WO1993023566A1

WO1993023566A1 - Nucleic acid detection and quantification

Info

Publication number: WO1993023566A1
Application number: PCT/GB1993/001058
Authority: WO
Inventors: Julia Elizabeth Stickland; Anna Louise Lisa Ramshaw; Colin Gerald Potter; James O'donnell Mcgee
Original assignee: Isis Innovation Limited
Priority date: 1992-05-21
Filing date: 1993-05-21
Publication date: 1993-11-25
Also published as: GB9210916D0

Abstract

A method of quantifying a first nucleic acid sequence in a test sample containing a mixture of nucleic acid sequences by providing labelled sequence-specific probes for the first nucleic acid sequence and a second nucleic acid sequence, incubating the test sample with the two probes under hybridising conditions, measuring the extent of hybridising of each probe, and determining a ratio of the two measurements as indicative of the first nucleic acid sequence. Also a method of detecting the presence or amount of a viral nucleic acid sequence in a test sample by simultaneous hybridisation of labelled probes specific for the viral nucleic acid sequence and a second nucleic acid sequence.

Description

NUCLEIC ACID DETECTION AND QUANTIFICATION

The present invention is concerned with a novel method of quantifying a particular nucleic acid sequence in a mixture of nucleic acids, using a dual- label hybridisation technique. The method described can be applied particularly to measurement of gene dosage and detection of allelic loss, which may be associated with cancer, but is also useful in quantification of other specific nucleic acid sequences such as particular messenger RNA (mRNA) sequences or viral nucleic acid in tissue and cell samples.

The usual methods used for screening tumour DNA for oncogene amplification are restriction fragment length polymorphism (RFLP) analysis and single-label DNA dot-blots. These have been widely used to study mutations in cancer, but each method has its limitations.

1. RFLP analysis employs Southern transfer of DNA onto a nylon membrane but uneven transfer can occur, especially of large DNA fragments, giving the possible false impression of differences in gene dosage. 2. The dot blot technique compares filters which are separately hybridised with the gene of interest or with a control probe. Differences in the hybridisation procedure and in the amount of each DNA sample on the filter lead to large variations in the gene ratios measured. Sequential hybridisation takes no account of DNA losses occurring during removal of the previous probe from the filter. 3. In both RFLP analysis and dot blots, hybridisation signals are detected by autoradiography, which may be quantified by densitometry. When autoradiographic films are hypersensitised by pre- flashing, the signal intensity is not always linearly related to the amount of probe bound. This is particularly significant when small or large quantities of DNA are to be measured, and may limit the detection of low levels of amplification in tumour DNA. It is also difficult to determine optimal autoradiograph exposure, thereby obtaining consistency between, experiments. A direct result of these problems is reluctance to score homozygous individuals for low levels of amplification or indeed allele loss by RFLP analysis. 4. Both techniques often fail to analyse the patient's constitutional DNA. If only tumour samples are analysed, it must be assumed that every patient's constitutional genotype at the control locus is similar. While this may be the case for most individuals, the occasional patient may be hemizygous ^at the control locus. Without constitutional DNA as a control, such patients would be scored erroneously as having a doubling in the dosage of the gene in the tumour.

For quantifying specific mRNA sequences, densitometry of autoradiographs (from electrophoretic gels, transfer membranes and dot blots) has been used to compare unknown samples with RNA standard nucleic acids (Laskey 1980). This, however, lacks linearity when used over a large range of nucleic acids and of radioactivity.

Alternatives to densitometry, are regular liquid scintillation counting and flat-bed scintillation counting (Warner, et al. 1985, Potter, et al. 1986) . In the latter, radioactive gel fragments, for example, from ribonuclease protection assays, can be measured by sandwiching them between two glass fibre sheets in plastic bags with liquid scintillant (Potter, et al. 1991). This sandwich technique gives a 32P spectrum more typical of that found in direct liquid scintillation counting, and has been extended to the analysis of RNA dot blots on charged nylon filters hybridised with 32P-labelled probes. The radioactive filters may be counted while still water-wet with similar high counting efficiency as with scintillant, permitting successive hybridisations with a number of probes in order to quantify different mRNAs within the same sample area (Potter, et al. 1991). This is a useful technique, but small differences in hybridisation efficiency over different parts of a filter could give rise to unacceptable errors when stringently evaluating the ratios of activity between sequentially applied probes for the quantification of mRNA.

Detection of virus in a sample is usually by Southern blot, dot blot or the polymerase chain reaction (PCR) . Dot blot hybridisation for detection of human papilloma virus (HPV) in cervical swabs, is well established. But whatever method is employed, there are problems of false negatives due to poor clinical samples in which the nucleic acid may be degraded or is absent altogether. The extreme sensitivity of PCR can give false positives through contamination. In research and clinical survey, more reliable methods suitable for screening and quantification purposes, are required. It is clear, therefore, that a reliable method of quantification is lacking in molecular biology. A dual-labelling hybridisation technique has now been developed to provide a sensitive means of detecting quantifying nucleic acids. The present invention therefore provides a method of quantifying a first nucleic acid sequence in a test sample containing a mixture of nucleic acid sequences including a second nucleic acid sequence, by providing a first labelled sequence-specific probe capable of hybridising with the first nucleic acid sequence and a second labelled sequence-specific probe capable of hybridising with the second nucleic acid sequence, which method comprises incubating the test sample with the two probes under hybridising conditions, measuring the extent of hybridising of each of the two probes to its respective nucleic acid sequence, and determining a ratio of the two said measurements as indicative of the quantity of the first nucleic acid sequence in the test sample. In the method according to the invention, each nucleic acid sequence may be a DNA sequence such as a gene or a part of a gene, or it may be an RNA sequence, such as mRNA. One may be a DNA sequence and the other an RNA sequence. Either or both nucleic acid sequences may be comprised of viral DNA or RNA.

