Global cDNA Amplification Combined with Real-Time RT–PCR: Accurate Quantification of Multiple Human Potassium Channel Genes at the Single Cell Level

We have developed a sensitive quantitative RT–PCR procedure suitable for the analysis of small samples, including single cells, and have used it to measure levels of potassium channel mRNAs in a panel of human tissues and small numbers of cells grown in culture. The method involves an initial global amplification of cDNA derived from all added polyadenylated mRNA followed by quantitative RT–PCR of individual genes using specific primers. In order to facilitate rapid and accurate processing of samples, we have adapted the approach to allow use of TaqMan™ real-time quantitative PCR. We demonstrate that the approach represents a major improvement over existing conventional and real-time quantitative PCR approaches, since it can be applied to samples equivalent to a single cell, is able to accurately measure expression levels equivalent to less than 1/100th copy/cell (one specific cDNA molecule present amongst 108 total cDNA molecules). Furthermore, since the initial step involves a global amplification of all expressed genes, a permanent cDNA archive is generated from each sample, which can be regenerated indefinitely for further expression analysis.


Introduction
Potassium channels are a large family of ion channels that are present in essentially all cell types and mediate a wide range of physiological functions. Physiological, pharmacological and expression studies have demonstrated that the behaviour of a variety of tissues is in¯uenced by the pattern of potassium channels present. For example, the large conductance calcium-dependent potassium channel BK Ca (also known as Maxi-K or slo channel) plays a key role in a range of cellular functions, including neuronal spike shaping and neurotransmitter release (Kaczorowski et al., 1996), setting contractile tone (Nelson et al., 1995), modulating myometrial activity (Anwer et al., 1992(Anwer et al., , 1993 and electrical tuning of cochlear hair cells (Art et al., 1995;Ramanathan et al., 1999). With the overall aim of examining mRNA levels of a panel of potassium channels in human tissue samples, we set out to develop an approach with the following characteristics. First, the method should provide a means of determining the approximate number of mRNA molecules/cell in each sample studied, thereby establishing the likely physiological consequence of the detected expression. Second, the method should be capable of detecting a wide range of expression levels, including those less than one part in a million (equivalent to less than one copy/ cell). Third, it should be applicable to small samples, including single cells, making it possible to correlate physiological measurements using patch-clamping with expression pro®les. Fourth, the method should allow the examination of multiple genes from the same sample in order to establish the presence and level of a range of potassium channels and their subunits. Finally, the method should ideally allow retrospective screening of accumulated samples in order to facilitate analysis of novel potassium channels.
Clearly, classical techniques for determining mRNA levels, such as Northern blot analysis and RNase protection assays, are not capable of ful-®lling these criteria, since they require large amounts of total RNA. With the advent of the polymerase chain reaction (PCR) and its application to the analysis of mRNA [reverse transcriptase PCR (RT±PCR)], it is now possible to analyse mRNA levels in small samples including individual cells (Brady et al., 1990;Brady and Iscove, 1993;Rappolee et al., 1989). Although RT±PCR can be used to quantify expression, existing methods are generally time-consuming, restricted to a small number of target genes and not suitable for the examination of samples as small as a single cell. To overcome this limitation we have developed an approach based on an initial global ampli®cation of cDNA copies of all polyadenylated mRNAs followed by real-time quantitative PCR directed towards speci®c genes. The resultant approach exhibits the precision associated with real-time PCR (Preudhomme et al., 1999;Steuerwald et al., 1999;Wang and Brown, 1999) and in addition has several valuable features which are not part of current real-time or standard quantitative techniques. These features include the ability to analyse samples as small as a single cell and the ability to determine absolute expression levels as low as one speci®c target cDNA in amongst 10 8 total cDNAs. Furthermore, since the initial step is a global ampli®cation, the method generates a renewable archive of representative cDNAs, which can also be used for retrospective screening of accumulated samples. To validate the approach we have examined the IK Ca potassium channel, which is highly expressed in placenta (Ishii et al., 1997;Logsdon et al., 1997) and BK Ca (a-and b subunits) which is both physiologically active (Khan et al., 1993) and detectable at the level of protein and mRNA (Song et al., 1999) in uterus.

