Convex lens-induced nanoscale templating SEE COMMENTARY
Daniel J. Berarda, François Michauda, Sara Mahshida,b, Mohammed Jalal Ahameda, Christopher M. J. McFaula, Jason S. Leitha, Pierre Bérubéb, Rob Sladekb, Walter Reisnera,1, and Sabrina R. Lesliea,1
aDepartment of Physics, McGill University, Montreal, QC, Canada H3A 2T8; and bDepartment of Human Genetics, McGill University, Montreal, Canada H3A 0G1
Edited by Robert H. Austin, Princeton University, Princeton, NJ, and approved July 15, 2014 (received for review December 5, 2013) We demonstrate a new platform, convex lens-induced nanoscale These challenges have limited the practical range of nanochannel templating (CLINT), for dynamic manipulation and trapping of sin- dimensions to 40–50 nm (1). Moreover, the high hydraulic re- gle DNA molecules. In the CLINT technique, the curved surface of sistance of nanoscale features (for a slit of depth h, the hydraulic a convex lens is used to deform a flexible coverslip above a sub- resistance scales as 1/h3, compared with 1/h for electrical re- strate containing embedded nanotopography, creating a nano- sistance) (5) requires that electrophoretic actuation be used to scale gap that can be adjusted during an experiment to confine drive DNA into sub-100-nm nanochannels. This in turn neces- molecules within the embedded nanostructures. Critically, CLINT sitates the use of special high-salt electrophoresis buffers [2–5× has the capability of transforming a macroscale flow cell into a Tris/borate/EDTA (TBE)], which reduces DNA extension (6) and nanofluidic device without the need for permanent direct bond- constrains the imaging buffer used. One answer to these challenges ing, thus simplifying sample loading, providing greater accessibil- is to develop specialized grayscale lithography approaches (7, 8) ity of the surface for functionalization, and enabling dynamic that can create gently funneling channel dimensions, reducing the manipulation of confinement during device operation. Moreover, free-energy barrier. Gray-scale approaches, although they are as DNA molecules present in the gap are driven into the embedded feasible technologically, are still highly challenging to implement topography from above, CLINT eliminates the need for the high and remain limited in the range of confinement that can be varied pressures or electric fields required to load DNA into direct-bonded continuously in both lateral and vertical dimensions. nanofluidic devices. To demonstrate the versatility of CLINT, we To overcome the challenges faced by classical nanofluidic confine DNA to nanogroove and nanopit structures, demonstrating technology, we have developed a new approach for introducing DNA nanochannel-based stretching, denaturation mapping, and tunable nanoscale confinement to trap and align DNA molecules partitioning/trapping of single molecules in multiple embedded cav- for optical analysis. Our confinement-based imaging technology ities. In particular, using ionic strengths that are in line with typical combines nanotemplated substrates with a single-molecule im- biological buffers, we have successfully extended DNA in sub–30- aging technique called convex lens-induced confinement (CLIC) nm nanochannels, achieving high stretching (90%) that is in good (9). Fig. 1 illustrates a flow cell implementation of CLIC mi- agreement with Odijk deflection theory, and we have mapped ge- croscopy, in which molecules are initially loaded into a planar nomic features using denaturation analysis. micron-scale chamber (10). To form the CLIC imaging chamber, the upper chamber surface is subsequently pressed into contact single-molecule manipulation | polymer confinement | genomic mapping | with the lower surface using the curved surface of a lens (Figs. 1B CLIC imaging | nanotechnology and 2). The final vertical confinement profile varies gradually away from the contact point, typically increasing by tens of anoconfinement-based manipulation is a powerful approach nanometers over a field of view of a hundred microns. When