Solution NMR studies reveal no global flexibility in the catalytic domain of CDC25B

Abstract

The CDC25B phosphatase is a critical regulator of the cell cycle and has been validated as an important therapeutic target in cancer. Previous studies using molecular dynamics simulations have concluded that the catalytic domain of CDC25B may experience a significant degree of dynamics or be partially disordered in solution, a finding that has a pronounced impact on the structure-based development of CDC25B inhibitors. We have probed the backbone dynamics of the CDC25B catalytic domain in solution using NMR relaxation experiments and found that the core of the protein is relatively rigid and does not experience any large-scale dynamics over a broad range of time scales. Furthermore, based on RDC measurements we have concluded that the conformation in solution is very similar to that observed in the crystal form. Importantly, these findings rationalize the application of the CDC25B crystal structure in structure-based drug development.

Introduction

The CDC25 family of dual-specificity phosphatases are key regulators of the cell cycle, and their role in cancer has led to their active targeting for small molecule inhibitor development for over two decades^1–4. CDC25 phosphatases are critical activators of CDK/cyclin complexes by removing inhibitory phosphorylations on CDK family kinases⁵. As a downstream target of the DNA damage response pathway, the CDC25 phosphatases play an important role in maintaining genomic stability after DNA damage. All three phosphatases, CDC25A, CDC25B and CDC25C have been shown to be overexpressed in primary samples from a variety of tumor types, often correlating with a poor patient prognosis^6,7. Because of its potential as a therapeutic target in cancer, studies into CDC25 family catalytic mechanism and structure are of vital importance. Previous molecular dynamics (MD) simulations combined with bioinformatics analysis have suggested that approximately 20 C-terminal residues of CDC25B are partially unfolded or disordered in solution^8,9. It has been hypothesized that dynamics of the C-terminus allows for the formation of transient pockets near the active site ^8,9. This finding may directly impact development of CDC25B inhibitors and may present a warning against using the crystal structure of CDC25B for structure-based drug design⁹. Therefore, the need for experimental characterization of the dynamics of CDC25B has been emphasized in the literature⁹.

In this study, we have characterized the dynamics of the CDC25B catalytic domain using NMR relaxation experiments on various time scales, as well as evaluated the structure of CDC25B in solution using residual dipolar couplings. In contrast to the predictions from MD simulation^8,9, we found that CDC25B is a relatively rigid protein with a very well defined structure of the C-terminal helical fragment and the overall conformation in solution is very close to that observed in the crystal structure. Ultimately, our results have important implications for the structure-based design of small molecule inhibitors of CDC25B.

Methods

Expression and purification of the CDC25B catalytic domain

The CDC25B catalytic domain (372–551) and variants were expressed in E. coli BL21 (DE3) with an N-terminal GST tag. The CDC25B cDNA with an N-terminal TEV cleavage site was purchased from Genscript and cloned into a pGST-21a vector with NcoI/XhoI. Cells were grown to in either LB or labeled M9 medium. After 16 hr induction with 0.5 mM IPTG at 18 °C, E. coil cells were lysed by cell disruption in a buffer containing 50 mM Tris (pH 8.0), 150 mM NaCl, 1 mM TCEP, and 0.5 mM PMSF. The cell lysate was subjected to centrifugation, after which the soluble fraction was incubated with glutathione resin. The resin was then washed with lysis buffer, and eluted with lysis buffer containing 50 mM L-glutathione. The eluate was proteolytically cleaved with TEV protease, followed by S-75 size exclusion chromatography in buffer containing 50 mM Tris (pH 8.0), 50 mM NaCl, and 1 mM TCEP. Pure fractions were pooled and frozen at −80 °C.

NMR spectroscopy and assignment of CDC25B catalytic domain

Samples for backbone assignment were made with ¹³C¹⁵N-labeled CDC25B C473S prepared in a buffer containing 50 mM Tris (at either pH 7.0 or 8.0), 50 mM NaCl, 1 mM TCEP and 5 % D₂O. Spectra were acquired at 30 °C on a 600MHz Bruker Avance III spectrometer equiped with cryoprobe, running Topspin version 2.1. Backbone assignment was done using a series of triple-resonance experiments including HNCACB, CBCA(CO)NH, HNCA, HN(CO)CA, HNCO, HN(CA)CO, and ¹⁵N-separated NOESY-HSQC^10–12. Processing and spectral visualization was performed using NMRPipe¹³ and Sparky¹⁴.

