Chronic neutrophilic leukemia: 2022 update on diagnosis, genomic landscape, prognosis, and management

2.1 Clinical features

CNL is a clinically heterogeneous disease. It most commonly presents as incidental neutrophilic leukocytosis in an asymptomatic subject. In a series of 14 molecularly defined CNL patients, 71% had antecedent chronic leukocytosis lasting a median of 12.5 months (range 5–84 months).¹⁰ CNL may also manifest with a spectrum of constitutional symptoms, fatigue being the most common, as well as bone pain, pruritus, easy bruising, or gout.⁵ Clinical examination may reveal splenomegaly (present in 36% of cases in one report¹⁰) and/or hepatomegaly, though lymph node involvement is relatively uncommon.¹² There are substantial data linking CNL to a bleeding diathesis,^13-16 including an appreciable incidence of cerebral hemorrhage. Mechanistically, this has been attributed to thrombocytopenia, platelet dysfunction, and/or vascular wall infiltration by neoplastic cells.^{14, 17} Patients exhibiting bleeding tendencies should undergo testing for acquired von Willebrand's disease along with other acquired coagulation and platelet function disorders.

2.2 Laboratory findings

The hallmark of CNL is persistent chronic, mature neutrophilia. CNL is distinguished by more mature granulopoietic forms compared with chronic myeloid leukemia (CML), with the preponderance of granulocytes (≥80%) at the segmented or band stage and minimal to no circulating blasts. There is also notable absence of monocytosis, basophilia, and eosinophilia. The majority of CNL patients also present with mild anemia (median hemoglobin ~11 g/dL)^{8, 9} and/or thrombocytopenia. In a 2015 series of WHO-defined CNL cases, peripheral blood parameters at diagnosis, expressed as median (range) were hemoglobin 10.8 g/dL (8.5–16.0), leukocyte count 54 × 10⁹/L (21.7–176.2), platelets 201 × 10⁹/L (25–476), neutrophils 85% (78–94), and immature cells (myelocytes and metamyelocytes) 6% (0–11).¹⁰ Worsening thrombocytopenia along with increasing splenomegaly often herald disease progression or blast crisis.⁸ Lactate dehydrogenase (LDH) levels are typically elevated, as is the leukocyte alkaline phosphatase (LAP) score, in contrast with the low LAP score observed in CML. Vitamin B12 levels are also often elevated due to transcobalamin release from granulocytes and their precursors.⁵ Of interest, low serum granulocyte-colony stimulating factor (G-CSF) levels have been documented in CNL patients,^18-20 although testing is not routinely performed.

By subdividing CNL patients into CSF3RT618I-mutated versus “other” CSF3R-mutated subgroups, two phenotypically and prognostically distinct subsets have emerged.¹¹ CSF3RT618I-mutated individuals have been found to cluster with adverse clinical and laboratory characteristics, with trends toward older age at diagnosis, higher white blood cell (WBC) counts and lower hemoglobin values, as well as significantly lower platelet counts at diagnosis and more frequently abnormal karyotype. As will be discussed, these features translate to significantly decreased overall survival (OS) compared to subjects harboring other CSF3R mutations.

2.3 Bone marrow morphology

Bone marrow is uniformly hypercellular (>90% cellularity) in CNL due to expanded neutrophilic granulopoiesis, with an increased myeloid to erythroid ratio which may exceed 20:1.²¹ The majority of granulocyte precursors are present in the metamyelocyte to segmented stages of maturation. There are fewer than 5% myeloblasts and an absence of Auer rods. There should also be no dysplastic features. Reticulin staining, though typically normal, may show minimal fibrosis but should not exceed a grade of 1+.⁸ Erythroid maturation is normoblastic and megakaryocytes, while morphologically normal, may be normal or slightly increased in number.⁹

2.4 Genetic predisposition

Although the true prevalence of constitutional CNL is undefined, there are currently three suspected familial cases of CNL reported^{26, 27}; only the most recently described one, however, represents inherited transmission of CSF3RT618I confirmed through targeted high throughput sequencing of germline DNA.²⁸ Germline CSF3R status should therefore be assessed in suspected familial cases. A recent study also identified germline CSF3R variants as novel predisposition genes in cases of both myeloid and lymphoid malignancies, though the mechanisms underpinning this risk remain unclear.²⁹

