Blastic natural killer cell lymphoma/leukemia (CD56-positive blastic tumor)

Prognostication and categorization according to anatomic sites of involvement

Authors


Abstract

BACKGROUND

Blastic natural killer (NK) cell lymphoma/leukemia (BNKL) is an immature CD56-positive neoplasm, which was recognized recently and characterized by systemic proliferation of tumor cells including skin, lymph node, and bone marrow.

METHODS

The current study analyzed 47 patients with BNKL (27 had leukemias and 20 had lymphomas). Patient data were collected for the survey of the NK-Cell Tumor Study Group.

RESULTS

There were 33 males and 14 females, with a median age of 53 years (range, 3 months to 89 years). There were few clinicopathologic differences between the leukemia and lymphoma types. Cutaneous involvement was noted at diagnosis in 28 patients, who presented a tendency for older age of onset (median: 56 vs. 46 years, P = 0.11) than patients with noncutaneous BNKL. Cutaneous BNKL showed less frequent mediastinal involvement (4% vs. 53%, P = 0.0002) and less severe thrombocytopenia (P =0 .03). Phenotypic characteristics were also different, with cutaneous BNKL favoring CD4 and HLA-DR expression, and noncutaneous BNKL favoring CD16 and CD34 expression. Both groups responded well to chemotherapy for lymphoid malignancies, but disease recurrence was frequent. The prognosis of patients with noncutaneous BNKL was significantly poorer than that of patients with cutaneous BNKL (median survival: 15 vs. 25 months, P = 0.02). Multivariate analysis confirmed that cutaneous involvement was a significant and independent prognostic factor for BNKL, as were age of onset and leukocyte count.

CONCLUSIONS

These findings suggested that BNKL is a heterogeneous disease and contains at least two subtypes. Although further investigations are needed to settle a marker for distinction, the presence of cutaneous involvement is a useful prognostic factor. Cancer 2005. © 2005 American Cancer Society.

The conceptual view of neoplasms derived from natural killer (NK) cells is changing and evolving.1 CD56-positive immature hematolymphoid tumors, which do not match any known disease entities, have been identified and postulated as precursor NK cell neoplasms.2, 3 Although the absence of specific markers makes it difficult to clearly establish the origin of these malignancies, the World Health Organization (WHO) classification system lists blastic NK cell lymphoma (BNKL) as a tumor of uncertain lineage and stage of differentiation.4

After its recognition in the Hong Kong workshop for NK cell lymphomas held in 1996,5 BNKL became well recognized as a disease entity during the last decade.6 Although BNKL is a relatively rare disease, numerous studies comprising series or case reports have attempted to characterize this unique hematolymphoid malignancy.7–9 BNKL is a systemic proliferation of CD56-positive immature cells, and is characterized by a predilection for skin/soft tissue involvement and an aggressive clinical course.7–9 Phenotypically, BNKL is characterized by the expression of CD4 and terminal deoxynucleotidyl transferase (TdT),4 but such expression is not exclusive. Another distinguishing feature is acute lymphoblastic leukemia (ALL) of NK cell lineage, which may represent a leukemic form of BNKL.10–18 From comparative studies of the leukemic and lymphomatous forms of precursor T and B-lymphoblastic leukemia/lymphoma,19, 20 it has been concluded that the two forms should be categorized as single entities.21 Although leukemic manifestations of bone marrow (BM) and/or peripheral blood (PB) involvement often appear during the course of disease,22–26 no attempts have been made to subject BNKL to such a comparison until now.

Several authors have proposed recently that CD4-positive CD56-positive immature leukemia/lymphoma is derived from plasmacytoid monocytes or plasmacytoid dendritic cell (pDC) precursors.27–29 The phenotypic and clinicopathologic spectra of BNKL and CD4-positive CD56-positive immature leukemia/lymphoma appear to overlap to some extent. For a better understanding of CD56-positive immature hematolymphoid tumors, the NK-Cell Tumor Study Group conducted a nationwide surveillance in Japan,30, 31 and analyzed the results for identification of the characteristics of CD56-positive blastic tumors (BTs).

