Super-resolution SERS Imaging of Single Particles in Cells

Abstract

The ability to locate and identify molecular interactions in cells has significant importance for understanding protein function and molecular biology. Functionalized metallic nanoparticles have been used as probes for protein tracking and drug delivery due to their ability to carry therapeutic agents and readily functionalized surfaces. In this work, we present a super-resolution surface-enhanced Raman scattering (SERS) approach for imaging and tracking membrane receptors interacting with peptide functionalized gold nanostars (AuNS). The α_vβ₃ integrin receptors in colon cancer cells are successfully targeted and imaged using AuNS with the high affinity amino acid sequence arginine-glycine-aspartic acid-phenylalanine-cysteine (RGDFC) attached. The RGDFC peptide interaction with the integrin receptor provides a bright and fluctuating SERS signal that can be analyzed with localization microscopy algorithms. Additionally, the observed SERS spectrum is used to confirm protein-peptide interaction. Experiments with functionalized and bare AuNS illustrate specific and non-specific binding events. Specific binding is monitored with a localization precision of ~6 nm. The observed spatial resolution is associated with tight binding, which was confirmed by the slower diffusion coefficient measured from 4.4 × 10⁻¹¹ cm²/s for the AuNS-RGDFC compared to 7.8 × 10⁻¹⁰ cm²/s for the bare AuNS. Super-resolution SERS images at different focal planes show evidence of internalized particles and suggest insights into protein orientation on the surface of cells. Our work demonstrates super-resolution SERS imaging to probe membrane receptor interactions in cells, providing chemical information and spatial resolution with potential for diverse applications in life science and biomedicine.

Graphical Abstract

Protein receptors^1,2 located in the membrane barrier between cells and their environment facilitate communication between extracellular medium and intracellular components. When external molecules or ligands bind to one of these receptors, it can trigger changes within the cells. Activating and blocking protein receptors are common strategies pursued in drug development^1-4. While there are significant strides being made in biological engineering and pharmacotherapy toward discovering new drug targets and appropriate ligands for drug delivery and disease treatments, the development of techniques to track and locate these (bio)molecules, essential to understanding drugs interaction and delivery, remains very challenging⁵.

A suite of advanced techniques have emerged to provide high resolution biological images such as cryo-electron microscopy (cryo-EM)⁶ and X-ray crystallography.⁷ Additionally fluorescence-based methods like resonance energy transfer (FRET)⁸ and DNA points accumulation for imaging in nanoscale topography (DNA-PAINT)⁹ have been demonstrated with optical techniques. These approaches provide valuable information about biochemical interactions and enable scientists to track and locate (bio)molecules with sub-diffraction limited resolution. While these techniques have had success, they rely on time-consuming sample preparation steps, require significant technical expertise, and/or often rely on labels that have autofluorescence and photobleaching issues. Surface-enhanced Raman scattering (SERS) is another approach that has been used to image biomolecules in cells^10,11. SERS labels and probes have been extensively used as an important bioimaging strategy due to their stable signals over time, reduced sample background (autofluorescence) and, narrow spectral bands that facilitate multiplexing compared to common fluorophores^10,12. However, SERS is high-sensitivity technique that can also be used as a label-free method to provide a molecular fingerprint and also obtain spatial information^13-18. In addition to providing chemical information about the cellular components, SERS is minimally invasive and has the versatility to work in adverse environments, e.g. in vitro and in vivo^19-24, which have led to the emergence of SERS as a widely applied and important analytical tool for biomedical applications²⁵.

The α_νβ₃ integrin is a cell surface glycoprotein receptor that is known to bind with high specificity to short amino acid sequences, such as Arginine-glycine-aspartic acid (RGD) containing peptides. α_νβ₃ integrin is widely targeted in cancer treatment because if its increased expression in cancer cells^1-4. Recently, our group reported that the SERS response from a gold nanostar (AuNS) functionalized with the peptide RGD-phenylalanine-cysteine (RGDFC) interacting with the α_νβ₃ receptor could be monitored by the distinct Raman signal observed and that the peptide on the AuNS inhibited the irreversible adsorption of proteins and biomolecules (corona formation), which is believed to be a key to successful in vitro and in vivo targeting^26,27. The Raman signal enhancements in SERS arise from the interaction of a few molecules (or a single-molecule) located in the highly localized electromagnetic field at the nanoparticle surface that arises from the excitation of localized surface plasmon resonances (LSPR).^13,28,29 The molecule enhancements are known to arise from localized areas of high intensity, commonly called “hotspots” on the plasmonic nanostructures^30,31. Previously, our group has shown that the signal obtained from tip-enhanced Raman spectroscopy (TERS)^32,33 and SERS^26,27 of the AuNS-RGDFC particle can be used to differentiate the binding of the RGDFC peptide to different receptors in cell membranes. In our prior TERS work, 25-50 nm spatial resolution was achieved over a 1-2 μm square region. In contrast, SERS can map an entire cell, approximately a 10 × 10 μm square area; however, the spatial resolution was limited by the 1 μm step size and the diffraction limit for the focal spot (~400 nm), which impedes determining the precise location or internalization of the nanoparticle within different compartments of the cell. Nonetheless, the spectrum observed in both the SERS and TERS experiments generated the same chemical signatures associated with the binding interaction. The intensity of the SERS signals obtained from the AuNS-RGDFC interaction is substantially bright to suggest it can be detected and analyzed, for example, using localization microscopy techniques for improved spatial resolution wide-field imaging³⁴.

