Genome‐Scale Community Model‐Guided Development of Bacterial Coculture for Lignocellulose Bioconversion

ABSTRACT

Microbial communities have shown promising potential in degrading complex biopolymers, producing value‐added products through collaborative metabolic functionality. Hence, developing synthetic microbial consortia has become a predominant technique for various biotechnological applications. However, diverse microbial entities in a consortium can engage in distinct biochemical interactions that pose challenges in developing mutualistic communities. Therefore, a systems‐level understanding of the inter‐microbial metabolic interactions, growth compatibility, and metabolic synergisms is essential for developing effective synthetic consortia. This study demonstrated a genome‐scale community modeling approach to assess the inter‐microbial interaction pattern and screen metabolically compatible bacterial pairs for designing the lignocellulolytic coculture system. Here, we have investigated the pairwise growth and biochemical synergisms among six termite gut bacterial isolates by implementing flux‐based parameters, i.e., pairwise growth support index (PGSI) and metabolic assistance (PMA). Assessment of the PGSI and PMA helps screen nine beneficial bacterial pairs that were validated by designing a coculture experiment with lignocellulosic substrates. For the cocultured bacterial pairs, the experimentally measured enzymatic synergisms (DES) showed good coherence with model‐derived biochemical compatibility (PMA), which explains the fidelity of the in silico predictions. The highest degree of enzymatic synergisms has been observed in C. denverensis P3 and Brevibacterium sp P5 coculture, where the total cellulase activity has been increased by 53%. Hence, the flux‐based assessment of inter‐microbial interactions and metabolic compatibility helps select the best bacterial coculture system with enhanced lignocellulolytic functionality. The flux‐based parameters (PGSI and PMA) in the proposed community modeling strategy will help optimize the composition of microbial consortia for developing synthetic microcosms for bioremediation, bioengineering, and biomedical applications.

This study demonstrated a genome‐scale community modeling approach to assess the inter‐microbial interaction pattern and screen metabolically compatible bacterial pairs for effective co‐culture design. The authors constructed community models of six lignocellulolytic microbial strains and then looked at the pairwise growth support index and metabolic assistance. This helped to identify nine beneficial bacterial pairs for developing mutualistic bacterial co‐culture systems. The model‐driven bacterial co‐cultures exhibited better enzyme activities on filter paper, carboxymethyl cellulose, and xylan than monocultures.

1. Introduction

The specialized microbial guild residing in the wood‐feeding termite gut environment has been considered the most effective lignocellulose degraders (Brune 2014). The broad spectrum of enzymatic cocktails produced by these specialized microbial communities helps execute the complex biological processes of lignocellulose bioconversion. (Auer et al. 2017; Bredon et al. 2018; Brenner, You, and Arnold 2008). Several genomics and metagenomics studies have revealed the structural and functional diversity of the termite gut microbiota essential to lignocellulose bioconversion (Rossmassler et al. 2015; Warnecke et al. 2007). The enriched gene pools of the termite gut microbial symbionts for efficient lignocellulose bioconversion (Arora et al. 2022) have also been revealed through metagenomics and metatranscriptomics. A diverse set of bacterial genes involved in the cellulose and xylan hydrolysis has also been detected through the metagenomics analysis of the hindgut paunch of wood‐feeding higher termite species (Hu et al. 2019; Warnecke et al. 2007). Moreover, the combination of shotgun sequencing, fosmid libraries, and biochemical assays enables the identification of around 219 functional genes encoding crucial carbohydrate‐active enzymes (CAZymes) (Liu et al. 2019). The complementary CAZymes cocktails were essential for the effective digestion of cellulose, hemicellulose, and pectin by the termite gut microbial populations (Marynowska et al. 2023).

Despite the characterization of numerous microbial genera and their gene pools, the basis of metabolic interdependency in microbial communities has remained indispensable. Inside a microbial consortium, the exchange of metabolic compounds in the community environment defines the ultimate phenotypic properties of an individual microbe. These interactions lead to collective biochemical properties of the microbial guild through metabolic interdependencies, cross‐feeding, and division of labor (Faust and Raes 2012; Seth and Taga 2014). Although the basis of the bacterial synergisms is not entirely decoded, several recent studies have shown the phenomenon of metabolic cooperation and division of labor in the co‐existing microbial community utilizing lignocellulosic material (Ali et al. 2023; Detain et al. 2022). The collaborative metabolic activities of a five‐member microbial community have been developed earlier for the effective degradation of the lignocellulosic material (Kato et al. 2005). Further, based on the division of metabolic labor, a co‐existing microbial consortium of Trichoderma reesei and Escherichia coli has been developed (Minty et al. 2013). In this coculture system, the enzymatic activity of T. reesei transformed the lignocellulosic biomass into soluble sugars, and E. coli fermented these into isobutanol. However, the microbial metabolic activities in a co‐existing community do not always lead to cooperative interaction. In a complex microbial consortium like termite gut microbiota, diverse metabolic activities of the microbial species can lead to positive, negative, or neutral outcomes (Faust and Raes 2012; Kundu et al. 2019). Moreover, the combinatorial assessment of different inter‐species interaction states can escalate the degree of interaction types, even with a small set of microbial entities. Therefore, pinpointing the fundamental interactions is crucial to identify the beneficial or compatible microbial pairs for improved functionality. Assessing the complex microbial interaction pattern with experimental coculture strategies may be challenging and tedious. Hence, implementing in silico modeling strategies can provide essential knowledge to screen the synergistic microbial entities and develop mutualistic communities for complex metabolic tasks.

Systems biology tools like genome‐scale community metabolic modeling help investigate the biochemical pathways of individual microbes and microbial consortia for optimizing their metabolic activities (Fondi and Liò 2015; Ibrahim, Raajaraam, and Raman 2021). The genome‐scale community models (GCMs) essentially provide the biochemical flux distribution patterns of individual and community‐level metabolic networks optimized in different environments (Beura et al. 2022; Perez‐Garcia, Lear, and Singhal 2016). Hence, community modeling has been widely used to study the intricate metabolic activities and interactions among diverse microbial members and provide essential biochemical knowledge for establishing effective coculture systems. Recently, a model‐assisted coculture setup followed by constrained‐based flux simulations of the human gut microbial species Phocaeicola dorei and Lachnoclostridium symbiosum predicted high lactate and succinate cross‐feeding fluxes when growing in inulin or xylan (Hirmas et al. 2022). Further, a GEM‐guided cocultivation of Pichia stipitis and Saccharomyces cerevisiae showed the cross‐feeding of ethanol to derive maximum advantage in the community environment (Ravikrishnan et al. 2020). Moreover, in the case of more extensive and complex microbial consortia, the community models help to track the combinatorial effect of the metabolic cross‐talk to find the metabolic and growth compatibility between the community members (Kumar et al. 2022; Kundu and Ghosh 2023). Hence, assessing the inter‐microbial metabolic assistance in the community environments helps to find the growth and metabolically compatible microbial pairs. The enzymatic and biochemical synergies of these compatible microbial communities can then be tested with experimental setups.

