SEARCH

SEARCH BY CITATION

Keywords:

  • autofluorescence;
  • flow cytometric resonance energy transfer (FCET);
  • indirect immunofluorescence;
  • cyanine 3 and 5 donor–acceptor pair;
  • CD45 isoforms;
  • HLA class I molecules

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED
  8. APPENDIX

Background

Flow cytometric fluorescence resonance energy transfer (FCET) is an efficient method to map associations between biomolecules because of its high sensitivity to changes in molecular distances in the range of 1–10 nm. However, the requirement for a dual-laser instrument and the need for a relatively high signal-to-noise system (i.e., high expression level of the molecules) pose limitations to a wide application of the method.

Methods

Antibodies conjugated to cyanines 3 and 5 (Cy3 and Cy5) were used to label membrane proteins on the cell surface. FCET measurements were made on a widely used benchtop dual-laser flow cytometer, the FACSCalibur, by using cell-by-cell analysis of energy transfer efficiency.

Results

To increase the accuracy of FCET measurements, we applied a long wavelength donor–acceptor pair, Cy3 and Cy5, which beneficially affected the signal-to-noise ratio in comparison with the classic pair of fluorescein and rhodamine. A new algorithm for cell-by-cell correction of autofluorescence further improved the sensitivity of the technique; cell subpopulations with only slightly different FCET efficiencies could be identified. The new FCET technique was tested on various direct and indirect immunofluorescent labeling strategies. The highest FCET values could be measured when applying direct labeling on both (donor and acceptor) sides. Upon increasing the complexity of the labeling scheme by introducing secondary antibodies, we detected a decrease in the energy transfer efficiency.

Conclusions

We developed a new FCET protocol by applying long wavelength excitation and detection of fluorescence and by refining autofluorescence correction. The increased accuracy of the new method makes cells with low receptor expression amenable to FCET investigation, and the new approach can be implemented easily on a commercially available dual-laser flow cytometer, such as a FACSCalibur. Cytometry 48:124–135, 2002. © 2002 Wiley-Liss, Inc.

Protein–protein interactions are critical to most biological processes, extending from the formation of cellular macromolecular structures and enzymatic complexes to the regulation of signal transduction pathways. The cellular machinery is completely dependent on the exchange of signals, ions, and nutrients between the extracellular environment and the interior of cells. Preexisting and ligand-induced associations of membrane proteins are crucial to such processes. Although generally applied biochemical and immunological approaches (i.e., cocapping, coprecipitation, chemical cross-linking) have provided valuable information about the topography of cell surface protein interactions, such techniques have several disadvantages. For example, the necessary application of extraction or isolation procedures prevent the investigation of proteins in their natural environment and may, on the one hand, disrupt protein–protein interactions or, on the other hand, induce the formation of artificial protein aggregates, the latter being a main concern in case of high expression of cell surface glycoproteins (1). In contrast, fluorescence resonance energy transfer (FRET) offers a suitable and convenient alternative because it can be performed on live cells without major interference with the physiologic condition of the cells. FRET is a process by which an excited donor dye transfers its energy to a nearby acceptor dye through radiationless dipole–dipole interaction (2). The energy transfer efficiency, which is a function of the inverse sixth power of the distance between the donor and acceptor, changes steeply in the range of 1–10 nm and can be used to assess molecular associations (2).

Flow cytometric FRET (FCET) statistically can provide very accurate information on the cell surface distribution of membrane proteins and conformational changes of biologically active molecules. This technique has been applied successfully to a wide range of biological systems, such as monitoring the association state of membrane proteins in immunologically competent cells and various tumor cells; for review, see Szöllősi et al. (3). However, FCET cannot be used easily in cellular systems that show autofluorescence (3). Although FCET does provide information on the lateral organization of molecules with high sensitivity, the technique is not widely applied to cellular systems for two main reasons. First, advanced dual-laser flow cytometric instruments with excitation wavelengths specific for the donor–acceptor pair are required. Although the measurement of FRET-induced donor quenching does not need dual-laser instruments or complicated evaluation, quenching cannot be used for cell-by-cell data analysis of FRET efficiency (3). Second, accuracy and reproducibility of FCET measurements are compromised if the ligands for the fluorescently labeled probes are expressed at low levels. In such cases, the contribution of autofluorescence may be significant relative to the specific signal.

Several reports have offered solutions to diminish the interference of high cellular autofluorescence with the detection efficiency of the specific signal. The high autofluorescence of alveolar macrophages excited with blue light has been quenched by crystal violet (4). Crystal violet, however, may weaken the specific signal. Furthermore, sodium tetrahydridoborate has been used successfully to reduce the autofluorescence of immunochemically stained cells in suspension (5). Alternatively, time-resolved detection of the fluorescence of europium chelates can improve the signal-to-noise (S:N) ratio even in glutaraldehyde fixed samples, where autofluorescence is always high. This is achieved by selective detection of the long-lived europium fluorescence through a time window in which short-lived autofluorescence has already decayed (6). In addition, a mathematical model based on the possible correlation between the autofluorescence and probe fluorescence has been developed for autofluorescence correction in flow cytometry (7). This model, however, is not based on a cell-by-cell approach, and it corrects only the distribution histograms to obtain a better discrimination between the so-called positive and negative cells (7). Importantly, multicolor compensation of autofluorescence on a cell-by-cell basis has been developed for flow cytometric analysis with the use of single-laser (8–11) and dual-laser (12, 13) excitations. These procedures are based on the observation that the excitation and emission spectra of autofluorescence and immunofluorescence are different. In fact, cellular autofluorescence is higher in the blue and green spectral regions than in the red spectrum, so excitation and detection at longer wavelengths reduce the contribution of autofluorescence (14, 15). Newly developed fluorophores with excitation and emission spectra in the red region are therefore superior to the more popular fluorescein and rhodamine fluorophores in FRET experiments. The carbocyanine dyes 3, 5, and 7 (Cy3, Cy5, and Cy7) are good candidates in this respect, and their application has been successful in microscopic FRET measurements (16–19).

