Physiology of insulin

This chapter is related to Section U1(ii) from the 2017 CICM Primary Syllabus, which asks the exam candidate to  "describe the physiology of insulin, glucagon and somatostatin". It was initially rather tempting to conflate this section with U2(ii), "understand the pharmacology of insulin preparations",  but that seems to be blueprinted as a completely separate syllabus item (i.e. there are specific exam questions that ask about the pharmacology of insulin but not its physiological effects). Similarly, it was challenging to dissect a boundary between this section and U1(iii), "explain the control of blood glucose", but the large number of highly specific SAQs that ask to explain the control of blood glucose also suggested that it should be treated separately.

What, then, would be left in this chapter on insulin physiology, if all discussion of insulin preparations and glycaemic control were carefully excised? Surely, just a quick summary, a light jaunt over the important points?

Ahem.

The design of this structure and content was made more difficult by the vagueness of the syllabus item wording, and by the fact that there do not seem to be any specific SAQs about it, which makes it difficult to guess what CICM consider to be relevant here. It is possible that Question 8 from the second paper of 2010 is somehow meaningful, because it asked what might happen if you did not have insulin. From this, it was possible to characterise some rough borders for the discussion that follows. Additionally it was helpful to review the construction of famous works by established authors which deal with insulin physiology, of which the most celebrated would have to be George Cahill's 1971 lecture, Physiology of Insulin In Man. The author hopes some credibility was accidentally transferred along with Cahill's borrowed structure in this page.

Production and storage of insulin by β-cells

Insulin is produced and stored in the pancreatic islet β-cells, which are unique to the islets of Langerhans - endocrine micro-organs which lay scattered along large vessels within the parenchyma of the pancreas. As if it were somehow relevant to the CICM trainee, it still for some reasons feels important to describe them:  these cells are polygonal, with an average diameter of 13–18 μm, and full of about 10,000 insulin-containing granules. That number might seem like one of those "two tennis court" sensationalisations of physiological data, but it is probably factual, and seems to come from the early electron microscope work by P.M. Dean (1973) who performed a series of morphometric studies on mouse β-cells. From these, we know that 20% of the interior of an average β-cell is occupied with endoplasmic reticulum, 4% with mitochondria, 10% with nucleus material and 11% with insulin-containing granules, which in mice take on a characteristic "fried egg" appearance when observed with transmission electron microscopy:

These granules fill with insulin from the endoplasmic reticulum of the β-cell, where it is synthesised from proinsulin and preproinsulin. Weiss et al (2015) have a highly regarded Endotext article on how it is manufactured, and to distract from their excellent work by offering an alternative explanation would be a disservice to the reader.  Still, that's exactly what we do here at Φ, so:

• The 1500-base insulin gene is transcribed into MRNA
• MRNA is transcribed to form preproinsulin, which is a 12,000 Da protein
• Preproinsulin comes tagged with a 24-residue signal peptide of hydrophobic residues, which is a classical peptide signal common to protein molecules destined for exocytosis. This residue is removed in the endoplasmic reticulum, forming proinsulin (an 86-amino-acid protein). The residue itself is then degraded and lost, meaning we never again see this 24-amino-acid fragment. Apparently, this whole process takes no more than 2-4 minutes
• Proinsulin is then folded into a characteristic tertiary structure in the rough endoplasmic reticulum, which requires the formation of disulfide bonds 
• The folded proinsulin is then packaged into Golgi apparatus vesicles where it is finally converted into ready insulin by endoproteases PC1 and PC3. This proteolysis yields the final form of insulin (a 51-amino-acid protein with a molecular weight of 5808 Da) and C-peptide, a 31-amino-acid polypeptide fragment that ends up packaged along with mature insulin in the β-cell granules. The main reason for mentioning this is that C-peptide is then released into the circulation along with insulin, acting as detectable biochemical evidence of normal insulin synthesis. Unlike insulin itself, C-peptide emerges intact from the molecular woodchipper that is the liver, and C-peptide levels can be measured in the systemic circulation, such that a very low level probably indicates Type 1 diabetes.

Or, as if there was a shortage of high-quality illustrations for the process of insulin synthesis, here's a sausage-themed graphical interpretation of the process (to complement the fried egg allegory of the insulin granules):

The vesicles which end up with all this insulin are then stored inside the β-cell, ready to fuse with the exterior membrane and release their contents into the extracellular fluid of the islet (which is basically continuous with the portal venous plasma, as the islet cell capillaries are famously porous). Obviously insulin is not secreted immediately as it is produced, and various reputable sources tend to repeat the same factoid, that the pancreas of a normal adult contains something like 200 units of insulin stored in β-cell granules. As far as a short Google search could reveal, this value seems to be a scaled extrapolation from a 1931 animal study by Scott & Fisher, who arrived at a figure of about 1.7-1.8 units per gram of pancreatic tissue.

