A Love Affair with Vitamins

I was born in Cambridge, England, where my father, David Green, had done the research for his Ph.D. in the laboratory of Sir Frederick Gowland Hopkins and where he remained as a Beit Fellow. My mother was English, whereas my father was an American citizen. When war broke out in Europe in 1940, America remained neutral, and my father was forced to leave England or give up his American citizenship. I grew up in Boston, New York, and then Madison, WI, where my father held a succession of academic positions. My earliest exposure to biochemistry, and to vitamins, was meeting the stream of scientists who joined our family for dinners. My mother was famous for her ability to prepare a dinner for guests with very limited notice and for her warm hospitality. Among those whose visits I remember were Luis Leloir and Sarah Ratner. I of course also met the members of my father's laboratory at the Enzyme Institute in Madison, and sometimes on Saturdays, I accompanied him to the laboratory and played with pH indicator dyes in little dimpled porcelain trays. I liked science and mathematics in school but never seriously imagined a career in science, most particularly because at that time there were few women in science, and most of those whom I met had chosen science to the exclusion of marriage and families. 
 
During one summer when I was in high school, my father employed me on the “night shift,” charged with preparing beef heart mitochondria for the next day's research. We would process vast amounts of beef heart from the local slaughterhouse in giant Waring Blendors and then isolate the mitochondria by differential centrifugation using a bank of centrifuges in the basement of the Enzyme Institute. I was amazed to find that I liked the work and enjoyed the camaraderie of the night shift. Nonetheless, I went to Radcliffe College in 1956 with the stated goal of becoming a social worker rather than a scientist! By that time, Radcliffe had merged with Harvard in all but name, and our classes were held jointly and graded jointly, even though our degrees remained separate. My cousin Barbara Green was a senior in Bertram Hall, where I was assigned as a freshman, and she was majoring in Biology. She spent hours telling me about the revolution in biology that was occurring, some of it at the Biology Laboratories at Harvard. So in the second semester of my freshman year, I signed up for my first science course, Biology 1. In the laboratory that accompanied this course, I met my husband-to-be, a Harvard sophomore named Larry Matthews. 
 
By the end of the first year of college, I had decided to major in Biology. The next 2 years were spent taking the required preparatory courses. I loved organic chemistry and calculus and hated introductory physics for non-majors, which seemed incredibly dry and unrelated to biology; at that time, the course was all about levers and pulleys, with almost nothing about atomic structure and nothing at all about quantum mechanics. In my junior and senior years, my biology courses were fascinating, e.g. a seminar in genetics taught by R. P. Levine and a course in biochemistry taught by George Wald. Wald was an incredibly inspiring lecturer, whose enthusiasm for science was infectious. A few months before graduation and my marriage to Larry, I broke my leg rather badly in a skiing accident, which greatly impaired my hunting for a job after college. Wald offered me a job as a research assistant in his laboratory. In retrospect, I find it amazing that I did not even consider applying to graduate school despite the fact that I graduated summa cum laude in Biology, but in 1960, there were almost no female role models on the faculty at Harvard, and I think it is fair to say that most male faculty members did not expect women, and especially married women, to seriously pursue academic careers in science.

I was born in Cambridge, England, where my father, David Green, had done the research for his Ph.D. in the laboratory of Sir Frederick Gowland Hopkins and where he remained as a Beit Fellow. My mother was English, whereas my father was an American citizen. When war broke out in Europe in 1940, America remained neutral, and my father was forced to leave England or give up his American citizenship. I grew up in Boston, New York, and then Madison, WI, where my father held a succession of academic positions. My earliest exposure to biochemistry, and to vitamins, was meeting the stream of scientists who joined our family for dinners. My mother was famous for her ability to prepare a dinner for guests with very limited notice and for her warm hospitality. Among those whose visits I remember were Luis Leloir and Sarah Ratner. I of course also met the members of my father's laboratory at the Enzyme Institute in Madison, and sometimes on Saturdays, I accompanied him to the laboratory and played with pH indicator dyes in little dimpled porcelain trays. I liked science and mathematics in school but never seriously imagined a career in science, most particularly because at that time there were few women in science, and most of those whom I met had chosen science to the exclusion of marriage and families.
During one summer when I was in high school, my father employed me on the "night shift," charged with preparing beef heart mitochondria for the next day's research. We would process vast amounts of beef heart from the local slaughterhouse in giant Waring Blendors and then isolate the mitochondria by differential centrifugation using a bank of centrifuges in the basement of the Enzyme Institute. I was amazed to find that I liked the work and enjoyed the camaraderie of the night shift. Nonetheless, I went to Radcliffe College in 1956 with the stated goal of becoming a social worker rather than a scientist! By that time, Radcliffe had merged with Harvard in all but name, and our classes were held jointly and graded jointly, even though our degrees remained separate. My cousin Barbara Green was a senior in Bertram Hall, where I was assigned as a freshman, and she was majoring in Biology. She spent hours telling me about the revolution in biology that was occurring, some of it at the Biology Laboratories at Harvard. So in the second semester of my freshman year, I signed up for my first science course, Biology 1. In the laboratory that accompanied this course, I met my husband-to-be, a Harvard sophomore named Larry Matthews. By the end of the first year of college, I had decided to major in Biology. The next 2 years were spent taking the required preparatory courses. I loved organic chemistry and calculus and hated introductory physics for non-majors, which seemed incredibly dry and unrelated to biology; at that time, the course was all about levers and pulleys, with almost nothing about atomic structure and nothing at all about quantum mechanics. In my junior and senior years, my biology courses were fascinating, e.g. a seminar in genetics taught by R. P. Levine and a course in biochemistry taught by George Wald. Wald was an incredibly inspiring lecturer, whose enthusiasm for science was infectious. A few months before graduation and my marriage to Larry, I broke my leg rather badly in a skiing accident, which greatly impaired my hunting for a job after college. Wald offered me a job as a research assistant in his laboratory. In retrospect, I find it amazing that I did not even consider applying to graduate school despite the fact that I graduated summa cum laude in Biology, but in 1960, there were almost no female role models on the faculty at Harvard, and I think it is fair to say that most male faculty members did not expect women, and especially married women, to seriously pursue academic careers in science.

