The neuroradiology of progressive multifocal leukoencephalopathy: a clinical trial perspective

Abstract

Progressive multifocal leukoencephalopathy (PML) is an opportunistic infection of the CNS caused by the JC virus, which infects white and grey matter cells and leads to irreversible demyelination and neuroaxonal damage. Brain MRI, in addition to the clinical presentation and demonstration of JC virus DNA either in the CSF or by histopathology, is an important tool in the detection of PML. In clinical practice, standard MRI pulse sequences are utilized for screening, diagnosis and monitoring of PML, but validated imaging-based outcome measures for use in prospective, interventional clinical trials for PML have yet to be established. We review the existing literature regarding the use of MRI and PET in PML and discuss the implications of PML histopathology for neuroradiology. MRI not only demonstrates the localization and extent of PML lesions, but also mirrors the tissue destruction, ongoing viral spread, and resulting inflammation. Finally, we explore the potential for imaging measures to serve as an outcome in PML clinical trials and provide recommendations for current and future imaging outcome measure development in this area.

Progressive multifocal leukoencephalopathy is a serious opportunistic infection of the central nervous system, for which no approved treatment exists. Baldassari, Wattjes et al. review current knowledge regarding the use of brain imaging in PML, and discuss the potential applications of imaging in PML clinical trials.

Introduction

Progressive multifocal leukoencephalopathy (PML) is an opportunistic infection caused by JC virus (JCV), involving an infection of white matter (principally oligodendrocytes) and neurons in the brain, and leading to irreversible demyelination and neuroaxonal damage.¹ The immunosuppressive conditions associated with PML are broad and heterogeneous, ranging from HIV-AIDS [relatively common before highly active antiretroviral therapy (HAART) was available] to disease-specific immunosuppressive therapies including monoclonal antibodies.^2-6 PML diagnostic criteria are based on clinical presentation, brain MRI findings, and demonstration of JCV DNA, either via polymerase chain reaction (PCR) in CSF or by histopathology.⁷ The initial description of PML clinical and imaging findings was mainly based on AIDS-related PML cases, as the rise of AIDS coincided with the initial widespread application of clinical MRI. Recently, available data from the various underlying immunosuppressive conditions associated with PML have broadened the spectrum of clinical and imaging findings, particularly in the setting of monoclonal antibody therapies.^6,8–10 In addition, the spectrum of JCV tropism has been expanded to include additional JCV-related disease entities partially overlapping with PML pathology and imaging findings, such as granule cell neuronopathy (GCN) and JCV encephalopathy.^11–14

Brain MRI is probably the most sensitive diagnostic tool for PML detection, as it is capable of detecting lesions even in the absence of suggestive clinical symptoms (‘asymptomatic PML’) and/or detectable JCV DNA in the CSF.^15–17 The diagnosis of PML by brain MRI is based on a multisequence protocol, in which each sequence can show specific aspects of PML pathology. In addition, brain MRI is routinely used to monitor PML lesions, to detect early imaging findings of immune reconstitution and assess the effect of treatment.^18,19 Indeed, imaging-based PML monitoring is increasingly important during treatment, as promising therapeutic strategies, including checkpoint inhibitors and allogeneic T-cell therapies, are now being reported.^18,20–23

Despite the persistent risk of PML among patients with immunosuppressive conditions and a growing risk of PML among patients treated with immunomodulatory agents, no therapeutics are approved for the treatment of PML. Recognizing this unmet need and the absence of an established paradigm for product development for this rare and complex disease, US Food and Drug Administration reviewers and PML experts at National Institutes of Health/National Institute of Neurological Disorders and Stroke initiated a collaboration in 2019 with the goal of stimulating product development for PML. This collaborative effort sought to address scientific and regulatory challenges associated with PML clinical trials and to develop PML clinical trial designs that will be acceptable to PML patients, clinicians, regulators, and industry.

This review summarizes published literature on the utility of commonly applied MRI techniques for the evaluation of PML at all stages, ranging from early diagnosis to clinical monitoring to long-term survival, and considers approaches that could be useful in the context of future clinical trials, especially with respect to the suitability of imaging measures as clinical trial end points for PML. The review also includes a section on PET imaging of PML, as well as an introductory section on the histopathology of PML that provides context for the imaging-focused summaries and recommendations that follow.

Materials and methods

The initial phase of this PML clinical trial design project included establishing four working groups tasked with assessing available data and identifying knowledge gaps in four key areas: JCV biomarkers, brain imaging, patient-focused drug development, and clinical outcomes. The first phase of these collaborative efforts yielded several distinct bodies of work that represent a unique, PML clinical trial-focused perspective that is currently absent from the published literature. The PML Brain Imaging Working Group was established as part of this collaboration and was tasked with reviewing existing data to determine the suitability of imaging measures as clinical trial end points for PML. JCV biomarkers, patient-focused drug development, and clinical outcomes were addressed by other working groups.

Data availability

Data sharing is not applicable to this article as no new data were created or analysed in this study.

Results

Histopathology of PML and its implications for imaging

PML, in general, is caused by immunosuppressive conditions that provoke reactivation of a latent JCV infection. This reactivation leads to a spread of virus within the CNS with multiple demyelinating lesions and subsequent severe tissue destruction. Although the term ‘leukoencephalopathy’ indicates primary involvement of white matter, specifically oligodendrocytes and astrocytes, JCV also infects neurons in the cortical grey matter and in the cerebellum (granule cells).²⁴ Endarterioles terminate at the grey-white junction, and small lesions are often found at this location (Fig. 1A).²⁵ Involvement of superficial (‘subpial’) cortical grey matter, which is generally observed in multiple sclerosis, is not typical in PML.^25,26 For unknown reasons, the optic nerve and spinal cord are generally not involved in PML, though rare cases have been reported.^2,27

Figure 1.

