Browse Articles

90-day oral toxicity study of a salmon nasal cartilage extract containing undenatured collagen and proteoglycan in Sprague-Dawley rats

Salmon nasal cartilage is a rich source of proteoglycan and collagen that is widely used in food products. In Japan, a novel extraction method has been developed and patented that enable the simultaneous production of proteoglycan and undenatured collagen from salmon nasal cartilage. This study evaluated the subchronic oral toxicity of this extract mixture (SCP Complex-LS, containing 40% proteoglycan and 40% undenatured collagen) in a 90-day repeated toxicity study in Sprague-Dawley rats, following by a 14-day recovery period. Rats (20/group; 10 males and 10 females) were administered SCP Complex-LS once daily by oral gavage at doses of 0 (vehicle), 10.3, 20.6, or 41.2 mg/kg body weight/day, with additional recovery groups (10/groups; 5 males and 5 females) receiving vehicle or the high dose. Clinical endpoints included mortality, clinical observations, body weight, food consumption, estrous cyclicity, ophthalmoscopy, clinical pathology, organ weights, gross pathology, and histopathology. No deaths or test item-related clinical signs were observed. Sporadic changes in hematological, biochemical, urinary, or organ weight parameters occurred in some dose groups but were small in magnitude, showed no consistent dose-response relationship, and were not corroborated by histopathological alterations. Histopathology revealed only minimal findings, such as mild inflammation or congestion in the liver, kidneys, and lungs, occurring at low incidence and with similar frequency in both control and high-dose groups. Estrous cycles remained within normal limits, and recovery groups showed no evidence of delayed or irreversible toxicity. Based on these findings, the No Observed Adverse Effect Level (NOAEL) for SCP Complex-LS was determined to be 41.2 mg/kg body weight/day in Sprague-Dawley rats under the conditions of this study, supporting its safety for use as a food ingredient within the expected range of human intake.

Categorization of household drinking water and sanitation service levels and associated determinants in Uganda

Abstract Achieving universal access to safe drinking water and sanitation remains a major challenge in Uganda, where persistent socioeconomic and spatial inequalities limit progress toward Sustainable Development Goal 6. While previous studies have largely focused on localized settings, national-level evidence applying the WHO/UNICEF Joint Monitoring Programme (JMP) service ladder framework is limited. This study aimed to address this gap. This study used secondary data from the 2018–2019 Uganda Malaria Indicator Survey. A total of 8,925 households were included, with data drawn from the Household Recode file. The dependent variables were household sources of drinking water and types of toilet facilities and these were classified according to the WHO/UNICEF JMP service ladder framework into safely managed, basic, limited, unimproved, and no service categories. Independent variables comprised a range of individual and household-level characteristics. Descriptive statistics were used to summarize key variables, while chi-square tests examined bivariate associations between the dependent and independent variables. Multivariable ordinal logistic regression was then applied to estimate the adjusted effects of the determinants on water and sanitation service levels. All analyses accounted for the complex survey design and sampling weights and were conducted using Stata version 18. Most Ugandan households relied on basic drinking water services (67.8%), while safely managed water remained low (8.8%) and was almost absent in regions such as Karamoja. Sanitation access was dominated by limited services (49.7%), with only 43.6% achieving safely managed sanitation. The significant determinants associated with JMP service levels were identified. Higher education, household wealth, male headship, and residence in central and western regions strongly increased the likelihood of accessing higher service levels. Rural areas and refugee settlements consistently exhibited the lowest service levels, reflecting structural deficits in infrastructure, affordability, and service reliability. This study shows that safely managed drinking water and sanitation remain limited in Uganda, with persistent inequalities driven by education, wealth, gender, and region. Targeted investments and pro-poor, equity-focused policies are urgently needed to improve service levels in underserved rural areas and refugee settlements. Strengthening infrastructure, reducing affordability barriers, and supporting context-specific water and sanitation programming will be essential for accelerating progress toward universal and safely managed services.