In this study, we have demonstrated a genome‐scale community modeling approach with newly introduced parameters, like pairwise growth support index (PGSI) and pairwise metabolic assistance (PMA), to screen the mutualistic bacterial communities for effective coculture development. The individual GEMs of six lignocellulolytic bacterial strains, i.e., Micrococcus luteus P1, Kluyvera sp. P2, Cellulomonas denverensis P3, Oceanobacillus sojae P4, Brevibacterium sp. P5 and Niallia circulans P6 were initially created to evaluate the flux distribution in metabolic pathways. Further, the individual GEMs were integrated to reconstruct community models for all the possible bacterial pairs. We have defined the PGSI and PMA to predict growth compatibility and metabolic assistance through the GCM analysis. The synergistic bacterial communities predicted through PGSI and PMA were tested by setting up the coculture experiments with the lignocellulosic substrates. Here, the experimentally derived degree of enzymatic synergism (DES) in filter paper (FP) (total cellulase), carboxymethyl cellulose (CMC) (endoglucanase), and xylan (xylanase) have been compared with the GCMs‐derived PMA values. The maximum DES has been achieved by cocultured bacterial communities C. denverensis P3 and Brevibacterium sp. P5, while growing in FP. The experimentally obtained DES of bacterial coculture correlated well with the model‐driven PMA values. Hence, the community model analysis helps design the optimal bacterial combination to overcome the substrate‐specific limitation of individual strains for efficient lignocellulose bioconversion. Overall, this in silico methodology highlighted implementing systems biology tools to predict the compatible microbial coculture systems, leading to the development of efficient lignocellulolytic bacterial consortia.

2. Methods

2.1. Isolation of Lignocellulolytic Bacteria

The wood‐feeding termites were collected from the IIT Kharagpur campus, West Bengal, India, and immediately brought to the laboratory. The termites were then surface sterilized with 70% ethanol (v/v) and subjected to dissection aseptically in a laminar airflow chamber. The gut extract was dissolved in 2–3 mL sterile phosphate buffer saline (PBS) solution. One milliliter of the homogenized gut extract was then inoculated in Erlenmeyer flasks containing 100 mL of peptone cellulose solution (PCS) medium (0.5% peptone, 0.5% NaCl, 0.1% yeast extract, 0.15% CaCO3, and 0.5% cellulosic material) at pH 7.2 under aerobic conditions supplemented with 0.5%–1% CMC and FP as cellulosic substrates (Kato et al. 2005; Samir Ali et al. 2019). The culture media has been supplemented with cycloheximide (0.02%) to restrict fungal growth. The inoculated enrichment medium was then incubated at 37°C on a rotary shaker (Innova, New Brunswick, USA) at 120 rpm for 2–3 days. Culture Samples collected from different time points of the sub‐cultivation were subjected to serial dilution to isolate individual microbial species. Here, 100 mL of the culture was serially diluted up to 10^−12, and 200 mL of each diluted culture was spread on PCS agar plates containing 0.5% (w/v) CMC. The visible bacterial colony has been observed by incubating the agar plates at 37°C for 48–72 h. A total of 19 individual microbial colonies were initially picked (Table S1), and the colony morphologies, i.e., color, shape, elevation, and margin, were assessed (Bhaduri et al. 2016; Singh et al. 2019). These microbial colonies were transferred to PCS agar plates and made into pure cultures through repeated streaking. The pure bacterial isolates were cryopreserved in nutrient broth with 15% glycerol at −80°C.

2.2. Screening the Cellulolytic Activity of Bacterial Isolates

The plate‐based technique has been acquired for the initial screening of the cellulolytic and xylanolytic activity of the individual bacterial isolates (Hendricks, Doyle, and Hugley 1995). Here, the individual isolates were grown in different patches on PCS media supplemented with 0.5% CMC (w/v) or 0.5% corn comb xylan for 48–72 h at 37°C (Kato et al. 2005; Samir Ali et al. 2019). After the incubation, properly grown bacterial patches were immersed in a 0.1% aqueous Congo red solution for 10–15 min. After removing the excess stain, the plates were washed with 1 M NaCl for 10 min, then dried overnight at ambient temperature. The clear zone (“Halos”) has been observed around several bacterial colonies. The diameter of halos (cm) was measured to determine cellulose degrading potential semiquantitatively (Dar et al. 2015; Hendricks, Doyle, and Hugley 1995). The hydrolytic capacity (HC), i.e., the ratio of the diameter of the zone of clearance to the diameter of the bacterial colony, was measured to estimate the cellulose‐degrading potential of the positive isolates (Table S1).

The bacterial isolates with higher HC have been identified by amplifying the 16S rRNA gene (Chen et al. 2015) followed by sequencing (see Supporting Information for detailed methodology). The generated sequences were analyzed, followed by a similarity search in the NCBI to identify the closest matches to the bacterial isolates (Table S2). Based on the sequence similarly, the six best‐performing bacterial strains were identified to be Micrococcus luteus P1, Kluyvera sp. P2, Cellulomonas denverensis P3, Oceanobacillus sojae P4, Brevibacterium sp. P5 and Niallia circulans P6. The nucleotide sequences were then submitted to the NCBI GenBank repository under the accession number OR825811‐ OR825819.

2.3. The Coculture Systems and Growth Synergism

The pairwise bacterial cocultures were developed by combining two distinct strains in the liquid PCS media with CMC as a metabolic substrate. Firstly, each bacterial culture was grown to log phase in the monoculture at 37°C. Based on the in silico‐predicted compatible bacterial pair, 0.5 mL of bacterial cultures (from two distinct monocultures) were added to the fresh PCS media. Here, we measured the growth rate of the cocultured bacterial community and compared it with the respective monoculture growth in the PCS CMC media to understand the degree of growth synergism. The growth has been calculated as follows.

$$\mathrm{Degree}\mathrm{of}\mathrm{growth}\mathrm{synergies}\left(\mathrm{DGS}\right)=\frac{\text{Growth in cocultured community}}{\text{Sum of the individual growth rate}}.$$

Further, these bacterial cocultures have also been tested with CMC, FP, and xylan as metabolic substrates to measure the total cellulase (FPase), endoglucanase (CMCase), and xylanase enzyme activities, respectively. For these, all the cultures were maintained at 37°C for 72 h with an agitation rate of 120 rpm. All the experimental measurements were done in triplets.

2.4. Reconstruction and Refinement of the Genome‐Scale Metabolic Models

The whole‐genome sequence (GCA_006094415.1, GCA_000735365.1, GCA_002240455.1, GCA_003351985.1, GCA_014863525.1, and GCA_003726115.1) of the best‐performing bacterial strains, i.e., Micrococcus luteus P1, Kluyvera sp. P2, Cellulomonas denverensis P3, Oceanobacillus sojae P4, Brevibacterium sp. P5 and Niallia circulans P6, were collected from the NCBI database. The genome sequences of these individual bacterial strains were annotated using the RAST server in modelSEED (v6.2.1) to reconstruct the draft GEMs with proper gene‐protein‐reaction (GPR) information (Henry et al. 2010). The model analysis, refinement, and manual curation have been executed in manipulation in COBRApy (Ebrahim et al. 2013) using Gurobi (Academic License Version 8.0) and GLPK solver in Systems Biology Markup Language (SBML) format (Hucka et al. 2003). We have used PubChem and the BiGG database to correct and assign the metabolites' charge and biochemical formula. Further, the gap‐filling of the unassigned metabolic pathways was done through manual curation using the KEGG pathway database (Kanehisa et al. 2017), ModelSeed reaction databases (Henry et al. 2010), BiGG Database (Schellenberger et al. 2010), and existing research knowledge. The individual draft models generated through ModelSEED lack the metabolite's charged formula. Hence, we have assigned the correct chemical formula and charge for every metabolite by referring to CHEBI (Chemical Entities of Biological Interest), PubChem (NIH), and the BiGG Database (UCSD) (Schellenberger et al. 2010). Subsequently, the databases (CHEBI, BiGG) were also used to balance the identified proton‐imbalanced reactions by the addition/deletion of one or multiple protons on either the reactant or product side. However, in the case of the remaining imbalanced reactions, the reaction stoichiometry was revised to ensure that the atoms on both sides of the reaction balanced out. Moreover, the individual draft models contain a complete medium (containing up to 70 metabolites) to produce the biomass precursors. Because several carbon sources and unbound reactions are available, each model shows an unrealistically high biomass formation rate (biomass flux up to 106.4 mmol/GDW/h). Hence, we first curated the metabolic pathway associated with biomass precursor formation to ensure biomass formation with minimal media (up to 20 metabolites). To curate each metabolic pathway, we have used the previously published model of the closely related microbe, the KEGG pathway database, and the biochemical information from the published literature. To reflect the cellulose degradation potential of the microbial species, we added reactions for endoglucanase (cellulose to cellodextrin), exoglucanase (cellodextrin to cellobiose), and beta‐glucosidase (cellobiose to glucose) into the GEMs. Since the cell cannot directly transport cellulose, these reactions convert it to glucose in the extracellular space. The final degradation results in the production of glucose, which the GEMs can then import as substrate. We have experimentally determined the glucose uptake rate of the individual strains against the dry cell biomass. Detailed information on individual metabolic model curation has been provided in Table S4 of the supplementary data.

