The Variable Influence of Dispersant on Degradation of Oil Hydrocarbons in Subarctic Deep-Sea Sediments at Low Temperatures (0–5 °C)

The microbial degradation of petroleum hydrocarbons at low temperatures was investigated in subarctic deep-sea sediments in the Faroe Shetland Channel (FSC). The effect of the marine oil dispersant, Superdispersant 25 on hydrocarbon degradation was also examined. Sediments collected at 500 and 1000 m depth were spiked with a model oil containing 20 hydrocarbons and incubated at ambient temperature (5 and 0 °C, respectively) with and without marine dispersant. Treatment of sediments with hydrocarbons resulted in the enrichment of Gammaproteobacteria, and specifically the genera Pseudoalteromonas, Pseudomonas, Halomonas, and Cobetia. Hydrocarbon degradation was faster at 5 °C (500 m) with 65–89% of each component degraded after 50 days compared to 0–47% degradation at 0 °C (1000 m), where the aromatic hydrocarbons fluoranthene, anthracene, and Dibenzothiophene showed no degradation. Dispersant significantly increased the rate of degradation at 1000 m, but had no effect at 500 m. There was no statistically significant effect of Superdispersant 25 on the bacterial community structure at either station. These results show that the indigenous bacterial community in the FSC has the capacity to mitigate some of the effects of a potential oil spill, however, the effect of dispersant is ambiguous and further research is needed to understand the implications of its use.

The depletion of oil reserves onshore and in shallow waters has forced the industry to explore progressively deeper waters for as yet unexploited petroleum reserves. In addition, global warming has led to a dramatic shrinking of the sea ice in the Arctic, and hitherto inaccessible arctic oil reserves are now considered for exploration. Understanding the environmental implications of an oil spill in the cold and deep ocean is therefore urgent in order to improve oil contamination monitoring and optimise mitigation measures 1 . Most of our recent knowledge on the fate of hydrocarbons in the deep ocean stems from research following the catastrophic Deepwater Horizon (DWH) blowout in the Gulf of Mexico which occurred at a depth of 1500 m and resulted in the largest single marine oil spill to date. During DWH, approximately 4.9 million barrels of crude oil were released at depth and it has been estimated that between 2-15% of this oil was deposited onto deep-sea sediments by sedimentation of oil-contaminated marine snow 2,3 . Once in contact with the sediment, hydrocarbons can become entrapped for long periods via absorption into sediment organics [4][5][6][7] and impact sediment ecosystem function services [8][9][10][11][12][13] . The persistence in sediments of hydrocarbons recalcitrant to degradation may also have negative impacts on sediment communities biodiversity and pose significant health risks as accumulated hydrocarbons slowly enter the food web 14 .
Microbial biodegradation is a key process for the removal of oil from the marine environment [15][16][17][18] . Hydrocarbon-degrading microbes exist as part of the rare biosphere in the ocean, forming "seed populations" that can respond rapidly to hydrocarbon exposure 19 . In the case of DWH, a succession of hydrocarbon-degrading Background FSC sediment community. The natural background bacterial communities from both stations were dominated by Proteobacteria (81% of all reads) of which the class Gammaproteobacteria was the most dominant (63% of all reads). Only 15% of the OTUs were shared between the two stations and permutational multivariate analysis of variance (PerMANOVA) revealed that the stations were significantly different to each other (p = 0.008, R 2 = 0.26, n = 5) ( Supplementary Fig. 2).

Microbial dynamics during hydrocarbon degradation.
There was a shift in bacterial OTUs in all treatments during the incubation experiment (Fig. 5a), evidenced by the significant explanatory power of time in per-MANOVA analysis (p = 0.001, R 2 = 0.04, N = 77). The presence of model oil modified the shift of bacterial OTUs (p = 0.001, R 2 = 0.2, perMANOVA, N = 77) and the mass of hydrocarbon degraded was significantly predicted by Scientific RepoRts | 7: 2253 | DOI:10.1038/s41598-017-02475-9 ordisurf()(See Methods, Statistical Analysis) for position on the nMDS (p = 0.001, R 2 = 0.4, N = 77) with a trajectory from left to right corresponding to higher mass of hydrocarbon degraded (Fig. 5b). Superdispersant 25 did not have a significant effect on the bacterial community (p = 0.08, R 2 = 0.01, perMANOVA, N = 77) but sampling location did (p = 0.002, R 2 = 0.06, perMANOVA, N = 77) reflecting the differing natural communities used as starting points ( Supplementary Fig. 2). The station effect persists through the experiment, despite the similarity in the dominant OTUs in the model oil treated incubations and is visualised as a separation between the stations in the y-axis of the nMSDS plot (Fig. 5b).
