Design of a protein tag and fluorogenic probe with modular structure for live-cell imaging of intracellular proteins† †Electronic supplementary information (ESI) available. See DOI: 10.1039/c5sc02351c

Quick and no-wash labeling of intracellular proteins was achieved in live cells using a PYP-tag mutant and a membrane-permeable fluorogenic probe with modular structures.


Materials and instruments
General chemicals were of the best grade available and supplied by Tokyo Chemical Industries, Wako Pure Chemical Industries, Sigma-Aldrich Chemical Co, and Kishida Chemical Co, and were used without further purification. Enzymes for subcloning experiments were purchased from Takara Bio and New England Biolabs.
Oligonucleotides were purchased from Greiner Bio-One or Life technologies. The plasmid pTag-BFP-actin was purchased from Evrogen.
NMR spectra were recorded on a Bruker AVANCE III HD 500 instrument at 500MHz for 1 H and 125 MHz for 13 C NMR. High-resolution mass spectra (HRMS) were recorded on a JEOL JMS-700. ESI-TOF MS was recorded on a Waters LCT-Premier XE.
Fluorescence spectra were measured using a Hitachi F7000 spectrometer with a photomultiplier voltage of 700V. UV-vis absorption spectra were obtained using a Shimadzu UV-2450 spectrometer. Fluorescence microscopic analyses were performed using a confocal laser scanning microscope (Olympus, FLUOVIEW FV10i) equipped with a 60× lens.

Preparation of plasmids
Plasmids for the single cationic mutants (pET21b-PYP-D53R, PYP-D71R, PYP-E74R, and PYP-D97R) were prepared using a QuikChange Mutagenesis Kit, using the primers listed below. CGC was selected as the triplet sequence coding for arginine. The mutation site is highlighted in red. For construction of PYP-4R, PYP-D71RE74R was first prepared, a point mutation was induced at D97R, and finally a mutation was induced at D53R, using above-mentioned primers. DNA sequencing of all the constructs confirmed that the genes were correctly inserted in the target plasmids.

PYP-E74R, PYP-D97R, and PYP-4R)
PYP-WT and single cationic mutant proteins were expressed in BL21(DE3) using the above-mentioned plasmids and purified as previously described. 1 His-tags were introduced for purification by Ni-NTA affinity chromatography and were not removed from the final proteins.
A modified procedure was employed for the preparation of PYP-4R. pET21b-PYP4R was expressed in BL21(DE3), harvested, and lysed following a previously described procedure. 3 The cell lysate was purified on an affinity column loaded with IMAC resin (Roche, cOmplete His-Tag Purification Resin), following the manufacturer's protocol.
Protein purity and size were assessed by SDS-PAGE. The purified proteins were dissolved in assay buffer (20 mM HEPES, 150 mM NaCl (pH7.4)), flash-frozen with liquid nitrogen, and stored at -80 °C.

Protein labeling reactions visualized by SDS-PAGE
SDS-PAGE analysis confirmed covalent binding between FCANB and each of the mutants ( Figure S3). PYP-tag (10 M) was reacted with FCANB (15 M) in assay buffer at 37°C for 1 hour. The reaction mixtures were subsequently analyzed by SDS-PAGE. After fluorescence images were obtained using the AE-6935B VISIRAYS (ATTO Corporation), the gel was stained with Coomassie Brilliant Blue.

UV-Vis spectra of FCANB and PYP WT
The absorption spectra of FCANB with or without PYP were determined in assay buffer.
FCANB (10 M) and PYP WT (15 M) were reacted at 37 °C for 6 h prior to measurement. using Antechamber in AmberTools 13. We used a representative structure from the top-ranked cluster as the input. REDUCE 3.03 was used to add hydrogen atoms to tFCANB before running Antechamber. The AM1-BCC charge method was selected in

Molecular dynamics of PYP-tFCANB binding in explicit solvent
The WT and D97R mutant forms of the six representative structures were used as the starting conformations for the MD simulations in which the tFCANB ligand was in a "bound" conformation. For simulations starting from an "unbound" conformation, we applied random rotations and translations to the tFCANB molecule, retaining those that satisfied the following criteria: less than three heavy atoms had a minimum distance of < 3.5 Å between tFCANB and PYP; the minimum distance from both the Cα atom of Cys-69 and Glu-or Arg-97 to any heavy atom in tFCANB was less than 15 Å. We generated two conformations for each selected structure, and the same relative tFCANB position and orientation were used for both WT and D97R simulations.
The details of the MD trajectories were as follows. We prepared topology files by tleap in AmberTools 13, 4 and transformed these into Gromacs 5 format by acpype. 6 The systems were solvated in a TIP3P explicit water box with a 10 Å edge distance. Sodium or chloride ions were added to neutralize the system as needed. The AMBER99SB forcefield was used for components of the system other than tFCANB. We used Gromacs 4.5.5 for all MD simulations. The particle mesh Ewald (PME) method was used to calculate the electrostatic potential. The cutoff distances for non-bonded interactions and the timestep for MD were set to 10 Å and 2 fs, respectively. The LINCS 7 algorithm was utilized to constrain distances between hydrogens and bound heavy atoms. The Parrinello-Rahman 8 and v-rescale methods were used for pressure and temperature coupling, respectively, if not otherwise specified below. The MD protocol was as follows. First, 5000 steps of energy minimization were performed by the steepest descent method followed by 20ps of MD at 300K to equilibrate waters and ions in the NVT ensemble, while applying 1000 kJ mol -1 nm -2 isotropic positional restraints to heavy atoms in the protein and the ligand. Then, we applied successive 1000 step energy minimizations while decreasing the weight of the positional restraints, from 1000 to 1 kJ mol -1 nm -2 . The system was heated from 50 to 300 K during 100 ps of MD at 1 atm without restraints. Finally, a production MD run for 100 ns was performed at 300 K and 1 atm.

