Saturday, January 20, 2018

Metformin (07) glargine 50iu/kg causes diabetes

The group which demonstrated that exogenous insulin induces insulin resistance in T1DM NOD mice went on to demonstrate that exogenous insulin induces T2DM in normal* mice on a normal chow diet.

*If you can describe Bl/6 mice as normal, with their failure to assemble mitochondrial super complexes and all of the potential implications that has on all sorts of metabolic effects. But I digress, as always.

Exposure to excess insulin (glargine) induces type 2 diabetes mellitus in mice fed on a chow diet

So you can imagine that I quite like this group. But they are naughty and the naughtiness is very annoying.

If you read about insulin administration in the current paper you will be presented with this bollocks:

"The dose of glargine was determined according to our previous experiments (Liu et al. 2009a )..."

What they did in Liu et al 2009a was to titrate the dose of insulin determir upwards to just achieve normoglycaemia. Different doses were needed in individual mice because that's what diabetes is like in NOD mice. That is NOT what they did in this current paper to the Bl/6 mice. Here they used glargine and they went straight in at a massive supraphysiological dose:

"C57BL/6 (B6) mice (male, 6–9-week old) from Charles River were treated with either saline or a long- and slow-acting insulin reagent, glargine (50 Unit/kg body weight, s.c. injection, once a day), for 8 weeks".

Even allowing for metabolic scaling from humans down to mice this is a massive dose of glargine. It's not remotely what they did with the NOD mice.

EDIT: Going back through the first paper the NOD T1 mice did actually end up with the average group dose of detemir being 25iu/kg twice daily. The difference in protocol is that the group is assuming that glargine 50iu/kg once daily is equivalent to detemir 25iu/kg twice daily. A big assumption. And that this would be fine for all mice. And that going in at the full dose rate on day one rather than titrating up over two weeks would also be equivalent. But the dose rate is more reasonable than I expected. END EDIT.

So there are fibs in the methods. Can you really trust these folks? Not much choice really...

None of the mice died on the glargine 50iu/kg dose, so I think we can assume that they developed a marked and rapid onset insulin-induced insulin resistance.

When you wade through the IRS1 serine phosphorylation, IRS1 tyrosine phosphorylation, Akt phosphorylation, total Akt etc etc etc in the results section the end conclusion is that the liver became insulin resistant but the gastrocnemius muscle (representing what I called "systemic" tissues) did not. Bummer for my nice, plausible and apparently incorrect ideas.

So to ease my cognitative dissonance I went back to our initial paper and waded through the IRS1 serine phosphorylation, IRS1 tyrosine phosphorylation, Akt phosphorylation, total Akt etc etc etc and discoved that, lo and behold, that under a clinical protocol the gastrocnemius did become insulin resistant. So did the liver but I can live with that, after all the liver does have a (bloody enormous) arterial blood supply in addition to receiving flow from the portal vein.

Phew, biases in-tact.

It's interesting to see in the paper that exogenous glargine actually destroys the pancreas. Beta cell mass falls, their mitochondria undergo marked oxidative stress and this allows the failure to deal with the hepatic insulin resistance induced by glargine 50iu/kg, hence the induction of diabetes.

I think the take home message is that taking glargine 50iu/kg won't help your own clinical diabetes, should you be so affected. It's also too high a dose if you are attempting to do harm!

Anyhow. Growth hormone and insulin resistance next, probably.


Friday, January 19, 2018

Metformin (06) Insulin-induced insulin resistance is real

When I started reading about insulin-induced insulin resistance I began with this paper:

Insulin Is a Stronger Inducer of Insulin Resistance than Hyperglycemia in Mice with Type 1 Diabetes Mellitus (T1DM)

It's a nice paper. They took NOD mice which had developed their NOD mouse version of T1DM and either treated them with insulin detemir, or didn't. They had a third group which never developed T1DM so were never treated and these served as a control group.

The treated diabetic NOD mice were gradually stabilised over a two week period then kept normoglycaemic for a further two days. They were assessed for insulin sensitivity using an insulin tolerance test, where a dose of neutral insulin is injected then you track what happens to the blood glucose concentration. The more insulin sensitive the animal, the more the glucose level drops:

It's pretty obvious that the detemir treated mice (top line) have absolutely no response to neutral insulin and that both non-treated diabetic mice and never-diabetic mice drop their blood glucose levels by about 50% on this particular dose of neutral insulin.

