Acid-Fast Positive and Acid-Fast Negative Mycobacterium tuberculosis: The Koch Paradox

ABSTRACT

Acid-fast (AF) staining, also known as Ziehl-Neelsen stain microscopic detection, developed over a century ago, is even today the most widely used diagnostic method for tuberculosis. Herein we present a short historical review of the evolution of AF staining methods and discuss Koch’s paradox, in which non-AF tubercle bacilli can be detected in tuberculosis patients or in experimentally infected animals. The conversion of Mycobacterium tuberculosis from an actively growing, AF-positive form to a nonreplicating, AF-negative form during the course of infection is now well documented. The mechanisms of loss of acid-fastness are not fully understood but involve important metabolic processes, such as the accumulation of triacylglycerol-containing intracellular inclusions and changes in the composition and spatial architecture of the cell wall. Although the precise component(s) responsible for the AF staining method remains largely unknown, analysis of a series of genetically defined M. tuberculosis mutants, which are attenuated in mice, pointed to the primary role of mycolic acids and other cell wall-associated (glyco)lipids as molecular markers responsible for the AF property of mycobacteria. Further studies are now required to better describe the cell wall reorganization that occurs during dormancy and to develop new staining procedures that are not affected by such cell wall alterations and that are capable of detecting AF-negative cells.

INTRODUCTION

Mycobacterium tuberculosis possesses a unique cell wall architecture that is distinct from both Gram-negative and Gram-positive bacteria. The cell wall consists of a thick, lipid-rich outer layer composed primarily of mycolic acids (1) (Fig. 1). This lipid layer lies on top of a layer of peptidoglycan and the polysaccharide arabinogalactan, which, in turn, are anchored to the inner lipid membrane common to all bacteria (2–4). The overall thick waxy coat renders acid-fast (AF) mycobacteria resistant to Gram staining. When stained with alternative dyes, the cell wall is resistant to decolorization with acid alcohol, thus giving these bacteria their sobriquet “acid-fast.” This unique AF property has been the basis for the continuous development of staining procedures over the past century and remains the cornerstone for the diagnosis of tuberculosis (TB), especially in low-income and middle-income countries where more than 90% of TB cases occur (5). The Ziehl-Neelsen (ZN) stain, also known as the AF stain, which is used in microscopic detection of M. tuberculosis, was originally developed independently by Ziehl and Neelsen, who improved on the early work of Koch, Rindfleisch, and Ehrlich (see below).

FIGURE 1.

Chemical structures of the major mycolic acids of M. tuberculosis. Cyclopropane rings and methyl branches are shown and annotated with the S-adenosyl methionine-dependent methyl transferases responsible for their synthesis. c, cis; t, trans.

Acid-fastness has been attributed to a number of mycobacterial cell wall components, including outer lipids, arabinogalactan-bound mycolic acids, and lipoglycans (6–8). The underlying theme in all of the proposed mechanisms is the presence of a lipid-rich, hydrophobic barrier that can be penetrated by phenol-based stains but is resistant to decolorization by acid-alcohol, although the precise molecular component(s) responsible for this unique staining property remains undetermined. Early studies of the effects of the first-line anti-TB drug isoniazid on staining characteristics of M. tuberculosis demonstrated a loss of acid-fastness after growth in the presence of isoniazid (9). Because isoniazid is known to inhibit the synthesis of mycolic acids (10), this observation suggested that these lipids may be the components responsible for AF staining and that mutants defective in mycolic acid production would be suitable candidates to study the phenomenon of acid-fastness. This stimulated more recent studies using defined M. tuberculosis mutants deleted in mycolic acid biosynthetic genes, culminating in the discovery that deletion of the β-ketoacyl acyl carrier protein (ACP) synthase encoded by kasB caused alterations in mycolic acid structure ultimately resulting in a loss of acid-fastness (11).

Nevertheless, it has been known for a long time that conversion of actively replicating AF-positive bacilli into dormant AF-negative bacteria can occur during infection in both human patients and in animal models (12). This loss of acid-fastness, also known as Koch’s paradox, has recently been linked to physiological metabolic changes as well as modifications of the cell wall composition and/or architecture that must occur in the dormant bacilli. This may have important consequences for the diagnosis of TB as well as in clinical epidemiology.

This review article will mainly focus on Koch’s paradox, highlighting recent contributions of new in vitro and cellular models to address the metabolic changes characterizing and linking persistence with loss of acid-fastness.

