Fluorescence strategies for high-throughput quantification of protein interactions

Abstract

Advances in high-throughput characterization of protein networks in vivo have resulted in large databases of unexplored protein interactions that occur during normal cell function. Their further characterization requires quantitative experimental strategies that are easy to implement in laboratories without specialized equipment. We have overcome many of the previous limitations to thermodynamic quantification of protein interactions, by developing a series of in-solution fluorescence-based strategies. These methods have high sensitivity, a broad dynamic range, and can be performed in a high-throughput manner. In three case studies we demonstrate how fluorescence (de)quenching and fluorescence resonance energy transfer can be used to quantitatively probe various high-affinity protein-DNA and protein-protein interactions. We applied these methods to describe the preference of linker histone H1 for nucleosomes over DNA, the ionic dependence of the DNA repair enzyme PARP1 in DNA binding, and the interaction between the histone chaperone Nap1 and the histone H2A-H2B heterodimer.

Figure 1.

Calculating spectrally-corrected FRET values (F_corr) is essential to determining accurate binding affinities. (A) A plot showing the fluorescent intensities (F.I.) of the raw (black filled circles with solid lines) and overlap corrected (F_corr; on left black filled sqaure with dashed line and right red filled sqaure with dashed line y-axis) FRET values; graphics correspond with the same colored axis. Data were obtained from binding nPARP1 to DNA at 250 mM NaCl. (B) The same data as in (A), but x-axis plotted on a log-scale. (C) A plot showing what happens when efficiency transfer (E) is used to calculate binding isotherms. Various theoretical curves were generated by changing γ and/or E-values; the curves are based on a K_D of 10 nM for a γ = 1.

Figure 2.

H1 binds 207 bp nucleosomes with higher affinity than free DNA. (A) Cartoon representation depicting the fluorescence (de)quenching of labeled histone H1 upon binding of DNA. (B) Pseudo-color overlay of a microplate assay showing fluorescence enhancement of H1 (green) upon binding to nucleosome or DNA. A total of 0.5 nM Alexa-647 (red) was added to each well for easier visualization and as pipetting control. (C) H1 binding curves to DNA (black filled square with solid lines) and 207 bp nucleosomes (red filled circles with solid lines) derived from data in (B). Data were globally fit with equation (3) with R² values being or exceeding 0.91. Norm. F.I. is normalized fluorescence intensity. (D) Native PAGE of samples taken from a microplate assay visualized by fluorescence (top) and ethidium bromide (bottom). Numbers 1–4 indicate band positions for H1-nucleosome, H1-DNA, nucleosome, and DNA, respectively. (E) Measurements at higher concentrations reveal 2:1 DNA:H1 (black filled square with solid lines) and 1:1 nucleosome:H1 (red filled circles with solid lines) stoichiometries. Points and error bars represent the average and SD of four replicates. Nuc., nucleosome.

Figure 3.

FRET analysis reveals salt-dependent binding of PARP1 to DNA. (A) A cartoon representation showing FRET as a result of protein–DNA interaction. (B) Representative microplate images showing the binding of donor-labeled nPARP1_Donor to DNA_Acceptor. Top three images are raw data showing acceptor, donor and FRET channels for acceptor only samples (A) and samples containing a donor–acceptor pair (DA). F_corr is the FRET image which has been mathematically corrected for spectral overlap using ImageJ software; a high contrast image of F_corr exemplifies the lack of signal in the DNA_Acceptor only titration and that signal arises only from nPARP1_Donor being present. (C) Representative binding curves showing the interaction between nPARP_Donor to DNA_Acceptor using FRET. Curves show binding reactions performed in 200 (red filled circles with solid lines) and 250 mM (black filled squares with solid lines) NaCl, respectively. Data were fit to a single binding curve (Equation 3). R² values for shown curves meet or exceed 0.97. (D) A log–log plot of salt concentration versus binding affinity reveals a linear dependence on binding between nPARP_Donor and DNA_Acceptor, indicating an ionic dependence to binding. The data are fit to a line to extract the ionic dependence on binding. (E) Stoichiometric measurements of nPARP_Donor to DNA_Acceptor performed at elevated protein concentrations (200 nM) at 175 mM NaCl. Each of the three linear phases was fit to lines, with intersections indicating both 1:1 and 2:1 DNA:PARP stoichiometries. Points and error bars represent the average and range of two replicates.

Figure 4.

(H2A–H2B)–Nap1 competition as a tool for identification of the binding interface. (A) A cartoon representation showing a competition experiment where the interaction between FRET partners is lost upon addition of unlabeled competitor protein. (B) Binding curve obtained from FRET-based binding between H2A–H2B_Donor and Nap1_Acceptor. Data were fit to a single exponential (Equation 3) with a Hill-coefficient. (C) Representative competition curves of unlabeled wild-type Nap1 (black filled square with solid lines), Nap1_1–365(blue filled triangles with solid lines), Nap1_74–417 (red filled circles with solid lines) and Nap1_74–365 (violet filled inverted triangles with solid lines) to the (H2A–H2B_Donor)–Nap1_Acceptor complex. H2A–H2B_Donor and Nap1_Acceptor remained constant at 10 nM and 50 nM, respectively, with the unlabeled Nap1 protein titrated. Points and error bars represent the average and range of two experimental replicates. R² values for shown meet or exceed 0.94. (D) Raw images of data from the competition experiment between H2A-H2B_Donor and Nap1_Acceptor with unlabeled Nap1. From top to bottom; Donor (green), FRET (red) and a pseudo-color overlay of Donor (green) and FRET (red) signals obtained from competitive binding between the FRET pair and unlabeled Nap1. (E) Data plotted from a representative competition experiment showing the raw data (black filled circles with solid lines), background corrected data (blue filled squares with solid lines) or F_corr (red filled triangles with solid lines) values. Little difference is observed in signal intensity after background correction, but a significant change is observed after spectral overlap subtraction. (F) The same data plotted as in (A), but normalized to highlight the impact of not correcting for spectral overlap. The uncorrected data significantly deviates from the F_corr curve giving a non-normal IC₅₀ curve.