Bmc Infectious Diseases Analysis of Multiple Single Nucleotide Polymorphisms Closely Positioned in the Ovine Prnp Gene Using Linear Fluorescent Probes and Melting Curve Analysis

Background: Resistance and susceptibility to scrapie has been associated with single nucleotide polymorphisms located within codons 136, 154 and 171 of the ovine prion protein gene (PRNP). Dual-labelled HyBeacon probes were developed to analyse single and clustered polymorphisms within these and neighbouring codons.


Background
Scrapie, one of several transmissible spongiform encephalopathies (TSEs), is a neurodegenerative and fatal disease that affects sheep and goats. Scrapie is infectious and thought to be transmissible between animals via an iso-form of the prion protein (PrP sc ) [1]. The cellular prion protein (PrP c ) is 256 amino acids in length and is encoded by the host prion protein gene (PRNP). Resistance and susceptibility to scrapie has been associated with single nucleotide polymorphisms (SNPs) affecting the amino acids specified by codons 136, 154 and 171 of the PRNP gene [2,3]. Five common alleles have been identified, commonly referred to by the single letter amino acid codes for these three codons (i.e. ARR, ARQ, VRQ, AHQ and ARH) [4]. TRQ and ARK alleles have also been described, arising from additional SNPs in codons 136 and 171 respectively [5,6]. Certain prion protein genotypes confer resistance to scrapie and selective breeding programmes have sought to increase the number of sheep that are genetically resistant and eventually eradicate the disease from sheep flocks [7] particularly via national screening programmes within Europe. In addition to classical scrapie, cases of atypical scrapie have recently been identified and linked to a polymorphism within codon 141 [8,9]. Genotypes such as ARR/ARR that are considered to be genetically resistant to classical scrapie appear to be susceptible to atypical scrapie infection [10][11][12].
Direct sequencing accurately genotypes samples with respect to SNPs in codons 136, 154 and 171 and will reliably identify any additional polymorphisms in neighbouring codons. However, genotyping technologies such as sequencing and restriction fragment length polymorphism (RFLP) analysis do not lend themselves to highthroughput analysis of large numbers of samples. Highthroughput genotyping technologies benefit significantly from automated methods involving robotics and LIMS (laboratory information management systems) to track samples throughout the process. The analysis of large numbers of ovine samples has been performed using techniques such as fluorescent primer extension [13,14], ARMS (amplification refractory mutation system) [15] and MALDI-ToF (matrix assisted laser desorption ionisation time of flight) mass spectrometry [16]. Such technologies require multiple technical steps that are both timeconsuming and costly. In contrast, homogeneous assays are performed in a single tube and typically reduce handson time and thereby the overall cost of the test. Homogeneous detection technologies frequently utilise the sequence specific hybridisation of fluorescent oligonucleotides and an associated property, such as the exonuclease dependent degradation of hydrolysis probes (also known as TaqMan), to monitor the real-time amplification of specific DNA sequences. Variations in probe hybridisation stability are exploited to differentiate closely related target sequences, such as PRNP polymorphisms [17].
Additional nucleotide polymorphisms have been identified in codons neighbouring 136, 154 and 171, with sheep possessing alleles such as AT 137 RQ, AR 143 RQ, AC 151 RQ, ARL 168 Q and ARQE 175 [18]. Whilst these haplotypes are relatively rare, the additional polymorphisms may compromise PRNP genotyping using primer extension and real-time PCR technologies since they will desta-bilise the hybridisation process and potentially cause miscalls.
We have developed homogeneous high-throughput diagnostic tests, employing the fluorescent HyBeacon ® probe technology [19,20], to reliably genotype sheep with respect to polymorphisms within codons 136, 154 and 171 polymorphisms in a background of neighbouring and potentially interfering SNPs and mutations. Analysis of the leucine/phenylalanine polymorphism within codon 141 [9] is also described.
Probes are included in PCR assays and emit greater amounts of fluorescence when hybridised to complementary target sequences than when single-stranded. Following amplification, the presence and identity of target sequences is determined by melting curve analysis. Duallabelled probes, possessing two fluorescent dye labels attached to internal nucleotides, exhibit considerably larger signal-to-noise ratios and melting peaks compared with single-labelled probes of identical sequence [21]. The stability and melting temperature of hybridised probes depends on the degree of homology between HyBeacons and their target sequences. Increasing the reaction temperature above the melting temperature (Tm) of the HyBeacon causes probe/target duplexes to dissociate and the amount of fluorescence emission to decrease. Sequences differing by as little as a single nucleotide may be detected and differentiated on the basis of melting peak Tm. Furthermore, multiple polymorphisms may be analysed using a single probe, enabling the presence of additional polymorphisms flanking PRNP codons 136, 141, 154 and 171 to be detected through distinct shifts in probe Tm. The analysis of extracted DNA and unpurified sheep blood samples is described here.

