Antibodies to Enteroviruses in Cerebrospinal Fluid of Patients with Acute Flaccid Myelitis

The presence in cerebrospinal fluid of antibodies to EV peptides at higher levels than non-AFM controls supports the plausibility of a link between EV infection and AFM that warrants further investigation and has the potential to lead to strategies for diagnosis and prevention of disease.

The cause of AFM remains elusive. An infectious agent is only rarely detected in cerebrospinal fluid (CSF). In the initial CDC investigation, the presence of EV RNA by reverse transcription-PCR (RT-PCR) was demonstrated in the CSF from only 1 of 55 children with confirmed AFM. However, more than 40% of children with AFM had PCR evidence of EV RNA in respiratory or fecal samples, suggesting a possible role for EV in AFM (4). Failure to detect viral RNA in CSF by PCR may represent absence of virus, low levels of viral template, or sequence mismatch between viral template and primers or probes. Unbiased high-throughput sequencing can address the challenge of sequence mismatch; however, it is less sensitive than real-time PCR because abundant host RNA competes with viral RNA in sequencing reactions. Another alternative is the VirCapSeq-VERT, a positive-selection high-throughput sequencing system that detects viral genetic material with sensitivity equivalent to real-time PCR (5). A third alternative is to focus on immunological responses to pathogens that may not be present in the materials sampled (6,7). In this study, we used both viral-capture high-throughput sequencing (VirCapSeq-VERT system) and peptide microarray approaches to examine the role of EV in AFM, using specimens collected from AFM patients and three control groups of patients with either non-AFM central nervous system (CNS) or inflammatory illnesses.  All samples were deidentified and met criteria for nonhuman subject research at Columbia University and the University of California, San Diego. For samples contributed by CDC, the work was determined by the CDC National Center for Immunization and Respiratory Diseases to constitute public health surveillance rather than human subject research. Y, yes; N, no; Neg, negative; NA, not available; PML, progressive multifocal leukoencephalopathy. b Combined results of PCR-based EV testing performed by the CDC and CII of Columbia University.
CSF Enterovirus Antibodies in Acute Flaccid Myelitis ® peptides with 11 overlapping amino acids (aa) that span the capsid proteins (VP1, VP2, VP3, and VP4) of all human EVs (species EV-A, EV-B, EV-C, and EV-D) and the polyprotein of all West Nile viruses (WNVs), using sequences available from the National Center for Biotechnology Information (NCBI) through 31 January 2019 (Table 3; see Table S1 in the supplemental material). The AFM-SeroChip-1 also contains 1,000 random 12-aa peptides to detect nonspecific binding. Fourteen CSF samples from AFM cases and 27 CSF samples from the 3 control groups (NAC, KDC, and AC) were tested for IgM and IgG antibodies to EV and WNV using AFM-SeroChip-1. Sera were available for testing from only a subset of these 14 subjects (Table 1). CSF samples from the 14 individuals with AFM and 6 members of the NAC group were also tested on arrays that detect antibodies to 8 tick-borne pathogens present in the United States, including Anaplasma phagocytophilum (agent of human granulocytic anaplasmosis), Babesia microti (babesiosis), Borrelia burgdorferi (Lyme disease), Borrelia miyamotoi, Ehrlichia chaffeensis (human monocytic ehrlichiosis), Rickettsia rickettsii (Rocky Mountain spotted fever), Heartland virus, and Powassan virus (7). Antibodies to WNV and tick-borne pathogens were surveyed due to community concerns that these agents might be implicated.

Identification of immunoreactive conserved peptides in the VP1 capsid proteins of enterovirus (EV-A to EV-D).
High-density peptide microarrays were used to detect immunoreactivity against a comprehensive EV capsid proteome. We identified an 18-aa region (7 overlapping 12-mer peptides from each EV species) ( Fig. 1) in the VP1 capsid proteins of EV-A, EV-B, EV-C, and EV-D that was immunoreactive with both serum and CSF from AFM patients ( Fig. 2; see Fig. S1 in the supplemental material). This region is highly conserved among EVs and is typically used to infer broad EV immunoreactivity (8,9). It includes the motif PALXAXETG, with the exception of EV-D, where PSL is substituted for PAL ( Fig. 1). CSF samples from 11 of 14 AFM patients (79%) were immunoreactive to this region, compared to those from only 1 of 5 (20%) NAC patients, 0 of 10 (0%) KDC children, and 2 of 11 (18%) AC patients (P ϭ 0.023, 0.0001, and 0.0028, respectively).
