Total Synthesis of a Mycolic Acid from Mycobacterium tuberculosis

Abstract In Mycobacterium tuberculosis, mycolic acids and their glycerol, glucose, and trehalose esters (“cord factor”) form the main part of the mycomembrane. Despite their first isolation almost a century ago, full stereochemical evaluation is lacking, as is a scalable synthesis required for accurate immunological, including vaccination, studies. Herein, we report an efficient, convergent, gram‐scale synthesis of four stereo‐isomers of a mycolic acid and its glucose ester. Binding to the antigen presenting protein CD1b and T cell activation studies are used to confirm the antigenicity of the synthetic material. The absolute stereochemistry of the syn‐methoxy methyl moiety in natural material is evaluated by comparing its optical rotation with that of synthetic material.


Introduction
Mycobacterium tuberculosis (Mtb),the causative agent of tuberculosis,isbyfar the most lethal member of the family of Mycobacteriaceae,c ausing around 10 million new infections yearly. [1] Thedifficult diagnosis and treatment of tuberculosis is reflected in high mortality rates (1.5 million in 2018), which makes it the leading cause of death by as ingle infectious agent worldwide. [1] This difficult treatment is in part caused by the presence of the mycomembrane in the cell envelope of Mtb,w hich consists largely of a-alkyl, b-hydroxy long-chain (C70-C90) fatty acids,k nown as mycolic acids. [2,3] These highly lipophilic fatty acids are categorized according to the presence of unsaturations or cyclopropyl groups ("a-mycolic acid"), amethoxy,oraketo function in the main chain. As the number of methylene units between these substituents is variable,mycolic acids are found as an inseparable mixture of homologues.M ycolic acids are found either as the free acids or esterified to the arabinogalactan layer, trehalose,g lucose (glucose monomycolate;GMM) or glycerol.
Theoverall molecular structure of mycolic acids has been elucidated in the late 1960s, [7][8][9] and several members have been prepared by the group of Baird [10][11][12][13] including two methoxy mycolic acid diastereomers (Figure 2, 1c and 1d). Thec omplete elucidation of their stereochemistry,h owever, has so far not been achieved. Moody and co-workers demonstrated that R,R stereochemistry of the a-alkyl, bhydroxy carboxylic acid is crucial for Tc ell recognition of glucose monomycolate. [14] Thes tereochemistry of the amethyl methoxy moiety in methoxy mycolic acids was inferred by Asselineau et al. to be S,S by comparison of molar optical rotations of natural samples with those of reference compounds. [7] Fort his,t he authors applied the principle of optical superposition, originally hypothesized by Va n 'tH off. [15] This hypothesis states that stereocenters that are remotely located do not influence each otherso ptical rotation, which therefore can be added up to provide the overall optical rotation. Thea nalysis was supported by the synthesis work of Baird et al.,b ut it remained uncertain whether this hypothesis was valid for mycolic acids as the optical rotation of the cis-cyclopropyl ring in model systems was immeasurably small, making the determination of its absolute stereochemistry using this strategy impossible.D ue to the lack of other viable analytical techniques,t he stereochemistry of the cyclopropyl ring still remains to be determined.
Methoxy mycolic acids ( Figure 1) appear,w ith some exceptions,t obe present only in pathogenic mycobacteria [3] and are antigenic in serological assays [4] and Tc ell assays. [5,6] Upon binding to the antigen-presenting protein CD1b, mycolic acids and their glucose mycolates activate Tcells. [16,17] Forg lucose monomycolates the type and length of the mycolate does not influence the Tcell stimulatory capacity. [17] Crystallography of the trimolecular complex of CD1b, glucose monomycolate and Tc ell receptor showed that the glucose forms the Tc ell epitope on top of the CD1b protein and stabilizes the complex. [18] No crystallographic data of CD1b-mycolic acid are available,but it is known that the type of functional groups in the chain and its length determine the level of Tc ell activation. [5,6] We hypothesized that the stereochemistry of the cyclopropyl and the a-methyl methoxy moiety of mycolic acid might influence its capacity to bind CD1b and activate Tcells.
Thed evelopment of mycolic acids and glucose monomycolates as potential antigens in vaccine development and diagnostics requires synthetic material strictly free from immunologically-active impurities.T his was our incentive to undertake the total synthesis of four stereoisomers of methoxy mycolic acid (Scheme 1, so S,S and R,R-stereochemistry in the a-methyl methoxy moiety plus R,S and S,Rstereochemistry in the cis-cyclopropyl moiety) and their glucose esters.T his endeavour required ad rastic decrease in the number of synthetic transformations compared to the previous synthesis by Baird et al. [11] Therefore we opted for incorporation of the Suzuki-Fu cross-coupling on several places in the design. Although the Suzuki-Fu cross-coupling [19] has hardly been applied in natural product synthesis, we advocate the use of this Pd-catalysed sp 3 -sp 3 crosscoupling reaction as it avoids the various functional group transformations and basic reaction conditions required for Wittig-type and Julia-type reactions.
In order to gain access to the four diastereomers in aconvergent manner and on gram scale,weretrosynthetically dissected the molecule in three fragments,e ach containing achiral segment (Scheme 1). Both enantiomers of fragments Ba nd Ch ad to be prepared, and combined with the natural enantiomer of fragment A. Whereas fragment Band Cwere planned to be connected via aS uzuki-Fu cross-coupling reaction, unification of C-B to Ausing this reaction could not be achieved since,a sm odel studies showed, with ac yclopropylmethyl bromide the reaction failed. AJ ulia-Kocienski olefination was projected instead.

