The Diagnostic Work-Up of Hypereosinophilia

Eosinophilia is defined as an increase of eosinophils in peripheral blood (PB) or tissues above what is considered to be the normal range. Eosinophils in PB of normal individuals range from 0.0 to 6.0% of leukocytes and have an absolute eosinophil count of 0.05–0.5 × 10⁹/L. The normal range for eosinophils in bone marrow (BM) is 1–6%. A definition of BM eosinophilia has been proposed that requires ≥20% of marrow cells to be eosinophils, with or without PB eosinophilia [1, 2]. In normal tissue and organs, eosinophils are either absent or scattered, depending on the sites. Tissue eosinophilia is defined as increased eosinophils or signs of eosinophil degranulation in extramedullary sites such as the gastrointestinal tract, lung, thymus, spleen or lymph nodes. Eosinophils are  normally controlled by cytokines interleukin (IL)-5, GM-CSF, and IL-3 produced by T-lymphocytes, mast cells, and stromal cells [3]. Upon activation, eosinophils release their granules, such as eosinophil peroxidase, eosinophil cationic protein, major basic protein, and cytokines like TGF-β [3] that may lead to thrombosis and tissue fibrosis and injury [3].

Hypereosinophilia (HE) is defined by a marked increase in eosinophils in PB, ≥1.5 × 10⁹/L [4]. If HE is persistent (≥6 months) [5, 6], and there is associated tissue damage, the disorder would be classified as hypereosinophilic syndrome (HES). A history of 6 months may not be necessarily enforced if the diagnostic work-up is adequate and treatment is needed to minimize organ damage caused by the eosinophilic infiltrate. Therefore, it is generally accepted that in the presence of tissue injury, if blood absolute eosinophil count > 1.5 × 10⁹/L on 2 occasions in an interval ≥1 month, HE could be considered as “persistent” [1]. On the other hand, if a patient is asymptomatic and an underlying cause is not identified, a minimal duration of 6 months would be needed to consider a case to be idiopathic HE.

Eosinophils in clonal myeloid disorders are often derived from hematopoietic progenitors bearing the same molecular genetic aberrations. These clonal myeloid neoplasms can be further categorized into 3 large groups: (1) myeloid/lymphoid neoplasms with eosinophilia and rearrangements of PDGFRA, PDGFRB, FGFR1 or provisionally PCM1-JAK2 [7-9]; (2) HE associated with another well-defined myeloid neoplasm, such as chronic myeloid leukemia (CML); and (3) chronic eosinophilic leukemia (CEL) not otherwise specified (NOS). The WHO classification of CEL, NOS is a diagnosis of exclusion that is characterized by a clonal eosinophilic proliferation [6]. The definition and classification of HE is shown in Table 1, and the diagnostic criteria for CEL, NOS and idiopathic HES are shown in Table 2 and Table 3, respectively.

The clinical work-up of HE can be challenging due to the broad and complex underlying causes of HE, either reactive or neoplastic. An algorithm that may facilitate the identification of the underlying cause is recommended in the evaluation of patients with HE [5] (Fig. 1). An accurate diagnosis and classification are essential for proper patient management.

Secondary HE is a reactive expansion of eosinophils driven by a primary disease process, such as parasitic infestation, drugs, allergies, autoimmune diseases, or malignant tumors. Patients with a recent travel history with a specific geographic location and recent exposure to animals, insects, raw food, or untreated water, a search for parasites in skin, respiratory secretions, or stool is necessary [10]. Repeated testing for ova and parasites is often needed, considering the life cycle of parasites. Other laboratory tests for suspected parasite infestation include serology, tissue biopsy, molecular testing, and imaging studies [11]. For patients who present with eosinophilia that is highly suspicious of helminthic infections, empirical treatment with ivermectin is both therapeutic and diagnostic [12]. Drug-induced eosinophilia [13] often shows a causative relation to recent use of drugs, although a delayed onset (2–6 weeks) may be seen in some cases [14]. Drug reaction with eosinophilia and systemic symptoms (DRESS) syndrome is a serious drug reaction characterized by eosinophilia, skin rash, and systemic symptoms that include fever, lymphadenopathy, pneumonia, hepatitis, and renal dysfunction [13, 14]. A number of medical conditions, including allergy and hypersensitivity, collagen-vascular disease, pulmonary eosinophilic disease, allergic gastroenteritis, and adrenal insufficiency, may present with eosinophilia. Tissue eosinophila is defined as tissue/organ damage due to an eosinophilic infiltrate that may or may not have PB eosinophilia. Pulmonary eosinophilic diseases [15, 16] comprise a heterogeneous group of disorders that are characterized by abnormally increased numbers of eosinophils within the pulmonary airways and parenchyma. The causes include infection, allergy (especially bronchopulmonary aspergillosis), toxins, and collagen vascular disease. Some are idiopathic. Churg-Strauss syndrome [17] is small-vessel necrotizing vasculitis characterized by chronic rhinosinusitis, asthma and prominent PB eosinophilia. A positive ANCA is found in 30–60% patients. A biopsy that demonstrates an eosinophilic infiltrate, necrotizing granulomas, and eosinophilic or giant cell vasculitis is considered to be the gold standard in confirming a diagnosis.

