Catalyzing decarboxylation by taming carbon dioxide

Introduction

The ability of microorganisms to convert carbohydrates into CO₂ and ethanol has served as the basis for the production of bread, wine, and fermented foods for centuries [1]. The end points of most metabolic pathways that produce energy in higher organisms involve decarboxylases that convert carboxyl groups into CO₂ [2]. The process of decarboxylation is formally an electrophilic substitution process in which a carboxyl group is replaced by an external electrophile, most commonly a proton (Scheme 1).

Scheme 1:

Generalized decarboxylation.

In contrast to those rapid processes that produce CO₂, carboxylic acids in solution are not subject to decarboxylation reactions under ordinary conditions. Where reactions do occur, the reactant has access to a class of mechanism that lowers the barrier to C–C bond cleavage [3]. An example is the decarboxylation of β-ketoacids that react according to the mechanism suggested by Pedersen [4] and enhanced by Guthrie [5] (Scheme 2).

Scheme 2:

The special mechanism of decarboxylation for β-ketoacids.

Thus, the seemingly simple electrophilic substitution of a proton for CO₂ is normally a chemically indelicate proposition, requiring extended heating or specialized reagents [6–8]. On the other hand, the diverse collection of rapid reactions promoted by decarboxylases indicate that mechanisms for decarboxylation are potentially more generally available [9]. The observation of diverse enzyme reactions assures us that the apparent resistance of carboxylic acids to decarboxylation in solution results from a kinetic barrier that enzymes have evolved to overcome, rather than the free energy difference between reactants and products. In a similar manner, enzymes that promote hydrolysis of proteins and nucleic acids overcome the kinetic barrier by providing pathways that do not exist in the normal milieu in which the compounds exist [10]. The possible similarity of the mechanisms for catalysis in hydrolytic and decarboxylation pathways has only recently begun to be recognized [11, 12]. In this review our objective is to illustrate how decarboxylation is hindered by the nature of CO₂ and how that can be overcome by alternatives that achieve the same outcome in terms of the conversion of reactants to products by altering the nature of the precursor to CO₂ in the reaction mechanism.

The nature of investigation of the mechanisms of decarboxylation reactions has been heavily influenced by a common and measurable product being “CO₂” as deduced from comparison of the molecular formulas of the reactants and products (Scheme 1). The product that is not CO₂ is derived from what is left behind as indicated by subtracting “CO₂” from the molecular formula of the reactant. If we then make the logical assumption that CO₂ is one product, all matters of acceleration by enzymes have to do with alterations to the reactant itself or to stabilization of a transition state with the portion of the reactant (intact or partially transformed) from which CO₂ separates [9]. However, our recent observations of catalysis of decarboxylation in solution suggested that the departure of CO₂ itself can be the source of low reactivity that is overcome by alternative pathways that promote separation of the products [12, 13]. With that knowledge, we can consider previously undiscovered routes for catalyzing decarboxylation [12–17].

We realized the possible existence of alternatives to CO₂ itself when we were confronted with observations that could not fit within mechanisms where CO₂ forms directly by cleavage of a C–C bond [11]. If direct formation of CO₂ in such a pathway is problematic, what is the nature of that problem and how can it be overcome by catalysis? Answering these questions provides a framework that leads to an understanding of the overall process in solution and within enzymes.

We began our investigations of decarboxylation in order to obtain quantitative comparisons of enzymic processes and very similar ones that occur without a protein [18, 19]. We reasoned that the more similar the enzymic reaction and nonenzymic reactions are, the more precisely we would know the extent to which the protein enables the reaction to occur with a lower barrier [20–24]. We found that there is a delicate balance in the steps that promote the reaction and that this balance is highly dependent on the specific reactant and the defined conditions in a solution [25]. In examining the mechanism of enzyme-catalyzed decarboxylases, we note that many of the best-studied decarboxylases are cofactor-dependent [9]. This adds a useful commonality to reaction patterns where the same type of covalent intermediate that is derived from the cofactor occurs with reactions promoted by different enzymes. Quantitative assessments are then readily accessed because nonenzymic analogues of the enzymic intermediates can be considered in terms of comparable reactions [12, 13, 26].

Thiamin diphosphate is the cofactor associated with enzymes that promote the decarboxylation of α-ketoacids [23, 27]. The carboxylic acid substrates are completely resistant to decarboxylation in the absence of enzymes. Those enzymes utilize bound thiamin diphosphate to form a covalent intermediate derived from addition of thiamin diphosphate’s thiazolium ring to the keto group of the ketoacid [23, 25, 28–31]. The resulting addition product is much more prone to decarboxylation (Scheme 3). Research in our laboratory focused on producing precise synthetic analogues of those intermediates, allowing a detailed comparison with the reactivity of similar intermediates that are enzyme-bound [19, 32].