All this intragranular insulin is not stored as a lake of sloshing hormone solution, but as a condensed crystalline form, bound into insoluble hexamer aggregates with the aid of zink and calcium ions (of which there is an abundance, in the order of 20-30 mmol/L), as well as a relatively low pH (~6.0). These condensed insulin crystals are the "yolks" of the fried egg granules. The finer detail of how and why this happens is probably a topic far beyond the needs of the CICM exam candidate, but the curious reader can avail themselves of Germanos et al (2021), whose excellent paper is enticingly titled "Inside the Insulin Secretory Granule". It will suffice to say that the main reasons for storing insulin in this form are related to the improvement in density: the condensation allows about five times more molecules to be stored in each granule, achieving a final intragranular insulin concentration of 40 mmol/L, which translates into 17.3g/L (or something like 500,000 IU/L of insulin).

Also, it appears the zinc and calcium seem to have various paracrine signalling properties when they are released along with the insulin. When the granules fuse with the external cell membrane, the pH changes and the metal ion concentration is diluted, causing the hexamers to spring open, releasing six soluble insulin monomers into the circulation.

Which brings us to:

Mechanism of insulin secretion by β-cells

The normal reaction of β-cells is to release their granule content into the bloodstream in response to an increase in the concentration of various nutrients, mainly blood glucose but also other soluble macronutrients. This is obviously a response that should be graded, proportional, and quickly responsive to changing conditions. Being wrapped in highly porous well-perfused vessels helps islet cells keep a close eye on the nutrient concentration of the blood and there are minimal barriers for insulin diffusion, allowing a rapid delivery of it directly to the liver via the portal circulation, and then to the rest of the organs and tissues. 

Glucose sensing by β-cells

It should be reasonably easy to explain the steps of how this happens at a molecular level, but unfortunately most authors seem to resort to acronym-rich diagrams, and to look upon them can invite despair. Instead, it may be better to describe it using words. So, here's a point-form paraphrase of Fu et al (2013):

• Glucose diffuses into β-cells through the GLUT-2 glucose transporter, which is a non-insulin-dependent transport protein expressed on their surface. There is enough of these pores on the cell surface that glucose rapidly equilibrates its concentration on both sides of the membrane. How quickly? Johnson et al (1990) measured the rate of diffusion and found that the intracellular concentration was about 90% of the extracellular concentration within about 60 seconds.
• Inside the β-cells, glucose is phosphorylated by glucokinase, an enzyme that is related to all your other hexokinases, except for having an unusually low affinity for its substrate (it requires about 4-6 mmol/L of glucose to get working)  and the distinct lack of normal regulatory behaviour (as in, it does not get inhibited by its own product). That means it is inactive within a normal range of glucose concentrations, but it starts working and keeps working as the glucose concentration increases beyond the normal range of BSL. This is the key feature that makes this enzyme a glucose sensor: it is turned off by hypoglycaemia. 
• Thus, where BSL is in the higher range of values, the glucose which has diffused into the β-cells ends up immediately phosphorylated into glucose-6-phosphate. From here, its path is reasonably predictable: it plugs into normal metabolic machinery and is consumed as a metabolic fuel source, producing ATP. It can also be metabolised into intermediates leading to the generation of diacylglycerol (DAG) which acts as an intracellular second messenger. The bottom line is that the end result of increased intracellular glucose-6-phosphate in β-cells is the increased availability of the products of glucose metabolism, of which the most important ones are ATP and DAG. This is where amino acid levels and free fatty acids also join the pathway: free fatty acids can act as a substrate for the synthesis of DAG, and amino acids (well, mainly glutamine) can enter the tricarboxylic acid cycle and produce ATP.
• ATP and DAG are then the main intermediate signal molecules that trigger insulin release from β-cell granules. They do this in slightly different ways:
  • ATP influences potassium flux across the β-cell membrane by inhibiting ATP-sensitive potassium channels, which stops the potassium leak out of β-cells and therefore promotes their depolarisation (as intracellular cations are now accumulating). These channels are worth knowing about mainly because they are the pharmacological target of sulfonylureas, which inactivate them, promoting β-cell depolarisation.
  • Depolarisation of the β-cell membrane leads to calcium influx, and calcium influx leads to the activation of all the important exocytosis-mediating membrane proteins. Thus, increased intracellular ATP leads to insulin release.
  • DAG seems to mainly influence the activity of protein kinase C, which does not seem to be essential for normal nutrient-related insulin release.

