Nanoparticle Tension Probes Patterned at the Nanoscale:Impact of Integrin Clustering on Force Transmission

Because cell membrane receptorsreside at the interface between a cell and its external surroundings,these molecules have evolved to sense and transduce both chemicaland physical cues with high fidelity. One common mechanism to triggeror modulate surface receptor activation involves ligand-induced clustering,which in turn contributes to a robust biochemical response. For example,T-cell receptors,^1,2 Fc-ε receptors,³ EphA2 receptor tyrosine kinases,⁴ Notch receptors,⁵ and epidermalgrowth factor receptors (EGFR)⁶ oligomerizeat the plasma membrane upon activation. Intriguingly, the majorityof oligomerized ligand–receptor complexes subsequently couplewith the cytoskeleton and are actively transported by retrograde flow.^4,7,8 Many of these assemblies experienceresistance during active transport through the plasma membrane, resultingin mechanical tension that is likely to modulate signaling outcome.Accordingly, the coupling between receptor clustering, mechanicaltension, and signal transduction at the plasma membrane is receivingincreased interest.⁹⁻¹³ However, one of the greatest challenges in this area pertains tothe lack of methods that can control clustering while also reportingon molecular tension during the signaling activity of live cells.In this paper, we report the development and application of an approachto address this need, thus showing the ability to simultaneously controlreceptor clustering with nanometer precision while also recordingreceptor mechanical forces with pN force resolution.

Integrinsare α-, β- heterodimeric cell surface receptorsthat span the plasma membrane and recognize specific ligand moleculeswithin the extracellular matrix (ECM).¹⁴ At the sites of activated integrin receptors, hundreds of differentstructural and adapter proteins will assemble into a three-dimensionalcross-linked structure that spans many microns in length and is describedas the focal adhesion (FA).¹⁵ Importantly,FAs regulate many cellular processes such as migration, differentiation,and proliferation.¹⁶ By providing a physicallinkage bridging the FA and the ECM, integrins experience mechanicalforces that are exerted by the cell and countered by the ECM.¹⁷ These forces play critical roles in integrinfunction and activation, allowing cells to recognize and respond tospecific physical features of their microenvironment.^18,19 Another essential parameter that modulates integrin activation isligand spacing, where it has been shown that the interligand spacingneeds to be at least 60–70 nm in order for FA maturation toproceed.²⁰⁻²² Therefore, it seems intuitive to conclude that thereis an intimate connection between integrin receptor clustering andforce transmission²³ but the details remainobscure.

To visualize the forces exerted by cell surface receptors,we recentlydeveloped molecular tension fluorescence microscopy (MTFM), whichgenerates pN-range force maps with high spatial and temporal resolution.²⁴ The probe in MTFM is comprised of a flexiblelinker molecule flanked by a donor fluorophore and ligand at one terminusand a quencher or a second fluorophore at the other terminus. TheMTFM probe is typically immobilized onto a surface, such that theflexible linker is in a collapsed resting state, where the donor fluorophoreis highly quenched. Upon the application of mechanical tension, thelinker is extended, thus separating the fluorophore from the quencherand increasing the fluorescence intensity by 10–20-fold. Recently,we developed integrin-specific MTFM probes by conjugating a fluorescentcyclized Arg-Gly-Asp-dPhe-Lys (c(RGDfK)) peptide at one terminus ofa polyethylene glycol (PEG)-linker and using a thiol at the otherterminus to immobilize this probe onto a 15 nm gold nanoparticle (AuNP).²⁵ The AuNP provides a physical scaffold to anchorthe probe and also efficiently quenches fluorescence when the PEGis in a relaxed conformation. To our surprise, we also found thatintegrin-specific MTFM probes immobilized through biotin–streptavidinare ruptured due to integrin-mediated forces that dissociate the biotin–streptavidincomplex.²⁶ Therefore, the use of thiol-goldbinding minimizes the possibility of force-induced dissociation ofthese probes. To the best of our knowledge, MTFM²⁴ and its recent variants²⁵⁻¹⁰⁰ provide the only known methodto visualize the pN forces exerted by cell surface receptors.

