Increased Dkk‐1 plasma levels may discriminate disease subtypes in myeloproliferative neoplasms

Abstract Alterations in the bone marrow niche induced by abnormal production of cytokines and other soluble factors have been associated with disease progression in classical BCR‐ABL1 negative myeloproliferative neoplasms (MPN). Variations in circulating proteins might reflect local disease processes and plasma proteome profiling could serve to identify possible diagnostic and prognostic biomarkers. We employed a human cytokine array to screen for 105 distinct analytes in pooled plasma samples obtained from untreated young MPN patients (<35 years) with different clinical phenotypes and driver mutations, as well as from healthy individuals. Among molecules that exhibited significantly increased levels in MPN patients versus controls, the top of the list was represented by Dickkopf‐related protein 1 (Dkk‐1), which also showed the highest potential for discrimination between MPN subtypes. In the next step, a quantitative ELISA was used to measure plasma Dkk‐1 levels in 30 young‐onset MPN—10 essential thrombocythemia (ET), 10 polycythemia vera (PV), 10 pre‐fibrotic primary myelofibrosis (pre‐PMF)—and 10 controls. The results suggested that plasma Dkk‐1 levels could differentiate ET from pre‐PMF, in JAK2 V617F‐positive as well as in CALR‐positive patients, and also ET from PV in JAK2 V617F‐positive patients.


| INTRODUCTION
Philadelphia-negative classical myeloproliferative neoplasms (MPNs) represented by polycythemia vera (PV), essential thrombocythemia (ET) and primary myelofibrosis (PMF) are clonal disorders that emerge in the haematopoietic stem cell compartment (HSC). They are defined by an excessive production of functional mature blood cells from one or more myeloid lineages. 1 Three classes of somatic activating driver mutations localized within specific exons of Janus Kinase 2 (JAK2), thrombopoietin receptor (MPL/TPOR) and calreticulin (CALR) genes are key players in the majority of MPNs 2 leading to an improper activation of JAK2 signalling. 3 The single HSC that harbours the oncogenic mutation-MPN stem cell-has to acquire a selective advantage over its normal  5 and/or interleukin-1b (IL-1b). 6 The alterations induced in the bone marrow niche, namely neural damage, angiogenesis, expansion of abnormal osteolineage cells and reticulin fibrosis, have been associated with MPN progression. 6 Both malignant and non-malignant cells of haematopoietic system are involved in the abnormal cytokine production via activation of JAK-STAT3 pathway. 7 Plasma/serum profiles of various cytokines and other soluble proteins might reflect local disease processes and could serve as noninvasive diagnostic or prognostic tools in patients with myeloproliferative disorders. 8 In this study, we used a human cytokine array to measure plasma levels of multiple proteins in young MPN patients (<35 years) with different clinical phenotypes and driver mutations. Our aim was to identify possible biomarkers for discriminating between different MPN subtypes.

| Patients and samples
The study was approved by the local ethics committee, and research was conducted according to the World Medical Association Declaration of Helsinki. We used platelet-poor EDTA plasma samples and genomic DNA (gDNA) isolated from peripheral blood granulocytes previously obtained by standard procedure from 30 MPN patients and stored in the biobank of Stefan S Nicolau Institute of Virology, Romania. Patients were referred from the Hematology Services of "Coltea" and "Fundeni" Clinic Romanian Hospitals to the Institute of Virology for detection of JAK2 V617 F mutation by allele-specific PCR. In JAK2 V617F-negative MPN patients, we analysed CALR exon 9 mutations using gDNA samples and a previously described protocol for direct Sanger sequencing. 9 Clinical data were provided for each case. All the patients gave an informed consent at the time of blood collection. For comparing results, we also included 10 age-matched healthy controls.

V617F mutational load quantification
For the quantification of JAK2 V617F mutation in gDNA from granulocytes, we used the ipsogen JAK2 MutaQuant Kit (Qiagen) with a measured limit of detection of 0.1% following the protocol recommended by the manufacturer for Applied Biosystems 7300 Real-Time PCR instrument (Thermo Fisher Scientific). The copy number of mutant and wild-type JAK2 alleles was calculated from the standard curves using SDS software version 1.4.1 (Thermo Fisher Scientific).
JAK2 V617F allele burden was expressed as the percentage of JAK2 V617F copy number relative to the total number of JAK2 copies (wild-type plus mutant).

| Statistical analysis
Statistical tests were performed with GraphPad Prism 6.01 for Windows. To compare the fold change in analyte expression in MPN patients versus control, one-way ANOVA test was employed for multiple groups and Student's unpaired t test for two groups.
Differences in Dkk-1 levels among MPN subtypes and controls were analysed by nonparametric Kruskal-Wallis test H followed by Mann-Whitney U test for one-to-one comparison. Also, Kruskal-Wallis test H was used to assess the differences in JAK2 V617F allele burden among ET, PV and pre-PMF patients. Correlation between plasma levels of Dkk-1 and JAK2 V617F allele burden was evaluated using the nonparametric Spearman rank correlation coefficient. P values <.05 were considered statistically significant.

