Aims To describe the UK NEQAS for Molecular Genetics scheme for gastrointestinal stromal tumour (GIST) and to report and interpret the findings of four rounds of circulation of this quality assurance programme for KIT/PDGFRA mutation analyses.
Methods Samples of GISTs from formalin-fixed paraffin-embedded tissue blocks were circulated to registered participants of the UK NEQAS for Molecular Genetics scheme for GIST. Three samples were provided per annual circulation from 2008 to 2011 inclusive. The participants were required to analyse the samples for KIT and/or PDGFRA mutations using their routine protocols, and the anonymised participants' reports were assessed and an annual scheme report issued.
Results The genotyping error rates for the 2008, 2009, 2010 and 2011 circulations were 13%, 33%, 19% and 4%, respectively. These errors were either missed or incorrectly described mutations. There was an overall false negative rate of 2% and false positive rate of 0%. The main recommendations that arose from these circulations were: (1) inclusion of reference accession numbers in reports; (2) avoidance of the term ‘heterozygous’ when analysing DNA from tumour tissue unless there was certainty that only neoplastic DNA was studied; and (3) the need to screen KIT exons 9, 11, 13 and 17 and PDGFRA exons 12, 14 and 18 before classifying a GIST as ‘wild type’.
Conclusions The UK NEQAS for Molecular Genetics scheme emphasises the potential complexities of KIT/PDGFRA mutation analyses for GISTs and provides recommendations to help optimise such genotyping and reporting. The scheme has also demonstrated its educational value among participating laboratories.
- Gastrointestinal stromal tumours
- kit proto-oncogene protein
- genotyping techniques
- quality assurance
- molecular pathology
- gastrointestinal disease
Statistics from Altmetric.com
- Gastrointestinal stromal tumours
- kit proto-oncogene protein
- genotyping techniques
- quality assurance
- molecular pathology
- gastrointestinal disease
KIT and PDGFRA mutation analyses have several potential uses in the clinical management of gastrointestinal stromal tumour (GIST). Such mutation testing can be used to help confirm or refute a diagnosis of GIST, test for germline mutation to investigate a potential hereditary GIST family, guide choice of chemotherapy and distinguish synchronous GISTs from a single primary with metastases.1–3 As a general rule, only one so-called primary mutation will exist in each GIST which has not been treated with receptor tyrosine kinase (RTK) inhibitors (so-called chemo-naïve GIST).1 However, this primary mutation may arise in one of four KIT exons (9, 11, 13 or 17) or in one of three PDGFRA exons (12, 14 or 18).1 Further, there is a bewildering range of mutation types which may occur, particularly in the KIT exons, ranging from substitutions to indel mutations involving large portions of an exon.1 The failure to demonstrate an activating mutation in the four KIT exons and three PDGFRA exons is used to define a GIST as ‘wild type’.1 Such a genotype carries several important clinical implications including: (1) an association with certain syndromes (eg, neurofibromatosis 1 or Carney-Stratakis syndrome); (2) a potentially different prognosis especially when arising in children and young adults; and (3) likely resistance to the current main RTK inhibitor used (ie, imatinib).1 ,2 Some specific mutation types have specific clinical correlates. A good example is the PDGFRA Asp842Val mutation which confers resistance, at least in the in vitro setting, to both imatinib and a second RTK inhibitor, sunitinib.4 ,5
In view of the above, it is clinically crucial that genotyping of GIST is performed accurately and that laboratories providing this service receive guidance on how to prevent errors. One way of minimising such errors is through participation in a quality assessment scheme. Two such schemes for GIST mutation testing have recently been described. The scheme reported by Merkelbach-Bruse and colleagues initially involved six German laboratories with three arms to the testing (testing of DNA samples, testing of samples requiring DNA extraction and finally testing with different primer pairs).6 This was followed by a round of testing involving 19 laboratories. However, a repeat cycle of this German scheme has not yet been reported and the scheme only required testing of three exons (ie, KIT exons 9 and 11 and PDGFRA exon 18). The quality control scheme reported by Hostein and colleagues involved 11 laboratories predominantly through affiliation with Conticanet (Connective Tissue Cancers Network to integrate European experience in adults and children).7 This scheme required testing of only DNA samples. Each of 50 DNA samples was analysed by four centres which meant that every sample was not tested by all 11 participating laboratories. Furthermore, only one round of the scheme was reported.
