2.1.

Spectral Analysis

Specific spectral analysis of protein-fluorophore conjugates used in our experiments was performed. First, 500 μl of 1 μM of pIgA-R ligands ([Fab]₂ pseudoligands against pIgA-R) conjugated to Alexa488 (Molecular Probes, Eugene, OR) or Cy3 (Amersham Life Science, Pittsburgh, PA) were prepared in PBS, as described previously.¹³ Luminescence intensity measurements to obtain the individual spectra were carried out using a Spex Fluorolog 2+2 spectrofluorimeter. All data were collected with right angle detection at room temperature (22±2°C) in air-saturated solutions. Excitation and emission intensity data were evaluated at the maximum.

2.2.

Growing MDCK Cells on Filter Insert

MDCK cells stably transfected with pIgA-R were grown to confluence in 100-mm cell culture dishes. After 4 days, the cells were trypsinized, centrifuged, and resuspended in DMEM/10 FBS/Pen-Strep.¹³ Then, 120 μl of the cell suspension was placed on top of an inverted Transwell Clear insert (Corning Costar, Cambridge, MA), i.e., on the outside of the membrane.¹⁴ After 4 h, the inserts are placed the normal way into multiwell dishes containing DMEM/10 FBS/Pen-Strep. After 3 days in culture these cells are fully polarized and are used immediately according to specific internalization protocols.¹³

2.3.

Internalizing Fluorophore-Labeled Ligands

The inserts with polarized MDCK cells are washed with PBS and equilibrated with DMEM/HEPES/BSA at 17 °C. Next, 160 μg/ml Alexa488- or Cy3-labeled pIgA-R ligands were co-internalized from apical and basolateral PM, respectively. Cells are incubated at 17 °C for 4 h to allow internalization of pIgA-R-ligand complexes into the subapical region by basolateral-to-apical transcytosis and apical endocytosis. At 17 °C, delivery to the apical PM is blocked and receptor-ligand complexes, co-internalized from opposite PM surfaces, localize predominantly in the apical endocytic compartment, which is located ∼1 to 5 μm below the apical PM.¹³ Our images were collected 3.5 μm below the apical PM (Fig. 2; FRET). Then, the cells are washed with PBS to remove unbound ligands and immediately fixed with 4 paraformaldehyde/PBS.

Three different samples were used during our experiments: the double-labeled specimen, containing apically internalized Alexa488-pIgA-R-ligand complexes and basolaterally internalized Cy3-pIgA-R-ligand complexes, plus corresponding single-label reference samples containing either Alexa488 or Cy3, which were used to establish the contamination levels.

2.4.

1-P and 2-P FRET Microscopy

For the 1-P data collection, we used a Nikon PCM 2000 laser scanning confocal microscope [Fig. 3(a)], equipped with a 60× water immersion lens (1.2 NA), argon (488 nm) and green HeNe (543 nm) lasers, and emission filters (515/50 nm and 590 nm LP for our standard experiments), respectively. For the experiments comparing different emission filters, we have used 580/20, 580/30, 580/40, 580/50, and 610/75, provided by Chroma Corp. (Brattleboro, VT), and our standard 590 LP. Simple PCI software (Compix, Cranberry Township, PA) was used to drive the hardware, image acquisition, and processing [Fig. 3(a)]. Bleaching is undetectable when the argon laser is only used for one scan to collect the final image. PCM is set to collect data simultaneously in both channels at a 1024 ×1024 pixel image.

Figure 3

1-P and 2-P microscopy systems. Schematic illustration of the (a) laser scanning confocal 1-P and (b) multi-photon 2-P excitation fluorescence microscopy systems integration.

For the 2P-FRET experiments we used a VERDI pumped turntable mounted titanium:sapphire laser, coupled to the laser port of a Biorad MRC600 (Biorad, Hercules, CA). The MRC600 scan head is coupled to a Nikon TE300 (Eclipse) epi-fluorescent microscope (Nikon, Melville, NY), linked to a personal computer and monitor [Fig. 3(b)]. For data collection, the same specimens were used as in the laser scanning confocal experiment. The objective was a 60× water immersion lens (1.2 NA), with the laser excitation wavelength at 790 nm and a 4.0 power level. We used this laser excitation wavelength to excite specifically the donor molecules. We used a 535/50-nm donor emission filter and our standard 610/60-nm acceptor filter, plus, for comparative measurements, the above-mentioned 580/20-30 series. Fluorescence levels are presented simultaneously on a split screen, one representing the donor channel and the other the acceptor channel.

