The research team, led by Tong Wu and Jiashuo Huang of Beijing Technology and Business University, set out to answer a deceptively simple question: what actually happens inside a cup when coffee meets milk? To find out, they prepared four different coffee-milk systems, combining coffee with ultra-high-temperature treated and pasteurized milk, in both whole and skim varieties. This design allowed them to isolate the effects of two variables that matter enormously to the dairy industry and to consumers: how intensely the milk has been heat-treated, and how much fat it contains.

The first surprise came from just how thoroughly coffee transforms milk’s internal architecture. When milk was added to coffee, the mixture’s pH rose, its color became lighter, and its viscosity increased. The particles suspended in the liquid grew larger, and spectroscopic analysis using Fourier-transform infrared spectroscopy revealed distinct shifts in the chemical signatures of the mixture. Under the electron microscope, the researchers could see the evidence directly: aggregates had formed, clusters of milk proteins bound together with coffee components into structures that neither ingredient possessed on its own.

To understand why these aggregates form at all, the team turned to molecular docking, a computational technique more commonly associated with drug discovery than with breakfast beverages. The simulations revealed that coffee components latch onto milk proteins primarily through hydrogen bonding and pi-pi stacking interactions, the latter being a form of molecular attraction between flat, ring-shaped chemical structures. These weak but numerous bonds act like microscopic Velcro, stabilizing the complexes that form when the two liquids merge. It is a molecular handshake, repeated billions of times throughout the cup, that gives the drink its new physical identity.

Here is where the physics becomes genuinely counterintuitive. One might assume that larger, more numerous particles would make a drink feel smoother, the way cream enriches a soup. The opposite occurred. The structural reorganization triggered by coffee actually enhanced interactions between particles in the liquid, and this crowded, interactive environment reduced the lubrication at the interface between the beverage and the surfaces of the mouth. The measurable consequence was an increase in the friction coefficient, a tribological parameter that quantifies how much resistance arises when the liquid slides across a surface, in this case a simulated tongue.

Tribology, the study of friction and wear, has quietly become one of the most powerful tools in modern food science. Traditional rheology measures how a liquid flows in bulk, but it cannot capture what happens in the thin films that exist between the tongue and the roof of the mouth during swallowing. Those thin-film dynamics are where mouthfeel lives. A beverage can have identical viscosity readings yet feel dramatically different in the mouth if its particles interact differently with oral surfaces. By measuring friction coefficients across their four coffee-milk systems, the researchers could connect the invisible molecular architecture of the drink directly to the felt experience of drinking it.

The role of fat emerged as the decisive counterweight. In the systems made with whole milk, fat globules physically disrupted the aggregate structures that coffee components and milk proteins had built together. The fat essentially wedged itself into the protein-coffee networks, breaking them apart and restoring lubrication to the oral interface. The result was lower friction and, presumably, the smoother, rounder mouthfeel that whole-milk coffee drinkers know and love. To confirm that fat was genuinely the causal agent rather than a bystander, the team performed fat add-back experiments, reintroducing fat into the systems and watching the friction coefficients fall in response. The evidence was direct: fat is the friction-reducer, the natural lubricant of the coffee-milk world.

This finding carries real weight for an industry under pressure. Reduced-fat dairy products are in growing demand as consumers seek healthier options, but skim-milk coffee has never fully replicated the sensory appeal of its full-fat counterpart, and this study now explains why in physical terms. Remove the fat, and the protein-coffee aggregates remain intact, interparticle interactions intensify, lubrication drops, and friction rises, producing the thinner, sometimes astringent sensation that makes skim lattes disappointing. The researchers frame their work as a step toward optimizing the mouthfeel and sensory quality of reduced-fat coffee-milk beverages, suggesting that if manufacturers can find other ways to disrupt aggregates or restore interfacial lubrication, they might engineer low-fat drinks that feel indulgent without the fat.

The study also highlights how much sensory science has matured as a discipline. Rather than relying solely on taste panels to describe differences that panelists may struggle to articulate, researchers can now trace those differences down through multiple levels of organization: the molecular bonds between coffee polyphenols and milk proteins, the microscopic aggregates visible under electron microscopy, the bulk properties like viscosity and particle size, and finally the tribological behavior at the oral interface. Each level feeds into the next, and the sensory experience that reaches the consumer is the integrated output of this entire cascade. It is a reminder that even the most everyday pleasures rest on remarkably intricate physical foundations.

