Fast rebinding increases dwell time of Src homology 2 (SH2)-containing proteins near the plasma membrane

Abstract

Receptor tyrosine kinases (RTKs) control a host of biological functions by phosphorylating tyrosine residues of intracellular proteins upon extracellular ligand binding. The phosphotyrosines (p-Tyr) then recruit a subset of ∼100 Src homology 2 (SH2) domain-containing proteins to the cell membrane. The in vivo kinetics of this process are not well understood. Here we use total internal reflection (TIR) microscopy and single-molecule imaging to monitor interactions between SH2 modules and p-Tyr sites near the cell membrane. We found that the dwell time of SH2 modules within the TIR illumination field is significantly longer than predictions based on chemical dissociation rate constants, suggesting that SH2 modules quickly rebind to nearby p-Tyr sites after dissociation. We also found that, consistent with the rebinding model, the effective diffusion constant is negatively correlated with the respective dwell time for different SH2 domains and the dwell time is positively correlated with the local density of RTK phosphorylation. These results suggest a mechanism whereby signal output can be regulated through the spatial organization of multiple binding sites, which will prompt reevaluation of many aspects of RTK signaling, such as signaling specificity, mechanisms of spatial control, and noise suppression.

The Src homology 2 (SH2) domain (1) controls many cellular activities by binding specifically to tyrosine-phosphorylated peptides (2). The tyrosine phosphorylation signal often originates from the cell surface through activations of receptor tyrosine kinases (RTKs). Humans, for example, have 58 known RTKs (3), each of which could generate diverse phosphorylation patterns at their cytosolic tails as well as at other associated proteins. These phosphorylation patterns are “read” mainly by a repertoire of effector proteins containing the SH2 domain (2, 4, 5), via recruitment of these effector molecules to the phosphotyrosine (p-Tyr) sites. For example, the epidermal growth factor receptor (EGFR), a well-known model system for studying phosphotyrosine signaling, is activated by epidermal growth factor (EGF), a small protein ligand. The binding of EGF turns on the kinase activity of EGFR molecules, which then autophosphorylate themselves at more than 20 tyrosine sites on the cytoplasmic tail. The phosphotyrosines are thought to serve as docking sites for SH2-containing molecules, such as Grb2, which bind to these sites with varying affinity. Downstream signals are propagated by the SH2 proteins, which in many cases are enzymes whose substrates are localized to the plasma membrane. In other cases such as Grb2, the SH2 protein serves to recruit additional downstream effector proteins such as the Ras activator Sos. EGFR family members also play critical roles in cancer development and progression and are frequently disregulated in tumors (6).

During the last decade, an expanding assembly of quantitative datasets has accumulated regarding tyrosine phosphorylation signaling, including the topological pattern of phosphorylation (7–10), the temporal dynamics of phosphorylation (10, 11), and characterization of the SH2/p-Tyr binding affinity (12, 13). On the other hand, the in vivo assembly and turnover kinetics of the SH2 complexes at p-Tyr sites remain poorly understood, partly because existing methods measure mostly the ensemble properties of the system. To this end, measurements on the single-molecule basis provide an advantage in understanding the dynamics of these heterogeneous interactions. With a total internal reflection (TIR) microscope, we can monitor the turnover of individual SH2 molecules by tracking the molecule from its initial docking to a p-Tyr site at the cell’s plasma membrane to its dissociation from the membrane and leaving the TIR illumination field. The dwell times of single SH2 molecules within the TIR field thus can be measured and analyzed statistically. Here we report that our experimental results from EGF-stimulated cells can be quantitatively explained only when fast multiple rebinding of SH2 molecules to p-Tyr sites is taken into consideration.

Results

Measuring Dwell Time of SH2 Molecules Near the Cell Membrane with TIR Illumination.

We expressed in A431 cells the SH2 domain of Grb2 (Grb2_SH2), as a chimeric fusion to tdEos, a photoactivatable fluorescent protein. Single molecules were detected after a small fraction of tdEos was photoactivated (14). Without extracellular ligand stimulation, very few Grb2_SH2 molecules could be seen under TIR illumination (Fig. 1A and Fig. S1), but the number of molecules quickly increased upon the addition of EGF to the media (Fig. 1 A and B), presumably because the activation of EGFR led to an increased number of p-Tyr sites at the membrane. During the time-lapse imaging (Movie S1), new molecules continuously appear at the plasma membrane, undergo diffusive motion for at most a few seconds, and then disappear from the membrane. We attribute the appearance of a molecule to the initial binding reaction to a p-Tyr site and the disappearance to the escape of the molecule from the TIR illumination field or to photobleaching.

