Introduction
Heterogeneity is abound for biomolecules; even nominally identical proteins have folding variations, conformational changes, and post-transcription modifications and exist at various stages of the enzymatic cycle (Martin-Baniandres et al., Reference Martin-Baniandres, Lan, Board, Romero-Ruiz, Garcia-Manyes, Qing and Bayley2023). With single-molecule approaches, it is possible to resolve these differences as well as observe dynamics without the need for synchronization (Moerner, Reference Moerner2002; Eisenstein, Reference Eisenstein2012; Ha, Reference Ha2014; Miller et al., Reference Miller, Zhou, Shepherd, JMWollman and Leake2017). It is possible to observe variations particular to the location or interaction state of a biomolecule and observe new or rare events that are obscured by ensemble averaging. Single-molecule studies are also the pinnacle of sensitivity and allow for analysis in crowded physiological environments. Single-molecule methods can observe interaction kinetics at equilibrium by relying on discrete “on” and “off” events, whereas ensemble measurements monitor the approach towards an equilibrium point (Al Balushi and Gordon, Reference Al Balushi and Gordon2014b; Fineberg et al., Reference Fineberg, Surrey and Kukura2020).
Among optical methods, fluorescence has played a dominant role in the analysis of single biomolecules. Fluorescence-based methods have allowed for analysis of super-resolution structural information, particle tracking, folding/unfolding dynamics, and interactions (Chung et al., Reference Chung, McHale, Louis and Eaton2012; Diezmann et al., Reference Diezmann, Shechtman and Moerner2017). Despite this overwhelming success, attaching fluorescent labels can significantly modify the biophysics. Surface plasmon resonance studies have shown a 3–4 times difference in the equilibrium dissociation constants with streptavidin-peptide binding upon labelling with Cy3 (YS Sun et al., Reference Sun, Landry, Fei, Zhu, Luo, Wang and Lam2008). Competitive binding studies have also shown variations in dissociation constants (up to two orders of magnitude) for DNA and proteins Dietz et al., Reference Dietz, Wehrheim, Harwardt, Niemann and Heilemann2019. The impact varies with different fluorescent labels used in surface plasmon resonance imaging studies for binding to cells (Yin et al., Reference Yin, Wang, Wang, Zhang, Zhang and Tao2015). At the single-molecule level, using a hybrid plasmonic platform, changes were observed for DNA-protein interactions impacting the diffusion coefficient, the on and off binding rates, the surface potential, and molecular weight (Liang et al., Reference Liang, Guo, Hou and Quan2017). Labelling can introduce variations in protein dynamics (Weisgerber and Knowles, Reference Weisgerber and Knowles2021), interactions with lipid bilayers (Hughes et al., Reference Hughes, Rawle and Boxer2014), artefacts in tracking due to non-specific binding (Zanetti-Domingues et al., Reference Zanetti-Domingues, Tynan, Rolfe, Clarke and Martin-Fernandez2013), and destabilization/collapse of proteins (Riback et al., Reference Riback, Bowman, Zmyslowski, KevinWPlaxco and Sosnick2019).
In addition to modifying the biophysical properties, labelling adds complexity and cost: the labels have to be attached, and optical filters are required. Labelling limits observation time due to photobleaching and other effects. It also limits time resolution due to photon counting and precludes studies of species where labelling sites may not be known (or similarly, multiple labelling sites may be present, and so the attachment location is not deterministic). This may be complemented using X-ray crystallography and diffraction (Ringe and Petsko, Reference Ringe and Petsko1985; Ilari and Savino, Reference Ilari and Savino2008) or by dynamic NMR (Kay, Reference Kay2011) to get accurate protein structures, but this is highly specialized and inaccessible. Photobleaching impacts the time bandwidth, limiting the dynamic range between fastest and slowest timescales to three orders of magnitude (Schmid and Dekker, Reference Schmid and Dekker2021). There have been improvements in labelling due to fluorescence fusion proteins (Luo et al., Reference Luo, Dai, Chen, Yue, Andrade-Powell and Chang2020; Reja et al., Reference Reja, Minoshima, Hori and Kikuchi2021). By genetically encoding fluorescent protein genes, proteins of interest can be fused with fluorescent tags, enabling real-time tracking of their dynamics and studying protein–protein interactions in the cellular environment. However, this technique requires the labels to be genetically encoded before protein expression, adding cost and complexity. With some fluorescence-based studies, there have been conflicting reports of conformational changes since Förster resonance energy transfer (FRET) relies on the fluorophore distance, which can be the same in different conformations (Hanson et al., Reference Hanson, Duderstadt, PWatkins, Bhattacharyya, Brokaw, Chu and Yang2007; Henzler-Wildman et al., Reference Henzler-Wildman, Thai, Lei, Ott, MagnusWolf-Watz, Pozharski, Wilson, Petsko, Karplus, Hübner and Kern2007; Li et al., Reference Li, Liu and Ji2015). Therefore, it is of interest to develop techniques that can observe biomolecules, their dynamics, and their interactions without fluorescent labels.
Techniques that avoid tethering to surfaces are similarly desired because tethering has been shown to hinder binding depending upon where the attachment is, and it also constrains the molecule against diffusion, which may alter the interaction or impact the molecular stability (Shental-Bechor and Levy, Reference Shental-Bechor and Levy2008; Grawe and Knotts, Reference Grawe and Knotts2017). Simulations predicted that tethering in the wrong position can dramatically alter the folding of proteins (Arviv and Levy, Reference Arviv and Levy2012; Carmichael and Shell, Reference Carmichael and Shell2015). Molecular dynamics simulations predicted either stabilization or destabilization of protein structure due to tethering (Levy, Reference Levy2017). Tethering may modify the energy landscape of the protein domains (Busto, Reference Busto1998). It was demonstrated experimentally that enzyme properties changed when immobilized to solid surfaces, leading to a decrease in its activity (Hanefeld et al., Reference Hanefeld, Cao and Magner2013; Rodrigues et al., Reference Rodrigues, Ortiz, Berenguer-Murcia, Torres and Fernández-Lafuente2013). Therefore, as with labelling, tethering can obstruct the natural structure and dynamics of proteins, and techniques that operate free from tethers can be explored to alleviate these points.
Optical scattering approaches are emerging as ways to measure single biomolecules and probe their dynamics, interactions, and properties. Methods that make use of direct scattering, like nanoparticle tracking analysis, are presently not sensitive enough to probe particles below 10 nm in size. Other methods can make use of shifts in resonances of microcavity, plasmonic, or hybrid platforms to enhance the sensitivity. They can also be based on interference scattering, making use of interference between a surface and particle scattering (iScat), or surface plasmon (surface plasmon resonance imaging – SPRi, plasmon scattering microscopy – PSM, or plasmon-enhanced protein tracking through interference – PEPTI) or a nanochannel in nanofluidic scattering microscopy (NSM) (Špačková et al., Reference Špačková, Moberg, Fritzsche, Tenghamn, Sjösten, Šıpová-Jungová, Albinsson, Lubart, van Leeuwen, Westerlund, Midtvedt, Esbjörner, Käll, Volpe and Langhammer2022). Nanoaperture optical tweezers (NOTs) have emerged as a way to analyze single proteins in solution and make use of the optical forces to prevent the protein from diffusing away. Here, we will provide an overview of each of these techniques as well as describe their application in the study of proteins. Figure 1 shows schematic and representative data for various approaches to light scattering measurement of single proteins and other biomolecules.
