Crystal structure of Venus, a yellow fluorescent protein with improved maturation and reduced environmental sensitivity

Yellow emission variants of green fluorescent protein (GFP) have been found useful in a variety of applications in biological systems due to their red-shifted emission spectrum and sensitivity to environmental parameters, such as pH and ionic strength. However, slow maturation properties and new requirements for more intense fluorescence necessitated further mutagenesis studies of these proteins. Venus, a new variant with improved maturation and brightness, as well as reduced environmental dependence, was recently developed by introducing five mutations into the well-characterized variant, EYFP. In this paper, we present the crystal structure of Venus at 2.2 Å resolution, which enabled us to correlate its novel features with these mutation points. The rearrangement of several side chains near the chromophore, initiated by the F46L mutation, was found to improve maturation at 37 o C by removing steric and energetic constraints which may hinder folding of the polypeptide chain, and by accelerating the oxidation of the C α -C β bond of Tyr66 during chromophore formation. M153T, V163A and S175G were also found to improve the rate of maturation by creating regions of greater flexibility. F64L induced large conformational changes in the molecule, leading to the removal of halide sensitivity by preventing ion access to the binding site.


Introduction
Green fluorescent protein (GFP), originally isolated from the jellyfish Aequorea victoria, has been the subject of continued interest since its gene was first cloned in 1992 (1). The high stability of mature GFP over various environment conditions, the spontaneous autocatalytical generation of the GFP's fluorophore, and the possibility of spectral manipulation by mutagenesis, make GFP and its related proteins attractive tools for numerous biological applications (2).
Maturation of GFP proceeds in three major steps, beginning as the 238-residue single GFP polypeptide folds into its nearly native conformation. Residues 65-67 of the folded protein then undergo several chemical reactions necessary for chromophore formation, including cyclization and dehydration (3). The final rate-limiting step in the maturation process involves the oxidation of the Cα-Cβ bond of Y66 by aerial oxygen. This maturation process takes approximately 3 hours at room temperature and its efficiency further decreases at 37 o C, hindering the use of GFP in some biological applications. To improve maturation and performance at 37 o C, mutations such as F99S, M153T, and V163A have been introduced, creating valuable GFP variants for a wider range of applications (4), (5).
Thorough understanding of the relationship between the protein sequence and its physicochemical properties requires 3D structure determination of GFP at atomic-resolution. Crystal structures of both the monomeric and dimeric forms of GFP have been previously solved (6), (7). The wild-type GFP (wtGFP) chromophore, which exists in equilibrium between two ground states, produces two maxima in the light absorption spectrum. The near-UV absorption peak at 395 nm originates from the neutral chromophore, while the ionic chromophore absorbs at a longer wavelength and produces a smaller peak at 475 nm (3). Green light is emitted from wtGFP at 503 nm. Spectral properties of GFP were shown to be influenced by its dimerization state (8) which is characterized by a dissociation constant of ~100 µM (9).
In the case of YFP, a yellow variant of GFP, the main absorption maximum of 514 nm corresponds to the chromophore ionic form and is red-shifted relative to wtGFP. Unlike wtGFP, the neutral form does not emit fluorescence efficiently and the higher-energy peak is largely suppressed (10). The equilibrium between the neutral and ionic forms of the YFP chromophore can be shifted in by guest on July 8, 2020 http://www.jbc.org/ Downloaded from favour of the latter form by the following two mutations: S65G, which prevents hydrogen bond formation between residues G65 and E222, and T203Y, which produces additional polarizability around the chromophore. Also, the π-π interaction between Y203 and the chromophore phenol ring reduces the excited state energy, thereby increasing the excitation and emission wavelengths (6).
When compared to the green-emission variants of GFP, the absorbance level of YFP and its derivatives display a strong environmental sensitivity. The ratio between the two absorbance peaks, and therefore fluorescence efficiency, are dependent upon pH and the halide ion concentration (10), (11). This sensitivity makes YFPs efficient pH and halide sensors (12), (13), but for applications requiring stable fluorescence in a wide range of environment conditions, it is necessary to have proteins that are insensitive to pH and halides. Griesbeck and colleagues (14) developed a less sensitive variant of yellow fluorescent protein, citrine, which contained a Q69M mutation in addition to the four mutations used to create the original yellow variant EYFP(S65G/V68L/S72A/T203Y). This mutation helped decrease the pH and halide sensitivity and facilitated expression at 37 o C. With a relatively low value of pKa = 5.7, citrine is particularly applicable in systems of lower pH and in a wide range of Clconcentration.
