Microbial Indicators of Diesel Fuel Toxicity in Polar Soils

MICROBIAL INDICATORS OF DIESEL FUEL TOXICITY IN POLAR SOILS

Josie van Dorst

A thesis in fulfilment of the requirements for the degree of Doctor of Philosophy

School of Biotechnology and Biomolecular Sciences Faculty of Science UNSW Australia

TABLE OF CONTENTS

TABLE OF CONTENTS ...... ii ABSTRACT ...... vi ACKNOWLEDGEMENTS ...... viii LIST OF PUBLICATIONS ...... ix LIST OF FIGURES ...... xi LIST OF TABLES ...... xivv LIST OF ABBREVIATIONS………………………………………………………………………………………………………………xvi

1 INTRODUCTION ...... 1 1.1 HUMAN IMPACTS IN THE ANTARCTIC REGION ...... 1 1.2 PERSISTENCE OF PETROLEUM HYDROCARBONS IN TERRESTRIAL ANTARCTIC AND SUBANTARCTIC ECOSYSTEMS ...... 2 1.3 RESPONSE OF SOIL MICROBIAL COMMUNITIES TO PETROLEUM HYDROCARBONS ...... 5 1.4 FUNCTIONAL GENES ASSOCIATED WITH HYDROCARBON DEGRADATION ...... 7 1.5 BIOREMEDIATION OF PETROLEUM HYDROCARBONS IN POLAR SOILS ...... 7 1.6 EXISTING REGULATIONS IN ANTARCTICA ...... 10 1.7 GUIDELINES SPECIFIC TO PETROLEUM HYDROCARBON CONTAMINATION IN POLAR SOILS 11 1.8 ECOTOXICOLOGY IN POLAR SOILS ...... 13 1.9 TOWARDS DEVELOPMENT OF MICROBIAL INDICATORS ...... 15 1.10 RESEARCH SITES ...... 18 1.10.1 Macquarie Island ...... 18 1.11 AIMS ...... 21 2 NOVEL CULTURING TECHNIQUES AND THE MICROBIAL RESPONSE TO DIESEL FUEL IN A SUBANTARCTIC SOIL...... 22 2.1 INTRODUCTION ...... 22 2.2 METHODS ...... 23 2.2.1 Site and sampling methods ...... 23 2.2.2 Cultivation on low nutrient artificial media ...... 24 2.2.3 Micro-cultivation with SSMS ...... 25 2.2.4 Epi-fluorescent microscopy (EFM) ...... 28

2.2.5 Secondary cultivation on low nutrient artificial media ...... 28 2.2.6 DNA extraction of pure isolates ...... 28 2.2.7 Identifying unique isolates with restriction fragment length polymorphisms (RFLP) .. 29 2.2.8 Sanger sequencing of pure cultured isolates ...... 29 2.2.9 DNA Extraction from SSMS membrane...... 30 2.2.10 DNA Extraction from soil ...... 30 2.2.11 Pyrosequencing of the 16S rRNA gene ...... 30 2.2.12 Calculation of diversity estimates from pyrosequencing data ...... 31 2.2.13 Dose-response modelling ...... 32 2.2.14 Response of individual genera to SAB ...... 32 2.3 RESULTS...... 33 2.3.1 Comparison of the methods used to capture bacterial diversity ...... 33 2.3.2 Cultivation from soil and micro-cultivation from SSMS onto artificial media ...... 35 2.3.3 Bacterial diversity from SSMS enrichments ...... 39 2.3.4 Bacterial diversity directly from soil ...... 41 2.3.5 Response of individual genera to TPH ...... 42 2.3.6 Total bacterial community dose-responses to TPH ...... 45 2.4 DISCUSSION ...... 47 3 COMPARATIVE SENSITIVITY OF FINGERPRINTING AND PYROSEQUNCING TO EVALUATE SOIL BACTERIAL COMMUNITIES...... 50 3.1 INTRODUCTION ...... 50 3.2 METHODS ...... 52 3.2.1 Site descriptions and sampling design...... 52 3.2.2 Physical and chemical data collection...... 53 3.2.3 DNA Extraction from soil samples...... 55 3.2.4 Bacterial ARISA...... 55 3.2.5 Roche 454 FLX Titanium pyrosequencing...... 56 3.2.6 Bacterial T-RFLP...... 56 3.2.7 Correlation of biological data between methods...... 57 3.2.8 Correlation with environmental variables...... 57 3.2.9 Modification of matrices...... 58 3.3 RESULTS...... 59 3.3.1 Intra-sample variation...... 59 3.3.2 Recovered OTU Abundances between methods...... 60

iii

3.3.3 Correlation between pyrosequencing and fingerprinting methods...... 61 3.3.4 Differentiating a priori groups...... 65 3.3.5 Environmental predictors of bacterial composition ...... 66 3.3.6 Modified data matrices...... 69 3.3.7 In Silico verses experimental T-RFLP...... 72 3.4 DISCUSSION ...... 72 4 BACTERIAL TARGETS AS POTENTIAL INDICATORS OF DIESEL FUEL TOXICITY ON SUBANTARCTIC SOILS ...... 75 4.1 INTRODUCTION ...... 75 4.2 METHODS ...... 77 4.2.1 Soil sampling and diesel fuel spiking ...... 77 4.2.2 Barcoded amplicon pyrosequencing targeting the 16S rRNA gene...... 79 4.2.3 Multivariate data analysis...... 82 4.2.4 Correlation of environmental variables ...... 82 4.2.5 Dose-response of individual OTUs, genera and phyla across the TPH range ...... 82 4.2.6 Targeting the nitrogen cycle with quantitative PCR...... 83 4.2.7 Dose-response modelling...... 86 4.3 RESULTS...... 86 4.3.1 Bacterial community composition ...... 86 4.3.2 Environmental predictors...... 92 4.3.3 Community indices...... 93 4.3.4 Abundance of functional genes...... 95 4.3.5 Dose response modelling...... 96 4.4 DISCUSSION ...... 100 5 APPYING MICROBIAL INDICATORS TO CONTAMINATED SITES ON MACQUARIE ISLAND ...... 103 5.1 INTRODUCTION ...... 103 5.2 METHODS ...... 105 5.2.1 Soil Sampling ...... 105 5.2.2 Community fingerprinting ...... 114 5.2.3 Pyrosequencing ...... 114 5.2.4 Community indices...... 115 5.2.5 Quantitative PCR ...... 115 5.2.6 Phylogenetic diversity across TPH concentrations...... 116 5.2.7 Significantly inhibited and stimulated genera within sites ...... 116

5.2.8 Environmental correlations to bacterial distribution ...... 116 5.3 RESULTS...... 117 5.3.2 Community fingerprinting profiles and the significance of soil sampling factors ...... 119 5.3.3 Samples selected for pyrosequencing...... 121 5.3.4 Microbial community diversity estimates across sites...... 121 5.3.5 Oligotrophic to copiotrophic ratio across sites ...... 123 5.3.6 Targeted bacterial indices using qPCR ...... 123 5.3.7 Phylogenetic diversity within each site...... 125 5.3.8 Significantly inhibited and stimulated genera across contaminated soils...... 127 5.3.10 The correlation of bacterial communities and soil properties...... 132 5.4 DISCUSSION ...... 135 6 DISCUSSION AND CONCLUSIONS ...... 139 6.1 METHOD SELECTION ...... 139 6.1.1 Novel culturing methods verses culture independent methods ...... 139 6.1.2 Relative sensitivity of community fingerprinting and pyrosequencing technologies . 139 6.2 BACTERIAL DIESEL FUEL TOXICITY IN SUBANTARCTIC SOILS ...... 141 6.2.1 Functional redundancy of microbial communities in polar soils ...... 141 6.2.2 Acidobacteria to Proteobacteria ratio ...... 142 6.2.3 Ammonium oxidation inhibition ...... 143 6.3 RECOMMENDED MICROBIAL INDICATORS AND ASSOCIATED PROTECTIVE CONCENTRATIONS ...... 145 6.4 PHYLOGENETIC DIVERSITY WITHIN SUBANTARCTIC SITES ...... 146 6.5 ENVIRONMENTAL PREDICTORS OF BACTERIAL COMMUNITY COMPOSITION IN POLAR AND SUB POLAR SOILS ...... 148 6.6 EFFICIENCY OF BIOREMEDIATION STRATEGIES ON MACQUARIE ISLAND ...... 149 6.6.1 Evidence of effective bioremediation ...... 149 6.6.2 Limiting factors to effective ongoing bioremediation ...... 150 6.7 CONCLUSIONS ...... 152 6.8 FUTURE RESEARCH ...... 153 7 REFERENCES ...... 155

ABSTRACT

Petroleum hydrocarbon contamination in polar regions occurs commonly as a result of resource extraction, human habitation and subsequent activities. Toxicity information relating to the effects of petroleum hydrocarbon contamination on terrestrial ecosystems in polar and sub-polar regions is limited, yet available evidence suggests that the effects of oil spills are more damaging than in temperate regions due to low temperatures and slower ecosystem recovery. In order to satisfy the obligations of the Protocol on Environmental Protection of the Antarctic treaty, remediation efforts have begun throughout the Antarctic region with a strong focus on bioremediation strategies. However, remediation efforts are currently hindered by a lack of established universal guidelines on remediation triggers and end point concentrations.

The primary aims of this research were to (1) evaluate the utility of the bacterial community as indicators of petroleum hydrocarbon toxicity in polar soils, (2) identify microbial indicators and associated site-specific toxicity thresholds for the subantarctic site Macquarie

Island and (3) apply the identified microbial indicators to soils from chronically contaminated sites undergoing active bioremediation, to evaluate the extent of the bacterial communities recovery.

Following method selection investigations, an acute / sub-acute spiking ecotoxicology experiment was set up across 4 soil types on Macquarie Island. Utilising broad and targeted community indices, as well as functional genes involved in the nitrogen cycle, the bacterial community structure and its functional potential was investigated in response to Special

Antarctic Blend (SAB) diesel fuel contamination. The primary effect of diesel fuel toxicity observed was a reduction in species richness, evenness and phylogenetic diversity. The decline in richness and phylogenetic diversity was linked to disruption of the nitrogen cycle,

vi with species and functional genes involved in nitrification significantly reduced. The bacterial amoA gene, indicative of potential ammonium oxidation was identified as the most suitable indicator of toxicity from the 11 evaluated. Dose-response modelling for this target gene generated an average effective concentration (or protective concentration) responsible for

-1 20% change (EC20) of 155 mg kg .

The identified targets were applied to soils from 3 chronically contaminated sites on

Macquarie Island that have been subject to an in situ bioremediation program since 2009.

The three sites were unique in their soil type, contamination source and in the microbial response to the bioremediation. Highly variable total petroleum hydrocarbons (TPH) concentrations meant that statistically significant reductions were only observed at one of the sites. However, other lines of evidence including increased respiration rates and fuel degradation signatures supported a level of effective biodegradation occuring at the sites.

Despite the highly variable TPH concentrations, the microbial indicators provided evidence of microbial community recovery. In 2 out of 3 sites, the richness, evenness and phylogenetic diversity increased in soils with declining TPH concentrations and at all sites there was a level of recovery in the abundance of the amoA gene. The MPH South, the most recently contaminated site with the lowest organic carbon content, appeared to have a greater disruption to the community, with no recovery of diversity estimates observed in low TPH soils.

Here, in both the spiking experiments and characterisation of chronically contaminated sites undergoing remediation, soils with low levels of organic carbon were found to be more sensitive to disturbance and were slower to recover. This finding is particularly relevant for

Antarctic soils with characteristic nutrient limitations.

vii

ACKNOWLEDGEMENTS

First and foremost, I would like to thank my supervisor Dr Belinda Ferrari who has provided unfailing support and guidance throughout my candidature. I would also like to acknowledge

Professor Ian Snape for his initial persuasion to begin this journey, and the continuing support since. The most rewarding experiences and collaborations have, without exception, been a product of his inspiring personal drive and capacity to connect people, knowledge and resources. I would like to thank all of the extended team involved in the laboratory, field and support roles at the AAD, UNSW and beyond. Without them, this research would not be possible. In particular I would like to thank Tom Mooney, Dan Wilkins, Greg Hince,

Tristrom Winsley, Cath King, Lauren Wise, Ben Raymond, Tim Spedding, James Walworth,

Steve Siciliano, Mark Brown and Andrew Bissett, all of whom have shared their time, expertise and friendship freely. Finally I would like to thank my family for their love and unwavering support in all my journeys and Justin, who has not only graciously encouraged me to experience and enjoy long months away on the amazing Macquarie Island, but has also provided the constant friendship, laughter and love that can make anything possible.

viii

LIST OF PUBLICATIONS

1. van Dorst JM, Siciliano S, Winsley T, Snape I, Ferrari BC. Bacterial targets as Potential

Indicators of Diesel Fuel Toxicity in Subantarctic Soils. Applied and Environmental

Microbiology 2014, DOI:101128/AEM.03939-13

2. van Dorst JM, Bissett A, Palmer AS, Brown M, Snape I, Stark JS, Raymond B, McKinley

J, Ji M, Winsley T, Ferrari BC. Community fingerprinting in a sequencing world. FEMS

Microbiology Ecology 2014, DOI:10.1111/1574-6941.12308

3. Siciliano SD, Palmer AS, Winsley T, Lagerewskij G, Lamb E, Bissett A, Brown MV, van

Dorst JM, Ji M, Ferrari BC, Grogan P, Chu H, Snape I. Soil fertility is associated with fungal and bacterial richness whereas pH is associated with community composition in polar soil microbial communities. Soil biology and Biochemistry 2014, DOI:10:1016/j.soilbio

.2014.07.005

4. Winsley T, Snape I, Mckinlay J, Stark J, van Dorst JM, Ji M, Ferrari BC Siciliano SD.

2014. The Ecological Controls on the Prevalence of Candidate Division TM7 in Polar

Regions. Frontiers in Microbiology 2014, DOI:10.3389/fmicb.2014.00345

5. Winsley TJ, Brown MV, van Dorst JM and Ferrari BC. Capturing greater 16S rDNA sequence diversity within the Bacterial domain. Applied and EnvironmentalMicrobiology.

2012; 78: 5938-5941

6. Ferrari BC, Zhang C and van Dorst JM, 2011. Recovering greater fungal diversity from pristine and diesel fuel contaminated sub-Antarctic soil through cultivation using both a high and low nutrient media approach. Frontiers in Microbiology 2011; 217.

LIST OF SUBMITTED REPORTS AND BOOK CHAPTERS

Wilkins D, van Dorst JM, Rayner J, King C, Snape I, Wasley J, Hince G, Wise L, Mooney

T, Mumford K, Winsley T, Lagerewskij G, Stark S, (2013), Macquarie Island Integrated Risk

Assessment (MIRA), Australian Antarctic Division. Submitted to the Tasmanian State

Government, July, 2013.

Ferrari BC, Winsley TJ, Bergquist PL, van Dorst JM. Flow cytometry in environmental microbiology: a rapid approach for the isolation of single cells for advanced molecular biology analysis. Methods in Molecular Biology. 2012;(881):3-26.

LIST OF FIGURES

Figure 1.1 Abandoned fuel drums in Antarctica...... 2

Figure 1.2 Locations of Antarctic research stations highlighting the prominence of human activity on ice free areas...... 4

Figure 1.3 Bioremediation strategies currently in use in Antarctica and the subantarctic ...... 9

Figure 1.4 King penguins at Gadgets Gully rookery, Macquarie Island ...... 19

Figure 1.5 Macquarie Island research station ...... 21

Figure 2.1 Soil substrate membrane system (SSMS) set up ...... 27

Figure 2.2 The number of unique phyla and genera recovered with culture dependent and culture independent techniques ...... 34

Figure 2.3 Relative abundance of pure cultured phyla across diesel fuel concentration ranges.

...... 36

Figure 2.4 Relative abundance of major phyla ...... 39

Figure 2.5 Relative abundance of major genera within Proteobacteria ...... 40

Figure 2.6 A volcano plot of the genera significantly inhibited or stimulated with increasing diesel fuel concentrations...... 43

Figure 2.7 DRC modelling of community diversity estimates against diesel fuel concentration gradient for SSMS and total soil gDNA...... 46

Figure 3.1 Map of sampling site locations ...... 54

Figure 3.2 Correlation between technical replicates generated with ARISA...... 59

Figure 3.3 Rarefaction curves calculated for each site based on the 454 pyrosequencing, T-

RFLP and ARISA results...... 61

Figure 3.4 Correlation of the OTU Richness and PCO1 co-ordinates between 454 pyrosequencing data and ARISA results, 454 pyrosequencing and T-RFLP results and

ARISA and T-RFLP results...... 63

Figure 3.5 Comparison of the PCO plots generated with the 454 pyrosequencing, ARISA, and

T-RFLP results...... 64

Figure 3.6 Correlation of the environmental parameters with the biological datasets for 454 pyrosequencing, ARISA, and T-RFLP datasets ...... 68

Figure 3.7 Comparison of the PCO pots generated from the reduced 454 pyrosequencing matrices of OTUs contributing >1%, >10% and > 15% of the total abundance ...... 71

Figure 4.1 Ordination plots of samples from P2 based on ARISA fragments and pyrosequencing data...... 80

Figure 4.2 The average measured TPH concentration and bacterial community similarity of samples across SAB spiking fuel concentration ranges and soil plots ...... 87

Figure 4.3 Relative abundances of genera present in soil samples from low to high special

Antarctic blend (SAB) diesel fuel concentrations...... 91

Figure 4.4 The effect of increasing SAB fuel concentrations on richness, diversity, phylogenetic indices and key functional genes within the nitrogen cycle ...... 94

Figure 4.5 Comparison of EC20 values derived from community measures across variable carbon contents ...... 99

Figure 5.1 Location of contaminated sites...... 104

Figure 5.2 Map of MPH...... 107

Figure 5.3 Map of FF ...... 108

Figure 5.4. Soil sampling methods ...... 109

Figure 5.5 Average TPH concentrations per site ...... 118

Figure 5.6 Ordination plots of the bacterial communities based on Automated ribosomal intergenetic spacer analysis (ARISA) profiles and pyrosequencing results...... 120

Figure 5.7 Bacterial community diversity estimates across the TPH range ...... 122

Figure 5.8 Targeted bacterial community indices across the TPH concentration range...... 124

xii

Figure 5.9 The average relative abundances of individual Phyla across TPH concentrations..

...... 126

Figure 5.10 Volcano plots of individual genera significantly inhibited or stimulated with high

TPH at each site ...... 129

Figure 5.11 Correlation of environmental variables to bacterial community composition within each site ...... 133

Figure 5.12 Environmental correlations with the distribution of the bacterial community ... 134

xiii

LIST OF TABLES

Table 1.1 Soil quality clean-up criteria guidelines for petroleum hydrocarbon contamination adopted by circumpolar countries. The units of concentration are mg kg-1 ...... 12

Table 1.2 Key sensitivity data...... 15

Table 2.2 Unique phyla recovered with culture dependent and culture independent methods 33

Table 2.3 Recovered isolates direct from soil to artificial media and SSMS to artificial media culturing methods...... 37

Table 2.4 The most significantly inhibited and stimulated genera present in gDNA recovered from the SSMS enrichments or directly from soil...... 44

Table 2.5 EC20 values generated from dose-response curves and the corresponding dose- response model parameters...... 47

Table 3.1 Calculated diversity estimates for each polar site...... 60

Table 3.2 MANTEL tests results ...... 62

Table 3.3 Distance groups delineated a priori based on sampling distances within sites...... 65

Table 3.4 ANOSIM results testing the ability of each method to distinguish between a priori groups ...... 66

Table 3.5 The consistent environmental variables obtained from the distance based redundancy models ...... 67

Table 3.6 Modified MANTEL tests results ...... 69

Table 3.7 Modified data ANOSIM test results ...... 70

Table 4.1 Analysis of similarity (ANOSIM) results testing for significant differences between fuel categories based on the ARISA and pyrosequencing results...... Error! Bookmark not defined.

Table 4.2 Summary of measured physicochemical parameters...... 81

xiv

Table 4.3 Details of primers used in quantitative PCR (qPCR) to target functional genes within the nitrogen cycle...... 84

Table 4.4 ANOVA and ANOSIM results testing the amplicon pyrosequencing community

UniFrac and bray-curtis dissimilarities between fuel categories and soil plots...... 88

Table 4.5 Operational Taxonomic Units (OTUs) significantly (p <0.05) stimulated and inhibited across the SAB diesel fuel spiking range...... 89

Table 4.6 Distance Linear Model (DistLM) results indicating strength of environmental variable as a predictor of the biological distribution and patterns of bacterial communities across all samples...... 92

Table 4.7 Generated Effective Concentration values (EC20) across a range of community indices...... 97

Table 5.1 Physical and chemical parameters of soil samples ...... 110

Table 5.2 Soil sampling at each contaminated site ...... 11512

Table 5.3 qPCR reaction efficiencies and standard curve properties ...... 119

Table 5.4 ANOSIM results for sampling factors ...... 1107

Table 5.5 Summary of inhibited and stimulated genera at each contaminated site...... 1286

Table 5.6 The 10 most significantly inhibited and stimulated genera at each site...... 1308

LIST OF ABBREVIATIONS

454 454 tag pyrosequencing platform

AAD Australian Antarctic Division

ANARE Australian National Antarctic Research Expeditions

ANOSIM Analysis of Similarity

ARISA Automated ribosomal intergenic spacer analysis

C_AFH Canada, Alexandra Fjord Highlands

DistLM Distance-based linear model

DL Minimum detection limit

DRC Dose-response Curves

EC Effective concentration

FF Fuel Farm gDNA genomic DNA

LC Lethal concentration

NGS Next Generation Sequencing

N_SS Norway – Longyearbyen Slijeringa

N_SV Norway – Longyearbyen Vestpynten

MPH Main Power House

OTU Operational Taxonomic Unit

xvi

PAH Polycyclic aromatic hydrocarbons

PCO Principal Component Ordination

PCR Polymerase Chain Reaction qPCR Quantitative Polymerase Chain Reaction

RFLP Restriction fragment length polymorphism

RT Robinsons Transect

SAB Special Antarctic Blend

SD Standard Deviation

SE Standard Error

SSMS Soil Substrate Membrane System

T-RFLP Terminal restriction fragment length polymorphism

UCM Unresolved Complex Mixtures

UNESCO United Nations Educational Scientific and Cultural Organisation

xvii

1 INTRODUCTION

1.1 HUMAN IMPACTS IN THE ANTARCTIC REGION

Petroleum hydrocarbons and, in particular, diesel fuel has been identified as the most extensive and environmentally damaging of all the contaminants in Antarctica (1-3).

Inadequate or non-existent records on previous petroleum hydrocarbon spills means the true extent of petroleum hydrocarbon contamination is unknown but estimates suggest between 1-

10 million m3 of soil may be contaminated (1) (Figure 1.1). The necessity of fuel for all activities in Antarctica including transport, heating and energy generation has led to the widespread legacy of the contaminant.

Humanity’s fascination and interaction with the Antarctic region began in the 18th century with sealers and the early Antarctic explorers. The harvesting of seals, whales and penguins by sealers was a particularly exploitative period leading to the local demise of many species and the near-extinction globally for others (2). Since that time human presence in Antarctica, regardless of intention, has continued to have an impact on the environment through the necessity of fossil fuels for survival, interaction with flora and fauna, the introduction of foreign species (4, 5) and the generation of chemical and biological wastes (2, 3). Although commonly perceived as one of humanity’s last pristine frontiers, the rapidly expanding numbers of tourists and national research programs, fishing exploits throughout the Southern

Ocean, and the influence of anthropogenic-driven global processors such as global change and ozone depletion all continue to contribute to major, persistent and in some cases permanent disruption of Antarctic ecosystems (2, 3).

Figure 1.1 Abandoned fuel drums in Antarctica (photo by Ian Snape).

1.2 PERSISTENCE OF PETROLEUM HYDROCARBONS IN

TERRESTRIAL ANTARCTIC AND SUBANTARCTIC

ECOSYSTEMS

When petroleum hydrocarbons are released into the environment, research has found that apart from high-energy marine environments where disposal is rapid, contaminants have an extended longevity (6-9). For example, fuel spills from the Bunger Hills in Antarctica were found to be essentially unchanged more than a decade after the spill events (7) and aromatic compounds were detected in the soil at the Scott Base and Marble Point in Antarctica, more than 30 years after any known contamination events (6). Petroleum hydrocarbon spills near

Casey station, Antarctica, were found to have fuel half-life of decades, despite substantial levels of evaporation and weathering immediately following the spill event (8, 10). At

Macquarie Island, the persistence of high fuel concentrations and relatively unweathered fuel

2 signatures led investigators to estimate a typical spill half-life it the order of tens to hundreds of years (9).

Once petroleum hydrocarbons reach the environment, impacts have been reported across several tropic levels, ranging from bacteria to vertebrates (11-14). Marine sediments contaminated with petroleum hydrocarbons and heavy metals originating from the terrestrial environment, have greater abundances of polychaetes and gastropods compared to uncontaminated sediments (14), and higher occurrences of pathological anomalies have been reported in fish (12). In soils, the occurrence of petroleum hydrocarbons selectively stimulates hydrocarbon degrading species, altering microbial community structure and decreasing overall diversity (15, 16). Indirect effects on the soil systems have also been reported, with a reduction of soil quality through changes in pH, interference with water retention and movement (17), restriction of oxygen movement (9), reduced availability of nitrogen and phosphorus, and interference with geochemical nutrient cycles (15, 18, 19) .

Less than 1% of the Antarctic continent is ice free, yet it is within this limited ice free terrestrial environments where the majority of wildlife, human activity and subsequent contamination events are concentrated (2, 3) (Figure 1.2). Antarctic soils share characteristic nutrient limitations and low moisture which in turn support limited invertebrate and microbial biodiversity (20). It is the relative simplicity of Antarctic soil ecosystems that make them ideal ecological archives and at the same time particularly vulnerable to perturbations (2).

The rarity of soil ecosystems further exacerbates the significance of the direct and indirect impacts of petroleum hydrocarbons on soil systems.

Figure 1.2 Locations of Antarctic research stations highlighting the

prominence of human activity on ice free areas.

Soils from the Antarctic Peninsula and subantarctic regions are more developed than the soils from the continent (3). They have higher organic matter content and are comprised of more complex systems with the presence of flora and a greater numbers of macro and micro invertebrates. Nutrient limitation still prevails, but is usually site and vegetation dependent.

Although representing relatively more diverse and complex systems, the low temperatures, low nutrients and extreme conditions continue to have a limiting impact on turnover rates and metabolic activity of the soils in comparison to temperate regions (9, 21, 22). As a result, the bio-attenuation rates of petroleum hydrocarbons remain significantly slower and more damaging than in temperate regions.

1.3 RESPONSE OF SOIL MICROBIAL COMMUNITIES TO

PETROLEUM HYDROCARBONS

In the species-poor Antarctic environments, microorganisms constitute the bulk of the biomass and underpin the majority of nutrient cycling (20). Biodegradation via indigenous microorganisms is also the primary mechanism responsible for residual hydrocarbon removal in the environment (23). In the presence of petroleum hydrocarbons, culture dependent and culture independent methods have identified a loss of microbial diversity and enrichment of culturable heterotrophs and hydrocarbon degrading species (6, 15, 16, 22, 24, 25). Many of the hydrocarbon degrading bacterial species isolated from polar soils are consistent with temperate soils and are dominated by members within Proteobacteria and Actinobacteria

(15, 16, 26-30). Within Proteobacteria, hydrocarbon degrading species most commonly belong to the Pseudomonas or Sphingomonas genera, whilst within Actinobacteria, species capable of hydrocarbon degrading most commonly belong to the genus Rhodococcus. Despite the consistency of associated genera there remains a level of divergence between the isolated strains and their potential metabolic activity at different sites.

Pseudomonas has been isolated from contaminated polar soils from the Arctic, Antarctic and subantarctic. In the Arctic, Pseudomonas strains B17 and B18 were isolated from petroleum- contaminated soil and were capable of utilising C5-C12 n-alkanes and the aromatic compounds naphthalene and toluene (26, 27). In contrast, the Pseudomonas strains isolated from Arctic tundra near the northern coast of Ellesmere Island (28) were capable of utilising either only aromatics in the form of toluene and benzene (Pseudomonas IpA-92 and IpA-93) or the alkanes of octane and dodecane (Pseudomonas DhA-91). In Antarctica near Scott Base and Marble point, Pseudomonas strains Ant 5, Ant 9 and 7/22 were isolated from soils chronically contaminated with a mixture of oils (31). All of the strains were capable of

5 degrading JP8 jet fuel and other aromatic hydrocarbon compounds, yet the specificity of potential substrates varied between strains with only Ant 5 and Ant 9 capable of utilising naphthalene and 2-methylnapthalene and Ant 9 and 7/22 were capable of utilising xylene, and

1,2,4-trimethlybenzene.

Sphingomonas strains have been isolated from many of the same contaminated sites throughout the polar soils as Pseudomonas (26, 31, 32). Unlike Pseudomonas, the individual strains of Sphingomonas isolated and tested appear to utilise only aromatic compounds for growth, including m-xlene, naphthalene, fluorene and phenanthrene, along with the aromatic fractions of jet fuel and diesel fuel. Conversely, the hydrocarbon degrading Rhodococcus isolated from contaminated polar soils appear to utilise alkanes exclusively, despite evidence of aromatic degradation from temperate regions. The compounds readily utilised by

Rhodococcus include but are not limited to; JP8 jet fuel (16, 29, 30) , C6-C20 n-alkanes (29), pristine (16, 29), dodecane and hexadecane (16, 30).

To date, the majority of studies assessing the impact of petroleum hydrocarbons on polar soil microbial populations have focused on the stimulated portion of the bacterial community, primarily through culturing and Sanger sequencing of clone libraries (15, 16, 33-35). The aim of these studies has been to establish the natural attenuation capacity of a soil, or the effects of active bioremediation in these cold environments. By comparison, little information exists on the species sensitive to petroleum hydrocarbon associated toxicity. Of those studies available, assessments are confined to microbial processes such as denitrification (36) or methanogenesis (37) with little to no information on the species or mechanisms involved.

1.4 FUNCTIONAL GENES ASSOCIATED WITH HYDROCARBON

DEGRADATION

Hydrocarbon degrading genes are routinely found in both clean and contaminated soils (38-

40), but are in greater abundance at contaminated sites, decreasing with declining levels of hydrocarbons (41). Genes encoding for the alkane hydroxylase system have been used as proxies for alkane degradation with homologues of the alkane monooxygenase typically sourced from Psuedomonas putida (Pp alkB) or Rhodococcus spp (Rh alkB1 and Rh alkB2)

(38, 41, 42). Aromatic degradation potential is similarly investigated with the ndo and nahAC genes encoding naphthalene dioxygenases, the xylE and nahH genes encoding 2,3-catchol dioxygenases, and phnAc gene encoding aromatic dioxygenase (38-40, 43). The functional genes nirK, nirS and nosZ encoding for enzymes within the denitrification pathways have also been linked to hydrocarbon degradation. Numbers of nosZ in particular have reported to increase with hydrocarbon contamination and bioremediation strategies. The inflated denitrifying population is thought to be indicative of anaerobic hydrocarbon degradation (44).

1.5 BIOREMEDIATION OF PETROLEUM HYDROCARBONS IN

POLAR SOILS

Due to the exceptionally high costs and the disruption to the environment of dig and haul remediation options, on-site bioremediation through optimising the activity of indigenous organisms has been recommended as the most appropriate remediation strategy in polar soils

(15, 45). The slow attenuation rates and persistence of hydrocarbons in the soil, despite the presence and proven activity of hydrocarbon degrading species, has motivated a number of bio-remediation trials and related research across the Arctic (43, 46, 47), Antarctic (1, 26, 41) and subantarctic regions (9, 21, 48-50) . The cold temperatures, nutrient limitations and

7 availability of liquid water are all thought to contribute to the slow attenuation rates of contaminants. As such, the majority of bio-remediation strategies above have thus far focused on variations of in-situ and ex-situ nutrient amendments, with some trials combining additional aeration (9) or heating (51, 52).

In 2009 at Macquarie Island, a micro-bioventing system was installed in-situ that delivers aeration and nutrient addition throughout the soil matrix through sparge points arranged in a grid, approximately 1m apart (9, 48, 53) (Figure 1.3). The nutrient addition is usually limited to 2-3 times over the Antarctic summer, while the aeration is controlled remotely in a switching system throughout the year. At Casey station in Antarctica, approximately 600 m3 of soil contaminated by Special Antarctic Blend (SAB) diesel, ATK (aviation fuel) and other hydrocarbons (10) was excavated to construct 6 biopiles. The ex situ bioremediation strategy includes nutrient addition and aeration systems to stimulate hydrocarbon degradation (Figure

1.3).

The type of nutrient addition in bioremediation programs has varied from fish compost (50), slow release fertilizer (54), to commercially available Inipol EAP-22 (55). The amount of nutrient amendments is also far from standardised. However, evidence has shown that nitrogen concentration should be calculated as a function of soil water, rather than as a function of dry soil weight (48). Insufficient soil water can lead to excessive nitrogen concentrations in the soil which are inhibitory to degradation rates. Overall, nutrient addition irrelevant of type has consistently improved degradation rates in polar soils, yet the improvement has been found to be inconsistent across hydrocarbon classes (26). The greatest improvement in degradation rates has been observed in alkanes and aliphatics, with small to negligible improvements of degradation rates for heavy polycyclic aromatic hydrocarbons

(PAHs) (50). Extended bioremediation trials in polar soils utilising in situ (21, 46) and ex situ

8 methods still exhibit microbial toxicity and a presence of heavy PAHs (21, 46, 50). More research is needed to evaluate the identity and toxicity of the residual compounds and the best strategies for optimising their degradation. The lack of clean up protocols or consensus on remediation guidelines has also been highlighted as a major hindrance to the planning and execution of bioremediation programs (2, 26).

Figure 1.3 Bioremediation strategies currently in use in Antarctica

and the subantarctic. (A) Nutrient amended and aerated biopiles at Casey

Station. (B) Installation of in-situ aeration system at Macquarie Island (C)

Installation of in-situ spikes for aeration and nutrient addition at Macquarie

Island.

