Immune and Epstein-Barr virus gene expression in cerebrospinal fluid and peripheral blood mononuclear cells from patients with relapsing-remitting multiple sclerosis

Patients were recruited at the MS centers of the University Hospital San Luigi Gonzaga in Orbassano, University of Cagliari, and University of Firenze. The study was approved by the ethic committees of the three participating MS centers and of the Istituto Superiore di Sanità, and carried out according to the Declaration of Helsinki. Written informed consent was obtained from all study participants.

Thirty-one patients with RRMS were included in this study [45, 46]. None of the patients received immunomodulatory or immunosuppressive treatment at the time of CSF and blood sample collection and had not received such treatments for at least 12 months. Demographic and clinical information were derived from medical records and are summarized in Table 1. MS disease onset was defined as the first episode of focal neurological dysfunction indicative of MS; relapses were defined as the development of new or recurrent neurological symptoms not associated with fever or infection and lasting for at least 24 h [46]. Patients were categorized on the basis of the presence of a relapse or a condition of remission at the time of sampling. The expanded disability status scale (EDSS) score was calculated on the basis of a complete neurological examination by a neurologist expert in MS.

All patients were examined by a routine brain MRI protocol before or after sample collection (median time interval =26 days; range 1–90 days). The time interval between sample collection and MRI was significantly shorter in patients in clinical relapse (median =11 days, range 1–37 days) than in patients in clinical remission (median =42 days, range 2–90) (p = 0.0015 by Student’s t test). MRI scans (T2-weighted and T1-weighted pre- and post-gadolinium administration, slice thickness 5 or 3 mm) were obtained in all patients using a standardized scanning protocol [47] with 1.5 T MR scanners.

All CSF and peripheral blood samples were obtained for routine diagnostic work-up. CSF and blood from each patient were always drawn on the same day. CSF samples were processed according to the BioMS-eu consortium guidelines [48]. A total of 3 to 17 ml of CSF (median =11 ml) were obtained by lumbar spinal tap with an atraumatic needle. Within 30 min after lumbar puncture, CSF samples were centrifuged at 1200 rpm for 10 min at room temperature to separate the cellular component from cell-free supernatant; the cell pellets were stored at −80 °C in RNAlater (Qiagen) or RNA was immediately extracted (see below). Blood (10 ml) was drawn in EDTA tubes, and PBMC were isolated using Lymphoprep, preserved in RNAlater and frozen at −80 °C. CSF samples were routinely analyzed for cell counts. Quantitative (IgG index) and qualitative (oligoclonal bands) analysis of intrathecal IgG synthesis after lumbar puncture was performed using standard methods.

Total RNA was extracted from CSF cells (median =4.5 × 10⁴, range 7 × 10³–5 × 10⁵) and PBMC (4 × 10⁵) using the AMBION RNAqueous micro kit (Life Technologies, Grand Island, NY, USA) according to the manufacturer’s instructions, including genomic DNA digestion. Total RNA from PBMC was quantified by Nanodrop 2000 (Thermo Fisher Scientific, Waltham, MA, USA) and 200 ng was reverse transcribed for each sample. Because of the very low and highly variable RNA yield from CSF cells, the entire volume (15 μl) of RNA extracted from each CSF sample was reverse transcribed. Reverse-transcription (RT) was performed using the high capacity reverse transcription kit with RNase inhibitor (Life Technologies). The resulting cDNA was diluted to a final volume of 50 μl and splitted into four 12.5 μl aliquots. To increase the number of targeted copies, each cDNA aliquot was amplified for the specific gene assays by pre-amplification reaction (14 cycles) using the TaqMan PreAmp Master Mix (Life Technologies) and pooled gene-specific primers, and the reaction conditions indicated by the manufacturer. Inventoried and self-designed TaqMan gene expression assays were used to study cellular and EBV genes, respectively (see Additional files 1 and 2). Cellular gene assays were pre-amplified together with the housekeeping gene GAPDH; viral gene assays were pre-amplified separately together with GAPDH and the B-cell-specific genes CD19 and CD20. The pre-amplification product was diluted 1:5 up to 250 μl in TE buffer, and 4 μl of this dilution was used as template for a single real-time PCR analysis. Quantitative PCR experiments were performed in triplicates with the same inventoried or self-designed TaqMan assays used in the pre-amplification step (250 nM probe and 900 nM each primer), using the 7500 Real-Time PCR System (Life Technologies) for cellular genes and the StepOne Plus Real-Time PCR System (Life Technologies) for viral genes. Thermocycling parameters were 50 °C (2 min), 95 °C (10 min), followed by 40 cycles of 95 °C (15 s) and 60 °C (1 min) for both cellular and viral genes. The results of gene expression analysis are expressed as Ct values (Ct = threshold cycle of PCR at which the amplified product is detected). The ΔCt is the difference in Ct values derived from the gene of interest and the reference gene GAPDH; the factor 2^-ΔCt is used to express the ratio between the gene of interest and the internal reference gene. To rule out cross-contamination of reagents and primers, all RT, pre-amplification, and real-time PCR experiments included a NTC sample, containing all the components of each reaction except for the template. Considering that 12 μl of pre-amplified cDNA was analyzed for each transcript and that the available volume of each pre-amplified aliquot was 250 μl, we were able to analyze in triplicates up to 20 transcripts per aliquot.

