Method Development for the in Situ Screening of Uncultivable Bacteria for Antibiotic Production

Method Development for the in situ Screening of Uncultivable Bacteria for Antibiotic Production

Daniel Blicher Holst Hansen

Department of Life Sciences, Imperial College London

A thesis submitted for the degree of Doctor of Philosophy

December 2018

Acknowledgements

I would like to thank my two supervisors Harry Low and Martin Buck for their continued guidance and support during the past three years. Your guidance made me a better scientist and gave me a skill set that I hope to use for the rest of my professional career - and for that I am deeply grateful. Also thank you to my two examiners Dr Thomas Bell and Dr Mark Paget. I hope you enjoy reading my thesis.

I would like to acknowledge the students I have supervised during my studies. Thank you, Belen Sola Barrado, for your help, especially in developing a M. smegmatis luminescence strain. Thank you to Bernard Cassar Torreggiani and Daniel Corredera Nadal for your help screening for novel antibiotic producers. Finally, a special thank you to Andrew Morrison, who on multiple occasions went above and beyond on a project that needed all the hands it could get. Andrew’s biggest contributions included helping me develop the B. subtilis reporter strains and optimising the microscopy images. The facility managers involved in this project also deserve recognition and gratitude. This includes Jane, Catherine, and Jess in flow cytometry and Paul Brown from physics. Steve Nelson and Stephen Rothery both contributed with unique skill sets that would have been hard to be without.

I am as always grateful to my family for their continued support, without which this would not have been possible: my mum, dad, brother and sisters, especially. In that regard, I would like to tell Dagmar Scholes and Sönke Mehrgardt that I truly appreciate their help and the degree to which they made me feel part of the family. I would also like to tell all my colleagues at Imperial College London how much I appreciated their help and support. This is especially true of Eli Cohen, Cian Duggan, Josie Ferreira, and, of course, Olga Bohuszewicz.

I am grateful to the anonymous donors that made an applied, yet dreamy, project possible. I hope this thesis shows that significant progress has been made, and that the money given might therefore end up helping a lot of people.

Finally, a heartfelt thank you to Natalie Scholes who keeps helping me be better.

Abstract

Multidrug-resistant bacteria are one of the most critical threats facing human health. Despite an urgent need for novel treatments, antibiotic discovery has stalled. Consequently, no novel antibiotic class has been discovered in the last 40 years. Most existing antibiotics were discovered by screening the 1 % of bacteria that can be cultivated in standard laboratory conditions. The 99 % residual bacteria represent a bioresource that could yield novel antibiotics; but at present this bioresource remains largely unscreened. Here I present the development of a novel high throughput in situ screening array (“ISSA” platform), which screens soil bacteria for antibiotic production in their native environment. The developed platform can isolate a known number of soil strains and screen them against specific clinically relevant pathogens. The platform consists of 4080 discrete compartments, with each compartment containing an average of one soil bacterium and an adjacent luxCDABE bioreporter strain that turns off when antibacterial compounds are secreted. The in situ incubation in the ISSA platform significantly boosts cultivation rates of previously uncultivable bacteria, resulting in the isolation and cultivation of novel bacterial strains. The ISSA platform overall leads to an efficient identification process of soil bacteria that secrete antibacterial compounds. So far, the platform has identified seven antibacterial compound producing strains. Three of these strains are novel antibacterial compound producers, which demonstrates that the developed ISSA platform can be a valuable tool in the race to discover novel antibiotics.

Declaration of originality

I declare that this thesis titled “Method Development for the in situ Screening of Uncultivable Bacteria for Antibiotic Production” is original work conducted by myself for the degree of Doctor of Philosophy. Work performed by others has been acknowledged and credited. All resources and ideas from the work of others has been appropriately referenced.

The copyright of this thesis rests with the author. Unless otherwise indicated, its contents are licensed under a Creative Commons. Attribution Non Commercial 4.0 International Licence (CC BY- NC). Under this licence, you may copy and redistribute the material in any medium or format. You may also create and distribute modified versions of the work. This is on the condition that: you credit the author and do not use it, or any derivative works, for a commercial purpose. When reusing or sharing this work, ensure you make the licence terms clear to others by naming the licence and linking to the licence text. Where a work has been adapted, you should indicate that the work has been changed and describe those changes.

Please seek permission from the copyright holder for uses of this work that are not included in this licence or permitted under UK Copyright Law.

Contents Acknowledgements ...... 2 Abstract ...... 3 Declaration of originality...... 4 Copyright declaration ...... 4 List of figures ...... 9 List of tables ...... 12 Abbreviations ...... 13 Chapter 1. Introduction ...... 14 1.1 Defining antibiotics ...... 14 1.2 The historic importance of antibiotics ...... 14 1.3 Increase in bacterial resistance to antibiotics ...... 15 1.4 Most urgent threats posed by bacterial pathogens ...... 16 1.5 The history of antibiotics discovery ...... 17 1.6 Re-evaluating bacteria as a bioresource for novel antibiotics ...... 22 1.7 Uncultivable bacteria - metagenomic analysis of bacterial diversity in soil ...... 22 1.8 Uncultivable bacteria - metagenomic functional analysis of secondary metabolites in soil ...... 25 1.9 Cultivating unculturable bacteria – understanding the required growth conditions ...... 26 1.10 Cultivating unculturable bacteria – existing methods ...... 27 1.11 Using uncultivable bacteria as a resource for novel antibiotics ...... 31 1.12 Biosensors as in vivo antibiotic secretion detectors ...... 32 Chapter 2. Aims ...... 33 Chapter 3. Materials and methods ...... 36 3.1 Materials ...... 36 3.1.1 Media ...... 36 3.1.2 Bacterial strains ...... 37 3.1.3 Vectors ...... 38 3.1.4 Primers ...... 38 3.1.5 General lab reagents and kits ...... 38 3.1.6 General materials ...... 38 3.1.7 Machines and microscopes ...... 39 3.1.8 Software ...... 39 3.3 Methods ...... 39 3.3.1 Growth of bacteria for quantification of proxy species in soil ...... 39

3.3.2 Staining bacteria in soil with CFDA-SE/water and CFDA-SE/EDTA ...... 40 3.3.3 Preparing CFDA-SE stained bacteria for microscopy ...... 40 3.3.4 Quantification of bacteria stained using flow cytometry ...... 40 3.3.5 Machining ISSA arrays ...... 41 3.3.6 Accuracy testing of FACS Sorter ...... 41 3.3.7 Accuracy testing of ECHO 555 liquid handler ...... 41 3.3.8 Preparing “proof of principle” loading media...... 41 3.3.9 Dipping loading protocol ...... 42 3.3.10 Mechanical shearing loading protocol ...... 42 3.3.11 Vacuum-assisted loading protocol ...... 42 3.3.12 Sequential loading protocol ...... 42 3.3.13 Proof of principle harvesting protocols – centrifugation ...... 43 3.3.14 Proof of principle harvesting protocol – mechanical disruption ...... 43 3.3.15 ISSA protocol ...... 43 3.3.16 Cultivating uncultivable bacteria using a 384-compartment ISSA array with no reporter strain ...... 47 3.3.17 16S rRNA species identification ...... 47 3.3.18 Reporter strain development – Pseudomonas putida and Pseudomonas fluorescens (P1H2 and FH1) ...... 47 3.3.19 Reporter strain development – Bacillus subtilis ...... 48 3.3.20 Preparation of electrocompetent Mycobacterium smegmatis ...... 48 3.3.21 LuxCDABE integration of M. Smegmatis ...... 48 3.3.22 Measuring light intensities of newly developed reporter strains ...... 49 3.3.23 Breaking up mycelia (mycelia break up protocol) ...... 49 3.3.24 Classical Waksman detection of antibiotic production using S. aureofaciens or S. venezuelae - overlay ...... 49 3.3.25 Classical Waksman detection of antibiotic production – cut out ...... 50 3.3.26 Broad assay of antibiotic production using various producer strains, reporter strains, and media in 96-well plates (mixed) ...... 50 3.3.27 Detecting antibiotic production using S. aureofaciens and Mini Bac Pveg Lux in mixed/layered formats ...... 51 3.3.28 In vitro ISSA platform detection of antibiotic production ...... 51 3.3.29 In vivo ISSA platform detection of antibiotic production ...... 52 Chapter 4. Quantifying soil bacteria ...... 53 4.1 Introduction ...... 53 4.2 Results ...... 55 4.2.1 Selectively staining soil bacteria ...... 55 6

4.2.2 Testing the scope of CFDA-SE across the bacterial kingdom ...... 56 4.2.3 Quantification of soil bacteria using flow cytometry ...... 65 4.2.4 Measuring bacterial population of soil dilutions filtered with 35 μm and 5 μm filters 66 4.2.5 Further controls ...... 67 4.2.6 Quantifying proxy species in soil...... 67 4.2.7 Sensitivity and selectivity of the CFDA-SE/EDTA quantification protocol ...... 69 4.3 Discussion ...... 70 Chapter 5. ISSA platform design ...... 72 5.1 Introduction ...... 72 5.1.1 Design considerations for the physical ISSA platform ...... 72 5.1.2 Loading protocols considerations ...... 73 5.1.3 Incubation and harvesting considerations ...... 74 5.2 Results ...... 75 5.2.1 Platform design ...... 75 5.2.2 Developing a loading protocol ...... 76 5.2.3 Sequential loading ...... 81 5.2.4 Developing a harvesting protocol ...... 83 5.2.5 Sterility experiments ...... 85 5.2.6 Crosstalk experiments ...... 85 5.2.7 In vivo experiments to establish the cultivation ability of the ISSA platform ...... 87 5.2.8 Building a library of uncultivable bacteria ...... 87 5.3 Discussion ...... 89 Chapter 6. Development of luminescent reporter strains ...... 91 6.1 Introduction ...... 91 6.1.1 Environmental bacterial bioreporters ...... 91 6.1.2 Bioreporter strain selection ...... 91 6.2 Results ...... 95 6.2.1 Testing prior published strains ...... 95 6.2.2 Protocol for development of lux integrants ...... 96 6.2.3 Development of secondary reporter strain for in vitro experiments ...... 100 6.3 Discussion ...... 101 Chapter 7. Understanding antibiotic secretion ...... 103 7.1 Introduction ...... 103 7.2 Results ...... 105 7.2.1 Comparing S. aureofaciens and S. venezuelae as producer strains ...... 105

7.2.2 Protocol development for reduction of hyphae by Streptomyces species grown in liquid media...... 106 7.2.3 Exploring antibiotic production in different setups when the reporter strain and the producer strain are introduced simultaneously ...... 108 7.2.4 Exploring the effect of layered and mixed setups on platform sensitivity to antibiotic production ...... 110 7.2.5 Identifying significant variables using multivariable regression ...... 111 7.2.6 Further analysis of the multivariable regression results ...... 115 7.3. Discussion ...... 117 Chapter 8. In vitro and in situ screening of antibiotic production in soil bacteria using the developed ISSA platform...... 119 8.1 Introduction ...... 119 8.2 Results ...... 120 8.2.1 In vitro screening ...... 120 8.2.2 Harvesting of in vitro hits ...... 124 8.2.3 Confirming antibacterial compound production by isolated hits ...... 126 8.2.4 In situ screening ...... 127 8.3 Discussion ...... 130 Chapter 9. Final summary & future perspectives ...... 132 Chapter 10. Bibliography ...... 135 Appendix I ...... 146 A1.1 Pre-processing image macro...... 146 A1.2. Create grid macro ...... 146 A1.3 Quantify luminescence macro ...... 149 Appendix II – Terms for using previously published work ...... 150 AII.1. Oxford University Press ...... 150 AII.2. American society for Microbiology ...... 152

List of figures

Fig 1.1. Overview of the Waksman platform – page 18

Fig 1.2. Overview of the steps in the Waksman platform compared to the steps in high- throughput screening – page 21

Fig 1.3. Essential steps in antibiotic screening using the Ichip procedure – page 31

Fig 2.1. ISSA overview – page 34

Fig 2.2. Overview of the chapter development of the ISSA platform – page 35

Fig 3.1. Key components for loading an ISSA platform – page 44

Fig 4.1. Image of CFDA/SE-stained soil isolate next to autofluorescent sediments – page 54

Fig 4.2. Autoclaved and pristine environmental samples stained using CFDA-SE – page 56

Fig 4.3. Composites of stained autoclaved soil and pure cultures stained with CFDA-SE – page 58

Fig 4.4. Panel of S. venezuelae and M. smegmatis stained with CFDA-SE with water or EDTA. - page – page 60

Fig 4.5. Composites of fluorescent and light images of pure cultures stained with CFDA-SE with EDTA – page 62

Fig 4.6. Autoclaved and pristine soil stained with CFDA-SE/EDTA and filtered using a 200 μm filter – page 63

Fig 4.7. Pristine soil stained with CFDA-SE/EDTA and filtered using 35 μm and 5 μm filters – page 65