PML pathology. Two demyelinated PML lesions located at the grey-white matter junction are indicated by arrows [initial lesions shown in A, Luxol fast blue/periodic acid Schiff (LFB/PAS) staining]. Lesions may enlarge either by extension (not shown) or expansion due to viral progression at the lesion edge [B; LFB/PAS (blue/pink) plus anti-VP1 (viral protein 1) (brown)]. Numerous virally infected cells are found at the lesion edge, as shown in the inset. Finally, the lesions fuse to form large demyelinated areas (pale regions marked with asterisk in B) and band-like demyelination at the border between the cortex and white matter (C; LFB/PAS plus anti-VP1). Typical histological features include virally infected glial cells with ground glass oligodendrocytes (D; haematoxylin and eosin staining; ground glass oligodendrocytes are indicated by arrows) and bizarre, pleomorphic astrocytes [E; anti-glial fibrillary acidic protein (GFAP); astrocytes are indicated by arrows]. Lesions may be destructive, leading to a pronounced axonal damage in the lesion centre (F; Bielschowsky silver stain) and even necrotic lesions (G; haematoxylin and eosin). The inflammatory infiltrate within PML lesions consists of numerous macrophages (H; KiM1P staining) and typically few lymphocytes (I; anti-CD3 staining), due to immunosuppression. In contrast, numerous lymphocytes are found with immune reconstitution (J; anti-CD3). JCV infection may also cause granule cell neuronopathy (K; Pab 2003 staining for large T-antigen). Scale bars = 1 mm in A–C; 50 μm in D and E; 20 μm in G and H; 100 μm in F and I–K. Counterstain in E and H–K: haematoxylin. Ctx = cortex; NAWM = normal-appearing white matter; WM = white matter.

Once a lesion has been initiated, its evolution is thought to occur by direct extension and expansion, which may result in fusion with other lesions.²⁸ Extension frequently follows white matter tracts, leading to small and discontinuous lesions along those tracts; whether this occurs by viral spread through the myelin sheath itself is unknown. Lesions may also expand locally via viral propagation at the lesion edge (Fig. 1B). In the final stage, enlarging lesions fuse to form large, confluent demyelinated areas (Fig. 1B).²⁸ Band-like demyelination may be observed at the grey-white junction (Fig. 1C). De novo lesion formation in brain areas remote from the original lesion may also occur.

Within areas of active infection, a prominent feature is the presence of ‘ground-glass’ oligodendrocytes, which are virus-infected oligodendrocytes with ongoing intranuclear viral replication and accumulation (Fig. 1D).²⁹ Affected oligodendrocytes eventually succumb, and since oligodendrocytes are myelinating cells, oligodendrocyte loss leads to demyelination (Fig. 1A and B). Ground-glass oligodendrocytes predominate at the lesion edge, whereas in the lesion core, which harbours a later stage of tissue damage, oligodendrocytes and oligodendrocyte progenitor cells may be completely absent; pronounced and irreversible axonal damage is seen instead (Fig. 1F). End-stage lesions may appear necrotic with severe and irreversible tissue destruction (Fig. 1G).³⁰

Infected astrocytes show pleomorphic changes, including striking enlargement and bizarre nuclear transformation, resembling neoplastic cells²⁹; but they are less likely to die than oligodendrocytes (Fig. 1E).

PML lesions are also infiltrated by numerous macrophages that phagocytose myelin debris (Fig. 1H). In immunosuppressed patients, very few T cells, B cells, and plasma cells are typically observed (Fig. 1I), but upon immune reconstitution, a dense inflammatory lymphoid infiltrate may ensue (Fig. 1J).³¹ This is frequently the case, for example, in natalizumab-associated PML following discontinuation of natalizumab therapy.³²

In general, histopathological findings in PML are comparable in different diseases associated with PML. However, lesions may differ in the amount of inflammation associated with the infection as well as in the viral load. Thus, fewer inflammatory cells and more virally infected cells are expected in severely immunosuppressed patients (e.g. HIV-AIDS patients not receiving HAART) in comparison to patients with an immune system that is less compromised (e.g. natalizumab-treated patients even prior to treatment discontinuation).

Not surprisingly, microscopic pathological alterations occur before MRI changes are observed, which in turn may precede clinical symptoms.⁴ The earliest pathological lesions observed by histology are small lesions with infected glial cells and microglial/macrophage activation, followed by oligodendrocyte loss and demyelination. Asymptomatic PML cases may have small lesions featuring glial changes with or without demyelination, which are therefore considered the earliest pathological characteristics.²⁵ Interestingly, asymptomatic PML may also present with fully developed PML lesions indistinguishable from those found in a symptomatic PML patient.

In the setting of immune system reconstitution, e.g. through combination antiretroviral therapy in HIV-AIDS or after withdrawal of immunosuppressive treatments, a massive influx of inflammatory cells may ensue, leading to clinical worsening and the so-called immune reconstitution inflammatory syndrome (IRIS). The prominent lymphocytic infiltrate consists mainly of CD8+ cytotoxic T cells and plasma cells, which can effectively eliminate the virus.³² During IRIS, the blood–brain barrier demonstrates inflammation-related changes in permeability.

Long-term survivors who mount a sufficient immune response may clear the virus and often harbour severely damaged and gliotic white matter as sequelae, with relative preservation of the cortical ribbon.²⁴ In patients who are unable to fight the virus, infection continues with lesion enlargement and development of new lesions, finally leading to a widespread lesion distribution and pronounced tissue destruction.