Simplified representations of label-free, single protein techniques and accompanied results with a ≈ 66 kDa protein. (a) Nanoaperture optical tweezers, schematic showing laser tweezer microscope with an aperture in a gold film where the transmission is monitored with representative trapping signal of Bovine Serum Albumin below (Pang and Gordon, Reference Pang and Gordon2012). (b) Whispering gallery mode, schematic showing evanescent coupling to spherical resonator through an optical fiber with a tunable laser and measurement of the intensity coupled via a photodiode upon attachment of a biomolecule, with representative frequency shift data shown below (Yu et al., Reference Yu, Jiang, Lin and Lu2016). (c) Plasmonic-WGM, schematic showing introduction of gold nanorod to enhance sensitivity and representative wavelength shift data shown below (Note, 44 kDa protein shown) (Toropov et al., Reference Toropov, Houghton, Yu and Vollmer2023). (d) iScat (Dahmardeh et al., Reference Dahmardeh, Dastjerdi, Mazal, Köstler and Sandoghdar2023), schematic showing interference scattering from a biomolecule and surface observed by image subtraction on a camera, and a typical image subtraction shown below. (e) Photonic crystal-plasmonic hybrid (Liang et al., Reference Liang, Guo, Hou and Quan2017), schematic of a 1D photonic crystal cavity with a gold nanoparticle and protein-DNA interactions observed with single molecule sensitivity and representative frequency shift data shown below. (f) Plasmonic scattering microscopy, schematic showing total internal reflection excitation of surface plasmons and imaging from the top, with representative data shown below (Wan et al., Reference Wan, Ma, Zhang and Wang2022). (g) Surface plasmon resonance imaging, showing illumination of a planar gold film exciting an evanescent wave that interferes with the scattering of a protein, imaged on the opposite side of illumination to remove the laser background, with a typical image of single proteins below (Ma et al., Reference Ma, ZijianWan, Zhang, Wang and Tao2020). (h) Nanofluidic scattering microscopy, schematic showing the diffusion of a protein through a nanochannel with interference between protein scattering and nanochannel scattering, accompanied by a representative time/position trace of bovine serum albumin below (Špačková et al., Reference Špačková, Moberg, Fritzsche, Tenghamn, Sjösten, Šıpová-Jungová, Albinsson, Lubart, van Leeuwen, Westerlund, Midtvedt, Esbjörner, Käll, Volpe and Langhammer2022). (i) Plasmon enhanced protein tracking through interference, schematic showing the tracking of a single protein through interference of surface plasmon excited by scattering a laser off a nanoaperture with protein scattering, with representative image of a bovine serum albumin (Peters et al., Reference Peters, McIntosh, Albu, Ying and Gordon2023).
Light scattering
The scattering of light from a single protein can be characterized by the polarizability, which is the ratio of the dipole moment to the electric field. The polarizability of an ellipsoid particle is given by:
$$ \alpha ={\epsilon}_0{\epsilon}_1V\frac{\left({\epsilon}_2-{\epsilon}_1\right)}{L\left({\epsilon}_2-{\epsilon}_1\right)+{\epsilon}_1} $$
where
$ V $
is the particle volume,
$ {\varepsilon}_1 $
is the background relative permittivity,
$ {\varepsilon}_2 $
is the particle relative permittivity,
$ {\varepsilon}_0 $
is the free-space permittivity,
$ L $
is the depolarization ratio, which for a prolate spheroid is:
$$ L=\frac{1-{\xi}^2}{2{\xi}^3}\left(\ln \frac{1+\xi }{1-\xi }-2\xi \right)\simeq \frac{1}{3}-\frac{2{\xi}^2}{15}-\frac{2{\xi}^4}{35}+O\left({\xi}^6\right) $$
where
$ \xi =\sqrt{1-\frac{b^2}{a^2}} $
and
$ a $
and
$ b $
are the major and minor axes.
From this equation, we can see that the elongated object of higher refractive index than the background (protein) has a higher polarizability since
$ L $
decreases as the particle becomes more prolate from Eq. 2, and this can be used to obtain information about the shape and dynamic changes of a protein. For the case of a sphere L = 1/3; therefore, for roughly spherical proteins it has been found that the polarizability scales with the mass of the protein (Becker et al., Reference Becker, Peters, Crooks, Helmi, Synakewicz, Schuler and Kukura2023):
$$ \alpha =\delta \alpha \cdot m $$
which is expected since the polarizability scales with the volume. Published values for this proportionality for proteins have varied as 239 Å3/kDa (Thiele et al., Reference Thiele, Pfitzner and Kukura2023), 460 Å3/kDa (Špačková et al., Reference Špačková, Moberg, Fritzsche, Tenghamn, Sjösten, Šıpová-Jungová, Albinsson, Lubart, van Leeuwen, Westerlund, Midtvedt, Esbjörner, Käll, Volpe and Langhammer2022), and 724 Å3/kDa (Becker et al., Reference Becker, Peters, Crooks, Helmi, Synakewicz, Schuler and Kukura2023). This variation can be accounted for somewhat by whether the background permittivity is included in the polarizability of Eq. 1, but also is impacted by being close to a high refractive index substrate (Bobbert and Vlieger, Reference Bobbert and Vlieger1987). Single- and double-stranded DNA have produced values that are 93% and 80% of these values (Li et al., Reference Li, Struwe and Kukura2020).
For 445 nm laser, the scattering cross section of an average sized protein is 0.5 × 10−11 μm2. So around 1 in 100 billion of the photons incident are scattered from a 0.5 micron squared spot (approximately the diffraction limit), and for 1 mW, this is 20,000 photons/second, with a comparable and typically larger count coming from surrounding regions. As a result, the signal is small, and homodyne (interference) detection schemes can be used to enhance the signal, as typified by interference scattering.
Interference scattering
iScat makes use of the interference between the scattering from a protein (or other object) at a surface and the surface itself (reference signal). This technique has been commercialized and is gaining widescale adoption as a protein characterization tool. The basic theory of how this operates is described below.
iScat signal
The total scattered electric field can be written as:
$$ {E}_t={E}_r+{E}_s $$
where the subscripts
$ r $
and
$ s $
are for reference (from the surface) and scattered photons (from the protein). The intensity detected is proportional to the magnitude squared of this field:
$$ {I}_t={I}_r+{I}_s+2\left|{E}_r{E}_s\right| cos\phi $$
where
$ \phi $
is the relative phase.
Since both the surface scattering and the protein scattering are proportional to the incident light, we can write this as
$ {E}_r=r{E}_i\hskip0.35em \mathrm{and}\hskip0.35em {E}_s=s{E}_i $
. Usually the scattering is small so that the second term can be neglected in Eq. 5. In this limit, the shot noise is proportional to the square root of the reference signal, so that the signal to noise ratio is given by:
$$ SNR=2\mid s\mid {E}_i $$
which is twice the square root of the number of scattered photons and is proportional to
$ \alpha $
, or the protein mass. The reference signal can come from reflection off the surface that the protein lands on. A factor of
$ 1/\surd 2 $
can be added to this relation to account for frame subtraction, which is used to detect changes when a protein lands on the surface (Becker et al., Reference Becker, Peters, Crooks, Helmi, Synakewicz, Schuler and Kukura2023). As compared with dark-field approaches, where the reference field is removed and the signal is ideally coming from the scattered particle only, the SNR also scales as the scattered field, and so these should have similar limits of detection; however, practical implementations, such as nanoparticle tracking analysis (Filipe et al., Reference Filipe, Hawe and Jiskoot2010), are limited to particles larger than 10 nm in size by undesired background scattering. Therefore, making use of the interference term has the primary practical benefit of increasing the signal relative to the unwanted background.
