ARTICLE pubs.acs.org/ac
Detection of Copy Number Changes in DNA from Formalin Fixed Paraffin Embedded Tissues Using Paralogue Ratio Tests Gerald Saldanha,* Linda Potter, Lovesh Dyall, Danielle Bury, Nisreen Hathiari, Abdlrzag Ehdode, Edward Hollox,† and James Howard Pringle Department of Cancer Studies and Molecular Medicine, University of Leicester, LE2 7LX, United Kingdom ABSTRACT: The purpose of this study was to evaluate whether paralogue ratio tests (PRT) using real-time PCR can accurately determine the DNA copy number (CN) using formalin fixed paraffin embedded (FFPE) tissue. Histopathology diagnostic archives are an enormous resource of FFPE tissue, but extracted DNA is of poor quality and may be unsuitable for CN assessment, thus representing a missed opportunity for studies of genetic association and somatic change in cancer in large cohorts of easily accessible samples. Assays with paralogues on chromosomes 18 and 20 (18|20 PRT) and chromosomes 13 and X (13|X PRT) were tested using archived FFPE pathology samples with known CN, including tonsil, placentae, and FFPE melanoma cell lines. The assay proved accurate over the dynamic range from 1:1 to 1:3 and gave precise results when repeated four times over several weeks. The precision of the assay was marginally reduced once the CT value for 10 ng of FFPE DNA increased above 30 cycles, reflecting importance of DNA quality. The assays distinguished changes in CN ratio with high sensitivity and specificity. The 13|X PRT could detect cells with distinct genotypes microdissected from within the same FFPE sample. Therefore, PRTs are suitable for analyzing CN in FFPE tissues.
T
he purpose of this study was to demonstrate proof of concept that a novel application of the paralogue ratio test (PRT) assay1,2 using real-time TaqMan-based measurement of PCR amplification products can measure DNA copy number accurately, precisely, and with a wide dynamic range in formalinfixed paraffin-embedded (FFPE) tissue. The analytical validity of PRT has been described by one of us previously using good quality genomic DNA prepared from fresh frozen tissues or cells,1,3 and the assay performs well under these conditions. However, there have been no attempts to rigorously test PRT performance using degraded DNA from FFPE tissue, and the analytical validity in this situation is unknown. Measurement of DNA copy number changes is important in human disease. Germ line copy number variants (CNVs) influence susceptibility to multifactorial diseases4 such as psoriasis,3 and DNA copy number changes occur as somatic events in cancer cells as part of genomic instability.5 Some of these are of potential diagnostic, prognostic,68 and therapeutic relevance.9 Nevertheless, few assays that use DNA extracted from FFPE tissue have robust evidence of analytical validity,10 namely, the demonstration that a test measures what it is supposed to, that it does so with accuracy, and that the same sample yields the same result on different occasions, i.e., the test is precise. Using FFPE tissue is difficult because the combination of formalin fixation and processing into paraffin wax leads to DNA degradation and cross-linking. Other variables such as the time the sample spent in formalin and the age of a clinical sample mean that DNA analysis can be extremely capricious.11 The small number of analytically validated assays that work on archived FFPE samples limits potential studies using this valuable clinical resource, and this is a missed opportunity because these samples are easily accessible and r 2011 American Chemical Society
numerous. Development of new analytically validated assays to measure DNA copy number using FFPE tissue is therefore a priority. For example, these samples could be used to investigate the effect of candidate CNVs on multifactorial disease susceptibility.12 Use of archived clinical samples would mitigate the requirement for prospective collection of fresh material, making large studies simpler and more cost-effective. In addition, large numbers of cancer specimens reside in clinical archives, and they could be used to investigate somatic DNA copy number variation related to cancer pathogenesis. Use of FFPE archived diagnostic tissue is particularly attractive when one wishes to assess whether a CNV is associated with a timedependent end-point such as disease progression or death, where it is relatively simple and quick to assemble a large, powerful retrospective cohort. Various technologies have been used to analyze DNA copy number changes in FFPE