Rewiring of the Fruit Metabolome in Tomato Breeding.

Humans heavily rely on dozens of domesticated plant species that have been further improved through intensive breeding. To evaluate how breeding changed the tomato fruit metabolome, we have generated and analyzed a dataset encompassing genomes, transcriptomes, and metabolomes from hundreds of tomato genotypes. The combined results illustrate how breeding globally altered fruit metabolite content. Selection for alleles of genes associated with larger fruits altered metabolite profiles as a consequence of linkage with nearby genes. Selection of five major loci reduced the accumulation of anti-nutritional steroidal glycoalkaloids in ripened fruits, rendering the fruit more edible. Breeding for pink tomatoes modified the content of over 100 metabolites. The introgression of resistance genes from wild relatives in cultivars also resulted in major and unexpected metabolic changes. The study reveals a multi-omics view of the metabolic breeding history of tomato, as well as provides insights into metabolome-assisted breeding and plant biology.

AttentiveSkin: To Predict Skin Corrosion/Irritation Potentials of Chemicals via Explainable Machine Learning Methods.

Skin Corrosion/Irritation (Corr./Irrit.) has long been a health hazard in the Globally Harmonized System (GHS). Several in silico models have been built to predict Skin Corr./Irrit. as an alternative to the increasingly restricted animal testing. However, current studies are limited by data amount/quality and model availability. To address these issues, we compiled a traceable consensus GHS data set comprising 731 Corr., 1283 Irrit., and 1205 negative (Neg.) samples from 6 governmental databases and 2 external data sets. Then, a series of binary classifiers were developed with five machine learning (ML) algorithms and six molecular representations. For 10-fold cross-validation, the best Corr. vs Neg. classifier achieved an Area Under the Receiver Operating Characteristic Curve (AUC) of 97.1%, while the best Irrit. vs Neg. classifier achieved an AUC of 84.7%. Compared with existing in silico tools on external validation, our Attentive FP classifiers showed the highest metrics on Corr. vs Neg. and the second highest accuracy on Irrit. vs Neg. The SHapley Additive exPlanation approach was further applied to figure out important molecular features, and the attention weights were visualized to perform interpretable prediction. Structural alerts associated with Skin Corr./Irrit. were also identified. The interpretable Attentive FP classifiers were integrated into the software AttentiveSkin at https://github.com/BeeBeeWong/AttentiveSkin. The conventional ML classifiers are also provided on our platform admetSAR at http://lmmd.ecust.edu.cn/admetsar2/. Considering the data deficiency and the limited model availability of Skin Corr./Irrit., we believe that our data set and models could facilitate chemical safety assessment and relevant studies.

In Silico Prediction of Skin Sensitization for Compounds via Flexible Evidence Combination Based on Machine Learning and Dempster-Shafer Theory.

Skin sensitization is increasingly becoming a significant concern in the development of drugs and cosmetics due to consumer safety and occupational health problems. In silico methods have emerged as alternatives to traditional in vivo animal testing due to ethical and economic considerations. In this study, machine learning methods were used to build quantitative structure-activity relationship (QSAR) models on five skin sensitization data sets (GPMT, LLNA, DPRA, KeratinoSens, and h-CLAT), achieving effective predictive accuracies (correct classification rates of 0.688-0.764 on test sets). To address the complex mechanisms of human skin sensitization, the Dempster-Shafer theory was applied to merge multiple QSAR models, resulting in an evidence-based integrated decision model. Various evidence combinations and combination rules were explored, with the self-defined Q3 rule showing superior balance. The combination of evidence such as GPMT and KeratinoSens and h-CLAT achieved a correct classification rate (CCR) of 0.880 and coverage of 0.893 while maintaining the competitiveness of other combinations. Additionally, the Shapley additive explanations (SHAP) method was used to interpret important features and substructures related to skin sensitization. A comparative analysis of an external human test set demonstrated the superior performance of the proposed method. Finally, to enhance accessibility, the workflow was implemented into a user-friendly software named HSkinSensDS.

In Silico Prediction of Chemical Acute Dermal Toxicity Using Explainable Machine Learning Methods.