In FSC1000 the natural background community was still present at the start of the incubations (day 0 and 2) however, in FSC500 there was a notable shift in the bacterial community structure between the natural background and at the start of the incubations (Fig. 5a). Subsequently to this, the bacterial OTUs shifted in a similar fashion for both stations, indicating that the natural oil degraders were not lost from FSC500 during transport and storage.
The control incubations in both stations were initially dominated by Colwellia psychrerythraea. This OTU remained dominant in FSC1000 until the end of the study whereas in FSC500 it was largely replaced by OTUs from the orders Marinicellales and Thiohalorhabdales. OTUs related to Colwellia aestuarii and Pseudoalteromonas were also predominant in FSC1000 controls from day 30 onwards but did not replace Colwellia psychrerythraea. In the model oil-treated incubations, Colwellia psychrerythraea was also initially dominant. However, bacterial communities in oil treatments diverged from those in controls with time and OTUs from the genera Pseudoalteromonas, Cobetia, and Halomonas predominated as hydrocabons were degraded. This shift was delayed in the FSC1000 model oil only treatment reflecting the later start in oil degradation (Figs 1-4). There were no specific OTUs related significantly to the presence of Superdispersant however, by causing the degradation process to proceed faster, the presence of dispersant affected the timing of bacterial succession. Links between bacterial taxa and hydrocarbon degradation. LEfSe analysis (See Methods, Statistical Analysis) was carried out to identify which taxa were significant biomarkers of hydrocabon degradation (Fig. 5c,d) 39 . This revealed general shifts at the class level with Gammaproteobacteria associated with the model oil treatments and Beta/Deltaproteobacteria associated with the control. At a lower taxon level, the Beta/ Deltaproteobacteria in controls was Nitrosomonas and Myxococcales. LEfSe also revealed differences between the control and model oil groups at lower taxonomic groups within Gammaproteobacteria. The order Alteromonas (specifically the genera ZD011, Moritella, and Colwellia) was dominant in controls while Oceanospirillales was associated with the hydrocabon treatments. Within the Oceanospirillales order, the genera Cobetia and Halomonas were associated with model oil while Marinomonas and Amphritea were associated with the control. The genera Pseudomonas and Pseudoalteromonas were also associated with model oil.

Discussion
Despite the wealth of information on the biotic response and environmental consequences of oil release and dispersant use in the deep sea following DWH, our knowledge on the environmental factors that regulate microbial hydrocarbon degradation and bioremediation efficiency remains incomplete 19 . Here, the first data is presented on the structure and function of HDB communities in the deep FSC, an area of active exploration and exploitation of oil reservoirs in deep waters and of particular oceanographic importance due to its role in the global thermohaline circulation. Furthermore, we investigated the effect on hydrocarbon degradation of the marine dispersant, Superdispersant 25, which is approved by the UK Food and Environment Research Agency and the most likely to be employed in the event of an oil spill in UK waters. Results showed that, without dispersant, the degradation of all hydrocarbons was considerably higher at 500 m (5 °C) compared to that at 1000 m (0 °C). Lower biodegradation of alkanes and PAHs at 0 °C compared to 5 °C has also been observed in seawater incubations from a Norwegian fjord 40 . It is not possible from results here to determine whether higher biodegradation at 500 m is purely due to the higher temperature or if differences in the bacterial community and sediment characteristics also contributed. However, basin-wide investigations have recently identified temperature as the main factor determining the oil biodegradation potential of bacterial communities in geographically separated oil-polluted sediments in the Mediterranean 41 . Regardless, our results demonstrate that oil will persist longer in the deeper sediments of the FSC and suggest that the impact of oil release in the deep FSC or contamination of deep-water sediments as a result of water mass circulation would be higher due to the sub-zero temperatures prevailing below 600 m.
