Introduction

Hot springs serve as natural geothermal laboratories that offer unparalleled insights into microbial ecology, particularly in extremophile adaptation, metabolic plasticity, and early life evolution. The varying levels of dissolved ions and minerals give rise to a spectrum of water types, categorized as bicarbonate-rich, chloride-rich, sulfate-rich, or mixtures of two ions, such as sulfate-chloride-rich hot springs, as categorized in Pipet diagrams. Over the years, extensive research has focused on the microbiota in bicarbonate- and sulfate-rich hot springs from various countries1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26. Some countries may have both bicarbonate- and sulfate-rich hot spring sites due to variations in volcanic activity, geochemical inputs, hydrogen sulfide availability, limestone presence, and other local factors. Hot springs in Southern Asia, for instance along a geographic transect from southern Thailand through Peninsular Malaysia to Singapore tend to be neutral to alkaline (with the absence of sulfate-rich hot springs) and are typically characterized by high bicarbonate concentrations27. Geochemical variations play a crucial role in shaping planktonic microbial diversity, biofilm formation, and metabolic diversity, and microbial adaptation within these ecosystems16,23,28.

Microbial biofilms in hot springs exhibit stratified community structures and pigment-based coloration, which vary based on factors such as the primary energy source utilized by the microbial community. For example, photosynthetic biofilms, typically recognized by their green pigmentation, derive energy directly from sunlight. Although these photosynthetic biofilms are all dominated by Cyanobacteriota, geochemical variation among sites has led to differences in microbial composition and metabolic profiles. Other examples include biofilms reported in Sipoholon hot springs, Indonesia29; Octopus and Mushroom Springs, Yellowstone National Park (YNP), USA30,31; Nakabusa Spring, Japan32; Garga spring, Russia33; Sungai Klah, Malaysia19; and other hot springs worldwide6,34,35,36.

Non-photosynthetic biofilms are typically found in hot springs where temperatures exceed the upper limit for bacterial photosynthesis. High-temperature streamer biofilm communities (SBCs) are a prominent example, commonly observed at temperatures ranging from 70°C to near boiling in shallow thermal channels with turbulent water flow37,38. High-temperature SBCs were first documented in 196739. These biofilms are characterized by filamentous structures that are predominantly white or pink, though shades of grey or beige (tan) are also common3,4,5,6,7,8,11. High-temperature SBCs are often dominated by members of the phylum Aquificota, including Thermocrinis ruber, a pink filament-forming species isolated from the alkaline waters of the White Creek area in YNP40. The microbial compositions of high-temperature SBCs from various hot springs indicate that microbial communities are shaped by site-specific geochemical conditions37,41. Acidic, high-sulfide springs were enriched in Hydrogenobaculum spp., which oxidize molecular hydrogen and reduce elemental sulfur. Circumneutral to alkaline springs were dominated by Thermocrinis-like species in low-sulfide environments, which use reduced sulfur or molecular hydrogen as electron donors and oxygen as the terminal electron acceptor, or bySulfurihydrogenibium spp. in high-sulfide settings37. Furthermore, Sulfurihydrogenibium has been found to thrive in sites with elevated concentrations of dissolved carbon dioxide, hydrogen (H₂), and hydrogen sulfide (H₂S). Members of Aquificota, such as the chemoautotroph Thermocrinis, are capable of aerobic oxidation of sulfide and thiosulfate. CO₂ fixation has been demonstrated, with key enzymes for the reverse tricarboxylic acid (rTCA) cycle identified in metagenome-assembled genomes (MAGs) of Thermocrinis and Hydrogenobaculum28,41,42.

The Dusun Tua (DT) hot spring, located near Kuala Lumpur, Malaysia, is a bicarbonate-rich, non-volcanic hot spring and a pH of 7.6. The water typically reaches temperatures of approximately 75 °C. Patches of grey-tan biofilm are found attached to plant litter (e.g., twigs and leaves) and stones along the shallow spring. Given their high temperatures and filamentous morphology, these biofilms likely represent a form of SBC. Preliminary analysis of the microbial community in DT biofilms suggests that the dominant phyla differ from those reported in previously described high-temperature SBCs. Most existing knowledge on such biofilms is derives primarily from studies of sites such as YNP, Rehai, Chukotka, and Kunashir, which have largely focused on sulfate-rich waters dominated by Thermocrinis or Sulfurihydrogenibium (phylum Aquificota). However, differences in water chemistry, as well as the distinct coloration and microbial diversity of the DT biofilm, suggest that this environment supports distinct microbial populations and metabolic characteristics. Despite the presence of over 60 hot springs in Malaysia, grey-tan SBCs have not been reported from other locations in the region. High-temperature SBCs in tropical, bicarbonate-rich hot springs remain poorly characterized, particularly in Southeast Asia, where genomic datasets are limited. To address these gaps, we employed a multi-faceted approach combining 16 S rRNA amplicon sequencing, shotgun metagenomics with assembled contigs, and genome-centric analysis of MAGs generated using an Illumina–Nanopore hybrid strategy. Our aim was to investigate the microbial architecture and functional potential of this biofilm. A total of 61 MAGs were recovered and analyzed to elucidate their roles in carbon, nitrogen, and sulfur cycling, as well as extracellular matrix biosynthesis. These findings provide key insights into thermophilic biofilm dynamics within a tropical bicarbonate-rich ecosystem, thereby extending the current understanding of SBCs beyond traditional model sites.

