Research ArticlePlant biology

Real-time dynamics of peptide ligand–dependent receptor complex formation in planta

See allHide authors and affiliations

Science Signaling  04 Aug 2015:
Vol. 8, Issue 388, pp. ra76
DOI: 10.1126/scisignal.aab0598


The CLAVATA (CLV) and flagellin (flg) signaling pathways act through peptide ligands and closely related plasma membrane–localized receptor-like kinases (RLKs). The plant peptide CLV3 regulates stem cell homeostasis, whereas the bacterial flg22 peptide elicits defense responses. We applied multiparameter fluorescence imaging spectroscopy (MFIS) to characterize the dynamics of RLK complexes in the presence of ligand in living plant cells expressing receptor proteins fused to fluorescent proteins. We found that the CLV and flg pathways represent two different principles of signal transduction: flg22 first triggered RLK heterodimerization and later assembly into larger complexes through homomerization. In contrast, CLV receptor complexes were preformed, and ligand binding stimulated their clustering. This different behavior likely reflects the nature of these signaling pathways. Pathogen-triggered flg signaling impedes plant growth and development; therefore, receptor complexes are formed only in the presence of ligand. In contrast, CLV3-dependent stem cell homeostasis continuously requires active signaling, and preformation of receptor complexes may facilitate this task.


In multicellular organisms, growth and development is based on the coordination of cell proliferation and differentiation between single cells and within tissues. In every developmental step, communication between cells is essential. In plants, several signal transduction pathways coordinate growth or responses to various stimuli. Many of these pathways involve small signaling peptides, which are perceived by receptor-like kinase (RLK) proteins that transduce the signal into the cells. The CLAVATA (CLV) pathway is an example of RLK-mediated peptide signal transduction in Arabidopsis thaliana. Another example is the plant defense response initiated by the presence of the bacterial peptide flg22, which involves the receptors FLAGELLIN-SENSITIVE 2 (FLS2) and BRI1-ASSOCIATED KINASE 1 (BAK1). Here, we investigated the initial events that occurred at the receptors in each of these pathways using a multiparameter fluorescence imaging spectroscopy (MFIS) approach, combining fluorescence lifetime imaging with fluorescence polarization and anisotropy microscopy over time. With this approach, we simultaneously measured changes in protein concentration and both homomeric and heteromeric interactions between the receptors with pixel-wise resolution. This provided high spatial and temporal resolution of the interaction states of the receptors over time in individual living plant cells and in response to peptide treatments, which enabled the detection of rapid or transient changes in complex formation, arrangement, and intracellular localization.

The CLV pathway is the key regulatory pathway for stem cell homeostasis in the floral and shoot apical meristems of A. thaliana. CLAVATA3 (CLV3) encodes a precursor protein that is processed into a 13–amino acid peptide, which is further modified by the addition of sugar moieties to hydroxyproline residues (1). The mature CLV3 peptide is secreted from stem cells and perceived by the CLAVATA1 (CLV1), CLAVATA2 (CLV2), and CORYNE (CRN) receptor-like proteins in underlying cells of the organizing center (OC) (26). In the OC, the signal is transmitted intracellularly to repress the expression of WUSCHEL (WUS), encoding a stem cell regulatory transcription factor (7, 8).

CLV1 encodes an RLK with an extracellular leucine-rich repeat (LRR) receptor domain that binds the peptide CLV3 (9), a transmembrane domain that integrates into the plasma membrane, and an intracellular kinase domain for downstream signaling. Protein interaction studies showed that CLV1 preferentially forms homomers at the plasma membrane (10). CLV2 also has a transmembrane domain for localization in the plasma membrane and an extracellular LRR receptor domain, which may possibly interact with a range of different peptides of the CLV3-related CLE family (11); however, direct binding of CLV3 to the purified LRR-domain of CLV2 could not be detected in (12). CLV2 carries a short juxtamembrane domain on the intracellular side but lacks a kinase domain. CRN is also localized to the plasma membrane through its transmembrane domain but lacks an extracellular LRR receptor domain (5). CLV2 and CRN interact through their transmembrane domains, and this interaction is required for the export of both proteins from the endoplasmic reticulum (ER) and delivery to the plasma membrane (10). CRN may act as a co-receptor for CLV2; however, CRN may be a pseudokinase because the kinase domain does not autophosphorylate and is structurally atypical (13).

Mutants in any of the genes CLV1, CLV2, CLV3, and CRN produce supernumerary stem cells because of a lack of WUS repression. In floral meristems, this can be quantified as the number of carpels that fuse to form the silique. Using this readout, none of the clv1, clv2, or crn single mutants is as strong as the clv3 mutant, and clv1/clv2 or clv1/crn double mutants show additive effects that reach the severity of the clv3 mutant. In contrast, the phenotype of the clv2/crn double mutants resembles that of the clv2 or crn single mutants, indicating that CLV2 and CRN function together but in parallel and independently of CLV1 (25). CLV1 and CLV2 do not interact directly with each other (10). However, recent evidence suggests that crosstalk exists between the two pathways, which could be mediated by a direct interaction of CRN with CLV2 and CLV1 (10). Interaction studies using fluorescence resonance energy transfer (FRET) between fluorescently labeled receptor proteins show that CRN might act as the central component in a multimeric complex consisting of CLV1, CLV2, and CRN, which is detectable in the absence of CLV3.

