Metabolic Modelling Facilitates the Design of Synthetic Communities by Simplifying Natural Microbial Communities

ABSTRACT

Natural microbial communities generally have complex compositions and unclear metabolic interactions, posing constraints on their applications. Clarifying these intricate interactions within microbial communities is challenging for traditional experiment‐based methods. Here, we developed a simulation‐based approach to design synthetic communities (SynComs) by simplifying complex microbial communities through metabolic modelling. We constructed genome‐scale metabolic models (GSMMs) and curated them based on data obtained from straightforward experiments, ensuring these models precisely characterized metabolic features of each strain. By simulations utilizing multi‐strain metabolic models encompassing various strain combinations, we identified helper strains capable of enhancing the degradation efficiency of degrader strains and predicted optimal strain combinations that achieved a simplified community structure while maintaining high pollutant‐degrading efficiency. The simulations also unravelled cross‐feeding of glucosamine, amino acids and organic acids between the degrader and helper strains, which boosted the pollutant‐degrading efficiency of SynComs. Furthermore, helper strains rapidly degraded the toxicant intermediate, thereby alleviating its inhibitory effect on degrader strains. These predictions were further verified experimentally, demonstrating the accuracy and feasibility of metabolic model‐based simulations. Our study establishes a framework for designing simplified SynComs without sacrificing degradation efficiency and highlights the often‐underestimated role of microbial interactions in biodegradation.

A simulation‐based approach was developed to design synthetic microbial communities (SynComs) by simplifying complex microbial communities through metabolic modelling. The simulation‐based approach was used to construct a simplified SynCom for pesticide degradation and elucidate the intricate metabolic interactions among the members of the simplified SynCom, thereby providing theoretical and methodological foundations for SynCom construction.

1. Introduction

Currently, the global organic pollution problem is escalating, and the remediation of organic pollutants has emerged as a worldwide concern (Morin‐Crini et al. ²⁰²²; Ruan et al. ²⁰²³; Xu and Jiang ²⁰²⁴; Chen et al. ²⁰²⁵). In recent years, there has been a notable increase in the number of reported pure‐culture strains with the ability to degrade organic pollutants (Huang et al. 2017; Ruan et al. ²⁰²²; Zhu et al. ²⁰¹⁹; Hu et al. ²⁰²⁰). The discovery of these degrading strains has greatly enriched the strain resource library for organic pollution degradation. However, in natural environments, the complex molecular structures of organic pollutants make it difficult for a single strain to accomplish the degradation process independently. Moreover, the widespread presence of composite pollutants has an adverse effect on the survival and degradation efficiency of a single degrading strain, leading to suboptimal pollution degradation efficiency when using single strains (Xu and Jiang 2024; Lawson et al. ²⁰¹⁹). In contrast, microbial communities have many advantages. These include enhanced environmental adaptability, the capacity to achieve complete degradation of complex organic pollutants, improved degradation efficiency, and the potential to simultaneously degrade multiple pollutants (Chen et al. 2025; Dejonghe et al. ²⁰⁰³; Liu, Huang, et al. ²⁰¹⁹; Wang et al. ²⁰²³; Li, Chen, et al. ²⁰²⁵; Ma et al. ²⁰²⁵). Therefore, microbial communities play a crucial role and hold broad application prospects in organic pollution biodegradation.

The functions of microbial communities are largely determined by the interactions between strains (Kato et al. 2005; Cordero and Datta ²⁰¹⁶; Dal Co et al. ²⁰²⁰; Liu, Chen, et al. ²⁰¹⁹). For example, Kato et al. (2005) constructed a cellulose‐degrading microbial consortium consisting of five bacterial strains, including Clostridium straminisolvens CSK1, Clostridium sp. FG4, Pseudoxanthomonas sp. M1‐3, Brevibacillus sp. M1‐5 and Bordetella sp. M1‐6 (Kato et al. 2005). Removal of strain CSK1 from the consortium completely abolished the cellulose degradation capacity of the consortium. When strain FG4 instead of CSK1 was eliminated, the cellulose degradation efficiency was improved, accompanied by reduced ethanol accumulation and increased oligosaccharide content. The removal of M1‐3 exerted no remarkable effects on the degradation capability but increased acetate accumulation. These results demonstrate that synergistic interactions among community members are critical for efficient cellulose degradation and the final products. Therefore, designing and optimizing microbial communities often requires a comprehensive understanding of the complex metabolic interactions within the microbial communities. However, the metabolic interaction processes within microbial communities are complex and highly susceptible to alterations of community members and environments. Traditional experiment‐based methods are difficult to comprehensively clarify the complex metabolic interactions within microbial communities. For example, conventional research methods for exploring interactions in a microbial community require the combination of multiple experimental techniques to capture both exchanged and non‐exchanged metabolites and to detect the metabolic interactions among strains (Ponomarova and Patil 2015). Nevertheless, this multi‐technique experimental strategy demands a large amount of experimental data and is challenging to implement for the dynamic monitoring of metabolic interactions among strains. As a result, it is only applicable to some model systems and cannot be widely employed in the research of complex microbial communities (Embree et al. 2014; Großkopf and Soyer ²⁰¹⁴).

Metabolic modelling based on computer simulation is an emerging technology developed in recent years for investigating the interactions within microbial communities (Zomorrodi and Segrè 2016; Muller et al. ²⁰¹⁸; Xu et al. ^{2019, 2024}; Ruan et al. ²⁰²⁴). The genome‐scale metabolic model (GSMM), presenting the entire metabolic network relationship of a strain through mathematical methods, integrates information on all enzymes, biological reactions, reaction directions, and stoichiometric relationships of metabolites in the cell (Zomorrodi and Segrè 2016; Heirendt et al. ²⁰¹⁹). The simulation of microbial communities using the multi‐strain metabolic model (i.e., community model) is the latest model technology developed based on high‐throughput sequencing technology and GSMM. This microbial community model can not only present all the metabolic reactions of each strain but also track the metabolic exchanges between strains (Ruan et al. 2024; Henry et al. ²⁰¹⁶). Through microbial community modelling, it is feasible to analyse the metabolic exchanges and functional divisions among community members, clarify metabolic interactions within communities, and predict the growth and pollutant degradation efficiency, thus providing guidance for design of synthetic microbial communities (SynCom). In addition, compared with the traditional method of extensive experimental exploration, the metabolic modelling technology can significantly reduce the time and economic costs associated with SynCom design and exploring metabolic interactions within SynComs.

Rac‐Dichlorprop (the racemic form of dichlorprop, rac‐DCPP) is a commonly used phenoxycarboxylic acid herbicide (Ma et al. 2012; Hu, Zhao, et al. ²⁰²¹). After its application, the majority of DCPP migrates into the surrounding soil, while only a minor fraction is absorbed by plants. Due to its low volatility and difficulties in biodegradation and direct photolysis, DCPP easily accumulates in the environment (Gintautas et al. 1992; Malaguerra et al. ²⁰¹²). Moreover, the high water‐solubility of DCPP enables it to easily enter water bodies from the soil via rainwater or field irrigation (Feld et al. 2016; Paszko et al. ²⁰¹⁵; Zhang et al. ²⁰²³). In the past few decades, DCPP residues have been detected in the surface water and groundwater of numerous countries and regions, posing a significant threat to the ecological environment and human health (Paszko et al. 2015; Loos et al. ²⁰⁰⁹; Marriott et al. ²⁰⁰⁰; Hu, Liu, et al. ²⁰²¹; Zhu et al. ²⁰²⁰). Consequently, the residual levels in the environment and its removal have attracted considerable attention (Paszko et al. 2015).

