Cloning Expeditions: Risky but Rewarding

Abstract

In the 1980s, a good part of my laboratory was using the then-new recombinant DNA techniques to clone and characterize many important cell surface membrane proteins: GLUT1 (the red cell glucose transporter) and then GLUT2 and GLUT4, the red cell anion exchange protein (Band 3), asialoglycoprotein receptor subunits, sucrase-isomaltase, the erythropoietin receptor, and two of the subunits of the transforming growth factor β (TGF-β) receptor. These cloned genes opened many new fields of basic research, including membrane insertion and trafficking of transmembrane proteins, signal transduction by many members of the cytokine and TGF-β families of receptors, and the cellular physiology of glucose and anion transport. They also led to many insights into the molecular biology of several cancers, hematopoietic disorders, and diabetes. This work was done by an exceptional group of postdocs and students who took exceptionally large risks in developing and using novel cloning technologies. Unsurprisingly, all have gone on to become leaders in the fields of molecular cell biology and molecular medicine.

INTRODUCTION

Throughout my entire adult life, I have been fascinated by red blood cells and by cell membranes and their proteins. It all began in the summer of 1958, after 11th grade, when I spent the first of three summers at Western Reserve (now Case Western Reserve) Medical School with Robert Eckel studying potassium transport in red blood cells. We were trying to determine the intracellular glycolytic intermediates that powered K⁺ uptake, and among other techniques I used flame photometry to measure the K⁺ concentration in red cells. This led to my first scientific publications (1, 2), and I have had membranes and red blood cells constantly on my mind ever since!

But first I took a detour, as I majored in mathematics and chemistry at Kenyon College. My Ph.D. thesis under Norton Zinder at the Rockefeller focused on a genetic analysis of the RNA bacteriophage f2, generating and analyzing amber (nonsense) and temperature-sensitive mutants; I identified mutations in three phage genes—for the coat protein, a subunit of the RNA polymerase, and an assembly protein. My work as a postdoctoral fellow under Sydney Brenner and Francis Crick focused on understanding the regulation of translation of the three f2 genes (3–5), and my early work as a Massachusetts Institute of Technology (MIT) faculty member focused on the mechanism and regulation of initiation of translation of the α and β globin genes (6, 7). I recently reviewed these projects in a “Reflections” piece in the Journal of Biological Chemistry (8); I realized I am indeed joining the “senior scientist” set when I was asked to write this piece, yet another reminiscence!

BIOGENESIS OF MEMBRANE PROTEINS: THE 1970s

Whether by accident or design I still do not know, but upon arrival at MIT I was given an office next door to David Baltimore, an old friend from Rockefeller days, and we shared three large research laboratories. David and his postdoc (then wife) Alice Huang introduced me to the study of vesicular stomatitis virus (VSV). One VSV gene, encoding the G protein, or glycoprotein, became invaluable in studies David Knipe carried out in the early 1970s defining the endoplasmic reticulum (ER)-to-Golgi compartment-to-plasma membrane pathway for biosynthesis of the G protein as a model for all cell surface glycoproteins (9–11). Later, in collaboration with Günter Blobel's group, Flora Katz and Jim Rothman developed in vitro cell-free protein-synthesizing systems where they could translate the VSV G mRNA and insert it into ER membranes (12). Jim then used this system to demonstrate obligatory cotranslational insertion of this transmembrane glycoprotein into the endoplasmic reticulum membrane and cotranslational attachment of the two asparagine-linked oligosaccharides (13, 14).

Contemporaneously, we worked on the biogenesis of several erythrocyte “membrane” proteins—that is, the major proteins in a purified red cell membrane pellet, or “ghost.” We showed that several proteins, now known to be cytoskeletal proteins, are made on membrane-free polysomes (15, 16). One of my favorite experiments demonstrated that the major red cell membrane and cytoskeleton proteins are made at different times during development (17). This involved injecting a live mouse with several millicuries of [³⁵S]methionine (the pulse), then (chase) bleeding it every ∼12 h for a few days, and preparing membrane ghosts followed by SDS gel electrophoresis and autoradiography. The logic was that the last proteins to be made during the multiday developmental period would be the first to be found in mature red cells released into the blood. Old-timers will recognize this as a whole-organism version of the “Dintzis” experiment (18).

