Identification of active isoprene-degrading bacteria using DNA-SIP
Diversity of bacteria from soils in the vicinity of oil palm trees
Analysis of 16S rRNA gene amplicon sequences showed that the unenriched bacterial community from soils in the vicinity of oil palm trees was very similar across replicates, confirming that extraction and handling procedures were consistent (Fig S1 and Fig S2 show the relative abundance (RA) of 16S rRNA genes in these environmental samples). The unenriched soil microbial community (S T0) was mainly composed of Proteobacteria (40.8 ± 0.5% RA), Actinobacteria (13.1 ± 0.7%), Bacteroidetes (11.2 ± 1.4%) and Acidobacteria (10.8 ± 0.6%, Fig S1), all of which are dominant phyla in soils [43,44,45,46]. The most abundant genera were Rhodoplanes (5.9 ± 0.1%) and Flavobacterium (4.0 ± 0.9%; Fig S2).
Soils were then enriched with 13C-isoprene to identify the active isoprene degraders in this environment through DNA-SIP (see the “Methods” section). Sequencing of the 16S rRNA genes of the 13C-heavy fractions showed that, although there was considerable inter-sample variability, Rhodoblastus (10.2–33.7%) and Pelomonas (14.2–54.9%) were highly enriched in all soil replicates (S 13C H; Fig. 1). Novosphingobium was one of the major genera labelled in the 13C-heavy fractions of replicates 2 (S 13C H R2) and 3 (S 13C H R3) representing 47.8% and 24.5%, respectively. Finally, Sphingomonas dominated the isoprene-degrading community of replicate 3 (S 13C H R3) with a RA of 42.4% (Fig. 1). These four genera had 19- to 90-fold higher RA in 13C-heavy (S 13C H) than in the 13C-light (S 13C L) soil fractions and constituted 28.7–40.2% of the total microbial community of the unfractionated soils incubated with 13C-isoprene (13C UF; Fig S2), which strongly suggest that they are active isoprene degraders. As expected, Novosphingobium, Pelomonas, Sphingomonas and Rhodoblastus also dominated the 12C-isoprene-incubated microbial community (S 12C L), and each genus had a very similar RA to those of the unfractionated 13C-samples (S 13C UF; Fig S2).
Previous DNA-SIP experiments and cultivation-dependent studies have identified members of Sphingomonadaceae (Sphingopyxis) and Comamonadaceae (Ramlibacter and Variovorax) as isoprene degraders with a functional IsoMO [32, 36, 38]. However, this is the first evidence that other genera of these families such as Sphingomonas, Novosphingobium and Pelomonas are likely to be able to metabolise isoprene. In addition, Rhodoblastus is the first member of the Beijerinckiaceae family to be implicated in isoprene degradation. Therefore, it will be interesting to attempt to isolate representative strains of this genus in future studies in order to confirm this ability.
DNA-SIP and 16S rRNA gene amplicon sequencing showed that the oil palm soil harbours a distinct isoprene-degrading bacterial community from soils beneath temperate trees that emit high levels of isoprene, such as willow. DNA-SIP experiments using willow soil incubated with 13C-labelled isoprene identified Rhodococcus, Ramlibacter and Variovorax as the major genera labelled in the 13C-heavy fractions [32]. However, these genera represented < 1% of the 13C-heavy (S 13C H) fractions from oil palm soil. Also, other well-characterised isoprene-degrading microorganisms, such as Gordonia, Nocardioides, Mycobacterium and Sphingopyxis species [32, 47, 48], constituted only a small part of the isoprene-degrading community (< 1%) from soils taken from the vicinity of oil palm trees.
Both unenriched samples and heavy DNA fractions from these soil incubations were also subjected to metagenomic sequencing. Community composition of raw reads was assessed with MetaPhlAn2 [49]. As MetaPhlAn2 uses a range of clade-specific marker genes to assess the phylogeny of the metagenomics reads, results differed slightly from those obtained using 16S rRNA gene amplicon sequencing analysis. According to the phylogenetic analysis of the soil metagenomes, the unenriched soil (S T0) community was dominated by Proteobacteria (81.5%), Actinobacteria (7.8%) and Acidobacteria (6.3%; Fig S1), confirming the results obtained by the analysis of the 16S rRNA gene amplicon sequencing data. The most abundant bacteria in the unenriched soils that could be classified at the genus level belonged to Cupriavidus (14.2%), followed by Pseudogulbenkiania (13.2%) and Burkholderia (11.7%; Fig. 1).
