Metabolomic Analysis of the Cerebrospinal Fluid in Latent Syphilis and Neurosyphilis Patients

Purpose: The infection rate of syphilis continues to rise, and the diculty in detecting and treating neurosyphilis promptly needs to be resolved. The metabolic proles of cerebrospinal uid (CSF) of different patients were analyzed to understand the pathogenesis of syphilis better. Method: The metabolic proles of 88 CSF samples from patients were analyzed by UPLC-Q Exactive-MS. The experimental process was evaluated by PCA, PLS-DA, and HCA. T-test statistics were used to compare levels of metabolites to determine signicant differences between groups. Pathway analysis was based on the KEGG database. Result: In total, 272 metabolites based on 3937 features obtained in ESI- mode and 252 metabolites based on 3799 features in ESI+ mode were identied. A clear separation between latent syphilis and neurosyphilis was found. Levels of lipid and linoleic acid metabolites, such as 9-OxoODE and 9,10,13-TriHOME, were increased in syphilis patients. In patients with neurosyphilis, signicant changes in levels of 5-hydroxy-L-tryptophan (5-HTP) and acetyl-N-formyl-5-methoxykynurenamine (AFMK) in the tryptophan-kynurenine pathway were also detected. Only one metabolite, theophylline, differed signicantly between symptomatic and asymptomatic neurosyphilis patients. Additionally, KEGG analysis revealed signicant enrichment of tryptophan metabolism pathways, indicating a high correlation between tryptophan metabolism and syphilis symptoms. Conclusions: Levels of linoleic acid metabolites, 5-HTP, AFMK and theophylline were signicantly altered in different patients. The role of these differential metabolites in the development of syphilis is worthy of further exploration, probably improving the treatment and diagnosis of neurosyphilis and occult syphilis in the future.


Introduction
Syphilis is a sexually transmitted disease (STD) caused by Treponema pallidum associated with signi cant complications if left untreated and can facilitate the transmission and acquisition of HIV infection. Without effective treatment, T. pallidum usually invades the central nervous system and causes neurosyphilis [1]. Neurosyphilis used to be a complicated disease in clinical research. With the discovery of penicillin, the prevalence of neurosyphilis and syphilis was controlled [2]. However, since 2000, the number of cases of syphilis has begun to increase again, with more than 100,000 new cases in 2018 alone; by contrast, only approximately 60,000 new cases of gonorrhoea occurred in 2018 [3]. The number of patients with neurosyphilis has also risen with the resurgence of syphilis. Merritt et al. reported that approximately 30% of all patients developed neurosyphilis, of whom 30% were asymptomatic [4].
Latent syphilis and neurosyphilis are di cult to distinguish when the early clinical symptoms are not obvious, which is the best treatment time. Nevertheless, diagnosis is complicated, and there is no gold standard. A neurosyphilis diagnosis is mainly based on clinical manifestations, speci c and nonspeci c serological tests for syphilis, abnormal cerebrospinal uid (CSF), and occasionally neuroimaging [1,5].
Serology of neurosyphilis is usually detected by traditional methods, such as the nontreponemal test (NTT) using the Venereal Disease Research Laboratory (VDRL) or rapid plasma reagin (RPR), and treponemal tests, such as the T. pallidum agglutination test (TPPA), to con rm positive results [2,6]. Even this negative result does not entirely rule out neurosyphilis, as VDRL may be negative in 30% to 70% of cases [7]. Therefore, more speci c and sensitive diagnostic markers should be found.
Metabolomics, which can directly re ect altered metabolites and their interaction with various stimulating factors, has great potential in disease screening and diagnosis biomarker identi cation. CSF is a biological uid that is greatly affected by the dysfunction of the central nervous system, and analysis of CSF can well re ect changes in the nervous and biochemical states of the body [8,9]. Indeed, CSF metabolomics has been widely used in various neurodegenerative diseases and brain tumours, and studies have reported potential CSF diagnostic markers for Alzheimer's disease and malignant glioma [10,11]. Nevertheless, the use of CSF-based metabolites for a probable diagnosis of neurosyphilis remains mostly unexplored.
In this study, UPLC-Q Exactive-MS was conducted to analyze the metabolic pro les of CSF from 88 patients with syphilis, neurosyphilis, and non-syphilis, and more than 400 metabolites were identi ed in two modes. By evaluating the original data by multistatistical analysis, signi cant differences in the levels of CSF metabolites between syphilis and neurosyphilis were detected. T-test analysis showed that 20 metabolites differed between syphilis and non-syphilis patients, with ve between neurosyphilis patients and latent syphilis patients. Only one metabolite, theophylline, was signi cantly different between symptomatic and asymptomatic neurosyphilis patients. The result may reveal the potential molecular mechanisms associated with syphilis and neurosyphilis, which may promote the development of syphilis biomarkers.

