Leaf collection and aquarium set-up

In July 2021, 35 green leaves of the water lily Nymphaea sp. were gathered from a large freshwater pond at the Bonn University Botanic Garden. The floating leaves were collected several meters away from shore using a rowboat. Upon collection, the leaves were immediately placed in the aquariums (S1 Fig). The pond is approximately 1.5 meters deep and fed primarily by a small stream and rainfall.

Pond water and mud from the pond bottom were also collected. The pond mud for the mud-substrate aquarium was sterilized before the start of the experiment. Pond water was collected in sterilized containers at the start of the experiment and immediately placed in the aquariums.

To determine the substrate minerology of each substrate, X-ray powder diffraction data were recorded on a PANalytical Empyrean diffractometer using CuKα radiation (λ = 1.54060 Å) at 40 kV and a tube current of 40 mA. Samples were hand-ground in an agate mortar and scanned on a rotating stage at 2°2θ/min (step size 0.02°2θ) from 5–80°2θ. The qualitative and quantitative mineralogical analysis was performed using the HighScore Plus PANalytical software. The pond mud contains a mineral composition of 30% quartz (SiO₂), 62% illite ((K, H₃O) (Al, Mg, Fe)₂(Si, Al)₄O₁₀[(OH)₂·(H₂O)]), and 8% alkali feldspar (KAlSi₃O₈). Two other substrates were tested, kaolinite clay (Al₂Si₂O₅(OH)₄) and fine-grained quartz sand (SiO₂), which were purchased commercially. The grain size of the substrates was not as factor in this experiment, but instead emphasized the differences in the mineral composition of the individual substrates.

The sediments were sterilized using an autoclave to ensure the starting microbial community was derived only from the leaves and pond water. The experiment was run in four 60 L aquaria. One aquarium contained pond water with no substrate as a control, while the other three aquariums contained pond water and one of the following fine-grained substrates: kaolinite clay, pond mud, or fine sand. Each aquarium was maintained at room temperature (~ 20°C) and was exposed to the natural daylight cycles in Bonn, which was 14 hours of daylight on average during the experiment.

Each water lily leaf was placed into a hand-sewn mesh pouch to ensure that all the tissue from individual leaves remained together during decay. Eight leaves were placed in each aquarium at the water–substrate interface, held down by some substrate along the edges of the pouches, but the leaves themselves were not allowed to overlap. Henceforth, for the sake of linguistic convenience, we will refer to the leaves in the various treatments as green leaves (leaves collected straight out of the pond in the Botanic Garden), control leaves (leaves in the aquarium with only pond water without any substrate), kaolinite clay leaves (leaves on the kaolinite clay substrate), mud leaves (leaves on the pond mud substrate), and sand leaves (leaves on the sand substrate).

Sampling of aquarium leaves

Water lily leaves were sampled after one (T1), two (T2), or three (T3) months. Three additional leaves were collected from the pond and used as the T0 control (S1 Fig). Due to the large size of the water lily leaves and the limited space in the aquariums, only two leaves could be sampled from each treatment at T3. As time progressed, the leaves sunk and settled directly on the substrate. By T3, the substrate coated the top and bottom of the leaf surfaces. After the leaves were collected, a small section of each leaf was preserved in formaldehyde-acetic-alcohol (FAA: 70% ethanol, 4% formaldehyde, 2% acetic acid) for subsequent microscopy. Another small section of the leaf was dried and pressed between newspapers to preserve the leaf in its degraded state for later study. The rest of the leaves were preserved in TE Buffer at -20°C until DNA extraction.

Scanning electron microscopy

The FAA-preserved leaves were transferred into a 70% ethanol solution and subsequently dried using 22 cycles in a Leica EM CPD300 critical point dryer (CPD) at the Nees Institute for the Biodiversity of Plants (now part of the Bonn Institute of Organismic Biology), University of Bonn. The CPD leaves were mounted on aluminum stubs and coated with a thin layer of palladium. Scanning electron microscopy (SEM) was carried out using a Tescan Vega 4 LMU Scanning Electron Microscope at the Division of Paleontology, Institute of Geosciences (now part of the Bonn Institute of Organismic Biology), University of Bonn.

