Concatenation of 14-3-3 with partner phosphoproteins as a tool to study their interaction

Regulatory 14-3-3 proteins interact with a plethora of phosphorylated partner proteins, however 14-3-3 complexes feature intrinsically disordered regions and often a transient type of interactions making structural studies difficult. Here we engineer and examine a chimera of human 14-3-3 tethered to a nearly complete partner HSPB6 which is phosphorylated by protein kinase A (PKA). HSPB6 includes a long disordered N-terminal domain (NTD), a phosphorylation motif around Ser16, and a core α-crystallin domain (ACD) responsible for dimerisation. The chosen design enables an unstrained binding of pSer16 in each 1433 subunit and secures the correct 2:2 stoichiometry. Differential scanning calorimetry, limited proteolysis and small-angle X-ray scattering (SAXS) support the proper folding of both the 14-3-3 and ACD dimers within the chimera, and indicate that the chimera retains the overall architecture of the native complex of 14-3-3 and phosphorylated HSPB6 that has recently been resolved using crystallography. At the same time, the SAXS data highlight the weakness of the secondary interface between the ACD dimer and the C-terminal lobe of 14-3-3 observed in the crystal structure. Applied to other 14-3-3 complexes, the chimeric approach may help probe the stability and specificity of secondary interfaces for targeting them with small molecules in the future.


Results
Design of the 14-3-3-pB6 chimera. The C-terminal intrinsically disordered peptide of 14-3-3 does not affect the phosphopeptide-binding principles and is often removed for structural studies 33,43,44,46 . The last residue 231 of the structured 14-3-3 core is located in the vicinity of its amphipathic groove (AG). Based on these observations, we have recently constructed a C-terminal fusion of the HSPB6 peptide LRRAS 16 APL to 14-3-3σ, wherein the Ser16 residue could be phosphorylated by PKA. This chimeric protein termed pCH1 was soluble and could be crystallised. The X-ray structure showed phosphopeptide-14-3-3 interactions equivalent to those of 14-3-3 in complex with the synthetic, unfused HSPB6 phosphopeptide 33,44 . Given the N-terminal location of the 14-3-3-binding motif in the HSPB6 primary structure (a common feature of many 14-3-3 partners) and the native 2:2 stoichiometry of the 14-3-3-pHSPB6 complex 33 , the peptidic chimera construct was extended to a major part of HSPB6 including the NTD starting with residue 12 as well as the core ACD, yielding a chimera phosphorylatable at Ser16 of the HSPB6 part which we named "14-3-3-pB6" (Fig. 1A).
The fusion was expressed in E. coli in a soluble form in the presence or the absence of PKA and could be readily purified following the approach 44 developed for the peptidic chimera (Fig. 1B). The unphosphorylated version was produced in much higher yield and was efficiently phosphorylated by PKA in vitro (Fig. 1C). As a result, milligram amounts of the electrophoretically homogenous phosphoprotein could be obtained in three days including expression.

Oligomerisation of the 14-3-3-pB6 chimera.
To study the oligomeric state of the chimera, we employed size-exclusion chromatography (SEC) at different loaded protein concentrations of either unphosphorylated or phosphorylated version ( Fig. 2A,B). Importantly, at low protein concentration both versions displayed on the elution profile a single symmetrical peak with the apparent M W of ~85 kDa. The phosphorylated version had a slightly lower M W value, which is likely because phosphorylation leads to the AG occupation by the HSPB6 phosphopeptide and, therefore, to protein compaction. At higher concentrations, a second peak with an apparent M W of 157-168 kDa started to appear, again showing a slightly more compact particle size for the phosphorylated chimera, in which case also the amplitude was appreciably higher (Fig. 2). These data are in good agreement with the theoretical values for the dimer and tetramer (M w = 83.4 kDa and 166.8 kDa, respectively). More accurate M W estimates of the 14-3-3-pB6 chimera were obtained by SEC-MALLS: two peaks with M w = 82.8 (76.1% mass fraction) and M w = 143.8 kDa (23.9% mass fraction) were detected (Fig. 2C). The first peak with a negligible polydispersity (±1.05%) almost perfectly matched a dimer, while the second corresponded to a heterocomplex (Fig. 2D). Natural 14-3-3 proteins form highly soluble and rather stable dimers maintained by extensive contacts involving N-terminal α-helices, and have short flexible C-terminal tails 6,8 . A C-terminal fusion partner would not likely affect dimerisation of the 14-3-3 core, however, the question remains whether the ACD part of the fusion is folded as in native HSPB6, whose oligomeric status is dictated by the ACD dimerisation 34,47 . Since the highest-order most stable oligomeric species is a dimer for both 14-3-3σ and the HSPB6 ACD, we presumed that the observed chimera tetramerisation occurs via domain swapping with the preservation of the 14-3-3 dimer and ACD dimer interfaces (Fig. 2D). Such swapping has been observed in some crystal structures of the peptidic 14-3-3 chimeras 44 and is possible only if both the 14-3-3 dimer and ACD dimer are folded within the chimera and display their inherent propensity for dimerisation.