In silico evaluation of bioactive compounds from the sea urchin Temnopleurus toreumaticus: Potential multifunctional agents against mosquito vectors and tropical pathogens

This study investigates the ethanolic extract of the sea urchin Temnopleurus toreumaticus (test and spine) for its larvicidal efficacy against Aedes aegypti , cytotoxicity, and in silico interactions supporting potential antimalarial and antiviral activities. Zoochemical profiling by Fourier-transform infrared spectroscopy (FT-IR) showed the presence of functional groups, while gas chromatography-mass spectrometry (GC-MS) analysis identified 24 bioactive compounds with insecticidal and enzyme-inhibitory properties. The extract exhibited strong larvicidal (LC₅₀ = 164.18 µg/mL) and pupicidal (LC₅₀ = 209.91 µg/mL) effects on A. aegypti in a concentration‑dependent manner, supported by significant acetylcholinesterase inhibition (IC₅₀ = 151.49 µg/mL). Morphological examinations showed epithelial disorganization and structural damage in treated larvae and pupae. In silico docking confirmed that the identified zoochemicals exhibited binding affinities ranging from –7 to –8 kcal/mol for juvenile hormone binding protein (PDB 5V13), –5 to –6 kcal/mol for acetylcholinesterase (PDB 1DX4), –4 to –5 kcal/mol for Plasmodium falciparum lactate dehydrogenase (PDB 1CEQ), and –4 to –5 kcal/mol for chikungunya nsP2 protease (PDB 3TRK), indicating their multitarget larvicidal, antimalarial, and antiviral potential. Yeast cell-based assays indicated cytotoxic activity (EC₅₀ = 159.27 µg/mL) with a strong dose-response relationship. Overall, the ethanolic extract of T. toreumaticus is a promising lead for next‑generation, environmentally safe biocontrol agents targeting vector‑borne diseases such as malaria and chikungunya, while clearly emphasizing that the antiviral and antimalarial properties are currently supported only by in silico evidence and require further in vitro and in vivo validation.

Assessing impacts of rainfall intensity and slope gradient on runoff process and dissolved organic carbon loss via surface flow and interflow under simulated rainfall

Dissolved organic carbon (DOC) is an indispensable component of the global carbon cycle. Previous research on runoff process and DOC loss mainly focused on surface flow, with few reports of the hydrological pathway of interflow or DOC loss via interflow. To address this deficiency, a series of rainfall simulations were conducted with three rainfall intensities of 60, 90, and 120 mm h -1 (R60, R90, and R120) and three slope gradients of 5, 15, and 25° (S5, S15, and S25) of purple soil. The initial time of surface flow was faster under high rainfall intensity and steep slope, and the initial time of interflow increased with increased rainfall intensity under gentle slope. In general, the surface flow rates increased first, and reached a steady state within 10–35 min. The interflow curves were single-peak curves for R60-S5 and R90-S5, but exhibited a continued rising trend for other treatments. The interflow volume occupied 69.2% of the total runoff volume under R60-S5, and the percentages of interflow decreased as the rainfall intensity and slope increased. These results indicated that interflow was an important hydrological pathway of purple soil. The DOC concentration of the surface flow decreased with rainfall duration, with opposite trend for DOC concentration of interflow. The DOC concentrations in the interflow were 1.35–2.34 times higher than those in the surface flow. However, the rainfall intensity and slope had little effect on DOC concentrations in both surface flow and interflow. Furthermore, the DOC loss fluxes via surface flow and interflow were 3.77–26.94 g and 0.41–13.73 g, respectively, and the ratios of interflow DOC loss fluxes to the total DOC loss fluxes gradually decreased with the increase of rainfall intensity and slope. Under R60, DOC loss via interflow was the major DOC loss pathway, accounting for 51.0%−78.4% of the total DOC loss, whereas for R90 and R120, DOC loss via surface accounted for >90%. Moreover, runoff volume was positively linearly correlated with the corresponding DOC loss fluxes in both the surface flow (R 2  = 0.93, P  < 0.01) and interflow (R 2  = 0.99, P  < 0.01). These results provide a scientific basis to estimate the fluxes of DOC loss and control carbon loss in the purple soil area of China.

DualSightNet: A novel dual architecture for visual quality control of railway infrastructure

