Today our work on the total synthesis of the araiosamines is out in JACS. While the chemistry (success and unexpected failures) was detailed in the paper and SI, I would like to share some stories from the science behind the construction of these molecules.
On the first day I joined the lab, I told Phil that my motivation was to get extensive training in natural product total synthesis in a top tier synthetic chemistry lab. Phil immediately suggested the araiosmianes (isolated from a sponge collected from Vanuatu in 2011), saying, “if you want an education in total synthesis, these molecules would be an excellent option.” When I had a first glance at these molecules in the isolation paper, I thought they were only trimers of three bromoindoles, and there were only six carbons in the skeleton, no big deal at all! However, a very talented graduate student, Ming Yan, had been struggling with these molecules for two years. When involved in the project, I found these molecules are really tough to make, to say the least.
During the first two weeks, we continued to work on the strategy, which employed an Achmatowicz reaction. However, due to the failure of indole installation, Phil decided to abandon this route. We met Phil in his office early on the following morning and tried to come up with new ideas. After one hour of discussion, we left his office without feasible plans in mind. We thought it was the time to move on. Maybe a guanidine related methodology project? Five minutes later, we received an email from Phil, asking, “did we consider the sulfones?” We immediately went to the library to mine the literature and found that α-amido sulfone is actually a stable acylimine precursor for the Mannich reaction. Eventually, the amido sulfone approach saved the project and opened the door to a completely new strategy—stepwise construction of the linear carbon skeleton.
The first Mannich reaction gave two diasteremers, however, favoring the undesired one. When performing the DIBAL reduction with either one, we always got the aldehyde as two inseparable diastereomers (poor yields, irreproducible and variable dr). We thought epimerization might be an unavoidable problem of the aldehyde. The next aldol reaction with the aldehyde as a mixture of diastereomers indeed gave the trimer product, but again in irreproducible and variable yields. During the attempts for the second Mannich reaction through nitrile hydrozirconition (SI), we found the nitrile could be reduced to the aldehyde with the Schwartz reagent without epimerization. This tricky reaction (the concentration and stoichiometry are both critical) must be quenched by loading the reaction mixture on the TLC plates (fortunately 1 TLC plate could quench about one gram scale reaction). Trace of EtOAc in the starting material, or quenching the reaction by silica gel powder, or aqueous solution, would induce epimerization.
With pure aldehydes prepared, we observed their interesting reactivities towards the subsequent aldol reaction: the desired diastereomer 4 worked well to give the trimer product, whereas undesired diastereomer 3 gave a complex mixture, which could not be identified. Fortunately, what we needed to do was to epimerize the undesired nitrile to the desired 2 with t-BuNH2. Moreover, this unusual reactivity also inspired us to solve a key problem of C-H gunidinylation.
In parallel to the skeleton construction, we also attempted the key C-H guanidinylation by indole oxidation of 6 prepared from the undesired Mannich product as a model substrate, due to the scarcity of the desired, but minor Mannich product. After screening ca. 100 conditions (including DDQ), only mixture of diastereomers in extremely poor yields was observed. When we figured out that the aldol reaction worked only with the desired diastereomer, the question arose: what about this C-N guanidinylation with the desired diastereomer? Surprisingly, DDQ mediated C-N guanidinylaiton worked very well with the desired diastereomer in quantitative yield and with complete stereoselectivity!
To make a long story short, we combined both a linear and a cyclic strategy to finally make the mesylate 10 for the SN2 reaction. The subsequent reaction sequence of substitution, reduction, and guanidinylation worked very well. It seemed that ariaosmiane C would be conquered very soon. However, two months of effort did not lead to any trace of the natural product. That was the most difficult time for us. We were so close to the natural product: the mass was always correct, but the NMR spectra never matched! We began to doubt the stereochemistry—the only reaction that could give the wrong stereochemistry would be the SN2 reaction, because all other stereogenic centers were confirmed by X-ray crystallography of the precursors. The 2-D NMR spectra did not help very much in this case. Some signals supported the right stereochemistry whereas some did not. Phil said we would definitely need X-ray crystallography, given that Palau’amine’s structure was misassigned by NOE, which was not conclusive for such a complex structure. We decided to determine the stereochemistry of azide 11 by X-ray. However, to grow a crystal of such a late stage intermediate was not an easy task. We spent a lot of effort on the scale up and obtained 20 mg of the azide 11. Unexpectedly, too many bromines hindered the crystallization process, probably due to their too much lipophilicity. After extensive solvent screening, we eventually grew a crystal from a solvent mixture of MeCN, MeOH and H2O. After obtaining the X-ray crystallography, we were completely shocked—the configuration was retained when the azide was introduced in the displacement step. We also found that the six membered N,O-acetal ring has a perfect chair conformation while the unexpected axial bromoindole and azide have an antiperiplanar conformation. Clearly, due to the neighboring group participation of the axial bromoindole, a double inversion took place.
