Practicing Biochemistry without a License

I grew up in a New Jersey suburb twenty miles east of Manhattan. My passion for science began early in grade school, prompted in part bymy father’s training in chemical engineering and his work at Union Carbide Corporation during the development of the plastics industry. My first chemistry set and home laboratory were followed by a series of summer jobs in industrial laboratories. My fondest memory was making a ball-and-stick model of the most complicated organic molecule I could find in my chemistry textbook and showing it to my father. He instantly identified it as sulfanilamide. From thatmoment on, I knew Iwas learning a new language and joining my father in a serious undertaking. I was sure I would become an organic chemist. Myhigh school chemistry and physics teacherwas concerned that I had opted out of our biology course. He recommended a program for high school students at the Roscoe B. Jackson Memorial Laboratory in BarHarbor,Maine, then and now the world’s premier facility for the study ofmouse genetics. That summer changed my career path, giving me my first exposure to first-class scientists, including two future Nobel laureates. George Snell introduced us to the genetics of histocompatibility. Howard Temin was one of the college students in our group and later a co-discoverer of reverse transcriptase. I went to Harvard College that fall with the hubris that my summer at the Jackson Laboratory obviated the need to take the introductory biology course. Instead, in my freshman year, I enrolled in Carroll William’s physiology course and in a course on evolution led by Alfred Romer. These charismatic teachers failed to dissuade me from majoring in chemistry but left me with a nagging interest in biology and an inkling to go to medical school. My most memorable chemistry professors were Leonard Nash, Paul Doty, and George Kistiakowsky. “Kisty” was President Eisenhower’s science advisor. In mid-fall, he said to us, “Sorry, fellows [we had two women in a class of about fifty], I can’t lecture to you on Monday. I have to go to Washington to stop a third world war!” After getting an M.D. degree in 1961 from the University of Pennsylvania School of Medicine, I completed a three-year medical residency at New York Hospital, Cornell Medical Center. While there, I was the primary caregiver during the last two years in the life of a very gallant young teenager with Cooley’s anemia or -thalassemia major, an inherited defect in the synthesis of -globin subunits. Alfred and his illness impelled me to go into hematology. From 1964 through 1967, I had the good fortune to complete a research fellowship at the ThorndikeMemorial Laboratory and HarvardMedical Unit of the Boston City Hospital under the mentorship of James (Jim) Jandl, arguably the leading experimental hematologist of that era. Jim’s research focused on the red blood cell and disorders thereof. No other cell in the body can match the erythrocyte in physiologic importance, simplicity of design (it loses its nucleus before entering the bloodstream), and ease of access. Thus, it was a logical target for the ever-burgeoning discipline of biochemistry. THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 288, NO. 7, pp. 5062–5071, February 15, 2013 © 2013 by The American Society for Biochemistry and Molecular Biology, Inc. Published in the U.S.A.


Getting Started in Research
My time at the Thorndike Memorial Laboratory was the beginning of a twenty-five-year infatuation with all aspects of hemoglobin: structure, function, comparative biology, and pathophysiology. In those days, a considerable amount of biomedical research exploited the increasing availability of radioisotopes. I began by preparing 59 Felabeled hemoglobin to examine the stability of the hemeglobin linkage. I found that human oxy-, carboxy-, and deoxyhemoglobins held on to their hemes with equal tenacity. In contrast, when the heme iron was oxidized (methemoglobin), heme readily dissociated from globin and bound to a receptor protein (albumin) or exchanged hemes when mixed with another chromatographically separable unlabeled hemoglobin. The rate of dissociation of ferriheme from ␣-globin subunits was eight times slower than that from ␤or ␥-globin subunits. These experiments led to a better understanding of the pathogenesis of hemolytic anemia seen in patients with certain mutant hemoglobins and also provided some insight into the steps underlying the processing of hemoglobin in red cells infected with malaria.