MATERIALS AND METHODS

Patients

Between 1994 and 1998, 47 patients were diagnosed with BNKL of either the lymphomatous or the leukemic form at collaborating institutions of the NK-Cell Tumor Study Group. The current study was approved by the working committee of the NK-Cell Tumor Study Group and the institutional review boards of participating institutions. An initial diagnosis was made at one of the institutions, and was reviewed whenever the original histologic material was available. Specimens were reviewed by two expert hematopathologists (S.N. and J.S.) and clinical data were evaluated by the diagnostic committee (R.S., K.K., and K. Oshimi). Only patients diagnosed with BNKL according to the WHO classification were included.4 The patients' follow-up data were reviewed in December 2002.

Histopathology and Immunophenotyping

Tissue specimens were fixed in 10% formalin and embedded in paraffin. Sections (5-mm thick) were stained with hematoxylin and eosin, periodic acid-Schiff, Giemsa, and Gomori silver impregnation. Immunohistochemical studies were performed as described previously.32 Briefly, expressions of various antigens in the paraffin-embedded sections were examined by using the avidin-biotin peroxidase complex method. The antibodies comprised Ber-H2/CD30 (Dako, Santa Fe, CA), LeuM1/CD15 (Becton-Dickinson, Sunnyvale, CA), L26/CD20 (Dako), CD79a (Dako), UCHL1/CD45RO (Dako), MT1/CD43 (Bio-Science Products, Emmenbrucke, Switzerland), CD3 (Dako), CD4 (Novocastra Laboratories, Newcastle, Uniteg Kingdom), CD8 (Dako), E29/EMA (Coulter Immunology, Hialeah, FL), CD56 (Novocastra), Leu7/CD57 (Becton-Dickinson), LMP-1 (Dako), DO-7/p53 (Dako), bcl-2 (Dako), TIA-1 (Coulter Immunology), granzyme B (M onosan, Uden, The Netherlands), bF1 (T Cell Science, Cambridge, MA), ALK1 (Dako), p80 (Dr. Mori S, The University of Tokyo, Japan), and HECA-452/cutaneous lymphocyte-associated antigen (CLA; Dr. E.C. Butcher, Stanford University, CA).33, 34 Flow cytometric immunophenotyping was performed, as described previously.35 The presence of Epstein–Barr virus (EBV) infection was determined by means of in situ hybridization with the EBV-encoded RNA probe on paraffin sections by means of the Dako hybridization kit with a cocktail of fluorescein isothiocyanate-labeled EBER oligonucleotides.3

Diagnosis

The diagnosis of BNKL was made according to both histologic and immunophenotypic characteristics as described in the WHO classification system.4 Patients were diagnosed with BNKL when the tumor cells from the biopsy specimen showed lymphoblast-like morphology and evidence of an NK cell lineage phenotype. They must express ≥ 1 NK cell lineage antigen but not any other B, T, or myeloid lineage-committed antigens (surface CD3, CD8, CD13, CD20, and CD33). CD4 was not required as a mandatory marker, because many CD4-negative patients have been reported.7, 18, 22, 36–9 Patients with > 30% tumor cells in the BM or PB were categorized as having the leukemic form of BNKL or blastic NK cell leukemia, and the others as the lymphomatous form. To exclude precursor T-lineage lymphoblastic lymphoma (LBL)/leukemia, tumor cells had to be negative for surface CD3 as demonstrated by flow cytometry and in the germline configurations of the T-cell receptor (TCR) genes.

Chemotherapy

For chemotherapeutic regimens, acute myeloid leukemia (AML)-type chemotherapy was defined as including high-dose cytosine arabinoside (ara-C) or as including ara-C administration for ≥ 5 consecutive days accompanied by anthracyclines for ≥ 3 days, as previously described.40 Regimens including multiple and weekly repeated vincristine and anthracycline administration supplemented by corticosteroids were categorized as ALL/malignant lymphoma (ALL/ML)-type chemotherapy. The ALL/ML-type regimens were further divided into CHOP (cyclophosphamide, doxorubicin, vincristine, and prednisone)/CHOP-like regimens and more intensive ALL/third-generation regimens, as described elsewhere.8

Statistical Analysis

The lactate dehydrogenase (LDH) index was calculated at each of the institutions from a patient's serum LDH level divided by the upper limit of the serum LDH level. Correlations between the two groups were examined with the chi-square test, the Fisher exact test, and the Mann–Whitney U test. Patient survival data were analyzed with the method of Kaplan and Meier, and were compared by means of the log-rank test. Univariate and multivariate analyses used the Cox proportional hazard regression model, and variables were selected with the stepwise method. Data were analyzed with Stata statistical software (College Station, TX).

