Droplet digital PCR for the quantification of Alu methylation status in hematological malignancies

This study included a total of 46 patients affected by hematologic malignancies, subdivided into three groups: a) thirty patients affected by chronic lymphocytic leukemia (CLL) (Twenty-three males and seven females; median age at diagnosis 59 years, range 28–79 years); b) seven patients affected by myelodysplastic syndromes (MDS), at intermediate/high risk according to the International Prognostic Scoring System (IPSS) score [23] (six males and one female, median age at diagnosis 68 years, range 61–81 years); c) nine patients with chronic myelomonocytic leukemia (CMML) (six males and three females, median age at diagnosis 71 years, range 62–83 years). For CLL patients, the presence of the 11q, 13q and 17p deletions, and trisomy 12 (del(11q), del(13q), del(17p), and + 12, respectively), was identified by Fluorescent In Situ Hybridization (FISH), as previously reported [24, 25]. CLL patients without the above-mentioned alterations were globally classified as “normal karyotype”. The 13q14.3 deletion was detected as the sole cytogenetic alteration in 6/18 (33%) patients, whereas in the remaining cases it was present in association with del(17p) (28%), del(11q) (28%), + 12 (22%) and del(6q) (5%). Del(11q) and del(17p) were present as the sole cytogenetic abnormality in #18, #12 and #29 cases, respectively. CLL patients were also analyzed for mutational status of the IgV_H and NOTCH1 gene hotspot c.7541_7542delCT by Sanger Sequencing and allele-specific oligonucleotide PCR (ASO-PCR), respectively. For MDS patients, bone marrow (BM) samples were analyzed at diagnosis and during AZA treatment. CMML patients at diagnosis were previously profiled for mutations in ASXL1 exon 12 and the SRSF2 hotspot gene mutation (c.284C>D) by SS and ASO-PCR, respectively.

The most important patients’ clinical and molecular characteristics are summarized in Additional files 1-3: Tables S1-S3. For MDS patients, the number of AZA cycles received up to the time of BM aspiration is indicated (Additional file 2: Table S2). Ten healthy donors (HD) were included in our analysis as controls (six males and four females, median age 58 years, range 55–81 years). This study was approved by the local ethics committee, and all patients provided written informed consent to take part in this project.

Acute promyelocytic leukemia (APL)-derived NB4 cells (DSMZ, Braunschweig, Germany) were also included in the study. As previously described, the PML/RARα translocation in APL is associated with an overall increase in methylation [26], as well as in untreated NB4 cells [27]. Cells were plated using 2 normal T25 culture flasks (1.5 × 10⁷ cells in each flask) with 15 mL of RPMI 1640 supplemented with 10% heat inactivated FBS, 1% penicillin/streptomycin and 375 μl DEC 30 μM (final concentration of 0.75 μM) or 375 μL of PBS, and were incubated for 3 days at 37 °C, 5% CO₂. Four experiments of our ddPCR assay on NB4 cultures, untreated or treated with DEC, were done on four different days. DEC concentration used in cellular experiments were compliant with the maximum concentrations reached in human plasma at clinically routine dosages [28, 29].

Genomic DNA (gDNA) was directly isolated from peripheral blood (PB) using the QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany), and from BM and cell cultures (collected cells were first resuspended in RLT buffer and PBS solution, respectively) using the DNeasy Blood & Tissue Kit (Qiagen). Extracted gDNA samples were quantified with the Qubit 2.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA).

According to the manufacturer’s digestion-ligation procedures (EpiJET DNA Methylation Analysis kit, T4 DNA Ligase, Thermo Scientific), we used 250 ng of gDNA input for each reaction. For each DNA sample two aliquots of 250 ng of gDNA were simultaneously digested with 1 unit of either MspI or HpaII restriction enzymes, and ligated to a previously prepared synthetic adaptor [18] in parallel in two separate tubes. The synthetic adaptor was previously prepared [18] by incubating two complementary oligonucleotides at 65 °C for 2 min, and then at 18 °C for 35 min.

