Azaphiles, unite! Azaphobes, run! It’s time to delve into the synthesis of tetrazoles (review) in this two-part post. Whatever you think you know about making these hard-core heterocycles, I’ll wager you’ll still find something interesting here. Are you thinking that all tetrazoles are made with azide chemistry and thus you’ll lose a limb to an explosion? Read on.
Full disclosure: I love molecules with several nitrogens in a row. My graduate work involved adding organometallics to azides to produce triazenes and a good deal of my group’s research at the University of Michigan involved azides, triazoles, triazolines, etc. Good times.
When considering tetrazoles, most organic chemists will likely think first of 5-substituted tetrazoles, long known to be good carboxylic acid biomimetics/bioisosteres in the drug discovery endeavor, their acidity being much the same as those of carboxylic acids (see Meanwell’s excellent review). But there are other important types of tetrazoles that are nonacidic, e.g., disubstituted tetrazoles and fused bicyclic tetrazoles.
In the current post, we’ll consider making 5-substituted tetrazoles using azide chemistry. Online IR studies by the process R&D group at Merck Frosst are instructive in dealing with the dangers of hydrazoic acid.
In the next post, we’ll look at the tetrazole-azidoazomethine equilibrium for the synthesis of disubstituted and fused-bicyclic tetrazoles, where we’ll find that one can avoid sodium azide and hydrazoic acid issues entirely. Well, almost entirely.
5-Substituted tetrazoles – dealing with the dangers of hydrazoic acid
A popular procedure for making 5-substituted tetrazoles is the reaction of sodium azide with a nitrile, often in the presence of an ammonium salt. The example shown below is from Organic Syntheses (Novartis Process R&D and Ley’s group at Cambridge), providing the useful enantiocatalyst shown on an 80 mmol scale. The excess sodium azide was destroyed with sodium nitrite and sulfuric acid, which converts hydrazoic acid into nitrogen and nitrous oxide gases.
While the above procedure may be popular, any time you use sodium azide you should be thinking, “hydrazoic acid can be generated, it’s explosive and toxic, and I need to take the appropriate safety precautions.” That’s precisely what happened during some recent process R&D work at Merck Frosst on the steroyl-CoA desaturase inhibitor MK-8245. The discovery chemistry route used NaN3/pyridinium chloride as shown below, but the process group felt that the potential for significant amounts of hydrazoic acid generation was too high.
Armed with the ability to detect hydrazoic acid in the headspace above the reaction mixture using online IR, the Merck Frosst researchers surveyed alternatives. Sharpless’s zinc bromide procedure, proposed to minimize hydrazoic acid formation by control of the pH, led to a reading of 2000 ppm of HN3 in the headspace, which is below the detonation threshold of 15,000 ppm but was still felt to be undesirable. In their own survey of conditions, the Merck Frosst scientists found something quite new and significant: Reaction with sodium azide in the presence of a catalytic amount of zinc oxide in aqueous THF (pH 8) proceeded efficiently, and most notably, with only 2 ppm of HN3 in the headspace! They were able to make 7 kg of the tetrazole in one run in nearly quantitative yield. Nice!
I’d be remiss if I didn’t mention Bu3SnN3 and Me3SiN3/Cu(I) as sodium azide surrogates, sometimes used on large scale. Shown below is an application to valsartan (see here and here) with recycling of the tin by-products. The intermediate stannyl tetrazole and leftover Bu3SnN3 were converted with HCl to Bu3SnCl, which was then converted to the fluoride, which was removed by filtration and recycled to Bu3SnCl.
Next: The tetrazole-azidomethine equilibrium and its use in tetrazole synthesis.
It’s time for some “oldie-but-goodie” heterocyclic chemistry, namely the Knorr pyrrole synthesis. What’s left to be said about this venerable route to pyrroles? Well, I’d like to point out that it should probably be termed the Knorr pyrrole syntheses (plural). As usual with chemistry, things are more complex than they first appear.
