Quantitative analysis of highly parallel transfection in cell microarr
http://www.100md.com
《核酸研究医学期刊》
Service de Génomique Fonctionnelle CEA, 2 Rue Gaston Crémieux CP22, 91057 Evry Cedex, France and 1 IMSTAR, 60 Rue Notre Dame des Champs, 75006 Paris, France
*To whom correspondence should be addressed. Tel: +33 1 60 87 34 74; Fax: +33 1 60 87 34 98; Email: xavier.gidrol@cea.fr
ABSTRACT
As more genomes are sequenced, we are facing the challenge of rapidly unraveling the functions of genes. To that end, cell microarrays have recently been described that transfect thousands of nucleic acids in parallel and can be used to analyze the phenotypic consequences of such perturbations. As many parameters can influence the efficacy of transfection in such a format, we describe some important features in manufacturing cell microarrays that may improve reliability and efficiency of both plasmid DNA and siRNA transfection. We have also developed image analysis software that allows automatic detection of cell clusters, quantification of transfection efficiency and levels of expression/extinction of genes. Along with cell microarrays, this bioinformatic tool should expedite functional exploration of the human genome.
INTRODUCTION
With the complete sequencing of the human genome, research priorities have shifted from the identification of genes to the elucidation of their function. High throughput technologies are a key feature of functional genomic experimentation. In the mid 1990s, DNA arrays made it possible to significantly increase throughput of gene expression analysis by simultaneously monitoring tens of thousands of genes (1–4). Before that technological development, biologists were studying gene expression of a few genes at a time by northern blots and RT–PCR. They are now able to monitor expression at the genomic scale and the entire human genome can be analyzed in a single array.
Similarly, methods to characterize gene function are also utilized, such as transgenic or knockout mice. They are based upon gain or loss of protein function and analysis of the resulting phenotypes to infer a potential role for the protein under scrutiny. In such approaches, DNA constructs that direct overexpression of a gene product or, on the contrary, eliminate its synthesis are introduced into the cell. Indeed, such phenotypic analysis gives a good idea of the potential function of the gene product. Until now, these methods were time consuming and only a few genes at a time could be analyzed. It was recently demonstrated that chemically synthesized short (<30 nt long) double-stranded siRNA (small interfering RNA) molecules, homologous to a target gene, could specifically inactivate gene function when introduced into the cell (5). RNA interference is a natural process for sequence-specific, post-transcriptional gene silencing initiated by double-stranded RNA (6,7). Thus, RNAi offers the possibility of high throughput ‘knockdown’ studies for the analysis of thousands of genes of unknown function (8,9).
To speed up the functional exploration of the human genome, there is a need for high throughput technologies allowing transfection of thousands of nucleic acids in parallel and the simultaneous analysis of thousands of resulting phenotypes. Ziauddin and Sabatini have described a cheap and flexible cell-based microarray system for the high throughput analysis of gene overexpression (10). Others have used this technology with siRNA (11,12), however, there remain several parameters that impact on the quality and reproducibility of transfection in such a cell microarray. In this report, we describe procedures and key features for manufacturing cell microarrays that generate reproducible and highly parallel transfection. Furthermore, to precisely quantify efficacy of transfection, level of expression or extinction of genes, image analysis software was also developed. This cell array format and automated image analysis system have the potential to be used in extensive analysis of gene function at the genome scale.
MATERIALS AND METHODS
Plasmid and small interfering RNA (siRNA)
The pEGFP-C1 plasmid expressing enhanced green fluorescent protein (EGFP) was obtained from Clontech (Paolo Alto, CA). Escherichia coli cells were transformed with pEGFP-C1 and plasmids were purified with a Midi Prep Qiagen Plasmid Kit (Qiagen, Hilden, Germany). Plasmid concentrations were assessed by UV absorbance; the OD 280/260 nm ratio was always >1.8. Synthetic siRNAs specific to lamin A/C (sense CUGGACUUCCAGAAGAACAdTdT, antisense UGUUCU UCUGGAAGUCCAGdTdT) (5) and EGFP, modified at their 3'-end with rhodamine (sense GCAAGCUGACCCUG AAGUUCAU, antisense GAACUUCAGGGUCAGCUUG CCG) (13) were purchased from Qiagen. For transfection, siRNAs were solubilized for 1 h at 37°C in a resuspension buffer (30 mM HEPES–KOH pH 7.4, 100 mM potassium acetate, 2 mM magnesium acetate) to a final concentration of 0.3 μg/μl.
Cell array printing
The general procedure was inspired by Ziauddin and Sabatini’s work (10) and was optimized to achieve better reproducibility of EGFP transfection. Five microliters of pEGFP-C1 at 0.1 μg/μl was diluted with 6.5 μl of EC buffer (Effectene kit; Qiagen). Two microliters of Enhancer supplemented with 1.2 μl of a 1.5 M sucrose solution and 2 μl of Effectene reagent was successively added to the mixture. After a 15 min incubation at room temperature, 12 μl of a 0.5% gelatin solution (Sigma G-1393 gelatin diluted in deionized water) was added and the solution was transferred to a 96-well plate for microarray printing. Regarding siRNA microarrays, the general procedure was the same except for slight modifications of the plasmids, siRNAs and formation of lipid complexes. One microliter of pEGFP-C1 at 0.6 μg/μl was mixed with 0.15–0.6 μg siRNA, specific to EGFP or lamin A/C. Mixtures were diluted in 11 μl of EC buffer supplemented with 2 μl of a 1.5 M sucrose solution, and 3.3 μl of Enhancer plus 3.3 μl of Effectene reagent was next added to the samples. After 15 min at room temperature, 6 μl of a 1% gelatin solution was added and samples were transferred to a 96-well plate for microarray printing. Poly(lysine)-coated (PolysineTM; Menzel-Gl?ser) microscope slides (25 x 75 x 1.0 mm) were printed with a Microgrid II Biorobotics (Cambridge, UK) arrayer at room temperature equipped with Biorobotics quill pins. The spots were 200 μm in diameter (verified by printing of fluorescent DNA) and were spotted with a 400–550 μm pitch. In most experiments, 600 features per slide were printed to facilitate the image analysis, but >5000 spots can be printed onto a single slide. Slides were stored at room temperature in a dessicator for several weeks.
Cell culture and reverse transfection
Human embryonic kidney cell line HEK293T and HEK293T cells stably expressing EGFP (kindly provided by Didier Poncet, INRA, Jouy-en-Josas, France) were grown at 37°C in a 5% CO2 humidified atmosphere in Dulbecco’s modified Eagle’s medium containing 4.5 g/l glucose supplemented with 10% fetal calf serum, 100 000 U/l penicillin, 50 mg/l streptomycin and 200 mM glutamine. Cells were seeded onto the microarray in 100 mm Petri dishes to reach confluence 48 h after transfection (100–150 000 cells/cm2 with HEK293T cells, 80 000 cells/cm2 with HaCaT cells and 60 000 cells/cm2 with HeLa cells) and cultured for 24–48 h at 37°C/5% CO2. Slides were fixed following a standard protocol (4% paraformaldehyde for 15 min at room temperature) and stored in phosphate-buffered saline at +4°C for several weeks.