In a previous study, we successfully enriched a highly efficient DCPP‐degrading microbial community from contaminated soils. Subsequently, we isolated five strains from the enrichment culture, namely Sphingopyxis sp. DBS4, Sphingopyxis sp. DCP‐4, Bosea sp. DCP‐2, Pigmentiphaga sp. DCP‐6d, and Achromobacter sp. DCP‐11d and evaluated their ability to degrade DCPP. Further investigations of these strains revealed that strain DBS4 can completely degrade rac‐DCPP, while strain DCP‐4 is only capable of degrading (R)‐DCPP. Strains DCP‐2, DCP‐6d, and DCP‐11d can all degrade the intermediate metabolite 2,4‐dichlorophenol (2,4‐DCP) of DCPP. Notably, the DCPP degradation efficiency of this enriched microbial community is significantly higher than that of the single strain DBS4. This suggests the existence of “helpers” and metabolic interactions between degraders and helpers within the community, which can boost the DCPP degradation efficiency. Therefore, elucidating the metabolic interactions within the DCPP‐degrading microbial community will improve our understanding of the mechanisms underlying the high‐efficient degradation of pollutants by microbial communities and facilitate the engineering of SynComs for biodegradation.

Currently, two principal strategies are employed for the design of SynComs: top‐down and bottom‐up (Lawson et al. 2019). The top‐down strategy starts from complex natural microbial communities. Then targeted selection pressures are applied to enrich specific microbial consortia with desired functions. In contrast, the bottom‐up strategy begins with individual strains. These strains with distinct functions are combined to construct a defined consortium that performs the intended biological functions under given conditions. Since communities derived from conventional top‐down methods are often highly complex with unclear interaction networks, their further application is limited. How to simplify these complex communities to construct SymComs without sacrificing their functions is essential for the application of microbial communities. Here, we provide a pipeline that employs the bottom‐up strategy to simplify the microbial communities initially obtained through top‐down methods.

In this study, we sequenced the genomes of strains isolated from the natural DCPP‐degrading communities. Subsequently, we constructed high‐quality single‐strain GSMMs and various community models covering different strain combinations for the strains involved in DCPP degradation (Figure 1). The objectives of this study are: (1) to analyse the synergistic metabolic relationships between strains through community modelling, deepening our understanding of the cooperative behaviour among microorganisms; (2) to explore the mechanism underlying the enhanced DCPP degradation efficiency of microbial communities compared to a single strain and (3) to establish an efficient approach for simplifying microbial communities without compromising their degradation efficiency, so as to design a highly efficient DCPP‐degrading SynCom. Our results provide new solutions for SynCom design by simplifying complex microbial communities without sacrificing pollutant‐degrading efficiency (Figure 1), which will facilitate the application of SynComs in the bioremediation of polluted environments.

FIGURE 1.

Workflow of metabolic modelling‐based SynCom construction by simplifying complex microbial communities. The workflow begins with functional communities obtained from the top‐down enrichment strategy. Each single‐strain GSMM is curated based on experimental results, and the high‐quality GSMMs are used to construct multi‐strain models with different strain combinations. Simulations are carried out to compare the performances of different strain combinations, and predict the optimal SynComs that simplify the community structure while maintaining functions similar to those of complex communities. Metabolic interactions within SynComs are also documented through simulations, which helps to explore the underlying mechanisms enhancing the functions of SynComs. All simulation‐based predictions can be easily verified by experimental results. C/N represents carbon or nitrogen sources; F_ex represents the flux of exchange reactions.

2. Material and Methods

2.1. Chemicals, Medium, Bacterial Strains, and Culture Conditions

Rac‐DCPP (CAS RN 120–36‐5, > 99% purity), (R)‐DCPP (CAS RN 15165–67‐0, > 99% purity) and 2,4‐DCP (CAS RN 120–83‐2, > 99% purity) were purchased from Aladdin Reagent (Shanghai, China) Co. Ltd. (S)‐DCPP was obtained by separation of the rac‐DCPP at Chiralway Biotech Co. Ltd. (Shanghai, China). All other chemicals and solvents were of analytical or HPLC grade. Mineral medium (MM: 1.0 g NH4Cl, 0.5 g KH2PO4, 1.5 g K2HPO4, 1.0 g NaCl, and 0.2 g MgSO₄·7 H₂O per litre of water, pH = 7.0) and Luria‐Bertani (LB) medium were used in this study. Five strains, including Sphingopyxis sp. DBS4, Sphingopyxis sp. DCP‐4, Bosea sp. DCP‐2, Pigmentiphaga sp. DCP‐6d and Achromobacter sp. DCP‐11d were used in this study. These strains were cultured at 30°C in the LB broth.

2.2. HPLC Analysis of Rac‐DCPP

Rac‐DCPP was extracted using an equal volume of dichloromethane. The resulting extract was evaporated and re‐dissolved in methanol. Subsequently, the solution was filtered through a 0.22‐mm‐pore Millipore membrane. The rac‐DCPP in the filtered solution was analysed using an HPLC system (UltiMate 3000 Titanium) equipped with a C₁₈ reversed‐phase column (5 μm, 250 × 4.6 mm). A mixture of 80% methanol (methanol:water, v/v) and 1.0% glacial acetic acid was used as the mobile phase. The flow rate was 1.0 mL/min, and the column temperature was maintained at 30°C. The injection volume was 20 μL, and rac‐DCPP was detected at a wavelength of 230 nm.

2.3. Genome Sequencing

Whole‐genome sequencing of four strains (DCP‐2, DCP‐4, DCP‐6d and DCP‐11d) was carried out at Shanghai Majorbio Bio‐pharm Technology Co. Ltd. (Shanghai, China) using both Illumina MiSeq and PacBio platforms. The sequencing procedure, read assembly, de novo gene prediction, and functional annotation were conducted following previously described methods (Zhang et al. 2023). Canu (v2.3) software was used to perform the initial assembly of the long‐read data generated by PacBio sequencing (Koren et al. 2017). The short‐read data obtained from Illumina sequencing were then further aligned with the contigs from the initial assembly. Pilon (v1.24) software was employed to refine the sequences (Walker et al. 2014), and LRScaf was used to connect the contigs into longer scaffolds (Qin et al. 2019). De novo gene prediction was performed on the assembled genome using Prokka (v1.13) software (Seemann 2014). The predicted genes were compared against the NR, Swiss‐Prot, COG, KEGG, and GO databases using BLAST (BLAST +2.14.1) for functional annotation. The GenBank accession numbers for the complete genome sequences of strains DCP‐2, DCP‐4, DCP‐6d and DCP‐11d are SUB15680332, SUB15680937, SUB15681044 and SUB15681330, respectively. The genome sequence of strain DBS4 was retrieved from the GenBank under accession numbers CP102384–CP102388.