More recently, Nagai et al. introduced a new variant of yellow-emission protein, called Venus, with reduced pH and halide dependence and an improved maturation rate (15). These properties were achieved in Venus by various mutations which differ from those used in citrine (14). Instead of Q69M mutation used in citrine, Venus contains five other mutations (F46L, F64L, M153T, V163A and S175G) in addition to the previously described four mutations characteristic to EYFP. The F46L mutation helps accelerate the oxidation step in chromophore formation, an effect which is specific to yellow-emission proteins. The remaining four mutations are known to improve folding at 37 o C and at least one w a s found to remove proton and chloride sensitivity (15) ( Table 1). The aim of our work was to solve the 3dimensional structure of Venus, shedding light on the structural basis of the observed molecular properties. We compared this crystal structure of Venus at 2.2 Å resolution with previously published structures of wtGFP (7) and EYFP (10) to pinpoint mutation-dependent structural changes. Our ultimate goal is to use systematic structural analysis of GFP mutants to provide future guidelines for by guest on July 8, 2020 http://www.jbc.org/

Protein expression and purification:
Venus was subcloned into a pRSET b plasmid and expressed in the E. coli strain BL21(DE3) gold . Cells were grown using standard culture medium (LB-broth) in a shaker incubator at 37 o C until the optical density of 0.7 was reached. Upon induction with IPTG, the temperature was lowered to 25 o C and protein was expressed for 7-8 hours. Protein containing the N-terminal poly-histidine tag was purified using nickel-affinity chromatography. Subsequently, ion exchange and gel filtration were employed to achieve satisfactory levels of purity. The N-terminal six-histidine tag was cleaved using enterokinase.

Structure determination:
The protein solution was concentrated to 33 mg/ml in 20 mM Tris hydrochloride (pH 7.9), with 50 mM NaCl and 0.5 mM DTT. Yellow fluorescent rod-shaped crystals (P3 1 12), with approximate dimensions of 0.5 x 0.08 x 0.08 mm, were grown in hanging drops containing 2 µl protein and 2 µl of mother liquor at room temperature for 3-7 days. The mother liquor contained 2.1 M ammonium sulfate, 100 mM Tris hydrochloride at a pH of 8.4 and 4% (v/v) PEG 400. The crystals were cryo-protected in 20% glycerol and 5% (v/v) PEG400 then flash-cooled to 100 K in a stream of N 2 gas. Diffraction data from a single crystal were collected in-house on a Mar345 imaging plate detector using Rigaku RU200 rotating anode as a generator of Cu-Kα X-rays. 0.5 degree oscillations were collected in 300 s exposures. Data were collected to 97.7% completeness at 2.2 Å resolution. Unit cell dimensions were a= 82.704, b= 82.704, c= 72.555 Å. The EYFP coordinate file (10), which served as a search model for phasing by molecular replacement, was edited by removing mutated residues and the chromophore in order to minimize bias. The parameters for the anionic chromophore were obtained from the Hetero-compound Information Centre, Uppsala (HIC-up). Rigid-body refinement was carried out using the CNS program (20). Electron-density maps (2Fo-Fc and Fo-Fc) were inspected using the program O (21) and water molecules were added using CNS. Nine rounds of refinement interspersed with manual building in O, including introduction of water and PEG molecules, resulted in R cryst =22% and R free =25%. The data collection and refinement statistics are summarized in Table 3. The following PDB codes were used for by guest on July 8, 2020 http://www.jbc.org/

Chromophore environment and spectral properties
Venus was crystallized as a dimer with one molecule per asymmetric unit. The refined structure of Venus shows an 11-strand (labelled A to K from the N-terminus) β-barrel, typical of GFP-derived fluorescent proteins. Calculated for α-carbons of residues 5-60 and 70-225, the r.m.s.d. of a single Venus β-barrel relative to wtGFP and EYFP β-barrels is 0.32 Å and 0.35 Å, respectively. In all three structures compared, the cross-section taken perpendicularly to the long axis of the barrel is not perfectly circular but rather oval in shape. While the longer diameter of the oval cross section of all three proteins is identical (23.8 Å; measured from distances of four Cα pairs), the shorter diameter differs between EYFP (22.3 Å) and Venus (21.9 Å). The difference in the oval shape between these yellow variants is closely related to the packing of interior amino acid residues and influences the integrity of dimer interface, as discussed below.