CLONING BY ANTIBODIES: LAMBDA GT11

But to further understand how membrane proteins were made, we needed to know the amino acid sequences of normal cell membrane proteins and to have molecular techniques to study individual membrane proteins in detail. In these early days, the only membrane protein whose sequence was known was glycophorin, and no membrane protein had been cloned (remember, even globin mRNAs were cloned only in 1977!). So I picked three membrane proteins that are expressed in reasonable abundance—the liver asialoglycoprotein receptor studied by Gil Ashwell and two major red cell membrane proteins, namely, Band 3, the anion exchanger that could be more or less purified by SDS-PAGE of red cell ghosts, and the glucose transporter that had been purified by Gus Lienhard. But aside from doing a lot of peptide purification and amino acid sequencing and making degenerate pools of DNA to screen a cDNA library by hybridization, which was the only existing cDNA cloning technology and a forbidding undertaking with little guarantee of success, there was no obvious way to clone these proteins.

The solution came from Rick Young, a postdoc at Stanford whom we were recruiting to the MIT and Whitehead faculty, who developed a general technique for using antibodies as probes for proteins encoded by cloned DNA (Fig. 1). His vector, lambda gt11, allowed insertion of libraries of foreign DNA in frame into the beta-galactosidase structural gene lacZ and promoted synthesis of chimeric proteins. Induction of lysogens of these libraries produced large quantities of antigens, and clones could be detected by blotting filters of colonies with specific antibodies (19). However, no one had actually used these technologies to clone mammalian proteins and we were unaware of the technical hurdles that awaited us. But three postdocs—Mike Mueckler, Ron Kopito, and Martin Spiess, all of whom joined my laboratory in the early 1980s—were up to the challenge, especially given all of the help Rick provided us.

Fig 1.

Use of λ expression cloning to identify a cloned DNA based on binding of the encoded protein to a specific antibody. The λgt11 vector was engineered to express the E. coli protein β-galactosidase at high levels. The only EcoRI recognition site (red) in this vector lies near the 3′ end of the β-galactosidase gene. If a cDNA, or protein-coding fragment of genomic DNA, is inserted into this EcoRI site in the correct orientation and proper reading frame, it is expressed as a fusion protein in which most of the β-galactosidase sequence is at the N-terminal end and the protein sequence encoded by the inserted DNA is at the C-terminal end. Plaques resulting from infection with recombinant λgtll contain high concentrations of such fusion proteins. These proteins can be transferred and bound to a replica filter, which then is incubated with an antibody (blue) that recognizes the protein of interest. Rinsing the filter washes away antibody molecules that are not bound to the specific fusion protein attached to the filter. Bound antibody is usually detected by incubating the filter with a second radiolabeled antibody that binds to the primary antibody. Any signals that appear on the autoradiogram are used to locate plaques on the master plate containing the gene of interest. (Adapted from reference 65 with permission.)

GLUCOSE TRANSPORTERS

Using an antibody provided by Gus, Mike Mueckler showed that the “erythrocyte glucose transporter” was expressed in the HepG2 hepatoma cell line. Then, from a HepG2 cDNA library he made in lambda gt11, Mike was quickly able to clone the protein now known as GLUT1. But DNA sequencing was still in its infancy—we did this in our own laboratory using the Sanger technique and with huge gels that were dried and subjected to autoradiography—and the errors in reading the gel autoradiograms were many. We were afraid that our initial sequence contained errors and thus that our derived amino acid sequence would be wrong. So we enlisted the help of Howard Morris and his group in London, one of the pioneers of protein mass spectrometry (another technique in its infancy). They obtained sufficient amino acid sequences from peptides generated from Gus's purified protein to show that our derived amino acid sequence was correct. Indeed, the amino acid sequence and the 12 membrane-spanning alpha helixes we predicted have stood the test of time (20).

Jeff Flier, a visiting scientist, then worked with Mike to show that fibroblast cell lines transformed with activated ras or src oncogenes, as well as several tumor cell lines, had elevated levels of glucose uptake and, importantly, elevated GLUT1 expression. This provided the first partial explanation of the Warburg effect—that tumor cells have high rates of aerobic glycolysis fueled by glucose imported by the additional GLUT1 protein (21).

