Synthesis Golf III: Chloramphenicol

A full FAQ post has been written on meta.chem.SE, explaining the premise of synthesis golf and the 'rules'. Please take a look at this before answering (if you haven't already).

This third round of golf concerns the synthesis of chloramphenicol 2,2-dichloro-*N*-[(1⁠*R*,2⁠*R*)-1,3-dihydroxy-1-(4-nitrophenyl)propan-2-yl]acetamide, an antibiotic on the World Health Organisations list of essential medicines:

InChI=1S/C11H12Cl2N2O5/c12-10(13)11(18)14-8(5-16)9(17)6-1-3-7(4-2-6)15(19)20/h1-4,8-10,16-17H,5H2,(H,14,18)/t8-,9-/m1/s1

In addition to being a useful target, chloramphenicol was chosen as during previous rounds of , several people commented that they'd like to see targets amenable to biosynthetic transformations, and scaleable process routes. This target very much fits both of those criteria (the drug is manufactured on a huge scale, and several enzymatic routes have been developed).

The challenge is to propose a route to chloramphenicol, including a method for setting the absolute and relative stereochemistry of the 1,2-hydroxyamine functionality. You may start from anything commercially available in the Sigma Aldrich catalogue, but to keep things interesting, no starting material may contain more than 8 carbon atoms.

Approach:

The approach taken in this proposed synthesis of chloramphenicol is largely based on Greene's synthesis of paclitaxel, using dihydroxylation methodology proposed by Sharpless.

Whilst not particularly elegant (I was hoping for some kind of asymmetric hydrogenation to set the stereocentres), it does get to the product in a respectable 7 steps (including setting both stereocentres, as required), and avoids having to introduce protecting/directing groups at any stage.

Forward synthesis:

Step 1: Horner Wadsworth Emmons olefination

The first step in this proposed synthesis is a Horner Wadsworth Emmons olefination to introduce the (E)-alkene required for the planned dihydroxylation. Without the nitro-group, the exact reaction has been done many times and with good selectivity, so no reason to assume that this HWE would be particularly challenging (the nitro group clearly makes the aldehyde electron poor).

Step 2/3: Sharpless dihydroxylation, epoxidation sequence

With the (E)-alkene formed, a Sharpless asymmetric dihydroxylation can take place using AD-MIX-B (taking the aryl group as 'large' and the ester as 'medium'). This step (without the nitro) is present in the Greene paclitaxel synthesis (above) and initially gives 82% e.e, though a single recrystallisation of the product brings this up to >95%.

Once recrystallised, treatment with mesyl chloride and subsequent heating (same pot) affords the necessary epoxide.

Step 4/5: Epoxide opening and reduction

Now comes the slightly dubious step....

In the paclitaxel synthesis, the epoxide is opened with sodium azide at the benzylic position (expected, the aromatic ring can stabilise the developing partial positive charge here), however in our case, we have an incredibly electron withdrawing nitro group in the 4-position, completely shutting down any stabilisation at the benzylic position.

Based on this lack of benzylic stabilisation, we could reasonable assume that it will take place with the opposite regioselectivity to the paclitaxel case. There is some evidence for this in the literature, (including where there is an ester present on the other side, but sadly not using azide as the nucleophile), and a cursory flick through an organic textbook will tell you that SN2 adjacent to carbonyls is generally fast.

Assuming it works fine, a Staudinger reduction converts the azide to the secondary amine with retention of stereochemistry.

Step 6: Ester reduction

Lithium borohydride is a mild and selective reagent, able to reduce esters in the present of various other functionality, including nitro groups. One draw-back of this is that lithium borohydride is significantly slower than its counterparts (not to mention more expensive), but, using LAH or other highly reactive hydride species would result in side reactions involving the nitro group.

Step 7: Amide bond formation

The required acyl chloride is commercially available, and should do the required reaction in place of the possibly competitive ester formations (nitrogen LP higher in energy).

Conclusion:

7 steps overall starting from the cheap, commercially available 4-nitrobenzaldehyde. The majority of steps have good precedence on similar systems (ignoring the epoxide opening, which may/may not be fatal), and all of the proposed reactions can (and have been) demonstrated to be possible on large scale (again, ignoring the epoxide opening, but TMS-azide is often substituted in the case of process syntheses involving azide).

