How Do Enzymes Really Work?

Myunderstanding is that “reflections” can cover amultitude of sins. I have previously written a brief history ofmy career and donotwant to duplicate that exposition (1). Instead, this “reflection” will trace a primary focus of my laboratory for many decades, namely how enzymes achieve their remarkable catalytic efficiency. (Indeed, this has been a focal point of the field of biochemistry for a much longer period.) This discussion will include personal interjections that trace my progress from undergraduate years to the present, as well as describe some of the other areas pursued in my research program. The answer to the question posed inmolecular terms has been elusive, even though it has been a central theme of many research programs, including my own. A complete review of this subject is far beyond the scope of this article. Instead, I will review this subject as reflected in my personal interests and research program, although some historical perspective is included. I apologize in advance for not including all of themany noteworthy people and references that have significantly advanced the understanding of enzyme catalysis. My entry into the field of enzymology as an independent investigator occurred in 1960 when I was hired as an instructor at the Massachusetts Institute of Technology (MIT). My background included a graduate degree in Physical Chemistry from the University of Wisconsin, where I did my graduate work in the laboratory of Robert Alberty, a wonderful mentor as well as a leading figure in the field of steady-state enzyme kinetics. This was followed by a postdoctoral year in Göttingen in the laboratory of Manfred Eigen, who developed the field of relaxation methods and received the Nobel Prize in 1967. These methods revolutionized the field of solution kinetics, and I was very fortunate to have the opportunity to work closely with him. At this juncture, the importance of kismet should be noted. My choice of the University of Wisconsin for graduate school was not based on a careful analysis of graduate schools. It was a knee jerk reaction to my undergraduate years at Princeton University. Princeton provided a great education, but it was a social desert with nowomen coeds in the area and cars banned for students. This was especially difficult for a student coming from a high school in a small town inWisconsin. Aftermy Princeton experience, I wanted to return to the warmth (not referring to temperature) of theMidwest.My selection of a research supervisor was based on a single conversationwith Robert Alberty after I arrived at graduate school. My choice of postdoctoral study was based on a seminar I heardManfred Eigen give at the University ofWisconsin; he was relatively unknown at that time, but I found the work very exciting. Perhaps today’s students are worried too much about these decisions and their importance. My faculty position at MIT was also very fortunate. Walter Stockmayer, a senior professor atMIT and a brilliant scientist, visited Eigen’s laboratory, and I had lunch with him. A job offer followed soon afterward. The incredible catalytic efficiency of enzymes had been long recognized. Indeed, this efficiency of billions, trillions, and even more relative to the uncatalyzed reaction is unparalleled in chemistry. Consequently, the interest of many chemists, particularly organic chemists, was aroused. By 1960, the understanding of enzyme catalysis had advanced beyond the lock and key hypothesis of Emil Fisher, in which the substrate (key) was presumed to fit exactly into the enzyme (lock). This hypothesis dealt with the specificity of enzymes rather than their efficiency. The efficiency of enzymes was first postulated to be due primarily to the close proximity of the amino acid side THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 283, NO. 33, pp. 22337–22346, August 15, 2008 © 2008 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in the U.S.A.

chains of the enzyme participating in catalysis to the substrate, thereby raising the effective concentration of the catalyst, and to the proper orientation of the catalytic groups of the enzyme. However, the effective concentration of the enzyme cannot be a dominant effect: the concentration of pure water is only 55 M, so an enhancement greater than ϳ100 is not possible. Thus, proper orientation of the catalytic groups was assumed to be of predominant importance. In studies of acid-base catalysis of model "substrates," Tom Bruice and collaborators demonstrated that freezing in the rotational degrees of freedom of the substrates greatly enhanced catalysis, in some cases by more than a factor of a million (2). This caused a flurry of rethinking of enzyme catalysis, and a popular way of viewing this process was to view the enzyme as an "entropy trap" that froze the substrate into a particular configuration relative to translation and rotation (3). Many talented organic chemists attempted to create small molecule catalysts that contained acid-base groups required for catalysis and a binding site to dock the substrate. However, to this day, such attempts have not produced a small molecule catalyst that comes close to achieving the efficiency of enzymes. A related approach is to try to proteolyze enzymes to make them smaller than the original molecule of biological origin. This approach has also not been successful: the minimum molecular mass for an efficient enzyme is ϳ10,000 Da. In other words, an enzyme must be a macromolecule.