Super-resolution SERS imaging can provides very fast acquisitions, on the order of microseconds (μs) to milliseconds (ms), and with nanometer spatial resolution even using wide-field illumination from a narrow linewidth laser source^34-37. Similar to photo-activated localization microscopy (PALM)^38,39 or stochastic optical reconstruction microscopy (STORM)⁴⁰, in super-resolution SERS-STORM^35-37,41,42, sequential images (frames) are recorded using a 2-dimensional (2D) array, such as a charge-coupled device (CCD) or complementary metal-oxide–semiconductor (CMOS) camera, where the spatial resolution is the convolution of the point spread function (p.s.f.) and the pixel size of this camera. However, rather than the “on-off” blinking and switching behavior of single fluorophores^40,43 in fluorescence microscopy, the natural time-dependence of the SERS signal in single-molecule regime^13,14,16-18, SERS intensity fluctuations (SIFs)^42,44 can provide a temporally fluctuating signal appropriate for localization algorithms. The origin of those SIFs is not fully understood, but has been associated with the “on-off” diffusion of molecules visiting the hotspots in solution³¹ or surface reconstruction of atomic features that confine light into so-called “picocavities” in air^18,42. The reconstruction of these picocavities is believed to occur on sub-millisecond time scales at room temperature^18,42,45. Therefore, the signal variance from frame to frame contains information about the molecule probing the hotspot solved in space. The p.s.f of the SIFs can be fit by a traditional 2D Gaussian function and, therefore, the position (or centroid) of the emission can be determined within the diffracted-limited spot to a precision on the order of nanometers^18,34,46. This means information from the near field is accessed in the far-field, in our case, from a wide-field illumination (λ = 659 nm). Thus far, there are only a few applications of super-resolution SERS imaging for biological samples: macrophage cells⁴⁷, bacteria⁴¹ and DNA fragments^36,37. Most of these reports were performed on structured surfaces, and we are unaware of any reports taking advantage of nanoparticle-based SERS as labels or probes to achieve super-resolution images.

In this paper, we demonstrate in vitro super-resolution SERS imaging and tracking of nanoparticle-membrane receptor interactions in intact SW620 colon cancer cells. The functionalized nanoparticles interact via modification of the surface with an RGDFC peptide that binds specifically to the α_vβ₃ integrin receptor. The SERS spectrum obtained further confirms the interaction of RGDFC-AuNS bound to α_vβ₃ integrin receptor. Using the temporal fluctuations of the Raman signal in the image, a localization precision of ≤6 nm is achieved with the RGDFC-AuNS. The diffusion coefficient determined experimentally suggests the AuNS-RGDFC are strongly bound. Our results suggest that SIFs are coming from the interaction and conformation of the AuNS-RGDFC-α_vβ₃ receptor at the membrane surface. By taking images at different focal planes, a Z-stack provides a 3D perspective of AuNS position and suggests internalization of the nanoparticles within the cells.

EXPERIMENTAL SECTION

Chemicals.

Tetrachloroauric (III) acid (HAuCl₄), cetyltrimethylammonium chloride (CTAC), sodium borohydride (NaBH₄), silver nitrate (AgNO₃), hydrochloric acid (HCl), and ascorbic acid (C₆H₈O₆) were all purchased from Sigma Aldrich (USA). Milli-Q water was used. Cyclic(-Arg-Gly-Asp-D-Phe-Cys) was purchased from Peptide international. PBS, FBS, and RPMI-1640 medium were purchased from Thermo Fischer Scientific (USA).

Nanofabrication of silver nanoislands (AgI).

Briefly, the coverslips (1 inch²) were previous cleaned as previously described⁴⁸ and then placed in the thermal evaporator for deposition of the silver layer. A thin film of 15 nm of silver nanoislands (AgI) was deposited over the cleaned coverslips.