Fig 4.8. Isolating soil bacteria using flow cytometry – page 66

Fig 4.9. Isolating soil bacteria from environmental samples using CFDA-SE/EDTA and 35 μm and 5 μm filters – page 67

Fig 4.10. Comparing the efficiency of CFDA-SE/water and CFDA-SE/EDTA when staining S. venezuelae and P. putida using flow cytometry – see page 68

Fig 4.11. Sorting S. venezuelae using an BDFACs Aria III cell sorter – page 70

Fig 5.1. Design of the ISSA prototype – page 75

Fig 5.2. 0.5 mm grade 2 titanium sheet laser cut with decreasing compartment diameter (ø) in a standard 384-well plate format – page 76

Fig 5.3. Different designs of the ISSA platform – page 77

Fig 5.4. Colorimetric accuracy test of the FACS ARIA dispensing capabilities – page 78

Fig 5.5. Colorimetric accuracy test of the ECHO 555 dispensing capabilities – page 79

Fig 5.6. Manual loading protocols and efficiencies– page 81

Fig 5.7. Sequential loading protocol – page 82

Fig 5.8. Luminescence exhibited by reporter strains in sequential loading protocol – page 83

Fig 5.9. Harvesting protocols – page 85

Fig 5.10. Crosstalk experiments – page 86

Fig 5.11. Cultivated phyla from ISSA platform when run without reporter strain – page 88

Fig 6.1. E. coli bioreporter development – page 93

Fig 6.2. Plasmid maps for integration of luxCDABE operons in M. smegmatis and B. subtilis – page 95

Fig 6.3. Luminescence of GC2 bioreporter before and after incubation in soil – page 96

Fig 6.4. Selecting luminescent integrants – page 97

Fig 6.5. Protocol for the selection of integrants with high and stable luminescence – page 99

Fig 6.6. Luminescence of P1H2 before and after in situ incubation – page 100

Fig 6.7. Luminescence over time of Mini Bac Pveg Lux and M. smegmatis G13 Lux – page 101

Fig 7.1. Clearance zones caused by antibiotic secretion by S. venezuelae and S. aureofaciens after 5 days in YM media – page 106

Fig 7.2. S. aureofaciens grown in various conditions – page 108

Fig 7.3. Testing the ability of model producer strains to kill/inhibit the reporter strains P1H2 and Mini Bac Pveg luxCDABE – page 109

Fig 7.4. Platform sensitivity to antibiotic production in layered and mixed setups – page 111

Fig 7.5. Standardised luminescence over time for layered and mixed setups – page 117

Fig 8.1. Blank in vitro luminescent array results – page 121

Fig 8.2. 1.0 bacterium in vitro luminescent array results – page 123

Fig 8.3. Analysis of select tail end observations at day 0 and day 7 – page 124

Fig 8.4. Purification of producer strains – page 125

Fig 8.5. Clearance zones by M. varians and S. maltophilia – page 126

Fig 8.6. Antibiotic clearance of P1H2 and Mini Bac Pveg Lux caused by P. parabrevis – page 127

Fig 8.7. In vivo luminescent array results for blank array and 1.0 bacterium per compartment arrays – page 129

Fig 8.8. Clearance zones of Mini Bac Pveg Lux caused by B. mycoides – page 130

List of tables

Table 1.1. Main bacterial threats associated with antibiotic resistance - page 17

Table 1.2. Important classes of antibiotic – page 19

Table 4.1. Strains tested for CFDA-SE staining efficiency – page 57

Table 4.2. Population estimates using CFDA-SE/water, CFDA-SE/EDTA, and plating – page 69

Table 5.1. Poisson distribution values for λ = (1-5) – page 74

Table 5.2. Species designation of domesticated strains using the ISSA platform without reporter strains – page 88

Table 7.1. OD600 of S. aureofaciens using different detergents – page 107

Table 7.2. Multivariable regression of standardised luminescence from layered & mixed setup – page 113

Table 7.3. Multivariable regression of standardised luminescence using only data from the layered setup – page 114

Table 7.4. Multivariable regression of standardised luminescence using only data from the mixed setup – page 115

Table 8.1. Harvested hits from in vitro ISSA platform screenings – page 126

Abbreviations

AU Arbitrary units BD Becton Dickinson bp Base pair cAMP cyclic AMP CDC Centers for Disease Control and Prevention (US) CFU Colony forming unit CFDA Carboxyfluorescein diacetate CFDA-SE Carboxyfluorescein diacetate succinimidyl ester CTC 5-cyano-2,3-ditolyl tetrazolium chloride DAPI 4′,6-diamidino-2-phenylindole DIL Dioctadecyl Tetramethylindocarbocyanine Perchlorate DMSO Dimethyl Sulfoxide GFP Green Fluorescent Protein h Hour HTP High throughput HTS High-throughput sequencing Ichip Isolation chip ISSA In situ Screening Assay M Molar mM Millimolar Mb Megabase MIC Minimum inhibitory concentration min Minute Mini Bac Super-competent IIG-Bs20-11 B. subtilis ml Millilitre nl Nanolitre OD Optical density P1H2 Developed P. putida LuxCDABE reporter strain PBS Phosphate-buffered saline PCR Polymerase Chain Reaction RLU Relative luminescence unit RPM Revolutions per minute rRNA Ribosomal RNA sec Seconds U. University U.o.A University of Aberdeen ULGTA Ultra-low gelling temperature agarose W Watts WHO World Health Organisation β Beta λ Lambda Ø Diameter μg Microgram μl Microlitre μm Micrometre

Chapter 1. Introduction

1.1 Defining antibiotics The term “antibiotics” has a wide range of definitions [1]. Historically, the standard definition of antibiotics reflected Waksman’s definition from 1947 which defined an antibiotic as:

“ a chemical substance, produced by micro-organisms, which has the capacity to inhibit the growth of and even to destroy bacteria and other micro-organisms" [2].

The clinical success of antibiotics as in vivo treatments against bacterial infections has, however, meant that the term “antibiotics” has increasingly shifted towards being defined as :

“a medical chemical compound that at low doses is able to selectively kill or inhibit growth of targeted bacteria with little to no impact on the patient”.

The latter definition has become increasingly prevalent in scientific literature [1, 3-5], and will be used in this thesis unless stated otherwise. The latter definition changes the meaning of antibiotics in three important ways. Firstly, the latter use of the term antibiotics does not distinguish between compounds that have been derived naturally or synthetically. Secondly, it only encompasses chemicals with mechanism of action that are sufficiently selective to have low toxicity and high efficiency even at low concentrations. Thirdly, the definition excludes antifungals and antiviral drugs from the definition of antibiotics.

Based on the latter definition, it follows that discovered compounds that have antibacterial properties will not be described as “antibiotics” in this thesis. Instead such compounds will be referred to “antibacterial compounds” until the compounds have been isolated and their toxicity and efficiency for use in medicine or veterinary care have been examined.

1.2 The historic importance of antibiotics At the beginning of the 20th century, bacterial infections such as pneumoniae, tuberculosis, and diphtheria were among the main causes of death in Western societies. In 1900 pneumonia and

14 tuberculosis alone contributed to > 20 % of all deaths in the United States [6]. Antibiotics provided a treatment option that had the potential to cure affected patients in a matter of days. Antibiotics therefore had an immediate effect on death rates associated with bacterial infections. One example of the efficacy of antibiotics is the introduction of streptomycin. After the introduction of streptomycin in 1945, the mortality rates associated with tuberculosis decreased from 39.9 deaths per 100,000 population in 1945 to 9.1 deaths per 100,000 between 1945 to 1955 [7]. On a broader scale, it is hard to quantify the exact effect of antibiotics on fatality rates, as antibiotic discovery coincided with other major developments, including increases in wealth and a higher level of adherence to antiseptic techniques [8, 9]. The combined effect of the above factors, however, resulted in bacterial infections no longer being among the main killers in Western society [6]. The effective and quick treatment provided by antibiotics also greatly aided the development of modern Western healthcare [10]. By concentrating people with compromised health states in centres of excellence for specialized procedures, hospitals become vulnerable to a concentration of opportunistic bacterial pathogens such as Pseudomonas aeruginosa (P. aeruginosa) or Actinetobacter baumannii (A. baumannii) [11, 12]. Procedures such as advanced surgery [13], implants [12], bone-marrow transplantations [14], and chemotherapy [15] further increase the short-term vulnerability of patients to microbial invasion. Antibiotics mitigate this risk and thereby allow hospitals to carry out majorly invasive or immune compromising procedures that might otherwise be considered too risky.

1.3 Increase in bacterial resistance to antibiotics The widespread use of antibiotics, however, places strong selective pressure on pathogens to evade antibiotics, which has resulted in the development of multi-resistant pathogens across the bacterial kingdom [4, 11, 16]. Bacterial resistance refers to the ability of bacteria to continue to grow even after exposure to therapeutic levels of the antibiotics. Acquired antibiotic resistance happens mainly through three mechanisms [3]. Firstly, resistance to an antibiotic can happen through the modification of the antibiotic target by genetic mutation or post-translational modification of the target. Secondly, by minimizing the intracellular concentrations of an antibiotic, either through reducing the level of penetration into the bacterium, or by pumping out the antibiotics via efflux pumps. Thirdly, antibiotic resistance can happen through the inactivation of antibiotics by hydrolysis or modification [3]. The acquisition of resistance mechanisms occurs mainly through either spontaneous mutations or horizontal gene transfer. It has e.g. been

15 documented that mutations in Enterobacteriaceae, Pseudomonas species (spp.), and Acinetobacter spp. that lead to reductions in porin expression significantly contribute to the resistance to carbapenems and cephalosporins [3, 17]. Furthermore, bacteria, including Enterobacteriaceae, P. aeruginosa, and Staphylococcus aureus (S. aureus), can acquire resistance through the overexpression of efflux pumps, which is often the result of mutations in local or global regulators [17]. For example, it was found that mutations to the mtrR promoter region in Neisseria gonorrhoeae resulted in the overexpression of the MtrC-MtrD-MtrE efflux pump system, which conferred increased resistance to macrolide antibiotics [18]. Looking at acquired resistance through horizontal gene transfer, the blaCTX-M14 gene, which codes for an extended-spectrum β- lactamase is an example of a gene transferred across species via the IncK plasmid in Enterobacteriaceae [19-21]. Horizontal gene transfer poses a specifically potent threat to modern antibiotics, as the occurrence of resistance plasmids can lead to the rapid loss of antibiotic activity across a wide spectrum of bacterial species [4].

The resulting global fatalities associated with acquired antibiotic resistance is estimated to be 700,000 per year. The regional numbers for Europe and the US are 25,000 per year and 23,000 per year respectively [22, 23]. Furthermore, the Centers for Disease Control and Prevention (CDC) estimates that > 2,000,000 cases of illnesses yearly in the US can be directly attributed to antibiotic resistance [45], leading to prolonged periods at home or in hospitals, and consequently significant productivity losses for the overall economy. The reliance on antibiotics in the global food production and health industries means that there is a continual selective pressure on bacteria to evade antibiotics. This pressure means that the level of antibiotic resistance among bacterial pathogens is expected to rise, resulting in an estimated 10,000,000 fatalities per year globally by 2050 [24].

1.4 Most urgent threats posed by bacterial pathogens The World Health Organisation (WHO) recently published a list of the greatest estimated threats of antibiotic resistance [25]. The list aligns closely with a list previous published by the CDC [23] and outlines 12 threats that pose critical, high, and medium clinical risks for patients. These threats exist across a variety of cultivable bacterial species, which demonstrates that antibiotic resistance is widespread within the bacterial kingdom.

Table 1.1. Main bacterial threats associated with antibiotic resistance [26]

The three threats categorised as critical, A. baumannii, P. aeruginosa, and ESBL producing Enterobacteriaceae, are all Gammaproteobacteria associated with carbapenem resistance and are typically associated with hospital infections. Recently, the urgency of the threat posed by Gram- negative bacilli has reached new levels as polymyxin resistance transmitting plasmids have been reported [27]. The polymyxins (colistin and polymyxin B) are considered the last line of defence against many Gram-negative bacilli. The spread of polymyxin resistance could therefore lead to a situation where large populations of patients cannot be treated by any available antibiotics [27].

1.5 The history of antibiotics discovery New antibiotics are clearly needed to alleviate the growing international antibiotic resistance crisis. Extensive screening for antibiotic has, however, already been carried out using a variety of approaches. Modern antibiotic discovery was spurred by mainly two developments. Firstly, Ehrlich set up a theoretical framework for the systematic screening of chemical compounds for antimicrobial activities against a target pathogen [28]. This framework led to the discovery of arsphenamine in 1909 [28]. Secondly, Alexander Fleming discovery of the ability of Penicillium notatum to inhibit growth of S. aureus showed that specific antimicrobial compounds could be found in the microbial world [29, 30]. Since then, two approaches have dominated the search for novel antibiotics (i) the Waksman platform and (ii) high throughput screening of compound libraries against identified bacterial targets [31, 32].

