Cell Wall Structure of a Mutant of Mycobacterium smegmatis Defective in the Biosynthesis of Mycolic Acids*

A mutant strain of Mycobacterium smegmatis defective in the biosynthesis of mycolic acids was recently isolated (Liu, J., and Nikaido, H. (1999) Proc. Natl. Acad. Sci. U. S. A. 96, 4011–4016). This mutant failed to synthesize full-length mycolic acids and accumulated a series of long chain β-hydroxymeromycolates. In this work, we provide a detailed characterization of the localization of meromycolates and of the cell wall structure of the mutant. Thin layer chromatography showed that the insoluble cell wall matrix remaining after extraction with chloroform/methanol and SDS still contained a large portion of the total meromycolates. Matrix-assisted laser desorption/ionization and electrospray ionization mass spectroscopy analysis of fragments arising from Smith degradation of the insoluble cell wall matrix revealed that the meromycolates were covalently attached to arabinogalactan at the 5-OH positions of the terminal arabinofuranosyl residues. The arabinogalactan appeared to be normal in the mutant strain, as analyzed by NMR. Analysis of organic phase lipids showed that the mutant cell wall contained some of the extractable lipids but lacked glycopeptidolipids and lipooligosaccharides. Differential scanning calorimetry of the mutant cell wall failed to show the large cooperative thermal transitions typical of intact mycobacterial cell walls. Transmission electron microscopy showed that the mutant cell wall had an abnormal ultrastructure (without the electron-transparent zone associated with the asymmetric mycolate lipid layer). Taken together, these results demonstrate the importance of mycolic acids for the structural and functional integrity of the mycobacterial cell wall. The lack of highly organized lipid domains in the mutant cell wall explains the drug-sensitive and temperature-sensitive phenotypes of the mutant.

The lack of effective treatment for mycobacterial infections by most of the broad spectrum antibiotics has been largely attributed to the extremely low permeability of the mycobacterial cell wall (1)(2)(3). This cell wall has a unique structure, consisting of three covalently linked polymers: peptidoglycan, arabinogalactan (AG), 1 and mycolic acid (4). The peptidoglycan is attached to AG via a phosphodiester bridge (5). About twothirds of the nonreducing termini of the AG polysaccharide are esterified with mycolic acids (6). This covalently linked skeleton of cell wall is often described as the mycolyl-arabinogalactan-peptidoglycan complex. One of the most remarkable features of the mycobacterial cell wall is that up to 60% of its weight is composed of lipids, including mycolic acids. In addition to lipids in the covalently linked skeleton, several types of "extractable lipids" are present in various mycobacterial species, including trehalose-containing glycolipids, phenolic glycolipids, glycopeptidolipids (GPLs), lipooligosaccharides (LOSs), phosphatidylinositol mannosides (PIMs), phosphatidylethanolamine (PE), and triacylglycerols (TAGs) (4,7,8).
Mycolic acids are long, complex ␣-alkyl-␤-hydroxy fatty acids that are unique to mycobacteria and the closely related genera (3,4,9). Mycolic acids play a critical role in the structure and function of the mycobacterial cell wall. They constitute the inner leaflet of the lipid bilayer of cell wall and have extremely low fluidity (3,10,11). It is this nature of mycolic acids that accounts for the exceptionally low permeability of the mycobacterial cell wall and explains, in large part, the natural resistance of mycobacteria to many antibiotics and chemotherapeutic agents (1)(2)(3). This mycolic acid-based permeability barrier also shields mycobacteria from environmental stress and contributes to disease persistence during infection (12). In addition, mycolic acid-containing glycolipids such as trehalose dimycolate (cord factor) have been implicated in the pathogenesis of Mycobacterium tuberculosis (9,13).