tissues, including fluorescence in situ hybridization (FISH),13 comparative genomic hybridization (CGH), array CGH,14 digital PCR,15 and multiplex ligation-dependent probe amplification.16 FISH is the most well-known because of its clinical use for assessing HER2 status in breast cancer17 but is limited by the number of loci that can be tested in a given experiment and involves estimation of copy number in only a small sample of all the nuclei of a specimen.13 Digital PCR using the Fluidigm system is very promising and may afford good accuracy and dynamic range even with degraded DNA from FFPE tissue but has significant hardware requirements.18,19 Next generation sequencing is also promising but routine screening of hundreds of FFPE Received: January 19, 2011 Accepted: March 21, 2011 Published: March 30, 2011 3484
dx.doi.org/10.1021/ac200153j | Anal. Chem. 2011, 83, 3484–3492
Analytical Chemistry
ARTICLE
Table 1. Details of Cell Lines Used and Chr:13|X and Chr:18|20 PRT Primers and Probes with Genomic Location cell line
age (years)
sex
population
site of originating lesion
A375
cell holdings
54
female
unknown
unknown
RPMI 7951
ATCCa
18
female
European origin
lymph node
Sk-Mel 2
ATCC
60
male
European origin
thigh
Sk-Mel 5
ATCC
24
female
European origin
axillary lymph node
Sk-Mel 28
ATCC
51
male
unknown
unknown
primer/probe
a
source
sequences
location
chain ID
13_X F primer
GTGTTCTGGCTGCTACAGGA
chr13:41380942, chrX:141214329
227470b
13_X R Primer Chr.13 Probe
AAGCACCAAGAATTTATTTCTCA VIC-TCTTCTAAGCAGCCCTGC-MGB
chr13:41381026, chrX:141214420 chr13:41380979
227470b
Chr.X Probe
FAM-CTAAGCAGATCACAGCCCT-MGB
chrX: 141214370
18_20 F primer
CCAGCTACACCTGGCAGTTTGA
chr18:72998567,chr20:51871094
184624c
18_20 R primer
GGGYGGTGAGCTGCTGCA
chr18:72998662,ch20:51871189
184624c
Chr.18 Probe
VIC-TTGAGGATCTGCGCCT-MGB
chr18:72998612
Chr.20 Probe
FAM-CTTTAAGATCTGGGACTT-MGB
chr20:51871140
ATCC = American Type Culture Collection. b Assembley: GRCh37/hg19. c Assembley: NCBI35/hg17.
samples by targeted or whole genome sequencing at sufficient read depth to accurately infer copy number would be very difficult at the present time. Here, we show that PRT can accurately and precisely assay DNA copy number in genomic DNA from FFPE tissue. To develop a PRT assay, a pair of paralogues is identified, one at the test locus and one elsewhere that serves as the reference locus.1,2 By careful primer design, both test and reference paralogues can be amplified by a single primer pair because of shared sequence identity, yet each paralogue can be distinguished because of sequence difference, allowing specific detection of each paralogue with differently labeled TaqMan probes. Thus, the ratio of the two amplicons can be measured in the same PCR using a single primer pair, minimizing the effects of variable PCR efficiency and amplification kinetics and so providing an accurate indication of the relative copy number of the test locus. PRT can be used to determine the presence or absence of aneuploidy, with any significant deviation from a ratio of one consistent with aneuploidy.2 Aneuploidy is present in malignant melanoma but not in the benign counterpart, melanocytic naevus,20 and FISHdetected aneuploidy has been used to classify equivocal melanocytic tumors as melanoma.6 An effective diagnostic test based on aneuploidy might require a series of PRTs at different loci across the genome to produce sufficient overall sensitivity. Importantly, PRT assays have significant advantages of simplicity and cost compared to a whole genome approach for aneuploidy detection using aCGH. For other types of studies, a very accurate estimation of copy number is required. For example, assessment of CNVs in genotyping studies typically requires exact assessment of copy number so that it can be associated with a phenotype. In this scenario, a battery of analytically validated PRTs targeting the same locus could be developed. The data from each PRT could then be used together to accurately infer integer copy number using statistical analyses such as maximum likelihood.21 To determine whether PRT using FFPE tissue might be an analytically valid assay that could be further developed for these applications, we assessed two different, novel PRT assays for accuracy and precision using samples with known DNA copy number and then used the assays to identify cancer-related copy number alterations in FFPE cell lines and archived clinical
samples after microdissection, the latter being the most challenging application of an assay to measure DNA copy number.
