OBJECTIVES: Accurate identification of risk factors for precocious puberty is essential for clinical diagnosis and management, yet the performance of natural language processing methods applied to unstructured electronic medical record (EMR) data remains to be fully evaluated. This study aims to assess the performance of a prompt engineering method for extracting individual risk factors of precocious puberty from EMRs. METHODS: Based on the capacity and role-insight-statement-personality-experiment (CRISPE) prompt framework, both simple and optimized prompts were designed to guide the large language model GLM-4-9B in extracting 10 types of risk factors for precocious puberty from 653 EMRs. Accuracy, precision, recall, and F1-score were used as evaluation metrics for the information extraction task. RESULTS: Under simple and optimized prompt conditions, the overall accuracy, precision, recall, and F1-score of the model were 84.18%, 98.09%, 81.99%, and 89.32% versus 97.15%, 98.31%, 98.16%, and 98.23%, respectively. The optimized prompts achieved more stable performance across age (<9 years vs ≥9 years) and visit-time (<2023 vs ≥2023) subgroups compared with simple prompts. The accuracy range for extracting each risk factor was 60.03%-97.24%, while with optimized prompts, the range improved to 92.19%-99.85%. The largest performance improvement occurred for "beverage intake" (60.03% vs 92.19%), and the smallest for "maternal age of menarche" (97.24% vs 99.23%). In comparing distributions among simple prompts, optimized prompts, and ground truth, statistically significant differences were observed for snack intake, beverage intake, soy milk intake, honey intake, supplement use, tonic use, sleep quality, and sleeping with the light on (all P<0.001), while exercise (P=0.966) and maternal menarche age (P=0.952) showed no significant differences. CONCLUSIONS Compared with simple prompts, optimized prompts substantially improved the extraction performance of individual risk factors for precocious puberty from EMRs, underscoring the critical role of prompt engineering in enhancing large language model performance.
Humans ; Puberty, Precocious/epidemiology* ; Risk Factors ; Electronic Health Records ; Female ; Child ; Natural Language Processing

L-theanine is an important natural non-protein amino acid that is widely used in food and medicine. Although in previous studies, a microbial fermentation method for L-theanine without the addition of ethylamine has been developed, the conversion rate of this process needs to be further improved. In this study, we constructed a de novo synthesis pathway of L-theanine with glucose as the substrate. First, an in vitro transformation pathway containing ω-transaminase (TA) and γ-glutamylmethylamide synthetase (GMAS) was designed, optimized, and introduced into the chassis strain Escherichia coli K12 W3110 to achieve de novo synthesis of L-theanine. To improve the synthesis efficiency through metabolic engineering, we increased the copies of the GMAS gene gams and the TA gene spuC and enhanced the expression of the aldehyde dehydrogenase gene eutE to provide sufficient acetaldehyde substrate, knocked out the lactate dehydrogenase gene ldhA and the pyruvate formate lyase gene pflB to block bypass metabolism, and introduced the alanine dehydrogenase gene alD to recycle alanine. Furthermore, we over-expressed the phosphoenolpyruvate carboxylase gene ppc to enhance the carbon flux of the TCA cycle, knocked out the succinyl-CoA synthase gene sucCD to reduce the loss of downstream flux of TCA, and integrated the glutamate dehydrogenase gene gdh to enhance the supply of L-glutamate. Finally, the polyphosphate kinase gene ppk was introduced to the ATP cycle, which enhanced the energy supply in L-theanine production. The recombinant strain Tea11 produced 22.60 g/L L-theanine in a 5 L fermenter in 28 h, with a conversion rate of 41.71%. This synthetic pathway in this study balanced the relationship between the supply of ethylamine and the production of theanine, providing a new idea for metabolic engineering of microorganisms to produce L-theanine.
Glutamates/biosynthesis* ; Metabolic Engineering/methods* ; Escherichia coli/genetics* ; Fermentation ; Transaminases/metabolism* ; Amide Synthases/metabolism* ; Glucose/metabolism*