We report the development of a highly sensitive FCET method that takes advantage of long emission wavelengths of cyanine dyes and a new technique for cell-by-cell correction of autofluorescence. This new FCET method can be applied in cellular systems where low S:N ratio limits the application of the conventional technique (20). We found that the spectral characteristics of Cy3 and Cy5 as a FRET dye pair sufficiently conform to the optical parameters of the dual-laser FACSCalibur flow cytometer. The dual-laser FACSCalibur is used commonly in daily clinical routine and scientific research, thus facilitating the implementation of the FCET technology. The use of the long emission wavelengths of cyanine dyes markedly reduces the cellular autofluorescence, which in combination with a cell-by-cell correction of autofluorescence using the broad spectrum of autofluorescence, significantly increases the accuracy of energy transfer data analysis.

To date, FRET results obtained by different researchers using distinct labeling protocols (21) are hard to compare. The increased sensitivity of our modified FCET technique allowed us to address important issues, namely the applicability of the method in cellular systems with high autofluorescence and the effect of different direct and indirect immunolabeling approaches on the measured FRET efficiencies. We could detect homoassociation of CD45 isoforms in cells having low expression levels comparable with cellular autofluorescence. We also examined the effects of direct and indirect immunofluorescent labeling with whole immunoglobulin G (IgG) or antigen-binding fragments (Fab) on the efficiency of an intramolecular energy transfer process occurring between the heavy and the light chains of HLA class I molecules. The highest FRET values were measured with directly labeled donor and acceptor antibodies. Increasing the complexity of the labeling scheme by introducing the secondary antibody clearly showed a decrease in the energy transfer efficiency.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED
  8. APPENDIX

Cells

The Epstein-Barr virus–transformed JY human B-lymphoblast cell line (a kind gift from Frances Brodsky, University of California, San Francisco) and the CD45 HPB-ALL T-cell lines transfected with the R0 isoform of CD45 were cultured in RPMI 1640 medium supplemented with 10% fetal calf serum in 5% CO2 atmosphere at 37°C (22). Peripheral blood lymphocytes from healthy donors were isolated and gene-transduced with a single-chain T-cell receptor (TCR):CD3ζ chimeric gene construct (23). Human T lymphocytes activated with anti-CD3 monoclonal antibodies (mAbs) were transduced with retrovirus containing the chimeric TCR gene produced by the stable packaging cell line Phoenix (kindly provided by Y. Soneoka, Oxford, UK). T lymphocytes were cultured and expanded in RPMI 1640 medium supplemented with 25 mM Hepes, L-glutamine, 10% human serum, and 360 IU/ml of recombinant interleukin-2 (Proleukin, Chiron, Amsterdam, The Netherlands) and stimulated every 2 weeks with a mixture of irradiated allogeneic feeder cells, as described elsewhere (24).

Antibodies

W6/32 (IgG2aκ) mAb with specificity for the heavy chain of class I HLA A, B, and C molecules and L368 (IgG1κ) mAb specific for β2 microglobulin were provided by F. Brodsky (University of California, San Francisco). Affinity-purified rabbit anti-mouse IgG2a (γ2a chain specific) and rabbit anti-mouse IgG1 (γ1 chain specific) were obtained from Rockland (Gilbertsville, PA). Cy3- and Cy5-conjugated and nonconjugated affinity-purified Fabs of goat anti-mouse IgG (H + L chain specific) were obtained from Jackson ImmunoResearch (West Grove, PA). Anti–Pan-CD45 (CD45.2, IgG1) was obtained from S. Meuer (Heidelberg, Germany), OKT-3 (anti-CD3) mAb was obtained from the European Collection of Cell Cultures (Salisbury, UK), and TCR Vβ1 IgG1 mAb was obtained from Immunotech (Marseille, France).

Fab Preparation

Fabs were prepared from mAbs as described previously (25). Briefly, IgG mAbs were dialyzed with phosphate buffered saline (PBS; 100 mM Na2HPO4, 150 mM NaCl, 1 mM ethylene-diamine-tetra-acetic acid, pH 8.0) and digested with activated papain at 37°C for 11 min. The enzyme activity was terminated by the addition of iodoacetamide. The reaction mixture was passed through a Sephadex G-100 fine column and the collected Fabs were separated further from intact Ig by using a protein A sepharose column.

Conjugation of Abs and Fabs With Fluorescent Dyes

Antibodies were labeled with fluorescent dyes as described previously (1), with slight modifications. Briefly, covalent binding of the bifunctional succimidyl ester derivatives of sulfoindocyanine dyes (Cy3 and Cy5; Amersham, Braunschweig, Germany) to the lysyl-ϵ amino groups of antibodies was carried out in bicarbonate buffer at pH 8.3. Dyes dissolved in bicarbonate buffer were added to antibody at a 100-fold excess, and the reaction mixture was incubated on ice for approximately 15 min. Unreacted dye molecules were removed by gel filtration through a Sephadex G-25 column. The dye:protein labeling ratio was determined by spectrophotometry, and its value was approximately 3:1 in the case of whole IgGs and approximately 1:1 in the case of Fabs for Cy3 and Cy5.

