The Microbial and Molecular Diversity of Australian Biocrusts

THE MICROBIAL AND MOLECULAR DIVERSITY OF

AUSTRALIAN BIOCRUSTS

A DISSERTATION SUBMITTED

BY ANGELA MARY CHILTON B. Biological Science (Hons.)

IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR

THE AWARD OF DOCTOR OF PHILOSOPHY

AT THE

SCHOOL OF BIOTECHNOLOGY AND BIOMOLECULAR SCIENCES

THE UNIVERSITY OF NEW SOUTH WALES

SYDNEY, AUSTRALIA

APRIL 2018

THE MICROBIAL AND MOLECULAR DIVERSITY OF

AUSTRALIAN BIOCRUSTS

A DISSERTATION SUBMITTED

BY ANGELA MARY CHILTON B. Biological Science (Hons.)

IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR

THE AWARD OF DOCTOR OF PHILOSOPHY

AT THE

SCHOOL OF BIOTECHNOLOGY AND BIOMOLECULAR SCIENCES

THE UNIVERSITY OF NEW SOUTH WALES

SYDNEY, AUSTRALIA

SUPERVISORS

PROF. BRETT A. NEILAN AND BELINDA FERRARI

SCHOOL OF BIOTECHNOLOGY AND BIOMOLECULAR SCIENCES

THE UNIVERSITY OF NEW SOUTH WALES

SYDNEY, AUSTRALIA

The University of New South Wales Thesis/Dissertation Sheet

Surname or Family name: CHILTON

First name: ANGELA Other name/s: MARY

Abbreviation for degree as given in the University calendar: PHD

School: Biotechnology and Biomolecular Faculty: Science Sciences

Title: THE MICROBIAL AND MOLECULAR DIVERSITY OF AUSTRALIAN BIOCRUSTS

Abstract 350 words maximum: (PLEASE TYPE)

The increased accessibility of genomics has enabled unparalleled exploration of the microbial world. Taxonomic diversity is proving to be vast beyond expectations, with extrapolations from genomes indicating functional diversity to be even greater. In the face of multi-drug resistance, this promising frontier is galvanising the pursuit of microbes for novel anti-infectives and other socially beneficial compounds. Given the immense scope of the microbial world, there is an imperative to identify and prioritise promising avenues for future work. Arid soils have been identified as diversity hotspots for the nonribosomal megasynthase genes responsible for producing a large array of microbial bioactive compounds (secondary metabolites). Biocrusts are top-soil microbial niches within arid lands that present as compelling candidates for bioprospecting given their extremotolerance, enrichment with cyanobacteria, and use as model systems. This thesis takes direction from prior research to specifically target arid Australian biocrusts and profile their genetic capacity for small molecule production. Next generation sequencing was used to target the 16S rRNA gene as well as conserved domains within two nonribosomal biosynthetic pathways. Examination of biocrusts at the local scale showed visually distinct stages were defined by distinct bacterial communities. However, no difference was found in the biosynthesis genes between the stages. At the intra-continental scale, seasonality of precipitation was found to control biocrust community assembly and drive the distribution of secondary metabolite genes, suggesting possible functional roles involving water cycles. Overall, functional richness correlated with taxonomic richness. Network analysis was used to identify influential bacteria and genes as targets for future bioprospecting efforts. This thesis identified underlying patterns in taxonomic and functional diversity to further narrow the search field and contribute to a biosynthetic atlas of Australia.

Declaration relating to disposition of project thesis/dissertation

I hereby grant to the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or in part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all property rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation.

I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstracts International (this is applicable to doctoral theses only).

4th April 2018

Signature Date Witness Signature

The University recognises that there may be exceptional circumstances requiring restrictions on copying or conditions on use. Requests for restriction for a period of up to 2 years must be made in writing. Requests for a longer period of restriction may be considered in exceptional circumstances and require the approval of the Dean of Graduate Research.

FOR OFFICE USE ONLY Date of completion of requirements for Award:

ORIGINALITY STATEMENT

‘I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, or substantial proportions of material which have been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where due acknowledgement is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis. I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project's design and conception or in style, presentation and linguistic expression is acknowledged.’

Signed:

Date: 4th April 2018

Thesis abstract

Table of Contents

List of Figures ...... vi List of Tables ...... ix Abbreviations ...... xi CHAPTER 1: INTRODUCTION AND LITERATURE REVIEW ...... 1 THE EXTREMOBIOSPHERE ...... 2 BIOCRUSTS AS AN EXTREME MICROBIOME ...... 3 CASE STUDY: ARID AUSTRALIA ...... 4 BIOCRUST FORMATION AND MICROBIAL ECOLOGY ...... 5 CYANOBACTERIA ...... 9 A NOTE ON CYANOBACTERIAL TAXONOMY ...... 9 CYANOBACTERIA OF BIOCRUSTS: A METHODS PERSPECTIVE ...... 11 OTHER PROKARYOTES OF BIOCRUSTS ...... 17 BIOCRUST ADAPTATION FOR EXTREMOTOLERANCE ...... 19 WET-DRY ADAPTATION ...... 20 UV RADIATION TOLERANCE ...... 22 SPECIALISED MICROBIAL METABOLITES ...... 24 MICROBIAL NRPS AND PKS MEGASYNTHASES ...... 27 NONRIBOSOMAL PEPTIDES AND POLYKETIDES ...... 28 THE GENOMICS OF NRPS AND PKS ...... 29 EXTREMOTOLERANT BIOCRUST ORGANISMS AS POTENTIAL SOURCES OF NOVEL SPECIALISED MICROBIAL METABOLITES ...... 32 SCOPES AND OBJECTIVES OF THIS THESIS ...... 33 CHAPTER 2: BIOCRUST MORPHOLOGY IS LINKED TO MARKED DIFFERENCES IN MICROBIAL COMMUNITY COMPOSITION ...... 35 INTRODUCTION ...... 36 METHODS ...... 39 STUDY AREA AND FIELD SAMPLING ...... 39 MOLECULAR ANALYSIS ...... 40 STATISTICAL AND NETWORK ANALYSIS ...... 42 RESULTS ...... 43

iii

RICHNESS AND DIVERSITY OF MICROBIAL TAXA ...... 43 COMMUNITY COMPOSITION ...... 44 MICROBIAL INDICATORS OF BIOCRUST STAGE...... 44 NETWORK ANALYSIS ...... 48 DISCUSSION ...... 50 BIOCRUST STAGES AS A PROXY FOR MICROBIAL COMMUNITY STRUCTURE ...... 50 CYANOBACTERIAL “BLOOMS” PROMOTE BIOCRUST FORMATION ...... 51 BIOCRUST STAGES DEFINED BY MICROBE-MICROBE ASSOCIATIONS ...... 53 CHAPTER 3: BIOCRUSTS ASSEMBLY IS DRIVEN BY SEASONALITY OF PRECIPITATION ON AN INTRA-CONTINENTAL SCALE ...... 55 INTRODUCTION ...... 56 METHODS ...... 57 SAMPLE COLLECTION AND PROCESSING ...... 57 STATISTICAL ANALYSIS ...... 58 RESULTS ...... 61 BIOCRUST DIVERSITY AND COMMUNITY COMPOSITION ...... 61 BIOCRUST PHYLOGENY AND BIOGEOGRAPHY ...... 66 DISCUSSION ...... 68 INTRA-CONTINENTAL PATTERNS OF BIOCRUST MICROBIOME DIVERSITY ...... 68 BIOGEOGRAPHY OF BACTERIA WITHIN AUSTRALIAN BIOCRUSTS ...... 70 SEASONALITY OF PRECIPITATION EFFECTS BIOCRUST COMPOSITION ...... 71 CHAPTER 4: BIOCRUST SECONDARY METABOLISM IS DRIVEN BY SEASONALITY OF PRECIPITATION ON AN INTRA-CONTINENTAL SCALE74 INTRODUCTION ...... 75 METHODS ...... 76 AMPLIFICATION AND SEQUENCING OF NRPS AND PKS GENES ...... 76 BIOINFORMATIC ANALYSIS ...... 77 STATISTICAL ANALYSES ...... 78 RESULTS ...... 80 SECONDARY METABOLITE DIVERSITY WITHIN AUSTRALIAN BIOCRUSTS ...... 80

TAXONOMIC AND FUNCTIONAL CLASSIFICATION OF C AND KS DOMAINS ...... 82 COMMUNITY STRUCTURE AND BIOGEOGRAPHY OF KS AND C DOMAINS 86 RESERVOIRS OF BIOSYNTHETIC POTENTIAL AND INFLUENTIAL OTUs ... 88 DISCUSSION ...... 91 INTRA-CONTINENTAL DISTRIBUTION OF BIOSYNTHETIC GENES IS LINKED TO SEASONALITY OF PRECIPITATION ...... 91 RESERVOIRS OF BIOSYNTHETIC POTENTIAL AND INFLUENTIAL OTUs ... 94 POSSIBLE NRP AND PK FUNCTION IN BIOCRUSTS ...... 95 CHAPTER 5: CONCLUSIONS AND FUTURE DIRECTIONS ...... 98 RESEARCH MOTIVATIONS AND OBJECTIVES ...... 99 KEY FINDINGS ...... 100 BIOCRUSTS ARE DYNAMIC AT LOCAL AND INTRA-CONTINENTAL SCALES ...... 100 BIOSYNTHETIC CAPACITY IS LINKED TO SEASONALITY OF PRECIPITATION ON AN INTRA-CONTINENTAL SCALE ...... 101 SECONDARY METABOLITE POTENTIAL OF BIOCRUSTS ...... 102 FUTURE DIRECTIONS ...... 103 REFERENCES ...... 106 SUPPLEMENTARY MATERIAL ...... 145

List of Figures

CHAPTER 1

Figure 1.1 Photo of biocrust coverage at the Kalgooleguy common, Cobar, NSW, Australia Figure 1.2 Map of Aridity Index across Australia

Figure 1.3 Modular organisation of NRPS and PKS systems

Figure 1.4 Molecular structures of NRPs and PKs

CHAPTER 2

Figure 2.1 Collection site and photos of biocrust stages

Figure 2.2 nMDS of Cobar Bare, Early, Mid, and Late stage biocrusts

Figure 2.3 Relative abundance of the major bacterial classes grouped by phylum for Bare, Early, Mid, and Late stage biocrusts Figure 2.4 Co-occurrence networks for Bare, Early, Mid, and Late stage biocrusts CHAPTER 3

Figure 3.1 Sample site locations across Australia with seasonal precipitation gradient overlay Figure 3.2 Sample site diversity measures for taxa and for cyanobacteria and non-cyanobacteria Figure 3.3 Relative abundance of major bacterial classes grouped by phyla across Australia Figure 3.4 Heatmap of OTUs contributing to a cumulative of 5% of pairwise variation identified via SIMPER Figure 3.5 nMDS of biocrust samples according to site

Figure 3.6 nMDS of biocrust samples according to season of predominant rainfall Figure 3.7 Canonical analysis of principal coordinates (CAP) of Bray-Curtis dissimilarity with the factor Site set as a priori

Figure 3.8 nMDS of weighted UniFrac distances between biocrust samples

Figure 3.9 Linear regression of community differences based on composition and phylogeny to geographical distance. CHAPTER 4

Figure 4.1 Richness, evenness and diversity measures for condensation and ketosynthase domain OTUs from biocrusts across Australia Figure 4.2 Linear correlations between diversity measures of NRPS and PKS, NRPS and 16S rDNA, and PKS and 16S rDNA Figure 4.3 Dendrogram clustering of biocrust samples based on NRPS Bray- Curtis dissimilarity with taxonomic and domain class classification of OTUs Figure 4.4 Dendrogram clustering of biocrust samples based on PKS Bray- Curtis dissimilarity with taxonomic and domain class classification of OTUs Figure 4.5 Similarity of C domain sequences and KS domain sequences to the respective NaPDoS databases Figure 4.6 NMDS of Bray-Curtis dissimilarity distance for the Cobar data subset for NRPS and PKS OTUs Figure 4.7 NMDS of Bray-Curtis dissimilarity and UniFrac distance for whole NRPS and PKS OTU datasets Figure 4.8 Linear regression of community differences based on composition and phylogeny to geographical distance Figure 4.9 Scale free network of significant correlations of 16S rDNA to NRPS/PKS OTUs Figure 4.10 Ratio of relative abundance to number of network nodes at order level Appendix

Supplementary Co-occurrence networks for Bare stage biocrusts Figure A1.1

Supplementary Co-occurrence networks for Early stage biocrusts Figure A1.2

Supplementary Co-occurrence networks for Mid stage biocrusts Figure A1.3

Supplementary Co-occurrence networks for Late stage biocrusts Figure A1.4

vii

Supplementary nMDS of sample sites according to collection season, site climate Figure A1.5 class, which kit used for extraction, and collection year

viii

List of Tables

CHAPTER 2

Table 2.1 Mean richness, diversity and evenness of microbial OTUs across Bare, Early, Mid, and Late stage biocrusts Table 2.2 Indicator Species Analysis for biocrust stages

Table 2.3 Topology metrics and C-score measures derived from biocrust co- occurrence networks CHAPTER 3

Table 3.1 Sample site locations with symbol key and collection details

Table 3.2 PERMANOVA and PERMDISP results for factors

CHAPTER 4

Table 4.1 PERMANOVA and PERMDISP results for Bray-Curtis and UniFrac matrices Table 4.2 RELATE values for comparison of sequence datasets according to different distance methods Appendix

Supplementary Contribution of phyla to nodes and edges in co-occurrence network Table A1.1 Supplementary Significant differences between site 16S rDNA diversity measures Table A1.2 determine via ANOVA multiple-comparisons. Supplementary NRPS and PKS PCR conditions Table A1.3 Supplementary Primers used for degenerate PCR of NRPS Condensation and PKS Table A1.4 Ketosynthase domains Supplementary Significant differences between site NRPS diversity measures Table A1.5 determine via ANOVA multiple-comparisons Supplementary Significant differences between site PKS diversity measures Table A1.6 determine via ANOVA multiple-comparisons

Supplementary Taxonomy and classification of abundant NRPS and PKS OTUs Table A1.7 Supplementary Taxonomy and classification of highly connected 16S rDNA, NRPS, Table A1.8 and PKS network OTUs Supplementary Taxonomy and classification of most central 16S rDNA, NRPS, and Table A1.9 PKS network OTUs

Abbreviations

A adenylation domain ACP acyl-carrier protein domain ANOVA analysis of variance AT acyl-transferase domain C condensation domain CAP canonical analysis of principal coordinates EPS exopolymeric substances KS ketosynthase domain NGS next generation sequencing nMDS non-metric multidimensional scaling NRP nonribosomal peptide NRPS nonribosomal peptide synthetase OTU operational taxonomic unit PCR polymerase chain reaction PERMANOVA permutational multivariate analyses of variance PERMDISP permutational multivariate analyses of dispersion PK polyketide PKS polyketide synthase SIMPER similarity percentages breakdown UV ultraviolet radiation

1 CHAPTER 1: 2 INTRODUCTION AND LITERATURE REVIEW 3

4 THE EXTREMOBIOSPHERE

5 Prior to the mid-20th century, it was a common assumption that life on Earth was restricted to

6 environments observing a set of limited, anthropocentric conditions. However, in line with our

7 empirical exploration beyond these limits has been the discovery of an abundant and diverse

8 array of organisms, forcing the physical and metabolic boundaries of life to be successively

9 pushed further. These environments now encompass what is known as the extreme biosphere,

10 or extremobiosphere, and are populated by extremophiles (Rothschild and Mancinelli 2001).

11 There are many parameters by which an environment may be considered extreme including

12 temperature, pH, salinity, pressure, radiation, and osmotic stress. In many cases, extreme

13 environments present multiple stresses (Stan-Lotter et al 2013). Each factor has consensus

14 thresholds which define the extreme conditions (Rothschild and Mancinelli 2001). To be

15 considered a true extremophile, the organism must have its optimal growth within one of these

16 conditions. Organisms which have mesophilic optimal growth but are able to survive in

17 extreme conditions are considered extremotolerant. In keeping with the current terminology of

18 extremophiles, these organisms are here referred to as extremotolerants.

19 Microbial adaptation to extreme conditions manifests in a number of cellular and metabolic

20 strategies (Stan-Lotter et al 2013). Revelation of these molecular adaptations has had extensive

21 impact on the biological sciences. Of note has been the isolation of Taq, a thermostable DNA

22 polymerase from the extreme thermophile Thermus aquaticus (Chien et al 1976) which

23 revolutionised polymerase chain reaction (PCR) technology. Other examples of extreme

24 adapted biomolecules and strategies include osmolytes, increased membrane integrity, DNA

25 repair mechanisms, protein stability, UV pigments, polysaccharides and nutrient scavenging

26 (Wilson and Brimble 2009). Accordingly, extremophiles and extremotolerants present as a rich

27 source of novel biochemistry (Charlesworth and Burns 2015, Mandal and Rath 2014, Wilson

28 and Brimble 2009).

29 BIOCRUSTS AS AN EXTREME MICROBIOME

30 Biocrusts (also known as biological soil crusts) are complex communities of microorganisms

31 and non-vascular plants which form a continuous organic matrix within the top millimeters of

32 surface soil (Belnap et al 2003). Here, this definition excludes rock patinas (desert varnish),

33 above ground moss carpets and non-continuous rock-associated microbial communities such

34 as endoliths and hypoliths. Biocrusts may be considered as dry counterparts of microbial mats

35 (Angel and Conrad 2013, Gundlapally and Garcia-Pichel 2006, Rossi and De Philippis 2015).

36 Composed of a varying assemblage of cyanobacteria, lichen, fungi, heterotrophic bacteria,

37 algae and mosses (Bates et al 2012, Maier et al 2014, Nagy et al 2005, Zhao et al 2016),

38 biocrusts perform multiple ecosystem services such as carbon sequestration, atmospheric

39 nitrogen fixation, nutrient capture, soil stabilisation and also affect hydrological processes

40 (Belnap 2002, Bowker et al 2013a, Felde et al 2014, Zhang 2005, Zhao et al 2016). Biocrusts

41 are rich in phototrophic organisms which require direct access to sunlight and are often

42 constrained to sites where abiotic stress is high and vascular plant growth is limited (Bowker

43 2007), such as arid lands. Arid lands are typified by low and variable rainfall, high

44 evapotranspiration, high solar radiation, extreme temperature range and oligotrophic soils

45 (Pointing and Belnap 2012, Yaalon 1986), conditions which approach the boundaries for life

46 (Navarro-González et al 2003). Such environmental stresses result in characteristic

47 heterogeneous landscapes where biocrusts are able to colonise large expanses of open, inter-

48 plant areas (Bowker 2007). Globally, biocrusts cover approximately 12% of Earth’s terrestrial

49 surface (Rodriguez-Caballero et al 2018). Within arid lands, biocrusts are often the dominant

50 soil cover (Pointing and Belnap 2012) (Figure 1.1).

52 Figure 1.1 Photo of biocrust coverage at the Kalgooleguy common, Cobar, NSW, Australia.

53 CASE STUDY: ARID AUSTRALIA

54 As Australia migrated north towards the equator, breaking apart from the super-continent of

55 Gondwana, great inland seas occupying the centre dried up leaving a legacy of ancient salt

56 beds. The forests receded and were replaced by arid and semi-arid desert and grasslands.

57 Secure on a single tectonic plate, Australia has experienced very few soil enriching

58 geomorphological processes such as mountain uplifts, glaciation or volcanic activity.

59 Weathering has resulted in an exceptionally flat land mass. Subsequently, Australia has a high

60 proportion of ancient, weathered, saline, oligotrophic soils (McKenzie 2004). Prevailing south

61 westerly winds pass over cold ocean currents which repress cloud formation and the

62 subtropical high-pressure belt restricts cloud movement into the centre. From the east, the

63 Great Dividing Range creates a rain shadow. Often, the only precipitation systems with

64 enough power to reach the centre are cyclonic (Yaalon 1986). Together, these influences result

65 in arid and semi-arid conditions for approximately 70% of the Australian land mass (Figure

66 1.2) where top soils are dry for extended periods. Due to the coverage of latitudes, seasonal

67 precipitation discrepancies are observed where northern Australia experiences predominantly

68 summer rainfalls while the southern states receive most precipitation during winter months

69 (Eldridge 2001). The climate (scarcity of cloud cover), geology (flatness), and equatorial

70 location contribute to extreme levels of ultra violet (UV) radiation. In association, arid

71 Australia experiences high temperatures. Australian soils are primarily classified as thermic or

72 hyperthermic regimes (Reich 1997), being between 15-22°C and >22°C respectively, although

73 temperatures can reach 60°C and can undergo large changes over diurnal cycles (Pointing and

74 Belnap 2012). In light of the combination of multiple stresses imposed, arid and semi-arid top

75 soils may be considered extreme environments (Navarro-González et al 2003, Pointing and

76 Belnap 2012). As a community, biocrust organisms are able to thrive during metabolically

77 favourable wet periods and enter a state of protective dormancy during dry intervals and thus

78 are classified as extremotolerants.

79 BIOCRUST FORMATION AND MICROBIAL ECOLOGY

80 Biocrusts are typically formed through the stabilisation of soil surface particles via

81 exopolymeric substances (EPS) secreted by resident free-living bacteria and fungi (Garcia-

82 Pichel and Wojciechowski 2009 and see figures in Thomas and Dougill 2007). Fine, nutrient

83 rich aeolian particles also bind to the adhesive secretions and a thin surface layer of biological

84 cement forms (Zhang 2005). Upon stabilisation, additional microorganisms are able to

85 colonise and community biomass and diversity generally increase (Garcia-Pichel et al 2001,

86 Thomas and Dougill 2007, Yeager et al 2004). The presence and ratio of subsequent colonising

87 organisms is dependent on several factors which operate at different spatial scales(Bowker et al

88 2016). At microsite and local scales, biotic, topographic, and edaphic properties effect the

89 distribution and occurrence of biocrust constituents (Garcia-Pichel and Belnap 2001).

90 Although only millimetres to centimetres in depth, biocrusts have demonstrated stratified

91 organisation, with bacteria orientating themselves to microsite preferences (Angel and Conrad

92 2013, Garcia-Pichel et al 2003, Steven et al 2013a). At intra-continental scales, climatic forces

93 such as

95 Figure 1.2 Map of Aridity Index (AI, dimensionless) across Australia where AI=MAP/MAE 96 and MAP=Mean Annual Precipitation, MAE = Mean Annual Evapotranspiration. Based on 97 UNEP values. Map generated via Atlas of Living Australia (Williams et al 2012). 6

98 temperature and rainfall regulate the presence of lichens and mosses (Eldridge 2001, Ullmann

99 and Budel 2001). Globally, the degree of land mass isolation and dispersal limits are primary

100 factors explaining variance between hot and cold deserts around the world (Bahl et al 2011).

101 Biocrust morphology is highly variable and may be differentiated according to several features:

102 their dominant constituent organism (cryptogamic for non-vascular plants, microphytic for

103 algae, microbiotic for microbial) (Eldridge and Greene 1994, Wu et al 2009); their surface

104 topology (smooth, rugose, rolling, pinnacled)(Belnap 2001, Belnap et al 2001); their colour

105 (light and dark pigmented)(Belnap et al 2008a, Belnap et al 2008b); microstructure (Felde et al

106 2014); or combinations thereof (Pietrasiak et al 2013, Pócs 2009). These parameters are

107 intrinsic descriptors of developmental stage (Thomas and Dougill 2007). Generally, early-stage

108 crusts are dominated by filamentous cyanobacteria which form a thin, smooth, lightly coloured

109 surface crust. Mid-stage crusts are thicker and appear darker in colour due to an increase in

110 biomass and the addition of pigmented species of cyanobacteria as well as early colonisation by

111 mosses and lichens (Pietrasiak et al 2013, Zhang 2005). Late-stage biocrusts are typically

112 dominated by macro-components which increase thickness and surface roughness. At each

113 stage, wet-dry and freeze-thaw cycles can affect surface topography (Belnap 2001, Beraldi-

114 Campesi and Garcia-Pichel 2011). In many cases, notably where moisture events occur during

115 high temperatures which burn lichen thalli and moss rhizoids, biocrusts do not progress past

116 early- or mid-stages (Eldridge 2001).

117 A primary motivation for biocrust research has been the elucidation of the ecological services

118 provided by biocrusts and their influences in arid lands. Illumination of these roles has

119 provided outcomes such their use as indicators of arid land health, as model systems, as

120 strategies to control dust formation and the recognition of the importance of their

121 rehabilitation post disturbance (Belnap et al 2016a). A prominent finding has been that

122 community composition significantly affects function. In biocrusts, atmospheric nitrogen

123 fixation is performed by diazotrophic bacteria, most notably by heterocystic cyanobacteria

124 such as Nostoc and Scytonema (Pepe-Ranney et al 2016). A positive correlation between

125 enrichment with heterocystic cyanobacteria and increase in nitrogenase activity has been

126 observed, with dark or cyanolichen crusts exhibiting the highest rates compared to light or

127 moss dominated biocrusts (Belnap 2002, Su et al 2011, Yeager et al 2004). A similar

128 observation is seen with carbon fixation where more mature crusts have higher photosynthetic

129 rates (Housman et al 2006, Li et al 2012), likely due to greater chlorophyll a content from

130 increased diversity and biomass (Kidron et al 2015).

131 Through binding of the soil particles, biocrusts protect the soil surface from the erosive forces

132 of wind and water, preventing dust emission (thus nutrients loss) as well as stabilising rolling

133 dunes (Bowker 2007). Resistance to wind erosion increases with developmental stage and the

134 addition of macro-components (Belnap and Gillette 1998). Late stage crusts are also more

135 resistant to mechanical disturbances which fragment the protective crust layer, exposing loose

136 grains below to erosion (Leys and Eldridge 1998). The effect of biocrust upon hydrological

137 cycles is more complex (Felde et al 2014), involving several factors such as texture,

138 hydrophobicity, pore systems, surface roughness and sub-crust interactions which are

139 themselves dependent on the community profile. There is, therefore, an imperative to appraise

140 the community composition and natural histories of biocrusts to effectively understand their

141 ecosystem functions.

142 The macro component of biocrusts are reasonably well studied due to simpler taxonomy and

143 achievable, in situ identification and quantification of lichens and mosses (Eldridge et al 1997,

144 Rosentreter et al 2007). However, although research into the less conspicuous microbial

145 members has grown, greater diversity and complexity as well as difficulty in microbial

146 detection and classification has meant much of the microbial component of biocrusts remains

147 ill-defined. As pioneering members and major contributors to the organic matrix,

148 cyanobacteria are the most studied bacterial phylum of biocrusts.

149 CYANOBACTERIA

150 Cyanobacteria are an oxygenic photoautotrophic, monophyletic, deep-branching bacterial

151 phylum believed to be responsible for the Great Oxidation Event as well as the endosymbiosis

152 which lead to the evolution of photosynthetic eukaryotes. They are ubiquitous colonisers of

153 almost all habitats on Earth (Soo et al 2014), a phenomenon due in part to their diverse

154 morphologies and adaptive metabolisms. Although their primary metabolism is oxygen-

155 evolving photosynthesis, many are also able to metabolise storage polymers or environmental

156 sugars during dark periods (Stal and Moezelaar 1997). The capacity for nitrogen fixation is

157 observed unevenly across the phylum and is achieved either via specialised, oxygen-excluding

158 cells called heterocysts or by non-heterocystic cyanobacteria in localised anaerobic conditions

159 formed within sheaths (Stal 2001).

160 A NOTE ON CYANOBACTERIAL TAXONOMY

161 The nomenclature and systematics of cyanobacteria remain undecided, deliberations of which

162 are discussed elsewhere (Castenholz 2015, Komárek and Golubić 2004, Oren 2004, Pinevich

163 2015). Briefly, due to their initial classification as algae, the identification and taxonomy of

164 cyanobacteria fall under both the International Code for Botanical Nomenclature (ICBN) and

165 Nomenclature of Bacteria (ICNB). Although similar, each Code provides a separate system of

166 identification and classification which divide the phylum into 5 comparable Subsections 9

167 (Bacterial), or Orders (Botanical): The unicellular Subsection I (Chroococcales) and Subsection

168 II (Pleurocapsales), as well as the filamentous Subsection III (Oscillatoriales), Subsection IV

169 (Nostocales) and Subsection V (Stigonematales). While lower taxonomic ranks have

170 experienced much revision, the 5-division system has provided a relatively stable method in a

171 field where molecular data had largely been lacking. In recent decades, however, growth in

172 cyanobacterial entries in nucleotide and genome databases has enabled phylum wide

173 phylogenetic scrutiny of established groupings. This process has revealed the phylogenetically

174 unsupported nature of seemingly homogenous morphological features used to group taxa,

175 where common features may not have common descent but have possibly arisen and/or been

176 lost several times. Subsections (Orders) were revealed to be polyphyletic and not supported as

177 evolutionary lineages. In response, Hoffman et al (2005) proposed a polyphasic approach

178 which combines phylogenetic analysis supported by morphologically stable features such as

179 thylakoid arrangement to provide a new, botanical-based classification system. Four subclasses

180 are defined: Gloeobacterophycideae, Synechococcophycideae, Oscillatoriophycideae and

181 Nostocophycideae. The authors acknowledge this is yet to be a unifying system due to the still

182 insufficient data and low number of representatives for several clades. With the isolation of

183 new species and provision of additional genetic information, future revisions will be required.

184 The discord in cyanobacterial classification is not limited to the identification of field samples

185 and pure cultures. Given the low yield of cultivatable bacteria from the environment,

186 researchers are moving to culture-independent methods such as whole genome and whole

187 microbiome analysis through targeted and shot-gun metagenomics. Here, large datasets are

188 annotated via comparison to nucleotide databases and researchers may select from several

189 options, largely depending on the microbial community at question. While programs exist to

190 conduct mass heuristic queries of public databases such as the National Centre for

191 Biotechnology Information (NCBI) GenBank (Dowd et al 2005), there are several issues

192 regarding the validity of public submissions (Ashelford et al 2005, DeSantis et al 2006) which

193 result in potentially erroneous taxonomic assignments (Garcia-Pichel et al 1996). Also, a large

194 percentage of submissions are not annotated, named as ‘environmental’ or ‘unclassified’

195 (McDonald et al 2012). This is especially the case for novel environments where there is a

196 scarcity in cultured representatives (Abed et al 2012). To avoid misinformation in the

197 annotation of ribosomal RNA gene datasets, several groups have assembled curated databases,

198 including SILVA(Pruesse et al 2007), Ribosomal Database Project (RDP)(Cole et al 2014) and

199 GreenGenes (DeSantis et al 2006). Concerning cyanobacteria, each of these employ different

200 classification systems, with GreenGenes referencing the latest hierarchy proposed by

201 Hoffmann et al 2005 (McDonald et al 2012).

202 CYANOBACTERIA OF BIOCRUSTS: A METHODS PERSPECTIVE

203 Cyanobacteria are founding members and key engineers of biocrusts world-wide. The global

204 community is specific but diverse, with the majority of cyanobacteria falling within select

205 polyphyletic clades. Non-heterocystic simple filamentous cyanobacteria of the poorly resolved

206 morphogenera Microcoleus/Phormidium/Leptolyngbya/Oscillatoria are the major contributors to

207 biocrust biomass and structure, with several strains exhibiting the capacity for building

208 architecturally important supra-cellular ropes (Garcia-Pichel and Wojciechowski 2009). In

209 terms of abundance and distribution, Microcoleus, specifically Microcoleus vaginatus

210 (Oscillatoriales), is the most widely reported cyanobacterium of biocrusts globally (Starkenburg

211 et al 2011). The heterocystic and darkly pigmented Nostocalean Nostoc, Calothrix and Scytonema

212 are conspicuous members within established biocrusts and are attributed with darkening crust

213 colour (Belnap 2002). The culmination of early analyses of cyanobacterial diversity, primarily 11

214 based on morphological traits, is summarised in Chapter 12 Synopsis: Comparative Biogeography of

215 Soil Crust Biota within Biological Soil Crusts: Structure, Function and Management by Jayne Belnap and

216 Otto L. Lange (Büdel 2003). More recently, the scope of biocrust cyanobacteria was updated in

217 Biological Soil Crusts: An Organizing Principle in Drylands (Belnap et al 2016a). Across the globe,

218 over 70 crust-associated genera have been identified via morphological and molecular

219 approaches.

220 In early studies, researchers applied a range of conventional molecular methods to profile and

221 assess biocrust microbial diversity, often using multiple techniques in conjunction with

222 microscopy and phylogenetic analysis of cultured isolates. Molecular fingerprinting such as

223 denaturing gradient gel electrophoresis (DGGE) and terminal restriction fragment length

224 polymorphism (TRFLP) provide rapid assessment to compare the diversity of communities,

225 though are limited in resolution and their direct identification of individual members. DGGE

226 analysis of biocrusts using bacterial primers typically produces less than 30 bands (gross

227 genomic polymorphisms) per sample reflecting a lower diversity when compared to mesophilic

228 soil environments (Garcia-Pichel et al 2003, Zaady et al 2010). Sequencing of bacterial DGGE

229 PCR bands supports visual assessments of biocrusts as a cyanobacteria-enriched niche with

230 30-50% of band sequences reporting from the phylum. Comparison of DGGE sequences to

231 nucleotide databases confirms the ubiquity of simple filamentous types, notably Microcoleus

232 (M. vaginatus, M. steenstrupii), as well as the heterocystic genera Nostoc and Scytonema (Garcia-

233 Pichel et al 2001, Garcia-Pichel et al 2003, Gundlapally and Garcia-Pichel 2006, Nagy et al

234 2005, Zaady et al 2010). Molecular fingerprinting also provides a useful tool to track the

235 recovery of disturbed crusts or the development of artificial biocrusts (Bates et al 2012, Kuske

236 et al 2012, Li et al 2014, Nagy et al 2005, Zhang et al 2014). However, these methods have low

237 resolution and true diversity is likely masked by the dominance of highly abundant organisms

238 (Castillo-Monroy et al 2011, Dunbar et al 2000, Kuske et al 2002, Nubel et al 1999). The

239 construction of ribosomal RNA gene libraries (clone libraries) enables larger-scale screening of

240 individual members, improving resolution and allowing for direct identification. Ribosomal

241 RNA gene clone libraries of biocrusts using bacterial and cyanobacteria-specific primers have

242 been employed across the globe, including America (Redfield et al 2002), Africa (Dojani et al

243 2014), Middle East (Hagemann et al 2015), and China (Zhang et al 2014). These studies

244 document higher abundance levels for cyanobacteria (50-80%) than DGGE though different

245 sites and crust types were studied making comparisons between the two methods ineffectual.

246 A clear advantage is seen, however, where a marked increase in the number of ribosomal RNA

247 gene sequences generated (100’s compared to 10’s) has resulted in greater rates of species

248 detected within crust types (Abed et al 2010, Hagemann et al 2015, Yeager et al 2004). Rarer

249 crust-associated cyanobacterial genera identified include Acaryochloris, Brasilonema, Petalonema

250 and Pleurocapsa (Hagemann et al 2015).

251 More recently, conventional methods relying on capillary sequencing of segregated PCR

252 products are being succeeded by next generation, high through-put direct sequencing

253 technologies (next generation sequencing, NGS). These technologies compensate shorter read

254 lengths with the generation of thousands of sequences per sample, an advantage for evaluating

255 complex microbial communities (Kozich et al 2013, Sogin et al 2006). Steven et al (2012)

256 demonstrated the higher resolving and discerning power of increased sequence reads through

257 comparing targeted and shotgun sequencing approaches (see Fig 1 within reference). The

258 distinct bacterial communities of biocrusts and the root zones of the arid bush Larrea tridentate

259 were compared via small subunit ribosomal DNA libraries generated both by 454

260 pyrosequencing using targeted bacterial primers and by recruiting the ribosomal DNA

261 component of shotgun libraries. The targeted approach (pyroSSU) generated an average of

262 14460 reads per sample while an average of 1106 ribosomal DNA reads were retrieved from

263 the shotgun libraries (metaSSU, 0.2% of total sequences). A crucial difference was observed in

264 the relative abundances detected by each method, where pyroSSU revealed consensus ratios of

265 cyanobacteria (over 40% of community reads) while metaSSU apparently under reported

266 (approx. 14%) for biocrusts. Dissimilarity analysis (Bray-Curtis) showed that metaSSU also

267 resulted in weaker distinguishing power between the two communities. In this case, the

268 discrepancies noted may in part be due to the different annotation approaches in place for the

269 two methods. Public 16S ribosomal gene databases contain over 1 million sequences for direct

270 identification of targeted datasets where as full genome databases required for shotgun

271 sequence assignment are comparatively deficient. However, shotgun libraries allow for

272 concurrent functional analyses and are useful in identifying community responses to changing

273 environmental conditions.

