CD4+/CD56+ haematodermic neoplasm or ‘early’ plasmacytoid dendritic cell leukaemia/lymphoma (pDCL) was described as a disease entity in the last World Health Organisation/European Organisation for Research and Treatment of Cancer classification for cutaneous lymphomas. These leukaemia/lymphomas co-express CD4 and CD56 without any other lineage-specific markers and have been identified as arising from plasmacytoid dendritic cells. Despite a fairly homogeneous pattern of markers expressed by most pDCL, numerous distinctive features (e.g. cytological aspects and aberrant marker expression) have been reported. This may be related to the ‘lineage-independent developmental’ programme of dendritic cells, which may be able to develop from either immature or already committed haematopoietic progenitors. This highlights the need for specific validated markers to diagnose such aggressive leukaemia. Here, we propose –among others (e.g. T-cell leukaemia 1) – blood dendritic cell antigen-2 and high levels of CD123 expression as potential markers. In addition, we propose a multidisciplinary approach including several fields of haematology to improve pDCL diagnosis.
The ontogenic origin of these disorders has not been clearly identified, with much debate over myelomonocytic or mixed NK/myelomonocytic precursors. However, Lucio et al (1999) had suggested a dendritic origin for an unusual case of lymphoma. Subsequently, three studies (Chaperot et al, 2001; Feuillard et al, 2002; Petrella et al, 2002) demonstrated that these leukaemic cells shared the same characteristics as plasmacytoid dendritic cells (pDC). CD4+CD56+ leukaemia could now be considered as the malignant counterpart of normal pDC (Chaperot et al, 2001; Feuillard et al, 2002; Jacob et al, 2003). The term pDC-derived acute leukaemia (pDCL) was thus proposed; this term is used throughout this review. Current proposals that update the classification of T-cell lymphoma indicate clearly that blastic NK-cell lymphoma has been incorrectly categorised and corresponds to pDCL (Nava & Jaffe, 2005). Finally, the recent WHO/European Organisation for Research and Treatment of Cancer (EORTC) classification for cutaneous lymphomas classified this entity as CD4+/CD56+ haematodermic neoplasm or ‘early’ plasmacytoid dendritic cell leukaemia/lymphoma (pDCL) (Slater, 2005; Willemze et al, 2005). Despite this classification and an increasing number of case reports, biological data that enable a clear definition of this entity, and an unequivocal diagnosis in particular, are too scarce. This is an important point; although pDCL is relatively sensitive to multiagent chemotherapy, relapses are frequently observed with an aggressive clinical course (survival median: 9–12 months, Reimer et al, 2003). Indeed, only allogeneic haematopoietic cell transplantation can lead to complete remission (Reimer et al, 2003). However, the clinical course is not always aggressive (Ng et al, 2006) especially in young patients (Falcao et al, 2002; Rossi et al, 2006). This review, with an emphasis on critical data, aims to aid pDCL diagnosis. In addition to the review by Jacob et al (2003), who first proposed the diagnostic criteria for pDCL, we discuss the potential importance of new markers [such as T-cell leukaemia 1 (TCL1), MxA protein] and report our own experience with ‘pDCL-specific’ markers (e.g. CD123). Some of these markers (e.g. CD33) may represent a new direction for the therapeutic management of this haematological disorder.
Prevalence and clinical presentation
Plasmacytoid dendritic cell leukaemia is a rare phenomenon, representing <1% of acute leukaemia cases (Jacob et al, 2003) and 0·7% of cutaneous lymphomas (Ng et al, 2006). Its clinical presentation is relatively homogeneous. However, pDCL predominantly affects males, with a sex ratio of 3:1. It mainly affects the elderly with an average of 67 years (range <1–90 years), although some paediatric cases have also been reported (Feuillard et al, 2002) with a less aggressive clinical course (Rossi et al, 2006).