Participants
Informed consent was obtained from all subjects in this study. All experiments were performed following the approved guidelines. The patients included those with latent syphilis (Y, n=39), those with asymptomatic neurosyphilis (W, n=9), and those with symptomatic neurosyphilis (Z, n=33) as well as non-syphilis patients (F, n=7). Brie y, neurosyphilis was de ned based on positive treponemal test results and the toluidine red unheated serum test (TRUSRT) [12,13]. Patients were considered to have symptomatic neurosyphilis if they have obvious clinical symptoms of neurosyphilis, such as meningitis, stroke, acute changes in mental status, abnormal hearing or vision; neurosyphilis patients with no neurological symptoms have asymptomatic neurosyphilis. Latent syphilis was de ned based on a positive serological test for syphilis but a negative CSF test and no neurological symptoms or classic syphilis symptoms. Non-syphilis patients were negative based on serological and CSF tests.

Sample preparation
Approximately 1 mL of CSF was collected from each participant via lumbar puncture. The samples were transported on ice immediately after collection, and a syphilis serological test was carried out. CSF proteins and white blood cells were counted within six hours after CSF collection and stored at -80℃ until use. CSF samples were thawed, and 100 μL of sample and 300 μL of methanol were transferred to a 1.5-mL Eppendorf tube and vortexed for 30 s. All samples were kept at -40°C for 1 h, vortexed for 30 s and centrifuged at 12000 rpm and 4°C for 15 mins. Next, 200 μL supernatant and 5 μL DL-ochlorophenylalanine (internal reference, 140 μg/mL) were transferred to vials for HPLC-MS analysis.

MS data processing and identi cation
Raw data were acquired and aligned using Compound Discover (version 3.0, ThermoFisher Scienti c, Waltham, MA, USA) according to the m/z value and the retention time of the ion signals. In the case of some large or small variables, normalization was often performed after alignment, and a line plot was used to evaluate the methodology. The con dence interval of 95% of the sample value was considered to be stable and feasible.
Normalized data were imported into the SIMCA-P program (version 14.1, Umetric, Umea, Sweden) for principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), and the calculation of variable importance in the projection (VIP). As an unsupervised pattern recognition method for processing metabonomic data, PCA can classify metabolic phenotypes according to all input samples. All data from differentially expressed compounds were used to build PCA models. Different colours and icons represent different groups. PLS-DA was employed to supervise regression modelling of the data set to identify potential biomarkers. The quality of the models was described by relevant R2 and Q2 values. R2 displays the variance explained in the model and indicated the quality of the t. Q2, as calculated by a cross-validation procedure, indicates the predictability of the model.
Fold change (FC) analysis and independent-sample t-test statistics were rst applied for comparison of metabolite levels to determine statistically signi cant differences among the four groups. The cut-off criteria for screening differentially expressed metabolites were FC>2, VIP>1.5, and P<0.05. According to databases, the chemical structures of important metabolites were then identi ed, such as the Human Metabolome Database (http://www.hmdb.ca), using accurate mass and MS/MS fragment data.
HCA was performed and visualized by using the embedded module of MetaboAnalyst 4.0 [15]. By applying the Euclidean distance measure and Ward clustering algorithm, dynamic changes in signi cantly different metabolites were compared, and the ratio obtained was used to draw an HCA heat map. Colours in the heat maps correlate with the degree of increase (red) and decrease (blue) relative to the mean metabolite ratio.
Based on the differentially expressed metabolites, metabolic pathway and metabolite biofunction analyses were performed using the network database (KEGG pathway http://www.genome.jp/kegg/) to investigate the bioprocesses affected by T. pallidum infection. In brief, the enrichment level was calculated by the t-test, and metabolic pathways with P values less than 0.05 were considered statistically signi cant.