In the digestion-ligation reaction, 250 ng of gDNA, 1 μL of a synthetic adaptor 0.1 μM, 1 μl of MspI or HpaII (corresponding to 1 U), and 2 U of T4 ligase were added, in addition to 2 μl of ligase buffer 10X and 2 μl of MspI/HpaII buffer 10X, and nuclease-free water to a final volume of 40 μl. Thermal conditions for digestion and ligation were: 1 h at 37 °C and 2 h at 16 °C.

The digestion-ligation mixtures (final concentration 6.25 ng/μl) were subsequently serially diluted through sequential dilutions to obtain a final concentration of 2 pg/μl corresponding to a final amount of 6 pg distributed in three wells.

Ten samples were diluted to two final concentrations, 10 pg/μl and 2 pg/μl, to test the use of smaller amounts of DNA samples and evaluate the reproducibility of Alu sequences methylation data starting from different amounts of digested/ligated gDNA. One single test was executed on patients’ samples, because the initial amount of gDNA was the limiting factor.

Like QUAlu [18], this assay is based on the selective amplification of Alu sequences containing a CpG site within the Alu consensus sequence AACCCGG.

We defined the precise amount of gDNA as input of our Alu assays using serial diluted mixtures, to obtain a final concentration of 10 pg/μL or 2 pg/μL, so as to get smaller amounts of template and to avoid the droplet saturation limit in ddPCR. For each sample the two final dilutions of digestion-ligation mixtures of MspI and HpaII were analyzed.

Prior to proceeding further with ddPCR tests, we verified that the primers used did not form any dimers during the amplification step nor aspecific sequences, by performing qualitative PCR.

ddPCR experiments were performed using the QX-200 instrument (Bio-Rad, Hercules, CA). ddPCR experiments were conducted in triplicate for both the digestion-ligation mixtures. The 20 μL ddPCR reaction mixture was then loaded into the Bio-Rad DG8 droplet generator cartridge. A volume of 70 μL of droplet generation oil was loaded for each sample. The cartridge was placed in the QX200 droplet generator. Thermal-cycling conditions were 95 °C for 5 min (1 cycle), 95 °C for 10 s (ramp rate 2 °C/second, 40 cycles), 65 °C for 7 s (ramp rate 2 °C/second, 40 cycles), 4 °C for 5 min (ramp rate 2 °C/second, 1 cycle), 90 °C for 5 min (ramp rate 2 °C/second, 1 cycle), and 4 °C hold. The expected amplicon size was 57 bp.

After amplification, the 96-well PCR plates were loaded on the Bio-Rad QX200 droplet reader and ddPCR data were analyzed with QuantaSoft analysis software (version 1.7.4, Bio-rad). Since the amplicon size was slightly below the range generally recommended in ddPCR, for wells flagged as “No Call” visual inspection and manual setting of the threshold value was performed. The target concentration in each sample was expressed as digested consensus Alu copies/μL.

Statistical analysis was performed using conventional pipelines in R 3.1.2 (ww.​r-project.​org). To test for differences between two distributions, the Wilcoxon Rank Sum Test for matched (MDS patients) or independent samples was used, whereas the Kruskal-Wallis and Dunn tests were used for comparing more than two groups of patients. Spearman correlation was used for correlation analysis. The significance level was set at p < 0.05 for all analyses.

According to the GRCh37/hg19 human genome assembly, genomic coordinates and fasta sequences of Alu repeats were retrieved from UCSC Genome Browser (https://​genome.​ucsc.​edu/​index.​html). Alu consensus sequences were retrieved by searching the Alu consensus motif AACCCGG. ChIPseeker and org.​Hs.​eg.db R packages were used for genomic distribution analysis of Alu consensus sequences. Since our ddPCR assay involved Alu sequences containing the AACCCGG consensus motif, we retrieved Alu genomic coordinates and fasta sequences, and filtered those containing this motif.

From a total of 1,142,278 Alu sequences, 171,702 (about 15%) had the consensus sequence and were the real target of our assay. Considering a region range of Transcription starting site (TSS) of 3000 bp, 74.4% of Alu sequences were annotated as “distal intergenic”, and about 5% were in the promotor regions (Additional file 4: Table S4).

We also observed a positive correlation between the chromosome length and Alu sequences distribution (rho Spearman correlation = 0.74, p = 5.64 × 10^− 5), indicating a globally homogeneous distribution of these target sequences along all the human chromosomes.

