CHAPTER 2. DYE-BASED BIOSENSORS TO QUANTIFY
2.3.11 Correlation of biosensor properties with
in protein stability caused by the sum of multiple mutations (Fig 2.26). The crystal
structures of active ERK2 (PDB: 2ERK) and inactive ERK2 (PDB: 1ERK) were used as
the input protein structures. Backbone structure pre-relaxation prior to simulations and
the backbone flexibility was modeled in most cases. The Medusa force field was used to
model the structures and the free energy of the input protein computed by Eris was the
153
hydrogen-bonding interactions, and backbone-dependent statistical energies . The last
term is of great importance because it detects the strain on the backbone induced by
multiple mutations and these effects are taken into consideration by Eris. Positive changes in the free energy ∆∆G (∆∆G = ∆Gmutant
- ∆Gwild-type) of the mutant of interest indicates that destabilization effects are introduced by the mutations. Strong stabilization by mutations might also cause malfunction of the mutant proteins, thus the ideal ∆∆G values should be close to zero. It is clear that the labeling control, the ERK2 mutant with
no exposed intrinsic cysteines (ERK2 Cys-free, ERK2 C63S/C159S/C164S/C252S), was strongly destabilized as compared to wild type ERK2; ∆∆G between inactive wild type ERK2 and inactive mutant is 2.42 kcal/mol and ∆∆G between active wild type ERK2 and active mutant is 4.22 kcal/mol. Only the ERK2 I254C (ERK2 C63S/
C159S/C164S/C252S/I254S) construct has a slightly lower ∆∆Ginactive of 2.04 kcal/mol
and ∆∆Gactive of 2.14 kcal/mol. However these values still imply strong destabilization
effects. ERK2 F181C (ERK2 C63S/C159S/C164S/F181C/C252S), ERK2 G228C (ERK2
C63S/C159S/C164S/G228C/C252S), ERK2 C252 (ERK2 C63S/C159S/C164S), and
ERK2 F329C (C63S/C159S/C164S/C252S/F329C) were predicted to have more than 10
kcal/mol destabilization in free energy as compared to inactive wild type ERK2.
Compared to wild type ERK2, these strongly destabilized ERK mutants still retained the
same features in secondary structures but the orientation of individual residues was very
distinct as indicated by structural alignment of wild type ERK2 and computed models
(data not shown). In conclusion, protein stability of dye-labeled ERK2 needs to be
154
on the stability of biosensor proteins should also be modeled and compared using models
compatible with unnatural amino acids.
Figure 2.26 Calculated ΔΔG for all tested biosensor constructs.
Results were obtained from the Eris protein stability predictor using pre-relaxation of flexible backbones of protein structures of inactive ERK2 (PDB: 1ERK) and active ERK2 (PDB: 2ERK). The change in free energy difference was calculated as: ∆∆G = ∆Gmutant
- ∆Gwild-type. Except ERK2 C63, ERK2 C164, and ERK2 C252, all other mutants contain C63S, C159S, C164S, C252C mutations.
Correlation between estimated protein stability change (∆∆Ginactive) versus average
protein yields, average labeling efficiency, average biosensor brightness, average
biosensor response to MEK binding, and average biosensor specificity of ERK2 cysteine
mutants are plotted in Fig 2.27. ∆∆Ginactive, the change in protein stability between
inactive conformations of wild type ERK2 and an ERK2 mutant, was used for correlation
because these purified ERK2 proteins are mainly unphosphorylated, as confirmed by
Western blotting. The experimental values of ERK2 I29C, ERK2 S151C, ERK2 C164,
ERK2 G228C, ERK2 C252, ERK2 I254C, ERK2 N255C, ERK2 F329C, and ERK2
L333C were used for correlation analysis. Average protein yields were measured as
amounts of soluble proteins purified from a 1 L culture using similar expression and
purification procedures. A low correlation coefficient was found between ∆∆Ginactive and
155
assays only once (Fig 2.18). Multiple smaller-scale protein productions or the use of
ratios of soluble and insoluble fractions as assessed by SDS PAGE and Coomassie blue
staining should yield more consistent data points for correlation analysis. Average
labeling efficiency of each ERK2 cysteine mutant was obtained using the dye-to-protein
ratios of mero53-labeled, mero87-labeled, and mero221-labeled ERK2 sensor proteins.
A correlation coefficient of 0.49 was found between labeling efficiency and relative
stability of inactive biosensor proteins. Since dye concentrations for calculation of
labeling efficiency include contributions from both dye-labeled proteins and free dye that
may be non-covalently associated with the surface of biosensor proteins, one possible
explanation of low correlation between labeling efficiency and protein stability may be
that, due to misfolding, the surface of some destabilized mutant proteins becomes more
hydrophobic than that of wild type ERK2, which results in increased association of free
dye through hydrophobic interactions. These low-molecular weight non-covalently
associated free dyes would dissociate from dye-labeled proteins upon denaturing, thus
ratios of free dyes and dye-labeled proteins could be quantified by SDS PAGE and
fluorescence gel scanning. Average brightness and average biosensor response of these
dye-labeled proteins showed no correlation with predicted protein stability of biosensor
proteins because dyes were attached to different regions on the ERK surface. The
correlation value would be useful when comparing ERK mutants with an identical site of
dye attachment, but different combinations of cysteine mutations for removal of intrinsic
exposed cysteines. Incorporation of protein stability predictions and measurements of
156
understanding how these sensor proteins behave and whether they are truly able to report
events of interest.
Figure 2.27 Correlation of experimental data with predicted protein stability of biosensor proteins.
∆∆Ginactive
, the change of free energy difference after mutations calculated by Eris, represents the predicted protein stability value of ERK mutants. A. Correlation between average protein yields and predicted protein stability. Amounts of purified soluble proteins from a 1 L culture were used to calculate average protein yields. B. Correlation between average labeling efficiency and predicted protein stability. Labeling efficiency was defined as the relative concentrations of dyes and proteins of dye-labeled ERK measured in phosphate buffer. C. Correlation between average biosensor brightness and predicted protein stability. Average biosensor brightness values measured the average dye emission intensity values of mero53-, mero87-, or mero221-labeled ERK. D. Correlation between average biosensor response and predicted protein stability. Average biosensor response was defined as the maximum fluorescence change upon addition of constitutively active MEK1. E. Correlation between average biosensor specificity and predicted protein stability. Average biosensor specificity is defined as the ratio of
157
fluorescence response towards active MEK1 versus the total amounts of fluorescence response against all test proteins. F. A table of correlation coefficients of the above comparison.