Open AccessCCS ChemistryRESEARCH ARTICLE5 Sep 2022Probing Allosteric Modulation of Membrane Receptor in the Native State by Data Mining-Integrated Tracking Microscopy Bin Xiong, Jia Wu, Jinhui Shang, Yancao Chen, Yan He, Xiao-Bing Zhang and Weihong Tan Bin Xiong *Corresponding author: E-mail Address: [email protected] Molecular Science and Biomedicine Laboratory, State Key Laboratory of Chemo/Biosensing and Chemometrics, College of Chemistry and Chemical Engineering, Hunan University, Changsha 410082 Google Scholar More articles by this author , Jia Wu Molecular Science and Biomedicine Laboratory, State Key Laboratory of Chemo/Biosensing and Chemometrics, College of Chemistry and Chemical Engineering, Hunan University, Changsha 410082 Google Scholar More articles by this author , Jinhui Shang Molecular Science and Biomedicine Laboratory, State Key Laboratory of Chemo/Biosensing and Chemometrics, College of Chemistry and Chemical Engineering, Hunan University, Changsha 410082 Google Scholar More articles by this author , Yancao Chen Molecular Science and Biomedicine Laboratory, State Key Laboratory of Chemo/Biosensing and Chemometrics, College of Chemistry and Chemical Engineering, Hunan University, Changsha 410082 Google Scholar More articles by this author , Yan He Department of Chemistry, Tsinghua University, Beijing 100084 Google Scholar More articles by this author , Xiao-Bing Zhang Molecular Science and Biomedicine Laboratory, State Key Laboratory of Chemo/Biosensing and Chemometrics, College of Chemistry and Chemical Engineering, Hunan University, Changsha 410082 Google Scholar More articles by this author and Weihong Tan Molecular Science and Biomedicine Laboratory, State Key Laboratory of Chemo/Biosensing and Chemometrics, College of Chemistry and Chemical Engineering, Hunan University, Changsha 410082 Google Scholar More articles by this author https://doi.org/10.31635/ccschem.021.202101541 SectionsSupplemental MaterialAboutAbstractPDF ToolsAdd to favoritesDownload CitationsTrack Citations ShareFacebookTwitterLinked InEmail As a general mechanism for governing the bioactivity of membrane receptors, allosteric modulation is critical in cell signaling and cell communication but remains difficult to measure in situ. Herein, we introduce a data mining-integrated tracking microscopy (DMITM) to investigate allosteric modulation of membrane receptors in the native state in live cells. Using K-means clustering-based hidden Markov modeling to uncover the ligand binding and unbinding events with diffusivity variations of ligand-conjugated nanoprobes as observations, DMITM enables the analysis of receptor activity modulation from the viewpoint of ligand binding kinetics and free energy landscape simultaneously. As a demonstration, we applied this technique to uncover the details of how Ca2+ and Mn2+ function in site-specific allosteric regulation of integrin activity in live cells. It was found that both the receptor accessibility to ligands and the stability of integrin-ligand complex can be synchronously modulated upon allosteric stimulation. Interestingly, the stability change contributes more to the allosteric effect as compared with the accessibility variation, suggesting a differential regulation paradigm in the site-specific regulation of integrin activity. This technique provides a new opportunity to study protein activity regulation and uncover the underlying mechanisms in live cells. Download figure Download PowerPoint Introduction Allosteric modulation of membrane receptors by various metal ions and small-molecule modulators has been considered a general mechanism for the regulation of receptor activity and function in live cells.1 The occupation of an allosteric site by a modulator can trigger conformational change of membrane receptor,2 leading to variation in the biological activity of membrane receptors toward orthosteric ligands.3,4 Such a paradigm for receptor activity modulation plays a significant role in cell signaling and cell communications,5–7 which are involved in various diseases such as neurological diseases, cardiovascular disorders, cancers, and so on.1,8–10 Therefore, investigating the allosteric modulation of membrane receptor activity is critical to understanding the underlying molecular mechanisms of receptor function and regulation in live cells. Previous investigations of allosteric modulation of membrane receptors mainly focused on the characterizations of the allosteric regulatory site of receptors as well as corresponding conformational changes with techniques such as cryogenic electron microscopy (Cryo-EM), X-ray diffraction (XRD), nuclear