In short, glucose metabolism by β-cells is the main mechanism that triggers insulin release. One might have expected a system as critically important as this to have some sort of direct glucose sensor with glucose as the ligand (a glucose-gated sodium channel would have been nice, or at least something coupled to a G-protein). But there is none, and to ask why would enter the realm of speculative fiction. Instead, glucokinase must play the role of the sensor, and the affinity of glucokinase for glucose is the key property that determines the homeostatic range.

Insulin exocytosis from β-cell granules

Insulin secretion is ultimately mediated by the appearance of intracellular calcium in the β-cells. According to Fu et al (2013), with enough intracellular calcium, the rate of membrane fusion and release can be up to 500 insulin granules per second. There would be little reason to go into much detail about the exact mechanisms of granule priming and SNARE protein mediated membrane fusion, as it would be unlikely to ever become the topic of an exam question, and there are no pharmacological targets in that pathway. It is, however, worth knowing that some granules do not require much calcium for their exocytosis, and these are constantly being emptied into the bloodstream, supplying a constant low-dose basal insulin concentration. Which is a logical segue into:

Pattern of normal insulin secretion

It is probably most convenient to divide insulin secretion into two main components:

• Basal insulin secretion, which occurs constantly during the fasted state, and
• Stimulated insulin secretion, which occurs in response to the intake of nutrients. 

You could fairly call the latter "postprandial" or "glucose-induced" or some such, but "stimulated" is a more accurate term because there are multiple possible stimuli to insulin secretion, of which glucose is only the most important and potent.

Stimuli for, and inhibitors of, insulin secretion

There's a chance somebody somewhere will one day need to list the things that influence insulin secretion which are not glucose. The list below was put together using a chapter from Introduction to Psychoneuroimmunology by Daruna (2012):

• Insulin secretion is inhibited by:
  • Adrenaline and noradrenaline
  • Sympathetic stimulation
  • Parathyroid hormone
  • Somatostatin
  • Cortisol (acutely)
  • Pancreatic polypeptide
  • Inflammatory cytokines
• Insulin secretion is stimulated by:
  • Parasympathetic stimulation (vagal tone)
  • Growth hormone
  • Cortisol (chronically)
  • Prolactin
  • Gonadotropins

And in the absence of any such stimuli, in the fasted state, islet β-cells will continue to secrete insulin, albeit unenthusiastically. 

Cortisol is listed in Daruna (2012) as a stimulant for insulin release mainly because chronically elevated cortisol levels (or, for that matter, sustained exposure to exogenous corticosteroids) will eventually give rise to a state of insulin resistance which ultimately results in a baseline hyperinsulinemia (Protzek et al, 2016). Under normal circumstances however cortisol is a counter-regulatory hormone and reduces both the secretion and the effects of insulin.

Basal insulin secretion

One sometimes hears this described as "postabsorptive" insulin secretion, as opposed to the secretion of insulin which occurs when you are absorbing something.  Practically speaking, the prescriber will recognise this as the physiological equivalent of the pharmacological long-acting insulin which accounts for about 50% of the total daily insulin dose. This long acting insulin dose is meant to supplement the baseline secretion of insulin which is missing in the diabetic. For the non-diabetic, this secretion rate is something like 0.5-1.0 units per hour. It is continuous, but by no means constant - the level of insulin seems to oscillate in a pulsatile manner, with one "ultradian" cycle lasting 1-2 hours, and even shorter rapid oscillations with a period of 5-10 minutes.  It really is one of these things that are better explained with images, so here's a composite created by combining work from Schmitz et al (2008) with  Simon & Brandenberger (2002):

In case the suspicious reader is wondering whether those hourly peaks of insulin are the result of the experimental subject's uncontrollable frequent snacking, be reassured: that chart from Simon and Brandenburger was measured in a patient receiving a stable continuous stream of enteral nutrition. This cyclical rhythmicity of insulin secretion seems to be completely independent of any fluctuations in glucose delivery.