To ensure the operational safety of trains, it is essential to monitor the condition of the rails. In order to detect problems ranging from wear and tear to possible sabotage, no comprehensive and continuous monitoring is carried out today. The objective of this study is to develop and robustly validate a deep-learning model capable of reliably classifying a broad range of safety-critical rail surface defects. The use of deep learning to classify defect types based on optical images is a promising approach, but the existing literature does not yet achieve a robust, high-performing classification for a broad range of failure types. Many approaches rely mainly on local feature extraction, but this carries the risk of overlooking global relationships, which are crucial for distinguishing certain defect types. This study addresses this gap by utilizing the specific features of this problem domain (small, local, and global defect types). To this end, DualSightNet is introduced as a hybrid architecture enhanced by an attention module for classifying a broad range of railway track surface defects. The model achieves a five-fold cross-validated average balanced accuracy of 97.55 % on a peer-reviewed, real-world dataset of 5,153 images covering seven defect types, recorded directly from an inspection vehicle under operational conditions, indicating strong generalization across the diverse real-world variations represented in the dataset. Compared to existing CNN- or Transformer-based approaches, DualSightNet is the first approach that combines local and global feature extraction through a gating-based fusion mechanism and then enhances the fused representation using an attention module, which enables substantially more robust multiclass defect recognition. This sets a new benchmark for our problem domain, surpassing previous approaches, which either lack broad defect coverage or do not employ rigorous cross-validated evaluation. Our results have far-reaching practical implications, proving that by leveraging problem-specific features, neural networks are able to robustly classify a broad range of defect types. The inference time of the proposed system (3.00 ms per image) makes DualSightNet suitable for deployment in automated inspection vehicles and real-time monitoring scenarios.

Refining estimation techniques for the Two-Sided Power Distribution: A data-sensitive perspective

This study proposes a novel method aimed at achieving more reliable parameter estimates for the Two-Sided Power Distribution (TSPD), particularly under small-sample conditions. The proposed approach enhances the flexibility and data sensitivity of the distribution by redefining it as a convex combination of two independent uniform components. A new estimation formula is introduced for the probability P r { Y < X } , which holds critical importance in fields such as system reliability and stress–strength modeling. Compared to classical theoretical expressions, this new formulation produces more accurate and stable results in small-sample settings. Simulation studies and real-data applications demonstrate that the proposed method reduces estimation error and yields values closer to empirical observations. Furthermore, the method provides a balanced and computationally feasible alternative to conventional techniques such as maximum likelihood estimation. The results reveal that the proposed approach offers a significant advantage for the reliable computation of key performance metrics such as reliability probabilities and related measures.

I Told You So! Verbal cue beliefs are associated with truth detection, but not lie detection

This study aimed to 1) replicate and extend previous findings on the relationship between accurate beliefs about the diagnosticity of verbal deception cues and actual truth/lie discrimination as reported in Bogaard and Meijer (2017), and 2) examine the role of presentation modality. Participants (N = 246) listed their beliefs about deception cues and then judged the credibility of truthful and deceptive autobiographical statements presented in one of three modalities: audio-only, transcript, or audio-visual. As in the original study, participants provided continuous credibility ratings. In addition, we asked participants to make binary truth/lie judgments, allowing us to assess discrimination accuracy. We replicated the original finding, namely a small but significant positive correlation between accurate verbal cue beliefs and the credibility ratings of truthful, but not deceptive, statements. A small association was also observed between accurate cue beliefs and truth discrimination accuracy. Contrary to our expectations, modality did not significantly influence truth/lie discrimination or credibility judgments. These findings suggest that while cue knowledge may support the recognition of truthful statements, it does not aid in the identification of lies, nor does the presentation format substantially impact veracity judgments.

Does road safety cointegrate with socio-economic conditions in rich developing countries?

Despite significant progress in road safety in developed countries, it remains a persistent and critical challenge in the developing world. This study investigates the long- and short-term relationships between socio-economic conditions and road safety performance in affluent developing countries, using the United Arab Emirates (UAE) as a case study. Employing an autoregressive distributed lag (ARDL) cointegration error-corrected model with data from 1980 to 2024 (sourced from the UAE Federal Government, the World Bank, and UN World Population Prospects), the analysis examines the link between the road crash severity index (fatalities to total injuries) and core socio-economic variables—GDP per capita, unemployment rate, and population density—while controlling for traffic law enforcement via fines. The findings confirm a long-term equilibrium, with an error correction term indicating road safety adjusts to socio-economic shocks at a rapid annual rate of 60%. Granger-causality tests further establish that these socio-economic factors significantly influence road safety outcomes, a concern underscored by an identified upward trend in crash severity. We conclude that socio-economic conditions are a fundamental determinant of road safety, highlighting the necessity for policy interventions that move beyond traditional engineering solutions. Consequently, road safety must be reframed not solely as a transportation concern but as an integral objective of public health and socioeconomic policy, which requires a collaborative, multi-sectoral approach to forge a resilient, safe system.

