ENZYMOLOGY WITH A SPIN-LABELED PHOSPHOLIPASE C: SOLUBLE SUBSTRATE BINDING BY ³¹P NMR FROM 0.005 TO 11.7 TESLA

Abstract

³¹P NMR relaxation studies from 0.005 to 11.7 T are used to monitor water-soluble inositol 1,2-(cyclic)-phosphate (cIP) binding to phosphatidylinositol-specific phospholipase C spin-labeled at H82C, a position near the active site of the enzyme, and to determine how activating phosphatidylcholine (PC) molecules affect this interaction. We show that, in the absence of an interface, cIP binding to the protein is not rate-limiting, and that lower activation by PC vesicles as opposed to micelles is likely due to hindered product release. The methodology is general and could be used for determining distances in other weakly binding small molecule ligand/protein interactions.

Phosphatidylinositol-specific phospholipase C (PI-PLC) enzymes catalyze the hydrolysis of phosphoinositides in two steps: an initial phosphotransferase reaction to produce diacylglycerol and inositol 1,2-(cyclic)-phosphate (cIP), followed by hydrolysis of the cIP to inositol-1-phosphate. For PI-PLC from Bacillus thuringiensis, water-soluble cIP hydrolysis is slow with a K_m >50 mM (1,2), but the specific activity is significantly enhanced by the addition of phosphatidylcholine (PC) interfaces, either micelles such as diheptanoyl-PC (diC₇PC) or vesicles of 1-palmitoyl-2-oleoyl-PC (POPC) (2,3). For example, at 5 mM cIP, the addition of diC₇PC leads to a 40–50-fold increase in specific activity (4); POPC SUVs added yield a 3–4-fold increase. While both PC interfaces decrease K_m, the V_max is dramatically different (Fig. 1).

Figure 1.

Dependence of PI-PLC specific activity, at 25°C, on cIP concentration with 5 mM diC₇PC micelles (●) or 5 mM POPC SUVs (❍) added. Note the 10-fold different specific activity scales for cIP with diC₇PC micelles (left) or POPC SUVs (right).

These cIP kinetics raise two interesting questions: (i) does the decreased K_m for PC-enhanced cIP hydrolysis reflect a tighter binding of substrate to the enzyme, and (ii) why is V_max lower with a vesicle activator compared to a micelle surface? To address these, we have used an alternative to standard NMR methods (high resolution field cycling ³¹P NMR spectroscopy or fc-P-NMR) in which the ³¹P spins are prepared, and their signals detected, at standard high fields, but their relaxation back toward equilibrium occurs at a lower field, as described in the Supporting Information and elsewhere (5–7). Resonances for cIP and PC species are well separated (cIP at 17 ppm and diC₇PC at 0 ppm), at the observation field of 11.7 Tesla (T), and spin-lattice relaxation rates (R₁ = 1/T₁) can be obtained over a wide range of magnetic field strengths (6,7). The field dependence of R₁, from 0.005 up to 11.7 T, can then be analyzed with standard theory (6,7) to obtain correlation times.

For small phosphorus-containing molecules free in solution, the dipolar contribution to ³¹P relaxation is small, and the ³¹P R₁ is dominated by chemical shift anisotropy (CSA) over most of the accessible field range. But if the small molecule spends part of its time ligated moderately weakly to a larger complex such as an enzyme, then its observed R₁ is the weighted average of its small free R₁ and the much larger R₁ that it would have if bound permanently. The effect is field-dependent and depends on the ratio of ligand/protein as well as proximity of the ³¹P to the nearby proton dipoles, in either the small molecule or the enzyme, that relax it. A way of enhancing the added effect is to further add an electron-spin label to the protein near a suspected binding site. The much larger magnetic dipole of the electron can have a useful effect on the ³¹P even with a high ligand/protein ratio.

To explore cIP binding to PI-PLC with fc-P-NMR, we used a mutant protein where a key active site residue (His82, the general acid in the proposed mechanism that protonates the DAG anion initially produced in the phosphotransferase step (8)) was altered to cysteine. The H82C enzyme is inactive but should still bind substrate. A nitroxide spin-label (1-oxyl-2,2,5,5-tetramethylpyrroline-3-methyl-methanethiosulfonate) was attached to this cysteine (H82C-SL) as a way of introducing a much larger dipole in the region where the cIP should bind. The field dependence profiles for cIP (5 mM) with spin-labeled PI-PLC (0.5 mg/ml, 14.4 µM) in the absence and presence of with diC₇PC (5 mM) micelles are shown in Fig. 2A. In the absence of protein, cIP exhibits behavior typical for small phosphate esters – the relaxation rate is more or less constant and low below 1 T, with a rise at higher fields due to CSA that follows a square law dependence.

Figure 2.

