Cystic fibrosis (CF) is the most common life-shortening recessive disease in the caucasian population and results from mutation of the cystic fibrosis transmembrane conductance regulator (CFTR) gene (1). CF is a multiorgan disease, and disease progression is highly variable between patients. Recent data indicate that CFTR deficiency in immune cells may further contribute to the heightened inflammatory responses and increased pathogen pressure (2, 3). However, still little is known on CFTR function in various immune cell subsets, and more methods to analyze CFTR function in such cell types are needed.
The most prevalent loss-of-function CFTR mutation is the deletion of phenylalanine 508 (F508del), but over 1500 mutations have been associated with CF (www.genet.sickkids.on.ca / www.cftr2.org). These mutations have been classified according to their impact on CFTR function (4). In general, Class I mutations affect CFTR translation (stop codons, e.g., G542X), Class II mutations involve the folding and apical trafficking of CFTR (e.g., F508del). Classes III (e.g., G551D), IV (e.g., R117H), and V affect CFTR gating, conductance, and splicing, respectively. Novel drugs are being developed that target mutation-specific defects of CFTR from which correctors that improve apical trafficking of CFTR such as VX-809 (5) and potentiators that improve CFTR gating, such as VX-770 (6), are highly promising (7).
Multiple techniques have been described to study CFTR function including electrophysiological approaches such as patch-clamp recordings and short-circuit current measurements in Ussing chambers or influx and efflux measurements of radioactive ions (8). From these assays, iodide-mediated quenching rates of an ectopically expressed mutant form of yellow fluorescent protein (YFP) has been widely used as a relatively simple assay to measure CFTR function upon introduction or removal of iodide (9–11). This assay is optimally suited for CFTR function measurement in plate-bound adherent cells, and accurate CFTR function measurement are achieved through normalization for YFP expressing level at the start of the experiment.
Here, we developed a ratiometric sensor to measure CFTR function based on the halide-sensitive YFP and a halide-insensitive red fluorescent protein. Our ratiometric indicator effectively corrects for differences in sensor expression level between experimental units such as individual cells and can accurately measure CFTR-dependent iodide in- and efflux. Our sensor is particularly suited to study CFTR function when variability in sensor expression levels cannot be normalized in a specific experimental setup as we show here for flow cytometry. The method described here may also be used to study CFTR function in suspension cells or yeast cells, which are used to express CFTR for crystallization studies (12).
Materials and Methods
Cell Culture and Plasmids
Baby Hamster Kidney (BHK) cells stably expressing CFTR wild type (wt) (BHK-CFTRwt) or CFTR-F508del (BHK-F508del) (13) were maintained in DMEM (Invitrogen) supplemented with 8% heat-inactivated FCS, penicillin and streptomycin (Invitrogen), and methotrexate at 37°C and 5% CO2. Venus-constructs were made by overlap extension polymerase chain reaction (PCR) (14) from Venus constructs previously described (15). pcDNA3-Venus-linker-dsRed was constructed by PCR-amplification of Venus and dsRed and sequential ligation into pcDNA3, which already contained the linker (15). Halide-sensitive Venus was created by introducing the H148Q/I152L mutations (9) by QuickChange™ (Stratagene, La Jolla, CA) site-directed mutagenesis method using Phusion® polymerase (New England Biolabs, Ipswich, MA). pcDNA3-Venus-E2A-dsRed was created by replacing the linker in Venus-linker-dsRed by the autocleavable peptide E2A (16). pcDNA3-Venus-E2A-mKate was constructed by replacing dsRed by mKate2 (17). pcDNA3-YFP-E2A-mKate was constructed by replacing Venus for YFP, which is constructed from pEYFP-N1 (Clontech, Mountain View, CA) with the additional H148Q/I152L and F46L (18) mutations for increased iodide sensitivity and fluorescent intensities, respectively. CFTR mutants were made by QuickChange site-directed mutagenesis on pcDNA3-CFTRwt using Phusion polymerase. Primers were designed to extent approximately 15 bp up- and downstream of the site of mutagenesis but always ending with C or G. The PCR program used for the CFTR-mutants was as follows: 98°C for 60 s; followed by 25 cycles 98°C for 25 s, 55°C for 25 s, and 72°C for 10 min; final elongation step was at 72°C for 10 min. All constructs were sequence verified. BHK cells were lentivirally transduced using Phage2-YFP-E2A-mKate carrying a puromycin selection cassette. Phage2 backbone was a gift from Dr. D. D'Astolfo (Hubrecht Institute, Netherlands).
