Nanoaperture fabrication via colloidal lithography for single molecule fluorescence imaging

In single molecule fluorescence studies, background emission from labeled substrates often restricts their concentrations to non-physiological nanomolar values. One approach to address this challenge is the use of zero-mode waveguides (ZMWs), nanoscale holes in a thin metal film that physically and optically confine the observation volume allowing much higher concentrations of fluorescent substrates. Standard fabrication of ZMWs utilizes slow and costly E-beam nano-lithography. Herein, ZMWs are made using a self-assembled mask of polystyrene microspheres, enabling fabrication of thousands of ZMWs in parallel without sophisticated equipment. Polystyrene 1 μm dia. microbeads self-assemble on a glass slide into a hexagonal array, forming a mask for the deposition of metallic posts in the inter-bead interstices. The width of those interstices (and subsequent posts) is adjusted within 100-300 nm by partially fusing the beads at the polystyrene glass transition temperature. The beads are dissolved in toluene, aluminum or gold cladding is deposited around the posts, and those are dissolved, leaving behind an array ZMWs. Parameter optimization and the performance of the ZMWs are presented. By using colloidal self-assembly, typical laboratories can make use of sub-wavelength ZMW technology avoiding the availability and expense of sophisticated clean-room environments and equipment.


Introduction
Single molecule fluorescence techniques are valuable tools in biophysical research. By avoiding the averaging inherent in bulk measurements, they can distinguish subpopulations of molecules, directly observe the trajectory and timing of enzymatic reaction steps without needing to synchronize a population, and can enable study of rare events and conformational fluctuations. [1] These techniques include superresolution microscopy to track single molecule motion, fluorescence resonance energy transfer (FRET) to detect nanometer-scale distance changes, and polarized total internal reflectance (polTIRF) microscopy that measures angular changes of a macromolecule. [2][3][4] All single-molecule fluorescence techniques require careful optimization of signal-to-noise ratio due to their inherently limited signal.
The two most widely adopted schemes for reducing fluorescence background from out of focus fluorescent probes are total internal reflection fluorescence (TIRF) microscopy and confocal microscopy.
Both techniques create small optically defined volumes of detection. In the case of TIRF incident light reaches the glass-water interface at an angle greater than the critical angle for total internal reflection and is reflected. This creates a near-field standing wave (the evanescent wave) on the aqueous side that decays exponentially with distance from the boundary. This localized oscillating electromagnetic field excites only the fluorescent molecules very close (<100 nm) to the glass-water interface, significantly reducing background fluorescence. In confocal microscopy, a confocal pinhole in a conjugate image plane is used in conjunction with a focused laser spot to define a femtoliter scale detection volume.
Despite the advantages offered by these arrangements, background fluorescence from labeled substrates in solution remains a technical challenge, requiring fluorescent substrates to be used at nano-/pico-molar concentrations. However, biological processes often occur at concentrations up to a million-fold greater (micro-/milli-molar). A potential solution is nanofabricated zero-mode waveguides (ZMWs). [5][6][7] arrays of holes (30-300 nm diameter) in a thin, opaque metal layer (typically aluminum, chromium, or gold) on a glass substrate. [7] These sub-wavelength apertures do not propagate optical modes in the visual spectrum, but illumination will create an evanescent wave, like in TIRF, that decays inside the well. [8,9] For fluorescence microscopy, ZMWs are most applicable when their size is sufficiently small that the excitation light decays within the waveguide (λ > λc). By constraining the observation volume physically (in the plane of the slide) and optically (along the optical axis), ZMWs give observation volumes of attoliter scale or smaller, allowing use of micro-/milli-molar fluorescent substrates without prohibitive background fluorescence. [7] At µM concentrations of fluorescent substrates, only one or a few probes are simultaneously within the ZMW, exchanging rapidly with the non-excited bulk phase.
ZMWs have been produced by directly patterning the metal layer using ion beam milling, [10,11] via electron beam lithography and subsequent dry-etching [5,12] , or metal lift-off. [13][14][15][16] Although the presence of thousands of wells on each substrate allow parallel collection of data during an experiment, each well of a ZMW array is either fabricated in series, a slow and expensive process, or in parallel with advanced deep-UV lithography. [17] In contrast to these top-down methods, nanosphere (or "colloidal") lithography (NSL) uses the self-assembly of an ordered crystal to create a mask for nanopatterning.
NSL was first introduced nearly 40 years ago, [18][19][20] and has been used to fabricate many different nanostructures, including dry-etched SiO2 nanoplates for polymer imprinting, [21] large scale arrays of silicon nanowires, [22] ordered arrays of gold nanoparticles for catalysis of ZnO nanowire formation, [23] and sensors based on surface enhanced Raman spectroscopy (SERS). [24] Complex geometries can be obtained with multiple overlapping layers of micron and nanoscale spheres, [25] or by tilting the masks relative to the metalizing vapor beam to change the projection geometry. [26] Nonetheless, NSL remains less widely used than electron-beam or photolithography.
To take advantage of NSL's simplicity, flexibility and low cost, a method of fabricating ZMWs through colloidal templating of polystyrene beads was developed, using the self-assembly of an ordered structure to pattern arrays of nanoscale wells. This technique can be used to fabricate ZMWs from many different metals, including the commonly-used aluminum [5,6,27] and gold. [28] Importantly, this technique requires few specialized fabrication tools, and should be accessible to many labs that perform single-molecule experiments. ZMWs were fabricated using colloidal lithography according to the schematic diagram in Figure 1.