Biodegradation of the hydrocarbon mixture after 50 days in 0 °C incubations was lower than that measured in arctic seawater incubations with a medium grade crude oil at −1 °C over 60 days 42 ; in the latter case 58-61% of total measurable oil was lost over the incubation period and specific components (heptadecane, octadecane,  Table 1. Summary of total hydrocarbon degradation for each group after 50 days and results of comparison between rates between model oil and model oil + dis treatments with ANCOVA. n = 15 with 3 per time point for ANCOVA and n = 3 for total degradation after 50 days. naphthalene, phenanthrene, dibenzothiophene) reached complete degradation 42 . In our study, the same components showed less than 50% degradation after 50 days (Table 1) with the most extreme difference observed in dibenzothiophene degradation which was zero. It is, therefore, reasonable to suggest that PAHs, such as fluoranthene, anthracene, and Dibenzothiophene, may become entrapped in the subsurface sediments of the deep FSC and could be periodically released by bioturbation or incorporated into faunal biomass resulting in long term impacts. Hydrocarbon degradation without Superdispersant 25 at 5 °C after 50 days was >75% for both aliphatic and aromatic hydrocarbons (except anthracene, Table 1) with most activity observed after 30 days. This lag period in the onset of degradation may be the reason why Baelum et al. 9 found merely 25% removal of total hydrocarbons in 5 °C incubations of Gulf of Mexico deep seawater with Macondo oil after 20 days. Higher biodegradation of alkanes and PAHs (~90%) at 5 °C were measured by Campo et al. 32 in 42-day incubations with light crude but these experiments were performed in artificial seawater inoculated with an enriched oil-degrading bacterial culture which could have accelerated the degradation process 43 . The environmental lifetime of hydrocarbons in FSC sediments predicted by linear regression ranged between 2 and 5 months (at 5 °C and 0 °C respectively, Supplementary Table 4), at least for those components subject to bacterial degradation. It should be noted however that degradation rates from laboratory incubations (although a vital tool) cannot be extrapolated to natural ecosystems without caution 8 . Hydrocarbon degradation appears to progress much slower in situ; in comparison, Liu et al. 44 found only light to moderate ( < 25%) weathering of oil in heavily contaminated sediments 1 year after the DWH spill and subsequent studies in the Gulf of Mexico have confirmed the presence of Macondo oil in deep-water sediments 4 years after the DWH spill although reduced in mass by 80-90% compared to levels immediately after contamination 45 . Migration of oil in deeper sediment layers where anaerobic conditions prevail may partly explain slower biodegradation rates in the field 30,46 . In the case of the Exxon Valdez and the West Falmouth oil spills, hydrocarbons have been shown to persist for decades after contamination in the subsurface layers of coastal sediments 47,48 . Hydrocabon contamination of FSC sediments here resulted in the enrichment of Gammaproteobacteria, specifically the order Oceanospirillales and the genera Pseudoalteromonas and Pseudomonas. Oceanospirillales and Pseudomonas taxa also became enriched in the DWH deep water plume and were most abundant during the initial, unmitigated flow phase of the spill when the concentration of hydrocarbons and the fraction of insoluble n-alkanes and cycloalkanes were highest 20,49,50 . The enriched Oceanospirillales taxon in the DWH plume was an unclassified and uncultivated OTU within the family Oceanospirillaceae 20 whereas in our study the enriched taxa belonged to Halomonadaceae. Therefore, although at higher taxonomic levels, such as class and order, the main HDB may be similar between geographic locations; site-specific research is required for fine low taxon detail.