| Patient characteristics and plasma pool preparation
To avoid bias from co-morbidities and treatment, we selected 30 untreated patients that received a diagnosis of MPN at young age (<35 years): 10 ET, 10 PV and 10 pre-PMF patients. JAK2 V617F mutation was present in 20 patients (66.6%). The other 10 patients harboured CALR mutations, classified in type 1 (del52, n = 6) and type 2 (ins5, n = 4) mutations. 9 The diagnosis of each MPN subtype was established according to the World Health Organization (WHO) 2016 criteria. 10 For the initial proteomic screening, we prepared 5 pools of plasma samples from patients sharing the same MPN subtype and driver mutation and 1 pool from controls (10 age-matched subjects with normal blood counts). Clinical characteristics, haematological parameters, type of driver mutation and also JAK2 V617F allele burden of MPN patients corresponding to each pool are summarized in Table 1.

| Relevant plasma proteins detected in MPN patients by the human cytokine array
The preliminary multianalyte screening for the differentially expressed cytokines among various MPN patients and controls was performed in 6 plasma pools (5 from MPN patients and 1 from controls), and 6 individual plasma specimens from each pool, using 12 array membranes.
Examples of array membranes displaying visual differences in signal intensity are shown in Figure 1A.   Among analytes that exhibited a significant fold increase in MPN patients versus controls, Dkk-1 showed the highest potential for discrimination between MPN subtypes and was selected for further study (P = .0031, one-way ANOVA, Figure 1C). develop complications, such as progression to overt PMF and leukaemic transformation, as well as a higher mortality rate. Therefore, the differentiation between the two MPN phenotypes is very useful in practice. At the moment, this is based mainly on bone marrow histology. 12,13 Secondly, we found that circulating levels of Dkk-1 might distinguish also ET from PV in JAK2 V617F-positive patients. In our group of young MPN patients, prodromal or "masked" PV (mPV) predominated. As shown previously, the distinction between mPV and ET exhibits both prognostic and therapeutic relevance, to address the higher incidence of thrombosis in PV versus ET and the need of phlebotomies in PV for obtaining a haematocrit less than 45%. 14 Even when applying the new WHO 2016 thresholds for haemoglobin and haematocrit, it was found that in 14% of cases the PV, the diagnosis was missed.

| Plasma levels of Dkk-1 in MPN subtypes
Thus, serum erythropoietin levels and bone marrow biopsy are mandatory for establishing a diagnosis of PV in these cases. 15 The Dkk-1 plays an important role during embryogenesis and it was involved in bone homoeostasis in adulthood. 18 From a pathological perspective, it was suggested that Dkk-1 released by platelets and endothelial cells might be involved in the inflammatory reactions within the atherosclerotic plaques. 11 As a potent inhibitor of canonical Wnt signalling-a pathway aberrantly activated in many types of cancer-Dkk-1 was initially considered a tumour suppressor 19 However, many studies have found that Dkk-1 is overexpressed in malignant tissue and/or serum of patients with various solid cancers (breast, cervical, endometrial, oesophageal, gastric, colorectal, liver, urothelial, lung, etc) being associated with tumour growth, angiogenesis and unfavourable prognosis. 19 Also, increased Dkk-1 levels in bone marrow plasma and blood plasma of multiple myeloma patients were correlated with the presence of osteolytic lesions. 20 Interestingly, in a recent study, it was shown that exosomes produced by acute myeloid leukaemia cell lines and patient-derived blasts stimulated the expression of Dkk-1 in bone marrow stromal cells that contributed to the transformation of normal haematopoietic niche into a leukaemia cell supportive microenvironment. 21 In our study, significantly higher plasma Dkk-1 levels in PV and pre-PMF compared to ET patients could be related to bone marrow A particular finding of this study was that JAK2 V617F allele burden did not show statistically significant differences among MPN subtypes. These could be a feature of the analysed group consisting of young MPN patients, as in a recent study, it was reported that adolescents and young adults with PV and PMF displayed significantly lower JAK2 V617F allele burdens compared to older patients having a similar disease phenotype. 25 Accordingly, Dkk-1 levels in plasma were not correlated with the mutated allele load suggesting that they might reflect early changes in bone marrow microenvironment rather than a direct relation with clonal burden.
Thus, plasma DKK-1 levels could represent a potential noninvasive biomarker for differentiation between MPN subtypes, especially in young patients that often display an overlap in clinical onset of disease.

ACKNOWLEDG EMENTS
We gratefully acknowledge the funding from the project Competi-

CONFLI CT OF INTEREST
The authors confirm that there are no conflict of interests.