To test the relationship between integrin tension and ligand spacing,it is necessary to nanopattern MTFM probes. In principle, this couldbe achieved using a number of methods ranging from microcontact printing²⁸ to dip-pen nanolithography²⁹ and e-beam lithography.³⁰ However,few approaches are amenable to rapid prototyping, soft matter patterning,high-end fluorescence microscopy compatibility, as well as sub-10nm resolution registry over cm² areas.^31,101 Fulfilling these requirements is needed to control integrin spacingat the relevant nanometer length scales, while also providing thethroughput necessary to study the inherent heterogeneity of cellsand to permit simultaneous biochemical analysis.

Addressingthis need, Spatz and colleagues have developed elegantmethods to nanopattern RGD ligands.³² Thisapproach, named block copolymer micellar nanolithography (BCMN), generatesarrays of immobilized AuNPs with tunable spacings that range from∼30 nm up to ∼300 nm across the entire substrate (cm-scales).RGD-decorated AuNPs confine the integrin receptors to minimum distancesdefined by the interparticle spacing. Herein, we combine MTFM withBCMN (Scheme 1) to provide evidence showingthat the mechanism of ligand density sensing is force-mediated; onlysufficiently spaced ligands (<60–70 nm) allow for the transmissionof myosin-generated tension to individual integrin receptors.

Scheme 1

Combining Block Copolymer Micellar Nanolithography (BCMN) with AuNP-BasedMolecular Fluorescence Tension Microscopy (AuNP-Based MTFM) for Investigatingthe Role of Ligand Density in Modulating Integrin Tension

(a) Illustration showing theprocedure used to generate BCMN patterned MTFM tension probes. (b,c)Schematic showing the expected mechanism of how cell-generated forcesactivate the tension probe. (d) Chemical structure of the MTFM tensionprobe ligand that was synthesized (see, SupportingInformation) and used in this work.

Given that the 50 and 100 nm interparticle spacings are known topromote and destabilize FA formation, respectively, we tuned the dip-coatingspeed in BCMN to generate substrates with these AuNP spacings.²⁰ The hexagonal arrangement of the AuNP patternas well as the interparticle distances and heights were evaluatedby atomic force microscopy (AFM) (Figure 1a,b) and scanning electron microscopy (SEM) (Figure S1, Supporting Information). On the basis of thisanalysis, the AuNPs height was 8.4 ± 1.0 nm (Figure S2, Supporting Information), and the spacing on thetwo types of substrates was determined to be 99 ± 12 nm and 49± 9 nm.

Figure 1

(a,b) Representative AFM topography images of BCMN-patterned9nm AuNP arrays with a mean spacing of 99 nm ± 12 nm and 49 nm± 9 nm. Scale bar, 500 nm. (c) NSET calibration plot showingquenching efficiency of Cy3B fluorophore as a function of distancefrom AuNP surface as set by a range of DNA duplexes (Table S1, Supporting Information). The data was fit tothe R⁴ NSET model and d₀ (50% quenching distance) was determined to be 14.5 ±0.5 nm. (d) Theoretical plot showing the change in fluorescence asa function of applied tension based on combining the WLC and NSETmodels. The dynamic range of the probe corresponds to quenching efficiencyvalues ranging from 90 to 10%.