Results

Site description

The Dusun Tua (DT) hot spring is located adjacent to the Langat River in Malaysia. Figure 1 presents photographs and illustrations that provide a visual overview of the site and its biofilms. A Piper plot, constructed using average water chemistry data from 2020 to 2024, indicates that the water consistently remains rich in bicarbonate, with the overall composition of major dissolved ions remaining relatively stable over the years. The DT hot spring has an average temperature of 75 °C and a pH of 7.6.

Fig. 1
Fig. 1
Full size image

Dusun Tua hot spring sampling site. (A) Piper plot summarizing the water chemistry. (B) Map showing the location of the hot spring (red dot). (C) Photograph of the site before the flood event. (D) Photograph after the flood, showing structural disturbance and changes to the spring margins. Biofilms were sampled from submerged stones and plant litter within 1 to 1.5 m of the spring head and pooled for each replicate. (E, F) Physical appearance of the grey-tan biofilm attached to submerged surfaces.

Microbial taxonomy of the DT biofilm based on microbial amplicon sequence variants (ASVs)

The water sample without visible biofilm (labelled as DT_W) was collected from the spring head. The predominant amplicon sequence variants (ASVs) in this sample belonged to Aquificota, Desulfobacterota, Deinococcota, and Pseudomonadota (Fig. 2). Biofilm samples were randomly collected approximately one meter downstream from the spring head and labelled as DT_7a and DT_7b for the July sampling trip, and DT_11a and DT_11b for the November trip. The grey-tan biofilms were filamentous and measured several centimeters in length. The DT biofilms were primarily composed of Aquificota (35%), Chloroflexota (27%), and Desulfobacterota (10%). Within Aquificota, the dominant genus was Hydrogenobacter, accompanied by a secondary taxon initially classified as a Venenivibrio-like ASV. The Venenivibrio is an acidophilic genus reported exclusively from Aotearoa-New Zealand43,44. Further analysis suggests that the ASV detected in the DT biofilms is unlikely to be a true member of this genus, despite its assignment to Venenivibrio by QIIME 2. The Chloroflexota population was mainly represented by members of the class Anaerolineae, while Desulfobacterota was largely composed of the family Thermodesulfobacteriaceae. Other major phyla present in the biofilms included Bacteroidota (Candidatus Kryptonia), Deinococcota (Thermus), and Candidatus Hydrothermae (Fig. 2; Table 1). In addition, the biofilm samples were co-dominated by archaea from Thermoproteota (Thermoprotei) and Micrarchaeota (Candidatus Micrarchaeum).

Fig. 2
Fig. 2
Full size image

Taxonomic distribution based on 16 S rRNA amplicon sequencing of the water sample (DT_W) and grey-tan biofilm samples collected in July (DT_7a and 7b) and November (DT_11a and 11b). (A) Major bacterial phyla. Aquificota dominated both water and biofilm samples, while Chloroflexota showed higher relative abundance in the November biofilms. (B) Major bacterial families. (C) Major archaeal phyla. (D) Major archaeal families.

Table 1 Selected reports on Aquificota-rich samples in high-temperature hot spring.
Full size table

Table 1 shows that filamentous SBC biofilms occurring at temperatures between 70 °C and near boiling are often dominated by Aquificota, with relative abundances ranging from approximately 30% to as high as 98% in some sites. Thermocrinis (Aquificota) species are often the major taxon in pink filamentous biofilms, while Sulfurihydrogenibium (Aquificota) may contribute to the white to grey hue of filamentous biofilms3,37,38. These biofilms are often accompanied by other thermophilic taxa, such as Thermotogota, Deinococcota, and Thermoproteota. In contrast, the DT biofilms are distinguished by a substantial co-dominance of Chloroflexota (27%), alongside Aquificota (35%) and Desulfobacterota (10%), suggesting a potentially unique microbial community structure influenced by local geochemical conditions.

Microbial taxonomy of the DT biofilm based on metagenomic shotgun contigs

Given the similarity in ASV composition between DT_7 and DT_11 (Fig. 2), genomic materials from both samples were pooled and subsequently subjected to Illumina shotgun metagenomic sequencing and Oxford Nanopore sequencing, followed by a co-assembly approach. This resulted in 126,999 contigs larger than 1,000 bp, with a N50 of 1,304 bp and the largest contig being 287,008 bp. After filtering out non-cellular components and unassigned contigs, 86% of the total DT contigs belong to the domain Bacteria, with Aquificota (25.0%) and Deinococcota (24.8%) being the largest, followed by Pseudomonadota (13.6%), Bacillota (7.9%), and Actinomycetota (6.1%) (Fig. 3).

