Dennis
Gillingham
* and
Na
Fei
University of Basel, St. Johanns-Ring 19, Basel, CH-4056, Switzerland. E-mail: dennis.gillingham@unibas.ch; Fax: +41 (0)61 267 09 76; Tel: +41 (0)61 267 11 48
First published on 13th February 2013
Catalysed X–H insertion reactions into diazo compounds (where X is any heteroatom) are a powerful yet underutilized class of transformations. The following review will explore the historical development of X–H insertion and give an up-to-date account of the metal catalysts most often employed, including an assessment of their strengths and weaknesses. Despite decades of development, recent work on enantioselective variants, as well as applying catalytic X–H insertion towards problems in chemical biology indicate that this field has ample room for innovation.
Dennis Gillingham and Na Fai | Dennis Gillingham (left) was born in Newfoundland, Canada. He studied at the Memorial University of Newfoundland with Graham Bodwell, receiving a BSc in 2001. After completing a PhD at Boston College with Amir Hoveyda, he moved to the ETH Zürich for a post-doctoral stint in Don Hilvert's lab. In November 2010 he took up his current position as an assistant professor at the University of Basel. His research interests broadly encompass new synthetic methods, but with a particular focus on nucleic acids. |
Na Fei (Right) was born in Rizhao, China. She received her Bachelor's degrees in Chemistry from Shandong Agricultural University in 2008. Later she joined Shaozhong Wang's group at Nanjing University for her MSc studies. She started her PhD work in Prof. Dennis Gillingham's group at the University of Basel in 2011, where she is developing new methods for the alkylation of nucleic acids. |
Key learning points1. X–H insertion is an underexploited process with great potential for further development.2. Metal–carbenoids are less reactive than is typically believed and are compatible with a variety of common functional groups. 3. The availability of diverse ligand classes has reinvigorated the study of X–H insertion. Many metals behave poorly or are completely inactive without the right ligand. 4. Recent breakthroughs in copper- and iron-catalysed X–H insertion should stimulate further development with these metals. 5. X–H insertion can be carried out in water, opening the door to applications in chemical biology. |
Scheme 1 General scheme for XHI. |
Given its long history, it is surprising that XHI remains underdeveloped. After the seminal observations3 and the subsequent demonstration by Merck that XHI could be used industrially (see panel D in Fig. 1),4 it would seem logical that all aspects of this reaction would have been quickly refined, yet its development throughout the 80's and 90's was sporadic. The predominant focus in the carbenoid field was instead cyclopropanation and C–H insertion. A major reason for the sluggish development of XHI is likely that there are good ways to achieve the substitution of polar X–H bonds through classical uncatalysed nucleophilic displacement reactions. Over the past two decades, however, the catalysis field has proven in countless cases that the properties of templating and turn-over offered by a catalyst can deliver benefits that no stoichiometric process can. Tuning catalyst structure can offer the prospect of controlling chemo-, diastereo-, enantio-, regio-, and site-selectivity. A good catalyst serves as a template and chaperone for the substrates, guiding them over a single slice of the reaction hypersurface. Therefore, good stoichiometric methods to achieve a certain reaction do not diminish the importance of catalyst development. Chemists today are working with larger, more diverse, and structurally more complex molecules than ever before. Achieving selective reactions on these large molecules presents a task in chemoselectivity not likely to be overcome through uncatalysed processes. Progress in the XHI field on this front, i.e. tackling catalysis in complex systems, is discussed in the last section on applications to chemical biology.
