Time-scale separation: why a quasi-steady complex keeps changing

A supplied enzyme can change its output while its complex remains nearly in balance. Recall the calculation, expose the apparent contradiction, and use singular perturbation theory to justify the reduction.

CCBS · EST 5053 · Lecture 6 · core page (full) · 2026-09-17 (Thu) 10:40–12:15 · students present, lecturer-backed

Lecture 3 gave us a reduced enzyme model that turns substrate supply into product output. Its key step was to set the complex derivative approximately to zero. Yet a change in supply changes the required complex concentration. We will rebuild that example from the beginning, identify what the approximation means, and return with a justified calculation of its changing state.

0–6 min · §1

Recall the supplied-and-drained enzyme.

6–11 min · §2

Why does a quasi-steady complex keep changing?

11–18 min · §3

Totals identify what binding cannot change.

18–26 min · §4

The general singular perturbation framework.

26–33 min · §5

Compare binding and catalytic times.

33–40 min · §6

The complex follows the changing substrate total.

40–56 min · §7

Complete the open enzyme, then recover Michaelis–Menten.

56–68 min · §8

Product competition makes the binding constraint implicit.

68–80 min · §9

Self-repression turns binding into a Hill law.

80–86 min · §9a

A fast stochastic promoter can also be eliminated.

86–91 min · §10

Assemble the common differential-algebraic model.

91–95 min · §11

A steady state must balance the slow fluxes too.

Twelve teaching points, including the short §9a bridge, on a 95-minute spoken route. Complete balances remain available for study. Detailed Jacobians, polynomial expansions and noise-moment algebra are worked in the exposition. The playground is outside the lecture clock.

1Recall the supplied-and-drained enzyme

Lecture 3 used an enzyme to convert a continuing substrate supply into an outgoing product stream. Here is the complete model again. In a well-mixed, fixed-volume vessel, free enzyme $E$ binds free substrate $S$ to form complex $C_{ES}$. Catalysis converts the bound substrate into free product $P$ and returns the enzyme.

Substrate enters at concentration flux $v_\mathrm{in}$. Product leaves at flux $k_\mathrm{out}P$. Association has rate constant $k_+$, dissociation $k_-$, and catalytic conversion $k_\mathrm{cat}$. Enzyme has no source or drain.

The totals count free and bound material: $q_E=E+C_{ES}$, $q_S=S+C_{ES}$, and $q_P=P$. Binding only redistributes each constituent. The total balances are therefore

These total equations still need the complex concentration. Lecture 3 closed the calculation by setting its formation and removal approximately in balance:

Substitute $E=q_E-C_{ES}$ and $S=q_S-C_{ES}$. The balance becomes $(q_E-C_{ES})(q_S-C_{ES})\approx K_MC_{ES}$. Its physically allowed root supplies the approximate complex at each substrate total:

Inserting this root into the total balances gives a usable two-variable model. It predicts substrate accumulation, product output and the response to a change in supply. The remaining question is why the approximation is allowed during that response.

Reconstruct the open enzyme model and identify the one approximate step that closes its exact total balances.

2Why does a quasi-steady complex keep changing?

The approximation seems to set the complex derivative to zero, yet the predicted complex changes with the input. At any steady processing state, total balance requires $C_{ES}^*=v_\mathrm{in}/k_\mathrm{cat}$. If the supply changes and another processing state is reached, the complex must change between them. Starting with no complex also requires a nonzero derivative while binding forms it.

The same puzzle appears even when supply and drain are turned off. Set $v_\mathrm{in}=k_\mathrm{out}=0$ to obtain a closed vessel. The complex rises after mixing. As substrate is converted to product, the available substrate decreases and the complex eventually falls toward zero. This simpler case lets us develop the method before returning to the driven system.

Use illustrative rates $k_+=10\,\mu\mathrm{M}^{-1}\mathrm{s}^{-1}$, $k_-=10\,\mathrm{s}^{-1}$, and $k_\mathrm{cat}=0.5\,\mathrm{s}^{-1}$. The affinity scale is $K_d=k_-/k_+=1\,\mu\mathrm{M}$. Unbinding takes $0.1\,\mathrm{s}$, while catalysis takes $2\,\mathrm{s}$. Association at free concentration $1\,\mu\mathrm{M}$ contributes a rate of $10\,\mathrm{s}^{-1}$.