Although it is known that AuNPs with greater diameters aremoreefficient at quenching fluorescence,³³ weavoided larger AuNPs because of the potential for multiple integrinbinding to each particle. On the basis of structural data and previousliterature reports, we selected 9 nm AuNPs because this is the mostefficient quenching AuNP that would still ensure a maximum of oneintegrin engaged to each particle.³² Accordingly,we measured the nanometal surface energy transfer (NSET) radius betweena 9 nm AuNP and Cy3B dye and found this to be 14.5 ± 0.5 nm byusing highly packed duplex DNA as a molecular ruler (Figure 1c, Table S1 and Figure S3, SupportingInformation). This NSET radius is in agreement with a valueof 14.7 nm that was reported for 8 nm AuNPs and Cy3B dye.³³ Note that the Cy3B dye was used in this workto leave the enhanced green fluorescent protein (eGFP) fluorescencechannel available for genetically encoding markers of FAs. On thebasis of this NSET radius and the predicted wormlike chain (WLC) modelof PEG₈₀,³⁴ we would expecta force dynamic range that saturates at 27 pN, assuming the abilityto detect quenching efficiency values from 90 to 10% (Figure 1d).

To prevent nonspecific protein adsorptionand cell binding, theplasma-treated AuNP array substrate was passivated using a 0.1 mg/mLsolution of poly(l-lysine)-graft-poly(ethyleneglycol) (PLL-g-PEG)(PLL(20 kDa)-g[3.5]-PEG(2 kDa)) in 0.1 M HEPES buffer for 1 h. Subsequently, unboundPLL-g-PEG was removed by rigorously rinsing withDI water. We found that this protocol minimized the nonspecific interactionof NIH/3T3 fibroblasts to substrate (Figure S4, Supporting Information).

The final step of substratefabrication involves modifying AuNPswith the molecular tension ligand. To maintain the collapsed mushroomconformation of the tension ligand, it was necessary to functionalizethe AuNP with low densities of the fluorescent probe. It was alsoimportant to block the remaining AuNP surface, thus minimizing potentialnonspecific protein interactions. Accordingly, the AuNP was modifiedwith a binary mixture of the tension ligand and the short thiolatedPEG, SH(CH₂)₂(OCH₂CH₂)₈COOH. The synthesis of the SH-PEG₈₀-c(RGDfK)-Cy3Bmolecular tension ligand was adapted from our previous work (FiguresS5 and S6, Supporting Information). Briefly,a terminal cysteine residue that presents an amine and thiol was incorporatedin the c(RGDfK) peptide. The amine group was initially modified withan NHS-azide with high yield (>90%). In the following two steps,themaleimide-Cy3B dye and alkyne-terminated polyethylene glycol (Alkyne-PEG₈₀-SH, MW 3400) were further coupled to the thiol and azidegroups, respectively. After HPLC purification, thiolated MTFM ligandswere allowed to self-assemble onto the surface of the AuNP. By varyingthe concentration of tension ligands from 400 to 20 nM, while maintaininga constant thiol concentration of 40 μM using SH(CH₂)₂(OCH₂CH₂)₈COOH, wetuned the density of tension probe ligands per AuNP. By empiricallytesting the cell adhesion onto these different substrates, we foundthat the 200 nM ligand concentration was the minimum concentrationsufficient for allowing significant cell adhesion and spreading (FigureS4, Supporting Information). Given thatlower ligand densities are desirable for minimizing background signal,we selected this concentration for subsequent cell studies.

To quantify the number of molecular tension ligands per AuNP weperformed a quantitative fluorescence calibration and found that particlesincubated with a 200 nM concentration of tension probe (39.8 μMSH(CH₂)₂(OCH₂CH₂)₈COOH) had an average of 5.1 ± 0.5 probes per AuNP (FigureS7, Supporting Information). Because ofthe significant excess of the thiol ligand compared to the concentrationof AuNPs, this average number is valid for both 50 and 100 nm spacedsubstrates. When these particles are immobilized onto the glass coverslip,only part of the Au surface is available for sensor modification dueprimarily to the steric blocking of the surface bound PLL-g-PEG brush. The estimated thickness of the PLL-g-PEG layer in the hydrated state is approximately 4–6nm,³⁵ which is comparable to the size ofAuNP radius. Therefore, we assumed that at most only half of the AuNPsurface area was available for binding tension sensors, thus allowinga maximum average number of 2.5 probes per particle. This number stronglysuggests that each AuNPs was loaded with a low density of the tensionprobe, thus ensuring that the PEG was in the collapsed mushroom conformationat resting.³⁶