The biosynthesis of mycolic acids has been the focus of intense research for a number of years, primarily because of the presumption that enzymes involved in the synthesis of this unusual lipid are attractive targets for the development of novel chemotherapeutic agents. Identifying new drug targets is now particularly important in view of the prevalence of multidrug-resistant tuberculosis (14) and the increasing association of atypical mycobacteria such as the Mycobacterium aviumintracellulare complex with AIDS patients (15). In an effort to elucidate the biosynthetic pathway of mycolic acids, we recently isolated a mutant of Mycobacterium smegmatis that had a defect in the synthesis of mycolic acids (16). This mutant, 155NS1, was generated by chemical-induced random mutagenesis of the wild type (WT) strain mc 2 -155. The 155NS1 mutant was unable to synthesize full-length mycolic acids and accumulated a series of ␤-hydroxymeromycolates with sizes ranging from 36 to 48 carbons (16). This mutant was more permeable than the WT strain to hydrophobic agents and exhibited hypersensitivity to various hydrophobic compounds such as novo-biocin, rifampicin, erythromycin, and crystal violet. Also, this mutant was temperature-sensitive; i.e. although it grew at a normal rate at 30°C, it did not show any visible growth at 37°C. In this work, the structure and composition of the cell wall of this mutant were studied in detail. We show that the ␤-hydroxymeromycolates that accumulate in the mutant strain are covalently attached to AG of the cell wall, but they do not form highly organized structural domains, and that mutant cells have an abnormal colony morphology and cell wall ultrastructure.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-The strain 155NS1 is a mycolic acid-defective mutant of M. smegmatis mc 2 -155 isolated previously (16). Since this mutant cannot grow at 37°C, strains 155NS1 and mc 2 -155 both were grown at 30°C in all experiments described in this study. Mycobacteria were grown in Middlebrook 7H9 broth (Difco) supplemented with 0.2% glycerol and 10% Middlebrook OADC enrichment (Difco). Cells were harvested at midexponential phase.
Isolation and Fractionation of the Cell Wall-The isolation and purification of cell wall were as described previously (10,17). Briefly, mycobacterial cells were disrupted by sonication, and the cell wall fraction was separated from the cytoplasmic membrane by centrifugation in a sucrose step gradient. It was previously shown that this method minimized contamination from the cytoplasmic membrane (17). Purified cell wall (0.75 g, wet weight) was extracted twice with 10 ml of CHCl 3 /CH 3 OH (2:1, v/v) for 1 h at 50°C, which released the extractable lipids including GPLs, LOSs, and PIMs. After centrifugation at 10,000 ϫ g for 30 min, the supernatant was collected and analyzed by TLC (see below). Delipidated cell wall was treated twice with 2% SDS in 20 mM potassium phosphate buffer (pH 7.0) at 90°C for 1 h and each time pelleted at 10,000 ϫ g for 30 min, which removed most proteins and lipoarabinomannan. The cell wall pellet was then washed successively with water, 80% acetone, and acetone to remove SDS. The resulting insoluble cell wall matrix is known as the mycolyl-arabinogalactan-peptidoglycan complex in WT cells, in which all soluble proteins, extractable lipids, and carbohydrates were removed (6,18). Since we did not know whether the ␤-hydroxymeromycolates in the mutant strain were attached to the cell wall, we referred to this fraction of cell wall as the insoluble cell wall matrix.
Smith Degradation of the Cell Wall-The insoluble cell wall matrix, defined above as the product remaining after extraction with chloroform/methanol and SDS, was subjected to Smith degradation by a method previously described (6). Briefly, the cell wall was sonicated to a fine suspension in 10 ml of 0.1 M NaIO 4 in 50 mM sodium acetate buffer (pH 5.0) and kept in the dark for 5 days. After extensive washing in water, the insoluble residue was reduced by stirring with 130 mg of NaBH 4 in 10 ml of water for 5 h. The reaction mixture was titrated to pH 5 with acetic acid, pelleted, and treated with 1 N HCl at room temperature overnight. The insoluble residue was washed with water and extracted with CHCl 3 . The CHCl 3 -soluble products were subjected to analysis by electrospray ionization (ESI) and matrix-assisted laser desorption/ionization (MALDI) mass spectroscopy.
