Introduction

Systemic mastocytosis (SM) is characterized by abnormal proliferation and accumulation of mast cells in diverse tissues such as bone marrow (BM), skin, liver, spleen, or gastrointestinal tract. Diagnosis is based on the presence of one major (multifocal dense infiltrates of mast cells in BM biopsies and/or in sections of other extracutaneous organs) and at least one minor, or the presence of three out of four minor criteria (>25 % atypical cells on BM smears or spindle-shaped mast cell infiltrates, the presence of a mutation in the tyrosine kinase (TK) KIT in BM or other extracutaneous organs, expression of CD2 and/or CD25 by mast cells in BM, peripheral blood (PB) or another extracutaneous organ and baseline serum tryptase concentration >20 μg/l) [1–4].

Diagnosis of aggressive SM (ASM) is based on the presence of one or more C-findings [cytopenia (ANC < 1 × 109/l, Hb < 10 g/dl, or platelets < 100 × 109/l), hepatomegaly with impaired liver function, palpable splenomegaly with signs of hypersplenism, malabsorption with significant hypoalbuminemia, significant weight loss >10 % over the last 6 months and/or large osteolyses). Mast cell leukemia (MCL) is defined by the presence of at least 20 % mast cells on BM smears. Organ damage, C-findings and markedly elevated STL >200 μg/l are frequently observed [5–7]. In a subset of patients, an associated clonal hematologic non-mast cell disorder is present (SM-AHNMD), most frequently chronic myelomonocytic leukemia (SM-CMML) or hypereosinophilic syndrome/chronic eosinophilic leukemia (SM-HES/CEL). Smoldering SM (SSM) is a recently defined subgroup, which requires at least two of three B findings (hepatomegaly and/or splenomegaly, BM mast cells >30 %, STL >200 μg/l, increased marrow cellularity/dysplasia or organomegaly) in the absence of C findings. Patients with typical BM infiltrates but not fulfilling diagnostic criteria for ASM, MCL, SM-AHNMD, or SSM are diagnosed as indolent SM (ISM). Median survival for SM-AHNMD, ASM/MCL, SSM, and ISM is 24, 41, 120, and 198 months, respectively [8–10].

Depending on subtype, adequate material, and assay sensitivity, a gain of function mutation from aspartate to valine at codon 816 within the TK domain of KIT (KIT D816V), the receptor for stem cell factor, can be identified in cells from BM or other affected tissues in >80 % of patients [10–12]. Very rarely, other somatic (D815K, D816F, D816H, D816Y, and V560G) and germline (F522C, A533D, K509I, and del419) KIT mutations have been reported [10, 13]. However, accurate numbers on the overall frequency of KIT D816V in the diverse SM subtypes remain largely unknown because of (1) the rarity of disease, (2) highly variable mast cell burden in BM and PB, and (3) inadequate sensitivity of disparate assays. Conventional sequencing (CS) has a sensitivity of only 10–20 % and repeatedly fails to detect the mutation in those patients in which only a small fraction of cells is affected. Because mast cells are usually not present in PB, the use of BM is recommended for diagnostic procedures. However, KIT mutations may not be restricted to mast cells and may thus also be found in other non-mast cell lineages suggesting that the expansion of clonal mast cells derives from an uncommitted hematopoietic stem cell [14–16].

Using a highly sensitive allele-specific quantitative real-time PCR (RQ-PCR), we here demonstrate that the KIT D816V mutation is present in more than 97 % of SM patients. If positive in BM, it is found in PB from all patients with advanced SM (SM-AHNMD, ASM/MCL) and in 46 % of ISM patients. There is a strong correlation between the KIT D816V expressed allele burden (KIT D816V EAB) and disease activity, disease subtype, and consequently also survival. In addition, serial measurement of the KIT D816V EAB is useful for monitoring of residual disease after chemotherapy or allografting.

