Ulva fasciata was randomly sampled from Nanao Island, Shantou, Guangdong, China (about 100 m from the coast) (116.53°E 23.11°N). In the seedling stage, U. fasciata were sampled once a week on November 9th, 16th, and 23th, that corresponding to seedling stages 1, 2, and 3, respectively. During the growth stage, U. fasciata were sampled on December 27, February 2, and March 28, which corresponding to growth stages 1, 2, and 3, respectively. At maturity, U. fasciata were collected on April 2, May 3, and June 3, corresponding to maturity stages 1, 2, and 3, respectively (Figure 1A). The samples were collected in three replicates, washed with sterile seawater, and stored in 50-mL sterilized centrifuge tubes.

During nine sampling, the in situ environmental parameters (temperature, salinity, pH, and dissolved oxygen concentration) were recorded (Supplementary Table 1). At the same time, 1 L seawater samples were collected, filter through 0.45 and 0.22 μm membranes to remove impurities and collect bacteria. The filtrates (0.22 μm filter membrane) were stored in sterile sampling bags and the filtered water was cryopreserved for the determination of physical and chemical parameters, including dissolved organic carbon (TOC), inorganic carbon (TIC), total nitrogen (TN), and total carbon (TC) according to the methods described by Urgun-Demirtas et al. (2008). All samples were placed in a foam box with ice cubes and transferred to the laboratory within 4 h. In the laboratory, the microbiota harboring the surface of U. fasciata was then immediately collected. The U. fasciata samples in a 50-mL sterilized centrifuge tube was mixed with an appropriate amount of sterile seawater and treated with ultrasonic (1 min, 50 W, 5s/5s). Algae were removed, and then the suspension was centrifuged at 4,000 × g for 10 min to collect the pallet, which was then used for bacterial DNA extraction.

Bacterial DNA was extracted using the CTAB method (Zhou et al., 1996). DNA quality and integrity were evaluated by gel electrophoresis. Samples with high-quality DNA were stored at −80°C until further use. The DNA of all samples was sent to Shanghai Majorbio Bio-Pharm Technology Company for sequencing and data analyses.

To determine the species and abundance of bacteria, the V3–V4 region of the 16S rRNA gene was amplified using primers 806R (5′-GGACTACHVGGGTWTCTAAT-3′) and 338F (5′-ACTCCTACGGGAGGCAGCAG-3′), and the PCR products were sequenced using Illumina MiSeq sequencing platform. The sequencing data were stitched by Flash V1.2.11, and then the optimized Fastq files were obtained after quality control by Fastp V0.19.6. The sequences were preprocess using the software Usearch 7 (parameter setting: minsize 2), and the sequences were grouped into operational classification units (OTUs) for species level classification using 97% sequence similarity. After obtaining the OTU classification table, those OTUs assigned to chloroplasts or mitochondria and eukaryotes were excluded from the data set. OTU counts were normalized, the original count was divided by the total number of each sample, and then multiplied by the median to calculate the relative abundance. Using QIIME V1.9.1, the taxonomic information was assigned to each representative OTU sequence and filtered OTUs were used to calculate alpha diversity indexes, including Shannon, Chao1, and Simpson. Calculate the Bray-Curtis distance for all pairs of samples to determine the amount of diversity shared between communities (β-diversity) (Liu et al., 2021). Principal Co-ordinates Analysis (PCoA) sorting was performed according to Bray-Curtis distance. Adonis based on Bray distance was used to calculated the significant difference between bacterial community and the null hypothesis (Anderson, 2001). Dispersion of microbial communities was evaluated using the pairwise permutation tests with Tukey’s HSD. The Kruskal–Wallis test was used to determine whether there were significant differences in the abundance of bacterial classes. Bacterial taxonomic biomarkers for bacterial classification were obtained by referring to the random forest algorithm (Zhang et al., 2018). The Cytoscape software package was used for visualization. All of the above software used default parameters.