magnetic resonance (NMR) spectroscopy, and fluorescence resonance energy transfer (FRET).11–14 These techniques have advanced the fundamental understanding of receptor regulation and underlying mechanisms by analyzing the structure of receptor–modulator complexes, enabling us to approach the next question of how these allosteric modulators regulate the receptor activity in live cells. To this end, the flow cytometry assay has previously been employed for the analysis of allosteric modulation by simply detecting the changes in receptor affinity toward the ligands,15,16 which can provide limited information about the kinetics and thermodynamics of receptor responses. Hence, the in-depth investigation of allosteric modulation of membrane receptors in the native state remains challenging due to the lack of techniques that enable the in situ analysis of receptor activity modulation in live cells. Single-particle tracking techniques in combination with robust data processing methods have exhibited great promise for the investigation of biomolecular interactions at the single-molecule level owing to their capability of monitoring ongoing biological events with unprecedented resolution.17–24 In this work, we report a data mining-integrated tracking microscopy (DMITM) to study the activity modulation of membrane receptor in the native state in live cells by analyzing the changes in receptor responses toward the orthosteric ligand from the viewpoint of binding kinetics and free energy landscape. To achieve the in situ analysis of activity modulation of membrane receptor, four critical steps were involved in the analytical framework of DMITM technique (Figure 1). First, using the ligand-conjugated gold nanoparticles (AuNPs) as optical probes, the diffusion dynamics of single probes over the nanoparticle-cell interaction processes were tracked under a time-lapse darkfield microscope at nanometer-level localization precision and millisecond-scale time resolution ( Supporting Information Figure S1). Second, the underlying ligand binding and unbinding interactions were uncovered with the K-means clustering-based hidden Markov modeling (KHMM) by interpreting the probe diffusivity transitions over the interaction process with single receptors. Third, the kinetic rates were extracted from single-molecule kinetics analysis for the binding and unbinding interactions, respectively. Finally, the allosteric modulation of receptor activity was determined according to the changes in receptor responses toward the ligand from the viewpoint of ligand binding kinetics and free energy landscape simultaneously. As a proof of concept demonstration, we applied this DMITM technique for in situ analysis of metal-ion-dependent adhesion site (MIDAS) involved integrin activity modulation in single live cells. Upon the allosteric stimulation by the treatment of Mn2+ addition or Ca2+ depletion, integrin exhibited a moderately enhanced ligand binding rate and a notably suppressed ligand unbinding rate, indicating that Mn2+ acts as an allosteric activator but Ca2+ (at the millimolar level) plays a role of allosteric inhibitor for integrin activity regulation, respectively. The allosteric inhibitor reduced the energy barrier of dissociation interaction by ∼0.59 kBT and heightened the energy barrier of association interaction by ∼0.43 kBT. On the contrary, the allosteric activator raised the energy barrier of dissociation interaction by ∼0.74 kBT and lowered the energy barrier of association interaction by ∼0.30 kBT. The changes in the free energy landscape due to allosteric modulation measured by single-particle analysis were further validated by bulk measurements with the binding affinity assay. From the analysis of receptor responses from the viewpoint of ligand binding kinetics and free energy landscape, a differential regulation paradigm for MIDAS-involved allosteric modulation of integrin-ligand complex stability and receptor accessibility to the ligands was uncoverd. These demonstrations suggest that this DMITM technique could provide a new opportunity for investigating protein activity regulation by active molecules (e.g., drug, cofactor, and small-molecule modulator) and uncovering the underlying mechanisms in live cells. Figure 1 | Analytical framework of DMITM for probing allosteric modulation of membrane receptor activity in live cells by analyzing receptor responses from the viewpoint of ligand binding kinetics and free energy landscape simultaneously. Download figure Download PowerPoint Experimental Methods Synthesis of AuNPs The AuNPs used in this study were prepared with a seed-mediated growth