The study also examined whether the microbial community associated with human skin changes between childhood and early adulthood. The team analyzed skin samples from 304 healthy volunteers: 150 children with a mean age of 5.45 ± 0.50 years and 154 young adults with a mean age of 19.48 ± 0.88 years. The overall microbial structure and alpha diversity differed significantly between the groups, and principal-coordinate analysis separated the cohorts.

Bacillota and Actinomycetota abundance did not differ significantly, whereas Bacteroidota was more abundant and Pseudomonadota less abundant in young adults. The study was cross-sectional, so it measured differences between age groups at one point rather than following individual microbiomes over time. At the genus level, Bacteroides was the predominant genus in both cohorts, accounting for an average relative abundance of 15.36% in children and 26.67% in young adults, as determined by 16S rRNA sequencing.

The abundances of Bacteroides and Alistipes increased significantly in young adults, whereas Prevotella 9 was more abundant in children. Several Pseudomonadota genera, including Klebsiella , Moraxella and Pseudomonas , were significantly less abundant in the adult group. In children, Bacteroidota and Pseudomonadota abundance were inversely correlated; the same association was not detected among young adults.

The authors proposed that the age-associated increase in Bacteroides might contribute to the exclusion of potentially pathogenic Gram-negative bacteria, although the human data establish an association rather than causation. To test that possibility, the researchers established a mouse skin co-colonization model using B. fragilis ATCC 25285 and bacterial isolates from skin or clinical specimens. Concurrent topical application of live B. fragilis and K. pneumoniae at a 1:1 ratio significantly reduced K. pneumoniae colonization.

The same bacterium also reduced the loads of Moraxella catarrhalis and Pseudomonas fulva . By contrast, colonization by the Gram-positive organisms Staphylococcus aureus and S. epidermidis was not significantly affected, indicating selectivity for the tested Pseudomonadota in this model. Timing also mattered: pre-inoculating mice with B. fragilis two hours before K. pneumoniae was effective, whereas introducing B. fragilis after K. pneumoniae was not.

Experiments with cultured cells indicated that the effect could not be explained simply by soluble antimicrobial products or impaired bacterial adhesion. Concentrated B. fragilis culture supernatant inhibited several organisms in vitro, but the pattern did not match the mouse results, and S. aureus was inhibited in culture without a corresponding reduction in vivo. In an immortalized human epidermal cell model, B. fragilis did not alter K. pneumoniae , M. catarrhalis or P. fulva adhesion to keratinocytes.

Exposure to K. pneumoniae , but not to B. fragilis alone, increased transcription of TNFα , IL36α and IL10 , while the antimicrobial peptide LL37 was inhibited by either bacterium. These observations pointed away from direct competition or inflammatory signalling in keratinocytes as the main explanation for the reduced colonization seen in mice. The central mechanistic evidence came from macrophages expressing LC3-GFP.

Cells exposed to B. fragilis alone or co-exposed to B. fragilis and K. pneumoniae accumulated more LC3-positive puncta than cells infected with K. pneumoniae alone. Western blotting showed increased LC3-II and reduced p62, a protein that is selectively degraded during autophagic flux, following co-infection. Transmission electron microscopy revealed double-membrane autophagosomes and vacuoles containing cellular debris.

Treatment with 3-methyladenine, an inhibitor of autophagy initiation, abolished the reduction in intracellular K. pneumoniae , M. catarrhalis and P. fulva , linking the antibacterial effect to the autophagic response under the tested conditions. Purified B. fragilis outer membrane vesicles reproduced several of these findings. The vesicles were spherical, had a bilayer membrane and averaged 162.1 ± 18.9 nanometres in size, with a mean zeta potential of −17.768 ± 2.238 millivolts.

Particle-to-protein ratios remained consistent across different protein concentrations, and the vesicle preparation displayed a different protein profile from whole-cell lysates. At 5 micrograms per millilitre, the vesicles increased LC3 puncta, LC3-II accumulation and autophagic structures in macrophages and reduced the intracellular loads of all three tested Pseudomonadota species. The effect disappeared when autophagy was inhibited with 3-methyladenine.

The vesicles did not significantly inhibit bacterial growth when tested directly in vitro, arguing against a simple direct-antibiotic explanation. Evidence from genetically modified mice extended the mechanism to skin colonization. Live B. fragilis lowered K. pneumoniae loads in wild-type C57BL/6J mice but not in homozygous lc3a -deficient mice, with six animals per group.

Heat-killed bacteria were ineffective. Skin immunohistochemistry showed increased LC3A immunoreactivity and reduced p62/SQSTM1 immunoreactivity after treatment with live B. fragilis , consistent with enhanced autophagic activity and substrate turnover. Purified intact outer membrane vesicles also reduced K. pneumoniae colonization in wild-type but not lc3a -deficient mice, with four animals per group.

Neither disrupted B. fragilis vesicles nor S. epidermidis extracellular vesicles had the same effect, suggesting that vesicle integrity and bacterial source were important under the tested conditions. Live B. fragilis and intact vesicles also reduced skin expression of the inflammatory cytokines TNFα and Il36α induced by K. pneumoniae . The authors described the mechanism as a possible explanation for the inverse relationship between Bacteroidota and Pseudomonadota observed in children.

The experiments also show that the response depends on the integrity of the outer membrane vesicles and on host LC3A-associated autophagy. Whether this mechanism operates at the same scale or strength in healthy human skin remains unresolved. Several limitations constrain how broadly the results can be applied.

