The Power of Proficiency Testing: Unraveling Single-Nucleotide Polymorphism Interference, With Potential Impact on Clinical Testing of Spinocerebellar Ataxia Type 3

The challenge samples that made up the MGL2 Survey consisted of purified DNA obtained from established cell lines from the biorepository of the Coriell Institute for Medical Research (Camden, New Jersey). Selected cell lines have had their disease-specific genotype verified by the Centers for Disease Control and Prevention Get-RM program or by CAP/ACMG Committee member laboratories. A total of 32 laboratories participated in the 2015 MGL2 Survey. Each laboratory tested 3 provided samples for SCAs 1, 2, 3, 6, and 7, entered the observed genotype, and selected the appropriate interpretation from a short list of provided responses. Results were reported as the number of CAG repeats identified for alleles 1 and 2 for each of the 5 SCA genes for the 3 provided samples.

Results of the 2015 MGL2 Survey showed a discrepancy in the test results for SCA3. We therefore attempted to characterize this discrepancy. To identify the Coriell cell line selected as the 2015 MGL2-09 spinocerebellar challenge sample, the records of the CAP/ACMG Committee were searched. It was determined that this sample corresponded to Coriell sample GM14982. Available sample demographic data were collected, and an aliquot was obtained for additional experiments performed at Mayo Clinic, Rochester, Minnesota, to investigate the observed discordances for SCA3 test results.

The presence of a potential SNP within a primer-binding site, which may lead to allele dropout, was assessed by using the SNPCheck v3.1 (NGRL Manchester, Manchester, United Kingdom) and Alamut visual, version 2.7 rev. 2 (December 2015; Interactive Biosoftware, Rouen, France) programs. These programs were used to check the forward (5′-CCAGTGACTACTTTGATTCG-3′) and reverse (5′-GGCCTTTCACATGGATGTGAA-3′) primer sequences used to amplify the polyglutamine region in exon 10 of the ATXN3 gene (GenBank accession No. NM_004993.5), as originally described in the seminal study that identified the CAG repeat expansion in this gene as the underlying molecular basis of SCA3.⁶  The forward and reverse primers of this set are referred to as published forward primer and published reverse primer, respectively.

Targeted direct Sanger sequencing was performed by using polymerase chain reaction (PCR) primers with 5′ universal sequencing primer (USP) tails⁷  designed to flank the published forward and reverse primer sequences (forward: 5′-USP-TGTGATGAAAATACCTACCT-3′; reverse: 5′-USP-CCAGGGAAATTTAGTAGATTAC-3′). Template PCR was performed with the KAPA2G PCR system (KAPA Biosystems, Wilmington, Massachusetts) in a 10-μL reaction. The standard KAPA2G PCR setup was used, with the addition of 5% dimethyl sulfoxide and 2 μL of the KAPA Enhancer I solution. The reaction was amplified with a 2-step thermal cycler program. The first step consisted of a stepdown program that decreased the annealing temperature from 68°C to 58°C during 15 cycles. The second step included a 20-cycle reaction with an annealing temperature of 58°C. The resulting PCR product was purified by using Agencourt AMPure XP DNA purification reagent (Beckman Coulter Life Sciences, Indianapolis, Indiana) to remove unincorporated primers and nucleotides from the reaction. The PCR amplicons were sequenced bidirectionally, with universal forward and reverse sequencing primers and the BigDye Terminator v1.1 cycle sequencing kit (Applied Biosystems, Thermo Fisher Scientific, Foster City, California). Sequencing reaction products were purified with the Agencourt CleanSEQ purification reagent (Beckman Coulter Life Sciences). Diluted purified sequencing products were detected on the ABI 3730xl DNA Analyzer (Thermo Fisher Scientific) capillary electrophoresis instrument. Sequencing traces were analyzed using the Mutation Surveyor software (SoftGenetics LLC, State College, Pennsylvania). The CAG repeat number included CAG variants CAA and AAG, Chr14 (GRCh37):g.92537385-92, as originally described.⁶ 

