Expand this Topic clickable element to expand a topic
Skip to content
Optica Publishing Group
Journal Home About Issues in Progress Current Issue All Issues Feature Issues

Influence of FRET and fluorescent protein maturation on the quantification of binding affinity with dual-channel fluorescence cross-correlation spectroscopy

Varun K. A. Sreenivasan, Matthew S. Graus, Rashmi R. Pillai, Zhengmin Yang, Jesse Goyette, and Katharina Gaus

Open Access Open Access

Abstract

Protein-protein interactions at the plasma membrane mediate transmembrane signaling. Dual-channel fluorescence cross-correlation spectroscopy (dc-FCCS) is a method with which these interactions can be quantified in a cellular context. However, factors such as incomplete maturation of fluorescent proteins, spectral crosstalk, and fluorescence resonance energy transfer (FRET) affect quantification. Some of these can be corrected or accounted for during data analysis and/or interpretation. Here, we experimentally and analytically demonstrate that it is difficult to correct the error caused due to FRET when applying dc-FCCS to measure binding affinity or bound molecular concentrations. Additionally, the presence of dark fluorescent proteins due to incomplete maturation introduces further errors, which too cannot be corrected in the presence of FRET. Based on simulations, we find that modalities such as pulse-interleaved excitation FCCS do not eliminate FRET-induced errors. Finally, we demonstrate that the detrimental effect of FRET can be eliminated with careful experimental design when applying dc-FCCS to quantify protein-protein interactions at the plasma membrane of living cells.

© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement

1. Introduction

Biophysical methods enable investigating protein-protein interactions, which are indispensable cellular processes, with minimal perturbation in a complete cellular context. Amongst these methods, fluorescence cross-correlation spectroscopy (FCCS) promises to robustly quantify binding affinity of these interactions. In its original form [1], the dual-channel FCCS (dc-FCCS) measures the coupled diffusion (or co-diffusion) of two molecules as the signature of a bound state. The extent of coupled- and uncoupled- diffusion of two spectrally distinct fluorescently labelled molecules through a diffraction-limited focal volume is monitored by fluorescence excitation with two lasers and confocal fluorescence detection with two sensitive detectors. The concentrations of the two molecular species are then estimated from the amplitude of autocorrelation, while cross-correlation between the fluorescence detected by the two detectors can be used to estimate the extent of binding or the binding affinity.

Despite the capabilities and promise of FCCS, its application to quantify binding has mostly been reported by laboratories specialized in FCCS, regardless of the availability of commercial microscopes with built-in FCCS modules [2–9]. One of the reasons that hinder wider uptake could be the various instrumental and sample factors that cause the measured correlation amplitudes to deviate significantly from the expected theoretical values. This introduces complexities in data curation, analysis, and interpretation. For example, the instrumental factors include the difference in the confocal volumes of the two spectral channels [1], imperfect overlap of the two confocal volumes [10], spectral crosstalk between the channels [11], and optical background and noise. Sample factors include unlabeled molecules (e.g., endogenous proteins) [2], labelled but non-fluorescent molecules [e.g., incomplete maturation of fluorescent proteins (FPs) and photobleached fluorophores] [2], change in fluorescence brightness upon binding [e.g., fluorescence resonance energy transfer (FRET) or quenching] [12,13], photobleaching, anomalous diffusion (e.g. due to domain formation) [14], competitive binding, multivalent binding, homo-oligomerization, etc. Approaches to counter these include advanced FCCS implementations, correction factors, or careful interpretation. For example, models have been developed to correct for the factors including confocal volume, non-fluorescent molecules, and homo-oligomerization all at once [15]. Advanced implementations of FCCS include single-wavelength FCCS (SW-FCCS) [16] and two-photon FCCS [12], which use a single laser to excite both fluorophores to ensure perfectly overlapped confocal volumes in the two detection channels; and pulse-interleaved excitation-FCCS (PIE-FCCS) [17] and alternating laser excitation-FCCS (ALEX-FCCS) [8,18], which eliminate spectral crosstalk. On the other hand, statistical filtering-based FCCS methods, such as single-color fluorescence lifetime cross-correlation spectroscopy (sc-FLCCS), combine these two advantages by requiring a single laser and by allocating the fluorescence to the two channels based on photon arrival time instead of spectral filtering [13,19]. Correction methods to counter spectral bleed-through and background have also been well-accepted [11]. Anomalous diffusion can be studied with approaches such as focal size-varying FCS [14,20]. On the other hand, countering photophysical factors such as photobleaching, quenching, and FRET requires careful consideration of fluorophores, sample preparation, and data-acquisition parameters. Crucially, in practice, many of these parameters occur simultaneously and further complicate the analysis.

In this article, we thoroughly analyze the artefacts that occur due to FRET and incomplete FP maturation on the correlation amplitudes, which lead to incorrect binding affinity estimates. FRET is known to affect the correlation amplitude [21], as has been discussed in detail in the context of protease-based cleavage of tandem FPs [12,22]. FRET introduces non-linearity in the relationship between the cross-correlation amplitude and the concentration of bound molecules, which allowed the measurement of molecular brightness in addition to concentrations. However, such an approach is only possible if the concentrations of the two cleaved molecular species are equal due to the nature of the experiment. In comparison, the detrimental effect of FRET in the context of applying dc-FCCS to measure binding affinity has not yet been demonstrated or analytically quantified.

Here, we experimentally demonstrate non-linear relationships between the correlation amplitudes and the molecular concentrations in a system containing green and red FPs freely diffusing in two dimensions on a supported lipid bilayer (SLB) membrane, representative of cell membranes. SLBs allowed us to control the molecular concentrations and eliminated artefacts due to competitive binding by endogenous molecules. Additionally, by ensuring that errors from other factors such as spectral crosstalk, homo-oligomerization, confocal volume overlap, and fluorescence background were corrected or are negligible, we found that the source of non-linearity was FRET. Additionally, we found that when FPs were used to label molecules, their incomplete maturation further added to the nonlinearity and increased the error while estimating molecular concentrations and binding affinity. Using simulated data, we show in our model samples that PIE-FCCS offered a significant improvement over dc-FCCS to overcome FRET, but it did not fully eliminate errors. Finally, we demonstrate how careful experimental design can help eliminate the detrimental effect of FRET in dc-FCCS experiments to quantify the binding of G-protein coupled receptors to heterotrimeric G-proteins.

2. Theory

2.1 Setup: correlation amplitude in ideal conditions

A detailed discussion of the theory of dc-FCCS can be found elsewhere [1,12]. The autocorrelation and cross-correlation between the detected fluorescence intensity time traces are defined as:

$${G_{i \times j}}(\tau )= \frac{{\left\langle {\delta {I_i}(t)\delta {I_j}(t + \tau )} \right\rangle }}{{\left\langle {{I_i}} \right\rangle \left\langle {{I_j}} \right\rangle }}$$
where, ${G_{i \times j}}(\tau )$ is the normalized average correlation between the fluorescence intensities ${I_i}(t)$ in the channel i and ${I_j}(t + \tau )$ in the channel j, τ-seconds later. $\left\langle { {} \rangle } \right.$ denotes time averaging with respect to t and assumed to represent ensemble averaging. Channels i and j in most dc-FCCS setups are green and red channels, i.e., $i,j \in \{{g,r} \}$. In an ideal dc-FCCS experiment in which the correlation decays temporally with τ only because of 2D diffusion, the amplitude of the autocorrelation in the channel i is given as [1,12,23]:
$$\mathop {\lim }\limits_{\tau \to 0} {G_{i \times i}}(\tau )= {G_{i \times i}}(0 )= \frac{{\sum\nolimits_k {\left\langle {{C^k}} \right\rangle } {\eta _i}^{{k^2}}}}{{{A_i} \times {{\left( {\sum\nolimits_k {\left\langle {{C^k}} \right\rangle } {\eta_i}^k} \right)}^2}}}$$
where, $\left\langle {{C^k}} \right\rangle$ is the ensemble average of 2D molecular density of the species k and ${\eta _i}^k$ is its molecular brightness in the i-th channel. Ai is the confocal detection area in the i-th channel, defined as ${A_i} = \pi \omega _i^2$, where ωi is the 1/e2-radius of the gaussian confocal area. In a sample, where two molecular species G and R, which are labelled with green and red fluorophores respectively, co-exist with their bound state, GR, the autocorrelation in green and red channels can be derived from Eq. (2) as:
$${G_{g \times g}}(0 )= \frac{{\left\langle {{C^G}} \right\rangle \eta _g^{{G^2}} + \left\langle {{C^{GR}}} \right\rangle \eta _g^{G{R^2}}}}{{{A_g} \times {{\left( {\left\langle {{C^G}} \right\rangle \eta_g^G + \left\langle {{C^{GR}}} \right\rangle \eta_g^{GR}} \right)}^2}}}$$
$${G_{r \times r}}(0 )= \frac{{\left\langle {{C^R}} \right\rangle \eta _r^{{R^2}} + \left\langle {{C^{GR}}} \right\rangle \eta _r^{G{R^2}}}}{{{A_r} \times {{\left( {\left\langle {{C^R}} \right\rangle \eta_r^R + \left\langle {{C^{GR}}} \right\rangle \eta_r^{GR}} \right)}^2}}}$$
Note, that terms $\eta _g^R$ and $\eta _r^G$, the molecular brightness of red fluorophores in the green channel and vice­-versa, have been assumed to be zero due to negligible spectral cross-excitation and bleed-through of fluorescence emission (Fig. 1, Table S4). Similarly, the amplitude of the cross-correlation between the green and red channels is:
$${G_{g \times r}}(0 )= \frac{{d \times \left\langle {{C^{GR}}} \right\rangle \eta _g^{GR}\eta _r^{GR}}}{{{A_{g \times r}} \times \left( {\left\langle {{C^G}} \right\rangle \eta_g^G + \left\langle {{C^{GR}}} \right\rangle \eta_g^{GR}} \right)\left( {\left\langle {{C^R}} \right\rangle \eta_r^R + \left\langle {{C^{GR}}} \right\rangle \eta_r^{GR}} \right)}}$$
where, ${A_{g \times r}}$ is the effective confocal area, defined as ${A_{g \times r}} = ({{A_g} + {A_r}} )/2$, and the factor, d ≤ 1, accounts for non-concentricity of the confocal volumes between the two channels [24]. In reaching Eq. (3a,b) and (4), it is explicitly assumed that binding or unbinding events do not occur while the molecules diffuse through the confocal area (∼0.1-100 ms for lipid-anchored proteins, Fig. 1). When the fluorescence brightness values are assumed to be unaffected by binding (i.e., $\eta _g^G = \eta _g^{GR}$ and $\eta _r^R = \eta _r^{GR}$), Eq. (3a,b) and (4) simplifies. The auto-correlation amplitudes relate to the total concentration of green- and red- fluorescent species, and the cross-correlation amplitude relates to the bound concentration as follows.
$${G_{g \times g,noFRET}}(0 )= \frac{1}{{{A_g} \times \left( {\left\langle {{C^G}} \right\rangle + \left\langle {{C^{GR}}} \right\rangle } \right)}}$$
$${G_{r \times r,noFRET}}(0 )= \frac{1}{{{A_r} \times \left( {\left\langle {{C^R}} \right\rangle + \left\langle {{C^{GR}}} \right\rangle } \right)}}$$
$${G_{g \times r,noFRET}}(0 )= \frac{{d \times \left\langle {{C^{GR}}} \right\rangle }}{{{A_{g \times r}} \times \left( {\left\langle {{C^G}} \right\rangle + \left\langle {{C^{GR}}} \right\rangle } \right)\left( {\left\langle {{C^R}} \right\rangle + \left\langle {{C^{GR}}} \right\rangle } \right)}}$$
The cross-correlation amplitude, when normalized to the autocorrelation amplitude in the green channel, is proportional to the fraction of bound red-labelled molecules [Eq. (6)]. Similarly, the cross-correlation amplitude when normalized to the autocorrelation amplitude in the red channel is proportional to the fraction of bound green-labelled molecules. That is, the normalized cross-correlation amplitudes (NCCAs) are:
$$\frac{{{G_{g \times r,noFRET}}(0 )}}{{{G_{g \times g,noFRET}}(0 )}} = y\frac{{d \times {A_g}}}{{{A_{g \times r}}}} \qquad \textrm{or} \qquad \frac{{{G_{g \times r,noFRET}}(0 )}}{{{G_{g \times g,noFRET}}(0 )}} \propto y\textrm{ }$$
$$\frac{{{G_{g \times r,noFRET}}(0 )}}{{{G_{r \times r,noFRET}}(0 )}} = x\frac{{d \times {A_r}}}{{{A_{g \times r}}}} \qquad \textrm{or} \qquad \frac{{{G_{g \times r,noFRET}}(0 )}}{{{G_{r \times r,noFRET}}(0 )}} \propto x\textrm{ }$$
Where, $x = \left\langle {{C^{GR}}} \right\rangle /\left( {\left\langle {{C^G}} \right\rangle + \left\langle {{C^{GR}}} \right\rangle } \right)$ is the fraction of green-labelled molecules bound to red-labelled molecules, and similarly, $y = \left\langle {{C^{GR}}} \right\rangle /\left( {\left\langle {{C^R}} \right\rangle + \left\langle {{C^{GR}}} \right\rangle } \right)$. In ideal conditions, x and y are directly quantifiable from the dc-FCCS data using Eq. (6a,b, left), where the parameters d, ${A_g}$, ${A_r}$, and ${A_{g \times r}}$ describing the confocal volume geometry are determined by instrumental calibration (Materials and methods) [10]. x and y can also be determined by normalizing the NCCAs to that of a positive control [4,11] to account for factors such as incomplete maturation when FPs or incomplete labelling [Fig. 3(a,b,e,f)], as below.
$$\frac{{\frac{{{G_{g \times r,noFRET}}(0 )}}{{{G_{g \times g,noFRET}}(0 )}}}}{{\max \left( {\frac{{{G_{g \times r,noFRET}}(0 )}}{{{G_{g \times g,noFRET}}(0 )}}} \right)}} = {\left. {\frac{{{G_{g \times r,noFRET}}(0 )}}{{{G_{g \times g,noFRET}}(0 )}}} \right|_{norm}} = y\textrm{ }$$
$$\frac{{\frac{{{G_{g \times r,noFRET}}(0 )}}{{{G_{r \times r,noFRET}}(0 )}}}}{{\max \left( {\frac{{{G_{g \times r,noFRET}}(0 )}}{{{G_{r \times r,noFRET}}(0 )}}} \right)}} = {\left. {\frac{{{G_{g \times r,noFRET}}(0 )}}{{{G_{r \times r,noFRET}}(0 )}}} \right|_{norm}} = x$$
Where, $\max ({{G_{g \times r,noFRET}}(0 )/{G_{g \times g,noFRET}}(0 )} )$ and $\max ({{G_{g \times r,noFRET}}(0 )/{G_{r \times r,noFRET}}(0 )} )$ are the NCCAs for the positive control sample. In effect, the normalization by the positive control accounts for the proportionality constants in Eq. (6a,b), because the x and y values of a positive control sample are unity. If the negative control exhibits a non-zero cross-correlation amplitude, it can also be used for normalizing the NCCAs, depending on the origin of the cross-correlation [4]. In the following section, we describe the effect of FRET on the relationship between NCCAs, x, and y.

 figure: Fig. 1.

Fig. 1. Positive and negative controls for dc-FCCS on SLB. (a,b) Schematic of the SLB system, where biotinylated lipids anchor streptavidin, which binds b-FPs. Tandem labelled b-mCherry-mEGFP was used as the positive control, whereas a mixture of b-mEGFP and b-mCherry was used as the negative control. (c,d) Representative auto- and cross-correlation curves fit to Eq. (10) along with residuals of the fit. Legends in (d) are also applicable to (c).

Download Full Size | PDF

2.2 Influence of FRET and maturation of fluorescent proteins on estimating fractional binding

When present, FRET causes an increase of the molecular brightness of the acceptor (red) at the expense of the donor’s (green), resulting in $\eta _g^{GR}/\eta _g^G = {f_g} \le 1$ and $\eta _r^{GR}/\eta _r^R = {f_r} \ge 1$. Additionally, incomplete maturation of FPs results in factors mG, mR, which are the fractions of green and red fluorophores that are fluorescent. The factor by which the NCCAs are affected by FRET and incomplete maturation is derived in the Supplement 1, using an approach similar to that of Kohl et al. [12]. The final expression is:

$$\frac{{{G_{g \times r,FRET}}(0 )}}{{{G_{g \times g,FRET}}(0 )}} = y\frac{{d \times {A_g}}}{{{A_{g \times r}}}} \times \frac{{{m^G} \times {f_g}{f_r} \times [{1 - x + x({1 - {m^R} + {m^R}{f_g}} )} ]}}{{[{1 - y + y({1 - {m^G} + {m^G}{f_r}} )} ][{1 - x + x({1 - {m^R} + {m^R}{f_g}^2} )} ]}}$$
Similarly,
$$\frac{{{G_{g \times r,FRET}}(0 )}}{{{G_{r \times r,FRET}}(0 )}} = x\frac{{d \times {A_r}}}{{{A_{g \times r}}}} \times \frac{{{m^R} \times {f_r}{f_g} \times [{1 - y + y({1 - {m^G} + {m^G}{f_r}} )} ]}}{{[{1 - x + x({1 - {m^R} + {m^R}{f_g}} )} ][{1 - y + y({1 - {m^G} + {m^G}{f_r}^2} )} ]}}$$
Thus, the effect of FRET on the NCCAs is nonlinear and is dependent on the fraction of binding (x, y), the FRET-induced brightness changes to the acceptor and donor (fr, fg), as well as the FP maturation factors (mG, mR). The maturation factors, which in the absence of FRET are simply proportionality constants, augment the FRET-induced non-linearity and can only partly be corrected using a positive control. The number of factors that the nonlinearity depends on makes it impossible to estimate x and y from the dc-FCCS measurements unless the values of fg, fr, mG, and mR are known a priori.

