Backgrounds and leakages between channels were corrected during analysis as described elsewhere (3,5,6,8). Traces in Figure 2, C–E, were corrected for background and leakage but not for γ. We determined γ factors from anti-correlated donor and acceptor photobleaching events using γ = ΔIA/ΔID, where (ΔIA = IAbefore bleach – IAafter bleach) and (ΔID = IDafter bleach – IDbefore bleach) as described elsewhere (1,23,24). γ was determined from the data for Figure 2F to be 2.16 for the Quadview. We confirmed that the same sample when measured with our two-color smFRET microscope yielded consistent FRET efficiency results (0.31) when using γ specific for that instrument (0.94). Emission leakage between spectral channels was measured using single-labeled DNA samples. Transmission efficiencies of the blue, green, red, red with extra 640 × 100 filter, and infrared channels in our Quadview system were 60%, 49%, 66%, 50%, and 65% of their value, respectively, without the Quadview dichroics and filters in place (bypass mode). Forty percent of Alexa 488 signal level leaked into the green channel, 9% leaked into the red channel, and none leaked into the infrared channel. No TAMRA signal was detected in the blue or infrared channels, 15% of the TAMRA signal leaked into the red channel. No Alexa 555 emission was detected in the blue or infrared channels, but 9% leaked into the red channel. There was no detectable emission from Alexa 647 in the blue, green, and infrared channels or from Cy7 in the blue, green, and red channels.Observation protocols
Because blue dye photobleaching was limiting, we sometimes used a time-lapsed, shuttered blue illumination. This illumination scheme simplified interpretation of FRET between Alexa 555 (or TAMRA) and Alexa 647, because we could calculate FRET when the blue laser was not active. Undesired direct excitation of TAMRA and Alexa 555 by blue light can complicate FRET measurements (3,5,6,8).
Figure 4 illustrates several excitation patterns. Sequential excitation is used first to excite Alexa 647 (635 nm) followed by the excitation of Alexa 488 (473 nm) to colocalize the acceptor and the auxiliary molecule (Figure 4B). Next, the donor Alexa 555 is excited (532 nm) to observe FRET between the SNARE domains of Syntaxin 1A as reported by the high Alexa 647 emission. In Figure 4C, a laser sequence that allows for semicontinuous monitoring of Alexa 488–labeled Munc-18 is used. The 635-nm laser is used to identify the acceptor Alexa 647 dye and then is shuttered off at frame 5. The donor Alexa 555 dye is excited (532 nm) at frame 10 and FRET to the acceptor is observed. The 473-nm laser is shuttered on for one frame at every 15th frame. The flashing laser pattern allows periodic monitoring of the Munc-18 colocalization while both prolonging the life of the blue dye and leaving frames for FRET calculation that are unaffected by the blue laser. The frames with 473-nm illumination are omitted from display in the red and green signals. Note the bleaching of Alexa 488 after 8 s.
Results and discussion
The sequential and simultaneous FRET and colocalization experiments were performed with a prism-type TIRM, a Quadview splitter containing three dichroic mirrors, and an emCCD, allowing the simultaneous excitation and detection of the fluorescence from four dyes with emission spectra characteristic of Cy2, Cy3, Cy5, and Cy7 (Figure 1 and Materials and methods). We first demonstrated this instrument by examining DNA bending induced by yMutSα. MutS homologs are responsible for recognizing and signaling repair of base-base mismatches and insertion-deletions in newly replicated DNA (25). MutS-mismatch complexes adopt multiple conformations with different degrees of DNA bending (from unbent to bend angles of ~90°C) (20,26,27). It has been suggested that each bending state may have a different repair signaling potential (20,25,27).
Biotinylated DNA (50 bp) labeled with a FRET donor (TAMRA) and acceptor (Cy5) separated by 19 bp with a CC mismatch approximately halfway between the two fluorophores was tethered to a streptavidin-coated quartz surface (Figure 2, A and B). Because yMutSα contains >30 cysteines, site-specific labeling using maleimide dye methodology is not feasible. Consequently, 6His-tagged yMutSα (tagged on the N-terminal of Msh2) was noncovalently labeled using Tris-NTA-OG (OG-yMutSα) (14). OG-yMutSα complexes were prepared before addition to CC-mismatch-DNA coated surfaces (see “Materials and methods” section). DNA bending was monitored by continuously exciting TAMRA with 532-nm illumination and measuring FRET to Cy5. The presence of OG-yMutSα was monitored by measuring OG emission under either continuous (data not shown) or pulsed (Figure 2, C–E) 473-nm illumination. Despite using an oxygen scavenging system, a shutter-pulsed scheme was required to allow longer observation intervals before OG photobleached.