The research on acute dermal toxicity has consistently been a crucial component in assessing the potential risks of human exposure to active ingredients in pesticides and related plant protection products. However, it is difficult to directly identify the acute dermal toxicity of potential compounds through animal experiments alone. In our study, we separately integrated 1735 experimental data based on rabbits and 1679 experimental data based on rats to construct acute dermal toxicity prediction models using machine learning and deep learning algorithms. The best models for the two animal species achieved AUC values of 78.0 and 82.0%, respectively, on 10-fold cross-validation. Additionally, we employed SARpy to extract structural alerts, and in conjunction with Shapley additive explanation and attentive FP heatmap, we identified important features and structural fragments associated with acute dermal toxicity. This approach offers valuable insights for the detection of positive compounds. Moreover, a standalone software tool was developed to make acute dermal toxicity prediction easier. In summary, our research would provide an effective tool for acute dermal toxicity evaluation of pesticides, cosmetics, and drug safety assessment.

In silico prediction of ocular toxicity of compounds using explainable machine learning and deep learning approaches.

The accurate identification of chemicals with ocular toxicity is of paramount importance in health hazard assessment. In contemporary chemical toxicology, there is a growing emphasis on refining, reducing, and replacing animal testing in safety evaluations. Therefore, the development of robust computational tools is crucial for regulatory applications. The performance of predictive models is heavily reliant on the quality and quantity of data. In this investigation, we amalgamated the most extensive dataset (4901 compounds) sourced from governmental GHS-compliant databases and literature to develop binary classification models of chemical ocular toxicity. We employed 12 molecular representations in conjunction with six machine learning algorithms and two deep learning algorithms to create a series of binary classification models. The findings indicated that the deep learning method GCN outperformed the machine learning models in cross-validation, achieving an impressive AUC of 0.915. However, the top-performing machine learning model (RF-Descriptor) demonstrated excellent performance with an AUC of 0.869 on the test set and was therefore selected as the best model. To enhance model interpretability, we conducted the SHAP method and attention weights analysis. The two approaches offered visual depictions of the relevance of key descriptors and substructures in predicting ocular toxicity of chemicals. Thus, we successfully struck a delicate balance between data quality and model interpretability, rendering our model valuable for predicting and comprehending potential ocular-toxic compounds in the early stages of drug discovery.

Candidate Gene Identification and Transcriptome Analysis of Tomato male sterile-30 and Functional Marker Development for ms-30 and Its Alleles, ms-33, 7B-1, and stamenless-2.

Male sterility is a valuable trait for hybrid seed production in tomato (Solanum lycopersicum). The mutants male sterile-30 (ms-30) and ms-33 of tomato exhibit twisted stamens, exposed stigmas, and complete male sterility, thus holding potential for application in hybrid seed production. In this study, the ms-30 and ms-33 loci were fine-mapped to 53.3 kb and 111.2 kb intervals, respectively. Tomato PISTILLATA (TPI, syn. SlGLO2), a B-class MADS-box transcription factor gene, was identified as the most likely candidate gene for both loci. TPI is also the candidate gene of tomato male sterile mutant 7B-1 and sl-2. Allelism tests revealed that ms-30, ms-33, 7B-1, and sl-2 were allelic. Sequencing analysis showed sequence alterations in the TPI gene in all these mutants, with ms-30 exhibiting a transversion (G to T) that resulted in a missense mutation (S to I); ms-33 showing a transition (A to T) that led to alternative splicing, resulting in a loss of 46 amino acids in protein; and 7B-1 and sl-2 mutants showing the insertion of an approximately 4.8 kb retrotransposon. On the basis of these sequence alterations, a Kompetitive Allele Specific PCR marker, a sequencing marker, and an Insertion/Deletion marker were developed. Phenotypic analysis of the TPI gene-edited mutants and allelism tests indicated that the gene TPI is responsible for ms-30 and its alleles. Transcriptome analysis of ms-30 and quantitative RT-PCR revealed some differentially expressed genes associated with stamen and carpel development. These findings will aid in the marker-assisted selection for ms-30 and its alleles in tomato breeding and support the functional analysis of the TPI gene.

All-flesh fruit in tomato is controlled by reduced expression dosage of AFF through a structural variant mutation in the promoter.

The formation of locule gel is an important process in tomato and is a typical characteristic of berry fruit. In this study, we examined a natural tomato mutant that produces all-flesh fruit (AFF) in which the locule tissue remains in a solid state during fruit development. We constructed different genetic populations to fine-map the causal gene for this trait and identified SlMBP3 as the locus conferring the locule gel formation, which we rename as AFF. We determined the causal mutation as a 416-bp deletion in the promoter region of AFF, which reduces its expression dosage. Generally, this sequence is highly conserved among Solanaceae, as well as within the tomato germplasm. Using BC6 near-isogenic lines, we determined that the reduced expression dosage of AFF did not affect the normal development of seeds, whilst producing unique, non-liquefied locule tissue that was distinct from that of normal tomatoes in terms of metabolic components. Combined analysis using mRNA-seq and metabolomics indicated the importance of AFF in locule tissue liquefaction. Our findings provide insights into fruit-type differentiation in Solanaceae crops and also present the basis for future applications of AFF in tomato breeding programs.

