Novel markers of normal and neoplastic human plasmacytoid dendritic cells

Hematologists are familiar with the major cell lineages generated in the bone marrow and with the neoplastic disorders to which they give rise. However, less attention has been focused on minor hematopoietic subpopulations that, despite their low numbers, have important functional roles and are of clinical relevance because of their capacity to undergo neoplastic transformation.

One such lineage comprises the cell type known as plasmacytoid dendritic cells (pDCs), a population that is typically found in cell clusters (or as isolated cells) in T cell–rich interfollicular areas in peripheral lymphoid tissue.^1,2  Their plasmacytoid morphology reflects a rich content of rough endoplasmic reticulum whose major product comprises type I interferons (mainly interferon-α). They respond to a variety of stimuli (including viruses, and hypo- and nonmethylated bacterial DNA sequences^3-6 ) and carry receptors, including a number of Toll-like receptors, capable of binding a spectrum of pathogen-associated molecules.⁷  They therefore represent a strategically positioned first-line defense against viral and other pathogens entering lymphoid tissue from the circulation.

A number of studies have documented the molecules expressed at the RNA and/or protein level by pDCs, although more data are available for the mouse than for man.⁸  Knowledge of the phenotype of pDCs is not only of relevance to an understanding of their origin and function but is also of potential clinical importance in the field of hemato-oncology, since it has been recognized for a number of years that pDCs can undergo neoplastic transformation. In early reports it was noted that such neoplasms were associated with a myeloproliferative disorder (principally acute or chronic myelomonocytic or monocytic leukemia),⁹  but more recently a distinctive tumor type, CD4⁺C56⁺ hematodermic neoplasm, involving both the skin and peripheral blood/bone marrow, has been documented.¹⁰ 

The diagnosis of these tumors in biopsy samples is based principally on markers such as CD4, CD56, CD123, and TCL1,^10-12  but most of these are present on other cell types. Furthermore, difficulties in diagnosis can also arise when a molecule is absent or is aberrantly expressed. There is therefore a need for robust additional markers of pDCs detectable in routine biopsies that are expressed on their neoplastic counterparts but are not found on tumors with which they may be confused.

In this paper we report an extensive immunophenotypic characterization of human pDCs and document a range of molecules (eg, involved in cell signaling and gene transcription) that have not previously been demonstrated in pDCs in routine biopsy material. Many of these are also expressed on neoplastic pDCs, and one molecule, the adaptor protein CD2AP (CD2-associated protein), is not expressed to a comparable degree by other peripheral white cells, making it a valuable new marker for detecting normal and neoplastic pDCs in tissue biopsies and peripheral blood.

Paraffin-embedded tissue sections of reactive human tonsils, lymph nodes with histologic features of Castleman's disease (no. 4), of Kikuchi's disease (no. 4), skin biopsies of cutaneous lupus erythematosus (no. 6), lichen planus (no. 5) and normal bone marrow trephines (no. 2) were obtained from the authors' institutions. Cryostat sections of human tonsils from the same source were also used for this study.

Tissue sections from 47 pDC-derived neoplasms were obtained from the author's institutions (M.D., F.F., P.G.I., P.T., T.M., K.R., and H.S.) and comprised (1) cutaneous, bone marrow, or splenic tumors, which had been diagnosed as blastic natural killer (NK)–cell lymphoma,¹³  CD4⁺CD56⁺ hematodermic neoplasm¹⁰  or simply (in the case of 2 cutaneous neoplasms) as pDC tumors (41 cases); and (2) neoplasms diagnosed as pDC proliferations associated with myeloproliferative disorders⁹  (6 cases).

