Oxygen-dependent biofilm dynamics in leaf decay: an in vitro analysis

Biofilms are important in the natural process of plant tissue degradation. However, fundamental knowledge of biofilm community structure and succession on decaying leaves under different oxygen conditions is limited. Here, we used 16S rRNA and ITS gene amplicon sequencing to investigate the composition, temporal dynamics, and community assembly processes of bacterial and fungal biofilms on decaying leaves in vitro. Leaves harvested from three plant species were immersed in lake water under aerobic and anaerobic conditions in vitro for three weeks. Biofilm-covered leaf samples were collected weekly and investigated by scanning electron microscopy. The results showed that community composition differed significantly between biofilm samples under aerobic and anaerobic conditions, though not among plant species. Over three weeks, a clear compositional shift of the bacterial and fungal biofilm communities was observed. The alpha diversity of prokaryotes increased over time in aerobic assays and decreased under anaerobic conditions. Oxygen availability and incubation time were found to be primary factors influencing the microbial diversity of biofilms on different decaying plant species in vitro. Null models suggest that stochastic processes governed the assembly of biofilm communities of decaying leaves in vitro in the early stages of biofilm formation and were further shaped by niche-associated factors.


Bacterial biofilm communities
Taxonomic richness was measured as alpha diversity (observed number of ASVs) and Simpson's evenness for biofilm communities under both oxygen conditions (Fig. 2).Alpha diversity of bacterial biofilm communities increased over time (ANOVA, p < 0.001) in the aerobic assay, and decreased under anaerobic conditions.Differences in prokaryotic richness between aerobic and anaerobic conditions were statistically significant (ANOVA, p < 0.021) with the higher observed richness of samples in aerobic conditions.Conversely, the evenness was similar between both oxygen conditions.
The alpha diversities of prokaryotic biofilm communities from the three different plants were not statistically different and they shared 20.1% of all ASVs (Fig. S4B).Biofilms from Acer leaves had the highest percentage of exclusive ASVs (1043 ASVs, 23.2%) compared to Lonicera (966 ASVs, 21.5%) and Hedera (880 ASVs, 19.5%) plants.Interestingly, the richness of the total biofilm plant communities was much lower than the diversity of planktonic bacteria in the water (Fig. 2 & Supplementary Fig. S5).
In total, 1209 ASVs (26.9%) were unique to all plant-biofilm samples at the beginning of the experiment, while 756 ASVs (16.8%) were exclusive to microbial communities sampled after three weeks of study (Fig. 3A).The Venn diagram shows that microbial communities changed over time with a consistent core number of community members (257 ASVs, 5.7%) which were present during the whole experiment (Fig. 3A).
Unconstrained ordination of the prokaryotic biofilms (Bray-Curtis distance-based PCoA-Principal Correspondence Analysis) revealed that the bacterial communities were structured by the oxygen conditions (Supplementary Fig. S6A).The PERMANOVA analysis (p = 0.021) confirmed that leaf biofilms were significantly different under aerobic and anaerobic conditions (Table S1).In addition, leaf biofilm communities varied significantly compared to the planktonic microorganisms in the fresh lake water that had been used as inoculum and medium in the experiment (Supplementary Fig. S6A).
The changes in bacterial biofilm community over time (Fig. 3B) indicate that all biofilm samples followed the same temporal trajectory, although there was no significant difference in microbial biofilm communities between the plants (p = 0.235).
Although differences in the biofilm community structures between treatments and during the experiment were evident, the microbial communities contained many shared taxa (Fig. 4).Taxonomic analysis of the 16S rRNA ASVs showed that Proteobacteria was the dominant phylum, representing 50% of the total prokaryotic biofilm communities (Fig. 4, Supplementary Fig. S7) followed by Firmicutes, Bacteroidota, Desulfobacterota, Verrucomicrobiota, and Actinobacteriota.In addition, Acidobacteriota, Cyanobacteria, and Patescibacteria were present with relative abundances > 1.5% of the total bacterial communities in the 72 biofilm samples (Supplementary Fig. S7).
The genera Vogesella, Tolumonas, Rhodoferax, Sphaerotilus, and Aeromonas decreased in relative abundance over time under aerobic and anaerobic conditions, whilst Clostridium, Desulfovibrio, and Caproiciproducens increased (Fig. 4, Supplementary Fig. S7).Furthermore, ASV1 within Prevotellaceae, and ASV2 assigned to Acidaminococcaceae, Paludibacter, Anaerosius, and Rhizobium increased in abundance over time.Some orders, most notably Bacteroidales, Ocillospirales, Acidaminococcales, and Clostridiales were more abundant in anaerobic conditions.Overall, only 2 ASVs within Rhodoferax and Aquabacterium were observed in high abundance (> 1%) in the plant biofilms and water samples with a decreasing trend over time.The exception is Aquabacterium which increased abundance over experimental time in the Acer biofilm samples.
To investigate how time and oxygen conditions affected the abundance of individual ASVs in leaf biofilms, we used DESeq2 which determines ASVs with statistically significant differences (p < 0.01) in abundance over time (Fig. 5).
Several ASVs within Undibacterium and Rhodoferax were prevalent in the plant biofilms at the beginning of the experiment.In addition, Hypnocyclicus and Flectobacillus ASVs were highly abundant in many biofilm samples and decreased with time.After three weeks of the experiment, leaf biofilms in vitro were enriched with Anaerosius, Clostridium, Terrimicrobium, Desulfovibrio, Ruminiclostridium, Rhizobium, Acetobacteroides and Azovibrio.Comparing the different oxygen conditions, several ASVs classified as Paludibacter, Caproiciproducens, and Clostridium sensu stricto 1 were significantly more abundant under both oxygen conditions (Fig. S8).

