Microfluidic Transfection for High-Throughput Mammalian Protein Expression

Mammalian synthetic biology and cell biology would greatly benefit from improved methods for highly parallel transfection, culturing and interrogation of mammalian cells. Transfection is routinely performed on high-throughput microarrays, but this setup requires manual cell culturing and precludes precise control over the cell environment. As an alternative, microfluidic transfection devices streamline cell loading and culturing. Up to 280 transfections can be implemented on the chip at high efficiency. The culturing environment is tightly regulated and chambers physically separate the transfection reactions, preventing cross-contamination. Unlike typical biological assays that rely on end-point measurements, the microfluidic chip can be integrated with high-content imaging, enabling the evaluation of cellular behavior and protein expression dynamics over time.

Chemical transfection can be achieved with widely used reagents such as cationic lipids, which neutralize the negative charge of DNA and facilitate entry into the cell by endocytosis. The integration of lipid-based transfection with contact spotting has enabled high-throughput transfection. In this technique, called reverse transfection, purified cDNA samples are mixed with transfection reagent and spotted onto glass slides [2]. The arrays are next seeded with cells, which undergo transfection in situ, resulting in exogenous gene expression. Unlike protein microarrays [3][4][5], this method does not require individual purification of each sample and proteins can be analyzed in the natural cellular context. More than 5,000 samples can be printed on a single glass microscope slide using standard techniques [2,6,7]. This throughput enables massively parallel characterization of complex synthetic networks and the screening of genome-wide RNAi and cDNA libraries [8,9].
Although reverse transfection has been optimized for a variety of genetic materials [10,11] and cell types [12], methods for cell manipulation on the arrays still stand to be improved. Cell seeding and culturing is performed manually, and crosscontamination is a concern because spots on the array are not physically separated from one another. Standard reverse transfection offers poor control over the cell microenvironment and cannot support sophisticated downstream experiments.
Microfluidics could compensate for these shortcomings by providing a means to enclose each position on the array in a cell culture chamber. Microfluidic devices are readily fabricated using standard photolithography and soft lithography techniques [13] and contain micromechanical components such as valves that execute complex fluidic manipulations. These devices are characterized by high throughput, automation, small sample requirements, and compatibility with other analytical techniques. Standard microfluidic chips designed for mammalian cells are capable of a variety of functions, including long-term perfusion culture and the ability to individually address cell chambers [14].
This protocol describes the high-throughput transfection of mammalian cells on a microfluidic chip [15]. The device consists of two components: a glass slide patterned with transfection reagent, and a microfluidic device with chambers that support cell culturing (Fig. 1). The 280 microfluidic chambers have a capacity of ~600 cells each. The chambers of the chip are aligned to the transfection array so that each chamber contains a unique transfection mixture. The arraying of DNAcontaining transfection mixtures is coupled to the arraying of poly-L-lysine (PLL), which serves to anchor the DNA to the glass slide. This process has been optimized to enable highly efficient transfection with low cross-contamination. The chip is operated and cultured in a microscopy setup, facilitating the interrogation of protein expression dynamics and synthetic gene network performance. • Heidelberg VPG200 laser lithography system (Heidelberg Instruments Mikrotechnik GmbH).
• EVG150 coater and developer system for positive resist (EV Group).
• Glass slide rack and dish.

PLL and transfection mixture preparation
• NaOH.

Microscope setup
• Nikon Ti-E Eclipse automated epi-fluorescence microscope. • Cube air heater and incubation chamber to enclose the microscope (Life Imaging Services).
• NIS Elements, Fiji, and MATLAB imaging and image processing software.

Methods overview
High-throughput microfluidic transfection incorporates several techniques including photolithography, soft lithography, microarraying, operation of a microfluidic setup, and cell culture (Fig. 1) The assembled device is connected to a microfluidic setup. A suspension of cells is loaded into the channels, and valves are actuated to segregate individual chambers.
Medium is perfused through channels running parallel to the cell chambers, eliminating sheer stress during culturing. Diffusion from the channels introduces nutrients into and eliminates waste from the chambers. The cell chip is cultured directly on a microscope stage encased in an incubation chamber, enabling manipulation and monitoring of transfection over several days.

Mask and wafer fabrication
All mask and wafer fabrication steps are performed in a class 100 clean room. For researchers without access to a clean room or PDMS fabrication facilities, many commercial mold and device fabrication services are available. The design files of the transfection device are available for download from zenodo.org (https://zenodo.org/record/820777#.WVSrbhPfrOQ).

Chip design
1. The microfluidic device is designed using CAD software. The chip measures 1.6 x 5.8 cm and contains 280 culturing and transfection chambers (Fig. 2a).
Two molds should be designed: one for the control layer (30 µm height), which contains the valves, and another for the flow layer (30 µm height), which contains the channels and chambers necessary for reagent introduction and cell culturing (Fig. 2b,c, Note 3).
2. The flow layer mold contains two patterns (SU-8 and AZ9260) on the same wafer, so alignment marks are added to the design to enable alignment of the layers during the exposure step.
3. The control layer is scaled by 101.5% to account for PDMS shrinkage during curing.