The United Kingdom National External Quality Assessment Scheme (UK NEQAS) for Molecular Genetics scheme for GIST was launched in 2008 and has had four annual circulations up to the time of this paper's submission. A principal aim of the scheme was to assess the quality of laboratory testing under conditions as close to real-life practice as possible. Therefore, tissue sections prepared from the same tumour block were distributed to all participating laboratories for DNA extraction and subsequent molecular analysis. This analysis of the same tumour tissue permitted interlaboratory comparisons, and the unique experience of multiple circulation rounds of the scheme has enabled an analysis of the effect of feedback from earlier rounds. The following report describes this UK NEQAS scheme, its findings over four rounds of testing, and the recommendations which have emerged from this experience.
The UK NEQAS for Molecular Genetics scheme for GIST has been run annually since 2008. Initial invitations were emailed to all laboratories registered with the UK NEQAS for Molecular Genetics website (http://www.ukneqas-molgen.org.uk), although participation requests from other interested laboratories were also accepted. All participating laboratories were provided with a secure password-protected UK NEQAS for Molecular Genetics website account for scheme registration, submission of results, accessing scheme scores, downloading scheme reports and recording laboratory performance.
Three cases were selected for each round of circulation. All cases were derived from formalin-fixed, paraffin-embedded tissue blocks. Only tissue blocks containing >70% neoplastic tissue were chosen. The tissue used for the 2008 round derived from the Department of Histopathology, Queen Elizabeth Hospital, Birmingham, UK, that for the 2009 round derived from the Department of Pathology, Christie Hospital, Manchester, UK, whereas the tissue used for the 2010 and 2011 rounds derived from the Department of Histopathology, Bristol Royal Infirmary, Bristol, UK. The scheme required testing for and recording of activating mutations only. The KIT/PDGFRA genotype of each case was validated independently by two laboratories prior to distribution. The samples were distributed in August of each year and participating laboratories were permitted a maximum of 5 weeks to submit their reports. All results were submitted electronically using the UK NEQAS for Molecular Genetics website.
All participant submissions were marked by two independent molecular genetics assessors against marking criteria ratified annually by the UK NEQAS for Molecular Genetics Steering Committee. Submissions were assessed anonymously as the reports were identifiable only by a unique laboratory code. A final scheme report along with individual laboratory feedback was distributed to all participating laboratories in November. Laboratories were invited to submit feedback regarding the scheme report. If applicable, a response to this feedback was returned to the enquiring laboratory before the end of December.
In the 2008 round there were five participating laboratories (all UK-based), in the 2009 round eight participating laboratories (six UK-based), in the 2010 round 14 participating laboratories (six UK-based) and in the 2011 round 16 participating laboratories (six UK-based). The identities of these laboratories remain anonymous as a condition of participation in this NEQAS scheme, although the majority of the non-UK participating laboratories were located in Europe.
Because this was the first pilot round, no formal marking scheme was applied to the scheme. All five laboratories received, for each case, 10 rolls of 10 micron thick sections in an Eppendorf tube; one laboratory also requested a single unstained 5 micron thick slide-mounted section.
The mutations that had been independently validated for each case are shown in table 1. The case 1 mutation was correctly identified and described by all five laboratories. One laboratory missed the case 2 mutation in PDGFRA exon 18; this was attributed to a low level of mutant DNA not being visualised in an electrophoretogram with prominent background peaks. Two of the four laboratories that detected the case 2 mutation described the protein changes as two separate mutations rather than a single indel mutation. One laboratory missed the case 3 mutation in KIT exon 11; this was attributed to the laboratory's forward primer binding too close to the intron 10/exon 11 junction, leading to poor visualisation of this upstream mutation in the forward sequencing reaction. Therefore, as a whole, two of 15 (5 laboratories × 3 cases) submissions (ie, 13%) contained genotyping errors. However, these errors did not include any false positive results (ie, reporting a mutation that did not exist).