Nfor controlling the conformation of single DNA molecules CLIC is performed over a surface containing nanotemplated on chip. When single polymer chains are squeezed into envi- ronments confined at length scales below their diameter of gy- Significance ration in free solution, the polymer equilibrium conformation will be molded by the surrounding nanoscale geometry. Nano- Convex lens-induced nanoscale templating (CLINT) represents channel arrays can be used for massively parallel extension of a conceptual breakthrough in nanofluidic technology for sin- DNA across an optical field, serving as the basis for a high- gle-molecule manipulation. CLINT solves a key challenge faced throughput optical mapping of genomes (1, 2). More varied by the nanofluidics field by bridging the multiple-length scales manipulations can be performed based on the design of the sur- required to efficiently bring single-molecule analytes from the rounding nanotopology, such as using nanocavities embedded in a pipette tip to the nanofluidic channel. To do this, CLINT loads nanoslit to trap single DNA molecules (3). Nanoconfinement- single-molecule analytes into embedded nanofeatures via dy- based manipulation, compared with competing techniques for namic control of applied vertical confinement, which we have single-molecule manipulation such as tweezer technology and demonstrated by loading and extending DNA within nano- surface/hydrodynamic-based stretching, has three key advantages channels. CLINT offers unique advantages in single-molecule DNA (4): (i) It is highly parallel, providing the high throughput essential mapping by facilitating surface passivation, increasing loading for mapping gigabase-scale mammalian genomes (1); (ii)itcanbe efficiency, obviating the need for applied pressure or electric efficiently integrated with microfluidics to rapidly cycle molecules fields, and enhancing compatibility with physiological buffers through the channel arrays for upstream/downstream pre- and and long DNA molecules extracted from complex genomes. postprocessing of DNA; and (iii) it does not require applied flow PHYSICS or electric force to maintain the DNA extension. Author contributions: W.R. and S.R.L. designed research; D.J.B., C.M.J.M., J.S.L., and S.R.L. performed research; D.J.B., F.M., S.M., M.J.A., C.M.J.M., J.S.L., P.B., R.S., W.R., and S.R.L. Nanoconfinement-based approaches have, however, a key dif- contributed new reagents/analytic tools; D.J.B., F.M., C.M.J.M., and S.R.L. analyzed data; ficulty inherent to the use of nanoscale dimensions: the need to and D.J.B., F.M., R.S., W.R., and S.R.L. wrote the paper. bridge length scales differing by up to 5 orders of magnitude The authors declare no conflict of interest. – (submillimeter scale of a pipette tip to channels in the 10 100 nm This article is a PNAS Direct Submission. range) in the same fluidic device. This introduces two challenges in See Commentary on page 13249. device design and manufacture: (i) the need to drive single-mol- 1To whom correspondence may be addressed. Email: [email protected] or reisner@ ecule analytes across a very high free-energy barrier at the edge of physics.mcgill.ca. nanoconfined regions and (ii) inefficient fluid transport due to the This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. BIOPHYSICS AND
high hydraulic resistance of channels with nanoscale dimensions. 1073/pnas.1321089111/-/DCSupplemental. COMPUTATIONAL BIOLOGY
www.pnas.org/cgi/doi/10.1073/pnas.1321089111 PNAS | September 16, 2014 | vol. 111 | no. 37 | 13295–13300 Downloaded by guest on September 25, 2021 AB volume interactions. If the width D of a square cross-section nanochannel is larger than the 50 nm persistence length P of the DNA, the polymer is described by the de Gennes confinement regime and will coil up into multiple blobs along the nano- 10 m 50nm channel axis (4). When D < P, coiling within the channel is suppressed and the molecule undergoes periodic deflections D L g along the walls with no back-bending (Odijk regime) (12).