Finally, in addition to PML, the spectrum of JCV tropism can rarely involve the meninges (JCV meningitis), pyramidal cells (JCV encephalopathy), or cerebellar granule cells (Fig. 1K).^11–14,33 GCN may occur with or without concomitant demyelination of the cerebellum or other areas of the brain. Infection leads to a loss of granule cell neurons and cerebellar atrophy.^11,33

Multisequence brain MRI for PML diagnosis and monitoring

Brain MRI for PML diagnosis and monitoring is performed as a multisequence image acquisition protocol, taking advantage of the fact that each sequence has value regarding the detection of certain aspects of PML lesions. The optimal MRI protocol depends on the clinical setting. For PML screening (e.g. pharmacovigilance for patients on monoclonal antibody therapy), an abbreviated protocol without contrast-enhanced T₁-weighted sequences, but including fluid-attenuated inversion recovery (T₂-FLAIR), T₂-weighted, and diffusion-weighted imaging (DWI), has been recommended.³⁴ In particular for PML monitoring purposes, contrast-enhanced T₁-weighted imaging is mandatory to detect PML-IRIS. Representative images demonstrating PML findings on the sequences discussed below are presented in Figs 2 and 3. Additionally, Fig. 2 demonstrates histopathological correlations of key MRI findings in PML. The imaging pattern of PML described in the following sections may vary depending on the type of immunosuppression leading to PML, particularly when comparing HIV-associated PML and monoclonal antibody-associated PML.³⁵ However, this observation needs further validation by larger comparative studies.

Figure 2.

Histopathological correlations of MRI findings in classic and inflammatory PML. Brain MRI and corresponding histopathology (from different cases) of classic PML (top row) and inflammatory PML (bottom row). Histopathology for both classic and inflammatory PML is characterized by positive SV40 T antigen staining, indicating JCV infection. However, inflammatory PML lesions typically show lower numbers of infected cells. In classic PML, few CD8+ T cells and CD138+ plasma cells are present, whereas in inflammatory PML, there is pronounced infiltration of CD8+ T cells and CD138+ plasma cells. T₂-FLAIR images show the extent of the PML lesion. T₂-weighted turbo/fast spin echo sequences may show intralesional vacuoles (bottom row). In early stages, the T₁-weighted signal intensity can be normal (bottom row), while in advanced stages, it is low (top row). In inflammatory PML, contrast-enhanced T₁-weighted images often show enhancement (bottom row). High signal intensity on DWI in classic PML represents areas of active inflammation and viral replication leading to swelling of oligodendrocytes. Lesions show heterogenous ADC values, but rarely is diffusion markedly restricted. Scale bars = 50 µm.

Figure 3.

Imaging findings of PML on multisequence MRI. Axial brain MRI images of a patient presenting with PML lesions in bilateral occipital lobes, primarily in the subcortical white matter and adjacent cortical grey matter. These images are consistent with a classic MRI pattern of PML, with high signal intensity on T₂-weighted images and DWI, and low signal intensity on the contrast-enhanced T₁-weighted images (arrows). On SWI, bands of low signal intensity adjacent to cortical grey matter can be observed, suggesting deposition of paramagnetic compounds such as iron (arrows).

T₂-weighted MRI

T₂-weighted sequences, including T₂-FLAIR, are standard in the screening, diagnosis and monitoring of PML in clinical practice, and T₂-FLAIR is particularly sensitive in the detection of PML lesions. Although not formally investigated yet, it can be anticipated that T₂-FLAIR (particularly 3D FLAIR) has the highest sensitivity in the detection of PML lesions. However, conventional T₂-weighted images are useful for describing lesion size and location (most frequently observed in the frontal and parieto-occipital lobes and to a lesser extent in the temporal lobes, the basal ganglia region, and posterior fossa), identifying lesion borders that are sharp towards the cortex but blurred towards the white matter, perilesional nodules, grey matter involvement, and the presence or absence of intralesional vacuoles.^9,36–39 The spinal cord and optic nerves are not involved. These characteristics are very important to exclude other diagnoses or comorbidities, such as existing multiple sclerosis lesions or new multiple sclerosis disease activity.⁴⁰ Though the presence of a central vein may be specific to multiple sclerosis lesions,⁴¹ there are no published data regarding the central vein sign in PML, so it is uncertain whether the presence of a central vein distinguishes multiple sclerosis from PML lesions.

Another characteristic imaging feature potentially observed on T₂-weighted sequences is the presence of multiple small punctate lesions outside the main PML lesions. Some of these lesions may show contrast enhancement. The histopathology of these lesions is not completely understood, but it has been suggested that these lesions represent inflamed perivascular spaces.⁴² Depending on the image plane, this imaging feature is referred to by the descriptive terms ‘punctate pattern’ and ‘milky way appearance’.^43,44

Descriptors of the extent of T₂ lesion dissemination (e.g. unilobar, multilobar, widespread) are considered prognostic markers.⁴⁵ Lesion volume based on T₂-weighted images has also been shown to correlate with CSF JCV load; a cross-sectional study demonstrated that the smaller the MRI T₂ lesion volume at the time of diagnosis, the lower the viral load and the higher the chance that JCV DNA is not detected in the CSF.¹⁶ For this reason, in cases where PML is strongly suspected on clinical grounds, lesion evolution on T₂-weighted MRI has been used to support a presumptive diagnosis of PML even when CSF JCV PCR is negative.^9,19

Once the diagnosis of PML has been established, T₂-weighted MRI remains critical to assess lesion progression.²³ The time interval between follow-up scans depends on the clinical situation, including the patient’s underlying immunosuppressive condition, comorbidities, and age. PML lesions usually expand continuously in patients without immune reconstitution. On the other hand, in patients with partial or complete immune reconstitution, PML lesions tend to expand slowly for weeks to months, after which PML-IRIS may develop. In PML-IRIS, lesions undergo rapid expansion, accompanied by signs of inflammation (e.g. perilesional oedema, mass effect, enlargement of perivascular spaces, and contrast enhancement).^21,22,46 Once PML becomes controlled, usually through immune reconstitution, lesion volume ceases to correlate with CSF viral load.¹⁸

T₂-weighted MRI has been applied in case reports and series reporting various treatment strategies being investigated for PML. The most recent therapeutic strategies using immune checkpoint inhibitors or allogeneic BK virus-specific T cells have shown a remarkable effect on lesion progression; an example is shown in Fig. 4, which illustrates the potential for longitudinal MRI to be applied in the treatment setting.⁴⁷ The use of MRI in prospective interventional PML clinical trials is reviewed later in this article. In most reports of PML’s response or non-response to therapies, brain MRI, particularly T₂-weighted imaging, was used to monitor disease progression during treatment (Supplementary Table 1). Most examples include only a qualitative or semiquantitative description of lesion load and evolution during treatment; in only a few case series were the T₂-weighted scans used to quantify lesion volume, with facilitation via deep learning algorithms.⁴⁸

Figure 4.