Fine Mapping of the Ph-2 Gene Conferring Resistance to Late Blight (Phytophthora infestans) in Tomato.

Late blight is a devastating tomato disease. Breeding new varieties with multiple resistance (R) genes is highly effective for preventing late blight. The Ph-2 gene mediates resistance to Phytophthora infestans race T1 in tomato. In this study, we used an F2 population derived from a cross between Solanum lycopersicum Moboline (resistant) and LA3988 (susceptible) cultivars for the fine mapping of Ph-2. Two flanking markers, CAPS-1 and CC-Ase, mapped Ph-2 to a 141-kb genomic region containing 21 projected genes, 5 of which were identified as putative R genes. The Solyc10g085460 coding sequence varied significantly between the parents. The markers developed and candidate genes identified in this study shall be useful for the molecular breeding of tomato exhibiting increased late blight resistance and for the cloning of the Ph-2 gene.

Anthocyanin Fruit encodes an R2R3-MYB transcription factor, SlAN2-like, activating the transcription of SlMYBATV to fine-tune anthocyanin content in tomato fruit.

Anthocyanin fruit (Aft) and atroviolacea (atv) were characterized in wild tomato and can enhance anthocyanin content in tomato fruit. However, the gene underlying the Aft locus and the mechanism by which Aft and atv act remain largely unknown. In this study, the Aft locus was fine-mapped to an approximately 145-kb interval on chromosome 10, excluding SlAN2 (Solyc10g086250), SlANT1 (Solyc10g086260) and SlANT1-like (Solyc10g086270), which have previously been suggested as candidates. Thus, the R2R3-MYB transcription factor SlAN2-like (Solyc10g086290) was considered the best candidate gene for Aft. The CRISPR/Cas9-mediated SlAN2-like mutants show a much lower accumulation of anthocyanins associated with the downregulation of multiple anthocyanin-related genes compared to the wild-type tomato, indicating that SlAN2-like is responsible for the Aft phenotype. The repressive function of SlMYBATV also was confirmed through the CRISPR/Cas9 approach. A yeast-two-hybrid assay revealed that SlMYBATV interacts with the bHLH protein SlJAF13. Furthermore, yeast-one-hybrid and dual-luciferase transient expression assays showed that Aft directly binds to the SlMYBATV promoter and activates its expression. The results herein provide candidate genes to enhance anthocyanin content in tomato fruit. This research also provides insight into a mechanism involving the Aft-SlMYBATV pathway that fine-tunes anthocyanin accumulation in tomato fruit.

Fruit weight is controlled by Cell Size Regulator encoding a novel protein that is expressed in maturing tomato fruits.

Increases in fruit weight of cultivated vegetables and fruits accompanied the domestication of these crops. Here we report on the positional cloning of a quantitative trait locus (QTL) controlling fruit weight in tomato. The derived allele of Cell Size Regulator (CSR-D) increases fruit weight predominantly through enlargement of the pericarp areas. The expanded pericarp tissues result from increased mesocarp cell size and not from increased number of cell layers. The effect of CSR on fruit weight and cell size is found across different genetic backgrounds implying a consistent impact of the locus on the trait. In fruits, CSR expression is undetectable early in development from floral meristems to the rapid cell proliferation stage after anthesis. Expression is low but detectable in growing fruit tissues and in or around vascular bundles coinciding with the cell enlargement stage of the fruit maturation process. CSR encodes an uncharacterized protein whose clade has expanded in the Solanaceae family. The mutant allele is predicted to encode a shorter protein due to a 1.4 kb deletion resulting in a 194 amino-acid truncation. Co-expression analyses and GO term enrichment analyses suggest association of CSR with cell differentiation in fruit tissues and vascular bundles. The derived allele arose in Solanum lycopersicum var cerasiforme and appears completely fixed in many cultivated tomato's market classes. This finding suggests that the selection of this allele was critical to the full domestication of tomato from its intermediate ancestors.

Identification of Candidate HY5-Dependent and -Independent Regulators of Anthocyanin Biosynthesis in Tomato.