2.5 Revised 2016 WHO diagnostic criteria

The identification of CSF3R mutations in the majority of patients with CNL³⁰ revolutionized our understanding of the pathogenesis of this once biologically obscure disease. In 2016, the WHO endorsed the presence of CSF3RT618I or other activating CSF3R mutations as a key diagnostic criterion for CNL.³¹ The remaining criteria include peripheral blood leukocytosis of ≥25 × 10⁹/L (of which >80% are segmented neutrophils plus band forms and <10% are neutrophil precursors with rare myeloblasts), monocyte count <1 × 10⁹/L and absence of dysgranulopoiesis, as well as hypercellular bone marrow with granulocyte hyperplasia, normal maturation and <5% myeloblasts.³¹ The remaining components are exclusionary and include the absence of fulfillment of WHO criteria for another MPN and absence of rearrangements in PDGFRA, PDGFRB, FGRF1, or PCM1-JAK2.³¹ The WHO also recognizes CSF3R-negative disease if supported by chronic neutrophilia (minimum 3 months), splenomegaly, and absence of reactive leukocytosis. If there is an underlying plasma cell disorder, myeloid clonality must be demonstrated to make the additional diagnosis of CNL.³¹ Although the absence of a CSF3R mutation does not preclude the possibility of CNL, it should prompt careful review of the diagnosis. The 2016 WHO diagnostic criteria for CNL are summarized in Table 1.

2.6 Differential diagnosis

Neutrophilic leukocytosis may underlie a wide variety of diseases, both benign and malignant. An accurate CNL diagnosis requires the exclusion of potentially confounding entities such as reactive neutrophilia/leukemoid reaction, CML, neutrophilic-CML (CML-N), MPN/myelodysplastic (MDS) overlap disorders such as atypical chronic myeloid leukemia (aCML) and chronic myelomonocytic leukemia (CMML), as well as other myeloid neoplasms.

Distinguishing CNL from a leukemoid reaction may be challenging as both may present with significant neutrophilia, bone marrow hypercellularity, normal cytogenetics, and absence of BCR-ABL1 fusion gene. WBC count may be more modestly elevated in leukemoid reaction, though there have been reports of WBC counts up to 100 × 10⁹/L.³² Detailed history-taking and a thorough clinical examination are critical in excluding occult malignancy or infection. The demonstration of clonality, including identification of a CSF3R mutation or other molecular or cytogenetic abnormality, supports the diagnosis of CNL.

CML is invariably associated with a BCR-ABL1 fusion gene and manifests with a higher proportion of myelocytes as well as more frequent basophilia, thrombocytosis, and/or eosinophilia. While the absence of the Philadelphia chromosome is a provision to CNL diagnosis, a rare form of CML termed neutrophilic-CML, or CML-N, has been described with features that include prominent neutrophilia, overlapping with CNL.³³ CML-N is characterized by an uncommon BCR-ABL translocation resulting in the transcription of an e19/a2 type BCR-ABL messenger RNA yielding a 230-kD BCR-ABL protein (p230).^{8, 34} Clinically, CML-N presents with lower total WBC counts, less severe anemia, less frequent/severe splenomegaly, and delayed blast transformation compared to CNL.³³

Atypical CML is an uncommon and heterogeneous disease with shared features of both myeloproliferative and MDS neoplasms. It is characterized by leukocytosis (neutrophils and precursors comprising ≥10% of leukocytes) and prominent granulocytic dysplasia, useful in distinguishing it from CNL. Atypical CML also does not meet criteria for BCR-ABL1-positive CML, demonstrates absence of PDGFRA/B and FGFR1, and presents with minimal basophilia and monocytosis. The updated 2016 WHO criteria for aCML are outlined in Table 1.

CMML is another MPN/MDS overlap disorder whose diagnosis requires persistent (≥3 months) peripheral monocytosis >1 × 10⁹/L, absence of BCR-ABL1 fusion or PDGFRA/B rearrangements, fewer than 20% blasts or promonocytes in blood and bone marrow, and evidence of dysplasia or clonal abnormality (or persistent monocytosis lasting ≥3 months with exclusion of all other causes). The chronic monocytosis and presence of dysplasia in CMML are central in distinguishing it from CNL.

The diagnosis of CNL also requires exclusion of other myeloid malignancies including polycythemia vera (PV), essential thrombocythemia (ET), and primary myelofibrosis (PMF), each of which display characteristic molecular and morphological features well defined in the 2016 WHO classification of myeloid neoplasms.³¹

Rarely, paraneoplastic leukocytosis may result from ectopic production of G-CSF by solid tumors, mimicking the neutrophilic leukocytosis of CNL. Such cases have been reported in association with a variety of malignancies including urological,³⁵ lung,³⁶ mesothelium,³⁷ and thyroid,³⁸ among others, and may be associated with more aggressive tumor cell proliferation, perhaps owing to G-CSF's role in autocrine growth stimulation.³⁹