Samples of U. fasciata were collected at different times, rinsed and immediately transferred to the laboratory. The blade surface was gently scraped with a sterile knife, and the scraped materials were gradually diluted and spread on the marine agar plate 2216E, and placed in an incubator at 25°C for 3–4 days. Single colonies were selected, streaked, purified, and inoculated in marine broth 2216E. The 16S rRNA gene of bacteria was amplified with universal primers 27F and 1492R, and the samples were sent to Bioengineering (Shanghai) for sequencing. The phylogenetic names of the isolates were identified by EzBioCloud database. A total of 228 pure strains were obtained. These strains were cultured on marine broth 2216E to logarithmic growth phase. The bacterial suspension was mixed with 40% sterilized glycerol in the ratio of 1:1. These strains were selected as candidate strains for bacterial treatment. To test the promoting effect of bacterial treatment on plant growth, U. fasciata without surface bacteria (WSB) was cultured with isolates.

Ulva fasciata without surface bacteria was prepared as follow:

The surface of U. fasciata was washed with sterile seawater and ultrasonic waves (1 min, 50 KHz, 5s/5s) for three times to remove bacteria. The mixture of antibiotics (including 100 μg/mL streptomycin sulfate, 0.25 μg/mL amphotericin, 100 μg/mL ampicillin, and 0.1 μg/mL imidazole) was added to the culture tube and incubated for 2 days; the same dose of antibiotics was added on the third day. The surface of U. fasciata was quickly scrubbed with a sterile cotton swab dipped in a multi-enzyme cleaning solution (3M), and then rinsed with sterile seawater. The obtained SBR was evaluated through three methods: (1) PCR detection of seawater collected from the last rinsing of U. fasciata and algal abrasive solution in the range of about 1500 bp using primers 27F and 1492R, (2) the seawater collected from the last time rinsing of U. fasciata was used to culture on marine agar 2216E to confirm the growth of bacteria in 7 days, (3) the sterile surface of U. fasciata was stained with 1% (v/v) of DAPI for 20 min, the excess DAPI was removed with sterile water, and then the confocal laser scanning microscopy (CLSM) was used to check whether there were bright blue bacterial spots (bacteria) in the sample. Samples were imaged with CLSM (LSM 700, Zeiss, Germany) and x100 oil immersion objective. Select DAPI channel 400–630 nm. Four random locations were scanned on each sample.

The experiment consisted of three bacterial treatment groups, sterile seawater (C group), Bacillus cereus U5-30 (P group) and Hyunsoonleella sp. HU1-3 (H group). The bacteria were incubated in marine broth 2216E at 25°C for 2 days, centrifuged, washed twice with sterile seawater, and then resuspended to a density of 10⁷ cells per mL. The volume of bacterial liquid added to each bacterial treatment was 5 mL. The mixed bacteria were added to a bottle connected to filtered air and contained 800 mL of sterile seawater, and then co-cultured with U. fasciata (WSB) for 7 days. At the end of the experiment, the co-cultured U. fasciata was washed with sterile seawater, and the surface bacterial strains of U. fasciata were obtained by ultrasonic shock. The 2216E marine agar plate was used to test whether the bacterial treatment strains survived after 7 days.

To determine the growth rate, at the beginning of the co-culture experiment, young U. fasciata with similar growth trend and size were selected, weighed (set up as W_t) and sterilized. The vase was placed in a light incubator for 7 days (temperature, light intensity, photoperiod and air supply time were 20°C, 4000 LX, 12 h:12 h, and 24 h, respectively). After 7 days of co-culture, U. fasciata was taken out, the surface seawater was removed and weighed (W₀). Growth rate was calculated by the formula of [(W_t/W₀)^1/t-1] × 100% (t is days of culture) (Yong et al., 2013). During the experiment, each bacterial treatment group was guaranteed at least three biological replicates. The contents of soluble protein, soluble sugar, phycocyanin, and chlorophyll-a (Schwenzfeier et al., 2011) were determined to characterize the biomass of U. fasciata.