L-glutamic acid is the world's largest bulk amino acid product that is widely used in the food, pharmaceutical and chemical industries. Using Corynebacterium glutamicum G01 as the starting strain, the fermentation by-product alanine content was firstly reduced by knocking out the gene encoding alanine aminotransferase (alaT), a major by-product related to alanine synthesis. Secondly, since the α-ketoglutarate node carbon flow plays an important role in glutamate synthesis, the ribosome-binding site (RBS) sequence optimization was used to reduce the activity of α-ketoglutarate dehydrogenase and enhance the glutamate anabolic flow. The endogenous conversion of α-ketoglutarate to glutamate was also enhanced by screening different glutamate dehydrogenase. Subsequently, the glutamate transporter was rationally desgined to improve the glutamate efflux capacity. Finally, the fermentation conditions of the strain constructed using the above strategy were optimized in 5 L fermenters by a gradient temperature increase combined with a batch replenishment strategy. The glutamic acid production reached (135.33±4.68) g/L, which was 41.2% higher than that of the original strain (96.53±2.32) g/L. The yield was 55.8%, which was 11.6% higher than that of the original strain (44.2%). The combined strategy improved the titer and the yield of glutamic acid, which provides a reference for the metabolic modification of glutamic acid producing strains.
Glutamic Acid ; Corynebacterium glutamicum/genetics* ; Ketoglutaric Acids ; Metabolic Engineering ; Alanine

γ-aminobutyric acid can be produced by a one-step enzymatic reaction catalyzed by glutamic acid decarboxylase. The reaction system is simple and environmentally friendly. However, the majority of GAD enzymes catalyze the reaction under acidic pH at a relatively narrow range. Thus, inorganic salts are usually needed to maintain the optimal catalytic environment, which adds additional components to the reaction system. In addition, the pH of solution will gradually rise along with the production of γ-aminobutyric acid, which is not conducive for GAD to function continuously. In this study, we cloned the glutamate decarboxylase LpGAD from a Lactobacillus plantarum capable of efficiently producing γ-aminobutyric acid, and rationally engineered the catalytic pH range of LpGAD based on surface charge. A triple point mutant LpGAD^{S24R/D88R/Y309K} was obtained from different combinations of 9 point mutations. The enzyme activity at pH 6.0 was 1.68 times of that of the wild type, suggesting the catalytic pH range of the mutant was widened, and the possible mechanism underpinning this increase was discussed through kinetic simulation. Furthermore, we overexpressed the Lpgad and Lpgad^{S24R/D88R/Y309K} genes in Corynebacterium glutamicum E01 and optimized the transformation conditions. An optimized whole cell transformation process was conducted under 40 ℃, cell mass (OD₆₀₀) 20, 100 g/L l-glutamic acid substrate and 100 μmol/L pyridoxal 5-phosphate. The γ-aminobutyric acid titer of the recombinant strain reached 402.8 g/L in a fed-batch reaction carried out in a 5 L fermenter without adjusting pH, which was 1.63 times higher than that of the control. This study expanded the catalytic pH range of and increased the enzyme activity of LpGAD. The improved production efficiency of γ-aminobutyric acid may facilitate its large-scale production.
Glutamate Decarboxylase/genetics* ; Lactobacillus plantarum/genetics* ; Catalysis ; gamma-Aminobutyric Acid ; Hydrogen-Ion Concentration ; Glutamic Acid