There may also be provided a control sample which is also tested by the method as defined, and the ratio obtained for the test sample is compared to the ratio obtained for the control sample as indicative of the quantity of the first nucleic acid sequence in the test sample. In a particular method, the test sample comprises tumour DNA from a patient and the control sample is of constitutional DNA of the same patient. In a preferred method according to the invention, the first sequence-specific probe is labelled with 2-phosphorus and the second sequence- specific probe is labelled with 35-sulphur. Other radioactive labels may also be used. Preferably, the two radiolabels are counted simultaneously in a flat- bed scintillation counter. -

The invention also provides a method of detecting the presence or amount of a viral nucleic acid sequence in a test sample containing a mixture of nucleic acid sequences including a second nucleic acid sequence, by providing a first labelled sequence- specific probe capable of hybridising with the viral nucleic acid sequence and a second labelled sequence- specific probe capable of hybridising with the second nucleic acid sequence, which method comprises incubating the test sample with the two probes under hybridising conditions, measuring the extent of hybridising of each of the two probes and comparing the two measurements to determine the presence or amount of the viral nucleic acid sequence.

The dual-label technique is described herein in detail for quantification of gene dosage in breast cancer, for quantifying T-cell receptor mRNA expression and for detecting and quantitatively measuring human papilloma virus in cervical cells.

In one example of the dual-label technique f°^r gene dosage, DNA is transferred to a nylon membrane and hybridised simultaneously to two radioactively labelled DNA or RNA probes. A probe for the locus of interest is labelled with (o- 32P] dCTP and as an internal control, a probe to the β-actin locus is labelled with [α- 35S] dATP. The bound radioactivity is measured on a flat-bed scintillation counter. Simultaneous measurement of the counts arising from the 32P and 35S-labelled probes is possible by using conventional pulse height analysis to separate the activities of the 32P and 35S. Comparison of the 32P/35S count ratios from the tumour and blood samples gives an accurate estimate of the relative level of the gene of interest. Although 35S gives a weak signal, the presence of multiple copies of the control locus (in this case the β-actin gene) results in hybridisation of sufficient label to be easily detected. It is wise to have a multiple copy control locus in this instance as gene loss which frequently occurs in cancer, might otherwise distort results. At a single-copy control locus, mutations may reduce allele dosage from two to one, or even to zero, with dramatic effects on the dosage of the locus of interest.

Example 1 below shows studies of the amplification of the int-2 gene in breast cancer by a dual-labelling dot blot method. The human int-2 locus is known to be amplified in breast cancer (Callahan, 1989; Callahan and Campbell 1989). First of all, the efficacy of dual-labelling was compared with single- label hybridisations. It was essential to begin by ensuring that two differently labelled probes would not interfere with each other when employed at the same time. There was found to be no interaction.

The dual-label technique as used here is suitable for quantifying gene dosage in cancer, and can be used to study gene amplification in one of two ways. In the first of these methods, a number of paired tumour and constitutional DNA samples is studied together. The population of constitutional DNAs provides a conservative internal control for variation caused by experimental method. This approach is suitable for determining the frequency and level of amplification and allele loss in a population. If only a few samples are available or the amplification levels of individual patients are required, a number of replicate samples are analysed to provide a measure of statistical confidence in the amplification levels that are determined. Ideally, both techniques would be used together, but time and sample shortages prevent this in practice. There are significant advantages with the dual-label blot technique of the invention. It requires only small amounts of DNA (1 μg per dot), allowing more cases to be studied, frequently in replicate. The control probe provides a measure of the amount of DNA loaded per dot, so that the absolute quantity of DNA bound to the membrane is of less importance. The method is rapid and needs neither restriction endonuclease digestion nor agarose gel electrophoresis. Accurate quantification of radioactivity by flat-bed scintillation counting overcomes some of the problems of densitometric analysis as it permits the detection of a wide range of radioactivity with great sensitivity.

Another important area of application of the invention lies in the study of allele loss in tumours. Here, work is frequently restricted by the need to study informative (heterozygous) individuals: sometimes, fewer that 25% of cases are heterozygotes. The use of dual-label dot blots would render every case informative, greatly increasing the data gained from a sample of tumours. Loci, such as p53, with few informative cases but potential importance in tumourigenesis, could be studied in small population samples. Quantification of gene dosage by this technique may permit the biological significance of mutations to be assessed with more confidence and provide more accurate associations between molecular and clinical variables. It is envisaged that precise information on allelic loss will provide a basis for developing gene therapy, cytokines or drugs which interfere with cancer-associated alterations in receptor structure or numbers.

A further application of the present technique for DNA is in the prenatal diagnosis of genetic disorders such as Downs Syndrome. Post-natal diagnosis of certain genetic abnormalities may also be carried out using the method of the invention.