RESULTS

Patient Characteristics

The characteristics of the 47 patients are listed in Table 1. Eight of the patients have been included in previous reports.41–45 Seventeen patients presented without BM tumor cells at initial diagnosis, 4 with tumor cells < 20%, and 26 with tumor cells > 30%. No patients showed BM involvement ranging from 20% to 30%. Accordingly, patients were subdivided into the lymphomatous and leukemic form of BNKL at 30% of BM involvement, but in 1 patient, tumor cells accounted for 18% of BM, but 32% of PB. This patient was included for the leukemia from of BNKL. Therefore, 20 patients were categorized as having lymphoma and 27 were categorized as having leukemia. Both groups showed a similar age of onset (median 50 vs. 55 years) and male predominance. Although the leukocyte counts was comparable for the BNKL and the leukemic group, the PB tumor cell percentage was higher and anemia and thrombocytopenia were more severe for the latter. The sites of involvement were comparable for the lymphoma and leukemia groups except for hepatosplenic diseases, which were recognized only in the leukemia group (Table 1). Extranodal involvement was observed in all patients except for two with lymphoma. Lymph nodes and skin were the most frequently affected sites.

Table 1. Patient Characteristics at Initial Onset
CharacteristicsBlastic NK cell lymphoma (n = 20) (%)Precursor NK-ALL (n = 27) (%)P value
  1. NK: natural killer cells; ALL: acute lymphocytic leukemia; PB: peripheral blood; Hb: hemoglobin; PLT: platelet.

Median age in yrs (range)50 (14–89)55 (0–79)0.97
Sex (males/females)16/417/100.33
 Median leukocyte count (μL) (range)6290 (2540–8700)5700 (1100–85,800)0.60
 Median PB blast (%), (range)0 (0–0)13 (0–85.0)<0.0001
 Hb level (g/dL) (range)13.3 (11.5–16.9)10.5 (3.9–14.1)<0.0001
 Median PLT level × 104/μL (range)22.5 (9.2–40.4)6.6 (0.8–25.4)<0.0001
 Bone marrow3 (15)27 (100)<0.0001
 Peripheral blood0 (0)19 (70)<0.0001
 Extramedullary20 (100)24 (89)0.25
Lymph node16 (80)18 (67)0.35
Liver0 (0)6 (22)0.03
Spleen0 (0)8 (30)0.01
Nasopharynx3 (15)8 (30)0.31
Skin12 (60)16 (59)0.96
Mediastinum6 (30)5 (19)0.49
Others3 (15)7 (26)0.48

Histopathology

Morphologically, all of the patients showed features of LBL. Tumor cells diffusely infiltrated into T-zone areas in the form of monotonous proliferation of immature mononuclear cells. The component cells showed a blastic morphology.

Immunophenotype

Results of immunophenotyping are listed in Table 2. All patients were positive for CD56, and negative for surface CD3, by definition. Those examined were proved to be uniformly negative for T- (CD8 and TCRs), B- (CD20 and surface immunoglobulins), and myelomonocytic markers (CD13, CD14, CD15, and CD33), as well as for CD25, CD57, and granzyme B. Most of the patients were positive for HLA-DR and TdT, many for CD4 and CD7, and 4 of 27 patients for CD16, an NK cell marker. Conversely, most of the patients were negative for CD1, CD10, TIA-1, and CLA. EBV was not detected in any of the 35 patients examined. Although CD5 expression was statistically more frequent for lymphoma-type BNKL, the expression of other markers was comparable between the lymphoma and leukemic forms of BNKL.

Table 2. Phenotypic Characteristics of BNKL
CharacteristicsaBNKL (%)Leukemic BNKL (%)P value
  • BNKL: blastic natural killer cell lymphoma/leukemia; cyCD3: cytoplasmic CD3; HLA: histocompatibility antigen; TdT: terminal deoxynucleotidyl transferase; TIA: xxxx; CLA: cutaneous lymphoid antigen; NS: not significant.

  • a

    The following antigens were negative in all patients examined: surface CD3, CD8, CD13, CD14, CD15, CD20, CD25, CD33, CD57, surface immunoglobulins, T-cell receptors (αβγδ), and granzyme B.