Metagenomic sequencing revealed that 13C-heavy fractions from replicates 1 and 2 (S 13C H R1-2) were dominated by Thiomonas (34.5%) and Gordonia (18.2%), whereas replicate 3 (S 13C H R3) had a higher abundance of Gordonia (47%) and Sphingobium (21.8%; Fig. 1). However, these genera represented < 1% of the 13C-heavy fractions in the 16S rRNA gene amplicon sequencing data. It is not surprising to find members of Gordonia dominating the isoprene-degrading community, since strains from this genus have been shown to contain a complete isoprene degradation gene cluster [48]. However, this study provides the first evidence that Thiomonas and Sphingobium species may be also able to catabolise isoprene.
Diversity of bacteria from the phyllosphere of oil palm trees
The bacterial community of unenriched oil palm leaf (L T0) samples was dominated by Proteobacteria (74.5 ± 0.3%, Fig S1), which is not surprising since Proteobacteria have been found to be the most abundant phylum in the phyllosphere of several plant species [50,51,52,53]. Firmicutes also constituted a major component of the unenriched bacterial community from oil palm leaves (22.1 ± 0.2%, Fig S1), as has been reported for some trees and agricultural plants [51, 52, 54, 55]. The most abundant genera in the oil palm phyllosphere were Acinetobacter (26.4 ± 0.7%), followed by Clostridium (22.0 ± 0.2%) and Enterobacter (11.6 ± 0.2%; Fig 2).
The 16S rRNA gene amplicon sequencing data showed that the diversity of the isoprene-degrading community of the samples incubated with 13C-isoprene (L 13C H) was highly consistent between replicates, with Gordonia (51.4 ± 9.4%) and Zoogloea (12.3 ± 2.2%) being the most abundant genera (Fig. 2). The RA of Gordonia and Zoogloea was 84.9 and 58.2-fold higher in the 13C-heavy (L 13C H) compared to the 13C-light (L 13C L) fraction, respectively (Fig S3), indicating that they are active isoprene degraders. In addition, these two genera constituted 10.8% of the total microbial community of the 13C-unfractionated (L 13C UF) samples and, as expected, were also highly abundant in the 12C-isoprene-incubated (L 12C L) microbial community (13.6%; Fig S3).
Strains of Gordonia that grow on isoprene as sole carbon and energy source have been isolated previously from leaves of an oil palm tree in the Palm House of Kew Gardens, London [32]. However, although a number of SIP experiments with 13C-isoprene have been performed with samples from a wide range of environments, including the phyllosphere, estuaries and soils, no members of the Zoogloea genus or the order Rhodocyclales have been identified as active isoprene degraders [32, 36, 38, 48]. Here, the identification of Zoogloea as an isoprene degrader indicates that the variety of microorganisms able to metabolise this important climate-active gas is greater than previously known.
Rhizobium also had a relatively high RA (8.5 ± 2.2%) in all replicates from 13C-heavy fractions (L 13C H) compared to the unenriched (L T0) samples (Fig. 2). However, its RA was 2.2-fold higher in the 13C-heavy (L 13C H) than the 13C-light (L 13C L) fractions, which is the same ratio observed between the 12C-heavy (L 12C H) and the 12C-light (L 12C L) fractions (Fig S3). Therefore, based on these data, and since no strains of this genus have been isolated from this environment to corroborate its ability to degrade isoprene, Rhizobium spp. cannot be yet confirmed as isoprene degraders.
Previous SIP experiments exploring the phyllosphere of other high isoprene-emitting trees from temperate regions, such as poplar, identified Rhodococcus and Variovorax as the major players in isoprene degradation [38]. However, in our experiment, the phyllosphere from tropical oil palm trees yielded a distinct profile of active isoprene degraders, with Gordonia and Zoogloea being the main genera enriched in the 13C-heavy fractions and Rhodococcus and Variovorax showing a low RA (2.1 ± 0.4% and < 1%, respectively). The RA of other well-characterised isoprene degraders such as Sphingopyxis, Ramlibacter, Nocardioides or Mycobacterium [32, 48] also represented < 1% of the labelled bacterial community from oil palm leaves (L 13C H).
Phylogenetic analysis of the unenriched leaf (L T0) metagenomes confirmed that the unenriched bacterial community of the oil palm phyllosphere was overwhelmingly dominated by Proteobacteria (99.1% RA; Fig S1). At the genus level, metagenomics analysis also supported the 16S rRNA gene amplicon sequencing data, since Acinetobacter (40.8% RA) and Enterobacter (12.7% RA) were highly abundant in the unenriched phyllosphere community, together with Pantoea (15.5% RA; Fig. 2).
Metagenomic data showed that Gordonia constituted 93.7% of the isoprene-degrading community of oil palm leaves (L 13C H R1-3; Fig. 2), in accordance with the 16S rRNA gene amplicon sequencing results. However, no Zoogloea sequences were identified in the 13C-heavy fractions in the metagenomic analysis probably due to the different approach that MetaPhlAn2 uses to assign the phylogeny of the reads compared to the 16S rRNA gene amplicon analysis.
