The correlation of next-generation sequencing-based genotypic profiles with clinicopathologic characteristics in NPM1-mutated acute myeloid leukemia

The purpose of this study was to analyze the association between next-generation sequencing (NGS) genotypic profiles and conventional clinicopathologic characteristics in patients with acute myeloid leukemia (AML) with NPM1 mutation (NPM1mut). We selected 238 NPM1mut patients with available NGS information on 112 genes related to blood diseases using the χ2 and Mann-Whitney U tests and a multivariable logistic model to analyze the correlation between genomic alterations and clinicopathologic parameters. Compared with the NPM1mut/FLT3-ITD(−) group, the NPM1mut/FLT3-ITD(+) group presented borderline frequent M5 morphology [78/143 (54.5%) vs. 64/95 (67.4%); P = 0.048], higher CD34- and CD7-positive rates (CD34: 20.6% vs. 47.9%, P < 0.001; CD7: 29.9% vs. 61.5%, P < 0.001) and a lack of favorable−/adverse-risk karyotypes (6.4% vs. 0%; P = 0.031). In the entire NPM1mut cohort, 240 NPM1 mutants were identified, of which 10 (10/240, 4.2%) were missense types. When confining the analysis to the 205 cases with NPM1mut insertions/deletions-type and normal karyotype, multivariable logistic analysis showed that FLT3-ITD was positively correlated with CD34 and CD7 expressions (OR = 5.29 [95% CI 2.64–10.60], P < 0.001; OR = 3.47 [95% CI 1.79–6.73], P < 0.001, respectively). Ras-pathway mutations were positively correlated with HLA-DR expression (OR = 4.05 [95% CI 1.70–9.63], P = 0.002), and KRAS mutations were negatively correlated with MPO expression (OR = 0.18 [95% CI 0.05–0.62], P = 0.007). DNMT3A-R882 was positively correlated with CD7 and HLA-DR expressions (OR = 3.59 [95% CI 1.80–7.16], P < 0.001; OR = 13.41 [95% CI 4.56–39.45], P < 0.001, respectively). DNMT3A mutation was negatively correlated with MPO expression (OR = 0.35 [95% CI 1.48–8.38], P = 0.004). TET2/IDH1 mutations were negatively correlated with CD34 and CD7 expressions (OR = 0.26 [95% CI 0.11–0.62], P = 0.002; OR = 0.30 [95% CI 0.14–0.62], P = 0.001, respectively) and positively correlated with MPO expression (OR = 3.52 [95% CI 1.48–8.38], P = 0.004). In conclusion, NPM1mut coexisting mutations in signaling pathways (FLT3-ITD and Ras-signaling pathways) and methylation modifiers (DNMT3A and TET2/IDH1) are linked with the expressions of CD34, CD7, HLA-DR and MPO, thereby providing a mechanistic explanation for the immunophenotypic heterogeneity of this AML entity. Supplementary Information The online version contains supplementary material available at 10.1186/s12885-021-08455-7.


