|

 

Gene-quantification.biz

Published online August 25, 2004
Nucleic Acids Research, 2004, Vol. 32, No. 15 RNA expression microarrays (REMs), ahigh-throughput method to measure differencesin gene expression in diverse biological samples Charles E. Rogler*, Tatyana Tchaikovskaya, Raquel Norel, Aldo Massimi1,Christopher Plescia2, Eugeny Rubashevsky, Paul Siebert3 and Leslie E. Rogler Department of Medicine and Marion Bessin Liver Research Center, 1Department of Molecular Genetics, Albert EinsteinCollege of Medicine, Bronx, NY, USA, 2Department of Neurosciences, Mt Sinai College of Medicine, New York, NY,USA and 3BD Biosciences-Clontech, Palo Alto, CA, USA Received April 12, 2004; Revised June 4, 2004; Accepted July 30, 2004 have a vital function. Using cDNA microarrays, researcherscan simultaneously measure steady-state mRNA levels in all We have developed RNA expression microarrays the known genes and thousands of expressed sequence tags (REMs), in which each spot on a glass support is com- (ESTs) expressed in a cell (4–6).
posed of a population of cDNAs synthesized from a Genome-wide expression analysis has advanced transcrip- cell or tissue sample. We used simultaneous hybrid- tional based research in all areas of biology. In mammalian ization with test and reference (housekeeping) genes biology for example, cDNA microarray approaches have iden- to calculate an expression ratio based on normaliza- tified novel genes involved in the cell cycle (7,8), specific tion with the endogenous reference gene. A test REM differentiation programs (9,10) and specific disease states containing artificial mixtures of liver cDNA and dilu- (11,12) to mention just a few. In cancer biology, an important tions of the bacterial LysA gene cDNA demonstrated application has been the identification of distinct subtypes of the feasibility of detecting transcripts at a sensitivity tumors, such as subtypes of breast tumors (13,14), lymphomas(15), kidney tumors (16), melanomas (17) and other tumor of four copies of LysA mRNA per liver cell equivalent.
Furthermore, LysA cDNA detection varied linearly In all the above cases, the genome-wide scans have led to across a standard curve that matched the sensitivity the identification of candidate disease-specific genes in defined of quantitative real-time PCR. In REMs with real sam- sets of samples. Follow-up studies, on the expression of the ples, we detected organ-specific expression of albu- best candidate genes in a much larger sample base, are then min, Hnf-4 and Igfbp-1, in a set of mouse organ cDNA needed to quantitate and validate their involvement in specific populations and c-Myc expression in tumor samples biological processes or diseases. Experimental approaches to in paired tumor/normal tissue cDNA samples. REMs investigating the broader ‘expression niche' of candidate extend the use of classic microarrays in that a single genes include, among others, the use of tissue northern REM can contain cDNAs from hundreds to thousands blots, RNA dot blots (19) and tissue microarrays (20). Acqui- of cell or tissue samples each representing a specific sition of samples and sample processing for large sets of sam-ples are often the rate-limiting step in this process. Therefore, physiological or pathophysiological state. REMs will there is a need for an experimental tool that provides large sets extend the analysis of valuable samples by providing of samples that can be probed for the expression of specific a common broad based platform for their analysis genes in a high-throughput manner. In addition, the tool should and will promote research aimed at defining gene use small amounts of valuable samples so that the effective use functions, by broadening our understanding of their of those samples can be extended.
expression patterns in health and disease.
In this report we describe the development and use of a new microarray technology, called RNA Expression Microarrays(REMs), that addresses the above needs. REMs are producedby spotting cDNAs synthesized from the poly(A)+ mRNAs of a tissue. REMs have the advantage of precise internal normal- Evolutionary selection pressure functions both at the organ- ization, quantitative comparisons between the samples and a ismal level, and at the molecular level, by precisely tuning the capacity for high-throughput analysis of thousands of diverse regulatory properties of enhancers and promoters, so that each samples simultaneously. We have validated the technology gene product is produced when and where it is needed and in using artificial mixtures and compared it with the leading sufficient quantities to supply its required function (1–3). Con- quantitative expression analysis method, and applied REM sequently, the temporal and spatial pattern of expression of a technology toward biologically relevant questions in develop- gene is a catalog of biological processes in which a gene can mental and cancer biology.
*To whom correspondence should be addressed. Tel: +1 718 430 2607; Fax: +1 718 430 8975; Email: [email protected] Nucleic Acids Research, Vol. 32 No. 15 ª Oxford University Press 2004; all rights reserved Nucleic Acids Research, 2004, Vol. 32, No. 15 chamber. Approximately 20 ml of prehybridization solution(prehybridization solution is 35% formamide, 4· SSPE, Preparation of fluorescent probes 0.5% SDS, 2.5· Denhardts and 0.2 mg/ml salmon sperm Gene-specific sense and antisense primers
500 bp apart are DNA) is added over the arrayed samples and the coverslip identified near the 30 end of the cDNA sequence of selected is placed over the samples avoiding bubbles. The slide is gene. A T7 promoter sequence is attached to the antisense incubated in the hybridization chamber that is humidified primer and the cDNA fragment is PCR amplified, purified by adding 10 ml of water in each corner, for 1–2 h at 50C.
using the Qiaquick PCR purification kit and the product is After incubation the coverslip is removed by dipping in water, sequence verified. An antisense RNA is synthesized using the slide is dried by centrifugation as above, dust is removed as T7 RNA polymerase according to the Epicenter AmpliScribe described, the slide is returned to the chamber, covered by T7 Flash transcription kit protocol, (Epicentre Cat. no.
hybridization solution containing a mixture of Cy3- and Cy5- ASF3257), except that the reaction is carried out at 42C labeled probes and by coverslip, and incubated in humidified for 1 h. The antisense RNA is purified using a RNeasy hybridization chamber for 16–20 h at 50C.
Mini kit from Qiagen (Cat. no. 74104). Five micrograms of After hybridization, the cover slip is removed by immersing antisense RNA, at 0.3 mg/ml in H2O, is annealed at 70C for REM in 100 ml of 2· SSC/0.1% SDS then washed with several 5 min with 6 mM sense primer. After annealing, Cy3- or Cy5- hundred milliliters of 0.2· SSC/0.1% SDS with stirring for 10– labeled sense strand cDNA is synthesized using 10 U/ml 15 min at room temperature, washed with 0.2· SSC and then Invitrogen Superscript III reverse transcriptase, in 50mM with 0.1· SSC for every 15 min. Slide is dried by centrifuga- Tris–HCl, pH 8.3, 75 mM KCl, 3 mM MgCl2, 10 mM DTT, tion as above, stored at room temperature, in dark, until scan- 1 U/ml RNaseOUT (Invitrogen, Cat. no. 10777-019), 500 nM ning (preferably the same day).
