5D Library › Insulin-Glucagon Bistability, Visceral Lipolysis Kinetics and Linear-Programming Diet Optimization · 中文
Chapter 2. The Lipase: Two Conditions, and an Overshoot
What has to happen before fat can leave a cell
Stored fat is a triglyceride, and a triglyceride cannot cross a membrane. It has to be cut, and the enzymes that cut it are held under hormonal control. Lafontan and Langin’s review sets out the sequence in human tissue: insulin activates phosphodiesterase 3B, which clears cyclic AMP; with cyclic AMP low, protein kinase A is quiet; with the kinase quiet, hormone-sensitive lipase sits unphosphorylated at its activating sites and the droplet stays shut (Lafontan and Langin 2009).
The popular summary, that falling insulin “dephosphorylates and activates” the lipase, has the direction of the main step backwards. The activating sites are phosphorylated, not dephosphorylated, and protein kinase A does it. Anthonsen and colleagues identified those sites and showed which of them govern activation (Anthonsen et al. 1998). What the popular summary is half remembering is a separate site. Garton and colleagues showed that AMP-activated protein kinase phosphorylates a different serine and that this phosphorylation restrains the enzyme rather than activating it (Garton et al. 1989). So the enzyme needs two conditions at once: the activating sites on, and the restraining site off.
Evidence
Published. Lafontan and Langin review the control of lipolysis in human adipose tissue and the place of phosphodiesterase 3B in insulin’s antilipolytic action (Lafontan and Langin 2009). Anthonsen and colleagues map the activating phosphorylation sites of hormone-sensitive lipase (Anthonsen et al. 1998). Garton and colleagues show that AMP-activated protein kinase phosphorylates a distinct site whose effect is inhibitory (Garton et al. 1989).
Two conditions that move on different timescales do something a single condition cannot. p03 gives the restraining site a slower, insulin-dependent course of its own and then steps insulin down from a post-meal level to a fasted one.
Calculation
# Chapter 2. p03: What has to happen to hormone-sensitive lipase before
# stored fat can leave a fat cell. Insulin activates phosphodiesterase
# 3B, which clears cyclic AMP, so protein kinase A is quiet and the
# enzyme's activating sites stay unphosphorylated. Falling insulin lets
# cyclic AMP rise and the kinase phosphorylate those sites. A separate
# site that the energy sensor AMP-activated protein kinase phosphorylates
# holds the enzyme off until it is dephosphorylated, so the enzyme needs
# both: the activating sites on and the restraining site off. Insulin is
# in pmol/L; the rates are scaled to a maximum of one.
# The two conditions follow a fall in insulin on two timescales. The
# activating path (phosphodiesterase, cyclic AMP, kinase) relaxes toward
# its new level with a time constant TAU_ACT; the restraining site, set
# by the slower energy sensor, relaxes with TAU_RES. Both constants are
# illustrative, not fitted; only TAU_ACT < TAU_RES matters for the shape.
CAMP0, KI, HI = 1.0, 90.0, 1.8 # cyclic AMP fall as insulin rises
KPKA, HPKA = 0.45, 2.2 # kinase response to cyclic AMP
KS565, HS565 = 55.0, 1.4 # restraining site vs insulin
TAU_ACT = 11.0 # minutes: activating path follows
TAU_RES = 40.0 # minutes: restraining site follows
FED, FASTED = 350.0, 35.0 # insulin after a meal and when fasted
def camp(ins):
return CAMP0 / (1.0 + (ins / KI) ** HI)
def pka(c):
return c ** HPKA / (KPKA ** HPKA + c ** HPKA)
def free565(ins):
# share of enzyme whose restraining site is dephosphorylated
return 1.0 / (1.0 + (KS565 / max(ins, 1e-9)) ** HS565)
def active(ins):
return pka(camp(ins)) * free565(ins)
if __name__ == "__main__":
print(" insulin cyclic AMP kinase site free lipase active")
for ins in (350.0, 200.0, 120.0, 90.0, 60.0, 45.0, 35.0, 25.0):
print(" {:<10.0f}{:<13.3f}{:<9.3f}{:<12.3f}{:.3f}"
.format(ins, camp(ins), pka(camp(ins)), free565(ins),
active(ins)))
a_fed, a_fast = active(FED), active(FASTED)
print(" active lipase fed {:.4f}, fasted {:.3f}, ratio {:.1f}"
.format(a_fed, a_fast, a_fast / a_fed))
print(" insulin at which the lipase reaches half its fasted value:")
lo, hi = FASTED, FED
for _ in range(60):
mid = 0.5 * (lo + hi)
if active(mid) > 0.5 * a_fast:
lo = mid
else:
hi = mid
half = 0.5 * (lo + hi)
print(" {:.1f} pmol/L".format(half))
print(" minutes after insulin steps from fed to fasted:")
print(" minute kinase site free lipase active share of fasted")
c0, c1 = camp(FED), camp(FASTED)