Fig. 1 Seminal observations (panels A–D) and a recent application (panel E) in XHI chemistry. |
A number of excellent reviews have appeared on the topic of XHI but the comprehensive ones are more than a decade old,5 and the recent ones tend to focus on isolated aspects of the reaction.6 Furthermore, applications of XHI geared towards chemical biology have never been covered. This review will introduce XHI by referring to seminal studies and thereafter give a whistle-stop tour through the field, discussing only the highlights. Examples will be limited to XHIs where X = N, O, and S. This exclusion is not meant to dismiss the importance of other XHIs, but the focus of the present review is on education, the chosen examples are meant to be representative of the field. The reader is referred to the older reviews for a comprehensive look at the historical development of XHI and the full variety of potential insertion processes.5
Scheme 2 General mechanism for XHI. |
Difficulties notwithstanding, a number of experimental and computational studies lend insight into the rhodium(II)-, copper(I)-, and iron(III)-catalysed reactions. These will be discussed in turn in the following paragraphs. For the remainder of the present review, unless explicitly stated otherwise, any reference to step A, B, C, or D refer to the steps shown in Scheme 2.
Scheme 3 Mechanistic analysis of copper(I) catalysed XHI reaction. |
As mentioned earlier, it seems a number of the transition states in the elementary steps for XHI with copper complexes are close in energy. Fructos and coworkers have discovered an example of how this can be exploited to control selectivity in copper–carbenoid transfer.17 A common problem in metal–carbenoid reactions is a rapid dimerization of the diazo starting material. However, they have shown that if the catalyst is the copper(I)-N-heterocyclic carbene complex shown in Scheme 4 no reactivity towards ethyldiazoacetate (EDA) is observed (top line in Scheme 4). Only when both substrates are present does a rapid consumption of EDA ensue, leading to the expected XHI or cyclopropanation products (bottom line in Scheme 4). An explanation for these observations might be that coordination of the substrate lowers the energy of the transition state leading to copper–carbenoid. This is a stark contrast to most other catalysts, where coordination typically inhibits carbenoid formation. Although no further mechanistic studies have been reported on this system, these observations highlight the ability of ligands to dictate the course of catalytic reactions.
Scheme 4 Controlling copper catalysis with ligands. |
Scheme 5 Kinetic analysis of the effect of inhibitors on rhodium-catalysed reactions. |
Although they have never been observed, the relevance of ylides in rhodium(II)-based XHI processes (step C in Scheme 2) can be inferred from examples where the putative ylide intermediate has been trapped with imine electrophiles19 or even coaxed into rearrangement pathways rather than the [1,2]-proton shift shown in step D of the mechanistic scheme.20 These experiments not only provide evidence for the existence of ylide intermediates, but also suggest that, in line with computational results,14 the [1,2]-proton shift has a surprisingly high activation barrier.
As shown in Scheme 6 the iron(III) complexes can deliver N–H insertion products with up to 1000 turnovers in a matter of minutes. Intriguingly, EDA in the presence of the iron catalyst is completely unreactive. Only when the amine is added does a rapid evolution of nitrogen occur. In addition other classes of substrates that typically participate in reactions with metal–carbenoids, such as olefins or alcohols, are either completely unreactive or many orders of magnitude slower than the N–H insertion process. The possibility that the nitrogen-containing substrate is simply an ancillary ligand that promotes rapid carbene formation seems unlikely given that competition experiments where olefins and amines are present lead almost exclusively to N–H insertion.21 Gross contends that this suggests a carbenoid-free pathway involving direct nucleophilic displacement by the amine of the dinitrogen leaving group on an iron-bound EDA molecule. On the other hand, Woo has observed that homo-coupling of EDA is accelerated in the presence of Lewis basic amines, and this would seem to support an amine-accelerated carbenoid pathway.22 Further experiments are needed to clarify this mechanistic dichotomy.