Scanning and image capture
Cell microarrays were scanned using a ScanArray? 5000 scanner (Packard BioChip Technologies) and quantified using Genepix? software (Axon Instruments, CA). To measure EGFP fluorescence of transfected cells, images were acquired with a PathfinderTM OSA instrument (14) (IMSTAR, http://www.imstar.fr), an automated optical scanning system equipped with a high resolution (1300 x 1000 pixels) CCD camera with a dynamic range of 12 bits or 4096 levels of intensity per pixel, an adapted filter set and various automated devices. This instrument allows the registration of multiple fluorescence images of the entire cell microarray over a wide range of excitation/emission wavelengths and at a final resolution of 0.34 μm/pixel. Two processing steps were performed on the raw data coming from the camera: dark current subtraction and shading correction. The first consists of pixel-by-pixel subtraction of the image obtained in the absence of light from every captured image. The second consists of dividing the captured images by the image captured from an empty field with a statistically significant exposition time, the goal of this second step being to compensate for light excitation uniformity across the camera view field. When necessary, fluorescence microscopes were used and array images were captured and analyzed with either an Olympus IX70 or a Zeiss Axiovert microscope. Both have fluorescent light emission systems for EGFP detection (excitation 488 nm, emission 520 nm) and a digital camera for picture image acquisition.
Quantification of cell arrays
Fluorescent images were analyzed with a PathfinderTM OSA image analysis commercial software package. This software comprises an integrated image database management component that underlies all other functions from automated image acquisition through protocols of image processing and image segmentation to the final results tables and statistical analysis. The database structure is flexible to support any image-to-image relationship. It allows image browsing by displaying the results of a query on the database and lets the user recognize the image by its reduced resolution view. Furthermore, image processing macro functions, called protocols, can be applied to any number of images selected by a query on the database, using a single command. Finally, the results table, which accumulates the calculated parameters for all objects from all images from the query, can be displayed and analyzed. This allows complete analysis in record time, with minimal user interaction.
Total level of EGFP expression. The level of EGFP expression was estimated by measuring the mean EGFP fluorescence intensity corresponding to each cluster, after automatic recognition and contour detection. A digital mask containing six grids of 10 x 10 elements, each corresponding to the pattern of EGFP-expressing DNA constructs printed on the slide, were plotted from the acquired microarray image by software. The fluorescence mean intensity corresponding to each cluster was calculated within a circle of diameter Dspot = 300 μm in each square grid. The mean fluorescence intensity inside the circle was background corrected by subtracting the median fluorescence intensity of pixels located in the grid square but outside the circle. The values corresponding to each block are represented as the average fluorescence intensity of 64–100 spots.
Transfection efficiency. To estimate the EGFP plasmid transfection efficiency, the ratio of transfected cells (within the Dspot = 300 μm circle) to total cells was calculated. Detection of non-transfected cells was performed by Hoechst 33342 staining, followed by capture of either Hoechst 33342-specific fluorescence images or bright field microarray images. The images obtained had a resolution of 0.7 μm/pixel. The analysis was based upon superimposed images corresponding to Hoechst 33342 and EGFP fluorescence. The detection and segmentation of Hoechst 33342-stained individual cells was performed on a bright field image using PathfinderTM software where fluorescent spots, defined as cell clusters, are recognized as well as cell morphology in each spot. Basically, for the present study the segmentation procedure developed associates mathematical morphology algorithms and measurements of intensity, size and shape and consists of a multi-step protocol in order to generate a binary image of the individual cells in each image: (i) a gray level threshold, for a first step segmentation; (ii) a watershed-based separation of touching cells detected as a single object by the previous step; (iii) a morphology filter eliminated the artifacts. Morphology criteria included area, roundness and distance function, resulting in a fully segmented (called binary mask) cell image, accurately detected on the non-uniform background. The binary image and fluorescence intensity merged at the subcellular level provided a statistical analysis for multiple parameters of cell populations, automatically generated and displayed as a histogram of distribution. Then, the average EGFP fluorescence intensity was measured in every cell on the corresponding EGFP image and a positive/negative status was assigned to all cells. The ratio of the number of EGFP-positive cells to the total number of cells within a definite circle was used to evaluate the percentage of transfected cells.
RESULTS
The expression vector pEGFP-C1, which allows expression of genetically engineered GFP with a high fluorescence quantum yield, was used as a reporter construct to optimize conditions for highly parallel transfection of HEK293T cells in a cell array format. We first tested several concentrations of gelatin for spotting, in order to spatially confine DNA or RNA molecules within spots. We have printed blocks of 100 features for each concentration of gelatin. As demonstrated in Figure 1, we observed the best transfection efficiencies between 0.25 and 0.5% gelatin. At lower concentrations, clusters of transfected cells are less definite and at higher concentrations efficiency of transfection was slightly reduced and less localized over the printed area.
Figure 1. Effects of gelatin concentration on transfection efficiency and EGFP expression. Enlarged view of 4 x 10 printed features after transfection into HEK293T cells. Plasmid pEGFP-C1 was complexed with the transfection reagent and mixed with a range of gelatin solutions. From left to right and top to bottom: 1, 0.50, 0.25, 0.10, 0.05 and 0.025% (w/v) gelatin solutions. HEK293T cells were then plated onto the array and the scan was recorded 48 h after transfection.
The effect of concentration of the DNA–lipid–gelatin complex printed on the array was also tested. Transfection efficacy increased when higher concentrations of DNA were used (data not shown). However, when the complex was printed in excess (by increasing the number of hits for each pin at the same place on the microarray), we observed leakage of the expression vectors outside the spotted area and transfected cells were observed all over the array (Fig. 2). There is, therefore, a balance between transfection efficacy and spatial restriction. In the end, with one hit and 0.5% gelatin we were able to manufacture reliable arrays with thousands of gelatin spots containing the pEGFP-C1 plasmid ready to use in transfection assays. The diameter of printed spots after robot arraying was 200 μm. The distance from center to center was 500 μm (Fig. 3). We have been testing different transfection agents: Effectene? (Qiagen), LipofectamineTM 2000 (Invitrogen) and jetPEITM (Polyplus Transfection). Effectene? was by far the most efficient for transfecting HEK293T cells. In addition to HEK293T cells, we efficiently transfected HeLa cells and HaCaT human keratinocyte cells, optimizing conditions for each cell type (data not shown).
Figure 2. Effects of biopolymer–DNA complex amount on cell transfection. The plasmid DNA (pEGFP-C1) was first complexed to the transfection reagent and then mixed with the gelatin. The arrayer printed this solution either five times (a), three times (b) or once (c) on the polylysine-coated microscope slide. The number of hits (i.e. number of times one pin hits the slide at the same place) is directly proportional to the amount of solution printed onto the microarray (1 nl solution/hit). The amount of solution printed onto the microarray (1 nl solution/hit) is directly proportional to the number of hits (i.e. number of times one pin hits the slide at the same place).