Deciding to Be a Scientist
I worked in George Wald's laboratory while my husband completed his M.D. degree at Harvard Medical School. It was a revelation for me. The Wald laboratory studied the role of vitamin A in visual transduction, focusing on the events that occurred when visual excitation was triggered by the light-induced isomerization of vitamin A aldehyde (retinaldehyde) bound to rhodopsin. George's laboratory was small, and he offered each of us great independence in studying the steps in the bleaching of rhodopsin that were initiated by light. My bench mate, Toru Yoshizawa, devised a means of measuring the early intermediates in bleaching at very low temperature and discovered pre-lumirhodopsin. John Dowling was studying the physiological role of vitamin A acid (retinoic acid), which was unrelated to vision. George suggested to me that I should examine some rather peculiar spectral changes that accompanied the bleaching of frog rhodopsin at room temperature and then left to spend the summer at Woods Hole. I was to see whether similar changes might accompany bleaching of rhodopsin isolated from cattle retinas. The earlier studies in the Wald laboratory had elucidated a series of irreversible steps in bleaching, from rhodopsin to pre-lumirhodopsin, to lumirhodopsin, and thence to metarhodopsin, but what I saw was that there was a reversible equilibrium between two forms of metarhodopsin that differed in their spectral properties (1). The equilibrium could be shifted by varying the temperature or the pH or by adding glycerol to the solution. Classic metarhodopsin was red, with an absorption maximum at 478 nm, indicating that the Schiff base linkage between the protein and retinaldehyde was protonated, whereas what I termed metarhodopsin II was yellow, suggesting that this form had a deprotonated Schiff base. However, the pH depend-ence of the equilibrium was exactly opposite, with low pH favoring metarhodopsin II, suggesting that a major conformational change in the protein was associated with this transition. In collaboration with Paul Brown, who had designed and built a microspectrophotometer in the Wald laboratory, we were able to make measurements on isolated frog retinas. These measurements confirmed the formation of metarhodopsin II in the retina, and to our delight, they suggested that the appearance of metarhodopsin II coincided with the initiation of a neuronal signal. What we know now, but did not know then, is that metarhodopsin II is R*, the activated form of the G proteincoupled receptor rhodopsin.
These experiments were thrilling for me; trying to develop a coherent framework to explain my observations kept me awake at night and let me experience firsthand the excitement of science. Thus began the reorientation that led me to decide to become an independent scientist. Another factor in this transition was the atmosphere in the Wald laboratory. Each day, the entire group met for lunch in a room dedicated for that purpose. We took turns shopping for lunch supplies and shared the costs. Lunch conversations ranged from science to politics to art and music. We were often joined by guests, who contributed vigorously to our discussions. I remember visits from John Edsall, whose laboratory was upstairs, and from Jeffries Wyman, who had just finished walking across the Khyber Pass from India to Afghanistan.
Larry and I were frequent guests at the Walds' home, not just for laboratory parties but also more casually. Sometimes we were the only guests for dinner, and sometimes we were joined by guests with expertise and interests that ranged from Mayan archaeology to biology. Ruth Hubbard, George's wife, was not only a fine scientist in her own right but also a wonderful hostess and the mother of a young child. Of all the scientists who influenced my career, I think Ruth had the greatest impact. In apportioning her efforts between science and family, Ruth seemed to have the right balance, and she made it clear that one could do first-rate science and be a wife and mother.
I made the decision shortly thereafter that I would go to graduate school. Larry was completing his M.D. degree at Harvard Medical School and seeking a residency in orthopedics. We chose to go to the University of Michigan, he to residency and I to graduate school in biophysics.

The Training Years
For a woman who hated introductory physics, one may wonder why I chose a graduate major that involved more years of study of physics. My experience with the interplay of light and rhodopsin had convinced me that it was important to understand biology at the molecular level and that this required quantitative analyses of the experimental data. By this time, I was familiar with the elegant analyses of Edsall and Wyman. Wyman's mathematical analysis of hemoglobin had played a major role in the development of the theory of allosterism (2) Massey was studying a broad spectrum of enzymes that used flavin cofactors, and he was particularly interested in understanding how the protein modified the intrinsic reactivity of the flavin cofactor. Massey rapidly assembled a large and very active laboratory, populated with postdoctoral fellows from around the world and with graduate students from both biochemistry and biophysics. He was a little unsure whether a currently single mother of a young child could successfully complete a Ph.D., but he gamely agreed to let me continue in his laboratory. He wanted me to study the old yellow enzyme, "das alte gelbe ferment," an enzyme that had first been identified by Warburg and Christian and had been purified and characterized by Theorell.
