The classical myeloproliferative neoplasms (MPNs), namely polycythaemia vera (PV), essential thrombocythaemia (ET) and primary myelofibrosis (PMF), have overlapping clinical, morphological and molecular features. Primary myelofibrosis can be classified further into a pre-fibrotic and overt fibrotic stage. Cases with features of a MPN, which do not meet all criteria for a specific MPN entity, are classified as MPN unclassifiable.¹

The canonical MPN driver gene variants include specific variants in Janus kinase 2 (JAK2), calreticulin (CALR) and the thrombopoietin receptor gene (MPL).² Janus kinase 2 p.V617F accounts for 97% of PV and 55% of ET and PMF cases.³ Furthermore, JAK2 exon 12 variants are reported in the majority of PV patients who are negative for the JAK2 p.V617F variant.⁴ Calreticulin variants account for 25% of ET and 30% of PMF cases.³ Calreticulin variants can be subtyped as type 1 (52 bp deletion; c.1092_1143del52;p.Leu367Thrfs*46) and type 2 (5 bp insertion; c.1154_1155insTTGTC) CALR variants.⁵ In ET, type 1 and 2 CALR variants are more evenly distributed (51% for type 1 vs 39% for type 2) than in PMF (70% for type 1 vs 13% for type 2).⁶ Thrombopoietin receptor gene variants account for 5% of ET and 8% of PMF cases.³ The absence of MPL variants in some studies could indicate that MPL variants are not equally prevalent among all population groups.⁷

The absence of a classical driver gene variant in otherwise typical MPN suggests that there are alternative low-frequency non-canonical MPN variants which can result in a similar phenotype.^8,9 Cases of triple-negative MPNs (negative for the canonical variants in JAK2, CALR and MPL) occur in ET and PMF in 10% to 15% of cases.¹⁰

Driver variants in JAK2, CALR and MPL affect the MPN phenotype and prognosis. Primary myelofibrosis patients with JAK2 p.V617F are more likely to present with portal vein thrombosis, while those with CALR variants are younger, have a higher platelet count, and a lower leukocyte count when compared with other PMF patients.^11,12 The risk of blast transformation in CALR-mutated PMF is lower than in JAK2-mutated or triple-negative PMF.^11,12 There may be population-specific differences in the prognostic implications of a genetic driver. Calreticulin-mutated PMF was shown to have improved overall survival compared with JAK2-mutated PMF in the non-Asian population, while in the Asian group, JAK2-mutated PMF performed better.¹³ Such findings highlight the need for better description of the MPN mutational landscape in different populations.

The heterogeneity in variant frequency of JAK2, CALR, MPL and triple-negative MPNs between regions indicates that epidemiological studies in sub-Saharan Africa are required. The variant heterogeneity could be compounded in countries such as South Africa, where extensive genomic diversity is present.¹⁴ In addition, to our knowledge, there have been no published epidemiological studies assessing the frequency of JAK2, CALR or MPL variants in MPNs in South Africa. Understanding the genetic drivers has both diagnostic and therapeutic implications.

The aim of this study was, therefore, to describe the genetic variants found in MPNs in a South African population. The study objectives were to determine the frequency of PV, ET and PMF; to determine the frequency of JAK2 p.V617F, JAK2 exon 12, CALR and MPL variants in MPNs; and to compare full blood count parameters between variants within each MPN group.

A total of 757 patients were tested for variants in JAK2 p.V617F, JAK2 exon 12, CALR and/or MPL over the 12-year period (Figure 1). All patients were initially tested for JAK2 p.V617F. Of these, 143 (19%) were positive for the variant and were not assessed for other variants based on the workflow of the testing laboratory.

Diagnostic bone marrow examination was not performed in 420 (55%) of the patients who had genetic analysis. Of these, 16 met the diagnostic criteria for PV without a bone marrow examination.¹ Bone marrow examination was performed on 337 (45%) patients (Figure 1). In 63% of these (212/337), an MPN-associated variant was not detected and the final diagnosis on bone marrow examination was not in keeping with a classical MPN. In 4% (13/337), a variant for a classical MPN was detected; however, on bone marrow examination, the diagnostic criteria for a classical MPN were not met.

There were 128 patients diagnosed with an MPN and included for further analysis, with 16 of these meeting diagnostic criteria for PV without bone marrow aspiration and trephine (Figure 1). The average age of MPN diagnosis was 62 ± 14 years with a male to female ratio of 1:1 (64 male patients to 64 female patients). Of the 128 patients diagnosed with an MPN on bone marrow examination, 49 (39%) had PV, 26 (20%) had ET, 17 (13%) had pre-PMF, and 25 (19%) had overt-PMF. In addition, 11 patients (9%) were classified as MPN unclassifiable. The ages at presentation of PV, ET and PMF were similar (Table 1).

In PV, JAK2 p.V617F was seen in the majority of patients (96%). The remaining two (4%) JAK2 p.V617F-negative patients with PV were also negative for JAK2 exon 12 variants (Figure 2). The diagnosis of PV was based on the minor criterion of low erythropoietin level in these cases. In ET and PMF, the most common variant was JAK2 p.V617F, seen in 65% (ET) and 71% (PMF) (Figure 2). Patients who were JAK2 p.V617F negative were tested for CALR and MPL variants; that accounted for seven of the nine ET patients and six of the 12 PMF patients. The patients who were not tested for CALR or MPL variants were investigated for an MPN before testing for CALR and MPL variants became available at our centre. Three patients (2%) had a CALR 52 bp deletion (Type 1 mutation) – all of these were PMF cases. Five patients (5%) had a CALR 5 bp insertion (Type 2 mutation) – four were in patients with ET and the other one in a patient with PMF. A novel CALR frameshift variant was detected in one patient with ET [NM_004343.3(CALR):c.1104_1143del40 (p.E369Qfs)]. All tested patients were negative for an MPL variant. There were five triple-negative cases, two of which were ET, two PMF, and one case was MPN unclassifiable (Figure 2).