Fig. 1. Illustration of the DNA-loading procedure. (A) When the chamber height has microscale vertical dimensions, DNA molecules are unconfined A and take on coiled conformations. (B) When the push-lens is lowered, the imposed vertical nanoscale confinement causes DNA molecules to align in the Z-axis Piezo nanochannels, their energetically preferred state. Actuator Lens Rotation Mechanism structures such as nanocavities and nanogrooves, the vertical Z-axis Coarse Adjuster confinement imposed by CLIC drives the single-molecule ana- XY Translation “ lytes into the embedded topology. We call this approach convex Stage lens-induced nanoscale templating” (CLINT). Note that in CLINT, the molecules are loaded gently by imposing confine- ment from above, eliminating the need for high pressures or electric fields to introduce single-molecule analytes into the Push-Lens confined region of the device. This ease of loading ensures that Flow Cell the CLINT approach, when used to confine macromolecules in Microscope structures with dimensions smaller than their persistence length, Objective is compatible with a much wider range of ionic conditions than BC classic direct-bonded devices that rely on electrophoresis for loading. Loading molecules from above can also reduce the 50 3 mm 40 sensitivity of the technique to fabrication defects that can lead to 30 clogging of direct-bonded channels loaded from the side. We 20 demonstrate that CLINT can efficiently load DNA into nano- h (nm) 10 channels less than 30 nm in size, imposing subpersistence length 0 -100 -50 0 50 100 confinement at which stretching is high, polymer back-folding is Displacement energetically unfeasible, and thermal fluctuations are suppressed (µm) (i.e., the Odijk regime). The ability to operate devices in this DE regime is critical as the suppression of back-bending and thermal 240 µm 240 µm fluctuations leads to more efficient alignment of optically map- ped DNA fragments to a reference genome (11). Finally, unlike enclosed classical nanofluidic devices, the CLINT device interior can be easily exposed, leading to greater reusability and ease of access. This feature can be used, for example, to load single cells at precise locations on the chip adjacent to nanofluidic features and to apply surface coatings to suppress nonspecific interactions of fluorescent probes and proteins with exposed device surfaces. In CLINT imaging, DNA molecules are loaded into a chamber formed by two transparent surfaces, such as a fused silica coverslip and substrate, separated by a spacer (typically 5- to 30-μm–thick F G adhesive tape). Molecules loaded into the chamber are initially unconfined (Fig. 1A). A convex lens mounted on a nanopositioner 5 µm 5 µm presses down on the upper coverslip, deforming it (Figs. 1B and 2B). The upper surface bows downward until it comes into contact with the lower planar coverslip at a single point, forming the CLIC 500 nm 500 nm chamber geometry (Fig. 2C). The applied vertical confinement causes DNA molecules to favor extended conformations due to self-exclusion interactions in the region surrounding the contact point, where the chamber height is less than the diameter of gy- Fig. 2. Schematic of the CLINT imaging platform. (A) Schematic of the CLIC ration Dg of DNA in free solution (4). The lower surface of the device. (B) Close-up showing flow cell deformation by the push-lens, with imaging chamber contains embedded nano-lithography (e.g., aqueous sample inserted before deformation. Adhesive tape thickness ex- nanochannels) subjecting the DNA to additional transverse con- aggerated to 100 μm (in reality ∼30 μmor10μm). (C) Fourth-order poly- finement. When the chamber height is small enough, the mole- nomial fit to the chamber height profile (10). (D) An array of high-resolution cules’ free energy is minimized when they are maximally extended images (80 μm) is acquired to span a large region (720 μm). Interferometry in the channels (Fig. 1B), which we observe to occur spontane- pattern corresponds to Newton’s rings. The contours of the chamber height ously. This technique can be applied to other nanotopologies (e.g., fit (green) are overlaid on the intensity minima. (E) Fluorescence image of a dye solution in the same CLINT chamber. Because fluorescence intensity is nanopits) and other nanomaterials (e.g., actin, microtubules, DNA proportional to the chamber height, the microchannels are bright. The nanotubes, nanowires). nanochannels, which run from one microchannel to the other, are not visi- When DNA is confined to a nanochannel of dimensions less ble. SEM images of nanopits (F) and nanochannels (G)(Insets are close-ups) than Dg, it will stretch out along the channel axis due to excluded- embedded in the lower coverslip.