method.25 First, the 18 nm AuNPs were obtained according to the classical Frens method and used as seeds for subsequent growth procedures. Then, 0.27 mL of 24.28 mM HAuCl4 was added into 50 mL stocking solution containing 1 mL of seed solution, followed by dropwise addition of 0.12 mL ascorbic acid (0.1 M) under stirring. The reaction lasted 1 h until the color change was completed. The obtained AuNPs were characterized with UV-vis spectroscopy (Shimadzu UV-1800, Kyoto, Japan) and TEM (JEM 1230, JEOL, Tokyo, Japan). Nanoprobe preparation and characterization For the preparation of ligand-coated nanoprobes, 1 mL of the as-prepared AuNPs (10 pM) was incubated with 10 μL of mixture containing CH3O(CH2CH2O)16CH2CH2SH (PEG16, 150 μM) and HS(CH2CH2O)42CH2CH2NH2 (PEG42, from 2.5 to 10 μM) overnight. After removing the residual reagents by centrifugation, the AuNPs were redispersed in 10 mM phosphate buffer (PB) (pH 7.4), followed by adding 10 μL of 5 mM 3-maleimidopropionic acid N-hydroxysuccinimide ester (MBS, freshly prepared and dissolved in 10 mM of PB buffer with pH 7.4) for an incubation of 6 h. After centrifugation, the functionalized AuNPs were redissolved in water containing 5 mM triethylamine, followed by the addition of 10 μL of 5 mM cyclic Arg-Gly-Asp-D-Phe-Lys-(Cys) (cRGD) or cyclic Gly-Gly-Gly-Gly-Gly-Lys-(Cys) (cGGG) peptide solution for incubation of 6 h. The surface-functionalized AuNPs were characterized with UV–vis spectroscopy and dynamic light scattering (Nano ZS, Malvern, Worcestershire, UK). To examine the ligand number on single AuNPs, a rhodamine-labeled cRGD peptide was used to replace the unlabeled ligand during the nanoprobe fabrication processes. By plotting the fluorescence of rhodamine-labeled cRGD peptide at different concentrations as the calibration curve, we determined the number of ligands on the surface of single nanoparticles by measuring the fluorescence intensity after dissolution of the AuNPs using excess potassuim cyanide.25 To ensure the cRGD ligands with sufficient density and reduced steric effect on the nanoparticle surface, the concentration of PEG42 during the binary polyethyleneglycol (PEG) modification process was selected as 5 μM to prepare nanoprobes for imaging experiments. Cell culture and cell imaging A human cervical cancer (HeLa) cell line was obtained from the American Type Culture Collection (ATCC, Virginia, USA). The cells were maintained in Dulbecco’s modified Eagle’s medium (Gibco) and supplemented with 10% fetal bovine serum (Gibco) at 37 °C, 5% CO2 in a humidified atmosphere. The cells were cultured on a cleaned coverglass in a plastic cell culture dish. To investigate the allosteric modulation of integrin activity by single-particle tracking, the cell-attached coverglass was transferred onto a concave microscope slide with cell culture medium in the absence or presence of allosteric modulation Ca2+ depletion by C14H24N2O10 (EGTA, 2 mM) or Mn2+ addition (1 mM), respectively. After the addition of [email protected] (10 pM), the sample was observed under the time-lapse darkfield microscope, consisting of a darkfield microscope (Nikon 80i, Tokyo, Japan) and a color CCD camera (Olympus, DP72, Tokyo, Japan). The exposure time was 70 ms in the imaging experiments. After focusing on the cell surface, the rotating knob of the darkfield microscope was locked to avoid image drift in our single-particle tracking experiments. All the nanoprobe trajectories in the sequential darkfield images were analyzed with ImageJ software. To test the activity of the prepared nanoprobes, the cells were separately incubated with [email protected], [email protected], and [email protected] at a final concentration of 1 pM. The inhibitory experiment was performed by adding 1 mM of cRGD into the cell culture medium, followed by the addition of [email protected] Nanoparticles captured by the cells at different times from 30 to 90 min were examined with darkfield microscopy. The nanoprobe-based cell incubation assay was conducted by separately incubating the [email protected] and [email protected] with cells for 1 h in the presence of EGTA (2 mM) or Mn2+ (1 mM). The allosteric effect of the modulators on mediating integrin activity was then examined with darkfield microscopy. Classical MSD analysis and confinement area calculation The motion of individual nanoparticles when interacting with cells was recorded by time-lapse darkfield microscopy, and the corresponding trajectories were extracted with the Particle Tracker plug-in in ImageJ. The mean-square displacement (MSD) for each trajectory was then calculated as17 MSD ( n τ ) = 1 N − n ∑ i = 1 N − n [ ( x ( i + n ) τ − x i τ ) 2 + ( y ( i + n ) τ − y i τ ) 2 ] (1) where x and y are the centroid position of the particle, nτ is the time lag for calculating MSD, and τ is the acquisition time. The confined diffusion coefficient (D) was then obtained by fitting the MSD curve with the power law MSD ( n τ ) = 4 D ⋅ ( n τ ) α , where the exponent α indicates the nonlinear relationship of MSD with time. The square displacement (SD, r t 2 ) was calculated according to the transient displacement of nanoprobes, r t 2 = ( x n τ − x ( n − 1 ) τ ) 2 + ( y n τ − y ( n − 1 ) τ ) 2 (2)where v x and v y are the partial velocity along coordinate axes. According to previous reports,26,27 the characteristic confinement length (L) for a confined diffusion can be extracted by fitting the MSD with the expression, MSD ( t ) = L 2 3 ( 1 − exp ( − t m τ ) ) (3)where the τm represents the equilibration time. With the characteristic confinement length, the equivalent diameter (d) can be then calculated according to the equation, d = 2 3 L . Thus, the characteristic confinement area of individual particles at the confined diffusion state can be calculated by Sc = πd2/4. KHMM The KHMM algorithm for hidden state annotations using the SD value as observation,20 was implemented in Python 3.8.3 in our work. The Python codes are available from the corresponding author upon request. The detailed description for the theoretical framework of KHMM can be found in the Supplementary Note 1 in Supporting Information. To evaluate the robustness of this data mining technique, we analyzed the trajectories with designated diffusive states and calculated the accuracy of state assignment and dwell time determination. Single-molecule kinetics and free energy landscape analysis To conduct the analysis of binding dynamics, the states with dwell time less than 0.21 s (3 frames) possibly due to occasional perturbation or collision with pericellular matrix were not considered as bound or unbound states and not counted in the dwell time distribution analysis. With the extracted binding and unbinding events from the nanoprobe diffusivity changes as well as the corresponding durations, the rate constants of ligand binding or unbinding interaction can be obtained by fitting the probability as a function of time. The theoretical investigations of single-molecule kinetics as well as modulator-induced changes in the free energy landscape are described in the Supplementary Notes 2 and 3 in Supporting Information. Investigating the allosteric modulation of integrin activity from the binding affinity assay The cells in suspension were first incubated with rhodamine-labeled cRGD peptides at different concentrations from 0 to 20 μM for 1 h in the presence of EGTA (2 mM) or Mn2+ (1 mM), respectively. After removing the residual ligands by centrifugation three times, the cells were washed with Dulbecco’s phosphate buffered saline, and the fluorescent intensity was then measured by multimode microplate reader (SpectraMax iD3, San Jose, USA). By fitting the binding curves of cRGD of fluorescently labeled cRGD to integrin on the cell surface, the affinity constants were obtained. Then the variation in Gibbs free energy change for receptor–ligand interactions under allosteric modulation by Ca2+ deletion and Mn2+ addition were determined according to the relative changes in affinity constants. Results and Discussion Tracking the dynamic responses of the membrane receptor to the ligand-coated nanoprobe in live cells As a proof of concept, the αvβ3 integrins that function as adhesion receptors in cells were selected as a common model, and the cRGD peptides were used to mimic the cell-adhesive ligands.25,28 A nonspecific cGGG peptide was used as the reference in this study. For the particle tracking studies in this work, AuNPs, which have strong light scattering at the resonance frequency and enable long-term imaging under a darkfield microscope,29–32 were used to construct the nanoprobes ( Supporting Information Figure S2). The cRGD and cGGG peptide ligands were separately conjugated onto the surface of AuNPs using the PEG linker at a moderate density (∼625 peptide ligands per AuNP, Supporting Information Figures S3 and S4 and Table S1), where the PEG linker could prevent the ligand denaturation from nonspecific absorption of ligands on the nanoparticle surface in various biorelated applications.33–35 After being incubated with the cells, the as-prepared nanoprobes exhibited specific interactions with integrins on the cell membrane ( Supporting Information Figure S5), suggesting the nanoprobes are bioactive and can be used for investigating