The advent of electron microscopy (EM) has proved to be a powerful tool that can provide high-resolution images [9, 10]. Nevertheless, stringent experimental requirements such as high vacuum conditions, sample staining using external chemical, sample dehydration, powerful electron beam etc. are inherent drawbacks of EM that could be detrimental for imaging soft, delicate biological samples. On the other hand, scanning probe microscopic (SPM) techniques such as atomic force microscopy (AFM) has enabled us to obtain valuable insights into the morphological and topographic details of biological samples in the nanoscopic regime. However, AFM lacks the wealth of information that is obtained from the optical microscopic techniques [11–13]. In recent years, various super-resolution imaging techniques have been developed that have substantially overcome the diffraction-limit. Fluorescence-based microscopy methods such as stimulated emission depletion microscopy (STED) [14, 15], saturated structured-illumination microscopy (SSIM) [16], photoactivated localization microscopy (PALM) [17], fluorescence photoactivation localization microscopy (FPALM) [18], stochastic optical reconstruction microscopy (STORM) [19] and direct STORM (dSTORM) [20] have been successfully utilized in improving the spatial resolution by an order of magnitude and unravelling biological structural details.

Apart from the aforementioned super-resolution imaging methods, near-field scanning optical microscopy (NSOM) is another high-resolution SPM-based technique that bridges the AFM and optical microscopy, thus, resulting in a simultaneous generation of topographic and optical images of a sample [1–7]. In NSOM experiments, light is sent through an aperture made up of tapered optical fiber, which is used as a source of illumination and the diameter of the optical probe (aperture) is much smaller than the incident wavelength. The aperture is brought very close (≤10 nm) to the sample surface at a distance much smaller than a wavelength (near-field) and the sample is scanned by employing a shear-force feedback principle that controls the tip movement relative to the sample surface. During the scan, evanescent waves emanating from the aperture interact with the sample surface before any interference due to the diffraction of light occurs, and elicit optical response (absorption, fluorescence, Raman scattering) from the sample while a topographic image of the sample is also simultaneously generated. The evanescent field decays very sharply as a function of increasing distance hence generating a maximum fluence of incident radiation near the sample surface [6, 7]. Depending on the sample characteristics and experimental requirements, the optical signals are collected by suitable detection assembly and recorded electronically in either transmission or reflection geometries. Typically, the transmission mode is utilized if the sample is transparent whereas the reflection mode is preferred for non-transparent samples. The resolution of the optical images is independent of the incident wavelength but depends on the diameter of the sub-wavelength aperture that governs the XY-resolution (~50–100 nm) as well as the distance between the probe and the sample surface (that determines Z-resolution; ~10 nm). The other added benefit of NSOM is the absence of out-of-focus light which is in contrast to that commonly observed in conventional far-field microscopy [6, 7]. Taken together, NSOM is a non-invasive imaging technique which provides high-resolution, sub-diffraction optical and topographical images and has been successfully employed to elucidate the structural details of a wide range of biological systems. Additionally, infra-red (IR) and Raman-based imaging methods in which the spectroscopic technique has been coupled with scanning probe microscopy namely, apertureless near-field scanning infrared microscopy (ANSIM) [21, 22], coherent anti-stokes Raman scattering (CARS) [23] and tip-enhanced Raman scattering (TERS) [24–27] have also been utilized to extract both topographical and molecular structural details of a variety of biological systems. Recently a “nanoscopic tool”, comprising a combination of AFM and STED microscopy, for imaging biological samples has been reported [28]. In the current review, we discuss a few recent applications of nanophotonics in unravelling the structural details of an important class of biological supramolecular assemblies, namely amyloids that are made up of proteins and are the hallmarks of a variety of neurodegenerative disorders. In the following section, we provide a brief overview on amyloids.

