Introduction

Castleman's disease was first described in a case report by Castleman and Towne1 in 1954, which was followed by a series in 19562. It is a unicentric or multicentric disease of the lymph node with or without polyclonal proliferation of B cells. Histopathology identifies a hyaline-vascular type, a plasma-cell variant or a mix of both.3 A number of high-quality reviews on Castleman's disease have been published.4,5,6,7,8,9,10,11,12 This Review discusses its stepwise disease progression as a consequence of chronic viral infection, which involves cytokine-driven switches between latency and replication in two reservoirs of differing cell proliferation dynamics.

Histopathology

Castleman's disease is characterized by significant architectural changes in all lymphatic compartments.Lymph nodes are enlarged and often transformed into scar-like areas.1,2 Consequently, a surgical sample of a diseased lymph node is most likely to provide the needed material for comprehensive histopathology while cytopathology is of limited value.13 Lymphoid follicles in Castleman's disease are increased in number and display fundamental abnormalities, the diagnosis of which has resulted in several redundant disease names such as follicular lympho-reticuloma,14 angiofollicular lymph node hyperplasia,15 angiomatous lymphoid hamartoma,16 benign giant lymphoma,17 and lymphoid or lymphnodal hamartoma;18 these terms should be superseded by the unifying term Castleman's disease. Two main histopathology variants have been identified (Figure 1 and Table 1): the hyaline-vascular type and the plasma-cell type. Frequent transitions between types have led to the identification of the mixed type that is reported in 15% of cases.3 The other major pathological classification scheme is that of unicentric versus multicentric disease. Keller et al.3 determined that unicentric disease is not equivalent to unifocal disease: “in each plasma-cell case, multiple discrete lymph nodes made up the clinically observed mass, unlike the single rounded mass found in most of the hyaline-vascular cases”, a finding confirmed later.11 In the absence of HIV infection, unicentric disease is most frequent, while HIV-positive patients invariably present with multicentric disease.

Figure 1: Histopathology of Castleman's disease.
figure 1

Hyaline-vascular Castleman's disease with a | eosinophilic hyaline scars and b | onion skin lesions. Plasma-cell variant Castleman's disease with c | hyperplastic germinal centers and d | sheets of mature plasma cells and paracortical plasma cell proliferation filling the sinusoids. Representative markers in plasma-cell variant Castleman's disease showing e | tight clusters of dendritic cells in the follicle centers with no meshwork expansion and f | presence of human herpesvirus type 8-infected plasmablasts in the vicinity of the germinal centers.

Full size image
Table 1 Histopathological features of Castleman's disease
Full size table

B-cell proliferation in Castleman's disease is varied and is initially polyclonal in origin.19,20 A key difference between hyaline-vascular disease and plasma-cell disease relates to the follicular dendritic cell network. This network is normal in plasma-cell disease, while it is expanded or disrupted in hyaline-vascular disease with multiple tight collections of follicular dendritic cells.21 Morphological findings in hyaline-vascular disease are close to pathognomonic. However, it is difficult to differentiate the plasma-cell disease from other conditions such as reactive changes in autoimmune disease, Castleman's disease-like changes associated with HIV infection, plasmacytic lymphoma, marginal zone lymphoma with plasmacytoid differentiation, plasmacytoma, and plasmablastic lymphoma. Under such circumstances, clinical information becomes crucial to establish the diagnosis: the coincidence of hyalinized germinal centers together with an intense interfollicular plasmacytic hyperplasia in a HIV-positive patient with accompanying fever, lymphadenopathy, splenomegaly and cytopenia is highly suggestive of multicentric Castleman's disease.10 Indeed, the lymph nodes of these individuals frequently contain Kaposi lesions, supporting the diagnosis.22 Detection of human herpesvirus type 8 (HHV8) latency-associated nuclear antigen (LANA) is crucial23 because multicentric disease is invariably associated with a positive HHV8 status. Absence of HHV8 virtually excludes HIV positive-related Castleman's disease.24,25 Up-to-date reviews on classification11 and viral association26 provide more detail on these issues.