2.4. Reconstruction of Metabolic Models for Single Strains

The draft metabolic models of the five strains were initially reconstructed independently by using the Model SEED to analyse the annotated genome sequences (Xu et al. 2019; Henry et al. ²⁰¹⁰; Seaver et al. ²⁰²¹). Subsequently, the growth of each strain in MM medium supplemented with various carbon sources was experimentally tested. The obtained experimental results were then used for model curation. The curation of the draft models was carried out in MATLAB using the COBRAToolbox—3.0 (Heirendt et al. 2019). Based on the experimental findings, if a strain could grow under a specific carbon source while the corresponding draft model failed to simulate this growth, potential missing reactions were identified and incorporated into the draft model. The FastGapFill function was utilized to detect and fill network gaps in each model. These models were further refined manually with reference to public databases, including KEGG (Kanehisa et al. 2016), UniProt (The UniProt Consortium ²⁰¹⁹), BiGG (Norsigian et al. 2020), IMG (Chen et al. 2019) and MetaCyc (Caspi et al. 2020). Invalid or incorrect energy‐generating loops were removed. To ensure consistency, all reaction IDs from different databases were standardized. Elementally imbalanced reactions based on chemical formulae were carefully checked and balanced. The draft models were iteratively calibrated to ensure that the reconstructed models could synthesize all biomass components in MM medium with alternative carbon sources. The models of the five strains used for simulations are provided in Data S1.

2.5. Construction of Multi‐Compartment Models for Bacterial Communities

To model bacterial communities, we integrated the five single‐species models into different multi‐compartment models. To simulate the bacterial cells where reactions take place, each single model was considered as a distinct compartment. An extracellular environment shared among all strains was simulated by establishing a common community compartment. For each single model, transport and exchange reactions were added to enable the absorption or secretion of metabolites from the extracellular environment. Thus, within the community compartment, the exchange of metabolites among species was permitted, mimicking the metabolic interactions in the co‐culture medium. The bacterial community was constructed using the “createMultipleSpeciesModel” function in COBRAToolbox—3.0. The biomass of the community was defined as the cumulative biomass of all individual strains. In the simulations, biomass was employed as the optimization objective. Although biomass does not directly equate to degradation activity, it frequently exhibits a positive correlation with degradation performance, thereby serving as a reasonable proxy for estimating potential degradation capacity in bioremediation systems. As DCPP was supplied as the sole carbon source, all accumulated biomass originated from DCPP conversion. Therefore, a higher biomass yield directly reflects a greater DCPP transformation, supporting the use of biomass as an indicator of degradation efficiency. Importantly, the validity of this approach was subsequently confirmed through experimental validation. The initial inoculation ratio was set to 1:1 in model simulations which is a commonly applied strategy for simulation (Ruan et al. 2024; Han et al. ²⁰²⁴). The optimal strain combinations were determined using the community objective function in flux balance analysis (FBA). Consistently, the same 1:1 inoculation ratio was applied in experimental validation, and other ratios were not tested in this study. The specific combinations within these communities are presented in Table 2. The community models are available in Data S2.

TABLE 2.

Statistics for the multi‐strain models.

No.	Multi‐strain model	Total strains	Total reactions	Total metabolites
②	DBS4&DCP‐4	2	2967	3028
③	DBS4&DCP‐2	2	3137	3157
④	DBS4&DCP‐6d	2	3098	3198
⑤	DBS4&DCP‐11d	2	3119	3187
⑥	DBS4&DCP‐4&DCP‐2	3	4597	4652
⑦	DBS4&DCP‐4&DCP‐6d	3	4558	4693
⑧	DBS4&DCP‐4&DCP‐11d	3	4579	4682
⑨	DBS4&DCP‐2&DCP‐6d	3	4707	4801
➉	DBS4&DCP‐2&DCP‐11d	3	4710	4772
⑪	DBS4&DCP‐6d&DCP‐11d	3	4695	4837
⑫	DBS4&DCP‐4&DCP‐2&DCP‐6d	4	6166	6295
⑬	DBS4&DCP‐4&DCP‐2&DCP‐11d	4	6169	6266
⑭	DBS4&DCP‐2&DCP‐6d&DCP‐11d	4	6277	6413
⑮	DBS4&DCP‐4&DCP‐6d&DCP‐11d	4	6155	6332
⑯	DBS4&DCP‐4&DCP‐2&DCP‐6d&DCP‐11d	5	7736	7907

Note: “①” refers to a single‐strain model DBS4.

2.6. Metabolic Model‐Based Simulations of Growth and DCPP Degradation

The fluxes were calculated using the parsimonious FBA method (pFBA). This method optimizes the biomass function while minimizing the flux of each nutrient exchange reaction through the model (Ruan et al. 2024; Xu et al. ²⁰²⁴). The investigations were conducted in MM medium where DCPP was the sole carbon source. The biomass of each combination of strains (different communities) was compared to identify the best‐predicted consortium. The absorption or secretion of metabolites was revealed by the negative and positive fluxes of exchange reactions, respectively. Therefore, exchanged compounds between two strains were identified as metabolites with a negative flux of the exchange reaction in one strain and a positive flux of the exchange reaction in the other strain.

Dynamic modelling was performed by the dynamic FBA (dFBA) algorithm to simulate the temporal growth of the bacterial community and the dynamic degradation processes of DCPP by the optimal degrading bacterial community (Xu et al. 2019; Popp and Centler ²⁰²⁰). The investigations were conducted under two specific experimental conditions: one in a MM medium where DCPP was the sole carbon source, and the other in a MM medium where the carbon sources were DCPP and glucose at a ratio of 1:1.

2.7. In Vitro Experiments Testing Predictions by Modelling

To experimentally validate the computational predictions, the degradation of DCPP by different microbial consortia was measured both in MM medium and soil environments. The degradation levels of DCPP by all simulated consortia in MM medium were determined. Each strain in corresponding SynComs was prepared with a concentration of 1.5 × 10⁶ CFU/mL and inoculated into 20 mL of MM medium containing 30 mg/L of DCPP. An un‐inoculated medium served as the control. The cultures were then incubated on a rotary shaker at a temperature of 30°C and a rotational speed of 180 rpm. The DCPP levels in the medium were measured every 3 h using HPLC following the method described above.

Soil samples containing 5 mg/kg of DCPP were employed for the DCPP degradation test. The single‐strain DBS4, as well as the optimal two‐ (DBS4&DCP‐4), three‐ (DBS4&DCP‐4&DCP‐11d), four‐ (DBS4&DCP‐4&DCP‐2&DCP‐11d) and five‐strain (DBS4&DCP‐4&DCP‐2&DCP‐11&DCP‐6d) consortia were inoculated into the soil, with each strain reaching a final concentration of approximately 10⁶ CFU/g of soil. Soils without any strain inoculation were used as controls. To extract DCPP from the soil, 5 g of soil samples from different treatments were collected into 50 mL centrifuge tubes. Then, 10 mL of an extraction solution (methanol/water/acetic acid at a ratio of 49:49:2) was added to each tube. After shaking at 30°C and 180 rpm for 1 h, the supernatant was obtained by centrifuging at 6000 rpm for 10 min. Dichloromethane was used to further extract DCPP from the solution. The extract was evaporated, re‐dissolved in methanol, and then filtered through a 0.22‐mm‐pore Millipore membrane. The DCPP level was detected by HPLC as described above.

