Eur. J . Biochem. 189, 167-17. (1990) 0 FEBS 1990

Structural and immunological properties of the phenolic glycolipids from Mycobacterium gastri and Mycobacterium kansasii Martine GILLERON, Anne VENISSE, Jean-Jacques FOURNIE, Michel RIVIERE, Marie-Ange DUPONT, Nicole GAS and Germain PUZO Centre de Recherche de Biochimie et de Gttnttique Cellulaires du Centre National de la Recherche Scientifique, Toulouse, France

(Received October 10, 1989) - EJB 89 1221

Mycobacterial species-specific antigens belong to the three following classes : phenolic glycolipids (Phe GI), acyltrehalose-containing lipooligosaccharides and polar glycopeptidolipids. These antigens have been chemically defined and alkali-labile epitopes were found to characterize the lipooligosaccharide antigen type.

In the present study the major Mycobacterium kansasii phenolic glycolipid epitope namely Phe G1 K-I was delineated as the distal monoacetylated disaccharidic residue : 2,6-dideoxy-4-0-methyl-ct-~-arabino-hexopyrano- syl-(1 + 3)-2-0-methyl-4-0-acetyl-~-~-fucopyranose. This acetoxy group is required for K-I epitope recognition demonstrating that alkali-labile epitopes also occur in the phenolic glycolipid antigen class. Using immunoelectron microscopy, the Phe GI K-1 epitope was localized around the electron-transparent layer on the M . kansasii cell- wall surface. Furthermore, two new phenolic glycolipids namely Phe G1 K-111 and Phe GI K-IV were discovered in minute amounts. They were purified and characterized by their retention time in direct-phase column HPLC. These molecules are also M. kansasii antigens, whose epitopes differ from that of Phe G1 K-I. The complete family of phenolic glycolipids Phe G1 K-I, K-11, K-111 and K-IV was found in both rough and smooth variants of both M. kansasii and Mycobacterium gastri species.

Mycobacterial cell walls are endowed with many im- munogenic glycolipids. Some of them are common to most mycobacterial species and two are structurally similar to the gram-negative bacterial lipopolysaccharides, arabinogalac- tan-mycolic acid complex and. lipoarabinomannan [l]. Other cell-wall glycolipids were found to be species-specific and even subspecies-specific antigens [2, 31. They belong to the three following classes : phenolic glycolipids (Phe GI), acyl- trehalose-containing lipooligosaccharides and polar glyco- peptidolipids.

The renewal of interest in the phenolic glycolipid class mainly arose from the discovery in Mycobacterium leprae of a phenolic glycolipid, PGL-I, with serological properties [4]. It was the only species-specific M . leprae antigen and its epitope is now successfully used in ELISA for screening lepromatous leprosy [5]. Moreover, it has been suggested that PGL-I carbohydrate epitope regulates the activation of lepromatous-leprosy-specific T-suppressor cells [6]. This as- sumption is controversial today since it has been shown that PGL-1 induces a general mitogenic suppression unrelated to

Correspondence to G. Puzo, Centre de Recherche de Biochimie et de Gttntttique Cellulaircs du CNRS, 118 route de Narbonne, F-31062 Toulouse Cedex, France

Abbreviations. PGL, major phenolic glycolipid from Mycohac- teriurn leprae; Phe GI K-I, major phenolic glycolipid from Mycohac- terium kansasii; Phe G1 K-11, Phe GI K-111, Phe GI K-IV, quantitat- ively minor phenolic glycolipids from M . kansasii; Phe GI G-I, major phenolic glycolipid from Mycobacteriurn gastri; Phe GI (3-11, Phe GI G-111, Phe G1 G-IV, quantitatively minor phenolic glycolipids from M . gastri; Phe GI B, mycoside B, major phenolic glycolipid from Mycobacteriurn bovis BCG; Phe GI M, mycoside G, major phenolic glycolipid from Mycobacteriurn marinurn; Fuep, fucopyranosyl; Rhap, rhamnopyranosyl; BSA, bovine serum albumin.

~~

the leprosy type [7, 81. Finally, this phenolic glycolipid could play a protective role in the intramacrophagic survival of M . leprae [9] in agreement with the localization of the phenolic glycolipid molecule in a capsule surrounding the bacillus [lo].

Prior to the discovery of this phenolic glycolipid, related glycolipids were described in M . kansasii (mycoside A) [l 1 , 121, Mycobacteriulri bovis BCG (mycoside B) [13, 141 and Mycobacterium marinurn (mycoside G) [15]. More recently, the exact structure of the M . kansasii phenolic glycolipid, renamed Phe GI K-I, was established [16- 181 and further- more, phenolic glycolipid was also characterized in Myco- bacterium tuberculosis strain Canetti [19].

