Spatially Resolved Total Internal Reflection Fluorescence Correlation

correlation microscopy (TIR-FCM) system is constructed with an electron ... multiplying charge-coupled device (EMCCD) camera in a total internal refle...
1 downloads 0 Views 469KB Size
Anal. Chem. 2007, 79, 4463-4470

Spatially Resolved Total Internal Reflection Fluorescence Correlation Microscopy Using an Electron Multiplying Charge-Coupled Device Camera Balakrishnan Kannan,† Lin Guo,† Thankiah Sudhaharan,§ Sohail Ahmed,§ Ichiro Maruyama,| and Thorsten Wohland*,†

Department of Chemistry, 3, Science Drive 3, National University of Singapore, Singapore 117543, Centre for Molecular Medicine, 20 Biopolis Way, #07-01 Centros, Singapore 138668, and Molecular Neuroscience Unit, Okinawa Institute of Science and Technology, Suzaki 12-2, Uruma, Okinawa 904-2234, Japan

A spatially resolved total internal reflection fluorescence correlation microscopy (TIR-FCM) system is constructed with an electron multiplying charge-coupled device (EMCCD) camera. The system was used to determine diffusion coefficients of lipid molecules in a planar lipid bilayer, and lipids and epidermal growth factor receptor (EGFR) proteins on cell membranes of Chinese Hamster Ovary (CHO) cells. The evaluation of the “cross talk” between neighboring pixels suggests that a higher degree of multiplexing can be achieved than was previously proposed [Kannan, B. et al. Anal. Chem. 2006, 78, 344451] using the same camera with a focused laser excitation. The best time resolution possible with this system is 4 ms for a region of interest comprising 20 lines in the CCD and is good enough to determine membrane diffusion in lipid bilayers and of membrane proteins in living cells. In this work, using a TIR-FCM setup, 1600 autocorrelation functions were measured simultaneously with a time resolution of 4.8 ms. This area corresponds to a 40 × 40 pixel region of interest with a dimension of 11.3 × 11.3 µm2 and is sufficiently large to allow the measurement of the lower membrane of a whole cell simultaneously.

multiplexing and could measure only 1-4 spots simultaneously.7-11 On the other hand, image correlation spectroscopy12-14 (ICS) which employs spatial correlation of fluorescence signals gives access to the spatial distribution of particles. In contrast to FCS, ICS has a very limited time resolution which is restricted by the maximum frame rate of the imaging device. This problem was partly solved recently when the first ICS technique, called k-space ICS,15 could reach a time resolution of 50 ms using an electron multiplying charge-coupled device (EMCCD) camera in a total internal reflection fluorescence (TIRF) setup. A compromise between the two techniques was found with raster image correlation spectroscopy16 (RICS) and spatiotemporal image correlation spectrsocopy17 (STICS) where the spatial and temporal information within a confocal picture was used simultaneously to calculate spatial and temporal correlations. Because of the process of sequential laser scanning, however, the temporal information was not uniform. Recently, an EMCCD camera was used18-19 to measure FCS with a focused laser, and it was shown that multiplexing of FCS measurements is possible and allowed a time resolution of 4 ms for 20 × 20 pixel areas18 and 20 µs for line measurements.19 In another development a spinning disk confocal microscope in combination with an EMCCD camera was used to

Fluorescence correlation spectroscopy1 (FCS) is a powerful spectroscopic tool with single molecule sensitivity.2 This temporal correlation technique has been used to measure the kinetic parameters of molecular systems3-4 including biological cells.5-6 However, FCS despite its good time resolution is limited in