1.6 EXISTING REGULATIONS IN ANTARCTICA

International regulation of the impacts of human activities in Antarctica began in 1948 with the International Convention for the Regulation of Whaling (signed 1946) followed relatively closely by the Antarctic treaty (signed in 1959, coming into force in 1961). The Antarctic treaty originally prohibited the military activities and nuclear testing in the area south of 60

°S with the Treaty parties further obligated to preserve and conserve the living resources in

Antarctica. Since 1961, the Antarctic treaty has been expanded to include the conservation of

Flora and Fauna (signed 1964, coming into force 1982), the convention for the conservation of Antarctic seals (signed 1972, coming into force 1978) and the more recent Protocol on

Environmental Protection to the Antarctic Treaty (1991). Management of fisheries was additionally initiated by the convention for the conservation of Antarctic Marine Living

Resources (CCAMLR) (signed 1980, coming into force 1982). It is the Protocol on

Environmental Protection that prohibits mineral resource activities and further stipulates the protection of the Antarctic environment;

“The protection of the Antarctic environment and dependent and associated ecosystems and intrinsic value of Antarctica, including its wilderness and aesthetic values and its value as an area for the conduct of research, in particular research essential to understanding the global environment, shall be fundamental considerations in the planning of all activities in the

Antarctic Treaty area”

Importantly the Protocol on Environmental Protection also stipulates that past and present waste disposal sites on land and abandoned work sites of Antarctic activities shall be cleaned up by the generator(s) of such wastes and the user(s) of such sites. However, no universally accepted standards for remediation currently exist in Antarctica and there is no authority

10 governing the rehabilitation of sites. The Annex VI to the Environmental protocol regarding

Liability Arising from Environmental Emergencies signed in 2005 is also yet to be ratified.

1.7 GUIDELINES SPECIFIC TO PETROLEUM HYDROCARBON

CONTAMINATION IN POLAR SOILS

Unlike Antarctica, a number of soil quality guidelines currently exist throughout the Arctic, yet the derivation of the guidelines differs between circumpolar countries with the resulting values varying enormously (Table 1.1). For example, the clean – up criteria for diesel fuel in

Norway is 100 mg kg-1, compared to 2000 mg kg-1 in Canada. Without available toxicology data for local Arctic or Antarctic species, there is no evidence to predict which targets most reasonably represent risk for the Arctic or Antarctic regions. A more sophisticated understanding of the polar terrestrial systems is crucial in moving towards universally accepted guidelines, or protocols to establish site-specific guidelines.

Table 1.1 Soil quality clean-up criteria guidelines for petroleum

hydrocarbon contamination adopted by circumpolar countries. The units

of concentration are mg kg-1. Table adapted from Snape et al. (2008).

Product Finland1 Norway2 Sweden3 Canada4 Alaska5 Gasoline 100 7 5-100 1000 100 Diesel 200 100 100 2000 200 Crude oil 100 100 2500 200 Residual 600 100 100 2500 2000 Polycyclic Aromatic Hydrocarbon Analysis Anthracene 50 2.5a (32c) 3800 Benzo(a)anthracene 40 0.4 6.2ai 6.6 Benzo(b)fluoranthene 0.7 6.2 ai 6.6 Benzo(k)fluoranthene 40 0.4 6.2 ai 66 Benzo(a)pyrene 40 0.1 0.4 20 (72) 0.66 Chrysene 40 0.5 6.2 ai 660 Dibenzo(a,h)anthracene 20 0.23d Fluorene 20 0.6 50a (180c) 2500 Indenol(1,2,3-c,d)pyrene 40 0.4 2.7 d 6.6 Naphthalene 100 0.8 8.8 ai 1900 Pyrene 40 0.1 0.6 7.7 ai 1900 Petroleum Hydrocarbon Range ac C6-C10 (volatile) 500 7 100 30-230 1400 ac C20-C25 (light extractable) 2000 100 100 150-2200 12500 ac C25-C38 (heavy extractable) 2000 100 100 400-5000 13700 1 Soil quality guidelines from Finnish Minister for the Environment 2 Guidelines from Norway Pollution Control Authority 3 Soil quality guidelines from Swedish Environmental Protection Agency 4 Soil quality guidelines for PAH analytes from Canadian Council of Ministers of the Environment (CCME, 2010) 5 Soil clean-up levels from the Alaska Department of Environmental Conservation (2008) a Agriculture/residential land use c Commercial/industrial land use d Protection of potable drinking water i Soil and food ingestion ac Range reflects criteria for four land-use categories (agriculture, residential, commercial and industrial

1.8 ECOTOXICOLOGY IN POLAR SOILS

Established environmental quality standards in temperate regions are primarily based on concentration-response data obtained from single species toxicity bioassays. Bioassays are designed to identify concentrations of single or mixed contaminants that led to significant negative effects on organisms (ECx), or no significant effects in the case of NOEC, NEC

(56). Traditional toxicity organisms used in bioassays are selected on the basis of sensitivity and survival in a laboratory setting (57), and can have a variety of endpoints including survival, growth, and reproduction (56). Established guidelines exist surrounding the generation of toxicity tests suitable for use in environmental quality standards, for example the Australian and New Zealand water guidelines propose use of a minimum of 8 species from at least 4 taxonomic groups for use in sensitivity distribution models, with more weight given to higher tropic levels and chronic tests over acute tests (58).

The validity of using single “model” species to represent sensitivities of entire ecosystems has been questioned (59, 60) with some reports suggesting even a battery of mono-specific assays are not capable of predicting the effects of a toxic substance on natural ecosystems

(56). In Antarctica, the majority of terrestrial ice free areas are characterised as cold desserts

(3). The limited biomass consisting of microorganisms, lichens, mosses and relatively few invertebrates (springtails, mites, nematodes, tardigrades and rotifers), are generally slow growing, in low abundance, and have limited diversity (20). Key traditional indicator species, in contrast, are generally fast-growing temperate species that respond well to ecotoxicology tests in the laboratory (60). Very few traditional indicator species are present in cold regions and with those present possessing very different life strategies, the uncertainty surrounding their capacity to accurately represent toxicity in cold region ecosystems is further

13 compounded. Evidence from studies on the Antarctic marine organisms, Sterechinus neumayeri (Antarctic sea urchin) and Phaeocystis antarctica (microalgae), show significantly higher levels of sensitivity to metal contaminants and copper (61) in comparison to similar temperate species.

Considerable effort has been made in recent years to remedy the lack of relevant toxicity data for petroleum hydrocarbons in polar soils. Emphasis has been placed on utilising indigenous invertebrates, such as earthworms and mites (62), microbial process based assays (17, 63), and the development of environment-specific tests to accommodate species with slow metabolisms and extended life stages (61). The greatest progress has been made at Macquarie

Island, with sensitivity data obtained for the indigenous earthworm Microscolex macquariensis, invertebrate community composition (62), fungal diversity (64), and a range of microbial process based assays evaluated (17) (Table 1.2). Process based toxicology tests have targeted key processes within the nitrogen cycle and found that particular pathways are inhibited (nitrification) while others are stimulated (denitrification). These investigations have gone a long way in identifying potential microbial indicators suitable for polar soils and potential functional impacts on the soil ecosystem (17, 63). However, these toxicity assays remain black box investigations, with no information on the species or mechanisms involved.

Table 1.2 Key sensitivity data including effective concentrations, ECx

(concentrations that causes a response on a certain % of test organisms)

and lethal concentrations, LCx (concentrations causing x% mortality of

the population).

Fuel concentration (mg kg-1) Toxicity Test Response Antarctic Sub- Reference Antarctic Nitrification (PNA1) in soil EC25 1000 Schaefer 2009 Bacterial biomass EC25 1700 Schaefer 2009 Microbial community EC25 670 Schaefer 2009 composition Total soil respiration2 EC25 2400 Schaefer 2009 Free phase fuel (sheen) Presence 50 Rayner et al. 2007 Nitrification (PNA1) in soil EC20 190 Schafer 2007 Denitrification in soil EC20 950 Schafer 2007 Total soil respiration2 EC20 220 Schafer 2007 Microbial carbohydrate EC20 16 Schafer 2007 utilisation Fungal diversity Reduced 250-500 Ferrari et al. 2001 Earthworm3 avoidance Behaviour 181 Mooney 2013 Earthworm3 reproduction4 EC10 95 Mooney 2013 Earthworm3 survival LC50 103 Mooney 2013 Invertebrate community EC200 285 Mooney 2014 composition5 1 Potential nitrification activity 2 Respiration caused by sucrose 3 Test species is Microscolex macquariensis 4 Measured as juvenile production and survival 5 Based on collembolan (springtail) assemblage

1.9 TOWARDS DEVELOPMENT OF MICROBIAL INDICATORS

Microbial populations have been proposed as ideal indicators of polar soil health (17, 65, 66).

They constitute the bulk of the biomass in polar soils and as such are the primary providers of ecosystem services, such as decomposition, mineralisation and inorganic nutrient turnover 15 and pollutant removal (67). They respond to changes in the physical and chemical environment and in turn influence the physical and chemical environment underpinning ecosystem function through biogeochemical cycling. It is this dynamic feedback system between responding to their environment and subsequent influence on it that make microorganisms ideal, albeit complex, bio-indicators of ecosystem health.

A range of methods have been successfully utilised to link perturbations to shifts in microbial structure and function, including, community fingerprinting with automated ribosomal intergenetic spacer analysis (ARISA); terminal restriction fragment length polymorphism (T-

RFLP) or denaturing gradient gel electrophoresis (DGGE) (68-70); community structural diversity with phospholipid fatty acid (PLFA) profiles (71, 72); functional gene analysis with quantitative PCR (36, 73); and combinations of microbial biomass, enzymatic assays, and community profiling and / or sequencing (17, 63, 74-78).

The structure of microbial communities is consistently observed to be altered in response to perturbations, yet the impact on overall microbial diversity is inconsistent between investigations depending on the type of disturbance and the methods used. For example, microbial diversity has been reported to decline in response to heavy metals (71, 79), have no impact in an agricultural land use and recovering metal contamination stream investigations

(78, 80), or recover quickly after acute exposure to diesel fuels (76). The site and contaminant specificity highlights the need for understanding microbial responses to contaminants in light of environmental and population interactions in order to be able to predict and therefore protect microbial communities and the ecosystems they support (67).

One of the main challenges for developing microbial indicators is the high variability of microbial communities observed naturally across temporal and spatial scales. In Antarctica,

16 previous studies have linked the distribution of microbial diversity to variations in geological and environmental factors including pH, moisture, location, elevation, proximity to the ocean and proximity to wildlife breeding sites (81-85). Similarly in temperate regions, environmental variables have been linked to microbial distribution with pH identified as a major driver (86, 87). Recent investigations have further reported that environmental properties of terrestrial environments such as vegetation cover, soil organic carbon, pH, and nutrient availability not only had direct and indirect effects on the microbial community but also on the community’s response to perturbation (74, 78, 79).

For successful integration of microbial indicators into environmental assessments or standards, there is a need to characterise the microbial response to contaminants beyond community ‘shifts’. Concentration dose-responses to toxicity or negative effects are an established method in identifying potentially harmful concentrations in the environment and can be applied to microbial responses. In Schafer et al. 2007, 2009, the potential nitrification, potential denitrification and carbohydrate utilisation toxicity assays were performed across an increasing contaminant range in order to evaluate the dose-response effect of diesel fuel on subantarctic and Antarctic soils. The responses were successfully fitted to the most appropriate dose-response curves on order to calculate corresponding EC20 and EC25

(concentrations that result in a 20 or 25 % effect of a measured response) (17, 63). Dose- response modelling can provide a means of comparing sensitivities to contaminants between sites and also provides an opportunity to integrate microbial toxicity information into established risk assessment frameworks.

1.10 RESEARCH SITES

Despite differences, the Arctic and the Antarctic located at opposite poles share more similarities with each other than any other regions in the world (88). Both polar regions are dominated by significant ice coverage and share characteristic cold, dry climates with long dark winters and brief light summers. They also share paleoclimate and oceanography systems of significant importance. The Arctic and Antarctica are both subject to ongoing contamination threats and historical contamination legacies (54). Although not completely analogous, it is thought that progress made at individual contaminated polar sites will have the potential to inform and benefit other contaminated sites throughout the polar regions (54).

While the following thesis utilises samples from the high Arctic and East Antarctic, there is a focus on guidelines for the Antarctic region. The primary field site is located at the subantarctic Australian territory, Macquarie Island.

1.10.1 Macquarie Island

Macquarie Island is a remote Australian subantarctic territory measuring 23 km long and 5 km wide, located approximately 1500 km South of Tasmania, Australia. Dominated by its oceanic position, the temperatures vary little between seasons with minimum average temperatures ranging from 3 to 8 °C and cold, wet and windy conditions are persistent throughout the year. As an extremely remote land mass, the island provides a sanctuary for migratory animals of the Southern Ocean, with thousands of seals and millions of seabirds coming ashore each year to breed and moult. Species utilising the island include the

‘vulnerable’ New Zealand Fur Seal, the Black Browed and Grey-headed Albatrosses and the

‘critically endangered’ Wandering Albatross (Australian Environmental Protection and

Biodiversity (EPBC) Act 1999).

After a tumultuous history of exploitation by sealers, the island was declared a wildlife sanctuary in 1933 and in 1972 it was made a State Reserve under the Tasmanian National

Parks and Wildlife Act. In 1977 it was declared a Biosphere reserve under the UNESCO

Biosphere program, and in 1997 the island was declared a World Heritage Site due to its unique ocean floor geography and its natural beauty. In 1999 the Australian Government created the Macquarie Island Marine Park, with highly protected zones extending 3 nautical miles from the island. Finally, in 2002 the Macquarie Island Nature reserve was listed on the register of Critical Habitat under the EPBC Act, as habitat critical to the survival of the Grey- headed and Wandering Albatross. It was in 1948 that the Australian Antarctic Division

(AAD) established a National Antarctic Research Expeditions (ANARE) research station on the Island and the research station has been permanently occupied since.

Figure 1.4 King penguins at Gadgets Gully rookery, Macquarie Island.

As a result of the continual human occupation, contaminants are now present in the air, sea and soil. Petroleum hydrocarbons have been identified as one of the most prolific of the contaminants on the island with three major petroleum hydrocarbon contaminated sites

19 identified surrounding the fuel storage and power generation facilities (9, 89). The most recent of these sites originated from a fuel spill in 2002. The AAD has an obligation to reduce the impact of hydrocarbon contamination on the Macquarie Island environment under the

Island’s Environmental Management and Pollution Control Act 1994, and to satisfy World

Heritage, Biosphere and Critical Habitat values, and a further obligation to satisfy its own

Environmental Policy to:

“demonstrate leadership in environmental management across all its activities in

Australia and Antarctica” (AAD 2011)

Since the fuel spill in 2002, considerable effort has been made to remediate the recent and historical contaminated sites. After feasibility and optimisation trials (9, 48, 53) a bioremediation program began in 2009, employing nutrient addition and bioventing to stimulate the petroleum hydrocarbon degradation potential of the indigenous microbial population. The bioremediation program is ongoing but as yet, no guidelines or remediation endpoints have been set by the regulatory authority, the Tasmanian Environmental Protection

Agency (TEPA).

Figure 1.5 Macquarie Island research station located on the Northern

Isthmus

1.11 AIMS

Microorganisms have a promising potential as ecosystem indicators, particularly in polar soils where traditional toxicity species are rare. With developments in sequencing technologies, the phylogenetic and functional understanding of environmental microbial communities is rapidly expanding, and in turn promoting the development of molecular microbial indicators as integrated measures of ecosystem health (90). The primary aims of this research were to (1) evaluate the utility of the microbial community as indicators of petroleum hydrocarbon toxicity in polar soils, (2) identify microbial indicators and associated site specific toxicity thresholds for the subantarctic site Macquarie Island and (3) apply the identified microbial indicators to soils from chronically contaminated sites undergoing active bioremediation in order to evaluate the extent of the bacterial communities recovery.

2 NOVEL CULTURING TECHNIQUES AND THE

MICROBIAL RESPONSE TO DIESEL FUEL IN A

SUBANTARCTIC SOIL.

2.1 INTRODUCTION

The enormous disparity between the diversity recovered with culture dependent and culture independent techniques was highlighted in 1995, with estimates of less than 1% of the known bacterial diversity considered culturable (91). The limited capacity of culturing techniques to reflect the depth and breadth of microbial diversity hinders our understanding of metabolic functions and microbial interactions responsible for ecosystem services (92, 93). In contaminated soils understanding such interactions has the potential to inform and enhance bioremediation efforts. Furthermore, it is difficult to gauge the impact of toxicity to microbial populations, without understanding the full extent of phylogenetic and metabolic diversity.

Recent advances and decreasing costs of molecular techniques have rapidly expanded the accessibility of high throughput sequencing technologies (94). Pyrosequencing targeting the

16S rRNA gene can now provide rapid and comprehensive taxonomic information within uncultured complex environments such as marine, soils and sediments (95-97). Rapid sequencing through the use of next generation sequencing (NGS) platforms has already been utilised to evaluate the impact of petroleum hydrocarbons on bacterial community diversity in mangroves (98), oil fields (99) and in the Gulf of Mexico after the Deepwater Horizon oil spill (100). However, understanding the functional relevance of phylogenetic shifts remains problematic without cultured representatives.

The uncultured majority are thought to have oligotrophic growth strategies (101) and grow predominantly as micro-colonies (102-104). By mimicking the oligotrophic conditions found in a specific environment, novel micro-culturing techniques have begun to bridge the gap between the cultured and the uncultured majority (105-110). Novel culturing attempts have included limiting nutrient availability, increasing incubation times and providing contact with the source environment (102, 103, 109). Many of these novel approaches were designed to mimic aquatic environments and have resulted in the successful growth of bacteria from many divisions including those with no previous cultured representative including SAR11

(106) and OP10 (102, 105). The soil substrate membrane system (SSMS) is a novel micro- culturing technique targeted at terrestrial systems (111). The technique attempts to mimic in- situ growth conditions of the soil environment by using environmental soil as the growth substrate, replacing the excessive nutrients characteristic of artificial media and has been successful in substantially increasing the culturability of bacteria from soil (107).

In the following chapter we utilised the SSMS to recover isolates from the environment and to determine the proportion of the bacterial soil community that could be enriched in micro- colonies. Our aims were (1) to test the capacity of SSMS to characterise the microbial populations within subantarctic soils against direct culturing with artificial media and (2) against a culture independent method and finally, (3) to evaluate the capacity of SSMS to identify subantarctic microbial toxicity thresholds.

2.2 METHODS

2.2.1 Site and sampling methods

At Macquarie Island, (54°37’53”S, 158°52’15”E) a soil sample (500 g) was collected from the top 30 cm of an uncontaminated plot approximately 170 m away from the largest

23 contaminated site at the Main Power House (MPH) (9). The soil was sub-sampled and spiked with SAB diesel fuel to nominal fuel concentrations from 0-20 000 mg kg-1 (Table 2.1). The soil samples were homogenised and stored for 3 months at ambient temperature (10 °C). To determine the measured TPH concentrations, sub-samples (10 g) from each spiked soil sample were extracted with hexane and assessed by gas chromatography at the AAD, according to the method described by Schafer et al. (2007). Spiked soils were grouped according to their final measured concentrations into the low (DL-400 mg kg-1), medium

(400-5000 mg kg-1), and high (>5000 mg kg-1) diesel fuel concentration ranges (Table 2.1).

After the 3 month incubation, soil samples were sieved aseptically through a 2 mm metal sieve to remove large particulate matter in preparation for cultivation and micro-cultivation.

Table 2.1 Nominal, measured and fuel concentration ranges for spiked

soils.

Nominal Measured Diesel fuel TPH TPH concentration mg kg-1 mg kg -1 ranges 0

2.2.2 Cultivation on low nutrient artificial media

To create the bacterial inoculums a sub-sample of each sieved soil was diluted 1:200 in filtered distilled H2O, vortexed for 30 s and allowed to settle. Serial dilutions of the inoculum

-1 -2 -3 were created with filtered distilled H2O at 10 , 10 and 10 . An aliquot (100 µl) of each

24 serial dilution was spread onto low nutrient artificial media (0.1 x RAVAN) plates in triplicate. RAVAN media was prepared from the stock solution (10X); 5 g glucose, 5 g peptone, 5 g yeast extract, 5 g sodium acetate, 5 g tri-sodium citrate, 2 g pyruvic acid and made up to 1 litre with milli-Q water (101, 111). Plates were sealed with parafilm to prevent moisture loss and then incubated aerobically and anaerobically at 10°C to maximise potential recovery of isolates. For anaerobic conditions the plates were placed in an anaerobic chamber

(Becton Dickinson, North Ryde, Australia) with anaerobic gas packs (BD). The resulting mixed colonies were sub-cultured (at least 3X) until pure isolates were obtained, with particular attention to unique morphologies.

2.2.3 Micro-cultivation with SSMS

For novel micro-cultivation, 50 µl of the 1:200 inoculums from above (2.2.22), was added to

10 ml of distilled H2O, then filtered onto a 0.2 µm, white, isopore PC membrane (Millipore,

Australia) with a vacuum pump filtration manifold as described in Ferrari et al. (2008). The inoculated PC membranes were carefully transferred to an inverted 25 mm tissue culture insert (TCI) (Millipore, USA) and placed on top of the fixed 0.02 µm sterile membrane

(Figure 2.1). Before the PC membranes were added, 3 grams of soil was added into the inverted TCI with approximately 750 µl of pre-filtered distilled H2O (Figure 2.1A). The H2O was added slowly to the soil and gently vortexed to create a moist soil slurry texture that could maintain contact with the underside of the fixed TCI 0.02 µm membrane. The top of the sterile TCI membrane served as a barrier between the non-sterile soil slurry and the underside of the sterile PC membrane (Figure 2.1B). The end result permits transfer of compounds less than 0.02 µm only, between the bacterial inoculums and the non-sterile soil slurry substrate.

To maximise the potential recovery of bacterial diversity from the environment, six replicate

SSMS were set up in 6 well plates for each soil sample (Figure 2.1C): three for aerobic and three for anaerobic incubation. To maximise growth rates whilst remaining within a relevant temperature range 10°C was selected for incubation. The optimal incubation time of 16 days was determined by evaluating growth membranes for micro-colonies, without overcrowding, as determined through epi-fluorescent microscopy (EFM) as described below (2.2.4). For aerobic conditions, the final culture vessels (6 well plates) were hydrated and sealed to prevent moisture loss during incubation (111). For anaerobic conditions, the vessels were hydrated and placed in an anaerobic chamber with anaerobic gas packs as described above.

After the 16 days of incubation the growth membrane was removed and checked for growth confirmed with EFM. Once growth was confirmed the remaining membrane was cut into quarters for DNA extraction and secondary culturing with artificial media. Despite the separate culturing conditions, aerobic and anaerobic results were combined for further analysis in order to present the response of the whole community.

Figure 2.1 Soil substrate membrane system (SSMS) set up. (A) An inverted tissue culture insert (TCI) with a soil slurry created to saturate the membrane of the TCI. (B) An SSMS set up with an inoculated polycarbonate membrane placed on top of the inverted TCI. (C) A complete set up of the

SSMS cultivation system. Up to six TCIs can be placed inside a six-well plate.

2.2.4 Epi-fluorescent microscopy (EFM)

The PC growth membranes were removed carefully from the top of the TCIs and cut into small sections as described in Ferrari et al. (2008). A solution of low setting point agarose

(Sigma-Aldrich, Castle Hill, Australia) was heated to boiling then allowed to cool to 30-40

°C. The small sections of membrane were dipped into the solution and allowed to dry at 37

°C, fixing the cells to the membrane. Once the cells were fixed, the membranes were mounted onto a microscope slide with 10 µl of Vectashield mounting medium (Vector laboratories, USA) and 1 µl of SYBR Green II RNA gel stain (Invitrogen). A cover slip was placed over the membrane on top of the microscope slide. The slides were visualised using an

Olympus BH2 microscope with a Nikon DXM 1200F digital camera (Olympus, North Ryde,

Australia) with appropriate filters for SYBR Green emission wavelength (λmax = 520 nm).

2.2.5 Secondary cultivation on low nutrient artificial media

The tips of two quarter growth membranes were secured in the lid of 1.5 ml tubes and vortexed with 1 ml of filtered Tris-EDTA (TE) buffer to remove the cells from the membrane. The buffer was centrifuged for 60 s at 12 000 revolutions per minute (RPM) to create a pellet and the supernatant was removed. The pellet was resuspended in 300 µl TE buffer and 100 µl was used as an inoculum to spread in triplicate on low nutrient media plates

(0.1 x RAVAN). This process was repeated for all the samples. The resulting mixed colony growth was sub-cultured until pure cultures were obtained.

2.2.6 DNA extraction of pure isolates

DNA was extracted from isolates through boiling of sub-samples of cells at 99 °C for 10 min.

Approximately 10 ng of the resulting DNA lysate was added to each 50 µl PCR reaction using primers F27 and R1492 to amplify the 16S rRNA gene (Lane 1991). The PCR protocol

28 consisted of initial denaturation at 94 °C for 5 min, followed by 35 cycles of 94 °C for 30 sec,

60 °C for 30 sec and 72 °C for 30. All reactions were terminated with a final step of 72 °C for

5 min. Reactions were carried out on a My-cycler Thermocycler (BIO-RAD, USA). Products were visualised on a 2% agarose gel following staining with SYBR safe according to manufacturer’s instructions.

2.2.7 Identifying unique isolates with restriction fragment length polymorphisms (RFLP)

A portion (15 µl) of the 16S rRNA PCR amplicons were digested with the restriction enzymes

HinFl and Rsal (Promega, USA) following the manufacturer’s instructions. Digested products were visualised on a 2% agarose gel after staining with SYBR safe. Unique RFLPs were selected for further 16S rRNA gene sequencing. For RFLPs that were observed in high numbers, PCR products were sourced from more than one isolate.

2.2.8 Sanger sequencing of pure cultured isolates

The PCR products of the selected isolates were purified using a PCR purification kit (Qiagen,

USA). Approximately 30 ng of the purified PCR product was utilised in a sequencing reaction with 1 µl BigDye Terminator (Applied Biosystems) and 1.5 µl 5X buffer, per 20 µl reaction. The sequencing reaction protocol consisted of 25 cycles of 96 °C for 10 sec, 50 °C for 5 sec and 60 °C for 4 min. Reaction products were cleaned with ethanol and EDTA

(Ethylene diamine tetraacetic acid) and analysed with an ABI 3730 sequence scanner at the

Ramaciotti Centre for Gene Function Analysis (UNSW Australia). All sequence data from the ABI sequencer was checked for chimeric artefacts using http://foo.maths.uq.edu.au/~huber/doc/doc/bellerophon.pdf. The data was then matched using the NCBI nucleotide Basic Local Alignment Search Tool (BLAST) for nearest matches and closest cultured isolates in the Genbank Database.

2.2.9 DNA Extraction from SSMS membrane

The mixed bacterial community present on the enriched SSMS growth membranes were used for DNA extraction. Two membrane quarters were added to a 2 ml lysing matrix tube from the FastDNA SPIN kit for soil (MP Biomedicals, USA) and extracted according to the manufacturer’s instructions. To confirm the presence of DNA in the extracts a portion of each sample (6 µl) was electrophoresed on a 2% agarose gel.

2.2.10 DNA Extraction from soil

After preliminary optimisation extractions of gDNA from polar soils, the FastDNA SPIN kit for soil (MP Biomedicals, Seven Hills, NSW, Australia) was selected over other commercial soil DNA extraction kits based on higher DNA yields, and the amount of soil used for the extractions was reduced from 0.5 g to approximately 0.25-0.30 g (data not shown). To account for potential variation or influence of extraction biases, each soil sample was extracted in triplicate. Quantification of the genomic DNA (gDNA) for the 3 technical replicates was determined with a Picogreen assay according to the manufacturer’s instructions (Life Technologies, Mulgrave, VIC, Australia), and the absorbance was measured on a fluorescence plate reader (SpectraMax M3 Multi-Mode Microplate Reader,

Molecular Devices, Sunnyvale, CA, USA).

2.2.11 Pyrosequencing of the 16S rRNA gene

The gDNA extracted from the membranes and directly from the soil samples was used as a template for amplification of the 16S rRNA gene, with the universal primers, 28F and 519R

(112). Raw data was provided in the form of standard flowgram format (sff) files. The sequences and flowgrams were extracted from the sff files, de-multiplexed and error-checked via the Pyronoise algorithm (113) in MOTHUR (114). Further quality screening of the

30 sequence data included removing short reads (<150bp), homoploymers (>8bp repeats) and truncation of 16S reads greater than 450bp, followed by de-replication for ease of computation. Chimera checking with the UCHIME algorithm (115) allowed removal of any other erroneous sequences and sequences were pre-clustered at 1% to account for 454’s titanium instrument error rate (116). Bacterial seed sequences were aligned to the curated

SILVA secondary structure alignment (117). Aligned 16S sequences were then clustered into

OTUs based on 96% sequence similarity as the best definition for species-level OTUs in this region of the 16S gene (118). Taxonomic assignment of the identified bacterial OTUs was performed using the Greengenes 2011 database trimmed to the same region as our amplicons

(V1-V3)(119). An OTU abundance-by-sample matrix was generated from the bacterial dataset with MOTHUR (114). As an additional stage of quality control, global 16S rDNA gene sequence singletons were removed from the dataset. primers 28F and 519R (Lane

1991).

2.2.12 Calculation of diversity estimates from pyrosequencing data

To evaluate the overall bacterial diversity recovered from the SSMS following pyrosequencing samples, aerobic and anaerobic sequences were combined for each fuel concentration analysed. Multivariate data analysis was conducted with the software packages

PRIMER v6 and Permanova+ (120). To account for the log-normal distribution of the data, the abundance-by-sample matrix was square root transformed. After transformation the skewness and kurtosis was reduced closer to 0. For sample comparison the abundance-by- sample matrix was than standardised to express the OTU abundances as relative abundances.

The community diversity estimates of richness, Pielou’s evenness coefficient (J), Shannon

(H’) and Simpson (λ) were calculated in PRIMER.

2.2.13 Dose-response modelling

The dose response of the microbial communities to the diesel fuel was evaluated by fitting dose response curves to microbial community indices across the diesel fuel gradient. A large number of data points are recommended over high replication to best capture potential shifts in the curve (178, 179). To maximise the confidence surrounding the dose-response curves, individual samples and their actual TPH concentrations were utilised instead of an average of the duplicates within concentration ranges. The dose response curves were generated within the “drc” r package (180). To establish the most suitable model, the quality of fit was assessed with Akaike’s information criteria and ANOVA comparisons of models. Effective concentration values (including standard error and confidence intervals) were calculated from the best fitting dose response curve.

2.2.14 Response of individual genera to SAB

The response of individual genera across the spiking range was further interrogated by a script in R, which plotted the log abundance against the log TPH concentration. The script fitted the response data to a linear equation, determining if each genus was significantly inhibited or stimulated with increasing TPH, and then calculating the slope of the line. The results were summarised in a volcano plot where the x-axis indicated the slope of the line

(negative values indicated inhibition and positive values indicated stimulation), and the y-axis indicated the – log (P value). Calculating the –log of the P values allowed the most significant changes to be towards the top of the plot, and individual genera with the most substantial increases or decreases were positioned furthest from the middle. Significance was tested at (P<0.05).

2.3 RESULTS

2.3.1 Comparison of the methods used to capture bacterial diversity

Molecular approaches sequencing total gDNA (directly from the soils or from the SSMS enrichments) recovered far greater numbers of species, genera and phyla than the pure culture dependent methods alone (Figure 2.2). Of the pure culture dependent methods, the SSMS and artificial media combination recovered more genera and species than culturing directly from the soil, yet both pure culture methods recovered the same 4 phyla; Proteobacteria,

Firmicutes, Actinobacteria and Bacteroidetes (Table 2.22).

Table 2.2 Unique phyla recovered with culture dependent and culture

independent methods

Direct culturing SSMS culturing gDNA from SSMS gDNA direct from SOIL enrichment Proteobacteria Proteobacteria Proteobacteria Proteobacteria Firmicutes Firmicutes Firmicutes Firmicutes Actinobacteria Actinobacteria Actinobacteria Actinobacteria Bacteroidetes Bacteroidetes Bacteroidetes Bacteroidetes Gemmatimonadetes Gemmatimonadetes Acidobacteria Acidobacteria Chlorobi Chlorobi Chloroflexi Chloroflexi Verrucomicrobia Verrucomicrobia Spirochaetes Spirochaetes Planctomycetes Planctomycetes TM7 17 x Candidate divisions 18 x Others

The sequencing of gDNA directly from the soils recovered approximately double the number of phyla and two thirds more genera than sequencing from the SSMS enrichment. The most abundant 11 phyla were consistent between the SSMS and soil gDNA methods yet there were

33 an extra 17 candidate divisions and 18 other phyla recovered directly from the soil. With both total gDNA sequencing methods, the mid-range fuel concentrations had the highest number of genera. The difference in phyla and genera numbers observed between the low, medium and the high fuel ranges was greatest for extracted gDNA directly from the soil.

Figure 2.2 The number of unique phyla and genera recovered with

culture dependent and culture independent techniques. Numbers

recovered are grouped by diesel fuel concentration ranges; low (DL-400 mg

kg-1), medium (401-5000 mg kg-1) and high (>5000 mg kg-1). The total

extraction of gDNA from SSMS enrichments and directly from soil retrieved

considerably more species than the culturing techniques. The pyrosequencing

results revealed a peak in unique phyla and genera at mid-level diesel fuel

concentrations.

2.3.2 Cultivation from soil and micro-cultivation from SSMS onto artificial media

A total of 300 isolates were recovered directly from the soil using artificial media and from the combination of SSMS and artificial media. From the 300 isolates, 69 RFLPs were identified as unique. Representative isolates from each unique RFLP were then sequenced. A total of 46 species from 24 genera were identified (Table 2.33). The SSMS technique recovered a greater number of species in comparison to the traditional culturing efforts directly from soil with 32 species from 18 genera isolated compared to 18 species and 11 genera. Of the 69 unique profiles identified, 10 were unable to be successfully sequenced.

The majority of isolates recovered belonged to the genera Arthrobacter, Spingomonas and

Pseudomonas with 89, 60 and 43 isolates respectively. There were 5 novel isolates unable to be accurately identified beyond genera (Achromobacter xylosoxidans 96%, Arthrobacter psychrochitiniphilus 96%, Arthrobacter stackebrandtii 96% Hymenobacter ocellatus 96%

Sphingomonas faeni 95% and Yersinia enterocolitica 96%) and 2 novel isolates were unable to be identified beyond family (Adhaeribacter terreus 91%, Arthrobacter sulfureus 94%)

(Table 2.33).

Of the pure cultured genera, 4 were more dominant in the low TPH contamination samples including Acidovorax, Brevundimonas, Microbacterium and Sporosarcina, while the genera

Rhodococcus were particularly prevalent in the mid-range TPH samples (Figure 2.2). The genera Arthrobacter, Burkholderia, and Sphingomonas were dominant across all TPH ranges.

Pseudomonas and Rhodanobacter were also present in all TPH ranges but increased their prevalence in the samples with high TPH, along with Achromobacter. Of the genera recovered more than once, Microbacterium, Paenibacillus and Burkholderia were recovered only with the SSMS to artificial media combination and Acromobacter was recovered only with the direct culturing method from soil.

Figure 2.3 Relative abundance of pure cultured phyla across diesel fuel concentration ranges. (A) Direct culturing, (B) SSMS culturing. The prevalence of Proteobacteria increased across the TPH range, however all concentrations were dominated by Proteobacteria.

Table 2.3 Recovered isolates direct from soil to artificial media and SSMS

to artificial media culturing methods.

