Measuring Insulin Levels, Hyperinsulinemia and Insulin Resistance

A slight detour in my current endeavors to put forth some information regarding the etiology of diabetes, insulin resistance and β-cell function, before I get to the two major reviews that prompted this.

To review, in my last post, discussing this paper, a seminal observation/conclusion can be summed up as follows:   Insulin is formed in stepwise fashion from a larger protein (preproinsulin) that has a terminal signaling chain cleaved forming proinsulin that is then folded, cross-linked and has one of the three main chains cleaved in the last step(s) resulting in a protein with two parallel chains, insulin.  I like this newer representation I found below because it provides the numbers in the amino acid sequence at which the cuts are made.  This is important to better understanding the papers I'm going to discuss here.

SOURCE

Proinsulin is the entire B-C-A chain.  I suppose since the C-chain is "disposed of" (this is c-peptide), they designate the chains of insulin A and B by length, but as you can see, the numbering starts at (1) on the B-chain, continues at (31) to the C-chain, and on to the A-Chain at amino acid (66).  Thus the C-chain or c-peptide is amino acids 31-65.   Properly formed insulin contains just the A&B chains, proinsulin contains the intact C-chain, and we can conceivably have proinsulin-like molecules that are incompletely split -- with the C-chain "hanging on" at the 31 end or the 65 end.  The latter are often referred to as split-proinsulin molecules.

Of key importance is that only fully formed insulin is biologically active.  The various precursors have little if any insulin function.

This makes the basic finding (and summary of the findings of others) of Tirgoviste all the more compelling.  This is that based on 170,000+ cases of diabetes, of both T1 & T2, the major finding was that proinsulin levels are elevated compared with non-diabetics.  It is proposed that the defect is in the assembly process of the insulin resulting in formation of "immature" secretory vesicles containing large amounts of proinsulin.  From that paper, it is learned that normally the proinsulin:insulin (P:I) ratio is less than 0.1 or a P:I of at most 1:10.  In diabetics the P:I ratio can exceed 1.  This would mean more proinsulin than insulin requiring proinsulin to be greater than 50% of the molecules.

This would likely result in "hypersecretion" or an overactive pancreas to compensate.  I've constructed a small table of some possible scenarios.  Forget units and let's use a stimulus of 100 that elicits insulin secretion.  Normally 100 insulin "units" are secreted which is what is required to "clear" the full 100 units of stimulus.  Along with this let's use the upper limit of 10 "units" of proinsulin for a total secretory response of 110 "units".  This is the top row.  In the second row, let's say β-cell function is slightly impaired so that the P:I ratio is 1:5.  For a 110 "unit" secretory response, only 92 "units" of insulin is secreted (along with 18 of proinsulin).  This insulin response clears most, but not all of the stimulant leaving 8 "stimulant units" in circulation.  The 8 "stimulant units" require 8 insulin "units" but at the 1:5 P:I, this requires 9 total secretory "units".  Thus the total secretory response is 119 "units" which is 8% higher than normal.  It gets worse, however for the more compromised β-cell.   Following the same math, when function declines to a 1:1 ratio, the total secretory response is increased by 75%, and at 3:2 (1.5X as much proinsulin as insulin, 60% proinsulin/40% insulin), the secretory response is more than doubled, increased by 116%.

So, as I have intimated in recent comments, the role of hyperinsulinemia in obesity and diabetes may have been complicated -- perhaps horribly so -- by the methods used to determine insulin levels.  Especially in older studies, though it is unclear if this has consistently been addressed and corrected.  Still, much of the current understanding of the etiology of Type 2 diabetes relies on  studies from the 80's that resulted in the concept of insulin resistance leading to hyperinsulinemia.

So I recently got my hands on the full text of this study:  RADIOIMMUNOASSAY MAY OVERESTIMATE INSULIN IN NON-INSULIN-DEPENDENT DIABETICS  (note, this was published in 1989, but bears a 2008 date in some searches because that is the online publish date).   The summary was enough to peak my interests:

We have compared insulin concentrations measured by radioimmunoassay (RIA) in plasma from 50 fasting non-insulin-dependent diabetics (NIDDM) with those measured by a new monoclonal antibody-based two-site immunoradiometric assay (IRMA) of insulin (which has no significant cross-reaction with proinsulin-like molecules. We find that the RIA measures the sum of the insulin and proinsulin like molecules and that the IRMA insulin concentrations are 38% of those measured by the RIA in those diabetic subjects. We conclude that the importance of insulin deficiency in NIDDM may have been obscured by this error.