The prevalent method for studying the kinetics of enzyme mechanisms in 1960 was steady-state kinetics. Such studies are extremely useful in characterizing the specificity of the active site of the enzyme, in some cases giving a reasonable picture of the active site, and in determining the overall turnover number of the enzyme. Studies of the pH dependence also provide useful information about the approximate pK values of groups involved in catalysis, but identifying the specific side chain is hazardous, as pK values are frequently shifted in enzymes relative to model compounds. Chymotrypsin, in particular, was intensively studied with steady-state kinetic methods, with literally hundreds of such studies being published. Unfortunately, steady-state kinetic studies provide little direct information about reaction intermediates, as the enzyme concentration is too low to detect reaction intermediates directly. In a landmark study, Peller and Alberty (4) provided an insightful analysis of steady-state kinetics that indicated exactly what information steady-state kinetics can provide with regard to reaction intermediates and pK determinations. (This is not to negate the use of isotope rate effects and isotope exchange, which have proved extremely useful in mechanistic studies of enzymes, although they, too, are limited with regard to the direct information obtained about reaction intermediates (5).)

Transient Kinetics
The obvious way to circumvent the shortcomings of steady-state kinetics is to study enzyme catalysis at high concentrations of enzymes. The difficulty with this approach is that because of the extreme catalytic efficiency of enzymes, the reactions become very fast, typically occurring in times Ͻ1 s, in fact, usually in the millisecond and microsecond ranges. Stopped-flow methods were available at this time, which extended the observable time range to milliseconds, but their use was not widespread because appropriate equipment had to be home-built. Chance and collaborators, among others, developed this method and found transitions between reaction intermediates in enzyme catalysis, especially with redox systems having visible chromophores such as flavins and porphyrins (6). Again, chymotrypsin became a special target with the use of nitrophenol ester substrates: the solution becomes yellow when nitrophenolate ion is formed (7). Such experiments provided clear kinetic evidence for the now well known acyl-enzyme intermediate, and its isolation further documented the existence of this reaction intermediate (8). (It should be noted that the extensive studies of chymotrypsin were not due to intrinsic interest in this enzyme, but were motivated primarily by the ease of preparation of large quantities of the enzyme and the relative ease of preparing hundreds of substrates because the enzyme is both a protease and esterase.) The time resolution of the stopped-flow method is limited to a few milliseconds (although with special adaptations, modern instrumentation has achieved submillisecond time resolution). This situation was remedied with the development of relaxation methods by Manfred Eigen and collaborators (9). These techniques extended the observable time range into the subnanosecond regime. The temperature jump method is especially useful for enzymes, as microsecond time resolution is easily achieved with joule heating via electrical discharge, and submicrosecond times with lasers. These methods also permitted the study of reactions previously deemed to be immeasurably fast, such as protolytic reactions and hydrogen bonding. Studies of model compounds in this instance are important for understanding the role of these elementary steps in enzyme catalysis.
When my laboratory entered the field of enzyme mechanisms, we were particularly interested in elucidating the elementary steps in enzyme catalysis using transient kinet-

REFLECTIONS: How Do Enzymes Really Work?
ics and thermodynamics. By direct observation of reaction intermediates, it was possible to arrive at the details of enzyme mechanisms not previously achieved and to generalize this information into a global picture of enzyme catalysis. I will not dwell on the use of x-ray crystallography in this endeavor. Obviously, this method is essential to arrive at a molecular description of enzyme catalysis. However, as has been often stated, in most cases, static structures of enzymes are observed, and their role in catalysis, a dynamic process, must be inferred.
In addition to studying enzyme mechanisms, we devoted a considerable effort to determining the rates of the elementary steps involved in protein conformational changes and the role of cooperativity in the rates of protein conformational changes. This knowledge is critical in developing a general model for enzyme catalysis.

Ribonuclease
Although my laboratory worked with a large number of different enzymes, I will discuss only the results obtained with ribonuclease A and aspartate aminotransferase, as these were studied in considerable detail and illustrate the type of information that can be obtained with thermodynamic and transient kinetic methods. Ribonuclease was selected for study because it is a relatively small enzyme (ϳ14,000 Da) and could be obtained in pure form commercially. It had been extensively studied with physical and chemical methods, so considerable information was available about its structure and function (10). In addition, the structure of the enzyme was being determined in two laboratories with x-ray crystallographic methods. Although physiologically ribonuclease catalyzes the breakdown of RNA, small molecule substrates could be obtained relatively easily, either commercially or by synthesis. The breakdown of a dinucleoside occurs in two steps: formation of a 2Ј,3Ј-cyclic phosphate, followed by hydrolysis into a 3Ј-nucleotide phosphate. The enzyme has a strong preference for cytidine and uridine in the cyclic phosphate. Steady-state kinetic studies of the breakdown of dinucleosides into the cyclic phosphate intermediate and of the hydrolysis of the cyclic phosphate yielded turnover numbers in the range of 2000 -3000 s Ϫ1 and Michaelis constants in the range of 1-5 mM.