One CSF sample from a member of the NAC group, who had an acute CNS infection with E-25, was also immunoreactive to this conserved EV region (Tables 1 and 2). Antibodies to EV peptides were found in CSF from 2 of 11 AC patients (NC-A-25 and   S1). Immunoreactivity against an EV-D68-specific VP1 peptide in AFM patients. Although 70% (18/26) of both AFM and control sera were immunoreactive with at least one VP1 EV-D conserved peptide (Fig. S1), we identified a 22-aa linear peptide sequence in the carboxyl terminus of the EV-D68 VP1 protein that was immunoreactive only in AFM patients (Fig. 1). In AFM patients, this peptide was immunoreactive with 6 of 14 (43%) CSF samples and 8 of 11 (73%) serum samples (Fig. 3). In contrast, this peptide was not immunoreactive with CSF or serum samples from 5 NAC, 10 KDC, or 11 AC patients. Statistical analysis indicated differences between AFM versus NAC, KD, and NC controls for both CSF (P ϭ 0.008, 0.0003, and 0.035, respectively) and serum (P ϭ 0.139, 0.002, and 0.009, respectively).

DISCUSSION
Since the summer of 2014, Ͼ560 cases of AFM have been confirmed in the United States. The uncertainty underlying the cause for the majority of AFM cases precludes important public health interventions such as development of diagnostics, specific therapy, and preventive measures. Growing evidence suggests that infections with EV, including EV-D68 and EV-A71, may contribute to AFM (4,10). A potential link to EV-D68 and EV-A71 was proposed based on (i) the presence of viral RNA in some respiratory and stool specimens, (ii) the observation that EV-D68 infection can result in spinal cord infection, necrosis, and paralysis in mice (11,12), and (iii) the temporal association of respiratory outbreaks of EV-D68 with outbreaks of AFM (4). However, EVs have only rarely been detected in CSF of AFM patients (4 of 567 total confirmed cases studied at the CDC [https://www.cdc.gov/acute-flaccid-myelitis/afm-surveillance.html]).
The potential link to EV, combined with the inability to directly detect its nucleic acids, encouraged us to search for indirect evidence of EV infection in CSF by testing for the presence of EV-specific antibodies. Since EV infection is common in children, traditional immunoassays such as enzyme-linked immunosorbent assay (ELISA) are impeded by cross-reactivity. Virus neutralization assays are considered the "gold standard" but are resource intensive and impractical to implement for Ͼ100 types of EV. Accordingly, we employed a microarray system that can simultaneously screen thousands of peptides from all known human EVs with high resolution. Across the entire EV capsid proteome tiled with 160,000 peptides, the conserved EV VP1 peptides were consistently immunoreactive in CSF of 79% (11/14) of children with AFM, the majority of whom had no EV RNA in CSF. In comparison, NAC patients without evidence of EV RNA in any specimen and KDC patients did not have antibodies to EV peptides in CSF. This conserved peptide region reflects broad EV immunoreactivity, suggesting an association with EV infection and AFM. It is also consistent with EV pathogenesis in mouse models, wherein EVs, including EV-D68, can cause paralytic illness (11,12).
The demonstration that antibodies to EV peptides are intrathecally produced, such as using an intrathecal ratio of CSF to sera, would provide stronger evidence of CNS infection. We do not have access to sufficient numbers of simultaneously collected CSF and serum specimens to exclude trafficking of EV antibodies from the peripheral circulation into the CNS. However, we consider this unlikely to explain the presence of antibodies to EV peptides in CSF of AFM patients. Of 16 control patients who had antibodies to EV peptides in serum (3 NAC, 8 KDC, and 5 AC patients), only 2 adults and 1 child also had antibodies against EV peptides in CSF. The child with antibodies against EV peptides in CSF had an active CNS infection, with E-25 detected in the same CSF sample. One of the 2 adults had multiple sclerosis, and the other had idiopathic epilepsy. Furthermore, in a study of unexplained encephalitis in Gorakhpur, India, using We demonstrated immunoreactivity to EV-D68-specific peptides (located within the VP1 C terminus) in CSF from 43% of AFM patients and no immunoreactivity in any of the CSF samples from NAC, KDC, and AC control groups. Based on structural homology, the VP1 C terminus of the EV-D68 strain has been predicted to form neutralization immunogens and conformational epitopes for putative neutralizing antibodies (13,14). The same EV-D68-specific peptides were immunoreactive in 8 of 11 (73%) sera from AFM patients, compared to none from the three control groups (NAC, KDC, and AC). These findings may appear to be at odds with seroprevalence studies in children and adults in the United States that indicate near 100% infection rates using neutralization assays (13,15). One potential explanation is that neutralization assays may detect additional nonlinear conformational epitopes, whereas peptide arrays present only linear epitopes. Reactivity from one peptide region may not be comparable to that of the neutralization assay using the whole virion. In work with sera from children exposed to Zika virus, antibodies to the NS2B protein were specific but were not detected at 6 months in 45% of documented infections (6). Understanding why this EV-D68-specific linear peptide is highly reactive in AFM cases but not in controls may shed light on the immune response to EV-D68.
