Dear Editors,

Acute myeloid leukemia (AML) can be classified into multiple genetic subtypes based on recurrent pathogenic structural variants (SVs), copy number abnormalities (CNAs), and single nucleotide variants (SNVs) that inform prognostication and clinical management [1, 2]. Karyotype analysis is considered mandatory in the evaluation of AML. If karyotype analysis fails, fluorescence in situ hybridization (FISH) can be used as an alternative technique per ELN 2022 [2]. Recent reports have proposed whole genome sequencing (WGS) as an alternative methodology to karyotyping and FISH [3, 4]. MPseq is one example of a WGS technique optimized for the detection of genome-wide SVs and CNAs [5]. However, previous studies have not directly assessed the prognostic utility of WGS approaches in identifying genetic abnormalities above gold-standard cytogenetic approaches within an unbiased, prospective setting.

Here, we performed a prospective evaluation of MPseq in comparison to karyotyping and FISH combined with panel sequencing in the genetic characterization of 105 cases of AML from the Mayo Clinic from August 2017 to December 2018 (Fig. 1A, Supplementary Figs, 1, 2, Supplementary materials and methods). The median age of the cohort was 65 years (range 1–90) with 10 (9.5%) patients under the age of 30 years. Just over half of the cases represented de novo AML (n = 55, 52%), 23 (22%) had AML with myelodysplasia-related changes (AML-MRC), 21 (20%) had relapsed AML and 6 (6%) had therapy related AML (Supplementary Tables 1, 2). The most prevalent cytogenetic result based on all three methodologies was a normal karyotype (37 cases, 35%), followed by deletions of chromosomes 5q and/or 7q (25 cases, 24%). Seven had a simple, non-complex karyotype with non-subtype defining abnormalities, 6 had trisomy 8, 5 had a complex karyotype without 5q or 7q deletions (atypical complex), 6 had a NUP98 (11p15.4) rearrangement and 7 had a KMT2A (11q23.3) rearrangement with gene partner MLLT10 (10p12.31) in 3 cases, ELL (19p13.11) in one case, MLLT6 (17q12) in one case and MLLT3 (9p21.3) in 2 cases. Four had t(15;17)(q24;q21), 3 had inv(3)(q21.3q26.2) or t(3;3)(q21.3;q26.2) including a single case with inv(3) with BCR::ABL1, 3 had inv(16)(p13.1q22) or t(16;16)(p13.1;q22), and a single case each with either t(6;9)(p23;q34.1) or KAT6A rearrangement involving 8p11.2 (Fig. 1B).

Fig. 1: Genetic characterization and abnormal distribution of AML cohort.
figure 1

A 105 cases from patients with a diagnosis of AML. Karyotype, FISH, and MPseq was performed to identify the structural variation (SV) including copy number abnormality (CNA), results were tabulated, and cases were divided into the following subtypes based on karyotype, FISH, or MPseq results. NGS-based panel sequencing was also performed to identify pathogenic or likely pathogenic single nucleotide variants (SNVs). B Pie chart displaying the relative distribution for each cytogenetic subtype in 105 AML cases. Normal Karyotype (NK, 35%), 5q deletion (5q del) and/or 7q deletion (7q del) (24%), simple karyotype (7%), Trisomy 8 (6%), NUP98 rearrangement (NUP98r, 6%), KMT2A rearrangement excluding t(9;11)(p22;q23) (5%), atypical complex karyotype (CK) (5%), t(15;17)(q24;q21) (4%), inv(3)(q21.3q26.2) (3%), inv(16)(p13.1q22) (3%), t(9;11)(p22;q23) (2%), t(6;9)(p23;q34.1) (1%) and KAT6A rearrangement (1%). The t(9;11) rearrangements are separated from other KMT2A rearrangements due to their differential influence on outcome. C Distribution of discrepant cases are indicated in the boxes. No karyotype data are indicated in light grey. Abnormalities detected by FISH, karyotype and MPseq are indicated in dark grey, abnormalities detected by FISH and MPseq but not karyotype are indicated in purple, abnormalities detected by karyotype but not FISH or MPseq are indicated in red, abnormalities detected by karyotype and MPseq but not FISH are indicated in green, abnormalities detected by FISH and karyotype, but not MPseq are indicated in yellow and abnormalities detected by MPseq, but not FISH and karyotype are indicated in blue. Abnormalities not detected by karyotype, FISH and MPseq are indicated in white. Blast by flow cytometry or morphology are indicated. D Percentage of concordance between MPseq vs. karyotype + FISH, MPseq vs. FISH, FISH vs. karyotype and MPseq vs. karyotype for AML-related genomic events described in C. E Percentage of concordance between each MPseq vs. karyotype + FISH, MPseq vs. FISH, FISH vs. karyotype, and MPseq vs. karyotype for individual genomic events.