After two years in Jim Jandl's laboratory, I was obliged to serve a two-year tour of duty in the United States Army. I was lucky enough to be assigned to a research laboratory at Fort Knox in Kentucky, which was the largest center for blood procurement and shipment during the Vietnam War. The transfusion of mismatched blood can result in massive lysis of red cells and release of hemoglobin into the plasma of the recipient. My first laboratory project was to investigate how the kidney handles circulating hemoglobin. Clinicians were aware of a seeming paradox. The blood plasma of normal humans contains 4 -5 g/dl albumin, but none of it can be found in urine. In contrast, comparable levels of free hemoglobin in the plasma result in dark red urine. The two proteins have similar molecular weights. Why is the renal glomerulus permeable to hemoglobin but not albumin? A series of single-author Journal of Biological Chemistry articles by Guido Guidotti (1,2) made it clear that under physiologic conditions, the oxyhemoglobin tetramer (␣ 2 ␤ 2 ) dissociates readily into stable ␣␤ dimers: ␣ 2 ␤ 2 7 2␣␤, with K d ϳ 2 M. The 32,000-kDa dimer would be expected to readily traverse the glomerular barrier. To test this hypothesis, I treated hemoglobin with bis(N-maleimidomethyl) ether, which Simon and Konigsberg (3) had shown bound covalently to the hemoglobin free sulfhydryl group at cysteine ␤93 and a second unidentified site, effectively cross-linking the tetramer. When either dogs or rats were infused with untreated hemoglobin or hemoglobin treated with a non-cross-link-ing sulfhydryl reagent, they passed the same voluminous amounts of dark red urine, whereas the urine of animals given bis(N-maleimidomethyl) ether was the normal pale yellow (4). This work has some practical implications. In the ongoing effort to develop free hemoglobin as an oxygen-transporting blood substitute, renal excretion is prevented by a variety of cross-linking reagents.

2,3-Bisphosphoglycerate and Comparative Hemoglobin Biochemistry
The primary mission of our army laboratory was to improve the stability and function of stored blood. Shortly before I arrived there, Ruth and Reinhold Benesch had reported that the oxygen affinity of hemoglobin in human red cells is tightly regulated by 2,3-bisphosphoglycerate (2,. This glycolytic intermediate is in roughly molar equivalence with the hemoglobin tetramer. 2,3-BPG lowers oxygen affinity because it binds tightly to deoxyhemoglobin but not to oxyhemoglobin. I showed that during storage in a variety of blood banking preservatives, the rapid decay of red cell 2,3-BPG closely paralleled a marked increase in oxygen affinity (5). This may pose a problem in a battlefield setting, where very large amounts of blood are often administered.
Upon discharge from the army, I felt I needed an apprenticeship in a strong biochemistry laboratory before initiating an independent research program. The Benesches were aware of experiments I had done at Fort Knox and gave me the opportunity of joining their laboratory at Columbia University. Six weeks after my arrival, I faced the first and only bona fide crisis in my career. I telephoned Helen Ranney at the Albert Einstein College Medicine and told her that Reinhold had just fired me from his laboratory for reasons that I was at a loss to explain and that continue to elude me. Helen, who at that time was a heavy smoker, replied, "Wait a minute. Let me go get a cigarette. I can see this conversation is going to take a while." She then arranged for me to join Robin Briehl's laboratory at Einstein in the Bronx.
While in Robin's laboratory, I showed that the addition of 2,3-BPG resulted in the expected marked lowering of the oxygen affinity of Hb A (␣ 2 ␤ 2 ) but weakly impacted the oxygen affinity of Hb A Ic (␣ 2 ␤ 2 N-glucose) and Hb F (␣ 2 ␥ 2 ) and had had no effect on Hb F in which the ␥ N termini were acetylated (6). After reviewing papers by Max Perutz, I made the educated guess that negatively charged 2,3-BPG bound to positive charges on deoxyhemoglobin at the ␤ N-terminal amino groups and at histidine ␤143 (position ␥143 is serine). I communicated these results to Perutz and several months later received a gratifying REFLECTIONS: Practicing Biochemistry without a License FEBRUARY 15, 2013 • VOLUME 288 • NUMBER 7 handwritten reply stating that he and Jonathan Greer did model fitting and found that "this seems quite a reasonable position for BPG to take up in deoxyhemoglobin. In oxyhemoglobin, on the other hand, this position is unfavorable because the internal cavity is too narrow." In his classic 1970 Nature article on hemoglobin cooperativity, Perutz states that "Fortunately, Bunn and Briehl have obtained biochemical data that go far to pinpointing the [2,3-BPG] binding site" (7). Three years later, while in Perutz's laboratory, Arthur Arnone reported x-ray diffraction patterns that confirmed this binding site (8).
