Satiety (06) The MCAT mice

Tucker emailed me this paper while I was writing the last post:

In adipose tissue, increased mitochondrial emission of reactive oxygen species is important for short-term high-fat diet-induced insulin resistance in mice

I was particularly excited by it because they report actual H2O2 production in pmol min-1 [mg wet weight]-1. Or, in olden-speak, pmol/min/mg tissue. Shrug.

This section of Figure 5 superficially appears to confirm all of my biases. Respiration supported by lipid oxidation generates more ROS from D12942 fed mice than chow fed mice.

Except that's not a question I would have asked. These are cells in very specific cell culture conditions and the white and black bars are both from cells being "fed" on exactly the same concentration of palmitoyl-CoA. Theoretically the ROS production should be the same, given the same fatty acid. Plenty from palmitate, a little less from linoleate.

But here the question they asked is about cells extracted from a mouse exposed to D1294 for a week versus those from a mouse exposed to chow for a week and now both are currently being exposed to an identical level of palmitate.

It's very probable that the D12942 derived cells have more mtETFdh complexes than chow fed and that these complexes generate more ROS than cells from mice fed chow. We'll never know because they measured levels of complexes I, II, III and V, but not the complex delivering the primary lipid-derived FADH2 input to the ETC, mtETFdh. Go figure.

However.

When you read the methods section you realise they are not looking at physiology at all. These are not cells fed on simple substrates and then the ROS production is being measured. They are cells whose ETC chain is fully inhibited at the ATP synthase level by oligomycin at 6.7μg/ml. That is a solid inhibitory level. So, essentially, they are "pressurising" the ETC to a high membrane potential using succinate (10mM, also not a low level) and then looking at the ROS leakage under these conditions as the test substrate is applied, in addition to the succinate. The mtETFdh complex is famous for ROS generation under high membrane potential in far-from-physiological preparations.

This is interesting but has absolutely nothing to do with physiology and certainly gives no information about fatty acid mediated ROS generation at low physiological mitochondrial membrane potential. The whole point of ROS generation from fatty acid oxidation is that it does NOT require an elevated membrane potential.

Aside: Red flag warning. They also looked at ROS generation under state 3 respiration conditions, in which the cell preparation is flooded with ADP, which allows ATP synthase to dissipate membrane potential while forming ATP, so reduces ROS generation, which it did. Here outlined in red from section C of Figure 5:

But you can't dissipate membrane potential through complex V using ADP when that complex V is 100% blockaded by 6.7μg/ml of oligomycin. ADP only consumes membrane potential when ATP synthase is active. Flooding a cell preparation with a generous supply of ADP when ATP synthase is inhibited by oligomycin *should* generate zero uncoupling, so there will still be an high membrane potential and high ROS production. Yet ADP clearly reduces ROS generation in this model. I cannot see why this should be the case. It shouldn't.

I think the word is "dodgy". End aside.

So I'm not exactly overwhelmed by their permeabilised cell ROS generation model. It tells us nothing about in-vivo physiology.

The MCAT part of the study is even worse, in it's own way. These are real live mice which have been engineered to produce human catalase in their mitochondrial matrix (in addition to their own catalase) and they do show a marked preservation of insulin sensitivity when fed on D12492.

This is reported as a Good Thing.

They are tinkering with the second core physiological messenger, H2O2, (the first messenger is superoxide and this should be unaltered by catalase) within the mitochondrial matrix, and you would expect this to cause detectable physiological (or pathological) changes.

On chow it doesn't. They used a 10% of energy from fat chow and, not surprisingly, the reduction in mitochondrial matrix H2O2 had no measurable effect on insulin sensitivity. There is no need for grossly detectable physiological insulin resistance when only 10% of dietary calories are coming from fat. Insulin signalling preservation is an essential net benefit here.

The knock-in mice demonstrated some effects in some adipose depots under D12942 feeding, rendering them apparently less resistant to insulin's ability to phosphorylate AKT than they should be, ie more like chow fed mice. I have a 10 slide PowerPoint deck dissecting the problems with their data presentation, which I will spare you. But ultimately the effect is much smaller than presented but it is still present. Mostly.

What we actually want to know is whether the decrease of the normal ROS signal from FAO in the MCAT knock-in mice makes them behave as if they have an inadequate ROS signal, as mice fed 18% linoleate do. Do they get *extra* obese on D12942?

We don't know.

MCAT mice on low fat chow don't have any change to their insulin sensitivity compared to wild type. They have a normal, very carbohydrate based metabolism which requires minimal insulin resistance at low mitochondrial membrane potential.

Wild type mice on a 60% fat diet should, excepting the linoleate effect (which is initially huge), resist insulin and stay slim, they should be soourcing 60% of their calories from fatty acid oxidation and limiting the rump of their calorie generation from the small amounts of carbohydrate and protein in D12942. MCAT mice, like linoleate fed mice, shouldn't.