Analysis of water chemistry

Concurrent with the sampling of leaves at T1 and T2, water samples were collected in 50 ml Falcon tubes for chemical analysis. The water was measured for concentrations of iron, manganese, silica, calcium, magnesium, sodium, and potassium, as well as for electrical conductivity, pH, and temperature. The composition and concentration of both cations and anions were measured using inductively coupled plasma mass spectrometry (Agilent ICP-MS 7700) and ion chromatography (Metrohm IC 930 Compact Flex 947 linked to a Professional UV/VIS Vario detector) at the Institute for Hygiene and Public Health (University Hospital Bonn).

DNA extraction and sequencing

Using the leaves preserved in TE Buffer, leaf biofilms were separated from leaf surfaces with sonication in an ultrasound bath (Branson 1210 Ultrasonic Cleaner; Emerson, St. Louis, Missouri, USA) for 2 minutes. Subsequently, the samples were vacuum filtered using a 0.25-micrometer pore filter (mixed cellulose ester membrane; Berrytec GmbH, Grünwald, Germany). The filters with the biofilm microbes were then cut into small pieces and kept in 2 mL microcentrifuge tubes with TE buffer and stored at -20°C until further processing.

DNA extraction from the collected biofilm samples was performed using the FastDNA Spin Kit for Soil, following the manufacturer’s protocol. This particular kit had previously been confirmed to be efficient for biofilm DNA extraction from water lily leaves [17]. The elution step was carried out using 50 microliters of Dnase-free water. The concentration and quality of the extracted DNA were assessed using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, Massachusetts, USA).

For microbial community analysis, the V4 variable region of the 16S rRNA gene was amplified using the primers 16s-515F (GTGCCAGCMGCCGCGGTAA) and 16s-806R (GGACTACVSGGGTATCTAAT) [54]. The ITS (Internal transcribed spacer) region was amplified using the primers ITS1F (CTTGGTCATTTAGAGGAAGTAA) and ITS2R (GCTGCGTTCTTCATCGATGC) [55]. This allowed for the quantification of both bacterial (16S) and fungal (ITS) microbial communities. The polymerase chain reaction (PCR) reaction was a single-step PCR with HotStarTaq Plus Master Mix Kit (QIAGEN). The PCR consisted of an initial denaturation step at 95°C for 5 minutes, followed by cycling at 95°C for 30 seconds, 53°C for 40 seconds, and 72°C for 1 minute for 30–35 cycles. A final elongation step was performed at 72°C for 10 minutes.

Paired-end sequencing (bTEFAP) was conducted by MR. DNA (Shallowater, Texas, USA) using the Illumina MiSeq sequencing platform, following the manufacturer’s guidelines. Raw sequence data were processed using QIIME2 with default parameters [56]. Sequence quality control, denoising, and chimera removal were performed using DADA2 to generate amplicon sequence variants (ASVs) [57].

Taxonomic classification of the 16S rRNA sequences was accomplished using the Silva 138 database [58], while the Unite database was used for the ITS sequence classification [59]. ASVs and taxonomy data derived from the QIIME2 and DADA2 pipelines were integrated using R version 4.2.1 [60].

For 16S sequences, additional classification was performed using GreenGenes2 [61], which was employed for phenotypic inference using BugBase [62]. The BugBase phenotypic determination approach was employed to identify potential biofilm-forming bacteria present on the water lily leaves.

Microbial community analysis

ASV richness and Simpson’s evenness were calculated separately for bacteria (16S) and fungi (ITS) using “vegan” [63]. Each dataset (16S and ITS) was rarified to the lowest number of sequences in each dataset (43,836 and 14,654, respectively). Statistical differences in richness and evenness were calculated between substrate and time using ANOVAs. We used “ampvis2” to create heat maps of the most abundant bacteria and fungi genera [64]. Similarities in the ASV composition for bacteria and fungi were visualized using Venn Diagrams and the package “ggvenn” [65] We used the “anosim” function in “vegan” to visualize the ASV diversity within and between groups.

We used “microeco” to visualize the community composition of each substrate and time point using an non-metric multidimensional scaling (NMDS) plot and Bray-Curtis distances [66]. The statistical difference in microbial community composition was calculated using a PERMANOVA (the “adonis2” function in “vegan”). Pairwise comparisons between substrates and time were calculated using the “pairwise.adonis2.”

The correlation of the mineral content and other parameters of the water samples with microbial community composition was assessed using the “envfit” function in “vegan” and plotted using a dbRDA (distance-based redundancy analysis) in “microeco”. Only environmental vectors with a statistically significant correlation (P<0.05) were inclided. Correlations between the fifty most abundant genera based on DESeq2 differential abundance calculations and water chemistry were visualized using Pearson’s correlation plot in “microeco” with a significance cutoff of P = 0.05.

