Nano-bead calibration

First, we investigated the effect of a confocal volume mismatch on the obtained coincidence. To resemble the spatial profile of a fluorophore pair attached to a (bio)molecule (complex), we used a custom-designed nano-bead DNA-origami structure carrying a distribution of 5–10 dye copies of each fluorophore (Alexa 488 and Atto 647N) at a distance of approximately 10nm. By labeling each molecule with multiple dye copies we ensure 100% dual-labeling of all molecules and exclude a full darkening because the dyes blink independently.

Due to the stochastic character of the diffusion process (random walk) many different trajectories through the confocal detection volumes occur (Fig. 1a). When a molecule transits through the volumes it will emit a burst of fluorescence photons. Each burst appears as a dip in the inter-photon lag (IPL) and is detected by means of a burst threshold (dotted and dashed line in Fig. 1b for blue and red channel, respectively). The distribution of the number of photons per burst for the red channel is depicted in Fig. 1c. Central trajectories (trajectory C) that correspond to a high number of photons per burst are rarely observed. In contrast, borderline and peripheral trajectories (trajectory A and B, respectively) that correspond to bursts with a small to medium number of photons are rather frequent. The same observations can be made for the blue channel (see Supplementary Fig. 2a). If a borderline trajectory is followed (trajectory A), the IPL time trace of the red channel drops below the burst threshold, whereas the IPL time trace of the blue channel does not (see solid lines in Fig. 1b). As a consequence, no coincidence is detected although the molecule is in fact dual labeled. In the case of a peripheral trajectory (trajectory B), it is possible but not certain that the molecule is detected in the blue channel to confirm the coincidence. Only a central trajectory through both detection volumes leads to a pronounced burst in both detection channels, meaning that only bursts whose number of photons exceeds a certain brightness threshold should be considered for the coincidence analysis.

Figure 1d shows how the fraction of coincident bursts evolves when the brightness threshold n_br (normalized to the mean number of photons per burst) is increased. The curves for both colors are depicted, as the brightness threshold can either be applied to the red channel and the coincidence to all burst in the blue channel is evaluated or vice versa. Both curves show a similar trend, i.e. a steep increase of coincidence for low brightness thresholds followed by a plateau for larger brightness thresholds. The coincidence fraction of the red channel starts at a value of ~40%, which means that only 40% of the trajectories within the red volume do additionally penetrate the blue volume. This is a direct consequence of the larger red confocal volume. Accordingly, the blue coincidence fraction starts at a higher value of ~55%. The fact that the initial blue coincidence fraction deviates from 100% confirms a relative shift between both volumes (see Supplementary Fig. 3). The same reasoning explains why the blue coincidence saturates at lower brightness thresholds than the red coincidence. However, only for brightness thresholds exceeding a value of 0.75 both coincidence fractions reach the expected value of 100% which underlines the need to brightness-gate TCCD.

Single dye pair calibration

Next, we investigated the impact of single dye fluorescence emission on the coincidence because most experimental assays rely on single dye labeling. For this purpose, we custom-designed a double stranded (ds) DNA molecule which was aimed to be completely labeled with a pair of Alexa 488 and Atto 647N. Incomplete dual-labeling due to potential dsDNA dissociation was minimized by attaching both dyes to the same strand (see Methods). The obtained results are summarized in Fig. 2.

Open in a separate window

Fig. 2

Benchmark of brightness-gated two-color coincidence detection with single dye pair dsDNA reference.

a Histogram of number of photons per burst for red channel (Atto 647N). b Fractions of coincident bursts as a function of normalized brightness threshold n_br. Obtained saturated coincidence fractions are 0.94 and 0.92 for red and blue channel, respectively. Burst and analysis statistics are given in Supplementary Notes 1 and 2. c Concept of global coincidence analysis. Number of selected red and blue bursts (solid lines) and those which show coincidence with the complementary color (dashed lines) are shown as a function of brightness threshold n_br. The optimal brightness threshold n_br,opt is depicted for each color separately (see Supplementary Note 5). The brightness threshold of the color with the higher number of coincident bursts is increased (here red) until the number of coincident bursts are equal, i.e. N_RB=N_BR as indicated by the gray dashed line. The number of red-only and blue-only bursts N_R0 and N_B0, respectively, are indicated by arrows. d Differentiation between color-specific coincidence fractions and global coincidence.

The distribution of photons per burst is similar to that of the nano-bead sample (cf. Figure 2a and Fig. 1c, match obtained intentionally by adjusting laser power). However, the coincidence fractions as a function of the brightness threshold (cf. Figs. 1d and ​and2b)2b) exhibit differences in the y-intercept (without brightness threshold, i.e. n_br=0), which we attribute to the smoothing of the IPL time trace and different burst thresholds. Nevertheless, a similar steep increase for brightness thresholds lower than 0.8 is observed which turns into a shallow plateau for larger brightness thresholds. The determined coincidence ratios are 94±1% and 92±1% for the red and the blue channel, respectively, and show a slight deviation from full coincidence attributed to certain incompleteness of the labeling. In addition, we performed multi-point calibrations with mixtures of different fractions of single and double-labelled dsDNA (see Supplementary Note 3). The results demonstrate the high reliability in determining coincidence fractions over the whole range of possible values.