Here’s the upshot: There are two fundamentally different pyrrole connectivities that are produced under the Knorr umbrella. Let’s look at what they are and then delve a little deeper into the connectivity that is often overlooked.
What is the conventional Knorr synthesis?
If you were to consult a reference source or ask someone to describe “the Knorr pyrrole synthesis,” you’d probably find something like this:
The Knorr is usually considered to be the condensation of an α-aminocarbonyl compound with another carbonyl compound, typically an active methylene compound such as a β-ketoester, in the fashion shown above, generally proceeding by (i) condensation of the amine with the other carbonyl compound and (ii) an intramolecular aldol (Knoevenagel) condensation. Using an organizational system that I employ in my heterocyclic chemistry course, it’s a [3+2]-a,c approach, i.e., it makes the bonds at the “a” and “c” faces of the pyrrole by combining a three-atom component and a two-atom component.
Getting specific, one of my favorite examples is Hamby and Hodges’ 1993 work at Parke-Davis, featuring a Weinreb amide approach to the α-aminocarbonyl compound and reductive deprotection to the amine in the presence of the β-ketoester:
Most [3+2]-a,c Knorr examples produce pyrroles with an electron-withdrawing group (EWG) at the 3-position, but there are examples without it.
The Fischer-Fink variant of the Knorr pyrrole synthesis
If you look a bit further into the Knorr synthesis, you find examples that I would classify as [3+2]-a,d variants. Here, the α-aminocarbonyl compound contributes two pyrrole atoms rather than three; the other carbonyl compound (typically a 1,3-diketone) contributes three atoms rather than two. These reactions proceed by (i) condensation of the amine with the other carbonyl compound to form bond a and (ii) an intramolecular aldol condensation to form bond d. If the amine bears two electron-withdrawing groups (as it does in much of the early work), one is lost during the reaction.
This variant of the Knorr synthesis is perhaps best termed the Fischer-Fink variant after the discoverers of traces of these compounds in traditional Knorr syntheses that employ ethyl acetoacetate (H. Fischer and E. Z. Fink, Hoppe-Seyler’s Z. Physiol. Chem. 1944, 280, 290 and 1948, 283, 152 and this link).
It was Kleinspehn in 1955 who figured how to obtain a majority of the [3+2]-a,d connectivity: simply use diethyl malonate rather than ethyl acetoacetate. Paine and Dolphin later improved Kleinspehn’s method by preforming the amine rather than carrying out an oxime reduction in situ. The Paine/Dolphin method is notable in that it works regioselectively with unsymmetrical 1,3-diketones and has found wide application in porphyrin synthesis.
Let’s look at a couple of my favorite examples of the Fischer-Fink variant of the Knorr pyrrole synthesis.
Elliott and co-workers at BioCryst used an α-cyano aldehyde rather than a 1,3-diketone, producing a 2-carboxy-3-aminopyrrole, an intermediate in their synthesis of some PNP-inhibitory pyrrolopyrimidinones (9-deazaguanines):
Prashad and co-workers in the process group at Novartis later used the Elliott chemistry to prepare a related PNP-inhibitor. Both the BioCryst and Novartis groups used cyanamide as the cycloguanidinylating agent to produce the pyrrolopyrimidinones, a method that can be traced back to the work of my early mentor Robert McKee at UNC-CH, the subject one of my earlier posts.
The Fiesselmann-type modification of the Fischer-Fink variant
Okay, that’s a mouthful, but I think it’s worth treating this [3+2]-a,d method separately. In the Fischer-Fink chemistry, a 2-aminomalonate is used, one of the esters being lost in the cyclization reaction. Starting in the 1980s, examples began to appear where a simple glycinate is used; the second ester is omitted. Thus, the condensation of simple α-amino esters with 1,3-diketones or their equivalents produces 2-carboxypyrroles as shown in the following example by Mataka (Synthesis, 1983, 157-159):
As far as I can tell, there isn’t a standard way to refer to these reactions. They remind me of the Fiesselmann thiophene synthesis, which involves the condensation of α-mercaptoacetates with 1,3-diketones and their equivalents, so I’ll call these Fiesselmann-type modifications of the Fischer-Fink variant of the Knorr pyrrole synthesis.