I remained in close contact with Perutz until he died. He was totally bereft of any form of scientific elitism. Indeed, he seemed to take great pleasure in supporting the careers of junior colleagues. He always visited my laboratory when he came to Boston, and as I got to know him better, we shared common interests in music and art (Fig. 1).
Even though my primary charge as a fledgling hematologist at Harvard Medical School was to focus on human blood and disorders, I could not resist the urge to explore how 2,3-BPG impacts hemoglobin function in other mammals. I first studied domestic animals whose blood was easiest to access. I found that they could be divided into two groups. Like man, rat, rabbit, dog, guinea pig, and horse all had high levels of 2,3-BPG in their red cells and hemoglobins whose oxygen affinities decreased markedly after the addition of 2,3-BPG. In contrast, sheep, goat, cow, and cat all had barely detectable levels of red cell 2,3-BPG and hemoglobins of intrinsically low oxygen affinity that were totally unresponsive to the addition of 2,3-BPG (9). I then had the opportunity to extend this study thanks to a collaboration with two indefatigable zoologists, Alan Scott and Ulysses Seal. Ulysses had a carte blanche entre into all zoos in the United States. Armed with a blow dart laced with a sedative and a syringe, he could be trusted to draw blood on any animal irrespective of size and ferocity with a virtual total absence of iatrogenic mishaps. This bonanza enabled us to extend the scope of our project to seventyone mammalian species covering fourteen orders (10). The results closely mimicked the earlier study mentioned above on domestic mammals. The pig and seven ruminant species, as well as ten species of Felidae (cats) and Viverridae (hyenas and civits), had low oxygen affinity hemoglobins and very low red cell 2,3-BPG. Thus, it seems likely that early in mammalian evolution, 2,3-BPG regulation became imposed on all orders except artiodactyls and a subset of carnivores.
Even though I functioned as a physician at Brigham and Women's Hospital, I was very seldom beset with night call for sick patients. However, I did respond with great alacrity to a telephone call at 3 a.m. from a veterinarian telling me that one of his horses was about to foal. This was the first of a number of emergency trips to elite horse farms on Boston's North Shore. I was curious to learn why, in all mammals that had been studied to date, the oxygen affinity of fetal blood was consistently higher than that of maternal blood. In humans and other primates, this difference is due to the presence of a fetal hemoglobin (␣ 2 ␥ 2 ) that, as mentioned above, binds weakly to 2,3-BPG because of serine rather than histidine at position ␥143 (6). In ruminants, this difference can be explained by the presence of fetal hemoglobins that have intrinsically high oxygen affinity. In contrast, other mammals lack structurally distinct fetal hemoglobins. Hyram Kitchen and I found that among twelve pairs of foals and mares, the level of red cell 2,3-BPG in the newborn was 36% lower than that in the mother but rose to normal levels within 5 days (11). Subsequently, the same findings were observed in a number of other species that lacked fetal hemoglobins. Thus, Mother Nature has devised three independent ways for facilitating oxygen uptake in the developing and actively metabolizing fetus.