To detect the smallest objects, there are two considerations: maximize the signal to noise ratio by increasing the incident intensity so that as many scattered photons can be detected as possible, and do not saturate the detection or unduly increase background by reducing the reference intensity (Becker et al., Reference Becker, Peters, Crooks, Helmi, Synakewicz, Schuler and Kukura2023). To reduce the reference signal, various apertures and attenuators have been used (Liebel et al., Reference Liebel, Hugall and Van Hulst2017; Becker et al., Reference Becker, Peters, Crooks, Helmi, Synakewicz, Schuler and Kukura2023). Integrating over time also increases the detected scattered photon number, at the expense of the time resolution Liebel et al., Reference Liebel, Hugall and Van Hulst2017. In a well-engineered solution (Becker et al., Reference Becker, Peters, Crooks, Helmi, Synakewicz, Schuler and Kukura2023), the typical time resolution of iScat is of the order of 0.1 second, and the typical smallest proteins that can be detected are 40 kDa (Becker et al., Reference Becker, Peters, Crooks, Helmi, Synakewicz, Schuler and Kukura2023; Dahmardeh et al., Reference Dahmardeh, Dastjerdi, Mazal, Köstler and Sandoghdar2023), considering what is achievable experimentally. Proteins below 10 kDa are possible with machine learning, pushing the SNR close to 1 through anomaly detection (Becker et al., Reference Becker, Peters, Crooks, Helmi, Synakewicz, Schuler and Kukura2023; Dahmardeh et al., Reference Dahmardeh, Dastjerdi, Mazal, Köstler and Sandoghdar2023). This surpasses the conventionally defined limit of detection given by SNR = 3 (Becker et al., Reference Becker, Peters, Crooks, Helmi, Synakewicz, Schuler and Kukura2023). While the analysis has focused on the noise that is from the reference (mainly surface scattering), there is an additional time-varying noise from background speckle fluctuations and drift even in a pure water sample that limits the detection to around 5 kDa even with integration (Ortega Arroyo et al., Reference Ortega Arroyo, Andrecka, Spillane, Billington, Takagi, Sellers and Kukura2014; Liebel et al., Reference Liebel, Hugall and Van Hulst2017; Dastjerdi et al., Reference Dastjerdi, Dahmardeh, Gemeinhardt, Mahmoodabadi, Köstler and Sandoghdar2021; Becker et al., Reference Becker, Peters, Crooks, Helmi, Synakewicz, Schuler and Kukura2023). The signal is usually defined in terms of the contrast, which is
$ 2\mid s\mid /\mid r\mid $
. Variations in the local reflection from the surface change the signal, and this leads to reduced resolution that can be accounted for theoretically (Becker et al., Reference Becker, Peters, Crooks, Helmi, Synakewicz, Schuler and Kukura2023). iScat has been commercialized by Refeyn, formed in 2018.
Plasmonic scattering microscopy, evanescent scattering microscopy and surface plasmon resonance imaging
In PSM, Kretschmann (total internal reflection) excitation of surface plasmon waves is used to excite scattering from nanoscale objects on a metal film. This approach combines the usual surface plasmon resonance geometry with microscopy, which makes use of interference scattering between the surface waves and the scattering, both from surface roughness and objects bound or immobilized at the surface (Zhang et al., Reference Zhang, Ma, Dong, Wan, Wang and Tao2020). The surface plasmon waves are guided along the surface, and therefore not detected by the microscope above unless there is scattering out of the plane. Initial calibration of the approach showed a transition from sixth power size dependence expected from direct scattering to third power dependence expected from interference scattering as the particle size was reduced. The approach allows for discrete binding analysis of individual antibodies onto surfaces containing proteins, which makes use of established surface immobilization protocols for SPR (Zhang et al., Reference Zhang, Ma, Dong, Wan, Wang and Tao2020). While SPR is not a tether-free technique, PSM and SPRi make use of tethering to image free-solution biomolecule binding. The sensing characteristics reported were SNR of 11 for IgA around 400 kDa with a 50 ms integration time, which is comparable to iScat. There was some indication that using a metal surface would make the approach sensitive to charge (Foley et al., Reference Foley, Shan and Tao2008; Shan et al., Reference Shan, Patel, Wang, Iglesias and Tao2010; Liu et al., Reference Liu, Yang, Wang, Wang, Gao, Wu and Tao2017); however, the impact of calcium ions with calmodulin was barely detectable at the single molecule level (Zhang et al., Reference Zhang, Ma, Dong, Wan, Wang and Tao2020). Similar to plasmonic scattering, it is possible to use total internal reflection to scatter off of proteins at a surface without surface plasmons, and the interference with scattering off the rough surface can be used to detect the presence of individual proteins by image subtraction (Zhang et al., Reference Zhang, Zhou, Wang, Zhou, Jiang, Wan and Wang2022). The image is detected from above the surface, and the excitation field is incident from below.
SPRi uses the Kretschmann geometry, where both the exciting laser and imaged light are collected from below the sample. While SPRi predates PSM, the ability to resolve small particles with SPRi was not demonstrated until recently by integral scattering (Sun et al., Reference Sun, Wang, Zeng, Yang, Zhang, Li and Yu2023). In that work, interference between the scattered field and the plasmon was used to extract the location and size of the scattering object, making use of a well-defined scattering pattern and wavevector filtering. The detection of BSA was achieved with SNR = 3, making this approach comparable to iScat. Earlier than this, oscillating a protein with a PEG linker to a surface allowed for “locking in” to the scattering signal at the oscillation frequency and thereby removing background noise. This allowed for detecting proteins as small as 14 kDa (Ma et al., Reference Ma, ZijianWan, Zhang, Wang and Tao2020).
Nanofluidic scattering microscopy
NSM again makes use of the interference between a protein to be tracked and a reference signal, except that the reference signal comes from scattering off of a nanochannel in glass that contains the protein (Špačková et al., Reference Špačková, Moberg, Fritzsche, Tenghamn, Sjösten, Šıpová-Jungová, Albinsson, Lubart, van Leeuwen, Westerlund, Midtvedt, Esbjörner, Käll, Volpe and Langhammer2022). This makes use of dark-field excitation and allows for tracking diffusion of particles; that is, they do not have to land on the surface. Passivation of the surface with a supported lipid bilayer was used to allow for tracking positively charged nanoparticles; for negatively charged nanoparticles, interaction with the surface was rare.
Holography
While the common path configuration of iScat has inherent stability, it is possible to have the reference beam take a different pathway. This allows for holographic reconstruction that gives a greater tracking volume than nanoparticle tracking analysis (albeit for larger particles than proteins so far, and also at lower concentrations than typical for nanoparticle tracking analysis) (Ortiz-Orruño et al., Reference Ortiz-Orruño, Quidant, van Hulst, Liebel and Arroyo2022). Stable phase extraction has been used with a four-camera approach for holographic extraction of single proteins (Thiele et al., Reference Thiele, Pfitzner and Kukura2023). This first demonstration was still limited to fairly large proteins (∼90 kDa).
Nanoplasmonic scattering
Thermal effects
The ability to detect single protein binding events was demonstrated by backscattering from a single gold nanorod (Zijlstra et al., Reference Zijlstra, Paulo and Orrit2012). In that work, the backscatter of a 693 nm (off resonance) laser from the nanorod with binding in the presence of a 785 nm heating laser (100 ms integration) was observed, as well as taking a full spectrum (15 s integration time). Plasmonic heating changes the local refractive index (photothermal effect), and this gives a larger wavelength shift than binding alone, which enabled the sensitivity to detect single streptavidin (53 kDa) binding. This technique allows you to adjust the sensitivity by making modifications in the heating laser power.