Materials and methods
Global Ampli®cation of cDNA copies of all polyadenylated mRNAs (PolyAPCR) Tissue RNAs were obtained from commercial suppliers (human lung, heart, placenta and breast from Ambion, and remaining tissue RNAs from Clontech). PolyAPCR was carried out essentially as previously described . In brief, for each sample 0.5 mg total RNA in a volume of 0.5 ml was added to 10 ml ®rst lyse buffer [600 U/ ml Prime RNase Inhibitor (5k±3k incl.), 50 mM Tris±HCl (pH 8.3), 75 mM KCl, 3 mM MgCl 2 , 0.5% Nonidet P-40, 10 mm dNTPs, 23 nM dT24 oligo], incubated at 65uC for 1 min, allowed to cool to 22uC (room temperature) and transferred to wet ice. After adding 12.5 units AMV reverse transcriptase (Boehringer-Mannheim), the reverse transcriptase reaction was initiated by transferring the samples from ice to 37uC. After 15 min incubation at 37uC the reverse transcriptase was inactivated by heating to 65uC for 10 min and samples were returned to ice. For the addition of the 5k priming site, a terminal transferase reaction was carried out by adding 10 ml 2r tailing buffer [200 mM potassium cacodylate (pH 7.2), 4 mM CoCl 2 , 0.4 mM DTT, 1 mM dATP] containing 10 U terminal deoxynucleotide transferase (Gibco BRL) and incubating the samples at 37uC for 15 min followed by heat inactivation at 65uC for 10 min. Following return of the samples to wet ice, 40 ml PCR mix [23.5 mM Tris±HCl (pH 8.3), 117.4 mM KCl, 8.2 mM MgCl 2 , 2.2 mM dNTPs, 0.23% Triton X-100, 47 mg/ml BSA, 85 units/ml Taq polymerase, 8.33 mm oligo Not140] was added to all samples and PCR was performed using two linked sets of 25 cycles. The ®rst 25 cycles were: denaturation at 94uC for 1 min, primer annealing at 42uC for 2 min and extension at 72uC for 6 min; and the second 25 cycles involved denaturation at 94uC for 1 min, primer annealing at 42uC for 1 min and extension at 72uC for 2 min. Following evaluation of the PolyAPCR products (PolyAcDNA) by agarose gel electrophoresis and densitometry, all samples were adjusted to a ®nal concentration of 0.5 mg/ml.

Gene-speci®c PCR
For each gene studied, PCR primer pairs directed towards the mRNA sequence present within 300 bases of the polyA addition site were designed as previously described (Nu Â ne Ãz et al., 2000). Sequences for PCR primers presented in the 5k±3k orientation, together with the Accession Nos of the target genes, are listed in Table 1. PCR reactions were carried out in a 22 ml reaction volume containing 1 ng PolyAcDNA, 0.33 mm each oligonucleotide, 0.5 units TaKaRa Ex Taq (TaKaRa), 0.25 mM dNTPs in buffer supplied by the manufacturer. PCR was performed without mineral oil in a Techne PHC-3 thermocycler ®tted with a heated lid by heating the samples for 5 min at 95uC, followed by 20±30 cycles of 30 s at 94uC, 30 s at 56uC and 1 min at 72uC.