Formation and Annealing of Polystyrene Bead Mask
Suspensions of 1 µm polystyrene beads in 1:400 (v/v) TritonX100:ethanol were pipetted onto the centers of the coverslips, which had been kept at 85-90% relative humidity (see Experimental Section Figure S1).
When spherical particles are partially immersed in a liquid on a horizontal substrate, the liquid meniscus between them is deformed, which causes attraction between adjacent spheres via surface tension. In addition, as solvent evaporates from those menisci, convective flux further drives the particles into an ordered phase. [29] Thus, the bead suspensions spread into approximately 2 cm circular puddles in which 2D hexagonal lattices (Figure 1b, 2a-d) self-assemble, with thin voids between domains having different lattice orientations (Figure 2c,d).

Figure 2.
Macroscopic images of 2D crystalline arrays of polystyrene beads displaying rainbow structural coloration (a,b) and SEM images of those hexagonal arrays (c,d). SEM images of beads that were not heated (e), or were heated at 107 °C for 5 s (f) or 20 s (g), with corresponding interstitial areas selected (cyan) and Feret diameter histograms for the measured areas. (h) Interstitial Feret diameter from SEM images as a function of melting time. Error bars represent standard deviation.
The large size of these grains (~5-15 µm across, Figure 2c,d) uniformly covering 2 cm areas may represent a useful feature not only for the fabrication of ZMWs, but possibly for the fabrication of materials for batteries and energy storage, [30] as well as nanowires for semiconductor devices. [31] The use of ethanol rather than water for the colloidal suspension differentiates this procedure from other published methods, [32] allowing the evaporation to occur in only 2-3 minutes rather than two hours. Relative to other methods, [32] this more robust deposition process only requires precise control of relative humidity, rather than of temperature, and surface tilt (as required by other methods).
When the lattice has properly assembled, the periodic surface of the bead array causes wavelengthdependent interference in reflected incident light and a brilliant, rainbow iridescence ( Figure 2b) similar to that of peacock feathers and butterfly wings. [33,34] This striking structural coloration allows an immediate, macroscopic assessment of 2D lattice formation.
Since the interstices between the beads ultimately determine the cross-sectional shape and size of the ZMWs, the wells were made narrower and more cylindrical by briefly heating the polystyrene beads close to their glass transition temperature (approximately 107 °C), [35] allowing them to fuse with one another at their contact points (Figure 1c). Similar results have previously been obtained using microwave pulses, [36,37] but the use of a standard laboratory hot plate simplifies the process. This procedure also partially decouples the size of the holes in the polystyrene mask from the size of the spheres used, which controls the ZMW spacing. Simply substituting smaller beads for the 1 µm ones would create smaller interstitial gaps, but would also reduce the spacing between the wells and prevent them from being completely resolved in optical images. The timing of this heating step needs careful adjustment. For a range of treatment times from zero to 30 seconds (at which duration the beads fuse together completely), there is a reproducible relationship between melting time and size. The resulting pores can be tailored to produce diameters in the range of 350-100 nm (Figure 2h), which is essential for constructing waveguides with cutoff wavelengths in the visible range appropriate for experiments with the most commonly used probes for single molecule fluorescence. In addition, a round cross-section is also important; in non-centrosymmetric waveguides, the transmission is polarization sensitive, which could compromise the attenuation of the ZMWs [38] or cause their fluorescence to be sensitive to the orientation of the macromolecules. [39]