• Even if you form the ester in the last step, you might get transfer of the acyl group onto the nitrogen since that's an intramolecular process. I've had this cause problems when I wanted the amino ester instead...
– Zhe
Jul 9 '17 at 18:38
• Could you use LiBH4 to reduce the azide along with the ester? Saves a step and you don't have to get rid of Ph3PO which is always a pain. Jul 9 '17 at 20:20
• Regarding the epoxide opening, Clayden gives typical $k_\mathrm{rel}$ for SN2 at (1) benzylic positions (2) alpha to C=O; the latter is faster by several orders of magnitude. Jul 15 '17 at 16:38
• I believe you mean the Staudinger reduction will give you the primary amine, and not a secondary amine. Dec 12 '18 at 22:29

Some literature searching led me to a different linear synthesis with 8 steps.

Overall discussion of strategy

A different, and perhaps more direct, route to aminoalcohols is provided by the Sharpless asymmetric aminohydroxylation:

Of course, it's never that simple. Styrenes tend to undergo this reaction with the wrong regioselectivity, i.e. the nitrogen tends to add to the benzylic position. In particular, Kurti and Czako write:1

cinnamate esters react to give preferentially the β-amino ester product...

A solution to this is to use a different chiral ligand, instead of the usual DHQ–PHAL and DHQD–PHAL ligands. In 1998, Sharpless reported the "reversal of regioselection in the asymmetric aminohydroxylation of cinnamates" , using an anthraquinone (AQN) core instead of the phthalazine (PHAL) core.2

There is another catch, in that electron-poor aryl groups don't work well in this reaction, for whatever reason:

Electron poor cinnamates proved less suitable with the AQN iigand system. For example, the AA of methyl 3-nitrocinnamate gave a 1:1 mixture of regioisomers and significant amounts of the diol by-product...

I don't really view that as a problem per se, though. The removal of the N-Cbz group is usually done with a hydrogenation, which is already incompatible with the nitro group.3 Instead, the entire problem can be circumvented by using the 4-chloro derivative, which was reported to give a 77:23 ratio of regioisomers B:A with an accompanying ee of 92%. The chloro group can then be converted to the nitro group near the end of the synthesis, using palladium chemistry developed by the Buchwald group.4

Okay - enough talking, now for the synthesis...

Forward synthesis

Methyl 4-chlorocinnamate is commercially available, but the restriction on starting materials to 8 carbons or fewer means that we need to make it. Of course, there are lots of valid possibilities. The most common way seems to be a Heck reaction, but olefin metathesis and the Horner–Wadsworth–Emmons have also been used. This cinnamate can then undergo the Sharpless aminohydroxylation as described previously, which creates the key C–O and C–N bonds with the desired stereochemistry.

The next step is DIBAL reduction of the ester to the alcohol, followed by protection of the 1,3-alcohol as the acetonide. I inserted this protection step as I'm afraid that the free hydroxyl/amino groups adjacent to each other would poison the Pd catalyst in the nitration step.

The synthesis is completed by:

• hydrogenolysis of the Cbz group (the use of ammonia as a solvent inhibits benzyl ether cleavage5, eliminating any possible risk of destroying the acetonide).
• acylation on N (the use of the acetonide also nicely sidesteps any potential acylation on O). I chose to do this before the nitration, just in case the free amine interferes with the Pd.
• Pd-catalysed nitration, which seems to have quite a broad substrate scope, and
• removal of the acetal with aqueous acid, preserving the amide.

Notes and references

1. Kürti, L.; Czakó, B. Strategic Applications of Named Reactions in Organic Synthesis; Elsevier: Amsterdam, 2005; pp 404–405.

2. Tao, B.; Schlingloff, G.; Sharpless, K. B. Reversal of regioselection in the asymmetric aminohydroxylation of cinnamates. Tetrahedron Lett. 1998, 38 (17), 2507–2510. DOI: 10.1016/S0040-4039(98)00350-5.

3. The use of a Boc carbamate (instead of Cbz) is not reported in the paper, but I wonder if it could be done, as the deprotection with TFA hopefully wouldn't mess up anything in the compound. However, the poor reactivity of the nitro cinnamate would then become a problem, necessitating some other workaround, so I didn't pursue this line of thought any further.

4. Fors, B. P.; Buchwald, S. L. Pd-Catalyzed Conversion of Aryl Chlorides, Triflates, and Nonaflates to Nitroaromatics. J. Am. Chem. Soc. 2009, 131 (36), 12898–12899. DOI: 10.1021/ja905768k.

