A Biophysics Toolbox for Reliable Data Acquisition and Processing in Integrated Force–Confocal Fluorescence Microscopy

Integrated single-molecule force–fluorescence spectroscopy setups allow for simultaneous fluorescence imaging and mechanical force manipulation and measurements on individual molecules, providing comprehensive dynamic and spatiotemporal information. Dual-beam optical tweezers (OT) combined with a confocal scanning microscope form a force-fluorescence spectroscopy apparatus broadly used to investigate various biological processes, in particular, protein:DNA interactions. Such experiments typically involve imaging of fluorescently labeled proteins bound to DNA and force spectroscopy measurements of trapped individual DNA molecules. Here, we present a versatile state-of-the-art toolbox including the preparation of protein:DNA complex samples, design of a microfluidic flow cell incorporated with OT, automation of OT-confocal scanning measurements, and the development and implementation of a streamlined data analysis package for force and fluorescence spectroscopy data processing. Its components can be adapted to any commercialized or home-built dual-beam OT setup equipped with a confocal scanning microscope, which will facilitate single-molecule force–fluorescence spectroscopy studies on a large variety of biological systems.

In a typical dual-beam OT-confocal fluorescence imaging measurement, the trapping lasers are used to trap two beads, after which one traps an individual DNA molecule between them.The resulting bead-DNA complex is then steered towards a separate compartment to load proteins onto the DNA, and/or to carry out force spectroscopy or fluorescence imaging measurements.To avoid intermixing between the bead, DNA, and buffer solutions, it is highly preferable to maintain them in distinct compartments.A widely used compartmentalization strategy is to create distinct channels using laminar flow (1), (2), as this obviates the introduction of physical barriers (see the bead, nucleic acid, and buffer channels in Figure 3A).While highly convenient, this approach does come with the limitation that the flow has to be maintained to ensure separation between the channels during long measurements or incubation.Furthermore, constant flow increases the shear stress on the DNA substrate, which facilitates dissociation of protein:DNA complexes, and can complicate force spectroscopy measurements.
To overcome these issues, we have designed a flow cell with relatively large buffer reservoirs (~0.5 µL) separated from the main channel by a long thin connecting neck (1.5 mm long, 0.25 mm wide, Figure 3A) to reduce the rate of diffusion between them.This allows the user to turn off the flow in all channels and carry out more precise imaging and force spectroscopy measurements there.
The inclusion of several such reservoirs provides the flexibility to include different storage reservoirs for reagents used in different incubation and/or imaging steps.For example, a protein can be loaded onto the tethered DNA molecule in one reservoir at the single-molecule level (Sample preparation for investigating protein:DNA interactions section), and subsequent measurements can be carried out in a different reservoir, if necessary.In our instrument, steering the DNA molecule from one reservoir to an adjacent reservoir by moving the sample stage relative to the trapped beads takes ~30 s (moving at a speed of 0.2 mm/s).At this movement speed, the drag force applied by the surrounding fluid on the beads (1.7 µm diameter) is estimated to be around 3 pN, which is sufficiently low to present minimal influence on the stability of the beads in the OT and the DNA molecules trapped between them.

Simulation of blurred point spread function
To quantify the motion blurring occurring in 2D confocal scan images, we have simulated the scan process based on the following parameters: Our simulations show that the blurring produces the elongation (in the moving direction of the fluorescent spot) and rotation of the PSF function.These two quantities are our metrics to determine the motion blurring.We use a rotated Gaussian fitting function to measure them: Here, A represents the amplitude, B the noise background, (x0, y0) the center coordinates, (σx ,σy ) the standard deviations, and θ the rotation angle.
For example, we simulated the unidirectional movement of a fluorescent spot at 1.7 nm/s (corresponding to a eukaryotic replicative DNA helicase) (3), which gives rise to negligible motion blur.In contrast, a fast moving fluorescent spot (200 fold faster, 340 nm/s), gives rise to a clearly rotated Gaussian point spread function, as shown in the table and figures below.

Fitting of spot intensity profile
The intensity profiles of detected fluorescence spots are fitted by 2D Gaussian function with a background term: where the fitted parameters are: A: The amplitude (intensity value corresponding to the mean possition) of the detected spot.
(x0, y0): The x and y coordinates of the detected spot, respectively.
(σx ,σy ): The standard deviation in x and y of the detected spot, respectively.

B:
The intensity correction due to background noise.
The estimated intensity value of any spot corresponds then to the sum of photons in a circular mask (given a detection radius), taking into account the background noise B.

Fluorophore bleaching step size calibration
To calibrate the minimum and average bleaching step size (∆  and ∆  , respectively) of a fluorophore, we typically singly label dCas9 with this fluorophore, attach the labeled dCas9 to a unique location on the DNA molecule through the use of an appropriately selected guide RNA, and continuously image the dCas9-bound fluorophore until it bleaches.In this case, provided that the fluorophore is well-behaved, the overwhelming majority of fluorescent spots should exhibit one bleaching step.Quantification of its associated mean intensity provides an excellent basis for starting the calibration process.The user inputs an initial guess for the minimal step size ∆  for the dCas9 dataset, after which CTrapPy generates fitted intensity plots and a step size distribution graph.The step size distribution has a mean ∆  and a standard deviation  ∆ .The user updates the initial guess until two conditions are met: • A large majority of the fitted intensity plots look good upon visual inspection; • ∆  ≤ ∆  − 2 ∆ , meaning that at least 95% of bleaching steps will be larger than the minimum step size.
In the case of multi-color measurements, it is important to check the crosstalk between different channels, and correct the fluorescence intensity taking the crosstalk into consideration.The method to correct for crosstalk is described in Colocalization of tracks and intensity correction section.

