High-Density Self-Contained Microfluidic KOALA Kits for Use by Everyone

Device Fabrication and Preparation

The lids and bases are fabricated by CNC milling (PCNC 770, Tormach, Waunakee, WI) with standard microscope slide dimensions (25 mm × 75 mm). Bases are composed of three layers: (1) a post layer, (2) a channel layer, and (3) a cover layer ( Fig. 2C ; Suppl. Fig. S1). (1) The post layer is milled from 2 mm thick polystyrene (PS; #ST313200, Goodfellow, Coraopolis, PA) and encompasses an inlet and an outlet post for each channel. Both posts have a horseshoe-like cross section with an outside diameter of 1 mm. The inlet post for each channel is 1.5 mm tall with a slot width of 0.8 mm, and the outlet post is 1 mm tall with a slot width of 0.5 mm. Lengthwise, the spacing of the posts matches that of the channels; however, widthwise, the posts and channels are staggered by 3 mm to facilitate close packing. (2) The channel layer is milled from 1.2 mm thick PS (#ST313120, Goodfellow) and encompasses 60 microchannels spaced by 1 mm (from channel centers). Each channel is 0.5 mm wide, 2 mm long, and 0.3 mm deep. In the center of each channel is a microwell that is 0.5 mm wide, 4 mm long, and 1.2 mm deep, with bevels at both ends to facilitate fluid flow. (3) The cover layer is 25 mm ×75 mm and is cut from 0.190 thick PS (#ST313190, Goodfellow). Prior to assembly, all three layers are treated with oxygen plasma (Femto, Thierry Corp., Royal Oak, MI) for 50 s at 50 W. Layers are assembled by acetonitrile bonding. ⁸ In brief, acetonitrile (#271004, Sigma Aldrich, St. Louis, MO) is applied dropwise to one layer of the device, which is then promptly mated with a second layer. The excess acetonitrile is aspirated from the channels, and the mated pair is gently pressed together on a hot plate preheated to 70 °C. After 30 s, the process is repeated for the third layer. After bonding, the device is plasma treated again with the same settings, and hydrophobic recovery is induced between the posts using a wiper. ⁹

Lids are milled from 2 mm PS. Each lid contains a microwell for each channel (i.e., 60 microwells) and a slot for absorbent pads. Each microwell has an elliptical shape, which is 1.5 mm wide, 4 mm long, and 2 mm deep and is spaced to match the pattern of the input posts. The slot for the absorbent pad is 10 mm wide, 72 mm long, and 2 mm deep. The lids are assembled by first applying an optical adhesive cover (#04 729 757 001, Roche, Basel, Switzerland) to the back of the lid and then placing two layers of absorbent pad (#CFSP223000, Millipore, Billerica, MA) in the designated slot. Both bases and lids are ultraviolet sterilized for 30 min, prior to cell culture, and lids are loaded with reagents prior to each application.

Fluid Handling

Fluid exchanges are characterized by measuring (1) duration, (2) washing efficacy, and (3) cross-contamination during fluid replacements. Fluid replacements are initiated by mating the lid with the base, thereby inserting the series of posts into the microwells. (1) To assess time required for fluid replacement, the microwells are filled with 15 µL of deionized (DI) water, and the lid is clicked on the base. Time is measured from the point of initial fluid contact between the post and the fluid to the point at which the microwell is emptied. (2) Washing efficacy is assessed by measuring the depletion of a fluorescent dye after individual wash steps. Channels are preloaded with a fluorescent dye (Alexa Fluor 488, #D-22910; Life Technologies, Carlsbad, CA) and then washed by applying lids containing 15 µL of DI water per microwell. Channels are imaged after each wash is completed (IX-70; Olympus, Shinjuku, Tokyo, Japan), and fluorescence intensities are measured using ImageJ. The background autofluorescence, inherent to the PS, is measured, averaged (n = 8 locations), and subtracted from the channel fluorescence. After subtraction, all channels are normalized to the average of the starting fluorescence intensity and averaged (n = 60 channels; error bars represent 1 standard deviation). (3) To assess cross-contamination, alternating microwells are filled with 15 µL of either Alexa Fluor 488 or Texas Red (#D-3328; Life Technologies). After completing the fluid exchange, each channel is imaged for both dyes, and fluorescence intensity is quantified using ImageJ. Autofluorescence noise is subtracted from each channel for both wavelengths.