dATP, dCTP, dGTP and 200 nM dTTP, plus either Cy3- orCy5-labeled dUTP at 100 nM, at 50C for 2 h (reaction volume Preparation of single-stranded LysA antisense cDNA normally 40 ml). After completion of the reaction, an equal and dilution into antisense liver cDNA volume of 17.5 mM MgCl2 and 250 mM Tris–HCl, pH 7.4, A 1 kb segment of a bacterial clone for diaminopimelate containing 4 U of RNAse H, is added and incubated for 30 min decarboxylase (LysA, ATCC accession number 87482) was at 37C, followed by the treatment with 0.5 U/ml of RNase 1 subcloned into pBluescript II KS+. The clone contained a 60 nt and RNase 1 buffer (10 mM Tris–HCl, pH 7.5, 5 mM EDTA artificial poly(A) tail at its 30 end. A 1.1 kb DNA fragment was and 200 mM sodium acetate (Promega no. 4261) for 10 min at amplified from the plasmid using antisense T7 and senseT3 37C. Probe solutions containing either Cy3 or Cy5 are com- primers homologous to plasmid sequences and the PCR pro- bined and purified together using a Qiaquick PCR purification duct was sequence verified. Sense strand LysA aRNA was kit. Final purification is accomplished by elution from the synthesized using an Ambion MEGAscript T3 RNA polymer- Qiagen columns using 10 mM Tris–HCl (pH 8.5) as elution ase kit (Cat. no. 1338). Antisense cDNA was synthesized from buffer (Qiagen protocol), the combined probes are precipitated the full-length aRNA using an oligo(dT) primer and Super- by adding one-third volume of 7.5 M ammonium acetate, script II reverse transcriptase followed by removal of the RNA followed by 2.5 vol of absolute ethanol, and precipitation at template with RNase 1 and the purification of single-strand 80C for 20 min. The precipitates are collected by centri- antisense cDNA over a Qiagen PCR purification column (Cat.
fugation at 13 K in a microcentrifuge for 15 min and the pellets no. 28104). Purified products were measured by OD are washed with 75% ethanol and air-dried.
checked for correct size.
Immediately before REM hybridization, the pellet contain- A large batch of single-stranded liver cDNA was synthe- ing the combined Cy3- and Cy5-labeled probes is dissolved in sized from 2.5 mg of total RNA from a C57/Bl6 female mouse 20 ml of hybridization buffer containing 35% formamide, 0.5% and used as the carrier for all the LysA dilutions. LysA 1.1 kb SDS, 2.5· Denhardts solution, 4· SSPE, 0.2 mg/ml yeast antisense LysA cDNA was mixed with liver cDNAs at 12 tRNA, 0.1 mg/ml poly(dA) and 2.5 mg/ml mouse/human Cot levels each representing a 2-fold dilution of LysA per liver 1 DNA. The probe is boiled at 95C for 2 min, snap-cooled, cell cDNA equivalent. Our mixtures were based on a 50 mg/ml spun down in a microcentrifuge at 13 K for 5 min and pre- solution of 1000 bp segment of single-stranded DNA contain- hybridized at 50C for 1 h.
ing 9.1 · 1013 molecules of DNA per milliliter (21). Based onthe above standard, we made series dilutions in which LysA Preparation of REMs for hybridization was varied from 9000 to 4 copies per liver cell equivalent. Allmixtures were prepared in 3· SSC solution. Mixtures were Dust from the slide is removed with air from a Fisherbrand also based on 0.2 pg mRNA per liver cell.
super friendly Air'IT (Cat. no. 23-022523). The array face ofthe REM is moisturized over boiling water for 5 s and the DNA Preparation of LysA sense Cy dye-labeled probe for is immediately crosslinked to the slide with 250 mJ of UV REM hybridization irradiation in a Bio-Rad UV GS GENE LINKER. The slide isre-moisturized over steam for 5 s and placed (array side up) on A set of nested primers were used to generate a 533 bp sub- a 100C hot plate for 3–5 s. Then the slide is rinsed in 0.1% fragment of LysA from the 1.1 kb antisense cDNA produced SDS for 10–20 s, followed by ddH2O for 10–20 s and then above. The gene-specific primers for this PCR fragment were incubated at 95C in ddH2O for 3–5 min. The slide is dipped in as follows: 50-CGAGCAAAGCATTCTCATCA-sense and absolute ethanol and excess ethanol is removed by centrifuga- 50-T7 linked antisense primer TAATACGACTCACTATAG- tion in a 50 ml tube at 1000 r.p.m. for 4 min. The slide is GGCTCCTCCAAGATTCAGCAC. T7 RNA polymerase was placed, array side up, in a microarray slide hybridization used to generate an antisense LysA aRNA, the 533 bp Nucleic Acids Research, 2004, Vol. 32, No. 15 fragment [this aRNA does not contain either oligo(dT) or T7 Alb1 gene probe (NM_009654 mouse albumin 1), 50-GA- polymerase promoter sequences]. The final,
513 bp, sense CAAGGAAAGCTGCCTGAC-forward, 50-T7 (GTAATAC- strand Cy dye-labeled LysA probe was synthesized from 5 mg antisense aRNA using the sense strand primer, at 30 pmol, and reverse, product size 750 bp.
reverse transcriptase in the standard probe synthesis conditions Gapdh gene probe (NM_008084 mouse glyceraldehyde-3- described above.
Quantification of LysA by real-time PCR TGTGAGGGAGATGCTCAGTG-reverse, product size 599 bp.
TaqMan probe and primers were designed with Primer Primer sequences for UBI, Hnf4 and Igfbp1 are available upon Express Software (Applied Biosystems) and synthesized by request to CER. All specific PCR products were sequence- Operon (Qiagen) as follows: 50-GAAACGGGTCACTC- verified and used as templates for antisense RNA synthesis by CATCGA-forward primer; 50-AGTCATGCGTATGCGCT- in vitro transcription followed by labeling procedure as TCTAC-reverse primer; and 50-6FAM-TTCTTCTTCGGA- described above.
Gapdh Control Reagents containing VIC-labeled probe and Quantitative real-time PCR assay for human MYC primers (P/N 4308313; Applied Biosystems) were used to quantify a reference gene expression. Series dilutions of SMARTTMcDNAs (22–24), were diluted to obtain template LysA cDNA mixed with mouse liver cDNA were prepared amounts per reaction of 100, 200, 400 or 800 pg. These in a way such that a particular reaction mostly contained the amounts matched the amount of SMART cDNA printed on same amount of corresponding dilution that was printed on the the REM. For the five pairs of lung tumor/normal SMART REM slide. TaqMan Universal Master Mix (P/N 4304437) cDNAs, we tested three replications of each sample at the was used to prepare reaction mixtures containing 900 nM 400 pg per spot level and for the amplification efficiency of each primer and 250 nM of appropriate TaqMan probe.
two individual tumor/normal samples in all four different We performed a single gene reaction for LysA or Gapdh in concentrations were tested too.
each well. For each data point, we had three repetitions andused 96-well optical PCR plates (P/N 4306737; Applied TaqMan primers and probes. Assay on demand gene expres- Biosystems). The plates were sealed, spun down and reactions sion reagents were from Applied Biosystems. Each assay con- run in an ABI PRISM 7000 Sequence Detection System under sisted of forward and reverse primers and MGB (Minor default conditions: 50C for 2 min, 95C for 10 min, and 40 Groove Binder) probe with 6FAM at the 50 end and non- cycles of 95C for 15 s and 60C for 1 min.
fluorescent quencher at the 30 end mixed in 20· dilutions.