Preparation of gene-speci®c quantity standard (GSQS)
Following successful gene-speci®c ampli®cation, PCR products were puri®ed using GFX PCR DNA and Gel Band puri®cation Kits (Pharmacia Biotech) as recommended by the manufacturer. Puri®ed PCR products were resolved by agarose gel electrophoresis, visualized by ethidium bromide staining and the concentration of each band calculated by densitometric analysis of the gel images using a GS-700 Imaging Densitometer (Bio-Rad) and Molecular Analyst software, Version 1.5 (Bio-Rad). A gene-speci®c quantity standard (GSQS) for each PCR product was generated by using the puri®ed bands to create a dilution series, where the number of DNA molecules ranged from 1.5r10 10 to 1.5r10 4 /ml (Nu Âne Ãz et al., 2000). For example, the 180 bp human glyceraldehyde-3-phosphate dehydrogenase (GAP) fragment was ®rst diluted to 2.97 ng/ml (1.5r10 10 molecules/ml) and from this solution serial 10-fold dilutions were prepared in TE (10 mM Tris±HCl, pH 8.0, 1 mM EDTA) containing 0.5 m/ml sonicated carrier lDNA.

Hybridization analysis of PCR products
For each gene, PCR (20±35 cycles) was applied to 1 ng PolyAcDNA (equivalent to 3r10 9 total cDNA molecules based on an average molecular size of 300 bases) from each sample and 2 ml from six relevant gene-speci®c quantity standards equivalent to a range of 3r10 7 ±30 molecules. PCR products were resolved by agarose gel electrophoresis, visualized by ethidium bromide staining, photographed and nylon ®lters prepared as previously described (Brady et al., 1995). For each gene an internal oligonucletide probe (Table 2) was endlabelled with¯uorescent dUTP (Amersham) using terminal transferase (Gibco Life Sciences) and hybridized to the relevant ®lter at 45uC for 16 h in 5rSSC, 0.1% SDS, 1 : 20 blocking agent (Amersham), 5% dextran sulphate. Following the removal of excess probe, ®lters were washed 2r30 minutes in 0.5 SSC+0.1% SDS and signals detected using a Gene Images CDP-Star detection module (Amersham Life Sciences) as described by the manufacturer. For each gene, separate PCR reactions were carried out for 20, 25, 23 and 35 cycles, all products hybridized as described above, and the resultant autoradiographs quanti®ed by densitometric analysis using a GS-700 Imaging Densitometer (Bio-Rad) and Molecular Analyst software, Version 1.5 (Bio-Rad). Preparation of polyAcDNAs from 10-and one-cell dilutions One thousand freshly growing human lymphoblastoid CCRF CEM C7A cells (Norman and Thompson, 1977) were directly lysed in 100 ml ®rst lyse buffer and two serial 10-fold dilutions carried out to yield cell lysates at concentrations of 100 and 10 cells/ml. To establish that global ampli®cation is feasible at the level of 10 and one cell(s), three independent PolyAPCR reactions were prepared from 10 ml aliquots taken from the 100 and 10 cell/ ml cell lysates. Following PolyAPCR, the resultant PolyAcDNAs from both 10 and one cell triplicates were analysed as described below.
TaqMan 2 real-time quantitative PCR TaqMan 2 (PE Applied Biosystems) PCR primers and probes were designed for each gene using Primer Express 2 and are listed in Table 2. Primers were supplied by either PE Applied Biosystems (PE ABI), Genosys, or MWG and¯uorescently labelled probes [5k 6-carboxy¯uorescein (FAM); 3k 6-carboxytetramethylrhodamine (TAMRA)] were supplied by either MWG or PE ABI (see Table 2). For each gene, TaqMan 2 PCR was applied to 1 ng PolyAcDNA (equivalent to 3r10 9 total cDNA molecules) from each tissue and 2 ml of all six relevant gene-speci®c quantity standards (equivalent to range of 3r10 7 /300 molecules) using a TaqMan 2 Gold kit, as recommended by the manufacturer. Samples were analysed using an ABI Prism 7700 Sequence Detection System (PE Applied Biosystems) as recommended by the manufacturer.

