Extending SEQenv: a taxa-centric approach to environmental annotations of 16S rDNA sequences

Sugarcane dataset

The environmental terms for the sugarcane dataset are illustrated in Fig. 2. Samples belonging to rhizosphere showed the environmental term “soil” as being the most prevalent (Fig. 2A). Other similar terms were also observed, such as “rhizosphere”, “forest soil”, “prairie” and “agricultural soil”. Of importance was the occurrence of the environmental terms such as “activated sludge”, “garden” and “contaminated soil” as more taxa with these metadata were prevalent in these datasets. Soil samples illustrated a similar collection of environment terms, with the “soil” term being the most significant environmental term observed (Fig. 2B). However, “forest soil” was observed relatively strongly here compared to rhizosphere results.

For the stem samples, the environmental term “garden” was strongly observed compared to other terms, with exception of “soil” (Fig. 2C). Other important terms included “forest soil”, “biofilm” and “garden soil”. Lastly, the root samples showed similarity to both soil and rhizosphere samples, where the environmental term “soil” was the most observed term (Fig. 2D). While the plots show similar list of environmental terms, with exception of the few most strongly observed terms, the ranking of the terms themselves vary across these habitats.

Across these different samples, SEQenv was unable to generate environmental terms for some ENVO IDs. This included 1000196, which stood for “coniferous forest biome”, 446, which was “terrestrial biome” and lastly 447, which was “marine biome”, 1000181, which was “mangrove biome”, 428, which was simply “biome” and 2030, which was “aquatic biome”.

Overall, the environmental term “soil” was prevalent across all sub-habitats, however other terms were ranked differently. Stem sub-habitat was more unique compared to soil, rhizsophere and root. This is also illustrated in Table 2, where the top 10 environmental terms are ranked according to their abundances for each sub-habitat, with the differences between sub-habitats highlighted in bold.

Marine dataset

For the coral atoll marine samples, the environmental term “sea water” was the most observed term, with “sea” coming up after that (Fig. 3A). A few other terms of importance include “bay”, “coral reef”, “ocean” and “sediment”. Southern ocean samples also showed a similar list of environmental term (Fig. 3B). Here as well “sea water” and “sea” terms were the most observed; however, “brine pool” was relatively strongly observed here compared to samples from coral atoll.

Similar to samples from sugarcane dataset, SEQenv was unable to determine the environmental term for the IDs 428, which was “biome”, 447, which was “marine biome” and 1000047, which was “mediterranean sea biome”.

While both marine samples showed a similar list of environment terms, these differed in the ranking of the terms themselves, which is illustrated in Table 3. Here, the ranking of top level environmental terms were the same for both coral atoll and southern ocean samples, however differences were observed in the lower ranked terms where “coral reef” was observed for coral atoll samples while southern ocean had environmental terms like “Mediterranean” and “marine bulk water”, which were absent in coral atoll samples.

Sugarcane dataset

While the environment terms “Soil” and “Forest Soil” were similar, the sequences that contribute to these terms differed (Fig. 4A and 4B respectively). This was quite apparent in the differences between both word clouds where the most abundant taxon for “soil” term included Acidothermus and Chloroplast, which is potentially a misclassified Cyanobacteria while Variibacter and Acidobacteriaceae were more strongly related to the “forest soil” term.

“Rhizosphere” environmental term had Burkholderia as being the most abundant taxon while Acidothermus was almost non-existent in this case as shown in Fig. 4C. Burkholderia was followed by Catenulispora sp. Neo1, Acidobacteriaceae (Subgroup 1) and Dyella.

The “garden” environmental term had distinct taxa, which were not observed in other environmental terms (Fig. 4D), with the members of Pantoea genus making up the collection of taxa.

The “Contaminated soil” term is an example of significantly different collection of taxa (Fig. 5A). While not listed in the top 10 environmental terms for the sugarcane dataset, it and “waste” environmental term consists of important collection of taxa. Here, Sphingomonas, Pseudomonas and Undibacterium were more abundant, in that order.