13296 | www.pnas.org/cgi/doi/10.1073/pnas.1321089111 Berard et al. Downloaded by guest on September 25, 2021 A Experimental Methods We implement CLINT imaging using the microscopy device shown in Fig. 2A SEE COMMENTARY (10), mounted on an inverted fluorescence microscope (SI Text, Imaging Conditions). The CLIC imaging chamber (Fig. 2B) is formed between two confining surfaces. The lower surface is a fused silica substrate containing embedded nanotopography. The upper surface is a coverslip with two small 10 holes sand-blasted into the corners for fluid insertion and recovery. Double- sided adhesive tape (30- or 10-μm thick; Nitto Denko) separates the upper Lowering Push-Lens and lower chamber surfaces, and is laser-cut (PBS Engraving) to create channels for liquid to flow into a main central chamber (SI Text, Substrate Preparation). Computer-controlled syringe pumps insert and retrieve the B sample from the imaging chamber, facilitating serial measurements and sample recovery. Chemically inert fluorinated ethylene propylene tubing connects the pump outlets to the holes in the upper coverslip and a seal is formed between them by a thick silicone gasket. To form the CLIC imaging chamber, a lens mounted on a nanopositioner 10 presses down on the upper coverslip, causing it to deform around its curved surface (Fig. 2B)(SI Text, Experimental Procedure). The chamber height increases very gradually away from the contact point, e.g., by 15 nm in μ μ – Fig. 3. Sequence of frames for a λ-DNA molecule extending in a nano- height over a 60- m distance from the contact point (using 30- m thick channel while the push-lens is lowered. (A) Extension in a 27-nm channel. adhesive tape, as in Fig. 2C). DNA extension experiments were performed μ – The time between frames is 364 ms. (B) Extension in a 50-nm channel. The using 30- m thick adhesive, and denaturation mapping experiments with μ – time between frames is 910 ms. See Movie S1 of DNA loading and SI Text for 10- m thick adhesive. a description of the experimental loading procedure. In studying DNA confinement, it is necessary to measure the chamber height profile precisely (SI Text, Chamber Height Characterization and Figs. S1 and S2). The chamber height profile is measured using both interferometry and fluo- Previous methods that were used to confine DNA in channels in rescence (10). To perform interferometry, we removed the emission filter from the Odijk regime, particularly in channels with D < 30 nm, have the imaging system and directly imaged the excitation field. When the chamber formed, we observed Newton’srings(Fig.2D), which result from the required the use of exceptionally high pressures or strong electric interference of reflections of the laser light from the upper and lower con- fields (13, 14), resulting in rapid movement of the DNA through fining surfaces. We image this interference profile over a wide field by taking the channel and low yield during loading. We demonstrate that a series of high-resolution images while moving the sample in a horizontal we can overcome these challenges using CLINT and conse- plane relative to the microscope objective. The intensity of the interference quently facilitate parallel genomic analysis, rendering it a viable pattern for wavelength λ is minimum at chamber heights corresponding to integer multiples of λ/(2n cos θ), where n is the index of refraction of the biomedical technology. ≈ θ We expect a DNA chain to enter the Odijk confinement re- medium (in this case water, for which n 1.33) and is the angle of incidence < of the laser illumination (45°). Next, we measure the fluorescence intensity of gime when it is confined in channels with D 50 nm. In the an in situ dye solution (Alexa Fluor 647, Fig. 2E), which emits into a spectrally Odijk regime, the extension R of a DNA molecule of total separate channel from the DNA. The dye fluorescence intensity is proportional contour length L and persistence length P confined to a recti- to the local chamber height. By fitting the dye intensity profile, and using this function to fit the interference pattern subsequently, we determine the linear nanochannel of horizontal dimension Dx and vertical di- chamber height profile using the method described in ref. 10. mension Dy is given by (12, 15) The lower surface of the imaging chamber contains nano- and microscale " !