2.3 Influence of FRET and fluorescent protein maturation on estimating binding affinity

The non-linear dependence of the NCCAs on the bound fractions, x and y, leads to inaccurate affinity quantification. Binding affinity, defined as ${K_A} = \left\langle {{C^{GR}}} \right\rangle /\left\langle {{C^G}} \right\rangle \left\langle {{C^R}} \right\rangle$ can be written in terms of the NCCAs as below ( Supplement 1).

$${K_A} = \frac{{{G_{g \times r}}(0 )\times {A_{g \times r}}}}{{d \times \left( {1 - {{\left. {\frac{{{G_{g \times r}}(0 )}}{{{G_{r \times r}}(0 )}}} \right|}_{norm}}} \right)\left( {1 - {{\left. {\frac{{{G_{g \times r}}(0 )}}{{{G_{g \times g}}(0 )}}} \right|}_{norm}}} \right)}}$$

3. Materials and methods

3.1 Preparation of unilamellar liposomes

18:1 (Δ9-Cis) PC (DOPC), 18:1 DGS-NTA(Ni), 18:1 Biotinyl Cap PE, and 18:1 PEG5000 PE (Avanti Polar Lipids, Inc) were dissolved in chloroform and stocks were prepared by mixing them in the following molar ratio DOPC:DGS-NTA: Biotinyl-Cap-PE: PEG5000-PE::96.5:2:1:0.5. The mixture was stored as small aliquots in brown glass vials at −80°C, after being purged with argon gas. Liposomes were prepared weekly from these stocks as follows. An aliquot was thawed, the chloroform was evaporated under nitrogen gas to obtain a lipid film, and the vial placed overnight in a vacuum chamber. The film was rehydrated with MilliQ water for 0.5-1 hour at room temperature to a final net lipid concentration of 1 mg.mL−1. The solution was vortexed and extruded through a 100 nm pore-size filter using the Avanti Mini Extruder kit to create unilamellar liposomes [25]. The liposomes were stored at 4°C and used within 1 week.

3.2 Preparation of supported lipid bilayer with anchored fluorescent protein

SLBs were prepared on glass coverslips (#1.5, 24 × 60 mm Menzel-Gläser) using a modified version of the protocol in Ref. [26]. Coverslips were cleaned sequentially with acetone, 1-M potassium hydroxide, and absolute ethanol under sonication for 20 minutes each at room temperature. The coverslips were rinsed with MilliQ water between the cleaning steps. The coverslips were dried in an oven and treated using air-plasma (Harrick Plasma). The silicone walls of the resealable 8-well chambers (DKSH) were peeled off the glass slide provided by the manufacturer and attached to the plasma-treated coverslip.

Unilamellar liposome solution containing 1-mg.mL−1 lipid was diluted 5-fold in MilliQ water, to which CaCl2 was added to a final concentration of 10 mM. Immediately, 150 µL of this liposome mixture was added to each of the wells and incubated for 10-20 minutes at room temperature (RT) in the dark to allow the formation of planar lipid bilayer at the coverslip-water interface. From here onwards, care was taken to avoid exposure of the bilayers to air and to minimize light-exposure. Note, because of residual solvent volume in the wells all the absolute concentrations listed below are approximate, but the relative concentrations of co-incubated reagents remain robust. The bilayers were rinsed with PBS using a wash-bottle and incubated with 1-mM EDTA along with 10-µM streptavidin prepared in PBS for 10 minutes at RT. Bilayers prepared without EDTA resulted in samples with a non-reproducible density of FPs (measured using FCS), which we attribute to multilamellar planar bilayers (data not shown). The bilayers were rinsed with PBS and incubated with a solution containing biotin-tagged FPs (b-FPs) in PBS together with 1% (w/v) BSA to minimize non-specific binding. The bilayers were rinsed with PBS in preparation for imaging and dc-FCCS measurements. Mobility of the lipids within the bilayers prepared in this way was verified for every batch of extruded liposomes using < 1-µg.mL−1 Alexa Fluor 546-labelled streptavidin instead of streptavidin. Mobile lipids in the SLB resulted in a shimmering appearance when observed using a TIRF microscope.

3.3 Expression and purification of biotin-tagged fluorescent proteins

N-terminal Avitag-mEGFP, Avitag-mCherry, Avitag-dark-mCherry (Y72S mutant [27]), or Avitag-mCherry-mEGFP were cloned into pTRC-HisA between the NheI and HindIII restriction sites. Chemically competent BL21(DE3) E. coli cells (Agilent Technologies) were transformed with the Avitag-FP constructs and grown on ampicillin-containing (50 µg.mL−1) LB agar plates overnight at 37°C. The following day colonies were selected and inoculated into LB media with 50-µg.mL−1 ampicillin and grown in a shaker incubator overnight at 37°C. 10 mL of this starter culture was then inoculated into 1 L of LB media and the cells were grown in a shaker incubator at 37°C until the optical density at 600 nm was 0.6. The temperature was then decreased to 18°C and IPTG (0.5 mM final concentration from a 1-M stock) and biotin (50 µg.mL−1 final concentration) were added to the culture media. The protein was left to induce and biotinylate with endogenous BirA enzyme overnight, after which the cells were pelleted by centrifugation and stored at −80°C until protein extraction and purification was performed. Protein was extracted by thawing cells, resuspending in 50-mM NaH2PO4, 300-mM NaCl, pH 7.5 (2×PBS), sonicating, and pelleting debris by centrifugation at 15000 xg for 15 minutes. The clarified lysate was passed through 2-mL Ni2+-NTA agarose resin in a gravity-fed column, which was then washed with 10 column volumes of 2×PBS, then with 10 mL of 2×PBS with 10 mM imidazole. b-FPs were eluted with 150 mM imidazole pH 7.5. Eluate was concentrated to 0.5 mL with a 30 kDa spin concentrator (Amicon) before a final polishing step of size-exclusion chromatography on a HiPrep 16/60 Sephacryl S-200 HR (GE Healthcare), equilibrated in PBS, was performed. Purified protein was mixed with glycerol to a final concentration of 10% (v/v) and aliquots were stored at −80°C until used.

3.4 Estimating concentrations of biotin-tagged fluorescent proteins

Three measures of the molar concentrations of b-FPs are possible: concentration of the actual protein, the concentration of fluorescent fraction due to incomplete maturation, or the concentration of biotinylated fraction due to incomplete biotinylation of AviTag. The biotinylation concentration was deemed to be the relevant measure of concentration since biotin mediates the attachment of the b-FPs to the bilayer. The concentration of biotinylated dark mCherry (b-d-mCherry) was determined using a biotin-4-fluorescein assay [28]. Briefly, a known amount of streptavidin was mixed with 3 molar equivalents (since streptavidin is tetravalent) of purified b-d-mCherry (estimated concentration), incubated for 30 minutes at room temperature and the remaining biotin-binding sites were measuring using the biotin-4-fluorescein assay. The biotinylation concentration of b-d-mCherry was deduced from the difference between the number of binding sites measured with and without premixing b-d-mCherry. The biotinylation concentrations of all other b-FPs were measured relative to that of the b-d-mCherry using a competitive, bead-based, flow-cytometry assay. To prepare SLBs on silica beads, unilamellar liposomes were diluted in MilliQ water to a final concentration of 0.4 mg.mL−1 and mixed with silica beads (Bang Laboratories Inc., SS05003) at a ratio of 2:1 (v/v). The mixture was incubated for 5 minutes, after which 15-fold volume excess of PBS was added. The beads were rinsed thrice with PBS by centrifugation at 600×g for 1 minute and incubated with 0.5-mM EDTA and 1-mg.mL−1 streptavidin in PBS for 15 minutes at room temperature. The beads were rinsed thrice with PBS by centrifugation. A solution containing b-FPs and 1% BSA, prepared in PBS, was added and incubated for 10 minutes. b-mEGFP, b-mCherry, or b-mCherry-mEGFP were titrated with the known concentration of b-d-mCherry resulting in competitive binding data to deduce their biotinylation concentrations relative to that of b-d-mCherry. The beads were rinsed thrice with PBS by centrifugation and the bead fluorescence was measured by flow cytometry, from which the biotinylation concentrations were estimated ( Supplement 1, Fig. S1, Table S1).

3.5 Bilayer and cell membrane dc-FCCS measurements

dc-FCCS and single-channel fluorescence correlation spectroscopy (sc-FCS) measurements were carried out on a Zeiss Laser Scanning Microscope 780 (Zeiss LSM 780) equipped with a Confocor3 module. CW 488-nm argon-ion and/or 561-nm DPSS lasers were used to excite mEGFP and/or mCherry, respectively, via a 100×, NA 1.4, oil-immersion, plan-apochromat Zeiss objective lens. The fluorescence emission from the focal volume was collected through the same objective lens and directed to the PMTs via a dichroic beam splitter (MBS488 for green only; or MBS488/561 for red-only or for dual channel), tunable spectral filters with transmission in 499-553 nm and 606-695 nm spectral windows for the green and red channels respectively, and a pinhole. The pinhole diameter was set at 1 AU ( = 79 to 80 µm), and its lateral position was optimized daily using calibration fluorescent dyes. The laser powers were set at 0.1 and 0.05 for 488 and 561 lasers, respectively, on the software, which resulted in 180 nW and 800 nW at the objective and gave rise to counts per molecule of 150-350 Hz in the green channel and 250-400 Hz in the red channel for mEGFP and mCherry, respectively. At this excitation setting, both mEGFP and mCherry were operating well below fluorescence saturation (Fig. S2). In comparison, the background count rates were 20-30 Hz in the green and 50-60 Hz in the red channels. Default settings were used for all other acquisition parameters. For dc-FCCS measurements on SLB, focusing was performed by maximizing the photon counts while adjusting the z-position. For dc-FCCS measurements on cells, the basal membrane of the cell was brought into focus in the confocal imaging mode. Measurements were acquired in sets of 15-second runs and saved as *.fcs files. Definite focus functionality of the Zeiss LSM 780 microscope ensured that any axial drift was actively corrected. However, the runs containing discernable changes in the fluorescence intensity due to active drift correction were discarded. Equilibration of the microscope at 37°C (SLB) or 25°C (cells) for approximately 1-2 hours ensured minimal drift. Background fluorescence was acquired with MilliQ water as the sample with the focus a few microns above the coverslip-water interface.