Research on joint vehicle routing optimization considering multiple distribution centers

In order to address the problem of efficiently distributing to multiple demand points within the city and multiple distribution centers on the urban fringes, this paper considers decision-making issues such as the selection of distribution centers and the planning of delivery routes. With the objective of minimizing the total transportation time, an integer programming model for the urban vehicle delivery network is constructed. Moreover, a class of column generation algorithms is designed for the characteristics of the problem. CPLEX is used to compare and analyze the solutions of several examples. The results verify the feasibility of the proposed column generation algorithm, and the practicality of the planning model is demonstrated through the analysis of examples generated from actual data.

Periplaneta americana extract ameliorates recurrent oral ulcers in rats by enhancing the intestinal epithelial barrier and regulating gut microbiota

Background Recurrent oral ulcers (ROUs) of oral mucosa disease are difficult to cure and relapse easily. Periplaneta americana extract (PAD), a raw material used in Kangfuxin Liquid and Yunnan Baiyao toothpaste, contains a variety of growth factors such as polypeptides and sticky sugar amino acids that promote tissue repair; this can encourage the growth of granulation tissue and reduce inflammation on wound surfaces. Objective In this study, we used a rat model of ROU induced by an antigen emulsifier to assess the ameliorative effects of PAD on rat ROUs and to explore its mechanism of action. Methods The effect of PAD in rats was evaluated by an enzyme-linked immunosorbent assay (ELISA) kit, hematoxylin-eosin staining, immunohistochemistry, reverse transcription quantitative PCR (RT-qPCR), and Western Blot (WB). 16S rRNA sequencing and gas chromatography-mass spectrometry (GC-MS) were used to detect the changes in intestinal flora and its metabolite short-chain fatty acids (SCFAs) in the feces of rats, respectively. Results PAD significantly reduced the infiltration of local inflammatory cells and significantly downregulated interleukin (IL)-6 and tumor necrosis factor (TNF-α), while upregulating IL-2, IL-10, and vascular endothelial growth factor (VEGF). In addition, PAD altered the diversity and abundance of the gut microbiota and increased fecal short-chain fatty acid levels. Notably, PAD can improve the pathological injury of the colon, enhance the intestinal barrier of the colon, and reduce the apoptosis of colon cells. Conclusion The oral administration of PAD may effectively treat ROUs via regulating metabolites and altering the composition of the intestinal microbiota, thereby improving the intestinal barrier function.

Conplastic FVB/N-mt129S6/SvEvTac mice: A new tool for cancer research

FVB/N mice, which are commonly used for cancer studies, have accelerated onset of endometrial cancer following developmental estrogenic chemical exposure. These mice also have a polymorphism in the mitochondrial gene, mt-Atp8 , leading to increased production of reactive oxygen species. We hypothesized that this polymorphism contributes to the enhanced endometrial cancer phenotype in FVB/N mice. To test this idea, we generated conplastic FVB/N-mt129S6/SvEvTac mice (FVB/N nuclear genome; 129S6/SvEvTac mitochondria: FVB/N-mt129). The impact of 129S6 versus FVB/N mitochondrial genomes on endometrial cancer development following neonatal exposure to the xenoestrogen, diethylstilbestrol, was tested by comparing the cancer phenotypes of FVB/N mice to FVB/N-mt129 mice. There was no difference in cancer incidence regardless of mitochondria source, but cancer grade was higher in the conplastic strain. Additionally, while the FVB/N genetic background is considered non-permissive for generation of pluripotent mouse embryonic stem cells, blastocysts from the conplastic background readily generated mouse embryonic stem cell clones that supported gene editing in culture and subsequently generated germline competent chimeric founder mice. FVB/N-mt129 mice are a potentially powerful resource for generating germline competent embryonic stem cells with an FVB/N nuclear genome and for studying cancer phenotypes.