Before moving forward, a quick summary of what an RIA is. This image on the right seems a fairly accurate general representation. The "Y" is the antibody and Ag stands for antigen (in this case insulin).  The antibodies are incubated with radioactive-labeled antigen and these bind -- a calibration curve can be constructed for various known concentrations of insulin and measuring bound and unbound radioactivity.  To determine the insulin concentration of an unknown sample (patient's in this graphic), these are incubated together and the unlabeled antigen displaces some of the radioactive antigen.  This reduces the amount of bound radioactivity and increases the free radioactivity.  The more insulin in the sample, the more bound radioactivity is displaced as the unlabeled antigen binds to the antibody in its place.  Cross-reactivity is the term for when an antigen other than the one being measured can also bind to the antibody.  Thus in this study, they demonstrated that proinsulin-like molecules bind to the antibody -- "cross react" -- resulting in inflated levels of insulin in less sensitive assays.

So, again using some hypothetical round numbers, if a stimulus elicited a 100 unit secretory response, RIA would have registered an insulin level of 100 while the more specific IRMA would have measured a level around 40 for diabetics as in this study.  This means that 60 units are proinsulin or proinsulin-like molecules that are "cross-reactive" in the RIA and thus "seen" as insulin by that analytical method.  This is not small potatoes we're talking about here!

It gets a bit more confusing because we cannot assume 100% cross-reactivity, nor can we assume equivalent cross-reactivities for proinsulin and the split proinsulins.   The term binding affinity is one that refers the the "efficiency" with which two molecules bind.  It is conceivable that not only does proinsulin bind to the antibody, but it might even bind preferentially.  This would displace even more radioactivity per molecule yielding a deceptively higher measured concentration.

OK, enough eye-glazing math.  I'll repeat the two critical passages from the summary:

• the RIA measures the sum of the insulin and proinsulin like molecules
• insulin concentrations are 38% of those measured by the RIA 

Some other excerpts:

The problem of specificity is unlikely to be serious in normal subjects, where it is thought that proinsulin-like molecules account for only 10-20% of the immunological insulin-like molecules in plasma.   However, in NIDDM, circulating proinsulin is known to be disproportionately elevated compared to insulin, both fasting and following an oral glucose load (Yoshioka et al., 1988). The insulin status of NIDDM therefore remains uncertain (de Fronzo et al., 1983; Nesher et al., 1987).

So here's some study info and results:

• n=50 Type 2 diabetics,  26 Caucasian and 24 Asian 
• Age range: 35 to 70 
• Body mass index (BMI) range: 18.6 to 40.1 kg/m2,  27 BMI under 26 ,  23 BMI over 26.5 (obese). 
• HbA1c range:  6.6 to 13.4% 
• Duration of diabetes: 6 months to 16 years. 
• Treatment:  19 diet alone, 19 diet & sulphonylurea, 5 diet & metformin, 7 diet, sulphonylurea & metformin. 

"We have shown that concentrations of 65-66 split proinsulin are very low in normal subjects and NIDDM.   In this study, we have measured concentrations of IRMA insulin, intact proinsulin, and 32-33 split proinsulin, in NIDDM and compared the results with insulin concentrations measured by RIA. "  All levels were measured from venous blood drawn in the fasted (12 hours overnight) state.  The results shown at right.  Interestingly, "sulphonylureas, obesity and poor glycaemic control made no significant difference to the percentage of proinsulin molecules in the plasma."  