How could we prevent the neighboring group participation? We tried many approaches including indole protection as the most straightforward, but none of these worked. The final idea was to employ a reductive amination, albeit not anticipated to be stereospecific. Interestingly, hydroxylamine was the only nucleophile that could condense with the ketone. Fortunately, after extensive experimentation the stereo- and chemoselective oxime reduction was achieved with SmI2 in the presence water. We were quite lucky because, initially we used the methoxy-substituted compound 14 as the substrate to investigate the reduction. In a later study, we found OTBS substituted compound 16 completely reversed the stereoselectivity of this reduction. Possibly, OMe directed the protonation from bottom face to give the desired stereochemistry. The desired product 15 showed very broad 1H-NMR peaks (some are missing) compared with the undesired isomer probably due to various conformers. Our tremendous effort spent on characterization of the undesired isomer was quite helpful. Without the X-ray of the undesired azide 11, how could we confirm the stereochemistry of desired amine 15 from its low quality NMR spectra? And the amine 15 was decidedly more difficult to crystallize.
With the correct stereochemistry established, after guanidinylation we thought the natural product araiosamine C could be obtained immediately after exposure of 18 to TFA. Indeed, we observed a clean conversion to a product showing the mass of araiosamine C. While we were planning our celebration, misery beset us again, but not without company. The NMR spectrum did not match that of the natural product. It was actually the elimination product enamine 19. Hoping to cyclize the guanidine though enamine-iminium equilibrium, we subjected 19 to various acidic conditions. However, no reaction took place. This intermediate’s inertia was confusing. We also attempted cyclization by mesylation of the anomeric alcohol of 22. In addition to the enamine product 23, we unintentionally choreographed an indole dance (see 24)! Apparently, the pesky neighboring group participation happened again, but this time at another position. With this result, we finally came to the conclusion that cyclization via an iminium intermediate would not be possible, because the guanidine could not outcompete the anchimeric indole.
At this time, an idea of ring-chain tautomerization between cyclic hemiaminal 25 and acyclic aldehyde 26 emerged. We proposed that a carefully controlled Boc-deprotection of 22 would equilibrate to araiosamine A. After a discussion on the morning of once de Mayo, 2016, I said to Ming, “maybe today we could actually make the natural product.” But both of us weren’t too optimistic, because we hoped so many times, and were subsequently disappointed. Again, the result was frustrating, as Boc-deprotection in TFA/DCM induced instant dehydration to give again the enamine 19. The hemiaminal ring was not opened to allow equilibration to araiosamine A.
