Developments Before 1991

Because lineage infidelity in leukemic blasts could not be described until lineage-associated markers became widely available, recognition of the more peculiar varieties of leukemia is, in the history of hematological oncology, a relatively recent development. This development is closely tied to the advent of routine cytochemical evaluation of marrow specimens in the 1970s, introduction of monoclonal antibodies in the early 1980s, and increasing use of multiparameter flow cytometry in leukemia phenotyping during the late 1980s and early 1990s.

Before 1980, clinicians and morphologists assigned leukemia diagnoses on the basis of clinical presentation and cell morphologic features, such as the cytoplasmic appearance and presence or absence of Auer rods, supplemented by basic cytochemical stains, such as Sudan Black B, chloroacetate esterase, and the periodic acid-Schiff reaction. The lineage and corresponding cytologic appearance and immunophenotype of blast cells in acute leukemia was believed to cleanly reflect malignant transformation and maturation arrest at a particular stage of hematopoietic differentiation, with acute lymphoid malignancies originating in lymphoid progenitors and acute myeloid malignancies originating in early myeloid series cells.

Several observations in the 1980s challenged this assumption, however. Expression of the terminal deoxynucleotidyl transferase (TdT) enzyme, thought to be an exclusive marker of lymphoid lineage for several years following its discovery by McCaffrey and colleagues in blasts from children with ALL in 1973, was found by 1981 to occur in 10% to 20% of cases of AML. (In fact, if more sensitive techniques are used, TdT activity is detectable in blasts in as many as 55% of AML cases, albeit usually at a much lower level of expression than is typically seen in TdT-positive ALL. ) Similarly, in the 1980s morphologists also learned that myeloperoxidase (MPO) expression, considered to be a myeloid lineage-defining marker after its initial description in the 1940s, is detectable by either cytochemistry or immunohistochemistry in more than 20% of B-ALL cases.

Additionally, in the late 1970s and early 1980s, a series of case reports and small series appeared in which investigators described detecting either 2 distinct populations of blasts of differing putative origin simultaneously in the same patient (“bilineage” leukemias), or a homogeneous population of blasts expressing markers of more than one lineage (“biphenotypic” leukemias). The relationship between bilineage and biphenotypic leukemias remained unclear, and in some instances leukemic blasts that expressed markers most suggestive of one cell differentiation program at the time of initial diagnosis would subsequently exhibit a different immunophenotype consistent with another lineage at relapse (eg, B-cell ALL at diagnosis, but AML without lymphoid lineage markers at relapse following ALL-directed chemotherapy).

In 1985, Mirro and colleagues assessed 123 consecutive pediatric leukemia cases diagnosed at St Jude Children’s Research Hospital in Memphis, Tennessee, or in Calgary, Alberta, Canada, using a panel of myeloid and lymphoid-associated antibodies, and found that ALL blasts coexpressed at least one myeloid marker in 19% of cases, whereas the leukemic blasts in 25% of AML cases expressed at least one lymphoid-associated marker. Although this landmark study highlighted the frequency of lineage infidelity in leukemia, the antibodies used to assign lineage in the study by Mirro and colleagues (ie, CD2 [T-11 antibody], CD3e [T-3], CD5 [T-101], and CD10 [J5 CALLA] for lymphoid; CD11b [Mo1 antibody], CD13 [MCS.2 and SJ-D1], CD15 [My-1], and CD36 [5F1] for myeloid) would, with the exception of CD3e, no longer be considered lineage-defining markers.

Investigators compiling adult leukemia series soon noted that a large proportion of patients with AML or ALL express at least one lineage infidelity marker, but that the presence of only a single such antigen does not alter clinical behavior or response to treatment. Further complicating matters, some markers initially thought to be tightly lineage-associated later proved to be so commonly expressed in malignancies of other lineages as to no longer even be considered aberrant. One such example is the CD15 carbohydrate adhesion molecule, which is expressed on normal neutrophils and on blast cells in most AML cases so was once considered myeloid-associated, but is also very commonly detected in B-cell ALL lacking CD10 expression and in other non-myeloid neoplasms (eg, Hodgkin lymphoma).

Development of European Group for the Immunologic Characterization of Leukemias Criteria

A growing number of case reports and small series in the 1980s highlighted increasing diagnostic confusion related to mixed lineage acute leukemias, and the need for more formal definitions of these somewhat puzzling disorders became increasingly clear. Catovsky and a group of European colleagues proposed the first scoring system for defining biphenotypic acute leukemia (BAL) in 1991, which weighed various cytochemical/immunohistochemical and immunophenotypic markers according to their perceived lineage specificity.