(A) Field dependence of 5 mM cIP ³¹P R₁ in the absence (+) and presence (❍) of 0.5 mg/ml H82C-SL, as well as with both H82C-SL and 5 mM diC₇PC (●) in 50 mM HEPES, pH 7.5, with 1 mM EDTA. (B) Field dependence of the cIP ³¹P R₁ with 5 mM diC₆PC (❒) or 5 mM PC SUVs (■) and H82C-SL. For comparison the relaxation profile for cIP in buffer is shown (+). The inset in (B) shows the very low field profile for cIP with the PC SUVs and H82C-SL present.

The presence of 14.4 µM non-spin-labeled H82C (with an excess of DTT present) yielded no significant change in cIP relaxation rates, presumably because the fraction of cIP that binds to enzyme is very small and the ³¹P-¹H interaction in the bound state is relatively small as well. However, when the same amount of spin-labeled PI-PLC (H82C-SL) is added to a sample with cIP alone, there is a substantial increase in R₁ below 2 T for cIP (❍, Fig. 2A).

While the binding of cIP to protein is weak, it is specific. In a mixture of three different water-soluble phosphates, cIP, glucose-6-phosphate, and diC₄PC, in the presence of H82C-SL (Supplement, Fig. S1), the profiles for the diC₄PC and glucose-6-phosphate were identical with or without the H82C-SL indicating they did not bind significantly to the enzyme (or if they did were >20 Å from the site of the spin-label). Only the cIP showed strong relaxation by spin-labeled protein.

The part of the ³¹P relaxation at each field due to the spin label, ΔR₁, was determined by subtraction of the non-spin-label contribution. The residual relaxation data were analyzed using standard relaxation theory (5, 6), modified for a system in fast exchange on the on-enzyme-turnover timescale:

$$\mathrm{\Delta }{\mathrm{R}}_1=\mathrm{\Delta }{\mathrm{R}}_{\mathrm{P}–\mathrm{e}}(0)/(1+{{\mathrm{\omega }}_{\mathrm{P}}}^2{{\mathrm{\tau }}_{\mathrm{c}}}^2).$$ (1)

where ΔR_P−e(0) is the low-field limit of ΔR₁. In eq. 1 we omit the usual terms of theory that involve the electron angular frequency because these are negligibly small owing to the large magnitude of the electron’s frequency compared to ω_P. As expected, the τ_c values so obtained did not depend on total enzyme concentration E_o, and these are presented in Table 1 for the samples shown in Figure 2. In the absence of the diC₇PC micelles, the cIP τ_c increases from 0.019 to 7.1 ns when spin-labeled protein is added. The ΔR_P−e(0) of 0.14 s⁻¹ clearly shows that bound cIP must be near the spin-labeled residue. When diC₇PC micelles are added, there is a larger effect on ΔR_P−e(0), but τ_c also increases. When POPC SUVs are present, the τ_c for bound cIP is even longer, 24 ns (■, Fig. 2B), and there is a further increase in R₁ below 0.01 T with ~1 µs correlation time (inset, Fig. 2B). This additional low field rise in R₁ was not observed for cIP in any of the samples with diC₇PC micelles as the activator.

Table 1.

Parameters extracted from fc-P-NMR for cIP (5 mM) with H82C-SL (14.4 µM) in the absence and presence of different PC species (5 mM).

PC	ΔRP−e (s−1)	τc (ns)
-	0.14±0.01	7.1±1.0
diC7PC	0.32±0.01	16.8±1.5
diC6PC	0.21±0.01	9.8±1.2
POPC	0.20±0.02	24.4±5.9

The values of the parameter ΔR_P−e(0) obtained by fitting the data depend on concentrations of total cIP (denoted by [cIP]_o) and of total enzyme E_o:

$$\mathrm{\Delta }{\mathrm{R}}_{\mathrm{P}-\mathrm{e}}(0)=([\mathrm{H}82\mathrm{C}-\text{SL}•\text{cIP}]/{[\text{cIP}]}_{\mathrm{o}})$$ (2a)

$$\times ({\mathrm{S}}^2{\mathrm{\tau }}_{\mathrm{c}}{{\mathrm{r}}_{\mathrm{P}-\mathrm{e}}}^{-6})(0.3{\mathrm{\mu }}^2{(\mathrm{h}/2\mathrm{\pi })}^2{{\mathrm{\gamma }}_{\mathrm{P}}}^2{{\mathrm{\gamma }}_{\mathrm{e}}}^2).$$ (2b)

Here the term on line (2a) is the fraction of cIP bound to the enzyme relative to the total cIP. In term (2b), S² is the order parameter of the electron spin-³¹P dipolar interaction which we take as unity because of the long distance r_P−e between the phosphorus of cIP and the electron spin compared to the size of local picosecond motions. We will eventually be interested in the values of r_P−e for the different complexes. The last term of (2b) contains standard constants of relaxation theory defined elsewhere (5,6).