High quantities of anthocyanins in plants confer potential protective benefits against biotic and abiotic stressors. Studies have shown that the bZIP transcription factor HY5 plays a key role in controlling anthocyanin accumulation in response to light. However, in hy5 mutants, residual anthocyanins have been detected, indicating that other regulators exist to regulate anthocyanin biosynthesis in an HY5-independent manner. Here, we employed the CRISPR/Cas9 (clustered regularly interspersed short palindromic repeats/CRISPR-associated protein 9) system specifically to induce targeted mutagenesis of SlHY5 in the purple tomato cultivar 'Indigo Rose'. The T2 generation of tomato plants homozygous for the null allele of the SlHY5 frameshift mutated by a 1 bp insertion contained a lower anthocyanin content. Transcriptional analysis showed that most of the anthocyanin biosynthesis structural genes and several regulatory genes were down-regulated in the hy5 mutant lines. With transcriptome analyses of the various tissues from hy5 mutant lines, eight candidate transcription factors were identified that may regulate anthocyanin biosynthesis in an HY5-independent manner. These findings deepen our understanding of how light controls anthocyanin accumulation and facilitate the identification of the regulators of anthocyanin biosynthesis in an HY5-independent manner.

Fine mapping and molecular marker development of the Sm gene conferring resistance to gray leaf spot (Stemphylium spp.) in tomato.

KEY MESSAGE: The tomato gray leaf spot resistance gene Sm was fine-mapped in a 185-kb region through a map-based cloning strategy and genome-wide association study; a candidate gene was proved to be involved in Sm-mediated resistance through transient gene silencing. Gray leaf spot, caused by Stemphylium spp., is a warm weather foliar disease in tomato (Solanum lycopersicum L). Resistance against gray leaf spot is conferred by a single incompletely dominant gene (Sm) located on chromosome 11. This study aimed to map and identify molecular marker tightly linked to the Sm gene for the use of marker-assisted selection in breeding. Using an F2 population derived from a cross between the resistant line '9706' and the susceptible line 'Heinz 1706', the Sm gene was mapped to a 185-kb interval between two markers, InDel343 and InDel-FT-32 on chromosome 11, which was consistent with the result of a genome-wide association study using 289 diverse accessions. An ORF predicted in this region was proved to be involved in Sm-mediated resistance through transient gene silencing and seems to be a good candidate of the Sm locus. To clone the Sm gene, a bacterial artificial chromosome (BAC) library was screened and one BAC clone B80B15 containing the predicted ORF was identified. The analysis of sequence and structure characteristics demonstrated that the candidate gene was not a typical type resistance gene. Additionally, a co-dominant marker Sm-InDel, which produced a 122-bp or 140-bp fragment for resistant or susceptible alleles, respectively, was developed. This marker was validated in 289 germplasm and could be used in marker-assisted selection for gray leaf spot resistance.

B-class MADS-box TM6 is a candidate gene for tomato male sterile-1526.

KEY MESSAGE: Tomato male sterile-1526 locus was fine-mapped to an interval of 44.6 kb, and a B-class MADS-box gene TM6 was identified as the candidate gene. Male sterile lines have been widely used for hybrid seed production in many crop plants. The tomato male sterile-1526 (ms-1526) mutant displays abnormal stamens and exerted stigmas and is suitable for practical use. In this study, the ms-1526 locus was fine-mapped to a 44.6 kb interval that contained four putative genes. Thereinto, Solyc02g084630 encodes tomato B-class MADS-box gene TM6 (syn. TDR6), which plays an important role in stamen development. Sequencing revealed that there was a 12.7 kb deletion in the ms-1526 region, where the promoter and first four exons of the TM6 gene were absent. ms-1547, an allele of ms-1526, also contained the same deletion in the TM6 gene. And the other allele ms-15 mutant contained a single-nucleotide polymorphism (SNP, C to A) in the coding region of the TM6 gene, which led to a missense mutation (G to W). The codominant insertion/deletion (InDel) marker MS26D and codominant derived cleaved amplified polymorphic sequence (dCAPS) marker MS15C were developed based on the deletion and SNP, respectively. A real-time quantitative reverse-transcription PCR showed that expression of the TM6 gene was barely detectable in the flowers of the ms-1526 and ms-1547 mutants. In addition, other floral organ identity genes, pollen development marker genes, and pistil marker genes were differentially expressed between wild type and mutant flowers. These findings may facilitate functional analysis of the TM6 gene and help in the marker-assisted selection of ms-15 and its alleles in tomato breeding.

Identification and Characterization of EI (Elongated Internode) Gene in Tomato (Solanum lycopersicum).