My foray into comparative "hemoglobinology" was augmented by the opportunity to take a mini-sabbatical on the Amazon River to study hemoglobins in freshwater fish. Austen Riggs organized a group of physiologists and biochemists to spend several months on the Alpha Helix, a large boat funded by the National Science Foundation and outfitted with remarkably compact and well equipped laboratory facilities. Many of these fish have a hemoglobin with an exaggerated Bohr effect (Root effect), indicating markedly enhanced lowering of oxygen affinity with falling pH. Hemoglobins with a Root effect facilitate the unloading of oxygen into the swim bladder, enabling fish to control their buoyancy. Using an anaerobic gel electrofocusing apparatus, I measured isoelectric points of hemoglobins in sixteen species of Amazon fish and showed that the difference in pI between carboxy-and deoxyhemoglobins provided an accurate measurement of the Bohr and Root effects (12).

Non-enzymatic Glycation
In the mid-1970s, my laboratory worked on non-enzymatic glycation of hemoglobin and other proteins. A decade earlier, Sam Rahbar showed that a minor component (Hb A Ic ) was elevated in red cells of diabetics (13). Following earlier work of Bob Bookchin and Paul Gallop at the Albert Einstein College Medicine (14), Paul and I showed that glucose formed a stable adduct with the N-terminal amino group of ␤-globin by a ketoamine linkage (Scheme 1) (15).
To study the biosynthesis of Hb A Ic in vivo, I clandestinely infused myself with serum transferrin bound to 59 Fe of high specific activity and then monitored the incorporation of radioactivity into the major and minor hemoglobin components (16). As expected, labeling of the major Hb A component rose to a maximum by approximately ten days after infusion and remained nearly constant dur-ing the 120-day red cell life span (Fig. 2). In contrast, the labeling of Hb A Ic rose slowly and continuously, exceeding that of Hb A after about day 60. This rather impetuous foray into human self-experimentation showed that Hb A Ic is formed continuously during the red cell's life span and that the ketoamine linkage is virtually irreversible. Thus, measurement of Hb A Ic could be and indeed has proved highly useful in monitoring therapeutic control of hyperglycemia in diabetic patients independent of fluctuations of blood glucose levels (17).
We utilized reduction with [ 3 H]borohydride to measure the on and off rates for the formation of the Schiff base aldimine linkage in Hb A Ic and the rate of conversion to the stable ketoamine adduct (18). These in vitro kinetics provided an accurate prediction of the formation of Hb A Ic in vivo (Fig. 2). We also identified ketoamine-linked glucose at other sites on hemoglobin ␣ and ␤ subunits as well as on human serum albumin (19) and proteins in the red cell membrane (20), lens crystallins (21), and basement membrane collagen in renal glomeruli (22). Non-enzymatic glycation in all of these proteins was elevated in patients with diabetes. These studies, along with work done in other laboratories (23,24), suggested the possibility that non-enzymatic glycation contributes to the longterm complications of diabetes.
Glucose, along with other aldose monosaccharides, is in equilibrium between the ring structure and the open chain structure in which an aldehyde or a ketone group is able to form a covalent linkage with amino groups on proteins. Compared with other isomers, glucose has the most stable ring structure (only 0.001% in the open form). We found that there was a very tight correlation between the fraction of an open chain in both aldose and ketose monosaccharides and the rate of condensation with hemoglobin, leading us to speculate that during evolution, glucose emerged as the universal metabolic fuel because it is far less prone than other aldose isomers to form unwanted and potentially toxic adducts with proteins (25).

Mutant Hemoglobins
Much has been said (although less has been written) casting doubt on the rigor of laboratory science done by physicians. Although bench research always has been the primary focus of my time and effort, I have tried my best to maintain a clinical role as a hematologist at Brigham and Women's Hospital. Through the years, I and many other physician-scientists have shared the angst of "robbing Peter to pay Paul." There have been gloomy times when I felt I was not doing justice to either my laboratory or my clinic. For me, a saving grace was that on many gratifying occasions, my research informed my clinical work and vice versa. This has been particularly true in the investigation of patients with mutant hemoglobins. Since 1970, I have exchanged correspondence and E-mail with physicians regarding patients with hemoglobin abnormalities, often resulting in the shipment of blood specimens to our laboratory and sometimes the opportunity to see the patients.
