The biology of fasting
Autophagy
Also called: Macroautophagy, Self-eating
Written by Aaron CuhaReviewed Sep 2026
The cell's recycling system: damaged proteins and worn-out structures get wrapped in a membrane, delivered to the lysosome, and broken back down into parts the cell reuses.
What it is
Autophagy is how a cell takes out its own rubbish and reuses it. A double membrane forms around whatever is being disposed of, closes into a sac called an autophagosome, and fuses with a lysosome, which is the cell's acid-filled digestion chamber. The contents are broken down to amino acids, fatty acids and sugars, and released back into the cell to be used again.
It runs constantly at a low level in every cell you have. It is not a switch that fasting turns on from zero. What changes with nutrient supply is the rate.
Yoshinori Ohsumi won the 2016 Nobel Prize in Physiology or Medicine for working out the genetics of this process in yeast, which is where nearly all the machinery was first identified.
The machinery
The initiating step is the ULK1 complex, a kinase that gets the membrane forming. ULK1 sits directly downstream of the two nutrient sensors: mTORC1 phosphorylates ULK1 at serine 757 to hold it off when food is plentiful, and AMPK phosphorylates it at other sites to switch it on when energy is short.
Downing of mTORC1 therefore releases the brake, and rising AMPK presses the accelerator. Both happen when you stop eating.
The growing membrane is decorated with LC3, which starts as LC3-I in the cytosol and gets a lipid tag added to become LC3-II, which anchors into the autophagosome membrane. LC3-II is the marker almost every study measures, because the amount of it tracks the number of autophagosomes present.
p62, also called SQSTM1, is a cargo receptor. It grabs tagged rubbish, binds LC3, and gets destroyed along with the cargo, so p62 going down is usually read as autophagy running.
What switches it on
Falling insulin, falling amino acids (leucine especially), and a rising AMP to ATP ratio. Each acts on a different part of the same control point. Exercise raises the same signals and is a genuine comparator: some of the clearest human autophagy measurements come from exercise studies rather than fasting studies.
On the timing
There is no verified human study putting a clock hour on when autophagy starts. What has been measured in people is the direction of change at particular durations and in particular tissues, and those measurements disagree with each other depending on which tissue you sample. Blood cells and skeletal muscle do not tell the same story, which is the single most useful thing to know about this literature and the thing most often left out.
What has been measured
In people
Autophagy has been measured in living people many times over, including with assays that measure true flux rather than a static snapshot. The results are real, modest, and tissue-dependent, and reading them requires knowing which studies measured a rate and which measured a stock.
- Autophagic FLUX, the actual rate rather than a count of autophagosomes, was measured in 121 people with obesity randomised for six months to standard care, calorie restriction, or intermittent fasting plus time-restricted eating. At six months the fasting arm differed significantly from standard care (P = 0.04, post hoc). Calorie restriction did not. The authors note the effect may be driven partly by autophagy declining in the control group, and there was no significant rise from baseline within the fasting arm itself (Journal of Physiology, 2025).
- In 11 overweight adults doing a four-day randomised crossover of eating 8am to 2pm versus 8am to 8pm, the early eating window raised expression of the autophagy gene LC3A and the ageing gene SIRT1 in whole blood cells in the morning before breakfast, and raised MTOR expression in the evening (Nutrients, 2019).
- In 50 women fasting 24 hours three days a week for eight weeks, with vastus lateralis biopsies taken after both 12 and 24 hour fasts, the 24 hour fast raised SQSTM1, while BECLIN1, SQSTM1 and LAMP2 mRNA fell in the reduced-energy arm after an overnight fast. The authors concluded autophagy markers in human muscle were reduced (Nutrition, 2022). Read that alongside the measurement caveat below, because a falling marker is not the same as falling autophagy.
- The measurement caveat, and it is the most important thing on this page. In human skeletal muscle, LC3B-II protein fell 24% immediately after exercise. The same team then ran an ex vivo flux assay on that same human tissue and found flux was not falling at all: effect sizes indicated a modest to large INCREASE in autophagy flux lasting up to 24 hours. Their stated conclusion is that exercise-induced decreases in LC3B-II do not reflect autophagy flux level. In rats the same protocol moved the markers the opposite way, LC3B-I up 109% and LC3B-II up 97%. So a study reporting that human muscle autophagy markers went down has not shown that autophagy went down (International Journal of Molecular Sciences, 2022).
- After a 72 hour fast in 8 healthy men with forearm tracer kinetics and muscle biopsies, mTOR phosphorylation fell by roughly 50%, with reduced phosphorylation of its downstream targets 4EBP1, rpS6 and ULK1, the kinase that initiates autophagosome formation (PLoS One, 2014).
- Skeletal muscle autophagy after a 36 hour fast has been compared between trained and untrained people (Journal of Applied Physiology, 2018).
In other species and in cell culture
The mechanism itself was worked out in yeast and is characterised in far more detail in mice than in people, including the tissues that are hardest to biopsy in a living human.