Fungal biofilm communities
Fungal communities shared 147 ASVs (16.9%) among different plant species with the highest number of unique ASVs (32.2%, 280 reads) belonging to the Lonicera biofilm samples.Moreover, 31.9% (277 ASVs) were shared between all leaf biofilm samples from both experimental treatments, but only 83 ASVs (9.6%) were present during the entire duration of the experiment in biofilms on all leaves (Fig. 6A, Supplementary Fig. S9).www.nature.com/scientificreports/Beta diversity PCoA plots revealed that fungal biofilm communities did not differ between oxygen treatments (Supplementary Fig. S6B) or between different plants.However, the fungal community changed over time in a Bray-Curtis distance matrix (Fig. 6B).
The alpha diversity of leaf biofilms in vitro showed significant changes in the fungal community over time.Fungal richness decreased during the aerobic incubation (ANOVA, p < 0.002), whereas it remained relatively stable under anaerobic conditions during the three weeks of the experiment (Fig. 7).In addition, the low fungal biofilm diversity was confirmed by an average Simpson index lower than 0.9.The Simpson indices for fungal leaf biofilms of different plants were Hedera > Lonicera > Acer (Supplementary Fig. S5B).Fungal richness was lower in the water samples than in the leaf biofilms.
Ascomycota was the dominant phylum in leaf biofilms in both treatments with a relative abundance of 78.51%, followed by Basidiomycota (10.43%).At the genus level, a high abundance of Aureobasidium, Alternaria, Taphrina, and Dothiora was observed in all leaf biofilm samples.Furthermore, the ASV5062 within Pleosporales and one unclassified ASV (phylum incertae sedis) were detected in high abundance across all leaf biofilms (Fig. 8).
Interestingly, the abundance of ASVs assigned to Laetiporus and Trametes increased in leaf biofilms over time in both oxygen treatments.According to DEseq2 (Fig. 9), Trametes, Sarocladium, Laetiporus, Cystobasidium, and Penicillium were enriched in the leaf biofilms after three weeks of the experiment.Furthermore, ASVs affiliated with the major plant pathogenic fungi Alternaria and Ramularia were dominant in the leaf biofilms at the beginning of the experiment.Alternaria increased in abundance over time except for Lonicera biofilms from anaerobic assays.In contrast, five ASVs assigned to the fungus Ramularia decreased in abundance over time.