Here are a few more examples of the Fiesselmann-type pyrrole syntheses. Note the complementary regiochemistry of the last two examples.
The greater Knorr synthesis umbrella produces two basic connectivities resulting from [3+2]-a,c or [3+2]-a,d constructions.
Knorr-o-philes please weigh in.
[Jan. 21, 2012: Edited Fischer-Fink citation to include earlier work.]
In principle, the acid-catalyzed cyclodehydration of amino alcohols is an attractive way to make saturated nitrogen heterocycles since it avoids having to make the alcohol into a leaving group such as a halide or sulfonate. And acid is cheap.
Unfortunately, this reaction doesn’t work well for underivatized amines because the amine is converted to a non-nucleophilic ammonium salt.
Enter the Mitsunobu reaction, which works quite well with simple amino alcohols (despite their low pKa) as well as their derivatives (amides, sulfonamides, etc.) The literature is filled with examples of such intramolecular Mitsunobu reactions.
However, if you play your cards right, acid-catalyzed cyclodehydrations can be practical and even preferred over the Mitsunobu method. The rule of thumb is that the alcohol should be able to form a reasonable carbocation. Park and co-workers recently published a nice example that compares these two methods in the context of 2,3-disubstituted indoline synthesis.
An attempted Mitsunobu reaction on the amido alcohol shown gave the indoline in low yield; the major product was elimination of water. Resorting to aqueous HCl in dioxane afforded the desired indolines in good yield with a small amount of elimination. Inversion at the secondary alcohol was observed. Substrates with a secondary benzylic or allylic alcohol worked, whereas those with a secondary alcohol (or an electron-poor benzylic alcohol) failed.
The authors propose a rather implausible nitrogen-complexed cation intermediate and don’t speculate on the why acid catalysis beats the Mitsunobu reaction. I suspect it’s related to leaving group ability and subtleties wherein the amide carbonyl oxygen is involved in a balancing act between assisting the departure of the leaving group by neighboring group participation (for hydronium ion) and acting as an internal acceptor of the benzylic proton in an E2-like mechanism (for oxyphosphonium ion). Your thoughts?
The starting materials were made by a diastero- and enantioselective reaction of the sparteine-complexed dianion shown below. While the authors don’t show it this way, I’ve taken the liberty of drawing the organolithium as internally complexed. It is believed that the benzylic organolithium equilibrates to the most stable diasteromeric sparteine complex, which then reacts with inversion of configuration with aldehydes to give the desired secondary alcohols in good yield with high er and dr. As a reminder, most configurationally stable organolithiums react with retention of configuration, but it’s not a certainty; there are numerous examples where inversion is observed.
In summary, if your alcohol is reasonably capable of ionization, the acid-catalyzed method is worth a shot. Otherwise, the Mitsunobu is generally a nice way to make saturated nitrogen heterocycles. Of course, classic chemistry involving halides and sulfonates is also viable if you don’t mind the extra work.
The aminomethylation of arenes and electron-rich aromatic heterocycles typically involves iminium ion chemistry, i.e., Mannich-type reactions. But when the heterocycle is electron-poor, what then? Recent work offers an attractive approach involving nucleophilic α-amino radicals.
David Mitchell and co-workers at Lilly recently published the scaleup of LY2784544, a JAK2 inhibitor. Their paper is chock-full of interesting chemistry and is highly recommended for a read, but let’s look at just one slice: the installation of the morpholinomethyl group using a radical addition reaction.
In a first-generation approach, the intermediate shown below was subjected to Minisci’s method for radical alkylation. Phthalimide-protected glycine was used as a source of pthalimidomethyl radicals. Around 3.5 kg of the CH substitution product was obtained, but the route was abandoned due to reproducibility problems, relatively low regioselectivity, insoluble by-products, and the need for large amounts of silver nitrate.