The residues surrounding the Venus chromophore are similar to those surrounding the chromophore of EYFP ( Figure 1A). More specifically, there are no significant changes between Venus and EYFP in the distances between the chromophore and residues T62, L68, Q69, Q94, R96, N146, R168, Q183, Y203 and S205. Electron density studies of Venus show an 11 o angle between the planes of the chromophore and Y203, while the same angle measured in EYFP is 11.5 o -12.3 o , making the plane of the Venus chromophore slightly more parallel than that of EYFP. This small difference may account for a minor change in the absorption spectrum of Venus relative to EYFP, as seen at a pH range of 4.6-8.6 and a 50 mM NaCl concentration. Although the position of the main absorption maximum remains unchanged (516 nm) in both Venus and EYFP, a smaller peak resulting from the neutral (protonated) chromophore is shifted in Venus by 20 nm producing a peak at 413 nm versus 393 nm in EYFP. The slightly more parallel orientation of the aromatic rings of Y66 and Y203 might also help improve the π-π interaction, thus reducing the excited state energy of the neutral chromophore. If this is the case, we can hypothesize that at lower pH, which is favourable to chromophore protonation, the angle between the rings would be smaller than 11°.
One of the changes observed around the Venus chromophore is a 0.3 Å increase in the by guest on July 8, 2020 http://www.jbc.org/ Downloaded from distance between the Cα-Cβ bond and the OE1 atom of the E222 side chain, compared to EYFP. A result of this increased distance is a 0.5 Å shift of the chromophore away from the protein surface and the E222 side chain. Following this shift, an increased rate of oxidation is observed, attributed to enhanced oxygen access to the oxidation site in Venus. Other factors facilitating this reaction will be discussed in the section Venus-specific mutations.
Another difference in the Venus chromophore area, when compared to that of EYFP, is the presence of a shorter hydrogen bond (by 0.2 Å) between the chromophore phenolate and H148. This may result from a slight rotation in the H148 imidazole making it less parallel to the β-barrel surface.
However, this effect may also be due to the higher pH at which the crystals were grown (pH 8.5 for Venus, pH 6.9 for EYFP). The previously observed pH effect on S65T suggests that at a lower pH this distance becomes larger and hydrogen bonds are broken, consistent with our observations of Venus as well as past studies of EYFP (16).
The mutual position of the two chromophores within the Venus dimer differs from that of wtGFP and EYFP. The chromophore-chromophore distance in the Venus dimer is shorter than that of others by aproximately 2 Å when measured using the phenol oxygens ( Figure 3A). These differences may be attributed mainly to the different orientation of subunits in the dimeric structure, coupled with the overall closer inter-subunit contact in Venus than in EYFP ( Figure 2 and Table 2).

Dimer structure and halide sensitivity
The subunit orientations in the Venus dimer ( Figures 1B and 1C) are similar to those in the wtGFP dimer, but differ from those of EYFP. When the backbone chain of one Venus subunit is superimposed on that of wtGFP, the second subunits display a similar orientation with respect to the superimposed subunits. In contrast, following superimposition of the first subunit of Venus on that of EYFP, the second subunits display a significantly different orientation (Figure 2A). The r.m.s.d. for the aforementioned second subunits is 1.15 Å between Venus and wtGFP and 4.97 Å between Venus and EYFP. By using the angular parameters described in Table 2, we can quantitatively define the unique orientation of the two subunits within the dimeric fluorescent proteins. The 'tilt' angle between two subunits of Venus is similar to that of wtGFP, but differs from that of EYFP (Table 2). These differences and similarities in domain orientations, depicted in Figure 2A, clearly show the differences in the position of one subunit when the other subunit is superimposed.