Microbial community assembly
Our results showed that stochastic processes (β-NTI < 2, ANOVA, p < 0.01, Fig. S10) determine the assembly of leaf biofilm communities.The null model of betaNRI (Net Relatedness Index) revealed temporal phylogenetic turnover in bacterial and fungal leaf biofilm communities (Supplementary Fig. S10).To understand which stochastic processes dominate the assembly of leaf bacterial and fungal biofilm communities, we used Bray-Curtis-based Raup-Crick calculations within the 'microeco' package.The results showed that drift was the dominant ecological process driving the colonization of the leaves on day 1 by prokaryotic and fungal communities (Fig. 10, Supplementary Table S2).Later the deterministic factors shaped the formation of the bacterial leaf biofilms (47.78-60.95%)and fungal biofilm communities (72.38-77.77%).Moreover, homogeneous dispersal (26.19-34.47%)and dispersal limitations (9.84-13.33%)contributed to the bacterial biofilm community succession.In addition, deterministic processes like variable and homogeneous selection influenced bacterial and fungi leaf biofilm assembly.However, normalized stochasticity ratio (NRT) and standard effect size (SES) results indicate that deterministic factors were dominant in mature biofilms after one week of the experiment.Interestingly, the relative contribution of niche-based deterministic processes (variable and homogeneous selection) in the assembly of leaf biofilms was higher for fungal communities compared to prokaryotes.

Discussion
In this study, we detailed differences in the successional dynamics of the bacterial and fungal biofilm communities in aerobic and anaerobic in vitro environments.The results of this study show that the availability of oxygen and incubation time were the main factors influencing the microbial diversity of biofilms on different decaying plant species.The type of leaf, whether evergreen or deciduous and its biological affinity did not play an important role.

Changes in microbial richness and diversity during biofilm formation
The biofilm communities on the leaves were distinctly different from the planktonic community in the fresh water at the beginning of the experiment with a significantly higher abundance of bacterial and fungal taxa which were not observed in the water.This suggests that members of the leaf-colonizing microbiota were early biofilm-forming microorganisms that established a primary biofilm environment for the subsequent colonizers.However, it should be noted that plant leaves had been incubated for 24 h before the first sampling, it is possible that a certain number of microorganisms from the water colonized the surface of the leaves, but they were not present in high abundance.Interestingly, the bacterial richness and diversity of leaf biofilms in our study were lower than the bacterial richness in the water.This result agrees with previous studies where bacterial community diversity of plant phyllosphere samples was lower than in soil, marine environments, or, the rhizosphere 17 .However, the microbiome of harvested leaves used in this experimental study in vitro cannot be compared with the microbiome of decaying or fresh leaves in nature due to the closed environment of the study.At the beginning of the experiment, the bacterial biofilm community on the leaves of each plant species consisted primarily of Proteobacteria and Firmicutes.Over time, Bacteroidota and Desulfobacterota increased in abundance under both oxygen conditions and it has to be assumed, that the oxygen was partly depleted during the incubation also in the aerobic experiment and especially in the lower layers of the biofilms.This result supports previous studies where Proteobacteria predominate the phyllosphere of distinct plant species [18][19][20] and Firmicutes and Bacteriodota increased during the decomposition of the macrophyte leaf litter 20 .Interestingly, the alpha diversities of prokaryotic and fungal biofilm communities from the three different plants were not statistically different.
We observed a significant increase in bacterial alpha diversity richness over time and a decrease in diversity for the fungal biofilm members.Various studies showed an increase in alpha diversity and abundance over time under aerobic conditions that follows patterns of species-time relationships observed in bacterial communities of phyllosphere and many other microbial ecosystems 19,21,22 .As the biofilm matures, bacteria and fungi start to metabolize the cell polymers and excrete metabolites creating more environmental niches and causing changes in the biofilm community 23 .www.nature.com/scientificreports/