1.5.1 The Waksman Platform The foundation for the Waksman platform was laid when Ernest Chain and Selman Waksman continued Fleming’s work by systematically screening soil microbes for antibiotic secretion [5, 33].

The Waksman platform plates soil isolates onto agar petri dishes. Soil bacterial isolates that can inhibit neighbouring cell growth are selected for further analysis [30]. The focal point of the Waksman platform is the visual identification of antibiotic secretion through the observation of clearance zones (also labelled zones of inhibition) of adjacent bacterial strains (see Fig 1.1 C) [30]. In more developed forms the Waksman platform incorporates Ehrlich’s idea of targeting specific pathogens, so that fungal/bacterial isolates (producer strains) are screened against a known pathogen (reporter strain). This can happen either through cocultivation on a single petri dish or by spotting extract from the producer strain onto a growing lawn of a selected reporter strain [31, 32, 34]. The set up means that the platform does not determine the mode of action during the initial screening. Compounds are instead selected through their ability to inhibit growth of a target organism. The mode of action is determined after the compound has been identified and isolated.

Fig 1.1. Overview of the Waksman platform. A The Waksman platform consists of mixing soil/microbial isolates with a strain of interest to detect clearance zones. B Zoom of clearance zone: if the soil strain secretes metabolites that can kill the reporter strain, it creates visibly identifiable areas with no bacteria. C Photo shows clearance zones from Waksman’s paper that originally set out the protocol that became the basis for the Waksman platform [30].

The Waksman platform was the basis for what has been labelled the golden age of antibiotic discovery, and was responsible for the discovery of most present-day antibiotics [33]. The antibiotic discovery process was, however, mostly driven by pharmaceutical companies and much

18 of the generated data was consequently never standardised or published [31, 32]. The few published accounts from companies involved in antibiotic discovery during this period point to an exhaustive search of the microbial world [31, 32]. A large group of pharmaceutical companies even encouraged employees to collect soil samples when the employees were abroad [32]. Soil samples were plated out and bacteria were isolated, before being grown in a variety of growth conditions to test for secondary metabolites. The conditions ranged in nutrient composition, temperature, and state of fermentation (solid/liquid) [32]. While initial screenings focused around using Actinobacteria and especially Streptomyces species as producer strains, the later stages screened a broader range of bacteria for antibiotic/antifungal properties [32].

Over a period of 30 years 1000-10,000 natural products that had antibacterial or antifungal activity were identified, including 50-100 compounds that were ultimately approved for therapeutic use [31]. These include some of the most important classes of antibiotics (see Table 1.2).

Table 1.2. Important classes of antibiotics Classes of antibiotics Target Mode of action Examples

Sulfonamides Folate synthesis Bacteriostatic Sulfanilamide

β-Lactams Cell-wall synthesis Bactericidal Penicillins Cephalosporins Carbapenems

Aminoglycosides Protein synthesis Bactericidal Spectinomycin Kanamycin Neomycin

Tetracyclines Protein synthesis Bacteriostatic Tetracycline Doxycycline

Chloramphenicols Protein synthesis Bacteriostatic Chloramphenicol

Macrolides Protein synthesis Bacteriostatic Erythromycin Clarithromycin

Glycopeptides Cell-wall synthesis Bactericidal Vancomycin Teicoplanin

Oxazolidinones Protein synthesis Bacteriostatic Linezolid

Ansamycins RNA synthesis Bactericidal Rifamycin

Quinolones DNA synthesis Bactericidal Ciprofloxacin

Streptogramins Protein synthesis Bactericidal Pristinamycin

Table 1.2 shows 11 of the most prominent classes of antibiotics. The two classes in cursive font (sulfonamides and oxazolidinones) were synthetically derived. The nine other classes of antibiotics were discovered by screening soil microbes. Eight of these classes were first identified through the screening of bacteria, showing the significance of soil bacteria in antibiotic discovery. The order Actinomycetales has been especially significant, as it was the host order for the discovery of seven of the above mentioned 11 antibiotic classes. When a new antibiotic class was identified, it was used as a chemical scaffold to create new semi-synthetic antimicrobial alleles through chemical modification [35], thereby leading to a plethora of therapeutics. Over time the Waksman platform, however, started exhibiting diminishing returns, as an increasing proportion of hits were identified as already known compounds downstream [36, 37].

1.5.2 High throughput screening targeting known bacterial targets The diminishing returns of the Waksman platform led to pharmaceutical companies switching towards mode-of-action specific screenings [31]. As stated earlier, the Waksman platform screens for the effect on bacterial growth and the mode of action is only determined afterwards. The naturally derived antibiotics found using the Waksman platform, however, only targeted a few critical cellular functions. These functions include the synthesis of cell walls (e.g. β-lactams and polymyxins), RNA synthesis (e.g. rifampicins), and translation of mRNA into protein (e.g. aminoglycosides, macrolides, and tetracyclines) [38]. Fluoroquinolones, which were derived synthetically, had shown that different targets could be found (fluoroquinolones target chromosome supercoiling dynamics in replication and transcription) [5]. Antibiotic research consequently increasingly targeted mechanisms of action that were deemed vital in bacteria, but non-existing, non-essential, or widely divergent in mammalian cells [31]. Glaxo Klein Smith (GSK) for example, identified 70 such molecular targets. These targets were screened against libraries of natural products and synthetic compounds in high-throughput screens [31]. The high-throughput screens therefore reversed the order of identification of the Waksman platform, as compounds were identified as having effects on modes of action, before being screened against bacterial targets (see Fig 1.2). It is estimated that between 1996 and 2004 34 companies carried out 125

20 screens on 60 different bacterial targets [31, 39].

Fig 1.2. Overview of the steps in the Waksman platform compared to the steps in high- throughput screening.

GSK estimates that between 1995-2001, GSK alone ran 67 high throughput campaigns against antibacterial targets using a compound collection of between 260,000-530,000 compounds [31]. A literature review by Silver suggests that the combined effort yielded no new clinical approved antibiotics [37]. Silver instead argues that there has been no successful discovery of a novel class of antibiotics since 1987 [37].

The lack of success has been attributed to a variety of factors. From a chemical standpoint, it has been argued that the libraries used for screening of new antibiotics lacked sufficient quality and diversity [31]. As the Waksman platform is target neutral, looking at the natural product component of compound libraries; natural compounds that efficiently targeted critical pathways would most likely already have been discovered during the Waksman platform era [33]. Looking at the synthetic component of compound libraries; it has been argued that these compounds lacked the diversity of natural products [31]. According to this argument, corporate compound libraries tended to follow Lipinski’s rule of 5 [40], as this made it more likely that any identified hits could later be turned into a commercial drug. Lipinski’s rule of five states that a compound is less likely to work as orally deliverable drug when it has more than five hydrogen bond donors, 10 hydrogen bond acceptors, the molecular weight (MWT) is greater than 500 g/mol and the calculated Log P is greater than five [40]. Known antibiotics, however, represent a much larger space of chemical diversity [41], with more than 50 % of commercially available antibiotics not following Lipinski’s rule of 5 [31]. More damaging, later re-evaluations of the identified vital bacterial genetic targets showed that these targets often either had redundancy or were poorly conserved across species [31]. Additionally, when hits were identified, they often turned out to be non-specific membrane-active agents (detergents or uncouplers) [31]. Finally, bacterial membranes turned out to be a bigger obstacle than anticipated. Compounds that worked well in

21 vitro often turned out to be ineffective in vivo, as they either failed to penetrate the bacterial membrane systems or were actively pumped out by efflux pumps [3, 31].

1.6 Re-evaluating bacteria as a bioresource for novel antibiotics As stated above, the shift away from the Waksman platform and towards high-throughput screening of compound libraries was largely driven by the belief that bacteria had been exhausted as a biological resource for antibiotics [31, 36]. There is increasing evidence that this assumption is incorrect. There is increasing evidence that antibiotic secretion in nature is highly regulated and that antibiotic secretion is consequently contingent on specific stimulus [42]. This would mean that a high proportion of antibiotics could have been missed using the Waksman platform. This hypothesis has been supported by studies into secondary metabolites. It was found that when using standard laboratory techniques < 10 % of secondary gene clusters are expressed in sufficient quantities to be detected [43]. A similar study found that only three of the predicted 25 secondary metabolite gene clusters in Streptomyces avermitilis have been shown to be active in laboratory conditions [44, 45]. Latent gene expression is especially pertinent in the Streptomyces genus, where species generally have large genomes, of which, a high proportion of genes are associated with secondary metabolites [46]. Streptomyces rapamycinicus e.g. has 3.0 megabases (Mb). This constitutes 24 % of total 12.7 Mb genome and codes for 48 secondary metabolite gene clusters [47]. Furthermore, the Waksman platform mainly screened organisms that were readily available. Biomes that are not easily available to researchers have therefore been under sampled. Efforts to sample these biomes have had some success. The marine strain Verrucosispora MG-37 isolated from a deep-sea sediment was found to produce the novel compound abyssomicin, which exhibits antibacterial properties [48]. This showed that expanding the microbial biomes screened for antibiotics can lead to novel potential hits.

1.7 Uncultivable bacteria - metagenomic analysis of bacterial diversity in soil It has been known for more than a century that visual observations of bacteria in soil/water samples are 2-3 orders of magnitudes higher than subsequent colony forming unit (CFU) counts after incubation in laboratory conditions [49]. The 0.1-1.0 % of bacteria that form colonies are the bacteria that have been used as a resource for the Waksman platform. The 99-99.9 % of bacteria that cannot easily grow outside their native environment (uncultivable bacteria) remain largely unscreened for antibiotic production [33]. These uncultivable bacteria represent the largest

22 untapped bioresource for antibiotic screening. Estimating the biodiversity of uncultivable bacteria has been facilitated by metagenomics. Metagenomics encompasses the study of microbial communities without the need of a priori cultivation in the laboratory and combined with 16S rRNA (ribosomal RNA) sequencing, it has greatly increased our understanding of the bacterial kingdom. The 16S rRNA gene codes for the component of the 30S small subunit of prokaryotic ribosomes that binds to the Shine-Dalgarno sequence [50]. Because of the essential function of 16S rRNA in transcription, it is ubiquitous in the bacterial and archaeal genomes, although the number of copies vary [51]. The gene sequence of 16s rRNA is divided into nine highly conserved regions flanked by nine variable regions [52, 53]. The combination of highly conserved and variable regions makes the 16S rRNA sequence ideal for phylogenetic analysis. Primers can be designed to fit highly conserved regions and therefore bind across a high proportion of the bacterial kingdom [52]. Conversely, the hypervariable regions provide variations that make 16S rRNA sequences unique down to the species level. The level of variation between two species is correlated with their evolutionary distance and 16S rRNA sequence comparison can consequently be used to construct phylogenetic trees across the bacterial kingdom [53, 54]. 16S rRNA has therefore become the standard sequence for phylogenetic analysis of bacteria [55]. A sequence similarity of 98.2-98.7 % has been established as the upper threshold for new species [55, 56]. A species that has a sequence similarity of < 98.2 % to all other known species therefore likely constitute a new species. Similarly the taxonomic border for sequence similarity using 16S rRNA has been argued to be 94.5 % for genus, 86.5 % for family, 82.0 % for order, 78.5 % for class and 75.0 % for phylum [55]. High-throughput sequencing (HTS) platforms, such as Roche 454, Illumina MiSeq, and lately Oxford Nanopore Technologies’ MinION, have made it possible to estimate the composition of soil bacteria at much higher resolution than before [57]. The resulting data yields the composition of a biome at a resolution that is close to the species level [58]. Furthermore, the above-mentioned platforms can accurately estimate the relative abundance of different phyla. As the cost of metagenomic sampling has fallen, bacterial community data has accumulated. The results show that that uncultivable bacteria represent a biological resource that is phylogenetically significantly different from cultivable bacteria. It has been shown that a gram of rhizosphere soil contains somewhere between 107-109 bacteria and possibly 106 distinct taxa [59, 60]. In their recent study, Delgado-Baquerizo et al. estimate that Proteobacteria is the most abundant phyla globally, making up about 30-40 % of all soil bacteria. Other dominant phyla include Actinobacteria (25-30 %) and Acidobacteria (10-15 %). Planctomyces, Chloroflexi, Verrucomicrobia represent > 3.0 % of bacterial

23 abundance and Bacteroidetes, Gemmatimonadetes, and Firmicutes represent > 1.0 % of bacterial abundance [61]. The above study largely confirmed earlier estimates of the global distribution of bacterial phyla [62]. Breaking down the data, the nine phyla mentioned above represent > 85 % of global bacterial abundance with the < 15 % of global bacterial abundance consisting of another 30- 80 phyla [55, 63, 64]. Locally, several factors can influence biodiversity profiles. Soil pH has been shown to be the most significant variable in determining bacterial biodiversity, with neutral pH soil producing the greatest taxonomic diversity [65, 66]. Other indicators include temperature, soil texture [67], nitrogen availability [68], and water concentration [68]. Comparing water concentration, non-dessert soils tend to have higher species diversity. Of particular interest to this project, soils with a lower risk of drought have been shown to have higher concentrations of antibiotic producing gene clusters - possibly due to higher interspecies competition [69]. Reversely, desert species are less likely to have antibiotic gene clusters, but more likely to have genes associated with dormancy and heat stress [58, 69, 70]. Finally, 16S rRNA metagenomic sequencing has shown that spatial separation of 1 cm can create significantly different local community profiles, showing that comprehensive sampling is needed to capture the biodiversity of a local biome [67]. Comparatively 16S rRNA analysis indicates that 30 bacterial phyla [54, 71] and three of 26 archaeal currently recognised phyla have cultured representatives [72]. Furthermore, at present 90 % of all cultivated bacteria are affiliated with just four phyla; Proteobacteria, Firmicutes, Actinobacteria, and Bacteroidetes [72]. Many phyla, while common in nature, consequently, have no model representatives.