The findings described above raise the question of how the assembly or reassembly of these different receptor complexes, consisting of CLV1 homomers (CLV1/CLV1), CLV2/CRN heteromers, and CLV1/CLV2/CRN multimers, is guided. Previous studies used methods such as coimmunoprecipitation experiments, genetic interaction studies, FRET acceptor photobleaching (APB) measurements, and structural analyses of protein crystals. A drawback of these methods is that they only reflect the static situation at the specific time point when the experiment is performed but lack any temporal dimension. Therefore, interaction dynamics over time could not be recorded. Furthermore, because most of these experiments were not performed in the intact living cell, all spatial information is lost, hence interactions taking place at the plasma membrane, in specific subdomains of the plasma membrane, or in other membranous compartments such as the ER or vesicles involved in receptor recycling cannot be discriminated.

By applying MFIS to Nicotiana benthamiana expressing the receptors in the CLV pathway, which were fused to fluorescent proteins, we showed that the CLV receptors were organized in preformed complexes before the addition of CLV3, indicating that the receptors exist in a “ready” state for rapid perception of the signal. We found that in the absence of CLV3, the CLV1/CLV1 homodimers and the CLV2/CRN heterodimers were evenly distributed along the plasma membrane, whereas larger multimers that contained all three receptors accumulated in small clusters along the plasma membrane. The addition of CLV3 triggered additional receptor clustering into more numerous and larger multimers, within specific domains along the plasma membrane.

In plant defense against bacterial pathogens, two LRR-RLKs, FLS2 and its co-receptorBAK1, detect the presence of potential pathogens by binding the bacterial peptide flg22 (14, 15). Both receptor proteins consist of an extracellular LRR domain, a transmembrane domain that integrates into the plasma membrane, and an intracellular kinase domain that transmits the signal within the cell (16, 17). The presence of flg22 triggers the formation of receptor complexes consisting of FLS2 and BAK1 (FLS2/BAK1), whereas the two receptors are kept separate when flg22 is not present (14). Whether BAK1 is in a monomeric or dimeric state before complex formation with FLS2 is unknown. Different groups have reported monomeric or homomeric complexes for FLS2 in the absence of flg22: Using coimmunoprecipitation experiments from whole seedling tissue, Sun et al. (18) showed that some FLS2 molecules formed homomeric FLS2/FLS2 complexes, whereas, using FRET and fluorescence recovery after photobleaching (FRAP), Ali et al. (19) could not detect FLS2 homomeric complexes in protoplasts.

Applying our MFIS technique, we monitored the assembly of BAK1/FLS2 heteromeric complexes at the plasma membrane over time in living N. benthamiana cells before and after the addition of flg22. We found that BAK1/FLS2 heteromers were not present before ligand addition and that BAK1/BAK1 or FLS2/FLS2 homomers were not detectable at the plasma membrane before flg22 addition. We monitored the formation of the FLS2/BAK1 receptor complex over the course of 1 hour after the addition of flg22 and observed higher-order complexes with at least two BAK1 molecules, probably connecting two FLS2 molecules.

Our analysis indicated that the CLV and flg22 pathways exhibited distinct receptor behavior with the receptors of the CLV pathway, which is constitutively active throughout plant growth and development, existing in a preassembled ready state before ligand perception, and with the receptors of the flg22 pathway exhibiting ligand-induced receptor complex formation.


Application of MFIS to monitor receptor complex dynamics in living cells

MFIS is used in mammalian and plant cells to study molecular interactions (20). MFIS, which is based on the detection of FRET between two fluorescent proteins or fluorophores that are fused to the proteins of interest (20), enables monitoring of various fluorescence parameters simultaneously and over time in living cells. We chose the FRET pair green fluorescent protein (GFP) and mCherry fused to the different A. thaliana receptor proteins and transiently coexpressed the fusion proteins in N. benthamiana with a β-estradiol–inducible system (10). Although receptors could act differently in their native context than in the N. benthamiana system, we previously found good correspondence between observations on receptor interactions made for A. thaliana proteins expressed in N. benthamiana and A. thaliana (10, 21). Furthermore, with the inducible system, we can perform measurements with low protein concentrations, thereby reducing the occurrence of overexpression artifacts.

To measure the effects of the ligands on the interaction state of their cognate receptors, we infiltrated a 1 μM peptide solution into the leaves. We assumed that the concentration of peptide initially reaching the receptors is much lower than 1 μM and continuously increases to a final concentration approaching 1 μM during the time course of the experiment.

After selecting multiple cells coexpressing the fusion proteins, we performed MFIS recordings of these cells every 10 min for 1 hour to monitor the interaction states of the receptors. We analyzed two main values: the fluorescence lifetime (τ) and the fluorescence anisotropy (r). The fluorescence lifetime of a fluorophore describes the time that the fluorophore remains in the fluorescent state after being excited by a laser pulse. Hetero-FRET between the GFP and mCherry quenches the fluorescence of GFP, which results in a shortened fluorescence lifetime for GFP (22). The fundamental anisotropy (r0) of free GFP in an aqueous solution and in the absence of any rotation is ~0.38. This anisotropy is altered by the rotational freedom of a fluorophore. Unbound GFP is free to rotate at all angles, although it rotates with a rather slow rotational diffusion time, ρ ≈ 15 ns, because of its large hydrodynamic radius. With a fluorescence lifetime of τ = 2.6 ns, GFP has a steady-state anisotropy of r = 0.32, which is derived from the Perrin equation [r = r0/(1 + τ/ρ)] (23). If the GFP is fused to a protein, the rotational freedom of GFP is reduced and, therefore, the value of r increases. In contrast, indirect excitation of a GFP by another juxtaposed GFP (homo-FRET) leads to a depolarization of the signal and, therefore, a reduction in total GFP anisotropy. Because of these properties, anisotropy measurements provide additional information about the FRET state of a fluorophore (24). The results of measurements at different time points were plotted in two-dimensional MFIS plots as steady-state anisotropy r versus lifetime τ (Fig. 1A).