If ACD would be losing its dimerisation propensity, the highest possible distinct oligomer would always be dimer connected via 14-3-3 intersubunit contacts. These data suggest that the 14-3-3 and ACD parts are folded within the obtained chimera, prompting us to study this directly by differential scanning calorimetry (DSC) and limited proteolysis.
Domain structure of the chimera studied by DSC and limited proteolysis. DSC was employed to study the thermal unfolding of the chimera compared to that of individual dimers of 14-3-3 and HSPB6 and to check for the presence of folded domains in its structure. In line with the previous reports 44,48,49 , the thermogram for the 14-3-3 dimer showed the cooperative transition with T m = 61.9 °C, whereas HSPB6 revealed a transition with a higher T m of 65.8 °C (Fig. 3). The chimera unfolded in at least two stages showing two distinct peaks on the thermogram characterised by T m values of 64.2 and 66.9 °C. Taking into account that phosphotarget binding increases thermal stability of 14-3-3 44 , we assume that the first peak corresponds to the phosphopeptide-bound 14-3-3 core, whereas the second, poorly separated peak most likely corresponds to the ACD dimer stabilised by 1.1 °C due to the covalent attachment to the 14-3-3 core and possible intramolecular interactions. Therefore, the DSC data indicate that the ACD dimer is folded within the chimera and that the phosphopeptide binding stabilises the 14-3-3 core.
Limited trypsinolysis of the 14-3-3-pB6 chimera led to a slow disappearance of the initial band with the concomitant accumulation of two bands with apparent M W values consistent with those of 14-3-3σ∆C and HSPB6∆N56 (~26 and ~11 kDa, respectively) (Fig. 4). In agreement with the presence of several Arg in the NTD of HSPB6 between the 14-3-3 core and ACD (e.g., Arg27, Arg32, Arg56), which are targets for trypsinolysis 33,48 , a series of fragments between 42 and 26.5 kDa could also be detected (Fig. 4). The 26.5 kDa band remains resistant to further trypsinolysis, in line with the known stability of 14-3-3 dimers.
The main site for chymotrypsinolysis in HSPB6 resides at the Tyr53-Tyr54 dipeptide in the disordered NTD, with the ACD featuring an immunoglobulin-like fold being more resistant to cleavage 48 . Incubation of the 14-3-3-pB6 chimera with chymotrypsin resulted in the accumulation of the ~11-12 kDa peptides matching HSPB6 54-153 (10.8 kDa) and HSPB6 33-153 (12.8 kDa) fragments. The remainder of the chimera (~42 kDa) accumulated on the gel as ~30 kDa and ~26 kDa chymotryptic products exceeding and matching 14-3-3σ∆C control (~26 kDa), www.nature.com/scientificreports www.nature.com/scientificreports/ respectively (Fig. 4). This is in line with the cleavage at either Phe33 or Tyr53/54 in the NTD of HSPB6. Thus, the relative resistance of the 14-3-3 and ACD parts to trypsinolysis and chymotrypsinolysis further confirmed the foldedness of these domains within the chimera.

Isolated 14-3-3 and ACD domains show lack of direct interaction.