There is something almost poetic in the conclusion. Every morning, billions of people perform an unwitting experiment in soft-matter physics, pouring milk into coffee and setting off a cascade of hydrogen bonds, pi-pi stacking interactions, and aggregate formation that will register on their tongues as nothing more than a pleasant smoothness. The Beijing team’s work transforms that daily ritual into a map, showing exactly which molecular events produce which physical consequences and which sensory outcomes. For the coffee industry, it offers a blueprint for building better low-fat beverages. For everyone else, it offers a new appreciation of the hidden engineering in the cup, and of the humble fat globule, quietly dismantling protein networks one interface at a time so that your morning coffee glides instead of drags.

Subject of Research: Structural reorganization of milk matrices in coffee-milk systems and its effect on oral friction and sensory perception

Article Title: Milk matrix-driven structural reorganization modulates oral friction and sensory perception in coffee-milk systems

Article References: Wu, T., Huang, J., Hu, Y., Guo, K., Zeng, H., & Wang, Y. (2026). Milk matrix-driven structural reorganization modulates oral friction and sensory perception in coffee-milk systems. npj Science of Food. https://doi.org/10.1038/s41538-026-01173-z

Image Credits: AI Generated

DOI: 10.1038/s41538-026-01173-z

Keywords: coffee, milk, mouthfeel, tribology, food science, milk fat, molecular docking, oral friction, sensory perception, reduced-fat beverages, protein aggregates, npj Science of Food

The research team, led by scientists at the Ocean University of China working with collaborators at the BYHEALTH Institute of Nutrition and Health and other Chinese institutions, set out to investigate whether chronic, low-level antibiotic exposure could trigger adaptive responses in this non-spore-forming bacterium. Unlike spore-forming species such as Bacillus, which can build tough, desiccated spores to ride out hostile conditions, Lactobacillus species have long been considered relatively fragile outside their comfortable niches. Bacterial dormancy offers an alternative strategy: a metabolically suppressed but fully reversible state that allows non-spore-forming bacteria to withstand adverse conditions without constructing specialized survival structures. The researchers exposed the sp1.1 strain of L. delbrueckii subsp. bulgaricus to concentrations of rifampicin that stressed but did not kill the cells, then tracked what happened to their physiology, their stress resistance, and their behavior in living animals.

The results were striking. Rifampicin-treated cells entered a state the authors characterize as dormancy, marked by sharply reduced ATP levels, the cellular energy currency, and impaired cell division. The bacteria essentially throttled themselves down, halting growth and conserving resources. Crucially, the state was reversible: when the antibiotic stress was removed, colonies recovered and normal growth resumed. This reversibility distinguishes dormancy from death and from irreversible injury, and it suggests a programmed, controlled response rather than simple damage. In the food industry, where starter cultures routinely face freezing, drying, acidification, and thermal processing steps that whittle down viable cell counts, a reversible dormant state that confers cross-tolerance to multiple stressors at once would be an enormously valuable property.

And cross-tolerance is exactly what the team observed. Dormant cells tolerated acid, alkali, and heat stress significantly better than untreated controls, meaning a single trigger, sublethal rifampicin exposure, produced broad-spectrum resistance rather than protection against just one threat. The functional consequences extended beyond the laboratory bench. When the researchers administered the treated bacteria to mice, the dormant cells persisted longer in the intestinal tract than normal cells, a property known as prolonged intestinal retention. They also exerted a greater impact on the structure of the gut microbial community, indicating that the dormant state changes not just survival but ecological behavior once the bacteria reach the gut. For probiotic formulations, where the central challenge is delivering enough live cells through gastric acid and bile to produce a benefit, these findings point to a potentially powerful new lever.

The mechanism behind this transformation turned out to be one of the most intriguing aspects of the study. When the researchers disrupted the protein aggregates that formed in the dormant cells, either by adjusting pH or by treating the cells with 1,6-hexanediol, a chemical that dissolves certain types of protein condensates, the enhanced tolerance vanished entirely. That causal experiment established that the aggregates are not a byproduct of dormancy but its functional engine. Rifampicin, which targets RNA polymerase and inhibits transcription, triggered a metabolic shift toward transcriptional and translational inhibition, and as protein synthesis slowed, proteins began to clump together into aggregates within the cells. What could have been cellular garbage, however, turned out to be something far more purposeful.