Secondary eosinophilia can be associated with a malignancy. Examples are lymphoblastic leukemia/lymphoma [18], peripheral T-cell lymphoma [19, 20], and classical Hodgkin lymphoma [21]. Eosinophilia can also be associated with solid tumors such as squamous cell carcinoma of the head and neck, renal cell carcinoma, small cell carcinoma of the lung, transitional cell carcinoma, and breast cancer. The cause of eosinophilia in a patient with a primary malignancy can be multifactorial, which can be a part of paraneoplastic syndrome (cytokines produced by tumors) [22], drug-induced HE, or a concomitant infection.

Lymphocyte variant HES (L-HES) is an entity characterized by HE associated with clonal circulating CD3-CD4+ T-cells. The CD3-CD4+ T-cells are often CD7dim/negative, CD2bright+, and CD5bright+ (Fig. 2). The median percentage of CD3-CD4+ lymphocytes of total lymphocytes is reported to be 17% (range 0.5–96%). These T-cells have a Th2 profile and are able to produce eosinophilopoietic cytokines such as IL-4, IL-5, and/or IL-13 [23], leading to overproduction of eosinophils [24]. In most cases of L-HES, the T-lymphocytes are clonal. Recently, a gain-of-function STAT3 mutation was found in a patient with L-HES, and by cell sorting, the mutation was identified in the CD3-CD4+ T-cells [25]. Patients with L-HES usually present with skin lesions and rheumatologic symptoms and less commonly with lymphadenopathy, gastrointestinal, and pulmonary symptoms. With long-term follow-up, about 5–20% patients reportedly develop T-cell lymphoma [26]. It is noteworthy that a significant number of angioimmunoblastic T-cell Lymphoma (AITL) patients have a small number of circulating CD3-CD4+ T-cells [27]. However, unlike AITL, CD10 expression and T-follicular helper cell markers are often not aberrantly expressed in L-HES but can be observed in patients with L-HES transformed to AITL [28].

In patients with HE, the work-up for a primary BM neoplasm should start as early as the first visit if there is significant cytopenia(s), cytosis, and abnormal circulating cells. The well-defined WHO myeloid neoplasms that may be associated with eosinophilia include systemic mastocytosis, CML, acute myeloid leukemia (AML) especially AML with inv(16) (p13q22)/CBFB/MYH11, myelodysplastic/myeloproliferative neoplasms (MDS/MPN), Philadelphia chromosome-negative myeloproliferative neoplasms, and acute lymphoblastic leukemia/lymphoma (ALL). Mild eosinophilia (> 5% eosinophils) at presentation is common in CML and observed in up to 78% patients [29]. However, the term “eosinophilic variant CML” [30, 31] is specifically used to describe a variant of CML with prominent eosinophilia with no significant neutrophilia or basophilia. Although these eosinophilic variant CML cases respond to TKI therapy, HE can cause heart, lungs, and CNS damage that may be severe and irreversible. Less than 1% of cases of ALL are associated with HE [32, 33], and these cases often have t(5; 14) (q31;q32) and chromosome 5 deletion involving IL3 gene that may lead to an increased eosinophil production. It is noteworthy that in all cases of HE, it is critical to screen for PDGFRA, PDGFRB, or FGFR1 or PCM1-JAK2 rearrangements before rendering a final diagnosis of “idiopathic” or a myeloid neoplasm that bears no disease-defining molecular genetic alternations, since myeloid lymphoid neoplasms with PDGFRA, PDGFRB, or FGFR1 or with PCM1-JAK2 can present as MPN, MDS/MPN, AML, or ALL. The differences in molecular genetic features and involvement sites between primary (clonal) HE versus secondary HE are shown in Table 4. The clinicopathological parameters needed for a comprehensive assessment of patients with HE are listed in Table 5.

