Glycine substitution in SH3-SH2 connector of Hck tyrosine kinase causes population shift from assembled to disassembled state

Abstract

A combination of small angle X-ray scattering (SAXS) and molecular dynamics (MD) simulations based on a coarse grained model is used to examine the effect of glycine substitutions in the short connector between the SH3 and SH2 domains of Hck, a member of the Src-family kinases. It has been shown previously that the activity of cSrc kinase is upregulated by substitution of 3 residues by glycine in the SH3-SH2 connector. Here, analysis of SAXS data indicates that the population of Hck in the disassembled state increases from 25% in the wild type kinase to 76% in the glycine mutant. This is consistent with the results of free energy perturbation calculations showing that the mutation in the connector shifts the equilibrium from the assembled to the disassembled state. This study supports the notion that the SH3-SH2 connector helps to regulate the activity of tyrosine kinases by shifting the population of the active state of the multidomain protein independent of C-terminal phosphorylation.

Introduction

Src kinases play a role in critical biological processes including cell growth and proliferation [1]. Activity is tightly regulated and failures in these mechanisms lead to diseases such as cancer [2]. Src-kinase family members share a conserved multidomain structure, which comprises three domains: Src homology 3 domain (SH3), Src homology 2 domain (SH2) and a highly conserved kinase domain [3, 4] as shown in Figure 1. SH3 and SH2 domains serve multifunctional roles; activity of Src kinases may be controlled by binding of signaling peptides or intramolecular interactions [3, 4]. Crystal structure of the assembled (downregulated) Hck (1QCF) shows a compact form stabilized by interactions of the N- and C- terminus of the catalytic domain with the SH3 and SH2 domains [5]. Cellular factors may also bind these domains during kinase activation and transition to the disassembled state.

Figure 1:

Crystal structure of Hck from the crystal structure from Schindler et al [5] (pdb id: 1QCF). The polypeptide SH3-connector-SH2-linker-[catalytic domain]-C-terminus is rendered with SH3 (yellow), SH3-SH2 connector (orange), SH2 (green), the linker between SH2 and catalytic domain (magenta), catalytic domain (cyan) with the activation A-loop (yellow), and C-terminus tail (red).

Intramolecular interactions between the SH2 domain and the C-terminal tail as well as the interactions between the SH3 domain and the SH3 binding motif on the SH2-kinase linker inhibit the activity of Src kinases [3, 4] by maintaining the assembled state. The connector between the SH3 and SH2 domains was originally thought to be irrelevant to activation [6]. However, Young et al [7] found that mutation of three residues to glycine in the connector, ~50 Å away from the catalytic site, could make c-Src kinase constitutively active. On the basis of molecular dynamics simulations and the mutational data, they proposed that the connector acts as a structural “inducible snap-lock” that is intrinsically rigid and required for the SH3 and SH2 modules to serve as an effective “clamp” in the assembled state of the enzyme, inhibiting the activity of the catalytic domain. Glycine mutations in this region render the connector too flexible and unable to retain the correlated motions of the SH3 and SH2 domains [7], lessening the ability of the SH2-SH3 clamp to inhibit the kinase in the context of the assembled state. In other words, by reducing the effectiveness of the SH2 and SH3 modules to inhibit the catalytic domain in its assembled state, the glycine mutations in the SH3-SH2 connector undermine the autoinhibitory character of the assembled state. This conjecture assumes, not that the glycine mutants cause a destabilization of the assembled state leading to a catalytically active disassembled state, but that the increased kinase activity is caused by changes in internal fluctuations of the SH2-SH3 domains.

The mechanism proposed by Young et al focused entirely on the internal dynamics of the assembled state. The implicit assumption is that the effect of the glycine mutations in the connector on the disassembled state imagined to correspond to a set of highly fluctuating conformations should be negligible. Representative configurations of the disassembled state are illustrated in Figure 2 [8]. An alternative explanation was proposed by Faraldo-Gómez and Roux [9] based on molecular dynamics (MD) potential of mean force (PMF) and free energy perturbation (FEP) simulations. The PMF calculations showed that, while the glycine mutations in the SH3-SH2 connector negligibly affect the assembled state, they significantly increase the fluctuations of the disassembled (active) state. On the basis of these results, it was argued that the glycine mutations would necessarily shift the equilibrium from the assembled (inactive) state toward the disassembled (active) state. This explanation implies increased flexibility of the connector lends more conformational entropy to the disassembled state.

Figure 2:

The nine clusters of conformation states generated by Yang et al [8] used to analyze the small angle X-ray scattering SAXS data on Hck kinase. The color coding for the different structural elements are: SH3 (yellow), SH3-SH2 connector (orange), SH2 (green), the linker between SH2 and catalytic domain (magenta), catalytic domain (cyan) and C-terminus tail (red).

