|Home | About | Journals | Submit | Contact Us | Français|
The anisotropic nanofluidic filter (nanofilter) array (ANA) is a unique molecular sieving structure for separating biomolecules. Here we describe fabrication of planar and vertical ANA chips and how to perform continuous-flow bioseparation using them. This protocol is most useful for bioengineers that are interested in developing automated multistep chip-based bioanalysis systems and assumes prior cleanroom microfabrication knowledge. The ANA consists of a two-dimensional periodic nanofilter array, and the designed structural anisotropy of the ANA causes different sized- or charged-biomolecules to follow distinct trajectories under applied electric fields, leading to efficient continuous-flow separation. Using microfluidic channels surrounding the ANA, the fractionated biomolecule streams are collected and routed to different fluid channels or reservoirs for convenient sample recovery and downstream bioanalysis. The ANA is physically robust and can be reused repeatedly. Compared to conventional gel-based separation techniques, the ANA offers the potential for faster separation, higher throughput, and more convenient sample recovery.
Compared to conventional bioanalysis and diagnostic methods, direct analysis of biologically-relevant molecules (e.g., proteins and nucleic acids) can potentially enhance speed, accuracy, and sensitivity1,2. Moreover, direct biomolecule observations and manipulations help investigators probe fundamental molecular processes in biochemistry and biophysics that are often obscured in ensemble assays3–5. Therefore, nanofluidic systems with characteristic dimensions comparable to molecular scale can provide new opportunities for direct observation, manipulation, and analysis of biomolecules, and these nanofluidic systems can potentially provide innovative platforms to achieve ultra-sensitive and high-resolution biosensing and detection6,7. Inspired by this concept, over the past ten years, there has been a surge of research effort from different scientific disciplines to design and fabricate different nanofluidic systems (with characteristic size scales between 10 to 100 nm) for biological and biomedical applications6,7. Among this surge, one notable research thrust is to design efficient, regular artificial molecular sieving structures that can be a potential alternative to the conventional gel-based separation methods to improve the speed and resolution of biomolecule separation8–10.
Development of efficient nanofluidic sieving structures is essential for optimizing biomolecule separation methods in a chip format within complete microanalysis environments8–10,11. These nanofluidic structures can provide flexible designs and offer precise control over the constraining geometries ideal for molecular sieving and separation. Moreover, the current research thrust in microfluidic based scaled-down analytical processes is stimulating development of different chip-based methods where bioanalysis can be carried out in a faster and lower cost fashion7,12,13. The random nanoporous gel materials that are conventionally used in routine bioseparation applications unfortunately have intrinsic difficulties being integrated into automated multistep chip-based bioanalysis systems. Therefore, the development of a versatile regular nanofluidic sieving structure that can be monolithographically integrated within highly complex bioanalysis microsystems will have profound implications for different biological and biomedical applications.
Over the past decade, there have been exciting developments in the field of artificial regular sieving structures due to the advance of micro- and nanotechnology8–10. So far, a myriad of regular sieve designs have been demonstrated with varying degrees of success in biomolecule separation14–18. These reported artificial sieving structures have been primarily successful for the separation of large biomolecules such as viral DNA (as reviewed in Ref. ). Recently, we have introduced a unique regular molecular sieving structure called the anisotropic nanofluidic filter (nanofilter) array (ANA)19,20, and have demonstrated its implementation for high-resolution continuous-flow separation of a wide range of DNA fragments (between 50 to 23,000 base pairs (bp)) and proteins (between 11 to 400 kDa) within a few minutes. In this protocol we describe the detailed fabrication of the ANA chips and how to perform continuous-flow bioseparation using these chips.
Separation of biomolecules in a biology laboratory is currently routinely achieved with gel-based separation techniques, such as gel-exclusion chromatography and slab-gel electrophoresis21,22. These gel-based separation techniques use gelatinous materials that consist of cross-linked three-dimensional nanometer-sized pore networks. Despite the popularity of both the gel-based separation methods, they still exhibit several limitations that make neither method optimal in separating complex mixtures for downstream analysis21. For example, the chief limitations of gel-exclusion chromatography are that the separation can be slow and that the resolution of the emerging peaks is limited, and additionally, relatively large amounts of samples are necessary to obtain visible and well-resolved bands. In slab-gel electrophoresis, recovery of the separated sample is often problematic and inefficient: the location of the band in the gel must be physically cut out, and additional washing steps are necessary to extract the desired sample from the gel, which often lead to significant sample loss21.