Clinical information for the study participants
In addition to seven non-syphilis patients, 42 patients with neurosyphilis and 39 patients with latent syphilis were recruited in this study, including nine asymptomatic neurosyphilis patients. Because CSF samples from healthy people are challenging to obtain, samples from seven control groups were obtained, namely, from neurologic patients who were T. pallidum negative. The clinical characteristics of the patients are summarized in Table 1.

Quality control and overall metabolic pro le
Data were subjected to a data integrity check before subsequent analysis, and no missing values were detected. The relative standard deviation (RSD) calculated from the ion features of the QC samples showed RSD values mostly less than 30%, indicating that the analysis program was reliable and could be used for subsequent sample analysis. The %RSD distribution is presented in Fig.1(a, b). The base peak intensity (BPI) chromatograms of the sample are illustrated in Fig.1(c, d). Among samples, 3799 and 3937 features were obtained in ESI+ and ESI-modes, respectively.
PCA was carried out using the molecular features of all the groups, and the distribution of metabolic pro les for the test samples in PCA is shown in Fig.2(a). The samples in each group were tightly clustered, but the difference between the groups was not large enough for a clear distinction. To eliminate any nonspeci c effects of the technique and con rm the biomarkers for differentiating the metabolite pro les of non-syphilis controls and other syphilis patients (including latent syphilis and neurosyphilis patients), a supervised PLS-DA model was established focused on the actual class discriminating variation. As depicted in Fig.2(b), due to the individual differences of the control samples, the distribution in the larger picture was scattered, but it was separated from the patient samples, with a clear distinction; R2Y was calculated to be 0.986 and Q2Y 0.632, which are greater than 0.4. To investigate dynamic changes in CSF metabolic pro les from syphilis to neurosyphilis, metabolomic data for the 39 latent syphilis and 42 neurosyphilis samples were analyzed by PLS-DA. As shown in Fig.2(c), we observed a clear separation between the two groups. Speci cally, R2 and Q2 were 0.992 and 0.873 for differentiating syphilis patients and neurosyphilis controls, respectively. In addition to the obvious clinical manifestations, there was a signi cant difference in the metabolic spectrum between asymptomatic neurosyphilis and symptomatic neurosyphilis. According to the PLS-DA results (Fig.2d), there was a signi cant separation between the two groups with Q2Y=0.818 and R2Y=0.995.