Classification and clinical staging

The classification of Castleman's disease has historically been based on clinical presentation (unicentric disease versus multicentric disease) and histopathogenic type (hyaline-vascular disease versus plasma-cell disease).3,11 This defines unicentric hyaline-vascular disease and multicentric plasma-cell disease; however, significant overlap of these disease features establishes diagnosis of a mixed type in 15% of patients. Many patients are assigned to a type on the basis of predominant clinical findings rather than total concordance with either end of the spectrum. Coexistence of hyaline-vascular disease and plasma-cell disease has been identified in most patients in whom multiple lymph nodes have been examined in vivo or on autopsy.27 This fosters the argument of Keller et al.3 that hyaline vascular features might represent scars of plasma-cell disease and thus reflect the evolution of Castleman's disease over time rather than represent absolute categories of the disease.

Unicentric and multicentric disease are hence part of a continuous spectrum ranging from unifocal disease (one lymph node), and unicentric disease (one main lymph node surrounded by a group of satellites),2,3 to multicentric disease (multiple groups of lymph nodes in multiple locations).11 Accordingly, unicentric disease based in a solid organ, such as the spleen, is also identified in draining locoregional lymph nodes.22

A model of progressive viral infection

HHV8 (also called Kaposi sarcoma-associated herpesvirus) is a rhadinovirus.28 A large genome enables these viruses to modulate and evade the immune control of their hosts with frequent transitions between latent and replicating lytic viral disease.28 This transition explains the clinical variability between unapparent, smoldering, and progressive disease, focal and systemic illness, and benign scarring and malignant transformation. We consider Castleman's disease as a model case of this pleiomorphic process of adaptation and immune evasion, which depends on host factors and is deeply influenced by complementing viral infections.

HHV8

The prominent role of HHV8 in the pathogenesis of Castleman's disease is undisputed.12,29,30,31 HHV8 infection is ubiquitous in HIV-positive Castleman's disease, and present in the majority of HIV-negative patients with multicentric disease. Mechanisms of molecular viral pathogenesis are well understood on the cellular level, but less so on the level of tissues, organs and the whole body (Figure 2). HHV8 was found to be causally related to Kaposi sarcoma in HIV-positive and HIV-negative patients in the mid 1990s.25,32 The target of HHV8 infection in Kaposi sarcoma is vascular endothelial cells, which are subsequently transformed into the abundant spindle cells embedded in fibrovascular stroma typically containing hyalinized granules.33 The infection is mostly latent and viral replication occurs at a low level.31 HHV8 infection is also associated with primary effusion lymphomas or body-cavity-based lymphomas.31,34 These are B-cell-derived lymphomas that are often coinfected with Epstein-Barr virus (EBV).35

Figure 2: Viral pathogenesis of Castleman's disease.
figure 2

Circles correspond to the number of observed patients. Arrow width corresponds to mean IL-6 levels (the wider the arrow, the higher the IL-6 level). HHV8 particle numbers correspond to virus status. HIV particle numbers correspond to prevalence of infection status. Abbreviations: HHV8, human herpesvirus type 8; IL-6, interleukin 6.

Full size image

Kaposi sarcoma and Castleman's disease share a number of features: both conditions are inflammatory, neoplastic, related to HHV8 infection, associated with hyalinizing processes, produce a fibrovascular stroma, and are often associated with HIV infection. In HIV-positive patients with Castleman's disease, the lymph nodes are often also involved by Kaposi sarcoma,22 which usually affects endothelial cells.33 It has been hypothesized that the association of Castleman's disease and Kaposi sarcoma in a single lymph node is due to lytic HHV8 infection of B-lymphoid cells. This exposes susceptible endothelial cells to extreme HHV8 viral loads resulting in the formation of Kaposi sarcoma tumorlets in multicentric Castleman's disease-related lymph nodes.22 In our opinion, the pathogenic pathway might work both ways. Double immunostaining with antibodies directed towards latent nuclear antigen-1 (LNA-1) and a viral analog of interleukin 6 (vIL-6; encoded by HHV8 ORFK2)36,37,38 in both multicentric Castleman's disease and Kaposi sarcoma identifies a further important similarity: both HHV8-infected vascular cells and HHV8-infected lymphoid cells serve as infective reservoirs.39 In Castleman's disease, specific double immunostaining for HHV8 products LNA-1 and vIL-6 identifies infected lymphovascular epithelial cells, some of which exclusively express LNA-1 (corresponding to latency), while some coexpress LNA-1 and vIL-6.40 Expression of vIL-6 is induced by the lytic program of viral replication.40,41 The lytic cycle of HHV8 in lymphovascular cells leads to destruction of these cells resulting in a hyaline scar, while production of vIL-6 causes neo-angiogenesis with formation of penetrating vessels via basic fibroblast growth factor and VEGF.42 The result is an expansion of the lymphovascular epithelial cell reservoir of HHV8-infected cells (Table 1). In addition, the chemotactic effect of human IL-6 leads to cluster formation of plasmablasts around small vessels (Figure 3).