TaqMan Universal PCR master mix (P/N 4304437; Applied Primers for specific gene probes synthesis Biosystems) was diluted 2-fold with water and appropriate All primers were selected using Primer 3 public available amount of assay mixture, and aliquots of 20 ml were dispensed program which can be found at into wells on the reaction plate (P/N 4306737; Applied Biosystems). An aliquot of 5 ml, containing designated primers sequences were double checked for gene specificity amounts of SMART cDNA were added to the reaction mix- using available gene databases. The following genes and tures. Target gene and reference gene assays were run as single primers were selected.
reactions on the same plate. The following assays were used: MYC gene probe (NM_002467 Homo sapiens v-myc Hs00153408_m1 for MYC oncogene (NM_002467); 50- myelocytomatosis viral oncogene homolog) (avian), 50-AGAG- GCAGCGACTCTGAGGAGGAACAAGA, reporter position AAGCTGGCCTCCTACC-forward, 50-T7 (GTAATACGAC- is between exon 2 and 3; Hs00187842_m1 for beta-2-microglobulin (NM_004048), product size 632 bp.
forward primer 50-AGGCTATCCAGCGTACTCCAAAGAT, GP gene probe (X58295 plasma glutathione peroxidase 3), reporter position is between exon 1 and 2; 50-CATCTGACCGCCTCTTCTGG-forward, 50-T7 (GTAAT- Hs99999903_m1 for beta-actin (GenBank mRNA X00351), forward primer 50-TCGCCTTTGCCGATCCGCCGCCCGT, reverse, product size 308 bp.
reporter position is at exon 1.
ACTB gene probe (X00351 Homo sapiens cytoplasmic beta- Synthesis of cDNAs for printing Total RNA was isolated using the Qiagen RNA purification GCCGATCCACACGG-reverse, product size 384 bp.
procedure (Qiagen no. 75144) according to the manufacturer's B2M gene probe (NM_004048 Homo sapiens beta-2- instructions. RNA quality was monitored using an Agilent 2100 bioanalyzer (LabChip, Caliper Technologies Corp.).
Invitrogen Superscript III reverse transcriptase (Cat. no.
TTGCCAGCCCT-reverse, product size 578 bp.
180080-044) was used to synthesize cDNA from 100 mg of A 23 kDa highly basic protein (X56932 Homo sapiens total RNA using an Oligo dT primer. After synthesis was ribosomal protein L13A) (RPL13A), 50-TAAACAGGTACTG- completed, the samples were heated at 94C for 2 min, and CTGGGCCGGAAGGTG-forward, 50-T7 (GTAATACGA- then treated with 0.5 U/ml of RNase I and RNase I buffer [10 mM Tris–HCl (pH 7.5), 5 mM EDTA and 200 mM TAGC-reverse, product size 483 bp.
sodium acetate] (Promega no. 4261) for 10 min at 37C.
Nucleic Acids Research, 2004, Vol. 32, No. 15 Single-stranded cDNA was separated using the Qiaquick PCR pins, part no. SMP3, arranged in a 4 · 4 array, each producing purification protocol (Qiaquick Spin Handbook, p. 18), except a nominal 100 mm diameter spot.
that an additional 35% guanidine hydrochloride wash step wasincluded after binding cDNA to the Qiaquick column. cDNA Dot spacing. Each of the 16 pins forms a domain which was was eluted with 10 mM Tris–HCl, pH 8.5, and precipitated programmed to generate a uniformly spaced 12 · 12 square dot with one-third volume of 7.5 M ammonium acetate and pattern, with a center-to-center dot spacing of 365 mm.
2.5 vol of absolute ethanol. cDNA was pelleted, washed Printing parameters. The printing program was configured to with 75% ethanol and dissolved in water. Concentrations produce four replicates of each sample for every microscope were adjusted to 100, 200, 400, or 800 ng/ml in 3· SSC for slide. This subdivides each domain area into four subdomains printing. cDNA quality was monitored by running samples on containing 3 · 12 unique dots. With each pickup, each pin 1% agarose gels and checking for a smear of cDNAs from produces four equally spaced spots per domain, one each per
500–3000 bases in length.
subdomain, from the same sample. The on-slide dwell timewas 100 ms while the HEPA filtered environment was main- AECOM microarray printing procedure for REM tained at 25C and 50% RH.
Microscope slides. The substrate used was the Corning GAPS The REM microarrays were produced with the custom built II amino silane coated slides.
microarray printer at the AECOM Microarray Facility.
Details of the equipment can be viewed on our website ). Following is the printerconfiguration and parameters used for printing.
REMs are a reverse format microarray, in which the high-complexity ‘target' is bound to a solid support and labeled Printhead and pins. Telechem SPH48 printhead with pins probes from at least two genes are hybridized simultaneously spaced 4.5 mm center-to-center, populated with 16 split-tip to the microarray (Figure 1). The cDNA printed on glass Figure 1. Overview of REM technology. (A) Illustrates the probe preparation protocol starting with PCR using forward and reverse gene-specific primers linked to aT7 promoter. T7 RNA polymerase produces an anitsense RNA (blue line), and then reverse transcriptase produces a sense strand Cy3- or Cy5-labeled cDNA probe(magenta plus red or green Cy dye). The Cy3 and Cy5 probes are made single-stranded with RNases and then mixed prior to REM hybridization. (B) REM productionand processing. Total RNA from tissues or cells serves as template for reverse transcriptase to synthesize a cDNA primed by Oligo(dT) or a SMART cDNA. cDNAsare printed on Corning GAP slides. High stringency hybridization is carried out with mixed Cy3 and Cy 5 probes, in a humidified hybridization chamber followed bywashing, scanning and processing of the data using custom made scripts in a Linux operating system.
Nucleic Acids Research, 2004, Vol. 32, No. 15 microscope slides can either be single-stranded antisense Table 1. Quantitative analysis of albumin, Hnf-4 and Igfbp-1 expression in cDNA produced by reverse transcription, or the cDNA can mouse organs and Gapdh hybridization to murine organ cDNAs printed on a be rendered double-stranded by amplification using SMARTTM REM at 400 pg/spot DNA technology (22–24). The kinetics of hybridization are Ratio gene/Gapdh – SD assumed to be similar to that of RNA dot or northern blots in which the cloned probe is in hybridization excess around the complex mRNA (cDNA) that is bound to the solid support.
During the development of the technology, we tested sev- eral different solid supports and different densities of sample printing per spot. We tested total RNAs, poly(A)+ RNAs, A RNAs (25), single-stranded cDNAs and SMART cDNAs (22– 24). While RNA samples were successfully hybridized, they were very sensitive to RNase degradation, and cDNAs were found to be much more stable substrate for printing and hybrid- ization. Therefore, in this report we only present data using our current REM protocol that involves printing cDNAs on silane- coated glass microscope slides at cDNA spotting densities representing
1000–4000 cell equivalents per spot. REMs are hybridized simultaneously with a test probe, usually labeled with Cy5 (red) fluorescent dye and a housekeeping gene probe, usually labeled with a Cy3 (green) fluorescent dye (Figure 1). Hybridization signals are measured with a laser scanner (26), and fluorescence data are processed using gene pix software (Axon, Garden City, CA). Data sorting and analysis are carried out using customized computer scripts written using a Linux operating system, and plotted using Gnu Plot software.
