The Per-Meal Leucine Floor: A Shift-Worker Protein Protocol

<!-- forge-cite-audit-2026-05-18 -->

Marcus is 41. Hospital security supervisor, level-one trauma center, overnight shift. Two kids at home. His relief partner Reggie has been working the same five-on rotation and is up nine pounds since February.

It is 02:47 on a Tuesday in the family lounge. Reggie slides a paper plate across the laminate. Cafeteria lasagna. 740 kcal, 14 grams of protein.

Hour eight of twelve. Marcus's container is open in front of him — 38g protein, ~4g leucine, 36g slow carb. He looks at Reggie's lasagna. Reggie looks at his container. Reggie says what every shift partner has said at hour eight for as long as cafeterias have existed:

"I don't know how you do that for five nights in a row."

The answer is not willpower. The answer is the per-feeding leucine math underneath it.

TL;DR

• The protein-priority appetite model (Raubenheimer & Simpson) keeps reproducing: when dietary protein density falls, total caloric intake climbs to compensate.
• **Total daily protein is the wrong unit.** Per-feeding leucine is the unit that decides whether muscle protein synthesis actually fires.
• The threshold for a clean Sestrin2 / mTORC1 signal lands around **~0.36 g/kg leucine per feeding** in younger adults and somewhat higher (~0.4 g/kg) in adults over 40 — based on Moore 2009, Witard 2014, and the broader leucine-threshold literature.
• The cofactor floor is non-negotiable: methylated B vitamins, magnesium glycinate, alpha-lipoic acid. Strip any one and the BCKDC complex bottlenecks.
• Night shift amplifies the problem on three rails — cortisol blunts mTORC1, melatonin tanks insulin sensitivity at 03:00, and sleep debt jacks AgRP excitability.
• The 5-feeding architecture: **07:08 post-shift, 13:30 wake-up, 16:42 pre-shift, 22:30 in-shift, 03:30 hour-eight defense.**

The headline number hides the real number

The protein-priority appetite model says one thing the food industry would prefer you forget. Humans do not eat to a calorie target. We eat to a protein target.

When dietary protein density drops below ~12% of energy, total intake climbs to compensate. We keep eating until enough absolute protein crosses the gut wall to satisfy a regulatory set point.

That dose-response curve is steep at the bottom. Dropping from 14% to 11% costs more in compensatory overshoot than dropping from 18% to 14%. Shift workers carry an additional multiplier on top of that — cortisol, sleep restriction, and melatonin disruption all push appetite upward at the same time they make synthesis less efficient.

Here is the part most podcasts did not quote. Distributing protein across multiple feedings, each of which clears the per-meal leucine threshold, blunts the overshoot even when total daily protein is held constant.

That is the operating principle. Per-feeding leucine on a circadian-aware schedule is the unit that decides whether the cafeteria run happens at hour eight.

Sestrin2 is the switch. FGF21 is the drive.

The mechanism walks two parallel rails — one in muscle, one in the brain.

The muscle rail

Wolfson and the Sabatini lab (Nature 2016) reported that Sestrin2 is the direct cellular leucine sensor. Leucine binds Sestrin2 → Sestrin2 releases GATOR2 → GATOR1 stops blocking RagA/B → mTORC1 docks at the lysosome and switches on muscle protein synthesis.

The latch has a sharp threshold. Below ~0.36 g/kg leucine per feeding (Moore 2009 AJCN n=24; Witard 2014 AJCN n=48; Murphy 2018 BJN n=20 older adults at ~0.42), Sestrin2 does not fully open. The meal becomes substrate without the anabolic signal.

The brain rail

When circulating BCAAs fall, the liver sensor GCN2 phosphorylates eIF2α. ATF4 binds the FGF21 promoter. Hepatic FGF21 climbs, crosses the blood-brain barrier, and acts on KLB-expressing AgRP neurons in the arcuate nucleus to drive a protein-specific appetite (Solon-Biet et al., Cell Metabolism, 2014; Hill et al., Cell Metabolism, 2019).

The 03:14 vending machine is not willpower. It is a hardwired loop closing on a missing feeding.

BCKDC is the ceiling

The leucine signal is half the equation. What happens to the carbon skeletons after the signal fires is the other half.

BCKDC — the branched-chain alpha-ketoacid dehydrogenase complex — has four cofactor dependencies:

• **Thiamine pyrophosphate** (needs thiamine + magnesium)
• **Riboflavin-5-phosphate** (needs riboflavin)
• **NAD+** (needs niacin)
• **Alpha-lipoic acid** covalently bound to the E2 subunit

Strip any one and the complex bottlenecks. You can hit the per-feeding leucine number and still leak the signal as wasted nitrogen.

Night-shift workers run depleted on B1, B2, and B6 at baseline. A large fraction of the population is also heterozygous for MTHFR C677T (Botto & Yang, Am J Epidemiol, 2000, pooled cohort), which is one of the practical reasons methylated and bioactive forms are non-negotiable. Folic acid is not L-methylfolate. Cyanocobalamin is not methylcobalamin. P-5-P is not pyridoxine HCl — there are documented paradoxical-neuropathy concerns with pyridoxine HCl at the higher doses you would need to backfill a depleted shift worker.