2.7 Disease associations

2.8 Cytogenetics and clonality

Cytogenetic abnormalities have been reported in a minority of CNL patients at diagnosis and/or during clonal evolution.^{8, 69, 70} In a 2002 review by Reilly,⁷⁰ 37% of CNL cases presented cytogenetic aberrations consisting primarily of trisomy 8, trisomy 21, deletion 11q and deletion 20q. Subsequently a CNL series of 40 patients identified cytogenetic abnormalities in 32.5% (13/40 patients).⁸ These were detected at baseline in 20% of patients and during clonal evolution in the remaining 12.5%, and included deletion 20q, trisomy 21, deletion 11q, and deletion 12p.⁸ Less commonly described abnormalities include tetraploidy 21, trisomies 7, 8, and 9, translocation 1:20, deletion Y, deletion 6, add 5p, deletion 15, and monosomy 2,⁷¹ and are considered nonspecific, yet nonrandom findings in myeloid malignancies.⁷²

3.1 CSF3R mutations in CNL: A biological unifying feature

G-CSF is a cytokine growth factor that regulates basal as well as stimulated granulopoiesis and promotes granulocyte differentiation.⁷³ The pivotal role of G-CSF in granulopoiesis has been underscored by studies demonstrating severe neutropenia in CSF3 or CSF3 receptor-deficient mice.^{74, 75} The G-CSF receptor is a single chain cell-surface receptor belonging to the cytokine receptor type I superfamily (chromosome 1p34.3).⁷³ Its cytoplasmic domain contains distinct functional regions: the membrane-proximal region is functionally ascribed to mitogenic signaling while the carboxy-terminal is involved in maturation signaling and regulation/suppression of proliferation.⁷⁶ On binding to its receptor, G-CSF exerts its effects via classical downstream pathways Janus kinase (JAK)-signal transducer and activator of transcription (STAT), SRC family kinases (notably LYN),^{77, 78} nonreceptor tyrosine kinase SYK, Ras/Raf/MAP kinases, and PI3K/Akt pathways, inducing neutrophilic differentiation, proliferation, and survival as well as stimulating neutrophil function.⁷⁹ The landmark 2013 discovery of CSF3R mutations in a high proportion of CNL patients represented a turning point in our understanding of CNL molecular pathogenesis.³⁰ Prior to this, CSF3R mutations had been reported in other myeloid disorders including severe congenital neutropenia (SCN), hereditary chronic neutrophilia, and rarely, myeloid leukemias, as described below.

SCN is a clinically heterogeneous inherited condition characterized by early onset of severe bacterial infections. Acquired somatic truncating mutations in CSF3R, analogous to those found in CNL,³⁰ have been described in up to 40% of SCN patients⁸⁰ concurrently with inherited, pathogenically significant mutations such as ELANE or HAX1.⁸¹ Although these acquired, mostly nonsense CSF3R mutations were initially thought to underpin SCN pathogenesis,^{82, 83} they were later demonstrated to define a subset of ~80% of SCN patients at high risk of leukemic transformation.^{80, 83-87} These truncating mutations are neither required nor sufficient for leukemic conversion in SCN, as supported by knock-in murine models.⁸⁸ It is postulated that through co-operation with other oncogenes,^{85, 89} these CSF3R mutations provide a clonal advantage with prolonged cell survival.⁹⁰

Another distinct CSF3R mutation has been involved in the pathogenesis of hereditary chronic neutrophilia. The germline, autosomal-dominant transmembrane domain CSF3RT617N mutation was described by Plo et al.⁹¹ as inducing dimerization of the CSF3R transmembrane domain, thus promoting constitutional receptor activation in a family with chronic neutrophilia, splenomegaly, and increased circulating CD34-positive myeloid progenitors. This receptor-independent granulocyte proliferation was reportedly sensitive to JAK2 inhibition.⁹¹

In 2013, Maxson et al.³⁰ identified disease-defining oncogenic CSF3R mutations in the majority of patients with CNL. In this study, CSF3R mutations were found in 89% and 40% of CNL and aCML cases, respectively. Two mutational variants of CSF3R were reported: the more common membrane proximal mutations consisting primarily of T618I and T615A point mutations, and frameshift or nonsense mutations leading to premature truncation of the cytoplasmic tail of CSF3R (D771fs, S783fs, Y752X, and W791Z). The truncation mutations, which also occur in SCN, largely coexisted with membrane proximal or transmembrane CSF3R mutations as compound mutations.^{30, 92} Mechanistically, the membrane proximal mutations prevent O-glycosylation of CSF3R resulting in increased active dimeric configuration, ligand-independent receptor activation, and constitutive downstream signaling through JAK2^{30, 93} (and potentially Bruton's tyrosine kinase according to one report⁹⁴). On the other hand, receptor truncation involves a loss of negative regulatory motifs including the dileucine sorting motif, which plays a role in receptor internalization,⁹⁵ and binding sites for the suppressor of cytokine signaling 3 (SOCS3) which targets the receptor for degradation.⁹⁶ Truncation mutations thus disrupt receptor trafficking, resulting in delayed receptor internalization and/or degradation,^{92, 95} increased CSF3R cell surface expression, sustained STAT5 activity, and increased cell proliferation.⁹⁷

