THE CHEMISTRY OF THE NATIVE CONSTITUENTS OF THE ACETONE-SOLUBLE FAT OF MYCOBACTERIUM

In the course of their pioneering studies on the chemical composition of the lipides of Mycobacterium tuberculosis, Anderson and his coworkers developed a general extraction procedure (1). According to their scheme, the cultures are exhaustively extracted with a mixture of ethanol-ether (1: 1) and then with chloroform. By further solvent fractionation the ethanol-ether-soluble lipides are then separated into the “acetone-soluble fat” and the acetone-insoluble “crude phosphatide,” while the chloroform extracts are divided into the methanol-soluble “soft wax” and the methanolinsoluble “purified wax.” None of these fractions, however, is chemically homogeneous. The discovery of the cord factor by Bloch (2) stimulated new interest in the chemistry of these fractions and led to a search for better methods of purification. Progress toward the isolation of the pure native lipides was made by the introduction of chromatography (3-6) and infrared spectroscopy (5, 7). The application of these methods resulted in the isolation and subsequent elucidation of the chemical structure of the toxic cord factor (8-11) and of several other native constituents of Anderson’s purified wax (12, 13). The present investigation extends our previous work on the composition of the chloroform extracts (12) to include the acetone-soluble fat fraction of the methanol-ether extracts. In contrast to these previous studies, the present paper and Paper II, however, deal exclusively with fractions obtained from the virulent human strain BrBvannes. This qualification seems indicated since Anderson, who in his careful and comprehensive work has laid the ground for all subsequent studies in this field, already emphasized the dependence of the lipide composition upon the nature of the

In the course of their pioneering studies on the chemical composition of the lipides of Mycobacterium tuberculosis, Anderson and his coworkers developed a general extraction procedure (1). According to their scheme, the cultures are exhaustively extracted with a mixture of ethanol-ether (1: 1) and then with chloroform. By further solvent fractionation the ethanol-ether-soluble lipides are then separated into the "acetone-soluble fat" and the acetone-insoluble "crude phosphatide," while the chloroform extracts are divided into the methanol-soluble "soft wax" and the methanolinsoluble "purified wax." None of these fractions, however, is chemically homogeneous.
The discovery of the cord factor by Bloch (2) stimulated new interest in the chemistry of these fractions and led to a search for better methods of purification.
Progress toward the isolation of the pure native lipides was made by the introduction of chromatography (3)(4)(5)(6) and infrared spectroscopy (5,7). The application of these methods resulted in the isolation and subsequent elucidation of the chemical structure of the toxic cord factor (8)(9)(10)(11) and of several other native constituents of Anderson's purified wax (12,13).
The present investigation extends our previous work on the composition of the chloroform extracts (12) to include the acetone-soluble fat fraction of the methanol-ether extracts. In contrast to these previous studies, the present paper and Paper II, however, deal exclusively with fractions obtained from the virulent human strain BrBvannes. This qualification seems indicated since Anderson, who in his careful and comprehensive work has laid the ground for all subsequent studies in this field, already emphasized the dependence of the lipide composition upon the nature of the hf. TUBERCULOSIS (BREVANNES). I strains and the cultural conditions (14). These observations were confirmed in later investigations, notably by Lederer and coworkers (15), by Kubica et al. (16), and in our studies on the native constituents of the Wax C fraction (12).
With use of the strain Brevannes and by applying chromatographic methods, Aebi et al. initiated the chemical study of the native constituents of this fraction (6). The present two reports give additional information obtained mainly by extensive infrared analysis of the chromatographically purified components.
The first paper deals with the fractionation procedures and the characterization of some glyceride and phosphoglycolipide fractions, while the subsequent report describes a new crystalline naphthoquinone related to vitamin Kz.

Isolation of Acetone-Soluble Fat
Cultural and extraction procedures have been described in a previous paper (8). 3 to 4 week-old surface cultures of a virulent human strain of M. tuberculosis (Brevannes) were used. The methanol-ether (2: 1) extracts from various batches were concentrated in vacua; the residue was dark brown and salve-like.
The acetone-soluble fat was prepared by solvent fractionation according to Aebi et al. (6) and had theappearance of a brown viscous oil.
Chromatographic Partitioning Further partitioning of the acetone-soluble fat was effected by chromatography on magnesium silicate-Celite (17) and silicic acid (Mallinckrodt) columns.
The preparation of the magnesium silicate-Celite and the chromatographic procedures have been described (5). Fractions' eluted from the columns were examined by infrared spectrophotometry (Tables  I and II).
