Fatal Encephalopathy with Wild-Type JC Virus and Ruxolitinib Therapy

Abstract

Objective:

JC virus (JCV) infection is a lytic infection of oligodendrocytes in progressive multifocal leu-koencephalopathy; less common forms of central nervous system manifestations associated with JCV infection include granule cell neuronopathy, encephalopathy, and meningitis. Presented is the first case of fatal JCV encephalopathy after immunosuppressive therapy that included ruxolitinib.

Methods:

Postmortem analysis included next generation sequencing, Sanger sequencing, tissue immunohistochemis-try, and formalin-fixed hemisphere 7T magnetic resonance imaging.

Results:

JCV DNA isolated from postmortem tissue samples identified a novel 12bp insertion that altered the transcription site binding pattern in an otherwise “wild-type virus,” which has long been thought to be the nonpathogenic form of JCV. Anti-VP1 staining demonstrated infection in cortical neurons, hippocampal neurons, and glial and endothelial cells.

Interpretation:

This expands the spectrum of identified JCV diseases associated with broad-spectrum immunosuppression, including JAK-STAT inhibitors, and sheds light on an additional neurotropic virus strain of the archetype variety.

The classic form of central nervous system (CNS) JC virus (JCV) infection is lytic infection of oligodendrocytes in progressive multifocal leukoencephalopathy (PML). Less common forms of CNS manifestations associated with JCV infection include granule cell neuronopathy, encephalopathy, and meningitis. PML has classically been associated with rearrangements in the early noncoding regulatory region of JCV, and atypical infections have been associated with either wild-type (archetype) JCV or novel mutations in the protein coding regions of agnoprotein or VP1 viral capsid.^1,2 Four PML cases have been associated with STAT1 gain-of-function mutations.³ STAT1 is a crucial transcription factor that regulates interferon response in the JAK-STAT immune regulatory pathway. In 2013, the JAK1/2 inhibitor ruxolitinib was associated with a case of fatal PML; more recently, ruxolitinib was associated with fatal JCV meningitis.^4,5 Here, we describe the first case of fatal atypical JCV encephalopathy after immunosuppressive therapy that included ruxolitinib.

Case Report

Standard Protocol Approvals, Registration, and Patient Consent

The patient was studied at the NIH in Bethesda, Maryland under NIH protocol 08-H-0046 approved by the institutional review board.

Viral Sequencing and Analysis

JCV DNA was isolated from 1mm³ frozen postmortem tissue using the commercial EZ1 XL DNA Tissue Kit (Qiagen, Valencia, CA; Cat No. 953034) according to the manufacturer’s published protocol. Samples sequenced on Illumina were phenol/chloroform extracted, ethanol precipitated, and resuspended in deionized water. Illumina sequence data (MiSeq) in fastq format were generated for 10M read pairs (2 × 150bp) from total DNA (100ng) extracted from human postmortem cerebellar cortex tissue (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE123125) and run in tandem against a known JCV/BK virus (BKV) positive control sample. Conversion of the data from fastq format to unaligned bam format was accomplished using the FastqToSam command (http://broadinstitute.github.io/picard/). Sequences were referenced mapped by sample to the current instance of the human genome (hg38), restricting mapping to only annotated known mRNA. Read pairs that did not map to known human mRNA were remapped against a reference pool of complete JCV genome sequences downloaded from the National Center for Biotechnology Information (NCBI; n = 581). The alignment consensus from the postmortem and control samples were exported from Integrative Genomic Viewer and used to identify the top 100 viruses sequence-related to it by megablast search (https://blast.ncbi.nlm.nih.gov/). The control sample top 2 by percentage length covered by the reads mapped to JCV strains AT-2 (AB048569) and HR-5 (AB048572), whereas the consensus postmortem cerebellar sample mapped to JCV GH-3 (AB048545) and GH-1 (ABO38252). BKV was present in the control but was not identified in the postmortem sample.

The JCV primer set was designed using Primer3Plus (http://www.bioinformatics.nl/cgi-bin/primer3plus/primer3plus.cgi) based on NCBI reference sequence NC_001699.1 (https://www.ncbi.nlm.nih.gov/nuccore/NC_001699.1). Primers are given in the Supplementary Table. The whole JCV DNA genome of varying postmortem tissues was amplified with the following polymerase chain reaction (PCR) cycling protocol: 95°C for 2 minutes; 95°C for 30 seconds; 55°C for 30 seconds; 72°C for 2 minutes; 34x cycles; 72°C for 5 minutes; 4°C end (see Supplementary Table). PCR reaction products were purified with HighPrep PCR beads (MagBio Genomics, Gaithersburg, MD; Cat No. AC-60050) according to the manufacturer’s protocol and sent for Sanger sequencing through Eurofins. SeqMan Pro (DNAStar) was used for consensus assembly and BLAST for comparison analysis (https://blast.ncbi.nlm.nih.gov).

