Progressive multifocal leukoencephalopathy and promyelocytic leukemia nuclear bodies: a review of clinical, neuropathological, and virological aspects of JC virus-induced demyelinating disease

Progressive multifocal leukoencephalopathy is a distinct neurologic entity established in 1958 [5]. Two cases of chronic lymphatic leukemia and one case of Hodgkin’s lymphoma were described as having similar clinical outcomes and sharing a characteristic neuropathology among postmortem brains. Uniquely, a mixture of small and larger demyelinating lesions was disseminated in the white matter and to a lesser degree, also in the gray matter. The small demyelinating lesions appeared to merge to form larger ones, an arrangement that was distinct from other demyelinating diseases known at that time. The fusion of multifocal demyelinating lesions was regarded as a progressive process and thus, the disease was named progressive multifocal leukoencephalopathy (Fig. 1a).

The diagnostic hallmark of progressive multifocal leukoencephalopathy is the presence of oligodendrocytes with enlarged nuclei containing intranuclear amphophilic inclusions. The presence of abnormally bizarre astrocytes has been also described. A link to viral infection was suggested [19] and two independent groups subsequently revealed the presence of polyoma-like virions in round or filamentous forms by electron microscopy [121, 150]. Isolation of the virus was difficult but primary human fetal glia cells were eventually identified as the first cell culture system that could support viral propagation. Human polyomavirus isolated from autopsied brain tissue was named JC virus after the patient, Mr. John Cunningham [98]. Progressive multifocal leukoencephalopathy was rare in those days, and regarded as a “slow virus infection” of the central nervous system, similar to scrapie, Kuru, and Creutzfeldt-Jakob disease.

Progressive multifocal leukoencephalopathy develops in patients with impaired immunity. The number of reported cases of this previously rare disease has significantly increased since the 1980s due to the pandemic of acquired immunodeficiency syndrome (AIDS). In the south of Florida in the United States, for example, there was a 12-fold increase in the frequency of diagnosis in 1990–1994 when compared with 1980–1984 [8]. Introduction of highly active antiretroviral therapy (HAART), initiated in France in 1996, has significantly decreased the morbidity and mortality of AIDS [56]; however, progressive multifocal leukoencephalopathy remains a major AIDS complication resulting in death [52]. Paradoxically, some patients develop a deterioration associated with HAART treatment known as immune reconstitution inflammatory syndrome (IRIS). IRIS can result from reconstitution of the immune system to recognize pathogens/antigens, and can be accompanied by toxoplasmosis, cryptococcosis, and tuberculosis, as well as progressive multifocal leukoencephalopathy [21, 51].

In non-AIDS patients, advancements in therapy of underlying diseases and the increasing use of immunomodulatory drugs have markedly changed the clinical background. Among cases associated with lymphoproliferative disorders, Hodgkin’s disease was the most frequently reported underlying disease from 1958 to 1989; in contrast, from 1990 to 2004, more reports linked the disease to B cell chronic lymphocytic leukemia and non-Hodgkin’s lymphomas [45]. This may be in part because for the latter cases, an increasing number of patients undergo high-dose chemotherapy with hematopoietic stem cell transplantation [48, 67] or treatment with purine analogs such as fludarabine widely used since 1997 [45]. Transplant recipients of solid organs, including kidneys, livers, hearts, or lungs, can also develop demyelination, likely due to immunosuppressive treatments used to prevent graft rejection [120]. The disease can also develop in persons with rheumatic diseases related or unrelated to immunosuppressive drug treatment; indeed, systemic lupus erythematosus is one of the most common underlying diseases [15, 17, 87, 95]. In 2005, development of progressive multifocal leukoencephalopathy was reported with the use of natalizumab, a monoclonal antibody that blocks inflammatory cell entry into the brain and is used for multiple sclerosis and Crohn’s disease [70, 79, 135, 148]. Children with congenital immunodeficiency disorders [2] and the elderly (in association with age-related immunosuppression) are also at risk. Thus, progressive multifocal leukoencephalopathy is now relatively common and can develop in association with a large number of different diseases, treatments, and conditions that lead to immunodeficiency.

