Home » Still In The RBF

Thapsigargin   

3 June 2007 13,837 views 31 Comments

thapsigargin.jpg

Ley, Andrews, Ball, Wierschem, Cleator, Oliver, Högenauer, Simic, Antonello, Hünger and Smith. Chem. Eur. J., 2007, Early View. DOI: 10.1002/chem.200700302; PNAS, 2004, 101, 12073-12078. DOI: 10.1073/pnas.0403300101.

Yep, I’ve finally got this one together, after much procrastination, and a mild leukascandolide diversion. Anyway, many of you will have seen the PNAS article from back in 2004; however, this is the real deal – a full paper on the synthesis of five Thapsigargins. The biological activity is discussed in quite some detail, explaining that the first use of this family in medical preparations was noted by Hippocrates; back in 400 BC! Its activity boils down to sub-nanomolar inhibitor of molecular
Ca2+ pumps, and has been used in conjunction with a His-Ser-Ser-Lys-Leu-Gln peptide, resulting in was specific cytotoxicity to prostate tumours. Loads more information in the paper…

thapsigargin_1.jpg

So, to the retrosynthesis; disconnecting the acyl groups is a fairly trivial move, but the practical aspects of acylating selectively might be a little daunting. From thence we get the 5,7,5 core of the beast, and extensive manipulations of the hydroxyl groups on the carbocyclic skeleton. The seven membered ring was built from an RCM step, providing an olefinic handle for dihydroxylation. Lastly, and quite neatly, they built the cyclopentane from a ring-contraction, starting with an elaborate cyclohexanone.

thapsigargin_2.jpg

And thus we start with some old-school rearrangement chemistry, doing a rather tasty Favorskii rearrrangement on the cyclohexanone (built itself from an epoxidation on carvone, and then opening with chloride). This also generated a fourth stereocentre, creating a rather complex cyclopentane very quickly.

thapsigargin_3.jpg

They then did anUpjohn oxidation on the olefin and cleavage with periodate to provide a ketone, alkylated with Felkin-Ahn control to give the corresponding alcohol, using titanium(IV) isopropoxide and then addition of allylmagnesium bromide. Elaboration of the lower sidechain gave them an enol-ether, which was Grubbsed to give the seven member ring, with that enol ether still intact. Dihydroxylation conditions then oxidised this to the α-hydroxy ketone; not a method I’d seen before, but quite logical.

thapsigargin_4.jpg

Skipping a chunk of work, they used a HWE reaction to bolt on an α,β-unsaturated lactone unit, which was opened and dihydroxylated. This, in fairly quick succession, lead to tetraol moiety, which was selectively oxidised using Ley’s own methodology, resulting in re-lactonisation. Top stuff.

thapsigargin_5.jpg

To introduce another hydroxyl centre stereoselectively, they took a derivative of the latter lactone in which the cyclopentane has been oxidised to a cyclopentanone, and form the TMS enol-ether. Substrate controled epoxidation (see Houks computational studies on this) then released the TMS group and left the desired α-hydroxy cyclopentane.

thapsigargin_6.jpg

Lastly, reduction of that cyclopentane with sodium borohydride gave attack from the wrong face wrt the target. However, simply moving to the zinc analogue gave the desired stereochemistry in a pretty decent selectivity. Not too surprising, but that must have come as a hell of a relief! They were then able to selectively deprotect in order to acylate the compound to the natural product, taking advantage of orthogonal protection and some cunningly hindered hydroxyls. Particularly impressive was the final esterification, which provided Thapsigargin A via preferential esterification of one secondary hydroxyl over another in 92% yield.

An impressive example of a complex polyoxgenated natural product, and a fine read (they also note that the average yield in the synthesis was 88.6% per step!!!).

1 Star2 Stars3 Stars4 Stars5 Stars (No Ratings Yet)
Loading ... Loading ...

31 Comments

  • boronman says:

    88.6% average per step. No way is this reality.

  • thirsty scholar says:

    you’ve got some texas carbons in the last three structures. excellent entry, otherwise!

  • McPostdoc says:

    Why on Earth do they need 40 equiv. of NMO for that TPAP oxidation?

  • pi* says:

    funny place to publish the full paper no?