Water treatment and E. coli in drinking water: Household responses to (invisible) water quality risks

Background: In 2024, an estimated 4 billion people lacked access to safely managed drinking water, with the greatest risks for vulnerable groups such as young children. The objective of this paper is to 1) document the share of households exposed to E. coli contamination at the water source, 2) examine whether households are more likely to treat their water when the source is E. coli contaminated, and 3) assess how household water treatment relates to E. coli levels in stored drinking water. Methods: This paper analyzes nationally representative Multiple Indicator Cluster Surveys (MICS) data from 59,633 households in low- and middle-income countries across 25 countries in Sub-Saharan Africa, Latin America, and Asia, all of which conducted standardized water-quality tests both at the source and in stored household drinking water. The main analysis documents how household water-treatment behavior differs by the level of E. coli contamination in their water source using regression methods by controlling for the country-specific fixed effects and household characteristics. Results: The study reveals that 78% of households do not treat their water, While 20% relying on sources with high E. coli contamination levels (>100 MPN/100 ml). Despite this risk, the presence of E. coli is positively associated with a higher likelihood of treating water by just 3 to 5 percentage points, depending on contamination levels, from an average baseline of 21%. The association is stronger in Latin America and weaker in Africa. Moreover, analysis of stored household drinking water reveals that a substantial share of households that report treating their water still exhibit moderate to high E. coli contamination in stored samples. Conclusion: Households’ water-treatment behavior shows only a modest association with the invisible risk of contamination, and many households still have E. coli contaminated drinking water even after treating it. These results highlight the need for WASH strategies that go beyond promoting access to treatment—specifically, strategies that inform households about contamination risks and emphasize correct and consistent application to ensure safe drinking water.

Extended-spectrum β-lactamase and carbapenemase-producing Enterobacterales among adult patients and their family members at Tikur Anbessa Specialized Hospital, Addis Ababa, Ethiopia

Background Extended-spectrum β-Lactamase and Carbapenemase-producing Enterobacterales cause severe infections and currently, they are spreading beyond hospitals and becoming a serious global health concern. They often colonize the gut silently, facilitating the transmission of resistant bacteria between patients and family members. Objective We sought investigate the prevalence and molecular characteristics of Extended-spectrum β-Lactamases-producing Enterobacterales (ESBL-PE), Carbapenem-resistant Enterobacterales (CRE), and factors associated there in among admitted adult patients and their family members at Tikur Anbessa Specialized Hospital, Addis Ababa, Ethiopia. Methods A case-control study was conducted among 100 patients and their respective 100 family members from February 2023 to October 2023. Stool specimens were collected and processed using standard microbiological techniques. Antimicrobial susceptibility testing and ESBL production were determined using VITEK 2 system. Carbapenemase production was tested using modified carbapenem Inactivation method, and detection of resistance genes was performed by PCR. Result Intestinal colonization with ESBL-PE was higher in patients (39.0%) than their respective family members (24.0%) (P = 0.028). Among patients, ESBL production was common in E. coli , 40.3% and K. pneumoniae, 34.7% than their family members 24.2% and 22.2%, respectively. Of the ESBL-PE isolates, 84.6% from patients and 100% from family members carried at least one ESBL encoding gene, with bla CTX-M being the predominant. Colonization with CRE and Carbapenemase-Producing Carbapenem-Resistant Enterobacterales was found to be 19.0% and 10.0%, respectively. These were identified only among patients, with bla NDM and bla OXA-48 are the most prevalent genes. Older age (>53 years) (P = 0.02) and previous ICU admission (P < 0.001) showed significant association with ESBL-PE colonization. Conclusion ESBL-PE colonization was more prevalent in patients compared to their family members, with bla CTX-M identified as the most common gene. Exclusive detection of carbapenemase genes among patients, and the association of previous ICU admission with ESBL-PE colonization, highlights the need for targeted screening and strengthened infection prevention.

Briefing Chat: The canny cow that can use tools, and how babies share their microbiomes

Daily briefing: Why cancer might protect against Alzheimer’s disease

Integrated optimization of spatiotemporal resources at the intersection for delay minimization using genetic algorithm

Integrated optimization of spatiotemporal resources at the intersection (IOSTRI) is crucial for traffic signal control, where both the lane allocation and signal control plans are optimized in a unified framework. This paper addresses the IOSTRI problem with delay minimization, formulating it as a binary mixed-integer nonlinear program (BMINLP) model that fully incorporates all possible uses of shared lanes and lane utilization adjustments. A genetic algorithm tailored to the model’s characteristics is designed, where four modules named lane converter, signal plan converter, flow calculation function and delay calculation function are used to calculate the fitness of each solution. Numerical results show the proposed model and algorithm’s ability to adapt to diverse traffic flow distribution patterns. High-quality solutions are obtained within 40–55 seconds, representing a significant improvement over previous studies and satisfying the requirements for real-time adaptive control of a single intersection.

