Hepatic Glucose Production and Insulin-Mediated Suppression

Why Separate Hepatic Production Into Two Fluxes?

Hepatic glucose production is not a single biochemical route. It includes glucose released through glycogenolysis, the breakdown of stored glycogen, and glucose formed through gluconeogenesis, the synthesis of glucose from non-carbohydrate precursors. Both contribute to glucose entering the circulation, but their contributions need not respond equally to insulin or glucagon. Treating them as distinct fluxes gives a model room to represent that difference.

In an organ-resolved glucose mass balance, hepatic production is a source term for circulating glucose. Here, the focus is the internal decomposition of that source. Define $J_{\mathrm{gly}}$ as the hepatic glucose-equivalent flux from glycogenolysis and $J_{\mathrm{gng}}$ as the flux from gluconeogenesis. Their sum is total hepatic glucose production, $J_{\mathrm{hep}}$. Fluxes must use the same units and refer to the same time interval and hepatic boundary.

Definition: $J_{\mathrm{gly}}$ and $J_{\mathrm{gng}}$ are nonnegative hepatic glucose-production rates attributable to glycogenolysis and gluconeogenesis, respectively. Total hepatic glucose production is $$ J_{\mathrm{hep}}=J_{\mathrm{gly}}+J_{\mathrm{gng}}. $$ For example, both component fluxes could be reported in mmol glucose/min. If rates are normalized by body mass, both must use the same normalization.

This accounting assumes that each glucose equivalent is assigned once to its production route. A model should not count the same carbon flow as both glycogen-derived and newly synthesized glucose. In practice, component contributions may require tracer measurements or other independent information; a measurement of total output alone does not automatically reveal the two parts.

A Bounded Hormone-Response Model

Let $I$ denote the effective insulin concentration and $A$ the effective glucagon concentration available to the model. Both are nonnegative and must be expressed in units consistent with their corresponding half-maximal concentration parameters. A Hill function maps each concentration to a dimensionless response between zero and one:

S(X;K,h)=\frac{X^h}{K^h+X^h}

Here $K>0$ is the concentration producing half of the maximum response, and $h>0$ is the Hill coefficient controlling the steepness of the concentration-response curve. Thus, $S(0;K,h)=0$, $S(K;K,h)=1/2$, and $S(X;K,h)$ approaches one as $X$ becomes large.

One compact component-resolved model is

J_k=C_k\left[1+\alpha_k S(A;K_{A,k},h_{A,k})\right] \left[1-E_{I,k}S(I;K_{I,k},h_{I,k})\right], \qquad k\in\{\mathrm{gly},\mathrm{gng}\}.

For each pathway $k$, $C_k$ is a positive reference production capacity in flux units; $\alpha_k\geq 0$ is the maximum fractional glucagon stimulation in this formulation; and $E_{I,k}$, with $0\leq E_{I,k}<1$, is the maximum fractional insulin suppression. The $K$ parameters have concentration units, and the Hill coefficients are dimensionless. The product form assumes that glucagon stimulation and insulin suppression act as separable modifiers of the pathway-specific rate. It is a modeling choice, not a claim that hepatic signaling is literally independent.

At zero concentrations of both hormones, $J_k=C_k$. At high insulin, the insulin multiplier tends to $1-E_{I,k}$, so production remains nonnegative. At high glucagon, the glucagon multiplier tends to $1+\alpha_k$. Parameters may differ by pathway: a shared hormone concentration does not require shared sensitivity or maximum effect.

What the Model Implies

The model has useful directional properties, provided its parameters obey the stated bounds. Since $S$ increases from zero to one, a nonnegative $\alpha_k$ makes the glucagon multiplier nondecreasing in $A$. Since $E_{I,k}$ lies between zero and one, increasing $I$ reduces the insulin multiplier while keeping it positive. These implications can be checked directly rather than assumed from the equation’s appearance.

\frac{\partial S(X;K,h)}{\partial X} =\frac{hK^hX^{h-1}}{(K^h+X^h)^2}>0 \quad (X>0)

For either pathway, the insulin derivative is

\frac{\partial J_k}{\partial I} =-C_k\left[1+\alpha_k S(A;K_{A,k},h_{A,k})\right] E_{I,k}\frac{\partial S(I;K_{I,k},h_{I,k})}{\partial I} \leq 0.

The inequality is strict when $E_{I,k}>0$ and $I>0$. Similarly, the glucagon derivative is nonnegative when $\alpha_k\geq0$. Total hepatic production inherits these directions because it is the sum of the two component rates. This is a model-based directional statement; real data can depart from it if relevant mechanisms, time delays, or changing physiological conditions are omitted.