Synthesis and Characterization of AuNS.

The AuNS synthesis procedure described by Sloan-Dennison²⁷ was utilized in this study. Rather than using HPLC water, Milli-Q water (18.2 MΩ cm) was used in our synthesis. AuNS were characterized via UV-vis, DLS, Nanosight, and zeta potential measurements.

AuNS functionalization with RGDFC.

Cyclic RGDFC (200 μM) was added to AuNS with an LSPR of 657 (1 mL) and left to shake overnight. The sample was centrifuged to remove excess RGDFC. RGDFC-AuNS were resuspended in 100 μL of water and kept in the fridge until used.

Cell culture.

Human SW620 colon cancer cells (ATCC, Manassas, Va.) were passaged at approximately 80% confluency in RPMI-1640 medium supplemented with 10% FBS. The cells were cultured in a humidified atmosphere containing 5% CO₂ with a temperature of 37°C.

Cell incubation with AuNS-RGDFC.

Cells and RPMI-1640 medium were transferred to a 35 mm diameter culture flask (2 mL) with AgI coverslip. The culture flask was left to incubate for two days, allowing the cells time to adhere to the AgI coverslip. 200 μL of bare AuNS-RGDFC was added to the culture flask and was left to incubate for two hours. The cell media was then removed and the adherent cells on the AgI coverslip were washed with sterile PBS (1 mL) three times. Next, they were fixed by immersing the AgI coverslip in 4% formaldehyde for 15 minutes. The AgI coverslip was then removed from formaldehyde, washed with PBS and H₂O, and left to dry.

Super resolution SERS Imaging.

The sample was imaged using our homebuilt wide-field Raman microscope, which combines an upright Olympus BX microscope with a CMOS camera (ORCA-Flash 4.0 V3, Hamamatsu) for high speed image acquisition. In a typical experiment imaging a fixed cell in air, the excitation laser consisted of a 659 nm single longitudinal mode laser with a power density ~0.2 kW/cm². The laser provided wide-field illumination onto the sample, the Raman scattering was collected using a 100 × objective (NA = 0.9) in air, filtered using an edge filter (660 nm), and images were recorded at 100 ms per frame. In addition to recording wide-field imaging, our microscope can divert the collected Raman scattering to an f/2 spectrometer equipped with a CCD detector (Andor) to record the Raman spectrum. The spectral resolution was estimated as ~15 cm⁻¹, limited by the diameter of the optical fiber used to transmit the collected Raman signal into the spectrograph.

RESULTS AND DISCUSSION

Super-Resolution SERS Experiment.

Figure 1 illustrates the approach used in this work to perform super-resolution SERS imaging of functionalized nanostars interacting with integrin receptors. Gold nanostars (AuNs) covalently modified with the RGDFC peptide (AuNS-RGDFC) through Au-S bonding with the cysteine residue were used as SERS probes (Figure 1A). After peptide functionalization on the AuNS (See supporting information for more experimental details), AuNS-RGDFC were then incubated with SW620 colon cancer cells, Figure 1A. The SW620 cells were incubated with AuNS-RGDFC in RPMI-1640 medium® over a semi- reflective/transparent silver nanoisland film (AgI, thickness of 15 nm) (Figure 1B). The AuNS were added to the media, and after 2 hours, allowing for the AuNS to interact with cell receptors, the cells were fixed with glutaldehyde for imaging. The thin AgI film substrate provided increased contrast compared to a plain coverslip (Figure S2). Kneipp et al.⁴⁹ have demonstrated that the SERS signal from cells on a Au nanoisland film is quite intense, which helps differentiate between signals from the bottom of the cell and nanoparticles interacting at the top or within the cells.

Figure 1. Super-resolution SERS imaging of AuNS-RGDFC in cancer cells.

(A) RGDFC-functionalized AuNS were used as label free SERS probes. AuNS had an average size of 110±2nm and the LSPR peak max was at 657 nm (See supporting information, Figure S1 and Table S1). RGDFC functionalized AuNS were incubated with SW620 colon cancer cells in RPMI media to enable the AuNS-RGDFC to bind with α_υβ₃ integrin receptor as previous demonstrated. (B) The cells were fixed on Ag island covered coverslips. (C) Wide-field SERS images were collected using a 659 nm laser, with a power density of ~0.2 kW/cm², and a 100× objective (NA 0.9) and 100 ms per frame. The inset image shows a bright-field image of a cell observed in this experiment. (D) A series of images recorded by collecting the Stokes scattering use an edge filter to reject the Rayleigh scattered light. (E) Super-resolution algorithms are applied to the wide field SERS images (separated in frames) to increase the localization of the AuNS probes interacting with the cell. The data shown was processed using ThunderSTORM⁵⁰, with the following parameters adjusted for our camera: 65 nm per pixel, 1.6 photoelectrons per count, and camera base level of 50. The inset shows the improvement in the resolution when compared with the images in (D). The inset scale bar is 100 nm. (F) A typical Raman spectrum collected in this experiment is shown. The spectrum is obtained by directing the collected light through an optical fiber and then to a spectrograph and CCD. The integration time for the spectrum shown was 30 s.