Gel-based sieving structures have been recently successfully incorporated into different chip-based bioanalysis systems to separate both nucleic acids and proteins23,24. These gel-based microsystems have provided convincing evidences for separation speed and resolution. However, as mentioned earlier, these gel-based microsystems still pose the intrinsic difficulties for integration with other bioanalysis components and packaging as a whole lab-on-chip system. Most of these gel-based microanalysis systems are not reusable, economically not an ideal choice for large-scale screen for different systems-biology applications. In contrast, the ANA can be monolithographically integrated within bioanalysis microsystems as an upstream sample preparation component to separate and purify complex biological samples, and the ANA is physically robust and can be reused repeatedly19,20.
Compared to the gel-based sieving structures, two major limitations currently exist for the ANA: the separation resolution and the sample throughput. The ANA structure needs to be further optimized for separation resolution to fully realize its promise for on-chip based proteomic research and biomarker discovery. The current size selectivity of the ANA structure is about 2–3 nm, which corresponds to the end-to-end distance of 10 bp DNA; however, it is still not optimal for protein separation. In the future, an improved ANA structure should have size selectivity comparable to the gel-based techniques and should discriminate and separate proteins (either native or denatured) with a molecular weight difference of about 5 kDa. Two approaches might be undertaken to further improve size selectivity and therefore separation resolution of the ANA structure. First, the size selectivity of the ANA should be improved by scaling down the nanofilter structures (period, gap size, etc.) with advanced sub-100 nm resolution lithography techniques25,26. We have proved both theoretically and experimentally that the size selectivity of the ANA is inversely proportional to both the nanofilter period (pitch size) and the nanofilter gap size27,28. Second, it is possible to extend the separation functionality of the ANA structure by utilizing the Debye layer, electro-osmosis, and surface chemistries, together with the geometrical constraints of the ANA, to achieve biomolecule separation based on a suite of molecular properties (e.g. size, charge, or hydrophobicity)10. For example, we have recently demonstrated that by actively controlling the buffer ionic strength and therefore the Debye length, switchable size- or charge-based separation of proteins can be achieved within the ANA19.
The sample throughput of the ANA needs also to be improved for its future implementations in different biological and biomedical applications, even though this concern is alleviated greatly by the ever increasing number of on-chip biosensing and detection mechanisms and their much improved detection limits. For the current ANA design, the maximum volume throughputs for protein samples are in the range of μL/h. In the future, large-scale bioseparations in a chip format with high sample throughput (in the range of μL/min) would be desired. A key for this function goal would be to devise a convenient and inexpensive method to fabricate robust molecular sieves with highly parallel nanopores.
Different regular micro/nanofluidic sieving structures have been reported recently in the literature to separate biomolecules such as viral DNA with fast speed and great resolution14–18. One notable example was the “DNA Prism” devised by Huang et al. to continuously separate long DNA fragments (61–209 kilo-base pairs (kbp)) within a few minutes15, a speed much faster than conventional pulsed-field gel electrophoresis (PFGE) and pulsed-field capillary electrophoresis. Han and Craighead recently designed an entropy trap array device to separate long DNA ladder samples (5–50 kbp) in about 30 min14. The ANA structure described in this protocol is a more recent development, and it offers several distinct advantages when compared to these previously reported micro/nanofluidic sieving structures, rendering it a promising generic molecular sieving structure for an integrated bioanalysis microsystem29–31. These unique advantages of the ANA include: (a) the capability to continuously separate physiologically-relevant molecules including proteins in a few minutes; (b) the versatile separation mechanisms (e.g., Ogston sieving, entropic trapping, and electrostatic sieving) that can take effect in the ANA to separate biomolecules covering broad biological size ranges and based on different molecular properties; (c) the continuous-flow operation of the ANA that facilitates the integration of the separation step with upstream or downstream analysis steps; thus, allowing the bioseparation to be performed “in-line” with other continuous flow processes; (d) the continuous-flow operation of the ANA allows downstream sampling (either of a detection signal or of the separated biomolecules themselves) to be time-integrated to improve the detection limit. In our opinion, these advantages of the ANA can have profound implications for proteomic research and biomarker discovery on a chip format32,33.
