We present the setup and experimental procedure to obtain smFRET data from large donor-acceptor networks with a TIRF microscope. The step-by-step analysis of these measurements with the Bayesian inference software Fast-NPS yields high-resolved structural information via the application of adapted dye models.
Single-molecule Förster Resonance Energy Transfer (smFRET) can be used to obtain structural information on biomolecular complexes in real-time. Thereby, multiple smFRET measurements are used to localize an unknown dye position inside a protein complex by means of trilateration. In order to obtain quantitative information, the Nano-Positioning System (NPS) uses probabilistic data analysis to combine structural information from X-ray crystallography with single-molecule fluorescence data to calculate not only the most probable position but the complete three-dimensional probability distribution, termed posterior, which indicates the experimental uncertainty. The concept was generalized for the analysis of smFRET networks containing numerous dye molecules. The latest version of NPS, Fast-NPS, features a new algorithm using Bayesian parameter estimation based on Markov Chain Monte Carlo sampling and parallel tempering that allows for the analysis of large smFRET networks in a comparably short time. Moreover, Fast-NPS allows the calculation of the posterior by choosing one of five different models for each dye, that account for the different spatial and orientational behavior exhibited by the dye molecules due to their local environment.
Here we present a detailed protocol for obtaining smFRET data and applying the Fast-NPS. We provide detailed instructions for the acquisition of the three input parameters of Fast-NPS: the smFRET values, as well as the quantum yield and anisotropy of the dye molecules. Recently, the NPS has been used to elucidate the architecture of an archaeal open promotor complex. This data is used to demonstrate the influence of the five different dye models on the posterior distribution.
Determining the structure of a biomolecule is a key prerequisite for understanding its function. Two well-established methods for structure determination are cryo-electron microscopy and X-ray crystallography1,2. Today, both methods provide high-resolution structural information with a resolution down to the angstrom level. These two methods have been used extensively to elucidate the structure of large biomolecules such as protein complexes. Though the existing methods have constantly been improved throughout the last decades, the complexity of biological structures still poses a major challenge to structural biology, in particular when large, dynamic and transient complexes are investigated3.
In order to study the dynamics of macromolecular complexes and the structure-function relationship in particular, single-molecule methodologies have provided useful information4. Several new strategies were developed providing an orthogonal approach on acquiring structural and dynamic information. Examples are high speed AFM5, mechanical manipulation6, fluorescence localization microscopy7, as well as single-molecule Förster Resonance Energy Transfer (smFRET)8,9. Since quite early on FRET has been termed a molecular ruler, due to the distance dependence on the length scale of biomacromolecules10.
One particularly interesting application of smFRET is to use the distance information obtained from smFRET measurements to infer structural information11,12,13,14,15,16,17,18,19,20,21,22,23. Due to the high time resolution of smFRET, the position of mobile parts of a protein structure can be localized. However, in order to extract quantitative information from smFRET data important correction parameters about the dye molecules need to be determined during the measurement24. With these correction factors, the FRET efficiency EFRET can be calculated using the formula
,
where IA and ID are the fluorescence intensities of the donor and the acceptor molecule, respectively (see Figure 2). The β-factor accounts for cross-talk, the leakage of donor emission into the acceptor channel and is calculated by
where I'A and I'D are the fluorescence intensities of the donor and the acceptor molecule after photo bleaching of the acceptor molecule.
The γ-factor corrects the difference in the relative detection efficiencies in the two channels as well as the differences in the fluorescence quantum yield of the donor and the acceptor dye. It is calculated from every individual time trace by
Note, that this description neglects direct excitation of the acceptor molecule, which sometimes becomes important and would need to be corrected for as well. For determining these correction factors it is useful to excite both the donor as well as the acceptor in an alternating scheme25 in order to differentiate between photo-physical changes and structural dynamics.