To purify the enzyme from yeast autolysates, I adapted a simple batch purification step using DEAE-cellulose, followed by chromatography on DEAE and then on calcium phosphate. To my amazement, old yellow enzyme was not yellow but green! The spectrum indicated a charge transfer complex. Such charge transfer complexes frequently occur when small molecules are bound in close proximity to the flavin cofactor of a flavoprotein. I used anaerobic dialysis of reduced old yellow enzyme to separate the putative small molecule from the protein and was able to identify a peak with absorbance at 330 nm by chromatography on Bio-Gel P-2 (3). This fraction had "regreening" activity. I could isolate compounds from yeast extract that gave rise to similar but not identical spectral changes when added to old yellow enzyme. In the course of purifying these compounds, I caught a whiff of vanilla in the lyophilisate. If I had only looked up the chemical structure of vanilla, I would have saved a year of dissertation research! The mysterious regreening factor was shown to be p-hydroxybenzaldehyde by mass spectral analysis (4). The University of Michigan did not have a sensitive mass spectrometer for biological work at that time, and so I traveled to Michigan State to collaborate with Charles Sweeley in characterizing my ligand. Vanilla, which is p-hydroxy-3-methoxybenzaldehyde, also forms a charge transfer complex with old yellow enzyme! Despite extensive characterization, the physiological function of old yellow enzyme in yeast and its physiological oxidant remain unknown. We had hoped that the identification of the regreening factor would provide a clue as to the oxidizing substrate, but alas, this did not happen.
By the time I defended my Ph.D. dissertation, I was pregnant with our second child. My husband and I had planned to pursue joint fellowships in Stockholm, but on my arrival, I discovered that Sweden provided 9-month maternity leaves and had absolutely no facilities for care of a newborn, so my fellowship had to be delayed until our return to the United States. Larry had promised to return to Michigan as a faculty member upon completion of his research fellowship, and I was offered an opportunity to do my postdoctoral fellowship with Charles Williams at Michigan. The Williams laboratory also studied flavoproteins, especially flavin-disulfide oxidoreductases, and efforts to determine the x-ray structures of these important enzymes were under way in other laboratories. I was particularly interested in measuring the pH dependence of the reduction potential of the flavin cofactor of lipoamide dehydrogenase. The two-electron reduced form of lipoamide dehydrogenase, in which the redox-active disulfide is reduced but the flavin remains oxidized, exhibits characteristic absorbance at 530 nm indicative of a charge transfer complex between a thiolate (presumably from the reduced disulfide bond) and the oxidized flavin. Our studies showed that reduction of the disulfide by dihydrolipoamide involved transfer of two electrons and two protons throughout the pH range between 5.5 and 7.6. Thus, if the charge transfer absorbance is due to a thiolate interacting with the oxidized flavin, it is necessary to invoke a base that accepts the second proton upon reduction as shown in Fig.  1. We postulated that an active-site residue, perhaps a histidine, is protonated upon reduction of the disulfide and that this conjugate base stabilizes the thiolate anion of the reduced disulfide (5).
I next wanted to see whether this putative active-site base played a role in activation of dihydrolipoamide for disulfide interchange with the redox-active disulfide of dihydrolipoamide reductase. My colleague David Ballou had an anaerobic stopped-flow spectrophotometer, which permitted rapid measurements of the enzyme absorbance after the oxidized dihydrolipoamide dehydrogenase was mixed with dihydrolipoamide or the reduced enzyme was mixed with lipoamide. We measured the reaction in both directions as a function of pH. Our experimental results were consistent with a mechanism in which the active-site base accepts a proton from dihydrolipoamide to activate the thiol for nucleophilic attack on the enzymatic disulfide bond (6).
Shortly after this paper was published, Charles Williams attended a symposium in La Paz, Baja California Sur, Mexico, at which Georg Schultz presented the x-ray structure of the related flavin-disulfide oxidoreductase, glutathione reductase. To our delight, a histidine was indeed present in the active site, perfectly positioned to stabilize the thiolate in the two-electron reduced enzyme. This x-ray structure, which illuminated so many of our experiments, influenced me profoundly. It was clear that our detailed kinetic and thermodynamic experiments could help to interpret the structure that they provided and that their beautiful structure could help us visualize the active site in ways that immediately suggested further experiments.
The Williams laboratory was a very exciting place to be at that time. Colin Thorpe, trained as a inorganic chemist, was a postdoctoral fellow, and Keith Wilkinson and Mike O'Donnell were graduate students in the laboratory. We discussed our experiments on a daily basis, and together, we examined the structure of glutathione reductase and discussed its implications for our own work. The entire laboratory met daily for lunch, and our discussions ranged from science to politics. Charles was also an amazing men-tor, who gave each of us independent and non-overlapping projects and who also allowed us to move our projects in new directions. Because Larry and I were not easily moveable since he was now an Assistant Professor in Orthopedics, Charles arranged for me to receive an appointment as a non-tenure track assistant professor and coaxed the Veterans Affairs Hospital into giving me a small laboratory and a technician so that I could begin independent research. I cannot emphasize enough how daring a decision that was for Charles, the Veterans Affairs Hospital, and the Chair of Biological Chemistry at the University of Michigan, Jud Coon. Only a few years before, anti-nepotism rules would have forbidden my appointment in most departments in the country, and yet here these brave men were encouraging me to begin an independent academic career.