Fig. 3
Fig. 3
Full size image

Taxonomic distribution determined using metagenome contigs at (A) domain level and (B) phylum level.

Metabolic pathway reconstruction using metagenome-assembled genomes (MAGs)

Functional inference from 16 S rRNA gene sequences can provide a general overview but is constrained by limited accuracy and taxonomic resolution. To overcome this, we utilized MAGs for genome-level metabolic reconstruction. MAGs were generated, and only medium- and high-quality MAGs were analyzed (Fig. 4, Supplementary Data A). Gene analysis of major metabolic pathways and genome-centric investigations were conducted using METABOLIC-G, with the potential capabilities of each MAG in nitrogen, sulfur, and other biogeochemical cycles summarized in Fig. 5. The microbial community likely depends on chemotrophy, fermentation, or anaerobic respiration for energy generation. Pathways linked to acetate oxidation, carbon fixation, ethanol oxidation, fermentation, and organic carbon oxidation were identified, suggesting that the microbiome may employ these carbon cycling strategies when conditions are favorable. Considering the finite availability of resources and the turbulent flow of high-temperature water, we hypothesize that not all metabolic processes predicted by METABOLIC are equally active under such harsh conditions. Given the high temperatures of the DT biofilm environment, photosynthetic autotrophy is absent, as evidenced by the lack of associated genes in nearly all MAGs. The METABOLIC data suggest a mixotrophic lifestyle, where the community alternates between chemosynthetic carbon fixation and heterotrophic organic substrate utilization, aligning with an earlier hypothesis5.

Fig. 4
Fig. 4
Full size image

Phylogenetic tree of MAGs recovered from hybrid-metagenomes. (A) Bacteria and (B) archaea. MAG quality is indicated as high (“H”) or medium (“M”) based on CheckM assessment. The major phyla are Aquificota (orange), Chloroflexota (cyan), Desulfobacterota (moss green), Bacteroidota (royal blue), and Deinococcota (brown).

Fig. 5
Fig. 5
Full size image

Heat map showing predicted metabolic potentials of individual MAGs, based on key marker genes identified using METABOLIC-G. Pathways are grouped by biogeochemical cycles: carbon, nitrogen, sulfur, and other elemental transformations. “Other” includes genes associated with arsenic redox processes, iron reduction, and selenate reduction.

To identify the primary metabolic processes, we further examined selected MAGs affiliated with major phyla identified from the amplicon data: Aquificota, Chloroflexota, Desulfobacterota, Bacteroidota, and Deinococcota. A total of seven MAGs were associated with Aquificota, with DT_Bin_08 linked to the order Hydrogenothermales, while the remaining six are associated with the order Aquificales. Among them, DT_Bin_26 is associated with Hydrogenobacter, whereas DT_Bin_41 is linked with an unidentified Aquificaceae bacterium. The remaining four MAGs (DT_Bin_25, DT_Bin_43, DT_Bin_49 and DT_Bin_52) are closely related to Aquificaceae bacterium with an NCBI strain identifier UBA1109645. DT_Bin_26, attributed to Hydrogenobacter from Aquificota, potentially encodes multiple abilities within the nitrogen cycle, including nitric oxide reduction, nitrate oxidation and reduction, nitrogen fixation, and nitrous oxide reduction. Previous culture-based studies of Aquificota suggested that these organisms typically oxidize inorganic compounds such as hydrogen gas or sulfur compounds (e.g., hydrogen sulfide) to generate energy, with oxygen serving as the terminal electron acceptor in their respiratory chain46,47,48,49,50.

The second most abundant phylum in the DT biofilm is Chloroflexota represented by MAGs DT_Bin_23, DT_Bin_35, DT_Bin_46, all associated with the class Anaerolineae. Gene mining suggests that Anaerolineae in the DT hot spring may adopt a chemomixotrophic lifestyle, utilizing organic compounds and oxygen for energy, supported by the presence of genes related to hydrogenase, carbon fixation, respiratory complexes, and the TCA cycle. Additionally, they may also carry out chemoautotrophic oxidation of inorganic compounds for CO2 fixation. Chloroflexota may utilize the oxidation of organic compounds from plant litter attached to trapped within their biofilms as an energy source. Notably, DT_Bin_35 encodes genes potentially involved in Fe(II) oxidation to Fe(III), further highlighting the metabolic versatility of this group.

The phylum Desulfobacterota, represented by family Thermodesulfobacteriaceae (DT_Bin_39 and DT_Bin_60), is another abundant member of the DT biofilm community. Members of this group are widely recognized for their role in sulfur cycling, particularly under anaerobic and thermophilic environments. The two MAGs appear to encode multiple sulfur transformation pathways (both oxidative and reductive routes), such as sulfur oxidation, sulfate reduction, sulfite oxidation, sulfite reduction, and thiosulfate disproportionation, which may support energy generation under fluctuating redox conditions. As a group of sulfate-reducing bacteria, Desulfobacterota is recognized for its adaptation to the extreme hot spring conditions, generating energy through diverse, thermophilically adapted metabolic strategies51.