(7) Hwang, L. C.; Leutenegger, M.; Go¨sch, M.; Lasser, T.; Rigler, P.; Meier, W.; Wohland, T. Opt. Lett. 2006, 31, 3010-3012. (8) Heinze, K. G.; Jahnz, M.; Schwille, P. Biophys. J. 2004, 86, 506-516. (9) Burkhardt, M.; Heinze, K. G.; Schwille, P. Opt. Lett. 2005, 30, 2266-2268. (10) Go ¨sch, M.; Serov, A.; Anhut, T.; Lasser, T.; Rochas, A.; Besse, P. A.; Popovic, R. S.; Blom, H.; Rigler, R. J. Biomed. Opt. 2004, 9, 913-921. (11) Go¨sch, M.; Blom, H.; Anderegg, S.; Korn, K.; Thyberg, P.; Wells, M.; Lasser, T.; Rigler, R.; Hard, A. M. S. J. Biomed. Opt. 2005, 10, 054008. (12) Peterson, N. O.; Hoddelius, P. L.; Wiseman, P. W.; Seger, O.; Magnusson, K. Biophys. J. 1993, 65, 1135-1146. (13) Wiseman, P. W.; Peterson, N. O. Biophys. J. 1999, 76, 963-977. (14) Wiseman, P. W.; Brown, C. M.; Webb, D. J.; Hebert, B.; Johnson, N. L.; Squier, J. A.; Ellisman, M. H.; Horwitz, A. F. J. Cell Sci. 2004, 117, 55215534. (15) Kolin, D. L.; Ronis, D.; Wiseman, P. W. Biophys. J. 2006, 91, 3061-3075. (16) Digman, M. A.; Sengupta, P.; Wiseman, P. W.; Brown, C. M.; Horwitz, A. R.; Gratton, E. Biophys. J. 2005, 88, L33-L36. (17) Hebert, B.; Costantino, S.; Wiseman, P. W. Biophys. J. 2005, 88, 36013614. (18) Kannan, B.; Har, J. Y.; Liu, P.; Maruyama, I.; Ding, J. L.; Wohland, T. Anal. Chem. 2006, 78, 3444-3451. (19) Burkhardt, M.; Schwille, P. Opt. Exp. 2006, 14, 5013-5020.

* Corresponding author. E-mail: [email protected]. Fax: +65-6779 1691. † National University of Singapore. § Centre for Molecular Medicine. | Okinawa Institute of Science and Technology. (1) Magde, D.; Webb, W. W.; Elson, E. L. Biopolymers 1972, 29, 705- 709. (2) Rigler, R.; Mets, U ¨ .; Widengren, J.; Kask, P. Eur. Biophys. J. 1993, 22, 169175. (3) Krichevsky, O.; Bonnet, G. Rep. Prog. Phys. 2002, 65, 251-297. (4) Elson, E. L. J. Biomed. Opt. 2004, 9, 857-864. (5) Rigler, R.; Pramanik, A.; Jonasson, P.; Kratz, G.; Jansson, O. T.; Nygren, P.; Stahl, S.; Ekberg, K.; Johansson, B.; Uhlen, S.; Uhlen, M.; Jornvall, H.; Wahren, J. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 13318-13323. (6) Brock, R.; Vamosi, G.; Vereb, G.; Jovin, T. M. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 10123-10128. 10.1021/ac0624546 CCC: $37.00 Published on Web 05/10/2007

© 2007 American Chemical Society

Analytical Chemistry, Vol. 79, No. 12, June 15, 2007 4463

perform spatially resolved FCS measurements.20 For the study of diffusion along a planar lipid bilayer21 or diffusion of transmembrane proteins22 using FCS, it is advantageous to measure in total internal reflection (TIR) configuration.23 With a penetration depth of a few hundred nanometers above the glass-water interface provided by the evanescent wave,24 TIR fluorescence detection has a good inherent axial resolution and is suitable for surface/ membrane studies. Because of the exponential decay and therefore limited penetration depth of the excitation field, the background signal from the bulk is strongly reduced and consequently cross-talk between the pixels is limited. Up to now in TIR-FCS,24-28 a point detector like an avalanche photodiode or a photomultiplier tube was used so that FCS was measured only at one point on the sample at a time. In the present work, the FCS detection with an EMCCD camera has been extended to TIR configuration as a follow-up to our previous work.18 This system enables multiplexing of FCS measurements in TIR configuration combined with imaging which finds immediate applications in the study of lipid bilayers and transmembrane proteins. Measurements are performed on lipid molecules diffusing on a lipid bilayer and diffusion of lipids and epidermal growth factor receptor (EGFR) proteins in Chinese Hamster Ovary (CHO) cell membranes. The system allows measurement of the whole chip of more than 260 000 points (512 × 512 pixels) at a resolution of 35 ms and for subregions of 1600 points (40 × 40 pixels) with a time resolution of 4.8 ms, sufficient for the observation of membrane dynamics. Evaluation of the correlation between pixels indicates that cross talk is reduced compared to a confocal system, and pixels with a distance of 1.42 µm (5 pixels) do not show any appreciable cross talk. Differences of the spatial diffusion coefficient distribution dependent on the probe and membrane system are shown, and distinctions of defective locations on a lipid bilayer and nondiffusive regions containing aggregates on a cell membrane can be made using this system. EXPERIMENTAL SECTION The TIR-FCM system is built around an inverted microscope (Axiovert 200M, Carl Zeiss, Singapore) as shown in Figure 1. Laser light from a diode pumped solid-state laser (Dual Calypso, Cobolt, PhotoniTech Pte Ltd, Singapore) with a dual wavelength output (491 and 532 nm) passed through a laser-line selection filter F1 (XL08, Omega, Brattleboro, VT) to select the 532 nm wavelength. The selected green laser light was beam-expanded three times and collimated by a pair of lenses, L1 and L2. For alignment purposes, the parallel beam of light is then reflected at 90° twice consecutively by two mirrors, TM1 and TM2 (fixed on (20) Sisan, D. R.; Arevalo, R.; Graves, C.; McAllister, R, Urbach, J. S. Biophys. J. 2006, 91, 4241-4252. (21) Burns, A. R.; Frankel, D. J.; Buranda, T. Biophys. J. 2005, 89, 1081-1093. (22) Pick, H.; Preuss, A. K.; Mayer, M.; Wohland, T.; Hovius, R.; Vogel, H. Biochemistry 2003, 42, 877-884. (23) Axelrod, D. Traffic 2001, 2, 764-774. (24) Thompson, N. L.; Burghardt, T.; Axelrod, D. Biophys. J. 1981, 33, 435454. (25) Thompson, N. L.; Axelrod, D. Biophys. J. 1983, 43, 103-114. (26) Hansen, R.; Harris, J. L. Anal. Chem. 1998, 70, 2565-2575. (27) Hassler, K.; Leutnegger, M.; Rigler, P.; Rao, R.; Rigler, R.; Go¨sch, M.; Lasser, T. Opt. Exp. 2006, 13, 7015-7023. (28) Ohsugi, Y.; Saito, K.; Tamura, M.; Kinjo, M. Biophys. J. 2006, 91, 34563464.