Our serological findings warrant further investigation into the association of AFM with enteroviruses, including EV-D68. Reports of respiratory illness consistent with viral infection typically occurred several days before onset of weakness in the majority of AFM cases. Our inability to detect EV-D68 RNA in the CSF of AFM patients may reflect lack of virus shedding from CNS parenchyma into the intrathecal compartment. Alternatively, CSF specimens may have been collected after virus was cleared from the CNS.
An important limitation of our study is that controls were not optimally matched to the AFM cases with respect to age, year, or season of collection. Although we tried to test for specificity of findings through use of three control groups, prospective studies with appropriately matched controls from the same EV outbreak season will be critical to testing their validity. Conclusions. While other etiologies of AFM continue to be investigated, our study provides further evidence that EV infection may be a factor in AFM. In the absence of direct detection of a pathogen, antibody evidence of pathogen exposure within the CNS can be an important indicator of the underlying cause of disease. Our findings warrant additional investigation, including testing more specimens, developing confirmatory assays, and designing optimally controlled prospective studies. These initial results may provide avenues to further explore how exposure to EV may contribute to AFM as well as the development of diagnostic tools and treatments.

MATERIALS AND METHODS
Virologic testing for AFM surveillance. In 2018, CDC received specimens from 381 patients suspected to have AFM (https://www.cdc.gov/acute-flaccid-myelitis/afm-cases.html). Classification criteria for a confirmed AFM case are acute flaccid limb weakness and magnetic resonance imaging evidence of a predominantly gray matter lesion that spans at least one spinal segment (https://www.cdc.gov/ acute-flaccid-myelitis/hcp/case-definition.html) (16). Suspect cases were considered non-cases if they failed to meet the confirmed case definition or had an alternative diagnosis to explain their symptoms. As part of diagnostic testing for AFM surveillance, CDC tested CSF, respiratory, and fecal specimens from suspected AFM cases for EV using a real-time reverse transcription-PCR (rRT-PCR) EV assay (17) and an EV typing assay by VP1 seminested PCR and Sanger sequencing (8). Specimens were also tested for parechoviruses using rRT-PCR (18). Respiratory specimens were further tested using an EV-D68specific rRT-PCR assay (https://www.fda.gov/medicaldevices/safety/emergencysituations/ucm161496 .htm#enterovirus).
Patients and specimens. For investigation using VirCapSeq-VERT and peptide microarray approaches, we selected a subset of samples (based largely on available CSF volumes) from AFM cases with onset in 2018, including 13 children and 1 adult. Three groups of controls (Table 1) (19,20), and (iii) 11 adults with non-AFM CNS diseases (adult control [AC] group). One additional NAC case from 1 child with autoimmune encephalitis (AC-KT-076) during 2019 was also included. NAC and AC controls were included to examine the possibility that antibodies in the peripheral circulation might traffic into the CSF in children and adults with CNS diseases who may have increased permeability of the blood-brain barrier. Paired CSF and serum samples were available for 11 children with AFM, 2 children from the NAC group, 8 children from the KDC group, and 5 adults from the AC group. As an internal positive control for VirCapSeq-VERT performance, we included two RNA extracts from stored CSF specimens of paralytic type 1 wild poliomyelitis cases from Republic of Congo in 2010 (21).
VirCapSeq-VERT analysis. The VirCapSeq-VERT capture probe library comprises approximately 2 million oligonucleotides that tile the coding regions of genomes for all viral taxa that contain at least one virus known to infect vertebrates (5). Total nucleic acid was extracted from 240 l of CSF using the NucliSens easyMAG automated platform (bioMérieux, Boxtel, The Netherlands). Illumina libraries were prepared, pooled, and hybridized with the VirCapSeq-VERT probe set prior to sequencing (Illumina HiSeq 4000) (5,22). Human genomic and ribosomal sequences were subtracted, and the remaining sequences were analyzed for viral sequences using MegaBlast and BLASTX against the GenBank nonredundant nucleotide and protein databases, respectively.