Kryptonia (DT_Bin_37), a member of the phylum Bacteroidota, Thermus (DT_Bin_58), from the phylum Deinococcota, together with other members of Bacteroidota (DT_Bin_7, DT_Bin_21, DT_Bin_45, DT_Bin_51, DT_Bin_56), are likely heterotrophic. Their presence suggests active roles within the biofilm, particularly in recycling organic matter derived from cell lysis, microbial exudates, or detrital inputs such as plant litter. Such heterotrophs may contribute to nutrient turnover and support the maturation of the biofilm matrix by releasing byproducts that are accessible to other community members. Amplicon data indicate that Candidatus Hydrothermae is a major member of the biofilm community; however, we were unable to construct its MAG. The candidate phylum Hydrothermae is primarily associated with hydrothermal vents but has also been reported in a few hot springs52.

Macromolecules related to biofilm formation

This section explores the role of key macromolecules in biofilm formation by major thermophilic phyla, as inferred from MAG analyses. Key proteins such as pellicle biosynthesis protein C (PelC) and pellicle biosynthesis protein G (PelG) were detected in Aquificota MAGs, indicating their potential role in biofilm formation and exopolysaccharide transport in matrix production53. The presence of bifunctional glycoside hydrolase GH114 (polysaccharide deacetylase family) proteins suggests the ability to modify polysaccharides, potentially enhancing matrix stability. Although flagellar genes are present, the apparent absence of quorum-sensing involvement implies that Aquificota relies on physical motility and extracellular matrix production in biofilm development. The detection of type IV pilus proteins such as PilT and PilQ suggests that twitching motility may support biofilm expansion and maintenance in Aquificota cells54.

Analysis of Chloroflexota MAG genomes suggests a reliance on polysaccharide biosynthesis as the primary strategy for biofilm formation. These MAGs encode polysaccharide-related proteins, including oligosaccharide flippases and polysaccharide biosynthesis enzymes. The absence of flagella and pili-related genes implies limited motility, with exopolysaccharide production likely serving as the primary mechanism for matrix development. Proteins such as lipid II flippase and PIG-L family deacetylases may contribute to cell wall remodeling, thereby enhancing biofilm integrity55.

Genome analysis of Desulfobacterota MAGs reveals a diverse set of flagellar and type IV pilus proteins, suggesting that motility may play a key role in biofilm formation. Proteins such as FlgB, FliM, and MotB suggest flagella-mediated motility and swarming behavior that facilitate surface attachment and biofilm development54. Prepilin peptidase and associated cleavage proteins support type IV pilus assembly, enabling twitching motility. Curli-associated proteins, including CsgF and CsgB, indicate potential roles in curli-mediated adhesion and matrix formation. The presence of CsgG/HfaB family proteins in both Desulfobacterota and Aquificota indicates that curli fibers may be a shared strategy for enhancing biofilm cohesion and stability. These amyloid-like proteins may form scaffolds that resists shear forces (e.g., high current flow in hot springs) and environmental stress, enhancing resilience beyond that provided by capsular and polysaccharide-rich biofilms56.

Bacteroidota-MAG genomes suggest that exopolysaccharide production plays a central role in biofilm formation. Capsular polysaccharide and lipopolysaccharide biosynthesis proteins, along with Vi polysaccharide biosynthesis enzymes, underscore their capacity to produce complex polysaccharides for biofilm integrity and adhesion. In the absence of pilin genes, the presence of gliding motility proteins (e.g., GldD, GldL, GldE) may provide an alternative mechanism for surface attachment and movement, supporting biofilm formation without flagella57. Deinococcota (e.g., Thermus sp. MAGs in DT biofilms) likely rely on extracellular matrix production, as indicated by polysaccharide-modifying proteins such as pyruvyl transferase (CsaB) and prepilin-type proteins. These enzymes may contribute to matrix stability and structural cohesion. The presence of type IV pilus proteins, including PilT and PilB, suggests twitching motility may play a key role in biofilm formation and maintenance, potentially compensating for the absence of flagella54.

Mining of carbohydrate-active enzymes (CAZymes) sequences

We conducted an analysis of carbohydrate-active enzyme (CAZyme) sequences to investigate the microbial community’s potential to hydrolyze cellulose for energy. Plant litter, rich in cellulose and hemicellulose, requires specific enzymes to break down these polymers into simpler sugars. CAZymes from thermophilic organisms are particularly valued for their enhanced thermostability, which is advantageous in biotechnological applications. In the context of biofilm metabolic, it is hypothesized that the microbial community may express CAZymes to access cellulose-derived carbon. To explore this potential, we examined CAZymes-coding genes within the MAGs. The analysis targeted cellulolytic enzyme sequences from glycoside hydrolase (GH) families GH6, GH48, and GH148, as well as hemicellulolytic enzymes from GH43, GH52, GH62, GH67, GH113, and GH120. Additionally, enzymes from mixed-function GH groups, including GH1, GH2, GH3, GH5, GH8, GH9, GH10, GH12, GH16, GH26, GH30, GH39, GH44, GH51, and GH116, were incorporated into the analysis58.