What is this, and why are we looking at it? The short answer is, nobody seems to know.  It is not clear what the point of this pattern is, or whether it is beneficial - but it does not seem to be critical to supporting life, insofar as the peristaltic drug pump loaded with insulin does not make any attempt to duplicate this pattern,  with no apparent harm to the patient. Still, apparently, about 75% of the total daily insulin "dose" is secreted in this fashion, so it is probably worth knowing about. The mechanism appears to be the electrical coupling of neighbouring islet cells which all reach a threshold together and all depolarise at once (which produces a burst of insulin)- instead of each cell gradually releasing a steady stream of granule content (which they also do, but apparently not enough to support basal insulin requirements). Peercy & Sherman (2022) describe this with a lovely metaphor, as a  voting/democratic paradigm, where the islet properties are a nonlinear average of the cell properties, with no ‘conductor leading the orchestra’"

Insulin secretion associated with carbohydrate intake

Glucose is the most potent stimulus of insulin secretion. All the other listed stimuli are merely regulatory by comparison, i.e they gently encourage the β-cells to secrete more, or less, insulin. By comparison, glucose is some steps above that in the scale of graded assertiveness. Glucose commands insulin release.

It is not clear whether it would be important for an intensivist to know that this release occurs in a biphasic pattern, but it is. Observe this plot of insulin concentration over time, collected by Blackard & Nelson (1970) following the rapid infusion of 25 grams of glucose.

As you can see, within seconds of glucose appearing in the bloodstream, insulin is dumped into the portal circulation. Henquin et al (2002) assert that this represents the release of insulin from a ready releasable pool of granules, mediated by a calcium-based mechanism as described above. These would be some highly calcium-sensitive exocytosis, i.e. only a modest elevation of intracellular calcium is required to pull this trigger. The slower delayed release of insulin is thought to be due to the amplifying effects of glucose on the calcium sensitivity of the exocytosis processes, making more and more granules ( a reserve pool) available for exocytosis at a lower calcium concentration. From the same paper, this stolen vandalised graphic demonstrates that early insulin secretion is coupled to a vigorous peak in intracellular calcium, but the second phase of insulin secretion remains robust despite a rather modest calcium level.

What point is there in knowing any of this, the pragmatic intensivist might bellow while pounding the table, when most of the insulin in my patients will come from a drug pump? A fair comment. Much is made of glycaemic control in critical care literature, as hyperglycaemia is clearly bad for basically all patients in the ICU population, at the same time as critical illness itself has a tendency to produce it by various stress response mechanisms. However, it would be unfair to say that in the ICU the intensivist controls the insulin. Duška & Anděl (2008) measured plasma insulin and C-peptide levels of some critically ill trauma patients and determined that in fact most of the insulin of their subjects was endogenously secreted, i.e. the exogenous insulin infusion only supplemented them with a fraction of their natural production rate.

Insulin receptors

This next section more properly resembles the "pharmacodynamics" of insulin. It is properly described as an "anabolic" peptide hormone, using the proper biological definition of the term, because it produces constructive metabolism and has various tissue growth-enhancing effects. It is also sometimes described as "pleiotropic", i.e. having multiple simultaneous effects. The CICM trainee may relish the opportunity to appear clever by using both of these words to describe insulin in the written paper, but in fact the examiners may not know what "pleiotropic" means, and they might resent the suggestion to go look it up, which means the effort may be wasted.

Insulin receptors

The insulin receptor is a huge chunky transmembrane protein, about 320 kDa in mass, the intracellular domain of which acts as a tyrosine kinase. It belongs to a huge family of receptor tyrosine kinase proteins which are highly conserved from an evolutionary perspective, and are present in basically all multicellular animals in some form or another. Insulin receptors are present on all mammalian cells, but their distribution is unequal - Watanabe et al (1998) counted only forty or so per every erythrocyte, whereas white adipose tissue cells had something like 300,000 each. 

It is possible to get through the insulin questions in the CICM papers without getting involved in too much molecular biochemistry, but if the reader is the sort of person who might retort that there is no such thing as too much molecular biochemistry, they are politely redirected to this Endotext paper by Peter De Meyts (2016). For the rest of us, it will suffice to say that insulin binding to this receptor leads to a signal transduction cascade that culminates in the activation of numerous protein kinases and gene transcription factors. Of these the most important are the PI3K signalling pathway (which leads to all the metabolic effects) and the MAPK pathway which leads to all the gene expression and cell growth/differentiation effects. 

Intracellular molecular effects of insulin on glucose uptake

The effect of insulin binding on the traffic of glucose is probably the most important mechanism for the exam candidate to study. It can be summarised as follows: 

• Target tissue cells contain vesicles which, on their inner membrane surface, express GLUT4 glucose transport proteins. 
• In response to the binding of insulin to its receptors, these cells perform a sort of exocytosis, where these vesicles migrate to the surface membrane and fuse with it
• The end result is the appearance of more GLUT4 transport proteins on the cell surface

Insulin increases the availability of GLUT4 glucose transport proteins on the surface of cells (Chang et al, 2004). Glucose is a relatively small molecule (180 Da), but it is relatively hydrophilic, which means that passage through the lipid bilayer is a major challenge, and so the rate of glucose transport into cells is completely dependent on these various forms of facilitated transport. Bernard Thorens (1993) describes the function of the available transporters in a little more detail, in case anybody needs that. GLUT4 is only one of the available GLUT isoforms, present mainly on the surface of specialised endosomes in skeletal muscle, cardiac muscle, and adipose tissue. Brain tissue can and does express insulin-responsive GLUT4 transporters but is not dependent on these, and will continue to sip glucose through all the other GLUTs even in the absence of insulin stimulation.