Figure 3: Pathogenesis of Castleman's disease in the perspective of a dysfunctional local cytokine network.
figure 3

In hyaline-vascular disease the virus is latent in the VEC and vIL-6 production occurs at a low level. In plasma-cell disease there is crossinfection of significant numbers of perivascular plasmablasts inducing massive viral replication resulting in high levels of vIL-6 and vIL-6-induced bFGF and VEGF. The multiple arrows in plasma-cell disease correspond to the exponential increase in IL-6 with an increase in HHV8 replication. Abbreviations: bFGF, basic fibroblast growth factor; HHV8, human herpesvirus type 8; IL-6, interleukin 6; VEC, vascular epithelial cell; vIL-6, viral IL-6.

Full size image

IL-6

The IL-6-like cytokines activate the glycoprotein 13 chain, which results in Janus kinase-mediated activation of the signal transducers and activators of transcription (STATs).43 Alternatively, IL-6 activates the mitogen-activated protein kinase (MAPK) cascade.44 IL-6 production has a triple role in Castleman's disease: it induces B-cell proliferation, mediates inflammatory clinical symptoms, and induces VEGF and angiogenesis.45,46 Germinal centers of hyperplastic lymph nodes produce large quantities of IL-6 and removal of such lymph nodes results in clinical improvement and a decrease in IL-6 levels.47 Subsequent dissection of the lymph node revealed increased IL-6 expression in interfollicular areas (other cytokines are absent), whereas elevated IL-6 gene expression inside follicles only occurs in the localized form of the disease associated with systemic manifestations.47 Administration of antibodies directed towards IL-648 or the IL-6 receptor49 alleviate systemic symptoms in patients with Castleman's disease. It is unclear whether the IL-6 excess predominantly derives from human or viral sources (IL-6 versus vIL-6) as expression can be independent of HIV status.30 However, it is known that vIL-6 activates angiogenesis and hematopoiesis pathways in a similar manner to human IL-6 (hIL-6),50 and increases expression of hIL-6.51 In contrast to human IL-6, the effects of vIL-6 are not dependent on, but are enhanced by, the gp80 path of the IL-6 receptor.52 Thus, vIL-6 is a 'pirate' cytokine with optimized and unregulated IL-6 functions, and is the key mediator for expansion of both infective compartments in Castleman's disease: plasma cell precursors and lymphovascular endothelial cells (Figure 3).

B-cell populations

B-cell immunity depends on specialized lymphoid compartments containing distinct B-cell populations.53 Maturation pathways of the extrafollicular T-zone generate plasma cell precursors with unmutated immunoglobulin genes. Activated B cells are primed in the germinal center to become memory cells or plasma cells with a mutated immunoglobulin gene.54 The mantle zone around the germinal center contains mostly naive B cells that secrete high levels of IgD and are prepared for mitosis. These cells are hypothesized to be the origin of mantle-cell lymphoma.55 By contrast, B cells of the splenic marginal zone are IgM-high, IgD-low and express CD27.56 In Castleman's disease, HHV8 virus particles are predominantly, but not solely, identified in the mantle zone of the lymph node.6 In lymphoid cells, vIL-6 is expressed around lymphoid follicles and within the follicular dendritic reticulum network. Detection of IL-6 in apoptotic bodies suggests that a complete lytic cycle is achieved in a subset of CD138− lymphoid cells.39 HHV8-positive plasmablasts express interferon-regulatory factor 4 (IRF4, also called MUM1) and B-lymphocyte-induced maturation protein 1 (BLIMP1). They are negative for paired box protein 5 (PAX5) and CD30, and positive for the proliferation marker Ki67.10 The expression of CD20 is not universal. In unicentric Castleman's disease HHV8 virus infection might be restricted to CD19+ cells, while it might involve the T-cell population in multicentric disease.57

The target cell for HHV8 infection is a plasma cell precursor or plasmablast at the stage of non-mutated immunoglobulin production, and infection seems to occur in the mantle zone of the lymph node or around vascular structures of the lymph nodes. Here, plasma cell precursors cluster around the vessels, possibly as a result of chemotaxis induced by IL-6 or vIL-6.

