Transcription
Thank you very much for the introduction, and to all of you for coming here this afternoon. Ralph LT was a friend of mine throughout almost all of my career in diabetes research. Whenever I would come to Stockholm, I would let him know well in advance, in the hope that he could allocate a few hours to hang out with me and talk about our various interests in the diabetes field in general. This took place in his home, where he always had a plate of delicious gravlax, which he made himself, or so he claimed. On this visit to Stockholm, I won't have that pleasure, but I do have the great honor of giving this lecture in his memory.
I want to address a problem for science in general, a problem that we refer to as Dogma displacement inertia, or DDI. By that, I mean that the flood of new information and facts in all branches of science far exceeds the ability of the scientists to assimilate and use the new information to displace dogmas that are no longer valid. The most extreme example of that was in the field of astronomy, where the dogma of geocentrism, born at the time of Aristo, was challenged and disproven in the 14th and 15th century by Copernicus and Galileo, but was not displaced for 450 years after that. So I have no illusions that in the next 45 minutes I will be able to displace any diabetes dogma that came into this room with you a few moments ago, but there's no harm in [Music] trying.
Diabetes dogmas are particularly difficult to displace because many of them arose in 1922 as a result of a near-magical moment in the history of medicine, when Banting, Best, and Collip injected Canadian children dying in diabetic ketoacidosis with the first insulin preparations. On this photograph on the left, you see a three-year-old boy dying in diabetic ketoacidosis, weighing only 15 lbs. He was injected with insulin 5 days before Christmas, and two months later you can see he's a little overweight, but otherwise a healthy little, um, person who can look forward to years of near-normal life. This event was so dramatic that within 360 days of the first injection, Frederick Banting was in Stockholm to receive the Nobel Prize. Had Frederick Banting worked in 2014 in in our institution, he would have been lucky to get IRB approval to give the injection in 360 days.
Now let's address one of the classic dogmas of type 1 diabetes that arose from Banting's work, and that is the idea that insulin is essential for life because it is the direct regulator of all metabolism. Well, this dogma is disproven in the very first slide. Here you see two mice, both of them completely insulin deficient, without any detectable insulin or C-peptide in the circulation. The mouse on the left is obviously very sick; you don't need a statistical analysis to show—showed that the P value is in the sawdust—indicating severe gly- uh, polyuria. Both of them have islet pumps; the one on the right contains only buffered saline, the one on the left contains a glucagon suppressor. And if we look at glucagon levels in the two animals, we see the sick animal has very marked hyperglucagonemia, while the one on the left with the glucagon suppressor has normal plasma glucagon levels. So this difference was so remarkable, we decided to uh, explore this in further detail, and to do this we obtained glucagon receptor knockout mice from Marc Shroder at Albert Einstein School of Medicine in the Bronx, and those mice came to Dallas. We sacrificed some of them to measure the glucagon receptor messenger RNA in their liver and found it virtually undetectable; their glucose levels were normal. So we then tried to destroy all their beta cells by giving double or triple doses of streptozotocin, and when we did that we saw no increase in their glucose levels, even though wild-type mice treated with this amount of streptozotocin developed severe ketoacidosis and died within two or three weeks.
To see the role of glucagon more carefully, we gave an injection of adenovirus containing the glucagon receptor cDNA, and within two days glucagon receptor peaked at very high levels, but this expression of glucagon receptor was transient; it soon dropped off and disappeared. During the rise in uh, glucagon receptor expression, hypoglycemia appeared for the first time, reaching close to 500 before it too uh, started to disappear, returning completely to normal levels when the glucagon receptor messenger RNA was no longer detectable. These results told us that what caused the syndrome of type 1 diabetes was not the lack of insulin directly, but rather the lack of glucagon action.
