2.1. Receptor Densities Determined by dSTORM
We developed a novel and simple approach to determine receptor densities and ligand-binding affinities to membrane receptors on single cells using single-molecule super-resolution imaging. Most membrane receptors are present at copy numbers too high to allow discerning single receptors with conventional diffraction-limited imaging such as confocal laser scanning microscopy. To get an idea on the molecular density of receptors at the plasma membrane, we immunolabelled the target proteins with the photoswitchable dye Alexa Fluor 647, performed dSTORM measurements and counted single spots (in whole cells or large sections of cells imaged) which represent single receptor sites. dSTORM images of MET and TNF-R1 are shown in Figure 1. Our experiments yielded membrane receptor densities in HeLa cells of 6.5±0.6 (0.6=standard deviation, s.d.) receptors per μm2 for MET and 1.3±0.3 (s.d.) receptors per μm2 for TNF-R1. We determined the membrane surface of HeLa cells using confocal microscopy to about 1600±380 (s.d.) μm2 (N=13). With that estimate, we calculated the average number of receptors per cell to be between 7900 and 12 900 (MET) and 1500 to 2900 (TNF-R1), respectively. The number of MET determined here is in reasonable agreement with published data from other cell lines.22–24 For TNF-R1, receptor copy numbers derived from TNFα-binding sites were reported in the range of 128±29 on neuroblastoma cells25 and about 7500 for FS-4 cells.26 In addition, receptor densities determined from super-resolution images can resolve variations between different cells. We observed that the receptor density ranged between 5.5 and 7.1 molecules per μm2 for MET (N=7), whereas we found a range of 0.8 to 1.8 molecules per μm2 for TNF-R1 (N=9). However, we note that our measurements cannot resolve oligomers, for example, dimers in the case of MET27 or dimers and trimers for TNF-R1.28 In this sense, the resulting receptor densities are to be considered as a lower estimate. Receptor densities from these experiments can complement other experimental data, such as from photobleaching experiments27, 29 or quantitative PALM,30–32 where the oligomeric state (dimers, trimers, etc.) and the heterogeneity of oligomeric states can be determined.
Figure 1. dSTORM images of Alexa Fluor 647 immunolabelled membrane receptors. Super-resolution images are shown for a) MET receptor and b) TNF-R1 distribution on the cell membrane of HeLa cells, respectively. The insets show the wide-field images (scale bar 2 μm).
Download figure to PowerPoint
In a second experiment, we determined molecular densities of receptor-specific ligands by quantifying the number of fluorophore-labelled ligands (InlB-ATTO647N, TNFα-ATTO647N) bound to the respective receptor at the single-cell level. Serum-starved HeLa cells were incubated with a fluorophore-labelled ligand and single-molecule imaging was performed. We observed a high degree of glass adsorption for both ligands, which interfered with single-molecule imaging on the cell membrane. We adapted our experimental protocol by staining cells with the respective ligands. After fixation, we gently scraped cells from the surface, followed by washing and seeding cells on a new glass slide (for details, see the Experimental Section). This approach reduced the background fluorescence signal significantly. Cells were incubated at different concentrations of labelled ligand. With single-molecule imaging, we counted the corresponding ligand number bound to receptors on single cells and determined ligand densities at saturating condition by fitting with a model for binding (see the Experimental Section). In the case of InlB binding to MET, we observed ligand densities between 1.25±0.24 (s.d.) (no scraping) and 2.52±0.26 (s.d.) molecules per μm2 (with scraping) (Table 1) under ligand saturation conditions. The lower ligand density in the case of cells that were not scraped can be explained by the higher background fluorescence due to the surface adsorption of the ligand, which interferes with the signal from ligand bound to the receptor. Thus, accurate fitting of fluorescent spots during the data analysis is impaired. For InlB only the MET receptor was found as receptor.20 The low receptor occupancy by InlB-ATTO647N can also be partly explained by preferential binding of the unlabelled ligand species to the receptor.