A BRIEF HISTORY OF AF STAINING

On 24 March 1882, Robert Koch gave his breakthrough lecture “Die Aetiologie der Tuberkulose” (13) on the discovery of the bacillus M. tuberculosis, in which he described the first successful attempt to stain M. tuberculosis by spreading mycobacteria-infected specimens on coverslips. After fixing the specimens, he dipped the coverslips into a hot, alkaline, ethanolic solution of methylene blue for 1 hour and then covered them with a solution of vesuvin (Bismarck brown) before rinsing with distilled water. This procedure allowed, for the first time, the visualization of M. tuberculosis as blue bacilli on a brown background. This important discovery in detecting the bacilli in infected human and animal specimens was quickly followed by several improvements on Koch’s technique (14). On 1 May 1882, P. Ehrlich announced at the Association of Internal Medicine in Berlin that he had developed a staining protocol whereby Koch’s methylene blue was replaced with an ethanolic solution of fuchsin or methylene violet dissolved in water saturated with aniline oil, followed by decolorization with 30% aqueous nitric acid solution and counterstaining with a yellow (Bismarck) or blue (methylene blue) dye (15). The bacilli appeared as a distinctive purple/blue on a yellow background or red on a blue background. Ehrlich’s method was deemed a “great improvement” over Koch’s staining protocol by none other than Koch himself.

On 12 August 1882, F. Ziehl published a modified version of Ehrlich’s staining protocol in which carbolic acid (phenol) was used instead of aniline oil along with a weaker acid for decolorization. The bacilli appeared as strikingly colored as in Ehrlich’s method but with the advantage that the carbol-fuchsin solution was more stable than Ehrlich’s aniline-fuchsin solution. Rindfleisch then proposed in October 1882 to enhance Koch’s protocol by heating the slides instead of dipping them in hot water, as reported by Koch. On 14 July 1883, F. Neelsen published a modified Ziehl’s staining protocol that described the use of a 0.75% fuchsin solution in 5% carbolic acid followed by decolorization with a 25% sulfuric acid solution. It took only 16 months for this method of staining of the tubercle bacilli to be standardized and to become widely recognized a decade later as ZN staining, although the technique was based on the complementary work of five researchers: Koch, Ehrlich, Ziehl, Rindfleisch, and Neelsen. From then on, M. tuberculosis would be known as an AF bacillus for its ability to resist decolorization by an ethanolic acid wash (Fig. 2).

FIGURE 2.

Staining of M. tuberculosis using ZN (left) and auramine O (right). Magnification, ×100.

Despite the successful staining of M. tuberculosis with the ZN technique, numerous researchers continued modifying the ZN protocol. In 1887, H. S. Gabbett proposed shortening the protocol by combining the decolorization with sulfuric acid and the counterstaining with methylene blue (16), with the whole process taking no more than 6 minutes to complete. In 1915, J. Kinyoun published his procedure to examine sputum from TB-infected patients using a twist on the staining method that has become known as the “cold staining” technique (17). Kinyoun stated that Gabbett’s decolorization solution did not work well on thick, unevenly spread slides and recommended the following protocol: (i) increased concentrations of fuchsin (3.1%) and phenol (6.25%) in the primary stain, which was done at room temperature, and (ii) decolorization with 3% hydrochloric acid instead of sulfuric acid (17). This protocol is the basis of the Kinyoun cold stain method, yet several modifications of this method can be found in the literature, which are described below.

V. Hallberg noticed that all the primary stain dyes used previously to stain M. tuberculosis were either water-soluble or formed semicolloidal solutions in water and decided to test the “Nachtblau” dye that forms colloidal solutions in water (18). Using a carbol-Nachtblau solution as the primary stain, followed by the ZN decolorization technique with carbol-fuchsin as the counterstain resulted in the bacilli being seen as dark blue on a red background. The new staining solution had the advantage of being stable for at least a year, and Hallberg and others reported higher sensitivity when compared to conventional ZN staining (19). However, F. Tison observed stability issues with Hallberg’s Nachtblau staining solution and proposed a cold method in which the primary stain with the Nachtblau solution was performed for 24 hours at room temperature followed by decolorization with an ethanolic solution of 25% hydrochloric acid for 3 minutes and counterstaining with orange G (20). The bacilli appeared bright blue on an orange background, which gave color-blind people, who cannot distinguish between red and blue, the ability to read AF staining slides. In 1962, T. T. Hok proposed combining the Kinyoun primary stain with Gabbett’s decolorization/counterstaining solution to produce a faster (4.5 minutes as opposed to 8 minutes for the traditional ZN staining protocol) and easier “cold staining” protocol (21). Experiencing problems with the decolorizing steps in the Hok’s cold staining protocol, J. L. Allen devised a modified hot ZN staining protocol and compared it to the traditional hot ZN staining and Hok’s cold staining methods. The modified hot ZN protocol consisted of combining the decolorizing and counterstaining steps (22). In her study of 122 pathology specimens, the majority (104) were smear-negative/culture-negative, 16 were smear-positive by the ZN method and culture-positive, and two were smear-positive by the modified ZN but culture-negative. Of the 18 smear-positive samples, the cold staining protocol detected only one. Both the ZN and the Kinyoun staining protocols are still currently used in clinical settings. The ZN primary stain consists of a 0.3% ethanolic fuchsin solution in 5% aqueous phenolic solution, whereas the Kinyoun primary stain is a 4% ethanolic fuchsin solution in 8% aqueous phenolic solution. Both methods use a 3% aqueous hydrochloric acid solution for decolorization and a 0.3% aqueous methylene blue solution for counterstaining.