TPH Isolate No of Conc. % number RFLPS artificial/SSMS range closest cultured representative similarity 66 1 Artificial High Achromobacter xylosoxidans 98 20A 4 Artificial High Achromobacter xylosoxidans 99 20B 4 Artificial High Achromobacter xylosoxidans 96 54 1 SSMS Control Acidovorax radicis strain 98 60 3 Artificial Control Acidovorax wohlfahrtii 99 61 1 Artificial Med Adhaeribacter terreus 91 Arthrobacter 36 2 Artificial Med psychrochitiniphilus 96 3A 3 SSMS Low Arthrobacter stackebrandtii 96 Arthrobacter 3B 3 SSMS High psychrolactophilus 99 Arthrobacter 3C 3 Artificial Control psychrolactophilus 99 Arthrobacter 1A 79 SSMS High psychrolactophilus 99 1B 79 SSMS Low Arthrobacter stackebrandtii 97 31 2 SSMS Low Arthrobacter stackebrandtii 99 33 1 SSMS Low Arthrobacter stackebrandtii 98 52 1 SSMS Low Arthrobacter sulfureus 94 55 1 SSMS Control Bosea massiliensis 98 25 1 SSMS Low Brevundimonas bullata 99 26 1 Artificial Low Brevundimonas staleyi 98 27 1 SSMS Control Burkholderia fungorum 99 15 1 SSMS High Burkholderia fungorum 99 24 1 SSMS Med Burkholderia fungorum strain 98 21 1 SSMS High Cellulomonas cellasea 99 11 1 Artificial Med Crocinobacterium jejui 99 Herminiimonas 41 1 SSMS Low arsenicoxydans 96 7 1 SSMS Med Hymenobacter ocellatus 96 45 1 Artificial Low Kaistia adipata 99 16B 3 SSMS High Knoellia sinensis 99 53 1 SSMS Low Microbacterium lacus 99 16A 3 SSMS Low Microbacterium lacus 100 38 1 SSMS Low Paenibacillus tundrae 100 56 1 SSMS Med Paenibacillus wynnii 99 13A 15 SSMS High Pseudomonas brenneri 99 13B 15 Artificial High Pseudomonas putida 99 13C 15 Artificial High Pseudomonas putida 99 13E 15 SSMS High Pseudomonas migulae 99

Pseudomonas 10 7 Artificial Low frederiksbergensis 99 Pseudomonas 46 2 Artificial Low frederiksbergensis 99 4A 21 Artificial Control Pseudomonas thivervalensis 98 Pseudomonas 4B 21 Artificial Low frederiksbergensis 99 4C 21 SSMS High Pseudomonas mandelii 100 12 8 SSMS High Pseudomonas migulae 98 19 1 Artificial High Pseudomonas thivervalensis 98 30 1 SSMS Low Pseudomonas veronii 99 Pseudoxanthomonas 44 1 Artificial Med yeongjuensis 99 22 1 SSMS High Rahnella aquatilis 98 14 2 SSMS Low Rhodanobacter fulvus 99 67 1 SSMS High Rhodanobacter ginsengisoli 98 68 2 SSMS High Rhodanobacter spathiphylli 98 9 1 Artificial Med Rhodococcus baikonurensis 100 18A 2 Artificial Med Rhodococcus baikonurensis 99 18B 2 Artificial Med Rhodococcus baikonurensis 100 58 1 SSMS Med Rhodococcus erythropolis 99 43 3 Artificial Med Rhodococcus erythropolis 99 2 39 Artificial Med Rhodococcus globerulus 99 40 1 SSMS Low Sphingomonas echinoides 99 37A 3 SSMS Low Sphingomonas aquatilis 99 37B 3 SSMS Low Sphingomonas echinoides 99 17B 34 SSMS Low Sphingomonas echinoides 100 17C 34 Artificial High Sphingomonas faeni 100 17D 34 Artificial High Sphingomonas faeni 99 17G 34 SSMS High Sphingomonas faeni 100 57 1 SSMS High Sphingomonas faeni 95 17F 34 SSMS Low Sphingomonas rhizogenes 99 6 3 Artificial Low Sphingomonas oligophenolica 99 49 1 SSMS Low Sphingomonas oligophenolica 99 17H 34 SSMS Low Sphingomonas oligophenolica 99 8 1 SSMS High Sphingomonas oligophenolica 99 5B 3 Artificial High Sphingomonas oligophenolica 100 62 8 SSMS Control Sphingomonas sanxanigenens 99 69 4 SSMS Low Sphingomonas sanxanigenens 99 28 1 SSMS Low Sphingomonas sanxanigenens 96 5A 3 SSMS Low Sphingomonas sanxanigenens 99 59 1 SSMS Control Sphingopyxis flavimaris 100 29 1 SSMS Control Sporosarcina globispora 99 48 1 Artificial Low Sporosarcina globispora 99 51 1 SSMS Control Sporosarcina psychrophila 99 23 1 SSMS High Yersinia enterocolitica 96 Highlighted isolates have <97% similarity to cultured isolates. 38

2.3.3 Bacterial diversity from SSMS enrichments

A total of 34308 16S rDNA sequences across the 14 (aerobic and anaerobic unamended controls plus aerobic and anaerobic duplicates in each concentration range) samples (average

2450 per sample) were obtained. After filtering out the low quality sequences a total of 24494

(average 1749 per sample) remained. At the phyla level the SSMS growth membranes were dominated by Proteobacteria, contributing 88-90% of the total relative abundance.

Firmicutes was present in the range of 4-8 % of the total relative abundance, while

Actinobacteria, Bacteroidetes, and Acidobacteria were also detected but in relative abundances less than 5 % (Figure 2.4A).

Figure 2.4 Relative abundance of major phyla. Total gDNA recovered

from SSMS enrichments (A) and directly from soil (B). For both methods

Proteobacteria was the dominant phyla, yet more phyla was observed in total

gDNA recovered directly from the soil.

Within Proteobacteria, the control, low and medium diesel fuel concentrations were dominated by Rhodoferax (26-36 %) and Simplicispira (14-27%), yet within the high diesel fuel range both genera were barely detected. Pseudomonas was prevalent across all concentrations but dominated the high-range diesel fuel samples contributing over 65% of the total relative abundance. Within the high diesel fuel range Dyella (7%), Parvibaculum (3%) and Rhodanobacter (0.8%) were also detected (Figure 2.5A). Although not in the same proportion, the SSMS enriched for approximately 1/3 of the total genera recovered directly from the soil (Table 2.2).

Figure 2.5 Relative abundance of major genera within Proteobacteria.

Genera recovered from SSMS (A) and directly from soil (B).

Pseudomonas was the dominant genus across all concentrations and both

methods. The total gDNA recovered directly from the soil had greater

numbers and evenness of genera.

2.3.4 Bacterial diversity directly from soil

A total of 45322 16S rDNA sequences across the 9 soil samples (3 x control, plus duplicates in each concentration range) were obtained. After filtering out the low quality sequences a total of 32358 sequences (average 3595 per sample) remained. As with the pyrosequencing from the SSMS membranes, the direct soil diversity was dominated by species within

Proteobacteria (Figure 2.54B). However, the detected prevalence of Proteobacteria was lower than the SSMS membranes ranging from 49-51% of the total relative abundance for the control, low and medium diesel fuel concentrations (Figure 2.54B). Other phyla also found to be abundant in the control, low and medium diesel fuel ranges included Firmicutes (9-10%),

Actinobacteria (10-16%), Bacteroidetes (5-11%), Gemmatimonadetes (2-5%), Acidobacteria

(2-5%), and Chloroflexi (4-6%). The phyla Planctomycetes, Verrucomicrobia,

Cyanobacteria, TM7 and the remaining phyla denoted ‘Others’ were also detected, contributing less than 5% each of the total relative abundance. In the high diesel fuel ranges the composition changed dramatically with Proteobacteria contributing over 80% of the total relative abundance and apart from Firmicutes (7%) in the highest fuel concentration; all other phyla contributed less than 5% each to the total relative abundance.

Within Proteobacteria, the community composition of the soil gDNA of the control, low and medium diesel fuel concentration ranges was found to be very different to the SSMS membranes enriched from the same soil samples (Figure 2.5B). The genera Simplicispira and

Rhodoferax were barely detected and instead Delfuviicoccus was the most abundant genera contributing between 9-20% of the total relative abundance. A greater number of genera were detected in relatively low abundance in soil, including Thioflavicoccus (5-9%), Rudaea (6-

7%), Pelagibius (4-9%), unclassified species within Xanthomonadaceae (3-6%), unclassified species within β-Proteobacteria (4%), Rhodanobacter (3-5%) and Dyella (3-4%). There was

41 also an additional 305 genera represented by ‘Others’ all contributing less than 1% each towards the total relative abundance.

In the highest diesel fuel samples there was a decline in the number of detected genera

(Figure 2.5B). The genera recovered were consistent with those extracted from the SSMS membranes of the same high TPH samples, although present in different proportions.

Pseudomonas contributed up to 30%, Rhodanobacter (5%), Parvibaculum (22%) and Dyella

(9%).

2.3.5 Response of individual genera to TPH

The SSMS enrichments had 12 genera significantly inhibited and 6 significantly stimulated (p

< 0.05) (Figure 2.6, Table 2.44). The diversity present in gDNA extracted directly from the soil had a more disproportionate number of inhibited genera across the fuel range with a total of 71 genera significantly inhibited and only 5 genera significantly stimulated. As highlighted in the Figure 2.5, many of the genera found to be significantly stimulated with increasing diesel fuel were consistent between the two methods, whereas the abundance and identities of the inhibited genera varied considerably (Table 2.44, Figure 2.6).

Figure 2.6 A volcano plot of the genera significantly inhibited or stimulated with increasing diesel fuel concentrations. Blue diamonds indicate significance at < 0.05. Pink diamonds were non-significant. On the x- axis, negative slope values indicate inhibition and positive values indicate stimulation. A greater number of inhibited genera were detected with the culture independent methods.

Table 2.4 The most significantly inhibited and stimulated genera present

in gDNA recovered from the SSMS enrichments or directly from soil.

relative relative abundance abundance in control in high % Slope P value % (average) (SD) SSMS Inhibiteda Rhodoferax -0.41 0.02 13.3 0.2 0.3 Simplicispira -0.35 0.02 15.0 1.1 1.6 unclassified_Comamonadaceae -0.24 0.01 3.8 undetected n.a unclassified_Rhodocyclaceae -0.21 0.02 4.4 undetected n.a unclassified_Bacteroidetes -0.18 0.03 2.9 undetected n.a Methylophilus -0.15 0.04 1.2 undetected n.a Malikia -0.15 0.02 1.4 undetected n.a Lactobacillus -0.14 0.01 8.8 2.7 0.9 Flexithrix -0.11 0.03 1.3 undetected n.a Dechloromonas -0.10 0.02 1.0 undetected n.a SOIL Inhibiteda Opitutus -0.24 0.02 4.5 0.5 0.6 Unclassified Bacteria -0.23 0.04 8.2 1.4 0.1 unclassified_Xanthomonadaceae -0.21 0.01 3.6 0.5 0.3 unclassified_γ-proteobacteria -0.20 0.02 4.7 0.5 0.0 Steroidobacter -0.20 0.01 3.1 0.5 0.3 unclassified_Chloroflexi -0.19 0.02 6.2 1.4 0.2 Geminicoccus -0.19 0.04 2.7 0.2 0.2 Gemmatimonas -0.19 0.02 7.4 2.6 0.7 unclassified_β-proteobacteria -0.18 0.04 4.3 0.8 0.3 unclassified_Planctomycetaceae -0.18 0.02 2.7 0.3 0.1 SSMS Stimulated Herbaspirillum 0.26 0.02 0.3 3.0 1.0 Dyella 0.23 0.04 0.7 2.3 2.5 Janthinobacterium 0.20 0.01 0.7 3.7 0.4 Parvibaculum 0.19 0.02 0.0 2.3 1.2 Pseudomonas 0.14 0.01 8.7 19.1 2.3 Azotobacter 0.05 0.05 0.0 0.3 0.1 SOIL Stimulated Pseudomonas 0.34 0.01 0.5 8.8 1.8 Parvibaculum 0.27 0.00 0.7 3.4 3.3 Herbaspirillum 0.15 0.03 0.1 1.1 1.3 Dyella 0.13 0.03 1.5 3.4 2.3 Halotalea 0.13 0.01 0.0 0.7 0.8 a Only the 10 most inhibited genera from each method were included in this list.

2.3.6 Total bacterial community dose-responses to TPH

The diversity of the total bacterial communities recovered from the SSMS and directly from the soil was evaluated for their response to TPH. As expected the species richness was higher in the soil community in comparison to the enriched SSMS community (Figure 2.7).

The species evenness (J’), Shannon (H’) and Simpson (λ) diversity estimates were also greater in the soil communities. For the soil communities, all of the community estimates measured declined with increasing TPH. For the SSMS communities the response was negligible, with small increases in the evenness and Simpson estimates and a decline in the species richness. Both methods were capable of generating dose-response curves. However, the results generated directly from the mixed soil communities’ generated more accurate dose-response curves and EC20 values with less associated error (Table 2.4). The EC20 values generated from the evenness, Shannon and Simpson data for the SSMS community were extremely low (0.02 – 1.4x10-76) and unrealistic in the context of the contaminant with minimum detection limits 20 to 160 depending on the method used (Table 2.5)

Figure 2.7 DRC modelling of community diversity estimates against diesel fuel concentration gradient for SSMS and total soil gDNA. The

SSMS enriched for 17.9% of the total species richness recovered directly from the soil. A decline in species richness, evenness, Shannon and Simpson diversity indices was seen with both methods. The decline in diversity measures from low and medium TPH to high TPH soils was greatest in soil communities.

Table 2.2.5 EC20 values generated from dose-response curves and the

corresponding dose-response model parameters.

b c f Population gDNA Model S.E. d.f. EC20 Model parameters Measurea source Conc.d S.E.e B d E

Sobs SSMS W2.3 89.5 4 3650 6377 5.15 227.2 8.0 SOIL W1.4 110 3 2493 13910 43.7 1140.2 3.62 Evenness SSMS LL.3 0.05 4 2.9x10-17 1.2x10-15 -45.2 2.90 1.7 (J’) SOIL W1.4 0.02 3 189 142 9.60 5.9 3.1 Shannon SSMS LL.3 0.37 4 1.4x10-76 1.1x10-73 -28.0 0.6 1.8 (H’) SOIL W1.4 0.05 3 459 244 6.5 0.8 3.0 Simpson SSMS W1.3 0.05 4 0.03 0.28 -40.9 0.9 1.73 (λ) SOIL W1.4 0.01 3 505 480 10.3 1.0 3.1 a Diversity measure, b The best fitting dose response curve for the data as determined by AIC (W2.3 = weibull 2 curve with 3 parameters, W1.4 = weibull 1 with 4 parameters, LL3 = log logistic with 3 parameters). c Degrees of freedom, d Concentration that corresponds to 20% e f effect on the community. Standard error of the generated EC20 concentration, Model parameters; b = relative slope, d = upper limit, e = point of inflection.

2.4 DISCUSSION

The SSMS and artificial media combination recovered greater numbers of genera and species and a greater number of novel isolates than culturing directly onto artificial media alone.

However, both methods did not recover any isolates belonging to phyla outside of

Proteobacteria, Firmicutes, Actinobacteria and Bacteroidetes (Table 2.22). Of the species recovered there was a selection towards species known to be associated with petroleum hydrocarbons in polar sites including Pseudomonas sp, Rhodococcus sp, and Spingomonas sp

(15, 16, 121). Species not commonly associated with petroleum hydrocarbons were isolated in low numbers. Furthermore, the phylogenetic diversity that was recovered bore little resemblance to the diversity recovered with the culturing independent methods.

Sequencing of the enriched mixed communities present on the SSMS revealed that a considerable proportion (17.9%) of the diversity recovered directly from the soil was able to be micro-cultivated using the approach. This suggests a promising potential for the SSMS to be combined with single cell applications such as flow cytometry (122, 123) and micromanipulation (123) to recover previously uncultured species. While the SSMS enrichments goes some way in bridging the gap between culture dependent and culture independent methods, in control, low and medium TPH concentration ranges, there was limited similarity of community structure and diversity between the SSMS enrichments and the sequencing of total gDNA direct from the soil.

The soils with high TPH contamination range showed the greatest similarity between the

SSMS and total soil diversity. All methods recovered Pseudomonas and Rhodanobacter and the sequencing of the mixed SSMS and soil communities also recovered Dyella and

Parvibaculum. Further, a decline in richness for the high TPH concentration range was observed with all methods. As with traditional culturing methods (101), the novel culturing methods used here appear to maintain a selective bias towards heterotrophic species with fast growth strategies, in particular species with known hydrocarbon degrading potential.

Sequencing the mixed communities directly from the soil extracts recovered the largest number of phyla (including 17 candidate divisions) and genera. The sequencing results also captured the large proportion of genera significantly inhibited with increasing diesel fuel, despite low relative abundances. There is evidence that many important soils ecosystem services are carried out by organisms in low abundance (124), therefore in order to best protect the complex soil systems; it is important to be able to capture organisms beyond the dominant few. Without providing adequate information on the inhibited and unknown proportion of the community, we suggest the novel culturing techniques are inadequate to

48 illuminate the whole communities’ response to petroleum hydrocarbons. The greater and more realistic coverage of the bacterial community structure, and the ability to generate reliable dose-response curves from the data led us to conclude that culture independent applications directly from the soil matrix are the most appropriate methods to pursue microbial indicators of petroleum hydrocarbon toxicity in polar soils.

3 COMPARATIVE SENSITIVITY OF FINGERPRINTING

AND PYROSEQUNCING TO EVALUATE SOIL

BACTERIAL COMMUNITIES

3.1 INTRODUCTION

The microbial ecology landscape has been revolutionised by culture-independent methods.

Key developments in sequencing technologies such as high-throughput capabilities, multiplexing technologies and miniaturisation of sequencing chemistries continue to reduce the time and costs involved with sequencing (125). The greater affordability and accessibility of next-generation sequencing (NGS) technologies has seen the volume and the application of sequencing dramatically increase. Examination of natural and engineered microbial ecosystems using NGS is now prolific throughout the literature, with the trend likely to continue (126-129). Yet, despite being more affordable than ever, delivering a large scale, well replicated, multivariate microbial diversity investigation with NGS remains an expensive and time consuming option, requiring significant downstream bioinformatic processing power and skill.

Genotypic fingerprinting techniques are an established set of culture-independent molecular tools, used to rapidly profile microbial communities in a variety of systems (68, 130, 131).

They include single-strand conformation polymorphism (SSCP), automated ribosomal intergenic spacer analysis (ARISA), temperature gradient gel electrophoresis (TGGE), denaturing gradient gel electrophoresis (DGGE) and terminal restriction fragment length polymorphism (T-RFLP). Unlike sequencing technologies, genotypic fingerprinting does not provide direct taxonomic information, though in the case of DGGE and less frequently T-

RFLP (132) and ARISA (133), the data output can be analysed further to obtain

50 identification. Limited capacity to capture the presence of rare taxa and the inaccuracy of genotypic fingerprinting based abundance estimates are well described limitations of fingerprinting techniques (134). Variations between analyses and semi-quantitative results have also been suggested to hinder robust comparisons between studies and even investigators (135, 136).

Despite these limitations, the use of fingerprinting methods to provide rapid and relatively inexpensive community profiles remains widespread (137-139). A number of small scale investigations have recently demonstrated consistent microbial community structures generated from fingerprinting and pyrosequencing methods (132, 140, 141). Gillevet et al.

(2009) used 454 pyrosequencing based analysis of the 16S rRNA gene (or “16S tag sequencing”) to confirm that ARISA fungal profiles may represent multiple taxa and a given taxon may be represented by more than one operational taxonomic unit (OTU), yet the fungal community trends associated with four salt-marsh plants across six time points (n=24), were still consistent between ARISA and 16S tag sequencing methods (140). Similarly, Pilloni et al. (2012) observed that T-RFLP and 16S tag sequencing recovered similar microbial community structures from a contaminated aquifer and highly reproducible read abundances across biological and technical replicates (n=9) (132). Cleary et al. (2012) found consistent patterns across composite samples of four rhizosphere environments (nursery, nursery transplant, native and mangrove) using DGGE data and 16S tag sequencing data (141).

The aim of this chapter was to investigate the sensitivity and hence utility of one commonly used NGS platforms - 454 Titanium pyrosequencing (used in Chapter 2) against two genotypic fingerprinting techniques, T-RFLP and ARISA. We chose to use a large sample set

51 of 225 soils obtained from across East Antarctica and the high Arctic, hence incorporating local and global spatial scales. We compared the ability of these culture-independent techniques to detect similarities in biological assemblages and to correlate these similarities with variations in environmental variables. T- RFLP was selected because it targets the same genetic marker (16S rRNA gene) as our NGS approach, allowing direct comparison between the two. ARISA, which targets the ITS region, was selected based on its reportedly greater taxonomic resolution then T-RFLP (142). We hypothesised that the NGS approach would provide far greater resolution than either of the fingerprinting methods, but that all methods would provide similar community structure patterns.

3.2 METHODS

3.2.1 Site descriptions and sampling design.

In total, eight polar locations were selected (Figure 1). Five sites from Antarctica (Figure

3.1B); Mitchell Peninsula (66°31’S, 110°59’E), Casey station (66°16’S, 110°31’E),

Robinson’s Ridge, denoted RT (66°22’S, 110°35’E), Herring Island (66°24’S, 110°39’E) and

Browning Peninsula (66°27’S, 110°32’E). Three sites from the high Arctic (Figure 3.1A);

Alexandra Fjord Highlands, Canada, denoted C_AFH (78°51’N, 75°54’W) Spitsbergen

Longyearbyen Slijeringa, Norway, denoted N_SS (78°14’N, 15°30’W) and Spitsbergen

Longyearbyen Vestpynten, Norway, denoted N_SV (78°14’N, 15°20W). At each site a variable-lag distance geospatial design was used to collect 93 samples. Three parallel transects 300 m long and 2 meters apart were established as outlined in Bajeree and Siciliano

(2012) (Figure 3.1C). Soil samples were collected at 31 points (0, 0.1, 0.2, 0.5, 1, 2, 5, 10, 20,

50, 100, 100.1, 100.2, 100.5, 101, 102, 105, 110, 120, 150, 200, 200.1, 200.2, 200.5, 201,

202, 205, 210, 220, 250 300m) along each transect. This design allowed fine-scale (0 to 1 m), medium-scale (1 to 10 m), and large-scale (10 to 300 m) spatial patterns of microbial

52 communities to be captured. For this investigation all samples from the Mitchell Peninsula were characterised (n = 93) and a sub-selection of samples were included from the other seven sites (n = 18 / 24 per site), resulting in a total of 225 samples.

3.2.2 Physical and chemical data collection.

Physical and chemical data were collected concomitantly for each soil sample and analysis was conducted at the AAD. The pH and conductivity was measured using a 1 in 5 soil to distilled water suspension. Grain size of the soils was measured by removing particles larger than 2 mm (gravel) then separating particles based on size into silt (<63 µm) and sand (63-

2000 µm). The particle size ranges were measured using laser scatter and the percentage of the total soil for each fraction calculated. Total Kjeldahl digested phosphorus and nitrogen were measured as well as total carbon by combustion and NDIR gas analysis (143). The water extractable chemicals (NO2, Br, NO3, PO4, SO4, and NH4) were obtained with a 1 in 5 dilution of soil to distilled water on a dry matter basis [mg/kg]. Anions were analysed by ion chromatography and ammonium cation concentration was measured by the colorimetric indophenol method. Measurement of major elemental concentrations (SiO2, TiO2, Al2O3,

Fe2O3, MnO, MgO, CaO, Na2O, K2O, P2O5, SO, and Cl) was performed using X-ray fluorescence (XRF) analysis. Additional site parameters including slope gradient, slope aspect and elevation were derived from the GPS information using GIS software and a site digital elevation model.

Figure 3.1 Map of sampling site locations. (A) Global map of the polar sampling sites from the high Arctic and East Antarctic. (B) Location of the five East Antarctic sampling sites. (C) Transect and sample locations at Mitchell Peninsula.

3.2.3 DNA Extraction from soil samples.

Approximately 0.25-0.30 g of soil was used for DNA extraction using the FastDNA SPIN kit for soil (MP Biomedicals, NSW, Australia) as described in 2.2.10. Each of the 225 polar soils was extracted in triplicate and the gDNA was quantified with Picogreen (Life Technologies,

VIC, Australia) and fluorescence measured on a fluorescence plate reader (SpectraMax M3

Multi-Mode Microplate Reader, Molecular Devices, CA, USA).

3.2.4 Bacterial ARISA.

The three replicate DNA extracts for each of the 225 samples (n=675) were titrated to a standard working concentration (10 ng/µl). The DNA was amplified by PCR for the ITS with the universal primers 1392f; 5’ GYACACACCGCCCGT3’ and a 5’MAX labelled 23Sr

5’GGGTTBCCCCATTCRG3’ (144, 145). The 25 µl reaction contained 0.5 µM of each primer, 1 unit of Taq polymerase (Promega), 1X buffer (Promega), 2.5 mM MgCl2

(Promega), 0.25 mM of each dNTP (Promega), 0.8 µg/µl of BSA. The reaction was held at

94°C for 2 min followed by 30 cycles of amplification at 94°C for 30 sec, 55°C for 30 sec and 72°C for 30 sec, with a final extension at 72°C for 5 min (146). Fragment length separation was performed at Macquarie University on an Applied Biosystems 3730 fragment analyser (Life Technologies) and subsequently interpreted via Genemapper software (Life

Technologies). The fragments were further filtered to separate and remove background fluorescence from true peaks in the T-REX (T-RFLP analysis Expedited) online software using the default parameters (147). The filtered data was used as input to a custom fragment length binning script (148) in R v2.10.1 (149). Fragments were assigned to bins of 2 base pairs (± 1 b.p) for fragments up to 700 base pairs length, bins of 3 base pairs for fragments between 700 and 1000 base pairs length and bins of 5 base pairs for fragments above 1000

55 base pairs (133). The final data output was transformed into a peak area verses sample matrix, with each peak assumed to represent an OTU (133, 145).

The intra-sample similarity between replicates was evaluated through a Bray-Curtis similarity matrix, nMDS ordinations and cumulative rank abundance curves in PRIMER 6 and by comparing the shared and unique OTUs between replicates. After confirming a strong correlation of the technical replicates, one DNA extract from each of the 225 soils was selected at random as a representative for further analysis with 454 sequencing and T-RFLP.

3.2.5 Roche 454 FLX Titanium pyrosequencing.

The gDNA from each of the 225 polar soils was sent to an external facility (Research and

Testing Lab, Lubbock, Texas, USA) for amplicon sequencing on the 454 FLX titanium platform as described in (2.2.11). The primers, data processing software and processing conditions were consistent with those outlined in 2.2.11.

3.2.6 Bacterial T-RFLP.

Genomic DNA from each of the 225 soils was used as a template in T-RFLP analysis. Again, the 28F and 519R primers were used to amplify the 16S rDNA gene as described in (130).

Each 50 µl PCR reaction contained gDNA template, 1 mM MgCl2, HPLC-purified primers, buffer, 1 U of Taq DNA polymerase (Qiagen), 50 mM of each dNTP (Promega). The PCR comprised 1 cycle of 94 °C for 3 min; 30 cycles of 94 °C for 30 s, 55 °C for 30 s, and 72 °C for 60 s, and a final extension step of 72 °C for 10 min. PCR amplicons were isopropanol precipitated, quantified as above and 70 ng digested with the restriction endonucleases HinFI.

Digests were isopropanol precipitated, resuspended in 10 ml formamide containing LIZ600 size standard (Applied Biosystems), denatured at 95 °C (3 min), and separated by capillary

56 electrophoresis (ABI PRISM3130xl Genetic 110 Analyser, Applied Biosystems). The fragment outputs were interpreted with Genemapper software (Life Technologies) and further analysed with TREX and the interactive binner script as above. Any peaks greater than the fragment size (500bp) were removed, resulting in a peak verses sample matrix, with each peak assumed to represent an OTU.

3.2.7 Correlation of biological data between methods.

The biological data sets generated from the three methods were imported into PRIMER 6 for multivariate analysis (120). The OTU abundance-by-sample matrices were transformed

(square-root) and a series of alpha diversity indices calculated. Margalef’s index (d), species richness (S), Pielou’s evenness (J’), Shannon index (H’) and the Simpson index (λ) were used to evaluate the variability of diversity indices between datasets. Rarefaction curves were calculated for each site. The OTU abundance-by-sample matrices were then standardised to give relative OTU abundances (%). A dissimilarity matrix for each data set was calculated using the Bray-Curtis coefficient and Principal Component Ordination (PCO) plots were created to visualise the ordering of the samples in reduced (2D) space. The co-ordinates of the first principal component (PCO1) were compared to evaluate the correlation between

PCO plots. To further test the correlation of the similarity matrices, Mantel correlations were calculated with the RELATE function in PRIMER. The potential to distinguish between a priori groups was evaluated with analysis of similarity (ANOSIM) (150), using 999 permutations and assuming  = 0.05.

3.2.8 Correlation with environmental variables.

The physical and chemical data was transformed based on the distribution of the data.

Variables that were right-skewed (elevation, slope, conductivity, PO4, SO4, SiO2, Al2O3,

MnO, CaO, P2O5, SO3, Cl, moisture, NH4, N and P) were log transformed, while the water extractable NH4 and Fe2O3 were transformed by the square root. The resulting environmental data matrix was normalised (mean/SD) to reduce the data to a consistent scale. The correlation of the environmental parameters with the biological distributions in each PCO plot was calculated with Pearson’s coefficient and displayed in a radar plot. The values from some environmental parameters that were logically related and generated similar results were combined as an average, as indicated on the plot (CaO & MgO, TiO2 & SiO2, Na2O & K2O,

Al2O3 & Fe2O3). Further, a Distance-based redundancy analysis (DistLM) (151) was utilised to determine the predicator capacity of each variable analysed individually (marginal) and in combination with other variables (sequential). The selection criteria used was adjusted R2 and the model selection procedure was step-wise. A step-wise procedure allows each variable to be added and eliminated with a conditional test performed at each stage to identify any improvements in the selection criteria. The step-wise selection procedure only stops once no further improvements in the selection criteria can be made. The values were added in order of correlation from above.

3.2.9 Modification of matrices.

To evaluate the influence of OTU abundances, all of the data was transformed to binary presence/absence matrices and the analyses were repeated. To investigate the proportion of the community driving the major biological patterns, four new reduced matrices were created with OTUs contributing > 1%, > 10%, > 15% and > 20% of the total abundance. To further investigate similarities and dissimilarities between the T-RFLP data and the pyrosequencing data, an in silico evaluation was generated in TRiFle (152) based on the pyrosequencing sequences. The resulting predicted data was compared to the experimental data and presented as modified data matrices.

3.3 RESULTS

3.3.1 Intra-sample variation.

There were strong biological similarities between replicate ARISA samples observed in the

NMDS ordinations (Figure 3.2A). Comparisons of shared and unique OTUs between replicate ARISA samples further confirmed a significant correlation with 98% of the total abundance consistent between all replicates. Those OTUs found in 2 or fewer replicates contributed < 2% of total abundance (Figure 3.2B). A strong correlation of the replicates across the whole data set was also observed in the cumulative rank abundance curves (Figure

3.2C).

Figure 3.2 Correlation between technical replicates generated with

ARISA. (A) The biological similarity between replicate samples for

Alexandra Fjord Highland. (B) Relative abundance of shared and unique

OTUs between replicates across the entire dataset. (C) The cumulative rank

abundance curves for replicates A, B and C across the entire dataset.

3.3.2 Recovered OTU abundances between methods.

The pyrosequencing approach retrieved greater numbers of OTUs than either of the fingerprinting methods by approximately an order of magnitude (Table 3.1). The average number of reads for all the sites processed with the 454 pyrosequencing was 3705 (491.5 SE), with an average species richness (S) of 585 (51.5 SE). The average number of

OTUs/fragments obtained per site with genotypic fingerprinting was 99.8 (0.1 SE) for

ARISA and 95 (0.5 SE) for T-RFLP, with an average species richness of 48 (3 SE) and 65

(1.6 SE) respectively. Overall the genotypic fingerprinting captured 8-11% of the OTU richness obtained with pyrosequencing.

Table 3.1 Calculated diversity estimates for each polar site.

Unique Total Diversity Evenness Shannon 1- OTUs seqs/frag. (M) a (J')b H'(loge)c Simpson 454 Mitchell 562 4895 66.10 0.75 4.71 0.97 Casey 506 6631 57.69 0.66 4.08 0.93 RT 465 2834 58.44 0.79 4.83 0.98 Herring 422 3562 51.50 0.75 4.49 0.96 Browning 480 3249 59.20 0.77 4.75 0.97 N_SS 776 2807 97.48 0.85 5.61 0.99 N_SV 670 2491 85.64 0.85 5.51 0.99 C_AFH 800 3174 98.82 0.83 5.53 0.99 ARISA Mitchell 53 99 10.28 0.72 2.78 0.88 Casey 36 100 11.41 0.78 3.06 0.92 RT 48 100 7.51 0.75 2.60 0.85 Herring 60 100 10.21 0.74 2.86 0.90 Browning 44 100 12.82 0.82 3.34 0.95 N_SS 37 100 9.28 0.77 2.85 0.90 N_SV 54 100 7.71 0.74 2.53 0.84 C_AFH 51 100 11.45 0.78 2.94 0.89 TRFLP Mitchell 63 100 10.89 0.74 2.88 0.89 Casey 63 94 13.63 0.70 2.91 0.90 RT 73 96 15.78 0.76 3.21 0.93 Herring 60 95 12.84 0.69 2.83 0.89 Browning 63 95 13.61 0.70 2.91 0.91 N_SS 62 95 13.48 0.72 2.95 0.92 N_SV 60 96 12.94 0.73 2.99 0.93 C_AFH 68 95 14.77 0.71 2.98 0.91 a Margalef’s Diversity index, b Pielou’s Evenness index, c Shannon Diversity index.

The rarefaction curves did not reach an asymptote for any of the sites indicating the current sequencing effort was not sufficient to capture the total diversity present (Figure 3.3). The

Antarctic sites (Casey Station, Browning Peninsula, Robinsons Ridge (RT), Mitchell

Peninsula and Herring Island), were closer to reaching the asymptote of the rarefaction curves than the high Arctic sites (C_AFH, N_SS and N_SV) suggesting lower biodiversity. This lower diversity present at the Antarctic sites compared to the Arctic sites was only reflected in the 454 pyrosequencing generated diversity, richness and evenness estimates.

Figure 3.3 Rarefaction curves calculated for each site based on the 454

pyrosequencing, T-RFLP and ARISA results.

