5D Library › Ketone Metabolism, β-Hydroxybutyrate Signalling and the Bioenergetics of Intermittent Fasting · 中文
Chapter 1 The Metabolic Switch: from Glucose to Ketones
What happens overnight
Mattson and colleagues call the change of fuel between feeding and fasting the metabolic switch: from a state dominated by glucose, supported by glycogen stored after meals, to one dominated by fatty acids and ketone bodies (Mattson et al. 2018). The trigger is the fall in liver glycogen. This chapter first settles the chemical accounts of these fuels, and then describes the time course of the switch with a simple two-compartment model.
Sugar and fat compete: the Randle cycle
In 1963 Randle and colleagues found, in isolated heart and diaphragm muscle, that oxidising fatty acids inhibits the use of glucose; they called it the glucose–fatty acid cycle, now usually known as the Randle cycle (Randle et al. 1963). Roughly, acetyl-CoA and NADH from fatty-acid oxidation inhibit pyruvate dehydrogenase, accumulating citrate inhibits phosphofructokinase, and glucose oxidation slows. Reviewing the theory, Hue and Taegtmeyer pointed out that it has a reverse half too: when glucose is plentiful, malonyl-CoA blocks the entry of fatty acids into mitochondria and fatty-acid oxidation is inhibited (Hue and Taegtmeyer 2009).
This competition has an important meaning in fasting: as liver glycogen falls and insulin drops, muscle and heart use more fatty acids, leaving the limited glucose to the brain and red blood cells, which need it.
Where ketone bodies come from
There are three ketone bodies: acetoacetate, D-β-hydroxybutyrate (BHB below) and acetone. The liver makes them from fatty acids. In the mitochondria of liver cells, β-oxidation turns fatty acids into acetyl-CoA; when acetyl-CoA is produced faster than the citric-acid cycle can handle it, two molecules of acetyl-CoA are joined through the HMG-CoA pathway into one of acetoacetate. Part of the acetoacetate is reduced to BHB by β-hydroxybutyrate dehydrogenase using NADH, and part decarboxylates spontaneously to acetone, which is mostly breathed out (Laffel 1999).
The liver lacks the key enzyme for using ketones, succinyl-CoA:3-oxoacid CoA-transferase (SCOT), so it makes ketones without using them and exports them to other tissues.
Calculation
The chemical accounts of several fuels: ATP per molecule (with modern P/O ratios of 2.5 ATP per NADH and 1.5 per FADH2), ATP per molecule of oxygen consumed, respiratory quotient (carbon dioxide over oxygen), and the balance sheet when the liver turns one molecule of palmitate into ketones.
# Stoichiometry of fuels. ATP yields use the modern P/O ratios (2.5 ATP per NADH, 1.5 per FADH2),
# 10 ATP per acetyl-CoA through the citric acid cycle, and count activation costs.
# Glucose: 2 ATP + 2 NADH (glycolysis) + 2 pyruvate -> 2 acetyl-CoA (2 NADH) -> ~32 ATP
# (30 if cytosolic NADH enters by the glycerol-phosphate shuttle).
# Palmitate (C16): activation costs 2 ATP equivalents; 7 rounds of beta-oxidation give 7 FADH2,
# 7 NADH and 8 acetyl-CoA -> 106 ATP.
# D-beta-hydroxybutyrate: BHB -> acetoacetate gives 1 NADH; acetoacetate is activated by
# succinyl-CoA:3-oxoacid CoA-transferase (SCOT), which uses up the succinyl-CoA that
# would otherwise have made 1 GTP; then 2 acetyl-CoA -> 21.5 ATP.
# Acetoacetate: as BHB without the NADH -> 19 ATP.
FUELS = { # name: (ATP per molecule, O2 consumed, CO2 produced, molar mass g/mol)
"glucose": (32.0, 6.0, 6, 180.16),
"palmitate": (106.0, 23.0, 16, 256.42),
"BHB": (2.5 + 20 - 1, 4.5, 4, 104.10),
"acetoacetate": (20 - 1, 4.0, 4, 102.09),
}
print("fuel ATP/molecule ATP per O2 RQ (CO2/O2) ATP per gram")
for name, (atp, o2, co2, mm) in FUELS.items():
print(f"{name:13s} {atp:12.1f} {atp / o2:12.2f} {co2 / o2:13.2f} {atp / mm:14.3f}")