HyBeacon probes
The probes employed for the analysis of PRNP polymorphisms were synthesised by ATDBio Ltd (Southampton, UK) and were each labelled using two fluorescein dT monomers (Glen Research, Virginia, USA). Polymorphisms within codons 136, 154, 171 and 141 were investigated using the dual-labelled HyBeacons detailed in table 1.

DNA extraction from sheep blood samples
8918 purified DNA samples were analysed using the HyBeacon method, comparing 136, 154 and 171 codon data with homogeneous MassEXTEND™ (hME) and MALDI-ToF (Sequenom, Hamburg, Germany) results obtained from the same sample set. The 141C3 probe was evaluated using a further 400 DNA samples, again comparing genotype calls with MALDI-ToF data. EDTA-blood samples were aspirated from sealed vacutainers (BD, Oxford, UK), transferring 50μl to 96 square well blocks (QIAGEN, Crawley, UK) using a Tecan Genesis RSP 150 robot (Tecan, Berkshire, UK). Blood samples were transferred to 91 plate wells, leaving 5 wells available for positive and negative controls. DNA extraction was performed using a Chemagen Chemagic magnetic separation module I robot and Chemagic DNA Blood100 kits (Auto Q Biosciences, Berkshire, UK), eluting DNA into 250μl of tissue culture water (Sigma Aldrich, UK) in 96 low-well plates.

Analysis of unpurified blood samples
3663 unpurified blood samples were analysed using the HyBeacon methodology, comparing genotype calls with hME and MALDI-ToF data obtained from extracted DNA samples. 9μl of each blood sample was diluted in 710μl of tissue culture water using 96 deep well plates and a Tecan Genesis RSP 150 robot. Blood samples were transferred to 91 plate wells, leaving 5 wells available for positive and negative controls. Samples were mixed for 10 minutes using a Grant-bio PMS-1000 microplate shaker (Fisher Scientific, Leicestershire, UK).

Polymerase Chain Reaction
PCR reaction volumes were 5μl, containing 1μl of purified DNA or diluted blood sample, 1× QIAGEN PCR buffer, 0.05μM forward primer, 0.5μM reverse primer (table 1), 0.25 units HotStarTaq polymerase (QIAGEN, Crawley, UK), 3 mM total MgCl 2 , 1 mM dNTPs (GE Healthcare, Amersham, UK) and 100 nM of HyBeacon probe. Asymmetric PCR was employed to generate an excess of the target strand such that probe hybridisation was favoured over annealing of amplified sequences. Four 96 well plates of extracted DNA or diluted bloods were analysed simultaneously. 500× PCR mastermixes were prepared such that sufficient reagents were available for robotic preparation of 384 well PCR plates. Amplification of target sequences was performed using a 384 well Tetrad thermal cycler (MJ Research). The 136F2 and 171R2 primers detailed in table 1 were employed to amplify a PRNP target sequence of 198 bp, comprising codons 119 through to 185. Following an initial denaturation to activate the hotstart enzyme (95°C 15 minutes), targets were amplified using 35 cycles (or 40 cycles for diluted blood samples) comprising denaturation (95°C 15 seconds), primer annealing (55°C 30 seconds) and extension of products (72°C 30 seconds). Following amplification, reactions were incubated at 72°C for 2 minutes prior to a denaturation (95°C 1 minute) and cool (20°C 10 minutes).