The bacteria strains were incubated in 2216E liquid medium at 25°C for 3 days, with or without 100 μg/mL of L-tryptophan, and then centrifuged. After that, the supernatant was diluted to OD600 = 0.8 with sterile seawater, and 10 mL of culture were added to a bottle containing 800 mL of sterile seawater with U. fasciata (WSB) and inoculated for 7 days. In the experiments, six supernatant were used, including sterile seawater (A group), 2216E liquid medium (B group), supernatant of B. cereus strain U5-30 living in 2216E liquid medium (C group), supernatant of Hyunsoonleella sp. strain HU1-3 living in 2216E liquid medium (D group), supernatant of B. cereus strain U5-30 living in 2216E liquid medium and L-tryptophan (E group), supernatant of Hyunsoonleella sp. strain HU1-3 living in 2216E and L-tryptophan (F group). Six groups of supernatants were separately cocultivated with U. fasciata (WSB), and the growth rate of U. fasciata was measured by the same method as above.

In order to illustrate how Hyunsoonleella strain HU1-3 promotes the growth of U. fasciata, the extraction of its genome was sent to BGI Corporation for genome test, and double-terminal sequencing was performed based on Illumina PE150 platform. Raw data is processed by low-quality filtering. Low quality bases of more than 40 bp were removed (quality value 38 or less), N bases were up to 10 bp, and the overlap between junctions was more than 15 bp. The repeated contamination is then removed to obtain valid data. The genome size was estimated by K-mer statistical analysis, and the raw sequencing reads were trimmed and assembled by SOAPdenovo v2.04 software¹. The FastQC software was used to analyze sequence quality (Brown et al., 2017). GeneMark software (V.4.17) (Besemer et al., 2001) was used for comparison and genomic protein-coding genes were obtained. These gene sequences were compared with the known Nr database (non-redundant protein database) (Li et al., 2002). For the comparison results of each sequence, the highest score (identity ≥40%, coverage ≥40%) was selected for gene function annotation.

The bacteria strains were incubated in 2216E liquid medium at 25°C for 3 days. 1-mL incubation fluid was centrifuged, the cell precipitates were washed twice with PBS, and then resuspended to 107 cells/mL. 1 mL of the suspension was mixed with 10 mL of liquid medium (with and without 100 μg/mL of tryptophan). After 3 days, 50 μL of supernatant was collected and mixed with 50 μL of Salkowski reagent (50 mL 35% HClO₄ + 1 mL 0.5 mol/L FeCl₃). After 25 min, the absorbance value was read at 530 nm (Bric et al., 1991) to measure the production of bacterial IAA. Bacteria were inoculated in 50 mL sterilized liquid NBRIP medium at a dose of 0.2% and cultured in a flask at 25°C for 7 days. The culture was centrifuged at 5,000 RPM for 20 min, and the precipitation of P₂O₅ was determined by molybdenum blue method according to Galhardo and Masini (2000) to measure the solubility of phosphate. All measurements were conducted in triplicate.

In order to investigate the variations of microbes harboring the surface of U. fasciata under natural conditions, seawater samples and U. fasciata samples, a total of 2,969,125 high-quality sequences were obtained from all samples. The bacterial community structure over time in different samples is illustrated in Figures 1B,C. There was a significant difference (P < 0.01) between the microbial community living in seawater and the U. fasciata surface community (Figure 1B). PCoA analysis also showed that there were different microbiota harboring the surface of U. fasciata at different growth stages (R² = 0.68, P < 0.01, adonis) (Figure 1C). This indicated that the growth stage plays a significant role in the establishment of microbiota inhabiting the surface of U. fasciata. Mature stage samples were close to one another, while seedling stage samples were relatively far away from one another (Supplementary Figure 1). The microbial community in mature stage overlapped with that in seedling stage and growth stage, indicating that the microbial community is in a transitional state between mature and seedling/growth stages. Similar results can be obtained by calculating the community diversity index (Shannon) and community richness index (Chao and ace) (Table 1), in which the alpha-diversity index of surface bacteria was significantly lower than that of seawater bacteria, while the diversity index of U. fasciata surface bacteria remained relatively stable.

