Multiple RNA- and DNA-binding proteins exhibit direct transfer of polynucleotides with implications for target-site search

Significance Classically, the lifetime of a protein–ligand complex is presumed to be an intrinsic property, unaffected by competitor molecules in free solution. By contrast, a few oligomeric nucleic acid–binding proteins have been observed to exchange competing ligands in their binding sites, and consequently their lifetimes decrease with competitor concentration. Our findings suggest that this “direct transfer” capability may be a more general property of nucleic acid– binding proteins. Thus, many DNA- and RNA-binding proteins could reduce the dimensionality of their search for their target sites by direct transfer to nucleosome DNA, instead of relying entirely on three-dimensional diffusion. Furthermore, direct transfer from nascent RNA to DNA may explain why so many DNA-binding proteins also bind RNA.

Recombinant murine TREX11-242 protein (catalytic core) was overexpressed and purified as previously detailed (2). Briefly, protomers were expressed as a fusion protein using a pLM303x vector that encodes maltose-binding protein (MBP) linked N-terminally to TREX1 with a rhinovirus 3C protease (PreScission Protease) recognition site. Plasmid was transformed into E. coli Rosetta2(DE3) cells (Novagen) for overexpression, cells lysed with an Emulsiflex C3 homogenizer (Avestin), then protein purified via sequential amylose column chromatography, overnight protease cleavage, and phosphocellulose (p-cell) column chromatography. The reported wild-type plasmid was subjected to site-directed mutagenesis to introduce R174A/K175A mutations or a C-terminal FLAG tag (DYKDDDDK), then the plasmids' identities were validated by sequencing. The mutated plasmids were used to obtain mutant and FLAG-tagged protein similarly to wild-type protein, with the exception that mutant protein did not bind the p-cell column and was left as an MBP + TREX1 mixture. Preparations were determined via SDS-PAGE to be >95% purity. Active-site (protomer) concentrations were determined by spectroscopy with ε280 = 24,142 M -1 cm -1 (TREX1 and TREX1-FLAG) or ε280 = 90,300 M -1 cm -1 (MBP-TREX1 R174A,K175A ), and the equivalency of TREX1 concentrations between wild-type and mutant preparations was validated by SDS-PAGE.
Recombinant C. elegans FBF-2164-575 protein (RNA binding domain) was purified by Chen Qiu (National Institute of Environmental Health and Safety, Lab of Traci Hall) as previously described (3). Reported concentrations are for activesites (monomer), unless otherwise indicated.

FP-Based Kd Determination
Pre-reaction mix was prepared with 5 nM ligand molecule in the indicated binding buffer (see Binding Buffer Compositions), then dispensed in 36 µL volumes into the wells of a 384-well black microplate (Corning #3575). Protein was prepared at 10X the reported concentrations via serial dilution in binding buffer. Binding reactions were initiated by addition of 4 µL of the respective protein concentration to the corresponding pre-reaction mix and then incubated for 30 min at room temperature. Wells with binding buffer only were also included for blanking. Fluorescence polarization readings were then taken for 30 min in 30 s intervals with a TECAN Spark microplate reader (Ex = 481 ± 20 nm, Em = 526 ± 20 nm). Each experiment had 2 or 4 technical replicates per protein concentration (as indicated), and at least three experiments were performed per protein-polynucleotide interaction. Protein concentrations are defined previously (see Purification of Proteins).
Raw data were analyzed in R v4.1.1 with the FPalyze function (FPalyze v1.3.0 package). Briefly, polarization versus time data were calculated for each reaction, the last 10 data points for each reaction were averaged to generate an equilibrium polarization value, and equilibrium polarization values were plotted as a function of protein concentration. Plot data were regressed with Eq. 9.1-2 to calculate Kd app and n for the interactions. Values from regression with Eq. 9.1 are reported in Table 1, but Eq. 9.2 regression values are provided in Supp. Fig. 1a.