GLUT4 proteins are described as "insulin regulated" mainly because they only appear in response to insulin, and not because insulin has some direct effect on their intrinsic function. These are, for all their marvellous complexity, dumb tubes: they allow glucose to slide down its concentration gradient into the cells. This concentration gradient is of course maintained by hexokinase, which quickly snapples up any intracellular glucose and converts it into glucose-6-phosphate. 

So: insulin comes along and binds its transmembrane receptor, activating its intracellular tyrosine kinase domain. The signal amplification cascade that follows is complex enough that we might leave it to Stöckli et al (2011) to properly describe. For the CICM exam candidate, the most important signalling pathway is probably the PI3K (phosphatidylinositol 3-kinase) cascade, which generates the secondary messenger phosphatidylinositol 3,4,5-trisphosphate (PIP₃) and which becomes important later in this discussion (as it is involved in several other non-glycaemic effects of insulin). For glucose uptake, the net effect of the intracellular signalling pathways is the translocation of GLUT4-rich endosomes to the surface of the cell, where they merge with the exterior cell membrane. In this fashion, the density of GLUT4 transporters on the cell surface increases, and this produces an increase in the permeability of the membrane for glucose. 

Specific physiological effects of insulin 

"Specific physiological effects of insulin of interest to the intensivist" would have been a better section heading. What follows is a brief discussion of how insulin does all the things we ICU doctors rely on it for: manipulate blood glucose levels, push electrolytes around, lower dangerously elevated blood lipids, and support cardiac contractility in cases of β-blocker overdose. There's a lot of infuriatingly complex molecular biochemistry in there, which would be entirely excessive for an ICU primary exam candidate, and so instead of patiently unpacking all the proteinaceous intracellular gubbins, the readers will be offered short summaries for the now, and links to detailed references for the future.

Glucose uptake

Without labouring the point any further, yes insulin causes the uptake of glucose into tissues. It is generally said that the dose required to drop your BSL by 2 mmol/L is 1 unit of insulin, which represents an intracellular movement of about 10-15g of carbohydrate (equivalent to 200-300ml of 5% dextrose). This is the sort of quotable dose-response relationship that is inevitably wrong in 90% of people, but most individuals fall within a certain safe range near these values, which allows crude intern-level insulin sliding-scale prescriptions to remain non-lethal. 

This dose-response relationship is actually a composite of several individual tissue dose responses, as not every cell type has the same appetite for glucose, and not all of them express GLUT4 channel proteins. Specifically, the insulin-mediated uptake is into "insulin-sensitive" tissues, which are skeletal muscle, liver and adipose tissue. Crudely reapproximating some data from a study by Lang et al (1990) which surely generated a massive pile of slightly sweetened rat corpses, the following dose-response curves can be traced for different tissues and organs:

From this, it might seem as if the myocardium is the main destination for insulinergic glucose shifts, but no - quantitatively it is a minor player. Skeletal muscle probably accounts for the majority of the uptake quantitatively - some authors estimate 75-80% of the total glucose movement - mainly because there is a lot more of these cells in the body (30kg or so, as compared to a paltry 1kg of hepatocytes). 

Glycogen and fat storage

The pancreas is perfectly positioned to command over the liver: its hormones are secreted directly into the portal circulation, where their concentration is often orders of magnitude above what might be expected in the systemic bloodstream. It is perhaps because of this that the liver can afford to have such a sluggish dose-response relationship with insulin (from the diagram above, the reader will see that it is second even to the skin and kidney). One does not need to be especially sensitive to a stimulus when one is constantly being beaten about the head with it. It is no wonder that Edgerton et al (2006) used the word "dominate" to describe how insulin influences hepatic glucose metabolism.

