Sunday, September 15, 2019

The paradoxical fat mice (1)

This paper is very interesting. I think I picked it up via Raphi on twitter. It comes from Jim Johnson's lab.

Caloric Restriction Paradoxically Increases Adiposity in Mice With Genetically Reduced Insulin

The background is in these two papers:

Phenotypic alterations in insulin-deficient mutant mice

Compensatory Responses in Mice Carrying a Null Mutation for Ins1 or Ins2

The paradoxical mice all had the Ins2 gene fully knocked out and in addition to this some mice also had one allele for the Ins1 gene knocked out (Ins1+/-). So the mice in the study had either a half or a quarter of the normal mouse insulin gene complement. Some mice were fed ad-lib, some were 40% calorie restricted (CR).

The CR, lowest insulin gene group (Ins1+/-) had significantly elevated total fat mass and a significantly elevated percentage of bodyweight as fat. That's a paradox to the insulin hypothesis of obesity and so really interesting. The Ins1+/+ group also had a (ns) increase in percentage body fat but not in absolute fat mass, so the trend is there too, but only a trend.

Metabolically, the split is between ad-lib vs CR groups.

All mice had the same maximal insulin response to a 2g/kg intraperitoneal glucose tolerance test but the CR groups had a very significantly reduced peak and AUC for glucose, ie they were much more insulin sensitive. The intra-peritoneal insulin tolerance test result might be worth a post in its own right, it's paradoxical too but there won't be space to cover it today.

So let's have a look at energy expenditure (EE) from Fig3 C.

To make things a bit clearer I've copy pasted the light period from the left half of the graph on to the end of the dark period to give more of an idea of the EE curves are really like during dark to light transition:

The red line starts horizontally with no significant difference in EE between ad-lib fed mice or CR mice. There is a modest increase during the dark (active) period when the CR mice get their three very small meals, as indicated. After the last meal a precipitous and highly significant fall in EE occurs. The mice enter torpor, a state of extreme lassitude and hypothermia. At around two hours in to the next light period the mice wake up and EE returns to just below that of the ad-lib mice and the cycle repeats. The ad-lib mice behave like normal mice.

The CR mice have a profound hypometabolic period every day. You could argue, if you are a cico-tard, that this is why they store excess fat. They eat all the food they can get but expend relatively little energy so they become fat: CICO. But I would disagree.

Here's my guess as to what is happening. Speculation warning.

We know that the CR mice are exquisitely insulin sensitive. They are that way because they have a low number of insulin genes and they never get enough food to trigger a major insulin spike anyway. The CR is the dominant factor but it needs the genetic background to get the paradox to occur. Insulin-induced insulin resistance, acute or chronic, does not occur due to lifetime low insulin exposure. The fact that all mice are capable of producing the same maximal insulin response to an IPGTT does not mean that the CR group experience an equivalent insulin exposure to the ad-lib group during their routine lives. They never get enough food to trigger a maximal insulin response.

The CR mice spent the bulk of the light period with a slightly low EE. Dark period arrives and with it food. As the food is eaten there is an upward trend in EE followed by a drop. The second small meal arrives, again an upswing followed by a drop. The third and final meal gives the same upswing but the drop in EE which follows just goes on downward. The mice enter torpor, a state of profound lassitude and hypothermia.

I think torpor happens because the mice simply have no accessible calories.

This is despite the fact that it occurs immediately after the third of their calorie restricted meals. Their problem is that the meals generate an insulin response. The mice are so insulin sensitive that calories are lost in to adipocytes (and probably hepatocytes) under the over-effective action of insulin.

They lose calories in to adipocytes. These are calories out. The adipocytes get bigger with the lost fat.

Torpor occurs BECAUSE the mice have become fatter.

This is the equivalent of the hunger which follows for a human under a euglycaemic (or even hyperglycaemic) hyperinsulinaemic clamp. There is no hypoglycaemia but fatty acids become locked in to adipocytes by the hyperisulinaemia and hunger follows due a lack of available calories. I posted about it here.