Background
The human NPM1 gene, located on chromosome 5q35.1 and containing 12 exons, encodes a nucleolar phosphoprotein that possesses multiple functions, including chromatin remodeling, ribosome biogenesis, genomic stability, and regulation of tumor suppressors and transcription factors [1][2][3]. Given its important role in biological significance, the functional category of NPM1 belongs to a separate category according to The Cancer Genome Atlas (TCGA) data [4].
NPM1 gene abnormalities are involved in fusion [5], deletion [2] and mutation, among which the mutation is the most largely studied. The incidence of NPM1 mutation (NPM1 mut ) accounts for approximately one-third of the cases of de novo acute myeloid leukemia (AML) and up to~50% of normal karyotype (NK) AML [6,7]. The initial presentations of NPM1 mut AML are characterized by multiple clinicopathologic aspects. For instance, its French-American-British (FAB) morphologies commonly have monocytic differentiation (M4 or M5) [8,9] and are likely to have cup-like nuclei [10]. Immunophenotypically, most NPM1 mut cases show CD34 negativity [11]. According to the analysis of myeloid blast population, nearly half of NPM1 mut patients show an acute promyelocytic leukemia (APL)-like antigen expression feature represented by CD34 (−) /HLA-DR (−) /MPO (str+) [12]. NPM1 mut AML mainly arises in an NK situation and is mutually exclusive with recurrent cytogenetic abnormalities [7,13]. NPM1 mut AML has unique gene expression profiles, especially the overexpression of HOX family members [14].
Because NPM1 mut AML is mainly seen in intermediaterisk cytogenetics, especially in the NK background, we hypothesize that the diversity of leukemic phenotypes depends to a certain extent on the heterogeneity of coexisting gene mutations in this subtype of AML. Whole genome or exosome sequencing revealed an average of 13 mutations in AML [7], indicating the interplay between mutations as an important pathomechanism of leukemic development and overt onset.
In addition, NPM1 mut in association with prognostication is generally described as insertions and/or deletions (indel), which are predominantly characterized by a 4 base-pair insertion in the C-terminus within exon 12 and a resultant frameshift consequence. However, data involving other types of NPM1 mut have scarcely been reported. Moreover, types of NPM1 mut were not specifically designated in AML classification and treatment guidelines [26,27]. The development of large-scale parallel sequencing technology, with its enlargement of higher throughput and wider coverage, is bound to detect more diversified mutational loci and types within the NPM1 gene as well as more concurrent mutations.
In this study, we selected newly diagnosed patients with de novo NPM1 mut AML and evaluated the correlations of clinicopathologic features with next-generation sequencing (NGS)-based genetic alterations in 112 genes related to blood diseases, aiming profoundly to understand the clinicopathological heterogeneity of this AML subtype.

Detection of mutations by NGS and conventional methods
Genomic DNA extraction (Qiagen, Germany), quality control and quantification measurement (Nanodrop Technologies, USA), ultrasonic fragmentation (Covaris, USA), library construction and target enrichment (SureSelect, Agilent Technologies, USA; Illumina, USA) were conducted according to the manufacturer protocols. High-throughput targeted measurement of gene mutations was performed on an Ion torrent PGM™ (Life Technologies) or MiSeq/HiSeq (Illumina) sequencer platform with an average sequencing depth of 800×. The custom-designed panel consisted of 112 potentially mutated genes which are involved in hematological disorders and are related to the following functional categories: signaling pathways, epigenetic regulators, transcription factors, spliceosomes, cohesin complex, tumor suppressors, NPM1 and others. Single nucleotide variants (SNVs) and short fragment indels in protein coding sequences (CDSs) were analyzed by using Ion Reporter™ and Variant Reporter pipelines and annotated referencing the dbSNP, 1000 Genomes, Polyphen-2 and COSMIC databases. NPM1 (exon 12), FLT3-ITD, and potential complex indels in CEBPA (TAD and bZIP domains) were additionally examined by PCR followed by direct sequencing as previously reported [30][31][32].

Statistical analysis
Descriptive statistics are presented as median (range) for non-normally distributed variables and frequency (incidence) for categorical variables. The χ2 test and Mann-Whitney U test were used to calculate the significance of associations between coexisting mutations and clinicopathologic features. To extract independent factors, those with a P-value < 0.15 were included as covariates in the multivariate logistic model using the forward stepwise selection procedure. The results are expressed as odds ratios (ORs) together with 95% confidence intervals (CIs). All calculations were performed applying IBM SPSS v26.0 for Windows. In all analyses, P-values < 0.05 were considered significant. GraphPad Prism 8.4.2, Circos-0.69-9 and R version 4.0.4 were also used for figure plotting.