2.5.

Data Collection

For both 1-P and 2-P microscopy, the specimen is positioned in a small chamber created by a coverslip between two metal rings and filled with a small amount of PBS. This whole assembly is placed on the microscope stage. We first select an appropriate area of the specimen, check the cell height (15 to 20 μm), and set a focal plane of 3.5 μm below the apical PM. For 1-P this is done with only the green HeNe laser in operation, and for 2-P at 790 nm; in both cases in an area of the specimen that is not used later for image acquisition. Optimal PMT settings are established in the pre-image-acquisition phase.

2.6.

Post-Acquisition Data Generation and Analysis

3.1.

Spectral Overlap of Alexa488- and Cy3-labeled pIgA-R Ligands

For energy transfer to occur, the donor and acceptor fluorophores need to possess sufficient spectral overlap between their emission and excitation spectra, respectively. In our FRET experiments, Alexa488 acts as the donor and Cy3 as the acceptor (Fig. 1). For this particular fluorophore pair the highest efficiency of energy transfer is at a distance of 67.5 Å, when 50 of excited donors are deactivated by FRET.⁴ Below and above 67.5 Å, FRET occurs at a reduced rate. To assess the extent of spectral overlap between Cy3 and Alexa488-pIgA-R ligands, a correct spectrogram for this fluorophore-protein pair was determined (Fig. 1). This fluorophore pair provides an excellent fit for FRET microscopy showing a strong spectral overlap. Unavoidably, background fluorescence, i.e., acceptor bleed-through and donor crosstalk, also occurs. When donor excitation takes place at 488 nm, some acceptor fluorophores will be excited due to the spectral overlap (acceptor bleed-through), thus contaminating the FRET signal (Fig. 1). Equally, that part of the donor emission spectrum that overlaps with the acceptor emission and is not curtailed by filters (donor cross-talk) will also add to the FRET spillover contamination (Fig. 1).

3.2.

Algorithm-Based Correction Method for FRET

For biological systems where the fluorophore pairs are not always separated by a consistent distance and where FRET occurs over a wide range of fluorescence intensities at a membrane plane, the importance of a sensitive, finely tuned correction system cannot be overstressed. In the absence of reliable correction methods, small changes expressed in differential FRET signals may be misinterpreted, particularly where the FRET signal approaches the fluorescence level of the background contamination. Recently, the PFRET algorithm method was developed to correct the FRET spillover contamination in a pixel-by-pixel manner.¹²

The PFRET correction algorithm for 1-P microscopy requires seven images to correct the double-label 1-P FRET images, as described previously.¹² Those include images of single-label reference donor and acceptor, as well as double-labeled specimens with comparable fluorescence ranges taken under identical imaging conditions (data not shown). Applying the algorithm with different reference specimens produces near-identical results, indicating that different single-label controls do not change the correction level (data not shown).

1-P images showing a two-dimensional Z-section (i.e., in the xy plane) at ∼3.5 μm below the apical PM were collected from double-labeled (Fig. 4, 1-P acceptor, donor and uFRET panels) and single-labeled (data not shown) images using 580-50 bandwidth emission filters. These images were then processed by the PFRET correction algorithm method¹² to generate the PFRET image (Fig. 4, 1-P PFRET panel), which shows the energy transfer levels. The acceptor (Fig. 4, 1-P acceptor panel), the quenched donor (Fig. 4, 1-P donor panel), and the PFRET (Fig. 4, 1-P PFRET panel) images are then used to calculate the three experimental parameters: acceptor, uD, and E values (Figs. 7 and 8) used during data analysis. Comparing the uFRET and PFRET panels (Fig. 4, 1-P panels), it is clearly visible where contamination has been removed by treating the uFRET image with the PFRET algorithm.

Figure 4

1-P and 2-P imaging acceptor, donor, uFRET, and PFRET distributions of pIgA-R-ligand complexes in apical endocytic membranes. Double-labeled MDCK polarized cells, containing co-internalized Alexa488- and Cy3-pIgA-R-ligand complexes from opposite PM surfaces, were imaged by 1-P and 2-P microscopy at an xy (Z-section) focal plane ∼3.5 μm below the apical PM using 580-50 emission filter (wavelength range 555 to 605 nm). These images were modified in Adobe Photoshop at the same rate to a higher level of contrast for better visualization. Images shown (overall size 15×7.5 μm) contain two ROIs of similar size (7.5×7.5 μm), each one containing a complete cell. Panel acceptor: acceptor excitation/acceptor channel shows acceptor fluorescence intensities (only in 1-P). Panel donor: donor excitation/donor channel shows the quenched donor fluorescence intensities. Panel uFRET: donor excitation/acceptor channel represents uFRET, which includes energy transfer levels plus the two contaminants in the FRET signal: donor crosstalk and acceptor bleed-through. Panel PFRET: uFRET image was processed by the PFRET algorithm, which removes donor crosstalk and acceptor bleed-through. The resulting image represents the actual energy transfer levels.