The influence of water chemistry, time, and substrate on the microbial community composition was evaluated using a variation partitioning analysis (VPA) within “vegan”. First, we created a correlation matrix with the water chemistry data and removed those minerals with collinearity greater than 0.8, which resulted in a model with Fe, Ca, and K (S1 Table). Hence, the VPA consisted of substrate, time, and Fe, Ca, and K. The data were transformed using the “Hellinger” method. The resulting VPA shows the percent of the variation in the microbial community composition that is explained by these three minerals. We then performed separate redundancy analyses (RDAs) with ANOVAs to identify the individual impacts of substrate, time, and water chemistry on community composition.

Finally, we used the BugBase data (16S only) to determine differences in the abundance of biofilm-formers that developed between substrates over time. BugBase uses an algorithm to predict functional pathways such as biofilm formation. We filtered the BugBase OTUs to only include those tagged as “Forms Biofilms” and calculated the richness of biofilm formers for each substrate and time point.

SEM images show differences between substrate treatments

Immediately after removal from the aquarium, there was evident decay of the kaolinite and sand leaves at T3 (S1 Fig). However, the differences in the level of decay between leaf samples from each substrate over time became distinctly clear after SEM analysis and the study of the pressed leaves themselves.

The control leaves at T1 were covered in bacteria and fungi. On the control leaves at T2, there was an increase in fungi, and microbially produced EPS could be observed. On the control leaves at T3, there were mineral particles, perhaps deriving from the pond water, on the leaf surfaces, and the leaves were coated in a biofilm with fungal hyphae. As observed on the pressed leaves, the leaf cuticles were still intact after 3 months (Fig 1). The pressed leaves consisted of pieces of leaf not preserved for SEM or used for DNA extraction and therefore do not necessarily represent the level of degradation of the full leaves.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Photos and micrographs of water lily leaves over time.

Conventional photos of leaves (left) and scanning electron micrographs of corresponding leaf surfaces (right). Leaves pressed in newspaper immediately after collection at all time points for the control (A–C), kaolinite clay (D–F), pond mud (G–I), and sand (J–L) treatments. Text boxes on the bottom corners of the micrographs describe the primary coverage of the leaf surface.

https://doi.org/10.1371/journal.pone.0315656.g001

The kaolinite clay leaves at all three time points were heavily coated in sediment, which made it difficult to discern any bacteria or fungi in the SEM images. When lifted out of the aquarium, all kaolinite clay leaves simply disintegrated when touched. The pressed leaves samples showed a thick layer of kaolinite clay, and little of the leaf structure was preserved (Fig 1).

The mud leaves at T1 had flakes of mud on their leaf surface, and there were visible remains of bacteria and fungi. The number of bacteria and fungi increased at T2, whereby the leaf surfaces were covered in fungal hyphae with bacteria embedded in the EPS between the fungi. At T3, the leaf surfaces were covered in sediment, much like the kaolinite clay treatment, and it was difficult to discern between bacteria and fungi. However, unlike the kaolinite clay samples, the mud leaves retained intact cuticles which had not completely disintegrated by the end of the experiment (Fig 1).

The sand leaves were covered in bacteria and fungi at T1, and like the leaves in the other sediments, the quantity of fungal filaments had increased by T2. At T3, fungi and EPS coated the leaf surface (Fig 1). The pressed sand leaves were not as degraded as the kaolinite clay leaves but showed more signs of decay compared to the mud and control leaves. Judging from only these visual observations, the sand leaves had also been colonized by microbial communities quite different from those on the mud, clay, and control leaves.

Substrate type impacts microbial community composition

The 16S rRNA (bacteria and archaea) sequencing totaled 9,339,968 high-quality sequences, with an average of 490 ± 298 ASVs per sample. The ITS (fungi) sequencing yielded 2,574,143 high-quality sequences with an average of 126 ± 55 ASVs per sample.

Regarding prokaryotic richness according to substrate type, the richness of the biofilm on the mud leaves was greater (1,576 ASVs) compared to that of leaves of all other substrates. The biofilm richness changed over time, peaking at T2 (Fig 2 and S2 Table). In contrast, Simpson’s evenness remained unchanged between substrates and over time (Fig 2 and S2 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Richness and evenness of microbial communities.