’ EXPERIMENTAL SECTION Samples. FFPE tissue blocks were from the Pathology archive of the University Hospitals of Leicester between 1983 and 2008. These included placental tissues with confirmed trisomy for chromosomes 13 and 18 alongside placentae from normal pregnancy. Reactive tonsil tissues were also obtained from this archive. Samples were not “cherry-picked” based on DNA quality. The only exclusion criterion was if there was insufficient material remaining in the tissue block. Melanoma cell lines (Table 1) were obtained from the American Type Culture Collection (ATCC) or Cell Holdings and were cultured as recommend by the repository. Melanoma cell lines were formalin fixed and paraffin embedded using the Shandon cytoblock method (Shandon Cheshire).22 Microdissection and DNA extraction from FFPE tissues and cell lines were performed as described previously.23 The eluted DNA was quantified using the Nanodrop ND-1000 Spectrophotometer (Thermo Scientific, Wilmington, USA) and stored at 4 °C. Local Ethics Committee approval was obtained for this study. Design of Paralogue Ratio Test. Paralogous sequences were identified for two loci using the online genome browser from the University of California Santa Cruz (UCSC Genome Browser).24 UCSC Genome Browser was programmed to show the full human chain self-alignments. To produce these, the human genome was compared to itself using blastz, trivial alignments were filtered out, and remaining alignments were formatted and arranged into chains to display all alignments between a single target chromosome and a single query chromosome.25.26 These were used to select suitable paralogous sequences for each assay. Quantitative PCR Design and Validation. Primer Express v2.0 software (Applied Biosystems) was used to design minor groove-binding (MGB) hydrolysis probe assays for each paralogous sequence. Areas of homologous sequences were located to determine common priming sites, and areas of mismatch sequences were used to develop specific probes for each locus. Table 1 shows the primer and probe sequences for PRT assays 3485
dx.doi.org/10.1021/ac200153j |Anal. Chem. 2011, 83, 3484–3492
Analytical Chemistry
ARTICLE
Figure 1. TaqMan PCR is shown in the top left for 18|20 PRT. In a diploid cell, the chromosome 18 (green) and chromosome 20 (red) trace is seen to overlap because of equal amplification. In contrast, the chromosome 18 trace appears earlier due to increased copy number relative to chromosome 20 in a trisomy 18 sample. The PCR primers and probes are shown at the bottom.
along with their genome location and chained self-alignment ID for the version of the human genome assembly. We confirmed the specificity of these assays by conducting a primer BLAST search (http://www.blast.ncbi.nlm.nih.gov) and UCSC Genome Browser In-Silico PCR analysis. Potential single nucleotide polymorphisms and copy number polymorphisms were also investigated using the NCBI dSNP database and the International HapMap project database (The International HapMap Consortium).27 PCR reactions were performed as previously described for ABI (MGB) hydrolysis probe assays using an ABI 7500 or ABI StepOne Fast Real Time PCR Systems.28 Briefly, the concentration of probe was titrated (50200 nmol/ L) to reduce background fluorescence. The PCR efficiencies were evaluated using a series of 2-fold dilutions of pooled tonsil DNA samples (200.156 ng/reaction) for each probe individually. Reactions containing 1020 ng of genomic FFPE DNA were pipetted into duplicate wells in a 96-well plate, and each plate included positive controls and no template controls. PRT assays were carried out in 10 μL reactions, and each well contained the following: 3.4 μL of genomic DNA (1020 ng), 5 μL of Genotyping MasterMix (Applied Biosystems, Foster City, CA), 1.6 μL mix of forward and reverse primers (1020pmol/L) with probes for each locus (50200 nmol/ L). All PCR reactions were prepared as master mixes and were transferred to 96-well plates with an electronic multichannel pipettor. The plates were sealed, PCR amplification was carried out, and real-time fluorescence data was collected. The thermal profile consisted of incubation at 95 °C for 60 s followed by 40 cycles of 95 °C for 15 s and 60 °C for 60 s. Fluorescence data were collected during the 60 °C annealing/extension cycles, and cycle threshold (CT) values were calculated with the automatic CTanalysis settings on an absolute quantification setting. An example plot is shown in Figure 1. The Amelogenin PCR DNA sex test
was used to verify the sex of fetal tissue and was carried out using the primers developed by Sullivan et al.29 Data Analysis. The difference in CT value between the target and reference loci was calculated (ΔCT = CT target CT reference) for each duplicate, and this value was used to compare samples. Tonsil samples were used as a control diploid population and a ΔΔCT value for each cell line or placental tissue sample was calculated by subtracting the ΔCT value for the sample from the mean tonsil ΔCT. The PRT can be calculated from the following equation: PRT= 2ΔΔCT. Statistical analysis was performed in SPSS v14.0 and Microsoft Excel 2007. Graphs and other statistical diagrams were designed in GraphPad Prism version 5.02 and Microsoft Excel 2007.