# lithography. Substrates containing arrays of 27 × 27 × 200,000 nm3,50× 50 × 2 2 3 + 3 200,000 nm3 nanochannels, and 50 × 600 × 600 nm3 and 50 × 900 × 900 nm3 R = L 1 − A Dx Dy : [1] 2 nanopits were fabricated. These features were patterned in fused silica using P3 electron-beam lithography (VB6 UHR EWF; Vistec Lithography) and etched using reactive ion etching (RIE). One-micron-deep microchannels, which can be Here, parameter A is a constant that depends on the details of = ± easily seen in Fig. 2 D and E, were defined using contact UV photolithography the channel cross-section. Yang et al. find that A 0.09137 (EVG620; EVG) and etched using RIE. The microchannels were designed to 0.00007 for a square nanochannel (15). Note that in our enable buffer exchanges when the imaging chamber is compressed, but experiments the effective nanochannel vertical dimension is were not used in these experiments. Shown in Fig. 2 F and G are SEM images given by Dy = d + h, where d is the groove depth and h is the of the 900-nm nanopits and 50-nm nanochannels, respectively. The spacing CLINT adjustable gap height. Although the deflection theory betweennanopitsis4μm and the spacing between nanochannels is 2 μm. holds only for subpersistence length channels, it is possible to These nanoscale features were aligned with the push-lens using an XY micropositioning stage before an experiment and before the chamber develop a model valid up to around 2P (100 nm) by including was formed. departures from the Odijk limit due to back-bending within the The chamber was first wet using a solution of 45 mM Tris·base, 45 mM molecule and at the molecule ends (16). Back-bending leads to boric acid and 1 mM EDTA (0.5× TBE), with 3% (vol/vol) β-mercaptoethanol the formation of S loops within the chain and C loops at chain (BME) added as an antiphotobleaching agent. λ-Phage DNA (48.5 kbp; New ends that can be included via an Ising-type transfer matrix for- England Biosciences) was used at a concentration of 50 μg/mL for nano- channel experiments and T4-phage DNA (166 kbp; New England Biosciences) malism. The loops have a nucleation free energy determined by PHYSICS was used at a concentration of 5 μg/mL for nanopit experiments. Experi- bending and an additional length-dependent cost due to the ments were carried out in 0.5× TBE, with 577 pM Alexa Fluor 647 (to mea- enhanced excluded-volume interactions along the extent of the sure the chamber height profile) and an antiphotobleaching system loop. Monte Carlo simulations can be used to fit key scaling consisting of 3% BME, 5 mM protocatechuic acid, and 500 nM proto- prefactors in the model. Although this model was originally de- catechuate dioxygenase. The DNA was stained with YOYO-1 fluorescent dye veloped for a channel of circular cross-section, we have extended (Life Technologies) at a ratio of one dye molecule per 10 bp. YOYO-1 μ ± the approach to our rectilinear channels with nonunity aspect increases the DNA contour length at this staining ratio from 16.5 m to 19.0 0.7 μmforλ-DNA and from 56.4 μmto65± 2 μm for T4 DNA (SI Text, Sample ratio by assuming that bending will take place primarily within Preparation) (17). Denaturation mapping experiments were performed using
> BIOPHYSICS AND the larger height dimension Dy Dx where the free-energy cost is λ-DNA at a concentration of 25 μg/mL, 0.25× TBE buffer, 3% BME, and 50%
lower (e.g., so the bending penalty will scale as kBTP/Dy). formamide (2). COMPUTATIONAL BIOLOGY
Berard et al. PNAS | September 16, 2014 | vol. 111 | no. 37 | 13297 Downloaded by guest on September 25, 2021 Results and Discussion A 160 DNA Extension in Nanochannels. We demonstrate CLINT micros- copy by dynamically loading λ-DNA into square nanochannels of 80 D = 27 nm (Fig. 3A) and 50 nm (Fig. 3B). As schematically portrayed in Fig. 1, DNA molecules evolve from bundled to h (nm) 0 extended conformations as the push-lens is lowered into contact with the lower surface of the chamber. The DNA in Fig. 3B is partially confined to two separate channels in the first and sec- 0145 90 135 80 ond frames because the spacing between channels is only 2 μm. displacement (µm) When the push-lens is lowered further, the molecule becomes completely confined to a single nanochannel (Movie S1). Eq. 1 B 3µm 3µm 3µm predicts maximum extended lengths for λ-DNA of 16.8 ± 0.6 μm in a 27-nm channel and 15.7 ± 0.6 μm in a 50-nm channel, consistent with the measured extensions. Here we have assumed a persistence length lp ≈ 51 ± 3 nm (17).