Figure 3. Fluorescence of HEK293T cells transfected with an EGFP coding plasmid. The expression vector was complexed to the biopolymer/transfection reagent prior to printing on the microarray. (a) Laser scan image of 2 blocks containing 10 x 10 spots each. (b) Enlarged view of 3 x 4 clusters of fluorescent EGFP-expressing cells (green). (c) Enlargement of a single cluster of transfected cells (microscope images). Merged images of both transfected and non-transfected cells from light microscopy (blue) and fluorescent EGFP-expressing cells (green). The EGFP-expressing cells are localized in definite circles, 250 μm in diameter (actually 200 μm is the average diameter of the printed spots, but during automatic image analysis the chosen diameter for spot analysis was 300 μm).
To precisely quantify the efficacy of transfection into HEK293T cells, one needs image analysis with single cell resolution for thousands of array features. The resolution of conventional microarray scanners, usually 5 μm/pixel, was insufficient for phenotypic analysis and we had to use an automated high resolution microscope-based digital image acquisition system of 0.5 μm/pixel resolution, as well as custom algorithms for quantitative image analysis of cell microarrays. The software allowed us to automatically quantify the intensity of fluorescence within a cluster. As demonstrated in Figure 4, a computer-designed square mask was positioned over the array image and each cluster of transfected cells was localized within a circle. The diameter of circles in the grid was 300 μm, slightly larger than the printed features (200 μm). Background fluorescence was calculated within the square, but outside the circle (Fig. 4a). The population of cells within one circle was segmented to identify all individual cells and images were analyzed to extract parameters from each cluster of cells. A schematic representation of the array gave a mean fluorescence intensity value for each cluster of cells as well as a colored scale of expression levels (Fig. 4b). Efficacy of transfection was then quantified using a HEK293T cell array. All cells present within one cluster were quantified using both phase contrast and Hoechst 33342 staining of nuclei (Fig. 5b and c). Cells expressing EGFP were then counted and efficacy of transfection was calculated. As demonstrated in Figure 5d, the average number of cells within a 300 μm circle was 486 ± 65 and the average number of transfected cells per cluster was 83 ± 31, a transfection efficiency slightly over 17%. The number of transfected cells was plotted versus the mean fluorescence intensity per cluster (Fig. 5e), showing a weak correlation (coefficient of correlation = 0.67). This observed dispersion resulted from the fact that we could control neither the number of plasmid copies within a cell nor the level of EGFP expression. This was confirmed by the large range of fluorescence intensities per cell (Fig. 5f), showing that a few cells very strongly expressing EGFP generated large variability.
Figure 4. PathfinderTM OSA image of a 1 μm resolution fragment of a cell microarray describing the algorithm of background subtraction and data representation. (a) Acquired image with the computer designed grid. (b) Table of mean fluorescence values for each cluster (the background fluorescence was previously taken into account) and schematic representation of the mean fluorescence intensity of cell clusters using a colored scale (blue for the lowest intensities and red for the highest).
Figure 5. Fluorescent signal detection, automatic cell counting and fluorescence quantification on cell microarrays of HEK293T cells transfected with an EGFP coding plasmid. (a) PathfinderTM OSA image of block containing 8 x 8 spots. (b) Enlarged view of a single cluster: merged images of transfected and non-transfected cells from light microscopy (gray), fluorescent Hoechst 33342 (blue) and EGFP-expressing cells (green). (c) Enlargement of a single cluster of transfected cells representing merged images acquired by light microscopy (grey) and EGFP fluorescence (green). Automatically detected cell contours are displayed in red. The expressing cells are localized in definite spots, 300 μm in diameter, shown in green. Reliable cluster detection was obtained with light microscopy images due to their very high homogeneity and contrast. (d) Automatically counted and analyzed numbers of cells per cluster (both transfected and non-transfected) averaged from a 100 clusters analysis. The calculated percentage values of transfected cells are shown in parentheses. (e) Correlation of the number of transfected cells per cluster with the mean fluorescence intensity per cluster (analysis of a 100 clusters block). (f) Histogram showing the distribution of the cell number within a 300 μm circle for each level of fluorescence intensity (measured for 100 clusters).
We then tried to characterize whether we could quantify any loss of function using this microsystem. An EGFP-C1-expressing plasmid was co-transfected with an effective siRNA specifically targeting EGFP mRNA. Double-stranded RNA was directly printed into spots along with plasmid DNA. We have been testing different transfection agents; Transmessenger? (Qiagen), RNAiFectTM (Qiagen), OligofectamineTM (Invitrogen) and Effectene? (Qiagen) to get siRNA into the cell. Using Effectene?, a very significant dose-dependent siRNA extinction of EGFP expression was observed on cell microarrays (Fig. 6). This extinction was very specific, as siRNA targeting lamin A/C had only a slight effect on EGFP expression (Fig. 6). With our image analysis software, we quantified, for each siRNA concentration, the mean fluorescence intensity for each of 64 cell clusters. We observed a significant dose-dependent decrease in fluorescence intensity. At 0.15 μg siRNA, only 11% of expression was still observable. Nearly 97% of EGFP expression was suppressed using 0.6 μg siRNA (Fig. 6b). We observed good reproducibility of extinction over 64 spots printed and transfected under similar conditions.
Figure 6. Highly parallel gene silencing with EGFP targeting siRNA co-transfected with the pEGFP-C1 plasmid into HEK293T. (a) Six blocks of pEGFP-C1 (0.6 μg) ± EGFP (or lamin A/C) targeting siRNA were printed on a cell microarray and fluorescence was quantified after 48 h. Each block was made of a 10 x 10 surrounding square of control sample (pEGFP-C1 alone) and a 8 x 8 filled square of spots per sample . The slide was fixed and scanned after 48 h incubation. (b) Normalized relative values of EGFP fluorescence for each sample. Fluorescence of the 64 clusters was averaged after background subtraction (as described in Materials and Methods) and plotted versus the siRNA concentration (empty columns for EGFP targeting plasmids and filled column for the lamin A/C control). Error bars depict the standard deviation (n = 64).
We also suppressed EGFP expression by transfecting siRNA into HEK293T cells stably expressing EGFP. Double-stranded siRNA, directed against EGFP transcripts, was printed on the array. After transfection, one could observe clusters of cells no longer expressing EGFP (Fig. 7a). A close look at this area showed that, within this dark zone, rhodamine fluorescence associated with siRNA was observed in the cells (Fig. 7b). A control using phase contrast microscopy demonstrated that living cells were present and healthy where EGFP expression was ‘knocked down’ (Fig. 7c).
Figure 7. Endogenous gene silencing in HEK293T cells stably expressing EGFP (HEK293T/EGFP). (a) Focus on three spots of EGFP-specific siRNA printed on a cell microarray and reverse transfected into HEK293T/EGFP cells. An area of fluorescence extinction was observable 48 h post- transfection (yellow arrows). (b) Fluorescence and (c) light microscopy pictures of one spot containing HEK293T/EGFP reverse transfected with rhodamine-coupled EGFP-specific siRNA leading to a local knockdown of EGFP.