Materials and methods

Patients, samples, and cell lines

Overall, 210 samples from 147 patients (male, n = 74; female, n = 73; median age, 59 years; range, 26–82) with SM [ISM, n = 63 (43 %); SSM, n = 8 (5 %); SM-AHNMD, n = 16 (11 %, sAML, n = 2; MDS/MPN-u, n = 2; MPN, n = 1; CEL/HES, n = 9; CMML, n = 2); ASM/MCL ± AHNMD, n = 60 (41 %)] according to the World Health Organization 2008 diagnostic criteria were studied (Table 1). CML patients in major molecular remission (n = 65) and healthy volunteers (n = 20) served as controls. In 62 patients, samples were collected contemporaneously from BM and PB (ISM, n = 28; SSM, n = 4; SM-AHNMD, n = 5; ASM/MCL, n = 25). All patients gave informed consent according to the Declaration of Helsinki. The study was approved by the institutional review board as part of the “German Registry on Eosinophilia and Mastocytosis.” For assay sensitivity calculation, cell dilutions of D816V positive HMC-1.2 cell line (kindly provided by Dr. Butterfield, Mayo Clinic Rochester, MN, USA) [17] in KIT wild-type NB4 cells (DSMZ Braunschweig, Germany) and diluted patient RNA in healthy control BM RNA were performed (Clontech, Mountain View, CA, USA).

Table 1 Patients' characteristics
Full size table

RNA extraction and cDNA synthesis

Total RNA was extracted using the commercial available TRIzol Reagent (Life Technologies, Darmstadt, Germany) or the CsCl gradient centrifugation method as previously described [18]. Complementary DNA (cDNA) was synthesized using random hexamers and MMLV reverse transcriptase (Life Technologies).

RQ-PCR and D-HPLC

Two RQ-PCR assays amplifying both the mutant KIT D816V and the KIT wild-type (wt) allele were designed (for primer sequences see Supplementary Information Table 1). The total KIT assay (KIT D816V plus KIT wt) served as internal control for cDNA quantity in addition to a previously described ABL control gene assay [19]. PCR was performed using the universal “mastermix” (LightCycler Faststart plus set hybridization probes, Roche Diagnostics, Mannheim, Germany) on a LightCycler instrument 1.5 (Roche Diagnostics) in a final volume of 20 μl with 2 μl cDNA or plasmid product (500 nm primer; 250 nm probes). RQ-PCR cycling conditions were as follows: 95 °C (10 min), 45 cycles: 95 °C (1 s), 60 °C (10 s), and 72 °C (26 s). A plasmid standard curve was used for absolute quantification. [Supplementary Information (Fig. 1, Methods)]. KIT D816V EAB was calculated as ratio between mutant KIT D816V and total KIT transcripts (mutant KIT + wt KIT expressed as % KIT D816V/KIT). In order to validate the novel KIT D816V EAB PCR assay, samples and results were compared to a recently published RQ-PCR on DNA level [20].

For denaturing high-performance liquid chromatography (DHPLC), KIT was amplified by nested RT-PCR [Supplementary Information (Table 1, Methods)] using the high fidelity Optimase Polymerase (Transgenomic, Omaha, NE, USA). KIT amplicons were analyzed after heteroduplex formation at 58 °C using a Wave 3500HT System (Transgenomic). Equal amounts of PCR product and an estimated 1 % mutant (diluted HMC1 cell line) were analyzed to confirm low mutant or high mutant proportions of the KIT D816V mutation. PCR products were sequenced bidirectional using the amplification primers.

Statistical analysis

Correlation between variables was investigated by Spearman's rank correlation coefficient, scatter plots or Mann–Whitney U test (t-approximation). Survival analysis was calculated using the Kaplan–Meier method. Significance level was 0.05 for all statistical testing. Analyses were performed with Graph pad prism 5.