Mass Spectrometry Analysis of Smith Degradation Products-MALDI mass spectra were acquired on a Voyager-DE STR MALDI-TOF mass spectrometer (PerSeptive Biosystems, Framingham, MA), equipped with a pulsed nitrogen laser emitting at 337 nm. Linear DE mode was used for acquiring MS spectra. The accelerating voltage was 20 kV. MALDI-MS spectra were acquired in a 2,5-dihydroxybenzoic acid matrix. For the sample preparation, 1 l of sample in CHCl 3 was mixed with 1 l of 10 mg/ml 2,5-dihydroxybenzoic acid dissolved in CHCl 3 and methanol (1:1), and 1 l of mixture was applied on the sample plate and allowed to evaporate to dryness in air. ESI-MS spectra were acquired on an API III ϩ triple quadrupole mass spectrometer (PE Sciex, Thornhill, Canada) fitted with an articulated pneumatically assisted nebulization probe. Samples were introduced into the electrospray ionization source at a flow rate of 30 l/min using a liquid chromatography pump. The electrospray needle was operated at Ϫ3.8 kV, orifice voltage was set at Ϫ50 V, and nitrogen was used as the nebulization gas. All mass spectra were obtained at the Mass Spectrometry Laboratory, Molecular Medicine Research Center, University of Toronto.
NMR Analysis of Arabinogalactan-The AG polysaccharide was released from the peptidoglycan as described (18). The insoluble cell wall matrix was treated with 2 N NaOH for 16 h at 80°C. The supernatant obtained by centrifugation (27,000 ϫ g, 30 min) was neutralized with acetic acid and dialyzed to remove salt. The supernatant was brought to 80% with ethanol and kept at Ϫ20°C to precipitate AG. The AG was solubilized in 0.5 ml of D 2 O, and the nondissolved materials were removed. 1 H NMR was performed on a Varian UNITYplus NMR (500 MHz) (NMR Laboratory, Molecular Medicine Research Center, University of Toronto). The D 2 O signal at 4.773 ppm at 25°C was used as the internal standard for assignment.
Electron Microscopy-Preparation of specimens was as described previously (26). Cells were fixed in 4% paraformaldehyde, 2.5% glutaraldehyde and postfixed in Karnovsky's 0.5% OsO 4 , 0.8% K 3 Fe(CN) 6 followed by 1% tannic acid. Cells were stained with 1% uranyl acetate, dehydrated in a graded ethanol series, and embedded in Spurr's resin. Thin sections were cut with an RMC MT-7000 ultramicrotome, poststained first in 1% uranyl acetate and then in Reynold's lead citrate. Electron microscopy was performed with a Philips CM-10 transmission electron microscope operating at 80 kV (Microscopy Branch, Rocky Mountain Laboratories of NIAID, National Institutes of Health).

Location of Meromycolates in the Cell Wall of 155NS1-We
previously showed that the ␤-hydroxymeromycolates were accumulated in the 155NS1 mutant (16). To examine any potential involvement of the meromycolates in the cell wall function, the localization of meromycolates in mutant cells was determined, which was achieved by fractionating the cell wall into several components and analyzing for the presence of meromycolates in these different materials. We first examined whether the meromycolates present in the cell wall of the mutant could be extracted with chloroform/methanol. Equal amounts of the extractable lipid fractions and the delipidated cell wall suspensions were saponified, derivatized quantitatively with p-bromophenacyl bromide, and analyzed by a reversed phase HPLC (16). The molar ratio of meromycolates in various fractions was obtained directly from the integrated peak areas compared with that of an internal standard (e.g. oleic acid). Results showed that 48% of the meromycolate residues in the original cell wall of the mutant were found in the organic phase, compared with 32% of mycolic acids in the cell wall of the WT strain in parallel experiments.
We further examined whether the meromycolates are present in the insoluble cell wall matrix of the mutant. TLC showed that upon saponification and extraction with CHCl 3 , ␤-hydroxymeromycolates were released from the insoluble cell wall matrix of the mutant, in amounts comparable with that of mycolic acids released from the WT cell wall (Fig. 1). These results suggested that the ␤-hydroxymeromycolates accumulated in the mutant were attached to the cell wall, presumably by a linkage to AG. No meromycolates were detected in the growth medium (data not shown).