The ability of CSF3RT618I, the most common CSF3R mutation, to promote CNL leukemogenesis was established by Fleischman et al.⁹⁸ using a murine bone marrow transplant model. The CSF3RT618 mutation induced a lethal myeloproliferative disorder which recapitulated the clinical features of CNL, providing proof of concept to Maxson et al.'s⁹⁸ seminal report. Moreover, the CSF3RT618I mutation was confirmed to act via the JAK–STAT signaling pathway as granulocytosis and splenomegaly were sensitive to treatment with JAK inhibitor, ruxolitinib.⁹⁸ An analogous study demonstrated the leukemogenic potential and likewise favorable response to ruxolitinib of one of the alternate transmembrane domain mutations, CSF3RT640N.⁹⁹

The proportions of CNL/aCML patients harboring membrane-proximal only versus membrane-proximal concurrent with truncating CSF3R mutations (compound mutants) are 75% and 25% respectively.¹⁰⁰ Given the scarcity of truncation-only CSF3R mutations in CNL, Maxson and Tyner¹⁰¹ postulated that, contrary to membrane proximal mutations, truncation variants alone may be insufficient to drive leukemogenesis. This was later confirmed in studies using knock-in mice expressing truncated CSF3R.⁸⁸ This attenuated pathogenicity has been posited to be due to inefficient MAP kinase signaling.¹⁰² Importantly, in Maxson et al.'s³⁰ landmark study, the two prototypical CSF3R mutations showed differential susceptibilities to tyrosine kinase inhibitors: membrane proximal mutations responded preferentially to the JAK inhibitor ruxolitinib, while truncation mutations were sensitive to inhibition with SRC kinase inhibitor dasatinib. This critical finding corroborated the concept of distinct downstream signaling dysregulation profiles in CNL, and provided a rationale for molecularly directed therapeutic targeting.

A follow-up study by Pardanani et al.⁶⁵ analyzed CSF3R mutations in strictly WHO-defined patient subsets. Of 35 suspected cases of CNL, 12 were subsequently confirmed to have true WHO-defined CNL with 100% of these harboring CSF3R mutations, while no cases of WHO-defined aCML, MG-associated CNL, PMF, or CMML were CSF3R-mutated.⁶⁵ CSF3RT618I was the most prevalent mutation, detected in 10 of 12 CNL patients, while the remaining two harbored alternative CSF3R mutations, M696T and I598I.⁶⁵ This study vitally endorsed the CSF3RT618I mutation as a sensitive and specific molecular marker for CNL, and the authors correspondingly recommended its inclusion into contemporary diagnostic criteria.⁶⁵ Later studies corroborated the lower mutational frequency of CSF3R mutations in aCML, which ranges from 0% to 40%, and confirmed the high prevalence of CSF3R mutations in WHO-defined CNL (80%–100% of patients).^{30, 103, 104} It is of note that CSF3R mutations are uncommon in AML (typically reported <3%–4%),^{105, 106} however, when present, they are often associated with core-binding factor gene abnormalities and double-mutated CEBPA,¹⁰⁷ with the significance of this association remaining unclear.

Up to 30% of CSF3R-mutated CNL cases exhibit compound mutations consisting of both membrane-proximal and truncation mutations on the same allele.¹⁰⁰ Rohrabaugh et al.¹⁰² performed elegant studies demonstrating the ability of CSF3R compound mutations to induce aggressive lethal leukemia in mice. Interestingly, BaF3 cells expressing CSF3R compound mutations were resistant to both JAK and SRC inhibitors in vitro.¹⁰² Both proximal and compound CSF3R mutations were shown to rely on enhanced MAPK signaling for oncogenic transformation and leukemia development in mice expressing either of these variants, which was effectively abrogated by targeting Mek1/2 with trametinib.¹⁰² Overall, these data establish a potential role for enhanced MAP kinase signaling in CSF3R-mutated myeloid malignancies and convey potential JAK inhibitor resistance in patients harboring compound mutations.¹⁰²

3.2 Additional pathogenic mutations in CNL

Concurrent mutations in CSF3R-mutated CNL patients are observed with variable frequencies and may be categorized into those impacting epigenetics such as ASXL1, mutations in SETBP1, spliceosome complex mutations such as SRSF2, as well as signaling mutations such as JAK2 (less frequently reported).^{10, 65, 101, 104, 108-110} Furthermore, ~10%–20% of CNL cases are negative for CSF3RT618I and other membrane-proximal mutations suggesting that additional genetic lesions contribute to the leukemic phenotype. Distinct models for mutation acquisition in CNL according to whether the CSF3R mutations occurs either a primary or secondary event have been proposed.¹⁰¹ The most frequently occurring comutations will be detailed below and are summarized in Table 2.