The LEfSe analysis identified two biomarker OTUs within the Halomonadaceae, Cobetia and Halomonas, which were associated with hydrocabon contamination in our incubations. The genus Cobetia has been previously isolated from oil-contaminated coastal waters of the Persian Gulf and is known to produce biosurfactants and degrade hydrocarbons including phenanthrene and dibenzothiophene which were present in the model oil   51,52 . However, Cobetia was not identified among HDB in the Gulf of Mexico. Halomonas was one of the dominant HDB in contaminated surface waters and, to a lesser extent, the deep water plume during the DWH spill as revealed by DNA stable isotope probing and cultivation-based methods 53 . The relative abundance of Pseudoalteromonas, which also became enriched in our model oil incubations, increased in the DWH deep water plume during partial capture as petroleum hydrocarbon concentrations decreased and the more dilute plume consisted of more BTEX relative to alkanes 20 . The decrease of ZD0117 in oil-enriched treatments in this study has also been observed in Antarctic seawater samples following contamination with oil 54 . The genus Colwellia is thought to play a predominant role hydrocarbon degradation in the deep sea. This is supported by the psychrophilic lifestyle 55 and the ability of members of Colwelliaceae to produce EPS under low temperature and high pressure 56 along with the metabolic potential for hydrocarbon degradation 50 . Indeed, Colwellia was one of the major taxa that became enriched in the DWH deep water plume and was the dominant bacteria in flocs and contaminated deep water sediments in the Gulf 9, 10, 57 . In this study, Colwellia was one of the most abundant taxa in both control and hydrocabon-contaminated incubations. This is not surprising since Colwellia is a heterotrophic group that does not depend solely on external hydrocarbon supply for growth. Yang et al. 24 reported the presence of Colwellia in both oily and non-oily Gulf sediments collected 5-7 months after the DWH spill (September, October and November 2010) and also concluded that Colwellia could be autochthonous to deep surficial sediments and cannot be unambiguously linked to oil contamination. The fact that Colwellia was abundant in control incubations here but not in the natural background samples may suggest that it is a strong competitor in enrichments/microcosms. There was a shift in the FSC500 bacterial community between background and day 0 incubations, while this was not the case in FSC1000 (Fig. 5). The shift is mainly related to the genre Colwellia, further demonstrating its robustness during sampling and storage, and could be attributed to the longer interval (one week) between sampling FSC500 and initiation of the incubation experiment compared to FSC1000. Nevertheless, the similar response of bacterial communities in FSC1000 and FSC500 to hydrocarbon contamination suggests that results are representative. It is however important to stress that though there are many benefits to the microcosm strategy employed in this study, not least statistical power gained from replication and destructive sampling, many questions remain as to how hydrocarbon degradation would proceed in the field. This study should be seen as a first step and further study in the field is required.
Studies on dispersants have generated confounding results, most likely because of the non-specific metrics used to define biodegradation (e.g. mineralisation to CO 2 , chemical disappearance quantified by GC-MS) and differences in the state and concentration of oil, as well as concentration and perhaps most significantly type of dispersant used in laboratory incubations 28,29 . Previous studies about microbial dispersant impacts have generally focused on Corexit, the dispersant used after DWH. In this study we used Superdispersant 25, as it is approved for used in the FSC by the EU and UK and very little is known about its environmental effects. We do not know to what extent the effects of Corexit can be compared with Superdispersant 25, and to our knowledge this is the first study to investigate its effects on the degradation of hydrocarbons. It is known that Superdispersant 25 is toxic to a range of marine animals, both with and without the presence of hydrocarbons 58, 59 , although its toxicity is less than that of Corexit 9527 59 , In the case of Corexit 9500, hydrocarbon degradation by deep water bacterial communities in the Gulf of Mexico (~5 °C) was either unaffected 32 , enhanced 60 or even inhibited 27 . In the latter case, suppressed degradation of certain classes of hydrocarbons, particularly alkanes, was associated with lower bacterial protein synthesis and exoenzyme activities in dispersant treatments. Arctic seawater microorganisms from the Chukchi Sea degraded similar amounts of oil at −1 °C with and without the dispersant Corexit 9500 42 . In this study with Superdispersant 25, we observed both no difference (5 °C) and significantly increased (0 °C) amount of hydrocarbons degraded by day 50. The presence of dispersant diminished the effect of temperature on the relative rates of hydrocarbon degradation, i.e. the amount of aliphatic and aromatic hydrocarbons degraded after 50 days was comparable between 0 °C and 5 °C when dispersant was present whereas degradation was significantly depressed at 0 °C without dispersant. Despite the apparent lack of impact of dispersant on the final amount of hydrocarbons degraded at 5 °C, the onset of degradation of several less soluble compounds was initiated earlier when dispersant was present (hydrocarbon groups 2, 3, and 4). In contrast, more soluble hydrocarbons (n-decane and 1-decene, Group 1) exhibited similar degradation patterns with and without dispersant suggesting that only the less water soluble hydrocarbons have their bioavailability increased by Superdispersant 25. Nevertheless, it seems that the effect of dispersant on degradation is specific for individual hydrocarbons 25,61 and depends on the characteristics of the dispersant to be used 61 . These observations highlight the challenge of predicting the consequences of dispersant use to mitigate oil spills in the natural environment where a number of physicochemical and biological factors may also contribute to hydrocarbon biodegradation 43,62 . Differences in hydrocarbon degradation patterns between treatments here were not accompanied by shifts in the bacterial community structure that could be specifically linked to dispersant application, although there could have been differences at species, strain or genome level not detected with our sequencing strategy. Previous studies have demonstrated taxa-specific and concentration-dependent effects of dispersants on microorganisms; certain Colwellia and Alcanivorax taxa, for example, responded to dispersants or oil-dispersant mixtures 9,27,63,64 whereas Marinobacter 27 and Acinetobacter 63 were suppressed by dispersants. In contrast to most studies, however, relatively low concentrations of hydrocarbons and Superdispersant 25 were applied in our incubations to simulate marine biodegradation of chemically dispersed oil at the low end of the range of dispersant concentrations found after a spill, which could explain the lack of effect of Superdispersant 25 on the bacterial community composition 65 .