The method according to the invention can also be used to quantify mRNA sequences, in a way which overcomes the problems encountered with successive hybridisations to the same filter. Example 2 demonstrates successful quantification of RNA by dual-labelling of RNA dot blots, where the mRNA of interest and the internal standard are labelled with

32P and 35S respectively. Again it was first of all necessary to show that it is possible to measure simultaneously two different probes hybridised to the same filter with an efficiency similar to that observed in single-label hybridisation reactions. The invention as used for mRNA quantification also has wide application, and potential importance in the investigation of large numbers of samples. It can be used to measure mRNA expression of T-cell mRNA^'s, oncogenes, cytokines and other growth factors (Ramshaw et al., work in progress), which are known to change in relative abundance in diseased tissue. Organ transplant rejection, for example, could be monitored for differential expression of various genes, and this may be important in clinical diagnosis (Dallman, et al.

1991) . It was found that the 35S counts were approximately half that of the 32P counts for the same cDNA labelled probe. It is preferable, therefore that the higher copy number mRNA should be hybridised with the S-labelled probe and the lower copy number mRNA species should be hybridised with the 32P-labelled probe. In some instances, where the sequence of interest has a very low copy number, the β-actin probe could be labelled with 35S to a lower specific activity

7 than normal (» 1-5 x 10 dpm/μg) . This would ensure that the low level 32P-counts for the message of interest are not less than half those of the S- labelled β-actin control probe, so that they may be efficiently separated by the counter. Alternatively, the sensitivity for 35S-labelled mRNA in low copy number may be enhanced by quantitative PCR, followed by double labelling and counting as described.

The present method is also suitable for detecting and quantifying viral nucleic acid in tissue samples. Recently, certain of the human papilloma viruses (HPV) , particularly HPV 16, have been implicated as at least part causative agents in cervical cancer. The amount of virus present seems to be correlated with risk of disease so accurate quantification of virus may give an early diagnosis. Example 3 demonstrates dual-labelling for detection of HPV16, where the ratio of the measurements for viral and host DNA can be determined to indicate the amount of virus present per total cellular mass in the sample.

Example 1

MATERIALS AND METHODS

Dual-label DNA dot-blots

Total DNA was prepared from 35 pairs of samples from breast tumour and patient^'s peripheral blood (Sambrook et al. , 1989). The concentration of each sample was estimated by optical densitometry

(OD__βo) and agarose gel electrophoresis. 1μg of each DNA sample was alkali-denatured by heating at 45°C for 10 minutes in 0.125M NaOH. Samples were immediately placed on ice and diluted in 0.125M NaOH/0.25x SSC, to give a final concentration of 0.125M NaOH/0.125x SSC. A dot blot manifold suitable for the counter format (Schleicher and Schull) was prepared for sample filtration by a single wash in 0.01% SDS, followed by three rinses in distilled water. The positively charged nylon transfer membrane (Wallac, 1205-403) and two absorbent backing sheets were pre-treated by soaking in 0.4M Tris (pH7.6) and mounted in themaifold. Each well of the manifold was washed through with 0.4M Tris (pH7.6), and the denatured DNA samples were then filtered by gravity onto alternate squares of the membrane, as shown in Figure 1. Six additional squares were left blank to measure the bbaacckground levels of non-specifically bound 32P and 35 S, labelled probe. Each DNA sample transferred to the nylon membrane was washed with filtration buffer (0.125M NaOH/0.125x SSC); weak suction was applied after 30 minutes to complete passage of the buffer through the membrane. The manifold was disassembled and the nylon membrane removed. The filters could be stored for up to 2 weeks in a sealed bag at 4°C, before hybridisation.

Probes were labelled by random priming (Feinberg & Vogelstein, 1983). A 0.9Kb £a£l insert of the int-2 DNA probe ss6 (Casey et al., 1986), was radiolabelled with [α- 32PjdCTP (specific activity 3000 Ci/mmol, Amersham U.K.). The control probe, a 0.66Kb

PCR product of the human β-actin gene subcloned into pBluescribe was labelled with [α- S]dATP (specific activity > 1000 Ci/mmol, Amersham U.K.) Unincorporated nucleotides were removed by Sephadex G50-300 spin

9 columns, specific activities of «2x 10 dpm/μg for

32 P- and «5x108 dpm/μg for 35S-labelled probes were routinely obtained. Probes were heat denatured at

100°C for 15 minutes prior to use and added to the hybridisation buffer at »1ng/ml. The choice of human β-actin as the control is important, as it hybridises to a number of related genes and pseudogenes (detecting about 15 alleles by

RFLP analysis of BamHI restricted DNA) . The resulting

35 S-signal bound to 1μg of DNA is sufficient for detection by the counter. A single-copy locus provides enough 32P counts.

Filters were pre-hybridized overnight at

42°C in 50% formamide, 10x Denhardt^'s solution, 1.5x

SSPE, 0.1% SDS and 0.5mg/ml denatured sheared herring sperm DNA. Hybridisation was carried out overnight at 42°C in the same solution with the addition of 5% dextran sulphate and the labelled probes. The hybridised filters were washed twice for 10 minutes at 22°C in 2x SSPE/0.2% SDS and once for 5 to 20 minutes in 0.2xSSPE/0.2% SDS at 65°C. The filters were thoroughly air dried and placed in thin plastic counting bags with 1ml of scintillant (Wallac

2 Betaplate Scint) per 60cm of filter. The filters were sandwiched between two glass fibre sheets each sealed in a plastic bag with 10ml of scintillant. These were suitable conditions to separate the pulse height spectra of 35S and 32P, as described for mRNA

(Potter et al. , submitted).