ASXL1 regulates histone modification and has been shown to be mutated with variable but, in certain cases, high frequency in CNL. Reported mutational frequencies range from 30% to 81%.^{10, 11, 104, 113, 114} In a critical study by Elliott et al.,¹⁰ the presence of ASXL1 mutations (along with thrombocytopenia) was found to be independently predictive of reduced survival in multivariable analysis. This finding was reiterated in another larger series of 19 CNL patients from the Mayo Clinic.¹¹ Furthermore, evolution of CNL to CMML was observed in patients harboring ASXL1 and lacking SETBP1 mutations, denoting a potential pathogenic role for ASXL1 in transformation.¹⁰ TET2 mutations appear to be less prevalent. In a cohort of 14 CNL patients, Meggendorfer et al.¹⁰⁴ found close to 30% of subjects were TET2-mutated. Sporadic mutations in other epigenetic modifiers such as EZH2 and KDM6A have also been reported.¹¹³ In recent years, the involvement of TET2 and ASXL1 mutations in clonal hematopoiesis of indeterminate potential (CHIP)¹¹⁵ has prompted the theory of CHIP providing a backdrop for a later-occurring CSF3R mutation, ultimately leading to a CNL phenotype.¹⁰¹

The prevalence of SETBP1 mutations in CNL has been reported between 14% and 56%^{65, 100, 104, 108, 116}; these are enriched in CNL patients carrying CSF3R mutations.¹⁰⁰ Pardanani et al.⁶⁵ and Cui et al.,¹⁰⁸ respectively, found that 40% and 75% of CSF3RT618I-mutated CNL patients coexpressed mutations in SETBP1. Elliott et al.¹⁰ demonstrated that all five (100%) of 13 CNL patients carrying a SETBP1 mutation coexpressed CSF3RT618I. In a 2017 review of 16 WHO-defined CNL cases, all of whom harbored the CSF3RT618I mutation, SETBP1 mutations were identified in 63% (n = 10/16).¹¹⁴ Conversely, Meggendorfer et al.¹⁰⁴ found that, although not statistically significant, SETBP1 and CSF3R mutations appeared to be mutually exclusive in their CNL cohort (n = 0/6). The prognostic relevance of concurrent SETBP1 and CSF3R mutations remains to be determined, though some evidence suggests this to be a detrimental alliance.⁶⁵ Elliott et al¹⁰ observed that both of the CNL patients (n = 2/14) who developed blast phase disease carried CSF3R-SETBP1 (but not ASXL1) comutations. There are also conflicting reports on the response to JAK inhibitors in cases of CSF3RT618I and SETBP1 co-occurrence, with either drug resistance,^{117, 118} hematological, clinical, and molecular improvement,¹¹⁹ or no significant prognostic impact detected.¹²⁰

Mutations in spliceosome complex components such as SRSF2 and U2AF1 have less commonly been reported in CNL. Meggendorfer et al.¹⁰⁴ found a mutational frequency of 21% for SRSF2 mutations in their cohort. Dao et al.¹¹³ found more frequent mutations in U2AF1 in a cohort of subjects with CSF3RT618I–mutated myeloid neoplasms (n = 4/10). The cooperative role of these and other spliceosome mutations in the pathogenesis of CNL, as well as their prognostic implications remain obscure.

Another rarely occurring comutation in CNL is calreticulin (CALR). Lasho et al.¹²¹ provided the first account of this in 2014 (CALRE398D c.1194 G > T). Cui et al.¹⁰⁸ also reported concomitant CSF3RT618I and CALR exon 9 frame-shift mutations in one subject (c.1154-1155insTTGTC). Thus far, no clear prognostic or therapeutic implications of CALR mutations in CNL have been disclosed.

3.3 Next generation, whole-exome, and RNA sequencing studies

Results of next generation sequencing (NGS) studies in CNL were initially reported by Langabeer et al.¹¹¹ in a cohort of four CSF3RT618I-mutated patients. Additional mutations were identified in all subjects as follows: SRSF2 (n = 4/4), SETBP1 (n = 3/4), NRAS (n = 1), and CBL (n = 1). Zhang et al.¹¹² more recently published comprehensive genomic and transcriptomic profiling data from 158 cases of neutrophilic leukemias of ambiguous diagnosis, including 39 patients with CNL (the remainder consisting of aCML, MPN-unclassified, MDS/MPN, and ambiguous diagnosis). The most prevalently mutated gene pathways were found to be those involved in chromatin modification (ASXL1, 65.8%; EZH2, 19%; and ASXL2, 3.2%). Mutations in genes involved in DNA modification were also frequently observed (TET2, 33.5%; and DNMTA, 5.7%), as well as in spliceosome complex members (SRSF2, U2AF1, SF3B1, ZRSR2, and RPRF8) with a total incidence of 55.7%.¹¹² CSF3R and SETBP1 mutations were identified more frequently in CNL, consistent with previous data.³⁰ Significant co-occurrence of CSF3R and SETBP1 as well as ASXL1 with CSF3R mutations was reiterated. A high level of genetic complexity was manifested by a high median number of mutations per patient (3.6; range, 0–8). More than half of the patients harbored mutations in ≥3 major pathways collectively involving ASXL1/2, TET2/GATA2, a signaling gene, and/or splicing complex member with ≥15 different combinatorial patterns. CSF3R was specifically demonstrated to co-occur with NRAS, CBL, PTPN11, SH2B3, NTRK2, and ABL1 mutations. The patterns characterizing CNL were similar to those of CMML, but distinct from those involved in age-related clonal hematopoiesis or other chronic MPN. Finally, drug sensitivity assays conducted in cells with concomitant CSF3R and NRAS mutations disclosed reduced drug response to either ruxolitinib or trametinib, a MEK inhibitor, but sensitivity to the drug combination targeting both pathways. In their conclusion, the authors suggest that CNL, aCML, MDS/MPN-U, and CMML belong to a disease “continuum,” and that while CNL is currently classified as an MPN, it more closely resembles the former group of diseases in its molecular constitution.