Temperature jump studies of ribonuclease produced immediate results: the native enzyme was found to exist in two conformations that interconvert in the microsecond time domain. Furthermore, this interconversion is pH-dependent and is associated with a pK of ϳ6.1. (The specific amino acid involved is histidine 48.) Studies of the association of pyrimidine 3Ј-phosphates with the enzyme indicated that the expected bimolecular reaction is followed by a conformational change of the enzyme occurring in the microsecond time domain. Investigating the interaction of dinucleosides and pyrimidine 2Ј,3Ј-cyclic phosphates with ribonuclease was a bit trickier, as the overall hydrolysis of dinucleosides is essentially irreversible. To circumvent this problem, an entirely new method was devised, namely a stopped-flow temperature jump, in which the substrate is rapidly mixed with the enzyme, and a temperature jump is applied within milliseconds of the mixing.
At this point in time, I moved to Cornell University as a Professor of Chemistry (1965), and I enjoyed 23 wonderful and productive years. I was fortunate to be recruited to Cornell by Harold Scheraga, who, in addition to becoming a close friend, was a leader in the study of the chemical and physical properties of ribonuclease.
The results obtained with the stopped-flow temperature jump revealed a pattern similar to that obtained in studies of the pyrimidine 3Ј-phosphates, namely the bimolecular reaction is followed by a conformational change. These results were noteworthy for two reasons. First, the bimolecular rate constant for the reaction between enzyme and substrates is very large, approaching the limit for a diffusion-controlled reaction. Such large bimolecular rate constants appear to be typical for enzymes in general. Second, the binding of substrates is followed by a conformational change, again a general finding for enzyme reactions. A schematic view of the overall mechanism is shown in Scheme 1 (11). This mechanism already displays a very important attribute of enzyme mechanisms in general, namely that many intermediates and conformational states of the enzymes are involved in the catalytic process. A detailed modeling of the pH dependence of the hydrolysis of cytidine 2Ј,3Ј-cyclic phosphate revealed an even larger number of potential pathways, including parallel paths, in the mechanism (12). SCHEME 1. Schematic representation of the minimum reaction mechanism for the hydrolysis of a dinucleoside by ribonuclease A (as discussed in text). PypN, pyrimidine dinucleoside; Py2Ј:3Јp, pyrimidine cyclic phosphate; Py3Јp, pyrimidine 3Ј-phosphate; E and EЈ, different enzyme conformations.
These kinetic results can be rationalized in terms of the structure of the enzyme that emerged after the kinetic studies were largely completed. The molecule is kidneyshaped with short helices packed against a central ␤-pleated sheet. The enzyme also has a long N-terminal helix. The overall structure is stabilized by four disulfide bonds. The substrate-binding pocket is a groove, as might be envisaged for an enzyme that hydrolyzes RNA. Knowledge of the enzyme structure has led to the development of a structural model for the reaction. Conformational changes that occur as the reaction proceeds are very small. The dynamic equilibrium found in the unliganded enzyme can be envisaged as an opening and closing of the groove. Histidine 48 is at the top of the hinge, and its protonation state is important for this process. More direct evidence for the role of this residue in the conformational change of the isolated enzyme and in the catalytic reaction has been obtained recently with nuclear magnetic resonance methods (13). When the substrate binds, the hinge closes, squeezing out water and creating a hydrophobic environment. The reaction then proceeds through a series of proton transfer reactions involving histidines 12 and 119 and a pentacoordinated intermediate. The details of these reactions will not be considered here. The hinge then opens to release the product.
An important finding for later considerations is that iodination of tyrosines far from the active site alters the catalytic properties of the enzyme (14). This suggests that the integrity of the entire structure is important for catalysis and that communication occurs between the active site and residues quite distant from the site.

Aspartate Aminotransferase
A second enzyme selected for detailed study was aspartate aminotransferase. The reason for selecting this enzyme was entirely happenstance, although in retrospect, it turned out to be a great choice. When I was a postdoctoral fellow in Germany, I used some of my National Science Foundation travel money to visit the laboratory of Professor Rossi-Fanelli in Rome. He spoke no English, and I spoke no Italian. After a brief discussion in German, he turned me over to an assistant professor, Paolo Fasella, who spoke fluent English. I also discovered that he was married to the sister of a classmate of mine at Princeton. Paolo was working on aspartate aminotransferase, and thus, a long-term collaboration and lifelong friendship began. This involved many visits back and forth between our laboratories, including an extended stay by Paolo and his family when I was at MIT.
