Clinical features of CD56+ lymphoma. A, Solitary tumor of blastic lymphoma on the anterior thigh in patient 2. B, Nasal-type natural killer–cell/T-cell lymphoma manifested as diffuse plaques and nodules in patient 1. C, Subcutaneous movable tumors of subcutaneous panniculitislike lymphoma on the arm of patient 3.
Histologic features of CD56 lymphoma in the skin. A, Homogeneous infiltrate of blastic cells in patient 4 with blastic CD56+ lymphoma. B and C, Nasal-type natural killer–cell/T-cell lymphoma in patient 1 with an infiltrate composed of pleomorphic small, medium, and large cells. D-F, Subcutaneous panniculitislike CD56+ lymphoma in patient 3. Hematoxylin-eosin used for panels A-D, CD3 stain for panel E, and CD56 for panel F; original magnification ×100 (panel B) and ×400 (panels A, C-F).
Kaplan-Meier survival plots for CD56+ lymphoma. The effect of the lymphoma type (A), age in years (B), and degree of involvement (C). Censored events are marked with tick marks on the step curves. SPL indicates subcutaneous panniculitislike lymphoma.
Estimated survival curves for patients with CD56+ lymphoma. A, Survival probability for patients with CD30+ and CD30− lymphomas. The variables included in the model are age, type of lymphoma, presence of extranodal dissemination, presence of Epstein-Barr virus, and CD4 status. B, Survival curves for CD4+ and CD4− lymphomas. CD30 status was removed from this regression model. Censored events are marked with tick marks on the step curves.
Gniadecki R, Rossen K, Ralfkier E, Thomsen K, Skovgaard GL, Jønsson V. CD56+ Lymphoma With Skin InvolvementClinicopathologic Features and Classification. Arch Dermatol. 2004;140(4):427-436. doi:10.1001/archderm.140.4.427
Extranodal lymphomas expressing CD56 (neuronal cell adhesion molecule) are characterized by a high incidence of cutaneous involvement and a very aggressive clinical course. Knowledge about the prognosis and clinicopathologic features of CD56+ lymphomas with skin involvement is very limited.
To determine survival and prognostic factors for extranodal CD56+ lymphomas with skin involvement and to describe their clinicopathologic features.
Retrospective literature survey and case studies.
A total of 181 patients with CD56+ lymphoma involving the skin: 177 cases from the literature and 4 new cases.
Main Outcome Measure
Survival and its dependence on the following putative prognostic factors: staging, histopathologic findings, lymphocyte markers, T-cell receptor gene rearrangement, Epstein-Barr virus infection, treatment modality.
Three major subtypes of CD56+ lymphoma in the skin were distinguished: blastic lymphoma, nasal-type natural killer–cell/T-cell lymphoma, and subcutaneous panniculitislike lymphoma. The disease disseminated readily, mainly to lymph nodes, bone marrow, the central nervous system, and the liver, but 45% of patients had a purely cutaneous disease at presentation. All subtypes had a very aggressive course with a median survival of 14 months. The main risk factors were age older than 55 years (hazard ratio [HR], 2.5; 95% confidence interval [CI], 1.8-3.2), systemic dissemination at presentation (HR, 2.0; 95% CI, 1.5-3.3), and lack of CD30 (HR, 3.8; 95% CI, 1.4-4.9) or CD4 expression (HR, 1.56; 95% CI, 1.06-2.57). The different treatment modalities did not improve survival.
CD56+ lymphomas involving the skin are rare and extremely aggressive regardless of their histologic presentation and the extent of skin involvement. No effective treatment is available. The risk of death is particularly increased in older patients with CD30−CD4− lymphomas.