YFP Quenching by Confocal Microscopy
BHK cells stably expressing CFTR and YFP-E2A-mKate were plated 2 days before their use in confocal experiments. For ectopic expression of Venus/YFP-constructs, BHK cells were plated 1 day before transfection. Cells were transfected using polyethylenimine [PEI, linear (MW 25,000); Polysciences, Warrington, PA] as described (19). Two days after transfection BHK cells were washed twice with an iodide-lacking buffer (in mM: 137 NaCl, 2.7 KCl, 0.7 CaCl2, 1.1 MgCl2, 1.5 KH2PO4, and 8.1 Na2HPO4). Cells were stimulated with forskolin (25 μM) and genistein (50 μM) or VX-770 (10 μM) for 20 min and placed in a buffer containing iodide (similar buffer but NaCl is replaced by equimolars of NaI). At least 3 s were recorded to establish a baseline prior the addition of NaI-containing buffer. YFP and dsRed were excited using a 25- to 40-mW argon laser (488 nm, at 2.8% power of low intensity setting), and mKate was excited using 50-mW diode-pumped solid-state laser (561 nm at 0.6% laser power). Pixel dwell time was 1.27 μs, and two scans were averaged. Changes in fluorescence intensity were monitored by a Zeiss LSM 710 confocal microscope via a spectral detector set to measure from 488 to 728 nm with 9-nm intervals; images were recorded every 1.5 s.
Data Analysis for Confocal Microscopy
ZEN 2009 (Carl Zeiss MicroImaging GmbH, Jena, Germany) software was used for the separation of YFP and mKate spectra. For YFP, the region 493 to 551 nm and, for mKate, the region 619 to 709 nm were used. Per condition, 20 randomly selected cells were used to calculate the rate of YFP/mKate quenching. The YFP/mKate values were calculated per cell in time and averaged per experimental condition. For rate of iodide influx calculation, GraphPad Prism (version 5.03 for Windows) was used to fit a one-phase decay curve through the YFP/mKate values. Rate constant K was used as measure of iodide influx.
YFP Quenching by Flow Cytometry
BHK cells were plated 1 day before transfection using polyethylenimine as transfection reagent with Venus/YFP-constructs and or CFTR constructs. Two days posttransfection, the BHK cells were suspended using PBS containing 5 mM EDTA and washed twice in a buffer containing or lacking iodide and finally resuspended in the buffer also used for confocal experiments supplemented with forskolin (25 μM) and genistein (50 μM). After 10 min, the cells were transferred to a buffer with or without iodide, and the changes in fluorescence were monitored in time by a BD FACSCanto II flow cytometer equipped with FACSDiva software for acquisition and data analysis. YFP and dsRed were excited using the 488-nm solid-state laser (20 mW), and mKate was exited using the 633-nm HeNe laser (17 mW).
Data Analysis for Flow Cytometry
Cells were selected based on forward and side scatter plots, and nontransfected cells were excluded based on dsRed or mKate fluorescence, measured in PE (585/42 nm) or APC (660/20 nm) channels, respectively. YFP fluorescence was recorded in the FITC channel (530/30 nm). YFP/dsRed or YFP/mKate ratio's are calculated and monitored using the FACSDiva software. Before each influx/efflux measurement each sample was measured to obtain a baseline measurement. For ratiometric measurements, data were pooled per 5 s and the change in fluorescence over the first 15 s was used as measure for CFTR activity.