Quantitative Analysis of Extended DNA. We characterize the phys- ical extension of DNA molecules in the CLINT device as a function of applied vertical confinement. This quantitative characterization is essential to understand the properties of DNA for optical genomic analysis and to optimize the design of our experimental setup and assay conditions. Movies of DNA molecules were acquired at several locations in the CLIC im- 30 aging chamber at a frame rate of 11 Hz, for a total of 100 frames. 15 The intensity profile of a DNA molecule extended in a nano- frames 0 channel is expected to be a step function convolved with a 16 17.2 18 7 8.5 10 3 5.0 7 Gaussian point-spread function. This convolution yields a model for the intensity profile I(x, t) of a linearly extended polymer end-to-end length (µm) along the channel that is described by the difference of two error C 1.0 functions (4). To extract the extension of individual molecules 0.8 stretched in the channels, the error function model for the in- tensity profile was fit to each frame, using a least-squares fitting 0.6 algorithm. See SI Text, DNA Length Determination and Fig. S3 for 0.5 greater detail. 0.4 We visualize DNA extensions and fluctuations in the channels Odijk theory as a function of chamber height (see Fig. 4B for representative 0.3 Odijk theory+S-,C-loops molecules in 27-nm channels at a series of heights). The per-
molecule time variation of DNA length is shown in the histograms Fractional extension of Fig. 4B: These are influenced by the chamber temperature and 0.2 could be minimized by cooling the sample. 4 8 12 20 30 40 60 80 120 200 Fig. 4C plots DNA fractional extension as a function of con- D h (nm) finement. Molecules which were obviously photonicked or stuck 0.2 0.2 0.4 to one of the confining surfaces were not included in the analysis. The DNA are, on average, slightly more extended in the 27-nm 0.1 0.1 0.2 channels than in the 50-nm channels at any given chamber height, with the longest molecules located at chamber heights less than 50 nm and having time-averaged lengths of 17.18 ± 0.05 0 0 0 μ ± μ 0 15 30 0 10 20 0 8 16
m and 17.37 0.03 m, respectively. The majority of the DNA densityProbability fragments were trapped at a chamber height less than ∼100 nm. end-to-end length (µm) Fig. 4C compares our measurements directly to Odijk theory and
13298 | www.pnas.org/cgi/doi/10.1073/pnas.1321089111 Berard et al. Downloaded by guest on September 25, 2021 A Lowering Push-Lens a genomic barcode pattern by imaging the DNA (2). The DNA must be heated to a temperature at which the AT-rich regions SEE COMMENTARY melt but the GC-rich regions do not. To perform denaturation 600 nm mapping, the temperature of the flow cell was raised to 42 °C using a heater located above the push-lens, which corresponded to an approximate sample temperature of 38 °C (SI Text, De- naturation Mapping and Fig. S6). The push-lens was then lowered to confine DNA to the nanochannels. After ∼10 s, a barcode- melting pattern consisting of a single dark spot near the center of 900 nm the molecule became clearly visible. After t ≈ 40 s, a second dark spot appeared as more YOYO-1 was released from a second denatured area (Fig. 6A). Fig. 6 B and C shows a partially melted molecule that has reached a steady state ∼1 min after lowering the push-lens. Fig. 6D shows the theoretical denaturation barcode for comparison, assuming a helicity of 0.7 (2), confirming good agreement with the experimental data for a single DNA molecule. B 240 Conclusion 180 Classical nanofluidic technology faces several simultaneous challenges: (i) efficient introduction of analytes from macroscale
h (nm) 120 to nanoscale dimensions, (ii) ensuring appropriate surface pas- 60 sivation in nanoscale environments dominated by surface inter- 0 actions, and (iii) extreme sensitivity to fabrication defects in 0 40 80 120 160 200 sub-50-nm channels. The CLINT approach, demonstrated here, displacement (µm) addresses these challenges by introducing dynamically control- lable confinement. This allows fluidics devices to be transformed h = 51 nm h = 104 nm h = 214 nm in situ from initially macroscale flow cells that enable easy in- troduction of analytes and passivation chemistries into nanoscale imaging devices, enabling direct single-molecule manipulation and analysis. We demonstrate that DNA can be dynamically loaded from the top and fully extended in sub-30-nm-size channels, as opposed to classic nanochannel approaches which require loading h = 47 nm h = 82 nm h = 204 nm from the side. Top loading, versus side loading, enables in- troduction of DNA into nanoconfined dimensions with (i)greater efficiency, (ii) control over the rate of analyte introduction by altering the rate of descent of the push-lens (9), and (iii)reduced sensitivity to fabrication defects: One defect can render a side- 10µm 10µm 10µm
Fig. 5. CLINT demonstration using nanopits. (A) Loading T4 DNA into nanopits (600- and 900-nm width, 50-nm depth, 4-μm spacing). Time between A frames is 1.1 s. See Movies S3 and S4 for single-molecule loading movies. (B) 10 Representative images of T4 DNA at equilibrium, confined to 600-nm pits at different heights within the imaging chamber.