3.3.3 Correlation between pyrosequencing and fingerprinting methods.

The OTU richness for each of the 225 samples was compared between methods. The richness showed no correlation between the 454 pyrosequencing and either of the fingerprinting methods, nor did the OTU richness correlate between the two fingerprinting methods (Table

3.1, Figure 3.4). Similarly, the fingerprinting methods did not generate the same diversity trends as the 454 pyrosequencing, or each other (Table 3.1).

There was a strong resemblance between the principal component plots (Figure 3.5) and a strong correlation of the PCO1 coordinates between all methods (r2 = 0.88, 0.63, 0.60)

(Figure 3.4). The complexity and volume of the data sets make visualisation in low dimensionality (2D) difficult, with only 22-29% of the total variation within the resemblance matrix explained by PCO1 and PCO2 for the complete dataset, and 29.7-40 % for the

Mitchell Peninsula subset (Figure 3.5).

Comparisons of the similarity matrices with MANTEL tests confirmed significant resemblances between the data sets (Table 3.2). The matrices from the 454 pyrosequencing and ARISA platforms were most similar (Rho=0.62, p < 0.001). The similarity matrices from

Mitchell Peninsula were also correlated, with the 454 and ARISA technologies again generating the most similar matrices (Rho = 0.72, p < 0.001). Simplifying the data to presence/absence increased the correlation between the 454 and the T-RFLP data sets from

Rho=0.46 to Rho=0.62 (p < 0.001). Only small variations were seen for other binary matrices.

Table 3.2 MANTEL tests results (testing the similarities of the matrices)

Matrices transformed transformed (Square root) (Presence/absence) All Data Rho value P Rho value P 454 v ARISA 0.62 < 0.001 0.59 < 0.001 454 v TRFLP 0.46 < 0.001 0.62 < 0.001 TRFLP v ARISA 0.42 < 0.001 0.52 < 0.001 Mitchell Peninsula Rho value P Rho value P 454 v ARISA 0.72 < 0.001 0.73 < 0.001 454 v TRFLP 0.60 < 0.001 0.63 < 0.001 TRFLP v ARISA 0.50 < 0.001 0.53 < 0.001

Figure 3.4 Correlation of the OTU Richness and PCO1 co-ordinates. (A)

Correlation between 454 pyrosequencing data and ARISA results, (B) 454 pyrosequencing and T-RFLP results and (C) ARISA and T-RFLP results.

Figure 3.5 Comparison of the PCO plots generated with the 454 pyrosequencing, ARISA, and T-RFLP results. The PCO plots on left represent all samples, separated by sites. The following three plots on right represent all the samples within Mitchell Peninsula site, separated by the a priori distance groups.

3.3.4 Differentiating a priori groups.

All methods were capable of separating (α = 0.05) samples based on site (Table 3.4, Figure

3.5). The 454 data had the greatest resolution at the global scale, generating the highest R value of 0.90 (p < 0.001), with 8/20 of the pair-wise results generated the maximum R statistic of 1 for individual site comparisons (Table 3.4). The global R values for the ARISA and T-RFLP were 0.67 and 0.760 (P< 0.001). The reduction of the data to presence/absence did little to vary the results. All methods were capable of detecting differences between distance groups (based on sampling distances within sites), with global R values of 0.64 to

0.57 (p < 0.001) (Table 3.4). However, the distance groups 4-10 (Table 3.3) could not be significantly separated in individual pair-wise comparisons. The reduction of the data to presence/absence slightly improved the global value for the 454 pyrosequencing data

(R=0.61, p< 0.001) and the T-RFLP data (R=0.68, p < 0.001), yet the distance groups 4-10 were still not significantly resolved (Table 3.4).

Table 3.3 Distance groups delineated a priori based on sampling distances

within sites.

Group number Distance range Sampling distances included (x 3 (m) Transects) 1 0-2 0, 0.1, 0.2, 0.5, 1, 2 2 5-20 5, 10, 20 3 50 50 4 100-102 100, 100.1, 100.2, 100.5, 101, 102 5 105-120 105, 110, 120 6 150 150 7 200-202 200, 200.1, 200.2, 200.5, 201, 202 8 205-220 205, 210, 220 9 250 250 10 300 300

Table 3.4 ANOSIM results testing the ability of each method to

distinguish between a priori groups

Matrices transformed transformed Pair-wise (Square root) (Presence/ results absence) sites R P R P Lowest R Highest R values values 454 0.90 < 0.001 0.87 < 0.001 0.36 1.0 (RT - Mitchell) (8/20 results) ARISA 0.67 < 0.001 0.69 < 0.001 0.14 0.85 (RT – Mitchell) (N_SV Mitchell) TRFLP 0.76 < 0.001 0.76 < 0.001 0.54 0.97 (RT – Mitchell) (N_SS– Herring) ARISA 0.70 < 0.001 0.66 < 0.001 0.18 0.90 all (RT-Mitchell) (N_SV-Herring) D. groupsa R p R P Lowest R Highest R values values 454 0.582 < 0.001 0.61 < 0.001 -0.12 1.0 (D. group 5-6) (D. group 2-6) ARISA 0.568 < 0.001 0.56 < 0.001 -0.26 0.92 (D. group 6-9) (D. group 2-7) TRFLP 0.635 < 0.001 0.68 < 0.001 -0.23 0.99 (D. group 5-6) (D. group 2-4) a Distance groups (2 = 5-20 m, 4 = 100-102 m, 5 = 105-120 m, 6 = 150 m, 7 = 200-202 m, 9 = 250 m) (see Table 3.3).

3.3.5 Environmental predictors of bacterial composition

With similar resemblance patterns between the biological data sets, similar correlations to the environmental data were expected. The T-RFLP data was found to have the weakest correlation with the environmental data, particularly with conductivity and SO4 (Figure 3.6).

In contrast, the correlations with phosphorus, PO4, P2O5 and PO4 were greater in the T-RFLP data than in the other two data sets. The most significant environmental variables generated within the sequential DistLM (Grain size [min], pH, Latitude, TiO2, MnO, Elevation, PO4,

Cl, Carbon, Aspect, Slope and Longitude), were consistent across the methods, although there were some slight variation in the ranking of variables (Table 3.5). The total variance explained by all 47 measured variables was 62.6 % (454), 51% (ARISA) and 59% (T-RFLP).

Environmental correlations with the PCO1 axis co-ordinates were calculated individually, while the sequential tests in DistLM generated a parsimonious subset of variables that best explains proportions of the variation. DistLM placed less emphasis on longitude, moisture, mud, NO3, Na2O, K2O and other grain size measurements suggesting a level of co-variance with other variables and more emphasis on elevation, PO4, aspect, slope and carbon.

Table 3.5 The consistent environmental variables obtained from the

distance based redundancy models (as calculated in a Sequential

DistLM).

454 sequencing ARISA T-RFLP Adj Pseudo- Prop. Adj Pseudo- Prop. Adj Pseudo- Prop. Variable R^2 F % R^2 F % R^2 F % Grain size (min) 0.14 38.13 14.70 0.10 24.93 10.50 0.09 23.00 9.40 pH 0.21 19.92 7.10 0.19 7.97 3.00 0.13 12.07 4.70 Latitude 0.26 14.70 4.90 0.17 8.05 3.10 0.26 12.78 4.20 TiO2 0.29 12.40 3.90 0.25 5.24 1.80 0.29 10.49 3.30 MnO 0.33 11.96 3.60 0.24 5.79 2.10 0.32 7.87 2.40 Elevation 0.35 9.32 2.70 0.30 3.86 1.30 0.22 14.54 5.10 PO4 0.37 6.89 2.00 0.26 4.61 1.60 0.35 6.67 1.60 Cl 0.39 6.41 1.80 0.32 2.52 0.80 0.37 6.21 1.80 Carbon 0.41 5.11 1.40 0.27 4.14 1.40 0.40 4.17 1.10 Conductivity 0.42 4.73 1.20 0.33 2.33 0.70 0.48 1.32 0.35 longitude 0.44 3.30 0.08 0.34 2.14 0.65 0.17 11.88 4.40 Aspect 0.43 3.61 0.90 0.29 3.97 1.30 0.34 7.64 2.30 Slope 0.44 4.65 0.12 0.30 3.85 1.30 0.38 5.37 1.50 Total variationa 62.60 51.00 59.00 All environmental variables listed were significant (α = 0.05). Variables explaining more than 1 % of the proportional biological variation are listed in bold. a Total cumulative variation explained for the best DistLM model generated.

Figure 3.6 Correlation of the environmental parameters with the biological datasets for 454 pyrosequencing, ARISA, and T-RFLP datasets. Correlation values were calculated against the PCO1 axis with

Pearson’s correlation.

3.3.6 Modified data matrices.

The reduction of the data to a binary form did little to alter the MANTEL and ANOSIM results (Table 3.2, Table 3.4, Table 3.6, Table 3.7). Sub-sampling the pyrosequencing data set to species contributing >1 %, >10%, >15% and >20% showed that the major biological patterns were defined by very few, highly abundant species (Table 3.6, Table 3.7). Retaining only those OTUs contributing more than 15% of the total abundance (17 OTUs) led to a deterioration of the patterns and the ordinations no longer bore any resemblance to the original data. However, the Mantel tests and ANOSIMs still produced similar patterns to the full data sets. With only those OTUs responsible for more than 20% of the total abundance

(10 OTUs), the patterns had completely broken down and differentiation between a priori groups was no longer possible.

Table 3.6 Modified MANTEL tests results (testing the similarities of the

modified data matrices)

Modified Matrices Square root P/Aa transformed transformed Rho value p Rho p value TRFLP TRFLP in silico v 454 0.79 < 0.001 0.50 < 0.001 TRFLP in silico v TRFLP exp.b 0.46 < 0.001 0.30 < 0.001 454 454 sub-sampled to sp.c > 1% 0.98 < 0.001 0.97 < 0.001 (558)g 454 sub-sampled to sp.d > 10% 0.83 < 0.001 0.80 < 0.001 (41) g 454 sub-sampled to sp.e > 15% 0.73 < 0.001 0.56 < 0.001 (17) g 454 sub-sampled to sp.f >20% 0.29 < 0.001 0.36 < 0.001 (10) g a Presence / Absence, bT-RFLP Experimental, csub-sampled to species contributing greater than 1% of total abundance, d >10% of total abundance, e >15% of total abundance and f >20% of total abundance, g number of remaining OTUs

Table 3.7 Modified data ANOSIM test results (testing the ability to

distinguish between a priori groups)

Modifie Matrices transformed transforme Pair-wise results d data (Sqr rt) d data (P/A) Similarity between sites R p R p Site separation TRFLP TRFLP in silico 0.72 < 0.001 0.52 <0.001 28/28 (0.17-0.92) 454 454 sub-s to sp.a > 1% 0.86 < 0.001 0.87 <0.001 28/28 (0.41-1.0) (558)e 454 sub-s. to sp. b > 10% 0.84 < 0.001 0.80 <0.001 28/28 (0.24-1.0) (41) e 454 sub-s. to sp. c > 15% 0.70 < 0.001 0.44 <0.001 27/28 (17) e (0.074-0.96) 454 sub-s. to sp. d > 20% 0.12 0.004 0.11 0.004 21/28 (10) e (-0.08- 0.91) Similarity b/w dist. groups R p R P Group separation TRFLP TRFLP in silico 0.49 < 0.001 0.40 <0.001 26/45 (-0.175-0.94) 454 454 sub-s to sp. a > 1% 0.58 < 0.001 0.56 <0.001 29/45 (558) e (-0.132-1.0) 454 sub-s. to sp.b > 10% 0.42 < 0.001 0.40 <0.001 22/45 (41) e (-0.185-0.97) 454 sub-s. to sp. c > 15% 0.34 < 0.001 0.29 <0.001 9/45 (17) e (-0.22-0.46) 454 sub-s. to sp. d > 20% 0.34 < 0.001 0.34 <0.001 20/45 (10) e (-0.125-0.97) a sub-sampled to species contributing greater than 1% of total abundance, b >10% of total abundance, c>15% of total abundance and d>20% of total abundance, e number of remaining OTUs.

Figure 3.7 Comparison of the PCO pots generated from the reduced 454 pyrosequencing matrices of OTUs contributing >1%, >10% and > 15% of the total abundance. The first three PCO plots on the left represent all samples, separated by sites. The following three plots on right represent all the samples within Mitchell Peninsula site, separated by a priori distance groups.

3.3.7 In Silico verses experimental T-RFLP.

The in silico T-RFLP results varied from the actual experimental T-RFLP results, yet the two were still significantly correlated (Rho = 0.46, p < 0.001) (Table 3.6). As expected, the in silico T-RFLP results bore a much closer resemblance to the 454 data (Rho= 0.79, p < 0.001) and had a similar capacity to distinguish between groups as the modified 454 matrix reduced to species contributing >10% or more. The numbers of OTUs from the 454 pyrosequencing data represented by each fragment size ranged from 1 to 1512, with an average of 100 OTUs per fragment. A total of 182 individual fragment sizes were generated, ranging from 20 to

308 bp. In comparison, the final experimental T-RFLP and ARISA generated 350 and 661 fragment sizes ranging from 150-500 bp and 350-1500 bp respectively.

3.4 DISCUSSION

Currently, 454 pyrosequencing is the most widely used NGS method used for characterising microbial communities (153, 154). We found that ARISA and T-RFLP displayed comparable sensitivity to 454 pyrosequencing, exhibiting the ability to differentiate between polar sites at a global scale and metres to 100’s of metres at a local scale. The fingerprinting methods were also highly reproducible as seen by the strong biological correlation observed between the technical replicates (Figure 3.2). In comparison with the NGS approach, fingerprinting was a cheap and rapid alternative, ideal for ensuring appropriate replication in large microbial ecology studies.

The fingerprinting methods generated approximately 10% of the OTU numbers recovered with pyrosequencing. We demonstrated here that the saturation point of the analysis at 100+ fragments per sample limited the ‘dynamic range’ (155) which restricted the upper ranges of the richness gradient. Thus, preventing any correlation between the fingerprinting and

72 pyrosequencing generated richness estimates (Figure 3.4). While previous investigations have reported higher richness and diversity estimates with ARISA over T-RFLP (142), we found here that the two fingerprinting methods generated variable yet statistically equivalent diversity estimates with no correlation to the pyrosequencing results, or to each other (156).

In line with previous investigators (134, 140), we have confirmed that diversity estimates from fingerprinting methods likely severely underestimate alpha diversity and thus are inappropriate for comparisons across microbial ecology investigations.

Despite less resolution at the global scale, the fingerprinting methods had a similar capacity to capture significant biological patterns and correlate environmental variables with biological distributions as the pyrosequencing data. By progressively reducing the pyrosequencing data to the most abundant sequences we found that the observed biological patterns were driven by a very small proportion of the most abundant species, with the major biological patterns disintegrating only after all but 10-17 of the original 17 000 OTUs remained (~0.1%). This was further supported by the fact that binary transformation (up weighting the importance of rare species) had no real effect on the observed biological patterns, re-enforcing the practical potential of molecular fingerprinting methods to capture major biological patterns. It is important to note that we did not saturate the rarefaction coverage curves, and that if we did rarer OTU’s may contribute more to general patterns. Yet, even very large sequencing efforts often fail in this regard (157) and will likely continue to fail to fully sample the community in the near future. Additionally, it is important to note that distance based analyses such as Bray-Curtis are dominated by the most variable species or

OTUs. It has been reported that as mean abundances increase, variance also tends to increase.

Therefore, the resulting resemblances even in the full data set may be disproportionally

73 driven by abundant organisms, despite the transformations, as a result of an underlying mean- variance relationship (158).

Fingerprinting methods are capable of detecting significant spatial (as seen here), temporal

(140) and treatment (141) shifts in communities similarly to pyrosequencing data, suggesting full scale sequencing is not necessary to monitor community structural changes (159). We conclude that genotypic fingerprinting provides an independent and legitimate molecular tool for the generation of large, statistically robust biological patterns (160). Additionally it offers a rapid, relatively inexpensive option to judiciously select appropriate samples for further phylogenetic analysis, based on statistically significant pattern information (161). We suggest a combination of community fingerprinting and sequencing could achieve a cost effective and robust method to analyse large numbers of samples in multivariate ecotoxicology studies.

4 BACTERIAL TARGETS AS POTENTIAL INDICATORS

OF DIESEL FUEL TOXICITY ON SUBANTARCTIC SOILS

4.1 INTRODUCTION

Toxicity information relating to the effects of petroleum hydrocarbon contamination on terrestrial ecosystems in polar and sub-polar regions is limited (20), yet available evidence suggests that the effects of oil spills are more damaging than in temperate regions due to low temperatures and slower ecosystem recovery (1, 2, 162). Currently no contamination or remediation guidelines exist in Antarctica and existing remediation trigger values throughout other polar and sub-polar regions (in particular the Arctic) are largely based upon countries domestic guidelines for diesel fuel. The result is highly variable values, ranging between 100 mg kg-1 and 2000 mg kg-1, with no evidence based guidance for site-specific modifications

(54).

Single species tolerance testing, used commonly for ecotoxicological assessments, is resource and time intensive. The confidence surrounding five to ten model organisms accurately representing the sensitivity of an ecosystem is also questionable (59, 60). Further, the selection criterion for model organisms has resulted in a bias towards temperate and northern hemisphere species, to the exclusion of rare and often sensitive species (60). In the Antarctic and subantarctic regions, traditional toxicology indicator species such as earthworms and large invertebrates are sparse or non-existent, further compounding the uncertainty surrounding the application of traditional tolerance testing ‘model’ organisms. In 2008,

Hickey et al. suggested that new, rapid tolerance testing that targets a suite of organisms across many taxa were required to increase the confidence in resulting toxicity estimates.

Rapid tolerance testing has been broadly defined as toxicity testing that aims to reduce the

75 time and resources per species, by reducing replication, the number of treatments and testing species concurrently. Rather than small numbers of precise estimates, rapid tests are expected to promote greater numbers of species assessed more approximately, with increased relevance to the ecosystem, thereby delivering more accurate estimates of community sensitivity overall.

Microorganisms are ideal targets for monitoring soil health due to their integral roles in biogeochemical cycles and ecosystem sustainability (17, 65, 66). Yet key challenges exist for developing microbial indicators, including the high variability of microbial communities observed across temporal and spatial scales. Many studies have linked this microbial variation to changes in the physical and chemical properties of soil with factors such as carbon content and pH particularly important (86, 87). For the subantarctic, bacterial and alkane-degrading diversity has already been correlated with organic carbon and total nitrogen availability (163). However, it is unclear how variability in soil types and the resulting microbial community composition will affect the resilience of soil bacterial communities when exposed to petroleum hydrocarbons.

Another key challenge for implementing microbial monitoring specific to microbial monitoring of petroleum hydrocarbons is the concurrent stimulation and inhibition effect of petroleum hydrocarbons on bacteria, which complicates toxicity assessments. To counter this, key nutrient-cycling parameters can be targeted that are sensitive to petroleum hydrocarbon toxicity but not stimulated by increases in available carbon (17). For example, the chemolithic process of ammonia oxidation is not affected by increases in organic carbon, yet it is inhibited by hydrocarbons both through the general mechanism of biochemical toxicity

(non-polar narcosis), and through specific inhibition of ammonium monooxygenase (164,

165). This process can be monitored by targeting the bacterial and archaeal gene for ammonia

76 monooxygenase, amoA. As bacteria and archaea are primarily responsible for the conversion of ammonia to nitrite, toxicity or inhibition of amoA suggests overall ecosystem impairment.

Similarly, functional genes including nifH and nosZ that are responsible for key enzymatic steps within the nitrogen cycle have been targeted to assess functional aspects of soil microbial communities (78, 166).

Our aim was to identify the microbial indicators of fresh diesel fuel toxicity in Macquarie

Island soils using next generation sequencing and qPCR. The effects of SAB on the bacterial community were evaluated with broad and targeted community indices, as well as the abundances of functional genes encoding key enzymes within the nitrogen cycle. We utilised the resulting dose-response curves to further test the utility of microbial targets as toxicity indicators. We hypothesised that the bacterial community would change with increasing fuel contamination, with a loss of diversity and a selection towards species capable of hydrocarbon degradation. We further hypothesised that soil parameters would influence the susceptibility of the soils to petroleum hydrocarbon toxicity, and that the targeted portions of the bacterial community would provide more sensitive indicators than community-wide diversity estimates.

4.2 METHODS

4.2.1 Soil sampling and diesel fuel spiking

Four uncontaminated bulk soil plots consisting of sandy to peaty soil were selected from the isthmus, approximately 100-150 metres from the major contaminated site (54° 37’ 53” S,

158° 52’ 15” E). To target the most active microbial zone and to correspond with related invertebrate studies (62), 500 g of soil was collected to a depth of 30 cm from each plot (a total of 2 kg). Each of the four bulk soils were homogenised by mixing, separated into ten

77 individual subsamples of 50 g, and placed in 100 ml amber jars. One control sample for each of the four bulk soils was left unamended. The remaining soils were spiked with SAB diesel fuel to three target concentration ranges; low (0 - 400 mg kg-1, n=3), medium (401 - 5000 mg kg-1, n=3) and high (5000 - 20 000 mg kg-1, n=3) (Table 4.2). Bulk soils were chosen to reduce the variability of soil properties within plots and to have control over the range of concentrations. All soils were incubated aerobically, in the dark. Three of the bulk soils (P1-

P3) were incubated for 21 days (short exposure), while the remaining bulk soil was incubated for 18 months (long exposure; P4). Whilst no comment can be made on the short-term response of P4, the extended exposure was set up to determine if microbial community responses to the diesel fuel were observed between both short and long-term samples (29).

For the extended exposure, the lids of the amber jars were opened every four weeks to maintain aerobic conditions. Although not highly aerobic these conditions will provide aerobic and anaerobic pockets within the jars, consistent with the soil matrix in-situ.

Nutrients and soil parameters were analysed after the incubation period at the AAD according to 3.2.2 (Table 4.2). Briefly, the total carbon was determined by loss-on-ignition and expressed as a percentage. The concentrations of anions and cations in the soil were measured on a water extract (1 g: 5 ml water) and expressed as mg kg-1. Conductivity and pH were also measured on the same water extract. A 10 g subsample of soil from each sample was extracted with hexane and assessed by gas chromatography to determine the total petroleum hydrocarbons (TPH) concentration (17). The detection limit (DL) for TPH concentrations was 20 mg kg-1. TPH concentrations below the detection limit were estimated with a substitution method based on half the detection limit, i.e. estimated TPH concentrations of 10 mg kg-1 (Table 4.2). Average measured TPH concentrations were calculated for each of the fuel categories within the four plots (Table 4.2, Figure 4.2). Dry

78 matter fraction was determined gravimetrically using the same samples analysed for TPH.

We have reported all nutrient and TPH concentrations on a dry matter basis unless stated otherwise.

4.2.2 Barcoded amplicon pyrosequencing targeting the 16S rRNA gene.

After the short (21 days) and long (18 months) incubation periods, total soil gDNA was extracted from 50 mg subsamples of each soil sample in triplicate with the FastDNA® SPIN

Kit for soil according to (2.2.10). In total, 120 DNA extracts (40 samples in triplicate) were titrated to a standard working concentration range of 5-10 ng µl-1. The technical replicates were analysed with ARISA to evaluate inter/intra sample similarity according to 3.2.4. Inter- sample replication was evaluated with ANOSIM in PRIMER v6. After inter-sample similarity was confirmed (Figure 4.1,Table 4.1), the extracted gDNA from a randomly selected replicate was used as a template for barcoded tag pyrosequencing.

The V1, V2 and V3 hypervariable regions of the small-subunit (SSU) ribosomal gene were targeted using the 27F and 519R universal primers according to 2.2.11. Although systematic errors remain with barcoding sequencing technologies, stochastic replication (3X) was performed to limit some of the PCR bias. The sff files, provided from the sequencing facility were processed with the software and conditions outlined in 2.2.11. An OTU abundance-by- sample matrix was generated from the resulting output, and various diversity indices were calculated. A neighbour-joining tree was created through MOTHUR (114) with the Clearcut program addition (168). The neighbour-joining tree was used to run the weighted UniFrac algorithm (169).

Table 4.1 Analysis of similarity (ANOSIM) results testing for significant

differences between fuel categories based on the ARISA and

pyrosequencing results.

Differences between fuel groups Global R p value

ARISA 0.397 0.001 Pyrosequencing 0.691 0.004

Figure 4.1 Ordination plots of samples from P2 based on ARISA

fragments (A) and pyrosequencing data (B). After Analysis of similarity

(ANOSIM) testing of the ARISA results, three replicates from each fuel group

were randomly selected for pyrosequencing.

Table 4.2 Summary of measured physicochemical parameters.

P1 (Low C) P2 (Med C) P3 (High C) P4 (Long exposure) Con.a Lowb Med.c Highd Con.a Lowb Med.c Highd Con.a Lowb Med.c Highd Con.a Lowb Med.c Highd (n=3)* (n=3) (n=3) (n=3) (n=3) (n=3) (n=3) (n=3)* (n=3) (n=3) (n=3) (n=3)* TPH (avg.)e

5000 mg kg-1), e (mg kg-1), f (% LOI) g Dry matter fraction, h (µS/cm) * One sample in this group had substantially less sequencing reads than the other samples; as such they were excluded from further phylogenetic analysis.

4.2.3 Multivariate data analysis.

Multivariate data analysis was conducted with the software packages PRIMER v6 and

Permanova according to (2.2.12). The resulting OTU/abundance matrixes were converted into resemblance matrices, calculated with both the Bray-Curtis similarity coefficient and the weighted UniFrac measurements. To test the null hypothesis of no differences between plots,

ANOSIM was performed on the resemblance matrices with 999 permutations. To test the null hypothesis of no differences between diesel fuel categories, a one-way analysis of variation

(ANOVA) was used to test for significant differences in the microbial communities between fuel categories within each plot.

4.2.4 Correlation of environmental variables

To evaluate the influence of environmental variables, the measurements from the chemical and physical analysis of the soils were transformed in PRIMER according to (3.2.8), generating an environmental data matrix. The relationship of the environmental variables to the biological distribution was analysed with a DistLM in PERMANOVA+ according to

(3.2.8). The DistLM was based on the null hypothesis of no relationship between the environmental variables and the biological resemblance matrix. All statistical tests were considered significant at P <0.05.

4.2.5 Dose-response of individual OTUs, genera and phyla across the TPH range

Individual OTUs were evaluated across the SAB spiking range to determine if they were significantly inhibited, or stimulated with increasing TPH concentration. After OTUs present in only one sample were removed, the log-transformed abundance of individual OTUs was plotted against the log-transformed TPH data. The resulting dose-responses of each OTU were then fitted to a linear equation in ‘R’ (http://www.R-project.org/). The number of OTUs

82 significantly stimulated or inhibited with TPH was determined based on the quality of fit (p

<0.05) and slope of the line. The percentage abundance of each OTU was aggregated into genera to create a heatmap within ‘R’. Representative OTUs that were unable to be reliably classified to the genera level were listed with the closest classification level. The OTUs were then aggregated into phylum level phylogeny. All phyla were analysed for positive or negative correlations to increasing fuel concentrations. The ratio of

Acidobacteria/Proteobacteri, considered representative of the ratio of oligotrophs/copitrophs in the environment, was calculated and used as an additional community index (170).

4.2.6 Targeting the nitrogen cycle with quantitative PCR.

Quantitative PCR (qPCR) was used to measure the abundance of the nifH, amoA and nosZ genes present. The same DNA extracts used in the barcoded pyrosequencing analysis were utilised for the qPCR analysis and each of the selected genomic DNA extracts were analysed by qPCR in triplicate. Samples were analysed using the QuantiTect Fast SYBR Green PCR

Master Mix real-time PCR kit (Qiagen, Doncaster, VIC, Australia) run on the ABI 7500 real- time PCR machine (Applied Biosystems). Each 20 µl reaction contained 12.5 µl of master mix, 1.25 µl of template DNA and 1 µM of the forward and reverse primers. The thermal cycling program consisted of 94 °C for 5 min, 45 cycles of 94 °C for 20 sec, 54 °C for 50 sec, followed by a melt-curve step from 50 °C to 95 °C. The quantitative fluorescence data were collected during the 54 °C step. To test for the potential presence of PCR inhibitors, a four to five-point curve (in duplicate) of different DNA concentrations for each soil sample was analysed according to Ma et al. 2008. At the DNA concentrations used, no inhibition was detected. Quantification data were collected only when; there was no detected PCR inhibition, no amplification in the ‘no template control’, a single peak in the melt curve consistent with specific amplification, and a reaction efficiency of 100 ± 10%.

Subunits of the nitrogenase enzyme are encoded by the genes nifH, nifD and nifK. The nifH gene is the most widely sequenced and utilised marker for nitrogen fixation. Here, we used the primers IGK3 and DVV (172) to target the nifH gene as a proxy for potential nitrogen- fixation activity (Table 4.3). A standard curve was generated with the IGK3/DVV amplified

PCR product from a control subantarctic soil. The number of copies was determined spectrophotometrically. The standard curves were linear over six orders of magnitude with amplification efficiencies of 93.8 - 95.1% and an R2 value of 0.997.

Table 4.3 Details of primers used in quantitative PCR (qPCR) to target

functional genes within the nitrogen cycle.

Process Target Primers Ref.a Primer sequence Positive gene control Nitrogen nifH IGK3 (172) GCIWTHTAYGGIAAR gDNA from fixation GGIATHGGIA soil DVV ATIGCRAAICCICCRC AIACIACRTC Ammonium B.amoAb amoA1F (173) GGGGTTTCTACTGGT gDNA from oxidation GGT soil amoA2R CCCCTCKGSAAAGCC TTCTTC A.amoAc A.amoAF (174) STAATGGTCTGGCTT gDNA from AGACG soil A.amoA GCGGCCATCCATCTG R TATGT Denitrification nosZ nosZ2F (175) CGCRACGGCAASAA Pseudomonas GGTSMSSGT stutzeri nosZ2R CAKRTGCAKSGCRTG GCAGAA a Reference. bBacterial amoA, cArchaeal amoA

Nitrification refers to the oxidation of ammonia into nitrite and the subsequent oxidation of nitrite into nitrate. As PCR primers targeting the entire nitrite oxidation functional group are not currently available, the ammonium oxidising step was used as a proxy for potential nitrifying activity. The ammonium oxidising step can be performed by ammonium oxidising

84 bacteria (AOB) and ammonium oxidising archaea (AOA). We evaluated both groups with primer sets targeting the AOB (amoA1F/amoA2R) (173) and AOA

(Arch.amoAF/Arch.amoAR) (174) amoA genes (Table 4.3). Standard curves were generated with the amplified PCR products from a control subantarctic soil, with the number of copies determined spectophotometrically. For bacterial amoA, the standard curve was linear over seven orders of magnitude with amplification efficiencies of 91.6 - 92.0%, r2 = 0.998. The archaeal amoA standard curve was linear over six orders of magnitude with amplification efficiencies of 90.6 – 94.8%, R2 = 0.974.

Classified genes known to be present in denitrifying microorganisms include nir (nitrate reductase) and nos (nitrous oxide reductase). We chose to target the nosZ gene as a proxy for potential denitrification with the primer set nosZ2-F/nosZ2-R (175). These primers primarily target Proteobacteria, excluding denitrifiers within Firmicutes. Although not present in all denitrifying species, the nosZ gene has been widely evaluated and applied in polar soils due to its relevance in nitrous oxide production (a potent greenhouse gas). Further, in Siciliano et al. (2007) the applicability of quantifying nosZ abundance from mixed templates was specifically tested with positive results (176). In another study nosZ was more sensitive to environmental changes than nirS and nirK (177). To determine the copy numbers of the nosZ gene, a standard curve was generated using the nosZ2-F/nosZ2-R (175) amplified PCR product from Pseudomonas stutzeri - a well-known denitrifying species. The number of copies of the nosZ gene was determined spectrophotometrically (Table 4.3). The standard curve was linear over five orders of magnitude; the amplification efficiency determined from the slope of the standard curve was 99.8%, R2 0.96.

4.2.7 Dose-response modelling.

For dose-response modelling, a large number of data points are recommended over high replication to best capture possible shifts in the curve (178, 179). Hence we used individual samples and their TPH concentrations instead of the concentration ranges to maximise the confidence surrounding the dose-response curves as in 2.2.13. The dose-response measurements were calculated as a percentage of the control to enable comparison between soil types (P1-P4). The dose response curves were generated within the ‘drc’ ‘R’ package

(180). This included plotting the data and selecting the most suitable model as evidenced by

Akaike’s information criterion (AIC) and ANOVA comparisons of models. Effective concentration values (ECx) (including standard error and confidence intervals) were then calculated from the curve generated by the most suitable model for each data set. The resulting EC20 values (concentrations that effect 20% of the population) calculated from the dose-response curves were used to compare the relative sensitivities of soil types and community measures.

4.3 RESULTS

4.3.1 Bacterial community composition

A total of 127,053 quality-checked sequence reads were obtained averaging 3304 (± 1626) sequence reads per sample. For further comparative analysis the sequences were sub-sampled at 1,450 sequences resulting in approximately 14,000 per plot and 4,350 sequences per SAB treatment. Three samples (within P1-low, P3-med and P4-high) had exceptionally low numbers of sequence reads at <800 and were excluded from further analysis. The distribution of the bacterial communities was used to create two similarity matrices, based on the Bray-

Curtis correlation coefficient and UniFrac distance. Utilising the weighted UniFrac similarity matrix, a one-way ANOVA confirmed significant differences between bacterial communities

86 according to their TPH concentration ranges; control, low (10-400 mg kg-1), medium (401-

5000 mg kg-1) and high (>5000 mg kg-1). Significant dissimilarity between the control and the treatments was observed in all four soil plots (Figure 4.2B, Table 4.4). The soil plots, P1-

P4, were also found to be significantly different in a one-way ANOSIM global R value =

0.73 (p <0.001) (Table 4.4).

Figure 4.2 The average measured TPH concentration and bacterial

community similarity of samples across SAB spiking fuel concentration

ranges and soil plots. (A) The average measured TPH log concentration for

the spiking fuel ranges within each soil plot. The dashed line indicates the

TPH detection limit (DL). (B) The average community similarity within each

plot based on the weighted UniFrac distance. Error bars account for the

standard deviation between biological replicates. *, significant decline in

community similarity was found (P<0.05)

Table 4.4 ANOVA and ANOSIM results testing the amplicon

pyrosequencing community UniFrac and bray-curtis dissimilarities

between fuel categories and soil plots.