Nanoaperture optical tweezers
Optical tweezers use the changes in momentum of photons (light) scattered off an object to manipulate that object. Since this depends on the polarizability, large laser powers are required to trap small objects like single proteins, and this makes the approach impractical. The diffraction limit also makes the trapping volume large. Therefore, nanoapertures in metal films have been used to enhance the trapping efficiency and provide a smaller trapping volume. The double-nanohole/bowtie/coaxial-shaped apertures in metal films have been used by several groups to trap and analyze single proteins (Pang and Gordon, Reference Pang and Gordon2012; Yoo et al., Reference Yoo, Gurunatha, Choi, Mohr, Ertsgaard, Gordon and Oh2018; Peri et al., Reference Peri, Sabnani, Raza, Ghaffari, Gimlin, Wawro, Lee, Kim, Weidanz and Alexandrakis2019; Verschueren et al., Reference Verschueren, Shi and Dekker2019; Ying et al., Reference Ying, Karakaci, Bermudez-Urena, Ianiro, Foster, Awasthi, Guha, Bryan, List, Balog, Acuna, Gordon and Mayer2021; Yang et al., Reference Yang, van Dijk, Primavera and Dekker2021). The continuous metal film surrounding the aperture helps to remove heat, and so typical temperature increases around 1 K/mW have been observed (Jiang et al., Reference Jiang, Rogez, Claude, Moreau, Lumeau, Baffou and Wenger2020; Verschueren et al., Reference Verschueren, Pud, Shi, Lorenzo De Angelis and Dekker2018; Xu et al., Reference Xu, Song and Crozier2018); this introduces a thermal gradient, which has a tendency to repel proteins (at room temperature or above) due to the typical positive Soret coefficient (thermophobic behavior). Nevertheless, the optical trapping potential can be large enough to allow trapping when a protein diffuses into the vicinity of the aperture under focused laser illumination.
The protein trapping is typically accompanied by a step in the transmitted power through the aperture. The transmitted power fluctuates due to thermal motion and conformational changes. The amplitude of the thermal fluctuations scales as the size of the particle being trapped Wheaton and Gordon, Reference Wheaton and Gordon2015. As with conventional optical tweezers, the power spectral density of these fluctuations has a corner frequency that scales as the trap stiffness divided by the drag. Since the trap stiffness is proportional to the polarizability, this also typically scales as the volume, and the Stokes drag scales as the radius, this gives a 2/3 power scaling of the corner frequency with mass, or − 2/3 if we consider a characteristic timescale Wheaton and Gordon, Reference Wheaton and Gordon2015. Plasmon-enhanced protein-tracking through interference (PEPTI) allows for tracking the protein prior to trapping by a nanoaperture due to scattering (Peters et al., Reference Peters, McIntosh, Albu, Ying and Gordon2023).
Microcavities and hybrid plasmonics
High-quality optical cavities have been investigated for label-free single protein detection by noting resonant shifts; however, initial reports were later revised to have not achieved the required sensitivity (Armani et al., Reference Armani, Kulkarni, Fraser, Flagan and Vahala2007; Lu et al., Reference Lu, Lee, Chen, Herchak, Kim, Fraser, Flagan and Vahala2011). A perturbation formulation can be used to estimate the wavelength shift of the resonance as well as the linewidth broadening (from the imaginary part) when a polarizable nanoparticle is introduced at position
$ \boldsymbol{ri} $
(Arnold et al., Reference Arnold, Khoshsima, Iwao Teraoka and Vollmer2003):
$$ \frac{\delta \omega}{\omega}\simeq \frac{-\alpha {\left|\boldsymbol{E}\left({\boldsymbol{r}}_{\boldsymbol{i}}\right)\right|}^2}{2\int \epsilon {\left|\boldsymbol{E}\left(\boldsymbol{r}\right)\right|}^2\; dV} $$
where
$ \boldsymbol{E}\left(\boldsymbol{r}\right) $
is the field of the unperturbed cavity at position
$ \boldsymbol{r} $
. Strictly speaking, this integral is only valid for closed lossless systems and diverges for open cavities; this issue can be resolved by using quasinormal mode theory with the appropriate unconjugated orthogonality relations (Kristensen and Hughes, Reference Kristensen and Hughes2014; Wu et al., Reference Wu, Gurioli and Lalanne2021; Franke et al., Reference Franke, Ren and Hughes2023).
Detection via frequency locking, or optomechanic effects, has improved the sensitivity to the single protein level (Su et al., Reference Su, Goldberg and Stoltz2016; Yu et al., Reference Yu, Jiang, Lin and Lu2016). By adding plasmonic nanoparticles or nanorods, it is possible to improve the sensitivity of these microcavities to the single DNA (∼2 kDa) (Baaske et al., Reference Baaske, Foreman and Vollmer2014; Liang et al., Reference Liang, Guo, Hou and Quan2017), protein (Dantham et al., Reference Dantham, Holler, Barbre, Keng, Kolchenko and Arnold2013), and even single ion levels (Kim et al., Reference Kim, Baaske, Schuldes, Wilsch and Vollmer2017). This arises because of the local field enhancement at the detection point with extreme subwavelength (plasmonic) localization. It is also possible, as noted with PSM, that charge is playing a role to shift the resonance through electrostatic interactions with the metal. A photonic-crystal plasmonic-particle cavity was used to establish the changes to the biophysical properties due to labelling at the single molecule level (Liang et al., Reference Liang, Guo, Hou and Quan2017). At higher powers, heating can take over and can have the opposite shift for hybrid platforms (Toropov et al., Reference Toropov, Houghton, Yu and Vollmer2023). Using a high-finesse fibre-based Fabry-Pérot microcavities and Pound-Drever-Hall cavity locking, detection of biomolecules down to a 1.2 kDa protein, Myc-tag, was achieved (Needham et al., Reference Needham, Saavedra, Rasch, Barber, Schweitzer, Fairhall, Vollbrecht, Wan, Podorova, Bergsten, Mehlenbacher, Zhang, Tenbrake, Saimi, Kneely, Kirkwood, Pfeifer, Chapman and Goldsmith2024). A 2D signal of temporal and intensity data allows this technique to distinguish between mixed protein samples and mixtures of DNA isomers of identical mass but different sequences. The detection relies on a refractive index change as the biomolecule displaces the water molecules of lower index, leading to resonance shifts of 1–49 kHz, 20 times greater than WGM resonator estimates. The high resolution is attributed to high passive stability, active low-frequency stabilization, creation of a velocity discrimination window, and the use of dynamic photothermal distortion of the resonance line shape.
Applications
The applications of light scattering single-molecule techniques to the label-free analysis of single proteins and other biomolecules are described in this section. Various biophysical parameters can be obtained through these techniques. Briefly, kinetic data such as protein oligomerization is obtainable via NOT, iScat, PEPTI, and holography; small molecule and antibody interactions with NOT, PC-Hybrid, SPRi, PSM, iScat, and WGM methods; and thermodynamic constants from NOT and PNP. When choosing which method to use in solving a specific question, one must consider not only the information desired but also the concentration range, temporal resolution, throughput, and accessibility of the technique. A technical comparison of the various approaches is given in Table 1.