From well established correlations the main components can be classified according to their infrared spectra2 as belonging to one of the following groups: I, glycerides; II, phosphoglycolipides; III, naphthoquinone derivatives.
1 Chromatographic fractions are referred to by the roman numerals used for the chromatograms represented in Tables I to III,    Group I. Glycerides That the lipides of M. tuberculosis are particularly rich in mono-, di-, and triglycerides of various long chain fatty acids has been shown in studies on the native constituents of the Wax C fraction of Anderson's purified wax (12) and Dubos' "toxic lipide" (13).   which were characterized by infrared spectroscopy (16) and more recently reported to be glycerides (18).
Fractions If and IId, which were spectroscopically identical with Smith's Compound D (Fig. 1 Furthermore, the infrared spectrum of Compound D is similar to the spectrum of 1-monostearin in the &crystalline form published by Chapman (22), who showed that the profound effect of crystallinity upon the in- 3 The chromatographic adsorption affinity of neutral lipides is predominantly determined by the number of free hydroxyl groups (12), and hence the elution sequence for glycerides is tri-, di-, monoglycerides (13). This order manifests itself by the increasing appearance of an OH band between 3500 and 3350 cm.? in the infrared spectra of subsequent eluates (Fractions Ic to If). 908 M. TUBERCULOSIS (BREVANNES). I frared spectra of glycerides can be used to differentiate between the various polymorphic forms.4 Infrared spectra of glycerides taken from solutions show broad and relatively unspecific absorptions not suitable for the identification of individual compounds, while the spectra of crystalline films (Fig. 1, Spectra 1 to 3), especially of the most stable /3 form, present a striking fine structure which consists of numerous closely spaced sharp bands characteristic for each compound.
Chapman's data also show that the spectra of cryst,alline monoglycerides esterified on carbon 1 differ grossly from those of the corresponding 2 isomers, whereas variations in the chain length of the fatty acid residues within either series of isomers produce only relatively small spectral changes in certain regions.
In view of these correlations it can be concluded that Smith's Compound D is a 1-monoglyceride of either stearic or palmitic acid. Although their glyceride nature was confirmed by hydrolysis, Compounds A and D were not further investigated by chemical methods.
Fractions Id and IIb and IIc, which had an adsorption affinity corresponding to that of a diglyceride, had an infrared spectrum similar to the one of Compound C published by Smith et al. (18). The chemistry of this compound, which represents 20 to 30 per cent of the acetone-soluble fat, was investigated in detail.
It will be referred to as Smith's Compound C. Isolation of Smith's Compound C and a-Monoglyceride of Mycolic Acid-For further purification the fractions containing Compound C (Id and IIb and 11~) were extracted exhaustively with boiling methanol.
Upon being cooled, a white crystalline precipitate separated which was filtered on Whatman paper No. 41-H and dried in vucuo. The infrared spectrum of the crystalline precipitate ( Fig. 1, Spectrum 1) was identical with that of Smith's Compound C. The methanol-insoluble residue remaining after extraction with boiling methanol was dissolved in ether and precipitated with excess methanol.
After being filtered and dried, it yielded a white powder which was identified by its infrared spectrum and melting point with the a-monoglyceride of mycolic acid previously isolated from Wax C and from Dubos' "toxic lipide" (12,13).
SaponiJication-100 mg. of Compound C were saponified with KOH in moist isopropyl alcohol, as described for cord factor (3  In Fig. 2 the progression bands of the acidic material from alkaline hydrolysis of Compound C (Spectrum A) are compared with those of palmitic acid (Spectrum P), stearic acid (Spectrum S), and an equimolar mixture of the two acids (Spectrum P + S, broken line). An analysis of these spectra shows that Spectrum A contains the progression bands of both palmitic and stearic acid and is identical with Spectrum P + S, indicating that the ether-soluble saponification product represents an equimolar mixture of palmitic and stearic acid. This conclusion is corroborated by the data from titration and ele- The solution was deionized by being filtered through Amberlite MB-3 and taken to dryness in vacua. The resulting colorless viscous liquid was identified as glycerol by infrared spectroscopy and paper chromatography (butanol-acetic acid-water, 25: 6 : 25). Spraying with aniline phthalate gave no spots and development with AgN03 and ethanolic NaOH (26) revealed only one spot with an Rp value identical to that of the glycerol control.
Reduction with LiAM-75 mg. of Compound C were reduced in anhydrous ether with 93 mg. of LiAlH,. 58 mg. (77 per cent) of an ethersoluble neutral reaction product were isolated in the usual way.
It had a melting point of 51-52' and its infrared spectrum had the typical absorptions of long chain paraffinic alcohols.