Scheme 6 Iron(III) corrole and porphyrin complexs display unique reactivity consistent with a non-carbenoid pathway, or a carbene pathway that is activated by the presence of amine ligands. |
Fig. 2 Metals for carbenoid-transfer reactions. The font scaling is meant to qualitatively convey the effectiveness of each metal. |
Rh2(OAc)4 was first used by Teyssie in 1973 to decompose EDA to produce a rhodium–carbenoid intermediate, subsequent O–H insertion then delivers ethers.11 Later they reported that rhodium carboxylates were also efficient catalysts for N–H and S–H insertion.24 These seminal observations set the stage for forty years of intensive study with rhodium(II) catalysts. As a result the substrate scope has been widely expanded to include aliphatic amine, aniline, amide, alcohols, phenols, thiols and silanes (see Table 1).3
Metal | Strengths | Weaknesses | |||
---|---|---|---|---|---|
a The narrow substrate range with these metals may simply be a reflection of underdevelopment. | |||||
Rh | • High turn-over numbers and frequencies | • Expensive | |||
• Broad substrate scope | • Competing C–H insertion and β-elimination | ||||
• Biomolecule modification | • Moderate stereocontrol | ||||
Cu | • Cheap | • Susceptible to inhibition by Lewis bases | |||
• Chemoselective | |||||
• Enatioselective variant available | |||||
• Reactivity highly sensitive to ligands | |||||
Rua | • Chemoselective | • Expensive | |||
• Rich coordination chemistry | • Narrow substrate scopea | ||||
Fea | • Cheap | • Narrow substrate scopea | |||
• Low toxity | |||||
• Reactivity highly sensitive to ligands | |||||
• Enantioselective variant available |
Attesting to the practicality of rhodium(II) systems is the Merck synthesis of thienamycin shown in panel D of Fig. 1. Nevertheless, despite such early promise, rhodium(II) chemistry has not been a panacea for XHI applications. The seeming preference for a metal-free ylide pathway (see Fig. 3) has thwarted attempts to create enantioselective processes. Additionally the facility with which rhodium carbenoids undergo C–H insertion29 and β-elimination30 is a problem with certain substrates.
Fig. 3 Comparison of the metal-associated and metal-free transition states for proton transfer in copper- and rhodium-catalysed O–H insertion. *Note that the energy of the rhodium-associated species has been adjusted to place it on the same energy scale as the copper structure. |
The above mentioned issues with rhodium have led to a renaissance in the application of copper catalysts to tackle current unsolved problems in XHI. Although first reported to be useful for XHI reactions in the early fifties,7,8 XHI chemistry with copper remained underutilized until recently – likely as a result of the harsh reaction conditions, low insertion yields, and sparing solubility of the copper complexes. Research with copper was therefore largely abandoned for rhodium(II) catalysts. In 2002, a report from Pérez and co-workers demonstrated that copper(I) complexes with homoscorpionate ligands could catalyse the insertion of EDA into N–H bonds of amines and amides in high yields under mild conditions.31 The electronic interaction between the copper and the heterocyclic ligand not only enhanced its stability, but also improved its reactivity and selectivity in the XHI reaction as a result of its unique structure. The great strides made in catalysis over the past fifty years are leveraged largely on the development of new ligands that can tune the properties of its bound metal. The report that the homoscorpionate ligands can improve copper(I)–carbenoid chemistry served to remind the community that these catalysts had a great deal of unrealized potential. Since then developments with copper(I) have been rapid and substantial.6,25,32
Ruthenium, one of rhodium's direct neighbours in the periodic table, was first introduced to catalyze ethyl diazoacetate insertion into S–H and N–H bonds in 1997.33 Recent results on ruthenium catalysed N–H34 and O–H insertion35 reactions, reported by the Che and Lacour groups respectively, demonstrate that ruthenium complexes can sometimes offer unique reactivity in comparison with other catalysts. Although a relative newcomer in XHI, the favourable properties of ruthenium (its similar reactivity to rhodium, lower cost, more available oxidation states, and rich coordination chemistry) suggest a bright future.
Given the costs associated with the rarer transition metals, processes based on iron would be well-received by potential users. The Woo and Gross groups have independently shown that iron(III)–corrole and iron(III)–porphyrin complexes are excellent catalysts for N–H insertion into a variety of amines and diazo substrates.21,36 Furthermore, the recent application of iron–spirobisoxaline complexes in highly efficient enantioselective O–H insertion should stimulate further developments with this practical alternative to the precious metal catalysts.37
The substrate range and specificity of rhodium(II)5 and copper(I)38 systems have already been reviewed or are covered in other sections of this review. Therefore the remainder of this section will simply give an overview in tabular format of the breadth of substrates and reaction types each catalyst has so far been applied to (see Table 1).