3.4 Mutation order and clonal evolution

Mutation order in CNL and other neutrophilic leukemias was assessed by Zhang et al.¹¹² in the aforementioned 2019 study. Variants in EZH2, SETBP1, TET2, U2AF1, and SF3B1 were proposed to be early mutations, while those in ASXL1, SRSF2, CSF3R, CBL, and NRAS, displaying a wider range of variant allele frequencies (VAFs), occurred irregularly either in the founder clone or in later subclones. Altogether, these data support a linear, as opposed to branching, mutation acquisition pattern in CNL. Clonal evolution in CNL has been addressed in several additional studies.^{8, 10, 111, 119} Langabeer et al.'s¹¹¹ NGS study identified clonal evolution in all patients at blast transformation, documented by increasing CSF3RT618I allele frequency or gain/loss of mutations. Karyotypic aberrations in the form of acquisition of monosomy 5 and 7 were noted in a case of CNL with blast transformation.¹⁰ Another report disclosed de novo trisomy 21, deletion 12p, and monosomy 7 anomalies at the time of blast transformation in three of 12 CNL patients, all of whom were cytogenetically normal at baseline (it is, however, disputable whether hydroxyurea therapy may have contributed to this).⁸ Most recently, Nooruddin et al.¹¹⁹ provided valuable insight into the clonal evolution of CNL, as de novo mutations in KIT and GATA2, as well as increased RUNX1 allele frequency, were present in an index CNL patient at the time of disease progression, and are posited to have driven clonal evolution.

4.1 Prognostic variables and risk stratification

Historically, phenotypic features such as leukocyte and platelet counts, were used to broadly align prognosis CNL; however, emerging genetic data are redefining risk determination.^{10, 11, 112} In their study of 16 WHO-defined CNL patients, Cui et al.¹¹⁴ found WBC >50 × 10⁹/L to be prognostically detrimental (median OS of 11 vs. 39 months in those with WBC <50 × 10⁹/L). In a study of 14 CSF3R-mutated CNL cases by Elliott et al.,¹⁰ multivariable analysis revealed ASXL1 mutations and thrombocytopenia to confer a significant survival disadvantage. Notably, only four (29%) patients were alive after a median follow-up of 77.6 months, none of whom harbored ASXL1 mutations. Interestingly, though SETBP1 mutations have been reported to contribute to disease transformation and treatment resistance,^{65, 117} they had no impact on survival in this study, nor in a contemporary meta-analysis by Shou et al.¹²⁰

An operational risk model predicting long-term survival in CNL was developed from data captured from 19 molecularly annotated, CSF3R-mutated CNL patients from the Mayo Clinic.¹¹ Multivariable analysis established the following prognostically informative variables with points commensurate to relative risk: platelets <160 × 10⁹/L (2 adverse points), leukocytes >60 × 10⁹/L (1 point), and presence of an ASXL1 mutation (1 point). When patients were allocated to low-risk (0–1 points; n = 9) or high-risk (2–4 points; n = 10) categories, respective OS times were “not yet reached” versus 22.4 months (p = .0016). From a practical standpoint, it was suggested that high-risk individuals undergo closer monitoring and possibly earlier consideration for hematopoietic stem cell transplant (HSCT). Another significant finding in this study was that CSF3RT618I-mutated individuals exhibited reduced life expectancy compared to other “variant” CSF3R-mutated patients, evoking the concept that T618I mutants may represent the definitive CNL entity from a molecular perspective. Recent genomic profiling has identified additional prognostically relevant combinatorial mutation patterns in CNL, with NRAS, ASXL1, GATA2, and DNMT3A mutations clustering with a trend toward shorter OS and, conversely, CBL mutations appearing to predict more favorable survival.¹¹²