Magnetic Resonance Image Acquisition and Analysis of Autopsy Tissue

Postmortem magnetic resonance imaging (MRI) was performed on an intact brain autopsy specimen fixed in 10% formalin for at least 14 days. The specimen was then placed in a custom-built holder filled with Fomblin for MRI. High-resolution imaging was performed using a 7T magnet with 32-channel transmit-receive (TX/RX) head coil at isotropic resolutions of 150μm (Constructive Interference in Steady State (CISS) sequence), 420μm (T2 multiecho gradient echo), and 310μm (inversion prepared T1-weighted fast low angle shot (FLASH)) sequences. After imaging, Fomblin was washed off with saline and tissue was returned to 2% formalin. The MRI was interpreted by a board-certified neuroradiologist.

Electron Microscopy/Ultrastructural Analysis of Human Cerebellar Tissue

For morphological studies using transmission electron microscopy, postmortem brain samples were fixed with 4% glutaraldehyde (EMS, Fort Washington, PA) in 0.1M sodium cacodylate buffer at pH 7.4.⁶ Fixed samples were further dissected into 1mm³ slices and immersed in the same fixative for at least 24 hours at 4°C. Slices were then washed with cacodylate buffer and treated with 1% osmium tetroxide in cacodylate buffer on ice for 1 hour. Treated specimens were block stained with 0.25% uranyl acetate in acetate buffer at pH 5.0 overnight at 4°C. The following day, block stained specimens were dehydrated in graded ethanol solutions and embedded in epoxy resin. Embedded blocks were sectioned approximately 70nm thick and mounted on mesh copper/nickel grids. Mounted sections were counterstained with uranyl acetate and lead citrate. Stained grids were assessed with a JEOL (Tokyo, Japan) 1200 EXII transmission electron microscope, and images were photographed using a bottom-mounted digital CCD camera (AMT XR-100; Advanced Microscopy Techniques, Danvers, MA).

Detection of VP1 Protein by Western Blot Analysis

Brain samples were dissected from the occipital and frontal lobes. In the occipital lobe, dissection was performed to differentiate between gray and white matter. Brain samples (50mg) were homogenized in 500μl of N-PER buffer (Thermo Fisher Scientific, Waltham, MA) in tubes with ceramic beads using Precellys 24 homogenizer. Samples were lysed for 10 minutes on ice and centrifuged at 10,000 × g for 10 minutes. Supernatants were collected, and protein concentration was measured with the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Concentration was adjusted to add 20μg of protein per sample. Sodium dodecyl sulfate–10% beta mercaptoethanol was added to the samples, and they were denatured at 95°C for 5 minutes and agitated at 900rpm in a thermo shaker. Samples were resolved in a 4 to 12% Bis-Tris gel (Invitrogen, Carlsbad, CA) at 200V for 45 minutes and then transferred to a polyvinylidene fluoride (PVDF) membrane (iBlot PDVF stack; Thermo Fisher Scientific) with an iBlot system (Thermo Fisher Scientific) for 7 minutes. Membranes were blocked in 5% skimmed milk–phosphate-buffered saline (PBS) for 1 hour and incubated at 4°C with anti-JCV capsid protein VP1 antibody (8E8; ab34756; Abcam, Cambridge, MA) and antivinculin antibody (Abcam; 1:1,000 in Tris-buffered saline, 5% albumin, 0.02% sodium azide) overnight. Blots were washed 3 times in PBS–0.05% Tween-20 for 5 minutes and incubated with antimouse and rabbit–horseradish peroxidase (HRP) antibodies (Cell Signaling Technology, Danvers, MA) for 1 hour (1:2,500 in PBS–0.05% Tween, 1% skimmed milk). Blots were washed again 3 times with PBS–0.05% Tween. Chemiluminescence was developed with the Supersignal West Dura extended duration signal kit (Thermo Fisher Scientific), and the blots were imaged in a Protein Simple (San Jose, CA) imager. Chemiluminescence was quantified with ImageJ software (NIH, Bethesda, MD).

