Part 2/15 : How the Human Immune System Recognizes Bacteriophages: TLR3, TLR9 and cGAS–STING in Phage Therapy

Bacteriophages do not infect human cells, but this does not make them immunologically invisible. The distinction between cellular tropism and immune recognition is fundamental to understanding phage therapy because the mammalian innate immune system does not determine whether a viral particle is capable of completing a productive replication cycle before responding to it. Instead, it detects molecular structures that appear in particular cellular compartments and interprets their presence as evidence of microbial invasion, cellular damage or abnormal nucleic acid localization. A bacteriophage entering a human tissue is therefore exposed to an immune surveillance system organized around molecular patterns and intracellular geography rather than around the biological category of the particle itself. DNA confined within the nucleus or mitochondria is expected, whereas DNA appearing in the cytosol can trigger innate immune signaling. 

The Phage Therapy, Copyright 4.0

Nucleic acids delivered into endosomal compartments can be detected by Toll-like receptors, while viral or microbial proteins exposed at the cell surface may engage membrane-associated receptors before the particle is internalized. The relevant question is consequently not whether the human immune system possesses a specific “phage receptor”, but whether the molecular components of a particular phage become accessible to pattern-recognition receptors in the cellular compartments where those receptors are active.

This issue is particularly important for double-stranded DNA phages, which represent a substantial proportion of therapeutic candidates. Their genomes can contain tens or hundreds of kilobases of foreign DNA, yet this DNA is densely packaged within a protein capsid and may remain physically inaccessible to intracellular DNA sensors. The immunological properties of purified phage DNA cannot therefore be assumed to reproduce those of an intact virion. Toll-like receptor 9, or TLR9, is frequently invoked in discussions of phage immunity because it detects DNA in intracellular vesicular compartments. TLR9 is localized primarily within the endosomal system, where it can encounter microbial DNA following cellular uptake and degradation of extracellular material. Its activation recruits the adaptor protein MyD88 and initiates signaling involving IRAK family kinases and TRAF proteins, ultimately converging on transcription factors such as NF-κB and IRF7. 

Depending on the responding cell, this pathway can promote inflammatory mediator production and type I interferon-associated programs. Experimental evidence has directly implicated TLR9 in specific phage-associated immune responses. Gogokhia and colleagues showed that experimental expansion of intestinal bacteriophage populations altered mucosal immunity and aggravated colitis in mice through a mechanism requiring TLR9 and IFN-γ [1]. The inflammatory phenotype was substantially altered in TLR9-deficient animals, indicating that the phage-associated response could not be explained solely by secondary changes in bacterial abundance. These findings established that bacteriophage exposure can, under defined conditions, participate in a mammalian immune process involving an intracellular DNA-sensing receptor [1].

The presence of phage DNA inside a mammalian cell, however, does not automatically mean that TLR9 is activated. This distinction was illustrated particularly clearly in experiments with bacteriophage T4. T4 is a large virulent Escherichia coli phage with a 168,903-base-pair double-stranded DNA genome and a virion approximately 200 nm in length [2]. Bichet and colleagues exposed human A549 lung epithelial cells and other mammalian cell systems to highly purified T4 particles at concentrations reaching 10^9 phages per millilitre and demonstrated that the virions were internalized and accumulated within membrane-bound intracellular compartments [2]. 

Despite the presence of large numbers of foreign DNA-containing particles inside the cells, intact T4 did not activate the NF-κB-dependent reporter system used in the study and did not induce detectable activation of an IFN-β promoter reporter [2]. The authors proposed that the T4 capsid remained sufficiently stable during intracellular trafficking to prevent the viral genome from becoming accessible to TLR9. The receptor was therefore potentially present in the relevant cell, and the foreign DNA had entered the intracellular environment, but the ligand and the receptor remained physically separated. This observation is central to phage immunology because it demonstrates that cellular internalization and immune sensing are distinct biological events. A phage may enter a mammalian cell without exposing its genome, and the structural integrity of the virion can determine whether an intracellular nucleic acid sensor ever encounters the material it is capable of recognizing.

The same principle applies to cytosolic DNA sensing through the cGAS–STING pathway. Cyclic GMP–AMP synthase, or cGAS, binds double-stranded DNA in the cytosol and catalyses the synthesis of the second messenger 2′3′-cyclic GMP–AMP, commonly referred to as cGAMP [3]. cGAMP binds STING, an adaptor protein associated with the endoplasmic reticulum, and promotes STING trafficking and signaling through TBK1. TBK1 phosphorylates IRF3, allowing the induction of type I interferon-associated genes and a broader antiviral transcriptional program [3,4]. From a purely molecular perspective, bacteriophage DNA is fully capable of acting as a substrate for this pathway if it reaches the cytosol in an accessible form. The T4 study again provided direct experimental evidence for this distinction. Intact T4 phages did not activate cGAS–STING signaling in bone marrow-derived macrophages, whereas DNA extracted from the same phage and experimentally transfected into the cytosol induced a strong STING-dependent IFN-β response [2]. 