Fig. 1 Changes in average BAK1-GFP fluorescence lifetime and anisotropy over time.

(A) Left: Exemplary two-dimensional (2D) MFIS plot demonstrating the effects of various combinations of hetero- or homo-FRET on fluorophore lifetime and anisotropy. The blue sphere represents the fluorophore. The average donor fluorescence weighted lifetime (average 〈τDf), shortened to “lifetime (τ),” is plotted on the x axis, and the anisotropy (r) on the y axis. Hetero-FRET results in a decrease in τ and an increase in r. Hence, the data points would shift from the position of the blue to the violet sphere. Homo-FRET results in a decrease in r with no effect on τ. The data point would shift to the yellow sphere. A combination of homo- and hetero-FRET results in a decrease in both τ and r. The data point would shift to the orange sphere. Right: MFIS plot for BAK1-GFP in cells coexpressing FLS2-mCherry exposed to the indicated peptides. τ is the average of six cells per time point. The dashed line represents r and τ for free GFP according to the Perrin equation (23). min, minutes after peptide infiltration. Data are plotted as the average ± SE. (B) The top shows a magnified section of a cell, coexpressing BAK1-GFP and FL2-mCherry, exposed to flg22 with the BAK1-GFP lifetime (τ) represented over a 60-min time period. The bottom shows the anisotropy of BAK1-GFP in the same area. Scale bar, 10 μm.

Furthermore, we quantified the distribution of different complexes, which exhibit different fluorescence lifetimes, by determining how heterogeneous the measured lifetimes from all pixels of the fluorescence lifetime imaging microscopy (FLIM) images were. If the same complexes are formed everywhere along the plasma membrane, then all the pixels will have comparable lifetimes and the heterogeneity of the sample will be low. If treatment with the receptor ligand results in the formation of different or unevenly distributed complexes, the lifetime heterogeneity will increase over time. This lifetime heterogeneity in a sample can be quantified as theta (θ). Finally, we analyzed if a change in average fluorescence lifetime resulted from the formation of more numerous FRET-active complexes (complexes containing misfolded mCherry are “FRET-inactive”) or from an altered arrangement of the measured molecules in which the two fluorophores are brought into closer proximity, which would affect the efficiency of FRET. To distinguish between these two possibilities (more FRET-active complexes or rearranged complexes with higher FRET efficiencies), we determined the fraction of FRET-active complexes (xFRET) and the FRET efficiency (E) by pixel-integrated MFIS-FRET analysis (see Materials and Methods for details on the technical procedures).

Detection of flg22-dependent stepwise formation of multimeric BAK1/FLS2 complexes at the plasma membrane

Although MFIS measurements have been applied to plant cells before, real-time imaging over time has not (20). As a proof of principle, we monitored the ligand-triggered interaction between the flg22 receptors FLS2 and BAK1 (14). We chose this interaction because it is well studied, and we therefore have a clear expectation: when treated with the flg22 peptide, the two receptors should interact and form complexes at the plasma membrane. Consequently, we should be able to monitor how the receptors change from a noninteracting into an interacting state over time after the addition of peptide.

We first tested if we detected preassembled BAK1 (BAK1/BAK1) and FLS2 (FLS2/FLS2) before flg22 is perceived. We performed FRET-APB measurements, which only detect hetero-FRET between GFP and mCherry, with GFP- and mCherry-tagged versions of the proteins. Although C-terminal fusions to BAK1 impair the protein’s signaling capacity, ligand-dependent complex formation with FLS2 is not impaired by the tag (25). We measured donor dequenching, which we quantified as apparent FRET efficiency (ap. E%) after mCherry photobleaching (26), to detect the interactions between BAK1-GFP and BAK1-mCherry (or FLS2-GFP and FLS2-mCherry). We measured, as a negative control, the ap. E% of BAK1-GFP (or FLS2-GFP) expressed without an mCherry fusion protein and, as a positive control, the ap. E% of a version of BAK1 (or FLS) tagged with both GFP and mCherry fused directly together. Control measurements resulted in an ap. E% of ~0 for the negative and ~18 for the positive controls (Table 1). Values for BAK1/BAK1 or FLS2/FLS2 were comparable to negative controls, indicating that these receptors did not interact.

Table 1 BAK1-BAK1 and FLS2-FLS2 interactions analyzed by FRET-APB.

Measurements for BAK1-GFP or FLS2-GFP alone are negative. BAK1-GFP-mCherry and FLS2-GFP-mCherry measurements are positive controls.