5. Sajiki, H. Selective inhibition of benzyl ether hydrogenolysis with Pd/C due to the presence of ammonia, pyridine or ammonium acetate. Tetrahedron Lett. 1995, 36 (20), 3465–3468. DOI: 10.1016/0040-4039(95)00527-J.

I started my route from D-serine. Not as available as biological L-serine, but still fairly cheap.

(a) TBS protect the alcohol. Otherwise, I'm going to get annoyed later trying to remove a benzyl ether.

(b) Exhaustive benzylation leads to dibenzylamine.

(c) Reduction with LAH

(d) Swern oxidation with dimethylsulfoxide and oxalyl chloride.

(e) I found this nice paper from Lüdtke (JOC, 2017, 82, 3334). The conditions are to use an aryl boronic acid (here, 4-nitrophenylboronic acid) and diethyl zinc. Unfortunately, I don't have journal access, so I can't check the substrate table to see if this particular combination works and gives good selectivity. As a general note, Reetz reports that these types of dibenzyl aminoaldehydes generally afford the Felkin-Ahn product, which is not what we want. This method with diethyl zinc affords the chelation controlled product, which provides the desired syn selectivity.

(f, 2 steps) Time to remove protecting groups. 10% palladium on carbon with ammonium formate should remove both benzyl rings. $\ce{TBAF}$ should pop off the silyl group. This product has a good chance of being crystalline, so we might consider shoring up purity with a recrystallization.

EDIT: The hydrogenation is probably the most questionable step in this synthesis since I have to do it in the presence of a nitrobenzene. In the worst case, we might be able to oxidize the aniline back to a nitrobenzene.

(g) Finally, dichloroacetyl chloride with $\ce{DMAP}$ and triethylamine should provide the desired product.

Eight steps in the longest linear sequence. Not my best work, but I think it works.

• I found the nitro group incredibly irritating too. Ugh. I had everything planned out and that nitro group was the only thing stopping a hydrogenation of Cbz + benzyl ether... Jul 9 '17 at 17:22
• Fingers crossed that the benzyl deprotection doesn't just wreck the nitro group... @orthocresol
– Zhe
Jul 9 '17 at 17:23
• Intermediate 5 is really intriguing. (1) it looks tantalisingly like it might be possible by some desymmetrisation of a 1,3-propane diol derivative, Krische style maybe, and (2) because it looks worryingly like it would epimerise Jul 9 '17 at 18:52
• If instead of benzyl you use dimethoxy benzyl for your N protection then it will come off more easily. However one should remember Murphy's Law applies. Jul 9 '17 at 20:16
• @Waylander Good idea. The references I found for that selective addition to the aldehyde are from the dibenzyl amine, but it's unlikely that substitution on the benzyl rings would kill the selectivity...
– Zhe
Jul 9 '17 at 20:19

I solemnly swear I came up with this on my own...

Retrosynthesis

The aminoalcohol motif (to me) strongly suggests the addition of an azide to an epoxide. The requisite chirality can be introduced with a Sharpless asymmetric epoxidation; then, the second stereocentre is easily set because the epoxide opening is stereospecific.

This target is relatively simple, so the retrosynthesis can probably fit in one scheme. I think it's also (mostly) pretty self-explanatory.

Forward synthesis

The choice of protecting group P in the starting material 5 is largely dictated by the presence of the nitro group in 6. Quite a lot of protected 2-bromoethanols are available in the Sigma–Aldrich catalogue; the obvious choice in terms of cost is probably the methyl ether (CAS 6482-24-2, 5 g for £20).

A standard Wittig reaction with 4-nitrobenzaldehyde 6 (CAS 555-16-8, 10 g for £19) should give the (Z)-alkene. Demethylation with $\ce{BBr3}$ and the appropriate Sharpless epoxidation should hopefully get us to 3:

and now I run into the same problem as described in NotEvans's answer: it would seem that the preferred regioselectivity of epoxide opening is at the benzylic position, where the $\mathrm{S_N2}$ process is accelerated.

My workaround would be to Swern oxidise the alcohol; open the epoxide (now the position α to the carbonyl group should definitely be favoured); and re-reduce the aldehyde (sodium borohydride):

The endgame is of course a Staudinger reduction and amide formation. As has been mentioned many times, dichloroacetyl chloride is commercially available (CAS 79-36-7, 5 g for £8).