DISCUSSION
Based on the reverse transfection format (10), we have developed a cell microarray to simultaneously transfect thousands of different nucleic acid molecules (expression vectors or siRNA) and software, PathfinderTM OSA image analysis capture, to analyze and quantify phenotypes resulting from either gain or loss of gene function. We have defined optimum conditions not only for reproducible transfection of expression vectors and siRNA, but also co-transfection of both siRNA and expression vectors into HEK293T cells. Using a regular DNA arrayer, we have been able to print, then simultaneously transfect high numbers of nucleic acid molecules. We describe the right equilibrium conditions between efficacy of transfection and spatial restriction. This cell microarray format is as easy to manufacture as DNA arrays, because most of the devices developed to analyze DNA arrays can be used for this technology as well. As the cell microarrays are also printed by the same robots as regular DNA arrays, massively parallel transfection of up to 5000 cell clusters per slide could be feasible, as the size of printed features is 200 μm. The potential of this cell array for functional exploration of genes of unknown function is clear, as one could generate clusters of cells overexpressing or silencing defined gene products on the same array and analyze the resulting phenotypes. Moreover, this format is cost effective as minimal amounts of expensive materials are required: expression vectors, siRNA or short hairpin RNA (shRNA) and transfection reagents. Several slides can be printed in duplicate or triplicate to increase statistical reliability. To date, the main limiting factor in manufacturing pan-genomic arrays of expression vectors, shRNA or siRNA is the lack of an exhaustive collection of full-length cDNA of all open reading frames in the human genome, as well as large siRNA or shRNA libraries targeting human genes on the genomic scale.
The use of this cell array format has recently been proposed to screen for effective siRNA or shRNA probes for inhibition of target gene expression (11,12). In both studies, they used HeLa cells and Lipofectamine as the transfection reagent. Apparently, silencing of EGFP was higher under our conditions, as we could observe up to 97% ‘knock-down’ of EGFP expression while they observed up to 30% residual expression when tranfecting siRNA (11). The use of HEK293 cells might explain this difference, unless Effectene was more efficient in tranfecting siRNA. Surprisingly, siRNA-specific transfection reagents such as Transmessenger or RNAiffect were not as efficient in a cell microarray format. With respect to the transfection of expression vectors in a microarray format, we also found that Effectene was the most efficient transfection reagent. In both cases (siRNA and expression vector), it could be detrimental to print excessive quantities of nucleic acid or transfection reagent.
Even though regular scanners and software dedicated to DNA microarrays could be used to generate images (Figs 1 and 2) and to quantify fluorescence intensity of each feature, this could be limiting for more complex analysis, such as transfection efficiency or subcellular phenotype, requiring the use of a microscope-based system. To be able to quantify fluorescence in cell clusters using fluorescence microscopy, we have developed a software module that automatically captures images and quantifies signals. This software is flexible and different types of information can be generated, such as the number of cells within a feature and the overexpression or extinction efficiency. We could observe a very reproducible number of cells per cluster, while the number of transfected cells in each cluster was more variable. The percentages of transfection reported in this paper cannot be directly compared to values obtained in a classical format (i.e. 24- or 96-well plates). Indeed, to make sure that all transfected cells are taken into account, the diameter of the area under scrutiny was deliberately larger (300 μm) than the actual printed feature (200 μm), thus including cells not in direct contact with the DNA–gelatin complex. In the regular transfection format, plasmid DNA is added to the cells and each one is in direct contact with DNA so that the percentage of transfected cells can be calculated. In the reverse transfection format that we have been using, cells are added over DNA. It is extremely difficult to estimate the number of cells in direct contact with DNA, which reduces the significance of the transfection percentage. Also, the fluorescence intensity in each cell varied largely, which is not surprising as we cannot control either the number of plasmids entering the cell or the intensity of transient expression of EGFP within the cell. This fluorescence intensity variability points to the necessity to print hundreds of features to ensure sufficient statistical analysis of expression. Highly parallel transfection is then very useful as one can generate numerous clusters of transfected cells. Thanks to the image capture system and the image analysis software, single cells could be analyzed and small increases in signal versus background could be quantified, as well as small changes in signal intensity. That could be an advantage of the cell array format over other technologies. The cell array and bioinformatic tools we have developed should facilitate functional exploration of the human genome.
REFERENCES
Cheung,V.G. and Spielman,R.S. (2002) The genetics of variation in gene expression. Nature Genet., 32 (suppl.), 522–525.
Slonim,D.K. (2002) From patterns to pathways: gene expression data analysis comes of age. Nature Genet., 32 (suppl.), 502–508.
Schena,M., Shalon,D., Davis,R.W. and Brown,P.O. (1995) Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science, 270, 467–470.
Lockhart,D.J., Dong,H., Byrne,M.C., Follettie,M.T., Gallo,M.V., Chee,M.S., Mittmann,M., Wang,C., Kobayashi,M., Horton,H. et al. (1996) Expression monitoring by hybridization to high-density oligonucleotide arrays. Nat. Biotechnol., 14, 1675–1680.
Elbashir,S.M., Harborth,J., Lendeckel,W., Yalcin,A., Weber,K. and Tuschl,T. (2001) Duplexes of 21-nucleotide RNAs mediate RNA interference in cultured mammalian cells. Nature, 411, 494–498.
Fire,A., Xu,S., Montgomery,M.K., Kostas,S.A., Driver,S.E. and Mello,C.C. (1998) Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans. Nature, 391, 806–811.
Hammond,S.M., Caudy,A.A. and Hannon,G.J. (2001) Post-transcriptional gene silencing by double-stranded RNA. Nature Rev. Genet., 2, 110–119.
Pothof,J., van Haaften,G., Thijssen,K., Kamath,R.S., Fraser,A.G., Ahringer,J., Plasterk,R.H. and Tijsterman,M. (2003) Identification of genes that protect the C. elegans genome against mutations by genome-wide RNAi. Genes Dev., 17, 443–448.
Kamath,R.S., Fraser,A.G., Dong,Y., Poulin,G., Durbin,R., Gotta,M., Kanapin,A., Le Bot,N., Moreno,S., Sohrmann,M. et al. (2003) Systematic functional analysis of the Caenorhabditis elegans genome using RNAi. Nature, 421, 231–237.
Ziauddin,J. and Sabatini,D.M. (2001) Microarrays of cells expressing defined cDNAs. Nature, 411, 107–110.
Kumar,R., Conklin,D.S. and Mittal,V. (2003) High-throughput selection of effective RNAi probes for gene silencing. Genome Res., 13, 2333–2340.
Mousses,S., Caplen,N.J., Cornelison,R., Weaver,D., Basik,M., Hautaniemi,S., Elkahloun,A.G., Lotufo,R.A., Choudary,A., Dougherty,E.R. et al. (2003) RNAi microarray analysis in cultured mammalian cells. Genome Res., 13, 2341–2347.
Caplen,N.J., Parrish,S., Imani,F., Fire,A. and Morgan,R.A. (2001) Specific inhibition of gene expression by small double-stranded RNAs in invertebrate and vertebrate systems. Proc. Natl Acad. Sci. USA, 98, 9742–9747.