A similar protocol is followed for 2-P microscopy with the exception that it only requires four images, including images of the single-label reference donor as well as double-labeled specimens with comparable fluorescence ranges taken under donor excitation wavelength with identical imaging conditions, since the acceptor Cy3 is not excited by the 790-nm donor excitation wavelength.¹² As a precaution, we took an image of the single-label acceptor at the 790-nm donor excitation wavelength to confirm that there is no spillover into the FRET channel (data not shown). It is important to note that the 2-P uFRET image does not contain acceptor bleed-through contamination and thus only the donor crosstalk contamination is removed by treating the uFRET image with the PFRET algorithm. Consequently, the difference between the intensity of 2-P PFRET and uFRET images is reduced when compared to that of 1-P images.

The images in Fig. 4 contain two representative ROIs of similar size, each corresponding to one cell, which show the typical punctate pattern of apical endocytic membranes located at the level of the apical cytoplasm.

3.3.

FRET Analysis Using Emission Filters with Different Bandwidths and a FRET Spillover Algorithm-Based Correction Method

The establishment of a correct spectrogram is important for the selection of the emission filters, which in turn determines how much background fluorescence is collected, along with the FRET signal. Furthermore, it is critical to strike the right balance between reducing donor crosstalk and acceptor bleed-through FRET contamination and yet capture most of the FRET signal. FRET emission filters with different bandwidths allow different levels of FRET signals to be collected, however, with increasing FRET fluorescence the spillover contamination increases (Fig. 5).

Figure 5

Different emission filters with wider bandwidths lead to increased 1-P and 2-P FRET signals and corresponding spillover contamination. Six different emission filters: 580/20 (range 570 to 590 nm), 580/30 (range 565 to 595 nm), 580/40 (range 560 to 600 nm), 580/50 (range 555 to 605 nm), 610/60 (range 580 to 640 nm) and 610-75 (range 572.5 to 647.5 nm) were used to collect FRET images, which were then subjected to processing using PFRET correction algorithm.

We hypothesized that our algorithm-based correction method described earlier should demonstrate its utility by different levels of correction accompanied by comparable E results. Our results support this assumption where wider bandwidths result in a higher percentage of correction (ANOVA p-value for 1-P=8.4E-18; 2-P=5.5E-14) while producing statistically indistinguishable E values (ANOVA p-value for 1-P=0.19; 2-P=0.37) (Fig. 6). As expected, the average correction level is lower in 2-P than 1-P microscopy, due to the fact that we do not observe any acceptor bleed-through, as the donor wavelength does not excite the Cy3 fluorophore (acceptor).¹² We have no explanation for the higher than expected correction of the 580-20 filter in 2-P microscopy other than to point at the consistent E figures, which are not affected by whatever caused the higher correction.

Figure 6

PFRET correction levels increase with wider emission filter bandwidth, while resulting in comparable E levels. Six different emission filters with wider bandwidth were used to test the PFRET correction analysis. PFRET correction levels ( correction) increase with the use of emission filters with wider bandwidth while E values show comparable results. Correction levels are higher in 1-P than in 2-P since there is no acceptor bleed-through contamination of the FRET signal in 2-P. Nevertheless, comparable E levels are obtained in 1-P and 2-P, thus validating our PFRET correction approach.

With the use of our correction algorithm, the choice of emission filter has now become less critical, both in 1-P and 2-P microscopy. For consistency, we continued to use the 590 LP filter in 1-P microscopy for the other experiments shown below. The 590LP filter catches the second of the two emission peaks of the acceptor Cy3 while at the same time limiting the background contamination. Based on our results, the second peak provides a strong enough signal to detect FRET (Figs. 3 and 4).

3.4.