View this table:

For all further experiments with a pixel-wise resolution, we used MFIS measurements of BAK1-GFP and FLS2-mCherry in control peptide (mock)–treated (fig. S1) or flg22-treated (fig. S2) cells over time (Fig. 1, A and B). The steady-state anisotropy of BAK1-GFP at the first time point was 0.35 for both treatments (Fig. 1, A and B, and figs. S1 and S2). This value is higher than the anisotropy value of free GFP (0.32), suggesting that the rotational freedom of BAK1-GFP is restricted. Furthermore, this indicated the lack of homo-FRET between two or more GFPs, and hence no BAK1-GFP homomer formation, which is consistent with the FRET-APB results showing that BAK1 did not interact with itself before complex formation with FLS2 (Table 1). Plotting the r-τ time series for cells expressing BAK1-GFP and FLS2-mCherry treated with the control peptide, we found no change in lifetime or anisotropy over time, indicating that BAK1 and FLS2 did not interact (Fig. 1A and fig. S1). However, upon addition of flg22, both fluorescence lifetime and anisotropy decreased over time, indicating the formation of BAK1/FLS2 complexes (Fig. 1, A and B, and fig. S2). Notably, the lifetime decrease preceded the decrease in anisotropy, indicating that the formation of BAK1/FLS2 heterodimers preceded the formation of BAK1/FLS2 multimers: At ~17 min after peptide treatment, the lifetime of BAK1-GFP was reduced compared to that of the control, indicating that a fraction of BAK1-GFP molecules was in a complex with FLS2-mCherry. We detected a reduction in the anisotropy of BAK1-GFP after ~37 min, indicating that the occurrence of homo-FRET (and therefore BAK1/BAK1 homodimerization) occurred after BAK1/FLS2 heterodimerization (Fig. 1A). These results indicated that in a first step after flg22 perception, BAK1/FLS2 heterodimers are formed, which would then in a second step aggregate to form larger complexes through interaction of the BAK1 molecules.

The measured lifetimes and anisotropy values are averages of all photons from all pixels collected in one image. This explains why the lifetime is only different from the negative control ~17 min after peptide infiltration, whereas complexes can be detected by coimmunoprecipitation after a few seconds (27). Here, the number of molecules in a complex must be high enough to alter the mean of the entire pixel. The flg22-induced receptor complexes are internalized after ligand binding, with vesicles detectable after ~25 min (28). Nonetheless, we found that the total fluorescence intensity of BAK1-GFP or FLS2-mCherry was stable over the course of 1 hour (figs. S1 and S2), most likely because only a small number of vesicles were targeted for degradation during this time. To verify that the reduced lifetime observed resulted from the formation of new complexes, rather than a change in FRET efficiencies of preexisting complexes, we determined the FRET efficiencies (E) and fraction of FRET-active complexes (xFRET). Over time, the FRET efficiency was unchanged (fig. S3, A and B), whereas the number of molecules in FRET complexes increased in the flg22-treated sample (Fig. 2A and fig. S3A). Furthermore, we found that the lifetime was homogeneous along the entire membrane and did not change during the observation time of 1 hour, indicating that BAK1/FLS2 complexes were evenly distributed along the membrane (Fig. 2B). Thus, the MFIS technique reliably reported the interaction states of receptor proteins in living plant cells over time.

Fig. 2 Quantification of BAK1-GFP and CRN-GFP lifetime heterogeneity (θ) and FRET efficiencies.

(A) Time series for fractions of FRET-active complexes (xFRET) in cells expressing the indicated receptors and exposed to the indicated peptides. n = 6 cells from two independent experiments for the cells expressing CRN-GFP, CLV2, and CLV1-mCherry exposed to CLV3; n = 4 for cells expressing CRN-GFP, CLV2, and CLV1-mCherry exposed to the control peptide (M); n = 6 for cells expressing BAK1-GFP and FLS2-mCherry exposed to flg22; n = 5 for cells expressing BAK1-GFP and FLS2-mCherry exposed to the control peptide (M). Tinted areas represent the SEM. (B) Lifetime heterogeneity (θ) for cells expressing the indicated receptors exposed to the indicated peptides. n = 6 cells from two experiments for cells expressing BAK1-GFP and FLS2-mCherry exposed to flg22; n = 5 for cells expressing BAK1-GFP and FLS2-mCherry exposed to the control peptide (M); n = 3 for cells expressing CRN-GFP and CLV2 exposed to CLV3; n = 6 for cells expressing CRN-GFP, CLV2, and CLV1-mCherry exposed to the control peptide (M) or CLV3. Data are plotted as average ± SE.

Detection of preformed receptor complexes in the CLV pathway

We then used the MFIS technique to monitor the interaction states of the CLV3 receptors to investigate if the three reported complexes all formed at the plasma membrane at the same time and in the same regions, if the receptor complexes are evenly distributed at the plasma membrane, and how CLV3 alters the behavior of the receptors at the membrane. Similar to our approach for BAK1 and FLS2, we transiently coexpressed CRN-GFP, CLV2, and CLV1-mCherry in N. benthamiana and exposed the cells to either CLV3 or an inactive, but closely sequence-related, control peptide (mock). Because CRN-GFP is only exported from the ER to the plasma membrane when in a complex with CLV2, we interpreted the localization of CRN-GFP at the plasma membrane as indicator that CLV2, which did not have a fluorophore attached, was expressed. The initial lifetime of CRN-GFP, coexpressed with CLV1-mCherry, at the first time point was ~2.37 ns, independent of the treatment (Fig. 3, A and B, and figs. S4 and S5), compared to 2.57 ns when CRN-GFP and CLV2 were expressed without CLV1-mCherry (Fig. 3A), indicating that the three proteins interacted and were present as part of the preassembled CLV1/CLV2/CRN receptor complexes at the plasma membrane even in the absence of CLV3.

Fig. 3 Changes in average CRN-GFP fluorescence lifetime (τ) and anisotropy (r) over time.

