A specific cartoon is responsible for a lot of bad programming. It goes like this: endurance work switches on an enzyme called AMPK, AMPK shuts off mTOR, mTOR is the master switch for building muscle, therefore cardio cancels your gains. The molecule even has a villain's name. From that single arrow on a slide, the field built an entire commandment — never run on leg day — and a lot of coaches enforce it as if it were thermodynamics.
We want to take the AMPK story apart and see what holds. Because concurrent training — pairing resistance and endurance work in the same program or session — is one of the most-studied questions in exercise physiology, and the molecular picture is both more interesting and less alarming than the slide suggests. The enzyme is real. The arrow is real. The size of the arrow is the part nobody puts on the slide.
What AMPK actually senses
AMP-activated protein kinase is a fuel gauge, not a cardio detector. It responds to the ratio of AMP and ADP to ATP inside the cell. When a muscle contracts hard enough to burn ATP faster than it can be regenerated, AMP and ADP accumulate, and AMPK gets phosphorylated and switched on. Its job is to restore energy balance: it promotes glucose uptake, fatty acid oxidation, and mitochondrial biogenesis (largely through PGC-1α), and it throttles back expensive anabolic processes the cell can't afford mid-crisis. One of the things it throttles is protein synthesis, partly by acting on the mTORC1 pathway and the molecule TSC2.
That last sentence is the entire basis for the interference fear. And on its own terms it is correct. The question coaches actually care about is different: how much, for how long, and under what conditions does that braking translate into less muscle on a human powerlifter eighteen weeks later.
Three mechanisms, judged on the same criteria
The interference effect — the observation that adding endurance training can blunt strength or hypertrophy gains — has at least three candidate explanations. They are not mutually exclusive. We can compare them on three honest criteria: how well the mechanism is demonstrated in cells, how well it survives in trained humans, and how directly a coach can manipulate it.
Mechanism 1: AMPK–mTOR signaling
In isolated muscle and rodent models, the signaling logic is clean. Activate AMPK pharmacologically (with AICAR) and you can suppress mTORC1 signaling and protein synthesis. Atherton's 2005 work in rat muscle (J Physiol) is the canonical demonstration that a single "endurance-like" stimulus and a "resistance-like" stimulus drive divergent signaling — the so-called AMPK–PKB switch. The mechanism is well-established at the bench.
In trained humans, it gets soft. AMPK activation after a single endurance bout is transient, often returning toward baseline within an hour or two. Several acute human studies that paired endurance and resistance work failed to find the predicted suppression of mTOR signaling or muscle protein synthesis at all. So: strong in cells, inconsistent in humans, and only indirectly manipulable — you cannot dial AMPK down, you can only avoid stacking its activation against the anabolic window.
Mechanism 2: Glycogen depletion
Low muscle glycogen independently elevates AMPK activity and appears to amplify endurance-type signaling. The "train-low" literature, much of it from John Hawley's and Keith Baar's groups, shows that exercising with depleted glycogen boosts the oxidative adaptation signal. For an endurance athlete that is a feature. For a lifter chasing hypertrophy it is a plausible saboteur, because the same depleted state that sharpens AMPK may blunt the resistance stimulus that follows.
This one is appealing because it is directly manipulable — carbohydrate availability is something a coach controls. But the demonstration that glycogen-mediated interference meaningfully reduces strength gains in humans is thinner than the mechanistic enthusiasm for it. Plausible, partially supported, easy to act on.
Mechanism 3: Residual fatigue and fiber recruitment
The least molecular explanation may be the most practical. Endurance work — especially long or high-impact running — produces peripheral and central fatigue, muscle damage, and acute reductions in force-producing capacity that can persist for hours to days. A fatigued athlete recruits high-threshold motor units less effectively and trains the subsequent resistance session at lower quality. No signaling cascade required. Wilson's 2012 meta-analysis (J Strength Cond Res) found the interference effect scaled with the modality, frequency, and duration of endurance work, with running producing more interference than cycling — a pattern that fits a fatigue-and-damage explanation more cleanly than a pure-signaling one.
This mechanism is the best supported in real training studies and the most directly manipulable. It is also the one the AMPK cartoon ignores entirely.
The AMPK cascade, in the order it happens
Walk it through once, in sequence, because the timing is where the fear falls apart.
You start an endurance bout. Within minutes, ATP turnover outpaces resynthesis, AMP and ADP rise, and AMPK is phosphorylated at Thr172. Active AMPK acts on TSC2 and on the mTORC1 component raptor, nudging the anabolic signal down while it redirects the cell toward energy restoration and oxidative gene expression. You stop exercising. Energy charge recovers. AMPK phosphorylation falls back toward baseline over the next one to two hours. The acute brake releases.
Now you lift. If you lifted while AMPK was still elevated — say, resistance work immediately after a depleting endurance session — you plausibly blunted the anabolic response of that one session. If you lifted after the gauge reset, the molecular conflict the slide warned about has largely passed, even though the fatigue from mechanism three may not have.
That is the crux. AMPK's suppressive window is measured in hours. A training program is measured in months. The cartoon collapses those timescales into a single arrow.
What the verdict looks like once you weigh them
| Mechanism | Demonstrated in cells | Survives in trained humans | Coach can manipulate it |
|---|---|---|---|
| AMPK–mTOR signaling | Strong | Inconsistent | Indirectly (timing) |
| Glycogen depletion | Strong | Thin but plausible | Directly (carbs) |
| Residual fatigue / recruitment | n/a | Best supported | Directly (modality, spacing) |
Read down the columns and a verdict assembles itself. AMPK is a genuine molecular reality and a fine explanation for what happens inside a cell in the first hour. It is a weak explanation for why a powerlifter who jogs would gain less strength over a block, because its effect is transient and the human signaling data refuses to behave. The variables that actually move the needle in training studies — endurance modality, session proximity, total endurance volume, and fuel state — are mostly the ones AMPK theory underweights.
It is worth saying plainly that some studies find no interference, and a few find concurrent training enhances hypertrophy, likely through improved capillarization, nutrient delivery, and work capacity. The interference effect is real but conditional, and AMPK is one input, not the gatekeeper.
An honest rule of thumb
If the goal is maximal strength or hypertrophy and you are also doing endurance work, separate the two sessions by at least six hours where the schedule allows, and do not perform the higher-impact, longer-duration endurance work in the hours immediately before a hard lower-body lift. Prefer cycling or other low-eccentric modalities when concurrent training is unavoidable, keep endurance volume to what the actual goal requires rather than a default, and do not lift in a glycogen-depleted state if hypertrophy is the priority. None of that requires fearing the enzyme. It requires respecting fatigue and timing.
What this doesn't answer
We still cannot tell you, for a given athlete, how large the interference cost is in pounds on a total — the between-person variation in published work is wide, and genetics, training age, nutrition, and sleep all modulate it in ways no acute signaling study captures. We also do not know how chronically elevated AMPK from high endurance loads, repeated over months, integrates with the resistance stimulus, because most mechanistic work is single-session.
The most useful next reading is the human side rather than the cell side: Baar's reviews on molecular concurrent-training programming, the Wilson 2012 meta-analysis for the modality and dose patterns, and any longitudinal trial that measures strength outcomes alongside signaling rather than instead of it. The molecule earned its place on the slide. It did not earn the commandment built on top of it.