The sample was imaged using our homebuilt wide-field Raman microscope for high speed image acquisition (Figure 1C). The laser was defocused to provide wide-field illumination onto the cells, and the Raman scattering was recorded on the CMOS array using an edge filter to collect only the SERS stokes signal (images and spectrum from 200 – 1800 cm⁻¹). Our system allows us to illuminate a large area (diameter of illuminated area ≈ 50 μm), which is larger than the size of a single cell of approximately 10 × 10 μm. The wide-field images were processed to achieve super-resolution SERS images (Figure 1E). In addition to recording wide-field images, the observed Raman spectrum was recorded (Figure 1F).

Super-resolution SERS images from colon cancer cells using AuNS-RGDFC.

Figure 2 shows results obtained from two SW620 cells incubated with AuNS-RGDFC. In the data, a persistent background is observed from the fixed cells. The background may arise from autofluorescence of other cellular components. In addition to this background, there are bright and intense fluctuations observed at discrete locations in different frames of the recorded images, as indicated in Figure 2B. The fluctuating signals are attributed to SIFs arising from the previously reported intense SERS associated with the interaction of RGDFC-AuNS binding to integrin receptors on the cell membrane.^26,27,32,33. The time dependent SIFs are used to differentiate the AuNS probes from the cellular background (Figure S3, see supplementary video 1). For example, one of these fluctuating bright spots is observed in frame #126 (Figure 2B, indicated by the green arrow), is not observed earlier in frame #01. The “on-off” SIFs is characteristic of single-molecule SERS experiments^17,18,42 observed in Figure 2C (the black arrows indicate the “on” intensities above a previously established user defined threshold of ~17). We can therefore use super-resolution SERS-STORM^{18,28,34,35,37,41} methodology to localize these fluctuating signals.

Figure 2. Super-resolution SERS results for fixed cells incubated with AuNS-RGDFC recorded at 10 Hz.

(A) Bright-field image of two SW620 colon cells. (B) Individual frames, #01 and #126, and the average wide-field SERS image (n = 460 frames). The green arrow in frame #126 highlights a particularly active cluster. (C) Analysis of the active cluster noted by the green arrow generates an integrated SERS intensity trajectory for the cluster (25x26 pixels). The red line in (C) shows the threshold used to eliminate the noise and low-level background fluctuations. The black arrows indicate the “on” SIFs. (D) The resulting averaged super-resolution SERS image for Figure 2B is shown. Analysis of the zoom image indicates the active cluster is composed of three or more particles on the cell (See Figure S4 in SI) The blur in the image is partially attributed to motion of the AuNS from frame to frame. (E) The SERS spectra are shown collected from fixed cells incubated with AuNS-RGDFC (yellow) and bare AuNS (blue), as well as AuNS-RGDFC on glass (red). The scale bars are 2 μm for (C) and (D), and 150 nm for the insets.

Figure 2D shows the resulting super-resolution SERS image obtained from 460 frames. Analysis of the fluctuating spot noted in Figure 2D suggests at least three scattering centers were found, which are not discernable in the diffraction limited image (See Figure S4). Given the spacing of the centers, greater than 100 nm, we believe these signals arise from different AuNS probes. While the entire cell is illuminated by the laser (Figure 2B, C), non-fluctuating signals do not get included in the super-resolved image (Figure 2D). Pixels with signal fluctuations below a set threshold are treated as noise, are not fit, and therefore not included in the super-resolved image. A video illustrating the ThunderSTORM fitting for each frame (#1 to 1000) is shown in Supplementary Video S2. The crosses in the video illustrate the centroids determined for each fluctuation, attributed to individual nanoparticles, in time.^{18,28,34,35,37,41}