Proteomic analysis of the aggregates revealed a striking selectivity. Rather than capturing random cellular proteins, the aggregates were enriched for proteins involved in translation, RNA metabolism, and DNA repair, precisely the functions a cell would need to restart growth once conditions improve. The authors propose a model in which these aggregates function as molecular safe houses, sequestering and protecting key proteins from degradation or damage during the dormant period. By concentrating translation and DNA repair machinery in one protected location, the aggregates may locally enhance the efficiency of those processes when the cell revives, while also serving as a protein reservoir that allows rapid recovery of growth. In this view, the aggregate is less like a junk pile and more like a sealed emergency kit, packed with the tools needed for reconstruction.

Two intrinsically disordered proteins, identified in the study as Gene1622 and Gene1909, emerged as likely architects of this process. Disordered proteins lack a fixed three-dimensional structure and are known drivers of biomolecular condensate formation in many organisms. In the rifampicin-treated bacteria, both proteins were upregulated in whole-cell measurements and were enriched within the aggregates, suggesting they help nucleate or organize the condensates. The involvement of disordered proteins links this bacterial phenomenon to a broader and rapidly growing body of research on biomolecular condensates, membrane-less compartments that cells across all domains of life use to organize their biochemistry. That a food bacterium uses condensate biology to survive antibiotic stress adds a new ecological dimension to what has largely been studied in model organisms and human cells.

The study also uncovered hidden diversity within the bacterial population. Using single-cell RNA sequencing, the researchers identified a subpopulation of cells that sustained expression of alaS and gatA, genes involved in translation, even while global protein synthesis was being suppressed. These genes were downregulated in bulk RNA sequencing, which averages signals across the whole population, but single-cell analysis revealed that a distinct cluster of cells kept them active. This population heterogeneity means that not all cells respond identically to antibiotic stress; some appear to hedge their bets, maintaining critical functions while their neighbors go fully dormant. Such bet-hedging strategies are well known in microbial ecology, and this study demonstrates them at single-cell resolution in a commercially important probiotic, showing how a population can prepare for recovery even while most of its members are shut down.

The implications cut in two directions. On the constructive side, the findings provide a theoretical basis for engineering more robust starter cultures and probiotics. If manufacturers can deliberately induce and control this reversible dormant state, perhaps through stress conditioning rather than antibiotic exposure, they could produce cultures that survive processing better, retain viability longer on the shelf, and persist more effectively in the gut. The authors note that the work was supported by the National Natural Science Foundation of China and the BYHEALTH Nutrition and Health Research Foundation, reflecting industry interest in exactly these applications. On the cautionary side, the study raises ecological questions about stress-induced adaptation in the food chain. Antibiotic residues at sublethal levels are a known feature of some food production environments, and this research shows that such exposure can make food bacteria hardier and more persistent in the gut, where they exert stronger effects on microbial community structure. Whether that enhanced persistence is beneficial or disruptive to gut microbial homeostasis remains an open question, and one the authors flag explicitly. As the boundaries between food microbiology, antibiotic stewardship, and gut ecology continue to blur, this study suggests that the microbes in our yogurt may be far more adaptable, and far more responsive to their chemical environment, than anyone assumed.

Subject of Research: Sublethal antibiotic-induced dormancy and protein aggregation in Lactobacillus delbrueckii subsp. bulgaricus

Article Title: Sublethal rifampicin enhances the tolerance of Lactobacillus delbrueckii subsp. bulgaricus to food processing and intestinal retention through protein aggregation

Article References: Liu, L., Li, Z., Di, C., Ma, C., Huang, Y., Hao, X., Fu, Z., Yi, H., Zhang, Z., Li, P., Li, L., Ze, X., He, R., Zhang, L., & Gong, P. (2026). Sublethal rifampicin enhances the tolerance of Lactobacillus delbrueckii subsp. bulgaricus to food processing and intestinal retention through protein aggregation. Microbiome. https://doi.org/10.1186/s40168-026-02517-3

Image Credits: AI Generated

DOI: 10.1186/s40168-026-02517-3

Keywords: Lactobacillus delbrueckii subsp. bulgaricus, rifampicin, bacterial dormancy, protein aggregates, probiotics, starter cultures, gut microbiome, stress tolerance, single-cell RNA sequencing, intrinsically disordered proteins, food fermentation, intestinal retention