FCET calculations are based on the “effective” labeling ratio, which takes the quantum efficiency changes of the dye moiety in the fluorescent conjugate into account. Therefore, the relative quantum efficiency values of the fluorescently labeled IgGs and Fabs were determined by spectrofluorometry. First, the absorbance of the labeled antibody solution at the excitation wavelength of the dye was determined. Second, the corrected fluorescence emission spectrum of the labeled antibody was recorded and integrated throughout the whole range of emission wavelengths. As a standard, fluorescein dissolved in 0.1 M NaOH was used, which has a fluorescence quantum yield of 0.92, independent of wavelength under these conditions. The unknown fluorescence quantum yield was calculated according to equation 1:

  • equation image(1)

where Edye and Est are the integrated fluorescence emission spectra of the labeled antibody and the fluorescein standard, respectively; Adye and Ast are the absorbance values of the antibody and standard dye solutions, respectively, at the wavelength used for excitation during the measurement of the fluorescence emission spectra; and Qst is the fluorescence quantum yield of the standard.

Labeling of Cells With Abs and Fabs

Freshly harvested cells were washed twice in ice-cold PBS (pH 7.4). The cell pellet was suspended in 100 μl of PBS with 1% bovine serum albumin (1 × 107 cells/ml) and incubated with Cy3- or Cy5-conjugated mAbs at a final concentration of approximately 100 μg/ml for 40 min on ice in the dark. For indirect staining, cells were incubated with nonconjugated mAbs at a final concentration of approximately 100 μg/ml. After washing the cells twice in PBS, they were incubated further with Cy3- or Cy5-conjugated polyclonal rabbit anti-mouse antibodies against the IgGγ1 or γ2a chain. Subclass-specific second-step antibodies had no detectable cross reactivity with other IgG subclasses as verified by corresponding control samples (i.e., first antibody IgG1, second antibody anti-IgG2a). In some cases, polyclonal goat anti-mouse Fabs were used against the H + L chain of IgG at a final concentration of approximately 20–30 μg/ml. In case of indirect staining with Fabs, the following labeling sequence was applied: (a) labeling with nonconjugated W6/32 mAb; (b) Cy3-conjugated Fabs of goat anti-mouse IgG; (c) unlabeled Fabs of polyclonal goat anti-mouse antibodies were added (50 μg/ml concentration) to block the free binding sites remaining on W6/32 mAbs; (d) cells were labeled with L368 mAb; (e) followed by Cy5-conjugated Fabs of goat anti-mouse IgG as a second-step antibody. The Cy5-conjugated second-step reagent showed no significant binding to W6/32 mAb as verified by appropriate control samples (e.g., primary mAb, Cy3-conjugated secondary Fabs, blocking step, Cy5-conjugated secondary Fabs). The labeled cells were washed with ice-cold PBS and then measured in PBS without fixation or fixed with 1% formaldehyde to allow the storage of cells overnight until data acquisition. Data obtained with fixed cells did not differ significantly from those of unfixed, viable cells.

Flow Cytometric Energy Transfer Measurements

A Becton Dickinson FACSCalibur flow cytometer (Becton Dickinson, San Jose, CA) was used throughout the study. Four fluorescence intensities were measured. Three of them were excited at 488 nm and detected at 530 ± 15 nm, 585 ± 21 nm, and above 670 nm, respectively, and the fourth fluorescence intensity was excited at 635 nm and detected at 661 ± 8 nm. The fluorescence intensities excited at 488 nm were detected spatially and temporarily separately from those excited at 635 nm; thus, the fluorescence intensities can be measured independently even in partly overlapping spectral ranges. Forward and side scattering were used to gate out debris and dead cells. All data were stored in list mode format. Energy transfer efficiency (E) was calculated from the four fluorescence intensities on a cell-by-cell basis (20). E values are presented as mean values from approximately normally distributed, unimodal energy transfer histograms of 10,000 cells calculated from at least four independent measurements. The cell-by-cell correction for autofluorescence is described in the Results.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED
  8. APPENDIX

Correction Algorithm for Autofluorescence in FCET Measurements

The use of fluorescein and rhodamine (or dyes with similar spectral characteristics) in FCET without autofluorescence correction (20) is a powerful tool for investigating inter- or intramolecular distance relationships on the cell surface in cases of high receptor density, where the contribution of autofluorescence is insignificant relative to the specific signal. To improve the applicability of the FCET method in very low S:N systems, we applied the red-shifted donor–acceptor pair Cy3 and Cy5. In addition, we supplemented the FCET method with a mathematical treatment allowing cell-by-cell correction for autofluorescence. Since individual cells have different autofluorescence intensities, subtraction of the mean autofluorescence intensity of nonlabeled cells from each cell in the labeled sample, as used in the conventional FCET method (20), may falsify the specific fluorescence values. For the conventional method, the 488-nm line of an argon-ion laser was used to excite direct donor fluorescence and sensitized acceptor emission, whereas a second laser line at 514 nm excited direct acceptor fluorescence. Altogether, three fluorescence intensities were measured (20). In the modified FCET method, a fourth fluorescence intensity was introduced as an independent parameter for the level of cellular autofluorescence. Because we used Cy3 and Cy5 as a donor–acceptor pair, the 488-nm line of an argon-ion laser and the 635-nm line of a red diode laser could efficiently excite the donor and acceptor dyes, respectively. The emission maximum of Cy3 (565 nm) falls in the range of the FL2 detection channel, whereas the maximum of Cy5 emission (667 nm) is in the range of the FL3 and FL4 channels. The FL1 channel, which does not overlap with the emission spectrum of Cy5 at all and has very little overlap with the spectrum of Cy3, was used for detecting autofluorescence.