At two hours in to the light period insulin drops low enough to allow lipolysis. The mice wake up.

That's all.

Except: Why do the CR mice have paradoxically (although ns) elevated fasting insulin cf the ad lib mice? There are two reasons. Here are the blood sample times added to the EE graph. The arrows are not quite in the correct clock times, as detailed in the methods, but the times related to feeding/metabolism are approximately correct.

The green arrow is the sampling time for the ad-lib fed mice. It is about six hours in to the light period and the mice would normally have been asleep during the hours leading up to it. Light-period snacking, from the respiratory exchange ratio (RER) graph in Fig3 D, would not normally have started by this time so it's a very simple physiological fasting sample.

The blue arrow for the fasting CR mice just hits the end of torpor. I'm not sure these mice ever have a time when they wouldn't eat, given the chance, but here they are in their hypometabolic phase and have minimal access to calories. At this time insulin is actually a little (ns) higher than for the ad-lib groups (Fig1 D). Higher insulin means fat stays in adipocytes. Why is insulin high?

Calorie restriction does many things in addition to dropping metabolic rate. If you fast a hard working group of humans for 5 days they develop a post prandial increase in GIP (glucose-dependent insulinotropic hormone). This was found in the CR mice in both the fasting and fed state (Fig4 C). GIP facilitates insulin release, hence insulin is a little higher the CR mice and loss of calories in to adipocytes more severe, necessitating torpor.

It's interesting as to why GIP might be elevated under hunger conditions. Possibly generating and saving fat becomes a priority when calories are low. The Ins1+/- CR mice certainly have the highest RER (>1.05) after their third meal, suggesting that they prioritise the conversion of glucose to fat. This DNL, should it occur in the liver, might go some way to explaining the elevated triglycerides in both of the CR groups. Maybe. Accentuated DNL in people who have undergone massive weight loss via gastric bypass surgery is routine during an OGTT. Like these people.

Anyway, I'll stop now. This post is about 1/4 the length it started out as, so if corners seem a bit cut then mea culpa.



Alex said...


Do I get you right? Your theory goes like this: the difference between insulin states determines the effect of insulin on adipocytes. So, the threshold for "effectivity" is relative to baseline? Large delta, large effect (all exponential of course)? The one that eats mostly carbs will not suffer grave consequences when increasing insulin temporarily, but the keto-one will put on weight just by looking at ice-cream?

How can Ins2-/- mice cope with this mutation? Is Ins2 not the mouse's brain variant of insulin? Does Ins1 compensate fully given CR? To what extent?

It is weird that the mice show lethargy/torpor since CR mice without mutations do actually the opposite: they are a wee bit more active than ad-lib. Fig3F: CR mice move less (they point this out) That is strange.

The higher insulin might, very carefully, suggest that physiological insulin resistance developed. Are the adipocytes greedy or just opportunistic?

What if physIR coupled with insulin deficiency screws up the brain? Can the mice mount the appropriate low-energy stress response? It does not seem so.

Best wishes!

altavista said...

Doesn't explain why my 40% corn oil mice have 2x insulin and 5x leptin. Not a unified theory yet mwahaha

Peter said...

Hi Alex,

I believe both of the Ins genes in rats/mice are probably whole animal expressed, certainly both are in the pancreas and it is in the pancreas that the visible adaptation to knockout occurs. This gives the idea (not read all of it). The links at the start of the post are to the folks who developed the model.

Exactly. Normal CR mice are hungry all the time and look for food. They will almost certainly have physiological insulin resistance to maintain normoglycemia in the presence of restricted food combined with extended periods of no food. Their ability to keep calories out of adipocytes and in the blood stream is in-tact so they remain functional.

The post-torpor mice still have mildly elevated insulin at the 12 hour mark, though it’s lower than at the 4h mark. All ns of course. Hows and whys beyond GIP are an open question… You might get some answers by looking at insulin signalling rather than insulin levels of course but that would need a whole lot more mice.

alta, unified theory: the holy grail….


Alex said...

