Slow-binding & two-step isomerization kinetics — from curved progress traces to kobs replots, pre-incubation IC50 shift, and koff residence time
The starting observation of every time-dependent inhibition (TDI) study is deceptively simple: a continuous enzyme assay whose product-formation trace bends. For a classical fast-equilibrium competitive or noncompetitive inhibitor, product accumulates linearly with time once the E+I⇌EI equilibrium is established within the assay dead time (typically <5–10 s for diffusion-limited binding). But for a growing class of pharmacologically important inhibitors — including many covalent kinase inhibitors, cathepsin K inhibitors, and slow-onset protease inhibitors — the enzyme-inhibitor complex itself evolves after formation, and the progress curve visibly curves downward from an initial fast rate v_i toward a slower steady-state rate v_s.
A typical TDI progress-curve experiment uses a model protease or kinase (e.g., a caspase cleaving Ac-DEVD-AMC, or cathepsin K cleaving Z-FR-AMC) at a substrate concentration close to Km so that substrate depletion over the assay window stays below ~15%, keeping the underlying steady-state kinetics quasi-constant. Enzyme is added last to initiate the reaction directly in the presence of inhibitor (no pre-incubation) across a dilution series spanning roughly 0.1–5× the expected Ki, plus a DMSO/vehicle-only control defining v_0, the uninhibited rate.
Fluorescence or absorbance is recorded continuously at 5–15 s intervals for 30–60 minutes on a plate reader or stopped-flow instrument. For a fast-binding inhibitor, each trace is a straight line from t=0, and a plot of the resulting steady rates against [I] simply yields IC50/Ki by standard dose-response fitting. For a time-dependent inhibitor, each trace instead shows an initial, steeper phase (rate v_i, reflecting only the rapidly formed low-affinity EI encounter complex) that decays smoothly into a shallower terminal phase (rate v_s, reflecting the tighter, slowly formed EI* complex). The curvature becomes more pronounced — and v_s drops further below v_i — as [I] increases, because higher inhibitor concentrations drive the initial binding equilibrium further toward EI, accelerating the observed approach to the new, tighter steady state.
A critical control is distinguishing genuine mechanism-based slow binding from artifacts: enzyme instability (progressive loss of activity independent of inhibitor, checked with vehicle-only traces), inner-filter/fluorophore inhibition at high inhibitor concentration, and substrate depletion mimicking curvature. Analysts typically fit both a straight line and the full four-parameter exponential model (below) to every trace and use an extra-sum-of-squares F-test; a statistically preferred 2-phase fit (p<0.01) across a concentration series, with the degree of curvature scaling with [I], is the operational definition of time-dependent inhibition worth carrying into full kinetic analysis.
Once curvature is confirmed, every individual progress curve is fit by nonlinear regression to the closed-form integrated rate law for slow-onset inhibition. This single equation compresses an entire time-course into three biologically meaningful numbers per inhibitor concentration: how fast the enzyme starts (v_i), how slow it ends up (v_s), and how quickly it gets there (k_obs) — the raw material for every mechanistic conclusion that follows.
The standard model fit to each trace is P(t) = v_s·t + [(v_i − v_s)/k_obs]·(1 − e^(−k_obs·t)) + d, where d is a small vertical offset absorbing dead-volume product formed before the first read. Fitting proceeds by nonlinear least squares (Levenberg–Marquardt), typically in GraphPad Prism or, for more complex or multi-step mechanisms, by numerical integration of the full differential rate equations in KinTek Explorer, which avoids assuming any closed-form solution and instead simulates E, I, EI, EI*, S, and P simultaneously and optimizes the microscopic rate constants directly against the raw fluorescence traces.
A key subtlety arises when the inhibitor concentration approaches or falls below the enzyme concentration used in the assay (common for very tight, sub-nanomolar Ki* inhibitors): free [I] can no longer be approximated as constant and equal to total [I], because a significant fraction is sequestered in EI/EI*. In that tight-binding regime the simple exponential fit is replaced by the Morrison quadratic (tight-binding) equation, which explicitly accounts for ligand depletion, or by numerical simulation. Ignoring this regime systematically underestimates potency — a common pitfall when comparators include sub-nM tool compounds.