Axial T₂-weighted and contrast-enhanced T₁-weighted MRI of a PML patient treated with BK virus-specific allogeneic T-cells. This patient’s MRIs demonstrated lesion growth and contrast enhancement at early stages (progressing) but stability and/or retraction, with resolution of contrast enhancement, weeks after treatment initiation.

T₁-weighted MRI without and with gadolinium

On unenhanced T₁-weighted images, PML lesions are initially isointense or mildly hypointense, reflecting early stages of disease. Disease progression is characterized by a further decrease in T₁-weighted signal intensity, approaching that of CSF. The lack of reversion to T₁-weighted isointensity⁴⁹ and the resulting encephalomalacia suggest that there is prominent tissue destruction and no significant remyelination in PML lesions.

Gadolinium enhancement in PML was initially assessed in HIV-AIDS patients, where contrast enhancement was rare; this has been termed classical PML.^50,51 However, an inflammatory response termed inflammatory PML, characterized by lesion enhancement, vasogenic oedema, mass effect, lymphocyte infiltration, and clinical worsening, was described in patients with HIV following treatment with HAART.⁵² Inflammatory PML has subsequently been described in HIV-positive patients not receiving HAART, as well as non-HIV-associated PML (e.g. natalizumab-associated PML).⁵² Interestingly, in natalizumab-treated patients, even asymptomatic and early PML lesions can demonstrate contrast enhancement, and this phenomenon is thought to be more common in this patient population.⁹ Contrast enhancement in these cases frequently demonstrates a perivascular distribution pattern, manifesting as punctate lesions, and may reflect incomplete immune suppression (Supplementary Table 2).^9,43,44,53 Histological evidence of T-cell infiltration in patients with inflammatory PML further supports the hypothesis of partial immune reconstitution contributing to this imaging finding.^1,52

Inflammation can be observed throughout the course of PML, including at the time of presentation if the patient’s underlying immune suppression is only partial. Immune reconstitution occurs if the patient’s PML-predisposing condition is reversed and is a dynamic process that can occur slowly or rapidly, depending on the underlying condition. For example, rapid and dramatic immune reconstitution generally occurs in patients with HIV/AIDS who initiate HAART, or in natalizumab-treated multiple sclerosis patients who undergo plasmapheresis.⁵⁴ On the other hand, slow, spontaneous immune reconstitution can occur in patients on immunosuppressive treatments that can be withdrawn, but not rapidly cleared or reversed.

The inflammation often observed in the setting of immune reconstitution, which is generally a desired component of PML management, is necessary for the successful control of viral infection. However, the degree of inflammation following immune reconstitution is a spectrum that can vary with the rapidity and magnitude of reconstitution. PML-IRIS is a clinical syndrome in which an exuberant immune response becomes the dominant pathophysiology (rather than virus-related tissue destruction). Patients with PML-IRIS experience clinical worsening due to an exaggerated immune response. PML-IRIS can result in extensive tissue damage, irreversible neurological deficits, and considerable morbidity and mortality. Corticosteroids can dampen the immune response during IRIS,⁵⁵ but premature treatment can also impair the immune response and thereby escalate viral replication.⁵⁶

Though some MRI features, such as pronounced mass effect, midline shift, or even impending herniation, may be suggestive of PML-IRIS, MRI is not able to distinguish reliably between desirable inflammation and inflammation associated with PML-IRIS. Therefore, in the clinical trial setting, MRI should be part of criteria defining IRIS, but this information should be interpreted in the clinical context, as PML-IRIS is a clinical (rather than radiological) diagnosis.

To identify a characteristic enhancement pattern in natalizumab-associated PML-IRIS, Wattjes et al.¹⁹ retrospectively evaluated PML patients and found that patchy or punctate enhancement at the lesion border was the most consistent finding. However, in a larger follow-up study, differences in the enhancement patterns observed in inflammatory PML compared to PML-IRIS were not always discernible.⁴⁶

Diffusion MRI

DWI applies diffusion-sensitizing magnetic field gradients in at least three orthogonal directions, followed by reconstruction of images that can sensitively detect and quantify the motion of water within tissue. In the context of acute stroke, reduced diffusion of water molecules may indicate compromise of normal metabolism due to the failure of energy-dependent ion pumps.⁵⁷

The borders of PML lesions, where viral replication and subsequent oligodendrocyte swelling occurs, commonly demonstrate increased signal on DWI (b-1000 images, which are generally preferred for this application). Often, an intermediate or low apparent diffusion coefficient (ADC) value can be observed in the same area. This bright rim can track the leading edge of lesions as they expand into previously unaffected white matter. Interestingly, use of b-values greater than the 1000 s/mm² typically used in clinical practice may improve the contrast-to-noise ratio, thereby improving visualization of the rim-and-core pattern.⁵⁸ On ADC maps, which quantify the rate of water diffusion, the areas of increased DWI signal are usually isointense to extralesional (uninvolved) white matter but may occasionally show decreased signal, which might indicate the combined presence of reduced diffusion and oedema. In the lesion centre, on the other hand, diffusion is usually facilitated (i.e. high ADC values), corresponding to loss of both myelin and axons. These findings and their dynamic changes create a fairly unique rim-and-core pattern on DWI (Fig. 2),⁵⁹ which can be helpful diagnostically, particularly by comparison to multiple sclerosis (where such a pattern is rare).⁶⁰

In the setting of PML-IRIS and the accompanying opening of the blood–brain barrier, vasogenic oedema results in higher ADC values. The rate of this increase has been used to detect effects of therapeutic intervention in HIV-PML, differentiating responders from non-responders.⁶¹ PML-IRIS also lacks the aforementioned expansion of the DWI-hyperintense PML lesion rim.⁹

Susceptibility MRI

Susceptibility-weighted MRI (SWI), available in several variants on clinical scanners, leverages the magnetic susceptibility differences of brain tissue constituents to generate image contrast.^62–64 Specifically, compounds that distort the local magnetic field, such as iron, calcium, myelin, and blood products, alter the phase of the MRI signal, and to a lesser extent its magnitude (so-called T₂*w images).