The carbohydrate structure of the M . kansasii major phe- nolic glycolipid (PheGI K-I) was determined as a mono- acetylated tetrasaccharide containing at the distal end a mono- saccharide uniquc in nature. Its structure was determined as 2,6-dideoxy-4-O-methyl-a-~-arabino-hexopyranose which was further supported by the synthesis [20] of its methyl glycoside derivative. Despite the unique structure of Phe G1 K-I, this component was also identified by serological and analytical techniques in M . gastri and was called Phe G1 G-I [21].

In the present study, the Phe G1 K-I epitope was delineated and further localized by immunoelectron microscopy of the mycobacterial cell wall. Besides Phe G1 K-I, other glycolipids in smaller amounts were characterized in M . kansasii, the structure of one of them, Phe GI K-11, has already been deter- mined [22]. Two unknown phenolic glycolipids, namely Phe GI K-I11 and Phe GI K-IV, were purified from M . kansasii cultures. Their occurrence in M . gastri was checked by sero- logical and chromatographic procedures in order to dis- tinguish between M . kansasii and M . gastri species on the basis of specific glycolipid epitopes.

168

EXPERIMENTAL PROCEDURES Culture conditions

M . kansasii strains ATCC 12478 and strains S890175 and S890370 from the Pasteur Institute, Paris and M . gastri strains W 471, HB 4362 and HB 4389 were grown as previously described [16]. The rough or smooth characteristics of the culture were defined on Jensen medium, except for the M . kansasii S890175 and S890370 strains which were examined on 7H10 agar.

Extraction of glycolipids From a four-week-old (1.5 1) culture of each strain, ap-

proximately 2 g lipid extract were obtained [16]. This extract was flash-chromatographed by two successive runs on a silicic acid column (KG60, 230-400 mesh, Merck, Darmstadt, FRG). The first run used petroleum ether/CHCl, ( l : l , by vol.), then pure CHC1, followed by 1% CH30H in CHC1, to release the PheG1 K-I (55 mg). The 8% CH30H in CHCl, fraction (65 mg) contained the minor phenolic glycolipids (Phe G1 K-11, Phe GI K-111, Phe G1 K-IV together with con- taminating Phe G1 K-I). This fraction was re- chromatographed by a second flash chromatograph and eluted by 1% CH30H in CHC13 to remove PheG1 K-1 (15 mg). The 2% CH30H in CHC13 eluate (10 mg) contained Phe G1 K-I1 with contaminating Phe G1 K-I. Phe GI K-111 and PheG1 K-IV appeared in the 3% and 4% CH,OH in CHCl,, still contaminated by PheG1 K-I1 (15 mg). Separation was checked by TLC (CHCl3/CH30H 95: 5) and sulfuric anthron spray to reveal the sugars.

For final purification, each fraction was chromatographed by HPLC equipped with a direct-phase column (5 pm Spherisorb, 4.6 mm x 250 mm) eluted at a flow rate of 1 ml/ min using an elution solvent gradient in the range 0 ~ 20% CH30H in CHC1,. Phenolic glycolipids were detected at 275 nm. Approximately 50 mg Phe G1 K-I, 1.5 mg Phe G1 K- 11, 1 mg Phe G1 K-111 and 1 mg Phe G1 K-IV were obtained in this way from each of the strains studied.

Analytical methods HPLC was conducted on a Gilson (Gilson, France S.A.,

Villiers-le-bel, France) apparatus composed of two model 303 pumps, an 802 manometric module and an 81 1 dynamic mixer. The elution gradient and the data acquisition were controlled with Gilson GME 714 Software and a Gilson model 621 data module interface, connected to an OPAT Normerel computer. Elution of the products was monitored with a Gilson model ultraviolet 11 6 variable wavelength detector.

TLC was conducted on commercial silica gel plates (DC Alurolle, Kieselgel 60 pF254 Merck, Darmstadt, FRG).

Preparation ojantisera Antibodies which reacted against pure Phe G1 K-I or

whole mycobacteria were raised in rabbits by repeated intra- dermal injections of 1 -ml emulsions prepared with Freund's incomplete adjuvant, 150 mM NaCl, 10 mM K2HP04/ KH2P04 (NaCl/P,) and pure Phe Gl K-I in ethanol or whole mycobacterial suspension respectively (5 : 4: I , by vol.) as pre- viously described [17].