s0, s1 = free565(FED), free565(FASTED)
def course(t):
e1 = 2.718281828459045 ** (-t / TAU_ACT)
e2 = 2.718281828459045 ** (-t / TAU_RES)
k = pka(c1 + (c0 - c1) * e1)
s = s1 + (s0 - s1) * e2
return k, s, k * s
for t in (0, 10, 20, 30, 45, 60, 90, 120, 180):
k, s, a = course(t)
print(" {:<9d}{:<9.3f}{:<12.3f}{:<16.3f}{:.3f}"
.format(t, k, s, a, a / a_fast))
peak_t = max(range(0, 301), key=lambda t: course(t)[2])
print(" peak at minute {:d}: {:.3f}, {:.3f} times the fasted value"
.format(peak_t, course(peak_t)[2], course(peak_t)[2] / a_fast))
print(" one time constant for both paths: largest share {:.3f}"
.format(max(
pka(c1 + (c0 - c1) * 2.718281828459045 ** (-t / TAU_ACT)) *
(s1 + (s0 - s1) * 2.718281828459045 ** (-t / TAU_ACT))
for t in range(0, 301)) / a_fast))
# insulin cyclic AMP kinase site free lipase active
# 350 0.080 0.022 0.930 0.020
# 200 0.192 0.133 0.859 0.114
# 120 0.373 0.399 0.749 0.299
# 90 0.500 0.558 0.666 0.371
# 60 0.675 0.709 0.530 0.376
# 45 0.777 0.769 0.430 0.331
# 35 0.846 0.800 0.347 0.278
# 25 0.909 0.825 0.249 0.205
# active lipase fed 0.0203, fasted 0.278, ratio 13.7
# insulin at which the lipase reaches half its fasted value:
# 184.9 pmol/L
# minutes after insulin steps from fed to fasted:
# minute kinase site free lipase active share of fasted
# 0 0.022 0.930 0.020 0.073
# 10 0.596 0.801 0.478 1.720
# 20 0.738 0.701 0.517 1.864
# 30 0.778 0.622 0.484 1.744
# 45 0.795 0.536 0.426 1.536
# 60 0.799 0.477 0.381 1.373
# 90 0.800 0.408 0.327 1.177
# 120 0.800 0.376 0.301 1.084
# 180 0.800 0.353 0.283 1.019
# peak at minute 17: 0.520, 1.874 times the fasted value
# one time constant for both paths: largest share 1.258Read the first table down the insulin column. Cyclic AMP rises from 0.080 at an insulin of 350 to 0.909 at 25, and the kinase follows it from 0.022 to 0.825, as the textbook account requires. But the restraining site is not free at the bottom: the share of enzyme with that site clear falls from 0.930 to 0.249 as insulin falls, because the energy sensor that phosphorylates it is itself activated by fasting. The product of the two, which is the active enzyme, is therefore not monotonic. It peaks near an insulin of 60, where it reaches 0.376, and it is lower at 25, where it is 0.205.
The time course makes the consequence visible, and here the two timescales matter. In p03 the activating path follows a step down in insulin with a time constant of 11 minutes and the restraining site follows with one of 40 minutes; both values are illustrative, and only their order matters. Ten minutes after the step the kinase is already at 0.596 while 0.801 of the enzyme still has its restraining site clear, so active lipase is 1.720 times its eventual steady fasted value. It peaks at minute 17, at 1.874 times that value; it is still 1.744 times it at thirty minutes, 1.373 at sixty and 1.084 at 120, and it settles back as the restraining site catches up. With a single time constant for both paths the largest share would be only 1.258, the static peak of the first table, so most of the overshoot comes from the lag between the two paths. The enzyme overshoots and then gives some of it back. The fasted state is not the state of maximum lipolytic capacity; the transition into it is.
Hypothesis · not established
The author’s interpretation (not yet tested): because capacity overshoots during the transition and settles lower, a day built from several moderate insulin troughs can mobilise more than a day built from one deep and prolonged trough at the same mean insulin. What would show it wrong: if, at matched twenty-four hour mean insulin and matched energy intake, glycerol appearance measured by tracer does not differ between a several-trough and a single-trough day, or is higher in the single-trough day, the overshoot has no whole-body consequence.
Why the abdomen gives way first
The fat inside the abdomen and the fat under the skin are not the same tissue. Bjorntorp’s argument was anatomical before it was biochemical: the visceral depot drains into the portal vein, so whatever it releases arrives at the liver first and at a concentration nothing else sees (Bjorntorp 1990). It also differs in its receptors. It carries more of the machinery that switches lipolysis on and responds less to the insulin that switches it off, which means it is the depot insulin cannot silence. Arner and colleagues, dating the lipid in human fat cells by radiocarbon, showed in addition that the turnover of stored lipid differs between people and depots rather than being a constant of the tissue (Arner et al. 2011).