Results

Assay sensitivity of RQ-PCR and D-HPLC

The maximum sensitivity of RQ-PCR and D-HPLC was calculated by analyzing serial cell dilutions of KIT D816V-positive HMC1 cells in NB4 cells and total RNA dilutions from KIT D816V-positive patients (KIT D816V EAB 50 and 80 %) with total RNA from BM of healthy individuals. The D-HPLC was optimized for high sensitivity detection of the most common KIT D816V mutation. D-HPLC detected mutant KIT D816V down to 0.5–1 % of mutant in normal cells based on cell line and RNA dilutions. For the RQ-PCR assay, enhanced sensitivity of 0.01–0.1 % mutated cells assessed by cell line dilutions could be achieved. However the KIT D816V EAB of the analyzed samples in the present data set was higher than 0.01 % in all cases and typically achieved minimal values of  0.1 %. The inter- and intra-assay variations were calculated by repeated analyses (eight and ten times tested) of a cDNA cell line dilution of the KIT D816V-positive HMC1 cell line with the KIT wt NB4 cell line corresponding to a mutation ratio of 0.1 and 25 %. The levels of assay variation were high in the samples with low mutation levels (11 and 35 %) and low in the samples with high mutation levels (9 vs. 15 %).

The RQ-PCR assay for KIT D816V and total KIT showed linearity over five orders of magnitude (from 40 to 4,000,000) as assigned by standard plasmid dilutions. In order to analyze the nonspecific cross-reaction of the wt allele with the KIT D816V mutation-specific assay, cDNA samples from 85 KIT D816V negative healthy volunteers (n = 20) and CML patients in major molecular remission (BCR-ABL/ABL ratio <0.1 %, n = 65) were analyzed. A nonspecific cross amplification (≥36CP) was observed in 34 of 85 (40 %) samples by RQ-PCR. We therefore assigned samples with less than ≥35CP by RQ-PCR as negative. Analyzing the same cell and mRNA dilutions by CS, an assay sensitivity of only 10–15 % mutant proportion could be achieved (Supplementary Information Fig. 1).

Comparison between CS, D-HPLC, and RQ-PCR

The three assays were applied to 210 samples from 147 SM patients with various SM subtypes. BM samples were available from 83 of 147 (56 %) and PB samples from 127 of 147 (86 %) patients. Four patients (ASM/MCL, n = 2; ISM, n = 2) were negative for KIT D816V but positive for rarer mutations at codon 816 (D816Y and D816H). These patients had an abnormal elution profile in the D-HPLC assay and the KIT D816H/Y mutations were detected by CS (Supplementary Information Fig. 1). One patient with SM-HES/CEL was negative for the KIT D816V or other rarer KIT D816 mutations in BM and PB using all available techniques. Diagnosis in this patient was defined by mast cell aggregates and immunohistochemistry (positivity for CD25). These five patients were consequently excluded from further analysis of KIT D816V expression levels in different disease subtypes of SM. RQ-PCR was the most sensitive technique and identified the KIT D816V mutation irrespective of disease subtype in 100 % of BM samples and 78 % of PB samples. All patients were KIT D816V-positive in BM. In PB, 46 % of ISM patients were KIT D816V positive (Fig. 1). For detection of KIT D816V, the overall sensitivities of CS (BM, 74 %; PB, 64 %) and D-HPLC (BM, 99 %; PB, 65 %) were inferior to RQ-PCR (BM, 100 %; PB, 78 %).

Fig. 1
figure 1

KIT D816V expressed allele burden (EAB) in SM-subtypes: frequency of KIT D816V positivity according to disease stage and sample material. Significant differences of the KIT D816V EAB in PB and BM were observed for indolent (ISM/SSM) vs. advanced SM (ASM/MCL ± AHNMD)

Full size image

Comparison RNA/cDNA vs. DNA

We compared the KIT D816V EAB by quantitative RT-PCR with a recently established allele-specific quantitative PCR on genomic DNA [20]. In 25 samples with KIT D816V EAB ranging between 0 and 54 %, there was a strong correlation between both assays (Spearman r = 0.899, p < 0.001). In an additional 12 patients who were KIT D816V-positive in BM but negative in PB by quantitative RT-PCR, 3 patients were also tested negative by genomic PCR, whereas 9 patients were tested positive by genomic PCR. The median genomic allele burden in those nine patients was 0.067 % (range, 0.007–0.13 %), thus confirming the 0.1–0.01 % sensitivity threshold of quantitative RT-PCR.