Flux Balance Analysis (FBA) (Orth, Thiele, and Palsson 2010) has been performed using the Gurobi solver in Python 3.8 for model testing, validation, and analyzing the flux distributions at different sections of the work. The FBA optimization formulation is as follows:

$$\mathit{Maximize}\left(v_j\right)v_{\mathit{Biomass}},$$

$$\mathit{Subjected}\mathit{ }\mathit{to}\sum _{jϵJ}{S}_{\mathit{ij}}.v_j=0\forall iϵI,$$

$${\mathit{LB}}_j\le v_j\le {\mathit{UB}}_j\forall \mathrm{j}\in J,$$

I and J represent the sets of metabolites and reactions in the reconstructed model. S _ij is the stoichiometric coefficient of metabolite ‘i’ in reaction ‘j’. The flux value of reaction j is denoted with v _j . The LB _j and UB _j denote the minimum and maximum permissible fluxes for reaction j, respectively. The biomass reaction flux is represented with v _Biomass .

2.5. Reconstruction of Genome‐Scale Community Metabolic Model

The refined GEMs of the individual strains have been integrated with the compartmentalized approach (Diener and Gibbons 2023; Kundu and Ghosh 2023) to build the GCM. Each bacterial species in GCM represents a compartmented metabolic network where the essential metabolites can be exchanged through the ‘e0’ compartment representing the extracellular environment. The maximization of the community biomass formation has been set as the objective function for the GCM. The community biomass was defined by summing the stoichiometric coefficients of the individual model's biomass reactions using the following optimization formula (Diener and Gibbons 2023; Kundu and Ghosh 2023; Zampieri, Efthimiou, and Angione 2023).

$$Z=\max \left(v_{\mathit{Biomass}}^{\mathit{Community}}\right),$$

$$\mathit{where}\mathit{ }\mathit{ }{v}_{\mathit{Biomass}}^{\mathit{Community}}=\sum _{k=1}^N{v}_{\mathit{Biomass}}^k.$$

Here, $v_{\mathit{Biomass}}^{\mathit{Community}}$ Represent the community biomass reaction. In GCM, the individual model's biomass coefficient has been rescaled based on their relative growth rate or abundance. Thus, the following equation has been used to maximize the Community biomass.

$$\mathit{Community}\mathit{ }\mathit{Biomass}=\sum _{k=1}^N{a_k\times v}_{\mathit{Biomass}}^k.$$

Here, a _k is the relative growth rate of a particular bacterial strain. Therefore, each individual biomass reaction is present in the community model with a specific coefficient based on the specific growth rate or abundance. For example, in the pairwise community model of M. luteus P1 and Kluyvera sp. P2, the individual biomass (cpd11416_c0) was tagged with microbe‐specific identifiers (cpd11416_c0ml for P1 and cpd11416_c0ksp for P2). The GCM can predict how the individual biomass formation and flux distribution alter in the community environment while maximizing the community biomass.

2.6. Pairwise Growth Support Index

Microbial entities can enhance or restrain the growth of fellow community members while residing in a consortium (Faust and Raes 2012; Kundu, Mondal, and Ghosh 2022; Perez‐Garcia, Lear, and Singhal 2016). To estimate the growth‐enhancing or growth‐limiting effect of a microbial entity on its fellow community members, we have formulated the PGSI. PGSI was calculated with each GCM to examine the growth compatibility of all microbial combinations. Considering a microbial pair P and Q, the effect of microbial species P on the Q's growth rate was calculated as:

$${\text{PGSI}}_{Q\to P}=\frac{v_{\mathit{bioP}|\mathit{PUQ}}-v_{\mathit{bioP}|P}}{v_{\mathit{bioP}|P}},$$

where $v_{\mathit{bioP|PUQ}}$ represents the biomass formation rate of P in community PUQ and $v_{\mathit{bioP|P}}$ is the biomass production rate of $\mathit{P}$ when growing individually. Thus, PGSI _Q→P  = 0 denoted that the microbe Q does not affect the growth rate of microbe P. Whereas PGSI _Q→P  > 0 signifies that the microbe Q positively influences microbe P, and PGSI _Q→P  < 0 indicates microbe Q has a net negative impact on microbe P's growth rate. Therefore if $v_{\mathit{bioPQ}}\ge {(v}_{\mathit{bioP}}+v_{\mathit{bioQ}})$, the microbial community PUQ can be called growth‐compatible. Alternatively, if $v_{\mathit{bioPQ}}<(v_{\mathit{bioP}}+v_{\mathit{bioQ}})$, then the community can be called growth‐incompatible.

To estimate the PGSI in a three‐membered community, we calculated the change (increase or decrease) in each individual bacterial growth within the community. Therefore, we can represent the PGSI as follows:

$${\text{PGSI}}_{X\to P}=\frac{v_{\mathit{bioP}|X}-v_{\mathit{bioP}|P}}{v_{\mathit{bioP}|P}}.$$

The biomass formation rate of P in community X, which includes the other two bacterial entities, is represented by ${\mathrm{v}}_{\text{bioP|X}}$. We calculated the PGSI that each of the three individual microbes received from their fellow community members, and then took the average to represent the overall PGSI of the community.

2.7. Flux‐Based Estimation of PMA

A bacterial strain's metabolic benefit when growing with another community member has been quantified through PMA. The PMA of a specific bacterial strain P growing in a pairwise community P ∪ Q has been estimated by considering the fraction of flux‐range‐expanded and activated metabolic reactions in an individual microbe P (Kumar et al. 2022; Kundu and Ghosh 2023).

$$\mathit{PMA}\left(P\right|P\cup Q)=\frac{\left(\frac{n_{\mathit{expanded},P|P\cup Q}}{N_P}\right)+\left(\frac{n_{\mathit{opened},P|P\cup Q}}{N_P}\right)}{2},$$

where $n_{\mathit{expand}}$ represent the number of reactions with an expanded flux range observed through FVA. $n_{\mathit{opned}}$ denoted the number of activated reactions that were initially blocked in the individual GEM and $N_P$ is the total number of reactions in the metabolic network of microbe P. Hence, a $\mathit{PMA}\left(P\right|P\cup Q)=1$ suggested that the bacterial strain P received 100% metabolic benefit while growing in a pairwise community with Q. Whereas a $\mathit{PMA}\left(P\right|P\cup Q)=0$ indicates no metabolic benefits obtained by P while growing with Q.

