Cells were inspected using wide-field inverted microscope DM IRE 2 with mechanical stage, integrated into the AS TP system (Leica Microsystems GmbH, Wetzlar, Germany), in either a bright field mode using the Leica modulation contrast or in a GFP fluorescent mode using the I3 filter cube (excitation filter BP 450/490, dichromatic mirror 510, suppression filter LP 515). Images were obtained using either Coolpix 4500 photo camera (Nikon Russia, Moscow, Russia) or Cascade II 512 video camera (Photometrics, Tucson, AZ, USA). Optical sections and side views of fluorescing cells and microcolonies in the GFP fluorescent mode were obtained using confocal microscope TCS SPE DM2500 and LAS AF Version 2.1.0 build 4316 software (Leica). Where indicated, non-GFP–producing cells were stained with a broad range fluorescent ink by spotting the side of the coverslip facing the gel with a CD/DVD/BD marker (LINER 2616; Centropen, Prague, Czech Republic; www.centropen.cz/product-catalogue/for-the-office/12-3-cddvdbd-markers/38-cddvdbd-liner-2616).

Microcolonies were grown in inverted merged gels whose agarose layer contained an appropriate complete growth medium with or without 0.5 mg/mL bovine skin collagen (cat. no. C4243; Sigma-Aldrich). The gels were incubated upside down in a 40-mm Petri dish at 37°C and 5% CO₂ under 3 mL growth medium. Microcolonies were extracted by sucking with a disposable 70-m inner diameter glass micropipet made from a borosilicate glass capillary tubing (cat. no. BF100-50-10;Sutter Instrument Company, Novato, CA, USA) using P-97 MicropipetPuller (Sutter Instrument). The micropipet was operated with the CellTram Air microinjector (Eppendorf Austria GmbH, Vienna, Austria) mounted on the TransferMan NK 2 micromanipulator (Eppendorf) integrated into the Leica AS TP system.

Results and discussion

We prepared 2-D cell arrays using merged PAA/agarose gels (18, 19). To this end, a suspension of cells in molten low-gelling temperature agarose was poured in a shallow well made in a microscope slide having a previously cast and dried polyacrylamide gel covalently attached. When the PAA gel swells by absorbing the liquid, it displaces cells toward the coverslip. Upon formation, the agarose gel becomes partly merged with the PAA gel and immobilizes cells concentrated on the PAA gel surface. Sometimes, especially for the experiments on cell growth and cloning, we used inverted merged gels. In this case, the dry PAA matrix was covalently attached to a glass slip covering a well filled with molten agarose containing cells.

To become immobilized, cells must be submerged in the agarose gel at some depth; hence, the agarose must harden before it is completely sucked up by the PAA matrix. However, in this case, cells are not layered into a monolayer; rather, they populate the entire space above the PAA surface (Figure 1A). We found that a nearly perfect monolayer can be generated by briefly spinning the slide in a centrifuge while the agarose is still liquid; this forces all the cells to sediment on the swelling PAA bed. When the PAA gel was made of 7% acrylamide and 0.07% bisacrylamide, the cells were submerged at a depth of 10 m (Figure 1B). Decreasing the acrylamide concentration to 4% and increasing the bisacrylamide/acrylamide ratio to 1/20 increased the depth to ╛25 and ╛20 m, respectively (Supplementary Figure S1). Apparently, the lower percentage of acrylamide makes the PAA gel more compressible during centrifugation, whereas the higher percentage of the cross-linker reduces the rate of PAA swelling. The resulting 2-D arrangement allows all the cells within a microscopic field to be observed simultaneously (Figure 1C).

Figure 1. HeLa cells in merged gels. (Click to enlarge)

The merged gels can produce unordered 2-D cell arrays of a high density, which allows the cells to be screened at high-throughput using a manually operated microscope. Figure 2A displays a fragment of a monolayer of ╛200,000 HEK-293 cells transfected by a GFP-encoding plasmid and arrayed in a 14-mm-diameter gel. This corresponds to the surface density of ╛1300 cells/mm2. Figure 2B shows that ╛40% of the cells seen in Figure 2A fluoresce. To estimate the screening throughput, we prepared a 2-D array of the same density upon a 10,000-fold dilution of the cells with nontransfected HEK-293 cells (Figure 2C). Eight GFP-producing cells were detected by eye in less than 30 min using a 40× objective, which corresponds to the screening speed of >100 cells/s. Although not as rapid as with an automated flow cytometer, the screening includes the option of image analysis, thereby increasing the reliability of detection and allowing cells to be distinguished from each other and from noncellular particles. Moreover, as each cell has a unique address in the array, the cells of interest can be subjected to a detailed analysis subsequent to the initial fast screening. The screening could be made faster and producing more information by using currently available automated instruments and software (20, 21).