Internode length is an important agronomic trait affecting plant architecture and crop yield. However, few genes for internode elongation have been identified in tomato. In this study, we characterized an elongated internode inbred line P502, which is a natural mutant of the tomato cultivar 05T606. The mutant P502 exhibits longer internode and higher bioactive GA concentration compared with wild-type 05T606. Genetic analysis suggested that the elongated internode trait is controlled by quantitative trait loci (QTL). Then, we identified a major QTL on chromosome 2 based on molecular markers and bulked segregant analysis (BSA). The locus was designated as EI (Elongated Internode), which explained 73.6% genetic variance. The EI was further mapped to a 75.8-kb region containing 10 genes in the reference Heinz 1706 genome. One single nucleotide polymorphism (SNP) in the coding region of solyc02g080120.1 was identified, which encodes gibberellin 2-beta-dioxygenase 7 (SlGA2ox7). SlGA2ox7, orthologous to AtGA2ox7 and AtGA2ox8, is involved in the regulation of GA degradation. Overexpression of the wild EI gene in mutant P502 caused a dwarf phenotype with a shortened internode. The difference of EI expression levels was not significant in the P502 and wild-type, but the expression levels of GA biosynthetic genes including CPS, KO, KAO, GA20ox1, GA20ox2, GA20ox4, GA3ox1, GA2ox1, GA2ox2, GA2ox4, and GA2ox5, were upregulated in mutant P502. Our results may provide a better understanding of the genetics underlying the internode elongation and valuable information to improve plant architecture of the tomato.

A putative R3 MYB repressor is the candidate gene underlying atroviolacium, a locus for anthocyanin pigmentation in tomato fruit.

Anthocyanins are potential health-promoting compounds in the human diet. The atv (atroviolacium) locus, derived from the wild tomato species Solanum cheesmaniae, has been shown to enhance anthocyanin pigmentation in tomato fruit when it co-exists with either the Aft (Anthocyanin fruit) or the Abg (Aubergine) locus. In the present study, the atv locus was fine-mapped to an approximately 5.0-kb interval on chromosome 7. A putative R3 MYB repressor was identified in this interval and is hereby designated as SlMYBATV. The allele of SlMYBATV underlying the atv locus harbored a 4-bp insertion in its coding region, which is predicted to result in a frame-shift and premature protein truncation. The other candidate R3 MYB and R2R3 MYB repressors of anthocyanin biosynthesis were also identified in tomato via a genome-wide search. Transcriptional analysis showed that most of the structural genes and several regulatory genes of anthocyanin biosynthesis were up-regulated in the tomato SlMYBATV mutant lines. These findings may facilitate the elucidation of the molecular mechanisms underlying anthocyanin pigmentation in tomato fruit and help in the marker-assisted selection of anthocyanin-enriched tomato cultivars.

Human preferences for sexually dimorphic faces may be evolutionarily novel.

A large literature proposes that preferences for exaggerated sex typicality in human faces (masculinity/femininity) reflect a long evolutionary history of sexual and social selection. This proposal implies that dimorphism was important to judgments of attractiveness and personality in ancestral environments. It is difficult to evaluate, however, because most available data come from large-scale, industrialized, urban populations. Here, we report the results for 12 populations with very diverse levels of economic development. Surprisingly, preferences for exaggerated sex-specific traits are only found in the novel, highly developed environments. Similarly, perceptions that masculine males look aggressive increase strongly with development and, specifically, urbanization. These data challenge the hypothesis that facial dimorphism was an important ancestral signal of heritable mate value. One possibility is that highly developed environments provide novel opportunities to discern relationships between facial traits and behavior by exposing individuals to large numbers of unfamiliar faces, revealing patterns too subtle to detect with smaller samples.

Preparation and characterization of monodisperse molecularly imprinted polymer microspheres by precipitation polymerization for kaempferol.

A new kind of molecularly imprinted polymer (MIP) microspheres for the selective extraction of kaempferol was prepared by precipitation polymerization using 4-vinylpridine (4-VP) and ethylene glycol dimethacrylate (EDMA) as functional monomer and cross-linker respectively. The synthesis conditions, such as ratios of 4-VP/EDMA and polymerization time were discussed in detail. Results showed that the 2% was the optimal concentration of co-monomers to obtain monodisperse MIP microspheres, the best ratio of 4-VP/EDMA was 1:2, and 24 h was considered as the proper polymerization time. Compared with the MIP agglomeration or coagulum particles, monodisperse MIP microspheres showed the better adsorption capacity: the saturated adsorption capacity of monodisperse MIP microspheres was 7.47 mg g-1, the adsorption equilibrium could be obtained in 30 min. Finally, the adsorption performances of the optimal MIP microspheres were evaluated by kinetic adsorption, adsorption isotherm, and selective adsorption experiments, which indicated that the adsorption mechanism were chemical single layer adsorption and the separation factor was up to 3.91 by comparing with the structure similar compound (quercetin). The MIP microspheres exhibit prospects in the kaempferol efficient and selective separation.