To experimentally verify the predicted metabolic interactions among the strains of the best‐predicted consortium, the optimal consortium (DBS4&DCP‐4&DCP‐11d) was used to detect the secreted metabolites of the strains. The strains DBS4, DCP‐4, and DCP‐11d, each at a concentration of 1.5 × 10⁶ CFU/mL, were co‐cultured in MM medium containing 30 mg/L of DCPP as the sole carbon and nitrogen source at 30°C and 180 rpm for 15 h. The secretions of the DBS4&DCP‐4&DCP‐11d consortium were collected and screened by LC–MS. Pure compounds of these predicted exchanged metabolites were used as reference standards. The LC–MS process was carried out according to the method described by (Ruan et al. 2024).

3. Results

3.1. Whole‐Genome Sequencing and Analysis of DCPP Metabolic Pathway

In addition to strain DBS4 (Zhang et al. 2023), we successfully isolated four strains from the enrichment culture, namely Sphingopyxis sp. DCP‐4, Bosea sp. DCP‐2, Pigmentiphaga sp. DCP‐6d and Achromobacter sp. DCP‐11d. Substrate spectrum determination demonstrated that strain DBS4 was capable of degrading rac‐DCPP (both the (R)‐ and (S)‐enantiomers of DCPP), whereas strain DCP‐4 only degraded the (R)‐DCPP. Strains DCP‐2, DCP‐6d and DCP‐11d were all able to degrade 2,4‐DCP, an intermediate metabolite of DCPP (Figure 2 and Table S1).

FIGURE 2.

Microbial community involved in rac‐DCPP degradation. (A) The degradation pathway of rac‐DCPP in the microbial community composed of Sphingopyxis sp. DBS4, Sphingopyxis sp. DCP‐4, Bosea sp. DCP‐2, Pigmentiphaga sp. DCP‐6d, and Achromobacter sp. DCP‐11d. The key enzymes catalysing the corresponding reactions are indicated by the coloured arrows. (B) Degradation of rac‐DCPP by the single strain DBS4 (a rac‐dichlorprop degrader) and the microbial community. The strains in the microbial community are DBS4, DCP‐4, DCP‐2, DCP‐6d and DCP‐11d.

The genome of strain DBS4 had been previously sequenced, which consisted of one chromosome (4,357,663 bp) and four plasmids (200,350, 227,036, 36,803 and 84,199 bp, respectively) (Zhang et al. 2023). In this study, we sequenced the genomes of the other four strains (Table S2). The complete genome of strain DCP‐2 contained two replicons, namely one chromosome (5,710,710 bp) and one plasmid (174,221 bp). The complete genome of strain DCP‐4 was composed of one chromosome (4,355,352 bp) and three plasmids (295,072, 200,257 and 72,970 bp, respectively). The whole genomes of strains DCP‐6d and DCP‐11d each had one chromosome (6,356,200 bp for DCP‐6d and 6,602,652 bp for DCP‐11d) without any plasmids. The whole genomes of strains DCP‐2, DCP‐4, DCP‐6d, and DCP‐11d were predicted to have 5659, 4678, 5850, and 6193 genes, respectively (Table S2).

Our previous study indicated that an α‐ketoglutarate‐dependent dioxygenase, SpoA, encoded in the DBS4 genome, converted (S)‐DCPP into 2,4‐DCP and pyruvate (Zhang et al. 2023). Moreover, two Fe²⁺/α‐ketoglutarate‐dependent dioxygenases, RdpA and SdpA, were identified in both Sphingobium herbicidovorans MH and Delftia acidovorans MC1 (Nielsen et al. 2017). SpoA shared only 34.8% amino acid sequence identity with SdpA from MH and MC1, yet both enzymes convert (S)‐DCPP to 2,4‐DCP and pyruvate, indicating functional conservation despite low homology (Zhang et al. 2023; Nielsen et al. ²⁰¹⁷). In contrast, DCP‐4 harboured the RdpA gene, which encodes an enzyme responsible for converting (R)‐DCPP to 2,4‐DCP and pyruvate. Subsequently, 2,4‐DCP was further catabolized via a classic pathway catalysed by the genes tfdBCDEF (Zhang et al. 2023). We explored the homologous genes of the above‐mentioned DCPP‐degrading genes in the genomes of strains DCP‐2, DCP‐4, DCP‐6d and DCP‐11d, respectively (Figure 2). The results showed that all the strains possessed the tfdBCDEF genes, while only strain DCP‐4 had the RdpA gene (Figure 2A). Notably, the efficiency of DCPP degradation by the five‐strain microbial community was significantly higher than that of the single strain DBS4 (Figure 2B), suggesting that there were metabolic interactions among the community members that could enhance the DCPP degradation efficiency.

3.2. Construction of Metabolic Models for Single Strains and Microbial Communities

Based on the sequenced genomes, the raw GSMMs of the five strains, namely DBS4, DCP‐2, DCP‐4, DCP‐6d and DCP‐11d, were constructed through gene‐protein‐reaction (GPR) associations (Table 1). The raw GSMMs obtained from the initial reconstruction for strains DBS4, DCP‐2, DCP‐4, DCP‐6d and DCP‐11d contained 1277, 1385, 1302, 1377 and 1472 reactions, respectively, along with 1416, 1522, 1444, 1592 and 1568 metabolites (Table 1). Generally, the initial GSMMs require revision because some metabolic reactions may be missed due to imperfect genome annotation.

TABLE 1.

Overall genome and GSMM statistics for individual strains.

Mode	DBS4	DCP‐4	DCP‐2	DCP‐6d	DCP‐11d
Genome size (Mb)	4.91	4.92	5.88	6.6	6.36
Gene number	4851	4678	5662	6193	5850
Initial reactions	1277	1302	1385	1377	1375
Gap‐filling reactions	169	152	180	180	186
Total reactions	1446	1454	1565	1557	1561
Biochemical reacions	1321	1328	1370	1389	1352
Transport reactions	66	66	105	89	112
Exchange reactions	59	60	90	79	97
Total metabolites	1475	1494	1569	1636	1604

To identify potentially missing reactions, we investigated the growth of each strain in minimal medium (MM) with NH₄ ⁺ as the nitrogen source (MM‐NH₄ ⁺ medium). Strain DBS4 was capable of growing in the MM‐NH₄ ⁺ medium supplemented with glucose, D‐xylose, or maltose as the carbon source. Strain DCP‐2 could grow in MM‐NH₄ ⁺ medium when provided with glucose, galactose, trisodium citrate, D‐xylose, or succinic acid as the carbon source. Similarly, strain DCP‐4 could grow in the MM‐NH4⁺ medium with glucose, succinic acid, mannitol, or trisodium citrate as the carbon source; strain DCP‐6d could utilize glucose, galactose, mannitol, trisodium citrate, or maltose; and strain DCP‐11d could grow with glucose or trisodium citrate. Evidently, all five strains were able to grow in MM supplemented with NH₄ ⁺ as the nitrogen source and glucose as the carbon source (MM‐G‐NH₄ ⁺ medium). These experimental results were employed for identifying missing reactions and curating the raw models.

