Mechanism for pH-dependent gene regulation by amino-terminus-mediated homooligomerization of Bacillus subtilis anti-trp RNA-binding attenuation protein

Recombinantly Expressed AT Is Produced in Both N-Formyl and Deformylated Forms.

High yields of recombinant U-[¹⁵N]-AT were obtained from E. coli cells transformed with a codon-optimized synthetic Bsu AT gene cloned into the pET24a plasmid and grown in defined minimal media (23). Purification of the protein under native conditions by ion exchange chromatography resulted in samples that were homogeneous based on SDS-PAGE with Coomassie staining. However, these initial samples of AT yielded ¹H-¹⁵N heteronuclear single quantum correlation (HSQC) spectra that exhibited doubling of several of the amide resonances, suggesting either that two forms of AT were present or that some of the residues were experiencing slow conformational exchange between two distinct states. Further purification of chemically denatured AT by reversed phase or anion-exchange chromatography revealed two species whose mobilities and electrospray mass spectra (Fig. S1) were consistent with AT (5,648.0 Da observed, 5,649.5 Da expected; ProtParam, ExPASy), and N-formyl-AT (28 Da greater; 5,676.0 Da observed). This observation is consistent with incomplete removal of the formyl group from the N-formyl-methione used to initiate bacterial protein translation.

The chemical identity of N-formyl-AT was further confirmed through analysis of its heteronuclear NMR spectra. While assigning backbone resonances for this protein we found signals for an amide in the HNCA and HNCACB spectra that correlated only to C^α and C^β shifts matching those assigned to Met1. In the HNCO spectrum, this amide proton correlated to two resonances in the carbon dimension centered at 166.6 ppm and separated by 199 Hz (Fig. 1A). In an HNCO spectrum, usually a single peak is observed correlating each amide ¹H (ω₃) to its directly attached ¹⁵N (ω₂) and the attached ¹³C (ω₁) of the carbonyl group in peptide bonds (32). However, for the formyl-methionine amide, the aldehyde proton will lead to splitting of the carbonyl resonance because ¹H decoupling is not generally applied during the ¹³C evolution period. The magnitude of the observed splitting is consistent with published values for one-bond couplings in formamide groups (33, 34). This amide also exhibits NOEs (Fig. 1B) to the H^α of Met1 and a resonance at 8.11 ppm that matches a unique resonance in the ¹³C HSQC spectrum with a ¹³C shift of 166.5 ppm (Fig. S2), consistent with ¹H and ¹³C shifts for trans-formyl-methionine of 8.12 ppm and 166.3 ppm (Human Metabolome Database, http://www.hmdb.ca/, HMDB01015). Thus, our recombinant AT is produced as a mixture of mature AT and unprocessed fAT bearing an N-formyl-Met; further N-terminal processing of AT in vivo by methionine aminopeptidase was not observed.

Effect of the N-formyl Group on Oligomerization.

It seemed unlikely that the mere presence of the N-terminal formyl group could account for the large number and magnitude of shift differences observed in the ¹H-¹⁵N spectra of the two species of AT (Fig. 2A). Backbone resonance assignments were obtained from triple resonance NMR spectra recorded for highly concentrated samples of both species of AT (∼8 mM AT and 3 mM fAT protomer concentrations) (23), and the chemical shift differences were mapped to the structure of the crystallographically observed AT dodecamer (Fig. 2B). Notably, the largest shift differences between AT and fAT map to the interface between trimers in the dodecamer. Because chemical shifts are particularly sensitive probes of intermolecular interactions (35), these data suggested that the oligomeric states of the AT species differed.