Its clinical presentation at the time of diagnosis usually consists of an isolated cutaneous lesion that rapidly evolves (in a matter of months) to multiple sites, proliferates into the blood, bone marrow (BM), lymph nodes (LN) and other areas, such as the spleen, liver, central nervous system (CNS), tonsils, lungs, kidneys and muscles (Willemze et al, 2005). However, a few rare cases of isolated cutaneous forms without clinical evolution for >6 months have been reported (Brody et al, 1995). Skin lesions, present in 90% of patients, are the most common feature and are often the main motivation for seeking medical advice. Lesions are gathered in one area or may be diffuse, lacking a clear topographical pattern. They range in size from a few millimetres to over 10 cm, and while they invade the dermis, the epidermis is unaffected. Lesions are often described as sub-cutaneous blotches, papules, tumefactions or nodules; or as being erythematous, highly pigmented, purpuric, necrotic, dark red or purplish (Petrella et al, 2002). The histopathological analysis of cutaneous lesions shows an infiltration of mononuclear cells into the dermis without angiotropism, that does not damage the vessels reaching the adipic tissue and always spares the epidermis (Petrella et al, 2002; Ng et al, 2006). This preferential localisation of pDCL is associated with the expression of markers frequently encountered in other cutaneous lesions, such as CD56, CD43 (Reichard et al, 2005; Urosevic et al, 2005) or Cutaneous Lymphocyte Associated antigen (CLA) (Petrella et al, 2004a). The overall health of patients at presentation is usually quite good, but the clinical course is very aggressive and rapidly fatal.
Cytology – cytochemistry
At diagnosis, most patients present with cytopenia (thrombocytopenia in two-thirds of cases, anaemia and neutropenia in the remainder) associated with BM infiltration (in >80% of cases) at extremely variable rates (Jacob et al, 2003). Hyperleucocytosis is infrequent but the presence of blasts in the blood is more commonly observed. Blast morphology is pleomorphic (Fig 1) with cell size varying from small to large with a regular-shaped round or oval nucleus. The chromatin nucleus is thin and blast-like. The cytoplasm is of average abundance, slightly basophilic, non-granular but it displays a heterogeneous structure, with a ring or ‘pearl necklace’ of microvacuoles beneath the cytoplasmic membrane (Fig 1). This may correspond to the pinocytosis vacuoles observed in normal pDC (Kohrgruber et al, 1999). The cytoplasmic membrane often exhibits pseudopods. Cytochemical reactions for peroxidase and butyrate esterase are negative (Feuillard et al, 2002). This is an important test to perform in order to complete phenotypic analysis. Association of myelodysplasia – concerning one to three lineages – with pDCL or a history of myelodysplastic syndrome before pDCL were frequently reported (Feuillard et al, 2002; Khoury et al, 2002; Kazakov et al, 2003). This may suggest a common origin of pDC and the myeloid lineage.
Markers expressed by cells of plasmacytoid dendritic cell leukaemia
Plasmacytoid DCL cells were characterised initially by the co-expression of CD4+ and CD56+, without B-, T-lymphoid, NK-cell, or myeloid lineage specific markers. However, pDCL can express markers (or antigens) from different lineages (Fig 1; Jacob et al, 2003). The expression of such markers can be due to aberrant marker expression – a frequent feature of leukaemic cells (Garand & Robillard, 1996) – or related to the pDC origin. Interpretation is rendered difficult as markers expressed by normal pDC are not fully characterised or may vary from one site to another (Briere et al, 2002), as well as after the maturation process (Bendriss-Vermare et al, 2001). In addition, single studies have never reported >25 cases of pDCL and the reported cases were not homogenous in terms of clinical presentation or cell morphology (Chaperot et al, 2001; Feuillard et al, 2002). Moreover, marker expression was analysed by different techniques (immunohistochemistry on skin biopsies or flow cytometry on cells in suspension as well as molecular biology at transcript levels) and at different sites (blood, BM and skin). Based on literature data (Chaperot et al, 2001; Feuillard et al, 2002; Jacob et al, 2003; Chaperot et al, 2004; Garnache-Ottou et al, 