Proteins are biological macromolecules which act as efficient miniature engines in the living systems and carry out a multitude of important functions in their correctly folded form. However, changes in solution conditions and/or mutations in the protein sequence result in the predominance of intermolecular interactions between the protein molecules over intramolecular interactions. As a consequence, aberrantly-folded or misfolded states of the protein are formed that subsequently undergo self-assembly and accumulate as cross β-sheet-rich amyloid aggregates [29–34]. The cross β-structure is a distinct signature of amyloids wherein the β-sheets are oriented parallel to the fibril axis and the constituent β-strands in each β-sheet are oriented perpendicular to the fiber axis (Figure 1A). This is further demonstrated by X-ray diffraction pattern of amyloids which shows two discrete and intense reflections at 4.7–4.8 Å (meridional; spacing between adjacent β-strands) and 10 Å (equatorial; “face-to-face separation” between β-sheets) (Figure 1B). Amyloid aggregates are implicated in a plethora of debilitating human disorders such as Alzheimer’s and Parkinson’s, Huntington’s and prion diseases and systemic amyloidosis. Studies have revealed that the transition from a normal functional protein to an abnormal, misfolded form involves a major conformational change that serves as the pivotal step in protein aggregation leading to amyloid fibril formation [29–31]. It has been documented that various factors such as the amino acid sequence govern the stability and the morphology of amyloids [29]. Additionally, the process of amyloid fibrillation comprises an inherent stochasticity owing to the formation and accumulation of transient oligomeric species that are eventually sequestered into amyloid aggregates [30, 31]. The protein amyloids and their precursors are cytotoxic and localize within the cell membranes, thus resulting in membrane disruption and subsequent cell death [35–38]. However, a complete understanding of protein aggregation and amyloid assembly still remains elusive. Also, the investigation is further complicated due to the existence of amyloid strain diversity that arises due to heterogeneous packing of β-sheets within the amyloid fibrils [39–41]. Hence, high-resolution structural elucidation of protein amyloids is expected to shed light into the conformational heterogeneity that is manifested in amyloid polymorphism and the precise nanoscale localization and architectural details of amyloid-membrane assemblies.

Figure 1

(A) Ribbon diagram of an amyloid fibril based on a structural model derived from a combination of solid state NMR and electron microscopic data. The amyloid fibril is viewed parallel (left) and perpendicular (right) to the fibril axis. Reprinted with permission from ref. [32] (authors of this article were from the NIH). (B) X-ray diffraction pattern of a cross β-sheet-rich amyloid fibril showing two discrete reflections at 4.7–4.8 Å (meridional) and 10 Å (equatorial). Reprinted from ref. [33], Copyright (2012) with permission from Elsevier.

Although the near-field technologies were developed several decades ago, it is only recently that the capabilities of these methodologies are being utilized in amyloid research in order to address a variety of important issues. In a recent report, the binding mechanism and the orientation of amyloid-specific dyes namely, Thioflavin-T (ThT) and its neutral analogue, 2-[4-(dimethylamino)phenyl]-benzothiazole (BTA-2), to insulin-derived amyloid fibrils was investigated using polarization-fluorescence-based NSOM [42]. The authors proposed that BTA-2 and ThT, being structurally similar, might exhibit similar binding pattern to amyloid fibrils, however, the “binding motif” of BTA-2 (neutral dye) might differ from that of ThT (ionic dye). Additionally, efforts were directed to ascertain whether the fluorescence emission is from monomeric or excimer or micellar-forms of the bound dyes. It was speculated that if monomeric dye binds to the β-sheet-rich grooves having a parallel orientation to the fibril axis, the fluorescence emission will be polarized along the fibril axis. Whereas, if excimer is formed, the emission is expected to be red-shifted compared to that of the monomeric dye with a simultaneous increase in the fluorescence lifetime. If micellar dye binds to the amyloid fibrils, the polarization will be lost and an isotropic emission will be observed. In order to identify which binding model is more appropriate, the concentrations of ThT and BTA-2 were chosen much higher than their respective critical micelle concentrations to investigate whether micellar-dye binds to the β-sheet-rich amyloids. Before embarking on the NSOM experiments, steady-state and time-resolved fluorescence experiments were carried out on insulin fibril solution containing either of the dyes which revealed that the species of ThT and BTA-2 bound to the fibrils is predominantly monomeric, thus excluding the possibility of excimer formation and micellar dye binding. Prior to NSOM experiments, the fibril containing dye solution was loaded on a chemically-modified glass coverslip. During NSOM imaging, the fluorescence emission from the dyes was divided into vertically- and horizontally-polarized fluorescence using a polarizing beam splitter. The vertical- and horizontal-polarized fluorescence images showed vertically and horizontally-oriented fibrils, respectively indicating that the dye fluorescence is polarized (Figure 2). The absence of any vertically-oriented fluorescent fibril in the horizontal-polarized fluorescence image and vice-versa revealed that the amyloid-staining dyes, ThT and BTA-2, bind to the β-sheet-rich grooves in such a way so that the molecular axis of the dye is parallel to the fibril axis. Additionally, such observation reaffirmed that the binding of the dye to amyloid fibrils is primarily governed by the molecular geometry (shape and size) of the dye and is independent of any electrostatic interactions between the dye and the fibrils.

