Developing deep LSTMs with later temporal attention for predicting COVID-19 severity, clinical outcome, and antibody level by screening serological indicators over time.

OBJECTIVE: The clinical course of COVID-19, as well as the immunological reaction, is notable for its extreme variability. Identifying the main associated factors might help understand the disease progression and physiological status of COVID-19 patients. The dynamic changes of the antibody against Spike protein are crucial for understanding the immune response. This work explores a temporal attention (TA) mechanism of deep learning to predict COVID-19 disease severity, clinical outcomes, and Spike antibody levels by screening serological indicators over time. METHODS: We use feature selection techniques to filter feature subsets that are highly correlated with the target. The specific deep Long Short-Term Memory (LSTM) models are employed to capture the dynamic changes of disease severity, clinical outcome, and Spike antibody level. We also propose deep LSTMs with a TA mechanism to emphasize the later blood test records because later records often attract more attention from doctors. RESULTS: Risk factors highly correlated with COVID-19 are revealed. LSTM achieves the highest classification accuracy for disease severity prediction. Temporal Attention Long Short-Term Memory (TA-LSTM) achieves the best performance for clinical outcome prediction. For Spike antibody level prediction, LSTM achieves the best permanence. CONCLUSION: The experimental results demonstrate the effectiveness of the proposed models. The proposed models can provide a computer-aided medical diagnostics system by simply using time series of serological indicators.

Isoflavone glucoside genistin, an inhibitor targeting Sortase A and Listeriolysin O, attenuates the virulence of Listeria monocytogenes in vivo and in vitro.

As a common intracellular facultative anaerobic Gram-positive bacterium, Listeria monocytogenes (L. monocytogenes) exhibits strong resistance to extreme environments, such as low temperature and a wide range of pH values, causing contamination in food production and processing. Sortase A (SrtA) and listeriolysin O (LLO), two crucial virulence factors of L. monocytogenes, are widely recognized as potential targets for the development of anti-L. monocytogenes infection drugs. In this study, we found that genistin simultaneously inhibits the peptidase activity of SrtA and the hemolytic activity of LLO without affecting the growth of L. monocytogenes, alleviating concerns about developing resistance. Furthermore, we demonstrated that genistin reduces L. monocytogenes biofilm formation and invasion of human colorectal cancer (Caco-2) cells. Subsequent mechanistic studies revealed that genistin inhibited LLO-mediated Caco-2 cell damage by blocking LLO oligomerization. Fluorescence quenching assay revealed the potential binding mode of SrtA and LLO to genistin. Genistin might bind to the active pocket of SrtA through residues Leu33, Asn29, and Met40, interacting with D1 domain of LLO involved in oligomerization and pore formation through residues Asn259. Studies in infection models revealed that genistin reduces mortality and pathological damage in mice infected with L. monocytogenes. These results indicate that genistin is a promising anti-virulence agent that could be considered an alternative candidate for the treatment of L. monocytogenes infection.

DILS: depth incremental learning strategy.

There exist various methods for transferring knowledge between neural networks, such as parameter transfer, feature sharing, and knowledge distillation. However, these methods are typically applied when transferring knowledge between networks of equal size or from larger networks to smaller ones. Currently, there is a lack of methods for transferring knowledge from shallower networks to deeper ones, which is crucial in real-world scenarios such as system upgrades where network size increases for better performance. End-to-end training is the commonly used method for network training. However, in this training strategy, the deeper network cannot inherit the knowledge from the existing shallower network. As a result, not only is the flexibility of the network limited but there is also a significant waste of computing power and time. Therefore, it is imperative to develop new methods that enable the transfer of knowledge from shallower to deeper networks. To address the aforementioned issue, we propose an depth incremental learning strategy (DILS). It starts from a shallower net and deepens the net gradually by inserting new layers each time until reaching requested performance. We also derive an analytical method and a network approximation method for training new added parameters to guarantee the new deeper net can inherit the knowledge learned by the old shallower net. It enables knowledge transfer from smaller to larger networks and provides good initialization of layers in the larger network to stabilize the performance of large models and accelerate their training process. Its reasonability can be guaranteed by information projection theory and is verified by a series of synthetic and real-data experiments.

