Background
Syphilis, caused by
Treponema pallidum pallidum, is an identifiable and curable infection [
1,
2]. The diagnosis of syphilis mainly relies on serological testing for reactivity to both non-treponemal (cardiolipin) and treponemal antigens [
3]. Treponemal assays are primarily used to confirm the presence of past or current syphilis, whereas non-treponemal assays are mostly used to monitor the “activity” of infection [
4]. Direct detection methods like dark-field microscopy, direct fluorescent antibody testing, nucleic acid amplification testing, and examination of cerebrospinal fluid can also be used, however they are often insensitive and difficult to access [
5].
There are limitations to serologic testing for syphilis. Depending on the stage of infection and severity of immunosuppression [
6], serologic testing has varying degrees of sensitivity and specificity [
4]. It is estimated that commonly used treponemal and non-treponemal tests have a sensitivity ranging from 76 to 100% and specificity ranging from 97 to 99% when compared to direct detection during symptomatic syphilis [
7]. Additionally, the optimal clinical management of the 15% to 41% of persons who remain serofast following treatment for syphilis can be unclear [
8]. The serofast phenomenon occurs because non-treponemal assays detect host antibodies (cardiolipin), which can have long half-lives [
8,
9]. False-positive non-treponemal tests can also occur among those with febrile illnesses, immunizations, hepatitis C virus infection, connective tissue disease, intravenous drug use, malignancy, older age, malaria, Chagas disease, tuberculosis, and leprosy [
7,
10,
11]. The antibodies that treponemal assays detect can remain present for life, even after successful treatment for syphilis. False-positive test results can also be due to due to other spirochete infections (
Borrelia spp.) and from commensal microorganisms [
7,
12]. To improve upon the limitations of serologic testing, diagnostic algorithms have been created to increase testing sensitivity and specificity [
4]. Those algorithms are based on serologic testing with a non-treponemal assay with a verification treponemal assay. However, those algorithms can miss previously treated, early untreated, and late latent cases of syphilis due to non-reactive non-treponemal assays. Current diagnostic algorithms are unable to distinguish currently active from previously treated syphilis.
Studies using specific inflammatory cytokines as potential markers of infection resolution among patients with primary and secondary syphilis have found promising results [
13,
14]. However, those studies tested few cytokines and did not have measurements of cytokine concentrations prior to incident syphilis. We investigated whether cytokine profiles might provide insight into the “activity” of syphilis by measuring 63 plasma cytokines in a longitudinal cohort of participants with no syphilis, incident syphilis, and treated syphilis. We aimed to address three questions: [
1] are changes in cytokine concentrations associated with changes in rapid plasma reagin (RPR) titers; [
2] does cytokine response depend on directionality of changes in RPR titers, i.e., increasing versus decreasing RPR titers; and [
3] can observing one or more cytokines better distinguish participants with active infection and treated syphilis.
Methods
From 2013 to 2015, a cohort of men who have sex with men (MSM) and male-to-female transwomen (transwomen) who engaged in behavior that put them at risk for syphilis were recruited and enrolled into an observational study in Lima, Peru [
15,
16]. Briefly, participants attended a sexually transmitted infection clinic every 3 months for a period of 2 years. During study visits, participants were interviewed, received routine syphilis and human immunodeficiency virus (HIV) infection testing, and had clinical specimens collected.
Collection and storage of blood serum
Whole venous blood samples were collected from participants every 3 months. Sampled blood was held at room temperature for clot formation and then underwent centrifugation. Sera from samples were frozen and stored at − 80 °C at the Universidad Peruana Cayetano Heredia Sexual Health Laboratory.
Syphilis testing among participants
Participants were tested for syphilis with serum RPR tests (BD Macro-Vue™ RPR Card Test Kit, Beckton Dickinson, Franklin Lakes, NJ) and Treponema pallidum particle agglutination (TPPA) tests (Serodia, Fujirebio Diagnostics Inc., Tokyo, Japan), using a cutoff value of ≥ 1:80. Participants were also tested for HIV infection (Alere Determine™ HIV 1/2, Alere Inc., Waltham, MA, USA and NEW LAV BLOT I, Bio-Rad, France).
Selection of samples for cytokine analysis
Incident syphilis was defined as an RPR titer that was ≥ 1:8 or a fourfold increase from the prior RPR titer, and a positive TPPA test. Treated syphilis was defined as a fourfold or larger reduction in RPR titer following syphilis treatment. We selected a serum sample collected during a clinic visit in which a participant was found to have incident syphilis, in addition to a serum sample collected from the same participant when they did not have syphilis from a temporally adjacent clinic visit. Samples collected from participants with incident syphilis were coded as “infection-yes” and samples collected from participants without syphilis were coded as “infection-no.” Additionally, samples were stratified by RPR titers, as well as whether they were collected prior to incident syphilis, i.e., “no syphilis,” and after the participant was treated for syphilis, i.e., “cured”. All samples were shipped on frozen carbon dioxide to the Human Immune Monitoring Center at Stanford University for analysis.