A number of case reports, case series, and cohort studies (Supplementary Table 3) have described susceptibility changes that are both common and characteristic of PML, manifesting as a paramagnetic leukocortical ‘rim’ or ‘band’ that can occur independently of the predisposing factor. Several of these studies have demonstrated longitudinal evolution of this band over time, including increases in thickness and continuous spread along the juxtacortical white matter over several months.^65–67 However, some studies have also demonstrated that a leukocortical band can occur in other disease states, including stroke and encephalitis.⁶⁸

Although the exact pathological basis and pathophysiology of the PML leukocortical SWI band are not yet known, iron accumulation in the normally iron-rich subcortical U-fibres, possibly within phagocytes,^65,67,69,70 blood–brain barrier disruption with leakage of haem products,⁶⁵ and cortical laminar necrosis⁷¹ have been proposed.

Other imaging modalities

Magnetization transfer MRI

Magnetization transfer MRI measures the exchange of magnetization between two proton pools: free protons and protons bound to the macromolecular fraction (which in brain consists primarily of myelin).⁷² The magnetization-transfer ratio (MTR) reflects the efficiency of the exchange and has been used in multiple sclerosis to quantitate the extent of demyelination and remyelination in lesions,^73–85 as well as diffuse microscopic areas of demyelination in normal-appearing white matter.^86–89

Because JCV lytic infection of oligodendrocytes causes profound and likely irreversible demyelination, magnetization transfer imaging could be a promising imaging tool for diagnosis and tracking disease progression in PML. Several small studies in HIV-AIDS patients with PML have been reported (Supplementary Table 4), one of which demonstrated that despite a similar appearance on T₂-weighted MRI, the mean MTR values of PML lesions in patients with HIV were lower than those of patients with HIV encephalitis.^90–92 However, another study utilizing region of interest measurements did not find a difference in MTR between multiple sclerosis lesions and PML lesions, though the stage of PML at the time of this analysis was not indicated.⁹³ Finally, serial longitudinal MTR in a parieto-occipital PML lesion in a patient with cutaneous T-cell lymphoma demonstrated a decrease in MTR at initial presentation, but no change in MTR at 6 or 9 months thereafter.⁹⁴ To date, no published interventional trials for PML have used magnetization transfer imaging.

Magnetic resonance spectroscopy

Multiple studies have used single voxel proton magnetic resonance spectroscopy (MRS) to evaluate PML lesions using short (20–35 ms) and long (135 ms) echo time acquisitions. Metabolite concentrations in lesions are expressed relative to the creatine peak at 3 ppm, which is stable and serves as an internal control.⁹⁵ A voxel placed in the central portion of the lesion is often compared to the contralateral normal-appearing white matter and also to healthy control subjects.^95–98

Typical spectra for PML lesions show reduction of the N-acetylaspartate (NAA) peak at 2.02 ppm, reflecting axon loss or dysfunction; an elevated choline peak at 3.2 ppm, representing cell membrane and myelin breakdown; and a prominent lactate doublet at 0.9 and 1.3 ppm, attributed to the presence of foamy macrophages, mitochondrial dysfunction, and anaerobic glycolysis.^99–101 At short echo time, a variable myoinositol peak at 3.6 ppm, if present in acute lesions, has been correlated with patient survival.¹⁰² Also, at short echo time, lipid peaks at 0.9 and 1.3 ppm, representing cell membrane breakdown and potentially activated lymphocytes,¹⁰³ are elevated at the onset of PML-IRIS in HIV- and natalizumab-associated PML.¹⁰⁴ Clinical spectroscopy-based studies are summarized in Supplementary Table 5.

Perfusion MRI

Data on perfusion MRI in PML are very limited, but hyperperfusion has been reported as a feature of active PML lesions. One study used arterial spin labelling (ASL), a perfusion method that magnetically traces arterial blood as it flows into tissue, to describe intense hyperperfusion in active PML lesions; this resolved once PML-IRIS developed and lesions stopped expanding.¹⁰⁵ Similar observations were presented in a longitudinal case study, in which hyperperfusion at the edge of PML lesions was also accompanied by high lactate on MRS and DWI hyperintensity, all of which suggest lesion activity.⁹⁸ To date, there are no relevant radiological-pathological correlative data, and the pathophysiology of hyperperfusion in evolving, active PML lesions in the absence of an immune reaction remains unclear. However, it is possible that hyperperfusion in active PML lesions reflects the influx of metabolically active macrophages that clear dead cells and myelin.

PET

PET is a non-MRI technique that uses radioligands—small, biologically active compounds labelled with radioactive elements—to create images with potentially exquisite molecular specificity. Although PET is inherently quantitative when used in conjunction with pharmacokinetic models, it has 5–10-fold lower spatial resolution than MRI, uses ionizing radiation, and is relatively slow (images are formed over a 60–90-min period). The most common radioligand is ¹⁸F-fluorodeoxyglucose (FDG), a glucose analogue that highlights metabolically active tissue.