It would be nice if MCAT mice on D12942 became "hyperphagic" as wild type mice do for a few days, but then kept it up for the full week and beyond, until their adipocytes simply became so full that they were damaged and releasing those inflammatory mediators that long term D12942 fed WT mouse adipocytes eventually produce. You might even pick this up in weight gain changes at one week, the MCAT mice on D12942 should be fatter than the wild type mice on D12942.

I guess the alternative is that the overlay of PKCε and Thr¹¹⁵⁰ might kick in to rescue matters, which would limit the excessive insulin sensitivity of the MCAT mice. And would be boring. Hard to say, but of PKCε is under redox control then then this route to salvation won't happen.

Sadly we'll never know about the weight gains. Despite the mice being weighed at the one week mark their weights are not reported.

As far as I can find no one else has fed MCAT mice on D12942 or its equivalent. They do very well long term on an unspecified chow diet but that is both expected and very, very different to feeding D12942.

Overall the paper turned out to be hugely disappointing. C’est la vie.

Peter

Satiety (05) Threonine/alanine and the fasting insulin resistance

Shulman had a "That's interesting" moment in his 2016 paper which unfortunately got filed under "everything else was as we expected", and placed on the penultimate page of the supplementary data.

Here's the moment:

"Basal and clamp plasma insulin levels were subtly increased in [chow fed] InsrT1150A mice (Supplemental Table 1)."

Terminology note and clarifications: InsrT1150A are the Thr¹¹⁵⁰A mutants discussed the previous posts, ie mice with an un-phosphorylatable alanine in the place of the Thr¹¹⁵⁰ of their insulin receptor. The during-clamp high insulin looks like a non related phenomenon, more a feature of altered insulin catabolism under clamp conditions. Which might be interesting, but not for today's discussion. Oh, and this didn't show up after a six hour fast (Supplemental Table 2) because there was still a glucose based metabolism at that time. I think the anomaly will be intrinsic to markedly fat based metabolism.

Here's Supplemental Table 1

The change in fasting insulin is in the wrong direction and it's the only statistically significant change, excepting the clamp insulin level.

Shulman is perfectly entitled to use the term "subtly" because the difference between 4.5μU/ml and 6.2μU/ml is unlikely to be of any physiological significance. It's like feeling uranium minerals are slightly warm to the touch, noting it and ignoring it. Whereas the actual follow on from warm uranium was Hiroshima.

These mice which have been engineered to invariably fail to phosphorylate the Thr¹¹⁵⁰ location on their insulin receptor are more insulin resistant, not less.

So, when assessed using a genuine fasting insulin level, these mice *do* resist insulin more than control mice do, despite their absolute lack of a phosphorylatable Thr¹¹⁵⁰.

Which brings us back to ROS.

I hope everyone recalls this paper:

Electron Transport Chain-dependent and -independent Mechanisms of Mitochondrial H2O2 Emission during Long-chain Fatty Acid Oxidation

discussed here, featuring this graph

Genuine fasting metabolism is based on fatty acids. These, irrespective of the mitochondrial membrane potential, will generate ROS and mediate redox dependent insulin resistance. No cellular caloric overload needed, fatty acid oxidation simply resists insulin signalling.

A layer on top of this is the enzymic phosphorylation of Thr¹¹⁵⁰ which will reduce insulin facilitated glucose ingress, and oxidation, in addition to the ROS signal, so you need less of an ROS signal.

The ROS signal is still there, you can't oxidise fatty acids without generating ROS. I think you can reduce the need for ROS using the supplementary Thr¹¹⁵⁰ system to also resist insulin. So, for a given level of ROS, the enzymic mechanism enhances glucose restriction and keeps tighter control over the inner mitochondrial membrane potential. Obviously if delta psi is high ROS are generated by any substrate oxidation at similar levels to those from fatty acid oxidation, as in the above graph.

So the ROS insulin signal is higher in the Thr¹¹⁵⁰/A mice which gives a mild increase in fasting insulin resistance. The ROS signal is higher either because there is more lipid oxidation occurring or (more likley) some glucose oxidation is occurring when the tissues are already energy replete from FAO. This will raise delta psi and increase ROS production from this source in addition to FAO, providing extra redox mediated insulin resistance.

Obviously once insulin acts on adipocytes to suppress FFA release during a clamp then the FFAs fall, so does the ROS signal and all behaves as normal and Shulman is happy.

That is where I was mentally sitting while thinking about the Thr¹¹⁵⁰/A mice before Tucker emailed me this paper, explicitly invoking ROS mediated insulin resistance:

In adipose tissue, increased mitochondrial emission of reactive oxygen species is important for short-term high-fat diet-induced insulin resistance in mice

Which is rather nice. The ROS are real.

Sadly the paper a bit like the curate's egg, good in places. I think it's worth a post in its own right.