To understand more fully the interactions of insulin and glucagon in vivo, in real life, we took a look at the pancreas of humans—normal humans on the right and type 1 diabetic humans on the left. This is the work of Leo Orci at the University of Geneva. As you can see, the human islet consists of a mixture of beta cells, shown in red—insulin-secreting beta cells—and glucagon-secreting alpha cells in green. You can also see that normally every alpha cell is juxtaposed to a beta cell. In the type 1 diabetic, obviously all the beta cells have been destroyed, and the alpha cells uh, sit in the islet virtually by themselves, uh, without any contact with insulin cells. Now, is this mixture and juxtaposition of the two cell types simply an accident of nature, or does it have functional implications? We feel very strongly that it has uh, vital functional implications for glucose homeostasis. This is how we established that—these experiments were done a long time ago, in about 1974. They consisted of isolating normal rat pancreas and perfusing it with a constant level of glucose—a high glucose level of 25 millimolar. The measurement of glucagon coming out of the pancreatic efferent vein uh, indicated uh, some suppression of the glucagon at a level below 1,000 picograms per ml. Then into the perfusate we introduced a potent neutralizing anti-insulin serum, hoping to capture and disarm all insulin coming out of the juxtaposed beta cells. What we found was a 150% increase in glucagon levels that lasted while the uh, antiserum was perfused, but returned quickly to normal or subnormal levels when the antiserum stopped. These results told us that the function of the juxtaposition of beta and alpha cells was to restrain glucagon secretion, and that insulin was in control of glucagon. In more recent studies, it was also found that insulin not only blocks glucagon secretion but blocks its expression in the beta cell, reducing it uh, drastically. These are cultured hamster cells obtained from Jacques Philippe in Geneva, exposed to uh, zero or high levels of insulin.
What does insulin do in terms of glucose metabolism? Certainly we have all been taught that insulin in the peripheral circulation controls glucose tolerance. What I've shown you so far would suggest that the only metabolic actions exerted by insulin are through its control of glucagon, but one exception to this might be its peripheral action on glucose tolerance. To test the role of glucagon in peripheral glucose tolerance, we took the glucagon receptor knockout mice and did an oral glucose tolerance test, and here you see the glucose levels perfectly normal for the glucagon receptor knockout mice. Now, now we destroy all our beta cells and compare the glucose tolerance before and after the destruction of beta cells, and we find very little difference. Is it possible that there was a difference in the uh, sensitivity to the streptozotocin that we used to destroy the beta cells? So we compared the insulin levels before beta cell destruction and after beta cell destruction, and indeed we had eliminated most of the beta cells, and yet the glucose tolerances were virtually the same. This suggests once again that glucagon is determining the shape of the glucose tolerance curve by its action on the liver, and insulin has very little influence—uence on this curve, even though it may be increasing peripheral uptake of glucose.
Well, if what I've told you thus far is true, one would expect to see glucagon elevated in every form of diabetes, and over many years we have in fact looked at every species that we could obtain with every form of diabetes that we could uh, induce—whether chemical, whether autoimmune, whether type 2 diabetes, lipodystrophic diabetes, and even total pancreatectomy. And the fact that totally depancreatized dogs, rats, and humans become diabetic was a a bit of ammunition for the skeptics that found it difficult to accept a glucagon-centric view of diabetes. The dogma that most of us were taught is that total pancreatectomy removes all alpha cells, and the fact that diabetes uh, follows was uh, the most serious obstacle to accepting the role of glucagon in the disease. So in the middle 70s, I spent a year here in the at the University of Geneva in the lab of Leo Orci, trying to solve this issue. We studied totally depancreatized dogs and found that our surgery was complete; their insulin levels were zero, they had hypoglycemia between 300 and 400, and the glucagon levels in their plasma, to our surprise and delight, were very high, even though the pancreas was no longer there. Well, where is this glucagon coming from? To identify the source, we sampled blood from various effluent vessels, and when we measured the glucagon in the gastric vein, we found that it was three or four times higher than in the vena cava, indicating that the glucagon in totally depancreatized dogs originated in the stomach. So Dr. Orci and I undertook a careful search of the stomach to see if we could find any alpha cells, and indeed we did. EM, which gives you a specific marker of the alpha cell based on the secretory granule consisting of an electron-dense but homogeneous core surrounded by an electron-lucent halo, were found not only in the pancreatic alpha cells but also in gastric alpha cells, which, like the pancreas, produced true glucagon in addition to glicentin, which is glucagon with an eight amino acid extension at the C-terminus.