Three laboratories described mutations as ‘heterozygous’. All laboratories used HGVS nomenclature although there was some variability in the naming of the complex mutation. Two of five (40%) laboratories did not cite any reference sequences in their reports.
The 2008 scheme report contained two main recommendations. First, description of a somatic ‘heterozygous’ mutation demonstrated from tumour-derived DNA was discouraged; this is because of uncertainty whether the demonstrated wild type allele derived from neoplastic cells or background non-neoplastic cells (eg, vascular constituents, lymphocytes). However, this recommendation need not apply if there had been absolute certainty that the DNA had derived only from neoplastic cells (eg, through use of laser microdissection). Second, citation of appropriate reference sequences was strongly encouraged. While not referred to in the 2008 scheme report, alternative forward primer sequences and hence binding sites were recommended in the individual EQA feedback report to the laboratory that had missed the case 3 mutation.
Three new laboratories joined the scheme. The scheme organisers decided to concentrate on genotyping accuracy with this round and therefore only this component was marked (one point for correct detection of a mutation and one point for correct description of the mutation). All eight laboratories received, for each case, 10 rolls of 10 micron thick sections in an Eppendorf tube; this was the only tissue format distributed in order to assess how well laboratories handled and analysed the same amount and type of tissue.
The mutations that had been independently validated for each case are shown in table 1. The genotyping accuracy of cases 1 and 2 was high. Only one laboratory demonstrated a genotyping error for case 1; this laboratory correctly described the nucleotide change but submitted an incorrect predicted protein sequence change. Only one laboratory demonstrated a genotyping error for case 2; this laboratory incorrectly described the nucleotide and predicted protein changes. Six laboratories incorrectly described the nucleotide and predicted protein change for case 3; all six laboratories did not submit the same genotyping error and this complex indel mutation had been validated independently at the laboratories of the two scheme assessors prior to the distribution. Therefore, as a whole, 8 of 24 (8 laboratories × 3 cases) submissions (ie, 33%) contained genotyping errors. However, these errors did not include any false positive results.
All laboratories used HGVS nomenclature although there was some variability in the naming of the more complex mutations. Two of eight (25%) laboratories did not cite reference sequences in their reports. The two laboratories that had not cited reference sequences in their 2008 round reports did cite these references in their 2009 reports.
The 2009 scheme report contained four main recommendations. First, description of non-activating mutations and polymorphisms was deemed unnecessary as they were and still are not known to have any clinical significance. Second, citation of reference sequences was re-emphasised. The third and fourth recommendations related to the spectra of exons analysed by the participating laboratories. The scheme documented that some laboratories screened all exons simultaneously whereas other laboratories used a two-phase approach, for example, KIT exon 11 and PDGFRA exon 18 for gastric GISTs and, if these two exons showed no activating mutations, then the remaining five exons. These variations in local protocols for GIST mutation testing were considered acceptable at that time (see Discussion regarding GISTs with double primary mutations). However, it was strongly recommended that KIT exons 9, 11, 13 and 17 and PDGFRA exons 12, 14 and 18 should all have been screened before a GIST was classified as ‘wild type’. Finally, one laboratory routinely analysed KIT exon 14 and another PDGFRA exon 10. The scheme assessors reported that existing literature indicated KIT exon 14 to be the site of only secondary mutations (ie, those induced by long-term RTK inhibitor therapy)1 and that, while homologous to KIT exon 9, PDGFRA exon 10 mutations had not thus far been demonstrated among GISTs.4 Screening chemo-naïve GISTs for KIT exon 14 or PDGFRA exon 10 mutations was not therefore recommended.
Six new laboratories joined the scheme. In addition to marking genotyping accuracy (as above), the scheme organisers decided to review the participating laboratories' clinical interpretation of their test data and to mark the clerical accuracy of the reporting in this round (two points for clerical accuracy). All 14 laboratories received, for each case, 10 rolls of 10 micron thick sections in an Eppendorf tube.