A genotype PCR assay was used to determine the SCA3 genotype status of the MGL2-09 sample and to investigate the effect of the rs372907618 SNP in the MGL2-09 sample. A new forward primer of identical sequence to the published forward primer was designed; however, it was truncated by the base at the 3′ end that corresponded to the rs372907618 SNP. The MGL2-09 sample was tested by using either the new or the published forward primer in conjunction with the published reverse primer. A 2-μL aliquot of the MGL2-09 sample DNA was amplified with a standard 25-μL PCR reaction containing 1.5 mM MgCl₂, 5 μL of a 5 M Betaine (Sigma-Aldrich Corp, St Louis, Missouri) solution, 80 pmol of 1:1 forward and reverse primer mix, and 1.25 units of AmpliTaq DNA Polymerase (Thermo Fisher Scientific). The PCR sample mix was predenatured at 98°C for 2 minutes, followed by 30 cycles of a denaturation step at 95°C for 30 seconds, an annealing temperature of 55°C for 30 seconds, and a 30-second extension step at 72°C, followed by a final extension step at 72°C for 10 minutes. Diluted product mixes (1:10) were separated and visualized by capillary electrophoresis on the ABI 3730xl using GS 500 ROX dye size standard (Thermo Fisher Scientific). Data analysis was performed by using a GeneMarker (SoftGenetics LLC) software analysis panel with predefined allele bins.

Subsequently, a genotype PCR assay for the MGL2-09 sample was performed by using 9 additional thermostable DNA polymerases either with 3′-5′ exonuclease activity (“proofreading”) or without 3′-5′ exonuclease activity (“nonproofreading”), including KAPA2G Robust HotStart (KAPA Biosystems), Klentaq1 (DNA Polymerase Technology, St Louis, Missouri), AmpliTaq Gold (Thermo Fisher Scientific), Platinum Taq (Thermo Fisher Scientific), FailSafe PCR Enzyme Blend (Lucigen Corp, Middleton, Wisconsin), TaKaRa LA Taq (Takara Bio USA, Mountain View, California), Platinum Taq High Fidelity (Thermo Fisher Scientific), Q5 High-Fidelity (New England Biolabs, Ipswich, Massachusetts), and Deep Vent DNA Polymerase (New England Biolabs). The published and new forward primers were used in conjunction with the published reverse primer. The PCR reactions for each enzyme and thermal cycling conditions were performed as recommended by each manufacturer.

On the subsequent 2016 MGL2 Survey, 2 supplemental questions regarding test methods for SCA3 were added to the SCA challenge to assess the relevance of the experimental findings reported here. These questions had yes/no answers, and were as follows:

1. For your SCA3 assay, do you use the PCR primers described in the paper that first reported the identification of the SCA3 (Machado-Joseph) gene?⁶  (Primer sequences were given in a footnote.)

2. For your SCA3 assay, do you use a thermostable polymerase that has 3′ to 5′ exonuclease activity? (Note: These enzymes are often sold as “high-fidelity polymerases.”)

Of the three 2015 MGL2 challenge samples used to assess the performance of SCA clinical testing, 2 were known to be positive: MGL2-07 had a pathogenic expansion in the ATXN1 gene (SCA1), and MGL2-09 had a pathogenic expansion in the ATXN2 gene (SCA2); both samples were negative for the remaining SCAs tested. The MGL2-08 challenge sample was negative, with no CAG expansion for any of the 5 tested SCA genes. Overall, testing of all 3 samples was considered satisfactory by the CAP/ACMG Biochemical and Molecular Genetics Committee.

Testing of the MGL2-09 sample for SCA2 showed consensus—31 of 32 laboratories reported an expanded allele. There was also consensus in normal results for SCA1, SCA6, and SCA7. In contrast, the reported results for SCA3 testing were inconsistent. Of the 32 participating laboratories, 18 reported this sample as homozygous for a normal allele, with a mean of 27 repeats (range, 26–30 repeats); 10 laboratories reported the sample as heterozygous for 2 differently sized normal alleles, with mean repeat numbers of 13.7 (range, 12–15) and 26 (range, 23–27); 3 laboratories did not report specific allele sizes but rather that the results were normal; and 1 laboratory did not provide a result for SCA3 (Table 1). Of the 28 laboratories that provided numerical results for the SCA3 repeat size, 18 (64%) reported homozygous results, and 10 (36%) reported heterozygous results. There was no such discordance for SCA3 testing of the other 2 challenge samples.