Western Blot Analysis
Western blot analysis was performed as described previously (20), the monoclonal antibody L12B4 (Chemicon International, Temecula, CA) was used to detect CFTR. Rabbit polyclonal HSP90 antibody was purchased from Ineke Braakman (Utrecht University, Netherlands). Neomycin phosphotransferase II was detected by using a rabbit polyclonal antibody (Merck Millipore, Billerica, MA). Primary antibodies were detected using polyclonal rabbit (DAKO, Glostrup, Denmark) or goat horseradish peroxidase-conjugated secondary antibody (Thermo Fisher Scientific, Rockford, IL). Visualization was performed using enhanced chemiluminiscence (GE Healthcare Europe GmbH, Freiburg, Germany) detected by x-ray films (Fuji Medical X-ray film, Tokyo, Japan).
Iodide-sensitive YFP has been used extensively to measure CFTR function. One disadvantage of this system is that YFP fluorescence intensity is both dependent of YFP expression level and intracellular iodide concentrations. We reasoned that co-expression of a halide-insensitive fluorescent protein from the same expression vehicle would act as internal control and allow for normalization of sensor expression level during measurements (schematically represented in Fig. 1A). In this approach, the ratio between YFP and another fluorescent protein is assayed upon substitution of a chloride-rich assay buffer for an iodide-rich buffer (influx rate) and subsequently for chloride-rich buffer (efflux rate). To measure CFTR-mediated iodide transport, cells are preincubated with specific agonists such as forskolin, which raises cyclic AMP to induce protein kinase A-mediated CFTR gating, and genistein, which potentiates CFTR channel activity.
We first selected a correct format for co-expression of two fluorescent proteins. We analyzed fluorescent intensities and subcellular distribution of Venus (a bright version of YFP) and dsRed upon co-expression of these proteins as a linked fusion protein or as separate entities by replacing the linker peptide in between Venus and dsRed for an autocleaving E2A peptide. We observed increased Venus intensities when both proteins were separately expressed upon E2A peptide-mediated cleavage (Fig. 1B, upper panel). When Venus and dsRed are physically linked probably, there is energy transfer between them. The fusion protein also formed aggregates more frequently upon ectopic expression in cells (Fig. 1B, lower panel). We therefore continued with an E2A cleavable peptide as optimal expression format.
The fluorescent emission spectra of Venus and dsRed when both excited at 488 nm are somewhat overlapping, which complicates their individual measurement in confocal microscopy (Fig. 1B, upper panel). Therefore, we replaced dsRed for mKate, a bright far red fluorescent protein that is optimally excited at 561 nm and displays a large Stokes shift (Fig. 1C). We engineered halide-sensitive versions of Venus and YFP by introducing H148Q and I152L mutations as previously described (9). We initially compared the fluorescent spectra of halide-sensitive Venus-E2A-mKate and YFP-E2A-mKate upon incubation of cells with increasing iodide concentrations in CFTRwt-expressing cells in the presence of CFTR activators (Fig. 1D). As expected, we observed that mKate fluorescence is unaffected by iodide concentration but that YFP is sensitive to iodide. As was previously indicated, we also observed that YFP was more sensitive to iodide than Venus (18). It is important to indicate that the apparent clear distinction between the YFP and mKate signals results from our experimental setup. We simultaneously analyze the entire emission spectra of the cells during our measurement, but emitted light is specifically blocked at the excitation wavelength (561 nm) for mKate. The small iodide-sensitive peak around 575 nm thus results from YFP. Together, we concluded that intracellular iodide could be most optimally measured by confocal microscopy using a dual-wavelength ratiometric sensor consisting of YFP and mKate separated by a cleavable peptide.