Cell Culture for IHC Assay

In this section, all fluid transfers, including cell seeding, are executed by filling each of the 60 microwells with 15 µL of the respective reagent and then mating the lid with the base. NMuMG cancerous epithelial cells are seeded in KOALA-HD channels at a surface density of 400 cells/mm² and cultured at 37 °C for a total of 72 h. For the first 24 h, cells are cultured in DMEM media with 4.5 g/L glucose (#10-017-CV; Corning Cellgro, Manassas, VA) supplemented with 10 µg/mL insulin and 10% fetal bovine serum. For the next 48 h, the cells are cultured in the same media supplemented with a 0, 2, 20, or 200 pM concentration of transforming growth factor β1 (TGF-β1; #100-21; Pepro Tech, Rocky Hill, NJ). The media are replaced every 24 h. Cells are washed with 1× phosphate-buffered saline (PBS), fixed with 4% paraformaldehyde (#43368; Alfa Aesar, Ward Hill, MA) for 15 min, and permeabilized with 0.1% Triton X-100 (#807426; MP Biomedicals, Santa Ana, CA) for 30 min. Channels are washed with 1× PBS and then blocked with 1× PBS supplemented with 3% bovine serum albumin (PBS+BSA) for 30 min. Primary antibody solutions are prepared by separately diluting anti-β-catenin (rabbit) to 2 µg/mL, anti-E-cadherin (mouse; #610182; BD Biosciences, San Jose, CA) at 1 µg/mL, anti-FAK (rabbit; #ab40794; Abcam, Cambridge, England, UK) to 2 µg/mL, anti-N-cadherin (rabbit; #ab18203; Abcam) to 6 µg/mL, or anti-vimentin (mouse) at 1:50 into PBS+BSA supplemented with 0.1% Tween 20 (PBST+BSA; #P1379; Sigma Aldrich). Primary antibody solutions are added to appropriate channels (with n = 3 for each concentration of TGF-β1) for 1.5 h and then washed twice with PBS+BSA. Secondary antibody solutions are prepared by diluting both Alexa 488 goat anti-mouse (#A-11029; Life Technologies) and Alexa 568 goat anti-rabbit (#A-11036; Life Technologies) to 4 µg/mL into a single solution of PBST+BSA, which is applied to all channels for 1 h. A Hoechst 33342 nuclear stain is applied to all channels for 15 min and then washed twice with PBS+BSA. Bases are imaged on an inverted fluorescent microscope (Eclipse Ti, Nikon, Tokyo, Japan) in three wavelengths: 345, 488, and 572 nm. Channel images are taken at 4× magnification and stitched automatically (NIS-Elements Ar, Nikon), and high-magnification images are taken at 10× magnification and overlaid using ImageJ. ImageJ is also used to measure the fluorescence intensity of the channels. Background fluorescence is taken from surrounding regions, averaged, and subtracted from the fluorescence values of the channels. Within each staining category (i.e., β-catenin, E-cadherin, etc.), the channels are normalized to the maximum intensity of their respective category and averaged (n = 3). Error bars represent one standard deviation.

Fluid Exchanges in KOALA-HD

All fluid-handling mechanisms necessary to operate KOALA are integrated into the KOALA lids and bases. Each lid is a one-time-use component composed of 60 microwells, which houses reagents and an absorbent pad to drive fluid flow ( Fig. 2C ). Similarly, each base is composed of 60 microfluidic channels; however, unlike the lids, the base is used for the duration of the assay. Each fluid replacement—performed by clicking a lid onto the base—is completed for all 60 channels within 15 s (see Supplemental Video). By measuring the average time required to empty each well, we find that the average flow rate through the channel is ~7 µL/s. Standard microfluidic techniques to produce alternative channel designs or absorbent pad compositions can be used to achieve different flow rates for other applications if needed. When the lid is applied to the base, fluid flows from the microwell (of the lid), through the channel (of the base), and into the absorbent pad (of the lid), thus completing a fluid exchange. After this, the lid is disposed of and the base is incubated until the next fluid replacement. The novel design of the base leverages horseshoe-shaped posts to enable fluid exchanges through several functionalities. ⁶ The posts provide a method of connecting the base to the lid to prevent air bubbles from being trapped in the channels while ensuring proper volumes during fluid exchanges. Posts are located at the channel inlet and outlet, which, when a lid is applied, make contact with the fluid in the microwell and the absorbent pad, respectively. Fluid flows through the channel initially driven by capillary forces in the channel and subsequently by wicking in the absorbent pad. When the microwell is emptied, the fluid pins in the bottom of the inlet post, and the outlet post severs the fluid flow by allowing a bubble to form within the horseshoe, thus disconnecting the absorbent pad from the channel. This process completes the fluid exchange.

To make KOALA suitable for HD applications, however, several obstacles must be overcome, including (1) determining a configuration suitable for densification, (2) eliminating cross-contamination, and (3) achieving complete and defined fluid exchanges in the new HD format.