One hypothetical explanation for such a discrepancy in the SCA3 testing results was possible SNP interference at a primer-binding site resulting in allele dropout and an apparent homozygous result. Because some of the participating laboratories may use the published primers, these primers were evaluated for a potential interfering SNP. Three SNPs were identified within the intronic region complementary to the published forward primer (rs372907618, rs746454122, and rs772351920). The rs372907618 SNP (Chr 14 (GRCh37):g.92537495 C>T; NM_004993.5:c.873-98G>A) is located at the 3′ end of the published forward primer-binding site, a location that has been shown to be important for primer binding and extension. This SNP has been reported in the human population at an overall minor allele frequency of 1.9% in the Exome Aggregation Consortium (ExAC) data and browser (beta)⁸  and a minor allele frequency of 3.7% in South Asian populations. Of note, the ethnicity for Coriell cell line GM14982 (MGL2-09 challenge sample) is listed as East Indian, a South Asian subpopulation. The rs746454122 and rs772351920 SNPs map to the 3′ third portion and to the penultimate 3′ base of the published forward primer, respectively. Additionally, both have very low minor allele frequency; rs746454122 had been described twice in a total of 57 040 alleles and rs772351920 only once in 61 868 alleles of the chromosome-aggregated ExAC database (dbSNP; accessed May 25, 2017). Four additional potentially interfering SNPs (rs779933798, rs757359624, rs751821844, and rs778200373) were identified in the exonic region covered by the published reverse primer. The rs779933798, rs751821844, and rs778200373 SNPs are missense variants, and the rs757359624 SNP is a synonymous variant. The minor allele frequency of these SNPs is extremely low; each SNP had been observed only once in the 121 412 alleles of the chromosome-aggregated ExAC database.⁸ 

Sequencing of the flanking region encompassing the published forward and reverse primer-binding sites revealed that the MGL2-09 challenge sample (GM14982 cell line) was indeed heterozygous for the rs372907618 SNP located at the 3′ end of the published forward primer (Figure 1). The 4 SNPs reported in the binding region of the published reverse primer sequence were absent in the MGL2-09 sample. Of note, an additional SNP, rs7158733 (Chr 14 (GRCh37):g.92537223G>T), located 90 bases downstream of the published reverse primer-binding region, was also identified in this sample (not shown). Additionally, it could be inferred from the sequencing traces that the shorter allele contains 14 CAG repeats and that the larger allele has 27 CAG repeats, which is consistent with the Survey responses for the heterozygous results.

The genotype PCR assay using the published primers demonstrated that the presence of the rs372907618 SNP in the MGL2-09 sample was associated with decreased PCR amplification efficiency (lower signal intensity) but not complete allele dropout of the shorter 14-repeat allele (Figure 2, A). This finding indicates that the shorter allele of the MGL2-09 sample harbors the rs372907618 SNP. This is consistent with the hypothesis that laboratories using this published forward primer may have experienced allele dropout of the 14-repeat allele and reported homozygosity for the larger allele. The set of primers including the new forward primer, in which the most 3′ base of the published forward primer (complementary to rs372907618) was trimmed, yielded a heterozygous result with higher signal intensity for the shorter allele, which is the typical result for heterozygous alleles because of increased PCR amplification efficiency of the shorter alleles (Figure 2, B).

Confirming the presence of the heterozygous rs372907618 SNP in the MGL2-09 sample and noting that it was at the binding site of the 3′ end of the published forward primer led us to suspect that many of the participating laboratories could be using the published primers, which therefore could have been subject to interference by the rs372907618 SNP with allele dropout. We further hypothesized that some laboratories that may also be using the published primers may have been spared the allele dropout by using thermostable DNA polymerases with 3′-5′ exonuclease activity (so-called proofreading or high-fidelity polymerases) due to excision of the mismatched 3′ base in the forward primer. This would thereby result in robust amplification of that allele despite the presence of an interfering SNP. We addressed the first hypothesis with supplemental question 1 on the 2016 MGL2 Survey. The second hypothesis was addressed in 2 ways: (1) with supplemental question 2 on the 2016 MGL2 Survey; and (2) experimentally by testing several proofreading and nonproofreading enzymes.