Equations 2–5 describe the composition of the four measured intensities in general, when each fluorescent component emits in each channel. The numbers in brackets refer to the excitation and detection wavelengths used to measure fluorescence in the FL1–FL4 channels. The background autofluorescence intensity (in channel 1), the unquenched donor fluorescence intensity (in channel 2), the directly excited acceptor fluorescence intensity (measured in channel 4), and the energy transfer efficiency are denoted by AF, ID, IA and E, respectively.

  • equation image(2)
  • equation image(3)
  • equation image(4)
  • equation image(5)

S1–S6 and B2–B4 factors characterize the “spectral overlap” between different channels (the ratios of the signals from, e.g., the donor-labeled, acceptor-labeled, or unlabeled samples detected in the different channels, see equations 6–14). Derivation of equations 2–5 is described in detail in the Appendix. The expression (εmath image εmath imagemath image εmath image) is the ratio of molar absorption coefficients of donor and acceptor dyes at the excitation wavelengths 488 and 635 nm. The α parameter describing the relative detection efficiency and quantum yield of the donor and acceptor is defined in the following section. S1, S3, and S5 are determined by using samples labeled only with Cy3 according to the following equations:

  • equation image(6)
  • equation image(7)
  • equation image(8)

S2, S4, and S6 are determined on cells labeled only with Cy5 according to the following equations:

  • equation image(9)
  • equation image(10)
  • equation image(11)

B2, B3, and B4 are determined on unlabeled cells according to the following equations:

  • equation image(12)
  • equation image(13)
  • equation image(14)

Both FL2 and FL3 showed good correlation with FL1 (r = 0.91 for the FL2–FL1 correlation and 0.85 for the FL3–FL1 correlation, P < 0.001 for both cases; Fig. 1). This finding justifies the application of B2 and B3 constants in the cell-by-cell autofluorescence correction. Good correlation also was observed between FL4 and FL1, but autofluorescence in the FL4 channel was so low that B4 was almost negligible. Nevertheless, B4 was used in our calculations to obtain more accurate FRET efficiency values.

thumbnail image

Figure 1. Correlation between autofluorescence signals detected in different channels. B factors (see equations 2–5 and 12–14), the ratios of the fractions of the autofluorescence spectrum in individual detection channels, are the slopes of the fitted lines. FL1 is measured at 488-nm excitation and 530 ± 15 nm emission; FL2 is measured at 488-nm excitation and 585 ± 21 nm emission; FL3 is measured at 488-nm excitation and >670 nm emission; and FL4 is measured at 635-nm excitation and 661 ± 8 nm emission.

Download figure to PowerPoint

Fortunately, using the filter and dichroic mirror setup of FACSCalibur resulted zero values for S3, S4, and S6; in addition, the ratio (εmath image εmath imagemath image εmath image) was zero, which simplified equations 2–5. For the sake of mathematical simplicity, we defined parameter A, and solving the equations yields:

  • equation image(15)

From this equation, the single-cell FRET efficiency E can be expressed as:

  • equation image(16)

Determination of the α Parameter

The factor α defines the detection sensitivity of fluorescence from an excited acceptor molecule with respect to the sensitivity to detect an excited donor molecule. The signal arising from sensitized emission of the acceptor is proportional to the number of excitation quanta transferred from the donor to the acceptor and is therefore proportional to IDE. The proportionality factor α can be determined experimentally from the ratio: the FL3 signal arising from N pieces of excited acceptor molecules to the FL2 signal arising from N pieces of excited donor molecules. Factor α depends on the fluorescence quantum yields of the used dyes (QD for the donor and QA for the acceptor) and the overall detection efficiencies of the respective detection channels as described in equation 17:

  • equation image(17)

where ηD and ηA are the detection efficiencies of channels 2 and 3 to photons with wavelength distributions of the donor and the acceptor emissions, respectively. Experimentally, α was determined by comparing the FL2 signal of a donor-only labeled sample with the FL3 signal of an acceptor-only labeled sample:

  • equation image(18)

where B denotes the mean number of receptors per cell labeled by the corresponding antibody, L denotes the mean number of dye molecules attached to an antibody molecule, and ε denotes the molar extinction coefficient of the dyes at 488 nm; subscripts A and D refer to acceptor and donor, respectively. The ratios of B, L, and ε correct for the different numbers of photons absorbed by the acceptor- and donor-only labeled samples.

The fluorescence quantum yields of the dyes may depend on the type of antibody they are attached to and even on the labeling ratio L, thereby affecting the value of α. The α factor determined for a given donor–acceptor antibody pair can be used for other antibody pairs labeled with the same dyes provided its value is corrected for the differences in the quantum yields:

  • equation image(19)

where subscript 1 refers to the antibody pair for which α had been previously determined, and subscript 2 refers to the new antibody pair.