4464 Analytical Chemistry, Vol. 79, No. 12, June 15, 2007

Figure 1. Schematic diagram of total internal reflection-fluorescence correlation microscopy (TIR-FCM) setup. F1, Laser line selection filter; F2, emission filter. L1, L2, lenses for beam-expansion; L3, lens in the illumination path; L4, tube lens; TM1, TM2, mirrors mounted on the tilting-mount; DM, dichroic mirror; OBJ, objective; BFP, back focal plane; oil, immersion medium; CS, coverslide; S, sample; EMCCD, electron multiplying charge coupled device camera; MM, MetaMorph software.

tilting mounts), which are kept at 45° to the beam. The reflected light from the second mirror TM2 is coupled into the illumination port of the microscope. Lens L3 in the microscope focuses the beam after reflection by a dichroic mirror DM (560DRLP, Omega) into the back focal plane BFP of an oil immersion objective OBJ (60×, NA 1.45, TIRFM, Olympus, Singapore). The mirror TM2 is used to control the angle under which the collimated beam enters the microscope and thus the angle under which the light beam exits from the microscope objective. If the beam enters at 0° to the optical axis of the microscope, we have a wide-field type (WF-type) configuration although with inhomogeneous illumination. By increasing the angle of the laser beam with the optical axis, the angle of incidence at the glass-water interface can be varied continuously to go beyond the critical angle (θc ) 61° for glass-water interface) for total internal reflection. For an objective with NA ) 1.45, the maximum angle achievable between light beam and optical axis24,27 is θmax ) sin-1 (NA/noil) ) 72.5° using an oil of refractive index noil ) 1.52. The evanescent field depth24,27 h (h ) λ/4π (nG2 sin2 θ - nw2)-0.5 with wavelength of light in vacuum λ ) 532 nm, refractive indices of glass nG ) 1.52 and water nw ) 1.33) varies from 75 to 300 nm for a span of 11.5° in incident angle using this objective. The illumination region has a Gaussian profile in the plane of incidence and is larger than the detection area in the present experiments. The intensity distribution within the maximum size (40 × 40) of the ROI over which measurements were performed is uniform within 4.3% (see Supporting Information) in TIR configuration with the field aperture partly closed. Fluorescence light emitted by fluorophores within the ROI was collected by the same objective and passed through the DM and the emission filter F2 (595AF60, Omega) before being collected by the EMCCD camera (Cascade II: 512, Photometrics, Tucson, AZ) mounted on the left port of the microscope. The microscope and the camera were controlled by the Metamorph software (Universal Imaging Corp., Downingtown,