FP-Based Competitive Dissociation Experiments
Pre-reaction mix was prepared with 5 nM ligand molecule and protein ≥ 2x KdP app (at 25°C) in the indicated binding buffer (see Binding Buffer Compositions), then dispensed in 36 µL volumes into the wells of a 384-well black microplate (Corning #3575). Competitor was prepared at 10X the reported concentrations via serial dilution in binding buffer or the respective carrier polynucleotide (Table 1) at a concentration equal to the highest competitor concentration. Pre-reaction mix and competitor dilutions were then incubated at the indicated temperature until thermal and binding equilibrium (4°C/90 min, 25°C/30 min, or 37°C/30 min). Competitive dissociation reactions were initiated by addition of 4 µL of the respective competitor concentration to the corresponding pre-reaction mix, then fluorescence polarization readings were immediately taken (the delay between initiation of the first reactions and the first polarization reading was ~90 s) at the indicated temperature (4/25/37°C) for 120 min (the streptavidin-biotin dissociation rate was so slow that the reactions had to be extended to 24 h with the plate placed in a humidity cassette to mitigate evaporation) in 30 s intervals with a TECAN Spark microplate reader (Ex = 481 ± 20 nm, Em = 526 ± 20 nm). Each experiment had 4 technical replicates per competitor concentration, and at least three experiments were performed per protein-polynucleotide interaction unless otherwise indicated. Specific protein concentrations used were (Protein-Prey = [Protein]): TREX1 R174A,K175A -ds-[F]d(N)60 = 250 nM, All Others = 100 nM. Protein concentrations were determined as described (see Purification of Proteins).
Raw data was analyzed in R v4.1.1 with the FPalyze function (FPalyze v1.3.0 package). Briefly, polarization versus time data were calculated for each reaction, the polarization data were normalized to the maximum and minimum polarization across all reactions, and each normalized reaction was fit with an exponential dissociation function (Eq. 10.1) to determine Nmin, λ, and koff obs (Eq. 10.1-2). Nmin values were plotted as a function of competitor concentration and regressed with Eq. 10.3 to determine IC50 for the competitors (Supp. Fig. 1b). Then koff obs values were plotted as a function of competitor concentration. Plot data (with background koff obs subtracted) were regressed via Eq. 5.1 then Eq. 5.2 with tuning parameters constrained to the Eq. 5.1 solutions, and the regression models were compared with the Bayesian Information Criterion (BIC) (5). Rate constants (k-1P and/or kθD) were reported from the best-performing regression model. If minimum polarization was not reached during competition experiments (e.g., due to a weak competitor), then it was manually defined with minimum polarization data from corresponding binding curve data (see FP-Based Kd Determination).

FP-Based Stoichiometry Experiments
Binding curve experiments were performed as described above (see FP-Based Kd Determination), with a few exceptions. (1) The ligand concentration was 250 nM. (2) Single experiments were performed with 4 technical replicates.