The compounded effect of < 1.0 % of all bacteria being cultivable in pure culture and < 10 % of secondary metabolites of cultivable bacteria being expressed in laboratory conditions means that only a very small fraction of the total secondary metabolites has been screened. Given that antibiotics make up a significant proportion of already captured secondary metabolites, accessing the larger proportion of secondary metabolites that are yet to be screened could lead to the discovery of a range of novel antibiotic classes. Several distinct approached have been explored to access these unscreened secondary metabolites. These approaches can be separated into culture- independent approaches (functional metagenomic approaches) and development of novel cultivation platforms.

1.8 Uncultivable bacteria - metagenomic functional analysis of secondary metabolites in soil Genomes can be accessed via culture-independent genome sequencing using environmental samples. While marker gene studies such as 16S rRNA can give highly accurate information regarding the taxonomy of an ecosystem, reconstructing genomes from environmental samples still poses significant difficulties. Assembly difficulty is highly correlated to the ratio of the sequencing read length and the length of repeats [73, 74]. In bacteria, intragenomic repeats are generally limited in length ( <10,000 bp [75]). Intergenomic repeats can, however, be nearly the entire chromosomes for closely related strains [73]. The high level of species abundance in soil microbiomes therefore results in an increased chance of false alignments, which leads to an increased rate of difficulty in constructing long genome contigs from metagenomic data derived from soil [76]. Howe et al. [77] e.g. generated 1.8 and 3.3 billion reads from two soil metagenomes. From these reads the maximum length of contigs they could assemble were 20,234 bp and 2,579 bp respectively, showing the difficulty in reconstructing genome sequences from shotgun data from short read platforms [77]. The present level of sequence depth and available bioinformatic approaches means that species with a relative presence of <1.0 % are often impossible to reassemble correctly from environmental samples [76, 78]. High-quality assembly of bacterial genomes using long read platforms, such as the PacBio RSII/Sequel and Oxford Nanopore MinION offers new possibilities, but have not been widely adopted due to cost, DNA quality requirements, and complexity of DNA preparation [73].

Another solution is single-cell genomics. Creating complete genomes from a single cell is, however, technically challenging, as amplification bias often leads to uneven coverage and fragmented assemblies [79, 80]. Nevertheless, using metagenomic information to carry out targeted functional investigations can still lead to valuable information. Spang et al. [81] e.g. used partial uncultured archaeal genome sequences derived from metagenomic data from marine sediments to provide strong evidence for the hypothesis that eukaryotes evolved from a bona fide archaeon. Another recent example is the culture free platform developed by Hover et al. [82] to detect antibiotics in soil microbes. The Hover et al. study shows many of the present possibilities and limitations of using metagenomic data for antibiotic discovery. Using the Asp-X-Asp-Gly motif, which is conserved in Calcium-dependent antibiotics, as the focus of their investigation, Hover et al. used sequence-guided screens of metagenomic data to detect similar secondary metabolite gene clusters [82]. Heterologous expression of the identified gene clusters allowed for testing of the

25 antimicrobial activity of the resulting natural products [82]. The study led to the discovery of malacidins, a distinctive class of calcium dependent antimicrobials that shows activity against the cell wall of multidrug-resistant Gram-positive pathogens, including methicillin-resistant Staphylococcus aureus [82]. The Hover et al. study was, however, based on motive similarity, something that cannot be assumed to exist for all novel antibiotic classes. Novel secondary metabolite clusters coding for new antibiotics might therefore not be characterized as such using in silico methods. Furthermore, the use of heterologous expression to test the function of identified gene clusters is problematic as the evolutionary distance of the natural host organism and the synthetic host organism has a significant effect on expression viability [78]. Gabor et al. [83] estimated that 60 % of the genes from 32 different bacterial and archaeal genomes contained gene expression signals that could not be recognized in E. coli. Missing co-factors and chaperones will lead to a further reduction of gene expression viability [84]. By genetically modifying the host organism, it has been possible to increase the metabolite space that can be explored using a single heterologous host organism [85]. Liebl et al. [86] engineered the EPI300 E. coli strain of the CopyControl kit from Epicentre to express additional sigma factors from Clostridium and Streptomyces, which resulted in an increase of 20-30% of heterologous metabolites derived from metagenomic data that could be expressed. Nevertheless, the expression of uncultivable secondary metabolite gene clusters is still restricted by the phylogenetic diversity of model organisms that have been investigated and the number of host organism that have been developed for heterologous expression [85]. Finally, antimicrobials pose a unique problem for heterologous expression, as antibiotics by definition are toxic to bacteria. Genes for the synthesis of antibiotics are therefore hard to induce and express in heterologous bacterial hosts, as the host cannot be assumed to have innate immunity to the non-native antibiotic, especially when potentially exposed to broad-spectrum antibiotics.

1.9 Cultivating unculturable bacteria – understanding the required growth conditions Cultivation platforms for uncultivable bacteria offer a promising path to accessing undiscovered classes of antibiotics. The causes that lead to “uncultivability” in different strains and phyla are, however, diverse and underexplored [87]. One leading explanation of uncultivability in standard media is the difference between nutrient compositions of standard laboratory media and soil. High nutrient laboratory conditions favour classical cultivable bacteria, which tend to be copiotrophs, which means that they can significantly increase their growth rate in response to high nutrient

26 conditions [88]. Soil on the other hand has low nutrient conditions, and as cell division is energetically expensive, most soil bacteria are oligotrophs [89, 90]. Up to 80 % of soil bacteria consequently need > 2 months to grow to the microcolony stage (15-20 members per colony) and often fail to grow in high nutrient conditions [91, 92]. For marine bacteria, the equivalent reported number is 99 % [93]. As a result, oligotrophs tend to be overlooked in classical cultivation studies [90, 94, 95]. Furthermore, many uncultivable strains have strict maximum population sizes. Growth rates in these uncultivable strains decrease as population sizes increase, and the cells seize dividing at total colony populations of <103 cells. The low maximum population number necessitates microscopic analysis for detection [93]. At first glance, a maximum population could seem a strange evolutionary response. In their natural environment bacteria are, however, subject to strong predation pressures by eukaryotes both in aquatic and soil environments [96]. After detecting a bacterial colony, nematodes can stay in an area until the entire bacterial population has been consumed. Mechanisms to ensure strict maximum population size have thus been postulated to minimize the risk of detection [97]. Furthermore, the complex ecosystems observed in surface soils lead to a high level of niche differentiation [98], which in turn leads to specific growth requirements. Members of the genera Abiotrophia and Granulicatell, for example, need pyridoxal or L-cysteine for growth [91]. Dependent on niche, many uncultivable bacterial strains therefore have specific requirements for pH, nutrition, temperature, soil moisture, and access to oxygen - with different uncultivable bacteria thriving in different conditions. Finally, the growth of some uncultivable strains is conditioned on the presence of certain growth factors [99]. When exposed to foreign environments, uncultivable strains can enter temporary states of low metabolic activity that is often mistaken for a lack of viability [99, 100]. Reversely, when exposed to certain familiar “bacterial cytokines” dormant bacterial strains can be coaxed to start growing [101]. Often these growth factors are secreted by neighbouring species. Nichols et al. [99] found that one isolate, MSC33 could only be cultured in standard Petri dishes when another local marine isolate MSC105 was present. They therefore categorised MSC105 as a “helper strain”. These helper strains are based on local conditions, and a helper strain will thus not necessarily facilitate the cultivation of strains outside its immediate ecosystem [99].

1.10 Cultivating unculturable bacteria – existing methods Many of the above factors have been incorporated into cultivation platforms in attempts to cultivate more representatives of uncultivable phyla. The developed platforms to culture and

27 domesticate a broader proportion of the bacterial kingdom can largely be divided into three categories; i) change in cultivation conditions in monoculture experiments, ii) cocultivation, and iii) in situ cultivation [102].

1.10.1 Monoculture experiments Changes in cultivation conditions attempt to cultivate monocultures in laboratory conditions in ways that more faithfully mimic the natural conditions of uncultivable bacteria. The change in conditions span a range of parameters. Köpke et al. [103] e.g. exposed samples from similar environments to changes in pH, nutrition, temperature, soil moisture, and access to oxygen. It was found that each parameter change lead to unique populations of growing uncultivable bacteria [103]. Another successful development in culturing uncultivable bacteria has been the shift away from nutrient rich-media towards more nutrient-poor media. Using low nutrient media has allowed oligotrophs to become more competitive [104]. It is important to note that the focus on slow growers also created changes to the experimental set-up, as incubation periods had to be increased substantially. Incubation times suggested in the literature include 8-12 weeks [88, 105] and 24 weeks [106]. In one study, incubation with low nutrient media over extended incubation times led to the isolation of new members of the phyla Acidobacteria, Gemmatimonadetes, Chloroflexi, and Planctomycetes [105]. Adding complexity to the media is another way to more faithfully mimic natural conditions, which in turn leads to higher cultivation rates. Uphoff et al. [107] found that when cultivation efforts were carried out with one carbon source, cultivated isolates from natural environments belonged almost exclusively to the Gammaproteobacteria [107]. In contrast, experiments including several carbon sources saw growth of four additional phyla [107]. Consequently, media such as SMS or R2A, which incorporate complex components such as malt extract or yeast, have gained prominence among microbiologists attempting to cultivate unculturable bacteria [108]. Using such media sources take away the knowledge of the exact chemical composition but has been justified by the increase in cultivability achieved through higher similarities with natural environments. A similar argument underlies the use of environmental extracts of natural environment as growth medias. Using seawater extract as growth media, it was possible to cultivate the previously uncultivable Alphaproteobacteria clade SAR11, which is a widespread group of free living bacteria found predominantly in seawater [109]. Similarly, in a study by Zengler et al. [109], seawater used as growth media facilitated the cultivation of previously uncultivated phylotypes within the Planctomycetales. Another promising

28 avenue for increased cultivation has been the addition of signalling molecules. The additions of the signal molecules cyclic AMP (cAMP) and acyl homoserine lactones were shown to significantly increase the cultivation rate of marine bacteria [110].

1.10.2 Cocultivation platform for uncultivable bacteria The ability of signalling molecules to promote growth of uncultivable species could be a result of synergies within bacterial communities. Mixed microbial populations have been shown to perform multiple-step functions, that are often not possible for individual species [102]. Breaking down cellulose from complex organic matter e.g. often involves community collaborations to be carried out [111]. While signalling factors have had some success in stimulating growth in uncultivable bacteria, syntrophic associations are often compulsory and this interdependence cannot easily be by-passed or suppressed by the addition of factors to the media [102]. Cultivability of some species will therefore be dependent of growth factors produced by neighbouring organisms in natural ecosystems. The most well-explored example in uncultivable bacteria is the need for external siderophore production for metal uptake. Vartoukian et al. [87] used this dependence to culture novel bacterial strains by selecting helper strains known to produce siderophores. This led to isolation of novel Bacteroidetes and Peptostreptococacceae bacteria [87]. Similarly, D'Onofrio et al. [112] showed that a Micrococcus luteus helper strain could induce the growth of the otherwise unculturable isolate Maribacter polysiphoniae. Further experiments revealed that it was the production of various siderophores, which were essential for in vitro cultivation. In the same study, D’Onofrio et al. showed that 10 % of all culturable isolates benefited from being cultivated with a random member from their community, which shows that cocultivation might boost the cultivability of uncultivable bacteria even without prior knowledge of the molecular mechanism of interdependence [113]. The vast amount of species within a biome, however, means that exploring all possible interdependencies within a bacterial network remains unfeasible in wet lab conditions. Instead, cocultivation experiment often require randomly picked pairs of colonies [112]. Ferrari et al. developed a cocultivation platform that allows for the simultaneous cocultivation of multiple species [88, 114]. In their studies, bacterial cultures were diluted in water and subsequently plated on the polycarbonate membranes, using soil extract as media. The cocultivation platform allowed for the detection of the live microcolonies on the membrane using viability staining. The colonies identified almost exclusively exhibited microcolony growth patterns, meaning that many species could be in close proximity to each other, yet remain physically discrete [88]. A fraction of these

29 microcolonies could subsequently be isolated and domesticated over time [114]. Domestication, here, refers to the process of conditioning uncultivable bacteria to laboratory conditions, with the goal to grow them in pure cultures. The purpose of domestication is to allow novel cultivated species to be analysed using modern molecular techniques. The Ferrari et al. cocultivation platform resulted in the isolation of a species belonging to previously uncultured clade TM7 that could then be grown in pure cultures [88].