The mechanisms proposed by Young et al [7] (the mutations mainly act on the assembled down-regulated form of the kinase by undermining the inhibitory effect of the SH2-SH3 clamp) and Faraldo-Gómez and Roux [9] (the mutations lead to an increase in the population of the disassembled active state) are fundamentally different. Here, we aim to characterize the structures of wild type Hck and the glycine mutant in solution. Small angle X-ray scattering (SAXS [10]) is a powerful technique to study the structure of macromolecules in solution, particularly for multidomain proteins or protein multimers [8, 10–16]. Recently, computer simulations based on coarse-grained models [8, 17] combined with SAXS were used to characterize the structures in solution in a quantitative way for Hck and ESCRT-III CHMP3 [13]. This integrated strategy [8] is used in the present study to determine how the configuration of Hck is affected by glycine substitutions in the SH3-SH2 connector.

Results and Discussion

A coarse-grained (CG) model of Hck was previously used to generate sets of conformations that were clustered into 9 possible states of Hck shown in Figure 2 [8]. The clustering analysis of the configurations collected in MD simulations of the CG model according to pair distances and calculated SAXS profiles is described in the Method section. In this model state 1 is the assembled state and is most similar to the Hck crystal structure 1QCF. Fitted and experimental SAXS profiles for wild type Hck are shown in Figure 3a. The composition of nine states in solution were calculated from the fitting: 75% state 1, 2% state 3 and 20% state 6 (Figure 3b). Figure 3c presents the fitted and experimental SAXS profiles for the glycine mutant. The populations of the mutant (Figure 3d) show that disassembled states 3–9 increase to 5%, 6%, 12%, 36%, 4%, 7% and 2% respectively.

Figure 3:

Analysis of the SAXS data. a) The fitted and experimental SAXS profiles for the wild type. b) The population of all states from SAXS fitting is shown. The populations are 0.753, 0.006, 0.017, 0.005, 0.008, 0.196, 0.004, 0.008 and 0.003for states 1–9 respectively. c) The fitted and experimental SAXS profiles for the mutant. d) The populations are 0.242, 0.026, 0.053, 0.058, 0.122, 0.363, 0.039, 0.072 and 0.024 for states 1–9 respectively.

Guinier analysis of the glycine mutant shows a larger radius of gyration (Rg), 29.61 Å compared with 28.04 Å for the wild type. This indicates the population of the disassembled state is increased, an observation that the BSS-SAXS analysis confirms; the population of state 1 drops to 24% in the glycine mutant compared to 75% for wild type (Figure 3a and 3b).

The population of the various accessible states in solution for the mutant can be roughly estimated based on the populations in the wild type protein by using free energy perturbation (FEP),

$$p_{i,mut}=p_{i,wt}{\langle e^{-\left(E_{mut}-E_{wt}\right)/k_{B}T}\rangle }_i/c$$ (2)

where p_i,mut is the population of state i for the mutant, p_i,wt is the population of state i for the wild type, c is a constant for the normalization of p_i,mut, and <…>_i represents an average over all configurations based on the unperturbed ensemble for state i. The torsion interactions at the glycine substitutions are turned off in the Gō model of the mutant to approximately account for the deletion of side chains that renders the backbone of the connector more flexible. The result of FEP analysis is shown in Table 1. The assembled state, cluster 1, gives the smallest value of $\langle e^{-\left(E_{mut}-E_{wt}\right)/k_{B}T}\rangle$, since the protein fluctuations around the crystallographic structure are restricted. State 3 is close to the assembled state, but the orientation of the SH3 domain is tilted and the binding between the SH3 domain and the PPII motif along the linker is reduced. Consequently, state 3 gives the highest value of $\langle e^{-\left(E_{mut}-E_{wt}\right)/k_{B}T}\rangle$ due to the twisted connector between SH3 and SH2. FEP simulations suggest that the population of the assembled state decreases to 0.57 for the mutant. The populations of disassembled states 3 and 6 increase to 0.08 and 0.26 respectively. Since FEP only works well for sufficiently small perturbations and the exponential average may not converge over one hundred configurations in each state, the result of FEP does not quantitatively match the populations obtained from our BSS-SAXS fitting for the mutant. However, it does show the same trend; the equilibrium shifts from the assembled state to the disassembled state.

Table 1.