Figure 2

(A) AFM image (5 μm×5 μm) of insulin amyloid fibrils formed at pH 2, 60°C. NSOM images (5 μm×5 μm) of Thioflavin-T (ThT)-stained insulin amyloid fibrils (B) topography image, (C, D) ThT fluorescence images in the vertical (indicated by a double-headed vertical arrow) and horizontal (indicated by a double-headed horizontal arrow) polarizations, respectively. In all of the NSOM images, the yellow and white boxes represent the horizontally- and vertically-oriented amyloid fibrils, respectively. Reprinted with permission from ref. [42], Copyright (2009) American Chemical Society.

Recently, we have performed super-resolution imaging of fluorescently-stained amyloid fibrils derived from the low pH form of β₂-microglobulin that is known to be associated with dialysis-related amyloidosis [43]. In our NSOM experiments, we used Nile Red as the fluorescent marker rather than more commonly used ThT due to the following reasons: (i) ThT has also been shown to bind to amorphous protein aggregates owing to the presence of inherent charge [44], (ii) ThT assay at neutral pH cannot be performed to detect amyloid fibrils that are formed at low pH and disintegrate at neutral pH [45], (iii) the low quantum yield of ThT prevents it from being utilized in single-molecule/particle imaging experiments [46] and (iv) ThT exhibits an emission maximum at ~482 nm upon binding to amyloid fibrils and hence, commonly used laser lines such as 488 and 532 nm cannot be used as excitation wavelengths. All these problems can be circumvented if Nile Red is used [47]. Hence, Nile Red was added into the solution containing β₂-microglobulin-amyloid fibrils to yield Nile Red bound-amyloid fibrils prior to our NSOM imaging experiments. We used transmission mode fluorescence-NSOM technique wherein an optical fiber of tip diameter ~100 nm was used and coupled with the 488 nm line of an Ar⁺ laser and the fluorescence emitted from the Nile-Red-stained fibrils was collected using the far-field optics (Figure 3). During the NSOM imaging, we obtained concurrent and spatially-correlated topographic (AFM) as well as fluorescence (NSOM) images (Figure 4A,B) and the strong spatial correlation was also evident from the 3D overlays of NSOM (fluorescence) over AFM (topographic) image and vice-versa (Figure 4C,D). Synchronous line profiling across the width of fluorescently-stained amyloid fibrils in both AFM and NSOM images provided the height of the fibrils (in nm) and Nile Red emission intensity (in kHz), respectively (Figure 5A). Additionally, the amyloid fibrils that were separated by ~75 nm could be spatially resolved in both AFM topography and NSOM fluorescence images (Figure 5B,C). Thus our near-field imaging experiment provides an optical map of fibrils at the nanoscale resolution that is well beyond the diffraction-limit [43]. Next, we addressed the key issues related to the structural heterogeneity/polymorphism that may arise due to the diversity in the supramolecular packing of protein molecules within each amyloid fibril. In order to address this issue, we carried out brightness analysis because we speculated that any alteration in the fibril fluorescence brightness within an individual fibril and among different fibrils could be associated with the variations in the binding affinity of Nile Red to fibrils. Such variability would reflect the heterogeneity of the supramolecular assembly. To perform brightness analysis, a large number of synchronous line profiling across the width of a single fibril was carried out, akin to multiple hypothetical segmenting, which generated a simultaneous optical (fluorescence) and topographic (height) profiles for a given cross-section (Figure 5A,D,E). Each of the pair of profiles, optical and topographical, obtained for the respective cross-section, was then integrated separately to obtain the total fluorescence counts (NSOM image) and the cross-sectional area (AFM image). The fluorescence brightness along each profile (or hypothetical segment) was estimated as the ratio of fluorescence counts to the respective cross-sectional area (fluorescence counts per unit area in kHz/nm²) and statistical analyses of these ratios demonstrated alterations in the fluorescence brightness within individual fibrils (Figure 5E). In order to probe whether different fibrils exhibit heterogeneity in the brightness, similar analyses (fluorescence per unit area) were carried out that depicted significant alterations among different fibrils. Additionally, to address the concern whether any plausible broadening in the brightness analysis occurred due to tip-induced convolutions, we performed another set of statistical analyses whereby the ratio of the maximum fluorescence counts to the maximum height (kHz/nm) within and among individual fibrils were taken into consideration. These two types of image analyses essentially lead to the same conclusion and thus reaffirm the diversity in supramolecular packing within the amyloid fibrils.

