Tension Sensing Nanoparticles for Mechano-imaging at the Living Non-living Interface

Abstract

One of the most significant challenges pertaining to understanding the interplay between mechanical forces and chemical reactions involves elucidating the magnitude of force experienced by specific molecules as a function of time and space.^2a-c To address this need, several pioneering groups in the area of mechanochemistry have developed force sensitive chromophores, or mechanophores, that respond to mechanical tension by undergoing covalent bond rearrangements that shift absorbance or fluorescence emission.³ Nonetheless, given the relatively large changes in free energy required to break covalent bonds, current mechanophore probes are sensitive to forces in the range of hundreds to thousands of pN (~10-100 kcal/mol, assuming a 10 Å displacement).⁴ Thus, current mechanophores are unable to probe forces in the range of 1-50 pN that can drive conformational changes in macromolecules and molecular assemblies.

Tension-driven conformational rearrangements underpin many of the fundamental processes that regulate living systems. For example, cell division,⁵ translation,⁶ and transcription⁷ require spatially and temporally coordinated low pN range forces to proceed. Accordingly, our group recently developed a method termed Molecular Tension-based Fluorescence Microscopy (MTFM), to measure pN forces exerted by cell surface receptors.^2b MTFM employs a ligand molecule linked to a polymeric “spring” and anchored to a surface. The linker is flanked by a pair of dyes utilizing fluorescence resonance energy transfer (FRET) to report on molecular forces that extend the polymer from its resting position. MTFM offers the only method to visualize pN forces exerted between membrane receptors and their extracelluar ligands.^2b We rationalized that, by developing a gold nanoparticle-based mechanophore, the force sensitivity of MTFM could be extended to measure receptor tension magnitudes that are not practically accessible by FRET-based approaches, such as the genetically encoded spider silk construct,^2a and our own FRET-based polyethylene glycol (PEG) tension sensors.^2b

Noble metal nanoparticles have revolutionized the field of chemical sensing due to their unique optical, electrical, electrochemical, and catalytic properties.⁸ Moreover, the relatively biocompatible nature of gold nanoparticles (AuNP) has lent itself to biological sensing applications for both in vivo and in vitro assays.⁹ In many of these applications, the superior fluorescence quenching ability of AuNPs is exploited to achieve high sensitivity turn-on detection.¹⁰ Compared to molecular quenchers, the effective quenching distance of AuNP can be as long as several tens of nanometers.¹¹ Theoretical and experimental studies have shown that the distance-dependent quenching of 1-20 nm AuNPs follows a 1/r ^{relationship, termed Nanometal Surface Energy Transfer (NSET),12} which provides a highly sensitive approach to measuring molecular distances in living systems.¹³

Herein, we report on an AuNP-based sensor for MTFM to visualize the pN-range force dynamics of integrin receptors during cell adhesion (Scheme 1). As a proof-of-concept, we target the α_Vβ₃ integrins using high affinity peptides, because integrins are the primary molecules to sustain large tensile loads supporting cell adhesion and cell migration.¹⁴ The AuNP MTFM sensor utilizes a calibrated NSET response to determine the molecular extension of an entropic polymer “spring”¹⁵ anchored to the AuNP scaffold. This distance information is then used to infer the corresponding molecular tension. Thus, this probe provides the first reversible nanoparticle mechanosensor for imaging integrin molecular tension.

Scheme 1

AuNP-based molecular tension fluorescence microscopy (AuNP-MTFM)

Scheme 1 describes the AuNP-MTFM approach. To synthesize the ligand (Figure S1), cyclic Arg-Gly-Asp-dPhe-Lys-(Cys) peptide (cRGDfK(C)) was first modified with an NHS-azide in high yield (>90%). This afforded the orthogonal reactive thiol and azide groups for further modification. Maleimide-Alexa488 dye and alkyne-terminated polyethylene glycol (PEG, MW 3400) were further coupled to the thiol and azide, respectively. The purity of the final product was confirmed by HPLC and MALDI-TOF MS (Figure S2). The mass distribution of the PEG polymer indicated a contour length of 28.0 ± 2.5 nm, assuming a monomer length value of 0.35 nm.¹⁶ Based on the Worm-like Chain (WLC) model for force-induced molecular extension,¹⁷ the fluorophore should rest at 4.9 ± 0.2 nm from the AuNP surface at low packing density (< 29 molecules/AuNP), and can be extended up to ~28 nm away from the gold surface. Thus, the sensor should generate a 10-fold increase in fluorescence signal, if subjected to sufficient tension.