Association between NPM1 mut coexisting mutations and immunophenotypic markers
Our results showed that the expressions of CD34 and CD7 were significantly associated with FLT3-ITD. Because NPM1 mut AML mostly occurs in the NK context, we hypothesized that diversities in antigen expression in leukemia cells to a certain extent are determined by the Fig. 2 a Number of mutated genes per patient and b the incidence of gene functional categories in NPM1 mut AML (by FLT3-ITD) Fig. 3 Mutational landscape for the relatively common genes mutated in > 5% of the entire NPM1 mut cohort, as well as for the members of spliceosomes and cohesion complex. Each row represents a different gene, and each column represents an individual patient; A colored cell indicates the presence of mutation, and a blank cell indicates wild type; Mutational types are grouped into five classifications as labeled by varying colors; The 27 individual genes are grouped into seven functional categories as listed in Fig. 2b, and the mutational incidence of each gene is listed on the left panel; Clinical data on cytogenetic risk are accordingly displayed on the bottom panel heterogeneity of coexisting mutations. To rule out the influence of abnormal karyotypes on the immunophenotype, as well as in view of the deductively insufficient pathogenicity of NPM1 mut missense mutations, only patients with NK and NPM1 mut indel-types were included for subsequent analysis. A total of 205 NPM1 mut patients fulfilling the above conditions were available for distributional crosstabulation between immunophenotypic markers and coexisting mutations. The significant results from the χ2 test and multivariate analysis are shown in Table 1.
Logistic analysis showed that in the entire NPM1 mut cohort, FLT3-ITD was positively correlated with the expressions of CD34 and CD7 (OR NS not significant, NA not applicable; An OR of > 1 or < 1 means an independently positive or negative association, respectively, for patients with coexisting mutations compared with those with wild-type FLT3-ITD (+) group (OR and P values are detailed in Table 1) and not in the NPM1 mut /FLT3-ITD (−) group. There were no significant correlations between NPM1 mut coexisting mutations and the expression of other antigens.

