|Home | About | Journals | Submit | Contact Us | Français|
Although polydimethylsiloxane (PDMS) microfluidic chips provide an alternative to more expensive microfabricated glass chips, formation of monolithic stationary phases in PDMS is not a trivial task. Photopolymerized silica sol-gel monoliths were fabricated in PDMS based microfluidic devices using 3-trimethoxysilylpropylmethacrylate (MPTMOS) and glycidyloxypropyltrimethoxysilane (GPTMOS). The monolith formation was optimized by identifying a suitable porogen, controlling monomer concentration, functional additives, salts, porogen, wall attachment methods, and rinsing procedures. The resulting monoliths were evaluated using scanning electron microscopy, image analysis, differential scanning calorimetry, and separation performance. Monoliths functionalized with boronic acid ligands were used for the separation of cis-diol containing compounds both in batch mode and in the microfluidic chip.
Extensive research has been performed developing monolithic stationary phases and integrating them into microfluidic systems [1–8]. Monolithic stationary phases are generally organic based or silica based. Recent reviews are available discussing both types [9–12]. However, the majority of reports utilize organic based monoliths in glass microfluidic chips due to the relative ease of adapting the formulations from the capillary format. Polymerization by UV light allows for the selective localization of the monolith within a microfluidic chip without the need for retaining frits. Immobilizing separation columns into microfluidic chips without the need for frits greatly simplifies the experimental protocol. However, many of the available chemistries developed in the capillary format have not been adapted to the microfluidic chip format.
A variety of separations have been performed using monoliths in microfluidic chips, although to a lesser extent than capillary based separations. Numerous separations whereby glycidyl methacrylate (GMA) based monoliths have been formulated in microfluidic chips and functionalized with various affinity ligands have been demonstrated. Anti-FITC antibodies have been used to extract FITC-labeled amino acids and proteins from mixtures . A GMA monolith with immobilized lectins allowed for the separation of chicken ovalbumin, turkey ovalbumin, and ovomucoid in a glass microfluidic chip . Li et al. functionalized a GMA based monolith with Cibacron blue and demonstrated the removal of albumin from cerebrospinal fluid in a glass microfluidic device.
The cost of fabricating glass microfluidic chips does not lend them to be used in disposable devices. PDMS has been established as a useful substrate for lab-on-a-chip devices[1, 17, 18] since it is inexpensive, relatively easy to fabricate microstructures via soft lithography, has good optical characteristics, and is biocompatible. Due to these benefits, numerous groups have used PDMS for microfluidic separation devices. However, compared to glass devices, which are analogous to capillaries, microfluidic chips fabricated from PDMS must overcome certain weaknesses. These include hydrophobicity of native PDMS, partitioning of small organic molecules into PDMS, and swelling of PDMS by numerous organic compounds. PDMS surface modifications have been developed to make the material more amenable to lab-on-a-chip devices but device reproducibility decreases as more steps are added to the fabrication process[21–24]. Monoliths fabricated in PDMS devices include a polyacrylamide gel functionalized with β-cyclodextrin used for the separation of amino acid enantiomers  and a hexyl acrylate monolith used for the separation of catecholamines. Kang et al. have employed monolith columns in an integrated multilayer PDMS/glass microchip for micro-extraction and electrophoresis. They also developed a 3D monolith-based platform in PDMS microfluidic channels for immunocapture and on-line immunoassay of proteins and viruses. Numerous factors influence the performance of monoliths for separations in PDMS microfluidic devices. The monolith formation should be homogenous within the channel and have an appropriately sized pore structure. Additionally, a suitably high ligand density must be achieved to thoroughly extract the desired analyte while minimizing non-specific binding.