’ RESULTS Accuracy of PRT Using DNA from FFPE Tissue. To determine whether PRT is accurate using DNA extracted from FFPE tissue, samples were selected with known DNA copy number at the genomic regions to be tested. We designed an assay using paralogues on chromosomes 18 and 20 (18|20 PRT) and analyzed 21 samples comprising 12 reactive FFPE tonsils that were presumed to be diploid, three FFPE placentae from normal pregnancies, three placentae from karyotype verified trisomy 13 (Patau syndrome) and known to be diploid at chromosomes 18 and 20, and three FFPE placentae from fetuses with karyotype verified trisomy 18 (Edwards syndrome). Assuming that the ratio of paralogues on chromosome 18 and 20 in tonsils, normal placentae, and trisomy 13 placentae would be 1:1, we expected to distinguish these from the trisomy 18 placentae, where the expected ratio should be 1:1.5. PRTs were performed on one to four occasions per sample, the number of assays relating only to the amount of available DNA rather than experimenter-chosen 3486
dx.doi.org/10.1021/ac200153j |Anal. Chem. 2011, 83, 3484–3492
Analytical Chemistry
ARTICLE
Figure 2. Paralogue ratio tests for chromosomes 18|20 and 13|X. Panel A shows the 18|20 PRTs for all samples, including repeat assays on the same sample. This shows that trisomy 18 samples with a ratio of 1:1.5 can be distinguished from diploid 18 samples with a ratio of 1:1, with no overlap in paralogue ratio between any samples. Panel B shows the same 18|20 PRT samples, but with average data for all those where more than one assay was done. Panel C shows the average 18|20 PRT data for placental tissue only. Panel D shows 13|X PRT samples used and their expected paralogue ratios. Panel E shows the results of the assays revealing that the X|13 PRT can distinguish each ratio.
exclusion criteria. The mean value of tonsils was normalized to a ratio of 1 and used to calculate a ΔΔCT value for each sample. A plot of every PRT, including those repeated on the same sample on different days, is shown (Figure 2A). For trisomy 18 placentae, the mean was 1.37, SEM 0.03, n = 12, while for samples that were diploid at chromosome 18, the mean ratio was 1.00, SEM 0.02, n = 50, t(60) = 11.09, p < 0.0001, r2 = 0.68. The range of trisomy 18 sample ratios was 1.25 to 1.62, while that for chromosome 18 diploid samples was 0.80 to 1.21, showing no overlapping values.
The slightly lower than expected mean for trisomy 18 placentae (1.37 rather than 1.5) is likely to be due to contamination by decidua, i.e., maternal cells, since a small amount of this was present in most of the placental samples. In Figure 2B, the same data are shown except that the mean values for samples assayed on more than one occasion is shown. The three placentae from fetuses with trisomy 18 had a mean paralogue ratio of 1.37, SEM 0.02, while the remaining 18 samples with diploid copy number at chromosomes 18 and 20 had a mean ratio of 1.00, SEM 0.02, 3487
dx.doi.org/10.1021/ac200153j |Anal. Chem. 2011, 83, 3484–3492
Analytical Chemistry t(19) = 7.30, p < 0.0001, r2 = 0.74. These data indicate that it is possible to distinguish small differences in ratio, i.e., 1:1 from 1:1.5, and suggest that high sensitivity and specificity might be possible, even when a given sample’s ratio is based on a “one-off” assay as might be performed in a clinical context (Figure 2A). Only considering the placentae, those with 2 copies of chromosomes 18 and 20 (normal pregnancy n = 3; Patau syndrome n = 3) had a mean paralogue ratio of 1.00, SEM 0.04, as shown in Figure 2C. This was significantly different from the three trisomy placentae, t(7) = 1.698, p = 0.0004, r2 = 0.85. Dynamic Range of PRT Using DNA from FFPE Tissue. To determine the dynamic range of PRT using DNA from FFPE samples, we analyzed nine FFPE reactive tonsils plus the FFPE placentae analyzed above. Thus, we had a range of expected ratios for an assay where paralogues are located on chromosomes 13 and X (Figure 2D,E), which we duly designed (13|X PRT). Diploid females had a mean paralogue ratio of 1.01, SEM 0.03; trisomy 13 females of 1.34, SEM 0.01; diploid males of 2.08, SEM 0.11; and trisomy males only comprised a single case whose ratio was 2.55. Because there were only two female trisomy samples and one male trisomy sample, all the groups were not formally compared. A comparison was made between female diploid samples and male diploid samples, where there were sufficient numbers of cases for valid analysis, t(13) = 11.29, p < 0.0001, r2 = 0.91. Taken together, the data from the 18|20 and 13|X assays suggest that PRT using DNA from FFPET reflects true DNA copy number. The assay proved accurate over the dynamic range tested. The two PRT assays detected very small changes in ratio (from 1:1 to 1:1.5) with good sensitivity and specificity, indicating that the assay would be suitable for simple categorization of aneuploidy presence or absence as used for diagnosis of melanocytic tumors.6 However, we only show evidence of accuracy over a dynamic range of copy numbers between 1:1 and 1:3, so further work may be required to validate the assay when assessment of a wider dynamic range is required, for example, in CNV association studies where high copy number variants exist. This will be important when investigators wish to compare PRT to competing technologies such as FISH or digital PCR.18,19 PRT’s main advantage is simplicity. In addition, PRT shows a trend toward underestimating copy number at ratios greater than 1:1. An important factor in this regard when placental tissue is used is contamination by maternal cells, which varies between samples. Effect of DNA Quality on PRT Precision Using DNA from FFPE Tissue. DNA from FFPE tissue is degraded. This results in shorter fragments and cross-links. Furthermore, typically, there is variability in the time between surgical excision and complete fixation in formalin. Rather than account for each of these variables, all of which have a bearing on DNA quality and quantity, we took the simple measure of assessing the amount of amplifiable DNA from a fixed starting quantity, which is likely to be a composite of all of these factors. Therefore, in PCRs containing 10 ng of DNA, we assessed the precision of the 13|X PRT assay as a function of the cycle threshold (CT) value, which was taken as a measure of the amplifiable DNA. To do this, we took the mean CT value for each paralogue within a reaction and plotted this against the difference between mean values in repeat assays that had been performed on different days. We found that precision was reduced once the CT value rose above 30 cycles, suggesting that under these circumstances the PRT result should be interpreted slightly more cautiously (Figure 3A). However, even then, the difference in mean CT value was never as much as 0.5 of a PCR cycle.