Now another classic dogma in type 1 diabetes is that insulin replacement, as carried out by Banting, Best, and Collip in 1922, is sufficient therapy for type 1 diabetes. Now we've had insulin monotherapy for over 90 years since that first injection, and we know that one can never duplicate the glucose homeostasis of non-diabetics, in which no amount of glucose in the diet or no amount of glucose utilization by exercise or starvation can move the glycemia above or below these narrow limits. Even the best type 1 diabetic controlled on insulin alone shows a pattern quite different from normal; many of the glucose levels are in dangerously low range, um, causing troubling hypoglycemia, while surges occur, particularly at mealtime, reaching very high levels. And this is the pattern of life of a type 1 diabetic, requiring multiple glucose samplings uh, through each day in an endless chase for normal glycemia that non-diabetics take for granted. Why is it that insulin replacement cannot duplicate this uh, type of glucose homeostasis? We think the big difference lies in the anatomy of the insulin that's secreted compared to the insulin that's injected. Secreted insulin begins its journey in the islets of Langerhans, and the juxtaposed uh, insulin cells are in direct contact with alpha cells; their first target, therefore, undiluted insulin reaches the alpha cell and suppresses the glucagon, as we showed a moment ago. Once the insulin leaves the islets of Langerhans and the pancreas, it enters the portal vein and reaches the liver, diluted by 1/14th from the concentration in the islet. Once it passes through the liver and gets in the peripheral circulation, there is another 10-fold dilution. Compare that with the concentrations with injected insulin; at the arbitrary dose I selected, you would get uh, identical concentrations at every one of the target levels. So obviously this cannot duplicate the um, effects of endogenously secreted insulin.
So what is the choice therapeutically, given this fact? Well, if you try to suppress glucagon by giving enough insulin to reach the paracrine levels uh, inside the islet, you would overwhelm the tissues outside the islet and cause a serious hypoglycemia. On the other hand, if you keep the levels physiologic uh, in a range of a non-diabetic, you will allow glucagon levels to rise abnormally and cause serious hyperglycemia. It really is a no-win situation for giving insulin alone. So we suggest another strategy. First, we want to get rid of the hypoglycemic levels, and this you don't have to be a rocket scientist to figure out how to do that, because if you measure the insulin levels when these nadirs of glycemia occur, you find that it's about 20 times the normal level. So we want to just drastically reduce the amount of exogenous insulin provided, and when we do that, a 90% reduction will cause the hypoglycemia to disappear. But when you do that, you further increase glucagon levels, and the hypoglycemic surges uh, increase in magnitude. So if you suppress the glucagon with a suppressor other than insulin, you can then move those hyperglycemic surges down into the normal range and provide the patient with a constant normal glycemia—stable normal glycemia. And this has been done with somatostatin in 1978, and here you see a u, a more recent glucagon suppressor uh, metformin, using a bihormonal approach to type 1 diabetes and obtaining superior stable u, glycemia. You can also achieve the same stable u, glycemia by neutralizing—blocking the glucagon receptor with a glucagon receptor antibody uh, here you see work still in progress uh, designed to see if the long-term effects on uh, glycemia are as good as with glucagon suppressors.
So now I'm going to turn from type 1 diabetes to type 2 diabetes, and there we don't have a dramatic initial event that bonds the concepts that we that we have about type 2 diabetes, and the field is much more confused than the type 1 field because the etiology of the disease—type 2 diabetes—is controversial. Among the causes proposed are genetic uh, defects, insulin resistance, beta-cell exhaustion, glucotoxicity, and a low redox, as well as obesity, and all of these uh, abnormalities do exist in the type 2 diabetic patient. But we recently have suggested a new concept for the etiology of type 2 diabetes, and that is that it is caused by lipotoxicity of the islets and can be reversed by reducing diet-induced hyperinsulinemia and lipogenesis.
Again, let's turn to the islets. The type 1 diabetic islet—you're all familiar with the fact that there are no juxtaposed beta cells to provide insulin. The type 2 diabetic islet is quite different; not only are there juxtaposed beta cells causing hyperinsulinemia, at least in the early years of this disease, but there seems to be more insulin-producing cells than in the non-diabetic. Why then would you have hyperglycemia? Well, we think that the answer is that you have insulin resistance in the alpha cells, just as you do in other targets of the hormone. If you follow mice or rats prone to develop type 2 diabetes and compare them with their wild-type controls—these are lean animals—you'll notice that the glucose levels are perfectly normal throughout the period of observation, and so is the islet content of fat. But when you look at uh, db/db mice that develop type 2 diabetes, you see quite a different pattern. Although in the pre-diabetic phase obviously glucose levels are normal, there is very little accumulation of lipids in the isolated islet until about the eighth week of life. At that point, there is a sharp increase in lipid content, and within a matter of days the blood glucose starts to rise, and you have overt diabetes. If you look at the beta cells before and after this rise in um, islet lipids, you see a striking difference. Before the rise, the mitochondria look perfectly normal, and there are normal insulin granules in the beta cell. After the rise in lipids, the mitochondria are almost unrecognizable; these are severely damaged mitochondria, and the islets are—the beta cells are degranulated.