Pore Creation
The subsequent text will primarily focus on the fabrication of aluminum ZMWs, which starts with the deposition of copper posts ( Figure 1). Where relevant, we also note the differences in the procedure used to fabricate gold ZMWs, which starts with the deposition of aluminum posts ( Figure S2).
After the hexagonal lattice of beads was formed and annealed, the resulting polystyrene masks were used during line-of-sight thermal or e-beam evaporative plating of 300 nm of copper or aluminum that reached the glass surface only through the interstices between the beads (Figure 1d, S2d). The mirrored top surfaces of the beads also enhanced the structural coloration, giving a vibrant rainbow appearance ( Figure   3a, S3a). with wider population distributions than at the post stage, since each successive processing step introduced some variability. In summary, the main population of wells were approximately 135 nm wide and 110 nm deep, similar to the dimensions of gold [28] and aluminum [27] ZMWs commonly made by other means.

ZMW Functionalization
After these aluminum or gold ZMW arrays were fabricated, their surfaces were passivated and functionalized for use in single-molecule fluorescence experiments using protocols reported earlier for devices made by e-beam lithography. The aluminum ZMW surfaces were passivated with poly(vinylphosphonic acid) (PVPA) to block non-specific binding, [5,6,27] while the gold ZMW surfaces were passivated with a self-assembled monolayer of methoxy-terminated, thiol-polyethylene glycol (PEG). [28] For both types, the glass surface at the bottom of the wells was functionalized with a chemically-orthogonal 16:1 mixture of methoxy-and biotin-terminated silane-PEG. [40] 3. Device Characterization

In Situ Transmission Measurements
The faint transmission of light through the ZMWs when back-illuminated with sufficient intensity, using the illumination tower of the inverted microscope ( Figure 4) allowed the optical properties of the guides to be characterized. Guides backlit with 500 nm light. Intensity data is collected from a set of pixels 2x2 in area centered on the guides (red boxes). The background level is collected from a box centered in the dark region between guides (green box). Unattenuated light intensity is collected from the closest large defect(yellow box). (c) Transmission data is fit to the model function T , grey ribbon of values corresponds to the uncertainty in the guide thickness z.
To assess the optical attenuation of the ZMWs, we measured the transmission of the guides at 50 nm wavelength intervals from 500-700 nm (see Experimental Section for details). As expected, transmission decreased as wavelength increased (Figure 4c). For the wavelength range measured, transmission through the wells was always <13% of the transmission through bare glass. This data was fitted using a simple model [41] for the transmission of a cylindrical waveguide (Figure 4c and Experimental Section) which yielded a cut-off wavelength λc = 441  8 nm. The consistency of the transmission data with this simple model provided an efficient quality control test.
Additionally, the unique pattern of bright defects caused by grain boundaries in the colloidal crystal mask was used to register the image coordinates of each ZMW (1500-2000 per field of view). These built in registration markers allowed stage drift to be compensated, and images to be registered across spectral channels.