PA). Detailed information on the EMCCD camera viz., exposure and readout times, data acquisition, background subtraction, and the calculation of autocorrelation functions (ACF) and crosscorrelation functions (CCF) are given elsewhere.18 See ref 29 for more information on the operating principle of EMCCD cameras. The camera noise was measured without any illumination (not shown), and the data does not give any correlation. The time resolution of the camera is limited by the frame read time (FRT) while being operated in overlap mode.18 FRT in turn is limited by the parallel (vertical) shift time (1.048 ms) and the serial conversion time (100 ns/pixel with 10 MHz digitizer). Since all the lines in the CCD have to be shifted after every exposure, the parallel (vertical) shift time (1.048 ms) is the same independent of the ROI size. For a ROI consisting of m lines and n columns, the serial conversion is performed for m × 512 pixels even if n < 512. This leads to the same time resolution for all the ROIs with same m but different n (1 e n e 512). The best time resolution is 4 ms for an ROI consisting of 20 lines in the CCD. The dimension of one pixel in the CCD chip is 16 × 16 µm2, corresponding to 284 × 284 nm2 in the sample. Typical laser powers (measured before the objective) used are 12 mW for the lipid bilayers and 6-9 mW for the CHO cells. Bilayer Preparation. Lipid bilayers were formed on cleaned glass cover slides by the method of vesicle fusion. Labeled small unilamellar vesicles (SUVs) of 1-palmitoyl-2-oleoyl-sn-glycero-3phospocholine (POPC) (cat. no.: 850457, Avanti Polar Lipds, Inc. Alabaster, AL) were prepared by first mixing 5.27 mM POPC and 8 µM 1,2-Dipalmitoyl-sn-Glycero-3-Phosphoethanolamine-N-(Lissamine Rhodamine B Sulfonyl) (Rhodamine-labeled DPPE) (cat. no.: 810158, Avanti Polar Lipids, Inc. Alabaster, AL) in 1:1 × 10-4 ratio. After vortexing for 5 min, the mixture was rotary evaporated for 2 h to remove the chloroform. The lipid layer on the inner wall of the flask was resuspended in deionized (DI) water (Arium 611VF, Sartorius, Singapore) to obtain a lipid concentration of 500 µM and vortexed for 5 min. Sonicating the resuspended lipid mixture for 15 min produces SUVs. The cover slides were cleaned by sonicating for 15 min each in (i) 2% Hellmanex, (ii) Piranha solution (7:3 H2SO4:H2O2), and (iii) analytical grade ethanol. After every sonication, the cover slides were rinsed copiously with DI water and finally stored in ethanol. The SUV solution was then dispensed (∼200 µL) on a cleaned dry cover slide and incubated at 60 °C for 3 h for the formation of bilayers. After incubation, the cover slides were left to cool for at least 30 min to reach room temperature. The excess unfused vesicles and the lipid aggregates in the solution above the bilayer were washed by replacing small volumes (∼200 µL) of the solution 10 times with DI water. For the characterization of the fluorescence intensity cross talk between pixles, intensity was measured on a single immobilized fluorescent bead (Fluospheres, Invitrogen, F-8786, radius ) 0.01 µm, λex ) 580 nm, λem ) 605 nm, 100 000× diluted in DI water). Cell Preparation. The N-terminus mRFP-EGFR construct was made by standard PCR method by inserting the monomeric red fluorescent protein (mRFP) gene (a kind gift from Prof. R. Tsien) into a XhoI site located about 75 bp from the N-terminus of the EGFR gene. The forward and reverse primer sequences are as follows respectively: 5′ GCGCGCCTCGAGATGGCCTCCTCCGA(29) Handbook of Biological Confocal Microscopy, 3rd ed.; Pawley, J. B., Ed.; Springer: New York, 2006.