TIRF Microscopy-Based Single-Molecule Experiments
PEG-biotin coated microscope slides were prepared as previously described (6). Slides were (1)  For single-label experiments, TREX1-conjugated slides were photobleached with a 640 nm laser for ~5 min, treated with 1 nM [Cy5]d(N)5 ligand ± 10 µM d(N)5 competitor in imaging buffer (3 mM Trolox, 1% v/v glucose, 1 mg/mL glucose oxidase, and 0.1 mg/mL catalase in reaction buffer), then immediately imaged with 640 nm laser excitation and a red wavelength bandpass camera filter. Data herein are from two independent days of experimentation with data collection at two different power/exposure settings for at least four replicate movies per condition: A-0.2s is movies from day-1 at 200 ms exposure and ~710 µW laser power, A-0.5s is movies from day-1 at 500 ms exposure and ~230 µW laser power, B-0.2s is movies from day-2 at 200 ms exposure and ~710 µW laser power, B-0.5s is movies from day-2 at 500 ms exposure and ~230 µW laser power. Representative composite images of collected movies can be found in Supp. Fig. 5a.
For dual-label experiments, TREX1-conjugated slides were photobleached with 532 nm and 640 nm lasers for 10 min, treated with 1 or 10 nM [Cy5]d(N)5 + 1 or 10 nM [Cy3]d(N)5 (days A or B, respectively) in imaging buffer (3 mM Trolox, 1% v/v glucose, 1 mg/mL glucose oxidase, and 0.1 mg/mL catalase in reaction buffer), then immediately imaged. Co-localization experiments used 532 and 640 nm laser excitation with green and red wavelength bandpass camera filters, and FRET experiments used 532 nm laser excitation with green and red wavelength bandpass camera filters. Colocalization data are from two independent experiments with four replicate movies each, collected with ~710 µW (red) and ~1.25 mW (green) laser powers. Co-localization movies were 5 min at 150 ms (day A) or 200 ms (day B) exposures. FRET data were collected from the same dual-label experiments as the co-localization, each with four replicate movies collected under similar laser powers as the co-localization data. FRET movies were 5 min at 150 ms exposure (day A) or 1 min at 50 ms exposure (day B).
For single-label experiments, each movie was analyzed in R v4.1.1 with the SMBalyze v2.0.6 package. Briefly, (1) identities of the tiff stacked image files were blinded with the blind.input function, (2) the id.spots function was used to create a composite image, identify particles, and calculate signal intensity over time for each particle, and (3) the refine.particles function was used to manually validate particle traces, refine binding event selections, and export residence times for all binding events. The numbers of total events selected for residence time calculations for each condition were n = 825 (0 µM) and n = 947 (10 µM). The residence times were then unblinded and analyzed in R v4.1.1 with a custom script to calculate koff obs for each replicate, and to calculate k-1P and kθD from koff obs data. Movies and RData output files from each step of analysis have been deposited to Zenodo (see Software, Data, and Materials Availability).
For dual-label experiments, each movie was analyzed in R v4.1.1 with the SMBalyze v2.0.6 package. Briefly, (1) broad-wavelength control images were used to align red and green camera images with the FRET.align function, (2) the FRET.id function was used to create composite images, identify particles, pair particles, and calculate signal intensity over time for each particle, and (3) the FRET.refine function was used to manually validate particle traces. For co-localization experiment data, the total number of apparent direct transfer events identified was n = 36 (day A) and n = 34 (day B) among 379 (day A) and 752 (day B) particle traces with dynamic binding states. For FRET experimental data, no stable FRET events were detected. Movies for all dual-label experiments described herein have been deposited to Zenodo (see Software, Data, and Materials Availability).
For photobleaching experiments, streptavidin-coated slides were treated with 10 pM [Cy5]d(N)5[Bio], then imaged with 640 nm laser excitation and red wavelength bandpass camera filters. Data are from a single experiment with single movies collected at each power setting on different fields of view. Laser power settings were from 0.14-2.28 mW with 600 ms exposure. Data were analyzed in R v4.1.1 with the SMBalyze 2.0.6 package. Briefly, (1) the id.spots function was used on each tiff stacked image file to create a composite image, identify particles, and calculate signal intensity over time for each particle, and (2) the bleach.calc function was used to concurrently analyze the particle traces from each movie to calculate signal over time at each power setting, perform exponential regression on signal-time data to determine their respective kbleach (Supp. Fig. 5b -left), and perform linear regression on kbleach-power data (Supp. Fig. 5b -right). The final linear regression gave the relationship of Eq. 13.1. Using this data via Eq. 13.2, and assuming a TIRF penetration depth (d) of 100 nm and cumulative antibody-TREX1 length (∆x) of 50 nm, our Fig. 3 experiments should have a corresponding photobleaching rate constant (kb * ) of 7.5-8.3 x10 -3 s -1 depending on laser power. Applying this and our Fig. 3 data to Eq. 13.3, we conclude that 5.4-19% (230 µW) or 6.0-21% (710 µW) of our apparent dissociation events should be attributable to photobleaching, depending on competitor concentration.