pKa calculations
We utilized the PYP crystal structure (PDB identifier 1OTB) and our models of the mutant PYP constructs. The pKa calculations were carried out using PROPKA 3.1 with default settings. 9

Explicit water simulations
The agreement between the density of the implicit solvent binding simulations and experimentally observed binding rates was remarkable for several reasons. First, the actual labeling experiment is irreversible, but our simulation was based on cumulative density of ligand positions. Second, we included the leaving group, which presumably causes some steric hindrance to the FCANB-PYP interaction, preventing perfect binding.
Third, we treated the water implicitly, which would be expected to result in artifacts in short-range electrostatic and hydrophobic interactions. Addressing the first issue (covalent bonding) is beyond the scope of our MD study, but the remaining two issues can be addressed by carrying out explicit water MD simulations of the FCANB without the leaving group (tFCANB) and comparing these to the implicit solvent results. Here, we initialized the ligand in the bulk at minimum distance of 6-15 Å from the protein surface close to the binding residue Cys-69 ( Figure S7) and again ran all 6 PYP constructs (4 point-mutants, 1 quadruple mutant, and WT) for a total of 6x30 25 ns simulations. When we computed the population of snapshots within a 6 Å threshold as above, we obtained a distribution very different from the longer implicit solvent result, which did not correlate well with the experimentally observed labeling rates. The lack of agreement is unsurprising because each MD run was too short to reach equilibrium even for such a small protein-ligand system. In addition, the biased system setup could introduce errors in the observed statistics. However, the mode of interaction between PYP and FCANB would be expected to be more realistic in the explicit water simulation than in implicit solvent. In order to assess if implicit solvent simulations contained artifacts that affected the resulting structures, we carried out the following exercise. First, we assumed that the densities derived from the implicit volume simulations were accurate. Second, we replaced snapshots within the 6 Å threshold from the implicit solvent runs by snapshots within the same threshold from the explicit solvent MD runs. The total number of explicit water snapshots was less, but the proportions for each mutant and WT were maintained. The resulting distribution, by definition, appears as a scaled-down version of Figure 4. We next assessed the binding propensity in more detail as follows. First, we computed the RMSD of the three reference atoms from each bound ligand in the 6-molecule ensemble. For each RMSD value, we computed a binding probability as with W = 6 Å. We then computed the average of this value over the 6 members of the bound ensemble. Finally, we summed this average probability over each explicit water snapshots. In spite of the use of explicit-water snapshots and the more rigorous treatment of binding propensity, the resulting distribution looked very similar to that obtained from the implicit solvent simulations ( Figure S8). Thus we concluded that the implicit water simulations did not contain severe artifacts that affected the structure of the FCANB-PYP complexes.

Distribution of probe-tag contacts
To further clarify the effects of the mutations PYP, we investigated contacts between PYP and different parts of FCANB. Since we expected that the explicit solvent simulations were more realistic in terms of the mode of binding, we investigated PYP-tFCANB interactions within the 6Å threshold. Here, we divided tFCANB into three parts; fluorophore, PYP ligand, and linker region of above two ( Figure S9), and counted the number of contacts of each atom group. We ignored the linker region from this analysis because it shouldn't be so important in electrostatics driven interactions.
The profiles and electrostatic field on the solvent accessible surface (SAS) of PYP constructs are shown in Figure S9B-C and S10. In principle, the distributions of other basic residues around Arg-53 (e.g. Arg-52). In contrast, the magnitude of these regions in D97R were more modest. Another example that demonstrates the importance of the electrostatic potential is D71R, where both fluorophore and PYP-ligand groups made contact with the mutated residue many times. In this case, the positive region of electrostatic potential was distributed on the outside of the binding pocket, and this likely made it difficult for FCANB to enter the binding pocket. In other words, the PYP-ligand group could not make strong contact with Cys-69 due to its attraction to Arg-71. E74R had few direct contacts between Arg-74 and FCANB, thus, we speculate that the main contribution to the acceleration above that of the WT was due to the difference in the net charge. The profile of the quadruple mutant (4R) indicated that the effect of mutations on the binding kinetics is not additive but cooperative. For example, Arg-74, which didn't interact well with FCANB as a point mutant, had many contacts with FCANB in the quadruple mutant, possibly due to the influence of the other two mutated residues (Arg-97 and Arg-71), which are close to Arg-74. Given the proximity of these three residues, it's natural that the electrostatic potential around them was changed drastically from that of the WT. Overall, for larger ligand binding, mutated residues should be selected so that the resulting electrostatic field is able to attract the ligand with a conformation similar to that of the bound state.

Implicit solvent MD simulation of RGT
To assess the robustness of the implicit solvent MD simulation, we applied the same protocol used for FCANB to the RGT ligand with a positively charged fluorophore (Fig.   S11). Since WT PYP has a net (-6) negative charge, a positively charged ligand has a much greater preference to bind. In the case of the 4R mutant, RGT binding rate was reduced, as expected. However, the difference between WT and 4R binding rates was not as large as in the case of FCANB, suggesting that RGT is not as sensitive to the overall charge as FCANB. This is consistent with the net ligand charge (+1 for RGT and -2 for FCANB). In addition, the distribution of residues on PYP is more suitable for binding RGT than FCANB.