Bold/italics show highest ratios for the test gene.
Organ-specific gene expression detected with REMs Our initial test of REM technology was to determine whetherwe could detect organ-specific hybridization of test probes.
We decided to use albumin as an example of an abundant liver-specific probe, Hnf4 as a liver preferential transcription factorprobe and insulin-like growth factor binding protein-1 (Igfbp-1)as a gene weakly expressed in the liver. Cy5-labeled albumin,Hnf4 and Igfbp-1 probes were synthesized along withCy3-labeled Gapdh probe. REMs were produced by printingsingle-stranded antisense cDNAs from a set of mouse organsat a density per spot that represented cDNA from 4000 cells(21). In the case of liver, we printed cDNAs from sixdifferent livers, representing one CD1 male, two C57Bl/6males, and one CD1 female and two C57Bl/6 females.
Each cDNA sample was printed in quadruplicate. Thus,the overall ratio of albumin/Gapdh for liver was calculatedfrom 48 quantitative fluorescence measurements (six liversamples, hybridized). Other mouse organs were also represented bymultiple samples and each was also quadruplicate spotted.
We hybridized a REM-containing cDNAs from 25 mouse organs with Cy5-labeled albumin plus Cy3-labeled Gapdhprobes (Table 1). The hybridization revealed strongly redspots for liver and green spots for all the other organs, asexpected. Examples of hybridized spots, viewed as the com-bined Cy5–Cy3 computer image, are shown in Figure 2A.
Using a customized computer script in Linux, we calculatedthe ratio of albumin signal versus Gapdh for the entire set ofmouse organs (Table 1). The ratio for albumin was 10.76 – 4.08 Figure 2. (A) Organ-specific hybridization of albumin to liver cDNA. A mouseorgan REM was hybridized with Cy 5-labeled albumin and Cy3-labeled Gapdh for liver, whereas the average ratio for the other organs was probes. The combined computer image shows red for liver cDNA spots and 0.25 – 0.1, clearly demonstrating the strong liver-specific green spots for all other organs. (B) Sorting of liver-specific albumin expression according to sex and genotype of donor mouse.
Nucleic Acids Research, 2004, Vol. 32, No. 15 However, the standard deviation for liver-specific hybridiza- 9000 copies per cell equivalent (i.e. 1.8 · 107 copies per tion was very high, suggesting an unexpectedly high level of 400 pg of sample) to approximately two copies per cell equiva- variability in albumin expression between the different liver lent (4 · 104 copies per 400 pg of sample) (21). The results samples. To investigate this, we sorted the liver data according from a set of standard mixtures, printed in quadruplicate and to sex, or genotype, of the mice from which the liver samples hybridized simultaneously with a green (Cy3) LysA probe and were taken. By this analysis, we observed no difference due to a red (Cy5) Gapdh reference probe are shown in Figure 3A.
mouse genotype; however, we observed a significant increase The computer combined image shows that spots containing the in albumin expression in female liver cDNAs (Figure 2B).
high level of LysA are green and those with a low or unde- This makes sense biologically since an important function tectable LysA level are red, representing solely Gapdh refer- of albumin is as a serum carrier protein for estrogen in females ence gene hybridization. A dye reversal experiment revealed a (27). These data demonstrate that the REM technology can reversed pattern of colored spots, demonstrating the accuracy reveal new information about gene expression differences due and reproducibility of the hybridization and detection technol- to sex and/or genotype.
ogy (Figure 3B).
A second mouse organ REM was hybridized with Hnf4 plus Quantitative analysis of the hybridization signals from Gapdh. Analysis of the hybridization again revealed liver pre- Figure 3A allowed us to calculate a ratio for the LysA gene ferential expression as expected along with expression in all versus Gapdh across the standard curve of 400 pg spots other organs tested (Table 1). Hybridization of a third REM (Figure 4). These data showed an increasing ratio from 4 to with Igfbp-1 and Gapdh probes revealed the strongest Igfbp-1 9000 copies of LysA per liver cell cDNA equivalent. A duplicate hybridization in the pancreas and spleen, in contrast to liver set of mixtures printed at 800 pg per spot produced a standard expression (Table 1).
curve that was virtually identical to that obtained with the 400 pgper spot series (Figure 4). Therefore, the ratio of test gene versus Normalization with housekeeping genes reference gene is independent of the density of spotting.
It is not possible to be certain that every spot on a REM isequally loaded with cDNA. Therefore, as stated above, it is Comparison with real-time quantitative PCR essential to have an internal housekeeping gene control. The We compared REM technology to quantitative real-time PCR data in Table 2 were generated from a separate REM printing, by analyzing six of the standard mixtures from the above in which several cDNA samples including muscle and brain analysis by both technologies. The real-time PCR data, were inadvertantly overloaded. Thus, for example, albumin expressed as a negative log of the Ct value (28), and the hybridization was very high in the muscle cDNA. However, REM data, expressed as the log 2 of the LysA/Gapdh ratio, after normalization with Gapdh, the ratio for albumin expres- are plotted on the y-axis in Figure 5. The two data sets are sion versus Gapdh in muscle is very low (0.08) as was compared across a set of known amounts of LysA, expressed expected. These data demonstrate that the test gene expression as the log 2 of the LysA copy number per reaction or spot.
can be accurately normalized against an internal reference for This plot shows a striking parallel from an abundance of quantitative analysis even when spots contain highly variableamounts of cDNA.
Standard curve generated with artificial mixtures We next set up an experiment to determine the accuracy ofREM technology for detecting rare transcripts in a complexliver cDNA mixture. We prepared artificial cDNA mixtures inwhich we spiked a liver cDNA preparation with various levelsof the bacterial gene, diaminopimelate decarboxylase (LysA;ATCC accession number 87482). These mixtures were printedon silane-coated glass microscope slides at 400 or 800 pg ofliver cDNA per spot. An aliquot of 400 pg cDNA representsthe cDNA from approximately 2000 hepatocytes and thelevels of spiked LysA cDNA ranged from approximately Table 2. Data demonstrating effectiveness of internal normalization approachfor calculating albumin expression in murine organs Albumin Intensity ALB Gapdh (ratio) Figure 3. Hybridization of a set of standard liver cDNA mixtures containing increasing amounts of bacterial LysA antisense cDNA. (A) LysA abundance varying from approximately 9 to 9100 copies LysA cDNA per cell equivalent are shown in this figure (left vertical labels). Mixtures of cDNAs were printed at 400 pg/spot (right vertical labels). LysA was labeled with Cy3 (green) and Gapdh with Cy5 (red), and the green image corresponds to high LysA. (B) Dye reversal experiment, LysA was labeled with Cy 5 and GAPDH with Cy 3, andhigh LysA is a red image.

Nucleic Acids Research, 2004, Vol. 32, No. 15 Another approach is the cDNA amplification method termedas SMART (switching mechanism at the 50end of the RNAtemplate). The SMART method (22,24) utilized a combinationof two primers in a single reverse transcription reaction. Atagged oligo(dT) primer is used to prime the first cDNA strandwhile the SMART oligonucleotide serves as a short, extendedtemplate at the 50 end of the RNA templates. When the reversetranscriptase reaches the 50 end of the mRNA, the enzymeswitches templates and continues replicating to the endof the SMART oligonucleotide. PCR amplification isnow initiated with primers complimentary to the 30 anchorand SMART oligonucleotide. This protocol uses a mini-mum number of PCR amplification cycles (
15) andSMART cDNAs have been shown to preserve the relativeabundance of different mRNAs in complex cDNA mixtures(33–36).