CD10/131/14NS
CD27/18 (39)7/27 (26)NS
CyCD310/14 (71)10/13 (77)NS
CD410/19 (53)13/23 (57)NS
CD55/18 (28)0/220.01
CD710/13 (77)17/25 (68)NS
CD101/17 (6)3/22 (14)NS
CD11b0/31/9 (11)NS
CD161/13 (8)3/18 (14)NS
CD191/17 (6)0/26NS
CD342/13 (15)3/16 (16)NS
CD384/4 (100)7/8 (88)NS
CD5620/20 (100)27/27 (100) 
CD683/12 (25)3/12 (25)NS
HLA-DR16/18 (89)19/24 (79)NS
TdT11/14 (79)11/16 (69)NS
TIA-I0/122/12 (17)NS
CLA0/51/3 (33)NS

Karyotype

Cytogenetic characteristics are listed in Table 3. The karyotype of tumor cells could not be determined in 11 patients, and showed the normal diploid type in 20 patients. The 16 patients with an abnormal karyogram displayed no specific recurrent cytogenetic abnormalities. Chromosome 1 and 13 aberrations were the most frequent and were found in 9 patients, with abnormalities of 1p10-13 in 5 patients. Although no numeric abnormalities were found for chromosome 1, monosomy 13 was detected in 3, trisomy 13 in 1, and 13q abnormalities in 5 patients, whereas 8 patients showed a variety of chromosome 7 abnormalities. Abnormalities of chromosome 6, 11, or 17 were recognized in 6 patients.

Table 3. Cytogenetic Characteristics of BNKL with Abnormal Metaphases
Patient no.aSkin lesionCD4Karyotype
  • BNKL: blastic natural killer cell lymphoma/leukemia; ND: not determined; NM: no metaphase.

  • a

    B and L represent blastic BNKL and leukemic BNKL cases, respectively.

B-1++51,X,add(X)(q2?),add(1)(q21),add(2)(q2?),+6,−7,add(7)(p1?),−8,−9,−9,add(10)(q22),add(11)(p1?),+13,+16,−17,+20,−21,add(22)(q13),+6mar
B-2++45,XY,t(1;2)(p11;q11),t(6;17)(qter;q11),−13
B-3++45,XY,del(4)(q27),add(11)(q23.1),−12
B-4++42−45,der(X)t(X;1)(p22.3;p13),−Y,−1,del(2)(q22q33),der(3)add(3)(q11.1),add(3)(q11.1),del(8)(q22),−9,der(11)t(11?)(p11.1),der(15)t(1;?;15) (q23;?;q22),−20,+der(?)t(?;Y)(?;q12),+1−3mar
B-547,del(X)(q13),Y,+del(1)(p13),i(17)(q10),add(19)(p13)
B-649,XX,+21,+22,+mar
L-1++44,XX,der(1;16)(p10;q10),del(3)(q21),der(6)t(1;6)(q23;q23),i(7)(q10),der(11)t(11;13)(q11;q12),der(12;16)(q10;p10),−13,−16,der(19)r(19;?)(p13q13;?)
L-2++45,XY,add(7)(p15),−10,add(13)(q34)
L-3++45,XY,der(7)t(1;7)(q11;q22),der(12;15)(q10;p10)
L-4++47,XY,add(1)(p32),del(6),(q?),+7,add(8)(q24),add(9)(q34)
L-5+44,XY,add(5)(q13),−7,add(12)(p11),add(13)(q?),−14,add(19)(p13),−22,+mar
L-642,X,−Y,add(1)(p11),t(1;8)(p13;q24),del(5q?),add(11)(p11),−13,−14,add(17)(p11),add(17)(q25),der(20;21)(q10;q10)
L-7+42,XY,−6,6q−,−7,−10,−10,11q−,−13,−14,−17,−17,+4mar
L-845,XY,7q+,18q−
L-948,XX,2q+,13q−,14q+,+2mar
L-1048,XY,1p−,2p−,2q−,6q−,9p+,12p+,12q−,13q+,−17,18q+,+21,+2mar