NMR-based relaxation measurements

All relaxation measurements were acquired at 30 °C using ¹⁵N-labeled CDC25B C473S or the wild type CDC25B in buffer containing 50 mM Tris (pH 7.0), 50 mM NaCl, 1 mM TCEP, and 5 % D₂O. T₁ and T₂ relaxation measurements, as well as the ¹H-¹⁵N heteronuclear NOE measurement were obtained using experiments similar to those described by Bax et al.¹⁵. The relaxation delays for the T₁ and T₂ experiments were 10, 30, 70, 150, 330, 750, 1500, 2200ms, and 17, 34, 51, 68, 85, 102, 136, 170ms, respectively. ¹H-¹⁵N heteronuclear NOE experiments were recorded as interleaved experiments with and without NOE saturation (Figure S3). Experimental relaxation delays were set at 6 seconds for T₂ and ¹H-¹⁵N heteronuclear NOE measurements, and 4 seconds for T₁ measurements. T₂ and ¹H-¹⁵N heteronuclear NOE experiments were performed in duplicate. T₁ and T₂ values were calculated by fitting relaxation curves in Sparky¹⁴.

Relaxation dispersion experiments were recorded at 30 °C at 600MHz using a series of CPMG pulse trains as reported previously¹⁶. Two-dimensional spectra were acquired with CPMG field strengths of 25, 50, 75, 150, 300, 750, and 1000 Hz. Peak lists with intensities were used as input into the program NESSY for calculation of R_2,eff and subsequent model fitting¹⁷.

Model-free analysis

Model free analysis was performed using the modelfree4 and FASTModelfree programs^18,19. The PDB structure used was prepared using the pdbinertia program²⁰, and the initial estimate for the diffusion tensor (D_par/per) was obtained using the r2r1_diffusion program²¹. An axially symmetric diffusion model was used with the following initial parameters: rotational correlation time τ_c = 11.4 μs, diffusion tensor D_par/per = 1.25, and rotation angles θ = 6.3 °, Φ = −220 °, which describe the re-orientation of the molecule to the principal axis system. The protocol for model selection was as described by Mandel et al²². FASTModelfree was employed to iteratively adjust the global parameters between optimizations of the internal parameters (S², τ_c, etc.). The global parameters converged after 8 iterations.

Measurement of residual dipolar couplings

Samples for measurement of residual dipolar couplings were made using ¹⁵N CDC25B C473S in buffer containing 50 mM Tris (pH 7.0), 50 mM NaCl, 1 mM TCEP, with 5 % D₂O and acquired at 30 °C. Charged polyacrylamide gels for alignment were prepared as described previously²³. Samples were prepared for isotropic measurements (gel-free) and two different anisotropic measurements, including a positively charged gel (50+M) and a negatively charged gel (50-S). ¹J_NH couplings were measured for isotropic and anisotropic using the IPAP experiment²⁴. Residual dipolar couplings were calculated by subtracting the ¹J_NH couplings of the isotropic sample from the couplings for the anisotropic sample. Comparison of experimental and predicted RDC values was performed with the DC program in NMRPipe¹³, using PDB code 2A2K²⁵ for the RDC back calculations.

Results

Assignment of the catalytic domain of CDC25B

In order to probe the backbone dynamics of the catalytic domain of CDC25B, we performed the assignment of the CDC25B backbone using triple resonance experiments. Our initial studies revealed that the ¹H¹⁵N-HSQC spectrum of CDC25B shows good dispersion of peaks, consistent with well-folded protein. However, we observed that the wild-type protein precipitated out of solution over longer periods of time, leading to reduced quality of NMR spectra. We found that point mutation of the active site cysteine to serine (C473S) substantially improved the long term protein stability, likely through reducing protein oxidation. We validated that C473S mutation did not affect protein structure and dynamics because both, the backbone chemical shifts (Figure S1) as well as the backbone ¹⁵N T1, T2 relaxation times (Figure S2) for the mutant and the wild-type protein are very similar. Furthermore, it has been shown that the structure of C473S mutant is identical to the wild type protein ²⁶. As a result of increased long-term stability, we selected CDC25B C473S variant for characterization using solution NMR.