Myeloid and lymphoid neoplasms with eosinophilia and rearrangements of PDGFRA, PDGFRB, and FGFR1 were recognized as a separate category in the 2008 WHO classification [9]; PCM1-JAK2 was added to this family as a provisional entity in the 2016 WHO classification revision [34]. The common features of these neoplasms are the overexpression of an aberrant tyrosine kinase as a result of a fusion gene, or rarely of a mutation. It has been shown that in these neoplasms, at least in some cases, the cell of origin is a mutated pluripotent (lymphoid–myeloid) stem cell. Eosinophilia is characteristic but not invariably present.

PDGFRA rearrangement, frequently with partner FIP1L1 [35], is the most common genetic lesion within this disease category, and accounts for approximately 10–20% of patients with unexplained eosinophilia in Western countries. On the other hand, about 70% of patients with PDGFRA rearrangement will present with eosinophilia [36]. The disease at presentation may resemble CEL, NOS, idiopathic HES, systemic mastocytosis with eosinophila, CML, or atypical chronic myeloid leukemia; BCR-ABL1-negative MPN; lymphoblastic leukemia (T-ALL or B-ALL) and rarely AML [37, 38]. Symptoms related to eosinophil activation commonly manifest as skin rash and less frequently as pulmonary, gastrointestinal tract, and/or cardiac involvement.

The BM is usually hypercellular with increased eosinophils, including eosinophilic precursors. Eosinophils may exhibit various morphological abnormalities. Megakaryocytes may not show significant abnormalities, but in some cases, there are large megakaryocytes with separated nuclear lobes or mixed small and large megakaryocytes. Spindled mast cells are often present but usually scattered and not in aggregates, with frequent aberrant expression of CD25 (around 60%). Some cases may show features fulfilling the criteria of systemic mastocytosis. These cases have been reported to have no KIT D816V mutations. However, microdissection of mast cells has revealed low levels of KIT D816V mutations in a subset of these cases [39]. Serum tryptase is often elevated. Myelofibrosis, of at least grade 1 of 3, is very common, being found in more than two-thirds of patients.

The interstitial deletion of 4q12 leads to FIP1L1--PDGFRA fusion. Since the CHIC2 gene is contained in the -deleted region, FIP1L1-PDGFRA fusion is often referred to as “CHIC2 deletion”. The deletion is cryptic by conventional cytogenetics; therefore, FISH is required to identify the -fusion/deletion. The fusion can also be -detected by RT-PCR [40]. PDGFRA with other partner genes is considered to be variant [41-45], including KIF5B-PDGFRA/ t(4; 10) (q12;p11), CDK5RAP2-PDGFRA/ins(9; 4) (q33;q12q25); STRN-PDGFRA/t(4; 12) (q12;p13.2); TNKS2-PDGFRA/ t(4; 10) (q12;q23.3) and ETV6-PDGFRA/ t(4; 12) (q12;p13), or BCR-PDGFRA/t(4; 22) (q12;q11.2) [46-49]. Recently, a case of B-ALL with FIP1L1-PDGFRA has been found to have a Ph-like ALL gene expression signature [50], which may be better classified as Ph-like B-ALL, rather than as myeloid lymphoid neoplasm with PDGFRA.

Almost all patients with the FIP1L1-PDGFRA fusion are sensitive to tyrosine kinase inhibitors (TKI) such as imatinib [40, 51, 52]. Primary or secondary resistance to imatinib is unusual [53, 54], but discontinuation of therapy may lead to relapse [55]. Resistance involves the T674I mutation within the ATP-binding domain of PDGFRα, analogous to the T315I in ABL1 mutation in CML and confers TKI resistance [53]. Cases with variant rearrangements are also sensitive to imatinib treatment. Chemotherapy is often needed when patients present with acute leukemia.