The mutations that had been independently validated for each case are presented in table 1. Six laboratories did not offer analysis of PDGFRA exon 14. Of the remaining eight laboratories, one missed the case 1 mutation in PDGFRA exon 14; no reason for this omission was offered by the laboratory. Six laboratories demonstrated a genotyping error for case 2; these errors varied from omission of the nucleotide change to incorrect nucleotide and predicted protein changes (eg, c.1662_1670del; p.Val555_Trp557del) to typographical errors. Three of the remaining eight laboratories described the protein changes as two separate mutations rather than a single indel mutation. The genotyping standard was very high for case 3 with none of the 14 laboratories falsely describing a mutation. Interestingly, one laboratory screened PDGFRA exon 14 for case 3 but not for case 1. Therefore, as a whole, 7 of 36 (14 laboratories × 3 cases) (six unavailable submissions for case 1) submissions (ie, 19%) contained genotyping errors. However, these errors did not include any false positive results.
All laboratories used HGVS nomenclature although there was some variability in the naming of the more complex mutations. Two of 14 (14%) laboratories did not cite reference sequences in their reports. The two laboratories that had not cited reference sequences in their 2009 round reports did cite these references in their 2010 reports.
The 2010 scheme report contained three main recommendations. First, it was re-emphasised that KIT exons 9, 11, 13 and 17 and PDGFRA exons 12, 14 and 18 should all have been screened before a GIST was classified as ‘wild type’. Second, citation of appropriate reference sequences was re-emphasised. The third recommendation was based on the general rule that only one primary mutation is demonstrable in any one chemo-naïve GIST.1 Describing a single unifying mutation is therefore favoured over two separate mutations. When the submissions were being reviewed for clinical interpretation, it quickly became clear that comments regarding prediction of response to RTK inhibitors were variable and would be difficult to mark consistently as part of an EQA scheme. The great diversity of RTK mutations among GISTs means there are insufficient data to reliably predict therapeutic response for many of these mutations, including the complex indel mutation of case 2. Finally, the formal marking of clinical interpretation would also be complicated by the existence of several potential clinical indications for KIT/PDGFRA mutation analyses (as outlined in the Introduction).
Five new laboratories joined the scheme and one of the laboratories that had participated during 2010 withdrew from the 2011 round. The difficulties of assessing clinical interpretation in the previous round were acknowledged. It was therefore decided to continue with assessment only of genotyping and clerical accuracy (marking scheme as before) for the 2011 round. Nonetheless, comments would be fed back to participants regarding any errors in clinical interpretation of their test data. To help further mimic the routine testing performed by participating laboratories, the laboratories were offered a choice of sample type: five rolls of 10 micron thick sections in an Eppendorf tube; five rolls of 10 micron thick sections in an Eppendorf tube with a single unstained 5 micron thick slide-mounted section; or 10 unstained 5 micron thick slide-mounted sections.
The mutations that had been independently validated for each case are presented in table 1. Two laboratories demonstrated a genotyping error for case 1. Both were presumed to represent typographical errors in that one laboratory submitted an incorrect nucleotide sequence (ie, c.2527_2535del) but a correct predicted protein change, whereas the other laboratory described the nucleotide change c.2526_2537del as c.2526-2537del (ie, a single nucleotide deletion located 2537 bp upstream from position 2526). The case 2 mutation was correctly identified and described by all 18 laboratories. The case 3 mutation was correctly identified and described by all 18 laboratories, although one laboratory did not submit a predicted protein change. Also, two laboratories described the protein changes as two separate mutations rather than a single indel mutation. Therefore, there was a high overall genotyping standard with only 2 of 54 (18 laboratories × 3 cases) submissions (ie, 4%) containing genotyping errors. Furthermore, these errors did not include any false positive results.
Two newly participating laboratories described mutations as ‘heterozygous’. All laboratories used HGVS nomenclature although there was some variability in the naming of the more complex mutations. Two of 18 (11%) laboratories did not cite reference sequences in their reports. One of the two laboratories that had not cited reference sequences in their 2010 round reports did cite these references in its 2011 reports whereas the other laboratory had withdrawn from the scheme.
