocerol dimycocerosate and phthiodiolone dimycocerosate, abbreviated as PDIM, will be used for all the families of the parent waxes without implying one particular stereochemistry for the component of the multimethylbranched acids. The term phthioglycol, a modifi cation of Stendal's original nomenclature ( 35 ), will be used to refer to the family of compounds, and the term phthiocerol will be reserved for the original 3-methoxy congener. Thus, the abbreviation of the phthiocerol backbone of a PDIM, for example, the 3-methoxy, 4-methyl, 9,11-dihydroxy hexaeicosane, is designated as mC 27:0 to refl ect the C 27 chain length with a methoxy group attached at C-3; glycol backbones with 3-keto, 4-methyl 9,11-dihydroxy hexaeicosane and with 3-hydroxy, 4-methyl 9,11-dihydroxy hexaeicosane are designated as kC 27:0 and as hC 27:0 , respectively. The two mycocerosate residues consisting of two tetramethylbranched C 30 -acyl chains located at 9 and 11, for example, are abbreviated as C 30 /C 30. Therefore, a phthioglycol with a respective mC 27:0 and kC 27:0 backbone and two C 30 -acyl chains is designated as C 30 /C 30 -mC 27:0 PDIM and C 30 /C 30 -kC 27:0 -PDIM, respectively.  , and so forth (intensity in the descending order), arising from cleavages of the C-C bonds of phthiocerol chain. The observation of these two ion series is consistent with the notion that the charge may primarily reside at C7, and cleavages of these C-C bonds may be similar to the "mobile proton" model ( 37,38 ), in which more than one charge sites are energetically and/or kinetically favored due to proton rearrangement which the proton may reattach to the carboxylate group of the remaining 32:0-FA ( Scheme 2 ). Further dissociation of the ion of m/z 1,000 (1,497 → 1,000; Fig. 2F ) eliminates the 32:0-FA substituent at C9 to give rise to the major ion of m/z 519, which may represent a protonated phthiocerol ion possessing a conjugated double bond with proton relocated at the methoxy side chain. This is followed by loss of a methanol molecule to yield the ion of m/z 487, in which the charge site may situate at C7 via hydrogen shift. These fragmentation processes leading to the ions of m/z 487 were supported by the MS 4 spectrum of the ion of m/z 519 (1,497 → 1,000 → 519; data not shown). The loss of the 32:0-FA substituent is also consistent with the observation at C-8 may also lead to the ion series in which the ion of m/z 375 formed by allylic cleavage becomes prominent. The cleavages from the various precursors with various locations of charge site due to proton rearrangement led to the formation of the whole array of ion series with CH 2 interval (14 Da). By contrast, the MS 5 spectrum of m/z 485 arising from a keto PDIM specimen contained the ion series that is 2 Da lighter and is characteristic to the family (discussed subsequently).  Fig. 3A ), 1,397.5 ( Fig. 3B ), respectively, are similar to those observed for methoxy-PDIM family as shown in Fig. 2 . The former spectrum ( Fig. 3A ) is dominated by the along the phthiocerol backbone. For example, C-C bond cleavages distal to the methyl side chain terminal from precursor ions possessing charge located at C7 gave rise to ions at m/z 193, 207, 221, 235, and so forth; while ions at m/z 445, 431, 417, 403, 389, and so forth arose from C-C cleavages from the methyl side chain terminal. Further dissociations of the ions of m/z 431, 417, and 403 also gave rise to ions of m/z 389, 375, and 361 by loss of propylene residue ( 39 ); while ions of m/z 151 (loss of C19-alkene), 137 (loss of C20-alkene), and 123 (loss of 21:0-alkene) can also arise from, for example, m/z 417 by consecutive fragmentation processes that eliminate an alkene residue ( Scheme 2 ). These fragmentation processes are supported by MS 6   (1,397 → 942; Fig. 3D ), and the MS 4 spectrum of the ion of m/z 503 (not shown).

CID MS
Further dissociation of the ion of m/z 485 (1,397 → 941 → 503 → 485; Fig. 3F ) gave rise to the ion series of m/z 457, 443, 429, 415, and 401, … , and so forth, in which the ion of m/z 457 is likely arising from loss of an ethane residue, in contrast to loss of a C 3 -alkane observed for the ion of m/z 487 originated from a methoxy PDIM compound as shown in Fig. 2G . The spectrum also contained the ion series of m/z 403, 389, 375, 361, and so forth, consistent with the notion that elimination of a water molecule involves the participation of the hydrogen at C-4.

Characterization of M. tuberculosis biofi lm PDIMs as [M + NH 4 ] + ions
A previous report indicates that M. tuberculosis produced mainly diesters of phthiocerol A and phthiodiolone A, and phthiocerol B diesters were present in very small amounts so that their detection was unreliable ( 10 ). High-resolution mass measurements (supplementary Fig. 3) on the NH 4 + adduct ion of these mycobacterial waxes clearly resolved phthiodiolone A dimycolates, which gave elemental compositions of C 86 H 168 (CH 2 ) n O 5 NH 4 (n = 1, 2, … , 12), from phthiocerol A dimycolates having elemental compositions of C 86 H 170 (CH 2 ) n O 5 NH 4 ( Table 1 ). More than one isomer was found in most of the species, and molecules in the phthiocerol B family were not detected. Examples for complete characterization of the entire PDIM family from biofi lm of M. tuberculosis ( Table 1 ) applying LIT MS n are given below.  Table 1 ; the structural assignments of the 29:0/32:0-k33:0, 29:0/30:0-k35:0, and 27:0/32:0-k35:0 isomers can be found in supplementary Fig. 2).
The MS 2 spectrum of the [M+ NH 4 ] + ions at m/z 1,397 ( Fig. 3B ), again, contained the prominent ion of m/z 941.9, arising from expulsion of 29:0-FA and NH 3 , and further loss of the remaining 29:0-FA gave rise to the ion of m/z 503 ( Fig. 3D ), in which the charge site is relocated at the carbonyl group ( Scheme 4 ). The ion at m/z 485 arose from additional loss of a water molecule ( Fig. 3D ), probably involving a prior 1,6-H shift to yield a carbonium ion with 3-OH side chain, whose charge may reside at C-7. This is followed by loss of a water molecule to form a stable allylic carbonium ion of m/z 485 with the participation of the adjacent hydrogen at C-4 ( Scheme 4 ). These fragmentation processes are supported by the MS 3 spectrum of the ion of m/z 941.9   Similar approaches were also applied to reveal the structural complexity of phthiodiolone dimycerosate family of the extract ( Table 1 ) Table 1 ).