Bernard Thorens was struck by the fact that Mike and Jeff had shown that GLUT1 was expressed, inter alia, in brain, kidney, and muscle but not in liver, intestine, or the islets of Langerhans, cells well known to take up glucose from the blood. By using low-stringency hybridization with a GLUT1 cDNA, Bernard was able to isolate from a rat liver cDNA library a cDNA encoding a protein—now known as GLUT2—that was 55% identical in sequence to the rat brain transporter and with a superimposable hydropathy plot. With Ron Kaback's help, Bernard expressed this protein in an Escherichia coli mutant defective in glucose uptake and showed that it was incorporated into the bacterial membrane and functioned as a glucose transporter (22). This began a long series of studies he conducted in my laboratory and his own on glucose transport and metabolism, especially in liver and pancreatic beta cells (23–25).

Maureen Charron, a new postdoc, suspected that neither GLUT1 nor GLUT2 was the transporter in adipocytes and muscle that in response to insulin signals moved from intracellular membranes to the cell surface membrane to increase the rate of glucose uptake. Using a rat GLUT1 probe and low-stringency hybridization, Maureen isolated a cDNA clone from a rat soleus lambda gt10 cDNA library whose sequence and predicted membrane structure were very similar to those of GLUT1 and GLUT2 (26). This protein, now known as GLUT4, was contemporaneously cloned by several other groups (27, 28) and was expressed predominately in tissues where glucose transport is sensitive to insulin, including striated muscle, cardiac muscle, and adipose tissue. This led to a long series of studies in my own laboratory on the regulation of glucose transport and glucose transporters in adipose and muscle cells, much of it in collaboration with Barbara Kahn (29–31).

ANION EXCHANGE PROTEINS

As part of a study on biosynthesis of Band 3, the major integral red cell membrane protein, Bill Braell, a Ph.D. student, had made a good antibody to the murine protein (32). Ron Kopito generated a cDNA library from anemic mouse spleen (an erythroid organ) in Rick's lambda gt11 vector and screened it with Bill's antibody. The amino acid sequence of murine band 3, deduced from the nucleotide sequence of the cDNA clone, confirmed that this membrane glycoprotein is composed of two major structural domains that correlate with its dual functions; its cytosol-facing N-terminal hydrophilic domain functions as the anchor for the erythrocyte cytoskeleton, and the hydrophobic C-terminal domain functions as the plasma membrane anion antiporter important for CO₂ transport by the red blood cell (33). Ron predicted that this highly hydrophobic domain crosses the plasma membrane at least 12 times, likely as alpha helixes, a prediction that is still thought to be correct but that awaits determination of a three-dimensional structure (34). Sam Lux, a visiting scientist and hematologist, then worked with Ron to clone human Band 3 (35). Human Band 3 was independently cloned and sequenced by Mike Tanner's group in Bristol (36).

Seth Alper, a nephrologist postdoc, worked with Ron Kopito and used low-stringency hybridization to clone an ortholog of Band 3 expressed in the kidney and other epithelial tissues (37). They also identified a splice variant of Band 3 highly expressed in the kidney (38).

Most importantly, they collaborated with Dennis Brown at Massachusetts General Hospital (MGH) to determine the cellular distributions of the kidney form of Band 3 in rat kidney collecting duct. In the medullary collecting duct, almost all intercalated cells expressed Band 3 on the basolateral membrane and the H⁺-ATPase on the apical membrane; these undoubtedly were the type A cells that secrete acid (HCl) into the kidney tubules as part of the regulation of blood pH (33). Thus, we had helped elucidate an important mechanism by which cells secrete acid, again one that has stood the test of time.

Other epithelial cells had basolateral H⁺-ATPase and no detectable Band 3; we recognized these as likely to be the bicarbonate-secreting cells. Years later, others showed that these cells express the apical anion antiporter pendrin, a member of the distinct SLC26 family of anion transporters. And in short order I actually became an expert in renal physiology (39, 40) since these were some of the first studies using molecular probes to elucidate fundamental aspects of kidney function.

