Model

Our computational workflow integrates sequence retrieval, ecological metadata, multiple sequence alignment, and phylogenetic reconstruction. Its aim is to analyse the convergent evolution of chitin degradation and identify robust chitinase candidates for heterologous expression.

This step aims to compile a curated, family‐labelled dataset of chitinase sequences for downstream phylogenetic and ecological analysis. Public database records were searched, cleaned, and classified into glycoside hydrolase families (GH18, GH19, …), followed by validation and summary visualisation.

Search and retrieval of Records from ENA

Initial retrieval of chitinase coding sequences from the European Nucleotide Archive (ENA) using keyword- and EC-based queries provided only partial coverage, insufficient to capture the full diversity of chitinolytic enzyme families. While ENA offers extensive nucleotide resources, annotation inconsistencies and limited query sensitivity restricted the recovery of sequences suitable for a comprehensive phylogenetic analysis.

Bibliographic identification of chitinase GH families

To address these limitations, we adopted a literature-guided framework. Curated reviews and reference databases (7–10) were used to identify the major glycoside hydrolase (GH) families associated with chitinolytic activity (GH18, GH19, GH23, GH48, GH75, GH80). This approach enabled targeted retrieval of representative sequences across taxa, ensuring broader coverage than repository searches alone.

This integrative strategy reflects a recurring challenge in enzyme evolutionary studies: the need to combine public database queries with curated bibliographic resources in order to overcome annotation gaps and achieve systematic representation of enzyme families (8,11).

This bibliographic framework defined the six GH families of interest and guided the targeted retrieval of ENA records. The outcomes of this retrieval, along with the subsequent curation and standardization of the associated ecological metadata, are presented in Section 3.

Based on the six glycoside hydrolase (GH) families identified from the literature (GH18, GH19, GH23, GH48, GH75, GH80), we retrieved corresponding records from the European Nucleotide Archive (ENA). Queries were formulated using family names and associated EC numbers, with the objective of capturing BioSample- and BioProject-linked entries rather than focusing solely on raw nucleotide sequences. Retrieved records included taxonomic identifiers, organism names, and ecological metadata fields such as country of origin, geographic coordinates, host information, isolation source, and collection date when available.

In addition to basic sequence descriptors, we maximized metadata retrieval by including all fields with potential ecological or phylogenetic relevance provided by the ENA API. These encompassed taxonomic information (scientific name, taxonomic identifier, complete lineage), ecological and environmental descriptors (country, location, altitude, isolation source, environmental sample status, environment biome, environment feature, environment material), and host-related information (host species and taxon identifier, host developmental stage, host sex, host tissue, host disease). Provenance and sampling-related fields were also retrieved, including collection dates (start, end, and standardized date), names of collectors and identifiers, as well as strain-level descriptors (strain, isolate, sub-species, sub-strain, cultivar, ecotype, variety). Where available, voucher information, culture collection identifiers, and links to study or sample accessions were also included. This exhaustive metadata set was designed to ensure that ecological and phylogenetic signals associated with chitinolytic enzymes could be captured with maximum resolution, while maintaining compatibility with ENA’s standardized field architecture.

To operationalize the retrieval process, we implemented a dedicated script (fetch.sh) that automates query submission to the ENA API for each of the six glycoside hydrolase families. For every family, the script generates a tab-delimited output file containing the full set of selected metadata fields, ensuring standardized and reproducible data collection across enzyme families. In parallel, we developed a companion script (fetch_fields.sh) to retrieve the authoritative list of available metadata fields for the sequence result type from the ENA API (returnFields endpoint). This script archives the complete field schema (sequence_fields.tsv), thereby capturing the exact metadata framework in use at the time of retrieval. Together, these scripts provide both the chitinase-related datasets and the contextual metadata architecture, ensuring that subsequent ecological and phylogenetic analyses are conducted against a fully documented and reproducible framework. Both scripts are made publicly available in the team’s GitHub repository.

In order to evaluate the completeness and diversity of the retrieved datasets, we implemented a Python script (gh_family_summary.py) that parsed the raw ENA metadata tables. For each GH family, the script calculated the total number of records, the count of distinct species represented, the number of environmental samples, and the number of distinct hosts. This summary step provided a quick quality check of the datasets, allowing us to confirm coverage and identify potential biases (e.g., overrepresentation of environmental samples in GH48).