Direct ELISA procedures 50 p1 glycolipid antigen stock solutions (5 pgiml in etha-

nol) were deposited in the wells of a polystyrene microtiter

plate (Nunc Imrnuno Plate I, Roskilde, Denmark) and im- mediately dried. These wells were then saturated with 200 pl 30% powdered skimmed milk with 0.1 % Tween 20 in NaCI/ Pi pH 7.2 for 2 h at 37°C. The plate was rinsed three times with 1% powdered skimmed milk in NaC1/Pi and 100 pl/ well of sera diluted in 0.4% powdered skimmed milk were deposited and left for 2 h at 37°C. After five rinses, 100- pl horseradish-peroxidase-linked, donkey anti-[rabbit (Fab)2] antibody (Amersham France, Les Ulis. France), diluted ac- cording to the supplier's instructions, were added to the wells and left for 2 h at 37°C. After five washings, 100 p1 2,2'- azinobis(3-ethylbenzothiazoline-6-sulfonic acid) substrate were added and absorbance at 405 nm was read 30 min later. Each value was measured in triplicate.

Inhibition ELISA

The inhibition of the binding reaction between 125 ng solid-phase Phe G1 K-I and specific rabbit anti-(Phe G1 K-I) antibodies diluted 1 : 100 was performed as follows. The inhibi- tory glycolipids were suspended at various concentrations in NaCl/Pi/O.l % sodium deoxycholate. 90 pl of these suspen- sions were mixed with 10 ~1 rabbit anti-(Phe G1 K-I) antiserum [l : 10 in NaCl/Pi/l % bovine serum albumin (BSA)] deposited in Phe G1 K-I-coated wells that had previously been blocked by 200 pl NaC1/Pi/5% BSA and were left for 2 h at 37°C before rinsing. The coloured reaction was obtained using the goat anti-[rabbit (FJ2] antibody conjugated with j-D- gdactosidase (Amersham France, Les Ulis, France) and o-nitrophenyl-P-D-galactoside substrate (Merck, Darmstadt, FRG), the absorbance being recorded 30 min later at 405 nm.

A control for no inhibition was performed using serum diluted 1 : 100 in NaCl/P,/O.l% sodium deoxycholate and a blank control was obtained by the development of the com- plete ELISA reaction in wells not coated with PheG1 K-I. Each value was measured in triplicate.

lmmunogold labelling and electron microscopy

About 50 p1 of a mycobacterial clump was suspended in an Eppendorf tube containing 1 ml NaCl/Pi/5% BSA/O.l YO sodium deoxycholate, further vortexed and heated for 1 h at 37°C in order to dissociate thoroughly all aggregates. After rinsing by centrifugation, 200 p1 rabbit serum 1 : 10 in NaCI/ Pi/ l% BSA were left for 1 h at 37°C. After five rinses, 100 p1 goat anti-(rabbit IgG) antibodies complexed with colloidal gold (size 10 nm, Janssen Life Science, Beerse) diluted 1 : 20 in Tris/l% BSA were added to the cell pellet and left for 1 h at 37 "C. After three rinses in buffer, the sample was concentrated to a 10-pl vol. and 2-pl aliquots were deposited on carbon- plated copper grids. These grids were viewed directly in a Jeol 1200EX electron microscope at 80 kV. In order to test the labelling specificity, some mycobacterial clumps were incu- bated with non-immune rabbit serum.

RESULTS

Epitope delineation o f the Phe GI K-I antigen

Rabbit polyclonal antibodies raised against Phe G1 K-I were obtained as previously described [23]. Their molecular specificity against Phe G1 K-I was determined by direct ELISA using phenolic glycolipids of known structure isolated from

169

- s. I

x > c ._ .- c

50 m C C 0

.-

.-

5 cc

K-l G-l B M PGL-I

Fig. 1. Reactivity in ELZSA of a polyclonal rabbit anti-(Phe GI K- I) antibodies against several phenolic glycolipids. ( W ) Rabbit anti- (Phe GI K-I) antiserum (1 : S O ) . (0) Rabbit pre-immune serum. K-I and G-I, Phe GI K-I and Phe GI G-I (major phenolic glycolipids from M . kansasii and M . gastri respectively), 2,6-dideoxy-4-0-Me-c-~- arabino-Hexp-( 1 +3)-2-0-Me-4-0-Ac-r-~-Fucp-(l+3)-2-0-Me-a- L-Rhap-(I +3)-2,4-di-O-Me-a-c-Rhap-(l +aglycon; B, Phe GI B (major phenolic glycolipid of M . bovis BCG), 2-O-Me-a-~-Rhap-(l +agly- con; M, PheG1 M (major phenolic glycolipid of M . marinum), 3-0- Me-a-L-Rhap-(I + aglycon; PGL-I (major phenolic glycolipid of M . leprae), 3,6-di-O-Me-P-~-Glcp-(I +4)-2,3-di-O-Me-u-~- Rhap-(I + 2)-3-0-Me-a-~-Rhap-(l + aglycon

mycobacterial species other than M . kansasii (Fig. 1). These phenolic glycolipids are mycoside B (Phe G1 B) [3, 24 - 271, mycoside G (PheG1 M) [IS], and PGL-I [4]. They share with Phe G1 K-I a common phenolphthiocerol dimycocerosate lipid core but differ in their carbohydrate structure (shown in the legend of Fig. 1). This figure shows that anti-(PheG1 K- I) antibodies do not react in direct ELISA with phenolic glycolipids other than Phe G1 K-I or Phe G1 G-I. These data clearly imply that the common phenolphthiocerol dimycocerosate core does not participate in the Phe G1 K-I epitope, but strongly support the relationship between the molecular specificity of the anti-(Phe G1 K-I antibodies and the unique Phe G1 K-I tetrasaccharide structure.