Some Comments:

• The first thing I notice is the wide ranges of fasting insulin -- both absolute values and as a percentage of total secretion. 
• By RIA, all subjects are demonstrably hyperinsulinemic (fasting insulin over 20 picomolar), but by the more specific assay, the IRMA, the lower ranges of both obese and non-obese insulin levels fall within  the range of normal levels.
• We don't know which subjects the range extremes refer to,  e.g. the lowest absolute insulin level may or may belong to the same subject as the lowest percent insulin secreted.  
• Although none would be considered "insulin deficient" by fasting insulin levels (even accurately measured as in the IRMA) there are some subjects with a total-PI:I ratio approaching  9:1!!  
• The subject producing 10% insulin would have absolute insulin levels measured almost 10X of normal by RIA. This seems to be the same subject with the max proinsulin and split proinsulin percents of 21 and 70% respectively.  This is some pretty serious β-cell dysfunction in the assembly department!
• The upper range of insulin production by percent -- 62 (obese) and 70 (non-obese) -- is already indicative of β-cell dysfunction with between 30 and almost 40% of secreted "insulin" being proinsulin-like molecules.  So even the "best performing" β-cells of the bunch were secreting from 2-3X normal amounts of proinsulin-like molecules.  

From the Discussion:

Cross-reactivity studies of the RIA method used in this study have shown that intact proinsulin and 32-33 split proinsulin cross-react with a potency very similar to insulin. Our studies of two other RIAs have shown similar high cross-reactivities with intact proinsulin and 32-33 split proinsulin.

Clearly these results do not entitle us to conclude that all RIA insulin measurements previously carried out in plasma taken from individuals with NIDDM are in error.  However, it may well be that the considerable confusion in the literature (de Fronzo et.al., 1983) concerning plasma insulin concentrations in this condition is due to the variable cross-reactivity of insulin RIAs with the proinsulin-related molecules. Proinsulin-like molecules have very low insulin-like biological activity (Peavy et al., 1985) and their confusion with insulin would lead to erroneous conclusions concerning the insulin status of NIDDM.

Now, I'm sure many of you reading this must be focusing on the dates here.  This paper was published in 1989 and references studies throughout the 80's.  Surely this has been addressed in the intervening quarter century?  Well, this is what I find a bit perplexing.  I'm not sure if perhaps I'm just not using the right search phrases, but while I find quite a bit on proinsulin and such, it is all from this time period.  And among the authors of the paper I've been discussing is none other than John S. Yudkin, clearly not some obscure research group whose findings might slip through the cracks!  I did look briefly at currently available RIA's for insulin, and one mentioned a 0.3% crossreactivity for synthetic proinsulin (good), but no mention of the more potentially complicating split proinsulin.  I'm also learning quite a bit more about the measurement error in these assays -- something that is not often mentioned in the studies, at least to acknowledge potential impact on the results.

So, some concluding thoughts for now.  In those with β-cell dysfunction, the incomplete maturation of proinsulin to insulin may be the cause of hyperinsulinemia in a few interrelated ways:

• Underproduction of active insulin in response to stimulus would lead to stimulus levels remaining elevated, thus stimulating further insulin production/secretion.   This results in a larger secretory response to produce the same amount of insulin. 
• Due to the phasic nature of insulin secretion, in playing "catch up" it is possible that secretion of insulin may exceed what is normally required (e.g. as calculated in my table of hypothetical scenarios above) to return stimulus levels to normal.  
• Whether the actual active insulin secretion is increased or not, the total insulin and proinsulin-like molecules would be greater, often significantly greater -- it would make sense that this increased production demand would involve initial increases in β-cell mass to meet demand.  But ...
• Insulin levels measured by less specific RIAs that see proinsulin or split proinsulin as insulin could result in dramatically inflated measured insulin levels (these were 3X+ at the high end of the range in this study!)

Thus hyperinsulinemia can be explained at the production level, both mechanistically and/or by measurement inaccuracies, and may not be completely compensatory for insulin resistance (this is not my thinking, this is what is presented in the upcoming reviews).  The elevated insulin levels may not be insulin at all and thus impaired glucose clearance may be insulin deficiency after all and not impaired clearance at the target cellular level (e.g. insulin receptors & GLUTs).  

This post is getting long so I will discuss at least one more related study in another post before moving on to the promised reviews.  The next study is this one:  Serum proinsulin levels at fasting and after oral glucose load in patients with Type 2 (non-insulin-dependent) diabetes mellitus.  In addition to looking at diabetics, they also compared normal and IGT groups.