PET studies in PML are confined to a few case reports and case series, describing a total of 17 cases to date (Supplementary Table 6). All patients were studied using FDG, but other radiotracers—including ¹¹C-methionine for amino acids (eight patients, four studies),^{106–109  11}C-flumazenil for neurons (one patient, one study),¹⁰⁹ and methyl-¹¹C-4’-thiothymidine for DNA synthesis (one patient, one study)¹⁰⁹—have also been used. Although these PET studies span more than 20 years and are technically heterogeneous, taken together they suggest that PML, whether active or inactive, is associated with low FDG uptake (suggesting hypometabolism) relative to healthy brain. A possible exception is in the setting of PML-IRIS, in which several of the reported studies suggest that tracer uptake can be high, regardless of gadolinium enhancement; however, these results are derived from very limited experience.

In addition to these studies, a recent PET study with ¹⁸F-GE-180 (a radioligand for the translocator protein 18 kDa, which binds to activated macrophages/microglia and astrocytes) in eight patients with natalizumab-associated PML, has shown promising results for the detection of immune reconstitution.¹¹⁰

Summary of imaging modalities in PML

Of all pulse sequences used in clinical practice, T₂-weighted sequences, in particular T₂-FLAIR, are very sensitive for the detection of PML lesions and should therefore be considered most important for diagnosis of PML.^9,111 Conventional T₂-weighted (turbo/fast) spin-echo sequences can confirm the findings on T₂-FLAIR and also demonstrate imaging characteristics highly suggestive of PML that are helpful in the differential diagnosis.³⁶ Both T₂-FLAIR and T₂-weighted are therefore recommended for both diagnostic and monitoring purposes as well as for PML screening protocols during pharmaco-vigilance for multiple sclerosis therapies.^34,112 If high quality 3D-T₂-FLAIR images are available, one can consider omitting conventional T₂-weighted acquisition.¹¹³

In the context of PML and PML-IRIS and supported by pathological evidence, gadolinium enhancement has traditionally been considered a sign of inflammation. However, the relationship between enhancement and inflammation is complex, and though it would be immensely useful to describe enhancement patterns that can reliably identify PML-IRIS, such a description has proven elusive, and the diagnosis of PML-IRIS remains clinical.¹¹⁴ In the context of a clinical trial, assessment of enhancement has value for baseline phenotyping (including stratification) and monitoring the effects of immune reconstitution, which is currently the aim of most therapeutic development efforts.

The typical rim-and-core pattern of active PML on DWI and its rapid centrifugal extension can refine the differential diagnosis of PML, and potentially detect treatment-related arrest of lesion expansion and active virus replication. Furthermore, DWI is a useful adjunct to confirm the onset of IRIS, and as such may find use as an outcome measure in clinical trials aimed at immune restoration.

Although the presence of a paramagnetic leukocortical band on SWI may not be specific to PML, it is characteristic and reflects consistent aspects of PML pathology. Available data suggest that the band is more typically observed in chronic PML, but it has also been documented in asymptomatic or acute PML and therefore may prove diagnostically useful in cases where the disease developed insidiously. On the other hand, the presence of this finding in PML survivors raises the possibility that it can track lesion consolidation in the context of treatment, which argues for further study and the potential inclusion of SWI in future clinical trials.

Because MTR is sensitive to demyelination and is widely available on clinical MRI scanners with a relatively short (<5 min) acquisition time, it is an attractive method for quantifying PML lesion volume and associated tissue damage in response to therapy. However, most reports have been cross-sectional, and longitudinal studies are required before definitive recommendations can be made.

MRS has led to some interesting insights regarding PML, including the suggestion that metabolic abnormalities may extend far beyond T₂-visible lesions, and as such it may be useful in characterizing selected PML cases. Although its chemical specificity makes it an attractive approach, MRS is rarely used in clinical practice because it is technically more difficult and time-consuming than conventional MRI, reproducibility across scanners is limited, and multicentre studies, which would be necessary for definitive clinical trials, are especially difficult due to challenges associated with standardization. These considerations therefore limit its potential use in clinical trials.

Additional study of perfusion MRI, and ASL in particular, is required to substantiate the results published to date. It is unknown whether these techniques can differentiate between active PML and potential treatment effects in the context of a clinical trial.

FDG-PET is not likely to be a useful imaging modality for diagnosing PML, as many pathologies are associated with hypometabolism, especially relative to the highly metabolic cortical grey matter. Additionally, recommendation of FDG-PET for use in the clinical trial setting would be challenging due to the lack of longitudinal studies for this modality in PML and the radiation exposure associated with each scan. Peer-reviewed data using other radioligands are even more sparse, and the published experience is mixed. Nevertheless, the possibility that FDG-PET may be able to identify inflammation is attractive, since the most successful therapeutic strategies to date involve immune restoration; further study is required.

Imaging outcomes in PML clinical trials

Investigations of treatments for PML are described primarily in case reports, small case series, and retrospective cohort reviews. Most of these include a qualitative description of imaging findings, reflecting the importance of MRI in both diagnosis of PML and monitoring of disease course. Optimal use of imaging measures, as prespecified end points or outcome measures in the context of prospective interventional trials in PML, has not been established.

To date, only nine prospective interventional trials in PML have been registered with clinicaltrials.gov (Supplementary Table 7), and only five have resulted in publication.^115–118 Two trials were terminated early following planned interim analyses that showed no expectation of treatment effect. Of the published studies, one did not report imaging findings,¹¹⁵ and one included MRI for safety monitoring only.¹¹⁷ The remaining three included specific MRI measures as secondary outcomes to assess efficacy.^116,118

Marra et al.¹¹⁹ reported a pilot study of cidofovir in 24 cases of HIV-PML. The imaging outcome was based on neuroradiological interpretation, using an MRI rank score derived by consensus of three neuroradiologists and ranging from 0 (normal) to 10 (most abnormal). Although no treatment effect was observed, a worsening MRI rank score correlated with worsening clinical scores as measured by the Karnofsky Performance Scale (P = 0.05).