A diagnosis of progressive multifocal leukoencephalopathy can be indicated by radiological imaging, but detection of JC virus is essential for a definitive diagnosis. JC viral DNA can be detected in the cerebrospinal fluid by polymerase chain reaction (PCR), but negative results do not completely rule out the possibility. For example, PCR results may be negative for HAART-treated AIDS patients, due to clearance of the viral DNA associated with improved immune responses [16, 46]. Therefore, pathological examination of a biopsy or autopsy sample may be required for a final diagnosis. The clinical prognosis for progressive multifocal leukoencephalopathy is usually poor, with most cases resulting in death within 6 months, although prolonged survival has sometimes been observed [7, 8].

Neuropathology of progressive multifocal leukoencephalopathy now shows histological diversity, likely associated with the expanded clinical spectrum of underlying diseases. As described earlier, the originally identified neuropathology was a typical arrangement of multiple well-demarcated demyelinating lesions, preferentially seen in the cerebral subcortical white matter [5] (Fig. 1a). However, for some recent cases, apparent cortical demyelination has been reported [11, 86, 127, 140]. The frontal lobes and parieto-occipital regions are most commonly affected, but demyelination may be also observed in deep gray structures, the cerebellum, the brainstem, and even in the spinal cord [9, 61, 62, 115].

The degree of histologically detected demyelination and of inflammation is also variable. Some cases display myelin pallor (Fig. 1b), whereas others show marked degeneration associated with axonal damage and the appearance of numerous foamy macrophages harboring phagocytolytic lipids or myelin debris. Such lesions are regarded as “burnt out,” and are often observed in cases associated with AIDS [52, 76]. A broad distribution of burnt out lesions may imply prolonged survival concurrent with advanced treatment [52]. The inflammatory response is usually poor; however, in cases of IRIS, severe inflammation may be present, including marked intraparenchymal and perivascular infiltration by macrophages and T lymphocytes [51, 85, 136].

The etiology of progressive multifocal leukoencephalopathy is JC virus lytic infection leading to the destruction of myelin sheaths. Therefore, detection of glial cells harboring JC virus inclusions is essential for pathological diagnosis. The characteristic cells are usually seen at the periphery of the demyelinating lesions, if the lesions are well demarcated (Fig. 1c). In contrast, if myelin pallor is broadly dispersed, the affected cells also tend to be scattered in a relatively broad area (Fig. 1d). The viral inclusions appear as amphophilic material dispersed throughout the nucleoplasm (full inclusions), by hematoxylin–eosin (HE) staining. However, the presence of numerous cells is also noted, which have a clearer nucleoplasm and marginated chromatin but lacking amphophilic material, despite nuclear enlargement (Fig. 1e–h). By immunohistochemistry, some of these cells display a distinct punctate distribution of promyelocytic leukemia (PML) protein, indicating the appearance of promyelocytic leukemia nuclear bodies (PML-NBs; note that for clarity, “PML” refers herein only to promyelocytic leukemia or PML protein, and not to the disease progressive multifocal leukoencephalopathy) (Fig. 1i). Both the JC viral DNA genome and viral capsid proteins show punctate signals by in situ hybridization and immunohistochemistry, respectively (dot-shaped inclusions) (Fig. 1j, k). Because PML protein and JC viral capsid proteins co-localize [117, 118] (Fig. 7), PML-NBs are likely the sites of JC viral replication. Dot-shaped inclusions are indicative of an earlier stage of viral infection, prior to the stage at which full inclusions would be observed (Fig. 1j–l).

By electron microscopy, JC virions are round or filamentous structures, approximately 35–40 nm in size. Although the intranuclear distributions of viruses are variable, many cells contain more viruses at the periphery than in center of the nucleus. Occasionally, round virions display a perfect alignment pattern, called crystalloid arrays (Fig. 2). This alignment pattern indicates that viral spread is associated with specific intranuclear structures, rather than being the result of viruses freely floating in the nucleoplasm. In 1980, Mazlo and Tariska [84] investigated the first phase of viral replication using electron microscopy and reported that viral progeny always appear in the vicinity of the inner nuclear membrane after nuclear enlargement of infected oligodendrocytes. Today, we better understand the development of intranuclear viral inclusions at a molecular level; nevertheless, even 30 years later these new findings are consistent with earlier results based on extensive ultrastructural studies.

