We have conducted a prospective cohort study of acute Lyme disease and post-treatment events (see Table 1). All patients enrolled in the study have untreated early Lyme disease that meets the Infectious Disease Society of America (IDSA) criteria for diagnosis [6], [11], [14]. Using samples collected from this cohort we employed a bead-based multiplex cytokine assay to measure the levels of 58 immune mediators and 7 acute phase proteins in the serum of patients with untreated early (acute) Lyme and matched controls. Displayed in figure 1A is a heat map of those mediators significantly elevated (q<0.1%) in early acute Lyme disease at the initial pre-treatment visit compared to controls. Most notable are elevated levels of several T cell chemokines (CCL19, CXCL9, CXCL10), acute phase inflammatory markers (CRP and serum amyloid A), several IL-1 cytokine family members (IL-1ra, IL-18, IL-33), inflammatory cytokines TNF-α and IL-6 and the T cell cytokine IL-2. Collectively, these cytokines and chemokines generate a novel signature that clearly distinguishes acute Lyme patients from normal controls.

Displayed in figure 1B is an unsupervised hierarchical clustering of the serum mediator profiles of the same early acute Lyme disease at the initial pre-treatment visit and control patients. This analysis allows us to distinguish two patient clusters of acute Lyme disease patients. One group termed mediator-high, display-elevated levels of the T cell chemokines and inflammatory markers during acute infection described above. A second group of Lyme disease patients (mediator-low) display mediator levels that result in their clustering among normal control samples. Since all patients met IDSA guidelines for acute Lyme disease, the mediator-low group represents a subset of Lyme patients that do not exhibit significant elevations in inflammatory mediators in blood.

Based on immune mediator levels during acute infection, two subgroups of Lyme patients could be distinguished. Table 2 presents a comparison of these two groups for a range of clinical outcomes. No differences in demographics, duration of illness or the distribution of single versus multiple erythema migrans lesions were noted between the two groups. However, significant differences were noted in the number of symptoms (p<0.005), absolute lymphocyte levels (p = 0.002), liver enzyme tests (p = 0.027) and the presence of detectable anti-Borrelia antibodies (p = 0.021). Lyme disease patients displaying high mediator levels at the initial pre-treatment visit (acute disease) had higher rates of seroconversion, but also had greater rates of lymphopenia and elevated liver enzyme.

A striking feature of the cytokine/chemokine profile was the upregulation of T cell specific mediators. Displayed in figure 2A–D are the measured levels of the chemokines CXCL9 (MIG), CXCL10 (IP-10), CCL19 and CXCL8 (IL-8) in sera obtained from patients at the time of their initial diagnosis of acute Lyme disease (pre-treatment), 4 weeks following diagnosis and treatment (post-treatment) and matched controls. The T cell chemoattractants CXCL9, CXCL10 and CCL19 are significantly elevated in serum during the acute infection but largely return to normal levels following treatment, resolution of the erythema migrans and recovery. Levels of CXCL9, CXCL10 and CCL19 are completely concordant with one another (Table 3). In contrast, the neutrophil chemotactic factor CXCL8 is not elevated during acute infection or at other observed time points (Figure 2D). Furthermore, the measured levels of CCL11 (eotaxin-1), CXCL1 (GROa), CCL2 (MCP-1), CCL7 (MCP-1), CCL3 (MIP-1a), CCL4 (MIP1b), CCL5 (RANTES) and CXCL12 (SDF-1a) are not significantly different as compared to the levels in matched controls (data not shown). Therefore in acute Lyme disease there appears to be a selective and coordinate elevation of T cell chemo-attractants.