5.1 Conventional therapy, oral chemotherapy, and interferon

Both splenic irradiation and splenectomy have been used since the 1970s for palliation of symptomatic splenomegaly in CNL,¹²⁶ including the index case reported by Tuohy.¹ However, due to reports of worsening neutrophilia postoperatively, splenectomy is currently not recommended for CNL.^{1, 5} Reports of use of oral chemotherapy in CNL also date back to the late 1970s,^{5, 126, 127} and though classically, these agents have been used to control leukocytosis, their effect is inevitably transient. Hydroxyurea has been the most commonly utilized agent in CNL with up to 75% of patients showing an initial response (reduced leukocytosis and/or splenomegaly) with median duration of therapy of 12 months (range: 6–87).⁸ In the Mayo Clinic's 2018 study of 19 molecularly annotated CNL patients, most were treated with hydroxyurea as a first-line agent (82%) and it was ultimately received by all (100%) over disease course.¹¹ The majority of subjects, however, required second-line therapy (53%) and nearly a third required three lines of therapy or more (32%). Other agents used in this series included interferon-alpha (IFN-a), hypomethylating agents, ruxolitinib, thalidomide, cladribine, and imatinib. A recent study by Dao et al.¹²⁵ similarly documented initial hydroxyurea treatment in 81.5% of patients who subsequently received ruxolitinib, confirming its preferential use as front-line agent in clinical practice.

IFN-a also has a long history of use in CNL and is the only agent offering the potential for durable remission, as published in limited case reports.^{10, 16, 122, 128} Meyer et al.¹²² described two CNL patients with progressive disease treated with IFN-a, both achieving long-term response after treatment durations of 16 and 26 months, respectively. One patient experienced slow progression after discontinuing IFN, though never required retreatment (follow-up 90 months), while the second maintained stable disease (follow-up 66 months). Additional studies similarly report long-term hematological and clinical remissions with IFN-a^{10, 16, 128} or pegylated IFN¹²⁹ in CNL, lasting from 24 to 41+ months. IFN thus remains a safe and effective agent for CNL and should be considered as first-line therapy in patients of childbearing age, or as a second or subsequent-line agent failing previous therapy.

5.2 JAK inhibitors

Ruxolitinib is a JAK1/2 inhibitor currently approved for treatment of patients with intermediate- or high-risk myelofibrosis, PV patients intolerant or refractory to hydroxyurea, and steroid-refractory acute graft-versus-host disease.^{130, 131} Due to its ability to inhibit oncogenic JAK–STAT pathway signaling, ruxolitinib was rationalized to provide clinical benefit in CNL. Although not Food and Drug Administration-approved for CNL, ruxolitinib has been evaluated in a number of murine models and patients with CSF3RT618I-mutated CNL and aCML. The original study by Maxson et al.³⁰ included a patient harboring the CSF3RT618I mutation in whom treatment with ruxolitinib induced a clinical response, which was maintained after 11 months of therapy.¹⁰⁰ Fleischman et al.⁹⁸ corroborated the salutary effect of ruxolitinib in a CSF3RT618I-expressing mouse model by demonstrating reduced cell proliferation and improvements in leukocytosis, spleen weight, and body weight. In a series of 19 CNL patients, four (all having previously been exposed to hydroxyurea) received treatment with ruxolitinib either as second-line (n = 3) or third-line (n = 1) therapy.¹¹ Reponses were varied: in one case, treatment was ongoing with favorable response but had been initiated recently (~2 months prior to data collection), in two other cases, there was an initial response but eventual worsening of leukocytosis requiring subsequent additional therapies (duration of response ~9.5 and 36 months, respectively), and in one case, ruxolitinib was received during blast phase as a bridge to transplant for only ~0.5 months; this patient ultimately had a favorable outcome and was alive at last follow-up ~46 months from diagnosis.¹¹

The results of a much-anticipated phase II prospective multicenter clinical trial evaluating safety and efficacy of ruxolitinib in patients with CNL (n = 21) and aCML (n = 23) were recently published in the Journal of Clinical Oncology.¹²⁵ Overall response rate was 35%, including four complete and nine partial responses in the CNL cohort, and 85% of patients met the criteria for at least one category of clinical benefit(s). A diagnosis of CNL (vs. aCML) and presence of CSF3R mutation significantly correlated with clinical response to ruxolitinib, as defined by control of leukocytosis and reduction of spleen volume. Notably, responders had longer median survival times compared to nonresponders (23.1 vs. 15.6 months).

The impact of JAK inhibitor therapy on CSF3R allele burden has been addressed in multiple studies with conflicting results: Dao et al.¹³³ found no effect (aCML patient with CSF3RT618I mutation), while Nooruddin et al.¹¹⁹ and Gunawan et al.¹³⁴ observed reductions in allele burden with ruxolitinib, though this did not systematically correlate with symptom improvement. In the aforementioned phase II trial of ruxolitinib in 21 patients with CNL, changes in allele frequency were inconsistent.¹²⁵ Similarly mixed results were derived from Stoner et al.'s¹³⁵ report, with three patients presenting allele burden reductions while two others displayed minimal change over time.