Data were collected from cells expressing CRN-GFP, CLV2, and CLV1-mCherry that were exposed to the control peptide (M) or CLV3 or from cells expressing CRN-GFP and CLV2 that were exposed to the control peptide. (A) Left: MFIS plot of data for the indicated cells exposed to the indicated peptides. The dashed line represents r and τ for free GFP according to the Perrin equation (23). Data are plotted as average ± SE. Right: Expanded view of the area containing the data from the cells expressing CRN-GFP, CLV2, and CLV1-mCherry. Error bars were removed for easier viewing. Dotted lines connecting the data points represent the chronological order. min, minutes after peptide infiltration. n = 6 cells per time point. (B) Top: A magnified section of a representative cell expressing CRN-GFP, CLV2, and CLV1-mCherry that was exposed to CLV3 with the CRN-GFP lifetime (τ) represented over a 60-min time period. Bottom: Anisotropy of CRN-GFP over the same area. Scale bar, 10 μm. (C) A magnified region showing subdomains with reduced lifetime of CRN-GFP in CLV3-treated cells expressing CRN-GFP, CLV2, and CLV1-mCherry. The asterisks mark the regions with reduced lifetime. Scale bar, 2.5 μm.

We evaluated cells exposed to CLV3 to investigate if ligand perception altered the formation or distribution of these preexisting receptor complexes. CLV3 led to a continuous decrease of GFP lifetime from 2.37 to 2.30 ns within 1 hour (Fig. 3, A and B, and fig. S4). A smaller decrease occurred in mock-treated cells, where the lifetime varied in the range of 2.35 to 2.37 ns (Fig. 3A and fig. S4). For both the CLV3- and mock-treated cells, the changes in lifetime were not accompanied by a visibly altered signal intensity for the GFP or mCherry channel, indicating that the total amount of receptor proteins detected at the plasma membrane remained constant (figs. S4 and S5). Because for this experiment, we averaged the lifetimes over all pixels in a given image, which does not allow to resolve local differences, we next performed a pixel-wise analysis.

Receptor complex clusters along the plasma membrane triggered by CLV3

When the lifetimes of all collected photons were overlaid onto the FLIM image of the cells, we noticed the emergence of small regions along the plasma membrane that exhibited reduced lifetimes 1 hour after addition of CLV3 (Fig. 3, B and C). However, because the lifetime of the GFP fusion proteins in the other areas of the plasma membrane was stable, the average lifetime of all pixels was only weakly affected by these regional differences. To quantify this regional effect of CLV3 treatment, we compared the lifetime heterogeneity (θ) of cells expressing only CRN-GFP and CLV2-mCherry that were exposed to CLV3 to that of cells expressing CRN-GFP, CLV2, and CLV1-mCherry that were exposed to either CLV3 or the control peptide (Fig. 2B). When expressed together without CLV1, CRN-GFP and CLV2-mCherry exhibited a smooth membrane distribution that was unaffected by the addition of CLV3, with an average θ close to 1. In contrast, in cells expressing CRN-GFP, CLV2, and CLV1-mCherry, exposure to the control peptide or CLV3 resulted in an increase in lifetime heterogeneity. However, the increase in lifetime heterogeneity for the cells exposed to the control peptide plateaued within 30 min at ~1.2, whereas the increase continued throughout the time course for cells exposed to CLV3, reaching 1.5 at 1 hour. These data indicated that CLV2/CRN heterodimers remained evenly distributed along the membrane (θ remained ~1 over time), whereas the CLV1/CLV2/CRN multimers localized slightly more heterogeneously in subdomains of the membrane (θ remained ~1.15 over time) and reacted to CLV3 with increasing clustering over time (θ increased up to 1.5 after 1 hour).

On the basis of this increase in fluorescence lifetime heterogeneity and the observed formation of regional lifetime differences along the plasma membrane in the images, we concluded that the binding of CLV3 to the receptor complexes triggered the clustering of more CLV1/CLV2/CRN multimers in subdomains of the plasma membrane. This observed CLV3-triggered clustering of smaller complexes into larger multimers should also be reflected in the measured FRET efficiencies and FRET-active complexes. In contrast to the BAK1/FLS2 interaction that we observed, the data indicated that the CLV3 receptors were already in FRET-active complexes before the addition of peptide. Accordingly, the fraction of FRET-active complexes (xFRET) should not change over time, but the FRET efficiency (E) should increase, because of more molecules being in closer proximity in these larger multimers. The FRET-active complex fraction xFRET remained unchanged over time in cells expressing CRN-GFP, CLV2, and CLV1-mCherry that were exposed to CLV3 or the control peptide (Fig. 2A), whereas the FRET efficiency between CRN-GFP and CLV1-mCherry in these cells exposed to CLV3 slightly increased from 0.36 to 0.45 (fig. S3). Therefore, we inferred that the molecular environment for CRN/CLV2/CLV1 multimers became more crowded after the addition of CLV3, suggesting the formation of larger clusters.


To test whether the MFIS technique is suitable to monitor the interaction state of receptor proteins in living plant cells over time, we monitored the complex formation between BAK1 and FLS2 after the addition of the ligand flg22. With MFIS, we recorded the expected formation of receptor complexes, but we also obtained previously unknown insights on the nature of receptor complex assembly. We found that both FLS2 and BAK1 form neither homomers nor heteromers before flg22 recognition at the plasma membrane. This is in accordance with the findings of Ali et al. (19), who monitored FLS2 at the plasma membrane using FRET and FRAP with a protoplast system. This is in contrast to the results of Sun et al. (18), who detected FLS2 homomers in the absence of the ligand by coimmunoprecipitation experiments with cell extracts. A possible interpretation of both apparently contradictory results would be that FLS2 does not form homomers in the absence of flg22 at the plasma membrane but can homomerize upon internalization, which would be detected only in the coimmunoprecipitation experiments.

