2.1 Patients

We included patients who visited Juntendo University Hospital and its affiliated institutions between January 2010 and December 2019 and were suspected of having myeloproliferative neoplasms (MPNs). Screening for chronic myeloid leukemia (CML) was performed by peripheral blood fluorescence in situ hybridization, bone marrow G-band analysis, or bone marrow real-time polymerase chain reaction (PCR). Congenital erythrocytosis was screened based on the presence of a family history. This study was approved by the ethics committee of the School of Medicine, Juntendo University (IRB#M12-0895) and was conducted in accordance with the 1975 Helsinki Declaration. Written informed consent was obtained prior to the use of the samples and the collection of clinical records.

2.2 Driver mutation and CREB3L1 expression analysis

JAK2 V617F, MPL exon 10, and CALR exon 9 mutations were analyzed as previously described.^8-10 Briefly, detection of JAK2 V617F was first performed by ABC-PCR, and in negative patients, allele-specific quantitative PCR was further conducted to confirm negativity. For detection of MPL exon 9 and CALR exon 9 mutations, two sets of primers (common forward: AGT AGG GGC TGG CTG GAT, W515L-specific forward: GGC CTG CTG CTG CTG AAG TT, W515K-specific reverse: (FAM)-ACC TGT AGT GTG CAG GAA ACT TCT T, and common reverse: (FAM)-GGT CAC AGA GCG AAC CAA GA) and a set of primers (forward: (FAM)-TGG TCC TGG TCC TGA TGT C and reverse: GGA ACA AAA CCA AAA TCC AC) were used to amplify the DNA fragment containing MPL exon 10 and CALR exon 9, respectively, and the amplicons were subsequently applied to capillary electrophoresis. For patients who were negative for these mutations, JAK2 exon 12 mutations were further screened using Sanger sequencing.¹¹ CREB3L1 mRNA levels in peripheral blood platelets (PLT) were measured by SYBR green-based reverse transcription-quantitative PCR with a set of primers (forward: GGA GAA TGC CAA CAG GAC, and reverse: ACC AGA ACA AAG CAC AAG G) as previously described.¹²

2.3 Data collection

Medical history, smoking habits, and weight at the time of first visit were collected from medical records. Heavy drinking was defined as daily consumption of ≥40 g of ethanol. Obesity was defined as body mass index (BMI) of ≥25 kg/m². Splenic enlargement was evaluated by palpation, abdominal ultrasonography, or abdominal computed tomography, and splenomegaly was diagnosed by the attending physician. End-stage renal disease patients receiving regular EPO injections were included in the study but were excluded from the EPO analysis.

2.4 Statistical analysis

Comparisons between the NNE and PV groups were performed using Fisher's exact test for categorical variables and the Mann–Whitney U test for continuous variables. All statistical analyses were performed using the R program (R Foundation for Statistical Computing). Statistical significance was considered as p < 0.05.

3.1 Defining acquired erythrocytosis

A total of 1979 individuals with suspected MPNs were screened. Among them, 121 patients fulfilled the following conditions; (1) hemoglobin (Hb) levels above the diagnostic criteria for PV (males, ≥165; females, ≥160 g/L), (2) Philadelphia chromosome-negative, and (3) available bone marrow specimens (Figure 1). The bone marrow specimens were centrally reviewed for the evaluation of the following: cellularity, myeloid/erythroid (M/E) ratio, megakaryocyte count, megakaryocyte nuclear atypia, progenitor cell rate, gelatinous transformation, and fibrosis grade. After ruling out CML and congenital erythrocytosis, a thorough analysis of bone marrow features (Figure 2A–H) and driver mutation status (described below) were carried out, resulting in diagnosis of 67 (55.4%) patients with NNE, 47 (38.8%) patients with PV, 3 (2.5%) patients with essential thrombocythemia (ET), 3 (2.5%) patients with prefibrotic primary myelofibrosis (pre-PMF), and 1 (0.8%) patient with atypical CML (aCML) (Figure 1).

Compared with PV bone marrow, NNE bone marrow exhibited hypo- to normo-cellularity, lower M/E ratio, fewer megakaryocytes, and less nuclear abnormalities (Figure 2A–D,I–L). In the NNE group, a quarter of patients exhibited gelatinous transformation characterized by atrophy of adipose tissue, focal loss of bone marrow hematopoietic cells, and extracellular deposition of a gelatin-like substance (Figure 2M).¹³ Despite elevated Hb levels, three ET and three pre-PMF cases were identified based on bone marrow features such as an increase in megakaryocytes and nuclear atypia. One patient was suspected to have aCML because of a hyperplastic bone marrow with trilineage dysplasia (Figure 2H).

In the driver mutation analysis, 42 and 5 of the 47 PV patients had JAK2 V617F and exon 12 mutation, respectively (Figure 1). Of the three ET patients, JAK2 V617F and CALR ins five mutations were found in two and one patients, respectively (Figure 1). All three pre-PMF patients harbored the JAK2 V617F mutation. All NNE and aCML patients were negative for JAK2 mutations (Figure 1). None of the patients had JAK2 mutation-negative PV.