Luminex assay and magnetic bead kits
The simultaneous measurement of 63 cytokines (eBiosciences/Affymetrix, San Diego, California, USA) in serum samples was performed using a multiplex bead-based Luminex assay. See Additional file
1: Table S1 for a complete list of tested cytokines.
Beads were added to 96-well plates and washed in a Biotek ELx405 washer (Biotek Instruments Inc., Winooski, Vermont, USA). Sera samples were added to plates containing mixed antibody-linked beads and incubated at room temperature for 1 h. Those samples were incubated overnight at 4 °C with shaking. Cold and room temperature incubation steps were performed on an orbital shaker at 500–600 rotations per minute. Between steps, plates were washed. Following overnight incubation, biotinylated-detection antibodies were added to plates for 75 min at room temperature, while shaking. Plates were incubated at room temperature with streptavidin-PE (eBiosciences/Affymetrix, San Diego, California, USA) for 30 min. A reading buffer was added to wells on each plate. Each sample was measured in duplicate. Plates were read using a Luminex 200 instrument (Luminex Corporation, Austin, Texas, USA) with a lower bound of 50 beads per sample per cytokine. The assay measures cytokine concentration and reports results in units of median fluorescence intensity (MFI). Four custom assay control beads (Radix Biosolutions, Georgetown, Texas, USA) were added to all wells.
Statistical methods
Linear models of titer and cytokines
We derived changes in RPR titer (log2) and cytokines by comparing the RPR titer value at time of incident syphilis to the RPR titer value taken from the prior visit. We then compared the linear model of RPR titer ~ infection (reduced model) to titer ~ infection + cytokine + infection × cytokine (full model), where titer and cytokine values are changes, deriving a p-value based on a partial F-test. Since this is a hypothesis generating study, we report unadjusted p-values. All statistical tests were performed in R (R, Vienna, Austria).
Comparison of changes in cytokine levels between “uninfected to infected” and “infected to uninfected” participant samples
To address the question of whether changes in cytokine concentrations are different when the patient is moving from an “uninfected to infected” as compared to “infected to treated,” we computed changes in cytokine concentrations (for example, a cytokine concentration was compared between incident syphilis and no syphilis). Samples were compared with other samples from an individual. We classified those changes as from “uninfected to infected” and from “infected to cured.” We compared those values using a t-test. Since this is a hypothesis generating study, we report unadjusted p-values.
Analysis of relationships between cytokine pairs
We considered all possible pairs of 63-cytokines (n = 1953) to identify relationships between pairs of cytokines. To do that, we divided values for each cytokine into 3 equal frequency tertiles or “bins” (low, medium, and high). Then, we focused on observations in low and high bins. For a particular pair of cytokines, we created a 2 × 2 array (cytokine 1, cytokine 2, low, high) and identified which 2 corners contained the most observations by comparing the number of observations in the lower left (low–low) and upper right (high–high) corners (positive correlation) to the number in the upper left (low–high) and lower right (high-low) (negative correlation). We next compared the number of observations associated among those with incident syphilis to those with no syphilis in the “left hand” and “right hand” corners. We calculated Chi squared values between those measures. We considered a p-value < 0.05 to be significant. We represented the significant relationships in a network graph using the R package, igraph. We then determined if those relationships could be used as classification rules, and identified which samples were correctly classified as “no infection” or “yes infection” by those rules. The resulting data were displayed in a heat map with hierarchical clustering (R method heatmap.2, with clustering method = complete). A similar analysis was used to identify which people were correctly classified by which clustering rules, with a correct classification for any sample considered to be a correct classification for each participant.
Analysis of variability
To compare the variability of samples drawn from our current study in Lima, Peru to a control population, we identified 16 healthy males without syphilis between the ages of 20 to 40 years who participated in a flu vaccine study in Palo Alto, California [
17]. Each of those control participants contributed 3 samples over a 30 days study period in either 1 or 2 years (3 or 6 total samples per person). All of those Palo Alto samples were analyzed on the same Luminex 63-plex kit (H51-2 and 12-10). Since the Peru cohort was analyzed on a different kit than control samples for age and sex, we also included control data from 4 healthy participants of a 14-week diet modification study (7 samples each) who were analyzed using the same Luminex kit as the Lima samples (H62-1 and 5-15). For each person and for each analyte, a percent coefficient of variation (%CV) was calculated as (standard deviation/mean) × 100. Differences between means of the 3 groups were tested using an ANOVA test without correction for multiple comparisons.