Worked Example: Calculating Both Fluxes at One Hormone State

Consider hypothetical parameters in mmol glucose/min, with hormone concentrations in nM. For glycogenolysis, let $C_{\mathrm{gly}}=4$, $\alpha_{\mathrm{gly}}=0.50$, $K_{A,\mathrm{gly}}=1$, $E_{I,\mathrm{gly}}=0.80$, and $K_{I,\mathrm{gly}}=2$. For gluconeogenesis, let $C_{\mathrm{gng}}=6$, $\alpha_{\mathrm{gng}}=0.30$, $K_{A,\mathrm{gng}}=1$, $E_{I,\mathrm{gng}}=0.50$, and $K_{I,\mathrm{gng}}=2$. Set all Hill coefficients to one, and suppose $A=2$ nM and $I=3$ nM.

The two response fractions are $S(A;1,1)=2/(1+2)=0.6667$ and $S(I;2,1)=3/(2+3)=0.6000$. Therefore, glycogenolysis is

J_{\mathrm{gly}} =4[1+(0.50)(0.6667)][1-(0.80)(0.6000)] =4(1.33335)(0.5200) =2.773\ \mathrm{mmol/min}.

For gluconeogenesis,

J_{\mathrm{gng}} =6[1+(0.30)(0.6667)][1-(0.50)(0.6000)] =6(1.20001)(0.7000) =5.040\ \mathrm{mmol/min}.

Thus total hepatic production is $2.773+5.040=7.813$ mmol/min. The components are not interchangeable: gluconeogenesis contributes more than glycogenolysis in this hypothetical state, even though the same hormone concentrations act on both.

Interpreting Changes in the Components

A change in total production can conceal very different changes in the two routes. For component $k$, the fractional suppression between a baseline state and a later state is $1-J_{k,\mathrm{later}}/J_{k,\mathrm{baseline}}$, provided the baseline flux is positive. Total suppression is instead calculated from the total rates. It is a baseline-flux-weighted combination of component suppressions, not generally their simple average.

1-\frac{J_{\mathrm{hep,later}}}{J_{\mathrm{hep,baseline}}} = \frac{J_{\mathrm{gly,baseline}}}{J_{\mathrm{hep,baseline}}} \left(1-\frac{J_{\mathrm{gly,later}}}{J_{\mathrm{gly,baseline}}}\right) + \frac{J_{\mathrm{gng,baseline}}}{J_{\mathrm{hep,baseline}}} \left(1-\frac{J_{\mathrm{gng,later}}}{J_{\mathrm{gng,baseline}}}\right).

This identity follows by substituting $J_{\mathrm{hep}}=J_{\mathrm{gly}}+J_{\mathrm{gng}}$ in the left-hand side and separating the two baseline terms. The weights are the component shares of baseline total production; they are nonnegative and sum to one.

Worked Example: Why Total Suppression Is a Weighted Result

Suppose hypothetical component-resolved measurements show baseline glycogenolysis of 3.0 mmol/min and gluconeogenesis of 5.0 mmol/min. During an insulin challenge, the respective rates are 1.5 and 4.0 mmol/min.

Baseline total is $3.0+5.0=8.0$ mmol/min; later total is $1.5+4.0=5.5$ mmol/min. The total fractional suppression is

\frac{8.0-5.5}{8.0}=0.3125=31.25\%.

Glycogenolysis falls by $(3.0-1.5)/3.0=50.0\%$, while gluconeogenesis falls by $(5.0-4.0)/5.0=20.0\%$. The baseline shares are $3/8=0.375$ and $5/8=0.625$. The weighted result is $(0.375)(50.0\%)+(0.625)(20.0\%)=31.25\%$, agreeing with the direct total calculation. This example describes a hypothetical observation; it does not establish that insulin caused the changes without a suitable study design.

Worked Example: Glucagon Stimulation Under Insulin Suppression

Use the parameters from the first worked example, but set $I=1.5$ nM and compare $A=0.5$ nM with $A=2.0$ nM. The insulin response is $S(1.5;2,1)=1.5/3.5=0.4286$. Thus, the insulin multipliers are $1-(0.80)(0.4286)=0.6571$ for glycogenolysis and $1-(0.50)(0.4286)=0.7857$ for gluconeogenesis.

At $A=0.5$, $S(A;1,1)=0.3333$. The predicted rates are

J_{\mathrm{gly}}=4[1+(0.50)(0.3333)](0.6571)=3.067\ \mathrm{mmol/min}, $$ $$ J_{\mathrm{gng}}=6[1+(0.30)(0.3333)](0.7857)=5.186\ \mathrm{mmol/min}.

