Part 4 – Can an enzyme make an irreversible reaction reversible?

Introduction

Currently, there is again a nice discussion on the Biocatalysis and Biotransformation group on LinkedIn. The question posed was “Can an enzyme make an irreversible reaction reversible?”, apparently with the idea to make good use of this reverse reaction. From the comments, it becomes clear that there still is misunderstanding about what an enzyme can do, so let me have a go at it too on this forum where I have some more room.

I think the answer revolves around understanding three concepts: irreversibility, thermodynamics, and probability. I’ll go over these one by one.

Irreversibility and analytics

Nothing is truly irreversible. There is no physical or thermodynamic law that states that you can’t unscramble an egg and have all the atoms move back into its shell. There is no thermodynamic law that has a time-direction component. This is the principle of time-symmetry: “what happens in one temporal direction can in theory also happen in the reverse”. The formulas to calculate particle movement are the same, irrespective of direction.[1] The concept of Microscopic Reversibility is widely accepted and means that atoms move along a reversible path along the reaction coordinate.

Thus, I think we should define what we mean by ‘irreversible’. Since this discussion is in the context of enzymatic catalysis, we should focus on what we feel is irreversible under practical chemical process conditions at scale (and not what can happen in the mass spectrometer, or in a sealed tube at 1000 °C). If we can’t detect the product of a reaction, we have to conclude that it did not form in the reaction. The term “irreversibility” therefore is connected to detection limits and current state-of-the-art analytics.[2]  A lot of irreversible reactions have been shown to be somewhat reversible by better analytical techniques.

Thermodynamics and time

Some reactions will have a high activation energy and may be prohibitively slow but given enough time, every system will reach a certain thermodynamic equilibrium. Let me use one example. As has been known for a long time, penicillin acylase can hydrolyze Penicillin G to phenyl acetic acid and 6-APA.[3] It can also catalyze the coupling of the two molecules. This is a true reversible reaction since it is easy to quantify the molecules. The thermodynamic equilibrium concentrations are the same in the case of a purely chemical reaction as in the enzyme case. If you happen to measure something different in the enzyme catalyzed reaction, you simply have not waited long enough [4], or you have inadvertently perturbed the system[5].

Figure 1: So-called ‘thermodynamic coupling’ to form Penicillin G.

Figure 2: Chemical equilibrium constant. Note the absence of an Enzyme-component. The equation for thermodynamic equilibrium takes a similar form.

Probability and time

Even if thermodynamics state that two molecules can react, there is only a certain probability that they will meet and actually do so. Outer space is virtually empty, the concentration of atoms lower than in the best vacuum ever produced on Earth. Yet, matter clustered and formed stars and planets – a low probability with enough time can lead to beautiful things.[6] The organic chemist’s trick at increasing the probability of a reaction is by increasing the concentration of reactants and often by inputting energy (stirring, heat) to increase diffusion and generate more collisions. A low probability and limited patience may make a reversible reaction seem irreversible.

Can an enzyme make an irreversible reaction reversible?

Well, that’s a moot point now since nothing is irreversible.  So, we have to be pragmatic about this and rephrase the question to “If we want to achieve an unlikely reaction to a measurable extent, how can an enzyme achieve things a process chemist alone cannot?”. Common process technological tricks revolve around changing macroscopic parameters in the product’s thermodynamic favor: substrate concentration, solvents, pH, precipitation of product, binding of product to a resin, extraction, etc and then applying a catalyst to reach that equilibrium faster. All of these process-tricks can be used in conjunction with an enzyme, assuming that the enzyme can be made robust enough (and there is ample evidence out there that shows that that can be achieved).

So, is an enzyme ‘just a catalyst’ that only helps attaining the thermodynamic equilibrium that is set by the macroscopic reaction parameters? In some cases the answer is “Yes” but sometimes it is a resounding “No”. An enzyme can perform some tricks the chemist can’t that make them incredibly efficient catalysts and increase the probability of a reaction occurring.[7]

Table 1: An enzyme is like a miniature reaction flask lined with functionality.

