Concordance of assays designed for the quantification of JAK2V617F: a multicenter study

Abstract

Background Many different techniques have been designed for the quantification of JAK2V617F allelic burden, sometimes producing discrepant results.Design and Methods JAK2V617F quantification techniques were compared among 16 centers using 11 assays based on quantitative polymerase chain reaction (with mutation-specific primers or probes, or fluorescent resonance energy transfer/melting curve analysis), allele-specific polymerase chain reaction, conventional sequencing or pyrosequencing.Results A first series of blinded samples (granulocyte DNA, n=29) was analyzed. Seven assays (12 centers) reported values inside the mean±2SD; the mean coefficient of variation was 31%. Sequencing techniques lacked sensitivity, and strong discrepancies were observed with four techniques, which could be attributed to inadequate standards or to different modes of expression of results. Indeed, quantification of JAK2V617F in relation to another control gene produced higher than expected values, suggesting the possibility of more than two JAK2 copies/cell. After calibration of assays with common 1% to 100% JAK2V617F standards (dilutions of UKE-1 cells in normal leukocytes), 14 centers tested ten new samples. JAK2V617F allelic burdens greater or equal than 1% were then reliably quantified by five techniques – one allele specific-polymerase chain reaction and four TaqMan allele-specific quantitative polymerase chain reaction assays, including one previously giving results outside the mean±2SD – with a lower mean coefficient of variation (21%). Of these, only the two TaqMan allele-specific quantitative polymerase chain reaction assays with primer-based specificity could detect 0.2% JAK2V617F.Conclusions Techniques expressing the allelic burden as JAK2V617F/total JAK2 and using a common set of standards produced similar quantification results but with variable sensitivity. Calibration to a reference standard improved reproducibility.

Introduction

The diagnosis of Philadelphia-negative myeloproliferative neoplasms has changed dramatically with the discovery of a somatic point mutation in the JAK2 gene (JAK2 1849G>T or JAK2V617F).1 Detection of this mutation is now a major criterion for the diagnosis of Philadelphia-negative myeloproliferative neoplasms.2 Whether JAK2V617F is the initial genetic event responsible for JAK2V617F-positive myeloproliferative neoplasms is still a matter of debate.3,4 Not all myeloid cells carry the mutation and, because of mitotic recombination, the mutation can be found on both JAK2 alleles.5,6 The variation in mutation burden thus reflects both the size of the myeloproliferative clone and the homozygous or heterozygous status of clonal cells. Homo/heterozygosity cannot be differentiated from JAK2V617F allelic burdens (%JAK2V617F) representing lower than 50% of total JAK2, but homozygosity can be inferred when %JAK2V617F is greater than or equal to 50%.

We previously reported that JAK2V617F burdens are higher in polycythemia vera than in essential thrombocythemia.7,8 In addition, a high %JAK2V617F has been associated with increased risk of thrombotic events in essential thrombocythemia; increased risk of cardiovascular events, larger spleen and greater treatment requirement in polycythemia vera; and higher leukocytosis and risk of acute transformation in primary myelofibrosis.9–11 Some of these findings, however, are controversial: Pemmaraju et al.12 found no correlation between the %JAK2V617F in neutrophils and occurrence of thrombotic events. Thus reproducible quantification of the JAK2V617F burden is of major interest in the evaluation of myeloproliferative neoplasms. Quantification of JAK2V617F may also be important in the follow-up of patients who have received a bone marrow transplant or treatment with interferon α-2a and the soon-to-come JAK2 inhibitors.13,14

Understanding the discrepancies between the results of the studies reported here is partly dependent on the comparison of the techniques used for quantification. We, therefore, undertook a study comparing 11 different techniques, carried out in 16 laboratories using various instruments. After a first round of analyses, we distributed dilutions of JAK2V617F-positive DNA to standardize the assays. With adjustment for the mode of expression of the results and the use of common standards, five techniques were similarly reliable for the quantification of JAK2V617F loads greater than or equal to 1% of total JAK2; two of these techniques could detect 0.2% JAK2V617F. The advantages and disadvantages of the five techniques are presented.