Comparison of isoprene degraders from soils and phyllosphere of oil palm trees
Studying the microbial diversity associated with plants is an essential step to understand host-microbiome interactions. However, only a few studies comparing microbial communities of phyllosphere and soils associated with the same plant species have been conducted to date (e.g. [54, 56, 57]). Here, we show that the unenriched bacterial communities from oil palm soils (S T0) and leaves (L T0) are distinct even at the phylum level (Fig S1), as reported for other plant species [54], although both soil and leaves are dominated by Proteobacteria. The 16S rRNA gene amplicon sequencing data also revealed that the active isoprene-degrading bacteria from soil samples were phylogenetically more diverse than those from the oil palm phyllosphere (see above).
When comparing the unenriched soil (S T0) and leaf (L T0) communities, it is interesting to note that although each major player in isoprene degradation was present in these contrasting environments at similar RA (Table S1), they responded differently to isoprene enrichment (S 13C H and L 13C H). This suggests that the physiochemical conditions and/or interactions with other groups of microorganisms shape the composition of the isoprene-degrading community in a particular environment.
In addition, unenriched soil and leaf oil palm metagenomes (S T0 and L T0) were analysed for the presence and relative abundance of isoA genes. Metagenomic data showed that isoA-containing bacteria were 5-fold more abundant in soil samples (1% of bacteria) than in the phyllosphere samples (0.2% of bacteria). Metagenomes obtained in previous studies of unenriched samples from high isoprene-emitting trees from temperate regions, such as poplar [38] and willow [32], were also analysed for comparison. Results showed that 0.7% of bacteria from soil beneath a willow tree and 0.02% of bacteria from poplar leaves contained isoA genes. These data, though sparse, showed the same trend observed in the oil palm environment, with soils containing greater numbers of bacteria with the genetic potential to degrade isoprene than the phyllosphere. This finding is surprising considering the greater availability of isoprene in the canopy than at ground-level [58] and indicates that soils could be a more important sink for isoprene than previously thought.
Recovery of metagenome-assembled genomes
Assembled contigs from soil and leaf metagenomes were used to reconstruct metagenome-assembled genomes (MAGs) using MaxBin2 [59]. A total of 20 MAGs from soil and 52 from leaf samples were obtained (Table S2). From these, two MAGs from soils and three from leaf 13C-heavy DNA metagenomes with > 75% completeness and < 10% contamination contained genes encoding homologous polypeptides to IsoABCDEF (E < 1e−40). MAGs containing IsoMO-encoding genes from soil incubations were taxonomically classified as Novosphingobium and Rhizobiales, and leaf MAGs were classified as Gordonia, Zoogloeaceae and Ralstonia (Table S3).
The Novosphingobium soil-associated MAG contained the full isoprene degradation gene cluster (isoABCDEFGHIJ) on a single contig along with aldH1, which encodes an aldehyde dehydrogenase [33]. However, no further accessory genes were recovered (Fig. 3). The products of these genes shared 76.2–100% amino acid identity (Table S4) with the corresponding polypeptides from Sphingopyxis sp. OPL5, a Sphingomonadales strain isolated from oil palm [32]. When the diversity and abundance of isoA genes in the 13C-heavy DNA fractions from soil samples were analysed by isoA amplicon sequencing, two amplicon sequence variants (ASVs), ASV44 and ASV11, were identified, closely related to the IsoA from the Novosphingobium MAG (> 99% amino acid identity; Table S5). These two ASVs represented 7.9% of the isoA genes of the 13C-heavy DNA fraction from replicate 2 and 11% from replicate 3, respectively (Fig. 4).
A Rhizobiales MAG was also reconstructed from 13C-heavy DNA soil samples. Although this MAG had high completeness (97.6%) and low contamination (2.5%), it showed a high strain heterogeneity (79%; Table S3), indicating that the MAG originated from DNA from one or more closely related microorganisms, thus making identification at a higher resolution difficult. Despite this, a complete isoprene degradation gene cluster (isoABCDEFGHIJ) plus aldH1 were located in a single contig (Fig. 3). When these genes were translated, they exhibited an amino acid identity of 54.8–84.9% to the homologous proteins from Sphingopyxis sp. OPL5, except for isoD, the product of which was more closely related to IsoD from Ramlibacter sp. WS9 [32] (Table S4). isoA gene amplicon sequencing analysis revealed that the ASVs closely related to the IsoA from the Rhizobiales MAG (> 71% amino acid identity; Table S5) dominated the isoA-containing bacterial community from all 13C-heavy DNA soil replicates, comprising a total of 64.9 ± 7.7% of the isoA genes in these samples (Fig. 4).
