3.4.1 Design strategy of MCD1 (RapidER)
MCD1 was designed based on several considerations including features of calcium bind- ing geometry and charged residue preference obtained from statistical analysis of different clas- ses of calcium binding proteins (66, 157, 158) and fast kinetics and calcium induced florescence change. We hypothesize that negatively charged ligand residues in a hemisphere coordination geometry with good solvent accessibility, electrostatic binding energy change, and the hydrogen bonding network of the chromophore are the key factors to control calcium binding affinity, ki- netics, and calcium binding dependent change of optical properties. First, red florescent protein mCherry was chosen because of its noted photo-stability, fast chromophore maturation rate and low pKa(109) that is important for reduce pH sensitivity when measuring calcium binding. Sec-
ond, we used the “half shell” with reduced coordination number to create the calcium binding site on the surface of the beta barrel to allow easy entry and release without the spatial barrier from the calcium binding site. With the half shell calcium binding site composed primarily of Asp and Glu residues as the most predominant calcium binding ligands, the calcium binding af- finity was expected to be lower than that of the classic EF-hand motifs. Differing from the clas- sic calcium binding pocket that is consisted with 6-7 oxygen atoms, and forms a bipyrimidal shape with high calcium binding affinity and selectivity, it has less coordinator number and forms an open bowl-like geometry. Third, to couple calcium binding with optical property change without interfering with chromophore formation, we finally decided to mount the calci- um binding site in pocket 1 according to located in the chromophore hydrogen-bonding network at the corresponding position in CatchER (Figure 3.1A and 3.32). Residue 145 is involved in flanking E144 and S146. The main chain oxygen of E144 forms a hydrogen bond with the chro-
mophore tyrosyl through a bridging water molecule, and the side chain hydroxyl group of S146 has two conformers, serving as the hydrogen bond donor for the chromophore tyrosyl directly. Fourth, based on the kinetic study of the electrostatically driven interaction between protein and ligands (169-171), we further reason that mCherry with a larger negatively charged surface around the designed calcium binding site may exhibit a faster calcium association rate than CatchER. Compared to CatchER, the calculated electrostatic binding energy change of MCD1 was greater than that of CatchER (Table 3.5). The calculated negatively charged solvent accessi- ble surface area (SASA) using PyMol in the designed calcium binding site was also larger in MCD1 than in CatchER.
Table 3.5 The electrostatic binding energy calculation
Proteins ΔGelec_binding (kT) Negatively charged SAA (Å 2
)
MCD1 -73.03 655
CatchER -59.48 589
3.4.2 Key factors for binding affinity and kinetic properties
Calcium binding accepts a flexible range of the coordination number ranging from 3-8. Because it is a soft metal, oxygen is preferred to be the coordinator. From the statistical analysis of the calcium binding sites in protein data bank, aspartic acid, glutamic acid, and water are the top three residues providing oxygen to bind calcium ion (157). The bipyrimidal shape formed by the oxygen atoms coordinating calcium is the classic one found in EF-hand motif. For the organ- ic calcium chelators or dyes, EGTA and Fura-2 have four carboxylate groups, where the calcium is coordinated by 8 atoms including two nitrogen atoms. In either EF-hand motif or tetracarbox- ylate compounds, calcium is surrounded by oxygen/nitrogen atoms and there is also a steric bar- rier formed by the “arms” of the coordinators. Such steric hindrance may limit the calcium disso- ciation rate and contribute to the high affinity. Renner and the co-workers also reported that in the EF-hand loop the side chain length and acidity affected the Ca2+ dissociation rate due to the steric block and electrostatic barrier, respectively (172). The calcium binding geometry made of fewer coordinating ligands is observed in proteins shown in Figure 3.1 such as 1FZC and 1BJR. For the single calcium binding site in 1FZC, the number of oxygen ligand from the protein/water is 5/1, and 4/4 and 5/0 for two calcium binding sites in 1BJR. In both proteins, the ligands from protein moiety form a bowl-like geometry and the calcium ion is partially exposed to the solvent. For organic compounds, the calcium binding moiety was modified with less coordinators in or- der to increase the calcium dissociation rate such as Mag-Fura-2. In Mag-Fura-2, the calcium binding pocket contains three carboxylate groups with 3 formal charges, and there is a channel available for calcium to escape. Thus, the calcium binding affinity of Mag-Fura-2 and other tri- carboxylate compounds is lower than those analogs with four carboxylate groups attributed by the higher calcium dissociation rate. There is another strategy to increase the koff of calcium
dyes, which is to add the electron attractive group like -NO2 or -F in the 5 position of o-
aminophenoxy ring. Such calcium dyes ended up with 5N or 5F in their name and have calcium binding affinity around 1-10 μM which is lower than the original tetracarboxylate calcium dyes but higher than those with tricarboxylate. The calcium binding association rate of these calcium dyes is in the magnitude of 108 M-1s-1. Therefore, the variance in Kd mainly results from the dis-
sociation rate koff as shown in Table 1.1. Clearly, the coordination number and the number of
formal charge have a stronger effect on both koff and Kd by determining the geometry and steric
hindrance.