Overall, this study is the first to demonstrate the capability of deep water, subarctic sediment microbial communities to degrade hydrocabons at temperatures of 0 and 5 °C. Hydrocarbon degradation was dependent on temperature with significantly slower degradation of both aliphatic and aromatic hydrocarbons at 0 °C compared to 5 °C. Certain PAHs, such as fluoranthene, anthracene and dibenzothiophene did not show detectable levels of degradation at 0 °C suggesting long-term impacts of oil contamination at near zero or subzero temperatures. The bacterial community response was consistent between 500 and 1000 m with the class Gammaproteobacteria, orders Oceanospirillales and Pseudomonadales, and genera Pseudoalteromonas, Pseudomonas, Halomonas and Cobetia predominating in model oil-contaminated treatments. The high degree of consistency between stations in terms of bacterial response to the model oil provides encouragement that oil contamination could be monitored by proxy of bacterial changes in the FSC. Robust and highly portable routine molecular methods, such as qPCR, could be developed to detect such changes. Due to the Arctic origin of bottom water masses in our study area, results here are relevant to oil biodegradation efficiency and bioremediation potential of bacterial communities in Arctic ecosystems where the ongoing exploration for oil and gas in offshore areas and a growing interest in developing the Northern Sea Route (NSR) as an alternative transportation route increase the risk of hydrocarbon pollution.

Methods
Sediment collection. Sediment was collected from two stations at 538 and 994 m (subsequently referred to as stations FSC500 and FSC1000, respectively) in the FSC using a day grab (Supplementary information 1). Sampling was conducted on the FRV Scotia (cruise number Sc201405) between 24/04/14-08/05/14. Sediment from five grabs collected at each station was pooled and stored at 1 °C under approximately 10 cm of ambient seawater. The overlying water was constantly aerated with an aquarium air pump (Airvolution AV2, Interpet, UK) and air stones until slurry incubations were initiated. Additionally, approximately 10 g of sediment from each grab were immediately stored at −80 °C for molecular analysis. Experiments were initiated after returning from the cruse, so were stored for 14-21 days before the start of the experiment, control incubations were run so we could account for any changes in the community due to storage and transport.

Preparation of Model oil.
A model oil containing 20 hydrocarbons was formulated with a mixture of aliphatic, aromatic, polyaromatic (PAH), and resin components proportions are shown in Supplementary Fig. 1.
The density of the model oil was 880 kg/m 3 , equivalent to a medium crude oil. The model oil was based on hydrocarbon molecules containing more than 8 carbon atoms as it is thought that these components are more likely to be deposited on the seabed 2, 66 . Although olefins/alkenes are present in crude oil (see ref. 67) these are usually present at lower concentrations than employed in the model oil prepared here. Our model was loosely based on the composition of a North sea crude (Schiehallion) however, this had to be manipulated in such a manner as to ensure mutual solubility of all components.  Table 1). The treatments are hereafter referred to as model oil (model oil only) and model oil + dis (model oil and dispersant). Each treatment was run in triplicate for each time point with a live control (sediment and seawater only) and sterile control (autoclaved sediment with model oil or model oil + dis) run in parallel. To prepare the incubations, 15 ml of ambient seawater was added to the vials which were then sterilised by autoclaving at 121 °C and 100 kPa for 15 min. After autoclaving, 5 ml of sediment was added to each vial and, depending on treatment, was spiked with 0.44 g of model oil and 0.015 g Superdispersant 25 (ratio of 30:1 oil: dispersant to mimic application at blowout location in accordance with manufacturer's instructions). The vials were then incubated in the dark on a shaker table at 150 rpm at 5 °C or 0 °C for FSC500 and FSC1000, respectively (reflecting the in situ temperature at each station). On days 10 and 30, the headspace in each vial was replenished with sterile air through the septa to ensure the conditions remained aerobic. Destructive sampling was carried out on days 0, 2, 10, 20, 30, and 50 for all treatments. Approximately 2 ml of the sediment were collected aseptically and stored at −80 °C until molecular analysis. The remainder was stored in the glass incubation vial at −20 °C until GC analysis.