To measure CFTR-dependent iodide influx, BHK cells expressing CFTRwt (BHK-CFTRwt) or CFTR-F508del (BHK-F508del) were lentivirally transduced with YFP-E2A-mKate to generate stable sensor-positive lines. BHK-CFTRwt cells were stimulated with forskolin and the CFTR potentiator genistein for 20 min in a chloride-containing buffer. Upon replacement of chloride for iodide, we observed a rapid decline in YFP/mKate ratio in time (Fig. 2A). We quantified these responses by measurement of the specific YFP and mKate signals in our emission spectrum (Supporting Information Fig. S1A). Forskolin in combination with genistein both stimulated rapid influx of iodide in BHK-CFTRwt cells (Fig. 2B; Supporting Information Fig. S1B shows responses using VX-770 to potentiate CFTR). We next assessed whether CFTR-F508del correction could be measured using our fluorescent sensor. Upon buffer exchange, no significant iodide influx was observed when BHK-F508del cells were unstimulated, but upon stimulation with forskolin and genistein, YFP/mKate ratios clearly decreased, which was more prominent when cells were preincubated with the CFTR corrector VX-809 (Figs. 2C and 2D, Supporting Information Fig. S1C indicates the individual responses of each experiment). Iodide-mediated YFP quenching in CFTR-F508del expressing cells was observed using forskolin and VX-770, which was stronger upon preincubation with VX-809 (Supporting Information Figs. S1D and S1E). We observed a linear relation between the YFP and mKate fluorescent intensities, indicating that expression of both proteins was similar, independent of sensor expression level (Supporting Information Figs. S2A and S2B). We also did not find a correlation between YFP/mKate ratio before iodide addition and YFP or mKate expression (Supporting Information Fig. S2C). Furthermore, we found no correlation between the YFP quenching rate and the initial YFP/mKate ratio (Supporting Information Fig. S2D). Together, these data indicated that wt and mutant CFTR function can be assessed using a dual wavelength ratiometric sensor by confocal microscopy.
In standard halide-quenching assays, differences in YFP expression between the experimental units are normalized by measurement of baseline YFP values before addition of iodide. Using ratiometric measurement, normalization for YFP expression level can be performed during the experiment that can lead to improved accuracy of CFTR function measurement depending on the experimental set up. Within our data set of VX-809-treated BHK-F508del cells stimulated with forskolin and genistein, we compared CFTR-dependent influx rates as measured by the decline in YFP levels or decline in YFP/mKate ratio, either with or without normalization for baseline values. As expected, the average decline in YFP levels and YFP/mKate levels were comparable, independent of normalization for baseline values (Fig. 3). Normalization for expression levels either by baseline correction (F/F0) or by ratiometry reduced the standard deviations of our measurement to similar levels. However, we did observed a slight reduction of 4% of standard deviations in our measurements when ratiometric data were also normalized for baseline when compared with baseline-corrected YFP measurements (n = 5, P < 0.05). This indicates that ratiometric measurement corrects for variability in sensor expression level and can, to a minor extent, improve accuracy when data are expressed relative to values before addition of iodide.
Our previous data suggest that ratiometric measurement is particularly suited to study CFTR function when sensor expression levels cannot be normalized for baseline values. We therefore assessed whether our ratiometric sensor would be suited to study CFTR function using flow cytometry, which is a preferred system for analysis of CFTR function in suspension cells and allows for easy separation and quantification of the fluorescent signals. However, as each cell is lost upon measurement, correction for baseline sensor levels is more difficult.
Using flow cytometry, we analyzed iodide responsiveness of our different sensors upon transfection in BHK cells (Fig. 4A). Again, the YFP-E2A-mKate was most sensitive to iodide, but Venus co-expressed with either mKate or dsRed demonstrated a more linear dependency on iodide compared with YFP. Hereafter, we measured the iodide influx rates in BHK-CFTR cells ectopically expressing the Venus-E2A-mKate. We selected cells based on FSC and SSC and expression of mKate (30–50% of transfection efficiency) and analyzed the Venus/mKate ratio in time upon addition of an iodide-rich buffer during online measurement (Fig. 4B). For BHK-CFTRwt cells, we observed a rapid iodide influx that was absent in cells expressing F508del (Fig. 4C). The highly sensitive YFP sensor was not optimally suited for flow cytometry due to rapid quenching of YFP after addition of an iodide-rich buffer to the cells, and the lag time of approximately 5 sec before measurements could be started (Supporting Information Fig. S3). Therefore, all following experiments were performed using the Venus-based fluorescent sensor.