Melting curve analysis
PCR plates were heated from 35°C to 75°C, at 0.1°C per second, using a LightTyper instrument (Roche Diagnostics, Lewes, UK). Fluorescence emission was monitored continuously to measure the stability of probe hybridisation and determine the identity of amplified target sequences. Melting peaks were constructed by plotting the negative derivative of fluorescence with respect to temperature (-d(Fluorescence)/dT) on the y-axis) against temperature (x-axis). Samples were genotyped with the LightTyper software version 1.5 using positive control samples to define peak standards. Peak Tm and area data was exported from LightTyper software version 1.1 to generate an independent genotype call.

Determination of PRNP genotype using HyBeacon melting peaks
Samples were genotyped using a combination of LightTyper software calls and exported peak data. The LightTyper software typically assigned genotypes to more than 95% of samples based on a set of defined standards employing reference control samples. Less than 5% of samples, for which the genotype was obvious by eye, required manual analysis when the LightTyper software assigned weak positive or unknown calls due to minor variations of melt curves from defined standards. A second, independent genotype call was obtained by exporting peak data into a Microsoft Excel spreadsheet that employed a combination of peak Tms and areas to genotype samples. Peak area ratios were employed to identify heterozygous samples that potentially possessed more than two copies of the PRNP gene [22,23]. A sample genotype was only reported if LightTyper/manual calls were concordant with exported spreadsheet data. Genotypes and rare sequence variants are presented using the notation described by Goldmann et al. [18].
Genotypes generated in HyBeacon tests were compared with hME and MALDI-TOF analysis performed in parallel. The hME assay employs MassEXTEND primers that anneal to target sequences adjacent to the polymorphism of interest. DNA polymerase is employed with a mixture of nucleotides and terminators to extend from the primer and through the polymorphic site. The resultant mass of the primer extension product is determined by mass spectrometry and is used to determine the sequence of nucleotides at the polymorphic site [16].

Melting curve analysis with complementary oligonucleotides
The affect of rare polymorphisms on probe Tm was predicted using complementary oligonucleotides. 100 nM of each probe was hybridised to 200 nM of oligonucleotide homologue in QIAGEN PCR buffer and a total of 3 mM MgCl 2 . Microseal white 384 well PCR plates (Bio-Rad) were heated to 95°C for 1 minute and cooled to 20°C for 1 minute, prior to melting curve analysis using a LightTyper instrument, heating samples from 35°C to 75°C, at 0.1°C per second. Peak Tm data was determined using LightTyper software version 1.1.

Results
Amplified targets were reliably detected and identified using the  1B). This T allele has not been reported in studies of UK flocks [18], but has been detected in other countries [6]. Identification of the  [18]. The E 175 allele reduced the Tm of the Q 171 peak such that it fell within the acceptance range of the H 171 peak. To prevent such miscalls during analysis, the polymorphism within codon 175 was neutralised using the universal base 5-nitroindole, which is a synthetic nucleotide analogue that hybridises equally with the four naturally occurring bases [24]. The 171R6 HyBeacon is unaffected by the E 175 SNP (