I could stop this post here. Exogenous insulin induces insulin resistance in T1DM mice, as it does in people. This is fact.

But of course you should not just accept this. The question is: Why?

Why does "enough" insulin as secreted by the pancreas to produce normoglycaemia (in the never-diabetic control group of mice) cause no insulin resistance whereas insulin detemir given to produce the same level of normoglycaemia induces striking insulin resistance in those treated NOD mice?

Recall that the hyperglycaemia in T1DM has little to do with the lack of insulin per se. The hyperglycaemia is caused by an excess of glucagon from the alpha cells of the pancreas. Insulin starts its control of hyperglycaemia by the suppression of pancreatic glucagon secretion, it's a local action within the islets. How high this concentration of insulin is under normal physiological conditions is quite hard to determine but it is likely to be a lot higher than the diluted insulin concentration in the portal vein, heading towards the liver.

The diluted insulin within the portal vein arrives at the liver where its next job is to suppress hepatic glucose output, again in antagonism to glucagon.

Finally, if glucose from the liver continues to enter the systemic circulation, the function of insulin here is to push that glucose in to any cells that will take it. Muscle and adipose tissue being two major targets.

So under normal physiology there is a gradient of insulin concentrations from very high within the Islets of Langerhans, to significantly lower at the hepatocytes, down to much lower in the systemic circulation.

Exogenous insulin produces no such gradient. It drains from its injection site into the systemic veins and is then redistributed, at a single concentration, throughout the body.

This will never effectively suppress alpha cell glucagon secretion and will only do a modestly effective job of suppressing hepatic glucose output. So glucose will be continuously secreted in to the systemic circulation. The dose of detemir used has to be enough to mop up this excess glucose supply, and it can only put it in to cells sensitive to insulin throughout the body. Muscle cells. Adipocytes.

Now look at it from recipient cell's point of view. Glucagon is high, hepatic glucose output is high and this continuous supply of glucose is being allowed in to systemic cells by the exogenous insulin. To these cells the glucose supply looks like a meal being digested (high glucose, high insulin). The cells rapidly realise that they have enough calories (High NADH levels, high mitochondrial delta psi, reduced electron transport chain). They don't want any more. Their solution: insulin-induced insulin resistance (ie reverse electron transport to generate H2O2 at complex I and so inactivate insulin signalling).

That's what happens.

So the very effective control of blood glucose in these NOD mice is at the cost of continuous exposure to supraphysiological insulin levels coupled with a supraphysiological glucose supply, because the systemic cells are "covering for" the failure to replicate the normal gradient from islet to liver to systemic circulation.

Exogenous insulin can never be physiological.

Aside: Except, of course, under deeply ketogenic eating where only minimal insulin is ever secreted, very little is metabolised, the gradients between alpha cells, hepatocytes and adipocytes flattens out and the correct physiology is for glucagon to be elevated with minimal insulin. I've posted this before. T1DM patients have no choice, ketosis is the only physiological state which can be fairly well mimicked using very low doses of exogenous insulin. End aside.

I would never suggest that exogenous insulin has no effect on pancreatic glucagon secretion or elicits no suppression of hepatic glucose output. It will always have some effect, but there will always be an abnormal emphasis of its effect on systemic tissues.

This is the situation in NOD mice and T1DM people. They become insulin resistant by simply using exogenous insulin to ensure normoglycaemia.

I guess the next question, which was asked by the same group whose paper we've just been looking at, is whether simply injecting exogenous insulin in to normal mice induces insulin resistance. That's the next paper, in the next post.

Spoiler: Of course it does.


Sunday, January 14, 2018

A wander off in to dietary protein calories

There is prize for developing the longest-lived mouse in the world. It was set up in 2003 and the first award went to Dr Bartke.

"On June 8th, 2003, the inaugural Methuselah Prize was awarded to Dr. Andrzej Bartke for the "Methuselah Mouse" that lived the equivalent of 180 human years".

You can read a bit more about growth hormone receptor knockout mice and other forms of dwarf mice in Dr Bartke's review, written soon after winning the prize:

Life extension in the dwarf mouse.