ZMW Performance Characterization with Labeled DNA
Wells were loaded with a biotinylated double stranded (ds) oligoDNA with Cy3 and Cy5 on opposite strands in positions known to give a FRET efficiency of ~0.7 using TIRF microscopy. [42] Before adding the oligo, the surface was first treated with 20 µM unlabeled streptavidin, which links the biotin on the PEG to that on the oligo using two of its four binding sites (see Experimental Section). After incubating a 100 pM concentration of labeled molecules for 10 minutes, a substantial number of wells contained single fluorescent DNA duplexes, as judged by single step-wise photobleaching. For both aluminum ZMWs ( Figure S4a) and gold ZMWs ( Figure S4c), single molecule smFRET traces were detected from the biotin, Cy3, Cy5-dsDNA oligos in the wells under illumination with a 532 nm laser, that gave FRET efficiencies of ~0.7 as expected ( Figure S4b,d).
The purpose of ZMWs is to attenuate background fluorescence intensity in the presence of a high concentration of labeled substrate diffusing freely in solution. In a typical TIRF setup without ZMWs, single-molecule traces cannot be detected above background fluorescence at probe concentrations above ~50 nM. After immobilizing biotin, Cy3-dsDNA oligos in ZMW wells and illuminating them with a 532 nm laser, however, we were able to detect individual immobilized fluorophores even at bulk concentrations of non-biotinylated Cy3-dsDNA as high as 1 µM (Figure 5a-d). PRE-translocation complexes were formed with Phe-tRNA Phe (Cy5) in the ribosomal P-site, and Val-tRNA Val (Cy3) in the A-site, where they exhibited FRET from Cy3 to Cy5 (Figure 6a). The pretranslocation complexes were stalled in the absence of the translocase EF-G·GTP, a condition that is known to allow transitions between "classical" and "hybrid" tRNA positions having high and moderate FRET efficiency. [43,44] In the presence of 500 nM free Phe-tRNA Phe (Cy5)·EF-Tu·GTP ternary complexes, smFRET traces showed ribosomes transitioning between a low FRET efficiency (~0.4) hybrid conformation and a high FRET efficiency (~0.7) classical conformation (Figure 6b) as previously observed. [43,44] Thus, these ZMWs fabricated via colloidal lithography allow single molecule fluorescence study of macromolecular dynamics at free concentrations of labeled substrates that would not be possible to approach using TIRF microscopy.

Conclusion
In single molecule fluorescence studies, such as smFRET, background fluorescence from labeled substrates often requires their use at concentrations several orders of magnitude less than present in vivo.
That shortcoming can be addressed through the use of zero-mode waveguides that attenuate background fluorescence by restricting the observation volume. However, the need for expensive and specialized nanofabrication equipment to fabricate ZMWs has precluded their widespread adoption by the biophysical community. Nanosphere lithography allows thousands of ZMWs to be fabricated in parallel, with sizes that are tunable via controlled fusing of polystyrene beads at their glass transition temperature. These robust and inexpensive structures can be made with minimal equipment.