274 Early NGS analyses were carried out using Roche 454 pyrosequencing technology. More

275 recently, a suite of Illumina instruments have come to monopolize the market, reflecting

276 continued progression in sequencing technologies. The majority of NGS studies are from the

277 Western deserts of America, e.g. Garcia-Pichel et al (2013), Steven et al (2015), Swenson et al

278 (2018), with several from China (Li et al 2014, Xiao and Veste 2017), Europe (Büdel et al 2014,

279 Maier et al 2014), Africa (Elliott et al 2014, Maier et al 2018), Australia (Abed et al 2012, Liu et

280 al 2017b), Israel (Angel and Conrad 2013), and the Arctic Circle (Steven et al 2013b). A variety

281 of different crust types have been profiled, from sand dunes (Liu et al 2014) to permafrost

282 soils (Steven et al 2013b)and from early (Elliott et al 2014) to late stages (Maier et al 2014).

283 These endeavours have contributed greatly to increasing the understanding of biocrust

284 community structure as well as revealing cryptic members and allowing for probing of the

285 factors affecting community differentiation. However, drawing direct comparisons and

286 contrasts of biocrusts globally is challenging. While sample collection methods appear to be

287 relatively standard, with studies collecting the top 0-2 cm of soil, there are several different

288 downstream approaches which can result in variable outcomes of operational taxonomic unit

289 (OTU) binning and taxonomy affecting overall interpretations. The bacterial 16S ribosomal

290 gene (16S rDNA) is approximately 1540 bp long (as per Escherichia coli U00096 notation) and

291 consists of nine hypervariable regions flanked by conserved motifs which serve as primer

292 targets for amplification. Due to the current generation limitations of amplicon lengths, NGS

293 technologies rely on amplifying partial regions incorporating one, two or three hypervariable

294 sections. Amplicon length, targeted variable region, and sampling depth can all affect diversity

295 measures (Cai et al 2013, Claesson et al 2010). Furthermore, all eubacterial 16S rDNA primers

296 have amplification bias and targeting different variable regions of the same community can

297 result in different relative taxon abundances (Claesson et al 2010). These issues are further

298 compounded by the choice available for classification datasets which can result in differing

299 accuracies for taxonomic assignments (Werner et al 2012). Among sampling factors, this may

300 account for the marked difference observed in abundance of cyanobacteria of biocrusts from

301 the same sites as detected by Steven et al (2013a) and Steven et al (2013b). It is therefore

302 difficult to ascertain to what extent differences observed reflect true community variation or

303 are manifestations of divergent data analysis.

304 Yet, consistent findings have emerged. NGS of biocrust communities has resulted in a

305 dramatic increase in the sampling depth achieved and indicates genotypic diversity is masked

306 by homogenous phenotypes, especially for cyanobacteria. Even so, the number of species

307 observed in biocrusts as well as OTU counts are still considered lower than for other, non-

308 extreme soil environments (Angel and Conrad 2013). Several studies from various

309 environments globally have shown clear niche differentiation between crust and sub-crust

310 microbial communities, supporting biocrusts as both a physically and phylotypically distinct

311 microbiome (Elliott et al 2014, Steven et al 2013a). Consideration of cyanobacterial

312 abundances show significant reductions from relatively high in the crust (8-45%) to generally

313 less than 1% of sequences in the sub-crust, highlighting the role of cyanobacteria as biocrust

314 defining members and their selection for direct access to light. In many cases, the probing of

315 sequencing datasets focus on functional or general community patterns and resolution to lower

316 ranking taxon levels of cyanobacteria is inconsistently performed. Where lower rankings are

317 quantified, filamentous morphotypes of the Oscillatoriales subsection retain their established

318 pre-eminence as significant contributors to community structure with Microcoleus vaginatus

319 remaining the most commonly reported cyanobacterium. Similarly, Scytonema and Nostoc prevail

320 as the top genera of the second most abundant subsection, Nostocales. Chroococcidiopsis appears

321 to be the sole representative of uni-cellular morphotypes, a genus relocated to the Subclass

322 Oscillatoriophycideae with phylogenetic and thylakoid relatedness to Microcoleus (Hoffmann et

323 al 2005). These findings may be considered congruous extensions of previous assertions

324 derived from earlier studies. However, the particular benefits of NGS data can be seen in the

325 statistical power available to detect community patterns and discern important factors affecting

326 microbiome structure. For example, Garcia-Pichel et al (2013) showed important differences

327 between the dominant cyanobacteria of biocrusts across a continental scale, whereby crusts of

328 cooler deserts were dominated by Microcoleus vaginatus while warmer desert crusts consisted

329 primarily of Microcoleus steenstrupii (Garcia-Pichel et al 2013). These findings have implications 16

330 for managing biocrusts and arid lands at the onset of climate change (Rodriguez-Caballero et al

331 2018).

332 As noted by the authors, these insights relied on prior culturing efforts. While high through-

333 put methods are becoming the gold standard, interpretations of molecular data are only as

334 robust as their annotation. Intensive culturing efforts of cyanobacteria from biocrusts by

335 several groups has contributed greatly to enriching annotation databases, providing ‘cultivation

336 anchors’ to improve the accuracy and efficacy of classifying sequences (Dojani et al 2014,

337 Garcia-Pichel 2008). Garcia et al (2013) further demonstrated the value of attaining isolates,

338 where the M. vaginatus/M. steenstrupii division observed through molecular data was validated

339 via microcosm experiments. In 2011, the genome of M. vaginatus FGP-2 was published,

340 enabling exploration of the genes and pathways responsible for arid land adaptation as well as

341 informing investigations of biochemical and physiological traits contributing to microbial

342 ecology (Baran et al 2015, Starkenburg et al 2011).

343 OTHER PROKARYOTES OF BIOCRUSTS

344 As the most conspicuous and functionally evident bacterial members of biocrusts,

345 cyanobacteria have received the greatest attention for study. The relegation of the non-

346 phototrophic prokaryotic component of biocrusts is exemplified by their absence in Biological

347 Soil Crusts: Structure, Function and Management (Belnap and Lange 2001). More recently, in

348 keeping with the development of technologies enabling their more accessible study, interest of

349 other bacterial phyla and their contribution to biocrust function has grown (Maier et al 2016,

350 Nunes da Rocha et al 2015). Over 30 bacterial phyla have been detected within biocrusts

351 (Steven et al 2013b), though the vast majority of sequences often belong to a select list.

352 Ubiquitous non-cyanobacterial phyla which feature prominently include Proteobacteria 17

353 (primarily Alpha, then few Beta, Gamma and Deltaproteobacteria with no

354 Epsilonproteobacteria), Actinobacteria, Acidobacteria and Bacteroidetes. Less prevalent and

355 abundant phyla include Deinococcus-Thermus, Chloroflexi, Firmicutes, Verrucomicrobia,

356 Planctomycetes and Gemmatimonadetes. These phyla represent a broad functional capacity of

357 chemoorganoheterotrophs as well ammonia-oxidizing chemoorganoautotrophs. While

358 cyanobacterial sequences identified in molecular studies generally have high similarity to

359 known submissions within databases, especially dominant Microcoleus types, other bacteria have

360 tended to have lower sequence similarities (Garcia-Pichel et al 2003). Culturing efforts

361 (Gundlapally and Garcia-Pichel 2006, Nunes da Rocha et al 2015) have served to enrich

362 biocrust representatives and has resulted in several novel bacteria isolated from biocrusts,

363 including Belnapia moabensis (Alphaproteobacteria)(Reddy et al 2006), Modestobacter versicolor

364 (Actinobacteria) (Reddy et al 2007), Dyadobacter crusticola (Bacteroidetes) (Reddy and Garcia-

365 Pichel 2007), Pseudomonas asuensis (Gammaproteobacteria) (Reddy and Garcia-Pichel 2015), as

366 well as Sphingomonas mucosissima and Sphingomonas desiccabilis (Alphaproteobacteria) (Reddy and

367 Garcia-Pichel 2007). Likely owing to their lower diversity observed within biocrusts, Garcia

368 (2015) recently demonstrated a high cultivation rate of non-phototrophic bacteria from

369 biocrusts, attaining 8% of representatives from metagenomic analysis (Nunes da Rocha et al

370 2015).

371 Several of the bacterial taxa routinely detected within biocrusts are considered

372 extremotolerants. Rubrobacteridae is a deep-branching Actinobacteria subclass which are

373 common to arid soils and rock surfaces (Holmes et al 2000). Sequences from American

374 (Gundlapally and Garcia-Pichel 2006, Kuske et al 2012), Israeli (Angel and Conrad 2013) and

375 Australian biocrusts (Abed et al 2010) have reported as most similar to Rubrobacter species,

376 organisms which exhibit tolerance of extremely high levels of ionising radiation while also

377 proving moderately thermophilic (Albuquerque and da Costa 2014). The phylum

378 Deinococcus-Thermus, observed in America (Nagy et al 2005), Oman (Abed et al 2010) and

379 China (Li et al 2013) also contains radiation tolerant bacteria. Indeed, biocrusts are rich in

380 phyla which have been grouped into the super-taxon Terrabacteria, consisting of

381 Cyanobacteria, Actinobacteria, Firmicutes, Chloroflexi and Deinococcus-Thermus (Battistuzzi

382 et al 2004, Battistuzzi and Hedges 2009). Terrabacteria is a proposed evolutionary group of

383 deep-branching phyla which are suggested first colonisers of ancient land and is enriched with

384 taxa which exhibit adaptation to environmental hazards. Archaea appear to be less prevalent in

385 biocrusts, contributing <1-5% of 16S rRNA gene copies and display low diversity (Angel and

386 Conrad 2013, Nunes da Rocha et al 2015, Soule et al 2009). The most prevalent Archean

387 phylum identified is Crenarchaeota which is enriched in the sub-crust profile.

388 BIOCRUST ADAPTATION FOR EXTREMOTOLERANCE

389 Arid soils are typified by low levels of biologically available carbon and nitrogen. Under these

390 nutrient limited conditions, it is evident that photoautotrophic and diazotrophic growth

391 strategies provide inherent advantage. Oxygenic photosynthesis is the primary mode of photon

392 harvesting within crusts where, bacterially, cyanobacteria are the sole oxygenic photosynthetic

393 phylum. The contribution of anoxygenic photosynthesis appears marginal, performed by

394 Alphaproteobacterial genera such as Methylobacterium, Belnapia, Muricoccus and Sphingomonas and

395 members of the phototrophic branch of the phylum Chloroflexi (Nagy et al 2005) under

396 microaerophilic conditions of the lower or sub-crust (Elliott et al 2014, Steven et al 2013a).

397 Nitrogen fixation follows a similar observation whereby cyanobacteria are attributed as the

398 major diazotrophic component and contribution by other phyla is provided at lower levels

399 (Pepe-Ranney et al 2016). Determining the global contributions of biocrusts to carbon and

400 nitrogen cycles is difficult due to both the variability in global community structure and

401 inconsistency of metabolically active time. Estimations of the contributions made by

402 cryptogamic covers, which include biocrusts place values at 7% of the annual global carbon

403 uptake by terrestrial vegetation and approximately 40% of the annual biological nitrogen

404 fixation (Elbert et al 2012).

405 WET-DRY ADAPTATION

406 The distillation of decades of extremobiosphere research establishes that despite the seemingly

407 overwhelming severity of physical or chemical extremes, where there is liquid water there is life

408 (Rothschild and Mancinelli 2001). The removal of water induces conformational changes in

409 biomolecules which can result in denaturation and cell death. Arid adapted organisms must

410 contend with extreme fluxes in water availability as moisture events are sporadic and

411 interspersed with long periods of drought where microorganisms may endure desiccation.

412 Biocrust organisms may be viewed as poikilohydric and demonstrate several dehydration

413 tolerance strategies. Cyanobacteria are major architects of biocrusts, contributing bulk organic

414 matter via excretion of EPS (Rossi and De Philippis 2015). Organic matter increases overall

415 soil water holding capacity and retention, while EPS which surrounds both producing and

416 associated microorganisms is able to capture and hold water in the localized cell environment

417 for prolonged metabolism. Cyanobacterial strains also demonstrate behavioural adaptation,

418 physically seeking water via hydrotaxis (Pringault and Garcia-Pichel 2004). To tolerate

419 desiccation, cells may enter a state of metabolic dormancy or reduced function (Rajeev et al

420 2013). Anhydrobiosis is a process where water is replaced with compatible solutes such as the

421 disaccharide trehalose, preserving the structural integrity of biomolecules (Billi 2012).

422 The response of dry biocrust communities to wetting has been examined from several aspects.

423 Early analysis showed an initial pulse in CO2, contrary to an intuitive expectation of the release

424 of O2 given the dominance of cyanobacteria and their primary metabolism of oxygenic

425 photosynthesis (Lange 2003). Crusts then took either one or two days to achieve full activity,

426 depending on the extent of the dry period prior. Angel et al (2013) observed a scarcity of O18

427 labelled cyanobacterial rRNA within dry and recently wetted biocrusts from Israel where

428 Sphingobacteriales and Actinomycetales were dominant. Cyanobacterial abundances only

429 reached anticipated levels one to three weeks later. This is in contrast to their 16S rDNA

430 qPCR results which supply immediate quantification of ribosomal genes and showed up to

431 50% of dry biocrusts qPCR reads as cyanobacterial. The authors note this discrepancy may

432 suggest cyanobacteria are present in abundant numbers within dry crusts though remain

433 dormant during initial periods of wetting. This suggestion is supported by observations from

434 Australian biocrust-associated cyanobacteria which resisted activation during the dry season,

435 despite experimental wetting (Williams et al 2014). Rather, the cyanobacteria were shown to

436 revive their photosystems during the wet season. Here, the authors propose a role of EPS as

437 an infiltration barrier, protecting the organisms from premature resurrection. Further to this,

438 the cyanobacterium Leptolyngbya ohadii isolated from Israeli biocrusts was shown to initiate

439 dehydration preparation upon exposure to dawn sunlight (Oren et al 2017), showing diurnal

440 cycles are also influential. In contrast to delayed re-activation, other findings have indicated a

441 rapid response to wetting (Strong et al 2013). Transcriptomic analysis of the cyanobacterium

442 M. vaginatus in response to diurnal cycles of wetting experiments revealed key stages in reviving

443 from dormancy (Rajeev et al 2013). The immediate transcriptional response showed induction

444 of DNA repair mechanisms and regulatory genes with photosynthetic pathways activated

445 within 1 h. The data also revealed dehydration preparation strategies, including induction of 21

446 oxidative and photo-oxidative stress responses, the production of storage molecules and genes

447 for membrane unsaturation.

448 Accurate responses to precipitation are critical for the survival of biocrust organisms (Reed et

449 al 2012). Two long-term studies have shown that altering the frequency and volume of

450 precipitation events significantly altered biocrust composition and reduced coverage over six

451 (Steven et al 2015) and fifteen years (Ferrenberg et al 2015). Research on desert mosses

452 suggests this is likely due to short wetting times resulting in failure to synthesis enough new

453 carbon storage molecules, effectively leading to carbon starvation for the next revival cycle

454 (Coe et al 2012). Together, these studies revealed an enigmatic complexity in the molecular

455 adaptation to extreme water fluctuations and the critical need to respond effectively to arid

456 conditions.

457 UV RADIATION TOLERANCE

458 Ultraviolet radiation between 280 nm and 400 nm reaches the earth’s surface where it can

459 interact with biological systems, causing molecular damage and cell death. Biocrusts are

460 restricted to environments with direct access to light and are obligatorily co-exposed to intense

461 harmful UV wavelengths. Biocrust communities have several strategies to counter the

462 deleterious effects of UV exposure. Many responses are related to DNA repair and reactive

463 oxygen species (ROS) detoxification (Rajeev et al 2013). It is interesting to note the co-

464 occurrence of radiation tolerance alongside desiccation tolerance of polyextremophilic

465 organisms. This is attributed to the similar molecular coping mechanisms (Billi et al 2000).

466 Within aquatic microbial mats, behavioural adaptations of endemic Microcoleus spp. have been

467 observed where organisms undertook vertical migration in order to protect themselves from

468 UVB exposure (Bebout and Garcia-Pichel 1995). Specialised microbial metabolites can act as

469 UV-absorbing pigments and are considered as bacterial sunscreens.

470 Scytonemin is the most commonly occurring indole-alkaloid within cyanobacteria,

471 documented in over 300 species across the phylum though has yet to be found in planktonic

472 species (Gao and Garcia-Pichel 2011, Rastogi et al 2015). The most prominent scytonemin

473 producing genera of biocrusts generally come from the Subclass Nostocophycideae, including

474 Scytonema, Nostoc and Calothrix. Scytonemin is excreted from the cell into the extracellular

475 sheath where its yellow-brown colour tints the appearance of the organism. Its structure has

476 been resolved as consisting of heterocyclic rings fused with carbon double bonds which allow

477 for strong UVA absorption. The biosynthetic pathway remains to be fully elucidated though

478 UVA has been shown to induce production (Gao and Garcia-Pichel 2011). Mycosporine-like

479 amino acids (MAAs or simply mycosporines) are also considered microbial sunscreens, though

480 consist of a diverse family of structures which have other proposed functional roles across the

481 domains of life. Mycosporines are small molecules consisting of a central cyclohexene core

482 where UV absorption occurs. There are three core genes in MAA biosynthesis which produce

483 Mycosporine-glycine, a template intermediary which is modified to produce a range of

484 structures. This pathway has been understood through the elucidation of the biosynthesis of

485 shinorine (Balskus and Walsh 2010). The effectiveness of scytonemin and MAAs as

486 environmental sunscreens has received the attention of biotechnologists for their potential

487 application in pharmaceuticals and biomedical research (Balskus and Walsh 2010, Rastogi et al

488 2015). However, the molecular potential of photoprotective pigments from microbiomes is

489 evidently under explored. For example, Modestobacter versicolor, an actinobacterium isolated from

490 biocrusts of the Colorado Plateau, has demonstrated pigment variation dependent on carbon

491 availability (Reddy et al 2007). The melanin-like compound, its biosynthesis and environmental

492 regulation have yet to be fully explored.

493 SPECIALISED MICROBIAL METABOLITES

494 Specialised microbial metabolites (SMs), commonly referred to as secondary metabolites, are

495 small molecules which comprise the ecologically-adapted branch of the host metabolome

496 (Davies and Ryan 2012). Secondary metabolite chemical scaffolds are extremely diverse and

497 encompass the major biomolecule groups (peptides, lipids and carbohydrates) as well as classes

498 such as polyketides and alkaloids (Wilson and Brimble 2009). Post-translational modification

499 incorporating a vast range of additional functional groups with different permutations of

500 stereochemical arrangements further increases chemical complexity. The resulting extensive

501 structural diversity gives rise to immense functional diversity and specialised adaptation,

502 including to extreme environments (Wilson and Brimble 2009) e.g. scytonemin (Gao and

503 Garcia-Pichel 2011). However, while the ecological functions of some SMs have been

504 identified (O'Brien and Wright 2011), in most cases the native function remains unknown or

505 speculative. This is largely due to many SMs exhibiting non-native bioactivity, focusing

506 research to their exploitation for pharmaceutical and bioindustrial applications rather than

507 determining their chemical ecology (Cragg and Newman 2013, Felnagle et al 2008). Due to

508 their bioactivity, many SMs are classed as natural products.

509 SMs as natural products have pervaded human history and altered the course of civilisation.

510 There are examples of traditional medicines from all cultural groups which were or still are

511 believed to have healing properties. However, while these crude treatments may have shown

512 some efficacy, the active agents remained unknown and thus their application limited. For

513 example, while 19th century Arab horse riders knew to rub the mould from damp saddles into 24

514 the wounds of their horses to help them heal, it was only through research in the 20th century

515 that the mechanism of action was elucidated to be via a small bioactive molecule, penicillin,

516 produced by the fungus Penicillium notatum (Schaefer 2014). This revelation ushered in the era

517 of antibiotics. En masse, microorganisms were fermented and compounds extracted, purified

518 and tested for inhibition against a multitude of pathogens. Essential anti-infectives were

519 discovered and classed according to their modes of action. Concurrently, SMs were found to

520 have therapeutic applications for a range of other health conditions with the development of

521 statins, antitumor agents, immunosuppressants and as biological probes (Newman et al 2015).

522 Between 1981 and 2014, 48% of the American Food and Drug Administration (FDA)

523 approved drugs were natural products or their derivatives (Newman and Cragg 2016).

524 However, many of the recently developed anti-infectives are for pathogens which have

525 evolved resistance to the first antibiotics employed. Medical practitioners often rely on

526 prescribing second or third generation variations in order to treat patients while an increasing

527 number of pathogens are deemed multi-drug resistant or untreatable. Disconcertingly, the rise

528 in antibiotic resistance was coupled with a decline in natural SMs discovery efforts.

529 There is a high financial cost associated with the natural product discovery pipeline as the

530 return rate of clinically relevant drugs from the exploration of hundreds of thousands of

531 compounds is notably low. In addition with high rates of rediscovery, pharmaceutical

532 companies shifted focus to procuring new lead compounds from combinatorial chemistry

533 approaches (Newman and Cragg 2012). Based on molecular targets, combinatorial chemistry is

534 able to produce large libraries of synthetic compounds for high through-put screening,

535 strategies which were anticipated to increase success rates while decreasing costs. Yet, the most

536 successful synthetic libraries were those which had basis in natural product scaffolds. In turn,

537 combinatorial libraries have become smaller and ‘more natural product like’.

538 Despite efforts, the proliferation of antibiotic resistance remains ahead of the advent of new

539 anti-infective compounds. Crucially, the development of new molecular and genetic techniques

540 has enabled a resurgence in SMs exploration and discovery (Harvey et al 2015). Breakthroughs

541 and reduced costs in sequencing technologies as well as identification of the genetic basis for

542 SMs biosynthesis has allowed us to begin tapping into the genetic treasure chest of microbes.

543 Metagenomic and whole genome sequencing has revealed a significant finding: the vast

544 majority of SMs genes are cryptic or silent, i.e. not expressed under so-called normal

545 conditions (Calteau et al 2014, Letzel et al 2013). When taken in conjunction with the

546 observation that the majority of microbes themselves remain averse to culturing, it is clear the

547 secondary metabolite potential of microbes remains vastly under characterised.

548 There are several methods to overcome limitations of microbial cultivability. Nicholas et al

549 2010 developed ichip technology; in situ cultivation of microbial communities which allows for

550 high through-put assay screening (Nichols et al 2010). Subsequently, Teixobactin, the first

551 antibiotic identified from a novel class in several decades, was discovered, (Ling et al 2015).

552 While these methods have proved fruitful, they do not adequately address the cryptic nature of

553 SMs production. A substantial number of SMs are synthesised non ribosomally via

554 nonribosomal peptide synthetases (NRPS) and polyketide synthases (PKS). The genes for

555 these megasynthases present as useful targets for a genomics-driven approach to enumerate

556 and assess the SMs genetic potential of whole microbial communities.

557 MICROBIAL NRPS AND PKS MEGASYNTHASES

558 NRPS and PKS employ an assembly-line order of biosynthesis, producing complex small

559 molecules from a select group of relatively simple building blocks. They share a multi-modular

560 template where NRPS and PKS modules comprise of functionally analogous core and

561 accessory catalytic domains (Figure 1.3). Modules are discrete and form genetically mobile

562 units able to assemble in vast permutations, including with each other to form NRPS/PKS

563 hybrid pathways (Fischbach and Walsh 2006, Mootz et al 2002). The order in which modules

564 catalyse chain elongation may either be a linear progression, producing a molecular sequence

565 which mirrors the module sequence (co-linearity rule), or may be iterative, where modules are

566 used multiple times in the production of one compound resulting in a molecular sequence

567 distinct from module order. According to this division, the megasynthetases are grouped as

568 NRPS Type A linear, Type B iterative and Type C nonlinear and PKS Type I non-iterative and

569 Type II iterative (Mootz et al 2002, Weissman 2009). PKS Type III is further distinguished

570 based on comprising a single domain, ketosynthase, with narrow substrate specificity (Shen

571 2003). Further information on the structures and mechanisms of NRPSs and PKSs is available

572 in specialised reviews (Fischbach and Walsh 2006, Meier and Burkart 2009, Strieker et al 2010).

573 Linear production of nonribosomal peptides (NRPs) and polyketides (PKs) requires three

574 module types: initiation, elongation, and termination. Elongation modules perform three core

575 processes for the addition of each substrate to the growing molecular chain: substrate

576 activation/loading, mobile tethering of substrates and intermediaries, and the formation of

577 new bonds between these entities. Each process is executed by single-reaction domains within

578 the elongation module. Within NRPS modules, these are the adenylation (A) domain, peptidyl-

579 carrier protein domain and the condensation (C) domain, respectively. Within PKS modules,

580 these are the acyl-transferase (AT) domain, acyl-carrier protein (ACP) domain and the 27

581 ketosynthase (KS) domain, respectively. Initiation modules are similar to elongation modules

582 though lack the bond-forming C and KS domains. Both NRPS and PKS systems have

583 termination modules comprising of a core thioesterase domain. Within each module, accessory

584 domains may also be present performing tailoring reactions such as cyclisation, oxidation,

585 reduction, dehydration and methylation.

586 NRPSs draw from a pool of over 500 monomer substrates including proteinogenic and non-

587 proteinogenic amino acids, fatty acids and alpha-hydroxy acids (Strieker et al 2010), resulting in

588 peptide products more structurally diverse than ribosomally produced peptides. Often, non-

589 proteinogenic residues are incorporated at the terminal ends of the peptide product resulting in

590 greater protection from proteolytic digest (Mootz et al 2002). Modular PKSs are enzymatically

591 similar to Type 1 fatty acid synthase systems and share evolutionary history (Arakawa 2012,

592 Jenke-Kodama et al 2005). Each incorporates Coenzyme A substrates, primarily malonyl-CoA,

593 and employ AT, ACP and KS domains. A distinction between fatty acid synthase and PKS

594 systems is the arbitrary inclusion of the reducing enzymes ketoreductase, dehydratase and

595 enoylreductase within PKS pathways resulting in greater structural diversity.

596 NONRIBOSOMAL PEPTIDES AND POLYKETIDES

597 The products of NRPS and PKS pathways, nonribosomal peptides (NRPs) and polyketides

598 (PKs), respectively, present as a double-edged sword to human interests. The list of anti-

599 infective and industrially significant NRPs and PKs is extensive and revolutionary. Major

600 therapeutic examples include penicillin, vancomycin, erythromycin and lovastatin (Figure 1.4a)

601 (Fischbach and Walsh 2006, Mootz et al 2002). However, the potent bioactivities of NRPs and

602 PKs also enact deleterious effects on humans, ecological systems and industry. Of particular

603 note are cyanobacterial toxins (cyanotoxins) which include potent neurotoxins, hepatotoxins,

604 cytotoxins and dermatoxins (Figure 1.4b) (Pearson et al 2010). In addition to serious health

605 impacts, local industries such as aquaculture, agriculture and tourism can suffer significant

606 economic losses (Woodhouse et al 2014). Beyond determining their health and industry

607 impacts, there is a need to understand the ecology and biosynthesis of NRPs and PKs. As with

608 the majority of SMs, the native function (or functions) of many NRPs and PKs is unknown or

609 speculative. Understanding why hosts produce SMs can identify warning signals for the

610 production of harmful compounds and assist in developing preventative management

611 strategies (Pearson et al 2016). Complementary to this is has been the deciphering of

612 biosynthetic pathways which has opened the lid on the nonribosomal peptide and polyketide

613 metabolism.

614

615 Figure 1.3 Modular organisation of NRPS and PKS systems, A=adenylation, ACP=acyle 616 carrier protein, AT=acyltransferase, C=condensation , DH=dehydrogenase, 617 ER=enoylreductase, KR=β-ketoreductase, KS=ketosynthase, T=thioester, , Te=thioesterase.

618

619 THE GENOMICS OF NRPS AND PKS

620 There are several advantageous features of NRPS and PKS genetics which lend them to

621 fruitful molecular biology and bioinformatic strategies. Their genes are encoded in large

622 operons or genes clusters allowing for whole megasynthetases to be identified, manipulated

623 and heterologously expressed (Ongley et al 2013). Core catalytic domains exhibit high genetic

624 conservation, such as the PKS ketosynthase (Jenke-Kodama et al 2005) and the NRPS

625 adenylation (Neilan et al 1999) and condensation (Rausch et al 2007) domains, which serve as

626 primer targets. Furthermore, in many cases the specificity-conferring code of the binding-

627 pocket moieties of A and AT domains mean substrate incorporation can be predicted from

628 amino acid sequence (Stachelhaus et al 1999). Megasynthase genetics and SM chemistry form a

629 complementary interface whereby each informs the other. The co-linearity rule and specificity

630 conferring code mean SM structure can be predicted and conversely, given a structure,

631 pathways and gene cluster arrangement can be proposed (Chu et al 2016). These conducive

632 attributes have enabled the development of sensitive PCR assays for the detection of toxic

633 cyanobacterial species (Jungblut and Neilan 2006), site-directed mutagenesis to exploit

634 megasynthase machinery for the production of “unnatural” natural products (Williams 2013),

635 and revealing the diversity and phylogeny of NRP and PK production (Calteau et al 2014, Shih

636 et al 2013). Sequencing technologies have been critical in this process and the proliferation of

637 genomic and metagenomic data on public databases has enabled powerful bioinformatic

638 strategies.

639 There are two main genomic approaches for NRPS and PKS biosynthetic gene cluster

640 discovery, targeting either single-organism genomes (Micallef et al 2015) or screening whole

641 microbiomes (Lemetre et al 2017). Bioinformatic mining of individual genomes revealed an

642 important finding; the genetic potential of microorganisms is much greater than their observed

643 chemistry (Calteau et al 2014). For example, the Actinobacterium Streptomyces avermitilis was

644 found to have 25 more gene clusters than products detected (Letzel et al 2013). Revelation of

645 646 Figure 1.3 Molecular structures of natural products a) with therapeutic activity b) as toxins c) 647 from extremophiles d) from biocrust-associated cyanobacteria.

648 these silent, or cryptic, gene clusters can enable genetic-guided isolation of novel compounds

649 (see references 51-57 in Letzal, 2013). However, it is well established that the vast majority of

650 microbes remain adverse to culturing and isolation, thus the true scope of novel SM

651 biosynthesis and chemistry is unknown. In order to assess the SM genetic potential of

652 recalcitrant organisms, bioprospecting of whole environmental DNA (eDNA) can be

653 performed using targeted PCR followed by high through-put sequencing (Charlop-Powers et al 31

654 2014, Charlop-Powers et al 2015, Lemetre et al 2017, Woodhouse et al 2013). Guided by these

655 findings, NRPS and PKS rich environments can be highlighted and targeted for further

656 analysis and isolation efforts.

657 EXTREMOTOLERANT BIOCRUST ORGANISMS AS POTENTIAL SOURCES OF 658 NOVEL SPECIALISED MICROBIAL METABOLITES

659 The exploration of extreme arid environments for novel bioactive compounds has focused on

660 attaining extremophile isolates, primarily Actinobacteria and some Cyanobacteria, revealing a

661 range of new chemical entities (e.g. Figure 1.4c) as well as NRPS and PKS genetic diversity

662 (Crits-Christoph et al 2016, Ding et al 2013, Gomez-Escribano et al 2015, Mohammadipanah

663 and Wink 2015, Okoro et al 2009). Recently, Charlop-Powers et al (2014) surveyed 96 distinct

664 soil microbiomes and found arid soils harboured the greatest NRPS and PKS richness,

665 directing further work to pursuing these microbiomes (Lemetre et al 2017)

666 Within the scope of arid soils, biocrusts present as expedient candidates for targeted

667 bioprospecting. Biocrusts are dominated by cyanobacteria, renowned prolific producers of

668 bioactive compounds. Scytonemin is a well-established example of a secondary metabolite of

669 biocrust-associated cyanobacteria (Figure 1.4d), occurring across several genera including

670 Scytonema, Nostoc and Calothrix (Couradeau et al 2016). These genera also comprise strains

671 which are known to produce saxitoxin (Smith et al 2012), the heptatoxin nodularin (Gehringer

672 et al 2012) and the cytotoxin calothrixin (Xu et al 2016), respectively. Recently, exometabolite

673 profiling of Microcoleus vaginatus has revealed this biocrust species produces thousands of

674 compounds (Baran et al 2015) that directly effect and shape the biocrust community (Swenson

675 et al 2018). From a practical aspect, the prevalence and accessibility of biocrusts lends them to

676 convenient examination. Furthermore, the relative microbial simplicity of biocrusts facilitates

677 their use as model systems to aid understanding of metabolite ecology (Baran et al 2015,

678 Bowker et al 2014). Together, these findings suggest biocrusts are a promising frontier for the

679 discovery of novel specialised microbial metabolites and warrant investigation.

680 SCOPES AND OBJECTIVES OF THIS THESIS 681 The rise of genomics is facilitating our exploration of the microbial world. An unprecedented

682 diversity is being revealed alongside critical functions underpinning all biological systems on

683 Earth. While microorganisms have long served as a ready source of beneficial and profitable

684 compounds, biosynthetic pathways, and environmental processes, it is evident much of their

685 promise remains untapped. Yet, as rapid progress is made to meet this challenge with DNA

686 sequencing technologies, there is a risk of drowning in the data generated. Furthermore, the

687 immense scope and diversity of habitats within nature offer an impractical number of possible

688 environments to survey. There is an imperative to identify promising niches to direct

689 bioprospecting efforts. Beyond simple gene surveys, understanding the natural histories of

690 microbiomes and their ecology can help direct the analysis of NRPS and PKS datasets to

691 maximise returns. This thesis seeks to explore patterns in taxonomic and functional gene sets

692 of Australian soil biocrust microbiomes. To this end, high through-put sequencing was used to

693 taxonomically profile biocrusts from across Australia before targeting the NRPS and PKS

694 genes present in these environments.

695 Chapter 2 will examine how biocrust communities change over successional stages in order to

696 determine whether visually distinct biocrusts from the same local environment should be

697 treated as separate sample types for specialised metabolite discovery. In Chapter 3, the intra-

698 continental patterns of biocrust distributions will be explored to determine large scale factors

699 driving community assembly. Chapter 4 conducts NRPS and PKS gene surveys of the

700 biocrusts from Chapters 2 and 3 to show how these genes occur across Australia and to

701 identify key target sites and organisms for future pursuit. These datasets will contribute to a

702 complex biosynthesis atlas of Australia.