The CD56 antibody recognizes the neuronal cell adhesion molecule, which in physiologic conditions is expressed on virtually all peripheral natural killer (NK) cells, a subset of monocytes and peripheral CD8+ T cells. Lymphocyte activation in the presence of several cytokines (interleukin 2, interferon [IFN] γ, and tumor necrosis factor α) may result in an increased expression of CD56 on both CD8+ and CD4+ cells (the so-called cytokine-induced killer cells), and this process may take place during the antitumor response.1,2
It has long been recognized that CD56 is expressed in a well-defined type of neoplasm, the NK/T-cell nasal lymphoma.3,4 This lymphoma is most common in populations from Asia (Korea, Hong Kong, and Japan) and South America and is consistently associated with the Epstein-Barr virus (EBV). It presents clinically with the destructive nasal or midline facial tumor. Recent molecular studies on the clonality of killer cell immunoglobulinlike receptors confirm the origin of this neoplasm from NK cells.5
In 1992, Kern et al6 and Wong et al7 reported expression of CD56 in extranodal lymphomas that are histologically similar to nasal lymphomas but do not occupy the nasopharyngeal region (nonnasal CD56+ lymphomas). These lymphomas run a very aggressive clinical course and have a weaker association with EBV than the true nasal lymphomas. Skin is often the site of the initial ivolvement. The cellular origin of this lymphoma type is unknown, but NK cells, T cells, or plasmacytoid monocytes have been proposed.4,8- 10
Owing to the extreme rarity of the disease, information about nonnasal CD56+ lymphoma is limited. Histopathologically, nonnasal CD56+ lymphoma has been divided into 2 types: (1) nasal-type NK/T-cell lymphoma (named by some authors as the pleomorphic, small, medium, and large cell type) and (2) blastic NK/T-cell lymphoma.3,4,11 Moreover, an additional entity of the aggressive NK-cell lymphoma/leukemia of an extremely malignant course has been proposed.11 Further complexity is added to the classification by the fact that CD56 may be expressed in other types of lymphomas and leukemias such as anaplastic large-cell lymphomas, lymphomatoid papulosis, hepatosplenic γδ T-cell lymphomas, subcutaneous panniculitislike (SPL) γδ T-cell lymphomas, B-cell and T-cell lymphoblastic lymphomas, or myeloid leukemias with or without skin involvement.2,12- 17
In the present article, we report 4 new cases of CD56+ lymphoma where the skin was the primary site of presentation. Moreover, we review the available literature on CD56+ lymphomas involving the skin and provide data on prognosis and clinical and pathologic variants. This article includes nonnasal lymphomas only. Other well-defined types of lymphomas expressing CD56 have also been excluded.
Since the inclusion of CD56 staining in the routine phenotyping of cutaneous lymphomas in 1997, we identified 4 cases of CD56+ lymphoma. In all patients, skin symptoms were the primary reason for referral. The cases were classified as CD56+ lymphomas on the basis of positive immunoperoxidase staining with the CD56 antibody and the inability to classify the tumor into other established T- or B-cell lymphoma categories.
Routine histopathologic studies were done on 4-µm-thick sections stained with hematoxylin-eosin and reviewed by 2 of us with expertise in hematopathology and cutaneous pathology (K.R. and E.R.). The immunophenotyping was performed on paraffin and cryostat sections using antibodies against CD34, terminal deoxynucleotidyl transferase, CD56, CD3, CD4, CD8, T-cell intracellular antigen-1, granzyme B, CD79a, and CD30 (Ber-H2).9,18
Interferon γ concentration was determined in a serum sample from patient 3 using the Quantikine human IFN-γ immunoassay (R&D Systems, Abingdon, England), according to the manufacturer's protocol. For the assessment of the T-cell receptor (TCR) rearrangement status, genomic DNA was amplified using consensus primers for TCR-γV (5′-CTTCCAACTTGGAAGGGAGAA-3′ covering TCR-γV1-5, 7, and 9) and TCR-γJ (5′-CCAGGCGAAGTTACTATGAGC-3′ covering TCR-γJ1 and 2) yielding an amplification product of approximately 350 base pairs that was subsequently analyzed by single-strand conformational polymorphism.19 For detection of EBV-encoded small nuclear ribonucleic acid (EBER), a mixture of fluorescein-conjugated oligodeoxyribonucleotides complementary to the 2 EBERs was used (EBER; DAKO, Glostrup, Denmark).