Acetylation-60 mg. of Compound C were treated with excess acetic anhydride in benzene-pyridine.
The reaction product was not soluble in hot methanol in contrast to the parent compound. Purification was effected by reprecipitation from ether with excess methanol. The product (55 mg.) was a white microcrystalline powder with a melting point of 49-50".
The infrared analysis of the acetyl derivative (Fig. 1, Spectrum 2) showed the disappearance of the free OH band of Compound C, indicating complete acetylation.
Chemical Structure-These results indicate that Smith's Compound C is a mixed diglyceride of stearic and palmitic acids corresponding to the formula C&YH,~O~. Further evidence was provided by elementary analysis of Compound C and its acetyl derivative. Of all the lipide fractions of M. tuberculosis, the purification of the phosphatides has proved to be most difficult.
Extensive studies in the past, summarized by Asselineau and Lederer (15), have shown that they are of a complex nature, and quite different from the phosphatides commonly found in animals and plants.
Hydrolytic degradation always liberated a considerable proportion of carbohydrates, nobably inositol, and some amino acids. It seems more appropriate, therefore, to use the term "phosphoglycolipides" in describing these fractions. Lederer and coworkers have succeeded by chromatography in partitioning the crude phosphoglycolipides into fractions of different composition (29,30). The homogeneity and structure of most of these fractions, however, have not been rigidly established.
The presence of phosphoglycolipides in some of the fractions obtained after chromatography of the acetone-soluble fat was detected by infrared spectroscopy and phosphorus assay (31). The phosphorus-containing material was concentrated in Fractions IIa and IIe.
Further purification was achieved by precipitation with acetone from ether or benzene and by rechromatography of the acetone-insoluble precipitates. The phosphorus content of the resulting fractions varied between 1 and 3 per cent. The details of these procedures and analytical data on the products will be reported later.
However, since this paper is concerned with a characterization by fractionation and infrared spectroscopy of the main components of the acetone-soluble fat, a short discussion of the infrared spectra of the phosphoglycolipides is included here. Infrared Xpectru-The infrared spectra of non-bacterial phospholipides have been studied by a number of investigators (32)(33)(34), who showed that the different classes of the lower molecular weight phospholipides give characteristic absorption spectra useful for identification (32). A characteristic feature of phospholipide spectra between 1235 and 1200 and 1100 and 1000 cm.-* is the presence of at least three strong bands, probably associated with covalent phosphate.
As evident from a comparison in Table III, the bands of a representative phosphoglycolipide sample agree well in this region with the corresponding phosphate absorptions of chemically defined phospholipides.
However, apart from this similarity with respect to the phosphate absorptions, the phosphoglycolipides from M. tuberculosis, because of their complex nature and high carbohydrate content, give usually less clearly defined infrared spectra than the known phospholipides of low molecular weight. Spectrum 1, shown in Fig. 3, is representative for a large number of partially purified phosphoglycolipide fractions prepared M. TUBERCULOSIS (BREVANNES). I from the acetone-soluble fat. Similar spectra were obtained with Yamamura's cavity-inducing proteolipide (35) (Fig. 3, Spectrum 2) and Negre's antigene methylique (36) (Fig. 3, Spectrum 3). As can be seen, the bands of the covalent phosphate group and the C-O stretching absorptions of the carbohydrate part overlap between 1200 and 1000 cm.-', causing strong, broad, and only partially resolved absorptions in this region.
In addition to the strong phosphate and C-O absorptions between 1200 and 1000 cm.-l, the spectra of these preparations show an intense ester C=O stretching band between 1750 and 1700 cm.-l. The shape and position of this band vary, indicating different degrees of hydrogen bond- ing. The ester C----O-C linkages absorb near 1240, thus contributing to the strong covalent phosphate band in this region (32). The spectra of some of these preparations show a pair of rather weak bands near 1650 and 1550 cm.-l characteristic of -CONH-, suggesting the presence of small amounts of peptide components.
The high carbohydrate content (10 to 40 per cent) in all of these samples accounts for the strong broad band at 3350 cm.-' characteristic for bonded OH. Despite the general similarity of the spectra of the various phosphoglycolipide fractions, there are a number of minor but distinct differences in band positions and in number and relative intensities of the maxima.
However, since these fractions are not homogeneous, no useful correlations for more detailed structural interpretations can be made. Yet, in a qualitative way these infrared spectra are sufficiently characteristic to distinguish the phosphoglycolipides from the lower molecular weight lipides obtained during the purification of the crude extracts.  (35), and Pound's wax fraction (38) with adjuvant properties. The degree of homogeneity and the chemical structure of these preparations remain largely unknown.