Commercialized artemisinin derivatives combined with colistin protect against critical Gram-negative bacterial infection.

The emergence and spread of the mcr-1 gene and its mutants has immensely compromised the efficient usage of colistin for the treatment of drug-resistant Gram-negative bacterial infection in clinical settings. However, there are currently no clinically available colistin synergis. Here we identify artemisinin derivatives, such as dihydroartemisinin (DHA), that produces a synergistic antibacterial effect with colistin against the majority of Gram-negative bacteria (FIC < 0.5) without induced resistance, particularly those carrying the mcr-1 gene. Mechanism analysis reveals the direct engagement of DHA with the active center of MCR-1 to inhibit the activity of MCR-1. Meanwhile, the results from transcriptome and electron microscope analysis show that DHA could also simultaneously affect the flagellar assembly and the energy metabolism of bacteria. Moreover, in the mouse infection models of Gram-negative bacteria, combination therapy shows remarkable treatment benefits, as shown by an improved survival rate, reduced morbidity, alleviated pathological injury and decreased bacterial loading. Due to the generally safe profile of specialized malaria medication administration in humans, artemisinin derivatives are a promising class of multi-target inhibitors on bacterial resistance and virulence that can be used to extend the usage life of colistin and to tackle the inevitability of serious bacterial infection with colistin.

Polyphosphate Kinase Is Required for the Processes of Virulence and Persistence in Acinetobacter baumannii.

Acinetobacter baumannii, one of the most successful bacteria causing severe nosocomial infection, was identified as a top-priority pathogen by the WHO. Thus, genetic manipulations to clarify the potential targets for fighting A. baumannii resistance and virulence are vital. Polyphosphate (polyP) kinase (PPK) is conserved in nearly all bacteria and is responsible for polyP formation, which is associated with bacterial pathogenicity and antibiotic resistance. In this study, ppk1-deficient (Δppk1::Apr), ppk1-complemented (Δppk1::Apr/PJL02-ppk1), and wild-type strains of A. baumannii ATCC 17978 were used to determine the influence of PPK1 on A. baumannii virulence and persistence mainly by polyP quantification, surface motility, biofilm formation, and bacterial persistence assays. Our work found that PPK1 is indispensable for polyP formation in vivo and that the motility of the PPK1-deficient strain was significantly impaired due to the lack of a pilus-like structure typically present compared with the complemented and wild-type strains. The deficiency of PPK1 also inhibited the biofilm formation of A. baumannii and decreased bacterial persistence under stimuli of high-concentration ampicillin (Amp) treatment, H2O2 stress, heat shock, and starvation stress. Furthermore, ppk1-deficient bacterium-infected mice showed a significantly reduced bacterial load and a decreased inflammatory response. However, complementation with PPK1 effectively rescued the impaired virulence and persistence of ppk1-deficient A. baumannii. In addition, metabonomic analysis revealed that PPK1 was associated with glycerophospholipid metabolism and fatty acid biosynthesis. Taken together, our results suggest that targeting PPK1 to control A. baumannii pathogenicity and persistence is a feasible strategy to fight this pathogen. IMPORTANCE A. baumannii was identified as a top-priority pathogen by the WHO due to its antibiotic resistance. Meanwhile, the pathogenicity of A. baumannii mediated by several vital virulence factors also cannot be ignored. Here, the role of PPK1 in A. baumannii was also explored. We found that the motility ability and biofilm formation of a PPK1-deficient strain were significantly impaired. Furthermore, PPK1 was essential for its persistence maintenance to resist stimuli of high-concentration Amp treatment, H2O2 stress, heat shock, and starvation stress. Metabonomic analysis revealed that PPK1 was associated with glycerophospholipid metabolism and fatty acid biosynthesis. In addition, ppk1-deficient bacterium-infected mice showed significantly reduced bacterial loads and a decreased inflammatory responses in vivo. Together, our results suggest that PPK1 is vital for A. baumannii pathogenicity and persistence.