Lymphovascular epithelial cells and B cells are involved at entirely different levels of intensity, in terms of the lytic cycles of HHV8 infection in Castleman's disease. For the lymphovascular epithelial cell population, viral replication results in cell death and the slow formation of neovasculature corresponding to the moderate mitotic potential of these cells. By contrast, the vIL-6 driven expansion in the plasmablast population implies exponential growth of a cell population with high mitotic potential for which IL-6 is a natural key proliferogenic factor. The critical difference between the processes in these two cell populations is the mitotic potential of the two compartments.

Anatomy and virus-based pathogenetic pathway

HHV8 infection targets two cellular compartments: the vascular endothelial cells and plasma precursor cells (Figure 3). Replication rate in vascular endothelial cells is as low as one copy per cell, resulting in low vIL-6 production. Perivascular chemotaxis and proliferation of plasmablasts results in apoptosis or cell lysis of both cell compartments with formation of the radial scars of hyaline-vascular disease. This follows temporary lymph-node swelling and eventual low grade systemic symptoms. Crossinfection of significant numbers of perivascular plasmablasts induces massive viral replication, which produces high levels of vIL-6 with paracrine stimulation of plasmablast proliferation. This amplifies the local target-cell population followed by enhanced viral infection, lysis, and vIL-6 production. The outcome is an exponential escalation of IL-6 secretion into the systemic circulation causing a systemic inflammatory syndrome. The transition of latency into aggressive replication occurs stochastically, explaining the coexistence of different forms of Castleman's disease in one patient.27

Unicentric plasma-cell disease consists of a cluster of lymph nodes rather than a single lymph node.11 This suggests that the infectious agent, as well as IL-6, is transported to the target cells primarily via the locoregional lymphovascular drainage system. The systemic rise of IL-6 eventually causes multifocal expansion of the vulnerable plasmablast pool resulting in extraregional multicentric disease. In HIV-negative patients, unicentric and multicentric Castleman's disease is observed in the absence of HHV8 infection, but with different clinical, pathological and laboratory findings. In HIV-negative patients with multicentric Castleman's disease, HHV8 infection is detected in about 50% of cases, while in unicentric Castleman's disease HHV8 infection has been detected in only a minority of patients (K.-M. Schulte & N. Talat, unpublished data).

We propose an anatomical model of disease progression initiated by primary viral infection at any level, most frequently involving the upper respiratory tract. The lymphatic fluid leaving the infected tissue is filtered in the locoregional draining 'sentinel' lymph node, which develops a lymphovascular endothelial infection provoking a T-cell-mediated immune response. Clinically, this initial infection presents as unicentric disease associated with local symptoms and is essentially curable by surgical excision. Eventually, lymphovascular endothelial containment of the virus in this sentinel lymph node is lost, engaging plasmablasts in a lytic viral infection. This causes B-cell infection and lymphatic transport into the surrounding lymph node and eventually into the blood, generating multicentric disease. At this stage symptoms are systemic and prognosis is poor. Disease progression corresponds to the biphasic progression of solid organ cancer with initial lymphatic and secondary blood-borne propagation. The induction process in the absence of HHV8 infection remains elusive, but causation by an unidentified viral agent with lower pathogenicity seems likely. Such a pathogen might also cause associated pathologies including 'paraneoplastic pemphigoid', predominantly identified in cases of Castleman's disease without HHV8 infection.

Concomitant viral infections

In Castleman's disease, HHV8 infects monotypic, but not polyclonal, naive B cells and is associated with a range of lymphoproliferative disorders, including microlymphomas and plasmablastic lymphomas.58 These observations add a second layer of insight into infection with HHV8; transformation into B-cell lymphomas is observed in diseases ranging from primary effusion lymphomas to the plasma-cell type of Castleman's disease. The context appears to be defined by the co-infecting agent, such as HIV or EBV.12 In humans, LANA is critical to induce B-cell hyperplasia and lymphoma.59 HHV8 is specifically associated with a variant of multicentric Castleman's disease in which lambda light-chain restricted, HHV8-positive plasmablasts localize in B-cell follicle mantle zones and, in some cases, form microscopic lymphomas.23 Animal models have helped to elucidate mechanisms of viral complementation. Rhesus rhadinovirus 17577 is a close phylogenetic relative of HHV8. In macaques previously infected with simian immunodeficiency virus, infection with rhesus rhadinovirus 17577 induces a phenotypically similar disease to human multicentric Castleman's disease in HIV-positive patients.60,61,62 Both Kaposi sarcoma and multicentric Castleman's disease are rare in HIV-2 patients and a first case of the latter was only recently reported.63 Complementary effects of viral co-infection are hence specific to HHV-8-specific CD8+ T cells rather than related to defects in the T-cell repertoire.64 HHV8-negative Castleman's disease does not exist in HIV-positive patients, which indicates that strong selection pressures related to HIV factors promote HHV8 infection. A molecular candidate is the transactivating HIV protein Tat that is capable of suppressing immunity and apoptosis. Tat synergizes with angiogenic factors and upregulates IL-6 expression by activating the IL-6 promoter.65 Overall, viral infection defines three subtypes of Castleman's disease: absence of HHV8 and HIV infection with excellent prognosis (unicentric Castleman's disease), presence of HHV8 and HIV infection with poor prognosis (multicentric Castleman's disease), and an intermediate disease subtype associated with HHV8 infection only.