To evaluate the effect of long-lived (ms-s) dark states on the outcome of our BTCCD, we exchanged Atto 647N to Alexa 647, a dye known to exhibit pronounced photo-switching²⁵. In this case the determined coincidence ratios are 95±1% and 79±3% for the red and the blue channel, respectively (see Supplementary Fig. 4). The coincidence values were robust against moderate laser power variations and showed a partial reduction at higher laser powers (see Supplementary Note 4).

Global coincidence analysis

For a global coincidence analysis, i.e. to determine the number of only-blue (N_B0), only-red (N_R0) and dual-labeled (N_RB,N_BR) molecules in an ensemble, we use the optimal brightness thresholds of the individual channels as initial parameters (see Supplementary Note 5 for determination of optimal brightness thresholds). These optimal brightness thresholds define the number of selected bursts N_B, N_R and the number of coincident bursts N_BR, N_RB in the blue and red channel, respectively. As the numbers of coincident bursts at the respective optimal brightness threshold are in general not equal, we increase the brightness threshold of the channel that exhibits the higher number of coincident bursts until N_BR and N_RB are equal. This reduces the absolute number of selected bursts, but barely changes the coincidence fraction (cf. plateau in Fig. 2b). In Fig. 2c the number of bursts in the red and blue channel (solid lines) and the number of coincident bursts (dashed lines) are plotted as a function of the brightness threshold. The optimal brightness thresholds $n_{\mathrm{br}}^{\mathrm{opt}}$ are 2.34 for the blue channel and 2.23 for the red channel (Supplementary Note 5). Here, the brightness threshold of the red channel is increased until the number of coincident bursts are equal (gray dashed line), making it possible to quantify the relative abundance of all three possible species of molecules (N_BR=N_RB, N_B0, and N_R0, see Fig. 2d and Table 1).

BTCCD can also be applied to molecular complexes which are linked to a pair of fluorescent proteins (FPs) of different colors. Such structures are often employed as FRET-based biosensors²⁶ and BTCCD can characterize the amount of individual chromophore maturation, a crucial parameter within these sensor constructs (see Supplementary Note 6).

Isolation and labeling of ribosomes

For the translation initiation experiments the RNase deficient E. coli K-12 strain CAN20/12E (RNase BN⁻, II⁻, D⁻, I⁻)⁴¹ was used. Ribosomes were isolated by sucrose gradient centrifugation using a zonal rotor as previously described⁴² and re-suspended in Tico buffer [20mM Hepes-KOH (pH 7.6 at 0°C), 10mM magnesium acetate, 30mM ammonium acetate, 4mM β-mercaptoethanol]. A reaction with Cy5-NHS-ester functionalized dye (GE Healthcare Life Sciences, Little Chalfont, UK) in labeling buffer [50mM Hepes-KOH (pH 7.5), 10mM MgCl₂, 100mM KCl] followed for 20min at 37°C, using a 20-fold excess of dye to minimize the unlabeled fraction of ribosomes. The excess of dye was removed by pelleting the ribosomes through a 1.1M sucrose cushion. The concentration of Cy5 and ribosomes was determined spectroscopically in a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, USA) using the absorption coefficients ε_cy5=2.5×10⁵M⁻¹cm⁻¹ (at λ=647nm) and ε_70S=4.2×10⁷M⁻¹cm⁻¹ (at λ=254nm) respectively. The label ratio was calculated to be ~6 Cy5 dyes per 70S ribosome. Labeled ribosomes were dissociated into 50S and 30S subunits by sucrose gradient centrifugation in dissociation buffer [20mM Hepes-KOH (pH 7.6 at 0°C), 0.97mM magnesium acetate, 200mM ammonium acetate, 4mM β-mercaptoethanol]. The concentration of ribosomal subunits was determined again spectroscopically using the absorption coefficients ε_cy5 (see above), ε_50S=2.8×10⁷M⁻¹cm⁻¹ and ε_30S=1.4×10⁷M⁻¹cm⁻¹ (the latter two at λ=254nm), respectively. The label ratio for the isolated subunits were calculated to be ~3 Cy5 dyes per 50S and per 30S subunit, each. Finally, the labeled subunits were re-associated with unlabeled counterparts to form empty 70S ribosomes, labeled only on the 50S or only on the 30S subunit. The re-association reaction was performed at 40°C for 1h using a 50% excess of 30S with respect to 50S⁴³. For analyzing the long term stability and the structural integrity of the 70S complex we also re-associated Cy5-labeled 50S with Alexa488-labeled 30S subunits to form double-labeled 70S ribosomes. Subsequently stability and integrity of labeled 70S ribosomes was validated by BTCCD measurements (Supplementary Fig. 5).