The state population of the mutant estimated from the population of wild type using FEP analysis

State	pi,wt	$\langle e^{-\left(E_{mut}-E_{wt}\right)/k_{B}T}\rangle$	−kTln()	pi,mut_FEP	pi,mut_Exp
1	0.753	33.41	−2.09	0.569	0.242
2	0.006	53.76	−2.38	0.007	0.026
3	0.017	211.01	−3.19	0.081	0.053
4	0.005	154.46	−3.0	0.019	0.058
5	0.008	177.86	−3.1	0.033	0.122
6	0.196	58.02	−2.42	0.258	0.363
7	0.004	111.68	−2.81	0.009	0.039
8	0.008	85.81	−2.65	0.016	0.072
9	0.003	132.71	−2.91	0.008	0.024

The regulatory mechanism described by Young et al [7] proposes that the glycine mutations mainly act on the assembled down-regulated form of the kinase by undermining the inhibitory effect of the SH2-SH3 clamp. Implicitly, this presumes a dependence upon the phosphorylation of the C-tail, which is required to establish the assembled form. However, our data (and that of Yang et al [8]) demonstrate that wild type Hck protein expressed in bacteria is predominantly in the assembled down-regulated conformation (Figure 3), even though the tyrosine in the C-tail is unphosphorylated. The BSS-SAXS analysis provides a view of the possible configurations of Hck in solution. As revealed by the analysis, substitution of residues in the SH3-SH2 connector with glycine alters the distribution of Hck across the clusters of the model. As shown in Figure 3, the disassembled state is more prominent in the glycine connector mutant than the wild type, in the absence of binding partners and phosphorylation of the C-tail. This observation indicates that the mechanism by which glycine substitutions in the SH3-SH2 connector increase activation does not require phosphorylation of the C-terminal tyrosine.

In ending this discussion, it is of interest to mention additional computational MD studies that have helped elucidate the nature of the allosteric coupling between the regulatory SH2 and SH3 modules and the active or inactive conformation of the kinase domain [18, 19]. The calculated free energy landscapes mapping the inactive-to-active conformation transition show that a significant thermodynamic penalty on the active conformation of the kinase domain is imposed when the SH3-connector-SH2-linker structural element and the catalytic domain are engaged to form the assembled state. Complete kinase activation is possible only if transient visits of the catalytic domain to active-like conformations are allowed [20], and the catalytically active state is subsequently locked by trans-autophosphorylation of Tyr416 in the A-loop [21]. In the assembled state, the inhibitory information from the SH2-SH3 tandem is allosterically transmitted via the linker region connecting the SH2 domain and the N-terminal portion of the kinase domain [18]. While the SH2 and SH3 domains are needed to stabilize the linker region in an “inhibiting conformation”, it is actually the linker/N-terminal region itself that is responsible for locking the kinase domain in an inactive conformation. For instance, the calculations showed that the catalytic domain is maintained in its inactive conformation by the linker restrained in its “inhibiting conformation”—even in the absence of the SH2 and SH3 domains [18]. This result suggests that the principal effect of the glycine mutations in the SH3-SH2 connector is unlikely to be on the assembled state. The important role of the linker region in the allosteric regulation of kinase activity highlighted by these computations is consistent with experiments [22]. The present analysis based on BSS-SAXS supports the shift in equilibrium mechanism proposed by Faraldo-Gómez and Roux [9].

Methods

Cloning and protein expression and purification.

Wild-type human HCK3D (residues 85–531, chicken c-Src numbering) or the Ser142Gly, Thr145Gly, Glu146Gly mutant of HCK3D were co-expressed in bacteria with YopH phosphatase and purified by the method of Seeliger et al [23] as described previously10. The pHCK3D and YopH Duet plasmids were provided by Markus Seeliger and John Kuriyan. pHCK3D is a pET-28a vector (Novagen) modified to yield a tobacco etch virus (TEV) protease cleavable N-terminal hexahistidine tag. It contains a fragment of human HCK spanning the SH3, SH2, and catalytic domains, and the C-terminal tail (residues 85–531, chicken c-Src numbering). YopH Duet is a pCDFDuet-1 vector (Novagen) containing full-length YopH phosphatase from Yersinia. The pHCK3D.142G.145G.146G mutant was generated using the QuikChange Site Directed Mutagenesis kit (Stratagene) with pHCK3D as a template and mutagenic primers from Integrated DNA Technologies. Recombinant clones were verified by sequencing across the cloning/expression region.