Increased Sensitivity of Autofluorescence-Corrected FCET

The modification of the original method of FCET, now incorporating autofluorescence background correction on a cell-by-cell basis, extends the applicability of the method to cells with a significantly lower receptor number. Two biological systems, one with a low expression level of surface molecule of interest and the other with a heterogeneous population of cells with respect to the association of interest between surface molecules that could not be accurately analyzed by the conventional FCET method, were tested with our modified technique. For instance, the expression level of the R0 isoform of the CD45 tyrosine phosphatase on a transfected HPB-ALL T-cell line was too low to accurately determine the extent of homodimerization of these molecules by the traditional FCET method because the mean of the FL2 histogram of cells labeled with Cy3-CD45 was only twice as high as the mean autofluorescence of unlabeled cells (signal:autofluorescence ratio of 1:1; data not shown). As shown in Fig. 2, cell-by-cell autofluorescence correction by the modified FCET method resulted in a more reliable measurement of the energy transfer histogram: the improved method yielded a shift to the positive range and a lower standard error of the mean (S.E.M.; curve a: mean ± S.E.M. = 4.9 ± 2.6% as compared with the result of the traditional way of background correction, i.e., the use of constant autofluorescence values (curve b: −0.1 ± 7.6%).

thumbnail image

Figure 2. Effect of cell-by-cell autofluorescence correction on the fluorescence resonance energy transfer (FRET) efficiency distribution at low signal-to-noise ratio. CD45 HPB-ALL T cells transfected with the CD45 R0 isoform were labeled with cyanine 3 (Cy3)- and Cy5-conjugated antigen-binding fragments of CD45 pan monoclonal antibody. Calculation of FRET efficiency was carried out with cell-by-cell autofluorescence correction (curve a; mean ± standard error of the mean: 4.9 ± 2.6%), or constant autofluorescence values were subtracted from the specific signals (curve b; −0.1 ± 7.6%).

Download figure to PowerPoint

Biological samples may contain different subpopulations of cells with respect to the association between surface molecules of interest. Discrimination between such subpopulations is essential but used to be difficult with the conventional FCET dye pairs (fluorescein and rhodamine) due to the high variance of the energy transfer distribution. Our new FCET technique can distinguish much more accurately between subpopulations of cells with different energy transfer efficiencies. FCET studies with primary human T lymphocytes transduced with single-chain chimeric TCR chains showed that the membrane topology of an exogenous TCR molecule (i.e., scTCR:CD3ζ as described in (23)) differs from that of full-length, nonmodified TCR chains (R. Debets et al., unpublished observations). Most transduced T lymphocytes express the scTCR:CD3ζ chain (i.e., cells positive for Vα12 and Vβ1), but the existence of a subpopulation of T lymphocytes, albeit a small subpopulation, that expresses an endogenous TCR chain recognized by the anti-TCR antibody (i.e., a Vβ1 chain) may hamper the intended FRET studies significantly. In this particular example, the energy transfer histogram had a bimodal rather than a unimodal frequency distribution when the endogenous TCRβ chain appeared to be present in a fraction of the cells. As shown in Fig. 3, the histogram of energy transfer between Cy5 labeled CD3 molecules and Cy3-labeled scTCR:CD3ζ constructs has two, clearly separated modes (peaks). The mean values of the segregated peaks are approximately 2% and 8%, implying that at least two subpopulations are present: cells expressing the exogenous scTCR:CD3ζ chain are represented by low energy transfer values, and cells expressing the endogenous TCRβ are represented by high energy transfer values.

thumbnail image

Figure 3. Increased sensitivity of our flow cytometric fluorescence resonance energy transfer (FCET) method allows for the discrimination of cell populations that differ in their energy transfer efficiencies. Human T lymphocytes were transduced with a gene coding for the chimeric scTCR:CD3ζ construct. Energy transfer was measured between cyanine 3 (Cy3)-labeled chimeric TCR:CD3ζ molecules and Cy5-labeled endogenous CD3 molecules using TCR-Vbeta1 mAb and Cy3-conjugated Fab fragments of Goat-anti-Mouse (GAM)-IgG and Cy5-conjugated OKT3 mAb. Gating the cells with higher fluorescence intensities on the dot plot of FL2 (Vbeta1+ Cy3-GAM-IgG) and FL4 (Cy5-OKT3) intensities (a) yielded a bimodal distribution of FRET efficiency (b). The gate was set to more clearly demonstrate a bimodal FRET efficiency distribution.

Download figure to PowerPoint

Geometry of Labeling Scheme Influences FRET Efficiency Values

FRET applications are often limited by the lack of appropriate, directly labeled mAbs to serve as a donor–acceptor pair. The use of fluorophore-conjugated secondary antibodies or Fabs may overcome this problem. However, the altered geometry of indirect versus direct labeling can significantly affect the observed FRET efficiency values. To check whether indirect labeling influences FRET efficiency measurements, we applied various labeling schemes to our model system of intramolecular energy transfer between two epitopes on a single HLA complex on JY cells and analyzed energy transfer with our modified FCET method.

Figure 4A illustrates labeling strategies in measurements conducted on JY cells by the FCET method. Red-emitting fluorescent dyes were indispensable to reliably discriminate between FCET histograms measured with different labeling protocols because autofluorescence would have made this resolution impossible with the conventional fluorescein–rhodamine pair. When staining the heavy and light chains (β2m) of HLA class I molecules with W6/32 and L368 mAbs, respectively, the direct or indirect labeling procedure differs slightly with respect to the distance between donor and acceptor fluorophores as a consequence of the different sizes of the used antibody complexes. In the case of secondary labeling, the concentrations of fluorescent secondary antibodies or their Fabs were set to ensure that the donor–acceptor ratio was close to one, as in the case of directly labeled primary antibodies, so that the FRET results were not influenced by the donor–acceptor ratio but rather by the distance between the fluorophores. The alterations in distance between fluorophores markedly affected energy transfer measurements (Fig. 4B). Labeling of both HLA epitopes with Cy3- and Cy5-conjugated secondary Fabs led to lower energy transfer efficiencies (15%; curve b/b) compared with direct labeling with fluorophore-conjugated primary mAbs or Fabs of these mAbs (approximately 30–28%; curves a/a and a′/a′). When cells were labeled indirectly with fluorophore-conjugated polyclonal whole IgG instead of Fabs, an even lower FRET efficiency (approximately 8%; curve c/c) was observed. The peak position of the energy transfer distribution curves decreased as we increased the size of the applied antibody complexes. As shown in Table 1, combining indirect labeling with Fabs on the donor side with direct labeling on the acceptor side caused an intermediate value of energy transfer efficiency (24%). Interestingly, combining direct labeling on either side and indirect labeling with whole IgG on the other side provided similar FRET values as indirect staining on both sides (i.e. approximately 9%). In JY cells the conventional FCET (where the autofluorescence was corrected for by using a constant value) provided practically the same results as the improved FCET method in which the autofluorescence was corrected for on a cell-by-cell basis because the expression level of HLA class I molecules is very high on JY cells (approximately 1 million molecules per cell) (25), resulting in a very good S:N ratio (data not shown).