Full size image

When evaluating cytogenetic results obtained by MPseq compared to karyotype plus FISH, 100 of 105 cases were concordant (95.2%) (Table 1, Fig. 1C, D). Of the 5 discordant cases, MPseq missed 4 abnormalities that were identified by karyotype and FISH. All 4 were found in a low level by FISH including a KMT2A rearrangement detected in 17% of cells in a case with t(11;17)(q23;q11.2). In 3 other cases, MPseq missed low-level trisomy 8 found in 3–5% of cells by FISH and detected by karyotype (Table 1, Fig. 1C, D). Since MPseq is validated to identify SVs >10% and CNAs >25% of the tumor clone [5], the missed KMT2Ar represents a false negative result. The final case of discordance included an atypical 7q deletion identified by MPseq that was missed by karyotype and FISH because it did not map within the FISH probe utilized (Table 1).

Table 1 Cases with discrepancies between karyotype, FISH and MPseq.
Full size table

When evaluating cytogenetic results obtained by each methodology individually, the concordance between MPseq and FISH was 93.3%, between FISH and karyotype was 85.7% and between MPseq and karyotype was 82.9% (Fig. 1D). FISH missed 4 abnormalities identified by MPseq involving deletions of 5q in 1 case and 7q in 3 cases because the deletions did not map within the FISH probe utilized (Table 1, Supplementary Fig. 3). Karyotype analysis failed to detect 11 abnormalities that were identified by FISH or MPseq (Table 1). These included 6 cases with a NUP98 rearrangement, 2 cases with a TP53 deletion, and one case each with a KMT2A::MLLT10 fusion and an inv(8) resulting in a KAT6A rearrangement, and a 7q deletion. Additionally, karyotype analysis detected 4 cases with a 17p deletion, which presumptively included loss of the TP53 locus, that were not identified by FISH or MPseq. There was 100% concordance for some AML-associated rearrangements with consistent and cytogenetically detectable breakpoints including t(9;22)(q34.1;q11.2), t(6;9)(p23.3;q34.1), inv(16)(p13.1q22), inv(3)(q21.3q26.2), and t(15;17)(q24;q21) (Fig. 1C) demonstrating that karyotype, FISH, and MPseq are similarly reliable methodologies to identify these recurrent rearrangements. However, as expected, all NUP98 rearrangements observed in this study were cryptic by karyotype and detected by both FISH and MPseq (Fig. 1E). Reduced concordance between karyotype and MPseq was also observed for 17p/TP53 deletions (57%), KMT2A rearrangements (75%) and trisomy 8 (81%) (Fig. 1E). FISH and MPseq have a higher sensitivity in comparison to karyotype analysis in detecting NUP98 and KMT2A rearrangements and TP53 deletions. For low-level abnormalities (<25%), FISH has increased sensitivity compared to MPseq if the abnormality is targetable by the available FISH probe.

Of the 5 cases with discordant results between MPseq compared to karyotype plus FISH (Table 1), the ELN risk stratification remained unchanged (Supplementary Table 3). No additional cryptic, prognostically defining genetic abnormalities per ELN 2022 were identified by MPseq in the remaining samples, including those with a normal or simple karyotype (Fig. 1C). Thus, MPseq did not alter the ELN risk stratification above information provided by karyotype combined with FISH demonstrating that by using the current risk stratification guidelines, karyotype in combination with FISH analysis remains a robust laboratory approach in the evaluation of AML.