(Left) Richness of (A) bacteria and (B) fungi, and (right) Simpson’s evenness. The first column in blue represents the green leaf control at T0. The other sets of columns plot the values for the control (green) and each substrate (orange, yellow, purple) through time (T1, T2, T3). Dashed lines separate the time trials.

https://doi.org/10.1371/journal.pone.0315656.g002

Fungal richness was significantly lower on the green leaves, compared to the mud and sand leaves (Fig 2 and S2 Table). Richness did not vary on the leaves on the other substrates. As with the bacteria, Simpson’s evenness remained constant in the leaf biofilms across all substrates throughout the experiment (Fig 2 and S2 Table).

Despite similarities in richness and evenness, the types of organisms colonizing the biofilms differed between substrates and over time. The most dominant bacterial phyla included Proteobacteria (44.8%), Firmicutes (19.8%), and Bacteroides (14.3%). It comes as no surprise, given its high richness value, that the biofilms of the mud leaves contained the greatest diversity of phyla (45), and that the low richness value of the green leaves meant that they had the fewest number (23). Ascomycota (12.7%) and Basidiomycota (4.9%) were the most dominant fungal phyla, although a large portion of the fungi could not be unidentified past the domain level (82%).

By identifying the most dominant genera among substrates and in the course of time, we can start to understand the community composition and the ecological role of specific genera within the biofilms (Fig 3). Desulfovibrio was the most abundant genus; it was found primarily on the control, kaolinite clay, and sand leaves at T1 and decreased in abundance over time. Pseudomonas was very abundant on the kaolinite clay leaves at T3, comprising more than 50% of the read abundance. It was also found on the mud leaves at T3. It should be noted that there were few Pseudomonas sequences at the early time points. Tolumonas showed a parabolic trend and peaked in abundance at T2 on both the control and kaolinite clay leaves. Malikia was most dominant on the mud leaves at T1, while Ralstonia was more abundant on the sand leaves at T1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Heatmaps of genera abundance.

Heatmaps for (A) bacterial and (B) fungal genera in each substrate at each time point. Red values indicate a high abundance, while blue values indicate a lower abundance. Each column represents one replicate.

https://doi.org/10.1371/journal.pone.0315656.g003

As stated previously, most of the fungi could not be identified at the generic level. However, among those that could be assigned to a genus, Malassezia was common on the green leaves, control leaves, and mud leaves. Sporobolomyces was abundant across samples at T1. Several genera were found primarily on the green leaves, including Aspergillus, Neoascochyta, Chaetomium, and Coniella. The genera Kazachstania and Phaeosphaeria were abundant on one mud leaf.

Given the differences in biofilm richness of the mud leaves compared to all that of leaves on all other substrates, as well as the variety of genera found in all leaves, it was not unexpected to find that community composition at each time point and across all substrates is unique (Fig 4). Only 4.3% of the bacterial ASVs were found in every substrate (excluding green leaf), compared to 12% of the fungal ASVs shared between substrates (Fig 4). The biofilm of the mud leaves had a more distinctive community composition at every time point, compared to biofilms on leaves on the other substrates. Based on the percentages of shared ASVs and the NMDS plot, the biofilms on the sand and control leaves had similar microbial communities, while those of the mud and kaolinite clay leaves were the most distinct from the green leaf. Based on the ANOSIM, there was a difference in the community composition between experimental conditions involving both substrate and time, although there was large variability in the community composition of a replication sand leaf at T1 which was considered during the interpretation here (S2 Fig). This variability was also evident in the NMDS plot (Fig 4 and S3 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Venn diagrams of shared ASVs.

Venn diagrams showing the number of ASVs shared between each substrate for (A) bacteria and (B) fungi. Sediment type is denoted by oval color: control (green), kaolinite clay (pink), pond mud (yellow), and sand (purple). (C) NMDS plot representing the influence of substrate (colors) and time (shapes) on microbial community composition. The polygons show the connections between samples within the same substrate at all time points. The stress value is located in the top right corner.