This Protocol is most useful for bioengineers and bioanalytical chemists that are interested in developing automated multistep chip-based bioanalysis systems7,12,13. Since the fabrication process of the ANA structure involves standard microfabrication techniques in a cleanroom environment, the researchers implementing this protocol need to receive necessary training from the cleanroom staffs about how to operate the required cleanroom equipments. Ideally, the researchers should either assume previous knowledge and experience in cleanroom microfabrication or have direct input of a cleanroom technologist in the research team. It is our goal that with the guidance of this protocol, the ANA structures can be fabricated in a fully-equipped and well-staffed cleanroom. It is also worth mentioning that, now that many different well-equipped microfabrication foundries are available worldwide for fabricating micro/nanoscale devices, even if the necessary microfabrication facility is not available locally or on campus, the designs of the ANA structure can still be out-sourced to fabrication foundries.
In the Protocol, the design and fabrication methods for two different types of the ANA-the planar ANA19 (Fig. 1a, as discussed in Procedure) and the vertical ANA20 (Fig. 1b, as discussed in Box 1)-will be discussed in detail. The design of both the ANA structures consists of a two-dimensional periodic nanofilter array, and the designed structural anisotropy of the ANA causes different sized- or charged-biomolecules to follow distinct trajectories, leading to efficient continuous-flow separation. The ANA structures are batch fabricated using conventional semiconductor microfabrication techniques on a silicon wafer. Using standard microfabrication techniques such as photolithography, reactive ion etching (RIE) or deep reactive ion etching (DRIE), and anisotropic potassium hydroxide (KOH) etching, the ANA structures with a nanofilter gap size down to about 10 nm can be fabricated27,34. Three different separation mechanisms have been demonstrated successfully with the ANA to separate both DNA and proteins based on either size (Ogston sieving or entropic trapping) or charge (electrostatic sieving)19. In the planar ANA structure, the relevant nanoscale constriction dimension of the planar nanofilter is etched by the RIE technique into the thickness direction of the silicon substrate (Fig. 1a); while in the vertical ANA, the nanoscale constriction of the vertical nanofilter is fabricated by taking advantage of the highly selective anisotropic KOH wet etching of the (110) silicon planes for high-aspect-ratio silicon structures with smooth and vertical sidewalls (Fig. 1b).
In the Experimental Design section, we first discuss the different separation mechanisms applicable for the ANA to separate biomolecules covering different size ranges and based on different molecular properties, and then we describe in detail the design guidelines of the ANA structure and the different microfabrication techniques involved in the ANA fabrication. Finally we describe the implementations of the ANA to separate different biomolecules such as DNA and proteins. In the Procedure section, a step by step description of the fabrication process is provided, followed by a troubleshooting table with information on how to troubleshoot the most likely problems encountered with the protocol. In the Anticipated Results section, we briefly describe the likely outcome of using the ANA to separate biomolecules for different applications.
Three different separation mechanisms (i.e., Ogston sieving, entropic trapping, and electrostatic sieving) have been applied in the ANA to separate biomolecules covering different size ranges and based on different molecular properties (see Ref.  for a detailed discussion of these separation mechanisms). For both Ogston sieving and entropic trapping, bioseparation should be conducted at high ionic strength where the Debye length becomes negligible compared to the nanofilter constriction size. Ogston sieving is effective for biomolecules with diameters smaller than the nanofilter constriction. Smaller molecules have a higher tendency to jump across the nanofilter and therefore assume a larger stream deflection angle in the ANA. In entropic trapping, diameters of biomolecules are greater than the nanofilter constriction size, and passage requires the molecules to deform to sneak through the nanofilter constriction. Since longer molecules have a greater probability to jump across the nanofilter constriction, they will assume a greater deflection angle in the ANA. For low ionic strength solutions where the Debye length becomes comparable to the nanofilter constriction size, electrostatic interactions (either repulsive or attractive) between charged biomolecules and charged nanofilter walls become prominent and start to dictate jump dynamics across the nanofilter. Therefore, similar sized biomolecules bearing different net charges are energetically favored to different degrees for passage through the nanofilter, resulting in efficient separation in the ANA. In practice, Ogston sieving and entropic trapping are most suitable for size separation of linear flexible biomolecules such as DNA and denatured proteins, while electrostatic sieving is most suitable for separation of native biomolecules such as globular proteins by both size and charge19. In addition, electrostatic sieving under the low ionic strength buffer can result in markedly increased size selectivity and therefore higher separation resolution for negatively charged biomolecules in the ANA35. This enhancement in size selectivity is likely due to an effective decrease in the nanofilter constriction size caused by electrostatic repulsion between negatively charged biomolecules and like-charged nanofilter walls.