In 1995, the European Group for the Immunologic Characterization of Leukemias (EGIL) refined this 1991 system ( Table 2 ) to make it somewhat more stringent, requiring a lineage-associated marker score of more than 2 points in each of 2 lineages to define BAL, as well as demoting several markers that Catovsky and colleagues had proposed as supporting lineage assignment (ie, immunoglobulin heavy chain gene rearrangement for B lineage; CD11b/c and CD12 for myeloid lineage) while adding or refining several others (ie, CD20 and CD79a for B lineage; TdT, CD1a, CD7, CD8, and CD10 for T lineage; CD13, CD64, and CD65 for myeloid lineage). The EGIL system was modified in 1998 to add CD117 as a putative myeloid-specific marker, but soon a few ALL cases were found that expressed CD117.

Clinicians and investigators quickly raised concerns about the EGIL criteria, such as the lack of specificity of certain EGIL-proposed markers, especially CD79a expression for defining B lineage and MPO positivity for defining myeloid lineage. In addition, the EGIL system used varying and arbitrary thresholds of cellular antigen expression required to qualify as “positive” (eg, 20% for most markers, but 10% for MPO, CD3, TdT, and CD79a), did not distinguish between biphenotypic and bilineage leukemia, did not assess intensity of antigen expression, and also failed to incorporate cytogenetic data. The latter proved to be a particularly troublesome limitation, because leukemia subtypes that were already well defined by routine G-banded karyotyping by the time the EGIL system was introduced (and, later, were further clarified using molecular genetic techniques) often express a degree of lineage infidelity, yet have other uniform features or may respond similarly to therapy regardless of immunophenotype. For example, t(8;21) AML often expresses a subset of B-cell antigens, whereas acute promyelocytic leukemia (APL), especially the hypogranular variant, frequently expresses T-cell markers, particularly CD2, and T-cell marker expression in APL has no effect on response to all-trans retinoic acid therapy. In addition, CD79a can be found in both T-ALL and myeloid leukemia, so lacks the B-cell specificity accorded to it by the EGIL system.

World Health Organization Criteria: Third and Fourth Editions

In the third edition of the WHO Classification of Tumours of Haematopoietic and Lymphoid Tissues proposed in 1999 and published in 2001, the WHO working group included a 2-page description of a classification of “acute leukemias of ambiguous lineage,” composed of 3 categories: bilineage, biphenotypic, and undifferentiated acute leukemias. Although the third edition WHO classification referenced the 1995 EGIL system (with the 1998 CD117 addition), the WHO third edition text unfortunately included a typographical error in its Table 4.03, stating that only 2 points, rather than more than 2 points, were necessary to assign a leukemia case to a given lineage. In one series, these slightly less stringent criteria tripled the number of cases defined as biphenotypic compared with EGIL criteria, and the reassignment was of dubious clinical value.

Perhaps because of this debacle, the 2008 fourth edition of the WHO classification did away with lineage scoring entirely in favor of descriptive but relatively specific requirements for assigning more than one lineage to a single blast population ( Table 3 ). Although the WHO fourth edition classification retained the category of acute leukemias of ambiguous lineage, the updated system also incorporated 5 major changes with respect to the third edition classification.

First, because of frequent ambiguity about whether mixed lineage markers are truly coexpressed by the same blasts or by 2 distinct populations of leukemic cells, WHO did away with the terms “bilineal” and “biphenotypic” in favor of the collective term “mixed phenotype acute leukemia” (MPAL). Second, the criteria that define the myeloid, T-lymphoid, and B-lymphoid components of MPAL were altered and simplified (see Table 3 ), although the same criticisms persisted about lineage specificity of proposed markers. Third, cases that are clearly acute leukemia with recurrent genetic abnormalities were excluded from MPAL, even if they express a mixed lineage phenotype that would otherwise qualify. Such entities include cases classified as AML with recurrent genetic abnormalities (eg, t[15;17], t[8;21], and inv[16]), but the WHO committee did not feel comfortable extending this “trumping” of immunophenotype by genetic and molecular markers to provisional entities, such as AML with mutated NPM1 or CEBPA. Similarly excluded regardless of lineage expression are cases known to represent chronic myeloid leukemia (CML) in blast phase, therapy-related AML, or AML arising from a myelodysplastic syndrome.