The response disappeared in STING-deficient macrophages, confirming that the phage genome itself was capable of activating the pathway when physical access to the cytosolic sensor was experimentally provided [2]. The difference between the intact virion and the purified genome was therefore not the chemical identity of the DNA but its intracellular accessibility. A therapeutic DNA phage must first undergo a cellular process that exposes its genome before TLR9 or cGAS can participate in recognition. Endosomal degradation may permit contact between phage DNA and TLR9, whereas capsid disruption followed by genome escape into the cytosol could theoretically expose the same genome to cGAS. A structurally stable capsid retained inside a vesicular compartment may allow a phage to remain intracellular for prolonged periods without activating either pathway.

Differences in capsid stability, virion architecture and intracellular trafficking may explain why immune responses vary considerably between phages. Bacteriophages are not a homogeneous class of particles. Their capsids differ in geometry, size, protein composition and physicochemical resistance, and their virions differ in surface charge, aggregation behaviour and association with mammalian membranes. These variables can alter the efficiency of cellular uptake and the fate of the particle after internalization. Zamora and colleagues provided an important demonstration of this heterogeneity by comparing four lytic Pseudomonas aeruginosa phages, OMKO1, LPS-5, PSA04 and PSA34, in human airway epithelial cells derived from a cystic fibrosis model [5]. 

The epithelial surface was exposed to 10^9 PFU/mL of each phage, yet the four viral particles showed distinct intracellular kinetics. Less than 0.001% of the initial inoculum was detected as internalized infectious phage after one hour, but subsequent persistence differed substantially between phages [5]. Between one and 24 hours, intracellular infectious titers of OMKO1 and PSA04 increased approximately tenfold relative to their one-hour values, whereas PSA34 showed an approximately 25-fold increase. Twenty-four hours after extracellular phages had been removed, approximately 50% of the initial intracellular infectious OMKO1 and LPS-5 populations remained detectable, compared with approximately 25% for PSA04 and PSA34. By 48 hours, intracellular infective phage levels had fallen to approximately 25% for OMKO1 and LPS-5 and around 10% for PSA04 and PSA34 [5]. The epithelial transcriptional response also differed between phages. 

Although fewer than 2% of the evaluated genes were altered overall, all four phages modified gene expression and a common group of 12 cytokine-signaling genes was affected across the experimental conditions. IFN-β secretion was observed after exposure, while other mediators, including IFN-λ1, IL-8, TNFSF13B and TNFSF8, differed according to the phage tested [5]. These results do not identify a universal receptor responsible for the epithelial phenotype, but they demonstrate that phages administered at the same concentration to the same mammalian cell type are not necessarily processed or sensed identically.

The interpretation of such cytokine responses requires caution because downstream immune readouts do not identify the initiating receptor. An increase in IFN-β does not by itself demonstrate activation of TLR9 or cGAS–STING, since several innate immune pathways converge on interferon regulatory factors. Similarly, NF-κB activation is not a unique molecular signature of one Toll-like receptor. Receptor involvement is best established using receptor-deficient cells, knockout animals, adaptor-deficient systems or selective perturbation of defined signaling intermediates. The interaction between the filamentous Pf bacteriophage of Pseudomonas aeruginosa and mammalian immunity provides one of the clearest examples of this type of mechanistic evidence. 

Sweere and colleagues demonstrated that Pf phage could be internalized by mammalian immune cells and trigger an antiviral immune response dependent on TLR3 and the adaptor protein TRIF [6]. TLR3 is an endosomal receptor classically associated with double-stranded RNA recognition and signals through TRIF rather than MyD88. TRIF-dependent signaling activates TBK1 and IRF3 and can induce type I interferon production, while parallel pathways influence NF-κB-associated inflammatory responses. Pf phage is a single-stranded DNA bacteriophage, but the investigators found that phage-associated RNA generated after internalization contributed to TLR3 activation [6]. 

The resulting type I interferon response reduced TNF production and impaired bacterial phagocytosis, thereby limiting clearance of P. aeruginosa [6]. The phage had not infected the mammalian cell and had not established a mammalian viral replication cycle. Instead, phage-derived nucleic acid became available in an intracellular compartment monitored by an antiviral receptor, and the resulting immune pathway altered antibacterial host defence in a manner that benefited the bacterial host of the phage.