Rate Constant Correlation Analysis
Data sourced from these studies (Fig. 5 -A-D) were the average values (Table 1) from initial experiments (Fig. 2 and Table 1). Data sourced from concurrent studies on PRC2 (Fig. 5 -E-F) are from 25°C isotherm data in Table 1 of the companion manuscript (7). The first published data (Fig. 5 -G) are taken from Table 2 of its reference (8). The second published data (Fig. 5 -H) are calculated from Table 1 values of reference (9), where k-1P is the dissociation rate at the lowest DNA concentration and kθD is the rate of change in dissociation rate between the highest and lowest DNA concentrations. The last published data (Fig. 5 -I-J) are taken from Figure 2 of reference (10).

Equations
For Eq. 1.1-10, terms are defined in Fig. 1c -Complete Reaction Scheme, equations give rates of change for indicated reactants as a function of time (t), and bracketed terms indicate concentrations.  Fig. 1c -Simplified Reaction Scheme, equations give rates of change for indicated reactants as a function of time (t), and bracketed terms indicate concentrations. For Eq. 2.6-10, apply Eq. 2.1-5 notation.
(Eq. 5.1) For Eq. 6.1-3, apply Eq. 2 notation, a1 and a2 are arbitrary constants, and δ is the relative dynamicity of a proteinpolynucleotide interaction. See Theoretical Background for conditions of simplification.
(Eq. 6.1) For Eq. 7, apply Eq. 6 notation, and HOPPàD is the "Hand-Off" Proficiency score for direct transfer of a protein from ligand to competitor. See Theoretical Background for context.  For Eq. 9, apply Eq. 2 notation, FPE is equilibrium polarization at a given [ET], FPmax is the maximum equilibrium polarization, FPmin is the minimum equilibrium polarization, [ET] is the total protein concentration, Kd app is the apparent equilibrium dissociation constant, and n is the Hill coefficient. For Eq. 10.1-2, apply Eq. 2 notation, Nt is relative polarization at a given time (t), Nmin is the minimum relative polarization, λ is the exponential rate constant, [DT] is the total competitor concentration, ID is the equilibrium signal at a given [DT] (Nmin for Eq. 10.1), Imax is the ID at [DT] = 0, Imin is the ID at [DT] à ∞, IC50 is the concentration of [DT] at which equilibrium [EP] is halved relative to [DT] = 0, h is the Hill slope, koff obs is the apparent dissociation rate, and DTP (Fig. 2f, 'Proportion of Direct Transfer') is the proportion of protein transfer between ligands that proceeds through a direct versus classic transfer pathway (Fig. 1b)  For Eq. 12.1-5, apply Eq. 2 notation.
(Eq. 12.1) (Eq. 12.2) (Eq. 12.3) (Eq. 12.4) For Eq. 13.1-3, apply Eq. 2 notation, kb * is the photobleaching first-order rate constant (s -1 ) for a [Cy5]d(N)5 ligand conjugated to a streptavidin-coated slide via biotin-labeled α-FLAG antibody and FLAG-tagged TREX1 protein, W is the excitation power for the laser (watts), ∆x is the additional separation between slide and fluorophore imparted by the antibody and TREX1 (nm), d is the TIRF penetration depth (nm), ε is the proportion of apparent dissociation events attributable to photobleaching, and 2. 8  b Simulations used 10-hour reaction window, due to low koff obs ; default was 2 hrs. c Analyses used simulation-provided baseline signal for koff obs calculations, due to incomplete competition. d Simulations used 20% measurement error; default was 5%. † Simulations generated 100 data sets for analysis; default was 20.