Table 1. Receptor and ligand densities on cells. For the determination of ligand densities, the number of InlB-ATTO647N and TNFα-ATTO647N bound to induced cells was determined. Standard deviations are given for each value.
|Receptor||Receptor density [μm−2]||Ligand density [μm−2]||Max. receptor occupation [%]|
|MET||6.5±0.6||1.25±0.24[a] 2.52±0.26[a,b]||19±7 39±5|
In order to determine the biological activity of TNFα-ATTO647N, a NF-κB reporter gene assay was performed (see the Supporting Information). Stimulation of U251 astroglioma cells with both native and ATTO647N-labelled TNFα resulted in significantly increased NF-κB-activity compared to untreated controls (Figure S1). No significant differences in biological activity between ATTO647N-labelled and native TNFα were observed. For TNF-R1, we found a ligand density of 0.45±0.0004 (s.d.) molecules per μm2 (with scraping) at saturation levels. We note that TNFα also binds to TNF-R2, however, the number of TNF-R2 receptors (compared to TNF-R1) in HeLa cells is negligibly small,33, 34 such that we do not further consider it in the following analysis.
We compared the numbers of receptors to the numbers of ligands bound under saturating conditions, which should give information on the average maximal occupation density. We found that the receptor occupation ranges between 19±7 (s.d.) % (no scraping) and 39±5 (s.d.) % (with scraping) for MET. TNF-R1 is occupied with 35±8 (s.d.) % by TNFα-ATTO647N. The low occupation levels could be accounted for by the use of polyclonal antibodies, which might have more flexibility in binding to the receptor than a corresponding ligand. Other reasons might be that a fraction of fluorophores (attached to the ligands) already resides in a non-fluorescent dark state or photobleaches before the beginning of the measurement. Also, a special receptor conformation or oligomerization state might be a prerequisite for ligand binding. However, this requires more experimental evidence and careful controls and is beyond the scope of this study. The receptor and ligand densities as well as resulting receptor occupation levels are summarized in Table 1.
2.2. Super-Resolution Imaging Reveals High-Affinity Binding of Ligands to Membrane Receptors on Intact Cells
We established a new method to determine binding affinities directly on cells using single-molecule imaging. Fluorophore-labelled ligand was titrated against cells expressing the target protein, single-molecule imaging was applied and ligand binding sites were counted. By plotting the number of labelled ligand bound to receptor against the total ligand concentration surrounding the cells, we derived binding curves. We applied a 1:1 binding model to determine the dissociation constants of ligand–receptor interactions on intact cells (Figure 2). This is legitimate, as for our studied receptor–ligand interactions, a 1:1 stoichiometry is predicted. For MET-InlB a 2:2 binding was found.35, 36 While TNF-R1-TNFα has a 3:3 stoichiometry, one TNFα trimer exhibits three binding sites for the receptor.28, 37, 38 In our experiments, plotting the total ligand concentration is justified, because the reduction of the ligand concentration caused by receptor binding is negligible (<0.01 %) and the dissociation constants determined are significantly larger than the receptor concentration. To test whether our method yields reliable data, we compared this approach to data obtained from fluorescence correlation spectroscopy (FCS).27 Specifically, we compared single-molecule data on binding of InlB-ATTO647N to the uninduced MET receptor (Figure S2, Table 2) to FCS data on binding of ligand to the MET ectodomain in vitro. For the uninduced case, receptors were fixed with formaldehyde prior to ligand incubation, that is, a biological reaction, for example receptor dimerization or co-receptor recruitment, cannot be induced. This approach rather shows the in vitro behaviour of the receptor than reflecting the in vivo system. We obtained similar dissociation constants for FCS and single-molecule imaging, that is, KD=5.0±0.8 (s.d.) nM (FCS)27 and KD=5.2±1.4 (s.d.) nM (single-molecule imaging). This good agreement of experimental results confirms the validity of the single-molecule approach. However, we note that the standard deviation is higher for the ex vivo measurements on cells, which is expected, as we deal with individual cells that naturally show a heterogeneous behaviour.