Design and Methods

Design of the study

From April to June 2007, laboratories with published assays designed for the quantification of JAK2V617F were identified via a PubMed search and contacted. Sixteen centers using 11 different techniques responded, agreeing to participate in a comparative study of JAK2V617F assays. Initially, a series of 29 blinded DNA samples from purified blood granulocytes of patients referred to three participating laboratories (centers 4, 5 and 7) were distributed to the 16 laboratories. Three of the 29 samples had been found negative; the JAK2V617F burden was lower than 10% in ten samples and greater than 10% in 16 samples. Each center was asked to process the sample according to their standard procedures and report the %JAK2V617F [(JAK2V617F/total JAK2) × 100].

The standards used by eight participating centers were analyzed in center 7 using an allele-specific quantitative polymerase chain reaction (AS-qPCR).7,15 Subsequently, a set of standards based on a JAK2V617F-positive, diploid cell line (UKE-1) was defined and distributed to all centers. Following the calibration of local standards, a second series of test samples, comprising ten samples from patients and a series of DNA dilutions of two cell lines (HEL/Karpas 299, prepared in center 9), were assessed blindly. The HEL/Karpas series included four samples with a JAK2V617F allelic burden lower than or equal to 2%, to provide further information on the sensitivity of the various techniques. Technical issues prevented centers 8 and 13 from participating in this part of the study.

Patients

The diagnosis of myeloproliferative neoplasm was established according to WHO guidelines for all JAK2V617F-positive patients. Diagnoses were, for the two series of samples, respectively, secondary erythrocytosis (1; 1), atopic hypereosinophilia (1; 0), essential thrombocythemia (16; 3), polycythemia vera (10; 5) and primary myelofibrosis (1; 1).

Cell line characterization and DNA preparations

UKE-1 cells, kindly provided by Dr. Walter Fiedler (Hamburg, Germany), were grown as described elsewhere.16 DNA content was assessed by flow cytometry after treatment of ethanol-permeabilized cells with RNAse and staining with propidium iodide. The karyotype was established using standard cytogenetic techniques, synchronization with synchroset (Amplitech, France), heat denaturation and Giemsa staining of R bands. For preparation of JAK2V617F standards, UKE-1 cells were counted with a hemocytometer and diluted with leukocytes (buffy coat white blood cells, WBC) obtained from normal healthy individuals. Pellets of each mix of UKE-1 and WBC (100% UKE-1, 75% UKE-1 + 25% WBC, 50% UKE-1 + 50% WBC, …) were kept frozen at −20°C. Genomic DNA was extracted using the QiaAmp kit. An additional cell line dilution series was obtained by diluting HEL into Karpas-299 cells. A mixture of 20% HEL + 80% Karpas 299 was found to have 50% JAK2V617F. Further dilutions were performed, taking this percentage into account.

Plasmids

Three sets of plasmids were used for quantification. Plasmids A and B were prepared in center 4 as described elsewhere.7 Briefly, a PCR fragment obtained by reverse transcriptase PCR of U937 or HEL (respectively wild type and mutant) with primers 3′-TTATGGACAACAGTCAAACAACA ATTCT-5′ and 3′-AACTCCTGTTAAAT-TATAGTTTACTGACACC-5′, was cloned into a TopoTA vector (Invitrogen, Carlsbad, CA, USA). Plasmids C and D were generated in center 8 by amplification of an 881 bp fragment with primers 3′-GATTCACTAATCATACCCAG-5′ and 3′-TTGAACCTGCCATAATCTCT-5′, cloned into pGEM T Easy (Promega, WI, USA). Plasmid E (mutant only) was prepared in center 15 as described by Hammond et al.20 The fragment obtained with primers 5′-TATCCCAAAGGACAGAAGTATTCATT-3′ and 5′-TTGTGTTTCCCTCCCT GTTGGA-3′ was cloned into Zero Blunt Topo (Invitrogen).

JAK2V617F quantification

The characteristics of the 11 techniques, used either for diagnosis/follow-up of patients with myeloproliferative neoplasms or for research, are presented in Table 1, while sequences of primers and probes can be found in the Online Supplementary Table S1. Most (13/16) centers had developed a quantitative real time PCR assay, predominantly (11/13) involving hydrolysis probe-based quantification. The same Taqman assay was used in five centers (centers 4, 5, 6, 7, and 8) with different real-time PCR machines; another assay was used in centers 9 and 10 on different qPCR platforms. Center 2 used fluorescent resonance energy transfer (FRET) probes, center 13 used SYBRgreen chemistry. Five assays based the specificity of detection on a primer, which at centers 9 and 10 was modified by inclusion of a mismatch and a locked nucleic acid. Center 12 quantified the products of a final point allele-specific PCR. Centers 3 and 16 performed DNA sequencing, with pyrosequencing and conventional Big Dye terminator chemistry, respectively. The assay used in center 1 was based on the TaqMan allelic discrimination assay principle. Results were expressed as percentages of mutated allele (%JAK2V617F) by all laboratories but center 15, which determined the ratio of JAK2V617F versus a control gene, human growth hormone (HGH). The different standards used for calibration included plasmids, cell lines or genomic DNA from patients.