On the Faculty at Last
I decided to study the role of folic acid (vitamin B 9 ) in one-carbon metabolism. The structure of the active reduced form of this vitamin is chemically very similar to that of dihydroriboflavin, and both can function in oxidation-reduction reactions (Fig. 2). However, in contrast to the roles of riboflavin-derived cofactors, which almost universally involve oxidoreduction, many enzymes that use folate cofactors catalyze reactions involving transfer of a one-carbon unit with no apparent involvement of oxidoreductions. I wanted to see whether "latent oxidoreductions" facilitated these one-carbon transfers.
Folate-dependent enzymes such as dihydrofolate reductase and thymidylate synthase had already been studied in detail because they were important targets for cancer chemotherapy. I focused my attention on the folate-dependent enzymes involved in methionine biosynthesis and regeneration (Fig. 3). The first step in this pathway, in which a methyl group is generated de novo by reduction of methylenetetrahydrofolate, is catalyzed by methylenetetrahydrofolate reductase (MTHFR). This enzyme was shown to be a flavoprotein with an FAD cofactor in Robert Stokstad's laboratory in Berkeley. Although the Stokstad laboratory was unable to purify the mammalian enzyme to homogeneity, their studies also  elucidated an extremely important allosteric modulation of MTHFR activity by S-adenosylmethionine (AdoMet) (7). AdoMet serves as the universal methyl donor in biological systems, and its availability regulates the de novo synthesis of methyl groups, which uniquely are transferred to homocysteine to form methionine, the precursor of AdoMet. Thus, the Stokstad laboratory had provided a classic example of feedback inhibition.
My major interest in this pathway pertained to the cellular control of one-carbon metabolism. Methylenetetrahydrofolate is essential for the de novo biosynthesis of thymidylate and is reversibly interconvertible with 10-formyltetrahydrofolate, the one-carbon donor for de novo purine biosynthesis. As shown by Stokstad, reduction of methylenetetrahydrofolate to methyltetrahydrofolate is an irreversible reaction that removes this one-carbon donor from the pool available for nucleotide biosynthesis. Under these circumstances, partitioning of one-carbon units should not be regulated just by demand for methionine but also by demand for purines and pyrimidines, but no mechanism for this partitioning was known.
The second step in the de novo generation of the methyl group of methionine is the transfer of the methyl group from methyltetrahydrofolate to homocysteine to form methionine. In prokaryotes such as Escherichia coli, this step terminates the pathway for methionine biosynthesis. In mammals, for which methionine is an essential amino acid because they are unable to synthesize homocysteine, this step conserves methionine by regenerating the methyl group. Thus, methionine must be supplied from the diet to support protein synthesis, but the demands for methionine for biological methylation reactions can be sustained by regeneration of methionine from the homocysteine produced as a result of these reactions.
Methionine synthase in mammals uses methylcobalamin, a derivative of vitamin B 12 , as a prosthetic group, and in E. coli, a highly homologous protein, the metH gene product, is synthesized when the bacteria are supplied with vitamin B 12 . The E. coli protein had been characterized, and the detailed mechanism was described by the laboratories of Herb Weissbach, who has recently written a Reflections article describing his studies (8), and of Frank Huennekens. The mechanism based on these studies is shown in Fig. 4. During catalysis, the cobalamin prosthetic group is alternately methylated with a methyl group derived from methyltetrahydrofolate and demethylated by transfer of the methyl group to homocysteine. Demethylation results in the formation of cob(I)alamin, in which the pair of electrons forming the carbon-cobalt bond in methylcobalamin remain on the cobalt. Cob(I)alamin is highly reactive toward oxygen, and during turnover, it undergoes oxidation about once in every thousand turnovers, forming cob(II)alamin. This form of the enzyme is inactive, and return to the catalytic cycle requires both reduction and remethylation. The Huennekens laboratory showed that the electron donor for reductive reactivation in the E. coli enzyme is flavodoxin, a small electron transport protein (9). The source of the methyl groups for reactivation is AdoMet rather than methyltetrahydrofolate (10).
I found this complicated reaction cycle very intriguing. Here is an enzyme that reacts with three different substrates and must distinguish between the reactivation and primary turnover methyl donors to avoid futile cycling. In the absence of any structural information, we had no clue about how this discrimination might occur.
My small laboratory group began our studies by developing methods for the efficient purification to homogeneity of both enzymes, mindful of Arthur Kornberg's commandment "do not waste clean thinking on dirty enzymes" (11). Purification of MTHFR was particularly challenging because in this case the enzyme had to be purified from a mammalian source since the bacterial enzyme lacked the complex allosteric regulation of the mammalian enzyme. We chose pig liver because it was readily available from a local slaughterhouse, but 3 kg of pig liver contain only 7 mg of MTHFR! Maria Vanoni, who was a postdoctoral fellow in the laboratory, placed a Chinese poster on our wall, "This is the year of the pig." Frozen pig livers were diced and blended into a slurry that was clarified by low speed centrifugation. Chromatography of this disgusting soup was impossible, so we devised a batch DEAE purification that gave us a 20-fold purification with a 99% yield and resulted in a clear solution containing our enzyme. Subsequent chromatographic steps resulted in a 32,000-fold purification with a 14% overall yield to produce 1 mg of homogeneous enzyme (12).