In order to not only obtain quantitative smFRET efficiencies but also quantitative structural information, the Nano-Positioning System (NPS) was introduced in 200826. The name was chosen based on its similarities to the satellite-based Global Positioning System (GPS). The NPS is a hybrid technique combining smFRET and X-ray crystallography data for the localization of unknown dye positions in biomacromolecular complexes. The crystal structure serves as a reference frame and the smFRET results are used to obtain distance information between an unknown fluorophore position (antenna) and a position known from the crystal structure (satellite). In consecutive experiments the distances between the antenna and several satellites are measured and the position of the antenna is determined by means of a statistically rigorous analysis scheme based on Bayesian parameter estimation. As a result, not only the likeliest position of the antenna is computed, but its complete 3D uncertainty distribution, the so-called posterior, visualized by credible volumes. Moreover, NPS was expanded to allow for the analysis of complete smFRET networks27.
The NPS has been used to solve a number of important questions in eukaryotic transcription, namely the course of the upstream DNA, the non-template DNA and the nascent mRNA within the RNA Polymerase II elongation complex12,28, also demonstrating the effect of transcription initiation factors26 and the dynamic architecture of an open-promotor complex29. Moreover, the NPS was used to elucidate the structure of the archaeal RNA Polymerase open complex30 and in particular the position of transcription initiation factor TFE, which binds competitively to the same site as transcription elongation factor Spt4/531.
Since then, a number of smFRET based structural approaches have been published15,18,21,23. When comparing different smFRET based structural methods, it becomes clear that the apparent precision of the method is highly dependent on the particular choice of dye models. One should note that dye molecules may exhibit different spatial and orientational behavior depending on their local environment.
To this end, Fast-NPS was introduced32. Fast-NPS uses an advanced sampling algorithm reducing the calculation times drastically. Furthermore, Fast-NPS allows one to perform a structural analysis and for each dye molecule the user can choose from a set of five different dye models which will be described next. The most conservative model, called classic, assumes that the dye occupies only one, but unknown, position. At this position, the fluorophore can rotate freely within a cone, whose size is determined from its respective (time-dependent) fluorescence anisotropy. The orientation of the cone is not known, which leads to large uncertainties when converting measured smFRET efficiencies into distances. In this respect, the model is conservative, since it will lead to the smallest precision compared to the other dye models. Only for very short distances should the assumptions made by the classic model lead to a noticeably incorrect position determination. For typical smFRET values, the correct position is always enclosed in the comparatively large credible volume.
However, since a higher precision is desirable, it is important to develop and test alternative dye models, which could help to improve the precision. If the dye rotates much faster than its inherent fluorescence lifetime, the so-called iso model can be applied. Here, the orientation factor κ2 (needed for calculating the characteristic isotropic Förster radius) is set to 2/3. As a result, the computed credible volumes are almost two orders of magnitude smaller as compared to those in the classic model32. In the case that the fluorophore is found in an environment that enables not only fast reorientation, but additionally fast motion all over its accessible volume, the meanpos-iso model should be used. In this model, the dye effectively occupies only one mean position, where the spatial averaging is accounted for by a polynomial distance conversion15. This model applies if for example the (commonly hydrophobic) dye is attached to a hydrophilic region, e.g., the DNA. Application of the meanpos-iso model leads to a further reduction in the size of the credible volumes by a factor of approximately two. However, a dye linked to a protein might bind reversibly to several hydrophobic patches in its sterically accessible volume (AV). A fluorophore that instantaneously switches between these regions, but within one region undergoes free rotation and fast localized motion is best described by the var-meanpos-iso model. For a similar situation in which the dye is not free to rotate the var-meanpos model applies. More details about these models can be found in our recent publication32.
These models provide an extensive repertoire to specifically account for the various environments a dye might encounter and applying them wisely optimizes its localization precision. In Fast-NPS every dye molecule attached to a specific position can be assigned to an individual model, such that FRET-partners are allowed to have different models. This enables limitless and close-to-nature modelling. However, it is important that one performs rigorous statistical tests to ensure that the result obtained by the final model combination is still in agreement with the experimental data. These tests are included in the Fast-NPS software.