Linkage between Meromycolates and Arabinogalactan-To obtain direct evidence for the attachment of meromycolates at the cell wall skeleton, the insoluble cell wall matrix of the mutant was subjected to the Smith degradation reaction (6). The Smith degradation products were extracted with CHCl 3 , and the resulting fractions were analyzed by mass spectroscopy. The positive ion MALDI-time-of-flight (TOF) mass spectrum of the Smith degradation products of the mutant cell wall is shown in Fig. 2A. The most abundant ions, (M ϩ H) ϩ , are at m/z 666 and 680. The peaks at 688 and 702 correspond to adduct of sodium (M ϩ 23) to the molecular ions, respectively. The signal at m/z 680 corresponds to a ␤-hydroxymeromycolylglycerol, and the molecular weight of this ␤-hydroxymeromycolate is 605, which was the most abundant meromycolate species detected in the mutant (16). The peak at m/z 666 may correspond to another ␤-hydroxymeromycolate (M r 591) substituted with glycerol. The Smith degradation products of the mutant cell wall can be explained if we assume that the ␤-hydroxymeromycolates in the mutant are attached to the C-5 positions of the terminal-and 2-linked arabinofuranosyl residues of AG (Fig. 2B). Accordingly, terminal-and 2-linked arabinofuranosyl units of AG substituted at C-5 with the meromycolate residues should produce CHCl 3 -soluble meromycolylglycerols. The expected molecular weights of meromycolylglycerols are in agreement with the result from MALDI-MS analysis; i.e. esterification of much of meromycolates with glycerol results in an increase in molecular weights by 75.
When the same sample was analyzed by ESI-MS with a negative ionization mode, more signals were evident (Fig. 2C). Two series of ions, (M Ϫ H) Ϫ , one at m/z 623, 651, 679, 707, 735, and 763, and the other at m/z 609, 637, 665, 693, 721, 749, and 777 were detected. Based on the above analysis, we believe that the series of peaks at m/z 623, 651, 679, 707, 735, and 763 correspond to ␤-hydroxymeromycolates with molecular weights of 549, 577, 605, 633, 661, and 689 (16) substituted with glyc-erol, respectively. We previously showed that these ␤-hydroxymeromycolates are likely to represent intermediates in the chain elongation cycles of the biosynthetic pathway of mycolic acids, and that each cycle adds a two-carbon unit to the meromycolates (16). The other series of ions at m/z 609, 637, 665, 693, 721, 749, and 777 in the ESI-MS spectrum may correspond to meromycolate homologs containing odd-numbered carbons substituted with glycerol. These meromycolate species were observed in our previous study, although they were present in a lower abundance (see Fig. 5 of Ref. 16). In mycolic acids, the odd-numbered carbon-containing homologs represent mycolic acids that contain a distal cis-double bond and a proximal trans-double bond with an adjacent methyl branch in the meromycolate chain (9,13). In the WT cell wall, it was previously shown that mycolic acids were attached to C-5 positions of the terminal and 2-linked arabinofuranosyl residues of AG (6,28) and that Smith degradation produced a CHCl 3 -soluble mycolylglycerol (Ref. 6; also see Fig. 2B). Consistent with this result, MALDI-TOF showed major peaks at m/z 1211, 1225, 1239, 1253, 1267, 1281, and 1295 (data not shown), which correspond to the free mycolic acids at m/z 1136, 1150, 1164, 1178, 1192, 1206, and 1220 substituted with glycerol, respectively. Based on these results, we concluded that the meromycolates were attached at the same position (C-5 of the arabinofuranosyl residues of AG) in the mutant cell wall as mycolic acids in the WT cell wall (Fig. 2B).
NMR Analysis of the Arabinogalactan-To examine whether the structure of AG in the cell wall of the mutant was altered, AG released from the cell wall skeleton by treatment with NaOH was analyzed by 1 H NMR. For comparison, AG from the cell wall of the WT strain was solubilized and analyzed in the same way. The proton NMR spectrum of the AG in the cell wall of the mutant was identical to that of the WT strain (data not shown), suggesting that the bulk structure of AG in the cell wall of the mutant is the same as that in the WT strain. An interesting question arising from these results is which enzymes are responsible for the transfer of meromycolates to AG of the cell wall. The antigen 85 complex (Ag85A, -B, and -C) is known to possess mycolyltransferase activity and has been shown to catalyze the transfer of mycolic acids to trehalose (29). It was suggested that the Ag85 complex was also responsible for deposition of mycolic acids on AG, although direct evidence has yet to be shown. Inactivation of the Ag85C gene was shown to transfer 40% fewer mycolates to the cell wall (30). It is not clear at present whether the same enzyme is responsible for the transesterification of the meromycolates to AG in the mutant cell wall.