Younger readers may wonder why we did not simply clone MTHFR, but this was the early 1980s, when mammalian cloning was in its infancy. Ironically, although we later collaborated with Rima Rozen in Montreal to clone human MTHFR, the only successful method of overexpression is in baculovirus-infected cells (13) (and very recently in Pichia pastoris (14)), despite efforts in many laboratories. The inability to successfully overexpress mammalian enzymes has greatly inhibited many efforts to study mammalian metabolism at the molecular level and remains one of the great barriers to progress.
About 10 years after our initial purification of porcine MTHFR, I received a call from Rima Rozen, a human geneticist at McGill University in Montreal. She thought that she had cloned human MTHFR and sent me some sequence data to look at. Unfortunately, the sequence was of an acyl-CoA dehydrogenase rather than the desired MTHFR. At that time, I had a graduate student, James Sumner, who was attempting to clone porcine MTHFR using peptide sequence data from the N-terminal catalytic domain and the C-terminal regulatory domain. Rima and I decided to join forces, and in a few months, human MTHFR had indeed been cloned (15,16). Rima was particularly interested in reports that a frequent polymorphism in human MTHFR was associated with elevated plasma homocysteine and might be a risk factor for cardiovascular disease in humans. Before sequencing of the MTHFR gene, this polymorphism was identified by the thermolability of the MTHFR protein in lymphocyte extracts (16). By analysis of the MTHFR gene from 114 humans, she located the polymorphism at nucleotide 677. At this locus, a C leads to an alanine in position 222 of human MTHFR, whereas a T leads to a valine in the same position. The polymorphism is extremely common, with ϳ10% of humans having the TT genotype. Rima showed that the 677C3 T polymorphism was indeed associated with significant thermolability of MTHFR, as assessed by heat treatment of lymphocyte extracts (17).
Frustrated by the inability to express human MTHFR at levels necessary for biochemical analysis, we decided to clone, overexpress, and purify the enzyme from E. coli. As mentioned earlier, this enzyme lacks the C-terminal regulatory domain of the mammalian enzymes. Christal Sheppard was able to overexpress and purify the wild-type bacterial enzyme (18), and she collaborated with the laboratory of Martha Ludwig to obtain a crystal structure of the enzyme (19). The bacterial enzyme is a tetramer of identical subunits that are arranged in a planar rosette. Each subunit is an ␣ 8 ␤ 8 (TIM) barrel, with the flavin of the FAD prosthetic group located at the mouth of the barrel with its adenine dinucleotide dripping over the edge of the barrel. Ala 177 (homologous to Ala 222 in the human enzyme) is located at the bottom of the barrel, far from the FAD cofactor. In studies of the Val 177 bacterial enzyme, Christal showed that the Val 177 enzyme was indeed thermolabile and that thermolability was associated with an increased propensity for loss of the flavin cofactor and dissociation of the tetramer into dimers when the mutant enzyme was diluted. The Ala 177 side chain lies inside a tight loop between helix ␣5 and strand ␤6 of the barrel, and Val cannot be accommodated without distortion of the structure. Residues at the opposite end of helix ␣5 interact with the phosphate and ribosyl groups of the FAD, rationalizing the enhanced propensity for flavin loss associated with the A177V mutation.
Dr. Kazuhiro Yamada, a postdoctoral fellow in the laboratory, succeeded in overexpressing human MTHFR in baculovirus-infected insect cells. He was then able to purify the human enzyme to homogeneity and to repeat the experiments that Christal had initially performed with the bacterial enzyme, demonstrating that the human MTHFR A222V mutant was accurately modeled by the bacterial A177V mutant (13).
Kazuhiro then discovered that human MTHFR is multiply phosphorylated on a serine-rich N-terminal extension of the catalytic domain that is absent in the bacterial enzyme (20). Phosphorylation is reduced when human HEK293 cells are cultured in medium lacking methionine or containing adenosine, which leads to a low ratio of AdoMet to S-adenosylhomocysteine. In vitro, activity is increased when the enzyme is treated with alkaline phosphatase. Thus, in mammalian cells, another layer of regulation has been added to the allosteric regulation of the enzyme activity by AdoMet. This type of dual regulation is quite common and was first documented in glycogen phosphorylase (21).
Our studies on the second enzyme involved in synthesis/regeneration of the methyl group of methionine, cobalamin-dependent methionine synthase, began a few years after I established my own laboratory. Although we were isolating an enzyme from E. coli rather than from a mammalian source, the purification of the enzyme was still incredibly tedious. In those bad old days before cloning and overexpression, we would order kilograms of dried cell paste from Grain Processing in Muscatine, IA, where the company operated a 2000-liter fermenter. When Ruma Banerjee joined my laboratory as a postdoctoral fellow, we decided that we would clone and sequence the metH gene specifying cobalamin-dependent methionine synthase (22). This was a bold decision on Ruma's part because she had trained as a chemist and had absolutely no experience in microbiology or genetics, but she rapidly learned to generate nested deletions and to pour and run sequencing gels. MetH is a 136-kDa protein, and Ruma estimated that she had sequenced 0.1% of the E. coli genome.