GGACGTC 3′ and 5′ GCGCGCCTCGAGACTCCCACTCCCTCCGGATCCGGCGCCGGTGGAGTGGCG 3′ (XhoI site shown in bold). The resulting PCR product and the EGFR gene (located in the pNUT vector) were digested with XhoI. In order to prevent self-ligation, the digested pNUT-EGFR vector was treated with shrimp alkaline phosphatase (SAP) (Roche) at 37 °C for 1 h. The reaction was stopped by incubating at 65 °C for 15 min. Subsequently, the digested mRFP gene was ligated to the SAPtreated pNUT-EGFR vector. The orientation and sequence of the insert was confirmed by automated DNA sequencing (ABI prism). CHO-K-1 cells were obtained from ATCC (Manassas, VA) and grown in 75 cm2 tissue culture flask up to 90% confluency in the complete growth media, 1× F-12 Nutrient mixture (Kaighn’s modification) media containing 10% fetal bovine serum qualified (FBS) and 1% antibiotics (penicillin and streptomycin). All the tissue culture reagents were obtained from Invitrogen (Singapore). Transient transfection was performed using Fugene 6 (obtained from Roche). For transfection, cells were seeded in a six-well tissue culture plate containing the 30 mm prewashed and sterilized cover glass at 1.5 × 105 cells/well 1 day before transfection. CHO cells were transfected with N-terminus EGFR plasmid by following the standard Fugene 6 protocol supplied by the manufacturer. After 24 h of transfection, cells were washed thrice with 1× PBS and measured in 1× PBS. For the diffusion of octadecyl rhodamine B chloride (R18) on CHO cell membranes un-transfected CHO cells were mixed with 20 nM R18 in PBS and incubated at room temperature for 15 min before the measurements. TIR-FCM Data Analysis. The ACF curves measured on a fluorophore solution in TIR configuration can be fitted to a model which describes the diffusion of molecules in a 3D volume where the intensity distribution is assumed to be uniform in the xy plane and exponentially decaying in the z direction27

G(τ) ) 1 +

(

1 τ 1+ 2 N ωτ

) [(

-1

×

z

1-

) (x ) x ]

τ w i 2τz

τ + 4τz

τ πτz

(1)

with

τz ) h2/4D

(2a)

τxy ) ωxy2/4D

(2b)

and

Here G(τ) is the autocorrelation function, N is the particle number in the probe volume weighted by the correction factor for the experimental geometry, τz, τxy are the diffusion times for the axial and in-plane diffusion, respectively, and D is the diffusion coefficient. ωxy and h are the distance in the radial and axial directions, respectively, at which the molecule detection efficiency decreases to e-2 of its value along the optical axis at the focal plane. For the calculation of D, ωxy ) 142 nm (half of 284 nm which is the dimension of one pixel in the sample plane) is used. The structure factor ω of the observation volume is defined by, Analytical Chemistry, Vol. 79, No. 12, June 15, 2007

4465

Figure 2. Maps of (a) particle number N and (b) in-plane diffusion time τxy measured on a lipid bilayer in TIRF configuration for an ROI of 20 × 20 pixels. One square pixel has a side of length 284 nm in the sample plane. The pixel marked “X” in parts a and b measures lipid aggregate which is seen in the time trace of the intensity (c) and gives an autocorrelation curve (e), reminiscent of a two-particle behavior. The pixel marked “+” in parts a and b measures diffusion of lipid molecules that is reflected in the time trace of intensity (d) and the autocorrelation curve (f). The fit to a two-component model in part e and eq 3 in part f are shown in dashed lines. The lower panels in parts e and f show the fit residuals. The time resolution is 4 ms.

ω ) ωxy/h. Using the definition of ω, τxy ) ω2τz. The function w is defined by w ) exp(-x2)erfc(-ix). The correction factor for the experimental geometry can be estimated by measuring a dye solution with known concentration. In the case of bilayers and cell membranes diffusion occurs in the xy plane and the z-component of τ is zero. ACF curves measured on bilayers and cell membranes were fitted to a 2D model given by

G(τ) ) 1 +

(

)

1 τ 1+ N τxy

-1

(3)

The ACF was evaluated for every pixel within the ROI. For a ROI of 40 × 40 pixels, 1600 ACF curves can be obtained. Every 4466

Analytical Chemistry, Vol. 79, No. 12, June 15, 2007

pixel will detect fluorescence from a defined detection volume. Since the detection volumina of neighboring pixels overlap, the number of independent measurements will differ from the total number of 1600 pixels, as will be discussed below. Fitting of correlation curves were carried out using self-written programs in IGOR Pro (version 4.09, Wavemetrics, Lake Oswego, OR). RESULTS AND DISCUSSION In our earlier work the EMCCD-based confocal FCS system was calibrated18 and was shown to be able to measure up to a maximum of 300 points at a time resolution of 35 ms. In the present work we extend the EMCCD system to include a TIRF configuration to achieve an even higher parallelization of FCS measurements. Diffusion in 2D (lipid molecules in a lipid bilayer,