  • Tot. Syn. says:

    Not too odd… Chem. Eur. J. is the full-paper companion to Angewandte, so is just like JOC is to JACS. But Steve hasn’t published much in JOC for the last few years, apparently because of an “altercation” involving a competitive publication, and has never published as the starred author in JACS. Interesting.

    http://pubs.acs.org/journals/oprdfk/promo/ley/articles.html

  • undergrad says:

    hate to hijack this thread.. but are any of you attending the national org symposium at duke? We outta go get some beers an talk about the marvels of modern organic chemistry

  • TWYI says:

    no 3.

    It is all about making sure there is no ‘reduced’ ruthenium about in the flask which can be bad news for sensitive substrates.

    The excess NMO regenerates the active ‘oxidised’ ruthenium very quickly.

  • pi* says:

    I would have submitted it as a jacs full paper…then again Ley does have more experience with this sort of thing than me

  • ddd says:

    what’s up with aztec chemicals?

  • willyoubemine says:

    Cool stuff. I love the Favorskii, you dont see that too much.

    why is this published twice?

  • Vindar Nut says:

    No. 1 : I think if you do the mathematics, that the average yield per step is correct at 88.6% and I am unsure as to why you think this is unbelievable…. aka “No way is this reality”. Maybe you could explain your problem with this statement

    No. 4 and No. 8 : I totally agree with pi* that this synthesis is worthy of being published as a full paper in JACS, however submissions to journals are rarely apolitical. I’m sure we all have stories about papers that should have been accepted and were not or vice-versa .
    Once bitten twice shy………

  • TWYI says:

    Worthy of JACS or not, it does not make it any less impressive just because it is in CEJ.

  • willyoubemine says:

    the initial cyclohexanone, where does that come from, chiral pool? carvone, or soemthing? its a nice strategy to the 5,7 bicycle.

  • boronman says:

    I am in no way saying the synthesis is not a good one. I realize the effert that all of the co-authors put into it must have been tremendous. After closer inspection of the paper, there are some very nice steps involved in making the 5-7-5 ring system. However and as someone who has done a ton of multi-step synthesis (beyond 30), I rarely beleive when one claims consistent yields of above 90%. Great work from the Ley group, but I suspect that a few yields may have been embellished a little.

  • Tot. Syn. says:

    I believe them. Taking yields I’ve got over the first fourteen steps in my total synthesis I currently have a average yield of 88.14%. The last six steps have only been completed once, so there’s actually room for optimisation.

    It’s perfectly possible, especially if you have protection / deprotection steps weighing the average towards the top-end. Take a simple three reaction sequence of Protect – Transformation – Deprotect, with yields of 96% – 70% – 92%. That averages out as 86%, and is quite reasonable…

  • Mytuppence says:

    For a linear reaction sequence, perhaps a more useful measure of “average yield” would be the nth root of the product of all individual steps.

    eg. Hypothetical two step synthesis with 80% and 20% yields. Overall yield 16%.

    Mean yield = 50%, however, two steps at 50% would give overall 25%, (always) gives an overestimate.

    But if average quoted = sqrt(0.8*0.2) = 40%, then two steps at 40% gives overall 16%

    The only slight issue is that this average is always smaller than the mean!

  • WillisWill says:

    It was always my experience that multiple runs of the same reaction never gave the exact same yeild. One day, it may be 90%, another day 80% – depending on the age/purity of the reagents/catalysts, the quality of purification, the hungover-edness of the grad student, etc…where I think some disinguenuity may come into play is when someone reports the “best” yield for each step in the sequence, even though they did not necessary happen in the same “run.” You can sometimes spot this in the experimental section when a reaction sequence A —> B —> C —> D reports the reaction of 2 mmol A, 0.2 mmol B, 2.5 mmol C, and 0.01 mmol D.

    I’ve not read Ley’s paper, and am in no way accusing his group of doing this for this paper or in any other.

    CEJ may not have the prestige of JACS currently, but if people like S.Ley continue to publish high quality full papers there, that will change. Certainly its better for everyone if there is more than one outlet for prestigious full papers. That may also be part of his motivation for submitting there.

  • Vindar Nut says:

    I believe the analysis completely.

    Interesting discussion, I had hoped to stimulate some.