Comparison of various label-free single biomolecule sensing techniques
References: 1. (Yu et al., Reference Yu, Jiang, Lin and Lu2016). 2. (Toropov et al., Reference Toropov, Houghton, Yu and Vollmer2023). 3. (Baaske et al., Reference Baaske, Foreman and Vollmer2014). 4. (Vincent et al., Reference Vincent, Subramanian and Vollmer2020). 5. (Kim et al., Reference Kim, Baaske and Vollmer2016). 6. (Yu et al. Reference Yu, Humar, Meserve, Bailey, Chormaic and Vollmer2021). 7. (Babaei et al., Reference Babaei, Wright and Gordon2023). 8. (Wheaton and Gordon, Reference Wheaton and Gordon2015). 9. (Ying et al., Reference Ying, Karakaci, Bermudez-Urena, Ianiro, Foster, Awasthi, Guha, Bryan, List, Balog, Acuna, Gordon and Mayer2021). 10. (Hacohen et al., Reference Hacohen, Ip and Gordon2018). 11. Yousefi et al., Reference Yousefi, Zheng, Zargarbashi, Assadipapari, Hickman, Parmenter, Bueno-Alejo, Sanderson, Craske, Xu, Perry, Rahmani and Ying2023. 12. (Zehtabi-Oskuie et al., Reference Zehtabi-Oskuie, Jiang, Cyr, Rennehan, Al-Balushi and Gordon2013). 13. (Kotnala and Gordon, Reference Kotnala and Gordon2014a). 14. (Al Balushi and Gordon, Reference Al Balushi and Gordon2014b). 15. (Al Balushi and Gordon, Reference Al Balushi and Gordon2014a). 16. (Yousefi et al., Reference Yousefi, Ying, Parmenter, Assadipapari, Sanderson, Zheng, Xu, Zargarbashi, Hickman, Cousins, Mellor, Mayer and Rahmani2023). 17. (Pang and Gordon, Reference Pang and Gordon2012). 18. (Wheaton et al., Reference Wheaton, Gelfand and Gordon2015). 19. (Dahmardeh et al., Reference Dahmardeh, Dastjerdi, Mazal, Köstler and Sandoghdar2023). 20. (Young et al., Reference Young, Hundt, Cole, Fineberg, Andrecka, Tyler, Olerinyova, Ansari, Marklund, Collier, Chandler, Tkachenko, Allen, Crispin, Billington, Takagi, Sellers, Eichmann, Selenko, Frey, Riek, Galpin, Struwe, Benesch and Kukura2018). 21. (Liebel et al., Reference Liebel, Hugall and Van Hulst2017). 22. (Heermann et al., Reference Heermann, Steiert, Ramm, Hundt and Schwille2021). 23. (Häußermann et al., Reference Häußermann, Young, Kukura and Dietz2019. 24. (Taylor et al., Reference Taylor, Mahmoodabadi, Rauschenberger, Giessl, Schambony and Sandoghdar2019). 25. (Fineberg et al., Reference Fineberg, Surrey and Kukura2020). 26. (Wu and Piszczek, Reference Wu and Piszczek2020). 27. (Soltermann et al., Reference Soltermann, Foley, Pagnoni, Galpin, Benesch, Kukura and Struwe2020). 28. (Sonn-Segev et al., Reference Sonn-Segev, Belacic, Bodrug, Young, VanderLinden, Schulman, Schimpf, Friedrich, Dip, Schwartz, Bauer, Peters, Struwe, Benesch, Brown, Haselbach and Kukura2020). 29. (Foley et al., Reference Foley, Kushwah, Young and Kukura2021). 30. (Küppers et al., Reference Küppers, Albrecht, Kashkanova, Lühr and Sandoghdar2023). 31. (Sun et al., Reference Sun, Wang, Zeng, Yang, Zhang, Li and Yu2023). 32. (Zhang et al., Reference Zhang, Ma, Dong, Wan, Wang and Tao2020). 33. (Wan et al., Reference Wan, Ma, Zhang and Wang2022). 34. (Zhang et al., Reference Zhang, Zhou, Wang, Zhou, Jiang, Wan and Wang2022). 35. (Baaske et al., Reference Baaske, Asgari, Punj and Orrit2022). 36. (Zhang et al., Reference Zhang, Zhou, Wang, Jiang, Wan and Wang2021). 37. (Ma et al., Reference Ma, ZijianWan, Zhang, Wang and Tao2020). 38. (Špačková et al., Reference Špačková, Moberg, Fritzsche, Tenghamn, Sjösten, Šıpová-Jungová, Albinsson, Lubart, van Leeuwen, Westerlund, Midtvedt, Esbjörner, Käll, Volpe and Langhammer2022). 39. (Peters et al., Reference Peters, McIntosh, Albu, Ying and Gordon2023). 40. (Liang et al., Reference Liang, Guo, Hou and Quan2017). 41. (Thiele et al., Reference Thiele, Pfitzner and Kukura2023). 42. (Alexandrakis et al., Reference Alexandrakis, Sasank Peri, Subnani, Raza, Ghaffari, Lee, Kim and Weidanz2020). 43. (Peri et al., Reference Peri, Sabnani, Raza, Urquhart, Ghaffari, Lee, Kim, Weidanz and Alexandrakis2020). 44. (Verschueren et al., Reference Verschueren, Shi and Dekker2019). 45. (Yang et al., Reference Yang, van Dijk, Primavera and Dekker2021). 46. (Peri et al., Reference Peri, Sabnani, Raza, Urquhart, Ghaffari, Lee, Kim, Weidanz and Alexandrakis2020). 47. (Huang et al., Reference Huang, Mousavi, Zhao, Hubarevich, Omeis, Giovannini, Schütte, Garoli and Angelis2019). 48. (Needham et al., Reference Needham, Saavedra, Rasch, Barber, Schweitzer, Fairhall, Vollbrecht, Wan, Podorova, Bergsten, Mehlenbacher, Zhang, Tenbrake, Saimi, Kneely, Kirkwood, Pfeifer, Chapman and Goldsmith2024).
* iSCAT is the only technique commercialized. Data were taken from technical notes from Refeyn Ltd.
+ In a physiological environment, proteins are typically present in the μM concentration range.
a Based on the authors’ preliminary data.
b The duration is dependent on the channel length and the diffusion of protein.
c Plasmonic nanopores.
Sizing
The relation between iScat contrast and mass was established by evaluating several proteins and other biomolecules landing on a glass surface (Young et al., Reference Young, Hundt, Cole, Fineberg, Andrecka, Tyler, Olerinyova, Ansari, Marklund, Collier, Chandler, Tkachenko, Allen, Crispin, Billington, Takagi, Sellers, Eichmann, Selenko, Frey, Riek, Galpin, Struwe, Benesch and Kukura2018). The mass derived from the contrast was accurate to within 5 kDa, and the precision for individual landing events was tens of kDa (Young et al., Reference Young, Hundt, Cole, Fineberg, Andrecka, Tyler, Olerinyova, Ansari, Marklund, Collier, Chandler, Tkachenko, Allen, Crispin, Billington, Takagi, Sellers, Eichmann, Selenko, Frey, Riek, Galpin, Struwe, Benesch and Kukura2018), which was later improved by accounting for local variations in the reflection of the glass interface to less than 10 kDa Becker et al., Reference Becker, Peters, Crooks, Helmi, Synakewicz, Schuler and Kukura2023. The smallest protein detected was 53 kDa in the initial work (Young et al., Reference Young, Hundt, Cole, Fineberg, Andrecka, Tyler, Olerinyova, Ansari, Marklund, Collier, Chandler, Tkachenko, Allen, Crispin, Billington, Takagi, Sellers, Eichmann, Selenko, Frey, Riek, Galpin, Struwe, Benesch and Kukura2018), and this has been improved to below 10 kDa with machine learning, albeit with an accuracy of 5 kDa and precision of 60% (Dahmardeh et al., Reference Dahmardeh, Dastjerdi, Mazal, Köstler and Sandoghdar2023). It was possible to detect a change in the mass of streptavidin with biotin binding (4 biotins per streptavidin), as well as modified biotin of different masses (Young et al., Reference Young, Hundt, Cole, Fineberg, Andrecka, Tyler, Olerinyova, Ansari, Marklund, Collier, Chandler, Tkachenko, Allen, Crispin, Billington, Takagi, Sellers, Eichmann, Selenko, Frey, Riek, Galpin, Struwe, Benesch and Kukura2018). The accuracy on these changes was at the kDa level and comparable to the change itself. Mass changes of lipid nanodisks and for proteins with glycosylation were also observed, as well as protein assembly by noting mass distributions with variations in protein concentration.
