The Reliability and Specificity of c-kit for the Diagnosis of Acute Myeloid Leukemias and Undifferentiated Leukemias

THE c-KIT PROTO-ONCOGENE encodes a 145-kD tyrosine kinase transmembrane receptor, which plays a key role in hematopoiesis, as it is the receptor for the stem cell factor (SCF). Expression of c-kit can be detected by immunostaining with monoclonal antibodies (MoAb) clustered under CD117 or by Northern blot or reverse transcriptase-polymerase chain reaction (RT-PCR), which allows detection of c-kit transcripts.

c-kit is expressed in a minority (<5%) of bone marrow progenitor cells, half of which express CD34 and some appear to be committed to the myeloid lineage.1-3 A variety of human leukemia cell lines, in particular those committed to the myeloid, erythroid, and megakaryoblastic lineages or with mutipotential capacity are also c-kit positive.2,4,5 In addition, c-kit expression has been documented in a minority of early human thymocytes located at the subcapsular cortex and at the intralobular septa of the thymus, which have a CD7⁺CD4^-CD8^-CD3^-phenotype, and are able to differentiate in vitro to the lymphoid and nonlymphoid lineages.2 Similarly, bone marrow CD34⁺CD2⁺CD7⁺ stem cells are able to differentiate towards the T-cell pathway and such differentiation is enhanced by the c-kit ligand.6 These findings suggest that CD34⁺ CD7⁺ c-kit⁺ cells are indeed pluripotent lymphohematopoietic cells and that c-kit plays a major role in early T-cell development.2,6 Similar data have been reported on murine prothymocytes, which are also c-kit positive and have germ-line configuration of the T-cell receptor (TCR) β/γ chain genes.7 Findings have been controversial as to whether a minority of bone marrow progenitor B cells express c-kit.1,8 Using sensitive techniques, reactivity with c-kit has been detected in 4% of bone marrow CD19⁺ cells from some, but not all, normal individuals,8 and although cord blood CD19⁺ lymphoid progenitors may have a moderate expression of the SCF receptor, the antibody reactivity has not been assessed in these samples.9 In addition, c-kit is expressed on a small fraction of natural killer CD56⁺ cells, which display both forms, the high and intermediate affinity of the interleukin-2 receptor.10 

In parallel to studies in normal hematopoietic cells, there have been a number of reports on c-kit expression in acute leukemia cells. Most of them have shown that a substantial proportion of cases, particularly acute myeloid leukemias (AML), are c-kit positive (for review, see Muroi et al11). Although the reactivity with c-kit has been mainly confined to the myeloid leukemias, its specificity for the myeloid lineage has been questioned, as some reports have observed expression of c-kit in few cases of early T-acute lymphoblastic leukemia (ALL)12 and weak reactivity in some B-cell lineage ALL using sensitive techniques.3 A letter by Muroi et al11 summarized the reported findings on c-kit expression in acute leukemias and based on the evidence that c-kit was, as a rule, negative in ALL, the investigators concluded that c-kit is a specific marker for the myeloid lineage. However, the number of ALL cases included in all of these reports was relatively low, with a total of 189.

On the EGIL proposals,13 the members did not dispute the myeloid specificity of c-kit, but felt that its reactivity needed to be tested on a large scale basis to validate or not the specificity of this marker, as well as its sensitivity for the myeloid lineage.

We report here findings obtained with MoAb against c-kit in a large group of cases with acute leukemias investigated in five European institutions or working groups and discuss the reliability of this marker for the diagnosis of AMLs.

Material from 1,937 acute leukemia patients diagnosed in institutions from five European countries has been analyzed. Most of the patients had been enrolled on national acute leukemia trials. The diagnosis was made by (1) standard morphology and cytochemistry of peripheral blood and bone marrow films according to the French-American-British (FAB) criteria14 and (2) immunophenotyping using a comprehensive panel of MoAbs against myeloid and lymphoid-associated antigens, as proposed by the EGIL group.13 Cases with a diagnosis of blast transformation of chronic myeloid leukemia and/or other myeloproliferative disorder were excluded. Of 1,937 patients, 1,211 were adults and 726 children. All of the samples contained over 30% of blasts and usually more than 50%.