Scheme 3:

Decarboxylation of an α-ketoacid involves addition of thiamin diphosphate to the carbonyl group of the substrate.

Our interest had focused on the intermediate derived from thiamin diphosphate and benzoylformate in benzoylformate decarboxylase, mandelylthiamin diphosphate, which is accurately modeled by mandelylthiamin (MTh) [21, 22, 33]. Without the protein, MTh undergoes decarboxylation with a rate constant that is at least 10⁶ times less than the rate constant of its enzymic counterpart (with the value of k_cat as a lower limit) [32].

The comparison of enzymic and nonenzymic rates of reaction has been systematized by the work of Wolfenden and coworkers [34, 35]. They set up standards to be used for comparison of reactions catalyzed by enzymes with rates of nonenzymic conversions of the same reactant to the same product. The favored comparative term, catalytic proficiency, is obtained from the ratio of the numeric value of the enzymic specificity function: k_cat/K_m (M^–1s^–1) to the first order rate constant for the nonenzymic reaction, k_non.(s^–1), giving a value in units of M^–1. According to Miller and Wolfenden, large values are indicative of “more proficient” enzymes: “These values vary from approximately ∼10⁹ M^–1 for carbonic anhydrase to ∼10²³ M^–1 for yeast orotidine 5′-phosphate decarboxylase (ODCase). ODCase turns its substrate over with a half-time of 18 ms, in a reaction that proceeds in its absence with a half-time of 78 million years in neutral solution [36].” The values for catalytic proficiency are expressed in inverse concentration units so the connection to reaction times is indirect. In this case, the nonenzymic reaction is formulated in terms of a direct loss of CO₂, leaving the residual carbanion at room temperature in neutral solution, based on an extrapolation from a reaction at high temperature.

An important contrast of the nonenzymic reaction and the enzymic decarboxylation of benzoylformate, upon cleavage of the C–C bond that produces CO₂ from MTh with no enzyme present, products form that are derived from the irreversible cleavage of thiamin (Scheme 4). In the enzymic counterpart, thiamin diphosphate is maintained intact [22, 32, 37]. The production of CO₂ from MTh leaves the C2α conjugate base of hydroxybenzylthiamin (HBnTh). Delocalization of that carbanion produces the corresponding enolamine, which undergoes fragmentation that destroys the core of thiamin [21, 22, 32]. This produces a phenyl thiazole ketone (PTK) and 2,5-dimethyl-4-amino-1,3-pyrimidine, (DMAP). The nonenzymic conversion of the decarboxylation product to the cleaved cofactor has a rate constant that is 100 times larger than the enzymic k_cat [38]. If the related product from which cleavage occurs would form on the related enzyme, how could the enzyme compete against the destructive reaction? [34, 35, 39]. Enzymes evolve to make reactions go faster, not slower and they control the specificity of the reaction [41]. Those forces of evolution would provide conditions that must also have prevented the fragmentation.

Scheme 4:

The decarboxylation of mandelylthiamin is accompanied by fragmentation of the thiamin core of the product.

In analyzing the products of the nonenzymic reactions of MTh, Qingyan Hu in our laboratory found that addition of Brønsted acids to reaction solutions drives the reaction toward formation of the products that parallel the enzymic reaction, HBnTh. The catalyst reduces the amount of the fragmentation products and favors the formation of HBnTh [32, 33]. We assume that with proximal Brønsted acids available in its active site, the enzyme would easily direct the reaction to the appropriate product. In addition to seeing the product directed to HBnTh by added Brønsted acids, Hu observed that protonated pyridine is particularly effective in producing HBnTh and reducing the amount of fragmentation [21, 22]. Furthermore, she found that protonated pyridine (and it C-alkyl derivatives) also accelerates the overall decarboxylation reaction of MTh. If protonation occurs after the formation of CO₂, it affects the products of the reaction but not the rate of their formation. Any Brønsted acid could potentially trap the carbanion and prevent the fragmentation reaction but it would not lower the barrier to the forward reaction. In order to consider the acceleration by protonated pyridine, we must modify the proposed mechanism in light of the observed increase in rate from protonated pyridine.

Theoretical work of Major and Gao showed that in the case of the decarboxylation of N-methyl picolinate, there is no barrier to the recombination of CO₂ that is formed in the reaction and the immediate residual product [42]. Applying these ideas, we see that if protonated pyridine is able to associate with MTh prior to formation of CO₂, it could trap the carbanion and accelerate the net forward reaction. Sauers, Jencks and Groh considered the carboxylation process for biotin, proposing that CO₂ will react very rapidly with an adjacent anion, which is precisely the situation after the C–C bond cleaves in a decarboxylation reaction [43]. Adapting the above considerations, we arrive at the mechanism in Scheme 5. The transfer of the proton of pre-associated protonated pyridine to the nascent carbanion not only will accelerate the reaction but will also prevent fragmentation of the product. Thus, evolution of catalytic activity coincidentally would preserve the cofactor [38].

