Burst experiments in solution

Data was collected for the double-labeled DNA hairpins in TEpH7.6 at salt concentrations of 10, 160, and 320 mMNaCl (for details see methods in chapter 4.2.2).

Assuming a two-state kinetic system, one expects a gradual change of the equilibrium from a completely open NDA hairpin conforma-

4.2 Influence of dye selection on DNA hairpin dynamics 75

Proximity Ratio (ε)

Figure 4.12: Proximity ratio distributions for hairpins labeled with the Cy3–Cy5 dye pair obtained from single-molecule burst analysis experi- ments. Comparison of the FRET efficiency distributions for the Cy3–Cy5 double-labeled hair-pins recorded at 0mM(dark blue), 10mM(light blue),

20mM(black), 40mM(gold), 80mM(green), 100mM(red), 160mM(pur-

ple), and 200mM(grey) NaCl concentration. Note that since for all salt con-

centrations only one populations exists an estimation of theγ factor was

not possible (see chapter 3.2.4.3 for details). Hence the presented Proxim- ity ratios have not been corrected for differences in fluorescence quantum yield, differences in detection efficiency, crosstalk and direct excitation.

76 New developments and applications

tion at low salt (low-FRET-state-populated) to a completely closed conformation at high salt concentrations (high-FRET-state-populated) due to a shielding of the DNA backbone charge by the salt ions. From the kinetic rates determined in the FCS analysis (figure 4.10) one can estimate 3–8 transitions during the average burst duration of ∼ 1 ms. Therefore, the mean FRET efficiencies per burst are likely to be averaged out and one expects only one distinct peak at any salt con- centration, with the position of the peak determined at the respective equilibrium value [90]. For a two state system this peak position is directly related to the ratio of the rates since the FRET efficiency of this peak is the result of a dynamic averaging over the states present throughout the duration of each burst

The hairpins labeled with the Cy3–Cy5 dye combination match this expected behavior, starting from a single low FRET population at low salt and reaching almost 100% population of the closed state at ∼100 mMNaCl (figure 4.12). In contrast, for a variety of combi- nations of 6-Tamra, Atto and Alexa dyes, different FRET histograms were observed (figure 4.13 A-F). While at low salt concentrations the histograms for all dye combinations show the expected shape, this is no longer true for elevated salt concentrations.

At 10mMNaCl all hairpins are found predominantly in the open state and the main peak of the histogram is at low FRET efficien- cies (∼10-20%) with few observations of high FRET bursts. For high salt concentrations (160 mM), the different dye pairs lead to distinct histograms. Under such conditions (at least) two distinct peaks are observed. These distributions cannot be explained using a simple two-state model, in agreement with deviations from the simplistic picture in the FCS fit residuals discussed above (see figure 4.11).

A method to analyze whether a population is static or dynamic has been suggested recently by Santoso et. al. [110][109]. This method relies on the calculation of numerous proximity ratios (ǫ) for each burst by using a sliding window that contains always a constant number of photons (here 10 photons). From these proximity ratios (ǫ) one can calculate a standard deviation of the meanσǫ for each burst. This σǫ can then be compared to the theoretical shot noise limited

4.2 Influence of dye selection on DNA hairpin dynamics 77 Atto 647N Alexa 647 6 -T a m ra A tto 5 3 2 A le x a 5 3 2 A B C D E F 10 mM NaCl 160 mM NaCl 320 mM NaCl

Figure 4.13: FRET distributions of DNA hairpins as a function of salt and choice of dye pair obtained from single-molecule burst analysis ex- periments. FRET-efficiency distributions for different FRET pair combi- nations measured at 100 µWlaser power with NaCl concentrations of 10

mM(blue), 160 mM(red) and 320 mM(green). Shown are burst analysis

data for Atto532–Alexa647 (A), Atto532–Atto647N (B), Alexa532–Alexa647 (C), Alexa532–Atto647N (D), 6-Tamra–Alexa647 (E), and 6-Tamra–Atto647N (F).

78 New developments and applications

ε ε ε ε

A B C D

Figure 4.14: Standard deviation analysis simulations (Adapted from [109]). A burst analysis was performed on simulated timetraces requiring at least 6 photons within 500msand a total of 30 photons per burst. Data shown was

additionally limited to bursts longer than 4ms. The theoretical shot noise

limitσǫSN(green parabola) was calculated for a 20 photon sliding window

as was also used for the calculation ofσǫpresented here as a function of

ǫ in a two dimensional contour plot. Simulated was a one state system

(A), an systems fluctuating between two states of ǫ = 0.5and ǫ = 0.7at rates much slower than the diffusion through the focal volume (17 s−1,

B), approximately equal to the diffusion time (166s−1, C) and faster than

the diffusion time (1660 s−1, D). For a full description of the simulation

experiment see [109].

standard deviationσǫSN (see Eq. 4.125).

σǫSN =

s

ǫ(1−ǫ)

1+GG+GR (4.12)

For a whole burst analysis dataset this is usually done in a two dimensional plot ofσǫ andǫ(see figure 4.14) [109].

From the simulations in figure 4.14 it becomes obvious what to expect from experimental data. Shot noise limited static bursts and bursts with dynamics slower than the diffusion time will appear close to the theoreticalσǫSN calculated for the respectiveǫ(green line). In

contrast dynamic bursts with rates faster than the diffusion time will show aσǫ exceeding the shot noise limit.

One drawback of this method however is the limitation to prox- imity ratios which, for different dyes, are not as easily comparable as FRET efficiencies since no corrections for the differing quantum

5_{Since a sliding window of 10 photons is used for the data presented hereGG+}

4.2 Influence of dye selection on DNA hairpin dynamics 79

yields, and spectra are applied. Hence the information gained from actual experiments (see figure 4.15) is limited to a comparison of the relative amount of dynamic events and populations clearly shot noise limited between the different datasets.

Here, especially the 6-TAMRA dye combinations and Alexa532- Atto647N are mostly distributed around the shot noise limit indicat- ing sub-populations stable on the timescale of the observation win- dow (_∼2ms) for these samples which are discussed in the following chapter.

In contrast Atto532-Alexa647, Alexa532-Alexa647 and to a lower extent Atto532-Atto647N show clear indications of dynamic averag- ing due to fast interchanging states as indicated by the burst popula- tions significantly exceeding the shot noise limit.

To further resolve the nature of the individual sub-states, in the next chapter in gel burst analysis experiments will be used to allow for a separation of open and closed hairpins.