PDGFRB is located at chromosome 5q32, and the most common rearrangement is t(5; 12) (q32;p13.2)/ETV6-PDGFRB. PDGFRB has been found rearranged to a large number of other partner genes [56-60] and such fusions are considered as uncommon variants. However, fusions that are typically associated with Ph-like B-lymphoblastic leukemia, such as EBF1-PDGFRB, SSBP2-PDGFRB, TNIP1-PDGFRB, ZEB2-PDGFRB, and ATF7IP-PDGFRB [61-63], are excluded from this category.

Myeloid and lymphoid neoplasms associated with PDGFRB rearrangement are rare [2, 64]. Most affected patients are adult men in their 50s [64-67]. Eosinophilia is common but may not be prominent or not present at all. The hematological findings may resemble CMML, CEL, NOS, atypical CML, BCR-ABL1 negative MPN, or rarely MDS. AML presentation is uncommon. T-lymphoblastic leukemia has been occasionally reported at presentation or at the time of progression.

It is generally believed that PDGFRB rearrangement is not cryptic and that if sufficient metaphases are obtained, all cases should show an abnormality involving 5q31–33 [9]. However, by RNA sequencing, PBGFRB rearrangement has been identified in cases with no karyotypic indications [68]. On the other hand, FISH analysis may not detect all rearrangements of PDGFRB [66]. These caveats suggest the need for a genome-wide or targeted RNA sequencing method in the detection of PDGFRB fusions. In some cases, a trial of imatinib is recommended for practical purposes [41]. It is noteworthy that not all t(5; 12) (q31–33;p12) harbor ETV6-PDGFRB fusion and FISH study is necessary.

Most hematolymphoid neoplasms associated with translocations of PDGFRB are sensitive to TKIs. Patients with these neoplasms respond to imatinib therapy with excellent hematopoietic and molecular responses. Long-term follow-up of PDGFRB-rearranged patients treated with imatinib reported a 96% response rate and a 10-year overall survival rate of 90% [69]. Primary or secondary resistance to imatinib is uncommon [65, 70]. Additional therapy is needed when patients present with acute leukemia or develop acute leukemia.

Myeloid/lymphoid neoplasms associated with FGFR1 rearrangement derive from pluripotent hematopoietic stem cells and cases can present as a MPN, AML, T- or B-lymphoblastic leukemia/lymphoma, or mixed-lineage acute leukemia and are often associated with PB and or BM eosinophilia [34]. The latter may represent blast crisis of an undiagnosed MPN. The median age is 30–40 years (3–84 years), and there is a slight male predominance (1.5: 1) [8]. FGFR1-rearranged diseases usually do not respond to first-generation tyrosine kinase inhibitor (imatinib) therapy and are clinically aggressive. High-dose chemotherapy followed by allogeneic stem cell transplant may improve patient survival [71]. Ponatinib in combination with high-intensity chemotherapy may further improve the response in these patients [71, 72].

There are 14 other partner genes reported and each may be associated with different disease manifestations. Patients with t(8; 13) (p11;q12)/FGFR1-ZMYM2 rearrangement often present with an MPN with eosinophilia and T-lymphoblastic lymphoma or blastic T/myeloid neoplasm [73]. Cases with t(8; 22) (p11.2;q11.2)/BCR-FGFR1 usually manifest with B-lymphoblastic leukemia either at presentation or at progression [74, 75]. FGFR1 rearrangement is not cryptic, and conventional cytogenetics demonstrates 8p11 abnormalities in almost all cases. FISH with break-apart probes for FGFR1 confirms the rearrangement. Concurrent RUNX1 mutations are very common and were reported in 78% patients in the MD Anderson series [71].