Global fitting — simultaneously fitting all [I] traces with shared parameters for v_0, Km, and the mechanistic rate constants (k_on, k_off, k5, k6) rather than fitting each curve independently — is now standard practice because it uses the full information content of the dataset and propagates realistic confidence intervals onto the mechanistic constants rather than onto derived, per-curve v_i/v_s/k_obs triplets. Quality control requires R²>0.98 per trace, randomly scattered residuals (no systematic curvature left unfit), and k_obs standard errors below ~15% of the fitted value; traces failing these criteria are usually symptomatic of enzyme instability or an assay window too short to capture the terminal steady-state phase.
The single most information-dense plot in TDI kinetics is k_obs replotted against [I]. Its shape alone distinguishes fundamentally different molecular mechanisms of slow binding, and its slope, intercept, and asymptote directly yield the microscopic rate and equilibrium constants that govern how tightly — and how slowly — an inhibitor engages its target.
Three canonical schemes dominate TDI literature, each leaving a distinct fingerprint on the k_obs vs [I] replot:
Scheme A — one-step slow binding (E + I → EI, no subsequent isomerization, but binding itself is intrinsically slow, e.g. due to a conformational gate): k_obs = k_on·[I] + k_off. The replot is a straight line; the slope is the true bimolecular association rate constant k_on (often 10³–10⁵ M⁻¹s⁻¹, far below diffusion-limited ~10⁸ M⁻¹s⁻¹, reflecting a required protein or ligand conformational change before binding), and the y-intercept is k_off, from which Ki = k_off/k_on.
Scheme B — two-step induced-fit/conformational-selection binding (E + I ⇌ EI ⇌ EI*): a rapid-equilibrium encounter complex EI (governed by Ki = k2/k1, typically diffusion-limited on-rate) is followed by a slow, often intramolecular isomerization to a tighter complex EI* with forward rate k5 and reverse rate k6. Here k_obs = k6 + k5·[I]/(Ki·(1+[S]/Km) + [I]) — a rectangular hyperbola in [I] that saturates at k_obs,max = k5 + k6 as [I]→∞ and has y-intercept k6. This is the shape simulated in the plot at left: note the pronounced curvature at low [I] transitioning to a plateau above roughly 3–5× Ki. From the fitted asymptote and intercept, the overall tight-binding constant follows as Ki* = Ki·k6/(k5+k6) — in this system, 850 nM · 0.00058/(0.045+0.00058) ≈ 12 nM, a ~71-fold tightening driven entirely by the isomerization step rather than the initial encounter affinity.
Scheme C — covalent, irreversible inactivation (E + I ⇌ EI →(k_inact) E–I, no measurable k6): k_obs increases hyperbolically with [I] but never plateaus at a true reversible asymptote; instead it approaches k_inact, and potency is reported as the ratio k_inact/KI (analogous to a bimolecular efficiency constant) rather than a true Ki, because activity is never fully recovered by dilution. Distinguishing Scheme B from Scheme C requires the jump-dilution reversibility test performed in Stage 5. Model selection in practice uses Akaike Information Criterion (AICc) comparison between the linear and hyperbolic fits together with visual inspection for a discernible plateau within the tested [I] range — extending the concentration range to ≥10× the apparent Ki is often necessary to unambiguously resolve the asymptote.
While full progress-curve analysis delivers the complete mechanistic picture, medicinal chemistry teams need a fast, scalable screen to flag time-dependent behavior across hundreds of analogs. The pre-incubation IC50 shift assay does exactly this: enzyme and inhibitor are held together for a variable pre-incubation period before a brief substrate pulse measures initial velocity, and IC50 is plotted as a function of that pre-incubation time.
Enzyme is combined with a dilution series of inhibitor in assay buffer and incubated at 25–37°C for a defined time — typically 0, 5, 15, 30, and 60 minutes across parallel plates or wells. At the end of each pre-incubation period, substrate is spiked in at a saturating pulse and the initial rate is captured over a short window (often <2 minutes, or read immediately by stopped-flow) chosen to be short relative to k_obs⁻¹ so that essentially no further EI→EI* isomerization occurs during the measurement itself — the read reports the instantaneous inhibitory state established during pre-incubation, not a new time-dependent process layered on top.