Rapid and reliable identification of tomato fruit weight and locule number loci by QTL-seq.

KEY MESSAGE: Bulk segregant analysis coupled with whole genome sequencing is a powerful approach and cost-effective method to identify loci controlling fruit traits in tomato. Domestication of fruit and vegetable crops was accompanied by selection for weight of the edible parts. Increases in fruit weight are controlled by multiple quantitative trait loci (QTL). To date, only two fruit weight genes have been cloned and a third has been fine-mapped. Genes that control locule number also impact fruit weight and two of them are known. To efficiently identify additional tomato fruit weight (FW) and locule number (LC) loci, six F2 populations were generated from crosses between closely related tomato accessions for which the alleles of the cloned FW and LC genes were known. We employed the bulk segregant approach coupled to whole genome sequencing (QTL-seq) which led to the identification of three highly significant and newly mapped FW QTL. fw11.2 was located in the distal part of chromosome 11 above the known loci fas and fw11.3; fw1.1 in the pericentromeric region of chromosome 1; and fw3.3 located ~1.6 Mb below the known fruit weight gene, SlKLUH/FW3.2. In addition, we mapped three LC QTL (lcn2.4, lcn5.1, and lcn6.1) although their significance was generally low. To confirm the location of the gene underlying fw11.2, we developed additional markers and conducted progeny tests. These results allowed us to narrow down the fw11.2 QTL to a region of ~750 kb corresponding to 66 candidate genes. Our research approach provided a cost-effective and time-efficient method for the identification of additional genes involved in FW and LC that could be used for both fruit development studies and crop improvement programs.

In Silico Prediction of ERRα Agonists Based on Combined Features and Stacking Ensemble Method.

Estrogen-related receptor α (ERRα) is considered a very promising target for treating metabolic diseases such as type 2 diabetes. Development of a prediction model to quickly identify potential ERRα agonists can significantly reduce the time spent on virtual screening. In this study, 298 ERRα agonists and numerous nonagonists were collected from various sources to build a new dataset of ERRα agonists. Then a total of 90 models were built using a combination of different algorithms, molecular characterization methods, and data sampling techniques. The consensus model with optimal performance was also validated on the test set (AUC=0.876, BA=0.816) and external validation set (AUC=0.867, BA=0.777) based on five selected baseline models. Furthermore, the model's applicability domain and privileged substructures were examined, and the feature importance was analyzed using the SHAP method to help interpret the model. Based on the above, it's hoped that our publicly accessible data, models, codes, and analytical techniques will prove valuable in quick screening and rational designing more novel and potent ERRα agonists.

Using a longitudinal network structure to subgroup depressive symptoms among adolescents.

BACKGROUND: Network modeling has been proposed as an effective approach to examine complex associations among antecedents, mediators and symptoms. This study aimed to investigate whether the severity of depressive symptoms affects the multivariate relationships among symptoms and mediating factors over a 2-year longitudinal follow-up. METHODS: We recruited a school-based cohort of 1480 primary and secondary school students over four semesters from January 2020 to December 2021. The participants (n = 1145) were assessed at four time points (ages 10-13 years old at baseline). Based on a cut-off score of 5 on the 9-item Patient Health Questionnaire at each time point, the participants were categorized into the non-depressive symptom (NDS) and depressive symptom (DS) groups. We conducted network analysis to investigate the symptom-to-symptom influences in these two groups over time. RESULTS: The global network metrics did not differ statistically between the NDS and DS groups at four time points. However, network connection strength varied with symptom severity. The edge weights between learning anxiety and social anxiety were prominently in the NDS group over time. The central factors for NDS and DS were oversensitivity and impulsivity (3 out of 4 time points), respectively. Moreover, both node strength and closeness were stable over time in both groups. CONCLUSIONS: Our study suggests that interrelationships among symptoms and contributing factors are generally stable in adolescents, but a higher severity of depressive symptoms may lead to increased stability in these relationships.