Essential clinical staging

The purpose of staging is to assign risk and prognosis, and to guide treatment decisions from initial presentation to long-term therapy. Practical staging of Castleman's disease needs to address three key clinical questions. Firstly, what is the extent of disease? Imaging by CT, MRI, or PET will clarify whether the disease is unicentric or multicentric. It will also reveal the eventual transition of the cells into solid forms of malignancy, such as lymphoma. Evaluation of the blood smear, differential analysis of blood cell counts, and an eventual bone marrow or effusion aspirate will demonstrate non-solid malignancy. Secondly, what is the histopathological type? Surgical biopsy of an accessible lymph node will classify the patient as having hyaline-vascular, plasma-cell or mixed type Castleman's disease. Finally, what is the viral etiology? Blood-based tests for HIV-1 antibody, HIV-2 antibody, and the viral load will clarify the HIV-status. Serology for EBV and PCR-based EBV detection may be useful in suspected primary effusion lymphoma. Immunostaining of lymphatic tissue for LANA, vIL-6, and eventually hIL-6 will clarify HHV8 association. Where available, information about HHV8 viral load may also help guide clinical decisions.

The choice of imaging

The main purpose of imaging in Castleman's disease is to facilitate disease staging, that is to clarify if the disease is unicentric or multicentric (a second vital purpose is to define surgical access and resectability). The best modality seems to be CT with analysis of the anatomical regions harboring the majority of the lesions, including the neck, chest, abdomen and pelvis.2,66,67 Mediastinal lymphadenopathy is frequently observed and easily identifiable on chest x-rays or CTs. No unique features allow safe differentiation from lymphoma, but the presence of hypervascularity is generally indicative of Castleman's disease.68,69 Calcification might be observed. The combination of mediastinal lesions with reticular or nodular interstitial lung changes, representing acute reticulonodular interstitial pneumonitis, might indicate Castleman's disease in HIV-positive patients.70 18F-fluorodeoxyglucose (FDG)-PET or PET-CT might identify multifocality.71,72

Clinical presentation

Clinical symptoms fall into four categories. First, local compression effects due to enlarged lymph nodes or extranodal Castleman's disease, with a mean lymph node size of 5.4 cm in unicentric disease. The mean size of the lymph nodes in multicentric disease is considerably smaller. Second, systemic B-cell-related symptoms due to cytokine activation with fever, weight loss, night sweats, weakness, and fatigue increase with tumor burden and histopathological type. Third, fluid retention in individuals with multicentric disease may be pronounced with edema, ascites, pleural or pericardial effusions, and seizures. Many such patients display multifocal lymphadenopathy and hepatosplenomegaly. Finally, symptoms of complications or associated diseases of Castleman's disease, such as lymphoma, paraneoplastic pemphigus, HIV, or polyneuropathy, organomegaly, endocrinopathy, monoclonal gammopathy and skin changes (POEMS) syndrome. Manifestations are nonspecific and the diagnosis is made by histopathology. No combination of symptoms safely differentiates Castleman's disease from a longer list of differential diagnoses.

As a hard and fast rule, unicentric hyaline-vascular Castleman's disease mostly causes local compression symptoms, while patients with the multicentric plasma-cell variant typically present with systemic symptoms. It is key, however, to carefully examine these patients in order to exclude diseases with a significantly overlapping presentation ranging from viral infection, to rheumatic or vasculitic disease.4 In this respect, a timely lymph-node biopsy is of the utmost importance.