THE ASIALOGLYCOPROTEIN RECEPTOR AND INTERNAL SIGNAL-ANCHOR SEQUENCES

Gilbert Ashwell had extensively characterized a galactose lectin on liver cells—the asialoglycoprotein receptor, so called since it bound and internalized glycoproteins from which the terminal sialic acid residues had been removed, exposing penultimate galactose residues (41). Alan Schwartz, a postdoc, purified this receptor from rat liver by solubilization and affinity chromatography on asialoorosomucoid-Sepharose and generated a very good receptor-specific antibody (42). In collaboration with Hans Geuze, Alan used double-label immunoelectron microscopy on ultrathin cryosections of rat liver to identify the site at which the asialoglycoprotein receptor and its ligand dissociate following their common endocytosis. We called this organelle CURL (compartment of uncoupling of receptor and ligand), but it is now known as the recycling endosome (43).

Alan had shown that the asialoglycoprotein receptor was highly expressed in the hepatoma cell line HepG2 (44), and then Martin Spiess, another postdoc, used Alan's antibody and Mike Mueckler's cDNA library from HepG2 cells in the lambda gt11 expression vector to clone one receptor subunit. The deduced amino acid showed that there was no cleaved N-terminal “signal sequence” then thought to be essential for insertion of all integral membrane proteins into the endoplasmic reticulum membrane. Rather, there was just one hydrophobic segment, residues 41 to 59, in the 291-amino-acid protein. We hypothesized that this was “an internal signal sequence, probably the membrane-spanning segment, … assumed to direct insertion of the carboxyl-terminal ligand binding portion of the receptor across the endoplasmic reticulum membrane” and that the hydrophilic amino terminus faces the cytoplasm, and the carboxyl terminus is exoplasmic (45). Indeed, Martin went on to show using our cell-free system that the membrane-anchor domain is both necessary and sufficient for membrane insertion of the nascent protein and is thus a signal-anchor sequence now known to be employed by all type II membrane proteins (33, 46).

Martin went on to clone a second subunit (H2) of the asialoglycoprotein receptor from the same HepG2 cDNA library, with a protein sequence homology of 58% to the first subunit, then termed H1 (47). Joyce Bischoff, another postdoc, went on to generate antipeptide antibodies that were specific for each polypeptide. A series of elegant protein chemistry experiments indicated that both subunits were components of the functional receptor (48), and thus, the asialoglycoprotein receptor became one of the earliest and best-characterized cell surface receptor proteins (49).

SUCRASE ISOMALTASE

Martin Spiess had done his Ph.D. thesis at the Eidgenössische Technische Hochschule (ETH) Zürich under Giorgio Semenza. Giorgio had completed an extensive series of studies on the structure and biogenesis of sucrase isomaltase, a major intrinsic protein in the apical plasma membrane of absorptive intestinal epithelial cells, where it plays a key role in the final steps of digestion of glycogen and starch. It is a large protein—260,000 Da—that reaches the apical membrane and is then split by pancreatic enzymes into a membrane-anchored isomaltase subunit and a soluble sucrase subunit that remains noncovalently bound to the isomaltase subunit. The plan was for Martin to return to Semenza's Biochemistry Department at the ETH and for the two to jointly supervise a graduate student in experiments to clone sucrase isomaltase.

So in the early fall of 1985, a new Ph.D. student at the ETH, Walter Hunziker, joined my laboratory to be trained in cloning technologies by Martin; after a year Walter was to return to the ETH to do the actual cloning with Martin. But we had both a good antibody to sucrase isomaltase and some cDNA purified from rabbit intestine. (I was told confidentially, but cannot verify, that someone had gone into the woods near Zürich, shot a rabbit, and then quickly extracted the small intestine and purified its RNA.) So Walter, who mastered all of the cloning technologies in a few weeks, started cloning sucrase isomaltase. Very quickly, he had generated the complete sequence of the protein—1,827 amino acids with 12 amino acids at the N terminus followed by a single 20-amino-acid segment that spans the bilayer once and serves as an internal signal anchor sequence. This was followed by a 22-residue serine/threonine kinase-rich, probably glycosylated stretch, presumably forming the stalk on which the globular, catalytic domains are directed into the intestinal lumen. There was a high degree of homology between the isomaltase and sucrase portions (41% amino acid identity), indicating that prosucrase isomaltase evolved by partial gene duplication.