The third staining technique for the detection of AF bacilli is auramine-rhodamine (AR) staining, which is a highly sensitive method but requires a fluorescence microscope. The use of the auramine dye to stain M. tuberculosis was introduced by Hagemann in 1938 (23) (Fig. 2). The auramine O staining method was more sensitive and faster in detecting tubercle bacilli than the ZN protocol, but the major drawback was the potential background fluorescence due to other bacteria/viruses or artifacts. J. Degommier recommended the use of two fluorochromes to easily distinguish M. tuberculosis from background artifacts, with auramine as the first stain, followed by decolorization and counterstaining with thiazine red, and resulting in M. tuberculosis bacilli appearing bright yellow while the artifacts or other microbes stained red (24). J. Augier instead proposed adding rhodamine to auramine to increase the contrast between tubercle bacilli and other microbes or artifacts (24). A counterstain with potassium permanganate that eliminated the fluorescence of all organisms/artifacts except for mycobacteria was later proposed (25). The authors also found that the AR combination was better at preventing the visualization of artifacts than auramine alone. The AR staining protocol, also called the Truant method, consists of staining with a solution of 1.2% auramine O, 0.6% rhodamine B, and 8% phenol in water/glycerol, followed by decolorization with 0.5% aqueous hydrochloric acid solution and counterstaining with 0.5% potassium permanganate.

Despite the apparent ease of use of these protocols, there are still numerous pitfalls to the staining of mycobacteria, which has encouraged researchers to continuously improve the staining techniques to make them faster, safer, and more reliable, especially since AF staining of sputum and sputum cultures is the primary tool for the diagnosis of active TB.

AF STAINING IN CLINICAL DIAGNOSIS OF TB

Considering the time required for sputum cultures to grow (6 to 8 weeks), analysis of sputum after AF staining is the quickest and easiest technique for diagnosing pulmonary TB. Nevertheless, AF staining faces numerous issues and, therefore, can lead to false-negative or false-positive results in clinical settings.

ZN staining is cost-effective, relatively simple, and fast to use. The processing of the samples, the thickness of the smears, the preparation and conservation of the reagents, the quality of the microscopes, the length of the primary and counterstaining, as well as the expertise of the technical staff play important roles in the sensitivity and specificity of AF staining (26–29). The sensitivity of the staining varies between 20 and 60% (30, 31), and the concentrations of the primary stain (carbol-fuchsin) and counterstain (methylene blue) were shown to be important for the detection of M. tuberculosis. The World Health Organization recommends using 0.3% carbol-fuchsin and 0.3% methylene blue, but it was suggested that, in clinical settings, staining with 1% carbol-fuchsin for 10 minutes and counterstaining with 0.1% methylene blue for 1 minute gives better results (27). Others failed to observe significant differences in sensitivity with concentrations ranging between 0.3 and 1% carbol-fuchsin, but decreasing the concentration of carbol-fuchsin to 0.1% resulted in a significant decrease in sensitivity (32).

The sensitivity of ZN staining is also related to the bacillary counts in sputum. To obtain a positive AF diagnosis from a TB patient’s sputum, around 5,000 to 10,000 bacilli per milliliter of sputum are required (33). Yet in extrapulmonary TB cases (34), TB patients coinfected with HIV (30, 35), or children infected with TB (36), the bacillary count might be lower than the minimum amount required for optimal ZN staining, resulting in poor sensitivity. To balance this issue, Shapiro and Hänscheid have shown that ZN-stained slides of unprocessed sputum, where M. tuberculosis was not easily detectable with bright-field microscopy, could be more readily observed by fluorescence microscopy (37). Other issues to consider are color blindness, given that the tubercle bacilli are seen as red on a blue background, as well as waste disposal. The ZN primary stain solution is phenol-based and is considered hazardous material requiring proper disposal.

The Kinyoun cold staining method has been found unreliable (22), giving more false-positive results than ZN staining (38, 39). The main advantage of this technique is that the duration of the different steps does not play such a critical role, although researchers have noted problems with the decolorization step and that far fewer bacilli are stained than when the samples are heated.