Statistics

The mean and standard deviation (SD) of results of the 16 centers (mean16 and SD16, respectively) were established for each sample tested. Results were then labeled as within 2 SD from the mean (mean16±2SD16) or as exceeding 2 SD from the mean. Centers with more than three results outside the mean16±2SD16 interval were excluded, and new mean and SD of values were then calculated. Similar analyses were conducted for the second series of samples, also using boundaries of 2 SD. Coefficients of variation (CV) were used to analyze the variability in results reported for each sample. Results from the different centers were also compared using a two factor analysis of variance (ANOVA).

Table 1.Overview of JAK2V617F assays employed by the participating centers.

Results

Assessment of %JAK2V617F in the first series of samples

A first series of 29 blinded DNA samples was proposed to the 16 participating centers (Table 2). There was one false positive (declared weakly positive at 0.1%) and six false negative results in two centers that used sequencing-based techniques. The mean and SD of results of the 16 centers were established for each sample tested. Centers 13, 14 and 15 had more than three results outside of the mean 16±2 SD16. In addition, centers 1 and 13 had results systematically lower than the mean16. A new, narrower mean 12±2 SD12 range was calculated from results generated by the remaining 12 centers. The mean CV in %JAK2V617F assessment by these 12 centers was 31%. Concordance improved with increasing mutational load: the mean CV was 63.4% for samples containing lower than 5% JAK2V617F, and 19.8% for samples with greater or equal than 5% JAK2V617F (p<0.001). Because the results of three centers were systematically low (centers 1 and 13) or high (center 15), we suspected under- and over-estimation of %JAK2V617F due to inaccurate calibration of standards. In contrast to other centers, center 15 expressed results as a percentage of number of copies of JAK2V617F compared to the number of copies of a control gene (HGH), instead of %JAK2V617F DNA of total JAK2 DNA. The 100% allelic ratio was defined as an equal number of JAK2V617F and HGH copies, while a ratio greater than 100% implied more JAK2V617F than HGH copies. Center 14, whose results were within the mean16±2SD16 for samples with lower than 10% JAK2V617F, but outside for JAK2V617F loads greater than 10%, also determined the %JAK2V617F by comparison with a control gene other than JAK2 (hemopoietic cell kinase, HCK). Results were then plotted against those of a cell line, UKE-1, with known JAK2V617F status (see below). These additional steps, although not hampering detection of low levels of mutant – the primary goal of this technique – were likely to be responsible for the variations in ratios observed in samples with high mutational loads.

Table 2.%JAK2V617F of total JAK2 n DNA of the first series of 29 clinical samples as determined by the 16 participating centers.

Comparison and optimization of standards

To confirm that part of the discrepancies could be attributed to the use of different references, standards were compared on the same apparatus with a single technique. Although there was a slight difference in crossing points obtained with linear (A,B) versus circular (C,D,E) plasmids (Figure 1), ratios of numbers of JAK2V617F/total JAK2 copies determined in unknown samples using both sets of plasmids yielded similar results. %JAK2V617F in standards from center 1 (HEL cells diluted in normal peripheral blood lymphocytes (PBL), or HEL/PBL standard) and center 11 (patient’s DNA), were quantified against sets of plasmids A-B and C-D, with similar results. However the observed %JAK2V617F in the HEL/PBL standard was consistently higher than that reported by center 1 (Online Supplementary Table S2). This was consistent with the fact that HEL cells have more than two copies of JAK2, and explained the systematic underestimation of the JAK2 mutational load by center 1.