Similarly, the community metabolic assistance (CMA) calculates the fractions of flux‐range‐expanded reactions and activated reactions in an individual microbe P while growing in multispecies community $X$ as:

$$\mathit{CMA}\left(P\right|X)=\frac{\left(\frac{n_{\mathit{expanded},P|X}}{N_P}\right)+\left(\frac{n_{\mathit{opened},P|X}}{N_P}\right)}{2},$$

where $X$ denoted the multispecies community, P denoted the individual metabolic network of microbe P, $n_{\mathit{expanded}}$ denoted the number of flux rage expanded reactions when growing in multispecies community $X$, $n_{\mathit{opned}}$ denoted the number of activated reactions in microbe P and $N_P$ represent the total number of reactions in P. Thus, a $\mathit{CMA}\left(P\right|X)=1$ indicates that microbe P derived 100% metabolic benefit when growing in a multispecies community. Whereas a $\mathit{CMA}\left(P\right|X)=0$ denoted no metabolic benefits towards P while growing with Q.

2.8. Estimation of Enzyme Activity and DESs

The total cellulase (FPase), endoglucanase (CMCase), and xylanase enzyme activity of the bacterial monoculture and coculture were estimated through the dinitrosalicylic acid (DNS) method (Dar et al. 2015; Miller 1959) using FP, CMC, and xylan, respectively. The crude extracellular enzymes were collected for the total cellulase assay by taking 2 mL of 24, 48, and 72 h old bacterial cultures growing in PCS with FP as a substrate. The collected samples were centrifuged at 7000 rpm for 8 min. The supernatant containing the crude enzyme was then added to tubes containing 50 mg of Whatman FP (no. 1). Further, 1 mL of citrate buffer was added to the tubes, followed by incubation at 50°C for 1 h in the water bath. After that, 3 mL of DNS reagent was added to each tube. The endoglucanase activity of the microbial cultures was done using CMC as the substrate. A total of 250 µL of the crud enzymes from each monoculture or coculture experiment were mixed with 750 µL of 1% CMC solution in citrate buffer. The mixture was kept in a 50°C water bath for 30 min. After the incubation, 3 mL of DNS reagent was added to terminate the reaction. For the xylanase assay, 500 µL of crud enzyme was mixed with 500 µL of 1% corn‐comb xylan solution, incubated at 50°C for 15 min, and finally terminated with 3 mL of DNS solution. All the enzyme‐substrate‐reagent mixtures were then kept in a water bath at 100°C for 10 min for color development. After that, the tubes were allowed to cool (room temperature), and the OD was measured at 540 nm (Miller et al. 1960). The concentration of the enzymes has been estimated against the glucose standard. Based on the reduced sugar concentration, the enzyme activities have been calculated as follows (Biz et al. 2014; Singh et al. 2019):

$$\text{Enzyme activity}(U/mL)=\frac{C_{\mathit{rs}}\left(\mathrm{t}\right)\times \text{RxnVol}\times \text{DlF}}{t_{\mathit{incubation}}\times \text{EnzVol}\times {\mathit{MW}}_{\mathit{rs}}}\times 1000.$$

Where C_rs (t) is the concentration of the reducing sugar at time t, RxnVol is the total volume of the reaction, DlF is the dilution factor, and t_incubation is the incubation time. EnzVol is the enzyme volume, and MW_rs is the molecular weight of the reducing sugar. All the enzyme activities have been measured in triplets.

By comparing the FPase (total cellulase), CMCase (endoglucanase), and xylanase activity of the individual bacterial strain and respective cocultures, we have measured the DES in the pairwise coculture.

$$\text{Degree of Enzyme Synergies}\left(\text{DES}\right)=\frac{{\text{EA}}_{\text{PQ}}}{{\text{EA}}_{\mathrm{P}}+{\text{EA}}_{\mathrm{Q}}}.$$

Where EA_PQ is the combined enzyme activity of the microbes P and Q in the coculture systems, EA_P and EA_Q denote the enzyme activity of microbes P and Q in the monoculture. Hence, DES > 0 signifies the enzymatic synergisms between microbial pairs in the coculture system.

3. Results and Discussion

3.1. Screening Lignocellulolytic Bacterial Isolates From Termite Gut Microbiota for GEM Reconstruction

This study intends to design an optimized lignocellulose degrading bacterial community with termite gut isolates through genome‐scale community metabolic modeling. The enrichment of the bacterial species was initiated by inoculating the termite gut extract in the liquid culture with CMC as the metabolic substrate. Initially, 35 enriched colonies were subjected to detailed morphological characterization. Based on similar morphological traits (colony shape, color, elevation, and margin types), 16 bacterial isolates were excluded (Bhaduri et al. 2016; Singh et al. 2019). The remaining 19 bacterial isolates were used to screen the hydrolytic capacities (HC) with CMC and xylan. The positive bacteria isolated produce a zone of clearance (‘Halos’) around their colony in the agar plates containing PCS media while incubated at 37°C for 48–72 h (Figure 1A) (Table S1). The HC of the bacterial isolates has been determined by measuring the ratio of the diameter (cm) of the zone of clearance to the diameter (cm) of the colony (Dar et al. 2015; Hendricks, Doyle, and Hugley 1995). The bacterial strains showed varied hydrolytic potential for the different substrates, and a maximum HC of 3.64 and 3.11 was achieved in CMC and xylan, respectively. Based on the hydrolyzing potential of bacterial isolates, six best‐performing isolates have been selected for molecular phylogeny analysis. Amplification of the 16S rRNA genes (Figure S1) and the sequence similarity search with NCBI BLAST helped identify the six bacterial strains as Micrococcus luteus P1 and Kluyvera sp. P2, Cellulomonas denverensis P3, Oceanobacillus sojae P4, Brevibacterium sp. P5, and Niallia circulans P6. The 16S rRNA gene amplicon sequences (FASTA) of the bacterial isolates have been submitted to the NCBI GenBank repository under the accession number OR825811 ‐ OR825819 (Table S2).

Figure 1.

Screening the lignocellulolytic abilities of the bacterial isolates: (A) The hydrolytic capacities (HC) of the bacterial isolates in CMC and xylan. The positive bacteria isolated showed a zone of clearness (‘halos’) around their colony. The HC of the positive bacterial isolates was semi‐quantitatively measured by calculating the diameter ratio of the ‘halos’ and the colony measurements. The enzyme activity profile of the individual bacterial isolates. The enzyme activity of the six best‐performing bacterial strains in filter paper (B), CMC (C), and xylan (D) are represented in the vertical axis against different time points (24, 48, 72, and 96 h). All the data points represent the average value from triplicate experimental measurements of enzyme activities.