The design of the ANA consists of a two-dimensional periodic nanofilter array. Nanofilters with a constriction size between 10–100 nm are arranged in rows and are separated by deep channels (Fig. 2). When injected into the ANA, the biomolecule stream is fractionated into different streams that are collected at intervals along the ANA opposite edge. Microfluidic channels are designed to surround the ANA, and they connect the ANA to fluid reservoirs where voltages are applied. The microfluidic channels provide sample loading and collection ports, and they also serve as electric-current injectors to generate uniform electric fields over the entire ANA structure (for more discussion, please refer to Ref. [15, 36]).
The planar ANA contains planar nanofilters whose nanoscale constrictions are defined by the nanofilter shallow region and the top glass ceiling. The vertical ANA contains vertical nanofilters whose nanoscale constrictions are defined by the narrow gaps formed between adjacent silicon pillars. The key structural parameters of the planar ANA include the planar nanofilter width (wps), length (lps) and depth (dps), deep channel width (wpd) and depth (dpd), and rectangular pillar width (wpp) and length (lps) (Fig. 2d, top). The key structural parameters of the vertical ANA include the silicon pillar gap width (wvs), deep channel width (wvd) and depth (dvd), and rectangular pillar width (wvp) and length (lvs) (Fig. 2d, bottom). The structural parameters of the nanofilter depth dps and the deep channel depth dpd of the planar ANA and the silicon pillar gap width wvs and deep channel depth dvd of the vertical ANA are defined during the cleanroom fabrication process (for example, by controlling the RIE etching time), while the others are all defined during the photomask design. The size selectivity of the ANA is largely determined by the nanofilter constriction size and the nanofilter row pitch size (for the planar ANA: dps and wpd+ lps; for the vertical ANA: wvs and wvd+lvs). In principle, smaller nanofilter constriction size and nanofilter row pitch size will result in enhanced size selectivity for bioseparation. The sample throughput of the ANA is largely determined by the dimensions of the sample injection channels in the ANA. These key structural parameters can be optimized for different bioseparation applications.
The following techniques are involved in the ANA fabrication, and they need to be considered carefully: photomask manufacture, photolithography, RIE and DRIE, plasma-enhanced chemical vapor deposition (PECVD) and low-pressure chemical vapor deposition (LPCVD), KOH wet etching, and anodic bonding.
Continuous-flow bioseparation in the ANA is achieved by applying two orthogonal electric fields across the ANA to drive the biomolecules to be analyzed to migrate across the nanofilter array. The horizontal electric field, Ex, drives the biomolecules to jump across the nanofilter constrictions along the x-direction (Fig. 2d; also see Fig. 1 in Ref. ), and the vertical electric field, Ey, causes drifting of biomolecules in the deep (for planar ANA) or wide (for vertical ANA) microchannels (Fig. 2d). Separation speed and resolution of the ANA are mainly modulated by these two independent electric fields Ex and Ey, respectively: higher Ex leads to a greater separation resolution while higher Ey leads to a faster separation speed. Careful regulation of both Ex and Ey simultaneously will be necessary for a rapid separation with high resolution.