To date, more than 1000 structural variants of human hemoglobin have been identified. Most of these have little if any biochemical or physiologic significance. However, many hemoglobin mutants have been highly instructive "experiments of nature." The clinical and hematologic phenotype of affected individuals has provided insights that greatly enrich the information gained from structural and functional analyses of the purified mutant hemoglobin. Among the many mutants that our laboratory has investigated, three seem worth mentioning in this Reflections article.
Forty years ago, I was asked to see a patient because she had too many red blood cells. I found that her daughter also had erythrocytosis and that both had red cells containing equal amounts of normal Hb A and a positively charged mutant hemoglobin, Hb Kempsey (␣ 2 ␤ 2 99Asp3 Asn ) (26). This site is located at an interface between the ␤ and ␣ subunits that is critical in the concerted switch between the "oxy" (R) and "deoxy" (T) quaternary structures. Hb Kempsey had very high oxygen affinity and thus impaired oxygen unloading that led to intracellular hypoxia, increased erythropoietin production, and erythrocytosis. Experiments done in collaboration with Quentin Gibson showed that the mutation at position ␤99 greatly destabilized the T structure and thus enhanced the binding of heme ligands (26). Quentin was one of my heroes. I took some comfort from the fact that he began his pioneering and career-long work on hemoglobin when he was doing obstetrics in Ireland and happened upon a patient with congenital methemoglobinemia.
A year later, I saw a patient with compensated hemolysis due to the presence of an unstable mutant hemoglobin. Determining its structure proved to be quite a challenge. In those days, structural analysis began with two-dimensional maps of tryptic peptides. All peptides of ␤-globin Cranston were normal except that the C-terminal peptide was replaced by two unique peptides, one of which was very hydrophobic and stubbornly resisted purification until we jerry-rigged a manual countercurrent distribution. Edman sequencing of the two abnormal peptides indicated that a non-homologous crossover between ␤-globin genes had resulted in the insertion of an AG dinucleotide and thus a frameshift that encoded a polypeptide that was elongated by eleven residues (27). While this work was going on, my close friend Bernie Forget, in a neighboring laboratory, was using two-dimensional gels to analyze normal human ␤-globin mRNA and identified nucleotide sequences that matched perfectly the unique C-terminal amino acid sequence of ␤-globin Cranston (28).
After our laboratory started working on glycated hemoglobin, I received a number of letters and phone calls about anomalous measurements of Hb A Ic . Both falsely elevated and falsely low values can occur because of the presence of a mutant hemoglobin. On occasion, investigation of the mutant can be quite instructive. We were consulted about a nine-year-old child who, for unclear reasons, was tested for diabetes by a chromatographic method that revealed a very high Hb A Ic value, whereas his blood glucose levels were consistently normal. Jean-Paul Boissel showed that the boy was heterozygous for a novel type of mutant globin in which the ␤ N-terminal valine was replaced by a methionine (29). This substitution resulted in the retention of the initiator methionine, which is fully cleaved in all ␣-, ␤-, and ␥-globin mutants identified previously. Because the structural abnormalities involved uncharged residues, Hb South Florida had the same chromatographic and electrophoretic mobility as Hb A. The reason for the falsely high level of Hb "A Ic " was that ϳ20% of ␤ South Florida was acetylated at its abnormal N terminus, thus imposing a modest negative charge. These findings on Hb South Florida prompted us to explore in some detail the structural constraints that determine the subset of proteins that REFLECTIONS: Practicing Biochemistry without a License retain the initiator methionine and those that are acetylated at the N terminus. Jean-Paul used cell-free transcription and translation systems (reticulocyte and wheat germ) to express ␤-globin mutants in which the N-terminal valine was replaced by each of the other 19 amino acids (30). The initiator methionine was cleaved when the adjacent residue was relatively small, whereas it was fully retained when the adjacent residue was relatively large. The extent of N-terminal acetylation ranged from 0 to 100% depending on the N-terminal amino acid sequence. These results were in total accord with known structures of globins and also confirmed and extended conclusions that Sherman and his colleagues derived from yeast mutants of iso-1-cytochrome c (31).