3.6 Bilayer and cell membrane dc-FCCS data analysis

Data were analyzed with the open-source software PyCorrFit 1.0.1 [29]. First, the runs containing abnormal trends, e.g. due to focus drift, were manually discarded with the help of the inbuilt “overlay” view to identify outliers. The correlation curves were averaged over every six runs and background-corrected using the inbuilt function. The corrected curves were fit to the following inbuilt two-component 2D diffusion model with a confocal (gaussian) measurement scheme:

$${G_{i \times j}}(\tau )= \frac{1}{N} \times \left( {\frac{f}{{1 + \frac{\tau }{{{\tau_{D1}}}}}} + \frac{{1 - f}}{{1 + \frac{\tau }{{{\tau_{D2}}}}}}} \right)$$
Note, the equation only includes parameters of the fit. N is the total number of diffusing entities, f is the fraction of entities in the first diffusion component with an effective diffusion coefficient of τD1, while τD2 is the effective diffusion coefficient of the second component. Weighting inversely proportional to the variance of correlation amplitudes in the six runs at each τ-channel was applied using the inbuilt fitting option. The Levenberg-Marquardt algorithm was used for all fits.

3.7 Calibration of the instrument for sc-FCS and dc-FCCS

The confocal volumes in the green and red spectral channels were calibrated using nanomolar aqueous solutions containing carboxylic-acid terminated reference standards of Alexa Fluor 488 and 568 dyes (AF488, AF568, Life Technologies). sc-FCS measurements were acquired in sets of 30-s runs, with laser power adjusted so that the count rates were 8 - 11 kHz in the green and 5 - 21 kHz in the red channels for AF488 and AF568, respectively. The correlation data were fit to the following inbuilt single-component 3D diffusion and triplet model with a confocal (gaussian) measurement scheme:

$${G_{i \times i}}(\tau )= \frac{1}{N} \times \frac{1}{{1 + \frac{\tau }{{{\tau _D}}}}} \times \frac{1}{{\sqrt {1 + \frac{\tau }{{{\rho ^2}{\tau _D}}}} }} \times \left[ {1 + \frac{{T\;\textrm{exp} \left( { - \frac{\tau }{{{\tau_{trip}}}}} \right)}}{{1 - T}}} \right]$$
Note, the equation only includes parameters of the fit. τD is the effective diffusion coefficient, ρ is the structural parameter defining the axial elongation of the ellipsoidal gaussian confocal volume, T is the fraction of diffusing entities in the triplet state, and τtrip is the characteristic triplet-state lifetime. Based on the known diffusion coefficients of AF488 and AF568, the lateral radii of the confocal volumes were estimated to be 250 ± 20 nm (green channel) and 290 ± 20 nm (red channel), which scale appropriately with the wavelengths of the two channels ( Supplement 1, Table S2). The structural parameter, ρ, was 7 ± 3 and 7 ± 2 in the green and red channels, respectively. The concentricity of the green and red confocal volumes in the three dimensions was found to be nearly ideal (d = 1 ± 0.1) based on dc-FCCS measurements on 100-nm TetraSpeck beads (Life Technologies) suspended in MilliQ water ( Supplement 1, Table S3).

3.8 Fluorescence lifetime analysis of biotin-tagged fluorescent proteins

The fluorescence lifetime of mEGFP was measured using a PicoQuant MicroTime 200 fluorescence lifetime imaging system. The output of a 470 nm laser diode was focused into an aqueous droplet containing b-FPs using a 60× 1.2 NA water-immersion objective. The emitted fluorescence was collected in an epi-fluorescence fashion and spectrally filtered using a 520/30 nm filter to remove laser background and focused onto a single-photon avalanche photodiode. The laser power was kept to a minimum to ensure that fluorescence emission from b-mEGFP exhibited a mono-exponential decay following the excitation laser pulse. The lifetime data from b-mEGFP and b-mCherry-mEGFP were fit to Eq. (12a) and (12b), respectively.

$$F(t )= {N_D}\textrm{exp} \left( { - \frac{t}{{{T_D}}}} \right)$$
$$F(t )= {N_D}\textrm{exp} \left( { - \frac{t}{{{T_D}}}} \right) + {N_{DA}}\textrm{exp} \left( { - \frac{t}{{{T_{DA}}}}} \right)$$
where, F(t) is the fluorescence intensity detected at a time t after the excitation laser pulse, TD and TDA are the fluorescence lifetimes of the FRET donor (mEGFP) in isolation or in tandem with an acceptor (mCherry), and ND and NDA are the amplitudes of the corresponding exponential decay functions. The measurement data was exported and fit to the equations using MATLAB with the built-in non-linear least-square solver, lsqcurvefit. TD obtained from fitting the b-mEGFP lifetime data to Eq. (12a) was fixed to be the lifetime of one of the components when fitting the b-mCherry-mEGFP lifetime data to Eq. (12b).

3.9 Cells expressing dual-labelled µ-opioid receptor and fluorescence lifetime analysis

DNA constructs encoding dual-labelled µ-opioid receptor (MOR), mEGFP-MOR-mCherry (receptor flanked by the two FPs) and MOR-mCherry-mEGFP (receptor tagged with the two FPs at the c-terminal end), were synthesized and inserted into pcDNA5/FRT/TO vector by GenScript. DH5α competent E. coli cells (New England Biolabs) were transformed with the dual-labelled MOR constructs and grown overnight on LB agar plates containing 50-µg.mL−1 ampicillin at 37°C. Colonies were selected and inoculated into LB media containing 50-µg.mL−1 ampicillin and grown overnight in a shaker incubator (150-200 rpm) at 37°C. A miniprep was performed to purify the DNA according to the manufacturer’s protocol provided with QIAprep Spin Miniprep Kit (Qiagen) and the constructs were stored at 4°C for transfections.

AtT20 cells containing a Flp-In recombination site (AtT20 Flp-In cells, mouse pituitary adenoma cells, a gift from Prof. Mark Connor, Macquarie University, Australia) were cultured in Dulbecco’s Modified Eagle Media (DMEM) containing high-glucose, 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin (cell culture media). 100-µg.mL−1 Zeocin was added to maintain the Flp-In recombination site. Cells were grown and maintained in a humidified incubator at 37°C with 5% CO2. To generate stable cell lines expressing dual-labelled MOR, the cells were transfected with the respective pcDNA5/FRT/TO construct for recombination at the Flp-In site. For transfection, cells were plated in a 6-well plate at high confluency, grown for 24 hours, and replenished with fresh cell-culture media (without Zeocin). The dual-labelled MOR construct, pOG44 (Thermofisher, 9-fold excess to dual-labelled MOR construct), FuGENE HD (Promega) (4 µL Fugene per 1 µg DNA, for a total DNA concentration of 3.5 µg per 100 µL), and DMEM were thoroughly mixed, incubated at RT for 10 minutes, and added dropwise to the cells. Cells were incubated for 48 hours, after which 150-µg.mL−1 Hygromycin was added to the cell culture media for selection. The selection media (cell culture media + 150-µg.mL−1 Hygromycin) was renewed every three days until colonies were visible. Clones expressing dual-labelled MOR were pooled and cultured for five passages in selection media, after which the cells were cryogenically stored.

For fluorescence lifetime measurements, cells expressing either of the two dual-labelled MOR were plated on cleaned and plasma-treated glass coverslips (#1.5, 24 × 60 mm Menzel-Gläser) attached to 8-well chambers (DKSH) and grown for 24 hours to reach 50% confluence. The cell culture media was replaced with phenol red-free Leibovitz’s L-15 media without any supplements shortly before the measurements. The fluorescence lifetime of mEGFP in cells was measured in the same manner as for b-FP samples described earlier, but with a pinhole (diameter 150 µm) in front of the avalanche photodiode for confocal detection. The decay of mEGFP-fluorescence from mEGFP-MOR-mCherry and MOR-mCherry-mEGFP after the excitation laser pulse were fit to Eq. (13a) and (13b), respectively.

$$F(t )= {N_D}\textrm{exp} \left( { - \frac{t}{{{T_D}}}} \right) + {N_{AF}}\textrm{exp} \left( { - \frac{t}{{{T_{AF}}}}} \right)$$
$$F(t )= {N_D}\textrm{exp} \left( { - \frac{t}{{{T_D}}}} \right) + {N_{DA}}\textrm{exp} \left( { - \frac{t}{{{T_{DA}}}}} \right) + {N_{AF}}\textrm{exp} \left( { - \frac{t}{{{T_{AF}}}}} \right)$$
where, a notation similar to Eq. (12) was used. The addition term TAF is attributed to the lifetime of cell autofluorescence and NAF is its amplitude.