The bimodal pore structure and high surface area of silica monoliths makes them an attractive alternative to the widely researched organic monolith. However, the thermal polymerization requirement associated with silica monoliths makes it difficult to transfer capillary work into the microfluidic chip format since the stationary phase needs to be selectively localized. Thus only a small percentage of the available chemistries that have been developed in the capillary format have subsequently been adopted into the microfluidic chip format. Methods to selectively pattern monoliths into microfluidic devices have been developed. Jindal and coworkers developed the method of selective filling whereby differences in free energy between filled and unfilled portions of the channel allow for the patterning of the stationary phase. This was demonstrated in a quartz/PDMS microfluidic chip for thermally activated tetraethoxysilane sol-gel. Giordano et al. developed a technique to deposit tetramethoxysilane based sol-gels into quartz microfluidic devices and showed the separation of nitroaromatic and nitroamine compounds. Using channel geometry effects and precise fluid control, Ishida and coworkers were able to introduce an octadecylsilane modified tetramethoxysilane monolith into a glass microfluidic chip. Dulay and coworkers developed a UV polymerizable sol-gel that blends the sol-gel approach and the methacrylate free-radical polymerization approach of organic monoliths . After development of the appropriate stationary phases in the capillary format, 3-trimethoxysilylpropylmethacrylate (MPTMOS) was used as a monomer to fabricate a 4.7 cm long column in a glass microfluidic chip However, due to both required solvents and polymerization conditions, the formulation of silica sol-gel based monoliths in PDMS based devices is not a trivial task.
The current work focuses on adapting photopolymerized silica sol-gel based monoliths for use in PDMS microfluidic devices. This paper describes optimization of the polymerization conditions, characterization of the resulting monolith structures, functionalization with affinity ligands, and binding to target molecules.
MPTMOS, glycidyloxypropyltrimethoxysilane (GPTMOS), bovine serum albumin (BSA), conalbumin, 4-aminophenylboronic acid, glycerol diglycidyl ether, phenol, ethylene glycol diglycidyl ether, poly(ethylene glycol) diglycidyl ether, lithium chloride, calcium chloride, sodium acetate, reagent alcohol were purchased from Sigma-Aldrich (St. Louis, MO). Ammonium acetate, sodium hydroxide, and methanol were purchased from Fisher Scientific (Fairlawn, NJ). Alexa Fluor 488 was purchased from Invitrogen (Carlsbad, CA). Sylgard 184 silicone elastomer kit was purchased from Dow Corning (Midland, MI). SU-8 2025 and SU-8 developer were purchased from Microchem Corp (Newton, MA). Nanoport assemblies were purchased from Upchurch Scientific (Oak Harbor, WA). Aluminum hermetic sample pans were purchased from Instrument Specialists (Twin Lakes, WI). Irgacure 1800 was kindly donated from Ciba (Tarrytown, NY).
Monomer precursor was mixed with 0.1N HCl in the ratio of 1 mmol monomer per 42 µl HCl for 15 minutes at room temperature. Irgacure 1800, was mixed with the porogen at a concentration of 5 wt% of the final pre-polymer solution. Monomer solution was then added to the porogen solution (producing 15 to 50 wt % monomer) and was mixed until a homogenous solution was formed. For monolith containing salt additives, salt was dissolved in 85% of the total HCl added to the monomer. Monomer was prehydrolyzed with HCl that did not contain salt for 5 minutes. The HCl containing salt was added at 5 minutes and mixing continued for a total of 15 minutes. Drops were spotted onto PDMS slabs and polymerized for 5 minutes using UV 365 nm light using a UVL-56 Blak ray lamp (UVP, Upland, CA). Drops were then rinsed with ethanol, coated with platinum (Denton Desk IV, Denton Vacuum, Moorestown, NJ) and imaged using SEM (Supra 55 Thermal Field Emission SEM, Carl Zeiss SMT).
Molds were prepared from SU-8 photoresist on silicon wafers using procedures described elsewhere. PDMS was mixed according to the manufacturer’s directions, degassed, poured onto the SU-8 mold and baked at 70°C for two hours. PDMS flats were prepared by baking in flat petri dishes. Enclosed channels were formed by bonding PDMS containing the channel to a flat slab using plasma oxidation to form an irreversible seal (Tegal Plasmaline 211 barrel asher, Tegal, Petaluma, CA). Sol-gel pre-polymer solution was prepared in the same manner as for bulk screening experiments and introduced into the channel using a hand-held syringe interfaced with a Nanoport fitting. The channel is masked with electrical tape, monolith is selectively polymerized using UV light at 365 nm, and excess reactants are removed with air and ethanol to yield the monolithic sol-gel column. For SEM visualization, chips were sectioned into small pieces using a razor blade, coated with platinum to prevent charging, and imaged.