To check that all amplicons were amplified uniformly without bias, we performed pre-amplification uniformity experiments using non-limiting cDNA from a human non pathological pulmonary hilar lymph node (obtained from Dr. Egidio Stigliano, Institute of Pathological Anatomy, Policlinico A. Gemelli, Rome, Italy), as control for cellular genes, and from an EBV transformed B-lymphoblastoid cell line (LCL), as control for EBV genes. The EBV+ LCL (L5) was generated by infecting 5 × 10⁶ PBMC obtained from a patient with MS with B95.8 EBV strain in a medium containing cyclosporin A (1 μg/ml, Calbiochem); the outgrowth of B95.8-infected PBMC was monitored twice a week, and after 5 weeks post-infection, the LCL was permanently established. Amplification of pre-amplified cDNA from lymph node and EBV+ LCL was compared with that of non pre-amplified cDNA. Primer uniformity was calculated by the formula ΔΔCt = ΔCt (Preamp) − ΔCt (cDNA). A ΔΔCt value within ±1.5 is considered acceptable, as indicated in the manufacturer’s instructions. Pre-amplification uniformity values related to the reference gene GAPDH were very close to zero for all the investigated gene assays (mean ΔΔCt values ± SD were 0.33 ± 0.35 and 0.90 ± 0.41 for EBV and cellular transcripts, respectively), indicating optimal pre-amplification uniformity. PCR efficiency by direct and PreAmp real-time PCR was checked for viral genes and found to be similar (range 0.97–1.08; optimal efficiency = 100 ± 10 %) over serial dilutions of EBV+ LCL cDNA (from 100 to 0.1 ng, corresponding to approximately 10.000 to 1 cells). Importantly, similar data were obtained for each target gene after pre-amplification from pooled and single assays.

Demographic, clinical, radiological, CSF, and gene expression data of 31 MS patients were analyzed by univariate and multivariate statistical techniques. In univariate analyses, comparisons between groups of patients were carried out by Student’s t test and Mann-Whitney test for continuous variables, and by Fisher’s exact probability test for categorical variables. Correlation between variables was assessed by Spearman’s rank correlation coefficient. Means and SE, or medians and interquartile ranges, were used to summarize continuous data, and percentages were used for categorical variables.

To unravel complex gene interactions that may better capture pathological processes in MS, the collected gene expression data were analyzed using two multivariate statistical techniques: cluster analysis aiming to group subjects into clusters and factor analysis aiming to define artificial factors, that is underlying latent variables, adequately describing the correlation structure of the original variables. Cluster analysis was carried out by average linkage method with Euclidean similarity measure. Clustering of patients was visualized by dendrogram and the choice of number of groups was based on Calinski/Harabasz pseudo-F index and Duda/Hart index stopping rules. Factor analysis was carried out using the principal factor method. Factor loadings, that is, correlations of the original variables with factors, were used for interpretation of artificial factors. Scores of subjects on artificial factors were entered in further analyses. The influence of demographic, clinical, or radiological parameters on patient clustering and factor scores was investigated by univariate analyses. Finally, the discriminating power of artificial factors to predict patient clustering was assessed. Classification accuracy was assessed through receiver operating characteristic (ROC) curve analysis, by calculating the area under ROC curves (AUC) and its 95 % confidence interval (CI). These analyses were carried out separately on immune gene expression data obtained in CSF and PBMC samples.