The human study was cross-sectional, so differences in hygiene, lifestyle, hormonal development and other environmental or developmental factors may have contributed to the observed age-related pattern. Because 16S rRNA sequencing cannot establish bacterial viability, species identity, stable residence or the precise anatomical source of a sequence signal, transient deposition and gut-to-skin transfer could not be excluded. The animal work also relied on the laboratory-adapted B. fragilis ATCC 25285 strain because the researchers did not recover Bacteroides from healthy human skin samples under their culture conditions.

The occlusive dressings used in the mouse model may have increased local hydration and reduced oxygen exposure, potentially favoring the anaerobe. The specific vesicle-associated molecules that initiate autophagy remain unresolved, and their activity may differ among species or strains. Proteomic and lipidomic analysis of vesicles from skin-adapted isolates will be needed to identify the relevant components and determine whether this laboratory mechanism operates in healthy human skin.

Subject of Research: Bacteroides fragilis outer membrane vesicles, autophagy and skin colonization resistance

Article Title: Bacteroides fragilis outer membrane vesicle mediates Pseudomonadota colonization by inducing autophagy on the skin

Article References: Guo, Y., Nurxat, N., Wang, Q., Shen, Z., Zhao, N., Liu, Y., Jian, Y., Ye, B., Li, M., & Liu, Q. (2026). Bacteroides fragilis outer membrane vesicle mediates Pseudomonadota colonization by inducing autophagy on the skin. iScience, 29(11), Article 117969. https://doi.org/10.1016/j.isci.2026.117969

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117969

Keywords: Bacteroides fragilis, outer membrane vesicles, autophagy, skin microbiome, Pseudomonadota, colonization resistance, Klebsiella pneumoniae, host–microbe interactions, murine skin model, 16S rRNA sequencing, Bacteroides, fragilis

Outer membrane vesicles, or OMVs, are roughly 50 to 150 nanometers in diameter, cannot replicate, and carry multiple pathogen-associated molecular patterns that make them potent adjuvants. They have already been engineered to display influenza antigens, including the Matrix two protein extracellular domain, and recombinant OMV vaccines have protected mice against lethal influenza A challenge. Their weakness is thermal: OMVs are unstable above 37 degrees Celsius and are normally stored frozen or freeze-dried long term. The authors report their own observations that OMVs precipitate from solution within hours at room temperature and overnight at 4 degrees Celsius, underscoring how fragile these vesicles are in liquid form.

The team, led by Stephanie M. Curley in the Electrical, Computer, and Biomedical Engineering Department, generated OMVs from engineered ClearColi cells displaying four heterologous M2e epitope sequences fused to the bacterial protein ClyA with a C-terminal histidine tag. Vesicles were purified by vacuum filtration and ultracentrifugation, quantified by total protein content, and characterized by dynamic light scattering to confirm particle size distribution and concentration. Films were then prepared by drop casting 750 microliters of formulation into 12-well plates and drying for at least 24 hours at room temperature.

Formulation engineering was a central part of the study. The researchers screened solvents including water, phosphate-buffered saline, sodium chloride, and Tris buffers, finding that 20 millimolar Tris at pH 8.8 produced the most consistent film formation across excipients. Two molecular weights of hydroxypropyl methylcellulose, or HPMC, were tested, but only the 10 kilodalton version dissolved fully at the concentrations needed; the 86 kilodalton polymer failed to dissolve, and 1 percent HPMC yielded thin, brittle films while 5 percent and above proved unwieldy. The final films used 2.5 percent HPMC, combined with poly(vinyl) alcohol, poly(ethylene) glycol, sorbitol, glycerol, or the amphiphilic compound PMAL, and OMVs were loaded to a reconstituted total protein concentration of 0.03 milligrams per milliliter.

Physical characterization revealed how the vesicles integrate into the film matrix. Fourier transform infrared spectroscopy showed broader and more intense peaks in the approximately 3500 per centimeter region, which the authors attribute to enhanced hydrogen bonding between hydroxyl groups of the HPMC matrix and functional groups on OMV surface proteins and lipopolysaccharides. Hydrophilic, hydroxyl-dense additives such as PVA, sorbitol, and glycerol further strengthened these interactions. Scanning electron microscopy showed spherical protrusions on the surface of OMV-containing films that were absent from control films, and cross-sectional imaging established a film thickness of approximately 50 to 100 micrometers, thin enough for rapid dissolution.

Long-term stability testing produced the study’s most consequential findings. Four film types were stored at room temperature for up to eight weeks, then reconstituted in phosphate-buffered saline and assessed by particle count and protein analysis. Films containing PMAL, an amphiphilic compound often used to stabilize biomolecules, degraded the OMV structure completely within four weeks, an outcome the authors link to its detergent-like nature, since surfactants are routinely used to degrade OMVs in laboratory assays. In contrast, films containing PVA with sorbitol or PVA with glycerol maintained consistent particle counts over the full eight weeks, with only a slight reduction in the PEG-containing film at the final time point.

Protein stability told a more nuanced story. Dot blots probing the histidine-tagged ClyA-M2e antigen on intact vesicles showed that PMAL-containing films lost detectable antigen by two weeks, earlier than the four-week point at which particle degradation became apparent, suggesting that surface proteins interact directly with film components and degrade before the vesicle structure collapses. SDS-PAGE analysis indicated protein degradation in PMAL films as early as one week after film formation, although the low antigen concentration of approximately 0.03 milligrams per milliliter limited the resolution of the expected 50 kilodalton ClyA-M2e band. Western blotting, which captures antigen both within and on the surface of vesicles, showed that ClyA-M2e content remained relatively consistent across eight weeks in sorbitol-containing films and, to a large extent, in PEG-containing films, with PEG formulations providing the most protein stability overall and some reduction visible in glycerol-containing films at eight weeks.