Staining of TB patients’ sputa with auramine O is described as the most sensitive, reproducible, and specific method for detecting M. tuberculosis (40, 41). This increase in sensitivity has been attributed to mycolic acids in the mycobacterial cell wall that retain auramine O better than carbol-fuchsin (42), although others have reported that auramine O and fuchsin are bound to nucleic acids and not mycolic acids (43). The reading of auramine O-stained slides is conducted at a lower magnification, resulting in a larger field of view and rendering the process three times faster than ZN staining (44). In paucibacillary M. tuberculosis samples, such as from extrapulmonary TB cases, auramine O staining can detect M. tuberculosis even in culture-negative samples (45, 46). Staining with auramine O instead of the ZN staining protocol also generates a lower percentage of false-negative results (47). A combination of both protocols was tested using auramine and Kinyoun carbol fuchsin solutions as the primary stain, followed by acid wash and methylene blue counterstaining and scanning the slides on bright-field and fluorescence microscopes; although application of this dual staining solution was faster than staining two slides for both methods, the proposed protocol had specific issues such as decolorizing inefficiency and increased background fluorescence (48). Finally, color-blind personnel can easily identify the tubercle bacilli in auramine O-stained smears. The main drawbacks are false-positivity due to background staining, the requirement of a fluorescence microscope, and the classification of auramine O as a carcinogen (49, 50). However, the replacement of auramine O with acridine orange, a dye that stains nucleic acids in bacteria and eukaryotic cells, was shown to be as efficient as auramine O in detecting M. tuberculosis in specimen smears (51).

THE KOCH PARADOX

Non-AF tubercle bacilli were first reported a year after Koch published his technique for detecting and identifying M. tuberculosis from TB patients (52). Although these non-AF bacilli did not retain the primary stain upon acid wash, they were able to cause TB disease when used to infect animals, could be found in lesions of patients with caseous or miliary TB, and could be grown in laboratory cultures depending on the media and strains used (53). The non-AF bacilli were either fully virulent, were fully avirulent and behaving like saprophytic bacilli, or had lost some but not all of their virulence characteristics (54). Additionally, although initially non-AF, these bacteria could regain their acid-fastness upon passage in animals (52). The loss of acid-fastness was also observed in early studies of cultures treated with isoniazid. Rist et al. had noticed that, within 24 hours of isoniazid treatment, the bacilli appeared blue with ZN staining, whereas M. tuberculosis bacilli resistant to isoniazid were red (9). When it was later discovered that isoniazid inhibits mycolic acid biosynthesis (10), the field had the first indication that cell wall organization and specifically the mycolic acid composition were involved in the ability of mycobacteria to retain the primary stain and resist the acid wash.

The non-AF phenotype of M. tuberculosis had been attributed as early as 1910 to a specific stage in its life cycle: the “resting or latent form” (55), also termed the “L-form” (56), “mycococcus” (57), or “Much’s granule” form (58). Since then, M. tuberculosis has often been described as persisting in the host in a metabolically inactive, latent state as non-AF rods. In the next section, we will attempt to present how important cell envelope components can be linked to the AF property and how changes in the cell wall composition/structure may be altered in persistent bacilli.

ACID-FASTNESS, A UNIQUE ATTRIBUTE OF THE WAXY MYCOBACTERIAL CELL ENVELOPE

Current methods to visualize bacilli within infected tissue rely on (i) ZN and AR staining (41), (ii) detecting bacterial surface proteins by immunohistochemistry or immunofluorescence (12, 59), or (iii) detecting bacterial nucleic acid by in situ hybridization or intercalating dyes (60–62). Although AF stains have been around for decades, the exact cellular component(s) of M. tuberculosis recognized by the dyes is still being elucidated. Fuchsin, the main component of ZN and Kinyoun AF stains, has been shown to stain the vastly complex lipid portion of the mycobacterial cell wall (42, 63, 64). However, little is known regarding the specific target of the combined AR stain. Whereas auramine O is believed to bind to mycolic acid (42) and nucleic acids (43, 65), the exact target of rhodamine remains unknown. However, mycobacterial genetics has been particularly useful in recent years in searching for molecular target(s) responsible for the AF property of mycobacteria.

The Importance of Mycolic Acids

KasB represents one of the two β-ketoacyl-ACP synthases involved in the final elongation steps during biosynthesis of mycolic acids (66). Disruption of kasB in Mycobacterium marinum and M. tuberculosis resulted in the loss of cording and AF staining (11, 67, 68). Moreover, the M. marinum kasB mutant was found to be more sensitive to lysozyme and to human neutrophil defensin peptide, and during infection of macrophages, there was a partial loss of phagolysosomal fusion inhibition (67). The M. tuberculosis kasB mutant produced mycolic acid chains that were two to four carbons shorter than their wild-type counterparts and oxygenated mycolic acids that were defective in trans-cyclopropanation (11). Ultrastructural analyses by conventional transmission electron microscopy failed to reveal any detectable differences in the thickness of the cell envelope between the wild-type M. tuberculosis strain and the kasB deletion mutant. However, cryo-transmission electron microscopy indicated that the region between the inner and outer membranes of the mutant, mainly composed of cell wall-anchored mycolic acids, showed a notable decrease in electron density (68). It was therefore proposed that the kasB mutant cannot synthesize tight mycolic acid bundles, thus affecting the packing of the lipid-rich layer of the mycobacterial cell wall. The reduced bundle formation results in the loss of AF staining, whereas acid-fastness of the wild-type strain may be due to the rigid cell envelope structure provided by the densely packed mycolic acids. Importantly, the M. tuberculosis kasB deletion strain was strongly attenuated, did not cause disease in infected mice, and strikingly, was able to persist at constant low levels in the lungs and spleen of mice for 450 days post-aerosol infection (11).

