A Distinct Microbiome Signature in Posttreatment Lyme Disease Patients

Abstract

Lyme disease is the most common vector-borne disease in the United States, with an estimated incidence of 300,000 infections annually. Antibiotic intervention cures Lyme disease in the majority of cases; however, 10 to 20% of patients develop posttreatment Lyme disease syndrome (PTLDS), a debilitating condition characterized by chronic fatigue, pain, and cognitive difficulties. The underlying mechanism responsible for PTLDS symptoms, as well as a reliable diagnostic tool, has remained elusive. We reasoned that the gut microbiome may play an important role in PTLDS given that the symptoms overlap considerably with conditions in which a dysbiotic microbiome has been observed, including mood, cognition, and autoimmune disorders. Analysis of sequencing data from a rigorously curated cohort of patients with PTLDS revealed a gut microbiome signature distinct from that of healthy control subjects, as well as from that of intensive care unit (ICU) patients. Notably, microbiome sequencing data alone were indicative of PTLDS, which presents a potential, novel diagnostic tool for PTLDS.IMPORTANCE Most patients with acute Lyme disease are cured with antibiotic intervention, but 10 to 20% endure debilitating symptoms such as fatigue, neurological complications, and myalgias after treatment, a condition known as posttreatment Lyme disease syndrome (PTLDS). The etiology of PTLDS is not understood, and objective diagnostic tools are lacking. PTLDS symptoms overlap several diseases in which patients exhibit alterations in their microbiome. We found that patients with PTLDS have a distinct microbiome signature, allowing for an accurate classification of over 80% of analyzed cases. The signature is characterized by an increase in Blautia, a decrease in Bacteroides, and other changes. Importantly, this signature supports the validity of PTLDS and is the first potential biological diagnostic tool for the disease.

Keywords: Lyme disease; diagnostics; microbial communities; microflora; tick-borne pathogens.

FIG 1

Principal-coordinate analysis of unweighted UniFrac distances of 16S rRNA genes. All samples from patients in the ICU, those with PTLDS, and the IT-Healthy and AGP Healthy cohorts are represented. The sequencing platform, Illumina, Ion Torrent, or both, used for each sample is indicated by shape.

FIG 2

A simplified flow chart of the Qiime2 classifier model pipeline used to analyze the microbiome data.

FIG 3

(a) Receiver operating characteristic curve evaluating the ability of a random-forest classifier model to classify PTLDS, healthy, and ICU controls based on the fecal microbiome determined by 16S rRNA gene sequencing. Rounded area under the ROC curve values were 1.00 for all cohorts. Gray lines represent the null model or random chance. (b) Relative abundance plots of the 5 most important features (OTUs) for classification of PTLDS, healthy, and ICU controls based on the fecal microbiome. The first, third, and fifth ranked most important features were Blautia species (OTU IDs 4474380, 4465907, and 4327141); the relative abundance of the Blautia spp. were combined for clarity. S. aureus (OTU ID 446058) and Roseburia sp. (OTU ID 4481427) were the second and fourth most important features, respectively. Bars represent the mean relative abundance plus or minus the standard error of the mean. Statistical significance was determined using Kruskal-Wallis (nonparametric) test followed by Dunn’s multiple comparison. ****, P value < 0.0001. ns, not significant.

FIG 4

(a) Principal-coordinate analysis of unweighted UniFrac distances of 16S rRNA genes. All samples from patients in the ICU (red) and IT-Healthy and AGP Healthy (blue) cohorts are represented. The time since antibiotic (Abx) treatment before sample collection is indicated by symbol color for PTLDS samples. The most recently taken type of antibiotic taken within 1 week to 1 year of sample collection is indicated by the symbol shape for PTLDS samples. (b) Area under the receiver operating characteristic curve (AUROC) evaluating the ability of a random-forest classifier model to classify the fecal microbiome in the healthy, ICU, and PTLDS cohorts, separated into two groups based on antibiotic use within the last 1 week to 1 month or equal to or greater than 6 months.

FIG 5

(a) Relative abundances of Bacteroides in the fecal microbiomes of healthy and ICU control cohorts and patients with PTLDS. Bars indicate the mean plus or minus the standard error of the mean. Statistical significance was determined using the Kruskal-Wallis (nonparametric) test followed by Dunn’s multiple comparison (****, P value < 0.0001; ***, P value < 0.001; *, P value < 0.01). (b) Relative abundances of Blautia versus Bacteroides in the fecal microbiomes of healthy and ICU control cohorts and of patients with PTLDS separated into three groups, G1, G2, and G3. The groups were determined based on the relative abundances of Blautia and Bacteroides: G1, >10% Blautia and <15% Bacteroides; G2, >15% Bacteroides; and G3, <10% Blautia and <15% Bacteroides. (c) Relative abundances of Enterobacteriaceae in the ICU and PTLDS cohorts and the Ion Torrent subset of the healthy control cohort.

FIG 6

Ranked area under receiver operating characteristic curve (AUROC) reported by Duvallet et al. (50) for the classification of the fecal microbiome in each disease versus a healthy control cohort. ART, arthritis; ASD, autism spectrum disorder; CDI, Clostridium difficile infection; CRC, colorectal cancer; EDD, enteric diarrheal disease; HIV, human immunodeficiency virus; IBD, inflammatory bowel disease; LIV, liver disease; NASH, nonalcoholic steatohepatitis; nonCDI, non-Clostridium difficile infection; OB, obesity; PAR, Parkinson’s disease; T1D, type I diabetes.