ARTICLE
Precision of 18|20 PRT in FFPE Tissue. We next looked at more general features of precision of the two PRT assays. For the 18|20 PRT, there were six cases where the assay was repeated on four occasions over a ten week period, comprising three FFPE trisomy 18 placentae and three FFPE placentae from normal pregnancies. To check for date bias between each assay, repeated measures analysis of variance was performed with assay date as the independent variable (Figure 3B). This showed no significant difference and therefore no evidence of date bias, with the mean of each assay being 1.24, 1.27, 1.22, and 1.18, F (3, 15) 0.820, p = 0.503, r2 = 0.14. All four repeat assays, considered individually, showed statistically significant differences in paralogue ratio between trisomy 18 and diploid placentae (Table 2). The overall range of values for the 12 assays in total that were done on trisomy 18 samples was 1.25 to 1.62, while the range for the 12 assays on diploid samples was 0.89 to 1.11, indicating once again that the 18|20 PRT assay has the potential to achieve high sensitivity and specificity for separating samples with small differences in paralogue ratios of 1:1 and 1:1.5. In addition, there were 17 cases where two repeat 18|20 PRT assays were performed, comprising the three trisomy 18 placentae, three trisomy 13 placentae, three placentae from normal pregnancies, and eight tonsils. A paired t test revealed no bias between repeat assays, t(16) = 1.298, p = 0.213, r2 = 0.095. The intraclass correlation coefficient (ICC) was 0.77, indicating good agreement;30 the mean difference in paralogue ratio between repeat assays was only 0.04, and the limits of agreement (i.e., two times the standard deviation of the difference between paralogue ratios in each repeat pair of assays) were 0.29 to 0.21 (Figure 3C). Thus, 95% of differences are expected to lie within this range. The small size of the limits agreement is further support that this assay is sufficiently precise to be used routinely. Precision of 13|X PRT in FFPE Tissue. Next, the 13|X PRT was assessed for precision. PRT was conducted on two occasions on the same 18 samples analyzed for dynamic range above. A paired t test revealed no significant difference between mean ratios in the two repeat assays, t(17) = 0.583, p = 0.568, r2 = 0.020, meaning that there was no bias. The ICC was 0.96, indicating very good agreement, the mean difference in paralogue ratio between repeat assays was only 0.02, and limits of agreement were 0.36 to 0.31, supporting the proposal that this PRT has good precision (Figure 3D). Together, these data show that two different PRTs give precise results when the assays are performed on different occasions on the same sample, including in assays that were repeated four times over several weeks. These latter assays compared a small change in ratio, from 1:1 to 1:1.5, yet the assay results showed high sensitivity and specificity. 18|20 PRT Analysis of DNA Copy Number Change in FFPE Melanoma Cell Lines. DNA copy number changes in melanoma have potential diagnostic value when assayed with FISH.6 Chromosome 18q loss and 20q gain are found in melanoma;20 therefore, one might expect that 18|20 PRT in some melanomas would show a paralogue ratio greater than one (please note the ratio in this experiment is the inverse to that calculated in the above experiments to yield a ratio >1 in melanomas). DNA from four melanoma cell lines, SkMel5, SkMel28, A375M, and RPMI 7951, was prepared from different sources: fresh, fresh sonicated (to mimic the short fragments found in FFPE tissue), and FFPE cytoblocks. The paralogue ratio for these samples was estimated from array CGH data extracted from the Cancer Genome Project at the Sanger Institute Web site (http://www.sanger.ac.uk) based 3488
dx.doi.org/10.1021/ac200153j |Anal. Chem. 2011, 83, 3484–3492
Analytical Chemistry
ARTICLE
Figure 3. Precision of PRT. (Panel A) Thirty six PRT assays with paralogues on chromosomes X and 13 were performed on two occasions using male and female diploid and trisomy 13 samples. The x axis shows the mean CT value for the paralogue on X chromosome and the paralogue on chromosome 13. The y axis shows the difference in the mean CT values between the two repeats of the assay for a given sample. The mean difference (solid line) and 2 SD of the mean difference (broken lines) are shown. The difference between repeats is more marked once the average CT value is above 30. However, even under these circumstances, the difference in mean CT value was never more than 0.5 of a cycle. Panel B shows four repeat 18|20 PRT assays on six placentae. Panel C shows a Bland Altman plot demonstrating means vs differences for 17 18|20 PRT assays repeated on different days. There is no bias associated with magnitude of the mean paralogue ratio. Panel D shows a Bland Altman plot for the 13|X PRT for 18 repeat assays. The plot suggests increased variability as the magnitude of the mean paralogue ratio increases but the differences remain small. The open circle containing a cross represents three overlapping data points. Together, the data in the four panels demonstrate excellent agreement between repeat assays. ICC = intraclass correlation coefficient.