https://doi.org/10.1371/journal.pone.0315656.g004

Water chemistry and substrate interact and influence microbial community

The substrate in each aquarium greatly influenced the microbial community on the leaves. Substrate type also impacted the water chemistry in each aquarium, since the substrate added a distinctive set of minerals to the original chemistry of the pond water. The concentration of the water minerals measured varied between substrates and over time. Fe and Mg concentrations increased in the aquarium water with the kaolinite clay and pond mud substrates over time compared to the concentrations in the aquariums with the control and sand substrates. Due to its complex mineralogy, the water in the mud aquarium had greater concentrations of Mn, Ca, Si, and K, as well as higher conductivity compared to the other substrates. Na, pH, and temperature were similar across all substrates (S4 Table). Fe concentration and temperature were higher at T1, while Mg, Na, K, and pH peaked at T2. Furthermore, the concentration of minerals in the water correlated with microbial community composition (Fig 5). They are significantly correlated at T1 and T2, with greater mineral concentrations correlated with the community composition on the mud leaves. Fe, Mn, Si, Ca, K, and conductivity were significantly correlated with the Bray-Curtis distances (Fig 5 and S5 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Ordination plot including water chemistry.

A dbRDA depicting the correlation between water chemistry elements and microbial community composition. Each point represents one sample, which are grouped by substrate (color). A dot or triangle denotes time point T1 or T2, respectively.

https://doi.org/10.1371/journal.pone.0315656.g005

Likewise, several of the most abundant genera on the mud leaves were correlated with mineral concentration (Fig 6). All genera that were significantly correlated with the minerals in the mud substrate showed the same trend, aside from Na. All were negatively correlated with Fe, Mn, temperature, and conductivity, and all were positively correlated with Si, Ca, Mg, Na, and pH. These genera include WCHB1-32, Ruminococcus, Microbacter, Holophaga, Dechloromonas, and Azospirillum.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Correlation plot between water chemistry and genera.

A heatmap showing Pearson’s correlations between the water chemistry elements and the most abundant genera, as grouped by substrate. Red indicates a positive correlation, while blue indicates a negative correlation. * denotes a p-value <0.05, ** p-value < 0.01, *** p-value < 0.001.

https://doi.org/10.1371/journal.pone.0315656.g006

The abundant genera in the kaolinite showed the opposite trend. Desulfovibrio Anaerovorax, and Acidaminobacter were positively correlated with Si, Ca, Mg, Na, K, and pH, and were negatively correlated with Fe and Mn. Only Orbiliaceae gen. incertae sedis was significantly correlated with the concentrations of minerals in the control, and Na was correlated with those of the sand.

The VPA further confirms that substrate and water chemistry influence the microbial community composition (S3 Fig). Substrate alone best explains the variation in community composition between the groups (23.8%). The interaction between substrate and water chemistry (Fe, Si, and K) explains 18.3% of the variation, while water chemistry alone is responsible for 16.3% of the variation. Time produces the smallest effect (4.8%). However, 36.1% of the variation remains unexplained.

Leaves on a mud substrate are a hot spot for biofilm-forming microbes

In addition to characterizing the microbial community composition and assembly processes, we investigated the potential role of biofilm-forming microorganisms in leaf preservation by counting the OTUs correlated with biofilm-producing bacteria in Bugbase. At T1, the control leaves contained the lowest number of biofilm formers, followed by the kaolinite clay, mud, and sand leaves. In Fig 1, biofilms are evident in the control leaves at T1, and based on the microbial analyses, these biofilms were formed by a low diversity of biofilm-formers. Notably, the number of biofilm formers doubled on the mud leaves at T2, while a decline was observed on leaves on the other substrates. The SEM images also show an increase in biofilm abundance in the mud at T2, though the presence of biofilms in T3 is obscured by the substrate coating the leaf surface (Fig 1). By T3, the kaolinite clay leaves had the lowest number of microbial biofilm formers among leaves on all substrates (Fig 7). In addition to having the lowest number of biofilm formers, there was little evidence of biofilms on the kaolinite leaves in the SEM images, regardless of the time point (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. The abundance of biofilm-formers.

The number of OTUs that were classified as biofilm formers by BugBase on the leaves collected from each sediment over time. The different colors denote the control treatment and three substrate treatments.

https://doi.org/10.1371/journal.pone.0315656.g007

Over time, the genera of biofilm-forming microbes underwent notable shifts. OTUs in the fungal classes of Actinomycetes, Bacteroidia, Clostridia, and Spirochaetia primarily contributed to biofilm formation in the control leaves. At T2, only ASVs in the class Polyangia were identified as biofilm formers. By T3, members of the classes Polyangia, Ignavibacteria, and Spirochaetia contained biofilm-forming ASVs. On the kaolinite clay leaves, OTUs from Actinomycetes and Spirochaetia were identified as biofilm formers at T1, while Desulfobacteria, Gammaproteobacteria, Candidatus Kapabacteria, and Verrucomicrobiae contributed to biofilm formation at T2. Interestingly, only the phylum Candidatus Kapabacteria contained biofilm-forming ASVs at T3, albeit in low abundance.