Counting of radioactivity from the dual- labelled filters was performed using a flat bed scintillation counter (1205 Betaplate, Wallac). Background levels of 32P and 35S emissions were determined from the specific areas of the nylon membrane to which DNA had not been applied. The background counts were then subtracted from the actual counts to give corrected counts for each isotope. The dosage of the int-2 gene can be calculated according to the following formula, where each variable represents measured radioactivity per unit time, (usually cpm) corrected for background and overlap of emission spectra (Potter £_£ s ., submitted) . Level of amplification = 32P tumour/35S tumour

32 P constitutional/

35 S constitutional

RFLP Analysis

Paired DNA samples from 23 of the above 35 cases were analysed independently for int-2 gene amplification by conventional RFLP analysis. Genomic DNA was digested with BamH1, electrophoresed on 0.9% agarose gels and transferred to nylon membranes (Hybond-N, Amersham, U.K.) (Sambrook et al., 1989). Filters were pre-hybridised and hybridised with the int-2 probe under the same conditions as those described for the DNA dual-labelled dot-blots. The membranes were washed as above and exposed to autoradiographic film with intensifying screens at -80 C for 2 to 7 days. Amplification of int-2 was scored by visual inspection of the resulting autoradiographs. Potential unequal loading was controlled for by visual inspection of ethidium bromide-stained agarose gels of separated DNA samples.

RESULTS

Comparison of dual- and single-labelling techniques

Before analysing individual cases for int-2 gene amplification, it was necessary to determine whether dual-labelling was as effective as single- label experiments in determining gene dosage. In particular, we had to determine if there was any interaction between the int-2 and β-actin probes when used simultaneously and also if dual-labelling nroduced lower variation amonσ samples. For this test, DNA was prepared from the peripheral blood of a normal individual. Six replicate DNA dots were hybridised to the 35S-labelled human β-actin probe alone; to the 32P-labelled int-2 probe alone; and to both probes simultaneously (two replicate experiments) .

Scintillation counting showed very similar activity when single- or dual-labelling was employed. The 35S-single-labelled β-actin probe averaged 677.9 cpm with a coefficient of variation (CV) of 8.3%; for the 32P-single-labelled int-2 probe, the average cpm was 180.7 with a CV of 14.6%. There was, therefore, acceptable consistency between replicate samples for the single-label experiments. Variation in cpm may be largely attributable to pipetting errors between samples.

The two dual-label hybridisation experiments gave results close to the single-label experiments. In the dual-label experiments, 35S activity averaged 757.7 and 747.0 cpm in each of the replicates, whilst

32 P counts averaged 172.7 and 143.6 cpm respectively.

These results suggest that there is no significant interaction between the int-2 and β-actin probes in the hybridisation reaction, int-2 amplification cannot be calculated here, since constitutional DNA was used, which has by definition, an amplification level of one. However, the variation expected can be calculated from the 32P to 35S ratios. The results of the two single-label experiments can be compared by dividing the 32P counts by the equivalent 35S counts from each set of six replicate samples. If this is done, a CV of 17.7% is obtained for the mean ratio.

The equivalent CVs from the replicate dual-label experiments are less, at 6.9% and 14.4%. This shows that the CVs from the dual-label experiments can be substantially lower than those obtained from independent single-label experiments.

Quantification of int-2 oncogene amplification in a . ^*east cancer population

Using the dual-label dot blot technique, the amplification level of the int-2 gene was measured in

35 paired (tumour and constitutional) samples from breast cancer patients. Two separate filters were bybridised independently, one with 17 cases and one, with 18 cases. The results of this experiment are shown graphically in Figure 1. Consider first the ratio of 32P to 35S counts obtained in the constitutional DNA samples alone. Despite inevitable variation in the quantity of these DNAs loaded onto the membrane, the ratios obtained were very similar between the two hybridisation experiments (means of 0.444 and 0.458;

CVs 7.1% and 8.9%) and could be combined showing low overall variation (mean = 0.451, CV = 8.5%). There is therefore good evidence that the measured 35S and 32P counts are linear for DNA samples of *1μg and that there is consistency between experiments. The constitutional 32P/35S ratios follow a distribution not significantly different from normal. This provides evidence that no sub-group of the sample population has a significantly different constitutional genotype, such as hemizygosity at the int-2 locus. The group of constitutional DNAs can, therefore, be considered as 35 replicate samples drawn from the same statistical population. It follows that the mean 32P/35S ratio (0.451) provides a more precise measure for calculating levels of int-2 amplification than individual 32P/35S ratios. The ratio of 32P to 35S counts in the tumour samples (Table 1) showed a higher mean (0.618) than the constitutional DNAs (0.451), as a result of the presumed int-2 oncogene amplification in some patients. The high CV of 62.6% and skewed distribution (Figure 1 ) suggests high amplification in some cases, with limited amplification in others. If none of the tumours had shown any amplification, we assume the CV of the tumour samples would have been similar to that of the constitutional DNAs.

The level of measured int-2 amplification is calculated using the population mean ratio of the constitutional DNA in every case. Table 1 shows the amplification ratios of the 35 samples placed in rank order. This ranges from 4.85 (case 4) to 0.72 (case 106) . An estimate of gene dosage can be calculated by multiplying the calculated amplification ratio by two. The constitutional DNAs appear to form a homogeneous population with a nominal mean amplification level of one. The estimate of variation among these samples may be used to set confidence limits on the deviation of the tumour amplification ratio from one. A 99% confidence interval was chosen as suitable and may be calculated as 0.78 to 1.22 (see Table 1 ) . Assuming that the tumour population has a similar level of variation to the cons'titutional DNAs, these same confidence limits may then be extended to the tumour samples. Any case with an amplification level over 1.22 or below 0.78 is therefore a candidate for scoring as amplified or as having lost an allele respectively. Using the above procedure, 14 cases (numbers

1, 4, 6, 8, 12, 18, 24, 103, 104, 107, 108, 109, 110, 119) are categorised as amplified at the int-2 locus. Two cases (numbers 16, 106) are scored as allele loss (Table 1). This give a frequency of int-2oncoσene amplification of 40% and a frequency of allele loss of 5.7%. RFLP analysis

RFLP analysis was used to give an independent assessment of int-2 gene dosage. Sufficient DNA for RFLP analysis was only available from 23 of the 35 cases studied by the dual-label method.