Figure 4. Standard curve for hybridization of increasing bacterial LysA gene We printed SMART cDNAs synthesized from mRNA iso- versus constant Gapdh in liver cDNA mixtures (from Figure 3B). The ratio of lated from five tumor/normal pairs from five major tumor the LysA fluorescence signal intensity versus the Gapdh signal intensity is types, including kidney, breast, uterus, lung and ovary. The plotted as the log 2 of LysA/Gapdh fluorescence intensities (y-axis).
A 2-fold increase in the LysA abundance in the liver cDNA is plotted on SMART cDNAs were printed at 100, 200, 400 and 800 pg per the x-axis as the log 2 of the LysA copy number per cell equivalent. The spot, and each SMART cDNA sample was printed in log 2 values on the x-axis represent the following LysA copy numbers per quadruplicate. Hybridization of a REM-containing SMART liver cell equivalent of cDNA, log 2.1 = 4, 3.1 = 9, 4.1 = 18, 5.1 = 36, etc. cDNA cDNAs, with a single-stranded antisense Cy5-labeled probe mixtures were printed at two densities, either 400 pg total liver cDNA per spot(diamonds) or 800 pg liver cDNA per spot (‘X'). An aliquot of 800 pg spots to MYC, and a Cy3-labeled probe to beta 2 microglobulin represent approximately 4000 cell equivalents of liver cDNA.
(b2M), as a housekeeping reference, produced significantfluorescence signals across the whole range of printing den-sities (Figure 6A). The quadruplicate printing of each sampleenabled us to calculate confidence intervals for each sampleand draw a conclusion whether MYC was up- or down-regulated in the tumor from each tumor/normal pair. In thecase of lung tumors, shown in Figure 6B, we concluded thatMYC was up-regulated in all five tumors (100%). Up-regulation of MYC in the lung tumor samples was confirmedusing quantitative real-time PCR. We calculated the ddCTvalue (37) for each tumor sample versus its matching normalsample (numbers above each tumor/normal pair in Figure 6B).
A negative ddCT means that MYC was more abundant in thetumor sample compared to its matching normal sample.
The ratio for MYC expression versus b2M was calculated Figure 5. Comparison of REM technology with quantitative real-time PCRREM data (diamonds); quantitative real-time PCR (circles). Left y-axis: log for all 200 SMART cDNA spots representing 25 tumor/normal 2 of the dCt value for LysA concentration by real-time PCR. Right y-axis: log 2 pairs, quadruplicate spotted. This survey showed a predomi- of the ratio of LysA/Gapdh fluorescence intensities for standard mixtures from nant up-regulation of MYC in lung and ovary tumors and REM data (Figure 4). x-axis: log 2 of the LysA cDNA copy number per spot.
down-regulation in kidney tumors (Table 3). In contrast, how-ever, MYC was predominantly unchanged in our group ofbreast and uterus tumors (Table 3). The highest MYC up- approximately four copies (log 2, 15.12) to 9100 copies (log 2, regulation was found in two lung tumors that had 4.3- and 25.12) of LysA per cell equivalent. Therefore, over a 2000- 5.7-fold increases, and the most significant down-regulation of fold change in LysA abundance, REM technology is equal to MYC was in kidney tumors.
quantitative real-time PCR in accuracy and sensitivity.
Different housekeeping genes and printing density yield SMARTTM cDNAs used for REM production and gene similar results in REM technology expression analysis The choice of reference gene may be important in certain A key feature of REM technology is its ability to represent a samples because routinely used housekeeping genes, such broad range of pathophysiological paradigms on a single pro- as Gapdh, are themselves regulated in certain instances. There- duct. However, valuable biological samples, such as biopsies fore, in order to measure variability in REM data with different or samples obtained by laser capture micro-dissection provide reference genes, we chose three different housekeeping genes, only small amounts of mRNA (29). This requires a method beta Actin, Ubiquitin and 23 kDa basic protein and hybridized for amplification of the mRNA population while maintaining them against a common test gene, glutatione peroxidase (GP), the relative balance in abundance between mRNA species in three separate REMs. In addition, we prepared a mixture of (30–32). One approach is the production of ARNA (25).
the three reference probes and hybridized the mixture against Nucleic Acids Research, 2004, Vol. 32, No. 15 Figure 6. (A) Combined fluorescence image of a REM hybridized simultaneously with red (MYC) and green (b2M) probes. Image from segment of REM containingSMART cDNA samples of paired tumor/normal tissues is shown. Horizontal rows: images of four replicate sets of each tumor/normal pair (8 spots/row). Verticalcolumns: 12 tumor/normal pairs printed at either (a) 800 pg or (b) 400 pg or (c) 200 pg or (d) 100 pg SMART cDNA/spot. (B) Histogram of MYC expression in lungtumor and normal samples using b2M as the reference gene. Error bars represent standard deviation determined from four measurements of the cMYC/b2M ratio foreach sample on the REM. The ddCT values determined by real-time PCR are above each pair. A negative ddCT value means MYC was higher in tumor tissuecompared to companion normal tissue by quantitative real-time PCR analysis. ddCT represents the difference in the number of real-time PCR cycles to reachmaximum rate of amplification between tumor and normal paired samples.
Table 3. Quantitative analysis of MYC expression in a panel of 25 tumor/normal pairs SMART cDNAs from five pairs each of lung, uterus, breast, ovary and kidney tumors are shown. Common reference gene was b2M. Data are from the400 pg/spot series.
GP. The aim of this fourth REM was to determine whether the lung tumor/normal pairs from four REMs was representative mixture of reference probes accurately reflected the results of all the tumor types and is shown in Figure 7.
with each individual reference probe. All the probes were This analysis showed that GP was lower in lung tumor/ human genes and were hybridized to a REM containing the normal pairs 1–4 and nearly the same in lung tumor/normal human tumor/normal pairs of SMART cDNAs. The data for pair 5, in each of the four REMs. This confirmed that different Nucleic Acids Research, 2004, Vol. 32, No. 15 Figure 7. Common GP expression profiles obtained using three different reference genes and a mixture of the three reference genes. Data shown for five lung tumor/normal pairs from four separate REMs hybridized with the designated probes. Upper left, GP versus beta actin (ACTB); upper right, GP versus ubiquitin (Ubi); lowerleft, GP versus 23 kDa basic protein; and lower right, GP versus a mixture of all three reference genes. For each sample, the ratio of GP/reference signal is plotted.
Standard errors are calculated from quadruplicate spotting of each sample.
reference genes can be used with qualitatively similar results.
