From kinase to cancer

I was convinced that the phosphoinositide produced by the oncoprotein-associated PI kinase was structurally distinct from PI-4-P and PI-4,5-P2.¹ I suspected that this enzyme might be placing a phosphate on the 3 position of the inositol ring to produce phosphatidylinsitol-3-phosphate (PI-3-P), but I realized that this would be heretical to the field and would require rigorous chemical proof, since PI-3-P had not been previously described.

So, at the Cold Spring Harbor meeting I sought out Bob Michel and Peter Downes, two of the most experienced biochemists in this field, and told them of our findings. While both were skeptical that a novel phosphoinositide had been missed after more than 35 years of research in this field, they both offered advice about how to prove the structure and Downes agreed to collaborate in this endeavor. With Downes' help, we were able to unambiguously define the structure and thereby reveal a new phosphatidylinositol 3-kinase (PI3K) signaling pathway.² Its significance, however, remained unclear.

The 1980s were an exciting time for studying protein kinases and lipid kinases. Although enzymes that place phosphates on proteins and on lipids were known to exist since the 1950s, the diversity of these enzymes had been dramatically underestimated. By the end of the 1980s it was clear that hundreds of protein kinases existed and that they were playing critical roles in regulation of cell growth, cell cycle, cell morphology, cell migration, and cell metabolism.

Progress in elucidating the role of lipid kinases in cellular regulation lagged far behind the work on protein kinases. It wasn't until the discovery in the early 1980s that PI-4,5-P2 is the precursor of two second messengers (diacylglycerol and inositol-1,4,5-trisphosphate) that there was any interest in characterizing the enzymes that phosphorylate phosphatidylinositol. Even then, most of the interest was in how hormones and growth factors stimulate the lipases that hydrolyze PI-4,5-P2. Since PI-4,5-P2 is present in unstimulated cells and does not dramatically change in response to growth factors and hormones, the kinases that generate this lipid were not thought to play an important role in cellular regulation.

My own interest in PI kinases was piqued by some unpublished results from my laboratory indicating that the activity of the plasma membrane sodium pump was differentially affected by reconstitution in synthetic membranes containing phosphorylated forms of phosphatidylinositol. As an assistant professor at Harvard in the late 1970s and early 1980s, I had been investigating the mechanism by which growth factors and hormones regulate the flux of cations into and out of cells. Growth factor receptors were just being purified at that time, and there was much excitement about the discovery that they had intrinsic protein-Tyr kinase activity.

My focus was on finding a link between the protein-Tyr kinase activity and the changes we were seeing in sodium, potassium, calcium, and proton fluxes across the membrane. Since the receptors, the ion-transport systems, and the lipid PI-4,5-P2 were all intrinsic components of the plasma membrane, I was attracted to the possibility that protein-Tyr kinases directly regulate phosphorylation of phosphatidylinositol and that phosphorylated forms of this lipid regulated the activity of cation transporters. The discovery by another group that PI-4,5-P2 was converted to the second messenger, inositol-1,4,5-trisphosphate, in response to growth factors and that this regulates cytosolic calcium³ further stimulated my interest in this lipid.

It was an attempt to find a direct link between protein-Tyr kinases and phosphatidylinositol kinases that led to the first step in our discovery of PI3K. In 1983, Raymond Erikson moved to Harvard, taking space one floor below me in The Biological Laboratories. His laboratory had purified the pp60v-src protein-Tyr kinase and was characterizing its enzymatic activity. They reported that in addition to the ability to phosphorylate tyrosine residues, the purified pp60v-src preparation could phosphorylate glycerol, which had been added for stabilization during freezing. The importance of this reaction was unclear, since the KM for glycerol was far above physiologic concentrations of this molecule. Upon reading this paper, I became excited about the possibility that the true substrate of this "glycerol kinase" was phosphatidylinositol. The inositol head group of this lipid resembles two glycerol molecules joined head to tail to form a cyclic structure.

In the fall of 1983, a former postdoctoral fellow in my laboratory, Ian Macara, was visiting on his return from a meeting on phosphatidylinositol turnover. Macara, Whitman, and I convinced each other that this wild idea - that the glycerol kinase activity in pp60v-src was really a phosphatidylinositol kinase activity - was worth pursuing. Whitman and I went downstairs to Erikson's laboratory to set up a collaboration to test this idea on purified pp60v-src, and Macara returned to his laboratory at the University of Rochester to test the idea using a different retroviral tyrosine kinase. Both laboratories found that a PI kinase activity had copurified with the protein-Tyr kinases. Importantly, the KM for phosphatidylinositol was 1000-fold lower than that for glycerol, indicating that phosphatidylinositol was likely to be a true in vivo substrate.⁴

While this discovery changed the course of research in my laboratory, we were still quite naive about the complexity of the connection between tyrosine kinases and PI kinases. First, we incorrectly assumed in this first paper that the PI kinase activity was intrinsic to the pp60v-src protein. This idea was supported by several facts: pp60v-src had been purified more than 1,000-fold and appeared as a single band on the gel; a shift to the nonpermissive temperature of a pp60v-src mutant resulted in loss of both protein-Tyr kinase activity and PI kinase activity; and, a pp60v-src inhibitor, quercetin, inhibited the tyrosine kinase activity and PI kinase activity with similar KIs.