Kaempferol-Driven Inhibition of Listeriolysin O Pore Formation and Inflammation Suppresses Listeria monocytogenes Infection.

Listeria monocytogenes remains a nonnegligible cause of foodborne infection, posing a critical threat to public health. Under the global antibiotic crisis, novel alternative approaches are urgently needed. The indispensable role of listeriolysin O (LLO) in the intracellular life cycle, barrier penetration, colonization, and systemic dissemination of L. monocytogenes renders it a potent drug target, which means curbing L. monocytogenes via interfering with LLO-associated pathogenic mechanisms. Here, we identified kaempferol, a natural small molecule compound, as an effective LLO inhibitor that engaged the residues Glu437, Ile468, and Tyr469 of LLO, thereby suppressing LLO-mediated membrane perforation and barrier disruption. Moreover, we found that kaempferol also suppressed host-derived inflammation in a distinct way independent of LLO inhibition. The in vivo study revealed that kaempferol treatment significantly reduced bacterial burden and cytokine burst in target organs, thereby effectively protecting mice from systemic L. monocytogenes infection. Our findings present kaempferol as a potential therapeutic application for L. monocytogenes infection, which is less likely to induce drug resistance than antibiotics because of its superiority of interfering with the pathogenesis process rather than exerting pressure on bacterial viability. IMPORTANCE Currently, we are facing a global crisis of antibiotic resistance, and novel alternative approaches are urgently needed to curb L. monocytogenes infection. Our study demonstrated that kaempferol alleviated L. monocytogenes infection via suppressing LLO pore formation and inflammation response, which might represent a novel antimicrobial-independent strategy to curb listeriosis.

Nonconvex Nonlocal Tucker Decomposition for 3D Medical Image Super-Resolution.

Limited by hardware conditions, imaging devices, transmission efficiency, and other factors, high-resolution (HR) images cannot be obtained directly in clinical settings. It is expected to obtain HR images from low-resolution (LR) images for more detailed information. In this article, we propose a novel super-resolution model for single 3D medical images. In our model, nonlocal low-rank tensor Tucker decomposition is applied to exploit the nonlocal self-similarity prior knowledge of data. Different from the existing methods that use a convex optimization for tensor Tucker decomposition, we use a tensor folded-concave penalty to approximate a nonlocal low-rank tensor. Weighted 3D total variation (TV) is used to maintain the local smoothness across different dimensions. Extensive experiments show that our method outperforms some state-of-the-art (SOTA) methods on different kinds of medical images, including MRI data of the brain and prostate and CT data of the abdominal and dental.

Experimental study on the desorption effect of penetrant on gas-containing coal.

Two non-ionic reagents, polyethylene glycol 4000 and Tween-80, two anionic reagents, sodium dodecyl benzenesulfonate and sodium lauryl sulfate, and a mixture of these non-ionic and anionic reagents were used as penetrants. The processes of replacement desorption and relief-pressure desorption of gas-containing coal were studied, the influence of the penetrant on the amount of gas replacement desorption and relief-pressure desorption was explored, and the change rule of the amounts of gas replacement desorption and relief-pressure desorption was analysed. The results show that the increase rate of the replacement desorption amount of the mixed penetrant is 11.81%-34.75%, and the decrease rate of the relief-pressure desorption amount is 51.68%-72.69%, which are higher values than those with a single penetrant. As the mass fraction of penetrant increases within the range of 0.5%~2%, the capacity of gas replacement desorption and hindering gas relief-pressure desorption will increase. At the same mass fraction, the effect of the mixed penetrant is better than that of the anionic penetrant, which in turn is better than that of the non-ionic penetrant.