A series of data sets like those of Fig. 2A were obtained for a several total concentrations of enzyme E_o and [cIP]_o, the latter chosen so that the R₁ of the ³¹P of cIP would be within the range permitted by the field-cycling apparatus. We assume that binding of cIP to enzyme is a simple bimolecular reaction with binding constant K_d. We define a new parameter v_P−e which is equal to ΔR_P−e(0) times [cIP]_o, divided by total enzyme. The new parameter v_P−e is then given by eq. 1 and eq. 2 with the first term (2a) replaced (after one familiar step using mass-action, and the fact that [cIP]_o >> E_o) by (1+ [K_d/cIP]₀)⁻¹ as shown in eq. 3:

$${\mathrm{v}}_{\mathrm{P}-\mathrm{e}}=(1+{\mathrm{K}}_{\mathrm{d}}/{{[\text{cIP}]}_0)}^{-1}\times \mathrm{c}.$$ (3)

where c is comprised of the terms in (2b). The quantity v_P−e is an NMR version of specific activity in enzyme kinetics, and the asymptotic value of v_P−e at high concentration of cIP is an NMR analog of an enzymatic V_max. In Fig. 3, v_P−e is plotted versus [cIP]_o for the cIP binding to H82C-SL in the presence of either diC₇PC micelles or POPC vesicles. The values of K_d for cIP binding to inactive H82C-SL obtained by fitting these curves (25±5 mM with diC₇PC and 14±7 mM for POPC SUVs) are roughly twice the K_m values obtained from the kinetics (10±1 mM with diC₇PC, 8.2±0.3 mM for POPC SUVs). The asymptotic values of v_P−e, which are better determined than the K_d values, differ by a factor of 2 with diC₇PC micelles leading to more effective relaxation. Dividing the average τ_c times the constants in (2b) by the asymptote, and taking the sixth root, we get tentative r_P−e distances of 7.6±0.1 Å for cIP bound to H82C-SL with diC₇PC present and 9.2±0.6 Å when POPC SUVs serve as the activating interface. (They are tentative distances because, in this case, we do not know exactly what motion of the spin-label and ³¹P that τ_c represents.) The two distances are significantly different when assessing the experimental error in the v_P−e asymptote in Fig. 3, and, more importantly the difference in τ_c values for the two systems. If we dock cIP into the myo-inositol position in the B. cereus PI-PLC crystal structure (1PTG (10)), and replace His82 with a Cys to which the flexible spin label is attached, the ³¹P-electron distance is predicted to be in the 5–9 Å, depending on the orientation of the nitroxide group, consistent with the fc-P-NMR analysis.

Figure 3.

Variation of v_P−e with total cIP concentration with diC₇PC (●) or POPC SUVs (❍) present.

So what do these results imply about the low activity of PI-PLC hydrolyzing cIP without a PC interface? τ_c/ΔR_P−e(0) is proportional to r⁶ multiplied by the fraction of cIP bound to the enzyme. Since τ_c/ΔR_P−e(0) in Table 1 is the same in the absence and presence of diC₇PC, the fraction of cIP at the active site must be similar. Therefore, the lower value of K_m in the presence of a PC interface is not the result of increased affinity for the cIP. The PC interface likely promotes a conformational change of the PI-PLC•cIP complex critical for hydrolysis. For ‘monomeric’ diC₆PC activation of PI-PLC, the increased τ_c indicates that the cIP bound to H82C-SL with diC₆PC present forms a complex. The larger τ_c would be consistent with a complex where more than a single diC₆PC was interacting with the protein to produce a mini-interface.

As to the lower activation by PC SUVs, there are two notable differences in cIP behavior with micelle versus vesicle. (i) The r_P−e is longer (by ~1–2 Å) with POPC SUVs, and (ii) cIP bound to the spin-labeled protein/POPC SUV complex exhibits an increased relaxation rate at very low fields that is not detected for the cIP and protein with diC₇PC micelles. PI-PLC binds tightly to PC SUVs with K_d in the 0.04–0.06 mM range (11). Displacement of the His82 catalytic residue could cause the poorer cyclophosphohydrolase activity of the enzyme when POPC SUVs serve as the activating interface. Alternatively, the further increase in the relaxation rate for cIP below 0.02 T (Fig. 2B, inset) with a correlation time that resembles SUV tumbling rates (~1 µs (12)) indicates that the cIP off-rate from the POPC•H82C-SL•cIP complex must be less than 10⁶ s⁻¹, an observation that would be consistent with slow release of IP from the POPC•PI-PLC•IP complex. Interestingly, given that the K_d values for cIP binding to H82C-SL with either interface are similar, a shorter off-rate with SUVs as activators would mean that the on-rate of cIP has been reduced – possibly indicating some occlusion of the active site when the protein is anchored to the vesicle.

Finally, this novel methodology holds promise for identifying binding proximities in other situations. For example, a small amount of protein spin-labeled at a particular site added to a mixture of phosphorus-containing ligands should relax only those ligands that bind near the spin-labeled site – a potentially useful way of identifying non-active-site binding ligands.