The level of confidence was set at 0.05 and statistical significance was assessed by adopting the Bonferroni correction for multiple testing. Number of comparisons considered patient subgroups differing for demographic, clinical, and radiological characteristics, and resulting from cluster analysis. For correlation analyses, multiplicity due to two inflammatory CSF parameters (IgG index and CSF cell count) was considered. Stata 11 was used for statistical analyses.

In preliminary experiments, the TaqMan® PreAmp Master Mix technique was applied to cDNA obtained from human lymphoid tissue and an EBV+ lymphoblastoid cell line (L5), as positive controls for immune-related and viral genes, respectively. Robust pre-amplification of multiple-pooled Taqman ABI inventoried gene assays (41 cellular genes listed in Additional file 1) and self-designed gene assays (7 EBV genes listed in Additional file 2) was obtained with a mean improvement of 4.6 ± 0.4 cycles (range 4.1–5.1) (p < 0.0001 compared to the Ct values obtained with the direct real-time RT-PCR). Figure 1 shows the specificity and increased sensitivity of PreAmp RT-PCR and the lower limits of detection of four of the seven EBV gene expression assays tested in the EBV+ LCL.

PreAmp real-time RT-PCR was then applied to cDNA from CSF cells and PBMC collected from 31 therapy-free RRMS patients. Thirty-one CSF and 29 PBMC samples were eligible for RNA analysis. The demographic, clinical, and CSF characteristics of the MS cohort are presented in Table 1. Twelve patients (38.7 %) experienced a clinical relapse at the time of CSF and blood sampling, and 13 patients (41.9 %) had gadolinium enhancing lesions on the brain MRI.

The selected set of immune-related genes includes the following: T-lymphocyte (CD4, CD8, forkhead box P3 (FoxP3)), B-lymphocyte (CD19, CD20, CD138, and B-cell maturation antigen (BCMA), the receptor for B-cell activating factor (BAFF) and a proliferation-inducing ligand expressed on B-lineage cells), natural killer (NK) cell (CD56, NK cell p46-related protein (NKp46)), monocyte/macrophage (CD68) and plasmacytoid dendritic cell (pDC) (blood dendritic cell antigen 2 (BDCA-2)) markers; granzyme B and perforin, the lytic enzymes mediating T cell/NK cell cytotoxic activity; cytokines of T-cell and/or NK-cell origin (interferon-γ (IFN-γ), interleukin (IL)-2, IL-4, IL-5, IL-17A) and of predominantly macrophage/DC origin (tumor necrosis factor (TNF), IL-1β, IL-6, IL-10, p40 (cytokine subunit shared by IL-12 and IL-23), IL-15); major histocompatibility complex (MHC) class II, involved in antigen presentation; the enzyme MMP-9 involved in extracellular matrix and myelin degradation; the enzyme nicotinamide phosphoribosyltransferase (NAMPT) involved in inflammation, metabolic, and stress responses; COX-2, the enzyme responsible for prostaglandin synthesis at sites of inflammation; inducible nitric oxide synthase (iNOS), the enzyme involved in the generation of nitric oxide; the BAFF; and C-X-C motif chemokine 10 (CXCL10), an IFN-inducible chemoattractant for activated T cells and NK cells, and CXCL13, a B-cell chemoattractant; molecules involved in type-1 IFN production (interferon regulatory factor 7 (IRF7)) and binding (IFN-α-inducible protein 6 (IFN-αR1)) and induced by type-1 IFN (interferon-stimulated exonuclease gene 20 kDa (ISG20), myxovirus (influenza virus) resistance protein (MxA), 2′-5′-oligoadenylate synthetase 1 (OAS1), protein kinase R (PKR), IFN-induced protein with tetratricopeptide repeats (IFIT1), IFN-α-inducible protein 6 (IFI6), ubiquitin-specific peptidase 18 (Usp18)). Most of the investigated immune transcripts were detected in all or the majority (>80 %) of CSF and PBMC samples (Table 2). A few transcripts, like IL-17A, IL-4, p40, CXCL13, and iNOS, were detected in a lower percentage of CSF cell and/or PBMC samples; only IL-5 was always undetectable (Table 2).