Following hybridisation of the RFLP filters to the int-2 probe, the majority of the population showed more than one band on the resulting autoradiograph.

To decide whether one band was amplified, the other hybridising band were used as internal controls for DNA loading. In cases where only one band hybridized to the probe, potential unequal loading was controlled for by visual inspection of ethidiu bromide-stained agarose gels of separated DNA samples. Six of the 23 cases analysed showed clear amplification of the int-2 gene; where the density of the 8.4Kb allele was increased (Table 1). One case, number 110, was scored as ^'possible amplification' of the 2.8kb (smallest) band. A further tumour, number 106, showed diminution of the intensity of the single hybridising band and this was scored as possible allelic loss. All other cases were scored as no amplification and no deletion. These results indicate amplification of the int-2 gene in 26-30% of the breast tumours analysed, consistent with previous studies (Tsuda, et al. 1989, Machotka, et al., 1989). These figures compare with a measured frequency of 40% by the dual-label dot blot method. RFLP analysis found possible allelic loss in 1 case (4.3%); the equivalent figure using the dual-label technique was 2 cases (5.7%) . The dual-label technique showed good consistency with RFLP analysis although, it found higher frequencies of int-2 amplification and allele loss than did RFLP analysis. Of the top seven cases in the rank order of amplification, six were independently scored as amplified by RFLP analysis. One (case 110) was scored as a possible amplification. It appears, therefore, that the dual-label experiment was able to place this case into its correct category. One tumour (case 106) showed possible allele loss by RFLP analysis. This case was bottom in the rank order of amplification (3.25 SD^'s below 1) and again the dual-label technique has permitted its correct categorisation.

Quantification of int-2 oncogene amplification in individual breast cancer patients

Constitutional DNA samples provide a useful control for experimental variation when determining the frequency and level of oncogene amplification in a population. Sometimes, however, determination of amplification levels in individuals may be required and for this the amplification ratios of samples must be determined in replicate. This permits a more accurate determination of dosage and provides confidence limits to be calculated for each individual case.

We therefore analysed int-2 amplification in 5 selected cases in triplicate. All had been studied in the 35-case experiment. Two had been scored as not amplified; one had an amplification level just greater than the upper 99% confidence limit; and two were highly amplified. A comparison was made within the triplicate experiment and between the triplicate and 35-case experiments. Table 2 shows the amplification levels determined in the 5 cases studied in triplicate. Relative standard deviation ratios within triplicates "">re typically 6% to 7%. In one of the 5 cases, ^■aver, this value was 20%. All the 5 cases showed air., _^ification levels that were not significantly different from the ratios determined in the 35-case experiment (Table 2) . The two highly amplified cases (numbers 108, 109) and two patients without significant amplification (numbers 116, 118) were all scored into the same category as before. Case 104 - previously scored as amplified but close to the 99% confidence limit - was scored by the triplicate experiment as no amplification.

DISCUSSION

We have shown that the dual-label dot blot technique can measure gene amplification in tumours. The amplification levels were calculated assuming that the counts arising from each isotope are proportional to the quantity of DNA on the filter and that their ratio is proportional to the average number of copies of the control locus or of the locus of interest. In the sample of 35 patients, amplification of the int-2 oncogene was measured in 14 cases at a frequency of 40% and allele loss was detected in 2 cases (5.7%). By comparison, visual scoring of RFLP autoradiographs showed only 6/23 cases (26%) having clear int-2 amplification. In part, this difference may result from our decision to score autoradiographs by eye, because of the inaccuracy of densitometry. All of these 6 patients showed significant amplification using the dual-label technique. Two further cases were scored by RFLP analysis, one as possible amplification and another as possible allele loss. The dual-label method allowed both of these cases to be firmly ascribed to their respective categories. The maximum amplification level among the 35 cases studied was 4.85. Previous studies have reported a wide range of amplification levels, consistent with this result (Donovan-Peluso et al. , 1991). The measured levels of amplification may result from four causes: (1) an increase (or decrease) in copy number of the int-2 gene only; (2) tumour aneuploidy and polymorphism; (3) the presence of normal tissue in tumour samples; and (4) variation resulting from experimental methods. Of these, the first two causes are biological in origin and the reasons for this study. The third cause is minimised when sampling tumours, but is unavoidable and leads to underscoring of mutations in quantitative techniques. It is important, however, to control for the fourth source of variation. In dual-labelling experiments, a single DNA dot is used to measure both the control and int-2 loci. This removes errors caused by pipetting variations. Counts due to non-specific background hybridisation of the probe are subtracted by the counter, to give the corrected counts for each sample. In addition, a more precise measure for calculating levels of int-2 amplification in the 35 cases can be provided by using the mean ratio of all 35 constitutional DNAs.

SxainpJle ?