Hello Peter,
I was wondering because insulin is necessary for the proper functioning of brain cells. Not for glucose uptake but excitability, regulation of appetite and fuel distribution and others. I was speculating whether these mutations still guarantee sufficient insulin for the brain during CR. These mice do not behave in a normal way. They are, as you implied, energy-deprived.

They did not measure ketones. A pity.

Anyway, I am looking forward to your next topic.


Passthecream said...
This comment has been removed by the author.
Peter said...


Have you looked at the adipocytes?


Passthecream said...

Hi Peter

I apologise in advance for my inability to hold all the different pieces of this puzzle in my head, at the same time!

The other 60% of AV's rat diet above would surely be key? For instance, if it was a significant % of sucrose.

(Corn oil btw is approx 13% saturated fats and 27% mufa. )

You would have this, quoting Protons(38)

'' we also ought to think of the situation under a large, uncontrolled fructose input through mtG3Pdh occurring at the same time as saturated fatty acids are being oxidised. That gives us this scenario:
... ... squiggle ... ...
Having two inputs reducing the CoQ couple (as well as a little input from SDH) is a perfect recipe for driving extreme reverse electron transport through complex I with the production of completely unreasonable quantities of superoxide and H2O2. This is the scenario of free radical mediated damage combined with serious insulin resistance. D12079B anyone? The problems are less severe with PUFA fats ... ... "

I am supposing that this would also give rise to significant insulin production in beta cells. Then there will be increased beta oxidation from the insulin resistance lowering the inhibitory effect of insulin on hsl and so on.

Can it be this simple or am I missing something?

karl said...


The diet they ate is at


This demonstrates what I've been speculating about - insulin sensitivity is what matters more than insulin levels.

I've also been thinking more about an adipocyte with a 2019 sized load of PUFA - the cell itself has to burn a bit of fat to power itself - would the change in insulin sensitivity effect LPL and HSL? (I think so?) - could there be some genetic variance in this effect?

LPL moves fat in to and HSL moves fat out of fat cells.
insulin is known to activate LPL in adipocytes
HSL is inhibited by insulin.


Another thought - I have a cat that has eaten little else than industrial chicken - full of LA - he is close to 10 years old - and not fat. My hunch is it is because he doesn't eat any carbs. So I'm speculating that it is the combination of a PUFA diet and the carbs that is fueling the T2D/obesity pandemic.

Peter said...

Pass, yes I think you are saying that it is the same drivers which cause both insulin secretion and resistance. The two fit together and control the level of insulin and so control bodyweight. But the cost can be of elevated insulin and not all of insulin's signalling can be resisted. Also I think adipocytes, as eluded to by karl, will be the last cells in the body to resist insulin, especially if oxidising PUFA... NB adipocytes are utterly packed with mitochondria. The lipid droplet is not "in" the cytoplasm per se. The cytoplasm is full of mitochondria.


Passthecream said...

Peter: " NB adipocytes are utterly packed with mitochondria. The lipid droplet is not "in" the cytoplasm per se. The cytoplasm is full of mitochondria."

That is interesting, it paints a picture of adipocytes as 'dark satanic mills' --- big factories with lots of fires burning in the dark, smoky (ketones?) and dangerous oxidising chemicals stored in close proximity to flammable substances.

But seriously, I should have been more clear that I was commenting on your reply to AltaVistas 40% corn oil observation ratger than the main topic. Mea Culpa.

So what you say there about high insulin being a problematic result is in accord with what AltaVista is observing ( given whatever the other 60% of diet possibly is.) I'm a bit slow to catch on sometimes (obviously!) but my excuse is that this is a strange type of mental arithmetic. My understanding of the concept of pufa being unable to generate much insulin resistance is that, on their own, they provide more caloric input than any resistance they can generate will be able to manage ie inadequate negative feedback

However, is it reasonable to think that when there is a lot of pufa being burned, pushing nadh without so much fadh2, but that is combined with a quantity of electrons inserted by eg mtg3pdh then there should be a suitably high level of qh2 in the pool to generate enough ret -> superoxide to generate adequate insulin resistance?