The 2011 scheme report contained three main recommendations. First, it was re-emphasised that describing a single unifying mutation is favoured over two separate mutations. Second, citation of reference sequences was re-emphasised. Third, for reasons outlined above and with the proviso also outlined above, description of a somatic ‘heterozygous’ mutation demonstrated from tumour-derived DNA was discouraged.
We have reported the findings of four rounds of circulation of the UK NEQAS for Molecular Genetics scheme for GIST. One potential limitation of this scheme is the relatively small number of cases tested for each round of circulation. However, GISTs are relatively rare neoplasms with an annual incidence of 10–15 per million8 compared with, in the UK, more than 350 per million for colorectal carcinoma.9 Therefore, requiring laboratories to analyse fewer samples in greater depth (testing potentially four KIT exons and three PDGFRA exons) was considered to mimic more closely real-life practice than testing more samples with a restricted exon panel and/or without a need to extract DNA. Complete simulation of day-to-day practice would require distribution of tissue blocks themselves to participating laboratories as this is how cases are usually received for GIST genotyping, at least in the UK. However, using different blocks from the same GIST would introduce an extra level of variability with, for example, potential differences in neoplastic DNA yield across different blocks. Studying tissue from only one block would minimise such variability, but distributing that one block among multiple laboratories would complicate the temporal organisation of the scheme and would run the risk of the block completely cutting through before it reached the last participating laboratories.
This is the third GIST genotyping quality assessment scheme to be reported and it is therefore worth comparing its detailed findings with those of the first two reported schemes. Merkelbach-Bruse and colleagues' scheme only assessed KIT exons 9 and 11 and PDGFRA exon 18, and its first arm required testing of 10 DNA samples by the five participating laboratories.6 Findings of this arm included missed mutations due to high background peaks in Sanger sequencing electrophoretograms (as was also found in the UK NEQAS scheme) and a false PDGFRA exon 18 mutation due to reamplification of PCR products. The second arm of the scheme required DNA extraction and then mutation analyses of 12 samples derived from FFPE tissue. One-third of these cases were falsely reported as ‘wild type’, and two cases which initially generated uncertainty regarding the presence of a mutation were found to be ‘wild type’ when a DNA extract of the case was distributed for analysis. The third arm of the scheme involved testing of different primer sets, and illustrated that binding sites which were either too far upstream or downstream from the target exon could produce suboptimal or misleading genotyping (as was also found in the UK NEQAS scheme). The recommendations following these three arms included: (1) assessment of DNA quantity and quality by agarose electrophoresis; (2) avoidance of PCR cycles of >40; (3) avoidance of reamplification of PCR products; and (4) manual reading of electrophoretograms. The end of this paper described an “external trial” in which 30 samples were tested by 19 laboratories. Each of these laboratories received five cases to genotype and also 10 electrophoretograms from which to report. Eleven of 19 laboratories were described as having performed the trial “successfully”, although details of what defined a successful performance were not given.
The quality assessment scheme of Hostein and colleagues used 50 DNA samples although each sample was analysed by only four out of 11 participating laboratories.7 There was an overall false negative rate of 4% (eight cases being reported inaccurately as ‘wild type’) and a false positive rate of 2% (one case being falsely reported as harbouring a KIT exon 11 mutation). In comparison, the four circulation rounds of the UK NEQAS scheme revealed an overall false negative rate of 2% (3 out of 129 submissions) and false positive rate of 0%. Hostein and colleagues' scheme reported that use of screening tools such as ‘length analysis of PCR product’ and denaturing high performance liquid chromatography were potentially not sufficiently sensitive, leading to rare cases of mutations being missed. A main recommendation arising from their scheme was a standardised GIST mutation report.
