The unexpected pathway to the creation of the HbA1c test and the discovery of AGE’s

In a previous study for this journal I discussed how failure in one project might give insight into another project, resulting in remarkable discoveries with significant impact, [1]. In this study, I will discuss the unpredictable road that led to the creation of the diagnostic test, HbA1c. I have found in my lifetime of scientific endeavours, that trusting the truth of clues arising from experiments, can lead to unexplored, unexpected and exciting new understanding of disease processes and treatments. My deep motivation has always been to understand the causes of diseases and how to treat them; yet the persistent pursuit of this has led me across many different scientific disciplines. The flexibility and willingness to enter areas that are not areas of my expertise, aligned with my dogged persistence, have been and are presently crucial. In this study, the story of the hope for defining a new treatment for patients with sickle cell disease, let to the discovery of a new diagnostic test, HbA1c, that has become the worldwide mainstay for the treatment and diagnosis of diabetes. It is especially comforting to know that using this test, many women with type 1 diabetes have delivered healthy babies. This story begins at a social affair in New York for a group of haematologists. At that time I was an Assistant Professor at the Rockefeller University trying without success to isolate erythropoietin from the urine of patients with aplastic anaemia. One of the participants, Alan Schecter, mentioned that there had been a recent report of the successful treatment for sickle cell anaemia crisis in patients with large amounts of urea [2]. The rationale that was given by the investigators was that urea would disrupt the polymerized haemoglobin S in the red cells that cause entrapment in capillaries and in turn cause severe pain and tissue damage. I responded that I could not question the validity of the clinical trials, but I could not believe that one could ever give enough urea to achieve this because the amounts of urea needed to break hydrogen bonds are in the molar range. However, the possibility existed that the clinical effects could be the result of the carbamylation of HbS with cyanate because urea in solution is in equilibrium with ammonium and cyanate ions. The finding that urea solutions were in equilibrium with cyanate that could react with proteins had been first observed with the enzyme ribonuclease [3]. Working with James Manning and Martha Fedorko, we were able to demonstrate that cyanate could react with the amino terminal valines of haemoglobin S and prevent the sickling of cells following deoxygenation in vitro [4] and in vivo [5]. The question was whether this observation could be translated into a therapeutic modality for these patients. We were encouraged to go forward because urea is present in vivo and small amounts of carbamylated haemoglobin could be measured in normal people. At first, the clinical results were very encouraging [6]. The cyanate-treated patients showed less red blood cell destruction, higher haemoglobin levels and reduced incidence of crisis. However, after longer periods of time (several years) on the drug, several patients developed peripheral neuropathy and cataracts similar to that seen in diabetics. We immediately stopped administering the drug to all the patients and continued to follow the patients. Fortunately, the neuropathy and the cataracts reversed with time. It was obvious that the oral administration approach to treat the disease was not going to work. Subsequently, Deiderich et al. [7] reported that extracorporeal red cell carbamylation of patients with sickle cell disease had positive haematological and clinical responses, but the commercial development of the apparatus did not take place. Needless to say the failure of the project was quite devastating, but gave me a very good scientific clue. As time went on I kept thinking about the fact that the neuropathologist and the ophthalmologist both noted that the cyanate-induced damages were similar to that seen in diabetes. I also remembered that there were reports [8, 9] of patients with diabetes having an increased amount of the minor haemoglobin called HbA1c. The structure was unknown, but appeared to be a sugar attached to the amino terminus of the beta chain of haemoglobin – the same place that cyanate reacts. This led me to consider that, perhaps, there was a chemically reactive compound in the blood of diabetics that could react with haemoglobin to form haemoglobin A1c and with other body proteins to explain the pathogenesis of diabetic complications. As I pondered over this possibility, I happened to visit with a friend, Scott Grundy, and told him my thoughts. During dinner, we talked about whether this could be true and how to prove it. He thought it was an interesting idea and encouraged me to follow up on this hypothesis. On returning to New York, I suggested to Ron Koenig, who was in the first class of students in the Rockefeller/Cornell MD/PhD programme, that we do a few simple experiments to see whether this idea was correct. The first experiment was to see whether diabetic mice also had the mouse equivalent