With the goal of delineating this epitope, native PheG1 K-I was degraded by controlled hydrolysis, as previously described [16], into three different products identified as diglycosylated, triglycosylated phenolic glycolipid and deacet- ylated tetrasaccharide phenolic glycolipid. Their affinities for the anti-(Phe G1 K-I) antibodies were measured by ELISA inhibition technique. As shown in Fig. 2 , none of these Phe G1 K-I products inhibited the Phe G1 K-I binding by the anti-(Phe G1 K-I) antibodies. Thus, the removal of the acetoxy group located on C4 of the fucosyl residue abolishes the reac- tivity of the Phe GI K-I epitope. However, the monoacetylated triglycosylated phenolic glycolipid (a natural phenolic glycolipid from M . kansasii which differs from Phe G1 K-I in the absence of the distal Phe G1 K-I monosaccharide residue) partially inhibited the Phe GI-K-I-specific binding with a ten- fold-reduced affinity constant compared to that of Phe G1 K- I ( 1 PM).

m r -- 0 5 10 15 20 25

[Glycolipidl (pg l rn l )

Fig. 2. Inhibitory activity by the following glycolipids of the reaction between specific rabbit anti-(Phe GI K-I) antibodies (I: 100) and 125 ng coated Phe GI K-Z. Inhibiting glycolipids: (0) Phe GI K-I; ( A ) diglycosyl-Phe G1 K-I; (0) deacetylated triglycosyl-Phe GI K-I; (M) deacetylated Phe GI K-I; (0) monoacetylated triglycosyl-Phe GI K-I

Taken together, these data are in agreement with previous work in which the involvement of the distal monosaccharide in the Phe G1 K-I epitope was demonstrated [20]. The affinity of the synthetic methyl glycoside derivative for anti-(Phe GI K-I) antibodies was found to be 1 mM, as measured by a similar inhibition procedure. So, the Phe G1 K-I epitope can be delineated as the distal monoacetylated disaccharide part, i.e. 2,6-dideoxy-4-0-methyl-u-~-arabino-hexopyranosyl- (1 + 3)-2-0-methyl-4-0-acetyl-u-~-fucopyranose.

Phe G1 K-I localization on the M. kansasii cell wull

Phe G1 K-I was localized by immunogold labelling and electron microscopy. Freshly grown bacteria were suspended in NaCl/Pi containing BSA and deoxycholate, as mentioned in Experimental Procedures, to dissociate aggregates. The bac- teria were incubated in the presence of the highly specific anti-(Phe G1 K-I) antibodies then labelled with anti-(rabbit Ig) labelled with colloidal gold prior to electron microscopy. The typical appearance of the Phe GI-K-I-labelled M . kansasii cells is presented in Fig. 3 A. This figure shows that the Phe G1 K- I molecules surround the well known electron-transparent layer. The specificity of the anti-(Phe G1 K-I) antibody binding is supported by the quasi-absence of significant immunogold labelling when M . kansasii is incubated with pre-immune rab- bit serum (Fig. 3B). However, we can distinguish the non- uniform distribution of the Phe G1 K-I on the cell surface by a zebra pattern which occurs on most of the observed bacilli.

Purification and characterization of unknown phenolic glycolipids from M. kansasii

From a CH30H/CHC13 extract of M . kansasii, the crude acetone-soluble fraction contained the phenolic glycolipids among other lipid molecular species. From this mixture, only the Phe G1 K-I and no other quantitatively minor glycolipids, such as the Phe GI K-11, were clearly identifiable on silicic acid TLC. This difficulty in detecting minor glycolipids is amplified by the overlapping of lipid components on TLC. So, in order to detect more polar unknown glycolipids the acetone-soluble mixture was fractionated by two successive steps of silica-gel flash chromatography. In the first, the column was irrigated by petroleum ether and CHC13 to eliminate most of the apolar