MATERIALS, METHODS AND RESULTS

Preparations of Dot Blots, Labelling, Hybridization and Washing

Eight identical filters were prepared each having six samples. These included RNA from the HUT- 78 T-cell line (ATCC) , 10μg/spot, the MCF-7 breast tumour cell line (ATCC) , 10μg/spot, peripheral blood mononuclear cells, 6μg/spot, inflamed aortic tissue from patients with atherosclerosis 8μg/spot, and controls of yeast t-RNA (Sigma, U.K.), 10μg/spot, and buffer only.

Cytoplasmic RNA was extracted by acid guanidinium thiocyanate-phenolchloroform, dissolved in DEPC-treated water and glyoxal denatured. RNA was filtered by gravity, onto a positively charged nylon membrane (Pharmacia-Wallac, 1205-403) using a dot blot manifold (Schleicher & Schuell) suitable for flat bed scintillation counting on a Betaplate TM counter.

'Cross-talk^' caused by lateral emission of light between adjacent samples, was minimized by using a filter with a printed grid and by applying the RNA in alternate sample positions on the filter.

After UV cross-linking using a

ΦM

Stratalinker (Stratagene, San Diego) , filters were hybridized with 35S- and 32P-labelled probes, either singly or together.

Probes were labelled by random priming.

50μCi[α-³²P]dCTP (specific activity 3000 Ci/mmol; Amersham U.K.), or 25 μCi[α- 35SJdATP (specific activity >1000 Ci/mmol; Amersham U.K.) were used to label 50 ng of T-cell receptor β-C region cDNA, and mouse β-actin cDNA. Unincorporated nucleotides were removed by Sephadex G-50-300 chromatography spin

Q columns (Sigma, U.K.). Specific activities of »2x10 dpm/μg for 32P-, and *5x108 dpm/μg for 35S-labelled probes were obtained. Probes were heat denatured for

10 minutes and added to the hybridization mixture at

«2x10 dpm/ml.

Filters were prehybridized overnight at 42°C in 50% formamide, 10xDenhardt's 1.5xSSPE, 1% SDS, 0.5 mg/ml denatured sheared herring sperm DNA, and hybridized as above with the addition of 5% dextran sulphate and the labelled probe(s). Hybridized filters were washed in 2xSSC, 0.5%SDS and 0.5xSSC, 0.1% SDS for 15 minutes each at room temperature followed by a high stringency wash in O.lxSSC, 1% SDS for 10 to 30 minutes at 65°C. The water-wet filters were covered in Saran wrap and exposed to autoradiographic film overnight. The filters were air dried and prepared for scintillation counting (see below) .

Flat bed Scintillation counting: Calibration and Setting the Discriminator Gates

The Betaplate TM counter detects light produced by radioactivity on filters which are sealed in plastic bags with scintillation fluid, these filters, sandwiched between two scintillant soaked glass-fibre sheets are placed in a rigid cassette positioned in the counter. The cassette moves automatically and radioactivity is detected with high counting efficiencies.

The counter has six pairs of photomultiplier detectors. Slight differences in counting efficiency and spectrum shape for each pair of detectors was determined and allowances made by normalization before quantitative counting of samples were carried out. The differences were stored and the calibration applied in a software programme for use in dual-label measurements. Normalization was carried out by preparing six 10μl samples of either 3~^ςS- or 32P-labeiled probe

(20,000 cpm/sample) which were spotted in a row, fixed by U V crosslinking and air dried. Scintillant was added and the standards were counted by the 'sandwich method' as described below. The Betaplate TM machine has a pre-set programme which allows a single detector to count each standard, and the relative count rate is stored. This can be calculated to compensate for any pipetting errors incurred while preparing the standards. The filter was rotated and each of the six pots counted separately by the six detectors, in such way that the machine could calculate the relative counting efficiency of each pair of detectors for each isotope.

32 P and 35S have different energy emissions which can be shown by pulse height analysis in up to 1024 channels in this machine. 32P has a spectrum between channels 100 and 900, and S with a lower energy emission between channels 50 and 600. 32P/35S channel spillovers were determined for the six sets of photomultipliers and incorporated into the automatic programmed for dual-labelling.

Discriminator gates were chosen for this analysis so that about 20% of the 35S spectrum was included in the 32P counts (channels 500-850) and about 10% of the 32P spectrum in the 35S counts (channels 300-600) . The lower gate for 35S was placed at channel 300 which helped to reduce ^'cross-talk', which appeared mostly as low-energy counts below channel 300. The counting efficiency of 35S was only reduced by about 15% by this restriction, but 'cross¬ talk^' was reduced to <0.1%.

To show that the differences were due to differences in detectors rather than the particular sample spot and filter characteristics, the filter was inserted so that the sample counted by detector 1 was moved to position 6 and recounted, detector 2 to position 4 and so on.

Counting and Detection of RNA Dot Blots

Filters were placed in thin plastic counting bags (Pharmacia-Wallac) with 1 ml of scintillant

2 (Betaplatescint, Pharmacia Wallac) per 60 cm of filter. They were sandwiched between two glass fibre sheets (Pharmacia-Wallac 1205-401) each sealed in a plastic bag with 10 ml Betaplatescint TM (Pharmacia- Wallac) which made it possible to separate the pulse- height spectra of 35S and 32P.

Six samples were counted simultaneously for

60 minutes. This counting time was necessary to increase the counting precision for low count rates. The count rates were corrected by the computer programme according to the stored calibration data, to give the corrected cpm for each isotope. The programme also took into account the different isotope half-lives. Theoretical standard error (attributable to the number of events counted) was computed taking into account all of the dual-label and ^'normalization' calculations. An error could therefore be calculated for ratios of isotope activities. The measurement in cpm was proportional to the amount of probe bound to the filter and hence the amount of specific mRNA sequences in a given sample area.

