The Science of Hunger & Medications to Combat Obesity | Dr. Zachary Knight

At the highest level, Knight frames the brain's regulation of food intake as two interacting systems rather than one: a short-term system operating 'on the timescale of a meal, so 10, 20 minutes', and a long-term system operating 'on the timescale of sort of weeks to months to years' that 'tracks levels of body fat'. They are 'primarily localized to different parts of the brain', and they interact so that 'these short-term behaviors we do, eating, are matched to our long-term need for energy'. This is the scaffold for everything that follows in the exchange.

The anchor evidence is an experiment by Harvey Grill 'about 50 years ago', the decerebrate rat. Grill 'made a cut in the rat brains' separating the brainstem, 'the most posterior part of the brain', from the entire forebrain, and 'got rid of 80% of the rat's brain'. The resulting 'zombie rats' could do very little, but 'one thing they can still do is regulate the size of a meal'.

Knight immediately adds a methodological caution: 'we have to be careful how we talk about this', because food has to be placed in the animals' mouths. They swallow, then 'eventually, at some point, they'll start spitting it out' - 'an indication that in some sense, they're becoming sated' using only the remaining brainstem. Gut signals are sufficient: 'gastric stretch, hormones that come from your intestine in response to food intake, like CCK' can be injected or the gut manipulated to change intake 'in an appropriate way'. What the decerebrate rat cannot do is 'respond to longer-term changes in energy need': fast it 'for a couple of days' and 'the amount that it eats doesn't change'.

- Brainstem and most posterior brain: neural circuits controlling a single meal on a 10-to-20-minute timescale, 'deciding when a meal should end'. - Forebrain, 'primarily in the hypothalamus': circuits tracking 'what is my overall level of energy reserves, what is my level of body fat', fluctuating 'on a timescale of, say, days when you're fasting'. - Direction of control: 'those forebrain centers feedback to talk to the brainstem and modulate those brainstem circuits' so the two timescales match.

Knight closes the framework by flagging its limits himself: 'So that's at the highest level how I think about the neural circuitry that controls feeding. There's obviously a lot more going on underneath that.' The two-system picture is explicitly a high-level sketch, not a finished mechanism.

Huberman presses on a choice built into the model: why does the brain track 'body fat' specifically, and not 'body fat and muscular mass or body fat and overall body weight'? Knight's answer is that 'body fat is unique because it represents this energy reserve'. He contrasts it with the thirst system his lab also studies: 'you don't have a reserve of water in your body', so for fat alone 'it's very important that the brain know how much remains and then adjust behavior in accordance with that so that you know how urgent it is to get the next meal'.

The 'major signal of the level of body fat' is leptin, 'a hormone' that 'was cloned in 1994' by 'Jeff Friedman at Rockefeller University' - identified as Knight's own 'postdoctoral advisor' - 'although its history goes back way before 1994'. Functionally it is 'part of this negative feedback loop from the fat to the brain that basically tells you about your level of body fat reserves and how urgent it is to find the next meal'.

- Jackson Labs in Maine, raising and selling mice 'since the 1920s', breeding 'thousands of mice'. - 'at some point in the 1950s', spontaneous mutants appeared: 'extremely fat', 'three times the size of a normal mouse', 'all body fat', eating constantly - 'extremely hyperphagic'. - Several strains shared the phenotype, yet 'they could tell even in the 1950s that these mutations were on different chromosomes'. Two were labeled 'obese' and 'diabetes', 'but they were basically the same'.

Doug Coleman at Jackson Labs proposed joining the circulations of the two strains to test for 'a circulating factor, a hormone that is produced by one of these strains'. When ob is attached to db, 'The ob mouse, that strain dramatically loses weight' - 'within a couple of weeks, it looks like a normal mouse' - while 'The db mouse, nothing really happens'. From that single result Coleman hypothesized 'these two mutations were mutations in a hormone and a receptor': ob lacks the fat-derived hormone, db lacks its receptor. Knight notes this 'was just an idea, it was a hypothesis' until 'in the 1980s' molecular biology made cloning possible: Friedman 'cloned leptin and showed that, in fact, Doug was exactly right', and 'Millennium Pharmaceuticals showed that the db mutation is in fact a receptor'.