We next evaluated whether MPseq could uncover AML-associated abnormalities not currently incorporated into the ELN guidelines. Using a list of 109 genes implicated in AML or MDS [6], we identified an average of 7.5 individual gene aberrations per case including an average of 3.9 losses, 3.1 gains or amplifications and 0.5 SVs (Supplementary Fig. 4). Overall, MPseq identified additional aberrations in 43 cases (40.1%) involving AML genes undetected by karyotyping (Supplementary Table 4). The most frequently deleted genes included KMT2C, CUX1, and EZH2, located on 7q and the most frequently gained genes included MYC and/or CCDC26, RAD21, and TRPS1, located on chromosome 8. MECOM was the most frequently rearranged gene in our AML cohort (Supplementary Figs 4–5). MPseq identified 11 cases (10.5%) with significant regions of gain that are not currently prognostic per the current ELN guidelines including a case (NK-34) of a KMT2A partial tandem duplication (Supplementary Fig. 6A), a case (Simple K-129) with 1–28 double minute chromosomes (extrachromosomal circular DNA fragments) with amplification of the MYC gene and 3 cases (NK-11, NK-19, 7q-55) with focal gains identified by MPseq of 8q24 involving MYC and/or the nearby long noncoding RNA CCDC26. The partial gain of CCDC26 in 7q-55 was also associated with a cryptic ins(14;8)(q32.2;24.21) resulting in an insertion between BCL11B into the CCDC26. We also observed 2 cases with iAMP21 by MPseq (case 5q-65 and 5q-147) (Supplementary Fig. 6B). Finally, 2 cases (5q/7q-94 and 5q/7q-96) had amplifications of both MECOM (with a rearrangement) and KRAS. Rare or novel SVs were identified in 20 cases (19.0%) including case KAT6Ar-128, which was found to have a cryptic inv(8) resulting in a KAT6A::SORBS3 fusion. A single case (simple K-132) had a ZMYND11::MBTD1 fusion. SVs disrupting NF1 were found in 2 cases (Atypical CK-146, 5q/7q-92), RUNX1 in 2 cases (5q-147, 5q/7q-85), non-GATA2 MECOM rearrangements in 6 cases (7q-56, 5q-148, 5q/7q-84, 5q/7q-94, 5q/7q-96, KMT2A-112) and an ASXL1 rearrangement resulting in a partial deletion of the gene in one case (5q/7q-87). Finally, MPseq characterized each of 6 NUP98 rearrangements revealing KMD5A partner gene in 3 cases and NSD1 partner in 3 cases. Recurrent focal deletions of AML associated genes not appreciated by karyotyping were common (Supplementary Figs 4, 5). Excluding genes located on chromosomes 5 and 7 and TP53, the most frequently deleted gene was NF1 in 9 cases.

In summary, we identified five cases (4.8%) of discordance between MPseq compared to karyotype with FISH when evaluating recurrent AML-associated genetic abnormalities. Greater discordance was identified when comparing karyotype to FISH (14.3%) or karyotype to MPseq (17.1%) individually. These findings contrast previous studies which have reported limited value of FISH in AML when an adequate karyotype is available [7, 8]. Six of our discordant cases included NUP98 rearrangements, which are often cryptic and not evaluated in previous studies. Thus, for specific aberrations that are difficult to detect karyotypically such as TP53 deletions and NUP98 rearrangements, the added value of MPseq is not strong if a comprehensive FISH panel including probes targeting NUP98 (especially in pediatric AML), TP53 and KMT2A are used. Alterations in TP53 are associated with significantly inferior outcomes and treatment responses in AML and biallelic alteration in TP53 AML confers the worse outcomes among CK AMLs [9]. Therefore, accurate characterization of the TP53 locus is important for prognostication and may soon impact clinical management.

Although we demonstrate that MPseq does not currently add additional prognostic value above FISH when compared to a comprehensive AML FISH panel, risk stratification guidelines will likely evolve to include novel abnormalities of prognostic significance detected by NGS. For example, growing evidence suggests NUP98 rearrangements are associated with poor outcome [10], but NUP98 rearrangements have not yet been incorporated into ELN guidelines. Similarly, KMT2A partial tandem duplications have been reported in 10% of AML and MDS and are associated with poor outcome [11]. Other rare abnormalities not incorporated into ELN guidelines identified here include MYC amplification, associated with a CK and poor outcome [12], BCL11B::MYC rearrangements reported as a driver in ambiguous leukemia [13], ZMYND11::MBTD1 fusion [14] and iAMP21 AML [15].

We demonstrate that karyotype analysis when combined with FISH remains a robust laboratory approach in the evaluation of AML. While the use of MPseq did not add significant value above FISH using current ELN guidelines, NGS technologies such as WGS will continue to identify novel high-risk AML subgroups, which will likely enhance future risk stratification guidelines.