TLC Analysis of CHCl 3 -extractable Lipids-The organic solvent-extractable lipids of the cell wall were analyzed by TLC. It was shown that the cell wall of M. smegmatis contains GPLs, LOSs, PIMs, PE, DPG, and TAGs (4,7,8). To determine whether these lipids are present in the mutant, we applied TLC developing systems that were previously established to examine each group of lipids (see "Experimental Procedures"). Twodimensional TLC system A was used for analysis of apolar lipids (e.g. TAGs). For analysis of polar lipids such as PIMs, PE, and DPG, two-dimensional TLC developed with solvent system B was used. In all experiments, lipids extracted from the cell wall of the WT strain under the same conditions were compared. The results are summarized in Table I. The cell wall of the mutant contained some of the extractable lipids including PIMs, PE, TAGs, and DPG but lacked GPLs and LOSs.
DSC of Cell Wall from the Mutant-We previously showed that the mycobacterial cell wall forms an asymmetric lipid bilayer with a mycolic acid-containing inner leaflet covered by an outer leaflet composed of extractable lipids (10, 17). We have  shown above that the ␤-hydroxymeromycolates accumulated in the mutant were attached to AG of the cell wall and that the mutant cell wall contained some of the extractable lipids. To examine whether these lipid components form an organized structure in the mutant cell wall, DSC analysis of the purified cell wall of the mutant was performed. No cooperative thermal transitions were observed in the mutant cell wall (Fig. 3A), indicating that lipids in the cell wall of the mutant did not form highly organized structural domains. DSC of the WT cell wall had major peaks between 30 and 60°C and reflected the melting of tightly packed mycolic acids ( Fig. 3B; also see Refs. 10, 11, and 27).
Altered Colony Morphology of Mutant Cells-The cell morphology and the cell wall ultrastructure of the mutant were directly examined by transmission electron microscopy (Fig. 4). Mutant cells appear to have a significantly larger average radius and tend to be more amorphous and less rodlike than WT cells (data not shown). The cytoplasm of mutant cells was centered with large electron-translucent voids, and they were often detached from the cell wall materials (Fig. 4). Such features were not evident in WT cells. The cytoplasm of WT cells possessed a uniform appearance of heavily stained ribosomes that were difficult to distinguish in the cytoplasm of mutant cells. These differences in cytoplasmic ultrastructure were previously observed in WT cells processed by different techniques (e.g. dehydration with acetone rather than ethanol during the embedding of the specimens) and reflected perturbation of the ultrastructural features during sample preparations (31). These data suggest that the structural integrity of mutant cells is compromised by the fixation procedure, which is expected, considering the altered cell wall of the mutant (see below).
There were striking differences in the cell envelope profiles between the mutant and the WT strain. Although mutant cells had a similar plasma membrane and a peptidoglycan layer as WT cells, they appeared to lack the typical thick electrontransparent zone outside of the plasma membrane (Fig. 4). The electron-transparent zone constitutes the hydrophobic domain of the cell wall and has been thought by many workers to correspond to mainly the mycolic acids covalently bound to AG (3,4,31). Transparency to electrons in this ultrastructure is explained by the extremely hydrophobic nature of mycolic acids, which excludes water-soluble, electron-dense heavy metal salts such as the uranyl acetate used in this study. In our previously proposed cell wall bilayer model (10), the mycolic acids extend upward from AG to fill the electron-transparent zone of the cell wall and interact with the extractable lipids. The arabinan chains are anchored fairly close to the reducing end of galactan, which itself is linked to peptidoglycan via linker disaccharide phosphate (3,10). This model is in agreement with the results presented here. The disappearance of the electron-transparent zone in the mutant cells agrees with the lack of mycolic acids and with the fact that although ␤-meromycolates were present in the cell wall they did not form a highly organized structure. These results were also consistent with a recent study that showed that treatment of M. avium with sublethal concentrations of isoniazid sufficient to inhibit the synthesis of mycolic acids resulted in the complete loss of this electron-transparent zone (26).