Castleman's disease and HIV infection

HIV infection does not cause Castleman's disease. It is, however, a permissive factor and is associated with typical epidemiological and clinical features of Castleman's disease as well as a marked enhancement in disease progression.5,8 Oksenhendler10 has recently proposed the following clinical indicators to identify Castleman's disease in HIV-positive individuals: fever, diffuse lymphadenopathy, splenomegaly, severe cytopenia, high serum C-reactive protein, high HHV8 DNA levels in peripheral blood mononuclear cells, intense plasmocytosis in the lymph nodes or bone marrow, nasal obstruction, respiratory symptoms, Kaposi sarcoma lesions, prior identical episode with spontaneous resolution, positive direct agglutination test (Coombs test), and hemophagocytic syndrome. Patients with Castleman's disease may become systemically severely ill from unusual pulmonary complications, such as acute reticulonodular interstitial pneumonitis, and develop acute respiratory distress syndrome.73 Administration of therapy specific for Castleman's disease is generally sufficient to treat these complications.

Immune reconstitution inflammatory syndrome74 might explain the observation that some patients develop Castleman's disease within weeks of starting highly active antiretroviral therapy.75 Seemingly, a degree of immune competence is necessary to establish the signs and symptoms of Castleman's disease. This coincides with the observation that CD4 counts and HIV viral titers are not correlated or are inversely correlated with the development of Castleman's disease.5 A review5 of 25 studies including 84 HIV-positive cases coincides with clinical findings outlined by Oksehendler. Focal pathological changes involving Kaposi sarcoma were found in 72% of the patients and were in the respiratory tract of 34% of patients.

Associated conditions

Kaposi sarcoma

Kaposi sarcoma was identified in a minority of Castleman's disease cases without reported HIV-positive status and in about 50% of all reported HIV-positive patients. It is exclusively associated with a positive HHV8 status.23,76

Non-Hodgkin lymphoma

The high incidence of HHV8-associated non-Hodgkin lymphoma,77 supports the idea of clonal expansion following polyclonal activation.10 Non-Hodgkin lymphoma has been reported in <5% of Castleman's disease cases without reported HIV-positive status, and in <20% of those who are HIV positive. A prospective cohort study of 60 HIV-positive patients with Castleman's disease found a 15-fold increase in lymphomas in comparison with HIV-negative patients. These included HHV8-positive and EBV-positive primary-effusion lymphoma, HHV8-positive and EBV-negative visceral large-cell non-Hodgkin lymphoma, and plasmablastic lymphoma/leukemia.77 The latter is a large B-cell lymphoma, which is different from classic plasmablastic lymphoma. It originates from naive B cells with IgM expression at the stage preceding immunoglobulin hypermutation.78 HHV8 infection is frequently observed in HIV-related solid immunoblastic/plasmablastic diffuse large B-cell lymphoma.79 Ki76-staining might differentiate benign from malignant B-cell proliferation.80 Overall, the association of Castleman's disease with Hodgkin lymphoma remains elusive.77,82,83,84,85,86,87

Paraneoplastic pemphigus

Paraneoplastic pemphigus presents as debilitating chronic blistering of the skin and mucous membranes, including oral mucositis and cutaneous lichenoid lesions. It is associated with lymphoproliferative neoplasms. Autoantibodies against plakin are useful diagnostic markers.88

POEMS syndrome

POEMS syndrome is a plasma-cell dyscrasia with excess levels of VEGF.89 More than 95% of patients have one or more monoclonal lambda sclerotic plasmacytomas or bone marrow infiltration.89 Biopsy of an osteosclerotic lesion or a lymph node typical of Castleman's disease are considered essential to establish the diagnosis and may be guided by FDG-PET.90 HHV8 infection is strongly associated with POEMS syndrome,91 but is not essential for the establishment of this condition; there are at least seven reported cases of POEMS in the absence of HHV8 infection.91,92,93

Follicular dendritic cell tumors

Follicular dendritic cell tumors originate from the dendritic cell network of the lymph node21 and may be either polyclonal or clonal.94,95,96 Such tumors can become malignant and form dendritic cell sarcomas in either the lymph node or the skin.97,98 This cancer is not associated with HHV8 infection.99 However, it is of note that dendritic cell sarcomas appear in the context of paraneoplastic pemphigus,100 a condition that has only been reported in HHV8-negative patients. This observation supports the hypothesis that a hitherto unidentified virus is responsible for driving Castleman's disease and associated diseases in these patients.

Treatment modalities—outcomes

Once a diagnosis of Castleman's disease has been established, it is of vital importance to determine whether a patient has unicentric or multicentric disease. The former responds to local approaches, while the latter mostly requires systemic therapy. Prospective, randomized data are sparse and evidence has mainly been derived from small series; as such, the following paragraphs should be read with this important caveat in mind.