Now, does this apply to humans? Well, Lydia Shapan, who uses magnetic resonance spectroscopy—MRS—to quantify the lipid content of various organs, has compared the pancreatic lipid content in lean individuals, in obese individuals, and in obese individuals at the time they develop impaired glucose tolerance, and she sees a big increase in the lipid content when the obese non-diabetic develops impaired tolerance, simulating the results in the—in the rodents. Also, there is published evidence that came from our lab many years ago that the alpha cells of normal uh, humans are relatively sensitive to a rise in insulin; the drop—the glucagon levels go down by three picograms for every micro-unit uh, rise in insulin, whereas in type 2 diabetes they're much less sensitive.
Now, to understand how the accumulation of lipids might damage islets, we incubated normal rat islets with the levels of palmitate that circulate in in their plasma, and the blue lines show that palmitate labeled with tritium winds up 4 hours later as tritiated ceramide. These are in normal animals. If you take animals that will develop type 2 diabetes, such as the db/db mice, and do the same experiment, you see a much greater incorporation of palmitate into ceramide, and this is because the enzyme serine palmitoyltransferase that condenses palmitate and serine is expressed at much higher levels in the db/db mouse. This is what the ceramide does to the beta cells that uh, synthesize it; you see the complete loss of chromatin in the nucleus of those beta cells that are making ceramide, and we call this lipoapoptosis, since it's lipid-induced cell death. Now, ceramide interferes with the action of insulin on the alpha cells. Here there is no ceramide and no insulin, and this is the level of glucagon messenger RNA in the normal beta cells. If we add to the medium one unit of insulin, we suppress glucagon expression uh, profoundly, but if there's ceramide present, then one unit of insulin will not suppress glucagon expression. In other words, ceramide is causing insulin resistance of the alpha cell.
Well, now let's take a look at the normal paracrinology of insulin and glucagon in response to glucose. Here we have isolated rat pancreas perfused with 5 millimolar—this is the insulin—this is the glucagon levels uh, at at that—that concentration. When we raise the glucose to 15 millimolar, insulin takes off and forms a dramatic spike, but look at the time length of that spike; it lasts only a minute or two, so that the amount of insulin that uh, was added to the perfusion effluent was very, very small—metabolically insignificant—less than 10% of the total insulin released. While this would have no effect on peripheral metabolism, inside the islet this is very big—a big insulin response—and we think what it does in the islet is suppress the glucagon, because the very first rise in insulin is associated with a profound reduction in glucagon. The insulin keeps on suppressing glucagon, and this insulin-glucagon ratio that results is over seven, which is very high, and this insulin-glucagon ratio tells the liver that it better get busy and suck up all the incoming glucose and store it as glycogen. Now this—this is the normal relationship of insulin and glucagon to one another and to their actions on the liver. Now let's look at type 2 diabetes. You do the same experiment in pancreas isolated from type 2 diabetic animals, and the first thing you note is that the dramatic spike in insulin is gone, but insulin levels are higher than normal because the insulin spike is gone. It's not surprising that there is no suppression of glucagon, and then, despite the high levels of insulin, glucagon levels are also high because they're insulin resistant. The net result is to give an insulin-glucagon ratio of of less than one. What this tells the liver is: don't store incoming glucose, continue to manufacture glucose, keep producing glucose because we're starving, even though a big meal has been ingested. In other words, the wrong information is being transmitted to the liver, and this is why glucose tolerance is so abnormal and hyperglycemia persists; the liver continues to produce uh, glucose even though it doesn't need any glucose.
Well, one of the things we notice when we compare glucagon receptor wild-type mice—+/+—with glucagon receptor knockout mice, and we put them both on the same high-fat diet, is that the glucagon receptor knockout mice do not get obese. Here's why we think they don't get obese: the knockout mice do not get hyperinsulinemic on the diet compared to the wild-type animals that get sustained chronic hyperinsulinemia. Now, insulin is vital for lipogenesis; its role as a lipogenic hormone is underplayed, but we know that without insulin you can't get fat. So if you measure body fat in the glucagon receptor knockout mice, it doesn't increase very much on the high-fat diet compared to the wild type, and if you look at the glucose levels, the wild-type uh, animals become mildly hypoglycemic, but not the glucagon receptor knockout mice. In other words, taken literally, this means that the hyperinsulinemia stimulated by the diet is responsible for the obesity and for the very mild type 2 diabetes that may occur, and that losing glucagon action prevents this from happening. Well, if this is true, and we made the glucagon receptor knockout mice just as hyperinsulinemic as the wild type by giving them exogenous insulin, they should get fat, and that's exactly what happens when we treat these mice with insulin; they get just as fat as the wild-type mice. And here you see that um, at five weeks of age, the four-fold increase in body fat with insulin, and at uh, 10 weeks it's a five-fold increase.