Study on parameter optimization of deep hole cumulative blasting in low permeability coal seams.

Coal seam gas extraction is an important means of exploiting and utilizing gas resources, as well as a means of preventing coal mine disasters. To improve gas extraction efficiency in high gas and low permeability coal seams while ensuring blasting security, deep hole cumulative blasting parameters were optimized. ANSYS/LS-DYNA software is used to establish a 3-dimensional cumulative blasting model. By comparing and analyzing the blasting stress nephograms, stress time-history curves, and crack expansion curves, the optimal blasthole diameter, charge position, and charge length are obtained. Based on the numerical simulation results, a field test was carried out in the No. 10 coal seam of the Pingdingshan coal mine. The test results show that after cumulative blasting, the gas concentration was increased by an average of 2.25 times, the gas purity was increased by an average of 3.78 times, the permeability coefficient of the coal seam was increased by 21 times, and the effective radius of blasting was up to 7 m. The positive effects of deep hole cumulative blasting parameter optimization on the pressure relief and permeability enhancement of a high gas and low permeability coal seam were determined, which can provide a reference for other similar working faces to implement this technology.

Crystallization Kinetics Control Enabled by a Green Ionic Liquid Additive toward Efficient and Stable Carbon-Based Mesoscopic Perovskite Solar Cells.

Carbon-based mesoscopic perovskite solar cells (MPSCs) are becoming one of the most competitive photovoltaic technologies owing to their lower manufacturing cost and excellent stability. In this work, methylammonium acetate (MAAc), an ionic liquid additive, is added into methylammonium lead triiodide (MAPbI3) perovskite and is used to fabricate high-performance MPSCs. Systematic and detailed studies have shown that the MAAc interacts with PbI2 preferentially to form a MAPbI3-x(Ac)x intermediate phase that can effectively control the crystallization kinetics of MAPbI3 in the triple-mesoscopic layer. MAPbI3 films with an appropriate amount of MAAc exhibit higher crystallinity, lower defect density, and dense pore filling, which effectively reduce carrier non-radiative recombination loss in MPSCs. As a result, a champion power conversion efficiency (PCE) of 13.54% is obtained based on the optimized MAAc-engineered MPSCs. The PCE is 24% higher than 10.90% of the control devices. Moreover, unencapsulated MAAc-engineered MPSCs retain 90% of their initial PCE after being stored in the dark for 50 days under ambient atmosphere, which demonstrates much better air stability than control devices. This work provides an effective strategy for developing efficient and stable carbon-based MPSCs with an eco-friendly ionic liquid additive.

Synergistic Effect of Defect Passivation and Crystallization Control Enabled by Bifunctional Additives for Carbon-Based Mesoscopic Perovskite Solar Cells.

The emerging carbon-based mesoscopic perovskite solar cells (MPSCs) are known as one of the most promising candidates for photovoltaic applications thanks to their screen-printing process and excellent stability. Unfortunately, they usually suffer from serious defects because it is challenging to realize sufficient mesopore filling of the perovskite precursor solution throughout the triple-mesoporous scaffold. Herein, a bifunctional additive, biuret, endowed with both carbonyl and amino groups, was designed to realize a convenient fabrication approach for controllable crystallization of the precursor solution. Owing to the strong coordination ability with perovskite components, the incorporation of biuret can not only regulate crystallization kinetics allowing for the growth of high-quality perovskite crystals but also associate with uncoordinated ions for defect passivation to enhance the overall photovoltaic performance of MPSCs. A champion power conversion efficiency (PCE) of 13.42% with an enhanced short-circuit current density of 19.49 mA cm-2 and a much higher open-circuit voltage of 0.96 V was achieved for the device doped with 3 mol % biuret, which is 26% higher than that of the control device (10.66%). Moreover, the unencapsulated devices with biuret incorporation demonstrated superior stability, maintaining over 90% of the original PCE after 50 days of storage under ambient conditions. This work helps exploit bifunctional additive strategies for simultaneous defect passivation and crystallization control toward high-efficiency and long-term stability of carbon-based MPSCs.