We performed growth simulations for the five strains individually using the MM‐G‐NH₄ ⁺ medium. None of the raw GSMMs could generate a biomass flux, indicating that these raw models were unable to synthesize one or more essential biomass components under the specified MM‐G‐NH₄ ⁺ medium conditions. Conversely, through wet‐lab experiments, we demonstrated that all five strains were indeed capable of growth in the MM‐G‐NH₄ ⁺ medium (Figure S1). To address the discrepancy between the model predictions and experimental observations, we systematically identified the missing metabolic reactions. Subsequently, these reactions were manually added into the raw models. This iterative process continued until the models produced the biomass of the five strains in the MM‐G‐NH₄ ⁺ medium. Specifically, we added 169, 180, 152, 180 and 186 missing reactions to the initial raw models of strains DBS4, DCP‐2, DCP‐4, DCP‐6d and DCP‐11d, respectively (Table 1). The final models were able to synthesize all necessary biomass components in the MM‐G‐NH₄ ⁺ medium, which was consistent with the experimental results.

To simulate the growth performance and DCPP‐degrading efficiency of different SynComs composed of various strains, we constructed a series of strain combinations covering all possible combinations. As the only strain capable of degrading rac‐DCPP, strain DBS4 was included in SynComs as a functional strain in all simulated combinations, while the other strains served as helper strains. In total, 15 SynCom models were constructed, including 4 two‐strain models, 6 three‐strain models, 4 four‐strain models, and 1 five‐strain model (Figure 3A and Table 2).

FIGURE 3.

Simulations of DCPP degradation performances by SynComs and metabolic interactions. (A) Schematic representation of simulations based on multi‐strain (community) models. A set of multi‐strain models with various strain combinations is constructed. For each multi‐strain model, each single‐strain model is considered as a separate compartment, and a common extracellular compartment is added to enable metabolite exchange among strains. F_ex represents the flux of exchange reactions. F_tr ^a and F_tr ^b represent the flux of transport reactions of strain A and B, respectively. For simplicity, only the input and output reactions are shown. (B) Predicted biomass of different strain combinations. The grey and white cells in the grids indicate whether a strain is included or not in the combination, respectively. The bars on the right side of the grids represent the predicted biomass. (C) Predicted metabolic interactions among strains DBS4, DCP‐4 and DCP‐11d by community modelling. (D) Simulations of glucose‐enhanced DCPP degradation by the SynCom DBS4&DCP‐4&DCP‐11d. Simulations were conducted in two media: One containing DCPP as the sole carbon source (DCPP) and the other with DCPP and glucose as carbon sources (DCPP+glucose).

3.3. Simulation‐Based SynCom Design and Metabolic Interaction Analysis

The simulations were conducted in MM medium containing DCPP and NH₄ ⁺ as the carbon and nitrogen sources (MM‐DCPP‐NH₄ ⁺ medium). Strain DBS4 was used as the control, which showed the lowest biomass production of 10.40 (Figure 3B). All four two‐strain SynComs showed enhanced biomass production compared to the control. Among them, the combination of DBS4 and DCP‐4 yielded the highest biomass production of 11.06, highlighting the crucial role of strain DCP‐4 in DCPP degradation within the SynCom (Figure 3B). Consistently, among the six three‐strain SynComs, the combination of DBS4, DCP‐4 and DCP‐11d (DBS4&DCP‐4&DCP‐11d) showed the maximum biomass production of 11.43 (Figure 3B). Moreover, all four‐strain and five‐strain SynComs that included DBS4, DCP‐4 and DCP‐11d displayed biomass levels comparable to those of the optimal three‐strain SynCom DBS4&DCP‐4&DCP‐11d (Figure 3B). Collectively, these results suggested that strains DCP‐4 and DCP‐11d could enhance the growth and DCPP degradation of strain DBS4, and the SynCom DBS4&DCP‐4&DCP‐11d exhibited the best performance in the MM‐DCPP‐NH₄ ⁺ medium (Figure 3B). Notably, the similar performance of the optimal three‐strain SynCom to the optimal four‐strain (DBS4&DCP‐4&DCP‐6d&DCP‐11d) and five‐strain SynComs (DBS4&DCP‐2&DCP‐4&DCP‐6d&DCP‐11d) implies the potential to simplify complex microbial communities without compromising their functionality (Figure 3B).

To elucidate the underlying mechanisms by which the SynCom DBS4&DCP‐4&DCP‐11d enhanced DCPP degradation and improved bacterial growth, we employed simulations to explore the metabolic interactions among strains DBS4, DCP‐4 and DCP‐11d in the MM‐DCPP‐NH₄ ⁺ medium (Figure 3C). By community modelling, we calculated the reaction fluxes in all strains, thereby determining the exchanged metabolites and their flow directions. A total of 7 predicted exchanged metabolites within the DBS4&DCP‐4&DCP‐11d consortium during DCPP degradation were selected, including D‐glucosamine, α‐ketoglutaric acid, L‐phenylalanine, N‐acetyl‐D‐glucosam, L‐lysine, stearic acid and 2,4‐DCP. Strain DBS4 took up both (S)‐ and (R)‐DCPP, degraded them, and subsequently secreted 2,4‐DCP, which was utilized by strain DCP‐11d (Figure 3C). Additionally, strain DBS4 secreted N‐acetyl‐D‐glucosam and nicotinamide‐ribonucleotide for strain DCP‐4 to consume (Figure 3C). Strain DCP‐4 absorbed and metabolized (R)‐DCPP and secreted α‐ketoglutaric acid, D‐glucosamine and L‐phenylalanine, which were then utilized by strain DBS4 (Figure 3C). After taking up and metabolizing the substances secreted by strain DBS4, strain DCP‐11d released L‐lysine and stearic acid for strain DBS4 to use. The L‐lysine secreted by strain DCP‐11d was also utilized by strain DCP‐4 (Figure 3C). These simulation results indicated the existence of complex cross‐feeding relationships among strains DBS4, DCP‐4 and DCP‐11d, and this synergistic mutualism enables them to efficiently degrade DCPP.

Bioaugmentation (inoculation of strains into polluted environments) and biostimulation (adding nutrients to polluted environments) are two fundamental strategies for bioremediation. To explore the feasibility of biostimulation, that is, whether the addition of an extra carbon source could further enhance the efficiency of DCPP degradation by the SynCom, we utilized the dynamic flux balance analysis (dFBA) algorithm to simulate the dynamic degradation process of the optimal three‐strain SynCom DBS4&DCP‐4&DCP‐11d in the MM‐DCPP‐NH₄ ⁺ medium supplemented with glucose (MM‐DCPP‐G‐ NH₄ ⁺ medium, with a glucose:DCPP ratio of 1:1). The results revealed that the addition of glucose accelerated the rate of DCPP degradation by the SynCom DBS4&DCP‐4&DCP‐11d (Figure 3D).

