Strains used in this study include Segniliparus rotundus strain CDC 1076^T ( = ATCC BAA-972^T = CIP 108378^T), Segniliparus rugosus sp. nov. type strain CDC 945^T ( = ATCC BAA-974^T = CIP 108380^T) [10], and Mycobacterium tuberculosis strain H37Rv ( = ATCC 27294^T).

Bacteria were grown in sterile filtered (0.2 µm) Middlebrook 7H9 medium containing 0.2% glycerol (w/v) and 10% (v/v) Middlebrook ADC enrichment. Cultures were incubated at 37°C for 4 days after which cells were collected by centrifugation at 3,500 rpm, 20°C for 15 min. Cells were washed three times, resuspended in 10 ml fresh growth medium, and cellular debris was allowed to settle. Five ml of suspended fine growth culture was transferred to 45 ml medium and incubated for 4 days, after which time cells were collected by centrifugation and sterilized with gamma irradiation at 2.0 megarad.

For scanning electron microscopy, cells were fixed with 5% glutaraldehyde in cacodylate buffer (0.067 mol, pH 6.2) for 4 hours at room temperature. Suspensions were filtered using a 0.1 um polycarbonate filter (Porectics Corporation, Livermore, CA). Filters were dehydrated in a graded series of ethanol washes and immersed in Hexamethyldisilazane (Polysciences, Warrington, PA) overnight at room temperature. Finished specimens were mounted on aluminum stubs with silver paint, coated with 30 nM of gold and observed using a Philips XL Environmental Scanning Electron Microscope (FEI Company, Portland, OR). For transmission electron microscopy, cells were placed into formalin for 1 hr, washed 3 times for 2 min each in 0.1 M phosphate buffer and transferred to a 1% solution of OsO4 for 1 hour. Cells were washed with water and dehydrated in a graded series of ethanol washes, collected and placed into embedding molds with 100% resin solution. The resin was polymerized at 50°C for 8 hours, cooled and cut into sections on a copper grid. Uranyl acetate and lead citrate were added, dried and the specimens were observed with a Phillips Tecnai™ G² Electron Microscope [14].

Total mycolates were isolated by saponification as previously described, using new, clean glassware and HPLC grade solvents [15]. Briefly, the wet cell pellet (50 mg) was transferred into 2 mL potassium hydroxide (25%, w/v) in 1∶1 methanol: water (v/v) was added and heated in at 100°C for 3 hrs. After cooling, 2 ml chloroform was added, followed by 1.5 ml hydrochloric acid in water (50%, v/v), and the fatty acids were separated by centrifugation, dried under nitrogen, treated with 100 ul potassium bicarbonate (2%, w/v) in 1∶1 (v/v) methanol and water and centrifuged. After removing supernatant, all free fatty acids were dried and re-dissolved in 1 ml tetrabutylammonium hydrogen sulfate (3.39%, w/v) in water followed by 1 ml methylene chloride and 100 ul iodomethane (Sigma) at room temperature for 30 min, centrifuged at 1000 rpm for 5 min, treated with 1 ml 1 M HCl. The lower phase was collected, dried down under nitrogen and analyzed by TLC (Silica Gel 60, Macherey-Nagel) using 5 developments with 94∶6 (v/v) petroleum ether: ethyl acetate [16], [17] and developed with 8% (v/v) phosphoric acid and 3% (w/v) cupric acetate and charring. Purification of mycolate subclasses was achieved by preparative TLC, after which bands were scraped, extracted with chloroform and used for further analyses.

Total and preparative MAMEs were analyzed by LC-ESI0 MS using an Agilent Technologies 6520 Accurate-Mass Q-Tof (Agilent, Santa Clara, CA) connected to an Agilent 1200 series HPLC system using a Varian Monochrom 3 µm×150 mm×2 mm diol column. The samples were suspended at 1 mg/ml in solvent A [Hexane:Isopropanol 70∶30 (v:v), 0.02% formic acid, 0.01% ammonium hydroxide] and filtered. Ten µg of lipid was injected, and the column was eluted at 0.15 mL/min with a binary gradient from 0% to 100% solvent B [isopropanol: methanol 70∶30 (v:v), 0.02% formic acid, 0.01% ammonium hydroxide]: 0–10 min, 0% B; 17–22 min, 50% B; 30–35 min, 100% B; 40–44 min, 0% B, followed by additional 6 min 0% B post-run as described in detail [18]. The ionization source was maintained at 325°C with a drying nitrogen gas flow of 5 L/min, a nebulizer pressure of 30 psi and a capillary voltage of 3000 V. Spectra were collected in the positive-ion mode from 100 to 3000 m/z at 1 spectra/s. Internal calibration was performed using proprietary standards of 121.050873 and 922.009798 a.m.u. Collision induced dissociation (CID)-MS experiments were performed on a LXQ 2 dimensional ion-trap mass spectrometer (Thermo Scientific) equipped with an electrospray ionization (ESI) source. For the positive-ion mode analysis, 1-D TLC purified samples were dissolved in chloroform/methanol (1∶1, v/v) at ∼10 µg/ml. Five µl of sample were mixed with a solution of 1 mM sodium acetate in methanol and loaded into a nanospray tip and analyzed by electrospray ionization-MS. Bacterial pellets of S. rugosus and S. rotundus were extracted with chloroform, which were subsequently diluted with methanol to a final concentration of ∼10 µg/ml, loaded into a nanospray tip (5 µl), and analyzed with a spray voltage and capillary temperature were set to 0.8 kV and 200°C. The collision energy was 28∼35% of maximum and the trapping of product ions were carried out with a q value of 0.2 or 0.25.