As expected the absolute ratios were different for each REM In this report, we have validated REM technology for measur- due to the different levels of hybridization of each housekeep- ing the expression of test genes in a diverse spectrum of ing gene. However, the relative differences between tumor and biological samples in a high-throughput manner. We have normal samples were very similar across the set of four inde- demonstrated that the REM technology can detect organ pre- pendently hybridized REMs. The REM hybridized with the ferential gene expression of both abundant transcripts such as combined set of three reference genes closely reflected albumin in the liver, and rare transcripts such as hepatocyte the data from each reference gene singly. Therefore, the nuclear factor 4 (Hnf4) in the liver and insulin-like binding use of a combined reference probe may be preferable for protein 1 (Igfbp1), expression in pancreas and spleen (Table 1).
REM hybridizations.
We also detected MYC oncogene expression in both tumor Finally, we compared the tumor/normal data across sets of and normal human tissue samples and confirmed the differ- identical samples that were printed at different printing ential regulation with quantitative real-time PCR.
densities. Data from a representative set of breast tumor/ One feature of REM technology is that it can be used to normal pairs printed at 200, 400 and 800 pg per spot and measure gene expression differences that are due to sex of the the average data for all three spotting densities are shown in individual. For example, in a prototype REM, we included Figure 8. Overall, there is a striking similarity between the liver cDNAs from male and female mice that were either datasets, supporting the conclusion that all three printing C57Bl6 or CD1 genetic backgrounds. Using the REM, we densities are suitable for REM analysis. In this example, showed that albumin expression was not different in livers GP was down-regulated in tumors in pairs 1 and 3, equal of mice from different genetic backgrounds; however, the to normal in tumor/normal pairs 2 and 4 and very slightly up sex of the mouse had a significant effect on albumin expression in tumor from pair 5. Each sample is quadruplicate spotted (Figure 2). The data showed that albumin expression is sig- and therefore each bar in the average profile (D) represents nificantly higher in female liver. This is not generally appre- 12 data points obtained at 3 densities for each sample. This ciated, however, it is consistent with the biological functions type of multiple sampling is a unique strength of REMs that of albumin which include being the major serum binding facilitates accurate standard error measurements and the protein for the female hormone, estrogen (27). Second detection of small differences between tumor and normal generation REMs, which contain 5–10 replicate organ samples, from both male and female mice of different genetic Nucleic Acids Research, 2004, Vol. 32, No. 15 Figure 8. Common expression profiles obtained for tumor/normal samples printed at different densities. Data shown for five breast tumor/normal pairs hybridizedwith GP and ubiquitin (UBI). SMART cDNAs were printed at three printing densities. (A) An aliquot of 200 pg of SMART cDNA per spot; (B) 400 pg of SMARTcDNA per spot; (C) 800 pg of SMART cDNA per spot; and (D) average data from all three levels. The REM containing human tumor/normal pairs was hybridizedwith human GP probe plus human ubiquitin probe. For each sample, the ratio of GP/ubiquitin signal is plotted. Standard errors are calculated from quadruplicatespotting of each sample. Data shown are for five kidney tumor/normal pairs.
backgrounds will have the unique ability to detect previously function well for samples that were apparently 10 or more fold unappreciated gene expression differences due to sex and different in loading per spot (Table 2). Second, when we purposely loaded different amounts of the same samples on Any microarray technology that utilizes printing of nucleic a REM, we repeatedly observed sets of ratios of test gene to acids must have a means to control for variability in printing normalization gene that were virtually identical across cDNA density that invariably occurs between samples. The need for spotting densities from 200 to 800 pg/spot (Figure 8). The controls for differential loading are one of the main limitations same degree of reproducibility and sensitivity of gene expres- of earlier RNA and cDNA dot blots (19). Using nylon arrays, it sion differences detected through this multiple spotting has been necessary to elute a first probe and rehybridize the approach has not been previously reported for microarrays.
array with a second housekeeping gene in order to normalize In a pilot experiment, we tested whether REMs could be the signals for the first probe. The use of a glass printing format eluted and re-hybridized with two new probes. We success- and co-hybridization with two fluroescent dyes has eliminated fully eluted the samples and re-hybridization yielded signifi- the need for re-hybridization for REM technology.
cant signals that generally had similar ratios of test to reference Two lines of evidence in this report support the conclusion genes. However, the signals were reduced in intensity com- that REMs can be accurately normalized by co-hybridization pared to the first hybridization (data not shown). Therefore, the with a housekeeping gene. First, we showed that cDNAs from pilot data strongly suggest that conditions will be found that muscle and other organs, which do not express albumin, when will enable the re-hybridization of REMs, thus greatly extend- normalized against Gapdh hybridization, reveal a very low ing their usefulness.
ratio of albumin/Gapdh that is essentially equivalent to back- Another possible expanded use of REMs includes the use of ground (Table 2). We showed that this internal normalization more than two labeled probes per hybridization. Theoretically, Nucleic Acids Research, 2004, Vol. 32, No. 15 the number of probes that could be hybridized simultaneously a mixture of reference genes produced datasets that were will only be limited to the number of fluorescence signals that virtually indistinguishable from those of REMs hybridized can be distinguished by a laser detector. Therefore, it may be with only a single reference gene (Figure 7). Therefore, the possible to simultaneously hybridize REMs with sets of probes use of standard mixtures of reference genes can control for that detect as many as 5–10 genes in a particular pathway, both loading differences and small variations in expression of housekeeping genes.
The sensitivity and accuracy of REM technology was We investigated the expression of the oncogene, MYC in the demonstrated by the use of artificial mixtures. Data from a SMARTTMcDNA tumor/normal pairs. Data from quadru- prototype REM that contained standardized mixtures of a plicate spotted samples enabled us to calculate confidence bacterial gene, LysA into a liver cDNA showed that a specific intervals for the ratios for each tumor and normal sample.
hybridization signal was detected when as few as four copies Therefore, we were able to draw conclusions as to differential of the LysA cDNA were present per liver cell cDNA equiva- oncogene expression in the paired samples at a level of sensi- lent. Furthermore, the LysA/Gapdh ratio increased in a near tivity not previously possible for RNA dot blots (19). In a linear pattern until 9000 copies per cell was reached. This single REM containing 25 tumor/normal pairs from five relationship was repeated whether we printed the liver tumor types, we were able to calculate the relative MYC cDNAs at 400 or 800 pg/spot demonstrating again the wide expression and reach conclusions as to whether MYC was range of the experimental sample loading for which accurate up, down or unchanged in the whole panel of tumor samples measurements can be obtained (Figure 4).