Especially orienting the molecules in the active site which can lead to a different product mixture is probably the reason for the confusion about ‘enzymes can affect the thermodynamic equilibrium’. Look at Figure 3 where the results of chemical ring opening of styrene oxide with azide differs very much from the enzymatic (halohydrin dehalogenase) ring-opening products. [8] Chemical reaction is relatively fast and occurs mainly at the benzylic (alpha) position. Azide is a poor leaving group and as such, the equilibrium exists of mostly a 70:30 mix of azido alcohols, with a negligible amount of styrene oxide. On the other hand, if we use enough enzyme to out-compete the chemical reaction, the mix is 2:98. Now, did the enzyme change the thermodynamic equilibrium?

Figure 3: Ring opening of styrene oxide with an enzyme gives a totally different product mix.

The point that people seem to miss is that in the case of enzyme catalysis, there can indeed be a DIFFERENT chemical equilibrium achieved but that that is NOT the thermodynamic equilibrium. Since the reverse reaction is so unmeasurably slow due to the azide being such a poor leaving group, it takes a very long time before true thermodynamic equilibrium is reached. The enzyme catalysis has resulted in what is called a ‘kinetic equilibrium’, a seemingly stable plateau on the way to the thermodynamically most favored product mixture.

Granted, reactions where regioselectivity issues are not in play, like making a simple ester from an acid and an alcohol, are much easier to understand and follow more closely the thermodynamics.

Conclusions

If we want to achieve an unlikely reaction, we don’t always need to worry about having to “change the thermodynamics which you can’t do”. It is enough to direct substrate and product mix towards a chemical equilibrium from which the product can be isolated. Modern beta-lactam antibiotics are all made on industrial scale using penicillin acylase in a kinetic coupling process from phenylglycine amide derivatives and the penicillin or cephalosporin nucleus. If you wait too long, all the antibiotic is hydrolyzed but there is a decent window of operation to isolate the desired product.

The fact that a reaction cannot be done in a round bottom flask in the normal pH and temperature range should not deter you from trying an enzyme which might create quite a different environment within the active site.

A thorough understanding of the enzyme mechanism will tell you which approach to take in case you want to accelerate the reverse reaction.

A cautionary word. Proof that a reverse reaction occurs is no guarantee it will be an efficient economical process at scale.

References

[1] For a really digestible discussion on these incomprehensible matters, read Brian Greene’s “The fabric of the cosmos  –  space, time, and the texture of reality”. Vintage Books, a division of Random House, Inc., New York, 2005. Chapter 6 Chance and the arrow – does time have a direction?

[2] Technology keeps progressing. Look at my good friend GeorgeFarquar’s work at the Lawrence Livermore National Lab on Single-Particle Aerosol Mass Spectrometry (SPAMS). https://www-pls.llnl.gov/?url=science_and_technology-life_sciences-spams

[3] V Svedas, AL Margolin, IV Berezin. Enzymic synthesis of b-lactam antibiotics: a thermodynamic background. Enzyme Microb. Technol. 1980, 2, pp. 138-144.

[4] Penicillin acylase is severely inhibited by phenyl acetic acid and thus, the enzyme kinetics are such that it is not much of a catalyst under these conditions and the equilibrium takes a long time to establish.

[5] Maybe the pH or ionic strength changed by introducing the enzyme solution, maybe the molecules associate with the enzyme and have an altered solubility. There are many things that can go wrong in this kind of experiments.

[6] Although not everybody believes that, like Republican Congressmen Paul Broun who told a Baptist church in Oct 2012 that “evolution, embryology and the Big Bang theory were lies spread by scientists out to erode people's faith in Jesus Christ.” He also claimed the Earth is roughly 9,000 years old, a view held by fundamentalist Christians based on biblical accounts of creation. The sad thing is that he serves on the House Science, Space and Technology Committee. http://www.huffingtonpost.com/2012/10/17/paul-broun-charles-darwin_n_1974054.html

[7] Of course, homogeneous and heterogeneous catalysis claims some of these features too, with a certain degree of success.