Identi cation of differential CSF metabolites and pathway analysis
Based on the database search, 272 metabolites were identi ed in ESI-mode and 252 in ESI+ mode.
Metabolites with VIP greater than 1.5 were included in t-tests to assess signi cant differences in CSF metabolites between different patients. A total of 20 metabolites obtained from the metabolic spectrum in both modes were ltered out compared to syphilis and non-syphilis patients, including ascorbic acid, acetaminophen, and fucose, which play an essential role in the immune system or nervous system. Next, ve signi cantly different metabolites, nicotine, trihydroxycoprostanoic acid, 4-hydroxybenzoic acid (4-HBA), 5-hydroxy-L-tryptophan (5-HTP), and acetyl-N-formyl-5-methoxykynurenamine (AFMK), were found between patients with latent syphilis and patients with neurosyphilis. However, only theophylline showed a signi cant difference in comparison of the metabolic pro les of symptomatic neurosyphilis and asymptomatic neurosyphilis. Differences in expression of 5-HTP and AFMK between the two types of patients were also found, with no signi cant differences (0.05<P<0.1). Levels of both metabolites were up-regulated more than three times in the samples from patients with symptomatic neurosyphilis. It is worth noting that the level of 5-HTP differed between the non-syphilis control group and latent syphilis, symptomatic neurosyphilis, and asymptomatic neurosyphilis groups. The signi cant differential metabolite components are described in Table 2. According to the differential metabolites screened, a heat map of HCA was generated based on Euclidean distance and the average clustering algorithm, as shown in Fig.3.
Next, we performed a KEGG analysis of the differentially expressed metabolites to examine the effect of T. pallidum infection on metabolic pathways. The results showed that these differential metabolites were enriched in amino acids and metabolic intermediates of various acids. The metabolic pathways of enrichment are mainly related to amino acid metabolism, energy metabolism, and phosphatidic acid metabolism, such as the tricarboxylic acid cycle. Fig.4 shows that there were marginally signi cant differences (P<0.1) between the groups: the citric acid cycle, gluconeogenesis, tryptophan metabolism, ubiquinone biosynthesis, and caffeine metabolism. In particular, tryptophan metabolism was signi cantly different (P<0.01) between neurosyphilis and latent syphilis and between symptomatic neurosyphilis and asymptomatic neurosyphilis.

