CSF Has No Resident Microbiome — 176 Samples Settle the Debate

A growing body of research has suggested that oral bacteria — particularly the periodontal pathogen Porphyromonas gingivalis — might colonize the cerebrospinal fluid (CSF) and contribute to Alzheimer's disease (AD) and related dementias (AD/ADRD). This hypothesis implies that the CSF, long considered a sterile privileged environment, may harbor a resident microbiome accessible to invading oral organisms. Researchers at Columbia University set out to rigorously test this idea using one of the largest CSF microbiome datasets assembled to date.

The team analyzed 176 consecutively archived CSF samples drawn from the biobank of the Neurological Institute of New York. Donors were community-dwelling individuals, 77% of whom were between 61 and 80 years old, and 57% had impaired cognitive status including diagnoses of AD or related dementias. An additional CSF sample from a confirmed bacterial meningitis patient served as a positive control. To ensure technical validity, all DNA extractions included a spike-in of known microbial controls (ZymoBIOMICS Microbial Community Standard), allowing the team to confirm that the workflow could reliably detect bacteria when present.

DNA was extracted from each sample, sequenced on an Illumina MiSeq platform targeting the V3–V4 hypervariable region of the 16S rRNA gene, and processed using standard bioinformatics pipelines including DADA2 for amplicon sequence variant calling and SILVA for taxonomic classification. The spike-in bacteria were consistently detected across samples, confirming assay sensitivity. Streptococcus pneumoniae was correctly identified in the meningitis positive control. However, across the 176 study samples, only negligible numbers of reads mapping to bacterial taxa were detected — a pattern consistent with sporadic contamination or sequencing artifact rather than a true resident microbiome. No enrichment of oral bacteria, including P. gingivalis, was observed in cognitively impaired donors compared to those with normal cognition.

Quantitatively, the bacterial read counts across CSF samples were vanishingly low and showed no statistically meaningful community structure. The diversity metrics and taxonomic profiles were dominated by the spike-in controls and background noise rather than any indigenous bacterial signal. Importantly, there was no difference in the sparse bacterial signal between samples from donors with AD/ADRD versus those with intact cognition, directly challenging the premise that oral bacteria accumulate in the CSF as a function of dementia status or periodontal disease burden.

The findings carry significant implications for the AD–oral microbiome field. While epidemiological associations between periodontitis and AD risk remain plausible and biologically interesting — potentially mediated through systemic inflammation, immune activation, or bacterial metabolite dissemination — the specific mechanism of bacterial colonization of the CSF appears unsupported. The authors conclude that CSF is a genuinely sterile compartment even in elderly, multimorbid individuals, and that prior positive findings in this space most likely reflect contamination during collection, processing, or sequencing. This important negative result should recalibrate the field toward other mechanistic explanations for the oral–brain axis in dementia.