The ¹H-¹⁵N HSQC spectrum for the CDC25B C473S shows 155 backbone amides, which constitute 83% of the all expected resonances. We have assigned 92 % of observed backbone amide resonances by employing triple resonance experiments (Figure 1). The majority of unobserved amide resonances correspond to N-terminal residues as well as residues located in two loop regions, corresponding to residues 461–467, and 495–504. These regions likely are unobserved due to rapid exchange of the amide protons with water.

Figure 1. Assignment of CDC25B catalytic domain.

A) The ¹H¹⁵N-HSQC spectrum of the CDC25B catalytic domain is shown with assignments labeled. B) Cutout of the overlapped region of the spectrum with assignments.

The CDC25B catalytic domain is relatively rigid in solution

To explore the backbone dynamics of the CDC25B C473S on the pico-second to nano-second timescale, we measured the R₁ and R₂ relaxation rates as well as the steady state ¹H-¹⁵N heteronuclear NOE using methodology as previously described¹⁵. We then used Lipari-Szabo model-free analysis²⁷ to quantitatively assess fast time-scale backbone motions of CDC25B C473S. This analysis yielded an overall averaged order parameter (S²) of 0.94, indicative of a well ordered backbone with limited internal dynamic motion on the ps-ns timescale (Figure 2a). Such a relatively high average order parameter may result from a lack of signals for several amides in loop regions, likely due to their fast exchange with water.

Figure 2. The CDC25B catalytic domain is relatively rigid in solution.

A) The generalized order parameter (S²) is shown for the entire sequence as determined by model free analysis, with a box around C-terminal helix denoting the helix previously reported to be dynamic. A schematic of CDC25B secondary structure is shown above. B) Additional exchange processes (R_ex) on the millisecond timescale, predicted by model free analysis. A box around the C-terminal helix is also shown. C) Representative relaxation dispersion plots for K420 and L453 with R_2,eff shown as a function of CPMG pulse frequency in Hz. D) Crystal structure of CDC25B (PDB: 2A2K) highlighting residues experiencing exchange in the relaxation dispersion experiments. Size of sphere indicates relative magnitude of R_ex calculated by relaxation dispersion curve modeling. The C-terminal helix is denoted by the dashed ellipse.

The global correlation time determined by the analysis converged to 11.4 μs, consistent with a monomeric protein. Only a few resonances show order parameters of less than 0.85, and these are exclusively found either at one of the CDC25 termini, or within the internal loop region 461–467. No amides experience an S² of less than 0.7, indicating that no section of the assigned protein, including the C-terminal helix, experiences any significant dynamics on this timescale. The model-free analysis also indicates that several residues experience additional exchange processes in the micro-second to milli-second timescale which contribute to the observed relaxation data (Figure 2b). With the exception of E450 and R488, the remaining residues modeled with R_ex contributions appear immediately adjacent to or within loop regions, and all of the residues are further than 10 Å from the active site. To ensure the dynamics of CDC25B was not altered by the C473S mutation, we measured both R₁ and R₂ relaxation rates for the wild type protein (Figure S2). Data obtained from these experiments are highly consistent with the data recorded for the mutant, showing an average deviation of 0.048 s⁻¹ and 1.57 s⁻¹ for R₁ and R₂ respectively.

To further address the possibility of conformational exchange within the CDC25B C473S mutant catalytic domain on slower, μs-ms time scales, we carried out ¹⁵N relaxation dispersion measurements using Carr–Purcell–Meiboom–Gill (CPMG) experiments¹⁶. We detected that only a few residues experienced significant exchange in this analysis (Figure 2c–d). The majority of these residues are within or adjacent to loop regions (Figure 2d). Several residues form small clusters, for example L388-Q389-T390 or D419-K420-V422 that have relatively similar exchange rates, representing local motion of these groups in the milli-second timescale. Importantly, no motion on μs-ms time scale was observed for any residues within 10 Å of the active site or within the C-terminal helix.