Dissolution testing addressed the platform’s intended route of administration. Because only the first two formulations, HPMC-PVA-sorbitol and HPMC-PVA-glycerol, were mechanically stable enough to be removed intact from the casting molds, these films were tested in solutions mimicking the salivary environment, containing alpha-amylase, lipase, and lysozyme alone and in combination in cell culture medium at 37 degrees Celsius. All dissolution times fell below roughly 400 seconds, with most films dissolving in under five minutes, and alpha-amylase produced the fastest dissolution of the three single enzymes. The authors note this is unexpected, since cellulose compounds inhibit alpha-amylase activity, and suggest instead that amylase binding to the cellulose surface may physically disrupt the film structure and accelerate dissolution.

The implications reach beyond the laboratory bench. Lower respiratory infections are currently the fourth leading cause of death worldwide, and influenza alone causes an estimated 389,000 deaths annually. Systemic injection often fails to promote immunity at the oral and nasal mucosa, the primary site of infection for respiratory pathogens, whereas mucosal vaccines can induce both local and systemic responses and eliminate needles entirely. Prior modeling work cited in the study predicts that improving the thermostability of a single pentavalent vaccine would make it available in almost every part of developing countries, and thin film platforms have previously achieved up to 100 percent recovery of infectious and immunogenic agents while preserving active and inactive viruses, bacteria, and proteins without refrigeration.

The authors are candid about limitations. Background scattering from reconstituted HPMC particles complicated dynamic light scattering analysis, and they suggest nanoparticle tracking analysis with an instrument such as the NanoSight300, which counts individual particles, may be better suited to future work. Formulations three and four frequently tore during removal from the casting molds, and the team is investigating silicone molds, which have provided a greater success rate in film removal and could yield films with fewer defects and more uniformity. The next step is animal testing to determine whether OMV immunogenicity remains consistent after long-term room temperature storage. If it does, a thin film that dissolves on the cheek within minutes and survives eight weeks on a shelf could bring mucosal nanovaccines to populations that ultra-cold freezers have left behind.

Subject of Research: Biomaterials Science

Article Title: Outer membrane vesicle loaded thin films as a stable room temperature vaccine platform

Article References: Randall, N. M., Squissato, J., Numanoglu, C., Dubois, W., Lee, H., & Curley, S. M. (2026). Outer membrane vesicle loaded thin films as a stable room temperature vaccine platform. Frontiers in Biomaterials Science, 5, Article 1913356. https://doi.org/10.3389/fbiom.2026.1913356

Image Credits: AI Generated

DOI: 10.3389/fbiom.2026.1913356

Keywords: outer membrane vesicles, thin films, vaccine thermostability, cold chain, mucosal vaccination, biomaterials, Outer, membrane, vesicle, loaded, thin, films

Citrobacter species are facultative anaerobic, Gram-negative bacteria of the order Enterobacterales, common in water, soil, food, and the intestines of humans and animals. Although they are not generally regarded as frequent pathogens, certain species cause urinary tract and bloodstream infections that can be acquired in hospitals or in the community. Immunocompromised patients and neonates are particularly vulnerable. Reported cases include a bloodstream infection caused by C. braakii in an immunocompromised allogeneic stem cell transplant patient, and an analysis of nine newborns with C. freundii sepsis in which one infant died within seven hours and three developed meningitis, with multidrug resistance present in every isolate. Against this backdrop, hospitals act as ecological niches that select for resistant bacteria, making hospital sewage a predictable reservoir of multidrug-resistant organisms.

The Greifswald strain was actually discovered by accident. It appeared within the inhibition zone of an imipenem-sensitive C. portucalensis isolate during routine testing, a hint that something in the sample could survive the antibiotic. Imipenem belongs to the carbapenems, β-lactam antibiotics with broad-spectrum activity against pathogens including Pseudomonas, streptococci, Neisseria, and Haemophilus. Carbapenems resist hydrolysis by most β-lactamases and are typically reserved for severe nosocomial infections caused by extended-spectrum β-lactamase-producing Enterobacteriaceae and Pseudomonas aeruginosa. The World Health Organization now lists carbapenem-resistant Enterobacterales among the bacteria most threatening to human health, which makes any new isolate carrying such resistance worth dissecting at the molecular level.

Resistance to carbapenems can arise through several mechanisms. Some Enterobacterales carry carbapenemases, β-lactamases capable of directly cleaving the drugs. Others combine porin loss or dysfunction with ESBL or AmpC β-lactamase production, modify their penicillin-binding proteins so the antibiotics bind poorly, reduce drug uptake through altered outer membrane proteins, pump drugs out via systems such as AcrAB-TolC, or remodel their biofilms. To determine which of these strategies GW-Imi-1b1 employs, the researchers subjected it to antimicrobial susceptibility testing using the Kirby-Bauer disc diffusion method and E-tests, following EUCAST guidelines, alongside microscopy, motility assays, and comprehensive proteomic analysis.