In a second-generation route, iminium ion chemistry was explored, but none of the desired material was formed. In their survey of iminium ion techniques, however, the Lilly group found one outlier: Hwang and Uang’s method using N-methylmorpholine-N-oxide and VO(acac)2. In the Uang work, electron-rich arenes such as phenols and naphthols were aminomethylated; there were no examples of electron-deficient heterocycles. Nonetheless, after some optimization, the Lilly group was able use the Uang method to produce the desired aminomethylated material shown above in good yield on a 44 kg scale.
Mechanistically, Mitchell and co-workers believe the Uang chemistry is substrate-dependent. For electrophilic substrates such as the current imidazopyridazine, the reaction proceeds by a radical mechanism involving the addition of a relatively nucleophilic α-amino radical to the pyridazine ring. With electron-rich systems such as those found in Uang’s work, an iminium ion mechanism is probably operational. Further mechanistic work is underway.
For those of you in drug discovery, do you think it would be interesting to carrying out such radical aminomethylations on existing drugs or related compounds? I’m reminded of the recent bevy of direct trifluoromethylation reactions by Baran, MacMillan, and Qing, featured at C&EN and In The Pipeline. Yes, aminomethyl groups and trifluoromethyl groups serve greatly different ends, but direct introduction of the former at rather unreactive sites would seem to be a nice option.
I always get excited when I see an unexpected glowing spot on a TLC plate. Fluorescence! I’ve sent many a graduate student on a wild goose chase to try to identify the source of that glow.
My curiosity about these compounds blossomed during my time in industry, where we worked on fluorescent compounds and quenchers of fluorescence. Adding to the fun was the fact that most small-molecule fluorophores (as opposed to materials like quantum dots) are heterocyclic compounds.
Let’s look at a current application of fluorescent heterocycles and then a bit of chemistry.
Check out this recent TEDMED talk by Quyen T. Nguyen, M.D., a Professor of Surgery at UCSD, who gives an entertaining talk on the use of fluorescent dyes to aid in finding and removing every last bit of a tumor. Her collaborator, Roger Tsien, received the Nobel Prize in Chemistry in 2008 for his work on green fluorescent protein (GFP).
In this work, cell-penetrating peptides (CPPs) composed of polycationic sequences of 6-12 consecutive arginines are covalently attached to a fluorophore (Cy5, more on that later). These labeled CPPs can penetrate into pretty much any mammalian cell without requiring specific receptors. To target them to the tumor, they make activatable CPPs (ACPPs). Generalized cellular uptake of the CPPs is blocked by fusing them to a polyanionic peptide domain (featuring glutamates) via a protease-cleavable linker, neutralizing the polycation domain via the formation of an intramolecular hairpin. For specificity, the linker is designed so that it is cleaved by certain matrix metalloproteinases that are overexpressed in tumors. After injection, the ACPPs travel to the tumor, are cleaved by the proteinases, and accumulate in the adjacent tumor tissue. Shine some light on the area and you can see the tumor, including the normally difficult-to-find margins. Great stuff!
What are the fluorescent dyes being used?
Nguyen and Tsien use indocyanines (or simply “cyanines” for our purposes), which have been long-known for their fluorescent properties and are popular for photography, laser printing, non-linear optical materials, and of course labeling biomolecules. They have the typical push-pull electronic structure of many fluorophores; the conjugated system has an electron donor at one end and an acceptor at the other.
Back to Nguyen and Tsien: Take a look here and here for some of their work. For many of their applications, they use Cy5 (originally Pharmacia Biotech, now GE Healthcare), a pentamethine cyanine dye shown in the scheme below, which absorbs at 650 nm and emits at 670 nm in the red region of the visible spectrum. This is well beyond the region of background fluorescence that is found in tissue, so the signal-to-noise is good. They’re also quite bright. Another advantage of these long-wavelength emitters is that the light can travel through quite a bit of tissue.