To quantify the distance between the Alexa488 dye and AuNP surface, we measured the d₀ (distance at which the quenching efficiency = 0.5) by using highly packed duplex DNA as a molecular ruler. Briefly, a series of DNA duplexes conjugated with 5’Alexa488 at the terminus of one strand and a 5’ thiol at the terminus of the complementary strand were immobilized to the AuNP surface while the salt concentration was titrated to 0.3 M NaCl (Figure 1a-b). To minimize self-quenching, the dye labeled DNA duplex was diluted in a 1:9 ratio using identical duplexes that lacked the 5’ fluorophore. We used the OliGreen assay and the calibrated fluorescence intensity to provide two independent methods to determine the DNA loading number per particle, which is reported in Figure 1b and Table S1.¹⁸ These values suggest that the duplexes present orientations perpendicular to the gold surface.¹⁹ We then measured the fluorescence intensity of the four AuNP-dsDNA-Alexa488 conjugates before and after dissolution of the AuNP. The quenching efficiency of triplicate AuNP samples was plotted as a function of distance and fitted to the NSET equation (Figure 1c), providing a d₀ of 13.6 ± 0.3 nm. This value is in agreement with literature precedents that reported d₀ of 10.4 nm and 15.7 nm for 10 nm AuNPs displaying Atto647, and Cy3, respectively.¹¹,20 Importantly, we can combine this calibrated NSET equation and the WLC relation, which has been experimentally and theoretically validated for PEG, to generate a theoretical curve depicting quenching efficiency as a function of applied tension (Figure 1d). Based on this plot, the quenching efficiency decreases from 0.9 to 0.1 when a force of 25 pN is applied, affording a wider dynamic range than our reported FRET-based PEG tension sensor,^2b and a 5- to 10-fold improvement in signal-to-background when compared to the genetically encoded spider-silk tension sensors construct.^{2a, 2c}

Figure 1

NSET calibration. (a) To determine d₀, particles were functionalized with a binary mixture of 1:9 labeled:unlabeled dsDNA. (b) Table showing the DNA sequences used, their measured density, and predicted lengths. (c) Plot showing the quenching efficiency as a function of distance from the AuNP surface. The data was fitted to the NSET model and d₀ was determined to be 13.6 ± 0.3 nm. (d) A theoretical force-quenching efficiency plot based on combining the WLC and NSET models. The force dynamic range corresponds to quenching efficiency ranging from 0.9 to 0.1.

Given that the ligand density dictates its conformation on the AuNP surface,²¹ and consequently influences sensor response, we quantified the density of the ligand as function of its incubation concentration. We found that the number of surface bound ligands per AuNP ranges from 700-100 when the ligand concentration was varied from 80 μM to 1 μM (Figure 2a). To passivate the bare gold surface and increase particle stability, we introduced other competing thiolated ethylene glycol molecules to form binary PEG coated AuNPs. Among these, we found that COOH-EG₈-SH to be advantageous due to increased particle stability. The dual PEG modification strategy offered significant improvement in controlling the number of ligand molecules per particle, especially at low ligand densities, eg. 1-10 ligands per NP (Figure 2b). This ensured that the polymers adopt their predicted Flory radius of 4.9 nm, thus maintaining the relaxed mushroom conformation and generating a more reproducible response.¹⁶

Figure 2

PEG conformation on AuNP surface. Plot of A488-PEG₈₂-SH loading as a function of incubation with single ligand (a), and binary mixture of ligands of A488-PEG₈₂-SH and COOH-EG₈-SH (b). The total ligand concentration was maintained at 40 μM, and the ratio was varied from 0 to 0.99.

We next verified the conformation of the polymer for particles displaying 8.2 ± 0.3 Alexa488-PEG₈₂-SH per NP. Using the measured d₀ of 13.6 nm, the quenching efficiency data (Figure S3) indicated that the dye was 8.4 nm from the NP surface. This distance is in general agreement with the predicted dye-AuNP distance of 7.4 nm, which is the sum of the expected contour length of COOH-EG₈-SH (2.8 nm),²² and the Flory radius of the partially crowded Alexa488-PEG₈₂-SH (4.6 nm). We further verified the PEG conformation using negative staining TEM of samples with different surface loading densities (Figure S4).

Typically, AuNPs are immobilized to a substrate using primary amines or thiols, which provide sufficient binding for most chemical sensing applications. However, to measure molecular forces, we needed to irreversibly immobilize AuNP conjugates, such that particles would not translocate in response to external tension (Figure S5). Therefore, we used lipoic acid ligands to provide strong affinity to the gold surface due to multivalent binding.²³ Moreover, this method allowed immobilization of AuNP-based MTFM sensors within a few minutes, thus minimizing the aggregation due to extended incubation times. To generate these surfaces, aminefunctionalized substrates were reacted with a mixture of 5% w/v mPEG-NHS (MW 2000) and 0.5% w/v lipoic acid-PEGNHS (MW 3400) in 0.1 M sodium bicarbonate buffer, overnight. AFM was then used to determine the surface density of anchored AuNPs (Figure 3a), and showed that the average inter-particle distance was ~110 nm. Although particle spacing was disordered across the substrate, this type of packing is known to allow cell focal adhesion formation.²⁴

Figure 3

Mechano-imaging of integrins in live cell. (a) AFM image showing the typical AuNP sensor distribution. After 1 hr of cell culture on the sensor surface, the brightfield (b) and RICM image (c) were captured to visualize cell adhesion. RICM (c) clearly revealed the adhesion sites (dark region) mostly at the edge of the cell. (d) TIRF image (500 ms acquisition time) of the same cell shown in (b-c). The raw fluorescence data was converted to a quenching efficiency map using bulk quenching efficiency of AuNP sensor. A force map was then calculated using the WLC model. Inset shows rod-like focal adhesions with a high level of force. Scale bar, 20 μm. (g) Representative images showing that paxillin colocalizes with tension signals within focal adhesions. Scale bar, 15 μm.