A total of 407 protein sequences associated with these GH groups were identified from the DT biofilm MAGs. Only ~ 35% of these sequences were associated with the dominant phyla—Aquificota, Chloroflexota, Desulfobacterota, Bacteroidota, and Deinococcota, which together accounted for over 75% of the community based on amplicon data. Surprisingly, non-dominant phyla contributed a greater number of these enzyme sequences, despite being minority members of the community. This observation suggests that non-dominant community members may disproportionately contribute to polysaccharide degradation, possibly serving as specialized degraders or opportunistic scavengers in the biofilm ecosystem.

However, the dynamic hydrology of the DT hot spring presents a significant challenge for extracellular enzyme function. The high flow rate and steep channel slope likely dilute or wash away freely secreted enzymes before they could act efficiently on their substrates. For enzymatic hydrolysis of plant material to be ecologically feasible under such conditions, we reasoned that these enzymes must either be anchored to cell surfaces or retained within the structural framework of the biofilm. To test this, all 407 cellulase and hemicellulase sequences were analyzed for extracellular localization using SignalP59, and for transmembrane domains indicative of membrane-bound enzymes using TMHMM60 and Phobius61. Surprisingly, only 83 enzymes were predicted to be extracellular, and none were identified as cell-bound, as their sequences lacked transmembrane regions. These findings suggest that instead of membrane anchoring, the community may rely on retaining enzymes within the mature biofilm matrix for effective polysaccharide hydrolysis.

Discussion

Aquificota-rich filamentous SBCs have a long history, particularly those first reported from hot springs in YNP. Early studies employed 454 metagenome sequencing with limited depth, which later followed by Illumina amplicon and shotgun sequencing (Table 1). Direct comparison ASVs from the DT biofilm with those obtained using other methods (e.g., clonal libraries, Roche 454, or Illumina iTag) may not be appropriate due to technical differences in primer, target regions, library preparation, sequencing depth, and data processing pipelines.

In high-temperature hot springs with high sulfide concentrations, Aquificota is typically dominated by the genus Sulfurihydrogenibium2,3,8,24, consistent with reports of other sulfur-rich SBC biofilms where sulfur or its derivatives may support the persistence of sulfur-oxidizing taxa. In high-temperature hot springs with relatively low sulfide content, Aquificota populations often have a high proportion of the genera Hydrogenobacter and Thermocrinis4,6,11,18,24,62. As the DT hot spring is a carbonate-type system with relatively low sulfide levels, the Aquificota population was primarily composed of Hydrogenobacter. It is notable that Thermocrinis was not detected above 1% ASV abundance, despite the temperature and pH of the DT site being within its optimal growth range. Although we were unable to construct a MAG for Thermocrinis through a hybrid approach using Illumina and Nanopore reads, approximately 2% of the total contigs generated from MK1c Nanopore sequencer were assigned to Thermocrinis. This further supports the conclusion that Thermocrinis is a minor constituent of Aquificota in DT biofilms.

Typically, Chloroflexota and Cyanobacteriota are commonly found in green or brown biofilms in circumneutral to alkaline hot springs due to their capacity for both aerobic and anaerobic photosynthesis58. During field sampling, a green-brown biofilm patch measuring approximately 10 × 5 cm was observed on terrestrial surfaces intermittently splashed by spring water near the source. The localized temperature at this site was approximately 45 °C, substantially lower than the 75 °C measured in the main outflow channel, where the submerged grey-tan biofilm was predominantly found. Sequencing of the green-brown biofilm using the V3–V4 primer set revealed that Chloroflexota and Cyanobacteriota together comprised nearly 60% of the total ASVs (data not shown). Unlike other Aquificota-rich SBCs (summarized in Table 1), the co-dominance of Aquificota and Chloroflexota in SBCs has not been previously documented. We hypothesize that the presence of Chloroflexota in the DT biofilm may results from dispersal processes. Water splashes, rainfall, and occasional shifts in spring water flow likely facilitated the downstream transport of cells from the upstream green-brown biofilm, facilitating their co-colonization of the downstream grey-tan biofilm. Cyanobacteriota were negligible in the grey-tan biofilm, likely inhibited by high temperatures, which limit bacterial photosynthesis.