Immunohistochemistry

Formalin-fixed paraffin embedded tissue was cut into 5μm sections and placed on slides. Slides were baked at 60°C followed by xylene and dehydration with graded alcohols for deparaffinization. Antigen recovery was performed in heat-activated antigen retrieval buffer of pH 6.0 (Dako, Carpinteria, CA) using a steam pressure cooker (Pascal, Dako). Slides were moved to an automated immunohistochemical stainer (Dako), where the endogenous peroxidase activity was blocked with 3% H₂O₂ for 10 minutes followed by staining with a polyclonal VP-1 antibody (Abcam) at 1:250 dilution for 1 hour. The sections were then incubated with EnVision⁺ Dual Link System-HRP (Dako) for 30 minutes, visualized with DAB⁺ (3, 3′-diaminobenzidine, Dako) for 10 minutes, counterstained with hematoxylin, and analyzed using light microscopy. Negative controls included mouse immuno-globulin G isotype control, and omission of the primary antibody was performed. An immortal SV40 infected podocyte cell line was used as a positive control.

Transcription Factor Binding Site Analysis

Reference sequence JCV GH-3 (AB048545) and patient sequence (JCV MDaV) were imported into Genomatix. Within this program, MatInspector was used to search for all known transcription factor binding sites in the given sequences. Data output from the tool was constructed with each sequence annotated with the binding matrix sites observed and imported into DNASTAR Lasergene to visualize.

Results

Clinical Course

A 16-year-old Brazilian girl with refractory aplastic anemia underwent allogeneic stem cell transplant at the NIH Clinical Center (March 17, 2015). She received unrelated umbilical cord blood combined with haploidentical CD34⁺ progenitor cells from a relative. Her post-transplant course was complicated by Epstein–Barr virus reactivation requiring rituximab (May 25, 2016) and late onset acute skin graft-versus-host disease treated with multiple immunosuppressive therapies, including low-dose corticosteroids and ruxolitinib by November 2016. Within 2 months of starting ruxolitinib, she was admitted for respiratory syncytial virus upper respiratory tract infection and a seizurelike episode. Electroencephalography (EEG) did not show seizure activity. Computed tomography (CT) of the head was normal; cerebrospinal fluid (CSF) showed 1 white blood cell/mm³ and elevated protein (98mg/dl; upper limit of normal = 45mg/dl). JCV in the CSF was 445,490 copies/ml (Fig 1). Ruxolitinib therapy was discontinued. Concurrent hemophagocytic lymphohistiocytosis (HLH) was diagnosed due to elevated IL-2R (3,230pg/ml), elevated ferritin (59,326 mcg/L), pancytopenia, and persistent fevers requiring additional immunosuppressives. MRI of the brain was unremarkable. Over 3 weeks, no CSF pleocytosis was seen, protein serially increased to 313mg/dl, and JCV in the CSF increased to 1,779,922 copies/ml. EEG showed progressive cortical signal suppression (not shown). Three weeks after the initial neurologic event, CT of the head showed global cerebral edema with herniation.

FIGURE 1:

Timeline of clinical course and postmortem JC virus (JCV) findings. (A) Serial brain imaging over hospital course, showing in vivo imaging and time-course of lab results: (i) computed tomography (CT) of the head with no pathologic changes, (ii) T2 fluid-attenuated inversion recovery magnetic resonance imaging (MRI) with no obvious abnormalities, (iii) CT of the head demonstrating global cerebral edema and loss of gray–white differentiation, (iv) cerebrospinal fluid (CSF) JC viral load, (v) plasma ferritin, and (vi) fever course throughout hospitalization. (B) Timeline of immunosuppressive therapies received by the patient from prior to hospitalization until death. MTX is defined as methotrexate (C) Postmortem brain with evidence of diffuse edema and tonsillar herniation (arrow). (D) Postmortem 7T MRI T2*-weighted imaging demonstrating hazy hyperintensity (arrows) in the periventricular white matter, likely due to diffuse brain edema. (E) Electron microscopy of postmortem frontal lobe tissue demonstrating ~40nm viral inclusions within the endoplasmic reticulum, characteristic of polyomavirus infection.

Postmortem MRI and Histopathology

Postmortem 7T MRI identified no white matter abnormalities to suggest PML (see Fig 1D), and histopathology showed positive VP1 staining in hippocampal and cortical neurons as well as in cortical glial and endothelial cells (Fig 2). VP1 was also detected by Western blot analysis in brain samples dissected from the occipital and frontal lobes. VP1 levels were higher in gray than in white matter. JCV load per milliliter of CNS autopsy tissue surpassed 1 billion copies in all sampled areas. Electron microscopy of postmortem frontal lobe tissue was consistent with polyomavirus infection (see Fig 1E). Postmortem autopsy tissue demonstrated positive JCV VP1 staining in cortical neurons, hippocampal neurons, and glial and endothelial cells (Fig 3).

