INTRODUCTION
Lyme disease, a tick-borne infection caused by
Borrelia burgdorferi, affects approximately 300,000 people annually in the United States (
1,
2). The symptoms of acute Lyme disease are highly variable, and when untreated, it can progress in severity over time from malaise and flu-like symptoms to neurological disorders, cardiac complications, and, in late stages, arthritis (
3). Antibiotic intervention typically cures Lyme disease; however, approximately 10 to 20% of Lyme patients develop posttreatment Lyme disease syndrome (PTLDS) with symptoms, including myalgias, chronic fatigue, and cognitive difficulties for more than 6 months after completion of antibiotic treatment (
4–8). The etiopathology of PTLDS is unknown, but it presents with symptoms that overlap those of other diseases, including chronic fatigue syndrome (
5,
9), depression, fibromyalgia (
5), and multiple sclerosis (
10).
Along with unknown etiopathology and a diverse range of symptoms, diagnosing PTLDS remains challenging. Although a clinical case definition proposed by the Infectious Diseases Society of America (IDSA) in 2006 has served as a specific research tool, there is no biological method to diagnose PTLDS (
11). While clinical biomarkers associated with PTLDS have been observed, a suitable diagnostic method and therapy remain elusive. A positron emission tomography (PET) brain imaging study among patients with PTLDS demonstrated elevated microglial activation compared to that of controls, congruent with localized inflammation (
12). Additional research has shown that a greater
B. burgdorferi-specific plasmablast response prior to treatment favors a resolution of symptoms versus the development of PTLDS, which indicates that even before treatment, a patient’s immunological landscape plays an important role in the development of PTLDS (
13). Compared to healthy controls, patients with PTLDS have significantly elevated expression of interferon alpha, greater antibody reactivity to brain antigens (
14), increased levels of the chemokine CCL19 (
15) and the cytokine interleukin 23 (IL-23) (
16), and a decrease in the CD57 lymphocyte subset (
17,
18). Furthermore, patients have a higher risk of developing new-onset autoimmune joint diseases after a Lyme erythema migrans rash (
19). Therefore, while the etiopathology is still unknown, these markers indicate biological abnormalities among patients with PTLDS.
The microbiome has been implicated in many diseases with symptoms that overlap those of PTLDS, including autoimmune diseases. The pathogenesis of autoimmune disease is affected by environmental (
20) and genetic (
21) factors, as well as the gut microbiome (
22,
23). The gut microbiome plays an important role in human health and has been shown to strongly influence host metabolism (
24,
25) and the immune (
22,
26) and nervous (
27) systems, as well as provide crucial colonization resistance against a range of intestinal pathogens (
28,
29). Further, microbiome compositional changes can alter immune tolerance (
22,
23). For instance, members of the intestinal microbiota have been characterized as contributing to the development of the long-term sequelae of acute infection events upon disruption of tissue and immune homeostasis (
30). Studies have found the microbiome to be on par with and often superior to the human genome in predicting disease states (
31,
32). Indeed, many microbiome-wide association studies have established correlation, and sometimes causation, of the gut microbiome in diseases such as multiple sclerosis (
33,
34), rheumatoid arthritis (
35), and systemic lupus (
36). Patients with PTLDS often undergo extensive antibiotic treatment (
5) which likely causes adverse alterations to their microbiomes. A potential parallel exists in autoimmune disease, in which antibiotic use has been linked to an increase in disease frequency because of the dramatic impact antibiotics have on the microbiome (
37).
Since PTLDS symptoms present similarly to diseases in which the microbiome is implicated, we reasoned that the same may be true for the gut microbiome of patients with PTLDS. We analyzed the gut microbiome of subjects with PTLDS from the John Hopkins Lyme Disease Research Center’s Study of Lyme Immunology and Clinical Endpoints (SLICE) cohorts, specifically drawing from a cross-sectional cohort of patients meeting the IDSA-proposed case definition for PTLDS. We performed 16s rRNA gene sequencing on stool samples from this cohort and report a gut microbiome signature associated with PTLDS. These data present a novel biomarker and potential diagnostic tool for PTLDS, as well as suggest a therapeutic avenue for PTLDS.
DISCUSSION
A conservative estimate of the prevalence of PTLDS in the United States in 2020 is 800,000 (
51). The etiopathology of PTLDS is unknown, and the disease lacks suitable laboratory diagnostics and therapeutic methods (
5). We aimed to address this ambiguity by investigating the microbiome, which has been shown to play an important role in diseases with symptom overlap with PTLDS. Through 16S rRNA gene sequencing, we identified alterations in the gut microbiome in a cohort of well-characterized patients with PTLDS compared to two healthy control cohorts and an ICU control cohort. The majority of the ICU patients were on antibiotic treatment at the time of collection; this served as a control for antibiotic use among patients with PTLDS. Using a receiver operating characteristic analysis based on a random-forest classifier model, we found that the microbiome of PTLDS patients is distinct from those of ICU patients and healthy controls.
Blautia species represented three of the five most important features for this classification, and the relative abundance was elevated in the PTLDS cohort compared to those in the healthy and ICU cohorts. Interestingly, an increased relative abundance of
Blautia has been observed in several other diseases. In patients with type 1 diabetes, elevated
Blautia abundance was observed and was correlated with increased IA-2 tyrosine phosphatase autoantibodies, important markers of autoimmunity (
52). Increased
Blautia has also been seen in obesity (
53), Alzheimer’s disease (
54), nonalcoholic fatty liver disease (
55), and multiple sclerosis (
56).