Royal et al.¹¹⁶ reported a pilot study of topotecan in 11 evaluable patients with HIV-PML. The imaging outcome utilized a semi-automated segmentation method.¹²⁰ The authors measured ‘time to radiographic response’, defined as ≥10% decrease in lesion volume; stable disease corresponded to no new lesions or no increase of ≥25% for at least 8 weeks, and progressive disease to an increase in MRI lesion volume ≥25%. All three PML survivors demonstrated radiographic improvement with time to radiographic response of ∼2 months, whereas no non-survivors met criteria for radiographic response.

Clifford et al.¹¹⁸ reported a phase 1/phase 2 study of mefloquine. The study was terminated early based on a planned interim analysis showing lack of treatment effect. At the time of termination, the study had recruited 37 patients, 29 of whom had underlying HIV; of these, only 12 completed the planned treatment course. The MRI outcome in this study was based on detection of gadolinium enhancement and on T₂-weighted/T₁-weighted lesion volumes calculated using a semiautomated edge-detection contouring thresholding technique,^121,122 which had previously been utilized for analysis of multiple sclerosis lesions. Baseline lesion volumes were compared to lesion volumes at Week 4, with gadolinium enhancement considered as a dichotomous covariate. Consistent with the lack of treatment effect on the primary outcome (change in JCV DNA copy number in CSF from baseline to Week 4), there were no differences in MRI measures between treatment groups, and no MRI data were reported.¹¹⁸

Summary and recommendations

To date, only three published clinical trials in PML have included a clearly defined imaging end point among the secondary outcomes. Subjective MRI-based rank scores are easily applied and may correlate to a true clinical end point but are less informative for determining treatment effect sizes at the group level or for evaluating individual responses to treatment, and generalizability across studies and patient cohorts is limited. Standardized quantitative measures, such as methods based on lesion segmentation, may overcome these limitations, and fully automated deep learning-based techniques may facilitate their application.⁴⁸ Unanswered questions include the relevance of lesion location and the definition of clinically meaningful radiological changes. Clearly, some brain regions (such as the frontal lobes) can tolerate relatively large degrees of injury compared to others (such as the brainstem). Hence, application of volumetric assessments of lesion burden as surrogate outcomes for clinical trials will likely require development of anatomically based scaling methods.

Discussion and overall recommendations

In this review, we have summarized available data regarding the application of neuroimaging techniques to PML. Although there is no basis for direct comparison of these approaches with respect to clinical use (diagnosis and disease monitoring) or therapeutic trials, several conclusions and recommendations are possible. Recommendations for use of these pulse sequences in the clinical trial setting, features specific to PML, and associated advantages and limitations, are summarized in Table 1.

Table 1.

Imaging approaches used for PML

Pulse sequence	Advantages and imaging characteristics	Limitations	Recommendation
MRI T2 FLAIR	Signal intensity: hyperintense Highest sensitivity for PML lesion detection: used in PML screening protocols Useful for diagnosis and disease monitoring	Quantitative measurements necessary for clinical trials Lesion borders are not blurred Multifocal lesions require manual tracing Difficult to segregate from confluent multiple sclerosis lesions Occasionally ambiguous (decrease in T2/FLAIR volume associated with encephalomalacia)	Recommended as outcome measure for clinical trials
MRI conventional T2	Signal intensity: hyperintense Particularly useful for PML characteristics (e.g. punctate pattern, vacuoles) Can confirm imaging findings seen on FLAIR	See T2 FLAIR limitations above	Adjunct measure for FLAIR to characterize lesion pattern
MRI T1-weighted	Signal intensity: isointense (early stage), hypointense (advanced stages) Demonstrates the degree of tissue destruction based on signal intensity Identifies laminar necrosis	Qualitative measure not useful as clinical trial outcome	Useful for comparison to post-contrast T1-weighted
MRI T1-weighted + contrast	Contrast enhancement reflects host immune response against the virus Predicts onset of PML-IRIS in immunocompromised patients Identifies inflammatory PML	Contrast enhancement variable Inhomogeneous	Useful for staging the disease and evaluating therapeutic response with IRIS
MRI diffusion-weighted	Signal intensity: variable, ranging from hyperintense to isointense to hypointense Hyperintense lesions (w/low ADC) reflect active demyelination and viral replication Useful for diagnosis and disease monitoring Rapid sequence (<5 min) available on all MRI platforms Reversible change: improvement associated with reversal to isointensity DWI hyperintense signal larger than lesion area on FLAIR/T2	Scanner sensitivity Occasionally confounding hyperintense signal in white matter tracts and multiple sclerosis hypointense lesions	Useful for locating areas of active disease
MRI susceptibility-weighted	Useful for identification of characteristic leukocortical band/rim	May not be specific to PML Unclear whether finding portends favourable prognosis or indicates lesion chronicity Unknown correlation of evolution with clinical change and prognosis Quantitation by QSM not readily available Unlikely to be useful for lesions in the posterior fossa	Research only
Magnetization transfer	Decrease in MTR reflects extent of demyelination MTR of PML lesions and global brain tissue (grey and white matter) can be determined Short sequence widely available on most MRI scanners Reversible change can be demonstrated	Postprocessing required MTR of PML lesions versus multiple sclerosis lesions may overlap Diffuse changes in NAWM of multiple sclerosis patients may confound the results	Longitudinal studies required
MR spectroscopy	Global reduction of NAA indicative of axonal damage can precede PML diagnosis Presence of lipid peaks may indicate on IRIS onset Elevated choline suggest cell membrane and myelin turnover	Weak diagnostic specificity for PML Technically challenging and time consuming Limited reproducibility across scanners Imperfect voxel positioning and variable results depending on lesion age Longitudinal evaluation requires a single scanner	Research only; however, may be useful if reduction of NAA precedes PML lesion appearance Accelerated brain atrophy may be an alternative measure
PET	Limited studies but potentially useful for PML-IRIS	Limited availability of PET scanners Optimal PET ligand uncertain PML-IRIS patients are frequently in critical care setting, precluding PET scanning	Research only

QSM = quantitative susceptibility mapping.