After ligand perception, we showed that heterodimeric FLS2/BAK1 complexes formed initially and subsequently aggregated to form larger complexes. These larger complexes likely consist of two BAK1 and two FLS2 molecules, with the two BAK1 molecules interacting directly with each other, flanked by FLS2 on each side to form an FLS2/BAK1/BAK1/FLS2 tetrameric arrangement (Fig. 4A). Each of the two FLS2/BAK1 units would be active in both extracellular signal perception and intracellular transduction. This tetrameric arrangement could explain why Ali et al. (19) did not detect FLS2/FLS2 homomers after flg22 treatment: their position on the flanks of the complex would distance the FLS2 molecules too far apart for FRET. Our results also indicated that the composition of the formed complexes appeared the same along the entire plasma membrane because the BAK1-GFP fluorescence lifetime distribution was homogeneous.

Fig. 4 Models for stepwise assembly of complexes and clusters in the flg22 and CLV3 signaling pathways.

(A) The flagellin pathway. In the absence of flg22, the two receptors FLS2 and BAK1 do not interact. When flg22 is present, complexes consisting of one BAK1 and one FLS2 molecule are formed in the first step. In the second step, these dimers form larger complexes consisting of two central BAK1 molecules with one FLS2 molecules on each flank. (B) The CLV pathway. Without CLV3 present, CLV1 forms preferentially homomers, CRN and CLV2 form heteromers, and only few CLV1/CLV2/CRN multimers are formed. When CLV3 is present, CLV1/CLV2/CRN multimers cluster in membrane subdomains.

We then applied MFIS to monitor the interaction states of the receptor proteins in the CLV signaling pathway. The CLV pathway is the key regulatory pathway in plant stem cell homeostasis. Plants maintain a constant number of stem cells in their aboveground stem cell niches, which are the shoot apical, axillary, and floral meristems. The key to the maintenance of these stem cell pools is a tight balance between stem cell proliferation and differentiation of their descendents. This regulation operates continuously, requiring the presence of receptor complexes at all times. We found that three different types of complexes were preformed at the plasma membrane and had the capacity to bind CLV3: these are CLV1/CLV1 homomers, CLV2/CRN heteromers, and CLV1/CLV2/CRN multimers. Of these three types of complexes, the CLV1/CVL1 homomers and CLV2/CRN heteromers were evenly distributed along the plasma membrane. Taking into account that the CLV2 receptor domain does not directly bind CLV3 and the atypical nature of the kinase domain of CRN, we expect that these CLV2/CRN heteromers require another active RLK to be fully functional. MFIS analysis indicated that CLV3 triggered the formation of the larger CLV1/CLV2/CRN multimers, preferentially in membrane subdomains (Fig. 4B). Within these complexes, CLV2/CRN may act as co-receptors for CLV1, thereby increasing the specificity of the interaction between CLV3 and CLV1, or CLV2/CRN may aid in the assembly of signaling-competent complexes. Sequestration of receptor complexes into membrane subdomains might facilitate the assembly of additional factors necessary for downstream signaling, including the phosphatases that limit the kinase activities of RLKs (2931). Our model with both an active CLV1/CLV1 homomer and a CLV1/CRN/CLV2 heteromer is not in contrast to the published genetic data, which suggested two parallel and independently acting pathways (5) on the basis of the observation that clv1/clv2 and clv1/crn double mutants are additive, whereas clv2/crn double mutants are not (5). This is consistent with all three receptors contributing to CLV3 signaling as components of CLV1/CLV1 homomers, CLV2/CRN heteromers, and CLV1/CLV2/CRN heteromeric complexes. An alternative function for the rapid clustering of CLV1/CLV2/CRN complexes in subdomains in the presence of ligand may be to sequester the receptors and enable their inactivation. In this case, the CLV1/CLV1 homomers and CLV2/CRN heteromers would signal in parallel and mostly independently, whereas the multimeric CLV1/CLV2/CRN complexes in the subdomains would be inactive. Clustering of the two otherwise independent signaling complexes in larger multimeric aggregates would provide a simple but effective means to facilitate the rapid and parallel down-regulation of both pathways in a situation with excess CLV3. The steady increase of clusters over time could then reflect the increasing amount of peptide that reaches the receptors, leading to their activation and subsequent sequestration. Such a mechanism may protect meristems from terminal stem cell loss after a transient surge in CLV3 abundance. Here, it is not clear how the membrane regions in which these multimers cluster are defined. These clusters may form at the contact sites of the cytoskeleton-guided ER or Golgi strands at the plasma membrane or within detergent-resistant membrane fractions (32).

A similar clustering of receptors upon ligand stimulation has been proposed for the epidermal growth factor receptor (EGFR) in mammalian cells (33). By combining microscopy, image correlation spectroscopy, phosphorylation assays, and computational modeling of mass action kinetics, Kozer et al. (33) suggested that EGFRs are localized at the plasma membrane as preformed dimers and that ligand leads to formation of higher-order oligomers in clusters along the plasma membrane, which then participate in transphosphorylation (33). EGF-triggered clustering of EGFR is mediated by the scaffolding protein flotillin-1 (flot-1, also known as reggie-2), which localizes to lipid rafts in the plasma membrane and subsequently aids in the formation of larger EGFR-containing complexes that activate a downstream mitogen-activated protein kinase (MAPK) signaling cascade (34). In Arabidopsis, the protein family related to flot-1 is the HIR family, consisting of four proteins. HIRs localize to microdomains in the plasma membrane and are involved in LRR receptor–mediated signaling pathways (35). Furthermore, there is evidence that the CLV receptors regulate shoot meristem homeostasis partially through signaling through the MAPK cascades. CLV1 functions as a CLV3-dependent negative regulator of the activity of the MAPK MPK6 activity, whereas CLV2 appears to counter this effect, thereby providing another mechanism to fine tune CLV3-dependent signaling (36).

