There are several biochemical and assay-artifact reasons your ATP curve can look like it returns upward at very high ATP — the most likely culprits are (1) the PK/LDH coupling system behaving differently at very high ATP (allosteric effects, competition, or saturation), (2) Mg²⁺/ATP chemistry changing free ligand/enzyme-available species, and (3) optical / matrix artifacts (turbidity, refractive index, pathlength errors) at high nucleotide concentrations. Below I list the plausible mechanisms, how each would produce the effect you saw, and a targeted troubleshooting plan (quick controls & longer checks) so you can pinpoint the cause.

Why it can happen — possible mechanisms

1. Coupling-system problems (PK / LDH)

PK and LDH are not infinitely fast or insensitive: high ATP can affect PK activity (allosteric inhibition or competition with ADP), or push PK/LDH toward unusual equilibria. If PK becomes less efficient at regenerating ATP from ADP (or its kinetics change), the coupled NADH signal will not reflect the primary enzyme linearly.

Paradoxically, depending on how PK is affected and on PEP/ADP levels, the apparent rate can be distorted upward at extreme ATP because the coupling equilibrium shifts or ADP/ATP exchange dynamics change.

2. Mg²⁺ / ligand speciation

Most ATP in solution is MgATP; the fraction of free Mg²⁺ versus MgATP changes with total [ATP]. If you haven’t maintained excess Mg²⁺, increasing ATP can sequester Mg²⁺ and change the active species concentrations for both your enzyme and PK. That can alter both the real enzymology and the apparent kinetics. Often you want [Mg²⁺] > [ATP] (by ~1–2 mM) for stable behavior.

3. Ionic strength / buffer / pH changes

Very high nucleotide concentrations raise ionic strength and can change pH or enzyme stability. That shifts Km/Vmax or causes partial denaturation/aggregation that gives non-Michaelis behavior.

4. Optical / plate-reader artefacts

High molar solutes change refractive index and pathlength, can cause light scattering or baseline shifts. If you use a microplate reader, pathlength corrections and blank subtraction become important. ATP/stock buffer components could be slightly turbid at high concentrations and change apparent NADH absorbance slopes.

5. Viscosity / diffusion effects

Very high solute concentration increases viscosity — this usually lowers apparent rates (diffusion-limited), but can interact with coupling kinetics to produce non-intuitive shapes.

6. Substrate-species effects (real multi-site behavior)

Some enzymes show classical substrate inhibition over one range but then a second binding site (or binding to a different conformational state) gives partial activation at very high [S] — a real, biochemical biphasic behaviour (less common but possible).

7. ATP stock contamination / imbalance

ATP stocks sometimes contain ADP/AMP or salts that affect the reaction at high total additions. Also PEP in the coupling mix could become limiting at high ATP, shifting the coupling equilibrium.

8. Instrument non-linearity / NADH regeneration limits

If NADH absorbance change approaches detector limits, small errors at end points produce apparent rate errors. Or the coupling enzymes become limiting so the relationship between primary enzyme rate and NADH oxidation is no longer linear.

How to test which is happening — quick controls (do these first)

1. Blank controls: measure NADH absorbance vs time with all reagents except your primary enzyme at the high ATP concentrations (12 and 16 mM). If you see a slope or baseline drift, it’s an optical/matrix artifact (subtract it).

2. ATP-only & ADP spike test: in separate wells with coupling system present but no primary enzyme, add a known small amount of ADP (e.g., 10–50 µM) at low and high background ATP (0.1, 8, 16 mM). See whether PK/LDH converts that ADP to NADH consumption equally. If coupling efficiency drops or changes sign at high ATP, it implicates the coupling system.

3. Vary Mg²⁺: repeat the assay at high ATP but increase MgCl₂ (so [Mg²⁺] > [ATP] by ~1–2 mM). If the weird upward deviation disappears, Mg chelation was the problem.

4. PEP excess / coupling enzyme excess: increase PK and LDH concentrations (or PEP) so the coupling is definitely non-limiting. If the curve then follows expected inhibition, the coupling was limiting.

5. Check ATP stock: analyze (or run HPLC if possible) or at least compare a freshly prepared ATP stock vs the one used. Prepare ATP in your assay buffer and retest.

6. Instrument linearity: check the reader linearity with known NADH concentrations and confirm pathlength correction is used if applicable. If using cuvettes, check for turbidity with OD scan 250–600 nm for samples containing 16 mM ATP.

7. Repeat in cuvette vs plate: run the same high-ATP sample in a cuvette spectrophotometer (if available). If behavior differs between plate and cuvette, suspect pathlength/plate artefact.

8. Time-course inspection: plot raw NADH traces for each replicate (not only initial rates). Do the curves remain linear for the same time window? Non-linear early phase suggests mixing or fast secondary effects.

Longer checks / confirmatory experiments

Direct Pi or ADP measurement (e.g., malachite green for Pi or HPLC quantify ADP): bypass coupling to check if the primary enzyme really hydrolyses ATP more at 12–16 mM or if the coupling/reporting is failing.

Fit alternative kinetic models: try a substrate-inhibition model and also a two-site model (biphasic) to see whether a second-binding-site activation term improves fit. But don’t over-interpret fits without experimental controls.

Test other coupling systems: use a different coupling enzyme (e.g., hexokinase + G6PDH system) to see whether the anomaly is PK/LDH-specific.

Check enzyme stability at high [ATP]: incubate your enzyme with 16 mM ATP for the assay time and then assay activity at a standard ATP (e.g., 1 mM) to test for activation/deactivation by ATP exposure.

Practical recommendations (what I would do next, in order)

1. Run the blank (no primary enzyme) with 0.1–16 mM ATP and subtract.

2. Do the ADP spike test at low/high ATP to test coupling linearity.

3. Repeat top ATP points with +2 mM extra MgCl₂ (or ensure [Mg²⁺] > [ATP]).

4. Run 12/16 mM ATP samples in a cuvette (or with pathlength correction) to exclude plate optics.

5. If coupling looks suspect, increase PK/LDH and PEP, or switch coupling system.

6. If still unexplained, measure product (Pi or ADP) directly.

A few concrete numbers & model reminders

Substrate inhibition model:

If very-high-[S] points deviate upward, consider adding a second high-[S] activation term or a two-state model — but only after ruling out artefacts.

Mg²⁺ rule of thumb: keep [MgCl₂] ≥ [ATP] + 1 mM to avoid major chelation effects for many ATP-dependent systems.

If you like, paste one of the raw NADH time-course traces (CSV or a small table of time vs Abs340 for each ATP point) and I’ll inspect shapes and suggest the best initial window for rate fitting and whether the high-[ATP] traces look like baselines shifts, coupling nonlinearity, or real increases.