Comparing the Alu methylation level of 7 out of 10 CLL patients, we did not observe any statistically significant difference starting from 10 or 2 pg/μL (p = 0.6). As observed in Fig. 1, the Alu methylation assay by ddPCR starting from 10 or 2 pg for well showed a good linearity (R² = 0.9627); however, in 3 out of 10 cases analyzed, we observed droplet saturation starting from 10 pg of gDNA for well. Indeed, QuantaLife software was not able to discriminate between positive and negative results and returned “No Call” for wells with too many positive droplets. For this reason, we decided to perform all the subsequent ddPCR experiments using 2 pg of digested/ligated gDNA for each ddPCR reaction, and calculated the sum of the three values of positive droplets.

We assessed our assay in CLL patients at diagnosis. ddPCR data from 30 CLL patients were compared with the values obtained from 10 age-matched HD, showing a statistically significant decrease of Alu methylation in CLL patients compared to the HD values (p < 0.05, Fig. 2a).

We also investigated the global Alu methylation level in relation to different cytogenetic risk groups. To this aim, CLL patients were classified in the following three groups according to the karyotypic alterations identified by FISH: low-risk (with isolated del(13q)), intermediate (with normal karyotype or + 12), and high risk (with del(11q), del(17p) or more than two chromosomal aberrations detected by FISH analysis). For CLL patients harboring two cytogenetic aberrations, the highest-risk alteration observed was considered [30]. Alu methylation status of the low-risk and high-risk groups was significantly reduced compared to HD (p < 0.05), whereas considering intermediate-risk patients the difference was not evident (Fig. 2b).

We also extended the analysis to other prognostic factors such as immunoglobulin heavy-chain variable (IgV_H) mutational status and the presence of NOTCH1 hotspot c.7541_7542delCT, but no statistically significant difference in ALU methylation levels was observed according to these molecular parameters (data not shown).

We performed four experiments of our ddPCR assay on the gDNA extracted from NB4 cultures (Fig. 3a), and compared the global Alu methylation level of untreated cells to the levels in those treated with DEC hypomethylating agent. The mean of the four experiments was 84.07 and 64.35 for untreated and DEC treated cells, respectively (standard deviation, SD = 1.58, and 7.21, respectively). We observed a significant decrease of the global Alu methylation level in DNA extracted from NB4 cells treated with DEC 0.75 μM, as compared to untreated cells (p < 0.05) (Fig. 3b). In Fig. 3c a replicate of our ddPCR assay performed on the NB4 cell line in the two different conditions is shown.

These data suggested the potential ability of ddPCR to successfully detect global Alu methylation variations, and its potential use in the molecular monitoring of onco-hematological patients undergoing hypomethylating treatment.

On the basis of the preliminary results obtained analyzing NB4 cells, we decided to test our ddPCR assay on genomic DNA from MDS patients treated with AZA.

Paired samples from 7 MDS patients at diagnosis and after some cycles with AZA were retrospectively analyzed according to the availability of BM samples (Additional file 2: Table S2). Comparing the global paired Alu methylation levels at diagnosis and after AZA treatment, we observed a statistically significant decrease of Alu sequences methylation after therapy (p < 0.05) as compared to diagnosis (the median Alu methylation level was 73 and 85% for pre- and post-treatment patients, respectively; Fig. 4a-b). Secondary AML from MDS occurred in 3 patients (Case #1–3), probably correlating with the observation of a new increase of the global Alu methylation level (Fig. 4a).

Finally, we attempted to test the assay in 9 CMML patients; 2 patients were mutated for ASXL1 and 2 for SRSF2 (exon 12 and the c.284C>D hotspot, respectively; Additional file 3: Table S3), two frequently mutated genes in CMML [31‐33].

Overall, CMML patients showed no differences in Alu methylation levels as compared with HD (p > 0.05). Among CMML patients, a decrease of Alu sequences methylation was observed in those harboring the main SRSF2 hotspot compared to patients without this mutation (p < 0.05, Fig. 5), whereas no significant difference was observed according to the presence of mutations in ASXL1 exon 12 (data not shown).