1H-NMR spectra were recorded on a Barnett Instrument 500 MHz at room temperature in 100 µg/ml deuterated chloroform in 200×5 mm 535-pp Wilmad NMR tubes with tetramethyl silane as a proton reference. Chemical shifts are expressed in ppm downfield from the signal of the TMS (0 ppm). The peak assignment for protons in cyclopropyl, cis unsaturations, trans unsaturations and α-branched, β-hydroxyl systems were based on comparison of the spectra obtained with prior published spectra [19], [20], [21], [22].

S. rotundus and S. rugosus show differing smooth and rough colony forms, and display somewhat distinct appearance by electron microscopy (Fig. 1) [10]. Transmission electron microscopy (TEM) identified two layers showing the characteristic appearance of formed lipid membranes: triple bands composed of alternating electron dense and light regions. The ultrathin TEM images for Segniliparus rugosus are interpreted as an electron transparent periplasmic space, enclosed by a triple-layer of alternating dense and transparent bands typical of a mycolate membrane and surrounded by a less electron dense outermost capsule layer. The capsular layer was irregularly shaped and was present in S. rugosus but not in S. rotundus. The multilayered appearance of the Segniliparus envelope recapitulated most known aspects of mycobacterial structures.

We methylated the free carboxyl groups of saponified mycolic acids and separated the individual mycolic acid methyl ester (MAME) classes using normal phase, one-dimensional thin layer chromatography (1D TLC). As expected, M. tuberculosis MAMEs resolved into three spots, which comigrated with commercial standards for keto-, methoxy- or α-mycolates, which contain keto, methoxy or no oxygen-containing functional groups on the meromycolate chains (Fig. 2a, left). M. tuberculosis produced ions corresponding to the expected mass of protonated adducts of MAME with keto (m/z 1280.315) or methoxy (m/z 1268.314) functional groups, as well as ammoniated α-mycolates (m/z 1169.219) (Fig. 2b, inset). These molecules of known structure provided standards for comparison with unknown forms of mycolic acids.

For Segniliparus rotundus, we detected three spots, but these did not precisely co-migrate with mycobacterial mycolates. For S. rugosus, we consistently detected two spots. To determine the chemical basis for TLC separation of Segniliparus MAMEs into two or three subclasses, we carried out positive-ion mode LC-ESI-MS with high accuracy TOF detection (Fig. 2b, Fig. S1 and S2). S. rotundus and S. rugosus produced mycolic acids varying over a much wider mass range, as compared to M. tuberculosis. This difference in molecular size was particularly apparent for S. rotundus, which were detected from m/z 946.978 to 1461.452, which was more than twice the mass variance for individual molecular species of MAMEs observed in M. tuberculosis, which spanned from m/z 1152.191 to 1352.406 (Fig. 2b). Further, the overall ion profiles showed MAMEs from S. rotundus were segregated into three groups, comprised of three alkane series that differed in the deduced number of carbon atoms, whereas S. rugosus forms only two such groups. These overall patterns suggested that the three nearly separate bands in S. rotundus likely correspond to long, medium and short chain mycolates, whereas S. rugosus produced two spots, likely corresponding to medium and long chain MAMEs.

To test this hypothesis directly, each band was scraped and analyzed by preparative TLC and mass spectrometry. For S. rotundus, the most intense ions derived from the low, middle and high migrating bands were detected at m/z 975.021, 1127.175 and 1391.453, respectively (Fig. S3). It is notable that the middle and upper spots on TLC appear as a doublet and share ions in the intermediate mass range (Fig. S3). For S. rugosus, the middle and high bands showed distinct mass profiles with the most intense ions observed at m/z 1141.192 and 1363.424, respectively (Fig. S4). Thus, preparative TLC with mass spectrometry analysis confirmed that the three discretely eluting classes of Segniliparus rotundus and two classes of S. rugosus MAMEs represent three and two chain length variants respectively. Thus, mycobacteria and segniliparus species have different organizing principles of mycolate class structure, with the former based on oxygen substitutions and the latter based on the carbon chain length. This general hypothesis was further tested by detailed analyses of individual MAMEs in each subclass.