As revealed in the global serum profiling shown in figure 1, the cytokine IL-6 and the innate immune acute phase factors C-reactive protein (CRP) and serum amyloid A (SAA) are elevated during acute Lyme disease. A more detailed analysis of the levels over time is presented in Figure 3. Increased levels of CRP (p = 0.0091), SAA p<0.0005) and IL-6 (p = 0.0282) were seen during the acute phase. CRP and SAA levels return to normal control levels following treatment and remain so throughout the follow-up period. IL-6 levels remain elevated and only return to normal levels months after infection and treatment. As is the case for the T cell chemokines, serum CRP and SAA significantly correlated with one another (Table 3, p = 0.016). Surprisingly, IL-6, a known inducer of CRP and SAA did not correlate with these acute phase proteins (p = 0.09 and 0.54 respectively, Table 3).

T cell chemokine levels in acute Lyme patient sera, while clearly elevated, do show heterogeneity with a subset displaying levels similar to controls (Figure 2A–C). A similar case can be made for the inflammatory biomarkers CRP and serum amyloid A (Figure 3). This heterogeneity suggests that individual chemokine and/or inflammatory markers may correlate with clinical parameters.

The erythema migrans (EM) lesion is the primary site of inflammation in acute Lyme disease and it can vary in size as well as number. This lesion is characterized as the site of active bacterial growth and the accumulation of immune inflammatory cells that is dominated by T cells [9], [15], [16]. The chemokines CXCL9 and CXCL10 are present at high levels within the EM lesion likely produced by fibroblasts and endothelial cells in an interferon-γ dependent manner [15]–[17]. Based on this we reasoned that the levels of CXCL9 and/or CXCL10 might correlate either with the size or degree of dissemination of the EM lesion. Within the Lyme cohort, a significant proportion of patients had disseminated lesions and the size of the lesions was varied (Table 1). Surprisingly, when we compared these variables with the levels of serum CXCL9 and CXCL10 there was no statistical association (Figure S1 and Table S1).

Previous studies have also described changes in liver enzyme levels during acute Lyme disease [18], [19]. This pattern is also observed in the current cohort of Lyme disease patients (Table 1). When we compared the levels of serum CXCL9/CXCL10 among the acute Lyme patients with high liver enzyme levels, there was a significant association of high CXCL9/CXCL10 and CCL19 levels with elevated liver enzymes in the acute phase of Lyme disease (Figure 4).

The finding that liver enzymes and acute inflammatory markers are elevated in acute Lyme disease suggested that there these two markers may be related. As shown in Figure 5, CRP and SSA but not IL-6 (data not shown) levels correlate with liver enzyme levels.

In this patient cohort, 35.7% of Lyme patients fail to test positive either at diagnosis or through seroconversion following antibiotic treatment (Table 1). This is consistent with previous studies that demonstrated that a significant fraction of Lyme patients that exhibit EM along with other symptoms of infection do not seroconvert when assayed by the current two-tiered testing [11], [12]. When we examined whether various elevated mediator levels could be correlated with seroconversion status, a significant association was observed with elevated CXCL9 and CXCL10 but not CCL19 levels (Table 4). When other clinical parameters such as duration of illness or size/dissemination of EM lesions (data not shown) were analyzed, there was no association with seroconversion status.

The chemokines CXCL9 and CXCL10 bind to a common receptor CXCR3 expressed largely on T cells [20]. We reasoned that high serum levels of these chemokines may drive T cells into inflamed tissues and as a result levels of CXCR3 expressing T cells may be altered during acute Lyme disease. Levels of CXCR3⁺ CD4 T cells were determined by polychromatic flow cytometry (Figure 6A), and found to be significantly lower in the blood of patients with acute Lyme disease versus controls (Figure 6B). In addition, serum levels of CXCL10 (Figure 6C) and CXCL9 (data not shown) were inversely related to the frequency of CXCR3⁺ CD4⁺ T cells in the peripheral circulation.