The results revealed a broad resistance spectrum. GW-Imi-1b1 proved resistant to tobramycin, ceftazidime, chloramphenicol, the fluoroquinolones ciprofloxacin, levofloxacin, and ofloxacin, and ertapenem, with an average ertapenem minimum inhibitory concentration of 12 µg/mL. It remained susceptible only at increased exposure to imipenem and meropenem, and fully susceptible to amikacin and trimethoprim/sulfamethoxazole. Proteomic analysis identified a striking culprit: a plasmid-encoded GES-5 β-lactamase, a carbapenem-hydrolyzing class A enzyme, ranked among the ten most abundant proteins in the whole-cell proteome even when no antibiotic was present. When the team grew the strain in meropenem concentrations of 0.064, 0.125, or 20 µg/mL, the abundance of this enzyme did not change significantly, indicating constitutive rather than inducible expression. A carbapenem-sensitive C. portucalensis strain used as a control lacked the enzyme and was readily inhibited by low meropenem concentrations, supporting the link between GES-5 and the resistance phenotype.

The plasmid location of the bla_GES-5 gene raises particular concern. The researchers confirmed the gene’s position on plasmid CP115731 by plasmid isolation and PCR. Because plasmids can move between bacteria through conjugation or transformation, this resistance mechanism may be transferable within the wastewater microbiota. Case reports have already documented plasmid-encoded GES-5 spreading among different species, including E. coli, Klebsiella oxytoca, and Enterobacter cloacae, across patients and hospital settings, and conjugation experiments with a GES-5-bearing plasmid in Pseudomonas knackmussii increased resistance to several carbapenems. The proteome also revealed all subunits of the AcrAB-TolC efflux pump, which exports multiple antibiotics and likely contributes additional resistance capacity beyond the β-lactamase alone.

Beyond resistance, the characterization uncovered features relevant to virulence. Scanning electron microscopy showed rod-shaped cells roughly 2 µm long and 0.5 µm wide that produce outer membrane vesicles, or OMVs, spherical blebs shed from the bacterial surface. OMVs in other Gram-negative pathogens such as Helicobacter pylori, P. aeruginosa, and Citrobacter rodentium carry lipopolysaccharides, peptidoglycans, enzymes, and toxins that can trigger inflammatory responses. They may also ferry antibiotic resistance genes to susceptible cells, and in some bacteria they act as decoys that bind phage particles and shield the cell from infection. Whether GW-Imi-1b1’s OMVs serve these roles remains unresolved; the authors note that sequencing the vesicle cargo and comparing it with the bacterial genome would be needed to identify potentially transferred genes. Motility assays, meanwhile, showed the strain is completely non-motile in swimming, swarming, and twitching tests, and electron microscopy confirmed the absence of flagella, despite the genome containing 92 putative motility-related domains. No protease secretion was detected either, in contrast to some other Citrobacter species.

The solution half of the story came from the same wastewater. In November 2023, the team collected water from the inflow to the Greifswald sewage treatment plant, treated it with chloroform to eliminate bacteria, and incubated it with GW-Imi-1b1. Plaques on a double agar layer lawn yielded a pure phage after three rounds of isolation. Sequencing revealed a circular double-stranded DNA genome of 39,389 base pairs with a GC content of 50.04 percent, closely matching its host’s 51.81 percent, a similarity consistent with close co-evolution. The genome encodes 57 potential proteins, 26 of unknown function, and contains no tRNAs, CRISPRs, or integration and excision modules, marking it as exclusively lytic. Electron microscopy showed an icosahedral capsid about 54 nm in diameter with a short, non-contractile tail of roughly 12 nm, podovirus morphology that places the phage in the class Caudoviricetes and the genus Kayfunavirus.

Functional testing showed the phage is a specialist with a fast lifestyle. Its lytic host spectrum was restricted to GW-Imi-1b1 itself; none of the other tested Citrobacter, Klebsiella, or E. coli strains, whether ESBL-producing or carbapenem-resistant, were lysed. Adsorption was nearly complete within five minutes, the latent period lasted about five minutes, and complete phage release occurred within the following ten, giving an infection cycle of roughly 20 minutes and a burst size of 12 plaque-forming units per cell. That burst size is small compared with the 35 to 55 PFU per cell typical of other Kayfunavirus phages, but small bursts are often paired with short lysis times, allowing rapid reinfection of new host cells. The phage also proved environmentally robust, retaining infectivity from 4 to 40 °C and across pH values from 6 to 10, though it lost infectivity at 50 and 60 °C and at acidic pH, with no infectious particles remaining at pH 2.

The narrow host range cuts both ways. It spares the rest of a microbiome, an advantage over broad-spectrum antibiotics, but it also means monophage therapy would fail against polymicrobial infections or mismatched strains, a limitation that phage cocktails can partially overcome. The authors caution that all assays were performed in vitro, and that preclinical trials testing ex vivo and in vivo lytic activity and potential cytotoxicity toward eukaryotic cells are required before vB_CbrP_HGW_001 could be considered a treatment option. Still, the study demonstrates a practical workflow: sewage, the very environment that breeds resistant bacteria, can be mined for phages that attack them, and global proteomics can expose the actively expressed resistance and virulence factors of problematic pathogens. Combining such phages with antibiotics, an approach that has reduced hospitalizations in enterococcal infections, may further extend the toolkit against multidrug-resistant organisms. For now, the Greifswald pair stands as a model system showing that for every resistance problem emerging in wastewater, a solution may be waiting in the same water.