A calibration curve for DSC measurements was constructed using silica particles with pore sizes ranging from 22–150 angstroms. Particles were saturated with water, weighed, and placed in aluminum hermetic pans, and sealed with a press. Temperature was ramped to −25°C at 10°C/min; −10°C at 10°C/min; 5°C at 1°C/min (DSC-Q100, TA Instruments, New Castle, DE). DSC experiments were performed in the same way on bulk monolith samples.
Image analysis of SEM micrographs was performed by importing images into Image J software. Based on the grayscale, images were converted to binary and pores were recognized by the software.
Monoliths were prepared in bulk or in the microfluidic channel as discussed previously. Aminophenylboronic acid was dissolved in water (0.78 wt %) and adjusted to pH 8.3 with sodium hydroxide. For bulk samples, sol-gel was reacted with boronic acid solution for 18 hours at 50°C in an oven. For chip samples, boronic acid solution was introduced into the channel using a hand held syringe interfaced to a Nanoport fitting. Large reservoirs were placed at the channel inlet and outlet and the chip was heated for 18 hours on a hot plate.
Monolith was polymerized in uniform 20 µl drops spotted on PDMS slabs. Monolith was progressively washed with ethanol, water, and buffer followed by loading with 5 mg/ml catechol or quinol solutions. The supernatant was extracted, multiple buffer rinses were loaded and extracted, and elution buffer of 100 mM sodium acetate, pH 3.8 was loaded an extracted. All supernatants were analyzed using UV 280 nm absorbance (HTS 7000 plate reader, Perkin Elmer, Waltham, MA) and the amount eluted from each step was determined by mass balance.
Monolith was formed in PDMS microfluidic chips using the conditions described previously. Rinsing was performed using a PC77 Pressure Injection Cell (Next Advance, Averill Park, NY). Alexa Fluor 488 labeled protein was introduced to the chip under electroosmotic flow controlled by an in-house built setup consisting of high voltage power supplies (0–3 kV, Bertran Associates Inc., Syosett, NY) controlled by LabView software. Electrical contact to the microfluidic chip was provided by platinum wires (Alfa Aesar, Ward Hill, MA). Flow was monitored using a confocal fluorescence microscope (Zeiss LSM 510 Meta, Carl Zeiss SMT). Fluorescence values were quantified using Image J software (NIH).
The role of a porogen is to serve as both a template for the formation of throughpores and a solubilizer for the silane monomer. Furthermore, a porogen should solubilize the monomer components and photoinitiator, but not the resulting polymer. To this end, numerous potential porogens were screened. Initial screenings tested the solubility of the photoinitiator in the porogen and its compatibility with PDMS. Numerous organic compounds were selected with most being glycidyl ether based since such porogens had previously been used in literature for sol-gel networks in PDMS devices. Four porogen candidates (ethylene glycol diglycidyl ether, glycerol diglycidyl ether, and phenol, poly(ethylene glycol) diglycidyl ether) resulted in the polymerization of a solid gel and were subsequently visualized by SEM (Figure 1). As can be seen in the figure, the most favorable monolith morphology resulted when ethylene glycol diglycidyl (EGDE) ether was used as a porogen, which produced suitable 1–2 µm macropores and a globular sol-gel monolith network structure.