Total production is $8.252$ mmol/min. At $A=2.0$, the glucagon response is $0.6667$, giving

J_{\mathrm{gly}}=4[1+(0.50)(0.6667)](0.6571)=3.505\ \mathrm{mmol/min}, $$ $$ J_{\mathrm{gng}}=6[1+(0.30)(0.6667)](0.7857)=5.657\ \mathrm{mmol/min}.

Total production is $9.162$ mmol/min, an increase of $0.910$ mmol/min, or about $11.0\%$, relative to the lower-glucagon prediction. Insulin still suppresses both rates relative to their glucagon-modified values without insulin suppression; the glucagon increase raises production within this model. This calculation assumes the same parameters and effective insulin concentration at both states.

Assumptions, Diagnostics, and Alternatives

The equations are most defensible over a defined physiological and experimental range. They assume that effective hormone concentrations adequately represent hepatic exposure, that the component capacities and response parameters are fixed over the modeled interval, and that an instantaneous response is adequate. The last assumption may be unsuitable when the data span different time scales: acute changes in pathway flux and slower changes in enzyme expression need not track the same concentration-response curve. Glucagon stimulation of gluconeogenesis can depend on substrate availability and duration, so its fitted effect should not be interpreted as a universal constant.

Other important limitations follow from the model structure. It does not explicitly track glycogen stores, precursor pools, or glucose cycling. If a simulation is extended over a period in which substrate availability changes substantially, constant capacities can predict implausible sustained fluxes. The model also omits interactions beyond its multiplicative form, such as signaling cross-talk or other regulators. These are reasons to refine or constrain the model when the data require it, not reasons to assume that every added mechanism is identifiable.

A central diagnostic is whether the available observations can distinguish $J_{\mathrm{gly}}$ from $J_{\mathrm{gng}}$. If only $J_{\mathrm{hep}}$ is measured, many pairs of component fluxes can have the same sum. Separate pathway effects may then be weakly identified even when the total-output fit looks excellent. Tracer-based estimates, independent pathway measurements, or informative constraints can help, but their measurement model and uncertainty also matter. When fitting, inspect component predictions as well as total output; check that rates remain nonnegative and parameter estimates remain within scientifically plausible bounds; and examine whether different parameter combinations produce nearly indistinguishable predictions.

The main alternatives answer different questions. A single total-output model can be adequate when only total hepatic production matters, but it cannot attribute a change to either route. An additive hormone model can represent opposing effects as separate terms, but without constraints it may predict negative production at extreme concentrations. The bounded multiplicative model used here preserves nonnegative rates and represents pathway-specific modulation compactly, but it imposes a particular interaction structure. Model choice should reflect the intended interpretation, measurement design, and concentration range rather than convenience alone.

Common Pitfall

Common mistake: “Total hepatic production fell by 30%, so glycogenolysis and gluconeogenesis each fell by 30%.” The conclusion does not follow from a total measurement. The total change is weighted by the baseline contributions, and the component changes may differ substantially. Use component-specific evidence to describe component-specific responses; otherwise report only the total change and acknowledge that the decomposition is not identified.

A related caution is to distinguish a fitted model effect from a causal conclusion. The equations say how predicted rates vary with their hormone inputs under the model assumptions. A causal interpretation additionally depends on the design, relevant confounding, and whether the model adequately represents the intervention and biological system.

Check Your Understanding

Answer each question, then compare with the brief solutions.

If $J_{\mathrm{gly}}=2.4$ mmol/min and $J_{\mathrm{gng}}=6.1$ mmol/min, what is total hepatic glucose production?
For a component with $C=5$ mmol/min, $\alpha=0.40$, $E_I=0.50$, and both response fractions equal to $0.50$, calculate the predicted rate.
Baseline total production is 10 mmol/min, comprising 4 mmol/min glycogenolysis and 6 mmol/min gluconeogenesis. Later component rates are 2 and 4.5 mmol/min. Calculate total fractional suppression and explain why it is not the unweighted average of component suppressions.
Why can total hepatic output data alone fail to identify pathway-specific insulin sensitivities?

Solutions. (1) $2.4+6.1=8.5$ mmol/min. (2) $5[1+(0.40)(0.50)][1-(0.50)(0.50)]=5(1.20)(0.75)=4.50$ mmol/min. (3) Later total is $2+4.5=6.5$ mmol/min, so suppression is $(10-6.5)/10=35.0\%$. The component suppressions are $50.0\%$ and $25.0\%$; weighting by baseline shares gives $(4/10)(50\%)+(6/10)(25\%)=35.0\%$, not their unweighted mean of $37.5\%$. (4) The measured total supplies only the sum of the two rates. Different pairs of pathway fluxes and sensitivities can produce the same total, so additional component information or constraints are needed.