The logic of the discovery begins with Huntington’s disease itself. The disorder is caused by an abnormal expansion of CAG trinucleotide repeats in the Huntingtin gene, which encodes an elongated poly-glutamine tract at the beginning of the protein. This expanded stretch drives misfolding, and the misfolded protein forms aggregates that disrupt transcription, intracellular transport, metabolism and, crucially, the ubiquitin-proteasome system, the cellular machinery responsible for shredding damaged or unwanted proteins. In patients, the chymotrypsin-like and caspase-like activities of the proteasome are measurably reduced in the striatum and cortex, the very brain regions that atrophy as the disease progresses. Eliminating the aggregates, therefore, has long been considered one of the most promising directions for therapy, but the question of how to do it safely has remained open.

MicroRNAs offered a tantalizing lead. These short non-coding RNAs regulate gene expression by binding to the 3′ untranslated regions of target messenger RNAs, blocking translation and thereby suppressing protein production. In the central nervous system, they tune everything from neurotransmitter receptors to ion channels, and several have documented neuroprotective effects in stroke, Alzheimer’s and Parkinson’s models. miR-196a had already earned attention in Huntington’s research: previous work from the same group showed that it improves neuronal morphology through the RANBP10 and IGF2 pathways, exerts antioxidant effects via the USP15/NRF2 axis, and, most strikingly, reduces mutant huntingtin aggregates and rescues pathological phenotypes in cell, mouse and induced pluripotent stem cell models. Yet the huntingtin gene contains no miR-196a binding sites, leaving the mechanism behind the aggregate reduction a genuine mystery.

To solve it, the team turned to unbiased proteomics. They transfected mouse neuroblastoma N2a cells with a fluorescent mutant huntingtin fragment containing 84 glutamine repeats, together with either miR-196a or a irrelevant control microRNA, and then analyzed the entire proteome by liquid chromatography-tandem mass spectrometry. Of 2,681 proteins identified, 147 rose and 144 fell significantly in the miR-196a-treated cells. Applying progressively stricter fold-change cutoffs and intersecting the results with five independent microRNA target-prediction databases, the researchers narrowed the field to six candidates. Western blot validation confirmed that two of them, Rad23b and Snap91, were significantly downregulated by miR-196a, and Rad23b, with its known ties to proteasome function, became the focus.

Confirming direct regulation came next. The researchers cloned the 3′ untranslated region of Rad23b downstream of a luciferase reporter and showed that miR-196a markedly suppressed luciferase activity, an effect abolished when the predicted binding site was mutated. The relationship held in living animals as well: cerebral cortex from miR-196a transgenic mice showed a moderate but significant reduction in Rad23b protein. Meanwhile, analysis of a public RNA sequencing dataset from directly reprogrammed striatal neurons revealed that Rad23b mRNA was significantly elevated in neurons derived from symptomatic Huntington’s patients compared with presymptomatic carriers, hinting that Rad23b tracks with disease progression rather than merely correlating with the mutation itself.

The functional experiments that followed delivered the study’s central surprise. When the researchers overexpressed Rad23b alongside mutant huntingtin in N2a cells, aggregates flourished, both the insoluble clumps and the soluble protein rose on Western blots, and cell death, measured by propidium iodide staining, climbed significantly. Conversely, knocking Rad23b down with short hairpin RNA or deleting it entirely using CRISPR-Cas9 shrank the aggregates dramatically. The effect was not an artifact of the fluorescent tag: untagged mutant huntingtin behaved the same way. When Rad23b was reintroduced into cells already treated with miR-196a, the microRNA’s suppressive effect on aggregates was reversed, establishing Rad23b as a key mediator of miR-196a’s neuroprotection.

Structural dissection explained how Rad23b physically engages the toxic protein. Fluorescently tagged Rad23b formed puncta that colocalized with huntingtin aggregates in cultured cells, and endogenous Rad23b did the same in the cortex and striatum of Huntington’s transgenic mice, while remaining evenly distributed in healthy animals. Co-immunoprecipitation confirmed a direct physical interaction in both directions. Mutating any of the three lysine residues at positions 6, 9 and 15 in the N-terminus of huntingtin, the likely sites of ubiquitination, reduced the Rad23b-driven aggregation, and deleting both of Rad23b’s ubiquitin-associated domains abolished its effect, pinpointing the ubiquitin-binding machinery as the critical interface.