Sections were dewaxed and rehydrated, digested with proteinase K, and postfixed for 20 minutes in 4% paraformaldehyde. Subsequently, the sections were allowed to prehybridize for 60 minutes at 45°C in a solution consisting of 50% formamide, 20% dextran sulfate, 0.2% polyvinyl pyrrolidone, 0.2% Ficoll, 0.1% sodium pyrophosphate, 5mM sodium edetic acid, and 50mM Tris hydrochloride at pH 7.6. After overnight hybridization at 37°C, the sections were incubated with alkaline phosphatase-conjugated rabbit anti-FITC (fluorescein isothiocyanate–conjugated) antibody (Roche Diagnostics, Basel, Switzerland) and developed with Fast Red TR (Sigma, St Louis, Mo). As positive controls, we used sections from lymph nodes with known EBV-associated Hodgkin disease.
The aim was to find all published cases of CD56+ lymphomas presenting in the skin. The PubMed database was searched up to September 2002 using MeSH terms "Antigens," "CD56 AND Lymphoma," and "Non-Hodgkin AND (skin OR cutaneous OR subcutaneous)" (38 references) and the phrase "CD56 AND Lymphoma AND (skin OR subcutis OR cutaneous OR subcutaneous OR panniculitis)" (127 references). Obviously irrelevant articles, such as those presenting no clinical data or reporting on lymphomas without skin involvement, were rejected. The selected 73 articles were read by 1 of us (R.G.) who further selected the articles reporting as a minimum the main clinical features, survival and the cause of death of individual patients, CD56 positivity of the lymphoma, and histopathologic description. At this stage, 50 articles were found. An additional reference (congress abstract) was identified from the literature lists of these articles, but this communication contained data already included in a separate publication.
Cases of CD56+ lymphoma with skin involvement were classified as previously described.4 Blastic lymphomas were defined as those showing a dense nonepidermotropic dermal infiltrate composed of cells resembling lymphoblasts with finely granular chromatin, inconspicuous nucleoli, and a sparse to moderate amount of cytoplasm. The lymphomas showing dermal and sometimes subcutaneous infiltrates of atypical pleomorphic cells (small, medium, or large) containing variable amounts of cytoplasm were considered to represent nasal-type NK/T-cell lymphomas. The SPL lymphomas showed similar histologic features to the nasal-type NK/T-cell lymphomas and were defined by the presence of clinical features of panniculitis and subcutaneous tumors. In all cases, the positivity of tumor cells for CD56 was required for diagnosis.
The stardard error and confidence intervals (CIs) for proportions were computed as described by Altman.20 The analysis of 2 × 2 frequency tables was done by the χ2 test with the Yates correction. Analysis of survival times was done by life table analysis and the log-rank test and by the Cox backward stepwise regression model. Disease-related survival was defined as the number of months from the establishment of the diagnosis to the death event that was related to the lymphoma. Patients who died due to unrelated diseases or were lost to follow-up were considered to be censored at the time of the last contact. Median survival time, defined as the time when the cumulative probability of survival equals 0.5, was established from the Kaplan-Meier step-function curves. Calculations were performed with the Unistat statistical package (Unistat Corp, London, England), GraphPad Prism (GraphPad Software, San Diego, Calif), and SPSS statistical software (SPSS Science, Chicago, Ill).
The crucial characteristics of our 4 cases of CD56+ lymphoma are summarized in Table 1 and illustrated in Figure 1 and Figure 2. The data gathered for further analysis came from 177 additional cases retrieved from published reports (Table 2).8,9,11,14,21- 66
During the preliminary analysis of the data, we confirmed the practical feasibility of the previously suggested classifications of blastic lymphomas and nasal type NK/T-cell lymphomas.4 Moreover, a group of patients presented with CD56+ lymphoma that manifested as panniculitis. Cytologically, the malignant cells of these tumors resembled those of pleomorphic lymphoma, but since this type had specific clinical features, we decided to analyze these cases as a separate subtype, CD56+ SPL lymphoma. The major characteristics of these 3 categories are summarized in Table 3. There was a marked male predominance in the cases of blastic and pleomorphic lymphoma, whereas the SPL type seemed to affect mostly women. The median age at diagnosis was 55 years and did not differ significantly between lymphoma subtypes.