Comparison of the buried surface areas for wtGFP, EYFP and Venus, indicates that the dimerization interface differs significantly between the three proteins ( Table 2). While the residues involved in the dimer interactions are essentially identical between Venus and wtGFP, significant differences are observed between Venus and EYFP. In both Venus and wtGFP, E142, Y145, N146, N149, R168, N170, E172, Y200 and S202 lie within the buried surface between the two subunits, whereas the same EYFP residues are solvent-accessible ( Figure 2B). These differences leave Venus with 5% more buried surface area relative to EYFP. In addition, the average inter-subunit distance between atoms located within the buried surface of the Venus dimer is somewhat shorter than that of EYFP (Table 2). These concurring results indicate that Venus has a larger dimerization interface than EYFP.
Wachter and coworkers (10), (17) have shown that the fluorescence intensity of YFP is inversely related to the concentration of halide ions in solution. This group identified a halide binding site proximal to the EYFP-H148Q chromophore which suppresses chromophore fluorescence upon ion binding. The halide binding cavity is adjacent to the backbone carbonyl groups of Y66 and G67 ( Figure   4B). Two mutations (T203Y and V68L) located at the buried halide ion binding cavity of EYFP ( Figure   4B) have been implicated in halide sensitivity (17). YFPs' halide ion sensitivity have lead to their use as halide ion sensors (13). Despite the presence of these mutations in Venus, no halide sensitivity has been observed.
The introduction of H148Q into EYFP enhances the halide affinity due to a better access of the ion to its binding cavity via a solvent channel, formed at a β-bulge located close to the dimer interface (10), (17). Venus has a cavity essentially identical to that of EYFP-H148Q in both size and side chain configuration. However, analogous mutagenesis performed on Venus, led to no changes in halide affinity (Nagai and Miyawaki, unpublished). Supported by structural information available on Venus, this data indicates that the β-bulge is inaccessible to halide ions due to rearrangement of the inter-subunit by guest on July 8, 2020 http://www.jbc.org/ Downloaded from interface. This rearrangement centered at the β-bulge is most likely initiated by F64L, a mutation that induces a relocation of many β-strand residues at the dimer interface.
Consistent with the aforementioned observation on the Venus β-bulge, the wtGFP β-bulge is also inaccessible to halide ions (7). The β-bulge residues Y145, N149 and R168 participate in dimer contacts in both Venus and wtGFP, while the same residues are fully solvent exposed in EYFP ( Figure   2B). These results suggest that the inter-subunit interface could influence the chromophore environment, linking the dimerization mechanism to the fluorescence property of GFP and its variants (8).

Venus-specific mutations
As previously stated, Venus contains five additional mutation sites relative to EYFP ( Table 1).
Comparison of the corresponding crystal structures helped explain the effects of these mutations on Venus properties. M153T, V163A and S175G were found to improve the rate of maturation, while mutations F46L and F64L induced a number of local and global structural changes. Here we describe the individual effect of each mutation in detail.
F46L. In the EYFP structure, the F46 phenyl ring exhibits an antiparallel π-π stacking to the phenyl ring of F64 ( Figure 4A). Substitution of each aromatic side chain with an alkyl side chain creates a vacancy at residue 46. This change in side chain orientation results in close hydrophobic contact between residue L46 and the neighbouring residue L44 as well as an associated change in the L42 side chain. The F46L mutation was found to improve maturation at 37 o C (15) by removing steric and energetic constraints, which may hinder folding of the polypeptide chain (Table 1). Furthermore, the Venus F46L substitution accelerates the oxidation of the Cα-Cβ bond of Y66 during chromophore formation (15). Interestingly, major functional changes caused by the F46L mutation resemble those seen in GFP-S65T (18), (6) and DsRed N42H/Q (19), which show an increased rate of oxidation relative to their 'precursor' proteins, wtGFP and wild-type DsRed, respectively. In GFP-S65T, the introduction of the threonine side chain at chromophore position 65, repositions the L44 and L42 side chain. In DsRed-N42H/Q, the more bulky side chain of H42 or Q42 shifts towards a position similar to L42 in Venus, leading to a repositioning of Q66, an equivalent to residue 65 in Venus and EYFP. In brief, the introduction of F46L causes a series of changes at residues 42, 44 and 65, facilitating an increased rate of maturation. In addition, a change in residue 65 (66 in DsRed) in the Venus chromophore may lead to an increased rate of oxidation.