The lignocellulose degrading potential of six termite gut bacterial isolates, i.e., Micrococcus luteus P1 and Kluyvera sp. P2, Cellulomonas denverensis P3, Oceanobacillus sojae P4, Brevibacterium sp. P5 and Niallia circulans P6 have been assessed using the dinitro salicylic acid (DNS) method (Dar et al. 2015; Miller 1959). The total cellulase (FPase), endoglucanase (CMCase), and xylanase enzyme activity have been measured for the bacterial isolates at different time points using FP, CMC, and xylan, respectively. The bacterial strains M. luteus P1, C. denverensis P3, and N. circulans P6 showed a constant increase in the total cellulase potential up to 72 h, achieving the maximum activity of 0.1503 U mL⁻¹ (±0.012), 0.1163 U mL⁻¹ (±0.006), and 0.1604 U mL⁻¹ (±0.012), respectively (Figure 1B) (Table S3). On the other hand, the FPase activity of Bervibacterium sp. P5 reached maximum [0.1636 U mL⁻¹ (±0.02)] at 48 h, and a slight decrease in activity was observed afterward. Overall, M. luteus P1, N. circulans P6, and Bervibacterium sp. P5 were the best‐performing individual strains in the FP substrate with the highest enzymatic activities. However, despite the high FPAse activity, M. luteus P1 and Bervibacterium sp. P5 exhibits relatively lower degradation potential in the CMC with an average activity of ~0.1 U mL⁻¹ (±0.009) between 24 and 48 h. As the CMC degradation potential majorly depends on the presence of only endoglucanase enzyme (Singh et al. 2019), the low production of endoglucanase may lead to decreased CMCase activity in strains P1 and P5. In contrast, the total cellulase (FPase) activity is the combined potential of β‐glucosidase, endo‐1,4‐β‐d‐glucanase (endoglucanase), and exo‐1,4‐β‐d‐glucanase (exoglucanase) enzymes of a microbial species. Hence, the strains P1 and P5 may produce higher amounts of β‐glucosidase and exoglucanases (not estimated in this study), which leads to higher FPase activity in P1 and P2. Using CMC as the metabolic substrate, O. sojae P4, C. denverensis P3, and N. circulans P6 were found to be the best‐performing strains that achieved an enzyme activity up to 0.184 U mL⁻¹ (± 0.006) (Figure 1C). However, despite the high CMCase activity, P4 showed a lower FP degradation (FPase) potential (Table S3). This could be due to the higher secretion of endoglucanase enzymes and the lower production of β‐glucosidase and exoglucanase (not quantified in this study) by O. sojae P4. Therefore, the enzymatic potential of the isolated bacterial strains varied significantly depending on the substrates. This trend has also been observed for the xylan‐supplemented media where Kluyvera sp. P2 showed good enzymatic potential 0.25 U mL⁻¹ (± 0.011) in the xylan‐supplemented media, but a lower activity was obtained in CMC and FP substrates (Figure 1D). Hence, the selected bacterial isolates showed significant enzyme activity in one of the given substrates, but their hydrolytic potential greatly varied with the substrate‐specific condition (Figure 2). Strains exhibiting good degradation potential in one substrate may not necessarily perform similarly in an alternative substrate. A Kruskal–Wallis test (Kruskal and Wallis 1952) and Tukey's HSD analysis showed significant differences [H _stat  = 10.04, p = 0.0066] between the FP, CMC, and xylan hydrolysis activity by the individual strains. The low hydrolytic potential of individual strains on specific types of lignocellulosic substrates signifies their substrate‐specific limitation that hinders the complete bio‐conversion of lignocellulosic material. This substrate‐dependent limitation of the individual bacterial activity can be overcome by developing synergistic bacterial coculture. The division of labor and cooperative metabolic activities in the microbial community help overcome the metabolic burden of fellow community members for a more efficient lignocellulolytic activity, as observed in previous studies (Kato et al. 2004; Lü et al. 2013; Singh et al. 2019). However, designing pairwise coculture systems by combining a large set of bacterial strains may pose several challenges. Moreover, because of the complex metabolic interactions between microbial species, implementing the model‐guided predictive method may help assess the growth and metabolic compatibility between microbial pairs. In this context, genome‐scale metabolic modeling is considered an effective tool that provides systems‐level insight into the biochemical flux distribution patterns of microbial entities. Thus, the implementation of genome‐scale modeling can predict the beneficial metabolic interaction patterns in microbial communities through in silico flux analysis to find synergistic bacterial pairs.

Figure 2.

The Maximum enzymatic activity of bacterial isolates in different lignocellulosic substrates: The horizontal axis represents the average maximum FPase (black), CMCase (red), and xylanase (blue) activities of the selected bacterial strains (vertical axis) derived from the triplicate experimental measurements. The error bar represents the standard deviation.

3.2. Reconstruction of Pairwise GCMs for Identifying Growth‐Compatible Bacterial Pairs

Assessment of the lignocellulolytic enzyme activities of the termite gut bacterial isolates helped identify the six best‐performing strains in terms of FP, CMC, and xylan degradation. The genome‐scale community modeling strategy has been implemented to understand the systems‐level biochemical activities of the individual species and screen the synergistic bacterial pairs for developing an effective coculture system. The genome‐scale metabolic models were initially reconstructed for six best‐performing bacterial isolates, i.e., M. luteus P1, Kluyvera sp. P2, C. denverensis P3, O. sojae P4, Brevibacterium sp. P5 and Niallia circulans P6. The individual GEMs were first developed using the GPR association information from the annotated genomes of each species in the ModelSeed platform (Faria et al. 2018). The chemical imbalance, reaction reversibility, and other metabolic anomalies of the reconstructed GEMs were corrected manually using PubChem (NIH), ModelSeed reaction databases (Henry et al. 2010), and BiGG Database (Schellenberger et al. 2010). The individual models were further refined by gap‐filling several microbe‐specific biochemical pathways based on the KEGG database (Kanehisa et al. 2017). The refined models contain an average of 1460, 1726, and 1314 reactions, genes, and metabolites, respectively (Table 1). The individual GEMs were optimized using FBA with biomass as the objective function. To validate the in silico models, the glucose uptake rate of the individual microbes was measured experimentally. This glucose uptake rate was used to constrain each model, followed by the estimation of the in silico biomass formation rate (Table S5). Since the biomass formation rate of the in silico models correlates well (average Pearson r = 0.92, p = 0.008) with the experimentally measured growth values (Figure S2), these individual GEMs have been used for the community model (GCM) reconstruction.

Table 1.

Information on the structural properties of the genome‐scale metabolic models (GEMs) of the bacterial species identified from the termite gut microenvironment.

Microbe Id	Strain information	Reactions	Gene	Metabolites	In silico growth rate	Experimental growth rate
						Mean	SD
P1	Micrococcus luteus P1	1120	1172	1087	0.188	0.179	0.0137
P2	Kluyvera sp. P2	1757	1996	1545	0.147	0.166	0.0089
P3	Cellulomonas denverensis P3	1260	1242	1235	0.085	0.0840	0.007
P4	Oceanobacillus sojae P4	1559	2446	1348	0.069	0.075	0.0098
P5	Brevibacterium sp. P5	1489	1562	1314	0.198	0.208	0.124
P6	Niallia circulans P6	1577	1942	1355	0.113	0.126	0.0068

To examine the metabolic activities and flux distribution of each bacterial pair, the individual GEMs of these six bacterial species have been integrated into 15 (⁶C₂) possible pairwise GCMs. The individual GEMs were represented as specific compartments in the GCM, where the community biomass was defined by summing the stoichiometric coefficients of the individual model's biomass reactions. The FBA has been performed with each community model, subjected to the maximization of the community biomass formation rate. In each pairwise GCM, an individual microbe can enhance, decrease, or have no effect on the growth of fellow community members when growing together. Therefore, a PGSI calculation has been introduced to assess how the individual microbial growth is altered in the presence of another microbe, ultimately affecting the community growth rate. Thus, considering two bacterial species (P and Q) growing together, if the community biomass is greater than the sum of the individual growth ($V_{\mathit{bioPQ}}\ge V_{\mathit{bioP|P}}+V_{\mathit{bioQ|Q}}$), the bacterial pair can be called growth‐compatible. Hence, PGSI analysis will help track the positive, negative, and neutral interactions between individual entities, leading to enhanced, decreased, and unchanged community growth ($V\mathit{bioP|P}+V\mathit{bioQ|Q}$), respectively. The growth compatibility assessment of 15 pairwise GCMs showed 8, 3, and 4 positive, negative, and neutral PGSI, respectively (Figure 3A). For the 8 positively interacting bacterial pairs, an average PGSI of 0.189 has been observed. These growth‐compatible bacterial pairs show mutualistic (+/+) and commensalic (+/0) interactions, leading to a higher (> 1%) community growth (V _bioPQ ) than the sum of individual GEMs' growth (V _bioP|P + V _bioQ|Q ). For instance, the C. denverensis P3 and O. sojae P4 showed an average PGSI of 0.65 in the community model GCM_P3‐P4 (Figure 3B) (Table S6). Because of this high average PGSI and mutualistic interactions between these bacterial strains, the community model growth has increased by 69.5%. On the other hand, bacterial pairs in GCM_P2‐P3 (PGSI = −0.249), GCM_P2‐P4 (PGSI = −0.173), and GCM_P2‐P6 (PGSI = −0.049) showed negative growth support on each other (Figure 3C). This negative PGSI leads to competitive (−/−) and amensalism (−/0) interactions in the community, where the growth has decreased up to 24.36% compared with the individual GEMs’ growth rate. Apart from the positive and negative growth impact, a neutral effect has also been noted for GCM_P1‐P3, GCM_P1‐P6, GCM_P5‐P6, and GCM_P4‐P6, where no significant difference (<1%) has been observed in the community environment. Hence, the PGSI analysis helps assess positive, negative, and neutral effects on the inter‐microbial growth rate in the community to track the growth‐compatible bacterial pairs. Overall, the nine positive and four neutral inter‐microbial interactions have been revealed through the pairwise growth compatibility assessment of the GCMs. These growth‐compatible microbial pairs can have metabolic synergisms that help support the growth of fellow community members and enhance their metabolic potential.