4. Results and discussion

4.1 Establishing a supported lipid bilayer system

SLBs were used as model membranes to experimentally investigate the robustness of dc-FCCS to FRET or incomplete FP maturation while quantifying molecular binding. b-mEGFP, b-mCherry and b-mCherry-mEGFP, were anchored to SLB via streptavidin-linkage (Fig. 1(a,b)). Figure 1(c,d) show representative dc-FCCS data for bound (b-mCherry-mEGFP) and unbound (b-mEGFP & b-mCherry) FPs, which represent positive and negative control samples. 50- to 100-fold excess b-d-mCherry reduced the effective valency of streptavidin to bind b-mEGFP/b-mCherry/b-mCherry-mEGFP to 0 or 1, thus minimizing the possibility of b-mCherry and b-mEGFP binding to the same streptavidin molecule in the negative control sample. In preliminary experiments, the presence of excess b-d-mCherry resulted in reproducible samples with excellent control of b-FP density on the SLB (Fig. S3). Maximum NCCAs obtained from the positive control were measured to be $\max [{{G_{g \times r}}(0 )/{G_{g \times g}}(0 )} ]$ = 0.59 ± 0.07 and $\max [{{G_{g \times r}}(0 )/{G_{r \times r}}(0 )} ]$ = 0.39 ± 0.04. The minimum NCCAs obtained with the negative control were effectively zero at ${G_{g \times r}}(0 )/{G_{g \times g}}(0 )$ = 0.02 ± 0.02 and ${G_{g \times r}}(0 )/{G_{r \times r}}(0 )$ = 0.02 ± 0.02, suggesting negligible spectral crosstalk and negligible interaction between b-mEGFP and b-mCherry either directly or indirectly (Fig. 2(a)). The spectral crosstalk was also separately verified to be negligible ( Supplement 1, Table S4).

 figure: Fig. 2.

Fig. 2. (a) Experimental and simulated NCCAs as a function of fractional bound molecular species on SLBs (x and y). For the models, fg and fr were set at 0.44 and 1.09 (FRET) or 1 and 1 (no FRET), while mG and mR were set at 0.7 and 0.7 for both. The biotinylation concentrations of the samples are shown in Table S5. Data collected from three independent experimental repeats were pooled. (b) The fluorescence lifetime of mEGFP in b-mEGFP and b-mCherry-mEGFP in solution along with respective one- or two-component exponential fits and residuals. A representative of three independent experiments.

Download Full Size | PDF

4.2 Effect of FRET and incomplete FP-maturation on NCCAs

Equation (8) predicts that FRET in combination with incomplete FP maturation results in a nonlinear relationship between the bound molecular fractions (x or y) and NCCAs. To experimentally observe this effect, a mixture of b-mEGFP and b-mCherry was titrated with b-mCherry-mEGFP to mimic various degrees of binding and incubated on the SLBs, on which dc-FCCS measurements were carried out (Table S5, Fig. 2(a)). The nonlinear relationship between the bound fraction (x and y) and the NCCAs as predicted by theory can be observed. To verify if the data fits with the theory, NCCA data was simulated based on Eq. (8). Since the efficiency of FRET in b-mCherry-mEGFP is the most important factor contributing to the non-linearity, we separately estimated FRET efficiency by fluorescence lifetime measurements to be used for simulation. Based on the equation ηFRET = 1-TDA/TD, where TDA = 1.19 ns and TD = 2.72 ns represent the fluorescence lifetimes of the donor in a donor-acceptor pair (DA, b-mCherry-mEGFP) or in isolation (D, b-mEGFP), the FRET efficiency was estimated to be 56 ± 1% (Fig. 2(b)). Therefore, fg in Eq. (8a,b), defined as the fractional molecular brightness of the donor after FRET, is 0.44 ± 0.01, based on which the NCCA data was simulated. Note, fitting the data to the set of Eq. (8a,b) to determine the parameters fr, mG, and mR did not converge. Therefore, these parameters were varied to visually match the simulated data to experimental observation. Based on this, the values of fr, mG, and mR were found to be 1.09, 0.7 and 0.7 (Fig. 2(a)). The maturation of mCherry (mR, the FRET-acceptor) was independently estimated from the fluorescence lifetime data to be 0.689 ± 0.003, based on the equation ${m^R} = ({{N_{DA}}{T_D}/{T_{DA}}} )/[{{N_D} + {N_{DA}}{T_D}/{T_{DA}}} ]$, which is in agreement with the model prediction. This provided further confirmation that Eq. (8a) and (8b) sufficiently explain the non-linearity in the dc-FCCS data. To portray the detrimental effect of FRET on the NCCA, data were simulated with the parameters fg, fr, mG, and mR set to 1, 1, 0.7, and 0.7, respectively (Fig. 2(a)). FRET not only results in a nonlinear relationship between the bound molecular fractions and the NCCAs, but also a reduced slope. Below, we describe how FRET affects the estimation of binding.

4.3 Error in estimating bound molecular fractions, x and y, due to FRET

Normally in dc-FCCS, the NCCAs are assumed to be proportional to the bound molecular fractions, x and y [Eq. (6)] [4]. But, when present, FRET-induced non-linearity leads to erroneous estimation of bound and unbound molecular fractions due to this assumption. Unlike other artefacts affecting dc-FCCS measurements which can be corrected analytically, the non-linearity of Eq. (8) makes it impossible to derive an analytical correction factor to accurately estimate x and y from the measured NCCAs. Moreover, numerically solving for x and y from the auto- and cross-correlation amplitudes requires the prior knowledge of fg, fr, mG, and mR, which may not be always measurable. Nevertheless, it is crucial that the extent of the error introduced by FRET and incomplete FP maturation is known so that measurements can be accordingly interpreted. To quantify and to visualize the error, we computed the NCCA values using Eq. (8) for combinations of x and y values ranging from 0.1 to 1 (or 10% to 100%) to simulate various degrees of binding. As is normally done in dc-FCCS experiments, the bound fractions were estimated from the simulated NCCA values using Eq. (6) and compared to the actual bound fractions in terms of percentage deviation (Fig. 3). First, we considered a scenario where FPs are used as fluorophores in the absence of FRET and calculated the errors in measuring x and y, when the Eq. (6, left) is used without correcting the NCCAs by normalization to a positive control (Fig. 3(a,e)). As can be seen, this can result in significant errors due to incomplete FP maturation. 70% FP maturation (estimated in our SLB experiments) resulted in the underestimation of x and y by 30%. In comparison, when NCCAs are appropriately normalized to a positive control using Eq. (7), the error from incomplete FP maturation is eliminated in the absence of FRET (Fig. 3(b,f)). It is worth noting that the FP maturation factors (mG and mR) should be as consistent as possible between samples and the positive control for this normalization to be meaningful. This can be ensured, for example, by expressing the control and the sample in the same cell system under similar conditions. Next, when FRET equivalent to what was observed in our experiments was added to the simulation, it gave rise to significant errors in estimating the bound fractions. Crucially, the magnitude of the error depended on the binding itself (Fig. 3(c,g)). Underestimation of binding by over 30% can occur when the fraction of mEGFP that is bound to mCherry is less than 30% (x < 0.3). Slight overestimation can also occur in extreme cases when the fraction of mEGFP that is bound to mCherry is greater than 90% (x > 0.9).

 figure: Fig. 3.

Fig. 3. Percentage error in estimating (a-d) x or the fraction of mCherry-bound mEGFP and (e-h) y or the fraction of mEGFP-bound mCherry due to incomplete maturation (a,b,e,f) alone or (c,d,g,h) in the presence of FRET. The simulated NCCAs were either not normalized (a, e) or normalized (b-d, f-h) to that of a positive control. The extent of FRET and maturation are shown in square brackets in the format [fg, fr, mG, mR], where the values represent hypothetical situations (a, b, e, f), experimental observations (c, g), and theoretical expectations for PIE-FCCS (d, h). Dotted diagonal lines indicate equal concentrations of red and green molecular species.

Download Full Size | PDF

Next, we tested if advanced implementations of FCCS can eliminate such errors. For example, PIE-FCCS developed by the Lamb group [17] was proposed to eliminate spectral crosstalk and FRET. This is because alternatingly exciting the green and red fluorophores with short nanosecond-laser pulses and synchronizing the detection allows the emitted photons to be allocated to the correct channel according to the excitation pulse. This strategy fully eliminates spectral crosstalk and the gain in the brightness of the acceptor due to FRET. However, due to the unequal quantum efficiencies of the donor (QE­mEGFP = 0.6) and the acceptor (QE­mCherry = 0.22) fluorophores [30,31], the photons not emitted by the donor due to FRET are not fully compensated by the increased acceptor emission [32]. The fg in the case of PIE was estimated to be 0.65 for b-mCherry-mEGFP, based on the QEs and FRET efficiency of 56%. Data simulated based on these values (Fig. 3(d,h)) show that the underestimation of x and y in PIE-FCCS is improved in comparison to dc-FCCS, but not eliminated. Indeed, if green and red fluorophores with matching QEs are used, PIE-FCCS will theoretically eliminate FRET induced non-linearities.