It's now 2018 and no one appears to have improved on the Laron mouse model which won that initial prize. Over the last 15 years there has been a lot of interesting research but no numerical progress. I think it is worth noting that Laron mice are not GH deficient, they have tons of the stuff. They simply do not have the receptor to do anything with it. Which, in particular, means they cannot generate IGF-1.

How do Laron humans fare? The best studied group live in Colombia. They're of very short stature. They have no recorded cases of diabetes and only one recorded cancer, which was non lethal*. Their every biochemical parameter is exemplary, especially insulin level and HOMA score. Do they all live to be centenarians? Apparently not. Being a dwarf in Columbia requires alcohol in large amounts to render life tolerable, plus accidental trauma is another huge problem. Quite what would happen if these people lived under similar conditions to the Laron mice in Dr Bartke's laboratory is a question which is unlikely ever to be answered! Longevity in the real world vs what works under ideal conditions...

*Dr Laron has reported two cases of Laron Syndrome people developing diabetes (and there are others), including the complications such as atherosclerosis, renal disease and diabetic retinopathy. This is an interesting observation and might be worth a post on its own some time.

There are tantalising suggestions from other GH modifying mutations in humans. One of the better studied of these is carried by the "Little people" of Krk in Croatia. More from Dr Laron:

Do deficiencies in growth hormone and insulin-like growth factor-1 (IGF-1) shorten or prolong longevity?

Longevity of the hypopituitary patients from the island Krk: a follow-up study

They have a mutation which causes multiple pituitary hormone deficits, ACTH secretion excepted. There are too few documented people with this genetic problem to say a great deal about longevity but ages of 68, 77, 83 and 91 years have been recorded in the four individuals to have died since detailed observations began. The equivalent syndrome in mice under lab conditions promotes longevity.

One of the nicer studies looking at human height (viewing this as a GH/IGF-1 signalling surrogate) and longevity is this one:

Shorter Men Live Longer: Association of Height with Longevity and FOXO3 Genotype in American Men of Japanese Ancestry

It found, as you might expect, an inverse relationship between height at enrolment and longevity. They also tied the relationship, observationally, to a down-regulating SNP of the FOXO3 gene, FOXO genes being major controllers of the insulin/IGF-1 signalling system.

Which genes you have is not under your control. What you do with then might well be...

Let's finish this post with the LoBAG diet. It's modest (20% of calories) in carbohydrate, has 30% of calories from protein and the rest as fat. It's being compared to a diet with similar carbohydrate content, 15% of calories from protein, with the rest as fat. Lots of details in here:

The metabolic response to a high-protein, low-carbohydrate diet in men with type 2 diabetes mellitus

As they say in the discussion:

"The present data indicate that the increase in IGF-1 is the result of the increase in protein content. The further decrease in carbohydrate did not result in a further increase in IGF-1. In fact, the increase was approximately the same (138% and 136%, respectively)".

What interested me initially (and had made me chase the paper) was the effect on GH itself. The LoBAG diet actually drops GH levels, admittedly by a ns amount. What turns out to be a much more interesting incidental finding is that, despite the downward trend in GH, IGF-1 rises by a statistically significantly and possibly by a biologically significant amount. Especially when you consider a whole slew of cancers sprout IGF-1 receptors on their surface.

Brief aside. You have to be very careful with GH and IGF-1 levels in papers like this one as both hormones come with a whole load of plasma binding proteins which very few people, including the LoBAG folks, ever measure. These may well alter the effective concentration of the hormone either upwards or downwards. Caution is needed with simple measurements like those in this paper. End aside.

So. Folks should eat whatever they feel comfortable with, protein-wise. I probably eat a little more protein than I would prefer, but then I'm no perfectionist. What I wouldn't do is to add protein gratuitously to any meal...

But that's just me I guess.


Sunday, December 24, 2017

Metformin (05) Insulin Resistance

Happy Christmas all. Would have been Happy Solstice but one of our cats died that day. Anyhoo.

Back to this image from the Japanese paper of the last post, open circles are without metformin, filled circles are the same people taking metformin 500mg tid.