To determine the oligomeric states of these chemically distinct forms of AT, we carried out ¹⁵N relaxation measurements to obtain their overall rotational correlation times (32). NMR spectra were recorded at 55 °C with 3.1 mM fAT and 8.6 mM AT (protomer concentrations). Analysis of the ¹⁵N relaxation data (Fig. 3) revealed a mean T₁ for AT of 1.91 ± 0.25 s, whereas that in fAT is 0.631 ± 0.091 s. The mean ¹⁵N T₂ of AT is 31.9 ± 4.5 ms, whereas that of fAT is 85 ± 13 ms. Fitting these data to an isotropic tumbling model (32) yielded overall rotational correlation times of 25.2 ± 2.8 ns for AT and 8.2 ± 1.0 ns for fAT. Hydrodynamic calculations (36) using coordinates derived from the AT₁₂ crystal structure (21) predicted an isotropic rotational correlation time at 55 °C of τ_m = 28.7 ns for AT₁₂, and an axially symmetric rotational diffusion tensor for AT₃ with correlations times of and . Hydrodynamic properties were also measured in one-dimensional stimulated echo gradient diffusion NMR experiments (37), which yielded diffusion coefficients of and for AT and fAT, respectively (Fig. S3). These values compare favorably with values calculated (36) from the coordinates of AT₁₂ and AT₃, 5.66 × 10^-7 cm²/s and 9.26 × 10^-7 cm²/s, respectively. The agreement between these experimental and calculated values indicate that under the experimental conditions deformylated AT forms dodecamers, whereas fAT is trimeric.

pH-Dependent Oligomerization of Deformylated AT.

The observation that the N-formyl group prevents formation of AT dodecamers is consistent with the crystallographically observed ion pair involving the amino terminus of each protein chain and the sidechain carboxylate of Asp7 from another trimer in the dodecamer (Fig. 4) (21, 30). The presence of the ion pair predicts the oligomeric state of AT is pH dependent. The effect of pH and concentration on the oligomeric state of AT was examined by recording a series of 2D ¹H-¹⁵N correlation spectra (Fig. 5) over a range of pH values spanning 6.5 to 8, taking advantage of the differences in the ¹H-¹⁵N shifts between the spectra between the trimeric and dodecameric forms as indicative of oligomeric state (Fig. 2). Protein concentrations for these experiments varied from 268 μM to 638 μM. At low pH and these concentrations, signals corresponding to both AT₃ and AT₁₂ are observed, with the dodecameric state being more highly populated. At higher pH the relative intensity of the AT₁₂ peak decreases, whereas that due to AT₃ increases; only signals from AT₃ are observed at pH 8 (Fig. 5). No oligomerization was observed for fAT over this pH range, even at higher concentrations (> 2 mM), indicating that the formyl group prevents formation of dodecameric AT. These data show that AT₃ and AT₁₂ are in slow exchange on the NMR time scale and that low pH favors oligomerization to the dodecamer.

We interpreted the pH and concentration dependence of AT₃ oligomerization to AT₁₂ by considering the following coupled equilibria in which k_a is the microscopic acid dissociation constant of the N-terminal amine:

[1]

[AT₃], [], [], and [] are the concentrations of differently protonated states of the trimer, and [AT₁₂] is the concentration of the dodecamer; K_a1, K_a2, and K_a3 are the macroscopic acid dissociation constants of the trimer. The pK_a of the N-terminal amine is defined as the microscopic pk_a = - log(k_a). The microscopic equilibrium constants for deprotonation of each N terminus were assumed to be equal and independent. K₁ is the association constant for four triprotonated trimers to form dodecamer. Implicit in the model is that only the triprotonated form of AT₃ is able to oligomerize; equivalently, protons do not dissociate from AT₁₂. We rationalize this approximation based on structural arguments (once assembled into the dodecamer the close proximity of the N-terminal amine and carboxylate would strongly favor protonation; Fig. 4), relatively weak stability of the oligomer (deprotonating one or more trimers would likely lead to disassembly), and the highly cooperative nature of oligomer assembly (no intermediates are observed, indicating all favorable contacts must be present to stabilize the dodecamer).

This model was used to fit the relative volumes of the AT₁₂ peaks of several amides with dispersed timer and dodecamer frequencies in the ¹H-¹⁵N correlation spectra (Val10, Ala28, Cys29, Gly31, Gly33, Val34, Ile35, and Thr42; Figs. S2 and S4). The most likely values of the parameters are found to be log K₁ = 15.9 ± 0.6 (K₁ in units of M^-3) and pk_a = 7.38 ± 0.09 (Fig. S4a). These parameters were found to accurately model the measured fractions of AT₁₂ as a function of pH and concentration (Fig. S4b). These linked equilibrium constants allow the determination of the apparent dissociation constant of the dodecamer, i.e., , the concentration of free AT₃ at which half of the protein is trimeric and the other half is oligomerized to AT₁₂. Due to the linked protonation equilibrium, this value is very sensitive to pH: at values below 6, where AT is essentially fully protonated, has an almost constant value of 3 μM, whereas at pH 7, increases to 13 μM, and to 2 mM at pH 8 (Fig. S5). As a consequence of this thermodynamic linkage (29), just as pH alters the effective equilibrium constant for oligomerization, protein concentration affects the protonation state of the protein. As the total protomer concentration increases, protonation is favored and increases.