The scattered signal from single proteins in NOT experiments can be related to the size by the root mean squared deviation (RMSD) of the aperture transmission signal and the autocorrelation or power spectral density. The RMSD scales linearly with protein mass and gives the optical size similar to iScat (Wheaton and Gordon, Reference Wheaton and Gordon2015). The autocorrelation function measures the similarity between a signal and its time-delayed version, whereas the power spectral density function gives the distribution of average power in the frequency domain, and these are related through the Fourier transform. Equation 8 relates the corner frequency obtained from the power spectral density to the hydrodynamic radius to the power of −2/3 (Kotnala and Gordon, Reference Kotnala and Gordon2014b; Wheaton and Gordon, Reference Wheaton and Gordon2015; Babaei et al., Reference Babaei, Wright and Gordon2023). Interestingly, this expression contains both the optical size and the hydrodynamic size,
$ {r}_h $
through the Stokes drag.
$$ {f}_c=\frac{\kappa }{\gamma}\propto \frac{\alpha }{6\pi \eta {r}_h} $$
Sizing of single proteins with PSM revealed two size regimes: one for large nanoparticles that follows a d 5.6 size dependence and for smaller nanoparticles that follows d 3 size dependence. That method showed sizing down to single BSA proteins (66 kDa) (Ma et al., Reference Ma, ZijianWan, Zhang, Wang and Tao2020; Zhang et al., Reference Zhang, Ma, Dong, Wan, Wang and Tao2020; Wan et al., Reference Wan, Ma, Zhang and Wang2022; Zhang et al., Reference Zhang, Zhou, Wang, Zhou, Jiang, Wan and Wang2022). The molecular weight determination of single proteins using NSM was achieved through the integrated optical contrast being linearly dependent on the polarizability of the biomolecule, which is linearly dependent on the molecular weight. Similar to PSM, they detected a protein of 66 kDa while also being able to measure DNA and vesicles (Špačková et al., Reference Špačková, Moberg, Fritzsche, Tenghamn, Sjösten, Šıpová-Jungová, Albinsson, Lubart, van Leeuwen, Westerlund, Midtvedt, Esbjörner, Käll, Volpe and Langhammer2022). The size sensitivity demonstrated for various techniques is shown in Figure 2.
Molecular weight sensitivity of label-free single biomolecule technique advancements over time. Whispering gallery mode (WGM) (Yu et al., Reference Yu, Jiang, Lin and Lu2016), nanoaperture optical tweezer (NOT) (Pang and Gordon, Reference Pang and Gordon2012; Wheaton and Gordon, Reference Wheaton and Gordon2015; Babaei et al., Reference Babaei, Wright and Gordon2023), interferometric scattering (iScat) (Piliarik and Sandoghdar, Reference Piliarik and Sandoghdar2014; Liebel et al., Reference Liebel, Hugall and Van Hulst2017; Young et al., Reference Young, Hundt, Cole, Fineberg, Andrecka, Tyler, Olerinyova, Ansari, Marklund, Collier, Chandler, Tkachenko, Allen, Crispin, Billington, Takagi, Sellers, Eichmann, Selenko, Frey, Riek, Galpin, Struwe, Benesch and Kukura2018; Hajdusits et al., Reference Hajdusits, Suskiewicz, Hundt, Meinhart, Kurzbauer, Leodolter, Kukura and Clausen2021; Dahmardeh et al., Reference Dahmardeh, Dastjerdi, Mazal, Köstler and Sandoghdar2023), plasmonic scattering microscopy (PSM) (Zhang et al., Reference Zhang, Ma, Dong, Wan, Wang and Tao2020; Wan et al., Reference Wan, Ma, Zhang and Wang2022), surface plasmon resonance imaging (SPRi) (Ma et al., Reference Ma, ZijianWan, Zhang, Wang and Tao2020), nanofluidic scattering microscopy (NSM) (Špačková et al., Reference Špačková, Moberg, Fritzsche, Tenghamn, Sjösten, Šıpová-Jungová, Albinsson, Lubart, van Leeuwen, Westerlund, Midtvedt, Esbjörner, Käll, Volpe and Langhammer2022), plasmon enhanced protein tracking through interference (PEPTI) (Peters et al., Reference Peters, McIntosh, Albu, Ying and Gordon2023), photonic-plasmonic hybrid (PC-Hybrid) (Liang et al., Reference Liang, Guo, Hou and Quan2017), plasmonic-WGM hybrid (WGM-Hybrid) (Kim et al., Reference Kim, Baaske, Schuldes, Wilsch and Vollmer2017; Toropov et al., Reference Toropov, Houghton, Yu and Vollmer2023), WGM-Hybrid* measured nucleic acids (Baaske et al., Reference Baaske, Foreman and Vollmer2014), holography (Thiele et al., Reference Thiele, Pfitzner and Kukura2023). Corresponding histogram of human proteome size/frequency is shown to the right (Consortium, 2019), FP-MC (Needham et al., Reference Needham, Saavedra, Rasch, Barber, Schweitzer, Fairhall, Vollbrecht, Wan, Podorova, Bergsten, Mehlenbacher, Zhang, Tenbrake, Saimi, Kneely, Kirkwood, Pfeifer, Chapman and Goldsmith2024).
Oligomers and assembly
iScat was used to measure oligomer species formation of the MinDE system by using a supported lipid bilayer (Heermann et al., Reference Heermann, Steiert, Ramm, Hundt and Schwille2021). In the presence of ATP, MinD monomers were found to be the most prominent species in solution; however, on the lipid bilayer, MinD dimers were dominant. Although it should be noted that their imaging conditions do not allow for an accurate quantification of MinD monomers (33 kDa) on the lipid bilayer using iScat. At higher particle densities, they observed that MinD forms higher order-complexes on a crowded bilayer. Another study showed that the FWHM resolution of 20 kDa of iScat was able to resolve the oligomeric states of BSA, revealing the rare state of tetramers (0.25% abundance) (Hundt, Reference Hundt2021). Mass photometry has also been used to study the oligomeric equilibria of 2-cysteine peroxiredoxins in both humans and plants. Their results showed conserved features among both as well as species-specific features (Liebthal et al., Reference Liebthal, Kushwah, Kukura and Dietz2021). This technique has also been used to characterize immunoglobulin heavychain binding protein self-oligomerization and its dependence on temperature, showing the monomeric form is stabilized at higher temperature as well as ATP-induced monomer stabilization at low temperature (Rivera et al., Reference Rivera, Burgos-Bravo, Engelberger, Asor, Lagos-Espinoza, Figueroa, Kukura, Ramírez-Sarmiento, Baez, Smith and Wilson2023).
NOTs have been used to probe the disassembly of single ferritin proteins under different pHs (Yousefi et al., Reference Yousefi, Zheng, Zargarbashi, Assadipapari, Hickman, Parmenter, Bueno-Alejo, Sanderson, Craske, Xu, Perry, Rahmani and Ying2023). At pH 2, ferritin underwent a stepwise fragmentation with critical fragments occurring at dimer, tetramer, 12-mer, and 22-mer subunits.
Interactions
DNA-protein
Unzipping of 10 base pair DNA-hairpins was seen with tumor suppressor p53 protein-DNA complex, showing a longer DNA unzipping time than freely trapped DNA hairpins. A mutant p53-DNA complex was also trapped, showing that a single point mutation of Cys135Ser causes p53 to lose the ability to suppress DNA unzipping (Kotnala and Gordon, Reference Kotnala and Gordon2014a). Characterization of DNA binding with forkhead box protein P2 was performed using mass photometry and was compared with fluorescence proximity sensing, showing agreement with free energy measurements (Häußermann et al., Reference Häußermann, Young, Kukura and Dietz2019).