Owing to incurable castration-resistant prostate cancer (CRPC) ultimately developing after treating with androgen deprivation therapy (ADT), it is vital to devise new therapeutic strategies to treat CRPC. Treatments that target programmed cell death protein 1 (PD-1) and programmed death ligand-1 (PD-L1) have been approved for human cancers with clinical benefit. However, many patients, especially prostate cancer, fail to respond to anti-PD-1/PD-L1 treatment, so it is an urgent need to seek a support strategy for improving the traditional PD-1/PD-L1 targeting immunotherapy. In the present study, analyzing the data from our prostate cancer tissue microarray, we found that PD-L1 expression was positively correlated with the expression of heterogeneous nuclear ribonucleoprotein L (HnRNP L). Hence, we further investigated the potential role of HnRNP L on the PD-L1 expression, the sensitivity of cancer cells to T-cell killing and the synergistic effect with anti-PD-1 therapy in CRPC. Indeed, HnRNP L knockdown effectively decreased PD-L1 expression and recovered the sensitivity of cancer cells to T-cell killing in vitro and in vivo, on the contrary, HnRNP L overexpression led to the opposite effect in CRPC cells. In addition, consistent with the previous study, we revealed that ferroptosis played a critical role in T-cell-induced cancer cell death, and HnRNP L promoted the cancer immune escape partly through targeting YY1/PD-L1 axis and inhibiting ferroptosis in CRPC cells. Furthermore, HnRNP L knockdown enhanced antitumor immunity by recruiting infiltrating CD8⁺ T cells and synergized with anti-PD-1 therapy in CRPC tumors. This study provided biological evidence that HnRNP L knockdown might be a novel therapeutic agent in PD-L1/PD-1 blockade strategy that enhanced anti-tumor immune response in CRPC.

As a model industrial host and microorganism with the generally regarded as safe (GRAS) status, Corynebacterium glutamicum not only produces amino acids on a large scale in the fermentation industry, but also has the potential to produce various new products. C. glutamicum usually encounters various stresses in the process of producing compounds, which severely affect cell viability and production performance. The development of synthetic biology provides new technical means for improving the robustness of C. glutamicum. In this review, we discuss the tolerance mechanisms of C. glutamicum to various stresses in the fermentation process. At the same time, we highlight new synthetic biology strategies for boosting C. glutamicum robustness, including discovering new stress-resistant elements, modifying transcription factors, and using adaptive evolution strategies to mine stress-resistant functional modules. Finally, prospects of improving the robustness of engineered C. glutamicum strains ware provided, with an emphasis on biosensor, screening and design of transcription factors, and utilizing the multiple regulatory elements.
Amino Acids/metabolism* ; Corynebacterium glutamicum/metabolism* ; Fermentation ; Metabolic Engineering ; Synthetic Biology

L-asparaginase hydrolyzes L-asparagine to produce L-aspartic acid and ammonia. It is widely distributed in microorganisms, plants and serum of some rodents, and has important applications in the pharmaceutical and food industries. However, the poor thermal stability, low catalytic efficiency and low yield hampered the further application of L-asparaginase. In this paper, rational design and 5' untranslated region (5'UTR) design strategies were used to increase the specific enzyme activity and protein expression of L-asparaginase derived from Rhizomucor miehei (RmAsnase). The results showed that among the six mutants constructed through homology modeling combined with sequence alignment, the specific enzyme activity of the mutant A344E was 1.5 times higher than the wild type. Subsequently, a food-safe strain Bacillus subtilis 168/pMA5-A344E was constructed, and the UTR strategy was used for the construction of recombinant strain B. subtilis 168/pMA5 UTR-A344E. The enzyme activity of B. subtilis 168/pMA5 UTR-A344E was 7.2 times higher than that of B. subtilis 168/pMA5-A344E. The recombinant strain B. subtilis 168/pMA5 UTR-A344E was scaled up in 5 L fermenter, and the final yield of L-asparaginase was 489.1 U/mL, showing great potential for industrial application.
Asparaginase/genetics* ; Bacillus subtilis/genetics* ; Industrial Microbiology ; Protein Engineering ; Rhizomucor/enzymology* ; Sequence Alignment

2-Hydroxybutyric acid (2-HBA) is an important intermediate for synthesizing biodegradable materials and various medicines. Chemically synthesized racemized 2-HBA requires deracemization to obtain optically pure enantiomers for industrial application. In this study, we designed a cascade biosynthesis system in Escherichia coli BL21 by coexpressing L-threonine deaminase (TD), NAD-dependent L-lactate dehydrogenase (LDH) and formate dehydrogenase (FDH) for production of optically pure (S)-2-HBA from bulk chemical L-threonine (L-Thr). To coordinate the production rate and the consumption rate of the intermediate 2-oxobutyric acid in the multi-enzyme cascade catalytic reactions, we explored promoter engineering to regulate the expression levels of TD and FDH, and developed a recombinant strain P21285FDH-T7V7827 with a tunable system to achieve a coordinated multi-enzyme expression. The recombinant strain P21285FDH-T7V7827 was able to efficiently produce (S)-2-HBA with the highest titer of 143 g/L and a molar yield of 97% achieved within 16 hours. This titer was approximately 1.83 times than that of the highest yield reported to date, showing great potential for industrial application. Our results indicated that constructing a multi-enzyme-coordinated expression system in a single cell significantly contributed to the biosynthesis of hydroxyl acids.
Escherichia coli/genetics* ; Formate Dehydrogenases ; Hydroxybutyrates ; Threonine Dehydratase