The diagnostic criteria for myeloid/lymphoid neoplasms with FGFR1 rearrangement are proposed in the 2016 revised WHO classification of hematolymphoid neoplasms as follows: (1) An MPN or MDS/MPN with prominent eosinophilia and sometimes with neutrophilia or monocytosis; OR (2) AML, T- or B-lymphoblastic leukemia/lymphoma (T-LBL), or mixed-phenotype acute leukemia (usually associated with PB or BM eosinophilia) AND the presence of t(8; 13) (p11.2;q12) or a variant translocation leading to FGFR1 rearrangement, demonstrated in myeloid cells, lymphoblasts, or both.

The recognition of hematolymphoid neoplasms presenting with eosinophilia associated with tyrosine kinase alterations has led to discoveries of new recurrent molecular genetic alterations in cases lacking PDFGFRA, PDGFRB, or FGFR1 rearrangements. An example is eosinophilia associated with t(8; 9) (p22;p24.1)/PCM1-JAK2 [76], which has been incorporated in the group as a new provisional entity in the 2016 WHO revision.

The t(8; 9) (p22;p24.1) with PCM1-JAK2 rearrangement should be distinguished from t(8; 9) (p11;q33)/ CEP110-FGFR1. This rare entity is characterized by a combination of peripheral eosinophilia and a MPN. BM often shows characteristic large immature erythroid islands/nodules, lymphoid aggregates, and often myelofibrosis, mimicking primary myelofibrosis (PMF). Cases of PCM1-JAK2 may also rarely present as T-cell or B-cell lymphoblastic leukemia or AML. While response to tyrosine kinase inhibitors is poor, targeted therapy with JAK2 inhibitors may offer potential benefit [77]. Some patients have achieved an excellent outcome after allogeneic stem cell transplant [7].

Several “variants” of this provisional entity have been suggested that JAK2 is fused to other genes other than PCM1 including t(9; 12) (p24.1;p13.2)/ETV6-JAK2, and t(9; 22) (p24.1;q11.2)/BCR-JAK2. However, these “variant” cases are clinicopathologically heterogeneous. Cases of B-ALL associated with ETV6-JAK2 may have the features of BCR-ABL1–like B-ALL [62].

After a thorough and extensive work-up that excludes reactive eosinophilia, eosinophilia associated with a well-categorized hematopoietic neoplasm as well as cases with rearrangements of PDGFRA, PDGFRB, or FGFR1 or with PCM1-JAK2, the challenge remains to distinguish true myeloid neoplasm from the rest of cases in which a clear cause is not identified. The latter is referred to as idiopathic HES, defined as persistent HE (≥6 months) and tissue/organ damage with no known causes (Table 3).

CEL, NOS is currently classified as a MPN characterized by persistent eosinophilia [9]. Organ damage due to eosinophils releasing cytokines or enzymes may occur but is not required for the diagnosis. CEL, NOS is a diagnosis of exclusion that requires the exclusion of reactive HE as well as other well-defined molecular genetic or clinical entities. On the other hand, CEL, NOS is also a diagnosis of inclusion in that BM and PB should demonstrate features of a myeloid neoplasm [78]. In the WHO criteria [6, 79], the neoplastic nature of CEL, NOS is defined by the presence of increased blasts in BM or PB, or proof of clonality of hematopoietic cells (Table 2). In the past, clonality had been mainly determined by chromosomal analysis or mutations well known to occur in MPNs, such as JAK2, MPL, CALR, or KIT. However, it has been shown that these mutations are extremely uncommon in CEL, NOS [80, 81]. In recent years, using NGS technology, somatic mutations associated with myeloid neoplasms have been detected in 25–30% of patients who have a normal karyotype and no increase in blasts and who would otherwise be considered to be idiopathic HES. Mutations by NGS have been found mostly in genes involved in DNA methylation and chromatin modification, such as ASXL1, TET2, EZH2, and DNMT3A [81, 82]. While a positive mutation provides evidence of clonality, these mutations have been reported in aging individuals lacking evidence of a myeloid neoplasm [83, 84]. Morphological features of the BM have been found to be extremely valuable in identifying cases with clinical and biological features of a myeloid neoplasm and are strongly recommended to be incorporated in the diagnosis of CEL, NOS (Fig. 3) [85].