170

Fig. 3. Gold-immunolabelled M. kansasii using as u primary untibody either the rubhit polyclonal anti-(Phe GI K-I) antiserum ( A ) or a pre- immune rabbit serum (Bj ,,followed by incubation with colloidal-gol~--coniu~ated anti-(rabbit IgG) untibody. Bar, 0.5 pm

compounds with an Rf higher than that of K-I. Then the PheG1 K-I (55 mg) was mainly eluted by 1 % CH30H in CHCI3, while the 8%0 CH30H fraction contained K-I1 con- taminated by K-I and unknown polar glycolipids. This last fraction was then re-chromatographed and two unknown compounds were recovered in both 3% and 4% CH30H in CHCI3 eluates which were pooled. Upon staining with anthron, two well-defined spots of Rf = 0.33 and 0.14 were observed (lower than K-11, Rf = 0.42; K-I, R f = 0.66) whose blue color characterized the glycolipids.

Finally, the fractionation of K-111 and K-IV was ac- complished using HPLC equipped with a 5-pm spherisorb analytical column. The mobile phase consisted of two succes- sive linear gradients of CH30H in CHC13: 0-4% CH30H in 9 min; 4- 15% CH30H in 11 min. Elution of glycolipids K-111 and K-IV was monitored at 275 nm (the absorption wavelength of the phenolic nucleus) and eluted as a single peak (Fig. 5). These were named Phe GI K-111 and K-IV. A slight shoulder on the PheG1 K-I11 peak indicated that this glycolipid was still slightly contaminated. However, Phe GI K- I11 and Phe G1 K-IV yielded homogeneous spots in silicic acid TLC using as eluent CHC13/CH30H (9:l by vol.; Fig. 4). Thus, since only small amounts (1 mg) of K-111 and K-IV were collected compared to K-I (50 mg), these new glycolipids, characterized by their retention times in gradient HPLC (K-111, 12.17 min; K-IV, 14.6 min) were not further purified.

Preliminary structural studies were performed to confirm the phenolic glycolipid nature of K-111 and K-IV. Ultraviolet absorption spectra of both compounds were identical (intense

1 2 3 4 5

Fig. 4. TLC of purifedphenolic glycolipids from M. kansasii. (1) Phe- nolic fraction enriched in minor phenolic glycolipids; (2) Phe G1 K-I, standard; (3) PheG1 K-11, standard; (4) PheGl K-111, purified by HPLC; ( 5 ) PheG1 K-IV, purified by HPLC; plates were developed in CHC13CH30H (9: 1 ) revealed with sulfuric anthron and heated

171

K-l

Solvent

5 i o 15 20 Retention time fminl

aJ u) C 0 Q u)

E L

c 0 V aJ aJ 1

n

Fig. 5. Direct-phase HPLC ofpurifiedphenolic glycolipids K-I, K-II, K-III, K-IV. A 5-pm spherisorb analytical column was eluted by a mixture of CH,OH (solvent B) in CHCl3 (solvent A) using a linear gradient with two slopes

K-l

K-Ill

1 2 3 1 5 1 2 3 1 5 1 2 3 4 5 1 2 3 L 5 1 6

Fig. 6. ELISA reactivity of purified phenolic glycolipids K-I, K-II, K-III and K-IV against serum from a rabbit immunized with the whole M. kansasii cell. (2 - 5 ) ELISA with 1 : 50, 1 : 100, 1 : 250 and 1 : 500 anti-(M. kansasii) antiserum dilutions, respectively, in wells coated with 250 ng glycolipid; (1) background reactions of pre-immune serum binding to 250 ng coated phenolic glycolipids, K-I, K-11, K-I11 and K-IV. (0) Columns 1 and 6 correspond respectively to pre- immune and immune sera without coated antigens

peak at 220 nm and two peaks of lower intensity at 276 nm and 282 nm), characteristic of an aromatic group [16]. More- over, their 'H-NMR spectra showed two doublets centered at 6 = 6.95 ppm and 6 = 7.14 ppm typical of a p-substituted phenolic group [16]. The multiplet centered at 6 = 4.82 ppm, assigned to two methine protons, also characterizes the pres- ence of two ester linkages of the phenolphthiocerol with two fatty acids. These data unambiguously support the phenolic glycolipid structural assignment for K-111 and K-IV com- pounds.

Serological properties and species-specificity analysis of the Phe GI K-111 and Phe G1 K-IV

The presence of the specific anti-(Phe GI K-111) and anti- (Phe GI K-IV) antibodies was demonstrated in the serum from a rabbit immunized with M . kansasii (Fig. 6). Thus, as pre-

Fig. 7.