At about the same time, Brown, Goldstein, and colleagues reported the sequence of the low density lipoprotein receptor and showed that the gene was a mosaic of parts shared with various other genes (23). We were incredibly envious because our sequence had no homologies or analogies in the extant sequence data base. It was only the second sequence for a cobalamin-dependent enzyme to be reported. To make the sequence talk to us, Ruma began what was to be a long process of dissecting the enzyme into functional modules. She showed that limited proteolysis of the native enzyme led to the formation of a 28-kDa peptide that retained the ability to bind cobalamin and identified this cobalamin-binding domain as extending from residues 643-900 of the 1227-amino acid protein.
Subsequently, Celia Goulding was able to dissect the enzyme into two halves: the N-terminal half contained determinants for binding homocysteine and methyltetrahydrofolate, whereas the C-terminal half contained determinants for binding cobalamin, AdoMet, and flavodoxin (24). Amazingly, when the two halves were mixed in the presence of substrates, they catalyzed the overall reaction efficiently but failed to form any detectable complex.

Histidine Again
In 1988, Cathy Drennan entered the biochemistry graduate program at Michigan and decided to study x-ray crystallography in Martha Ludwig's laboratory. Cathy had studied at Vassar, where Miriam Rossi was one of her teachers. Dr. Rossi had studied with Dorothy Hodgkin at Oxford and had determined the x-ray structure of methylcobalamin (25). Martha's research at Michigan was at that time focused on flavoproteins, especially flavodoxin, but Cathy convinced Martha to let her tackle the crystallization and structure determination of methionine synthase. In so doing, she initiated 20 years of collaboration between our two laboratories.
Cathy succeeded in obtaining crystals of the wild-type methionine synthase in the methylcobalamin state, and from their beautiful red color, we could tell that we had indeed crystallized a cobalamin-containing protein. However, when we analyzed the crystals by electrophoresis, it became obvious that she had crystallized a smaller fragment containing the determinants for cobalamin binding (26). During the long incubation at room temperature necessary for crystallization, the enzyme apparently underwent proteolysis, and only this fragment crystallized. Its borders were nearly identical to those of the cobalamin-binding fragment that Ruma had first studied. However, spontaneous limited proteolysis is problematic: when we changed the column in our fast protein liquid chromatography system, the proteolysis stopped, and no more crystals were obtained. It took many enzyme preparations to break in the column and to reintroduce the contaminants that led to proteolysis! The path to a successful structure determination was long and arduous for Cathy. Solution of the structure required isomorphous replacement with heavy metals, which proved extremely challenging, and the 3-Å resolution limit of the crystals made developing maps extremely difficult, but Cathy persevered. When I went on vacation in December of 1993, she was still butting her head against a stone wall, but when I returned, she appeared at my office door, beaming, and we went down for my first look at the structure. To my amazement, the dimethylbenzimidazole nucleotide that coordinates to the cobalt of methylcobalamin in solution had been displaced, and a histidine from the protein was now coordinated to the cobalt (27). This histidine was absolutely conserved in all the MetH homologs that had been sequenced to date. In fact, it was part of an Asp-X-His-X-X-Gly sequence that had been identified by Neil Marsh upon comparing the sequences of various cobalamin-containing enzymes (28). Indeed, the coordinating His 759 was hydrogen-bonded to Asp 757 , which in turn was hydrogen-bonded to the oxygen of Ser 810 . The question now was why: what advantage was conferred by this dramatic substitution upon binding of cobalamin to the enzyme?
We first thought that substitution of dimethylbenzimidazole by histidine might facilitate catalysis. However, mutation of Ser 810 to alanine or Asp 757 to asparagine or glutamate led to unimpressive losses of catalytic activity (29). In contrast, mutation of His 759 to glycine led to an enzyme that was catalytically completely inactive. Curiously, as the catalytic activity decreased in this series of mutant enzymes, the rate of reductive reactivation with AdoMet as the methyl donor increased. Dr. Joseph Jarrett, who conducted these experiments, is an extremely keen and thoughtful observer, and he recognized that the properties of the mutants could be correlated with the equilibrium between His-on (coordinated) and His-off (dissociated) forms of the enzyme in the cob(II)alamin form, reflected in subtle changes in the UV-visible absorbance and more dramatic changes in the EPR spectra. As the hydrogen bonding of the ligand triad (Ser-Asp-His) was disrupted and finally as the histidine itself was removed, the enzyme shifted from His-on to His-off forms and simultaneously lost catalytic activity and gained in the rate of reductive remethylation. Joe made the great leap to recognizing that the ligand triad was controlling the conformation of the protein: as coordination of the histidine was weakened, the enzyme shifted from conformations in which the cobalamin was juxtaposed to the modules that bind homocysteine or methyltetrahydrofolate to the conformation required for reductive reactivation, which juxtaposed cobalamin with the AdoMet-binding module. Joe and David Hoover, a graduate student in Martha Ludwig's laboratory, showed that when the electron transfer protein flavodoxin was bound to wild-type methionine synthase (reduced flavodoxin serves as the electron donor in reactivation), the reactivation conformation was also favored (30).