Figure 2

Analysis of FA proteinsand integrin tension colocalization. (a)Representative TIRFM-488 (GFP channel, green) and Cy3B epifluorescence(integrin-tension channel, red) images of NIH/3T3 fibroblast cellscultured on randomly arranged AuNP sensor substrates for 1–2h. The cells were transiently transfected to express GFP β₃-integrin, paxillin, zyxin, and LifeAct, and this signal wasfound to colocalize with the integrin tension signal. (b–e)Representative zoom-in images showing the distribution of GFP-taggedβ₃-integrin, paxillin, zyxin, and LifeAct with theintegrin–tension signal. The integrin–tension signalwas quantified and found to highly colocalize with FA markers (seeline scan analysis). Tension values were dynamic (see below) and variedacross subcellular regions reaching maxima that ranged from ∼10–20pN. Note that the reported tension values represent the average tensionper ligand, thus likely underestimating the forces applied by integrins.

Toward investigating the relationshipbetween force transmissionand FA formation, we next demonstrated the compatibility of MTFM withgenetically encoded tagging of FA markers. NIH/3T3 fibroblast cellswere transiently transfected with β₃-integrin, paxillin,zyxin, and LifeAct and then cultured onto substrates modified withrandomly arranged 9 nm diameter AuNP tension sensors for ∼1–2h and subsequently imaged using total-internal reflection fluorescencemicroscopy (TIRFM) (Figure 2a). The densityof disordered AuNP sensors on these substrates is ∼100 nm,which is sufficiently broad to allow FA maturation and force transmission.²² In all cases, we found strong integrin tensioncolocalization with the early markers of FAs such as β₃-integrin, and paxillin. In contrast, the zyxin and LifeAct signalswere distributed across the entire cell but preferentially localizedto the actin bundles, which is in agreement with previous reports.³⁷ The integrin tension signal was mainly detectedat the cell perimeter, coinciding with the greatest zyxin and LifeActintensities at the tips of the actin bundles. In some cases, the signalobserved at the center of the cell was due to autofluorescence fromthe nucleus. However, in other cases the signal was due to focal adhesiongenerated forces. The distinction between the two types of signalis clearer upon examination of the β₃-integrin-GFPchannel.

Upon analysis of subcellular regions (Figure 2b–e), we found that the maximum integrintension within eachFA typically appeared near the center of the rod-shaped elongatedstructure. Note that in these zoom-in images we observed that thepeak position of tension can be offset from the peak position of thefocal adhesion, either proximal or distal, by submicron distances,or in some cases it may perfectly overlap with the peak position ofthe focal adhesion marker, thus demonstrating the dynamic tensionfluctuations during FA formation.³⁸ Byquantifying the quenching efficiency of the tension ligands at restingand employing the ligand density of 2.5 per AuNP, we were also ableto estimate the average force per ligand, which ranged from 1 to 20pN (Figure S8, Supporting Information),which is consist with the observation of integrin force mediated biotin–streptavidindissociation.²⁶ Note that this value issignificantly greater than that reported by Dunn and colleagues (1–5pN) and may be due to the limited dynamic range of their spider-silkbased probes or the nature of the linear RGD peptide used in theirstudies.²⁷