Table 2. Comparison 18|20 PRT Using Trisomy 18 and Diploid Placentae Assayed on Different Days
a
assay date
trisomy mean (SEM)
diploid meana (SEM)
test statistics
28/10/2009
1.48 (0.03)
1.00 (0.06)
t(4) = 7.525, r2 = 0.93, p = 0.017
18/12/2009 21/12/2009
1.53 (0.01) 1.44 (0.09)
1.00 (0.06) 1.00 (0.05)
t(4) = 9.306, r2 = 0.96, p = 0.0007 t(4) = 4.447, r2 = 0.83, p = 0.013
6/1/2010
1.35 (0.07)
1.00 (0.05)
t(4) = 4.101, r2 = 0.81, p = 0.0148
Each assay was normalized to the mean diploid placenta.
on the PICNIC algorithm.31 The array CGH-predicted PRT values were 2 for A375M, 2 for RPMI 7951, 2.6 for SkMel28, and 1 for SkMel5. Seventeen FFPE tonsils were also analyzed in the same experiment and used for normalization of cell line data. The paralogue ratio for each sample was derived from the mean of two independent assays. The four melanoma cell lines had a mean 18|20 paralogue ratio of 1.62, SEM 0.23, while the 17 tonsils had a mean ratio normalized to 1.00, SEM 0.04, which is consistent with
a mean relative gain of 20q in melanoma, t(20) = 4.446, p = 0.0002, r2 = 0.50 (Figure 4A). The type of melanoma cell line DNA extract appeared to have no effect on the paralogue ratio (Figure 4B). The ICC was 0.90, indicating very good agreement between the paralogue ratios measured from different types of DNA. The mean paralogue ratio for fresh DNA was 1.92, SEM 0.19, for sonicated DNA 1.92, SEM 0.14, and for FFPE DNA was 1.79, SEM 0.21, and was similar to ratios predicted from aCGH data. 3489
dx.doi.org/10.1021/ac200153j |Anal. Chem. 2011, 83, 3484–3492
Analytical Chemistry
ARTICLE
Figure 4. PRT using melanoma cell lines and microdissected tissue. Panel A shows the 18|20 paralogue ratio of 17 FFPE tonsils compared to four FFPE melanoma cell lines. The difference in paralogue ratio was statistically significant. Panel B shows the 18|20 paralogue ratio for each cell line broken down by source of the DNA. The crosses show the predicted ratio using the PICNIC algorithm based on Sanger Institute’s aCGH data on each cell line. Panel C shows a photomicrograph of a section of placenta. The arrowheads show decidua, which is of maternal origin, while the arrow shows placental villi, which are of fetal origin. Panel D shows a photomicrograph of a serial tissue section of the same placenta after microdissection. Panel E shows four different microdissected placentae, PL1PL4. PL1 was known in advance to be from a male. For each, there is a microdissected decidua (d) and villus sample (v). The gel shows amelogenin PCR where a single band indicates female and two indicates male. Thus, all microdissected decidua, being of maternal origin, are confirmed as female. In contrast, the villi show two males, PL1 and PL4, and two females, PL2 and PL3. The results of 13|X PRT are shown at the bottom of the gel. Panel F shows TaqMan PCR of the PL1 microdissected decidua and villi. The placenta in panels C and D is also PL1.