Imbalanced melting peaks
Each parent sheep typically contributes a single PRNP allele to its progeny, such that maternally and paternally The mean Tms for each probe melting peak are presented, where N is the number of peaks investigated and SD is the standard deviation. The peak Tms for the major PRNP alleles were obtained from six 384 well plates prepared on different working days, thereby accounting for sample-tosample and run-to-run sources of variation. Variant alleles exhibiting uncommon genotypes or additional SNPs are italicised. Acceptance criteria for these rare polymorphisms were estimated from the few samples analysed, minimising the potential for miscalling melting peaks.
inherited alleles exist in a 1:1 ratio in 'normal' genotypes. However, a small subset of sheep have been found to exhibit abnormal (also known as complex) genotypes, where the two alleles of heterozygous samples appear to be present in unequal ratios [22,23]. Fluorescent primer extension, hME/MALDI-ToF and HyBeacon analyses have all yielded imbalanced peak data when analysing blood samples. The generation of imbalanced peak data may derive from a recognised genetic phenomenon known as chimerism or could be caused by the presence of additional polymorphisms in target sequences. Analyses employing alternate sample types, such as hair roots, ear tissue plugs and saliva, are expected to resolve a large proportion of the abnormal results obtained from blood.
Samples that yielded imbalanced melting peak data were called "Unknown" by the LightTyper software. Melting peaks were also determined to be imbalanced by measuring the ratio between mismatched and matched peak areas. Acceptance criteria for normal, balanced, heterozygotes were determined using the area ratios obtained from 4560 purified DNA samples (i.e. twelve 384 well plates, excluding negative controls). Acceptance criteria for peak area ratios ( The estimated frequency of animals appearing to possess more than two copies of the PRNP gene is approximately 0.001 of the UK flock when analysing blood samples [22,23]. Our study of 8918 purified DNA samples was enriched for samples that had previously generated imbalanced peak data in MALDI-ToF analyses. Whilst 72 samples exhibited imbalanced MALDI peaks, only 43 heterozygous samples yielded imbalanced melting peaks.

Assay performance
The performance of HyBeacon PRNP assays was assessed through determination of first time pass rates (FTPR) and concordance with parallel homogeneous MassEXTEND™ and MALDI-ToF analysis. Samples that contributed to the FTPR generated high quality peak data in 136, 154 and 171 assays. Samples were treated as "fails" if melting peaks were absent from any one of the tests or if software/ analyst genotype calls were discordant with exported spreadsheet data. Samples were also excluded from FTPR  [22].

Direct blood analysis
HyBeacon assays may also be performed directly using unpurified sheep blood samples. Blood samples only need to be diluted in water to reduce PCR inhibition and lighten sample colour enabling fluorescence detection. Five additional PCR cycles were required to analyse unpurified blood samples due to the low DNA concentrations once diluted. High quality melting peaks were obtained in 136, 154 and 171 assays, enabling accurate determination of sample genotypes. Analyses employing unpurified blood samples were extremely efficient yielding high quality melting peaks with the majority of samples, demonstrating first time pass rates between 85% and 95% (mean = 91.2%). Pass rates using unpurified blood were considerably more variable (standard deviation = 3.95) than PCR plates employing purified DNA samples. HyBeacon and MALDI-ToF data was 99.9% concordant, where the 4 DNA samples that yielded incongruent calls possessed leucine at codon 168. Blood samples yielded elevated numbers of weak positives with melting peaks of reduced height. Our study of 3663 unpurified blood samples was not enriched for samples that had previously generated imbalanced peak data in MALDI-ToF analyses. Whilst 7 samples yielded imbalanced MALDI peaks (with extracted DNA), 153 heterozygous samples exhibited imbalanced melting peaks in at least one assay.

Predicting the affect of rare sequence variants
In addition to the T 136 , T 137 , L 168 , K 171 and E 175 polymorphisms described above, additional published polymorphisms in codons 138, 151, 154, 170 and 172 are covered by the diagnostic probes [14,18]. The affect of glycine at codon 170 (D170G; GAT>GGT) was investigated using two DNA samples previously genotyped as ARR/AHG 170 Q using a combination of hME, MALDI-ToF and sequencing analysis. The G 170 polymorphism reduced the Tm of the 171R6 melting peak to 40.0°C and is not expected to compromise the accuracy of PRNP genotypes. In the absence of suitable DNA samples, the affects of other rare polymorphisms were predicted by hybridisation of probes to complementary oligonucleotide sequences (table 4). Whilst melting peak Tms obtained with oligonucleotide and PCR targets are expected to vary (e.g. due to length and concentration differences), probe hybridisation to 200 nM of oligonucleotide homologue yielded melting temperatures that were comparable with PCR amplified target sequences.