The regulation of mycolic acid biosynthesis has only recently begun to be unraveled, and numerous studies have shown that most essential enzymes forming the central core of type II fatty acid synthase are phosphorylated by Ser/Thr protein kinases and that posttranslational phosphorylation inhibits the activity of these enzymes in vitro (69). These enzymes include the β-ketoacyl ACP synthases KasA and KasB, the β-ketoacyl-ACP reductase MabA, the β-hydroxyacyl-ACP dehydratases HadAB and HadBC, and the enoyl-ACP reductase InhA (70–74). Recent studies identified the phosphorylation sites of KasB as Thr334 and Thr336, and to investigate the in vivo role of KasB phosphorylation in regulating mycolic acid biosynthesis, a KasB phosphomimetic mutant of M. tuberculosis was constructed in which Thr334 and Thr336 were replaced by Asp residues (75). In this mutant, constitutive phosphorylation of KasB on both Thr334 and Thr336 negatively affected the condensing activity of KasB, resulting in an altered mycolic acid chain length and a defect in trans-cyclopropanation. Importantly, this mutant strain was found to be extremely attenuated in immunocompetent and immunocompromised mice and had lost AF staining (75) (Fig. 3).

FIGURE 3.

Ser/Thr kinase-dependent signaling cascade resulting in phosphorylation of KasB and loss of acid-fastness. Modification of the cell wall composition in response to exogenous cues is central for M. tuberculosis adaptation to different environmental conditions. In response to an external signal, mycobacterial Ser/Thr kinases phosphorylate the different FAS-II components, including the β-ketoacyl ACP synthase KasB involved in the addition of the last carbon atoms during the mycolic acid elongation step. Phosphorylation on Thr334 and Thr336 decreases the condensation activity of KasB, resulting in the production of shorter mycolic acids, which probably affects the packing of the lipid layer and also results in the loss of the AF property and severe attenuation in mice.

These results provided new insights into the in vivo contribution and importance of Ser/Thr kinase-dependent phosphorylation in the control of (i) the clinically important feature of AF staining and (ii) the physiopathology of TB, suggesting that M. tuberculosis regulates these two related phenotypes through a signal transduction pathway. Interestingly, phosphorylation of KasB in Mycobacterium bovis BCG was more pronounced in stationary cultures than in replicating cultures, suggesting that phosphorylation is a mechanism by which mycobacteria might tightly control mycolic acid biosynthesis under nonreplicating conditions. It is tempting to speculate that the loss of acid-fastness in persistent infections may be linked to signaling leading to increased phosphorylation of KasB. Further work is needed to elucidate the in vivo cues that activate the appropriate kinases under nonreplicating conditions. This knowledge may lead to a better understanding of the molecular signals that trigger reactivation and TB disease.

Whereas the above-mentioned results indicate that enzymatic and signaling pathways that control mycolic acid chain length are required to maintain the AF property, other enzymatic steps introducing additional structural elements to meromycolic acid are also essential to sustain both the integrity of the cell wall and acid-fastness. M. tuberculosis produces significant amounts of cyclopropanated mycolic acids: α-mycolic acids possess two cis cyclopropanes on the meromycolate chain, whereas oxygenated mycolates contain either a distal methoxy or ketone group and a proximal cis or trans cyclopropane (Fig. 1). The cyclopropane rings as well as the methyl branches of the lipids are synthesized by a family of S-adenosyl methionine-dependent methyl transferases (1) that are highly homologous in both primary sequence and three-dimensional structure (76). Despite their structural similarity, genetic deletion of each methyl transferase has revealed highly specific biosynthetic functions for each enzyme. To investigate the phenotypic consequences caused by the loss of meromycolic acid modification, a chemical inhibitory approach was applied in which dioctylamine was used to inhibit multiple mycolic acid methyltransferases in a dose-dependent fashion (77). Lipid analysis combined with extensive genetic characterization of mycolic acid modifications indicated that dioctylamine inhibited multiple sites of cyclopropanation and methylation catalyzed by MmaA2, MmaA3, MmaA4, CmaA2, and PcaA (Fig. 1). This inhibition resulted in decreased bacterial viability, pleiotropic alterations in the cell envelope structure, and loss of AF staining (77). M. tuberculosis mutant strains lacking any mycolic acid cyclopropanation were also found to exhibit less AF staining than the wild-type strain (7), confirming the results obtained with dioctylamine. A possible explanation of the phenotypes observed following either chemical or genetic inhibition of the mycolic acid methyl transferases involves dysregulation of membrane fluidity leading to impaired protein localization or cell division as well as altered cell wall permeability.