JC virus has a circular double-stranded DNA genome. The complete sequence was first determined for the Mad-1 strain (5,130 bp) [41]. The first Japanese strain, Tokyo-1 (5,128 bp) was isolated [90], and we determined its complete sequence [116]. JC virus shares approximately 70% DNA sequence similarity with SV40 and BK virus, and displays a similar tripartite genome organization with early and late coding regions and a regulatory region. The early coding region encodes multifunctional regulatory proteins, large T antigen and its splicing variants (t, T′₁₃₅, T′₁₃₆, and T′₁₆₅) [134]. The late region encodes three capsid proteins, VP1, VP2, and VP3, and a small regulatory protein called agnoprotein.

The JC viral regulatory sequence is located between the early and late coding regions, and the promoter-enhancer elements function bidirectionally. The regulatory sequence displays marked diversity, but the sequences can be roughly divided into two groups, “archetype” and “neurotropic type”. Viruses isolated from the urine of healthy individuals have consistently been found to contain a characteristic regulatory sequence known as the archetype, which is composed of five segments: 25-, 23-, 55-, 66-, 18-bp sequences (Fig. 3). In contrast, viruses isolated from demyelinating brains display divergent sequences designated as the “neurotropic type” (also referred to as the “PML-type” where “PML” means progressive multifocal leukoencephalopathy). The neurotropic type regulatory sequences appear to have arisen from the archetype via deletion and duplication. For example, in the Mad-1 sequence, the 23- and 66-bp segments are deleted, and the 25-, 55-, and 18-bp segments are duplicated, resulting in 98-bp tandem repeats. In the Tokyo-1 sequence, the 66-bp portion is largely deleted, most 55- and 18-bp segments are duplicated, with a unique 17-bp sequence insertion [83]. Sequence rearrangement from the archetype to the neurotropic type may increase restricted tissue tropism or virulence, thus affecting the overall of pathogenic potential of the virus [22]. However, it is also argued that sequence rearrangement does not change the JC virus biology [124]. The so-called “archetype hypothesis” will be described further below.

Early interest focused on the idea of “slow virus infection” and the question was frequently raised, how does JC virus persist asymptomatically but in some cases, cause opportunistic demyelinating disease? JC virus is ubiquitous in human populations and approximately 50–80% of adults present antibodies against the virus [72, 97]. Although seroprevalence continues to increase until the seventh decade of life, most individuals are thought to already have asymptomatic infections by early childhood. Primary infections may occur via respiratory inhalation [109, 113] or through consumption of contaminated food or beverages [12‐14]. The viruses are presumably latent or persistent in lymphoid organs such as the spleen or tonsils, as well as bone marrow, and circulate in association with peripheral lymphocytes, monocytes, or cell-free plasma [32, 33, 43, 57, 88, 133]. A small amount of virus seems to replicate in the kidney and the virus is occasionally excreted in the urine. Circulating viruses may reach the brain and if host immunity is compromised, the virus can cause progressive multifocal leukoencephalopathy.

In 1990, in what is arguably one of the most important and interesting studies in the field, the presence of the so-called archetypal viruses was advocated [147]. Archetypal viruses are frequently excreted in the urine of healthy individuals [69, 146, 147] and detected in urban sewage [14]. Thus, it seems to be the archetype that circulates in the human population. It has been proposed that subsequent to a primary infection, the archetype remains quiescent in the kidneys or other organs with occasional excretion to the urine. Some time later, the virus may transform to the neurotropic type via sequence rearrangement (in particular, deletion and duplication; see above). Neurotropic type viruses acquire neurovirulence and thus can cause progressive multifocal leukoencephalopathy (Fig. 4a).