Scheme 5:

A pre-associated acid traps the carbanion formed upon C–C bond cleavage.

Exactly how would protonation of the carbanion generated from cleavage of the C–C bond in MTh upon loss of CO₂ lead to acceleration of the decarboxylation? Protonation would occur after the bond has cleaved, which is the only step in the reaction sequence other than protonation. How can a step after the bond cleavage make the rate larger? [33]. First, we assume that the protonated pyridine is catalyzing the reaction by transferring the proton to the carbanion. There is no site on MTh to which protonated pyridine can transfer a proton prior to C–C cleavage to promote the reaction. If the proton transfer occurs after C–C cleavage, the observed rate would be accelerated only if C–C cleavage is reversible. In that case, protonation of the carbanion drives the reaction in the forward direction by blocking the recombination of CO₂ and the carbanion.

As noted earlier, in order for the protonated pyridine to promote the reaction by trapping the carbanion, it must be associated with MTh before the C–C bond breaks. This could be accomplished by π-stacking between pyridine and the aromatic rings of MTh, allowing the proton to be aligned for transfer [44, 45]. With protonated pyridine stacked in a position to provide a proton directly or indirectly to quench the carbanion to prevent fragmentation, CO₂ could be blocked from combining with the carbanion.

This hypothesis can make what Popper designated as a “risky prediction” [46] whose test provides its logical validation or refutation. It is well-established that C–C bond cleavage has an intrinsic carbon kinetic isotope effect (CKIE) [45, 46]. To the extent that the protonation step is also partially rate-limiting, the full CKIE will not be observed [9, 47]. In the presence of protonated pyridine, the proposed mechanism would lead to the separation step being enhanced by the catalytic pathway, making the observed CKIE greater without affecting the intrinsic CKIE. Scott Mundle in our laboratory established a collaborative effort with Professor Barbara Sherwood Lollar. Together, we developed methods to measure the CKIE under various conditions [48]. Mundle’s results established that protonated pyridine does indeed produce an observable increase in the CKIE, consistent with this reaction model and its risky prediction [16]. Objections to this mechanism (based on computations) suggested in one instance that the role of pyridine is to lower the energy of the transition state without any comment on the critical issue of the effect of the catalyst of the magnitude of the CKIE [49]. However, in detailed study of a reaction where the rate of a decarboxylation is increased by up to a factor of 10⁴ in less polar solutions, the CKIE remains invariant [52]. Thus, the variation in CKIE that we observe must arise from a change in the extent of reversion.

The concept of catalysis by pre-association was first proposed where hydrogen bonding of the reactant and catalyst precedes a rate-determining reaction step [51, 52]. It is difficult to see how this type of pre-associative mode would apply to the decarboxylation of MTh. On the other hand, the decarboxylation of carbonate monoesters can be seen as being set up for pre-association of an H-bonded catalyst. Even with strongly basic leaving groups, the formation of CO₂ from carbonate monoesters occurs rapidly and consistently [43].

In searching for other means of pre-association of Brønsted acids with MTh in addition to our proposal of π-stacking of pyridinium ion and aromatic rings, Graeme Howe in our laboratory considered the potential of electrostatic attractions to preassociation of a catalyst. Since MTh is cationic, an anion could pre-associate, allowing an acidic group elsewhere in the molecule to trap the nascent carbanion from MTh. Howe found that malonate monoanion is indeed a catalyst for the decarboxylation of MTh. However, when he tested the same reaction with acetic acid/acetate buffer, he saw acceleration as with malonate [11]. Further examination showed that the decarboxylation of MTh is subject to general base catalysis. The catalytic reaction has a relatively low β value (0.26), indicating that there is a real but small degree of proton transfer in the same step as C–C cleavage [11]. There is no mechanism that we could cite to explain these conclusions in terms of production of CO₂ from MTh in a single step. How could deprotonation of the reactant accelerate C–C cleavage?

One answer is that the appearance of general base catalysis (and subsequent measurement of a significant solvent isotope effect) is indicative of a mechanism involving initial hydration of the carboxyl group. Upon C–C cleavage with transfer of a proton to the Brønsted base, bicarbonate rather than CO₂ is formed when the C–C bond breaks [11]. This would help overcome internal return as bicarbonate is polar and much more soluble than CO₂ [53], with those properties driving the reaction forward. An alternative would have CO₂ trapped by an external nucleophile whose addition is assisted by pre-associated Brønsted bases.