Protein–protein
Quantification of affinities of tubulin monomers and heterodimers in the < μM range was shown using iScat (Fineberg et al., Reference Fineberg, Surrey and Kukura2020). That showed an αβ-tubulin dissociation constant of 8.48 ± 1.22 nM and, in the presence of GTP, a value of 3.69 ± 0.65 nM. The same group also showed that mass photometry was able to accurately count and distinguish proteins by molecular weight, revealing heterogeneity and abundances at the single molecule level (Soltermann et al., Reference Soltermann, Foley, Pagnoni, Galpin, Benesch, Kukura and Struwe2020; Sonn-Segev et al., Reference Sonn-Segev, Belacic, Bodrug, Young, VanderLinden, Schulman, Schimpf, Friedrich, Dip, Schwartz, Bauer, Peters, Struwe, Benesch, Brown, Haselbach and Kukura2020). Small molecules and Ions Single molecule dynamics of protein-small molecule interactions have been studied using NOTs (Al Balushi et al., Reference Al Balushi, Zehtabi-Oskuie and Gordon2013; Al Balushi and Gordon, Reference Al Balushi and Gordon2014a, Reference Al Balushi and Gordon2014b; Yousefi et al., Reference Yousefi, Ying, Parmenter, Assadipapari, Sanderson, Zheng, Xu, Zargarbashi, Hickman, Cousins, Mellor, Mayer and Rahmani2023). Biotin-streptavidin, biotin-monovalent streptavidin, and acetylsalicylic acid-cyclooxygenase 2 were used to show that it is possible to distinguish between the bound and unbound state of the protein from the optical scattering (Al Balushi et al., Reference Al Balushi, Zehtabi-Oskuie and Gordon2013; Al Balushi and Gordon, Reference Al Balushi and Gordon2014b). Tolbutamide-human serum albumin and phenytoin-human serum albumin showed agreement in the dissociation constant reported in the literature by observing residence times represented by different amounts of transmission through the aperture (scattering) (Al Balushi and Gordon, Reference Al Balushi and Gordon2014a). In a work looking at ferritin, real-time dynamics of iron loading and biomineralization within a single unlabelled protein complex were shown (Yousefi et al., Reference Yousefi, Ying, Parmenter, Assadipapari, Sanderson, Zheng, Xu, Zargarbashi, Hickman, Cousins, Mellor, Mayer and Rahmani2023). Differences in structural rigidity of the apo- and holo-ferritin were shown. In-situ iron loading was observed and attributed to the folding of 8 gated pores, causing dynamic instability while iron was loaded into the core of the protein. SPRi was used to study Ca2+ ions binding to calmodulin, showing that calcium binding altered the conformation of calmodulin, increasing the hydrodynamic radius by 13% (Ma et al., Reference Ma, ZijianWan, Zhang, Wang and Tao2020).
Antibody detection
Using a commercial mass photometry system (Refyn), the detection of CD16 and IgG binding was shown for one-site binding and human α-thrombin (HT) and monoclonal anti-HT for two-site binding. Association constants were calculated and showed good agreement with existing methods (Wu and Piszczek, Reference Wu and Piszczek2020). Antibodies secreted from cells were monitored using iScat (McDonald et al., Reference McDonald, Gemeinhardt, König, Piliarik, Schaffer, Völkl, Aigner, Mackensen and Sandoghdar2018), finding the rate of secretion and the size range of secreted proteins and particles.
BSA and anti-BSA interactions were shown in a co-trapping experiment using NOTs (Zehtabi-Oskuie et al., Reference Zehtabi-Oskuie, Jiang, Cyr, Rennehan, Al-Balushi and Gordon2013). A variation of the NOT was used to discriminate the specific binding of anti-RAH to RAH antigen compared to non-specific binding of an anti-WNV antibody. That work integrated a nanopore with the double nanohole but showed that electrical measurements alone were not enough to discriminate between specific and non-specific binding, whereas it was possible with optical measurements. They were also able to quantify the dissociation constant to be 58 ± 17 nM (Peri et al., Reference Peri, Sabnani, Raza, Ghaffari, Gimlin, Wawro, Lee, Kim, Weidanz and Alexandrakis2019).
Conformational changes
The NOT trapping of BSA showed repeated steps that were attributed to conformational changes, as verified by reducing the pH and forcing the protein into the open state, where only a single step was seen at the higher transmission intensity (Pang and Gordon2012). Conformational changes were observed as steps in the transmitted intensity due to interactions for four different protein systems: haemoglobin interacting with single oxygen molecules, calmodulin under heating with increasing laser power (stabilized with calcium ions), adenylate kinase, and citrate synthase interacting with their substrates (Ying et al., Reference Ying, Karakaci, Bermudez-Urena, Ianiro, Foster, Awasthi, Guha, Bryan, List, Balog, Acuna, Gordon and Mayer2021). Variations in the extension of PR65 with various point mutations were measured with NOT and compared with theoretical predictions (Bahar et al., Reference Bahar, Banerjee, Mathew, Naqvi, Yilmaz, Zachoropoulou, Doruker, Kumita, Yang, Gur, Itzhaki and Gordon2023).
Diffusion and transport
The motion of myosin on actin filaments was tracked with iScat (Ortega Arroyo et al., Reference Ortega Arroyo, Andrecka, Spillane, Billington, Takagi, Sellers and Kukura2014). NSM and PEPTI tracked single, unlabelled protein diffusion. In NSM, the diffusivity was measured by the statistical analysis of the movement using a particle tracking algorithm. By approximating the protein as a hard neutral sphere, the hydrodynamic radius could also be inferred; however, the movement was inherently restricted by the nanochannel dimensions (Špačková et al., Reference Špačková, Moberg, Fritzsche, Tenghamn, Sjösten, Šıpová-Jungová, Albinsson, Lubart, van Leeuwen, Westerlund, Midtvedt, Esbjörner, Käll, Volpe and Langhammer2022). Similarly, PEPTI measured the diffusion of proteins using a particle tracking algorithm, but likely due to the restriction of the protein being near a surface and strong optical and thermal forces, the measured diffusivity was smaller than expected from unconstrained 3D diffusion (Peters et al., Reference Peters, McIntosh, Albu, Ying and Gordon2023).
Vibrational modes
Probing the acoustic vibrational modes in the range of 0.7–10 cm−1 for single proteins was achieved using NOTs in a technique named Extraordinary Acoustic Raman Spectroscopy (Wheaton et al., Reference Wheaton, Gelfand and Gordon2015). Two identical lasers were used to trap the protein; one of the lasers was then thermally tuned to induce a small change in laser wavelength, creating a beat frequency between the two lasers. When the beat frequency was resonant with the proteins acoustic resonance, the protein was excited to oscillate resonantly, which increased scattered light intensity fluctuations. This technique was applied to conalbumin (76 kDa), cyclooxygenase (69 kDa), streptavidin (60 kDa), carbonic anhydrase (29 kDa), and trypsin inhibitor (21.5 kDa), with the measured acoustic modes matching well with theory (DeWolf and Gordon, Reference DeWolf and Gordon2016). This approach was also used to measure the vibrational modes of single-strand DNA of different lengths and bases, showing good agreement of the observed resonances with a simple 1D finite chain model (Kotnala et al., Reference Kotnala, Wheaton and Gordon2015).
