Metagenomic analysis of the complex microbial consortium associated with cultures of the oil‐rich alga Botryococcus braunii

Abstract Microalgae are widely viewed as a promising and sustainable source of renewable chemicals and biofuels. Botryococcus braunii synthesizes and secretes significant amounts of long‐chain (C30‐C40) hydrocarbons that can be subsequently converted into gasoline, diesel, and aviation fuel. B. braunii cultures are not axenic and the effects of co‐cultured microorganisms on B. braunii growth and hydrocarbon yield are important, but sometimes contradictory. To understand the composition of the B. braunii microbial consortium, we used high throughput Illumina sequencing of metagenomic DNA to profile the microbiota within a well established, stable B. braunii culture and characterized the demographic changes in the microcosm following modification to the culture conditions. DNA sequences attributed to B. braunii were present in equal quantities in all treatments, whereas sequences assigned to the associated microbial community were dramatically altered. Bacterial species least affected by treatments, and more robustly associated with the algal cells, included members of Rhizobiales, comprising Bradyrhizobium and Methylobacterium, and representatives of Dyadobacter, Achromobacter and Asticcacaulis. The presence of bacterial species identified by metagenomics was confirmed by additional 16S rDNA analysis of bacterial isolates. Our study demonstrates the advantages of high throughput sequencing and robust metagenomic analyses to define microcosms and further our understanding of microbial ecology.

may be cultured in open ponds or raceways, which are the most costeffective methods of culture for low value products such as biomass or biofuel (Richardson, Johnson, & Outlaw, 2012). However, highly selective conditions are required which favor hydrocarbon synthesis and maintain the dominance of the algal crop over other species.
While the association with microorganisms is not essential for hydrocarbon production in B. braunii, the addition of single bacterial species to previously axenic B. braunii culture increased algal biomass and hydrocarbon yields (Chirac et al., 1985;Wang & Xie, 1996). However, these effects were variable and dependent not only on the bacterium present in the algal culture, but also on the algal strain and culture conditions (Chirac et al., 1985;Jones, 1972). Furthermore, the nature of the bacterial association with B. braunii was different when the algae were grown as a planktonic culture or in a biofilm (Rivas et al., 2010).
It is clear, therefore, that microorganisms in the B. braunii consortium can have unpredictable and sometimes contrasting effects on algal biomass and hydrocarbon yield, which are essential considerations for sustainable biofuel production.
To investigate further the nature of the B. braunii microbial consortium, we profiled the microbiota of an established, laboratory culture of B. braunii under different conditions, using high throughput Illumina sequencing of metagenomic samples.

| Modification of Botryococcus braunii culture conditions
Understanding the structure of algal-bacterial interactions using metagenomics offers the potential to ensure more robust cultures and exploit, for industrial or nutritional use, a wider variety of species than currently deployed. To investigate the nature of the B. braunii microbial consortium, and to discover which taxonomic units (species/genera) are present in close association with the alga, we used a combination of metagenomic analysis and identification of cultured isolates to investigate the bacterial species present in a laboratory culture of B. braunii strain Guadeloupe (Race B) (Metzger, David, & Casadevall, 1986).
To drastically perturb the microbiota present in the B. braunii culture and identify bacteria that may be more closely associated with B. braunii, culture conditions were altered from those routinely used (Condition A) by either repeated centrifugation and rinsing (Condition B) or by addition of the antibiotic, ciprofloxacin (Condition C) to a laboratory culture of B. braunii (Figure 1).
Ciprofloxacin is a broad-spectrum, fluoroquinolone antibiotic that inhibits DNA gyrase or topoisomerase IV activity (Maxwell, 1997) resulting in incomplete DNA synthesis during DNA replication and inhibition of bacterial growth. Ciprofloxacin was selected from a range of 10 antibiotics covering a range of activities and targets because it was the only antibiotic tested that had no significant effect on B. braunii growth and chlorophyll content ( Figure 1).
Nile red reagent is widely used to stain and quantify hydrocarbons produced by B. braunii ((Cooksey, Guckert, Williams, & Callis, 1987;Elsey, Jameson, Raleigh, & Cooney, 2007;Lee, Yoon, & Oh, 1998); in the microbiota between treatments therefore had no effect on the amount or composition of neutral lipids and hydrocarbons produced by the algae, which is consistent with previously reported data (Chirac et al., 1985). It is therefore likely that under all conditions tested the bacteria present were not having a negative impact on the growth or hydrocarbon metabolism of B. braunii.

