Quantification of rare NPM1 mutation subtypes by digital PCR

Nucleophosmin (NPM1) mutations (NPM1mut) are found in 45–55% of patients with normal karyotype acute myeloid leukaemia and are favourable in the presence of wildtype FLT3 (Falini et al, 2005; Schnittger et al, 2005). More than 70% of NPM1mut are type A; 25% have rare subtypes (Alpermann et al, 2013). Quantitative real-time polymerase chain reaction (RQ-PCR) assays targeting NPM1mut are available for monitoring minimal residual disease (MRD) (Schnittger et al, 2009; Krönke et al, 2011; Shayegi et al, 2013). Different expression thresholds have been suggested to define prognostic groups during follow-up (Schnittger et al, 2009; Krönke et al, 2011; Shayegi et al, 2013). Quantitative monitoring was performed mainly on type A mutations. Quantification of rare NPM1mut is challenging due to the need for individual assay designs and standard curves. Recently, digital PCR, which allows absolute quantification (Oehler et al, 2009), was introduced for haematology. For absolute quantification plasmid standards can be avoided, which makes it potentially more convenient to quantify rare NPM1mut than RQ-PCR. We investigated digital PCR to quantify 37 different very rare NPM1mut (non-A, -B, -D) at diagnosis and during follow-up. Validation of the digital PCR assays was compared to a LightCycler-based assay (Schnittger et al, 2009). Furthermore, these assays were applied for MRD diagnostics in a cohort of 99 patients. Following mRNA isolation from mononuclear cells (Schnittger et al, 2002) NPM1mut were identified in diagnostic samples by melting curve analysis (LightCycler 2.0; Roche Diagnostics, Mannheim, Germany) (Schnittger et al, 2005). Mutations were characterized by Sanger sequencing using 3130 and 3730 Genetic Analysers (ABI, Darmstadt, Germany). Nomenclature of NPM1mut subtypes followed ENSEMBL transcript ID ENST00000296930 (Table 1). To quantify the NPM1mut load by RQ-PCR the %NPM1/ABL1 level was determined (Gorello et al, 2006; Schnittger et al, 2009). Digital PCR was performed by mutation-specific reverse primers (Table SI) using a hydrolysis probe-based assay on 12.765 Digital Arrays (Fluidigm, San Francisco, CA, USA), that separates each PCR reaction into 765 reaction chambers. Analysis was done by EP1 Data Collection v.3.1.2 software (Fluidigm). Positive signals were counted with digital pcr analysis v.3.1.3 software (Fluidigm) and transformed to a value of estimated targets. Taking into account the number of estimated targets, the amount of cDNA added and the dilution factor of the cDNA, the absolute copy number of target transcripts was calculated. Expression levels were calculated as %NPM1/ABL1. To evaluate the individual assays, serial dilutions of diagnostic cDNA samples (n = 9) were prepared and the sensitivity and specificity of each assay was determined (Table SII). %NPM1/ABL1 levels obtained by conventional RQ-PCR and digital PCR (n = 58 samples) were compared by Pearson's correlation. Primers and probes are given in Table SII. To assess the sensitivity and specificity of the new digital PCR assays, cDNA of nine different diagnostic samples with NPM1_A mutation (n = 1) and rare subtypes I (n = 2), J (n = 2), K (n = 2), N (n = 1) and R (n = 1) were diluted in 1:10 steps in an NPM1mut-negative cDNA background. These dilution experiments indicated a sensitivity of 10−4–10−5 for the different NPM1mut subtypes (in detail, see Table SI). Figure 1 illustrates a dilution experiment by both conventional (A) and digital RQ-PCR (B) for a diagnostic sample with NPM1_K. To test applicability for MRD monitoring we selected cases with NPM1_A (n = 4), I (n = 5) and R (n = 6), from whom follow-up samples and conventional RQ-PCR results were available (n = 58 samples in total). Samples were retrospectively analysed by digital PCR to show comparability of both methods. Mutation levels by both methods were positively correlated (Pearson's correlation r = 0·51, P < 0·0001; Fig 1C–E). Subsequently, 99 patients with 37 different