In the mud leaves, biofilm formers were associated with the bacterial classes Alphaproteobacteria, Bacilli, and Clostridia at T1. By T2, Alphaproteobacteria, Bacilli, Bacteroidia, Clostridia, Gammaproteobacteria, Candidatus Kapabacteria, and Syntrophobacteria contained biofilm-forming ASVs. At T3, biofilm formers from the classes Bacilli, Bacteroidia, Clostridia, Gammaproteobacteria, Syntrophobacteria, and Verrucomicrobiae were identified.

Remarkably, the mud leaves had the highest number of bacterial biofilm formers. However, the class Actinomycetes, which was found in all other substrates, was conspicuously absent on all mud leaves. Moreover, the mud leaves uniquely hosted biofilm formers of the classes Alphaproteobacteria, Bacilli, Desulfobulbia, and Syntrophobacteria, underscoring their distinct microbial composition. Conversely, the control leaves stood out as the only ones with biofilm formers from the class Polyangia and no representatives of the Verrucomicrobiae (Fig 7).

Decay patterns and the influence of sediment

Our investigation into microbial interactions and leaf preservation reveals that decay patterns and the microbial community composition of the biofilms on Nymphaea water lily leaves vary with sediment type and over time. Notably, the mud and control leaves exhibited the least decay over three months, while the kaolinite clay leaves underwent substantial decay. Furthermore, microbial community analysis reveals differences in the composition and assembly of microbial biofilms on each substrate over time.

Contrary to our results, previous research with animal carcasses found that carcasses buried in kaolinite had better preservation [67, 68], perhaps resulting from low bacterial abundance and diversity due to the interactions between the microbial community and the clay minerals, that resulted in lower bacterial abundance and diversity [68, 69]. However, similar to our results, a study with scallop carcasses buried in kaolinite indicates that burial in kaolinite leads to decay as muscles in kaolinite lost more mass compared to those buried in sand [70]. This strongly suggests that the sediment interactions, and consequently interactions with the microbial community, vary between organisms and potentially lead to substrate-selective preservation. Thus, it is important to understand the microbial community dynamics for individual groups of organisms.

The variations in bacterial genera across the different substrates and time points underscore the nuanced response of freshwater microbes to changing environmental conditions. Pseudomonas, while it appeared later, was abundant in the kaolinite clay and mud leaves, indicating its involvement in advanced stages of tissue decay. It is well-known that some strains of Pseudomonas are plant pathogens and can cause cankers, leaf and stem spots, blight, soft rot, and galls [71]. A previous experimental taphonomic study reported that Pseudomonas will form a biofilm around marine embryos, and while the bacteria will then consume the organic matter, the biofilm will retain the shape of the embryo [72].

Furthermore, some species of Pseudomonas can also form biofilms that replicate the external morphology and internal structure of embryos [73]. The appearance of Pseudomonas in the second month of our experimental decay series suggests that at this point, the environment provided the ideal conditions for Pseudomonas growth and may have prompted different responses from the original microbial community on the leaves. On the kaolinite leaves, Pseudomonas proliferated and became the most abundant genus in the biofilms. During this time, we also observe more decay on the kaolinite leaves compared to leaves on the other sediments. The biofilms on the mud leaves also contained Pseudomonas, though fewer ASVs. This indicates that although Pseudomonas was present on the mud leaves, the effect of decay may have been modulated by the high richness of other microbes on the leaves.

These other microbes include Malikia, Tolumonas, Uliginosibacterium, and Ralstonia, for example. Malikia, a biofilm former, was abundant on the mud leaves at T1. Elsewhere, it has been reported that Malikia performs well in aerobic environments [74]. Tolumonas, which was most common bacterium on the control and kaolinite clay leaves at T2, can be facultatively anaerobic and grow in anoxic freshwater [75, 76]. Uliginosibacterium was particularly abundant on the control and sand leaves at T2 and T3 and has previously been described from freshwater environments, such as sediment [77, 78] and lakes [79, 80]. The most abundant genus on the green leaves (but not in the aquariums) was Ralstonia, a genus known to cause plant diseases [81, 82]. That the bacterial communities were so variable between the sediments indicates that each sediment type attracted a differently adapted bacterial community.