TLR3 and TLR9 therefore illustrate two different versions of the same immunological principle. TLR9 can participate in recognition when phage DNA becomes accessible in the endosomal environment, whereas TLR3 can respond when phage-associated RNA reaches the endosomal compartment under conditions such as those demonstrated for Pf. Neither receptor detects “bacteriophage” as a biological category. Each receptor detects a molecular ligand appearing in a compartment that it monitors. More recent studies suggest that phage recognition may also begin at the cell surface before viral nucleic acids are exposed. TLR2 is a plasma membrane-associated receptor capable of recognizing structurally diverse microbial ligands, particularly lipoprotein-associated molecular patterns. In 2025, Kapoor and colleagues investigated therapeutic mycobacteriophages in macrophages infected with Mycobacterium abscessus and identified the phage Ph17 as a stimulus for TLR2-associated signaling [7]. 

Ph17 exposure was associated with NF-κB activation, increased production of reactive oxygen species and activation of the NLRP3 inflammasome. Caspase-1 activity increased and mature IL-1β secretion followed, while the intracellular growth of the smooth morphotype of M. abscessus was reduced [7]. The investigators also observed fusion between phage-containing vacuoles and bacteria-containing phagosomes, suggesting that intracellular trafficking could facilitate contact between phage particles and an intracellular bacterial target [7]. In this model, the phage did not act exclusively through direct bacterial lysis. Part of the antibacterial phenotype emerged through modulation of the macrophage, linking TLR2-associated recognition, inflammasome activation and intracellular bacterial control.

These studies suggest that the immunological identity of a phage may be distributed across several structural layers. An intact virion presents a protein surface composed of capsid proteins, tail proteins, tail fibres and accessory structural components. These molecular structures may influence membrane association, cellular uptake and receptor engagement before the genome becomes exposed. Following internalization, degradation of the particle can reveal nucleic acids to endosomal receptors. Escape of DNA into the cytosol may permit cGAS recognition. Processing of virion proteins can subsequently generate antigens for adaptive immune responses. The order, efficiency and timing of these events may differ substantially between phages. 

This heterogeneity was emphasized by a 2025 transcriptomic analysis of six Acinetobacter baumannii bacteriophages in A549 epithelial cells and RAW264.7 macrophages [8]. Phage internalization varied according to both the phage and the mammalian cell type, with a podovirus displaying the highest internalization rate and a myovirus the lowest in the experimental panel [8]. RNA sequencing revealed distinct transcriptional profiles that the authors described as phage immunobarcodes. A549 epithelial cells generally developed anti-inflammatory transcriptional signatures after phage exposure, whereas macrophage responses were more heterogeneous. The siphovirus CK02 and podovirus CK21 produced profiles associated with anti-inflammatory signaling, while the myovirus CK12 maintained a more pro-inflammatory transcriptional state [8]. Importantly, neither TLR9 nor cGAS–STING activation emerged as a universal feature of exposure to the six DNA phages. These results support a model in which individual phages possess distinct mammalian cellular phenotypes rather than a common innate immune signature.

Any analysis of phage receptor pathways must also address the problem of bacterial contamination. Therapeutic and experimental bacteriophages are propagated in bacterial cultures, and insufficiently purified preparations can contain bacterial DNA, membrane fragments, proteins and, when Gram-negative hosts are used, lipopolysaccharide. These contaminants are themselves potent ligands for pattern-recognition receptors and can dominate cytokine responses. Lipopolysaccharide activates TLR4 and can induce NF-κB-dependent inflammatory signaling at concentrations that make the contribution of the phage particle difficult to distinguish. A Pseudomonas phage preparation containing residual bacterial material may therefore appear strongly inflammatory even when the purified virion is immunologically inactive. 

This is why mechanistic studies increasingly require extensive purification and multiple process controls. In the T4 experiments, phage preparations were treated with nucleases to remove extracellular DNA and RNA and underwent endotoxin depletion, with final experimental conditions containing less than 1 endotoxin unit per millilitre [2]. The investigators also evaluated phage-free preparation controls and separately examined intact particles, capsid material and purified T4 DNA. Such controls are not merely technical refinements. They determine whether an observed TLR, interferon or NF-κB phenotype can genuinely be attributed to the phage rather than to material derived from the bacterial production host.