Based on the 16 S rRNA amplicon dataset, six bacterial phyla (Aquificota, Chloroflexota, Desulfobacterota, Bacteroidota, Deinococcota, and Hydrothermae) constitute a significant portion of the DT biofilm community. In several high-temperature SBCs (Table 1), Aquificota constitutes more than 90% of the microbial community, for instance, in the Mammoth Hot Springs of YNP3 and the Rehai hot springs in China11. These observations have led previous studies to assume that Aquificota plays a primary role in biofilm formation, with other taxa acting as secondary occupants11. However, in the case of the DT biofilm, Aquificota comprises, on average, only 35% of the total ASVs. Given this more balanced community composition, we propose that the biofilm matrix and extracellular polymeric substances in the DT biofilm are assembled by a consortium, rather than by Aquificota alone. This is further supported by the identification of various sequences associated with biofilm formation, EPS production, and attachment across multiple phyla in the MAGs, including those from both Aquificota and other abundant groups. Overall, it is likely that EPS, curli fibers, and other matrix components likely co-contribute to the structural integrity of biofilm, promoting resilience and persistence under extreme environmental conditions.

The DT hot spring is located near the Langat River in Malaysia, with its main source about 6 m from the riverbank. A severe flood caused by river overflow led to the accumulation of plant litter and various wastes, along the shallow thermal channels, thereby reducing water velocity and introducing new surfaces for microbial colonization. Following the flood, likely due to these hydrodynamic changes, grey-tan biofilms began to form. Prior to the flood, biofilm development was negligible in the outflow channels, likely due to the rapid, turbulent flow and elevated temperatures that inhibited stable colonization. Although the flood did not alter water chemistry, pH, or temperature, it slowed water flow, primarily due to the plant litter. Our initial thought was that plant litter, as a source of polysaccharides and fermentable sugars, could have stimulated early biofilm formation. If this hypothesis is valid, cells would be expected to secrete cell-bound extracellular CAZymes to break down complex polysaccharides into simpler sugars for transport against the water flow. While anchoring hydrolases to the cell membrane has been reported in thermophiles63, this strategy may be energetically costly and impractical for hyperthermophiles, particularly those with streamlined genomes adapted to minimum non-essential metabolic burdens. Closer examination revealed that key MAG members lacked relevant cell-bound extracellular CAZymes.

We propose that plant litter plays a foundational role in shaping microbial colonization in the DT hot spring by reducing water flow and offering binding surfaces for the initial attachment of microbial cells. In the early stages of biofilm development, Aquificota-dominated communities likely establish themselves on these surfaces, relying on chemoautotrophy to oxidize reduced sulfur compounds (e.g., sulfide, elemental sulfur, or thiosulfate) for energy and CO₂ fixation. However, the relatively low sulfur levels in the DT system may limit the long-term sustainability of this sulfur-based metabolism. Despite this constraint, Aquificota may persist by utilizing both aerobic and anaerobic respiration pathways, allowing them to maintain metabolic activity across a range of redox conditions. This physiological versatility is particularly advantageous during early colonization, enabling Aquificota to remain metabolically active under both oxic and anoxic conditions. In the DT biofilm environment, steep redox gradients are shaped by site-specific features, including turbulent flow, sediment accumulation, and the development of EPS-rich microzones within the biofilm matrix. Over time, the dispersal of Chloroflexota from upstream habitats may have facilitated their gradual incorporation into the biofilm. Their prevalence suggests that they may play an important role in carbon cycling and contribute to energy acquisition under fluctuating redox conditions. The EPS-rich biofilm matrix likely creates microenvironments where oxygen levels vary sharply, allowing Chloroflexota to persist in both oxic and suboxic zones. Their presence within the submerged grey-tan biofilm, despite originating from upstream phototrophic biofilms, suggests successful colonization of hotter downstream zones, potentially by adopting a heterotrophic or chemomixotrophic lifestyle. Additionally, their enrichment near detrital plant material suggests the possible utilization of low-molecular-weight organics released from decaying biomass. This functional plasticity likely enables Chloroflexota to establish long-term residence within the DT streamer biofilm, especially as the community transitions from autotroph-dominated to a more metabolically integrated consortium. This integration may have contributed to a community shift, with Aquificota populations declining as Chloroflexota, gained a competitive advantage due to their greater metabolic flexibility. In more mature biofilms, the accumulation of organic carbon from microbial exudates, decaying cells, and plant litter may have favored heterotrophic processes, further supporting the proliferation of Chloroflexota. Additionally, the microbial community may utilize alternative pathways such as nitrification and denitrification to exploit nitrogen compounds for energy and reduce H₂. As the biofilm thickens and develops a protective extracellular matrix, heterotrophs are expected to increasingly rely on organoheterotrophy, while chemoautotrophic activity continues in parallel. The mucus-rich matrix likely shields extracellular (non-cell-bound) enzymes, enhancing substrate retention and enzymatic efficiency by protecting them from dilution or washout. This scenario supports the idea that microbial communities in SBCs dynamically adjust their metabolic strategies based on the availability of carbon sources and electron donors (e.g., O₂, H₂, and sulfide), optimizing survival and growth under fluctuating environmental conditions5.