Figure 3. Evaluation of the cross talk in (a) TIR configuration and in (b) confocal FCS using the same camera. The normalized crosscorrelation amplitude Gx(0) (continuous line), obtained from the bilayer data by fitting the CCF curves between two pixels to eq 3 and the fluorescence intensity measured with a single fluorescent bead (dashed line) are shown as a function of the distance from the center pixel in part a. Both Gx(0) and the intensity decrease with increasing distance from the center pixel and reach a negligible value at a distance of 4 pixels. The normalized autocorrelation amplitude G(0) (continuous line) and the fluorescence intensity (dashed line) measured in confocal FCS are shown in part b. Although G(0) decreases faster, the intensity vanishes only at a distance of 30 pixels.

fluorescent dye on CHO cell membrane, and membrane proteins in CHO cells) is measured to demonstrate the feasibility of TIRFCM. (See Supporting Information for calibration data of the TIRFCM setup). Lipid Bilayer. The diffusion of rhodamine-labeled POPC lipid molecules along the POPC bilayer on a cover slide within a 20 × 20 ROI was measured. Figure 2 shows the map of (a) N and (b) τxy for the 400 pixels measured in a 20 × 20 ROI. Some of the pixels (8%) gave ACF curves which showed a two-component behavior, and the data could not be fitted to eq 3. This happens when aggregates of lipid molecules or unfused vesicles pass through a pixel. In Figure 2 we compare a pixel where aggregates are detected (marked ‘X’ in Figures 2a and 2b) with a pixel in which only lipid membrane diffusion is seen (marked ‘+’ in Figures 2a and 2b). The intensity traces of these pixels are shown in Figures 2c and 2d, and the corresponding ACF curves are shown in Figures 2e and 2f. The ACF curve in Figure 2e was fitted with a two-component model whereas the ACF curve in Figure 2f can be fitted with a one-component model based on eq 3. The fit residuals are shown in the bottom panels. The value of parameters for 30 pixels which showed a two-component behavior were excluded from further evaluation after ascertaining from their intensity traces that aggregates, detected as large intensity peaks, distorted these ACF curves. In Figures 2a and 2b their parameters were set to zero and hence appear dark blue. The average values of N and τxy for 370 curves are 4.6 ( 0.6 and 17.1 ( 10 ms, respectively, similar to measured τxy values (7.3 ( 5.5 ms) by confocal FCS on the same system. The diffusion coefficient along the bilayer Dxy averaged over 370 pixels is 3.9((2.6) × 10-9 cm2 s-1. The reported value of D from fluorescence recovery after photobleaching experiments30 on a POPC bilayer formed on a coverglass is 10((2) × 10-9 cm2 s-1. Cross Talk. For multiplexing of FCS measurements the cross talk of fluorescence signal arising from the overlap of detection volumina of neighboring pixels should be minimal. To characterize the distance at which the cross talk becomes negligible CCF (30) Brozell, A. M.; Muha, M. A.; Sanii, B.; Parikh, A. N. J. Am. Chem. Soc. 2006, 128, 62-63.

Figure 4. Autocorrelation curves G(τ), of (a) the organic dye molecules (R18) diffusing on a CHO cell membrane and (b) the diffusion of mRFP tagged EGFR on a CHO cell membrane. Fit to eq 3 are shown in dashed lines. The fit residuals are shown in the lower panels. A pixel where there is no diffusion, a pixel which is bleaching, and a pixel which is outside the cell boundary (marked “ND”, “B” and “O” in Figures 5e and 5f) are shown in part c as continuous lines. The fit to eq 3 are shown as discontinuous lines. All the autocorrelation curves in parts b and c emerge from the whole cell measurement. The laser power was 9 mW for the R18-CHO cells and 6 mW for the mRFP-EGFR-CHO cells. The time resolution for measurements on R18-CHO cells is 4 ms and for measurements on mRFP-EGFR-CHO cells is 4.8 ms.

curves were calculated between pairs of pixels with different distances using the data on a bilayer. The center pixel in the ROI was cross-correlated with pixels along the horizontal and vertical directions one at a time. Figure 3a shows the normalized crosscorrelation amplitude Gx(0), obtained by fitting the CCF curves to eq 3. At a distance of 4 pixels in either direction, the Gx(0) falls below the noise level (shaded area), i.e. below the size of the random variation introduced by noise on the ACF. The fluorescence intensity cross talk between pixels cannot be characterized using the bilayer data as the bilayer extends over a finite number Analytical Chemistry, Vol. 79, No. 12, June 15, 2007