Fig. 1.

Rebinding of the Grb2_SH2 domain to phosphotyrosine sites increase its dwell time at the cell membrane. (A) Recruitment of Grb2_SH2 (Left) and the nonbinding mutant control, mtGrb2_SH2 (Right). Epi-fluorescence image shows the expressed molecules distributed mostly in the cytosol (Top row). TIR illumination of unstimulated cells shows little recruitment of Grb2_SH2 to the cell membrane (Middle row). Upon EGF stimulation, a large number of mobile Grb2_SH2 molecules, but not mtGrb2_SH2, can be detected with TIR fluorescence (Bottom row). (Scale bar: 2 μm.) (B) The time course of Grb2_SH2 recruitment after EGF stimulation. The amount of the Grb2_SH2 recruitment was measured by integrated fluorescence intensity under TIR illumination (lines) or by counting the number of single molecules (bars). The different traces represent measurements from four different cells. (C) Survival-time distribution for Grb2_SH2, mtGrb2_SH2, and Myr-tdEos in A431 cells stimulated for 20 min with EGF. (D) Hazard function (lines) of Grb2_SH2 computed from survival-time distribution (bars). Solid lines are the simple moving average. (E) Schematic diagram illustrating the rebinding mechanism.

Histograms of the dwell times, defined as the time delay between a molecule’s appearance and its disappearance, are shown in Fig. 1C. As expected the dwell time of Grb2_SH2 falls between that of a nonbinding mutant control (mtGrb2_SH2), which we consider to be due to nonspecific binding, and that of myr-tdEos (tdEos with a myristoylation signal peptide sequence from Src that directs constitutive membrane association), which is due to photobleaching. However, because of the inevitable contribution from photobleaching, the dwell-time representation is not a direct measurement of the turnover kinetics. To overcome this problem, we calculated the hazard function (15), defined as

where p(t) is the probability distribution of the dwell times, estimated by its histogram. The hazard function is additive for independent processes, and thus the contribution of photobleaching can be simply subtracted (SI Theory). The physical meaning of λ(t) is the average escape rate for molecules that survived for at least time t. The value of λ(t) becomes a constant (independent of t) if the underlying kinetic process is a simple first-order process, in which case the hazard function is equal to the first-order rate constant.

After subtracting the contribution from photobleaching (Fig. S2), we found that the hazard function for Grb2_SH2 appears to be a constant (Fig. 1D) at ∼0.52 s⁻¹. This result was surprising in two regards: First, the Grb2_SH2 domain binds to multiple EGFR p-Tyr sites in vitro, with varying affinities (12), which presumably arise partly from different dissociation rates. Heterogeneity is known to affect the shape of the hazard functions (16). For dissociation reactions, heterogeneity results in a decaying hazard function because a molecule that survived for a longer time is more likely to be one bound to a high-affinity site and thus would have a lower average dissociation rate. Second, the measured average hazard function value, 0.52 s⁻¹, is ∼20 times slower than the corresponding dissociation rate constant for Grb2_SH2 previously reported from in vitro studies (17, 18). In light of this evidence, we suspected that the SH2 molecules might rebind efficiently, which can happen if there is a competition between diffusion and a high on rate between the SH2 and the p-Tyr (19, 20). In such a scenario (Fig. 1E), the Grb2_SH2 molecule rebinds multiple times before eventually escaping from the TIR field. The interaction of SH2 with multiple p-Tyr sites therefore eliminates heterogeneity by averaging affinities over many sites. Indeed, on the basis of a reaction-diffusion model (SI Theory), we can theoretically predict a constant hazard function under this condition (Fig. S3). Furthermore, with multiple rebinding events, the apparent “dissociation rate” measured by the hazard function is much slower than the prediction based solely on the chemical off rate. In fact, we found (SI Theory and Fig. S4) that the hazard function should be dependent not only on the off rate, k_d, but also on the on rate, k_a, as well as on the diffusion constant, D, of the molecule when hopping between sites,