2005a; Gopcsa et al, 2005), pDCL can be identified by the following phenotype: CD45low, CD4+, CD56+, CD116low, CD123high (IL-3α receptor), HLA-DR+, CD45RA+, CD45R0−, BDCA-2+ (blood dendritic cell antigen-2 or CD303+), BDCA-4+ (CD304+), ILT-3+ (immunoglobulin-like transcript-3). This phenotype is therefore identical to that of normal pDC (Jacob et al, 2003), with the exception of CD56 expression, a characteristic feature of pDCL. However, CD56 expression is present in a limited sub-population of normal pDC (Comeau et al, 2002; MacDonald et al, 2002) thereby representing the normal counterpart of pDCL. A recent study identified CD56 as a marker for murine immature pDC (Allman et al, 2006). It should be stressed that certain cases of pDCL reported in the literature lack CD56 expression (Lucio et al, 1999; Momoi et al, 2002; Bueno et al, 2004; Petrella et al, 2004b; Reichard et al, 2005). In one of the cases reported in the GEIL study, CD56 was expressed in only 10% of cells although leukaemic cells had been functionally characterised as pDCL (Chaperot et al, 2001). Thus, CD56 does not appear to be a critical marker for pDCL although this marker is very useful for diagnosis. It is therefore likely that cases of CD4+CD56− lin− acute leukaemia (or expressing certain lineage markers, but with a score <2 according to the European Group for the Immunological Characterisation of Leukaemia (EGIL) score; Bene et al, 1995) correspond to CD56− pDCL. This type of acute leukaemia is currently classified as undifferentiated ‘stem cell leukaemia’ (Testa et al, 2002) or myeloid leukaemia if the expression of some myeloid associated-antigens is detected. Moreover, the study by Trimoreau et al (2003) demonstrated the loss of specificity of the CD4+CD123+CD45RA+CD45RO−CD116− profile when CD56 was lacking. In these situations, pDCL-specific markers will be useful.
A case of pDCL expressing CD56 but not CD4 has also been described (Ng et al, 2006). This was observed in a skin biopsy by immunohistochemistry at the time of diagnosis. At relapse, the flow cytometry analysis of the pDCL cells documented CD4 expression (Ng et al, 2006). This illustrates how complex pDCL diagnosis can be. It would be useful to examine all these reported cases more closely using pDC-specific markers.
Identification of pDC specific markers is an evolving field with the characterisation, for instance, of BDCA-2 and BDCA-4 (Dzionek et al, 2000). The use of BDCA-2 and -4 markers, still restricted to a few studies, is potentially of great interest to identify pDCL. Indeed, BDCA-2 expression was found in six cases of functionally characterised pDCL of seven tested and a low BDCA-4 expression was found in all cases (Chaperot et al, 2004). Expression of BDCA-2 and -4 was also found in three paediatric cases (Rossi et al, 2006) and two adult cases of pDCL (Anargyrou et al, 2003; Garnache-Ottou et al, 2005a). Expression of pDC specific markers was also analysed by immunohistochemistry (Urosevic et al, 2005). Our preliminary results suggested that BDCA-4 is not specific for pDCL as 10% (7/69) of other acute leukaemias expressed BDCA-4 at the same levels as pDCL (Garnache-Ottou et al, 2005b). This was expected as BDCA-4 (also know as neuropilin-1) has been reported to be expressed by other cells than pDC (Dzionek et al, 2000; Kuwabara et al, 2003; Bruder et al, 2004). Neuropilin-1 mRNA was expressed by all of the five analysed acute myeloid leukaemia (AML) blast cells (Schuch et al, 2002). Unlike BDCA-4, BDCA-2 was never expressed in the 69 tested acute leukaemias (Garnache-Ottou et al, 2005b). However, BDCA-2 expression was not systematically detected in the tested pDCL (four negative cases of the 14 tested, Garnache-Ottou et al, 2005b), as reported by others (Anargyrou et al, 2003; Chaperot et al, 2004). More simply, the level of CD123 expression – determined by cytometry and assessed by fluorescence intensity ratio (FIR) – may discriminate pDCL from other acute leukaemias. Our preliminary results showed that other types of acute leukaemias express lower levels of CD123 [FIR: 22 (mean) ± 8 (SEM); range: 2–71; n = 69] when compared with pDCL (FIR: 171 ± 36; range: 71–353; n = 14) (Garnache-Ottou et al, 2005b). This contrasts with the WHO/EORTC recommendations (Willemze et al, 2005), which did not take into account the high expression of CD123 but only its presence.