SERS spectra were collected to verify the origin of the SIFs, Figure 2E. The SERS spectra were collected with laser power at the sample of ~4 mW and an integration time of 30 s. Our previous results indicate the α_vβ₃ receptor can be selectively detected and discriminated from the other cell membrane parts using the observed Raman signal^26,27,32,33. To validate the origin of the SIFs, SERS spectra were recorded from three samples: 1) bare AuNS incubated with cells (blue spectrum), 2) AuNS-RGDFC incubated in cells (yellow spectrum), and 3) AuNS-RGDFC on glass (red spectrum). The SERS spectra show distinct differences, the SERS spectrum from the AuNS-RGDFC and cells (yellow spectrum) show strong bands around 1559 and 1625 cm⁻¹, which are absent in the other spectra (blue and red spectra). Those bands are assigned to tryptophan and phenylalanine^{5, 6, 36, 37}, and are consistent with our previous reports of AuNS-RGDFC bound to α_vβ₃ receptor in cells. These results suggest that this super-resolution SERS imaging approach can be used to selectively monitor the AuNS-RGDFC with nanometer precision.

The chosen threshold in ThunderSTORM is an important parameter in the analysis that can improve the resolution and avoid a false positive identification^50,51; however in the presence of high background it can be difficult to avoid false positives and possibly miss real signals. The threshold used here was 17 (Figure 2C). This threshold is somewhat arbitrary and indicates that proper background correction is needed.

A comprehensive analysis of the ThunderSTORM output was carried out, as shown in Figure S5. The frame by frame ThunderSTORM output demonstrated that the bright spot and the heterogeneous background were both recovered in resulting super-resolution SERS image (Figure S5A&B). We suspect two possible origins for the observed background: scattering from the out of focus nanoparticles, and scattering/fluorescence from the other cell components that transmit through the filter and reach the detector. In either case, the signal can be comparable to the conventional SERS spectra recorded by a traditional Raman spectrometer, such as a scattering from the AuNS-RGDFC-integrin interaction, and requires proper correction^18,42.

We have incorporated two approaches to correct for the non-uniform background observed in our images. The first approach utilizes the single-molecule accurate localization by local background subtraction (SMALL-LABS)⁵² algorithm reported by Biteen and co-workers. Briefly, the SMALL-LABS algorithm estimates the “true” background from “off-frames”, where the molecule of interest is not emitting, or “off”, to detect the “on”, or emitting molecules, in the image sequence. By estimating the mean background from the “off-frames”, the background corrected image of each emitter is obtained by subtracting the off-frames background from each raw image. The super-resolution algorithm is then applied, and the detected signals are returned (green and pink) and false signals are also identified (red). Figure S5C shows the results obtained from applying the SMALL-LABS algorithm to the data in Figure 2. Most of the features in Figure 2D are identified as artifacts; however, the bright spot in #126 frame was inconsistently identified in the data (Figure S5C). We hypothesize that the absolute difference in signal intensity for SERS imaging is less than is typical for fluorescence imaging, where the SMALL-LABS algorithm was designed.

We also explored the extreme value-based emitter recovery (EVER)^53,54 algorithm for background correction. In the EVER algorithm^53,54, the emitters (or nanoparticles) and the background change at different frequencies and are fit by separate Poisson distributions. When the intensity fluctuations of the emitter are similar to the background, it can become difficult to differentiate them from the background. However, since the median of the intensities of the emitter and the background are found as extremes values in our data, a linear function can describe the contribution of the background in the observed signal. Once established, the “true” background can be easily separated from the foreground emitters/nanoparticles. Figure 3 shows the importance of the background correction to achieve a reliable super-resolution SERS image in this work. Figure 3B shows the average SERS wide-field image recovered after using the EVER algorithm. It is clear that the background is significantly reduced, and the previously noted bright spot is quite pronounced (Figures 3A&B). ThunderSTORM⁵⁰ analysis was then performed to detect the nanoparticles after correcting the background with EVER (Figure S5D). Figure S5D shows that nanoparticle was correctly detected, and no false positives were observed. Moreover, the resulting super-resolution SERS image in Figure S5E confirms that only the region surrounding the bright spot was recovered rather than the background previously observed (Figure S5B). Figure 3C shows a merged image plot of the average, EVER background corrected SERS image in blue and super-resolution image in green, highlighting the gain in resolution. The advantage of properly corrected the background in the super-resolution image demonstrated here is clear. This result demonstrates an advantage of our method in comparison to fluorescence-based super-resolution microscopy approaches. When the density of a labeled protein is high, it is challenging to precisely identify a single fluorophore inside the diffract-limited spot, which requires a multiemitter fitting^55,56. While here the entire cell is illuminated by the laser excitation, only a few regions exhibit SIFs from enhanced scattering, which is consistent with detection of the AuNS-RGDFC binding to α_vβ₃ receptor^26,27,32,33. The spurious signals, not originating from nanoparticle binding are removed after the background subtraction. Figure S6 shows similar results for a single cell, where three isolated nanoparticles were detected.