703

CHAPTER 2: BIOCRUST MORPHOLOGY IS LINKED TO MARKED DIFFERENCES IN MICROBIAL COMMUNITY COMPOSITION

This chapter has been published as:

Chilton, AM, Neilan, BA and Eldridge, DJ (2017) Biocrust morphology is linked to marked differences in microbial community composition. Plant and Soil. https://doi.org/10.1007/s11104-017-3442-3

1 INTRODUCTION 2 Biocrusts are complex associations of macroscopic, non-vascular organisms such as mosses,

3 lichens and liverworts, and microscopic organisms such as cyanobacteria, fungi, bacteria and

4 archaea, that form intimate associations with surface soils. Biocrusts are dominated by

5 phototrophic organisms that require direct access to sunlight. Consequently, they are

6 particularly common in the interspaces between patches of perennial plants in areas such as

7 drylands where vascular plant cover is sparse. Biocrusts are critically important for regulating

8 carbon, nutrient and hydrological cycles, stabilising surface soils, and providing habitat for soil

9 biota (Bowker et al 2013a, Neher et al 2009, Zhao et al 2016). In many landscapes, biocrusts

10 are significant contributors to biomass, biodiversity and ecological functioning. Efficient

11 functioning of biocrust communities relies heavily on the underlying microscopic components

12 of the crust. Critical to the formation and stability of biocrusts is the stabilisation of soil

13 particles resulting from the ecosystem engineering actions of bacteria, particularly large

14 filamentous cyanobacteria (Garcia-Pichel and Wojciechowski 2009). This stabilisation is

15 typically initiated by rain and is responsible for nutrient enrichment of the local soil profile

16 encouraging subsequent colonisation by additional microorganisms, particularly more

17 macroscopic mosses and liverworts. The important bioengineering role of cyanobacteria has

18 been shown to be a precursor to the development of stable functional soil surfaces (Garcia-

19 Pichel and Wojciechowski 2009, Rossi and De Philippis 2015, Zhang 2005).

20 Despite the large body of research carried out on the macroscopic, more visible components

21 of biocrusts (e.g. Belnap et al. 2016), comparatively little is known about the underlying

22 microbial community structure. Profiling of the bacterial community of biocrusts using

23 ribosomal gene sequencing is in its relative infancy, but has been performed for a number of

24 locations worldwide. The majority of biocrust microbiome research has been carried out on 36

25 soils from the western deserts of the United States, and more recently China (Li et al 2014),

26 Europe (Büdel et al 2014) and Africa (Thomas and Dougill 2007). In Australia, however, the

27 microbial signature of biocrusts has been poorly studied (Abed et al 2012). Together these

28 studies have revealed that biocrusts from drylands are dominated by cyanobacteria and are less

29 bacterially diverse than those from more mesic areas (Garcia-Pichel et al 2003, Zaady et al

30 2010). The most widely reported cyanobacterium in the global literature is Microcoleus,

31 particularly Microcoleus vaginatus (Starkenburg et al 2011). More recent studies have shown that,

32 although biocrusts are dominated by filamentous cyanobacteria, additional non-cyanobacterial

33 phyla are also common (Nunes da Rocha et al 2015). These include the ubiquitous

34 Proteobacteria (primarily Alphaproteobacteria), Actinobacteria, Acidobacteria and

35 Bacteroidetes. Less common phyla include Deinococcus-Thermus, Chloroflexi, Firmicutes,

36 Verrucomicrobia, Planctomycetes and Gemmatimonadetes. Together these phyla represent a

37 broad group of microbes that function as chemoorganoheterotrophs or ammonia-oxidizing

38 chemoorganoautotrophs (Delgado-Baquerizo et al 2016).

39 Biocrust communities can be characterised in many ways including: level of development

40 (Belnap et al 2008b), biocrust morphology type (Pócs 2009, Thomas and Dougill 2007), level

41 of pigmentation (Couradeau et al 2016) or constituent organisms (e.g. cyanobacterial vs

42 chlorolichen;(Budel et al 2009)). These characteristics are inherently descriptive and subjective,

43 but are often associated with a particular stage of biocrust maturity. Generally, biocrusts shift

44 from thin, lightly-coloured cyanobacterial-dominated crusts to thicker, darker, more complex

45 assemblages. Little is known, however, about whether the outward appearance or morphology

46 of biocrusts, often based on crust type, pigmentation and level of development, is a useful

47 proxy of the underlying microbial community structure. Linking the microbial community

48 structure to the outward appearance of biocrusts is critically important if we are to use

49 biocrusts as indicators of ecosystem health and functioning (Castillo-Monroy et al 2011) or as

50 model systems to examine effects of different stresses such as overgrazing or climate change

51 (Bowker et al 2014, Garcia-Pichel et al 2013).

52 Previous research has revealed that the composition of microbial communities within biocrusts

53 is complex and dynamic, and changes under different environmental settings. For example, in

54 the Kalahari Desert, bacterial community structure varied markedly with vegetation type, and

55 was distinct from the subsoil microbiome (Thomas and Dougill 2007). In the Negev Desert in

56 Israel, precipitation was found to be the strongest driver of cyanobacterial diversity and

57 abundance (Hagemann et al 2015). In the deserts of the western United States, studies of

58 biocrusts along a developmental gradient from thin cryptic species to dark mature biocrusts

59 showed that changes in biocrust composition can have a direct effect on the soil microbial

60 community. Warming was associated with the replacement of the keystone heat-sensitive

61 Microcoleus vaginatus by the more heat-tolerant Microcoleus steenstrupii (Couradeau et al 2016).

62 Although both morphology and bacterial community composition have been used to better

63 understand the functional role of biocrusts, relatively little is known about how these relate to

64 each other.

65 Here we examine the pattern of community composition of the microbial community

66 associated with biocrusts ranging in development from bare surfaces to highly developed,

67 floristically rich and deeply pigmented biocrusts containing lichens and mosses. We

68 hypothesised that differences in biocrust development would be reflected in substantial

69 differences in microbial community structure. We tested this hypothesis using indicator species

70 analysis and co-occurrence networks. The aim here was to deepen our understanding of the

71 links between biocrust form and function and improve our understanding of the use of

72 biocrusts as model systems and ecosystem indicators.

73 METHODS

74 STUDY AREA AND FIELD SAMPLING

75 The Kalgooleguy Regeneration Reserve is an area of crown land north-west of Cobar, New

76 South Wales Australia (−31.49° S, 145.84° E) covering an area of 4777 ha. The climate is

77 characterised by low and variable rainfall (mean annual rainfall 390 mm), high rates of

78 evaporation (~ 2200 mm yr.−1), hot dry summers (maximum 28– 39°C, minimum 14–24°C)

79 and cool to mild winters (maximum 13–20°C, minimum 2–8°C). The reserve is located on the

80 Cobar Peneplain, a low undulating plain punctuated by stony ridges and ranges and

81 characterised by well-drained red and red-brown clay loams and loams, with increasing clay

82 content with depth (Typic Haplargids or Red Earths), with variable amount of stones in the

83 profile. The reserve is dominated by eucalypt woodlands and Acacia shrublands dominated by

84 Eucalyptus populnea, Callitris glaucophylla, Acacia aneura and dense patches of shrubs (Dodonaea

85 viscosa, Eremophila longifolia, Senna artemisioides, Acacia spp.) varying in cover from <5% to about

86 50%. The soil surface is dominated by a variable cover of biocrusts ranging from

87 cyanobacterial films to well-developed lichen crusts. Within the reserve we identified three

88 stages of crusts based on thickness, pigmentation and composition (sensu Belnap et al. 2008).

89 Early stage crusts (hereafter ‘Early-stage’) were defined as thin, lightly-coloured smooth crusts

90 dominated by cyanobacteria and with little evidence of colonisation by mosses or lichens. Mid-

91 stage crusts (hereafter ‘Mid-stage’) were thicker and more pigmented (darker) and showed

92 evidence of colonisation by mosses and lichens. Late stage crusts (hereafter ‘Late-stage’) had

93 the greatest pigmentation and thickness and were dominated by lichens and mosses (Figure

94 2.1). These three crust types were compared with uncrusted, bare surfaces (hereafter ‘Bare-

95 stage’). In April 2013 we collected samples of the four different stages from three large sites

96 within the reserve. Sites were separated by distances of about 1.5 km. At each site we collected

97 four samples of each stage within large plots of about 50 by 50 m resulting in 16 samples per

98 site. The distance between samples within a plot ensured that all samples were spatially

99 independent at the scale of the organism studied. For each sample, 10 cm2 plots were collected

100 to the depth of the biocrust and stored in paper bags and transported to the laboratory and at

101 the University of New South Wales and stored at 4°C.

102

103 Figure 2.1 Left - Map of collection site, Top - Example photographs of a) Bare, b) Early, c) 104 Mid and d) Late stages. Each square is about 10 cm across, Bottom - e) Photograph of changes 105 in biocrust stages from Early (far left) to late (far right) stages. Scale is 12 cm across.

106 MOLECULAR ANALYSIS

107 Environmental genomic DNA was isolated from 500 mg of homogenised soil using the

108 FASTDNA Spin Kit for Soil (MP Bio Laboratories, USA) according to the manufacturer’s

109 instructions. The hypervariable regions V1-V3 of the 16S rRNA gene were amplified using

110 unique combinations of barcoded 27F/519R primers. The pooled DNA libraries were

111 submitted to the Ramaciotti Center for Genomics (UNSW, Australia). Sequencing was

112 performed on an Illumina MiSeq using a MiSeq Reagent Kit v3 with a 2x300 bp run format.

113 Sequencing data were received de-multiplexed via the Illumina cloud-computer BaseSpace and

114 are available on the NCBI Small Read Archive (SRA) under project PRJNA396825. For this

115 study, only amplicons generated from the forward primers (27F) were used in order to avoid

116 artificial inflation of diversity measures due to poor confidence in contig formation (Kozich et

117 al 2013, Nielsen et al 2016). Sequence reads were processed and analysed using Mothur version

118 1.34.0 (Schloss et al 2009) according to the standard operating procedure developed by Kozich

119 et al. 2013. Briefly, sequences were checked for quality using a threshold average quality score

120 of 30 over 50-base increments. Sequences less than 200 bases, with greater than 8

121 homopolymers or containing ambiguous bases were re-moved. Amplicons were aligned and

122 trimmed to a consensus region using a customised V1–3 version of the SILVA alignment

123 database (Quast et al 2013). Pre-clustering was performed where by rare sequences with ≤1

124 per 100 bp difference to abundant sequences were merged. Chimeras were detected and

125 removed using the in-built application UCHIME (Edgar et al 2011). Sequences were classified

126 using the GreenGenes database (version gg_13_8_99, August 2013) (DeSantis et al 2006) with

127 an 80% pseudobootstrap confidence score. Sequences not classified at kingdom level or

128 classified as Mitochondria, Archaea or Eukaryota were removed. Samples were rarefied to

129 14,434 sequences resulting in a curated dataset of 678,398 sequences across 47 samples with an

130 average length of 259 bases. Operational taxonomic units (OTUs) were generated at a 0.03

131 distance threshold via a distance-based matrix with average neighbour clustering performed at

132 Order level (Schloss et al 2011). OTUs were then assigned taxonomy using the same

133 GreenGenes database.

134 STATISTICAL AND NETWORK ANALYSIS

135 We calculated measures of richness, diversity and evenness using the Diverse function in the

136 Primer/PERMANOVA package (Anderson et al 2008). Differences in richness, diversity and

137 evenness were deter-mined using mixed-models ANOVA. Our model structure accounted for

138 differences among the three sites, the four stages and their interactions. Tukey’s Least

139 Significant Difference (LSD) tests were used to determine differences among the four stages.

140 We use the same model structure to examine differences in the composition of the 16S rRNA

141 OTUs in relation to stage. With PERMANOVA, pair-wise, a posteriori comparisons were

142 made, where necessary, using a multivariate analogue of the t statistic, the probability levels

143 being obtained by permutation. Homogeneity of spread for factor Stage was confirmed using

144 PERMDISP with 999 permutations (pseudo F = 2.12. P(perm) = 0.414). We then used non-

145 metric multidimensional scaling ordination (nMDS) to derive the first two dimensions of the

146 nMDS biplot on log-transformed OTU abundance data to represent the compositional

147 differences among the four stages. The 2D solution provided a suitable representation of the

148 bacterial data (stress = 0.16). Indicator Species Analysis (De Cáceres and Legendre 2009) was

149 used to determine the degree of association between OTUs and stage type. Operational

150 Taxonomic Units were randomized among the stages and a Monte Carlo procedure performed

151 with 999 iterations to determine the statistical significance of the indicator values generated.

152 Co-occurrence analysis of OTUs for network analysis was measured using the SparCC

153 command within mothur with 100 iterations and 10,000 permutations (Friedman and Alm

154 2012). Only OTUs contributing greater than 0.25% to a stage and occurring across 3 or more

155 sites were included to avoid spurious associations. False discovery rates were kept below 5%

156 using Benjamini-Hochberg corrected P-values (q < 0.0016) (Benjamini and Hochberg 1995).

157 Significant OTU-OTU correlations for each crust type were visualised as scale-free networks

158 using the Cytoscape package version 3.2.1 (freely available at: http://www. cytoscape.org/).

159 Non-random co-occurrence patterns for the network OTUs were checked with the checker-

160 board score (C-score) with the R package EcoSimR under a null model preserving row and

161 column sums with default settings (Gotelli and Entsminger 2015). For each network, overall

162 topological parameters of connectivity, centrality and density were calculated (Assenov et al

163 2008).

164 RESULTS

165 RICHNESS AND DIVERSITY OF MICROBIAL TAXA

166 Across all samples we recorded a total of 44,005 OTUs at a 0.03 distance threshold. Less than

167 3% of all OTUs accounted for 75% of total abundance and 60% of OTUs occurred as

168 singletons. For the total dataset, i.e. considering all OTUs, there were no significant differences

169 in richness (F3,6 = 4.15, P = 0.065), diversity (P = 0.064) or evenness (P = 0.59) among the

170 four stages. However, when we excluded singletons from the analyses, both richness (F3,6 =

171 5.60, P = 0.036) and diversity (F3,6 = 5.63, P = 0.035), but not evenness (P = 0.63) increased

172 significantly with increasing biocrust stage (Table 2.1).

173 Table 2.1 Mean (± SE) richness, diversity (Margalef’s index) and evenness (Pileau’s index) of 174 microbial OTUs across the four stages for samples excluding singletons. Bare Early Mid Late Attribute Mean SE Mean SE Mean SE Mean SE Richness 2134a 90.2 2408ac 82.8 2581bc 98.0 2720b 51.0 Diversity 233a 9.5 252ac 8.7 271bc 10.4 286b 5.4 Evenness 0.80a 0.01 0.79a 0.01 0.81a 0.01 0.82a 0.01 175 Different letters within an attribute indicate a significant difference among the four stages at P 176 < 0.05

177 COMMUNITY COMPOSITION

178 We found a significant difference in the composition of OTUs among the four stages (Pseudo-

179 F3,6 = 3.37, P(perm) = 0.002). Multiple comparison tests revealed that the composition of Bare

180 was significantly different to the other stages (t > 1.87, P < 0.016), and that Early was different

181 from Late (t = 1.72, P = 0.017; Figure 2.2). Phylotyping of all OTUs identified 16 abundant

182 bacterial classes across seven phyla and all biocrust stages (Figure 2.3). Alphaproteobacteria

183 was the dominant phylum within the Bare and Late stages, but Cyanobacteria was the

184 dominant phylum within Early and Mid stages.

185

186 Figure 2.2 Non-metric multidimensional biplot of the 48 sites from the four stages based on 187 composition of OTU with an abundance >1%.

188 MICROBIAL INDICATORS OF BIOCRUST STAGE

189 Indicator species analysis revealed 18 OTUs comprising 13 genera that characterised the Bare

190 stage. Herbiconiux (Actinobacteria) was the strongest indicator and Flavisolibacter (Bacteroidetes)

191 the most abundant (Table 2.2). The Late stage was the only biocrusted stage with a unique

192 OTU indicator species, the terrestrial green-algae Chloroidium. The three biocrusted stages were

193 characterized by eight OTUs from six genera, all of the phylum Cyanobacteria. An unclassified

194 Oscillatoriophycideae OTU was the strongest indicator of biocrusted soils and Phormidium the

195 most abundant genus. There were no indicators of Early or Mid stages. All biocrust indicator

196 OTUs were most similar to other biocrust submissions whereas the Bare stage indicator OTUs

197 had greatest similarity to non-biocrust environments.

198

199 Figure 2.3 Relative abundance (%) of the major bacterial classes grouped by phylum for Bare, 200 Early, Mid and Late stage biocrusts.

201

202 Figure 2.4 Co-occurrence networks for Bare, Early, Mid, and Late stage biocrusts. Nodes 203 represent OTUs contributing greater than 0.25% to a stage and occurring across 3 or more 204 sites. Edges represent significant (p < 0.05 after Benjamini-Hochberg procedure) positive and 205 negative correlations. (See supplementary figures for full-sized images)

206 Table 2.2 Bacterial taxa significantly associated (P < 0.01) with different biocrust stages and biocrust-stage combinations using Indicator 207 Species Analysis.

BLAST Stage and phylum Class Genus IV RA (%) Accession Similarity (%) Bare Actinobacteria Actinobacteria Herbiconiux 0.98 17 KT773540 96 Proteobacteria Betaproteobacteria Pseudoburkholderia 0.98 28 KX508248 100 Proteobacteria Betaproteobacteria Limnobacter 0.98 46 JF809120 97 Bacteroidetes [Saprospirae] Segetibacter 0.97 17 AB696124 99 Actinobacteria Actinobacteria Friedmanniella 0.90 12 AF409005 98 Proteobacteria Alphaproteobacteria Acidisphaera 0.89 12 <95 Proteobacteria Betaproteobacteria Pseudogulbenkiania 0.88 11 <95 Firmicutes Bacilli Bacillus 0.88 14 KJ600919 99 [Thermi] Deinococci Deinococcus 0.87 15 <95 Bacteroidetes [Saprospirae] Flavisolibacter 0.85 71 JX797411 96 Chloroflexi Ktedonobacteria unclassified 0.83 49 <95 Actinobacteria Actinobacteria Nocardioides 0.82 1 KT772263 99 Cyanobacteria Chloroplast unclassified 0.80 12 HM725590 100 Late Cyanobacteria Chloroplast Chloroidium 0.80 30 HM731584 99 Crusted Cyanobacteria Oscillatoriophycideae unclassified 0.98 21 <95 Cyanobacteria Nostocophycideae Toxopsis 0.97 18 GU362214* 100 Cyanobacteria Nostocophycideae Cylindrospermum 0.97 11 <95 Cyanobacteria Nostocophycideae Aphanizomenon 0.94 13 JQ383870# 100 Cyanobacteria Nostocophycideae Tolypothrix 0.91 36 GU362210* 100 Cyanobacteria Oscillatoriophycideae Phormidium 0.87 50 JQ383804# 98 208 IV Indicator Value; RA relative abundance *sequence from Oman biocrust # sequence from Nevada, USA, biocrust.

209 NETWORK ANALYSIS

210 Co-occurrence networks derived from abundant OTUs showed non-random assembly

211 patterns (C-scores, Table 2.3), with highly-connected hubs (a group of nodes exceedingly more

212 highly connected than the average) indicative of small world networks (Figure 2.4). The

213 number of nodes (OTUs) remained relatively consistent across each stage (Table 2.3). Network

214 densities and clustering coefficients (measures of node connectivity) were also stable, with the

215 exception of the Early stage which exhibited a large drop in connectivity that resulted in an

216 increased network diameter. Degree (number of connections per node) followed a power-law

217 distribution with a few highly connected nodes forming edge-dense hubs resulting in modular

218 network topologies. High clustering coefficient and density scores indicated the Bare stage

219 network had the greatest modularity (Eldridge et al 2015). Proteobacteria was the greatest

220 contributor to nodes and correlations across all stages with the exception of the Early stage,

221 where cyanobacteria comprised the greatest number of nodes. Phylum level patterns of node

222 correlations were observed. Cyanobacterial nodes shifted from within-phylum to cross-phyla

223 correlations from the Bare to Early stages whereas all non-cyanobacterial phyla consistently

224 formed more cross-phyla than within-phylum associations (Supplementary Table A1.1). The

225 number of negative interactions increased as biocrust stage advanced.

226

227

228

229

230

231 Table 2.3 Topology metrics and C-score measures derived from scale-free co-occurrence 232 networks of abundant OTUs (greater than 0.25% for each stage) for Bare, Early, Mid and Late 233 stage biocrust microbial communities.

Network metric Bare Early Mid Late

Number of Nodes (OTUs) 286 252 275 242

Number of Edges 1311 648 1064 947

Clustering coefficient 0.29 0.179 0.247 0.224

Density 0.032 0.02 0.028 0.032

Network centralisation 0.084 0.068 0.74 0.101

Network heterogeneity 0.712 0.794 0.698 0.839

Network Diameter 8 11 9 8 Average number of neighbours 9.2 5.1 7.7 7.8

Connected components 4 2 3 5

Negative interactions (%) 23 36 40 41

C-Score measures

Observed Index 1.07 0.86 0.97 0.65

Mean of Simulated Index 1.05 0.83 0.94 0.06

Standard Effect Size (SES) 1.91 9.96 9.70 20.76 234 Edges represent significant (p < 0.05 after Benjamini-Hochberg procedure) positive and 235 negative correlations

236

237 DISCUSSION 238 Biocrusts comprise a wide range of physical types or stages, ranging from thin cyanobacterial

239 layers to thick, highly developed crusts dominated by a rich community of lichens, mosses and

240 liverworts. The extent to which these different biocrust forms reflect differences in their

241 underlying microbial signatures is, however, poorly known. In this study we show that

242 increases in biocrust morphology and complexity corresponded with in-creased richness and

243 diversity of biocrust-inhabiting microbes. Bare surfaces had a different complement of

244 bacterial taxa to biocrusted surfaces, irrespective of their complexity. Network arrangement

245 also differed among the four stages, with greater heterogeneity and more negative interactions

246 with increasing biocrust development. Our results indicate that recognisable features of

247 biocrust surfaces such as differences in thickness, cover and development are associated with

248 marked differences in microbial communities. Thus different biocrust surface types are likely

249 to reflect differences in biocrust capacity to moderate critical soil and ecological processes.

250 BIOCRUST STAGES AS A PROXY FOR MICROBIAL COMMUNITY STRUCTURE

251 Gross morphological attributes such as colour, shape and thickness have been used widely to

252 differentiate biocrusts into discrete community types, by develop-mental stage (Belnap et al

253 2008b) or morphological group (Eldridge and Rosentreter 1999). Some of this differentiation

254 is based on the notion that form reflects function and therefore that different forms should be

255 indicative of different species assemblages with unique functions (Kidron et al 2015). In our

256 study we found that biocrust richness increased with increases in developmental stage, from

257 Bare to Late stages, but there were no differences in composition among the three biocrusted

258 (Early, Mid, Late) stages, which were largely dominated by cyanobacteria. We would expect to

259 detect substantial differences between bare and biocrusted stages for a number of reasons. The

260 Bare stage showed little evidence of surface differentiation, a generally hardened surface with 50

261 few cracks, and little incorporation of litter. Compared with Bare, the biocrusted surfaces had

262 relatively high surface microtopography, up to 10 mm, with variable cracks and greater

263 evidence of biological activity e.g. spider holes and small ant holes. A wide range of surface

264 characteristics would likely provide more refugia for microbial communities in dryland

265 ecosystems than bare soils, which are essentially homogeneous and hostile. Our work across

266 regional eastern Australia has shown that biocrusts at a later stage of development support a

267 richer community of mosses, lichens and liverworts, and richer vascular plant associates

268 (Eldridge et al 1997). This would likely provide opportunities for a larger range of bacterial

269 taxa. A richer plant community should support a greater range of plant root types, a wider

270 spectrum of root exudates (Berg and Smalla 2009) and therefore a greater range of

271 microhabitats for bacteria (Lamb et al 2011). Evidence from Australian drylands indicates that

272 microsite differentiation can modify the abundance of soil ammonia oxidizing bacteria, with

273 reductions on bare soils, but increases in areas of biological activity around structures such as

274 ant nests that are often found in well-developed biocrusts (Delgado-Baquerizo et al 2016).

275 Further, as increasing development is associated with a thicker surface biocrust, we would

276 expect greater levels of soil carbon and nitrogen, given that carbon and nitrogen are

277 concentrated in the uppermost biocrust layers (Mueller et al 2015, Steven et al 2013a).

278 CYANOBACTERIAL “BLOOMS” PROMOTE BIOCRUST FORMATION

279 The primary microbial drivers of biocrust formation are filamentous cyanobacteria. These

280 stabilise loose soil particles by producing exometabolites, modifying soil physical properties

281 and enriching the metabolic potential of the soil. We found a persistent population of diverse

282 filamentous cyanobacteria across all stages, including the Bare stage, that changed in relative

283 abundance as the crust developed. We propose that biocrust-forming cyanobacteria are

284 prevalent within arid top-soils and that, given favourable conditions, undergo a bloom in 51

285 population akin to those within aquatic systems (Fuhrman 2009). Increases in the relative

286 abundance of cyanobacteria between Bare and Early stages combined with the high production

287 of exopolymeric substances by cyanobacteria likely lead to increases in microbial biomass

288 necessary to form a cohesive biocrust layer. These early biocrusts likely originate from raindrop

289 impacted physical crusts, which are often precursors of biocrusts in highly erodible Australian

290 loams (Eldridge 2001). Interestingly, the greatest increase in cyanobacterial abundance was due

291 to Nostocophycideae types. Nostocophycideae are typically documented as later additions to

292 biocrust com-munities, and are attributed with major roles in nitrogen fixation and darkening

293 of the biocrust (Belnap et al 2008b, Yeager et al 2004). Here, they were detected within bare

294 soil and are an abundant component as soon as the soil is stabilised in thin, light-coloured

295 crusts. This is an important consideration for future work determining the nutrient-cycling and

296 ecological roles of these biocrusts.

297 A clear difference between Bare and crusted stages detected with Indicator Species Analysis

298 shows that the Bare stage is characterised by non-cyanobacterial OTUs (excluding

299 chloroplasts) whereas the crusted stages are characterised by cyanobacterial OTUs. The lack of

300 cyanobacterial indicators within the Bare stage, despite abundant and diverse representation,

301 indicates that cyanobacterial OTUs inhabiting the Bare stage are not an independent

302 population but likely originate from the surrounding biocrusts (Shade et al 2012). These may

303 represent residual populations following physical disturbance (Kuske et al 2012) or inundation

304 by sediment (Williams and Eldridge 2011). However, we found no evidence of a remnant

305 biocrust matrix on the Bare stages during our sampling. All cyanobacterial indicator OTUs

306 were most similar to other biocrust submissions whereas the Bare stage indicator OTUs had

307 greatest similarity to non-biocrust environments. This supports the finding that biocrusts are a

308 niche populated by specialised organisms able to form and sustain biocrusts (Elliott et al 2014).

309 Despite the high abundance and diversity of cyanobacteria in our samples, and the putative

310 world-wide distribution of the genus, no OTUs were assigned to Microcoleus. Rather, Phormidium

311 was the most abundant cyanobacterial genus and was found consistently throughout all the

312 stages, particularly the Early stage. Microcoleus vaginatus is often identified as the primary

313 cyanobacterium of biocrusts, particularly in the early stages and more often in North American

314 samples (Garcia-Pichel et al 2003). Microcoleus and Phormidium are poorly resolved

315 phylogenetically with-in the Phormidiaceae family, however, beyond taxonomic discrepancies,

316 these types share important biocrust-forming attributes such as the formation of long,

317 sheathed filaments with large cells, features which likely support the ability to form supra-

318 cellular ropes to stabilise soil grains. Cyanobacteria with these features are thought to be able

319 to travel large distances in bare soils.

320 BIOCRUST STAGES DEFINED BY MICROBE-MICROBE ASSOCIATIONS

321 Our network analyses indicated that there were major differences in connectivity among the

322 four biocrust stages, indicating differences in their capacity to 1) recover from disturbance, 2)

323 deviate from equilibrium, and 3) perform multiple functions (Allison and Martiny 2008, Bissett

324 et al 2013). The Bare stage network had the greatest modularity (formation of hubs),

325 suggesting high reactivity and low resilience (Ruiz-Moreno et al 2006). This community

326 structure may be an important trait for bacteria within oligotrophic arid and semi-arid soils for

327 the prompt uptake of nutrients and response to infrequent wetting events. High reactivity and

328 low resilience may also explain how cyanobacteria can colonise and dominate bare surfaces and

329 initiate biocrust formation. An essential factor in the formation and growth of biocrusts is the

330 presence of biocrust-forming bacteria, primarily cyanobacteria. We observed a cyanobacterial

331 hub within the Bare stage network that likely indicates a niche where members respond to

332 environmental stimuli in the same way (Fuhrman 2009). Phormidium was the main genus in this

333 hub, but several other cyanobacterial genera such as Brasilonema, Leptolyngbya and

334 Cylindrospermum were also present. This may indicate a degree of functional redundancy within

335 bare soil, and suggests that these genera are also implicated in biocrust formation.

336 A sharp decline in the number of edges from Bare to Early stages resulted in strong de-

337 centralisation of the Early stage network. We theorise that the Early stage microbial

338 community has yet to effectively adapt to the modified conditions induced by cyanobacteria

339 colonisation and biocrust formation. A shift from within-phylum to among-phyla correlations,

340 which was unique for Cyanobacteria, may be an ecological strategy that promotes biocrust

341 formation. By Mid and Late stages, node connectivity appeared restored, but many of these

342 were negative correlations. We suggest that this is a reflection of resource partitioning

343 (Fuhrman and Steele 2008), likely due to the substrate preferences of heterotrophic bacteria

344 (Baran et al 2015) and indicates biocrust maturity. Overall, a sequence of high to low network

345 connectivity was followed by a trend towards recovery of network complexity, a pattern

346 observed in salt marsh chronosequences, where loss of network complexity could be due to

347 loss of taxonomic diversity (Dini-Andreote et al 2014).

348

CHAPTER 3:

BIOCRUSTS ASSEMBLY IS DRIVEN BY SEASONALITY OF PRECIPITATION ON AN INTRA-CONTINENTAL SCALE

1 INTRODUCTION 2 Biocrusts occur globally across a range of arid lands and are, consequently, spatially variable.

3 On a local scale, biocrusts exhibit a patchiness similar to that observed for vegetation cover

4 (see Chapter 2)(Bowker et al 2013b). The primary factors acting at the micro- to local scales are

5 biotic, topographic, and edaphic. At larger scales, climate and biogeographical forces shape

6 communities (Bowker et al 2016). Much of the basis for these latter findings have been derived

7 from studies on the macro component of biocrusts (Bowker et al 2014, Concostrina-Zubiri et

8 al 2014, Martínez et al 2006). Within Australia, observable effects of climate on the distribution

9 of dryland lichens manifests through seasonality of precipitation where lichens were restricted

10 to regions of winter rainfall (Eldridge et al 1997, Rogers 1972). Indications from bacterial

11 profiling of biocrusts from deserts of Western America have shown cyanobacteria are affected

12 by temperature (Garcia-Pichel et al 2013) and precipitation frequency (Ferrenberg et al 2015,

13 Steven et al 2015). Similarly, morphological analyses along a transect in Southern Africa

14 showed winter rainfall areas were richer in cyanobacteria than areas with predominantly

15 summer rainfall (Budel et al 2009). Given the large scales at which these factors operate, the

16 direct influence they have on biocrust coverage and composition, and the predicted impacts of

17 climate change (Rodriguez-Caballero et al 2018), it is important to better understand how such

18 forces drive biocrust assembly.

19 Yet, despite their critical roles in biocrust formation and maintenance, intra-continental scale

20 studies have yet to be conducted for Australian biocrust microorganisms. Current

21 characterisation of Australian biocrust bacteria encompasses microscopic identification of

22 cyanobacteria (Aboal et al 2016, Strong et al 2013, Williams et al 2008, Williams and Büdel

23 2012, Williams et al 2014) and genomic profiling of two sites (Abed et al 2012, Liu et al 2017b).

24 Recently, Australia was ranked relatively low in terms of cyanobacterial diversity compared to 56

25 other continents (Büdel et al 2016), likely due to a paucity of research. Here, biocrusts from

26 across Australia have been examined at the intra-continental scale to examine factors affecting

27 bacterial assemblages. Determining landscape scale distribution patterns can inform on the

28 functional nature of biocrust communities and help delimit ecological regions. For example,

29 biocrusts enriched with heterocystic cyanobacterial species, such as Nostoc, tend to fix nitrogen

30 at higher rates than those dominated by Microcoleus spp. (Barger et al 2016). Furthermore,

31 identifying compositionally distinct biocrusts can highlight novel communities for future

32 research to elucidate ecological impacts (Garcia-Pichel et al 2013). Pertinent to this thesis,

33 there was also an imperative to characterise biocrusts across Australia to facilitate an informed

34 approach to bioprospecting, including the discovery of spatially and functionally diverse

35 microbiomes. Here, biocrusts from across Australia were sampled to test the assumption

36 bacterial communities will vary over different special scales.

37 METHODS

38 SAMPLE COLLECTION AND PROCESSING

39 Biocrust samples from a range of geographically distinct semi-arid biomes across Australia

40 (Figure 3.1) were collected (Table 3.1), transported dry and stored at -80°C until processing.

41 Samples from northern Australia were attained from previous research projects. Samples were

42 confirmed visually as biocrusts according to aggregation of soil into cohesive crust structures

43 and the presence of microbial filaments. Biocrust was homogenized and

44 500 mg taken for genomic DNA extraction using the FASTDNA Spin Kit for Soil (MP Bio

45 Laboratories, USA) according to the manufacturer’s instructions. Molecular libraries of the

46 hypervariable V1-V3 region were generated using the same method as per the Cobar samples

47 (Chapter 2). For the Cobar data set, replicates of the biocrust stages for each site were merged

48 resulting in three averaged replicates per biocrust stage. The raw sequencing data from all

49 samples were analysed together in the mothur pipeline as outlined in Chapter 2. Samples were

50 rarefied to 10180 sequences across 45 samples with an average length of 268 bp.

52 Figure 3. 1 Sample site locations across Australia with seasonal precipitation gradient overlay 53 (Williams et al 2012).

54 STATISTICAL ANALYSIS

55 For analysis, singleton and doubleton OTUs were removed and where required, the data

56 standardized by total. In-built calculators within PRIMER-6 (version 6.1.11, Primer-E Ltd,

57 UK) were used to determine richness (Margalef’s Index), evenness (Pielou’s Index) and

58 diversity (Shannon Index) (Anderson et al 2008). Differences between sites in richness,

59 evenness and diversity were tested for using one-way ANOVA followed by examination with 58

60 Tukey multiple comparisons for significant differences. To determine the broad taxonomic

61 sources of diversity for each site, cyanobacterial and non-cyanobacterial OTUs were examined

62 separately. Databases were rarefied and standardized and site means were compared using

63 t-tests assuming non-Gaussian distributions.

64 For beta-diversity analyses, a Bray-Curtis dissimilarity matrix was derived from square-root

65 transformed OTU abundance data. Unconstrained ordination of samples via non-metric

66 multidimensional scaling (nMDS) was used to determine post-hoc lines of questioning and

67 whether sample processing effected community patterns. The factors; Site, Precipitation

68 Season, Collection Year, Collection Season, site Climate Class, and Kit Extraction were tested

69 for significance using Permutational multivariate analyses of variance (PERMANOVA) with 9999

70 Monte Carlo permutations. For factor Precipitation Season, sample sites were classed on a

71 post-hoc basis as either Winter, Summer, or Even depending on when they received the

72 majority of precipitation using a seasonal precipitation overlay (Figure 3.1). For factor Site,

73 different biocrust types from the same sampling site were assigned the same location (Table

74 3.1). Significant PERMANOVA outcomes were tested with Permutational multivariate analyses of

75 dispersion (PERMDISP) where significant PERMDISP results (p(perm)<0.05) suggest differences

76 between sample groupings detected by PERMANOVA may be due to within-group heterogeneity

77 of dispersion rather than true community structure variation. The factor Site was selected for a

78 posteriori hypothesis testing using canonical analysis of principle coordinates (CAP) with

79 vector overlays of seasonal precipitation and temperature averages for each site according to

80 the Australian Government Bureau of Meteorology (Australian Government Bureau of

81 Meteorology 2013). All samples were correctly assigned within the category Site. Weighted-

82 UniFrac analysis was performed upon a phylip formatted matrix of uncorrected pair-wise

83 distances between aligned OTU-representative DNA sequences. OTUs accounting for the

84 greatest variance (to a cumulative value of 5%) between biocrusts were determined using the

85 SIMPER function within PRIMER-6 and visualised in a heatmap. To test the extent geographical

86 distance explained community variation, simple linear regression was performed comparing

87 community composition dissimilarity and phylogenetic distance to geographic distance.