KIT D816V EAB in BM and PB

In ASM/MCL ± AHNMD, the median KIT D816V EAB did not significantly differ between BM and PB (32 vs. 37 %; range, 1–99 vs. 0.2–74; p = n.s.), whereas the median KIT D816V EAB in ISM/SSM was significantly higher in BM as compared to PB [9 vs. 0.2 %; range, 1.1–50 vs. 0–43; p < 0.0001 (Fig. 1)]. This was confirmed in contemporaneously collected BM/PB samples of 63 patients (ASM/MCL ± AHNMD: BM 27 vs. PB 33 %, p = n.s.; ISM/SSM: BM 9 % vs. PB 0.1 %, p < 0.0001, data not shown).

Correlation with complementary laboratory parameters and clinical findings

KIT D816V EAB in PB clearly distinguished between indolent and advanced phase disease (Fig. 1). Within the ASM/MCL group, statistically significant different KIT D816V EAB levels were observed for patients with or without AHNMD (38 vs. 22 %, p = 0.019) with or without monocytosis >1 × 109/l (45 vs. 29 %, p = 0.0002) and with or without elevated serum levels of alkaline phosphatase and gamma glutamyl transpeptidase >2× upper limits of normal, 33 vs. 11 %, p = 0.0079; data not shown). No differences were found regarding eosinophilia. Seven patients with advanced SM had a KIT D816V EAB <5 %. A screen in three of these patients revealed mutations in additional genes (KIT-ASXL1-CBL-EZH2-TET2, KIT-RUNX1, and KIT-U2AF1-DNMT3A) consistent with advanced disease [21, 22]. In ISM, nine patients had a KIT D816V EAB ≥5 %. Five of these patients had signs of more advanced disease, e.g., borderline Hb level <12 g/dl (n = 3), thrombocytopenia <140 x 109/l (n = 2), monocytosis >1.0 × 109/l (n = 2) or eosinophilia >1.0 × 109/l (n = 1) without clear diagnosis of AHNMD in BM. In three of the nine ISM patients, comprehensive mutational profiling was performed, which showed no additional mutations apart from KIT D816V [21]. Overall, there was a good correlation (r = 0.59, p < 0.0001) between KIT D816V EAB and STL (Fig. 2). No correlation was observed between BM mast cell infiltration and KIT D816V EAB, but there was neither a correlation between BM mast cell infiltration and STL.

Fig. 2
figure 2

Correlation of KIT D816V EAB with serum tryptase level (STL). Significant correlation between KIT D816V EAB in PB (x-axis) and STL (y-axis) irrespective of disease subtype

Full size image

Monitoring of residual disease

A significant reduction of the KIT D816V EAB in BM and PB was observed in four ASM/MCL ± AHNMD patients after chemotherapy with cladribine (n = 1), aggressive chemotherapy (n = 1), or allogeneic stem cell transplantation (SCT, n = 2), respectively. In the cladribine-treated patient, the KIT D816V EAB decreased from 42 to 5 % within 9 months. The mast cell infiltration in BM correspondingly decreased from 80 to 10 %. At the time of relapse (+18 months), the patient presented with progressive organomegaly, thrombocytopenia, and a KIT D816V EAB of 35 %. The results were closely mirrored by decrease and increase of STL (Fig. 3). The second patient showed a history of SM for 24 months, when aggressive chemotherapy had to be initiated after progression to secondary acute leukemia. After 4 weeks, the KIT D816V EAB decreased from 59 to 22 % (data not shown). Prior to allogeneic SCT of two ASM/MCL patients, the KIT D816V EAB in PB was 35 and 45 %, respectively. Both patients remain in complete clinical and molecular remission without detectable KIT D816V mutation repeatedly performed up to 12 and 17 months after allogeneic SCT (data not shown).