To determine a setup that best supports the densification, we test two designs: a format that uses individual posts as described above and a common rail configuration (Suppl. Fig. S2). The lid for the common rail design comprises a single trench surrounded by half elliptical microwells, whereas the base comprises a single rail with several protruding fins. When the lid is applied to the base, the fins provide the conduit for wicking fluid from the lid into the channel. We found that a design with individual posts for each channel is superior for two reasons. First, the separation between the inlet posts prevents cross-contamination via Concus-Finn filaments. More explicitly, the corners between vertical and horizontal features act as a conduit for fluids, which leads to cross-contamination. Second, the small-diameter inlet and outlet posts prevent excessive displacement of fluid in the microwells and support improved contact with the absorbent pad.

To eliminate cross-contamination, hydrophilicity between the posts and fluid displacement in the microwells is decreased. After plasma treatment, we implement a novel approach, described by Guckenberger et al. ⁹ to recover surface hydrophobicity between the posts, thus mitigating hydrophilicity-based cross-contamination. Inadequate contact between the outlet posts and the absorbent pad lead to cross-contamination via rapid displacement of the fluid in the microwells. This issue is circumvented by ensuring proper connection between the outlet posts and the absorbent pad prior to inserting the inlet posts into the microwells. In the future, a small, reusable device for repeatably aligning and mating the lid and base (i.e., a jig) will be designed to facilitate this connection and prevent the possibility of user error.

To ensure complete and defined fluid exchanges for each channel, we leverage the designs of the microwells, the posts, and the absorbent pads. The volume allotted for each fluid exchange is dictated by the volume of the microwell. However, maximizing use of the fluid in the microwell, and thus minimizing dead volume, depends on the orientation of the inlet post and the size of the absorbent pad. We found that to minimize dead volume, the open side of the horseshoe should be directed toward the end of the microwell (Suppl. Fig. S3). In doing so, the dead volume of each microwell is reduced to <3 µL. Further, to ensure that all of the microwells are emptied entirely, the absorbent pad has to be sufficiently large to assure that it does not saturate during the fluid exchange. Put together, these solutions make KOALA-HD fluid exchanges repeatable and robust.

Evaluating KOALA-HD Function

To make a highly functional platform to be used for toxicology and cell-based studies, we assess the platform for its ability to (1) ensure proper fluid replacement in the channels; (2) ensure no cross-contamination between adjacent channels, a difficulty particularly compounded by higher throughput; and (3) perform several fluid exchanges without failure.

We quantify the depletion of a fluorescent dye after washing steps to assess the efficiency of fluid replacements ( Fig. 3A ). To fulfill this, a lid filled with fluorescent dye is applied to the channels, followed by two washing lids. We quantify the fluorescent depletion after each wash and normalize it to the starting intensity. We found that one washing lid is capable of removing ~60% of the initial fluid, and a second washing lid reduced the dye concentration by more than 98% (n = 60), which, based on the volumes of the channels and fluid exchanges used here, is comparable to washing of traditional microfluidic channels. ¹⁰

OPEN IN VIEWER

Figure 3.

(A) Washing efficiency of lids is quantified by measuring a decrease in fluorescent intensity after each lid (n = 60). (B) Cross-contamination is assessed by filling a lid with two different fluorescent dyes (depicted by food coloring) and applying that lid to a base. Fluorescent intensities for both red and green are measured and represented for each channel. No cross-contamination is evident.

Cross-contamination is assessed by quantifying fluorescence in each microchannel after applying a lid prefilled with two fluorescent dyes ( Fig. 3B ). The lid is strategically filled such that each microwell containing green dye is directly adjacent to four microwells with red dye (illustrated with food coloring), enabling detection of cross-contamination from both horizontally and diagonally adjacent microwells. This lid is applied to a base, and each channel in the base is quantified for both red and green fluorescence ( Fig. 3B , bottom). The results reveal no cross-contamination between channels based on no increase in fluorescence intensity between channels (n = 1), which is confirmed by observing no fluidic connection between the adjacent posts or microwells. Notably, these data are collected using high-surface tension fluid (water), but for low surface tension fluids (e.g., oil, high-surfactant reagents), the lid and base may be inverted. This result sets the stage for using the KOALA-HD in toxicological screens or pharmaceutical identification of natural products, including readouts such as protein expression.

Finally, we assess the ability to operate a fluidic assay without failure by applying a series of lids and verifying complete flow in each channel. A series of 20 lids are prefilled with dyed water and applied to a single base. For comparison, about 15 fluid replacements are typically required for a traditional immunochemistry staining. Each lid is filled with alternating colors to make channel failure obvious. All 20 fluid exchanges are completed without a single channel failure.