Subject of Research: Characterization of a carbapenem-resistant Citrobacter braakii strain and its lytic bacteriophage isolated from hospital sewage water

Article Title: Problems bring along their solutions: characterization of a carbapenem-resistant Citrobacter braakii strain and its lytic phage both isolated from sewage water

Article References: Pukall, J., Rode, W., Albrecht, J., Oerters, S., Ganbarzade, A., Krombholz, J., Schaufler, K., Eger, E., Schlüter, R., Hinzke, T., Zühlke, D., & Sievers, S. (2026). Problems bring along their solutions: characterization of a carbapenem-resistant Citrobacter braakii strain and its lytic phage both isolated from sewage water. Frontiers in Bacteriology, 5, Article 1864514. https://doi.org/10.3389/fbrio.2026.1864514

Image Credits: AI Generated

DOI: 10.3389/fbrio.2026.1864514

Keywords: Citrobacter braakii, carbapenem resistance, GES-5 beta-lactamase, bacteriophage, phage therapy, wastewater, Kayfunavirus, antimicrobial resistance, outer membrane vesicles, proteomics, Enterobacterales, hospital sewage

The research, coordinated by first authors Alex Kayongo, Theda Ulrike Patricia Bartolomaeus and Till Birkner, with Sofia K. Forslund-Startceva as senior author, enrolled 200 adults stratified by both HIV and COPD status. Crucially, 160 of these participants were followed for two years, allowing the team to move beyond a single snapshot and examine how airway microbial communities change over time. This longitudinal design matters because cross-sectional studies, which sample individuals at only one point, cannot distinguish stable disease-associated features from ordinary temporal fluctuation, a limitation that has hampered much of the previous microbiome literature in respiratory medicine.

Methodologically, the study stands out for the breadth of its integrated multi-omics framework. The researchers combined 16S rRNA gene profiling of sputum to characterise bacterial communities, quantification of bacterial outer membrane vesicles, which are small membrane-bound particles shed by bacteria that can carry inflammatory cargo, and detailed profiling of airway cytokines, the signalling molecules that orchestrate inflammation. On the host side, they deployed cytometry by time of flight, or CyTOF, to phenotype immune cells at high resolution, bulk transcriptomics of induced sputum, bronchoalveolar lavage and peripheral blood, and single-cell RNA sequencing of bronchoalveolar lavage cells, which was then used to guide computational deconvolution of cell-type proportions in bulk samples.

Perhaps the most striking finding is what the team did not find. Contrary to expectations, sputum microbial communities were shaped predominantly by temporal ecological turnover, meaning that the composition of the airway microbiome changed substantially over time within individuals, rather than being primarily determined by HIV status, COPD status, or the combination of the two. Microbial diversity, microbiome instability, outer membrane vesicle burden and inferred airway cellular composition were largely preserved across the clinical groups. This suggests that the airway microbial ecosystem in this cohort is more resilient, or at least more variable in ways unrelated to disease, than many researchers had assumed.

Yet the study did not come up empty-handed. When the authors looked more closely, they identified disease-specific microbial signatures associated with HIV-associated COPD, indicating that particular bacterial taxa or community patterns do track with the combined condition even though the overall architecture of the microbiome does not. Alongside these microbial signatures, the team detected inflammatory transcriptional remodeling that was concentrated in induced sputum, the sample type most directly reflecting the airway environment. In other words, the host’s gene expression programs in the airways, rather than gross microbial composition, appear to be where the immunological fingerprint of HIV-associated COPD is most clearly written.

The emphasis on non-linear host-microbe interactions is central to the study’s conceptual contribution. Traditional analyses often test for simple linear associations between the abundance of a microbial taxon and a host measurement. Such approaches can miss relationships that only emerge above certain abundance thresholds, that depend on the presence of other microbes, or that vary across immune contexts. By applying integrated analyses designed to capture these selective and non-linear interaction networks, the researchers were able to link discrete microbial taxa with specific inflammatory, immune and transcriptional programmes, mapping a web of relationships that would have been invisible to simpler statistical models.

These findings carry substantial implications for how microbiome research in respiratory disease should be interpreted and designed. If temporal turnover dominates microbial variation, then single-time-point studies risk misattributing random ecological drift to disease. Conversely, if the most robust disease signals reside in host transcriptional responses and in selective, context-dependent interactions with particular taxa, then therapeutic strategies aimed at broadly reshaping the microbiome, such as indiscriminate antibiotics, may be poorly matched to the underlying biology. More targeted interventions, informed by the specific host-microbe networks identified here, become a more plausible long-term goal.

The study is also notable for its setting and its collaborative structure. Conducted in Uganda with clinical sites including Makerere University Lung Institute, Nakaseke General Hospital and the African Community Center for Social Sustainability, and with analytical partners in Berlin, Vienna, Graz, Heidelberg, London, Miami and Baltimore, the work addresses a genuine gap in global health research. Most large-scale airway microbiome studies have been conducted in high-income countries, and the ecology of the airway in African populations, where HIV prevalence, environmental exposures and prior infectious disease histories differ substantially, has been poorly characterised. The funding came from the US National Institutes of Health Fogarty International Center, the European and Developing Countries Clinical Trials Partnership supported by the European Union, Novartis Global Health, the German Research Foundation, the German Federal Ministry of Education and Research and the European Union’s Horizon 2020 programme.

For clinicians managing the growing population of people living with HIV who develop chronic lung disease, the study offers both reassurance and a caution. The relative preservation of microbial diversity and immune composition suggests that HIV-associated COPD in this cohort is not simply a state of catastrophic ecological collapse in the airways. At the same time, the identification of disease-specific microbial signatures and concentrated inflammatory remodeling in induced sputum points to measurable molecular changes that could, with further validation, serve as biomarkers for disease activity or treatment response. Because the inflammatory signal was strongest in the airway compartment itself, sputum-based monitoring may prove more informative than peripheral blood markers.