Test Pairwise tests t/R statistic p value P1 global 17.4 0.006 Differences between fuel categories control, high 8.0 0.009 (1-way ANOVA) control, med 5.5 0.026 control, low 4.2 0.032 low, high 5.5 0.027 low, med 1.9 0.241 med, high 3.8 0.044 P2 global 57.9 0.001 Differences between fuel categories control, high 12.8 <0.001 (1-way ANOVA) control, med 9.7 0.001 control, low 4.7 0.016 low, high 11.4 <0.001 low, med 7.1 0.003 med, high 4.4 0.021 P3 global 213.6 <0.001 Differences between fuel categories control, high 28 <0.001 (1-way ANOVA) control, med 19.4 <0.001 control, low 15.6 <0.001 low, high 17.4 <0.001 low, med 6.2 0.007 med, high 9.4 0.001 P4 global 278.5 <0.001 Differences between fuel categories control, high 28.6 <0.001 (1-way ANOVA) control, med 13.1 <0.001 control, low 8.3 <0.001 low, high 28 <0.001 low, med 6.6 0.005 med, high 23.6 <0.001 Differences between plots a global 0.729 <0.001 (1-way ANOSIM) P1, P2 0.722 <0.001 P1, P3 0.807 <0.001 P1, P4 0.765 <0.001 P2, P3 0.781 <0.001 P2, P4 0.781 <0.001 P3, P4 0.767 0.002 a plots include P1 = low carbon soil, P2 = medium carbon soil, P3 = high carbon soil and P4 = aged soil. b Fuel categories refer to TPH concentrations; unamended control, low TPH concentration (range DL-400 mg kg-1), medium TPH concentration (range 401-5000 mg kg-1) and high TPH concentration (>5000 mg kg-1) c Significant differences are in bold, p was considered significant at <0.05.

Diesel fuel substantially reduced the similarity of communities from the controls largely by stimulating or reducing the relative abundances of key lineages rather than removing entire lineages of bacteria (Figure 4.3). The communities shifted from large numbers of species in relatively low abundance, to communities heavily dominated by only a few species. We found that the OTUs consisting of 26% to 79% of the relative abundance were significantly inhibited across the SAB spiking range, while only a small proportion of OTUs, 0.4% to

7.8%, were stimulated (Table 4.5).

Table 4.5 Operational Taxonomic Units (OTUs) significantly (p <0.05)

stimulated and inhibited across the SAB diesel fuel spiking range.

Inhibited OTUs Stimulated OTUs soil total number relative relative number relative Relative plot number OTUs abundance abundance OTUs abundance abundance unique inhibited in control in high Stimulated in control in high OTUs soils soils (SE) soils soils (SE) P1 386 28 39 6.5 (1.95) 7 0.4 68.0 (10.4) P2 723 178 56 5.3 (0.8) 8 7.8 73.8 (7.7) P3 712 77 26 2.5 (3.1) 11 0.6 51.5 (1.3) P4 715 119 79 15.9 (0.3) 14 0.8 52.5 (1.4)

The genus most stimulated was Pseudomonas, contributing a maximum of 5.5% relative abundance in the control soils compared to >60% relative abundance in the highest concentration samples for the low (P1) and med (P2) carbon soils and >20% in the long exposure spiked soils (P4) (Figure 4.3). By comparison, the high organic carbon soil (P3), exhibited a lower relative abundance of Pseudomonas (<1%) and maintained a relative abundance between 0.06% and 4% across the entire spiking range. Instead, the genus

Parvicubaculum, which contributed 0.2% of the total abundance in the control soils, increased substantially, contributing to 43% of the final relative abundance (Figure 4.3).

Bacteria involved in the nitrification process, specifically ammonium oxidation or the oxidation of nitrite to nitrate, were significantly inhibited following the addition of diesel fuel. A functional group for the nitrification species ‘All nitrification species.’ was consolidated consisting of Nitrosospira sp., Nitrosococcus sp., Nitrobacter sp., Nitrosomonas sp., Nitrospira sp., and unclassified species from the Family Nitrosomonadaceae and the

Order Nitrosomonadales. The group was significantly inhibited across the diesel fuel concentration range for each soil plot, (p<0.05) (Figure 4.3).

Figure 4.3 Relative abundances of genera present in soil samples from low to high special Antarctic blend (SAB) diesel fuel concentration.

Samples are sorted according to SAB fuel concentration within each soil plot and are labelled according to fuel category (control, low, medium and high).

The log SAB fuel concentration is on a scale of 0-4.5, with 4.8 ~ 31, 000 mg kg-1. Representative operational taxonomic units (OTUs) that were unable to be reliably classified to the genus level were listed with the closest classification level.

4.3.2 Environmental predictors.

Carbon was a significant predictor variable (F = 5.13, p = 0.001) when analysed with a marginal DistLM model (Table 4.6). However, in sequential DistLM tests organic carbon contributed only 4.1% to the total biological variation observed between soil samples. The environmental variables of pH, nitrate, phosphate and TPH all had a greater influence on the biological distribution than organic carbon. The soil pH was the greatest predictor value, accounting for 18% of the total variation between all samples as calculated with Pearson’s correlation (p <0.001) (Table 4.6).

Table 4.6 Distance Linear Model (DistLM) results indicating strength of

environmental variable as a predictor of the biological distribution and

patterns of bacterial communities across all samples.

Marginal testsa Sequential testsb Variable Pseudo-F P c Prop.d Pseudo-F P c Prop. d Cumul. e pH 7.28 <0.001 0.18 7.28 <0.001 0.18 0.18 Nitratef 5.09 <0.001 0.13 5.73 <0.001 0.12 0.30 Phosphatef 4.30 <0.001 0.12 6.45 <0.001 0.12 0.42 TPH f 2.28 0.019 0.06 3.99 <0.001 0.07 0.49 Carbong 5.13 <0.001 0.13 2.58 0.006 0.04 0.53 Nitritef 4.44 <0.001 0.12 2.22 0.013 0.03 0.57 Brominef 2.43 0.008 0.06 1.52 0.116 0.02 0.59 Sulphatef 2.77 0.005 0.08 1.69 0.079 0.02 0.62 Conductivityh 4.06 0.002 0.11 1.58 0.110 0.02 0.64 Chlorinef 5.33 <0.001 0.14 1.46 0.140 0.02 0.66 Ammoniumf 1.30 0.137 0.04 1.18 0.289 0.02 0.68 Dry matter fraction 5.40 <0.001 0.14 1.04 0.377 0.02 0.69 a Marginal test results are based on individual variables. b Sequential test results are based on the relative proportion of influence when all variables are considered. c Significant differences are in bold, p was considered significant at <0.05. d The proportion of variance predicted with environmental variable. e The cumulative proportion of variance explained. f (mg kg-1), g (% LOI) h (µS/cm)

4.3.3 Community indices.

With increasing diesel fuel concentration the total species richness, species diversity, species evenness, similarity indices and UniFrac measurements across most soils declined from the controls (Figure 4.4). The control communities within the low carbon soils (P1) consisted of substantially lower species numbers, diversity and evenness than the higher carbon soils, and as a result the effect of the diesel fuel on the communities was less severe. The species richness within the medium carbon soils (P2) was sensitive to increasing diesel fuel with

ACE, Chao1 and Sobs values decreasing by up to 20%. The Sobs in the higher carbon soils (P3) declined almost 50% from the control while ACE and Chao values were variable across all fuel concentrations. In the long exposure soils (P4), the loss of species richness from the control occurred at higher concentrations than in the acute soils for Sobs, ACE and Chao with a total decrease of 60% observed.

The Shannon (H’) diversity index, Simpson diversity index and Pielou’s evenness index (J’) in the low carbon soils (P1) were not significantly impacted by SAB concentrations below

5000 mg kg-1(Figure 4.4). In the medium carbon (P2) soils a gradual decline in the Shannon

(H’) diversity index was observed from 100 mg kg-1 resulting in a 70% decrease from the control. This decrease was less severe in the higher carbon and aged soils (P3 and P4, respectively) with a 40% decrease from the control observed. The Pielou’s evenness index

(J’) in high carbon soils (P3) decreased by approximately 20% from the control, with the decline detectable at concentrations below 100 mg kg-1. The P4 soil exposed to the contaminant for 18 months exhibited different sensitivity compared with the short exposure soils, with the evenness index particularly inhibited.

Figure 4.4 The effect of increasing SAB fuel concentrations on richness, diversity, phylogenetic indices and key functional genes within the nitrogen cycle. (nifH = nitrogen fixation, amoA = ammonium oxidation, nosZ = denitrification). Values are expressed as percentage change from the control. The best fitting dose-response regression model was used to calculate

EC20 values.

The UniFrac dissimilarity measurement provided a sensitive target for all soils (Figure 4.4,

Table 4.7). The phylogenetic distance measurement changed between 40% - 60% from the control soils for all four samples. The long exposure soils (P4) exhibited a higher level of similarity to the control soil and to soils spiked with <1000 mg kg-1 fuel, suggesting a higher tolerance or recovery in the genetic potential of the community than short exposure soils. The ratio of Acidobacteria:Proteobacteria, which is indicative of oligotrophic:copiotrophic species, was reduced, suggesting a shift towards faster opportunistic species within γ- /β-

Proteobacteria. The low and medium carbon soils and long exposure soils were sensitive to this ratio with 75% decreases observed compared to the control soils. The high carbon soils were not sensitive to this measurement with very high EC20 values generated.

4.3.4 Abundance of functional genes.

Of the functional genes evaluated, bacterial amoA (indicative of nitrification) was the most sensitive indicator of hydrocarbon toxicity in soil (Figure 4.3). The low carbon and aged soils

(P1 and P4, respectively) were most sensitive to the bacterial amoA measurement with a significant decline in copy numbers observed at very low concentrations. A decline of bacterial amoA copy numbers in the medium and high carbon soils (P2 and P3, respectively) was also observed for SAB concentrations >100 mg kg-1. The archaeal amoA gene was present in low to non-detectable levels in all but one of the soils (P2). Within P2, the abundance of archaeal amoA was an order of magnitude lower than the bacterial amoA gene ranging from 5.02 x 102 to 4.07 x 101 with abundances declining with increasing SAB concentrations. As such, only the bacterial amoA was pursued as an indicator and archaeal amoA was excluded from further analysis. The abundance of the nifH gene (representative of nitrogen fixation), was variable across all soils and diesel fuel concentrations analysed, thus no significant or ‘measurable’ impact was observed. The nosZ gene abundance (indicative of

95 denitrification) was stimulated in all soils spiked with diesel fuel. While the greatest overall increases were observed in the low and medium carbon soils, similar increases in the nosZ gene copy numbers occurred at low TPH concentrations for all soils and the low carbon and long-term exposure soils generated similar sensitive dose-response curves between 100-1000 mg kg-1.

4.3.5 Dose response modelling.

Within the low carbon soil (P1) the EC20 values, based on percentage change from the control, generated high values outside the experimental spiking range for the Simpson index, and no significant response was measurable for ACE, Chao1 and Sobs. Overall, the calculations of EC20s across the traditional community diversity measures including Sobs,

ACE, Chao, Shannon (H’), Simpson, and Pielou’s evenness (J’), generated results with high associated errors and large variability between soil types (Figure 4.4, Figure 4.5). In contrast, the abundances of the bacterial amoA and nosZ genes along with the community UniFrac measurements generated sensitive EC20 estimates with less associated error and a response that was consistent across all soil types (Figure 4.4, Figure 4.5). For bacterial amoA, EC20

-1 -1 -1 concentrations ranged from 10 mg kg to 400 mg kg with an average EC20 of 155 mg kg

-1 (SD 195 mg kg ). For the nosZ gene and UniFrac measurements, EC20 concentrations ranged between 250 mg kg-1 and 700 mg kg-1, with an average of 355 mg kg-1 (SD 330 mg kg-1) for nosZ and 250 mg kg-1 (SD 460 mg kg-1) for UniFrac (Table 6). For bacterial amoA, nosZ and the Acidobacteria:Proteobacteria ratio the soil plots with the lowest carbon content (P1 and

P4) generated the most sensitive EC20 values. For the UniFrac measurement, the most sensitive EC20 values were found in the low (P1) and high carbon soils (P3).

Table 4.7 Generated Effective Concentration values (EC20) across a range

of community indices.

e Population Treatment EC20 Model parameters Measurea Modelb Plot %C exposure Conc.c SEd B d e f f Sobs W2.3 P1 5 short n.a. n.a. -0.03 171.0 1.9 SE (25.32) P2 8 short 140 80 -0.44 91.9 403.4 df = 23 P3 36 short 30 340 -0.15 124.8 871.3 P4 6.7 long 1300 1000 -0.66 111.4 2763.7 ACE W2.3 P1 5 short n.a. f n.a. f 0.04 154.0 4867.1 SE (26.17) P2 8 short 20 220 -0.26 94.9 149.1 df = 23 P3 36 short n.a.f n.a. f -0.03 153.2 51.5 P4 6.7 long 1300 2000 -0.53 103.2 3118.4 Chao1 W2.3 P1 5 short n.a. f n.a. f 0.00 157.3 .00001 SE (20.5) P2 8 short 30 16 -0.28 93.9 146.6 df = 23 P3 36 short 11000 6400 -0.61 97.9 23398 P4 6.7 long 1500 620 -0.59 109.1 3334 Shannon W2.3 P1 5 short 8100 850 -1.91 98.8 10384 (H) SE (7.14) P2 8 short 250 110 -0.40 99.9 816.4 df = 23 P3 36 short 2200 1900 -0.28 98.5 11865 P4 6.7 long 3700 5300 -0.21 102.3 36297 Evenness W2.3 P1 5 short 10400 830 -1.13 100.5 15863 (J’) SE (3.30) P2 8 short 140 470 -0.24 102.0 9992.5 df = 23 P3 36 short 50 390 -0.09 125.4 11809 P4 6.7 long n.a. f n.a. f 50273 145.6 882.4 w. UniFrac W2.3 P1 5 short 10 1.28 -0.13 159.4 6.42 SE (7.41) P2 8 short 60 110 -0.26 91.9 348.4 df = 23 P3 36 short 10 0.16 -0.12 172.9 1.98 P4 6.7 long 940 750 -0.41 93.4 3027.9 Acido:Prot LL3 P1 5 short 10 56 0.38 1.3 184.1 Ratio SE (3.4) P2 8 short 40 110 1.03 1.04 159.5 P3 36 short 4500 2700 1.05 1.7 17195 P4 6.7 long 10 0.14 0.24 2.06 2.7 nifH W1.3 P1 5 short n.a.f n.a.f -0.02 282.5 15863 SE(25.39) P2 8 short n.a.f n.a. f 0.01 320.4 9992.5 df = 90 P3 36 short 10 0.1 0.13 244.0 11809 P4 6.7 long 10 0.1 0.09 219 882.4 amoA W2.3 P1 5 short 10 0.06 -0.16 214. 0.4 SE (17.0) P2 8 short 410 240 -0.40 104.2 1356.6 df = 90 P3 36 short 200 490 -0.25 113.7 1337.6 P4 6.7 long 10 0.06 -0.20 228.4 0.2 nosZ W2.3 P1 5 short 130 110 0.54 1244.0 2066.2 SE (17.37) P2 8 short 740 540 0.78 1702.2 5119.1 df = 90 P3 36 short 520 830 5.84 526.2 671.5

P4 6.7 long 30 41 0.66 605.6 306.8 a The community indices evaluated including; Sobs (Species Observed), ACE (abundance- based coverage estimation), Chao1, Shannon (H), Pielou’s evenness (J’), Weighted UniFrac, the Acidobacteria:Proteobacteria ratio and the abundance of functional genes within the nitrogen cycle (nifH, amoA, and nosZ) as determined through qPCR. bThe best fitting dose response curve for the data as determined by AIC. (W2.3 = Weibull 2 curve / 3 parameters, W1.3 = Weibull1 curve / 3 parameters, LL3 = log logistic curve / 3 parameters). cThe concentration that corresponds to a 20% effective change on the community. dThe associated e standard error of the EC20 estimate only. Model parameters are defined as b = relative slope, d = upper limit and e = point of inflection. fNo reliable dose response curve could be fitted.

Figure 4.5 Comparison of EC20 values derived from community measures across variable carbon contents. * no reliable dose-response model could be fitted.

4.4 DISCUSSION

The primary observation of diesel fuel impacts on subantarctic soil microbial communities was a reduction in evenness and richness, resulting in a bacterial community that was heavily dominated by a few specific genera. Our results suggest that the decline in diversity and richness observed after diesel fuel contamination was linked to the disruption of the nitrogen cycle and affected niche specific species (Figure 4.4). The sensitivity, low associated estimate errors, and sustained inhibition of the bacterial amoA gene across variable soil types suggest inhibition of potential nitrification activity is likely to be the best microbial indicator of soil sensitivity to diesel fuel for subantarctic soils. The nosZ gene abundance,

Acidobacteria/Proteobacteria ratio and UniFrac similarities are also potentially valuable

-1 -1 microbial targets. The average EC20 value [155 mg kg (SE 95 mg kg )], generated from the abundance of the bacterial amoA gene was within our low diesel fuel concentration range

(<400 mg kg-1). This concentration is consistent with previous ecotoxicology investigations

-1 at Macquarie Island, where an EC20 of 190 mg kg was generated from an acute potential nitrification enzyme assay and a protective concentration between 50 and 200 mg kg-1 was recommended based on avoidance, survival and reproduction tests of the endemic Macquarie

Island earthworm (17, 62).

High throughput sequencing and qPCR targeting functional genes involved in the nitrogen cycle were used here to simultaneously identify the stimulated and inhibited portions of the bacterial community. This combined approach enabled the vulnerable taxa within the bacterial community to be identified. Across the four subantarctic soil types, the overall phylogenetic diversity, potential nitrification species and the Acidobacteria phyla were found to be significantly inhibited in response to SAB diesel fuel. This was especially important in the low carbon soils where the broad community diversity estimates of Sobs, ACE, and Chao1

100 were unable to detect a significant response to the SAB, due to pre-existing low species richness. While the limitations of widely used diversity estimates have been previously reported (155, 181), it is important to note that without the targeted and functionally important indices used here, the establishment of sensitive species would have been missed due to the limited response in the low carbon soils, or masked by the dominance of the few stimulated species.

Pseudomonas, the most stimulated genera found here, has well known hydrocarbon degrading capabilities and its presence has been widely reported in petroleum hydrocarbon contaminated sites in Antarctica and Macquarie Island (15, 16, 121). The other genera most stimulated in response to high diesel fuel concentrations was Parvibaculum, which has also been reported to have the genetic potential for hydrocarbon degradation (40). While the rapid stimulation of potential hydrocarbon degraders that we observed has been well documented, the full extent of species inhibited with diesel fuel (contributing up to 80% of the total abundance in control communities) has not previously been highlighted (Figure 4.3, Table

4.5). Of those species most inhibited by diesel fuel, nitrifying bacteria were notably vulnerable with significant declines in both the relative abundance of nitrification species, as well as bacterial amoA gene copy numbers (Figure 4.3, Figure 4.4). The amount of organic carbon in soils is thought to mediate toxicity in terrestrial systems by binding toxic substrates, thereby limiting their bioavailability in the environment. Carbon present in a system may also provide an alternative carbon source for microorganisms unable to utilise petroleum hydrocarbons. Previously on Macquarie Island, organic carbon has been correlated to the microbial distribution and hydrocarbon degrading capacity of soils (163). Here we found the sensitivity to potential nitrification inhibition, denitrification stimulation, and decreases in the

Acidobacteria/Proteobacteria ratio were greatest in the soils with the lowest organic carbon content (Figure 4.4, Table 4.7). However, we found the carbon content did not mediate

101 sensitivity of soils to all indices, for example the UniFrac distance measurement was least sensitive in the chronically exposed soil, despite relatively low soil carbon content. Carbon was found to be significantly correlated with the biological distribution. However, pH, soil moisture and available nutrient levels had more significant effects on the resulting bacterial communities (Table 4.6). Furthermore, the soil characteristics within each soil plot had more influence on the bacterial distribution than the diesel fuel concentration alone. We can conclude that the response of the bacterial population was different across soil types, with toxicity more likely to be greatest in low carbon soils. The mode of action and extent to which other environmental characteristics mediate sensitivity or resilience, especially pH, moisture and nutrients needs to be further explored.

The application of high-throughput sequencing and qPCR technologies used in this chapter allowed for the studied evaluation of the sensitivity of more than 1700 species, across 266 families, from 39 separate phyla. Consistent with rapid tolerance testing approaches, we believe the species coverage and confidence in the ecosystem relevance is high. However, it must be noted that the lack of replication in our control samples may have introduces bias into the final analysis and increased replication, particularly for the control soils would improve confidence in the individual toxicity values obtained here. Given their integral roles in terrestrial ecosystem function and the capacity for high throughput analysis, we believe microbial populations have a lot to offer as biological indicators.

102

5 APPYING MICROBIAL INDICATORS TO

CONTAMINATED SITES ON MACQUARIE ISLAND

5.1 INTRODUCTION

Three major petroleum hydrocarbon contaminated sites associated with fuel storage and power generation facilities of the ANARE station have been identified at Macquarie Island

(Figure 5.1) (9, 89) . The 3 sites were distinct in their contaminants and soil properties as determined through chemical, physical (1994, 2003) (9, 89) and biological surveys (2008)

(163). The most recent spill occurred at Main Power House (MPH) South in 2002 and resulted in approximately 180 metric tons of moderately drained, sandy soil becoming highly contaminated with petroleum hydrocarbons in the SAB range of C9 to C18 (Figure 5.1C). An older site (<1994) Main Power House (MPH) East, was estimated to be comprised of approximately 100 metric tons of water-saturated peaty soil, moderately contaminated with a mixture of SAB (C9 to C18), diesel range organics (DRO) (C9 to C28) and Lube range (C9 to

C40) sources (Figure 5.1C) The third spill (<1994) at the Fuel Farm (FF) was estimated to contain approximately 600 metric tons of sandy soil, low-to moderately-contaminated with primarily SAB (C9 to C18) and DRO (C9 to C28) (Figure 5.1C) (9).

In 2003, pilot tests began to test the applicability of bioventing and or nutrient addition remediation strategies (9, 48). Subsequently, a bioremediation project began in 2009 combining both nutrient addition and a micro-bioventing system to stimulate the petroleum hydrocarbon degrading potential of the indigenous microorganisms. Whilst the bioremediation project remains ongoing, the ability to close the site remains hindered by a lack of available remediation target concentrations specific to the environment.

103

Figure 5.1 Location of contaminated sites. A) Macquarie Island in the

Southern Ocean, southeast of Tasmania. B) Macquarie Island; black box

showing the location of the station at the north of the island. C) Layout of the

station on Macquarie Island, highlighting the areas of contamination at the

Fuel Farm (FF) (yellow), MPH South (red), MPH East (orange).Buildings are

shown by black rectangles. This map has been adopted from (182).

At Macquarie Island, microbial nutrient cycling processes including nitrification and denitrification have successfully been utilised in earlier simulated spiking experiments to model protective values for diesel fuel toxicity (17). Using a range of acute and sub-acute ex-

104 situ experiments, we have now expanded this work to include absolute abundances of key functional genes underpinning these processes, combined with a range of community wide, diversity, phylogenetic and niche specific targets (Chapter 4.4). Of the original 11 microbial targets evaluated, 4 were identified as promising microbial indicators of diesel fuel toxicity in subantarctic soils including amoA, nosZ, the weighted UniFrac distance measurement and oligotrophic to copiotrophic ratios.

Our aim was to apply the microbial indicators identified through acute and sub-acute ex-situ assays, to the chronically contaminated sites on Macquarie Island. We hypothesized that the overall TPH concentrations will decrease over the remediation process. We further hypothesized that consistent toxicity trends to the simulated spiking experiments would be observed. However, we expected a greater level of variability in the soil properties, microbial communities and TPH concentrations. Subsequently we also predicted a higher level of variability in the microbial community response than the simulated experiments.

5.2 METHODS

5.2.1 Soil Sampling

To monitor the microbial response across the lifespan of the bioremediation project, soil samples were collected from the 3 contaminated sites (Figure 5.2, Figure 5.3) over a 5 year period from pre-remediation in February 2009 (MPH East only), to February 2013. Samples were collected throughout the soil profile via soil coring at the MPH east and soil pits at the

Fuel Farm and MPH South (Figure 5.4). The location of the soil cores/pits was selected randomly within a grid at each site, i.e. one sample within each 3m x 3m grid square

(avoiding any areas with previous soil disturbances to the soil profile from infrastructure instalment or earlier soil core/pits). Soil samples (50 g) were collected aseptically into 50 ml

105 vials and replicate 100 g samples were collected alongside in amber glass jars for soil physical, chemical and TPH analysis. The TPH and soil parameters were measured according to the methods in 3.2.2 (Table 5.1). In order to account for relative contributions of biodegradation and evaporation, we considered not only the concentration of TPH but also some of the indices known to change during biodegradation or evaporation, such as the ratios of resolved n-alkanes and isoprenoids (10, 183). Only one ratio for biodegradation (n-

C12/(R+UCM) 11.5-12.5) and one ratio for evaporation (n-C12/i-C14) was presented here (Figure

5.5). The depth of soil sampling was dictated by the site characteristics, in particular, the distance between surface and bedrock. As a result, the total soil profiles range from 0-70 cm at MPH East, 0-140 cm at MPH South and 0-200 cm at the Fuel Farm. Each soil core/soil pit was sampled 5-6 times, vertically down the soil profile. Only two of the soil samples from each soil core/soil pit (1 shallow and 1 deep) were used for analysis in this study; MPH East shallow: 10-30 cm, MPH East deep: 40-70 cm, MPH South shallow: 25-60 cm, MPH deep:

60-135 cm, Fuel Farm shallow: 25-70 cm, Fuel Farm deep: 70-140 cm (Table 5.2). A summary of the soil physical and chemical properties is provided in (Table 5.1).

106

496,030 496,040 496,050 496,060 496,070

"'3 a) Shallow (depth class 1) ~ Emergency power house .... 3 0 _,.., ~ 0

Legend «> mglkg Sample Information 3 Year, Pyr osequenced . 2013 • 2013, pyrosequenced ~ ~ . 2012 "'0 "'0

496,030 496.040 496.050 496.060 496.070 496,030 496,040 496,050 496,060 496,070 .... "' b) Deep (depth class 2) 3

Emergency power house ...... 0 -> _,..-> 0 3 "'3 ~ "'g _,.3 .... g 0 » "',.; "'3 3 3 "',.; Legend Sample Information «> 3 Year, Pyrosequenced . 2013 • 2013, pyrosequenced 0 . 2012 :g0 "'g 0 2012, pyrosequenced g "',.; 2011 "',.; Q 2011, pyrosequenced 2010 0 2010, pyrosequenced . 2009 • 2009, pyrosequenced 1m contour ~ 0 • Fuel tank 2m ... "'0 N "'0 ~ 0 Concrete bund 6'?;. ~ ,.; ,.; 0 Building • MPH East zone <5' '? • MPH South zone A 0 2.5 5 10 1m m

496,030 496,040 496,050 496,060 496,070

Figure 5.2 Map of MPH (a) Samples collected within the shallow range for each site (MPH East = 10-30 cm), (MPH South = 25-60 cm) (b) samples collected at depth, (MPH East 40-70 cm), (MPH South = 60-135 cm).

107

495,830 495,840 495,850 495,860 495,870 495,880 495,890 495,900 495,910

0 0 N N "'g a) Shallow (depth class 1) "'g ,.;"' ,.;"'

+ ~ I ~ Legend 0 + 0 "' / "' Sample Information ,.;"' + "'"? "',.; "' / "' Year, Pyrosequenced + • 2013 /

0 + 0 0 e 2013 pyrosequenced / 0 "'0 + "'0 2012 / "',.; + E "',.; "' / "' Q 2012 pyrosequenced + / • 2010 + 0 / 0 "'.... e 2010, pyrosequenced + ...."' 0 / 0 X - Fence "',.; + "',.; 1m contour 70 0 Building ~ - Fuel tank ~ 0 0 ,.;"' D Concrete bund N ,.;"' E ('")

0 0 ...... A 0 0 0 5 10 20 3 ,.;"' m E ,.;"'

495,830 495,840 495,850 495,860 495,870 495,880 495,890 495,900 495,910 495,830 495,840 495,850 495,860 495,870 495,880 495,890 495,900 495,910

0 0 N N "'0 b) Deep (depth class 2) "'0 "' "' "',.; '·~+/O+. s:>+, "',.; · 1630 mglkg+, + +, ~ / ~ "'0 + "'0 Legend / ,.;"' + "',.; "' Samp4e Information / "' + Year, Pyrosequenced / • 2013 + 0 / 0 .,0 + .,0 0 e 2013 pyrosequenced / 0 ,.;"' + E ,.;"' "' 2012 / "' + / Q 2012 pyrosequenced + 0 / 0 "'.... • 2010 + ...."' 0 / <64 mglkg 0 2010, pyrosequenced ,.;"' e + "',.; X- Fence E 70 1m contour "' 0 0 "'.... "'.... 0 D Building 746mg 0 "' N "' "',.; Fuel bund "',.;

R A R ...... 0 3 0 0 5 10 20 "' ~(t\ "' "',.; m E "',.; N

495,830 495,840 495,850 495,860 495,870 495,880 495,890 495,900 495,910

Figure 5.3 Map of FF (a) Samples collected within the shallow range for FF

(25-70 cm) (b) samples collected for FF at depth, (70-140 cm).

108

Figure 5.4. Soil sampling methods. (A & B)Soil cores obtained from MPH

East (2010) using the split corer. Distinctive hydrocarbon layers indicated by the red arrows were visible at depths of 15-25 cm (A) and 40-47 cm (B).

(C & D) A soil sampling pit from FF (2010) with a total depth of 130 cm.

Discrete petroleum hydrocarbon layers indicated by the red arrows were observed below the surface horizon at 30-40 cm (C) and at a depth of 85 –

125 cm (D).

109

Table 5.1 Physical and chemical parameters of soil samples

a sample Year Fuel Depth d.group Drymatter (uS/cm) pH Cl NO2 NO3 PO4 SO4 NH4 no MPH East 1 2009 20900 15-25 1 n.a n.a n.a n.a n.a n.a n.a n.a n.a 2 2009 <64 60-70 2 n.a n.a n.a n.a n.a n.a n.a n.a n.a 3 2009 11900 15-25 1 n.a n.a n.a n.a n.a n.a n.a n.a n.a 4 2009 <64 60-70 2 n.a n.a n.a n.a n.a n.a n.a n.a n.a 5 2009 29200 13-30 1 n.a n.a n.a n.a n.a n.a n.a n.a n.a 6 2009 1350 60-70 2 n.a n.a n.a n.a n.a n.a n.a n.a n.a 7 2010 1000 50-60 2 0.87 181.90 6.78 120.94 0.08 0.54 137.24 8.80 9.65 8 2010 1800 20-30 1 0.84 375.00 6.34 338.86 0.08 6.62 34.79 294.82 4.64 9 2010 640 60-70 2 0.91 129.20 5.83 118.18 0.08 22.35 26.68 57.20 0.68 10 2010 13000 10-18 1 0.72 191.40 6.22 176.20 2.29 20.52 19.41 170.21 32.61 11 2010 410 60-70 2 0.45 307.00 5.55 597.54 0.08 306.23 19.52 191.38 5.26 12 2010 12100 15-25 1 0.84 137.30 6.68 78.70 2.00 2.90 19.00 103.00 17.22 13 2011 6800 25-33 1 0.17 487.00 5.47 69.65 0.15 8.01 1.20 300.00 3.00 14 2011 2040 45-50 2 0.19 166.80 6.82 96.93 0.25 28.22 11.07 244.16 0.75 15 2011 4350 20-25 1 0.11 295.00 6.00 65.79 15.82 230.62 4.14 300.00 11.06 16 2011 <64 40-45 2 0.16 320.00 4.69 74.68 0.15 864.22 6.36 47.45 16.53 17 2012 4250 29-38 1 0.81 182.50 6.90 105.54 1.59 9.07 3.11 170.73 0.57 18 2012 1460 38-44 2 0.81 287.20 5.92 117.56 2.12 222.10 2.04 287.68 6.36 19 2012 1180 25-30 1 0.79 106.80 6.85 105.30 0.08 0.08 16.70 24.45 16.33 20 2012 74 55-60 2 0.83 151.30 7.05 82.77 1.60 102.56 76.26 42.12 45.94 21 2012 <64 21-32 1 0.76 223.70 6.05 333.11 0.08 13.33 19.38 61.90 0.15 22 2012 305 46-57 2 0.85 419.00 6.01 280.94 0.08 2.34 10.76 479.96 0.15 23 2013 8820 20-30 1 0.84 502.00 6.00 99.00 <3.28 <4.43 <3.07 1110.00 <1.29 24 2013 3630 40-50 2 0.63 482.00 4.90 107.00 <3.28 <4.43 <3.07 1090.00 <1.29

110

25 2013 1250 60-70 2 0.87 102.00 7.60 63.00 <3.28 <4.43 79.73 16.00 <1.29 26 2013 2690 10-20 1 0.68 249.00 6.80 188.00 <3.28 150.51 3.37 253.00 52.81 27 2013 2640 60-70 2 0.84 167.00 7.00 70.00 10.18 159.36 28.83 57.00 32.20 28 2013 3020 60-70 2 0.86 94.00 7.00 62.00 <3.28 <4.43 36.80 22.00 11.85 29 2013 6900 19-28 1 0.86 824.00 4.50 89.00 <3.28 <4.43 <3.07 2390.00 27.05 30 2013 <64 65-75 2 0.81 111.00 6.40 106.00 <3.28 44.27 42.93 24.00 <1.29 MPH South 31 2010 720 25-30 1 0.58 109.50 5.15 89.24 0.08 35.34 8.52 174.66 2.85 32 2010 5030 80-85 2 0.59 87.20 4.60 66.93 0.08 91.70 1.94 75.99 3.08 33 2010 6620 35-40 1 0.49 191.00 6.44 281.37 0.08 43.61 63.02 114.92 11.35 34 2010 11500 75-80 2 0.42 165.10 6.40 374.80 0.08 11.12 156.31 28.40 14.38 35 2011 <64 25-30 1 0.14 109.10 4.72 128.91 0.15 20.03 3.34 63.41 0.75 36 2011 <64 55-75 2 0.08 144.10 6.76 79.33 1.10 132.28 13.21 76.53 29.68 37 2011 <64 75-80 2 0.09 111.60 4.45 22.89 0.15 183.95 4.57 59.12 16.12 38 2011 1130 25-35 1 0.16 94.00 6.80 54.04 0.57 57.51 38.95 42.45 0.75 39 2011 748 25-36 1 0.13 186.80 6.09 43.72 15.18 371.04 4.48 87.15 38.61 40 2011 <64 70-80 2 0.06 72.90 4.91 16.29 0.15 129.11 0.00 33.82 1.65 41 2012 <64 60-68 2 0.95 102.40 5.82 14.51 3.58 135.83 56.53 18.12 41.04 42 2012 323 20-30 1 n.a n.a n.a n.a n.a n.a n.a n.a n.a 43 2012 <64 80-85 2 0.93 82.60 5.29 34.86 0.08 60.91 10.58 57.71 0.15 44 2012 440 25-30 1 n.a n.a n.a n.a n.a n.a n.a n.a n.a 110- 45 2012 73 115 2 n.a n.a n.a n.a n.a n.a n.a n.a n.a 46 2013 6600 35-50 1 0.85 236.00 6.30 84.00 <3.28 <4.43 4.60 443.00 <1.29 47 2013 <64 80-90 2 0.89 69.00 6.20 61.00 <3.28 <4.43 23.31 20.00 7.08 48 2013 226 30-50 1 0.90 112.00 5.90 62.00 <3.26 88.54 6.59 65.00 <1.27 49 2013 <64 85-135 2 0.95 55.00 5.10 13.00 <3.26 61.98 11.35 26.00 8.37 50 2013 334 25-35 1 0.92 85.00 4.60 25.00 <3.26 66.40 <3.05 66.00 <1.27 51 2013 1470 75-90 2 0.95 83.00 4.40 33.00 <3.26 75.26 <3.05 25.00 <1.27

111

52 2013 6905 35-60 1 0.90 96.00 4.90 69.00 <3.28 13.28 5.52 82.00 <1.29 53 2013 84 90-120 2 0.93 55.00 5.00 27.00 <3.28 28.33 3.99 40.00 <1.29 FF 54 2010 <64 40-45 1 0.88 128.50 7.21 53.85 1.91 9.95 38.60 86.69 24.01 55 2010 9550 85-90 2 0.87 89.30 6.88 50.94 0.08 1.31 48.15 30.98 7.97 56 2010 7620 50-60 1 0.82 135.60 7.19 86.06 0.08 1.03 74.48 38.45 34.26 57 2010 609 55-60 1 0.75 84.10 6.71 37.96 0.08 0.91 36.19 47.93 22.06 58 2010 <64 90-100 2 0.91 135.50 6.69 49.76 0.08 59.66 168.16 25.32 39.37 125- 59 2010 5510 130 2 0.84 106.40 6.53 27.03 0.08 1.04 75.51 120.93 11.68 60 2010 6220 30-40 1 0.83 71.70 6.78 50.73 0.08 0.68 50.31 7.56 3.27 61 2010 5040 83-90 2 0.89 97.50 6.48 44.58 0.08 1.29 89.39 8.30 2.79 62 2010 5170 50-55 1 0.85 60.20 7.18 22.91 0.08 0.64 24.96 23.67 2.98 63 2010 9590 90-95 2 0.85 107.20 6.79 23.83 0.08 2.56 206.24 8.76 15.22 64 2010 <160 25-30 1 0.79 121.70 6.95 58.80 0.08 0.96 163.21 18.43 0.92 65 2010 <160 70-75 2 0.92 63.00 6.96 28.12 0.08 6.31 104.08 11.26 0.49 66 2012 351 25-30 1 0.91 133.50 6.73 24.87 0.08 2.99 2.44 223.60 4.47 67 2012 5880 92-97 2 0.91 202.80 6.56 28.37 1.08 1.15 5.46 353.03 5.90 68 2012 <64 55-60 1 0.92 24.80 6.93 24.11 0.08 3.37 11.16 5.83 0.75 69 2012 765 45-50 1 0.91 198.30 5.50 64.52 0.08 273.83 24.05 70.57 8.11 110- 70 2012 92 115 2 0.83 47.60 6.60 34.26 0.08 4.32 41.92 19.75 0.15 71 2012 756 90-95 2 0.87 181.50 6.81 69.33 0.08 3.62 44.50 246.34 4.18 72 2012 5925 55-60 1 0.83 94.40 7.09 42.89 0.08 0.08 18.51 90.23 0.15 120- 73 2012 <64 125 2 0.91 154.10 6.74 30.93 0.08 217.32 77.32 23.61 1.34 74 2012 1080 60-65 1 0.91 36.40 6.76 42.37 0.08 6.93 8.02 10.63 0.69 132- 75 2012 <64 137 2 0.87 86.80 6.58 31.30 1.83 58.19 59.84 21.42 11.93 76 2013 748 30-60 1 0.88 88.00 7.40 23.00 <3.28 5.75 28.83 65.00 <1.29

112

77 2013 3630 70-100 2 0.90 94.00 6.40 11.00 <3.28 106.24 46.00 16.00 15.46 78 2013 11700 30-60 1 0.86 64.00 6.90 53.00 <3.28 <4.43 23.61 <2 <1.29 79 2013 8910 60-70 1 0.85 114.00 6.40 165.00 <3.28 <4.43 9.81 4.00 <1.29 80 2013 16400 40-35 1 0.93 75.00 7.00 17.00 <3.28 <4.43 4.91 54.00 <1.29 81 2013 5130 75-95 2 0.78 95.00 7.30 77.00 <3.28 <4.43 67.46 13.00 103.04 82 2013 1770 40-55 1 0.88 149.00 7.20 21.00 <3.28 <4.43 8.59 246.00 <1.29 83 2013 1820 85-120 2 0.89 142.00 7.10 17.00 65.69 132.80 73.60 24.00 69.55 84 2013 168 50-80 1 0.77 95.00 7.00 20.00 <3.28 9.30 18.71 37.00 <1.29 85 2013 1630 80-120 2 0.86 93.00 6.90 17.00 <3.28 <4.43 73.60 109.00 <1.29 86 2013 <64 40-60 1 0.88 64.00 6.90 28.00 <3.28 29.66 174.79 26.00 <1.29 87 2013 <64 80-110 2 0.88 65.00 7.00 32.00 <3.28 16.38 101.19 15.00 <1.29 a Depth group designated 1 for shallow, or 2 for deep. Depths are dependent on soil profiles of each site. Highlighted samples were further analysed with pyrosequencing on the 454 barcoding Roche platform.