3.2 Distinctive clinical characteristics of NNE

Comparison of the clinical features between 67 NNE and 47 PV patients showed the following results. NNE patients were significantly younger than PV patients (median age: 57 vs. 65 years, p < 0.01; Figure 3A; Table S1). In addition, the NNE group showed a male predominance (91%) which reached statistical significance when compared with the PV group (male/total: 61/67 vs. 27/47, p < 0.001; Figure 3B; Table S1). Splenomegaly was less frequent in the NNE group than in the PV group (10.0% vs. 32.5%). Thrombotic events, of which all were arterial in both groups, were less frequent in the NNE group than in the PV group (8 [11.9%] vs. 10 [21.3%], Figure 3C) but the difference was not statistically significant (p = 0.11). With respect to ischemic stroke, the prevalence was higher in the NNE group (8/67) than in the general population.¹⁴

Hemoglobin levels at the time of diagnosis were significantly higher in the NNE group than in the PV group, presumably reflecting the male predominance in the NNE group (Figure 3D; Table S1). White blood cell count, PLT count, and lactate dehydrogenase levels were significantly higher in the PV group compared to the NNE group (p < 0.001 for all of three categories, Figure 3E–G; Table S1). Platelet counts in all NNE patients were below the level required for a diagnosis of ET (450 × 10⁹/L), while the majority of PV patients showed levels higher than the diagnostic cutoff. Iron and ferritin levels were significantly lower in the PV group (p < 0.001; Table S1), which may be due to the neoplastic proliferation of erythroid cells. EPO was also significantly lower in the PV group (p < 0.001, Figure 3H; Table S1), and all but two patients in the PV group exhibited EPO levels below the lower range of normal (<4.2 IU/L). Notably, 18 (40.8%) of the 44 NNE patients with available EPO data showed low EPO levels, which is a trait that has not been previously reported in NNE.^{15, 16}

3.3 Lifestyle-related factors and pre-existing clinical conditions in NNE

Comparison of lifestyle habits and pre-existing clinical conditions are shown in Figure 4. Compared to the PV group, the NNE group showed a higher frequency of smoking (70.5% vs. 33.3%, p < 0.001), alcohol consumption (43.5% vs. 11.9%, p < 0.001), and type 2 diabetes mellitus (21.2% vs. 6.4%, p < 0.05). There was also a trend for higher rates of obesity (41.3% vs. 19.4%) in the NNE group. Notably, all seven patients with pronounced obesity demonstrating a BMI of ≥30 kg/m² all belonged to the NNE group.

The frequencies of hypertension, dyslipidemia, and cardiovascular disease were similar between the NNE and PV groups (Figure 4). Cardiovascular diseases can cause erythrocytosis through chronic hypoxia, but no obvious increase in EPO levels was observed in patients with NNE. Diabetes mellitus, respiratory diseases, liver dysfunction, renal dysfunction, and a history of cancer were more prevalent in the NNE group (Figure 4). In both the NNE and PV groups, all patients with diabetes mellitus had type 2 diabetes. Liver dysfunction in the NNE group was mostly due to alcoholic hepatitis or fatty liver. Respiratory diseases included sleep apnea syndrome (SAS) (n = 3), asthma (n = 3), lung cancer (n = 2), chronic obstructive pulmonary disease (COPD) (n = 1), interstitial pneumonia (IP) (n = 1), and drug-induced pneumonia (n = 1). Some of these clinical conditions may have independently induced erythrocytosis in a proportion of NNE patients. None of the above mentioned factors were found in four NNE patients (Figure 4), suggesting idiopathic erythrocytosis in these patients.

3.4 Clinical features associated with EPO and bone marrow features in NNE

Given that NNE includes patients with erythrocytosis caused by diverse mechanisms, we attempted to classify NNE according to biological features that might be distinct to a specific disease entity. We first studied EPO levels, which have been reported to be increased in reactive erythrocytosis.¹ However, there was only one patient in whom the EPO was above the normal range (44.3 IU/L, normal range: 4.2–23.7 IU/L), and this patient was a heavy smoker and also obese. All eight patients highly suspected of reactive erythrocytosis due to traits such as renal artery stenosis (n = 1), IP and lung cancer (n = 1), COPD (n = 1), chronic heart failure (n = 1), SAS (n = 1), heavy smoking (n = 5), and severe obesity (n = 1) had normal EPO levels. These findings were consistent with some reports demonstrating increased EPO levels only in a subset of reactive erythrocytosis cases.^{15, 17}

Therefore, we compared the clinical parameters and bone marrow features between suspected NNE patients with reactive erythrocytosis who had EPO levels within the upper 25% limit (n = 11) and those with levels within the lower 25% limit (n = 11). No significant differences in clinical parameters were observed between the two groups (Figure 5; Table S2). However, the frequency of heavy drinking was significantly higher in the low EPO subgroup (81.8% vs. 27.3%, p < 0.05; Figure 6). Cases with cardiac disease, renal disease, and pronounced obesity with a BMI of ≥30 kg/m² were not found in the low EPO subgroup.

Considering that PV presents with unique bone marrow features, we hypothesized that bone marrow morphology may be of aid in classification of distinct entities within NNE. Thus, we classified the NNE patients into three groups based on bone marrow cellularity: hypocellular (n = 30), normocellular (n = 32), and hypercellular (n = 5). However, comparison of clinical parameters between the first two groups that included most NNE patients showed no significant difference in any of the parameters (Figures S1 and S2). These results showed that patients with NNE could not be further classified according to bone marrow pathology.

3.5 CREB3L1mRNA in PLT is detected in PV patients but not in NNE patients

We recently reported that CREB3L1 expression in PLT mRNA is a reliable diagnostic marker for MPN.¹² Thus, we examined CREB3L1mRNA in PLT obtained from 12 NNE patients and 12 PV patients (9 patients with V617F and 3 patients with exon 12 mutations). CREB3L1mRNA expression was detected in all PV patients but in none of the NNE patients studied (Figure 7), confirming the diagnosis based on bone marrow pathology.