Whatever the mechanistic rationale, chiral copper complexes are currently the best choice for achieving a variety of enantioselective XHI processes. Shown in Table 2 are the ligand systems that have proven most effective thus far.6,25–28 From their complex structures it is clear that one of the drawbacks with current methods is that the ligand syntheses are demanding. Before a method can become practical, however, it is first essential that the concept is established, and the systems shown have indeed opened the door to new developments in the area of XHI with copper catalysts.
a See ref. 25. b See ref. 26. c See ref. 6. d See ref. 27. e See ref. 28. | |||||
---|---|---|---|---|---|
Ligand | |||||
Preferred metal salt | Cu(OTf)2 | CuBr | CuCl or Cu(MeCN)4PF6 | (CuOTf)2·PhH | CuCl |
Reactions studied | O–H insertiona | N–H insertionb | O–H,c N–H,c and S–H insertionc | O–H insertiond | N–H insertion (1° and 2° amines)e |
ee range (%) | 20–98 | 80–94 | Typically 60–90 | 50–90 | 70–98 |
The spiro-bisoxazoline ligands developed by Zhou have proven especially versatile in XHI processes.6 A look at the minimized 3D structures and the conformation that would be required for bidentate chelation with copper(I) presents a perplexing picture. In the minimal energy conformer (left-most structure in Fig. 4) the nitrogens prefer to be more than 6 Å apart, precluding bidentate chelation. In the rotomer that would be required for copper chelation there is only 3.0 Å between the two nitrogens (middle structure in Fig. 4). Based on typical Cu–N bond lengths in similar complexes, this space would be a tight squeeze for copper coordination. With most ligands a simple conformational adjustment would readily occur to accommodate metal binding, but the rigid spiro motif in these ligands prohibits this possibility. Recently the Zhou group has put forth an explanation for this peculiarity that involves a bimetallic copper complex which creates a C2-symmetric environment with the help of two ligand molecules (right-most structure in Fig. 4).6 The Cu–Cu distance in the X-ray structure is 2.78 Å, suggesting the possibility of a cooperative role for both metal centres in catalysis. Certainly two ligand molecules are involved since a small but measurable positive non-linear effect is observed. A kinetic analysis of the Zhou system such as the one conducted by Pirrung and Morehead13 for rhodium(II) would be of interest since it could identify whether both coppers were independently active, cooperative, or susceptible to mixed inhibition.
Fig. 4 Right: lowest energy conformer for the Zhou ligand (MM2 minimization). Middle: oxazole rotomers that would be required for bidentate chelation to copper. Right: crystal structure of the bimetallic complex determined by Zhou. |
Despite the preference for the metal-free ylide pathway in XHI, enantioselective processes based on rhodium(II) may yet be realized. A recent report demonstrating that an achiral rhodium(II) carbene can be intercepted by an imine activated with a chiral Bronsted acid indicates that cooperative rhodium(II)–Bronsted acid catalysis may offer an alternative approach towards enantioselective XHI.42 Indeed Saito and coworkers have found that cinchona alkaloid additives can deliver enantiomeric excesses of up to 50% in the O–H insertion of α-diazoesters with water catalysed by achiral rhodium(II) catalysts.41 The Zhou group have also recently discovered that BINOL-based chiral phosphoric acids deliver high levels of enantioinduction in N–H insertion reactions.42 These promising initial findings suggest that XHI coupled with enantioselective protonation is an area poised for further development.
Despite these advances, an area where significant development is still required is in the controlled chemical modification of native biomolecules such as proteins, carbohydrates, and nucleic acids. Bioorthogonal chemistry, first articulated by Bertozzi, is an approach that involves introducing unnatural functional groups into target substrates.43 The selective reaction between these functional groups must not be seen in natural systems, and must not interfere with endogenous biological chemistry. Once the bioorthogonal functional group is introduced into the biomolecular target, the selective chemical reaction can then be used to label, modify, probe, or pull-down the molecule of interest. While this strategy has proven useful, the inherent need to create bioorthogonal functional groups in both reaction partners can be cumbersome. In natural systems a great variety of enzymes act simultaneously on their cognate substrates without interfering with one another; such complete catalyst control is the ideal form of reaction orthogonality. Given the right catalysts and ligands it should be possible for chemists to imitate the kind of selectivity achieved by enzymes, but progress in creating artificial catalysts to selectively target native biomolecular structures has thus far been modest. Nevertheless, much of the progress achieved to date has been with catalysts for XHI.