3.4. Experimental Validation of Metabolic Model‐Based Simulations

We validated the metabolic model‐based predictions through both shake‐flask and pot experiments. In the MM‐DCPP‐NH₄ ⁺ medium, the degradation rates of the two‐strain SynComs were generally higher than that of the single strain DBS4 (Figure 4A). This indicates that the addition of ‘helper strains’ promoted the degradation of DCPP by strain DBS4. Among all the two‐strain SynComs, DBS4&DCP‐4 exhibited the highest degradation rate, suggesting that strain DCP‐4 was the key strain for enhancing DCPP degradation, which was consistent with the model prediction results (Figure 4A). For the three‐strain SynComs, the degradation rates of SynComs containing strain DCP‐4 were better than those without strain DCP‐4 (Figure 4B). This also implied that strain DCP‐4 was a crucial strain for efficient DCPP degradation, which aligned with the findings from the two‐strain SynComs. Additionally, the degradation rate of the SynCom DBS4&DCP‐4&DCP‐11d was higher than that of DBS4&DCP‐4&DCP‐2 and DBS4&DCP‐4&DCP‐6d, consistent with the model predictions (Figure 4B). Among the four‐ and five‐strain SynComs, the five‐strain SynComs showed the best degradation rate (Figure 4C). Notably, since the SynCom DBS4&DCP‐2&DCP‐6d&DCP‐11d did not contain strain DCP‐4, its degradation rate was the lowest, further highlighting the importance of strain DCP‐4 (Figure 4C). It was worth emphasizing that the degradation efficiency of the optimal three‐strain SynCom DBS4&DCP‐4&DCP‐11d was comparable to that of the four‐ or five‐strain SynCom, which was in line with the model simulation results (Figure 4D). This result suggested that similar degradation efficiency to that of complex microbial communities could be achieved by simplifying the microbial community composition.

FIGURE 4.

Experimental validation of simulations. The DCPP‐degrading efficiency of different strain combinations was tested in both medium (A–D) and soils (E and F) supplemented with DCPP. To facilitate the comparison of different SynComs' performances, the two‐ (A) three‐ (B) and four‐ and five‐strain (C) SynComs were compared separately. (D) Comparison of the performances of a single strain and the optimal two‐, three‐, four‐ and five‐strain SynComs. (E) DCPP‐degrading efficiency of a single strain and the optimal two‐, three‐, four‐ and five‐strain SynComs in soils. (F) DCPP‐degrading efficiency of the SynCom DBS4&DCP‐4&DCP‐11d in soils supplemented with DCPP or DCPP and glucose (DCPP + glucose).

To explore the DCPP degradation efficiency of different SynComs in soils, we inoculated strain DSB4, the optimal two‐ (DBS4&DCP‐4), three‐ (DBS4&DCP‐4&DCP‐11d), four‐ (DBS4&DCP‐4&DCP‐2&DCP‐11d) and five‐strain (DBS4&DCP‐4&DCP‐2&DCP‐11&DCP‐6d) SynComs into DCPP‐contaminated soils and measured the residual amount of DCPP (Figure 4E). The soils without any strain inoculation served as controls. The results showed that the DCPP degradation efficiency of the multi‐strain SynComs was significantly higher than that of the single strain. Initially, the DCPP degradation efficiency of the three‐strain SynCom was lower than that of the four‐ and five‐strain SynComs (Figure 4E). However, 7 days after inoculation, there was no significant difference in the DCPP residual amount among the soils inoculated with three‐, four‐, and five‐strain SynComs (Figure 4E). These results indicate that the optimal three‐strain SynCom could achieve a DCPP degradation efficiency similar to that of the four‐ or five‐strain SynCom, which was consistent with the simulation and shake‐flask experiment results.

The simulation indicated that the addition of glucose could further enhance the DCPP degradation efficiency of the optimal SynCom DBS4&DCP‐4&DCP‐11d, which was experimentally verified. In the shake‐flask experiment, the degradation rate of DBS4&DCP‐4&DCP‐11d in the MM‐DCPP‐G‐NH₄ ⁺ medium was significantly higher than that in the medium without added glucose (MM‐DCPP‐ NH₄ ⁺ medium, Figure S2). This result was also confirmed by the soil experiment (Figure 4F).

To validate the metabolic interactions within the SynCom DBS4&DCP‐4&DCP‐11d during the DCPP‐degradation process predicted by the model simulation, we co‐cultured strains DBS4, DCP‐4, and DCP‐11d in the MM‐DCPP‐NH₄ ⁺ medium. Subsequently, we detected the metabolites in the culture solution using LC–MS. Given that the MM‐DCPP‐NH₄ ⁺ medium initially did not contain any of the predicted metabolites exchanged among strains (D‐glucosamine, α‐ketoglutaric acid, L‐phenylalanine, N‐acetyl‐D‐glucosam, L‐lysine, stearic acid and 2,4‐DCP), the detection of these substances in the culture solution could confirm their involvement in metabolic interactions. Nicotinamide‐ribonucleotide was not detected due to its instability. Except for D‐glucosamine, the other six substances were successfully detected by LC–MS (Figure S3 and Table S3). The non‐detection of D‐glucosamine might be attributed to its high utilization efficiency, which did not allow it to accumulate in the medium.

The simulation predicted that during the degradation of DCPP by DBS4&DCP‐4&DCP‐11d, the DCPP‐degrading strain DBS4 utilized D‐glucosamine, α‐ketoglutaric acid, L‐phenylalanine, and L‐lysine provided by other strains, thereby enhancing the efficiency of DCPP degradation. To validate this prediction, we compared the DCPP degradation efficiency of DBS4 when the above four exchanged metabolites were added to the cultures. The cultures without any addition and those with 2,4‐DCP (the intermediate of DCPP degradation) were used as controls. The results demonstrated that the addition of the four exchanged substances significantly improved the DCPP degradation efficiency of DBS4, while the addition of 2,4‐DCP significantly inhibited the degradation efficiency of DBS4 (Figure S4). These results also suggest that the SynCom can alleviate the inhibitory effect of 2,4‐DCP on the degrading strain DBS4 by rapidly degrading 2,4‐DCP.

4. Discussion

The ‘top‐down’ strategy has been widely used to engineer microbial communities, aiming to optimize these communities so that they can perform the desired functions (Lawson et al. 2019; Li, Liu, et al. ²⁰²⁵; Jia et al. ²⁰²¹). At present, enrichment culture is the primary ‘top‐down’ approach for acquiring highly efficient SynComs capable of degrading organic pollutants. However, microbial communities obtained via this method typically possess complex compositions and the metabolic interaction relationships within them remain unclear. These drawbacks impose limitations on their practical applications. Microbial community metabolic models can effectively elucidate the complex metabolic interactions within a microbial consortium and analyse the division of labor among member strains. Therefore, this approach enables the identification of key functional strains and helper strains, facilitating the determination of optimal combinations of these strains. Consequently, it allows the simplification of complex consortia without compromising degradation efficiency. In this study, starting from the highly efficient degrading microbial community obtained by enrichment culture, we designed a simplified SynCom based on metabolic modelling to achieve a degradation efficiency comparable to that of the complex microbial community (Figure 1). Notably, the metabolic model technology based on simulations can effectively reduce the time and economic costs for designing and constructing SynComs. More importantly, these biodegradation strategies based on SynComs can be effectively applied to the degradation of other organic pollutants (Xu et al. 2024). Thus, the strategy and method for designing and constructing a simplified SynCom proposed in this study will effectively promote the application of microbial communities in the field of pollution bioremediation.