Fig. 3
figure 3

KIT D816V EAB and STL during treatment with cladribine. Monitoring of the KIT D816V EAB in a patient with ASM-AHNMD during treatment with cladribine revealed a 1 log decrease from 42 to 5 %. The patient relapsed 18 months later, which was reflected by increased KIT D816V EAB and STL in PB

Full size image

KIT D816V EAB and overall survival

The KIT D816V EAB is strongly correlated with disease activity (e.g., STL, p < 0.001), disease subtype (e.g., indolent vs. advanced SM, p < 0.001) and consequently survival. Any survival analysis based on different KIT D816V EAB levels, e.g., 0 vs. >0 %, <2 vs. ≥2 %, <5 vs. ≥5 %, <10 vs. ≥10 %, or <20 vs. ≥20 %, respectively, provides significant p values. The best overall discrimination is seen for the group 0 vs. >0 % (p < 0.0001). None of the 57 ISM/SSM patients with a median KIT D816V EAB of 0.2 % has yet died, whereas 16 (median KIT D816V EAB of 36 %) of 67 (median KIT D816V EAB of 37 %) patients with advanced SM have died (data not shown).

Discussion

The accuracy and reliability with which cytogenetic and molecular aberrations are detected in hematologic malignancies are highly dependent on the number and source of cells as well as the techniques employed. Until recently, the most frequently used assay for the detection of point mutations and insertion/deletions has been CS of PCR-derived amplicons. However, the level of sensitivity is only between 10 and 20 %, good enough for detection of acute leukemias but leading to potentially false negative results in disorders with low disease burden, e.g., JAK2 V617F-positive essential thrombocythemia or KIT D816V-positive ISM [23, 24].

While the source of cells is not important for most hematological malignancies at diagnosis, an exceptional situation is found in SM because mast cells usually entirely reside in the BM and are only very rarely found in PB [25, 26]. Consequently, it has been repeatedly suggested that BM is required for reliable detection of the KIT D816V mutation in SM. However, disease subtypes may greatly vary regarding mast cell burden and involvement of other lineages. In addition, assay sensitivities have been considerably improved down to 0.1–0.01 % using genomic DNA based approaches, including PCR amplification followed by restriction digestion [27], peptide nucleic acid-mediated PCR clamping technique [11], and qualitative [28] and quantitative allele-specific PCR assays [20, 29]. More sophisticated techniques include cell sorting [30] or enrichment of malignant cells via laser microdissection followed by melting curve analysis of PCR products [31, 32].

We have established a novel allele-specific quantitative RT-PCR (RQ-PCR) with an improved sensitivity down to 0.01 % and consequently detected the KIT D816V mutation in 97 % of patients with histologically confirmed SM. Because the RQ-PCR assay cannot detect mutations other than D816V, we employed a D-HPLC assay in D816V-negative patients, which allowed the identification of alternative mutations within the PCR-generated amplicon. Four additional patients were identified with two KIT D816H and KIT D816Y mutations each, overall leaving only 1 out of 147 patients not carrying a KIT D816 mutation. In BM, positive results were obtained irrespective of disease subtype. All patients in advanced phase were also positive in PB potentially indicating involvement of several myeloid and/or lymphoid lineages as recently reported in patients with SSM [33] or ASM/MCL [11, 30]. In contrast, only 46 % of ISM patients were also positive in PB. Key factors for qualitative (positive/negative) and quantitative (KIT D816V EAB/genomic AB) presence of KIT D816V in PB therefore include disease burden, disease subtype, multilineage involvement, and assay sensitivity. Since the KIT D816V EAB is the basis of disease burden, we also observed a dependent correlation with disease subtype and consequently survival.