Discussion
In both latent syphilis and neurosyphilis, there are no apparent symptoms at the initial stage of infection, and serological tests are prone to false-positive as well as false-negative results, making syphilis diagnosis a clinical challenge. T. pallidum invades the nervous system during early primary syphilis, causing problems for the daily life of patients [16,17]. Therefore, it is imperative to screen, diagnose and treat suspected cases in a timely manner. Previously, Yang's team conducted a metabolomic analysis of serum and CSF samples from patients with syphilis and neurosyphilis [18]. Several signi cantly expressed metabolites, i.e., L-gulono-gamma-lactone, D-mannose, N-acetyl-L-tyrosine, and hypoxanthine, were identi ed in CSF and trimethylamine N-oxide in serum from neurosyphilis patients. In this study, a more detailed metabolomic analysis was carried out on 88 cerebrospinal uid samples from patients with neurosyphilis (symptomatic and asymptomatic) and latent syphilis as well as those without syphilis.
The metabolic analysis detected more than 400 metabolites, of which 20 were signi cantly different between syphilis and non-syphilis patients and ve between neurosyphilis and latent syphilis patients. Only theophylline was signi cantly different between symptomatic and asymptomatic neurosyphilis patients. KEGG pathway enrichment analysis revealed that tryptophan metabolism pathways were signi cantly enriched, indicating a high correlation between tryptophan metabolism and syphilis symptoms.
Signi cant differences in metabolites and pathways were found between the syphilis group and the nonsyphilis control group. The levels of 9,10,13-TriHOME, 9-OxoODE, PS (16:0/14:0), azelaic acid, and PS (16:0/14:0) were found to up-regulated in the syphilis group, but 2-HB and GDP were signi cantly downregulated. Among the up-regulated metabolites, 9,10,13-TriHOME, 9-OxoODE, and PS (16:0/14:0) are lipid metabolites. PS is an active substance on the cell membrane, especially in brain cells, and its main function is to improve the function of nerve cells, sooth vascular smooth muscle cells and increase blood supply to the brain [19]. It has been con rmed that the membrane protein of T. pallidum contains phosphatidylserine lipoprotein, which may explain the increase in PS [20]. This might distinguish syphilis from false-positive VDRL caused by antiphospholipid syndrome due to antiphospholipid antibodies. In general, 9,10,13-TriHOME and 9-OxoODE are related to the oxidative metabolism of linoleic acid [21,22]. TriHOMEs are the end product of linoleic acid oxidation, which have a physiological role in maintaining the water-skin barrier, though other physiological effects are unclear [23]. Linoleic acid has a neuroprotective effect, and its de ciency has been found in patients with mild neurological disorders and Alzheimer's disease [24]. This may also explain the increase in linoleic acid metabolites in syphilis patients. Downregulated 2-HB is associated with dyslipidemia, involved in glutathione oxidative stress, and elevated in patients with depression [25].
Signi cant differences in 5-HTP and AFMK were found in neurosyphilis patients compared with latent syphilis, and it may constitute a marker of whether T. pallidum has invaded the nervous system. In some lymph node tumour cells, expression of indoleamine-2,3-dioxygenase (IDO1) is increased by activating regulatory T cells (Tregs) that escape the host immune system [26,27]. This enzyme is involved in the breakdown of indoleamine (tryptophan, 5-hydroxytryptamine, and melatonin). AFMK participates in the kynurenine pathway, one of the three main metabolic pathways of melatonin and has potent antioxidant and anti-in ammatory abilities [28]. The high levels of AFMK observed in this study may result from melatonin production by local immune cells. 5-HTP is decarboxylated by aromatic L-amino acid decarboxylase with vitamin B6 as a cofactor to form serotonin (5-hydroxytryptophan, the precursor of melatonin). 5-HTP contributes to serotonin production, promoting emotional and nervous system health.
Both melatonin and serotonin are important metabolites in regulating emotion. It has been suggested that the tryptophan-kynurenine pathway may be involved in depression by mediating the in ammatory response [25]. This may explain why some symptoms of neurosyphilis are consistent with depression; it also suggests that more attention should be paid to the mental state and emotional counselling of patients during treatment for neurosyphilis.
Only one metabolite was found to differ between symptomatic and asymptomatic neurosyphilis patients, which may be due to the different organs involved in symptomatic neurosyphilis. Theophylline is a metabolite of caffeine that can increase CSF secretion to regulate intracranial pressure and relieve headaches [29,30]. Theophylline also reduces smooth muscle tension, promotes endogenous epinephrine and norepinephrine release, inhibits calcium release from the endoplasmic reticulum of smooth muscle, and decreases intracellular calcium concentrations for respiratory tract dilation [31][32][33].
Due to the di culty of lumbar puncture and poor compliance of patients, it is challenging to obtain CSF samples. Besides, this study enrolled few control samples. Overall, only preliminary conclusions were obtained in this study, and the sample size needs to be expanded for follow-up research. Nonetheless, some interesting conclusions were obtained. In the comparison between syphilis patients and nonsyphilis patients, signi cant differences in lipids and derivatives were detected, indicating that lipid metabolism plays a vital role in syphilis. Thus, liposomics may be used as a further research direction. As neurosyphilis is usually the result of the late development of untreated syphilis, it is important to promptly distinguish neurosyphilis from latent syphilis. Both 5-HTP and AFMK are important metabolites for maintaining the health of the nervous system. Although differences in other comparisons were found for 5-HTP, they were not signi cant. The ndings suggest that more attention should be paid to the mental state and emotional counselling of patients under treatment for neurosyphilis. The metabolites identi ed should be further studied to determine their relationship with T. pallidum infection.
In conclusion, we performed a metabolomics analysis of CSF samples from patients with symptomatic neurosyphilis, asymptomatic neurosyphilis, and latent syphilis. Through comparative analysis, we found that levels of linoleic acid metabolites were up-regulated in patients with syphilis. 5-HTP and AFMK in the tryptophan metabolism pathway were also signi cantly altered in neurosyphilis. Theophylline levels were signi cantly up-regulated in patients with symptomatic neurosyphilis. The role of these differential metabolites in the development of syphilis is worthy of further exploration. This analysis may facilitate the development of biomarkers for syphilis and determine the underlying molecular mechanisms associated with neurosyphilis.     Heat map based on differential metabolites of latent syphilis (Y), asymptomatic neurosyphilis (W), symptomatic neurosyphilis (Z) and non-syphilis control (F).