Process optimization and mechanism study of acid red G degradation by electro-Fenton-Feox process as an in situ generation of H2O2.

Dye-contaminated wastewaters are industrial wastewaters that are difficult to treat using traditional biochemical and physicochemical methods. In the present work, the acid red G was removed as a model pollutant by the electro-Fenton process for the first time. The anode and cathode used by the electro-Fenton process were iron plate and graphite felt, respectively. It was concluded that under the optimal conditions of current density = 20 mA cm-2, pH = 3 and initial Na2SO4 concentration = 0.2 M, the removal rate of acid red G (ARG) with an initial concentration of 300 mg L-1 could reach 94.05% after 80 min of electrolysis. This reveals that the electro-Fenton-Feox process used in this work has an excellent removal efficiency on acid red G. The required reagents (Fe2+ and H2O2) were generated by the electrode reaction, while the optimal generation conditions and mechanism of •OH, H2O2, and Fe2+ were investigated. By testing •OH, H2O2, and Fe2+ agents at different pH and current densities, it was revealed that the electro-Fenton reaction was most efficient when the current density was 20 mA cm-2, and the pH was 3. Moreover, the removal rate of ARG is consistent with first-order reaction kinetics.

Improved Pore-Filling and Passivation of Defects in Hole-Conductor-Free, Fully Printable Mesoscopic Perovskite Solar Cells Based on d-Sorbitol Hexaacetate-Modified MAPbI3.

For hole-conductor-free, fully printable mesoscopic perovskite solar cells (MPSCs), it is difficult to achieve free and efficient diffusion of perovskite precursors in micron-scale porous structures. Thus, the wettability of the perovskite precursor is one of the most crucial factors that determine the performance of MPSCs. Here, d-sorbitol hexaacetate (DSHA) is introduced as an additive for fabricating hole-conductor-free, fully printable MPSCs based on methylammonium lead iodide (MAPbI3). The fabricated MPSCs exhibited an efficiency of 14.33%. Moreover, the influence of DSHA on the optical properties, morphology, and filling of perovskite in the MPSCs has been systematically investigated. The results revealed that DSHA effectively optimized the morphology, improved the pore-filling, and passivated the defects of perovskite films. Remarkably, the unencapsulated MPSCs retained 93% of their original power-conversion efficiency (PCE) after 45 days of storage in air with humidity of 50 ± 5%.

A glassy carbon electrode modified with a bismuth film and laser etched graphene for simultaneous voltammetric sensing of Cd(II) and Pb(II).

Polyimide (PI) sheets were laser etched to obtain graphene-based carbon nanomaterials (LEGCNs). These were analyzed by scanning electron microscopy, X-ray diffraction and Raman spectroscopy which confirmed the presence of stacked multilayer graphene nanosheets. Their large specific surface and large number of edge-plane active sites facilitate the accumulation of metal ions. A glassy carbon electrode (GCE) with an in-situ plated bismuth film was modified with the LEGCNs to give a sensor with satisfactory response for the simultaneous determination of cadmium(II) and lead(II) by means of square wave anodic stripping voltammetry. It appears that is the first report on an electrochemical sensor based on the use of laser etched graphene for determination of heavy metal ions. Figures of merit for detection of Cd(II) include (a) a low and well separated working potential of -0.80 V (vs. Ag/AgCl), (b) a wide linear range (from 7 to 120 µg·L-1), and a low detection limits 0.47 µg·L-1. The respective data for Pb(II) are (a) -0.55 V, (b) 5 to 120 µg·L-1, and (c) 0.41 µg·L-1. The modified GCE displays remarkable repeatability, reproducibility, selectivity and stability. The sensor was applied to the simultaneous determination of Cd(II) and Pb(II) in spiked real water samples. The results confirm that the laser etching technique is an efficient tool for the preparation of carbon nanomaterials with high quality and great sensing performance. Graphical abstract Bismuth film and laser etched graphene-modified glassy carbon electrode (BF-LEGCN/GCE) for the simultaneous determination of cadmium(II) and lead(II) by square wave anodic stripping voltammetry.