In order to apply Fast-NPS to experimental data the measurement of (only) three input parameters is required. First, the dye-pair specific isotropic Förster radii () have to be determined. Therefore, the quantum yield (QY) of the donor dye, the donor fluorescence emission spectra and the acceptor absorption spectra need to be measured. These measurements can be carried out in bulk, using a standard spectrometer and a standard fluorescence spectrometer. For each pair, the R0 is then calculated using the freeware PhotochemCAD and can be used in the NPS analysis. Moreover, the (time-resolved) fluorescence anisotropies of the dye molecules need to be obtained using a polarization (and time) sensitive fluorescence spectrometer. However, the most important input parameters for Fast-NPS are the smFRET efficiencies measured on a single-molecule fluorescence microscopy setup, such as a total internal reflection fluorescence microscope (TIRFM).
Here, we present a step-by-step protocol for obtaining smFRET data and applying Fast-NPS (Figure 1).
1. Prerequisites and Lab Equipment
NOTE: The assembly of the measurement chamber is depicted in Figure 3. The sandwich-design of the measurement chamber comprises three major components: a quartz glass (fused silica) slide, a sealing film and a coverslip that seals the flow chamber. The measurement chamber is mounted onto a customized sample holder. The dimensions of the sample chamber and the metal holder are geared to fit to a standard microscopy quartz glass slide (76 mm x 26 mm).
2. Mounting of Flow Chambers in a Custom Holder
3. smFRET Measurement on the TIRF Microscope
4. Acquisition of the Transformation Map ("beadmap")
5. Processing and Analysis of smFRET Data
6. Displaying smFRET Data in Histograms
NOTE: In order to extract the mean smFRET efficiency of all recorded smFRET data the frame-wise data or the molecule-wise data are plotted in histograms and analyzed using Gaussian fits to (multiple) peaks. In the following, the protocol uses a commercial data analysis software (see Materials List). However, any other available software can be used instead.
7. Measurement of the Quantum Yield
where n and nStd are the refractive indices of the solvent of the sample and the standard, respectively. f(λ) and f_Std (λ) are the fluorescence intensities of the sample and the standard at the wavelength λ. A(λex) and Astd (λex) are the absorbance of the sample and reference at the excitation wavelength and Φstd is the quantum yield of the standard.
8. Calculation of the Isotropic Förster Radius
9. Measurement of the Anisotropies
and use it to calculate the anisotropy value for each wavelength:
where Ixy indicates the intensity for excitation polarization x and emission polarization y.
10. Installation of Fast-NPS Software
11. Centering the pdb File
12. Setting up the Position Priors
NOTE: All values are considered in angstrom.
13. Defining the Network Geometry
14. Calculation
15. Visualization of Results
16. Consistency Check of Chosen Model Combination
Transcription is the first step in gene expression in all organisms. In Archaea, transcription is carried out by a single RNA polymerase (RNAP). Compared to eukaryotes, the archaeal RNAP bears a striking structural resemblance to their eukaryotic counterparts while having a simpler transcriptional machinery. Thus, Archaea can be used as a model system to study eukaryotic transcription initiation by RNA Polymerase II (Pol II). Recently, the complete architecture of the archaeal RNA polymerase open complex has been determined from single-molecule FRET and NPS. The data from NPS analysis was used to build a model of the complete archaeal open promotor complex, which provides useful insights into the mechanism of transcription initiation.
To elucidate this structure, smFRET efficiencies were measured between unknown antenna dye molecules located within the open promotor complex and several known satellite dye molecules that were incorporated at five reference sites in the RNAP, whose positions are known from crystallographic structures (pdb-ID: 2WAQ)37. The antenna dyes were attached to either one of different positions on the non-template DNA, TFB, TBP or TFE. The complete network used in this study consisted of more than 60 measured distances.
Figure 7 depicts the model of the complete archaeal open promotor complex built from the NPS analysis. It comprises the double stranded promotor DNA (light and dark blue), the RNA Polymerase (grey) and the transcription initiation factors TBP (purple), TFB (green) and TFE (yellow). The model is superimposed with the results from the NPS analysis, the credible volumes, which were calculated using the classic model (A), the iso model (B), the meanpos-iso model (C), the var-meanpos-iso model (D) and the var-meanpos model (E).