The authors note that the article was shared early to provide faster access to peer-reviewed, accepted research, with a final version of record to follow, and the work is published open access under a Creative Commons Attribution licence. As follow-up studies extend this framework to larger cohorts, longer follow-up periods and interventional settings, the central lesson of this research is likely to endure: understanding chronic lung disease in people living with HIV requires looking not just at which microbes are present, but at when they change, how they interact with host immunity in non-linear ways, and where in the airway the resulting inflammatory programs are actually expressed.

Subject of Research: Airway microbiome ecology and host immune interactions in HIV-associated COPD

Article Title: Microbial ecological instability and non-linear host interactions shape airway inflammation in HIV-associated COPD

Article References: Kayongo, A., Bartolomaeus, T. U. P., Hecker, J., Birkner, T., Vila, E., Ntayi, M. L., Markó, L., Löber, U., Olweny, G., Kyalo, E., Kyamulabi, D., Ndawula, J., Ssengooba, W., Nabatanzi, R., Kyobe, S., Nassali, F., Kigozi, E., Kalyesubula, R., Munana, R., … Forslund-Startceva, S. K. (2026). Microbial ecological instability and non-linear host interactions shape airway inflammation in HIV-associated COPD. Nature Communications. https://doi.org/10.1038/s41467-026-78400-4

Image Credits: AI Generated

DOI: 10.1038/s41467-026-78400-4

Keywords: HIV, COPD, airway microbiome, 16S rRNA profiling, outer membrane vesicles, CyTOF, single-cell RNA sequencing, cytokines, inflammation, multi-omics, Uganda, host-microbe interactions

The protein at the center of the story is a B-type dye-decolorizing peroxidase, dubbed DypB_(2985), produced by the soil bacterium Pseudomonas putida. Peroxidases of this family are enzymes that break down reactive oxygen species and can also degrade dye molecules and other aromatic compounds, which makes them interesting both as tools of bacterial physiology and as candidates for biotechnological applications such as bioremediation. What makes DypB_(2985) puzzling is that it lacks the canonical N-terminal signal peptide that would normally route a protein through the Sec pathway, the workhorse system that threads unfolded proteins across the bacterial inner membrane and into the periplasm, the compartment between the inner and outer membranes of Gram-negative bacteria. Without that label, the standard models of bacterial export offer no obvious route for the enzyme to reach the cell surface.

Liang and colleagues approached the problem by systematically dissecting the secretion process in P. putida, and their experiments converged on two key players in the outer membrane: a lipoprotein designated Lpp_(1528) and an outer membrane protein called OmpW_(4836). Lipoproteins and OmpW-family proteins are abundant, well-studied components of the outer membrane in Gram-negative bacteria, where they typically help maintain client proteins in a folded state or form pore channels that allow molecules to pass through. The new work shows that these proteins do something more surprising in the case of DypB_(2985): they act as sequence-specific adaptors that shepherd a signal-peptide-free cargo through two successive stages of export, each stage relying on a different membrane-bound system.

The first stage begins in the cytoplasm. Lpp_(1528), despite being an outer membrane lipoprotein itself, is initially synthesized with its own Sec signal peptide, which routes it through the Sec machinery into the periplasm before it is anchored in the inner leaflet of the outer membrane. From there, the researchers found, Lpp_(1528) performs a remarkable trick: it recognizes a hydrophobic region at the C-terminus, the tail end, of DypB_(2985) while the peroxidase is still inside the cytoplasm. This recognition allows the enzyme to be coupled to the Sec translocation machinery for transport across the inner membrane, even though DypB_(2985) has no canonical N-terminal signal peptide of its own. In effect, the lipoprotein substitutes for the missing address label, physically linking the cargo to the export channel that would otherwise ignore it.

Once the pair has crossed the inner membrane, the partnership appears to dissolve. The study indicates that Lpp_(1528) and DypB_(2985) dissociate in the periplasm, leaving the peroxidase free in that compartment and the lipoprotein free to return to its anchoring duties. This handoff point is where the second stage of the relay begins. A different outer membrane protein, OmpW_(4836), now takes over. OmpW_(4836) binds a hydrophobic region at the opposite end of the enzyme, the N-terminus, and uses that interaction to direct periplasmic DypB_(2985) into outer membrane vesicles, or OMVs, the small spherical blebs that Gram-negative bacteria constantly pinch off from their outer membranes and release into the surrounding environment.

Outer membrane vesicles have attracted intense interest in recent years because they are far more than cellular debris. Bacteria use them to deliver toxins to rival cells, to package enzymes that digest complex nutrients, to shed antibiotics and signaling molecules, and to interact with host organisms during infection. Because OMVs bud from the outer membrane, proteins that end up inside them are effectively secreted: when the vesicle travels away from the cell and eventually ruptures or delivers its contents, the cargo reaches the extracellular space. What the new study adds is a concrete mechanism for how a specific protein gets sorted into these vesicles rather than remaining free in the periplasm. The recognition of the N-terminal hydrophobic region of DypB_(2985) by OmpW_(4836) provides that sorting specificity, ensuring that the peroxidase is loaded into vesicles destined for export rather than left behind.