Given the foldedness of the 14-3-3 and ACD domains within the chimera, we questioned whether they can directly interact with each other as individual entities. Purified ACD dimer (residues 72-149) was used to probe the interaction with 14-3-3σ∆C or its chimera with the HSPB6 phosphopeptide (pCH1). We took advantage of the absence of tryptophan residues in ACD and their presence in the 14-3-3 core and employed fluorescence-assisted size-exclusion spectrochromatography (Fig. 5). As expected, 14-3-3σ∆C 44 and ACD 47 eluted as dimers with the apparent M w of 52 and 20 kDa, respectively, whereas the chimera pCH1 displayed dimers (~53 kDa) and also tetramers (~109 kDa) formed due to the interdimer phosphopeptide swap 44 . However, we failed to detect any direct binding even upon loading very high micromolar concentrations of species on the column. This suggested that this secondary interface is extremely unstable on its own, but may be formed in the native complex due to the multipoint stabilising contacts observed in the crystal structure (Fig. 6A). It was most intriguing whether tethering of 14-3-3 and ACD within the chimera would permit their interaction (Fig. 6B).
Structural analysis of the 14-3-3-pB6 chimera. We have studied the structural properties of the chimera using SEC-SAXS (Fig. 6). The SAXS profile showed linearity of the Guinier region, indicating the absence of particle interactions (aggregation or repulsion). M w estimates from the SAXS data were in excellent agreement with the MALLS-derived value (Table 1). Pairwise distance distribution function calculated from the SAXS profile of the chimera showed a curve similar in shape to that calculated from the full-atom model of the 14-3-3σ∆C-pHSPB6∆C 2:2 complex 33 (Fig. 6A,C); however, in case of the chimera, the right part of the plot extended further along the X-axis, resulting in the higher D max (12.5 nm compared to 10 nm in case of the complex) and R g value (3.51 ± 0.01 nm for the chimera vs. 3.13 ± 0.02 nm for the complex). The dimensionless Kratky plot showed the main bell-shaped peak typical of the globular proteins, a shoulder indicating the presence of another folded part/domain, and a gradual rise along the X-axis indicating the presence of flexible regions (Fig. 6D). We assume that the main peak corresponds to the chimera core based on the 14-3-3 dimer, whereas the distinct shoulder at sR g ~ 5 is derived from ACD, which is folded within the chimera according to the DSC, SEC, and limited proteolysis data.  www.nature.com/scientificreports www.nature.com/scientificreports/ Next, we have attempted a SAXS-based structural modelling of the chimera. During modelling, the 14-3-3 core (pCH1) and the ACD dimer were considered as rigid bodies supplemented by the connecting flexible regions in CORAL 50 . Modelling based on the fixed relative position of the 14-3-3 core and the ACD dimer mimicking that of the 5LTW structure resulted in satisfactory fits (best χ 2 = 3.8 in the full range of scattering data), although some oscillating residual differences between calculated and experimental scattering profiles were observed, and the smaller size of the models (R g ~3.2 nm) could not describe the low-angle part of the experimental SAXS profile indicating a less compact size (R g ~3.5 nm) (Fig. 6E). These results are in line with the differences observed in the p(r) plots (Fig. 6C). In contrast, relieving the 14-3-3/ACD interface (Fig. 6B) allowed us to obtain larger models with different ACD position relative to the 14-3-3 core which reconciled the discrepancy and provided excellent fits to the SAXS profile in the full range of scattering data (χ 2 ranged from 1.16 to 1.27; Fig. 6E).

Discussion
14-3-3 protein complexes are important regulatory nodes of PPI networks mediating multiple intracellular processes in norm and pathology. The constantly expanding interactome of 14-3-3 protein family requires adequate research efforts, preferably in the high-throughput format, and the ability to structurally characterise 14-3-3 complexes. This is necessary for unravelling the fundamental binding mechanisms and selectivity of 14-3-3 proteins and also for their use in drug discovery 13 . Yet, just a few structures of protein complexes involving 14-3-3 have been solved until today (PDB entries 1IB1, 5LTW, 5N6N, 6GN8). To aid in structural studies, Sluchanko et al. have recently proposed the chimeric approach and demonstrated its applicability to solve structures of 14-3-3/ phosphopeptide complexes, which is scalable to a high-throughput format 44 .
