Myeloproliferative neoplasms (MPN) are driven by specific mutations in one of the three genes – JAK2, MPL or CALR (1). These mutations cause uncontrolled activation of the JAK-STAT signaling pathway leading to the release of secondary mediators of pro-survival and proliferative nature (2).

The rapid development of sequencing techniques in recent years allowed for better characterization of molecular profiles in patients with hematological malignancies and significantly contributed to an understanding of disease biology. It is accepted that besides driver mutations, the molecular profile of MPN comprises variants in a variety of genes (3, 4). The presence and allelic burden of those mutations are suspected to impact the disease’s clinical course. However, in MPNs, the chronic nature of the disease hinders the research aimed at understanding the significance of molecular changes, contrary to entities of acute nature. Mutations in only a fraction of genes have been recently proposed as potentially influencing disease risk and included in mutation-enhanced international prognostic systems (MIPSS) for polycythemia vera (PV) and essential thrombocythemia (ET) (5).

In this study, we aim to assess the dynamics of the molecular profile in PV and ET by investigating the occurrence and allelic burden of both driver and passenger mutations at two-time points. We hypothesize that the molecular profile of patients with PV and ET evolves with time, accumulating mutations in expanding malignant clones. The chronic nature of those diseases implies the need to assess the risk multiple times. We believe studies on the topic may help predict patient outcomes in advance, giving the time to react and treat accordingly.

Study participants have been selected from the population of patients of the Outpatient Ward, Department of Hematology and Transplantology, Medical University of Gdansk, Poland. Initially, all alive patients diagnosed with PV or ET were identified. Next, individuals for whom the diagnosis was made at least five years prior to the study were selected. Among those patients, we identified individuals from whom DNA samples of sufficient quality and quantity were available from the time of diagnosis. Finally, 49 patients consented to participate in the study.

Diagnoses were verified with MPN 2016 WHO criteria (6). Two patients with apparent features of myeloproliferative neoplasms but undetected driver mutation were labeled triple-negative (TN). Each patient’s documentation was analyzed to collect data regarding treatment, thrombotic complications, and disease progression. The general characteristics of the study group are presented in Table 1 .

A peripheral blood sample was collected from each participant for molecular testing and comparison with corresponding archival samples. The median time between historical sample collection and a present sample was 104 months (range 78-141 months)

Genomic DNA was extracted from peripheral blood samples using QIAamp DNA mini kit (Qiagen) according to the manufacturer’s instructions. DNA was quantified by spectrophotometric method and stored at –20°C for further analysis.

Next-generation sequencing was performed using the Archer VariantPlex Core Myeloid kit (ArcherDX), Mid Output Kit (300-cycles), and MiniSeq (Illumina). It allowed performing a comprehensive analysis of 37 genes: ABL1, ANKRD26, ASXL1, BCOR, BRAF, CALR, CBL, CABPA, CSF3R, DDX41, DNMT3A, ETNK1, ETV6, EZH2, FLT3, GATA1, GATA2, IDH1, IDH2, JAK2, KIT, KRAS, MPL, NPM1, NRAS, PHF6, PTPN11, RUNX1, SETBP1, SF3B1, SRSF2, STAG2, TET2, TP53, U2AF1, WT1, ZRSR2. The results were analyzed using Archer Analysis v.6.0.3.2 software (ArcherDX), and 2,7% allele frequency (VAF) as a cutoff was applied.

Collected data was used to stratify patients into the risk groups during sample collection. Conventional risk stratification was made according to the revised IPSET-T system in ET and based on age and thrombotic complications in PV (7, 8). Acquired sequencing data was used to attribute patients into the low- intermediate- or high-risk groups according to MIPSS-ET or MIPSS-PV (5). Since cytogenetic data were incomplete in the study population, all PV patients were considered to have normal karyotypes.

An additional exploratory risk stratification, including 5- and 10-year event-free survival (EFS), risk of secondary myelofibrosis (sMF) or acute myeloid leukemia (AML) were assigned based on prognostication model developed by Grinfeld et al. (4). This stratification was made to assess the utility of the proposed MPN Personalized Risk calculator (available online at https://www.sanger.ac.uk/science/tools/progmod/progmod/). For this evaluation, all patients were considered unknown karyotypes.