Biofilm development and succession of bacterial communities
In this study, species of the genera Tolumonas, Aeromonas, Vogesella, Rhodoferax, Sphaerotilus, Idionella, Paucibacter, Aquabacterium, Undibacterium, and Flectobacillus were potential biofilm formers.The genus Aeromonas is a strong decomposer of soft tissues as well as a biofilm former 24 .These facultatively anaerobic bacteria were observed in high abundance during the initial stages of the experiment in all plant biofilm samples.Domination of Aeromonas species during initial decay processes has been reported in previous studies about the decomposition of soft tissue materials 14 .In addition, filamentous bacteria of the genus Sphaerotilus and motile, microaerophilic species of Aquabacterium 25 , capable of accumulating large amounts of slime, were detected in high abundance during the first three weeks of study in the aerobic and anaerobic samples.The sheaths of Sphaerotilus may be encrusted with iron and manganese oxides [26][27][28] .
The ASVs that were assigned to iron-oxidizing bacteria within Rhodoferax and Undibacter were enriched exclusively on the first sampling occasion, suggesting that a lack of iron ions in the incubation medium caused a microbial shift within leaf biofilms in vitro.Over time, many ASVs that had been abundant at the beginning of the experiment decreased in abundance while other ASVs increased, thus establishing more diverse biofilm communities.
After one week of incubation when the biofilm layers were well visible, Clostridium, Desulfovibrio, Propionivibrio, Acidaminococcaceae, Paludibacter, and Acetobacteroides had increased.Desulfovibrio is a sulfate-reducing genus that typically grows anaerobically, but certain strains tolerate the presence of oxygen.Interestingly, in an anoxic environment, Desulfovibrio can oxidize organic matter and generate hydrogen sulfide as a final product, which may react with iron and lead to the formation of iron sulfide and further formation of pyrite 29,30 .From fossilization studies, it is known that pyritization is important for the preservation of plant tissues 16,31,32 .
Although Paludibacter and Clostridia were enriched under both oxygen conditions, fermentative bacteria were dominant in anaerobic samples.The genus Clostridium is widely distributed in soils and plants, but only a few studies have shown that diverse strains can colonize plants 33 .In our experiment, several clusters of Clostridium were observed in the biofilms including Clostridium sensu stricto 1, Clostridium sensu stricto 11, and Clostridium sensu stricto 12. Sarkar et al. ( 2022) emphasize that clostridia are effective H 2 and acid producers using carbohydrates such as cellulose and sucrose as substrates 34 .Numerous studies showed that saprophytic clostridia are associated with the decay of plants, phytoplankton, animal and human tissues 14,35 .
The genus Acetobacteroides also increased in the Acer leaf biofilm samples over time.These carbohydratefermenting bacteria produce acetate by degrading butyrate and propionate comparable to some clostridia.Previous studies showed that acetobacteria are involved in plant and animal decay processes when an anaerobic environment has been established within the biofilm or in the environment 14,36 .Moreover, ASVs belonging to the fermentative family Prevotellaceae and the genus Caproiciproducens were observed in higher abundance in the anaerobic incubation in the later stages of the experiment.Prevotella is associated with glycan degradation 37 , whilst members of the genus Caproiciproducens can produce acetic acid, butyric acid, caproic acid, lactic acid, and H 2 as metabolic products 38,39 .In addition, ASVs assigned to Acidaminococcaceae, which use numerous amino acids as their carbon sources, were detected in the leaf biofilms from anaerobic assays.After three weeks, nitrogen-fixing bacteria Rhizobium and Xanthobacter were enriched in the leaf biofilm samples (Fig. S11).Both genera can produce large amounts of EPS and enhance the formation of calcium carbonate 40,41 .

Biofilm development and succession of fungal communities
Fungi have major roles in the decomposition of organic matter and plants 20,42 .Numerous studies showed that they are involved in mineral and metal deposition, transformations, and biomineralization [43][44][45][46] .In this study, fungi detected in the leaf biofilms were mainly saprotrophs and pathotrophs (Supplementary Fig. S12).
A PCoA plot based on the Bray-Curtis dissimilarity (Fig. 6B) and betaNRI plots (Supplementary Fig. S10) revealed a succession of fungal communities with the highest betaNRI in biofilms sampled after three weeks of the experiment.We observed that fungal richness decreased over time with higher Simpson´s evenness in the aerobic incubation.Therefore, we can assume that time was the main factor shaping fungal biofilm communities.The fungi Laetiporus and Trametes exert a strong antimicrobial effect 47 and were enriched in the biofilms at the later stages of the experiment.Members of both genera are wood-decaying fungi that decompose cellulose, hemicellulose, and lignin 48 .Furthermore, Trametes strains secrete extracellular fatty acid droplets 49 and corrode apatite under glucose-limited conditions 50 .Graz et al. 2023 confirmed the oxalic acid-degrading activity of Trametes and Laetiporus 51 .Oxalic acid is commonly secreted by the majority of fungi during the wood decay process, and it is linked with the ability to chelate metals 51,52 .The endophytic fungi Sarocladium and Penicillium were enriched in the biofilms according to Deseq2.These fungi are common plant pathogens, but several studies observed enhanced stress tolerance and plant growth-promoting activity 53 .In addition, Alternaria and Aureobasidium were prevalent in all leaf biofilm samples during the experiment.Aureobasidium is a ubiquitous yeast saprotroph that can promote plant growth 54 and produces volatile organic compounds with antifungal properties against Alternaria 55 .Interestingly, Hou et al. 2011 observed precipitations of calcium carbonate by Alternaria during nitrate removal from the medium 56 .