[8] I decided to ignore the fact that halohydrin dehalogenase from Agrobacterium radiobacter is also almost completely enantioselective and that actually out of the 4 possible product-enantiomers (R-a, S-a, R-b. and S-b), only one is formed. JH Lutje Spelberg , JE van Hylckama Vlieg, L Tang, DB Janssen, RM Kellogg. Highly enantioselective and regioselective biocatalytic azidolysis of aromatic epoxides. Organic Letters 2001, 3(1), pp 41-43.

Part 3 - Lead optimization and diversification, active metabolites and P450 enzymes

Introduction

Drug discovery and development is a long process and very expensive with a relatively short window of patent protection during which to recoup the investment and actually make some money. The patents are designed to cover the drug as broad as possible to prevent someone else from making a simple small change to the molecule and bypassing the patent protection.

A good patent covers the drug, its indication and ways of making it, its metabolites, its different salts, crystal forms, methods of delivery, dosage, combination therapy, etc etc.

Pretty rigorous, wouldn’t you agree? Well, there are some nice examples out there where someone could have done some more homework on the drug’s metabolism and made the company much more money.

Example 1 – Benzodiazepines

In the 1960s, Geigy launched Carbamazepine [1] as an anticonvulsant and mood-stabilizing drug for epilepsy and a few other disorders. [2] It is still being prescribed (Tegretol), however, it has some side effects. The main liver metabolite of Carbamazepine is Oxcarbazepine which has less side effects and is sold by Novartis [3]. By that time, Geigy was part of Novartis, so I guess the revenues went to the same coffers. The main metabolite of Oxcarbazepine, and the real active component, is its monohydroxy derivative Licarbazepine “MHD”, which is essentially just a reduced form. [4] Clever people at Sepracor figured out that the S-enantiomer worked better than the racemate [5] and launched Eslicarbazepine. Since it was shown that the S-enantiomer was the preferred drug, the Portugese company Bial developed Eslicarbazepine acetate as a pro-drug. They teamed up with Eisai and Sepracor for the development and marketing.

Figure 1: Better and better metabolites as marketed drugs. Finally, a pro-drug to bypass some IP.

Example 2 – Cetirizine

Hydroxyzine has been around for >60 years and is still used today.[6] Hydroxyzine is metabolized in the liver to Cetirizine which became a second-generation anti-histamine. [7] Again, it was found that only one of the enantiomers is the most active form: Levo-cetirizine. In this case, the company UCB has been successful in re-inventing the drug two times, now marketing this third-generation anti-histamine.

Figure 2: Three generations anti-histamines. We could have had Gen 3 already in the 50-ies.

Example 3 – Ezetimibe

Another drug where an active metabolite played a large role in the development is Ezetimibe (Schering Plough).[9-11] As can be seen in Figure 3, the initial hit that came from high throughput screening is a relatively undecorated molecule. While testing the candidate for effectiveness in the rat, the researchers found that the bile of the rat contained active metabolites that were more effective. Structure elucidation showed introduction of a polar group on the α-position, the (S)-alcohol being the most interesting.

A thorough follow-up metabolism study showed improved pharmacokinetic behavior when the methoxy groups were removed from the scaffold and when certain sites were blocked from P450 metabolism by exchanging hydrogen for fluorine atoms.

Figure 3: Development of lead to drug, the Ezetimibe story in one picture.

Discussion

The above examples clearly show the importance of rigorous study of metabolism of the drug candidate. The usual tools that are available are 1)  human and animal liver fractions like microsomes and hepatocytes, 2) the individual CYPs, 3) chemical methods of oxidizing molecules to mimic human metabolism.