In PMF, haemoglobin was significantly higher in those patients positive for JAK2 p.V617F compared to those negative for JAK2 p.V617F (p = 0.013). In addition, patients diagnosed with PMF who lack the JAK2 p.V617F variant were significantly younger than the JAK2 p.V617F-positive group (p = 0.048). There were no other significant differences in ET or PMF between the JAK2 p.V617F-positive and -negative groups (Table 1). There were too few PV patients who were negative for both JAK2 p.V617F and exon 12 (n = 2) for meaningful comparison between groups.

We found that PV was the most common MPN subtype, accounting for 38% of MPNs, followed by PMF in 33%. Essential thrombocythaemia was the least common MPN, at 20%. We found that JAK2 p.V617F was the most common variant across all MPNs, seen in 96% of PV, 65% of ET and 71% of PMF. No MPL variants were detected among the patients tested. In PMF, JAK2 p.V617F was associated with older age at diagnosis and higher haemoglobin.

The proportion of PV, PMF and ET cases found in our study compared well to a seven-year retrospective study which examined JAK2 p.V617F variants in the Gauteng Province of South Africa.¹⁶ Primary myelofibrosis was diagnosed in 43.7% of the Gauteng patients, and ET in 11.5%.¹⁶ In contrast, PMF is typically the least frequent of the three classical MPNs outside of South Africa.⁷ In a large meta-analysis of 52 studies, which included 5300 MPN patients from the United States, China, Brazil and Europe, only 18% of MPN cases were diagnosed with PMF and 49% with ET.⁷ The reason for this difference in the frequency of the MPNs between South African and international studies is unclear. It may reflect a true population-specific difference in MPN frequencies; however, it may also be that patients with PMF are more likely than those with ET both to seek medical care and to be referred from primary care facilities for further investigation because of the more pronounced symptom severity seen in PMF.

In our 49 patients with PV, 96% had the JAK2 p.V617F. The frequency of JAK2 p.V617F variant ranges from 37.8% to 100% in PV across different studies (Table 2). The average frequency of JAK2 p.V617F-negative PV has been reported as 5% in literature from Europe and the United States, although the number varies slightly between studies (Table 2).¹⁷ This may be related to sensitivity of testing modality and sample study size, rather than differing driver variants. Of note, many studies did not perform a test for JAK2 exon 12 variants. The only African studies investigating the frequency of JAK2 exon 12 variants in PV patients were a Sudanese study that reported a JAK2 exon 12 frequency of 8.1% and an Egyptian study with a frequency of 0% (Table 2).^18,19 The frequency of PV driver variants in sub-Saharan African countries has not yet been determined. We only had two PV patients who were negative for JAK2 p.V617F and both patients were negative for JAK2 exon 12 variants. Further large-scale South African studies are required to more accurately characterise the frequency of JAK2 exon 12 mutations within the local population.

Our findings showed that there were no MPL variants detected in the JAK2 p.V617F-negative ET and PMF patients. The only other African studies, which were performed in Egypt, also failed to detect MPL variants in ET and PMF patients.^18,20 Of note, in our study, there were 11 patients who did not undergo MPL testing after testing negative for JAK2 p.V617F. Their mutational status is therefore unknown. Our findings support revising our institutional standard operating procedure to perform JAK2 p.V617F testing upfront, given its high prevalence, followed by next-generation sequencing when JAK2 p.V617F is negative. This strategy will improve detection of non-canonical mutations and, importantly, enhance identification of low allele frequency CALR variants that are currently under-recognised in South Africa.

Janus kinase 2 p.V617F or CALR variants were present in the majority of ET and PMF patients in our study, a finding frequently reported in the literature (Table 3 and Table 4). The frequency of CALR variants in our study was on the lower end of the frequency distribution range, with the majority of our patients having a JAK2 p.V617F variant. In our study, the prevalence of JAK2 p.V617F is higher in PMF patients (71%) than ET patients (65%) (Table 3 and Table 4). Calreticulin 5 bp insertion mutations (n = 6) were more common in our patients than CALR 52 bp deletion mutations (n = 3), which contrasts with Europe (32% vs. 56%).⁶ There were only 10 CALR variant-positive patients in our study and larger studies are required to confirm this finding.

Triple-negative ET and PMF were relatively infrequent in our patients at only 6%, which is less than the 10% to 15% reported in international studies.¹⁰ The number of triple-negative patients in this study may be higher than reported if all the ET and PMF patients who tested negative for JAK2 p.V617F were also tested for other variants, namely those affecting CALR and MPL.

Our study showed that patients with PMF were statistically more likely to present with a higher haemoglobin level if they were positive for the JAK2 p.V617F variant. The JAK2 p.V617F positive PMF patients were also significantly older at presentation than the JAK2 p.V617F negative group. These findings correlate with the international literature.³⁵ Haemoglobin level and age in PMF were the only phenotypic parameters where a statistically significant difference was associated with a specific variant, despite the fact that laboratory features of JAK2 p.V617F have been distinguished from those of other variants in PV, ET and PMF.^8,35,36 Future studies on prognosis of different variants in PMF are required in a South African context.