Maybe Boc-deprotection in an aqueous acidic environment would suppress the dehydration. Thus, in another attempt the deprotection was performed in TFA-MeCN-H2O (1:5:4) at 90 °C. The LCMS showed a very complex mixture. Nevertheless, I still took the crude NMR spectrum, which looked hopelessly complicated. I thought it must be as usual that some isomers had the same mass as natural product but their structures could never be identified. When coming back from the NMR lab and comparing the NMR spectrum with that of natural araiosamine A, I was completely surprised—we made the natural product! (Later we found the crude pruduct was a mixture of three interconvertible compounds, 25, araiosamine A and epi-araiosamine A). Although it was around mid-night, I immediately called Ming to tell him this great news. He drove back to the lab, and we sent Phil an email together. I was too excited to sleep on that night. At 6 am we met Phil in his office. Phil said, in order to confirm it was the natural product we needed to scale up the reaction and get a 13C-NMR spectrum. During the scale-up, the hydrolysis product 22 was originally planned to be isolated before deprotection. Unfortunately, the hydrolysis reaction ended up with being heated up to 90 °C by accident (Another completely different reaction was planned to be performed at 90 °C at the same time. But I was too excited, and heated up this hydrolysis reaction mixture by carelessness). Again, the LCMS showed a major product of dehydration. We were not quite sure if it was the natural product or enamine 19. Meanwhile, the group had been waiting outside the NMR lab for celebrating our success. Much to our relief, we got a clean 1H-NMR spectrum of the dehydration product, which completely matched that of araiosamine C! (The mechanism of this cascade transformation is detailed in our manuscript) With some luck in the final step, we made the entire family of araiosamines in one pot. Employing the Ellman auxiliary, we also achieved the asymmetric synthesis of araiosamine C to establish the absolute stereochemistry of the natural product. Additionally, in stark contrast to the initial isolation report, we have found that these molecules are actually potent broad-spectrum antibacterial agents. This is a rare example of a natural product synthesis enabled discovery of bioactivity after the isolation chemists explicitly stated that this class of alkaloids had no observable activity!
Lastly, I would say, without the accident in the final step, we would have definitely made araiosmaines C and D after isolation of araiosamine A and epi-araiosamine A, and subsequent subjection to dehydration. However, as 11-step syntheses have been trending in our lab, we were happy to keep it that way.
Note from Ming: I remembered vividly during my first day of graduate school when Phil described to me the remarkable similarities between araiosamines’ ring-chain tautomerization to that of carbohydrates. These alkaloids can be viewed as having an outward experience of “sea sugar”, though their structures are way more mystifying compared to fellow “sea salt”. This “sugar coating” was rather deceiving, enticing me to this “sweet-looking” project which turned out to be a bitter pill at the outset.
Our earliest effort in the synthesis attempts to make araiosamine through direct trimerization of indolylacetaldehyde imines/enamines. This aldehyde which rapidly polymerizes in its neat form was later referred to by my colleagues as the “mingaldehyde”. It gave me an early exposure to “interdisciplinary research”: I would be making polymeric materials together with small molecules; the polymers thus produced have translational potentials from bench to the coal-tar industry. Together with Julian Shaw, an extremely talented visiting student, we surveyed an assortment of indolylacetaldehyde surrogates (discussed in the SI). We gained valuable insights on the reactivity and stability of various indole building blocks which would find use in our later efforts. But the end result of trimerization studies may be presented in a highly similar fashion as this legendary publication:
Thus, half a year into the project, we decided to target the chain topology of araiosamines, embarking on what we dubbed as the “cyclic logic” in the paper. Admittedly, the amount of black tar I produced every day was dramatically reduced but extraneous functional groups present a significant hurdle. This was when we decided to combine lessons from all these approaches and formulate a new strategy to araiosamines. Having worked solo for a while after Julian’s departure, I was fortunate to be joined by Marc.
I am amazed that DDQ was able to cleanly form that C-N bond in the presence of 2 other indoles. My experience with Yonemitsu has not been so kind. I noticed in the paper that you performed this reaction in the presence of a secondary alcohol (forming compound 28). I wonder if that compound was epimerized at the hydroxyl, would you instead obtain the tetrahydrofuran. Nice work.ReplyDelete
Thank you for the comment. Yes, we are also surprised by the selectivity, especially because in our previous strategies we are so cautious about protecting the other indoles. Unfortunately, we have not obtained Aldol products that's epimeric at the alcohol but yes, there's a possibility that a competing THF ring formation may occur. But what we do know is that if we carry the C2 epimer of the Aldol product through the same sequence, we observe similar selectivities for the DDQ oxidation. There's no over-oxidation in either case.Delete
Thanks for your comment and interest in our work! The selectivity (and no over-oxidation) of DDQ oxidation may be explained by the formation of charge transfer complex between DDQ and the only indole, which has the benzylic CH2 group (C6 in compound 27), and therefore has the least steric hindrance compared with the other two indoles. The two aldol products have the same stereochemistry of OH at C3, but different at C2. Both can be interconverted by epimerization under basic conditions (however, significant retro-aldol reaction also takes place). Another stereochemistry of OH at C3 was never observed in the aldol reaction. DDQ oxidation of another C2 epimer (synthesized from another diastereomer of aldol product following the same reaction sequence, but in lower yield compared with the desired C2 isomer) was much slower, and gave significant amount of compound 28 resulted from epimerization at C2 under DDQ oxidation conditions.Delete
I wonder, guys, is your synthesis robust enough to make new derivatives of the compound? For example having chlorine instead of Br or azaindole instead of indole etc.. I mean, one of the reason for the total synthesis (at least mentioned in literature) is to make derivatives of such compound, however, Ive seen only one example of something like that is Meyers et al, Nature 533, 338–345.ReplyDelete
Or is there even a reason to do something like that?