At each pre-incubation time, the initial-rate data across the inhibitor dilution series are fit to a four-parameter logistic (Hill) dose-response to extract an apparent IC50. For a fast-equilibrium inhibitor, IC50 is essentially invariant with pre-incubation time (typically <2-fold variation, within assay noise) because equilibrium is reached within the mixing dead time regardless of how long you wait afterward. For a genuine time-dependent inhibitor, IC50 falls progressively with increasing pre-incubation time as more of the population converts from the loosely bound EI state to the tightly bound EI* state, and — critically — the IC50 vs. time curve plateaus once pre-incubation exceeds roughly 5/k_obs, at an asymptotic value reflecting Ki* rather than Ki. The magnitude of the total fold-shift (IC50 at t=0 divided by the plateau IC50) is a convenient potency-tightening metric reported directly in SAR tables; shifts below ~2–3-fold are usually within experimental noise, while shifts of 10–100-fold (as simulated here, ~41-fold) are unambiguous evidence of a two-step slow-binding or covalent mechanism and materially change how the compound's in vivo dose and dosing interval should be modeled — an assay run without pre-incubation would systematically underestimate true achievable potency by more than an order of magnitude.
The final, decisive experiment both confirms reversibility and delivers the parameter increasingly used to rank drug candidates ahead of binding affinity alone: residence time, τ = 1/k_off. A pre-formed, saturating enzyme-inhibitor complex is rapidly and massively diluted into a substrate-rich buffer, and the slow recovery of catalytic activity is tracked directly, giving a first-principles measurement of how long the complex survives once free inhibitor is effectively removed.
The complex is first pre-formed by incubating enzyme with a high, saturating concentration of inhibitor (typically ≥20× Ki*) for a period well beyond 5/k_obs, ensuring near-complete conversion of EI to the tight EI* state. This mixture is then diluted 50- to 200-fold directly into an assay well containing substrate at Km, dropping the free inhibitor concentration to well below Ki (often below Ki*), so that any inhibitor that dissociates is diluted away rather than rebinding. Product formation is monitored continuously over the following minutes to hours; a control lacking pre-incubation (enzyme added directly to the diluted, low inhibitor concentration) defines the fully active, uninhibited rate.
For a reversible slow-binding inhibitor, activity recovers slowly following A(t) = A_final − (A_final − A_initial)·e^(−k_off·t), and fitting this trace yields k_off = k6 directly and independently of the forward isomerization rate k5 measured earlier — an internal consistency check against the kobs replot asymptote. If instead little or no activity recovers even after extended dilution (hours), the inhibitor is covalent/irreversible (Scheme C), and k_off is operationally zero; residence time is then effectively infinite until new enzyme is synthesized in vivo.
Residence time, τ = 1/k_off (here 1/5.8×10⁻⁴ s⁻¹ ≈ 1724 s ≈ 28.7 min, giving a half-life of dissociation t½ = ln2/k_off ≈ 19.9 min), has become a first-class optimization parameter in drug discovery because target engagement in vivo depends on how long the drug-target complex persists relative to the pharmacokinetic clearance of free drug — a long residence time can sustain target inhibition well after plasma drug concentration has fallen below Ki, decoupling efficacy duration from exposure and enabling reduced dosing frequency.
The cathepsin K inhibitor odanacatib is a textbook two-step slow-binding inhibitor: a rapid-equilibrium encounter complex (Ki in the low nanomolar range) isomerizes over minutes to a markedly tighter EI* complex with sub-nanomolar Ki* and a measured residence time exceeding several hours. That extended residence time — far outlasting the compound's plasma half-life — was central to the rationale for once-weekly oral dosing in clinical development, illustrating how a Stage 5-style jump-dilution measurement can directly justify a clinical dosing regimen rather than remaining a purely academic kinetic parameter.