Codon 141
A polymorphism within codon 141 has been linked to atypical scrapie [8,9,18]. Codon 141 falls outside the range of 136, 154 and 171 HyBeacons, so an additional probe was designed to analyse this common polymorphism. The C to T polymorphism of codon 141 (Leu [L]>Phe [F]; CTT>TTT) was analysed using the HyBeacon 141C3 (table 1), which possesses a C nucleotide at the polymorphic site. Melting curve studies with complementary oligonucleotides predicted that asparagine at codon 138 would be reliably detected along with L 141 and F 141 in all homozygous and heterozygous allele combinations (table 4). In contrast the R 143 polymorphism was predicted to cause L 141 R 143 alleles to be miscalled as F 141 . The R 143 polymorphism was therefore neutralised using the universal base 5-nitroindole as described above. Analysis of codon 141 was performed using the same primers employed for 136, 154 and 171 assays. The 141 HyBeacon generated a melting peak possessing a Tm of approximately 54°C when hybridised to the fully complementary leucine target sequence and exhibited a C/A mismatch and a Tm of approximately 44.5°C when hybridised to the phenylalanine target sequence. Samples that were heterozygous for L 141 and F 141 generated both 54°C and 44.5°C peaks. Sample genotypes were unaffected by the presence of arginine at codon 143, with HyBeacon and MALDI-ToF results in complete concordance.