1.10.3 In situ cultivation platforms In situ cultivation platforms isolate bacteria in individual compartments and cultivate them in their natural environment. The loading mostly happens through stochastic isolation with an average of 0.1-10 bacteria per compartment [115]. Once loaded, the compartments deny the influx/outflux of other microbes but facilitate passive diffusion of nutrients and metabolic compounds. Several platforms have been established to accomplish this goal. Aoi et al. [116] developed a hollow-fibre membrane chamber (HFMC) platform for the in situ cultivation of environmental microorganisms [116]. The HFMC system consisted of 48 to 96 semi-permeable (0.1 μm mean pore size) hollow- fibre membrane tubes connected to injectors. Tubes were placed in the natural environment of the isolated bacterial strains and 0.1-0.5 cells were injected into each tube. The platform was left for up to one month, before cultures were analysed using 16S rRNA sequencing. The results showed considerable growth of Bacteroidetes, Nitrospira, and Verrucomicrobia among other phyla inside the hollow-fibre membrane tubes. The experiment did not address whether the protocol could be used for domestication [116]. In a separate in vivo cultivation experiment Ben Dov et al. [117] cultivated microorganisms by encapsulating them in polysulfone-coated agar spheres. The spheres were incubated in the natural environment of the encapsulated species [117]. The spheres allowed for in vivo cultivation of novel species including Bacteroidetes and Planctomycetes, but subsequent attempts to domesticate the species and grow them using standard cultivation techniques largely failed [117]. Both platforms therefore allowed for isolated cultivation of bacteria in vivo, which facilitated further genetic analysis of secondary metabolites downstream. Importantly, neither platforms showed viability as domestication platforms. An alternative approach that explicitly focused on domestication was developed in a series of papers by the collaborators Epstein and Lewis [93, 108, 118]. The most recent method (the “Isolation chip”, or “Ichip”) by Epstein and Lewis isolates soil bacteria and suspends them in agar (see Fig 1.3) [108]. The agar solution is then suspended in individual 1.25 µL growth chambers with an average

30 of 1 bacterium per growth chamber. Each growth chamber (isolation chamber) is in turn part of a 384-growth-chamber Ichip, which allows high-throughput cultivation. After suspending bacteria in the growth chambers, the chip is sealed by 0.03 μm semi-permeable membranes that allow metabolite and nutrient exchange but prevent bacterial penetration [108]. The Ichip is buried in the native environment of the soil bacteria for 14-28 days before being retrieved [108]. The Ichip has been reported to achieve cell population growth in up to 50 % of the chambers [108], with similar populations showing 0.5 % growth on standard petri dishes. The Ichip has consequently been reported to allow for up to 100-fold increase in cultivability [108]. After retrieval, isolates can be plated out in standard laboratory conditions, passaged for further rounds of Ichip domestication attempts, or become part of coculturing experiments [99, 108, 118]. Nichols et al. [108] provided a breakdown of sequenced phyla, which showed that when applied in soil, the Ichip was able to domesticate two phyla (Proteobacteria or Firmicutes). The breakdown of phyla therefore suggest that the Ichip only boosts the cultivation of limited number of phyla.

1. Soil sample diluted in water 6. Ichip is buried in situ in the ground where soil sample originated, for ~2 weeks

2. Plastic Ichip dipped into diluted soil sample 7. Ichip is returned to laboratory and individual growth chambers checked for microcolonies (~ 50 % expected) 3. On average a single bacterial cell enters a 8. Microcolonies are plated and assayed for single hole/growth chamber in the Ichip laboratory growth and scale up

4. Individual growth chambers within the Ichip 9. Growing strains are up-scaled and secondary are sealed with a porous membrane metabolites extracted

5. Porous membrane allows nutrient/metabolite 10. Metabolites screened for antimicrobial exchange but does not allow entry or exit of properties bacteria to the growth chambers

Fig 1.3. Essential steps in antibiotic screening using the Ichip procedure. Figure modified from Nichols et al. [108].

1.11 Using uncultivable bacteria as a resource for novel antibiotics The Ichip domestication platform allowed for the construction of a library of uncultivable bacteria, which could then be screened for antibiotic production in a variety of media using the Waksman platform [118]. In one such study Ling et al. screened extracts from 10,000 isolates for antimicrobial activity on plates overlaid with S. aureus [118]. Extract from a new species of

Betaproteobacteria, provisionally named Eleftheria terrae, yielded the compound Teixobactin [118]. Teixobactin has since been shown to be a novel class of antibacterial compounds, which has a novel mechanism of action that exhibits dual activity against firstly the highly conserved motif of lipid II (a precursor of peptidoglycans) and secondly of lipid III (a precursor of cell wall teichoic acid) [118]. Teixobactin has shown efficiency against multi-resistant Gram positive bacteria, including Mycobacterium tuberculosis (M. tuberculosis) and is therefore a candidate for commercialization [119]. Teixobactin is one candidate discovered using the Ichip [120, 121] and analogues are also being developed [122].

Using non-cultivable bacteria as an upstream resource for the Waksman platform has therefore shown great promise. As discussed in Section 1.6, however, < 10 % of secondary metabolites are secreted in laboratory conditions. Furthermore, the domestication rate is lower than the cultivation rates in in situ platforms, meaning that the transfer of bacteria to the laboratory leads to a loss of biodiversity [117, 118]. Antibiotic screening in situ should therefore theoretically lead to a higher capture rate of biodiversity as local growth factors increase cultivation, and local stressors are expected to increase the range of antibiotics secreted.

1.12 Biosensors as in vivo antibiotic secretion detectors One way to measure antibiotic secretion in situ is via optical biosensors. Biosensors have been constructed using either fluorescent markers or luminescent markers [123, 124]. Luminescence provides an immediate indicator of metabolic activity since it is an ATP-dependent process [125]. Furthermore, Ritchie et al. [126] have developed luminescent biosensor strains of a non-toxigenic strain of E. coli O157:H7 which is chromosomally lux-marked (luxCDABE). The strain remains viable and luminescent over time in challenging conditions. In a survival study, starved populations of E. coli O157:H7 Tn5 luxCDABE remained viable and luminescent after being placed in a natural environment for > 2 weeks - albeit with a lower cell population and lower luminescence per cell [126]. Such biosensors could therefore be placed in compartments in in situ conditions to screen for antibiotic production. If antimicrobial concentrations reach MIC around the biosensor the luminescence will switch off. The lack of luminescence can then be detected downstream using luminescence detectors (for further discussion see Chapter 7).

Chapter 2. Aims

This project aims to develop a sensitive high-throughput platform for the in situ screening of uncultivable bacteria for antibiotic production (see Fig 2.1). To create such a platform, the aim is to merge the Waksman platform with concepts from high throughput in situ screening methods (see Fig 2.1). Each in situ compartment will encompass two species – an isolated soil bacterium and a reporter strain of choice. The cocultivation should thereby stimulate an increase in antibiotic secretion from the soil organisms. Local stressors should further increase the rate of antibiotic secretion. After an incubation period the ISSA will be retrieved. If the reporter strain is still alive, it will emit luminescence. Alternatively, if the soil bacterial strain has secreted a sufficient quantity of antimicrobials, it will have killed/inhibited the reporter strain, and bioluminescence will have ceased. Soil bacteria capable of killing/inhibiting the reporter strain will be domesticated and their metabolites analysed. The in situ detection of antibiotic producers means that domestication efforts can be concentrated on antibiotic producers, leading to substantial efficiency gains compared to other domestication platforms.

Fig 2.1. Overview of the theoretical differences between the Waksman discovery platform and the ISSA platform.

The workflow of the platform is therefore to firstly identify the bacterial population within environmental samples. Secondly, individual soil bacteria are isolated next to a chosen reporter strain before being incubated for an extended period in the natural environment of the soil bacteria. Thirdly, the ISSA platform is retrieved, and bacterial colonies that have secreted antibiotics are isolated, cultivated, and identified phylogenetically. Fourthly, the antimicrobial compound is isolated and characterised. This PhD project will focus on step 1-3. To enable antibiotic discovery the thesis chapters will aim to develop the required individual components chapter for chapter (see Fig 2.2).

Fig 2.2. Overview of the chapter development of the ISSA platform. The following components had to be developed for the ISSA platform to be successful:

1) Develop protocol for staining, quantifying and isolating soil bacteria from soil (Chapter 4). 2) Develop an optimized physical platform for the in situ screening of uncultivable bacteria for antibiotic secretion (Chapter 5). 3) Test the ability of the developed in situ domestication platform to domesticate uncultivable strains (Chapter 5). 4) Build a library of uncultivable bacteria (Chapter 5). 5) Develop clinically relevant model reporter strains with strong and consistent emission of luminescence (Chapter 6). 6) Develop luminescent reporter strains for downstream analysis of antibiotic capabilities of identified producer strains from the ISSA platform (Chapter 6). 7) Develop a model organism for antibiotic secretion (Chapter 7). 8) Use the developed model organisms to adjust the ISSA development for detecting antibiotic secretion (Chapter 7). 9) Start antibiotic discovery using the developed ISSA platform (Chapter 8). 10) Identify isolated bacteria that secrete antibiotics in vitro and in vivo (Chapter 8). 35

Chapter 3. Materials and methods

3.1 Materials

3.1.1 Media Table 3.1 details the growth media used in this thesis. All medias had their pH adjusted to 7.2 using a hydrochloric acid solution and sodium hydroxide solution when they deviated from the pH range 7-7.4. All medias were prepared as either liquid, solid with 2.0 % agar, or solid 3.0 % with ultra-low gelling temperature agarose (ULGTA), unless stated otherwise. When media was either liquid or solid using 2.0 % agar, the media was sterilised by autoclaving at 121°C for 15 mins, under a pressure of 2 bars. When using ULGTA, the media was first autoclaved as stated above. The media was then allowed to cool, before ULGTA was added. Subsequently, 200 ml of the solution was poured into 500 ml flasks and heated in the microwave at 300 W until boiling. The flask was then further heated for 5 min in the microwave at 300 W. When required, additives such as antibiotics were added at appropriate concentrations. These were sterilised by passing through a 0.2 μm polyethersulphone membrane syringe filter.

Media Composition per litre Luria-Bertani 10.0 g Tryptone (Pancreatic digest of casein), 5 g Yeast extract, 5.0 g Sodium (LB) chloride (supplier: Sigma Aldrich) 2xTY 16.0 g Tryptone (Pancreatic digest of casein), 10 g Yeast extract, 5.0 g Sodium chloride (supplier: Melford Ltd) GYM 4.0 g Glucose, 4.0 g Yeast extract, 10.0 g Malt extract MYM 4.0 g Maltose, 4.0 g Yeast extract, 10.0 g Malt extract YM 4.0 g Yeast extract, 10 g Malt extract MYM + CaCO3 4.0 g Maltose, 4.0 g Yeast extract, 10.0 g Malt extract, 3.0 g Calcium Carbonate SMS 0.125 g Casein digest, 0.1 g Potato starch, 1 g Casamino acids R2A 0.5 g Proteose peptone, 0.5 g Casamino acids, 0.5 g Yeast extract, 0.5 g Dextrose, 0.5 g Soluble starch, 0.3 g Dipotassium phosphate, 0.3 g Sodium pyruvate, 0.05 g Magnesium sulfate YEME 3.0 g Yeast extract, 5.0 g Bacto-peptone, 3.0 g Malt extract, 10.0 g Glucose, 170.0 g Sucrose ISP4 10,0 g Soluble starch,1.0 g Dipotassium phosphate, 1 g Sodium chloride, 2.0 g Ammonium sulfate, 2.0 g Calcium carbonate, 0.001 g Ferrous sulphate, heptahydrate, 0.001 g Manganous chloride, 0.001 g Zinc sulphate 7H9 0.5 g Ammonium sulfate, 2.5 g Disodium phosphate, 1.00 g Monopotassium phosphate, 0.10 g Sodium citrate, 0.05 g Magnesium sulfate, 0.0005 g Calcium chloride, Zinc sulfate, 0.001 g Copper sulfate, 0.04 g Ferric ammonium citrate, 0.50 g L-Glutamic acid, 0.001g Pyridoxine, 0.0005 g Biotin (Supplier: Appleton Woods)