The relative abundance of epiphytic microbiota of U. fasciata and seawater is illustrated in Supplementary Figures 2A,B. Seawater samples were dominated by marine bacteria, such as HIMB11 and NS5 marine groups (Supplementary Figure 2A). Microbial community abundance were significant differences in the surface of U. fasciata with the seawater (Supplementary Figure 2B). The relative abundance of Proteobacteria and Bacteroidetes on U. fasciata surface was significantly higher that in seawater. The results of Kruskal-Wallis analysis (Supplementary Figure 3A) showed that although the relative abundance of these species increased or decreased throughout the growth period, there was no significant difference. Actinobacteria and Firmicutes changed greatly throughout the growth stage. At the genus level, there were significant differences in the relative abundance of some species (Supplementary Figure 2C). For instance, at the seedling stage, the relative abundances of Rhodobacteraceae (13.0%, P < 0.01, Kruskal–Wallis), Acinetobacter (6.2%, P < 0.01, Kruskal–Wallis), and Vibrio (1.2%, P < 0.01, Kruskal–Wallis) was significantly lower, while the relative abundance of Ralstonia (2.3%, P < 0.01, Kruskal–Wallis), Erythrobacter (2.8%, P < 0.01, Kruskal–Wallis), and Rhodococcus (2.4%, P < 0.01, Kruskal–Wallis) were significantly higher. The composition of microbiota on the surface of U. fasciata varied greatly throughout the growth, but the dominant species of microbiota were Rhodobacteraceae, Ralstonia, Erythrobacter, Rhodococcus, Acinetobacter, Alteromonas, Vibrio, Halomonas, and Granulosicoccus.

The regression analysis between the relative abundance of bacteria on the surface of U. fasciata species and the growth stage was established and revealed that the model explained 56% of the surface of U. fasciata microbiota variance related to U. fasciata growth stage (Supplementary Figure 4). At the class level, the class number (n = 30) was stable and relative to the cross-validation error curve. Therefore, these 30 bacterial classes, especially Rhodococcus and Hyunsoonleella, were defined as biomarker groups in the model.

The relationship between environmental factors and the abundance of different phyla is shown in Figure 2. At the phylum level, the relative abundance of Patescibacteria, Firmicutes and Bacteroidetes was positively correlated with the physiological parameters of different carbon forms. As mentioned above, the U. fasciata growth stage is the best explanation for the variation of microbial community. However, the environmental and physiological parameters of U. fasciata at different growth stages have different explanations for the variation of microbial community (Table 2). Dissolved oxygen (DO) and Phycocyanin had no effect on the microbial community on the surface of U. fasciata (R² = 0.04 and 0.05, respectively, P > 0.05, Adonis). TOC, pH, soluble protein and TN had moderate effects on the microbial community on the surface of U. fasciata. IC and TC had the greatest effect on the microbial community on the surface of U. fasciata (R² = 0.13, 0.14, respectively, P < 0.05, Adonis). The results showed that environmental factors and physiological parameters of U. fasciata also affected the structure of surface bacteria.

A total of 228 bacterial strains were isolated from the surface of U. fasciata at different growth stages (Supplementary Figure 5). These strains were classified into three phyla, including Firmicutes, Bacteroidetes, and Proteobacteria. Although bacteria were still present in the U. fasciata (WSB) lapped samples (Supplementary Figure 6C) and intercellular bacteria were detected by CLSM (Supplementary Figure 6P), the quality of U. fasciata (WSB) was comparable to that of wild U. fasciata (Supplementary Figure 7) (P > 0.05, ANOVA).