Characterization of multiple-methyl-branched long-chain mycocerosic (mycoceranic) acid substituents
To assign the structure of mycoceranic acid substituents, released free acids by alkaline hdrolysis were converted to the AMPP derivatives and subjected to ESI MS n analysis in the positive-ion mode. Three major species were observed at m/z 647, 619, and 605, corresponding to C32-, C30-, and C29-FA AMPP derivatives, respectively. The MS 2 spectra of m/z 647 ( Fig. 6A ), 619 ( Fig. 6B ), and 605 ( Fig. 6C ) all contained abundant ions at m/z 183 and 169, which are characteristic of FA-AMPP derivatives (40)(41)(42)(43)(44). The MS 2 spectrum of the ion of m/z 647 ( Fig. 6A ) also contained the ion series at m/z 365, 323, 281, and 239, along with the ion series of m/z 351, 295, and 253, pointing to the position of the methyl side chains at C-2, 4, 6, and 8 of 32:0-mycoceranic acid (see insets for fragmentation pathways), and gave assignment of 2,4,6,8-tetramethyloctaeicosanoic acid structure. The MS 2 spectrum of the ion of m/z 619 ( Fig. 6B ) contained the similar ion series, indicating the presence of 2, 4, 6, 8-tetramethyl-hexaeicosanoic acid (30:0-mycoceranic acid). In contrast, the MS 2 spectrum of the ion of m/z 605 ( Fig. 6C ) contained the ion series of at C11 than that at C9 upon CID as observed in this study may be applicable for specifi cally defi ning the mycoceranic acid substituents on the phthiocerol backbone (at C9 or C11). However, more studies with standard compounds with two different mycoceranic acid chains are required to confi rm this fi nding. In the negative-ion mode, in contrast, the MS n spectra obtained from the [M + X] Ϫ adduct ions do not provide suffi cient informative ions applicable for structure identifi cation, despite that ions in the fashions of [M + X] Ϫ (X = Cl, HCO 2 ) are readily formed, and the elemental compositions derived from high-resolution MS are also distinguishable among the methoxy-PDIM and keto-PDIM families (data not shown).
Fragment ions arising from cleavages of C-C bonds along the phthiocerol backbone following loss of the mycoceranic acid substituents may involve the precursors in various resonance forms differed by the charge sites due to proton delocalization ( 37,38 ). Thus, charge-remote fragmentation processes may not be invoked ( 44,45 ), and other fragmentation processes become available ( 46 3F ) and of m/z 487 ( Fig. 2G ) lies on the notion that the former spectrum contains abundant ions at m/z 191, 177, 163, 149, 135, and so forth; while the analogous ions seen at m/z 193, 179, 165, 151, 137, and so forth in the latter spectrum are less prominent ( Fig. 3F ). The prominence of the ion series of m/z 191, 177, 163, 149, and so forth, arising from m/z 485 ( Fig. 3F ) may be attributable to the fact these ions consist of one more double bond than the analogous ions of m/z 193, 179, 165, and so forth, arising from m/z 487 and are more conjugated and more stable. Striking differences were also observed for the MS 5 spectra of the ions of m/z 459 arising from m34:0 ( Fig. 4F ) and of 457 arising from k33:0 backbones ( Fig. 5F ). These differences in the profi les of the MS n spectra between those arising from phthiocerol and from phthiodiolone families also provide useful information for their structural differentiation by MS.
The advantage of characterization of mycoceranic acid substituents using charge-switch formation of the FA-AMPP derivatives is that feature ions of m/z 183 and 169 are readily recognizable ( Fig. 6 ) ( 34,(40)(41)(42)(43), and ions from charge-remote fragmentations can locate the methyl side chains ( 34,43 ) and the functional groups unambiguously ( Fig. 6 , inset). By contrast, the traditional GC/MS method for identifi cation of mycocerosic acids is laborious, requiring acid methanolysis to form the methyl ester, separated by TLC, followed by reductive degradation to mycocerosic alcohol and another two-dimensional TLC purifi cation, before formation of the fi nal t-butyldimethylsilyl ether derivatives (23)(24)(25). The superb sensitivity gained for the acid detected as the AMPP derivative (as compared with its underivatized form) also facilitates structural identifi cation and quantitation ( 34,(40)(41)(42)(43). This aspect may deserve further investigation.
The insight into the detailed structures of the PDIMs ( Table 1 )  The results are consistent with the notion that methoxy-PDIM is formed from keto-PDIM ( 14,30,47 ). It is also notable that keto-PDIMs are more prominent than the methoxy-PDIM species, while phthiodiolone dimycocerosates were the minor components previously reported in the M. tuberculosis cells ( 30 ).