Table 1 summarizes the available phenotype and clinical data for each case of pDC neoplasia. The study also included tissue sections of (1) 24 acute myeloid leukemia in the skin (leukemia cutis), with and without CD56 expression; (2) bone marrow trephines from 7 chronic myeloid and 5 chronic myelomonocytic leukemias; and (3) tissue sections from B- and T-lymphoblastic leukemia/lymphomas (7 and 6 cases, respectively). This material was retrieved from the files of 6 authors (F.F., K.H., M.-L.H., T.M., D.Y.M., and S.A.P.). The diagnosis of all lymphoid and myeloid neoplasms was based on the criteria of the World Health Organization (WHO) classification.¹³  Approval from the Oxford Research Ethics Committee B was obtained for this study (Research Ethics Committee Reference number: C02.162).

Peripheral blood mononuclear cells (PBMCs) were isolated from human ethylenediaminetetraacetic acid (EDTA)–anticoagulated peripheral blood from healthy donors after informed consent by a conventional gradient centrifugation technique using Histopaque (Sigma-Aldrich, Gillingham, United Kingdom). The isolated PBMCs at 1.25 × 10⁶/mL were used to prepare cytospins according to a protocol described elsewhere.¹⁴ 

Peripheral blood mononuclear cells were isolated from buffy coats by Ficoll gradient (GE Healthcare, Little Chalfont, United Kingdom), and pDCs were magnetically sorted with blood dendritic cell (DC) Ag BDCA-4 cell isolation kits (Miltenyi Biotec, Bergisch Gladbach, Germany), as described previously.¹⁵  More than 90% of the purified cells expressed CD123 (assessed by flow cytometry, data not shown), confirming the plasmacytoid DC nature of the great majority of the isolated cells. Blood pDCs isolated in this way (10⁶ cells/mL) were cultured in medium containing 1000 U/mL recombinant human granulocyte-macrophage colony-stimulating factor (GM-CSF) (Myelogen; Schering-Plough, Dardilly, France) and 20 ng/mL IL-3 (ProSpec, Rehovot, Israel). pDCs were finally collected as cytospin and fixed in 95% ethanol for 5 minutes before immunostaining.

The antibodies used in this study for immunohistologic staining are detailed in Table 2.

Single immunostaining, double immunoenzymatic, and immunofluorescent labeling were performed on tissue sections and on cytospin preparations of peripheral blood mononuclear cells, as described previously.^14,16  In some experiments cytospin preparations were subjected to antigen retrieval in a microwave oven in EDTA buffer prior to immunostaining.

Paraffin-embedded tissue sections of reactive human tonsils containing abundant pDCs were screened with antibodies against a range of leukocyte-associated molecules to identify markers strongly and selectively expressed in these cells. Some of the molecules evaluated were chosen because they had been documented in pDCs in the literature (but usually only at the level of mRNA expression, often in mice) and none had been studied previously in human tissue by immunohistologic techniques (with the exception of BCL11A, reported recently in pDCs in a single publication¹⁷ ). The other molecules evaluated were randomly selected known leukocyte-associated markers.

The adaptor protein CD2AP emerged from this immunohistologic screening as a selective marker of pDCs, being expressed uniformly throughout the cytoplasm of these cells (Figure 1). It was not found in other cells in peripheral lymphoid tissue (with the exception of very weak labeling of mantle zone B cells in some samples and rare cells in germinal centers). CD2AP was originally cloned from T cells,^18-20  but in the present study peripheral T cells were consistently CD2AP negative (using 3 different antibodies) regardless of antibody dilution or whether the tissue section came from a paraffin-embedded or cryostat sample. Endothelial cells and tonsillar squamous epithelium were weakly to moderately positive. In addition, 18 other molecules were expressed by cells with the typical features of pDCs, as summarized in Table 3. These markers differed greatly in the degree to which they were expressed in cell types other than pDCs, but many molecules were also present in B cells (Table 3).

Double immunolabeling was performed to explore the relationship between the new cytoplasmic pDC marker CD2AP and the known pDC-associated transcription factor BCL11A. This revealed that BCL11A and CD2AP were expressed in the same cells (Figure 1), and in double immunoenzymatic labeling studies of the phenotype of pDCs (Figures 1,Figure 2,Figure 3–4) these 2 markers were therefore used interchangeably to reveal pDCs (eg, CD2AP was used in combination with nuclear markers and BCL11A in combination with cytoplasmic molecules).