Strikingly, there was no cleaved N-terminal signal sequence, which contradicted much published work by the Semenza group, but a few experiments quickly confirmed that Walter's derived amino acid sequence was correct. Walter spent only 1 year at the ETH finishing a few experiments for his paper (50) and received his Ph.D. in record time—under 2 years—and received a gold medal besides.

SOME EARLY LESSONS

As should be obvious, the successes of these early cloning projects depended on a collaboration with a faculty colleague and on postdocs who not only were smart and talented experimentally but also were willing to take risks—to use really unproven cloning technologies and work through all of the inevitable technical problems that arose.

These cloning expeditions led me to realize the enormous impact that molecular studies of key membrane proteins were going to have, not only for understanding of basic cell physiology but also for understanding of the molecular basis of human disease—including diabetes, the metabolic syndrome, and kidney disorders—and also for the eventual development of novel therapies.

FUNCTIONAL EXPRESSION CLONING OF THE ERYTHROPOIETIN AND TRANSFORMING GROWTH FACTOR β RECEPTORS—A JOURNEY INTO THE UNKNOWN

Despite the obvious risks and technical difficulties surmounted by the students and fellows I discussed above, in hindsight cloning the glucose transporters, band 3, the asialoglycoprotein receptors, and sucrase isomaltase was relatively easy. We at least knew some things about the protein we wanted to clone, such as its molecular size and its tissue distribution, and had some reagent to use, such as an antibody or a partial amino acid sequence.

In the case of the erythropoietin and transforming growth factor β (TGF-β) receptors, we had nothing.

EXPRESSION CLONING OF THE Epo RECEPTOR

By the mid-1980s, several companies had cloned the gene encoding erythropoietin (Epo) and expressed it by recombinant DNA techniques in cultured mammalian cells. Epo is the singular hormone that prevents apoptosis of terminal erythroid progenitor cells (the CFU-E cells) and promotes their terminal division and differentiation into erythrocytes. Most kidney dialysis patients and many cancer patients are anemic, and the anemia can be reversed by injection of recombinant Epo. Epo was well on its way to becoming a multi-billion-dollar-a-year antianemia drug, and there was much interest in cloning the Epo receptor.

Alan D'Andrea, an MD postdoc, set out to do this in my laboratory. But the only way Epo receptors could be reliably detected on cells was by binding of ¹²⁵I-labeled Epo of very high specific activity. Several groups tried to identify the Epo receptor by cross-linking bound radioiodinated Epo, but the results suggested the existence of multiple receptors that differed in different cell types, and none of these was certain to be the Epo receptor.

Alan started with an erythroleukemia cell line that had about 1,000 Epo binding sites on its surface with a tight binding affinity (K_D [equilibrium dissociation constant] = ∼200 pM). He recruited Gordon Wong, a scientist with experience in cDNA library construction working at the Genetics Institute (one of the companies that cloned Epo and that was engaged in a furious patent war over the drug.) The procedure they adopted became extremely complicated and difficult to execute but was easy to describe in a textbook (Fig. 2). The expression vector contains the simian virus 40 (SV40) origin of DNA replication and a strong promoter that drives expression of the inserted cDNA; after transfection into COS cells (which express the SV40 T antigen), the plasmid gets replicated multiple times. The idea was that cells expressing the desired Epo receptor cDNA would become very radioactive when incubated with ¹²⁵I-labeled Epo, but there was considerable background binding of ¹²⁵I-labeled Epo to nonexpressing COS cells. A culture dish transfected with the pool of 1,000 cDNAs that ultimately contained the Epo receptor generally had ∼1,200 cpm of bound ¹²⁵I-labeled Epo whereas plates of cells transfected with control cDNAs had ∼900 cpm. It required multiple repeats to be certain that the signal really was significantly above the background, and several times Alan became frustrated and wanted to give up. Expression cloning is hard to troubleshoot since you know that the technology is working only when you have the actual clone in hand. I was convinced that this insane technology would eventually work, and my role in the project became mainly that of cheerleader and statistician.

Fig 2.