the receptor responses under darkfield microscopy. Since the nanoprobes are multivalent, we first attempted to resolve the complicated interactions between nanoprobes and membrane receptors by analyzing the diffusive behaviors of individual nanoprobes. After adding the nanoprobes into the culture medium, we monitored the diffusion dynamics of single nanoprobes over the whole interaction process with integrins on the apical surface of adherent cells. By plotting the nanoprobe trajectories over the integrin-ligand interaction process, we found that the nanoprobes displayed confined diffusive behaviors after approaching the cell membrane from the culture medium, and the trajectory boundary was gradually reduced (Figure 2a and Supporting Information Movie S1). From the representative time courses of local diffusive behaviors for the nanoprobes, notable stepwise decrease in diffusivity attributable to the multiple integrin-ligand pairs binding could be observed (Figure 2b and Supporting Information Figure S6), which was consistent with the multiple-receptor-mediated endocytosis of nanoparticles as reported previously.36,37 A remarkable difference in transient diffusivity histograms of the nanoprobes at those stages could also be detected (Figure 2c and Supporting Information Figure S6). All these results suggest that the changes in the diffusivity and the degree of confinement (DOC) of the nanoprobes over the multivalent interaction processes are associated with the number of receptor–ligand complexes.38,39 Therefore, the complicated interaction of a nanoprobe with single or multiple receptors on the cell surface can be resolved from single-particle diffusivity analysis over the entire interaction process, as captured under tracking microscopy. Figure 2 | Resolving the dynamic interactions between nanoprobe and multiple receptors in live cells. (a) Tracking the diffusive behaviors of nanoprobes on cell membrane under the time-lapse darkfield microscope. The insert shows the trajectory of representative nanoprobe highlighted with purple circle. The stepwise decrease in the SD fluctuations (b) and corresponding SD distributions (c) of the representative nanoprobe correlate with the multiple receptor mediated endocytosis processes after approaching cell membrane (stages 1–3) from culture medium (highlighted with yellow in the time courses of SD fluctuation), implying the complicated interactions of a ligand-coated nanoprobe with single or multiple receptors on cell surface can be discriminated from single-particle diffusivity analysis. Download figure Download PowerPoint Uncovering the transient ligand binding and unbinding to single receptor with data mining technique To investigate the receptor activity regulation, we focused on the analysis of dynamic interactions between the nanoprobe and single membrane receptors. As revealed by analyzing the trajectory boundary and transient diffusivity fluctuations (Figure 3a and Supporting Information Figure S7 and Movies S2 and S3), the nanoprobes experienced repetitive transitions between two distinct diffusive states when they started to interact with single membrane receptors. However, such phenomenon could not detected for other cases including thermal fluctuations ( Supporting Information Figure S8), and nonspecific interactions with the pericellular matrix or cell membrane ( Supporting Information Figures S9 and S10 and Movies S4 and S5). Moreover, the simulation results also indicated that the binding of the ligand-coated nanoprobe to a receptor anchored on a two-dimensional (2D) membrane led to a remarkable decrease in the diffisivity of nanoparticles ( Supporting Information Figure S11). Hence, the characteristic transition of nanoprobe diffusive behavior should correlate with transient ligand binding and unbinding interactions to membrane receptor, making it possible to probe the dynamic responses of single receptors to the ligand-coated nanoprobes from single-particle diffusivity analysis. Figure 3 | Uncovering the transient ligand binding and unbinding to single receptor. (a) Darkfield images for the representative nanoprobe over the interactions with integrin on cell membrane at different time. The green square inserts are the corresponding trajectories of the nanoprobe highlighted with green cycle. (b) KHMM analysis for uncovering transient binding and unbinding interactions using the SD as observations. (c) Changes in the slope of cumulative square displacement (CSD). Analysis of diffusion coefficient (d) and characteristic confinement length (e) of individual nanoprobes suggested the distinct diffusive behaviors at the