| Metagenomics analysis of the algal consortia
The development of high throughput metagenomic sequencing has enabled a profound examination of the complexity and dynamics of microbial populations (Cooper & Smith, 2015;Jansson, Neufeld, Moran, & Gilbert, 2012;Tringe & Rubin, 2005;von Mering et al., 2007). Metagenomic analysis can deliver information on the identity of microbes present in cultures and also data regarding the relative representation of different microbes within cultures.
Genomic DNA was extracted from all of the organisms present in conditions A, B and C and sequenced on the high-throughput Illumina GA2 platform. After excluding low quality sequence reads, this analysis generated in excess of 11 million sequence reads per culture, namely 11,313,866 reads for condition A, 15,510,317 reads for condition B and 11,702,497 reads for condition C. The metagenomic content of the three conditions were analyzed using two bioinformatic approaches: 1. Similarity to nucleotide and protein databases using BLAST, and 2. Validation by sequence read alignment to genomes of inferred species identified using BLAST in step 1.

| Taxonomic assignment of metagenomic DNA using BLAST and MEGAN
To assess the range of bacteria at species level in all culture conditions, BLAST analysis was used to infer bacterial species present in the three experimental culture conditions. High quality sequence reads for Conditions A, B and C were analyzed by BLASTN using the NCBI nucleotide database (NT), and by BLASTX using the non-redundant protein database (NR).
Using BLASTN to identify similar sequences and MEGAN to assign aligned reads to taxa, enabled taxonomic binning of 2,024,504 reads. The number of predicted bacterial species in the metagenome is dependent on the number of sequence reads that are binned to a given species.
Sequences of viral origin were found exclusively in the initial (Condition A) stock culture ( Figure 3). These reads were almost entirely (99.8%) predicted to be from bacteriophages of the order Caudovirales.
No sequences were binned to the Archaea. Sequences of eukaryotic origin aligned to Opisthokonta (Fungi and Metazoa), red algal (Rhodophyta) and plant (Viridiplantae) lineages ( Figure 3). Reads binned to Viridiplantae (Figure 3) may be attributed to conserved B. braunii sequences and were predominantly assigned to Vitis vinifera, Zea mays and Populus trichocarpa, which could be ascribed to the abundance of nucleotide data present for these organisms in the BLAST databases.
Bacterial assignments consisted of 83% of the binned metagenome (1,670,779 reads), which correlates to 4.34% of the entire metagenome.
The combined stringency analysis of BLASTX and BLASTN MEGAN assignment has identified species from 29 possible genera.
As part of this investigation the high stringency and standard stringency analyses shared 19 identified species that may be present or similar to species which are present in the B. braunii consortium:

| Further bacterial species resolution
High stringency MEGAN analysis was validated by alignments to previously sequenced genomes retrieved from NCBI of the identified species and the number and genome distribution of mapped reads determined. The number of reads from the entire metagenome that map to each of the 33 individual bacterial genomes identified by high F I G U R E 1 Response of the Botryococcus braunii (Guadeloupe) consortium to antibiotics. (a) DAPI staining of live cells of B. braunii (Guadeloupe) consortium. Scale bar = 20 μm. (b) B. braunii (Guadeloupe) consortium grown in MCV media supplemented with or without antibiotics. Algae growth was calculated by measuring chlorophyll extracted from culture; the growth rate was calculated from the change in chlorophyll concentration over 6 days (mg l −1 day −1 ) and divided by the average growth rate of the control (without antibiotics) to give the relative growth rate, so that control = 1; bars show the mean and SEM, n = 3. (c) Growth curves of B. braunii (Guadeloupe) consortium grown in MCV: initial consortium (Condition A: circle), washed culture (Condition B: square) and ciprofloxacin-treated (Condition C: triangle). Biomass was calculated after drying algae at 60°C for 4-5 days. Data has baseline correction for different inocula and points represent the mean of three replicates stringency analysis were determined using Bowtie (Langmead, 2010;Langmead, Trapnell, Pop, & Salzberg, 2009).
To investigate the distribution and coverage of aligned reads across the genome, MOSAIK analysis revealed that none of the reference genomes was fully covered by Illumina reads (Table S4). The highest levels of coverage were for Lactococcus lactis (69.8%, condition A) and Lactobacillus sakei (48.1%, condition A). The reads were distributed evenly throughout the genomes rather than clustered together and resulted in a mean depth of coverage of 4.2× for L. lactis and 1.7× for L. sakei. This analysis confirms the presence of these or very closely related species in the microbiome.
The evidence from the combined high stringency analysis performed in this investigation indicated that the initial B. braunii culture (Condition A) contained a population of species from the orders Lactobacillales, Enterobacteriales, Pseudomonadales, Flavobacteriales, Clostridiales and one species in the Burkholderiales order, Ralstonia F I G U R E 2 Effect of culture treatments on hydrocarbon production in B. braunii. (a) Lipid bodies and oil-rich colony matrix (red) are distinguished from chloroplasts (green) in live cells of B. braunii using Nile red reagent. Images were captured using Zeiss LSM 510 META microscope (Carl Zeiss, Oberkochen, Germany) using a Plan-Apochromat 63x/1.40 oil DIC M27 lens. Cells were irradiated with an excitation wavelength of 458 nm; emission of Nile red reagent was filtered between 550 and 571 nm whereas emission from chlorophyll was filtered between 668 and 721 nm. Individual cells are embedded within the oil-rich matrix (M); lipid bodies are located within the cells (arrows). Scale bar = 20 μm. pickettii (Table S1, Figure 4). These microorganisms were removed by washing (Condition B) and failed to re-colonize the culture when ciprofloxacin was present (Condition C