Peripheral blood and/or bone marrow mononuclear cells were isolated by a gradient density centrifugation and washed twice in Hank's balanced salt solution before immunophenotyping. The MoAb YB5B8 (Immunotech, Marseille, France) against the c-kit receptor was used in 1,665 samples and the MoAb 95C3 (Immunotech) in the remaining 272. Fifty samples were tested with these two MoAbs and with 102B4E1 (a gift from Dr Buhring, Tubingen). Negative controls included the omission and/or substitution of the MoAb by a mouse Ig of a matched isotype. Analysis was performed by flow cytometry with an indirect immunofluorescence technique using a fluorescein isothiocyanate (FITC)-conjugated antimouse Ig as second layer. The blast population was gated using morphologic parameters (ie, size of cells as measured by forward scatter and granularity as measured with side scatter signals). A case was considered positive when over 20% of blasts reacted with the MoAb according to the EGIL criteria.13 

The 1,937 cases were distributed in the following immunologic subgroups: (1) ALL (n = 819), 612 B-cell lineage and 207 T-cell lineage. The B- and T-cell lineage ALL were further subclassified in four subtypes according to the level of differentiation of the blasts.13 Briefly, these comprised in B-cell lineage ALL (CD19⁺, CD79a⁺, and/or CD22⁺; TdT⁺); B-I (no expression of other B-cell differentiation antigens); B-II (CD10⁺); B-III (cytoplasmic IgM⁺) and B-IV (cytoplasmic or surface immunoglobulin light chain⁺) and, in T-lineage ALL (CD3⁺, TdT⁺), T-I (CD7⁺); T-II (CD2⁺ and/or CD5⁺ or CD8⁺); T-III (CD1a⁺) and T-IV (membrane CD3⁺).

(2) AML (N = 1,103) were found in which blasts were positive with one or more myeloid markers. According to the FAB classification, AML cases corresponded to 106 M0, 191 M1, 258 M2, 59 M3, 187 M4, 125 M5, 43 M6, and 13 M7; the FAB subtype was not available in the remaining 121 cases.

(3) Eleven cases were classified as biphenotypic acute leukemia (BAL) and these comprised nine cases with a B-lymphoid and myeloid phenotype, one with a T-lymphoid and myeloid phenotype, and one case with trilineage differentiation; all of the cases had scores >2 for the myeloid and lymphoid lineages according to the proposed criteria.13 BAL was defined when blasts scored over 2 points for the myeloid lineage and one for the lymphoid lineages. The number of points assigned to each marker is based on the specificity of the marker for the myeloid and lymphoid lineages, as shown in Table 1. We did not assign myeloid points to c-kit reactivity, as the aim of this study was to test the specificity of c-kit for the myeloid lineage.

(4) The remaining four cases were considered as acute undifferentiated leukemia (AUL), as they could not be classified by cytochemistry and by the expression of lymphoid and myeloid markers; blast cells from these four cases were only positive with CD38, HLA-DR, CD7, CD34, and/or TdT antibodies.

Blast cells from 783 of the 1,937 cases (40%) were c-kit positive. These included all four AUL, two thirds of AML, and one third of BAL, whereas only a minority of ALL were c-kit positive (Table 2).

Results on the reactivity with c-kit in the AML group according to the FAB subtypes are shown in Table 3. A varying proportion of cases with all FAB subtypes were c-kit positive with the higher percentage of positive cases in the subtypes with a myeloid component: M0, M1, and M2, while only one third of M5 and M7 expressed c-kit (Table 3). Of the four cases with BAL, which were c-kit positive, two corresponded to the only ones with a T-lymphoid and myeloid component (one case) and to the case with trilineage differentiation, while the other two cases had a mixed B-lymphoid and myeloid phenotype.