The latter shows that PRT data is supported by independently generated data from a different type of assay. PRT Analysis of DNA Copy Number Change in Microdissected FFPE Tissue. Finally, we assessed whether the most demanding application of PRT could be employed successfully: analysis of microdissected FFPE tissue containing cell populations with differing genotype. Four FFPE placentae not previously used in the study were selected because they were from normal pregnancies and contained both maternal decidua and
fetal placental tissue. The sex of the baby (male) was known beforehand for one of the placental samples. The sex was unknown for the other three, which were kept anonymous under the terms of the ethics approval for this study. We performed 13|X PRT on microdissected maternal decidua (expected ratio 1:1) and fetal placental villi (expected ratio 1:1 for female fetus and 1:2 for male fetus). Crude microdissection using a pipet tip was performed as described previously.23 In addition, amelogenin gene PCR was used for sex determination29 that takes advantage 3490
dx.doi.org/10.1021/ac200153j |Anal. Chem. 2011, 83, 3484–3492
Analytical Chemistry of a length variation in the X-Y homologues, AMELX and AMELY, whereby females show a single PCR product and males show two products. All microdissected maternal decidua samples showed the expected female genotype with both amelogenin PCR and 13|X PRT. Both methods also indicated the same sex for all four microdissected placental villus samples. In the one case where the sex of the baby was known in advance to be male, amelogenin PCR, 13|X PRT, and clinical data all agreed (Figure 4CF). Thus, the assay proved capable of distinguishing cells with different genotypes microdissected from the same sample despite the problem with degraded DNA being further compounded by of the possibility of cross-contamination between cell populations during microdissection. Overall, these data demonstrate accuracy and precision of PRT using degraded DNA from FFPE samples. The present study has some limitations which will need to be addressed in the future. Only samples processed in one clinical laboratory were used. Different laboratories may have slightly different processing. We did not formally assess the effect of such variables as ischemic time, time spent in formalin, or the age of the tissue block. However, the archival samples that we used were all at least 4 years old, with some of the tonsils being more than 10 years old. Rather than assessing each of these variables, we took the more simple measure of assessing the actual amount of amplifiable DNA, which is likely to represent a composite of each factor. This pragmatic approach has been used for array CGH, where the size of amplifiable PCR product has been used to determine likelihood of assay success.11
’ CONCLUSIONS We describe a new application for the PRT assay, namely its use in analysis of degraded DNA from FFPE samples. We have demonstrated the assay’s analytical validity by showing accuracy and precision using two novel PRT assays, thus showing proof of concept that this assay could be further developed for investigation of DNA copy number changes in FFPE samples. Furthermore, because this assay is real-time PCR-based, it has the advantages of simplicity, scalability, and cost-effectiveness and uses equipment present in most molecular biology laboratories. In addition, a tissue assay that requires no change in clinical specimen handling would have good acceptability in pathology laboratories. This means that PRT has potential as the basis for clinical assays. ’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected]. Fax: þ441162523274. Present Addresses †
Department of Genetics, University of Leicester, United Kingdom.
’ ACKNOWLEDGMENT We would like to thank Dr. Helen Porter for providing placenta samples and helping us to identify appropriate areas for microdissection. We also thank Dr. Mike Biggs for taking the placenta photomicrographs and Eleanor Graham for the amelogenin gene primers.
ARTICLE