| Validation of bioinformatics approaches by 16S rDNA sequencing of bacterial isolates
Following repeated streak-culture on solid LB or MCV media, 10 bacterial strains with different colony characteristics (morphology, growth rate, medium preference) were isolated from the B. braunii consortium grown under Condition A. The genera of seven distinct bacterial strains were identified by 16S rDNA homology (Marchesi et al., 1998) as Achromobacter, Asticcacaulis, Flavobacterium, Agrobacterium, Microbacterium, Shinella and Variovorax (Table 1). Although similarity below 98.7%-99.0% of the 16S rDNA gene sequences of two bacterial strains has been reported as confirming they belong to different species (Stackebrandt & Ebers, 2006), the recent increase in metagenomic and bacterial sequencing has highlighted that 16S rDNA analysis can only refine identification to the family/genus level (Tu, He, & Zhou, 2014). All of these culturable genera, barring the Microbacterium and

All bacterial isolates grew in LB medium. Shinella, Agrobacterium
and Variovorax isolates also grew in fresh algal MCV medium at 25°C.
Achromobacter, Asticcacaulis, Microbacterium and one unidentified colony (GG1) were unable to grow in fresh MCV but grew in "spent" MCV medium that was prepared from a filter-sterilized culture of actively growing B. braunii. Moreover, Shinella showed improved growth in spent MCV compared to fresh MCV. The Flavobacterium failed to grow on fresh and spent MCV, suggesting that this strain requires another factor provided by the consortium to sustain growth.
Some bacterial species were unable to grow in MCV medium but were able to grow in filtered medium after B. braunii culture (unidentified species GG1, Achromobacter sp., Asticcacaulis sp. and Microbacterium sp.), suggesting they rely on factors produced by the algae or the other bacteria present in the algal consortium for growth. Further work in this area is required to determine which species in the consortium are beneficial and which are detrimental to B. braunii. Rhizobium species have been implicated as a probiotic in cultures of B. braunii (Rivas et al., 2010). In an attempt to understand how some of these interactions affect growth and hydrocarbon production, Chirac et al. (1985) investigated single species of bacteria combined with axenic cultures of B. braunii. Generally, biomass and hydrocarbon yield were increased compared to axenic cultures when CO 2 was limiting, suggesting bacterially produced CO 2 was utilized by the algae, but when CO 2 was abundant both biomass and hydrocarbon yield were reduced. The precise F I G U R E 3 Distribution of taxa from the metagenome of B. braunii. BLAST was used to recognize similar sequences to each read in the NCBI nucleotide database and non-redundant protein database. Taxa were assigned using MEGAN software at both standard and high stringency. The size of the circle is scaled logarithmically to the number of reads supporting the taxon; the proportion of the wedge indicates the numbers of reads from each culture condition: A (red), B (blue) and C (green) F I G U R E 4 Taxonomic analysis of the B. braunii metagenome dataset.
Step 1: Taxa were assigned using MEGAN restricted to bacteria with a heat map displaying abundance of reads assigned using BLASTN and BLASTX. Black dots represent likely presence based on high stringency cut-offs; Step 2: Presence of taxa were identified using guided assembly of Illumina reads on to candidate genomes using Bowtie and Mosaik; mismatches were not allowed. Black dots represent likely presence based on cut-offs: MOSAIK positive ≥ 8% coverage, BOWTIE positive ≥1500 reads. Stars represent species that were identified using both standard and high stringency analysis  (Chirac et al., 1985). The reasons for these potentially antagonistic or converse effects are complex and remain to be investigated fully. The presence of Corynebacterium sp. raised the level of hydrocarbon of the algal biomass (Banerjee et al., 2002) and Bacillus sp. increased hydrocarbon yield (Wang & Xie, 1996), however the precise reasons for the increases remain unknown. We have not observed any changes in hydrocarbon production when the consortia were perturbed, but as the growth rate was not significantly altered and the CO 2 was not limiting, this is not necessarily unexpected.
Finally, to ascertain whether any of the isolated bacteria were resistant to ciprofloxacin, isolates were streaked on to LB-agar containing 10 μg ml −1 ciprofloxacin. Achromobacter, Flavobacterium and The culturable bacteria species likely to be from the genera Variovorax, Achromobacter and Flavobacterium were resistant to ciprofloxacin (10 μg ml −1 ) whereas the Agrobacterium sp., Shinella sp., Asticcacaulis sp. and Microbacterium sp. were sensitive to ciprofloxacin when tested in isolation. An unexpected observation in this study was the sensitivity of the Asticcacaulis sp. to ciprofloxacin when grown on LB plates, but growth was evident in the consortia to a higher proportion than was present in the initial culture. This suggests that some additional factor was important for the increased presence of Asticcacaulis sp. in the consortia despite the ciprofloxacin, although it is possible that a very low level of ciprofloxacin-resistant bacteria grew rapidly once the sensitive bacteria were inhibited or killed. An alternative possibility is that the Asticcacaulis sp. is protected in some way by the algae or perhaps these bacteria are attached to the Botryococcus extracellular matrix and the oil affords a barrier to the ciprofloxacin ( Figure 5). Asticcacaulis sp. have been recorded in consortia with another Trebouxiophyceae alga, Chlorella sorokiniana (Watanabe et al., 2008) despite being a genus that is infrequently observed or isolated (Garrity et al., 2006). Another rarely observed species identified in this study was Stenotrophomonas maltophilia. It has been hypothesized to be a mutualistic bacterium when in culture with the dinoflagellate, Scrippsiella trochoidea (Tan et al., 2015).