113

Table 5.2 Soil sampling at each contaminated site

Site Years Number of Soil Fuel range (mg Sample Depths -1 sampled samples sampling kg ) Shallow Deep method MPH 2009, 2010, 30 Split corer DLa -27900 10-30 40-70 East 2011, 2012, 2013

MPH 2010, 2011, 23 Sampling pit DL a -11500 25-60 60-135 South 2012, 2013 Fuel 2010, 2012, 34 Sampling pit DL a-16400 25-70 70-140 Farm 2013 a Minimum detection limit (DL) of TPH

5.2.2 Community fingerprinting

Total gDNA was extracted from the 87 shallow and deep soil samples (5.2.1) in triplicate and quantified according to 2.2.10. Once diluted to a standard working concentration (10 ng/µl), the gDNA was utilised in ARISA assays to generate bacterial community profiles for the entire 261 DNA samples (3.2.4). Ordination plots were created in PRIMER according to

(3.2.7). A range of statistical analysis including ANOSIM and DistLM were performed on the multivariate community data to establish the significant sampling factors and significant environmental drivers (3.2.7, 3.2.8). A sub-selection of samples was chosen for further analysis with pyrosequencing based on the ordination plots and the significance of sampling factors (P < 0.05) (Table 5.1).

5.2.3 Pyrosequencing

DNA from the replicate extracts were combined and sent for pyrosequencing for a final 30 samples across the 3 sites. The pyrosequencing was performed on the Roche 454 titanium platform with the same primers and conditions as outlined in 2.2.11. The sequence data was primarily processed using the MOTHUR software package as described in 2.2.11. Further

114 resemblance matrices, ordination plots, statistical and multivariate analysis were also conducted in PRIMER according to 3.2.7.

5.2.4 Community indices.

A range of broad and targeted community diversity estimates were calculated for the samples across the TPH concentration range according to the indices identified in chapter 4.4. The broad diversity estimates included species richness, evenness, Shannon diversity and

Simpson diversity. Based on the high level of associated error and the highly variable EC20 concentrations generated in 4.3.3, Sobs, Ace and Chao were excluded. The Acidobacteria /

Proteobacteria ratio was also calculated and used as an additional index as described in 4.2.5.

5.2.5 Quantitative PCR

Key processes within the nitrogen cycle were targeted with qPCR by amplifying and quantifying functional genes previously identified to be sensitive to TPH (Chapter 4); amoA, nosZ, nifH according to (4.2.6). As a way of quantifying relative estimates of bacterial abundance, the 16S rRNA gene was also targeted using the primers 338 and 519 following the conditions outlined in (4.2.6). All reaction efficiencies and standard curve properties of the qPCR reactions are surmised in Table 5.3

Table 5.3 qPCR reaction efficiencies and standard curve properties.

16S rRNA amoA nosZ nifH Reaction efficiency 100.1-100.5% 96.9-99.9% 90.1-91.7% 85.7-86% R2 of slope 0.992-0.996 0.990-0.997 0.994-0.998 0.997 Linear over (x) orders of magnitude 6 7 6 6

115

5.2.6 Phylogenetic diversity across TPH concentrations.

The phylogenetic distribution of bacterial communities present were grouped within fuel groups of undetected (< D.L.), low (D.L.-400 mg kg-1), medium (401-5000 mg kg-1) and high

(> 5000 mg kg-1) levels of TPH. The average relative abundances of phyla within each fuel concentration range were determined within each contaminated site and presented as pie charts (Figure 5.7). Fifty three phyla present in low relative abundances (< 0.1 %) were collated in a group denoted ‘Others’ and included; Planctomycetes, Verrucomicrobia,

Cyanobacteria, Chlorobi, and 30 candidate divisions.

5.2.7 Significantly inhibited and stimulated genera within sites

To elucidate the abundance of individual genera across fuel groups, those genera present at a site in two or more samples were analysed in a plot-fitting script in R according to 2.2.14.

OTUs significantly (P < 0.05) inhibited or stimulated with high concentrations of TPH were determined based on the slope of the line and how accurately the data fitted the linear dose- response model used. The results were plotted individually for each genus and a summary of the results was collated and presented in a volcano plot (Figure 5.10). The top 10 genera significantly inhibited or stimulated within each plot were listed in Table 5.6.

5.2.8 Environmental correlations to bacterial distribution

The measured environmental variables were transformed and normalised as described in

(3.2.8) and an environmental similarity matrix was created using the Euclidian coefficient in

PRIMER. Correlation between the environmental similarity matrix and the bacterial similarity matrix was tested using the Pearson coefficient (3.2.8). Individual variables were further tested for correlation with the DistLM function in PRIMER E (3.2.8). The null

116 hypothesis of no correlation between environmental variables and the bacterial communities was tested (P < 0.05).

5.3 RESULTS

As expected, the MPH East exhibited the highest levels of TPH contamination (Figure 5.5).

Over the bioremediation project the TPH levels at the MPH East were observed to significantly decrease. At MPH South a decline in TPH was observed but was only significant at depth. For both MPH sites, inherent variability between the TPH concentrations over various years was also observed. By comparison, the FF site had no significant decreases in TPH and high variability across depth and years high.

Significant evidence of biodegraded fuel signatures was present by 2010 at all sites. Evidence of evaporation was also present at all sites, yet the rate of evaporation was slower than the biodegradation, with elevated ratios not significantly reduced until 2011 (MPH South), or

2012 (MPH East, FF). Despite the lower average TPH concentrations, the mechanisms of bioremediation and evaporation appeared hindered in the deeper samples. At MPH East this effect was particularly noticeable. In contrast to MPH South and FF, the MPH East is characterised with water saturated, peaty soil which would limit the mechanisms of evaporation and also potentially limit diffusion, and therefore the effectiveness of the micro- bioventing and nutrient addition system.

117

Figure 5.5 Average TPH concentrations per site. Biodegradation and evaporation ratios were compared to ratios from several batches of reference SAB from 2001-2002. The shallow samples had a greater degree of biodegradation and evaporation than samples collected at depth. TPH concentrations were on average higher in the shallow samples.

118

5.3.2 Community fingerprinting profiles and the significance of soil sampling factors

According to the ordination plots (Figure 5.6) and ANOSIM results (Table 5.) generated from the bacterial community profiles, the TPH concentration had the greatest influence on bacterial community structure (R = 0.239, P <0.01) at all sites. Soil samples were also significantly separated by site (R = 0.17, P < 0.01), depth (R = 0.095, P <0.01) and year of collection (R = 0.073, P < 0.01). Samples with high TPH concentrations were consistently similar to other samples with high TPH, regardless of site, depth or year of sampling (Figure

5.6). Within individual sites TPH concentration remained the most influential sampling factor

(Table 5.). The depth of the sample was also significant at all sites, yet had a greater impact on microbial communities at the MPH East and the FF sites. The year of the sample collection was significant at all sites but was observed to be the least significant sampling factor tested (R = 0.073 – 0.163, P < 0.01).

Table 5.4 ANOSIM results for sampling factors including site, TPH

concentration, year of sampling and depth of sample.

All sites MPH East MPH South FF R value P R value p R value p R value P TPH concentration 0.239 <0.01 0.409 <0.01 0.199 <0.01 0.274 <0.01 Year 0.073 <0.01 0.163 <0.01 0.159 <0.01 0.097 <0.01 Depth 0.095 <0.01 0.308 <0.01 0.117 <0.01 0.171 <0.01 Site 0.17 <0.01

119

Figure 5.6 Ordination plots of the bacterial communities based on Automated ribosomal intergenetic spacer analysis

(ARISA) profiles (A-C) and pyrosequencing results (D-F). Ordination of samples labelled by site (A&D); MPH East, MPH

South and FF. Ordination of samples designated by fuel concentration range (B&E); group 1 5000 mg kg-1. 2-D bubble plots representative of TPH concentrations (C&F). Samples were primarily separated by fuel concentration. Only in samples with low TPH were samples separated by site.

120

5.3.3 Samples selected for pyrosequencing.

In order to gain taxonomic identification and more detailed community diversity and structure information, a sub-selection of samples were chosen, based on the ordination plots

(Figure 5.6) and ANOSIM results from the community fingerprinting above. Samples from each significant group were selected with priority given to the most significant factors i.e.

TPH concentration, site and depth. A total of 30 samples were selected for pyrosequencing and are listed in Table 5.1.

5.3.4 Microbial community diversity estimates across sites.

At the MPH East and the FF, the highly contaminated soils (> 5000 mg kg-1) had low species richness, species evenness, Shannon diversity and Simpson diversity estimates (Figure 5.7).

In uncontaminated soils and soils that had been successfully remediated (D.L. – 400 mg kg-1) the diversity estimates were observed to recover to higher values. At the MPH East, a spike in diversity estimates was also observed in the soils with TPH concentrations in the mid-level contamination (1000-5000 mg kg-1). Limited samples were sequenced from this fuel range at the FF, therefore we cannot comment if this spike in diversity estimates was repeated at that site (Figure 5.7). The MPH South also exhibited a similar spike in diversity estimates but at concentrations closer to 10000 mg kg-1. There was also no subsequent decrease in community estimates, or concentration threshold observed. However, it must be noted that the highest

TPH measured at MPH South was 11500 mg kg-1, in comparison to the 27900 mg kg-1 at the

MPH East. In uncontaminated and remediated soils, the MPH South had lower average species richness (400) than the MPH East (800) and FF (800) as well as lower species evenness, Shannon and Simpson estimates (Figure 5.7).

121

Figure 5.7 Bacterial community diversity estimates across the TPH range. The species richness, evenness, Shannon diversity and Simpson diversity estimates decreased with increasing TPH concentrations at all sites except MPH South where a stimulation effect at mid-high TPH ranges was observed.

122

5.3.5 Oligotrophic to copiotrophic ratio across sites

According to Fierer et al.2007, Proteobacteria are considered representative of copiotrophic species and Acidobacteria representative of oligotrophic species. In highly contaminated soils

(> 5000 mg kg-1) at the MPH East and FF the ratio of Acidobacteria to Proteobacteria was very low (0-0.05) (Figure 5.8). At the MPH East, this ratio had recovered in uncontaminated or remediated soils (D.L. – 400 mg kg-1) to 0.2-0.3 and showed some level of recovery in soils with mid-level contamination (1000-5000 mg kg-1) with ratios of 0.05-0.08. At the FF, recovery of the ratio was only seen in soils with no detectable TPH (0.2), while soils with limited TPH concentrations <100 mg kg-1 still had lower ratios of between 0.02-0.08. At the

MPH South no recovery of the ratio was observed and all of the uncontaminated or remediated soils had ratios lower than 1.2. Higher ratios were observed at the site but they were associated with higher TPH concentrations.

5.3.6 Targeted bacterial indices using qPCR

The presence of petroleum hydrocarbons led to increases in the total bacterial biomass as measured by qPCR targeting the 16S rRNA gene (Figure 5.8). The increases in the 16S rRNA gene copy numbers were observed to be dependent on sample depth, with only small increases in copy numbers observed in MPH South and FF soils collected below 70 cm. The

16S rRNA gene copy numbers were an order of magnitude lower at the MPH South site (3.0 x

107) in comparison to the MPH East (3.0 x 108) and FF (3.0 x 108).

The abundance of the bacterial amoA gene, indicative of nitrification was inhibited with high

TPH concentrations at all sites (Figure 5.8). The amoA copy numbers was highest at MPH

East (up to 1.6 x 106), followed by the MPH South (1.8 x 105) and the FF (6.0 x103). The nosZ and nifH genes, indicative of denitrification and nitrogen fixation were also quantitated.

123

However, the abundance of copy numbers was variable across sites and TPH concentrations, with no significant response observed (data not shown).

Figure 5.8 Targeted bacterial community indices across the TPH

concentration range. The solid lines represent dose-response curves that

could be fitted to the data. The 16S rRNA gene copy numbers were separated

by depth (shallow <70 cm, deep >70 cm).

124

5.3.7 Phylogenetic diversity within each site.

MPH East and FF exhibited similar phyla level distributions of bacteria, particularly in the high TPH soils (Figure 5.9). Both sites were dominated by Proteobacteria with the average relative abundance increasing from 48-56 % to 70-78% in soils with high TPH. Communities at the MPH East and FF sites also had lower relative abundances of Acidobacteria in soils with high TPH compared with soils that had no detected TPH, with relative abundances shifting from 14.2% to undetectable at MPH East, and 5.4-1.4% at the FF. Other phyla observed to decrease in relative abundance in soils with high TPH included Firmicutes (11.0 to 5.0% at MPH East, 14.0 to 4.0% at FF), Chloroflexi (4.5 to 1.3 % at MPH East, 6.7 to

2.1% at FF) and Gemmatimonadetes (2.7% to undetected at MPH East, 2.6 to 0.1% at FF).

The relative contribution of the low abundant phyla denoted ‘Others’ also decreased from 4.3 to 2.7% at MPH and 9.5 to 1.6% at FF. The concentration of TPH appeared to have no inhibitory effect on Actinobacteria and Bacteroidetes with relative abundances remaining stable or increasing across the TPH range.

The Phyla level community composition at MPH South was similar to MPH East and FF in soils with no detectable TPH (Figure 5.9). However, the phyla composition in low/mid and high TPH concentration soils was very different in MPH South compared to MPH East and

FF. Most notably Proteobacteria was less abundant in the high TPH soils, decreasing from

59.2 to 35.0%, while Firmicutes was more abundant increasing from 11.1 to 34.4%.

Chloroflexi also increased from 3.9 to 8.2%, and the relative contribution of low abundant phyla ‘Others’ increased slightly from 3.3 to 4.8%. Some observed trends were consistent between the sites with the relative abundance of Acidobacteria decreasing from the soils with undetected TPH to soils with high TPH (3.4 to 0.7%) and Actinobacteria and Bacteroidetes

125 were not inhibited with TPH, maintaining stable abundances from 11.0 to 8.1% and 6.9 to

6.9% respectively.

Figure 5.9 The average relative abundances of individual Phyla across

TPH concentrations. The TPH concentrations were grouped according to

samples with TPH concentrations below detection limit (D.L), Low and

medium (D.L - 5000 mg kg-1) and high (>5000 mg kg-1).

126

5.3.8 Significantly inhibited and stimulated genera across contaminated soils.

At the MPH East and FF sites, a greater number of genera were significantly (<0.05) inhibited than stimulated with high TPH concentrations (Figure 5.10). Although only 57/614 at MPH East and 69/555 genera at FF were inhibited (<0.05), the combined relative abundance of the genera in the uncontaminated soils was 48 % and 80.4% respectively (Table

5.5). Of the ten genera that were most inhibited with high TPH at the FF, there were 4 genera from candidate divisions and unclassified genera from Actinobacteria and Proteobacteria

(Table 5.6). Other genera significantly inhibited included the genus Nitrospira, responsible for nitrite oxidation, and Rhodoplanes a phototrophic purple non-sulfur bacteria. At the MPH

East the most abundant inhibited genera was unclassified genera within Acidobacteriaceae.

Consistent with the FF, unclassified genera within Rhodoplanes and Actinobacteria were also significantly inhibited. Other genera most inhibited included the candidate division TM7, genera within Betaproteobacteria, Gemmatimonadetes, Syntrophobacteraceae,

Thermogemmatisporaceae and small numbers of Dyella and Rhodanobacter. At the MPH

South site, only three genera were observed to be significantly inhibited (Figure 5.10). The inhibited genera from candidate divisions WPS, SC3 and Chitinophagaceae (Table 5.6) contributed only 2.2% of the relative abundance in soils with no detectable TPH (Table 5.5).

127

Table 5.5 Summary of inhibited and stimulated genera at each

contaminated site.

Fuel Farm MPH East MPH South Total genera 555 614 508 Significantly inhibited 69 57 3 Avg. abundance at fuel group 1 80.4 ± 4.5 48.0 ± 0.0 2.2 ± 1.9 Avg. abundance at fuel group 1 38.8 ± 34.3 49.4 ± 25.8 2.1 ± 0.0 Avg. abundance at fuel group 3 1.7 ± 0.0 21.1 ± 8.2 1.0 ± 0.0 Avg. abundance at fuel group 4 1.1 ± 0.9 1.9 ± 1.7 0.01 ± 0.02 Significantly stimulated 20 7 22 Avg. abundance at fuel group 1 0.6 ± 0.8 0.0 ± 0.0 1.2 ± 2.0 Avg. abundance at fuel group 2 3.9 ± 3.6 0.08 ± 0.09 2.1 ± 0.0 Avg. abundance at fuel group 3 1.7 ± 0.0 1.3 ± 1.2 1.0 ± 0.0 Avg. abundance at fuel group 4 37.4 ± 20.0 4.4 ± 3.6 13.9 ± 4.7

The MPH South and FF had similar numbers of significantly stimulated genera at 22 and 20 respectively, with the genera at FF exhibiting the most substantial increases in relative abundance. The significantly stimulated genera at FF were present in clean soils at low abundances totalling <1% and were stimulated up to 37% in the high TPH soils. The stimulated genera within MPH South were also present in low abundance in clean soils

(<1%) and were stimulated up to 13.9% in the high TPH soils. One of the most stimulated genera at both sites was UNC_Rhodocylaceae. Rhodocyclaceae was formally classified as

Psuedomonadaceae and contains primarily aerobic species capable of denitrification and species with versatile metabolisms that prefer oligotrophic conditions (184). Other genera most stimulated with TPH at the FF included Rhodoferax (stimulated up to 16.2%) and

Alicycliphilus (stimulated up to 4.9%). At MPH East there were only 7 genera observed to be significantly stimulated with TPH, of these only 2 genera were stimulated above 1% of the total relative abundance, UNC_Porphyromonadaceae and UNC_Desulfobulbaceae. In high

TPH soils the relative abundance of all significantly stimulated genera contributed just 3.6% of the total relative abundance.

128

Figure 5.10 Volcano plots of individual genera significantly inhibited or stimulated with high TPH at each site. Blue diamond’s indicate genera significantly inhibited or stimulated (P<0.05). Pink diamonds indicate genera with no significant response to TPH. Negative slope values indicate inhibition, while positive slope values indicate stimulation with TPH.

129

Table 5.6 The 10 most significantly inhibited and stimulated genera* at

each site.

P Relative abun. Relative abun. Slope Value in low (5000 mg kg-1) Average (SD) Average (SD) Inhib ited MPH East UNC_Acidobacteriaceae -1.109 0.000 7.8 0.7 0.0 0.0 Dyella -0.878 0.009 0.6 0.6 0.0 0.0 Rhodoplanes -0.867 0.002 2.8 3.5 0.0 0.0 Rhodanobacter -0.825 0.005 0.9 0.6 0.0 0.0 UNC_Thermogemmatisporaceae -0.782 0.045 2.3 3.3 0.0 0.0 UNC_Betaproteobacteria -0.734 0.034 2.5 1.5 0.0 0.0 Gemmatimonadetes -0.734 0.018 1.0 1.4 0.0 0.0 UNC_Syntrophobacteraceae -0.719 0.036 0.5 0.8 0.0 0.0 TM7-1 -0.695 0.002 0.3 0.2 0.0 0.0 UNC_Actinobacteria -0.670 0.019 1.1 0.6 0.0 0.0 MPH South UNC_Chitinophagaceae -0.673 0.035 1.0 1.0 0.0 0.0 WPS-2 -0.520 0.041 0.8 0.7 0.0 0.0 SC3 -0.484 0.048 0.4 0.4 0.0 0.0 FF UNC_Proteobacteria -0.854 0.001 10.7 1.7 0.0 0.0 iii1-15 -0.790 0.000 3.5 0.1 0.0 0.0 UNC_Actinobacteria -0.785 0.000 4.0 1.9 0.0 0.0 S085 -0.700 0.000 2.4 0.2 0.0 0.0 Gemm-1 -0.697 0.003 4.4 0.7 0.0 0.0 Rhodoplanes -0.669 0.000 4.2 0.2 0.1 0.2 MND1 -0.662 0.008 5.7 3.0 0.0 0.0 Nitrospira -0.634 0.007 4.5 3.0 0.0 0.0 UNC_Micrococcaceae -0.581 0.022 0.6 0.5 0.0 0.0 EB1017 -0.580 0.004 1.7 1.1 0.0 0.0 Stimulated MPH East Trichococcus 0.283 0.007 0.0 0.0 0.2 0.2 Turicibacter 0.315 0.023 0.0 0.0 0.1 0.1 Kaistia 0.377 0.049 0.0 0.0 0.3 0.4 UNC_Desulfobulbaceae 0.490 0.042 0.1 0.1 1.1 1.1 SMB53 0.495 0.015 0.0 0.0 0.4 0.0 Williamsia 0.589 0.019 0.0 0.0 0.6 0.1 UNC_Porphyromonadaceae 0.643 0.011 0.0 0.0 1.6 2.1

130

MPH South GIF9 0.458 0.005 0.0 0.0 0.2 0.1 Gaiellales 0.507 0.040 0.2 0.3 1.0 0.2 UNC_Christensenellaceae 0.517 0.039 0.1 0.1 0.6 0.2 SHA-116 0.518 0.025 0.0 0.1 0.5 0.1 Desulfovibrio 0.521 0.006 0.0 0.0 0.3 0.3 UNC_Dehalococcoidaceae 0.551 0.014 0.0 0.0 0.5 0.4 Syntrophus 0.566 0.032 0.0 0.1 0.8 0.8 Treponema 0.683 0.016 0.1 0.1 1.2 0.3 UNC_Rhodocyclaceae 0.773 0.034 0.6 1.0 5.0 3.2 Microbispora 0.779 0.018 0.1 0.1 2.2 1.3 FF Caulobacter 0.428 0.031 0.0 0.0 0.3 0.3 WCHB1 0.552 0.005 0.0 0.0 0.5 0.5 Treponema 0.608 0.002 0.0 0.0 0.4 0.3 Geothrix 0.623 0.023 0.0 0.0 0.6 1.0 BA008 0.650 0.001 0.0 0.0 0.4 0.3 Geobacter 0.692 0.013 0.4 0.6 2.3 2.2 UNC_Flavobacteriaceae 0.749 0.007 0.0 0.0 2.3 2.0 Rhodoferax 0.802 0.006 0.2 0.2 16.2 11.2 Alicycliphilus 0.804 0.007 0.0 0.0 4.9 5.7 UNC_Rhodocyclaceae 1.008 0.001 0.0 0.0 8.8 13.5 *Genera level diversity that could not be classified to the genera level is listed with the closest level of classification after UNC_(closest classification level).

131

5.3.10 The correlation of bacterial communities and soil properties.

Within individual sites (MPH East, MPH South and FF) the bacterial communities were correlated to a similar extent with TPH (0.57, 0.40, 0.47), phosphate (0.31, 0.39, 0.33), ammonia (0.20, 0.36, 0.32) and conductivity (0.40, 0.67, 0.54) (Figure 5.121). At MPH

South, the site was heavily influenced by pH (0.90), while MPH East and to a lesser extent

MPH South were influenced by chlorine (0.55, 0.36) and sulphate (0.49, 0.34). The bacterial communities from soils with no to low levels of detected TPH exhibited a strong correlation with pH (0.83) and chlorine (0.59) (Figure 5.112). For the soils with low TPH, pH (0.70) and chlorine (0.50) were also correlated, along with location (0.92), year of sample collection

(0.88) and to a lesser extent nitrite (0.64), sulphate (0.56) and phosphate (0.30). With small increases in TPH contamination levels the actual amount of TPH in low (0.54) and medium

(0.56) fuel groups became significantly correlated to bacterial populations (Figure 5.11). For the highly contaminated soils, the TPH concentration and chlorine were less correlated

(0.35). Instead nitrate (0.66), soil moisture (0.53), pH (0.50) and nitrite (0.42) exhibited a greater correlation with the bacterial populations.

132

Figure 5.11 Correlation of environmental variables to bacterial community composition within each site.

133

Figure 5.12 Environmental correlations with the distribution of the bacterial community.

134

5.4 DISCUSSION

The MPH East was the only site that we sampled pre remediation, and consequently also showed the greatest decline in TPH over the remediation period (Figure 5.5). At MPH South a decreasing trend in TPH over the remediation period was also observed, however high variability within and between sampling years at the MPH South and FF resulted in no significant decrease in TPH over the four year period. Measuring the TPH alone gives limited information of the relative contributions of biodegradation and evaporation. Biodegradation and evaporation markers based on ratios of n-alkanes and isoprenoids within the GC-FID fuel signatures have been identified (183). Representative ratios or indices were applied here and showed evidence of both biodegradation and evaporation mechanisms, yet greater levels of biodegradation, particularly within the first 12 months of the bioremediation. Unsurprisingly, the samples collected at depth exhibited less evidence of evaporation and bioremediation.

The ordination plots and ANOSIM results highlighted the dominance of TPH as a driving factor of chronically contaminated microbial communities (Figure 5.6, Table 5.). Soils with medium to high TPH shared bacterial community structure irrelevant of site, year or depth.

However, soils with undetected or low TPH reflected the heterogeneity of the soils with the bacterial populations significantly separated by site, depth of sample and year of sampling.

This separation of bacterial communities within low TPH soils only, suggests that for these chronically contaminated sites, TPH is a greater driver of bacterial populations than soil parameters or the influence of bioremediation.

The presence of petroleum hydrocarbons was observed to stimulate the microbial community as observed with increases in the 16S rRNA gene (Figure 5.8). The abundance of the 16S rRNA gene was further observed to be dependent on depth with bacterial abundance less

135 within subsurface (>70 cm) soils (Figure 5.8). Previously the abundance, diversity, and composition of bacterial communities have been found to be horizon specific, with estimated bacterial diversity (H′) higher in the surface horizons than in the corresponding sub-surface horizons (185). The disparity between abundances at depth was not observed at the MPH east site where the maximum depth of sampling was 70 cm, but was observed at the MPH South and FF, suggesting the limiting effect of depth on microbial numbers at Macquarie Island occurs beyond 70 cm. Will et al. (2010) found that diversity was correlated with organic carbon content, total nitrogen content and the C-to-N ratio which decreased with depth. It is expected these limiting factors would be soil dependent and therefore variable depending on the site (185). The decreasing bacterial abundance with depth may also be contributing to the lower biodegradation as exhibited in the n-C12/(R+UCM)11.5-12.5 ratios (Figure 5.5).

The impact of chronic TPH contamination on bacterial communities was variable between sites. The microbial community diversity indices previously identified in chapter 4 were applicable to the MPH East and FF sites with species richness, evenness, Shannon and

Simpson diversity estimates lowest in soils with high TPH (Figure 5.7). Consistent with the bacterial population response in Chapter 4, the majority of individual genera were also inhibited in high TPH samples from MPH East and FF (Figure 5.10). Oligotrophic ecological niches were particularly susceptible to inhibition in these two sites with declines observed in the Acidobacteria to Proteobacteria ratio and a large number of candidate divisions (Figure

5.8). As in Chapter 4, the severity of responses varied between sites and samples and few of the species most stimulated, or inhibited with high TPH were consistent between sites.

The MPH South site responded differently to the MPH East and FF sites, as well as to previous spiking experiments. In uncontaminated or remediated soils, the Phyla level distribution was not dis-similar to the MPH South and FF sites (Figure 5.9). However, across

136 the range of sampled TPH concentrations, greater numbers of genera were stimulated rather than inhibited (Figure 5.10). Consistent with this observation the overall estimates for species richness, evenness, Shannon and Simpson diversity were also highest in soils with high TPH

(Figure 5.7). The soils within the MPH South site are sandy with very low average organic carbon content. This low organic carbon content corresponds to a low bacterial loading, with the 16S rRNA gene abundances an order of magnitude lower than the MPH East and FF sites and the species richness estimates substantially lower at an average of 400 species, compared to approximately 800 in the other two sites (Figure 5.8, Figure 5.7). The stimulation effect resulting from the additional carbon supply provided by the petroleum hydrocarbons may be outweighing any inhibitory effects from petroleum hydrocarbon toxicity. The only response at the MPH South consistent with the other sites was the decline in amoA abundances (Figure

5.8), decline of Acidobacteria (Figure 5.7) and the increase in overall 16S rRNA gene abundances (Figure 5.8).

The ammonium oxidation process within the nitrogen cycle was found to be inhibited in soils chronically contaminated with high TPH at all sites, consistent with previous simulated investigations (17) (Chapter 4). The abundance of amoA genes was highly variable between sites with copy numbers ranging from 1.6 x 106 to 6.0 x 103 between the MPH East and FF sites. Despite the variable abundances, the amoA copy numbers were inhibited in soils with high TPH at each site, in comparison to the undetected or low TPH soils (Figure 5.8). There was some indication that this process does not recover as quickly as the other identified indices. At all sites there were some soils still exhibiting low or non-detectable amoA copy numbers, despite reduced TPH concentrations (Figure 5.8).

While the majority of the community indices from our previous ex-situ spiking experiments

(Chapter 4) were found to be relevant, the nosZ gene abundance was not as applicable as

137 expected. The nosZ was found to be variable within sites and across TPH concentration ranges. No significant inhibition or stimulation response was observed. We suspect the rapid proliferation of species with the nosZ may be an acute to sub-acute response that is negligible on a longer time scale. In support of this observation, (Yergeau et al. 2012) found that

Pseudomonas sp. which commonly share denitrification and hydrocarbon degrading potential were significantly stimulated immediately after contamination, at one month after the start of remediation, but then decreased after 12 months of remediation, when the residual soil hydrocarbons were almost depleted. (186). Based on our observations, we suggest the usefulness of nosZ as a toxicity indicator may be dependent on the stage of site management and remediation.

