Bleaching‐resistant, Near‐continuous Single‐molecule Fluorescence and FRET Based on Fluorogenic and Transient DNA Binding

Abstract Photobleaching of fluorescent probes limits the observation span of typical single‐molecule fluorescence measurements and hinders observation of dynamics at long timescales. Here, we present a general strategy to circumvent photobleaching by replenishing fluorescent probes via transient binding of fluorogenic DNAs to complementary DNA strands attached to a target molecule. Our strategy allows observation of near‐continuous single‐molecule fluorescence for more than an hour, a timescale two orders of magnitude longer than the typical photobleaching time of single fluorophores under our conditions. Using two orthogonal sequences, we show that our method is adaptable to Förster Resonance Energy Transfer (FRET) and that can be used to study the conformational dynamics of dynamic structures, such as DNA Holliday junctions, for extended periods. By adjusting the temporal resolution and observation span, our approach enables capturing the conformational dynamics of proteins and nucleic acids over a wide range of timescales.


Bi-exponential fitting of binding kinetics.
To give readers straightforward numbers for the hybridisation kinetics in the main text, we have calculated mean values of tof f and ton, and also the inverse values as kon and kof f , respectively.
From previous studies, we know, however, that the hybridisation kinetics are indeed better described by a bi-exponential decay with two independent decay constants.To further characterise the binding behaviour, we thus fitted a bi-exponential distribution to the dwell times (extracted as described in the main experimental section) of both r-labels using a maximum-likelihood-estimator in Matlab (MEMLET [1] ).The fitted function takes into account the minimum measurable dwell time (one frame) and the discrete nature of the observable values.Confidence intervals for fitted decay times were determined via bootstrapping (Table S2).The results show that the green label at 100 nM has on-rates of kon,1 =0.31 s -1 and kon,2 =0.031 s -1 with 75.8% of events following kon,1.The off-rate constant are kof f ,1 =0.48 s -1 and koff,2 =0.038 s -1 with 34.4% of all events following kof f ,1 (see Figure S1A, C).For the red r-label, onrates of kon,1 =3.50 s -1 and kon,1 =0.12 s -1 with 90.5% described by kon,1 and off-rates of kof f ,1 =0.90 s -1 (78.4%) and kof f ,1 =0.031 s -1 were observed at 20 nM (Figure S1B, D).
Characterising self-quenching R-label strands.We additionally tested a construct terminally labelled with two ATTO655, two ATTO647N and Dabcyl, and ATTO647N and BHQ3.
All constructs were 11-nt long, except the ATTO655 labelled probe, which had a length of 8 nt.
To characterise the level of quenching, we performed ensemble measurements, assessing the absorption spectra and the fluorescence of the quenched r-labels in absence and presence of up to 100-fold excess of complementary DNA.
The fluorescence spectra were measured at a scanning spectrofluorometer (PTI) using 1-s integration time per 1-nm-wavelength intervals using 100 μL r-labels in buffer (50 mM HEPES, pH 7.4; 200 mM MgCl2, 10 mM NaCl, 0.1 % BSA, the same buffer as used for single molecule measurements) to a final concentration of 100 nM.Complementary DNA was added stepwise to achieve different concentrations until saturation of the signal was observed (0.1-10 μM).
The emission spectra (Figure S2) were recorded in presence of 0 to 10 µM complementary DNA (complementary DNA was added until the fluorescence signal saturated).Upon hybridisation, the increased stiffness of dsDNA forces the dye-dye interactions apart and thus should de-quench the probes.All red r-labels (Figure S2A-D) show an increase in fluorescence upon addition of an increasing concentration of complementary DNA, most prominent the 2xATTO647N and the ATTO647N-BHQ3 probes (4-fold and 6-fold increase, respectively).The level of fluorescence in the ATTO647N-BHQ3 probe is, however, still lower than the fully quenched 2xATTO647N.We reasoned that there is significant FRET from ATTO647N to the BHQ3 even in the hybridised state, which prevents fluorescence emission and thus renders the probe unsuitable for our purpose.For further single-molecule experiments, we selected the 2xATTO647N r-label as a red label, since it proved to be the most suitable out of the tested selection.
For the green r-label, we observe a 16-fold increase in fluorescence intensity under the same conditions (Figure S2E), which suggests that indeed increasing the distance between dark quencher and fluorophore in the de-quenched state (compared to the ATTO647N-BHQ3 pair) is allowing for strong emission once hybridised with complementary DNA.   Figure S1: Characterisation of the binding kinetics for the r-labels.The data was fitted to a bi-exponential distribution using a MLE algorithm in Matlab (MEMLET [1] , red graph).For display purposes, the data were binned with increasing bin size, counts were normalised by the bin width.Error bars on bins are derived from bootstrapping (200 iterations).

Figure S3 :
Figure S3:Individual frames imaged at various concentrations of green r-labels (0-5μM) in the imaging buffer.At 0 nM, the observed spots come from immobilised Cy3B.The subsequence images show spots of green r-labels at various concentrations binding to complementary, immobilised docking strands.Through the fluorogenic nature of the r-lables, individual targets can still clearly be identified at 5μM of r-lables in the imaging buffer.

Figure S4 :
Figure S4: Conformational dynamics in the HJ observed with single r-labels.FRET distributions (C-D) and dwell time histograms (E-F) for the HJ with only (A, C, E) or green (B, D, F) r-label, as indicated by the schematics in A and B, respectively.Data from one representative experiment of 179 (red r-label) and 72 molecules (green rlabel).

Figure S5 :
Figure S5: Additional traces of the green r-label binding at 100 nM.Repeated binding can be observed by the signal rising to ≈450 counts.Occasionally, the signal reaches higher counts (≈ 800 counts), which indicates intervals of Cy3B fluorescence without functional BHQ2.

Figure S6 :
Figure S6: Additional traces of the red r-label binding (100 nM).Repeated binding can be observed by the signal rising to ≈1000 -1500 counts.Occasionally, the signal shows lower counts (≈ 750 counts), which indicates intervals of only one functional ATTO647N dye emitting.

Figure S7 :
Figure S7: Additional traces of the green (300 nM) and red r-label (100 nM) binding in the FRET regime on the HJ with AA (top, grey), DD, and DA channel (middle, green and magenta, respectively) and calculated FRET efficiency (E, bottom, blue).The anticorrelated fluctuations in the DD and DA channels (middle) and the fluctuating E trace (bottom) indicate a dynamic interchange between a high-FRET state (E ≈ 0.75) and a low-FRET state (E ≈ 0.25).

Table S1 :
Sequences of DNA strands.The HJ is formed by nucleotides in capitals, other nucleotides are involved in r-label binding.

Table S2 :
Parameter mean and 95% CI from bootstrapping (200 iterations) the binding kinetics of the r-labels.