Therapy and Clinical Course

Table 4 summarizes patients' treatments and clinical courses. Forty-three patients were treated with various chemotherapeutic regimens including anthracycline for lymphoid malignancies, 30 of who (70%) attained complete disease remission (CR). Twelve patients with lymphoma and seven patients with leukemia received CHOP or CHOP-like regimens, and 12 of them attained CR (63%). Patients with leukemia were treated more often with ALL-type or third-generation chemotherapy, and 13 of the total of 19 patients achieved CR (68%). There was no difference in the CR ratio between CHOP/CHOP-like regimens and ALL/third-generation chemotherapy. Three patients treated with AML-type chemotherapy and two with salvage chemotherapy including high-dose ara-C also achieved CR. Two of three elderly patients with localized disease who received palliative chemotherapy with or without radiotherapy attained CR. One pediatric patient could not be treated. In total, 34 patients achieved CR, but 22 of them suffered disease recurrence. During treatment, 11 patients received allogeneic or autologous hematopoietic stem cell transplantation (HSCT), 4 of them (2 allogeneic and 2 autologous) during the first CR, 1 (autologous) during the second CR, and 6 (3 allogeneic and 3 autologous) during the active disease stage. Three of those who underwent transplantation during the first CR and one during the second CR are now alive and well, but the remaining seven suffered disease recurrence and died. Patients who received transplantation showed better overall survival than those did not, but the difference was not statistically significant (4-year overall survival rate: 34% vs. 5.4%, P = 0.11).

Table 4. Therapy and Response
CharacteristicsBlastic BNKLLeukemic BNKLP value
  • BNKL: blastic natural killer cell leukemia/lymphoma; CR: complete disease remission; RT: radiotherapy; CHOP: cyclophosphamide, doxorubicin, vincristine, and prednisone; ALL: acute lymphoblastic leukemia; AML: acute myeloid leukemia.

  • a

    One infant could not be treated. Another patient attained CR by autograft.

CR rate (total)15/20 (75%)19/26 (73%)a0.88
Weak chemotherapy and Rt3/40/1 
CHOP/CHOP-like8/124/7 
ALL/3rd generation2/411/15 
AML-type03/3 
Salvage chemotherapy2/30/3 
Median survival (range)17.1 mos (2.5–33.0)20.4 mos (0.8–84.2+)0.33

Prognosis

The median survival of all patients was 19.8 months (range, 1–84 months) without statistically significant differences in overall survival between lymphoma and leukemia types (Fig. 1). Univariate analysis showed that age > 30 years (P = 0.01), leukocyte count at diagnosis (P = 0.02), and negative expression of TdT (P = 0.02) were significant prognostic factors, as was cutaneous involvement (P = 0.04) (Table 5). Mediastinal involvement and serum LDH level showed marginal significance. Multivariate analyses revealed that age, cutaneous involvement, and leukocyte count were significant and independent factors affecting overall survival (Table 5). Patient sex, PB blasts, platelets, clinical stage, BM involvement, lymph node involvement, performance status, presence of B symptoms, the international prognostic index for malignant lymphoma, cytogenetic abnormalities (normal vs. abnormal, chromosome 7 or 17 abnormalities), and CD4 expression did not have a significant prognostic impact on survival by Cox proportional hazard analyses.

Figure 1.

Prognosis of blastic natural killer (NK) cell neoplasm. Overall survival of patients with lymphoma and lymphoblastic leukemia showed no significant differences. Dotted line: leukemic blastic NK cell neoplasm; straight line: blastic NK cell lymphoma.

Table 5. Prognostic Factors Affecting Overall Survival
VariablesUnfavorable factorsUnivariateMultivariateb
Hazard ratio (CI)P valueHazard ratio (CI)P value
  • CI: confidence interval; yrs: years; TdT, terminal deoxynucleotidyl transferase; LDH: lactate dehydrogenase.

  • a

    Leukocyte and LDH were analyzed as continuous variables.

  • b

    Final model.