138

6 DISCUSSION AND CONCLUSIONS

6.1 METHOD SELECTION

6.1.1 Novel culturing methods verses culture independent methods

The overarching goal of this research was to evaluate if microbial targets could be reliably used as indicators of petroleum hydrocarbon toxicity in polar soils. Initially we compared the

SSMS to culture-independent methods to evaluate the suitability of novel culturing techniques to recover the microbial diversity within a Macquarie Island soil. We found the novel culturing technique was unable to capture the same extent of bacterial diversity as the culture independent sequencing methods and therefore concluded that SSMS was unsuitable as an ongoing technique to monitor the whole community response to petroleum hydrocarbon toxicity. However, we did find that the SSMS successfully enriched for 17.9% of the total species richness captured with pyrosequencing of the 16S rRNA gene, which goes some way in bridging the gap between culture dependent and culture independent techniques. While not suitable for our purposes we suggest novel culturing methods such as the SSMS hold promising potential for bioprospecting efforts to harness the microbial diversity found in the extreme and unique polar environments (187), which are still primarily based on the collection of new isolates (188). The SSMS has holds particular promise if coupled with single cell technologies such as micro-manipulation and flow cytometric cell sorting that can enable genetic information to be generated from single cells or cells in micro-cultivation

(189-191).

6.1.2 Relative sensitivity of community fingerprinting and pyrosequencing technologies

Developments in technologies and reduction in costs have rapidly broadened the accessibility of high-throughput sequencing options (94). However, delivering large scale, multivariate

139 and well-replicated experiments with sequencing still remains an expensive and time consuming option. If taxonomic identities are pursued, as they often are, at the cost of replication, there is a reduced or non-existent capacity to detect significant environmental patterns (159). Our aim was to compare community fingerprinting techniques against a high throughput sequencing platform to evaluate the most reliable, time and cost efficient method, for large scale, multivariate investigations. In the comparison we investigated two rapid and relatively inexpensive community fingerprinting techniques (ARISA and T-RFLP) and one of the most commonly utilised next generation sequencing platforms technologies; 16S rRNA barcoded tag pyrosequencing. Our results indicated that community fingerprinting techniques had a similar capacity as the pyrosequencing platform to detect significant spatial, temporal and treatment shifts in bacterial community structure, and to detect significant environmental drivers of community composition (Chapter 3).

Community fingerprinting methods have limited to no capacity for taxonomic identification, and community structure information is constrained to the most abundant organisms (155).

Therefore, even with comparable capacity to detect significant biological patterns, community fingerprinting methods remain a restricted molecular tool. The exclusion of rare species prevents community fingerprinting methods from generating reliable diversity estimates (134) and excludes a phylogenetically diverse portion of the population with unknown functional relevance. In Musat et al. (2008) organisms present in low abundance (~

0.3%) were shown to contribute more than 40% of the total uptake of ammonium and 70% of the total uptake of carbon from the environment (124). In investigations where only fingerprinting methods are utilised, responses from rare portions of the community are missed or masked by responses from the dominant few.

140

We concluded that community fingerprinting methods used in isolation were insufficient to conduct ecotoxicology experiments. However, the consistency of biological patterns observed between the fingerprinting and sequencing techniques suggests fingerprinting methods can be used to ensure statistical robustness in identifying patterns (spatial scale or environmental drivers), as well as to inform further targeted analysis using more expensive, higher resolution methods. Reports already exist in the literature where T-RFLP profiles were used to confirm that bacterial and archaeal community composition in sub-polar and Arctic waters were significantly correlated with environmental conditions, yet not by spatial distance (160). While in Bissett et al. (2010), fingerprinting with T-RFLP was used to establish spatial, environmental and geographic patterns before combining significantly similar samples to target with PhyloChip (161). Thus, large numbers of samples and replicates can be analysed with community fingerprinting to establish significant biological patterns, before pursuing in-depth community structure and taxonomic information from a sub-selection of representative samples.

6.2 BACTERIAL DIESEL FUEL TOXICITY IN SUBANTARCTIC

SOILS

6.2.1 Functional redundancy of microbial communities in polar soils

In simulated diesel fuel spiking experiments (chapter 2 & 4), a loss of bacterial species richness, diversity and evenness was observed. In animal and plant ecosystems, a loss of biodiversity is often linked to a decline in community functionality and a decreased resilience to disturbance (192). In the past, microbial ecosystems have been thought to be more resilient to disturbance than plant and animal ecosystems, due to their high functional redundancy

(193, 194). However, uncertainty on the extent of this functional redundancy in microbial communities is growing. For example, Allison and Marting in 2008 reviewed over 110

141 studies investigating the effects of heavy metals, hydrocarbons and fertilizer amendments, and found that microbial communities are functionally variable (194) and therefore not as resistant or resilient to disturbances as previously thought. More recently, Cravo-Laureau et al. (2011) reported that a moderate loss of microbial diversity resulted in a significant decline in functional phenanthrene degradation capacity, highlighting that a small reduction in diversity can result in significant impacts on microbial ecosystem functionality (195). The loss of biodiversity we observed, led to decreases in functional capacity as evidenced by inhibition of ammonium oxidation genes and species. There was also a significant shift from oligotrophic to copiotrophic species resulting in loss of niche-specific organisms within unknown associated functions. Thus, we suggest that functional redundancy in polar soils may be more limited than in temperate regions as a result of the overriding oligotrophic conditions and lower species richness.

6.2.2 Acidobacteria to Proteobacteria ratio

The changes in community structure appeared to particularly affect niche-specific species with diesel fuel selectively stimulating copiotrophs while inhibiting oligotrophs. One of the novel potential microbial indicators found to be sensitive to SAB diesel fuel contamination was the Acidobacteria to Proteobacteria ratio (Chapter 3). In 2007, Fierer et al. suggested ecological partitioning within the total bacterial diversity based on the ‘r’ and ‘K’ selection criteria (170). Although encompassing high levels of phylogenetic and physiological diversity, it was found that certain Phyla exhibit oligotrophic (slow growing, ‘K’ selection criteria) or copiotrophic (rapid growth, ‘r’ selection criteria) life strategies. High numbers of

Acidobacteria were found to correlate with low nutrient environments and slow growing ‘K’ selected species occupying specific environmental niches. Conversely, Proteobacteria, in

142 particular β- proteobacteria, correlated with copiotrophic conditions and species capable of rapid ‘r’ selected growth.

The ratio of Acidobacteria to Proteobacteria observed in chapter 4 was consistent with this ecological partitioning trend and generated sensitive EC20 values for P1, P2, and P4 between

-1 -1 10-110 mg kg , all within our low fuel concentration range (<400 mg kg ). The low EC20 values suggest that low nutrient subantarctic soils were particularly susceptible to reductions in niche-specific species (Figure 4.4). While, the high nutrient soils, with pre-existing high ratios of rapidly growing opportunistic species had a much higher sensitivity threshold at

4,500 mg kg-1. The chronically contaminated sites at MPH East and FF on Macquarie Island also exhibited low Acidobacteria to Proteobacteria ratios in soils with high TPH (Figure 5.6) and higher Acidobacteria to Proteobacteria ratios in remediated soils. A level of recovery was observed at MPH East in soils, despite still elevated TPH concentrations at 1000-5000 mg kg-1. At the FF, the same level of recovery did not occur until TPH concentrations had declined to ~ 100 mg kg-1. No level of recovery was observed at MPH South with ratios variable across concentrations. As with other indices there appeared to be a lack of connectivity and consistency throughout the MPH South site, indicative of greater structural perturbations to the community (196).

6.2.3 Ammonium oxidation inhibition

Disruption to the nitrogen cycle was observed with increasing TPH highlighted by declines in the amoA copy numbers and a reduction of genera attributed to nitrification. Doubt has been raised concerning the use of ammonia oxidisers to assess toxicity because these organisms are thought to recover from petroleum hydrocarbon pollution (197). In our spiking experiments we observed that in contrast to the control, the soils chronically exposed to SAB

143 in P4 for 18 months, still had low to non- detectable levels of the genera involved in nitrification, as well as the bacterial amoA copy numbers. At all three of the chronically contaminated sites we also observed many soils still exhibiting low to non- detectable levels of amoA, despite decreasing TPH concentrations (Figure 5.6). These results support the prolonged sensitivity of ammonium oxidation processors to petroleum hydrocarbons and the suitability of it this process as a sensitive microbial target, with direct ecosystem function relevance.

Ammonium oxidation to nitrite and the subsequent oxidation of nitrite into nitrate are crucial processes in soil. The inhibition of nitrification species and bacterial amoA copy numbers was consistent with previous reports, identifying nitrification as a sensitive measurement for microbial communities following contamination with petroleum hydrocarbons (17, 198), heavy metals (199) and solvents (200). The specific toxicity of petroleum hydrocarbons on nitrification potential has been attributed previously to competitive inhibition with the ammonia monooxygenase enzyme by small aliphatic compounds competing directly with ammonia for the active binding site (165), and non-competitive inhibition from compounds with larger molecular weight through binding to a hydrophobic region of the ammonia monooxygenase AMO enzyme, and influencing enzyme turnover (165, 201).

It is important to note that in many soils the AOA are more abundant than AOB, although that was not the case here. In soils where AOA are abundant the role of archaea in ammonium oxidation should be considered. Additionally, the bacterial and archaea amoA gene along with the nifH and nosZ genes were only targeted with one primer set each in this study. Many other primers capable of targeting these genes exist and while we are confident the trends would be consistent, different primers may generate slightly different results.

144

6.3 RECOMMENDED MICROBIAL INDICATORS AND

ASSOCIATED PROTECTIVE CONCENTRATIONS

The most obvious changes to the bacterial community following spiking were the shifts in community structure from diverse communities with many species in low abundance, to communities dominated by fewer abundant species (Figure 4.3). Community structure can be tracked with various community diversity estimates but as we observed here, soils with low carbon and nutrient loading supported simpler communities with lower species richness than those with higher nutrients. As a result, the effective concentrations generated from the dose- response modelling were highly variable, exhibited high associated errors and were not reflective of disruption to the communities’ functional capacity and resilience. Indices that are more reflective of the phylogenetic relatedness and the ecological partitioning between phylogenetic groups generated protective values with a greater consistency between soil types and less associated error (Figure 4.5). Of all the indices targeted, the most consistent microbial indicator was the abundance of amoA. It was also found to be sensitive to low levels of TPH and did not readily recover. We suggest the amoA which has a direct link to the ecosystem function and one or both of the phylogenetic indicators (Acidobacteria to

Proteobacteria ratio, or UniFrac similarity distance measurement) can provide the most robust targets. The combination of amoA and phylogenetic targets can inform on inhibition to known ecosystem functions and can also represent the loss of phylogenetic diversity with unknown functions and ecosystem significance.

The corresponding recommended protective concentrations in this study are 150 mg kg-1to

1140 mg kg-1 (amoA 155 mg kg-1, UniFrac 255 mg kg-1, and Acidobacteria to Proteobacteria

1140 mg kg-1). The values generated from the amoA and UniFrac indicators are consistent with previous ecotoxicology investigations at Macquarie Island, where an EC20 of 190 mg

145 kg-1 was generated from an acute potential nitrification enzyme assay and a protective concentration between 50 and 200 mg kg-1 was recommended based on avoidance, survival and reproduction tests of the endemic Macquarie Island earthworm (62). The average EC20 concentrations generated here from the Acidobacteria to Proteobacteria ratios were disproportionately high because of the values generated in the high carbon soils. An organic carbon content of 36% was unrepresentative of the level of carbon found in the chronically contaminated sites (and the majority of polar soils). Excluding this concentration would result in a considerably lower protective concentration < 50 mg kg-1. Based on the microbial community indicators developed here, we recommend an average protective concentration of

200 mg kg-1, or a protective concentration range of 50-400 mg kg-1 for diesel fuel toxicity on

Macquarie Island.

6.4 PHYLOGENETIC DIVERSITY WITHIN SUBANTARCTIC SITES

Proteobacteria were consistently stimulated and Acidobacteria consistently inhibited with increasing TPH in the acute to sub-acute spiking experiments at Macquarie Island. A group of core genera were also observed to be significantly stimulated with increasing TPH that included Pseudomonas, Parvibaculum, Dyella and Rhodanobacter (Chapter 2, 4). While these genera were consistent between sites, the OTUs present within the genera, along with the extent of stimulation were specific to individual sites. Pseudomonas has well known hydrocarbon degrading capabilities and has been identified at many petroleum hydrocarbon contaminated sites throughout the polar regions (15, 16, 121). Parvibaculum, Dyella and

Rhodanobacter are not as widely identified with hydrocarbon degradation. However,

Parvibaculum and Dyella have been associated with PAH degradation in soil (202), and mangrove sediments (203, 204), and Rhodanobacter, although not linked to direct

146 degradation, has been reported to be capable of utilising the metabolic products of aromatic degradation (205).

In the chronically contaminated sites, the same level of consistency between sites was not observed and very few of the above mentioned genera were significantly elevated in association with high TPH. Both the stimulated and inhibited genera were found to be unique to individual sites and included many candidate divisions and unclassified genera. The high number of novel species further highlights the depth and breadth of unknown phylogenetic and functional diversity found in extreme environments.

The chronically contaminated sites have been exposed to petroleum hydrocarbons for up to

10 years at the MPH South and >20 years at the MPH East and FF. The lack of consistency between genera at these sites and the spiking experiments is therefore not surprising. We would expect that Pseudomonas and other genera would have been present in highly elevated numbers when the spills were fresh and the fuel not yet degraded (206). We suggest changes to the concentration and quality of fuel over time has selected for different organisms better suited to tolerating and or degrading the residual fuel products. The greatest consistency we observed between all the Macquarie Island sites analysed, was the changes to community structure. The bacterial community structure characteristic of soils with low TPH included large numbers of niche-specific species in very low abundance. In soils with high TPH that structure was altered to communities dominated by relatively few, abundant species.

147

6.5 ENVIRONMENTAL PREDICTORS OF BACTERIAL

COMMUNITY COMPOSITION IN POLAR AND SUB POLAR

SOILS

The complex interactions of bacterial communities and the surrounding physical and chemical environment are not fully understood. However, we found consistent environmental variables to be significant predictors of the microbial community structure at the terrestrial polar and sub-polar sites. Excluding grain size variables, pH explained the most variation in bacterial community composition, predicting between 18 and 7.1% of the variation between the polar and sub polar sites (Table 3.5, Table 4.6). Phosphate and carbon were also important variables predicting between 12 and 1.4% of the community variation (Table 3.5,

Table 4.6). For the Macquarie Island chronically contaminated sites, we examined environmental drivers of contaminated and uncontaminated communities separately and found in uncontaminated or remediated soils pH was again a major diver of community composition, along with chlorine, moisture and phosphate (Figure 5.12). At low to mid-levels of contamination pH and chlorine remained important predictors of variation but the location of the samples, the year of sample collection, TPH, nitrite and sulphate were also observed to explain bacterial community composition. In the most contaminated soils the most important predictor of community composition was no longer pH but nitrate and moisture, and to a lesser extent, TPH, pH and year of sampling.

The pH of soil has been widely linked to bacterial community composition (86, 87) along with other nutrient or soil fertility factors such as organic matter, nitrogen and chloride content (207, 208). In a recent paper examining the same polar microbial communities from chapter 3, the soil edaphic factors of nitrogen, carbon and chloride were found to have the

148 greatest control on richness and evenness, while pH controlled the structural and among- community phylogenetic composition (85).

Toxicity can be influenced by a range of soil properties including pH and nutrient availability. However, it is primarily organic carbon concentrations that influence petroleum hydrocarbon bioavailability and therefore toxicity in soil (209). At Macquarie Island, soils with low organic carbon were associated with lower microbial biomass and species diversity.

The low diversity led to less severe decreases in diversity estimates in response to petroleum hydrocarbons (Figure 4.5). For the other microbial indicators; phylogenetic diversity, oligotrophic to copiotrophic ratios and amoA copy numbers, soils with low to medium organic carbon concentrations were more sensitive to petroleum hydrocarbons generating lower EC20 values in the range of 10-400 mg kg-1 (Chapter 4) and exhibiting less potential for recovery in soils undergoing remediation (Chapter 5).

6.6 EFFICIENCY OF BIOREMEDIATION STRATEGIES ON

MACQUARIE ISLAND

6.6.1 Evidence of effective bioremediation

The active remediation strategy at Macquarie Island involves a combination of nutrient

-1 addition to target concentrations below 1200 mg N kg Soil H2O (48) and aeration to the identified optimum target range of between 10-21% oxygen (53). While we observed a level of reduction in average hydrocarbon concentrations at the MPH East and MPH South, the concentrations were highly variable at all sites.

In addition to the overall average reduction of the hydrocarbons at the contaminated sites, other lines of evidence exists that support effectiveness of the active bioremediation program.

149

Over time the GC-FID fuel signatures (even in high TPH soils) changed, decreasing in light n-alkanes and isoprenoid ratios indicative of bioremediation (183). There was also a recovery of the majority of microbial indices. The recovery was site and microbial index specific with

MPH South showing very little restoration of community structure, and amoA gene copy numbers were only partially restored in concentrations below 5000 mg kg-1

Although not analysed here (182), rates of respiration were also found to increase within the active remediation zones. After an initial lag phase, the average respiration rate in the remediation zone was approximately 3 times higher than the background rate at the FF, approximately 6 times higher than background respiration rates (in less organic soils) at MPH

East, and approximately 2 times higher than background respiration rates at MPH South. This rate was calculated over the length of the remediation from oxygen data set collected at 5 minute intervals since the remediation began. The oxygen consumption or respiration rates were observed to decrease over time.

6.6.2 Limiting factors to effective ongoing bioremediation

Despite the overall decreases in average measured hydrocarbons and further lines of evidence of effective bioremediation, there remains high variability and continued elevated hydrocarbon concentrations at all three sites. Several factors exist that could be contributing to the disparity of hydrocarbon concentrations including soil heterogeneity, the radius of influence surrounding aeration points, falling rates in remediation rates and increasing ratios of unresolved complex mixtures (UCM) (182).

Soil heterogeneity The inherent soil heterogeneity throughout the sites are likely to influence biomass loading, non-aqueous phase liquid (NAPL) formation (9), fuel saturation

150 concentrations and bioavailability of residual fuel (209) which can all affect toxicity and the subsequent biodegradation rates.

Effective remediation radius In 2011 a 3-dimensional sampling program was designed to evaluate the effective radius of influence from an aeration point that had been active for 24 months (182). A clear pattern of decreasing concentration with increasing proximity to the aeration point demonstrated that there was an effective remediation radius.

The effective remediation radius was calculated at less than 50-60 cm within the coarse sandy soils at the FF and less than 60-80cm for the coarse sandy soils of MPH South, leaving many areas outside an effective remediation radius. Since that time, aeration spikes have been progressively shifted to treat the un-aerated pockets (182).

Hydrocarbon degradation rates Utilising the respiration data, oxygen consumption rates were converted to in situ hydrocarbon degradation rates (mg hexane /kg soil /day), according to the methods in (182, 210). After an initial lag period, progressive decreases oxygen consumption and therefore hydrocarbon degradation rates were observed over time, consistent with theoretical expectations of progressive hydrocarbon degradation (211).

Declining remediation efficiency was observed at the FF and the MPH with degradation rates appearing to asymptote at 5 mg hexane per kg soil per day in sandy soils.

Increasing proportion of UCMs The composition of fuel in the environment is known to change over time as a result of biological and physical processors. Microbial degradation, volatilisation and leaching all selectively alter fuel composition, resulting in increases of UCMs which are resistant to degradation and have relatively unknown toxicity

(212, 213). At the three chronically contaminated sites an increase in UCMs was observed as biodegradation and evaporation increased, even in soils with high fuel concentrations.

151

We can conclude that one or a combination of the above factors could be preventing active bioremediation, and or limit the ongoing effectiveness of the system. Some of the factors identified can be improved with alterations to the remediation system. In particular, the zone of effective remediation radius can be improved with the addition or movement of aeration spikes, and indeed increases in respiration were observed in areas of re-sparging. For the remaining factors efficiencies are likely to continue to decrease or plateau. In this scenario, microbial indices become increasingly important as a means of monitoring ecosystem recovery. Microbial indicators may also have an important role to play in future experiments investigation the unknown toxicity of UCMs and other residual petroleum hydrocarbon compounds as well as informing remediation optimisation options.

6.7 CONCLUSIONS

Correlating microbial community structure to ecosystem function perturbed or not, remains a key challenge in microbial ecology (67). The amount and quality of data investigating the response of microbial populations to perturbation is increasing, yet the knowledge of microbial communities and their associated functions lags behind data synthesis capabilities.

Subsequently, results from perturbation investigations are often confounding with effects, causes and direct links between variables remaining unclear. Targeting ecosystem specific and sensitive functions such as nitrification, or phylogenetic markers such as Acidobacteria to Proteobacteria ratios is a promising step towards understanding bacterial community response and resilience. Within wetland assessments, community indices that relate to ecosystem specific functionality such as the AOB/AOA ratio have likewise been identified as promising microbial indicators (90).

152

We believe microbial indicators have huge potential as indicators of toxicity, especially in cold regions where the diversity and abundance of other suitable test organisms is low. The ability to model complex microbial communities in dose-response (or stressor-response) models further promotes the utility of microbial indicators as integrated measures of ecosystem health and fit within the framework of established risk assessment frameworks

(90). As knowledge and understanding of microbes functional role in ecosystems improves, the utility and reliability of microbial indicators will also improve.

6.8 FUTURE RESEARCH

With increasing numbers of new and untested chemical compounds reaching the environment, there is an urgent need for rapid, site-specific ecotoxicology approaches that, unlike single species testing, are capable of accurately representing whole ecosystems. (59,

60). We have identified key microbial indicators indicative of diesel fuel at a subantarctic site that can be expanded to include other contaminated sites and greater numbers of soils throughout the polar regions. The microbial indicators and methods outlined here also have promising potential to be applied to temperate regions. While the identified indices may not be relevant for all regions, once site/ region and contaminant specific indices are established, the methods developed can be integrated directly into established risk assessment frameworks.

Ongoing investigations to elucidate the links between community structure and function are also needed. Utilising a range of available technologies such as microarray and network analysis in regions like Macquarie Island, where extensive invertebrate, microbial and toxicology information already exists, provides an ideal environment to further address this aim. Ultimately, there is a need to better understand the environmental abiotic and biotic

153 controls on community resilience and recovery to perturbation, in order to predict and manage critical aspects of soil microbial communities, and the processes they mediate (67).

154

7 REFERENCES

1. Snape I, Riddle MJ, Filler DM, Williams PJ. 2003. Contaminants in freezing ground and associated ecosystems: Key issues at the beginning of the new millennium. Polar Record 39:291-300. 2. Tin T, Fleming ZL, Hughes KA, Ainley DG, Convey P, Moreno CA, Pfeiffer S, Scott J, Snape I. 2008. Impacts of local human activities on the Antarctic environment. Antarctic Science 21:3. 3. Bargagli R. 2008. Environmental contamination in Antarctic ecosystems. Science and the Total Environment 400:212-226. 4. Frenot Y, Chown SL, Whinam J, Selkirk PM, Convey P, Skotnicki M, Bergstrom DM. 2005. Biological invasions in the Antarctic: extent, impacts and implications. Biological Reviews 80:45-72. 5. Cowan DA, Chown SL, Convey P, Tuffin M, Hughes K, Pointing S, Vincent WF. 2011. Non-indigenous microorganisms in the Antarctic: assessing the risks. Trends in Microbiology 19:540-548. 6. Aislabie J, Fraser R, Duncan S, Farrell RL. 2001. Effects of oil spills on microbial heterotrophs in Antarctic soils. Polar Biology 24:308-313. 7. Gore DB, Revill AT, Guille D. 1999. Petroleum hydrocarbons ten years after spillage at a helipad in Bunger Hills, East Antarctica. Antarctic Science 11:427-429. 8. Revill AT, Snape I, Lucieer A, Guille D. 2007. Constraints on transport and weathering of petroleum contamination at Casey Station, Antarctica. Cold Regions Science and Technology 48:154-167. 9. Rayner JL, Snape I, Walworth JL, Harvey PM, Ferguson SH. 2007. Petroleum- hydrocarbon contamination and remediation by microbioventing at sub-Antarctic Macquarie Island. Cold Regions Science and Technology 48:139-153. 10. Snape I, Ferguson SH, Harvey PM, Riddle MJ. 2006. Investigation of evaporation and biodegradation of fuel spills in Antarctica: II-extent of natural attenuation at Casey Station. Chemosphere 63:89-98. 11. Bossio DA, Scow KM, Gunapala N, Graham KJ. 1998. Determinants of soil microbial communities: effects of agricultural management, Season, and Soil Type on Phospholipid Fatty Acid Profiles. Microbial Ecology 36:1-12. 12. Evans CW, Hills JM, Dickson JMJ. 2000. Heavy metal pollution in Antarctica: a molecular ecotoxicological approach to exposure assessment. Journal of Fish Biology 57:8-19. 13. Negri A, Burns K, Boyle S, Brinkman D, Webster N. 2006. Contamination in sediments, bivalves and sponges of McMurdo Sound, Antarctica. Environmental Pollution 143:456-467. 14. Stark JS, Riddle MJ, Simpson RD. 2003. Human impacts in soft‐sediment assemblages at Casey Station, East Antarctica: Spatial variation, taxonomic resolution and data transformation. Austral Ecology 28:287-304. 15. Aislabie JM, Balks MR, Foght JM, Waterhouse EJ. 2004. Hydrocarbon spills on Antarctic soils: effects and management. Environmental Science & Technology 38:1265-1274. 16. Saul DJ, Aislabie JM, Brown CE, Harris L, Foght JM. 2005. Hydrocarbon contamination changes the bacterial diversity of soil from around Scott Base, Antarctica. FEMS Microbiology Ecology 53:141-155.

155

17. Schafer AN, Snape I, Siciliano SD. 2007. Soil biogeochemical toxicity end points for sub-Antarctic islands contaminated with petroleum hydrocarbons. Environmental Toxicology Chemistry 26:890-897. 18. Atlas RM, Bartha R. 1972. Degradation and mineralization of petroleum in sea water: limitation by nitrogen and phosphorous. Biotechnology and Bioengineering 14:309-318. 19. Ferguson SH, Franzmann PD, Revill AT, Snape I, Rayner JL. 2003. The effects of nitrogen and water on mineralisation of hydrocarbons in diesel-contaminated terrestrial Antarctic soils. Cold Regions Science and Technology 37:197-212. 20. Wall DH, Virginia RA. 1999. Controls on soil biodiversity: insights from extreme environments. Applied Soil Ecology 13:137-150. 21. Coulon F, Pelletier E, Gourhant L, Louis RS, Delille D. 2004. Degradation of petroleum hydrocarbons in two sub-antarctic soils: Influence of an oleophilic fertilizer. Environmental Toxicology and Chemistry 23:1893-1901. 22. Delille D, Pelletier E. 2002. Natural attenuation of diesel-oil contamination in a subantarctic soil (Crozet Island). Polar Biology 25:682-687. 23. Atlas RM. 1981. Microbial degradation of petroleum hydrocarbons: An environmental perspective. Microbiological Reviews 45:180-209. 24. Delille D. 2000. Response of antarctic soil bacterial assemblages to contamination by diesel fuel and crude oil. Microbial Ecology 40:159-168. 25. Aislabie J, McLeod M, Fraser R. 1998. Potential for biodegradation of hydrocarbons in soil from the Ross Dependency, Antarctica. Applied Microbiology and Biotechnology 49:210-214. 26. Aislabie J, Saul DJ, Foght JM. 2006. Bioremediation of hydrocarbon-contaminated polar soils. Extremophiles 10:171-179. 27. Whyte LG, Bourbonniere L, Greer CW. 1997. Biodegradation of petroleum hydrocarbons by psychrotrophic Pseudomonas strains possessing both alkane (alk) and naphthalene (nah) catabolic pathways. Applied Environmental Microbiology 63:3719-3723. 28. Yu Z, Stewart GR, Mohn WW. 2000. Apparent contradiction: Psychrotolerant bacteria from hydrocarbon-contaminated Arctic tundra soils that degrade diterpenoids synthesized by trees. Applied Environmental Microbiology 66:5148-5154. 29. Bej AK, Saul D, Aislabie J. 2000. Cold-tolerant alkane-degrading Rhodococcus species from Antarctica. Polar Biology 23:100-105. 30. Thomassin-Lacroix EJM, Yu Z, Eriksson M, Reimer KJ, Mohn WW. 2001. DNA-based and culture-based characterization of a hydrocarbon-degrading consortium enriched from Arctic soil. Canadian Journal of Microbiology 47:1107- 1115. 31. Aislabie J, Foght J, Saul D. 2000. Aromatic hydrocarbon-degrading bacteria from soil near Scott Base, Antarctica. Polar Biology 23:183-188. 32. Yu Z, Stewart GR, Mohn WW. 2000. Apparent contradiction: psychrotolerant bacteria from hydrocarbon-contaminated Arctic tundra soils that degrade diterpenoids synthesized by trees. Applied and Environmental Microbiology 66:5148-5154. 33. Hamamura N, Olson SH, Ward DM, Inskeep WP. 2006. Microbial population dynamics associated with crude-oil biodegradation in diverse soils. Applied and Environmental Microbiology 72:6316-6324. 34. Zhang D, Margesin R. 2014. Characterization of culturable heterotrophic bacteria in hydrocarbon-contaminated soil from an alpine former military site. World Journal of Microbiology Biotechnology 30:1717-1724.

156

35. Ruberto L, Dias R, Lo Balbo A, Vazquez SC, Hernandez EA, Mac Cormack WP. 2009. Influence of nutrients addition and bioaugmentation on the hydrocarbon biodegradation of a chronically contaminated Antarctic soil. Journal of Applied Microbiology 106:1101-1110. 36. Guo GX, Deng H, Qiao M, Mu YJ, Zhu YG. 2011. Effect of pyrene on denitrification activity and abundance and composition of denitrifying community in an agricultural soil. Environmental Pollution 159:1886-1895. 37. Deng H, Guo GX, Zhu YG. 2011. Pyrene effects on methanotroph community and methane oxidation rate, tested by dose-response experiment and resistance and resilience experiment. Journal of Soils and Sediments 11:312-321. 38. Whyte LG, Schultz A, Van Beilen JB, Luz AP, Pellizari V, Labbé D, Greer CW. 2002. Prevalence of alkane monooxygenase genes in Arctic and Antarctic hydrocarbon-contaminated and pristine soils. FEMS Microbiology Ecology 41:141- 150. 39. Vázquez S, Nogales B, Ruberto L, Hernández E, Christie-Oleza J, Lo Balbo A, Bosch R, Lalucat J, Mac Cormack W. 2009. Bacterial community dynamics during bioremediation of diesel oil-contaminated antarctic soil. Microbial Ecology 57:598- 610. 40. Flocco CG, Gomes NCM, Mac Cormack W, Smalla K. 2009. Occurrence and diversity of naphthalene dioxygenase genes in soil microbial communities from the Maritime Antarctic. Environ Microbiol 11:700-714. 41. Powell SM, Ferguson SH, Bowman JP, Snape I. 2006. Using real-time PCR to assess changes in the hydrocarbon-degrading microbial community in Antarctic soil during bioremediation. Microbial Ecology 52:523-532. 42. Margesin R, Labbe D, Schinner F, Greer CW, Whyte LG. 2003. Characterization of hydrocarbon-degrading microbial populations in contaminated and pristine alpine soils. Applied and Environmental Microbiology 69:3085-3092. 43. Whyte LG, Bourbonnière L, Bellerose C, Greer CW. 1999. Bioremediation assessment of hydrocarbon-contaminated soils from the high Arctic. Bioremediation Journal 3:69-80. 44. Powell SM, Ferguson SH, Snape I, Siciliano SD. 2006. Fertilization stimulates anaerobic fuel degradation of antarctic soils by denitrifying microorganisms. Environmental Science and Technology 40:2011-2017. 45. Snape I, Riddle MJ, Stark JS, Cole CM, King CK, Duquesne S, Gore DB. 2001. Management and remediation of contaminated sites at Casey Station, Antarctica. Polar Record 37:199-214. 46. Braddock JF, Walworth JL, McCarthy KA. 1999. Biodegradation of aliphatic vs. aromatic hydrocarbons in fertilized Arctic soils. Bioremediation Journal 3:105-116. 47. Bell TH, Yergeau E, Martineau C, Juck D, Whyte LG, Greer CW. 2011. Identification of nitrogen-incorporating bacteria in petroleum-contaminated arctic soils by using [15N]DNA-based stable isotope probing and pyrosequencing. Applied Environmental Microbiology 77:4163-4171. 48. Walworth J, Pond A, Snape I, Rayner J, Ferguson S, Harvey P. 2007. Nitrogen requirements for maximizing petroleum bioremediation in a sub-Antarctic soil. Cold Regions Science and Technology 48:84-91. 49. Delille D, Coulon F, Pelletier E. 2004. Effects of temperature warming during a bioremediation study of natural and nutrient-amended hydrocarbon-contaminated sub- Antarctic soils. Cold Regions Science and Technology 40:61-70.

157

50. Delille D, Duval A, Pelletier E. 2008. Highly efficient pilot biopiles for on-site fertilization treatment of diesel oil-contaminated sub-Antarctic soil. Cold Regions Science and Technology 54:7-18. 51. Schiewer S, Niemeyer T. 2006. Soil heating and optimized nutrient addition for accelerating bioremediation in cold climates. Polar Record 42:23-31. 52. Couto N, Fritt-Rasmussen J, Jensen PE, Højrup M, Rodrigo AP, Ribeiro AB. 2014. Suitability of oil bioremediation in an Artic soil using surplus heating from an incineration facility. Environmental Science and Pollution Research 21:6221-6227. 53. Walworth J, Harvey P, Snape I. 2013. Low temperature soil petroleum hydrocarbon degradation at various oxygen levels. Cold Regions Science and Technology 96:117- 121. 54. Snape I, Acomb L, Barnes DL, Bainbridge S, Eno R, Filler DM, Plato N, JS P, Raymond T, Rayner J, Riddle M, Rike AG, Rutter A, Schafer A, Siciliano SD, Walworth J. 2008. Contamination, regulation, and remediation: An introduction to bioremediation of petroleum hydrocarbons in cold regions.Cambridge University Press, Cambridge, United Kingdom. 55. Delille D, Coulon F, Pelletier E. 2004. Biostimulation of natural microbial assemblages in oil-amended vegetated and desert sub-Antarctic soils. Microb Ecol 47:407-415. 56. Perrodin Y, Boillot C, Angerville R, Donguy G, Emmanuel E. 2011. Ecological risk assessment of urban and industrial systems: A review. Science of The Total Environment 409:5162-5176. 57. Cairns Jr J. 1983. Are single species toxicity tests alone adequate for estimating environmental hazard? Hydrobiologia 100:47-57. 58. ANZECC/ARMCANZ. 2000. Australian and New Zealand guidelines for fresh and marine water quality. In Zealand AaNZEaCCAaRMCoAaN (ed.), National water quality management strategy paper, vol. 4, Canberra, ACT. 59. Hickey GL, Kefford BJ, Dunlop JE, Craig PS. 2008. Making species salinity sensitivity distributions reflective of naturally occurring communities: Using rapid testing and Bayesian statistics. Environmental Toxicology and Chemistry 27:2403- 2411. 60. Kefford BJ, Palmer CG, Jooste S, Warne MSJ, Nugegoda D. 2005. What is meant by "95% of species"? an argument for the inclusion of rapid tolerance testing. Human and Ecological Risk Assessment 11:1025-1046. 61. King CK, Riddle MJ. 2001. Effects of metal contaminants on the development of the common Antarctic sea urchin Sterechinus neumayeri and comparisons of sensitivity with tropical and temperate echinoids. Marine Ecology Progress Series 215:143-154. 62. Mooney TJ, King CK, Wasley J, Andrew NR. 2013. Toxicity of diesel contaminated soils to the subantarctic earthworm Microscolex macquariensis. Environmental Toxicology and Chemistry 32:370-377. 63. Schafer AN, Snape I, Siciliano SD. 2009. Influence of liquid water and soil temperature on petroleum hydrocarbon toxicity in antarctic soil. Environmental Toxicology and Chemistry 28:1409-1415. 64. Ferrari BC, Zhang C, van Dorst J. 2011. Recovering greater fungal diversity from pristine and diesel fuel contaminated sub-antarctic soil through cultivation using both a high and a low nutrient media approach. Frontiers of Microbiology 2:217. 65. Neilson MN, Winding, A., 2002. Microorganisms as Indicators of Soil Health. In Institute NER (ed.), NERI Technical Report No. 338, Roskilde, Denmark. .