The broadest heterogeneity of KIT D816V EAB was observed in SSM and SM-AHNMD, although it still correlated strongly with STL. The presence or absence of KIT D816V in the cell compartment representing AHNMD remains elusive and requires elaborate molecular tests on sorted or microdissected cells [34]. In SM-CMML, monocytes are KIT D816V-positive in >90 % of cases, whereas the status of eosinophils in SM-HES/CEL remains less clear [11, 32]. In line with these results, we observed a significantly different KIT D816V EAB in patients with or without monocytosis but no differences in patients with or without eosinophilia. In contrast to ISM, it was suggested that in advanced SM the KIT mutation occurs in a multi progenitor cell resulting in multilineage involvement [35]. Consequently, a high level of KIT D816V mutation-positive progenitor cells were found in PB of advanced SM as compared to ISM and healthy controls [29, 36]. Of interest, we observed significant KIT D816V EAB differences between patients with ASM/MCL with or without AHNMD (38 vs. 21.5 %, p = 0.019, data not shown).

With an assay sensitivity reported to down to 0.003 %, Kristensen et al. [29] detected the KIT D816V mutation in all 25 ISM patients tested. In four of these patients, the KIT D816V allele burden was <0.01 % in PB and <0.1 % in BM, respectively, which would not have allowed the detection of KIT D816V with our assay. One patient was diagnosed upon atypical morphology and positivity of CD2/CD25, while three patients fulfilled the major criterion of compact mast cell aggregates. Three ISM patients exhibited a genomic allele burden of more than 35 %, which we would only have observed in advanced disease. No detailed clinical data or STL were presented, which would have facilitated to explain potential differences of the study cohorts. Because the genomic assay has not yet been validated in patients with advanced disease, a direct comparison of both assays in 37 samples including all subtypes of SM was performed and provided a strong correlation. In patients with mutation levels below 0.1 %, the RQ-PCR is however less sensitive. Until recently, we have preferred RNA/cDNA for routine analysis because it also allows screening for fusion genes (e.g., BCR-ABL1 and FIP1L1-PDGFRA) and gene expression levels (e.g., array technology, quantitative RT-PCR for expression of PDGFRA and PDGFRB), particularly for patients with associated eosinophilia and monocytosis [37–39]. However, the recently shown importance of pathogenetically relevant, additional mutations, and the higher sensitivity of DNA assays in indolent disease may redirect the focus on analysis of DNA [21, 22]. In a small subset of ASM/MCL patients (n = 14), we now have also shown the previously lacking potential value of this DNA assay in advanced phase disease. Of interest, all these studies share the common observation of disparities between BM mast cell infiltration, STL, genomic/transcriptional allele burden, and C findings. They may, at least in part, be explained by the recently described molecular heterogeneity of advanced SM with >50 % of patients being positive for four and more pathogenetically relevant point/length mutations [21].

RQ-PCR is a fast, sensitive, reliable, and cost-effective method for routine screening of KIT D816V in all SM subtypes. The KIT D816V EAB is strongly correlated with disease burden, disease subtype, and survival. PB is sufficient for mutation analysis in patients with advanced disease, while a negative result from PB should be complemented by a BM analysis in indolent SM. In the very rare event of negativity in histologically confirmed SM, patients should be screened by D-HPLC for rare mutations at codon 816, e.g., D816H or KIT D816Y, since they are not detected by allele-specific RQ-PCR. CS on unfractionated cells is certainly inadequate for reliable detection of KIT D816 mutations. Compared to established response markers like improvement of blood counts and reduction of organomegaly, BM mast cell infiltration, or STL, RQ-PCR is likely to provide most of its complimentary benefits at low levels of residual disease, e.g., after intensive chemotherapy or allogeneic SCT. Similar to CML, standardized RQ-PCR results may become objective and reproducible prognostic markers and treatment endpoints in clinical trials and daily clinical practice [40].