Is that hepatic insensitivity to insulin important for the intensivist, one might ask, after rolling the question around for a while. The answer is probably no, because we tend to titrate our infusion to the end result, which is the systemic blood glucose concentration. True, the liver usually expects to see a high dose of insulin coming from the pancreas, and is underwhelmed by lower concentrations coming from the systemic circulation. When Edgerton et al switched their experiment from portal to systemic insulin infusion, the hepatic glucose synthesis rate immediately doubled, even though there was systemic hyperglycaemia and hyperinsulinaemia. Still, all one has to do is increase the dose rate of the systemic infusion, and one will be able to achieve one's goal BSL. Recall that the normal pancreas produces 0.5-1.0 units per hour, whereas a common systemic insulin dose in the ICU may be 3-5 U/hr. In that sense, when we prescribe insulin, we prescribe it specifically to the liver of the critically ill patient. 

Anyway: when insulin binds hepatocyte receptors, it:

• Reduces the rate of gluconeogenesis
• Reduces the rate of glycogenolysis
• Reduces the rate of free fatty acid oxidation (and therefore ketone production)
• Increases the rate of glycogen synthesis
• Increases the rate of synthesis of VLDLs
• Decreases the rate of hepatic urea synthesis (mainly by decreasing amino acid deamination for gluconeogenesis)

The rate of glucose entry into hepatocytes increases with insulin, but most glucose entry into hepatocytes is via the insulin-insensitive GLUT2 transporter. Insulin exerts more control over what happens to the glucose when it is inside the cell, mainly by regulating the activity of hexokinase which phosphorylates glucose into glucose-6-phosphate, trapping it in the cells. The increased rate of glucose phosphorylation maintains the concentration gradient which keeps glucose flowing into the cell.

Intracellular movement of phosphate

Insulin causes the intracellular movement of phosphate. This appears to be a sodium-dependent mechanism, and happens within minutes of insulin exposure, which suggests that it is associated with glucose uptake rather than anything else, but nobody seems to have a clear idea of what exactly mediates this process, or why it happens.

Most authors seem to believe that there is some kind of need for phosphate to accompany glucose.  True, entrained glucose is used to generate glucose-6-phosphate, and then ATP, which means it would be logical to expect the cell to want a phosphorus atom to go with every glucose molecule, but this is not observed: for example, Riley et al (1979) determined that the drop in phosphate was relatively modest with normal doses of insulin (it fell by 25% or so), and only became clinically interesting with doses in excess of 1u/kg/hr.

This process becomes more relevant in states of intracellular phosphate depletion, where the capacity for intracellular movement become greater - the classic example of this is DKA. From this, we can speculate that the theory about phosphate shifting intracellularly to participate in glucose metabolism is basically true, and you just don't see it very much when there's already plenty of phosphate in the cell.

Intracellular movement of potassium

Insulin, and an accompanying dose of concentrated dextrose, are a familiar medical instrument used to manage hyperkalemia. There are several mechanisms implicated in this, which Nguyen et al (2011) boil down to the following steps:

• An increase in the entry of sodium into cells via the Na⁺/H⁺  exchange channel occurs, and intracellular sodium concentration can rise from 17 to 30 mmol/L (Rosić et al, 1985).
• This produces an increase in the activity of Na⁺/K⁺ ATPase (Gourley et al, 1965), which  removes all the excess sodium and increases the intracellular movement of potassium
• At the same time there is an inhibition of potassium efflux from cells (Zierler ey al, 1966), which means all the potassium being pumped into the cell cannot escape.

The specific protein signalling cascade involved in this process seems to have eluded investigators, or their findings are tagged inappropriately, making their research difficult to find. Fortunately, one does not need to know these in order to safely use insulin as a potassium-lowering agent. Insulin results in a rapid and reliable dose-dependent potassium decrease, where 10 units of insulin and 50ml of 50% dextrose (25g) lower serum potassium levels by about 0.8-1.0 mmol/L.

The graph here comes from Ljutić & Rumboldt (1993), who published on their experience with this 10-and-50 protocol. In case it does not spontaneously strike the reader as obvious, you give the glucose first. Apparently prior to this study all the textbooks contained some completely random dose recommendations for this hyperkalemia cocktail, to the effect that hypoglycaemia was a common side-effect. By concurrently measuring C-peptide levels, these authors were able to demonstrate that, if you give the 50% dextrose over five minutes, a burst of endogenous insulin secretion assists the subsequent insulin bolus and achieves a greater reduction in glucose without the risk of hypoglycaemia.

This potassium reshuffling is obviously not a permanent solution, as you have not really removed any ions. Over the subsequent 4-6 hours, it redistributes back into its original arrangement, which means you need to think of something else. However it does buy time, for example "until the hospital's dialysis unit opens for regular working hours" (Emmett, 2001). 