Photopolymerized sol-gel monoliths were then formulated in PDMS microfluidic chips using a modified recipe from Dulay and coworkers and channel dimensions previously used in our group. Specifically, monomer solution containing MPTMOS with 5% GPTMOS (by mole silane content) was formulated as described in the experimental section. The monomer solution was then mixed with Irgacure 1800 (photoinitiator) and EGDE (porogen) in varying amounts of monomer to porogen. When the recipe from Dulay et al. is followed, except using EGDE as the porogen in a 200 µm × 40 µm PDMS microfluidic channel, sol-gel formation is only observed along the walls of the channel (Figure 2a). This indicates that there is insufficient monomer present to form the monolithic network. When the monomer content in increased to 44%, sol-gel forms across the entire cross-section of the channel but it lacks visible throughpores (Figure 2b). Additional studies were carried out to tune the optimum monomer content and 28% monomer was found to produce the best combination of cross-sectional filling and micron sized throughpores (Figure 2 c–d).
Since the above channel dimensions of 200 µm × 40 µm did not result in full cross-sectional coverage of the monolith, the channel dimensions were changed to 100 µm × 40 µm. By increasing the aspect ratio, the wall wetting of the pre-polymerization solution is improved, which leads to increased cross-sectional filling. This led to proper filling and some macropore formation; however, the pores were <500 nm instead of the desired 1–2 µm (Figure 2e). The monomer content in these channels was also varied (less monomer) but this resulted in less cross-sectional filling rather than larger throughpores. Furthermore, when the epoxide content was increased above 5%, macropores failed to form, as visualized by SEM.
Balancing cross-sectional filling and macropore formation could not be accomplished solely by varying monomer content and channel aspect ratio. The hydrophobicity of the PDMS compatible porogen, ethylene glycol diglycidyl ether, is less than that of toluene. Thus, phase separation between the sol and the porogen may not be sufficient to produce the desired micron sized macropores. Poly(ethylene glycol) has commonly been used as a throughpore template and silane solubilizer[37, 38]. However, in the photopolymerized sol-gel system under investigation, addition of PEG (MW 3350) did not show any increase in throughpore structure as visualized by SEM. To this end, numerous salts were screened as additives to increase the phase separation between the aqueous sol and the porogen and thus induce the formation of macropores.
Salt was dissolved in water (near the solubility limit) and mixed with EGDE porogen. Visual inspection of the solutions for cloudiness was conducted, which served as a qualitative measure of the amount of phase separation induced between the aqueous salt mixture and the porogen. The salts tested in this manner included: sodium chloride, sodium acetate, sodium bromide, sodium nitrate, sodium sulfate, sodium iodide, sodium thiocyanate, sodium phosphate, disodium hydrogen phosphate, magnesium chloride, calcium acetate, calcium chloride, potassium nitrate, dipotassium phosphate, lithium bromide, and lithium chloride. Based on the bulk observations, the following salts were incorporated into the sol-gel mixture: 12M lithium chloride, 3.8M calcium chloride, 0.67M sodium sulfate, 4M sodium acetate, and 5M disodium hydrogen phosphate.
As described in the experimental section, it is important to note that in order to incorporate the salts into the sol-gel monomer mixture, pre-hydrolysis of the MPTMOS monomer was necessary. Upon application of UV light, only sols containing lithium chloride or calcium chloride resulted in a mixture that successfully polymerized. Pre-polymerization solutions containing either 12M lithium chloride or 3.8M calcium chloride were then introduced into the microfluidic chip. Both formulations produced suitably porous monoliths; however, only the monolith formed with calcium chloride also resulted in good cross-sectional coverage (Figure 3).
Incorporating the maximum amount of GPTMOS into the sol-gel monolith is critical for chip based chromatographic separations since the epoxide functionality can be used to subsequently attach chromatographic ligands. Increasing the epoxide content serves to both increase the column capacity and minimize the effects of any undesired non-specific binding. If the total amount of GPTMOS is added at the start of the reaction, the mixture fails to polymerize when higher than five percent of the silane monomers is GPTMOS. If GPTMOS is added in too high a concentration it will preferentially cross-link with itself rather than with MPTMOS monomers. This causes the resulting oligomers to lack a UV curable group for subsequent polymerization. A modified protocol was thus developed whereby GPTMOS is titrated into the monomer mixture throughout the sol-gel mix time. Specifically, MPTMOS and HCl are mixed for 5 minutes to initiate the hydrolysis reaction. Next, 5 equal aliquots of GPTMOS are added to the reaction mixture every 2 minutes for the remainder of the 15 minute mix time. This approach forces GPTMOS to increasingly link with MPTMOS rather than with itself, which allows UV polymerization to occur at higher concentrations of GPTMOS. Given perfect mixing, up to 50% of the silane monomers could be the epoxide containing GPTMOS. In bulk experiments, polymerization into a solid gel occurred when up to 40% of the silane was GPTMOS.