Assessment of VirCapSeq-VERT efficiency in detection of enterovirus RNA. To determine the efficiency of VirCapSeq-VERT in EV genome recovery, we quantitated the amount of EV template in CSF by quantitative reverse transcription-PCR (qRT-PCR). A one-step qRT-PCR assay that targeted a conserved region within the 5= untranslated region (UTR) was performed on all CSF samples. Ten microliters of total nucleic acid extract was reverse transcribed using Superscript III (Life Technologies). The qRT-PCR assay was performed using 2 l of cDNA, EV-F primer (TCCTCCGGCCCCTGAATGYGGCTAAT), EV-R primer (GGAAACACGGWCACCCAAAGTA), and EV-probe (6-carboxyfluorescein [FAM]-GCAGCGGAACCGACT-MGB) in a Bio-Rad Touch-CFX 96 real-time PCR instrument at 95°C for 2 min followed by 45 cycles of qPCR at 95°C for 15 s and 60°C for 1 min.
SeroChip analyses. Samples were diluted in binding buffer (0.1 M Tris-Cl, 1% alkali soluble casein, 0.05% Tween 20, and water) and hybridized on the AFM array overnight (16 h) at 4°C on the flat surface. After incubation, arrays were washed 3 times for 10 min each on a Little Dipper processor (SciGene; catalog no. 1080-40-1) with 1ϫ TBST (Tris-buffered saline plus 0.05% Tween 20) at room temperature. Secondary antibodies were diluted to 1:5,000 in binding buffer at a concentration of 2 g/ml. Secondary antibody incubation was done in a plastic Coplin jar (Fisher Scientific; catalog no. S90130) for 3 h at room temperature with gentle shaking on a rocker shaker. First, arrays were incubated with Alexa Fluor 647 AffiniPure goat anti-human IgG that was Fc␥ fragment specific (Jackson ImmunoResearch Labs, Inc.; catalog no. 109-605-098), followed by three 10-min washes. Then arrays were incubated with AffiniPure goat anti-human IgM that was Fc5 fragment specific (Jackson ImmunoResearch Labs, Inc., no. 109-005-043). Incubation with secondary antibody was followed by three 10-min washes with 1ϫ TBST at room temperature and spin-drying. The array slides were scanned on a NimbleGen MS 200 microarray scanner (Roche) at a 2-m resolution, with excitation wavelengths of 635 and 537 nm simultaneously. In the images recorded with the laser scanner, the relative fluorescent unit (RFU) signals for all the probes were extracted using Roche Sequencing Solutions image extraction software. The RFU signals for all the probes were extracted from the images using Roche Sequencing Solutions image extraction software. The RFU signals were converted into intensity plots after quantile normalization, background and spatial correction, and deconvolution for redundant peptides. An epitope was considered reactive when the signal intensity for at least three contiguous 12-mer peptides was above the threshold. The signal threshold was defined for each reactive epitope by calculating the mean plus 3 standard deviations of the signal intensity for the same epitope in the negative-control samples.
The array data, in the form of relative fluorescence signal intensities from the scanned images (arbitrary units [AU]), were spatially and background corrected and quantile normalized. Signal data points were filtered to retain only immunoreactive peptides that showed the mean ϩ2 standard deviations (Ͼ10,000-AU signal for serum and Ͼ3,500-AU signal for CSF based on measured values for reactivity against scrambled nonspecific peptide sequences). A region was considered immunoreactive when the signal intensity for at least three contiguous 12-mer peptides was above the threshold. The signal threshold was defined for each immunoreactive epitope by calculating the mean ϩ2 standard deviations of the signal intensity for the same epitope in negative-control samples (6,7). Immunoreactive peptides were compared with a comprehensive EV protein alignment to scan for type-specific signals.
Assessment of immunoreactivity signal intensity from the high-density peptide microarray among cases and controls. For the EV VP1 conserved peptide and EV-D68-specific peptide sequences, comparison of immunoreactivity signal intensity between AFM, non-AFM, Kawasaki disease, and adult CNS groups were evaluated using the Wilcoxon test for paired samples (JMP, v13.0.0; SAS Institute, Inc., Cary, NC).