to human HbA1c. Using the same method used for human blood, he was able to show that diabetic mice had a minor haemoglobin fraction that corresponded to that seen with haemoglobin A1c in human red cells and that diabetic animals had 2.8 times more than nondiabetic mice. The red cell is unique in that all of the haemoglobin and other proteins in the cell are produced primarily when the cell resides in the bone marrow. The question was when was HbA1c made? To address this question, radioactive iron was injected into normal mice, and a few days, later the radioactive blood cells were obtained. Aliquots of this blood were then injected into normal and diabetic mice. At intervals, small amounts of blood were removed and the amount of radioactivity in the haemoglobin A and haemoglobin A1c fractions was determined. When the ratio of the amount of radioactivity of HbA1c over that in HbA was plotted over time, it was clear to see that HbA1c increased linearly over the life of the red cell when the cells were placed in normal animals. In contrast, although the rate of HbA1c production was also linear when the cells were put into diabetic animals, the rate was 2.8 times faster (Fig. 1). In this first article, we suggested that measurement of haemoglobin A1c would be related to metabolic control, and the reactive substance that reacted with haemoglobin could be responsible for the complications of diabetes [10]. Reticulocytes radio-labelled with Fe59 were injected into normal and diabetic animals. At the indicated times, aliquots of blood were removed and the ratio of radio-labelled HbA1c to HbA was determined. The diabetic animals had a linear increase in HbAc over time that was 2.8 times faster than the normal animals [10]. At the time this work was performed, there was an intense debate about whether the complications of diabetes were exacerbated by poor metabolic control. There were two camps of thought – the people who believed in ‘tight control’ and the other who did not believe metabolic control was important and thus promoted ‘loose control.’ We thought by measuring HbA1c in patients from their clinics that we could validate the idea that ‘tight control’ was important. We arranged to obtain blood samples from two institutions, which will go unnamed, representing each point of view. To our great surprise, the HbA1c values were between 11% and 13% from both institutions compared with values of 4–6% in nondiabetics. This implied that care of patients with diabetes was far from optimum in both places in spite of what the doctors believed. In collaboration with Joe Williamson, we measured HbA1c in a series of patients with diabetes and found that HbA1c correlated very well with fasting blood sugar and an oral glucose tolerance test [11]. At the time of this work in 1976, we thought that HbA1c would quickly replace the glucose tolerance test in clinical practice as a way to diagnose diabetes. This was not to be the case. It has taken over 30 years for the American Diabetes Association to embrace this test for diagnosis. The Rockefeller University Hospital, where this work was carried out, was an ideal place for us to admit patients and learn how best to manage diabetes in a controlled environment. Ron Koenig [12] under the clinical direction of Charles Peterson was able to identify a number of patients who agreed to stay on the metabolic ward for extended periods of time so that we could determine how best to control blood glucose and determine its effects on HbA1c and other metabolic and clinical parameters. One of these patients was my mother who had Type 2 diabetes. She stayed in the hospital for 6 months whilst we learned how best to manage diabetes. As I write this I realize that this would probably not be allowed in most institutions today, but she was proud to participate in these studies knowing that the knowledge we gained would help others avoid the terrible complications of diabetes. After 3 weeks of the initiation of better diabetic control, there was a fall in fasting blood sugar, urinary sugar and HbA1c. Figure 2 displays the fall in HbA1c over time in a patient with Type 1 diabetes following initiation of glucose control. The amount of glucose in the urine and the percentage of HbA1c were followed over time in a patient with diabetes. The initiation of better glucose control as evidenced by a decrease in urinary glucose was followed 3 weeks later by a decrease in HbA1c. The introduction of the HbA1c test to the medical community was met with considerable scepticism when it was first introduced. An example of this occurred when I suggested to the administrators at the Rockefeller that we should patent the measurement of HbA1c as a method to measure metabolic control in diabetics, I was told that this would be a waste of money because they were assured by several key diabetologists that the test would not be used because patients really did not want to know. I argued unsuccessfully that this was not the case and that a patent would promote an orderly development of new more rapid procedures, and standards so that comparison of data from different hospitals would be possible. As a result the fragmentation of the market into many providers has resulted in the absence of a primary standard that