It was found that the phosphorus-containing glycolipides are not completely removed by the initial precipitation with acetone during the preparation of the acetone-soluble fat. It is a common experience in the purification of highly complex lipide mixtures that solvent fractionation often fails to give complete separations, even of fractions with widely differing solubilities. This is apparently due to strong molecular interactions.
Likewise, many compounds are extracted by solvents in which they have little if any solubility in the purified state.
Biological work with lipides of M. tuberculosis in the past has often been based on the nature of the extracting solvent system for distinguishing the chemical nature of the various biologically active materials from one another.
The above results emphasize again that this is a poor criterion. II. Lower Molecular Weight Lipides-In contrast to these preparations of a relatively complex nature, the native lipides of lower molecular weight have now been well characterized spectroscopically and chemically. Their characteristic, clearly defined infrared spectra and distinct chromatographic absorption affinities facilitate their separation from the more complex components of higher molecular weight.
They seem to be devoid of toxicity or antigenic properties with the exception of cord factor which, from its chemical structure and molecular weight, is to be regarded as an intermediate between the two classes.
A large proportion of the lower molecular lipides are esters of saturated straight and branched chain fatty acids (ranging in size from Cl6 to Css) with various hydroxylated compounds such as glycerol, long chain alcohols (phthiocerol), and aromatic hydroxy ethers (probably similar to Anderson's leprosols (12,39,40)). The distribution of the latter type of compounds appears to be strain-specific, while most of the glycerides (Smith's Compounds A, C, and D) have been isolated from all the strains studied, including H37Rv and BCG as well as a number of atypical and saprophytic variants (41).
In agreement with Smith and his coworkers it was found that mono-, di-, and triglycerides are relatively abundant in the acetone-soluble fat fraction of the methanol-ether extracts. This parallels our previous findings on the composition of Wax C (12,13). The glycerides of the acetone-soluble fat, however, differ from those in Wax C by the predominame of shorter chains (Cl6 and C18) in their constituent fatty acids. This accounts for their preferential extraction with methanol-ether. The presence of some cY-monoglyceride of mycolic acid in the acetone-soluble fat shows that there is some overlapping between the latter fraction and Wax C. The observed abundance of glycerides, already reported by Smith, is in contrast to results obtained by Anderson and coworkers (42), as well as by Aebi et al. (6).6 According to these authors, trehalose serves in place of glycerol in these lipides. We were unable to detect esters of trehalose in the acetone-soluble fat. So far, trehalose-6,6'-dimycolate (cord factor) is the only chemically pure native glycolipide that has been isolated from M. tuberculosis.
The isolation of trehalose in hydrolysates of the acetone-soluble fat reported in the literature suggested the occurrence of structural analogues of cord factor in which trehalose was esterified with shorter fatty acids (11). The present findings fail to substantiate our earlier hypothesis.
It is possible, therefore, that the different results reported in those earlier investigations were due either to the presence of cord factor in these extracts or, as already mentioned, to differences in the production of the bacterial cultures.7 SUMMARY The chemical composition of the acetone-soluble fat of Mycobacterium tuberculosis, strain Brevannes, has been examined by infrared spectroscopy after chromatographic partitioning.
On the basis of their spectral and chemical characteristics the separated native lipide constituents of the acetone-soluble fat fraction were found to belong to one of the following three classes of lipides: I, mono-, di-, and triglycerides; II, complex phosphoglycolipides; and III, 1 ,bnaphthoquinone derivatives. One of the main components of the first group, Smith's Compound C, was identified as a mixed diglyceride of stearic and palmitic acids. The infrared absorptions and solubility properties of the complex phosphoglycolipides were investigated, methods for their separation from the phosphorusfreelipides of lower molecular weight outlined, and their relations to various 6 The influence of the glycerol-containing culture medium has been suggested as a possible explanation for the observed high glyceride content (12). ' The extracts described in this paper were prepared from cultures used for the production of cord factor. In contrast to the authors quoted who extracted with ethanol-ether (l:l), we used methanol-ether (2:1), in order to avoid in this first step the extraction of cord factor which is subsequently extracted with chloroform. Since we found that cord factor has a certain solubility in ethanol-ether (1: I), it is likely that at least part of the trehalose found by those previous investigators was derived from cord factor. On the other hand, strain differences could hardly account for the differences in glyceride content, since one of the strains used by Smith et al. (H37Rv) was identical with the strain employed by Anderson, and ours (Brevannes) with the one used by Aebi.
M. TUBERCULOSIS (BREVANNES). I biologically active preparations discussed. The naphthoquinone derivatives are the subject of a subsequent paper. BIBLIOGRAPHY