158

66. Avidano L, Gamalero E, Cossa G, Carraro E. 2005. Characterization of soil health in an Italian polluted site by using microorganisms as bioindicators. Applied Soil Ecology 30:21-33. 67. Bissett A, Brown MV, Siciliano SD, Thrall PH. 2013. Microbial community responses to anthropogenically induced environmental change: Towards a systems approach. Ecology Letters 16:128-139. 68. Sun MY, Dafforn KA, Brown MV, Johnston EL. 2012. Bacterial communities are sensitive indicators of contaminant stress. Marine Pollution Bulletin 64:1029-1038. 69. Kumar N, Shah V, Walker VK. 2011. Perturbation of an arctic soil microbial community by metal nanoparticles. Journal of Hazardous Materials 190:816-822. 70. Sawall Y, Richter C, Ramette A. 2012. Effects of eutrophication, seasonality and macrofouling on the diversity of bacterial biofilms in equatorial coral reefs. PLoS ONE 7. 71. Xie X, Liao M, Ma A, Zhang H. 2011. Effects of contamination of single and combined cadmium and mercury on the soil microbial community structural diversity and functional diversity. Chinese Journal of Geochemistry 30:366-374. 72. Bach EM, Baer SG, Meyer CK, Six J. 2010. Soil texture affects soil microbial and structural recovery during grassland restoration. Soil Biology and Biochemistry 42:2182-2191. 73. Sims A, Horton J, Gajaraj S, McIntosh S, Miles RJ, Mueller R, Reed R, Hu Z. 2012. Temporal and spatial distributions of ammonia-oxidizing archaea and bacteria and their ratio as an indicator of oligotrophic conditions in natural wetlands. Water Research 46:4121-4129. 74. Niemeyer JC, Lolata GB, Carvalho GMd, Da Silva EM, Sousa JP, Nogueira MA. 2012. Microbial indicators of soil health as tools for ecological risk assessment of a metal contaminated site in Brazil. Applied Soil Ecology 59:96-105. 75. Das S, Jean J-S, Kar S, Chakraborty S. 2012. Effect of arsenic contamination on bacterial and fungal biomass and enzyme activities in tropical arsenic-contaminated soils. Biology and Fertility of Soils 49:757-765. 76. Li H, Zhang Y, Kravchenko I, Xu H, Zhang Cg. 2007. Dynamic changes in microbial activity and community structure during biodegradation of petroleum compounds: A laboratory experiment. Journal of Environmental Sciences 19:1003- 1013. 77. Lapinskiene A, Martinkus P, Rebzdaite V. 2006. Eco-toxicological studies of diesel and biodiesel fuels in aerated soil. Environmental Pollution 142:432-437. 78. Bissett A, Richardson AE, Baker G, Thrall PH. 2011. Long-term land use effects on soil microbial community structure and function. Applied Soil Ecology 51:66-78. 79. Sun MY, Dafforn KA, Brown MV, Johnston EL. 2012. Bacterial communities are sensitive indicators of contaminant stress. Marine Pollution Bulletin 64:1029-1038. 80. Lear G, Ancion PY, Harding J, Lewis GD. 2012. Use of bacterial communities to assess the ecological health of a recently restored stream. New Zealand Journal of Marine and Freshwater Research 46:291-301. 81. Stomeo F, Makhalanyane TP, Valverde A, Pointing SB, Stevens MI, Cary CS, Tuffin MI, Cowan DA. 2012. Abiotic factors influence microbial diversity in permanently cold soil horizons of a maritime-associated Antarctic Dry Valley. FEMS Microbiol Ecology 82:326-340. 82. Ma D, Zhu R, Ding W, Shen C, Chu H, Lin X. 2013. Ex-situ enzyme activity and bacterial community diversity through soil depth profiles in penguin and seal colonies on Vestfold Hills, East Antarctica. Polar Biology 36:1347-1361.

159

83. Richter I, Herbold CW, Lee CK, McDonald IR, Barrett JE, Cary SC. 2014. Influence of soil properties on archaeal diversity and distribution in the McMurdo Dry Valleys, Antarctica. FEMS Microbiol Ecology. 84. Roesch LFW, Fulthorpe RR, Pereira AB, Pereira CK, Lemos LN, Barbosa AD, Suleiman AKA, Gerber AL, Pereira MG, Loss A, da Costa EM. 2012. Soil bacterial community abundance and diversity in ice-free areas of Keller Peninsula, Antarctica. Applied Soil Ecology 61:7-15. 85. Siciliano SD, Palmer AS, Winsley T, Lagerewskij G, Lamb E, Bissett A, Brown MV, van Dorst J, Ji M, Ferrari BC, Grogan P, Chu H, Snape I. 2014. Soil fertility is associated with fungal and bacterial richness whereas pH is associated with community composition in polar soil microbial communities. Soil Biology and Biochemistry

86. Lauber CL, Hamady M, Knight R, Fierer N. 2009. Pyrosequencing-Based Assessment of Soil pH as a Predictor of Soil Bacterial Community Structure at the Continental Scale. Applied Environmental Microbiology 75:5111-5120. 87. Fierer N, Jackson RB. 2006. The diversity and biogeography of soil bacterial communities. Proceedings of the National Academy of Sciences of the United States of America 103:626-631. 88. Anisimov OA, D.G. Vaughan, T.V. Callaghan, C. Furgal, H. Marchant, T.D. Prowse, H. Vilhjálmsson and J.E. Walsh,. 2007. Polar regions (Arctic and Antarctic). 89. Deprez P, Arens M, Locher H. 1994. Identification and preliminary assessment of contaminated sites in the Australian Antarctic Territory. 1. Casey Station. Australian Antarctic Division, Hobart. 90. Sims A, Zhang Y, Gajaraj S, Brown PB, Hu Z. 2013. Toward the development of microbial indicators for wetland assessment. Water Research 47:1711-1725. 91. Amann RI, Ludwig W, Schleifer KH. 1995. Phylogenetic identification and in situ detection of individual microbial cells without cultivation. Microbiological Reviews 59:143-169. 92. Giovannoni S, Stingl U. 2007. The importance of culturing bacterioplankton in the 'omics' age. Nature Reviews Microbiology 5:820-826. 93. Lewis K, Epstein S, D'Onofrio A, Ling LL. 2010. Uncultured microorganisms as a source of secondary metabolites. Journal of Antibiotics 63:468-476. 94. MacLean D, Jones JD, Studholme DJ. 2009. Application of 'next-generation' sequencing technologies to microbial genetics. Nature Reviews Microbiology 7:287- 296. 95. Roesch LFW, Fulthorpe RR, Riva A, Casella G, Hadwin AKM, Kent AD, Daroub SH, Camargo FAO, Farmerie WG, Triplett EW. 2007. Pyrosequencing enumerates and contrasts soil microbial diversity. ISME Journal 1:283-290. 96. Sogin ML, Morrison HG, Huber JA, Welch DM, Huse SM, Neal PR, Arrieta JM, Herndl GJ. 2006. Microbial diversity in the deep sea and the underexplored “rare biosphere”. Proceedings of the National Academy of Sciences 103:12115-12120. 97. Rusch DB, Halpern AL, Sutton G, Heidelberg KB, Williamson S, Yooseph S, Wu D, Eisen JA, Hoffman JM, Remington K, Beeson K, Tran B, Smith H, Baden- Tillson H, Stewart C, Thorpe J, Freeman J, Andrews-Pfannkoch C, Venter JE, Li K, Kravitz S, Heidelberg JF, Utterback T, Rogers Y-H, Falcón LI, Souza V, Bonilla-Rosso G, Eguiarte LE, Karl DM, Sathyendranath S, Platt T, Bermingham E, Gallardo V, Tamayo-Castillo G, Ferrari MR, Strausberg RL,

160

Nealson K, Friedman R, Frazier M, Venter JC. 2007. The Sorcerer II Global Ocean Sampling Expedition: Northwest Atlantic through Eastern Tropical Pacific. PLoS Biol 5:e77. 98. Dos Santos HF, Cury JC, do Carmo FL, Dos Santos AL, Tiedje J, van Elsas JD, Rosado AS, Peixoto RS. 2011. Mangrove bacterial diversity and the impact of oil contamination revealed by pyrosequencing: Bacterial proxies for oil pollution. PLoS One 6. 99. Kryachko Y, Dong X, Sensen CW, Voordouw G. 2012. Compositions of microbial communities associated with oil and water in a mesothermic oil field. Antonie van Leeuwenhoek, International Journal of General and Molecular Microbiology 101:493-506. 100. Liu Z, Liu J. 2013. Evaluating bacterial community structures in oil collected from the sea surface and sediment in the northern Gulf of Mexico after the Deepwater Horizon oil spill. Microbiology Open 2:492-504. 101. Watve M, Shejval V, Sonawane C, Rahalkar M, Matapurkar A, Shouche Y, Patole M, Phadnis N, Champhenkar A, Damle K. 2000. The 'K' selected oligophilic bacteria: A key to uncultured diversity? Current Science 78:1535-1542. 102. Kaeberlein T, Lewis K, Epstein SS. 2002. Isolating "uncultivable" microorganisms in pure culture in a simulated natural environment. Science 296:1127-1129. 103. Zengler K, Toledo G, Rappe M, Elkins J, Mathur EJ, Short JM, Keller M. 2002. Cultivating the uncultured. Proceedings of the National Academy of Sciences U S A 99:15681-15686. 104. Simu K, Hagstrom A. 2004. Oligotrophic bacterioplankton with a novel single-cell life strategy. Applied and Environmental Microbiology 70:2445-2451. 105. Connon SA, Giovannoni SJ. 2002. High-throughput methods for culturing microorganisms in very-low-nutrient media yield diverse new marine isolates. Applied and Environmental Microbiology 68:3878-3885. 106. Rappe MS, Connon SA, Vergin KL, Giovannoni SJ. 2002. Cultivation of the ubiquitous SAR11 marine bacterioplankton clade. Nature 418:630-633. 107. Rasmussen LD, Zawadsky C, Binnerup SJ, Øregaard G, Sørensen SJ, Kroer N. 2008. Cultivation of hard-to-culture subsurface mercury-resistant bacteria and discovery of new merA gene sequences. Applied Environmental Microbiology 74:3795-3803. 108. Sait M, Hugenholtz P, Janssen PH. 2002. Cultivation of globally distributed soil bacteria from phylogenetic lineages previously only detected in cultivation- independent surveys. Environmental Microbiology 4:654-666. 109. Janssen PH, Yates PS, Grinton BE, Taylor PM, Sait M. 2002. Improved culturability of soil bacteria and isolation in pure culture of novel members of the divisions Acidobacteria, Actinobacteria, Proteobacteria, and Verrucomicrobia. Applied and Environmental Microbiology 68:2391-2396. 110. Svenning MM, Wartiainen I, Hestnes AG, Binnerup SJ. 2003. Isolation of methane oxidising bacteria from soil by use of a soil substrate membrane system. FEMS Microbiology Ecology 44:347-354. 111. Ferrari BC, Binnerup SJ, Gillings M. 2005. Microcolony cultivation on a soil substrate membrane system selects for previously uncultured soil bacteria. Applied and Environmental Microbiology 71:8714-8720. 112. Dowd SE, Callaway TR, Wolcott RD, Sun Y, McKeehan T, Hagevoort RG, Edrington TS. 2008. Evaluation of the bacterial diversity in the feces of cattle using 16S rDNA bacterial tag-encoded FLX amplicon pyrosequencing (bTEFAP). 1471- 2180 8:125.

161

113. Quince C, Lanzen A, Davenport RJ, Turnbaugh PJ. 2011. Removing noise from pyrosequenced amplicons. BMC Bioinformatics 12:38. 114. Schloss PD, Westcott SL, Ryabin T, Hall JR, Hartmann M, Hollister EB, Lesniewski RA, Oakley BB, Parks DH, Robinson CJ, Sahl JW, Stres B, Thallinger GG, Van Horn DJ, Weber CF. 2009. Introducing mothur: Open-source, platform-independent, community-supported software for describing and comparing microbial communities. Applied Environmental Microbiology 75:7537-7541. 115. Edgar RC, Haas BJ, Clemente JC, Quince C, Knight R. 2011. UCHIME improves sensitivity and speed of chimera detection. 1367-4803 27:2194-2200. 116. Huse SM, Welch DM, Morrison HG, Sogin ML. 2010. Ironing out the wrinkles in the rare biosphere through improved OTU clustering. Environmental Microbiology 12:1889-1898. 117. Pruesse E, Quast C, Knittel K, Fuchs BM, Ludwig W, Peplies J, Glöckner FO. 2007. SILVA: a comprehensive online resource for quality checked and aligned ribosomal RNA sequence data compatible with ARB. Nucleic Acids Research 35:7188-7196. 118. Lane DJ. 1991. 16S/23S rRNA sequencing. Nucleic Acid Techniques in Bacterial Systematics:115–175. 119. McDonald D, Price MN, Goodrich J, Nawrocki EP, Desantis TZ, Probst A, Andersen GL, Knight R, Hugenholtz P. 2012. An improved Greengenes taxonomy with explicit ranks for ecological and evolutionary analyses of bacteria and archaea. The ISME Journal 6:610-618. 120. Clarke KR, Warwick , R. M. 2001. Changes in marine communities: an approach to statistical analysis and interpretation. PRIMER-E, Plymouth. 121. Powell SM, Ma WK, Siciliano SD. 2006. Isolation of denitrifying bacteria from hydrocarbon-contaminated Antarctic soil. Polar Biology 30:69-74. 122. Miteva VI, Brenchley JE. 2005. Detection and isolation of ultrasmall microorganisms from a 120,000-year-old Greenland glacier ice core. Applied and Environmental Microbiology 71:7806-7818. 123. Ferrari BC, Gillings MR. 2009. Cultivation of fastidious bacteria by viability staining and micromanipulation in a soil substrate membrane system. Applied Environmental Microbiology 75:3352-3354. 124. Musat N, Halm H, Winterholler B, Hoppe P, Peduzzi S, Hillion F, Horreard F, Amann R, Jørgensen BB, Kuypers MMM. 2008. A single-cell view on the ecophysiology of anaerobic phototrophic bacteria. Proceedings of the National Academy of Sciences of the United States of America 105:17861-17866. 125. Mardis ER. 2008. Next-generation DNA sequencing methods, p. 387-402, vol. 9. 126. Nowrousian M. 2010. Next-Generation Sequencing Techniques for Eukaryotic Microorganisms: Sequencing-Based Solutions to Biological Problems. Eukaryotic Cell 9:1300-1310. 127. Zhang H. 2011. Using pyrosequencing and quantitative PCR to analyze microbial communities. Frontiers of Environmental Science and Engineering in China 5:21-27. 128. Zhou X, Ren L, Meng Q, Li Y, Yu Y, Yu J. 2010. The next-generation sequencing technology and application. Protein and Cell 1:520-536. 129. Novais RC, Thorstenson YR. 2011. The evolution of Pyrosequencing® for microbiology: From genes to genomes. Journal of Microbiological Methods 86:1-7. 130. Singh BK, Nazaries L, Munro S, Anderson IC, Campbell CD. 2006. Use of multiplex terminal restriction fragment length polymorphism for rapid and simultaneous analysis of different components of the soil microbial community. Applied Environmental Microbiology 72:7278-7285.

162

131. Bissett A, Cook PLM, Macleod C, Bowman JP, Burke C. 2009. Effects of organic perturbation on marine sediment betaproteobacterial ammonia oxidizers and on benthic nitrogen biogeochemistry. Marine Ecology Progress Series 392:17-32. 132. Pilloni G, Granitsiotis MS, Engel M, Lueders T. 2012. Testing the limits of 454 pyrotag sequencing: Reproducibility, quantitative assessment and comparison to T- RFLP fingerprinting of aquifer microbes. PLoS One 7. 133. Brown MV, Schwalbach MS, Hewson I, Fuhrman JA. 2005. Coupling 16S-ITS rDNA clone libraries and automated ribosomal intergenic spacer analysis to show marine microbial diversity: Development and application to a time series. Environmental Microbiology 7:1466-1479. 134. Bent SJ, Pierson JD, Forney LJ. 2007. Measuring Species Richness Based on Microbial Community Fingerprints: the Emperor Has No Clothes. Applied Environmental Microbiology 73:2399-2401. 135. Osborne CA, Rees GN, Bernstein Y, Janssen PH. 2006. New threshold and confidence estimates for terminal restriction fragment length polymorphism analysis of complex bacterial communities. Applied Environmental Microbiology 72:1270- 1278. 136. Dunbar J, Ticknor LO, Kuske CR. 2001. Phylogenetic specificity and reproducibility and new method for analysis of terminal restriction fragment profiles of 16S rRNA genes from bacterial communities. Applied Environmental Microbiology 67:190-197. 137. Giebler J, Wick LY, Chatzinotas A, Harms H. 2013. Alkane-degrading bacteria at the soil-litter interface: Comparing isolates with T-RFLP-based community profiles. FEMS Microbiology Ecology. 138. Neumann D, Heuer A, Hemkemeyer M, Martens R, Tebbe CC. 2013. Response of microbial communities to long-term fertilization depends on their microhabitat. FEMS Microbiology Ecology. 139. Bissett A, Richardson AE, Baker G, Kirkegaard J, Thrall PH. 2013. Bacterial community response to tillage and nutrient additions in a long-term wheat cropping experiment. Soil Biology and Biochemistry 58:281-292. 140. Gillevet PM, Sikaroodi M, Torzilli AP. 2009. Analyzing salt-marsh fungal diversity: comparing ARISA fingerprinting with clone sequencing and pyrosequencing. Fungal Ecology 2:160-167. 141. Cleary DFR, Smalla K, Mendonça-Hagler LCS, Gomes NCM. 2012. Assessment of variation in bacterial composition among microhabitats in a mangrove environment using DGGE fingerprints and barcoded pyrosequencing. PLoS One 7. 142. Danovaro R, Luna GM, Dell'Anno A, Pietrangeli B. 2006. Comparison of Two Fingerprinting Techniques, Terminal Restriction Fragment Length Polymorphism and Automated Ribosomal Intergenic Spacer Analysis, for Determination of Bacterial Diversity in Aquatic Environments. Applied Environmental Microbiology 72:5982- 5989. 143. Rayment G, E., Lyons D, J. 2011. Soil chemical methods: Australasia. CSIRO Publishing, Collingwood, Victoria. 144. Hewson I, Fuhrman JA. 2004. Richness and diversity of bacterioplankton species along an estuarine gradient in Moreton Bay, Australia. Applied Environmental Microbiology 70:3425-3433. 145. Fisher MM, Triplett EW. 1999. Automated Approach for Ribosomal Intergenic Spacer Analysis of Microbial Diversity and Its Application to Freshwater Bacterial Communities. Applied Environmental Microbiology 65:4630-4636.

163

146. Slabbert E, Van Heerden CJ, Jacobs K. 2010. Optimisation of automated ribosomal intergenic spacer analysis for the estimation of microbial diversity in fynbos soil. South African Journal of Science 106. 147. Culman SW, Bukowski R, Gauch HG, Cadillo-Quiroz H, Buckley DH. 2009. T- REX: software for the processing and analysis of T-RFLP data. BMC Bioinformatics 10:171. 148. Ramette A. 2009. Quantitative Community Fingerprinting Methods for Estimating the Abundance of Operational Taxonomic Units in Natural Microbial Communities. Applied and Environmental Microbiology 75:2495-2505. 149. R Development Core Team 2008, posting date. R: A language and environment for statistical computing. [Online.] 150. Clarke KR. 1993. Non-parametric multivariate analyses of changes in community structure. Australian Journal of Ecology 18:117-143. 151. Legendre P, Anderson MJ. 1999. Distance-based redundancy analysis: testing multispecies responses in multifactorial ecological experiments. Ecological Monographs 69:1-24. 152. Junier P, Junier T, Witzel K-P. 2008. TRiFLe, a Program for In Silico Terminal Restriction Fragment Length Polymorphism Analysis with User-Defined Sequence Sets. Applied Environmental Microbiology 74:6452-6456. 153. Zinger L, Gobet A, Pommier T. 2012. Two decades of describing the unseen majority of aquatic microbial diversity. Molecular Ecology 21:1878-1896. 154. Claesson MJ, Wang Q, O'Sullivan O, Greene-Diniz R, Cole JR, Ross RP, O'Toole PW. 2010. Comparison of two next-generation sequencing technologies for resolving highly complex microbiota composition using tandem variable 16S rRNA gene regions. Nucleic Acids Research 38. 155. Bent SJ, Forney LJ. 2008. The tragedy of the uncommon: understanding limitations in the analysis of microbial diversity. ISME Journal 2:689-695. 156. Hartmann M, Widmer F. 2006. Community structure analyses are more sensitive to differences in soil bacterial communities than anonymous diversity indices. Applied Environmental Microbiology 72:7804-7812. 157. Gilbert JA, Field D, Swift P, Newbold L, Oliver A, Smyth T, Somerfield PJ, Huse S, Joint I. 2009. The seasonal structure of microbial communities in the Western English Channel. Environmental Microbiology 11:3132-3139. 158. Warton DI, Wright ST, Wang Y. 2012. Distance-based multivariate analyses confound location and dispersion effects. Methods in Ecology and Evolution 3:89- 101. 159. Prosser JI. 2010. Replicate or lie. Environmental Microbiology 12:1806-1810. 160. Winter C, Matthews B, Suttle CA. 2013. Effects of environmental variation and spatial distance on Bacteria, Archaea and viruses in sub-polar and arctic waters. ISME Journal 7:1507-1518. 161. Bissett A, Richardson AE, Baker G, Wakelin S, Thrall PH. 2010. Life history determines biogeographical patterns of soil bacterial communities over multiple spatial scales. Molecular Ecology 19:4315-4327. 162. AMAP. 1998. AMAP Assessment Report: Artic Pollution Issues. Artic Monitoring and Assessment Programme, Oslo, Norway. :859. 163. Powell SM, Bowman JP, Ferguson SH, Snape I. 2010. The importance of soil characteristics to the structure of alkane-degrading bacterial communities on sub- Antarctic Macquarie Island. Soil Biology and Biochemistry 42:2012-2021.

164

164. Hyman MR, Murton IB, Arp DJ. 1988. Interaction of ammonia monooxygenase from Nitrosomonas-Europaea with alkanes, alkenes, and alkynes. Applied Environmental Microbiology 54:3187-3190. 165. Keener WK, Arp DJ. 1993. Kinetic studies of ammonia moonooxygenase inhibition in Nitrosomonas-Europaea by hydrocarbons and halogenated hydrocarbons in an optimized whole-cell assay. Applied Environmental Microbiology 59:2501-2510. 166. Colloff MJ, Wakelin SA, Gomez D, Rogers SL. 2008. Detection of nitrogen cycle genes in soils for measuring the effects of changes in land use and management. Soil Biology and Biochemistry 40:1637-1645. 167. van Dorst J, Bissett A, Palmer AS, Brown M, Snape I, Stark JS, Raymond B, McKinlay J, Ji M, Winsley T, Ferrari BC. Community fingerprinting in a sequencing world. FEMS Microbiology Ecology. 168. Sheneman L, Evans J, Foster JA. 2006. Clearcut: a fast implementation of relaxed neighbor joining. Bioinformatics 22:2823-2824. 169. Lozupone C, Hamady M, Knight R. 2006. UniFrac--an online tool for comparing microbial community diversity in a phylogenetic context. BMC Bioinformatics 7:371. 170. Fierer N, Bradford MA, Jackson RB. 2007. Toward an ecological classification of soil bacteria. Ecology 88:1354-1364. 171. Ma WK, Bedard-Haughn A, Siciliano SD, Farrell RE. 2008. Relationship between nitrifier and denitrifier community composition and abundance in predicting nitrous oxide emissions from ephemeral wetland soils. Soil Biology and Biochemistry 40:1114-1123. 172. Gaby JC, Buckley DH. 2012. A Comprehensive Evaluation of PCR Primers to Amplify the nifH Gene of Nitrogenase. PLoS One 7:e42149. 173. Rotthauwe JH, Witzel KP, Liesack W. 1997. The ammonia monooxygenase structural gene amoA as a functional marker: molecular fine-scale analysis of natural ammonia-oxidizing populations. Applied and Environmental Microbiology 63:4704- 4712. 174. Francis CA, Roberts KJ, Beman JM, Santoro AE, Oakley BB. 2005. Ubiquity and diversity of ammonia-oxidizing archaea in water columns and sediments of the ocean. PNAS U S A 102:14683-14688. 175. Henry S, Bru D, Stres B, Hallet S, Philippot L. 2006. Quantitative detection of the nosZ gene, encoding nitrous oxide reductase, and comparison of the abundances of 16S rRNA, narG, nirK, and nosZ genes in soils. Applied Environmental Microbiology 72:5181-5189. 176. Siciliano SD, Ma W, Powell S. 2007. Evaluation of quantitative polymerase chain reaction to assess nosZ gene prevalence in mixed microbial communities. Canadian Journal of Microbiology 53:636-642. 177. Zhang X, Liu W, Schloter M, Zhang G, Chen Q, Huang J, Li L, Elser JJ, Han X. 2013. Response of the Abundance of Key Soil Microbial Nitrogen-Cycling Genes to Multi-Factorial Global Changes. PLoS ONE 8. 178. Cottingham KL, Lennon JT, Brown BL. 2005. Knowing when to draw the line: designing more informative ecological experiments. Frontiers in Ecology and the Environment 3:145-152. 179. Knezevic SZ, Streibig JC, Ritz C. 2007. Utilizing R Software Package for Dose- Response Studies: The Concept and Data Analysis. Weed Technology 21:840-848. 180. Ritz C, Streibig JC. 2005. Bioassay analysis using R. Journal of Statistical Software 12:1-22.

165

181. Curtis TP, Sloan WT. 2004. Prokaryotic diversity and its limits: microbial community structure in nature and implications for microbial ecology. Current Opinion in Microbiology 7:221-226. 182. Wilkins D, Dorst Jv, Rayner J, King C, Snape I, Wasley J, Hince G, Wise L, Mooney T, Mumford K, Winsley T, Lagerewskij G, Stark S. 2013. Macquarie Island Integrated Risk Assessment (MIRA). Australian Antarctic Division. Submitted to Tasmanian state Government, July, 2013. 183. Snape I, Harvey PM, Ferguson SH, Rayner JL, Revill AT. 2005. Investigation of evaporation and biodegradation of fuel spills in Antarctica I. A chemical approach using GC-FID. Chemosphere 61:1485-1494. 184. van der Zaan BM, Saia FT, Stams AJ, Plugge CM, de Vos WM, Smidt H, Langenhoff AA, Gerritse J. 2012. Anaerobic benzene degradation under denitrifying conditions: Peptococcaceae as dominant benzene degraders and evidence for a syntrophic process. Environmental Microbiology 14:1171-1181. 185. Will C, Thürmer A, Wollherr A, Nacke H, Herold N, Schrumpf M, Gutknecht J, Wubet T, Buscot F, Daniell R. 2010. Horizon-specific bacterial community composition of german grassland soils, as revealed by pyrosequencing-based analysis of 16S rRNA genes. Applied and Environmental Microbiology 76:6751-6759. 186. Yergeau E, Sanschagrin S, Beaumier D, Greer CW. 2012. Metagenomic analysis of the bioremediation of diesel-contaminated canadian high arctic soils. PLoS ONE 7. 187. Margesin R, Feller G. 2010. Biotechnological applications of psychrophiles. Environmental Technology 31:835-844. 188. de Pascale D, De Santi C, Fu J, Landfald B. 2012. The microbial diversity of Polar environments is a fertile ground for bioprospecting. Marine Genomics 8:15-22. 189. Woyke T, Xie G, Copeland A, González JM, Han C, Kiss H, Saw JH, Senin P, Yang C, Chatterji S, Cheng JF, Eisen JA, Sieracki ME, Stepanauskas R. 2009. Assembling the marine metagenome, one cell at a time. PLoS ONE 4. 190. Siegl A, Kamke J, Hochmuth T, Piel J, Richter M, Liang C, Dandekar T, Hentschel U. 2011. Single-cell genomics reveals the lifestyle of Poribacteria, a candidate phylum symbiotically associated with marine sponges. ISME Journal 5:61- 70. 191. Youssef NH, Blainey PC, Quake SR, Elshahed MS. 2011. Partial genome assembly for a candidate division OP11 single cell from an anoxic spring (Zodletone spring, Oklahoma). Applied and Environmental Microbiology 77:7804-7814. 192. Tilman D, Wedin D, Knops J. 1996. Productivity and sustainability influenced by biodiversity in grassland ecosystems. Letters to Nature 379:718-720. 193. Yachi S, Loreau M. 1999. Biodiversity and ecosystem productivity in a fluctuating environment: the insurance hypothesis. PNAS U S A 96:1463-1468. 194. Allison SD, Martiny JBH. 2008. Resistance, resilience, and redundancy in microbial communities. Proceedings of the National Academy of Sciences 105:11512-11519. 195. Cravo-Laureau C, Hernandez-Raquet G, Vitte I, Jézéquel R, Bellet V, Godon J- J, Caumette P, Balaguer P, Duran R. 2011. Role of environmental fluctuations and microbial diversity in degradation of hydrocarbons in contaminated sludge. Research in Microbiology 162:888-895. 196. Sun MY, Dafforn KA, Johnston EL, Brown MV. 2013. Core sediment bacteria drive community response to anthropogenic contamination over multiple environmental gradients. Environmental Microbiology 15:2517-2531. 197. Deni J, Penninckx MJ. 2004. Influence of long-term diesel fuel pollution on nitrite- oxidising activity and population size of nitrobacter spp in soil. Microbiol Research 159:323-329.

166

198. Sverdrup LE, Ekelund F, Krogh PH, Nielsen T, Johnsen K. 2002. Soil microbial toxicity of eight polycyclic aromatic compounds: Effects on nitrification, the genetic diversity of bacteria, and the total number of protozoans. Environmental Toxicology and Chemistry 21:1644-1650. 199. Pereira R, Sousa J, Ribeiro R, Gonçalves F. 2006. Microbial indicators in mine soils (S. Domingos mine, Portugal). Soil and Sediment Contamination 15:147-167. 200. Miller JL, Sardo MA, Thompson TL, Miller RM. 1997. Effect of application solvents on heterotrophic and nitrifying populations in soil microcosms. Environmental Toxicology and Chemistry 16:447-451. 201. Chang S.W. HMR, Williamson K.J. 2002. Cooxidation of napthalene and other polycyclic aromatic hydrocarbons by the nitrifying bacterium, Nitrosomonas europaea. Biodegradation 13:373-381. 202. Sipila TP, Keskinen AK, Akerman ML, Fortelius C, Haahtela K, Yrjala K. 2008. High aromatic ring-cleavage diversity in birch rhizosphere: PAH treatment-specific changes of I.E.3 group extradiol dioxygenases and 16S rRNA bacterial communities in soil. ISME Journal 2:968-981. 203. Chang BV, Chang IT, Yuan SY. 2008. Biodegradation of phenanthrene and pyrene from mangrove sediment in subtropical Taiwan. Journal of Environmental Science Health 43:233-238. 204. Muangchinda C, Pansri R, Wongwongsee W, Pinyakong O. 2013. Assessment of polycyclic aromatic hydrocarbon biodegradation potential in mangrove sediment from Don Hoi Lot, Samut Songkram Province, Thailand. Journal of Applied Microbiology 114:1311-1324. 205. Bacosa HP, Suto K, Inoue C. 2012. Bacterial community dynamics during the preferential degradation of aromatic hydrocarbons by a microbial consortium. International Biodeterioration & Biodegradation 74:109-115. 206. Yergeau E, Bokhorst S, Kang S, Zhou J, Greer CW, Aerts R, Kowalchuk GA. 2012. Shifts in soil microorganisms in response to warming are consistent across a range of Antarctic environments. ISME Journal 6:692-702. 207. Grayston S, Campbell C, Bardgett R, Mawdsley J, Clegg C, Ritz K, Griffiths B, Rodwell J, Edwards S, Davies W. 2004. Assessing shifts in microbial community structure across a range of grasslands of differing management intensity using CLPP, PLFA and community DNA techniques. Applied Soil Ecology 25:63-84. 208. Allison V, Yermakov Z, Miller R, Jastrow J, Matamala R. 2007. Using landscape and depth gradients to decouple the impact of correlated environmental variables on soil microbial community composition. Soil Biology and Biochemistry 39:505-516. 209. Chung N, Alexander M. 1998. Differences in sequestration and bioavailability of organic compounds aged in dissimilar soils. Environmental Science and Technology 32:855-860. 210. Davis GB, Johnston CD, Patterson BM, Barber C, Bennett M. 1998. Estimation of biodegradation rates using respiration tests during in situ bioremediation of weathered diesel NAPL. Ground Water Monitoring and Remediation 18:123-132. 211. Gouch MA, Rhead MM, Rowland SJ. 1992. Biodegradation studies of unresolved complex mixtures of hydrocarbons: model UCM hydrocarbons and the aliphatic UCM. Organic Geochemistry 18:17-22. 212. Frysinger GS, Gaines RB, Xu L, Reddy CM. 2003. Resolving the unresolved complex mixture in petroleum-contaminated sediments. Environmental Science and Technology 37:1653-1662. 213. Booth AM, Sutton PA, Lewis CA, Lewis AC, Scarlett A, Chau W, Widdows J, Rowland SJ. 2007. Unresolved complex mixtures of aromatic hydrocarbons:

167

Thousands of overlooked persistent, bioaccumulative, and toxic contaminants in mussels. Environmental Science and Technology 41:457-464.

168