The ANA can be used to separate different physiologically-relevant biomolecules such as DNA, proteins, and carbohydrates. In our experiments, DNA molecules shorter than about 1 kbp and longer than about 2 kbp can be rapidly separated by the ANA based on Ogston sieving and entropic trapping, respectively, both with Tris-Borate-EDTA (TBE) 5× buffer (see Materials section)19. The size selectivity achieved so far for DNA in the Ogston sieving and entropic trapping regimes is about 10 bp and 1 kbp, respectively. For proteins (either native or denatured), either Ogston sieving or electrostatic sieving can be used for separation, depending on the ionic strength conditions chosen for different applications. TBE 5× and TBE 0.05× buffers have been used for separation of proteins by Ogston sieving and electrostatic sieving, respectively19. Additional 0.1% wt/vol sodium dodecyl sulfate (SDS) needs to be added to TBE 5× buffer for separation of denatured proteins by Ogston sieving. Non-specific adsorption of negatively charged native proteins in the ANA was not found to be significant, possibly due to electrostatic repulsion from the like-charged hydrophilic ANA walls. As a proof of principle, we have also recently achieved in the ANA separation of proteins with small molecules such as fluorescence dyes and different sized carbohydrates (collaboration with Dr. Ram Sasisekharan at the Massachusetts Institute of Technology, data not shown) in TBE 5× buffer based on Ogston sieving.
To visualize in-situ separation of biomolecules in the ANA, the biomolecules can be fluorescent-labeled with different dyes and then detected with the fluorescence microscopy method. To further detect and indentify the fractionated biomolecules after purification and separation in the ANA, the biomolecules can be routed by the microfluidic channels to different on-chip downstream bioanalysis components such as the microfluidic enzyme-linked immunosorbent assay (ELISA) to detect the presence of an antibody or an antigen in a complex biological sample.
The Delrin gadget is designed to hold the ANA device during bioseparation experiments. The Delrin gadget contains four different machined parts: one Delrin rectangular cuboid, one stainless steel plate, one silicone rubber gasket, and one printed circuit board (PCB) with soldered Pt wires (Panel b of Fig 3). The Delrin cuboid contains ten drilled through-holes that can connect to the buffer access holes in the ANA device. During assembly, the stainless steel plate and the Delrin cuboid are screwed together with the ANA devices and the silicone rubber gasket to completely seal the ANA devices. Both the stainless steel plate and the silicone gasket have a square opening at the center to allow for observing bioseparation in the ANA using an inverted epi-fluorescence microscope. The Pt wires on the PCB serves as electrodes to provide an electric connection between the buffer solution and the external power supply.
CRITICAL The operational conditions of the microfabrication techniques used below depend strongly on the many process parameters (such as pressure, temperature, gas flows, and radio frequency (RF) power) and the specific machine used; therefore, the parameters listed in the Procedure are given only for guidance. These parameters should be optimized empirically for each type of application.
|Gas||O2 (10 sccm)|
|RF Power||100 W|
|Gas||O2 (40 sccm)|
|RF Power||300 W|
|Gas||Cl2 (20 sccm), HBr (20 sccm)|
|RF Power||300 W|
|Gas||NH3 (25 sccm), SiH2Cl2 (250 sccm)|
|Nitride deposition rate||3 nm/min|
|Gas||CF4 (8 sccm), O2 (6 sccm)|
|RF Power||250 W|
|Nitride etch rate||3.4 nm/s|
(Planar ANA discussed in the Procedure)
Patterning alignment marks (timing: 3.5 h): Step 1–10: 2 h; Step 11–16: 1.5 h.
Fabrication of nanofilter shallow regions (timing: 3.5 h): Step 17: 3.5 h.
Fabrication of nanofilter deep regions (timing: 3.5 h): Step 18: 3.5 h.
Fabrication of buffer access holes (timing: 20 h): Step 19–23: 5 h; Step 24–27: 2 h; Step 28–32: 12 h; Step 33: 1 h.
Thermal oxidation and anodic bonding (timing: 12 h): Step 34–36: 5 h; Step 37–42: 7 h.
Bioseparation with ANA (timing: 3 h): Step 43–48: 3 h.
(Vertical ANA discussed in the Box 1)
Patterning alignment marks to find (111) planes of (110) wafer (timing: 10 h): Step i–iii: 5 h; Step iv–vi: 1 h; Step vii: 1 h; Step viii–x: 3 h.
Patterning alignment marks (timing: 2 h): Step xi: 2 h.
Fabrication of ANA narrow channels (timing: 4 h): Step xii: 4 h.
Fabrication of ANA wide channels and other microfluidic channels (timing: 17.5 h): Step xiii: 0.5 h; Step xiv–xxii: 11 h; Step xxiii–xxv: 6 h.
Fabrication of buffer access holes (timing: 8.5 h): Step xxvi: 0.5 h; Step xxvii–xxxi: 8 h.
Bioseparation with ANA (timing: 3 h): Step xxxii: 3 h.