A number of B cell–associated transcription factors, namely ICSBP/IRF8, the products of the E2A gene (E12 and E47) and FOXP1 (Figure 1) were found in pDCs, and PU.1 was very weakly expressed in a minority of these cells. In contrast, 11 other transcription factors (eg, BCL-6, BOB.1, and PAX-5; Table 4), all but 3 of which are associated with the B-cell lineage, were absent (Figure 1). The expression of leukocyte-associated molecules involved in intracellular signaling was also explored and this revealed the presence of 5 B cell–associated molecules (the adaptor protein BLNK; the kinases BTK, Lyn, and Syk; and the PLCγ2 phospholipase) and also the transmembrane adaptor protein LIME (Lck-interacting membrane protein, which is found in T cells and plasma cells; Figure 2, Table 3). However, other T cell–associated signaling molecules (eg, TRIM, SLP76) were absent (Figure 2, Table 4). In addition, 3 signaling molecules that have not been studied previously in human tissues by immunocytochemistry were detected in pDCs, namely DAP12 (DNAX-activation protein 12, known also as KARAP), IRAK1, and TCB1D4 (Figure 2, Table 3). Furthermore, the Toll-like receptors TLR7 and TLR9 were clearly present in human pDCs (Figure 3, Table 3), in keeping with data from studies of murine cells.^21,22 

Two other cell types present in interfollicular areas in human tonsil, namely classical interdigitating dendritic cells (expressing DC-SIGN) and presumptive immature lymphoid cells (expressing TdT), were investigated by double labeling (immunofluorescence for DC-SIGN and immunoenzymatic for TdT) in combination with CD2AP (Figure 3). There was no overlap in expression of these markers, providing further evidence that CD2AP is confined to pDCs.

pDCs have been extensively evaluated in the past for classical surface/CD molecules associated with lymphoid and myeloid lineages and they are known to express only a small number, for example, CD4, CD68. Such lineage markers were therefore not explored in detail in this study, but we confirmed the expression of CD4 and CD68 in CD2AP-positive cells and also noted that CD79b (but not its partner, CD79a) was weakly expressed. Furthermore, we studied CD123 (a well-recognized marker of pDCs) in relation to the transcription factor BCL11A by double labeling and showed the latter molecule in all CD123-positive cells (Figure 3), further confirming its selectivity for pDCs. CD33, which has been reported in some pDC-derived neoplasms,^23,24  was also studied, and it was found on occasional pDCs (as defined by expression of BCL11A).

The pDCs that are known to accumulate in non-neoplastic conditions (Kikuchi lymphadenitis,^25,26  Castleman disease,^27,28  lupus erythematosus,^29,30  and lichen planus³¹ ) were all sharply delineated (mainly in the form of nodular aggregates but also as scattered cells) by immunostaining for CD2AP (Figure 3). In biopsies of cutaneous lupus lesions, the pDCs appeared identical to those seen in the tonsil, whereas some of the pDCs in Kikuchi disease were a little larger, with a more voluminous cytoplasm. The phenotypic profile of pDCs in the latter disease was investigated with the full panel of markers to see if it differed from that of tonsillar pDCs, but no differences were observed (Figure 3). It was also noted that CD2AP-positive cells in Kikuchi lymphadenitis were myeloperoxidase negative (observed by using double immunofluorescence), showing they were not immature monocytes (usually present in this disease and positive for myeloperoxidase).