Effects of the addition of municipal solid waste incineration fly ash on the behavior of polychlorinated dibenzo-p-dioxins and furans in the iron ore sintering process.

Raw materials were co-sintered with municipal solid waste incineration (MSWI) fly ash through iron ore sintering to promote the safe treatment and utilization of MSWI fly ash. To assess the feasibility of this co-sintering method, in this study, the effects of the addition of MSWI fly ash on the formation and emission of polychlorinated dibenzo-p-dioxins and furans (PCDD/Fs) were estimated via iron ore sintering pot experiments. During co-sintering, most of the PCDD/Fs in the added MSWI fly ash were decomposed and transformed into PCDD/Fs associated with iron sintering, and the concentrations of lower- and mid-chlorinated congeners increased. As there was a sufficient chlorine source and the sintering bed permeability was decreased by the addition of MSWI fly ash, the PCDD/F concentration in the exhaust gas increased. The mass emission of PCDD/Fs decreased; however, the emission of toxic PCDD/Fs increased beyond the total emissions from the independent MSW incineration and iron ore sintering processes due to the transformation of PCDD/F congeners. The co-sintering may be an important solution after technological improvements in the flue gas cleaning system and PCDD/F formation inhibition procedures.

One pot synthesis of dandelion-like polyaniline coated gold nanoparticles composites for electrochemical sensing applications.

In this work, we report a facile and green strategy for one pot and in-situ synthesis of a dandelion-like conductive polyaniline coated gold nanoparticle nanocomposites (Au@PANI). The Au@PANI was characterized by SEM, TEM, XRD, TGA, FTIR, UV-vis and conductivity measurement, respectively. Newly-designed Au@PANI materials possessed a significantly high conductivity and strong adsorption capability. Thus, the Au@PANI modified glassy carbon electrode (GCE) was utilized for construct a novel electrochemical sensor for the simultaneous assay of Pb2+ and Cu2+ using square wave anodic stripping voltammetry (SWASV). Under the optimized conditions, an excellent electrochemical response in the simultaneous of Pb2+ and Cu2+ with detection limit of 0.003 and 0.008 µM (S/N = 3), respectively. Moreover, the prepared sensors realized an excellent reproducibility, repeatability and long term stability, as well as reliable practical assays in real water samples. Besides, the possible formation mechanism and sensing mechanism of Au@PANI nanocomposites have been discussed in detail. We believe this study provides a novel method of fabrication of noble metal nanoparticles decorated conducting polymer materials for the electrochemical sensing applications.

Silver@Nitrogen-Doped Carbon Nanorods as a Highly Efficient Electrocatalyst for the Oxygen Reduction Reaction in Alkaline Media.

In recent years, various platinum-free catalysts for the oxygen reduction reaction (ORR) have attracted great attention due to the limited natural abundance and high cost of platinum. Herein, Ag@N-C (N-C: nitrogen-doped carbon) nanorods for the ORR were synthesized through chemical polymerization and pyrolysis methods by using pyrrole and silver nitrate as raw materials. Pyrolysis could significantly increase the specific surface area of as-synthesized catalysts and convert pyrrolic-N into graphitic-N and pyridinic-N. The results of electrochemical tests show that the Ag@N-C-900 catalyst (pyrolyzed at 900 °C) exhibits highly efficient ORR catalytic activity, improved stability, and better methanol resistance in comparison to that of Pt/C catalyst in alkaline media.