- In the same 2022 study that found human muscle markers falling, intermittent fasting raised hepatic LC3-I protein and Map1lc3b mRNA in mice on both chow and high-fat diets, and raised LAMP1 protein and Beclin1 mRNA in liver on chow. Mouse muscle, like human muscle, showed no activation. Liver is the tissue where fasting-induced autophagy is clearest, and it is the tissue nobody biopsies in a healthy volunteer.
- Nearly all ATG genes and the core membrane machinery were identified in Saccharomyces cerevisiae, the work recognised by the 2016 Nobel Prize.
Why it matters
Failing autophagy is one of the recognised hallmarks of biological ageing, and it is implicated in neurodegenerative disease, where undegraded protein aggregates accumulate. That is why the question of whether an ordinary person can raise it deliberately is worth asking at all.
The honest state of the answer is that one six-month randomised trial in 121 people found flux higher in a fasting arm than in standard care, and that is the strongest human result there is. It is a real finding. It is also a single post hoc comparison with no within-group increase, in blood cells rather than in brain or liver.
A claim you will see repeated
You will see a specific hour quoted as the moment autophagy switches on, most often 16 hours. We traced that number through the eight people most often credited with it, reading the primary sources. Not one of the four researchers closest to the autophagy literature gives 16 hours. Valter Longo says about five days and calls the short-fast version out directly. Peter Attia says it cannot currently be measured in humans at all. Satchin Panda names it as the open question rather than answering it. Rhonda Patrick put the 16 hour question to autophagy researcher Guido Kroemer, whose answer began "We don't know" and who said his own leukocyte work needed three to four days. Jason Fung gives no number anywhere in his written work. Mark Mattson's much-quoted 16 to 18 hours, in the New England Journal of Medicine, is advice on how to titrate a patient into a fasting schedule, and in that paper his hour figures are about ketones and never about autophagy. The number appears to originate in a 2018 mass-market book, where the sentence "Research shows that 16 hours is optimal" carries no citation and is followed by a reference to a study of 4-day and 5-day fasting-mimicking diet cycles that contains no 16 hour arm at all. The most-cited paper for short fasting and autophagy, Alirezaei 2010, is a mouse study that tested 24 and 48 hours. The closest human anchor, Jamshed 2019, used an 18 hour fast in 11 people and measured one gene.
Citations
- Human2025Intermittent time-restricted eating may increase autophagic flux in humans: an exploratory analysis
The Journal of Physiology
121 people with obesity randomised for 6 months to standard care, calorie restriction, or intermittent fasting plus time-restricted eating. LC3B-II flux in peripheral blood mononuclear cells within whole blood, a rate measurement rather than a static marker. Significant difference in change from baseline between the fasting arm and standard care at 6 months (P = 0.04, post hoc); calorie restriction versus standard care not significant; no significant change from baseline within the fasting arm.
- Human2019Early Time-Restricted Feeding Improves 24-Hour Glucose Levels and Affects Markers of the Circadian Clock, Aging, and Autophagy in Humans
Nutrients
11 overweight adults, 4-day randomised crossover of an 8am to 2pm window versus 8am to 8pm. The early window raised whole blood cell expression of the autophagy gene LC3A and of SIRT1 before breakfast (p < 0.04), raised MTOR expression in the evening (p = 0.007), and lowered mean 24-hour glucose by 4 mg/dL and glycaemic excursions by 12 mg/dL.
- Human2022Intermittent fasting activates markers of autophagy in mouse liver, but not muscle from mouse or humans
Nutrition
50 women, mean age 51, randomised to intermittent fasting (24-hour fast on 3 non-consecutive days per week) at 70% or 100% of energy requirements for 8 weeks, with vastus lateralis biopsies after 12 and 24 hour fasts. A 24 hour fast raised SQSTM1; BECLIN1, SQSTM1 and LAMP2 mRNA fell in the 70% arm after an overnight fast. Autophagy markers in human muscle were reduced, the authors attribute this to weight loss.
- Human2014Fasting increases human skeletal muscle net phenylalanine release and this is associated with decreased mTOR signaling
PLoS One
8 healthy men studied postabsorptive and after a 72 hour fast, with forearm amino acid tracer kinetics and muscle biopsies. mTOR phosphorylation fell about 50%, with reduced downstream phosphorylation of 4EBP1, rpS6 and ULK1, the kinase that starts autophagosome formation.
- Human2022Exercise and Training Regulation of Autophagy Markers in Human and Rat Skeletal Muscle
International Journal of Molecular Sciences
The keystone methodological result. In human skeletal muscle LC3B-II protein fell 24% immediately after exercise and returned to baseline by 3.5 hours, while in rat soleus the same protocol raised LC3B-I 109% and LC3B-II 97%. An ex vivo flux assay on the human tissue showed the fall did NOT reflect reduced flux: effect sizes indicated a modest to large increase in autophagy flux lasting up to 24 hours. The authors conclude that exercise-induced decreases in LC3B-II do not reflect autophagy flux level, and that rodents and humans move these markers in opposite directions.
- Human2018Training state and skeletal muscle autophagy in response to 36 h of fasting
Journal of Applied Physiology
Human skeletal muscle autophagy markers compared between training states across a 36 hour fast.