Cytospin preparations of normal peripheral blood mononuclear cells contained rare CD2AP-positive cells, accounting for less than 0.5% of all cells (in 2 healthy donors; Figure 4). Their morphology was relatively constant, that is, medium-sized cells, often with slightly eccentrically placed nuclei some of which were indented. These CD2AP-positive cells all coexpressed the pDC-associated transcription factor BCL11A, and they lacked CD3 and CD20 (Figure 4). Furthermore, pDCs isolated from peripheral blood using anti–BDCA-4 all showed strong staining for CD2AP (Figure 4) and CD123. BDCA-4 is a well-accepted marker for purification of pDCs,^32,33  and it was used previously by 2 authors of this paper (F.F. and S.S.) to purify pDCs, achieving a purity of 90% to 98%.³⁴ 

Bone marrow trephine biopsies also contained scattered CD2AP-positive cells, very similar in morphology to those seen in peripheral lymphoid tissue, and all coexpressed the transcription factor BCL11A (Figure 4).

A total of 47 pDC-derived neoplasms (Table 1) were investigated for the expression of the pDC-associated molecules listed in Table 3 (with the exception of IRF7, LIME, and PLCγ2). However, because of the limited material available, it was not possible to test each case with all markers. In addition, biopsies from cases of acute myeloid leukemia in the skin (leukemia cutis) were studied, since these can cause diagnostic confusion, together with samples of chronic myeloproliferative disorders and acute lymphoblastic leukemia (Table 5).

Most of the pDC markers we evaluated were expressed on most of the cases of pDC neoplasia (Figure 5, Table 5). One of the 16 pDC-associated molecules, CD2AP, appeared to have particular potential diagnostic value, since it was present in neoplastic pDCs in most cases (41 of 43), but absent in all but one of the 24 leukemia cutis cases analyzed (and in all other myeloid neoplasms studied). The B-cell–associated transcription factors BCL11A and ICSBP/IRF8 were also commonly found in neoplastic pDCs (42 of 44 and 34 of 36 cases, respectively), but they were also found in a minority of cases of leukemia cutis (6 and 5 of 24 cases, respectively). Many of the other pDC-associated molecules studied (eg, BTK, DAP12, Lyn) were found in all cases of pDC neoplasia, but were also expressed in most leukemia cutis samples. Furthermore, markers expressed in normal B cells (eg, BLNK, BTK, Lyn) were commonly found in B-cell acute lymphoblastic leukemia (Table 5).

Plasmacytoid dendritic cells were first identified 50 years ago by Lennert and his associates in human lymph nodes as a population of cells with morphologic features of plasma cells lying in interfollicular T-cell–rich areas.³⁵  It was subsequently shown that they express CD36 and CD68, suggesting a monocyte/macrophage origin, and they came to be known by many authors as plasmacytoid monocytes.³⁶  This was supported by reports that neoplastic proliferations involving these cells (eg, in lymph node, spleen, bone marrow) are always associated with a myeloproliferative disorder (principally acute or chronic myelomonocytic or monocytic leukemia).^37-41  This suggested that both proliferative processes share a common origin, and this has been supported more recently by reports of identical cytogenetic abnormalities in the 2 cell populations in cases of myelodysplasia,^42,43  and in a case of acute myeloid leukemia.⁴⁴ 

Plasmacytoid monocytes/T cells were studied by many pathologists in the following years, and prominent clusters of these cells were recognized as a feature of Kikuchi disease,^25,26  and the hyaline vascular subtype of Castleman disease.^27,28  However, they received relatively little attention until it was shown that they secrete large amounts of type I interferons (principally interferon-α),^1,45  a function that accounts for their extensive rough endoplasmic reticulum.²⁹  It was also reported that similar cells could be generated in vitro from mononuclear or CD34-positive cells in the peripheral blood and that activation by factors such as interleukin 3 (IL-3) or CD40L generates cells of dendritic morphology with antigen-presenting capability.^4,46,47  For this reason they have come to be widely known as plasmacytoid dendritic cells.^3,4  At the same time a murine counterpart was identified with similar properties,^48-51  and many published studies are based on pDCs from this species.