I .O

yl

-2

6

0.5

0 K - l K-ll K-Ill K-IV

S A reactivity of a polyclonal rabbit anti-(Phe C1 I ) anti- body against the purified phenolic glycolipids from M. kansasii. (.) Rabbit anti-(Phe G1 K-I) antiserum (1 : 50). (0) Rabbit pre-immune serum. Identical values to those of K-11, K-I11 and K-IV were found for G-11. G-I11 and G-IV

viously shown for Phe G1 K-I and Phe G1 K-I1[17,22], the K- 111 and K-IV phenolic glycolipids are M. kansasii cell-wall antigens. The weaker reactivity of K-111 and K-IV antigens compared to K-I suggests that K-111 and K-IV epitopes differ from that of K-I. This assumption is supported by the lack of reactivity of rabbit anti-(Phe G1 K-I) antiserum with K-111 and K-IV phenolic-glycolipid-coated wells (Fig. 7).

Despite the unique structure of Phe GI K-I, this major glycolipid was also found in M . gastri cell walls [21]. Thus, we wondered whether the quantitatively minor K-11, K-111 and K-IV phenolic glycolipid antigens were also present in M. gastri. The binding of these glycolipids was found to be iden- tical by ELISA using rabbit anti-(M. kansasii) and anti-(M. gastri) antisera. Thus, these data indicated the presence of all these glycolipidic epitopes in both mycobacterial species. This statement was supported by the chromatographic analysis of the M. gastri phenolic glycolipidic fraction. The phenolic glycolipids of M . gastri (strain W471) and M. kansasii (strain ATCC 12478) had the same behdviour on silicic acid TLC and

172

1 2 3 4 5 6 1 Fig. 8. TLC crude fractions of phenolic glycolipids of the following strains of M. kansasii and M. gastri. (1 - 3) M . kansasii ATCC 12478 (smooth morphology), S890175 and S890370 (rough morphology). (4) Mixture of the phenolic glycolipids K-I, K-11, K-111 and K-IV. (5-7) M . gastri W471 (smooth morphology), HB 4389 and HB 4362 (rough morphology). The combined CH3OH/CHCl3 (1 - 8%) frac- tions from silicic acid flash chromatography were applied to TLC plates, thus, the spot intensities are not in agreement with the relative true abundance

HPLC supporting the presence of Phe G1 K-11, Phe G1 K-111 and PheG1 K-IV in M. gastri (Fig. 8). These newly found phenolic glycolipids in M . gastri were called Phe G1 G-11, Phe G1 G-I11 and Phe G1 G-IV, respectively.

Recently, Brennan et al. [28] noticed that PheG1 K-I is found whatever the M . kansasii strain investigated, while the acyltrehalose-containing lipooligosaccharides were detected in smooth, but not in rough, M . kansasii strains. In an effort to widen these observations, we investigated the presence of K-11, K-111 and K-IV in rough and smooth strains of M . kansasii and M. gastri cultures. The fraction of crude phenolic glycolipids was obtained and analysed by silicic acid TLC (Fig. 8). The results show the presence of all the different phenolic glycolipids in both M. kansasii and M . gastri rough and smooth phenotypes.

DISCUSSION

It has been demonstrated that the M. leprue major phenolic glycolipid (PGL-I) epitope corresponds to its distal monosac- charide residue, 3,6-di-O-methyl-~-~-gh1copyranose [29, 301.

This statement was mainly based on the binding reactivity of anti-(PGL-I) antibodies to minor phenolic glycolipids lacking one methoxy group either at the distal (PGL-111) or penulti- mate (PGL-11) monosaccharide residue [31].

Our previous work indicated involvement of the distal monosaccharide in the Phe G1 K-I epitope [20]. The present study demonstrates that the complete Phe G1 K-I epitope de- lineation also includes the monoacetylated fucosyl residue since only the latter binds the anti-(Phe G1 K-I) antibodies with a tenfold reduced affinity constant compared with the native Phe G1 K-I. Thus, the Phe G1 K-I epitope is the distal monoacetylated disaccharide residue : 2,6-dideoxy-4-0-meth- yl-a-D-arabino- hexopyranosyl-(l+3)-2- 0-methyl-4- 0- ace- tyl-a-L-fucopyranose. Further support for this stems from the fact that (a) removal of the acetoxy group dramatically de- creases the specific Phe G1 K-I binding by anti-(Phe G1 K-I) antibodies, (b) the lack of anti-(Phe GI K-I) immunoglobulin binding to the Phe G1 K-I1 glycolipid, whose penultimate acetoxylated fucosyl unit is glycosidically linked to a 2,4-di- 0-methyl-a-D-mannopyranosyl residue.