88 Table 3. 1 Sample site locations with symbol key and collection details. .Sample Name (Sample Collection Date of Date of DNA Symbol Sample Location group) Season Collection extraction

JA Bare (JA) -30.5429, 132.1159 Winter July, 2014 December, 2015

JA Early (JA) -30.5429, 132.1159 Winter July, 2014 December, 2015

JA Late (JA) -30.5429, 132.1159 Winter July, 2014 December, 2015

Cobar Bare (Cobar) -31.4949, 145.8402 Autumn April, 2013 April, 2013

Cobar Early (Cobar) -31.4949, 145.8402 Autumn April, 2013 April, 2013

Cobar Mid (Cobar) -31.4949, 145.8402 Autumn April, 2013 April, 2013

Cobar Late (Cobar) -31.4949, 145.8402 Autumn April, 2013 April, 2013

Rolleston -24.34325, 148.7129 Spring September, 2011 December, 2015

Charters Towers -20.532817, 146.1167 Spring September, 2011 December, 2015

Cloncurry, Granada (Cloncurry) -20.15185, 140.48405 Spring October, 2011 December, 2015

Cloncurry, Tara (Cloncurry) -20.112267, 140.44155 Spring October, 2011 December, 2015

Mataranka -15.016367, 133.04985 Spring October, 2011 December, 2015

Paraburdoo, Cell (Paraburdoo) -23.004117, 116.963067 Spring October, 2011 December, 2015

Paraburdoo, CTS (Paraburdoo) -23.0239, 166.9816 Spring September, 2011 December, 2015

Cooloongup -32.3411, 115.7782 Spring October, 2014 December, 2015

89 RESULTS

90 BIOCRUST DIVERSITY AND COMMUNITY COMPOSITION

91 Biocrust communities from across Australia differed significantly in richness, evenness and

92 diversity (Figure 3.2, Supplementary Table A1.2). Particularly, samples from Cooloongup were

93 significantly less rich, even, and diverse than all other sites (p=0.0001, Supplementary Table

94 A1.2). Richness showed the greatest variance while community evenness and diversity were

95 more uniform. The taxonomic sources of each index were examined at the broad level of

96 cyanobacteria and non-cyanobacteria groups. Cyanobacteria contributed the greatest richness

97 at Charters Towers and both Cloncurry sites while non-cyanobacterial groups contributed

98 more for all other significant comparisons.

100 Figure 3.2 Diversity measures for sample site with all taxa (top) and cyanobacteria and non- 101 cyanobacteria (bottom). Significant between site differences for all taxa are in Supplementary 102 Table A1.2. Order of samples as per dendrogram arrangement in Figure 3.3

103 61

104 105 Figure 3.3 Relative abundance of major bacterial classes grouped by phyla. Dendrogram 106 derived from Bray-Curtis dissimilarity.

107 Overall, biocrusts from northern Australia were more diverse than those from southern

108 Australia regarding whole community (p=0.0065), cyanobacteria (p=0.0024) and non-

109 cyanobacteria (p=0.0342). Clustering of each sample based on Bray-Curtis dissimilarity showed

110 replicates grouped together according to sample site (dendrogram of Figure 3.3). The major

111 phyla of biocrusts from all sites were Cyanobacteria (15.6%-72.0%), Proteobacteria (12.5%-

112 36.0%), Actinobacteria (3.9%-23.7%), Chloroflexi (3.9%-14.2%) and Bacteroidetes (0.5%-

113 7.4%). There were seventeen rare phyla including candidate divisions AD3, BRC1, MVP-21,

114 TM7, WSP-2 and WS2. The relative abundance of each taxon varied across the sample sites

115 (Figure 3.3). Cooloongup was the most compositionally distinct, likely driven by the

116 dominance of Nostocophycideae. Samples from more northern regions of Australia had higher

117 abundances of unclassified cyanobacteria than sites from southern Australia. Biocrusts from

118 Cobar were enriched for the Chloroflexi class Ktendonobacteria. The vast majority of

119 cyanobacterial sequences were unclassified at genus level (77%) using the GreenGenes

120 database. The most abundant cyanobacteria identified (>1% of cyanobacterial sequences) to

121 genus level were Phormidium (7.8%), Leptolyngbya (4.4%), Brasilonema (3.5%), Acaryochloris (1.7%)

122 and Scytonema (1.2%). Abundant OTUs were most similar to environmental sequences when

123 compared via BLAST to the GenBank nucleotide database (Benson et al 2013) (data not

124 shown). Examination of OTUs contributing most to inter-sample variation showed

125 Cyanobacteria explained a considerable proportion of the segregation observed between

126 southern and northern sample sites (Figure 3.4).

127 Unconstrained ordination of each sample showed biocrust communities remained tightly

128 grouped according to sample site within two dimensions (Figure 3.5). The positioning of

129 samples according to Bray-Curtis dissimilarity corresponded with a geographical gradient.

130 Specifically, a general trend from left to right conformed with moving from southern to

131 northern Australia and from regions with predominantly winter rainfall to northern regions

132 with predominantly summer rainfall (Figure 3.6). PERMANOVA indicated both sample site and

133 seasonality of precipitation (factors Site and Precipitation Season, respectively) had significant

134 effects on sample community structures. However, for factor Site, within-group variability was

135 not equal across all sites which may account for the significant PERMANOVA finding (Table

136 3.2). To examine other factors and whether differences in sample collection and processing

137

138

139 Figure 3.4 Heatmap of OTUs contributing to a cumulative of 5% of pairwise variation 140 identified via SIMPER. Scale in blue shows percent abundance of OTU within each sample site. 141 OTUs grouped according to Actinobacteria (red), Bacteroidetes (brown), Chloroflexi (yellow), 142 Cyanobacteria (green) and Proteobacteria (purple). 64

143 had significant effects on community contribution patterns, collection year, collection season,

144 site climate class and extraction kit were also tested as factors (Supplementary Figure A1.1).

145 These factors poorly explained the sample ordination and significant PERMANOVA results were

146 likely due to significant sample dispersions and were not pursued.

147 148 Figure 3.5 nMDS of biocrust samples according to site.

149

150 Figure 3.6 nMDS of biocrust samples according to season of predominant rainfall.

151

152

153 Canonical analysis of principal coordinates (CAP) using the a priori of factor Site correctly

154 allocated 100% of samples to their sample site (Figure 3.7). Correlation with environmental

155 variables showed the community composition of samples from mid and northern Australia

156 were most effected by winter and summer temperatures as well as high summer rainfall.

157 Samples from this region also had higher overall diversity. Biocrusts sampled from southern

158 regions were shaped by high winter rainfall.

159 Table 3.2 PERMANOVA and PERMDISP results for factors.

PERMANOVA PERMDISP Res Unique Factor df Pseudo-F P(perm) P(MC) F df1 df2 P(perm) df perms Site 7 37 9.7461 0.0001 9800 0.0001 9.2568 7 37 0.0025 Precipitation 2 42 5.3391 0.0001 9889 0.0001 2.4463 2 42 0.2121 Collection Year 3 41 9.2651 0.0001 9847 0.0001 80.775 3 41 0.0001 Collection Season 2 42 9.4449 0.0001 9889 0.0001 77.166 2 42 0.0001 Kit 1 43 8.0255 0.0001 9900 0.0001 270.08 1 43 0.0001 Site Climate 3 41 6.099 0.0001 9865 0.0001 67.927 3 41 0.0001 Site (Unifrac) 7 37 8.7627 0.001 9831 0.0001 4.6693 7 37 0.02 Precipitation (Unifrac) 2 42 5.1537 0.0001 9896 0.0001 1.4529 2 42 0.3404

160 BIOCRUST PHYLOGENY AND BIOGEOGRAPHY

161 Phylogenetic relationships between samples were determined by calculating weighted UniFrac

162 distances and visualised within a nMDS plot. Smaller distances between samples compared to

163 Bray-Curtis dissimilarities suggested that communities, while compositionally distinct,

164 comprised related species (Figure 3.8). Significant permanova results indicated the geographical

165 factors of site and precipitation influenced community patterns, however, there was significant

166 intra-site variation (Table 3.2). Linear regressions used to examine distance decay relationships

167 showed community composition (Bray-Curtis distances) correlated more strongly than

168 phylogenetic distance with geographical distance (Figure 3.9). 66

169

170 Figure 3.7 Canonical analysis of principal coordinates (CAP) of Bray-Curtis dissimilarity with 171 the factor Site set as a priori. Vectors show direction of correlation with averaged summer and 172 winter temperature and rain values and with Shannon diversity values derived in this study.

173

174 Figure 3.8: nMDS of weighted UniFrac distances between samples.

175 176 Figure 3.9 Linear regression of community differences based on composition (left) and 177 phylogeny (right) to geographical distance.

178 DISCUSSION 179 Biocrusts occur globally in a wide range of primarily arid biomes. Accordingly, they exhibit

180 spatial variability in community structure. In this chapter, the bacterial components of

181 biocrusts from across Australia were examined at the intra-continental scale. Multivariate

182 analyses revealed patterns in community composition and phylogeny consistent with emerging

183 biogeographical models detailing the natural histories of biocrusts. Namely, that climatic forces

184 govern biocrust assembly and that communities, while compositionally distinct, comprised

185 related species.

186 INTRA-CONTINENTAL PATTERNS OF BIOCRUST MICROBIOME DIVERSITY

187 Overall trends in diversity showed biocrusts sampled from northern Australia with

188 predominantly summer rainfall were more diverse than biocrusts from southern Australia with

189 higher winter rainfall (Figure 3.7). This is in contrast to findings based on traditional

190 classification of cyanobacteria from southern Africa where greater morphological diversity was

191 found in sites with higher winter than summer rainfall (Budel et al 2009). The discord may be

192 due to the different approaches used, whereby community sequencing is more sensitive to

193 genotypic diversity masked by homogenous morphology. The dataset presented here also 68

194 included non-cyanobacterial groups which were significant sources of diversity (Figure 3.2 and

195 Supplementary Table A1.2). Higher rates of non-cyanobacterial diversity may be expected as

196 the cell size of heterotrophic bacteria tend to be magnitudes smaller than biocrust-associated

197 cyanobacteria (1-2 µm compared to 10 – 60 µm) and therefore have a greater cell density to

198 biomass ratio. Furthermore, cyanobacteria produce a suite of exometabolites in conjunction

199 with a protective sheath environment, promoting micro-niche differentiation and support of

200 metabolic and phylogenetic diversity (Baran et al 2015).

201 The abundant higher-level taxa identified here (Figure 3.3) are common constituents of arid

202 soils world-wide (Makhalanyane 2015) and specifically of biocrusts (Kuske et al 2012, Thomas

203 and Dougill 2007). The defining community signature distinguishing biocrust microbiomes

204 from non-biocrust arid microbiomes is a dominant abundance of Cyanobacteria and an

205 enrichment of Bacteroidetes (Kuske et al 2012, Steven et al 2013a, Thomas and Dougill 2007).

206 While also abundant within biocrusts, Proteobacteria, Chloroflexi, and Actinobacteria are

207 typically more enriched within sub-crust communities (Steven et al 2013a, Thomas and Dougill

208 2007). This stratification likely represents metabolic niches driven by light and photosynthesis

209 (Garcia-Pichel et al 2003). As expected, all sites sampled here were dominated by filamentous

210 cyanobacteria. As in Chapter 2, no OTUs were assigned to Microcoleus and the most abundant

211 Cyanobacterial genera was Phormidium. The finding of an alternate dominant cyanobacterium to

212 Microcoleus highlights the importance of this study in establishing an Australian framework for

213 biocrust research.

214 Unclassified cyanobacteria were a large component of the dataset and were strong drivers of the

215 distinction between northern and southern samples (Figure 3.4), with unclassified cyanobacterial

216 OTUs more abundant in northern samples (Figure 3.3). Lack of classification of the OTUs at

217 below phylum levels reflects a paucity of arid cyanobacterial sequences within the Greengenes

218 database. Comparison of the most abundant unclassified cyanobacterial OTUs to the Genbank

219 nucleotide database showed these sequences were most similar to undescribed environmental

220 sequences unlikely to be included in curated databases for taxonomic classification (data not

221 shown). Previous microscopy efforts have identified 23 species of cyanobacteria from these

222 regions (Williams and Büdel 2012), including Porphyrosiphon notarissi, Nostoc commune, Microcoleus

223 spp. and Scytonema hoffman-bangii, however, these have yet to be genetically verified. While these

224 genera are common to biocrusts from around the world (Büdel et al 2016), genetic profiling here

225 suggests northern Australian biocrusts contain novel cyanobacterial species.

226 BIOGEOGRAPHY OF BACTERIA WITHIN AUSTRALIAN BIOCRUSTS

227 In terms of community assembly, the exploration of compositional and phylogenetic data

228 patterns presented here indicate Australian biocrusts are structurally distinct from each other

229 yet phylogenetically similar. Sample ordinations derived from Bray-Curtis dissimilarities

230 showed biocrust microbiomes exhibit spatial variability on an intra-continental scale and

231 grouped strongly according to site (Figure 3.5). However, the same strength of grouping was

232 not seen for phylogenetic analysis where samples were more interspersed across the nMDS plot

233 (Figure 3.8). Similar findings have been reported for cyanobacteria of Northern American

234 deserts, for example, Tehuaćan biocrusts comprised assemblages distinct from other deserts

235 despite sharing cosmopolitan species (Garcia-Pichel et al 2013, Rivera-Aguilar et al 2006).

236 Comparison between Bray-Curtis dissimilarity and phylogenetic distance was explored further

237 using distance-decays graphs to quantify their relationships with geographical distance. Despite

238 sampling limitations resulting in incomplete coverage along the full geographical distance

239 examined, correlation analyses showed important patterns. While in each case biocrust

240 resemblance decreased over geographical distance (Figure 3.9), Bray-Curtis measures again 70

241 showed a stronger correlation with geographical distances indicating greater biogeographical

242 organisation of samples.

243 These findings begin to inform the factors which govern the assembly of microbiomes

244 (Nemergut et al 2013). The different patterns and rates of change between composition and

245 phylogeny suggests that as communities, biocrusts may first adapt to a new niche via

246 compositional changes, that is, the selected enrichment of dominant taxa before diversification

247 via evolutionary changes. The capacity for dispersal is an important consideration regarding

248 this. Current studies suggest that arid cyanobacteria do not disperse well (Abed et al 2012, Bahl

249 et al 2011). Temporal analysis by Bahl et al (2011) of the desert cyanobacterium Chroococcidiopsis

250 indicated the global population of arid cyanobacteria diverged evolutionary histories before the

251 formation of the modern continents. Specifically within Australia, taxonomic profiling of a

252 large dust storm by Abed et al (2012) showed biocrust-cyanobacteria represented 2% of the

253 dust microbiome despite representing over 50% of the source community. Alternatively,

254 filamentous, crust-forming cyanobacteria have been shown to migrate through tops soils at a

255 rate of up to 2 cm per month (Nathali Machado, personal communication). Given the ancient

256 histories of both cyanobacteria as colonisers of land and Australia as a geographically isolated

257 continent (Beraldi-Campesi and Garcia-Pichel 2011, Bowker et al 2016), Australian biocrusts

258 may comprise an endemic communal cohort of bacteria structurally differentiated over

259 ecological distances and timescales.

260 SEASONALITY OF PRECIPITATION EFFECTS BIOCRUST COMPOSITION

261 Seasonality of rainfall was a significant driver of community differentiation across an intra-

262 continental scale and is likely a major factor explaining biogeographical distributions of

263 bacteria within biocrusts (Table 3.2). While Australian biocrusts comprise bacteria of related

264 lineage, environmental conditions of different regions differentially enrich certain members

265 over others resulting in compositional distinctions. Such patterns have also been observed

266 within American biocrusts where temperature dictated the dominance of alternate Microcoleus

267 spp. in hot and cold deserts (Garcia-Pichel et al 2013). Here, biocrust samples from northern

268 Australia correlated with high summer rainfall and temperatures while samples from southern

269 Australia correlated with high winter rainfall. This trend supports a growing body of research

270 showing climatic patterns of precipitation and temperature significantly affect biocrust

271 structure and coverage (Bowker et al 2016, Coe et al 2012, Escolar et al 2015, Garcia-Pichel et

272 al 2013, Rogers 1972, Steven et al 2015). Specifically, the timing of precipitation events with

273 high temperatures have a critical influence on the presence and relative abundance of

274 photosynthetic organisms (Bowker et al 2016). Arid-adapted photosynthetic organisms often

275 undergo desiccation as a survival strategy during periods of drought or extreme evaporation.

276 Upon rehydration, carbon storage molecules are mobilised to power the reactivation of

277 photosystems, resulting in a pulse of CO2 detected (Lange 2003). However, if the conditions

278 after rehydration are not conducive to productive photosynthesis, e.g. too-short a wetting

279 period or high temperatures inducing heat stress, a cellular carbon deficit can result meaning

280 the organism will fail to resurrect at the next wet-dry cycle (Coe et al 2012). Due to the critical

281 nature of these systems, phenotypic variations of desiccation traits within the cohort of

282 Australian biocrust cyanobacteria are likely strong metabolic factors explaining patterns

283 observed here (Ferrenberg and Reed 2017).

284 Previous drought-tolerance trials of biocrust-cyanobacteria from across the northern

285 Australian savannah provide preliminary insights as to what traits may be selected for in the

286 high summer rainfall communities profiled here (Williams and Büdel 2012, Williams et al

287 2014). Savannah cyanobacteria were resistant to resurrection during the summer season despite

288 rehydration, a strategy that avoids premature reactivation of photosystems and heat stressed

289 photosynthesis (Williams et al 2014). The authors nominate a water-regulating role of EPS

290 while more recent work on cyanobacteria has shown timing of light exposure is also important

291 (Oren et al 2017). These strategies likely impart an advantage allowing strains to colonise

292 hostile soils with less competition from macro-components. Indeed, the restriction of lichens

293 and mosses to winter rainfall areas (Rogers 1972) may account for the higher diversity of

294 cyanobacteria observed in northern samples here.

295 Northern biocrusts were dominated and characterised by genetically unclassified cyanobacteria

296 (Figures 3.3 and 3.4). These poorly documented groups may prove important as climate change

297 takes effect and precipitation patterns are altered (Rodriguez-Caballero et al 2018). Rainfall

298 manipulation studies conducted over several years have shown increasing summer rainfall events

299 significantly modifies the structure and function of biocrusts (Steven et al 2015). Macro-

300 components were greatly impacted which reduced the visible coverage of biocrusts while, after

301 an initial loss, cyanobacteria recovered both in abundance and biomass at six years. For Australia,

302 northern biocrusts may serve as reservoirs of biological and functional diversity that prove

303 critical in maintaining biocrusts on climate-impacted arid lands (Garcia-Pichel et al 2013,

304 Rodriguez-Caballero et al 2018). Future work will benefit from both culture-based methods and

305 NGS approaches to enumerate and describe these bacteria (Büdel et al 2016).

306

CHAPTER 4:

BIOCRUST SECONDARY METABOLISM IS DRIVEN BY SEASONALITY OF PRECIPITATION ON AN INTRA- CONTINENTAL SCALE

1 INTRODUCTION 2 Organisms across the tree of life produce a vast catalogue of small molecules (secondary

3 metabolites) which function to adapt cells to their local environment (Davies and Ryan 2012).

4 While in many cases the ecological function of these compounds is unknown, they have

5 proved particularly useful for medicine and industry due to their non-native bioactivities

6 (Newman and Cragg 2016). Conversely, many of these secondary metabolites, also known as

7 natural products, pose as potent health risks to humans and agriculture (Woodhouse et al

8 2014). Many classes of natural products are synthesised by large, ribosomally-independent

9 megasynthases called non-ribosomal peptide synthetases (NRPS) and polyketide synthases

10 (PKS)(Pearson et al 2016). In search of novel biochemistry, genomic analyses have revealed

11 microorganisms harbour a greater genetic potential for natural product production than

12 observed via chemistry (Micallef et al 2015). Furthermore, many bacteria remain recalcitrant to

13 culturing, hindering identification of their biosynthetic gene clusters. Consequently, culture-

14 independent methods have been developed to target these cryptic genes.

15 Deep sequencing of NRPS and PKS genes within whole microbiomes is an established

16 approach to elucidating the potential of environments for novel compound discovery

17 (Charlop-Powers et al 2014, Charlop-Powers et al 2016, Lemetre et al 2017, Woodhouse et al

18 2013). Due to the difficulty in determining molecular function from amplicon sequences, often

19 the aim is to provide indications as to how to direct future efforts in exploiting especially rich

20 niches or taxa sequences (Charlop-Powers et al 2015, Ziemert et al 2012). This approach has

21 provided valuable insights as to the prevalence of biosynthetic genes and factors controlling

22 their distribution and community structure. For example, Charlop-Powers et al (2015)

23 determined arid soils were the richest source of bioactives across a wide range of soil biomes.

24 In pursuit of this potential, a follow up study focusing on arid soil samples from across 75

25 Australia revealed latitude was a primary factor driving the distribution of NRPS and PKS

26 genes (Lemetre et al 2017). However, the authors were unable to associate an environmental

27 factor with this observation. Linking the distribution patterns of NRPS and PKS genes with

28 environmental factors can provide valuable insights as to the potential ecological roles of their

29 products within microbial communities. Furthermore, it can indicate potentially novel

30 metabolic strategies involving novel biosynthetic pathways and compounds.

31 Biocrusts are a complex top-soil community of microorganisms and non-vascular plants which

32 occur across Australia and the globe. They are equipped with a range of environmental

33 strategies to endure extreme arid conditions and are enriched with cyanobacteria, a phylum

34 renown for producing a prolific array of secondary metabolites (Pearson et al 2016). Genome

35 surveying of biocrusts presents as an appropriate next-step to build on previous work in

36 elucidating the secondary metabolite potential of arid soil microbiomes. Here, findings from

37 previous chapters are used to probe NRPS and PKS datasets generated from the genomics of

38 biocrusts from across Australia. It is predicted that due to having ecologically adapted roles,

39 differences in the NRPS and PKS profiles of biocrust communities will exhibit

40 biogeographical variation.

41 METHODS

42 AMPLIFICATION AND SEQUENCING OF NRPS AND PKS GENES

43 Genomic DNA from samples in Table 3.1 were used as the template input for amplification of

44 the PKS ketosynthase (KS) and NRPS condensation (C) domains from biocrust samples from

45 across Australia. A two-step PCR approach was used to generate amplicon libraries for each

46 gene target which were then pooled and run together using a v3 Reagent Kit with 2x300 bp

47 cycling on an Illumina MiSeq instrument. The first round of PCR used degenerate primers

48 modified with a 5’T7 overhang on the forward and a 3’M13 overhang on the reverse strands

49 (Supplementary Table A1.4). These overhangs served to incorporate sample specific barcodes

50 in the second round of PCR. The KS domains of PKS modules were targeted using the

51 DKF/DKR primer pair (Moffitt and Neilan 2003) while the C domain of NRPS modules was

52 targeted using the CNDMF/DCCR primer pair (Woodhouse et al 2013) (See supplementary

53 Table A1.3 for PCR conditions). Target bands (NRPS = ~1000 bp, PKS = ~700 bp) were

54 excised from a 1% agarose gel after electrophoresis and recovered using a ZymoClean Gel

55 Extraction Kit (Zymo Research). The purified amplicon product served as the template for the

56 second, barcode-incorporating PCR (Supplementary Table A1.4). The barcoded amplicon

57 products were normalised and pooled using a SequalPrep Normalisation Plate (Thermo Fisher

58 Scientific) and submitted to the Ramaciotti Centre for Genomics (UNSW, Australia) for

59 sequencing.

60 BIOINFORMATIC ANALYSIS

61 Due to the short read length of the amplicon products for each gene target, only the forward

62 reads were used in order to avoid artificial inflation of diversity measures due to poor

63 confidence in contig formation (Kozich et al 2013, Nielsen et al 2016). Each raw dataset was

64 processed following standard procedures within the mothur pipeline (Kozich et al 2013)

65 except custom-made databases were used for alignment and classification steps. The custom-

66 made databases were modeled on the curated databases of Natural Product Domain Seeker

67 (NaPDoS) (Ziemert et al 2012). Condensation and ketosynthase datasets were downloaded

68 from NaPDoS (July, 2016) and used to query the GenBank nucleotide database using the

69 tblastn function (States and Gish 1994). Identity cutoffs of 60% were used. Sequences were

70 then trimmed to a consensus region and aligned using muscle v3.8.31 (Edgar 2004). For each

71 C and KS domain reference sequence, the Genbank organism source information was 77

72 maintained and used to classify the OTUs according to taxon groups. After trimming for

73 ambiguous base pairs and greater than 8 homopolymers, sequences were aligned to their

74 respective databases. The alignments were screened for reads less than 200 bp and then re-

75 aligned. For the C domain alignment, the consensus region was set at positions 1091 to 1765

76 while for the KS alignment, the consensus region was found by setting the end at position

77 2652 and removing any sequence that started after 90% of sequences. A pre-clustering step

78 was performed for each data set at 97% (equal to 3 differences per 100 bp) (Charlop-Powers et

79 al 2016). Chimeras were detected using the Uchime function within mothur. Datasets were

80 rarefied to the sample with the lowest number of reads (samples with less than 9,000 reads

81 were removed). For the C domain dataset, samples were rarefied to 12548 sequences, for the

82 KS dataset they were rarefied to 9398 sequences. OTUs were formed at 90% nucleotide

83 similarity and classified according to source organism using the unaligned custom database. To

84 achieve NRPS and PKS domain classification, nucleotide sequences of OTU representatives

85 were translated to amino acid sequences using Geneious version 7.1.9. Any translations with

86 stop codons were removed. In order to achieve the minimum 100 aa length required for

87 querying the NaPDoS database, a string of 35 Js were artificially added to the translated

88 sequences. These additions did not align with any part of the NaPDoS database and were not

89 included in the alignment length or percent identity metrics (Ziemert et al 2012).

90 STATISTICAL ANALYSES

91 For analysis, singleton and doubleton OTUs were removed and where required, the data

92 standardised by total. To examine diversity and variance at local and intra-continental scales,

93 the data was processed both as a whole set including all sites and as a sub-set including only

94 Cobar samples. In-built calculators within PRIMER-6 (version 6.1.11, Primer-E Ltd, UK) were

95 used to determine richness (Margalef’s Index), evenness (Pielou’s Index) and diversity 78

96 (Shannon Index) (Anderson 2008 Primer reference). Differences between sites in richness,

97 evenness and diversity were tested using one-way ANOVA followed by examination with

98 Tukey multiple comparisons for significant differences. To determine the broad taxonomic

99 sources of diversity at each site, cyanobacterial and non-cyanobacterial OTUs were examined

100 separately where each taxon dataset was rarefied and standardized and site means were

101 compared using t-tests assuming non-Gaussian distributions. Correlations between 16S rDNA

102 and KS and C diversity were visualised and tested using simple linear regression using values

103 derived from rarefied datasets.

104 For beta-diversity analyses, a Bray-Curtis dissimilarity matrix was derived from square-root

105 transformed OTU abundance data. Samples were visualized in 2D ordination via nMDS with

106 temperature and rain vector overlays. There were not enough replicates to perform CAP

107 analysis. For the whole data-set, weighted-unifrac was performed using representative

108 sequences from OTUs with 4 or more sequences. For each secondary metabolite domain,

109 RELATE function was used to compare the three different types of distances used: Bray-Curtis

110 (compositional) weighted-unifrac (phylogenetic) and geographical (kilometers). All OTUs from

111 the entire data set contributing greater than 0.05% of sequences to their respective gene targets

112 were tested for co-occurrence using the Sparcc function within mothur with 100 iterations and

113 10,000 permutations. False-discovery rates were maintained below 5% using Benjamini-

114 Hochberg corrected p values (Benjamini and Hochberg 1995). To pursue which bacteria

115 within the biocrust communities where most associated with NRPS and PKS genes, only 16S

116 rDNA-C domain and 16S rDNA-KS domain correlations with r values ±0.5 were further

117 examined. This resulted in 9913 correlations between 751 OTUs. Measures of betweeness

118 centrality and degree were used to identify top connected OTUs which were compared to

119 GenBank using then blastn algorithm.

120 121 RESULTS

122 SECONDARY METABOLITE DIVERSITY WITHIN AUSTRALIAN BIOCRUSTS

123 High through-put screening of condensation (NRPS) and ketosynthase (PKS) domains from

124 biocrusts across Australia displayed similar trends in diversity measures (Figure 4.1). For each

125 enzymatic domain, richness showed the greatest variation between samples while evenness was

126 more uniform. Cooloongup consistently had the lowest diversity, followed by samples from

127 Paraburdoo, Rolleston and Cloncurry. Charters Towers had the highest NRPS measures while

128 JA Bare had the highest PKS diversity (Supplementary Tables A1.5 and A1.6). The positive

129 correlation between NRPS and PKS counts for each index was strongest for richness

130 (R2=0.733) (Figure 4.2). Evenness was poorly correlated (R2=0.1157) showing within-sample

131 differences in the structure of the relative abundance of NRPS and PKS OTUs. A moderate

132 positive correlation was observed for diversity (R2=0.4056). Overall, biocrusts were

133 significantly more rich (p=<0.0001), even (p=0.0277) and diverse (p=0.0004) in NRPS than

134 PKS domains. To determine whether diversity trends observed between NRPS and PKS genes

135 were also true for between functional and taxonomic genes, linear correlation analysis was

136 performed using 16S rDNA data from Chapter 3. Comparative positive correlation patterns

137 were found between 16S rDNA and NRPS/PKS diversity measures (Figure 4.2). NRPS was

138 consistently more strongly correlated with 16S rDNA diversity than PKS diversity. No

139 significant difference (p>0.05) in diversity was found between biocrust stages at Cobar for

140 either gene target (data not shown).

141

142 Figure 4.1 Richness, evenness and diversity measures for condensation (NRPS) and 143 ketosynthase domain (PKS) OTUs from biocrusts across Australia. Whole community shown 144 in grey (top rows) and cyanobacteria and non-cyanobacteria in colour (bottom rows).

145

146 Figure 4.2 Linear correlations between diversity measures of NRPS and PKS, NRPS and 16S 147 rDNA, and PKS and 16S rDNA.

148 TAXONOMIC AND FUNCTIONAL CLASSIFICATION OF C AND KS DOMAINS

149 The C and KS domain gene sequence datasets were classified by both source organism

150 (taxonomic) and by domain class (functional) (Figure 4.3 and Figure 4.4). The majority of

151 sequences were assigned to cyanobacterial taxa with Actinobacteria, Firmicutes, and

152 Gammaproteobacteria also dominant across both datasets. Bacteroidetes and

153 Alphaproteobacteria were also prominent phyla for NRPS. Abundant groups were resolved to

154 order level for NRPS where Nostocales, Synechococcales, and Pseudomonadaceae harboured

155 the majority of sequences and taxonomic diversity. Taxonomic classification of PKS genes

156 revealed the majority of sequences belonged to a select group of cyanobacteria genera,

157 especially Symploca and Anabaena. For each catalytic domain type, Cobar samples had the

158 greatest taxonomic richness. For functional classification, the majority of protein sequences

159 had low similarity (<85%) to the NaPDoS database (Figure 4.5). This restricted the capacity to 82

160

161 Figure 4.3 Dendrogram clustering of biocrust samples based on NRPS Bray-Curtis 162 dissimilarity with taxonomic and domain class classification of OTUs. Size of circle is average 163 relative abundance within each biocrust sample type. Colour of circle is taxonomic OTU 164 richness. LCL= domains which catalyse the formation of a peptide bond between two L- 165 amino acids.

166

167

168 Figure 4.4 Dendrogram clustering of biocrust samples based on PKS Bray-Curtis dissimilarity 169 with taxonomic and domain class classification of OTUs. Size of circle is average relative 170 abundance within each biocrust sample type. Colour of circle is taxonomic OTU richness. 171 Modular=Cis-AT modular, KS=KS1 starter domains, Trans=Trans-AT, Hybrid=KS domains 172 which occur alongside NRPS components.

173 assign putative functional pathways and sequences were classified at domain class level. At the

174 taxonomic level Order (therefore including phylum level), 25% of NRPS and 7.5% of PKS

175 reads were unclassified. Unclassified PKS reads were significantly less similar to the NaPDoS

176 database (p>0.0001) while unclassified NRPS reads showed no difference. The vast majority of

177 NRPS domain classes were LCL which catalyse the peptide bond between two L-amino acids.

178 All KS sequences were from Type I PKSs. The ketosynthase dataset had greater domain class

179 diversity than the condensation dataset, with the majority having modular activity.

180

181 Figure 4.5 Similarity of C domain sequences (NRPS, left) and KS domain sequences (PKS, 182 right) to the respective NaPDoS databases.

183

184 Figure 4.6 nMDS of Bray-Curtis dissimilarity distance for the Cobar data subset for A) NRPS 185 and B) PKS OTUs.

186 COMMUNITY STRUCTURE AND BIOGEOGRAPHY OF KS AND C DOMAINS

187 No significant grouping of samples from the Cobar biocrust stages was detected via ordination

188 (Figure 4.6) or permanova (For NRPS: Pseudo-F=1.2047, P(MC)=0.2655. For PKS: Pseudo-

189 F=1.4269, P(MC)=0.1275). The biosynthetic profiles of biocrusts from all sites were examined

190 using distance-based methods which compared composition (Bray-Curtis), phylogenetic

191 relatedness (UniFrac) and geographical distance (kilometres). Multidimensional scaling revealed

192 NRPS and PKS genes displayed similar site-based clustering to each other and to the 16S

193 rDNA dataset (Figure 3.5) for both Bray-Curtis and UniFrac distances (Figure 4.7).

194

195 Figure 4.7 nMDS of Bray-Curtis dissimilarity and UniFrac distance for whole NRPS and PKS 196 OTU datasets. Shannon=Shannon diversity index.

197 In each case, winter rainfall was most strongly associated with Cooloongup, the most southern

198 site, while summer rain was most strongly associated with Mataranka, the most northern site.

199 Diversity correlated strongest with JA Bare samples. Summer and winter temperatures

200 correlated with northern samples with the exception of the PKS Bray-Curtis NMDS where

201 they were associated with JA samples. Permutational multivariate analysis of the ad-hoc factors

202 Site and Precipitation Season showed that timing of rainfall at a site significantly affected the

203 composition of functional genes within biocrusts (Table 4.1).

204 Table 4.1 PERMANOVA and PERMDISP results for Bray-Curtis and UniFrac matrices.

Factor PERMANOVA PERMDISP Uniqu Res Pseudo- e df df P(perm NRPS df df F P(perm) perms P(MC) F 1 2 ) Site 7 26 9.8348 0.0001 9830 0.0001 28.465 7 26 0.001 Precipitation 2 31 4.5518 0.0001 9904 0.0001 3.0606 2 31 0.178 Site (Unifrac) 7 26 10.356 0.0001 9823 0.0001 24.344 7 26 0.001 Precipitation (Unifrac) 2 31 3.9247 0.0001 9907 0.0002 9.9634 2 31 0.005 PKS Site 7 25 7.3345 0.0001 9836 0.0001 12.104 7 25 0.001 Precipitation 2 30 5.1404 0.0001 9894 0.0001 1.5562 2 30 0.429 Site (Unifrac) 7 25 6.017 0.0001 9852 0.0001 3.3766 7 25 0.382 Precipitation 0.4069 (Unifrac) 2 30 3.5837 0.0001 9910 0.0003 6 2 30 0.767 205

206 Phylogenetic variation of PKS genes was also significantly driven by both location and

207 precipitation. Visual similarities in the ordination of both taxonomic and functional datasets

208 were tested using Mantel-type correlations via the RELATE function within PRIMER. For all

209 combinations, there was a strong similarity in the clustering of biocrusts based on Bray-Curtis

210 distances, suggesting similar rates of shared OTUs (Table 4.2). Clustering based on phylogeny

211 was less strong but still showed a moderate correlation. Geographical distance was moderately

212 correlated with community composition but was poorly correlated with phylogenetic

213 relatedness (Figure 4.8). For both NRPS and PKS datasets, there were no shared OTUs across

214 all sites.