5.3 Tyrosine kinase inhibitors

Rationale for the use of the SRC kinase inhibitor, dasatinib, in CNL was initially provided by Maxson et al.'s³⁰ 2013 report in which CSF3R truncation mutations, documented to operate predominantly through SRC kinases, exhibited in vitro drug sensitivity to dasatinib. Scarce data, however, actually address dasatinib activity in CNL patients in vivo. Of note, a recent study described a favorable response to chemotherapy plus dasatinib in a patient with B-cell acute lymphoblastic leukemia harboring a CSF3R truncation mutation.¹⁴⁰ However, it is speculated that in the context of additional proximal membrane mutations, which frequently co-occur in CNL, dasatinib may not be sufficient to induce response.¹³⁷ Corroborating this is a case of a compound CSF3R-mutated CNL patient (T618I and Q749X mutations) found to be refractory to dasatinib and ultimately requiring ruxolitinib salvage.¹³⁷ Dasatinib's efficacy in vivo is thus unconfirmed, particularly in cases of concurrent truncation and proximal mutations, and should be subject to further study. While a trial may be reasonable in CNL cases with isolated cytoplasmic truncation mutations, close monitoring for lack/loss of response would be mandated.

5.4 Induction chemotherapy

Standard induction “7 + 3” chemotherapy has not been shown to induce hematological remission in CNL. There is one report of a patient in blast phase CNL attaining a second chronic phase following AML induction chemotherapy,⁵ however, most patients are either refractory or succumb to treatment-related mortality.^{8, 16, 72}

5.5 Hematopoietic stem cell transplant

Literature on HSCT in CNL remains, regrettably, scarce. The first report of HSCT in CNL was in 1996 by Hasle et al.,⁵ documenting long-term remission in two cases. Limited cases thereafter corroborated this extent of benefit.^{16, 71} Transplant has been described in all phases of disease in CNL. As expected, patients transplanted in blast phase have worse outcomes (i.e., increased chemotherapy regimen-related toxicity and/or early relapse).^{8, 141} In a review of transplant outcomes in CNL, 71% of patients undergoing HSCT in chronic phase had more durable remissions compared to those transplanted in accelerated phase.¹²⁴ A contemporary retrospective study of allo-HSCT in Japan reported data from 14 aCML and five CNL patients transplanted between 2003 and 2014.¹⁴² The majority received myeloablative preconditioning regimens. Transplants were principally from alternative donors (n = 14) versus HLA-matched related (n = 5). One-year OS rates were 54.4% in aCML and 40% in CNL, which, given the often-dismal prognosis with standard therapy, supports allo-HSCT as a practice conferring survival benefit in CNL. Further substantiating these findings are recent population-based data from Ruan et al.⁷ This large-scale retrospective study disclosed favorable long-term survival in CNL patients approached with frontline allo-HSCT, as both transplanted subjects remained alive after 5 years (though both were <65 years old with no major comorbidities; Charlson Deyo scores of 0).⁷ Overall, given the limited efficacy of currently available therapies and CNL's often rapidly fatal course, it is recommended that eligible patients be considered for HSCT, particularly if they display high-risk features.¹¹ Following HSCT, the CSF3R mutation may serve as a biomarker for disease relapse, and may be reasonable to monitor in this context.¹⁴³ Future additional studies are required to define optimal timing of HSCT and evaluate the use of alternative stem cell sources and nonmyeloablative approaches.

5.6 Novel therapeutic targets

The identification of NRAS mutations in a proportion of CNL patients has led to speculation that MEK inhibition with drugs such as trametinib, may provide clinical benefits in certain cases.^{102, 144, 145} At this time, however, there are no investigational agents, to our knowledge, formally undergoing study for use in CNL. As we gain further insight into the genomic backdrop of CNL, a molecularly conditioned, “personalized” treatment approach will likely emerge as the new therapeutic prototype, though the clinical reality of this is merely at a precursor stage.

5.7 Global approach to management

A proposed algorithm for CNL management is provided in Figure 1. Once diagnosis is confirmed, performing NGS-based screening for ASXL1 and SETBP1 mutations could be advocated in light of these mutations' potential prognostic significance. Risk stratification according to the Mayo Clinic model¹¹ detailed above may be useful in broadly discriminating between low versus high-risk individuals. Eligible candidates should be evaluated for allogeneic HSCT. All patients, regardless of disposition for and/or prospective transplant plans, should undergo careful clinical evaluation and monitoring. At any timepoint, pharmacologic therapy should be initiated for either uncontrolled myeloproliferation (a reasonable target being leukocyte count <25–30 × 10⁹/L) and/or associated symptoms.