# Ketogenesis in the liver: how many ketone bodies can one palmitate give?
# palmitate -> 8 acetyl-CoA; 2 acetyl-CoA -> acetoacetate (via HMG-CoA); so at most 4 AcAc.
# AcAc + NADH -> BHB (beta-hydroxybutyrate dehydrogenase); AcAc -> acetone + CO2 (spontaneous).
acetyl = 16 // 2
print(f"one palmitate -> {acetyl} acetyl-CoA -> at most {acetyl // 2} acetoacetate"
f" (+ {7} NADH and {7} FADH2 kept by the liver)")
# The liver keeps the beta-oxidation energy: 7 FADH2 + 7 NADH - 2 (activation) = 26 ATP if it
# exports acetoacetate; reducing all 4 to BHB spends 4 NADH (10 ATP), leaving 16.
liver_acac = 7 * 1.5 + 7 * 2.5 - 2
liver_bhb = liver_acac - 4 * 2.5
print(f"ATP kept by the liver per palmitate: {liver_acac:.0f} (exporting acetoacetate) or {liver_bhb:.0f} (exporting BHB)")
print(f"ATP other tissues get from 4 BHB: {4 * 21.5:.0f}; liver + tissues = {liver_bhb + 4 * 21.5:.0f}"
f" vs {106} burning palmitate whole (the gap = 4 GTP lost at the SCOT step)")
# fuel ATP/molecule ATP per O2 RQ (CO2/O2) ATP per gram
# glucose 32.0 5.33 1.00 0.178
# palmitate 106.0 4.61 0.70 0.413
# BHB 21.5 4.78 0.89 0.207
# acetoacetate 19.0 4.75 1.00 0.186
# one palmitate -> 8 acetyl-CoA -> at most 4 acetoacetate (+ 7 NADH and 7 FADH2 kept by the liver)
# ATP kept by the liver per palmitate: 26 (exporting acetoacetate) or 16 (exporting BHB)
# ATP other tissues get from 4 BHB: 86; liver + tissues = 102 vs 106 burning palmitate whole (the gap = 4 GTP lost at the SCOT step)A few points stand out. First, per molecule of oxygen, BHB (4.78) beats fatty acid (4.61) but falls short of glucose (5.33). So the claim that ketones are more oxygen-efficient than glucose is false; they are more efficient only than fatty acids. Second, the SCOT step uses up succinyl-CoA that would otherwise have made a GTP, so routing fat through ketones costs the body 4 ATP compared with burning the fatty acid directly (102 against 106), about 4 %. Third, the liver takes most of the energy of β-oxidation for itself as it makes ketones (keeping 16 ATP when it exports BHB), which is why it can support itself during a fast while also fuelling the brain.
Evidence
In 1967 Owen, Cahill and colleagues catheterised the cerebral vessels of obese volunteers during prolonged fasting and found that after several weeks ketone bodies could supply more than half of the brain’s oxygen consumption, with a large fall in its glucose requirement (Owen et al. 1967; Cahill 2006). This overturned the old view that the brain can use only glucose. Note that prolonged fasting was a medically supervised research setting and is no basis for personal practice.
How fast glycogen runs out
The liver of an adult stores some tens of grams to over a hundred grams of glycogen, depending on the person and on the previous day’s diet. Rothman and colleagues measured human liver glycogen directly with nuclear magnetic resonance during a fast and found that it fell steadily and roughly linearly during the first twenty-odd hours; over this period glycogenolysis supplied less than half of the body’s glucose output, the rest coming from gluconeogenesis (Rothman et al. 1991).
Calculation
A simplified two-compartment model: compartment one is liver glycogen, compartment two blood BHB. Glycogen breakdown slows as the store shrinks; an insulin-to-glucagon index falls as glycogen falls; the lower the index, the more free fatty acids adipose tissue releases and the weaker insulin’s brake on ketogenesis; tissue uptake of ketones is saturable. The parameters are illustrative, chosen only to reproduce the rough time course in the literature, not fitted.