While previous studies have employed modelling approaches to optimize community structure, most have not specifically targeted pollutant degradation, nor have they dissected in detail the mechanisms responsible for enhanced community metabolic activity (Peng et al. 2025; Ruan et al. ²⁰²⁶). To address this gap, we established a methodology for constructing functional microbial consortia and demonstrate its application to SynCom design. Our approach illustrates that metabolic modelling can effectively simplify complex consortia to guide the design of bioremediation strategies for bioaugmentation and biostimulation. Furthermore, we elucidate the specific mechanisms underlying efficient DCPP degradation by the SynCom. These include metabolic cross‐feeding between degrader and helper strains and the relief of intermediate‐induced inhibition of the degrader by helper strains. These findings provide theoretical support for the rational design and construction of effective SynCom for bioremediation.

Designing and constructing SynComs guided by metabolic modelling is an emerging method that is expected to be used for the precise engineering of natural microbial communities and the design of simple and efficient SynComs (Chen et al. 2025; Zomorrodi and Segrè ²⁰¹⁶; Paszko et al. ²⁰¹⁵). The accuracy of predicted results from these models is often highly dependent on the model's quality. In this study, we first constructed draft metabolic models of five strains involved in DCPP degradation (Figure 1). Subsequently, through very simple experiments, we determined the minimal medium for each strain and corrected the draft metabolic model accordingly (Figure 1). This ensured that the metabolic model could accurately characterize the physiological activities of the strains. Based on these high‐quality single‐strain metabolic models, we then constructed multi‐strain metabolic models which represent microbial community models encompassing various strain combinations (Figure 1). Through multiple simulations, we identified helper strains that could enhance the degradation efficiency of degraders and predicted performances of the combinations of degraders and helpers (Figure 1). These predictions were further verified experimentally, demonstrating the accuracy and feasibility of metabolic model‐based simulations (Figure 1). The design strategy of SynCom based on metabolic models has also been successfully applied to the construction of atrazine‐ and bromoxynil‐degrading microbial communities (Xu and Jiang 2024; Xu et al. ²⁰¹⁹; Ruan et al. ²⁰²⁴). These studies collectively indicate that simulations based on microbial community models composed of different high‐quality single‐strain models offer an effective solution for optimizing microbial communities and designing simplified SynComs.

Metabolic interactions are key determinants that shape functions of microbial communities, and a comprehensive understanding of these complex interactions can greatly facilitate the design and optimization of microbial communities (Cordero and Datta 2016; Dal Co et al. ²⁰²⁰; Ruan et al. ²⁰²⁴). However, elucidating the dynamic and complex metabolic interactions within microbial communities is often challenging for traditional experimental methods. In contrast, simulations based on metabolic modelling offer alternative solutions to this problem. By harnessing computational algorithms and mathematical models, community modelling can simulate and analyse the complex metabolic networks within microbial communities, enabling a more in‐depth exploration of the underlying metabolic interactions (Chen et al. 2025; Zomorrodi and Segrè ²⁰¹⁶; Widder et al. ²⁰¹⁶). For example, based on metabolic modelling, simulations of a two‐strain microbial community composed of Geobacter metallireducens and Geobacter sulfurreducens have revealed the competition and cross‐feeding mechanisms between community members (Nagarajan et al. 2013). This metabolic modelling‐based strategy is also applicable to three‐strain microbial communities (Harcombe et al. 2014; Louca and Doebeli ²⁰¹⁵). In this study, we extended the application of this approach by conducting modelling of more complex microbial communities, including those composed of four and five strains. Our findings indicate that metabolic modelling based on high‐quality GSMMs can be effectively employed to simulate the metabolic behaviour of complex microbial communities. These results highlight that metabolic model technology is a powerful tool for simulating microbial communities to explore complex metabolic interactions that are difficult to study using traditional experimental methods, opening up new avenues for understanding and manipulating microbial communities.

The reduction of DCPP in the soil mainly occurs through two ways: biodegradation and non‐biodegradation, with microbial degradation being the predominant pathway (Zhang et al. 2023, 2020). Microbial degradation offers several advantages, such as high efficiency, safety, low cost, and the absence of secondary pollution, and thus holds great promise for bioremediation (Xu and Jiang 2024; Chen et al. ²⁰²⁵). However, DCPP is a typical herbicide with a chiral structure, and pure‐culture microorganisms capable of completely degrading it are rare (Hu, Liu, et al. ²⁰²¹). Currently reported DCPP‐degrading strains include Sphingobium herbicidovorans MH, Delftia acidovorans MC1, Alcaligenes sp. CS1, Ralstonia sp. CS2, Stenotrophomonas maltophilia PM, Rhodoferax sp. P230, Sphingopyxis sp. DBS4 and Sphingobium sp. L3 (Hu, Liu, et al. ²⁰²¹; Zhang et al. ²⁰²⁰; Müller et al. ¹⁹⁹⁹; Horvath et al. ¹⁹⁹⁰; Ehrig et al. ¹⁹⁹⁷; Tett et al. ¹⁹⁹⁷; Schleinitz et al. ²⁰⁰⁴; Mai et al. ²⁰⁰¹; Smejkal et al. ²⁰⁰¹). Among these, the DCPP degradation pathways of strains MH, MC1, and DBS4 have been elucidated (Nielsen et al. 2017; Zhang et al. ²⁰²⁰; Müller et al. ¹⁹⁹⁹; Horvath et al. ¹⁹⁹⁰; Schleinitz et al. ²⁰⁰⁴). In the process of DCPP degradation, the dioxygenases encoded by genes RdpA and SdpA firstly catalyse the hydrolysis of the (R)‐ and (S)‐isomers, resulting in the formation of the non‐chiral compound 2,4‐DCP. Subsequently, 2,4‐DCP is completely degraded via the classic 2,4‐DCP degradation pathway and ultimately enters the tricarboxylic acid cycle. Notably, the model predicted the metabolic flow of the intermediate 2,4‐DCP from the degrader strain DBS4 to the helper strain DCP‐11d. Subsequent experiments confirmed that 2,4‐DCP inhibits the degradation activity of DBS4, while DCP‐11d alleviates this inhibition by consuming the toxic intermediate. This finding reveals the core mechanism underlying the optimal performance of the three‐strain SynCom and directly demonstrates that helper strains function through the detoxification of intermediate metabolites. Our results indicate that the cross‐feeding of metabolites unrelated to DCPP degradation between degraders and helpers within SynComs can significantly enhance DCPP degradation compared to that achieved by degraders only. This finding highlights the advantages of SynComs in the biodegradation of pollutants.

A common pattern exists in pesticide degradation: in most cases, degrader strains play the central role, while helper strains can further enhance their degradation efficiency. A major challenge in bioremediation is how to enhance the degradation efficiency of degrader strains at low cost. In this study, we constructed metabolic models to effectively simplify complex consortia and uncover the metabolic cross‐feeding between degrader and helper strains, obtaining the optimal combination of degrader and helper strains. Although only a single pesticide was employed as an example, the similar pesticide degradation pattern by microbial communities enables this tool to be extended to other pesticide systems. Moreover, this approach is also applicable to the design of synthetic consortia for combined pollution (Ruan et al. 2026). Accordingly, the established method exhibits extensive application potential in bioremediation. Notably, strains inoculated into soil typically decline to control levels within approximately 1–3 weeks (Chen et al. 2022), primarily due to competition with indigenous microorganisms, inadaptation to the soil environment, and depletion of the target contaminants. This characteristic, however, does not preclude the application of bioaugmentation. High metabolic activity can be sustained through repeated inoculation, enabling the rapid reduction of pollutants within a short time frame.