Fig. 5 Rhodium catalysed tryptophan labelling can be used to modify a variety of proteins. |
Fig. 6 Earliest example of carbenoid chemistry to modify proteins. |
It was many years later before the potential of rhodium(II) complexes to promote XHI reactions for applications in chemical biology was recognized. Antos and Francis reported in 2004 that tryptophans in myoglobin and subtilisin could be modified at low pH using α-diazo esters and dirhodium tetraacetate (see Fig. 5).46 The reaction seems to proceed by a mixture of XHI and cyclopropanation of the tryptophan indole moiety. During this study they also discovered that hydroxylamine hydrochloride was a uniquely effective buffer salt for the reaction. They speculated that its role was to bind the metal centre and stabilize reactive intermediates. In their follow-up study they identified t-butylhydroxyl amine as an even better buffer salt and with it they could carry out reactions over a broad pH range (3–9) with little loss in efficiency or selectivity.46 Again the role of t-butylhydroxyl amine remains mysterious. Their new reaction conditions lead to smooth alkylation of tryptophan residues on a variety of proteins, including a case with FKBP where a tryptophan-containing C-terminal melittin tag was engineered into a protein to allow selective labelling with rhodium–carbenoid chemistry (see Fig. 5). This strategy is conceptually similar to the classical approach of introducing a cysteine into a target protein for subsequent reaction. For the metal-mediated tryptophan labelling to proceed efficiently its side-chain must be solvent accessible. This can be achieved by selecting proteins with solvent accessible tryptophans, or by denaturing with pH or temperature.
While the selective alkylation of tryptophans is an important advance, the Francis system requires large amounts of the rhodium complex and is limited to tryptophans. It would be ideal to be able to modify nearly any desired amino acid side-chain with the site-selectivity controlled by the catalyst. The Ball lab has developed a technique where the reactive rhodium–carbenoid is generated with a rhodium complex bound between two carboxylate residues in a α-helix.47 The interaction of the catalyst containing helix with its partner helix in a coiled coil motif brings the catalyst into a microenvironment completely controlled by the coiled coil registry (see Fig. 7). The local concentration of the rhodium vis-à-vis its binding partner is in the molar range, although its concentration in bulk solution is high micromolar. The rate accelerations achieved through high effective molarity allowed these researchers to target not only aromatic residues, but also nearly every amino acid side-chain that contains a nucleophilic hetereoatom. It is important to reiterate that although many amino acid side-chains can be targeted, selectivity can be precisely controlled through applying the well understood binding rules of the coiled coil motif. When they tested variants where the helical-wheel model predicted a large distance between the reactive side-chain and the catalyst no reaction was observed (see top of Fig. 7). The Ball system leverages proximity induced catalysis as the control element rather than bioorthogonal reactivity.47 Therefore the potential applications of this technique are numerous. In principle modifying any new protein of interest would only require identifying a ligand that binds specifically to the target region in the protein. While this may sound difficult it is important to bear in mind that computational methods are at the level of being able to predict with high degrees of confidence protein tertiary structure and binding interactions.48
Fig. 7 Top: the coiled–coil interaction induces proximity between catalyst and substrate in the matched case, and disposes them away from each other in the mismatched case. Bottom: this proximity-induced selectivity enables the specific targeting of a variety of protein side-chains (adapted with permission from ref. 47). |
Ho and coworkers have developed a ruthenium(II) porphyrin complex for XHI reactions in aqueous media and applied it to the modification of the N-terminus or reactive cysteines in peptides and proteins.49 The large hydrophobic porphyrin core of this complex is rendered water soluble by virtue of four pendant β-D-glucose moieties. The system is highly active, delivering complete conversion in 1 h at biological pH and temperature with micromolar concentrations of both substrates and only 10 mol% of the catalyst. The ruthenium(II) porphyrin is not only effective for XHI but also a variety of other metal–carbenoid reactions. Unlike the iron–porphyrin system it seems that reactions catalysed by this ruthenium–porphyrin complex proceed by a metal–carbenoid intermediate.