Mask fabrication
1. A Heidelberg VPG200 laser lithography system with a 20 mm writing head and a UV-light source (i-line λ = 355 nm) is used to write each layer of photoresist on a separate mask (2 masks for the flow layer, 1 for the control layer).
2. Masks are developed using the DV10 instrument. The development dispenser is purged before the mask is placed inside the chamber and developer (a 1:5 MP 351:DI water mixture) is applied twice (15 s application followed by 45 s agitation), followed by rinsing and drying.
3. The chrome layer of the mask is etched in a perchloric acid solution for 120 s, then washed with water (quick dump rinse followed by an ultra pure water bath) and dried. The remaining photoresist is removed using TechniStrip 1316 (manual application to mask followed by complete immersion in bath for 10 min). The mask is washed with water (quick dump rinse followed by an ultra pure water bath) and air dried. 10. The features are measured using a Dektak XT surface profilometer.

Flow layer wafer fabrication
1. The flow layer is patterned with SU-8 (for the channel features) as described above. For the valve features, AZ9260 is next patterned on the same wafers as described below.  inks, the pin is washed with water for 500 ms and dried for 500 ms. Humidity in the spotter is set to 50%.

Orientation marks are placed on the slides (see Note 11) and 2 h after
arraying is complete, PLL arrays are washed 5x 30 s with Milli-Q water, dried using a nitrogen air gun, and stored in a desiccator. PLL slides should be used between 2 and 8 weeks post-coating, since the quality of the PLL has been shown to deteriorate beyond this time period [20].  Immediately afterwards, the functionalized surfaces are aligned (one-shot) using a stereomicroscope (see Note 16). Slight pressure can be applied to ensure that all parts of the chip are in contact with the glass slide (see Note 17). The assembly is baked for 1 h at 80°C. 4. The transfection devices are stored in a desiccator.

Microfluidic setup and operation
The microfluidic setup comprises pressurized manifolds for flow and control layer manipulation, a fluorescence microscope, an incubation chamber, and a camera (Fig. 3).  2. For large volume or long-term flow connections, PTFE tubing is used (see Note 19). A bottle containing filtered CO 2 -independent medium supplemented with 10% FBS, antibiotic-antimycotic, and glutamax is attached to the microfluidic setup (see Note 20). A piece of tubing is placed inside the bottle, running from top to bottom. The tube is connected to a pin that punctures the septa of the cap (this entire assembly is autoclaved prior to use). On the other side of the septa, the pin is connected to a longer piece of tubing that connects to the chip. Finally, a second piece of tubing is attached to a manifold at one end and punctures the septa on the other end. This setup pressurizes the bottle to drive flow into the chip (Fig. 3). 2. For the first hour after loading, medium is pulse perfused [21] to promote attachment and prevent cells from being washed out of the chambers (Fig. 4b, Note 25). Medium is flowed for 5 min through the channels with access to the chambers closed. For the next 5 min, the flow of medium is stopped and the valve joining the chambers and medium channels are opened.

Microfluidic operation
3. After 1 h of pulse perfusion, the top/bottom chamber valves are opened and medium is continuously flowed at a rate of ~1 ml/h (see Note 26).

To measure transfection efficiency, images are acquired 48 h (see Note 27)
after introducing cells onto the chip, using 20x magnification to capture an entire chamber in the field of view (Fig. 5a). Images can also be acquired continuously after cell loading to monitor the progression of protein expression. Chambers are imaged in fluorescence and brightfield modes to count both the number of fluorescent cells and the number of total cells.

Images can be stitched together (for example using the Grid/Collection
Stitching plugin in Fiji [22]) to reconstruct the entire array (Fig. 5b). Cells can be counted using programs such as CellProfiler or Matlab.  4. If problems with AZ9260 adhesion to the wafer are experienced, an HMDS bake may be performed prior to dehydration and coating. 5. Multiple layers of foil may be necessary to prevent holes. The foil prevents direct contact and bonding between the PDMS poured on the wafer and the glass dish. 6. To fabricate functional microfluidic chips, it is essential to work quickly, protect the exposed flow layer, and keep wafers and chips free of dust. Chip fabrication can alternatively be performed in a clean room environment.
Wafers should never be cleaned with water or organic solvents. 7. Wafer-containing glass petri dishes should not be stacked in the oven, since this results in sub-optimal heat distribution. 14. In the case of limited DNA samples, the lipid-DNA recipe can also be halved. 15. Depositing lipid-DNA multiple times per spot does not increase transfection efficiency, and as has been previously reported [17] results in imprecisely localized spots due to spreading of excess DNA. 19. PTFE tubing is used to supply long-term culturing medium because cell toxicity may be observed when using Tygon tubing, as has been previously reported [23]. 20. The medium bottle should be carefully handled since any frothing may introduce bubbles onto the chip.