Negative Dependence of uD:A Ratio and uD Levels on E as an Indicator of Receptor Clustering Using 1-P and 2-P FRET Microscopy

Theories on how to distinguish between a clustered and a random distribution of membrane-bound proteins suggest that energy transfer between donor and acceptor molecules is governed by different dynamics with respect to E, when the distribution is random or clustered.^{18 19 20} E’s independence from acceptor levels has been used as the main indicator for a clustered distribution pattern.^{18 19 20} Our 1-P results indicate that E is independent of the acceptor (data not shown). The phenomenon of decreasing E with an increase in uD:A ratio and uD levels can be used as another indicator of receptor clustering.^{18 19 20} We call this donor geometric exclusion, where increasing the number of donor molecules within a cluster prevents other donors from being in FRET distance of an acceptor causing E to decrease. We have developed a simplified model to describe our findings [Fig. 7(a)], where the share of total donor population rises, causing increasing steric hindrance. We have not included in our model the role of donor-donor competition and donor-donor energy transfer as contributing factors, which, while not appearing in the FRET channel, may nevertheless impact E. As shown in Fig. 7(b), E clearly decreases with increasing uD:A ratio. Correlation analysis substantiates these conclusions, with values of r=−0.77, suggesting that clustering of receptor-ligand complexes occurs in apical endocytic membranes.

Figure 7

(a) Donor geometric exclusion hypothesis. By definition, molecules in a cluster are in proximity. As the D:A ratio and/or the number of donor molecules increases, a phenomenon (donor geometric exclusion) occurs in which donors will block other donors from transferring energy to an acceptor due to increased steric hindrance, thus leading to a reduced E. (b) Negative dependence of uD:A ratio on E using 1-P microscopy. E was plotted against uD:A ratios (Panel 1-P; diamonds). Qualitatively, the experimental results match the donor geometric exclusion hypothesis, confirming a clustered distribution.

When comparing 1-P and 2-P results, we see identical trends in the reduction of E with increasing uD levels (Fig. 8). Correlation analysis confirms these conclusions, with values of r=−0.72 and r=−0.85, respectively, for 1-P and 2-P data. These results suggest that negative dependence of uD on E can be used as an indicator of clustered distribution of receptor-ligand complexes in endocytic membranes. The relationship between uD levels and E is especially useful for 2-P microscopy where we do not acquire acceptor data (Fig. 4), confirming a clustered distribution of receptor-ligand complexes.

Figure 8

Negative dependence of uD levels on E is comparable using 1-P and 2-P FRET microscopy. E was plotted against uD levels [(a) panel 1-P and (b) panel 2-P]. 1-P and 2-P microscopy, where we only collected donor information, show comparable results in which E decreases when uD levels increase.

4.1.

PFRET Algorithm to Correct Spectral Spillover Contamination

Different spillover corrections methods exist—each perfectly suitable, depending on the level of sensitivity desired.^{7 8 9 10 11} For advanced quantitative evaluation, very sensitive methods, such as the recently developed PFRET algorithm, are essential.¹² The algorithm method was confirmed by the standard method of bleaching the acceptor (data not shown), where E is calculated on the basis of donor emission before and after total photobleaching of the acceptor. Here, emission filters with different bandwidths were used, where increasing FRET fluorescence with filters of wider bandwidths caused increased spillover contamination. The PFRET algorithm corrects the spillover contamination at different levels of magnitude, accompanied by comparable and statistically indistinguishable E results. While the type of emission filter may be less important with the new spillover correction algorithm, it is prudent to select filters that will collect optimal levels of the FRET signal, which in turn will lead to greater differentiation between data points. For our particular experimental system, the 590 LP emission filter is the most adequate to acquire high levels of FRET signal with a corresponding reduced spillover contamination.

4.2.

Comparing 1-P with 2-P Microscopy

For our particular experimental system, both 1-P and 2-P microscopy produce comparable results. We are using a polarized epithelial MDCK cell line, where a monolayer of cells achieves a height of 15 to 20 μm. Imaging at a focal plane of ≈3.5 μm below the apical surface, the more discrete focal section achieved by 2-P microscopy does not appear to offer any advantages. For the determination of clustered distribution, the 2-P data analysis post-image acquisition only has to deal with donor crosstalk contamination, since the 790-nm donor excitation wavelength does not appear to excite the acceptor fluorophore¹² (data not shown). If acceptor information is required for specific experimental systems, a different excitation wavelength and/or donor-acceptor pair can be used. In summary, the FRET assay presented here is suitable to establish qualitatively and quantitatively whether clustering of membrane-bound complexes occurs in endocytic membranes.