The skin lesions included plaques and tumors, sometimes with hemorrhage and necrosis. Patients with SPL lymphoma also showed features of panniculitis. When all individuals with CD56+ lymphoma were analyzed as a single group, most patients (53%; 95% CI, 45%-61%) presented with widespread skin lesions. Only 30% (95% CI, 22%-38%) had a single lesion, and 17% (95% CI, 11%-24%) had several lesions confined to a single anatomic region. Virtually any region of the skin could be affected, but the CD56+ SPL lymphoma tended to be localized on the extremities and the buttocks. A very rare but clinically characteristic presentation of CD56+ lymphoma was bruiselike lesions that tended to occupy the face and in 1 case were associated with paraneoplastic hemorrhagic diathesis.21- 23 Bruiselike lesions seem to be characteristic of blastic lymphoma.
Virtually all patients with SPL lymphoma had systemic symptoms: weight loss, fever, general weakness, and arthralgia, which have been attributed to the increased synthesis of proinflammatory cytokines such as IFN-γ.24 Also, our patient with SPL lymphoma (case 3) had an increased concentration of IFN-γ in the blood (2859 pg/mL; normal concentration in human plasma is <15 pg/mL). Systemic symptoms were less frequent (17% of cases) in the 2 other types of CD56+ lymphoma. Extracutaneous involvement was often encountered (52% of patients). The most frequent sites of dissemination were bone marrow (26%) followed by lymph nodes and/or spleen (20%), visceral organs (mostly liver) (11%), and the central nervous system (7%). Visceral dissemination was slightly less frequent in cases of SPL lymphoma than in the 2 remaining subtypes (Table 3).
In the nasal-type NK/T-cell lymphomas and blastic lymphomas, the infiltrate of the malignant cells was localized in the dermis and often in the subcutis. Epidermotropism was noted in only few cases.25 Angiocentricity and blood vessel invasion was present in approximately 60% of nasal-type NK/T-cell lymphomas and SPL lymphomas, but was absent in blastic lymphomas. In general, the histopathologic features were in accordance with previous descriptions.3,4
Immunohistochemical and molecular features are listed in Table 4. No specific single set of markers could be associated with the lymphoma subtypes; nonetheless, some regularity was observed. Malignant cells did not express B-cell markers (CD19, CD20, or immunoglobulin gene rearrangement) or NK-cell marker CD57 and only rarely expressed the stem-cell marker CD34. Likewise, expression of CD8 was noted in only 4 patients,26- 28(patient 5),29(patient 2) and expression of CD2, CD5, and CD7 was highly variable. Only in 3 cases did the malignant cells demonstrate a mature T-cell phenotype, CD3+CD4+CD8−, and have rearranged TCR genes.30- 32(patient 31)
The most common phenotype of the blastic CD56+ lymphoma was CD3−CD4+ present in 78% of the cases (P = .001, χ2 test). Terminal deoxynucleotidyl transferase was sometimes detected in this subtype, but the expression was often equivocal, and the number of patients investigated for this marker was too small to draw meaningful conclusions. Epstein-Barr virus was present only in few cases (Table 3), and in 2 cases the TCR rearrangement was detected.
Nasal-type NK/T-cell lymphoma was often associated with EBV infection. The most common phenotype was CD3+CD4− (52.2%), which is significantly more than expected by chance association (P = .02, χ2 test).
The characteristics of SPL lymphoma were similar to those of the nasal-type NK/T-cell lymphoma: 84% of patients had the phenotype CD3+CD4− (P = .001, χ2 test), and a substantial proportion of patients had detectable EBV and rearranged TCR genes (Table 3 and Table 4). In all but 1 case, the TCR-γδ but not TCR-αβ genes were rearranged. In all cases of SPL lymphoma that were examined for expression of cytotoxic markers, the presence of at least 1 marker (T-cell intracellular antigen-1, granzyme B, or perforin) was noted.