Previous molecular dynamics simulations indicated that the CDC25B C-terminal residues (amino acids 531–550) are partially unfolded or disordered^8,9. On the contrary, our experimental studies of the backbone dynamics clearly demonstrates that these residues are well ordered, with an average S² of greater than 0.9 for residues in this region. Additionally, a lack of any correlated exchange in this region on the μs-ms timescale indicates that this helix does not undergo any partial unfolding.

Solution structure of CDC25B is very similar as in the crystal form

In order to assess whether the crystal structure of CDC25B correctly reflects the protein conformation in solution or whether the crystal structure is distorted by crystal packing, we measured residual dipolar couplings (RDCs). RDC measurements accurately reflect the average orientation of a bond vectors relative to a molecular frame of alignment, and thus have been shown to provide reliable data for analysis of structure in solution^28,29. Specifically, comparing calculated and experimental RDC values is a powerful tool in validating the crystal structure²⁸. In order to measure RDCs we aligned the protein using both positively and negatively charged polyacrylamide gels as described previously²³. We determined approximately 110 ¹D_HN RDCs for each alignment (Figure 3). Comparison between the experimental RDC values and the calculated values from the published CDC25B C473S catalytic domain crystal structure (PDB: 2A2K²⁵) yields a very good agreement, with correlation coefficients of 0.96 and 0.97 for the protein in positively and negatively charged alignment media respectively (Figure 3). Importantly, RDCs for both the active site residues and the C-terminal helix agree very well with the RDC values predicted from the crystal structure. For example, the correlation coefficients for the RDCs for residues in the C-terminal helix are 0.95 and 0.91 in positively and negatively charged gels, respectively. These results from two independent alignments clearly indicate that the structure of CDC25B C473S in solution is very close to the structure determined by X-ray crystallography. To additionally examine the CDC25B C473S conformation in solution, we performed an analysis of the chemical shifts for all assigned resonances using the TALOS+ program³⁰, which predicts secondary structure elements from chemical shift data. The resulting secondary elements predicted from this analysis closely match those found in the crystal structure (Figure S4).

Figure 3. Residual dipolar couplings indicate that CDC25B crystal structure is representative of its structure in solution.

A and B) Correlation of predicted and experimental residual dipolar couplings. Alignment tensor, Q-factor, and Euler angles are given. Closed circles represent the CDC25B catalytic domain sequence from 385 to 530, open circles represent the C-terminal helix from 531 to 551. A) Correlations for positively charged gel sample (50+M). B) Correlations for negatively charged gel sample (50-S).

Discussion

In this study, we found that the catalytic domain of CDC25B it is relatively rigid in solution, and that the structure in solution is very close to the crystal structure. This finding is in contrast to previous MD studies of this protein; specifically, that the 20 amino acids near the C-terminus that were predicted to be dynamic in solution by MD underwent no such dynamics experimentally^8,9. Experimental analysis of backbone dynamics on slow timescales shows no conformational exchange in either the C-terminal region or anywhere near the active site, and therefore is highly unlikely to undergo any partial unfolding as previously suggested. Because we observed no backbone dynamics near the active site, it is unlikely that there are any transient or cryptic binding pockets near this site.

In our studies we focused on experimental characterization of backbone dynamics, and did not explore any potential motion of the side chains. Also, several backbone amides are not observed in our experiments, likely due to fast exchange of the amide protons with water. However, majority of these residues are relatively distant from the active site and the C-terminal helix, which were predicted to be disordered via molecular dynamics simulations⁹.

Our study has important implications for structure-based small molecule design, since it provides strong evidence that the crystal structure of the CDC25B catalytic domain accurately reflects its conformation in solution. This study further demonstrates a need to experimentally validate findings from molecular dynamics simulations, particularly in cases where this may directly impact design of small molecule inhibitors.