We evaluated the performance of 9 other thermostable DNA polymerases in addition to AmpliTaq, which totaled 5 enzymes with and 5 without 3′-5′ exonuclease activity, by using both the published set of primers and the set with the new forward primer (Table 2). A total of 4 of the 10 enzymes (2 each with and without 3′-5′ exonuclease activity) yielded a homozygous result (ie, with allele dropout), and 6 (3 each with and without 3′-5′ exonuclease activity) resulted in a heterozygous pattern for the published forward primer. The use of the new forward primer consistently resulted in a heterozygous pattern for all tested enzymes regardless of the 3′-5′ exonuclease activity. Of note, the heterozygous results observed with the published forward primer showed decreased amplification efficiency (decreased signal intensity) of the shorter allele compared with the heterozygous results obtained with the use of a new forward primer for most of the enzymes (Platinum Taq and Deep Vent, enzymes with and without 3′-5′ exonuclease activity, respectively, being the exception; Table 2 and Supplemental Figures 1 through 9 [see supplemental digital content at www.archivesofpathology.org in the March 2019 table of contents]).

The response rate for the 2 supplemental questions to the 2016 MGL2 Survey was high (25 of 27 laboratories [93%]; Table 1). Of the 25 laboratories responding, 23 had participated in the 2015 MGL2 SCA challenge; 15 had reported a homozygous result, 7 had reported a heterozygous result, and 1 had not reported a numeric result (result reported as normal). For the first question (use of published primers), only 3 laboratories (all located outside the United States) disclosed current use of the published primers, and all 3 had reported a homozygous result. The other 12 laboratories that reported a homozygous result indicated the use of primers other than the published primers, as did all 7 laboratories that reported a heterozygous result and the laboratory that did not report a numeric result. For the second question (use of proofreading enzymes), 18 laboratories replied that they did not use a DNA polymerase with 3′-5′ exonuclease activity; 11 of these had reported homozygous results, 6 had reported heterozygous results, and 1 did not report a numeric result (only normal). Of the 5 laboratories that reported the use of proofreading enzymes, 4 had reported homozygous results, and 1 had reported heterozygous results. Two of the 3 laboratories that had used the published primers reported use of proofreading enzymes. The 2 (of 25) laboratories that responded to the supplemental questions but did not participate in the 2015 Survey reported the use of primers other than the published primers, and 1 of each disclosed the use of enzymes with and without 3′-5′ exonuclease proofreading activity.

Polymerase chain reaction–based mutation detection methods have revolutionized many fields of medicine, perhaps none more so than medical genetics. Despite the power of PCR, its methods are subject to certain caveats, one of the most important being the possibility of allele dropout due to an unsuspected sequence variant that interferes with the hybridization of one of the amplification primers (or detection probes).⁹  Here, we report a case of allele dropout that was detected via the CAP proficiency program. Of the 28 laboratories that provided a numeric result for SCA3 testing of the 2015 MGL2-09 sample, 18 (64%) reported a homozygous result (ie, with allele dropout), whereas only 10 (36%) reported the expected heterozygous result.