The two plasmids containing the kinase and the phosphatase were co-transformed into Escherichia coli BL21(DE3) cells. 1-liter cultures were grown in Terrific Broth with 50 μg/ml each of kanamycin and streptomycin at 37°C to an OD600 of 1.0, cooled for 1 hour with shaking at 18°C, and induced overnight at 18°C with 1mM IPTG. Cells were harvested by centrifugation at 4000 g for 10 minutes at 4°C. Pellets were resuspended in 30 ml of 20 mMTris (pH 8.0), 500 mMNaCl, 5% glycerol, 20 mM imidazole (NiA buffer) and lysed by 4 cycles of homogenization at 15,000 psi using an Avestin homogenizer. The lysate was cleared by centrifugation at 125,000 g for 50 minutes at 4 °C. The supernatant was loaded onto 5 ml NiNTA resin (Qiagen) equilibrated in NiA buffer. After washing with 10 column volumes of NiA buffer, the protein was eluted with 3 column volumes of NiB buffer (NiA buffer plus 500 mM imidazole). The eluted protein was cleaved with 1 mg TEV per 25 mg kinase at 4 °C overnight while dialyzing against 2 liters of 20 mMTris (pH 8.0), 100 mMNaCl, 5% glycerol, 0.5 mM EDTA, 0.5 mM TCEP in a 12–14kDa molecular weight cutoff membrane. The dialyzed protein was diluted 1.3 fold with water, spun at 14,000 g for 10 minutes at 4°C, and loaded onto an ion exchange column (HiTrap Q FF, GE Lifescience) equilibrated with 20 mMTris (pH 8.0), 0.5 mM TCEP (QA buffer). The kinase was eluted with a linear gradient of 0–50% QB buffer (QA buffer plus 1 M NaCl). The peak fractions were analyzed by SDS-PAGE and the fractions containing kinase were pooled. A 1.5 fold molar excess of the inhibitor PP1 (Tocris) was added to prevent aggregation and the pooled fractions were concentrated using an Amicon Ultra-15 centrifugal filter device with a 30kDa molecular weight cutoff membrane. The concentrated sample was loaded onto a size-exclusion column (Superdex 200 10/300 GL, GE Lifecscience) equilibrated with 50 mMTris (pH 8.0), 100 mMNaCl, and 0.5 mM TCEP. HCK was eluted at the volume expected for monomeric protein. No aggregation was detected in the void volume of the column. Sample concentrations were adjusted for SAXS data collection.

SAXS experiments and analysis.

SAXS data were collected at the ALS beamline 12.3.1 (SIBYLS) of Lawrence Berkeley National Laboratory. Data was collected on a CCD area detector. Buffer subtraction and integration were carried out through beam-line software developed for this purpose. The wavelength λ = 1.0 Å and sample-to-detector distances were set to 1.5 m. The scattering vector is defined as q = 4π sinθ/λ, where 2θ is the scattering angle. All experiments were performed at 15 °C, and data were collected and processed as previously described [24]. Protein buffer for SAXS experiments was 50mM Tris pH8.0, 100mM NaCl, 1mM DTT. Data were measured at three concentrations, 1, 2.5 and 5 mg/ml, and three exposures, 0.5, 1.0 and 6 seconds. The data at 1mg/ml and 1 second exposure were used for analysis. Concentration was not seen to affect scattering. Pair distribution functions, P(r), were calculated by the program GNOM [25].

Monte Carlo and FEP simulations.

As in previous work [8], a modified Gō model is used in exhaustive configuration sampling and all structures are clustered into nine states (shown in Figure 2) using pair distances and calculated SAXS pattern [8, 26] as the metric of structure similarity. See Yang et al [8] for more details about the method. The nine clusters of structures for Hck as shown in Figure 2 are used as a “basis set” for SAXS profile fitting. State 1 corresponds to the assembled inactive kinase, and other states correspond to partially or fully disassembled states that are presumably active. The basis-set supported SAXS (BSS-SAXS) analysis method introduced by Yang et al [8] was employed to interpret the experimental data. The following equation is used as an objective function for fitting the SAXS scattering profile,

$${\chi }^2\left(\left\{p_i\right\},c\right)=\sum _{q_{\text{min }}}^{q_{\text{max }}}{\left(\frac{c\sum _{i=1}^{N_s}{p}_i{I}_i(q)-I_{\text{exp}}(q)}{\sigma (q)}\right)}^2$$ (1)

where p_i is the population of cluster i, q is the scattering vector, I_i(q) is the average scattering profile for the state i, I_exp(q) is the experimental scattering profile, σ(q) is the experimental standard deviation, c is a scaling constant to be determined by fitting and N_s is the size of basis sets (N_s =9 in our study). As in our previous work [8], the fractional population p_i and their uncertainties were determined by a Bayesian-based Monte Carlo (MC) sampling of the probability distribution exp(−χ²). A total of 500,000 MC steps was used for sampling. The results are independent on the initial guess. To estimate the state populations of mutant based on the experimental data of the wildtype, 100 configurations were randomly selected in each of the 9 clusters (Figure 2) for FEP calculations The potential energies in wild type (E_wt) and glycine mutant (E_mut) were evaluated for each snapshot and the average of exp(−(E_mut -E_wt)/k_BT) was calculated over all snapshots selected in each cluster. The dihedral interactions along the backbone were turned off to mimic the triple-gly mutant for five dihedrals, 136-137-138-139, 137-138-139-140, 139-140-141-142, 140-141-142-143, and 141-142-143-144.