4.4 Effect of FRET on the estimation of affinity

Next, we estimated how the error in measuring the bound fractions (x and y) affects the estimation of binding affinity, which is often the goal of FCCS experiments. The percentage error in affinity measurement was calculated as the difference between the actual affinity (for 0.1 < x < 0.9 and 0.1 < y < 0.9) and the affinity that would be calculated from the data based on Eq. (9) (Fig. 4). When FPs are used in the absence of FRET, normalizing the NCCAs to positive control effectively eliminates artefacts arising due to incomplete FP maturation. However, not normalizing the NCCAs to a positive control results in significant errors in affinity estimates even in the absence of FRET (equation not shown, Fig. 4(a)). When FRET was introduced to the extent measured in our experiments, underestimation of affinity was found to be non-uniform. Depending on the fractional binding (x and y), the error varied from −35% to −90%. Therefore, the error in affinity estimation depends both on the affinity itself as well as the starting net concentrations of the two molecular species. We then investigated the improvement PIE-FCCS would have over dc-FCCS to estimate binding affinity in the presence of FRET and incomplete FP maturation (Fig. 4(d)). As before, the underestimation was slightly improved, but was still dependent on the binding itself and varied from −35% to −70%. In practice, the errors in underestimating binding manifest itself as increased spread in the data measured across experimental replicates. This is because the extent of underestimation is dependent on x and y, which will vary between cells even when the affinity is constant, because of differences in protein expression levels. The error in estimation of binding affinity caused by lower levels of FRET and incomplete maturation are shown in Supplement 1 (Fig. S4).

 figure: Fig. 4.

Fig. 4. The percentage error in the estimation of affinity due to incomplete FP-maturation and FRET, as a function of x, the fraction of mCherry-bound mEGFP, and y, the fraction of mEGFP-bound mCherry, when affinity is estimated from simulated dc-FCCS data based on Eq. (9). The NCCAs were either not normalized (a) or normalized (b-d) to a positive control. The extent of FRET and maturation are shown in square brackets in the format [fg, fr, mG, mR], where the values represent hypothetical situations (a, b), experimental observations (c), and theoretical expectations for PIE-FCCS (d). Dotted diagonal lines highlight the path where the net concentrations of red- and green- labelled molecular species are equal.

Download Full Size | PDF

4.5 Eliminating FRET in dc-FCCS on transmembrane proteins by design

To demonstrate the elimination of FRET-induced artefacts by employing an appropriate labelling strategy, we consider the application of dc-FCCS to quantify the interactions of G-protein coupled receptors (GPCRs). FCCS has been used to investigate GPCR-dimerization [33,34] as well as the coupling of GPCRs to the heterotrimeric G-proteins [35–37], but the detrimental effect of FRET on the correlation amplitudes are rarely considered in such studies [9]. For example, in investigating the coupling of GPCR to the G-proteins, fluorescently tagging the GPCR at the extracellular N-terminal eliminates FRET between the fluorescent tags in the GPCR and the G-proteins, which could otherwise result in incorrect quantification of binding induced by an agonist (Fig. 5(a), left). Not only is the labelling strategy important in the actual samples, but the design of the positive control is also critical for accurate binding estimates (Fig. 5(a), right). Consider the example of dual labelled µ-opioid receptor (MOR) as the positive control, where tagging the receptor with mEGFP and mCherry on the same intracellular c-terminal results in significant FRET (56% efficiency) (Fig. 5(b)). FRET decreases the NCCAs of the positive control [Eq. (8), Fig. 2(a), Fig. 5(c), top]. The ${G_{g \times r}}(0 )/{G_{r \times r}}(0 )$ of the mEGFP-MOR-mCherry is significantly higher than MOR-mCherry-mEGFP as expected. When NCCAs of samples are normalized to the NCCAs of the positive control, the presence of FRET in the positive control will result in an over-estimation of binding (even when FRET is absent in the sample). Thus, tagging MOR on either terminals separates the mEGFP and mCherry by at least the depth of the plasma membrane, eliminates FRET, and yields the true value of maximum NCCA. We attribute the absence of an increase in ${G_{g \times r}}(0 )/{G_{g \times g}}(0 )$ in mEGFP-MOR-mCherry compared to MOR-mCherry-mEGFP to the reduced maturation of extracellularly expressed mEGFP. While it may be possible to correct the systematic error in estimating molecular binding due to FRET in the positive control alone, FRET being present in the sample will render the data erroneous due to the non-linear relationship between binding and the NCCAs, as described above.

 figure: Fig. 5.

Fig. 5. (a) Schematic of labelling of MOR and G-protein (top) that results in FRET and (bottom) that avoids FRET. (b) The fluorescence lifetime of mEGFP in the two positive controls, mEGFP-MOR-mCherry and MOR-mCherry-mEGFP, along with two- or three-component exponential fits. The fluorescence lifetime of the component attributed to cell autofluorescence is not shown. A representative of two independent experiments. (c) NCCAs of MOR-mCherry-mEGFP and mEGFP-MOR-mCherry obtained from measurements in 19 and 28 cells, respectively, collected across four independent experimental repeats.

Download Full Size | PDF

5. Conclusions

Fluorescence cross-correlation spectroscopy is a promising method to measure molecular interaction in a minimally invasive fashion. Here, we experimentally and analytically demonstrated and quantified the effect of FRET and incomplete maturation of fluorescent proteins on the fluorescence correlation amplitudes. The linear relationship between normalized cross-correlation amplitudes and the bound fraction of molecules is affected by FRET and becomes non-linear. The non-linearity depends on the bound molecular fractions (x and y), FRET-induced fractional change in brightness of the fluorophores (fg and fr), as well as the fraction of fluorescent proteins that are mature and fluorescent (mG and mR). Use of a supported lipid bilayer model enabled us to prepare samples with varying x and y values and allowed the estimation of the FRET and maturation factors, which would not be possible in a real sample with unknown x and y values. The error induced by FRET in estimating the bound molecular fractions can be up to −35% when the FRET efficiency is the order of 50-60%. These errors further propagate to affinity estimates, where an underestimation of up to 90% can occur. Indeed, increased FRET efficiencies and reduced fluorescent protein maturation will cause the underestimation to further accentuate. While our study is described based on a supported lipid bilayer model resembling the plasma membrane of cell, where G-protein coupled receptors and other proteins diffuse in two dimensions, our findings are also applicable to three dimensional systems. Practically, when dc-FCCS is used to measure binding affinity in cells, FRET artefacts will manifest themselves as increased variability in the affinity measured across cells or regions within a cell with varying protein-expression levels. With simulation, we demonstrated that advanced implementations of FCCS, such as PIE-FCCS, can only eliminate the FRET-induced non-linearity if the quantum efficiencies of the green (donor) and red (acceptor) fluorophores are equal. Since we used mEGFP and mCherry with quantum efficiencies of 0.6 and 0.22, respectively, PIE-FCCS would only have resulted in a moderate reduction of error in estimating the affinity. It is crucial to point out that the estimates of the errors shown here are only valid if the FRET and maturation factors remain consistent between the samples and the positive control. If not, this will introduce further errors due to inappropriate normalization. Indeed, it is best to eliminate FRET by employing careful experimental design and labelling strategies. In the case of large proteins or transmembrane proteins, such as GPCRs, placing the two fluorophores on either side of the plasma membrane can eliminate FRET and remove non-linearities in the correlation amplitudes. If fluorescent proteins are used, an appropriate positive control is required to correct for incomplete maturation, but also only in the absence of FRET. Indeed, the use of SLBs enabled us to eliminate or reduce other factors that would normally also influence dc-FCCS measurements. While the instrumental factors such as the imperfect overlap of confocal volumes and background can be accounted for by using established correction factors, sample-related factors such as endogenous proteins, presence of competing interaction partners, and lipid domain formation are generally only qualitatively accounted for during data interpretation.

Funding

Australian Research Council (CE140100011, DE160100888); National Health and Medical Research Council (APP1155162, APP1163814).

Acknowledgements

We acknowledge funding from the Australian Research Council (DE160100888 to VS, CE140100011 to KG) and the National Health and Medical Research Council of Australia (APP1155162 to KG, APP1163814 to JG and KG). The microscopy component of this study was carried out using instruments situated in, and maintained by, the Biomedical Imaging Facility (BMIF) at UNSW. VS and RP thank Dr Alexander McMillan and Dr Elvis Pandezic for training, experimental help, and discussions. VS and RP thank Dr Marina Santiago and Prof. Mark Connor for AtT20 Flp-In cell lines and discussions regarding transfections. VS acknowledges Dr Vincent Brianne and Prof. Sudipta Maiti for discussions.

Disclosures

The authors declare no conflicts of interest.

See Supplement 1 for supporting content.