Piecewise linear motion fit
First, a Kalman filter is applied to the spot locations of each trace using the pykalman Python library, in order to reduce noise.The resulting locations are fit to linear segments using CPA, with a linear cost function, to fit multiple linear segments.The minimum segment size and penalty term can be set by the user After the fit, every detected spot has an associated fitted velocity vCPA, which has lower noise than an instantaneous velocity calculated by simply subtracting subsequent spot locations.Of course, the piecewise linear fit only makes sense for static traces (with or without drift), and traces exhibiting piecewise linear motion.
The mean velocity μv of vCPA for all traces in the static calibration dataset can be used for drift correction, using where σx is the location error.The standard deviation σv of the velocity distribution can be used as a cutoff to determine if a spot is static or is undergoing motion.

Anomalous diffusion analysis
After MSD analysis on the drift-corrected traces, we can fit an anomalous diffusion exponent α.
This exponent gives us a characterization of motion type for each trace.We start with: where Dα is the anomalous diffusion constant, τ is the lag time and σx is the location error.α ≪ 1 corresponds to spots exhibiting confined diffusion; for diffusing particles we expect α ≈ 1; α ≫ 1 indicates particles undergoing unidirectional motion.The actual fit of α is done using the formula so that we can use linear regression.We use least squares to fit up to a maximum lag time corresponding to one third of the length of the actual fit.After that, the MSD generally exhibits fluctuations that are too large to perform reliable fitting.

Supplementary Figures
Figure S1 Automated Figure S1 Automated trapping of DNA in dual-beam OT.A: Functionalized DNA molecules are brought in contact with the beads by the flow to have one end of the DNA attached to the beads.Those DNA molecules attached to Bead 1 are stretched by the flow and the other end of the DNA is brought close to Bead 2 by the flow.Bead 2 is constantly moved close to and away from Bead 1 on the x-axis.The force exerted on the beads is low and independent of the bead-to-bead distance when the DNA is not tethered between the beads.B: Once the other end of the DNA is attached to Bead 2, rightward movement of Bead 2 generates a force that exceeds a force threshold set by the user, indicating that one or more DNA molecules are tethered between the beads.C: The flow is then turned off, leaving the remaining DNA molecules attached to the beads in a coiled and collapsed conformation.

Figure S2
Figure S2 Projection of fluorescence intensities on the x-axis to monitor fluorophore bleaching.A: Two fluorescent spots are present in the 2D confocal scanning image (upper panel), giving rise to two peaks on the fluorescence intensity projection plot (lower panel) exceeding the preset intensity threshold.The signals close to the beads (beyond the white lines on the upper panel) are excluded from the analysis by specifying a bead margin.B: After the fluorophores bleach, the background fluorescence intensity falls below the threshold.The continuous confocal scan is subsequently terminated.

Figure S3
Figure S3 Construction of a kymograph.(left panel) 1D scans are repeated along a selected line along the x-axis.A kymograph is constructed by stacking these 1D scans to form a visual representation (right panel) in which one axis is the x-axis and the other axis represents time.The motion of the fluorescent molecule over time is shown by a green line (indicated by the green arrow in the right panel).

Figure S4
Figure S4 Comparison of the localization precision of 1D and 2D scans.A: A fluorescently labeled DNA oligo is hybridized to a single-stranded flap on a double-stranded DNA trapped between two beads to form a static fluorescent spot in the confocal scanning image.The fluorescent spot is detected by continuous 2D confocal scans with a pixel size of 50 nm and a frame rate of 1 s/frame until the fluorophore bleaches.The x-and y-localizations of the spot are indicated by the white dashed lines, respectively.B: The x-localization of the fluorescent spot in each 2D image is determined by 2D Gaussian fitting and plotted against time.Mean and standard deviation of the localization are indicated in the figures.The distribution of the x-localization is shown in the histograms at right.C and D: The x-localization of the fluorescent spot in each frame is determined by 1D Gaussian fitting to a selected line in the 2D confocal scan image (white dashed line in panel A) to mimic the line scans used to build kymographs.Two different y-coordinates close to the center of the fluorescent spot are selected to carry out this 1D scan analysis.The average locations and standard deviations are shown in the figures.The distribution of the x-localizations is shown in histogram.E: The time traces and distributions of x-localizations of the fluorescent spot determined by 2D images versus 1D scans are overlayed.The standard deviations of x-localizations determined by 1D scan analysis are approximately 2 fold higher compared to 2D scan analysis.Therefore, 2D scan analysis yields a ~2 fold higher fluorophore localization precision than 1D scans.