The differential display by PCR (DD-PCR) technique (1,2) was conceived to allow the identification and molecular cloning of differentially expressed genes. This technique was devised to amplify messenger RNAs and display their 3' termini on polyacrylamide gels. We have simultaneously compared the gene expression patterns of 12 different populations representing various tissues or in vitro-derived hematopoietic cells in order to identify genes which are regulated during hematopoietic development. The results of this gene search will be published elsewhere (3). In this report we document several important considerations in the use of DD-PCR. We have optimized the conditions to detect more bands per run and have demonstrated that DD-PCR allows the cloning of portions ofmRNAs other than just their 3' termini, due to priming by one or both primers. Importantly, we have shown that DD-PCR allows the molecular cloning of very low abundance genes. We demonstrate that DD-PCR can be applied, in a reproducible manner, to the study of questions related to temporo-spatial patterns of gene expression requiring comparisons among a large number of mRNA populations. ES cells and embryoid bodies were prepared (4), the fibroblastic STO cell line (American Type Cell Culture, ATCC CRL 1503), the neuronal Neuro-2a (American Type Cell Culture, ATCC CCL 131) and the multipotential hematopoietic precursor cell line FDCPmixA4 (5) were cultured (American Type Cell Culture, 4), and mouse embryos were dissected as described (6). RNA was isolated using RNAzol solution (Tel-test, Inc., Friendswood, TX) and DNase treatment was performed using the MessageClean Kit (GenHunter Corporation, Brookline, MA) following manufacturer's instructions. RNA samples were quantitated by absorbance at 260 nm and evaluated on formaldehyde gels. High quality RNA was found to be imperative for the success of the subsequent steps. For reverse transcription, PCR and polyacrylamide gel electrophoresis, the manufacturer's recommendations were followed (RNAmap Kit, GenHunter Corporation). Primers AP3, AP5 and T12MC, synthesized using the sequence information provided in the RNAmap kit (DNAX), AmpliTaq DNA polymerase (Perkin Elmer-Cetus, Norwalk, CT) and 35S-dATP (Amersham, Arlington Heights, IL) were used. PCR tubes were introduced in the thermocycler when the block was either at room temperature (cold start) or after pre-warming the block to 94°C (hot start). A duplicate reverse transcription reaction and PCR were performed for each sample and run side by side in the same polyacrylamide gel. A radioactively labelled DNA ladder was prepared (7). Denaturing polyacrylamide gels (Gel-Mix 6 solution, Gibco-BRL) were run (sequencing gel apparatus model S2, Gibco-BRL), dried and exposed as recommended (GenHunter Corporation). PCR products were isolated from long electrophoretic runs. DNA was extracted from the dried gel slice and reamplified by PCR (GenHunter Corporation). Reamplified PCR products were cloned (TA Cloning kit, Invitrogen, San Diego, CA). To exclude the possibility of more than one PCR product being represented as a single band in the polyacrylamide gel, three independent clones derived from each polyacrylamide gel slice were sequenced using the 70750 Reagent Kit For Sequencing With Sequenase T7 DNA Polymerase and 7-deazadGTP (Amersham, Cleveland, OH) and run in polyacrylamide gels as above. Poly(A) RNA was selected from total FDCPmixA4 RNA (Oligotex-dT mRNA kit, QIAGEN) and from day 8.5 yolk sac (FastTrack mRNA Isolation Kit, Invitrogen) following manufacturer's instructions. Five gg of poly(A) RNA were used for the cDNA synthesis (Superscript Plasmid System, GibcoBRL). Lambda Not-Sall Arms (Gibco-BRL) were ligated to the cDNA and packaged (X Packaging System, Gibco-BRL). Libraries were screened as described (7). PCR reactions were performed in duplicate using cDNAs generated in independent cDNA synthesis reactions. The results shown in Figure IA confirm the reproducibility and reliability of the results obtained by DD-PCR. DD-PCR should allow the display of all of the esimated 15 000 unique species ofmRNA htanscribed in the cell at any given time (1). The number of primer combinations necessary to visualize all of the different messages as bands in polyacrylamide gels has been suggested to range from 240 (1) to 312 (7). Using conditions previously described for DD-PCR (2) we observed an average of 45-55 bands per lane indicating the need to perform a minimum of 300 primer combinations. We sought to increase the number of bands visualized while keeping a reduced number of reverse transcription reactions (2). By increasing the concentration of the 5' primer from 2 to 30 gM, using 32°C annealing temperature, a cold start and 40 PCR cycles (Fig. IB), an average of 66 bands per lane were easily detected, without a significant increase in the background signal, representing a 25% decrease in the total number ofprimer combinations required to visualize 15 000 bands. Significant differences in gene expression may be undetectable with 40 cycles ofPCR due to template saturation. By reducing the