If acylation on oxygen occurs I'd guess it can be reverted with $\ce{K2CO3/MeOH}$, and the amide should hopefully stay put.

All in all it's 6 steps if the epoxide is cooperative, 8 steps if it isn't.

• $\ce{BBr3}$ on a non-aromatic methyl ether... you like to live dangerously! Nice route though, 6 steps Jul 9 '17 at 19:19
• @NotEvans. That's true... and it's also allylic... probably not the best idea. :( Any suggestions for a replacement? I was also thinking of using an MOM ether (sm also commercially available) and removing it with PPTS... (at least Greene said it was good for allylic alcohols!) Jul 9 '17 at 19:21
• Good old THP ether would be OK here Jul 9 '17 at 20:13

Ok, so I decided to try this synthesis golf this time.

I decided to use benzene, 3-hydroxypropanoyl chloride, and dichloroacetyl chloride as my raw materials.

I'm starting off by performing a Friedel-Crafts acylation of benzene.

I will then do a Wolff-Kishner to get rid of the carbonyl.

Performing allylic substitution,

Dehydrohalogenation,

Diol addition using $\ce{OsO4}$,

Nitrate the benzene ring,

Replace $\ce{-OH}$ with $\ce{-Cl}$,

Replace $\ce{-Cl}$ with $\ce{-NH2}$

And, that's it. Hopefully, this synthesis goes well.

• Your starting acyl chloride probably doesn't exist, it would react with itself. At the very least, I'm unable to find a reliable source selling me that compound. Jul 9 '17 at 6:57
• There are a handful of reports of making that compound, probably intended for immediate use. Even so, it seems improbable that the acyl chloride will react with the benzene ring first before the hydroxyl group (much more nucleophilic). Jul 9 '17 at 7:11
• Are you sure about the outcome of your HCl reaction with the triol? I'm not. Jul 9 '17 at 9:34
• In addition to the above, you might also possibly replace the $\ce{OsO4}$ step with an enantioselective variant in order to make a single enantiomer, as required. In any case, nice attempt! Jul 9 '17 at 14:13
• @NotEvans Thanks. I don't know much reactions anyway, I only knew this reaction. I guess I'll get better at Synthesis Golf as time passes, and I learn more reactions. Jul 9 '17 at 15:22

Starting materials: acrylic acid, benzene, protected ammonia, dichloroethanoyl chloride.

6 total steps.

Conditions:

(1) - I2, NaHCO3

(2) - Benzene, Lewis Acid (AlCl3)

(3) - NH2Ts, base

(4) - DIBAL-H

(5) - H2SO4, HNO3

(6) - Dichloroethanoyl chloride, base

Reactions:

(1) - Iodolactonization, the 3-membered ring forms because if the 4-membered ring formed, Baldwin's rules would be violated.

(2) - L.A. addition to the carbonyl, benzene attacks the electrophilic carbon. This creates a deprotonated alcohol alpha to an alkyl halide, therefore epoxide formation occurs with predetermined stereochemistry.

(3) - Opening of epoxide using protected ammonia, inversion of stereochemistry occurs.

(4) - Reduction of the carbonyl. Occurs with this stereochemistry because the -Tosyl group is bulky and is hindering an attack from above. Therefore the attack proceeds from below and the alcohol is formed facing upwards.

(5) - Nitration of this compound occurs using a nitrating mixture, due to the strong acids present in the mixture, the amine is deprotected (tosyl group is removed).

(6) - Amide formation, self-explanatory.

• I'm surprised the free OH groups are untouched by the nitration conditions; there seems to me to be plenty of scope for this going wrong Jul 15 '17 at 14:29
• I tried searching the literature for any precedent of the first step and couldn't find anything. It strikes me as being a little odd (at least, to my knowledge), as the alkene conjugated to C=O is more electrophilic than nucleophilic (and hence iodonium formation, which involves nucleophilic attack of C=C on iodine, is harder). Jul 15 '17 at 16:35
• Polymerization might be the problem, I know alpha propiolactones are generated from the closing of deprotonated halopropanoic acids, so the same product may be able to form from deprotonation of 2,3-dibromopropanoic acid. Also, regarding the first comment, the -OH groups could be protected, resolving the nitration issue that may arise. Jul 16 '17 at 0:28

A synthesis starting from a basic starting material. Asymmetric induction was through the Sharpless epoxidation.