Structurally, an interesting feature is the presence of the geminal dimethyl groups, which not only dictate the landscape of the conjugated pi-system, they keep the dye molecules from stacking, which would allow energy transfer and thus quenching of fluorescence. One can vary the number of methine groups in the central conjugated region to tune the wavelengths of absorption and emission (Cy3, Cy7, etc.)
How are cyanine dyes made?
The venerable Fischer indole synthesis is used to make the 2,3,3-trimethyl-(3H)-indole starting materials. N-alkylation with an alkyl halide produces a salt, which is used in an aldol-like double condensation reaction with 1,1,3-trimethoxy-2-propene, an equivalent of malondialdehyde.
If you would like to learn more about the chemistry of cyanine dyes, take a look at Mojzych and Henary’s review on the synthesis of cyanine dyes, or if you can’t access that, download the thesis of Jamie Gragg, who has a nice historical overview of the discovery and development of cyanines.
If fluorescent heterocycles interest you, check back here for occasional posts; there will be more!
Toward the Ideal Reaction – Part 2
In Part 1, the concept of an ideal reaction for biomolecule labeling was outlined: One step, no protecting groups, fast, efficient, specific, no catalyst or reagents, no external stimulus, and easy cleanup. Jäschke’s work on oligonucleotide labeling via the inverse-demand Diels-Alder cycloaddition of tetrazines with norbornenes was featured.
Let’s look at some other chemistry and invite an expert guest, Jack Hodges, to chime in. He also has a few things to say about big pharma versus small companies.
What about click chemistry? Well, about that copper…
When one thinks of simple addition reactions for biomolecule labeling, click chemistry comes to mind. Sharpless’s copper-catalyzed dipolar cycloaddition of azides with alkynes to produce triazoles is the de facto standard click reaction, but, well, it uses copper. Not ideal. Plus, the copper may cause problems in the applications we’re focusing on.
Naturally, ingenious chemists have jumped on that problem.
Copper-free click reactions – dial in some strain, maybe change the dipole
How do you speed up dipolar cycloadditions to alkynes without using copper?
You can fool around with the electronics of the alkyne, but then you run the risk of making it susceptible to unwanted reactions.
Alternatively, it’s well-known that dipolar cycloadditions are faster with strained dipolarophiles. Bertozzi harnessed this effect for click chemistry via cyclooctynes (so-called SPAAC reactions, for strain-promoted alkyne-azide cycloadditions). Others have joined the fray.
Here’s an example of what we’re talking about, using van Delft’s bicyclononynes (BCNs):
Insights from a guest expert: Dr. Jack Hodges
To give us an inside look at the state of the art in this area, we have with us my colleague Jack Hodges, ex-WLPD/Pfizer, who now leads the chemistry effort at Berry & Associates, Inc. (B&A), a firm that specializes in nucleic acids chemistry. Some of you may recall Jack’s prominence in the early days of combinatorial chemistry, where he published seminal articles on the use of polymer-supported scavengers. He has recently negotiated a scientific and business relationship between B&A and SynAffix, B.V. to provide a line of SPAAC reagents for the oligonucleotide field that are derived from van Delft’s BCN.
Will Pearson: Jack, what are some of the challenges of using click chemistry in the oligonucleotide field?
Jack Hodges: There are plenty of examples of people who have used Cu-catalyzed Click reactions on oligonucleotides but there are also other examples where the Cu(I) catalyst has been reported to cause partial degradation of the oligo. The partial degradation paper we always note in our product literature is Kanan, M.W., et al., Nature, 2004, 431, 545-9. There are probably other similar reports. This problem doesn’t seem too surprising to me since both Cu(I) and Cu(II) can coordinate to heteroatoms, and nucleic acids are full of heteroatoms. Whether or not you believe Cu(I) is a serious problem around oligos, my feeling is that if you don’t need it, why bother? There are numerous recipes for making Cu(I) in-situ for traditional Click reactions and often it takes a fair amount of investigation to find the one that will work for your application. The Cu-free approach does away with all that. Just put the strained alkyne and the azide together and the triazole forms. The old fashioned expression is “dump and stir”. Somehow “dump and stir” doesn’t have as nice a connotation as “Cu-free Click” or the sound of the quirky acronym “SPAAC” (short for strain-promoted azide/alkyne cycloaddition), but in practice it is just that simple.