We believe the synthesis should be quite amenable to different indole scaffolds-after all the bromoindole we used is pretty electron rich and reactive but have survived most of the synthesis in their unprotected forms.Delete
Thank you very much for your comment. Actually, desbromoindole was used as a model substrate until the carbon skeleton was successfully constructed. The DDQ oxidation also worked well with desbromoindole in a dimer substrate. In the later stage synthesis, the electronic properties of indole do not play an important role in the reactivity (in one case the indole protection prevented the neighboring group participation. However, the SN2 reaction still did not go due to steric hindrance. See SI). So employing the current route to make analogs with different indole derivatives should be possible. The only concern is the oxidation potential of other indole derivatives in the DDQ step.ReplyDelete
did you finish the synthesis with desbromoindole? If yes, did it have any biological activity? Im not sure how much of the compound you got (didnt have time to read the article in depth) but what is the possibility of running something like suzuki coupling on those bromines? Its probably more into medchem but might be interesting tooDelete
After successful construction of the carbon skeleton, we always used bromoindole for the later studies because we needed to compare the spectral of synthetic products with those of isolated natural products. The last step reaction was performed in a 10 mg scale. But it would be possible to scale up further if required. It is also possilbe to run coupling reactions, but I would suggest doing that on some intermediates such as compound 32, because the natural products are extremely polar and would be difficult to handle.Delete
Its a valuable content shared,would like to know more about it.
Thanks for your comment, Amisha. Please do not hesitate to contact us if you have any questions.Delete
Hi, I was interested by your Schwartz reagent reaction as it seems slightly voodoo... did you see batch variability with your Cp2ZrHCl and/or did you try an in situ generation using the Sniekus method (Cp2ZrCl2 and LiAlH(OtBu)3)? Can you comment on why trace EtOAc causes epimerization? & did you have any other ideas on how to improve the quench, obviously TLC plates worked well enough for your needs but it's clearly not ideal - did you use silica scraped from the TLC plates when you saw epimerization with silica powder and again was there any batch variability or were you just using one batch of plates?ReplyDelete
thanks in advance
Hi Sgbadger, thank you very much for your interest in our work. Indeed, the Schwartz's reagent reduction was extremely tricky, and we spent a lot of effort before finding the optimal procedure. We used the commercial Schwartz's reagent (mostly from Alfa Aesar and Strem Chemicals), which worked well and gave reproducible results without batch variability. Additionally, we did observe slightly lowered yields after storage of the reagent for several months. We did not try the in situ generation of the Schwartz's reagent, because the aluminum species might induce the epimerization of the aldehyde based on our experience of reduction with DIBAL-H. The reason for the epimerization caused by EtOAc is not clear, presumably due to the interaction between the reactive N-metalloimine intermediate and EtOAc.Delete
I agree with you that quenching by TLC plates was fine for multi-gram scale reaction, but it is not ideal for further scale up. In our case, we did not observe any batch variability of the TLC plates. The silica gel was scratched and eluted with EtOAc/hexanes. I think the successful quenching by TLC plates may be attributed to the excellent heat transfer and quick solvent evaporation on the TLC plates to avoid formation of by-products. For larger scale reactions, I would suggest quenching with silica gel powder at -78 °C before immediate evaporation of the solvent at this temperature.
Thanks, very interesting observations!ReplyDelete
p.s. I would think LiAlH(OtBu)3 has a marked difference in reactivity to DIBAL-H, but may still have been unsuitable for your purposesDelete
The epimerization is most likely from the work-up. I think quench of LiAlH(OtBu)3 and subsequent work-up may also induce epimerization, which is similar to the case of DIBAL-H.Delete