It is currently accepted as dogma that the species-specific and alkali-labile immunogenic glycolipids belong only to the acyltrehalose-containing lipooligosaccharide class [l]. To our knowledge, this is the first report of an alkali-labile phenolic glycolipid antigen. Consequently, the antigenic reactivity of any unknown glycolipids should be determined in the native molecules and not in their deacylated state. Thus, the use of deacylated phenolic glycolipids already proposed by Young and Buchanan [32] to increase phenolic glycolipid coating on ELISA microplates might have led to erroneous results. This procedure can however be applied to the structurally well- known phenolic glycolipids found in M . leprae, M . tubercu- losis strain Canetti, M . bovis BCG and 11.1. marinum, but must be avoided for those found in M . kansasii. The Phe GI K-I acetoxy group plays a key role in the tridimensional structure of the Phe GI K-I tetrasaccharide and its elucidation is clearly the next step in the understanding of the Phe G1 K-I epitope topography.

The Phe GI K-I was localized on the cell-wall surface by immunoelectron microscopy based on the use of the specific anti-(Phe G1 K-I) antibodies and an electron-dense marker. This approach has already been successfully applied [33] and appears to be more performant than immunofluorescence which requires optical microscopy. Both techniques have pre- viously been used for the localization of PGL-I on the M. leprae cell wall [lo, 341. The cell-wall mycobacterial structure is composed of a triple layer of a basal peptidoglycan sur- rounded by an intermediate electron-transparent lipid layer of and an outer dense layer. Recently, Rastogi et al. [35] noticed the presence of an additional polysaccharide outer layer, wrapped around the triple layer.

Our immunoelectron microscopy pictures show that Phe G1 K-I is scattered around the electron-transparent layer. Thus, we can infer that the Phe G1 K-I carbohydrate moiety contributes to the polysaccharide outer layer formation. Re- cently, it was shown by a similar approach that the M . kansasii trehalose-containing lipooligosaccharides are also located on the cell-wall surface [28].

The molecular complexity of the M . kansasii phenolic glycolipid fraction has required the development of specific chromatographic techniques for its fractionation. Using di- rect-phase HPLC with a CH3OH/CHCl3 gradient elution, the phenolic glycolipid fraction was successfully resolved into four phenolic glycolipids, K-I, K-11, K-I11 and K-IV, characterized by their HPLC retention times. Among them, two new phe-

173

nolic glycolipids, namely Phe G1 K-111 and Phe G1 K-IV of unknown structures, bearing epitopes differing from that of Phe G1 K-I, were purified from the M . kansasii cell walls.

The distinction between M . gastri and M . kansasii species remains unclear [36]. Their similarities are supported by the presence of Phe G1 K-I in both species. With the discovery of these new phenolic glycolipid antigens, we wondered whether these molecules were also present in M . gastri. Using serologi- cal and chromatographic techniques, we demonstrated that K-11, K-111 and K-IV phenolic glycolipids are also present in M . gustri species where they are called Phe G1 G-11, Phe G1 G-I11 and Phe G1 G-IV respectively. Thus, based on these phenolic glycolipids, our studies agree with the similarity of both species.

It has been shown [37] that the M. kansasii strains (TMC 1201, 1204, 1214, 1217) contain besides the phenolic glyco- lipid, a second class of antigenic glycolipids named acyl- trehalose-containing lipooligosaccharides. These antigens were only found in smooth and not in rough strains, while the Phe G1 K-I remains present [28]. Rough strains of M . kansasii intravenously injected into mice induced infection whereas the smooth ones were rapidly cleared [38]. Then a correlation between the virulence of the bacteria and the absence of lipooligosaccharides was established [28]. Our data support and broaden the previous Brennan observations showing that both smooth and rough M . kansasii strains contain the com- plete set of phenolic glycolipids K-I, K-11, K-I11 and K-IV. We also demonstrated that M. gastri contains the latter phenolic glycolipids whatever the strain morphology. Thus, from a taxonomic point of view, the characterization by either chromatographic tools (TLC, HPLC) or serological pro- cedures (ELISA) of the PheG1 K-I and the minor phenolic glycolipids unambigously allow the identification of M. kansasii and M. gastri.

We thank Drs Cecile Asselineau and Marie-Antoinette Laneelk for their kind gift of the mycobacterial culture and Dr Veronique Levy- Frebault for providing us with the rough strains of Mycobacterium kansasii.

REFERENCES 1. Brcnnan, P. J. (1988) Microbial lipids (Ratledge, L. & Wilkinson,

S. G., eds) vol. 1, pp. 203 - 298, Academic Press, London. 2. Brennan, P. J. (1984) Microbiology (Leive, L. & Schlessinger,

D., eds) pp. 366-375, American Societes for Microbiology, Washington.

3. Puzo, G. (1989) Crit. Rev. Microbiol., in the press. 4. Hunter, S. W., Fujiwara, T. & Brennan, P. J. (1982) J . Biol. Chem.