is used through out the world. Reviewing the arguments brought up by nonbelievers of the concept of measuring control today is not very informative other than to say that it was very frustrating. I have come to realize with age that this is a step that is necessary before adoption of new ideas. One of the first groups of clinicians to embrace the concept of monitoring HbA1c was obstetricians who cared for pregnant Type 1 diabetic patients. At that time, the incidence of birth defects in children born to these mothers was >20%. By promoting tight control of blood glucose through use of HbA1c measurements, it was possible to reduce the incidence of birth defects to that found in nondiabetics [13, 14]. Another group that very quickly began to measure Haemoglobin A1c were insurance companies. One day soon after the publication of the New England Journal article referred to above, I received a call from a commercial laboratory asking for technical help on measuring HbA1c. Out of curiosity I asked who would be using so many measurements. He then replied that insurance companies were routinely measuring HbA1c on the blood samples that they obtained when people were applying for insurance to detect patients with diabetes. I was horrified that a test developed to help diabetics could potentially be used against them. There is no question that the DCCT in type 1 diabetics and the UKPD trial in type 2 diabetics convincingly showed that tight control as assessed by HbA1c could prevent the complications of diabetes. The DCCT, which was carried out by the NIH from 1983 to 1993, showed that the onset and progression of diabetic complications of the eye, kidney and nerve damage were significantly reduced in the patients with Type 1 who had stricter glycaemic control as evidenced by HbA1c measurements [15]. A follow on study to the DCCT, called the EDIC, continued to follow these patients and showed that the risk of having a heart attack, stroke or death from cardiovascular disease was less in the patient group with lower HbA1c [16]. Similar results were observed in the UKPD with Type 2 diabetic patients. This trial was initiated by Bob Turner [17] in 1977 and ran until 1997. At the 10 year mark, there was a significant reduction in death, and diabetic eye and kidney disease. These studies are milestones in the field of diabetes. The evidence for the importance of tight control is overwhelming. As a reflective note I refer you to Fig. 3, which is a composite of these two studies in terms of the HbA1c values observed between the different degrees of control and the type of diabetes, as well as the range seen with nondiabetic people. As I mentioned earlier when we first studied HbA1c, the values we observed were often in the range of 10–18%. I presume that this difference reflects the increased knowledge of the importance of glucose control to patients or at least those that volunteered for clinical studies. The second observation is that relatively small differences in HbA1c between the groups over a period of 10 years results in significant differences in the development of complications. These two studies were critical in pointing to the importance of glucose control in patients with diabetes and resolving once and for all that tight control was important in the development of the complications of diabetes. The HbA1c values of the DCCT (Type 1 diabetics) and the UKPDS (Type 2 diabetics) for the intensively treated patients compared with a comparisonal group of patients are shown over a 9-year period of time. In addition to HbA1c being a biological marker, we believed that the chemistry of the nonenzymatic reaction of glucose with other body proteins was responsible for the complications of diabetes and possibly some aspects of ageing in general. But first let me review the chemistry of the reaction of glucose with the amino terminal valine of haemoglobin and the epsilon amino groups of lysines in proteins (Fig. 4). The reaction of the amino terminal valine of haemoglobin with glucose occurs so because this amino group has a lower At of these amino groups have no This with about of the epsilon amino groups of amino groups can react with the of glucose to form a the amino terminus has a higher of a This can a to form what is called an by The structure of HbA1c was found to be Haemoglobin A with an attached to the amino terminus of the beta chain This relatively can to glucose and This reaction for the of HbA1c to an 1 The of the of the The of glucose can form a with the amino groups of This can a to form the – a relatively The can to the or to form a series of that can that a similar reaction of glucose with other body proteins might be we began by the The the is not on for glucose and cataracts in and with diabetes. The of proteins with glucose in vitro led to and an increased to form that were similar to that seen in the cataracts of diabetic These results suggested that cataracts in patients with diabetes could be the result of glucose of As we studied the chemistry of the reaction of glucose with we were to the chemistry In the of the the noted that glucose could react with proteins and when to form a of called for the of the studied this reaction because of its importance in and of The can many to form an series