The calcium binding features of protein-based calcium indicators are mainly determined by the calcium binding moiety (calmodulin or troponin C) involved. The calcium binding affinity of both CaM and TnC is around 10-6-10-7 M. The stoichiometry for both calmodulin and troponin C is 1:4, two calcium ions in each terminal domain. The calcium binding follows a cooperative manner in each domain, which may contribute to the slow calcium dissociation rate around 0.2- 20 s-1. In addition, the two domains are relatively independent, resulting in the biphasic calcium binding curve observed in Cameleon (38).
Efforts have been taken to reduce the Kd of these GECIs. In Cameleon 3, the mutation E104Q locating in the third EF-hand motif in CaM domain eliminated the high calcium binding component of the CaM C-domain and the Kd value is 4.4 µM. For Cameleon 4, the mutation
E31Q located in the first EF-hand motif further decreases the binding affinity in CaM N-domain but does not significantly affect the high affinity C-domain. The resulting Kd of Cameleon 4 is
83 nM and 700 µM with a Hill coefficient 1.5 and 0.87 respectively (54). Palmer and co-workers switched the charged residue in the interface between M13 and CaM which decreased the appar- ent Kd (0.8 and 60 µM) and increased the koff (256 s-1) although the original purpose was to elim-
inate the perturbation of the normal calcium signaling by interaction with their intrinsic target proteins (38). A similar situation was observed in the recent version 6 of GCaMP, which was generated by altering the interface between M13 and CaM as well as the one between CaM and cpEGFP. GCaMP6f exhibited faster calcium response to action potentials in neuronal activity than other GCaMPs (91). This evidence shows that the apparent Kd of the calcium indicators in-
volving CaM and M13 peptide is not only determined by the calcium binding motifs but also the linker and domain interface.
In a different scenario of TnC involved GECIs, TnC also has two terminal domains like CaM and N-terminal domain binds Ca2+ with lower affinity than C-domain, which is a regulatory site. The TnC molecule undergoes structural rearrangement after Ca2+ binding (173). TN-XL with faster kinetics than TN-L15 was created by switching the residue N and D at position 109, 111, 145 and 147 at the third and forth EF-hand motifs in C-terminal domain of TnC, where the calcium binding affinity was lowered (Kd=2.2 µM, Hill coefficient 1.7) and magnesium interfer-
ence was abolished (174). The resulting off rate of TN-XL measured by stopped-flow spectrome- ter was around 5 s-1, which was approximately 5-fold greater than TN-L15 and TN-XXL.
The cooperative binding, the slow dissociation kinetics and the high calcium binding af- finity were signatures for native Ca2+ binding protein involved GECIs. In addition, the Kd was
not controlled by the Ca2+ binding alone due to the multiple steps required for fluorescence sig- nal change. Therefore, it is difficult to tune the binding affinity and kinetics as well as avoid co- operativity in the same construct by rational design. In contrast, CatchER was successful in fast kinetics, low calcium binding affinity and 1:1 stoichiometry by avoiding the native calcium bind- ing moiety. In such a simplified calcium indicator as CatchER, the calcium sensing relies on the local dynamics in the calcium binding site. As a starting point, the simple calcium binding site in
CatchER was modified to increase the kinetics further by reducing the positively charged resi- dues around the binding site.
Compared with CatchER, RapidER had a higher calcium binding affinity as well as faster kinetics, which is in agreement with the calculation of the electrostatic binding energy change and the negatively charged solvent accessible surface area. The faster calcium dissociation rates found in both CatchER and RapidER than in the EF-hand motif sensors is attributed to the ge- ometry of the designed half-shell calcium binding site, where there is little steric barrier for cal- cium release. A similar comparison was also observed in tetracarboxylate and tricarboxylate cal- cium dyes fura-2 and mag-fura-2. As we discussed above, the electrostatic attractive functional group such as NO2 and halide was added in the compound to reduce binding affinity and increase
koff without changing the coordination number and geometry. But it cannot be applied in proteins
because the conjugated system was not able to form throughout the side chains. However, there were three positively charged residues K74, K166 and R220 around the calcium binding site in MCD1 compared to CatchER where there was only one K42 nearby (Figure 3.31). The edge of the enlarged negative circle may also be neutralized by these positive residues and thus increase the dissociation rate. The evidence for the role of the positive residues around the binding site can be found in the crystal structure of Ca2+-CatchER, where there were two populations of cal- cium ion positions observed and both of them are coordinated by E147 but away from E223 and E225 which were close to K42. We hypothesized that the positively charged residues around the calcium binding site attracted electron density to reduce the electrostatic barrier.