Calculation
# Chapter 2. p04: Why the visceral depot gives way first. Two depots are
# modelled: the fat inside the abdomen, drained by the portal vein, and
# the fat under the skin. The visceral depot carries more of the
# catecholamine receptors that switch lipolysis on and fewer of the
# insulin receptors that switch it off, so at the same insulin level it
# releases more fatty acids per kilogram. Masses are illustrative for an
# adult with a large waist; the release rates are scaled so that the
# subcutaneous depot releases one unit per kilogram at fasted insulin.
MASS = {"visceral": 2.2, "subcutaneous": 12.0} # kg of fat
VMAX = {"visceral": 2.30, "subcutaneous": 1.00} # per kg, insulin free
IC50 = {"visceral": 110.0, "subcutaneous": 55.0} # pmol/L, half block
HILL = 1.6
PER_UNIT = 0.84 # g of fatty acid an hour per unit of release
def rate(depot, ins):
return VMAX[depot] / (1.0 + (ins / IC50[depot]) ** HILL)
def release(depot, ins):
return MASS[depot] * rate(depot, ins)
if __name__ == "__main__":
print(" insulin visceral subcutaneous visceral share")
for ins in (350.0, 200.0, 120.0, 90.0, 60.0, 45.0, 35.0, 25.0):
v, s = release("visceral", ins), release("subcutaneous", ins)
print(" {:<10.0f}{:<11.3f}{:<15.3f}{:.3f}"
.format(ins, v, s, v / (v + s)))
v35 = release("visceral", 35.0)
s35 = release("subcutaneous", 35.0)
v350 = release("visceral", 350.0)
s350 = release("subcutaneous", 350.0)
print(" visceral share at fasted insulin {:.3f}, after a meal {:.3f}"
.format(v35 / (v35 + s35), v350 / (v350 + s350)))
share_mass = (MASS["visceral"]
/ (MASS["visceral"] + MASS["subcutaneous"]))
print(" visceral mass is {:.3f} of the two depots together"
.format(share_mass))
print(" times its share of mass: fasted {:.2f}, after a meal {:.2f}"
.format((v35 / (v35 + s35)) / share_mass,
(v350 / (v350 + s350)) / share_mass))
print(" grams of fat released a day at a steady insulin level:")
print(" insulin visceral subcutaneous together")
for ins in (90.0, 60.0, 45.0, 35.0):
v = release("visceral", ins) * PER_UNIT * 24.0
s = release("subcutaneous", ins) * PER_UNIT * 24.0
print(" {:<10.0f}{:<11.1f}{:<15.1f}{:.1f}"
.format(ins, v, s, v + s))
print(" release is not loss: most of it returns to the depot unless")
print(" the day ends in an energy deficit, which p05 and p06 set out")
# insulin visceral subcutaneous visceral share
# 350 0.686 0.591 0.537
# 200 1.405 1.350 0.510
# 120 2.354 2.676 0.468
# 90 2.933 3.751 0.439
# 60 3.669 5.583 0.397
# 45 4.083 6.955 0.370
# 35 4.362 8.080 0.351
# 25 4.628 9.351 0.331
# visceral share at fasted insulin 0.351, after a meal 0.537
# visceral mass is 0.155 of the two depots together
# times its share of mass: fasted 2.26, after a meal 3.47
# grams of fat released a day at a steady insulin level:
# insulin visceral subcutaneous together
# 90 59.1 75.6 134.7
# 60 74.0 112.6 186.5
# 45 82.3 140.2 222.5
# 35 87.9 162.9 250.8
# release is not loss: most of it returns to the depot unless
# the day ends in an energy deficit, which p05 and p06 set outIn this fit the visceral depot is 0.155 of the mass of the two depots together, but it never contributes that little. At fasted insulin its share of release is 0.351, which is 2.26 times its share of mass; straight after a meal, at an insulin of 350, its share rises to 0.537, which is 3.47 times its share of mass. The direction is worth dwelling on, because it is the opposite of the way the claim is usually told. The visceral depot’s relative advantage is largest when insulin is high, not when it is low. When insulin is low, every depot is releasing near its maximum and the larger mass under the skin dominates the total: at an insulin of 35 the two release 87.9 and 162.9 grams a day, and together 250.8.
The last two lines of p04 are the ones to carry forward. Release is not loss. Fatty acids that are not oxidised are taken back up and re-esterified, and at the end of a day in energy balance the depot is where it started. What a deficit fixes is the total lost; what the pattern of eating can change is which depot it comes out of, and by how much. Chapter 3 computes that.
Summary
The lipase needs its activating sites phosphorylated by protein kinase A and its restraining site dephosphorylated, so the popular one-line summary is half right in a way that matters: the two conditions move on different timescales, and the enzyme therefore overshoots to 1.874 times its steady fasted value 17 minutes after insulin falls. The visceral depot gives up 2.26 times its share of mass when fasted and 3.47 times just after a meal, so its advantage is largest at high insulin, which is the reverse of the usual telling.
References
Fan, Y. (2026). Insulin-Glucagon Bistability, Visceral Lipolysis Kinetics and Linear-Programming Diet Optimization: Hepatic Glycogen Depletion Times, Triglyceride-Glucose Indices and Micronutrient-Constrained Menus [胰岛素胰高血糖素双稳态、内脏脂解动力学与线性规划配餐] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090248 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).