Figure 1: Workflow of the acquisition and processing of the parameters needed for the Fast-NPS calculation. Please click here to view a larger version of this figure.
Figure 2: Exemplary fluorescence intensity time trace of a smFRET event. The fluorescence intensities of the donor (green) and the acceptor molecule (red) showing the three characteristic phases, namely I: smFRET, II: donor fluorescence after acceptor photobleaching, III: background fluorescence after donor photobleaching. Please click here to view a larger version of this figure.
Figure 3: Schematic illustration of the flow chamber for smFRET experiments. The flow chamber is mounted onto a customized metal holder with acrylic glass holders. The sandwich-design of the flow chamber comprises a quartz glass (fused silica) slide with two holes for attaching inlet and outlet tubing, a sealing film and a coverslip that closes the flow chamber. The prism for TIRF illumination is mounted onto the lower half of the flow chamber. Hollow tab screws provide the inlet and outlets for the flow chamber. Please click here to view a larger version of this figure.
Figure 4: Preparation of the quartz glass slide and the sealing film. Mechanical drawing of the quartz glass slide indicating the positions of the holes (given in millimeters). Please click here to view a larger version of this figure.
Figure 5: Mechanical drawing of the flow chamber. The measures for the aluminum prism holder, acryl glass holders and aluminum mounting frame are given in millimeters. Please click here to view a larger version of this figure.
Figure 6: Schematic illustration of the prism-type TIRF setup used for smFRET experiments. Abbreviations for optical components: A, aperture; DM, dichroic mirror; F, emission filter; L, lens; M, mirror; O, objective; P, prism; PSD, position sensitive photo-diode; S, sample; PS, positioning stage; T, telescope. Please click here to view a larger version of this figure.
Figure 7: Simulation results of the different model assumptions. All pictures show the archaeal RNA polymerase (pdb-ID: 2WAQ, top view) together with the model for promoter DNA (tDNA and ntDNA in blue and cyan, respectively), TBP (purple), TFB (green) and TFE (yellow) in the archaeal open complex30. The credible volumes are superimposed for the NPS simulation results of (A) the classic model, (B) the iso model, (C) the meanpos-iso model, (D) the var-meanpos-iso model and (E) the var-meanpos model. All volumes are shown at 68% credibility. The classic and the var-meanpos networks are consistent with the smFRET data. In contrast, networks where for all dyes the iso, meanpos-iso or the var-meanpos-iso model is chosen are inconsistent with the measured data. Please click here to view a larger version of this figure.
We present the setup and experimental procedure to accurately determine FRET efficiencies between dyes attached via flexible linkers to biomacromolecules, i.e., nucleic acids and/or proteins.
In order to ensure precise smFRET measurements (Section 3), it is crucial to exclude air from the flow chamber at any time during the measurement. Furthermore, make sure to not overload the flow chamber with fluorophores. The fluorophores must be clearly separated to ensure correct analysis. As smFRET pairs, which do not show bleaching of the donor have to be excluded from the analysis, make sure that >80% of the molecules in the field of view are bleached at the end of the movie. To account for inhomogeneities in the sample the β-factor and the γ-factor, correcting the cross-talk and relative detection efficiencies of the donor and acceptor channel, respectively, are calculated for each FRET pair individually.
The camera settings (integration time, electron multiplier gain, pre-amplifier gain and readout rate described in Section 3.9) should be set to values giving the best tradeoff between signal to noise ratio, dynamic range and time-resolution. They need to be re-adjusted for different experiments or if different hardware is used. The numbers of frames need to be high enough to ensure that most of the donor molecules bleach within the observation time.
For the measurements on the fluorescence spectrometer (Sections 7 to 9) a good compromise between the signal intensity and the spectral resolution of the recorded data has to be found. To this end the slits in the excitation and emission pathway of the fluorescence spectrometer have to be adapted dependent on the instrument used and the sample concentration.