Hemoglobin Subunit Assembly
My laboratory's interest in the assembly of hemoglobin subunits began with measurements of tetramer-dimer dissociation equilibria in purified mutant hemoglobins with oxygen affinities that were markedly high (Hb Chesapeake) and low (Hb Kansas). My initial data utilizing gel filtration (32) were confirmed and extended when I arrived in Robin Briehl's laboratory and he taught me how to use the Model E analytical ultracentrifuge (33). Although these results were both sparse and somewhat crude, they were in accord with predictions made from Guidotti's osmotic pressure measurements (2) and with the two-state allosteric model of Jacques Monod, Jean-Pierre Changeux, and Jeffries Wyman.
We then addressed a question that was of interest to both clinicians and biochemists. Individuals heterozygous for stable hemoglobin mutants would be expected to have three major hemoglobin species: normal Hb A, the mutant hemoglobin, and an asymmetrical hybrid tetramer. Thus, individuals with the sickle trait should have Hb A (␣ 2 ␤ 2 A ), the hybrid (␣ 2 ␤ A ␤ S ), and Hb S (␣ 2 ␤ 2 S ) in roughly 1:2:1 proportions. However, as all clinical laboratories would readily affirm, on electrophoresis, only two species are seen: Hb A and Hb S. Guidotti (2) had predicted that the hybrid is indeed formed but that the hybrid tetramer dissociates into ␣␤ dimers of different charge that are separated by electrophoresis or chromatography. When hemoglobin is deoxygenated, the marked change in quaternary structure confers a huge increase in tetrameric stability. When we deoxygenated a mixture of Hb A and Hb S and separated the hemoglobins anaerobically, half of the hemoglobin appeared as a band of intermediate mobility and was shown to be composed of equal amounts of ␤ A and ␤ S (34). Art Nienhuis and I used this experimental stratagem to demonstrate the coexistence of normal (A) and "stress" (C) hemoglobins within sheep red cells (35).
Observations on clinical laboratory data lay the groundwork for further investigation of the assembly of hemoglobin subunits. Humans inherit two ␤-globin genes, one from each parent. Commonly encountered ␤-globin mutants such as S, C, and D Los Angeles are all transcribed and translated at normal rates and produce protein products that are as stable as Hb A. Therefore, heterozygotes would be expected to have equal amounts of normal and mutant hemoglobins. However, as shown in Fig. 3, we noticed that ␤-globin mutants with amino acid replacements resulting in increased positive charge were usually less abundant than Hb A, whereas mutants with increased negative charge tended to be more abundant (36). Moreover, in heterozygotes with decreased production of ␣-globin (␣-thalassemia), the proportion of positively charged ␤-globin mutants fell in proportion to the number of ␣-globin genes deleted. In contrast, the coexistence of ␣-thalassemia resulted in an increase in the abundance of ␤-globin mutants with increased negative charge. These observations led us to hypothesize that during hemoglobin synthesis in erythroid cells, the rate at which ␣-globin and ␤-globin monomers combine to form stable ␣␤ dimers is driven in part by electrostatic attraction: normal positively charged ␣-globin monomer would be expected to combine more slowly with mutant ␤-globin monomers that have gained positive charges and more rapidly with those that have gained negative charges. Moreover, differences in the formation of mutant and wild-type hemoglobins would be enhanced when the availability of ␣-globin is limited by ␣-thalassemia. These expectations were borne out by competition (37,38) and kinetic (39) experiments employing purified wild-type and mutant heme-intact globin subunits (Fig. 3).