Taken together, the findings sketch a stepwise pipeline that is elegant in its division of labor. One outer membrane protein, Lpp_(1528), solves the entry problem by tethering a signal-less cargo to the Sec machinery at the inner membrane. Another, OmpW_(4836), solves the exit problem by loading the cargo into vesicles at the outer membrane. The authors describe these proteins as molecular tethers, and the description fits: each one bridges a gap between a protein that lacks the standard credentials for export and a transport system that requires them. The crosstalk between the Sec pathway and the OMV system, mediated by these tethers, constitutes what the researchers characterize as an unconventional secretion route, one that expands the known repertoire of ways bacteria move proteins beyond their envelopes.

The implications reach beyond a single enzyme in a single species. Non-classical secretion, the export of proteins without signal peptides, has been documented across many bacterial groups, and the proteins involved often perform important extracellular functions, from oxidative stress defense to nutrient acquisition. Understanding the machinery behind one such case provides a template for hunting down similar mechanisms elsewhere. The stepwise logic uncovered here, in which distinct accessory proteins handle cytoplasmic entry and vesicular exit, suggests that other signal-less proteins may likewise depend on outer membrane components acting as adaptors, a possibility that could reframe how microbiologists interpret the growing catalog of proteins found in bacterial secretomes without obvious export signals.

There is also a practical dimension. Outer membrane vesicles are increasingly viewed as versatile biotechnological platforms: they can be engineered to display antigens for vaccine development, to carry enzymes for industrial catalysis, or to deliver therapeutic molecules. A major challenge in this field has been controlling which proteins end up inside the vesicles, since cargo loading has often appeared somewhat arbitrary. The OmpW_(4836)-dependent sorting mechanism identified in this study offers a handle for that problem. If a hydrophobic recognition sequence is what directs a protein into OMVs, then attaching similar tags to proteins of interest could, in principle, allow researchers to program the vesicular packaging of chosen cargoes, turning a naturally occurring bacterial export route into a customizable delivery system.

For P. putida itself, the ability to secrete a peroxidase without a signal peptide likely reflects an evolutionary solution to a specific survival need, allowing the bacterium to deploy oxidative enzymes outside the cell where they can help manage reactive compounds in the environment. The study by Liang, Zhu, and Lin, published on September 15, 2026, in PLOS Genetics, does not claim to have found the only route for unconventional secretion, and the authors frame their work as expanding understanding of non-classical protein export mechanisms and bacterial survival strategies rather than closing the book on them. But it does demonstrate, with unusual mechanistic clarity, that the boundary between the classical and unconventional secretion worlds is blurrier than once thought. A protein with no signal peptide can still ride the Sec pathway, provided a lipoprotein is willing to vouch for it, and it can still reach the outside world, provided an outer membrane protein is willing to pack it into a departing vesicle. In the bacterial cell envelope, it seems, the right molecular handshake can substitute for any address label.

Subject of Research: Unconventional secretion of a signal-peptide-free peroxidase in Pseudomonas putida via outer membrane protein-mediated crosstalk between the Sec pathway and outer membrane vesicles

Article Title: Outer membrane proteins mediate unconventional secretion of Pseudomonas peroxidase through crosstalk between the Sec pathway and outer membrane vesicles

Article References: Liang, C., Zhu, W., & Lin, L. (2026). Outer membrane proteins mediate unconventional secretion of Pseudomonas peroxidase through crosstalk between the Sec pathway and outer membrane vesicles. PLOS Genetics, 22(9), e1012311. https://doi.org/10.1371/journal.pgen.1012311

Image Credits: AI Generated

DOI: 10.1371/journal.pgen.1012311

Keywords: Pseudomonas putida, protein secretion, Sec pathway, outer membrane vesicles, lipoprotein, OmpW, peroxidase, DypB, unconventional secretion, Gram-negative bacteria, outer membrane proteins, molecular tethers

Bacteroides fragilis is one of the most abundant bacterial species colonizing the human colon, and its relationship with its host is famously double-edged. Certain strains carry a toxin gene and are strongly associated with promoting colorectal tumors, producing the so-called Bacteroides fragilis toxin that damages DNA and drives inflammation. Yet other strains, lacking this enterotoxin, have been linked in previous studies to beneficial or protective effects, including the suppression of tumor growth in experimental models. The Chinese team worked with a human-derived non-enterotoxigenic strain precisely because of this paradox: if the same species can both promote and restrain cancer, then the specific molecules each strain produces become the decisive variable. Identifying those molecules is the first step toward understanding the mechanism and, eventually, toward developing microbiota-based therapies.

The methodological heart of the study is an approach known as activity-guided fractionation, a strategy long used in natural product chemistry to find the active needle in a haystack of molecules. The researchers first grew their bacterial strain in liquid culture and collected the conditioned supernatant, the cell-free broth containing everything the bacterium had secreted into its environment. They then tested this crude material against HT-29 cells, a well-established human colorectal cancer cell line, to confirm that the supernatant indeed inhibited cell proliferation. With that baseline activity established, they began to chop the complex mixture into progressively simpler fractions, testing each one for anti-proliferative activity along the way. Only fractions that retained activity were carried forward, ensuring that the analytical effort concentrated on the parts of the mixture most likely to contain the bioactive molecules.