Figure 3.

Assessment of in silico community growth profile and PGSI. (A) Comparison between the pairwise community model growth (blue) and the sum of the individual (grey) model growth as obtained from the FBA analysis. A higher (> 1%) community growth rate compared to the sum of individual growth signifies the in silico growth compatibility between nine bacterial pairs. (B) The positive growth support (PGSI) between strains P3 and P4 leads to mutualistic interactions. Meanwhile, the negative PGSI between P3 and P2 (C) is responsible for inter‐microbial competition.

3.3. Assessment of the Lignocellulolytic Potential of the Model‐Assisted Bacterial Coculture System

Alongside the pairwise growth pattern, the metabolic benefits a bacterial species receives from its fellow community members are essential for developing synergistic relations. The metabolic compatibility between microbial entities has been measured through PMA. PMA (P | P∪Q) captures the number of flux‐range‐expanded reactions and activated reactions in an individual microbe P while growing in a duplet with Q (see Section 2). Thus, PMA (P | P∪Q) = 1 represents 100% beneficiary effect of microbe Q on P while growing together. Conversely, PMA (P | P∪Q) = 0 indicates that microbe Q does not provide any metabolic benefits to microbe P. The PMA has been calculated from all 15 GCMs and represented as an interactive network in Cytoscape (Figure 4). The average of the calculated PMA (~ 0.075) has been set as the cut‐off to identify the significant metabolic assistance between microbial pairs. The O. sojae P4 and C. denverensis P3 strains provided the maximum metabolic assistance (up to 0.16) toward different community members (Figure 4) (Table S7). In the community model GCM_P1‐P3, C. denverensis P3 provided a PMA of 0.15 towards M. luteus P1, whereas M. luteus P1 showed a PMA of 0.10 towards C. denverensis P3. The flux range of the lignocellulose degradation reactions was also increased in the GCM_P1‐P3 (Figure S3). Hence, with a high average PMA of 0.126, GCM_P1‐P3 can be considered a beneficial pairwise community where bacterial members provide positive metabolic assistance to each other. The pairwise community models GCM_P3‐P5, GCM_P3‐P6, and GCM_P1‐P4 also showed a significant PMA, which is higher than the average PMA values (>0.075), considering all 15 pairwise community models (Table S7). Interestingly, for the nine pairwise community model with higher PMA, we observed an expansion in the flux range associated with cellulose conversion reactions and positive PGSI (Figure S3). Therefore, growth compatibility and metabolic assistance are essential for establishing the synergistic relationship between bacterial pairs. This may also indicate that the collaborative degradation of the cellulose can be one major factor in establishing the synergistic relations between bacterial pairs. Based on in silico analysis, seven pairwise bacterial communities (GCM_P1‐P3, GCM_P1‐P4, GCM_P1‐P5, GCM_P3‐P4, GCM_P3‐P5, GCM_P3‐P6, GCM_P4‐P5) showed positive PGSI, and two communities (GCM_P1‐P3 and GCM_P5‐P6) exhibited neutral growth assistance with a high PMA value (Table 2).

Figure 4.

Pairwise metabolic assistance between individual microbes. The network represents the inter‐microbial metabolic assistance in the pairwise bacterial community as mapped using Cytoscape. Each node denotes the specific bacterial strains, and the colored edges represent the fraction of PMA value directed toward the metabolically dependent species from fellow community members.

Table 2.

Pairwise growth support index (PGSI) and metabolic assistance (PMA) of nine compatible bacterial pairs that have been tested with experimental coculture systems. The growth compatibility and enzymatic synergism (DES) derived for the experimental setup have also been presented.

Community model	Strain 1	Strain 2	Avg. PGSI	Avg. PMA	Coculture systems	Sum of individual growth rate (h−1)	SD	Community growth (h−1)	SD	DES
GCMP1‐P3	Micrococcus luteus P1	Cellulomonas denverensis P3	0.006	0.126	COMP1‐P3	0.263	0.010	0.291	0.023	1.106
GCMP1‐P4	Micrococcus luteus P1	Oceanobacillus sojae P4	0.307	0.097	COMP1‐P4	0.254	0.004	0.197	0.014	0.776
GCMP1‐P5	Micrococcus luteus P1	Brevibacterium sp. P5	0.031	0.069	COMP1‐P5	0.387	0.022	0.384	0.018	0.992
GCMP2‐P5	Kluyvera sp. P2	Brevibacterium sp. P5	0.011	0.054	COMP2‐P5	0.292	0.011	0.301	0.026	1.031
GCMP3‐P4	Cellulomonas denverensis P3	Oceanobacillus sojae P4	0.647	0.108	COMP3‐P4	0.159	0.016	0.157	0.014	0.987
GCMP3‐P5	Cellulomonas denverensis P3	Brevibacterium sp. P5	0.207	0.124	COMP3‐P5	0.292	0.017	0.371	0.04	1.271
GCMP3‐P6	Cellulomonas denverensis P3	Niallia circulans P6	0.289	0.125	COMP3‐P6	0.21	0.010	0.296	0.007	1.410
GCMP4‐P5	Oceanobacillus sojae P4	Brevibacterium sp. P5	0.015	0.089	COMP4‐P5	0.283	0.013	0.321	0.031	1.134
GCMP5‐P6	Brevibacterium sp. P5	Niallia circulans P6	0.006	0.065	COMP5‐P6	0.334	0.019	0.39	0.009	1.168

To validate the growth and metabolic compatibility predicted by the pairwise GCM, nine bacterial coculture systems, i.e., COM_P1‐P3, COM_P1‐P4, COM_P1‐P5, COM_P2‐P5, COM_P3‐P4, COM_P3‐P5, COM_P3‐P6, COM_P4‐P5, and COM_P5‐P6 have been tested experimentally. The endoglucanase, total cellulase, and xylanase activity of the model‐driven bacterial pairs have been assessed by conducting coculture with CMC, FP, and xylan, respectively. Among the nine bacterial pairs selected through the GCM analysis, seven pairs showed growth compatibility in the experimental coculture system. Hence, the GCM analysis showed a 78% accuracy in predicting synergistic growth patterns in pairwise bacterial communities. For the growth‐compatible bacterial pairs, the community growth rate either increased or remained unchanged compared to the sum of individual growth (Table 2). For example, a higher PGSI ranging from ~0.01 to 0.65 has been observed in the community models GCM_P1‐P3, GCM_P3‐P4, and GCM_P3‐P5. When the coculture systems (COM_P1‐P3, COM_P3‐P4, and COM_P3‐P5) of these bacterial pairs were developed, they showed a 9.03%–26.75% increase in experimental community growth rate compared to individual growth. Therefore, for the growth‐compatible bacterial pairs, the model‐driven PGSI showed a positive correlation (Pearson r = 0.86, p = 0.01) with the experimentally measured growth rate increment in the coculture (Figure 5A).