5 . Cho, S. N., Yanagihara, D. L., Hunter, S. W., Gelber, R. H. &

6. Mehra, V., Brennan, P. J., Rada, E., Convit, J. & Bloom, B. R.

7. Prasad, H. K., Mishra, S. R. & Nath, I. (1 987) J . Exp. Med. 165,

257, 15 072 - 15 078.

Brennan, P. J. (1983) Infect. Immun. 41, 1077-1083.

(1984) Nature 308, 194- 196.

239 - 244.

8. Fournit, J . J., Adams, E., Mullins, R. J. & Basten, A. (1989)

9. Neil, M. A. & Klebanoff, S. J. (1988) J. Exp. Med. 167, 30-42. 10. Young, D. B., Khanolkar, S. R., Barg, L. L. & Buchanan, T. M.

11. Smith, D. W., Randall, H. M., Gastambide-Odier, M. M. &

12. Gastambide-Odier, M. M. & Villt, C. (1970) Bull. Soc. Chim.

13. Randall, H. M. & Smith, D. W. (1953) J . Opt. Soc. Am. 43,

14. Demarteau-Ginsburg, H. & Lederer, E. (1963) Biochim. Biophys.

15. Villi., C. & Gastambide-Odier, M. M. (1970) Carbohydr. Res. 12,

16. Fournie, J. J., Riviere, M. & Puzo, G. (1987) J . Biol. Chem. 262,

17. Fournie, J . J., Riviere, M., Papa, F. & Puzo, G. (1987) J . Biol.

18. Fournie, J. J., Riviere, M. & Puzo, G. (1987) Eur. J . Biochem.

19. Daffk, M., Lacave, C., Laneelle, M. A. & Lantelle, G. (1987) Eur.

20. Gilleron, M., FourniC, J . J., Pougny, J . R. & Puzo, G. (1988) J .

21. Vercellone, A,, Rivikre, M., Fournik, J. J. & Puzo, G. (1988)

22. Rivitre, M., Fournie, J . J . & Puzo, G. (1987) J . Biol. Chem. 262,

23. Papa, F., Rivikre, M., Fournit, J. J., Puzo, G. & David, H. (1987)

24. Daffk, M., LanCelle, M. A,, Lacave, C. & Laneelle, G. (1988)

25. Chatterjee, D., Bozic, C. M., Knisley, C., Cho, S. N. & Brennan,

26. Vercellone, A. & Puzo, G. (1989) J . Biol. Chem. 264,7447-7454. 27. Vercellone, A,, Rivikre, M. & Puzo, G. (1989) Adv. Mass

28. Belisle, J. T. & Brennan, P. J. (1989) J . Bacteriol. 171, 3465-

29. Fujiwara, T., Hunter, S. W., Cho, S. N., Aspinall, 0. G. &

30. Cho, S. N., Chatterjee, D. & Brennan, P. J. (1986) Am. J . Trop.

31. Hunter, S. W. & Brennan, P. J. (1983) J . Biol. Chem. 258, 7556-

32. Young, D. B. & Buchanan, T. M. (1983) Science 221, 1057-

33. Fuerst, J. A. &Perry, J. W. (1988) J. Bacteriol. 170, 1488-1494. 34. Boddingius, J. & Dijkman, H. P. (1989) J . Histochem. Cytochem.

35. Rastogi, N., Frehel, C. & David, H. L. (1986) Curr. Microbiol.

36. Lind, A. & Ridell, M. (1984) The mycobacteria - a source book, part A (Kubica, G. L. & Wayne, L. G., eds) pp. 67-104, Marcel Dekker Inc., New York.

37. Hunter, S. W., Jardine, I., Yanagihara, D. L. & Brennan, P. J. (1985) Biochemistry 24, 2798 -2805.

38. Collins, F. M. & Cunningham, D. S. (1981) Infect. Immun. 32,

Infect. Immun. 57, 3653 - 3659.

(1984) Infect. Immun. 43, 183-188.

Koevoet, A. L. (1957) Ann. N . Y. Acad. Sci. 69, 145- 157.

Biol. 52, 679 - 693.

1086-1092.

Acta 70, 442-451.

97-107.

3174- 3179.

Chem. 262, 3180-3184.

168, 181 -183.

J . Biochem. 167, 155-160.

Carbohydr. Chem. 7,733-748.

Chem. Phys. Lipids 48, 129 - 134.

14879 - 14884.

J . Clin. Microbiol. 25, 2270 - 2273.

Biochim. Biophys. Acta 958,443 -449.

P. J. (1989) Infect. Immun. 57, 322-330.

Spectrom. Bord. l l B , 1350-1351.

3470.

Brennan, P. J. (1984) Infect. Immun. 43, 245-252.

Med. Hyg. 35, 167- 172.

7562.

1059.

37,455-462.

13,237-242.

61 4 - 624.