Accelerated metabolism of triglycerides

Berserk hyperlipidaemia, of the sort that produces pancreatitis and causes people to hoot excitedly at supernatant in their blood tubes, is often treated with an insulin infusion. Sure, you could also separate the fat using plasmapheresis, but its main advantage is really only speed and by the time you manage to get the machine started the insulin will have already been working for hours (to say nothing of the difficulty and expense involved in getting a hold of a plasma exchange device on the weekend, or in a remote rural setting). In any case, it's the free fatty acids that do all the damage, and the triglyceride levels are probably unimportant on their own. 

Anyway: this is not the right forum to air the author's bias against plasmapheresis in hyperlipidaemia. The mechanism of action of insulin in this scenario appears to be threefold. On one hand, insulin increases the activity of lipoprotein lipase in adipose tissue, which breaks down the circulating triglycerides, releasing free fatty acids (Sadur et al, 1982). On the other hand, insulin increases the uptake of triglycerides and free fatty acids into adipose tissue, increasing their clearance from the bloodstream. And on the ...third hand, insulin inhibits the hydrolysis of existing triglycerides in fatty tissue, preventing any further free fatty acid release. The net effect is a decrease in plasma triglyceride levels, dropping them by 50% within 24 hours in one case series by Afari et al (2015).

Haemodynamic effects of insulin

Acutely, insulin is an inodilator, and the halves of this portmanteau (the "ino" and the "dilator") each has a distinct physiological mechanism.

Insulin reduces afterload by decreasing the peripheral arterial resistance, acting mainly on skeletal muscle arteries. It appears to be doing this by some mechanism that increases the amount of endothelial nitric oxide, which would have been a pretty good guess even without any research (as that tends to be the final common pathway of many vasodilators). We know this because nitric oxide synthase inhibitors tend to largely abolish the vasodilatory effects of insulin (Steinberg et al, 1994). It probably also hyperpolarises the smooth muscle membrane by stimulating Na⁺/K⁺ ATPase (Kawasaki et al, 2000). Exactly how all of this happens at a molecular level remains to be established; some experts (eg. Engebretsen et al, 2011) blame the PI₃K pathway, but offer no references to support this. Most people tend to agree that the effect is substantial. Steinberg et al infused about one unit per hour directly into the femoral arteries of volunteers and found that leg blood flow increased by about 50%. 

Insulin is also an inotrope. We've known about this for some hundred years or so, as the very first pancreatic islet juice patiently extracted by Visscher & Müller (1926) produced visible changes to the circulatory environment of other experiments ("Our attention was drawn to the subject by entirely unexpected observations while injecting insulin into the heart-lung preparation", they remarked). Countless subsequent experiments on what sounds like an adorable collection of loveable pets, as well as data from presumably less adorable adult humans, has confirmed that insulin has a positive inotropic effect, which Klein & Visser (2010) likened to the effects of a low-dose dobutamine infusion. The comparison is probably because several studies from the eighties and nineties (when pulmonary artery catheters were mandatory in all ICU patients) compared the two drugs directly, as for example did Hiesmayr et al (1995):

A 16% improvement of cardiac index is nothing to scoff at, but note the dose. That'd be about 100 units per hour for most normal sized patients, i.e. roughly twenty times more than what you would normally infuse for routine blood sugar control in the ICU. This dose range more closely resembles the use of high dose insulin euglycaemic therapy (HIET), such as is used for calcium channel blocker and β-blocker toxicity.

How does insulin do this? The short answer is, nobody really knows. Most ICU consultants, when pushed to the wall and confronted with this question, will blather something about increasing the concentration of cyclic AMP, thereby bypassing the blockade of adrenergic G-protein-coupled receptors, but this is actually either totally incorrect or at best only part of the picture. There are probably several processes at play:

• Sympathoexcitatory effects: insulin is not just comparable to catecholamines, it literally causes an increase in catecholamine levels. Scherrer & Sartori (1997) list a whole host of different experiments where this has been conclusively demonstrated. Weirder still, this seems to be a central mechanism: for Lembo et al (1991), only systemic insulin caused the noradrenaline levels in a forearm to increase, while infusing insulin directly into the same forearm had basically no effect on plasma noradrenaline. Following from this, researchers have implied that chronic hyperinsulinaemia causes a chronic sympathetic activation that might be responsible for some of the cardiovascular badness associated with Type 2 diabetes.
• Increased myocyte calcium concentration: experiments on explanted trabeculae by von Lewinski et al (2005) demonstrated that intracellular calcium increases were essential for the inotropic effect of insulin, and that the increased availability of intracellular sodium (as substrate for the INCX sodium/calcium exchange current) is probably the key. Whatever the specific molecular machinery involved here, the main activator seems to be PI₃K, because disabling PI₃K seems to block the inotropic effects of insulin
• Increased myocardial carbohydrate metabolism is supposedly beneficial, as the stressed myocardium prefers to use carbohydrates, apparently. Engebretsen et al (2011) report that the increase in glucose entry into the cardiac myocytes is essential. The way this is explained by Lheureux et al (2017) is:
  • Myocardial cells normally prefer to metabolise fatty acids, which is an efficient process with a RQ of 0.7
  • Myocardial cells switch to the metabolism of carbohydrates under conditions of stress, where glucose and lactate become the preferred substrates, and the RQ shifts closer to 1.0. This seems to be mediated by the activation of AMP-sensitive protein kinases (Young et al, 1999), and the whole point seems to be to increase myocardial performance at the cost of efficiency (making more metabolic substrate available)
  • This metabolic pathway is less efficient, and required the delivery of more oxygenated blood and more glucose. But the myocardium may not be able to get more blood, because it is under stress and the contractility may be affected. 
  • The increased translocation of insulin-sensitive GLUT4 channels to the membrane may therefore increase the uptake of glucose by the myocardium in spite of fixed coronary blood flow, basically increasing the "glucose extraction ratio" of the coronary circulation.
• Increased coronary blood flow because of the arterial vasodilator effects of insulin already mentioned above, which may be responsible for the "recruitment" of previously idle myocardial wall segments in patients with chronic ischaemic heart disease
• Decreased afterload by the same vasodilatory mechanism, but systemically.

Effect of insulin on target tissues

Having just gone over the physiological effects of insulin in some detail, the need to rearrange these bits of information into a different structure only becomes apparent when one considers how examiners think. "List the effect of insulin on different tissues" or "describe the effect of insulin administration on specific organ systems" are a fair possibility. This would require the answer to be organised slightly differently. The best reference for something like this would have to be Newsholme & Dimitriades et al (2001), whose paper has the right level of clarity and detail. 

• Effect of insulin on the liver
  • Reduced gluconeogenesis and urea synthesis
  • Reduced glycogenolysis
  • Reduced free fatty acid oxidation
  • Reduced ketone production
  • Increased glycogen synthesis
  • Increased VLDL synthesis
• Effect of insulin on the skeletal muscle
  • Increased glucose uptake
  • Increased glycogen synthesis
  • Increased protein synthesis and decreased protein catabolism
• Effect of insulin on the myocardium
  • Increased carbohydrate utilisation 
  • Increased  contractility
  • Increased coronary blood flow
  • Decreased afterload
• Effect of insulin on adipose tissue
  • Increased glucose uptake
  • Increased synthesis of triglycerides and decreased lipolysis
  • Increased lipoprotein lipase activity
  • Increased uptake of free fatty acids
• Effect of insulin on the brain
  • Increased sympathetic nervous system activity
• Effect of insulin on the pancreas
  • Decreased glucagon release (your pancreas, wisely acknowledging that it would defy logic to secrete both of these hormones simultaneously, operates a negative feedback loop where one inhibits the secretion of the other.) 

Consequences of insulin deficiency

Finally, we arrive at something substantially related to the CICM First Part Exam. Question 8 from the second paper of 2010 is the only "physiology of insulin" question available, and it asked specifically for the physiological consequences of "an inability to produce insulin". This sounds like a question about endocrine function, but the college answer makes it seem like mainly a question about DKA in a Type 1 diabetic. Thankfully, this was back in 2010, when the examiners still recognised that assessment drives learning, and they included a page reference to the 10th edition of Guyton and Hall (2000, p.888) to explain what they were expecting.

Thus, working from this:

• Decreased insulin availability leads to:
  • Decreased glucose uptake into skeletal muscle
  • Increased reliance on glycogenolysis and gluconeogenesis in liver and muscle
  • Increased hormone-sensitive lipase activity, leading to an increased release of free fatty acids from adipose tissue
  • Lack of insulin regulation in the liver leads to increased fatty acid oxidation
  • Increased availability of acetyl CoA liberated through fatty acid oxidation causes increased ketone synthesis because of excess acetyl CoA being metabolised into acetoacetyl-CoA and ultimately into the ketones acetoacetate and β-hydroxybutyrate
• The consequences of this are:
  • Hyperglycaemia
    • Thus:
      • osmotic diuresis
      • volume depletion
      • compensatory cardiovascular changes in response to volume loss (eg. tachycardia)
    • With volume loss, electrolytes are also lost, leading to a total body potassium and phosphate deficit
  • Metabolic acidosis
    • Thus, increased respiratory rate to compensate by lowering systemic CO₂
    • Increased anion gap (ketones are unmeasured anions)