The serum and PBMC samples used have been generated as part of a prospective cohort study with age- and sex-matched controls enrolled from 2008–2013. This study includes a well-defined cohort of patients with acute Lyme disease enrolled from a Lyme endemic area of the mid-Atlantic United States. Only patients with untreated, confirmed early Lyme disease manifesting an active EM skin lesion at the time of enrollment, as defined by CDC case criteria are eligible [6], [11], [14]. Patients with a history of prior Lyme disease or the presence of confounding preexisting medical conditions associated with prolonged fatigue, pain or neurocognitive symptoms are excluded [45]. Controls are nonhospitalized age- and sex-matched and have no prior history of Lyme disease or any exclusionary medical conditions including lack of inflammatory disorders. The enrolled Lyme disease patients are followed from the time of acute infection longitudinally for a period of 2 years for a total of 7 study visits. The matched controls are followed for 2 years across 4 study visits. At each study visit, extensive clinical data and biological specimens are collected (see below).

At the pre-treatment study visit, a standard Complete Blood Count (CBC) and Comprehensive Metabolic Panel (CMP) are drawn and performed at an internal laboratory. In addition to these clinical tests, an additional SST tube is drawn to be sent to an outside, commercial laboratory for Lyme serology testing. Serology results are determined following the CDC's two-tier testing algorithm measuring both IgM and IgG, with time of symptom onset being determined by a structured interview with the patient at the pre-treatment study visit [6], [11], [14]. For those patients who are negative according to the two-tier serology at the first study visit, a subsequent serology is drawn at the second study visit after antibiotic treatment and sent to the same commercial laboratory for testing.

At each study visit, patients are given a physical exam, a structured interview of twenty signs and symptoms of disease, and undergo a blood draw. At the first visit, a measurement of the EM is taken and recorded.

We performed multiplex analysis of 58 cytokines/chemokines and 7 acute phase markers using the Bio-PlexTM bead array system as recommended by the manufacturer and using previously described optimized assay protocols [46]–[49]. Data processing was performed using Bio-Plex manager software version 4.4.1 and serum concentrations were interpolated from standard curves for each respective cytokine. This protocol and data generated were MAIME compliant and were deposited in the Gene Expression Omnibus Repository (accession number GSE55815).

Peripheral blood mononuclear cells (PBMCs) were isolated from fresh heparinized blood using Ficoll Hypaque. Monoclonal antibody reagents specific for CD3 (UCHT1), CD4 (RPA-T4), CD8 (SK1) and CXCR3 (IC6) were purchased from Becton Dickenson. PBMCs were first incubated with unlabeled human IgG to block nonspecific binding followed by incubation with fluorescent tagged monoclonal reagents. PBMCs were washed and polychromatic flow cytometry performed using a FACSAria instrument (Becton Dickinson, San Jose, CA). Lymphocytes were gated using forward and side scatter and the data analyzed using FlowJo software (Tree Star).

For descriptive analyses of the multiplex biomarker data as continuous variables, data was analyzed using the log of the “ratio to average”. The log of the ratio to average was calculated by setting all values less than 1 pg/mL to 1 pg/mL, and then calculating the log base 2 of [(value)/(average value in the cohort)]. The “ratio to average” values were input into SAM (Significance Analysis of Microarrays Version 4.0) [50]. The SAM output was sorted based on false discovery rates (FDRs, represented by the q value) in order to identify mediators with the greatest differences in levels between patient subgroups. We used hierarchical clustering software Cluster 3.0 to arrange the SAM results according to similarities among mediator levels, with no markers weighted in Figure 1A, and with weighting of CCL19 and Serum Amyloid A was used in the clustering for Figure 1B. The clustered results were displayed using Java Treeview (Version 1.1.5r2).

Categorical variables were analyzed using Fisher's Exact or Chi-square statistics. A standard ANOVA or unpaired t-test was used on two or more group comparisons for continuous variables. Pearson correlations were used for linear comparisons. For longitudinal continuous variables, a repeated measures ANOVA was performed. All statistical calculations were performed with SPSS software (IBM Corporation v 21).

All human subject studies in this manuscript were reviewed and approved by the Johns Hopkins School of Medicine Institutional Review Board (protocol NA_00011170). All subjects recruited in the study were adults and all were provided written informed consent.