The predicted functional profiles provide insights into the potential ecological roles of the dominant taxa within the biofilm. For example, the detection of CO₂ fixation pathways in Aquificota and several other phyla suggests that primary production may be distributed across multiple lineages, rather than confined to classical sulfur oxidizers. In comparison, the diversity of complete nitrogen cycling genes/pathways appears somewhat lower than that of carbon cycling, which may be attributed to the fragmented nature of some MAGs. Nevertheless, MAG DT_Bin_26 (classified as Aquificota, genus Hydrogenobacter) exhibits a broad range of nitrogen-related functions, including genes for nitric oxide reduction, nitrite oxidation, nitrite reduction, nitrogen fixation, and nitrous oxide reduction. This implies that at least some members of the community may contribute to internal nitrogen turnover and redox balance. The co-occurrence of hydrogenase and sulfur oxidation genes in several MAGs further points to possible redox flexibility, allowing microbes to utilize alternative electron donors depending on environmental conditions. Collectively, these traits suggest a metabolically diverse and potentially functionally complementary community that is well-suited to withstand thermal stress, nutrient limitation, and hydrodynamic disturbance.

Conclusion

The DT biofilm hosts a complex microbial community, dominated by Aquificota, Chloroflexota, Desulfobacterota, Bacteroidota, Deinococcota, and Candidatus Hydrothermae, with significant archaeal co-dominance by Thermoproteota and Micrarchaeota. While some high-temperature streamer biofilms are typically dominated by Aquificota, the DT system exhibits a richer phylogenetic diversity. The mechanisms that enable Aquificota to coexist with such a broad range of other phyla remain unclear, especially in contrast to the simpler compositions seen in other SBC systems. The relatively high abundance of Chloroflexota in the DT biofilm may stem in part from dispersal of Chloroflexota–Cyanobacteriota biofilms from upstream, contributing to a more heterogeneous community structure than those reported in other hot springs. Genome-resolved metagenomic analyses revealed diverse set of metabolic strategies, including carbon, nitrogen, sulfur, and other elemental cycles, indicating niche partitioning and metabolic complementarity among the coexisting taxa. Slower water flow following the flood appears to be the primary factor enabling biofilm establishment by providing stable surfaces for microbial attachment. This is supported by genome-centric and CAZyme analyses, which show the absence of membrane-bound cellulases and hemicellulases, indicating that in-situ plant-derived polysaccharides were not a key nutrient source during the initial colonization stage. Instead, biofilm development was likely initiated by microbial attachment and chemoautotrophic activity, primarily by Aquificota. Over time, the internal accumulation of organic matter—such as microbial exudates and decaying cells—facilitated the gradual integration of heterotrophs, including Chloroflexota, Bacteroidota, and others. As the biofilm matured and developed a protective extracellular matrix, extracellular CAZymes likely became increasingly effective at degrading trapped plant litter, further sustaining heterotrophic processes within the community. Additionally, the microbial community in the DT biofilm appears to adopt a cooperative strategy for polysaccharide degradation, wherein low-abundance taxa contribute specialized enzymatic functions, and the biofilm structure enhances enzyme retention and substrate accessibility. We hypothesize that enzymatic degradation of plant polymers may serve as an auxiliary energy source, supplementing the metabolic activities of the community as it adapts to environmental changes. This study did not include systematic temporal monitoring or spatial profiling along the full upstream–downstream gradient, nor did it incorporate transcriptomic or metabolomic analyses, which should be considered in future. In addition to ecological insights, the MAG-derived genes identified, such as thermotolerant CAZymes and redox-related enzymes, may hold biotechnological relevance. These include potential applications in high-temperature biomass degradation, biofuel production, and redox-driven wastewater treatment. The presence of stress-adapted traits also offers clues for designing robust microbial platforms for industrial processes. Together, these traits represent a valuable resource for future applied studies.

Methods

Biofilm sampling and DNA extraction

The Malaysian Dusun Tua (DT) hot spring is located at latitude 3.1368° N and longitude 101.8345° E. Water analysis was performed by a local certified service provider, Allied Chemists Sdn Bhd, using methods reported by the American Public Health Association (APHA). Sampling trips were conducted in July 2022 (labelled DT_7a & DT_7b) and November 2022 (DT_11a & DT_11b) to collect biofilms. The grey-tan biofilm samples were randomly collected into multiple tubes and immediately transported to the laboratory and frozen at − 20 °C within 12 h. The bulk genomes of the biofilms were extracted using SPINeasy™ DNA extraction kit (MP Biomedicals, Irvin, CA, USA) according to the manufacturer’s recommended protocol. DT biofilm samples were lysed using a TissueLyser (Qiagen, Germany) via bead beating, following 3 cycles (20 Hz) for 1 min each. The extracted DNA were evaluated using a Nanodrop™ 1000 spectrophotometer (Thermo Scientific, Wilmington, DE, USA), a Qubit 3.0 Fluorometer (Invitrogen, Merelbeke, Belgium) and with a 1% w/v agarose gel electrophoresis. Approximately 8 L of water was collected from the spring head of the hot spring, which showed no visible biofilm. The water was filtered through a 0.22 μm Sterivex filter unit (Merck, USA) until clogging occurred, and the filter was subsequently stored at − 20 °C until genomic DNA extraction.