The mechanism, however, turned out to be darker than a simple shuttle-to-destruction story. Rad23b is classically described as an ubiquitin shuttle protein, using its ubiquitin-like domain to dock at the proteasome and its ubiquitin-associated domains to ferry ubiquitinated cargo for degradation. Consistent with that role, Rad23b overexpression increased the ubiquitination of mutant huntingtin. Yet the ubiquitinated protein was not destroyed. A reporter assay using a destabilized GFP substrate revealed that Rad23b-deficient cells degraded proteasomal cargo faster than wild-type cells, and direct enzymatic measurements showed that deleting Rad23b significantly boosted the chymotrypsin-like activity of the proteasome, the very activity that is depressed in the brains of Huntington’s patients. Inhibiting the proteasome with MG132 restored aggregates in Rad23b-knockout cells, whereas blocking autophagy with chloroquine had no effect, confirming that the pathway in play is the ubiquitin-proteasome system, not autophagy. In short, Rad23b tags huntingtin for destruction but simultaneously gums up the destruction machinery, a molecular sabotage that tips the balance toward accumulation.

The in vivo evidence sealed the case. The team generated Rad23b transgenic mice by lentiviral transgenesis and crossed them with R6/2 Huntington’s mice, which carry an N-terminal human huntingtin fragment with 120 to 150 CAG repeats. The resulting double-transgenic animals fared measurably worse than their Huntington’s littermates: at eight and nine weeks of age they fell off an accelerating rotarod sooner and at lower speeds, and they showed elevated hindlimb clasping scores, a classic sign of worsening motor dysfunction. Biochemical analysis of the cortex at six weeks revealed significantly more mutant huntingtin aggregates in the double transgenics, and immunofluorescent staining of the retrosplenial cortex showed Rad23b-labeled aggregates accompanied by significantly more TUNEL-positive dying cells. Neither Rad23b nor miR-196a transgene expression alone produced any Huntington-like pathology, underscoring that Rad23b acts by amplifying the mutant protein’s toxicity rather than causing disease on its own.

The implications reach beyond Huntington’s disease. Rad23b has been found sequestered in neuronal inclusions across a striking range of disorders, including spinocerebellar ataxias, frontotemporal dementia, amyotrophic lateral sclerosis and Parkinson’s disease, yet its role appears context-dependent: in C9ORF72-linked disease, restoring Rad23b actually reduces poly-glycine-alanine aggregates. The authors suggest their mechanism fits poly-glutamine aggregates specifically, and they caution that bulk cortical lysates may dilute neuron-specific proteasomal changes, that behavioral testing here focused on motor readouts, and that miR-196a likely acts through multiple pathways, including metabolic ones hinted at by their proteomic enrichment analyses. Even so, the conclusion is compelling: downregulating Rad23b, or disrupting its interaction with mutant huntingtin, offers a concrete, mechanistically grounded strategy to clear toxic aggregates and slow disease progression, potentially applicable across the family of poly-glutamine disorders. For a disease that has resisted every therapeutic attempt for decades, a small RNA with a well-defined target is exactly the kind of lead the field has been waiting for.

Subject of Research: miR-196a-mediated suppression of Rad23b to reduce mutant huntingtin aggregates through the ubiquitin-proteasome system in Huntington's disease

Article Title: miR-196a reduces mutant Huntingtin aggregates by Rad23b-mediated degradation in Huntington’s disease

Article References: Tung, C.-W., Chan, S. C., Chen, Y.-C., Wu, P.-M., Cheng, P.-H., Chen, C.-M., & Yang, S.-H. (2026). miR-196a reduces mutant Huntingtin aggregates by Rad23b-mediated degradation in Huntington’s disease. Journal of Biomedical Science, 33(1), Article 90. https://doi.org/10.1186/s12929-026-01292-5

Image Credits: AI Generated

DOI: 10.1186/s12929-026-01292-5

Keywords: Huntington's disease, miR-196a, Rad23b, mutant huntingtin, protein aggregates, ubiquitin-proteasome system, microRNA, neurodegeneration, proteomics, R6/2 transgenic mice, chymotrypsin-like activity, gene regulation