The fourth change between third and fourth editions of the WHO classification was inclusion of 2 molecularly defined diagnostic categories in the acute leukemia of ambiguous lineage group: BCR-ABL1 –positive and MLL -rearranged acute leukemias. In many series of EGIL-defined BAL, t(9;22) and t(11q23) translocations have been overrepresented, probably because loss of ABL tyrosine kinase activity or MLL fusions with associated HOX dysregulation severely disrupts normal differentiation programs. Indeed, the MLL gene took its name from its association with “mixed lineage leukemia” or “myeloid-lymphoid leukemia.”

Unfortunately, although well intentioned, these molecularly defined acute leukemia of ambiguous origin categories quickly caused confusion. MLL-rearranged acute leukemias without a mixed lineage phenotype are elsewhere classified by WHO both as “AML with t(9;11)(p22;q23); MLLT3-MLL” and variants, and as “B lymphoblastic leukemia/lymphoma with t(v;11q23); MLL rearranged.” These entities may be seen with a much greater frequency to MPAL with MLL rearrangement, and the prognosis and treatment are determined by the genotype rather than the immunophenotype.

In addition, patients who previously had unrecognized chronic myeloid leukemia (CML) and first present to medical attention at a late stage (ie, in blast crisis), are not easy to distinguish from those with BCR-ABL1 –positive de novo ALL or AML, as p190 and p210 transcripts can be detected in either setting. Although the WHO classification states, “this diagnosis (ie, MPAL with t[9;22][q34;q11.2; BCR-ABL1]) should not be made in patients known to have had CML,” the stage at which the patient presents for medical attention is a poor criterion on which to base a disease classification, as delay in diagnosis does not reflect a genuine biologic difference and may be because of socioeconomic or other personal factors.

A fifth change between the 2001 and 2008 classifications is that in the 2008 edition, the WHO committee also included an “other ambiguous lineage leukemias” category for truly bizarre cells. For instance, leukemia cases that express T-cell lineage markers but not cytoplasmic CD3, or cases that express several myeloid markers but not MPO, do not fit into any of the other MPAL categories.

These cases are considered distinct from cases with no lineage-specific markers at all, which WHO designates as acute undifferentiated leukemia (AUL) and should be diagnosed only after a cautious, thorough evaluation. AUL cases should express CD34, HLA-DR, or CD38—something to exclude the possibility of a nonhematopoietic neoplasm—but, by definition, lack specific myeloid or lymphoid antigens. Some AUL cases have a t(9;9) translocation that results in fusion of SET, a nuclear phosphoprotein that inhibits activity of protein phosphatase 2A (PP2A), to the oncoprotein CAN/NUP214 involved in the more common t(6;9) in AML.

Finally, the difficult-to-define entity of natural killer cell lymphoblastic leukemia/lymphoma was included as a provisional category within the miscellaneous “other ambiguous lineage leukemias” category in the fourth WHO classification, whereas those cases more clearly designated as originating from plasmacytoid dendritic cells (also CD56 positive, like natural killer cells) were reclassified as a new entity, blastic plasmacytoid dendritic cell neoplasms (BPDCNs). By definition, BPDCNs coexpress CD4 and CD56, and usually also express CD123 or CD303 (BDCA2) (see Table 1 footnotes).

Limitations of Current Criteria

In addition to the concerns raised about lineage specificity of particular markers and variable intensity of marker expression, clinicians have expressed concerns that to some extent trying to pigeonhole leukemias of ambiguous origin is an exercise in futility. It is not clear whether, for example, T/myeloid leukemia and B/myeloid leukemia are genuinely distinct conditions requiring a different therapeutic approach, and whether the WHO 2008 criteria are really defining “real” entities in the sense of clinically actionable unique disorders.

There are also technical concerns. Currently, pathologists primarily perform immunophenotyping of new leukemia cases by flow cytometry, and flow is to some extent operator-dependent, with results varying depending on such parameters as gating and reagent quality. In addition, most clinical laboratories, even those at major academic centers, normally do not use all the flow markers in the EGIL panel or even all of the major WHO markers, so biphenotypic/bilineage expression may be missed. A conference convened in 2006 in Bethesda, Maryland, defined a consensus set of reagents suitable for general use in the diagnosis and monitoring of hematopoietic neoplasms, but did not recommend the full panel necessary for defining biphenotypic leukemia according to EGIL criteria. For example, many laboratories do not routinely perform cytoplasmic CD79a as part of a standard new leukemia diagnostic flow panel, in part because as stated, this “B-lineage–specific” marker also is regularly found in cases of AML (especially t[8;21] or APL ) and T-ALL.