Khomyakova,E.B., Dreval,E.V., Tran-Dang,M., Potier,M.-C. and Soussaline,F.P. (2004) Innovative instrumentation for microarray scanning and analysis: application for characterization of oligonucleotide duplexes behavior. Cell. Mol. Biol., 50, in press.(Sandrine Baghdoyan, Yoann Roupioz, Amand)
*To whom correspondence should be addressed. Tel: +33 1 60 87 34 74; Fax: +33 1 60 87 34 98; Email: xavier.gidrol@cea.fr
ABSTRACT
As more genomes are sequenced, we are facing the challenge of rapidly unraveling the functions of genes. To that end, cell microarrays have recently been described that transfect thousands of nucleic acids in parallel and can be used to analyze the phenotypic consequences of such perturbations. As many parameters can influence the efficacy of transfection in such a format, we describe some important features in manufacturing cell microarrays that may improve reliability and efficiency of both plasmid DNA and siRNA transfection. We have also developed image analysis software that allows automatic detection of cell clusters, quantification of transfection efficiency and levels of expression/extinction of genes. Along with cell microarrays, this bioinformatic tool should expedite functional exploration of the human genome.
INTRODUCTION
With the complete sequencing of the human genome, research priorities have shifted from the identification of genes to the elucidation of their function. High throughput technologies are a key feature of functional genomic experimentation. In the mid 1990s, DNA arrays made it possible to significantly increase throughput of gene expression analysis by simultaneously monitoring tens of thousands of genes (1–4). Before that technological development, biologists were studying gene expression of a few genes at a time by northern blots and RT–PCR. They are now able to monitor expression at the genomic scale and the entire human genome can be analyzed in a single array.
Similarly, methods to characterize gene function are also utilized, such as transgenic or knockout mice. They are based upon gain or loss of protein function and analysis of the resulting phenotypes to infer a potential role for the protein under scrutiny. In such approaches, DNA constructs that direct overexpression of a gene product or, on the contrary, eliminate its synthesis are introduced into the cell. Indeed, such phenotypic analysis gives a good idea of the potential function of the gene product. Until now, these methods were time consuming and only a few genes at a time could be analyzed. It was recently demonstrated that chemically synthesized short (<30 nt long) double-stranded siRNA (small interfering RNA) molecules, homologous to a target gene, could specifically inactivate gene function when introduced into the cell (5). RNA interference is a natural process for sequence-specific, post-transcriptional gene silencing initiated by double-stranded RNA (6,7). Thus, RNAi offers the possibility of high throughput ‘knockdown’ studies for the analysis of thousands of genes of unknown function (8,9).
To speed up the functional exploration of the human genome, there is a need for high throughput technologies allowing transfection of thousands of nucleic acids in parallel and the simultaneous analysis of thousands of resulting phenotypes. Ziauddin and Sabatini have described a cheap and flexible cell-based microarray system for the high throughput analysis of gene overexpression (10). Others have used this technology with siRNA (11,12), however, there remain several parameters that impact on the quality and reproducibility of transfection in such a cell microarray. In this report, we describe procedures and key features for manufacturing cell microarrays that generate reproducible and highly parallel transfection. Furthermore, to precisely quantify efficacy of transfection, level of expression or extinction of genes, image analysis software was also developed. This cell array format and automated image analysis system have the potential to be used in extensive analysis of gene function at the genome scale.
MATERIALS AND METHODS
Plasmid and small interfering RNA (siRNA)
The pEGFP-C1 plasmid expressing enhanced green fluorescent protein (EGFP) was obtained from Clontech (Paolo Alto, CA). Escherichia coli cells were transformed with pEGFP-C1 and plasmids were purified with a Midi Prep Qiagen Plasmid Kit (Qiagen, Hilden, Germany). Plasmid concentrations were assessed by UV absorbance; the OD 280/260 nm ratio was always >1.8. Synthetic siRNAs specific to lamin A/C (sense CUGGACUUCCAGAAGAACAdTdT, antisense UGUUCU UCUGGAAGUCCAGdTdT) (5) and EGFP, modified at their 3'-end with rhodamine (sense GCAAGCUGACCCUG AAGUUCAU, antisense GAACUUCAGGGUCAGCUUG CCG) (13) were purchased from Qiagen. For transfection, siRNAs were solubilized for 1 h at 37°C in a resuspension buffer (30 mM HEPES–KOH pH 7.4, 100 mM potassium acetate, 2 mM magnesium acetate) to a final concentration of 0.3 μg/μl.
Cell array printing
The general procedure was inspired by Ziauddin and Sabatini’s work (10) and was optimized to achieve better reproducibility of EGFP transfection. Five microliters of pEGFP-C1 at 0.1 μg/μl was diluted with 6.5 μl of EC buffer (Effectene kit; Qiagen). Two microliters of Enhancer supplemented with 1.2 μl of a 1.5 M sucrose solution and 2 μl of Effectene reagent was successively added to the mixture. After a 15 min incubation at room temperature, 12 μl of a 0.5% gelatin solution (Sigma G-1393 gelatin diluted in deionized water) was added and the solution was transferred to a 96-well plate for microarray printing. Regarding siRNA microarrays, the general procedure was the same except for slight modifications of the plasmids, siRNAs and formation of lipid complexes. One microliter of pEGFP-C1 at 0.6 μg/μl was mixed with 0.15–0.6 μg siRNA, specific to EGFP or lamin A/C. Mixtures were diluted in 11 μl of EC buffer supplemented with 2 μl of a 1.5 M sucrose solution, and 3.3 μl of Enhancer plus 3.3 μl of Effectene reagent was next added to the samples. After 15 min at room temperature, 6 μl of a 1% gelatin solution was added and samples were transferred to a 96-well plate for microarray printing. Poly(lysine)-coated (PolysineTM; Menzel-Gl?ser) microscope slides (25 x 75 x 1.0 mm) were printed with a Microgrid II Biorobotics (Cambridge, UK) arrayer at room temperature equipped with Biorobotics quill pins. The spots were 200 μm in diameter (verified by printing of fluorescent DNA) and were spotted with a 400–550 μm pitch. In most experiments, 600 features per slide were printed to facilitate the image analysis, but >5000 spots can be printed onto a single slide. Slides were stored at room temperature in a dessicator for several weeks.
Cell culture and reverse transfection
Human embryonic kidney cell line HEK293T and HEK293T cells stably expressing EGFP (kindly provided by Didier Poncet, INRA, Jouy-en-Josas, France) were grown at 37°C in a 5% CO2 humidified atmosphere in Dulbecco’s modified Eagle’s medium containing 4.5 g/l glucose supplemented with 10% fetal calf serum, 100 000 U/l penicillin, 50 mg/l streptomycin and 200 mM glutamine. Cells were seeded onto the microarray in 100 mm Petri dishes to reach confluence 48 h after transfection (100–150 000 cells/cm2 with HEK293T cells, 80 000 cells/cm2 with HaCaT cells and 60 000 cells/cm2 with HeLa cells) and cultured for 24–48 h at 37°C/5% CO2. Slides were fixed following a standard protocol (4% paraformaldehyde for 15 min at room temperature) and stored in phosphate-buffered saline at +4°C for several weeks.