Aspartate aminotransferase catalyzes the transfer of an amino group from aspartate to ketoglutarate to give glu-tamate and oxalacetate. This mitochondrial enzyme was nontrivial to prepare, as it required fresh pig hearts as the starting material. Obtaining fresh pig hearts from slaughterhouses was far from an enjoyable task. The advantages of the enzyme were 2-fold. 1) The reaction could be easily studied in two parts, namely the addition of the amino group from either amino acid to enzyme-bound pyridoxal phosphate and the subsequent transfer of the amino group from pyridoxamine phosphate to either keto acid. 2) The enzyme-bound cofactor served as a wonderful spectral probe of the reaction in the visible region of the spectrum (ϳ300 -500 nm). Our first temperature jump experiments with aspartate aminotransferase were encouraging in that they clearly showed the transition between the aldimine (360 nm) and ketimine (330 nm) intermediates, but were disappointing in that they did not reveal other reaction intermediates. Nevertheless, we continued to pursue characterization of the enzyme and its mechanism, and other intermediates soon emerged (15).
The culmination of the investigation of aspartate aminotransferase occurred when we found a substrate that slowed down the reaction somewhat, ␤-erythro-aspartic acid (16). This allowed us to detect a very large number of intermediates. The analysis of the 11 relaxation times observed proved challenging, but in the end, we were able to delineate a detailed mechanism with seven intermediates in the transition from the initial pryridoxal enzyme to the intermediate pyridoxamine enzyme (15 intermediates for the overall reaction!). As an added bonus, we were able to determine the spectra of all of the intermediates. The mechanism of the half-reaction obtained is summarized in Scheme 2. The initial formation of the enzyme with the substrate is followed by a conformational change prior to formation of an aldimine with the substrate and pyridoxal phosphate. The enzyme then forms a quinoid structure with maximum absorption at 490 nm. The ketimine form of the enzyme substrate follows, which in turn becomes SCHEME 2. Schematic representation of the mechanism for the transamination reaction catalyzed by aspartate aminotransferase. Asp, aspartate; Oa, oxalacetate; Kg, ketoglutarate; Glu, glutamate; EAld, substrate-pyridoxal phosphate aldimine; EQuin, enzyme-substrate quinoid; EKet, substrate-pyridoxal phosphate ketimine; E A , pyridoxal phosphate-enzyme internal Schiff base; E N , pyridoxamine phosphate-enzyme. The primed enzyme is the open conformation, and the unprimed enzyme is the closed conformation.
the enzyme-keto acid complex. A conformational change then occurs, and the keto acid and pyridoxamine substrate are formed. I do not know of any other enzyme for which so many intermediates have been directly observed. Some years after our kinetic studies, the crystal structure of the enzyme was determined, along with the structures of the enzyme complexed with various inhibitors and substrates (17). The enzyme is a homodimer with a subunit molecular mass of ϳ45,000 Da. The two identical catalytic sites are formed from two different (identical) polypeptide chains. The active site has two clearly defined conformations, open and closed. The mechanism proposed on the basis of the crystal structures exactly parallels the mechanism derived from the kinetic studies, but in addition, details of the molecular interactions between the protein and substrate are delineated. Following the initial binding, a significant conformational change occurs as the polypeptide chains close around the substrate. This closed conformation sequesters the amino acid substrate from the solvent, and the sequence of aldimine-substrate, quinoid, ketimine-substrate, keto acid-pyridoxamine occurs within this closed structure. The conformation then switches to the open form to permit dissociation of the keto acid.
Thus, the mechanism for aspartate aminotransferase has many of the same features as that for ribonuclease, namely multiple intermediates and conformational changes. In the case of the Escherichia coli form of this enzyme, it has been demonstrated that residues not at the active site influence the catalysis (18). This view of enzyme catalysis is buttressed by many other experiments with these and other enzymes that are not discussed here.