The hybridization efficiency for 35S and 32P using either the β-actin or the T-cell receptor β-C region probe was compared. As expected, β-actin mRNA was present in all of the samples, with the exception of yeast t-RNA which acted as a negative control. Large amounts of T-cell receptor mRNA were present in the T-cell line, and lesser amounts detected in the inflamed aortic tissue and PBMC. TCR mRNA was not detected in the MCF7 breast tumour line.

Overnight autoradiographs of the Saran wrapped water-wet dot blots showed strong signals for hybridized 32P, but little hybridized 35S was apparent. After counting, the scintillant soaked filters were re-exposed to film and after 3 days well- defined spots were also seen in the S singly- labelled filters. Light produced by the scintillant made autoradiography of the 35S much more effective.

Similar signals were found for the single and dual-labelled dot blots for the P- and the S- labelled probes for both the TCR and the β-actin.

Scintillation counting showed a very similar pattern of activity when either single or double- labelling was employed. Counts for β-actin for 35Ξ were up to 2500 cpm, and for 32P up to 7000 cpm. Counts for TCR for 35S were up to 4000 cpm, and for

32 P up to 10000 cpm. Background counts (buffer and yeast t-RNA, between 0 and 100 cpm) were subtracted from each of the RNA sample counts. Figure 2 shows the relative ratios of single and dual labelled dot blots for β-actin, for TCR and for the ratio of TCR over actin. Repeats for both combinations of dual- labelling hybridisation reactions were carried out and gave similar results.

Example 3

Materials, Methods and Results

Dual-labelled dot-blots for DNA were carried out, as above. A 32P-labelled probe for the nucleic acid of interest, a virus genotype, and a S-labelled probe for β-actin, were simultaneously hybridised to samples.

Initially, CaSki cells -that- contain 300-500 copies of HPV 16 per cell were hybridized with 32P- labelled whole viral genome in pAT153 (Seedorf et al, 1985). Good linearity was observed for HPV16 plotted against actin counts. HPV 16 was detected at a lower level in SiHa cells having 1-2 copies per cell but not in HeLa cells that contain HPV18 and which served as a negative control.

To test the feasibility of this system on clinical material, DNA samples from 13 patients were analysed without knowledge of their HPV status. From the measured radioactivity from the 32P-labelled HPV16 probe alone, clear positive results were evident for patients 2, 5, 10, 12 and 13. Patients 1 and possibly 8 gave a small signal. Patients 3, 4, 6, 7, 9 and 11 had no signal and usually it would not be known whether this was because DNA of any kind was absent or the sample was virus negative. The double label results are shown in Figure 3. Only a small amount of DNA is present for patient 1 but this is commensurate with the virus signal, as are the results for the other positive samples 2, 5, 10, 12 and 13. Sample 8 has little DNA so this result must be regarded as indeterminate; samples 3 and 11 are clearly false negatives as no human DNA is present. In clinical practice samples 3 and 8 would be repeated.

After these results were obtained they were compared with previous analyses. Measured radioactivity from the dot-blots was 'rated (- to +++) and compared with the results from PCR and non- isotopic insitu hybridisation (NISH), as shown in Table 3 below. Excluding samples 3 and 11, samples 7 and 9 were negative for HPV 16 but contained HPV 33 by NISH. Sample 4 was normal squamous epithelium. these samples showed the low level of non-specific binding in this assay. Sample 6, which was negative by dot- blot, had a NISH pattern characteristic of very small amounts of integrated virus (Cooper et al, 1991). This sample was negative by Southern blot but weakly positive by PCR. All the other positive samples had episomal NISH patterns except for the highly positive sample 12, with both episomal and integrated patterns. We therefore show that more rational decisions may be made for each sample of DNA as to whether they are positive or negative for virus and which samples have indeterminable status and need to 5 be repeated. Statistical criteria can be drawn up to provide a consistent and reliable system. The ratio between counts derived from virus and host DNA may give additional information as this gives the amount of virus present per total cellular mass in the sample. 10

T5

20

25

30

35 Table 1

Table 2

Table 3 ^'

HPV status in cervical cells shown by dual label dot blot and NISH.

* Based on cpm values shown in Figure 3. ** Processed as detailed in Cooper et al. 1991.

Figure 1- Frequency distribution of dual-label hybridization results for int-2 amplification levels of blood and tumour DNA from the 35-case population experiment.

Figure 2 Histogram showing the relative ratios between samples in single and dual- labelling. T-cell receptor cpm/actin cpm of dot blots with separate filters or with double label. The RNA sources are A. Hut 78, T-cell line; B. Aortic inflammatory tissue; C. MCF 7 breast cell line; D. Peripheral mononuclear cells.

Figure 3 Count rates for labelled probes hybridised to dot blots of DNA derived from cervical scrape preparations. Simultaneous measurement of activity using 32 P-labelled probe fc 35 probe for HPV16 and S-labelled probe for β-actin.

REFERENCES

Callahan, R. (1989). Breast Cancer Res. Treat., 13, 191-203.

Callahan, R. & Campbell, G. (1989). J. Natl. Cancer Inst., 81, 1780-1786.

Casey, G., Smith, R., NMcGillivry, D. & Dickson, C. (1986). Mol. Cell Biol., 6, 502-510.