Ecological processes governing the leaf biofilm assembly
The betaNRI values of biofilm communities under both oxygen conditions at the beginning of the experiment were, on average, more phylogenetically dispersed or random, and, after one week of the experiment, biofilm communities were significantly phylogenetically clustered.
In our experimental study in vitro, in addition to the leaf-colonizing flora, the harvested leaves were all inoculated with saprophytes from the lake water used as medium.Therefore, it could be assumed that microbial biofilm communities from different plant species would be phylogenetically clustered over time.In contrast, plant Vol:.( 1234567890 www.nature.com/scientificreports/microbial communities in nature often diverge over time 57 .Here, our results showed that stochastic processes dominated the assembly of leaf biofilm communities under aerobic and anaerobic conditions at the beginning of the experiment (Supplementary Fig. S10, Supplementary Table S2).Bray-Curtis-based Raup-Crick calculations confirmed that 77% of turnover in bacterial and 61% of turnover in fungal biofilm communities was assigned to drift.It is well-known that ecological drift is attributed to random processes like immigration or emigration of microbial taxa, and pronounced under conditions where microbial diversity is low 58,59 .A study of bacterial community temporal dynamics confirmed that stochastic processes were important during the colonization of the phyllosphere 60 .However, in our study, the standard effect size of non-random processes on leaf biofilm compositional dissimilarity (Bray-Curtis) was higher than two when compared between plants, between treatments, and between sampling occasions except for the first day of the experiment (Supplementary Table S2).This means that the observed biofilm compositional dissimilarity was higher than expected by a random assembly.Since decaying leaves in our study in vitro were inoculated from the same environmental microbial pool, homogeneous dispersal likely mediated the similarity of leaf biofilm microbial communities.However, we assume that drift was the major process at the beginning of the experiment when plant leaves were inoculated by microorganisms from the lake freshwater, whilst dispersal and selection further shaped biofilm microbial communities over time.Interestingly, several studies showed the same pattern during succession, where the early stages of biofilm formation were influenced by stochastic factors, while selection was dominant in mature biofilms 21,61 .

Conclusions
This study characterizes microbial community composition and temporal dynamics of decaying leaves under varying oxygen conditions in vitro.Biofilms have a major role in the decay and preservation of soft tissues and therefore understanding underlying microbial community structure and dynamics under different environmental conditions is important.We used harvested leaves to explore the microbial biofilm communities and the mechanisms shaping community assembly processes during decay under aerobic and anaerobic oxygen conditions.
Our results indicate that microbial biofilms were not influenced by plant species, but that oxygen conditions and time were the main factors shaping bacterial and fungal biofilm communities.In succession, bacterial communities increased in alpha diversity and formed distinct biofilms under aerobic and anaerobic conditions.The assembly of biofilm communities of decaying leaves in vitro was primarily governed by stochastic processes at the early stages of biofilm formation and further shaped by niche-based deterministic factors.The presence of decomposer microorganisms during biofilm formation and their succession with other bacteria and fungi capable of fermentation and biomineralization indicated the presence of niches that enable both processes.These results will act as a starting point for further investigations of leaf biofilms within the biomineralization and preservation context.