Entry | Sequence | Structure | Conva (%) | Product | Significance |
---|---|---|---|---|---|
a Conversion based on the oligonucleotide; the diazo substrate is typically completely consumed when nucleic acids are present. b Diazo substrate also not consumed. | |||||
1b | d(T)4 | Single-stranded | 0 | No reaction | Ts are not alkylated |
2b | r(U)4 | Single-stranded | 0 | No reaction | Us are not alkylated |
3 | d(TAT) | Single-stranded | 20 | N6-A alkylation | As are targeted |
4 | d(TCT) | Single-stranded | 16 | N4-C alkylation | Cs are targeted |
5 | d(TTTATTTGTTTCTTT) | Single-stranded | 37 | Multiple alkylations | Longer sequences can be targeted |
6 | d(CGAACGTTTTTCGTTCG) | 0 | No reaction | Double-stranded sequences are unreactive | |
7 | d(CGAACGTTTTTCGTTCGA) | 21 | 3′ Overhang regions can be targeted | ||
8 | r(CUAGCAUUUUUUGCUAGA) | 33 | Short hairpin RNAs can be targeted |
The dirhodium(II) tetraacetate catalyst we employed was selective for exocyclic N–Hs in the nucleobases, but otherwise delivers unspecific alkylation in single-stranded nucleic acids (compare entries 1–5 in Table 3). In contrast studies on hairpin sequences revealed that double-stranded stretches were unreactive (entry 6). It seems that if the N–H bonds are engaged in Watson–Crick base-pairing they are unavailable for reaction with the rhodium. This selectivity profile was exploited to target certain unpaired regions in DNAs and RNAs (see entries 7 and 8).
The turn-over frequency of the rhodium(II) catalyst was dramatically affected by the nucleic acid sequence. For example while reaction with d(TAT) required 24 h to deliver complete consumption of the diazo substrate, reaction with d(TCT) under otherwise identical conditions led to complete diazo consumption in 3 h. This is not simply an effect of the inherent nucleophilicity of the substrate since a mix experiment with both substrates also leads to sluggish conversion of the diazo compound. Based on the kinetic analysis of rhodium(II) catalysis shown in Scheme 4, these results suggest that either the pre-equilibriums leading up to catalysis, or the αKil or βkcat turnover pathways are affected by nucleic acid ligands. We are currently investigating if and to what extent these possibilities are operative.
The Romo group has recently developed a selection of diazo substrates designed for use in the synthesis of probe molecules in conjunction with XHI.52 They demonstrated that rhodium–carbenoid based O–H and N–H insertion can be used to create functional conjugates with a variety of biologically active natural products (see Fig. 8). Furthermore, the ester side-chains of the diazo substrates included alkynyl groups to facilitate further modification by the copper-catalysed azide alkyne cycloaddition. Shown in Fig. 8 is the collection of natural products that have been demonstrated to be compatible with this technique. Most of these molecules contain multiple potential reactive sites and yet practical yields of the mono-etherification products can be obtained.
Fig. 8 An assortment of complex natural products with diverse functionality can be converted to probe compounds in a single step using Rh-catalysed O–H and N–H insertion. |
Chemical intuition might suggest that metal–carbenoids are too reactive to be exploited in complex environments and/or with protic solvents. Counter to this intuition are the recent developments applying XHI to the modification of intact proteins, nucleic acids, and complex natural products. These results hint at a prolific future for the XHI reaction in fields such as chemical biology and nanotechnology, where selective catalysis in complex settings is essential.
This journal is © The Royal Society of Chemistry 2013 |