Discussion
Previous studies regarding the prognostication of NPM1 mut often depicted its mutational type as indels, and there was little information about other types of NPM1 mut . In our cohort of 238 NPM1 mut patients, 240 NPM1 mutant events were identified, among which the vast majority (232, 99.1%) were indel-types. All of these indels derange the tryptophan residues W288 and W290, which are indispensably responsible for NoLS [2]. Ten NPM1 mut missense mutations were clustered in the NPM1 mut /FLT3-ITD (−) group, and none of them disrupted the two loci of NoLS, nor were they involved in NPM1 posttranslational modification sites [3]. Moreover, all except one (9, 90.0%) missense mutation were accompanied by an AML subtype-defining favorable-or adverserisk genetic abnormality, indicating that NPM1 mut missense mutation may be insufficient to drive leukemogenesis and necessitate other well-characterized pathomechanisms. Consequently, the theme of prognosis concerning NPM1 mut AML should be in the context of its indel-types with emphasis or by default, instead of including missense-types of relative rarity and possibly inadequate pathogenicity.
In the present study, NK reached~90% in the entire NPM1 mut cohort with analyzable metaphases and accounted for 84.6% in the NPM1 mut /FLT3-ITD (−) group, similar to the finding of 82.4% in a large sample survey [13]. Moreover, recurrent cytogenetic translocations were uncommon, and FISH did not detect any KMT2A (MLL) translocation or TP53 deletion, implying that the leukemogenesis of frameshift NPM1 mut does not rely on chromosomal abnormalities. Nonetheless, all NPM1 mut indels arose together with coexisting mutations, especially those affecting epigenetic regulators and signaling pathways, which points to the necessity of interactivity of NPM1 mut with other genetic lesions to promote leukemic overt occurrence. The favorable-and adverse-risk abnormal karyotypes were only aggregated in the NPM1 mut /FLT3-ITD (−) group, implying possibly pathogenic independence between FLT3-ITD and those karyotypes in NPM1 mut AML.
Compared with the NPM1 mut /FLT3-ITD (−) group, the NPM1 mut /FLT3-ITD (+) group had higher incidences of CD34 and CD7 expression, similar to other reports [34]. FCM immunophenotyping is not only used in the differential diagnosis of AML but also has prognostic relevance. In terms of an individual immunomarker, CD34 (+) in NPM1 mut AML was associated with a poor prognosis [11,15]. CD123 was only expressed in leukemia and other neoplastic cells but hardly in normal hematopoietic cells [35]. A percentage of CD123 (+) cells in NPM1 mut patients divided by a cutoff of 52% was also reported to predict prognosis [36]. Going forward, the combination of multiple aspects of antigen expression could more potently predict survival. In particular, CD34 (+) /CD38 (−) /CD123 (+) , which represents an LSC phenotype, showed inferior prognosis [16]. In addition, most LSC phenotypes also present cross-lineage, antigen overexpression or asynchronous expression phenomena [16]. In our study, the positive incidences of stem cell antigen CD34 and cross-lineage antigen CD7 expression were higher in the NPM1 mut /FLT3-ITD (+) subset, which may be implied to encompass more LSCs at initial presentation. LSCs are in the relatively silent cell cycle G0 phase and highly express the drug-resistant efflux transporter P-glycoprotein (PGP) or multidrug-resistant protein (MDR1) [11,37]. Chen CY et al. [17] clustered immunophenotyping in 94 NPM1 mut patients and divided them into two categories according to CD34, CD7 and HLA-DR expressions, showing that the prognosis of type-II class characterized by CD34 (+) /HLA-DR (+) / CD7 (+) was significantly poorer versus the type-I class CD34 (−) /CD7 (−) . However, their results might be affected by the biased distribution of concurrent FLT3-ITD, which has a positive correlation with CD34 and CD7 expressions. Because of the limited number of cases, it was not clear whether the differential effect of class I and II features on prognosis was independent of FLT3-ITD, although a stratified analysis had been carried out.
We investigated the relationship between NPM1 mut coexisting mutations and immunophenotypic markers. In general, there was a distributional association of signaling and methylating mutations with CD34, CD7, HLA-DR and MPO expressions. The regulatory effect of Ras-pathway mutations on the expression of these antigens was only found in the NPM1 mut /FLT3-ITD (−) group but not in the NPM1 mut /FLT3-ITD (+) group, partly owing to the reciprocal exclusivity of FLT3-ITD with Ras-pathway mutations. DNMT3A mutation was positively correlated with the expressions of CD34, CD7 and HLA-DR in both genotypic groups, while TET2/IDH1 mutations were negatively correlated with those antigens specifically in the NPM1 mut /FLT3-ITD (+) group. In contrast, DNMT3A mutation was negatively correlated with MPO expression, while TET2/IDH1 mutations were positively correlated with MPO expression. These results suggested that DNMT3A and TET2/IDH1 mutations might play different roles in regulating the expression of these immunophenotypic markers.
In the NPM1 mut /FLT3-ITD (−) group, Ras-pathway mutations and DNMT3A-R882 were positively correlated with the expression of the monocyte marker HLA-DR and negatively correlated with the myeloid marker MPO, which is linked to the FAB morphology of monocytic differentiation (M4/M5) or granulocytic differentiation (M2). Comparatively, in the NPM1 mut /FLT3-ITD (+) group, although TET2/IDH1 mutations were negatively correlated with HLA-DR expression, the more commonly coexisting DNMT3A-R882, which was positively correlated with HLA-DR expression, might take precedence and be accountable for a more frequent M4/M5 morphology in this genotypic group.
Mason EF et al. [12] analyzed myeloid blast populations excluding monocytic differentiation in NPM1 mut patients. Nearly half of the cases (48%) had an APL-like phenotype represented by CD34 (−) /HLA-DR (−) /MPO (str+) , which could predict the presence of TET2 or IDH1/2 mutations, a result in line with our findings. Moreover, the authors demonstrated the APL-like phenotype beneficially impacted RFS and OS, and its combination with coexisting TET2 or IDH1/2 mutations was more explicit to refine prognostic subgroups. Our present study extended those findings. We additionally showed an independent negative association of Ras-pathway mutations with the APL-like phenotype only in the NPM1 mut /FLT3-ITD (−) group. Additionally, we showed a negative association of DNMT3A-R882 with this phenotype only in the NPM1 mut /FLT3-ITD (+) genotypic background. These results suggested that the interplay of NPM1 mut coexisting genetic lesions might jointly determine the trend of antigen expression, partly explaining the immunophenotypic heterogeneity in NPM1 mut AML.

Conclusions
In summary, NPM1 mut missense mutations may be of leukemogenic insufficiency and largely rely on other well-defined pathomechanisms in the development of overt leukemia. The correlation of coexisting mutations in signaling pathways and methylation modifiers with antigen expression (represented by CD34, CD7, HLA-DR and MPO) may partly explain the immunophenotypic diversity in NPM1 mut AML. Comprehensively evaluating the FCM immunophenotype and NGS landscape of genetic lesions allows us to gain insight into the clinicopathological heterogeneity of this distinct AML entity.