If we initially look at the fasting values for the RQ we see 0.8 without medication and 0.77 under metformin, so adding in metformin gives an increase in fat oxidation. A very simple explanation for this is that, via metformin induced blockade of mitochondrial glycerol 3 phosphate dehydrogenase (mtG3Pdh), there is less glycolysis derived electron input at the CoQ couple, so it is less reduced and so there is less tendency for reverse electron transport through complex I. This generates less of the superoxide necessary to trigger the initiation of insulin signalling. The cells then behave as if there is even less insulin present than the 25pmol/l measured, so free fatty acids are more available for oxidation due to this reduced insulin signalling.

The simple concept is of metformin as "LC eating in a pill".

The fed state is altogether different and diametrically opposite. Insulin levels are between 100 and 200pmol/l. Under these conditions the RQ is significantly higher under metformin, an RQ of 0.8 vs the non medicated RQ of 0.76, present at the two hour mark with the differential maintained at three hours. Under these conditions metformin is facilitating the oxidation of glucose while there are calories in excess of immediate needs available, many of them from the butter in the cookies.

If we consider that blockade of mtG3Pdh should blunt insulin signalling we have a paradox that by one hour after a meal insulin signalling appears to have been facilitated and at hours two and three this reaches statistical (and probably biological) significance. So how can the one drug have opposite effects under differing conditions?

My suspicion is that the drug is doing the same thing at all times but that insulin is doing different things at different nutrient availabilities. Re consider this graph from this post:

The initial conclusion here was that metformin only facilitates blood glucose reduction in the presence of insulin. Metformin should, theoretically, blunt the action of insulin. But if we consider that at high levels of insulin the function of that insulin is to limit its own action, I think it would be much better viewed as metformin blunts insulin induced insulin resistance. Insulin was bolused iv at 90 minutes. It will have given a massively supra-physiological plasma level. Insulin induced insulin resistance in the insulin treated group appears to be absent at 30 minutes (ie 120 minutes on the graph), to have started at 60 minutes (150 minutes on the graph) and to have gotten p to below 0.05 at 90 minutes (180 minutes on the graph). Of course under an-insulinaemic conditions there is no insulin signalling to facilitate or block, hence the zero to 90 minutes on the graph where metformin has no effect on blood glucose before insulin was bolused.

From Ivor Cummings (not sure where I got the actual paper from) we have this concept:

Insulin is a stronger inducer of insulin resistance than hyperglycemia in mice with type 1 diabetes mellitus (T1DM)

Insulin is an effective inducer of insulin resistance. Working on the basis that insulin induced insulin resistance is triggered by "excessive" (ie physiologically appropriate to limit nutrient ingress under high calorie availability conditions) levels of superoxide generation then blockade of mtG3Pdh will reduce this "excessive" level of superoxide so defer the onset of insulin induced insulin resistance. This will allow on going utilisation of glucose post meal in the Japanese paper and post insulin bolus in the rodent study.

You just have to wonder whether metformin reduces hunger by blunting insulin signalling at adipocytes, so supplying more calories for the brain to sense or by facilitating the action of insulin within the brain so higher levels of insulin derived from eating absolute crap no longer induce CNS insulin resistance. Maybe both.

Oh, and if you facilitate glucose ingress in to cells when fatty acids are providing an already reduced electron transport chain you will clearly divert pyruvate to lactate rather than having it enter the mitochondria. Glycolysis without pyruvate oxidation gives lactate generation. Just like metformin does... and without needing complex I blockade concentrations. As the Japanese paper commented:

"Post-prandial plasma lactate concentration was significantly increased after the metformin treatment in both healthy subjects and diabetic patients".

Note that the effect was only present under high insulin levels post prandially when the normal physiological response is to shut down excess calorie ingress by inducing insulin resistance. Prevent this response and the calories enter cells as un-needed glucose and exit as "waste" lactate, minimising ATP generation.


Saturday, November 04, 2017

Metformin (04) Pre and Post Prandial

The next metformin paper to look at is this one:

Beneficial effects of metformin on energy metabolism and visceral fat volume through a possible mechanism of fatty acid oxidation in human subjects and rats

Here are the RQ data from 16 healthy humans after an overnight fast and for the three hours following a mixed carbohydrate/fat meal tolerance test (type of carbohydrate and fat not specified).