Synthesis of the Fragments
The a-branched b-hydroxy acid moiety in fragment Awas installed via an anti-selective Abiko-Masamune asymmetric aldol reaction [20] between aldehyde 6 and chiral ester 9 (Scheme 2a). Thes ynthesis of 6 started with aD IBAL reduction of commercially available 2 to the corresponding lactol 3,d irectly followed by aW ittig olefination with 4 [21] resulting in 5 in 70 %y ield over these two steps.T he a,bunsaturated thioester 5 was then silylated, saturated, and reduced to the aldehyde in as ingle step in 82 %y ield by aF ukuyama reduction. Esterification of chiral auxiliary 8 with the acyl bromide of 7 gave 9 in nearly quantitative yield. Thes ubsequent Abiko-Masamune aldol reaction of 9 and 6 resulted in arewarding dr of 9:1, and partial hydrolysis of the TES ether.T herefore,t he crude was treated with aqueous HCl in THF and subsequently purified by column chromatography to provide diol 10 in 55 %y ield and dr > 97:3. Bis-TBS protection of 10,f ollowed by aS uzuki-Fu coupling reaction [19] with 1-hexadecene resulted in 11 in good yields, and effectively installed the required a-chain. Selective deprotection of the primary TBS ether with pyridine·HF, followed by DMP oxidation of the primary alcohol to the corresponding aldehyde resulted in fragment A.
Thesynthesis of fragment Bstarted with the alkyne zipper reaction [23] applied to commercial 12,f ollowed by protection of the alcohol to provide terminal alkyne 13 in 86 %y ield (Scheme 2b). Deprotonation of 13 with n-BuLi at 0 8 8C followed by addition to paraformaldehyde resulted in 14 in 70 %y ield. In our hands,L indlar reduction of 14 [24] to cis allylic alcohol 15 resulted in amixture with the trans isomer, inseparable on silica and Ag-impregnated silica. Although this problem was avoided by subjecting 14 to aP -2 nickel reduction, [25] this time 1 H-NMR indicated concomitant overreduction. Fortunately,s eparation of 15 from approximately 15 %o ft he aliphatic alcohol succeeded with as ingle run on aAg-impregnated silica column in 73 %yield. Alkene 15 was cyclopropanated with 95 % ee (see SI), mediated by dioxaborolane 16 a or 16 b according to Charettesp rocedure yielding the enantiomeric alcohols 17 a and 17 b, respectively. [26] Thealcohols 17 a and 17 b were subsequently converted into their pivaloyl esters,and desilylated using TBAF to yield both enantiomers of 18 in excellent yield over three steps, [22] and bromination resulted in both enantiomers of fragment B. Fragment Bdecomposed upon extended storage (longer than one month) at 0 8 8C, and was therefore used within one week after synthesis.
Thes ynthesis of fragment C( Scheme 3a)s tarted with adiastereoselective conjugate addition of MeLi to 19 a and its enantiomer 19 b. [11,27] Then, LiAlH 4 reduction followed by bromination of 21 resulted in bromide 22 in good yields over two steps.C hain extension was achieved by subjecting 22 to aS uzuki-Fu cross-coupling [19] with 1-hexadecene.A cid hydrolysis of the acetal, followed by aone-pot tosylation of the primary alcohol and intramolecular S N 2r eaction went smoothly,a nd yielded the corresponding epoxide 25 in excellent yield over two steps.E poxide opening with Grignard reagent 26 using catalytic CuCl minimized the formation of halohydrin by-products [28] and afforded the secondary alcohol 27 in good yield. Finally,methylation of the secondary hydroxyl using an excess of MeI and NaH resulted in both enantiomers of fragment Ci nn early quantitative yields.