Statistical analysis was performed using the computer program STATISTICA v.13.3. Data was statistically described in terms of mean or median and range. Comparison between the two groups was made using Mann–Whitney U test for continuous variables and the Chi-square test for categorical variables. P values were considered significant if less than 0.05.

In addition to known driver mutations, NGS-based studies allow for detecting numerous variants in various genes. Those findings must be considered with caution because not all variants are confirmed to be pathogenic. Since the molecular landscape of MPNs is undiscovered area, we performed a thorough literature search to look for associations of detected variants with pathogenicity. The two most frequently detected variants - CBL p.Asp460del and RUNX1 p.Glu422Ala - were described in various neoplasms, including hematologic, with no confirmed pathogenicity (9, 10). Additionally, the total number of mutations detected in 2^nd sample did not directly impact the disease’s phenotype, indicating the need for qualitative rather than a quantitative approach. On the other hand, our analysis allowed us to detect significant molecular changes in patients otherwise labeled as TN – one with TP53 p.Arg248Leu – a widely described, pathogenic hotspot gain of function mutation– and RUNX1 p.Arg423Gly variant - undescribed previously, here associated with apparent features of MPN (11). Additionally, we found two coexisting driver mutations in two patients – JAK2 p.Val617Phe along with MPL p.Trp515Leu and JAK2 p.Val617Phe with non-canonical MPL p.Ser505Asn. In both cases MPL variants had higher VAF than JAK2 p.Val617Phe, and the latter MPL variant is confirmed pathogenic (4, 12). In another patient, we found a non-canonical JAK2 p.Lys607Asn variant co-occurring with JAK2 p.Val617Phe, which was described in AML patients (13). In a routine workup done with the PCR method, identifying true driver mutation in such patients is challenging, underlying the importance of complete, thorough molecular evaluation in MPN patients in the future.

Unsurprisingly, the current study’s two most frequently mutated genes were TET2 and DNMT3A. Those mutations are widely described in hematologic neoplasms with inconsistent conclusions regarding their pathogenicity (14–17). Those and other mutations are also associated with clonal hematopoiesis of indeterminate potential (CHIP) - an effect of accumulation of specified mutations during life without a clear connection to morbidity (18). On the other hand, there are reports of increased genetic instability in TET2- or DNMT3A-mutated patients, possibly triggering further clonal hematopoietic expansion and contributing to acquiring additional HMR mutations (19–23). By analyzing the dynamics of those mutations in our study, we show that TET2-mutated patients are bearing this mutation from the diagnosis and acquisition of DNMT3A occurs later during the disease. While we did not find the exact correlations with those mutations with clinical phenotype, we confirmed that TET2 and DNMT3A mutations detected in the first sample resulted in a significantly more mutations detected in the second sample.

In PV and ET, conventional risk assessment is aimed at predicting the risk of thrombosis - one of the factors significantly influencing the mortality and morbidity of patients diagnosed with MPN (7, 8). Here, thrombotic complications occurring after diagnosis were recorded in six patients, with the majority classified as non-high-risk. In older patients, accumulation of cardiovascular risk factors, age and history of thrombosis are helpful variables for predicting the risk of thrombosis. However, in younger patients, otherwise classified as low- or very-low-risk, data from consecutive genetic analyses may help to predict thrombotic complications correctly. In our study, two young patients experienced thrombotic complications, despite not harboring conventionally accepted risk factors. One had a significantly increasing JAK2 p.Val617Phe allele burden, a process which was proven to be associated with the risk of thrombosis by Soudet et al (24). The second patient harbored ZRSR2 p.Ser447_Arg448dup variant, which is likely benign when detected in MDS, but connected with splanchnic vein thrombosis in MPN (25, 26).

Wider accessibility to modern diagnostic methods initiated the search for different variables predicting the outcomes of patients with MPNs. In 2019 Tefferi et al. proposed a new prognostication system, incorporating results from genetic analyses, for ET and PV - MIPSS-ET and MIPSS-PV, respectively (5). However, the authors acknowledge the need for further evaluation of these systems. In our study, we applied the MIPSS scoring system based on the results of two consecutive samples from each patient. In as many as five ET patients, detected mutations allowed for the increase in MIPSS score, placing those patients in the high-risk group and suggesting the possible utility of this system in assessing the risk in a dynamic fashion. Additionally, we show the utility of MIPSS-ET in predicting the risk of sMF. On the other hand, one young PV patient with a detected pathogenic TP53 mutation remained in the low-risk group according to MIPSS-PV. The appearance of high-risk variants at follow-up underlines the utility of sequential molecular evaluation of patients with MPNs.