Other markers, such as TCL1 may be useful for pDCL diagnosis (Willemze et al, 2005). TCL1 is expressed during lymphoid differentiation but is not expressed in mature B and T cells. In normal LN, TCL1 expression is restricted to certain immature B cells and pDC (Herling et al, 2003). In pathological situations, TCL1 is involved in the leukaemogenesis of mature T cells (Virgilio et al, 1994; Pekarsky et al, 2001) and B-cell malignancies (Narducci et al, 2000). Herling et al (2003) found that TCL1 was expressed by most CD4+CD56+ cutaneous pDCL. At relapse after blastic transformation of CD4+CD56+ pDCL in CMML, leukaemic cells co-express both TCL1 and myeloid markers (Herling et al, 2003). In contrast, TCL1 expression was not frequently encountered in de novo myeloid leukaemia (Herling et al, 2003), and these observation were confirmed by Petrella et al (2004a). This suggests that TCL1 is a good marker and that pDC could be related to a myeloid lineage.
Normal pDC exhibit a particular chemokine receptor profile responsible for the specific homing of such cells in the inflamed LN and in the skin. Leukaemic cells derived from patients with pDCL display a chemokine receptor pattern similar to that of circulating pDC (CXCR3+CXCR4+CXCR2+CCR7+low) and some cases express CCR2, CCR5 and CCR6 (Bendriss-Vermare et al, 2004). A recent study confirmed the expression of CXCR4 and CCR7 on pDCL cells (Gopcsa et al, 2005). However, the usefulness of these LN or skin homing markers in the pDCL diagnosis is limited since CD62L and CXCR3 expression were also found in 30 AML, acute lymphoblastic leukaemia and lymphoma, as also observed for CLA (Petrella et al, 2004a).
An extensive characterisation of markers expressed by pDCL will enable the possible use of monoclonal antibodies to treat patients presenting with pDCL. Most pDCL express CD33 (see above), suggesting that gemtuzumab ozogamicin can be used as proposed for CD33+ AML (Tsimberidou et al, 2006). Alemtuzumab (anti-CD52, Campath-1G) induces depletion of circulating DC (Klangsinsirikul et al, 2002). However, pDC depletion seems to be less efficient than myeloid DC (Klangsinsirikul et al, 2002) and pDCL cells from five patients did not express CD52 (Gopcsa et al, 2005). Gene expression profiling of pDCL identified the upregulation of FLT3 mRNA (Dijkman et al, 2007). This observation is promising for pDCL treatments using FLT3-specific inhibitors.
At present, data are too scarce and additional studies are required to confirm the specificity (and the relevance) of all these markers.
Functional studies: the gold standard to identify plasmacytoid dendritic cell leukaemia
Plasmacytoid DC leukaemia cells possess antigen-presenting cell (APC) function thereby allowing their classification as members of the dendritic lineage, especially to pDC (Chaperot et al, 2001). Culture experiments have shown that, in the presence of IL-3 and CD40L, or in the presence of a virus, leukaemic cells can differentiate into mature DC and stimulate allogeneic naive CD4 T cells (Chaperot et al, 2001; Garnache-Ottou et al, 2005a). As described for normal pDC, cells derived from a patient with pDCL were shown to take up viral antigens and present them to specific CD4 and CD8 T cells (Chaperot et al, 2004). Moreover, pDCL secreted IFN-α in response to influenza virus (Chaperot et al, 2004; Garnache-Ottou et al, 2005a). However, not all pDCL produced IFN-α and, in general, pDCL secreted lower levels of IFN-α than their normal counterparts (Chaperot et al, 2004; Garnache-Ottou et al, 2005a), suggesting that pDCL exhibit altered responses to viral stimuli or that they represent a different stage when compared with natural IPC. Whether normal circulating CD56+ pDC are good IFN-α producers in response to viruses is unknown and remains to be determined.
On the other hand, under different progenitor culture conditions on methylcellulose, pDCL proved incapable of differentiating into NK cells, B cells, myeloid cells or monocytes (Jacob et al, 2003) and therefore differentiation into mature pDC only was possible, thus confirming their affiliation with these cells and the pDCL denomination. To date, the functional characterisation of pDCL is the only way to unequivocally diagnose this leukaemia. This characterisation is however time-consuming, requires a large number of cells and specialised laboratories.