- The ob gene is 'expressed in fat', 'exclusively expressed in adipose tissue', and its expression is 'directly proportional to how much body fat you have' - 'in a linear manner' - so 'the level of leptin in your blood is a direct readout of your body fat reserves'. - 'the functional form' of the leptin receptor 'is expressed almost exclusively in the brain', in 'all of the brain regions that we knew from previous work were important for appetite' - in Knight's phrase, 'a map in the brain of the neurons that control hunger'. - Weight loss makes leptin 'fall' because adipose tissue is lost; 'The absence of that hormone' triggers 'this entire homeostatic response to starvation': 'increased hunger', 'decreased energy expenditure', 'decreased body temperature, even decreased fertility'.

Huberman recalls that 'Amgen Pharmaceuticals owned the patent for leptin in hopes that it would become the blockbuster diet drug'. The logic he states is that giving the hormone would let 'the brain would be tricked into thinking that there was more body fat, more energy reserves than there was', so 'people would basically be less hungry, eat less, and lose body fat'. Knight describes the mood when leptin was cloned: 'it was thought, basically, we've cured obesity', with 'an auction for the patent' that Amgen won for 'something like $20 million upfront payment plus royalties'.

The trial disappointed: 'they did a clinical trial, gave obese people leptin, subcutaneous injections of this hormone, and they didn't lose a lot of weight'. The reason, 'subsequently revealed', is that 'individuals who are obese do not have low levels of leptin for the most part. They actually have high levels of leptin', producing 'a state of leptin resistance'. Knight's analogy is explicit: 'it's analogous to someone who has type 2 diabetes. It's not because they lack insulin, it's because they actually have, over time, a high level of insulin, and so target tissues stop responding to insulin'.

- Reanalysis stratified participants 'according to their starting leptin level': 'the people with the lowest levels of leptin, they lost the most weight when you gave them this drug', and 'the people with the highest levels of leptin lost the least weight'. - This leaves 'a rationale' for two scenarios: a subset with 'lower levels of leptin' who 'aren't people with mutations like the obese mouse', or treatment 'after weight loss', when 'your leptin levels plummet' and 'a big part of the reason it's so difficult to keep weight off is because those leptin levels are so low'.

Knight is candid about the unresolved part: 'Why it never made it as a drug for that application, I really don't understand.' He speculates it 'has something to do, I think, with the pharmaceutical industry, with the economics, with a bunch of other issues that aren't necessarily scientific', and that leptin 'could come back' now that 'these GLP-1 drugs' have 'millions of people losing so much weight' who might 'transition to a different kind of drug to keep the weight off'. He frames these as possibilities about the industry, not as established findings.

Knight divides feeding and other motivated behaviors 'into appetitive and consummatory phases'. Appetitive is 'searching for food', 'foraging', 'all the actions that lead up to the actual behavior itself'; consummatory is 'actually putting the food in your mouth and eating it'. 'the general thought' is that 'these forebrain circuits in the hypothalamus are more important' for the appetitive phase and 'the brain stem circuits are more important for the consummatory phase', namely 'licking, chewing, swallowing'. Within the hypothalamus, 'a population of neurons called AgRP neurons' - 'agouti-related peptide' - are 'absolutely critical for that appetitive phase, for the searching for food, for the desire to find food and consume it when you're hungry'.

Huberman asks whether AgRP neurons 'connect to areas of the brain and body that stimulate the desire to move' - whether they drive the locomotion of getting up to find food. Knight pushes back: 'they're not directly linked to any of those circuits'; they are 'linked directly to other forebrain circuits involved in motivation'. In his account the neurons are 'not directly talking to the motor circuits', but instead create 'this general problem that the animal has to solve' - 'I'm hungry. I need to get food' - which 'the animal uses all of its mental capacities to solve that problem'.

they're just there to set the goal, not so much to direct the solution.

'a few thousand neurons at the base of the hypothalamus', 'the most ventral, the most bottom part of the forebrain' - 'a tiny population of cells, but outsized importance for the control of feeding behavior'. Stimulate them in a non-hungry animal and 'the animal will voraciously eat like it's starving'; silence them and 'animals will starve to death', refusing food voluntarily 'until basically you have to euthanize them because they've lost so much weight'. Their activity 'is thought to track the body's need for energy', partly because 'they express these receptors for leptin' and 'leptin inhibits AgRP neurons' - 'one mechanism by which leptin controls hunger'.