In the neuregulin (NRG) pathway, the EGFR-related ErbB4 is evenly distributed along the plasma membrane, with only some molecules in lipid rafts before ligand perception. Addition of the ligand NRG results in a relocalization and clustering of the existing protein complexes into lipid rafts, together with several associated signaling molecules, displaying a mechanism similar to our observations for the CLV3 receptors (37).

The observation that FLS2/BAK1 complexes only formed when the ligand was present is intriguing, especially when considered within the biological context. The FLS2/BAK1 complex is only required in the case of bacterial infection, and formation of this complex at the plasma membrane in the absence of ligand could result in basal activity of flagellin signaling and could compromise growth and development (17). Such inappropriate activation of immune responses would therefore incur a fitness penalty. Activation of the growth-promoting brassinosteroid receptor (BR) pathway, which is mediated by the receptor BRI1 and BAK1, inhibits immune signaling. This inhibition, however, does not occur through competition between FLS2 and BRI1 for the BAK1 co-receptor, but downstream of the receptors (38).

The observation that the flagellin receptors are kept separate and the CLV receptors form complexes independent of ligand availability seems to reflect the nature of the two pathways. In contrast to defense signaling, stem cell homeostasis is a continuous process, requiring the pathway to be active at all times. Our findings indicated that this physiological difference is reflected at the molecular level in the interaction properties of the signaling receptors. A similar observation was described by Bücherl et al. (39), who reported that ~7% of the BRI1 molecules interact with BAK1 in the absence of ligand. Addition of BR then results in an increase in heteromers. This is in accordance with our findings that in a constitutively active pathway involved in plant development, some receptors are maintained in a ready state to ensure ongoing signaling (39). Our study demonstrated that MFIS applied to living plant cells not only monitored the interaction states of several receptors over time but also detected changes in the composition of existing complexes, revealing a dynamic view of signaling in plants.


Plant reporter lines

N. benthamiana plants were grown in the greenhouse for 4 weeks before transient transformation. Transformation and expression were described before (10).

Construction of inducible receptor fusions

The CLV1, CLV2, CRN, and BAK1 expression vectors were described before (10). The FLS2 fusions were created from complementary DNA using the pENTR/D-TOPO and Gateway LR Clonase II cloning kits, as well as the destination vectors pABindGFP, pABindmCherry, and pABindFRET, as described previously (10).


The peptides flg22 (QRLSTGSRINSAKDDAAGLQIA) and CLV3 (RTV[Hyp]SG[Hyp]DPLHHH) and the inactive control peptide (LPQHPHGRSDVT) were ordered from Thermo Fisher Scientific. They were infiltrated into the transformed plant leaves immediately before imaging at an initial concentration of 1 μM in an infiltration medium using 1-ml flat-top syringes as described by Bleckmann et al. (10).


Measurements were performed using a multiparameter fluorescence detection setup as described previously (20, 40). Experiments were performed with a confocal laser scanning microscope (FV1000, Olympus) additionally equipped with a single photon counting device with picosecond time resolution (Hydra Harp 400, PicoQuant). GFP was excited at 485 nm with a linearly polarized, pulsed (32 MHz) diode laser (LDH-D-C-485, PicoQuant) at 0.8 μW at the objective [60× water immersion, Olympus UPlanSApo NA (numerical aperture) 1.2, diffraction-limited focus]. mCherry was excited at 559 nm with a continuous wave laser (FV1000) at 5.4 μW at the objective. The emitted light was collected in the same objective and was separated into perpendicular and parallel polarization with respect to excitation polarization. GFP fluorescence was then detected by an avalanche photo-diode (PDM50-CTC, Micro Photon Devices) in a narrow range of its emission spectrum (bandpass filter, HC520/35; AHF). mCherry fluorescence was detected by a hybrid photodetector (HPMC-100-40, Becker & Hickl), of which the detection wavelength range was set by the bandpass filters (HC 607/70, AHF). Images were taken with 20-μs pixel dwell time and a resolution of 103 nm/pixel. A series of 40 frames were merged into one image and further analyzed using custom-designed software (LabVIEW).

The FRET-APB measurements were done on a Zeiss LSM 780. GFP was excited with a continuous wave argon laser at 488 nm using a 40× water immersion objective (Zeiss C-Apochromat 40×/1.20 W Corr M27), and emission was detected between 498 and 524 nm by a GaAsP detector. mCherry was excited using a 561-nm continuous wave diode-pumped solid-state laser, and emission was detected between 578 and 639 nm. A series of 12 256 × 256 pixel frames with 0.18 μm pixel size, 47 μm2 image size, and 1.27 μs pixel dwell time was recorded. After five frames, mCherry was photobleached in a region of interest along the plasma membrane by 80 iterations with 100% laser power. The FRET efficiency (E%) was determined as the change in GFP intensity after photobleaching of the acceptor mCherry by [(GFPafter − GFPbefore)/GFPafter × 100].