# A two-compartment model of the metabolic switch during a fast (illustrative, not fitted to data).
# Compartment 1: liver glycogen G (grams). Compartment 2: blood D-beta-hydroxybutyrate K (mmol/L).
# Time t = hours since the end of the evening meal. Ideas built in:
# * liver glycogen is broken down to support blood glucose, more slowly as the store runs low
# (it is roughly exhausted after about 24-36 h of fasting in adults);
# * an insulin-to-glucagon index R falls as glycogen falls (R = 1 well fed, near 0 when exhausted);
# * as R falls, adipose lipolysis rises (free fatty acids F) and the liver's ketogenesis is
# released from insulin's brake (the Randle-cycle side: fatty acids then also spare glucose);
# * ketone uptake by tissues is saturable, so blood levels rise steeply once production exceeds it.
G0 = 90.0 # g liver glycogen after an evening meal (typical adult range ~ 70-110 g)
def rates(G, K, R_override=None):
glyco = 3.2 * G / (G + 20.0) # g/h glycogenolysis
R = R_override if R_override is not None else max(0.0, min(1.0, (G / G0) ** 1.5))
F = 0.35 + 1.15 * (1 - R) # mmol/L free fatty acids
keto = 0.25 * F / (1 + (R / 0.25) ** 2) # mmol/L per h ketone production
uptake = 1.0 * K / (K + 2.5) # saturable uptake and oxidation
return glyco, R, F, keto, uptake
def fast(hours, dt=0.01, R_override=None, K0=0.08):
G, K, out = G0, K0, {}
for i in range(int(hours / dt) + 1):
t = i * dt
glyco, R, F, keto, up = rates(G, K, R_override)
if abs(t - round(t)) < dt / 2 and round(t) % 6 == 0:
out[round(t)] = (G, R, F, K)
G = max(0.0, G - glyco * dt)
K = max(0.0, K + (keto - up) * dt)
return out
print(" hours | liver glycogen g | insulin/glucagon index | free fatty acids mM | blood BHB mM")
for t, (G, R, F, K) in fast(72).items():
print(f"{t:6d} | {G:16.0f} | {R:22.2f} | {F:19.2f} | {K:12.2f}")
# Same model with the insulin index held low by a very-low-carbohydrate (ketogenic) diet:
for R in (0.30, 0.20, 0.10):
K = fast(240, R_override=R)[240][3]
print(f"insulin/glucagon index held at {R:.2f} (ketogenic diet): steady blood BHB ~ {K:.2f} mM")
# hours | liver glycogen g | insulin/glucagon index | free fatty acids mM | blood BHB mM
# 0 | 90 | 1.00 | 0.35 | 0.08
# 6 | 75 | 0.75 | 0.63 | 0.03
# 12 | 60 | 0.54 | 0.88 | 0.07
# 18 | 46 | 0.36 | 1.08 | 0.17
# 24 | 33 | 0.22 | 1.24 | 0.37
# 30 | 22 | 0.12 | 1.36 | 0.69
# 36 | 13 | 0.06 | 1.44 | 1.03
# 42 | 7 | 0.02 | 1.48 | 1.27
# 48 | 3 | 0.01 | 1.49 | 1.40
# 54 | 1 | 0.00 | 1.50 | 1.46
# 60 | 1 | 0.00 | 1.50 | 1.48
# 66 | 0 | 0.00 | 1.50 | 1.49
# 72 | 0 | 0.00 | 1.50 | 1.50
# insulin/glucagon index held at 0.30 (ketogenic diet): steady blood BHB ~ 0.34 mM
# insulin/glucagon index held at 0.20 (ketogenic diet): steady blood BHB ~ 0.60 mM
# insulin/glucagon index held at 0.10 (ketogenic diet): steady blood BHB ~ 1.06 mM
In this model, after an overnight fast (12 hours) blood BHB is only about 0.07 mmol/L, hardly different from before dinner; about 0.4 at 24 hours; about 1 at 36 hours; and close to a plateau of 1.5 after 48 hours. The plateau is an artefact of the simple model; in prolonged fasting, blood BHB keeps rising for days to weeks, reaching several mmol/L (Cahill 2006). The shape reflects two things: while liver glycogen is plentiful, insulin does not fall far and ketogenesis stays inhibited; once glycogen has fallen far enough, ketone production grows faster than uptake and blood ketones rise quickly.
The model’s numbers agree with the broad ranges described in textbooks, but it must be stressed that people differ greatly. How much carbohydrate was eaten the day before, how much exercise, body size and insulin sensitivity all shift the curve markedly. Inferring your own ketone level from a population curve, or conversely treating a ketone reading as the test of whether a fast “worked”, is unreliable.
A ketogenic diet differs from fasting: carbohydrate intake stays very low, the insulin index stays low, and blood ketones settle on a plateau. In the model, an index held at 0.10–0.30 gives a steady BHB of about 0.3–1.1 mmol/L, overlapping the bottom of the range usually called “nutritional ketosis” (about 0.5–3 mmol/L).
Note
Nutritional ketosis is not ketoacidosis. In healthy people fasting or on a ketogenic diet, blood BHB usually stays below a few mmol/L, and acid–base balance is kept by residual insulin secretion together with the kidneys. Diabetic ketoacidosis occurs with severe insulin deficiency; ketones can reach well over ten mmol/L with acidification of the blood, a life-threatening emergency (Laffel 1999). People taking SGLT2 inhibitors can develop ketoacidosis without high blood glucose, which is why the list of people at risk at the start of this book names them specifically (Peters et al. 2015; U.S. Food and Drug Administration 2015).
Summary
The chemistry of the metabolic switch is the Randle cycle and hepatic ketogenesis. Per molecule of oxygen, ketones are more efficient than fatty acids but less efficient than glucose; routing fat through ketones loses about 4 % of the ATP. The time course is set by liver glycogen: an overnight fast produces almost no ketosis, blood ketones rise clearly only after 24–36 hours, and individuals differ widely. The next chapter takes up a more contested question: if ketones are not more oxygen-efficient, why do some believe they make the heart and brain work better?
References
Fan, Y. (2026). Ketone Metabolism, β-Hydroxybutyrate Signalling and the Bioenergetics of Intermittent Fasting: From the Randle Cycle and Epigenetic Inhibition to Autophagy Kinetics, Read Beside Classical Chinese Fasting Texts [酮体代谢、β-羟丁酸信号转导与间歇性禁食的生物能量学] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23753263 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).