We further validated CFTR dependency of YFP-quenching in this approach by preincubation with the specific CFTR inhibitor CFTRinh-172. We observed that iodide-mediated YFP quenching was inhibited by CFTRinh-172 in a dose-dependent manner (Fig. 4D). To assess the sensitivity of this assay, BHK cells stably expressing CFTRwt or F508del were mixed in different ratios, and YFP quenching was monitored in time (Fig. 4E). We could clearly distinguish 10% of CFTRwt cells in F508del background using this method. These data demonstrated the ability of our novel fluorescent sensor to measure CFTR activity by flow cytometry.
We next assessed iodide influx/efflux rates of BHK cells expressing different CFTR mutants belonging to various classes of mutations that exhibit different residual activity. We co-transfected the fluorescent sensor with different CFTR constructs in a 1:10 ratio to ensure that cells transfected with the fluorescent sensor were also transfected with CFTR. Gating strategy was similar as shown in Figure 4B. As expected, cells transfected with CFTRwt showed a rapid iodide influx and efflux, whereas CFTR-F508del responded at rates comparable with empty vector-transfected BHK cells. CFTR-G551D showed an intermediate response after stimulation with forskolin and genistein. Interestingly, the CFTR-R117H displayed iodide influx rates similar to CFTRwt, whereas the efflux rates were smaller than for CFTRwt (Fig. 5A, left panel). From these data, we also determined CFTR function by analysis of YFP data alone (Fig. 5A, right panel). As expected, larger variation was present in this data set, preventing accurate measurement on CFTR function. This shows the importance of the ratiometric measurements when cells (or experimental units) cannot be monitored in time.
As controls, Western blots were prepared to validate expression of our constructs (Fig. 5C). YFP-quenching was only observed in BHK cells expressing CFTRwt or CFTR-mutant proteins, which also showed expression of the complex-glycosylated CFTR C-band, which is associated with apical membrane expression of CFTR and residual function (21). F508del is not properly folded and retained within the ER membrane causing expressing of solely the core-glycosylated CFTR B-band that is rapidly degraded. The G542X mutation contains a premature stop codon, which is located before the glycosylation site and migrates as a single band at a lower molecular weight. HSP90 expression was used to confirm equal loading. Neomycin phosphotransferase II that is expressed from the same expression vector as CFTR was used as transfection control. In conclusion, these data show that ratiometric measurements are essential to accurately measure CFTR function by YFP quenching using flow cytometry.
The purpose of this study was to develop a novel approach to measure CFTR function by flow cytometry. Therefore, we adapted a halide-sensitive YFP quenching assay in which we co-expressed a nonhalide-sensitive fluorescent protein to correct for sensor expression level. Here, we used it to study CFTR-mediated iodide transport and found similar properties for our sensors and the previously published YFP-based sensors. This novel ratiometric sensor is ideally suited to study CFTR function when sensor expression levels cannot be corrected for at the level of individual experimental units.
We described two critical properties of the ratiometric sensor for optimal measurement of iodide transport. First, physical attachment of two fluorescent proteins can lower the sensitivity of the YFP probe as we observed higher YFP fluorescent levels for YFP-E2A-dsRed than for YFP-linker-dsRed. This is likely due to Förster resonance energy transfer (FRET) between YFP and dsRed when the molecules are in close proximity due to physical linkage. The potential drawback of separating the two fluorescent proteins is that their levels may change upon expression due to differences in turnover rates. However, since our measurements are short, it is unlikely that the ratio between YFP and mKate is altered during the experiment, and therefore we favor to separate the proteins to increase the fluorescence intensity of YFP.