Peter

Satiety (04) D12942 and insulin resistance(s)

These are just some of the illustrations I doodled out for the last post while thinking about insulin sensitivity/resistance in D12942 fed mice. I hope they make it clearer what Shulman was looking at in 2016 and what he moved on to look at in 2021, comparing D12942 feeding to control mice versus to his mouse model with the Thr¹¹⁵⁰ to alanine (Thr¹¹⁵⁰A) switch.

I've left various possible routes for the development of insulin resistance dashed for D12942 because that still needs a significant amount of discussion, see below. I've also added in yellow a line for the state of phosphorylation of the Thr¹¹⁵⁰A substituted mouse mutant. What the 30 minute clamp is looking at is the residual physiological insulin resistance at a time when hunger on D12942 has almost normalised and the level of phosphorylation of Thr¹¹⁵⁰ is approaching what it should be if D12942 was a physiological high fat diet.

This clamp induced change will be a generic pattern, here are all of the scenarios I imagined two posts ago, with the Thr¹¹⁵⁰A mutant added, just to reiterate the colour scheme:

and here's what I expect happens when you measure insulin sensitivity by glucose infusion rate during a mild insulin clamp at 30 minutes:

If I am correct it suggests that all of these cells are physiologically normal under normal clamp conditions. This is not "pathological" insulin resistance.

The next question is why cells under all of these conditions have differing levels of insulin resistance before the clamp started.

The obvious explanation is that the phosphorylation of Thr¹¹⁵⁰ is a surrogate for the activity of PKCε, which is a surrogate for cell membrane DAGs, which are a surrogate for free fatty acid availability. High fat diets provide high fatty acid availability, so high levels of phosphorylation of Thr¹¹⁵⁰. If you are oxidising fat it is essential that you reduce the oxidation of glucose by an appropriate amount.

The phosphorylation of Thr¹¹⁵⁰ is an enzymic mechanism of inducing insulin resistance, in insulin sensitive cells, in proportion to the availability of free fatty acids.

This is completely separate from the redox induced insulin resistance posited by the Protons hypothesis.

It has its uses. It has its limitations.

The first use is that it allows fatty acid induced insulin resistance without the need for those redox changes intrinsic in fatty acid oxidation. If you want a safety mechanism to limit ROS generation to tightly controlled levels this could be one method of achieving it. A fine tuning mechanism.

The second is as a safety net for when the redox signal from fatty acid oxidation fails, as it does under the beta oxidation of linoleic acid.

This latter is what I want to look at in more detail today.

When we change a mouse from chow to D12942 there is a sudden switch from a metabolic substrate mixture with a normal redox balance to one with reduced ability to generate ROS. Because PUFA are preferentially oxidised, the effect should occur rapidly.

The insulin secreted in response to a meal acts *too* effectively and activates both the pyruvate dehydrogenase complex and the acetyl-CoA carboxylase complex to allow the generation of malonyl-CoA in the cytoplasm, which inhibits fatty acid entry in to the mitochondria. This marked fall in FAO results in a marked fall in ROS generation and reduces ROS derived insulin resistance still further.

This is very straight forward from the ROS perspective. Insulin over-acts, fatty acid oxidation plummets, there is an hypocaloric state which cannot be corrected by fatty acid oxidation and so carbohydrate oxidation predominates. D12942 only provides 20% of calories as carbohydrate so the mice need to eat an awful lot of it to stay alive. Here's the acute ROS mediated pathological insulin sensitivity and associated hunger. The unusable fat is mostly (but not completely) dumped in to adipocytes by the same insulin sensitivity.

Over the following days the free fatty acid levels in plasma rise because they are not being oxidised (see Astrup 1998, discussed here) and this allows the progressive accumulation of DAGs in the plasma membrane where they activate PKCε, which phosphorylates Thr¹¹⁵⁰ *irrespective* of the levels of ROS being generated within a given cell. If D12942 behaved like a physiological high fat diet we should get this, just from phosphorylation of Thr¹¹⁵⁰:

But failed fatty acid oxidation derived insulin resistance has dropped the baseline from which Thr¹¹⁵⁰  insulin resistance has to start. Combining the immediate intrinsic ROS signal change with the slower rise in FFA/DAG mediated insulin resistance gives this pattern:

As Thr¹¹⁵⁰ mediated insulin resistance rises, irrespective of the failed ROS signal, it progressively limits the pathological *excess* insulin signalling and allows progressively more fatty acid oxidation to become possible. The hunger of the first day gradually reduces until by day seven, while hunger is still slightly above that of chow fed mice, it is no longer statistically significantly so.

The next situation we need to look at is the mouse with the Thr¹¹⁵⁰A engineered mutation when fed D12942 but I've written enough for today. Oh, and I think we also need to consider the situation once *genuine* pathological insulin resistance is established in control and Thr¹¹⁵⁰A mice.

Peter