4467

Figure 5. Results from measurements on a whole cell using an ROI of 40 × 40. (a) The average fluorescence intensity on the bottom membrane of a whole CHO cell with mRFP-tagged EGFR measured in TIRF configuration. The boundary of the cell membrane is visible (white dashed line) from the intensity distribution. (b) Map of goodness of the fit (χ2) to eq 3 of autocorrelation curves G(τ) obtained for all the 1600 pixels in the ROI. The pixels which are outside the cell membrane have a higher value of χ2 due to the poor signal-to-noise ratio. The pixels which are on the cell membrane have a good signal-to-noise ratio and hence have a lower value of χ2. The cell boundary (white dashed line) seen in part a fits well into the border separating the high χ2 region from the low χ2 region. Maps of N and τxy are shown in parts c and d. The cell boundary again separates the region with higher N values from the region with lower N values in part c. In part d the value of τxy range from 0 to 0.5 s for better contrast. Maps of N and τxy selected on the basis of χ2 (e3 × 10-5) are shown in parts e and f. While constructing parts e and f, an upper limit was set for N (e100) and τxy (e0.5 s) because a few pixels on the cell membrane were having relatively higher values for N and τxy compared to the rest of the pixels. The pixels which exceed this limit are shown in dark blue. ACF curves of these pixels have low amplitude and give rise to large values of N and τxy. This can be due to exclusion of labeled receptor molecules from these regions or to low mobility of receptor molecules in these membrane patches. Photobleaching also gives rise to high amplitude ACF curves with a long diffusion time. ND: Pixel with no diffusion, B: pixel showing bleaching, and O: pixel which lies outside the cell. Each square pixel has a side of length 284 nm in the sample plane. The time resolution is 4.8 ms.

of pixels. In this case the intensity measured by every pixel is a convolution of the intensity emanating from its detection volume 4468 Analytical Chemistry, Vol. 79, No. 12, June 15, 2007

and the intensity arising from the neighboring volumina in the sample plane. Because of this reason, intensity was measured on

a rectangular ROI of size 1 × 512 which contained a single immobilized fluorescent bead of 20 nm diameter (see Experimental Section) at the center of the ROI. The measured intensity as a function of the distance from the center pixel is shown in Figure 3a. It is clear that the intensity measured on a single fluorescent bead also decreases to a negligible value at a distance of 3-4 pixels. Taking every fifth pixel as a detection element it is possible, in principle, to have a 102 × 102 array of independent detection elements and hence 10 404 detection volumes with this EMCCD camera. The spatial resolution for FCS in this case is 1.42 µm, and the time resolution varies between 4 and 35 ms when recording between 400 and 10404 independent points, i.e., 20 × 512 to 512 × 512 pixels. This is a 36-fold increase in multiplexing in comparison to the EMCCD based confocal FCS18 where the multiplexing was proposed for every 30 pixels (10 pixels, each 3 × 3 binned). In the case of confocal FCS, G(0) decreases to a negligible value at a distance of 15 pixels (5 binned (3 × 3) pixels) as shown in Figure 3b, but the decay of the fluorescence intensity (Figure 3b) was slower than the decay of ACF amplitude G(0) with pixel distance.18 The intensity cross talk gives rise to an uncorrelated background and decreases the ACF amplitude. This in turn affects the evaluation of the particle number N. So for the confocal FCS multiplexing was proposed only for every 30 pixels. Another difference between TIR-FCM and confocal FCS using EMCCD is that in TIR-FCM the illuminated region is obtained from a single laser beam whereas in confocal FCS the single laser beam has to be split into the required number of spots to multiplex the focal volume. Although every fifth pixel can be an independent detection element in TIR-FCM, it still contains the cross talk from the neighboring excitation volumina. This is due to the fact that TIR illumination produces uniform excitation intensity over a region. For high throughput screening this cross talk can be avoided when separating samples by at least 5 pixels on a chip. For cell measurements a solution to this problem would be the development of a deconvolution algorithm to allow the full evaluation of all pixels with a resolution of 284 nm. Measurements on CHO Cells. Figure 4a shows the ACF curve of a pixel measuring diffusion of R18 on the bottom membrane of a CHO cell in TIRF configuration. Fitting of the ACF curve to eq 3 yielded N ) 24.1 ( 0.9 and τxy)14.8 ( 1.2 ms. The average value of these parameters for 400 ACF curves within the ROI of 20 × 20 pixels is N ) 26.4 ( 5.1, τxy)25.8 ( 9.5 ms. The diffusion coefficient is 2.15((0.85)×10-9 cm2 s-1. Diffusion of mRFP tagged EGFR on a whole CHO cell was measured on the bottom membrane in TIRF configuration using a 40 × 40 ROI (1600 pixels). In Figure 5, a comparison is made between different fit parameters and the goodness of the fit to eq 3 with reference to the intensity distribution measured in TIRF configuration. As the ROI includes pixels which are outside the cell membrane a selection procedure is developed to choose only those pixels which measure the cell membrane. A glance at Figures 5a to 5c reveals that the cell boundary (white dotted line) which is seen in the intensity distribution (Figure 5a) manifests itself in the maps of the goodness of the fit (χ2) (Figure 5b) and N (Figure 5c). Outside the cell boundary, the pixels have higher values of χ2 and N compared to the pixels inside the cell boundary. For the pixels outside the cell boundary, fluorescence