To image the tension exerted by living cells, human breast cancer cells (HCC 1143) were plated onto the AuNP sensor modified glass surfaces for 1 hour to allow the cells to form adhesions. To minimize auto-fluorescence and maximize the signal to background ratio, we performed total internal reflection fluorescence microscopy (TIRFM) using a laser coupled inverted microscope. At the resting state, we observed relatively low and uniform background fluorescence intensity (Figure S5c). After cells adhered and polarized on the surface, we used brightfield (Figure 3b) and reflection interference contrast microscopy (RICM, Figure 3c) to observe the binding between the cell and the surface. In the AuNP-MTFM channel (TIRF 488), we observed a strong fluorescence intensity increase (up to 7 fold over background) (Fig. 3d) that was strongly associated with the cell-binding pattern in RICM (Figure 3c). The signal to background ratio at the brightest spots of the image was ~20, which allowed direct and facile identification of areas of high tension. In some local regions, rod-like contact patterns formed (Figure 3d-f, red arrow, inset) suggesting the formation of mature focal adhesions.²⁵ Using the bulk quenching efficiency values, and the NSET model, we first determined the average PEG extension at each pixel of the image (Figure S7). Next, the WLC model was used to estimate the minimum ensemble tension exerted at each sensor ligand, thus providing a tension map (Figure 3f). While a number of techniques such as traction force microscopy and PDMS post arrays have been used to measure cell forces, our method detects pN tension, thus representing a 3-order of magnitude sensitivity improvement, and is exclusive to RGD-binding receptor molecules. The temporal and spatial resolution of MTFM are not different from standard fluorescence microscopy techniques, which exceed that of traction force microscopy, and especially if combined with super-resolution imaging or single molecule imaging approaches.²⁶

To validate that the tension signal is due to specific focal adhesion complexes, cells were cultured onto passivated PEG-only surfaces or non-sensing AuNP conjugate surfaces. Minimal cell attachment was observed under our experimental conditions, thus showing specificity of RGD binding (Figure S7). In another experiment, we treated the cells on AuNPMTFM sensor surfaces with 25 μM latrunculin B (LatB), which inhibits actin polymerization, and thus prevents force generation (Figure S8). After 10 min of treatment with LatB, the increased fluorescence completely dissipated and returned to background levels for all observed cells (Figure S8). In addition, we immunestained for paxillin, a marker of focal adhesions (see Methods in SI). Figure 3g shows a high degree of overlap between the AuNP-MTFM response and the paxillin-staining. Note that the tension signal significantly weakens once cells are fixed, likely due to nm scale cytoskeletal and PEG relaxation. Taken together, these experiments clearly show that the reversible fluorescence response is due to mechanical tension exerted by integrins the MTFM sensor.

To demonstrate dynamic imaging capabilities, we collected time-lapse images of a cell as it contacted a substrate and initiated focal adhesion formation (Figure 4, SI Movie 1). The tension signal rapidly translocated from the center of the cell to its periphery over the span of 10 min. This is the first direct evidence showing the presence of integrin tension during initial cell substrate contact, which was previously inferred by ligand translocation.²⁷ In another time-lapse video, we captured the tension dynamics of the reverse event, as a cell underwent contraction and the high-tension regions at the cell periphery dissipated (Figure S9, SI Movie 2). The alignments of tension signals are clearly perpendicular to the cell edge, corresponding to its distal orientation.²⁵ Finally, histogram analysis of force distribution within focal adhesions showed a mean tension of ~1.0 pN, which is in agreement with estimates from traction force microscopy and PDMS post arrays (Figure S10). Interestingly, some regions displayed tension values that are an order of magnitude greater than the average, which is compatible with integrin-RGD rupture forces measured using atomic force microscopy.²⁸

Figure 4

Initial tension dynamics during cell adhesion. Time-lapse images recording the first 10 min of cell-substrate interactions after the cell engaged the surface. Tension signal (TIRF 488 channel, 100 ms acquisition time) colocalized with cell binding locations as shown in the RICM channel. Scale bar, 15 μm.

In summary, we report the first nanoparticle-based molecular tension probe and use it to image the dynamics of integrin adhesions. We calibrated the NSET response and verified the sensor conformation on the AuNP surface. In principle, this sensor reports on molecular tension with a wider dynamic range than is fundamentally allowed by FRET-based methods. Note that the stability of the thiol gold bond exceeds that of integrin-RGD interaction, and thus the MTFM sensor is stable within our experimental conditions.²⁹ Surprisingly, we observed a wide distribution of tension magnitudes within adhesions. Given that nanoparticles can be spatially patterned onto surfaces, this suggests that one may be able to readily investigate the role of ligand density and topology on force transmission. AuNPs can be also tracked using label-free methods, which should allow for dual force-localization studies in 3D.

Footnotes

REFERENCES