unbound and bound states. Download figure Download PowerPoint To uncover the hidden stochastic binding–unbinding interactions underlying the transient diffusivity fluctuations, we proposed a data mining technique, called the KHMM method on the basis of previous reports.40–42 This algorithm achieves the diffusivity clustering according to difference in Euclidean distance from cluster centres and implements the hidden state annotations on the basis of the maximum likelihood estimate ( Supplementary Note 1). Using the simulated time-series diffusivity observations containing two assigned diffusive states, we demonstrated the robustness of the KHMM algorithm for tagging the hidden states as well as the corresponding durations with superior accuracy (∼98.7% as the diffusion coefficient ratio for the two states up to 3, Supporting Information Figures S12 and S13). Then we applied the KHMM analysis to unveil the most likely bound–unbound state evolutions for the dynamic interactions between the ligand-coated nanoprobe and membrane receptor (Figure 3b). The state transitions resolved from KHMM analysis were well matched with the changes in the slope of cumulative step length for the diffusive motion of the same nanoprobe (Figure 3c). Further, the diffusion coefficient of nanoprobes at the resolved unbound and bound states are ∼0.0143 and ∼0.0048 μm2/s (Figure 3d), as compared with the measured diffusion coefficient (∼0.01 μm2/s) of individual integrin receptors inside the confined zone on cell surface.43 That means the nanoprobes displayed faster diffusivity at the unbound state than that of the receptor itself, but the diffusion rate declined remarkably after the nanoprobes bound to the receptor due to the increased particle confinement. Both the remarkable difference in the diffusion coefficients ( D 2 / D 1 ≈ 3 ) and the characteristic confinement lengths at the resolved bound and unbound states also confirmed that the state annotations by KHMM are reliable (Figures 3d and 3e). Given the capability of KHMM analysis to uncover the underlying bound and unbound states as well as the dwell time of each state, the ligand binding and unbinding interactions can be dissected separately, making it possible to analyze the receptor activity regulation from the viewpoint of ligand binding kinetics. Investigating the allosteric modulation of receptor activity from ligand binding kinetics analysis Since the dwell time reflects a window of opportunity for observing the binding or unbinding interactions, the dwell time distributions for the unbound and bound states allow us to figure out the time-dependent association and dissociation probability, respectively. To understand the underlying mechanism of state switching and dwell time of individual states, we sought to investigate the single-molecule binding kinetics for receptor–ligand interactions. The theoretical expressions of association probability (Pa) and dissociation probability (Pd) as a function of time for ligand binding and unbinding interactions can be described as P a = e − k a ⋅ [ N L / ( N A ⋅ S C ) ] t and P d = e − k d ⋅ t ( Supporting Information eqs S10 and S13 in Supplementary Note 2), where ka and kd are the 2D binding and unbinding rate constants, respectively, NL is the number of ligands on the surface of individual AuNPs, NA is the Avogadro constant, and SC represents the area of confined zone in which the nanoprobe is accessible to bind with individual integrins. After calculating the time-dependent association and dissociation probability from the dwell time distributions at unbound and bound states, the binding and unbinding rate constants can then be extracted by fitting the corresponding time-dependent association and dissociation probability with the theoretically derived expressions for binding kinetics analysis, respectively. Hence, using single-molecule kinetics analysis, DMITM enables the analysis of receptor activity modulation from the changes in kinetic rates for receptor–ligand binding and unbinding interactions simultaneously. Previous studies have uncovered that the MIDAS can be involved in allosteric modulation of integrin function by divalent metal ions (e.g., Ca2+ and Mn2+) using different approaches such as theoretical calculation, X-ray crystallography, immunofluorescence analysis, isothermal calorimetry, and surface plasmon resonance analysis.44–48 Still, the understanding of integrin activity regulation as well as its underlying mechanism remain obscure. To address these questions, we implemented the DMITM technique to uncover the metal ions associated allosteric modulation of the αvβ3 integrin recep