The Gordonia phyllosphere MAG was identified to the species level as Gordonia polyisoprenivorans i37 [48] (average nucleotide identity; ANI: 96.7%) and contained all IsoMO-encoding genes isoABCDEF, along with upstream genes isoGHIJ (Fig. 3). This MAG also contained aldH2, CoA-DSR and gshB, which are accessory genes often found within the isoprene degradation gene cluster in Gram-positive bacteria and encode an aldehyde dehydrogenase, a CoA-disulfide reductase and a glutathione synthetase, respectively [33]. All the genes associated with the isoprene degradation pathway recovered in the Gordonia MAG encoded polypeptides that shared > 94% amino acid identity to the corresponding proteins from Gordonia polyisoprenivorans i37, except for IsoB (82.1%) and AldH1 (83%; Table S4). When the 13C-heavy DNA fractions from leaf samples were analysed by isoA amplicon sequencing, several ASVs with a high percentage of amino acid identity to the IsoA from the Gordonia MAG (> 93%; Table S5) were recovered. However, ASV1, which showed 100% amino acid identity to IsoA from the Gordonia MAG, overwhelmingly dominated the isoA-containing bacterial community, representing 91.9 ± 7.3% of the isoA genes in these samples (Fig. 4).
The remaining two phyllosphere MAGs, Zoogloeaceae and Ralstonia, contained genes that encoded homologous polypeptides to IsoABCDEF, although they showed a low amino acid identity to the corresponding IsoMO proteins from well-characterised isoprene degraders (34.1–52.5%; Table S4). In addition, no gene homologues to isoGHIJ were recovered from these MAGs (Fig. 3). While the absence of homologues to isoGHIJ and the low sequence identity to IsoABCDEF indicates that these bacteria may harbour a novel isoprene degradation pathway, especially considering that these MAGs were recovered from 13C-heavy DNA fractions metagenomes, we cannot be absolutely certain that they are from bona fide isoprene degraders. These isoA-like sequences were not identified by the isoA amplicon analysis, suggesting that the relatively high number of mismatches with the isoA primers prevented successful PCR amplification. Therefore, further targeted isolations of these bacteria and/or expression of these isoABCDEF genes in a heterologous host [38] are required to establish that these are genuine isoprene degraders.
Finally, if Zoogloeaceae and Ralstonia microorganisms could be confirmed as bona fide isoprene degraders, the abundance of bacteria with the potential to metabolise isoprene would increase from 1 to 2% in oil palm soil and from 0.1 to 0.2% in phyllosphere unenriched samples. Similarly, the RA of isoA-containing bacteria would also increase in both willow soil (from 0.7 to 1.3% of bacteria) and poplar leaf (from 0.02 to 0.05% of bacteria) unenriched samples, indicating that isoprene degraders could be more abundant in the environment than previously thought. However, more samples from contrasting ecosystems need to be explored to support this hypothesis.
Isolation and characterisation of Variovorax sp. OPL2.2
Cultures set up using material from the DNA-SIP experiments from soils and leaves were subcultured three times at 2-week intervals with 25 ppmv isoprene before plating onto minimal medium with isoprene as sole carbon source (see the “Methods” section). A strain belonging to the genus Variovorax isolated from leaf enrichments was able to grow on isoprene as sole carbon and energy source (Fig S4).
Although DNA-SIP experiments have shown that Variovorax plays an important role in isoprene degradation in the phyllosphere [38], Variovorax sp. OPL2.2 is the first strain of this genus isolated from a tropical environment. When genomic DNA of OPL2.2 was screened for isoA, it yielded a PCR product which had a translated sequence with 99.4% amino acid identity to IsoA from Variovorax sp. WS11, an isoprene degrader isolated from willow soil [32].
The genome of this new isolate from oil palm, Variovorax sp. OPL2.2, was sequenced using Illumina and Nanopore technologies to confirm that it contained a full isoprene degradation gene cluster. Assembly of both Illumina and Nanopore reads and downstream analysis with CheckM [60] revealed that the Variovorax sp. OPL2.2 genome assembly comprised of 50 contigs totalling 8.5 Mbp and had a 98.4% completeness, 1.1% contamination and a GC content of 67.4%. Finally, after automatic annotation by Prokka [61], 8200 predicted coding sequences were found in the Variovorax sp. OPL2.2 genome.
Genome analysis confirmed that Variovorax sp. OPL2.2 contained isoABCDEF encoding IsoMO. isoGHIJ and aldH1 genes, which are involved in the subsequent steps of isoprene metabolism, were located upstream isoABCDEF in an identical layout to those of many bona fide isoprene-degrading strains [38] (Fig. 3). garB, which encodes a glutathione disulfide reductase, was also located in the same gene cluster (Fig. 3). isoABCDEFGHIJ, aldH1 and garB encoded polypeptides with high amino acid identity (99.7–100%) to those from Variovorax sp. WS11. Indeed, ANI analysis (> 99.9%) revealed that Variovorax sp. OPL2.2 is the same species as Variovorax sp. WS11.