For the “waste” environmental term, Acidothermus was the most observed taxon, followed by Acidobacteriaceae (Subgroup 1) (Fig. 5B). Additionally, members of Chitinophagaceae family were also seen under this environmental term.

Marine dataset

Prochlorococcus dominated the taxa abundance for the environmental term “sea water” (Fig. 6A). The other taxa such as SAR11 clade and SAR86 clade were also observed, though at a lower abundances. On the other hand, while similar taxa was observed for the environmental term “sea”, the relative abundance of these taxa were significantly different (Fig. 6B). SAR11 clade and Synechococcus became more abundant, while Prochlorococcus was observed to be far less prominent than what was observed for the environmental term “sea water”.

“Ocean” environmental term showed Chloroplast becoming more abundant compared to per term taxa abundance for “seawater” and “sea” environmental terms (Fig. 6C). Similar behaviour was seen for Marinimicrobia (SAR406 clade) as well, which was very low in abundance in the aforementioned environmental terms.

SAR86 Clade and Prochlorococcus jointly dominated the “Brine Pool” environment term (Fig. 6D). Furthermore, a few taxa such as SAR324 clade (Marine group B) and Alteromonas were also observed, although at a very low abundance. Species diversity was observed to be quite low in this case as only a few taxa contributed to this environment term. Overall, most of the taxa belonged to Proteobacteria and Cyanobacteria phyla.

Sugarcane dataset

Per taxon environmental term relative abundance for Acidothermus and Burkholderia is illustrated in Fig. 7. While the environmental term “soil” dominated the list of terms for both genera, 52.6% for Acidothermus and 45.7% for Burkholderia, differences were observed for the lower ranked terms, where “Forest soil” was the second most observed term for Acidothermus at 9.9% (Fig. 7A), however for Bulkholderia the term “rhizosphere” was observed higher than “forest soil”, accounting for a significant portion of the environment terms at 20% and “forest soil” term accounting for 10% here, similar to Acidothermus.

The terms “woodland”, “waste” and “rice field” were ranked higher for Acidothermus (Fig. 7A) as well, at 5.5%, 4.9% and 4.8% respectively. For Bulkholderia, the “waste” term was not in the top 6 environmental terms (Fig. 7B), and furthermore the terms “woodland” and “rice field” were not observed for this genus.

“Field soil”, “peat swamp” and “sludge” environmental terms were observed for Bulkholderia at 3.1%, 2.6% and 2.1% respectively, however they were absent from the collection of top 6 terms for Acidothermus. Lastly, the remaining collection of environmental terms came at 15.5% for Acidothermus and 16.4% for Bulkholderia.

Overall, distinct differences were observed between both genera. For Acidothermus, with exception of the most abundant “soil” term, others gradually decreased in how much they accounted for in the list of environmental terms. However Bulkholderia showed the gradual decrease after the third ranked “forest soil” term.

Marine dataset

The per taxon term abundance pie charts for the genus Prochlorococcus and Synechococcus are illustrated in Fig. 8. For Prochlorococcus, the environmental terms “sea water” was the most observed term, accounting for 73.35% of environmental terms, an overall majority (Fig. 8A), which came down to third rank for Synechococcus, at 26.5% (Fig. 8B).

Furthermore, “reef” was observed strongly for Synechococcus at 27%, however the term was absent in the top 4 list for Prochlorococcus. Furthermore, “ocean” term was present for Prochlorococcus at 1.49%. Other differences included the term “brine pool” at 9.17% for Prochlorococcus, although it was absent for Synechococcus. Lastly, “coast” environmental term was observed only for Synechococcus at 2.7%.

Overall, environmental term distribution was different between both genera. A single “Sea water” term dominated Prochlorococcus list of environmental terms while Synechococcus saw three terms accounting for most of the environmental terms observed, on an almost equal level and where “reef” term was distinctly observed for Synechococcus.