Unicentric disease

Unicentric Castleman's disease responds to local ablation therapy. Careful preoperative staging is mandatory to prevent the predictable failure of local therapy in systemic disease. Active surveillance is often considered as an option for patients with asymptomatic unicentric disease. Surgery, in terms of local resection, is the option of choice for unicentric disease when it is unlikely to inflict locoregional complications. Frequently involved and readily accessible sites include cervical, axillary, and inguinal lymph-node stations, which should be cleared by a systematic lymph-node dissection. The appropriate set-up is paramount: an experienced surgeon, free-resection margins at reasonable cost as opposed to circumferential R1-excision, systematic locoregional excision as opposed to excision of the targeted lymph node is needed to ensure proper histopathological staging. This might identify otherwise unrecognized transgression into systemic disease indicating a risk of relapse and the need for systemic treatment. Neck, mediastinal, abdominal, pelvic and retroperitoneal sites account for the majority of unicentric locations. At present, such lesions can be targeted with low morbidity using open or endoscopic techniques. Radiotherapy is preferable in cases when anatomy or patient-related features render safe or complete excision unlikely.101,102

Multicentric disease in HIV-negative patients

Multicentric disease is almost invariably symptomatic, and so therapy is indicated and is generally agreed to improve outcome. Experience in dealing with HIV-negative patients with Castleman's disease before the availability of specific chemotherapy and immunomodulatory agents shows that persistent or worsening symptoms and secondary complications are strongly associated with untreated disease.103 Therapy aims to cure the underlying infection such as HHV8, to modify IL-6-related cytokine escalation, to reduce the proliferation of B cells, to debulk tumor mass, or to compensate for the immunodeficiency resulting from the loss of lymph-node functions.

Antiviral therapy directed at non-HIV viruses

Antiviral therapy might improve outcomes in HHV8-positive patients. Potential agents include ganciclovir, valganciclovir, or foscarnet. The efficacy of valganciclovir to suppress HHV8 replication was documented in a prospective randomized trial.104 There is no evidence that EBV-directed therapy has any role in treating Castleman's disease.

Modulation of the cytokine network

The nonspecific, immune-suppressive effects of corticosteroids might reduce lymph node size, clinical symptoms, and inflammatory markers in HIV-negative patients with Castleman's disease. However, their short-lived effect and significant side effects preclude their use beyond a transient setting. The monoclonal antibody tocilizumab blocks the human IL-6 receptor and antagonizes the crucial role of IL-6 in disease propagation and manifestation. Accordingly, this agent also reduces the size of involved lymph nodes.105

Cytostatic therapy

Thalidomide can be used with considerable success in HIV-negative patients with Castleman's disease, although the multiple biological effects related to use of this agent are incompletely understood.106,107 In the context of treating Castleman's disease, the teratogenic side effects of thalidomide are of secondary concern. Chemotherapy with cyclophosphamide, doxorubicin, vincristine and prednisone (CHOP) is frequently employed in the treatment of systemic disease. Cyclophosphamide, vinblastine and etoposide can also be used as single agents.

Multicentric disease in HIV-positive patients

HIV therapy alters the spontaneous evolution of Castelman's disease. Treatment outcomes in HIV-positive patients with Castleman's disease have recently been reviewed.5,10 At present, multimodal therapy achieves similar outcomes in multicentric HIV-positive patients with Castleman's disease in comparison with HIV-negative patients. Single-agent therapy rapidly resolves acute clinical and laboratory findings. Etoposide given at a single intravenous dose (120–150 mg/m2) followed by weekly oral maintenance therapy (100–120 mg/m2) provides good disease control.108 However, bone marrow suppression and oncogenic effects make this regimen unsuitable for long-term use.

Rituximab, an anti-CD20 monoclonal antibody, achieves high response rates in HIV-positive patients with Castleman's disease. A phase II trial using rituximab in 21 patients achieved 2-year overall survival and disease-free survival rates of 95% (95% CI 86–100%) and 79% (95% CI 49–100%), respectively.109 Plasma acute-phase proteins, immunoglobulins, and HHV8 viral load decreased following initiation of therapy. The prospective, open-label, ANRS 117 CastlemaB Trial demonstrated sustained remission in 24 chemotherapy-dependent patients using rituximab.110 Disease was stable under chemotherapy, sustained remission was observed in more than 90% of patients at 2 months and the overall survival rate was also above 90%, and patients tolerated rituximab remarkably well. Moderate infections accounted for the majority of adverse effects. Both studies identified reactivation of formerly controlled Kaposi sarcoma.109,110 As a caveat to these findings, recent reports identified a total failure of rituximab therapy resulting in rapid death,111,112 indicating that it is not safe as single agent, specifically in aggressive disease.