Enzyme Conformational Changes
As part of my goal of dissecting enzyme mechanisms into all of the individual steps and intermediates, knowledge of more elementary steps, such as protolysis, hydrogen bonding, and alterations in water structure, was essential. In a 1963 review, Eigen and I summarized current knowledge of these elementary steps (19). Eigen and collaborators had done an impressive job with protolytic reactions, reactions deemed too fast to measure before the advent of relaxation methods. In fact, for most protolytic reactions, the rates can be calculated at least as accurately as they can be measured because the reactions of protons and hydroxyl ions with acids and bases are diffusion-limited, except for a few exceptional cases. The implications of these findings for enzymes were 2-fold: an upper limit of ϳ10 3 s Ϫ1 for turnover numbers of enzymes utilizing water in the proton transfer reactions and an upper limit of ϳ10 5 -10 6 s Ϫ1 for enzymes utilizing direct proton transfer from enzyme to substrate. This discourse will not consider the details of proton transfer reactions in enzyme catalysis, although they are obviously of great importance. I will continue to focus on reaction intermediates and conformational changes.
At the time of the review, there was scant information regarding the second-order rate constant characterizing the combination of enzyme and substrate, although the limited information available suggested that this rate constant was close to its diffusion-controlled limit. At the present time, it seems quite clear that this is the case, but most of the second-order rate constants seem to fall somewhat short of the diffusion-controlled limit (20). It should be noted that for a reaction to be diffusion-limited, it is necessary for the step following the initial combination of enzyme and substrate to be faster than diffusion apart of the reactants; therefore, diffusion-limited reactions are indirect evidence for a conformational change following the combination of enzyme and substrate.
To characterize protein conformational changes better, we decided to study the kinetics of simpler systems such as hydrogen bonding and alterations in water structure (21). In addition, transformations such as the helix-coil interconversion in polypeptides were probed. The primary tool used initially was ultrasonic attenuation measurements. Later in time, temperature jump methods (many in other laboratories) would prove very useful. For simple hydrogen bonding reactions such as dimer formation with 2-pyridone, the reaction is always diffusion-controlled. This implies that the rate of hydrogen bonding after the monomers have diffused together is of the order of 10 11 -10 12 s Ϫ1 . The rate of water structure breakdown in pure water can be derived from the frequency dispersion of the dielectric constant and was known to be ϳ10 12 s Ϫ1 . On the other hand, the first-order rate constant for the breakdown of water structure around a simple polymer such as polyethylene glycol was found to be ϳ10 8 s Ϫ1 from ultrasonic attenuation measurements. The reason for this relatively small first-order rate constant is that water structure is formed around the polymer, and the breakdown and formation of this structure are cooperative, i.e. they involve multiple water molecules. This is one of the simplest examples of an organized or cooperative system, but demonstrates an important point: cooperative processes occur much slower than the individual elementary steps. In simple terms, this is because initiation of the change becomes rate-limiting. Perhaps a more relevant example of a cooperative process is the helix-coil transition in polypeptides: the rate constants associated with this process lie in the range of 10 7 -10 8 s Ϫ1 (22)(23)(24). The rate constant associated with formation and breakdown of ␤-sheet struc-tures is somewhat less, ϳ10 6 s Ϫ1 (25). Another cooperative process of importance in biological systems, namely the stacking and unstacking of bases in nucleic acids, has characteristic rate constants in the range of 10 6 -10 7 s Ϫ1 (26).
Based on the above analysis, the fact that the observed conformational transitions in enzymes are much slower than the elementary steps involved implies that these conformational changes are highly cooperative. Other evidence also suggests that these changes are highly cooperative. This evidence includes the following: 1) the alteration of enzyme activity and, as will be discussed below, conformational changes by modification of amino acids distant from the active site; 2) the fact that enzymes must be fairly large macromolecules to function (more than Ϸ10,000 Da); 3) failure to synthesize small molecules that approach enzymatic efficiency; and 4) theoretical analyses, also to be discussed below. This does not imply that all conformational transitions within proteins are cooperative. Proteins are dynamic structures, and rapid motions of single amino acids or of a few amino acids have been observed by a variety of methods, including nuclear magnetic resonance and fluorescence polarization. Such motions are generally not cooperative in nature and are generally too fast to be directly involved in enzyme catalysis. Instead, these motions will be equilibrated during the relatively slower steps involved in enzyme catalysis. However, both the rates and equilibrium characteristics of these motions may be altered by slower conformational changes and other catalytic events. Consequently, it is important to characterize all of the potential conformations of the protein in structural, thermodynamic, and dynamic terms.