Endgame of the Synthesis
With all three fragments in hand, we initiated the endgame by yet another Suzuki-Fu cross-coupling to connect the fragments Ba nd C( Scheme 3b). Reductive removal of the pivaloate,a nd as ubsequent Mitsunobu reaction with 1phenyl-1H-tetrazole-5-thiol afforded the thio-ether in good yields over three steps.U nfortunately,a mmonium heptamolybdate/H 2 O 2 oxidation to sulfone 29 proved to be very sluggish due to limited solubility of the starting material in EtOH/THF,w hereas oxidation with m-CPBAr esulted in ac omplex mixture of unidentifiable products.A lthough oxidation with RuO 4 was possible, [29] concomitant oxidation of the methoxy function to the ketone was hard to suppress. Fortunately,a cceptable rates in the ammonium heptamolybdate/H 2 O 2 oxidation were achieved by addition of n-BuOH as co-solvent, resulting in nearly quantitative yields of the sulfone.S ubsequently,w ea pplied aJ ulia-Kocienski olefination for the final coupling of 29 with fragment A. Although initially excellent yields (85-90 %) were achieved in this reaction, at some point we suffered from irreproducible yields,e ven when we applied starting materials that had provided satisfactory yields earlier. We could circumvent this problem by using an excess of lithiated 29,w hich was nearly quantitatively recovered. With the complete backbone in place,wesubjected 30 to adiimide reduction [30] of the olefin, followed by TBS deprotection with pyridine·HF in excellent yield. Theu se of n-BuOH as cosolvent proved again to be essential for acceptable conversion rates.A sw ew ere reluctant to store large quantities of the mycolic acids because of the potentially sensitive hydroxy acid moiety,the final step was carried out portion-wise at sub-gram scale.R emoval of the chiral auxiliary with LiOH, LiOH/H 2 O 2 or hydrogenolysis were unsuccessful, but hydrolysis with 2equiv of tetrabutylammonium hydroxide (n-Bu 4 NOH) in THF at rt resulted in complete and clean conversion, and yielded the desired mycolic acids in sufficient purity and quantitative yields after astraightforward pentane/acetonitrile extraction. Overall, the total synthesis of mycolic acid 1a was achieved in al ongest linear sequence of just 17 steps,i n1 5% yield. Thec orresponding glucose esters ( Figure 2, 32 a-d)w ere obtained in two steps using am ethod developed by Prandi [31] (see SI for experimental details and yields).

The Stereochemistry of Natural Methoxy Mycolic Acid
With all four diastereomers in hand, it was now possible to prove or disprove the ascription of the absolute stereochemistry (S,S)o ft he syn-methoxy methyl moiety by Asselineau et al. [7] An aliquot of the carboxylic acids was treated with (trimethylsilyl)diazomethane in toluene/methanol to provide the methyl esters (see SI). The[a]D of both methyl esters with S,S configuration in the a-methyl methoxy segment (1a and 1d)was 08 8,which corresponds with the specific rotation of the natural sample reported before. [7] Consequently,a lso the specific molar rotations [F] D were 08 8.T his supports the previous assignment of the stereochemistry by Asselineau, and Baird (for 1d). The[ a] D for the diastereomers with R,R configuration in the a-methyl methoxy segment was + 7.88 8 for 1band + 7.98 8 for 1cand similar to those reported by Baird and co-workers for (synthetic) diastereomer 1c. [11] The[ F] D of 1b and 1c was + 998 8 and + 1008 8,r espectively,a nd comparable with the sum of the values of the [F] D of the amethyl methoxy and the b-hydroxy acid segments prepared earlier by Asselineau (+ 908 8). Therefore,i ti sr easonable to conclude that the remotely located stereocenters in methoxy mycolic acid do not influence each otherso ptical rotation. Theprinciple of optical superposition is therefore applicable here and the assignment of the absolute stereochemistry by Asselineau is correct, although the stereochemistry in the cyclopropyl ring remains to be determined (see below).