Amplicon sequencing and data processing

The extracted metagenomic DNA from biofilm samples and water were used as templates for amplicon sequencing. We used the following primers: (a) bacterial 16S rRNA V3-V4 region (341F 5’-CCTAYGGGRBGCASCAG-3’ and 806R 5’-GGACTACNNGGGTATCTAAT-3’), and (b) archaeal 16 S rRNA V4 region (U519F ‘-CAGYMGCCRCGGKAAHACC-3’ and 806R 5’-GGACTACNSGGGTMTCTAAT-3’), with amplification carried out under standard conditions (55 °C annealing temperature and 25 PCR cycles) as performed by the sequencing service provider. Sequencing was conducted using Illumina NovaSeq 6000 platform (Illumina, San Diego, USA) at NovogeneAIT Genomics (Singapore). Paired-end libraries with read lengths of 250 bp were generated. Adapter sequence were trimmed and reads were filtered to ensure an average Phred score of 20 using Trimmomatics64. Subsequently, the resulting paired-end reads were demultiplexed and denoised using the DADA2 plugin65 with default parameters within the QIIME 266 pipeline. Subsequently, taxonomic classification was performed using the SILVA SSU 138.1 database. The generation of abundance table and subsequent data analysis were accomplished in R using data generated from the QIIME 2 pipeline66. Prior to diversity and community composition analyses, the feature table was rarefied to the sequencing depth of the lowest-depth sample to account for uneven sequencing effort across samples. Plots were generated using the ggplot2 library in R.

Shotgun sequencing and data processing

The duplicated extracted metagenomic DNA from biofilm samples collected at each sampling time were pooled prior to fragmentation by a Covaris sonicator (Covaris, Woburn, USA). Then, the fragmented DNA was used for dual-indexed, paired-end library construction following the Illumina DNA Prep kit protocol (Illumina, San Diego, USA). The constructed library samples were analyzed in Qubit 3.0 Fluorometer and Agilent Bioanalyzer 2100 (Agilent Technologies, Palo Alto, USA). Shotgun metagenome sequencing was carried out in an Illumina NovaSeq 6000 with the running mode of PE 150 (paired-end 150 bp). A minimum of 20 Gb (equivalent to approximately 66.5 million paired end reads) output was reserved for each sample. The resulting raw pair-end reads were trimmed and filtered using seqtk-v1.4. In addition, the identical genomic sample undergone Nanopore long-read sequencing. We conducted library preparation directly following the Nanopore-recommended protocol SQK-NBD112-24 (Q20 + chemistry on R10.4 flow cell), omitting genome shearing to prioritize obtaining longer reads. The Nanopore Ligation Sequencing Kit SQK-LSK112 protocol was adhered to, including AMPure treatment, adapter ligation, purification, and Qubit quantification. Samples were loaded onto a MinION R10.4 flow cell within Mk1C, operating with MinKNOW v.22.05.8. Basecalling was performed using MinKNOW coupled with Guppy v6.1.5, configured with FLO-MIN112-Super-Accurate settings. After basecalling, the process involved the removal of barcodes, trimming and filtering reads, followed by subjecting those meeting Q20 criteria (< 5% error) to de novo assembly using Flye v2.9 (parameters: --nano-hq --meta). Co-assembly was then performed with the pooled Illumina clean reads (DT_7 and DT_11) along with Nanopore long reads data using MEGAHIT v1.2967.

Metagenome-assembled genome and metabolic pathway reconstruction

Co-assembled contigs of Illumina shotgun data and Nanopore data were used to construct MAGs using the binning tools MetaBAT2 v2.12.168,69 CONCOCT v1.0.070, and MaxBin2 v2.2.671,72 followed by the bin refinement & reassembly module of metaWRAP v1.373. The quality of the MAGs were assessed using CheckM v1.2.274. Taxonomic classification of the MAGs was performed using GTDB-tk v2.3.2 with the Genome Taxonomy Database (GTDB) as a reference75,76. The phylogenetic trees of the MAGs were constructed using IQ-TREE v2.0.6 with default parameters77 and using ModelFinder for substitution model selection78. The phylogenetic trees generated were subsequently visualized using iTOL79. The metabolic functional profiles of all the MAGs were predicted using METABOLIC-G v4.080. KEGG modules with completeness greater than 75% were considered as a present. Gene presence was annotated using the HMMER search tool81as implemented in METABOLIC-G. When necessary, reported genomes of related MAGs were manually compared with those obtained in this study. Genes annotation of selected MAGs was performed using DFAST82.