BCG Immunotherapy Reprograms CNS Immunity and Alters Amyloid Dynamics in Older Adults
Key Insights
BCG vaccination induced trained immunity-like transcriptional responses in cerebrospinal fluid monocytes, with sustained gene expression changes persisting through 12 months post-immunization.
In non-AD participants, CSF Aβ42 levels declined while plasma Aβ42 increased, yielding a 38% reduction in the CSF-to-plasma Aβ42 ratio over 12 months; no such changes occurred in AD participants.
Plasma cytokine responses were more pronounced and sustained in AD participants, while CSF cytokine profiles shifted toward an immunoregulatory state across both groups without inflammatory elevation.
A pair of open-label clinical trials conducted at Massachusetts General Hospital has provided the first evidence in humans that Bacillus Calmette–Guérin (BCG) (search) immunotherapy can remodel the immune environment of the central nervous system (CNS). The findings, published in Communications Medicine, demonstrate that BCG vaccination induces durable, compartment-specific immune reprogramming in older adults and is associated with altered amyloid-β (search) (Aβ) dynamics in individuals without established Alzheimer's disease (search) (AD) pathology.
The year-long study enrolled 23 participants aged 55 and older across two protocols: NCT04507126 (cognitively unimpaired and mildly impaired older adults) and NCT05004688 (individuals with mild cognitive impairment or mild-to-moderate AD). Using k-means clustering of the CSF ptau181 (search)/Aβ42 ratio, participants were stratified into 11 non-AD and 12 AD biomarker-defined groups. All participants received two intradermal BCG vaccinations (Tokyo 172 strain, 1.8–3.9 × 10⁶ CFU) at baseline and one month later, with cerebrospinal fluid (CSF) and peripheral blood collected at baseline, 3 months, and 12 months.
Trained Immunity Signatures in CNS Monocytes
BCG immunotherapy was followed by robust peripheral cytokine recall responses. In PBMCs stimulated ex vivo with heat-killed BCG, IFN-γ production increased significantly, with non-AD participants showing an early peak at 3 months and AD participants exhibiting a delayed but sustained response that converged by 12 months. LPS stimulation induced strong and sustained increases in IL-1β, IL-6, and TNF-α, with all cytokines peaking at 6 months in the overall model.
Single-cell RNA sequencing of 254,292 high-quality cells (106,224 CSF; 148,068 PBMC) revealed that monocytes—the most abundant innate immune population in CSF—exhibited sustained transcriptional reprogramming. LPS stimulation of CSF monocytes at 3 and 12 months post-BCG revealed temporally structured gene expression programs: early-upregulated genes (e.g., HSD11B1, GAB3, IL6) were associated with metabolic remodeling, sustained-upregulated genes (e.g., SERPINB2, LAMP3, AIM2, IL1B, TNFA) implicated persistent cytokine regulation, and late-upregulated genes (e.g., OAS1, ISG15, CXCL10) suggested delayed interferon signaling and leukocyte recruitment.
Notably, while LPS elicited robust transcriptional responses in both PBMC and CSF monocytes, heat-killed BCG stimulation triggered strong gene upregulation in PBMC monocytes but failed to induce measurable responses in CSF monocytes. "Despite being the immunizing agent, HKBCG failed to induce measurable transcriptional responses in CSF monocytes, in contrast to the robust responses seen in PBMCs, indicating compartment-specific constraints on stimulus-specific recall rather than a global failure of immune training," the authors wrote.
Basal Transcriptional Remodeling and Pathway Enrichment
At 12 months post-BCG, gene set enrichment analysis (GSEA) of unstimulated CSF immune cells revealed significant enrichment of immune-metabolic pathways in AD participants. In CSF monocytes, BCG treatment enriched pathways including aerobic respiration and respiratory electron transport, interleukin-1 family signaling, and TNFR2-related NF-κB signaling. Both TNFR2 non-canonical NF-κB signaling and autophagy pathways—suppressed at baseline in AD—showed evidence of restoration following BCG.
CD4⁺ T cells demonstrated similar findings, with significant enrichment of immune-metabolic pathways in AD participants, including FCεRI-mediated NF-κB activation, antigen processing and cross-presentation, and the HSP90 steroid hormone receptor cycle—several of which were downregulated in AD at baseline. CD8⁺ T cells did not exhibit significant pathway enrichment at any time point.
Compartment-Specific Cytokine Responses
BCG immunotherapy produced divergent plasma and convergent CSF cytokine profiles. In plasma, AD participants showed selective early elevations in IL-2 and IL-10, followed by delayed rises in IL-1β, IL-6, and TNF-α—changes not observed in non-AD participants. In contrast, CSF cytokine profiles followed a shared immunoregulatory trajectory across both groups: IFN-γ and IL-2 declined significantly by 12 months (p = 0.010 and p = 0.006, respectively), while IL-1β, IL-6, and TNF-α remained low and showed no consistent longitudinal changes.
"The divergence between peripheral and CNS responses highlights a compartmentalized immune trajectory shaped by disease state," the researchers noted. "In AD participants, the muted CNS cytokine response—despite underlying pathology—may reflect compartment-specific immune adaptation or exhaustion."
Amyloid-β (search) Dynamics Differ by Disease Status
Amyloid-β (search) trajectories diverged significantly by AD status. In non-AD participants, CSF Aβ42 levels declined significantly over 12 months, accompanied by reductions in the Aβ42/40 ratio. Concurrently, plasma Aβ42 increased at 6 and 12 months. The CSF-to-plasma Aβ42 ratio declined by 38.2% over 12 months in non-AD participants, with no meaningful change in the AD group. These findings were supported by significant AD × Time interactions.
When participants were stratified by MoCA score rather than biomarker status, individuals scoring above 25 showed a similar pattern, with a 38.6% reduction in the CSF-to-plasma Aβ42 ratio at 12 months. No significant longitudinal changes were observed in CSF ptau181 (search), NFL, or GFAP levels in either group.
Safety and Tolerability
BCG immunotherapy was well tolerated. One non-serious case of injection site dermatitis was reported, and no other adverse events were attributed to BCG. Four cases of COVID-19 and two non-COVID respiratory infections occurred during the study period but were not considered related to study procedures. All participants developed a visible BCG scar, consistent with expected local vaccine take. Of the 23 enrolled participants, 17 (74%) completed the full 12-month protocol.
Limitations and Future Directions
The authors emphasized several important limitations. The study was exploratory, open-label, and included a modest sample size without a placebo control, precluding causal inference regarding BCG-specific effects. "Replication in larger, placebo-controlled cohorts will be essential to validate these findings," they wrote. The durability of effects beyond one year remains unknown, and the extent to which observed CNS changes reflect reprogramming of resident cells versus recruitment of trained peripheral monocytes or T cells could not be determined.
"Vaccines have traditionally been viewed through the lens of infectious disease prevention," said co-first author Marc Weinberg, MD, PhD. "Although more research is needed, these findings suggest they may also influence biological processes involved in brain aging and neurodegenerative disease."
Senior author Steven Arnold, MD, managing director of the Interdisciplinary Brain Center at the Mass General Brigham (search) Neuroscience Institute, added: "The immune system and the brain may be far more connected than we once thought. The next step is to test this rigorously in larger, controlled studies, particularly in prevention, where the hope would be to preserve brain health before significant Alzheimer's disease (search) develops."
The study was conducted under FDA IND 23605 at Massachusetts General Hospital between December 2020 and January 2024.