Aside: Here is the test "food" description: "meal tolerance tests (592 kcal, 75g of carbohydrate, 28.5g of fat; Saraya Co., Osaka, Japan)". It's great to know that there is a company called Saraya and that they have headquarters in Osaka. But I can't even find out what sort of "meals" Saraya make. Quite how anyone might replicate this study using the methods section is beyond me. In addition to these omissions the test "meal" is repeatedly described as "cookies". Go figure. Still, let's assume the measurements of RQ is numerically accurate, fingers crossed. End aside.

These healthy people, who haven't eaten overnight, have an RQ of 0.8 and the test meal produced a downward trend in RQ indicating that the "cookies", providing roughly 50% of calories as fat, tended to increase fatty acid oxidation or decrease carbohydrate oxidation. I can't be arsed to criticise their stats methods. Let's stick with the gross changes.

After two weeks on metformin at an eventual dose rate of 500mg three times daily there is a significant fall in fasting RQ indicating an increase in non-fed fat oxidation compared to the control state.

Under metformin the "cookies" produce a rising RQ, suggesting preferential metabolism of glucose in the immediate post prandial period.

So metformin promotes fat oxidation during fasting but promotes glucose oxidation during the first three hours after a plate of "cookies".


We should see if we can explain these effects on RQ in terms of mitochondrial glycerol-3-phosphate dehydrogenase (mtG3Pdh), electron transporting flavoprotein dehydrogenase (ETFdh) and the redox state of the CoQ couple driving reverse electron transport (RET) through complex I.


Succinate doesn't drive reverse electron transport. Maybe.

Mike Eades sent me this paper:

Reactive oxygen species are generated by the respiratory complex II – evidence for lack of contribution of the reverse electron flow in 
complex I

suggesting that RET through complex I, when driven by succinate oxidation at complex II, is a pure artefact of the pathologically high level of succinate used in the mitochondrial preparations involved. Bearing in mind that trying to work out exactly what the physiological concentration of succinate might be, in the region of the active site of a complex II in a working, oscillating, in-situ mitochondrion, involves an awful lot of guesswork.

However, the paper might well to be correct, within the limitations of the mitochondrial preparations they are using.

If you feed mitochondria with 5.0mmol/l succinate there is profuse ROS generation, 85% of which can be blocked by rotenone, ie this 85% is RET generated. The other 15% comes from other places, including complexes II and III, at least. But if you feed mitochondria with 0.5mmol/l succinate, or even 1.0mmol/l, there is no ROS generation at all. The case is made that ROS from RET are not a feature of "normal" levels of succinate driving the reduction of the CoQ couple.


But this is a mitochondrial preparation. It has no cytoplasm, no glycolytic enzymes, no source of glycerol-3-phosphate, no FFAs, no carnitine. You can't buy a vial of FADH2 bound to electron transferring flavoprotein to feed in at ETFdh. This makes manipulating the CoQ couple in a way which is physiologically significant very difficult. In the current study we have no input to the CoQ couple other than complex II using succinate.

Those folks like myself, who feel that the redox state of the CoQ couple is the main sensor of the energy status of the cell, would never expect a single input in to the CoQ couple to be the sole representative of energy status. Even during glycolysis there is some fatty acid oxidation providing electron transferring flavoprotein to ETFdh. And succinate from FFA derived acetyl-CoA will also supply to complex II during lipid oxidation. And conversely some glycolysis will occur, even when FFA oxidation predominates, supplying glycerol-3-phosphate to mtG3Pdh.

Until we can set preparations up in which these inputs can be adjusted we are not able to say much about what might be happening in-vivo to RET. And once you start smashing the mitochondria to pieces and reassembling them as inside-out vesicles (so you can supply metabolites to the intra-mitochondria binding sites that would normally be hidden away from your extra-mitochondrial culture fluid) you are a very, very long way from in-vivo indeed.

Just saying...


Metformin (03) In-vivo experiments require non-lethal dose rates!

Just before I move on to metformin-induced substrate oxidation changes in healthy volunteers, I think it's worth looking at this neoplasia paper in a little detail. It's fairly typical of the work done on metformin as an anti-cancer agent and focuses on the highly reproducible inhibitory effect of metformin on complex I.

