What a recovered output can hide

A steady flux can survive damage. A protein level can conceal several mechanisms. The Week 2 submissions turn reaction-network bookkeeping into a guide for deciding what to measure next.

CCBS · EST 5053 · Lecture 3 audience extension
Editorial synthesis of six Week 2 student submissions · prepared with Codex · 15 September 2026

A system can return to its old output without returning to its old condition. An enzyme pool can be permanently depleted while product formation recovers. A protein concentration can remain constant while production and removal both change. The observation is real. The interpretation needs another measurement.

Lecture 3 made a cell calculable by separating reaction changes, rate laws and conserved pools. The students' extensions ask what those equations allow an observer to infer. Their common pressure point is the difference between a balance and the machinery that maintains it.

We will develop that distinction in three steps. A finite enzyme pool gives an exact recovery-without-repair example. Energy and gene-expression models show why the problem survives a change of biological setting. Growth and protein burden then tell us when changing a steady level does, and does not, change the system's repertoire.

Student names link to the accepted versions used here. The connections, qualifications and new calculations are editorial synthesis, not quotations or a reconstructed classroom conversation. Read beside the core, exposition and lineage. These illustrative models are not fits to experimental data.

1A rate curve contains a capacity

A saturating rate law carries information about a finite resource. Yibing Li makes that resource explicit by recovering the enzyme complex hidden inside Michaelis–Menten kinetics.

Let $E$ and $S$ denote free enzyme and substrate concentrations, and $C$ the enzyme–substrate complex. The resolved elementary steps are

The active enzyme total is $q_E=E+C$. Each occupied enzyme contributes to product flux, so $v=k_{\mathrm{cat}}C$. Since $C\le q_E$, the exact capacity bound is $v\le k_{\mathrm{cat}}q_E$, before making any quasi-steady approximation.

If the complex relaxes sufficiently quickly, setting its leading-order rate of change to zero gives

Here $S$ is free substrate, not the total $q_S=S+C$. A valid reduction also needs an appropriate separation of timescales. The algebra alone does not supply that permission.

The distinction between $K_M$ and a dissociation constant already mattered to Briggs and Haldane in 1925. Their complex can disappear by catalysis as well as dissociation. A quasi-steady complex therefore need not be at binding equilibrium. Their paper also notes that the same overall progress curve need not determine how the two exit rates are divided.¹

Yibing's finite-pool argument gives us a useful diagnostic. More substrate can occupy unused enzyme, but it cannot create more enzyme. The next submission asks what happens when part of that capacity is removed permanently.

Separate the exact enzyme-capacity bound from the approximate rate law, and distinguish free substrate from its conserved or accumulated total.

2Recovery without repair

A damaged enzyme pool can regain its original throughput by working at a higher substrate concentration. Chengwei Jin constructs this counterexample with a finite inhibitor pulse and a constant substrate inlet.

The inhibitor concentration $H(t)$ is externally maintained during the pulse and washed out afterward. Inactivation removes active free and bound enzyme at the same rate per molecule. Inactivating a bound enzyme releases its substrate. No new active enzyme is synthesized.

Define the inactivation rate $\kappa(t)=k_{\mathrm{inact}}H(t)$ and the constant inlet flux $v_{\mathrm{in}}$. Those assumptions give the full balance model

The total-substrate equation is the important bookkeeping step. Binding changes free substrate but not $q_S$. The inlet adds substrate and catalysis removes it. This remains true during the inhibitor pulse because bound-substrate release was included in the model.

For a rectangular pulse of amplitude $H_0$ and duration $T$, the surviving enzyme fraction is exactly $a=\exp(-k_{\mathrm{inact}}H_0T)$. Washing out inhibitor stops further loss. It does not restore the lost enzyme.

Chengwei chooses concentrations in µM and time in minutes. The parameters are $k_+=4\,\mu\mathrm{M}^{-1}\mathrm{min}^{-1}$, $k_-=k_{\mathrm{cat}}=2\,\mathrm{min}^{-1}$, $k_{\mathrm{inact}}=0.1\,\mu\mathrm{M}^{-1}\mathrm{min}^{-1}$ and $v_{\mathrm{in}}=1\,\mu\mathrm{M}\,\mathrm{min}^{-1}$. Initially $q_E=1$, $C=0.5$ and $q_S=1.5\,\mu\mathrm{M}$.