Do mechanisms involving addition to the carboxyl group prior to C–C bond cleavage present a general solution to overcoming the problems associated with formation of CO₂ and reversion? We have found instances where it is almost certain that enzymes accelerate decarboxylation reaction by avoiding formation of CO₂ next to a carbanion. It is especially interesting that evidence for benzoylformate decarboxylase includes reactivity patterns that suggest that this is the case. Miriam Hasson and her co-workers reported that benzoylformate decarboxylase is irreversibly inactivated by benzoylphosphonate-derived analogues of benzoylformate [56]. They expected that the inhibition would be reversible, with ThDP adding to the keto group to form an analogue of the predecarboxylation intermediate. Instead, benzoylphosphonate irreversibly inactivates the enzyme. Analysis of the inactivated enzyme revealed that serine at the active site had been converted to a phosphate ester that which resists hydrolysis, inactivating the enzyme [54].

The authors suggest that serine is essential at the active site and that the phosphonate is able to inactivate the enzyme after being attacked by the hydroxyl group of the residue derived from serine (Scheme 3). It is not explained how this addition would function in the normal catalytic mechanism but it can be readily understood as a parallel to catalysis avoiding CO₂ formation by producing the carbonate monoester through addition of the hydroxyl group to the carboxyl of enzyme-bound MThDP [11, 12]. In terms of the normal substrate, the addition of the serine hydroxyl group to the carboxyl of MThDP would lead to C–C cleavage and avoid formation of CO₂ at the active site. The resulting alternative is the carbonate monoester derived from the side chain of serine. These monoesters decompose rapidly [41], permitting the enzyme to function while CO₂ is released from the side chain of the protein. If serine is not available, abundant water at the active site can serve as an alternative trapping agent.

In serine proteases, the hydroxyl of the active site serine acts as a nucleophile in a well-established global pathway [55]. Those enzymes utilize serine to add to the amide of proteins and form an intermediate that leads to production of an acyl enzyme derivative. This process has many features in common with the assistance that serine would provide in decarboxylation reactions. The serine addition route for proteases could have evolved along with a similar route for decarboxylases.

Dramatic unexpected evidence for the role of serine in enzymic decarboxylation was reported by McLeish, Petsko and coworkers. Benzaldehyde lyase is a ThDP-dependent enzyme that is not a decarboxylase (Scheme 6) [56, 57]. The enzyme becomes an active decarboxylase with benzoylformate as a substrate when serine is in place of an alanine residue at the active site [58]. This decarboxylation-active mutated enzyme is subject to inactivation by benzoylphosphonate through formation of the phosphate ester of serine-26, just as is seen with native benzoylformate decarboxylase (Scheme 3) [54]. This establishes that serine can also serve a critical role in decarboxylation in the mutant as it does in benzoylformate decarboxylase. Both situations provide a central role for serine as a catalytic nucleophile.

Scheme 6:

A proposed route for inactivation by a phosphonate and catalysis of decarboxylation by a serine side chain based on the inactivation of benzoyformate decarboxyalse by a phosphonate analogue of benzyolformate [56,57].

Soon after the potential parallel of proteases and decarboxylases was first noted, Professor Nicholas Williams (Sheffield U.K.) alerted us to coincidental proposals for a decarboxylase that functions by addition of water to a carboxyl group. Professor Ding and coworkers reported a high resolution structure of isoorotate decarboxylase [61]. The overall reaction is shown in Scheme 7. The enzyme catalyzes the loss of a carboxylate that would leave behind a localized carbanion that is very similar to the reactive core of the substrate orotidine5-phosphate decarboxylase. The substrate is similar to the N-methylpicolinate, which is the case that was studied by Major and Gao and which showed no barrier to internal return of CO₂ [40]. This enzyme could undergo reaction by prior addition of a nucleophile to the carboxyl. In this case, the proposed mechanism resulted from analysis of the crystal structure of the enzyme with an analogue of the substrate at the active site. This showed that the enzyme resembles zinc proteases, as exemplified by carboxypeptidase A [60]. Ding and coworkers propose a decarboxylation mechanism that involves addition of water to the carboxyl of the substrate, similar to what we propose for the base catalyzed mechanism of MTh that leads to formation of bicarbonate and the mechanism of benzoylformate decarboxylase [59]. The catalytic proficiency of the enzyme is nearly equal to that of OMP decarboxylase and the acceleration mechanism appears to come from an adaptation of the active site of a zinc protease [59]. The metal ion promotes C–C cleavage that follows hydration, overcoming the problems associated with departure of CO₂.

Scheme 7:

The decarboxylation of isoorotate.

Conclusions

The ability of enzymes to catalyze decarboxylation reactions by trapping of carbanions with a Brønsted acid or addition of water, serine, or other nucleophile to a carboxyl group, overcomes problems associated with internal return of CO₂ or other obstacles to its release from nascent carbanions. The catalytic patterns and CKIEs associated with the decarboxylation of mandelylthiamin provide accurate data for discovering mechanisms. The surprising similarities between proteases and decarboxylases present an opportunity to consider evolution of common mechanistic pathways.