Figure 5.

Correlation between in silico analysis and experimental data: (A) the pairwise growth support index of the pairwise GCMs showed a good correlation with the experimental community growth. (B) Similarly, the model‐derived PMA showed a good agreement with the experimentally measured degree of enzymatic synergisms.

This community‐level growth compatibility can affect the metabolic synergisms (PMA) of microbial pairs, ultimately affecting the coculture systems' enzyme activities. Therefore, how inter‐microbial growth compatibility (PGSI) and metabolic assistance (PMA) act as critical factors in facilitating the DES have been investigated through coculture systems. The DES represents the ratio of the enzyme activity in coculture to the cumulative sum of the monoculture enzyme activities. The bacterial coculture showed diverse activity at different time points (Figure S4) towards different lignocellulosic substrates, i.e., FP, CMC, and xylan. For instance, COM_P3‐P5 (C. denverensis P3/Brevibacterium sp. P5) and COM_P5‐P6 (Brevibacterium sp.P5/Niallia circulans P6) are the best‐performing communities in FP, which showed the maximum enzymatic activity (FPase) of 0.429 U mL⁻¹ (±0.06) and 0.411 U mL⁻¹ (±0.055), respectively (Figure 6A, Table S8). The community enzyme activity has been increased by 53%, with the highest DES of 1.53 between C. denverensis P3 and Brevibacterium sp. P5 in the community environment while growing in FP (Figure 6D, Table S9). On the other hand, when CMC was used as a substrate, COM_P1‐P3 (M. luteus P1/C. denverensis P3) and COM_P3‐P6 (C. denverensis P3/N. circulans P6) showed the maximum enzyme activity of 0.388 U mL⁻¹ (±0.041) and 0.45 U mL⁻¹ (±0.03), respectively (Figure 6B). The COM_P1‐P3 also showed good degradation ability in xylan [0.32 U mL⁻¹ (±0.014)], where the community enzyme activity was increased by 42% (DES 1.42) compared to the respective monocultures (Figure 6C,D). Consequently, M. luteus P1 and C. denverensis P3 also possess a high PMA of 0.126, as predicted through the community model (GCM_P1‐P3) analysis. Therefore, the experimental enzymatic synergies agree with the model‐predicted metabolic assistance between M. luteus P1 and C. denverensis P3. Overall, the model‐predicted PMA of the pairwise communities showed a good correlation (Pearson r = 0.86, p = 0.034) with the DESs in the seven growth‐compatible bacterial cocultures (Figure 5B). Hence, the in silico parameters showed good prediction capabilities in identifying the beneficial bacterial pairs. Overall, six pairwise communities, i.e., COM_P1‐P3, COM_P3‐P4, COM_P3‐P5, COM_P3‐P6, COM_P4‐P5, and COM_P5‐P6, have shown significant enzymatic synergies and increased community activity in one of the given substrates. Therefore, assessing the newly defined parameters of PGSI and PMA helps screen the synergistic consortia by reducing the number of testable bacterial combinations.

Figure 6.

Comparative analysis of the community enzyme activity and DES. The alterations in the FPase (A), CMCase (B), and xylanase (C) activities in the 9 pairwise coculture (horizontal axis) have been compared to the sum of the individual activity in the vertical axis. The higher activity in the coculture signifies the enzymatic synergism between microbial pairs. (D) The horizontal axis represents the degree of enzymatic synergism (DES) value of the 9 bacterial pairs (vertical axis) in different lignocellulosic substrates. All the data points represent the mean value from the triplicate experimental measurements, and the error bar represents the standard deviation of each measured data.

However, it has been observed that the best‐performing community on one substrate sometimes showed lower degrading potential on other substrates. For example, although COM_P1‐P3 showed efficient enzyme activity in CMC and xylan, it has shown very low degradation potential [0.28 U mL⁻¹ (± 0.01)] and DES (1.05) in FP (Figure 6D). Conversely, in FP (FPase), COM_P3‐P5 and COM_P5‐P6 were the best‐performing communities, with maximum DES activity of 1.53 and 1.27, respectively. However, despite demonstrating high FPase activity, COM_P5‐P6 underperforms in CMC, leading to a 17.67% decrease in activity (Figure 6D). Hence, further optimization of community composition and design in larger communities is required to achieve consistent degradation capabilities across all of the provided lignocellulosic substrates (FP, CMC, and xylan). The knowledge of growth and enzymatic compatibility derived from genome‐scale models and coculture experiments is essential for designing larger communities with the best‐performing candidates. Here, fellow community members can compensate for a particular strain's enzymatic limitation for efficient substrate utilization. Therefore, we have extended the proposed community modeling pipeline and flux‐based parameters to measure the biochemical and growth compatibilities in larger communities, thereby identifying compatible microbial combinations.

3.4. Designing GCM‐Guided Synergistic Consortia for Lignocellulose Degradation

The growth compatibility (PGSI) and metabolic assistance (PMA) derived from the community models and the development of bacterial coculture help to screen six synergistic bacterial pairs with lignocellulose degradation potential. Assessing the enzymatic activities of these model‐driven bacterial cocultures helps identify different best‐performing bacterial pairs in FP, CMC, and xylan. Overall, the pairwise consortia, COM_P3‐P5, COM_P4‐P5, and COM_P5‐P6, performed best in FP. On the other hand, COM_P1‐P3, COM_P3‐P4, COM_P3‐P5, and COM_P3‐P6 showed maximum degradation potential in CMC and xylan. Hence, to develop a more robust community that can degrade all the types of lignocellulosic substrates, three‐membered bacterial consortia have been developed by combining six best‐performing individual strains (P1‐P6) identified earlier. Combining these six individual strains can generate 20 distinct combinations (⁶C₃) of three‐membered community compositions. Therefore, a total of 20 three‐membered GCMs have been reconstructed to test the growth compatibility of microbial consortia with the PGSI parameter. Out of these 20 three‐membered GCMs, 7 models exhibited positive growth support, with an average PGSI ranging from 0.01 to 0.327. On the other hand, neutral and negative growth support have been observed for 5 and 8 community models, respectively (Table S10). The assessments of metabolic compatibility also showed good community metabolic assistance (CMA > 0.1) in the growth‐compatible GCMs. Among the 7 growth‐compatible communities predicted through the PGSI, 4 three‐membered bacterial consortia, i.e., COM_P1‐P3‐P4, COM_P1‐P3‐P5, COM_P3‐P4‐P5, and COM_P3‐P5‐P6 were tested experimentally to determine their synergistic lignocellulolytic ability.