Figure 3

Schematic representation of the NSOM-fluorescence set up utilized to image Nile Red-stained β₂-microglobulin amyloid fibrils, formed at low pH, in the transmission mode. An optical probe of tip diameter ~100 nm was coupled with the 488 nm line of an Ar⁺ laser which was used as a source of illumination. The fluorescence from the Nile Red was collected using far-field optics comprising a 50× long working distance objective and an appropriate combination of optical filters (488 nm notch filter and 550 nm long pass filter) that were used to block the scattered excitation light and allow the fluorescence emission to pass through. The fluorescence was detected by a highly sensitive avalanche photodiode (APD). Reprinted with permission from ref. [43], Copyright (2012) American Chemical Society.

Figure 4

NSOM images (10 μm×10 μm) of Nile Red-stained β₂-microglobulin amyloid fibrils. (A) Topographic (AFM) image and (B) the respective NSOM fluorescence image. (C, D) 3D overlays of NSOM (fluorescence) over AFM (topography) and vice-versa, respectively to demonstrate that the images are strongly spatially-correlated. Reprinted with permission from ref. [43], Copyright (2012) American Chemical Society.

Figure 5

Analysis of the topography and fluorescence of single amyloid fibrils observed during NSOM-fluorescence imaging. (A) NSOM fluorescence image of a few well-separated single amyloid fibrils (I-III). (B, C) AFM (topography) and NSOM (fluorescence) profiles along the white line (Figure 5A, bottom) show that the two fibrils separated by ~75 nm can be resolved by the NSOM technique. (D) Magnified NSOM images of the individual fibrils (I–III) and (E) the brightness analyses of corresponding fibrils depicting the variations in Nile Red fluorescence emission along a single fibril as well as among different fibrils owing to the heterogeneity in supramolecular packing. Reprinted with permission from ref. [43], Copyright (2012) American Chemical Society.

In regular or aperture-based NSOM, the image resolution is restricted by the aperture size. If the optical aperture is too small say, ~10 nm, sufficient light cannot pass through. Consequently, an image resolution <10 nm cannot be obtained. Hence, in order to achieve sub-10 nm resolution, the concept of apertureless NSOM was introduced [21]. Typically, semiconductor based/metalized AFM tips or tips made by electrochemical etching are used in apertureless NSOM experiments. The AFM tip acts as a “small antenna” and the evanescent field between the tip and the sample is magnified wherein the light is detected from the tip apex. If the tip apex radius is <10 nm, a sub-10 nm resolution can be achieved even in the infra-red (IR) region. During the scan, the amplitude is kept constant and the tip and the sample are maintained at an intermittent contact mode that is controlled by the shear-force AFM feedback loop. Additionally, the near-field between the tip and the sample surface is modulated in such a way so that the background signals are suppressed efficiently. When IR is coupled to the apertureless NSOM (ANSIM), it provides the structural information (vibrational frequencies) of a sample at the molecular level while the topographic map is simultaneously generated. However, the structural information is only limited to the sample surface. ANSIM was utilized to perform secondary structural characterization of single amyloid fibrils derived from the peptide fragment (²¹NFLNCYVSGFH³¹) belonging to β₂-microglobulin in combination with other techniques such as, AFM, TEM and FT-IR [22]. Both the AFM and TEM images of the amyloid fibrils revealed distinct morphologies of amyloid fibrils at different stages of fibril formation. In addition to the topographic investigation, FTIR spectra of the amyloid fibrils were recorded both in the presence and in the absence of an osmolyte namely, trimethylamine N-oxide (TMAO), because it has been observed that patients suffering from renal failure have elevated levels of TMAO. The fibrils in the absence of TMAO showed vibrational peaks at 1630 and 1668 cm^-1whereas that in the presence of TMAO showed peaks at 1630, 1665, and 1692 cm^-1. Next, ANSIM experiments were carried out to investigate the structural heterogeneity among each β₂-microglobulin fibril formed in the presence of TMAO wherein tunable CO laser was used as the IR source. The ANSIM set-up used in this study was based on interferometric reflection type apertureless NSOM. Briefly, a multi-mode AFM tip was used in the tapping mode that could generate the topographic features of the amyloid fibrils as well as the required near-field. The CO laser beam was focused onto the AFM tip apex wherein the polarized beam was oriented parallel to the probe axis to obtain an enhanced scattered field. The scattered field was further enhanced by homodyne detection whereby the homodyne field reflected from a piezocrystal was added to the scattered IR radiation from the sample resulting in maximum interference. The vibrational frequency range was set at 1600–1700 cm^-1for detection. At each wavenumber, the morphological features (fibril topography) and near-field (scattered) images of amyloid fibrils were obtained simultaneously (Figure 6). A significant variation in the scattered intensities of the fibrils was observed though the heights of the fibrils were predominantly uniform. Also, the comparison of the theoretically-predicted and experimentally-derived IR scattered fields from different amyloid fibrils showed a good agreement of the peak positions obtained from FT-IR experiments (Figure 6A). The study also revealed that the amyloid fibrils consist of parallel β-sheets and certain morphologically distinct fibrils might have higher binding affinity towards TMAO, thus providing an explanation of the presence of increased TMAO levels in patients. Hence, the ANSIM study provided detailed secondary structural insights into the individual amyloid fibrils.