Non-Mycolic Acid-Containing Components

The PhoPR two-component system plays a crucial role in the physiology and pathogenicity of M. tuberculosis as well as in regulation of global gene expression (78). Disruption of phoPR caused a robust growth attenuation in human and mouse macrophages as well as in infected mice and prevented growth at low magnesium concentrations (6). Genes that were positively regulated by PhoPR include those found in the pks2 and the msl3 gene clusters that encode enzymes required for the biosynthesis of sulfolipids and diacyltrehalose/pentaacyltrehalose, respectively. Consistent with these findings, lipid analysis revealed the absence of all three lipids in the phoP mutant (6, 79). Microscopic inspection of the phoP mutant not only revealed that the cells appeared smaller than the wild-type bacteria but that they had also lost AF staining, suggesting that sulfolipids and/or diacyltrehalose/pentaacyltrehalose could explain the phenotypic traits of the phoP mutant (6).

Although acid-fastness has essentially been attributed to the waxy nature of the cell wall outer membrane, recent studies highlighted the important contribution of other components, especially lipomannan/lipoarabinomannan (LM/LAM), which also participate in the immunomodulation of the host response. Ablation of the branch forming α-1,2-mannosyltransferase (MSMEG_4247) in Mycobacterium smegmatis leads to accumulation of branchless LAM and the complete absence of LM (80) and AF-negative bacteria (8). This strongly suggests that changes in the LM/LAM structures can affect the cell wall integrity and AF staining. However, an equivalent mutant in M. tuberculosis did not show a defect in AF staining, suggesting that the LM and LAM do not exert a significant impact on the AF staining of M. tuberculosis (8).

LIPID ACCUMULATION AND LOSS OF THE AF PROPERTY

Latent TB infection is characterized by the presence of M. tuberculosis bacilli that can persist in a nonreplicating state, known as the dormancy phase, inside lipid-rich foamy macrophages in granuloma (81, 82). Under these environmental conditions, persistence is favored by the storage of intracellular lipid inclusions (ILI) in the bacterial cytoplasm. These structures are essentially composed of triacylglycerols (TAG) resulting from the degradation of lipid bodies contained in foamy macrophages (83–85). ILI are thought to provide a source of carbon and energy prior to metabolic reactivation and replication, prerequisites ultimately leading to active TB (84). They are also found in bacilli derived from sputum of TB patients (86) and have been proposed as biomarkers for nonreplicating persistence, because a strong correlation between ILI and dormancy has been established (87). TAG degradation by M. tuberculosis involves a wide array of lipolytic enzymes, in the form of cell surface-associated/secreted enzymes for lipid body degradation (88–90) or as intracellular enzymes for ILI degradation (91, 92). Although the origin of the lipid accumulation within ILI has begun to be elucidated (83), identification of the enzymes involved in the transfer of lipids from lipid bodies to ILI remains elusive (88). One possible enzyme is LipY (Rv3097c), a specific TAG hydrolase that plays a major role in the degradation of TAG-containing ILI under growth conditions mimicking dormancy (91, 92). It has been demonstrated that a lipY deletion mutation lost the capacity to utilize stored TAG and to escape dormancy (93).

Because most in vitro dormancy models use single stress factors and fail to generate a truly dormant population, a novel multi-stress model has recently been developed by applying the combined stresses of low oxygen (5%), high CO₂ (10%), low nutrients (10% Dubos medium), and acidic pH (5.0), thereby mimicking conditions encountered in the host (94). Under these conditions, M. tuberculosis stopped replicating, accumulated TAG and wax ester, acquired phenotypic antibiotic resistance, and lost acid-fastness. Dual staining of M. tuberculosis with the combination of auramine O and Nile red has been used to reveal AF staining properties and neutral lipid accumulation in the same cell (94) (Fig. 4A). When synchronous cultures of M. tuberculosis were subjected to the multi-stress conditions for increasing periods of time, a steady decrease in auramine O-stained, green-fluorescent AF cells with a concomitant increase in Nile red-stained, red-fluorescent, ILI-containing cells was observed (Fig. 4B). After 18 days under multiple stresses, AF-positive cells decreased to about 30% of the population, while Nile red-stained cells with internal, red, spherical bodies (corresponding to ILI) increased from 10% to about 70% (94). This difference in dual-staining properties indicated the occurrence of at least three subpopulations under the multiple-stress condition: a subset of auramine O-positive cells (actively replicating), a second subset that stained with both auramine O and Nile red (presumably transitioning to a nonreplicating state), and a third subset that stained exclusively with Nile red (nonreplicating and dormant). Nile red-positive lipid droplets were found in M. tuberculosis cells from sputum samples, and these lipid-loaded bacteria from human patients were found to be dormant (86, 87).

FIGURE 4.