By the 1970s, it was well established that enzyme catalysis proceeds through a large number of intermediates and involves cooperative conformational changes. The question of how this creates efficient enzyme catalysis was still a matter of speculation, however. I referred earlier to the more obvious factors of a high effective concentration of the catalyst, exact orientation of catalytic groups, and restriction of substrate rotational freedom. A popular term for the conformational change following the initial binding of substrates to enzyme was "induced fit." This term was coined by Dan Koshland, an exceptional enzymologist and great wordsmith (27). The implication is that the conformational change creates a conformation that binds the substrate better than the free enzyme, and the enzyme adapts a shape fitting the substrate better. This is a very attractive picture but does not explain why the conformational change enhances catalysis. In point of fact, from the standpoint of thermodynamics, any change in conformation must bind the substrate better or it would not occur. An idea that is more far-reaching is that the conformational change drastically alters the environment of the active site. This is suggested by a number of crystal structures in which the water appears to be squeezed out of the active site when a substrate(s) binds. Some of the important consequences of this change in environment are as follows: 1) direct proton transfer without intervening water molecules can occur; 2) the effective dielectric constant is lowered significantly, thereby enhancing electrostatic interactions such as charge-charge and hydrogen bonding; and 3) pK values can be drastically changed, making proton transfer more favorable.
The confluence of reaction intermediates and the cooperative conformational changes suggest another important fact of enzyme catalysis. The enzyme breaks down the catalytic event into multiple steps with standard free energies of activation that are much less than would be found for a single-step reaction. This, of course, speeds up the overall catalysis. To accomplish this, the enzyme structure must optimize itself for each step: this conformational adaptability is possible only for a macromolecule.
In a speculative article, I suggested that perhaps the role of cooperative conformational changes is dynamic, in addition to the rather static roles postulated for the individual steps (28). The basic idea is that proteins have hundreds of relatively weak interactions within their structures. Making or breaking one or a few of these is unlikely to provide a catalytic advantage. However, if, for example, hundreds of weak interactions are formed simultaneously with the breaking of a substrate bond, the activation energy could be lowered considerably. This, of course, is the description of a cooperative process, i.e. the coordinated motion of the entire molecule. Such a compensation effect could be important in both the kinetics and thermodynamics of enzyme catalysis.
Thus, we have arrived at a fairly comprehensive picture of enzyme catalysis that incorporates information about the elementary steps involved and the role of multiple intermediates and cooperative conformational changes. As we shall see, this view of catalysis must be expanded in light of experimental results obtained many years later.

Interlude
Beginning in about 1970 and through 1989, my interest strayed from elementary steps in enzyme catalysis and its fundamental nature to more complex systems. I have always been interested in new methods and approaches, and we developed new approaches to looking at complex systems. These included new kinetic methods and extensive use of fluorescence, especially the use of fluorescence resonance energy transfer (FRET) to map out complex structures and fluorescence lifetime measurements to provide dynamic information about mechanisms and structures. We devoted considerable effort to allosteric enzymes, particularly aspartate transcarbamoylase (29). This work included binding studies to help establish the subunit structure of the enzyme, kinetic measurements, and structural mapping with FRET. The kinetic measurements revealed that regulation involves a myriad of conformational changes, some conforming best to the Monod-Wyman-Changeux model and others to the Koshland-Nemethy-Filmer model. Rabbit muscle phosphofructokinase presented an additional challenge, as enzyme association-dissociation appears to be involved in the regulatory process in addition to more traditional allosterism (30). We also did some kinetic studies of tryptophan synthetase, a fascinating enzyme that has two enzyme activities in one molecule (31).
The ATP-synthesizing enzyme from chloroplasts proved exceptionally interesting, as it gave my laboratory the opportunity to devise methods for working with membrane-bound enzymes. FRET was particularly useful in mapping out the overall topology. Thirteen distances were measured to provide a working model of the enzyme that proved to be reasonably consistent with the x-ray structure that appeared some years later (32,33). Much of this work was done in collaboration with my colleague Richard McCarty. In addition, we developed a phase-sensitive method that permitted the direct measurement of the rate of proton pumping in phospholipid vesicles (34). The stoichiometry of the ATP synthase proton pump reconstituted in phospholipids vesicles was also established (35).
The study of the pyruvate and ␣-ketoglutarate dehydrogenases posed some unique challenges, including mapping the distance between the three catalytic sites and providing evidence that transfer of substrate between lipoic acids occurred. We were also able to measure the rate of rotation of the lipoic acid between sites with dynamic fluorescence depolarization. This is the only measurement of this rotation rate that has appeared to date and demonstrates that it is fast enough to support the overall catalytic rate (36).
Finally, I should mention our studies of chicken liver fatty acid synthetase. In this case, we were able to measure the individual reaction rates for the sequence of reactions that occur, including the stereochemistry, as well as provide structural information through FRET measurements (37).