To understand the reason for the discrepancy and to alert the community about potential allele dropout interference due to a primer-binding site SNP in SCA3 clinical testing, we followed up this observation by identifying the Coriell cell line that was used in this SCA challenge and by sequencing the region surrounding the CAG repeat region in ATXN3 exon 10. Two heterozygous SNPs that are frequently observed in South Asian populations were identified in Coriell cell line GM14982, which is indeed derived from a person of South Asian descent. The rs7158733 SNP is located 90 bases downstream of the published reverse primer-binding region and, because it has been recognized since 2003 and is not located in the primer-binding site, was not believed to be a good candidate for the observed assay interference. The rs372907618 SNP was of more interest because it occurred at the binding site complementary to the 3′ end position of the published forward primer, a location in which mismatches have been shown to critically decrease PCR amplification efficiency.^10–13  In fact, we demonstrated that allele dropout can occur. A total of 3 of the 18 laboratories that had reported a homozygous result disclosed in the answer to supplemental question 1 that they used the published primers, which indicated that allele dropout in their assays was most likely a result of SNP interference in the primer-binding site. For the other 15 laboratories that reported a homozygous result, 12 reported the use of alternative primers and 3 gave no additional information. Because of this evidence that the rs372907618 SNP interferes with the published forward primer, laboratories are encouraged to also determine whether one of their alternative PCR primer-binding sites overlaps with this SNP.

Before receiving the responses to the supplemental questions, we hypothesized that many laboratories may have used the published or similar primers, which would explain the observed allele dropout. We also hypothesized that the rs372907618 SNP may not have interfered with the PCR amplification of the mismatched primer if the laboratory used a thermostable DNA polymerase that possessed 3′ to 5′ exonuclease activity. In this case, the 3′-end mismatched base of the published forward primer would be removed, thus leaving a perfectly matched primer, which would yield robust amplification and heterozygous results even in the presence of the rs372907618 SNP. Although our results did not segregate as expected according to our hypothesis (ie, with proofreading enzymes giving heterozygous results and nonproofreading enzymes being uniformly susceptible to allele dropout and yielding homozygous results), it is noteworthy that thermostable DNA polymerases with different 3′-5′ exonuclease proofreading activities did, in fact, affect the ability of the published forward primer to amplify the mismatched allele with the rs372907618 SNP. Several DNA polymerases gave robust amplification of the allele with the rs372907618 SNP (ie, heterozygous results), although usually with decreased efficiency. Given that only 3 laboratories used the published primers, it is not surprising that supplemental question 2 did not shed more light on the reason for the discordant results. Nevertheless, it is interesting that users of both proofreading and nonproofreading enzymes reported both heterozygous and homozygous results for the 2015 MGL2-09 sample. Thus, some enzymes with proofreading activity are able to amplify the mismatched allele and some are not; the same holds true for enzymes that lack proofreading 3′-5′ exonuclease activity.

Other factors also can influence the amplification of mismatched primers, such as reaction components and PCR cycling conditions,¹⁴  which were not systematically controlled in our experiments. It is clear, however, that the choice of thermostable DNA polymerase can determine whether complete allele dropout is observed in the case of a 3′-end mismatched primer, which indicates that the DNA polymerase exonuclease activity also influences allele dropout due to SNP interference. Interestingly, of the 6 laboratories that disclosed the use of a proofreading enzyme, 4 reported a homozygous result for the 2015 MGL2-09 sample, and only 1 reported a heterozygous result (1 laboratory did not participate in the 2015 challenge).

Clearly, the use of the published primers was not the only cause of allele dropout in the analysis of the 2015 MGL2-09 sample, although other laboratories may be using similar primers whose amplification efficiency is affected by the presence of this SNP. We hypothesize that the rs372907618 SNP is the most likely culprit for the observed allele dropout, even with redesigned forward primers. Alternatively, although we believed that the heterozygous rs7158733 SNP was not a good candidate for causing the allele dropout, it is possible that some laboratories have indeed designed reverse primers that would be subject to interference by this SNP or other SNPs that may be present in a region not interrogated by our targeted sequencing approach. We also did not investigate the presence of copy number alterations; therefore, it is possible that a copy number change not detectable by our targeted Sanger sequencing may be the cause of the observed allele dropout.

Although we were unable to completely explain the discordance in the SCA3 results for the CAP 2015 challenge MGL2-09 sample, this exercise highlights the unique ability of well-organized proficiency testing programs to detect systemic problems with clinical laboratory tests on a national and an international scale. Furthermore, we hope that this report will remind laboratorians of the need to understand the amplification characteristics of the thermostable DNA polymerases that they use for PCR-based assays and to periodically review frequently updated SNP databases for uncommon polymorphisms that could disrupt amplification of any given target sequence.