Loss of AF staining coincides with the accumulation of TAG-containing intracellular lipid inclusions. (A) Dual staining of M. tuberculosis grown under multiple stress conditions, using auramine O for AF-staining (green) and Nile red as a neutral lipid stain (red). Bacilli were observed by confocal laser scanning microscopy. Overlaid images of the dual-stained bacteria are shown. Bar = 4 μm. (B) Quantification of the number of AF-positive and lipid-stain-positive bacilli grown as in (A). Auramine O-stained and Nile red-stained positive cells were counted from multiple scans. (Adapted from Deb et al. PLoS ONE 4(6):e6077 with permission of the publisher.)

The finding that TAG accumulation within ILI of Nile red-positive bacilli correlated with reduced acid-fastness was further confirmed through the use of a tgs1 deletion mutant. The tgs1 gene encodes a TAG synthase that is the dominant contributor to TAG storage when M. tuberculosis is exposed to various single stress factors (95, 96). The tgs1 deletion mutant failed to accumulate TAG when subjected to the multi-stress treatment, and consistently lower proportions of Nile red-positive bacilli and higher percentages of AF-positive bacilli were observed than with the wild-type strain (94). Transcriptomic analyses of bacteria subjected to the multi-stress response revealed the achievement of a dormant state, the induction of stress-responsive genes, and the repression of energy generation, transcription, and translation machineries (94). Interestingly, among these genes, kasB was found to be downregulated, further substantiating the possible link between the loss of the AF property and the shutdown of mycolic acid biosynthesis.

Whether AF-negative bacilli and ILI accumulation occur as bacteria enter dormancy was subsequently addressed in granulomas using a biomimetic in vitro model of human TB granuloma (84). In a multi-stress model, granuloma sections and M. tuberculosis cells were subjected to the dual auramine O and Nile red staining. At day 0, M. tuberculosis exhibited few Nile red-stained but abundant auramine O-stained positive cells. In contrast, M. tuberculosis-infected granulomas contained a higher proportion of Nile red-positive cells at day 8 than at day 0. In addition, around 10% of the bacteria from the day 8 granuloma samples displayed tolerance to rifampicin compared to less than 1% at day 0. In this human TB granuloma model, M. tuberculosis presents features of dormant mycobacteria as judged by the (i) loss of acid-fastness, (ii) accumulation of TAG-containing ILI, and (iii) induction of drug tolerance.

Neutralizing tumor necrosis factor alpha (TNFα) signaling results in the disruption of the granuloma structure in vivo, allowing the bacilli to escape the granuloma and ultimately leading to the induction of active TB. To investigate whether M. tuberculosis within granulomas can emerge from dormancy following anti-TNF treatment, resuscitation of the bacilli was monitored by comparing the characteristic dormancy phenotypes, including the auramine O/Nile red-staining profile and tolerance to rifampicin after treatment with anti-TNF antibodies. The vast majority of the cells from granulomas treated with anti-TNF were AF-positive and failed to accumulate large amounts of ILI and also exhibited significantly less rifampicin tolerance than mycobacteria from granulomas treated with a control antibody (84). Under these conditions, the tgs1 mutant accumulated significantly fewer lipids in the form of ILI in the granulomas than did the control strain, and it was compromised in its ability to enter into a dormant state. Conversely, staining of the lipY-disrupted mutant from granulomas revealed a compromised ability to resuscitate and escape dormancy upon immunosuppression with anti-TNF treatment (84). Overall, these findings point to a critical role of the tgs1/lipY expression profile in influencing the lipid accumulation/consumption in M. tuberculosis, allowing the bacilli either to enter into dormancy with a loss of acid-fastness or promoting resuscitation and the escape from dormancy.

AF-NEGATIVE M. TUBERCULOSIS AND CELL WALL ALTERATIONS

The AF-negative phenotype of M. tuberculosis bacteria in established experimental animal infections (12) strongly suggests that cell wall changes are occurring. Consistent with these alterations, the cell wall of M. tuberculosis thickens when grown in vitro under hypoxic conditions, as revealed by transmission electron microscopy (97). These findings emphasize the need to understand the physical and spatial organization of the cell wall and the cell wall changes that occur during in vivo growth. Although the primary structure of the major cell wall components is fairly well established, details such as the degree of coverage of the peptidoglycan (PG) layer by covalently attached mycolic acids in the outer membrane, as well as the spatial organization of the components occurring in in vivo-grown bacilli, remain elusive. Thus, with the aim of addressing the spatial properties of the mycobacterial cell wall and to begin examining the differences between mycobacteria grown in cultures and in animals, the cell wall characteristics of M. tuberculosis grown in vitro were compared with those of Mycobacterium leprae grown in armadillos (98). The cell wall of M. leprae contained significantly more mycolic acids attached to PG than did the cell wall of in vitro-grown M. tuberculosis (mycolate:PG ratios of 21:10 versus 16:10, respectively). The greater coverage of M. leprae PG by mycolic acids may render this bacterium less permeable overall. However, whether similar changes may occur with in vivo-grown M. tuberculosis remains to be determined experimentally.