We need to distinguish rapid adjustment from slow change. Historically, Briggs and Haldane already allowed the quasi-steady complex to decline after its initial transient.¹ Our task is to derive the two time scales, identify the relevant slow variables, and justify a complex that keeps following them.

Explain why a quasi-steady complex must still change during substrate consumption or a response to altered supply.

3Totals identify what binding cannot change

Define the forward and reverse binding fluxes before adding their effects:

A positive net binding flux increases complex while consuming free enzyme and substrate. A negative flux reverses those changes. In the closed vessel, all four species balances are

Binding rapidly changes both free enzyme and free substrate. Neither is automatically a slow coordinate. Add free and bound amounts instead:

Here a $q$ counts a molecular constituent across its forms. Adding the corresponding species equations cancels binding exactly:

No small parameter has been used. Binding alone preserves these totals at any rate. Catalysis preserves enzyme total but converts substrate total into product total. The closed-system sum $q_S+q_P$ is conserved. A “binding invariant” need not be constant under the whole network.

Total coordinates make the separation visible: binding redistributes species within a fixed total. Slow chemistry changes the total itself. This is why total-variable reductions can handle appreciable sequestration.²

Derive exact binding-invariant totals and distinguish them from quantities conserved by the entire reaction network.

4The general singular perturbation framework

Singular perturbation theory separates fast adjustment from slow evolution within the same dynamical system. The aim is to predict the slow motion without resolving every fast transient. We first write the general equations, then ask what remains on each time scale.

Let $x^\mathrm{slow}$ and $x^\mathrm{fast}$ collect dimensionless slow and fast variables. Their reference times are $t_\mathrm{slow}$ and $t_\mathrm{fast}$. Write their dimensionless rate functions as $f$ and $g$, with explicit dependence on the time-scale ratio:

The small parameter compares times after the state scales have been chosen. In the model family under consideration, the rate functions remain finite on the region of interest as $\varepsilon$ decreases. The ratio can shrink because the fast process accelerates or because the slow process slows down. The two enzyme limits below realize these two possibilities.

On a fast interval, the slow variables barely change. Measure elapsed time in fast units, $t/t_\mathrm{fast}$. The equations become $dx^\mathrm{slow}/d(t/t_\mathrm{fast})=\varepsilon f$ and $dx^\mathrm{fast}/d(t/t_\mathrm{fast})=g$. At leading order the slow variables stay at their initial values. The fast variables relax according to

On a slow interval, fast balance becomes a constraint. Let $h(x^\mathrm{slow})$ be a selected stationary fast state, obtained by solving

The function $h$ reconstructs the fast state from the current slow state. Substituting it into the slow equation gives the reduced dynamics:

This is a singular limit because a differential equation has become an algebraic condition. The reduced model cannot specify an independent initial fast state. The initial fast relaxation accounts for the missing initial condition.

Fast attraction is what justifies the substitution. The selected stationary state must attract nearby fast states at each retained slow state. For smooth equations, a stable fast linearization whose decay rates stay bounded away from zero supports this reduction on suitable bounded regions. Start within its basin of attraction and follow a finite slow-time interval that stays in the region. After the initial transient, sufficiently small $\varepsilon$ permits the fast variables to follow the changing stationary state.³

The reconstructed fast state can therefore keep moving: $h$ depends on the slow variables. We will now identify this general structure in the enzyme and calculate that motion explicitly.

Identify the fast relaxation problem, the reduced slow dynamics, and the attraction condition needed to connect them.

5Compare binding and catalytic times

The enzyme fits this framework with totals as the slow variables and complex concentration as the fast variable. To display the separation, measure every concentration in units of the fixed affinity scale $K_d=k_-/k_+>0$. A bar changes the unit while preserving the quantity's name:

The free concentrations in these units are $\bar E=\bar q_E-\bar C_{ES}$ and $\bar S=\bar q_S-\bar C_{ES}$. Enzyme total $\bar q_E$ is constant. We continue to use physical elapsed time $t$.

Choose reference times that identify the two processes:

The binding reference is the unbinding time. It is also the association time at free concentration $K_d$, because $k_+K_d=k_-$. The actual binding relaxation time depends on the free concentrations. With our illustrative rates, $t_\mathrm{bind}=0.1\,\mathrm{s}$, $t_\mathrm{cat}=2\,\mathrm{s}$, and $\varepsilon=0.05$.