Knight's lab asked 'What is the natural sort of firing of this population of neurons when an animal eats a meal' - 'a very basic question' that 'Was not really addressable until about 10 years ago' because 'these are such a tiny population of cells, so deep in the brain'. In work by 'one of my first graduate students, Yiming Chen', the lab used 'fiber photometry', putting 'a fiber optic into the mouse's brain' to 'record fluorescence from these AgRP neurons' using 'a calcium sensor, so calcium as a surrogate for neural activity'. The expectation was 'that these AgRP neurons would gradually decline in activity as the animal eats and levels of hormones in the blood start changing, feeding back to inhibit these neurons'.

Zach, I gave the mouse the piece of food, but the weirdest thing happened. The neurons shut off almost immediately.

Knight assumed error - 'Yiming, you've made a mistake. It's okay. You're just starting off in graduate school. This happens. Go back and repeat the experiment' - but the result repeated 'every single time': 'I give a hungry mouse food and the AgRP neurons within just a few seconds, their activity has greatly diminished back to the level it would be in a fed mouse, even before they take the first bite of food'. By varying 'the kind of food he gave them or the accessibility of the food or how hungry the mouse was', Chen showed 'what the neurons were doing was predicting'. The mouse 'looks at how palatable it is, it matches how hungry the mouse is, how accessible it is', and 'within a few seconds these neurons predict how much food the mouse is going to eat in the forthcoming meal'.

these neurons know how much the mouse is going to eat before the mouse even takes the first bite

The quantitative result is unusually clean: 'you give the mouse different foods' and 'look at how much these AgRP neurons drop when the mouse sees and smells the food' - 'this drop happens in three seconds, four seconds, something like that' - then 'plot that against' 'how much does the mouse go on to eat in the next 30 minutes. You can just draw a straight line'. Knight's interpretation is that these circuits, 'what they're constantly doing is predicting the future': signals from the body 'are slow', and 'you don't want to wait 20 minutes from the food that you ingested to reach your stomach and then slowly start entering your intestine to figure out what was the nutrient content of the meal'. Animals 'learn, presumably through just experience' that a given smell and look 'has about this many calories', and that information is transmitted 'to start the process of satiation before the meal begins'. This is the concrete version of Huberman's introduction claim that 'you have brain circuits that actually determine how much you're likely to eat even before you take your very first bite'.

Knight refuses to over-read the finding: 'we don't fully know the answer'. One interpretation is that the neurons 'control foraging alone, they don't control eating', enabling the shift from the appetitive to the consummatory phase. 'But that doesn't seem to be the whole explanation', because if 'you artificially stimulate these neurons, so prevent that drop from ever happening', 'the mouse will just sit there and eat'. 'The reality is that the entire behavior is linked and you can't fully separate them.' He adds a general caveat about the limits of the enterprise: 'in biology it's always hard to answer why something happens'.

You can figure out what happens, but the reason why it evolved that way is challenging.

- 'starting the process of satiety before the meal begins'. - 'reducing this appetitive drive and allowing the transition to consummatory behavior'. - 'cephalic phase responses' that 'are necessary to prepare you for a meal': the Pavlov example where 'the ring of the bell alone causes the dog to salivate in the absence of any food', plus 'the secretion of insulin occurs in response to food cues, changes in gastric acid, gut motility'. - Knight's own conclusion: 'it probably is doing all of these things'.

Huberman supplies a first-person test case: at a meal, when more food kept arriving, 'suddenly my appetite for the appetizers went down because I knew there was more food coming'. His counterfactual is that 'Had I not known that there was more food coming' 'I think I would have consumed more of the appetizers', and he reads this as 'clearly, there's something going on with these AgRP neurons at the moment' - 'integrating based on new information'.

He then draws a clinical parallel from a solo episode on eating disorders: people with anorexia are 'unbelievably tuned to the caloric content of food', their visual system 'almost hyper-accurate calculators of the amount of calories in food', 'with a very small margin of error', and 'that drives, in that condition, obviously, food avoidance'. He names anorexia nervosa 'the most deadly of the psychiatric conditions, sadly' and frames the contrast as 'A pathologic dysregulation, a maladaptive dysregulation' versus ordinary regulation.