Results
Global ampli®cation and gene-speci®c PCR reveals a striking tissue-speci®c pattern for potassium channel genes Agarose gel electrophoresis and ethidium bromide staining was used to detect the ampli®cation products from 25 cycles of gene-speci®c PCR applied to PolyAcDNAs derived from human tissues and a BK Ca a serial standard ( Figure 1A). Based on the appearance of DNA bands in the ®rst ®ve BK Ca a serial standards ( Figure 1A), the threshold of PCR detection using agarose gel electrophoresis/ethidium bromide staining for these conditions was around 3r10 3 target molecules present at the start of the PCR reaction. Of the six tissue PolyAcDNAs examined, a strong band was detected in samples from uterus that was similar to, or brighter than, the band seen in serial standard containing 3r10 5 BK Ca a cDNA molecules. Faint but reproducible bands were also seen in the breast and placenta lanes ( Figure 1A). To increase the sensitivity and speci®city of PCR detection, internal oligonucleotide probes were used to detect gene-speci®c PCR products ( Figure 1B). As can be seen from comparing the BK Ca a bands in Figures 1A and 1B, detection of PCR products by Southern hybridization increased the detection sensitivity approximately 100-fold and revealed low-level expression of BK Ca a in lung and heart, whereas expression in foetal liver remained undetectable. Since annealing temperature, number of cycles and primer concentration were identical for both tissue and serial standard samples, an estimate of the abundance of BK Ca a transcripts in the starting tissue samples can be made by comparing ampli®cation products derived from the tissue PolyAcDNAs to the BK Ca a serial standards. For example, band intensities in both Figure 1A and 1B indicate that theBK Ca a PCR product from uterus is equivalent to at least 3r10 5 BK Ca a cDNA molecules present at the beginning of the PCR reaction. From a visual comparison of the bands present in the tissue samples and serial standard in Figure 1B, it can be estimated that there are around 3r10 3 BK Ca a cDNA molecules present in the breast and placenta samples, 300 molecules present in the heart and lung samples and undetectable levels (fewer than 30 copies) in the foetal liver sample. Since 1 ng PolyAcDNA is equivalent to 3r10 9 molecules (calculation based on 300 bp as the average size of PolyAcDNA, 1 bp=650 Da and Avagadro's number=6r10 23 ) the abundance of BK Ca a subunits molecules in the uterus sample can be expressed as >1 BK Ca a molecule/10 4 total cDNA molecules. Based on the assumption that a single animal cell expresses in the order of 5r10 5 mRNA Figure 1. RT±PCR analysis of tissue mRNAs. (A) Results of PCR using BK Ca a primers applied to 1 ng of tissue-derived PolyAcDNAs, a control sample in which no RNA was added to the initial reverse-transcriptase reaction and the BK Ca a serial standards. PCR products have been visualized by agarose gel electrophoresis, ethidium bromide staining and UV illumination. (B) Results of PCR using the indicated primers applied to 1 ng of tissue-derived PolyAcDNAs, control and the relevant serial standards. The PCR products were detected using labelled internal oligonucleotides, as described in Materials and methods. For both panels the numbers below the serial standards refer to the number of speci®c target cDNA molecules present at the beginning of the PCR reaction molecules (Alberts et al., 1997), the average expression levels for BK Ca a in the uterus would be in the order of 50 or more mRNA molecules/cell. Based on the same assumptions, expression levels for BK Ca a for the breast and placenta samples would be in the order of 0.5 molecules/cell, for the heart and lung samples an average of 0.05 molecules/cell and for foetal liver less than 0.005 copies/cell. As expected, the housekeeping gene GAP was expressed at a level of at least 5 copies/cell in all samples, whereas the potassium channel IK Ca showed a restricted tissue-speci®c expression pattern which was distinct from the pattern for BK Ca a ( Figure 1B).