At the time of diagnosis, two-thirds of pDCL cells showed cytogenetic anomalies (Leroux et al, 2002). In the majority of cases, karyotypes were complex and correlated within the same clone with an odd combination of recurrent anomalies affecting six chromosomal regions including: anomalies of the long arm of chromosome 5 targeting two regions (5q21 or 5q34), of the short arm of chromosome 12:12p13, of chromosome 13, loss of the long arm of chromosome 6 or deletion of 6q23-qter, monosomy 15p, and monosomy 9 (Leroux et al, 2002). Recurrent alterations in chromosome 4, 9 and 13 were also recently reported (Dijkman et al, 2007). These anomalies are all characteristic of lymphoid and myeloid pathologies and cannot be considered as specific of pDCL, although their association within the same clone is unusual.
The critical role of pDC in the control of viral infection by IFN-α release (Siegal et al, 1999) led to a search for a relationship with the viral status of the patients. A search for antigens associated with an infection by Epstein Barr virus (EBV) was performed and was negative except in two cases (for review, Jacob et al, 2003). This is in contrast with the association of EBV in a majority of mature NK pathologies. Relationships with other viruses, such as human immunodeficiency virus, hepatitis B and C virus, human herpes virus 8 and 6, cytomegalovirus and human T leukaemia virus-1 have not been established yet.
The differential diagnosis of pDCL warrants attention, especially those cases that do not exhibit a typical phenotype (see above) and when functional studies cannot be performed.
Nasal or extranasal CD56+ NK lymphoma are differentiated by the presence of azurophilic granules in their cytoplasm, by a frequent association with EBV infection. Although expression of CD56, CD2, CD7 or intracytoplasmic granzyme B can be common between these lymphoma and pDCL, these NK lymphomas never express CD4 (Harris et al, 2000).
CD33+ AML expressing CD4+ and CD56+ represent 10–20% of AML. In cases of AML expressing other myeloid markers and typical cytochemical reactions, AML diagnosis can be easily performed. However, it is warranted for very undifferentiated AML where ‘strong’ myeloid markers are weakly expressed or for undifferentiated monoblastic (esterase negative) AML. Identification of a pDC profile and pDC specific markers will be useful in such a situation.
Mixed myeloblastic/NK-cell leukaemia are types of myeloid leukaemia that co-express CD7 CD33 and CD56, markers that correspond to the proliferation of an immature precursor with myeloid and NK potential (Suzuki et al, 1997). The phenotype of such leukaemia is different from that of pDCL because, in all reported cases, CD4 and CD36 were not expressed although other myeloid markers (CD13, MPO and CD11b) were (Suzuki et al, 1997). Furthermore, mixed myeloblastic/NK-cell leukaemia cells always express CD34 whilst expression of CD34 by pDCL has only described in two cases (Khoury et al, 2002; Knudsen et al, 2002) among a hundred cases reported to date.
The final section of this review will discuss two critical aspects of pDCL: their ontogeny and their state of maturity.
Plasmacytoid dendritic cell leukaemia as a tool to reconsider models of haematopoiesis?
Despite the numerous studies published on normal pDC, the ontogenic origin of these cells is still a matter of debate. Arguments in favour of a lymphoid origin of pDC are widely reported in the literature (Table I and for review see Briere et al, 2002). Arguments in favour of a myeloid origin are provided mainly by data on pDCL. Normal and leukaemic pDC express myeloid-associated markers (e.g. CD33). Normal pDC were found to acquire myeloid markers CD11c, CD13 and CD33 in culture (Bendriss-Vermare et al, 2001). Although the first studies using clonogenic culture systems (for review see, Briere et al, 2002) showed pDC generation only from lymphoid progenitors, it was then shown that generation of pDC was also possible from a CD34+ progenitor expressing CD115 [macrophage colony-stimulating factor (M-CSF) receptor, thus associated to the myeloid lineage] (Olweus et al, 1997). The strongest argument for a myeloid origin comes from pDCL in relapse: they present as CMML or AML (Khoury et al, 2002; Herling et al, 2003). TCL1 expression attested that pDCL cells before relapse and blastic transformations (CMML or AML) were derived from a common cell since de novo AML never express TCL1 (Herling et al, 2003). The frequent association of myelodysplastic