Now what I want to show you is how gluc—the glucagon action is involved in the type 2 diabetic syndrome. Here we're looking at normal mice; nothing wrong with these mice, and we just need to show you the normal glucose level—obviously around 100—and this is the insulin level that coexists with this normal glucose level in a normal uh, rodent. Now let's look at the db/db mouse. The db/db mouse at about 8 to 12 weeks becomes severely hyperglycemic—between 400 and 500—despite levels of insulin that are um, many times higher than in the wild type. What would happen if they didn't have the a glucagon receptor? Well, to answer that question, we crossed the db/db mice with glucagon receptor knockout mice, so we have a uh, animals that—db/db animals, but they lack the glucagon receptor, and they go through life with a normal glucose level; they never get the diabetes that their um, wild-type littermates experience, and and insulin levels remain at a low value. Now the final experiment in this series is to replace in these animals lacking the glucagon receptor—to replace the glucagon receptor by giving them adenovirus containing the cDNA of the receptor, and when we do that we find within uh, two days of of of the injection, glucose levels are approaching 600 and insulin levels are sky-high.
So in summary, this tells us that glucagon action is required for type 2 diabetes. It also tells us that treating type 2 diabetes—insulin-resistant type 2 diabetes—by giving more insulin would be expected not to be beneficial to the patient, but harmful, and indeed in the ACCORD study there were 50 excess deaths when patients receiving intensive insulin treatment with u500 insulin. The way to treat this disease is to block glucagon action. This can be done with glucagon suppressors or, as shown in this slide, with an antibody to the glucagon receptor, and you get normal glycemia for about 10 days without uh, any other treatment.
So in summary, this is the pathway to lipotoxic type 2 diabetes that we uh, believe is consistent with the data that I've shown you: you have to overeat; caloric intake must be increased. In a normal person, that will cause hyperinsulinemia, provided glucagon action is available to enable this to occur; in its absence, this does not occur. The hyperinsulinemia upregulates the transcription factor in the liver necessary for lipogenesis, and you get increased synthesis of palmitoyl CoA. This enzyme is present to condense the palmitoyl CoA with serine and form ceramide. Ceramide accumulates if there's excess lipid in a tissue, and it does two things in the islets: one, it gets into the beta cells and causes the lipoapoptosis that I showed you a few moments ago; it knocks out that spike of insulin, but the other thing that it does is it causes insulin resistance in the alpha cell, just as it does in other targets of insulin, and this combination leads to the syndrome of type 2 diabetes.
Now I'd like to show you the diabetes dogma disprovers. Young Lee did the work in the type 2 diabetes; May Wong in the type 1, with the able assistance of Cindy U. Kim Cor has been in my lab since 1971, and most of that time as lab—as lab manager. She's not only involved in all of the uh, technical work, but she made these slides for me, and she keeps our lab going. Matthew Evans is in Michael Ross's lab, and he collaborated on these studies.
Now I'd like to return to my relationship with Ralph LT. In the early 1960s, he and I saw eye to eye on type 1 diabetes because we both believed in the insulinocentric concept of the disease, in which insulin lack was responsible for every manifestation. But if—by 1975 when I saw him again—I had given the Banting lecture in which I introduced the bihormonal concept, in which insulin lack was responsible for the manifestations uh, on the left side of the slide, while glucagon excess uh, created the abnormalities on the right. I wonder then how Ralph would have reacted had we been able to meet in 2014 after giving this lecture in which we're suggesting that insulin lack is responsible for glucagon excess, and all of the manifestations of type 1 diabetes result from glucagon in the absence of insulin.
In addition to uh, collaborators listed here, I want to acknowledge the contribution of my wife, Mariss, uh, who played an enormous role in making all of this possible. She hasn't been to Sweden since 1988 when she completed the Vätternrundan in the fastest time ever for a Texas woman, and that speed record stands even today. With that, I will close and acknowledge the wonderful hospitality that I've received in Stockholm, and I wish all of you well, and thank you for everything. Bye-bye.