Figure 3. Impact of background on super-resolution analysis.

(A) Average wide-field SERS image (n = 460 frames). A strong background also is observed. (B) The EVER algorithm is used to generate a background corrected image, where it is possible to visualize the enhancement of the probed area and background reduction. (C) Merged imaged of the EVER-corrected background (in blue) and super-resolution SERS image (green) shows the gain in resolution after the SERS fitting. The zoomed in- images show the active cluster is composed of a bright central emitter and possibly three other particles (C), which are not evident in the diffracted-limited spot in (A) and (B). The scale bars are 2 μm for (A), (B) and (C), and 400 nm for the insets.

To further evaluate our methodology, images were recorded at a rate of 100 Hz, or 10 ms per frame (Figure S7). We found that even with faster image acquisition, it was possible to image an entire single cell with a good S/N ratio and detect signals characteristic of single nanoparticles binding to the integrin receptor (Figure S7B). The full-width at half maximum (FWHM) of the super-resolution image was determined to be 50 nm (Figure S7C, D). By comparison, bare AuNS (Figure S8) exhibit a poorer resolution (FWHM = 140 nm) even with longer signal accumulation (rate = 10 Hz). The poorer localization suggests that bare AuNS are not successfully targeted to a specific receptor on the membrane, and therefore may retain some translational mobility. Some of the average super-resolution images (Figure S8) show a certain degree of asymmetry or elongation, which may suggest a nanoparticle dimer^46,57 is detected. This may indicate in the absence of the protein-ligand interaction, the scattered Raman signal is only strong enough to be detected from aggregates.

Super-resolution SERS nanoparticle tracking: AuNS-RGDFC vs bare AuNS.

The results demonstrated in the previous section indicate that we can locate and track the AuNS-RGDFC in cells with a higher localization precision than bare AuNS. This suggest that SERS images can use to pinpoint and identify single nanoparticles binding to receptors.^18,35,42 Figure 4 shows a comparison of individual functionalized (AuNS-RGDFC) and non-functionalized (bare AuNS) nanoparticles detected in cells. Figure 4A shows the centroids resulting from the 2D Gaussian fit to the emission of a very bright AuNS-RGDFC nanoparticle and Figure 4B shows the result for a bare AuNS. We estimated the lateral resolution using the standard deviation of centroids as a function of time, consistent with prior reports^{16, 23, 25, 26, 28, 35}. Again, the cells incubated with AuNS-RGDFC shows a superior precision within ~6 nm (Figure 4A) versus ~123 nm for the bare AuNS (Figure 4B). This change in localization precision is remarkable and it may result from strong binding between the RGDFC peptide and the integrin receptor shown in the scheme in Figure 1A. Furthermore, this precision is smaller than AuNS probe, which suggests we are only collecting SERS photons originating from the AuNS-RGDFC interacting with the α_vβ₃ receptor system. This was further confirmed by estimating the uncertainty of a Gaussian fit to be ~1.4 nm (sigma = 14 nm (precision), minimum photon intensity, N = 296; standard deviation of the background in photons, b = 3.77; and the pixel size of the EM-CCD/CMOS camera in nanometers, a = 65 nm) for the AuNS-RGDFC.⁵⁸ Prior work shows the AuNS-RGDFC bound to the α_vβ₃ integrin receptor gives a significant and specific enhanced Raman response^26,27,34,59. These results corroborate the previous observations using TERS^32,33 and later by a diffracted-limited illumination using SERS.^26,27 The increased localization is believed to derive specifically from RGDFC-integrin interaction on the AuNS. While incubating cells with bare AuNS has been shown to have a SERS signal, the enhanced bands observed in the SERS spectrum are heterogeneous and non-specific, suggesting random and weak interactions with the membrane of cells and emission from the entire AuNS. The quality of the fits further supports this analysis, as the residuals of the centroids plotted in Figure S9 and S10 indicate a single particle is fit in each frame.

Figure 4. Super-resolution SERS trajectories solved in time for individual nanoparticles.

SIFs trajectory over time for an individual (A) AuNS-RGDFC (localization precision = 6 nm) and (B) bare AuNS (localization precision = 123 nm) plotted in the same scale. The data for the bare AuNS (B) are spread over a larger area than the AuNS-RGDFC (A). The black arrow in the zoomed-in figure shows a vector (r = 68 nm) indicating the net direction of the movement of the centroids in time for the AuNS-RGDFC. No preferential direction is observed for bare AuNS in (B). (C) 3-dimensional (3D) plot shows the stochastic changes in the centroid trajectory over time for a different AuNS-RGDFC interacting in a cell (localization precision = 7 nm). (D) 3D plot for showing that the centroid trajectory is constant for the bare AuNS. The axis in (C) and (D) were normalized between −1 and 1 for comparison. (E) Diffusion coefficient estimated for AuNS-RGDFC and bare AuNS (n = 6 nanoparticles). The dashed circle indicates an outlier detected.