features in myeloid cells from the BM of patients with pDCL or a history of myelodysplastic syndrome (affecting the myeloid lineage) before the diagnosis of pDCL (Feuillard et al, 2002; Khoury et al, 2002; Kazakov et al, 2003) suggest a common origin of pDC and the myeloid lineage. One may propose that pDC are derived from both myeloid and lymphoid origin. A common precursor for the pDC and myeloid DC is strongly suggested by a report that both in vitro expanded myeloid DC and pDC exhibit the same chromosomal abnormality as that detected in the original myeloid leukaemic clone from which they arise (Mohty et al, 2001). Moreover, two pDC subsets that diverge at an early stage of differentiation were identified in murine BM: one expressing early lymphoid transcription factors (e.g. TdT) and the other expressing myeloid-associated CD115 mRNA (Pelayo et al, 2005). Thus, pDC may arise from different progenitors. Current models of haematopoiesis (Katsura, 2002) do not propose a binary vision of haematopoiesis (myeloid progenitors on one side and lymphoid progenitors on the other) but rather the presence of one or several myeloid/lymphoid progenitors that can give rise to myeloid, B-, T-lymphoid cells as well as DC. Thus, along the lymphoid lineages, myeloid potential is retained until quite late. This allows us to reinterpret data concerning pDC. Indeed, pDC may derive from one or more multipotent precursors, which explain the persistence of lymphoid lineage markers in these cells, and at the same time, a myeloid potentiality. The description of mouse IKDC (Taieb et al, 2006) that co-expressed markers related to pDC and NK lineage but diverged from a ‘classical’ NK developmental pathway sustains this hypothesis. Lastly, one has to consider the possibility, as proposed by Canque and Gluckman (2001) and Gluckman et al (2002), that DC do not belong to a particular haematopoietic lineage. Distinct DC populations develop from various haematopoietic progenitors (either immature or already committed to the lymphoid or the myeloid lineage) and even from terminally differentiated macrophages (Canque & Gluckman, 2001; Gluckman et al, 2002). This ‘lineage-independent developmental’ programme allows rapid leucocyte differentiation into APC. This does not render the diagnosis of pDCL any easier, because pDCL can express nearly all the lineage markers and display different morphological aspects.
Table I. Arguments in favour of the ontogeny origin of plasmacytoid dendritic cell leukaemia.
Arguments in favour of a
Ig, immunoglobulin; pre-Tα, the invariable pre-T cell antigen receptor α chain; ILT3, Immunoglobulin-like transcript 3; pDC, plasmacytoid dendritic cells; Id-2, inhibitor of DNA binding 2; Id3, inhibitor of DNA binding 3; pDCL plasmacytoid dendritic cell leukaemia; CMML, chronic myelomonocytic leukaemia; AML, acute myeloid leukaemia.
The ‘lineage independent developmental’ programme of pDC is in agreement with the absence of expression of haematopoietic progenitor markers (CD34, CD133, TdT and CD117). Thus, pDCL are not very immature. The term ‘immature’ should be considered here in the haematological sense and in the immunological sense (i.e. incompetent APC). This is why we propose to use the term pDCL rather than ‘early’ pDCL as proposed by the WHO/EORTC classification (Slater, 2005; Willemze et al, 2005).
The pDCL entity is currently characterised by phenotypic and functional criteria (Fig 1). We thus propose a multidisciplinary approach (Fig 1). To establish pDCL diagnosis, it is necessary not to exclude cases that share common characteristics with pDCL (CD56, CD4 expression, EGIL score <2) while presenting differences, such as myeloid marker expression, lack of CD56 expression, atypical cytological features, etc. Such cases require an analysis of pDCL-specific markers: CD123high or BDCA-2. These pDCL specific markers are currently being validated. Concurrently, larger studies of pDCL are required to validate these specific markers and identify new potential markers.
We thank Rod Ceredig and David Chalmers for reading the manuscript, Pascal Lepelley, Michel Degenne, Richard Garand and Sylvie Daliphard (all from the French pDCL network) for providing images of plasmacytoid dendritic cell leukaemia, the GEIL and the GFHC for providing pDCL cells. Our studies in this field are supported by grants from the Comité Départemental de la Ligue contre le Cancer du Doubs – Comité de Besançon, the GOELAMS and the Etablissement Français du Sang (nos 2003 and 2004-10).