where n is the p-Tyr density and a is the imaging depth of the TIR field. On the basis of this model, on average ∼20 rebinding events were detected in our experiment for each Grb2_SH2 molecule before it escaped. Furthermore, the equation predicts that factors that increase the binding rate, k_onn, will further decrease the ability of molecules to escape the TIR field; on the other hand, fast diffusion of the dissociated molecules, which is controlled by D, promotes their escape. The importance of rebinding can also be evaluated outside the context of the TIR experiment. Lagerholm and Thompson (21) proposed using a unitless parameter, , to evaluate the significance of rebinding: For systems with b << 1, rebinding can be safely ignored and the dissociation kinetics are dictated by the chemical rate constant k_d, whereas when b >> 1 the kinetics of the dissociation process are strongly influenced by rebinding. On the basis of the hazard function, we estimated the value of b for Grb2_SH2 to be ∼44.6, assuming a k_d value of 10 s⁻¹ (17) and a D value of 15 μm²/s, corresponding to the cytosolic diffusion rate of a small protein (22).

SH2 Mobility Is Correlated with Its Apparent Membrane Dwell Time.

To verify that the rebinding dynamics are indeed important to the interactions between SH2 domains and p-Tyr sites, we measured the hazard functions, λ(t), as well as the effective mobility, D_eff, of five different SH2 domains from four SH2 proteins that are thought to directly interact with EGFR (Movie S2 and Fig. 2). We hypothesized that, if SH2 domains dissociate from the membrane without rebinding, the mobility should solely be determined by the diffusion constant of the EGFR receptor protein; thus, the variation in mobility of different SH2 domains should be small. On the other hand, if the SH2 domains rebind frequently, the mobility is determined not only by the diffusion constant of the receptors, but also by the frequency and the duration of the “hopping” motion between p-Tyr binding sites. We found that D_eff varied by more than an order of magnitude among different SH2 domains (Fig. 2A), supporting a rebinding mechanism. Furthermore, all five SH2 domains we studied exhibited constant hazard functions, albeit at different average values (Fig. 2B), again consistent with a rebinding mechanism. Finally, the average hazard rates and the mobilities were correlated (Fig. 2C), as was predicted by the reaction-diffusion model (SI Theory):

Fig. 2.

Correlation between the effective mobility and the apparent dissociation rate. (A) Effective diffusion constant (D_eff) for different SH2 modules. The ensemble-averaged MSD-Δt curves were computed from single-molecule traces for five different SH2 modules (squares). The effective mobilities were obtained by computing the slope (straight lines) from the first to the fourth time lags in the MSD-Δt curves, corresponding to the ∼200-ms timescale. The results illustrated the significant differences of D_eff among different SH2 modules. (B) Hazard functions of different SH2 modules. The hazard functions were computed from survival-time distributions as in Fig. 1. (C) Correlation between D_eff and the average dwell time, calculated as the inverse of the average hazard function values (1/λ). The solid line is the theoretical prediction from Eq. 3, assuming D = 15 μm²/s, D₀ = 0.03 μm²/s, and a = 0.5 μm. The theoretical values agree well with the experimental data.

Here D₀ is the diffusion constant of EGFR, which is ∼0.03 μm²/s (23), and D is the diffusion constant of the unbound SH2, which is ∼15 μm²/s (22). Using these numbers, and assuming TIR illumination depth a ∼ 0.5 μm, we can directly calculated the value of D_eff as a function of λ. We found that the calculated values from this equation agree well with our experimental data (Fig. 2C).

Spatial Distribution and Clustering of Target Phosphotyrosine Binding Sites.