Further examination of centroid movement shown Figure 4A (zoomed-in), indicates preferential motion along the y-axis is observed in relation to movement along the x-axis. The black displacement vector (r = 68 nm) annotated on the zoomed-in image indicates that the centroid position of the nanoparticle is changing in time from frame #0 to #1000 (black to yellow color). Figure 4C shows this preferential translation from a second experiment visualized using a 3-dimensional (3D) plot (∣X / nm∣, ∣Y / nm∣ and time / s). In Figures 4C, D, we have normalized the displacement data to look for this preferential translation. This behavior is observed repeatedly for different AuNS-RGDFC. It is important to note that the unfunctionalized AuNS shows no preferential motion and in either direction over the same periods of time (Figure 4D).

The centroid displacement can be used to calculate the diffusion coefficient for the nanoparticles interacting with the cells (See Equation 2 SI). Figure 4E shows a boxplot built using the values in Table S2 for individual nanoparticles with AuNS-RGDFC and bare AuNS. The diffusion coefficient obtained for AuNS-RGDFC was 4.4 × 10⁻¹¹ cm²/s (n = 6) against 7.8 × 10⁻¹⁰ cm²/s (n = 6) for bare AuNS (Figure 4E). The slower diffusion coefficient for AuNS-RGDFC is consistent with nanoparticles bound to the membrane receptor, and is consistent with other results reported in literature^33,60. Interesting, by increasing the laser power from 2 mW to 7 mW at the sample, the measured diffusion coefficient increases one order of magnitude from 4.4 × 10⁻¹¹ cm²/s to 1.1 × 10⁻¹⁰ cm²/s (See Figure S11 for more details). The calculated Stokes-Einstein diffusion coefficient is 2.2 ×10⁻⁸ cm²/s for a 110 nm particle in water at 298K, clearly showing reduced mobility in our samples. The ~3× increase in the magnitude of diffusion coefficient would correspond to a temperature increase to ~900 K, or more likely, an increase in mobility suggesting a weakening of the binding interaction with the surface. The diffusions coefficients for the AuNS-RGDFC are still slower than the expected bulk diffusion, which further indicates that AuNS-RGDFC are interacting with the membrane surface.

Previous literature reports demonstrate a shift in the SERS centroid over time due to changes in the near field scattering properties of the nanoparticle.^34,35 These phenomena are dependent on the shape, size, interparticle spacing, and polarization angle of illumination. In our experiments, we are monitoring individual nanostars with near uniform excitation illumination. In contrast to the previous work where a change in the centroid could be mapped onto the structure of the nanoparticles, in our experiments the particle has the ability to move and/or reorient in space and in relation to the electric field polarization. Consequently, the changes in SERS intensities can also originate from the AuNS-RGDFC-α_vβ₃ conformation, which could be manifest as blur in the spots recovered in the previous section and the large FWHM observed before (Compare Figure S7, S8).

Figure S12A, B shows a revised plot of Figure 4A correlating the signal intensities with their positions and in time. The centroids are color coded with respect to the intensity thresholds indicated (Figure S12A, B). While low SERS intensities spread out around the center, while high SERS intensities localized in the middle of the pattern (Figure S12A, B). The intensity data indicates the highest intensities are localized to a very small areas (≤ 6 nm), which is similar to other reports^18,42. Figure S12C examines the time scales of the observed fluctuations. One explanation is there are changes in the interaction, and thus enhanced vibrational modes, between the AuNS-RGDFC and α_vβ₃ receptor^61-64. The AuNS-RGDFC- α_vβ₃ conformation may be tilting and altering the alignment with respect to the illumination polarization (Figure S12D). This is confirmed by the 2D plot of SERS spectra shown in Figure S13, where characteristic SIFs were observed for the Raman bands at 1559 cm⁻¹ (tryptophan) and 1625 cm⁻¹ (phenylalanine) (Figure 2E). The asymmetric structure of the AuNS is expected to have a polarization dependence with respect to the electric field and the orientation of the branches, consequently, a change in orientation could alter the intensity of the centroid of the nanoparticle over time (Figures 4A, C). The fact that the high intensity fluctuations and distribution appear to occur in the same physical space and are very localized supports the idea of a changing conformation relative to the polarization of the electric field as demonstrated before in single-nanoparticle SERS experiments⁴². This change can be easily visualized by the frequencies of SIFs in Figure S12C, where most of the SIFs rate presented were ≤25 Hz (white threshold), or ≥40 ms. This result is reasonable compared to the time scale of changes in protein conformation^65-67. This interpretation is also consistent with previous literature on integrin conformation^61-63 and thermal activation by the laser illumination. This observation suggests the protein conformation is active, as shown also in Figure S12C, however, it is turned-off after a few seconds due to the interconversion process^61-63. Figure S12D proposes a mechanism by which the orientation of the AuNS-RGDFC-α_vβ₃ conformation could change and give rise to fluctuations in the SIFs observed here in this work.