Common points and lessons emerged from these two previous quality assessment schemes for GIST genotyping.6 ,7 First, a low-level mutation may be missed if too much target DNA is initially amplified. Second, use of HGVS nomenclature is strongly recommended. Third, interpretation errors are common. Finally, primer choice is crucial in that suboptimal choice of primer binding sites may lead to mutations being missed. The last two points were also emphasised by the findings of the UK NEQAS scheme. Not surprisingly, interpretation errors were more commonly made with complex mutations assessed in the UK NEQAS scheme. In the 2008 round of this scheme, a suboptimally chosen KIT exon 11 forward primer binding site led to an upstream mutation in this exon being missed. There was only one major difference between the findings of the two previous quality assessment schemes and the UK NEQAS scheme—namely, none of the participants of the latter scheme submitted a false positive result (ie, reporting a mutation that did not exist).
A few particular points and lessons emerged from the four circulation rounds of the UK NEQAS scheme and were emphasised in the scheme's recommendations. First, there is considerable latitude in the naming of complex mutations within HGVS guidelines but, for reasons outlined above, description of one single unifying mutation is currently preferred whenever possible to two separate activating mutations. There have been rare reports of chemo-naïve GISTs harbouring two primary activating RTK mutations10 ,11; this has been particularly reported among ‘micro-GISTs’.10 We are currently reviewing the clinicopathological and molecular characteristics of all double primary mutant GISTs that have thus far been reported by the six UK laboratories performing KIT/PDGFRA mutation analyses. This review aims to assess the authenticity, frequency and potential clinical significance of such double primary mutations, and will guide future UK NEQAS for Molecular Genetics recommendations on this issue. The findings of this review may also guide recommendations as to whether all four KIT exons and three PDGFRA exons should be screened simultaneously or whether a two-phase screening approach is acceptable. A second point that was repeatedly emphasised in the UK NEQAS scheme recommendations was the avoidance of the term ‘heterozygous’ when describing the presence of both a wild type and mutant allele, for reasons outlined above. A few studies have shown that homozygosity of RTK mutations in GISTs associates with more aggressive clinical behaviour.12 ,13 However, such homozygosity has not been linked to any other clinicopathological feature,12 including clinical response to imatinib therapy14 and, at present, the demonstration of homozygosity would not lead to any change in patient management. Therefore, there is currently no clinical need to distinguish between true heterozygosity and homozygosity when reporting RTK mutations of GISTs. Finally, the data derived from the four rounds of circulation are unique in demonstrating that participating laboratories can be guided to optimise their practice, for example, with the citation of reference sequences.
In summary, this UK NEQAS for Molecular Genetics scheme has emphasised the potential complexities of KIT and PDGFRA mutation analyses for GIST and has demonstrated its educational value among participating laboratories. The scheme has provided participants with a mechanism to externally measure the quality of their service in comparison with other laboratories. It has emphasised, in particular, the importance of citing reference sequences and screening all the standard seven KIT/PDGFRA exons before classifying a GIST as ‘wild type’.
Gastrointestinal stromal tumour (GIST) genotyping can be complex, particularly because of the great range and variety of KIT or PDGFRA mutations that are found among these neoplasms.
The UK NEQAS for Molecular Genetics scheme for GIST has shown that the most common GIST genotyping errors are missed mutations or incorrectly described mutations.
The UK NEQAS scheme has compiled several recommendations that can help optimise the detection and accurate reporting of KIT/PDGFRA mutations in GIST.
We thank all the laboratories for their participation in the UK NEQAS for Molecular Genetics scheme for GIST and the UK NEQAS for Molecular Genetics Steering Committee for their support and guidance. We also thank the Department of Histopathology, Bristol Royal Infirmary and the National Genetics Reference Laboratory, St Mary's Hospital, Manchester for their help with sample preparation, validation and assessment.
Competing interests ZD has received educational grants from Astra Zeneca and Roche. NACSW has received educational grants from Novartis and has received lecturing honoraria from Novartis and Pfizer.
Ethics approval The paper reports how well participant laboratories analysed circulated tumour tissue samples for specific mutations. The tumours were derived from human patients but were anonymised at source. The genotyping was performed as part of a quality assurance program and not as a scientific study.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.