Discussion
The accuracy, reproducibility and robustness of HyBeacon assays was evaluated using in excess of 8900 purified DNA samples, comparing genotype calls with parallel homogeneous MassEXTEND™ and MALDI-ToF analysis. Samples were analysed in 384 well plates using a LightTyper instrument, achieving an average first time pass rate (FTPR) of 97.8%. HyBeacon and MALDI-ToF genotype calls were 99.8% concordant, where all discordant samples were sequenced and confirmed to possess leucine at codon 168. Fewer samples yielded imbalanced peak data in HyBeacon tests compared with hME and MALDI-ToF analysis of the same sample set, thereby reducing the number of samples requiring reanalysis.
Genotyping technologies that employ primer extension events to characterise downstream SNPs (e.g. SNaPshot, Applied Biosystems or hME, Sequenom) can be affected by the presence of additional polymorphisms within the primer sequences. Additional polymorphisms neighbouring the SNPs of interest can considerably reduce the stabil-ity of genotyping primers and affect the efficiency of extension, resulting in imbalanced peaks or allele dropout and potentially incorrect calls. The presence of leucine at codon 168 reduces the efficiency of primer extension with L 168 Q alleles in Sequenom hME assays, creating MALDI peak imbalances in ARR/ARL 168 Q and ARL 168 Q/ ARH heterozygotes. L 168 is detected using HyBeacons as a decrease in Tm, such that Q and L 168 Q targets may be reliably differentiated.
Alternate fluorescent probe technologies (such as hydrolysis probes), which rely on real-time fluorescence increases to detect and differentiate polymorphic sequences, will also be affected by the presence of additional SNPs in target sequences. Such real-time PCR technologies typically employ a separate probe for each sequence variant and would require seven probes just to analyse the common 136, 154 and 171 alleles [25]. The presence of additional SNPs within target sequences would decrease the stability of real-time probes, reducing or preventing the generation of fluorescent signal and potentially generating incorrect genotypes. The stability of HyBeacon hybridisation is also affected by the presence of rare polymorphisms. However, unlike the genotyping technologies described above, these additional SNPs are detected through decreases in peak Tm, generating novel melting peaks and correct calls. Post-amplification melting curve analysis of PRNP polymorphisms has also been presented for HybProbes that employ fluorescence resonance energy transfer (FRET) [26,27].
HyBeacon assays were able to detect the additional polymorphisms T 136 , T 137 L 168 and K 171 through distinct shifts in melting peak Tm. The rare polymorphisms R 138 , N 138 , C 151 , H 151 , G 151 , L 154 and D 172 [14,18], were not encountered during validation of the HyBeacon assays, but their effects were predicted using complementary oligonucleotide targets (  [18]. Further unpublished sequence polymorphisms could also give rise to incorrect genotype calls if they result in melting peaks that coincide with the expected Tms detailed in table 2. Several of the other reported polymorphisms, such as S 167 and K 176 fall outside the HyBeacon binding sites and have no effect on peak Tms. All genotyping technologies have the potential to miscall PRNP genotypes when certain rare sequence variants are present in primer/probe sequences. These miscalls may remain undetected unless all samples are sequenced or genotypes compared with an independent genotyping method. Undesirable polymorphisms, which have the potential to generate incorrect calls, may be neutralised using universal nucleosides such as inosine, 3-nitropyrrole and 5-nitroindole [24], as described here with R 143 and E 175 . Alternatively, degenerative bases such as dP and dK phosphoramidites may be introduced into oligonucleotides to base pair with A and G or with C and T respectively (Glen Research, Virginia, USA). A more cost effective method to "mask" additional polymorphisms might be to design probe sequences such that they comprise a mismatched nucleotide at the polymorphic site when hybridised to all possible target alleles [28]. The Tm for each sequence variant should ideally differ by less than 1°C to ensure detection of further polymorphisms that may flank the masked position.
The 136, 141, 154 and 171 assays were performed in separate PCR plates, each containing a single HyBeacon probe. The HyBeacon assays were not combined into multiplexes to ensure that additional SNPs and imbalanced melting peaks were reliably detected. The bottle-neck for analysis within our laboratory was the number of PCR plates that could be prepared per day. However, with two Hamilton LabSTAR robots in operation up to 1440 samples could be analysed each day using 136, 154 and 171 assays. Furthermore, since each of the HyBeacon tests employs the same primer pair, a single PCR plate can be prepared for each batch of samples, without inclusion of probes or Chill-Out liquid wax. Daughter plates prepared robotically post-amplification will mix a portion of the target amplicon with a specific HyBeacon probe. This approach considerably increases the number of samples that may be analysed each day. Melting curve analysis using a LightCycler 480 instrument (Roche Diagnostics) or similar real time PCR system would negate the requirement for Chill-Out liquid wax. The LightCycler 480 instrument will also permit colour-multiplexing with 384 well plates, enabling multiple probes labelled with spectrally distinct dyes to be employed simultaneously in each well [19,27]. HyBeacon assays may be performed in both high-throughput and ultra-rapid formats, where assays performed in glass capillaries using a LightCycler instrument are completed within 30 minutes.
The accuracy, reproducibility and robustness of blood assays was evaluated using in excess of 3600 unpurified samples. Whilst the average pass rate was in excess of 90%, the increased analysis times and repeat rates required for direct blood analyses negate much of the cost savings achieved through omitting DNA extraction. Further development is required to enable unpurified blood samples to be analysed in a high-throughput scenario. However, "pen-side" tests could be performed using unpurified blood samples and a portable PCR instrument.

Conclusion
The fluorescent probe assays described here provide an extremely accurate and robust method for the analysis of single and clustered PRNP gene polymorphisms. Target sequences were detected through the generation of melting peaks and differentiated on the basis of melting temperature. Genotyping assays were successfully validated through comparison with a MALDI-ToF method and DNA sequencing. The flexibility of the diagnostic tests enabled samples to be correctly genotyped in the presence of additional sequence variations flanking the polymorphisms of interest.