Pixel-wise fluorescence-weighted lifetime analysis

The histograms presenting the decay of fluorescence intensity after the excitation pulse were built for each pixel with 128 ps/bin. The fluorescence-weighted lifetime of the donor molecule in single pixel was determined using a model function containing only two variables (〈τDf) and scatter contribution [for details, see (21)], with maximum likelihood estimator (MLE). The instrument response function was measured with the back reflection of the laser beam and used for iterative reconvolution in the fitting process.

Pixel-wise anisotropy analysis

The steady-state anisotropy is given by rG=F||G·FF||+2·G·F, where F || and F are the average fluorescence count rates within a pixel, with a polarization parallel and perpendicular to that of the excitation light, respectively. Both were corrected for dead time of the detection electronics (41) and mixing of polarization in the high numerical aperture objective (42); F = F|| + 2GF is the total fluorescence intensity. Calibration measurements with Rhodamine 110 delivered the G-factor to correct the signal for orientational sensitivity differences of the detection system.

Fluorescence lifetime heterogeneity analyses

With MLE, the variance of fluorescence lifetime distribution σ02 is inversely proportional to the number of photons in the decay histogram N under shot noise–limited conditions: θ0=σ02·N. The constant θ0 that governs the width of the lifetime distribution can be calculated [for details, see (43)]. For the experiments in living cells, lifetime distribution is wider than what is expected under the ideal condition because of detector noise and the heterogeneous cellular environment in biological samples. This will be reflected by a larger θ value than θ0. To quantify the lifetime broadening, the θ value is calculated for each image according to θ=σ2·N¯, in which σ2 is the variance of the lifetime distribution and N¯ is the mean photon count. Considering that each individual cell may have a different environment, which results in different absolute θ values, we normalized θ values of a set of k time series measurements to that of the first measurement: θ(k)¯=θ(k)/θ(1). In this way, the time-independent lifetime broadening factors that exist in all the measurements are excluded by normalization, leaving only the time-dependent lifetime broadening due to peptide infiltration. After normalization, the starting normalized θ value is always 1, so that different sets of time series measurements can also be compared. An increasing θ value indicates that the lifetime distribution is wider, an indication that the whole population becomes more heterogeneous, and vice versa.

Quantitative MFIS-FRET pixel-integrated analysis

Noninteracting (donor-only) and interacting (donor-acceptor) species can be resolved in our lifetime-based FRET analysis. Two fluorescence lifetimes, τD01 and τD02, were assigned to donor-only species considering the typical biexponential characteristic of fluorescence proteins in vivo. When FRET occurs, the two donor lifetimes were quenched through (an) associated FRET rate(s): τDA1(2) = 1/[τDA1(2) + kFRET]. Using predetermined donor lifetimes from donor-only sample, FRET rate(s) can be fitted for a subensemble of pixels. For the cells expressing BAK1-GFP and FLS2-mCherry and treated with flg22, pixels located at the plasma membrane of a given cell were selected as a subensemble, and two FRET rates were used in the fit. However, as additional selection criteria, only the pixels with a lifetime shorter than 2.3 ns and a green-to-red intensity ratio below 6.2 were selected as a subensemble to determine heterogeneity for cells expressing CRN-GFP, CLV2, and CLV1-mCherry and treated with CLV3 or mock peptide. One FRET rate was required in this case. FRET efficiency was calculated as (E = 1 − 〈τDAx/〈τD0x), where 〈τDAx and 〈τD0x are the species-weighted donor lifetimes in the presence and the absence of acceptor, respectively.

Mean fraction of FRET-active complexes, 〈xFRETn

To compare the peptide-induced effect in different cells (fig. S3), we computed the scaled fraction of FRET-active complexes (〈xFRETn) for cell i at time point t: xFRET,S(t)=xFRET(t)·1ni=1n(xFRET(t=end))ixFRET(t=end), where n is the number of cells. The mean fraction of FRET-active complexes 〈xFRET(t)〉n over n cells is calculated as xFRET(t)n=1ni=1n(xFRET,S(t))i.


Fig. S1. Fluorescence intensity, lifetime, and anisotropy projections on a cell expressing BAK1-GFP and FLS2-mCherry that was exposed to the control peptide

Fig. S2. Fluorescence intensity, lifetime, and anisotropy projections on a cell expressing BAK1-GFP and FLS2-mCherry that was exposed to flg22

Fig. S3. Time series of FRET efficiencies (E) and fraction of FRET-active complexes (xFRET) of individual cells after peptide treatment.

Fig. S4. Fluorescence intensity, lifetime, and anisotropy projections on a cell expressing CRN-GFP, CLV2, and CLV1-mCherry that was exposed to the control peptide

Fig. S5. Fluorescence intensity, lifetime, and anisotropy projections on a cell expressing CRN-GFP, CLV2, and CLV1-mCherry that was exposed to CLV3


Acknowledgments: M.S. and R.S. thank P. Žádníková for helpful comments and discussion of the manuscript. Funding: Part of this work was financed by the Deutsche Forschungsgemeinschaft through CRC590 and EXC1028 (Cluster of Excellence on Plant Sciences). C.A.M.S. acknowledges the support of the CRC974. Author contributions: M.S., Q.M., S.W.-P., A.B., Y.S., R.S., and C.A.M.S. designed the research; M.S. and Q.M. performed the research; S.F. contributed new analytic tools; Q.M. and M.S. analyzed the data; M.S. and R.S. wrote the paper. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All reagents, chemicals, and bacterial and plant lines are available commercially. The plasmids are available upon request.
View Abstract

Stay Connected to Science Signaling

Navigate This Article