Moreover, we present the Fast-NPS analysis method to obtain structural information of transient or dynamic macromolecular complexes. NPS has been applied to reveal the path of the non-template DNA strand and the position of transcription initiation factors in the archaeal RNA polymerase open complex. Using the network of more than 60 different distance measurements, we showed that Fast-NPS, equipped with a newly implemented sampling engine (Eilert, T., Beckers, M., Drechsler, F., & Michaelis, J. in preparation), reduces the time needed for the analysis of this complex smFRET network by ≈2 orders of magnitude, as compared to the original global NPS method27. The algorithm's robustness is rooted in a Metropolis-within-Gibbs sampler combined with a parallel tempering scheme. Fast-NPS shows exact reproducibility of network results and is consistent with results published earlier30.
Several different methods have been published aiming to infer structural information from smFRET measurements11,12,13,14,15,16,17,18. All of these approaches provide only one specific dye model. Thus, dyes, that do not fulfill the assumptions made by the respective model, cannot be used or lead to false structural information. Fast-NPS, on the contrary, allows to select for each dye molecule a different model. This helps to account for different conformational behavior of both, the dye molecule itself, as well as the linker used for its attachment. The local molecular surroundings of the dye molecule, as well as its physical properties will determine which model is most appropriate.
For the analyzed smFRET network of the archaeal initiation complex, an isotropic assumption for all dye molecules leads to a drastic decrease in the size of the credible volumes as compared to the classic model. In combination with a dynamic position averaging for all dye molecules the median of all credible volume sizes (at 95%) reduces to less than 0.5 nm3. However, these dye molecule posteriors are no longer consistent with their smFRET measurements, indicating that the assumptions made lead to false structural information. In contrast, the posteriors determined in the classic model are consistent with the determined smFRET efficiencies.
As the assumption of isotropic and/or dynamic position averaging for all dyes lead to inconsistencies, Fast-NPS enables dye molecule priors in which each dye can be assigned one of the five models. Each model uses the same accessible volume. The algorithm for the calculation of the dye AVs makes several assumptions. At first, the fluorophore's spatial shape is approximated by a sphere. Thus, a diameter taking into account the fluorophore's width, height and thickness should be used (Section 12). Further, the linker's shape is approximated by a flexible rod. The values presented in Section 12 were computed for the dye Alexa 647 attached via a 12-C linker. To date, it is not possible to accurately determine a priori which model is most suited, given an experimental geometry, and thus all models should be tested. In general, one will choose the model which gives the smallest possible posterior size, while still being consistent with the data. To test whether a choice of models is consistent with the smFRET data, we calculate both the posterior and the likelihood. Consistency means that more than 90% of the samples collected from the posterior are within the 95% confidence interval of the likelihood.
While it is true that the lower the anisotropy, the smaller the distance uncertainty, in a smFRET network geometric arrangements of the dye molecules also have to be taken into account. Thus, while representing dye molecules with a low fluorescence anisotropy with an iso model is a typical first choice, the consistency test provides a more direct means for selecting the correct dye model. The optimal choice of dye models can lead to a drastic increase in localization precision and at the same time retain the network's consistency with its FRET data.
To summarize, Fast-NPS allows to gain structural and dynamic information of large macromolecular complexes. In contrast to common structural methods such as x-ray crystallography or cryo electron microscopy this allows for monitoring highly flexible or transient complexes, thus greatly widening our mechanistic understanding of complex biological processes.
The authors have nothing to disclose.
The authors thank B. Gruchmann for the mechanical drawings of the flow chamber. Further, we want to express our gratitude to Max Beckers and Florian Drechsler for insightful comments and discussions regarding NPS and the underlying sampling engine.