Age> 30 yrs4.1 (1.4–11.8)0.0113.2 (3.3–52.4)0.0002
Cutaneous involvementPresent2.1 (1.04–4.1)0.043.3 (1.5–7.1)0.002
Leukocyte counta1000/μL increment1.036 (1.006–1.067)0.021.089 (1.048–1.13)0.00002
TdTNegative2.9 (1.2–7.1)0.02 
Mediastinal involvementAbsent2.1 (0.98–4.6)0.06 
LDHa100 IU/L increment1.020 (0.997–1.043)0.09 

Categorization by Cutaneous Involvement

The results of a comparison of the clinical characteristics of the cutaneous (n = 28) and noncutaneous groups (n = 19) are summarized in Table 6. The cutaneous group showed a tendency of a later onset (median 56 vs. 46 years). Both groups showed male predominance, whereas the ratios of leukemia and lymphoma were similar. However, striking differences were found for the platelet count and mediastinal involvement, with the noncutaneous group showing significantly more severe thrombocytopenia (P = 0.03) and a marked preponderance of mediastinal disease (4% vs. 53%, P = 0.0002). Results of a comparison of phenotypic features by the presence of cutaneous involvement are summarized in Table 7. Expression of CD4 and HLA-DR occurred more frequently in the cutaneous group, whereas CD16 and CD34 expression was observed only in the noncutaneous group. Expression of CLA was positive for only one of the seven patients in the cutaneous group. The cutaneous type showed significantly better survival than the noncutaneous type (Fig. 2) (P = 0.02).

Table 6. Patient Characteristics: Cutaneous Versus Noncutaneous
CharacteristicsCutaneous BNKL (n = 28) (%)Noncutaneous BNKL (n = 19) (%)P value
  1. BNKL: blastic natural killer cell lymphoma; PB: peripheral blood; Hb: hemoglobin; PLT: platelet; CR: complete disease remission; mos: months.

Median age in yrs (range)56 (0–89)46 (0–78)0.11
Sex (male/female)19/914/50.75
Leukemia:lymphoma16:1211:80.96
Peripheral blood count   
 Median leukocyte count (μL) (range)5320 (1600–85,800)6000 (1100–28,200)0.36
 Median PB blast (%) (range)0 (0–71.0)0 (0–85.0)0.52
 Median Hb level (g/dL) (range)12.6 (11.5–16.9)11.5 (3.9–14.1)0.58
 Median PLT count (× 104/μL) (range)16.6 (3.6–25.4)5.4 (0.8–40.4)0.03
Sites of involvement   
 Bone marrow17 (61)13 (68)0.76
 Peripheral blood11 (39)9 (47)0.76
 Extramedullary28 (100)17 (89)0.16
  Lymph node18 (64)16 (84)0.19
  Mediastinum1 (4)10 (53)0.0002
  Liver3 (11)3 (16)0.67
  Spleen5 (18)3 (16)1.00
  Nasopharynx8 (30)3 (16)0.32
  Skin28 (100)0 (0)
Others4 (15)6 (32)0.28
CR rate20/28 (71%)13/19 (68%)0.83
Median survival (range)25.2 mos (0.9–84.2+)14.6 mos (0.8–41.4)0.02
Table 7. Phenotypic Features According to Cutaneous Involvement
CharacteristicsaCutaneous BNKL (%)Noncutaneous BNKL (%)P value
  • BNKL: blastic natural killer cell lymphoma/leukemia; cyCD3: cytoplasmic CD3; HLA: histocompatibility antigen; TdT: terminal deoxynucleotidyl transferase; TIA: xxxx; CLA: cutaneous lymphoid antigen.

  • a

    The following antigens were negative in all cases examined: CD3, CD8, CD13, CD14, CD15, CD20, CD25, CD33, CD57, surface immunoglobulins, T-cell receptors (αβ/γδ) and granzyme B.

CD28/26 (31)6/19 (32)0.95
cyCD313/18 (72)7/9 (78)1.00
CD422/27 (81)3/18 (17)< 0.0001
CD52/27 (7)3/16 (19)0.34
CD715/22 (68)12/16 (75)0.73
CD102/21 (10)2/17 (12)1.00
CD160/204/14 (29)0.02
CD340/185/14 (36)0.01
CD386/7 (86)5/5 (100)1.00
CD5628/28 (100)19/19 (100)
CD685/10 (50)1/5 (20)0.62
HLA-DR26/27 (96)9/15 (60)0.005
TdT15/20 (75)7/10 (70)1.00
TIA-12/16 (13)0/80.54
CLA1/7 (14)0/11.00
Figure 2.

Prognosis of cutaneous and noncutaneous blastic natural killer (NK) cell neoplasm. Patients with cutaneous blastic NK cell lymphoma/leukemia (BNKL) showed significantly better prognosis than did patients in the noncutaneous group (P = 0.02). Straight line: noncutaneous BNKL neoplasm; dotted line: cutaneous BNKL neoplasm.