Kaplan-Meier survival curves revealed that the median survival for all patients with CD56+ lymphoma was 14 months and did not differ significantly between the different subtypes (Figure 3). To determine whether any clinical characteristic predicts survival, we chose the following variables: age, sex, race, presence of extracutaneous dissemination at presentation, and extent of skin involvement (single lesions, local involvement, widespread lesions). These variables have been shown to be important in various types of non-Hodgkin lymphomas, including cutaneous T-cell lymphomas. Only age and the presence of extracutaneous dissemination were strong negative predictors of survival (Figure 3). Patients younger than 55 years (the median age of onset of CD56+ lymphoma) had twice as long survival as those older than 55 years (25 months vs 10 months; P = .003; hazard ratio, 0.53; 95% CI, 0.32-0.71). Lymph node involvement did not have a measurable negative effect on survival, while dissemination into bone marrow and/or visceral organs had a strong negative effect, decreasing survival from a median of 25 months to 8 months (P<.001; hazard ratio, 2.0; 95% CI, 1.48-3.31).
There was no survival difference between patients with CD3+ tumor cells and those with CD3− lymphomas. There were too few patients with CD8+ lymphomas to perform a reliable analysis. T-cell receptor TCR rearrangement did not have any prognostic importance either. However, EBV status, CD4 expression, and CD30 expression were related to survival in the single-variable model. To elucidate the importance of these phenotypic markers for survival, we performed the backward stepwise Cox regression analysis to account for possible associations between the variables. The following variables were selected: age at diagnosis, type of lymphoma, extranodal dissemination, and presence of EBV, CD4, and CD30. Of these, only age, CD4, and CD30 were identified as significant variables (P = .01, P = .03, and P = .004, respectively). Presence of CD30 was the strongest predictor of survival (Figure 4A). The importance of CD4 was also investigated (in a model from which the CD30 status was removed to increase the number of patients in the model). Interestingly, in this regression model, the presence of visceral dissemination was also a significant predictor (P<.001; Figure 4B).
Therapy did not seem to affect survival. The patients who received treatment were classified into 4 categories: (1) those who received CHOP (cyclophosphamide-doxorubicin-vincristine-prednisone combination therapy); (2) those who received another doxorubicin-based therapy; (3) those who received another combination chemotherapy; and (4) those who received local radiation, systemic steroids, or no therapy. There were no differences in survival among these 4 groups of patients.
The CD56+ lymphoma is known to be one of the most aggressive neoplasms affecting the skin. However, because of its rarity, little is known about its clinicopathologic features and prognostic factors. CD56+ lymphomas frequently affect the skin, and the dermatologist is often the first physician to see the patient. Unfortunately, knowledge about this lymphoma is not widespread, which is reflected in the fact that most published case reports originate from few academic centers with long-term interest in cutaneous oncology and malignant hematology. The true prevalence of CD56+ lymphoma is unknown because the CD56 antibody is not routinely used for lymphoma phenotyping by all centers. The fact that 4 cases were encountered over 3 years in a relatively small center like ours (population base, approximately 1.5 million) may suggest that this lymphoma is more prevalent than expected.
For the present article, we critically assessed 177 published cases of CD56+ lymphoma together with our own 4 cases. One must be aware of possible limitations of this design. There is a theoretical possibility that the same group will describe the same patient multiple times without making reference to other publications. In several cases we identified clinical data of the same patient in several communications, mostly posters or congress abstracts followed by full articles. We were aware of this possibility and worked to exclude patients who may have been described multiple times by checking for identical age, sex, survival, and clinical characteristics in patients reported in multiple publications by the same researchers.
Another limitation is that the characteristics of the described patients are not representative of all patients with CD56+ lymphoma. The CD56 marker is not routinely used by all dermatopathologists, and there might be a tendency to perform this staining for more aggressive or histopathologically unclear lymphomas. However, owing to the rarity of CD56+ lymphoma, prospective studies have not been performed. To date, the present article is the largest survey on this type of lymphoma.
The data confirm an extreme aggressiveness of this tumor and a median survival of approximately 1 year after diagnosis despite chemotherapy and radiation therapy. The data confirm the usefulness of the accepted classification of these lymphomas: blastic and nasal-type NK/T-cell lymphomas. In addition, CD56 is also expressed in a subset of SPL lymphomas (Table 5).