Sample Preparation.

Bsu AT was overexpressed from E. coli cells in defined minimal media. AT and fAT were purified under denaturing conditions by reversed phase HPLC and refolded by dilution and dialysis. pH and concentration dependence of the trimer-dodecamer equilibrium of AT was measured by sequentially dialyzing the same sample into buffers at pH 7.0, 7.5, 8.0, and 6.5 with NMR spectra taken after each dialysis; the concentrations of the fAT solutions varied between 2.14 and 4.74 mM, whereas that for AT varied between 268 μM and 628 μM. Additional details are provided in SI Text.

NMR Spectroscopy of AT.

NMR spectra were acquired on Bruker DRX 600 or 800 NMR spectrometers equipped with cryoprobes at a sample temperature of 55 °C. Samples for resonance assignments consisted of an 8.6 mM (protomer concentration) AT in 20 mM d₁₁-Tris at pH 7.8 (298 K), 250 mM NaCl, 300 μM ZnCl₂, 2 mM TCEP [tris(2-carboxyethyl)phosphine], and 0.02% (wt/vol) NaN₃ and a 3.1 mM sample of fAT in 20 mM d₁₁-Tris pH 7.5, 20 mM β-mercaptoethanol, 1 mM ZnCl₂, and 0.02% (wt/vol) NaN₃ unless otherwise specified. Resonance assignments of U-[¹³C, ¹⁵N]-formyl-AT (fAT) and triply labeled, U-[²H, ¹³C, ¹⁵N] AT were obtained by analysis of standard triple resonance NMR spectra (23); assignments for AT₃ have been deposited in the BioMagResBank. Backbone amide ¹⁵N relaxation data were recorded at 600 MHz, 55 °C using standard methods.

Transverse relaxation rates of amide protons were measured by a ¹⁵N-edited ¹H CPMG (Carr–Purcell–Meiboom–Gill) sequence. ¹H magnetization was excited with a 90° pulse, which was followed by the CPMG pulse train. Magnetization was then transferred to ¹⁵N for editing and back to ¹H for detection in the standard HSQC fashion (50). ¹H-¹⁵N multiple quantum transverse relaxation was measured similarly with multiple quantum coherence formed after the first INEPT transfer, which was refocused by a CPMG train on both ¹H and ¹⁵N. Magnetization was transferred back to ¹H for detection. In both measurements, pulse delays were 1 ms in the CPMG train ([τ-180°-τ]_n) and the number of 180° pulses was varied from 1 to 64. A reference spectrum with τ = 1 μs and a single 180° pulse were also measured. The amide region of the resulting spectra were integrated and fit to a single exponential decay.

Fitting of pH Dependence of Oligomerization to the Equilibrium Model.

To determine the equilibrium constants in the oligomerization model [Eq. 1], the volumes of resolved resonances from AT₁₂ and AT₃ (residues Val10, Ala28, Cys29, Gly31, Gly33, Val34, Ile35, and Thr 42) were integrated and corrected for the transverse relaxation during the coherence transfer portions of the pulse sequence (51). The fraction of AT protomers in AT₁₂ was obtained from the corrected peak volumes by

[2]

as a function of pH and total AT protomer concentration for each of the residues analyzed. The data from signals for each residue were fit to the oligomerization model (Eq. 1) by calculating the likelihood function (52) that the model parameters (pK_a and log K₁) describe the measured values of f_AT12. Simulated values of f_AT12 were calculated as a function of pK_a (pK_a varied from 6.8 to 8.0 in steps of 0.012) and log K₁ (log K₁ varied from 12 to 20 in steps of 0.08), and the most likely values of pK_a and log K₁, and their errors, were determined from the likelihood probability distribution (Fig. S5).