Quantitative measurement of PCR products
Although data presented in Figure 1A, B can be used to provide useful estimations of expression levels, it is also clear from an examination of the serial standard that there is a`plateau' effect which is effectively limiting PCR product in samples which contained 3r10 5 or more target molecules. In order to improve the precision of measurements using this system, two parallel approaches were compared. In the ®rst approach, following gene-speci®c PCR and Southern hybridization ( Figure 1B), abundance was estimated by densitometric analysis ( Figure 2C). In the second approach, TaqMan 2 real-time PCR was applied to both the serial standards and the tissue PolyAcDNAs analysed in Figure 1B. Figure 2A and B shows the results of TaqMan 2 analysis applied to IK Ca serial standards and Figure 2C shows the standardization curve generated from densitometric analysis. A comparison of TaqMan 2 and densitometric analysis applied to samples shown in Figure 2D revealed that, although both approaches gave comparable results, real-time PCR was able to detect low levels of IK Ca expression in heart which were not detected by densitometric analysis of the hybridized PCR products.

Real-time quantitative analysis of PolyAcDNA at the single cell level
Since PolyAPCR has been widely used for global ampli®cation of all expressed sequences from single cells (Berardi et al., 1995;Brady et al., 1990Brady et al., , 1995Dulac and Axel, 1995), we next examined whether the current quantitative adaptation could be used at the single cell level. Previous RT±PCR analysis of individual cells has shown that for all genes  (Brady et al., 1990(Brady et al., , 1995Brail et al., 1999). In order to examine variation due simply to the method, PolyAPCR was carried out in triplicate on dilutions of a fresh cell lysate prepared from 1000 human T lymphoma cells (for details, see Materials and methods), thereby avoiding the inherent differences in gene expression exhibited by individual cells. Following triplicate reverse transcriptase, tailing and PolyAPCR applied to serial dilutions equivalent to 10 and 1 cell(s), the level of the housekeeping genes GAP, lactate dehydrogenase A (LDHA) and initiation factor 2B (IF2B) were established using TaqMan 2 PCR ( Figure 3). As can be seen from Figure 3, high levels of GAP and approximately 100-fold lower levels of LDHA and IF2B were detected in each of the 10-and 1-cell PolyAcDNA samples. Variation between samples was more apparent at the 1-cell level, particularly in the case of LDHA. Despite the variation between samples, the average expression for all genes was similar for both the 10-cell and 1cell series (Figure 3). These results demonstrate that the approach can be applied to single cells and is sensitive enough to reliably detect genes expressed at low levels.

BK Ca a and b mRNAs are highly strongly expressed in human uterus
Having established that the TaqMan 2 approach is suitable for PolyAcDNA samples, we next used the approach to examine the levels of BK Ca a, BK Ca b and IK Ca mRNAs in a panel of human tissues. The results presented in Figure 4 show that the method can be used to detect and quantify levels of expression equivalent to one speci®c cDNA molecule in amongst 10 8 total cDNAs. For all samples, high levels of the housekeeping gene GAP and varying levels of the potassium channels were detected. In the case of individual genes, expression levels varying over four orders of magnitude could be accurately assessed (cf. IK Ca expression in the adult liver and placenta samples in Figure 4). IK Ca levels were found to be highest in placenta and both BK Ca a and b sub-units were highest in the uterus.