For α-mycolates, the general chemical structure is deduced to give an [M+H]⁺ ion corresponding to O(3), C(X+1), H[2(X+1-Y)+1], where X is the number of carbons and Y is the number of unsaturations and cyclopropyl groups. Ammoniated adducts correspond to O(3), N(1), C(X+1), H[2(X+1-Y)+4]. For example, S. rotundus ions detected at m/z 1374.427 and 1391.454 were deduced as C₉₅H₁₈₅O₃ and C₉₅H₁₈₈NO₃, respectively (Figs. S2 and S1). The former is a proton adduct, while the later is an ammonium adduct, with both molecules having a chain length of C94 (X = 94) with three unsaturations (Y = 3). The mass error for both ions is 0.005 atomic mass units (amu) compared with the calculated m/z 1374.432 and 1391.459. Furthermore, MAME with oxygenated R groups deviate from this prediction in defined ways, with the key prediction that four, rather than three oxygen atoms are present. Systemic application of these rules to all ions detected by high accuracy measurements from the TOF detector to the 62 molecular ions for S. rotundus and 46 ions in S. rugosus, which allowed assignment of the chemical composition of 37 molecular forms of S. rotundus and 28 molecular species of S. rugosus. The molecular species are summarized in Fig. 3 and ions are comprehensively detailed in Figs. S1 and S2. The chemical composition assignments were considered reliable because ions were detected within narrow ranges of error expected for the method, except for compounds at the outer ranges of the chain length spectrum, which produced trace signals for which larger mass errors are expected. Further evidence for the chemical assignments included the relative intensity of isotopes, which matched those of high carbon content molecules and received high scores for isotope matching (not shown). Further, most deduced neutral structures [M], were detected both as protonated and ammoniated species (Figs. S2 and S1).

The chemical compositions of these 65 molecules confirmed that the chemical features, which organize Segniliparus mycolates into 3 alpha mycolic acid subclasses. In all cases the deduced chemical formula for Segniliparus mycolates contains three oxygen atoms. Assuming that these are carried on the carboxyl ester and the β hydroxyl group, no oxygen-containing R groups are present on the meromycolate chains. The alpha mycolic acids separated into α⁺-mycolates (C84–C100, m/z 1251.302 to 1475.518), α-mycolates (C72–C84, m/z 1085.130 to 1253.305) and short α’-mycolates (C58–C70, m/z 890.920 to 1059.109). The average length of Segniliparus α-mycolates (C94) greatly exceeded those lengths of mycolates in M. tuberculosis (C79), which are normally known as long chain mycolates (Fig. 2b), leading to their designation as ultralong mycolates. The C100 mycolate described here is to our knowledge the longest known natural mycolic acid.

To confirm the length of the alpha branch, we carried out nanospray MS with collision-induced dissociation (CID) analysis (Fig. 4). In the positive-ion mode analysis, we found that ions corresponding to sodium adducts [M+Na]⁺ of C92 and C94 MAMEs from the S. rugosus α⁺ fraction, as well as C78 MAMEs, each generated a fragment of m/z 354 (Fig. 4a and data not shown). This fragment is attributed to a C22∶0 fatty methyl ester, which was cleaved at the C2–C3 position [22]. Because the product ions, including the meroaldehyde chain, are not always stable in the positive-ion mode, we sought to improve sensitivity by carrying out similar experiments on free mycolic acids (MA) in the negative ion mode [23], [24]. In both species, we isolated ions at m/z 1386 (C96 MA), 1330 (C92 MA), and 1136 (C78 MA), which were deduced as [M-H]⁻ of C96 MA, C92 MA and C78 MA. Additionally, for S. rotundus, we detected m/z 969, which was deduced as C66 MA. Each of these ions was subjected to CID-MS analysis. S. rugosus mycolates all generated single product ion 339 and no detectable chain length series (Fig. 4b). In contrast, ions 339 and 367 were found in S. rotundus (Fig. 4c). We interpret these patterns as α-chain derived fatty acyl chains of C22∶0 in S. rugosus, but C22∶0 and C24∶0 in S.rotundus. These results are in agreement with a prior pyrolysis analysis [10].