Only 4% of ALL cases were c-kit positive, and of these, two thirds were T-lineage ALL and the other third B-lineage ALL, representing, respectively, 11% c-kit positive cases within T-ALL and 2% c-kit positive in the B-cell lineage ALL group. The stage of maturation and immunophenotypic characteristics of the c-kit positive ALL are summarized in Table 4. Half of the T-ALL corresponded to pro-T–ALL (T-I) and the other half had a more mature T-cell immunophenotype (TII-TIV). All of the c-kit positive B-cell lineage ALL had either a common-ALL (B-II) or pre-B–ALL (B-III) immunophenotype, and only one case had a mature B-ALL (B-IV) immunophenotype. Cells from 21 of the 34 c-kit positive cases (62%) expressed one or two other myeloid markers: CD13, CD33, or both, but none qualified as BAL. Of these 21, 13 were T-lineage ALL and the other seven were B-lineage ALL (Table 4).

In the 50 cases in which the three MoAb against c-kit were investigated, there were no significant differences in the proportion of c-kit positive cells.

Overall, there were no substantial differences regarding the incidence of c-kit positive cases in pediatric and adult leukemias, although the proportion was slightly higher in adult AML, 69% c-kit⁺(613 of 883) versus 58% (128 of 220) in pediatric AML cases. In ALL, 4.6% (23 of 501) of pediatric samples and 3.4% (11 of 318) of adult samples, respectively, were c-kit positive. Considering BAL, c-kit expression was seen in both, children (1 of 5 [20%]) and adults (2 of 6 [30%]) and all four AUL were adults.

There was a correlation between expression of CD34 and c-kit in AML, as 76% of cases either expressed both antigens or were negative with the two MoAb, while the remaining 24% of cases expressed only one of the two antigens. In contrast, there was no evidence that c-kit⁺ AML had a higher frequency of lymphoid antigen expression, being the proportion of cases positive with CD2 (9%), CD7 (25%), CD10 (1%), and CD19 (7%) similar to c-kit negative cases.

Our findings, based on a large number of well characterized acute leukemia cases, clearly establish the high specificity of c-kit for leukemic cells committed to the myeloid lineage and thus, support its value as a diagnostic reagent in the characterization of acute leukemias. Although the sensitivity of c-kit to detect AML seems to be slightly lower than CD13 and CD33, particularly in cases with a monocytic component, its specificity is significantly higher, as the latter two markers are expressed in a significantly larger proportion of ALL cases (for review, see Drexler et al15). When considering AML, our data show that c-kit is more often expressed in CD34⁺ AML, but do not support that it is preferentially associated with a particular FAB subtype. Although c-kit is more often positive in cases with myeloid/granulocytic differentiation (eg,, M0 to M2) compared with those with proliferation of monoblasts, it cannot be taken as a marker discriminatory of the various AML subtypes.

The analysis of the immunophenotypic characteristics of the minority of c-kit positive ALL showed that two thirds of these cases were also positive with one or two myeloid markers and/or corresponded to pro-T-ALL. Thus, c-kit seems to identify a minor group of ALL, which has either early T-cell features and/or expresses other myeloid antigens. This suggests that such leukemias arise from primitive progenitor cells or represent the leukemic counterpart of the normal prothymocyte, which is c-kit⁺, CD7⁺. Indeed, the four c-kit⁺ ALL cases reported by Nishi et al12 had myeloid antigen expression, only two expressed CD3 in the cytoplasm, and all had the TCR β/γ chain genes in germ line configuration, features that fit well with a very early hematopoietic cell.

Of interest is our finding that all four cases with undifferentiated acute leukemia were c-kit positive. None of these cases expressed myeloid or lymphoid differentiation markers, but most were CD7⁺,CD34⁺. Such an immunophenotype is similar to that found in a small proportion of bone marrow and/or early thymic cells2,6 able to differentiate to myeloid and lymphoid cells and perhaps represent an earlier step in differentiation than that of the c-kit⁺ pro-T–ALL.

In summary, MoAb recognizing the c-kit receptor appear to be useful reagents recommended to be routinely used in the panel of markers for the diagnosis of acute leukemias to demonstrate or support the myeloid commitment of the blasts in both pediatric and adult patients. Due to its now well established specificity for the myeloid lineage, the c-kit marker should score for the myeloid lineage at least one point in the scoring system for the diagnosis of biphenotypic acute leukemias proposed by the EGIL.13