(Table 3). This survey serves as one example of how REM The accuracy of the standard curve data was tested by technology can be used in cancer research. Since thousands of analysis of the standard mixtures with quantitative real-time samples can be printed on a single REM, REM technology can PCR. In this analysis, quantitative real-time PCR and REM provide a high-throughput approach to testing candidate gene data closely paralleled each other from samples with 4 to 9000 expression in diverse tissues and tumors.
copies of LysA per liver cell cDNA equivalent. Therefore, we The only other array based approach designed to test sam- conclude that REM technology is equivalent to quantitative ples from multiple tissue types simultaneously is tissue micro- real-time PCR over at least a 2000 to 4000-fold change of arrays (20). These arrays contain thin sections of multiple tissues on a microscope slide, allowing an investigator to The SMART method has been successfully applied to the determine gene expression using antibodies to detect protein generation of full-length cDNA libraries (24), and as a source in cells. Also, in situ hybridization of tissue microarrays pro- for cDNA probes for GEMS from RNA obtained by laser vides information on the expression of a gene in specific cell capture microdissection (32). When we printed SMART types. However, antibody staining and in situ hybridization are DNAs on a REM at multiple printing densities of spotting, not quantitative technologies and are labor intensive. Further- the ratio of expression of test genes were nearly identical in the more, due to the nature of the experimental approach each range of spotting densities (Figure 7).
section cut from the tissue microarray is different from the There has been much discussion in the literature about the previous section. Also, these arrays are generally limited to preservation of original RNA representation after mRNA/ fewer than one hundred samples, whereas REMs can easily cDNA amplification. It is generally assumed that linear ampli- accommodate thousands of samples that are spotted more than fication is superior to exponential amplification due to biases once for quantitative analysis.
in abundance relationships (38,39). However, this assumption Therefore, REM Technology fulfills an important need for a does not hold up on closer examination of the recent literature.
high-throughput, sensitive, accurate and quantitative method Wang et al. (40) has pointed out that conventional T7-based to measure gene expression simultaneously in multiple tissues RNA amplification can introduce biases in the amplification or cell types, at a time in biological research when there is a because of a possible 50-under representation and because low strong emphasis on quantitative expression analysis. REMs stringency temperatures are applied during generation of the provide a platform on which to build libraries of samples double-stranded cDNA. SMART cDNA amplification from that can be used to characterize specific functions of genes total RNA was found to preserve representation of high, in specific biological contexts. In addition to general survey medium and low abundance mRNAs and compared favorably REMs that contain samples from organs, tissues and cell types, to quantitative northern-blot analysis (41). Additionally, specialty REMs that have experimental samples designed to SMART cDNA generated signals expressed nearly identical ask specific questions about regulation of a gene in specific patterns to unamplified total RNA probes upon hybridization cellular contexts, and developmental contexts, can be designed to 4600 arrayed genes in a GEM analysis (41).
and produced. The future content and number of specialty It is known that genes that are generally considered to REMs (such as liver, kidney, heart, tumor profiles, develop- be houskeeping genes, such as Gapdh, are differentially mental stages and gene knockout REMs) and their application expressed under various experimental conditions. Therefore, to biology is virtually unlimited. Therefore, we envision the it would be advantageous to be able to utilize a mixture of library of REMs as continuing to grow and the impact of housekeeping genes for normalization, in order to control for REMs on biological research to increase with time. The avail- minor variations in any one of the genes. In this report, we ability of REMs that have samples from classic experiments have utilized four different normalization genes including will provide researchers with access to relate their current Gapdh, beta Actin, Ubiquitin and 23 kDa basic protein. In research directly to historically validated paradigms. The all the cases, where the same test gene was tested against use of reference REMs designed to ask questions in the two or more reference genes, the data were qualitatively simi- area of toxicology and pharmaceutical research may also lar (Figure 7). Furthermore, data from REMs hybridized with gain use in the drug approval process. REMs can essentially Nucleic Acids Research, 2004, Vol. 32, No. 15 ‘immortalize' specific experiments that can be printed thou- 16. Takahashi,M., Rhodes,D.R., Furge,K.A., Kanayama,H., Kagawa,S., sands of times and be widely distributed.
Haab,B.B. and Teh,B.T. (2001) Gene expression profiling of clear cellrenal cell carcinoma: gene identification and prognostic classification.
Proc. Natl Acad. Sci. USA, 98, 9754–9759.
17. Bittner,M., Meltzer,P., Chen,Y., Jiang,Y., Seftor,E., Hendrix,M., Radmacher,M., Simon,R., Yakhini,Z., Ben-Dor,A. et al. (2000) Molecular classification of cutaneous malignant melanoma by geneexpression profiling. Nature, 406, 536–540.
The authors thank Dr Liang Zhu and Dr David Shafritz 18. Kitahara,O., Furukawa,Y., Tanaka,T., Kihara,C., Ono,K., Yanagawa,R., for critical reading of this manuscript. This work was Nita,M.E., Takagi,T., Nakamura,Y. and Tsunoda,T. (2001) supported by funding from the National Institute of Health Alterations of gene expression during colorectal carcinogenesis revealed by cDNA microarrays after laser-capture microdissection (5P30DK41296) and Einstein Comprehensive Cancer Center of tumor tissues and normal epithelia. Cancer Res., 61,3544–3549.
grant (CA13330-32).
19. Kafatos,F.C., Jones,C.W. and Efstratiadis,A. (1979) Determination of nucleic acid sequence homologies and relative concentrationsby a dot hybridization procedure. Nucleic Acids Res., 7,1541–1552.
20. Kononen,J., Bubendorf,L., Kallioniemi,A., Ba¨rlund,M., Schraml,P., Leighton,S., Torhorst,J., Mihatsch,M.G., Sauter,G. and Kallioniemi,O.P.
1. Brown,P.O. and Botstein,D. (1999) Exploring the new world of the (1998) Tissue microarrays for highthroughput molecular profiling of genome with DNA microarrays. Nature Genet. (suppl.), 21, 33–37.
tumor specimens Nature Med., 4, 844–847.
2. Lewin,B. (1997) Eukaryotic gene expression, chapters 28 and 29. In 21. Sambrook,J. and Russell,D.W. (2001) Molecular Cloning: A B.Lewin(ed.), Genes VI. Oxford University Press, Oxford, NY and Tokyo.
Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold Spring 3. Watson,J.D., Hopkins,N.H., Roberts,J.W., Steitz,J.A. and Weiner,A.M.
(1987) The steps in protein synthesis, chapters 13–15. In Gillen,J.R. (ed.), 22. Chenchik,A., Zhu,Y.Y., Diatechenko,L., Li,R., Hill,J. and Siebert,P.D.
Molecular Biology of the Gene. Benjamin/Cummings Publishing Co., (1998) Generation of high quality cDNA from small amounts of total Menlo Park, CA.
RNA by SMART PCR. In Siebert,P.D. and Larrick,J.W. (eds), Gene 4. Shalon,T.D. (1995) DNA microarrays: a new tool for genetic analysis.
Cloning and Analysis of RT–PCR. Eaton Publishing, Natick, MA, PhD Thesis, Stanford University, UMI Dissertation services, Ann Arbor, pp. 305–319.
MI, pp. 29–33.
23. Zumabayeva,B., Diatchenko,L., Chenchik,A. and Siebert,P.D. (2001) 5. Schena,M., Shalon,D., Davis,R.W. and Brown,P.O. (1995) Quantitative Use of SMART generated cDNA for gene expression studies in monitoring of gene expression patterns with a complementary DNA multiple human tumors. BioTechniques, 30, 1–7.
microarray. Science, 270, 467–470.
24. Zhu,Y.Y., Machleder,E.M., Chenchik,A., Li,R. and Siebert,P.D. (2001) 6. Schena,M., Shalon,D., Heller,R., Chai,A., Brown,P.O. and Davis,R.W.
Reverse transcriptase template switching: a SMARTTM approach for (1996) Parallel human genome analysis: microarray-based expression full-length cDNA library construction. BioTechniques, 30, monitoring of 1000 genes. Proc. Natl Acad. Sci. USA, 93, 10614–10619.