The cofactor floor decides whether the leucine you swallowed at 16:42 makes muscle or makes urea.

The cortisol window eats the cafeteria plate

Night shift amplifies the problem on three axes.

Cortisol blunts mTORC1. Peak cortisol in a phase-shifted supervisor lands in the early morning hours, exactly when the cafeteria run gets tempting. Sustained supraphysiologic cortisol meaningfully suppresses muscle protein synthesis even when leucine is present in circulation.

Melatonin tanks insulin sensitivity at 03:00. Scheer et al. (PNAS, 2009, forced-desynchrony n=10) showed the cafeteria plate's bechamel pushes glucose meaningfully higher at night than the same plate at noon. The insulin spike clears amino acids out of plasma faster than the brain can register a satisfied protein signal.

Sleep debt jacks AgRP excitability. Markwald et al. (PNAS, 2013, n=16) measured appetite-driving signaling climbing per night of restricted sleep. By night five of a five-on, Reggie's AgRP neurons sit on a hair trigger and the lasagna is the only available answer.

The two feedings that escape the cortisol window are pre-shift before 17:30 and post-shift after 06:30. Those are the windows where the Sestrin2 latch fully opens. The middle of the shift is for stable glucose, electrolytes, and a tactical low-leucine snack that does not pretend to drive synthesis it cannot deliver.

The 5-feeding architecture

Three feedings above the per-meal leucine threshold. Two strategic.

Feeding 1 — 07:08 post-shift

Casein-forward. 42g protein, ~4g leucine. Eaten before sleep. Drives MPS through the eight-hour fasted window of daytime sleep.

Feeding 2 — 13:30 wake-up

38g protein, ~4g leucine. Paired with the cofactor reload:

• Methylcobalamin 1mg
• L-methylfolate 800mcg
• P-5-P 50mg
• Thiamine HCl 100mg
• Riboflavin-5-phosphate 25mg
• Niacinamide 100mg
• Alpha-lipoic acid 300mg
• Magnesium glycinate 200mg

Feeding 3 — 16:42 pre-shift

The biggest leucine hit of the day. 49g protein, ~4.4g leucine from a slow-cooked chuck roast. 38g resistant starch from cold-stored jasmine rice. Three softgels of 2:1 EPA-to-DHA krill.

Cortisol is at the post-CAR diurnal floor. The Sestrin2 window opens cleanly.

Feeding 4 — 22:30 in-shift

25g whey isolate plus 5g free leucine in a shaker. The GCN2 reset that buys the next four hours and pulls FGF21 off the climb before AgRP gets loud.

Feeding 5 — 03:30 hour-eight defense

Not a meal. A defensive holding pattern.

• Hard-boiled egg pack
• Casein-forward bar prepped at home
• Electrolytes
• 200–300mg L-theanine to chip at the cortisol peak

Never the cafeteria lasagna. Never the family-lounge donuts. Never the 03:14 vending machine.

Then the 07:08 feeding becomes feeding one of the next 24-hour cycle and the loop holds.

What the AI coaching system actually does with this

This is the work no human coach can sit next to you and do at 04:18 inside a trauma elevator on the way to the unit.

Chiron is the always-on AI coach in Marcus's Telegram thread. DMs back in under 60 seconds. Holds full history. Knows the rotation calendar, knows tonight is night four of a five-on, knows the last A1c, knows the per-feeding leucine target at his bodyweight. At 16:55 it pings him with the protein math and the methylated stack reminder. Every shift. Never ghosts. Send a photo of the cafeteria plate at 03:14 and it tells him whether he is looking at feeding four or feeding five for the night.

HERMES is the research bot. When new protein-distribution work hits PubMed, HERMES ingests the supplementary tables, flags the load-bearing findings, and pushes updated numbers to Chiron with effect sizes and sample sizes intact.

Forge is the operations agent. Watches the seven-day HRV trend, the sleep debt across the five-on, and protein-per-feeding compliance. Auto-deloads the Wednesday lower-body session before Marcus would have asked. If body weight stalls for ten days, Forge runs the diet-break math.

A human coach — even a great one — cannot be in your pocket at 02:47 with the leucine math live and the last seven nights of HRV in working memory.

An AI coaching system that thinks like a strength coach plus a nutritionist plus a circadian biologist, with persistent memory across every conversation Marcus has ever had with it, can.

Reggie pushed the lasagna back across the laminate at 02:51 last Tuesday and asked what Marcus was eating. The answer is not what he is doing. The answer is who is doing it with him. You can meet the system at legacyinmotion.fit.

---

The data behind this

• Wolfson RL et al. Nature. 2016;536:7596.
• Moore DR et al. Am J Clin Nutr. 2009;89:161–168.
• Witard OC et al. Am J Clin Nutr. 2014;99:86–95.
• Murphy CH et al. Br J Nutr. 2018;120:984–994.
• Solon-Biet SM et al. Cell Metabolism. 2014;19:418–430.
• Hill CM et al. Cell Metabolism. 2019;30:18–26.
• Botto LD, Yang Q. Am J Epidemiol. 2000;151:862–877.
• Scheer FA et al. PNAS. 2009;106:4453–4458.
• Markwald RR et al. PNAS. 2013;110:5695–5700.
• Raubenheimer & Simpson, protein-priority appetite model — foundational and replicated work.