3.4.3 Positions of calcium binding sites
RFP has a larger conjugated system than that in GFP, which was extended to the back- bone of F65 before the cyclized tri-peptide (175, 176). The GFP was modified with several mu-
tations to improve the oligomer character, brightness, photostability and maturation time (177). Like GFP, the chromophore environment of RFP plays important roles in maintaining the fluo- rescence. The side chain orientation of the neighboring residues in the β sheets is usually oppo- site. Those projecting to the interior of the β-can form hydrophobic or electrostatic interactions to participate in or protect the chromophore environment, while the others facing the solvent assist to keep the protein from aggregation or degradation. The residues with side chains in the interior of the protein have more direct contact with the chromophore so as to affect the optical property directly. For example, mutation of the E215 in mCherry resulted in the blue shift of the spectrum, where the original E215 was protonated and formed a hydrogen bond between the protonated carboxyl group and the imidazolinone ring nitrogen (159, 178). mBanana has the mutation I197E based on mTangerine, which may contribute to the increase of pKa from 5.7 to 6.7 due to the redistribution of the chromophore electron density (110). However, calcium binding in the sur- face of the β-barrel mainly involves the side chains protruding outside. In order to change the spectral properties by calcium binding, the calcium coordinators need to influence those residues in the chromophore environment.
As shown in Figure 3.1, in the chromophore of mCherry, the chromophore phenol hy- droxyl group was close to the opening of the β-barrel, where a loop region is located. The corre- sponding location in GFP served as a tunnel to allow proton migration during excited state pro- ton transfer. This phenolate oxygen formed the H-bonds to the side chain of S146 (both states), to the main chain of E144 through a bridging water molecule (wat1), and to the side chain of Q163. As mentioned above, E215 in mCherry was proposed to be protonated to form H-bond with the imidazolinone nitrogen. The position of this residue was relatively rigid because a net- work was observed among the chromophore, E215, Q42, S69 and a water molecule (wat4 in
Figure 3.1). Both main chains and side chains of these three residues together with the chromo- phore had contact with each other, leading to a tight association. H-bonds were found from the side chains of R95 and Q64 to the imidazolinone oxygen. Since the peptide bond connecting F64 and the chromophore was part of the conjugated system in mCherry, the main chain oxygen in F64 was also under consideration, which formed H-bonds with the side chain Q109 and S111 via water molecules.
We selected three potential calcium binding sites to affect the phenol group. RapidER in pocket 1 was one of them, including the mutation A145E in between E144 and S146. The pocket 2 included E144 and the pocket 3 had the mutation E164 adjacent to Q163. Both pocket 1 and 2 had mutations R216E next to E215. Pocket 4 was located near R95 and Q109, including muta- tions K92E and T108E. The results suggested that only the variants in pocket 1 and 2 showed the calcium dependent fluorescence change. The equilibrium-dialysis assay confirmed that mcP6 (K92E/E94/T108E/D110), belonging to the pocket 6 variants, also bind calcium with the compa- rable binding affinity as ones in pocket 1 and 2. The lifetime of Tb3+-MCD1 FRET verified the calcium went to the expected position in MCD1. Calcium binding increase the lifetime of MCD1 and thus the quantum yield was enhanced. Therefore, the phenol hydroxyl group was more sen- sitive than others to the change of the electrostatic environment. It is the hot spot that has the major potential to affect the optical property.
The calcium dependent fluorescence change was decreased when it was expressed in the mammalian cells. Further modification was done to increase the dynamic range. For MCD15 (RapidER’), the mutation R220E on MCD1 outside the designed calcium binding pocket made the calcium induced fluorescence change detectable in situ, without altering the in vitro optical and biophysical properties. However, the dynamic range of MCD15 (10%) is lower than other
low-affinity calcium sensors (~30%). The limitation is possible to be overcome by either further suppressing the quantum yield of the calcium-free form or enhancing the brightness of the calci- um-loaded form. It would be promising to have such a red fluorescent protein with the property of calcium triggered-fluorescence enhancement.