Here we describe a 14-3-3 chimera with an almost complete phosphotarget, i.e. the relatively well-characterised partner HSPB6. We have shown that the chimera encompassing the 14-3-3 core and the major part of HSPB6 (residues 12-153) including phosphorylation of Ser16 can be readily produced in a soluble form in E.coli and has a controllable stoichiometry of binding partners (Fig. 1). Importantly, it is amenable for structural studies owing to the foldedness of the two interacting entities, the 14-3-3 dimer and the fused ACD dimer, with their dimeric interfaces preserved (Figs 2-4). The 14-3-3-pB6 chimera dimer reproduces the main structural features of the previously characterised native 2:2 complex of 14-3-3 and pHSPB6, at the same time, forming a somewhat more flexible structure. Indeed, in the crystal structure of the 14-3-3/pHSPB6 complex, the direct interface formed between one of the 14-3-3 subunits and the ACD dimer measures only 400 Å 2 , although a salt bridge between Arg224 14-3-3 and Glu86 ACD is present 33 . As we show here using SEC, this interface is rather unstable since an isolated ACD dimer does not bind to the 14-3-3 dimer (Fig. 5). This is in line with previous observations that also the full-length unphosphorylated HSPB6 (thus void of N-terminal binding in the AG) does not significantly interact with 14-3-3 [35][36][37] . Importantly, the crystal structure of the 14-3-3/pHSPB6 complex reveals that the very N-terminus of HSPB6 (residues 2-10) containing the VPV motif 34 patches the β4/β8 groove, while the immediately following phosphorylated motif is anchored in the AG of the 14-3-3 subunit (Fig. 6A). This arrangement restricts the position of the ACD relative to the adjacent 14-3-3 molecule. In the chimera, the C-terminus of 14-3-3 is linked to the HSPB6 sequence starting www.nature.com/scientificreports www.nature.com/scientificreports/ at position 12 only (Fig. 1A), making the patching of the β4/β8 groove on the ACD impossible (Fig. 6A). In line with that, our SAXS analysis suggests that, within the chimera, the relative position of the 14-3-3 dimer and the ACD dimer is variable (Fig. 6). In summary, our results reveal a complex dynamics of the 14-3-3/pHSPB6 association where hierarchy of stabilising factors are at play.
We assume that, like in the HSPB6 case, the secondary interfaces of other crystallographic 14-3-3 complexes may not be fully occupied in solution and require additional stabilising factors. However, this very feature could turn advantage in the future development of much more selective small-molecule stabilisers compared to targeting the primary, phosphopeptide binding regions characterised by the certain level of degeneracy and being rather similar in the most known 14-3-3/phosphotarget complexes.
Importantly, the chimeric approach may be applicable to many protein partners of 14-3-3 characterised by single, N-terminally located 14-3-3 binding phosphosites, whereas utilization of the known preferential heterodimerisation of certain 14-3-3 isoforms 3,4 can expand this approach to study multiply phosphorylated and ternary complexes. Co-expression of 14-3-3/client chimeras can be achieved in the presence of the appropriate protein kinase(s) 45 . These concepts could see yet further progression by benefiting from the recent development of engineered E.coli strains and modified translational machinery components which enable the production of phosphoproteins using amber codon suppression 51,52 .