of more than have been after the of of us have carried out the reaction by sugar on the of a or before and the In addition to the can form in the deep in the will become as a result of they are 1 in suggested that this chemistry would be important in and we could not that it did in and I to for in biological One of the of a small of is that they have a with a and When we human cataracts for this we observed that were the same as that had been with glucose we called this of glucose as as the had This to be because many of the studied by were under and To those under of we the A study of from people showed a linear increase in the amount of with age and that diabetic had an amount of that was to their age for Fig. the of as a of age and diabetes in the of the that over time in have a The amount of in the linearly over time in nondiabetics. have increased amounts of these compared with age nondiabetics. these there have been over in that have in as a of diabetes or Figure 6 is a of the human body where have been suggested to be in the development of the complications of diabetes or have been in the pathogenesis of and complications of diabetes and Another important that can react with is We were to this possibility because we that blood glucose could cause damage to the the first of A number of that with are to be The addition of to solutions led to the of that had the same noted for proteins and of the Subsequently, and showed that there was a increase in the rate of mouse which had the in a diabetic mother compared with a nondiabetic animals also over time. A linear increase in the number of in the was observed from birth to months that we was because of the reaction of with The of damage from is in of work. unexpected results occurred the years that we on the chemistry of the The first occurred in our work when we to an for HbA1c. this we observed that the data the data obtained with the values obtained by However, there was a of people who had lower values with the compared with the of the of these patients that this of people had a of We that in vivo by the of could be with the of HbA1c in a way that the could no longer we were able to show was that the of with red cells led to in a studies showed that the measurement of this could for a previous we thought it would be a several to the money to this were met with scepticism and not We work for At a laboratory we discussed in the possibility that could form an between the group of the and the amino group of valine of The importance of this structure was that it could explain our previous it could explain how this could and the from to form in vitro and in vivo (Fig. 4). In the of two the development of as and had In we were able to prove that the structure that we had was correct. We in this article that the for was the result of Another result occurred when we were in the of diabetic a that we could of in of patients with diabetes that had because of critical The finding was that a number of nondiabetic patients with also had amounts of in their In how this could out that the nondiabetic had been or were A possible was that reactive are present in that were being and through the and the blood the The of The of the which are with are in to form A simple experiment that through a solution led to the of on the to had their turn but also on and body As many of the of are a form of we put the concept that the ageing of could be the result of the of in the We to that would most of these reactive we thought that the would embrace an that did not turn out to be the and many good the time was not The on the of As in the to be of the in spite of to reduce I believe that the scientific and medical community should to decrease the damage by a that is the we to with or The first had an amino group attached to which is always at This the amino group from and thus it to be able to react with studies of with diabetic animals showed that we could and in the kidney A clinical trial in diabetic patients with was over two to there were positive clinical effects observed in the patients the drug, the drug treatment groups did not over the group for the primary of of Needless to say this was a The second was a compound that could break in vitro and in vivo and and showed a significant decrease in of the heart and blood However, a clinical study in patients with heart failure to show an in of both has been are the to be learned from these clinical The first obvious is that clinical are and to out, especially the have to be carried out for periods of time. The second is that we better to of diabetic complications with to the disease so that the amount of drug and the can be in 2 studies. A good example of the type of that I to is HbA1c. the HbA1c has become the standard for the and the for new to be to treat we have come a way in being able to treat diabetes, these patients go on to neuropathy and cardiovascular disease. New are needed more than ever to treat the complications of diabetes because of the of Type 2 diabetes in the A suggested that the reaction of glucose with proteins in the body could cause years from we will know how to decrease the damage by In it is important to that which is for is also the of ageing and of was I to and for their help in the of this