215

216 Figure 4.8 Linear regression of community differences based on composition (top) and 217 phylogeny (bottom) to geographical distance.

218 Table 4.2 RELATE values for comparison of sequence datasets according to different distance 219 methods.

Compared datasets Distance type Rho p value Bray-Curtis 0.862 0.0001 NRPS and 16S rDNA UniFrac 0.547 0.001 Bray-Curtis 0.828 0.0001 PKS and 16S rDNA UniFrac 0.627 0.001 Bray-Curtis 0.869 0.001 NRPS and PKS UniFrac 0.677 0.001 220

221 RESERVOIRS OF BIOSYNTHETIC POTENTIAL AND INFLUENTIAL OTUs

222 Co-occurrence analysis was performed to achieve OTU-level resolution of taxonomic groups

223 most associated with NRPS and PKS groups as well as to determine which NRPS and PKS

224 OTUs play central roles for biocrust community members. The majority of correlations were

225 positive (81.5%). Scale-free display of the OTU-OTU network formed three major, highly-

226 connected hubs with four smaller satellite hubs (Figure 4.9). One of the major hubs was

227 defined by PKS-cyanobacteria associations. The measures of betweeness centrality (BC) and

228 degree were used to identify integral nodes for each gene type within the network. For PKS,

229 many of the most influential nodes were assigned to Symploca sp. HPC-3, an isolate from the

230 hypersaline Shark Bay (Supplementary Table A1.9). The majority of central NRPS nodes were

231 Nostocophycideae and Synechococcophycideae. Greater phylum-level diversity was seen for

232 16S rDNA nodes with alphaproteobacteria and cyanobacteria being highly connected.

233 Alphaproteobacteria were the most over represented taxon in the network, contributing more

234 nodes to the network than proportionate to their relative abundance (Figure 4.10).

235

236 Figure 4.9 Scale free network of significant correlations (r>0.5) of 16S rDNA to NRPS/PKS 237 OTUs. Node size reflects betweeness centrality. Arrow indicates PKS-cyanobacteria hub. 89

Acidobacteriia [Chloracidobacteria] Solibacteres Actinobacteria Rubrobacteria Thermoleophilia Cytophagia [Saprospirea] Anaerolineae Chloroflexi Ktendonobacteria Chloroplast Nostocophycideae Oscillatoriophycideae Synechococcophycideae Cyanobacteria_unclassified Bacilli Alphaproteobacteria Betaproteobacteria Deltaproteobacteria Proteobacteria_unclassified Unclassified

-4 -2 0 2 4 6 8 10

238 239 Figure 4.10 Ratio of relative abundance to number of network nodes at order level.

240

241 DISCUSSION 242 Australian biocrust microbiomes have been shown to taxonomically vary according to

243 developmental stage (Chapter 2) and geographically on an intra-continental scale (Chapter 3).

244 Here, high through-put sequencing of the same communities targeting NRPS and PKS

245 domains showed these functional genes exhibited highly similar distribution patterns to

246 taxonomic genes on the intra-continental scale but not on the local scale. As with 16S rDNA

247 analyses, seasonality of precipitation was underscored as a driving factor distinguishing

248 biocrusts of differing locations. Strong positive correlations were found for measures of

249 richness between all gene types, signifying overall taxonomic richness of sites is a fair indicator

250 for small-molecule gene richness in Australian biocrusts. Network analysis highlighted several,

251 otherwise cryptic groups as central in the expression, mediation, or utilisation of NRPS and

252 PKS genes. Together, these findings contribute to establishing a complex biosynthesis atlas of

253 Australia.

254 INTRA-CONTINENTAL DISTRIBUTION OF BIOSYNTHETIC GENES IS LINKED 255 TO SEASONALITY OF PRECIPITATION

256 Within Australian soils, Lemetre et al 2017 found strong correlations of biosynthetic gene

257 organisation with latitude. However, the authors were unable to determine which

258 environmental factors drove this pattern. Data presented in this chapter suggests that, as with

259 16S rDNA biogeography, NRPS and PKS gene distributions are driven in part by seasonality

260 of precipitation and would be the primary factor explaining the observations made by Lemetre

261 et al (2017). Cooloongup, the southernmost site, was strongly correlated with winter rainfall

262 while Mataranka, the northernmost site was strongly correlated with summer rainfall in terms

263 of both composition and phylogeny of the microbial community present (Figure 4.7). This

264 trend was supported by significant PERMANOVA results for sites grouped according to

265 seasonality of precipitation (Table 4.1). Seasonal temperatures were less consistent, though

266 future work applying this factor with larger datasets may reveal associated patterns.

267 Alpha diversity correlations between both biosynthesis and taxonomic gene targets revealed

268 enlightening trends which may be used to direct future bioprospecting endeavours. Typically,

269 ecologists focus on diversity indexes derived from both richness and evenness measures as the

270 interplay between these two factors is important for ecosystem function and robustness.

271 However, in the search for environments harbouring the greatest potential for novel natural

272 products, community richness serves as a better indicator (Lemetre et al 2017). Such an

273 approach gives weight to the rarer functional diversity able to be captured by deep-sequencing

274 rather than to a few dominating producers. However, it is important to note that each unique

275 sequence or OTU detected does not necessarily represent a unique gene cluster and associated

276 product. Due to the multi-modular arrangement of NRPS and PKS biosynthetic gene clusters,

277 multiple C and KS domains may occur within a single cluster to produce one compound

278 (Kalaitzis et al 2009). Nonetheless, OTU data generated in this thesis showed a strong positive

279 correlation between taxonomic and functional richness (Figure 4.2). While arid soils have

280 previously been identified as biosynthesis hot-spots (Charlop-Powers et al 2014), this finding

281 indicates future natural product discovery efforts would benefit from specifically targeting

282 taxonomically rich arid top-soils. Due to the cryptic nature of biosynthetic gene expression,

283 correlations with environmental parameters may not always be evident. The work of this thesis

284 demonstrates the benefit of conducting 16S rDNA analyses alongside functional gene

285 profiling. Given this, substantial advances may be gained from integrating current, Australia-

286 wide microbiome profiling efforts (Bissett et al 2016) with biosynthesis surveys.

287 Beta-diversity analyses highlighted important differences in the relationships between

288 secondary metabolite composition (Bray-Curtis) and phylogeny (Unifrac) with 16S rDNA

289 profiles. While ordination of each distance matrix resulted in similar site-based clustering

290 (Figure 4.7), a lower Rho value was observed for phylogenetic relatedness whereby NRPS and

291 PKS genes maintained stronger site-based clustering compared to 16S rDNA data (Table 4.2).

292 This may speak to the dynamics between adaptation and speciation that affect biocrust

293 community assembly. Chapter 3 suggested biocrust communities initially adapt to new spatial

294 niches via compositional changes, followed by strain speciation. Data here builds on this

295 picture suggesting differentiation of the secondary metabolome may be an observable step

296 initiating speciation (Lemetre et al 2017). This is consistent with our understanding of the

297 genetic mobility of NRPS and PKS domains both within single organisms (Fischbach and

298 Walsh 2006, Mootz et al 2002) and across communities (Jensen 2010) as well as the proposed

299 ecological functions of their products (Davies and Ryan 2012). This is likely operating in

300 conjunction with environmental filtering where seasonality of precipitation acts as a selective

301 pressure enriching certain bacteria and their genomes over others.

302 Previous work has showed geographical distance is a key factor describing the relationships

303 between the genetic profiles for secondary metabolism of soil microbiomes (Charlop-Powers

304 et al 2014). Here, distance decay graphs were used to explore this concept further. Despite

305 limitations in sampling resulting in incomplete coverage along the full geographical distance

306 examined, correlation analyses showed important patterns (Figure 4.8). In support of work by

307 Charlop-Powers et al 2014, biocrusts closest together most strongly resembled each other in

308 composition of NRPS and PKS domains. The further biocrusts were away from each other,

309 the less similar their functional gene profiles (Figure 4.8, top row). However, the phylogenetic

310 relationship with distance was less pronounced and may be further reduced upon additional

311 sampling (Figure 4.8, bottom row). The relative conservation of C and KS genetic sequences

312 across an intra-continental scale may be reflective of the shared extreme environmental

313 conditions biocrusts often endure. Several ubiquitous small molecules have well established

314 roles in adaptations to arid niches. For example, compatible solutes serve as osmoprotectants

315 during desiccation cycles (Billi and Potts 2002, Rajeev et al 2013) while scytonemin and MAAs

316 absorb harmful UV radiation (Gao and Garcia-Pichel 2011). Correlation trends here suggest

317 there may be a suite of similarly common yet undescribed NRPS and PKS products involved

318 in critical cellular maintenance for arid extremobiosphere survival.

319 RESERVOIRS OF BIOSYNTHETIC POTENTIAL AND INFLUENTIAL OTUs

320 While overall trends correlated taxonomic richness with biosynthetic richness, diversity

321 measures highlighted certain sample sites as particularly rich sources of NRPS and PKS genes.

322 Charter’s Towers and Cobar were the richest sites in C domains while Charter’s Towers,

323 Mataranka, Rolleston, JA Bare, and Cobar Late were the richest in KS domains (Figure 4.1). In

324 addition to targeting rich sites, identifying rich microbes within biocrusts serves to further

325 narrow the search field and prioritise lines of enquiry. The cyanobacterial orders Nostocales

326 and Synechococcales, as well as the Gammaproteobacterium Pseudomonas harboured the

327 majority of NRPS abundance and richness (Figure 4.3). This is unsurprising given

328 cyanobacteria are renown prolific producers of secondary metabolites (Pearson et al 2010)

329 while the ecological roles and biotechnological potential of Pseudomonas secondary metabolites

330 have proved a productive area of research (Boruah and Kumar 2002, Deveau et al 2016, Spago

331 et al 2014). For PKS, Symploca and Anabaena cylindrica were primary reservoirs of KS domains

332 within biocrusts. Both organisms were detected via 16S rDNA analysis though were at low

333 abundances (0.76% and 0.06%, respectively. Data from Chapter 3). 94

334 In terms of higher level taxa, Alphaproteobacteria were notably over-represented within the

335 functional network, contributing more nodes than expected from their relative 16S rDNA

336 abundance (Figure 4.10). Oscillatoriophycideae and Synechococcophycideae were also

337 relatively over-represented while Nostocophycideae and Actinobacteria appeared less

338 influential in biocrust secondary metabolism. This is in contrast to previous findings where

339 NRPS and PKS gene richness within soils specifically correlated with actinomycetes (Charlop-

340 Powers et al 2014). While this disparity may be due to different niches being surveyed, it may

341 also be due to different primer sets targeting distinct domains. The authors who developed the

342 screening methods used here acknowledge that the primers likely have a bias against groups

343 with high GC content, which may skew the coverage against actinobacteria and firmicutes

344 (Woodhouse et al 2013). Network measures of centrality identified both low-abundance

345 organisms and genes with potentially important functional roles within biocrusts (Figure 4.9

346 and Supplementary Tables A1.7-A1.9). According to degree, OTUs with the greatest similarity

347 to uncultured cyanobacteria from the deserts of western America were the most highly

348 connected nodes (Supplementary Table A1.8), suggesting these strains are keystone functional

349 members.

350 POSSIBLE NRP AND PK FUNCTION IN BIOCRUSTS

351 Overall, the gene sequences generated here had low (<80%) similarity to public databases

352 (Figure 4.5 and Supplementary Tables A1.7 - A1.9)(Ziemert et al 2012), particularly for NRPS

353 reads. This is an important finding signifying biocrusts warrant pursuit as sources of novel

354 natural products. Further work is required to validate the growing list of biosynthetic gene

355 clusters via the linkage with their corresponding natural products, including chemical analyses

356 and gene cluster capture and expression (Moss et al 2016). Generally across Australia, the non-

357 cyanobacterial community was more rich in NRPS while the cyanobacteria was more rich in 95

358 PKS (Figure 4.1). The majority of cyanobacterial ketosynthase genes were of the Type 1

359 Modular PKS subgroup Cis-Acyltransferase (Cis-AT) domains while the majority of

360 condensation domains were LCL NRPS (forming condensation bonds between two L-amino

361 acids) (Figures 4.3 and 4.4). Each subgroup produces a broad range of products, however, due

362 to limited taxonomic resolution it is difficult to clarify the differential enrichment of NRPS and

363 PKS through understanding molecular function. Yet, given the considerable contributions of

364 cyanobacteria to biocrust formation and maintenance, the enrichment of PKS within this

365 taxon may speak to the relative importance of PKS products in biocrusts. This is supported by

366 network analysis which showed a highly-connected hub built primarily of PKS-cyanobacteria

367 connections (Figure 4.9). PKS nodes also had the highest betweeness centrality where PKS

368 OTU24 (of Symploca origin) was highlighted as a pivotal mediator of network information

369 (Supplementary Table A1.9).

370 Despite low sequence identities limiting confidence in taxonomic assignments, several classes

371 of compounds were consistently documented through-out the current dataset and types of

372 analyses so as to warrant further discussion here. For example, highly abundant and highly

373 central PKS OTUs were assigned to the curacin and jamaicamide pathways (Supplementary

374 Tables A1.7 - A1.9). Each are products of hybrid NRPS/PKS pathways produced by Lyngbya

375 majuscula (Moorea producens), a mat-forming marine cyanobacterium, and may share common

376 transcription factors (Pearson et al 2016). The ecological role of these compounds is currently

377 unknown, however, curacin is a potent anticancer agent while jamaicamide exhibits sodium

378 channel blocking activity (Pearson et al 2016). The taxonomy of these PKS OTUs was closest

379 to Symploca sp. HPC-3, a filamentous cyanobacteria isolated from the hyper-saline stromatolites

380 of Shark Bay (Burns et al 2005). From this isolate, the authors putatively identified the

381 production of cyanopeptolin S, a compound with protease inhibitory activity (Jakobi et al

382 1995). Meanwhile, 96.8% of NRPS sequences were assigned to the syringomycin product

383 pathway (Supplementary Table A1.7 – A1.9, full data not shown). Syringomycin E was

384 discovered from Pseudomonas syringae, and has been shown to act as a surfactant against plant

385 cell membranes leading to the leaking of ions (Hutchison et al 1995). Whether the products of

386 the genes assigned to syringomycin here have similar roles requires further elucidation,

387 however, such nutrient harvesting actions within oligotrophic soils would provide competitive

388 advantage. Nutrient scavenging and uptake is a commonly proposed ecological function of

389 cyanotoxins (Holland and Kinnear 2013).

CHAPTER 5:

CONCLUSIONS AND FUTURE DIRECTIONS

1 RESEARCH MOTIVATIONS AND OBJECTIVES 2 The democratisation of genomics has fuelled an explosion in the exploration of the microbial

3 world. Taxonomic diversity is proving to be vast beyond expectations, with extrapolations

4 from genomes indicating functional diversity to be even greater. In the face of multi-drug

5 resistance, this promising frontier is galvanising the pursuit of microbes for novel anti-

6 infectives and other socially beneficial compounds. Given the immense scope of the microbial

7 world, there is an imperative to identify and prioritise promising avenues for future work.

8 Arid soils have been identified as diversity hotspots for the megasynthase genes responsible for

9 producing a large array of microbial bioactive compounds (Charlop-Powers et al 2014) .

10 Biocrusts are top-soil microbial niches within arid lands that present as compelling candidates

11 for bioprospecting given their extremotolerance, enrichment with cyanobacteria, and use as

12 model systems. This thesis takes direction from prior research to specifically target arid

13 Australian biocrusts and profile their genetic capacity for small molecule production. It seeks

14 to identify underlying patterns in taxonomic and functional diversity to further narrow the

15 search field and contribute to a biosynthesis atlas of Australia.

17 KEY FINDINGS

18 BIOCRUSTS ARE DYNAMIC AT LOCAL AND INTRA-CONTINENTAL SCALES

19 Forces determining microbial community assembly operate on different spatial scales (Belnap

20 et al 2016b). This thesis has identified that biocrust developmental stage and seasonality of

21 precipitation shape biocrust microbiomes on local and intra-continental scales, respectively. In

22 Chapter 2, biocrust morphology, as a proxy for developmental stage, was shown to reflect

23 distinct bacterial communities. Previously, biocrusts of different morphological properties have

24 been shown to have different ecological effects, such as rates of carbon and nitrogen fixation

25 and hydrological influences (Felde et al 2014, Yeager et al 2004). The distinction between

26 biocrust stages detected here indicates these functional differences are due in part to

27 differences in bacterial composition (Couradeau et al 2016). This is a key finding for

28 establishing guidelines for implementing biocrusts as indicators of arid health, as model

29 systems, and as functional units (Bowker et al 2014). Further to the purpose of this thesis, it

30 indicated that visually distinct biocrusts within a localized sampling area should be treated as

31 different sample types for bioprospecting. A clear distinction was observed in community

32 structure between bare soil and the cyanobacteria-enriched crusted soil. This was driven by a

33 bloom in filamentous cyanobacteria, notably Phormidium and Nostocophycideae types, in the

34 transition from loose soil to a cohesive early stage crust. Communities became more

35 homogenous and diversity increased as developmental stage progressed.

36 At intra-continental scales, climatic forces have the greatest effect on biocrust assembly

37 (Belnap et al 2016b). Within Australia, Rogers 1972 showed the distribution of dryland lichens

38 was affected by amount and seasonality of rainfall. Chapter 3 showed the bacterial component

39 of biocrusts display similar climate driven ordinations. Specifically, seasonality of rainfall best

100

40 explained the sample groupings. Biocrust community profiles of northern Australia correlated

41 with high summer rainfall while those of southern Australia correlated with high winter

42 rainfall. The differential enrichment of some members over others likely stems from different

43 capacities to tolerate desiccation and to effectively generate storage molecules during short

44 wetting events experienced in northern regions of Australia. Of note, unclassified

45 cyanobacteria were characteristic of northern biocrusts and explained a large portion of the

46 variance between the geographical regions. This group may harbour novel extremotolerant

47 strategies pertinent for sustaining biocrust coverage in a warming climate (Garcia-Pichel et al

48 2013). Chapter 3 also considered the continent-wide 16S rDNA dataset from an evolutionary

49 perspective. The phylogenetic distances between sample sites was considerably less than those

50 observed for composition, indicating Australian biocrusts comprise a common cohort of

51 bacteria differentiated across sites based on the environmental filtering of precipitation

52 patterns . Chapter 3 delivers important findings showing different sites around Australia

53 comprise different biocrust profiles and identified seasonality of precipitation as a significant

54 force driving their assembly.

55 BIOSYNTHETIC CAPACITY IS LINKED TO SEASONALITY OF PRECIPITATION 56 ON AN INTRA-CONTINENTAL SCALE

57 Understanding the natural histories and ecologies of microorganisms can facilitate hypothesis

58 driven science. The insights gained from determining community assembly patterns in

59 Chapters 2 and 3 allowed for effective and relevant lines of inquiry to test the NRPS and PKS

60 data sets generated here. Recently, Lemetre et al. 2017 made compelling progress developing a

61 biosynthetic map of Australian arid topsoils. Latitude was identified as an organising factor for

62 samples, however, despite including mean annual measures of temperature and precipitation,

63 the authors were unable to identify biological forces driving sample patterns. In this thesis, 101

64 Chapter 4 showed that seasonality of precipitation was a significant factor explaining the

65 variance between the biosynthetic profiles of biocrusts. The rainfall distribution gradient in

66 Australia operates on a latitudinal basis where northern regions have higher summer rainfall

67 and southern regions have higher winter rainfall. While it is probable these factors co-vary

68 according to other variables not measured here, it remains likely that seasonality of

69 precipitation is an important biological factor informing on the observations made by Lemetre

70 et al. 2017. Given this strong influence, it can be postulated that the products of the genes

71 surveyed may have critical roles in adaption to precipitation patterns.

72 Examination of NRPS and PKS variance at the local scale supports these notions. Given their

73 prolificacy as producers of an array of bioactive compounds, it was expected the dry bloom in

74 cyanobacteria eliciting a shift in taxonomic structure would be reflected in the biosynthetic

75 profiles. However, no significant differences in composition or diversity were observed

76 between the Cobar stages. Based on indications from the intra-continental dataset, this may be

77 due to the samples experiencing the same climactic forces imposing uniform selective

78 pressures on the conservation of NRPS and PKS genes. An improvement to this study would

79 be to examine the stages at the transcriptomic level to determine how the expression of these

80 genes may change across biocrust formation and development. Within the buffering nature of

81 the organic matrix, there may be less expression of genes required for desiccation tolerance in

82 later biocrust stages.

83 SECONDARY METABOLITE POTENTIAL OF BIOCRUSTS

84 Arid biocrusts may be considered as extreme environments comprised of extremotolerant

85 organisms. Strategies for survival include the production of EPS to stabilise soil particles and

86 increase water retention, the production of UV absorbing pigments, and desiccation tolerance 102

87 (Chapter 1). Such strategies rely on the production of secondary metabolites to alter the local

88 environment and protect essential cellular machinery. In 2017, Lemetre et al showed arid soils

89 had the greatest NRPS and PKS richness compared to soils from mesic environments. Argued

90 here is the notion that the higher richness is a product of the multiple molecular adaptations

91 required to colonise a multi-stress environment. Specifically, based on findings regarding

92 seasonality of precipitation from this thesis, the control over responses to water availability

93 appear to be defining pathways involving small molecule regulation. Yet, the biochemistry of

94 biocrust molecular adaptations remain largely unknown, especially for under-explored niches

95 and undocumented bacteria (Reddy et al 2007). Building upon Lemetre et al (2017), this thesis

96 has shown arid biocrusts are rich sources of potentially novel NRPS and PKS genes. It has

97 highlighted northern Australian biocrusts as particularly rich, specifically Charter’s Towers in

98 Queensland. Furthermore, candidate taxa with a high abundance of NRPS and PKs genes as

99 well as high centrality were identified, including Symploca, Anabaena cylindrical and Pseudomonas.

100 These findings provide directions for future bioprospecting endeavours, noting to include rarer

101 species beyond dominant Phormidium types.

102 FUTURE DIRECTIONS 103 Natural product discovery is at the interface of multiple fields including genomics, chemistry,

104 proteomics, metabolomics, and microbiology. The next steps from the findings presented here

105 can branch out along all of these lines of enquiry. A simple first action would be to test current

106 NRPS and PKS gene datasets using seasonality of precipitation as a factor explaining

107 continental wide distributions. This would also benefit from conducting new sampling along

108 continental latitudinal and longitudinal transects to improve calculations to delimit biosynthetic

109 gene regions of soils. Additionally, rather than targeted sequencing, shot-gun metagenomic

110 sequencing may better capture the co-occurrence of secondary metabolite genes with other 103

111 environmental genes and taxonomy (Steven et al 2012). Given the likelihood that the high

112 selective pressures of arid environments lead to the maintenance of cryptic NRPS and PKS

113 genes within genomes and communities, metatranscriptomics may help resolve the difference

114 between different biocrusts (Angel and Conrad 2013). There may be fundamental differences

115 in the expression of NRPS and PKS genes as cyanobacteria establish and form the initial

116 organic matrix, as well as the interactions between early-stage communities and successional

117 mosses and lichens in later stages. At an organism level, genome mining of biocrust isolates

118 would provide insights as to the organisation of gene clusters as allow predictions of

119 compound structure (Micallef et al 2015).

120 Beyond genomic mining approaches, small molecule extractions from whole biocrust

121 communities as well as bacterial isolates allows for the detection and characterisation of

122 expressed compounds. Both crude and purified extractions can be used for bioassays against

123 pathogens(Alvin et al 2016), as well as employed in microcosm experiments to help determine

124 their ecological roles (Baran et al 2015). For example, biocrust cyanobacteria have been shown

125 to exert an effect on native Australian seeds, likely due to the production of secondary

126 metabolites (Muñoz-Rojas et al 2018). Meanwhile, new long-amplicon PCR cloning methods

127 have increased the capacity of capturing biosynthetic gene clusters for heterologous expression

128 methods (Greunke et al 2018), broadening the scope of the characterisation of large gene

129 clusters. Further downstream, gene editing technologies can tailor NRPS and PKS catalytic

130 domains to alter substrate incorporation, thus modifying molecular structures for optimal

131 bioactivity (Williams 2013). Ultimately, the aim of gene surveys such as conducted here, is to

132 identify niches and organisms with gene clusters of interest to feed into characterisation and

133 expression pipelines for high titre recovery for medical and industrial applications (Liu et al

104

134 2017a). This thesis has developed findings which contribute to the biosynthesis map of

135 Australia. It has highlighted biocrusts as a diverse source of potentially novel and fascinating

136 secondary metabolites, as well as provided guidance to extract greater understanding from

137 similar datasets and provided informed strategies targeting new environments and organisms.

138

105

139 REFERENCES

140 Abed RM, Al Kharusi S, Schramm A, Robinson MD (2010). Bacterial diversity, 141 pigments and nitrogen fixation of biological desert crusts from the Sultanate of Oman. 142 FEMS Microbiol Ecol 72: 418-428.

143

144 Abed RMM, Ramette A, Hübner V, De Deckker P, de Beer D (2012). Microbial 145 diversity of eolian dust sources from saline lake sediments and biological soil crusts in 146 arid Southern Australia. FEMS Microbiology Ecology 80: 294-304.

147

148 Aboal M, Werner O, García-Fernández ME, Palazón JA, Cristóbal JC, Williams W 149 (2016). Should ecomorphs be conserved? The case of Nostoc flagelliforme, an 150 endangered extremophile cyanobacteria. Journal for Nature Conservation 30: 52-64.

151

152 Albuquerque L, da Costa MS (2014). The Family Rubrobacteraceae. In: Rosenberg E, 153 DeLong EF, Lory S, Stackebrandt E, Thompson F (eds). The Prokaryotes: 154 Actinobacteria. Springer Berlin Heidelberg: Berlin, Heidelberg. pp 861-866.

155

156 Allison SD, Martiny JBH (2008). Resistance, resilience, and redundancy in microbial 157 communities. Proceedings of the National Academy of Sciences of the United States of 158 America 105: 11512-11519.

159

160 Alvin A, Kalaitzis JA, Sasia B, Neilan BA (2016). Combined genetic and bioactivity‐ 161 based prioritization leads to the isolation of an endophyte‐derived antimycobacterial 162 compound. Journal of Applied Microbiology 120: 1229-1239.

163 106

164 Anderson M, Gorley R, Clarke K (2008). PERMANOVA+ for PRIMER: guide to 165 software and statistical methods. 2008. Plymouth, UK.

166

167 Angel R, Conrad R (2013). Elucidating the microbial resuscitation cascade in 168 biological soil crusts following a simulated rain event. Environmental microbiology 15: 169 2799-2815.

170

171 Arakawa K (2012). Biosynthesis: Diversity between PKS and FAS. Nat Chem Biol 8: 172 604-605.

173

174 Ashelford KE, Chuzhanova NA, Fry JC, Jones AJ, Weightman AJ (2005). At least 1 in 20 175 16S rRNA sequence records currently held in public repositories is estimated to 176 contain substantial anomalies. Applied and Environmental Microbiology 71: 7724-7736.

177

178 Assenov Y, Ramirez F, Schelhorn SE, Lengauer T, Albrecht M (2008). Computing 179 topological parameters of biological networks. Bioinformatics 24: 282-284.

180

181 Australian Government Bureau of Meteorology (2013). Climate Data Online.

182

183 Bahl J, Lau MCY, Smith GJD, Vijaykrishna D, Cary SC, Lacap DC et al (2011). Ancient 184 origins determine global biogeography of hot and cold desert cyanobacteria. Nature 185 Communications 2.

186

107

187 Balskus EP, Walsh CT (2010). The genetic and molecular basis for sunscreen 188 biosynthesis in cyanobacteria. Science (New York, NY) 329: 1653-1656.

189

190 Baran R, Brodie EL, Mayberry-Lewis J, Hummel E, Da Rocha UN, Chakraborty R et 191 al (2015). Exometabolite niche partitioning among sympatric soil bacteria. Nat 192 Commun 6.

193

194 Barger NN, Weber B, Garcia-Pichel F, Zaady E, Belnap J (2016). Patterns and controls 195 on nitrogen cycling of biological soil crusts. In: Weber B, Budel B, Belanp J (eds). 196 Biological soil crusts: An organizing principle in drylands, 226 edn. Springer Nature. 197 pp 257-285.

198

199 Bates ST, Nash TH, Garcia-Pichel F (2012). Patterns of diversity for fungal 200 assemblages of biological soil crusts from the southwestern United States. Mycologia 201 104: 353-361.

202

203 Battistuzzi FU, Feijao A, Hedges SB (2004). A genomic timescale of prokaryote 204 evolution: insights into the origin of methanogenesis, phototrophy, and the 205 colonization of land. BMC Evolutionary Biology 4: 1-14.

206

207 Battistuzzi FU, Hedges SB (2009). A major clade of prokaryotes with ancient 208 adaptations to life on land. Molecular biology and evolution 26: 335-343.

209

210 Bebout BM, Garcia-Pichel F (1995). UV B-induced vertical migrations of cyanobacteria 211 in a microbial mat. Applied and Environmental Microbiology 61: 4215-4222.

108

212

213 Belnap J, Gillette DA (1998). Vulnerability of desert biological soil crusts to wind 214 erosion: the influences of crust development, soil texture, and disturbance. Journal of 215 Arid Environments 39: 133-142.

216

217 Belnap J (2001). Comparative structure of physical and biological soil crusts. In: Belnap 218 J, Lange OL (eds). Biological soil crusts: structure, function, and management. 219 Springer-Verlag: Berlin Heidelberg.

220

221 Belnap J, Kaltenecker JH, Rosentreter r, Williams J, Leonard S, Eldridge D (2001). 222 Biological soil crust: ecology and management. In: Peterson P (ed). Biological soil 223 crusts: Ecology and management. U.S. Department of the interior.

224

225 Belnap J, Lange OL (2001). Biological soil crusts : structure, function, and 226 management. Berlin

227 London : Springer: Berlin

228 London.

229

230 Belnap J (2002). Nitrogen fixation in biological soil crusts from southeast Utah, USA. 231 Biology and Fertility of Soils 35: 128-135.

232

233 Belnap J, Büdel B, Lange OL (2003). Biological Soil Crusts: Characteristics and 234 Distribution. In: Belnap J, Lange OL (eds). Biological Soil Crusts: Structure, Function, 235 and Management. Springer Berlin Heidelberg: Berlin, Heidelberg. pp 3-30. 109

236

237 Belnap J, Phillips SL, Flint S, Money J, Caldwell M (2008a). Global change and 238 biological soil crusts: Effects of ultraviolet augmentation under altered precipitation 239 regimes and nitrogen additions. Global Change Biology 14: 670-686.

240

241 Belnap J, Phillips SL, Witwicki DL, Miller ME (2008b). Visually assessing the level of 242 development and soil surface stability of cyanobacterially dominated biological soil 243 crusts. Journal of Arid Environments 72: 1257-1264.

244

245 Belnap J, Budel B, Weber B, SpringerLink (2016a). Biological Soil Crusts : An 246 Organizing Principle in Drylands. Cham : Springer International Publishing : Imprint: 247 Springer.

248

249 Belnap J, Lange OL, Bowker MA, Buedel B, Sannier C, Pietrasiak N et al (2016b). 250 Ecological studies: 173-197.

251

252 Benjamini Y, Hochberg Y (1995). Controlling the False Discovery Rate: A Practical and 253 Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society Series 254 B (Methodological) 57: 289-300.

255

256 Benson DA, Cavanaugh M, Clark K, Karsch-Mizrachi I, Lipman DJ, Ostell J et al 257 (2013). GenBank. Nucleic Acids Res 41: D36-42.

258

110

259 Beraldi-Campesi H, Garcia-Pichel F (2011). The biogenicity of modern terrestrial roll- 260 up structures and its significance for ancient life on land. Geobiology 9: 10-23.

261

262 Berg G, Smalla K (2009). Plant species and soil type cooperatively shape the structure 263 and function of microbial communities in the rhizosphere. FEMS Microbiology 264 Ecology 68: 1-13.

265

266 Billi D, Friedmann EI, Hofer KG, Caiola MG, Ocampo-Friedmann R (2000). Ionizing- 267 Radiation Resistance in the Desiccation-Tolerant Cyanobacterium Chroococcidiopsis. 268 Applied and Environmental Microbiology 66: 1489-1492.

269

270 Billi D, Potts M (2002). Life and death of dried prokaryotes. Research in Microbiology 271 153: 7-12.

272

273 Billi D (2012). Anhydrobiotic rock-inhabiting cyanobacteria: Potential for astrobiology 274 and biotechnology. In: Stan-Lotter H, Fendrihan S (eds). Adaption of Microbial Life to 275 Environmental Extremes: Novel Research Results and Application. Springer Vienna: 276 Vienna. pp 119-132.

277

278 Bissett A, Brown MV, Siciliano SD, Thrall PH (2013). Microbial community responses 279 to anthropogenically induced environmental change: Towards a systems approach. 280 Ecology Letters 16: 128-139.

281

282 Bissett A, Fitzgerald A, Meintjes T, Mele PM, Reith F, Dennis PG et al (2016). 283 Introducing BASE: the Biomes of Australian Soil Environments soil microbial diversity 284 database. GigaScience 5: 1-11. 111

285

286 Boruah HP, Kumar BS (2002). Biological activity of secondary metabolites produced 287 by a strain of Pseudomonas fluorescens. Folia microbiologica 47: 359-363.

288

289 Bowker MA (2007). Biological soil crust rehabilitation in theory and practice: An 290 underexploited opportunity. Restoration Ecology 15: 13-23.

291

292 Bowker MA, Eldridge DJ, Val J, Soliveres S (2013a). Hydrology in a patterned 293 landscape is co-engineered by soil-disturbing animals and biological crusts. Soil 294 Biology and Biochemistry 61: 14-22.

295

296 Bowker MA, Maestre FT, Mau RL (2013b). Diversity and Patch-Size Distributions of 297 Biological Soil Crusts Regulate Dryland Ecosystem Multifunctionality. Ecosystems 16: 298 923-933.

299

300 Bowker MA, Maestre FT, Eldridge D, Belnap J, Castillo-Monroy A, Escolar C et al 301 (2014). Biological soil crusts (biocrusts) as a model system in community, landscape 302 and ecosystem ecology. Biodiversity and Conservation 23: 1619-1637.

303

304 Bowker MA, Belnap J, Büdel B, Sannier C, Pietrasiak N, Eldridge DJ et al (2016). 305 Controls on Distribution Patterns of Biological Soil Crusts at Micro- to Global Scales. 306 In: Weber B, Büdel B, Belnap J (eds). Biological Soil Crusts: An Organizing Principle 307 in Drylands. Springer International Publishing: Cham. pp 173-197.

308

112

309 Budel B, Darienko T, Deutschewitz K, Dojani S, Friedl T, Mohr KI et al (2009). 310 Southern African biological soil crusts are ubiquitous and highly diverse in drylands, 311 being restricted by rainfall frequency. Microb Ecol 57: 229-247.

312

313 Büdel B (2003). Synopsis: Comparative Biogeography of Soil-Crust Biota. In: Belnap J, 314 Lange OL (eds). Biological Soil Crusts: Structure, Function, and Management. 315 Springer Berlin Heidelberg: Berlin, Heidelberg. pp 141-152.

316

317 Büdel B, Colesie C, Green TGA, Grube M, Lázaro Suau R, Loewen-Schneider K et al 318 (2014). Improved appreciation of the functioning and importance of biological soil 319 crusts in Europe: the Soil Crust International Project (SCIN). Biodiversity and 320 Conservation 23: 1639-1658.

321

322 Büdel B, Dulić T, Darienko T, Rybalka N, Friedl T (2016). Cyanobacteria and Algae of 323 Biological Soil Crusts. In: Weber B, Büdel B, Belnap J (eds). Biological Soil Crusts: An 324 Organizing Principle in Drylands. Springer International Publishing: Cham. pp 55-80.