The male preponderance of CD56+ lymphoma has been noted previously.11,25,28,33,34 This predominance holds true for blastic and nasal-type NK/T-cell lymphomas, while CD56+ SPL lymphoma seems mainly to affect women. However, sex is not a factor that affects survival.
The 3 most important independent predictors of survival are age at diagnosis and the expression of CD30 and CD4. Of these, CD30 expression has the strongest positive prognostic importance, independent of other possible confounders such as age, clinical stage, lymphoma subtype, and the presence of EBV in the lesions. This surface molecule belongs to the tumor necrosis factor receptor family and is also expressed in other cutaneous lymphomas such as lymphomatoid papulosis and the anaplastic large cell lymphomas that run a favorable clinical course.67 The role of CD30 in cutaneous lymphomas is puzzling because nodal CD30+ anaplastic lymphomas are very aggressive—especially those not carrying the t(2;5) chromosomal aberration—in contrast to their cutaneous counterparts. CD30 molecules exert pleiotropic biological effects on lymphoma cells, which depend largely on the cellular context.67 In anaplastic lymphoma, stimulation of CD30 leads to cell growth arrest and apoptosis via signaling mechanisms involving p21CIP1/WAF1 and antagonizing the activity of nuclear factor κB.68,69 In contrast, activation of CD30 has a mitogenic effect in Hodgkin lymphoma that constitutively expresses nuclear factor κB.67,68 The favorable prognostic importance of CD30 in CD56+ lymphomas has previously been suggested by Mraz-Gernhard et al25 and was confirmed in the present study. The reason for the predictive value of CD4 is unclear. It is possible that the absence of CD4 heralds a less mature and hence more aggressive phenotype.
In the Cox survival model based on 33 patients with known variables (age, lymphoma subtype, presence or absence of extranodal dissemination, EBV status, and expression of CD4 and CD30), the presence of extranodal dissemination was not a significant predictor. We were puzzled by this finding and considered the possibility that the number of patients was too low to detect this effect. Therefore, we constructed another regression survival model that did not include the CD30 status. This analysis was made on 129 patients and showed statistical significance for extranodal dissemination. Although we consider it likely that the clinical stage is predictive for survival, we cannot exclude the possibility that this variable is correlated with (and dependent on) the expression of CD30.
The cellular origin of CD56+ lymphomas is still a matter of debate; NK cells, T cells, and monocytes have been suggested as relevant cell types. Recent research has provided evidence that blastic CD56+ lymphomas originate from plasmacytoid monocytes that under normal conditions may act as precursors of antigen-presenting cells.10,34 However, in a recent case of CD56+ lymphoma,9 the expression of the NK-receptor complex CD94/NKG2 was found, which suggests an NK-cell origin of this neoplasm. The malignant cells in SPL lymphomas seem to be related to γδ T cells. In fact, the clinical and histologic features of CD56+ and CD56−γδ T-cell lymphomas are indistinguishable, and a significant proportion of SPL CD56+ lymphomas express TCR-γδ. It seems that TCR-γδ+ and TCR-γδ− lymphomas originate from a common T-cell precursor, but developmental arrest occurs earlier the TCR-γδ− type. Another observation supporting this concept is the coexistence of TCR-γδ+ and TCR-γδ− populations in patients with CD56+ leukemia/lymphoma syndrome.70 An interesting feature of SPL lymphomas, demonstrated for the first time by Burg et al24 and also shown in our case 3, is the presence of systemic symptoms (fever, weight loss, general weakness, and/or arthralgia) that may be related to the production of IFN-γ by tumor cells. Like αβ T cells, γδ cells may differentiate into IFN-γ (T helper 1–like) and interleukin 4 (T helper 2–like) producing cells. However, γδ T cells differentiate by default toward the type 1 cytokine–producing cells,71 and this maturational pathway may be caricatured in SPL lymphoma.
Corresponding author and reprints: Robert Gniadecki, MD, DMS, Department of Dermatology D, Bispebjerg Hospital, Bispebjerg bakke 23, DK-2400 Copenhagen NV, Denmark (e-mail: firstname.lastname@example.org).
Accepted for publication July 2, 2003.