Can bacterial toxins train innate immunity through the inflammasome to improve sepsis survival?

Trained immunity is a potentiated second response

Trained immunity (TI) is innate immune memory: after a first stimulus, innate cells such as macrophages respond more strongly to a later challenge, even though the second challenge is unrelated in specificity. The paper uses a strict operational test for this. Macrophages are exposed to a first stimulus, rested, then stimulated with the same amount of lipopolysaccharide (LPS), and the readout is how much more IL-1β, IL-6 and TNFα they express or secrete compared with untrained cells.

That design matters because it separates TI from acute inflammation. In bone marrow-derived macrophages treated with CNF1 for 18 hours and challenged with LPS six days later, the paper reports a dose-dependent increase in IL-1β, IL-6 and TNFα expression, and higher secreted cytokine levels than in cells given the catalytically inactive CNF1 C866S mutant or left untreated. The mutant control is central: it retains the purity and stability of wild-type CNF1 but fails to activate Rac, so the training is tied to the toxin's enzymatic activity rather than to contaminating material.

The training appears in vivo and in human cells

The in vivo test asked whether a single toxin exposure leaves a lasting mark on a relevant cell population. Mice received intravenous CNF1 at 250 ng/kg, and two weeks later splenic macrophages were isolated, cultured for eight more days, and then stimulated with LPS for 24 hours. Macrophages from CNF1-injected mice showed significantly higher IL-1β and TNFα expression than those from mice given the inactive mutant, with a trend toward higher IL-6. Splenic macrophages were chosen because they are sentinels for blood-borne pathogens during systemic infection.

The response was also cell autonomous: training persisted in macrophages removed from the animal and restimulated in culture, so it did not require a continuously present toxin or a live infection. The authors then tested conservation in human monocyte-derived macrophages from healthy blood donors. CNF1 training increased IL-1β, IL-6 and TNFα expression there as well, and this human response was reduced when Bcl2a1 was knocked down. This is conservation in primary human macrophages in culture, not evidence of a human survival benefit.

NLRP3 and IL-1 signaling are required for training

The mechanism runs through the NLRP3 inflammasome, a cytosolic multiprotein platform that matures pro-IL-1β and pro-IL-18 into secreted cytokines. In macrophages lacking Pak1, Nlrp3 or Casp1/11, the CNF1 dose-dependent training of IL-1β, IL-6 and TNFα was blunted. Macrophages lacking the IL-1 receptor (Il1r1) also failed to train, placing IL-1 signaling downstream of inflammasome activation rather than merely alongside it.

Two further experiments support the causal chain. Stimulating macrophages with recombinant IL-1β one week before LPS produced the trained phenotype, while recombinant IL-18 did not. Nigericin, a bona fide NLRP3 trigger, and the bacterial type III secretion system needle protein, which activates the NLRC4 inflammasome, each induced training as well. The authors therefore argue that training may be a general consequence of inflammasome activation and IL-1β secretion, not a CNF1-specific quirk. The paper also notes that nigericin produced a stronger trained response than CNF1, and that the two differ in how IL-1β is released; the authors present this as a hypothesis needing further work, not a settled explanation.

Chromatin remodeling points to BCL2A1 as a required node

TI is expected to involve epigenetic reprogramming, and the paper looked for it directly. ATAC-seq was performed six days after CNF1 exposure and before LPS, measuring chromatin accessibility; RNA-seq was performed 24 hours after LPS, capturing the transcriptional response of previously trained cells. CNF1-trained cells showed a distinct set of upregulated transcripts and changes in chromatin accessibility before the second stimulus, consistent with a pre-set state rather than an acute effect.

From the most accessible and upregulated genes, the authors ran a targeted siRNA screen in macrophages and measured IL-6 secretion as the training readout. Knockdown of Bcl2a1b produced the largest impairment, and knockdown of the other functional family members Bcl2a1a and Bcl2a1d also reduced IL-6, though to different extents. In Bcl2a1 knockout macrophages, CNF1 still activated Rac normally, but the training of IL-1β and IL-6 was significantly reduced, with TNFα following the same non-significant trend. The authors further report co-immunoprecipitation of NLRP3 with endogenous BCL2A1, and mapping that does not require the C-terminal LRR domain. That interaction is presented as a possible common mechanism, not as proof of how BCL2A1 enables training.

Survival improved in one E. coli sepsis model

The functional payoff came from a mouse sepsis experiment. Mice injected with CNF1 at 250 ng/kg were, 14 days later, given 10^6 colony-forming units of E. coli intraperitoneally, and survival was monitored. Toxin-induced TI enhanced host resistance and improved survival in this model. The paper also reports a transient increase in myeloid progenitor cells in the bone marrow after CNF1 injection, which the authors connect to proposed central trained immunity, in which reprogramming in the bone marrow produces trained innate cells.

The scope of that result is narrow and should be kept narrow. It is a single experimental E. coli sepsis model in mice, with one toxin dose and one challenge dose, and survival is the endpoint; the paper does not report a human infection outcome. The mechanistic work in macrophages and the survival experiment are linked by the same CNF1 pretreatment, but the survival benefit is not dissected cell by cell in the evidence presented here.

Transience is framed as a feature, not a defect

The trained state did not persist indefinitely. The authors describe CNF1-induced TI as transient and argue that this may be adaptive: a self-limited form of innate memory could provide temporary protection against recurrent infection while avoiding the risks of prolonged inflammatory reprogramming. They explicitly raise the counter-scenario, noting that persistent reprogramming of myeloid cells could predispose to exaggerated responses to sterile inflammatory stimuli and to inflammatory disease.

This is the paper's most consequential interpretive move, and it is an author interpretation rather than a measured outcome. The evidence for transience comes from the training phenotype declining over the experimental timeline, and the evidence for harm from persistent reprogramming is not tested here. Readers should treat the balance of short-term protection against long-term risk as a framework the data motivate, not a demonstrated trade-off.

A toxin-sensing route to temporary innate memory

The paper's distinctive claim is that sensing a bacterial virulence factor, not a vaccine or a microbial wall component, is sufficient to install inflammasome-dependent innate memory. The chain is specific: CNF1 enzymatic activity, NLRP3 and caspase-1/11, IL-1 receptor signaling, BCL2A1, chromatin remodeling, a potentiated LPS response, and improved survival in one mouse E. coli sepsis model. The human evidence is conservation of the trained phenotype and BCL2A1 dependence in cultured primary macrophages.

The practical priority that follows is comparative rather than therapeutic. Within this paper, CNF1 and nigericin both train through NLRP3 but differ in amplitude and in IL-1β release mode, and IL-1β, but not IL-18, was sufficient to induce training. Those differences, plus the transient time course, are the variables a reader should carry forward when judging whether inflammasome-driven training could be deliberately engaged without tipping into sustained inflammatory risk. The paper establishes the phenomenon and its molecular requirements; it does not establish a dosing window, a clinical strategy, or a human benefit.

Sources used in this answer