325

326 Burns BP, Seifert A, Goh F, Pomati F, Jungblut AD, Serhat A et al (2005). Genetic 327 potential for secondary metabolite production in stromatolite communities. FEMS 328 Microbiology Letters 243: 293-301.

329

330 Cai L, Ye L, Tong AHY, Lok S, Zhang T (2013). Biased Diversity Metrics Revealed by 331 Bacterial 16S Pyrotags Derived from Different Primer Sets. PLoS ONE 8.

332

113

333 Calteau A, Fewer DP, Latifi A, Coursin T, Laurent T, Jokela J et al (2014). Phylum- 334 wide comparative genomics unravel the diversity of secondary metabolism in 335 Cyanobacteria. BMC Genomics 15: 1-14.

336

337 Castenholz RW (2015). General Characteristics of the Cyanobacteria. Bergey's Manual 338 of Systematics of Archaea and Bacteria. John Wiley & Sons, Ltd.

339

340 Castillo-Monroy AP, Bowker MA, Maestre FT, Rodríguez-Echeverría S, Martinez I, 341 Barraza-Zepeda CE et al (2011). Relationships between biological soil crusts, bacterial 342 diversity and abundance, and ecosystem functioning: Insights from a semi-arid 343 Mediterranean environment. Journal of Vegetation Science 22: 165-174.

344

345 Charlesworth JC, Burns BP (2015). Untapped Resources: Biotechnological Potential of 346 Peptides and Secondary Metabolites in Archaea. Archaea 2015.

347

348 Charlop-Powers Z, Owen JG, Reddy BVB, Ternei MA, Brady SF (2014). Chemical- 349 biogeographic survey of secondary metabolism in soil. Proceedings of the National 350 Academy of Sciences of the United States of America 111: 3757-3762.

351

352 Charlop-Powers Z, Owen JG, Reddy BVB, Ternei M, Guimaraes DO, De Frias UA et 353 al (2015). Global biogeographic sampling of bacterial secondary metabolism. eLife 354 2015.

355

114

356 Charlop-Powers Z, Pregitzer CC, Lemetre C, Ternei MA, Maniko J, Hover BM et al 357 (2016). Urban park soil microbiomes are a rich reservoir of natural product biosynthetic 358 diversity. Proceedings of the National Academy of Sciences of the United States of 359 America 113: 14811-14816.

360

361 Chien A, Edgar DB, Trela JM (1976). Deoxyribonucleic acid polymerase from the 362 extreme thermophile Thermus aquaticus. Journal of Bacteriology 127: 1550-1557.

363

364 Chu J, Vila-Farres X, Inoyama D, Ternei M, Cohen LJ, Gordon EA et al (2016). 365 Discovery of MRSA active antibiotics using primary sequence from the human 366 microbiome. Nature Chemical Biology 12: 1004-1006.

367

368 Claesson MJ, Wang Q, O'Sullivan O, Greene-Diniz R, Cole JR, Ross RP et al (2010). 369 Comparison of two next-generation sequencing technologies for resolving highly 370 complex microbiota composition using tandem variable 16S rRNA gene regions. 371 Nucleic Acids Res 38: e200.

372

373 Coe KK, Belnap J, Sparks JP (2012). Precipitation-driven carbon balance controls 374 survivorship of desert biocrust mosses. Ecology 93: 1626-1636.

375

376 Cole JR, Wang Q, Fish JA, Chai B, McGarrell DM, Sun Y et al (2014). Ribosomal 377 Database Project: Data and tools for high throughput rRNA analysis. Nucleic Acids 378 Research 42: D633-D642.

379

115

380 Concostrina-Zubiri L, Pescador DS, Martínez I, Escudero A (2014). Climate and small 381 scale factors determine functional diversity shifts of biological soil crusts in Iberian 382 drylands. Biodiversity and Conservation 23: 1757-1770.

383

384 Couradeau E, Karaoz U, Lim HC, Nunes da Rocha U, Northen T, Brodie E et al 385 (2016). Bacteria increase arid-land soil surface temperature through the production of 386 sunscreens. Nat Commun 7.

387

388 Cragg GM, Newman DJ (2013). Natural products: a continuing source of novel drug 389 leads. Biochimica et biophysica acta 1830: 3670-3695.

390

391 Crits-Christoph A, Robinson CK, Ma B, Ravel J, Wierzchos J, Ascaso C et al (2016). 392 Phylogenetic and Functional Substrate Specificity for Endolithic Microbial 393 Communities in Hyper-Arid Environments. Frontiers in Microbiology 7: 301.

394

395 Davies J, Ryan KS (2012). Introducing the parvome: bioactive compounds in the 396 microbial world. ACS chemical biology 7: 252-259.

397

398 De Cáceres M, Legendre P (2009). Associations between species and groups of sites: 399 Indices and statistical inference. Ecology 90: 3566-3574.

400

401 Delgado-Baquerizo M, Maestre FT, Eldridge DJ, Singh BK (2016). Microsite 402 Differentiation Drives the Abundance of Soil Ammonia Oxidizing Bacteria along 403 Aridity Gradients. Frontiers in Microbiology 7.

116

404

405 DeSantis TZ, Hugenholtz P, Larsen N, Rojas M, Brodie EL, Keller K et al (2006). 406 Greengenes, a chimera-checked 16S rRNA gene database and workbench compatible 407 with ARB. Applied and Environmental Microbiology 72: 5069-5072.

408

409 Deveau A, Gross H, Palin B, Mehnaz S, Schnepf M, Leblond P et al (2016). Role of 410 secondary metabolites in the interaction between Pseudomonas fluorescens and soil 411 microorganisms under iron-limited conditions. FEMS Microbiology Ecology 92: 412 fiw107-fiw107.

413

414 Ding D, Chen G, Wang B, Wang Q, Liu D, Peng M et al (2013). Culturable 415 actinomycetes from desert ecosystem in northeast of Qinghai-Tibet Plateau. Annals of 416 Microbiology 63: 259-266.

417

418 Dini-Andreote F, De Cássia Pereira E Silva M, Triadó-Margarit X, Casamayor EO, 419 Van Elsas JD, Salles JF (2014). Dynamics of bacterial community succession in a salt 420 marsh chronosequence: Evidences for temporal niche partitioning. ISME Journal 8: 421 1989-2001.

422

423 Dojani S, Kauff F, Weber B, Budel B (2014). Genotypic and phenotypic diversity of 424 cyanobacteria in biological soil crusts of the Succulent Karoo and Nama Karoo of 425 southern Africa. Microb Ecol 67: 286-301.

426

427 Dowd SE, Zaragoza J, Rodriguez JR, Oliver MJ, Payton PR (2005). Windows .NET 428 Network Distributed Basic Local Alignment Search Toolkit (W.ND-BLAST). BMC 429 Bioinformatics 6: 1-14.

117

430

431 Dunbar J, Ticknor LO, Kuske CR (2000). Assessment of Microbial Diversity in Four 432 Southwestern United States Soils by 16S rRNA Gene Terminal Restriction Fragment 433 Analysis. Applied and Environmental Microbiology 66: 2943-2950.

434

435 Edgar RC (2004). MUSCLE: A multiple sequence alignment method with reduced 436 time and space complexity. BMC Bioinformatics 5.

437

438 Edgar RC, Haas BJ, Clemente JC, Quince C, Knight R (2011). UCHIME improves 439 sensitivity and speed of chimera detection. Bioinformatics 27: 2194-2200.

440

441 Elbert W, Weber B, Burrows S, Steinkamp J, Budel B, Andreae MO et al (2012). 442 Contribution of cryptogamic covers to the global cycles of carbon and nitrogen. Nature 443 Geosci 5: 459-462.

444

445 Eldridge D, Greene R (1994). Microbiotic soil crusts - a review of their roles in soil and 446 ecological processes in the rangelands of Australia. Soil Research 32: 389-415.

447

448 Eldridge D, Tozer M, Lepp H (1997). A practical guide to soil lichens and bryophytes 449 of Australia's dry country NSW Department of land and water conservation: Sydney.

450

451 Eldridge DJ, Rosentreter R (1999). Morphological groups: A framework for monitoring 452 microphytic crusts in arid landscapes. Journal of Arid Environments 41: 11-25.

118

453

454 Eldridge DJ (2001). Biological soil crusts of Australia. In: Belnap J, Lange OL (eds). 455 Biological soil crusts: structure, function, and management. Springer-Verlag: Berlin 456 Heidelberg.

457

458 Eldridge DJ, Woodhouse JN, Curlevski NJ, Hayward M, Brown MV, Neilan BA (2015). 459 Soil-foraging animals alter the composition and co-occurrence of microbial 460 communities in a desert shrubland. ISME J 9: 2671-2681.

461

462 Elliott DR, Thomas AD, Hoon SR, Sen R (2014). Niche partitioning of bacterial 463 communities in biological crusts and soils under grasses, shrubs and trees in the 464 Kalahari. Biodiversity and Conservation 23: 1709-1733.

465

466 Escolar C, Maestre FT, Rey A (2015). Biocrusts modulate warming and rainfall 467 exclusion effects on soil respiration in a semi-arid grassland. Soil biology & 468 biochemistry 80: 9-17.

469

470 Felde VJMNL, Peth S, Uteau-Puschmann D, Drahorad S, Felix-Henningsen P (2014). 471 Soil microstructure as an under-explored feature of biological soil crust hydrological 472 properties: Case study from the NW Negev Desert. Biodiversity and Conservation 23: 473 1687-1708.

474

475 Felnagle EA, Jackson EE, Chan YA, Podevels AM, Berti AD, McMahon MD et al 476 (2008). Nonribosomal peptide synthetases involved in the production of medically 477 relevant natural products. Molecular pharmaceutics 5: 191-211.

478 119

479 Ferrenberg S, Reed SC, Belnap J, Schlesinger WH (2015). Climate change and physical 480 disturbance cause similar community shifts in biological soil crusts. Proceedings of the 481 National Academy of Sciences of the United States of America 112: 12116-12121.

482

483 Ferrenberg S, Reed SC (2017). Biocrust ecology: unifying micro- and macro-scales to 484 confront global change. New Phytologist 216: 643-646.

485

486 Fischbach MA, Walsh CT (2006). Assembly-line enzymology for polyketide and 487 nonribosomal Peptide antibiotics: logic, machinery, and mechanisms. Chem Rev 106: 488 3468-3496.

489

490 Friedman J, Alm EJ (2012). Inferring Correlation Networks from Genomic Survey Data. 491 PLoS Computational Biology 8.

492

493 Fuhrman JA, Steele JA (2008). Community structure of marine bacterioplankton: 494 Patterns, networks, and relationships to function. Aquatic Microbial Ecology 53: 69-81.

495

496 Fuhrman JA (2009). Microbial community structure and its functional implications. 497 Nature 459: 193-199.

498

499 Gao Q, Garcia-Pichel F (2011). Microbial ultraviolet sunscreens. Nat Rev Micro 9: 791- 500 802.

501

120

502 Garcia-Pichel F, Prufert-Bebout L, Muyzer G (1996). Phenotypic and phylogenetic 503 analyses show Microcoleus chthonoplastes to be a cosmopolitan cyanobacterium. 504 Applied and Environmental Microbiology 62: 3284-3291.

505

506 Garcia-Pichel F, Belnap J (2001). Small-scale environments and distribution of 507 biological soil crusts. In: Belnap J, Lange OL (eds). Biological soil crusts: structure, 508 function, and management. Springer-Verlag: Berlin Heidelberg.

509

510 Garcia-Pichel F, López-Cortés A, Nübel U (2001). Phylogenetic and Morphological 511 Diversity of Cyanobacteria in Soil Desert Crusts from the Colorado Plateau. Applied 512 and Environmental Microbiology 67: 1902-1910.

513

514 Garcia-Pichel F, Johnson LS, Youngkin D, Belnap J (2003). Small-Scale Vertical 515 Distribution of Bacterial Biomass and Diversity in Biological Soil Crusts from Arid 516 Lands in the Colorado Plateau. Microbial Ecology 46: 312-321.

517

518 Garcia-Pichel F (2008). Molecular ecology and environmental genomics of 519 cyanobacteria. Caister Academnic Press: Norfolk, UK.

520

521 Garcia-Pichel F, Wojciechowski MF (2009). The evolution of a capacity to build supra- 522 cellular ropes enabled filamentous cyanobacteria to colonize highly erodible 523 substrates. PLoS ONE 4.

524

525 Garcia-Pichel F, Loza V, Marusenko Y, Mateo P, Potrafka RM (2013). Temperature 526 Drives the Continental-Scale Distribution of Key Microbes in Topsoil Communities. 527 Science 340: 1574-1577. 121

528

529 Gehringer MM, Adler L, Roberts AA, Moffitt MC, Mihali TK, Mills TJ et al (2012). 530 Nodularin, a cyanobacterial toxin, is synthesized in planta by symbiotic Nostoc sp. 531 ISME J 6: 1834-1847.

532

533 Gomez-Escribano JP, Castro JF, Razmilic V, Chandra G, Andrews B, Asenjo JA et al 534 (2015). The Streptomyces leeuwenhoekii genome: de novo sequencing and assembly in 535 single contigs of the chromosome, circular plasmid pSLE1 and linear plasmid pSLE2. 536 BMC Genomics 16: 1-11.

537

538 Gotelli NJ, Entsminger GL (2015). EcoSim: Null models software for ecology. Zenodo.

539

540 Greunke C, Duell ER, D’Agostino PM, Glöckle A, Lamm K, Gulder TAM (2018). 541 Direct Pathway Cloning (DiPaC) to Unlock Natural Product Biosynthetic Potential. 542 Metabolic Engineering.

543

544 Gundlapally SR, Garcia-Pichel F (2006). The community and phylogenetic diversity of 545 biological soil crusts in the Colorado Plateau studied by molecular fingerprinting and 546 intensive cultivation. Microbial Ecology 52: 345-357.

547

548 Hagemann M, Henneberg M, Felde VJ, Drahorad SL, Berkowicz SM, Felix- 549 Henningsen P et al (2015). Cyanobacterial Diversity in Biological Soil Crusts along a 550 Precipitation Gradient, Northwest Negev Desert, Israel. Microb Ecol 70: 219-230.

551

122

552 Harvey AL, Edrada-Ebel R, Quinn RJ (2015). The re-emergence of natural products for 553 drug discovery in the genomics era. Nat Rev Drug Discov 14: 111-129.

554

555 Hoffmann L, Komárek J, í, Ka, tovský J (2005). System of cyanoprokaryotes 556 (cyanobacteria) – state in 2004. Algological Studies 117: 95-115.

557

558 Holland A, Kinnear S (2013). Interpreting the Possible Ecological Role(s) of 559 Cyanotoxins: Compounds for Competitive Advantage and/or Physiological Aide? 560 Marine Drugs 11: 2239-2258.

561

562 Holmes AJ, Bowyer J, Holley MP, O'Donoghue M, Montgomery M, Gillings MR 563 (2000). Diverse, yet-to-be-cultured members of the Rubrobacter subdivision of the 564 Actinobacteria are widespread in Australian arid soils. FEMS Microbiology Ecology 565 33: 111-120.

566

567 Housman DC, Powers HH, Collins AD, Belnap J (2006). Carbon and nitrogen fixation 568 differ between successional stages of biological soil crusts in the Colorado Plateau and 569 Chihuahuan Desert. Journal of Arid Environments 66: 620-634.

570

571 Hutchison ML, Tester MA, Gross DC (1995). Role of biosurfactant and ion channel- 572 forming activities of syringomycin in transmembrane ion flux: A model for the 573 mechanism of action in the plant-pathogen interaction. Molecular Plant-Microbe 574 Interactions 8: 610-620.

575

123

576 Jakobi C, Oberer L, Quiquerez C, König WA, Weckesser J (1995). Cyanopeptolin S, a 577 sulfate-containing depsipeptide from a water bloom of Microcystis sp. FEMS 578 Microbiology Letters 129: 129-133.

579

580 Jenke-Kodama H, Sandmann A, Muller R, Dittmann E (2005). Evolutionary 581 implications of bacterial polyketide synthases. Molecular biology and evolution 22: 582 2027-2039.

583

584 Jensen PR (2010). Linking species concepts to natural product discovery in the post- 585 genomic era. Journal of Industrial Microbiology and Biotechnology 37: 219-224.

586

587 Jungblut AD, Neilan BA (2006). Molecular identification and evolution of the cyclic 588 peptide hepatotoxins, microcystin and nodularin, synthetase genes in three orders of 589 cyanobacteria. Archives of microbiology 185: 107-114.

590

591 Kalaitzis JA, Lauro FM, Neilan BA (2009). Mining cyanobacterial genomes for genes 592 encoding complex biosynthetic pathways. Natural Product Reports 26: 1447-1465.

593

594 Kidron GJ, Li XR, Jia RL, Gao YH, Zhang P (2015). Assessment of carbon gains from 595 biocrusts inhabiting a dunefield in the Negev Desert. Geoderma 253–254: 102-110.

596

597 : Guide to the nomenclature and formal taxonomic treatment of oxyphototroph 598 prokaryotes (Cyanoprokaryotes). International Association of Cyanophyte Research; 599 Luxembourg.

124

600

601 Kozich J, Westcott SL, Baxter NT, Highlander SK, Schloss PD (2013). Development of 602 a dual-index sequencing strategy and curation pipeline for analyzing amplicon 603 sequence data on the miseq illumina sequencing platform. Applied and Environmental 604 Microbiology 79: 5112-5120.

605

606 Kuske CR, Ticknor LO, Miller ME, Dunbar JM, Davis JA, Barns SM et al (2002). 607 Comparison of soil bacterial communities in rhizospheres of three plant species and 608 the interspaces in an arid grassland. Appl Environ Microbiol 68: 1854-1863.

609

610 Kuske CR, Yeager CM, Johnson S, Ticknor LO, Belnap J (2012). Response and 611 resilience of soil biocrust bacterial communities to chronic physical disturbance in arid 612 shrublands. ISME J 6: 886-897.

613

614 Lamb EG, Kennedy N, Siciliano SD (2011). Effects of plant species richness and 615 evenness on soil microbial community diversity and function. Plant and Soil 338: 483- 616 495.

617

618 Lange OL (2003). Photosynthesis of Soil-Crust Biota as Dependent on Environmental 619 Factors. In: Belnap J, Lange OL (eds). Biological Soil Crusts: Structure, Function, and 620 Management. Springer Berlin Heidelberg: Berlin, Heidelberg. pp 217-240.

621

622 Lemetre C, Maniko J, Charlop-Powers Z, Sparrow B, Lowe AJ, Brady SF (2017). 623 Bacterial natural product biosynthetic domain composition in soil correlates with 624 changes in latitude on a continent-wide scale. Proceedings of the National Academy of 625 Sciences of the United States of America 114: 11615-11620.

125

626

627 Letzel AC, Pidot SJ, Hertweck C (2013). A genomic approach to the cryptic secondary 628 metabolome of the anaerobic world. Nat Prod Rep 30: 392-428.

629

630 Leys JF, Eldridge DJ (1998). Influence of cryptogamic crust disturbance to wind 631 erosion on sand and loam rangeland soils. Earth Surface Processes and Landforms 23: 632 963-974.

633

634 Li H, Rao B, Wang G, Shen S, Li D, Hu C et al (2014). Spatial heterogeneity of 635 cyanobacteria-inoculated sand dunes significantly influences artificial biological soil 636 crusts in the Hopq Desert (China). Environmental Earth Sciences 71: 245-253.

637

638 Li K, Liu R, Zhang H, Yun J (2013). The diversity and abundance of bacteria and 639 oxygenic phototrophs in saline biological desert crusts in Xinjiang, northwest China. 640 Microb Ecol 66: 40-48.

641

642 Li XR, Zhang P, Su YG, Jia RL (2012). Carbon fixation by biological soil crusts 643 following revegetation of sand dunes in arid desert regions of China: A four-year field 644 study. CATENA 97: 119-126.

645

646 Ling LL, Schneider T, Peoples AJ, Spoering AL, Engels I, Conlon BP et al (2015). A 647 new antibiotic kills pathogens without detectable resistance. Nature 517: 455-459.

648

126

649 Liu R, Li K, Zhang H, Zhu J, Joshi D (2014). Spatial distribution of microbial 650 communities associated with dune landform in the Gurbantunggut Desert, China. 651 Journal of Microbiology 52: 898-907.

652

653 Liu T, Mazmouz R, Ongley SE, Chau R, Pickford R, Woodhouse JN et al (2017a). 654 Directing the Heterologous Production of Specific Cyanobacterial Toxin Variants. ACS 655 chemical biology 12: 2021-2029.

656

657 Liu YR, Delgado-Baquerizo M, Trivedi P, He JZ, Wang JT, Singh BK (2017b). Identity 658 of biocrust species and microbial communities drive the response of soil 659 multifunctionality to simulated global change. Soil Biology and Biochemistry 107: 208- 660 217.

661

662 Maier S, Schmidt TSB, Zheng L, Peer T, Wagner V, Grube M (2014). Analyses of 663 dryland biological soil crusts highlight lichens as an important regulator of microbial 664 communities. Biodiversity and Conservation 23: 1735-1755.

665

666 Maier S, Muggia L, Kuske CR, Grube M (2016). Bacteria and non-lichenized fungi 667 within biological soil crusts. In: Weber B, Budel B, Belnap J (eds). Biological soil 668 crusts: an organising principle in drylands. Springer International Publishing: Cham. 669 pp 81-100.

670

671 Maier S, Tamm A, Wu D, Caesar J, Grube M, Weber B (2018). Photoautotrophic 672 organisms control microbial abundance, diversity, and physiology in different types of 673 biological soil crusts. ISME Journal: 1-15.

674

127

675 Mandal S, Rath J (2014). Extremophilic Cyanobacteria For Novel Drug Development. 676 Springer International Publishing: Cham.

677

678 Martínez I, Escudero A, Maestre FT, De La Cruz A, Guerrero C, Rubio A (2006). 679 Small-scale patterns of abundance of mosses and lichens forming biological soil crusts 680 in two semi-arid gypsum environments. Australian Journal of Botany 54: 339-348.

681

682 McDonald D, Price MN, Goodrich J, Nawrocki EP, DeSantis TZ, Probst A et al (2012). 683 An improved Greengenes taxonomy with explicit ranks for ecological and evolutionary 684 analyses of bacteria and archaea. ISME J 6: 610-618.

685

686 McKenzie NJ (2004). Australian soils and landscapes: an illustrated compendium. 687 CSIRO: Collingwood, Australia.

688

689 Meier JL, Burkart MD (2009). The chemical biology of modular biosynthetic enzymes. 690 Chemical Society reviews 38: 2012-2045.

691

692 Micallef ML, D'Agostino PM, Al-Sinawi B, Neilan BA, Moffitt MC (2015). Exploring 693 cyanobacterial genomes for natural product biosynthesis pathways. Marine Genomics 694 21: 1-12.

695

696 Moffitt MC, Neilan BA (2003). Evolutionary affiliations within the superfamily of 697 ketosynthases reflect complex pathway associations. Journal of molecular evolution 56: 698 446-457.

128

699

700 Mohammadipanah F, Wink J (2015). Actinobacteria from Arid and Desert Habitats: 701 Diversity and Biological Activity. Frontiers in Microbiology 6: 1541.

702

703 Mootz HD, Schwarzer D, Marahiel MA (2002). Ways of assembling complex natural 704 products on modular nonribosomal peptide synthetases. ChemBioChem 3: 490-504.

705

706 Moss NA, Bertin MJ, Kleigrewe K, Leão TF, Gerwick L, Gerwick WH (2016). 707 Integrating mass spectrometry and genomics for cyanobacterial metabolite discovery. 708 Journal of Industrial Microbiology and Biotechnology 43: 313-324.

709

710 Mueller RC, Belnap J, Kuske CR (2015). Soil bacterial and fungal community responses 711 to nitrogen addition across soil depth and microhabitat in an arid shrubland. Front 712 Microbiol 6: 891.

713

714 Muñoz-Rojas M, Chilton A, Liyanage GS, Erickson TE, Merritt DJ, Neilan BA et al 715 (2018). Effects of indigenous soil cyanobacteria on seed germination and seedling 716 growth of arid species used in restoration. Plant and Soil 429: 91-100.

717

718 Nagy ML, Pérez A, Garcia-Pichel F (2005). The prokaryotic diversity of biological soil 719 crusts in the Sonoran Desert (Organ Pipe Cactus National Monument, AZ). FEMS 720 Microbiology Ecology 54: 233-245.

721

129

722 Navarro-González R, Rainey FA, Molina P, Bagaley DR, Hollen BJ, de la Rosa J et al 723 (2003). Mars-Like Soils in the Atacama Desert, Chile, and the Dry Limit of Microbial 724 Life. Science 302: 1018-1021.

725

726 Neher DA, Lewins SA, Weicht TR, Darby BJ (2009). Microarthropod communities 727 associated with biological soil crusts in the Colorado Plateau and Chihuahuan deserts. 728 Journal of Arid Environments 73: 672-677.

729

730 Neilan BA, Dittmann E, Rouhiainen L, Bass RA, Schaub V, Sivonen K et al (1999). 731 Nonribosomal Peptide Synthesis and Toxigenicity of Cyanobacteria. Journal of 732 Bacteriology 181: 4089-4097.

733

734 Nemergut DR, Schmidt SK, Fukami T, O'Neill SP, Bilinski TM, Stanish LF et al 735 (2013). Patterns and Processes of Microbial Community Assembly. Microbiology and 736 Molecular Biology Reviews : MMBR 77: 342-356.

737

738 Newman DJ, Cragg GM (2012). Natural products as sources of new drugs over the 30 739 years from 1981 to 2010. Journal of natural products 75: 311-335.

740

741 Newman DJ, Cragg GM, Kingston DGI (2015). Natural Products as Pharmaceuticals 742 and Sources for Lead Structures. The Practice of Medicinal Chemistry: Fourth Edition. 743 pp 101-139.

744

745 Newman DJ, Cragg GM (2016). Natural Products as Sources of New Drugs from 1981 746 to 2014. Journal of natural products 79: 629-661.

130

747

748 Nichols D, Cahoon N, Trakhtenberg EM, Pham L, Mehta A, Belanger A et al (2010). 749 Use of Ichip for High-Throughput In Situ Cultivation of “Uncultivable” Microbial 750 Species. Applied and Environmental Microbiology 76: 2445-2450.

751

752 Nielsen S, Needham B, Leach ST, Day AS, Jaffe A, Thomas T et al (2016). Disrupted 753 progression of the intestinal microbiota with age in children with cystic fibrosis. 754 Scientific Reports 6.

755

756 Nubel U, Garcia-Pichel F, Kuhl M, Muyzer G (1999). Quantifying microbial diversity: 757 morphotypes, 16S rRNA genes, and carotenoids of oxygenic phototrophs in microbial 758 mats. Appl Environ Microbiol 65: 422-430.

759

760 Nunes da Rocha U, Cadillo-Quiroz H, Karaoz U, Rajeev L, Klitgord N, Dunn S et al 761 (2015). Isolation of a significant fraction of non-phototroph diversity from a desert 762 Biological Soil Crust. Frontiers in Microbiology 6: 277.

763

764 O'Brien J, Wright GD (2011). An ecological perspective of microbial secondary 765 metabolism. Current opinion in biotechnology 22: 552-558.

766

767 Okoro CK, Brown R, Jones AL, Andrews BA, Asenjo JA, Goodfellow M et al (2009). 768 Diversity of culturable actinomycetes in hyper-arid soils of the Atacama Desert, Chile. 769 Antonie van Leeuwenhoek 95: 121-133.

770

131

771 Ongley SE, Bian X, Zhang Y, Chau R, Gerwick WH, Müller R et al (2013). High-Titer 772 Heterologous Production in E. coli of Lyngbyatoxin, a Protein Kinase C Activator from 773 an Uncultured Marine Cyanobacterium. ACS chemical biology 8: 1888-1893.

774

775 Oren A (2004). A proposal for further integration of the cyanobacteria under the 776 Bacteriological Code. International Journal of Systematic and Evolutionary 777 Microbiology 54: 1895-1902.

778

779 Oren N, Raanan H, Murik O, Keren N, Kaplan A (2017). Dawn illumination prepares 780 desert cyanobacteria for dehydration. Current Biology 27: R1056-R1057.

781

782 Pearson L, Mihali T, Moffitt M, Kellmann R, Neilan B (2010). On the chemistry, 783 toxicology and genetics of the cyanobacterial toxins, microcystin, nodularin, saxitoxin 784 and cylindrospermopsin. Marine Drugs 8: 1650-1680.

785

786 Pearson LA, Dittmann E, Mazmouz R, Ongley SE, D’Agostino PM, Neilan BA (2016). 787 The genetics, biosynthesis and regulation of toxic specialized metabolites of 788 cyanobacteria. Harmful Algae 54: 98-111.

789

790 Pepe-Ranney C, Koechli C, Potrafka R, Andam C, Eggleston E, Garcia-Pichel F et al 791 (2016). Non-cyanobacterial diazotrophs mediate dinitrogen fixation in biological soil 792 crusts during early crust formation. ISME Journal 10: 287-298.

793

794 Pietrasiak N, Regus JU, Johansen JR, Lam D, Sachs JL, Santiago LS (2013). Biological 795 soil crust community types differ in key ecological functions. Soil Biology and 796 Biochemistry 65: 168-171. 132

797

798 Pinevich AV (2015). Proposal to consistently apply the International Code of 799 Nomenclature of Prokaryotes (ICNP) to names of the oxygenic photosynthetic 800 bacteria (cyanobacteria), including those validly published under the International 801 Code of Botanical Nomenclature (ICBN)/International Code of Nomenclature for 802 algae, fungi and plants (ICN), and proposal to change Principle 2 of the ICNP. Int J 803 Syst Evol Microbiol 65: 1070-1074.

804

805 Pócs T (2009). Cyanobacterial crust types, as strategies for survival in extreme habitats. 806 Acta Botanica Hungarica 51: 147-178.

807

808 Pointing SB, Belnap J (2012). Microbial colonization and controls in dryland systems. 809 Nat Rev Micro 10: 551-562.

810

811 Pringault O, Garcia-Pichel F (2004). Hydrotaxis of cyanobacteria in desert crusts. 812 Microb Ecol 47: 366-373.

813

814 Pruesse E, Quast C, Knittel K, Fuchs BM, Ludwig W, Peplies J et al (2007). SILVA: A 815 comprehensive online resource for quality checked and aligned ribosomal RNA 816 sequence data compatible with ARB. Nucleic Acids Research 35: 7188-7196.

817

818 Quast C, Pruesse E, Yilmaz P, Gerken J, Schweer T, Yarza P et al (2013). The SILVA 819 ribosomal RNA gene database project: Improved data processing and web-based tools. 820 Nucleic Acids Research 41: D590-D596.

821

133

822 Rajeev L, da Rocha UN, Klitgord N, Luning EG, Fortney J, Axen SD et al (2013). 823 Dynamic cyanobacterial response to hydration and dehydration in a desert biological 824 soil crust. ISME J 7: 2178-2191.

825

826 Rastogi RP, Sonani RR, Madamwar D (2015). Cyanobacterial Sunscreen Scytonemin: 827 Role in Photoprotection and Biomedical Research. Applied Biochemistry and 828 Biotechnology 176: 1551-1563.

829

830 Rausch C, Hoof I, Weber T, Wohlleben W, Huson DH (2007). Phylogenetic analysis of 831 condensation domains in NRPS sheds light on their functional evolution. BMC 832 Evolutionary Biology 7: 1-15.

833

834 Reddy GS, Nagy M, Garcia-Pichel F (2006). Belnapia moabensis gen. nov., sp. nov., an 835 alphaproteobacterium from biological soil crusts in the Colorado Plateau, USA. Int J 836 Syst Evol Microbiol 56: 51-58.

837

838 Reddy GS, Garcia-Pichel F (2007). Sphingomonas mucosissima sp. nov. and 839 Sphingomonas desiccabilis sp. nov., from biological soil crusts in the Colorado 840 Plateau, USA. Int J Syst Evol Microbiol 57: 1028-1034.

841

842 Reddy GS, Potrafka RM, Garcia-Pichel F (2007). Modestobacter versicolor sp. nov., an 843 actinobacterium from biological soil crusts that produces melanins under oligotrophy, 844 with emended descriptions of the genus Modestobacter and Modestobacter 845 multiseptatus Mevs et al. 2000. Int J Syst Evol Microbiol 57: 2014-2020.

846

134

847 Reddy GS, Garcia-Pichel F (2015). Description of Pseudomonas asuensis sp. nov. from 848 biological soil crusts in the Colorado plateau, United States of America. Journal of 849 microbiology (Seoul, Korea) 53: 6-13.

850

851 Redfield E, Barns SM, Belnap J, Daane LL, Kuske CR (2002). Comparative diversity 852 and composition of cyanobacteria in three predominant soil crusts of the Colorado 853 Plateau. FEMS Microbiol Ecol 40: 55-63.

854

855 Reed SC, Coe KK, Sparks JP, Housman DC, Zelikova TJ, Belnap J (2012). Changes to 856 dryland rainfall result in rapid moss mortality and altered soil fertility. Nature Climate 857 Change 2: 752.

858

859 Reich P (1997). Soil temperature regimes Natural Resources Conservation Service, 860 United States Department of Agriculture: Washington D.C. America.

861

862 Rivera-Aguilar V, Montejano G, Rodríguez-Zaragoza S, Durán-Díaz A (2006). 863 Distribution and composition of cyanobacteria, mosses and lichens of the biological 864 soil crusts of the Tehuacán Valley, Puebla, México. Journal of Arid Environments 67: 865 208-225.

866

867 Rodriguez-Caballero E, Belnap J, Büdel B, Crutzen PJ, Andreae MO, Pöschl U et al 868 (2018). Dryland photoautotrophic soil surface communities endangered by global 869 change. Nature Geoscience 11: 185-189.

870

135

871 Rogers RW (1972). Soil surface lichens in arid and subarid south-eastern Australia. 872 III.* The relationship between distribution and environment. Australian Journal of 873 Botany 20: 301-316.

874

875 Rosentreter R, Bowker M, Belnap J (2007). A field guide to biological soil crusts of 876 western U.S. drylands: Common lichens and bryophytes. U.S. Government Printing 877 Office: Denver, Colorado.

878

879 Rossi F, De Philippis R (2015). Role of cyanobacterial exopolysaccharides in 880 phototrophic biofilms and in complex microbial mats. Life 5: 1218-1238.

881

882 Rothschild LJ, Mancinelli RL (2001). Life in extreme environments. Nature 409: 1092- 883 1101.

884

885 Ruiz-Moreno D, Pascual M, Riolo R (2006). Exploring network Space with Genetic 886 Algorithms: Modularity, Resilience, and Reactivity. In: Pascual M, Dunne JA (eds). 887 Ecological Networks: Linking Structure to Dynamics in Food Webs. Oxford University 888 Press.

889

890 Schaefer B (2014). Natural products in the chemical industry. Heidelberg : Springer.

891

892 Schloss PD, Westcott SL, Ryabin T, Hall JR, Hartmann M, Hollister EB et al (2009). 893 Introducing mothur: Open-source, platform-independent, community-supported 894 software for describing and comparing microbial communities. Applied and 895 Environmental Microbiology 75: 7537-7541.

136

896

897 Schloss PD, Gevers D, Westcott SL (2011). Reducing the Effects of PCR Amplification 898 and Sequencing Artifacts on 16S rRNA-Based Studies. PLOS ONE 6: e27310.

899

900 Shade A, Peter H, Allison SD, Baho DL, Berga M, Bürgmann H et al (2012). 901 Fundamentals of microbial community resistance and resilience. Frontiers in 902 Microbiology 3.

903

904 Shen B (2003). Polyketide biosynthesis beyond the type I, II and III polyketide 905 synthase paradigms. Current Opinion in Chemical Biology 7: 285-295.

906

907 Shih PM, Wu D, Latifi A, Axen SD, Fewer DP, Talla E et al (2013). Improving the 908 coverage of the cyanobacterial phylum using diversity-driven genome sequencing. 909 Proc Natl Acad Sci U S A 110: 1053-1058.