The prediction results of machine forest model suggest that members of Rhodobacteraceae and Flavobacteriaceae may play an important role in the growth of U. fasciata. Therefore, bacterial treatment experiments were carried out to investigate the growth-promoting effect of Rhodobacteraceae and Flavobacteriaceae on U. fasciata. The role of endophytic bacteria was ignored in the experiment bacterial treatments. Among the isolates, Hyunsoonleella sp. HU1-3 demonstrated the greatest growth promoting potential on U. fasciata. Compared with other bacteria of Hyunsoonleella, its 16S rRNA gene sequence similarity and average nucleotide identity values were less than 98.7 and 95%, respectively (Chun et al., 2018), so it is considered as a new species of the genus.

The four physiological indices and growth rate of U. fasciata of group H [co-culture of U. fasciata (WSB) and Hyunsoonleella sp. HU1-3] were significantly higher than those of group C (only seawater was added as the negative control) and group P [co-culture of U. fasciata (WSB) and B. cereus U5-30 as the positive control] (P < 0.05, ANOVA) (Figure 3). In fact, compared with group P, the positive effect of Hyunsoonleella HU1-3 on soluble protein, soluble sugar and phycocyanin of U. fasciata showed that the strain significantly increased the biomass of sterile U. fasciata (P < 0.05, ANOVA) (Figures 3A–D). The growth rate of group H (5.49% day^–1) and group P (4.05% day^–1) was significantly higher than that of group C (1.10% day^–1) (Figure 3E).

Six different groups of bacterial supernatants were cocultured with U. fasciata (WSB) (Figure 4). There was no significant difference in the growth rate of U. fasciata between group A (sterile seawater) and group B (sterile 2216E liquid medium) (P > 0.05, ANOVA). Interestingly, the supernatant of group C (B. cereus strain U5-30) and group D (Hyunsoonleella sp. HU1-3) significantly increased the growth rate of U. fasciata (1.52% day^–1 and 2.45% day^–1, respectively; P < 0.05, ANOVA). Surprisingly, the U. fasciata growth rate of group E and group F (2.77% day^–1 and 3.95% day^–1, respectively; P < 0.05, ANOVA) was significantly higher than that of group C.

Based on the genome of Hyunsoonleella HU1-3 (Figure 5 and Supplementary Table 2), several amidases (such as N-acetylmuramoyl-L-alanine amidase LytC and 2-oxoglutaramate amidase) and tryptophan related enzymes (such as tryptophan 2,3-dioxygenase and tryptophan-tRNA ligase) were identified, indicating that Hyunsoonleella sp. HU1-3 may produce IAA through IpyA and IAM pathways. Genes involved in the synthesis of cytokinins (CK) have also been annotated, such as tRNA-dimethylallyltransferase, tRNA-2-methylthio-N6-dimethylallyladenosine, and cytokinin-nucleoside 5-monophosphate ribose hydrolase. These results indicate that Hyunsoonleella sp. HU1-3 has the potential to synthesize a variety of CKs. Furthermore, seven phosphate synthases genes were found in the genome of Hyunsoonleella sp. HU1-3. Interestingly, Hyunsoonleella sp. HU1-3 contains genes involved in vitamin synthesis (thiamine-phosphate synthase and thiamine-monophosphate kinase). Other genes annotated as iron carrier enzymes (such as iron-sulfur cluster carrier protein and iron-dependent repressor IdeR) were also found in the genome.

Plant growth is known to depends on several transaminase proteins, especially growth hormones, such as IAA. In the absence of tryptophan, the IAA yield of B. cereus U5-30 was 209.70 ± 3.64 μg/mL. The addition of tryptophan increased the IAA production to 375.16 ± 3.59 μg/mL. The phosphate solubility of B. cereus U5-30 was 191.18 ± 2.87 μg/mL. On the other hand, the IAA yield and phosphate solubility of Hyunsoonleella sp. HU1-3 without tryptophan were 227.64 ± 0.82 μg/mL, and 213.98 ± 1.15 μg/mL, respectively. It is speculated that the IAA yield and phosphate solubility of Hyunsoonleella sp. HU1-3 are better than B. cereus U5-30 (Table 3).