Experimental Section
Slide cleaning. Borosilicate coverslips (No 1.5, Fisherbrand) were cleaned by 10 min sonication in acetone at 40 °C and rinsed three times with distilled water (diH2O), followed by a repeat of those steps. Bead annealing. A flat, milled aluminum plate (13.9 x 13.9 x 0.9 cm), used to provide a uniform temperature surface was placed on top of a hot plate heated to 107 °C (the glass transition temperature of polystyrene), as verified by a thermocouple junction placed on the aluminum plate. When the temperature was stabilized, coverslips with 2D crystal lattices of beads were placed (one-at-a-time) on the surface for 5-25 s and then moved to another aluminum plate at room temperature for rapid cooling. Scanning probe microscopy. Surface profiling of the metal posts and ZMWs was conducted using a Bruker Bioscope Catalyst AFM in tapping mode, with MikroMasch Opus 160AC-NA standard tips or Nanosensors AR5-NCL high aspect ratio tips.
Transmission measurements. To obtain the normalized transmission data used in Figure S4, a series of band-pass filters with 10 nm bandwidth were used with the microscope's illumination tower. The Melles-Griot filters used were 500 nm (03FIV006), 550 nm (03FIV008), 600nm (03FIV018), 650 nm (03FIV002), and 700 nm (03FIV024). The normalized transmission was measured as seen in Figure S4 using: where , is the average intensity of a 2 pixel x 2 pixel square selected to contain the guide emission.
is the average intensity of the dark area between guides (see Figure 4). This accounts for the non-zero A/D converter output from the camera even in the dark. is the average intensity measured from a rectangle placed in the closest defect, a region without any waveguide attenuation.
Aluminum ZMW functionalization. The following protocol describes the use of orthogonal PVPA chemistry to passivate the Al cladding, and silane chemistry to functionalize the glass bottoms of the wells, as adapted from published protocols. [27] The fabricated ZMWs were first cleaned with a 5 min sonication in acetone, followed by 3 washes with DI water 5 min isopropanol sonication, followed by 3 more washes with DI water. The Al ZMWs were dried with nitrogen and oxygen plasma cleaned for 5 min. A 4.5% (mass/mass) aqueous solution of PVPA was heated to 90 °C. The Al ZMWs were incubated in this preheated solution for 10 min. The ZMWs were then washed with DI water and dried with nitrogen. The ZMWs were finally annealed on a hot plate at 80 °C for 10 min. The Al ZMWs were then treated with 100:5:1 methanol : glacial acetic acid : 3-aminopropyl-triethoxysilane overnight, followed by three diH2O rinses, 1 min sonication in EtOH, and were allowed the air dry. The ZMWs then were treated for 3 hrs with 250 mg/mL polyethylene glycol (PEG) in 100 mM NaHCO3. 95% of the PEG used was mPEGsuccinimidyl valerate (MW 2,000, Laysan), while 5% was biotin-PEG-succinimidyl carbonate (MW 2,000, Laysan). The unbound PEG was rinsed with diH2O, followed by drying with N2. Functionalized ZMWs were vacuum sealed and stored in the dark at -20 °C until use.
Gold ZMW functionalization. The following protocol describes the use of orthogonal thiol chemistry to passivate the gold cladding, and silane chemistry to functionalize the glass bottoms of the wells, as adapted from published protocols. [28] The fabricated ZMWs underwent a repeat of the cleaning process used for the initial coverslips (see "Slide Cleaning"), but with no diH2O rinse following sonication in ethanol.
Instead, they were treated with 100:5:1 methanol : glacial acetic acid : 3-aminopropyl-triethoxysilane overnight, followed by three diH2O rinses, 1 min sonication in EtOH, and were allowed to air dry. The samples then were treated for 3 hrs with 250 mg/mL polyethylene glycol (PEG) in 100 mM NaHCO3.  [43] was used. An enzymatic deoxygenation system of 0.3% (w/v) glucose, 300 μg/ml glucose oxidase (Sigma-Aldrich), 120 μg/ml catalase (Roche), and aged 1.5 mM 6-hydroxy-2,5,7,8-tetramethyl-chromane-2-carboxylic acid (Trolox, Sigma-Aldrich) was added to the final single-molecule imaging solutions to reduce fluorophore photobleaching and blinking. [43] DNA duplex formation. tRNAs and ternary complex formation. tRNAs were prepared using the reduction and charging protocols previously described [43], [45] starting with E. coli tRNA Val , and yeast tRNA Phe purchased from Chemical Block (Moscow). tRNA Phe and tRNA Val were conjugated to Cy5 and Cy3 respectively via a dihydroUhydrazine linkage as described [46] . Ternary complex was formed by incubating 4 µM EF-Tu, 2 µM dye- Fluorescence microscope. A custom-built objective-type total internal reflection fluorescence (TIRF) microscope was used with the incidence angle adjusted to be near TIRF conditions unless otherwise stated. Data processing. Fluorescence traces were extracted from recorded image stacks using using lab-written ImageJ software [43] , and analyzed using Matlab. FRET efficiency was calculated according to [47] : where and are the raw fluorescence intensities of the donor and acceptor above background, is the cross-talk of the donor emission into the acceptor recording channel, and accounts for the ratios of quantum yield and detection efficiency between the donor and the acceptor. ImageJ was also used to identify the location of defects in the ZMW array using images of the microscope fields taken with the white light lamp on the illumination tower. The locations of those defects were used to exclude fluorescence traces originating from the defects.