Metformin inhibits mitochondrial complex I of cancer cells to reduce tumorigenesis.

Most of this work is very clever and very carefully done, but lives with the problem that the experiments usually use concentrations of metformin in-vitro which would be lethal in-vivo because, well, everybody does it and there is no effect if you don't... However the mouse xenograft studies have to use clinically relevant therapeutic doses of metformin otherwise the mice would be, well, a bit dead. There are other problems which will become apparent as we work through the data.

The figure I'd like to focus on is supplementary data section three of figure seven.

Graphs B and C look like this:

This is what they did to generate them. They took A549 tumour cells and injected them in to immuno-incompetent mice then measured the growth of the resulting tumour.  A549 cells are highly sensitive to metformin, so graph B comes as no surprise. Graph C is much, much cleverer. They wanted to prove that metformin was actually working on complex I. So they destroyed complex I with a shRNA targeting NDUSF3, an essential subunit of this complex. To keep the cell line functional they replaced complex I with our old friend the yeast derived NADH dehydrogenase NDI1. This enzyme does not bind metformin nor pump protons but does reduce NADH to NAD+ and does feed electrons to the CoQ couple and the downstream complexes. You can see from graph C that replacing complex I with NDI1 protects the A549 cell derived tumours from the growth slowing effects of metformin.

Look at B. Look at C. Protection from metformin in C. Yes?

Now, you have to ask: What is the effect of knocking down complex I in cancer cells? If you cannot reduce NADH to NAD+ then the TCA cannot turn. Citrate cannot be metabolised to alpha ketoglutarate so is exported from the mitochondria and can be used for tumour anabolism. The tumour becomes highly aggressive. Like this:

Down-Regulation of NDUFB9 Promotes Breast Cancer Cell Proliferation, Metastasis by Mediating Mitochondrial Metabolism

or this, blogged about many years ago:

Mitochondrial complex I activity and NAD+/NADH balance regulate breast cancer progression

This illustrates my marked discomfort with accepting complex I blockade as the mechanism of anti-cancer action of metformin. Blockading complex I will admittedly decrease ATP supply from oxidative phosphorylation but at the cost of supplying a large amount of citrate to the cytoplasm ready for anabolic processes, while glycolysis continues unabated, supplying cytoplasmic NADH and ATP.

So in the current paper, by knocking down NDUSF3, they should have generated an aggressive phenotype. They didn't, because they also engineered-in NDI1, which will reduce cytoplasmic NADH to NAD+ very effectively. Dropping the NADH to NAD+ ratio suppresses tumour aggressiveness in the above papers.

Does the engineered A549 NDUSF3 + NDI1 tumour in nude mice show reduced or increased aggressiveness compared to the A549 unmodified tumour? We are looking to compare the top line in graph B above (dark squares) with the pale squares in graph C. By eyeball they actually look pretty much the same.

Except for the x axes. Graph B is 40 weeks, graph C is 50 weeks. Hard to compare the two... But if we stretch graph C so that weeks 10-40 align with weeks 10-40 of graph B, then superimpose the two graphs we can generate the following, rather more informative, image:

It looks to me as if inserting NDI1 in to the mitochondria of a cell line, (probably) made aggressive by knockdown of NDUSF3, renders the in-vivo tumour growth rate much lower than the natural tumour cell line and remarkably similar to that of metformin treated natural tumour cell line. Probably by reducing the NADH:NAD+ ratio.

This doesn't automatically suggest that metformin might be acting by reducing the NADH:NAD+ ratio, though it might be, but it does illustrate how nicely you can still pull interesting snippets out of papers full of experiments with metformin at lethal concentrations.

The difference between isolated mitochondrial preparations and mouse models is that the mouse models have a supply of insulin, glycerol-3-phosphate and the enzyme to use cytoplasmic NADH to reduce the CoQ couple, facilitating insulin signalling and so cancer growth. This is much more likely to be the process which we can block with metformin at therapeutic concentrations.


Wednesday, August 02, 2017

Metformin (02) The dose makes the poison

Before the days of interest in metformin as an anti-neoplastic agent, a performance enhancing drug or a longevity promoter, it was just given to T2DM patients to help lower blood glucose levels. These folks, as a group, quite often have significant renal disease. Which can render metformin and lactate cumulative in the blood stream and lead to a life threatening lactic acidosis.