and lowered into contact with the bottom substrate, DNA molecules were driven into the nanopits from above. Due to excluded-volume interactions, DNA molecules adopt conformations with coiled seg- 10s ments stored in multiple pits. The number of pits occupied by each molecule varies with the geometry of the nanopits and the imposed vertical confinement. BCD Fig. 5A shows T4 DNA being loaded into 600- and 900-nm nanopits. Initially, the molecules are unconfined; as the con- 1600 finement is dynamically imposed by lowering the push-lens, the 1500 DNA molecules occupy an increasing number of nanopits. 1400 Molecules confined to 600-nm pits compared with 900-nm pits 1300
occupy a higher number of pits because the 900-nm pits can store PHYSICS
more of a polymer’s contour (3). Representative images of DNA Intensity (Counts) 1200 confined to pits at different chamber heights, at equilibrium, are 10s 20 40 60 80 shown in Fig. 5B. At smaller chamber heights, each DNA mol- Position (pixels) ecule extends over a larger number of pits. Fig. 6. Denaturation mapping of λ-DNA in 50-nm channels. (A)Kymograph DNA Denaturation Mapping. To demonstrate genomic mapping showing initial loading of DNA into a nanochannels and release of YOYO-1 λ dye to reveal the characteristic λ-DNA melting profile. (B) Kymograph taken ∼1 using CLINT we performed denaturation mapping of -DNA min after loading a molecule into a nanochannel. (C)DNAimageproducedby confined to 50 nm channels. AT-rich regions in the DNA melt at averaging all 497 frames from the kymograph in B.(D) Intensity profile along a lower temperature than GC-rich regions. The YOYO-1 dye the central axis of the molecule shown in C (blue) and theoretical melting BIOPHYSICS AND
unbinds from the melted regions making it possible to obtain profile (red), assuming a helicity of 0.7. COMPUTATIONAL BIOLOGY
Berard et al. PNAS | September 16, 2014 | vol. 111 | no. 37 | 13299 Downloaded by guest on September 25, 2021 loaded channel inoperable, whereas top loading can still use the reagent exchange so that continuous reactions can be performed portion of the channel away from the defect. Reducing sensitivity within the well (e.g., local sequencing). Moreover, CLINT operates to defects is critical to enable the cost-effective scaling of nano- with an all-glass structure, so that the fluorescence backgrounds are fluidic technology for easy dissemination and extension to 10-nm reduced, and the surfaces are both nanometer-scale smooth and dimensions. Previous nanofluidic devices have been based on the nonporous to oxygen, which facilitates stable imaging. Lastly, CLINT need to directly bond nanostructures via high-temperature steps, enables working with a much wider range of buffer conditions not a need that CLINT obviates. Direct bonding is static and com- limited by the need to prevent surface interactions and create strong plicates surface passivation approaches critical for working with electrokinetic driving forces. Thus, buffer conditions can be tuned to single-molecule analytes (which tend to stick nonspecifically to optimize imaging and binding conditions or match exact physiolog- surfaces). Moreover, in direct-bonded nanofluidic devices, mole- ical conditions. These advantages suggest that CLINT-based nano- cules are introduced into nanochannels from adjoining micro- fluidics would be an ideal platform for high-throughput mapping of channel reservoirs, requiring high electric or hydrodynamic forces DNA–protein interactions on extended genomes. to overcome the large free-energy barriers introduced at the abrupt change in device dimensions. ACKNOWLEDGMENTS. We thank Andrew Martin, Andrew Caleb Guthrie, One powerful advantage of CLINT, demonstrated by capturing and Alexander Verge for assistance in developing instrumentation control of DNA in nanocavity structures, is the ability to dynamically trap software and Bojing Jia for assistance in sand blasting coverslip holes. We single analytes in arrays of local nanowells that are closed off from also thank the Center for Physics of Materials and technicians Richard Talbot, the surrounding chemical environment, for extended imaging and Robert Gagnon, and John Smeros for technical support. For financial in the presence of high-reagent concentrations. For example, cap- support, we thank the National Science and Engineering Research Council turing bound molecular complexes from bulk solution into arrays of of Canada, the Canadian Institutes of Health Research, the Canadian nanowells enables their stoichiometry and distribution of properties Foundation for Innovation, and McGill Physics Department. D.J.B. and C.M.J.M. received research fellowships from the Bionanomachines Collaborative to be determined. Previous nanoreactor approaches using de- Research and Training Experience (CREATE) program and Science Under- formable PDMS lids are fundamentally limited in that the cavities graduate Research Awards, S.M. received a Dr. David T. W. Lin Fellowship exist in either open or closed states and cannot access a continuous from the Faculty of Medicine, and J.S.L. received research fellowships from range of confinement conditions (18). For example, CLINT can the Bionanomachines and Cellular Dynamics of Molecular Complexes couple molecularly thin chambers to embedded structures, enabling CREATE programs.