The prominence of SH2 rebinding suggests a high density of potential p-Tyr binding sites (∼10³ sites/μm² based on Eq. 2). We asked whether this density is the average density of the whole cell or a local density elevated due to microscopic domains of the receptors clustering (24). To answer this question, we quantified the effects of the duration of EGF stimulation on the hazard function values and the effective mobilities of Grb2_SH2. Indeed, as the stimulation duration increased, the dwell time of Grb2_SH2 molecules increased as well (Fig. 3A and Fig. S5), presumably due to an increase in p-Tyr density. The number of Grb2_SH2 molecules recruited to the membrane typically reached a plateau at around 5–10 min of stimulation (Fig. 1B), suggesting that the number of p-Tyr sites reaches a maximum value at that time. Results from other assays, such as far-Western analysis (Fig. S6) and mass spectrometry (11), led to the same conclusion. Interestingly, however, the SH2 dwell time continued to increase and stabilized only after >20 min (Fig. 3A), suggesting that the rebinding of SH2 molecules was indeed responding to changes in local density of the binding sites. Even though the number and thus the average density of p-Tyr sites no longer increased after 10 min, our data suggest that the local density continued to increase, most likely by clustering of the EGFR receptors. As expected if the local density of binding sites were increasing, the mobility of the Grb2_SH2 molecule decreased correspondingly over time, presumably because they spend less time in cytosol before rebinding (Fig. 3B). The changes in the mobility and the changes in the dwell time are negatively correlated with each other (Fig. 3C). These results highlight the importance of the spatial organization (local density) of the receptors, as opposed to their total expression, in regulating the membrane dwell time of SH2 molecules. Supporting this idea, we found that in cells (H226) with much lower EGFR expression than A431, the hazard function and the dwell time of Grb2_SH2 are almost the same as in A431 cells (Fig. S7).

Fig. 3.

Temporal evolution of Grb2_SH2 turnover dynamics. (A) Average λ-values decrease over time. The hazard function was measured every 1.5 min after addition of EGF. The average λ-values between 0.1 s and 2.0 s were calculated. The decreasing values indicate that the apparent dissociation rate is lower for cells stimulated for a longer time. (B) Effective mobilities (D_eff) decrease over time. The effective mobilities were measured as in Fig. 2. (C) The decrease in effective mobility and the increase in survival time are correlated. (D) Average fluorescence intensity of Grb2_SH2 molecules at the cell’s plasma membrane by TIR microscopy. The image shows the overall spatial distribution of the Grb2_SH2 molecules. (E–H) Spatial distribution of the binding sites during various time delays after EGF stimulation. The spatial location of each SH2 binding event was recorded and plotted as a black dot at the spatial location. The plotted region corresponds to the box in D. The same number of molecules was plotted for E–H. (I) Same as in E–H except the cell was expressing mtGrb2_SH2. The cell was stimulated for 40 min. (Scale bars: 5 μm.)

To understand the spatial distribution of the phosphotyrosine binding sites in cells, we recorded the first observed locations of each individual SH2 molecule detected at the cell membrane and plotted their coordinates (Fig. 3 D–I). We found that when the cell was stimulated for a short period, no apparent spatial pattern could be discerned from the plot (Fig. 3E). For cells that were stimulated for a longer time (e.g., >20 min), however, we found spatial regions to which the SH2 molecules were repeatedly recruited with a higher-than-average frequency, indicating clusters of p-Tyr sites on activated EGFR (Fig. 3 F–H). No clustering was observed for mtGrb2_SH2 (Fig. 3I). This result further lends support to the argument that the decrease in hazard function values is correlated with clustering of the receptors.

Rebinding Dynamics of Multivalent SH2 Molecules.

Next, we investigated the turnover dynamics of multivalent SH2 molecules. Most signaling proteins are multivalent: They have multiple modular binding domains that allow interaction with more than one molecule at the same time. Some SH2 proteins, such as Shp2 or PLCγ1, have two SH2 domains and can thus be recruited to the membrane through either monovalent or bivalent binding to p-Tyr (Fig. 4A). We measured the hazard function and the effective mobility of the Shp2_SH2(NC) molecule (Movie S3), which has both the N- and the C-terminal SH2 domains of the Shp2 phosphatase. In contrast to the monovalent Shp2_SH2(N) or Shp2_SH2(C), the hazard function of Shp2_SH2(NC) is no longer a constant (Fig. 4B), presumably due to a heterogeneous mixture of both monovalent and bivalent complexes. The monovalent complex was formed when an SH2 molecule was initially recruited to the membrane, and the bivalent complex was formed later. Thus, we attribute the decay in the hazard function to the gradual process of bivalent complex formation. The decay time constant is on the order of a few hundred milliseconds, indicating that the formation of the second connection is relatively slow compared with the formation of the first. The difference is probably because once the SH2 molecule is tethered to the membrane through the first p-Tyr-SH2 interaction, its diffusion rate decreases significantly, which results in a reduced reaction rate for the formation of the second binding.

Fig. 4.

