Harnessing the power of fermented foods as bioactive delivery systems
Cheese is usually appreciated as a finished product; a traditional functional food with a flavour and texture profile shaped by fermentation. Although cheese consumption has been associated with health benefits, the mechanisms through which this is achieved are still unclear. Our overarching hypothesis is that the cheese matrix functions as a structured delivery vehicle for live microorganisms and their metabolites to the gut, driving compositional shifts in the resident microbiota and localised production of short-chain fatty acids, particularly propionate, which can modulate hepatic lipid metabolism and confer blood lipid health benefits (1, 2).
Working with a local artisan cheese producer (Nettlebed Creamery in Oxfordshire), our team at the University of Reading (led by PhD student and cheesemaker, Sabrina Longley) applied a molecular phenotyping approach in order to characterise different cheeses; not just as a fermented food, but as a structurally and functionally organised dynamic bioactive delivery matrix, and to ask what a rigorous multi-omics snapshot could tell us about what actually survives the cheesemaking process to the point of consumption (3). Pairing 16S amplicon sequencing with ¹H-NMR spectroscopy, we analysed three distinct cheese styles produced at the Creamery across the full maturation arc of British artisan cheese: Bix (a soft, bloomy-rind), Highmoor (a semi-soft, washed-rind), and Witheridge (a semi-hard, hay-aged cheese).
Uncovering a distinct ecological succession
Each cheese style produced a unique ecological trajectory. Bix, aged just 8–9 days, stayed dominated by Lactococcus lactis with minimal diversification during maturation. Highmoor showed classic smear-ripened dynamics: an early starter lactic acid bacteria triad (L. lactis, S. thermophilus, L. delbrueckii subsp. bulgaricus) gave way to rind flora (Brevibacterium, Corynebacterium, Staphylococcus equorum) at mid-maturation, before diversity contracted again by maturity, providing evidence that late-stage competitive exclusion, and not continued growth, shapes the final microbial community in this cheese style. The result we found most striking (and exciting) was in hay-aged Witheridge. Here, we observed a 3.8-fold increase in species over time, as well as in comparison to the other cheeses, and the highest Shannon (5.55) and Simpson (0.90) diversity indices of any sample we measured; plausibly reflecting hay-derived proteins and fibres (potential prebiotics) supporting ecological niche expansion during the six months of anaerobic maturation.
An important point we were careful to flag in our paper was that at the point of consumption, it was S. thermophilus not L. lactis that dominated both aged cheeses; the relative abundance of L. lactis reduced to under 10% in Highmoor, and under 5% in Witheridge. Given that dietary L. lactis: A) survives gastric transit but is largely killed in the duodenum, with rapid lysis of non-viable cells (4), B) is cleared from human faeces within three days of ingestion (5), and C) behaves as a transient rather than colonising member of the gut community more broadly (6), we caution against claims that lactococcal abundance in aged cheese affects the gut microbiota.
The dynamic biochemical fingerprint
¹H-NMR spectroscopy resolved a clean metabolic signature over time: lactose was undetectable in all three cheeses by maturity, confirming complete conversion via LAB glycolysis and rendering these products functionally low-FODMAP live-microbe vehicles. More relevant to our hypothesis, propionate (the short chain fatty acid produced as a result of Propionibacterium freudenreichii metabolism) was absent in Bix, present at 0.032 mM in Highmoor, and highest in mature Witheridge (0.108 mM). Controlling for maturation, the raw, hay free Witheridge produced zero propionate, confirming that propionibacterial activity took place during extended maturation, rather than as a result of raw-milk carryover.
It is important to highlight that despite these interesting results, the dose of propionate delivered to the colon in a 30 g cheese portion is lower than that of other delivery vehicles (e.g. an inulin-propionate ester, which delivers roughly 2.36 g propionate to the colon per 10 g dose to provide metabolic benefits (7). So the mechanistic case we were evaluating, was not on propionate content per se, but on whether cheese delivers live propionibacteria capable of ongoing in situ production, and/or triggers resident microbiota remodelling.
Role of the cheese matrix in microbial activity
The exploration of cheese as a bioactive delivery system that motivated this work has independent mechanistic support in the wider literature. In vitro digestion models show the cheese matrix (versus liquid dairy) significantly protects P. freudenreichii’s immunomodulatory surface protein SlpB from proteolysis (8), while growth within Emmental cheese juice upregulates chaperonins and confers acid/bile-salt cross-tolerance beyond simple physical buffering (9). Separately, non-starter lactic acid bacteria in Cheddar cheese survive simulated gastric-duodenal passage at up to 10⁷ CFU/g, with select strains showing 64–79% epithelial adhesion (10). Taken together, matrix protection looks real but taxon-specific; strong for propionibacteria and non-starter lactic acid bacteria, but weak for lactococci, which is an important caveat to carry forward in future research.
One limitation of our own dataset follows directly from this. 16S amplicon sequencing is DNA-based and cannot distinguish live cells from dead ones, so while we can describe which taxa are present and abundant at the point of consumption, we have not yet established viable counts in a portion as eaten. Quantifying delivered CFU is a priority for the next phase of the work, and is one of the questions our current human study is designed to help answer.
From bench to bedside: an RCT assessing the impact on cholesterol levels
The findings from our paper have now led to a 12-week human study (NCT07351175) that is currently ongoing, where we are comparing 30 g/day of the hay-aged cheese against a cheddar cheese control, in hypercholesterolaemic adults (with total cholesterol as the primary endpoint). The study design was centred around addressing an important unresolved gap: cheese reliably lowers LDL-C versus butter (11). However, the leading candidate mechanism – calcium-driven faecal fat excretion, fails on direct testing (12). By pitting a propionate-rich, high-diversity hay-aged cheese against an equally aged, lactose-free comparator, we hope to isolate the microbial/metabolite variable(s) for the first time in this space. If the hypothesis holds, we expect the causal mechanism to run through gut microbial remodelling rather than direct SCFA delivery. We will be glad to report back to this community with the results once the trial is complete!
References:
(1) den Besten et al., 2013, J. Lipid Res. 54:2325
(2) Hosseini et al., 2011, Nutr. Rev. 69:245
(3) Longley, Gibson & Wijeyesekera, 2026, ACS Food Sci. Technol., DOI: 10.1021/acsfoodscitech.5c01243
(4) Drouault et al., 1999, Appl. Environ. Microbiol. 65:4881
(5) Klijn et al., 1995, Appl. Environ. Microbiol. 61:2771
(6) Derrien & van Hylckama Vlieg, 2015, Trends Microbiol. 23:354
(7) Chambers et al., 2015, Gut 64:1744
(8) Rabah et al., 2018, Food Res. Int. 106:712)
(9) Gagnaire et al., 2015, PLoS ONE 10(8):e0135780
(10) Leeuwendaal et al., 2021, J. Funct. Foods 83:104425
(11) de Goede et al., 2015, Nutr. Rev. 73:259
(12) Feeney et al., 2023, Eur. J. Nutr. 62:1755