Hydrocarbon extraction and analysis. To determine the extent of hydrocarbon loss due to evaporation, 1 g of model oil was incubated in glass vials (n = 3) at room temperature. A decrease in mass of less than 1% was detected after 50 days. Hydrocarbons were extracted from the slurries by liquid-liquid extraction with 3 × 10 ml dichloromethane (DCM) (VWR, UK). 1 µl of the DCM fraction was then subjected to GC analysis using toluene as an internal standard to correct for injector error (20 µl/ml toluene). Samples were analysed on a Varian CP3800 fitted with 30 m Zebron ZB-50 column and FID detector. The hydrocarbons were quantified against an external standard containing known amounts of the model oil components. The conditions for the analysis were as follows; injector and detector temperature was 330 °C and the injection split ratio was 80:1, initial oven temperature was 50 °C with a 3 min hold and then increased at 10 °C min −1 to 110 °C, followed by an increase to 200 °C at 5 °C min −1 with at 12 min hold, and finally ramping to 300 °C at 20 °C min −1 . All components showed recovery rates 90-95% after 50 days in the sterile controls (Supplementary Table 2). The exception to this was ethylbenzene and xylenes which showed recovery levels of < 75%. The partial loss of these specific components cannot be accounted for at this stage but probably reflect the greater relative volatility of these components and have been excluded from further analysis.
Microbial community analysis. Total

Statistical analysis.
Statistical analysis was carried out in R 78 and the package Vegan 79 .
To simplify results, hydrocarbons that had similar degradation profiles were grouped by principal component analysis. The raw data for individual components is given in Supplementary Data 1. Group 1 comprised the two shortest chain aliphatics, the alkane n-decane and the alkene 1-decene. Group 2 comprised the medium length aliphatic hydrocarbons tetradecane, pentadecane, hexadecane, heptadecane, icosane, and docosane. Group 3 comprised the aliphatic alkane dodecane, and the alkene octadecene. Group 4 contained tetracosane (Supplementary Table 3). To test whether degradation rates were significantly different between conditions, analysis of covariance (ANCOVA) was used to compare the slopes of the linear regressions of hydrocarbons degraded over time with the formula "hydrocarbon mass ~ day * treatment". Significantly different degradation rates were accepted at p < 0.05, or if slopes were incomparable (i.e. did not follow the same x, y relationship) comparison for sterile versus live incubations is given in Supplementary Table 4.
To correct for variation in sequencing depth between samples, rarefaction was carried out using the r function "rrarefy()". For beta diversity a distance matrix was created with the function "vegdist()" using the Jaccard index. This was visualised with non-metric multidimensional (nMDS) scaling using the function "metaMDS()". To test for differences between sampling groups, permutational multivariate analysis of variance (perMANOVA) using distance matrices was carried out using the function "adonis()" with 1000 restarts. To check if hydrocarbon concentrations were correlated with the nMDS, ordisurf()was used to fit vectors to the surface ordination plot with generalized additive models. Significance was accepted at p < 0.05. Linear Discriminant Effect Size analysis (LEfSe) was used to identify bacterial shifts associated with different treatment conditions using the Galaxy pipeline (available at https://huttenhower.sph.harvard.edu/galaxy/) 39 . First LEfSe identifies features (e.g. clades or OTUs) that are significantly differently abundant between groups with non-parametric factorial Kruskal-Wallis sum-rank tests; it then interrogates these features for biological consistency and relative importance with Wilcoxon rank-sum tests among subclasses and Linear Discriminant Analysis to determine the effect sizes. Data Availability. Sequences from this study are available through the European Nucleotide Archive under project PRJEB15950 (ERP017791). All other data generated or analysed during this study are included in the Supplementary Information file and tables.