To avoid variation of results due to differences in standards, we searched for a standard that could be used by all laboratories. Plasmid DNA standards allow reproducible and reliable quantification; however, not all diagnostic laboratories are able to produce plasmids. We reasoned that DNA from diploid cells characterized as heterozygous for the JAK2V617F mutation would be the best standard. DNA extracted from heterozygous erythroid colonies represents the ideal 50% standard. Unfortunately, single colonies not contaminated by non-mutated cells are difficult to obtain and provide limited amounts of DNA. As a substitute, we used the DNA of the UKE-1 cell line, generated from a transformed myeloproliferative neoplasm. This cell line bears a bi-allelic JAK2 mutation in the context of a roughly diploid genome, as demonstrated by cytogenetic and DNA content analyses (Online Supplementary Figure S1). Standard dilutions were generated by diluting UKE-1 cells with white blood cells (WBC) obtained from normal healthy volunteers. DNA was then extracted and the allelic ratio for each dilution (100%, 75%, 50%, 25%, 10%, 5% and 1% JAK2V617F) was found to be similar to the theoretical value using A-B and C-D plasmid dilutions as standards (Online Supplementary Table S2). To validate the UKE-1 standards definitively, the results from a 50% dilution (50% UKE-1 cells/50% normal WBC) were compared to results obtained with DNA from single colonies of JAK2V617F-positive patients. Although not always exactly identical, the %JAK2V617F observed in DNA from heterozygous colonies and in the UKE-1 50% sample were very similar (Online Supplementary Table S2).

Assessment of %JAK2V617F using common UKE-1 standards

Fourteen laboratories, using ten assays, participated in this second part of the comparisons. The UKE-1 standards were distributed to all centers, with the instruction to calibrate local standards if necessary. A new series of samples from ten patients plus a series of DNA dilutions of HEL/Karpas 299 cells including four samples with lower or equal than 2% JAK2V617F, were then blindly assessed.

Only one false positive result was reported (DNA pool of healthy individuals declared to contain 0.15% JAK2V617F) (Table 3). The sample with the lowest theoretical mutation load (0.15% JAK2V617F) was found to be positive by only four of the ten techniques. Three other samples with low mutation burden (0.2–1.8%) were declared negative by centers 3 and 16. Center 16 also reported two samples with 1.6% and 2.1% JAK2V617F (mean values) as negative.

Three centers reported three or more values outside the mean 14±2 SD14. These were centers 14 and 15, for the reasons mentioned above, as well as center 2. Means and SD (mean11, SD11) were re-calculated without these three centers. For these 11 remaining centers, the use of common UKE-1 standards reduced the variability of JAK2V617F quantification: no center reported systematically high or low values (Figure 2) and all values from center 1 were now within the mean11±2 SD11.

Hence, as summarized in Table 4, when using common standards, 11 centers using seven assays consistently reported values inside the mean±2 SD for samples with 5% or more JAK2V617F. The %JAK2V617F reported by these 11 centers for the UKE-1 and HEL/Karpas 299 DNA dilutions were found to be not significantly different from the theoretical %JAK2V617F. These seven assays detected JAK2V617F with different levels of sensitivity, ranging from 5% for conventional DNA sequencing to 0.15–0.20% for TaqMan assays with specificity based on primers. The mean CV in assessment of %JAK2V617F using the five assays with a sensitivity of 1% or less was reduced to 21%.

Figure 1.Comparison of JAK2 plasmids. TaqMan AS-qPCR were performed in center 7 to determine the crossing points for five concentrations of plasmids A, C [wild type (WT) A], and B,D, and E (JAK2V617F, B).

Discussion

This study highlights the importance of using defined, accurate standards to calibrate JAK2V617F quantitative assays. The semi-quantitative assay of center 1, which reported values outside the mean±2 SD in the first part of the study, reported values inside in the second part of the study thanks to the application of common standards. In addition, use of common standards reduced the variability of results. Although the choice of standards remains the prerogative of each center, we suggest combining plasmid DNA dilutions, which allow precise quantification of the number of copies of JAK2, with at least one well-calibrated genomic DNA sample, as an internal control. Dilutions of the diploid UKE-1 cell line, with two copies of mutated JAK2 per cell, represent useful internal genomic DNA standards.

Table 3.Quantification of JAK2V617F using common UKE-1 standards.

The study also underlined the necessity to use a single mode of expression of results, the percentage of total JAK2 represented by JAK2V617F (%JAK2V617F). Centers that expressed JAK2V617F as percentages of genes other than JAK2 consistently reported values outside the mean±2 SD. However, the %JAK2V617F values reported by center 15 that were more than two-fold the average values were exclusively found in samples from patients with polycythemia vera and primay myelofibrosis. Cells with more than two copies of wild type or mutated JAK2 may be found in cases of myeloproliferative neoplasms with chromosome 9 trisomy, which occurs in approximately one third of cases of polycythemia vera.23 Future studies with techniques evaluating the absolute number of copies of JAK2 per cell by quantifying against one or several control genes20,22 may bring useful information on the status and evolution of patients with myeloproliferative neoplasms.