To ensure reproducibility and transparency, we implemented the main analytical tasks as standalone scripts, deposited in the project’s public GitHub repository. Each script corresponds to a discrete step of the workflow, such as data retrieval, sequence alignment, and phylogenetic tree reconstruction, and produces standardized intermediate files for downstream use.

Retrieval of family-specific FASTA datasets (fasta_fetch.sh)

The first script automated the retrieval of nucleotide sequences from the ENA Portal and Browser APIs. For each target GH family (GH18, GH19, GH23, GH48, GH75, GH80), it:

1. Queried ENA with a description-based search to obtain accession identifiers.
2. Downloaded the corresponding FASTA sequences in bulk using the ENA Browser API.
3. Generated paired .tsv files (containing accessions) and .fasta files (containing sequences) for each GH family.

Empty lines were removed and sequence headers standardized. The result was a curated set of FASTA files, one per family, serving as the input for subsequent alignment and distance matrix construction.

Multiple sequence alignment and distance matrices (gh_fasta_msa.R)

The second script processed each GH-family FASTA file to produce pairwise identity distance matrices. The workflow included:

• Reading nucleotide sequences into R (Biostrings),
• Performing multiple sequence alignment with the msa package,
• Converting the alignment into seqinr format,
• Computing pairwise identity distances (seqinr::dist.alignment).

For every input FASTA, the script produced a <family>_matrix.tsv file containing the pairwise distance matrix. Hierarchical clustering dendrograms were also generated to provide preliminary visualizations of the relationships between sequences.

Visualization of dendrograms and NJ trees (plot_trees_pretty.R)

The third script consumed the <family>_matrix.tsv files and the corresponding raw metadata tables (<family>.tsv). Its functionality included:

• Robust reading and cleaning of distance matrices,
• Mapping accession identifiers to species/organism names (where metadata available),
• Label truncation to preserve readability,
• Construction of hierarchical clustering dendrograms and Neighbor-Joining trees (using the ape package),
• Export to standardized PDF figures (<family>_matrix_dendrogram.pdf, <family>_matrix_nj.pdf).

To avoid clipping, plotting parameters were adjusted with extended canvas widths and margins, ensuring that accession labels and scale bars remained fully visible across all families.

The search and classification process identified six glycoside hydrolase (GH) families with established chitinolytic activity: GH18, GH19, GH23, GH48, GH75, and GH80. Among these, GH18 and GH19 were the most abundant and broadly distributed, consistent with their widespread occurrence across bacteria, fungi, plants, and animals (7,9).

GH23, GH48, GH75, and GH80 were less represented but correspond to distinct enzymatic functions reported in the literature, including lysozyme-like chitinases (GH23), processive bacterial chitinases (GH48), chitin deacetylases (GH75), and exo-chitinases (GH80) (8,10).

In addition to these six major families, other GH families have been occasionally reported with marginal or secondary chitinolytic activity (e.g. GH20, GH46, GH73, GH85). These were not included in the present analysis due to their limited representation and lower functional significance compared to the six core chitinase families (7,8).

This analysis was carried out by iGEM MBG-Duth as part of the team’s systematic approach to mapping the enzymatic landscape of chitin degradation.

The bibliographic framework and ENA record search yielded datasets of varying size and composition across the six major GH families. GH48 was the most represented, with 517 records, the majority of which (491) originated from environmental samples, while species diversity was relatively limited (18 distinct species, 6 host taxa). GH18 also showed broad representation (121 records), spanning 47 distinct species and 7 host taxa, including both environmental and host-associated samples.

GH19, though smaller in size (52 records), displayed notable taxonomic breadth, encompassing 32 species and 7 hosts, suggesting recurrent emergence of this family across distinct lineages.

The remaining families (GH23, GH75, and GH80) were comparatively underrepresented, each with ~20 records. GH23 contained 22 records distributed across 16 species and 2 hosts, while GH75 (21 records) included 7 species and 2 hosts, with approximately half derived from environmental samples. GH80 was the smallest dataset (21 records), restricted to 5 species and a single host, highlighting its rarity relative to the other families.

Overall, these results underscore the strong dominance of GH18 and GH48 in public repositories, contrasted with the restricted distribution of GH23, GH75, and GH80. This uneven representation reflects both true biological prevalence and potential sampling biases in environmental sequencing efforts.

Analysis of the ecological metadata revealed distinct distributional patterns among the six GH families. GH18 emerged as the most ecologically widespread family, with records spanning diverse environments such as soil, rumen, rivers, lakes, and forests, and hosts ranging from humans to ruminants and crustaceans.