The concept that autophosphorylation of protein-Tyr kinases would allow them to form tight complexes with other signaling proteins had not yet been discovered. It was not until our collaboration a year later with Tom Roberts at the Dana Farber Cancer Institute and Brian Schaffhausen at Tufts University that we were able to show that protein-Tyr kinase and PI kinase activities resided on distinct enzymes that formed complexes that depended on the activity of the tyrosine kinase.

This was a difficult time for me, as I was not offered a tenure position at Harvard and was looking for other positions. Other laboratories had attempted and failed to reproduce our results that a PI kinase activity specifically associated with pp60v-src. This was because we performed our PI kinase assays with PI presented in sonicated lipid vesicles while other laboratories, lacking the sonifier needed to make the vesicles, used PI solubilized in detergent micelles as a substrate. PI 3-kinase does not phosphorylate PI presented in detergent micelles so they failed to detect the activity. Several manuscripts that we submitted at that time were rejected based on criticisms suggesting that results from my laboratory could not be trusted. An attempt to obtain funding from the National Institutes of Health for this work met with similar criticisms. I knew, however, that our results were correct, and many terrific people supported me.

In 1985, Irwin Arias, whom I had met when he was doing a sabbatical at Harvard, offered me a full professorship at Tufts' new Department of Physiology, which he was forming. Tom Roberts, Brian Schaffhausen, and I investigated PI kinase activity in a series of transformation-defective mutants of polyoma middle T. Being able to reproduce our findings in the two locations kept us confident that we were on the right track. In the end, our graduate students and postdocs would travel to other laboratories with a sonifier in their suitcases so that they could provide lipid vesicles to other laboratories and allow them to reproduce our findings.

Our discovery in 1987 that this novel enzyme was producing PI-3-P rather than PI-4-P, published in Nature in 1988, came at a time when we were just beginning to persuade some leaders in the tyrosine kinase field that our results were reproducible. However, this claim produced a similar round of skepticism and outright disbelief from the inositol lipid biochemists. I was told that one leader in this field would ?eat his hat' if our claim turned out to be correct.

Our attempt to publish the purification of PI3K to homogeneity in the Journal of Biological Chemistry⁵ was held up by a reviewer for nearly a year, not because of any significant criticism about whether we had purified the enzyme, but from criticism that our previous claims of the importance of this enzyme were not believable. Fortunately, our findings were quickly reproduced and accepted by a group of outstanding inositol lipid biochemists in England. By 1990 our results were generally accepted and I was able to get funding from the NIH to support this work. Before that, I kept the work going mostly from funds I had received to support our work on regulation of cation transport systems.

Along the way, we had refined our findings. It turns out, for example, that we were also naive about the complexity of phosphatidylinositol phosphorylation in 1984. Since the only phosphoinositides known to exist in the mid-1980s were PI-4-P and PI-4,5-P2, we assumed that the pp60v-src-associated PI kinase was producing PI-4-P. It wasn't until we became serious about purifying this enzyme that we realized that there were multiple phosphatidylinositol kinases in cells and that the ones that associated with protein-Tyr kinases were responsible for the phosphorylation of the 3 position of the inositol ring that we had stumbled upon in 1987 before that critical Cold Spring Harbor meeting. Further characterization of the pp60v-src-associated and polyoma middle T-associated PI3K by Leslie Serunian, a postdoctoral fellow, and Kurt Auger, a graduate student in my laboratory, revealed that this enzyme not only converted phosphatidylinositol to PI-3-P but also converted PI-4-P to PI-3,4-P2.

A bit of serendipity led to our discovery that PI3K could convert PI-4,5-P2 to PI-3,4,5-P3. Leslie was attempting to prove that a single enzyme in the polyoma middle T complex was responsible for converting phosphatidylinositol to PI-3-P and PI-4-P to PI-3,4-P2, but she was having trouble reproducing her findings. She showed me a series of thin-layer chromatograms where a radioactive spot appeared at the expected location of PI-3,4-P2 when PI-4-P was used as a substrate of the enzyme, and then a series of thin-layer chromatograms where no spot appeared. On careful examination, we noticed that a new radioactive spot appeared very close to the origin of the thin-layer chromatogram in the experiments where PI-3,4-P2 failed to be produced. I suggested that she stain the thin layer with iodine to visualize the (nonradioactive) substrate to make sure that PI-4-P was really present in the assay.