FIGURE 2:

Postmortem JC virus (JCV) distribution and sequencing data. (A; i) JCV GH-3 isolate AB048545.1 sequence; (ii) JCV Mad1 isolate NC_001699.1 sequence; (iii) patient frontal lobe viral Sanger sequencing consensus strand demonstrating a 12bp insertion in the noncoding mRNA viral transcripts at GH-3 viral position 2584 in autopsy samples; (a–d) chromatograph of insertion in corresponding transcripts from patient postmortem (a) frontal lobe, (b) occipital lobe, (c) meninges, and (d) dura.(B) Next generation sequencing Illumina MiSeq 2×150 1590 matched reads mapped against archetype JCV GH-3 isolate AB048545.1 from autopsy cerebellum sample flash-frozen and stored at −80°C. (C) JCV load per autopsy tissue region in copies/ml. (D) Left: Image of a Western blot stained with an antibody against JCV VP1 protein on 6 different brain postmortem patient samples (3 from occipital lobe and 3 from frontal lobe). In the occipital lobe, 1 sample corresponded to gray matter, 1 to white matter, and the other was a mixture of both. A positive control from a case of progressive multifocal leukoencephalopathy (PML) was included. Right: Quantification of the VP1 protein in gray and white matter in the occipital lobe samples.

FIGURE 3:

JC virus VP-1 immunohistochemical staining of neuronal, glial, and endothelial cells. VP-1 immunohistochemical staining at ×40 magnification on patient brain postmortem sections of cortex and hippocampus demonstrates neuronal staining. (A) Positive control of SV40 infected podocyte cell line. (B) Negative control with secondary antibody only in cortical gray matter. (C, D) VP-1 staining of hippocampal and cortical neurons, respectively. (E, F) VP-1 staining in glial (green arrows) and endothelial cells (blue arrows) in cortical white matter and hippocampus, respectively.

Postmortem Sequencing

Next generation deep sequencing of brain tissue identified 3 predominant archetype JCV strains, and Sanger sequencing confirmed a novel insertion in the noncoding portion of the early mRNA transcript (see Fig 2A). Within the noncoding early mRNA, there was a gain of 18 transcription binding factor sites in this unique viral sequence, 50% of which were found within or overlapping the 12bp inserted region (Fig 4).

FIGURE 4:

Expansion transcription factor (TF) binding sites in patient JC virus (JCV) strain due to noncoding region 12bp insertion. Comparison of TF binding site differences between JCV_MdAV and reference JCV_GH3 in the region of 12bp insertion sequence uses Genomatix and MatInspector analysis. (A) Left: Patient sequence JCV_MdAV. Right: Reference archetype GH_3 sequence; red boxes represent new TF binding sites found in JCV_MdAV, and green demonstrates retained TF binding sites present in both patient and reference sequence. (B) New (red, A, left) TF binding sites present in noncoding region between bp 2585 TCTTTACTTTTT and 2596 that map to the 12bp insertion in the patient JCV_MdAV sequence.

Discussion

JCV encephalopathy is a unique and rare manifestation of JCV infection that has not previously been associated with ruxolitinib. To our knowledge, this is the first fatal case of JCV encephalopathy associated with ruxolitinib and HLH immunosuppressive therapy. The clinical course consisted of progressive cortical suppression culminating in global cerebral edema with postmortem 7T MRI identifying no white matter lesions typical of JC viral infection. Histology confirmed JCV VP1 predominantly in glial cells, and hippocampal and cortical neurons. These results suggest a form of JCV encephalopathy without white matter lesions, even in advanced disease, which has not previously been reported.

Sequencing from multiple tissue compartments demonstrated a unique 12bp insertion in a noncoding region in the intergenic area between VP1 and large T antigen, not previously reported in CNS JCV. During transcription, this 12bp insertion is present in the early mRNA. Although sequencing in coding regions for major JCV proteins VP1 and large T antigen from this strain (termed JCV_MdAV) was found to be 99% identical to other reported JCV archetype strains, there was a gain of 18 transcription binding factor sites in this unique viral sequence, 50% of which were found within or overlapping the 12bp inserted region. This may explain the extraordinarily high levels of the virus and the expanded cell tropism in the brain. JCV encephalopathy has not previously been associated with JAK-STAT pathway modulation, although there are reported PML and JCV meningitis associations with ruxolitinib.^4,5

JAK1/2 kinase inhibitors are potent modulators of interferon-γ, which is known to suppress JCV reactivation. This might provide one explanation for the possible increased susceptibility to JCV CNS infection in this population, and in patients who are highly immune suppressed viral reactivation may allow for the production of unique viral mutations, including noncoding regions not previously known to be pathogenic, such as in this case, that allow for changes in viral transcription that increase pathogenicity through atypical manifestations of JCV such as JCV encephalopathy. Atypical JCV infection in the CNS is rare, and this case highlights the necessity of CSF JCV assessment in patients on ruxolitinib or combination immunosuppressive therapy who have clinical neurologic changes without MRI abnormalities.