References

1. P. Schwille, F. J. Meyer-Almes, and R. Rigler, “Dual-color fluorescence cross-correlation spectroscopy for multicomponent diffusional analysis in solution,” Biophys. J. 72(4), 1878–1886 (1997). [CrossRef]  

2. Y. H. Foo, N. Naredi-Rainer, D. C. Lamb, S. Ahmed, and T. Wohland, “Factors affecting the quantification of biomolecular interactions by fluorescence cross-correlation spectroscopy,” Biophys. J. 102(5), 1174–1183 (2012). [CrossRef]  

3. C. I. Maeder, M. A. Hink, A. Kinkhabwala, R. Mayr, P. I. H. Bastiaens, and M. Knop, “Spatial regulation of Fus3 MAP kinase activity through a reaction-diffusion mechanism in yeast pheromone signalling,” Nat. Cell Biol. 9(11), 1319–1326 (2007). [CrossRef]  

4. W. Sadaie, Y. Harada, M. Matsuda, and K. Aoki, “Quantitative in vivo fluorescence cross-correlation analyses highlight the importance of competitive effects in the regulation of protein-protein interactions,” Mol. Cell. Biol. 34(17), 3272–3290 (2014). [CrossRef]  

5. X. Shi, Y. H. Foo, T. Sudhaharan, S. W. Chong, V. Korzh, S. Ahmed, and T. Wohland, “Determination of dissociation constants in living zebrafish embryos with single wavelength fluorescence cross-correlation spectroscopy,” Biophys. J. 97(2), 678–686 (2009). [CrossRef]  

6. B. D. Slaughter, J. W. Schwartz, and R. Li, “Mapping dynamic protein interactions in MAP kinase signaling using live-cell fluorescence fluctuation spectroscopy and imaging,” Proc. Natl. Acad. Sci. U. S. A. 104(51), 20320–20325 (2007). [CrossRef]  

7. T. Sudhaharan, P. Liu, Y. H. Foo, W. Bu, K. B. Lim, T. Wohland, and S. Ahmed, “Determination of in vivo dissociation constant, KD, of Cdc42-effector complexes in live mammalian cells using single wavelength fluorescence cross-correlation spectroscopy,” J. Biol. Chem. 284(20), 13602–13609 (2009). [CrossRef]  

8. S. Ly, R. Altman, J. Petrlova, Y. Lin, S. Hilt, T. Huser, T. A. Laurence, and J. C. Voss, “Binding of apolipoprotein E inhibits the oligomer growth of amyloid-β peptide in solution as determined by fluorescence cross-correlation spectroscopy,” J. Biol. Chem. 288(17), 11628–11635 (2013). [CrossRef]  

9. S. Christie, X. Shi, and A. W. Smith, “Resolving Membrane Protein-Protein Interactions in Live Cells with Pulsed Interleaved Excitation Fluorescence Cross-Correlation Spectroscopy,” Acc. Chem. Res. 53(4), 792–799 (2020). [CrossRef]  

10. T. Weidemann, M. Wachsmuth, M. Tewes, K. Rippe, and J. Langowski, “Analysis of Ligand Binding by Two-Colour Fluorescence Cross-Correlation Spectroscopy,” Single Mol. 3(1), 49–61 (2002). [CrossRef]  

11. K. Bacia and P. Schwille, “Practical guidelines for dual-color fluorescence cross-correlation spectroscopy,” Nat. Protoc. 2(11), 2842–2856 (2007). [CrossRef]  

12. T. Kohl, K. G. Heinze, R. Kuhlemann, A. Koltermann, and P. Schwille, “A protease assay for two-photon crosscorrelation and FRET analysis based solely on fluorescent proteins,” Proc. Natl. Acad. Sci. U. S. A. 99(19), 12161–12166 (2002). [CrossRef]  

13. M. Štefl, K. Herbst, M. Rübsam, A. Benda, and M. Knop, Single-color Fluorescence Lifetime Cross-Correlation Spectroscopy in vivo,” bioRxiv 2020.2001.2023.917435 (2020)

14. M. D. N. Stolle and C. Fradin, “Anomalous Diffusion in Inverted Variable-Lengthscale Fluorescence Correlation Spectroscopy,” Biophys. J. 116(5), 791–806 (2019). [CrossRef]  

15. M. J. Kaliszewski, X. Shi, Y. Hou, R. Lingerak, S. Kim, P. Mallory, and A. W. Smith, “Quantifying membrane protein oligomerization with fluorescence cross-correlation spectroscopy,” Methods 140-141, 40–51 (2018). [CrossRef]  

16. L. Hwang and T. Wohland, “Dual-Color Fluorescence Cross-Correlation Spectroscopy Using Single Laser Wavelength Excitation,” ChemPhysChem 5(4), 549–551 (2004). [CrossRef]  

17. B. K. Muller, E. Zaychikov, C. Brauchle, and D. C. Lamb, “Pulsed interleaved excitation,” Biophys. J. 89(5), 3508–3522 (2005). [CrossRef]  

18. A. N. Kapanidis, N. K. Lee, T. A. Laurence, S. Doose, E. Margeat, and S. Weiss, “Fluorescence-aided molecule sorting: analysis of structure and interactions by alternating-laser excitation of single molecules,” Proc. Natl. Acad. Sci. U. S. A. 101(24), 8936–8941 (2004). [CrossRef]  

19. R. Macháň, P. Kapusta, and M. Hof, “Statistical filtering in fluorescence microscopy and fluorescence correlation spectroscopy,” Anal. Bioanal. Chem. 406(20), 4797–4813 (2014). [CrossRef]  

20. L. Wawrezinieck, H. Rigneault, D. Marguet, and P. F. Lenne, “Fluorescence correlation spectroscopy diffusion laws to probe the submicron cell membrane organization,” Biophys. J. 89(6), 4029–4042 (2005). [CrossRef]  

21. N. L. Thompson, Fluorescence Correlation Spectroscopy (Springer, 2002), pp. 337–378

22. C. Eggeling, P. Kask, D. Winkler, and S. Jager, “Rapid analysis of Forster resonance energy transfer by two-color global fluorescence correlation spectroscopy: trypsin proteinase reaction,” Biophys. J. 89(1), 605–618 (2005). [CrossRef]  

23. E. L. Elson and D. Magde, “Fluorescence correlation spectroscopy. I. Conceptual basis and theory,” Biopolymers 13(1), 1–27 (1974). [CrossRef]  

24. T. Weidemann and P. Schwille, “Dual-color fluorescence cross-correlation spectroscopy with continuous laser excitation in a confocal setup,” Methods Enzymol. 518, 43–70 (2013). [CrossRef]  

25. I. Avanti Polar Lipids. Preparing Large, Unilamellar Vesicles by Extrusion (LUVET), <https://avantilipids.com/tech-support/liposome-preparation/luvet> (

26. S. V. Pageon, T. Tabarin, Y. Yamamoto, Y. Ma, P. R. Nicovich, J. S. Bridgeman, A. Cohnen, C. Benzing, Y. Gao, M. D. Crowther, K. Tungatt, G. Dolton, A. K. Sewell, D. A. Price, O. Acuto, R. G. Parton, J. J. Gooding, J. Rossy, J. Rossjohn, and K. Gaus, “Functional role of T-cell receptor nanoclusters in signal initiation and antigen discrimination,” Proc. Natl. Acad. Sci. U. S. A. 113(37), E5454–E5463 (2016). [CrossRef]  

27. E. M. Schmid, M. H. Bakalar, K. Choudhuri, J. Weichsel, H. S. Ann, P. L. Geissler, M. L. Dustin, and D. A. Fletcher, “Size-dependent protein segregation at membrane interfaces,” Nat. Phys. 12(7), 704–711 (2016). [CrossRef]  

28. G. Kada, H. Falk, and H. J. Gruber, “Accurate measurement of avidin and streptavidin in crude biofluids with a new, optimized biotin-fluorescein conjugate,” Biochim. Biophys. Acta 1427(1), 33–43 (1999). [CrossRef]  

29. P. Muller, P. Schwille, and T. Weidemann, “PyCorrFit-generic data evaluation for fluorescence correlation spectroscopy,” Bioinformatics 30(17), 2532–2533 (2014). [CrossRef]  

30. N. C. Shaner, R. E. Campbell, P. A. Steinbach, B. N. Giepmans, A. E. Palmer, and R. Y. Tsien, “Improved monomeric red, orange and yellow fluorescent proteins derived from Discosoma sp. red fluorescent protein,” Nat. Biotechnol. 22(12), 1567–1572 (2004). [CrossRef]  

31. G. J. Kremers, S. G. Gilbert, P. J. Cranfill, M. W. Davidson, and D. W. Piston, “Fluorescent proteins at a glance,” J. Cell Sci. 124(15), 2676 (2011). [CrossRef]  

32. J. Wlodarczyk, A. Woehler, F. Kobe, E. Ponimaskin, A. Zeug, and E. Neher, “Analysis of FRET signals in the presence of free donors and acceptors,” Biophys. J. 94(3), 986–1000 (2008). [CrossRef]  

33. R. C. Patel, U. Kumar, D. C. Lamb, J. S. Eid, M. Rocheville, M. Grant, A. Rani, T. Hazlett, S. C. Patel, E. Gratton, and Y. C. Patel, “Ligand binding to somatostatin receptors induces receptor-specific oligomer formation in live cells,” Proc. Natl. Acad. Sci. U. S. A. 99(5), 3294–3299 (2002). [CrossRef]  