The staining revealed that the substrate did not migrate at the expected position of PI-4-P, but rather migrated at the position of PI-4,5-P2. The error had not been made by Leslie but by the manufacturer of the phosphoinositide, Boehringer Mannheim. Leslie had purchased a new bottle of PI-4-P and had switched to this substrate in the middle of her experiments. The company had mislabeled the bottle: The phosphoinositide inside was PI-4,5-P2, and this was being phosphorylated to produce the slowly-migrating spot on the thin layer. This mistake was fortunate for us because it led to the discovery of PI-3,4,5-P3 (PIP3).

Importantly, Leslie and Kurt showed that PI-3,4-P2 and PI-3,4,5-P3 were nominally absent in quiescent cells, appeared acutely in response to growth factors, and were constitutively elevated in transformed cells.^6,7 In addition, Leslie showed that these lipids were not substrates of the phospholipases that hydrolyze PI-4,5-P2 , and thus were unlikely to be involved in the production of soluble inositol polyphosphate such as IP3 and IP4, which were also emerging as regulators of cell function. By 1990 we were convinced that PI-3,4-P2 and PI-3,4,5-P3 were membrane-embedded second messengers (like diacylglycerol) that regulate cell growth and that, when overproduced, result in cell transformation.

My original motivation in pursuing the PI kinase activity associated with protein-Tyr kinases was to test the idea that synthesis of the lipids PI-4-P and PI-4,5-P2 played a critical role in regulation of cation transport systems in response to hormones and growth factors. However, our discovery of the new set of lipids phosphorylated at the 3 position of the inositol ring led us in a new direction. Ultimately I shifted my limited resources to pursue the importance of these lipids for cell growth and cell transformation, and I discontinued the work on regulation of cation-transport systems. Research over the past decade has, perhaps ironically, revealed that several plasma membrane cation transporters and channels are indeed directly regulated by contact with PI-4,5-P2 or by PI-3,4,5-P3.

When I moved my laboratory to the Beth Israel Hospital and Harvard Medical School in 1992, I decided to focus the research on elucidating the PI3K pathway. By that time the importance of this pathway in growth factor signaling was becoming generally accepted, and many large, well-funded laboratories had joined in and were beginning to make important contributions.

Some of the most compelling findings of the past decade, in our lab and others, have linked growth factor receptor stimulation of PI3K to activation of mTOR, Ser/Thr kinase and p70S6kinase.^8,9The collaboration with John Blenis that led to this link was particularly rewarding for me, because I have known John since 1983 when he was a postdoc in Erikson's laboratory. Even in those days John was trying to find the biochemical pathway that would link the activation of protein-Tyr kinases to the phosphorylation of ribosomal protein S6. He chose to start with S6 and work his way back, while I started with PI3K and worked my way forward. We played basketball together for many years, and debated models in the locker room. Had we realized how complicated it was and that it would take 20 years to link these observations, we might have given up.

Of course, many other laboratories made key discoveries that contributed to the signaling network that is now known to exist downstream of PI3K. This network participates in the control of cell growth, cell cycle entry, cell survival, cell metabolism, cell migration, and organism lifespan. Much of the network is conserved back to flies and worms and even to the yeast, Schizosaccharomyces pombe. Genetic studies of these organisms have strongly complimented and in some cases guided the biochemical studies in my laboratory.¹⁰ Yet, there are many aspects of the network that we don't yet understand.

In recent years my laboratory has become interested in the biochemical mechanism by which growth factor receptor protein-Tyr kinases and the PI3K pathway control normal cell growth, and how disregulation of this pathway results in cancers. In a collaboration with Seigo Izumo's laboratory we showed that transgenic expression of activated PI3K in cardiac myocytes results in enlarged hearts due to increased growth of individual myocytes.¹¹ This observation was consistent with studies in flies, in which modulation of PI3K affected cell size. We focused our attention on trying to understand how this pathway affected cell growth.

The critical question is: What are the outputs from this network that execute the growth response? As we learn more about transcriptional regulation downstream of PI3K, it is clear that this pathway controls genes that mediate glucose uptake and metabolism. Ironically, some of the earliest investigations into differences between tumors and normal tissues revealed changes in glucose uptake and in the ratio of lactate production to oxygen consumption, the so-called Warburg effect (based on the findings of Otto Warburg in 1929). For much of the first half of the 20th century, the focus of oncology research was to explain why tumors have increased glucose uptake, increased protein synthesis, and increased lipid synthesis compared to normal tissues. Interest in these questions waned as the genetic basis of cancers became revealed through studies of tumor-promoting retroviruses and endogenous oncogenes over the past 30 years.

As we finally begin to tease out the functions of these oncogenes, we begin to learn that many are in pathways that control cell metabolism. We have come full circle. Inhibitors of protein tyrosine kinases are already approved for treating cancers, and inhibitors of PI3K have gone into clinical trials. Our results suggest that drugs already approved for treating metabolic diseases, such as metformin (see "References