It has been reported in recent years that pDCs can give rise to a second tumor type that appears distinct from the rare neoplasms associated with myeloproliferative disorders.^23,52-58  This tumor entity was initially believed to arise from NK cells (because of its expression of CD56^59-61  and categorized as blastic NK cell lymphoma in the WHO classification, but it was subsequently renamed CD4⁺/CD56⁺ hematodermic neoplasm in the WHO/EORTC (European Organisation for Research and Treatment of Cancer) classification scheme for cutaneous lymphomas.^10,62  These tumors have many phenotypic features of pDCs^54,63,64  and are characterized by skin lesions, frequent involvement of other tissues (marrow, spleen, and/or lymph nodes), either initially or later in the course of the disease, and circulating neoplastic cells. The disease often responds initially to therapy, but overall it has a poor prognosis.^{23,55-57,61,63-67} 

In leukemic cases, the diagnosis is made principally by flow cytometry.^66,68  However, cases are commonly first diagnosed in a skin biopsy, requiring markers that can be detected in paraffin-embedded tissue.⁶³  The first markers to be used (CD4, CD56, HLA-DR, CD123, and CD45RA^54,63 ) have more recently been supplemented by TCL1 and CLA (HECA-452),^10-12,57  but all these markers are also present on neoplasms of non-pDC origin. A type II C-type lectin BDCA-2 has also been reported as a specific marker of pDCs⁶⁹  and evaluated on pDC tumors.^64,70-72  However, there is little information on its expression in non-pDC neoplasms, and it is only expressed in a proportion of pDC tumors. Siglec-H has also been reported as a surface constituent on pDCs^73,74  but not exploited as a diagnostic marker.

In this paper we report a number of new immunohistologic markers that can be used to detect normal and neoplastic pDCs in human tissue samples. The adaptor protein CD2AP is of particular interest because it was essentially restricted to pDCs. CD2AP is an 80-kDa molecule first identified through its binding to the terminal 20 amino acids in the cytoplasmic domain of the T-cell–associated molecule CD2.^18,19  CD2AP has also been shown to play a role in podocyte homeostasis in the kidney (CD2AP-deficient mice develop nephrotic syndrome and glomerulosclerosis^75,76 ). CD2AP is described in the literature as a component of normal T cells (and we expected to see expression in these cells when assessing its immunohistologic reactivity in peripheral lymphoid tissues). However, the only staining observed in T cells was in cortical thymocytes (data not shown) and 3 different antibodies (all of which react with cells transfected with the CD2AP gene; manuscript in preparation) gave identical immunohistologic labeling. Furthermore, we confirmed by double staining for BCL11A and CD2 (data not shown) reports from the literature that CD2 (the partner for CD2AP) is not detectable in pDCs in tissue sections²  (although it has been reported in a minority of cases of pDC neoplasia).^{2,53,57,63,64,66,77}  In consequence, CD2AP may associate with a hitherto unidentified partner other than CD2 in normal pDCs. It may be added that, while CD2AP expression has been documented in mouse T cells, its expression pattern in human T cells is unclear, having only been demonstrated to be expressed in the T-cell leukemia cell line, Jurkat cells.^18,19,78,79  It is therefore possible that there are species-specific differences in the pattern of CD2AP expression. It may be added that Northern blotting studies have suggested that CD2AP is widely expressed in human tissues (with the exception of brain),^18,80  but Dustin et al (1998) reported a more restricted range of protein expression (assessed by Western blotting).¹⁸  Whatever the explanation of the unexpected absence of immunostaining for CD2AP in peripheral T cells in human tissue samples, it clearly emerged in this empirical study as a selective and reproducible marker of pDCs.