Leucine dehydrogenase (LDH) is the key rate-limiting enzyme in the production of L-2-aminobutyric acid (L-2-ABA). In this study, we modified the C-terminal Loop region of this enzyme to improve the specific enzyme activity and stability for efficient synthesis of L-2-ABA. Using molecular dynamics simulation of LDH, we analyzed the change of root mean square fluctuation (RMSF), rationally designed the Loop region with greatly fluctuated RMSF, and obtained a mutant EsLDHD2 with a specific enzyme activity 23.2% higher than that of the wild type. Since the rate of the threonine deaminase-catalyzed reaction converting L-threonine into 2-ketobutyrate was so fast, the multi-enzyme cascade catalysis system became unbalanced. Therefore, the LDH and the formate dehydrogenase were double copied in a new construct E. coli BL21/pACYCDuet-RM. Compared with E. coli BL21/pACYCDuet-RO, the molar conversion rate of L-2-ABA increased by 74.6%. The whole cell biotransformation conditions were optimized and the optimal pH, temperature and substrate concentration were 7.5, 35 °C and 80 g/L, respectively. Under these conditions, the molar conversion rate was higher than 99%. Finally, 80 g and 40 g L-threonine were consecutively fed into a 1 L reaction mixture under the optimal conversion conditions, producing 97.9 g L-2-ABA. Thus, this strategy provides a green and efficient synthesis of L-2-ABA, and has great industrial application potential.
Aminobutyrates ; Escherichia coli/genetics* ; Leucine Dehydrogenase/genetics* ; Threonine Dehydratase

5-aminolevulinic acid (5-ALA) plays an important role in the fields of medicine and agriculture. 5-ALA can be produced by engineered Escherichia coli and Corynebacterium glutamicum. We systematically engineered the C4 metabolic pathway of C. glutamicum to further improve its ability to produce 5-ALA. Firstly, the hemA gene encoding 5-ALA synthase (ALAS) from Rhodobacter capsulatus and Rhodopseudomonas palustris were heterologously expressed in C. glutamicum, respectively. The RphemA gene of R. palustris which showed relatively high enzyme activity was selected. Screening of the optimal ribosome binding site sequence RBS5 significantly increased the activity of RphemA. The ALAS activity of the recombinant strain reached (221.87±3.10) U/mg and 5-ALA production increased by 14.3%. Subsequently, knocking out genes encoding α-ketoglutarate dehydrogenase inhibitor protein (odhI) and succinate dehydrogenase (sdhA) increased the flux of succinyl CoA towards the production of 5-ALA. Moreover, inhibiting the expression of hemB by means of sRNA reduced the degradation of 5-ALA, while overexpressing the cysteine/O-acetylserine transporter eamA increased the output efficiency of intracellular 5-ALA. Shake flask fermentation using the engineered strain C. glutamicum 13032/∆odhI/∆sdhA-sRNAhemB- RBS5RphemA-eamA resulted in a yield of 11.90 g/L, which was 57% higher than that of the original strain. Fed-batch fermentation using the engineered strain in a 5 L fermenter produced 25.05 g/L of 5-ALA within 48 h, which is the highest reported-to-date yield of 5-ALA from glucose.
Aminolevulinic Acid/metabolism* ; Corynebacterium glutamicum/metabolism* ; Fermentation ; Metabolic Engineering ; Rhodobacter capsulatus/enzymology* ; Rhodopseudomonas/enzymology*