Categorization by Other Characteristics

Although patients with positive TdT expression had significantly better prognosis than patients with negative expression of TdT (Table 5), no clinical characteristics were significantly different between these 2 groups (data not shown). Cytoplasmic CD3 was the only marker whose expression was different between the 2 groups (89% for the TdT-positive group and 33% for the TdT-negative group, P = 0.02).

Because the cutaneous involvement was closely correlated with CD4 expression and inversely with mediastinal involvement (Table 6), the expression of CD4 was lower for patients with mediastinal involvement (9%) than those without (71%). However, the difference in the CR rate (55% vs. 80%, P = 0.12) and prognosis (P = 0.06) did not reach statistical significance between the mediastinal and nonmediastinal groups. Prognosis also was not significantly different between CD4-positive and CD4-negative patients (Fig. 3).

Figure 3.

Prognosis of CD4-positive and CD4-negative blastic natural killer (NK) cell lymphoma/leukemia (BNKL). Overall survival of CD4-positive and CD4-negative BNKL patients showed no significant differences (P = 0.20). Straight line: CD4-positive BNKL; dotted line: CD4-negative BNKL.

DISCUSSION

We present 47 patients with BNKL. By using a cutoff value of 30% for BM/PB involvement, 20 patients were diagnosed with lymphoma and 27 were diagnosed with leukemia, but there was no clear difference between the 2 groups. As in the case of precursor-B and precursor-T ALL/LBL, the presentation of leukemia or lymphoma simply represents tumor localization and does not seem to indicate the existence of disease subtypes of BNKL.

We then focused on the cutaneous involvement, because the anatomic site of onset is closely related to the disease categorization of several types of lymphoid malignancies.46 Comparison of clinicopathologic features between the cutaneous and noncutaneous groups resulted in many noteworthy differences, including age of onset, disease localization, immunophenotype, and prognosis. Cutaneous BNKL rarely showed mediastinal involvement, but more than one-half of the patients with noncutaneous BNKL did. A high incidence of CD4 expression and an absence of CD34 expression for the cutaneous group were consistent with recent reports of cutaneous CD56-positive non-T-cell lymphoma with blastic morphology by other groups.9, 36, 37 The phenotypic differences between the cutaneous and noncutaneous BNKL groups suggest that these two subtypes have different cell origins with different maturation status because CD34 is one of the stem cell antigens. Although patients with both types responded well to chemotherapy for lymphoid malignancies, their different prognoses further justify different clinical management of these two diseases. Because cutaneous involvement and CD4 expression shaped such close correlation, categorization of the disease in terms of CD4 expression is also possible,38 but the prognosis of patients with CD4-positive and CD4-negative BNKL was not significantly different. These findings strongly suggest that the entity known as BNKL consists of two distinct diseases.

Because of the lack of lineage-specific antigens other than CD56, the true origin of BNKL/CD56-positive BT has not yet been determined. Two French groups recently claimed the origin of CD4-positive CD56-positive tumors to be the precursors of pDC27, 28 or plasmacytoid monocytes.29 They showed that the tumor cells express interleukin-3 (IL-3) receptor α (CD123) in all patients, acquire dendrite-like morphology with IL-3 plus CD40L stimulation, and produce interferon-α by influenza virus stimulation in some of the patients. These features are shared with those of pDCs, but there still remain several problems before the question of the origin of CD4-positive CD56-positive BTs can be settled. First, CD56, a key molecule for this malignancy, is not expressed in most normal pDCs.47–50 Previously reported CD4-positive/CD4-negative CD56-negative CD123-positive tumors with pDC features might represent true pDC malignancy.51 Second, CD123 is not an exclusive marker for pDCs. Because IL-3 is a cytokine that is capable of stimulating a variety of hematopoietic cells or colony-forming cells, its receptor CD123 is also expressed in a variety of hematopoietic malignancies.29, 53, 54 Third, normal pDCs usually reside in lymph nodes, PB, and BM, but not in cutaneous/subcutaneous regions. The reason for the preponderance of skin involvement in CD4-positive CD56-positive tumors cannot be accounted for if they are derived from pDCs. The occasional supervenience of myelomonocytic leukemias in a subset of CD56-positive BT suggests that this tumor might also have monocytic or multilineage potential,39 but the leukemic characteristics of BNKL are far different from those of acute myelomonocytic/monocytic leukemia.