Nine other signaling molecules were also shown for the first time in this study to be detectable at the protein level in normal and neoplastic human pDCs (Table 3). Some had been implicated previously at least as possible components of pDCs, usually in murine studies, that is, BLNK, Btk, DAP12, IRAK1, and Syk. However, others have not been associated with this cell type; for example, the GTPase-activating protein TBC1D4 (also known as AS160). Also the strong expression of TLR7 and TLR9 we found in human pDCs (Figure 3) is in keeping with reports of their transcripts in human and murine pDCs.^22,81,82 

There is evidence that pDCs can arise from both lymphoid and myeloid precursors,^83-88  and it has also been reported that they show features associated with several lineages, that is, T cells, B cells, and myeloid cells.^{58,77,85,89,90}  Furthermore, the lymphoid precursor-associated marker TdT is expressed at the mRNA level in murine pDCs and as protein in neoplastic human pDCs.^{8,56,63,65,91-93}  In consequence, the cellular origin of pDCs has been the subject of controversy, to which the present study might be expected to bring some new insight. It was striking that many markers detected in normal and neoplastic pDCs are expressed in B-lymphoid cells (Table 3), including 4 transcription factors (BCL11A, the products of the E2A gene, ICSBP/IRF8 and PU.1) and 5 signaling molecules (BLNK, BTK, Lyn, PLCγ2, and Syk). In addition, LIME and TBC1D4 are found in B cells at the plasma cell⁹⁴  and germinal center maturation stages, respectively (Figure 2). This suggestion that pDCs are related to B cells is in keeping with reports that signaling initiated from BDCA-2 (CD303) resembles the effect of B-cell receptor ligation (eg, both involve Syk and BLNK phosphorylation).^95,96  Interestingly, the consequence of this stimulation is to inhibit rather than to stimulate interferon (IFN) production.^69,97,98  However, 5 classical B cell–associated transcription factors we studied (eg, OCT2, PAX5) were not expressed by pDCs (Table 4). Furthermore, 3 other B cell–associated molecules we identified (Lyn, PU.1, and Syk) are also found in myeloid cells, and another novel marker, DAP12, is expressed in macrophages but not in B cells (Figure 2, Table 3).

Thus, the phenotypic studies in this paper shed only limited light on the controversial question of the cellular origin of pDCs. We also found no evidence to support the report by Pelayo et al⁸  that different populations of plasmacytoid dendritic cells (pDC1 and pDC2) can be distinguished on the basis of phenotypic differences. They reported that BCL11A, ICSBP/IRF8 and PAX5 are all positive in pDC1 and negative in pDC2, but in the present study we found that these markers (and indeed other markers evaluated) were either present (BCL11A, ICSBP/IRF8) or absent (PAX5, mb1/CD79a) in all pDCs. The homogeneity of pDC marker expression in the present paper therefore argues against the existence of subsets of human pDC1 comparable to those reported by Pelayo et al in mice.⁸  It has also been suggested that the expression of cell markers (eg, BDCA-4 and CD7) changes during pDC maturation and that the clinical behavior of pDC-derived tumors is related to the maturation stage from which they arise.⁷¹  However, the markers documented in the present study did not show variability suggestive of major phenotypic alteration during differentiation.

The other controversial topic concerns the degree to which circulating pDCs become cells with classical dendritic morphology, possessing the ability to present antigen, when they enter peripheral lymphoid tissue. This is a widely held belief and accounts for the use of the term “dendritic” and for the fact that on occasion circulating pDCs are referred to as “precursors,” implying that their morphology changes when they emigrate into tissues. In the present study the morphology of cells in peripheral lymphoid tissue carrying pDC markers appeared very similar to that of the cells seen in the bone marrow and in the circulation. Furthermore, although some of the numerous pDCs present in Kikuchi disease showed a slightly more voluminous cytoplasm, for the most part they lacked dendritic surface processes (Figure 3), and we detected no acquired novel phenotypic features. It is therefore possible that the ability of pDCs to transform into cells with the morphology and function of classical dendritic cells is an in vitro phenomenon rather than a common physiologic event in vivo.