This paper looked at a series of 10 hapless folk to whom this happened:

Metformin overdose causes platelet mitochondrial dysfunction in humans

The mean blood concentration which gets you an ITU bed was 32mg/l. Now this is a clinical paper, written by clinicians. Nothing wrong with that, except they use Noddy units which makes the metformin concentrations extremely difficult to relate to the vast body of metformin research, which uses units of millimolar or micromolar.

So we really need to take this image

and think of it in these terms when we're looking at research papers using mmol or micromol concentrations:

Bear in mind that these are very chronic exposure values and metformin is thought to be progressively cumulative within the mitochondria on chronic exposure. Of course, complex I is intra mitochondrial and there will be some dependency on cumulation in getting significant effects at this site. What we can say is that, in the above diagram, there is not enough inhibition of complex I to raise lactate production in platelets, an extra-hepatic tissue (hepatocytes may be slightly different), unless we are using near-death concentrations.

What is not hidden away inside the mitochondrial matrix is mtG3Pdh. It's on the outer surface of the inner mitochondrial membrane and will be exposed to whatever metformin concentration that manages to get inside the cell.

From the classic paper

Metformin suppresses gluconeogenesis by inhibiting mitochondrial glycerophosphate dehydrogenase

we have this graph from Figure 3, using a slurry of mashed up mitochondria and some glycerol phosphate:

Here we have a significant effect on the oxidation of glycerol-3-phosphate at micromolar concentrations. Admittedly by 50μmol we are looking at very much the upper end of therapeutic concentrations but an effect is clearly visible at this level. We can say from the platelet paper that exposure to 250μmol (black circles at the bottom of the graph), if sustained, will put you in the ITU with potentially fatal lactic acidosis.

Because mtG3Pdh is exposed to cytoplasmic (non cumulative vs mitochondrial) metformin levels it will see the drug at plasma concentrations (or slightly less) and it will see these concentrations as soon as metformin enters the blood stream.

If you want a performance enhancing drug for endurance exercise, say a cycle race taking about three hours, you can pop a single metformin 500mg tablet one hour before the start of the race and extend your time to exhaustion from 167 minutes to 191 minutes. That might make some difference to winning vs not finishing.

Metformin improves performance in high-intensity exercise, but not anaerobic capacity in healthy male subjects

Equally, there is no acute effect on lactate levels in the same study. This is no surprise as I find it difficult to envisage acute complex I blockade, to lactate generating levels, as a performance enhancing ploy.

TLDR: metformin probably works in the cytolasm on mtG3Pdh. Rising lactate may well indicate mitochondrial cumulation and some degree of complex I inhibition. Extrapolating benefits from studies based around millimolar concentrations in-vitro may well put you in to the ITU if you try them in-vivo.


Tuesday, July 25, 2017

Metformin (01) Insulin

This image is taken from the paper Insulin requirement for the antihyperglycaemic effect of metformin and it deserves a little consideration.

They are using BB/S rats which spontaneously develop T1DM if fed standard rodent chow. In the absence of exogenous insulin they die but giving them a little Ultratard twice daily keeps them alive for quite some time. Stopping the Ultratard allows exogenous insulin withdrawal to produce an acute, alive, an-insulinaemic rodent model. This is the model used and at the start of this experiment the rats had no detectable insulin in their blood.

At time point -60 these an-insulinaemic rats were given metformin intrajejunally. Over the next 60 minutes the metformin did nothing to lower plasma glucose. At time point zero they were given a small intravenous bolus of glucose. Metformin had no effect on the additional hyperglycaemia induced.

At time point +90 they were given neutral insulin intravenously. In the control group plasma glucose concentration dropped to a nadir of 20mmol/l at time point +150 but in the metformin treated rats the same dose of insulin continued to reduce the plasma glucose to 10mmol/l at time point +180, when p dropped below 0.05.


Insulin is essential to demonstrate any effect of metformin on blood glucose.

Any idea about how metformin works, be that via the inhibition of mtG3Pdh or via inhibition of complex I, has to accommodate the essentiality of insulin.

That's an interesting constraint.