LeafAI: Interpretable plant disease detection for edge computing

In real-world agriculture, healthy plant leaves are significantly more common than diseased ones. This natural class imbalance presents challenges in automated plant disease detection, as analyzing each leaf with computationally intensive deep-learning models is problematic, leading to inefficiency and increased resource consumption. To tackle this challenge and promote sustainable AI solutions, this study presents an iterative, hybrid AI approach that boosts computational efficiency, interpretability, and scalability for real-time disease detection. This hybrid system operates in two stages: first, a lightweight traditional machine learning classifier performs binary classification to quickly separate and exclude healthy leaves, followed by a deep learning model (ResNet, DenseNet, MobileNet, and EfficientNet) that classifies the specific disease in the smaller group of diseased leaves. This two-stage method minimizes computational load while maintaining high classification accuracy. Additionally, this study uses Explainable AI (XAI) methods, particularly Gradient-weighted Class Activation Mapping (Grad-CAM), to generate heatmaps. These heatmaps highlight the image areas that most significantly influence the model’s predictions, thereby improving transparency and refining the feature extraction process. The proposed hybrid model, comprising Logistic Regression and Mobilenetv3, offers up to 77.6% faster inference than conventional deep learning models with only about 3% accuracy loss. For a large-scale test of 1,227 images on an entry-level laptop, the hybrid model reduced the total inference time from 4,548 seconds to just 1,010.13 seconds, with minimal CPU load. By addressing class imbalance, optimizing inference efficiency, and incorporating explainable AI, this work contributes a scalable, sustainable, and trustworthy solution for plant disease detection in precision agriculture.

ZAK activation at the collided ribosome

Abstract Ribosome collisions activate the ribotoxic stress response mediated by the MAP3K ZAK, which in turn regulates cell-fate consequences through downstream phosphorylation of the MAPKs p38 and JNK 1 . Despite the critical role of ZAK during cellular stress, a mechanistic and structural understanding of ZAK–ribosome interactions and how these lead to activation remain elusive. Here we combine biochemistry and cryo-electron microscopy to discover distinct ZAK–ribosome interactions required for constitutive recruitment and for activation. We find that upon induction of ribosome collisions, interactions between ZAK and the ribosomal protein RACK1 enable its activation by dimerization of its SAM domains at the collision interface. Furthermore, we discover how this process is negatively regulated by the ribosome-binding protein SERBP1 to prevent constitutive ZAK activation. Characterization of novel SAM variants as well as a known pathogenic variant of the SAM domain of ZAK supports a key role of the SAM domain in regulating kinase activity on and off the ribosome, with some mutants bypassing the ribosome requirement for ZAK activation. Collectively, our data provide a mechanistic blueprint of the kinase activity of ZAK at the collided ribosome interface.

US science in 2026: five themes that will dominate Trump’s second year

3D-printed low-voltage-driven ciliary hydrogel microactuators

Abstract Micrometre-sized, densely packed natural cilia that perform non-reciprocal 3D motions with dynamically tunable collective patterns are crucial for biological processes such as microscale locomotion 1 , nutrient acquisition 2 , cell trafficking 3–5 and embryonic and neurological development 6–8 . However, replicating these motions in artificial systems remains challenging given the limits of scalable, locally controllable soft-bodied actuation at the micrometre scale. Overcoming this challenge would enhance our understanding of ciliary dynamics, clarify their biological importance and enable new microscale devices and bioinspired technologies. Here we show a previously unrecognized fast electrical response of micrometre-scale hydrogels, induced by voltages down to 1.5 V without hydrolysis, with bending motions driven by ion migration across a nanometre-scale hydrogel network 3D-printed by two-photon polymerization, occurring within milliseconds. On the basis of these findings, we print gel microcilia arrays composed of a soft acrylic acid-co-acrylamide (AAc-co-AAm) hydrogel (modulus of approximately 1,000 Pa) that respond to electrical stimuli within milliseconds. Each microcilium measures 2–10 µm in diameter and 18–90 µm in height, achieving 3D rotational bending motion at up to 40 Hz, mirroring the geometry and dynamics of natural cilia. These gel microcilia maintain functionality after 330,000 continuous actuation cycles with less than 30% performance degradation. The gel microcilia arrays can be integrated on flexible polyimide substrates and fabricated at large scale using conventional lithography techniques. They also offer individual dynamic control by means of microelectrode arrays and enable fluid manipulation and particle transport at the micrometre scale.