Figure 3

(a) Representative images of GFP-paxillinexpressing NIH/3T3 cellsseeded onto the 50 and 100 nm spaced AuNP substrates. Images are shownfor the 0.5 and 3 h time-points, highlighting the differences in integrintension, cell shape, and FA size at these two time points. (b) Plotof GFP-paxillin cluster size (which is indicative of FA size) as afunction of time for n = 10 cells. The plot showsa steady increase in FA size over 5 h after cell seeding on the 50nm-spaced substrate, which is in contrast to the 100 nm spaced substrate,which shows limited FA maturation. (c) Plot showing the average tensionper integrin ligand across the entire cell for n =10 cells. Integrin tension increased rapidly within the first hourand was then maintained for cells cultured on AuNP arrays spaced at50 nm. This is in contrast to cells cultured on the AuNP arrays spacedat 100 nm where tension decreased by the later time points. (d) Representativeimages showing the change in integrin tension before and after treatingthe same cell with Y-27632 (40 μM) and cytochalasin D (10 μM).Corresponding timelapse movies are included in Movies S2 and S3 (see Supporting Information). (e) Stepwise blebbistatinand cytochalasin D treatment of cells (n = 4) ledto significant reduction of mean ligand tension. (f) Blebbistatintreatment of cells (n = 4 cells) led to over 80%reduction in the total cell traction force and FA area.

We next investigated the relationship between integrinclusteringand tension by culturing GFP-paxillin transfected cells onto MTFM-patternedsubstrates with 50 and 100 nm spacing. The cells were continuouslymonitored using TIRFM (GFP-paxillin) and epi-fluorescence microscopy(integrin tension) for over 5 h. Representative cell images are shownin Figure 3a, and the data indicated that thefootprint of the cells cultured on the 100 nm spacing remained small(300–1000 μm²) in contrast to cells grownon the substrate with 50 nm spacing (2000–5000 μm²). The difference in cell spreading was observed at the earliesttime points (∼30 min) and became more pronounced at all latertime points (Figure 3a). Although cells startspreading almost immediately upon plating on the 50 nm spaced AuNParrays, only a few cells spread onto the 100 nm spaced substrate beforethe 30 min time point, which is in agreement with literature.²⁰ Surprisingly, both the average FA size (as measuredby GFP-paxillin, Figure 3b) and the averagetension per ligand (Figure 3c) were similarfor the cells cultured on both substrate spacings at the early timepoint of 30 min (each data point represents n = 10cells). At later time points, FA area and the average tension perligand diverged; cells on the 50 nm spacing formed significantly largerFAs with greater values of tension (Figure 3b,c, Figure S9, Supporting Information). It is notable that for the substrate of 50 nm spacing, the averagesize of FAs continues to grow over the time course of the experiment(from nascent focal adhesion to mature and elongated focal adhesion).However, the average tension per integrin ligand only rises to a thresholdlevel that is maintained across the 5 h experiment. To verify thisobservation, we added 30 μM oleoyl-l-α-lysophosphatidicacid (LPA), a stimulant of myosin-contractility, to cells that havebeen cultured for 5.5 h in serum supplemented media (Figure S10, Supporting Information). Statistical analysisshowed no significant increase in mean tension per integrin or totaltraction per cells. In contrast, addition of LPA to serum-starvedfibroblasts led to a significant increase in tension and FA area,recovering to levels similar to that of cells cultured in serum-supplementedconditions. This data clearly indicates that although the total tensiongenerated by the cell is growing, the tension per integrin ligandis maintained at a constant value; thus, individual integrin tensiondoes not increase continuously during FA maturation and cell spreading.Our finding has two implications. First, it suggests that the mechanismof how cells continuously increase the exerted traction force is throughincreasing the number of surface-engaged integrins (and cell area)rather than mounting greater force per ligand.³⁹ Second, this observed level of force maybe related to theuniversal peak tension that was recently reported by Wang and Ha.⁴⁰