Sugarcane dataset

While the results for the root sub-habitat were similar to soil and rhizosphere, differences were observed for the presence of environmental terms such as “sediment” and “biofilm”, which were ranked higher. This might be because of the taxa that belonged to these terms was more abundant in the root habitat due to plant-soil close association (Garbeva, Veen & Elsas, 2004). Furthermore, biofilms play an important role in plant-microbial interactions in the rhizsophere (Danhorn & Fuqua, 2007). Additionally “forest soil” environmental term was relatively more prominent for soil samples compared to rhizosphere results and this might be due to difference in abundance of taxa that are more prevalent in forest soils, which are located further away from the phytobiome system (Garbeva, Veen & Elsas, 2004).

The differences between stem samples and others, which was driven by terms such as “garden”, “glacier water” and “ground water” can be explained by the these terms being driven by taxa unique to the stem habitat and likely to be endophyte in nature. These taxa live within the plant biomass in a symbiotic relationship (Gouda et al., 2016) and therefore observed in the samples taken from the stem. This may be a result of comparatively lower number of sequences from these habitats exist in the database. Nonetheless, given that SEQenv (Sinclair et al., 2016) acquires isolation sources based on the sequences in the dataset, the differences in species found in the stem samples compared to other samples led to strong difference in environmental based tagging. Furthermore, as the stem samples were taken from the stem of sugarcane plants, the ranking of environmental terms in this case are a good representative of the type of the environment the microbial sequences came from. This highlights the value of SEQenv in discriminating between habitats.

Marine dataset

While most of the environmental terms observed for the two different marine based sub-habitats were similar, the ranking of the terms themselves were different and some environmental terms were uniquely observed such as “coral reef” environmental term for coral atoll sample, due to differences in the environment between these two sub-habitats and the variation in taxa abundance that comes with it (Jeffries et al., 2015). Some microbial communities in coral reef systems exist in a symbiotic relationship with coral polyps, playing a role in nutrient cycling as well as assisting in disease resistance for these organisms (Garren & Azam, 2012). Therefore taxa belonging to this environment are more likely to be observed for coral atoll samples. “Marine bulk water” was uniquely observed for southern ocean samples while being absent for coral atoll samples, due to the environmental characteristic of the ocean waters and the taxa that are prevalent in it.

For all the datasets used for analysis, it was apparent that SEQenv was able to determine intra-habitat differences and patterns even at environmental term level of information.

Sugarcane dataset

The difference in the abundance of taxa between “rhizosphere” and “soil” terms illustrate that while some taxa were common across different environment, the abundances observed were different.

Acidothermus, which was strongly observed in the “soil” environment term, is a thermophilic, acidophilic, cellulolytic bacterium, prevalent in acidic environments (Mohagheghi et al., 1986), while Acidobacteriaceae as observed more in the “forest soil” environment term, is a family of Acidobacteria which are ubiquitous in soil environment (Quaiser et al., 2003).

The “garden” environment term was significantly different from other terms in the case of the sequences that contributed to it where “Pantoea” was the most abundant taxa observed. It is well known that Pantoea Spp. lives in many plant tissue both as commensal and in some cases as pathogens (Pataky et al., 2000).

Members of the Sphingomonas genus were observed for the environmental term “contaminated soil” and bacteria belonging to this genus is well known to have the ability to degrade chemicals in contaminated soil as it is one of the best known genus for biodegradation of chemical contaminants (Alvarez et al., 2012; Schmidt et al., 1992; Ye et al., 1995). The most prevalent species of Sphingomonas was observed to be an “uncultured marine bacterium”. The presence of this bacterium here may be due to this taxa being prevalent in both contaminated soil and marine habitats.

Lastly, while taxa that contributed to the terms such as “contaminated soil” and “waste” were not as abundant as the aforementioned terms like “soil”, “forest soil” or “garden”, they were nonetheless very important as they provided taxa abundances under a specific environmental context. Therefore, for studies that may aim towards a specific goal in mind, such as bioremediation, this may help in targeting sequences that come from relevant environments.