In addition, approximately one-fifth of patients with PTLDS had a relative abundance of
Enterobacteriaceae over 10%, while the average relative abundance of
Enterobacteriaceae in the healthy cohort collected at Northeastern University was 1.14%, consistent with reports of
Enterobacteriaceae in the healthy gut microbiome (
57,
58). Members of the
Enterobacteriaceae family have a proinflammatory lipopolysaccharide (LPS) in the outer membrane and can exacerbate inflammation (
59). A high relative abundance of
Enterobacteriaceae is reported in inflammatory bowel diseases (
60), in metabolic disorders like type 2 diabetes, and in immune diseases like cancers (
49). Concomitantly, we report a depletion of
Bacteroides in the G1 and G3 subsets of the PTLDS cohort which together comprised 64.4% of the PTLDS cohort.
Bacteroides is a common member of the gut microbiome and plays important, symbiotic roles, such as modulation and regulation of the immune system, maintenance of intestinal integrity, and carbohydrate digestion (
46,
47). We have previously reported that γ-aminobutyric acid (GABA) production by human-derived
Bacteroides is widespread, and there is a correlation between brain signatures of depression and fecal
Bacteroides levels in patients with major depressive disorder (
48). Moreover,
Bacteroides organisms are major producers of short-chain fatty acids, which have been shown to support host immune homeostasis both locally and systemically (
61,
62).
While it is possible given the nature of these aberrations that the microbiome is causal or contributory to PTLDS, establishing this relationship is difficult, as no animal model of PTLDS exists. Our results suggest an intriguing opportunity to test causality by using fecal microbiota transplant (FMT) or defined symbiotic consortia to treat patients with PTLDS. FMT has been successfully used to treat
Clostridium difficile infection in patients (
63–65). Furthermore, FMT or the administration of symbiotic bacteria has also been shown to be efficacious in treating multiple sclerosis (
34), Parkinson’s disease (
66), Alzheimer’s disease (
67), and rheumatoid arthritis (
68) in animal models of disease.
As well as suggesting a potential diagnostic tool through the interrogation of the fecal microbiome, the robustness of these results reinforces the validity of PTLDS by showing strong distinctions between the fecal microbiomes of a rigorously curated cohort of patients with PTLDS, patients in the ICU, and healthy controls. Previously reported biomarkers further validate PTLDS and provide an opportunity for the field to progress. These biomarkers include quantitative immune alterations; patients with PTLDS present with elevated levels of the T-cell chemokine CCL19 compared to patients with acute Lyme disease who have returned to health (
15), an increase in the cytokine IL-23 (
16), a decrease in the CD57 lymphocyte subset (
17,
18), and a decreased plasmablast response prior to treatment (
13). Furthermore, a pilot study used [
11C]DPA-713 PET imaging to study cerebral glial activation and found that several brain regions had higher [
11C]DPA-713 binding in patients with PTLDS than in healthy controls (
12). In addition to these biomarkers, Fallon et al. (
69) developed a survey, the General Symptom Questionnaire-30 (GSQ-30), to assess symptom burden and changes; patients with PTLDS reported higher GSQ-30 scores before treatment and maintained these scores until 6 months posttreatment. The GSQ-30 could be a powerful tool to accompany biomarkers like the gut microbiome in PTLDS. The existence of these biomarkers, along with the microbiome signature that we report, contributes to the evidence for a biological basis for PTLDS. It also supports clinical and accumulating research evidence, first published for treatment trials and population-based studies (
70–72), that persistent symptoms after treatment of Lyme disease are common and can significantly impact quality of life. The lack of sensitivity of PTLDS symptoms such as fatigue, pain, and cognitive dysfunction can lead to the conclusion that they are not different than the background noise levels in the general population (
73). However, studies operationalizing the proposed case definition for PTLDS (
11) which utilize standardized symptom and quality-of-life measures have shown that the prevalence and magnitude of these symptoms are often more severe (
4,
74).
Antibiotic use likely affected the microbiome of patients with PTLDS; 9.2% of patients with PTLDS were on antibiotic treatment during the time of collection, and many had extensive antibiotic treatment in their recent health history. While antibiotics likely alter the microbiome composition in patients with PTLDS, our data show that the PTLDS microbiome does not cluster by antibiotic history, and having PTLDS is a better classifier than antibiotic history. Therefore, it is unlikely that antibiotic use alone explains the distinct PTLDS microbiome. Importantly, the microbiome of PTLDS patients was distinct from the microbiome of patients in the ICU. Regardless of the cause of the disruption observed in the microbiome in PTLDS, our data suggest that therapeutic intervention targeting the microbiome may ameliorate PTLDS symptoms. In conclusion, we report that a cohort of patients with PTLDS have microbiomes distinct from those of healthy and ICU controls. Furthermore, we show that through machine learning we can use the microbiome as a high-fidelity indicator of PTLDS. We reinforce the validity of this disease by showing strong distinctions between a rigorously curated cohort of patients with PTLDS and controls.