Angewandte Chemie
Research Articles 7558 www.angewandte.org Interleukine-2 (IL-2) (Figure 3a). Thel evel of activation by the four synthetic diastereomers,t he natural long glucose mycolate from Mycobacterium phlei,a nd the natural short glucose monomycolate from Rodococcus equi,w ere comparable.T his was consistent with previous observations that glucose monomycolate-specific Tc ells are insensitive to variations in the distal parts of mycolic acid, but highly sensitive to changes in or near the glucose. [17] As for the Tcell activation assays with free mycolic acid, it is known from studies with natural mycolic acid isolates that aTcell clone is typically more potently activated by acertain type of mycolic acid (a-, keto-, or methoxy-) than by another. Thep attern of antigen potencyi sd ifferent for each Tc ell clone.Inthe presence of CD1b-expressing antigen-presenting cells,a ll four synthetic mycolic acids stimulated the Tc ell clone 11 well at the highest concentration tested, 5 mgmL À1 (Figure 3b). Thet wo synthetic mycolic acids with the S,S configuration at the a-methyl methoxy segment, however, were better agonists than those with R,R configuration, and better than the natural mixture of Mtb-derived mycolic acids when tested at the suboptimal concentration of 0.08 mgmL À1 . No significant difference was found for the stereochemistry at the cyclopropylg roup.
As an alternative bioassay,m ycolic acid and glucose monomycolate can be loaded into fluorescently labelled CD1b tetramers in vitro.T etramerized, fluorescently labelled CD1b-lipid complexes ("tetramers") can be used to study glucose mycolate [32] and mycolic acid [6] specific Tc ells at single-cell level, in combination with fluorescently labelled antibodies,i nf low cytometry.T he fluorescence level of the cells is an indication of the strength of the interaction between Tcell receptor and antigenic target, CD1b-mycolate.
All synthetic glucose mycolates,w hen loaded into CD1b tetramers,s tained LDN5 well (Figure 3c). Am inor but noticeable difference in the mean fluorescence intensity (MFI) is observed depending on the stereochemistry of the a-methyl methoxy unit. TheMFI with the GMMs with the S,S configuration is twice as high as in those with the R,R configuration. No noticeable difference is seen between the tetramers loaded with the glucose mycolates that differ in the stereochemistry of their cyclopropyl group.
Fort he mycolic acids,h owever, more distinct differences are observed (Figure 3d). Te tramers loaded with mycolic acids with the S,S-methoxy methyl unit give much higher MFIs than their R,R counterparts.
In addition to the effects of the configuration of the methoxy methyl group,mycolic acids with an R,S-cyclopropyl unit have amoderately increased MFI compared to their S,R counterparts (MFI compound 1d> 1a and compound 1c> 1b in Figure 3d). If we assume that the naturally occurring configuration of the cyclopropyl provides the best CD1b loading or the highest affinity for the Tc ell receptor, this observation suggests the R,S-stereochemistry of the cyclopropyl group in natural mycolic acids.

Conclusion
In conclusion, the synthetic route presented here provides access to gram amounts of methoxy mycolic acid. The successful repetitive application of the Suzuki-Fu cross-  coupling and the anti-selective Abiko-Masamune asymmetric aldol reaction in this natural product synthesis are instrumental to this success.T he availability of all four diastereomers allowed to confirm the S,S stereochemistry of the methoxy methyl function in natural methoxy mycolic acid, as inferred by Asselineau and Baird. TheTcell receptor is sensitive to variations in the stereochemistry of the side chain of free mycolic acid, and not in that of glucose monomycolate. Although the absolute stereochemistry of the cis-cyclopropyl function could not be established, the CD1b-free mycolic acid tetramer staining experiments favour R,S stereochemistry. Thea vailability of large amounts of fully synthetic methoxy mycolic acid and glucose mycolate (GMM) allows further study into their application in vaccine development and diagnostics.