Remarkably, using common UKE-1 standards, seven assays – conventional sequencing, pyrosequencing, one AS-PCR assay and four TaqMan assays – were found to assess JAK2V617F similarly. The mean CV in assessing 1% or more JAK2V617F in granulocyte DNA using the TaqMan and AS-PCR assays was relatively low (21%). Hence, the choice of technique depends on the desired sensitivity and equipment available. Not surprisingly, the two techniques with the poorest sensitivity were sequencing techniques. Both techniques allow rapid processing of large series of samples; pyro-sequencing also enables the detection of potential additional mutations present less than six to eight bases downstream of 1849G>T (V617F). Less expected was the finding of three very different techniques providing values inside the mean±2 SD with 1% sensitivity: one was quantitative (TaqMan assay with sensitivity based on probes, center 11) but two were semi-quantitative assays (TaqMan with double dye probe, center 1; AS-PCR, center 12).

The best sensitivity – 0.2% JAK2V617F – was achieved by three assays. One was reliable only for mutation loads of less than 10% JAK2V617F. Thus only two TaqMan AS-qPCR assays with primer-based specificity, using plasmid dilutions for wild type and mutated JAK2 quantification, consistently provided values inside the mean±2 SD with a sensitivity estimated at 0.2%. Both assays require two PCR reactions. The assay using LNA-modified primers gave the only false positive result (less than 0.2%). One might question the necessity to detect less than 1% JAK2V617F. However, highly sensitive techniques are required to study residual disease, notably for those patients who have received allografts or treatment with interferon α-2a.13,14 Moreover, we and others also recently demonstrated that hydroxyurea treatment can reduce JAK2V617F burden to very low or undetectable levels.24,25 Obviously, for a less than 1% sensitivity to be meaningful, JAK2V617F assays should be performed on purified cells, not on the highly variable mix of cells of whole blood.26

Figure 2.Observed %JAK2V617F values compared to average and theoretical %JAK2V617F values. A and B. Results obtained by the various techniques giving values inside the mean±2 SD were compared to mean values (open diamonds) for samples from patients in the first (A) and second (B) series. Mean plus 2 SD (closed circles) and mean minus 2 SD (closed triangles) are indicated. C and D: results obtained by the various techniques giving values inside the mean±2 SD were compared to theoretical values (open diamonds) for dilutions of the Karpas (C) and UKE-1 (D) cell lines. Mean plus 2 SD (asterisk) and mean minus 2 SD (cross) obtained by all 14 laboratories or by the 11 laboratories producing values inside the mean±2 SD (closed triangles and closed squares) are represented.

Table 4.Advantages and disadvantages of the seven JAK2V617F assays providing results inside the mean±2 SD.

Finally, none of the assays tested can guarantee accurate quantification of JAK2V617F for all patients. Any unsuspected additional mutation occurring within the sequence of primers can potentially reduce or prevent amplification of the concerned allele, resulting in inaccurate quantification of JAK2V617F and possible false negative results. In these instances, a second assay based on primers or probes with non-overlapping sequences will confirm the initial assessment of %JAK2V617F.

Acknowledgments

the authors wish to thank Elizabeth Young, Katie Thomas, Lydie Lecucq, Sophie Salzes and Francis Belloc for expert technical help and the tumor bank of the CHU de Bordeaux for providing samples.

Footnotes

• Funding: this study was supported by a grant from the Comité Grand Ouest and Comité Aquitaine Charente of the Ligue contre le Cancer to SH and EL, respectively, and by contributions from Exeter Leukaemia Fund to KH.
• The online version of this article contains a supplementary appendix.
• Authorship and Disclosures EL: performed analyses, provided samples, analysed the data and wrote the paper; FG: performed analyses and provided samples; EH: performed analyses, provided samples and plasmids, NSR: performed analyses and provided samples; JJ, BF, KH, MH, CR, SS, VU, SC, VH, CM, DP, MS, MM, SE, NK, RH, JTP, RCS: designed assays and performed analyses; MM: performed statistical analyses; SH: designed the study, performed analyses, provided plasmids, and wrote the paper.
• The authors reported no potential conflicts of interest.

• Received June 11, 2008.
• Revision received August 14, 2008.
• Accepted August 25, 2008.