In light of these findings, there was the surprising observation that cortisone-forced reactivation of M. tuberculosis in infected guinea pigs, which had been previously treated with chemotherapy (rifampicin, pyrazinamide, and TMC207), revealed very weak AF staining despite the ability to cultivate very high numbers of bacilli from the lungs of the reactivating animals (99). Under these conditions, AF bacilli were sparse and difficult to see in the lung sections, further supporting the concept that despite drug treatment, the bacteria undergo physiological adaptation such as cell wall modification. By analogy with the kasB deletion mutant, which persisted indefinitely in mice and was characterized by shortened mycolic acids, loss of AF staining, and the inability to cause disease in mice (11), it appears reasonable to speculate that M. tuberculosis bacilli persisting in this model of chemotherapy may share similar properties with the kasB mutant, which could explain the basis for their poor AF staining in vivo.

LOSS OF ACID-FASTNESS AND PERSISTENCE

As already mentioned, ZN-negative cells correspond to M. tuberculosis bacilli in a dormant state displaying distinct cell wall alterations (12). In particular, it was demonstrated that the classical, cell wall composition-dependent staining with either ZN or AR was lost during persistent infection in mice. In contrast, detection of M. tuberculosis by cell wall composition-independent staining using a polyclonal, anti-M. bovis BCG serum was maintained during persistent infection (12). Because of its polyclonal nature, the antiserum recognizes multiple epitopes in the mycobacterial cell wall, and this recognition is independent of the spatial arrangement of cell wall components, thus explaining why the antiserum generated positive results at all times postinfection, even in tissue sections that were negative for ZN staining. These observations were further corroborated by analyzing histopathological lung sections from patients with acute TB, reactivated TB, or persistent latent TB. Whereas M. tuberculosis in tissue sections from the patients with either acute or reactivated TB was positive for both ZN staining and for immunohistochemistry using the anti-M. bovis BCG serum, bacilli in tissue sections from the patients with latent TB were positive for staining with the antiserum but remained ZN-negative (12). Therefore, it can be inferred that in both latent, experimental TB and in patients with TB, loss of cell wall composition-dependent ZN staining represents a specific attribute of dormant bacilli that can best be explained by mycobacterial cell wall alterations abolishing ZN staining. Alternatively, reorganization of the cell wall may prevent entry of the ZN stain into the bacilli.

A similar dual-staining approach has also been successfully used to study and compare the multiple phenotypic subpopulations in M. tuberculosis cultures and in lung sections of M. tuberculosis-infected mice and guinea pigs (100). This experimental protocol included the combination of fluorescent AF staining and AR that targeted the mycolic acid-containing cell wall and an immunofluorescence assay that targeted bacterial proteins using an anti-M. tuberculosis whole cell lysate, polyclonal antisera. Two phenotypically different subpopulations were found in stationary cultures, whereas three subpopulations were observed in hypoxic cultures and in lung sections. Bacilli were either exclusively AF-positive, exclusively immunofluorescent, or AF-positive and immunofluorescent. The finding of a subpopulation of AF-negative bacilli, corresponding to dormant M. tuberculosis, is consistent with earlier work (12). By applying both staining methods simultaneously, it now becomes possible to detect AF-positive and AF-negative bacteria in the same microenvironments in vitro and in vivo (100). The discovery of heterogeneous phenotypes of M. tuberculosis in the same biological samples reveals new challenges, prompting future studies to investigate the metabolic changes of the bacilli in these microenvironments using genomic, proteomic, metabolomic, and lipidomic approaches.

CONCLUSION AND PERSPECTIVES

Because in vivo growth has been shown to induce dormancy in substantial subpopulations of M. tuberculosis, it is very likely that the ZN-negativity of dormant mycobacteria leads to an underestimation of bacterial burden, which has important consequences for diagnosis of TB as well as for clinical epidemiology. From the dual-staining approaches, combining both ZN and polyclonal serum staining, it seems reasonable to assert that a ZN-negative but antibody-positive specimen points toward a dormant infection rather than the absence of infection. Furthermore, ZN-negative granulomatous pathologies of unknown etiology may result from persistent mycobacteria. Therefore, the current AF-staining methods are not highly reliable for the diagnosis of TB, and further staining improvements are needed to better detect AF-negative cases. Furthermore, this article emphasizes the critical role of mycolic acids in the AF property of M. tuberculosis. If loss of acid-fastness correlates with a reduction in mycolic acid chain length, perhaps as a direct consequence of KasB phosphorylation in persistent bacteria, then a more reliable staining procedure that is not dependent on mycolic acid chain length may be warranted. Therefore, more precise knowledge of the dormant state of M. tuberculosis may not only help to improve the detection of the latent forms of the bacilli but would also have important implications for chemotherapy and vaccine developments. This may now be possible thanks to the recent development of multi-stress in vitro granuloma models that induce M. tuberculosis to enter a dormant-like state.