The lignocellulosic bio‐conversion potential of these model‐driven three‐membered bacterial consortia has been tested experimentally with CMC, FP, and xylan as metabolic substrates. Overall, among the four compatible bacterial communities tested experimentally, two bacterial consortia, i.e., COM_P1‐P3‐P5 and COM_P3‐P5‐P6, have shown robust enzymatic degradation potential in all the provided lignocellulosic substrates. For instance, the three‐membered community model GCM_P1‐P3‐P5 showed an average PGSI of 0.116 between M. luteus P1, C. denverensis P3, and Brevibacterium sp. P5 (Table S11). Moreover, the good community metabolic assistance (CMA = 0.095) between these microbial species may also help to increase the collaborative metabolic and enzymatic potential of COM_P1‐P3‐P5. When this synergistic bacterial consortium (COM_P1‐P3‐P5) was developed experimentally, it showed a maximum enzymatic activity of 0.305 U mL⁻¹ (±0.041), 0.52 U mL⁻¹ (±0.068), and 0.459 U mL⁻¹ (±0.046) for FP, CMC, and xylan, respectively (Figure 7A). Hence, the GCM‐predicted growth and metabolic synergisms help to increase the enzymatic degradation potential of these bacterial consortia by lowering the substrate‐dependent limitation of the individual bacterial strains. During the monoculture experiment, the individual strains P1 and P5 showed moderate enzymatic activities [(0.1 U mL⁻¹) (±0.012)] in CMC. However, when they had grown in the three‐membered consortia COM_P1‐P3‐P5 with C. denverensis P3, the community showed a more proactive (~30% higher) enzymatic degradation potential against CMC. Therefore, the substrate‐specific limitation of the individual bacterial strains P1 and P5 has been covered by fellow community members (P3) in the consortia and provides a higher metabolic potential for efficient lignocellulose bio‐conversion. Compared to the pairwise coculture systems, the three‐membered consortium COM_P1‐P3‐P5 performed 15.6%, 30%, and 9.5% higher FPase, CMCase, and xylanase activities, respectively.

Figure 7.

The FP, CMC, and xylan degradation potential of three‐membered bacterial consortia: The FPase (black), CMCase (red), and xylanase (blue) activity of the bacterial consortia (A) COM_P1‐P3‐P5 (B) COM_P3‐P5‐P6 at different time point.

Moreover, the model‐driven information about the bacterial synergisms also helps to elevate the combined enzymatic activity of the individual strains in the experimentally developed consortium COM_P3‐P5‐P6. For instance, the individual strains, i.e., P3, P4, and P6 in COM_P3‐P5‐P6 exhibited good xylanolytic potential [(~0.2 U mL⁻¹) (±0.032)] in monoculture and pairwise coculture. When these three well‐performing strains were combined in COM_P3‐P5‐P6, the bacterial consortia showed the highest xylanase activity of 0.617 U mL⁻¹ (±0.088) (Figure 7B).

Overall, in the defined microbial consortium COM_P3‐P5‐P6, the FP, CMC, and xylan bioconversion potential were increased by 44.1%, 8.4%, and 3.5%, respectively. Consequently, the model‐predicted (GCM_{P3‐P5‐P6)} growth support index (PGSI = 0.22) and metabolic assistance (CMA = 0.13) were also found to be high in this defined consortium. Therefore, assessing the newly defined flux‐based parameters helps screen the synergistic consortia by reducing the number of testable bacterial combinations. The increase in the number of microbial species can exponentially elevate testable microbial combinations for the development of synthetic microcosms. Handling these large sets of microbial combinations and assessing their metabolic interlinks is challenging as well as highly time‐consuming. Hence, the model‐guided coculture system design approach aids in formulating the optimal bacterial composition with high enzymatic synergisms and growth compatibility while reaching higher activity on the provided lignocellulosic substrates. The community modeling method developed here can help manipulate the microbial community composition and metabolic activities to aid novel experimental strategies for bioengineering applications.

4. Discussion and Conclusion

Implementation of the genome‐scale modeling integrated with the conventional microbial coculture helped us develop a complete strategy for designing synergistic microbial consortia for lignocellulose bioconversion. Initially, the individual bacterial isolates (P1–P6) from the termite gut showed varied enzyme activities with substrate‐specific limitation in the provided lignocellulosic materials, i.e., CMC, FP, and xylan. The flux distribution via active biochemical pathways was assessed by constructing individual GEMs for the six best‐performing lignocellulolytic microbial strains. Further, the individual GEMs were integrated to reconstruct genome‐scale community metabolic models (GCMs) for 15 possible bacterial pairs. The maximization of biomass formation has been set as the objective function of the individual GEMs and pairwise community model. Cellulose (CMC and FP) and xylan have been used as metabolic substrates while optimizing the metabolic model. Therefore, the optimization of biomass formations will ultimately lead to the enhancement of substrate uptake or the breakdown of cellulose or hemicellulose. We have incorporated the extracellular degradation reactions of cellulose into the GEMs based on gene and biochemical information, as the cell cannot transport cellulose directly. For instance, we have incorporated reactions associated with endoglucanase, which converts cellulose to cellodextrin, exoglucanase, which converts cellodextrin to cellobiose, and beta‐glucosidase, which converts cellobiose to glucose, into the GEMs. Hence, the optimization of the biomass function provides essential information about the flux distribution patterns in the individual GEMs and pairwise community model while degrading the lignocellulosic substrate.

Implementing the flux‐based parameters, namely PGSI and PMA, helps evaluate pairwise growth compatibility and metabolic assistance through GCM analysis. The PGSI and PMA‐driven growth and metabolic compatibility of the bacterial pairs were found to be essential for a mutualistic bacterial consortium. A total of 9 metabolically compatible bacterial pairs predicted from the in silico GCM analysis have been validated by designing experimental coculture systems. The pairwise community model with novel flux‐based parameters accurately predicted growth and metabolic synergisms in seven bacterial cocultures for lignocellulosic bio‐conversion, demonstrating 78% accuracy. Moreover, the PGSI derived through the GCMs analysis correlated well (Pearson r = 0.86, p = 0.012) with the experimentally measured growth rate enhancement in bacterial coculture. The enzymatic activities of these bacterial cocultures have been tested with the FP, CMC, and xylan. Compared to the bacterial monoculture, the pairwise communities showed up to 53% higher activities in one of the given substrates (FP, CMC, and xylan). Interestingly, the model‐derived PMA of the bacterial communities showed good agreement (Pearson r = 0.79, p = 0.034) with the DESs in the coculture systems, which signifies the reliability of the in silico assessment of the metabolic synergisms.

Furthermore, we combined the six individual strains to form three‐membered communities, which enhanced the bacterial lignocellulolytic ability. To evaluate the growth compatibility among all the possible bacterial combinations, 20 (⁶C₃) three‐membered GCMs have been reconstructed. The assessment of the PGSI and CMA parameters facilitated the screening of seven compatible three‐membered bacterial communities among 20 possible combinations. The model‐predicted bacterial synergisms were further tested by developing compatible bacterial consortia with experimental setups. Here, the microbial consortia comprising C. denverensis P3, Brevibacterium sp. P5, and N. circulans P6 (COM_P3‐P5‐P6); and M. luteus P1, C. denverensis P3, and Brevibacterium sp. P5 (COM_P1‐P3‐P5) exhibited robust degradation potential and higher enzyme activities in FP, CMC, and Xylan. Therefore, the model‐driven evaluation of bacterial metabolic assistance and growth compatibility facilitates the design of synergistic bacterial consortia with enhanced lignocellulolytic enzyme activities. Overall, the proposed community modeling strategy and flux‐based assessment of the microbial synergisms can be extended to measure biochemical and growth compatibilities in larger communities, thereby identifying compatible microbial combinations. The model‐predicted synergistic microbial combination can be tested further with a lesser experimental hurdle. We can further enhance the model‐assisted microbial consortia by optimizing the microbial ratio in the seed culture or by incorporating new microbial members into the established consortia. This could pave the way for the large‐scale implantation of synthetic microbial consortiums in bioproduction.

Author Contributions

P.K. and A.G. conceived the study. P.K. and A.G. executed the model analysis and interpretations. P.K. executed the coculture experimental procedure and enzymatic assay. P.K. and A.G. wrote the manuscript. A.G. supervised the whole study. Both authors read and approved the final version of the manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting information.

Supporting information.

Supporting information.

Data Availability Statement

The data that supports the findings of this study are available in the Supporting Information material (data files) of this article. The associated models and codes are available in GitHub (https://github.com/itsamit/GCM).