Leave of Absence
In 1988, I made a career decision to move full time into academic administration. I had served as Chair of the Department of Chemistry at Cornell University from 1970 -1975, but had declined a second term. I was quite happy with this decision, but in 1983, Bob Barker, an excellent biochemist and friend, was Provost and lured me into becoming the first Director of the Cornell University Biotechnology Program. (I had worked hard to recruit Bob to Cornell several years earlier.) This proved to be enjoyable and challenging. In the end, we had a vigorous program and a large new building. Having made the decision to close my laboratory and move into administration, I reluctantly left Cornell to become Vice Chancellor for Academic Affairs at the University of California, Santa Barbara. This position proved, however, to be a transient intermediate, and I moved to Duke University in 1988 as Vice Chancellor for Medical Center Academic Affairs. I have often been asked whatever possessed me to forsake a well supported thriving research program. (Many claimed that this must have been temporary insanity or a midlife crisis.) The answer is that I wanted the opportunity to build academic programs on a more global basis. Enhancing the overall quality of our academic institutions, both in teaching and in research, is a legacy that is beyond what I could have done as an individual faculty member. Life as an administrator was challenging (and frustrating), but in the end, I felt that significant contributions were made to Duke University in terms of new programs, maintenance of high quality, and faculty development.

Back to the Bench
In 1999, I decided to return full time to the faculty in the Department of Biochemistry. During my years as Vice Chancellor, I taught in the graduate biophysical course. Classroom teaching was always enjoyable to me: at Cornell University, I taught freshman chemistry as well as graduate courses. After a brief gestation period, greatly aided by a Vallee Professorship at Harvard provided by my good friend Bert Vallee, I decided that I would try an entirely new research field. I looked for a field that required experimental innovation, consistent with my previous research experiences, and that showed promise of providing new information about how enzymes work. The field that attracted me the most was single-molecule fluorescence microscopy, even though I had not worked with a microscope since freshman biology. Studying the kinetics of single enzyme molecules has the potential to reveal steps in the mechanism that are not seen in ensemble experiments.
In ensemble experiments, an average rate constant is measured, whereas for single molecules, each molecule reacts at a different rate. The mathematical correspondence between the single-molecule lifetimes and ensemble rate constant is known, thus permitting a correlation between single-molecule and ensemble experiments.
Because commercial equipment was not available, I had to go into the lab myself to see what could be done. Here again, kismet provided me with a unique opportunity. The Department of Cell Biology happened to have two excellent microscopists, Mike Sheetz and Tobias Meyer. (Both have since left Duke, much to my dismay.) Although they were not studying single molecules, they provided sufficient expertise so that I could build an appropriate setup to view single molecules. This was done on a shoestring with a modest investment from my department and an early grant from the Petroleum Research Foundation (PRF; run by the American Chemical Society). I cannot say enough good things about the PRF, the only organization that was willing to support me without a massive amount of preliminary data. Eventually, I was able to obtain an NIH grant, but that would not have been possible without the PRF. In my mind, it was a modest miracle that I could put together working equipment after so many years away from the bench. Many blind alleys were encountered along the way! One of the attractive features of the field of single-molecule fluorescence microscopy was that it was not highly populated with regard to individual investigators, a situation that is no longer the case. I did not want to set up a large research program, but hoped to find a few systems that might further our understanding of enzymes. Once again, fortune fell my way, as my longtime friend Steve Benkovic, an outstanding enzymologist at Penn State, was interested in setting up a collaboration. I first met Steve when he was a graduate student at Cornell, in about 1965. Because we shared many common interests, we become very good friends over the years. The two systems that we decided to look at were dihydrofolate reductase (DHFR) and the T4 DNA replication system. I will not discuss the latter, although we obtained some very interesting results with this system, but will discuss DHFR, as it represents a return to my original theme, namely how do enzymes work?

Return to Fundamentals
DHFR is an important enzyme that is essential for the biosynthesis of purines, thymidylate, and several amino acids and has been the target of drug therapy for cancer. It catalyzes the reduction of 7,8-dihydrofolate by NADPH to form 5,6,7,8-tetrahydrofolate. The enzyme has been extensively studied by a variety of different methods, including kinetics, structure, nuclear magnetic resonance, and theory (38,39). From the standpoint of single-molecule fluorescence microscopy, it has several advantages: 1) it is very stable; 2) site-specific mutagenesis is relatively easy; and 3) removal of all of the sulfhydryl groups gives an active enzyme. This latter feature makes it possible to substitute a cysteine at one or more places along the polypep-tide chain, thus permitting specific labeling with fluorescent probes. Also important for this story is that a variety of assays are available, including measurement of the rate of the overall reaction, measurement of the rate of hydride transfer, and measurement of the rate of change in fluorescence of covalently bound probes.