Especially the chemical methods are rather crude since they mostly do not have any selectivity and always oxidize the easiest positions first. There are porphyrin systems that are designed to resemble a P450 that show some promise (I guess anything is better than bleach) but these alternatives cannot come close to the rich metabolism that enzymes can achieve. [12]

As such, it is surprising that the pharmaceutical industry has not embraced the diversity that is out there in Nature. There are many bacterial P450 enzymes that show overlap with human metabolism, yet also offer the distinct opportunity at making new non-human metabolites. The most studied and easy-to-use bacterial P450 enzyme is the CYP102A1 from Bacillus megaterium (“BM3”). In 2009, researchers from Eli Lilly and CalTech have written a very nice paper on the testing of the hypothesis that bacterial P450s have a place in drug development. [13] Figure 4 below is one of their examples, Verapamil, that shows that all major human metabolites are formed by a small set of mutant enzymes, as well as four new metabolites.

Figure 4: Verapamil metabolism by human CYPs (black box) and the broader BM3 metabolism (red box).

Conclusion

I don’t know about you, but I keep wondering what the Ezetimibe drug would have looked like if bacterial P450s had been used in the development.[14] Maybe the drug’s efficacy would be better if oxidation had occurred at the beta or gamma positions. This was not tested since there is no animal that displays that metabolism….

References

[1] My good friend Martin Mayhew (now at Verenium; www.linkedin.com/pub/martin-mayhew/b/4b7/673) commented on my previous post, mentioning that CYPs also can have synergistic and/or antagonistic effects, as another reason to do the microsome testing alongside individual CYP testing. For instance, Carbamazepine is metabolized by CYP 3A4 but induces (= “make more active”) other CYPs and this has an effect on other drugs if they are used together with Carbamazepine.

[2] http://en.wikipedia.org/wiki/Carbamazepine; http://www.ncbi.nlm.nih.gov/pubmedhealth/PMH0000620/

[3] http://en.wikipedia.org/wiki/Oxcarbazepine

[4] http://www.ncbi.nlm.nih.gov/pubmed/12959634

[5] A racemic mixture is a 1:1 mix of the enantiomers (mirror images, like left hand and right hand; http://en.wikipedia.org/wiki/Enantiomers ). See here for a nice paper on the two enantiomers: http://www.ncbi.nlm.nih.gov/pubmed/20578208. Sepracor became known for this trick, which was called “chiral switch”: http://www.usm.edu/phillipsgroup/CHE255_S10/chapter5/Chirality_at_work.pdf

[6] http://en.wikipedia.org/wiki/Hydroxyzine

[7] http://en.wikipedia.org/wiki/Cetirizine

[8] http://en.wikipedia.org/wiki/Levocetirizine

[9] SB. Rosenblum , T Huynh , A Afonso , HR Davis , Jr., N Yumibe , JW. Clader, and DA. Burnett. Discovery of 1-(4-Fluorophenyl)-(3R)-[3-(4-fluorophenyl)-(3S)- hydroxypropyl]-(4S)-(4-hydroxyphenyl)-2-azetidinone (SCH 58235):  A Designed, Potent, Orally Active Inhibitor of Cholesterol Absorption. J. Med. Chem., 1998, 41 (6), pp 973–980.

[10] Clader JW. The discovery of ezetimibe: a view from outside the receptor. J. Med. Chem., 2004, 47 (1), pp 1-9.

[11] http://en.wikipedia.org/wiki/Ezetimibe

[12] For instance: K Inami, M Mochizuki. Chemical models for cytochrome P450 as a biomimetic metabolic activation  system in mutation assays. Mutat Res. 2002, 519(1-2), pp 133-40. E Brulé, YR de Miguel. Supported manganese porphyrin catalysts as P450 enzyme mimics for alkene epoxidation. Tetrahedron Letters 2002, 43 (47), pp 8555–8558.

[13] AM Sawayama, MMY Chen, P Kulanthaivel, MS Kuo, H Hemmerle, and FH Arnold. A Panel of Cytochrome P450 BM3 Variants to Produce Drug Metabolites and Diversify Lead Compounds. Chem. Eur. J., (2009) 15, pp 11723–11729.

[14] Codexis sells these as licensed from CalTech: http://www.codexis.com/pdf/Codexis_MycroSys.pdf. They’re called MicroCYPs although you would not say so from the file name :-(.