Fungal communities exhibited different patterns over time and between sediment types, with Malassezia prevalent on green, control, and mud leaves. Malassezia is a pathogenic genus often found on the skin of animals [83]. The pond water used in our experiment came from a botanic garden freely open to the public, which may have been the source of the variety of human pathogens found growing on the leaves in the aquariums. In the first month, Sporobolomyces emerged as an abundant genus across all samples. It is a yeast that produces a red carotenoid pigment that has been shown to exhibit antimicrobial properties [84]. Furthermore, Sporobolomyces inhabits the phyllosphere [85] and may have come from within the leaf, instead of originating from the pond water. Other genera, namely, Aspergillus, Neoascochyta, Chaetomium, and Coniella, were found predominately on the green leaves, highlighting the diversity and abundance of fungi from the phyllosphere that may, in turn, impact the biofilm community during leaf decay.

Interestingly, the influence of elemental concentrations, particularly Fe, Mn, Si, Ca, K, and conductivity, on microbial community composition was significant, emphasizing the intricate relationship between environmental factors and microbial dynamics. Based on the correlation heatmap between water chemistry and the microbial community composition, communities on both the mud and kaolinite clay leaves correlated with several minerals in the water. However, these correlations were not necessarily the result of the ion concentrations but rather may have been influenced by the nutrients and organic matter present in the substrates. The pond mud was from a natural environment and may have contained more organic matter and nutrients that promoted the growth of different microbial communities, while the commercially purchased sand and kaolinite may have contained less naturally occurring organic matter and nutrients.

On the mud leaves, a correlation exists between the presence of Paludibacter, Microbacter, Holophaga, Dechloromonas, and Azospirillum with all minerals: there is a positive correlation with Si, Ca, Mg, Na, and pH, and a negative correlation with Fe, Mn, temperature, and conductivity. Paludibacter, Microbacter, and Holophaga, which were slightly more abundant on the mud leaves at T2, are anaerobic bacteria that excrete propionate and acetate [86–88]. Dechloromonas is a (per)chlorate-reducing genus [89], while Azospirillum is a microaerophilic, non-fermentative, and nitrogen-fixing genus [90]. The genera that correlate with the minerals in the water suggest that the environment on and within the biofilms was conducive to fermentation and anoxic lifestyles. Elsewhere, it has been noted that the excretion of acids may facilitate mineralization and aid in the preservation of soft tissues [91, 92].

On the kaolinite clay leaves, Desulfovibrio, Anaerovorax, and Acinetobacter were all positively correlated with Fe, Mn, and temperature. The dominance of Desulfovibrio, a sulfate-reducing bacterium, on the control, sand, and kaolinite clay leaves suggests these substrates provided an environment for Fe precipitation in the form of iron–sulfide solids (FeS) [93]. Previous studies have hypothesized that iron precipitation is a key process for leaf preservation, based on the observation of iron-encrusted plant fossils [6, 13]. However, the mud leaves—which had the least visible decay after three months—had lesser amounts of Desulfovibrio. Anaerovorax is strictly anaerobic bacterium with a fermentative metabolism, which excretes acetate and butyrate [94]. The genus Acinetobacter, on the other hand, can readily form biofilms [95] and furthermore can oxidize magnesium [96].

The application of the VPA further delineates the impact of substrate and water chemistry on microbial community composition. While the type of substrate is the most important factor, the interaction between substrate and water chemistry substantially contributes to much of the variation between the experimental treatments. Surprisingly, the variable of time exerted a relatively minor effect on community composition, reflecting the resilience of microbial communities to temporal changes.

Our study also identifies key biofilm-forming microbial OTUs associated with each substrate. The biofilms on the kaolinite clay leaves, for example, contained the sulfate-reducing, biomineralizing biofilm formers of Desulfobacteria [97, 98]. These biofilms also contained Candidatus Kapabacteria which has been found in anoxic, high-sulfide, wetlands [99], as well as in microbial mats in hot springs [100], indicating their ability to tolerate hostile environments. However, based on our observations of the rapid decay of the kaolinite leaves, it appears that the biofilm microbial community does not form a protective layer on the leaves that would prevent decay and encourage the formation of a mineral veil in the early taphonomic stages.