The fractionation pipeline was deliberately multi-stage. Ultrafiltration first separated the supernatant by molecular weight, dividing proteins and other large molecules from smaller metabolites. Solvent extraction then partitioned the material according to chemical properties, distinguishing hydrophilic components from those that dissolve in organic solvents. Finally, chromatography provided a finer resolution, separating molecules on the basis of their interactions with a stationary phase. At each step, the researchers returned to the HT-29 cell assay, tracking where the anti-proliferative activity went. This iterative process, in which biological activity is used as a compass for chemical purification, is what distinguishes the study from a purely descriptive catalog of bacterial proteins. It converts an overwhelming proteomic inventory into a focused set of suspects.

Once the active fractions had been identified, the team turned to liquid chromatography coupled with tandem mass spectrometry, or LC-MS/MS, to determine which proteins were present in each. In this technique, proteins are digested into peptides, separated by liquid chromatography, and then fragmented inside the mass spectrometer; the resulting fragmentation patterns allow researchers to infer the identity and sequence of the original proteins with high confidence. Comparative proteomic analysis across the active and inactive fractions revealed 41 candidate proteins consistently associated with the bioactive material. The comparison is the critical element: proteins found in active fractions but absent from inactive ones, or enriched in the former, become candidates, while proteins present everywhere are unlikely to explain the specific biological effect.

Functional annotation of these 41 candidates revealed striking patterns. Many of the proteins are involved in iron acquisition, a vital survival function for bacteria living in the iron-restricted environment of the gut, where hosts actively withhold this metal from microbes. Others belong to ATP-binding cassette transport systems, molecular pumps that use the energy of ATP hydrolysis to move substrates across bacterial membranes, and to protein secretion pathways, the machinery by which bacteria export molecules into their surroundings. The enrichment of these functional categories suggests that the bioactivity may be tied to the bacterium’s interactions with its environment, including its host, rather than to housekeeping functions inside the cell.

Perhaps the most intriguing bioinformatic finding concerns protein localization. A substantial proportion of the 41 candidates were predicted to reside in the bacterial outer membrane or to carry signal peptides, the short amino acid tags that direct proteins out of the cell. This distribution raises the possibility that some of the candidate proteins travel in or on outer membrane vesicles, small spherical blebs that Gram-negative bacteria such as Bacteroides release into their environment. Outer membrane vesicles have attracted intense interest in microbiome research because they can package proteins, lipids, and genetic material and deliver them to host cells, potentially modulating immune responses and signaling pathways. If the anti-proliferative molecules are indeed shuttled in vesicles, that would offer a plausible route by which a gut bacterium could influence the behavior of epithelial cells lining the colon.

The authors are careful to frame their findings as a candidate list rather than a proof of mechanism, and this caution is scientifically appropriate. Correlation between the presence of a protein in an active fraction and the fraction’s biological effect does not establish causation. Any of the 41 proteins, or some combination of them, could be responsible for the anti-proliferative activity, or the true effector could be a non-protein molecule that co-purified with them. The researchers explicitly state that the candidates should be regarded as associated with the bioactive fractions rather than confirmed functional effectors, and that the individual and combined roles of these proteins require direct experimental validation. Such validation would typically involve purifying each candidate, testing it alone and in combination on cancer cells, and knocking out or silencing the corresponding genes in the bacterium to see whether activity disappears.

The broader significance of the study lies in the growing recognition that gut microbiota-derived factors contribute to colorectal cancer progression and, potentially, to its prevention. Epidemiological and experimental work over the past decade has implicated bacterial metabolites, toxins, and cell surface components in virtually every stage of tumor development, from DNA damage to immune evasion. Yet the field has struggled with a fundamental problem: the gut contains hundreds of bacterial species producing thousands of molecules, and pinpointing which of them matter for a given disease outcome is enormously difficult. Studies like this one demonstrate a tractable path forward. By coupling a reductionist fractionation workflow with unbiased proteomics and rigorous bioinformatics, researchers can convert a vague observation, that a bacterium seems to influence cancer cells, into a concrete, testable molecular hypothesis.

Looking ahead, the framework established by the Zunyi Medical University team could be applied to other bacterial strains and other cancer models, and the 41 candidates themselves provide a rich starting point for mechanistic work. If one or more of these proteins is eventually confirmed to suppress colorectal cancer cell proliferation, it could inspire entirely new classes of therapeutics derived from the microbiome, whether as purified proteins, engineered bacterial strains, or vesicle-based delivery systems. Conversely, even negative results from follow-up validation would sharpen the field’s understanding of how bioactive fractions exert their effects. For now, the study stands as a careful, methodical piece of science: a reminder that the answers to some of the most pressing questions in cancer biology may be produced by organisms already living inside us, waiting to be identified one protein at a time.

Subject of Research: Proteomic identification of secreted Bacteroides fragilis proteins associated with anti-colorectal cancer activity

Article Title: Proteomic identification of secreted protein candidates mediating anti-colorectal cancer activity in a human-derived non-enterotoxigenic Bacteroides fragilis

Article References: Peng, Y., Huang, L., Chen, M., Yang, H., Gan, Y., Wang, N., Wang, M., Yang, Z., & Li, J. (2026). Proteomic identification of secreted protein candidates mediating anti-colorectal cancer activity in a human-derived non-enterotoxigenic Bacteroides fragilis. BMC Cancer. https://doi.org/10.1186/s12885-026-17063-2

Image Credits: AI Generated

DOI: 10.1186/s12885-026-17063-2

Keywords: Bacteroides fragilis, colorectal cancer, proteomics, microbiome, activity-guided fractionation, LC-MS/MS, outer membrane vesicles, HT-29 cells, secreted proteins, ABC transporters, iron acquisition, gut microbiota