1. Lam ET, et al. (2012) Genome mapping on nanochannel arrays for structural variation 10. Berard D, et al. (2013) Precision platform for convex lens-induced confinement micros- analysis and sequence assembly. Nat Methods 30(8):771–776. copy. Rev Sci Instrum 84(10):103704. 2. Reisner W, et al. (2010) Single-molecule denaturation mapping of DNA in nanofluidic 11. Wang Y, Reinhart WF, Tree DR, Dorfman KD (2012) Resolution limit for DNA barcodes channels. Proc Natl Acad Sci USA 107(30):13294–13299. in the Odijk regime. Biomicrofluidics 6(1):014101. 3. Reisner W, Larsen NB, Flyvbjerg H, Tegenfeldt JO, Kristensen A (2009) Directed self- 12. Reisner W, et al. (2005) Statics and dynamics of single DNA molecules confined in – organization of single DNA molecules in a nanoslit via embedded nanopit arrays. Proc nanochannels. Phys Rev Lett 94(19):196101 196105. 13. Menard LD, Ramsey JM (2013) Electrokinetically-driven transport of DNA through Natl Acad Sci USA 106(1):79–84. focused ion beam milled nanofluidic channels. Anal Chem 85(2):1146–1153. 4. Reisner W, Pedersen JN, Austin RH (2012) DNA confinement in nanochannels: Physics 14. Bilenberg B, et al. (2006) High resolution 100 kV electron-beam lithography in SU-8. and biological applications. Rep Prog Phys 75(10):106601–106610. Microelectron Eng 83(4):1609–1612. 5. Bruus H, Microfuidics C (2008) Theoretical Microfluidics (Oxford University Press, Oxford). 15. Yang Y, Burkhardt TW, Gompper G (2007) Free energy and extension of a semi- 6. Reisner W, et al. (2007) Nanoconfinement-enhanced conformational response flexible polymer in cylindrical confining geometries. Phys Rev E Stat Nonlin Soft of single DNA molecules to changes in ionic environment. Phys Rev Lett 99(5): Matter Phys 76(1 Pt 1):011804–011811. 058302. 16. Dai L, Ng SY, Doyle PS, van der Maarel JRC (2012) Conformation model of back- 7. Stavis SM, Geist J, Gaitan M, Locascio LE, Strychalski EA (2012) DNA molecules descending folding and looping of a single DNA molecule confined inside a nanochannel. ACS a nanofluidic staircase by entropophoresis. Lab Chip 12(6):1174–1182. Macro Lett 8(8):1046–1050. 8. Cao H, Tegenfeldt JO, Austin RH (2002) Chou SY Gradient nanostructures for interfacing 17. Günther K, Mertig M, Seidel R (2010) Mechanical and structural properties of YOYO-1 microfluidics and nanofluidics. Appl Phys (Berl) 81(16):3058–3060. complexed DNA. Nucleic Acids Res 38(19):6526–6532. 9. Leslie SR, Fields AP, Cohen AE (2010) Convex lens-induced confinement for imaging 18. Shon MJ, Cohen AE (2012) Mass action at the single-molecule level. J Am Chem Soc single molecules. Anal Chem 82(14):6224–6229. 134(35):14618–14623.
13300 | www.pnas.org/cgi/doi/10.1073/pnas.1321089111 Berard et al. Downloaded by guest on September 25, 2021