Clinically, CEL, NOS, and HES are rare diseases [86]. There is a male predominance in both disease entities; however, the median age is in the 4th decade for idiopathic HES and in the 6th decade for CEL and NOS [81]. Organ damage due to an eosinophilic infiltrate or eosinophil activation is, by definition, seen in all patients with idiopathic HES and frequently involves skin, lung, and gastrointestinal tract [87]. On the other hand, CEL, NOS patients frequently present with constitutional symptoms, anemia, abnormal platelet counts, and symptoms related to cytopenias, frequent organomegaly and elevated LDH. Symptoms related to eosinophilic activation, such as allergy, urticaria/rash, edema, asthma, myalgia/arthralgia, or eosinophil-mediated organ injury, are less frequent in CEL, NOS.

Patients with CEL, NOS who either have a karyotypic abnormality and/or increased blasts have a median overall survival of 15–22 months [81, 88]. Most CEL, NOS patients either die of AML progression or BM failure. In contrast, the survival of idiopathic HES patients is highly variable, most likely due to the difficulty in separating a neoplastic disease from reactive/secondary HES. The overall survival is reported as 80% at 5 years [81, 89]. And the most common causes of death of patients with HES are heart failure or brain infarction or complications of long-term corticosteroid use [9]. For patients with mutations identified by NGS, the clinicopathological features and survivals are intermediate between CEL, NOS with an abnormal karyotype or increased blasts and idiopathic HES with no mutations [81, 90].

BM morphology is not an integral part in the diagnosis of CEL, NOS. This is in contrast to the diagnosis of other myeloid neoplasms with no signature molecular genetic aberrancies. In Ph-negative MPNs, including essential thrombocythemia, Polycythemia Vera and PMF, BM morphology is one of the major diagnostic criteria [91]. The argument is that morphological changes in eosinophils are not unique to a neoplastic eosinophilic disorder [78, 79, 92-94]. Of note, a recent multicenter study that reviewed the BMs of 139 patients with a diagnosis of idiopathic HES or CEL, NOS [85], reported that the abnormal BM (and some PB) features are not only limited to eosinophils, but also include BM hypercellularity, abnormal megakaryocytes, dyserythropoiesis, dysgranulopoiesis, MF2 or greater fibrosis, and so on in a true neoplastic HE. These features were similar to those seen in MDS, MDS/MPN, and/or BCR-ABL1 negative MPNs. Cytological abnormalities in eosinophils could be seen in reactive processes and HES [95], but the abnormalities are much more severe and frequent in CEL, NOS [81, 85] (Fig. 3). On the other hand, the BM in HES is usually normal except for a slight hypercellularity due to increased eosinophils. Furthermore, abnormal BM morphology has been found to correlate with an abnormal karyotype and the presence of myeloid neoplasm related mutations and is an independent prognostic factor.

Even with the advent of NGS, the distinction between CEL, NOS and idiopathic HES remains challenging. The presence of a clonal karyotypic abnormality or increased blasts supports a diagnosis of CEL, NOS; however, as in the other MPNs, karyotypic abnormalities or increased blasts only present in a subset of cases. Myeloid neoplasm-associated somatic mutations can be used as evidence of clonality. While the detection of mutations are more specific for a myeloid neoplasm (TP53, EZH2, SETBP1, NRAS, CSF3R, JAK2) and the numbers of mutations (more than 1 mutations in myeloid neoplasm associated genes) are more in favor of a neoplastic process, one needs to be extremely cautious with the caveat of prevalent “clonal hematopoiesis of indeterminate potential” (CHIP) in the elderly patients [83, 84]. Approximately 20% of patients who are currently classified as idiopathic HES demonstrate abnormal BM morphology. It is advised that finding abnormal BM features should prompt NGS studies in helping the differential diagnosis of CEL, NOS versus idiopathic HES.

In summary, the causes of HE are complex; however, by following a proper algorithm and performing cost-effective molecular genetic testing, an underlying cause can be identified in the majority of the cases. However, there still remains a significant subset of cases in which the etiology of HE is not apparent, and these are currently classified as idiopathic HES. It is unclear if all these cases are reactive to an unidentified stimulus or whether a subset may be truly neoplastic even though they are not currently recognized by morphology and molecular genetic investigation. Future studies are required to uncover the pathogenesis of the underlying eosinophilic proliferation that will ultimately be needed to better manage these patients.