Scanning and image capture
Cell microarrays were scanned using a ScanArray? 5000 scanner (Packard BioChip Technologies) and quantified using Genepix? software (Axon Instruments, CA). To measure EGFP fluorescence of transfected cells, images were acquired with a PathfinderTM OSA instrument (14) (IMSTAR, http://www.imstar.fr), an automated optical scanning system equipped with a high resolution (1300 x 1000 pixels) CCD camera with a dynamic range of 12 bits or 4096 levels of intensity per pixel, an adapted filter set and various automated devices. This instrument allows the registration of multiple fluorescence images of the entire cell microarray over a wide range of excitation/emission wavelengths and at a final resolution of 0.34 μm/pixel. Two processing steps were performed on the raw data coming from the camera: dark current subtraction and shading correction. The first consists of pixel-by-pixel subtraction of the image obtained in the absence of light from every captured image. The second consists of dividing the captured images by the image captured from an empty field with a statistically significant exposition time, the goal of this second step being to compensate for light excitation uniformity across the camera view field. When necessary, fluorescence microscopes were used and array images were captured and analyzed with either an Olympus IX70 or a Zeiss Axiovert microscope. Both have fluorescent light emission systems for EGFP detection (excitation 488 nm, emission 520 nm) and a digital camera for picture image acquisition.
Quantification of cell arrays
Fluorescent images were analyzed with a PathfinderTM OSA image analysis commercial software package. This software comprises an integrated image database management component that underlies all other functions from automated image acquisition through protocols of image processing and image segmentation to the final results tables and statistical analysis. The database structure is flexible to support any image-to-image relationship. It allows image browsing by displaying the results of a query on the database and lets the user recognize the image by its reduced resolution view. Furthermore, image processing macro functions, called protocols, can be applied to any number of images selected by a query on the database, using a single command. Finally, the results table, which accumulates the calculated parameters for all objects from all images from the query, can be displayed and analyzed. This allows complete analysis in record time, with minimal user interaction.
Total level of EGFP expression. The level of EGFP expression was estimated by measuring the mean EGFP fluorescence intensity corresponding to each cluster, after automatic recognition and contour detection. A digital mask containing six grids of 10 x 10 elements, each corresponding to the pattern of EGFP-expressing DNA constructs printed on the slide, were plotted from the acquired microarray image by software. The fluorescence mean intensity corresponding to each cluster was calculated within a circle of diameter Dspot = 300 μm in each square grid. The mean fluorescence intensity inside the circle was background corrected by subtracting the median fluorescence intensity of pixels located in the grid square but outside the circle. The values corresponding to each block are represented as the average fluorescence intensity of 64–100 spots.
Transfection efficiency. To estimate the EGFP plasmid transfection efficiency, the ratio of transfected cells (within the Dspot = 300 μm circle) to total cells was calculated. Detection of non-transfected cells was performed by Hoechst 33342 staining, followed by capture of either Hoechst 33342-specific fluorescence images or bright field microarray images. The images obtained had a resolution of 0.7 μm/pixel. The analysis was based upon superimposed images corresponding to Hoechst 33342 and EGFP fluorescence. The detection and segmentation of Hoechst 33342-stained individual cells was performed on a bright field image using PathfinderTM software where fluorescent spots, defined as cell clusters, are recognized as well as cell morphology in each spot. Basically, for the present study the segmentation procedure developed associates mathematical morphology algorithms and measurements of intensity, size and shape and consists of a multi-step protocol in order to generate a binary image of the individual cells in each image: (i) a gray level threshold, for a first step segmentation; (ii) a watershed-based separation of touching cells detected as a single object by the previous step; (iii) a morphology filter eliminated the artifacts. Morphology criteria included area, roundness and distance function, resulting in a fully segmented (called binary mask) cell image, accurately detected on the non-uniform background. The binary image and fluorescence intensity merged at the subcellular level provided a statistical analysis for multiple parameters of cell populations, automatically generated and displayed as a histogram of distribution. Then, the average EGFP fluorescence intensity was measured in every cell on the corresponding EGFP image and a positive/negative status was assigned to all cells. The ratio of the number of EGFP-positive cells to the total number of cells within a definite circle was used to evaluate the percentage of transfected cells.
RESULTS
The expression vector pEGFP-C1, which allows expression of genetically engineered GFP with a high fluorescence quantum yield, was used as a reporter construct to optimize conditions for highly parallel transfection of HEK293T cells in a cell array format. We first tested several concentrations of gelatin for spotting, in order to spatially confine DNA or RNA molecules within spots. We have printed blocks of 100 features for each concentration of gelatin. As demonstrated in Figure 1, we observed the best transfection efficiencies between 0.25 and 0.5% gelatin. At lower concentrations, clusters of transfected cells are less definite and at higher concentrations efficiency of transfection was slightly reduced and less localized over the printed area.
Figure 1. Effects of gelatin concentration on transfection efficiency and EGFP expression. Enlarged view of 4 x 10 printed features after transfection into HEK293T cells. Plasmid pEGFP-C1 was complexed with the transfection reagent and mixed with a range of gelatin solutions. From left to right and top to bottom: 1, 0.50, 0.25, 0.10, 0.05 and 0.025% (w/v) gelatin solutions. HEK293T cells were then plated onto the array and the scan was recorded 48 h after transfection.
The effect of concentration of the DNA–lipid–gelatin complex printed on the array was also tested. Transfection efficacy increased when higher concentrations of DNA were used (data not shown). However, when the complex was printed in excess (by increasing the number of hits for each pin at the same place on the microarray), we observed leakage of the expression vectors outside the spotted area and transfected cells were observed all over the array (Fig. 2). There is, therefore, a balance between transfection efficacy and spatial restriction. In the end, with one hit and 0.5% gelatin we were able to manufacture reliable arrays with thousands of gelatin spots containing the pEGFP-C1 plasmid ready to use in transfection assays. The diameter of printed spots after robot arraying was 200 μm. The distance from center to center was 500 μm (Fig. 3). We have been testing different transfection agents: Effectene? (Qiagen), LipofectamineTM 2000 (Invitrogen) and jetPEITM (Polyplus Transfection). Effectene? was by far the most efficient for transfecting HEK293T cells. In addition to HEK293T cells, we efficiently transfected HeLa cells and HaCaT human keratinocyte cells, optimizing conditions for each cell type (data not shown).
Figure 2. Effects of biopolymer–DNA complex amount on cell transfection. The plasmid DNA (pEGFP-C1) was first complexed to the transfection reagent and then mixed with the gelatin. The arrayer printed this solution either five times (a), three times (b) or once (c) on the polylysine-coated microscope slide. The number of hits (i.e. number of times one pin hits the slide at the same place) is directly proportional to the amount of solution printed onto the microarray (1 nl solution/hit). The amount of solution printed onto the microarray (1 nl solution/hit) is directly proportional to the number of hits (i.e. number of times one pin hits the slide at the same place).