With respect to imaging of PML in routine clinical care, there is substantial experience supporting the use of T₂-weighted sequences (including T₂-FLAIR) for identifying and monitoring individual lesions; gadolinium enhancement for identifying PML-IRIS (albeit not reliably); DWI for determining whether a lesion is actively expanding; and potentially SWI for assessing lesion chronicity. Perfusion, magnetization transfer, spectroscopy, and PET remain research applications at this time. Quantification of atrophy and lesion volumes are emerging areas of interest in the field of PML, but only limited data are available. Techniques for brain and lesion volume quantification are now available but require further research regarding performance and clinical application.⁴⁸ In addition, there is no evidence supporting or refuting the application of ultra-high field MRI in PML. A limited number of case reports describe PML lesion characteristics and PML lesion evolution based 7 T MRI but do not report advances in lesion detectability and characterization relative to standard MRI field strengths.^123–125

With respect to therapeutic trials, imaging outcome measures should ideally be objective, quantifiable using standardized methods,¹²⁶ specific to the disease of interest, and correlated (both cross-sectionally and longitudinally) to clinically meaningful symptoms and disability. A surrogate end point should be predictive of the true clinical end point of interest and be in the causal pathway.¹²⁷ Imaging measures show promise in this regard, potentially as surrogates of either a clinical measure or event (e.g. survival, degree of physical disability, quality of life) or disease activity (e.g. resolution of infection). Though extent of involvement on MRI (e.g. unilobar or widespread/multilobar, as well as lesion burden) and the presence of supratentorial versus infratentorial lesions have been shown to be associated with survival and functional outcomes in some studies of PML,^128–131 these data are generally limited, heterogeneous, and confounded by other clinical features (e.g. age, underlying disease, and baseline disability) known to be associated with survival.

Successful validation of a surrogate end point could greatly increase trial efficiency by reducing study duration, sample size, and cost. Importantly, a simplified and shorter trial design would also be viewed more favourably by patients and families, thereby facilitating recruitment and retention, which are paramount in the study of rare diseases. Specific factors that should be considered when selecting imaging modalities for clinical trials are the underlying conditions of the patient population of interest and the mechanism of action of the therapeutic intervention (e.g. immune-stimulating versus antiviral).

Unfortunately, PML research has not yet coalesced around a set of imaging methods that have proved indispensable in the context of therapeutic trials. However, careful application of imaging approaches that are useful in clinical practice, augmented by judicious use of quantitative methods, may capture clinically meaningful information in the trial setting. One potential approach would be development of a composite score, based on PML lesion evolution, that accounts for the proportion of ‘active’ (new and/or expanding on T₂-FLAIR and showing hyperintense signal on DWI) versus ‘inactive’ (T₁-weighted hypointense with focal atrophy and a leukocortical T₂*w band) lesions. Additional methods that have shown promise in other diseases, such as quantification of brain structure atrophy, may also be relevant for PML, but this has not yet been explored.

Another consideration for the use of imaging outcomes for PML in both clinical practice and clinical trials is that of the timing for appropriate monitoring. In general, we recommend repeat brain MRI with gadolinium administration in the event of clinical worsening to evaluate for imaging signs potentially consistent with PML-IRIS. As stated previously, the time interval between follow-up scans depends on the clinical situation, including the patient’s underlying immunosuppressive condition, comorbidities, and age. However, in terms of routine surveillance to monitor PML evolution, particularly in the clinical trial setting, we generally recommend brain MRI with gadolinium administration at least on a monthly basis.

Development and validation of potential imaging outcomes for PML will likely depend upon well-designed natural history studies and post  hoc analyses of previously completed clinical trials. Synthesis of this information will be important to determine whether a particular pulse sequence, or combination of pulse sequences, has the characteristics desired in an imaging outcome measure for PML. In the interim, we recommend that clinical trials acquire multisequence MRI including, at minimum, the following core sequences for diagnosis and monitoring of PML: T₂-weighted/T₂-FLAIR, T₁-weighted with and without gadolinium, DWI, and potentially SWI. Future therapeutic trials will likely lead to valuable lessons in piloting new and exploratory imaging approaches, subsequently honing more efficient trial designs for future use.

Funding

No funding was received towards this work.

Competing interests

Views expressed in this manuscript should not be construed to represent the views or policies of either the FDA or NIH. All opinions, recommendations, and proposals are unofficial and non-binding on FDA and NIH. L.E.B. reports no competing interests. M.P.W. reports speaker and/or consultancy from Novartis, Sanofi-Genzyme, Genilac, Bayer Healthcare, Roche, Biogen, Biologix, Celgene, Merck Serono, Icometrix, Imcyse, IXICO, Medison, outside the submitted work. I.C.M.C. is supported by the Intramural Research Program of NINDS. A.G. has received honoraria for lecturing and financial support for research from Bayer, Biogen, Merck Serono, Novartis, Roche. I.M. reports personal fees from Biogen Idec, Bayer Healthcare, TEVA, Serono, Novartis, Genzyme, Roche, grants from Biogen Idec, Genzyme, Novartis, grants from N-RENNT 2 (Niedersachsen Research Network on Neuroinfectiology), outside the submitted work. T.Y. is supported by the National Institute for Health Research University College London Hospitals Biomedical Research Centre. D.S.R. is supported by the Intramural Research Program of NINDS. He has received research support from Vertex Pharmaceuticals and Sanofi-Genzyme, unrelated to the present work. N.R. has received consulting fees from Blueprint Medicine, Roche, Sanofi, and Bristol Myers Squibb.

Supplementary material

Supplementary material is available at Brain online.

Abbreviations

ADC

apparent diffusion coefficient

IRIS

immune reconstitution inflammatory syndrome

MTR

magnetization-transfer ratio

PML

progressive multifocal leukoencephalopathy

SWI

susceptibility weighted imaging