The mild pulse causes a permanent molecular loss and a transient output loss. A product assay taken sufficiently long after washout would report recovery. An enzyme assay would report damage. Both assays would be correct because they measure different quantities.

Use a conserved-total balance to distinguish restored throughput from restored molecular capacity after a perturbation.

3The price of keeping output fixed

Flux recovery is paid for by substrate accumulation. At a finite post-washout steady state, the substrate balance forces $k_{\mathrm{cat}}C^*=v_{\mathrm{in}}$. The complex balance then determines how much free substrate the surviving enzyme needs:

A positive finite solution requires $k_{\mathrm{cat}}q_E^\infty>v_{\mathrm{in}}$. This condition is exact for the steady states of the full model. The same formula that appeared as an approximate dynamical closure in §1 follows here from exact steady-state balances.

The condition also has a direct physical meaning. The surviving enzymes must be able to process substrate faster than it arrives when sufficiently occupied. At equality, full occupancy would be required at finite substrate, which this binding law cannot achieve. Below equality, incoming substrate outpaces even the maximum catalytic throughput.

The finite equilibrium is locally attracting within the post-washout enzyme slice. In coordinates $(q_S,C)$, the Jacobian has negative trace and determinant $k_{\mathrm{cat}}k_+E^*>0$. The simulations approach that equilibrium after the mild pulse. For the severe pulse, substrate instead accumulates without a finite steady state in this idealized constant-inlet model.

The recovery mechanism resembles output adaptation, but no dedicated sensor or controller was added. Accumulation increases enzyme occupancy until the inlet is matched. The output alone cannot tell us whether a regulatory feedback circuit restored function or a material balance forced the remaining machinery to work harder.

This gives an immediate experimental discriminator: measure the accumulating pool as well as the delivered flux. Near the capacity boundary, that pool is much more sensitive to enzyme loss than the final output is.

Derive the capacity threshold and the hidden pool cost before interpreting constant throughput as evidence of repair or dedicated regulation.

4More fuel meets another ceiling

Increasing supply can move the bottleneck without eliminating it. Shunkang Xue makes an auxiliary energy module explicit instead of treating ATP as an inexhaustible background condition.

Let $A$ denote ATP and $D$ ADP, with a fixed total $q_A=A+D=1$ in chosen concentration units. Use effective regeneration, work and leak fluxes

The regeneration capacity $J$ is a parameter, not the realized regeneration flux. Fuel-driven ATP regeneration is a composite process. This equation does not resolve its chemical mechanism or establish a thermodynamic efficiency. The leak is assumed to return ATP to the ADP pool, so the stated total remains conserved.

The net rate decreases strictly with $A$. More ATP means less ADP for regeneration and more substrate for work and leak. The endpoints direct motion into $[0,1]$, so every positive $J$ gives one attracting ATP level. Raising $J$ moves that level. It does not create bistability in this model.

Read alongside Chengwei's pulse, this calculation changes how we interpret a plateau. One plateau is imposed by a fixed inlet. Another is imposed by a downstream enzyme and its finite substrate pool. Both look like an output that stopped changing. Their useful interventions are different.

Distinguish a supply-capacity parameter from a realized flux, then combine all finite-pool bounds before identifying a throughput ceiling.

5One level, several explanations

A steady abundance generally does not identify which kinetic parameter changed. Fei Wang raises this problem through gene expression and the distinction between protein abundance and functional activity.

Let $M$ and $P$ be mRNA and protein concentrations. Let $f_{\mathrm{tx}}$ be the transcription flux, $k_{\mathrm{tl}}$ the translation rate per mRNA, $\delta_M,\delta_P$ the respective degradation constants, and $\lambda$ the specific growth rate. At constant parameters,

The measured protein level constrains a combination of parameters. Doubling transcription and halving translation leaves this steady protein level unchanged while doubling steady mRNA. Multiplying every rate by the same positive factor leaves both steady concentrations unchanged while changing every relaxation time. These are exact ambiguities of the declared model.

A second observable can separate some ambiguities. mRNA measurements distinguish the first example. A time course can distinguish the second. Neither automatically determines protein activity, which can depend on occupancy, modification, localization or another state not present in the two equations.

Fei's disease-motivated example is useful as an inference question, not as a fitted disease model. We do not infer a cancer mechanism from arbitrary slider settings. The justified conclusion is narrower: the same abundance can arise from different parameter combinations, and abundance is not itself an activity assay.