Monolith formation in the PDMS microfluidic channel was studied under increased GPTMOS content, using ethylene glycol diglycidyl ether as the porogen, and 3.8M calcium chloride in the aqueous HCl. Results indicate that as the epoxide content is increased, the gel becomes denser and lacks macropores suitable for bulk fluid flow (Figure 4a). Decreasing the monomer content does not lead to macropores; rather, it results in decreased cross-sectional monolith coverage in the channel. To increase the formation of macropores, a modified rinsing protocol was developed. Using a constant pressure device interfaced with Upchurch Nanoport fittings, the channel was first flushed with air after UV polymerization to ensure the removal of porogen. The channel was then rinsed with ethanol to ensure the solubilization and subsequent removal of the unreacted monomer. This protocol provides for large uniform macropores (1–3 µm) and good cross-sectional coverage of the sol-gel monolith in the microchannel for monoliths containing up to 20% epoxide content (Figure 4b–c).
Additional studies were conducted to further optimize to monolith structure and morphology. The effect of time that the pre-polymerized sol-gel solution was allowed to rest and further react in the PDMS microfluidic channel was investigated. Prior to UV irradiation, the solution was allowed to sit in the dark in the microfluidic channel for an additional fifteen minutes to two hours. Regarding macropore formation, it was found that an additional rest time between thirty minutes and one hour led to the most beneficial results. For rest times of less than fifteen minutes there was no noticeable change in macropore structure compared to without rest time. When the rest time was increased to two hours, the sol-gel network became too cross-linked, resulting in a dense gel structure.
The resulting photopolymerized sol-gel monoliths will be used for chip based separations. In order to have the most efficient separations, it is necessary that fluid flow through the monolithic network is uniform. Non-uniformities in fluid flow will result if the monolith is not properly anchored to the PDMS walls. Previous studies commonly use MPTMOS as an anchor to capillary or chip walls since it can form siloxane bonds with free silanol groups present on glass or plasma treated PDMS. However, in the current study it was important to introduce an additional means of wall attachment. To this end, photoinitiator was adsorbed to the channel walls before introducing the pre-polymerized sol-gel solution into the channel. This allows for a high concentration of photoinitiator at the walls in comparison to the channel bulk. Flow of fluorescent analytes was visualized in the resulting monoliths and it was observed that fluid flow bypassing due to improper wall attachment is minimized. However, it should be noted that local higher concentration of photoinitiator at the wall may lead to potential inhomogeneity of the resulting monolith.
Silica monoliths are known to have a bimodal pore size distribution. Both nano-sized mesopores and micron-sized macropores were characterized. Differential scanning calorimetry (DSC) has been used as an alternative to BET nitrogen adsorption and mercury porosimetry to determine pore sizes of materials[41–43]. DSC is advantageous because samples are measured in the wet state. Samples are saturated with water and the difference between the melting (or freezing) temperature of bulk water and water contained within the pores correlates to the average pore size. Silica particles with known pore sizes were used to generate a calibration curve. Mesopore size of representative monoliths is summarized in table 1. It can be noted that when the monolith containing 20% epoxide is functionalized with boronic acid, the decrease in the average pore size is comparable to the size of the aminophenylboronic acid molecule.