Tet opt 10 g Sucrose, 6 g Citric acid, Ammonium sulfate, 0.25g Magnesium sulfate, 0.015 synthetic g Monopossium phosphate, 1.25 g Calcium carbonate, trace elements. media 7H10 0.025 g Magnesium sulfate, 0.04 g Ferric Ammonium citrate, 0.4 g Sodium citrate, 0.5 g Ammonium, 1.5 g Monosodium Glutamate 1.5 g Disodium Phosphate, 1.5 g Monopotassium Phosphate, 17.0 g Agar, 0.85 g Sodium Chloride, 2.0 g Glucose, 5.0 g Bovine Albumin V, 0.004 g Catalase Pyridoxine, 0.001 g Zinc Sulfate, 0.001g Copper Sulfate, 0.0005 g Biotin, 0.0005 g Calcium Chloride, 5.0 g Glycerol, 0.05 ml Oleic Acid

3.1.2 Bacterial strains Species Supplier/identification number Streptomyces venezuelae Donated by M. Buttner (JIC) Streptomyces aureofaciens LMG 5968 Streptomyces fradiae DSM 40063 Streptomyces cattleya DSM 46488 Streptomyces griseus LMG 5974 Caulobacter cresentus Donated by M. Beeby (Imperial College London) Bordetella bronchiseptica Donated by M. Beeby (Imperial College London) Escherichia coli (k12) DSM 498 Myxococcus xanthus Donated by M. Beeby (Imperial College London) Amycolatopsis orientalis DSM 40040

Terriglobus roseus DSM 18391 Bryobacter aggregatus DSM 18758 Gemmata obscuriglobus DSM 5831 Luteolibacter arcticus DSM 102244 Gemmatimonas aurantiaca DSM 14586 Bacillus subtilis Mini Bac Donated by J. Altenbuchner (U. of Stuttgart) [127] Bacillus subtilis 168 DSM 402

Mycobacterium smegmatis ATCC 700084 E. coli S17-1 λpir mini-Tn5 Donated by Ken Forbes (U. of Aberdeen) [126] One Shot® TOP10 cells Thermo Fisher

LuxCDABE bioreporter strains Developed/Donated Pseudomonas putida P1H2 This study Pseudomonas flourescens FH1 This study Mycobacterium smegmatis G13 This study Mini Bac Pveg luxCDABE This study Bacillus subtilis Pveg luxCDABE This study E. coli O157:H7 RA1 Donated by Ken Forbes (U. of Aberdeen) [126] E. coli O157:H7 RA2 Donated by Ken Forbes (U. of Aberdeen) [126] E. coli O157:H7 GC1 Donated by Ken Forbes (U. of Aberdeen) [126]

E. coli O157:H7 GC2 Donated by Ken Forbes (U. of Aberdeen) [126] Pseudomonas putida luxCDABE Donated by Ken Forbes (U. of Aberdeen)[125] Pseudomonas fluorescens luxCDABE Donated by Ken Forbes (U. of Aberdeen) [125]

3.1.3 Vectors Name Supplier Pveg (promotor) Addgene (plasmid # 55173) [128] pBS3Clux Addgene plasmid (# 55172) [128] PlepA (promotor) Addgene plasmid (# 55175) [128] pUT mini-Tn5 luxCDABE Donated by Ken Forbes (U.o.A) [125] pBS-Int Addgene plasmid (# 50000) [129] pMV306DIhsp+LuxG13 Addgene plasmid (# 49999) [129]

3.1.4 Primers Name Sequence 27F-YM AGAGTTTGATYMTGGCTCAG 907R CCGTCAATTCMTTTGAGTTT 907R-A CCGTCAATTCATTTGAGTTT 907R-C CCGTCAATTCCTTTGAGTTT 1391R GACGGGCGGTGWGTRCA 1492R AAGTCGTAACAAGGTAACCG

3.1.5 General lab reagents and kits Name Supplier Agarose, ultra-low gelling temperature, molecular biology grade Sigma-Aldrich Molecular Biology Grade Agarose, 500 g Appleton Appleton Woods Ltd Taq DNA Polymerase, native Life Technologies Ltd Q5 High-Fidelity DNA Polymerase New England Biolabs KOD Hot Start DNA Polymerase Merck 2XYT BROTH, POWDER Melford Ltd Wizard Plus SV Minipreps DNA Purification System Promega Wizard SV Gel and PCR Clean-Up System Promega Picolourlock Gold Innova Biosciences

3.1.6 General materials Name Supplier Breathe-Easy® sealing membrane Sigma-Aldrich SealPlate® film sterile Sigma-Aldrich CELLSTARTM OneWell PlateTM Greiner Bio-One Whatman® Nuclepore(TM) Track-Etched Membranes L X W 8 in. x 10 Sigma-Aldrich in., pore size 0.03 mum Company Grade 2 Titanium (0.5 mm) Titek Ltd FILTER SYRINGE MINISART 0.2 μm Appleton Ltd Syringe filter 5.0 µm, Sartorius VWR International

Corning™ Falcon™ Test Tube with Cell Strainer Snap Cap (35 μm filters) Sigma-Aldrich 1mm acetal sheets theplasticshop.co.uk Spreaders, L-shaped, disposable VWR International Silicone Rubber sealant clear, 50 ml tube RS components Ltd Bead Glass Soda Lime Glass 2mm WVR International 0.5mm (24 AWG) - Grade 1 - Titanium Wire - Vacuum Annealed - 10 The Crazy Wire Metre Spool. Company

Corning 96 Well Flat Clear Bottom Black TC-Treated Microplates Individually Wrapped with Lid Sterile Sigma Aldrich

3.1.7 Machines and microscopes Name Supplier MPS Flexible (laser) Rofin Ltd ILS12.75 (laser) engraversnetwork.com Echo 555 Liquid Handler Labcyte Ltd BD FACSARIA III (FACS sorter) BD BD Fortessa (FACS analyser) BD Zeiss Axio Observer (microscope) Zeiss Axiocam ICc 5 (microscope camera) Zeiss ChemiDoc™ MP Imaging System (imaging system) Bio-Rad CLARIOstar (plate reader) BMG LABTECH Gene Pulser XcellTM Electroporation System Thermo Fischer

3.1.8 Software Name Version Supplier Flowjo v7 Flowjo, LLC Fiji (ImageJ) 2.00- rc-43/1.51p [130] Diva 8.0.1 BD Biosciences Office 365 2016 windows Microsoft Illustrator CC 2019 Adobe Image Lab 6.0.1 Bio-Rad MARS Data Analysis Software V5.40 R3 Bmglabtech Reader Control Software V3.32 Bmglabtech

3.3 Methods

3.3.1 Growth of bacteria for quantification of proxy species in soil Bacteria were grown in pure isolates, in their recommended growth media, until they reached stationary phase. 1 ml of the resulting culture was mixed with 5 ml of SMS and then poured onto 50 g of autoclaved soil in 250 ml a sterile glass flask. The glass flask was shaken manually and then incubated for 3-5 days at 28°C.

3.3.2 Staining bacteria in soil with CFDA-SE/water and CFDA-SE/EDTA 1 g of moist soil was resuspended in 10 ml of SMS media. The solution was shaken vigorously for 2 min. After shaking, the solution was filtered using 35 μm filters (Corning™ Falcon™ Test Tube with Cell Strainer Snap Cap) or 5 μm (Syringe filter 5.0 µm, Sartorius) if specified. The filtered solution was mixed with sterile water or 0.5 M EDTA solution in a 1:1 ratio. 1 ml of the solution was then mixed with 5 mM CFDA-SE (Vybrant® CFDA SE Cell Tracer Kit) and shaken at 220 rpm at 28°C for 20 min. The solution was placed under a microscope or quantified using flow cytometry dependent on purpose.

3.3.3 Preparing CFDA-SE stained bacteria for microscopy 50 μl CFDA-SE/EDTA stained soil solution/CFDA/SE-EDTA was mixed with 50 μl 1.0 % ULGTA (37°C) and vortexed. 2 μl of the suspension was then pipetted onto a microscopy slide and placed on a Zeiss Axio Observer microscope. Images were taken using 40x magnification. Phase contrast and the GFP fluorescent channel were used. For images with fluorescence, the laser intensity was set at 17 %. The camera used, was the Axiocam ICc 5. The images were analysed in FIJI and contrast and brightness settings were set at min/max = 300/1500 AU for fluorescent images of soil and 400/3000 AU for pure cultures fluorescent images.

3.3.4 Quantification of bacteria stained using flow cytometry 5 μl of CFDA-SE stained sterile soil solution was mixed with 995 μl sterile water. The solution was then added to a Fortessa analyser. 125,000 events were run in each experiment and each experiment was run in triplicate. The data was analysed using forward scatter (height, threshold = 200 volts) and the green excitation laser (487 nm), with captured emission at 530 ± 15 nm (height). A threshold gate was drawn, so that 0-2 bacteria per 125,000 events were inside the gate in the three replicates. Having established the threshold gate, 5 μl of stained soil solution containing bacteria was mixed with 995 μl sterile water and the solution was added to a BD Trucounttm Tubes (catalogue number 34033). 1000 Trucount beads were counted by the flow analyser by separating the Trucount beads from the soil solution using the 780_60 YG-H (height) channel and the 710 ± 20 nm (height) channel.

When single cell sorting took place, an ARIA III sorter was used. No counting beads were added to the solution. Events were single cell sorted into standard 96-well plates (96 Well Flat Clear Bottom Black Microplates) containing 200 µl of the appropriate growth media.

3.3.5 Machining ISSA arrays 0.5 mm grade 2 titanium sheets were used as base material for the array and protective plates. Outer frames were made of 3 mm stainless steel. All designs were designed in AutoCAD, before being lasered using the MPS Flexible laser. 384-compartment arrays had compartments with 0.6 mm ø unless stated otherwise. 5180-compartment arrays had compartments with 0.3 mm ø.

3.3.6 Accuracy testing of FACS Sorter Two purpose cut 1 mm acetal plates were designed according to standard format 1536-well plates, but with well diameters of 1.00 mm and 1.75 mm respectively. The plates were designed using Autocad and machined using an ILS12.75 laser. For each plate, 16 × 24 wells were filled with Picolourlock Gold in ratios prescribed by the manufacturer (Innova Biosciences). Five droplets, which together constituted ≈ 20 nL of PBS [131] were dispensed into each well using a BDFACs Aria III cell sorter. Green wells were then counted manually.

3.3.7 Accuracy testing of ECHO 555 liquid handler 50 ml PBS was dispensed onto 185 mm standard filter paper. The filter paper was then allowed to dry before being cut out and placed behind a 384-hole titanium plate that had either uniform diameter holes or holes with decreasing hole sizes (0.8 mm-0.1 mm). The plate and filter paper were screwed on to 10 mm Perspex skirts that insolated against the current when used in the ECHO 555. The construct was then placed in the ECHO 555 liquid handler and 6 x 2.5 nl drops were dispensed. If accurate, the drops cleared the titanium plate and created a green readout on the filter paper. The percentage of green “spots” was counted manually.

3.3.8 Preparing “proof of principle” loading media The luminescent GC2 strain was grown overnight in LB at 37°C at 220 rpm and diluted in 3.3 % ULGTA LB media (37°C) in a 1:10 ratio. The final cell concentration 5 × 108 - 109 cells/ml. The solution was shaken for 2 min to increase dispersion of the reporter strain.

3.3.9 Dipping loading protocol An array was dipped in a 500 ml beaker filled with proof of principle loading media. The excess was scraped off using disposable L-shaped spreaders. The compartment luminescence of the ISSA array was then analysed by placing the array in a ChemiDoc™ MP Imaging System for 600 sec.

3.3.10 Mechanical shearing loading protocol The room temperature was raised to above > 30°C. A bottom outer, a bottom protective plate, the membrane, the array plate, and the top frame were all assembled in that order, with the open side of the array facing upwards. The assembled components were then incubated at 37°C. 30 ml of proof of principle solution (37°C) was placed on top of the array. Small brushes were then used to physically dislodge the air bubbles. The top frame was subsequently removed, and the excess media was scraped off using disposable L-shaped spreaders. The compartment luminescence of the array was analysed by placing the array in a ChemiDoc™ MP Imaging System for 600 seconds.

3.3.11 Vacuum-assisted loading protocol The room temperature was raised to above > 30°C. A bottom outer, a bottom protective plate, the membrane, the array plate, and a 6 mm loading top frame were all assembled in that order, with the open side of the array facing upwards. The assembled components were then incubated at 37°C. 50 ml of proof of principle media (37°C) was poured on top of a modified loading platform. A vacuum was then applied using a desiccator. After 2 min, the vacuum was removed. The top frame was then removed, and the excess media was scraped off using disposable L-shaped spreaders. The luminescence of the array compartments was then analysed by placing the array in a ChemiDoc™ MP Imaging System for 600 sec.

3.3.12 Sequential loading protocol Sterile 3.0 % ULGTA SMS was loaded into a 5180-compartment array using the vacuum assisted loading protocol. After scraping off excess solution, the loading frame was replaced with a new sterile loading frame. The filled array was subsequently covered in 10 ml of SMS mixed in with the P1H2 reporter strain solution (final cell concentration 5 × 108 - 109 cells/ml) and left for 10 min at 37°C. 30 ml of 4.0 % ΜLGTA SMS (37°C) was then poured on top and mixed into the solution using a disposable L-shaped spreader and left to incubate for a further 30 min at 37°C. The array was then sealed, and the fully assembled ISSA was covered with filter paper saturated with SMS media

42 and left to incubate overnight at 28°C. The compartment luminescence of the array was then analysed by placing the array in a ChemiDoc™ MP Imaging System for 600 sec.