On the other hand, the genus Clostridia was also a prominent biofilm former on the mud leaves. Previous taphonomic work found Clostridia in the biofilms on decaying crayfish and hypothesized that Clostridium may be involved in adipocere formation and, consequently, instrumental in the pyritization of soft tissue fossils [101]. In leaves with less evident decay, like those in the mud, it is likely a consortium of microbes will tip the scales from decay towards a protective layer.

The shifting prevalence of biofilm formers over time and among different substrate types further emphasizes their dynamic roles in soft tissue preservation. It is worth mentioning again that the leaves on the mud substrate emerged as a hotspot for biofilm formers in our decay experiments. This, in combination with our observations of less morphological decay on the pressed mud leaves, suggests the biofilm-forming microbes likely slowed down leaf decay. Hence, mud substrates appear to support water lily leaf preservation better by facilitating the stronger development of more microbially diverse biofilms than kaolinite or sand clay substrates. The microbial-mediated biofilms would also facilitate mineral precipitation on leaf surfaces that will enclose and protect the leaf tissue from abrasion and damage during transport and from invertebrate herbivory, as well as slow down decay from further bacterial activity and preserve the fine-scale morphological features of the plant tissue [12, 102].

Limitations of the experimental design

Although this multi-aquarium experiment was successful in elucidating the role of sediments on the biofilm microbial community of water lily leaves, there are important limitations to consider. First, microbial communities are highly variable, and the community composition and the environmental conditions can change over time which could impact replicability. Future work should perform the experiment over multiple timeframes throughout the year to identify changes in the microbial community and the effect of the substrate. Second, the sand and kaolinite were purchased commercially for this experiment while the organic-rich pond mud was derived from a nearby pond. In order to improve applicability of this experiment to natural conditions, future work should use naturally derived substrates of varying mineral compositions. Furthermore, though this work focused on the microbial composition of the biofilms, future experiments should address how sediment impacts the mineral content within the biofilms, as this likely differs between the organic-rich pond mud and the other sediments. This would also allow future researchers to test the role of organic matter in leaf fossilization. Finally, this study emphasizes the role of biofilms and sediment in the preservation of water lily leaves. However, the underrepresentation of water lily foliage in the fossil record is due in large part to rapid decay [38]. Future experiments should be performed with other plant leaves abundant in the fossil record to identify if these patterns are consistent across plant species.

The preservation of water lily leaves in the fossil record

Here, we provide evidence that biofilm formation and biofilm microbial communities differ depending on the substrate in the early stages of leaf decay. These differences in community composition and the consequential metabolisms of the microbes could slow decomposition long enough for the formation of a mineral veil that can protect the leaf until it enters the ideal conditions for fossilization [13]. This veil may not become preserved in the fossil record, but act as a protective barrier until the conditions for preservation are met. Thus, the occurrence of water lilies in the fossil record may be influenced by the substrate type and, consequently, the biofilm microbial community on their leaves during decay.

In our survey of well-documented occurrences of water lilies in the geological record (Table 1), the fossil leaves are nearly always found in sedimentary rocks such as marlstone, claystone, mudstone, and siltstone [103]. Common to most examples, too, is the occurrence of a dark, carbonaceous, or organic-rich rock matrix, which may represent the muddy substrate at the bottom of an ancient pond or lake.

The preservation of the fossil water lily leaves in such sediments does not come as much of a surprise, given the biofilm-enhancing properties of a pond mud substrate, which is attributed to the mud substrate’s rich and complex mineralogy. It is also perhaps not a coincidence that the site with the greatest number of water lily specimens, totaling over 39 leaves, bears the extremely dark, oil-bearing shales of the middle Eocene Lake Messel in Germany [50] and is the most organic-rich locality of all those described here with fossil Nymphaeales and Nymphaeaceae foliage.

Thus, the sedimentology of these water lily leaf-bearing sites implies that a substrate such as pond or lake mud may be the most conducive for the fossilization of nymphaealean leaves. As shown, a mud substrate containing a complex suite of minerals may support the preservation of leaf tissues better than other substrates by facilitating the stronger development of microbially diverse, biomineralizing biofilms. To produce well-preserved fossil leaves, the development of strong, protective biofilms in the earliest stages of leaf decay, before the onset of substantial tissue degradation, may be crucial. The need for a protective biofilm has also been emphasized in several other studies [15, 18].