Sickle Hemoglobin
The remarkable ability of deoxy-Hb S to form rod-like intracellular polymers is fundamental to the pathophysiology of sickle cell disease. Clinicians were aware that SS (homozygous) patients had less severe disease if they were fortunate enough to have high levels of fetal hemoglobin (Hb F, ␣ 2 ␥ 2 ). We were curious to explore the role of the asymmetrical hybrid tetramer ␣ 2 ␥␤ S in the inhibition of polymerization by Hb F. Mark Goldberg and I found that the hybrid was much more effective than the ␣ 2 ␥ 2 homotetramer in inhibiting sickling (40). These experiments showed that the inhibition conferred by the ␥-globin is in trans to the valine ␤6 polymerization site.
In 1980, I used my first and only full sabbatical leave to go to the National Institutes of Health (NIH), and I worked first with Bill Eaton and then with Art Nienhuis. Bill arranged for me to be a Fogarty Scholar-in-Residence at the NIH. This gave me a wonderful opportunity to get to know a number of other scientists at the NIH and to visit their laboratories. Bill (Fig. 4) is one of the country's premier biophysicists, and over a thirty-year period, his laboratory has contributed more than any other to our current understanding of the kinetics and equilibrium of sickle hemoglobin polymerization. We decided to address a seemingly simple but clinically relevant question: why is the sickle trait (AS) benign, whereas SC patients have significant morbidity? Contrary to an early report involving rather crude measurements, Bill and I made extensive equilibrium and kinetic experiments that showed that polymerization of mixtures of Hb S and Hb C was identical to that of mixtures of Hb S and Hb A (41). However, we found that two other factors contributed about equally to enhanced sickling in SC patients. First, the level of Hb S in these patients is ϳ10% higher than that in AS individuals owing to differences in electrostatic attraction discussed above: ␣␤ S dimers assemble at about the same rate as ␣␤ C dimers but more slowly than ␣␤ A dimers. In addition and of equal importance, experiments done in collaboration with Carlo Brugnara (42,43) and Connie Noguchi and Alan Schechter (41) showed that polymer formation is favored in SC red cells because Hb C induces potassium efflux and water loss, resulting in higher intracellular hemoglobin concentration than in AS red cells.

Erythropoietin and Hypoxia-inducible Factor
During my sabbatical at the NIH, I decided that it was time to change to another area of research. As a gesture of farewell to my first love, I spent a good amount of time at the NIH working with Bernie Forget (at Yale University) on a comprehensive book on the molecular, genetic, and clinical aspects of hemoglobin (44).
My six months in Art Nienhuis' laboratory provided me with sorely needed hands-on tutelage in molecular genetic technology and enabled my laboratory to switch its focus from hemoglobin to erythropoietin (Epo). Genetics Institute, Inc., the company that first reported the cloning of the Epo gene, generously supplied us with expression constructs and assistance with radioimmune and biological assays. Jean-Paul Boissel, Danyi Wen, and I prepared a large number of site-directed mutants of Epo that helped to confirm its three-dimensional structure and also to determine the sites that bind to the Epo receptor (45,46).
Our transition to Epo research was further facilitated by the creative input of another postdoctoral research fellow, Mark Goldberg, who, as mentioned previously, had worked with me on sickle hemoglobin during college and medical school. After an exhaustive search, Mark found two human hepatoma cell lines that produce barely detectable Epo mRNA and protein when incubated at standard 20% oxygen but robust amounts of both in response to hypoxia (47). These cells were useful for identification of key elements in the Epo gene that are important in hypoxic induction. Most important is a hexanucleotide response element within a critical 3Ј-enhancer, which Wang et al. (48) showed binds to a heterodimeric hypoxia-inducible transcription factor (HIF). Kerry Blanchard (49) and Debbie Galson (50) in our laboratory identified a hormone response element in the Epo 3Ј-en- hancer, just downstream of the HIF site, that binds to the orphan receptor HNF-4 and contributes to high level hypoxic induction of Epo as well as localization of expression in the kidney and liver.