3.3.13 Proof of principle harvesting protocols – centrifugation A 384-compartment array (0.6 mm ø compartments) was filled according to the vacuum assisted loading protocol using proof of principle media. The array was then placed in between an upper protective plate and a lower funnel plate. The upper and lower plates only differed from the array plate by having 0.7 mm ø compartments. These three plates were placed on top of the 384-well plate filled with 1.0 % agar SMS. 10 ml of sterile water was subsequently poured onto a purpose cut filter paper and the filter paper placed on top of the upper protective layer. The complex was clamped together using autoclave tape. Centrifugation with 1789 x g was then applied for 10 min. The 384-well plate was separated from the array and sealed using a Breathe-Easy® sealing membranes. The rate of success was quantified by placing the 384-well plate in a MP Imaging system for 1200 seconds. Luminescent spots were then counted manually.

3.3.14 Proof of principle harvesting protocol – mechanical disruption A 384-compartment (0.6 mm ø compartments) array was filled according to the vacuum-assisted loading protocol using proof of principle media. The array was then placed in between an identical upper protective plate and an identical lower funnel plate. The upper and lower plates only differed from the array plate by having 0.7 mm ø compartments. These three plates were placed on top of the 384-well plate filled with 1.0 % agar SMS. A waterproof sterile membrane (SealPlate® film sterile) was applied on top of the protective plate and the complex was clamped shut using autoclave tape. Each plug was harvested by removing local strips of the impermeable membrane using sterile scalpels, before emptying the underlying compartments with purpose designed “harvesting pins” (0.5 mm titanium wire). The harvesting pins could pierce compartments but had a 90° angle kink that ensured that the pin could not penetrate the underlying SMS agar suspension in the 384-well plate. A new sterile pin was needed for each targeted compartment.

3.3.15 ISSA protocol As this protocol is novel, unorthodox, and a key component of the PhD thesis, the protocol has been included in greater detail.

Fig 3.1. Key components for loading an ISSA platform. 1 Inner array and two protective plates (grade 2 titanium) 2 Two semi-permeable membranes (Whatman® Nuclepore(TM) Track-Etched Membranes L X W 8 in. x 10 in., pore size 0.03 μm) 3 Silicone Rubber sealant in 2 ml syringe. Needle attached has 0.5 mm ø and had been cut down to decrease resistance 4 Tweezers for heating 5 Screwdriver 6 Outer frames (3 mm height, stainless steel) 7 Loading frames (6 mm height, aluminium). 8 Stainless steel 8 mm 3M screws.

DAY -1: 1. Clean one array and two outer protective plates, making sure no compartments are obstructed. If necessary, clear blocked wells using a 0.3 mm needle. 2. Place a small amount of silicone glue around the edge of the array plate. 3. Attach membrane. 4. Create screw holes in the membrane by heating tweezers with Bunsen burner and then pierce membrane (maximum 2 screw holes can be pierced between reheating tweezers). 5. Place an outer frame on the bottom, then a protective plate, and finally the array with the membrane facing downwards. 6. Place a loading frame on top of the membrane and screw together. 7. On top of this, add another chip, a piece of autoclavable paper, a membrane, another piece of paper, another chip, and a top frame - in that order.

8. Wrap the complex in tinfoil and secure the folds with autoclave tape. 9. Autoclave closed complex. 10. Prepare 10 ml of overnight culture of the reporter strain (PIH2) in LB and incubate in shaker at 28°C overnight at 220 rpm.

Place the following items in a 37°C oven: 1. Autoclaved ISSA platform (still wrapped). 2. 40 ml of 4.0 % ΜLGTA SMS. 3. 20 ml of 4.0 % ΜLGTA SMS. 4. 15 ml of SMS.

DAY 0: 1. Prepare soil bacteria: take 1 g of soil and dissolve in 10 ml of autoclaved water. 2. Mix well, then pipette 2 ml through 35 mm filter. Keep the filtered solution on ice. 3. Quantify bacteria using flow cytometry and add the appropriate soil suspension to 13.5 ml of SMS (37°C) 4. To this solution, add 40.5 ml of 4.0 % ΜLGTA SMS and mix (37°C). 5. Pour this mixture onto the loading platform in sterile conditions. 6. Place the loading platform in a desiccator for about 2 min (until the bubbles stop rising from the array). 7. Unscrew loading platform and scrape off the mixture before the mixture is fully solid. Make sure everything is removed. 8. Place a fresh loading platform on top of the array plate and place the entire complex in the fridge for 2 min to ensure the first layer has solidified (add sterile and non- permeable lid). 9. Pour 10 ml of reporter strain (P1H2) culture on top. Make sure it covers the entire surface. 10. Keep in 37°C oven for 10 min. 11. Add 30 ml of 4.0 % ULGTA SMS and mix well. 12. Cover with sterile and non-permeable lid and incubate for 45 min at 37°C. 13. Unscrew loading platform and scrape off the mixture before the mixture is fully solid. Again, make sure all excess agarose solution is removed.

14. Place a small amount of silicone glue around the edge of the array plate and attach membrane. 15. Pour 3 ml of SMS media on top of membrane and smooth gently using fingers. 16. Create screw holes in the membrane by heating tweezers with Bunsen burner. 17. Add a top frame and screw tight. (ISSA array is now fully assembled). 18. Wet purpose cut filter paper with SMS and place on top of ISSA array. 19. Place in humidity chamber overnight at 28°C.

Day 1:

1. Remove filter paper and place 10 ml of SMS on top of ISSA. 2. Use Clariostar MP imaging system to screen compartments for luminescence. 3. Return ISSA platform to environment where soil was sampled.

Day 21.

1. Retrieve ISSA chip and harvest compartments where luminescence has decreased significantly (P > 0.0001). 2. To harvest individual colonies, fill a CELLSTART OneWell PlateTM with 1.0 % agarose SMS and place a sterile 5180-compartment plate array on top. Place the retrieved ISSA array on top and empty compartments using 0.3mm sterile needles. After harvesting, retrieve plugs from one well plate and resuspend in 100 μl SMS. 3. Plate out the suspension on petri dishes filled with 2.0 % agar SMS. 4. Incubate at room temperature until colony growth can be detected.

Day 21-28.

1. Monitor petri dishes for growth. 2. When growth is detected, scan colonies for luminescence using ChemiDoc™ MP Imaging System for 600 sec. 3. Isolate colonies not exhibiting luminescence on new petri dishes. 4. Verify identity of purified soil colonies using 16S rRNA sequencing.

3.3.16 Cultivating uncultivable bacteria using a 384-compartment ISSA array with no reporter strain Bacteria were quantified using the CFDA-SE/EDTA quantification protocol. 0.1, 1.0 and 10.0 bacteria per compartments were loaded into three separate 384-compartment ISSA platforms. The ISSA protocol was modified by not adding luminescent strains and by incubating the ISSA platform immediately after loading. The ISSA platforms were incubated for 21 days, before being retrieved. All plugs were harvested using the mechanical harvesting protocol. Bacteria were then allowed to incubate in the 384-well plates for 7 days before compartment growth was measured. As a control, the soil solution at day 0 was resuspended on LB agar petri dishes in serial dilutions of 1:1, 1:102, and 1:104 ratios. The LB petri dishes were incubated at room temperature for 7 days and CFUs were counted. The domestication rates of ISSA array and standard plates were compared so that the proportional increase in domestication associated with in situ incubation in the ISSA platform could be estimated.

3.3.17 16S rRNA species identification Isolated soil bacterial colonies were identified using colony PCR. The colony PCR protocol was as follows: A scrape from an isolated colony was resuspended in 100 μl sterile water and boiled for 7 min. 25 μl PCR reactions contained 2.5 μl 10x KOD polymerase buffer, 200 mM dNTPs, 0.5 μM primers, 0.02 KOD polymerase Units, 5.0 % dimethyl sulfoxide (DMSO), 2 μl of bacterial lysate, and 20.5 μl Millipore water (or to 25 μl). 40 cycles were run with a 30 sec 95°C denaturation step, a 30 sec 50°C annealing step, and a 40 sec 70°C step. The protocol concluded with a final 10 min extension step. PCR products were purified with Wizard® PCR purification kit and subsequently with the SV Purification Gel Kit, before being dispatched for Sanger sequencing (Eurofins) using the 907R-A primer. If the 907R-A primer did not work, the PCR product was sequenced using the 907R-C primer. The resulting 16S rRNA sequences were analysed using the EzBioCloud database.

3.3.18 Reporter strain development – Pseudomonas putida and Pseudomonas fluorescens (P1H2 and FH1) The donor strain (E. coli S17-1 λpir mini-Tn5 luxCDABE) and the recipient strain (either Pseudomonas putida (P. putida) or Pseudomonas flourescens (P. flourescens)) were individually grown overnight in LB with ampicillin (100 mg/ml). The two cultures were subsequently diluted (1:100) and grown for 2.5 h at 37˚C with selective antibiotics. The donor and recipient strains were mixed and plated out on filter paper, which was placed on a LB petri dish with ampicillin

(100 mg/mL). The samples were incubated for 48 h at 37˚C. The resulting biofilm was resuspended in 2ml of LB. The resuspension was then serially diluted in the following ratios 1:102, 1:104, and 1:106. 100 µL of each dilution were spread on multiple 150 mm petri dishes containing 50 mg/ml kanamycin and 35 mg/ml chloramphenicol. The plates were analysed using a Bio-Rad ChemiDoc MP image reader and the top performing 292 colonies were resuspended in 200 µl LB in standard 96-well plates. Luminescence of samples was monitored for 7 days and analysed daily by a BMG Labtech Clariostar. The 17 selected integrants with the highest and most constant luminescence were “purified” using 1-2 rounds of plating and single colony picking using ChemiDoc™ MP Imaging System to pick the colonies with the highest associated luminescence.

3.3.19 Reporter strain development – Bacillus subtilis Promotors Pveg and PlepA were digested and ligated into the pBS3Clux plasmid using EcoRI and PstI. Ligation reactions were chemically transformed into One Shot® TOP10 cells, before validation and selection of correctly constructed plasmids for subsequent Bacillus subtilis (B. subtilis) transformation. B. subtilis 168 was transformed using the “Paris method” of transformation [132]. Spizizen’s minimal medium was used as growth media. Bacillus subtilis Mini Bac cells were induced to competency by adding 0.5 % mannitol as described by Rahmer et al. [127].

3.3.20 Preparation of electrocompetent Mycobacterium smegmatis 100 ml of Middlebrook 7H9 Broth media containing 0.05 % glycerol and 0.5 % Tween 80 was inoculated with Mycobacterium smegmatis (M. smegmatis). M smegmatis was grown to an OD600 0.6-1.0. The culture was resuspended in two 50 ml falcon tubes and centrifuged for 10 min at 1789 G. The supernatant was discarded, and each pellet was resuspended in 20 ml of 7H9 media before being pooled. The suspension was centrifuged again for 10 minutes at 1789 G. The pellet was then resuspended in 10 ml 7H9. This step was repeated three times. The competent cells were finally resuspended in 100 µl aliquots of 7H9 25 % glycerol media solution, before being flash frozen and stored at -80 °C.

3.3.21 LuxCDABE integration of M. Smegmatis pMV306DIhsp+LuxG13 and pBS-Int were propagated in E. coli according to the protocols described in Andreu et al. [129]. Plasmid DNA was purified with Wizard Plus SV Minipreps DNA Purification System® kit according to the manufacturing protocol. Electrocompetent M. smegmatis were

48 thawed on ice for 10 min. 5 µg of plasmid DNA was added to the electrocompetent cells, and the solution was transferred to a pre-cooled 0.4 mm cuvette. The cells were then electroporated using the Gene Pulser XcellTM Electroporation System (2.5 KV, 1000 Ω, 25 µF). Electroporated cells were resuspended in 1 ml of 7H9 media and allowed a recovery time of 4 h at 37°C at 220 rpm. Subsequently, 100 µl of the cell culture were plated on 7H10 broth media supplemented with Middlebrook ADC Enrichment plates. Plates were finally incubated for 72 h at 37°C. A ChemiDoc™ MP Imaging System was used to pick the colonies with the highest associated luminescence, which were then isolated, cultivated, and flash frozen.

3.3.22 Measuring light intensities of newly developed reporter strains Reporter strains were incubated overnight (M. smegmatis for 72 h) at 28˚C on petri dishes on LB petri dishes. The reporter strains were then serially diluted to produce the required concentration (ratios chosen were 1:103, and 1:106) before being mixed with the selected growth media (2xTY or SMS) containing 2.0 % agar. Each reporter strain was grown in 4-6 copies in 96-well plates, with each well containing 225 μl of the reporter strain solution. The 96-well plates were incubated at specified temperatures and luminescence measured periodically using a Clariostar plate reader. The gain was set to 2500 AU.