    I have often contemplated the most accurate statistic to really gain an appreciation of how “good” a synthetic sequence. My opinions for what they are worth :

    By quoting the average yield per step, low yields can be compensated for by high yields – that is true. However, overall if you have a large number of low yielding steps or a single really low yield (1%) it will be reflected by the average yield per step most accurately. My opinion is that the average yield per step reflects the relative efficiency of the whole sequence. ……kind of like the total greater than the sum of the parts…….

    The authors are not claiming that every steps works in 88% yield, but over the whole sequence, it appears to be equivalent to a route of X steps with every step
    having an average yield of 88%, which is not interperation but a fact!

    The facts are never wrong only the interperation of them!

    Another classic among the total synthesis community is longest linear sequence. I believe that doesn’t truly reflect the amount of labour associated with a given synthesis (measured in grad student /postdoc student sanity!).

    Take two synthetic sequences to prepare the next awe inspiring natural product :

    Route A, is convergent and relies on the coupling of two fragments, each takes 10 steps to make. (Total = 20 steps, longest linear sequence = 10 steps)

    Route B, is also convergent, but features a more elaborate fragment, this takes16 steps to prepare, this is then coupled with a commercially available material to complete the synthesis. (Total number of steps = 17 steps, longest linear sequence = 16 steps)

    Another classic deception, start counting your steps from compounds known in the literature, providing you reference them that okay…… right…… maybe….. not in my opinion…… not lying…. Really……. merely economical with the true….. hmmm

    I agree with WillisWill and Mytuppence, on the fact that even good reactions are not completely reproducible over a large number of experiments. However, are we going to believe that in total synthesis people don’t always report the best-ever yield for a given transformation – a little naïve if you don’t.

    It’s just human nature to represent the facts in the best possible light.

    Maybe everyone just needs to be a little more honest and not have the attitude “everyone else is doing it so why don’t we.”

    What the community needs is a statistic that reflects the total number of steps in the preparation of a molecules that somehow includes a measure of how good the steps are. Average yield per step seems like a good one to me.
    Maybe embrace some statistical method, variance , standard deviation, range of yields including scales. What is the most accurate picture of what we do?

    What do you guys think?

  • TheEdge says:

    Vindar, I’d take the shorter linear sequence every time. If the yields on the individual steps are similar, I’d need to start with significantly less material. Lets say you need 30 mmol of material at the end of your sequence (which, I admit, is a lot) and get an average yield of 80%. You’d need to start on 300 mmol for your convergent synthesis, and over 1 mol for the non-convergent.

    In the spirit of full disclosure that this discussion warrants, my project has three times as many total steps as linear steps, so I’m a little biased.

    I think the most accurate picture for yields is to give a range, or note that yields varied on scale. I’ve found that variable yields in my synthesis are due to either scale (and the exotherm associated with it) or choosing to proceed with an impure reagent, both of which could be mentioned in an experimental.

    Yields over an entire sequence are nice to mention, if the sequence is reasonably short (

  • willyoubemine says:

    All this discussion and the conclusions is based on the supposition that yields and number of steps decides the validity of a synthetic route. Maybe you agree with that, I dont.

    As Shakespeare wrote: “The strategy is the thing” -admittedly, an obnoxious point.

    After all is said and done, process chemistry is best left to industry. IMO, academia’s role is in expanding the scope of what process can do. New strategies, novel disconnects and clever methodologies are what decide the relevance and inherent beauty of a total synthesis.

  • Vindar Nut says:

    Nice point made by TheEdge and I’m just playing devil’s advocate but :

    But you would need 300 mmol of both starting materials (600 mmol) and that also assumes that the yield over both convergent sequence is equal. The number of individual operations you need to perform is more – it may be more efficient in terms of yield (over the longest linear sequence) but what about the time you invest in running (and developing) two independent ten step sequences.

    More often than not, one sequence will be far more efficient than the other so practically you will need to run at least one of these sequences on greater than 300 mmol to get a the required amount of your limiting material for your synthesis. Let imagine one ‘leg’ of your convergent route is half as efficient, then you need 600 mmol of one SM plus 300 of the other SM (total = 900 mmol) not too much different from the convergent route now and heavily dependent on the individual reaction yields in the sequences.