Substitution into the exact balances gives the general slow–fast pattern without renaming any concentrations:

Here $x^\mathrm{slow}=(\bar q_S,\bar q_P)$, $x^\mathrm{fast}=\bar C_{ES}$, and $t_\mathrm{bind}=\varepsilon t_\mathrm{cat}$. The total-equation right sides form $f$. The complex-equation right side is $g$.

On a binding interval, the total changes only slightly: $d\bar q_S/d(t/t_\mathrm{bind})=-\varepsilon\bar C_{ES}$. Binding alone leaves it exactly fixed. This is the enzyme's version of the general fast problem.

Take the limit by speeding association and dissociation together at fixed $K_d$. Hold the initial totals and catalytic rate fixed. The leading slow description retains the total dynamics and replaces the complex equation by binding balance:

Put the enzyme into the general singular perturbation form while keeping every concentration and reference time physically identifiable.

6The complex follows the changing substrate total

Continue with the rapid-binding limit from Section 5. Catalysis reduces substrate total, and fast binding makes the complex follow its changing value. To establish this, we need two facts: binding restores the complex at a fixed total, and its restored value increases with substrate total.

Return to the original concentration units. Write the complex reconstruction as $h_{C_{ES}}(q_S;q_E,K_d)$. The subscript identifies the species reconstructed by the general function $h$. With enzyme total $q_E>0$ and affinity $K_d>0$ fixed, write this as $h_{C_{ES}}(q_S)$. It is the physical solution of the binding constraint:

Binding restores this value when the total is held fixed. Below it, association exceeds dissociation and the complex increases. Above it, less enzyme and substrate remain free, while more complex can dissociate. The complex decreases. This verifies the fast attraction required by the general theory.

The restored complex increases with substrate total. Differentiate the binding relation while keeping enzyme total and affinity fixed:

The denominator is free enzyme plus free substrate plus $K_d$, so it is positive. The numerator is the positive free enzyme concentration. The exposition solves the quadratic and calculates the relaxation rate.

Catalysis therefore makes the reconstructed complex decrease. After the initial binding transient, the leading slow model is

Apply the chain rule to the complex reconstructed by this reduced model:

This negative derivative answers the question from Section 2. The complex changes on catalytic time. Its change over a binding interval is smaller by $t_\mathrm{bind}/t_\mathrm{cat}$. At complete depletion, both the complex and its rate of change approach zero.

The family of binding-stationary states is called the critical manifold. At sufficiently fast finite binding, trajectories approach a nearby attracting slow manifold and then move along it. The small displacement supplies the flux imbalance needed for a nonzero derivative. The exposition derives that displacement and distinguishes the slow manifold from the curve where the full complex derivative is zero.⁴

Calculate the slow change of a rapidly adjusting complex by differentiating its dependence on the changing total.

7Complete the open enzyme, then recover Michaelis–Menten

We can now justify the calculation that opened the lecture. Recall the complete mechanism: substrate enters, binds enzyme, becomes product, and leaves. The model is

Supply and removal have separate rate laws $v_\mathrm{in}$ and $k_\mathrm{out}P$. The binding flux is $v_\mathrm{bind}=v_\mathrm{bind}^{+}-v_\mathrm{bind}^{-}=k_+ES-k_-C_{ES}$. The full species equations and their exact total balances are

Two ways to separate the times

Route A accelerates binding at fixed affinity. Set $\varepsilon_\mathrm{bind}=k_\mathrm{cat}/k_-$. Bars still mean division by $K_d$, and $t_\mathrm{cat}=1/k_\mathrm{cat}$. Including the source and drain gives

Increase $k_+$ and $k_-$ together. Keep their ratio $K_d$, the initial totals, catalytic rate and boundary rates fixed. Supply changes the total negligibly during binding relaxation. The attracting fast constraint from Section 6 gives $C_{ES}\approx h_{C_{ES}}(q_S;q_E,K_d)$ after the initial layer.