A major aim of the present paper was to identify new markers present on neoplastic pDCs. We therefore analyzed a total of 43 cases of hematodermic neoplasia, and also 6 cases of pDC proliferation associated with a myeloproliferative disorder (Tables 1 and 5). The phenotype of these 2 types of pDC neoplasia matched published data, including the fact that in the latter disorder, TdT was not expressed and CD56 and TCL1 were less commonly expressed than in hematodermic neoplasia (Table 1). A significant diagnostic problem is posed by cutaneous deposits of diseases such as myeloid leukemia (leukemia cutis) since these are morphologically similar to hematodermic neoplasms and some express CD4 and CD56. We therefore evaluated the most selective pDC markers on 24 cases of this sort.

This analysis showed that many of the markers we documented on normal pDCs were expressed on their neoplastic equivalents. Most of these markers were also expressed on neoplasms of B-cell and/or myeloid lineage. However only a minority of leukemia cutis samples (the neoplasms most likely to be confused with the CD4⁺CD56⁺ hematodermic neoplasm) expressed BCL11A and BLNK, while CD2AP and ICSBP/IRF8 were even less commonly expressed (Table 5), indicating their potential value for the diagnosis of hematodermic neoplasms. It should be added that the frequency of CD2AP and ICSBP/IRF8 expression on myeloid neoplasms was comparable to, or lower than, that of the 2 recently proposed diagnostic markers TCL1 and CLA.^11,12,57 

The number of cases associated with a myeloproliferative disorder was small, but it may be of significance that some cases in this category lacked the transcription factors BCL11A (2 of 5), E47 (1 of 3), and ICSBP/IRF8 (2 of 3), whereas these molecules were present without exception in cases of hematodermic neoplasia. Conversely, PU.1 (a transcription factor found not only in B cells but also in macrophages) was present in each of 3 cases associated with a myeloproliferative disorder but absent in 11 of 16 hematodermic tumors.

In conclusion, we have documented a range of phenotypic markers of normal and neoplastic human pDCs in tissues and peripheral blood. Our observations do not resolve the controversy surrounding the cellular origin of these cells (beyond reinforcing reports that they share many molecular features with B cells), but they present a picture of the plasmacytoid dendritic cell as a cell that changes little in morphology or phenotype between the bone marrow and peripheral tissue, even when it accumulates in large numbers in diseases such as Kikuchi lymphadenitis or following neoplastic transformation. These new markers include molecules (in particular the adaptor protein CD2AP and the transcription factor ICSBP/IRF8) that may be of value for the diagnosis of pDC tumors.

The authors thank Mr Ralf Lieberz for his technical assistance, and Mrs Bridget Watson for her expertise in the preparation of the manuscript. The authors thank all the members of the French Study Group on Cutaneous Lymphomas for their collaboration.

This work was supported by Project Grant (no. 0382) and Program Grant (no. 04061) from the Leukemia Research Fund and by the Julian Starmer-Smith Lymphoma Fund.

Contribution: T.M. designed the project, analyzed and interpreted the data, and wrote the paper. J.C.P., E.B., and S.T. were responsible for cell and tissue preparations as well as performing all the immunohistologic experiments. K.K.R., K.H., M.D., M.-L.H., S.A.P., and T.P. provided tissue samples. M.J.D., I.D., and A.S.S. provided novel reagents. S.S. performed part of the experiments. H.S. and P.G.I. provided tissue samples and also reviewed the results. F.F. provided tissue samples, reviewed the results, and contributed to writing the paper. D.Y.M. contributed to the design of the study, interpreted the data, and wrote the paper.

The work was carried out in the Leukaemia Research Fund Immunodiagnostics Unit, Nuffield Department of Clinical Laboratory Sciences, John Radcliffe Hospital, Oxford, United Kingdom.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Teresa Marafioti, Leukaemia Research Fund Immunodiagnostics Unit, Nuffield Department of Clinical Laboratory Sciences, John Radcliffe Hospital, Oxford, OX3 9DU, United Kingdom; e-mail: teresa.marafioti@ndcls.ox.ac.uk.