As an exploratory objective, we used the MPN Prognostic Calculator created based on the work of Grinfeld et al. (4). Authors underline that the tool is rather a proposition and is not yet validated. This analysis showed significantly better prognosis across all evaluated scores for patients with PV compared to ET. It failed to distinguish patients with a high risk of developing sMF. Whether this is the result of the construction of the tool itself or the generalization of input genetic data, remains to be further investigated. Regarding MPNs, we encourage reporting and analyzing the impact of specific variants rather than stratifying the risk by the presence of mutation only.

Fibrotic or AML progression remains the most therapeutically challenging dilemma in PV and ET patients. Here, nine patients developed sMF during their disease. None of those patients displayed features of fibrosis on the initial bone marrow sample. However, the evaluation of the characteristics of the megakaryocytes in those samples might not be in accordance with the current knowledge, hence it cannot be ruled out that a proportion of those patients had an evolution from prefibrotic to an overt stage of MF. Nevertheless, we revealed that patients with fibrotic progression had a significantly higher mutational burden of their driver mutation at samples collected during the transformation compared to baseline and to patients without sMF. Additionally, those patients harbored specific variants previously attributed to fibrosis, including ASXL1 p.Gly646TrpfsTer12, RUNX1 p.Leu56Ser, ZRSR2 p.Arg169Ter and U2AF1 p.Gln157Pro (4, 16, 27–33). Apart from the abovementioned variants, other mutations were detected only in those patients who developed fibrosis but their importance requires further investigation ( Supplementary Table 2 ). While the sample size is relatively small, those findings support the idea of repeated monitoring, allowing for the detection of increasing VAF and pathogenic variants before fibrotic progression while being less unpleasant for the patient and more objective than trephine biopsy.

In PV and ET, the interest in the correlation between treatment and mutational landscape is growing. In our study, we performed analyses on two homogenous group of patients – those treated with HU monotherapy and patients that endured without any treatment during the study. Based on our results, the potential leukemogenic effect of prolonged exposure to HU remains questionable, since those patients did not present features of expanding mutational landscape. It is unclear whether this result is an effect of treatment efficacy or rather phenotypically stable disease, not requiring treatment change. In treatment-free patients, detecting non-pathogenic variants and stable VAFs of driver mutations correlated with genetically stable disease and clinical picture. The question remains whether their stable molecular profile was the phenotype of the disease itself or, that by not introducing the treatment we did not expose the malignant clone to further genetic instability (34, 35).

In conclusion, we emphasize the need for careful interpretation of molecular findings, particularly the assessment of the pathogenicity of specific variants. Extensive studies evaluating the molecular profile of patients and confronting it with the diseases’ phenotype are critical. However, to date, only few studies evaluated the molecular findings in a dynamic fashion. Patients with MPNs must be evaluated more than once throughout their disease and it is insufficient to assess the risk only once, especially in young patients. Based on numerous studies across all fields of hematology, it is evident that diseases’ phenotype is driven and modulated by various molecular changes in addition to basic factors, such as age and history of thrombosis. This field is open for discovery, and authors of the current study believe that a larger portion of this knowledge remains unknown, including transcriptomics, epigenetic modifications, interaction with the microenvironment and paracrine activity of both malignant and non-malignant cells.

The data can be accessed by the following link: https://datadryad.org/stash/share/MSNwkQvXlqHqOWyTvAiizUBZBS6UBJZaxqSqwLXtHaM.

The studies involving humans were approved by Bioethics Committee for Scientific Research, Medical University of Gdansk. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

PS – conducted the research, recruited patients, analysed the data and written the manuscript. BW and MŻ – performed NGS analysis and reviewed the manuscript. AL – performed DNA isolation and reviewed the manuscript. MB – supervised the research, coordinated the work, helped writing and reviewed the manuscript.