Methods
Cloning, protein expression and purification. Chimera containing the C-terminally truncated human 14-3-3σ (Uniprot ID P31947; residues 1-231, 14-3-3σ∆C) bearing on its N-terminus a His 6tag cleavable by 3 C protease and the phosphorylatable peptide of the small heat shock protein HSPB6 (residues [12][13][14][15][16][17][18][19], connected to the 14-3-3 core with the help of the GSGS linker (CH1), was described previously 44 . The 14-3-3σ sequence was modified to reduce surface entropy 53 by introducing the 75 EEK 77 → AAA amino acid replacements ("clu3" mutant) and, after 3 C cleavage, contained three extra residues GPH at Figure 6. Structural analysis of the 14-3-3-pB6 chimera using SAXS. (A) Crystal structure of the 14-3-3σ∆C-pHSPB6∆C complex (PDB 5LTW). The 14-3-3 dimer is shown as cartoon. The HSPB6 ACD dimer is shown as a molecular surface, with residues involved in the interface with 14-3-3 highlighted in light green. The partially ordered NTD of one HSPB6 chain is shown in magenta. In addition, the N-terminal arms of the peptidic pCH1 chimera (when superimposed on the full complex) are drawn as cyan and green lines. Phosphoserines are represented by red spheres. (B) Schematics showing that in the absence of the β4/β8 patching, the 14-3-3/ACD interface may be preserved or not preserved. (C,D) Comparison of the p(r) functions (C) and the dimensionless Kratky plots (D) for the chimera (experimental SAXS data processed by GNOM 63 ) and the 5LTW structure supplemented with the missing loops 33 (calculated from the model by CRYSOL 50 and GNOM 63 ). (E) The fits of the best among each of the two types of models depending on whether the 14-3-3/ACD is preserved or not preserved to the SAXS data and the associated residuals (∆/σ) shown on top.
Chemically competent cells of the Escherichia coli BL21(DE3) strain were transformed using the pET28_ His_3C_14-3-3σ(clu3)∆C_B6.12-153 (kanamycin resistance) and pACYC-PKA (chloramphenicol resistance) 45 plasmids simultaneously or, alternatively, using only the chimera plasmid to produce unphosphorylated protein. Protein expression was induced by the addition of IPTG to a final concentration of 0.5 mM and continued for 4 h at 37 °C. The overexpressed proteins were efficiently purified by subtractive immobilized metal-affinity chromatography (IMAC) and gel-filtration as described for the peptide chimeras 44 . The chimera was fully soluble and could be concentrated to above 10 mg/ml for storage at −80 °C. Given the much higher yield of the chimera expressed in the absence of PKA, it was further phosphorylated in vitro essentially as described elsewhere 33,44,54 . The most optimal conditions leading to a complete shift of the chimera band on the native-PAGE 55 included: 0.1 mg/ml chimera, 30 µM ATP, 0.01 mg/ml PKA, 4 mM MgCl 2 , 40 min at 37 °C. Protein concentration was determined spectrophotometrically using the sequence-specific extinction coefficient at 280 nm equal to 0.67.
The same setup was used to study direct binding between 14-3-3σ∆C or phosphorylated CH1 containing two Trp residues per chain and Trp-lacking HSPB6 ACD. In this case, the profiles were followed simultaneously by www.nature.com/scientificreports www.nature.com/scientificreports/ absorbance at 230 nm, and by Trp fluorescence (excited at 298 and recorded at 360 nm, at a low voltage) using fluorescence-assisted spectrochromatography on a Varian ProStar 335/ProStar 363 system. The experiments were performed three times with the most typical results presented.
The absolute masses of the species formed by the 14-3-3-pB6 chimera were analyzed on a Superdex 200 Increase 10/300 column (GE Healthcare) using multiparametric detection as described earlier 56 . Protein sample (75 µl; 10.14 mg/ml) was pre-incubated for 15 min at room temperature and then loaded on the column equilibrated by SEC buffer. The column was operated at 20 °C at a flow rate of 0.5 ml/min. Multi-angle laser light scattering (MALLS) was measured using a Wyatt Technologies miniDawn TREOS module coupled to an OptiLab T-Rex refractometer for protein quantitation (dn/dc = 0.185 was taken). The MALLS detector was calibrated relative to the scattering from toluene and, in combination with the refractometric signal, was used to determine the M W distribution of species eluting from the SEC column.