Channel Design for Cell Culture

To ensure suitability for cell culture, KOALA-HD is designed to (1) allow precise readouts of cellular functions (proliferation, protein localization, etc.), (2) contain enough media to maintain cell viability throughout the incubation period, and (3) be negligibly affected by evaporation. The channels are designed with a lowered region in the center that serves three purposes. First, the decreased plastic thickness enables better cellular readouts (e.g., higher magnification microscopy, less autofluorescence from the plastic). Second, the increased channel size reduces the shear stress by slowing fluid velocity. Third, the lowered region increases the volume and thus increases the reservoir of nutrients that are available for the cells. With this channel design, we are able to culture cells for 24 h without media exchanges.^11,12 However, for more highly active or sensitive cells, more frequent fluid replacements may be necessary. Evaporation is mitigated using previously described and characterized methods.^13,14

IHC Assays

We demonstrate the ability to perform high-density IHC assays and complex staining processes reliably and without the need to prepare antibody dilutions or source-specific reagents. We chose an IHC assay based on its ubiquity in screening applications and for the multitude of steps necessary to complete the assay (i.e., cell culture, fixation, permeabilization, staining, washes, etc.). We use a relevant cancer-specific assay in which NMuMG epithelial cells are treated with various dosages of TGF-β1, inducing an epithelial-to-mesenchymal transition. ¹⁵ Markers of this transition include down-regulation and localization of β-catenin and E-cadherin to cell-cell contact areas ¹⁶ and increased expression of FAK, ¹⁷ N-cadherin, ¹⁸ and vimentin. ¹⁹ Such complex readouts require cell culture, treatment with a drug, and IHC staining. We demonstrate cell culture in KOALA-HD by culturing NMuMG cells in 60 channels for 72 h. The process requires 15 sequential fluidic steps, all performed using the KOALA platform (i.e., cell seeding, media replacements, TGF-β1 dose treatments, fixing solution, washing buffer, primary antibody staining, and secondary antibody staining). The platform is able to perform each fluidic step in all channels simultaneously without failure or cross-contamination. Each fluidic exchange requires about 15 s with a simple mating of the required lid with the base containing cells.

Cells are treated in four groups of 15 channels with varying concentrations of TGF-β1, after which cells are fixed, permeabilized, stained with a nuclear stain, and stained in triplicate for β-catenin, E-cadherin, FAK, N-cadherin, and vimentin ( Fig. 4A ). We found that β-catenin decreases and N-cadherin increases with increasing levels of TGF-β1, as expected. We did not observe any decrease of E-cadherin for increasing levels of TGF-β1, but we did observe localization to the cell-cell contact areas. FAK decreases with increasing levels of TGF-β1, which is inverted from typical results but has been observed in NMuMG cells expressing D119A-β₃ integrin. ¹⁷ Finally, prevalence of vimentin, a mesenchymal marker, is evident in cells treated with TGF-β1, indicating a transition to a mesenchymal phenotype. Selected channels treated with the highest concentration of TGF-β1 (i.e., 200 pM) are imaged at high magnification ( Fig. 4B ). These images demonstrate localization of β-catenin and E-cadherin to the cell-cell contact areas. A noticeable feature in some of the images is the striated pattern formed by the cells arranging themselves along fabrication artifacts. These are the result of machining marks left by the CNC machine during the channel fabrication and can be removed by using other fabrication methods such as injection molding. All channels are stained with both the Alexa 488 and Alexa 568 dyes to detect nonspecific binding. Importantly, we demonstrate that KOALA-HD can—in the footprint of a glass slide—cover a broad range of treatments and readouts, providing high-content microscopy images at a high-throughput scale, without a need for automation or trained personnel. In future applications, this platform could be imaged with a scanner for higher-throughput analysis.^12,20

OPEN IN VIEWER

Figure 4.

(A) NMuMG cells are cultured in KOALA-HD channels and treated in 15-channel segments with varying concentrations of TGF-β1. Within each segment, channels are stained with primary/secondary antibodies for (a) β-catenin/Alexa-568, (b) E-cadherin/Alexa-488, (c) FAK/Alexa-568, (d) N-cadherin/Alexa-568, and (e) vimentin/Alexa-488 (Hoechst not shown). The bar graph demonstrates the average fluorescence intensities for each of the treatment conditions and antibody stains (n = 3). Error bars represent standard deviation. Analysis of variance indicated significant differences within β-catenin data (p < 0.01). p Values are indicated for protected Fisher’s least significant difference test as follows: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. (B) High-magnification images of each antibody stain taken from channels treated with 200 pM concentration of TGF-β1. Each channel is imaged for Hoechst, Alexa 488, and Alexa 568.