Rebinding of multivalent SH2 module Shp2_SH2(NC). (A) two different schemes of rebinding for bivalent SH2 molecules, depending on whether it forms a monovalent complex (Upper) or a bivalent complex (Lower). (B) The hazard functions for the Shp2_SH2(NC) module in an A431 cell stimulated by EGF (20 min), in comparison with its monomeric counterparts. The nonconstant λ-values for Shp2_SH2(NC) arise from the mixed population of monovalent complexes and bivalent complexes. The decay reflects the temporal process of bivalent complex formation. (C) MSD-Δt analysis of Shp2_SH2(NC) (Top), Shp2_SH2(N) (Middle), and Shp2_SH2(C) (Bottom). For each molecule, the MSD curves were calculated for the subpopulations with long (blue), intermediate (red), and short (black) survival times. Only Shp2_SH2(NC) exhibited different MSD curves depending on survival times.

We expect the monovalent complex to rebind via diffusion in cytosol, whereas the bivalent complex rebinds as a membrane-associated complex (Fig. 4A). Therefore, the rebinding model would predict that the molecules with longer survival time should have lower mobility than molecules with shorter survival time. Indeed with MSD-Δt analysis we found the mobility of molecules with long survival times (>5 s) is approximately eight to nine times less than that of molecules with short survival times (<1 s) (Fig. 4C and Fig. S8). Such a bifurcation in mobility was not observed for monomeric SH2 molecules (Fig. 4C and Fig. S8). Furthermore, the mobility of Shp2_SH2(NC) molecules with short survival times (<1 s) is roughly the same as that of their monomeric counterpart [Shp2_SH2(N)]. These results are consistent with the rebinding model. The hazard functions of two other bivalent SH2 molecules [PLCγ1_SH2(NC), which has the two naturally occurring SH2 domains of PLCγ1, and an engineered tandem Grb2 construct, Grb2_SH2-SH2] exhibited similar decaying values (Fig. S9), consistent with the above model.

Finally, we also measured the hazard function of full-length Grb2, in which the SH2 domain is flanked by SH3 domains that bind to proline-rich peptides in effectors such as Sos (Fig. 5). We found that the hazard function of full-length Grb2 exhibits very little decay, suggesting that the recruitment of Grb2 to the cell membrane depends predominately on the SH2 domain. That is, SH3-mediated interactions do not appear to be sufficient to tether Grb2 to the membrane. Consistent with this interpretation, we also found no significant difference in the recruitment kinetics between full-length Grb2 and the Grb2_SH2 domain (Fig. 5A), and a mutation in the SH2 domain alone (mtGrb2) was sufficient to prevent recruitment of full-length Grb2 to the cell membrane (Fig. 5A). We found that the average dwell time of full-length Grb2 was longer than that of Grb2_SH2, consistent with the previous finding that full-length Grb2 and Grb2/Sos complexes bind tighter to p-Tyr (17, 25). The flattened hazard function further indicates that the heterogeneities of multiple p-Tyr sites were averaged out, presumably due to multiple rebinding, indicating that the results we obtained from isolated SH2 domains can be extrapolated to full-length SH2 proteins.

Fig. 5.

Turnover kinetics of full-length Grb2 molecules. (A) Comparison of the recruitment kinetics of full-length Grb2, full-length mtGrb2, and Grb2_SH2 in A431 cells. Data were collected from A431 cells as in Fig. 1B. The amount of the recruitment was measured by integrated fluorescence intensity under TIR illumination after cells were treated with 25 ng/mL EGF. (B) Hazard functions of full-length Grb2 and Grb2_SH2 in A431 cells. Similar to that of Grb2_SH2, the hazard function of full-length Grb2 is a constant. The lower average value for full-length Grb2 signifies a longer average dwell time, consistent with a more stable binding to p-Tyr sites.