Route B slows substrate consumption by reducing enzyme abundance. Keep the microscopic constants fixed and choose a representative positive free substrate scale $S_\mathrm{ref}$. Define

The small parameter controls both sequestration and relative speed. Starting with little complex at free substrate $S_\mathrm{ref}$, the approximate complex formed is $q_ES_\mathrm{ref}/(K_M+S_\mathrm{ref})$. Its fraction of the substrate pool is $\varepsilon_\mathrm{QSSA}$. Substrate therefore changes little during complex formation when this ratio is small.⁵

Use fractional occupancy $C_{ES}/q_E$ as the fast variable, because it remains finite as enzyme decreases. Scale supply with capacity, keeping $v_\mathrm{in}/(k_\mathrm{cat}q_E)$ finite. The exact equations become

These are the general slow and fast equations from Section 4. On substrate time the fast derivative carries $t_\mathrm{complex}/t_\mathrm{substrate}$. At fixed substrate, its restoring rate is $k_+S+k_-+k_\mathrm{cat}>0$. The leading balance gives $C_{ES}/q_E=S/(K_M+S)$, even when catalysis is not slow relative to unbinding.

Keeping $S=q_S-C_{ES}$ in that balance gives $C_{ES}\approx h_{C_{ES}}(q_S;q_E,K_M)$. This total-QSSA closure is justified to leading order in the small-enzyme family just described. Its use outside that family needs its own separation check.²

The complex changes in the reduced open system

Both routes close the total equations through the same physical root, with a different constant $K$:

This completes the opening calculation. Increased supply can increase complex concentration. Depletion can decrease it. Fast adjustment keeps the complex near a relation that the changing total moves. A zero leading fast rate does not require a zero slow derivative.

Michaelis–Menten in total substrate needs one more approximation

The exact algebraic constraint contains free substrate $S=q_S-C_{ES}$. To replace free substrate by total substrate, bound substrate must occupy only a small fraction of that total. For positive substrate total, two sufficient conditions follow directly from the binding equations:

Either condition makes $q_S$ dominated by free $S$. Substitute $S\approx q_S$ into $C_{ES}=q_ES/(K+S)$ to obtain

Which species dominates a total determines the useful rate-law regime. If binding is tight and substrate is scarce relative to enzyme, much of $q_S$ can reside in complex. Then replacing $S$ by $q_S$ fails. In the tight-binding limit the complex approaches $\min(q_E,q_S)$: substrate limits occupancy on one side, enzyme on the other. The implicit quadratic covers both regimes.

The forcing and observation window belong to each approximation. Check that supply moves little substrate during fast relaxation. An abrupt input change can create a new short adjustment. If product is also treated as slow, its removal time must be on the retained scale. Here product does not feed back, so a faster product drain can instead retain its own differential equation. Product rebinding will change that conclusion in the next example.⁶

Justify the fast constraint first. A Michaelis–Menten law in total substrate additionally requires free substrate to dominate that total.

8Product competition makes the binding constraint implicit

Product can inhibit processing by occupying the enzyme that substrate needs. Add a product-bound complex $C_{EP}$, product binding and reversible conversion between the two complexes. This is a new model, so we specify its complete mechanism.

Define a net flux for each process:

The five species equations keep the two binding fluxes separate:

Count each bound constituent in its total. Binding then cancels exactly:

The product drain contains free $P=q_P-C_{EP}$. Replacing it by $q_P$ at this stage would remove bound product that the model protects.

Two fast variables and two algebraic constraints

Make both binding pairs fast while keeping their affinities $K_S=k_{-,S}/k_{+,S}$ and $K_P=k_{-,P}/k_{+,P}$ fixed. Choose $t_\mathrm{bind}=\max(k_{-,S}^{-1},k_{-,P}^{-1})$, $t_\mathrm{cat}=1/(k_\mathrm{cat}^{+}+k_\mathrm{cat}^{-})$ and $\varepsilon=t_\mathrm{bind}/t_\mathrm{cat}$. Bars in this example mean division by $K_S$. The singular perturbation form is

As both binding rates increase together, the scaled binding terms stay finite. Keep catalytic and boundary rates fixed. The two-complex fast system attracts its physical stationary state at fixed totals. Its Jacobian has negative trace and positive determinant, as derived in the exposition. The leading fast constraints are

These two equations determine the complexes needed by the slow total equations. Together they form a differential-algebraic system. We can integrate it by solving the binding constraints at each current pair $(q_S,q_P)$.