34. J. Petersen, S. C. Wright, D. Rodriguez, P. Matricon, N. Lahav, A. Vromen, A. Friedler, J. Stromqvist, S. Wennmalm, J. Carlsson, and G. Schulte, “Agonist-induced dimer dissociation as a macromolecular step in G protein-coupled receptor signaling,” Nat. Commun. 8(1), 226 (2017). [CrossRef]  

35. B. Caldwell, A. Ustione, and D. W. Piston, “Dopamine Receptor Signaling in MIN6 beta-Cells Revealed by Fluorescence Fluctuation Spectroscopy,” Biophys. J. 111(3), 609–618 (2016). [CrossRef]  

36. Y. Li, R. V. Shivnaraine, F. Huang, J. W. Wells, and C. C. Gradinaru, “Ligand-Induced Coupling between Oligomers of the M2 Receptor and the Gi1 Protein in Live Cells,” Biophys. J. 115(5), 881–895 (2018). [CrossRef]  

37. S. J. Briddon, L. E. Kilpatrick, and S. J. Hill, “Studying GPCR Pharmacology in Membrane Microdomains: Fluorescence Correlation Spectroscopy Comes of Age,” Trends Pharmacol. Sci. 39(2), 158–174 (2018). [CrossRef]  

Supplementary Material (1)

NameDescription
Supplement 1       Supplement 1

Cited By

Optica participates in Crossref's Cited-By Linking service. Citing articles from Optica Publishing Group journals and other participating publishers are listed here.

Alert me when this article is cited.


Figures (5)
Fig. 1.
Fig. 1. Positive and negative controls for dc-FCCS on SLB. (a,b) Schematic of the SLB system, where biotinylated lipids anchor streptavidin, which binds b-FPs. Tandem labelled b-mCherry-mEGFP was used as the positive control, whereas a mixture of b-mEGFP and b-mCherry was used as the negative control. (c,d) Representative auto- and cross-correlation curves fit to Eq. (10) along with residuals of the fit. Legends in (d) are also applicable to (c).

View in Article | Download Full Size | PDF

Fig. 2.
Fig. 2. (a) Experimental and simulated NCCAs as a function of fractional bound molecular species on SLBs (x and y). For the models, fg and fr were set at 0.44 and 1.09 (FRET) or 1 and 1 (no FRET), while mG and mR were set at 0.7 and 0.7 for both. The biotinylation concentrations of the samples are shown in Table S5. Data collected from three independent experimental repeats were pooled. (b) The fluorescence lifetime of mEGFP in b-mEGFP and b-mCherry-mEGFP in solution along with respective one- or two-component exponential fits and residuals. A representative of three independent experiments.

View in Article | Download Full Size | PDF

Fig. 3.
Fig. 3. Percentage error in estimating (a-d) x or the fraction of mCherry-bound mEGFP and (e-h) y or the fraction of mEGFP-bound mCherry due to incomplete maturation (a,b,e,f) alone or (c,d,g,h) in the presence of FRET. The simulated NCCAs were either not normalized (a, e) or normalized (b-d, f-h) to that of a positive control. The extent of FRET and maturation are shown in square brackets in the format [fg, fr, mG, mR], where the values represent hypothetical situations (a, b, e, f), experimental observations (c, g), and theoretical expectations for PIE-FCCS (d, h). Dotted diagonal lines indicate equal concentrations of red and green molecular species.

View in Article | Download Full Size | PDF

Fig. 4.
Fig. 4. The percentage error in the estimation of affinity due to incomplete FP-maturation and FRET, as a function of x, the fraction of mCherry-bound mEGFP, and y, the fraction of mEGFP-bound mCherry, when affinity is estimated from simulated dc-FCCS data based on Eq. (9). The NCCAs were either not normalized (a) or normalized (b-d) to a positive control. The extent of FRET and maturation are shown in square brackets in the format [fg, fr, mG, mR], where the values represent hypothetical situations (a, b), experimental observations (c), and theoretical expectations for PIE-FCCS (d). Dotted diagonal lines highlight the path where the net concentrations of red- and green- labelled molecular species are equal.

View in Article | Download Full Size | PDF

Fig. 5.
Fig. 5. (a) Schematic of labelling of MOR and G-protein (top) that results in FRET and (bottom) that avoids FRET. (b) The fluorescence lifetime of mEGFP in the two positive controls, mEGFP-MOR-mCherry and MOR-mCherry-mEGFP, along with two- or three-component exponential fits. The fluorescence lifetime of the component attributed to cell autofluorescence is not shown. A representative of two independent experiments. (c) NCCAs of MOR-mCherry-mEGFP and mEGFP-MOR-mCherry obtained from measurements in 19 and 28 cells, respectively, collected across four independent experimental repeats.

View in Article | Download Full Size | PDF

Equations (21)

Equations on this page are rendered with MathJax. Learn more.

G i × j ( τ ) = ⟨ δ I i ( t ) δ I j ( t + τ ) ⟩ ⟨ I i ⟩ ⟨ I j ⟩
lim τ → 0 ⁡ G i × i ( τ ) = G i × i ( 0 ) = ∑ k ⟨ C k ⟩ η i k 2 A i × ( ∑ k ⟨ C k ⟩ η i k ) 2
G g × g ( 0 ) = ⟨ C G ⟩ η g G 2 + ⟨ C G R ⟩ η g G R 2 A g × ( ⟨ C G ⟩ η g G + ⟨ C G R ⟩ η g G R ) 2
G r × r ( 0 ) = ⟨ C R ⟩ η r R 2 + ⟨ C G R ⟩ η r G R 2 A r × ( ⟨ C R ⟩ η r R + ⟨ C G R ⟩ η r G R ) 2
G g × r ( 0 ) = d × ⟨ C G R ⟩ η g G R η r G R A g × r × ( ⟨ C G ⟩ η g G + ⟨ C G R ⟩ η g G R ) ( ⟨ C R ⟩ η r R + ⟨ C G R ⟩ η r G R )
G g × g , n o F R E T ( 0 ) = 1 A g × ( ⟨ C G ⟩ + ⟨ C G R ⟩ )
G r × r , n o F R E T ( 0 ) = 1 A r × ( ⟨ C R ⟩ + ⟨ C G R ⟩ )
G g × r , n o F R E T ( 0 ) = d × ⟨ C G R ⟩ A g × r × ( ⟨ C G ⟩ + ⟨ C G R ⟩ ) ( ⟨ C R ⟩ + ⟨ C G R ⟩ )
G g × r , n o F R E T ( 0 ) G g × g , n o F R E T ( 0 ) = y d × A g A g × r or G g × r , n o F R E T ( 0 ) G g × g , n o F R E T ( 0 ) ∝ y  
G g × r , n o F R E T ( 0 ) G r × r , n o F R E T ( 0 ) = x d × A r A g × r or G g × r , n o F R E T ( 0 ) G r × r , n o F R E T ( 0 ) ∝ x  
G g × r , n o F R E T ( 0 ) G g × g , n o F R E T ( 0 ) max ( G g × r , n o F R E T ( 0 ) G g × g , n o F R E T ( 0 ) ) = G g × r , n o F R E T ( 0 ) G g × g , n o F R E T ( 0 ) | n o r m = y  
G g × r , n o F R E T ( 0 ) G r × r , n o F R E T ( 0 ) max ( G g × r , n o F R E T ( 0 ) G r × r , n o F R E T ( 0 ) ) = G g × r , n o F R E T ( 0 ) G r × r , n o F R E T ( 0 ) | n o r m = x
G g × r , F R E T ( 0 ) G g × g , F R E T ( 0 ) = y d × A g A g × r × m G × f g f r × [ 1 − x + x ( 1 − m R + m R f g ) ] [ 1 − y + y ( 1 − m G + m G f r ) ] [ 1 − x + x ( 1 − m R + m R f g 2 ) ]
G g × r , F R E T ( 0 ) G r × r , F R E T ( 0 ) = x d × A r A g × r × m R × f r f g × [ 1 − y + y ( 1 − m G + m G f r ) ] [ 1 − x + x ( 1 − m R + m R f g ) ] [ 1 − y + y ( 1 − m G + m G f r 2 ) ]
K A = G g × r ( 0 ) × A g × r d × ( 1 − G g × r ( 0 ) G r × r ( 0 ) | n o r m ) ( 1 − G g × r ( 0 ) G g × g ( 0 ) | n o r m )
G i × j ( τ ) = 1 N × ( f 1 + τ τ D 1 + 1 − f 1 + τ τ D 2 )
G i × i ( τ ) = 1 N × 1 1 + τ τ D × 1 1 + τ ρ 2 τ D × [ 1 + T exp ( − τ τ t r i p ) 1 − T ]
F ( t ) = N D exp ( − t T D )
F ( t ) = N D exp ( − t T D ) + N D A exp ( − t T D A )
F ( t ) = N D exp ( − t T D ) + N A F exp ( − t T A F )
F ( t ) = N D exp ( − t T D ) + N D A exp ( − t T D A ) + N A F exp ( − t T A F )
Select as filters
Select Topics Cancel
Top
       
© Copyright 2024 | Optica Publishing Group. All rights reserved, including rights for text and data mining and training of artificial technologies or similar technologies.
Privacy | Terms of Use