Sampling and experimental setup
The freshwater used in this study was collected at the Tongrube Lake, Röttgen, Germany (50°40′24.7′′N/7°04′29.6′′E).Tongrube Lake has a natural water inflow and outflow, and it is one of the largest water bodies in Bonn, Germany, with a maximum depth of six meters and an area of 1700 square meters.The lake bottom consists of a clay substrate.Water samples were collected in 1 L sterile glass bottles, then kept at 4 °C during transport to the laboratory.The freshwater was filtered through a sterilized coarse filter (grade 3hw; Binzer & Munktell Filter GmbH, Battenberg, Germany) for the removal of debris, sediment, and larger organisms.
Deciduous leaves of a maple tree, Acer sp., and evergreen leaves of ivy, Hedera helix, were gathered from plants growing around the lake in Röttgen, while evergreen leaves of Henry's honeysuckle, Lonicera henryi, were collected in a privately owned garden.We collected approximately 30 leaves of each plant species with sterile gloves and placed them into sterilized glass petri dishes.Plant leaf collection complied with relevant institutional, national, and international guidelines and legislation.
Both experimental treatments in this study were conducted at a constant temperature of 24 °C in a laboratory incubator in the dark for three weeks.We immersed 72 leaves in lake water in sterile glass petri dishes.Triplicates were kept together in the same petri dish.Half of the samples were exposed to aerobic conditions, and the other half was placed in anaerobic containers (BD GasPak™ EZ container).The lake water used for the anaerobic setup was treated with N 2 gas for 20 min and anaerobic conditions were established with a gas generating system, BD GasPak ™ (BD BBL™ CO 2 gas generators).Sampling in triplicates of each leaf species was performed after 24 h of incubation, then after 7 days, 14 days, and 21 days.Water samples in triplicates were taken directly from the lake on the first day of the study.To document biofilm formation and plant decay, all samples were examined and photographed using a Zeiss Stemi 2000 stereo microscope and a Tescan Vega 4 LMU Scanning Electron Microscope (SEM) after each sampling session.
A total of 72 leaf samples and three water samples were collected for DNA analysis.All samples were manually shaken in 50 ml of 1 × TE buffer and then sonicated in an ultrasound bath (Branson 1210 Ultrasonic Cleaner; Emerson; St. Louis, Missouri, USA) for two minutes to separate the biofilm and microorganisms attached to the leaf surface.Then, planktonic microorganisms from the freshwater samples or detached biofilm microorganisms from the plant samples were collected on PORAFIL NC filters (0.25 µm, Macherey-Nagel GmbH & Co, KG, Germany) using vacuum filtration.The filters were stored at − 20 °C until nucleic acid extraction.

DNA extraction, amplification, and sequencing
The total genomic DNA of each sample was extracted following the manufacturer's protocol using a Fast DNA spin kit for soil (MP Biomedicals), which had been found to be optimal for the extraction of DNA from biofilms on decaying leaves 24  www.nature.com/scientificreports/DNA was eluted in 50 µl Dnase/RNase-free water and the concentration of the extracted nucleic acids was measured using a Qubit 4.0 fluorometer (Thermo Fisher Scientific, USA) with the dsDNA HS assay kit (Invitrogen).The V4 hyper-variable region of the 16S-rRNA gene was amplified in duplicates, using the forward primer 16 s-515F (GTG CCA GCMGCC GCG GTAA) and the reverse primer 16 s-806R (GGA CTA CVSGGG TAT CTAAT) to cover both bacteria and archaea 62,63 .For the determination of the fungal rRNA sequences, the ITS region was amplified using the ITS1F (CTT GGT CAT TTA GAG GAA GTAA) and ITS2R (GCT GCG TTC TTC ATC GAT GC) primers 64 .PCR was conducted using the HotStarTaq Plus Master Mix Kit (QIAGEN) and carried out as follows: initial denaturation at 95 °C for 5 min, denaturation at 95 °C for 30 s, annealing at 53 °C for 40 s, and extension at 72 °C for 1 min for 30-35 cycles, followed by final elongation at 72 °C for 10 min.Paired-end sequencing (bTEFAP) was performed by Molecular research DNA (Shallowater, Texas, USA; http:// www.mrdna lab.com) on an Illumina MiSeq sequencing platform (Illumina, San Diego, California, USA) following the manufacturer's instructions 65 .16S-rRNA and ITS regions of biofilm and water samples were sequenced in triplicate (n = 72 from the leaves, n = 3 from the water).

