Autophagy — A Sardine Protocol Dossier

TL;DR

Autophagy is real biology and short fasts in humans almost certainly engage it, but the magnitudes, tissue distributions, and clinically meaningful endpoints are not pinned down to the level that the popular fasting discourse implies. The molecular biology is mapped ^{1 2}; rodent fasting → autophagy is unambiguous ^{3 4}; human short-fast data is thin and mostly relies on PBMC-marker proxies. We deliberately do not lead our protocol claims with autophagy, and this dossier walks through why — and what to track honestly across cycles. The Sardine Protocol's working position is that autophagy is a "probably-also-engaged" mechanism that adds confidence to the cycle bundle, not a load-bearing claim.

What we mean by autophagy

Autophagy ("self-eating") is a cellular recycling process. Damaged organelles, misfolded proteins, and other cellular debris are tagged with ubiquitin or LC3-II adducts, encapsulated in a double-membrane vesicle called an autophagosome, and delivered to the lysosome for hydrolytic breakdown. The constituent amino acids, lipids, and nucleotides feed back into the cellular pool. Autophagy operates constitutively at low levels in healthy cells and is upregulated under stress — nutrient deprivation, hypoxia, oxidative stress, and intracellular pathogen presence are the canonical triggers.

Three sub-types matter for the fasting story:

• Macroautophagy — the bulk-degradation pathway most often studied; "autophagy" without qualification almost always refers to macroautophagy.
• Chaperone-mediated autophagy (CMA) — selective protein degradation via Hsc70 chaperone targeting; less prominent in the fasting literature but mechanistically distinct.
• Mitophagy — selective autophagic clearance of damaged mitochondria; relevant to the mitochondrial-biogenesis story ¹.

The molecular machinery is well-characterized in Mizushima 2008 and the disease-implication framework is laid out in Levine & Kroemer 2008 — together these two reviews are the conceptual foundation for almost every modern autophagy paper.

Master regulation:

• mTORC1 is the master negative regulator. When nutrients (especially branched-chain amino acids), insulin, and growth factors are abundant, mTORC1 is active and autophagy is suppressed. mTOR signaling biology is comprehensively reviewed in Saxton & Sabatini 2017.
• AMPK is the master positive regulator. Low cellular energy charge activates AMPK, which both inhibits mTORC1 and directly phosphorylates ULK1, the autophagy-initiating kinase.
• TFEB is a transcription factor coordinating expression of autophagy and lysosomal genes; nuclear translocation is gated by mTORC1 phosphorylation.

The reasoning that drives popular fasting content is: short fasts lower mTORC1, raise AMPK, induce autophagy, therefore short fasts produce the disease-relevant benefits seen in autophagy-upregulated models. The first three steps are well-supported. The leap from "autophagy upregulates in cells/animals during fasting" to "fasting produces measurable disease-relevant benefits in humans via autophagy" is where the popular discourse outruns the human data, and where this dossier earns its keep by being calibrated rather than enthusiastic.

What the evidence says (the public preview cuts here)

Rodent fasting → autophagy is unambiguous.

Even short fasts (24 hours) produce measurable increases in autophagic flux across multiple tissues in mice and rats. Alirezaei 2010 showed striking neuronal autophagy induction after 24-hour and 48-hour fasts using LC3-II/I ratio markers and direct autophagosome counting — a finding that surprised the field at the time because brain was assumed to be relatively protected from autophagic stress. Bagherniya 2018 catalogs the rodent and limited human evidence comprehensively.

Human fasting → autophagy is plausibly engaged but methodologically constrained.

Three structural problems make human autophagy research hard:

1. Tissue access. The gold-standard markers (LC3-II/I ratio, p62 degradation, autophagosome quantification by electron microscopy) require tissue biopsy. Human studies typically settle for peripheral blood mononuclear cells (PBMCs) as a proxy, which may not reflect what's happening in metabolically relevant tissues like liver or skeletal muscle.
2. Static vs flux. Autophagy is a dynamic process; what matters biologically is the rate at which substrates flow through autophagosomes to lysosomes. Static measurements of LC3-II abundance are confounded by the balance of autophagosome formation and degradation. Flux assessment requires inhibitor blocking, which has ethical limits in healthy human subjects.
3. Sample size. Most human autophagy studies enroll under 20 participants; statistical power for nuanced effects is limited.

What the available data shows:

• Several small studies have measured autophagy markers in PBMCs after 24–72 hour human fasts and reported increases consistent with the rodent picture. Effect sizes are typically modest but directionally consistent.
• The Brandhorst & Longo 2015 fasting-mimicking diet study measured indirect autophagy markers in midlife adults across three monthly 5-day FMD cycles and reported changes consistent with autophagic engagement, alongside other systemic effects (IGF-1 reduction, immune cell turnover, inflammatory marker shifts). The authors are appropriately careful in attributing specific outcomes to autophagy versus the broader "metabolic switch" biology.
• The de Cabo & Mattson 2019 NEJM review catalogs autophagy among the proposed mechanisms of intermittent fasting's metabolic benefits, with appropriate caution about the rodent-to-human translation gap. The Mattson 2017 review on metabolic switching and the Anton 2018 "metabolic switch" framing similarly include autophagy in the proposed downstream-of-switching adaptive bundle.
• Hofer & Madeo 2024 extends the picture by examining how dietary spermidine — a natural polyamine present at varying levels across foods — interacts with autophagy induction and post-fast adaptation. This is one of the more sophisticated current attempts to reason about which compounds during refeed preserve or amplify autophagic adaptations rather than simply turn them off.

The βHB-as-signal angle.

Newman & Verdin 2014 and Veech 2004 raise the possibility that β-hydroxybutyrate, produced during ketosis, has direct signaling effects (HDAC1/2/3 inhibition, NLRP3 inflammasome modulation) that intersect with autophagy biology. The evidence here is largely preclinical; whether human ketosis at the levels reached during a sardine fast (1.5–2.5 mmol/L) produces autophagy-relevant signaling in humans is unstudied.

The protein-restriction angle.

Fontana 2008 demonstrated that calorie restriction with adequate protein intake did not lower IGF-1 in long-term CR practitioners, while protein-restricted CR did. Levine & Longo 2014 extends this with NHANES cohort data on protein intake and mortality in midlife. Both papers bear on autophagy because mTORC1 — the master negative regulator of autophagy — is most strongly suppressed by amino acid restriction, not by simple energy deficit. A short fast with moderate protein (a sardine fast) produces less mTORC1 suppression than a true water fast or a protein-restricted FMD; the autophagy implication is that the AMPK arm is fully engaged but the mTORC1 arm is partially attenuated.

The honest summary: rodent fasting → autophagy is a closed case; human short-fast → autophagy is probably engaged based on multiple converging lines of evidence (PBMC markers, indirect substrate evidence, mechanistic plausibility), but the magnitude, the tissues most affected, the duration of effect, and the clinically meaningful endpoint impact are not yet pinned down. The cleanest human substrate-biology backstop remains the Cahill 1970 starvation review showing that fasting transitions which would be expected to upregulate autophagy do occur in humans across the relevant timescales.