910

911 Smith FMJ, Wood SA, Wilks T, Kelly D, Broady PA, Williamson W et al (2012). Survey 912 of Scytonema (cyanobacteria) and associated saxitoxins in the littoral zone of 913 recreational lakes in Canterbury, New Zealand. Phycologia 51: 542-551.

914

915 Sogin ML, Morrison HG, Huber JA, Welch DM, Huse SM, Neal PR et al (2006). 916 Microbial diversity in the deep sea and the underexplored “rare biosphere”. 917 Proceedings of the National Academy of Sciences 103: 12115-12120.

918

137

919 Soo RM, Skennerton CT, Sekiguchi Y, Imelfort M, Paech SJ, Dennis PG et al (2014). 920 An Expanded Genomic Representation of the Phylum Cyanobacteria. Genome Biology 921 and Evolution 6: 1031-1045.

922

923 Soule T, Anderson IJ, Johnson SL, Bates ST, Garcia-Pichel F (2009). Archaeal 924 populations in biological soil crusts from arid lands in North America. Soil Biology and 925 Biochemistry 41: 2069-2074.

926

927 Spago FR, Ishii Mauro CS, Oliveira AG, Beranger JPO, Cely MVT, Stanganelli MM et 928 al (2014). Pseudomonas aeruginosa produces secondary metabolites that have 929 biological activity against plant pathogenic Xanthomonas species. Crop Protection 62: 930 46-54.

931

932 Stachelhaus T, Mootz HD, Marahiel MA (1999). The specificity-conferring code of 933 adenylation domains in nonribosomal peptide synthetases. Chem Biol 6: 493-505.

934

935 Stal LJ, Moezelaar R (1997). Fermentation in cyanobacteria. FEMS Microbiology 936 Reviews 21: 179-211.

937

938 Stal LJ (2001). Nitrogen Fixation in Cyanobacteria. eLS. John Wiley & Sons, Ltd.

939

940 Stan-Lotter H, Oren A, Seckbach J (2013). Polyextremophiles : Life Under Multiple 941 Forms of Stress. Springer: Dordrecht.

942 138

943 Starkenburg SR, Reitenga KG, Freitas T, Johnson S, Chain PS, Garcia-Pichel F et al 944 (2011). Genome of the cyanobacterium Microcoleus vaginatus FGP-2, a photosynthetic 945 ecosystem engineer of arid land soil biocrusts worldwide. J Bacteriol 193: 4569-4570.

946

947 States DJ, Gish W (1994). Combined use of sequence similarity and codon bias for 948 coding region identification. Journal of computational biology : a journal of 949 computational molecular cell biology 1: 39-50.

950

951 Steven B, Gallegos-Graves LV, Starkenburg SR, Chain PS, Kuske CR (2012). Targeted 952 and shotgun metagenomic approaches provide different descriptions of dryland soil 953 microbial communities in a manipulated field study. Environmental microbiology 954 reports 4: 248-256.

955

956 Steven B, Gallegos-Graves LV, Belnap J, Kuske CR (2013a). Dryland soil microbial 957 communities display spatial biogeographic patterns associated with soil depth and soil 958 parent material. FEMS Microbiol Ecol 86: 101-113.

959

960 Steven B, Lionard M, Kuske CR, Vincent WF (2013b). High Bacterial Diversity of 961 Biological Soil Crusts in Water Tracks over Permafrost in the High Arctic Polar Desert. 962 PLoS ONE 8.

963

964 Steven B, Kuske CR, Gallegos-Graves LV, Reed SC, Belnap J (2015). Climate change 965 and physical disturbance manipulations result in distinct biological soil crust 966 communities. Applied and Environmental Microbiology 81: 7448-7459.

967

139

968 Strieker M, Tanovic A, Marahiel MA (2010). Nonribosomal peptide synthetases: 969 structures and dynamics. Current opinion in structural biology 20: 234-240.

970

971 Strong CL, Bullard JE, Burford MA, McTainsh GH (2013). Response of cyanobacterial 972 soil crusts to moisture and nutrient availability. Catena 109: 195-202.

973

974 Su YG, Zhao X, Li AX, Li XR, Huang G (2011). Nitrogen fixation in biological soil 975 crusts from the Tengger desert, northern China. European Journal of Soil Biology 47: 976 182-187.

977

978 Swenson TL, Karaoz U, Swenson JM, Bowen BP, Northen TR (2018). Linking soil 979 biology and chemistry in biological soil crust using isolate exometabolomics. Nature 980 Communications 9.

981

982 Thomas AD, Dougill AJ (2007). Spatial and temporal distribution of cyanobacterial soil 983 crusts in the Kalahari: Implications for soil surface properties. Geomorphology 85: 17- 984 29.

985

986 Ullmann I, Budel B (2001). Ecological determinants of species composition of 987 biological soil crusts on a landscape scale. In: Belnap J, Lange OL (eds). Biological 988 soil crusts: structure, function, and management. Springer-Verlag: Berlin Heidelberg.

989

990 Weissman KJ (2009). Chapter 1 Introduction to Polyketide Biosynthesis. Methods in 991 Enzymology. Academic Press. pp 3-16.

140

992

993 Werner JJ, Koren O, Hugenholtz P, DeSantis TZ, Walters WA, Caporaso JG et al 994 (2012). Impact of training sets on classification of high-throughput bacterial 16s rRNA 995 gene surveys. ISME J 6: 94-103.

996

997 Williams G (2013). Engineering polyketide synthases and nonribosomal peptide 998 synthetases. Current opinion in structural biology 23: 603-612.

999

1000 Williams KJ, Belbin L, Austin MP, Stein JL, Ferrier S (2012). Which environmental 1001 variables should I use in my biodiversity model? International Journal of Geographical 1002 Information Science 26: 2009-2047.

1003

1004 Williams WJ, Eldridge DJ, Alchin BM (2008). Grazing and drought reduce 1005 cyanobacterial soil crusts in an Australian Acacia woodland. Journal of Arid 1006 Environments 72: 1064-1075.

1007

1008 Williams WJ, Eldridge DJ (2011). Deposition of sand over a cyanobacterial soil crust 1009 increases nitrogen bioavailability in a semi-arid woodland. Applied Soil Ecology 49: 26- 1010 31.

1011

1012 Williams WJ, Büdel B (2012). Species diversity, biomass and long-term patterns of 1013 biological soil crusts with special focus on Cyanobacteria of the Acacia aneura Mulga 1014 Lands of Queensland, Australia. Algological Studies 140: 23-50.

1015

141

1016 Williams WJ, Büdel B, Reichenberger H, Rose N (2014). Cyanobacteria in the 1017 Australian northern savannah detect the difference between intermittent dry season 1018 and wet season rain. Biodiversity and Conservation 23: 1827-1844.

1019

1020 Wilson ZE, Brimble MA (2009). Molecules derived from the extremes of life. Natural 1021 Product Reports 26: 44-71.

1022

1023 Woodhouse JN, Fan L, Brown MV, Thomas T, Neilan BA (2013). Deep sequencing of 1024 non-ribosomal peptide synthetases and polyketide synthases from the microbiomes of 1025 Australian marine sponges. The ISME Journal 7: 1842-1851.

1026

1027 Woodhouse JN, Rapadas M, Neilan BA (2014). Cyanotoxins. Cyanobacteria. John 1028 Wiley & Sons, Ltd. pp 257-268.

1029

1030 Wu N, Zhang YM, Downing A (2009). Comparative study of nitrogenase activity in 1031 different types of biological soil crusts in the Gurbantunggut Desert, Northwestern 1032 China. Journal of Arid Environments 73: 828-833.

1033

1034 Xiao B, Veste M (2017). Moss-dominated biocrusts increase soil microbial abundance 1035 and community diversity and improve soil fertility in semi-arid climates on the Loess 1036 Plateau of China. Applied Soil Ecology 117-118: 165-177.

1037

1038 Xu S, Nijampatnam B, Dutta S, Velu SE (2016). Cyanobacterial metabolite 1039 calothrixins: Recent advances in synthesis and biological evaluation. Marine Drugs 14.

142

1040

1041 Yaalon DH (1986). Ecosystems of the World, Vol. 12A. Hot Deserts and Arid 1042 Shrublands. Geoderma 39: 165-167.

1043

1044 Yeager CM, Kornosky JL, Housman DC, Grote EE, Belnap J, Kuske CR (2004). 1045 Diazotrophic Community Structure and Function in Two Successional Stages of 1046 Biological Soil Crusts from the Colorado Plateau and Chihuahuan Desert. Applied and 1047 Environmental Microbiology 70: 973-983.

1048

1049 Zaady E, Ben-David EA, Sher Y, Tzirkin R, Nejidat A (2010). Inferring biological soil 1050 crust successional stage using combined PLFA, DGGE, physical and biophysiological 1051 analyses. Soil Biology and Biochemistry 42: 842-849.

1052

1053 Zhang Y (2005). The microstructure and formation of biological soil crusts in their 1054 early developmental stage. Chinese Science Bulletin 50: 117-121.

1055

1056 Zhang Y, Cao C, Peng M, Xu X, Zhang P, Yu Q et al (2014). Diversity of nitrogen- 1057 fixing, ammonia-oxidizing, and denitrifying bacteria in biological soil crusts of a 1058 revegetation area in Horqin Sandy Land, Northeast China. Ecological Engineering 71: 1059 71-79.

1060

1061 Zhao Y, Zhang Z, Hu Y, Chen Y (2016). The seasonal and successional variations of 1062 carbon release from biological soil crust-covered soil. Journal of Arid Environments 1063 127: 148-153.

1064

143

1065 Ziemert N, Podell S, Penn K, Badger JH, Allen E, Jensen PR (2012). The natural 1066 product domain seeker NaPDoS: A phylogeny based bioinformatic tool to classify 1067 secondary metabolite gene diversity. PLoS ONE 7.

1068 1069

144

1070 SUPPLEMENTARY MATERIAL 1071

1072 1073 Supplementary Figure A1.1: Co-occurrence networks for Bare stage biocrusts

145

1074 1075 Supplementary Figure A1.2: Co-occurrence networks for Early stage biocrusts

1076

146

1077 1078 Supplementary Figure A1.3: Co-occurrence networks for Mid stage biocrusts

147

1079 1080 Supplementary Figure A1.4: Co-occurrence networks for Late stage biocrusts

1081

148

1082

1083 1084

1085 Supplementary Figure A1.5: nMDS of sample sites according to (a) collection season (b) site 1086 climate class (c) which kit used for extraction and (d) collection year.

1087

1088

1089

1090

1091

1092

1093

149

1094 Supplementary Table A1.1 Contribution of phyla to nodes and correlations (%). Correlations

1095 categorised into positive (+) and negative (-) within-phylum and among-phyla (%).

Phylum Nodes Correlations Within Phylum Among Phyla Bare + - + - Acidobacteria 1.7 1.9 0.0 0.0 70.0 30.0 Actinobacteria 19.6 17.8 21.4 6.8 52.4 19.4 Firmicutes 2.8 2.0 15.1 0.0 52.8 32.1 Bacteriodetes 9.8 14.9 33.8 1.5 52.7 12.0 Chloroflexi 16.1 13.6 41.0 1.1 38.5 19.4 Cyanobacteria 18.2 17.8 66.5 0.0 13.3 20.2 Deinococcus 0.7 0.8 0.0 0.0 75.0 25.0 Proteobacteria 30.4 30.3 28.9 6.5 45.4 19.1 Others 0.7 0.9 0.0 0.0 73.9 26.1 Early Acidobacteria 4.0 2.2 0.0 0.0 44.8 55.2 Actinobacteria 17.9 16.5 20.6 4.7 47.2 27.6 Firmicutes 0.0 0.0 0.0 0.0 0.0 0.0 Bacteriodetes 4.4 4.9 9.5 0.0 60.3 30.2 Chloroflexi 9.9 13.8 31.3 0.0 36.9 31.8 Cyanobacteria 32.1 25.7 36.0 4.8 27.6 31.5 Deinococcus 0.4 0.2 0.0 0.0 0.0 100.0 Proteobacteria 31.3 36.7 26.5 15.1 34.5 23.9 Others 0.0 0.0 0.0 0.0 0.0 0.0 Mid Acidobacteria 3.3 2.8 10.2 0.0 47.5 42.4 Actinobacteria 17.8 16.7 14.1 9.0 41.1 35.8 Firmicutes 2.9 1.8 15.8 0.0 23.7 60.5 Bacteriodetes 4.7 4.9 5.8 0.0 44.2 50.0 Chloroflexi 12.7 13.8 17.1 2.0 41.0 39.9 Cyanobacteria 26.9 28.1 36.8 4.7 28.3 30.3 Deinococcus 0.4 0.2 0.0 0.0 50.0 50.0 Proteobacteria 31.3 31.8 26.0 10.0 33.4 30.6 Others 0.0 0.0 0.0 0.0 0.0 0.0 Late Acidobacteria 5.0 4.6 6.9 0.0 60.9 32.2 Actinobacteria 16.5 14.0 11.3 3.0 40.8 44.9 Firmicutes 0.0 0.0 0.0 0.0 0.0 0.0 Bacteriodetes 4.1 4.1 5.2 2.6 54.5 37.7 Chloroflexi 12.4 20.6 27.1 0.0 41.9 30.9 Cyanobacteria 29.3 25.7 21.4 3.7 28.0 46.9 Deinococcus 0.4 0.6 0.0 0.0 36.4 63.6 Proteobacteria 31.8 29.0 21.1 10.2 38.2 30.5 Others 0.4 1.4 0.0 0.0 92.6 7.4 1096 150

1097 Supplementary Table A1.2: Significant differences between site 16S rDNA diversity measures 1098 determine via ANOVA multiple-comparisons.

Mean Adjusted P Richness Diff. 95.00% CI of diff. Significant? Summary Value Cooloongup vs. Paraburdoo, CTS -106.9 -207.8 to -5.926 Yes * 0.03 Paraburdoo, CTS vs Cobar Bare 108.8 7.86 to 209.7 Yes * 0.0253 Paraburdoo, CTS vs. Mataranka 104.4 3.46 to 205.3 Yes * 0.0372 Paraburdoo, Cell vs. Cobar Mid 114.1 13.19 to 215.1 Yes * 0.0156 JA Late vs. Cobar Mid -103.4 -204.4 to -2.493 Yes * 0.0404 JA Late vs. Cobar Late -110.9 -211.8 to -9.96 Yes * 0.021 Cooloongup vs. Parabudroo, Cell -141.5 -242.4 to -40.56 Yes ** 0.0011 Paraburdoo, CTS vs Cobar Early 124.8 23.83 to 225.7 Yes ** 0.0058 Paraburdoo, CTS vs. Charters Towers 119.3 18.36 to 220.2 Yes ** 0.0097 Paraburdoo, CTS vs. Rolleston 142.4 41.46 to 243.3 Yes ** 0.001 Paraburdoo, CTS vs. Tara 119.9 18.93 to 220.8 Yes ** 0.0092 Paraburdoo, Cell vs. Cobar Late 121.6 20.66 to 222.5 Yes ** 0.0078 Cooloongup vs. JA Early -164.2 -265.1 to -63.26 Yes *** 0.0001 Cooloongup vs. JA Late -152.2 -253.1 to -51.26 Yes *** 0.0004 Paraburdoo, CTS vs Cobar Mid 148.8 47.83 to 249.7 Yes *** 0.0006 Paraburdoo, CTS vs. Cobar Late 156.2 55.29 to 257.2 Yes *** 0.0003 Cooloongup vs. JA Bare -199.8 -300.7 to -98.86 Yes **** <0.0001 Cooloongup vs. Cobar Bare -215.7 -316.6 to -114.7 Yes **** <0.0001 Cooloongup vs. Cobar Early -231.6 -332.6 to -130.7 Yes **** <0.0001 Cooloongup vs. Cobar Mid -255.6 -356.6 to -154.7 Yes **** <0.0001 Cooloongup vs. Cobar Late -263.1 -364 to -162.2 Yes **** <0.0001 Cooloongup vs. Mataranka -211.3 -312.2 to -110.3 Yes **** <0.0001 Cooloongup vs. Charters Towers -226.2 -327.1 to -125.2 Yes **** <0.0001 Cooloongup vs. Rolleston -249.3 -350.2 to -148.3 Yes **** <0.0001 Cooloongup vs. Cloncurry, Tara -226.7 -327.7 to -125.8 Yes **** <0.0001 Cooloongup vs. Cloncurry, Grananda -205.5 -306.4 to -104.5 Yes **** <0.0001 Mean Adjusted P Evenness Diff. 95.00% CI of diff. Significant? Summary Value Cooloongup vs. Paraburdoo, CTS -0.1417 -0.27 to -0.01337 Yes * 0.0199 Paraburdoo, CTS vs Cobar Bare 0.1379 0.009571 to 0.2662 Yes * 0.0261 Paraburdoo, CTS vs Cobar Mid 0.1295 0.001138 to 0.2578 Yes * 0.0464 Paraburdoo, CTS vs. Cobar Late 0.1356 0.007304 to 0.264 Yes * 0.0305 Paraburdoo, CTS vs. Charters Towers 0.1503 0.022 to 0.2787 Yes * 0.0107 Paraburdoo, CTS vs. Rolleston 0.1478 0.0195 to 0.2762 Yes * 0.0128 Paraburdoo, CTS vs. Tara 0.1368 0.008504 to 0.2652 Yes * 0.0281 Paraburdoo, CTS vs. Mataranka 0.164 0.0357 to 0.2924 Yes ** 0.0039 Cooloongup vs. Parabudroo, Cell -0.1901 -0.3185 to -0.0618 Yes *** 0.0005 Cooloongup vs. JA Bare -0.255 -0.3833 to -0.1267 Yes **** <0.0001 151

Cooloongup vs. JA Early -0.2291 -0.3574 to -0.1007 Yes **** <0.0001 Cooloongup vs. JA Late -0.226 -0.3543 to -0.09767 Yes **** <0.0001 Cooloongup vs. Cobar Bare -0.2796 -0.4079 to -0.1513 Yes **** <0.0001 Cooloongup vs. Cobar Early -0.2544 -0.3827 to -0.126 Yes **** <0.0001 Cooloongup vs. Cobar Mid -0.2712 -0.3995 to -0.1428 Yes **** <0.0001 Cooloongup vs. Cobar Late -0.2773 -0.4057 to -0.149 Yes **** <0.0001 Cooloongup vs. Mataranka -0.3057 -0.4341 to -0.1774 Yes **** <0.0001 Cooloongup vs. Charters Towers -0.292 -0.4204 to -0.1637 Yes **** <0.0001 Cooloongup vs. Rolleston -0.2895 -0.4179 to -0.1612 Yes **** <0.0001 Cooloongup vs. Cloncurry, Tara -0.2785 -0.4069 to -0.1502 Yes **** <0.0001 Cooloongup vs. Cloncurry, Grananda -0.2548 -0.3831 to -0.1265 Yes **** <0.0001 Mean Adjusted P Diversity Diff. 95.00% CI of diff. Significant? Summary Value Paraburdoo, CTS vs. Cobar Early 1.13 0.05138 to 2.209 Yes * 0.0331 Cooloongup vs. Paraburdoo, CTS -1.317 -2.395 to -0.238 Yes ** 0.0067 Paraburdoo, CTS vs Cobar Bare 1.281 0.202 to 2.359 Yes ** 0.0092 Paraburdoo, CTS vs Cobar Mid 1.308 0.229 to 2.386 Yes ** 0.0073 Paraburdoo, CTS vs. Cobar Late 1.37 0.2917 to 2.449 Yes ** 0.0042 Paraburdoo, CTS vs. Mataranka 1.462 0.3837 to 2.541 Yes ** 0.0018 Paraburdoo, CTS vs. Charters Towers 1.397 0.3187 to 2.476 Yes ** 0.0033 Paraburdoo, CTS vs. Rolleston 1.433 0.3547 to 2.512 Yes ** 0.0023 Paraburdoo, CTS vs. Tara 1.295 0.2164 to 2.374 Yes ** 0.0081 Cooloongup vs. Parabudroo, Cell -1.758 -2.837 to -0.6794 Yes *** 0.0001 Cooloongup vs. JA Bare -2.379 -3.457 to -1.3 Yes **** <0.0001 Cooloongup vs. JA Early -2.098 -3.177 to -1.02 Yes **** <0.0001 Cooloongup vs. JA Late -2.041 -3.12 to -0.9627 Yes **** <0.0001 Cooloongup vs. Cobar Bare -2.597 -3.676 to -1.519 Yes **** <0.0001 Cooloongup vs. Cobar Early -2.447 -3.525 to -1.368 Yes **** <0.0001 Cooloongup vs. Cobar Mid -2.624 -3.703 to -1.546 Yes **** <0.0001 Cooloongup vs. Cobar Late -2.687 -3.766 to -1.608 Yes **** <0.0001 Cooloongup vs. Mataranka -2.779 -3.858 to -1.7 Yes **** <0.0001 Cooloongup vs. Charters Towers -2.714 -3.793 to -1.635 Yes **** <0.0001 Cooloongup vs. Rolleston -2.75 -3.829 to -1.671 Yes **** <0.0001 Cooloongup vs. Cloncurry, Tara -2.612 -3.69 to -1.533 Yes **** <0.0001 Cooloongup vs. Cloncurry, Granada -2.39 -3.469 to -1.311 Yes **** <0.0001 1099

1100

1101

1102

152

1103 Supplementary Table A1.3: NRPS and PKS PCR conditions.

Temperature (°C) Duration Number of Cycles 94 2 min 1 94 10 sec 55 (PKS) 52 (NRPS) 30 sec 30 72 1 min 72 7 min 1 4 hold Each PCR reaction had: 1 x Tris-HCL buffer, 2.5 mM MgCL2, 1.5 mM dNTP mix, 1 mM BSA, 0.2 U taq (all Bioline) and 1.25 uM of Forward and Reverse primers (IDT) as per Supplementary Table A1.4 1104

1105

153

1106 Supplementary Table A1.4: Primer name and sequences used for degenerate PCR of NRPS

1107 Condensation and PKS Ketosynthase domains (top) and PCR incorporating sample-specific

1108 indexes (middle) and Illumina sequencing primers (bottom).

Primer Sequence

T7Prom-DKF TAATACGACTCACTATAGGGGTGCCGGTNCC(AG)TGNG(TC)(TC)TC

M13R-DKR CAGGAAACAGCTATGACGCGATGGA(TC)CCNCA(AG)CA(AG)(CA)G

T7Prom-CnDm TAATACGACTCACTATAGGGATGCATCACATT(AG)TN(TC)(TC)NGA

M13R-DCCR CAGGAAACAGCTATGACGTGTTNAC(AG)AA(AG)AANCC(AGT)AT

DEGF.I5.T7P.A AATGATACGGCGACCACCGAGATCTACACAAGCAGCAGGTCGCTGACGCTAATACGACTCACTATAGGG

DEGF.I5.T7P.B AATGATACGGCGACCACCGAGATCTACACACGCGTGAGGTCGCGACGCTAATACGACTCACTATAGGG

DEGF.I5.T7P.C AATGATACGGCGACCACCGAGATCTACACCGATCTACGGTCGCTGACGCTAATACGACTCACTATAGGG

DEGF.I5.T7P.D AATGATACGGCGACCACCGAGATCTACACTGCGTCACGGTCGCTGACGCTAATACGACTCACTATAGGG

DEGF.I5.T7P.E AATGATACGGCGACCACCGAGATCTACACGTCTAGTGGGTCGCTGACGCTAATACGACTCACTATAGGG

DEGF.I5.T7P.F AATGATACGGCGACCACCGAGATCTACACCTAGTATGGGTCGCTGACGCTAATACGACTCACTATAGGG

DEGF.I5.T7P.G AATGATACGGCGACCACCGAGATCTACACGATAGCGTGGTCGCTGACGCTAATACGACTCACTATAGGG

DEGF.I5.T7P.H AATGATACGGCGACCACCGAGATCTACACTCTACACTGGTCGCTGACGCTAATACGACTCACTATAGGG

DEGR.I7.M13R.1 CAAGCAGAAGACGGCATACGAGATAACTCTCGTATACCGGTGTCGCCCAGGAAACAGCTATGAC

DEGR.I7.M13R.2 CAAGCAGAAGACGGCATACGAGATACTATGTCTATACCGGTGTCGCCCAGGAAACAGCTATGAC

DEGR.I7.M13R.3 CAAGCAGAAGACGGCATACGAGATAGTAGCGTTATACCGGTGTCGCCCAGGAAACAGCTATGAC

DEGR.I7.M13R.4 CAAGCAGAAGACGGCATACGAGATCAGTGAGTTATACCGGTGTCGCCCAGGAAACAGCTATGAC

DEGR.I7.M13R.5 CAAGCAGAAGACGGCATACGAGATCGTACTCATATACCGGTGTCGCCCAGGAAACAGCTATGAC

DEGR.I7.M13R.6 CAAGCAGAAGACGGCATACGAGATCTACGCAGTATACCGGTGTCGCCCAGGAAACAGCTATGAC

DEGR.I7.M13R.7 CAAGCAGAAGACGGCATACGAGATGGAGACTATATACCGGTGTCGCCCAGGAAACAGCTATGAC

DEGR.I7.M13R.8 CAAGCAGAAGACGGCATACGAGATGTCGCTCGTATACCGGTGTCGCCCAGGAAACAGCTATGAC

DEGR.I7.M13R.9 CAAGCAGAAGACGGCATACGAGATGTCGTAGTTATACCGGTGTCGCCCAGGAAACAGCTATGAC

DEGR.I7.M13R.10 CAAGCAGAAGACGGCATACGAGATTAGCAGACTATACCGGTGTCGCCCAGGAAACAGCTATGAC

DEGR.I7.M13R.11 CAAGCAGAAGACGGCATACGAGATTCATAGACTATACCGGTGTCGCCCAGGAAACAGCTATGAC

DEGR.I7.M13R.12 CAAGCAGAAGACGGCATACGAGATTCGCTATATATACCGGTGTCGCCCAGGAAACAGCTATGAC

154

DEG.READ1.SEQ.T7P GGTCGCTGACGCTAATACGACTCACTATAGGG

DEG.READ2.SEQ.M13R TATACCGGTGTCGCCCAGGAAACAGCTATGAC

DEG.INDEX.SEQ.M13R.RC GTCATAGCTGTTTCCTGGGCGACACCGGTATA 1109

1110 Supplementary Table A1.5 Significant differences between site NRPS diversity measures 1111 determine via ANOVA multiple-comparisons.

Mean Significant Summar Adjusted P Richness Diff. 95.00% CI of diff. ? y Value Paraburdoo, CELL vs. Cobar Early -88.95 -167 to -10.92 Yes * 0.0169 Paraburdoo, CELL vs. Cobar Mid -121.6 -199.6 to -43.55 Yes *** 0.0006 Paraburdoo, CELL vs. Cobar Late -114.8 -192.8 to -36.79 Yes ** 0.0012 Paraburdoo, CELL vs. Charters Towers -192.9 -278.4 to -107.4 Yes **** <0.0001 Paraburdoo, CTS vs. Cobar Bare -86.43 -164.5 to -8.402 Yes * 0.0218 Paraburdoo, CTS vs. Cobar Early -109.3 -187.3 to -31.27 Yes ** 0.0021 Paraburdoo, CTS vs. Cobar Mid -141.9 -220 to -63.9 Yes **** <0.0001 Paraburdoo, CTS vs. Cobar Late -135.2 -213.2 to -57.14 Yes *** 0.0002 Paraburdoo, CTS vs. Mataranka -96.2 -181.7 to -10.72 Yes * 0.0189 Paraburdoo, CTS vs. Rolleston -98.6 -184.1 to -13.12 Yes * 0.0151 Paraburdoo, CTS vs. Charters Towers -213.3 -298.7 to -127.8 Yes **** <0.0001 Cobar Bare vs. Cooloongup 150.9 72.88 to 228.9 Yes **** <0.0001 Cobar Bare vs. Charters Towers -126.8 -204.8 to -48.79 Yes *** 0.0004 Cobar Early vs. Cloncurry, Granada 85.4 7.369 to 163.4 Yes * 0.0241 Cobar Early vs. Cooloongup 173.8 95.75 to 251.8 Yes **** <0.0001 Cobar Early vs. Charters Towers -104 -182 to -25.92 Yes ** 0.0036 Cobar Mid vs. Cloncurry, Granada 118 40 to 196.1 Yes *** 0.0009 Cobar Mid vs. JA Early 87.08 9.052 to 165.1 Yes * 0.0204 Cobar Mid vs. JA Late 83.73 5.702 to 161.8 Yes * 0.0285 Cobar Mid vs. Cooloongup 206.4 128.4 to 284.4 Yes **** <0.0001 Cobar Mid vs. Cloncurry, Tara 100.5 22.5 to 178.6 Yes ** 0.0051 Cobar Late vs. Cloncurry, Granada 111.3 33.24 to 189.3 Yes ** 0.0017 Cobar Late vs. JA Early 80.32 2.285 to 158.3 Yes * 0.04 Cobar Late vs. Cooloongup 199.6 121.6 to 277.7 Yes **** <0.0001 Cobar Late vs. Cloncurry, Tara 93.77 15.74 to 171.8 Yes * 0.0103 Cobar Late vs. Charters Towers -78.08 -156.1 to -0.05191 Yes * 0.0497 Mataranka vs. Cooloongup 160.7 75.2 to 246.2 Yes **** <0.0001 Mataranka vs. Charters Towers -117.1 -202.5 to -31.57 Yes ** 0.0027 Rolleston vs. Cooloongup 163.1 77.6 to 248.6 Yes **** <0.0001

155

Rolleston vs. Charters Towers -114.7 -200.1 to -29.17 Yes ** 0.0034 Cloncurry, Granada vs. Cooloongup 88.38 2.901 to 173.9 Yes * 0.0386 Cloncurry, Granada vs. Charters Towers -189.4 -274.8 to -103.9 Yes **** <0.0001 JA Bare vs. Cooloongup 146.8 61.3 to 232.3 Yes *** 0.0002 JA Bare vs. Charters Towers -131 -216.4 to -45.47 Yes *** 0.0007 JA Early vs. Cooloongup 119.3 33.85 to 204.8 Yes ** 0.0022 JA Early vs. Charters Towers -158.4 -243.9 to -72.92 Yes **** <0.0001 JA Late vs. Cooloongup 122.7 37.2 to 208.2 Yes ** 0.0016 JA Late vs. Charters Towers -155.1 -240.5 to -69.57 Yes **** <0.0001 Cooloongup vs. Cloncurry, Tara -105.9 -191.4 to -20.4 Yes ** 0.0077 Cooloongup vs. Charters Towers -277.7 -363.2 to -192.3 Yes **** <0.0001 Cloncurry, Tara vs. Charters Towers -171.9 -257.3 to -86.37 Yes **** <0.0001 Mean Significant Summar Adjusted P Evenness Diff. 95.00% CI of diff. ? y Value -0.1421 to - Paraburdoo, CELL vs. Cobar Bare -0.07473 0.007395 Yes * 0.0214 -0.1401 to - Paraburdoo, CELL vs. Cobar Early -0.0728 0.005462 Yes * 0.0268 -0.1713 to - Paraburdoo, CELL vs. Cobar Mid -0.104 0.03666 Yes *** 0.0007 Paraburdoo, CELL vs. Cobar Late -0.1017 -0.169 to -0.03436 Yes *** 0.0009 Paraburdoo, CELL vs. Cooloongup 0.1306 0.05678 to 0.2043 Yes *** 0.0001 Paraburdoo, CELL vs. Charters -0.2462 to - Towers -0.1724 0.09863 Yes **** <0.0001 -0.1843 to - Paraburdoo, CTS vs. Cobar Bare -0.117 0.04965 Yes *** 0.0001 -0.1824 to - Paraburdoo, CTS vs. Cobar Early -0.1151 0.04771 Yes *** 0.0002 -0.2136 to - Paraburdoo, CTS vs. Cobar Mid -0.1463 0.07891 Yes **** <0.0001 -0.2113 to - Paraburdoo, CTS vs. Cobar Late -0.144 0.07661 Yes **** <0.0001 -0.1713 to - Paraburdoo, CTS vs. Mataranka -0.09755 0.02378 Yes ** 0.0039 Paraburdoo, CTS vs. Cloncurry, -0.1699 to - Granada -0.09615 0.02238 Yes ** 0.0046 -0.1724 to - Paraburdoo, CTS vs. JA Late -0.09865 0.02488 Yes ** 0.0035 Paraburdoo, CTS vs. Cooloongup 0.0883 0.01453 to 0.1621 Yes * 0.0107 Paraburdoo, CTS vs. Charters Towers -0.2147 -0.2884 to -0.1409 Yes **** <0.0001 Cobar Bare vs. JA Bare 0.08708 0.01975 to 0.1544 Yes ** 0.005 Cobar Bare vs. JA Early 0.08153 0.0142 to 0.1489 Yes ** 0.0096 Cobar Bare vs. Cooloongup 0.2053 0.1379 to 0.2726 Yes **** <0.0001 Cobar Bare vs. Cloncurry, Tara 0.1273 0.06 to 0.1947 Yes **** <0.0001 Cobar Bare vs. Charters Towers -0.09767 -0.165 to -0.03033 Yes ** 0.0014 Cobar Early vs. JA Bare 0.08515 0.01781 to 0.1525 Yes ** 0.0062 Cobar Early vs. JA Early 0.0796 0.01226 to 0.1469 Yes * 0.0121 Cobar Early vs. Cooloongup 0.2034 0.136 to 0.2707 Yes **** <0.0001

156

Cobar Early vs. Cloncurry, Tara 0.1254 0.05806 to 0.1927 Yes **** <0.0001 -0.1669 to - Cobar Early vs. Charters Towers -0.0996 0.03226 Yes ** 0.0011 Cobar Mid vs. Rolleston 0.0915 0.02416 to 0.1588 Yes ** 0.0029 Cobar Mid vs. JA Bare 0.1164 0.04901 to 0.1837 Yes *** 0.0002 Cobar Mid vs. JA Early 0.1108 0.04346 to 0.1781 Yes *** 0.0003 Cobar Mid vs. Cooloongup 0.2346 0.1672 to 0.3019 Yes **** <0.0001 Cobar Mid vs. Cloncurry, Tara 0.1566 0.08926 to 0.2239 Yes **** <0.0001 -0.1357 to - Cobar Mid vs. Charters Towers -0.0684 0.001062 Yes * 0.0444 Cobar Late vs. Rolleston 0.0892 0.02186 to 0.1565 Yes ** 0.0038 Cobar Late vs. JA Bare 0.1141 0.04671 to 0.1814 Yes *** 0.0002 Cobar Late vs. JA Early 0.1085 0.04116 to 0.1758 Yes *** 0.0004 Cobar Late vs. Cooloongup 0.2323 0.1649 to 0.2996 Yes **** <0.0001 Cobar Late vs. Cloncurry, Tara 0.1543 0.08696 to 0.2216 Yes **** <0.0001 -0.138 to - Cobar Late vs. Charters Towers -0.0707 0.003362 Yes * 0.0341 Mataranka vs. Cooloongup 0.1859 0.1121 to 0.2596 Yes **** <0.0001 Mataranka vs. Cloncurry, Tara 0.1079 0.03413 to 0.1817 Yes ** 0.0013 -0.1909 to - Mataranka vs. Charters Towers -0.1171 0.04333 Yes *** 0.0005 Rolleston vs. Cooloongup 0.1431 0.06928 to 0.2168 Yes **** <0.0001 -0.2337 to - Rolleston vs. Charters Towers -0.1599 0.08613 Yes **** <0.0001 Cloncurry, Granada vs. Cooloongup 0.1845 0.1107 to 0.2582 Yes **** <0.0001 Cloncurry, Granada vs. Cloncurry, Tara 0.1065 0.03273 to 0.1803 Yes ** 0.0015 Cloncurry, Granada vs. Charters -0.1923 to - Towers -0.1185 0.04473 Yes *** 0.0004 JA Bare vs. Cooloongup 0.1182 0.04443 to 0.192 Yes *** 0.0004 JA Bare vs. Charters Towers -0.1848 -0.2585 to -0.111 Yes **** <0.0001 JA Early vs. Cooloongup 0.1238 0.04998 to 0.1975 Yes *** 0.0002 JA Early vs. Charters Towers -0.1792 -0.253 to -0.1054 Yes **** <0.0001 JA Late vs. Cooloongup 0.187 0.1132 to 0.2607 Yes **** <0.0001 JA Late vs. Cloncurry, Tara 0.109 0.03523 to 0.1828 Yes ** 0.0011 -0.1898 to - JA Late vs. Charters Towers -0.116 0.04223 Yes *** 0.0005 -0.1517 to - Cooloongup vs. Cloncurry, Tara -0.07795 0.004185 Yes * 0.0324 Cooloongup vs. Charters Towers -0.303 -0.3767 to -0.2292 Yes **** <0.0001 Cloncurry, Tara vs. Charters Towers -0.225 -0.2988 to -0.1512 Yes **** <0.0001 Mean Significant Summar Adjusted P Diversity Diff. 95.00% CI of diff. ? y Value Paraburdoo, CELL vs. Cobar Bare -0.7732 -1.494 to -0.05223 Yes * 0.0287 Paraburdoo, CELL vs. Cobar Early -0.8505 -1.571 to -0.1296 Yes * 0.0123 Paraburdoo, CELL vs. Cobar Mid -1.135 -1.856 to -0.4142 Yes *** 0.0005 Paraburdoo, CELL vs. Cobar Late -1.168 -1.889 to -0.4469 Yes *** 0.0004 Paraburdoo, CELL vs. Cooloongup 1.425 0.6353 to 2.215 Yes **** <0.0001 157