’ REFERENCES (1) Armour, J. A.; Palla, R.; Zeeuwen, P. L.; den Heijer, M.; Schalkwijk, J.; Hollox, E. J. Nucleic Acids Res. 2007, 35, e19. (2) Deutsch, S.; Choudhury, U.; Merla, G.; Howald, C.; Sylvan, A.; Antonarakis, S. E. J. Med. Genet. 2004, 41, 908–915. (3) Hollox, E. J.; Huffmeier, U.; Zeeuwen, P. L.; Palla, R.; Lascorz, J.; Rodijk-Olthuis, D.; van de Kerkhof, P. C.; Traupe, H.; de Jongh, G.; den Heijer, M.; Reis, A.; Armour, J. A.; Schalkwijk, J. Nat. Genet. 2008, 40, 23–25. (4) Wain, L. V.; Armour, J. A.; Tobin, M. D. Lancet 2009, 374, 340–350. (5) Cahill, D. P.; Kinzler, K. W.; Vogelstein, B.; Lengauer, C. Trends Cell Biol. 1999, 9, M57–60. (6) Gerami, P.; Jewell, S. S.; Morrison, L. E.; Blondin, B.; Schulz, J.; Ruffalo, T.; Matushek, P. t.; Legator, M.; Jacobson, K.; Dalton, S. R.; Charzan, S.; Kolaitis, N. A.; Guitart, J.; Lertsbarapa, T.; Boone, S.; LeBoit, P. E.; Bastian, B. C. Am. J. Surg. Pathol. 2009, 33, 1146–1156. (7) Harbour, J. W. Am. J. Ophthalmol. 2009, 148, 823–829, e821. (8) Laxman, B.; Tomlins, S. A.; Mehra, R.; Morris, D. S.; Wang, L.; Helgeson, B. E.; Shah, R. B.; Rubin, M. A.; Wei, J. T.; Chinnaiyan, A. M. Neoplasia 2006, 8, 885–888. (9) Slamon, D. J.; Godolphin, W.; Jones, L. A.; Holt, J. A.; Wong, S. G.; Keith, D. E.; Levin, W. J.; Stuart, S. G.; Udove, J.; Ullrich, A.; Press, M. F. Science 1989, 244, 707–712. (10) Furness, P. N.; Zimern, R.; Wright, C.; Adams, M. PHG Foundation and The Royal College of Pathologists, 2008 (www. phgfoundation.org/file/3998/). (11) van Beers, E. H.; Joosse, S. A.; Ligtenberg, M. J.; Fles, R.; Hogervorst, F. B.; Verhoef, S.; Nederlof, P. M. Br. J. Cancer 2006, 94, 333–337. (12) Brookfield, J. F. BMC Biol. 2010, 8, 41. (13) Blokx, W. A.; van Dijk, M. C.; Ruiter, D. J. Histopathology 2010, 56, 121–132. (14) Pinkel, D.; Albertson, D. G. Nat. Genet. 2005, 37 (Suppl), S11–17. (15) McCaughan, F.; Dear, P. H. J. Pathol. 2010, 220, 297–306. (16) van Dijk, M. C.; Rombout, P. D.; Boots-Sprenger, S. H.; Straatman, H.; Bernsen, M. R.; Ruiter, D. J.; Jeuken, J. W. Diagn. Mol. Pathol. 2005, 14, 9–16. (17) Laudadio, J.; Quigley, D. I.; Tubbs, R.; Wolff, D. J. Expert Rev. Mol. Diagn. 2007, 7, 53–64. (18) Wang, J.; Ramakrishnan, R.; Tang, Z.; Fan, W.; Kluge, A.; Dowlati, A.; Jones, R. C.; Ma, P. C. Clin. Chem. 2010, 56, 623632. (19) Weaver, S.; Dube, S.; Mir, A.; Qin, J.; Sun, G.; Ramakrishnan, R.; Jones, R. C.; Livak, K. J. Methods, 2010, 50, 271276. (20) Bastian, B. C.; Olshen, A. B.; LeBoit, P. E.; Pinkel, D. Am. J. Pathol. 2003, 163, 1765–1770. (21) Hollox, E. J.; Detering, J. C.; Dehnugara, T. Hum. Mutat. 2009, 30, 477–484. (22) Dagg, B.; Eustace, D. L.; Han, X.; Money, S.; Heyderman, E. J. Clin. Pathol. 1992, 45, 1122–1123. (23) Saldanha, G.; Purnell, D.; Fletcher, A.; Potter, L.; Gillies, A.; Pringle, J. H. Int. J. Cancer 2004, 111, 705–710. (24) Rhead, B.; Karolchik, D.; Kuhn, R. M.; Hinrichs, A. S.; Zweig, A. S.; Fujita, P. A.; Diekhans, M.; Smith, K. E.; Rosenbloom, K. R.; Raney, B. J.; Pohl, A.; Pheasant, M.; Meyer, L. R.; Learned, K.; Hsu, F.; Hillman-Jackson, J.; Harte, R. A.; Giardine, B.; Dreszer, T. R.; Clawson, H.; Barber, G. P.; Haussler, D.; Kent, W. J. Nucleic Acids Res. 2010, 38, D613–619. (25) Schwartz, S.; Kent, W. J.; Smit, A.; Zhang, Z.; Baertsch, R.; Hardison, R. C.; Haussler, D.; Miller, W. Genome Res. 2003, 13, 103–107. (26) Chiaromonte, F.; Yap, V. B.; Miller, W. Pac. Symp. Biocomput. 2002, 115–126. (27) Nature 2003, 426, 789796. 3491
dx.doi.org/10.1021/ac200153j |Anal. Chem. 2011, 83, 3484–3492
Analytical Chemistry
ARTICLE
(28) Rose-Zerilli, M. J.; Barton, S. J.; Henderson, A. J.; Shaheen, S. O.; Holloway, J. W. Clin. Chem. 2009, 55, 1680–1685. (29) Sullivan, K. M.; Mannucci, A.; Kimpton, C. P.; Gill, P. Biotechniques 1993, 15 (636638), 640–631. (30) Altman, D. G. In Practical statistics for medical research, 1st ed.; Chapman and Hall: London, 1991, pp 404. (31) Greenman, C. D.; Bignell, G.; Butler, A.; Edkins, S.; Hinton, J.; Beare, D.; Swamy, S.; Santarius, T.; Chen, L.; Widaa, S.; Futreal, P. A.; Stratton, M. R. Biostatistics 2010, 11, 164–175.
3492
dx.doi.org/10.1021/ac200153j |Anal. Chem. 2011, 83, 3484–3492