Figure 3. Fluorescence of HEK293T cells transfected with an EGFP coding plasmid. The expression vector was complexed to the biopolymer/transfection reagent prior to printing on the microarray. (a) Laser scan image of 2 blocks containing 10 x 10 spots each. (b) Enlarged view of 3 x 4 clusters of fluorescent EGFP-expressing cells (green). (c) Enlargement of a single cluster of transfected cells (microscope images). Merged images of both transfected and non-transfected cells from light microscopy (blue) and fluorescent EGFP-expressing cells (green). The EGFP-expressing cells are localized in definite circles, 250 μm in diameter (actually 200 μm is the average diameter of the printed spots, but during automatic image analysis the chosen diameter for spot analysis was 300 μm).
To precisely quantify the efficacy of transfection into HEK293T cells, one needs image analysis with single cell resolution for thousands of array features. The resolution of conventional microarray scanners, usually 5 μm/pixel, was insufficient for phenotypic analysis and we had to use an automated high resolution microscope-based digital image acquisition system of 0.5 μm/pixel resolution, as well as custom algorithms for quantitative image analysis of cell microarrays. The software allowed us to automatically quantify the intensity of fluorescence within a cluster. As demonstrated in Figure 4, a computer-designed square mask was positioned over the array image and each cluster of transfected cells was localized within a circle. The diameter of circles in the grid was 300 μm, slightly larger than the printed features (200 μm). Background fluorescence was calculated within the square, but outside the circle (Fig. 4a). The population of cells within one circle was segmented to identify all individual cells and images were analyzed to extract parameters from each cluster of cells. A schematic representation of the array gave a mean fluorescence intensity value for each cluster of cells as well as a colored scale of expression levels (Fig. 4b). Efficacy of transfection was then quantified using a HEK293T cell array. All cells present within one cluster were quantified using both phase contrast and Hoechst 33342 staining of nuclei (Fig. 5b and c). Cells expressing EGFP were then counted and efficacy of transfection was calculated. As demonstrated in Figure 5d, the average number of cells within a 300 μm circle was 486 ± 65 and the average number of transfected cells per cluster was 83 ± 31, a transfection efficiency slightly over 17%. The number of transfected cells was plotted versus the mean fluorescence intensity per cluster (Fig. 5e), showing a weak correlation (coefficient of correlation = 0.67). This observed dispersion resulted from the fact that we could control neither the number of plasmid copies within a cell nor the level of EGFP expression. This was confirmed by the large range of fluorescence intensities per cell (Fig. 5f), showing that a few cells very strongly expressing EGFP generated large variability.
Figure 4. PathfinderTM OSA image of a 1 μm resolution fragment of a cell microarray describing the algorithm of background subtraction and data representation. (a) Acquired image with the computer designed grid. (b) Table of mean fluorescence values for each cluster (the background fluorescence was previously taken into account) and schematic representation of the mean fluorescence intensity of cell clusters using a colored scale (blue for the lowest intensities and red for the highest).
Figure 5. Fluorescent signal detection, automatic cell counting and fluorescence quantification on cell microarrays of HEK293T cells transfected with an EGFP coding plasmid. (a) PathfinderTM OSA image of block containing 8 x 8 spots. (b) Enlarged view of a single cluster: merged images of transfected and non-transfected cells from light microscopy (gray), fluorescent Hoechst 33342 (blue) and EGFP-expressing cells (green). (c) Enlargement of a single cluster of transfected cells representing merged images acquired by light microscopy (grey) and EGFP fluorescence (green). Automatically detected cell contours are displayed in red. The expressing cells are localized in definite spots, 300 μm in diameter, shown in green. Reliable cluster detection was obtained with light microscopy images due to their very high homogeneity and contrast. (d) Automatically counted and analyzed numbers of cells per cluster (both transfected and non-transfected) averaged from a 100 clusters analysis. The calculated percentage values of transfected cells are shown in parentheses. (e) Correlation of the number of transfected cells per cluster with the mean fluorescence intensity per cluster (analysis of a 100 clusters block). (f) Histogram showing the distribution of the cell number within a 300 μm circle for each level of fluorescence intensity (measured for 100 clusters).
We then tried to characterize whether we could quantify any loss of function using this microsystem. An EGFP-C1-expressing plasmid was co-transfected with an effective siRNA specifically targeting EGFP mRNA. Double-stranded RNA was directly printed into spots along with plasmid DNA. We have been testing different transfection agents; Transmessenger? (Qiagen), RNAiFectTM (Qiagen), OligofectamineTM (Invitrogen) and Effectene? (Qiagen) to get siRNA into the cell. Using Effectene?, a very significant dose-dependent siRNA extinction of EGFP expression was observed on cell microarrays (Fig. 6). This extinction was very specific, as siRNA targeting lamin A/C had only a slight effect on EGFP expression (Fig. 6). With our image analysis software, we quantified, for each siRNA concentration, the mean fluorescence intensity for each of 64 cell clusters. We observed a significant dose-dependent decrease in fluorescence intensity. At 0.15 μg siRNA, only 11% of expression was still observable. Nearly 97% of EGFP expression was suppressed using 0.6 μg siRNA (Fig. 6b). We observed good reproducibility of extinction over 64 spots printed and transfected under similar conditions.
Figure 6. Highly parallel gene silencing with EGFP targeting siRNA co-transfected with the pEGFP-C1 plasmid into HEK293T. (a) Six blocks of pEGFP-C1 (0.6 μg) ± EGFP (or lamin A/C) targeting siRNA were printed on a cell microarray and fluorescence was quantified after 48 h. Each block was made of a 10 x 10 surrounding square of control sample (pEGFP-C1 alone) and a 8 x 8 filled square of spots per sample . The slide was fixed and scanned after 48 h incubation. (b) Normalized relative values of EGFP fluorescence for each sample. Fluorescence of the 64 clusters was averaged after background subtraction (as described in Materials and Methods) and plotted versus the siRNA concentration (empty columns for EGFP targeting plasmids and filled column for the lamin A/C control). Error bars depict the standard deviation (n = 64).
We also suppressed EGFP expression by transfecting siRNA into HEK293T cells stably expressing EGFP. Double-stranded siRNA, directed against EGFP transcripts, was printed on the array. After transfection, one could observe clusters of cells no longer expressing EGFP (Fig. 7a). A close look at this area showed that, within this dark zone, rhodamine fluorescence associated with siRNA was observed in the cells (Fig. 7b). A control using phase contrast microscopy demonstrated that living cells were present and healthy where EGFP expression was ‘knocked down’ (Fig. 7c).
Figure 7. Endogenous gene silencing in HEK293T cells stably expressing EGFP (HEK293T/EGFP). (a) Focus on three spots of EGFP-specific siRNA printed on a cell microarray and reverse transfected into HEK293T/EGFP cells. An area of fluorescence extinction was observable 48 h post- transfection (yellow arrows). (b) Fluorescence and (c) light microscopy pictures of one spot containing HEK293T/EGFP reverse transfected with rhodamine-coupled EGFP-specific siRNA leading to a local knockdown of EGFP.