Flowchamber preparation | |||
Customized metall sample holder | self-built | n/a | |
quartz-glass slides, 76 x 26 mm | Technical Glass Products | 26007 | |
coverslips, 60 x 24 mm | Marienfeld | 101242 | |
detergent, Hellmanex II | Hellma | 320.001 | |
ultra-pure water from Synergy UV | Millipore | 2512600 | |
Zepto plasma cleaner | Diener | n/a | |
(3-aminopropyl)-triethoxysilane, p.a. | Sigma-Aldrich | A3648 | |
methoxy PEG-succinimidyl valerate, 5 kDa | Laysan Bio Inc. | MPEG-SVA-5000-1g | |
biotinylated PEG-succinimidyl valerate, 5 kDa | Laysan Bio Inc. | BIOTIN-PEG-SVA-5000 | |
Sodium biocarbonate | Sigma-Aldrich | S5761 Sigma | |
Sodium carbonate | Sigma-Aldrich | S2127 Sigma-Aldrich | |
sealing film (Nescofilm) | Fisher Scientific | 12981805 | |
Tygon Flexible Silicone Tubing, 0.8 mm ID, 2.4 mm OD | Saint-Gobain Performance Plastics | 720958 | |
Fine-Bore Polyethylene Tubing, 0.58 mm ID, 0.96 mm OD (Smiths Medical) | Fisher Scientific | 12665497 | |
Neutravidin | Life Technologies | A2666 | |
Name | Company | Catalog Number | Comments |
Total internal reflection fluorescence microscope | |||
Nd:YAG Laser, 532 nm | Newport Spectra-Physics | EXLSR-532-100-CDRH | |
diode-pumped solid-state laser, 491 nm, Calypso | Cobolt | 904010050 | |
diode laser 643 nm, iBeam smart | Toptica | iBEAM-SMART-640-S | |
dichroic mirror, 532 RDC | Chroma | F33-540 | |
dichroic mirror, 476 RDC | Chroma | F33-476z | |
acousto-optic tunable filter | AA Opto-Electronic | AOTFnC-VIS | |
plano-convex cylindrical lens, f = 75 mm | Thorlabs | LJ1703L1-A | |
plano-concave cylindrical, f = -300 mm | Thorlabs | ||
prism, PS 991 | Thorlabs | PS991 | |
focussing lens, f = 75 mm | Thorlabs | LA1608-B | |
syringe pump, PHD 2000 | Harvard Apparatus | 70-2002 | |
2 stepper motors, Z812B | Thorlabs | Z812B | |
piezoelectric actuator, PE4 | Thorlabs | PE4 | |
IR diode laser | Edmund Optics | CPS808 | part of the autofocus system |
dichroic mirror, 775 DCXR | Chroma | 775 DCXR | |
position-sensing detector (PSD), PDP90A | Thorlabs | PDP90A | part of the autofocus system |
water-immersion objective, Plan Apo 60X WI, NA 1.2 | Nikon | MRD07601 | |
dichroic mirror, 645 DCXR | Chroma | 645 DCXR | part of the emission pathway |
emission filter, 3RD550-510 | Omega Optical | 3RD550-510 | green channel in the emission pathway |
emission filter, 3RD660-760 | Omega Optical | 3RD660-760 | red channel in the emission pathway |
EMCCD camera, iXon+ DU897EBV | Andor | AND-20-00032 | |
EMCCD camera, iXon3 DU897D-BV | Andor | AND-20-000141 | |
Name | Company | Catalog Number | Comments |
Miscellaneous | |||
Varian 50 | Cary | UV-VIS spectrometer | |
Fluorolog2 | SPEX | fluorescence spectrometer | |
Solis (V4.15) | Andor | control software for the EM-CCD camera | |
Apt user utility (V1.022) | Thorlabs | control software for the piezo-motors | |
Norland Optical Adhesive 68 | Thorlabs | adhesive | |
PC-AFN-0.8 Nile red | Kisker Biotech | avidin-coated fluorescent multispec beads | |
Matlab | Mathworks | technical computing language for custon written software | |
Origin (V9.0) | Originlab | scientific graphing and data analysis software | |
Hellma 105-202-15-40 | Hellma | 105-202-15-40 | absorption cuvette of 1 cm path length |
Hellma 105-251-15-40 | Hellma | 105-251-15-40 | fluorescence cuvette with 3 mm path length |