Identification and isolation of a cDNA encoding a desired cell-surface receptor—here the erythropoietin receptor—by plasmid expression cloning. mRNA is extracted from cells that normally express the receptor—in this case an erythroleukemia cell line—and reverse transcribed into double-stranded cDNA. (a) The entire population of cDNAs is inserted into plasmid expression vectors between a strong promoter and a terminator of transcription. The plasmids are transfected into bacterial cells that do not normally express the receptor of interest. The resulting cDNA library is divided into pools, each containing about 1,000 different cDNAs. (b) Plasmids in each pool are transfected into a population of cultured cells (e.g., COS cells) that lack the receptor of interest. Only transfected cells that contain the cDNA encoding the desired receptor synthesize it; other transfected cells produce irrelevant proteins. To detect a pool containing the 0.1% of cells that express the desired receptor, a radiolabeled ligand specific for the receptor (here Epo) is added to the culture dishes containing the transfected cells; the cells are incubated for ∼30 min to allow receptor-mediated endocytosis of receptor-bound hormone. The cells are then fixed, and the amount of radioactivity on the entire plate is determined by scintillation counting. Most plates contain cells that do not express the receptor and have background levels of radioactivity; pools with positive clones have slightly more radioactivity bound to them, since positive cells synthesizing the specific receptor should have a lot of bound hormone. Plasmid cDNA pools giving rise to a positive signal are maintained in bacteria and subdivided into smaller pools, each of which is rescreened by transfection into cultured cells. After several cycles of screening and subdividing positive cDNA pools, a pure cDNA clone encoding the desired receptor is obtained (66, 67). (Modified from reference 68 with permission.)

Thankfully, after about 2 years of constant work Alan had cloned the Epo receptor (67). The derived amino acid sequence indicated a 507-amino-acid protein with a single membrane-spanning segment but with no sequence resemblance to any protein in any database. There were no sequences indicative of a kinase or other enzymatic domain. Fortunately, Gerry Fasman was a sabbatical visitor in the laboratory. The beta chain of the interleukin 2 (IL-2) receptor had just been cloned, and Gerry and Alan were able to show that the IL-2 and Epo receptor sequences had significant sequence homology; this result—published in the shortest Cell paper in memory, 1.5 pages—defined the cytokine receptor superfamily (51). This of course led to an explosion of research in many laboratories, including my own, on the mechanism of hormone binding to these receptors and the identification of downstream signal transduction pathways, including, inter alia, the new JAK-STAT pathway (52–55).

THE TGF-β RECEPTORS

With so many membrane proteins being studied in my laboratory, I had not paid much attention to receptors and signal transduction for the large TGF-β, activin, and bone morphogenetic protein (BMP) family of hormones. I did know that TGF-β inhibited the growth of many types of cells, mainly by inducing expression of genes encoding inhibitors of the cell cycle. I knew that there were at least three receptor subunits, since several groups had shown that radioiodinated TGF-β could be cross-linked to three proteins on the cell surface, uninspiringly labeled Band 1, Band 2, and Band 3. I also knew that my colleague Bob Weinberg was getting very interested in the TGF-β signaling pathway since many cancer cells were resistant to growth inhibition by TGF-β and seemed to lack one or more of these “Band” proteins. Xiao-Fan Wang, a postdoc in Bob's laboratory, was trying—unsuccessfully—to clone the major TGF-β binding protein, Band 3, by generating and sequencing tryptic peptides and making degenerate oligonucleotides to screen a cDNA library.

Until the clones were in hand, I actually did not know that Herbert Lin, an MD/Ph.D. student in my laboratory, was involved in this work. Herb had already set up our expression cloning system and, almost as a laboratory rotation project, used it to clone two G protein-coupled receptors—an adenylate cyclase-coupled calcitonin receptor and the endothelin receptor (56, 57). At Whitehead, we encourage our students to establish research collaborations with other laboratories, so without telling me Herb started working with Xiao-Fan to use an expression cloning strategy and radioiodinated TGF-β to clone Band 3, now known as betaglycan (58). Betaglycan was cloned contemporaneously and independently by Joan Massagué's group (59). This protein had a very short cytosolic domain and did not appear to be a signaling receptor; in fact, subsequent work in my laboratory showed that Band 3 expression facilitated the binding of TGF-β to the two other receptor subunits.