thumbnail image

Figure 4. Labeling strategy affects the sensitivity of flow cytometric fluorescence resonance energy transfer (FCET). A: Strategy of labeling. The heavy and light chains (β2m) of class I HLA molecules was labeled directly with fluorescently tagged primary monoclonal antibodies (mAbs; a) or antigen-binding fragments (Fabs; a′), indirectly with unlabeled mAb, and consequently with fluorescent Fabs of Goat-anti-Mouse (GAM)-IgG (b) or whole Rabbit-anti-Mouse (RAM)-IgG (c) molecules. We investigated whether the size of the applied antibody complexes affects the sensitivity of our FCET technique. We used fluorophore-conjugated mAbs (a/a, Cy3-L368 + Cy5-W6/32), Fabs (a′/a′; Cy3-L368 Fab + Cy5-W6/32 Fab), unlabeled mAb followed by fluorophore-tagged polyclonal Fabs (b/b; L368+Cy3-GAMIgG Fab and W6/32+Cy5-GAMIgG Fab), or unlabeled mAb followed by fluorophore-tagged polyclonal whole IgG (c/c; L368+Cy3-RAMIgG1 and W6/32+Cy5-RAMIgG2a). In the case of indirect staining (b, c), more than one polyclonal antibody or Fab can bind to a single primary antibody. For the sake of simplicity, only one secondary antibody or Fab is shown in the scheme. B: Representative histograms (one of five independent experiments) displaying fluorescence resonance energy transfer efficiency distributions measured between the light and heavy chains of class I HLA molecules with the use of direct labeling with whole IgG (curve a/a) or Fabs (curve a′/a′), indirect labeling with nonconjugated mAbs followed by fluorescently tagged polyclonal Fabs (curve b/b), or fluorescently tagged polyclonal whole IgG (curve c/c). Curve d represents the fluorescence resonance energy transfer distribution of cells in the absence of energy transfer when cells were labeled with only a Cy3-conjugated primary mAb. Histograms displayed were measured on the same day, and mean values of the curves do not exactly correspond to the data displayed in Table 1.

Download figure to PowerPoint

Table 1. Flow Cytometric Measurements of Intramolecular Energy Transfer Efficiency With Various Labeling Strategies*
DonorD/A labeling strategyAcceptorFRET (%) ± S.D.
  • *

    Energy transfer values between the specified donor and acceptor pairs are displayed. Energy transfer values were calculated from the means of energy transfer histograms of 10,000 cells. Standard deviations (S.D.) were calculated from five independent measurements. W6/32 monoclonal antibody or Fab acts against the heavy chain of HLA class I molecules; L368 monoclonal antibody or Fab acts against the β2 microglobulin. Cy, carbocyanine; D/A, donor/acceptor; Fab, antigen-binding fragment; FRET, fluorescence resonance energy transfer; GAM, Goat-anti-Mouse; IgG, immunoglobulin G; RAM, Rabbit-anti-Mouse.

  • a

    W632 + Cy3-GAMIgG Fab corresponds to indirect labeling, where the incubation of cells with the first antibody (W6/32) was followed by incubation with the secondary antibody (Cy3-GAMIgG Fab), as described in Materials and Methods. In the case of secondary labeling, the concentrations of fluorescent secondary antibodies or their Fab fragments were set to ensure that the donor:acceptor ratio was close to 1 (0.86 ± 0.26), as in the case of directly labeled primary antibodies, so that the FRET results were not influenced by the donor:acceptor ratio. We applied this approach for antibody combinations where we were certain that all antigens were labeled with at least one fluorescencently tagged IgG or Fab.

Cy3-W6/32a/aCy5-L36829.4 ± 1.9
Cy3-L368a/aCy5-W6/3230.5 ± 1.3
Cy3-W6/32 Faba′/a′Cy5-L368 Fab29.8 ± 5.2
Cy3-L368 Faba′/a′Cy5-W6/32 Fab27.1 ± 2.1
W6/32 + Cy3-GAMIgG Fabab/aCy5-L36824.5 ± 0.4
L368 + Cy3-GAMIgG Fabb/aCy5-W6/3223.4 ± 1.6
W6/32 + Cy3-GAMIgG Fabb/bL368 + Cy5-GAMIgG Fab19.9 ± 4.7
L368 + Cy3-GAMIgG Fabb/bW6/32 + Cy5-GAMIgG Fab13.3 ± 3.5
W6/32 + Cy3-RAMIgG2ac/cL368 + Cy5-RAMIgG110.8 ± 2.8
L368 + Cy3-RAMIgG1c/cW6/32 + Cy5-RAMIgG2a8.9 ± 2.5
W6/32 + Cy3-RAMIgG2ac/aCy5-L3688.1 ± 1.6
L368 + Cy3-RAMIgG1c/aCy5-W6/329.6 ± 0.5