These initial conclusions, based on mass measurements, were further tested by 1H-NMR analysis of each class. For the high migrating S. rotundus doublet, there is overlap in chain length of the α⁺ and α mycolates. Therefore, these two samples are considered enriched for homogenous chain length, but are not pure class-specific mixtures (Fig. S3). Therefore, most 1H-NMR analyses was carried out on purified preparations of S. rugosus α⁺ and α-mycolates, as well as S. rotundus α’-mycolates (Fig. 5a). The proton resonances of mycolates were assigned based on published studies [19], [20], [21], [22]. Strong signals corresponding to methylene (CH₂) protons in the mycolate chains were seen at 1.2–1.5 ppm, and the terminal methyl (CH₃) protons appeared at 0.88 ppm. Characteristic protons of the Cα branched (H_a, 2.43 ppm, 1 H, m), the hydroxylated Cβ (H_b, 3.65 ppm, 1 H, m), and the terminal methyl group in ester linkage to the mycolate (3.71 ppm, 3 H, s) were identified in all five lipid samples, confirming the presence of the α-branch, β-hydroxyl and carboxylate, which define these lipids as mycolic acids.

The presence, number and stereochemistry of oxygen-containing R groups, unsaturations or cyclopropyl groups in the meromycolate chain are potentially important because these deviations from otherwise straight alkane structure influence the packing and thereby affect the fluidity of the mycolate membrane [3], [25]. We determined the number and orientation of the functional groups in these purified α-mycolates by the ¹H-NMR chemical shifts and the integration of proton signals. As predicted by mass spectrometry, signals for methoxy-mycolate protons (3.3 ppm, 3 H) and for keto-mycolate, protons of CH₂ (∼2.4 ppm, 2 H) and CH (∼2.5 ppm, 1 H), were absent from all 5 preparations of Segniliparus MAMEs, confirming the lack of oxygenated structures in the meromycolate chain.

In all purified mycolates, characteristic olefinic protons were identified. The signals for cis-double bond protons (H_f, H_f’) were at 5.35 ppm and trans-double bond protons (H_e, H_e’) were at 5.31 and 5.22 ppm. Methylene protons adjacent to the double bonds were located at 2.0 ppm and 1.98 ppm for cis and trans, respectively (Fig. 5a, S5). The signals for branched methyl groups adjacent to a trans double bond resonated at 0.94 ppm. In addition to simple unsaturations, the signals at −0.34, 0.55 and 0.64 ppm, attributed to the protons on the cis di-substituted cyclopropane ring, were detected in ultralong α⁺ mycolates of both strains and α mycolates of S. rotundus (Fig. 5a top panel and Fig. S5). The 1H-NMR results clearly indicated that Sengiliparus mycolates contain only double bonds and cyclopropanation as functional groups in their meromycolate chains.

The relationship of olefinic protons, two for each double bond, and a total of four protons on the Cβ (3.65 ppm) and methyl group (3.71 ppm), provide the basis for integration to establish the number and configuration of double bonds in each class (Fig. 5a). Based on integration of proton signals, S. rugosus α⁺ mycolates contain 3 unsaturations, which are deduced to be 60 percent cis and 30 percent trans double bonds, with trace levels of cis cyclopropanation. S. rugosus α mycolates contain 2 unsaturations, which are deduced to be comprised of 80 percent cis and 20 percent trans double bonds. S. rotundus α’ mycolates contain one double bond, which gives signals corresponding to the cis-orientation only. NMR signals for S. rotundus α⁺ and α preparations independently support these conclusions, but cannot be interpreted quantitatively due to the enriched, rather than pure nature of these preparations (Figs. S3 and S5).

Mycolates are synthesized as a homologous series of molecules that differ by C₂H₄ in length. Trans- but not cis-double bonds introduce an extra methyl branch, which contributes to C₃H₆. Therefore, mixed MAMEs comprised of molecules possessing and lacking trans unsaturations or cis-cyclopropanations typically appear as two overlapping alkane series in which each molecule differs by CH₂. For the molecules only containing double bonds, these rules predict that odd chain mycolates are seen only in trans unsaturated molecules. Cross comparison of mass spectra (Fig. S3, S4) and NMR (Fig. 5a) indicate that the ratio of even to odd carbon numbers correlates with the percentage of cis unsaturated to trans unsaturated. For example, NMR of S. rotundus α’ mycolates indicates pure cis mycolates and is observed as a single series of molecules that differ by C₂H₄ from one another. Mixed cis and trans mycolates from S. rugosus α and α⁺ classes showed a minor series of odd-numbered mycolates in which the intensity of the minor odd chain species correlates with the predicted trans unsaturations from NMR data. The combined results of the mass spectrometry and NMR analyses identify chain length as the key feature, which separates these 3 classes and support the assignment of meromycolate structures within each class, as summarized in Fig. 5b.