Fluorescent enzymes were prepared with labels in various locations on the molecules, both near and far from the catalytic site. The enzyme has a flap that closes over the active site when substrates bind: the flap proved to be a convenient place to put fluorescent probes that were environmentally sensitive. In addition, probes were placed in positions not at the active site. These were coupled to the catalytic site via FRET pairs, which are sensitive to changes in distance between probes. In brief, the single-molecule experiments, along with ensemble experiments, revealed a myriad of conformational transitions. Some were associated with ligand binding. Some were directly coupled to hydride transfer, whereas others were not. Previous studies had shown that mutations far from the catalytic site can influence the activity of the enzyme and that these mutations also influence the nature of the conformational changes. These results again suggest that motions of the entire molecule are essential for the catalytic reaction. This harkens back to the earlier discussion of cooperative (coordinated) conformational changes.
Taken together, these results suggest that a multitude of conformations are involved in the catalytic reaction, some of them serially in the reaction sequence and some as parallel paths. For a single substrate (S), and a single product (P), this mechanism can be represented as shown in Scheme 3 (40). In this schematic representation, E i represents the multiple conformations of the enzyme, the A i , B i . . . X i represent conformations of enzyme-substrate complexes that are sequential in the catalytic pathway, and parallel catalytic pathways exist, e.g. A 1  In other words, for each step in the reaction sequence, the enzyme should be considered to exist as a large ensemble of conformations, and many of the members of the ensemble are capable of carrying out catalysis. Of course, this ensemble view has been utilized in considering protein folding for quite a while, but its relevance for enzyme catalysis was not obvious. Recent nuclear magnetic resonance experiments by Peter Wright and collaborators with DHFR also suggest the existence of ensembles of conformations (41).
Evidence of shifting conformations and the movement of amino acids during catalysis has also been found in theoretical studies carried out on DHFR (39,42). In this work, the average changes in distance between amino acids were calculated as the reaction proceeds over the standard free energy barrier. These studies were done by my daughter, Sharon Hammes-Schiffer, and collaborators. This is yet another example of kismet. She is a professor at Penn State and was persuaded to carry out theoretical work on DHFR by Steve Benkovic. We did not suspect that our interests in DHFR would converge. This convergence of research interests was a pleasant surprise.
In looking at this ensemble view of enzyme catalysis, it should be noted that some of the interconversion of conformations may be very fast and thus at equilibrium during catalysis, whereas others may be on the same time scale as catalysis or even rate-limiting. Furthermore, the mechanistic path followed during catalysis will undoubtedly change as conditions, such as pH, are altered. In looking at catalysis in terms of standard free energy diagrams, the conventional plot of standard free energy versus reaction coordinate needs to be revised. The free energy diagram should be presented in at least three dimensions, as shown in Fig. 1 (43). In this plot, substrate is on one side of the peaks, and product is on the other. The paths up and down the peaks have multiple minima (conformations) and are connected by hills along the conformation coordinate (parallel paths). Multiple transition states are also present. This type of representation is applicable to all enzymes, not just DHFR.
Thus, my odyssey has taken me from a simple Michaelis-Menten picture of enzyme catalysis to this more complex view, a catalytic network with multiple protein conformations playing a predominant role in understanding the catalytic efficiency of enzymes. FIGURE 1. Schematic representation of the standard free energy (G 0 ) landscape for the catalytic network of an enzyme reaction. If a plane parallel to the axis labeled Ensemble Conformations bisects the transition states at the maxima, the substrates (EϩS) are on one side of the plane, and the products (EϩP) are on the other. This figure illustrates the multiple populations of conformations, intermediates, and transition states, as well as the large number of productive parallel catalytic pathways. Conformational changes occur along both axes, and strong coupling occurs between the coordinates. The reaction path(s) can move along and between both coordinates. For real enzymes, the number of maxima and minima along the coordinates is expected to be greater than shown. The dominant catalytic pathways will be altered by external conditions and protein mutations. This has been reprinted with permission (copyright 2008 American Chemical Society) (43).

Epilogue
I have once again decided to discontinue my research program and move into semi-retirement. A few collaborations with other laboratories remain. In assessing my years in academia, the most satisfying and important contribution in my mind is the many students and postdoctoral fellows who have trained in my laboratory and gone on to successful careers. I was fortunate to have many outstanding students and postdoctoral fellows pass through my laboratories. The research accomplished was entirely due to their efforts. Second on my list are the students whom I may have influenced through advising and classroom teaching. The rest has been exciting and fun and I hope has advanced the cause of basic science.