DNA construct for cell-free protein synthesis (CFPS)

The DNA construct expressing the GFP variant Emerald (GFPem) together with the C-terminal SecMstr AP was described previously¹¹.

CFPS reactions

The PURE system⁴⁴ without ribosomes (PURExpress Δ ribosomes, NEB #E3313, New England Biolabs, Ipswich, USA) was used in all experiments. Reactions were performed according to the manufacturer’s protocol containing 500nM of labeled ribosomes and 5.5nM of linear DNA construct. Additionally 5nM of anti-sense tmRNA (AS-tmRNA)⁴⁵, complementary to the tmRNA sequence was added to suppress tmRNA-induced ribosome-nascent chain dissociation^46,47. For the translation initiation +50S reaction 2.5, 7.5 and 10μM of unlabeled 50S subunits were added to the reaction mix, which results in a 5-fold, 15 fold and 20-fold excess of unlabeled 50S. After completion of the synthesis reactions the diffusion properties were checked by fluorescence correlation spectroscopy and subsequently the solutions were diluted to a final concentration of labeled ribosomes in the order of a few picomolar for single-molecule measurements.

Confocal fluorescence measurements and data acquisition

All experimental details of the performed measurements were described previously, for the DNA samples in refs. ^40,48, for translation initiation with ribosomes in ref. ¹¹ and for FRET-based biosensors (see Supplementary Note 3) in ref. ³⁸. Briefly, all measurements were performed with a MicroTime200 (Picoquant, Berlin, Germany) equipped with pulsed laser diodes, an UPLSAPO 60X/1.2NA objective, a 75µm pinhole and with two/four silicon avalanche photo-diodes detectors. The lasers were operated at a frequency of 20MHz and individual excitation power intensities were set for the following samples. The 488nm and 640nm excitation was used for DNA nano-beads with 4.2 and 3.7μW, respectively; dsDNA with 21 and 18μW, respectively; and ribosome/GFP constructs with 21 and 16μW, respectively. A 440nm and 510nm excitation was used for the FRET-biosensors with 20 and 17μW, respectively. To avoid any by chance coincidences of diffusing molecules the sample concentrations were carefully checked by FCS before each BTCCD measurement. For this purpose, the average number of each dye (color) within the detection volume at a time was adjusted at <N>~0.01. Aliquots from stock solutions of all samples employed in this study were diluted down to above mentioned concentrations and were measured for 20min time intervals. A sequence of several subsequent 20min measurements were summed up until a few thousand up to a few ten thousand bursts of each detection color were measured (typically 20–60min data acquisition time for a data set). A more detailed discussion on the total data acquisition time is given in Supplementary Note 9. A typical time trace of a double-labeled sample and how the burst show up in plots which make use of the IPL time are given in Supplementary Fig. 6. The identification of single-molecule bursts is described in detail in ref.⁴⁰. While for the calibration and the biosensor samples a single sample aliquot was used to obtain a full sample data set, the ribosomes samples were replaced every 30min by a new sample aliquot. An inspection of crucial measuring parameters (e.g., number of burst per time, brightness of bursts) of all 20min measuring intervals within a full data set helped to validate the sample stability.

Two-color coincidence detection analysis

After identification of all bursts in both channels, burst that contain only a single photon are discarded because they would induce an artificially small dwell time and high molecular brightness. The start time of each burst corresponds to the macro time tag of the first photon of that burst and, accordingly, the end time of a burst is defined as the macro time tag of the last photon. The burst duration is defined as the difference between its start and end time. Bursts that last more than 100-fold longer than the average burst duration are also discarded because they might belong to contaminations or aggregates. For all remaining bursts the burst intensity, i.e. the number of photons between the start and end time, and the mean number of photons per burst is calculated. In the coincidence analysis the brightness threshold is normalized to that mean number of photons in a burst and continuously increased. The coincidence is calculated for each channel independently and for each value of the brightness threshold only bursts that have more photons as defined by the brightness threshold are considered for analysis. Subsequently, the coincidence of these bright bursts to all detected bursts in the other channel is calculated. Two bursts are considered as coincident if the start or end time tag of one burst is within the start and end time tags of the other burst. The fraction of coincident bursts is calculated for each brightness threshold (e.g., for red channel f_R=N_RB/N_R, see Supplementary Fig. 1) as the number of coincident bursts (N_RB) normalized to the number of selected bursts (N_R). The related values for all presented measurements are given in the Supplementary Note Tables N1.3-5. Here the error of the fractions of coincidence values is determined by the statistical error σ of the number of bursts as given by Poisson statistics with $\sigma =\sqrt{N}$ (see also Supplementary Note 2). The data analysis was performed using self-written Matlab routines (Mathworks, Natick, MA, USA). OriginPro (9.0.0G, 64bit) was used to produce the graphical presentation of the obtained results.

C.R. O.Y. and H.H. acknowledge support by the International Helmholtz Research School of Biophysics and Soft Matter. The authors would like to thank Prof. K.H. Nierhaus for many insightful discussions and Prof. M. Pohl as well as Dr. A. Boersma for providing us with FRET biosensor samples.