GH19 showed a more plant-associated profile, frequently linked to crops such as Solanum lycopersicum and Zea mays, yet it was also detected in clinical samples (plasma and serum), highlighting a dual association with both plant and human systems.

In contrast, GH23 and GH80 displayed a strong clinical bias, with a high proportion of records linked to human hosts and medical samples, consistent with a pathogenic or opportunistic niche. GH48 was overwhelmingly associated with rumen environments, with the majority of sequences originating from cow rumen, underscoring its specialization in gut-associated cellulose and chitin degradation.

GH75, although represented by a smaller dataset, showed a balanced distribution across soil environments and a few clinical sources, suggesting intermediate ecological breadth. Temporal coverage of the dataset extended from 2002 to 2024, offering a two-decade perspective on the sampling of chitinolytic enzymes, while isolated cases such as GH18 sequences from high-altitude ecosystems (2,808 m) further emphasize the ecological range of this family.

Together, these patterns illustrate a spectrum from generalist families (GH18) to highly specialized ones (GH48, GH80), with implications for both environmental and clinical contexts of chitinase activity.

The phylogenetic trees reconstructed for the six GH families revealed distinct clustering patterns. GH18 displayed broad diversification, with sequences forming multiple well-supported clades rather than a single lineage.

GH19 showed tighter clustering, with several well-defined groups and overall shorter branch lengths. GH23 and GH80 exhibited strong clustering with limited internal divergence, reflecting more compact phylogenetic structures.

GH48 sequences formed a large but dense clade, characterized by limited branching depth despite the high number of sequences included. GH75, although represented by a smaller dataset, demonstrated intermediate clustering, with several distinct subgroups but without the extensive diversification observed in GH18.

Taken together, these qualitative observations underscore a clear contrast among GH families, with GH18 showing extensive diversification, GH48 and GH80 displaying compact clustering, and GH19, GH23, and GH75 occupying intermediate positions.

The ecological distribution patterns observed across GH families provide practical insights for enzyme selection in a synthetic biology context such as iGEM MBG-Duth. Families with broad ecological presence, like GH18, represent robust candidates due to their demonstrated versatility across environments, hosts, and ecological niches. Such generalist profiles suggest functional robustness and adaptability, desirable traits for engineered biological systems.

In contrast, more specialized families such as GH48, predominantly linked to rumen ecosystems, may offer highly efficient but context-specific activities that could be harnessed in targeted applications, provided their ecological bias is acknowledged. Similarly, families with clinical associations (GH23, GH80) highlight potential biosafety considerations that should be carefully evaluated before integration into synthetic biology projects.

By situating enzyme choice within this ecological framework, our study directly supports the difficult decision-making process inherent in iGEM MBG-Duth projects, where teams must balance functional performance, ecological relevance, and biosafety constraints when selecting chitinolytic enzymes for deployment.

The phylogenetic reconstructions of the six GH families reveal marked differences in diversification patterns that are highly informative for understanding the evolutionary dynamics of chitinases. GH18 stands out as the most diversified family, forming numerous well-supported clades that indicate deep evolutionary branching and functional divergence. In contrast, GH19, GH23, GH75, and GH80 display more compact topologies, with tighter clustering and shorter branch lengths, consistent with narrower taxonomic distributions and more specialized functions.

GH48 represents an intermediate but particularly notable case: despite its very large dataset, the family exhibits dense clustering with limited branching depth, a pattern suggestive of recent gene duplications or horizontal gene transfer rather than gradual diversification. Collectively, these phylogenetic patterns indicate that while some chitinase families (notably GH18) have undergone broad and ancient diversification, others (GH23, GH80) remain relatively conserved, and still others (GH48) appear shaped by recent ecological expansions. Such variation underscores that chitinolytic capacity is not the result of a single evolutionary trajectory, but rather has emerged through multiple, independent lineages across the tree of life.

The patterns revealed in our phylogenetic analysis align with previous findings that chitinolytic enzymes are ecologically widespread and often found in unrelated lineages (1,2,12). However, unlike earlier work that examined convergence within specific taxonomic groups (4) or lifestyle-associated diversification in fungi (5), our approach contextualizes phylogenetic relationships with ecological metadata, enabling comparisons of functional traits across habitat types. This integration helps to highlight potential selective pressures, such as substrate availability or interspecies interactions, that may drive the repeated emergence of similar enzymatic functions.