The results also suggest that at earlytime points (∼30min), the mechanism of integrin force generation is independent ofintegrin clustering, and increasing the average tension beyond 2−3pN per ligand requires a greater ligand density (< ∼ 60nm spacing). On the basis of previous literature, the early time pointforces are likely generated by actin polymerization, rather than myosincontraction.⁴¹ To distinguish the contributionsof actin polymerization and myosin contractility to integrin tension,we imaged NIH 3T3 fibroblasts before and after treatment with theRho kinase inhibitor Y-27632, and cytochalasin D (Figure 3d, Figure S11, Movies S2 and S3, Supporting Information). Analysis from n =5 cells treated with Y-27632 for 30 min shows that integrin tensionsignal generated by mature FAs decreased significantly to values of∼2–3 pN per ligand and were exclusively localized toa submicron structure at the cell edge. In contrast, addition of 10μM cytochalasin D rapidly (∼5 min) abolished all integrintension signal to background levels, likely due to the disruptionof actin polymerization. To further confirm that actin polymerizationand myosin contractility are the two main contributors to integrintension, we performed a stepwise inhibition of both processes in thesame cells. We first performed timelapse imaging on fibroblasts (n = 4 cells) that were treated with a myosin II inhibitor(25 μM blebbistatin). In this experiment, the average ligandtension was reduced from ∼6 to ∼3 pN and reached a steadystate value within 5 min of adding the drug, which is similar to theeffect of Y-27632 and suggests the loss of myosin-driven tension atthis time point. Interestingly, this level of tension coincides withthe magnitude of ligand tension during initial cell spreading (Figure 3c, t = 30 min). When these cellswere further treated with 10 μM cytochalasin D, the mean integrinligand tension was immediately reduced (within 1 min) to approximately2 pN and was gradually reduced to ∼1 pN within 15 min (Figure 3e). This ∼2 pN decrease in tension is likelydue to the loss of actin-driven forces and associated membrane tension.Exclusively treating cells with blebbistatin led to an 80% decreasein FA size as well as >80% decrease in the total tension per cell(Figure 3f), which is in agreement with literatureprecedent.⁴² When comparing the total decreasein cell tension with the loss of tension per ligand, it is clear thatmyosin-inhibition leads to a decrease in the number of engaged integrinsand not only a decrease in integrin tension. Taken together, thisdata suggests that during initial FA formation, actin polymerizationdrives integrin tension to an average of 1−3 pN per ligand.This is closely followed by actomyosin-contractility that increasesthe average tension to ∼6–8 pN and is associated withFA maturation. Note that this value of mean ligand tension was alsoobserved in two additional cell lines, rat embryonic fibroblasts (REFs)and human bone osteosarcoma epithelial cells (U2OS) (Figure S12, Supporting Information).

To better understandthe relationship between FA maturation andforce transmission for the high ligand density substrate, we capturedtimelapse movies of integrin tension with F-actin dynamics (Figure 4a-b). At initial time points, we observed diffuseintegrin tension over the lamellipodium, which was similar to theintensity of tension in cells grown on the 100 nm spaced substrate.At subsequent time points, we observed high integrin tension punctathat localized to the tips of f-actin bundles (Figure 4b (white arrow), and Movie S1, SupportingInformation), which is the location of the linkage betweenthe FA and cytoskeleton.⁴³ We tracked asingle integrin tension puncta (white arrow, Figure 4b) and observed that within the time frame (from t = 25 to 41 min) the maximum integrin ligand tension within the singleFA increased 1 order of magnitude from ∼3 pN to ∼12pN. For this single FA, we found that integrin tension increased concomitantlywith FA growth (Figure S13, Supporting Information). Interestingly, micron-scale actin fiber assembly coincides withthe increase of integrin tension (Figure S14, Supporting Information), in agreement with literature suggestingthe importance of stress fibers as a template for tension mountingand FA maturation.⁴⁴

Figure 4

Integrin tension andactin dynamics during early FA maturation.(a) Representative brightfield, reflection-interference contrast microscopy(RICM), LifeAct GFP (TIRFM 488, red), integrin tension (epifluorescenceCy3B, green), and overlay of GFP and tension signals for a singleNIH/3T3 fibroblast cells adhered on sensor substrate immediately followingcell seeding. The full timelapse movie from t = 5min to t = 43 min after cell seeding is availableas Movie S1 in Supporting Information.(b) Zoom-in timelapse overlay images of integrin tension for t = 19 to 41 min.