The emerging evidence therefore argues against the existence of a single innate immune pathway responsible for bacteriophage recognition. TLR9-dependent intestinal inflammation [1], the absence of TLR9 and cGAS–STING activation by intact intracellular T4 [2], TLR3–TRIF signaling induced by Pf-associated RNA [6], TLR2-associated macrophage activation by Ph17 [7] and phage-specific epithelial and macrophage transcriptomic profiles [5,8] are not necessarily contradictory observations. They describe different phage particles exposed to different mammalian cells and undergoing different intracellular fates. The immune system encounters each phage as a physical nanoparticle with a particular capsid, genome, surface composition and stability. A phage may remain extracellular, undergo macropinocytosis, enter a phagosome, persist within a membrane-bound compartment, reach a lysosomal environment or experience partial structural degradation. Each transition changes the set of receptors capable of accessing the phage and therefore changes the potential immune outcome.

For phage therapy, this receptor-level complexity has direct implications. Therapeutic selection remains appropriately dominated by bacterial criteria such as host range, efficiency of plating, adsorption kinetics, bacterial killing, resistance emergence, genome sequence and manufacturing quality. Nevertheless, a phage intended for repeated intravenous administration, prolonged inhalation or delivery to chronically inflamed mucosal tissue will encounter mammalian cells at concentrations that can reach billions of viral particles. It is increasingly plausible that therapeutic phages possess distinct immunological phenotypes determined by the way their structural proteins and genomes are processed by human cells. 

One phage may remain largely silent because its capsid protects its genome during intracellular trafficking. Another may undergo rapid degradation and expose DNA to TLR9. Cytosolic release of phage DNA may create the conditions required for cGAS activation. A filamentous phage may generate or deliver RNA capable of activating TLR3. Surface or structural components of another phage may influence TLR2-associated signaling in an infected macrophage. The mammalian cell itself introduces further variation because respiratory epithelial cells, intestinal immune cells and tissue macrophages differ in receptor expression, endosomal organization, phagocytic capacity and inflammatory state.

The molecular recognition of bacteriophages should therefore not be reduced to the statement that phages are pro-inflammatory, anti-inflammatory or immunologically neutral. Those categories are too broad to describe a class of viruses with enormous structural and genetic diversity. Mammalian innate immunity appears to respond to the molecular consequences of phage exposure rather than to phage identity itself. 

The receptor sees accessible DNA, RNA or structural material in a defined cellular compartment, and the biological fate of the virion determines whether this encounter occurs. Understanding these pathways may eventually become relevant to therapeutic phage characterization because two phages with similar antibacterial activity can still interact differently with epithelial cells, macrophages or intracellular nucleic acid sensors. The immune system does not need to recognize a bacteriophage as a virus of bacteria. It only needs to encounter one of its molecular components in the wrong place.

Sources

[1] Gogokhia L, Buhrke K, Bell R, et al. Expansion of Bacteriophages Is Linked to Aggravated Intestinal Inflammation and Colitis. Cell Host & Microbe. 2019;25(2):285–299.e8. https://doi.org/10.1016/j.chom.2019.01.008

[2] Bichet MC, Adderley J, Avellaneda Franco L, et al. Mammalian cells internalize bacteriophages and use them as a resource to enhance cellular growth and survival. PLOS Biology. 2023;21(10):e3002341. https://doi.org/10.1371/journal.pbio.3002341

[3] Sun L, Wu J, Du F, Chen X, Chen ZJ. Cyclic GMP–AMP Synthase Is a Cytosolic DNA Sensor That Activates the Type I Interferon Pathway. Science. 2013;339(6121):786–791. https://doi.org/10.1126/science.1232458

[4] Ishikawa H, Ma Z, Barber GN. STING regulates intracellular DNA-mediated, type I interferon-dependent innate immunity. Nature. 2009;461:788–792. https://doi.org/10.1038/nature08476

[5] Zamora PF, Reidy TG, Armbruster CR, et al. Lytic bacteriophages induce the secretion of antiviral and proinflammatory cytokines from human respiratory epithelial cells. PLOS Biology. 2024;22(4):e3002566. https://doi.org/10.1371/journal.pbio.3002566

[6] Sweere JM, Van Belleghem JD, Ishak H, et al. Bacteriophage trigger antiviral immunity and prevent clearance of bacterial infection. Science. 2019;363(6434):eaat9691. https://doi.org/10.1126/science.aat9691

[7] Kapoor H, Maves AM, Bowder MA, Danelishvili L. Phage-mediated TLR2 signaling attenuates intracellular Mycobacterium abscessus survival in macrophages. Scientific Reports. 2025;15:28504. https://doi.org/10.1038/s41598-025-07320-y

[8] Muema CM, Kibii B, Zhong M, et al. Transcriptomic analysis of phage–mammalian cell interaction reveals diverse phage immunobarcodes. iScience. 2025;28:113513. https://doi.org/10.1016/j.isci.2025.113513