Second, it is critical that the fluorescent signal that is used for normalization is completely insensitive to iodide and thus also fully separated from YFP-derived signals. Otherwise, the ratiometric value would loose sensitivity compared with the YFP signal alone. We found increased difficulty to separate the YFP from dsRed as compared with mKate using confocal microscopy and propose that the latter would be optimal for confocal microscopy due to the possibility to perform dual wavelength excitation and the larger differences in emission spectra. For flow cytometry, optimal compensation is required to separate YFP from dsRed signals, but this can be easily accomplished.
Our sensor is especially suited to measure CFTR-dependent iodide influx when correction for sensor expression level cannot be performed, as we showed in a newly developed flow cytometric approach to analyze CFTR function. Our flow cytometric approach allowed us to quickly investigate CFTR dependent influx or efflux rates in transient expression experiments and relied on ratiometric normalization for accurate measurement. Compared with confocal microscopy, emission spectra were easier to separate, and fluorescent intensities were easily quantified. The major pitfall is a lag time in fluorescent measurement when the chloride buffer is changed for iodide buffer that results in a less clearly defined influx rate that probably leads to underestimation of cells that display high CFTR activity. This can be compensated by using a less iodide-sensitive sensor as shown here or by using a setup in which the iodide rich buffers can be added directly during online measurements. This may explain the relative high activity of CFTR-G551D (impaired gating) and CFTR-R117H (impaired conductivity) that we observed. For CFTR-R117H, however, it has previously been indicated to mediate currents comparable with CFTRwt (22). Intriguingly, our data would suggest that iodide import through CFTR-R117H is quicker as compared with iodide export through CFTR-R117H.
By studying YFP quenching rates of individual cells that displayed high or low sensor expression levels or high or low YFP/mKate ratios, we could validate that differences in sensor expression levels or quenching status before iodide addition do not impact iodide-induced CFTR-mediated quenching. By flow cytometry, we could also select high and low sensor-expressing cells and found identical CFTR-dependent quenching rates in cells stably expressing CFTRwt (Supporting Information Fig. S4). These data further validated the reliability of CFTR function measurements using genetic YFP-based sensors.
Ratiometric values correct for differences in sensor expression level between expression units in a specific experimental setup during the experiment and thereby reduce experimental variability and improve accuracy. Other dyes such as SPQ (6-methoxy-N-(3-sulfopropyl)quinolinium) (23, 24), LZQ (7-(beta-D-ribofuranosylamino)-pyrido[2,1-h]-pteridin-11-ium-5-olate) (25), and DIBAC4(3) (bis-(1,3-dibutylbarbituic acid)trimethine oxonol) (26, 27) have been used to study CFTR function using plate-bound assays. Both SPQ and LZQ can be quenched by iodide like YFP, whereas DIBAC4(3) is used to measure membrane potential, which changes upon activation of ion channels. A big advantage of these dyes over genetic sensors is that these can be easily added to cells to allow measurements. As loading and leakage efficacy between individual cells cannot be controlled using these single dyes, we expect that genetic sensors like our YFP/mKate sensor to be more ideal to detect small changes at low CFTR activity conditions using flow cytometry.
Ratiometric measurements to quantify intracellular chloride have been previously used. This sensor called Clomeleon specifically uses FRET to measure chloride concentration (28, 29). For flow cytometry, this sensor appears less suited as it depends on a 456-nm laser that is not present on most flow cytometers, although this approach may also be adapted for flow cytometry using a different FRET donor.
In conclusion, we designed a novel fluorescent sensor to measure CFTR function that allows for correction of expression levels between experimental units based on the already published halide sensitive YFP. Compared with the known YFP sensor, it is especially suited for applications in which expression level cannot be normalized as we show in a newly developed YFP quenching assay using flow cytometry. Our data indicate that our flow cytometric assay can be an important tool for analysis of CFTR function in suspension cells such as immune cells or yeast cells that can be used to purify functional CFTR protein (12, 30). In addition, our flow cytometric approach may potentially be used to sort cells with high CFTR function from heterogeneous cell mixtures or assess CFTR function in various cell types simultaneously by including cell surface markers.