| CONCLUSIONS
The presence of co-cultured bacteria is widely observed in laboratory cultures of microalgae and this may be associated with symbiotic associations that originate from the environment and persist in the laboratory. Studies on symbiotic associations of C. sorokiniana have revealed that microorganisms adhere to the surface of the algae directly or bind to the carbohydrate sheath produced by Chlorella providing the close proximity required for symbiotic association (Watanabe et al., 2005 Microbacterium sp. failed to grow in algal growth medium (MCV) but was able to grow in media after removal of the remaining B. braunii consortia by filtration, indicating that some factor produced by the consortium is required for maintenance of Microbacterium sp.
Other bacteria identified in this study have been previously observed in consortia with other green algae. Dyadobacter fermentans (Bacteroidetes) was present in all conditions in this study; D. fermentans has been identified in cultures of C. sorokiniana (Otsuka, Abe, F I G U R E 5 SEM images of Botryococcus braunii algae and associated bacteria. Botryococcus braunii consortia were imaged using cryogenic scanning electron microscopy. B. braunii cultures were washed with hexane to remove the hydrocarbons. Colonies were flash-frozen in liquid N 2 slush, transferred to a vacuum and coated in gold using the Gatan Alto 2100 system. Images were acquired using a JEOL JSM-6390 LV scanning electron microscope at 5 kV with a working distance of 10-12 nm Fukui, Nishiyama, & Senoo, 2008) and a Dyadobacter sp. has also been observed in culture with C. vulgaris (Lakaniemi, Hulatt, Wakeman, Thomas, & Puhakka, 2012). D. fermentans was originally isolated from the stems of Zea mays (Chelius & Triplett, 2000) and grows in nitrogenlimited media with no currently known beneficial or detrimental effect on the host plant.
Investigation of a Chlorella consortium indicated that different bacteria utilized carbon and nitrogen compounds secreted by the algae in characteristic ways. This observation illustrates the complexity of interactions between the alga, the associated microcosm and the environment. When numerous species are present, the complexity of interactions is multiplied accordingly and the balances within the system are more likely to be perturbed. In order to culture B. braunii cost effectively at scale, understanding the composition of the community is essential.
Our results indicate that sequences attributed to the co-cultured bacterial community were dramatically altered by the perturbations to the B. braunii culture. We reason that bacterial species least affected by the washing treatment (Condition B) were therefore more robustly associated with algal cells. These keystone species included members of Rhizobiales, comprising potentially symbiotic Bradyrhizobium and Methylobacterium, and representatives of the genera Dyadobacter, Achromobacter and Asticcacaulis.
These sequence data have been submitted to the NCBI SRA database under accession number SRP072490. Additional information on the methods used are available in Appendix S1.