Structures of the individual modules and the N-and C-terminal halves of the protein now were solved in the Ludwig laboratory (31)(32)(33), although a structure of the intact protein remains an elusive goal. The structure of the C-terminal half of the protein in the cob(II)alamin form was particularly informative (33). It was initially solved with the H759G mutant, which Joe Jarrett had shown to be locked in the reactivation conformation. This structure strongly suggested that His 759 would have to be dissociated for the enzyme to enter the reactivation conformation because interactions of the cobalamin corrin ring with the AdoMet-binding domain lead to a rotation of the corrin that moves the cobalt away from C-␣ of residue 759 by 2.3 Å, so the correlations between increased rates of reductive activation and weakened coordination of the histidine now made structural sense. We began to realize that the presence of a colored prosthetic group, whose spectral properties were sensitive to the conformation of the protein, was a unique advantage in exploring the conformational changes required to execute the complete catalytic repertoire of methionine synthase.
Dr. Vahe Bandarian, who had joined my laboratory as a postdoctoral fellow, decided to study the ligand triad mutants in the context of the C-terminal half of methionine synthase. He postulated that by reducing the number of stable conformations that were available, he would be better able to detect shifts in the equilibria between them. Of the four conformations shown in Fig. 5, removal of the N-terminal half of the protein would leave only the Cap-Cob and AdoMet-Cob conformations. When the enzyme is in the methylcobalamin form, if histidine dissociates from the cobalt in the AdoMet-Cob conformation, the color would be expected to shift from red to yellow, and thus, the contribution of this conformation to the ensemble could be estimated under different conditions.
Vahe was able to show that ligand triad mutations, particularly the D757E mutation, shift the equilibrium to favor the AdoMet-Cob form. He was also able to measure the effect of temperature and of substrates or substrate analogs on the equilibrium, allowing us to deduce a set of rules governing the interchange. Although the shifts were more subtle in the full-length protein, he was able to show that the same rules applied, and in addition, he could then measure the influence of methyltetrahydrofolate on the equilibrium. His experiments formed the basis for my inaugural paper in the Proceedings of the National Academy of Science of the United States of America (34). It became clear that the various stable conformations were energetically very similar such that differential ligand binding of Ͻ1 kcal/mol was sufficient to perturb the equilibrium.
Very recently, Supratim Datta, a postdoctoral fellow in my laboratory, constructed a C-terminal fragment containing His 759 and stabilized it in the reactivation conformation by introducing a disulfide bond. The structure of this stabilized C-terminal fragment, determined by Markos Koutmos, a postdoctoral fellow in Martha Ludwig's laboratory, provided yet another surprise (35). In this structure, His 759 is indeed dissociated from the cobalt, and now it makes intermodular hydrogen bonds with residues in the AdoMet-binding module, so the dissociation of His 759 is associated with stabilization of the reactivation conformation, "locking" the enzyme in this conformation until the cofactor can be reduced and methylated by AdoMet.
Angela Fleischhacker, a chemistry graduate student, examined the effect of the net positive charge on the cobalt of cobalamin on the His-on/His-off equilibrium in fulllength MetH. For the series of alkylcobalamins in solution, earlier studies in Ken Brown's laboratory had shown that the base-on/base-off equilibrium becomes less favorable (i.e. stabilizes the form in which dimethylbenzimidazole is dissociated and replaced by water) as the alkyl substituent becomes more electron donating in the order methyl, ethyl, propyl (36). Angela measured the corresponding equilibria in MetH and found a very similar trend (37). She showed that cob(II)alamin lies between ethylcobalamin and propylcobalamin in the series so that the His-on/Hisoff equilibrium for this species in considerably less favorable than for methylcobalamin. In the case of the protein, the equilibrium does not simply involve the dissociation of histidine but also a shift from the reactivation conformation into one of the catalytic conformations. Her results emphasize the delicate balance between the stability of the major conformers such that changes in the His-on/His-off equilibrium of a few kilocalories result in detectable shifts in conformation.
Our results now suggested that shifts between His-on and His-off conformations of the methylcobalamin cofactor reflected changes in conformation, with the reactivation conformation assuming the His-off form. However, Supratim Datta noted that the aquocob(III)alamin form of his disulfide-stabilized C-terminal fragment was also red, indicating a His-on conformation, even though the disulfide cross-link would be expected to lock the enzyme in the reactivation conformation. In collaboration with Katherine Patridge and Markos Koutmos in the laboratory of the late Martha Ludwig, he was able to crystallize this species. The results showed that the red form was indeed in the reactivation conformation but in a strained state in which the histidine has moved under the cobalt and broken the hydrogen bond interactions with the AdoMet-binding module (40). We imagine a tug of war in which the net positive charge on the cobalt competes for the histidine with the intermodular interactions that stabilize the reactivation conformation. In the cob(II)alamin state, binding of flavodoxin is sufficient to shift the equilibrium to favor the reactivation conformation, and the dissociated histidine forms the intermodular bonds that stabilize this conformation. In the cob(I)alamin state, which is dominantly base-off (four-coordinate) in solution, the enzyme remains in the reactivation conformation. However, upon methyl transfer from AdoMet to form methylcobalamin, the positive charge on the cobalt increases, and the histidine is drawn into the His-on state, breaking the intermodular contacts and allowing the enzyme to shift into one of the catalytic conformations. In aquocobalamin, there is even more positive charge on the cobalt, sufficient to stabilize the strained His-on state sufficiently to allow us to trap it. The Cap-Cob conformation is seen in the crystal structure of the cobalamin-binding module (27) and positions a four-helix bundle (the Cap) at the N terminus of that module above the cobalamin cofactor (Cob). The Fol-Cob and Hcy-Cob conformations are required to juxtapose methyltetrahydrofolate (Fol) and homocysteine (Hcy), respectively, above the cobalamin cofactor to allow methyl transfer in primary turnover. The AdoMet-Cob or reactivation conformation is the conformer that binds the electron donor flavodoxin most tightly and juxtaposes adenosylmethionine with the upper face of the cobalamin.