Moving quickly, Xiao-Fan and Herb then used a cell line that lacked expression of “Band 3” to generate a cDNA library and used the same expression cloning strategy to isolate the cDNA encoding “Band 2,” now known as TGF-β receptor type II (60). The TGF-ß type II receptor and the activin type II receptor, cloned contemporaneously by Joan Massagué's laboratory, turned out to be the first receptor serine kinases known (61, 62). These studies of type II receptors, and the subsequently cloned type 1 receptors, also receptor serine kinases, led to a flurry of activity in studies of mice, humans, and Drosophila to elucidate the downstream signaling molecules, culminating with the elucidation of the Smad signaling pathway. These, in turn, led to major insights into the mutations in this pathway contributing to oncogenesis; loss of Smad 4, for instance, was soon shown to occur in many pancreatic cancers.

MANAGING GRANTS AND POSTDOCS

All of these cloning projects were risky and without any guarantee that they would work. Thus, it should not surprise anyone that I had no grants from NIH to support any of these cloning projects. In fact, during my lectures at the time on the cloning and characterization of the Epo receptor I would show a slide entitled “Ten reasons why NIH would not support the Lodish laboratory to clone the Epo receptor.” The slide included statements like, “Lodish has no track record in the field.” “He has never made radioiodinated Epo and it is not clear he can do the binding assays.” “The expression cloning technology is unproven.” “It is well known that the Epo receptor has multiple subunits and expression cloning cannot be used to clone just one of them.” Needless to say, all of these points were nonsense, and yet there was no way any grant review panel then—or now—would approve a grant application with so many uncertainties.

Fortunately, I was able to assemble some outside funds and also to convince the program directors of my NIH grants to divert a few dollars toward these cloning projects. Of course, once we had these clones in hand I was able to get large grants from NIH to work on them; almost immediately after publishing the findings on the TGF-β receptors, for instance, I was awarded a large grant from the NCI to study TGF-β receptor interactions and downstream signaling pathways.

The moral here is clear; to get ahead in science—both as a postdoc and as a principal investigator—one occasionally must take on high-risk projects with potentially very high payoffs. Another important lesson I learned from these cloning expeditions was to let my postdocs take with them a large part of their research projects when they leave my group and start their own laboratories; this turned out to be a great attraction for recruiting to my laboratory extraordinarily talented and ambitious young researchers. And as a consequence, every few years I could reinvent part of my laboratory and start work in another area and take on another risky project with another talented student or postdoc. Thus, and as I wrote in a successful renewal of my NIDDK Merit Award, “I have been able to attract to my laboratory a spectacular group of postdoctoral fellows to work on these and other projects. Part of the inducement to come here is that I allow postdoctoral fellows to take with them major parts of their project when they establish their own academic research laboratories. As a consequence, many of the above projects are not continuing in my laboratory but are in the very good hands of these former fellows.” Another advantage of encouraging my former postdocs and not competing with them is that we remain in close touch and occasionally even collaborate (Fig. 3).

Fig 3.

Gathering at my house in 2009 of locals who were in my laboratory in the late 1980s to meet Bernard Thorens. Bernard was spending a sabbatical with former collaborator Barbara Kahn at the Beth Israel Deaconess Medical Center. Left to right: Naomi Cohen (my long-time laboratory manager), Kathryn John, Mike Shia, Chris Hwang, Bernard Thorens, Alan D'Andrea, Seth Alper (immediately below Alan), Joyce Bischoff, myself, Herbert Lin, Sam Lux, and Barbara Kahn.

I was deeply honored by the 2010 Mentoring Award from the American Society of Hematology because I was nominated and supported by many former fellows working in hematology. And yes, indeed I have been working on red blood cells for over 55 years, and the work is still exciting. New technologies, such as expression cloning and cloning by antibodies, are often what drives new discoveries and major advances in science. Indeed, using new deep RNA sequencing techniques we are discovering a large number of long noncoding RNAs (lncRNAs) that are expressed only in red cell progenitors and that are essential for normal erythropoiesis (63, 64). There are many more risky projects to undertake.