    I am not saying a short longest linear sequence is a bad thing to mention, but it should not be the be all and end of all when considering the merits of any synthetic route.

  • Vindar Nut says:

    No. 20 Great point willyoubemin!

    what makes a good synthesis is totally subjective and to use another quote from the literary giant and fellow countryman :

    ” Good Lord Boyet, my beauty, though but mean,
    Needs not the painted flourish of your praise:
    Beauty is bought by judgement of the eye,
    Not utter’d by base sale of chapmen’s tongues.”

    I guess the Beauty is in the eye of the beholder

  • McPostdoc says:

    Mytuppence,

    For a linear reaction sequence, perhaps a more useful measure of “average yield” would be the nth root of the product of all individual steps.

    This *is* how average yields are calculated in papers, right? In Ley’s paper, he claims 43 steps with an overall yield of 0.61%. He calculates the average yield from this: X^43 = .0061 – therefore X = .886, so the average yield is 88.6%.

    Tot. Syn. should have said his average yield in post 15 was 85.2%, since .852 = .96 * .70 * .92.

  • McPostdoc says:

    Huh. The ‘Latest Comments’ column on the upper right grabbed the first part of my post, including an open italics tag for when I was quoting Mytuppence’s post. But it doesn’t go far enough to catch the close italics tag, so it puts the entire front page and all posts completely in italics.

    Sweet! I now have the knowledge to change all the text on the front page of the Totally Synthetic blog to, say, bold, pink, size six letters. That’s pretty awesome.

  • TWYI says:

    Come on guys, protecting groups are so 2006 in total synthesis.

  • Tot. Syn. says:

    McPostdoc:

    I have your IP address, and know who you are. Even now, I’m watching you on google maps, and can follow you everywhere. Make one move against this blog and I’ll spike your silica…

  • TheEdge says:

    Yeah, Vindar Nut, I played fast and loose with my numbers. I also ignored the possibility of needing two equiv of one of the coupling partners in the final reaction. FWIW, I would report the yield of the poorer yielding sequence, but I’m sure not everyone would.

    I also think it’s easier to work out any number of ten step syntheses than a single 16 step synthesis. Your supply of material at step 5 is always going to be much larger than at step 12, making it that much easier to find things that work (and much easier to admit that the route just isn’t going to cut it and find a different set of bonds to build around).

    I agree that longest linear sequence isn’t the most important aspect of a synthesis, but I do think people need to pay more attention to things like convergency when planning syntheses. I think Baran’s work is most impressive because it is all relatively convergent. His syntheses are so tight because he rapidly increases complexity by bringing together relatively complex pieces and then not doing a lot of fiddling afterwards, and he does it on small molecules where the norm is to build a carbon skeleton and then install functionality one group at a time.

    And one last thing: I think this is a good synthesis paper. I appreciate their comments about the decomposed aldehyde, among other things. It is steppy, but I think any synthesis that sets out to make 5 different members of a class (and other analogues for testing) with such diversity in the acyl sidechains is going to get a little steppy.
    I’m also really weirded out by the zinc borohydride reduction. The zinc is certainly chelating to the SEM ether, but the product certainly isn’t what I would predict. Normally you get attack over the hydrogen on the chelating stereocenter.

  • pi* says:

    yield in steps = only important if your comparing two similar things, but who really cares.
    strategy=everything

    http://www.chem.harvard.edu/research/faculty/matthew_shair/files/JACS2000_7424.pdf

    look, scheme 2, step d = 31%, step e = 51%, step n = 12%.

    must be rubish, right?

  • spottospot says:

    #28
    peace

  • McPostdoc says:

    Tot. Syn.

    Fear not. Even though I could use a different IP address (after all, some of can access the intertubes from home, too), I will use this knowledge only for good.

    But seriously, that bug does make it a bit tougher to use HTML in posts, as is nice to do when quoting a post to which one wants to reply.

  • Grubbs the cat says:

    Stumbled across this just now… As a former postdoc on the project I can assure everybody that the yields were not faked by any means. The synthesis is amazingly robust and does scale – it was awsome to see by successor with 1 gram of step 26 where I had finished with 20 mg before!

    Anyway, I saw it triggered an interesting discussion on statistics!