CLA is a homing receptor of T cells that regulates lymphocyte trafficking to the skin,33 and has been recognized in mature NK cell lymphomas.34 However, its expression was only recognized in a few of our patients, including patients with cutaneous BNKL, which is consistent with the results for CD4-positive CD56-positive malignancy.27 Petrella et al.55 recently identified CLA expression in 27 of 29 patients with CD4-positive CD56-positive cutaneous malignancy.55 Although the CLA expression in cutaneous and noncutaneous CD56-positive BT needs further investigation, CLA seems to be unrelated to the cutaneous presentation of tumor cells.

The clinical characteristics of noncutaneous BNKL, such as younger age onset, frequent mediastinal involvement, and occasional expression of CD34, indicate the need for differential diagnosis from precursor T-cell LBL. The similarity of these two tumors is to be expected because T cells and NK cells share the same progenitor in ontogeny. However, noncutaneous BNKL differs from precursor T-cell LBL in several respects. The former shows germline configuration of TCR genes, occurs at a patient median age of 46 years, and mostly shows a poor prognosis. In contrast, patients with T-cell LBL/ALL usually have rearranged TCR genes, a much younger age at onset, and have a fair prognosis.19 A relatively high incidence of CD16 expression also supports the distinction of noncutaneous BNKL from precursor T-cell LBL.

The expression of TdT is one of the intriguing features of BNKL. TdT is associated with TCR and immunoglobulin gene rearrangement through its enzymatic function to add nucleotides at the junction of gene segments in normal T or B cells. Although TdT is usually expressed in TCR-rearranged LBL, several patients show no or very weak expression of TdT,56 suggesting that TdT is down-regulated after the completion of TCR rearrangement in these tumors. In the majority of patients with BNKL, TdT is expressed, but TCR and immunoglobulin genes are usually not rearranged except in rare cases. This discrepancy has also been reported for AML.57 Drexler et al.57 further showed an association between TdT and CD34 expression, and have concluded that TdT could be viewed as a marker of hematopoietic immaturity, and a part as a prognostic factor for AML. The present data for BNKL, however, could not identify the correlation between TdT and CD34 expression, and showed the inverse prognostic impact of TdT expression.

For the treatment of BNKL, Reimer et al.8 reviewed 97 CD4-positive CD56-positive cases in the literature, and identified the advantage of chemotherapy for acute leukemia compared with CHOP or CHOP-like regimens (CR rate: 94% vs. 55%).However, their results might be biased by age distribution, because the median age of the patients treated with chemotherapy for acute leukemia was > 10 years younger than that for patients who received CHOP or its equivalent (56 vs. 67.5 years), as well as the assignment of patients who underwent HSCT to another category. Our series also showed the advantage of intensive ALL-type chemotherapy, but the difference was not significant (63% vs. 68%). The poor prognosis of patients with this disease7–9, 36, 37 justifies the use of HSCT for patients with BNKL,2, 8, 22, 58–61 but its efficacy needs to be investigated in clinical trials.

In conclusion, clinical and prognostic analyses reveal that BNKL is a heterogeneous disease. Two subtypes, i.e., cutaneous type (with better prognosis) and noncutaneous type (often accompanied by mediastinal involvement and a poorer prognosis), comprise BNKL. The exact nature of these two entities and their accurate nomenclature need further clarification.

Acknowledgements

The authors thank the following collaborating institutions and their staffs for providing the patient data and specimens: Akita University School of Medicine; Tohoku University School of Medicine; Nippon Telephone and Telegram Kanto Hospital; Institute of Medical Science, Tokyo University School of Medicine; Tokyo Medical School; National Tokyo Medical Center; Yokohama City University School of Medicine; Niigata University School of Medicine; Shinshu University School of Medicine; Fukui University School of Medicine; Hamamatsu Medical Center; Gifu Municipal Hospital; Mie University School of Medicine; Osaka University School of Medicine; Japanese Red Cross Osaka Hospital; Osaka Municipal Medical Center; National Osaka Hospital; Kure Kyosai Hospital; Kokura Memorial Hospital; and Miyazaki Medical College.

Ancillary