F64L. In EYFP, the V68L substitution reorients both the central helix and the chromophore towards the protein surface (8). In Venus, however, the substitution of phenylalanine with leucine at position 64 counteracts any change in orientation caused by V68L, and the central helix and the chromophore remain roughly in their wild-type positions. Therefore, when compared to EYFP, the Venus chromophore is shifted approximately 0.5 Å toward the centre of the molecule ( Figure 3B and 4A). A similar dislocation of neighbouring residues Y145, Y203, L220, and E222 accompanies this slight movement of the central helix. Such structural changes also influence the position of β-barrel strands G, J, and K as well as part of strand H. The shifted β-strands contain residues involved in the dimer interface, such as N144, S147, and A206 ( Figure 4A). Relocation of these residues causes a small decrease in the average diameter of the β-barrel and geometric "flattening" of the β-barrel itself, increasing the surface area buried between the two monomers relative to EYFP. The β-strand movement occurs in close parallel to the movement of the chromophore and the central helix, and hence the distances between the residues surrounding the chromophore remain essentially unchanged in Venus relative to EYFP. The implications of the F64L mutation may therefore involve a reduction in halide ion binding to Venus (as discussed above in section Dimeric structure and halide sensitivity).
M153T and V163A. Both T153 and A163, commonly mutated to improve protein maturation, are located in the loop region between strands G and H. This region is located in close proximity to the halidebinding site but away from the dimer interface in the EYFP ( Figure 4B). There is no obvious difference in the orientation of side chains of residues 153 and 163 between Venus and EYFP, where the polar side chain of T153 remains exposed to the solvent, and the side chain of A163 continues to point towards the molecular interior. Based on its location near to the halide-binding site, the V163A mutation may be involved in eliminating chloride sensitivity in Venus ( Figure 4B) (17). S175G. The S175G mutation is located in a short turn between strands H and I. The removal of the hydroxyl group of serine drastically alters the backbone conformation of the region spanning residues 173-175. Such a change involves breaking the three-centered hydrogen bond found in EYFP between the side-chain carboxyl oxygen atom of D173 and the amide nitrogens of G174 and S175. This is caused by ~180° rotation of the carboxyl side chain of D173, which flips away from the backbone carbonyl of G174, resulting in its complete exposure to the solvent ( Figure 4C). The S175G mutation in cycle-3 GFP (5) and Venus (15) has been shown to facilitate the protein folding process in vivo. S175G also enhances the fluorescence intensity of ECFP two-fold in an in vitro experiment (Nagai and Miyawaki, unpublished). While the exact structural basis of this increased folding rate is unclear, it should be noted that a localized conformational change outside of the β-barrel ( Figure 4C) significantly impacts the properties of the chromophore within the β-barrel, including maturation.

Concluding Remarks
The five mutations that distinguish Venus from EYFP increase the rate of maturation and abolish the sensitivity to halide ions (15). The 2.2 Å X-ray structure of Venus reveals several structural differences when compared to EYFP and wtGFP, thus helping to elucidate the structure-function relationship of this fluorescent protein. Among the five mutations, those known as 'folding mutations' (4), (15) seem to facilitate the maturation process by introducing highly flexible, smaller side chains within the loop regions located outside the β-barrel (M153T, V163A, S175G) or near the chromophore (F64L). The F46L mutation promotes a local structural change near the chromophore, which consequently accelerates oxidation, a rate-limiting step in chromophore maturation. In addition, w e found that the F64L mutation is responsible for a major structural change inside the β-barrel, leading to an alteration of the inter-subunit orientation, ultimately preventing halide ion access to the binding cavity near the chromophore.        (20) with probe radius 1.4 Å. 2 Distance between monomers was obtained as a mean of distances between chemical groups belonging to different monomers in the buried surface (obtained from MolMol (23)) with a limit of 5 Å (± standard deviation).