Subsequent to the promotion of the archetype idea, how and where the viruses transform from the archetype to the neurotropic type became a subject of debate. If the virus is transmitted via respiratory inhalation, then it would stand to reason that the archetype would be detected in tonsilar tissues, an initial site of infection. In one report, however, viral sequences from such samples were identified as mostly of the neurotropic type [88], although archetypal sequences were detected in another study [63]. JC virus may be transmitted by consumption of contaminated food or beverages [12‐14] and consistent with this, JC viral sequences are frequently detectable in the human gastrointestinal (GI) tract [77, 106, 107]. However, according to one study, viral sequences detected in the GI tract were identified as belonging to the Mad-1-like neurotropic type [77]. JC virus may also circulate in the body in association with peripheral blood lymphocytes (PBLs), where sequence arrangement may occur. One study reports that the viruses in PBLs were mostly of the neurotropic type [31], and the presence of both archetype and neurotropic types was reported in another study [6]. Bone marrow has also been suggested as the site of transformation, as neurotropic type virus has been detected there as well [130]. Making the story yet more confusing, the presence of JC viral DNA has repeatedly been detected in postmortem brain tissue taken from patients who died from non-neurologic diseases [29, 36, 99, 141]. Unexpectedly, these viruses contain the rearranged, neurotropic type sequences and thus, the possibility arises that the brain is also a site of latency [29, 99, 141]. Repeated detection of neurotropic viruses in multiple organs of carrier individuals stimulated arguments against the archetype hypothesis. Importantly, pediatric patients with congenital immunosuppression display progressive multifocal leukoencephalopathy in which both archetype and neurotropic type viruses are present in different tissues. These data have been interpreted to mean that in some patients, demyelinating disease may develop after a primary infection rather than requiring a latency period [93, 94]. The archetype hypothesis has been widely accepted and indeed, the disease may occur by transformation from the archetype to neurotropic type. However, at least one alternative possibility remains, i.e. that neurotropic type viruses also circulate in normal human populations and moreover, may cause demyelinating disease upon primary infection (Fig. 4b).

Today, we know that JC virus infection is not specific to oligodendrocytes, as viral DNA or proteins have been repeatedly detected in multiple tissues. However, the disease-state outcome of infection, progressive multifocal leukoencephalopathy, selectively occurs in human oligodendroglias. Why JC virus establishes its most efficient lytic infection in this specific cell type and not others remains largely an open question.

The life cycle of polyomaviruses is comprised of early and late events. During early events, viral infection is initiated by attachment to a cell surface receptor, followed by trafficking from the cell membrane toward the nucleus. Next, T antigens are synthesized, and they assist replication of viral genomic DNAs in the nucleus. During late events, the viral capsid proteins VP1, VP2, and VP3 are synthesized in the cytoplasm, and they localize to the nucleus for production of viral progeny. Finally, viral progeny are released from the cells, at which time they can go onto infect additional host cells (Fig. 5a). Cellular susceptibility for lytic infection can be determined at any stage in the lytic cycle.

JC virus can bind to receptors present on a variety of cell types, but progeny production does not necessarily occur in all cell types. Viral entry is mediated by a ubiquitous receptor, an N-linked glycoprotein with α(2–6)-linked sialic acids, which associates with the serotonergic receptor 5HT2A on human glial cells [35, 81]. However, even when the virus attaches to cell surface receptors, or penetrates and reaches the nucleus, these events may not always lead to completion of the viral lytic cycle. The results of in vitro experiments demonstrated that JC virus entry into nuclei was observed for 15 different types of cell lines, but among them, virus replication occurred exclusively in human neuroblastoma cells [126]. The data indicate that the presence of cell surface receptors is essential but does not specifically determine cellular susceptibility for efficient lytic infection [34, 139] (Fig. 5b).

The 98-bp tandem repeats in the regulatory sequence of Mad-1 were found to be a transcriptional regulatory element with enhancer-like activity in human glial cells [65]. Since then, the results of many studies have supported the idea that restricted lytic infection by JC virus in human brain tissue depends on the presence of nuclear factors that are predominantly expressed in glial cells. The rearranged sequence of the regulatory regions markedly affects the levels of viral transcription and replication, as they encode binding sites for nuclear proteins that regulate viral DNA replication or transcriptional activity. The sequences include the TATA box, and binding sites for nuclear transactivators, including Sp1, YB1, sup2, NF-κB, Purα, Tst-1, c-Jun, and NF-1 (reviewed in [102]). Thus, efficient lytic infection in specific cells seems to be dependent on multiple nuclear factors that selectively bind to the rearranged regulatory sequence.