DISCUSSION
Based on the reverse transfection format (10), we have developed a cell microarray to simultaneously transfect thousands of different nucleic acid molecules (expression vectors or siRNA) and software, PathfinderTM OSA image analysis capture, to analyze and quantify phenotypes resulting from either gain or loss of gene function. We have defined optimum conditions not only for reproducible transfection of expression vectors and siRNA, but also co-transfection of both siRNA and expression vectors into HEK293T cells. Using a regular DNA arrayer, we have been able to print, then simultaneously transfect high numbers of nucleic acid molecules. We describe the right equilibrium conditions between efficacy of transfection and spatial restriction. This cell microarray format is as easy to manufacture as DNA arrays, because most of the devices developed to analyze DNA arrays can be used for this technology as well. As the cell microarrays are also printed by the same robots as regular DNA arrays, massively parallel transfection of up to 5000 cell clusters per slide could be feasible, as the size of printed features is 200 μm. The potential of this cell array for functional exploration of genes of unknown function is clear, as one could generate clusters of cells overexpressing or silencing defined gene products on the same array and analyze the resulting phenotypes. Moreover, this format is cost effective as minimal amounts of expensive materials are required: expression vectors, siRNA or short hairpin RNA (shRNA) and transfection reagents. Several slides can be printed in duplicate or triplicate to increase statistical reliability. To date, the main limiting factor in manufacturing pan-genomic arrays of expression vectors, shRNA or siRNA is the lack of an exhaustive collection of full-length cDNA of all open reading frames in the human genome, as well as large siRNA or shRNA libraries targeting human genes on the genomic scale.
The use of this cell array format has recently been proposed to screen for effective siRNA or shRNA probes for inhibition of target gene expression (11,12). In both studies, they used HeLa cells and Lipofectamine as the transfection reagent. Apparently, silencing of EGFP was higher under our conditions, as we could observe up to 97% ‘knock-down’ of EGFP expression while they observed up to 30% residual expression when tranfecting siRNA (11). The use of HEK293 cells might explain this difference, unless Effectene was more efficient in tranfecting siRNA. Surprisingly, siRNA-specific transfection reagents such as Transmessenger or RNAiffect were not as efficient in a cell microarray format. With respect to the transfection of expression vectors in a microarray format, we also found that Effectene was the most efficient transfection reagent. In both cases (siRNA and expression vector), it could be detrimental to print excessive quantities of nucleic acid or transfection reagent.
Even though regular scanners and software dedicated to DNA microarrays could be used to generate images (Figs 1 and 2) and to quantify fluorescence intensity of each feature, this could be limiting for more complex analysis, such as transfection efficiency or subcellular phenotype, requiring the use of a microscope-based system. To be able to quantify fluorescence in cell clusters using fluorescence microscopy, we have developed a software module that automatically captures images and quantifies signals. This software is flexible and different types of information can be generated, such as the number of cells within a feature and the overexpression or extinction efficiency. We could observe a very reproducible number of cells per cluster, while the number of transfected cells in each cluster was more variable. The percentages of transfection reported in this paper cannot be directly compared to values obtained in a classical format (i.e. 24- or 96-well plates). Indeed, to make sure that all transfected cells are taken into account, the diameter of the area under scrutiny was deliberately larger (300 μm) than the actual printed feature (200 μm), thus including cells not in direct contact with the DNA–gelatin complex. In the regular transfection format, plasmid DNA is added to the cells and each one is in direct contact with DNA so that the percentage of transfected cells can be calculated. In the reverse transfection format that we have been using, cells are added over DNA. It is extremely difficult to estimate the number of cells in direct contact with DNA, which reduces the significance of the transfection percentage. Also, the fluorescence intensity in each cell varied largely, which is not surprising as we cannot control either the number of plasmids entering the cell or the intensity of transient expression of EGFP within the cell. This fluorescence intensity variability points to the necessity to print hundreds of features to ensure sufficient statistical analysis of expression. Highly parallel transfection is then very useful as one can generate numerous clusters of transfected cells. Thanks to the image capture system and the image analysis software, single cells could be analyzed and small increases in signal versus background could be quantified, as well as small changes in signal intensity. That could be an advantage of the cell array format over other technologies. The cell array and bioinformatic tools we have developed should facilitate functional exploration of the human genome.
REFERENCES
Cheung,V.G. and Spielman,R.S. (2002) The genetics of variation in gene expression. Nature Genet., 32 (suppl.), 522–525.
Slonim,D.K. (2002) From patterns to pathways: gene expression data analysis comes of age. Nature Genet., 32 (suppl.), 502–508.
Schena,M., Shalon,D., Davis,R.W. and Brown,P.O. (1995) Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science, 270, 467–470.
Lockhart,D.J., Dong,H., Byrne,M.C., Follettie,M.T., Gallo,M.V., Chee,M.S., Mittmann,M., Wang,C., Kobayashi,M., Horton,H. et al. (1996) Expression monitoring by hybridization to high-density oligonucleotide arrays. Nat. Biotechnol., 14, 1675–1680.
Elbashir,S.M., Harborth,J., Lendeckel,W., Yalcin,A., Weber,K. and Tuschl,T. (2001) Duplexes of 21-nucleotide RNAs mediate RNA interference in cultured mammalian cells. Nature, 411, 494–498.
Fire,A., Xu,S., Montgomery,M.K., Kostas,S.A., Driver,S.E. and Mello,C.C. (1998) Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans. Nature, 391, 806–811.
Hammond,S.M., Caudy,A.A. and Hannon,G.J. (2001) Post-transcriptional gene silencing by double-stranded RNA. Nature Rev. Genet., 2, 110–119.
Pothof,J., van Haaften,G., Thijssen,K., Kamath,R.S., Fraser,A.G., Ahringer,J., Plasterk,R.H. and Tijsterman,M. (2003) Identification of genes that protect the C. elegans genome against mutations by genome-wide RNAi. Genes Dev., 17, 443–448.
Kamath,R.S., Fraser,A.G., Dong,Y., Poulin,G., Durbin,R., Gotta,M., Kanapin,A., Le Bot,N., Moreno,S., Sohrmann,M. et al. (2003) Systematic functional analysis of the Caenorhabditis elegans genome using RNAi. Nature, 421, 231–237.
Ziauddin,J. and Sabatini,D.M. (2001) Microarrays of cells expressing defined cDNAs. Nature, 411, 107–110.
Kumar,R., Conklin,D.S. and Mittal,V. (2003) High-throughput selection of effective RNAi probes for gene silencing. Genome Res., 13, 2333–2340.
Mousses,S., Caplen,N.J., Cornelison,R., Weaver,D., Basik,M., Hautaniemi,S., Elkahloun,A.G., Lotufo,R.A., Choudary,A., Dougherty,E.R. et al. (2003) RNAi microarray analysis in cultured mammalian cells. Genome Res., 13, 2341–2347.
Caplen,N.J., Parrish,S., Imani,F., Fire,A. and Morgan,R.A. (2001) Specific inhibition of gene expression by small double-stranded RNAs in invertebrate and vertebrate systems. Proc. Natl Acad. Sci. USA, 98, 9742–9747.
Khomyakova,E.B., Dreval,E.V., Tran-Dang,M., Potier,M.-C. and Soussaline,F.P. (2004) Innovative instrumentation for microarray scanning and analysis: application for characterization of oligonucleotide duplexes behavior. Cell. Mol. Biol., 50, in press.(Sandrine Baghdoyan, Yoann Roupioz, Amand)