Marine dataset

Overall, Marine habitats showed an interesting collection of taxa that come from a variety of marine environments. Similar to the sugarcane dataset, while the list of sequences contributing to each environment may seem similar at first, there were exceptions where unique sequences were observed to be more abundant in specific environments. Furthermore, the ranking itself varied across every environmental term. Additionally, similar to the differences observed for “soil” and “forest soil” environmental terms in the sugarcane dataset, “sea” and “seawater” exhibited the same pattern with respect to taxa observed.

Prochlorococcus, which was observed in multiple environmental terms such as “sea water”, “sea” and “ocean” in different abundances, is a very small marine cyanobacteria, which is one of the most abundant photosynthetic organism on the planet (Partensky, Hess & Vaulot, 1999), while bacteria belonging to SAR11 clade are accountable for methane dissolved in the oceans (Carini et al., 2014) and are dominant marine heterotrophs. They are cosmopolitan and abundant across marine habitats, particularly SAR11 (Brown et al., 2012), which is a highly abundant marine bacterium and was present for most environment terms at different abundances.

Synechococcus is a unicellular cyanobacteria that is prevalent in the marine environment and has been shown to dominate in this system (Jeffries et al., 2015). It was present for the environmental terms “sea” and “sea water”, while being absent in top 10 ranked list of taxa for the “ocean” term. SAR86 Clade, members of which are aerobic chemoheterotroph (Dupont et al., 2012), and the aforementioned Prochlorococcus jointly dominated the “Brine Pool” environment term.

The per environment term taxa abundance provided a more concise and relevant view of the environmental annotations. Linking sequences to environmental terms in such a manner would be more suitable than a list of environmental terms that SEQenv provides. This enhancement significantly improved the analysis capability of SEQenv system and provided a novel approach to contextual, taxa based environmental annotation, which was originally not present in the SEQenv pipeline. Furthermore, the integration developed here enabled a more thorough approach towards 16S rRNA sequence analysis and offers a single pipeline for both taxonomic and environmental annotation of sequences.

Sugarcane dataset

In accordance with per term taxa abundance result for “soil” environment term, “soil” dominated the list of terms for Acidothermus, which is a thermophilic and acidophilic microbe that is found in acidic environment (Mohagheghi et al., 1986). Other terms such as “forest soil” or “woodland” point towards these environments being favourable to its growth, as it has been observed in samples collected from forest environment (Kim et al., 2015; Meng et al., 2013).

Burkholderia occupies a variety of environmental niches (Compant et al., 2008) including soil (Janssen, 2006) and some strains of this genus can cause diseases for humans and animals (Coenye & Vandamme, 2003). Furthermore, the bacterium is observed to be prevalent in rhizsophere environment for plants (Caballero-Mellado et al., 2007), which may be the reason why the environmental term “rhizosphere” was strongly observed for it as compared to Acidothermus. Finally, the presence of the term “sludge” maybe be due to its potential and application for biodegradation (Zhang et al., 2013). Overall, this data support the widespread distribution in plant rhizosphere of these taxa in multiple niches.

Marine dataset

Prochlorococcus, one of the most abundant organism on the planet (Partensky, Hess & Vaulot, 1999), is typically observed in oligotropic oceans where nutrients availability is poor, in contrast to Synechococcus that favours nutrient rich environment (Whitton, 2012). Hence terms such as “ocean” and “brine pool” pointed towards prevalence of Prochlorococcus in these environments.

The list of terms for Synechococcus includes “reef” and “coast” which are nutrient rich environments compared to oceans. In fact the bacterium has been observed to be present in high abundance at coral reefs especially during summer time (Moriarty, Pollard & Hunt, 1985) as well as coastal regions such as the Portuguese coast (Martins et al., 2005).

Overall, the enhancement provided robust data on taxa-specific distribution in different habitats and highlights the usefulness of this approach for delineating the niches potentially occupied by specific taxon, in this case supporting the known distribution of these abundant marine autotrophs, which drive primary production (Christaki et al., 1999).