Small-angle X-ray scattering (SAXS) and modelling. The SAXS data (I(s) vs s, where s = 4πsinθ/λ, 2θ is the scattering angle and λ = 1.24 Å) for the 14-3-3-pB6 chimera were collected at 20 °C in parallel with the MALLS/RI detection in a SEC-SAXS format to ensure separation of the particles of interest from undesired oligomeric species and aggregates at the EMBL P12 beamline (PETRA III, DESY Hamburg, Germany 57 ). This was achieved by the equal division of the flow between the SAXS (3600 × 1 s frames) and the MALLS/RI detection modules 58 , enabling simultaneous data collection. Data reduction, radial averaging and statistics analysis were done using the SASFLOW pipeline 59 , the SEC-SAXS data were processed using CHROMIXS 60 . ATSAS 2.8 package 61 was used for data analysis and modelling. PRIMUS 62 was used to perform Guinier analysis from which the radius of gyration, R g , and extrapolated forward scattering, I(0), were determined (lnI(s) versus s 2 that were linear in the sR g range reported in Table 1). The pairwise real-space distance distribution function, p(r), was calculated using GNOM 63 that provided additional R g and I(0) estimates and the maximum particle dimension, D max . The Porod volume and other structural parameters in solution are presented in Table 1.
In order to model the 14-3-3-pB6 chimera against the SAXS data, the crystal structure of the 14-3-3 chimera with the HSPB6 phosphopeptide (residues 12-19) (PDB 5OKF) was first modified in Coot 64 to include the linker and some loop residues absent from the electron density map. This resulted in the full-atom model of the dimeric 14-3-3 core (GPH followed by residues 1-231 of human 14-3-3σ with clu3 mutations) tethered with the HSPB6 phosphopeptide (residues 12-20) by the GSGS linker (247 residues overall). The dimeric ACD part (residues 72-153) was based on the previous co-crystal structure with 14-3-3 (PDB 5LTW 33 ). The first scenario was based on the relative orientation of the 14-3-3 core and ACD dimer as in the 5LTW structure (with the 14-3-3/ACD interface preserved and the N-terminal tails of the HSPB6 set flexible). The second scenario implied free movements of the ACD dimer tethered by the N-terminal tail of the HSPB6 subunits to the 14-3-3 core. In all cases, the program CORAL 50 was used in multiple parallel runs to model the structure of the full dimeric 14-3-3-pB6 chimera (2 × 380 residues) by representing the missing parts of the N-terminal tail of HSPB6 (residues 21-71) in both chains (12.7% by mass) by a C α -trace to minimize the discrepancy between the experimental SAXS profile in a range of 0 < s < 0.3 Å −1 and that calculated from models. The obtained models were validated against the SAXS curve for the full range of scattering data using CRYSOL 50 . Two ways of interconnecting the phosphopeptides bound to 14-3-3 and the two ACD subunits have been attempted and yielded similarly fitting models.
The SAXS profile, the Kratky plot, and the pairwise distance distribution function for the full-atom model of the 14-3-3σ∆C-pHSPB6∆C complex based on the 5LTW crystal structure supplemented with missing loops were calculated using CRYSOL 50 and GNOM 63 . DSC. 14-3-3-pB6 chimera (1.1 mg/ml) or individual dimers of 14-3-3 and HSPB6 (1 mg/ml each) were dialysed overnight against a 20 mM HEPES-NaOH buffer, pH 7.5, containing 150 mM NaCl and subjected to DSC on a VP-capillary DSC (Malvern) at a heating rate of 1 °C/min. Thermograms were processed using Origin Pro 8.0 and transition temperature (T m ) was determined from the maximum of the thermal transition 48 .
Limited proteolysis. Proteolysis of the phosphorylated chimera (1 mg/ml) was performed at 37 °C in buffer P (20 mM Tris-HCl, pH 7.6, containing 150 mM NaCl, 15 mM β-mercaptoethanol) using either the TPCK-treated trypsin (mass ratio protease:substrate equal to 1:12,000) or the TLCK-treated chymotrypsin (mass ratio protease:substrate equal to 1:6,000) by incubating the mixtures for different times. The reaction was blocked by addition of the SDS-sample buffer containing PMSF up to the final concentration of 7 mM. The samples were then boiled and analyzed by SDS-PAGE on 15% polyacrylamide gels 65 . The apparent M W values were determined using GelAnalyzer 2010a (http://www.gelanalyzer.com/index.html) and compared to those of 14-3-3σ∆C and HSPB6 ACD used as controls as well as calculated masses obtained using massXpert 66 .