Discussion

In this study, we measured the apparent dwell time of SH2 domains within a TIR illumination field near the cell membrane. We found the average dwell time is longer than predicted on the basis of the chemical dissociation rate, and for monomeric SH2 domains the dissociation kinetics exhibited little heterogeneity. Although other possibilities cannot be completely excluded, the simplest and most likely explanation is that the interaction between SH2 domains and p-Tyr sites involves repeated rebinding. We propose that the rebinding is the result of competition between the very fast intrinsic chemical kinetic rate of the association reaction and the relatively slow diffusion rate of the biomolecules in vivo. Part of the reason that the chemical association rate is high is that there is an abundance of available p-Tyr binding sites, because each receptor molecule has multiple phosphorylation sites and because of the clustering of the receptors that further increases the local concentration of binding sites. Because the SH2 molecules do not directly dissociate from the receptors in our model, the traditional concept of turnover time has an ambiguous meaning in this case. There are, in fact, two different turnover times. The SH2 turnover time on an individual p-Tyr site is quite short, because the molecule quickly dissociates. The fast dissociation is consistent with the fact that p-Tyr sites can be very quickly dephosphorylated (26) and thus are not protected by the SH2 binding for a very long time. On the other hand, the SH2 turnover time with respect to the receptor cluster is relatively long, because the SH2 molecule “hops” between the binding sites. Theory predicts that clustering stabilizes the binding between p-Tyr sites and the SH2 molecules, which we verified in experiments.

The finding that the SH2 protein hops between clustered binding sites along the cell membrane suggests that a single SH2 molecule, when recruited to the membrane, could potentially interact with multiple p-Tyr sites that are near each other. This kind of mechanism may also play a role in other intracellular protein interactions. For example, studies of Cdc4 interaction with Sic1 revealed that Cdc4 can interact with multiple phosphorylated sites on Sic1 even though Cdc4 contains only one interaction site (27). Together with our findings, it seems plausible that dynamic equilibrium through multiple rebinding events is a common theme of phosphoprotein interactions. The rebinding mechanism lessens the importance of single high-affinity binding sites. In terms of RTK signaling, the SH2 dwell times depend on the average affinities of many phosphorylated sites; lack of a single high-affinity site will have only a moderate impact on the average duration of binding. Therefore, the rebinding mechanism may explain why mutation of RTK phosphosites predicted to bind specifically to particular SH2 domains had relatively little effect on downstream signaling. It was only when all sites were eliminated and individual phosphosites added back that specific effects on downstream signaling could be observed (28, 29).

Finally, these results underscore the importance of receptor clustering, as kinetic stabilization of the SH2/p-Tyr interaction depends on high local density of phosphorylated sites. Receptor clustering may further affect the downstream outcome of the signaling process (30). For example, it was proposed (31) that in T cells, prolonged ligand/receptor association suppresses “noise” from adventitious receptor activation (a kinetic proofreading mechanism). In a similar manner, the increased SH2 protein dwell time near the cell membrane, coupled to high in vivo dephosphorylation rates (26), may serve to dampen spurious noise associated with nonspecific binding of SH2 proteins to membrane molecules. Finally, because the recruited SH2 proteins do not associate with just one p-Tyr site, but instead “glide” over many p-Tyr sites, the chemical interaction of such an SH2 protein with downstream substrate or effector molecules at the cell membrane may not follow simple chemical kinetic laws. The details of such interactions should be an interesting subject of future research.

Experimental Procedures

Constructs and Cell Culture.

SH2 domain DNA sequences were obtained from the Human SH2 collection (Open Biosystems) and subcloned to the C terminus of tdEos coding sequence to form the chimeric fusion constructs. The myr-tdEos clone contains the myristoylation signal sequence from the N terminus of Src. A431 and H226 cells were maintained in DMEM culture medium and replated on freshly cleaned glass-bottom dishes (Matek) at medium to low density (<50% confluent) right before microscopy experiments.

Microscopy and Image Analysis.

All imaging experiments were carried out on a custom-made total-internal-reflection fluorescence microscope, equipped with a 532-nm laser (Crystal Lasers) for fluorescence excitation and a 405-nm laser (Coherent) for photoactivation. Fluorescent spots corresponding to single molecules were detected via intensity thresholding. The positions of the molecules were tracked over time, using the nearest-neighbor method. The scan area used during tracking when searching for the position of the molecule in the next time frame was chosen on the basis of estimated diffusion rate of the molecule so that the molecule has a >99.5% chance to be within the scan area. The spatial density of the fluorescence spots was kept low (0.05–0.1/μm²) to avoid mistracking. To estimate the effective mobility, ensemble averaged MSD was calculated as a function of time lags (Δt), which was subsequently used to determine the effective diffusion constant.

More details of the experimental procedure are included in the SI Experimental Procedures.