Cooper, K, Herrington, C. S., Graham, A. K., Evans, M. F. and McGee, J. O^'D. In situ human papillomavirus (HPV) genotyping of cervical intraepithelial neoplasia in South African and British patients: Evidence for putative integration in vivo. J. Clin. Pathol. 1991, 44:400-405.

Cooper, K, Herrington, C. S., Stickland, J. E., Evans, M. F. and McGee, J. O'D. Episomal and integrated human papillomavirus in cervical neoplasia shown by non-isotopic in situ hybridisation. J. Clin. Pathol. 1991, 44:960-966.

Dallman, M. J., Larsen, C.P. and Morris P. J. Cytokine gene transcription in vascularized organ grafts: analysis using semiquantitative polymerase chain reaction. J. Exp. Med. 1991; 174, 493-6.

Donovan-Peluso, M., Contento, A. M., Tobon, H., Ripepi, B. & Locker, J. (1991). Am. J. Pathol., 138, 835-845.

Elston, C. . and Ellis, I. O. (1991). Histopathology, 19, 403-410.

Feinberg, A. & Vogelstein, B. (1983). Anal. Biochem., 132, 6-13. Laskey, R. A. The use of intensifying screens or organic scintillators for visualising radioactive molecules resolved by gel electrophoresis. Methods Enzymol, 1980, 65, 363-7.

Maddox, J. (1992). Nature, 355, 201. Machotka, S. V., Garret, C. T., Schwartz, A.

M. & Callhaan, R. (1989). Clin. Chem. Acta. 184, 207-217. Peters, G., Brookes, S., Smith, R. & Dickson, C. (1983). Cell, 33, 369-367.

Peters, G., Brookes, S., Smith, R., Pllaczek, M. & (1989). Dickson, C. Proc. Natl. Acad. Sci. USA, 86, 5678-5682.

Potter, C. G., Warner, G. T., Yrjonen, T. and Soini, E. A liquid scintillation counter specifically designed for samples deposited on a flat matrix. Phys. Med. Biol. 1986, 31, 361-9. Potter, C. G., Tan, C. C. and Ratcliffe, P. J. Quantification of 32P-labelled samples in gel fragments, using the flat-bed liquid scintillation counter. Anal. Biochem, 1991, 197, 121-4.

Potter, C. G. and Le Jeune S. Quantitative analysis of nucleic acid dot blots using the flat-bed scintillation counter. Technique 1991; accepted for publication.

Sambrook, J., Fritsch, E. F. & Maniatis, T. (1985) . Cold Spring Harbor Laboratory, Cold Spring Harbor: NY.

Seedorf, K., Krammer, G., Durst, M., Suhai, S. and Rowenkamp, W. G. Human papillomavirus type 16 sequence. Virology 1985, 145:181-185.

Stark, G. R., Debatisse, M., Giulotto, E. & Wahl, G. M. (1989), Cancer Res., 57, 901-908.

Steel, M., Thompson, A. & Clayton, J. (1991). British Medical Bulletin, 47, 504-508.

Tsuda, H., Hirohashi, S., Shimosato, Y., Hiroti, T., Tsugane, S., Yamomot, H., Miyajima, N., Toyoshima, K., Yamomoto, T., Yokota, J., Yoshida, T., Sakamoto, H., Terada, M. & Sugimura, T. (1989). Cancer Res. 49, 3104-3108.

Warner, G. T., Potter, C. G., Yrjonen, T. and Soini, E. A new design for a liquid scintillation counter for micro samples using a flat-bed geometry. Int. J. Appl. Radiat. Isot. 1985, 36, 819-21.

Claims

1. A method of quantifying-a first nucleic acid sequence in a test sample containing a mixture of nucleic acid sequences including a second nucleic acid sequence, by providing a first labelled sequence specific probe capable of hybridising with the first nucleic acid sequence and a second labelled sequence specific probe capable of hybridising with the second nucleic acid sequence, which method comprises incubating the test sample with the two probes under hybridising conditions, measuring the extent of hybridising of each of the two probes to its respective nucleic acid sequence, and determining a ratio of the two said measurements as indicative of the quantity of the first nucleic acid sequence in the test sample.

2. A method of detecting the presence or amount of a viral nucleic acid sequence in a test sample containing a mixture of nucleic acid sequences including a second nucleic acid sequence, by providing a first labelled sequence-specific probe capable of hybridising with the viral nucleic acid sequence and a second labelled sequence-specific probe capable of hybridising with the second nucleic acid sequence, which method comprises incubating the test sample with the 2 probes under hybridising conditions, measuring the extent of hybridising of each of the two probes and comparing the two measurements to determine the presence or amount of the viral nucleic acid sequence.

3. A method as claimed in claim 1 or claim 2, wherein each nucleic acid sequence is a DNA sequence such as a gene or a part of a gene.

4. - A method as claimed in any one of claims 1 to 3, wherein there is also provided a control sample which is also tested by the method as defined, and the ratio obtained for the test sample is compared to the ratio obtained for the control sample as indicative of the quantity of the first nucleic acid sequence in the test sample.

5. A method as claimed in claim 4, wherein the test sample comprises tumour DNA from a patient and the control sample is of constitutional DNA of the same patient.

6. A method as claimed in claim 1 or claim 2, wherein each nucleic acid sequence is an RNA sequence.

7. A method as claimed in any one of claims 1 to 6, wherein the first sequence-specific probe is labelled with 32-Phosphorus and the second sequence- specific probe is labelled with 35-Sulphur.

8. A method as claimed in claim 7, wherein the two radiolabels are counted simultaneously in a flat- bed scintillation counter.