Figure 2. Binding study on ligand-induced HeLa cells via single-molecule imaging. Representative localisation images at different ligand concentration and the resulting titration data are shown for a) InlB-ATTO647N binding to MET (no scraping) and b) TNFα-ATTO647N binding to TNF-R1 (with scraping). Here, the x-axis represents the total ligand concentration. The y-axis corresponds to the number of counted ligand–receptor pairs. In the case of InlB-ATTO647N the bound ligand density was corrected for the DOL of 0.5. The binding curves were determined by least-square-fitting with a Langmuir-binding model. The single-molecule images on the left correspond to the data points with green and blue circles in the binding curves. For each ligand concentration, several cells (N=4–10) were analysed (scale bar 1 μm).
Download figure to PowerPoint
Table 2. Dissociation constants of different receptor–ligand systems.
|MET, induced, scraped||InlB-ATTO647N||3.3±0.8|
|TNF-R1, induced, scraped||TNFα-ATTO647N||15.69±0.03|
We now make use of our approach to determine ligand–receptor binding constants on intact cells and investigated how the dissociation constant changes when looking at ligand binding in living cells (induced receptor). We mimicked this interaction by first incubating live HeLa cells with endogenous MET/TNF-R1 expression with a fluorophore-labelled ligand, and fixing cells afterwards, prior to single-molecule imaging. Representative binding curves for the MET receptor and TNF-R1 are shown in Figure 2. Notably, we determined the same dissociation constant for induced MET for both cells measured without (Figure 2 A) and with cell scraping (Figure S3), which indicates that cell scraping and additional washing steps do not alter the results (see also Table 2). A possible explanation for the higher ligand density is the decrease in background fluorescence in the case of scraped cells which might overlap with the ligand signal. Accurate localisation of the fluorescent signal of labelled ligands can therefore be impaired.
The binding affinity of InlB for its receptor MET slightly increases in live cells, which is expressed in a decrease of the dissociation constant from 5.2±1.4 nM in uninduced cells (Figure S2) to 3.1±1.3 nM after induction (Figure 2 a). This observation could be explained for example by the presence of co-receptors involved in the binding reaction or co-receptors recruiting the ligand in the proximity of the receptor or by the additional contact formed upon dimerization of two InlB molecules.36 The binding curves show that the number of bound ligand varies significantly at higher ligand concentrations, again showing cell-to-cell variability. In the past, other techniques were used to determine the binding affinity of InlB to MET. Surface plasmon resonance yielded dissociation constants between 20 and 150 nM39 and ELISA yielded values of 1 nM and 5 nM.39–41
We also determined TNF-R1/TNFα binding affinity (Figure 2 b), and found a value of 15.69±0.03 nM in induced cells. Notably, literature reports a vast range of binding affinities for TNF-R1/TNFα in different cell lines, ranging from 3 to 920 pM using radioligand binding assays42–44 to 0.59 to 290 nM using SPR.45–47 We again point out that these large variations may in part be explained by the different techniques. It is known that TNF-R1 undergoes reorganization upon TNFα binding which may affect further ligand–receptor interactions; a receptor pre-assembly for ligand binding is discussed.48 This binding assay is performed on living cells in the case of ligand induction, thus taking effects of receptor organization on the cell membrane into consideration. In contrast, in vitro binding assays reflect the single-site ligand-binding affinity, excluding avidity effects.
In summary, the presented approach allows measuring dissociation constants directly on cells and under varying experimental conditions. Compared to other techniques, for example in vitro FCS, the results obtained for the MET receptor are in good agreement, but show a higher standard deviation. This standard deviation must partly be attributed to cell-to-cell variability, as expression levels of receptors differ. Other cell-based approaches like radiometric ligand binding assays average over a large number of cells so that information on individual cells are lost. Also, no differentiation between intact and damaged cells is made. In contrast, single-molecule binding studies are performed at the single-cell level. Simultaneously, by averaging over several cells information on average ligand numbers bound to receptors is gained. In this way, we could determine binding curves revealing moderate- to high-affinity binding of different ligand–receptor pairs.