A free-enzyme variable reduces the algebraic solve to one equation. From each total and its binding constraint,

The right side is strictly increasing for $E\geq0$. There is therefore one physical solution in $[0,q_E]$. Multiplication by the denominators gives a cubic in the generic case:

An explicit cubic formula is possible, but a bounded one-dimensional solve is easier to use. Equal affinities are a special case in which the equation simplifies. The important new structure is the coupled algebraic constraint, not a long root formula.

Product-bound enzyme dominance reveals inhibition

The fast constraints show directly when product occupies most of the enzyme. Divide all enzyme states by free enzyme and use their binding relations:

If $P/K_P\gg1+S/K_S$, then $q_E\approx C_{EP}$. Little enzyme remains available for substrate. In this regime $C_{ES}\approx q_E(K_P/K_S)(S/P)$, so more free product suppresses forward activity at fixed free substrate. This is a dominance statement about the enzyme total. It does not require free product to dominate the product total.

A batch-conversion test exposes the processing cost. Pause supply and drain, set $q_S(0)=10\,\mu\mathrm M$ and $q_P(0)=0$, and vary only product-binding affinity. All three cases start with the same enzyme and substrate. The complete reversible mechanism is retained.

At half conversion, the weak, intermediate and strong cases place about 1%, 45% and 98.5% of enzyme in $C_{EP}$. Their forward activities are about 99%, 56% and 1.5% of the product-free reference at that same substrate total. These comparisons identify inhibition independently of substrate depletion or reverse catalysis.

Restoring a fixed substrate input changes the question. Any finite processing steady state must satisfy $v_\mathrm{cat}=v_\mathrm{in}$. Product inhibition can then require a much larger substrate pool to sustain the same throughput. Equal final fluxes therefore do not measure equal enzyme availability. The batch test isolates the slowing that this open-system balance can conceal.

Free-substrate and free-product dominance gives an explicit rate law

If $C_{ES}\ll q_S$ and $C_{EP}\ll q_P$, replace free $S,P$ by their totals. The enzyme is divided among three states:

Substitution gives the competitive two-input flux and its approximate total dynamics:

Product affects two different parts of this expression. Its denominator term reduces forward catalysis by occupying enzyme. Its negative numerator term represents reverse catalysis. Product binding can inhibit the forward reaction even when reverse catalysis is negligible. Figure 8 uses the full model for its trajectories and the exact fast-binding constraints for its matched comparison. It does not assume free-product dominance.

Coupled fast binding gives an implicit DAE. Free-ligand dominance can simplify its algebraic constraints into an explicit competitive rate law.

9Self-repression turns binding into a Hill law

Binding can control a production rate as well as a catalytic flux. The enzyme example used the fraction bound to substrate. Gene repression uses the fraction of promoter still available to produce. We can derive both from the same total-and-binding calculation.

A protein can reduce its own production by binding its gene's promoter. Let $D$ denote free promoter and $R$ free repressor protein. The bound promoter $C_{DR}$ produces nothing. We omit RNA and model protein production directly at rate $\rho D$. Only free protein is removed, at rate $\gamma R$.

With $v_\mathrm{bind}=k_+DR-k_-C_{DR}$, every species balance is

Binding conserves promoter and total repressor. Their definitions and slow balances are

Choose $t_\mathrm{slow}=1/\gamma$, $t_\mathrm{bind}=1/k_-$, and $\varepsilon=\gamma/k_-$. Bars now mean division by $K_d=k_-/k_+$. Speed both binding directions at fixed $K_d$, $\rho/\gamma$ and $q_D/K_d$. Then

The fast equation has the same binding form as the enzyme. Its physical stationary state attracts because its derivative with respect to complex is negative. Fast binding therefore gives the DAE

The algebraic equation defines $C_{DR}=h_{C_{DR}}(q_R;q_D,K_d)$. In terms of free protein, promoter availability is $D=q_D/(1+R/K_d)$. If bound repressor stores little of the total, $R\approx q_R$, and the reduced equation becomes the familiar degree-one repression law:

This decreasing hyperbola uses the unbound fraction, while the enzyme rate used the bound fraction. The underlying binary-binding calculation is the same. Sufficient conditions for free-protein dominance are $q_D\ll K_d$ or $q_D\ll q_R$, by the bounds from Section 7.