Discussion
This study explored the relationship between changes in cytokine concentrations and syphilis infection state by measuring 63 cytokines in serum samples from a longitudinal sample of MSM and transwomen from Lima, Peru. Though pairwise analysis, 20 cytokine pairs could distinguish serum samples collected from participants with incident syphilis and cured syphilis. We identified three cytokine networks of interest: an Eotaxin–Rantes–Leptin network, a Mig-IL1ra-Trail-CD40L network, and an IL12p40-IL12p70 network. We found that sub-groups of participants had different cytokine responses to incident syphilis: Some participants grouped into an Eotaxin–Rantes–Leptin network, whereas others grouped into a Mig-IL1ra-Trail-IL12p70 network. Finally, our analysis indicated there might be differences in cytokine concentration responses, particularly Resistin, BDNF, CD40L, and IFNB, during incident syphilis versus cured syphilis. While some participants were infected with HIV, there were no convincing clustering effects around HIV infection status. Comparisons between our cohort and control cohorts showed differences in variability by cytokine, with some cytokines showing higher variability in the study cohort. High cytokine variability is suggestive that differences in cytokine concentrations observed in this study between active syphilis and cured syphilis are likely true phenomena versus random variation.
We were unable to find similar studies for a comparison of our findings. Our current study design greatly differs from our previous smaller cross-sectional pilot study, which compared cytokine values from samples collected from 5 patients with syphilis (in which the duration of syphilis was unknown) to samples collected from 5 patients without syphilis [
18]. While there were large methodological differences between those studies, our current study also found significant differences in Eotaxin, Leptin, IL12P70, and VEGF-D, which replicated some of the findings of our earlier study. We did not replicate findings of differences in GMCSF, IL10, IL15, IL1B, IL7, IP10, MCP3, MIP1B, NGF, PDGFBB, TNFA, and VEGF.
In addition to its role as an anorexigenic, leptin has also been found to be a proinflammatory factor [
19]. In acute inflammation and infection, leptin levels increase due to the presence of bacterial endotoxins and other signaling cytokines like TNF-a, IL-6, and IL-1b [
20,
21]. In experiments,
Klebsiella spp. pulmonary infections are more lethal in leptin-deficient mice when compared to wild-type mice. It was also found that leptin administration into leptin-deficient mice reduced bacteriemia, improved macrophage phagocytosis, improved polymorphonuclear leukocyte H
2O
2 production, reduced bacterial load in vitro, and improved survival [
22‐
26]. Leptin could have an important role in the immune response to syphilis, however it may also be elevated among our participants due to the treatment of syphilis with penicillin. A study among patients with the Jarisch-Herxheimer Reaction found that TNF, IL-6, and IL-8 were elevated 2 to 4 h after receiving penicillin treatment, with a return toward baseline values 12 h after receiving penicillin [
27]. Therefore observed elevations in inflammatory cytokines could also be observed due to the treatment of syphilis versus incident syphilis.
There were several limitations to our study. Our analysis is based on data collected from a medium sized cohort (103 samples from 44 persons) characterized by confounders such as prior syphilis. Detection of changes in cytokines were limited to 3-month intervals due to the parent study design. There may be differences in cytokine expression among patients infected with HIV infection [
28], however statistical analysis did not find differences between participants infected with HIV and participants not infected with HIV. There is variation in our data, as shown in Fig.
1. We were unable to determine if that variation was related to individual variability or to unidentified technical considerations. However, we performed a variability analysis that showed consistency in datasets when we compared our results with those from cohorts of matched healthy controls. There are several limitations in our identification of ratios for classifying samples. Our analysis did not consider correlation within multiple samples from the same person. In addition, the definition of tertiles was created from our data. Additional studies are needed to validate our identified rules and to derive general threshold values for low and high cytokine concentrations.
Our study has many strengths. In our study, we tested stored sera collected at regular intervals from a longitudinal cohort that allowed us to compare samples collected during incident syphilis to those collected 3 months prior. This allowed each participant to act as their own control. While prior studies have compared cytokine levels during an incident syphilis to those after treatment, we were able to compare cytokine levels prior to incident syphilis to those during incident syphilis. With that approach, we may have a better comparison group that is not influenced by variation between individuals. Additionally, quarterly testing allowed us to have a documented time frame in which incident syphilis took place. We performed an analysis using RPR titers as a continuous variable. Another strength of our analysis of ratios is that we identified pairs of cytokines that have the potential to distinguish a patient who does not need treatment for active syphilis from a patient who does. Those values can be determined in a single visit, unlike changes in cytokine concentrations over time. Future research is needed to confirm the presence of the cytokine networks we identified.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.