NPM1mut were analysed by digital PCR at diagnosis and during follow-up. Further patient details provided in the Supporting Information. Four hundred and ninety-eight samples were investigated, 83 at diagnosis, 415 during follow-up. The median number of samples per patient was 4 (range, 1–28). Mutations, nomenclature and patient numbers investigated are given in Table 1. Most frequent mutations were I (n = 22/99; 22·2%), K (n = 9; 9·1%), R (n = 8; 8·1%), J and L (n = 6; 6·1% each), C, G, and ZE (n = 5; 5·1% each). Clinical follow-up was available in 84/99 patients. Median expression levels at different intervals were: diagnosis (n = 83): 1351·2 (range, 127·0–17912·0), days 18–60 (n = 31): 0·9 (0·0–567), days 61–120 (n = 36): 0·0 (0·0–15290·0), days 121–365 (n = 46): 0·0 (0·0–590) and >day 365 (n = 48): 0·0 (0·0–4279). The 2-year overall survival (OS, Kaplan Meier) for patients investigated at diagnosis was 88·7%, 2-year event-free survival (EFS) 69·5% (median follow-up 16·8 months). By Cox regression in the interval 18–60 d a higher MRD level as continuous parameter was associated with worse OS (P = 0·098) and EFS (P = 0·049), whereas in other intervals the MRD level showed no significant impact. Further, a threshold of 0·01%NPM1mut/ABL1 was chosen according to our previous publication (Schnittger et al, 2009). During all intervals, patients with %NPM1mut/ABL1 levels ≥0·01 had worse outcome than patients with %NPM1mut/ABL1 levels <0·01 (Table SIII). At days 61–120 patients with ≥0·01%NPM1mut/ABL1 had significantly worse survival than patients <0·01 (2-year OS 66·7% vs. 100·0%, P = 0·009; 2-year EFS: 0·0% vs. 100·0%; P = 0·004). Patients with ≥0·01%NPM1mut/ABL1 at >365 d from diagnosis had worse 2-year EFS than patients <0·01%NPM1mut/ABL1 (44·4% vs. 91·8%; P = 0·001; Fig 1F). Within all other intervals analysed, patients with <0·01%NPM1mut/ABL1 also had superior survival, but significance was not reached (Table SIII). So far, MRD studies applying RQ-PCR have targeted only a limited number of NPM1mut subtypes (Schnittger et al, 2009; Krönke et al, 2011). MRD studies by next-generation sequencing (Thol et al, 2012) permit assessment of all NPM1mut but is resource-intensive. We showed the applicability of digital PCR for a large variety of NPM1mut without the need for plasmid standards. The digital PCR sensitivity (10−5) was comparable to that of conventional RQ-PCR (Schnittger et al, 2009). NPM1mut levels by digital and conventional RQ-PCR were significantly correlated at diagnosis and during follow-up. The investigation of 99 patients with rare NPM1mut confirmed a digital PCR threshold of 0·01%NPM1mut/ABL1 to be prognostically relevant according to our previous RQ-PCR MRD study (Schnittger et al, 2009). Thus, digital PCR enables MRD diagnostics in all rare NPM1 mutation subtypes tested and is reliable to identify patient risk during follow-up. We thank all physicians for sending patients′ samples and providing clinical information to our laboratory. FD, UB, CH, TA, and SuS performed data analysis. UB wrote the first manuscript draft. FD and SuS performed molecular analyses. DR performed statistics and created figures. All other authors contributed to data analysis, interpretation of data, characterization of samples, writing of the manuscript, and approved the final version. CH, WK, TH and SuS are part owners of the MLL Munich Leukaemia Laboratory GmbH. FD, UB, TA and DR are employed by the MLL Munich Leukaemia Laboratory GmbH. Data S1. Additional information to patients, primers and assays. Table SI. Primers and probes used for the different NPM1 mutation subtypes. Table SII. Characterization of the sensitivity and specificity of digital PCR performed on different NPM1 mutation subtypes for the establishment and the validation of the assays. Table SIII. Characterization of the sensitivity and specificity of digital PCR performed on different NPM1 mutation subtypes for the establishment and the validation of the assays. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.