Bioinformatics and statistical analysis
The sequence reads were processed using the QIIME2 66 .Sequence quality control, denoising, and chimera filtering were performed using the DADA2 v.1.16 67pipeline which determines amplicon sequence variants (ASVs).Taxonomy for the 16S rRNA sequences was assigned using the Silva database v.138 68 , whilst the Unite database was used for the ITS sequences 69 .The dataset was rarefied by subsampling each sample to 46,232 bacterial reads and 1153 fungal reads.ASVs and taxonomy derived from the Qiime2 and DADA2 pipelines for each database were merged using R version 4.2.1 70 .Multivariate statistics, diversity calculations, and data visualization were conducted within the R packages ampvis2 (version 2.7.3) 71 , vegan (version 2.5.1) 72, phyloseq (version 1.4.0) 73, picante 1.6.2 74, ecodist (version 2.0.9) 75, microeco (version 0.20) 76 , and ggplot2 (version 3.4.2) 77 .Beta diversity was estimated using pairwise Bray-Curtis dissimilarities and visualized using principal coordinate analysis (PCoA).The statistical differences in microbial community compositions were tested using permutational multivariate analysis of variance (PERMANOVA, adonis2 function).Additionally, differences in community richness and evenness were tested by a univariate statistical test Shapiro-Wilk's followed by a post-hoc Games-Howels test.The DESeq2 method 78 was used to test the differential abundance of ASVs over time and between different plant species and was carried out in R. Functional assignments of fungal traits at genus levels were determined using the FungalTraits database 79 within the "microeco" package.

Scanning electron microscopy (SEM)
A small piece of every leaf was excised with a sterile scalpel for SEM.These fresh leaf samples were fixed in 70% v/v ethanol + 4% v/v formaldehyde in water for at least 24 h.Later, samples were dehydrated with ethanol and stored in 100% ethanol until critical point drying.Samples were dried in 22 cycles using a Leica EM CPD300 Critical Point Dryer at the Nees Institute for Biodiversity of Plants.Dried samples were coated with a thin layer of palladium with a sputter coater (Cressington Sputter Coater 108 manual, Tesca GmbH, Germany).Coated samples were mounted on carbon tape on individual stubs and investigated under high-vacuum mode using an SE detector at an acceleration voltage of 20 kV in a Tescan Scanning Electron Microscope.

Figure 1 .
Figure 1.Biofilms developed on decaying leaves in vitro after three weeks of incubation under aerobic conditions: (A) Hedera helix (B) Lonicera henryi (C) Acer sp.

Figure 2 .
Figure 2. Alpha diversity of leaf prokaryotic biofilm communities over time and water samples.The boxes show median, 25, and 75 percentiles.The vertical lines show minimum values excluding outliners, and the dots show outliners.The colours represent the oxygen conditions and water.

Figure 3 .
Figure 3. (A) Venn diagram showing the shift of the bacterial communities in 72 biofilm samples over time, day 1, green; week 1, orange; week 2, violet; week 3, magenta.(B) Ordination by PCoA of bacterial community data based on Bray-Curtis dissimilarity; WS, lake water sampled on day 1.

Figure 4 .Figure 5 .
Figure 4. Heatmap of the 20 most abundant bacterial genera or ASVs in the leaf biofilms.The numeric values indicate the percent relative abundance in the sample which are colour-coded (orange indicates high relative abundance and blue indicates low relative abundance).Triplicate biofilm and water samples were merged.

Figure 6 .
Figure 6.(A) Venn diagrams of the fungal communities over time.(B) Ordination by PCoA of fungal biofilm data based on Bray-Curtis dissimilarity.

Figure 7 .Figure 8 .
Figure 7. Alpha diversity of leaf fungal biofilm communities from different plants over time.The boxes show median, 25, and 75 percentiles.The vertical lines show minimum values excluding outliners, and the dots show outliners.The colours represent the oxygen conditions and water.

Figure 9 .
Figure 9. Heatmap showing significantly different fungal ASVs abundant across time (week 0-week 3) based on DESeq2 results (p adj < 0.01).Samples have been grouped initially by time (week 0, week 3), then by plant genus (Acer, Hedera, Lonicera) and treatment (aerobic, anaerobic).ASVs are labelled according to phylum level taxonomy indicated on the left.

Figure 10 .
Figure 10.Plots showing the relative contribution (%) of various ecological assembly processes governing the turnover of bacterial and fungal leaf biofilm communities.