To search for a link between the immune-related gene signatures and/or the patient clusters identified by multivariate statistical approaches and EBV infection status, we then evaluated expression of five EBV latent (EBV-encoded small RNA 1 (EBER1), EBV nuclear antigen (EBNA)1, EBNA3A, latent membrane protein (LMP)1, LMP2A) and two EBV lytic (BZLF1, gp350/220) genes in all the analyzable CSF cell and PBMC samples. EBV transcripts were detected in only a few samples: 3 of 31 CSF cell samples (9.7 %) and 4 of 29 (13.8 %) PBMC samples from 5 of 31 patients (16.1 %) (Table 7). All five EBV+ patients belonged to cluster 1, as defined by cluster analysis on CSF gene expression data; of these, three were clinically relapsing/MRI active and two were clinically remitting/MRI inactive (Table 7). Relative to GAPDH, the frequency of viral transcripts was hundred- to thousand-fold lower than that of CD19 and CD20 (pre-amplified in the same cDNA aliquot). It is worth noting that GAPDH Ct values in the three EBV+ CSF samples ranged between 17.3 and 18.2, while the median GAPDH Ct value in the MS cohort was 20.0 (range 14.2–27.7), with values >18.5 being detected in 71 % (22 of 31) of CSF samples (Table 7). This suggests that, despite enrichment of target genes by pre-amplification, low frequency viral transcripts could be missed in most CSF samples because of insufficient material.

The pattern of EBV gene expression observed in five MS patients was highly variable (Table 7). In PBMC from two female patients, one relapsing/MRI active (FI01) and one remitting/MRI inactive (TO41), detection of the untranslated EBV transcript EBER1, in the absence of any other viral latent gene product, suggests latency 0 program. This EBV gene program is expressed in infected circulating memory B cells and allows the virus to escape immune surveillance [43]. Of these two patients, only TO41 also displayed EBER1 in CSF cells together with the EBV latent gene LMP2A. Expression of LMP2A in the absence of EBV nuclear antigens suggests latency II or default program [43]. EBV latency activation (latency II program) manifesting as expression of LMP1, another gene encoding a latently expressed membrane protein, was found also in CSF cells of a relapsing/MRI active female patient (TO32) and in PBMC of a remitting/MRI inactive female patient (CA07) (Table 7). LMP1 and LMP2A play an important role in supporting the survival and differentiation of EBV latently infected B cells by mimicking CD40 and B cell receptor stimulation, respectively [56‐58]. Only a relapsing/MRI active male patient (FI13) displayed detectable levels of EBV genes that are expressed during viral reactivation (Table 7). BZLF1, which is associated with the immediate early lytic cycle and plays a key role in the switch from latent to lytic infection, was detected in CSF cells while gp350/220, which is associated with the late lytic cycle and encodes a major EBV envelope glycoprotein [43, 59], was detected in PBMC suggesting virion production (Table 7). In this patient, profound deregulation of EBV infection in PBMC was also revealed by the presence of EBNA-1 and EBNA-3A RNA (latency III or growth program) [43] and strikingly elevated levels of CD20 RNA (25-fold higher that the median value in the cohort), suggesting new infection events and B-cell activation, respectively (Table 7). It is worth noting that the only patient showing EBV reactivation was the same clustering separately from the rest of the cohort and displaying the highest factor 2 score based on immune-related gene expression data in PBMC (see Fig. 7). This finding provides an explanation for the strong association of factor 2 with genes encoding NK cell-related and cytotoxicity markers (CD56, NKp46, perforin, IFN-αR1), pro-inflammatory molecules (NAMPT, MHC class II, IL-1β, CXCL13) and, as mentioned above, the B-cell marker CD20 as an indicator of EBV-driven B-cell expansion/activation.