Paraburdoo, CELL vs. Charters Towers -1.862 -2.652 to -1.072 Yes **** <0.0001 Paraburdoo, CTS vs. Cobar Bare -1.149 -1.87 to -0.4282 Yes *** 0.0005 Paraburdoo, CTS vs. Cobar Early -1.227 -1.947 to -0.5056 Yes *** 0.0002 Paraburdoo, CTS vs. Cobar Mid -1.511 -2.232 to -0.7902 Yes **** <0.0001 Paraburdoo, CTS vs. Cobar Late -1.544 -2.265 to -0.8229 Yes **** <0.0001 Paraburdoo, CTS vs. Mataranka -1.062 -1.852 to -0.2723 Yes ** 0.0033 Paraburdoo, CTS vs. JA Late -0.921 -1.711 to -0.1313 Yes * 0.0137 Paraburdoo, CTS vs. Cooloongup 1.049 0.2593 to 1.839 Yes ** 0.0037 Paraburdoo, CTS vs. Charters Towers -2.238 -3.028 to -1.448 Yes **** <0.0001 Cobar Bare vs. Cooloongup 2.198 1.477 to 2.919 Yes **** <0.0001 Cobar Bare vs. Cloncurry, Tara 1.013 0.2922 to 1.734 Yes ** 0.002 Cobar Bare vs. Charters Towers -1.089 -1.81 to -0.3679 Yes *** 0.0009 Cobar Early vs. JA Early 0.7365 0.01556 to 1.457 Yes * 0.0424 Cobar Early vs. Cooloongup 2.276 1.555 to 2.996 Yes **** <0.0001 Cobar Early vs. Cloncurry, Tara 1.091 0.3696 to 1.811 Yes *** 0.0009 Cobar Early vs. Charters Towers -1.012 -1.732 to -0.2906 Yes ** 0.0021 Cobar Mid vs. Rolleston 0.7332 0.01223 to 1.454 Yes * 0.044 Cobar Mid vs. Cloncurry, Granada 0.7737 0.05273 to 1.495 Yes * 0.0285 Cobar Mid vs. JA Bare 0.9567 0.2357 to 1.678 Yes ** 0.0038 Cobar Mid vs. JA Early 1.021 0.3002 to 1.742 Yes ** 0.0018 Cobar Mid vs. Cooloongup 2.56 1.839 to 3.281 Yes **** <0.0001 Cobar Mid vs. Cloncurry, Tara 1.375 0.6542 to 2.096 Yes **** <0.0001 -1.448 to - Cobar Mid vs. Charters Towers -0.7268 0.005896 Yes * 0.047 Cobar Late vs. Rolleston 0.7658 0.0449 to 1.487 Yes * 0.031 Cobar Late vs. Cloncurry, Granada 0.8063 0.0854 to 1.527 Yes * 0.02 Cobar Late vs. JA Bare 0.9893 0.2684 to 1.71 Yes ** 0.0026 Cobar Late vs. JA Early 1.054 0.3329 to 1.775 Yes ** 0.0013 Cobar Late vs. Cooloongup 2.593 1.872 to 3.314 Yes **** <0.0001 Cobar Late vs. Cloncurry, Tara 1.408 0.6869 to 2.129 Yes **** <0.0001 Mataranka vs. Cooloongup 2.111 1.321 to 2.901 Yes **** <0.0001 Mataranka vs. Cloncurry, Tara 0.926 0.1363 to 1.716 Yes * 0.013 Mataranka vs. Charters Towers -1.176 -1.966 to -0.3863 Yes ** 0.001 Rolleston vs. Cooloongup 1.827 1.037 to 2.617 Yes **** <0.0001 Rolleston vs. Charters Towers -1.46 -2.25 to -0.6703 Yes **** <0.0001 Cloncurry, Granada vs. Cooloongup 1.787 0.9968 to 2.576 Yes **** <0.0001 Cloncurry, Granada vs. Charters Towers -1.501 -2.29 to -0.7108 Yes **** <0.0001 JA Bare vs. Cooloongup 1.604 0.8138 to 2.393 Yes **** <0.0001 JA Bare vs. Charters Towers -1.684 -2.473 to -0.8938 Yes **** <0.0001 JA Early vs. Cooloongup 1.539 0.7493 to 2.329 Yes **** <0.0001 JA Early vs. Charters Towers -1.748 -2.538 to -0.9583 Yes **** <0.0001 158

JA Late vs. Cooloongup 1.97 1.18 to 2.76 Yes **** <0.0001 JA Late vs. Charters Towers -1.317 -2.107 to -0.5273 Yes *** 0.0003 Cooloongup vs. Cloncurry, Tara -1.185 -1.975 to -0.3953 Yes *** 0.0009 Cooloongup vs. Charters Towers -3.287 -4.077 to -2.497 Yes **** <0.0001 Cloncurry, Tara vs. Charters Towers -2.102 -2.892 to -1.312 Yes **** <0.0001 1112

1113 Supplementary Table A1.6 Significant differences between site PKS diversity measures 1114 determine via ANOVA multiple-comparisons.

Mean 95.00% CI of Adjusted P Richness Diff. diff. Summary Value Paraburdoo, CELL vs. Cobar Bare -55.2 -85.76 to -24.63 *** 0.0001 Paraburdoo, CELL vs. Cobar Early -46.23 -76.79 to -15.67 *** 0.0009 Paraburdoo, CELL vs. Cobar Late -71.13 -101.7 to -40.57 **** <0.0001 Paraburdoo, CELL vs. Cobar Mid -60.4 -90.96 to -29.83 **** <0.0001 Paraburdoo, CELL vs. Mataranka -106.3 -139.8 to -72.8 **** <0.0001 Paraburdoo, CELL vs. Rolleston -64.93 -98.41 to -31.45 **** <0.0001 Paraburdoo, CELL vs. JA Bare -82.08 -115.6 to -48.6 **** <0.0001 Paraburdoo, CELL vs. Cooloongup 35.29 1.809 to 68.77 * 0.0333 Paraburdoo, CELL vs. Charters Towers -90.78 -124.3 to -57.3 **** <0.0001 Paraburdoo, CTS vs. Cobar Bare -70.07 -108.7 to -31.41 *** 0.0001 Paraburdoo, CTS vs. Cobar Early -61.1 -99.76 to -22.44 *** 0.0006 Paraburdoo, CTS vs. Cobar Late -86 -124.7 to -47.34 **** <0.0001 Paraburdoo, CTS vs. Cobar Mid -75.27 -113.9 to -36.61 **** <0.0001 Paraburdoo, CTS vs. Mataranka -121.2 -162.2 to -80.14 **** <0.0001 Paraburdoo, CTS vs. Rolleston -79.8 -120.8 to -38.79 **** <0.0001 Paraburdoo, CTS vs. JA Bare -96.95 -138 to -55.94 **** <0.0001 Paraburdoo, CTS vs. JA Late -41.8 -82.8 to -0.7898 * 0.0433 Paraburdoo, CTS vs. Charters Towers -105.7 -146.7 to -64.64 **** <0.0001 Cobar Bare vs. Mataranka -51.08 -81.65 to -20.52 *** 0.0003 Cobar Bare vs. Cloncurry, Granada 40.21 9.648 to 70.78 ** 0.0044 Cobar Bare vs. JA Early 36.74 6.178 to 67.31 * 0.0106 Cobar Bare vs. Cooloongup 90.49 59.92 to 121.1 **** <0.0001 Cobar Bare vs. Cloncurry, Tara 52.26 21.7 to 82.83 *** 0.0002 Cobar Bare vs. Charters Towers -35.58 -66.15 to -5.02 * 0.0143 Cobar Early vs. Mataranka -60.05 -90.61 to -29.49 **** <0.0001 Cobar Early vs. Cloncurry, Granada 31.25 0.6815 to 61.81 * 0.0423 Cobar Early vs. JA Bare -35.85 -66.41 to -5.287 * 0.0133 Cobar Early vs. Cooloongup 81.52 50.96 to 112.1 **** <0.0001 Cobar Early vs. Cloncurry, Tara 43.3 12.73 to 73.86 ** 0.002 Cobar Early vs. Charters Towers -44.55 -75.11 to -13.99 ** 0.0014

159

Cobar Late vs. Mataranka -35.15 -65.71 to -4.587 * 0.0159 Cobar Late vs. Cloncurry, Granada 56.15 25.58 to 86.71 **** <0.0001 Cobar Late vs. JA Early 52.68 22.11 to 83.24 *** 0.0002 Cobar Late vs. JA Late 44.21 13.64 to 74.77 ** 0.0016 Cobar Late vs. Cooloongup 106.4 75.86 to 137 **** <0.0001 Cobar Late vs. Cloncurry, Tara 68.2 37.63 to 98.76 **** <0.0001 Cobar Mid vs. Mataranka -45.88 -76.45 to -15.32 ** 0.001 Cobar Mid vs. Cloncurry, Granada 45.41 14.85 to 75.98 ** 0.0012 Cobar Mid vs. JA Early 41.94 11.38 to 72.51 ** 0.0028 Cobar Mid vs. JA Late 33.47 2.908 to 64.04 * 0.0243 Cobar Mid vs. Cooloongup 95.69 65.12 to 126.3 **** <0.0001 Cobar Mid vs. Cloncurry, Tara 57.46 26.9 to 88.03 **** <0.0001 Mataranka vs. Rolleston 41.35 7.869 to 74.83 ** 0.0082 Mataranka vs. Cloncurry, Granada 91.3 57.81 to 124.8 **** <0.0001 Mataranka vs. JA Early 87.83 54.34 to 121.3 **** <0.0001 Mataranka vs. JA Late 79.36 45.87 to 112.8 **** <0.0001 Mataranka vs. Cooloongup 141.6 108.1 to 175.1 **** <0.0001 Mataranka vs. Cloncurry, Tara 103.3 69.86 to 136.8 **** <0.0001 Rolleston vs. Cloncurry, Granada 49.95 16.46 to 83.43 ** 0.0011 Rolleston vs. JA Early 46.48 12.99 to 79.96 ** 0.0025 Rolleston vs. JA Late 38.01 4.524 to 71.49 * 0.0179 Rolleston vs. Cooloongup 100.2 66.74 to 133.7 **** <0.0001 Rolleston vs. Cloncurry, Tara 62 28.51 to 95.48 **** <0.0001 Cloncurry, Granada vs. JA Bare -67.1 -100.6 to -33.61 **** <0.0001 Cloncurry, Granada vs. Cooloongup 50.28 16.79 to 83.76 ** 0.001 Cloncurry, Granada vs. Charters Towers -75.8 -109.3 to -42.31 **** <0.0001 JA Bare vs. JA Early 63.63 30.14 to 97.11 **** <0.0001 JA Bare vs. JA Late 55.16 21.67 to 88.64 *** 0.0003 JA Bare vs. Cooloongup 117.4 83.89 to 150.9 **** <0.0001 JA Bare vs. Cloncurry, Tara 79.15 45.66 to 112.6 **** <0.0001 JA Early vs. Cooloongup 53.75 20.26 to 87.23 *** 0.0005 JA Early vs. Charters Towers -72.33 -105.8 to -38.84 **** <0.0001 JA Late vs. Cooloongup 62.22 28.73 to 95.7 **** <0.0001 JA Late vs. Charters Towers -63.86 -97.34 to -30.37 **** <0.0001 Cooloongup vs. Cloncurry, Tara -38.23 -71.71 to -4.744 * 0.017 Cooloongup vs. Charters Towers -126.1 -159.6 to -92.59 **** <0.0001 Cloncurry, Tara vs. Charters Towers -87.85 -121.3 to -54.36 **** <0.0001 Mean 95.00% CI of Adjusted P Evenness Diff. diff. Summary Value -0.2139 to - Paraburdoo, CELL vs. Mataranka -0.1346 0.05519 *** 0.0002 -0.1975 to - Paraburdoo, CELL vs. JA Bare -0.1182 0.03879 ** 0.0011 160

Paraburdoo, CELL vs. Cooloongup 0.1534 0.07404 to 0.2328 **** <0.0001 -0.2306 to - Paraburdoo, CTS vs. Mataranka -0.1334 0.03621 ** 0.0028 -0.2142 to - Paraburdoo, CTS vs. JA Bare -0.117 0.01981 * 0.0105 Paraburdoo, CTS vs. Cooloongup 0.1546 0.05736 to 0.2517 *** 0.0005 -0.2217 to - Cobar Bare vs. Mataranka -0.1493 0.07682 **** <0.0001 -0.2053 to - Cobar Bare vs. JA Bare -0.1329 0.06042 **** <0.0001 Cobar Bare vs. Cooloongup 0.1387 0.06624 to 0.2111 **** <0.0001 Cobar Early vs. Mataranka -0.1754 -0.2479 to -0.103 **** <0.0001 -0.2315 to - Cobar Early vs. JA Bare -0.159 0.08659 **** <0.0001 Cobar Early vs. Cooloongup 0.1125 0.04007 to 0.185 *** 0.0007 Cobar Early vs. Charters Towers -0.09553 -0.168 to -0.02309 ** 0.0043 -0.2252 to - Cobar Late vs. Mataranka -0.1528 0.08036 **** <0.0001 -0.2088 to - Cobar Late vs. JA Bare -0.1364 0.06396 **** <0.0001 Cobar Late vs. Cooloongup 0.1352 0.06271 to 0.2076 **** <0.0001 -0.1453 to - Cobar Late vs. Charters Towers -0.0729 0.0004577 * 0.0477 Cobar Mid vs. Mataranka -0.1953 -0.2677 to -0.1228 **** <0.0001 -0.1561 to - Cobar Mid vs. Cloncurry, Granada -0.08362 0.01117 * 0.0154 Cobar Mid vs. JA Bare -0.1789 -0.2513 to -0.1064 **** <0.0001 Cobar Mid vs. Cooloongup 0.09268 0.02024 to 0.1651 ** 0.0058 -0.1878 to - Cobar Mid vs. Charters Towers -0.1154 0.04292 *** 0.0005 Mataranka vs. Rolleston 0.1343 0.05489 to 0.2136 *** 0.0002 Mataranka vs. Cloncurry, Granada 0.1117 0.03229 to 0.191 ** 0.0021 Mataranka vs. JA Early 0.1979 0.1185 to 0.2773 **** <0.0001 Mataranka vs. JA Late 0.1682 0.08884 to 0.2476 **** <0.0001 Mataranka vs. Cooloongup 0.288 0.2086 to 0.3673 **** <0.0001 Mataranka vs. Cloncurry, Tara 0.142 0.06264 to 0.2214 *** 0.0001 0.0005435 to Mataranka vs. Charters Towers 0.0799 0.1593 * 0.0475 -0.1972 to - Rolleston vs. JA Bare -0.1179 0.03849 ** 0.0012 Rolleston vs. Cooloongup 0.1537 0.07434 to 0.2331 **** <0.0001 -0.1746 to - Cloncurry, Granada vs. JA Bare -0.09525 0.01589 * 0.0108 0.006893 to Cloncurry, Granada vs. JA Early 0.08625 0.1656 * 0.0259 Cloncurry, Granada vs. Cooloongup 0.1763 0.09694 to 0.2557 **** <0.0001 JA Bare vs. JA Early 0.1815 0.1021 to 0.2609 **** <0.0001 JA Bare vs. JA Late 0.1518 0.07244 to 0.2312 **** <0.0001 JA Bare vs. Cooloongup 0.2716 0.1922 to 0.3509 **** <0.0001 JA Bare vs. Cloncurry, Tara 0.1256 0.04624 to 0.205 *** 0.0006

161

JA Early vs. Cooloongup 0.09005 0.01069 to 0.1694 * 0.0179 -0.1974 to - JA Early vs. Charters Towers -0.118 0.03864 ** 0.0012 JA Late vs. Cooloongup 0.1198 0.04039 to 0.1991 *** 0.001 -0.1677 to - JA Late vs. Charters Towers -0.0883 0.008943 * 0.0213 -0.2253 to - Cooloongup vs. Cloncurry, Tara -0.146 0.06659 **** <0.0001 Cooloongup vs. Charters Towers -0.2081 -0.2874 to -0.1287 **** <0.0001 Mean 95.00% CI of Adjusted P Diversity Diff. diff. Summary Value Paraburdoo, CELL vs. Mataranka -1.568 -2.209 to -0.9259 **** <0.0001 Paraburdoo, CELL vs. JA Bare -1.336 -1.978 to -0.6944 **** <0.0001 Paraburdoo, CELL vs. Cooloongup 1.322 0.6799 to 1.963 **** <0.0001 Paraburdoo, CELL vs. Charters Towers -0.967 -1.609 to -0.3254 *** 0.001 Paraburdoo, CTS vs. Mataranka -1.754 -2.539 to -0.9677 **** <0.0001 Paraburdoo, CTS vs. JA Bare -1.522 -2.308 to -0.7362 **** <0.0001 Paraburdoo, CTS vs. Cooloongup 1.136 0.3497 to 1.921 ** 0.0016 Paraburdoo, CTS vs. Charters Towers -1.153 -1.939 to -0.3672 ** 0.0013 Cobar Bare vs. Mataranka -1.229 -1.815 to -0.6435 **** <0.0001 Cobar Bare vs. JA Bare -0.9977 -1.583 to -0.412 *** 0.0002 Cobar Bare vs. Cooloongup 1.66 1.074 to 2.246 **** <0.0001 Cobar Bare vs. Charters Towers -0.6287 -1.214 to -0.04297 * 0.0287 Cobar Early vs. Mataranka -1.44 -2.026 to -0.8545 **** <0.0001 Cobar Early vs. JA Bare -1.209 -1.794 to -0.623 **** <0.0001 Cobar Early vs. Cooloongup 1.449 0.8631 to 2.035 **** <0.0001 Cobar Early vs. Charters Towers -0.8397 -1.425 to -0.254 ** 0.0017 Cobar Late vs. Mataranka -1.165 -1.751 to -0.5791 **** <0.0001 Cobar Late vs. JA Bare -0.9333 -1.519 to -0.3476 *** 0.0005 Cobar Late vs. JA Early 0.5952 0.009466 to 1.181 * 0.0443 Cobar Late vs. Cooloongup 1.724 1.138 to 2.31 **** <0.0001 Cobar Mid vs. Mataranka -1.489 -2.074 to -0.9028 **** <0.0001 Cobar Mid vs. JA Bare -1.257 -1.843 to -0.6713 **** <0.0001 Cobar Mid vs. Cooloongup 1.401 0.8148 to 1.986 **** <0.0001 Cobar Mid vs. Charters Towers -0.888 -1.474 to -0.3023 *** 0.0009 Mataranka vs. Rolleston 1.079 0.4369 to 1.72 *** 0.0003 Mataranka vs. Cloncurry, Granada 1.286 0.6444 to 1.928 **** <0.0001 Mataranka vs. JA Early 1.76 1.118 to 2.402 **** <0.0001 Mataranka vs. JA Late 1.522 0.8799 to 2.163 **** <0.0001 Mataranka vs. Cooloongup 2.889 2.247 to 3.531 **** <0.0001 Mataranka vs. Cloncurry, Tara 1.578 0.9364 to 2.22 **** <0.0001 Rolleston vs. JA Bare -0.847 -1.489 to -0.2054 ** 0.0042 Rolleston vs. JA Early 0.6815 0.0399 to 1.323 * 0.0313 Rolleston vs. Cooloongup 1.811 1.169 to 2.452 **** <0.0001 162

Cloncurry, Granada vs. JA Bare -1.055 -1.696 to -0.4129 *** 0.0004 Cloncurry, Granada vs. Cooloongup 1.603 0.9614 to 2.245 **** <0.0001 Cloncurry, Granada vs. Charters Towers -0.6855 -1.327 to -0.0439 * 0.0299 JA Bare vs. JA Early 1.529 0.8869 to 2.17 **** <0.0001 JA Bare vs. JA Late 1.29 0.6484 to 1.932 **** <0.0001 JA Bare vs. Cooloongup 2.658 2.016 to 3.299 **** <0.0001 JA Bare vs. Cloncurry, Tara 1.347 0.7049 to 1.988 **** <0.0001 JA Early vs. Cooloongup 1.129 0.4874 to 1.771 *** 0.0001 JA Early vs. Charters Towers -1.16 -1.801 to -0.5179 *** 0.0001 JA Late vs. Cooloongup 1.368 0.7259 to 2.009 **** <0.0001 JA Late vs. Charters Towers -0.921 -1.563 to -0.2794 ** 0.0017 Cooloongup vs. Cloncurry, Tara -1.311 -1.953 to -0.6694 **** <0.0001 Cooloongup vs. Charters Towers -2.289 -2.93 to -1.647 **** <0.0001 Cloncurry, Tara vs. Charters Towers -0.9775 -1.619 to -0.3359 *** 0.0009 1115

163

1116 Supplementary Table A1.7: Taxonomy and classification of abundant NRPS and PKS OTUs. RA=Relative Abundance.

Custom Database Taxonomy NaPDoS Taxonomy GenBank BLASTn RA OTU Sim. Domain Sim. (%) Phylum Genus Product Organism Accession (%) Class (%) NRPS Pseudomonas OTU01 9.16 Gammaproteobacteria 55 LCL syringomycin 74 Janthinobacterium agaricidamnosum HG322949 brassicacearum OTU02 2.35 Cyanobacteria unclassified 54 LCL syringomycin 82 Microcoleus sp. PCC 7113 CP003630 uncultured_bacterium_c OTU03 2.33 unclassified 51 LCL syringomycin 74 Scytonema sp. NIES-4073 AP018268 ontig00024 OTU04 2.14 Cyanobacteria Nostoc 61 LCL syringomycin 96 Scytonema sp. HK-05 AP018195 Xenorhabdus OTU05 1.51 Gammaproteobacteria 49 LCL syringomycin 72 Calothrix sp. NIES-2098 AP018172 miraniensis OTU06 1.28 Cyanobacteria Acaryochloris 49 LCL syringomycin 96 Scytonema sp. NIES-4073 AP018268 OTU07 1.24 Cyanobacteria Nostoc 55 LCL syringomycin 77 Scytonema sp. HK-05 AP018194 uncultured_bacterium_c OTU08 1.11 unclassified 52 LCL syringomycin 75 Nostoc carneum NIES-2107 AP018183 ontig00024 OTU09 1.03 Cyanobacteria unclassified 58 LCL syringomycin 78 Oscillatoria nigro-viridis PCC 7112 CP003614 OTU10 1.02 Cyanobacteria Chamaesiphon 59 LCL syringomycin 75 Scytonema sp. HK-05 AP018194 PKS OTU01 4.68 Cyanobacteria Symploca sp. HPC-3 73 modular curacin 81 Symploca sp. HPC-3 AY604655 Anabaena cylindrical 92 Scytonema sp. HK-05 AP018195 OTU02 4.67 Cyanobacteria PCC7122 69 modular myxothiazol OTU03 3.82 Cyanobacteria unclassified 76 modular curacin 86 Nostoc sp. NIES-4103 AP018288 OTU04 3.58 Cyanobacteria Symploca sp. HPC-3 79 modular curacin 81 Moorea producens JHB CP017708 OTU05 3.35 Gammaproteobacteria unclassified 74 modular myxothiazol 76 Streptomyces sp. K01-0509 JX545234 Anabaena cylindrical 97 Scytonema sp. NIES-4073 AP018268 OTU06 3.34 Cyanobacteria PCC7122 65 modular myxalamid Anabaena cylindrical 78 Nostoc sp. NIES-4103 AP018288 OTU07 3.24 Cyanobacteria PCC7122 74 modular myxalamid OTU08 2.90 Cyanobacteria Symploca sp. HPC-3 73 modular jamaicamide 84 Symploca sp. HPC-3 AY604655 164

OTU09 2.65 Cyanobacteria Symploca sp. HPC-3 76 modular curacin 79 Symploca sp. HPC-3 AY604655 OTU10 2.20 Cyanobacteria unclassified 69 modular myxothiazol 94 Cylindrospermum stagnale PCC 7417 CP003643 1117

1118 Supplementary Table A1.8: Taxonomy and classification of highly connected 16S rDNA, NRPS, and PKS network OTUs. Deg. = Degree 1119 (number of connecting edges from node), RA=Relative Abundance.

Custom Database Taxonomy NaPDoS Taxonomy GenBank BLASTn Deg. RA OTU Sim. Domain Sim. Accessio (%) (%) Phylum Genus Product Organism (%) Class (%) n NRPS OTU65 0.66 0.27 Cyanobacteria Hapalosiphon 51 LCL syringomycin 74 Scytonema sp. NIES-4073 AP018268 Gammaproteo- OTU05 0.66 1.51 Xenorhabdus 49 LCL syringomycin 72 Calothrix sp. NIES-2098 AP018172 bacteria Acaryochloris marina OTU21 0.64 0.71 Cyanobacteria 57 LCL syringomycin 77 Calothrix sp. NIES-2100 AP018178 MBIC11017 Acaryochloris marina Chroococcidiopsis thermalis OTU47 0.64 0.42 Cyanobacteria 51 LCL syringomycin 74 CP003597 MBIC11017 PCC7203 Oscillatoria nigro-viridis OTU09 0.64 1.03 Cyanobacteria unclassified 58 LCL syringomycin 78 CP003614 PCC7112 Oscillatoria acuminata (M.vag) OTU15 0.64 0.86 Cyanobacteria Gloeobacter kilaueensis JS1 54 LCL syringomycin 73 CP003607 PCC6304 OTU31 0.63 0.56 Cyanobacteria Gloeobacter kilaueensis JS1 54 LCL syringomycin 76 Scytonema sp. HK-05 AP018195 Gammaproteo- Janthinobacterium OTU01 0.62 9.16 Pseudomonas 55 LCL syringomycin 74 HG322949 bacteria agaricidamnosum Chamaesiphon minutus Chroococcidiopsis thermalis OTU13 0.62 0.94 Cyanobacteria 54 LCL syringomycin 76 CP003597 PCC6605 PCC7203 Unclassified Oscillatoria nigro-viridis OTU55 0.61 0.33 unclassified 52 LCL syringomycin 75 CP003614 Environmental PCC7112 PKS OTU24 0.80 0.92 Cyanobacteria Symploca sp. HPC-3 77 modular curacin 82 Symploca sp. HPC-10 AY604660 OTU01 0.73 4.68 Cyanobacteria Symploca sp. HPC-3 73 modular curacin 81 Symploca sp. HPC-3 AY604655

165

Gammaproteo- OTU05 0.69 3.35 Lysobacter capsici 74 modular myxothiazol 76 Streptomyces sp. K01-0509 JX545234 bacteria OTU37 0.65 0.44 Cyanobacteria Oscillatoria sancta 78 modular stigmatellin 83 Scytonema sp. PCC 7110 AY695321 Anabaena cylindrical OTU07 0.63 3.24 Cyanobacteria 74 modular myxalamid 78 Nostoc sp. NIES-4103 AP018288 PCC7122 OTU30 0.63 0.52 Actinobacteria Streptomyces sp. CNH365 58 Trans virginiamycin 71 Lysobacter enzymogenes AP014940 Gammaproteo- Uncultured bacterium clone OTU36 0.59 0.47 Legionella parisiensis 67 modular myxothiazol 74 GU362184 bacteria W102 Hapalosiphon welwitschii UTEX OTU26 0.58 0.77 Unclassified unclassified 82 KS jamaicamide 85 KF699060 B 1830 Corynebacterium ulcerans Uncultured bacterium clone OTU38 0.58 0.43 Actinobacteria 76 KS jamaicamide 74 GU362184 FRC11 W102 OTU193 0.58 0.06 Proteobacteria unclassified 54 KS jamaicamide 70 Streptomyces sp. Tue6075 CP010833 16S rDNA Alphaproteo- Methylobacterium Uncultured cyanobacterium OTU28 1.10 0.38 Not Applicable 100 JQ401959 bacteria unclassified clone CNY_02201 Uncultured bacterium clone OTU29 1.02 0.39 Cyanobacteria unclassified 100 JQ377866 NTS001Fastb1_11200 Alphaproteo- Methylobacterium Uncultured Rubrobacter sp. OTU64 1.01 0.23 99 JQ400555 bacteria unclassified clone CNY_00430 Uncultured Streptophyta clone OTU23 0.98 0.47 Cyanobacteria Chlorophyta unclassified 100 JQ402308 CNY_02607 Alphaproteo- Uncultured soil bacterium OTU66 0.97 0.22 Rhizobiales unclassified 100 HM131966 bacteria clone D2B007 Uncultured bacterium clone KU71350 OTU124 0.91 0.11 Actinobacteria Rubrobacter unclassified 100 DWTP1.3B.H08 5 Uncultured bacterium clone OTU41 0.91 0.33 Cyanobacteria Phormidium 100 GU362231 53 Alphaproteo- Rhodospirillaceae Uncultured bacterium clone OTU96 0.91 0.17 100 JF227222 bacteria unclassified ncd2592h03c1 Alphaproteo- Uncultured bacterium clone OTU114 0.86 0.13 Roseomonas 100 GU362216 bacteria 48 Uncultured bacterium clone OTU87 0.84 0.19 Cyanobacteria unclassified 99 JQ378337 NTS003Powerf8_11444 1120

166

1121 Supplementary Table A1.9 Taxonomy and classification of most central 16S rDNA, NRPS, and PKS network OTUs. BC = Betweeness 1122 Centrality, RA=Relative Abundance.

Custom Database Taxonomy NaPDoS Taxonomy GenBank BLASTn RA OTU BC Sim. Domain Sim. (%) Phylum Genus Product Organism Accession (%) Class (%) NRPS Acaryochloris marina OTU21 0.02477 0.71 Cyanobacteria 57 LCL syringomycin 77 Calothrix sp. NIES-2100 AP018178 MBIC11017 OTU12 0.01807 0.95 Cyanobacteria Nostoc sp. PCC 7107 61 LCL syringomycin 82 Microcoleus sp. PCC 7113 CP003630

OTU50 0.01741 0.36 Gammaproteobacteria unclassified 61 LCL syringomycin 78 Calothrix sp. NIES-2100 AP018178

Chamaesiphon minutus PCC OTU76 0.01617 0.23 Cyanobacteria 55 LCL syringomycin 83 Microcoleus sp. PCC 7113 CP003630 6605 Hapalosiphon welwitschii OTU54 0.01387 0.34 Cyanobacteria 52 LCL syringomycin 76 Calothrix sp. NIES-4105 AP018290 UTEX B 1830 Acaryochloris marina OTU23 0.01324 0.70 Cyanobacteria 47 LCL syringomycin 67 Nostoc carneum NIES-2107 AP018183 MBIC11017

OTU01 0.01278 9.16 Gammaproteobacteria Pseudomonas 55 LCL syringomycin 74 Janthinobacterium agaricidamnosum HG322949

OTU212 0.01039 0.07 Cyanobacteria Prochloron didemni P1-Palau 54 LCL syringomycin 78 Calothrix sp. NIES-4105 AP018290

OTU191 0.01035 0.08 Cyanobacteria Nostoc sp. GSV224 61 LCL syringomycin 89 Nostoc piscinale CENA21 CP012036 OTU30 0.01031 0.57 Cyanobacteria Nostoc sp. GSV224 58 LCL syringomycin 81 Nostoc punctiforme PCC 73102 CP001037 PKS OTU24 0.04627 0.92 Cyanobacteria Symploca sp. HPC-3 77 modular curacin 82 Symploca sp. HPC-10 AY604660 OTU08 0.01967 2.90 Cyanobacteria Symploca sp. HPC-3 73 KS jamaicamide 84 Symploca sp. HPC-3 AY604655 Hapalosiphon welwitschii UTEX B OTU26 0.01904 0.77 Unclassified unclassified 82 KS jamaicamide 85 KF699060 1830 OTU13 0.01630 1.41 Cyanobacteria Nostoc sp. CCAP 1453/38 73 modular curacin 78 Moorea producens PAL-8-15-08-1 CP017599

167

OTU01 0.01623 4.68 Cyanobacteria Symploca sp. HPC-3 73 modular curacin 81 Symploca sp. HPC-3 AY604655 OTU58 0.01604 0.23 Cyanobacteria Symploca sp. HPC-3 76 modular curacin 81 Symploca sp. HPC-3 AY604655

OTU33 0.01373 0.50 Cyanobacteria Symploca sp. HPC-3 78 modular curacin 79 Cylindrospermum licheniforme UTEX B KX682397

Hapalosiphon welwitschii UTEX B OTU28 0.01274 0.70 Cyanobacteria Symploca sp. HPC-3 74 modular curacin 85 KF699060 1830

OTU05 0.01111 3.35 Gammaproteobacteria Lysobacter capsici 74 modular myxothiazol 76 Streptomyces sp. K01-0509 JX545234

OTU29 0.01190 0.61 Actinobacteria Corynebacterium mustelae 74 KS jamaicamide 71 Uncultured bacterium clone V99 GU385446 16S rDNA Uncultured Rubrobacter sp. clone OTU64 0.03787 0.23 Alphaproteobacteria Methylobacterium unclassified Not Applicable 99 JQ400555 CNY 00430 OTU147 0.03530 0.09 Alphaproteobacteria Rhizobiales unclassified 96 Nostoc sp. NIES-3756 AP017295 Uncultured cyanobacterium clone OTU28 0.03428 0.38 Alphaproteobacteria Methylobacterium unclassified 100 JQ401959 CNY 02201 Uncultured bacterium clone BSL OTU35 0.03123 0.34 Cyanobacteria Tolypothrix distorta 100 KU884011 7642A01 Uncultured bacterium clone OTU33 0.02866 0.33 Cyanobacteria Cyanobacteria unclassified 99 JQ369494 NC2F1e9 19504 Uncultured bacterium clone OTU124 0.02174 0.11 Actinobacteria Rubrobacter unclassified 100 KU713505 DWTP1.3B.H08 Uncultured Streptophyta clone CNY OTU23 0.02158 0.47 Cyanobacteria Chlorophyta unclassified 100 JQ402308 02607 Uncultured bacterium clone garden OTU13 0.02090 0.66 Cyanobacteria Phormidium unclassified 99 KP157256 soil 58033 Uncultured bacterium clone OTU7 0.02013 0.86 Cyanobacteria Brasilonema unclassified 100 JQ377206 NT30e1 19723 Uncultured bacterium clone OTU6 0.01958 0.96 Cyanobacteria Cyanobacteria unclassified 100 JQ384935 NT51d6 14715 1123

1124 168