Conclusions and outlook
We have reviewed single-molecule techniques based on optical scattering, focusing on their wide applications in studying unmodified biomolecules. These techniques together can cover the size range of human proteins, with a detection limit down to 4 kDa (Babaei et al., Reference Babaei, Wright and Gordon2023). Currently, there are only a few approaches that can characterize individual, unmodified biomolecules in solution. Optical scattering is one primary approach. Other label-free techniques are mostly based on electrical measurements, including nanopores (Y.-L. Ying et al., Reference Ying, Hu, Zhang, YujiaQing, Maglia, Meller, Bayley, Dekker and Long2022; Wang et al., Reference Wang, Zhao, Bollas, Wang and Au2021) , nanochannels (Choi et al., Reference Choi, Jia, Riahipour, McKinney, Amarasekara, Weerakoon-Ratnayake, Soper and Park2021; Wang et al., Reference Wang, ZequnWang, Tong, Zhang, Fang, Xie, Liang, Yin, Yuan, Zhang and Wang2022), and nanowire field effect transistors (FETs) (Sorgenfrei et al., Reference Sorgenfrei, Chiu, Ruben, Yu, Kim, Nuckolls and Shepard2011; He et al., Reference He, Li, Ci, Qi and Guo2016; Liu et al., Reference Liu, Chen, Yin, Yang, Feng, Sun, Lai and Guo2023). While nanowire FETs typically require the analyte to be attached to the probe, they can detect single DNA signals without tethering when combined with nanopores (Xie et al., Reference Xie, Xiong, Fang, Qing and MLieber2012). Compared to light scattering approaches, nanopore sensing has been used in detecting a wider range of analytes, from subnanometers (metal ions (Roozbahani et al., Reference Roozbahani, Chen, Zhang, Wang and Guan2020) to hundreds of nanometers (virons (Taniguchi et al., Reference Taniguchi, Minami, Ono, Hamajima, Morimura, Hamaguchi, Akeda, Kanai, Kobayashi, Kamitani, Terada, Suzuki, Hatori, Yamagishi, Washizu, Takei, Sakamoto, Naono, Tatematsu, Washio, Matsuura and Tomono2021). Nanopore’s applications include DNA/RNA sequencing (Derrington et al., Reference Derrington, Butler, Collins, Manrao, Pavlenok, Niederweis and Gundlach2010; Deamer et al., Reference Deamer, Akeson and Branton2016), protein fingerprinting (Yusko et al., Reference Yusko, Bruhn, OliviaMEggenberger, Rollings, Walsh, Nandivada, Pindrus, Hall, Sept, Li, Kalonia and Mayer2017; Houghtaling et al., Reference Houghtaling, Ying, Eggenberger, Fennouri, Nandivada, Acharjee, Li, Hall and Mayer2019) and protein sequencing (Hu et al., Reference Hu, Huo, Ying and Long2021). Based on these applications, many companies have emerged, including Oxford Nanopore Technologies (Eisenstein, Reference Eisenstein2012), NabSys, and Figura Analytics and so forth. One challenge in nanopore sensing is that most proteins or DNA bases transit through the nanopore within several microseconds, too fast to be detected by the typical recording bandwidth (50 kHz (Houghtaling et al., Reference Houghtaling, Ying, Eggenberger, Fennouri, Nandivada, Acharjee, Li, Hall and Mayer2019), let alone to gather detailed information about the biomolecules. Recent efforts in electroosmotic trapping to slow down protein translocation have made it possible for nanopores to interrogate conformational changes of single proteins for longer durations (Galenkamp et al., Reference Galenkamp, Biesemans and Maglia2020; Huang et al., Reference Huang, Willems, Bartelds, Van Dorpe, Soskine and Maglia2020; Schmid et al., Reference Schmid, Stömmer, Dietz and Dekker2021). Despite significant advancements, nanopore sensing can only operate at high salt concentrations (typically 1 M) due to its resistance pulse sensing nature.
Unlike nanopore sensing, only one optical scattering-based method has been commercialized - iScat, which was commercialized as Mass Photometry by Refeyn Ltd in 2018. With a brief period of time, Mass Photometry has led to over 300 research publications involving protein sizing, protein–protein interaction, and more. While each technique has distinct advantages and limitations, due to the nature of the scattering signal, most of them overlap in limitations such as low temporal resolution, low throughput, and restricted operating concentration ranges below the physiological environment. Moreover, the specialized expertise and resources required for most techniques can slow down their widespread adoption in biological laboratories.
Recent advances in machine-learning based data analysis keep pushing the detection limit of the single protein scattering signal (Dahmardeh et al., Reference Dahmardeh, Dastjerdi, Mazal, Köstler and Sandoghdar2023). Machine learning also offers potential to streamline data analysis (Thomsen et al., Reference Thomsen, Sletfjerding, Jensen, Stella, Paul, Malle, Montoya, Petersen and Hatzakis2020), which can facilitate the commercialization of the techniques by making them more user-friendly. High-speed cameras will promote optical scattering techniques in studying single proteins. On one hand, they improve the time resolution of the techniques that offer high-throughput, like iScat, SM, SPRi, PEPTi, and holography, allowing them to capture the fast interactions between different proteins or conformational changes within a single protein. On the other hand, replacing high-bandwidth photodetectors with high-speed cameras in the single-channel techniques, such as NOT, Opto-PSM, and PNP, may greatly improve the throughput of those techniques.
Integrating different optical scattering-based techniques has the potential to address some limitations. For example, NOTs collect a scattering signal only through an nanoaperture and at the same time reduce the sensing volume by plasmonic resonance, making it possible to operate at a concentration comparable to a physiological condition (e.g. micromolar (Wheaton and Gordon, Reference Wheaton and Gordon2015). However, NOTs suffer from the low throughput. PEPTi (Peters et al., Reference Peters, McIntosh, Albu, Ying and Gordon2023), an approach combined NOT with iScat, shows promise in operating in high concentration while ensuring the throughput. Moreover, electrical signal-based approaches, such as the well-established nanopore technique, can offer complementary information to light scattering methods. Combining nanopores with NOTs has demonstrated potential for antibody detection (Peri et al., Reference Peri, Sabnani, Raza, Ghaffari, Gimlin, Wawro, Lee, Kim, Weidanz and Alexandrakis2019, Reference Peri, Sabnani, Raza, Urquhart, Ghaffari, Lee, Kim, Weidanz and Alexandrakis2020, DNA sequencing (Belkin et al., Reference Belkin, Chao, Jonsson, Dekker and Aksimentiev2015), single-molecule SERS (Huang et al., Reference Huang, Mousavi, Zhao, Hubarevich, Omeis, Giovannini, Schütte, Garoli and Angelis2019, and improving the capture rate of NOTs (Verschueren et al., Reference Verschueren, Shi and Dekker2019. The integration of iScat with nanochannels, forming NSM, shows promise in analysing complex biofluid samples (Špačková et al., Reference Špačková, Moberg, Fritzsche, Tenghamn, Sjösten, Šıpová-Jungová, Albinsson, Lubart, van Leeuwen, Westerlund, Midtvedt, Esbjörner, Käll, Volpe and Langhammer2022).
In the past decades, FRET (Schuler, Reference Schuler2013; Lerner et al., Reference Lerner, Cordes, Ingargiola, Alhadid, Chung, Michalet and Weiss2018) and single-molecule force spectroscopy (Schönfelder et al., Reference Schönfelder, Sancho and Perez-Jimenez2016; Li, Reference Li2023) have revolutionized the field of protein study and revealed features that are hidden in ensemble-level measurements (Michalet et al., Reference Michalet, Weiss and Jäger2006; Ziemba et al., Reference Ziemba, Li, Landgraf, Knight, Voth and Falke2014; Willkomm et al., Reference Willkomm, Jakob, Kramm, Graus, Neumeier, Meister and Grohmann2022). Label-free methods, particularly techniques based on optical scattering reviewed here, provide additional information beyond the labelled techniques. Given the groundbreaking research in complex protein samples (Heermann et al., Reference Heermann, Steiert, Ramm, Hundt and Schwille2021), long-term observation of some molecules by NOTs (Pang and Gordon, Reference Pang and Gordon2012; Yousefi et al., Reference Yousefi, Ying, Parmenter, Assadipapari, Sanderson, Zheng, Xu, Zargarbashi, Hickman, Cousins, Mellor, Mayer and Rahmani2023, plasmonic nanopores (Peri et al., Reference Peri, Sabnani, Raza, Urquhart, Ghaffari, Lee, Kim, Weidanz and Alexandrakis2020), and imaging macromolecules in live cells (Zhang et al., Reference Zhang, Zhou, Wang, Jiang, Wan and Wang2021; Ma et al., Reference Ma, Zhang, Zhou, Wan and Wang2022; Küppers et al., Reference Küppers, Albrecht, Kashkanova, Lühr and Sandoghdar2023), we believe that light-scattering based techniques will provide game-changing advances to life sciences.
Competing interest
The authors declare none.
Open access