Super-resolution SERS imaging in the Z-direction.

Using the localization precision demonstrated, we sought to address the question of whether nanoparticles stay on the cell surface or are internalized after binding to the receptors.^68,69 To answer this question, we collected wide-field SERS images in series while changing the focal plane every 1000 frames by 1 μm steps in the Z-direction to create a 3D-image stack through a full cell (Figure 5). Figure 5C shows the merged plots of wide-field SERS images (in red) after the EVER background correction and super-resolution SERS nanoparticle (in green) detected by ThunderSTORM⁵⁰ for different Z-slices. Four nanoparticles were detected on the cell in the highlighted areas in the −1, −2, −4 and −5 μm slices (Figure 5C). While the first nanoparticle was detected on top of the cell in the position −1 μm, the particles in the −2, −4 and −5 μm slices suggest possible internalization within the cells (Figure 5C). Although our spatial resolution is limited to −1 μm steps, these results suggest that different nanoparticles are being probed and detected from different x, y and z positions in the cell. An interesting fact in the super-resolution Z-stack SERS image in Figure S14 is that the nanoparticles on top (−1 μm) and at the bottom (−5 μm) of the cell present a crescent like shape pattern, which was not observed in the nanoparticle within the cell (−2 μm and −4 μm). This result is consistent with an oriented probe due to the curvature of the cell membrane, as the cell shape is not uniform.^57,70-72 The size of the crescent features is consistent with a single nanostar, and may indicated binding to two or more branches of the AuNS. These initial results suggest further improvements in z-imaging may provide additional insight into the conformation of the AuNS-peptide probes interacting with cells. In summary, the results showed here suggest that our method is not only suitable for locating and tracking binding events on the membrane of the cells but also for internalization of the nanoparticles and possible caveats about the probed biomolecule’s orientation.

Figure 5. Z-stack super-resolution image for fixed cells incubated with AuNS-RGDFC.

(A) Bright-field image of a single cell. (B) Z-stack mapping were carried out with steps of 1 μm for every 1000 frames totalizing 5 μm, where 0 μm position is the top and −5 μm is the bottom of the cell. (C) Merged image of background corrected image (in red) and super-resolution SERS images (in green) for each Z-stack slices, suggesting the internalization of some nanoparticles. The scale bars are 2 μm. (D) Merged image of bright field image of the cell (in blue) and super-resolution SERS images (in green). Zoomed-in image are the selected are in the cell and shows nanoparticles found in independent locations in x,y in the cell, where with the crescent shaped p.s.f. More details in supporting information and main text. The scale bars are 100 nm (#1) and 150 nm (#2 to 4).

CONCLUSIONS

A super-resolution SERS imaging method for monitoring integrin receptors in cells is demonstrated. AuNS-RGDFC binding to α_νβ₃ receptor on the membrane of the cell was visualized with a precision of up to ~ 6 nm, and the identity of the binding was further confirmed by the SERS spectrum. The slower diffusion coefficient for AuNS-RGDFC demonstrated that the nanoparticles were strongly bound on membrane receptor in comparison to bare AuNS. The nanoparticles show some translational diffusion over the surface, and possible orientation changes that suggest protein-conformation changes on the time scale of ≥40 ms. Z-stack images provide a 3D correlation between the super-resolution SERS images and evidence that some of the AuNS-RGDFC were internalized within the cells. Additionally, the p.s.f of some nanoparticles evinced a crescent shape on top of the cells, suggesting the orientation or possible clustering of AuNS-RGDFC-α_νβ₃ can also be elucidated. This method provides non-invasive, fast, and chemically specific high spatial resolution cellular imaging, with potential implications for new applications in life science and biomedical analysis. The ability to image receptor binding events with high precision and also obtain vibrational spectra of the chemical interactions on nanometer length scales provides researchers a new approach to understand important biochemical interactions and mechanism. This is an important advance toward understanding, for example, biochemical signaling and drug delivery interactions in biological samples.