Figure 6. Histogram of (a) particle number N and (b) diffusion time τxy measured on lipid bilayer (black lines), R18 on CHO cell membrane (filled pattern), and mRFP-tagged EGFR in CHO cell membrane (gray lines).

signal-to-noise ratio is poor and the ACF curves have low or vanishing amplitudes. They yield high values for χ2 and N when fitted to eq 3. Whereas the pixels on the cell membrane have a good signal-to-noise ratio, they yield low values for χ2 and N. Going by the above reasoning, it is possible to select the pixels within the ROI by setting an upper limit to the χ2 value. Maps of N and τxy both selected on the basis of χ2 (e3 × 10-5) for the pixels within the ROI are shown in Figures 5e and 5f. Including all the pixels which passed the selection procedure based on χ2, we have 561 pixels on the cell membrane (Figure 5f). The average diffusion coefficient D for these 561 pixels is 5.48 ((5.1) × 10-10 cm2 s-1. This shows that the diffusion of mRFP-EGFR is slower than the diffusion of lipid probe (R18) on the CHO cell membrane. The large standard deviation in D measured on mRFP-EGFR is consistent with the assumption of membrane heterogeneity. A typical ACF curve of a pixel which is on the cell membrane showing diffusion of mRFP-EGFR is shown along with the fit residuals in Figure 4b. In Figure 4c (i) a pixel (marked “ND” in Figures 5e and 5f) where no correlation is measured, (ii) a pixel (marked “B” in Figures 5e and 5f) where bleaching is seen, and (iii) a pixel (marked “O” in Figures 5e and 5f) which is outside the cell boundary are shown. Figure 6 shows the evolution of heterogeneity when going from a lipid bilayer to a cell membrane. The histograms of (a) N and (b) τxy for the lipid molecules diffusing in a bilayer (black) have a narrow distribution as they diffuse in a homogeneous environment. The distribution widens (filled pattern) for the lipid probe (R18) diffusing along the CHO cell membrane due to the membrane heterogeneity. mRFP tagged EGFR diffusion has the widest distribution (gray) due to the membrane heterogeneity and the receptor distribution along the membrane including possible interactions with extra-membranous structures, e.g., the cytoskAnalytical Chemistry, Vol. 79, No. 12, June 15, 2007

4469

eleton. From the foregoing results, it is clear that measurement of diffusion on a whole cell membrane coupled with imaging, both enabled by TIR-FCM using an EMCCD camera, can be achieved and allows the distinction of regions of different protein concentrations and mobilities on the cell membrane. CONCLUSION In the present work, we present total internal reflection fluorescence correlation microscopy (TIR-FCM) based on an electron multiplying charged-coupled device (EMCCD) camera as an imaging and spectroscopic detector. The system is shown to be able to measure diffusion on bilayers and can give a full FCS image of a cell. The time resolution of the system is 4 ms when measuring 20 lines or 4.8 ms when using 40 lines. A typical cell of 10 × 10 µm2 extent can be measured with 40 × 40 pixels ROI with a time resolution of 4.8 ms resulting in 1600 simultaneous FCS measurements. For a full chip measurement of 512 × 512 pixels, the time resolution drops to 35 ms. To avoid cross talk in high-throughput measurements, a distance between measurement points of at least 5 pixels should be used, yielding a maximum of

4470

Analytical Chemistry, Vol. 79, No. 12, June 15, 2007

about 10 000 points which can be measured in parallel. Despite the limited time resolution, the system allows measurements of transmembrane protein mobility on a whole cell simultaneously. With future advances in camera technology and the possible development of deconvolution algorithms for TIR-FCM, this system is a promising tool for high-throughput as well as quantitative cell biology applications. ACKNOWLEDGMENT The authors thank Yimian Hong for the assistance in the preparation of lipid vesicles and bilayers. SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review December 29, 2006. Accepted April 2, 2007. AC0624546