Figure 6

ANSIM data obtained for the amyloid fibrils derived from the peptide fragment (NFLNCYVSGFH) belonging to β₂-microglobulin. (A) The experimental scattered field is represented by the left y-axis and the calculated imaginary part of effective polarizability is represented by the right y-axis as a function of wavenumber. The ANSIM scattered field for the fibril in the absence of TMAO (•) and in the presence of TMAO (○) is represented. The dashed-lines serve as eye-guides. Simultaneous acquisition of (B) topography image and (C–F) near-field images collected at four different wavenumbers. The regions of fibrils used to obtain the ANSIM spectra (Figure 6A) are illustrated by the white boxes in the topography image. Reprinted with permission from ref. [22], Copyright (2011) American Chemical Society.

Apart from the aperture-based or apertureless NSOM, significant advances have been made towards understanding the structure and composition of prion- and insulin-derived amyloid fibrils by utilizing TERS imaging technique that provides a spatial resolution of ~10 nm [25–27, 48]. A TERS setup involves an amalgamation of AFM and a laser Raman spectrometer. Briefly, a single gold or silver metal nanoparticle is attached to the AFM tip apex and the evanescent field generated by the incident laser light excites the surface plasmons of the metal-coated tip that magnifies the vibrational modes of the molecules. During the topographic scanning of the specimen, an overall enhancement in the Raman signal is observed. The primary advantage of TERS imaging lies in the fact that different amino acids and the secondary structural contents can precisely be localized and assigned based on the Raman scattering from the amyloid fibrils without the involvement of any extraneous fluorescent labels.

Over the past decades, the technological advancements in the field of nanophotonics have enabled us to gather super-resolution optical images and to appreciate the resplendence of seemingly intricate biological systems and processes. These techniques are now beginning to unravel the important high-resolution details of complex amyloid assemblies. Recently, significant structural information on the amyloids has been amassed from state-of-the-art super-resolution optical imaging techniques. However, further developments in these imaging techniques are required to unravel the structural details and the underlying mechanisms at the molecular level that will eventually help us design the therapeutic strategies against the amyloid disorders. In addition to the rod-like linear fibrils, annular pore-like morphologies are also prevalent in many diseases and are known to efficiently permeabilize cell membranes. Unravelling the structural factors that regulate the formation of a linear or pore-like protein aggregates and the amyloid-membrane interactions on the cell surface by using multicolour illumination and detection strategies at the single-molecule/particle resolution would profoundly enrich our understanding of the biology of amyloids.

We thank the Mukhopadhyay lab members for their research contributions described in this article and for critically reviewing the manuscript and Shruti Arya for helping us with the figures. M.B. thanks the Department of Science and Technology (DST)-Women Scientists’ Scheme for research fellowship and grant. S.M. thanks the Council of Scientific and Industrial Research (CSIR) and IISER Mohali for financial support.