Discussion
The results presented in Figure 4 demonstrate that the approach described can be used to identify patterns of potassium channel expression that re¯ect the cellular composition and physiological status of target tissues. For example, high levels of BK Ca a and BK Ca b have been previously reported in uterus and ascribed to the preponderance of smooth muscle cells in this tissue (Khan et al., 1993(Khan et al., , 1997Song M. 1999;Wallner et al., 1995). Our ®nding that, among the 16 tissues examined, BK Ca mRNA levels (both a and b) were highest in the uterus sample and lowest (equivalent to <0.1 copy/cell) in tissues such as adult and foetal liver samples (Figure 4) are consistent with the proposed association of BK Ca expression and smooth muscle cells. Intriguingly, although the BK Ca b subunit is thought to act solely through moderating the activity of the BK Ca a subunit (Kaczorowski et al., 1996;Rama- Tseng-Crank et al., 1996), the ratio of BK Ca a to BK Ca b varied considerably between tissue samples, with similar levels of both subunits detectable in the three brain tissues compared to 40-fold higher levels of BK Ca b than BK Ca a detectable in the heart sample ( Figure 4). Furthermore, in the case of both the heart and lung samples, expression of BK Ca b is equivalent to around 5±10 copies/cell, whereas the level of BK Ca a in the same samples is far below one copy/cell, indicating that at least some cells in those tissues may express BK Ca b in the absence of BK Ca a. Expression of BK Ca b in the absence of BK Ca a would suggest either that BK Ca b is able to act alone or that it is able to interact with other potassium channels. These results highlight the importance of being able to determine both absolute and relative mRNA levels for multiple genes in the same sample.
In the case of IK Ca , 12 of the human tissues analysed in Figure 4 were previously independently analysed by two research groups using polyA mRNA and Northern analysis (Ishii et al., 1997;Logsdon et al., 1997). A comparison of the IK Ca Northern data (Ishii et al., 1997;Logsdon et al., 1997) and the TaqMan data in Figure 4 revealed that both PolyAmRNA Northern and the Poly-APCR TaqMan 2 identi®ed highest levels of expression in placenta followed by medium to high expression in prostate and lung. Furthermore, the PolyAPCR TaqMan 2 method was able to detect low to moderate expression levels in several of tissues such as uterus, testes and spinal cord ( Figure 4), which were undetectable using Northern analysis (Ishii et al., 1997;Logsdon et al., 1997). These results indicate that the detection limit of PolyAmRNA and Northern analysis used by both groups is >1 : 10 6 , which is 100-fold lower than the PolyAPCR TaqMan 2 method described here.
PolyAPCR has been used to examine mRNA levels and patterns in a wide variety of cells including highly puri®ed haemopoietic cell subpopulations (Cumano et al., 1992;Sauvageau et al., 1994), oocytes, eggs, and pre-implantation embryos from fertilization to the blastocyst stage (Rambhatla et al., 1995), micro-well cultures of cells (Yan et al., 1993), haemopoietic colonies (Brady et al., 1990;Cheng et al., 1996) and single cells (Brady et al., 1990;Brady et al., 1995;Dulac and Axel, 1995;Trumper et al., 1993). However these studies have been based largely on ®lter hybridization which at very best is limited to detecting 1 speci®c molecule present in amongst 10,000±100,000 total cDNAs (Brady et al., 1990;Cheng et al., 1996;Rambhatla et al., 1995). Here we demonstrate detection of levels of expression as low as one speci®c molecule present in amongst 10 8 total cDNAs (Figure 4), thereby improving the overall sensitivity at least 1000-fold. Based on the assumption that a single cell contains 10 5 ±10 6 mRNA molecules (Alberts et al., 1997), we are able to detect levels of expression equivalent to 1/100th of a copy per cell. Furthermore, unlike existing TaqMan 2 methods (Steuerwald et al., 1999;Wang and Brown, 1999) PolyAPCR TaqMan 2 can be applied to an unlimited number of genes from small  (Figures 3 and 4). It must be stressed that the PolyAPCR TaqMan 2 method is not intended as a direct alternative to mass expression screening methods such as high density cDNA array hybridization (Duggan et al., 1999;Hauser et al., 1998) since, at present, it is suitable for studying 10 s to 100 s of individual genes. However, PolyAPCR TaqMan 2 can be applied to samples several orders of magnitude smaller than those suitable for array hybridization and is able to detect expression levels at least 1000fold lower than currently detected by array methods. In other words, unlike PolyAPCR TaqMan 2 , current high density cDNA array hybridization methods are unsuitable for the analysis of samples less than 1000 cells and are unable to reliably detect mammalian genes expressed either at a level lower than a few copies per cell (Duggan et al., 1999). PolyAPCR TaqMan 2 therefore extends and complements existing mRNA pro®ling methods, since it can be used to provide sensitive and quantitative analysis of individual genes identi®ed by highdensity screening.