Several reports have documented the microbial degradation efficiency of DCPP (Feld et al. 2016; Zhang et al. ²⁰²³; Horvath et al. ¹⁹⁹⁰). For instance, (Feld et al. 2016) observed that microorganisms in a sand filter were able to degrade 30% of DCPP (0.14–0.17 μg/L) in the inlet groundwater within a two‐month period. Horvath et al. ¹⁹⁹⁰ discovered that Flavobacterium sp. MH could completely degrade 100 mg/L of DCPP within 70 h. In this study, leveraging metabolic modelling, we screened the optimal DCPP‐degrading microbial community (the SynCom DBS4&DCP‐4&DCP‐11d) from a complex natural microbial community, and verified the degradation ability of this three‐strain SynCom in both MM medium and soils. The results showed that it could reduce the concentration of 30 mg/L of DCPP to 1.245 mg/L within 17 h in MM medium. Under soil conditions, it was capable of degrading more than 75% of DCPP (at a concentration of 5 mg/kg) within 7 days. Notably, its degradation efficiency was comparable to that of the more complex five‐strain microbial community. Considering the strain culture conditions and costs, the simplified SynCom derived from the complex microbial community offers significant advantages and will facilitate the application of microbial communities in bioaugmentation. Furthermore, through simulation, we predicted that the exogenous addition of glucose could further enhance the degradation efficiency of the SynCom. This finding suggests that the addition of specific carbon sources can be employed for biostimulation. These predictions were subsequently validated by experimental results. Together, our findings demonstrate that metabolic model technology can be effectively applied to the design of bioaugmentation and biostimulation strategies for soils contaminated with organic pollutants.

We used the steady‐state assumption for performing FBA on GSMMs. The rationale for this assumption is grounded in the distinct timescales of cellular processes: metabolic reactions (e.g., substrate conversion) occur on timescales of milliseconds to seconds, while changes in biomass composition, gene expression, and environmental conditions occur over minutes to hours. Thus, over short‐to‐medium time frames, metabolic networks can be approximated as being at steady state, where fluxes adjust rapidly to maintain stable metabolite pools relative to these slower changes in biomass and regulation. Although this assumption simplifies the system and does not explicitly account for competitive or exclusionary interactions, its advantages are substantial. It obviates the need for kinetic parameters, which are largely unknown and required to solve dynamic mass balance equations. Importantly, despite its simplicity, steady‐state FBA has been shown to yield accurate predictions in numerous biological contexts (Heirendt et al. 2019). While a steady‐state‐based computational approach may appear inconsistent with the spatially heterogeneous and resource‐limited conditions typical of soil environments, consideration of temporal scales reconciles this. Over sufficiently short intervals relevant to metabolic flux analysis, the microenvironment surrounding a microbial consortium can be regarded as quasi‐constant, thereby aligning with the steady‐state premise.

Although our results suggest that enhanced degradation efficiency can be attributable to both cross‐feeding and intermediate detoxification, the relative contribution of each mechanism was not quantitatively resolved in this study. It is important to note that the primary purpose of the wet‐lab experiments was to validate the model predictions, not to further dissect the underlying mechanisms. Notably, intermediate detoxification itself constitutes a form of cross‐feeding: the helper strain efficiently consumes the toxic intermediate 2,4‐DCP, relieving its inhibitory effect on the degrader and thereby enhancing the degrader's metabolic activity. More broadly, intermediate metabolites can serve both as resources and as ecological stressors. For example, De Vos et al. found that 3‐hydroxypropionate can inhibit Anaerostipes rhamnosivorans, yet this compound can be converted by the same strain into propionate—illustrating that mutualistic relationships are not necessarily stable synergies, but may instead depend on a delicate balance among intermediate concentration, toxicity tolerance and subsequent transformation capacity (De Vos et al. 2024). Assessing this balance could help evaluate the stability of SynComs. Therefore, a quantitative dissection of these mechanisms warrants further investigation.

5. Conclusion

Our study provided a strategy for constructing efficient SynComs by simplifying complex microbial communities. Through this approach, we designed a SynCom for high‐efficient degradation of DCPP, and elucidated metabolic interactions among strains boosting the pollutant‐degrading efficiency of the SynCom. These results also highlighted the importance of helper strains in the SynCom. Furthermore, the addition of exogenous carbon sources such as glucose effectively enhanced the degrading performance of the SynCom. In summary, our findings demonstrated the promising application potential of community modelling in the design of SynCom to facilitate the utilization of natural complex microbial communities.

Author Contributions

Yahua Chen: investigation, supervision. Bingang Yang: validation, visualization. Zhenguo Shen: investigation, supervision. Chen Chen: conceptualization, writing – original draft, writing – review and editing, resources, project administration, supervision, funding acquisition. Jiandong Jiang: investigation, supervision. Wanxin Li: writing – review and editing, methodology. Shifeng Ding: validation, visualization, formal analysis. Xihui Xu: methodology, funding acquisition, project administration, resources, writing – original draft, writing – review and editing, conceptualization. Xinyu Lin: writing – original draft, formal analysis, visualization, validation, software, data curation.

Funding

This work was supported by the National Key Research and Development Program of China, 2024YFD1200202. National Natural Science Foundation of China, 32470113, 42477008. Outstanding Youth Foundation of Jiangsu Province, BK20250096. Jiangsu Agricultural Science and Technology Innovation Foundation, CX [24]1016. Earmarked fund for CARS‐10‐Sweetpotato.

Ethics Statement

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: Single‐strain GSMMs of the five strains.

Data S2: Community GSMMs of the fifteen SynComs.

Figure S1: Growth of strains in minimal mineral medium with various carbon sources.

Figure S2: DCPP‐degrading efficiency of the SynCom DBS4&DCP‐4&DCP‐11d in mineral medium supplemented with DCPP or DCPP and glucose (DCPP+glucose).

Figure S3: Identification of predicted exchange metabolites (Figure 2c) via LC–MS in a co‐culture of strains DBS4, DCP‐4, and DCP‐11d in mineral medium containing DCPP as the carbon source. (A) α‐Ketoglutaric acid; (B) L‐Phenylalanine; (C) N‐Acetyl‐D‐glucosam; (D) L‐Lysine; e Stearic acid; (F) 2,4‐DCP. The fragment peaks of compounds in the second‐order mass spectrum are shown. The quality spectrum of standard compound is presented in Table S3.

Figure S4: Degradation of DCPP by strain DBS4. The DCPP degradation was tested in three media: one containing DCPP as the sole carbon source (DBS4), one with DCPP and 2,4‐DCP as carbon sources (DBS4 + 2,4‐DCP), and one with DCPP and exchanged metabolites (DBS4 + exchanged metabolites). A mixture of D‐Glucosamine, α‐Ketoglutaric acid, L‐Phenylalanine and L‐Lysine was used as the exchanged metabolites.

Table S1: Substrate spectrum of the five isolated strains.

Table S2: General features of the four genomes.

Table S3: Quality spectrum of standard samples.

Data Availability Statement

The data that support the findings of this study are openly available in GenBank at https://www.ncbi.nlm.nih.gov/genbank/, reference number SUB15680332, SUB15680937, SUB1568104 and SUB15681330.