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED
  8. APPENDIX

Although FCET is often used to study molecular associations in intact cells (3, 16, 19), widespread application of the technique is limited by the need for a relatively high expression level of proteins to achieve a high S:N ratio. Cellular autofluorescence is one major source of noise in flow cytometry, thereby decreasing the S:N ratio. In this study we used two approaches to decrease the relative contribution of autofluorescence to the measured intensities: (a) application of long wavelength dyes, whose excitation and emission spectra show less overlap with those of cellular autofluorescence; and (b) a mathematical cell-by-cell correction for autofluorescence with the use of an independent fluorescence channel as a measure for cellular autofluorescence. We demonstrated that the lowest number of receptors, which can be measured reliably by FCET, is significantly lower if the above approaches are applied. In addition, these approaches can be implemented on commercially available instruments. Because secondary labeling strategies are often used to increase the signal in flow cytometry, we carried out a detailed analysis as to whether different labeling protocols change FCET measurements. Our results emphasized that FCET results should be compared with care, when obtained with different labeling methods.

We used Cy3 and Cy5 as the donor–acceptor pair for FCET measurements. Autofluorescence is less of a problem in the red region of the spectrum, which makes autofluorescence almost negligible in the case of Cy5. This makes the Cy3–Cy5 pair superior to the more widely used fluorescein–rhodamine pair for FRET applications. The extent of spectral overlap between Cy3 emission and Cy5 absorption is similar to that between fluorescein emission and tetramethylrhodamine absorption, resulting in similar R0 values (5.0 and 5.6 nm) (26, 27). Even if fluorescein-like dyes with higher fluorescence quantum efficiency and less photobleaching are available, the relatively high autofluorescence of cells is a problem in this spectral region, especially when less abundantly expressed proteins are studied.

Cellular autofluorescence originates mainly from the cytoplasm. Oxidative processes and certain molecules (e.g., reduced nicotinamide adenine dinucleotide (NADH) and catecholamines) may contribute to autofluorescence (28). Therefore, the autofluorescence level depends on the metabolic state of the cell and can display wide heterogeneity even among cultured cells derived from a single progenitor.

Because of the cell-by-cell variation in autofluorescence, methods using a population average value are not accurate to correct for autofluorescence. On the one hand, cells with a lower than average autofluorescence are overcorrected by this method, which, in the case of low specific signal, may even result in negative values. On the other hand, highly autofluorescent cells are undercorrected by this method. Inaccurate correction of fluorescence intensities significantly affects the calculated energy transfer values because the measured parameters are divided by each other, as is necessary in FCET calculations, which results in compounded errors in the derived parameters.

To improve the accuracy of FCET measurements, we used a cell-by-cell correction for autofluorescence. The autofluorescence values measured in the different channels correlated extremely well, which was a prerequisite for the correction on a single-cell basis (Fig. 1). Strongly positive correlations of autofluorescence signals were detected on other T-cell lines (Jurkat) and confluent breast carcinoma cells (SKBR-3, MDA-453; data not shown). Cell-by-cell correction for autofluorescence was especially important for the FL2 and FL3 channels because autofluorescence is not negligible in these channels as it is in the FL4 channel. In addition, miscalculation of the FL2 signal is the largest source of error in energy transfer calculations (29). Improved data analysis resulted in smaller errors in the calculation of energy transfer values of single cells. This contention is supported by the observation that FCET histograms showed a smaller coefficient of variation (Fig. 2), which made the calculation of the mean of FCET histograms more reliable. Reduction of autofluorescence-related errors increases not only the reproducibility of FCET measurements but also the accuracy of the method; i.e., FCET histograms not only get narrower but also may be shifted. This is crucial to the study of proteins with very low expression levels, such as the CD45 protein tyrosine phosphatase on a transfected HPB-ALL T-cell line (Fig. 2). Moreover, the significantly increased sensitivity of our new FCET method incorporating cell-by-cell correction of autofluorescence allowed the detection of small subpopulations with different energy transfer values (Fig. 3).

In many cases, the aim of FRET experiments is to decide whether the proteins of interest associate with each other, i.e., whether the measured FRET efficiency is significantly different from zero. FRET values above 5% are considered significantly different from zero in the case of the fluorescein–rhodamine donor–acceptor pair (25). Smaller coefficients of variation of FCET histograms and increased reproducibility and accuracy of the modified method make lower energy transfer values acceptable as significantly different from zero. Of course, this limit still depends on the S:N ratio, but in optimal conditions it can be as low as 2% (Pascal Batard, personal communication).

Indirect immunofluorescent labeling strategies may be applied to FCET measurements, if a suitable fluorophore-conjugated monoclonal antibody is not available or to enhance the specific fluorescence signal (21). Our intramolecular FCET measurements between the heavy and light chains of HLA class I molecules showed that the application of a larger antibody complex causes a decrease in FRET efficiency due to the geometry of the antibody complexes; i.e., when antibody or Fab complexes get larger, the actual distance between the donor and acceptor fluorophores increases. This explains the decreased FRET efficiency when fluorescent secondary Fabs are used on the donor and acceptor sides relative to primary fluorescent antibodies. It also explains the additional decrease in FRET efficiency when whole fluorescent secondary antibodies were used. The size of the applied fluorescent dyes also can affect the value of energy transfer. Energy transfer efficiencies are even lower when using phycoerythrin- and allophycocyanin-conjugated mAbs (Pascal Batard, personal communication) in comparison with Cy3- and Cy5-conjugated mAbs. Pfeiffer et al. (30) also obtained lower FRET efficiencies when applying phycoerythrin and Cy5 dyes as donor–acceptor pairs. Labeling with fluorescent secondary antibody on the acceptor side not only increases the donor–acceptor distance but also may increase the acceptor concentration because, instead of a single acceptor-labeled primary antibody, several acceptor-labeled secondary antibodies could be present. To rule out this possibility, the secondary antibodies and Fabs were applied in such a concentration that the donor and acceptor concentrations were kept constant. We applied this approach for antibody combinations where we were able to make sure that all epitopes were labeled with at least one fluorescencently tagged IgG or Fab to avoid the disturbing effect of empty epitopes in the FRET efficiency calculation. In these cases, the FRET values are not influenced by the donor–acceptor ratio but by the relative orientation of the fluorophores and the distance between them. Secondary labeling with fluorescent Fab or IgG on the donor side decreases FRET efficiency because the size of the labeling complex increases (Table 1). Although it is difficult to predict the value of energy transfer without molecular models and calculations, our results emphasized that FRET values cannot be directly compared with each other if they were obtained with different labeling strategies.