Notably, the phylogenetic incongruences we observed cannot be unambiguously attributed to convergent evolution without considering the role of horizontal gene transfer. Previous research on GH19 chitinases provides clear examples of such transfer events between bacteria and plants (3), underscoring the complexity of disentangling these evolutionary processes. This suggests that future studies could benefit from whole-genome context and synteny analysis to better discriminate between convergence and horizontal gene flow.

The integration of ecological and phylogenetic data provides a more nuanced view of how chitinolytic capacity has emerged across the tree of life. The broad diversification of GH18, reflected both in its multiple clades and its wide ecological distribution spanning soil, aquatic, host-associated, and even high-altitude environments, suggests that this family represents a highly adaptable “generalist” lineage.

By contrast, GH48, despite its vast number of sequences, is ecologically restricted to rumen environments and phylogenetically compact, a pattern consistent with recent ecological expansion rather than ancient diversification. GH19 displays intermediate characteristics: while its phylogenetic topology is relatively compact, its ecological distribution across both plants and clinical samples indicates repeated recruitment into distinct functional niches, possibly mediated by horizontal gene transfer.

Families with a strong clinical bias, such as GH23 and GH80, exhibit both compact phylogenies and host-restricted distributions, consistent with specialized, opportunistic roles. GH75 occupies an intermediate position, showing modest diversification alongside balanced representation across soil and clinical environments.

Taken together, these findings highlight a spectrum of evolutionary strategies, from generalist diversification (GH18) to ecological specialization (GH48, GH23, GH80), with intermediate modes exemplified by GH19 and GH75. This combination of ecological breadth and phylogenetic structure underscores that the capacity to degrade chitin has not arisen through a single evolutionary route but instead through multiple, independent trajectories shaped by both ecological opportunity and lineage-specific constraints, a hallmark of convergent evolution.

Beyond the evolutionary patterns themselves, the analysis highlights practical attributes of chitinolytic enzymes relevant to synthetic biology. By identifying enzymes that exhibit ecological stability across varied habitats, our results suggest a subset of candidates with potentially greater robustness when expressed in heterologous systems such as Bacillus subtilis. This functional and ecological screening adds a layer of applied relevance to the phylogenetic framework, linking evolutionary modelling to concrete design decisions in the wet lab phase.

Furthermore, the broader narrative of convergent evolution provides a compelling scientific context that strengthens the educational and public engagement aspects of the iGEM project, positioning the work not only as a technical investigation but also as a contribution to science communication. In this way, the dry lab component serves both as a foundation for theoretical understanding and as a practical guide that directly informs and supports the engineering workflow.

In practical terms, this approach bridges the gap between abstract evolutionary analysis and tangible engineering choices, ensuring that candidate enzymes are selected not solely for their catalytic efficiency, but also for their proven adaptability and resilience in diverse environmental contexts. This alignment between evolutionary robustness and engineering feasibility represents a critical step towards reliable, application-ready biotechnological solutions.

The genes selected for our project (BBa_K4349000 – endochitinase, CfcI – exochitinase) belong to the GH18 family. This means that all the ecological and evolutionary patterns we observed for GH18, such as its presence across diverse environments, organisms, and hosts, are also valid for our chosen genes.

The endochitinase and exochitinase we study represent characteristic examples of GH18 enzymes. It is also worth noting that most functional chitinases applied in biotechnology, agribiology, and synthetic biology originate from the GH18 family.

By highlighting this link, our project further connects the dry-lab bioinformatic findings with the wet-lab synthetic biology design, emphasizing the direct relevance of ecological and evolutionary insights to practical applications (13).

Several methodological challenges accompany this type of large-scale bioinformatic analysis. Clear separation between different chitinase families (e.g., GH18, GH19) is essential to ensure valid comparisons, as merging distinct catalytic mechanisms may obscure true evolutionary patterns. Distinguishing between horizontal gene transfer and convergent evolution is particularly complex without supplementary genomic context, and technical limitations such as alignment inaccuracies, model selection biases, and the computational demands of phylogenetic reconstruction in large and diverse datasets can affect the reliability of inferred relationships. Nevertheless, with careful tool selection, stringent quality control, and appropriate dataset curation, these limitations can be mitigated.

A further challenge lies in the availability and quality of environmental metadata in public databases. Incomplete or inconsistent annotation can obscure ecological patterns and reduce confidence in cross-habitat comparisons. Expanding curated datasets and standardizing metadata collection would improve the reliability of ecological inference in future analyses.