Subsequent studies in other laboratories showed that HIF can be induced in nearly all cells within all eukaryotes outside of yeast and that a large number of physiologically relevant genes are regulated by HIF. Eric Huang in our laboratory found that the levels of HIF-␣ and HIF-␤ mRNAs were not affected by hypoxia. However, Western blotting showed that equally abundant levels of HIF-␤ (aryl hydrocarbon receptor nuclear translocator (termed ARNT)) protein were present in both hypoxic and normoxic cells, whereas HIF-␣ protein was undetectable in normoxic cells but was readily induced by hypoxia (Fig. 5) (51). Moreover, when hypoxic cells were rapidly re-exposed to 20% oxygen, HIF-␣ was rapidly destroyed in the proteasome, with a t1 ⁄ 2 of Ͻ5 min. Eric identified an oxygen-dependent degradation (termed ODD) domain in HIF-␣ that was necessary and sufficient for rapid destruction in oxygenated cells (52). Soon thereafter, three laboratories independently demonstrated that in the presence of oxygen and iron, two highly conserved proline residues in the ODD domain are hydroxylated (53)(54)(55), enabling HIF-␣ to bind to von Hippel-Lindau protein, followed by docking of ubiquitin E3 ligase, polyubiquitination, and recognition by the proteasome. Peter Ratcliffe and his colleagues then identified and characterized HIF-␣ prolyl hydroxylases, first in Caenorhabditis elegans, and proved that these enzymes function as oxygen sensors for HIF regulation (56).

Ncb5or, a Novel Flavoheme Protein
Perhaps our laboratory's twenty-year devotion to hemoglobin research gave us tunnel vision. For a number a years, we suspected that the HIF oxygen sensor was a heme protein and indeed published preliminary evidence supporting this erroneous conclusion (57). However, the good news was that our continued search for novel heme proteins led Hao Zhu to the cloning and characterization of a novel 58-kDa protein with a cytochrome b 5 -like domain at the N terminus, a 90-residue spacer of unique sequence, and a flavin-binding cytochrome reductase-like domain at the C terminus (58). Hao showed that this protein is associated with the endoplasmic reticulum and functions primarily as a reductase able to donate electrons to a variety of substrates in vitro (59). Accordingly, it was named Ncb5or (for NADH-cytochrome b 5 oxidoreductase) (Fig. 6). The in vivo function of this protein was clarified by learning that a global knock-out produced mice with insulin-deficient diabetes and lipoatrophy (60). Subsequently, we found that our knock-out mice had a defect in ⌬ 9 desaturation of fatty acids, even when their diabetes was corrected by treatment with insulin (61). Thus, Ncb5or may function as a substitute for or a supplement to the classic cytochrome b 5 and cytochrome b reductase complex in mediating fatty acid desaturation. The diabetes and lipoatrophy in the knock-out mouse appear to be the consequence of lipotoxicity owing to an excess of saturated fatty acids. This abrupt turn in the direction of our laboratory's research made us newcomers in a highly competitive field. Thus, grant funding became a huge, although temporary, challenge. More importantly, it gave me the opportunity, in the late autumn of my research career, to enter the arcane world of lipid biochemistry and intermediary metabolism, challenges that recharged my somewhat run-down batteries.

Postlude
Four years ago, my research laboratory shut down, and I also stopped seeing both inpatients and outpatients. I currently work about three-quarters time, primarily teaching. I direct two hematology courses for second-year students at Harvard Medical School and also tutor graduate students in the Leder Program, an offshoot of an earlier curriculum that I initiated twelve years ago to introduce Ph.D. students to human biology and pathophysiology of disease. The remainder of my time is devoted to writing (review articles, journal critiques, and a recent book for medical students) and playing the piano.  Reinhold Benesch was fond of saying that "a molecular biologist is someone who practices biochemistry without a license." I plead guilty to having practiced both biochemistry and molecular biology without a license. The extent to which our laboratory and its research have met standards of rigor, reproducibility, originality, and importance can be credited to the wisdom and generosity of mentors mentioned in this Reflections article and in equal measure to a remarkable group of postdoctoral fellows, students, and colleagues who have made my fifty years in research as much fun as my boyhood chemistry set and home laboratory.