The stream deflection pattern during separation depends on the separation mechanisms applied in the ANA. For Ogston sieving, smaller molecules will assume a larger steam deflection angle in the ANA and will be collected by the microfluidic channels on the right-hand side of the ANA opposite edge. In contrast, for entropic trapping, since longer molecules assume a greater deflection angle in the ANA, they will be deflected towards the right-hand side of the ANA opposite edge and collected by the microfluidic channels. For electrostatic sieving, the separation pattern would be more difficult to predict. In principle, smaller biomolecules bearing less negative charges will be deflected more in the ANA, towards the right-hand side of the ANA opposite edge.
During separation, we can use the CCD camera attached to the microscope to visualize and record the migration trajectories of fluorescence-labeled biomolecules in the ANA. These fluorescence images can be further analyzed with the image processing software (e.g., IPLab, BD Biosciences Bioimaging, Rockville, MD) to generate fluorescence intensity profiles. Gaussian functions can be used for fitting these fluorescence intensity profiles to determine the means (the maximum intensity) as well as the widths of the fractionated biomolecule streams19. The separation efficiency of the ANA can then be further quantified by calculating both the size selectivity and the effective peak capacity under different electric field conditions22. Figure 3e–f shows examples of composite fluorescent photographs of continuous-flow separation of proteins (Fig. 3e) and DNA (Fig. 3f) through both the planar and the vertical ANA, respectively. Upon application of the electric field Ey along the y-axis of the ANA, the initial biomolecule stream is continuously injected into the deep (for the planar ANA) or wide (for the vertical ANA) channels on the top left of the ANA structure (Fig. 3fi). After the orthogonal electric field Ex is superimposed along the x-axis across the nanofilters, the drifting biomolecules in the deep channels start to jump across the nanofilters and separate from each other (Fig. 3fii&iii). With properly adjusted values of both Ex and Ey, both the protein mixtures and the long DNA digest can be based-line separated into distinct streams within a few minutes (Fig. 3e–f). The sample volume throughput of the ANA structure can be estimated based on the migration speed of biomolecules in the ANA and the dimensions of the ANA sample injection channels. For the protein separation shown in Fig. 3e, the sample volume throughputs for the planar and vertical ANA are in the range of nL/h and μL/h, respectively. The sample throughput of the ANA can be scaled up by parallelism with multi-device processing. For the vertical ANA, the sample throughput can be further enhanced by increasing the depth of the ANA structure to a maximum value of the silicon wafer’s thickness, which is about 500 μm for a 6″ wafer.
|Gas||NH3 (25 sccm), SiH2Cl2 (250 sccm)|
|Nitride deposition rate||3 nm/min|
|Gas||SiH4 (50 sccm), N2O (800 sccm)|
|RF Power||270 W|
|Oxide deposition rate||10 nm/sec|
|Gas||CHF3 (45 sccm), CF4 (15 sccm), Ar (100 sccm)|
|RF Power||200 W|
|Etch rate||3.5 nm/s|
|Etch mode||Passivation mode|
|Process time:||6 s||4.5 s|
|Overrun:||0.5 s||0 s|
|Platen generator power:||80 W||60 W|
|Coil generator power:||600 W||600 W|
|Gas:||SF6 (70 sccm)||C4F8 (35 sccm)|
|Etch rate:||1.47 μm/min||N/A|
|Gas||O2 (45 sccm)|
|RF Power||200 W|
|Gas||H2 (5 sccm), O2 (10 sccm)|
The authors acknowledge financial support from the National Institute of Health (EB005743), Korea Institute of Science and Technology–Intelligent Microsystems Center (KIST-IMC), and the Singapore-MIT Alliance (SMA-II, CE program). We also thank J. Yoo for his contribution in the experimental setup, H. Bow and S. Reto for helpful discussions. The MIT Microsystems Technology Laboratories is acknowledged for support in microfabrication.
COMPETING INTERESTS STATEMENT
The authors declare that they have no competing financial interests.
AUTHOR CONTRIBUTIONSJ. Fu, P. Mao, and J. Han conceived and designed the ANA chips. J. Fu and P. Mao fabricated the ANA chips. J. Fu and P. Mao designed and performed experiments, and analyzed data. J. Fu and P. Mao wrote manuscript.