WP: For copper-free click reactions of azides with alkynes, there are quite a few competing techniques. Would you please tell us about your path to selecting van Delft’s BCNs, giving us a little perspective on the various technologies that are available? What made BCNs your top choice for commercialization? (B&A also sells compounds based on Schultz & Pigge’s MFCOs.)
JH: There are two things that attracted us to BCN. First the BCN synthesis looks attractive from a commercial standpoint. It starts with 1,5-cyclooctadiene (which is about as cheap an 8-membered ring compound as one can buy) and requires relatively few synthetic steps, each of which look to be scalable. Second, BCN has the lowest calculated LogP value among the other popular cyclooctynes in the literature. This makes BCN desirable in terms of maintaining the viable biological properties for oligos and other biological macromolecules to which it is attached. DIFO is pretty close to BCN in terms of LogP but it is much harder to synthesize. To be fair, BCN-OH is still hugely more lipophilic than propargyl alcohol. So if you absolutely need to avoid lipophilicity, maybe the traditional Cu-catalyzed Click is your best bet.
WP: What should we know about IP and licensing in this field?
JH: So far as I am aware, the only issued US Patent for Cu-free Click reagents and methods belongs to UC Berkeley (US 7,807,619). There are other published patent applications that may well lead to additional patent coverage in this field. It seems pretty clear that SPAAC reagents and methods fall outside the broad coverage of the Scripps patents that cover Sharpless’s Cu-Catalyzed Click methodology.
WP: The classic ways to attach things to oligonucleotides involve amine acylation or thiol alkylation. Besides click chemistry, what other bioorthogonal conjugation methods do you believe are useful for the modification of oligonucleotides?
JH: Another Click reaction, oxime formation, is gaining steam with oligos. B&A now sells some oxyamine and aldehyde reagents that can be incorporated during oligo synthesis or via post-synthetic modification. Based upon their current popularity, we will be expanding our oxyamine product offerings.
WP: While I’ve got you here, is there anything you’d like to say about your experience of working in a small chemistry company versus your time in big pharma?
JH: On the whole both experiences have been very positive. I had a great run in the pharma industry when it was still a vibrant place for doing top notch research. For most of my 22-year career in pharma, drug companies embraced the financial risk of drug discovery. That made it fun to be a chemist working in big pharma. I don’t think the same fun is there today because drug companies got too big via mergers. As they grew, the lower productivity of research staff that occurred while they were distracted immediately following a merger (when there is an inevitable re-evaluation, repositioning, and reorganization of research endeavors) somehow became part of the justification for the industry to reduce the overall amount of research investment. To me, pharma companies today have all caught the same disease. They all seem downright fearful of research investment. To the scientists in the trenches, it can appear as if their management puts as high a priority on trying to minimize risk as it does trying to discover drugs. Perhaps Bruce Roth’s recent article in C&E News says it better than I do here. (WP: Links here and at Chemjobber here; see also Firestone’s perspective here and at In the Pipeline here.)
Today, I am very happy working in a privately-held specialty chemical company. The ‘you eat if you can sell it’ attitude of small company life certainly keeps a scientist’s daily activities interesting, especially in a world where foreign competitors seem to always have a lower labor cost. This environment drives Berry & Associates to work on the most difficult problems of our customers. There we can provide products of tremendous value to cutting edge biological technologies. To highly educated and experienced chemists, such work is as fun as it is profitable.
WP: You’ve been working with heterocycles for most of your career. Looking back so far, what accomplishment in that field do you feel is most significant?
JH: Ten years later, I still get inquiries about my lithiooxazole paper that was published in JOC in 1991. It didn’t feel like super-remarkable chemistry at the time but people still read that paper. Editorially speaking, I guess that is a darn good reason to publish results that initially surprise you. You can save others from the same surprises!