Interestingly, recent studies have revealed that JC virus also infects neurons in cerebellar granule layers and even pyramidal neurons in the cerebral cortex. In the cerebellum, lytic infection of granule cells was proposed as JCV granule cell neuronopathy, and in the cerebral cortex, fulminant JC virus encephalopathy [73, 143, 144]. These data suggest that efficient lytic infection proceeds even in neuronal cells, a process less well understood than lytic infection in glial cells.

Since establishment of a eukaryotic expression system to generate virus-like particles (VLPs) [119], we have been studying the molecular mechanisms underlying how JC virus develops intranuclear inclusions. The JC virus capsid would be composed of VP1, VP2, and VP3, likely in a specific ratio as in SV40 [80]. Proportional expression of the capsid proteins is in part regulated by alternative splicing, and at least four species of late RNAs are produced (M1–M4) [116]. Of these, it is M1 that encodes VP2/VP3 and M2 that encodes VP1, which is downstream of agnoprotein (Fig. 6) (note that because the complete amino acid sequence of VP3 is identical to the C-terminal two-thirds of VP2, we refer to them collectively as VP2/VP3).

Nuclear proteins typically contain nuclear localization signals (NLSs), most commonly a stretch of basic amino acids such as K (lysine) and R (arginine). JC virus VP1 is inefficient in nuclear transport, which is attributable to what is presumably a weak NLS in its N-terminal sequence, MAPTKRKGERKD. This sequence has the properties of an NLS; namely, basic amino acids (K ⁵ R ⁶ K ⁷ and R ¹⁰ K ¹¹ ) flank the intervening residues GE (superscripts indicate the positions of residues in an amino acid sequence). This can be compared with SV40 VP1, which includes the sequence MAPTKRKGSCPGAAPKKPKE (including K ⁵ R ⁶ K ⁷ and K ¹⁶ K ¹⁷ , K ¹⁹ ), which functions as an efficient, bipartite NLS [58]. It stands to reason, then, that with its functionally weak NLS, JC virus VP1 is distributed in both the cytoplasm and the nucleus, whereas with its efficient NLS, SV40 VP1 is exclusively localized to the nucleus. Exclusive nuclear localization of JC virus VP1 can be achieved upon co-expression with VP2/VP3, which contains two clusters of basic amino acids in the C-terminal region, GPNKKKRRK (cluster 1) and KRRSRSSRS (cluster 2). These sequences compensate for the inefficient nuclear targeting of VP1.

Interestingly, co-transported capsid proteins were found to accumulate at dot-shaped subnuclear domains preferentially detected at the nuclear periphery; that is, at PML-NBs. Confocal microscopic analysis has clearly demonstrated co-localization of VP1 with PML protein, a major component of PML-NBs. Co-localization in a dot-shaped distribution pattern was observed both in human brain tissues with progressive multifocal leukoencephalopathy and in COS-7 cells producing VLPs. In the nuclei, major and minor capsid proteins were found to stretch their intranuclear distributions along nuclear matrix-like structures (Fig. 7a, b).

By electron microscopy, along with viral gene manipulation in COS-7 cells, we have revealed that in the presence of VP2/VP3, colloidal gold particles representing the VP1 epitopes were clustered along the inner nuclear periphery, where numerous VLPs with both round and filamentous structures were also observed. The data are consistent with what has been seen for JC virus-infected glial cells in human brain (Fig. 8). Altogether, it appears that major and minor capsid proteins move cooperatively from the cytoplasm to the nucleus and then accumulate in PML-NBs, where they are assembled into progeny virions (Fig. 9).

The mechanisms of JC virus-mediated cell death remain largely unclear. JC virus-infected cells may undergo apoptosis [108, 145]. However, this is in contrast with another report, in which the authors suggest that JC virus-mediated cell death is a non-apoptotic, necrotic mode of death in progenitor-derived astrocyte cell culture [114]. A recent study also demonstrated higher expression of the anti-apoptotic protein survivin, and the authors claimed that viral-infected cells are anti-apoptotic, allowing for completion of the viral lytic cycle [100].

During development of full viral inclusions, PML-NBs may go on to disrupted [117, 118], and our current thinking is that disruption or even dysfunction of PML-NBs may induce JC virus-mediated cell death. Indeed, disruption or dysfunction of PML-NBs is associated with a variety of human diseases, including cancers, neurodegenerative disorders, and viral infections, as described below.