Acknowledgments
First and foremost, I wish to express my gratitude to my students and postdoctoral fellows and to my collaborators in the studies I have described: above all, my colleague and friend, the late Martha Ludwig (Fig. 6). In our 20-year collaboration, Martha and I served as foils for one another. She was coolly analytical and extremely cautious in reaching conclusions, which served to check my more intuitive and impulsive tendencies. We spent many hours each week going over data in minute detail and thinking jointly about what to do next and how to interpret it. I miss her terribly.
My account has focused on the research in our laboratory, but I should point out that parallel developments in laboratories around the world have contributed to our understanding. To begin with, genome sequencing has revealed that cobalamin-dependent methyltransferases or methyltransferases containing related corrinoid cofactors are abundant in both prokaryotes and Archaea. In many anaerobic organisms, these enzymes catalyze reactions in central energy-generating pathways such as methanogenesis. All of these enzymes cycle between Co(I) and MeCo(III) forms during catalysis, and it appears that almost all of them require reductive activation at intervals during catalysis (recently reviewed in Ref. 38). In all of these methyltransferases, the cobalamin or corrinoid cofactor is bound with the dimethylbenzimidazole nucleotide displaced (base-off), and in the majority of them, a histidine residue from the protein is coordinated to the cobalt when the cofactor is in the methylcobalamin form. In Methanosarcina, a wide variety of methyltransferase complexes have been identified that catalyze methyl transfers from simple one-carbon metabolites to coenzyme M (methylthiosulfonate) in an early step in methanogenesis. In contrast to methionine synthase, these complexes segregate the functions of donor and acceptor substrate binding and corrinoid binding into separate proteins. One of these enzyme complexes, methanol-coenzyme M methyltransferase, has been particularly well characterized both mechanistically and structurally. The corrinoid-binding component of this complex shows significant sequence homology to the analogous domain of methionine synthase, including the characteristic Asp-X-His-X-X-Gly motif indicative of a corrinoid cofactor with a histidine coordinated to the cobalt. Recently, a structure of the corrinoid-binding component (MtaC) in complex with one of its substrate-binding components, the methanol-binding protein (MtaB), was determined (39). This is the only structure of a methyltransferase corrinoid-binding domain in complex with one of its substrate-binding proteins or domains. MtaC is indeed structurally related to methionine synthase, and as in the structure of the C-terminal fragment of H759G methionine synthase, the fourhelix bundle or cap at the N terminus of the corrinoidbinding protein is displaced to allow juxtaposition of MtaB with the corrinoid cofactor. It remains to be determined whether the role of the axial histidine ligand in methionine synthase in orchestrating conformational changes during the catalytic cycle will also be the same in the coenzyme M methyltransferases.
This Reflections article necessarily omits much of the oeuvre of my laboratory and thus fails to name many of the wonderful students, postdoctoral fellows, and others who have contributed. I have been blessed with many truly outstanding colleagues, who form my second family. It is a delight to have a former associate visit, perhaps with a spouse or a child. When consulting or committee work takes me to a city where former colleagues live, I have enjoyed the opportunity to meet them for dinner and to catch up on their lives. Although quite a few of them are in academia or doing research in industry, a few of them have jobs that make use of their training in unique ways. One of my former graduate students, Christal Sheppard, is now Counsel to the House of Representatives Judiciary Committee and formerly was Democratic Staff to the House Science Committee. Kerry Fluhr is now Intellectual Property and Commercialization Manager at the Commonwealth Scientific and Industrial Research Organization, Australia's National Research Agency in Sydney. Mohan Amaratunga is Life Sciences Program Manager at GE Global Research. I feel extremely privileged to have been blessed with an exciting career in research and a fulfilling personal life. My husband is an orthopedic surgeon, and I was often asked by his colleagues and their spouses why I wanted to work so hard outside the home when his salary was more than enough to support our family. My answer was always, simply, that I loved the work. Although science is often frustrating, particularly at the bench level, it is occasionally exhilarating and always absorbing. One can never forget what my colleague Janet Smith calls the "aha" moments, the times when a set of disparate observations gels to provide a new insight. I have not tried opiates but suspect that aha moments have similar addicting qualities. I take particular pleasure in seeing my own passion for science reflected: in many members of my lab family; in my younger son, Keith, who is finishing graduate school at California Institute of Technology in theoretical astrophysics; and in my granddaughter Jennifer, who has discovered the pleasures of mathematics at the age of 12.