The combination of immunotherapy and chemotherapy, termed immunochemotherapy, might enhance safety or efficacy. A case series using a mix of anthracycline-based chemotherapy and rituximab in HIV-positive patients with Castleman's disease has shown good response rates.113 Polychemotherapy with CHOP has induced long-term remission, but the target group needs to be more precisely defined.5 Vinblastine has induced life-threatening drug interactions with ongoing antiretroviral treatment.114

Potential new treatment pathways

Immunotherapy with anti-CD20 antibodies depletes mature B cells, but not pre-B cells, certain peripheral B-cell subpopulations, most antibody-producing cells, or B-cell malignancies. Mature B cells can also be targeted with a monoclonal antibody to CD19.115 Indeed, the CD19+ cell population can remain unaffected following treatment with monoclonal anti-CD20 antibody.116 In unicentric disease, the infected cell population may be restricted to CD19+ cells,57 encouraging the consideration of monoclonal anti-CD19 antibody therapy in systemic disease.

Treatment failure can also be interpreted in the context of the two-compartment model of cellular infection. Latent HHV8 infection of the lymphovascular endothelial compartment escapes the effects of a complete eradication of the B-cell compartment, and infection there might manifest as latent antigen expression only. However, the alternative pathway of very low-copy viral replication is a likely origin of recurrent disease. There is little experience in how to target a latent viral reservoir in endothelial cells. Long-term therapy with antiviral-based analogs might be an option. The complex and targeted antiangiogenic effects of thalidomide on the endothelial compartment have not yet been fully elucidated,116,117 but the drug seems to be a viable treatment option given its effectiveness in other hematological malignancies.117 The evidence supports administration of rituximab followed by thalidomide for maintenance, which has produced complete regression in some cases.118

Bortezomib also targets the endovascular compartment and reversibly inhibits the 26S proteasome, which normally functions to degrade ubiquitinated proteins. This leads to cell-cycle arrest, apoptosis, and inhibition of angiogenesis. A specific feature of this agent is the inhibition of nuclear factor (NF)-kappaB. Bortezomib has been shown to reduce proinflammatory cytokine levels and transfusion dependency in a patient with multicentric Castleman's disease.117,119

Future therapy might target both cellular compartments in a sequential manner. Anti-CD19 therapy would comprehensively destroy the lymphatic compartment of Castleman's disease (the proliferating plasmablast and B-cell precursor pool) without damaging the stem-cell population (the pro-B cell of the bone marrow). Thalidomide or bortezomib could then contain the latent HHV8 infection, which causes inflammatory and angiogenic disease propagation in the lymphovascular compartment.

Conclusions

Castleman's disease is a paradigm of chronic viral infection manifesting as solid lymphatic disease. It exemplifies the potential of rhadinoviruses, with their large genomes, to sustain chronic disease through skillful immune evasion and low-copy replication. Multiple swaps between latency and lytic cycle affecting two fundamentally different cell pools are ectopically enhanced by secretion of the viral cytokine vIL-6, which becomes the pacemaker not only of viral replication itself but of the ever growing target-cell pool for viral infection. Co-infection with HIV transforms local disease into systemic disease on the basis of effective molecular inter-viral cooperation in the presence of HIV infection. The sustained presence of HHV8 infection can lead to specific lymphatic conditions due to the transforming potential shared by human rhadinoviruses. While initial and local disease can be effectively treated by the ablation of the effective focus, systemic modalities are needed to control regional and extra regional spread. Multi modal approaches targeting both infected cellular reservoirs, that is B cells and endothelial cells, seem more promising than attempts to contain infection in either compartment alone. Chronic rhadinoviral infection is overwhelmingly common, and 50 years of research into Castleman's disease has provided essential contributions to the understanding and therapy of related human disease.

Review criteria

We searched the NCBI PubMed database in September 2009 using the search term “Castleman's disease” without further constraints. This identified 1,791 articles in English, French, Spanish, and German. We read the abstracts of all articles and studied the full text of all cited references. These were also searched for further leads using the Thompson ISI Web of Science software.