Characterization of PAHs in size-fractionated submicron atmospheric particles and their association with the intracellular oxidative stress.

The submicron atmospheric particulate matters (PMs), consisting of 7 size-resolved fractions, were collected by two Dekati low pressure impactors from December 2015 to January 2016 in Hangzhou, China. The chemical analysis revealed the higher accumulation efficiency of polycyclic aromatic hydrocarbons (PAHs) in the finer submicron PMs, and 77.0% of particulate PAHs in PM1 were associated with PM0.4. Moreover, the BaP equivalent concentrations (BaPeq) were evaluated for size-fractionated submicron PMs, indicating that 77.2% of carcinogenicity of particulate PAHs in PM1 were associated with PM0.4. In the cytotoxic tests, a significant size-dependent decrease of glutathione (GSH) level was observed in the PMs-exposed human pulmonary epithelial cells (A549), demonstrating the greater intracellular oxidative stress induced by the finer submicron PMs. The results also suggested that 82.4% of PM1-induced oxidative stress were associated with PM0.4, and the intracellular oxidative stress was significantly correlated with the particulate PAHs. Therefore, besides PM10, PM2.5 and PM1, special attentions should be given to PM0.4 as well.

Preparation and adsorption property of xylan/poly(acrylic acid) magnetic nanocomposite hydrogel adsorbent.

Adsorbents based on natural polysaccharides have attracted increasing interest because of their low-cost and biodegradability, particularly, polysaccharide-based nanocomposite adsorbents. In this study the xylan/poly(acrylic acid) magnetic nanocomposite hydrogel adsorbent was prepared from wheat straw xylan and Fe3O4 nanoparticles, and its adsorption property was studied on methylene blue removal. The prepared hydrogel adsorbent had a semi-interpenetrating network structure and exhibited a macro-porous structure with interconnected porous channels. Super-paramagnetic characteristic behavior was observed from magnetic analysis using a vibrating sample magnetometer. The optimum condition for methylene blue adsorption on the adsorbent was found at pH 8 with an adsorbent dosage of 3g/L and an initial concentration of 400mg/L, and the removal percentage reached above 90%. The adsorption isotherm of methylene blue on the prepared hydrogel adsorbent was fitted to the Langmuir model, and the pseudo-second-order kinetic model could describe the adsorption process. All obtained results indicated that the prepared hydrogel adsorbent is promising for water treatment applications.

Corneal collagen fibril changes after ultraviolet a/riboflavin corneal crosslinking.

PURPOSE: The aim was to investigate the changes in collagen type 1 and type 3 in rabbit corneas undergoing corneal crosslinking with ultraviolet A and riboflavin and to analyze the possible mechanisms of corneal haze formation. METHODS: After removal of the central epithelium, the right corneas of 60 New Zealand rabbits were crosslinked with riboflavin and ultraviolet A, and 10 additional rabbits were used as the control group. The animals were killed 3, 7, 15, 30, 90, and 180 days postoperatively. Collagen type 1 and type 3 were analyzed using picrosirius red stain by means of polarized light microscopy. The biochemical changes in collagen type 3 at the time points indicated above were determined by Western blot analyses. RESULTS: Collagen type 3 was significantly increased 30 days after corneal crosslinking compared with that in the control cornea, gradually increased until reaching its maximum value 90 days after riboflavin and ultraviolet A crosslinking, and then decreased until it returned to the normal state 180 days after crosslinking. There were no significant changes in collagen type 1 over time after corneal crosslinking. In agreement with the picrosirius red staining results, the western blot analyses showed that collagen type 3 was detected 15 days after the crosslinking treatment and continued to be present. However, 180 days after the crosslinking treatment, collagen type 3 could not be found in the crosslinked corneas. CONCLUSIONS: These findings suggest that ultraviolet A/riboflavin crosslinking results in collagen type 3 synthesis and degradation, which may offer at least a partial explanation for the formation of corneal haze.