Growth also affects more than the dilution term in real bacteria. Klumpp, Zhang and Hwa combined measurements and models of growth-dependent expression machinery, gene dosage and other cellular parameters. Their analysis is a warning against identifying every growth-associated expression change with dilution alone.²

Find parameter combinations that preserve an observed steady state, then choose another observable or time course that makes those explanations distinguishable.

6The host changes the balance

A growth rate can be a dynamical consequence rather than an external constant. Haobo Zheng couples protein expression to a resource pool and asks how a nutrient pulse changes the response.

Let $x$ be protein, $r$ an effective resource concentration and $u$ the external resource setting. In model units, take

The protein consumes resource in the second equation. The resource sets growth in the third. Growth then removes protein concentration in the first. This closes a feedback path through the host even though the production flux stays constant.

At a positive steady state, the first equation gives $x^*=0.8/\lambda(r^*)$. Substituting into the resource balance eliminates the protein:

This positive solution requires $u>0.07333$. At $u=1$, the steady protein is 1.0432. At a sustained $u=1.8$, it is 0.8687. In this chosen model, a richer environment lowers protein concentration because faster dilution outweighs the unchanged production flux.

Haobo's comparison reveals a useful model audit. Before celebrating an invariant output, trace how the tested perturbation enters the equations. If the perturbation changes no term affecting that output, invariance may be built into the comparison by omission.

Trace the path from an environmental input to the measured variable before interpreting a flat response as robustness.

7A burden is not yet a switch

Changing a removal law does not automatically create multiple stable states. Runzi Yan brings protein crowding and aggregation into the addition–removal picture. The useful next question is whether those changes move a level or create a qualitatively new behavior.

For an illustrative soluble-protein variable $x\ge0$, define production $p(x)$ and removal $r(x)$ by

The production law decreases and the removal law increases. Therefore $p-r$ decreases strictly. It is positive at zero and negative for large $x$, so it has exactly one root. Flow points toward that root on both sides. No second stable state is hidden between the plotted samples.

The biological interpretation also depends on what $x$ counts. Removing soluble protein into aggregates can lower soluble abundance while increasing aggregate mass. A one-pool model cannot report both quantities. To ask about persistent aggregate burden, we would need an aggregate state and its formation and clearance balances.

This is the bridge to Lecture 4's audience chapter. A feedback path can make net removal nonmonotone and permit three crossings. The added mechanism, not the word “burden,” must do that mathematical work.

Use monotonicity to test whether an altered rate law changes only a steady level or can create multiple states, and keep soluble and aggregate pools distinct.

8Make the explanations disagree

The next experiment should separate explanations that already agree with the first observation. That is the common outcome of these six submissions.

The enzyme example says to measure accumulated substrate beside recovered product flux. The energy example says to measure the carrier pool and the realized work flux beside the nominal supply capacity. The expression example says to pair protein abundance with mRNA, activity or response time according to the ambiguity being tested.

The growth and burden examples add a model-side requirement. Make the input act through an explicit path. Name which pool the assay reads. Then ask whether the claimed behavior follows from conservation, monotonicity, an approximation, a numerical run or an actual measurement.

These distinctions sharpen rather than weaken the reaction-network description. Its balances identify what an observation must satisfy. Its hidden states identify what the observation leaves undecided. A recovered output becomes scientifically informative when we can say what else would have to recover for the proposed explanation to be true.

Design a second measurement or perturbation that distinguishes mechanisms compatible with the same observed output.

Models and provenance

All six contributions are accepted files from the pinned Week 2 corpus. The chapter uses their questions and constructions, with editorial connections and qualifications. It does not assign grades or imply student approval of the synthesis.

The instructor's local course project stores the input checksums, contribution ledger, independent Python calculations and figure data under lectures/research/week02-extension-synthesis/. The pulse calculation integrates each constant-forcing interval separately. All parameter values needed to reproduce the central examples are stated above. No student script was executed and no external script is needed to read this page.

References

1. G. E. Briggs and J. B. S. Haldane, “A note on the kinetics of enzyme action,” Biochemical Journal 19, 338–339 (1925). DOI. The full note distinguishes the steady-complex approximation from binding equilibrium.
2. S. Klumpp, Z. Zhang and T. Hwa, “Growth rate-dependent global effects on gene expression in bacteria,” Cell 139, 1366–1375 (2009). DOI. Full-text sections on global expression parameters and growth feedback provide the biological context, not a fit for the illustrative models here.