7. Iyer,V.R., Eisen,M.B., Ross,D.T., Schuler,G., Moore,T., Lee,J.C.F., 25. Eberwine,J. (1996) Amplification of mRNA populations using aRNA Trent,J.M., Staudt,L.M., Hudson,J.,Jr, Boguski,M.S., Lashkari,D., generated from immobilized oligo(dT)-T7 primed cDNA.
Shalon,D., Botstein,D. and Brown,P.O. (1999) The transcriptional BioTechniques, 20, 584–591.
program in the response of human fibroblasts to serum. Science, 283, 26. Cheung,V.G., Morley,M., Aguilar,F., Massimi,A., Kucherlapati,R. and Childs,G. (1999) Making and reading microarrays. Nature Genet., 21, 8. Robert,R., Klevecz,J.B., Forrest,G. and Murray,D.B.A. (2004) Genome wide oscillation in transcription gates DNA replication and cell cycle.
27. Andre,C., Jacquot,Y., Truong,T.T., Thomassin,M., Robert,J.F. and Proc. Natl Acad. Sci. USA, 101, 1200–1205.
Guillaume,Y.C. (2003) Analysis of the progesterone displacement of its 9. Plescia,C.P., Rogler,C.E. and Rogler,L.E. (2001) Genomic expression human serum albumin binding site by beta-estradiol using analysis implicates Wnt signaling pathway and extracellular matrix biochromatographic approaches: effect of two salt modifiers.
alterations in hepatic specification and differentiation of murine hepatic J. Chromotogr. B, 769, 267–281.
stem cells. Differentiation, 68, 254–269.
28. Winer,J., Kwang,C., Jung,S., Shackel,I. and Williams,P.M. (1999) 10. Yu,Y., Khan,J., Khanna,C., Helman,L., Meltzer,P.S. and Merlino,G.
Development and validation of real-time quantitative reverse (2004) Expression profiling identifies the cytoskeletal organizer ezrin and transcriptase–polymerase chain reaction for monitoring gene expression the developmental homeoprotein Six-1, as key metastatic regulators.
in cardiac anal. Biochemistry, 270, 41–49.
Nature Med., 10, 175–181.
29. Emmert-Buck,M.R., Bonner,R.F., Smith,P.D., Chuaqui,R.F., Zhuang,Z., 11. Mathiassen,S., Lauemoller,S.L., Ruhwald,M., Claesson,M.H. and Goldstein,S.R., Weiss,R.A. and Liotta,L.A. (1996) Laser capture Buus,S. (2001) Tumor-associated antigens identified by mRNA microdissection. Science, 27, 467–1001.
expression profiling induce protective anti-tumor immunity.
30. Luo,L., Salunga,R.C., Guo,H., Bittner,A.K., Joy,C., Galindo,J.E., Eur. J. Immunol., 31, 1239–1246.
Xias,H., Rogers,K.E., Wan,J.S., Jackson,M.R. et al. (1999) Gene 12. Kaminski,N., Allard,J.D., Pittet,J.F., Zuo,F., Griffiths,M.J.D., Morris,D., expression profiles of laser-captured adjacent neuronal subtypes. Nature Huang,Z., Sheppard,D. and Heller,R.A. (2000) Global analysis of gene Med., 5, 117–121.
expression in pulmonary fibrosis reveals distinct programs regulating 31. Luzzi,V., Mahadevappa,M., Raja,R., Warrington,J.A. and Watson,M.A.
lung inflammation and fibrosis. Proc. Natl Acad. Sci. USA, 97, (2003) Accurate and reproducible gene expression profiles from laser capture microdissection, transcript amplification, and high 13. Perou,C.M., Jeffrey,S.S., van de Rijn,M., Rees,C.A., Eisen,M.B., density oligonucleotide microarray analysis. J. Mol. Diagn., 5, Ross,D.T., Pergamenschikov,A., Williams,C.F., Zhu,S.X., Lee,J.C. et al.
(1999) Distinctive gene expression patterns in human mammary 32. Fink,L., Kohlhoff,S., Stein,M.M., Hanze,J., Weissmann,N., Rose,F., epithelial cells and breast cancers. Proc. Natl Acad. Sci. USA, 96, Akkayagil,E., Manz,D., Grimminger,F., Seeger,W. and Bohle,R.M.
(2002) cDNA array hybridization after laser-assisted microdissection 14. Perou,C.M., Sorlie,T., Eisen,M.B., van de Rijn,M., Jeffrey,S.S., from nonneoplastic tissue. Am. J. Pathol, 160, 81–90.
Rees,C.A., Pollack,J.R., Ross,D.T., Johnsen,H., Akslen,L.A. et al. (2000) 33. Wang,E., Miller,L.D., Galen,A., Ohnmacht,G.A., Liu,E.T. and Molecular portraits of human breast tumours. Nature, 406, 747–752.
Marinocola,F.M. (2000) High-fidelity mRNA amplification for gene 15. Allzaeh,A.A., Eisen,M.B., Davis,R.E., Ma,C., Lossos,I.S., profiling. Nat. Biotechnol., 18, 457–459.
Rosenwald,A., Boldrick,J.C., Sabet,H., Tran,T., Yu,X. et al. (2000) 34. Kestler,D.P., Hill,M., Agarwal,S. and Hall,R.E. (2000) Use of Distinct types of diffuse large B-cell lymphoma identified by gene bidirectional blots in differential display analysis. Anal. Biochem., 280, expression profiling. Nature, 403, 503–511.
Nucleic Acids Research, 2004, Vol. 32, No. 15 35. Becker,B., Vogt,T., Landthaler,M. and Stolz,W. (2001) Detection 38. Baugh,L.R., Hill,A.A., Brown,E.L. and Hunter,C.P. (2001) Quantitative of differentially regulated genes in keratinocytes by cDNA array analysis of mRNA amplification by in vitro transcription.
hybridization: Hsp27 and other novel players in response to artificial Nucleic Acids Res., 29, 1–9.
ultraviolet radiation. J. Invest. Dermatol., 116, 983–988.
39. Dixon,A.K., Richarson,P.J., Pinnock,R.D. and Lee,K. (2000) Gene- 36. Vernon,S.D., Unger,E.R., Rajeevan,M., Dimulescu,I.M., Nisenbaum,R.
expression analysis at the single-cell level. Trends Pharmacol., 21, 65–70.
and Campbell,C.E. (2000) Reproducibility of alternative probe synthesis 40. Wang,E., Miller,L.D., Ohnmacht,G.A., Liu,E.T.and Marincola,F.M.
approaches for gene expression profiling with arrays. J. Mol. Diagn., (2000) High-fidelity mRNA amplification for gene profiling.
Nat. Biotechnol., 18, 457–459.
37. Livak,K. and Schmittgen,T.D. (2001) Analysis of relative gene 41. Endege,W.O., Steinmann,K.E., Boardman,L.A., Thibodeau,S.N. and expression data using real-time quantitative PCR and the 2-DDCt method.
Schlegel,R. (1999) Representative cDNA libraries and their utility in gene Methods, 25, 402–408.
expression profiling. BioTechniques, 26, 542–550.

Source: http://gene-quantification.biz/rogler-array-2004.pdf