In conclusion, we provided a detailed description of a modified method to measure energy transfer in flow cytometry. The main value of the approach is reduction of autofluorescence-related errors, which is achieved by the application of long wavelength dyes, such as Cy3 and Cy5, as a donor–acceptor pair and by cell-by-cell correction of autofluorescence. Our modified FCET method allows energy transfer efficiency to be determined on cellular systems with a very low expression level of the molecules. Further, the lower variance of FCET distributions enables a much more accurate discrimination of subpopulations having distinct FRET efficiencies. The practical advantage of the new method is that it can be implemented easily on a commercially available dual-laser benchtop flow cytometer without the need for hardware modifications.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED
  8. APPENDIX

We greatly acknowledge Gergely Szentesi's fundamental role in writing and organizing the FLEX program capable of calculating fluorescence resonance energy transfer efficiency on a cell-by-cell basis. We also thank Ms. Gabriella Őri, Mrs. Teréz Lakatos, and Mrs. Tünde Terdik for their skillful technical assistance.

LITERATURE CITED

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED
  8. APPENDIX

APPENDIX

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. LITERATURE CITED
  8. APPENDIX

The following derivation forms the basis of equations 2–5 presented in the Results section. The specific fluorescence signal (not including autofluorescence) in the four detection channels is as follows:

  • equation image(A1)
  • equation image(A2)
  • equation image(A3)
  • equation image(A4)

ρ488 and ρ635 are the photon fluxes of the corresponding laser lines at the place of excitation; ε refers to the molar extinction coefficients of the donor (D) and acceptor (A) at the wavelengths indicated in the subscript; LDand LA are the dye:protein labeling ratios of the donor- and acceptor-labeled antibodies used to target the cell surface receptors; BD and BA are the numbers of receptors on the cell labeled by donor- and acceptor-tagged antibodies, respectively; E is the mean FRET efficiency for the given cell, QD and QA are the fluorescence quantum yields of the donor and acceptor dyes; and ηmath image and ηmath image are the detection efficiencies in the respective detection channels (i = 1, 2, 3, 4) for the photons having a wavelength distribution of the donor or the acceptor. ηi values comprise the transmission characteristics of the mirrors and filters in the detection pathways and the detection sensitivities of the detectors to the photons reaching the appropriate photomultiplier for the photon distribution arising from the donor or the acceptor.

The above set of equations can be solved for E, BD, and BA if all the other instrumental and spectroscopic factors present in the different terms are known. However, because the values of ηmath image and ηmath image cannot be easily determined, the introduction of the following parameters (all of which can be experimentally measured; see equations 6–11 and 17) is useful:

  • equation image(A5)
  • equation image(A6)
  • equation image(A7)

The factors S1, S3, and S5 can be determined with donor-labeled cells, and S2, S4, and S6 can be determined with acceptor-labeled cells (see equations 6–11 in the Results section). The value of α can be measured by comparing the FL3 signal of an acceptor-labeled sample and the FL2 signal of a donor-labeled sample (with additional correction for the “absorption side”; see equation 18):

  • equation image(A8)

Preferentially, the same type of donor- and acceptor-labeled mAbs are used for the calculation of α so that the BD/BA term is cancelled.

We define two more parameters, ID and IA, for the would-be FL2 signal of the nonquenched donor (if E = 0 were the case) and the FL4 signal of the acceptor:

  • equation image(A9)
  • equation image(A10)

Using these definitions, the equations for FL1 to FL4 take on the following form:

  • equation image(A11)
  • equation image(A12)
  • equation image(A13)
  • equation image(A14)

These equations describing the specific fluorescence can be extended to include the autofluorescence contributions yielding equations 2–5 in the Results section. The solution of equations 2–5 by using the method of determinants yields the cell-by-cell values for the FRET efficiency E, the specific fluorescences of the donor and the acceptor ID and IA, and the autofluorescence AF. Taking into account that the factors S3, S4, and S6 are zero, and so is the value of (εmath image εmath imagemath image εmath image), equations 2–5 are significantly simplified, providing the following solution:

  • equation image(A15)
  • equation image(A16)
  • equation image(A17)
  • equation image(A18)
  • equation image(A19)

These calculations can be carried out on the data of a selected cell population by using a custom-written software program or importing data into commercially available spreadsheet programs. We used a custom-made program (FLEX) written in Delphi, where one can define histograms and dot plots of fluorescence intensities, select populations, and store the mean values of fluorescence intensities or any derived parameters in an Excel spreadsheet inside the program. The program is available on the Web as freeware with help instructions. The Web page is www.biophys.dote.hu/kutatas/research.htm, and the link to follow is “instruments, software”.