Micron-sized macropores were characterized by SEM visualization and subsequent image analysis of the micrographs. Representative macropore size distribution histograms are shown in figure 5 for monolith formed in the microfluidic chip containing 20% epoxide content with and without immobilized boronic acid. Histograms are generated using pooled data from numerous SEM images of the same monolith chemistry. As expected, the macropore size distributions with and without the presence of boronic acid are qualitatively similar since the majority of the surface area of silica based monoliths is from the mesopores. Pore peaks are observed at 0.16 and 0.50 microns, with macropores present up to 2.5 microns in size. The macropore size distribution indicates that both pressure and electrically driven flow can readily occur.
Boronic acid can be used as an affinity ligand that binds carbohydrates and other cis-diol containing compounds. In mildly basic solutions, boronic acid is hydroxylated and becomes a tetrahedral boronate anion which can then undergo esterification and form a reversible covalent bond with molecules containing cis-diols. This bond is broken by decreasing the pH and/or adding competing cis-diol compounds (i.e. 100–500 mM sorbitol). Previous work by Potter et al. has shown the ability of immobilized boronic acid ligands to separate ribonucleotides from dexoyribonucleotides. Additionally, Xu et al. immobilized aminophenylboronic acid to highly ordered mesoporous silica and demonstrated solid phase extraction of glycopeptides from non-glycopeptides in a tryptic digest of horseradish peroxidase.
In the current work, sol-gel monoliths functionalized with boronic acid were evaluated for their ability to bind to cis-diol containing compounds. The small molecules catechol and quinol, which differ only in the location of attached hydroxyl groups, were used in batch studies to determine the selectivity of the monolith. As shown in figure 6, monoliths not functionalized with boronic acid did not show any selectivity toward catechol. However, when monolith containing 20% epoxide is functionalized with boronic acid, selectivity of catechol is increased five times. Further, the amount of catechol that can be eluted to the monolith corresponds to approximately 25% of the available epoxide groups present. Mass transfer limitations present during both the chemical attachment of boronic acid to the sol-gel matrix as well as during the binding and elution of the molecules from the monolith will affect the capacity of the material.
Proof-of-concept solid phase extraction of proteins was demonstrated on the boronic acid modified monolith in a PDMS microfluidic chip. Conalbumin labeled with Alexa Fluor 488 was used as a model glycoprotein for chip extraction experiments. A step injection of conalbumin was loaded onto the microfluidic chip under electrically driven flow (110 V/cm). After fluorescence was visualized in the column, a basic wash step was performed, followed by an acidic elution step. Average fluorescent values are extracted from the images using Image J and plotted. As shown in figure 7, additional conalbumin only eluted from the column under acidic conditions when the monolith was functionalized with boronic acid. When the monolith was not functionalized with boronic acid, all of the conalbumin was removed during the wash step. This demonstrates that photopolymerized sol-gel monoliths can be functionalized with boronic acid and used for the extraction of glycoproteins. However, the relatively low capacity of the monolith for the glycoprotein can be attributed to the mesopore size whereby steric limitations will prevent large proteins from entering many of the pores. To this end, ongoing research is being carried out to evaluate alternative ligand functionalization techniques and use these monoliths for the separations of glycopeptides from a complex protein digest.
Photopolymerized silica sol-gel monoliths have been developed for use in PDMS microfluidic devices. After selection of a suitable porogen, ethylene glycol diglycidyl ether, the monolith chemistry and morphology was optimized. In order to have good cross-sectional filling of the monolith in the microfluidic channel, it was found that 28% monomer to 72% porogen was optimal. Additionally, a channel aspect ratio on the order of 2.5 : 1 was superior to 5 : 1. In order to incorporate appreciable amounts of the epoxide containing silane monomer, GPTMOS, calcium chloride was added as a phase separation additive. Furthermore, the channel was interfaced with a constant pressure device to enable rinsing with air and ethanol to generate the macropore network. Macropore characterization was carried out with SEM and image analysis and mesopore characterization was determined using differential scanning calorimetry. Finally, monoliths were functionalized with boronic acid ligands to demonstrate binding to cis-diol containing compounds. Ongoing work is focusing on alternative functionalization techniques and employing the monolith chips for separations of glycopeptides from a protein digest.
This work was supported by NSF grant CBET 0522656