Loss of the electron-transparent zone in mutant cells coin-cides with the appearance of disorganized fibrils on the cell surface (Fig. 4). The identity of these thin fibrils is unknown, and they are probably composed of the ␤-hydroxymeromycolates accumulated in mutant cells, which were shown above to attach the cell wall onto AG in the same manner as mycolic acids. This is somewhat different from the situation of M. avium treated with isoniazid described above (26), where the loss of the electron-transparent zone did not coincide with the appearance of obvious shedding materials (the samples were processed for electron microscopy in identical ways; see Ref. 26). Instead, a more densely staining thick layer appeared. Correspondingly, no fatty acid intermediates with lengths equivalent to the ␤-hydroxymeromycolates accumulated in the 155NS1 mutant were detected in the cell wall of M. avium treated with isoniazid (26). In addition to meromycolates, some  3. Differential scanning calorimetry of the cell wall. Cell wall materials from the mutant (A) and the wild type (B) were prepared and analyzed by DSC as described previously (10,11,27). of the extractable lipids including TAGs, PIMs, DPG, and PE were also present in the mutant cell wall (Table I), and they are likely to be components of the fibrils in mutant cells. In WT cells, there is an outermost electron-dense layer that was also absent in mutant cells (Fig. 4). This layer was visualized by staining with ruthenium red in earlier studies (3,4). It varies in thickness (from negligible to massive), electron density, and appearance (fibrillar, granular, or homogeneous), which is attributable to differences in species, growth conditions, and preparation methods for microscopy (3,4). The fact that ruthenium red allowed this layer to be consistently visualized suggests that the minimal structure is probably composed of negative charged head groups of lipids. In the cell wall bilayer model, such lipids include many of the extractable lipids like GPLs, LOSs, and PIMs, which interact with mycolic acids and constitute the outer leaflet of the cell wall bilayer (3,10). Loss of the outermost electron-dense layer in the mutant may correlate with the lack of GPLs and LOSs in the cell wall (Fig. 4, Table I).
Lack of a highly organized lipid bilayer in the mutant cell wall explains many of the physiological phenotypes of mutant cells. Thus, mutant cells became hypersensitive to various hydrophobic drugs and were extremely permeable to hydrophobic compounds (16). The structural characteristics of the mutant cell wall may also explain why mutant cells cannot grow at 37°C. We previously showed that mycobacteria maintain proper cell wall fluidity by changing the structure and composition of mycolic acids in response to growth temperature. For example, an increase in the growth temperature from 20 to 45°C increased the chain length of mycolic acids by 2-4 carbons and the percentage of trans-mycolate from 19 to 59% in M. smegmatis (11). Both of these changes resulted in a decrease in the cell wall fluidity (11). Yuan et al. recently showed that M. tuberculosis H37Rv and M. bovis BCG strains that lack ketomycolates were severely defective for growth within macrophages, establishing a critical role for mycolate composition in proper cell wall function (32). The ␤-hydroxymeromycolates in the cell wall of the mutant, with sizes ranging from 36 to 48 carbons, did not form a organized structure and thus are unlikely to maintain the proper cell wall fluidity to support the growth of mutant cells at 37°C (16). On the other hand, the fact that the mutant did grow at a lower temperature, e.g. 30°C, is intriguing. It is generally thought that the structural integrity of the mycobacterial cell wall is essential for the survival of mycobacteria. This is true in M. tuberculosis. Isoniazid is a potent drug with exquisite specificity for M. tuberculosis. In M. tuberculosis, isoniazid inhibits mycolic acid synthesis by interfering with enzymes involved in the chain elongation reaction (16,33,34). The lethal effect of isoniazid on mycolic acid synthesis parallels the time course of loss of M. tuberculosis viability. However, it was recently shown that M. avium formed colonies at 37°C at isoniazid concentrations that are inhibitory for mycolic acid biosynthesis (26). Thus, M. smegmatis appears to like M. avium and is able to compensate for the loss of mycolic acids by unknown mechanisms.
In summary, the availability of a mycolate-defective mutant allows us to directly study the structure-function relationship of the mycobacterial cell wall in great detail. The results pre-sented in this paper demonstrate the importance of mycolic acids for the structural and functional integrity of the mycobacterial cell wall. Our better understanding of the structure of the mycobacterial cell wall, and of the contribution of various cell wall components to the cell wall structure and its role in drug resistance, will help in designing a strategy to overcome this permeability barrier.