3.3.23 Breaking up mycelia (mycelia break up protocol) S. aureofaciens was grown for 48 h in 10 ml GYM/MYM with 10-15 2 μm glass beads at 28°C. The culture was then passaged (1:30 dilutions) to 10 ml GYM/MYM media with 10-15 2 μm glass beads containing the chosen concentration of detergent. After 3 × 1 ml of the culture was left unfiltered, passages through 35 μm filters, or 5 μm filters. The OD600 was then measured and compared. In the final protocol MYM media was used as growth media and 0.5 % Brij was added in the second inoculation step.

3.3.24 Classical Waksman detection of antibiotic production using S. aureofaciens or S. venezuelae - overlay S. venezuelae and S. aureofaciens were grown for 48 h in 10 ml MYM with 10-15 2 μm glass beads at 28 °C. The culture was then passaged (1:30 dilutions) to 10 ml MYM media containing 10-15 2 μm glass beads and 0.5 % Brij-35. S. venezuelae and S. aureofaciens were then spun down and serially diluted (1:10, 1:102 and 1:104) onto petri dishes containing six solid medias that had been identified as antibiotic production medias in the literature (MYM, GYM, YM, ISP4, YEMES, and GI

49 production medium). Each serial dilution had three replicates. The serial dilutions were grown for five days at 28°C to allow for antibiotic production. The plates with 50-500 distinct colonies were selected and a 2.5 mL thin layer of 40°C LB 2.0 % agar mixed with the reporter strain GC2 (ratio 1:200) was placed on top of the producer strains in the petri dishes. The petri dishes were then incubated overnight at 28°C. After 8 h, antibiotic production could be detected by eye, as the agar was “clear” in affected areas.

3.3.25 Classical Waksman detection of antibiotic production – cut out Antibiotic producers were grown in specified growth media, before being plated out on 150 mm petri dishes. The producer strains were grown for 5-14 days at 28°C to allow for antibiotic production. 30 mm ø circles were periodically cut out from the plates and placed on a new petri dish. Overnight cultures of P1H2, GC2, or Mini Bac Pveg Lux were then mixed into 40°C LB 2.0 % agar (ratio 1:200) and the solution was poured around the producer strain cut out. The petri dish was incubated overnight at 28°C and clearance zones could be detected the next day. In later experiments, to increase clearance zones, petri dishes were placed in a fridge at 5°C and re- examined after three weeks.

3.3.26 Broad assay of antibiotic production using various producer strains, reporter strains, and media in 96-well plates (mixed) The producer strains S. fradiae, S. venezuelae, S. aureofaciens, and S. cattleya were grown according to the mycelia breakup protocol. The cultures were then spun down and resuspended in the six chosen growth medias (2xTY, R2A, YEMES, MYM, MYM + CaCO3, and ISP4) in a 1:10 ratio. The reporter strains P1H2, GC2, Mini Bac Pveg lux, and M. smegmatis G13 Lux were grown to stationary phase. The reporter strains were serially diluted into the six chosen growth medias (1:106 ratio for P1H2, GC2, 1:105 for Mini Bac Pveg lux, and 1:103 for M. smegmatis G13 Lux). 100 μl of the serially diluted producer strain and 5 μl of the serially diluted reporter strain were mixed with 120 μl of the chosen growth medias containing 2.0 % ULGTA. The co-inoculated solution was vortexed and incubated in a 96-well plate. A matrix was set up to test all combinations of producer strains, reporter strains, and media. Four 96-well plates were set up, with each 96-well plate having a unique reporter strain mixed into every chamber. On each plate, every two rows had a separate producer strain and every two columns had separate growth media. Finally, a “blank” plate was set up, in which 5 μl of each reporter strain was grown in 220 μl pure culture in all six growth

50 medias. On the blank 96-well plate every two rows had a separate reporter strain and every two columns had separate growth media each. The experiment was run in a standard incubator at 28°C for six days. A 37°C heat shock was carried out for the first hour to boost antibiotic production. The luminescence of the wells was measured periodically using a ChemiDoc™ MP Imaging System for 600 sec.

3.3.27 Detecting antibiotic production using S. aureofaciens and Mini Bac Pveg Lux in mixed/layered formats The producer strain S. aureofaciens was grown according to the mycelia breakup protocol. S. aureofaciens was then serially diluted in MYM media using ratios of 1:1, 1:10, 1:100, 1:1,000. The reporter strain Mini Bac Pveg Lux was grown in LB and serially diluted in MYM media using 1:102 and 1:105 dilutions. In the mixed setup, 5 μl of Mini Bac Pveg Lux was mixed with 100 μl of the producer strain media and 120 μl of growth media. The solution was then dispensed into wells in 96-well plates. In the layered setup, 100 μl of the producer strain media and 120 μl of the growth media were dispensed into 96-well plates and allowed to solidify. 5 μl of the serially diluted reporter strain was then resuspended in 1.0 % ULGTA MYM and dispensed on top of the solidified wells. A matrix was set up to test all combinations of serial dilutions of the producer strain and reporter strain in 96-well plates (96-Well Flat Clear Bottom Black Microplates). Each combination of producer strain dilutions and reporter strain dilutions were set up in quadruplicate. Quadruplicate “blank” wells were set up, in which 5 μl of each reporter strain dilutions were grown in 220 μl pure culture. The experiment was incubated in a humidity chamber. a 37°C heat shock was carried out for the first hour to boost antibiotic production, before the temperature was set to 28°C for 14 days. The luminescence of the wells was measured periodically using a ChemiDoc™ MP Imaging System for 600 sec. The experiment was run in triplicate. Each well was integrated using purpose scripted macros for ImageJ (see APPENDIX I for macro scripts) and an ANOVA regression was run using Excel 2016.

3.3.28 In vitro ISSA platform detection of antibiotic production Moist soil was retrieved from three locations (Hyde park (UK), Silwood Park (UK), and Contencie (France) and transported back in plastic bags to prevent the soil from drying out. The ISSA arrays were loaded with 1.0 or 0.0 bacterium per compartment according to the ISSA loading protocol. Scans were carried out using a ChemiDoc™ MP Imaging System for 600 sec. The platforms were

51 first scanned after 10 h (day 1) and daily afterwards until day 7. In between scans, the ISSA platform was incubated in a humidity chamber and wrapped in purpose cut damp paper covering both sides to prevent compartments drying out over time. During luminescence measurements a 3 mm layer of 1/10 strength SMS solution was placed on top. The ISSA array was divided into two halves, with the left half being scanned 5 min after nutrient addition and the right half being scanned 20 min after nutrient addition. The corresponding images were analysed by integrating the compartment luminescence using purpose written macros for ImageJ (see Appendix I) and the resulting data was analysed using Excel 2016.

3.3.29 In vivo ISSA platform detection of antibiotic production Moist soil was retrieved from Silwood Park (UK) and transported back in plastic bags to prevent the soil from drying out. The ISSA arrays were loaded with 1.0 or 0.0 bacterium per compartment using the ISSA loading protocol. Scans were carried out using a ChemiDoc™ MP Imaging System for 600 sec. The platforms were first scanned after 10 h (day 1) and at completion (day 21). In between scans, the ISSA platform was incubated in situ at Silwood Park. During transportation to Silwood Park (about 120 min from door to in situ location) the ISSA platforms were wrapped in tissue soaked in soil water. During luminescence measurements a 3 mm layer of 1/10 strength SMS solution was placed on top. The ISSA array was divided into two halves, with the left half being scanned 5 min after nutrient addition and the right half being scanned 20 min after nutrient addition. The corresponding images were analysed by integrating the compartment luminescence using purpose written macros for ImageJ (see Appendix I) and the resulting data was analysed using Excel 2016.

Chapter 4. Quantifying soil bacteria

4.1 Introduction This chapter focused on building a protocol to distinguish live soil bacteria from their external environment, and subsequently isolate and quantify them. The aim was to facilitate the loading of a defined number of soil bacteria into the later developed in situ screening array (ISSA platform). Most soil bacteria are, however, situated in bacterial communities on significantly larger soil particles, often resulting in mixed populations [133, 134]. A variety of techniques have consequently been used to dislodge soil bacteria from their habitat. Dislodging techniques can be divided into physical or chemical dispersion techniques [134]. Physical dispersion techniques that have frequently been used are centrifugation, blending, shaking, and mild ultrasonic treatment [134]. Chemical dispersion agents have often been used in combination with mechanical methods, and several substances have been used, including chelating agents [135], sodium charged ion exchange resins [135-137], and detergents [135]. The strong binding between the bacteria and the soil particles means that dislodging attempts have often led to cell damage [134]. Lindahl and Bakken investigated the relationship between cell damage and proportion of cells dislodged and found that while all the techniques listed above had a positive correlation between the level of damaged cells and dislodged cells, shaking and blending produced the highest level of dislodged bacteria that were still viable [134].

After being dislodged, soil bacteria had to be made discernible in the soil debris solution. This process was complicated because of the composition of soil, which consists of a complex mix of organic and inorganic compounds [138]. Soil bacteria have been found to be impossible to distinguish from the surrounding environment using light microscopy [138]. Consequently, microbiologists have used a variety of fluorophores. These include Acridine Orange direct count [133, 139] and 4,6-di-amidino-2- phenylindole (DAPI) [140]. Many stains, however, require fixing of the bacteria in question prior to staining. These stains therefore cannot determine the ratio of active/live/dead bacteria prior to fixation. Alternative stains, such as propidium iodide, while able to stain certain live bacteria, bind specifically to DNA or RNA and therefore compromise the genome and its expression in the stained organism [141]. Recently, 5-cyano-2,3-ditolyl tetrazolium chloride (CTC) has been used to identify active bacteria without compromising their viability [92, 140, 142, 143]. CTC is membrane permeable and is a direct indicator of oxidative metabolism.

During metabolism, the bacterial electron transport system reduces CTC, leading to a red fluorescent molecule that can be detected by fluorescent microscopy or flow cytometry [143]. Results from Ullrich et al. [144], however, showed that the above described redox reaction led to lower growth rates in natural aquatic bacterial communities. The reported lower cell growth therefore questioned the ability of CTC to preserve the viability of stained cells. Furthermore, the suggested staining protocol for CTC have often been lengthy [145], with optimal incubation times as high as 24 h reported in the literature [146]. 5-(and-6-)-carboxyfluorescein diacetate- succinimidyl ester or “CFDA-SE” is a recently suggested alternative [133, 147]. CFDA-SE is cell permeable and naturally non-permeable [147]. Native bacterial esterases cleave CFDA-SE, leading to a reaction that eventually results in CFDA-SE being converted into a green fluorophore (excitation 492 nm, emission 517 nm) [133]. CFDA-SE has therefore been used to identify metabolically active cells from their surroundings (see Fig 5.1) and has the added advantage of being relatively safe; both for the bacterial cell and the researcher carrying out the experiment. Furthermore, CFDA and CFDA-SE have both been used in combination with flow cytometry to detect live bacteria in aquatic environmental samples [133, 148].

Fig 4.1. Image of CFDA-SE-stained soil isolate next to autofluorescent sediments. Image taken from Fuller et al. [133].

To the best of the knowledge of this author, however, previous studies staining bacterial isolates have had a bias toward readily cultivable bacterial phylum. Consequently, there is a need to create a staining technique that can detect a significant proportion of the often-uncultivable terrestrial bacterial biome, while staying sufficiently selective to not stain the surrounding debris.

4.2 Results

4.2.1 Selectively staining soil bacteria To develop a protocol that could stain a broad variety of soil bacteria, while leaving surrounding soil debris unstained, moist soil was retrieved from the Presidents Garden at Imperial College London and immediately resuspended in sterile water in a ratio of 1 g / 10 ml. The sample was subsequently manually shaken for 30 sec, before being filtrated using 200 μm, 35 μm, or 5 μm filters. A variety of fluorescent stains, including Sybr Green, Syto9, Hoechst, Fm-464, Acridine orange, DAPI, Dioctadecyl Tetramethylindocarbocyanine Perchlorate (DIL), and CFDA-SE were then applied to separate samples. Of these applied stains, CFDA-SE most accurately identified distinct foci without staining surrounding debris. This suggested that CFDA-SE staining was highly specific to bacteria; a finding that confirmed results from Fuller et al. [133]. To confirm that CFDA- SE stained live bacteria, soil samples were isolated and subdivided into autoclaved and non- autoclaved soil, filtered with a 200 μm filter, and imaged using a fluorescent microscope. The corresponding 16 bit images (range 0 – 65535 arbitrary units (AU)) were analysed using FIJI [130]. If CFDA-SE selectively stained live bacteria, autoclaved soil should exhibit no signal when using fluorescent microscopy, as live bacteria would have died during autoclaving. During experiments, autoclaved soil consistently exhibited no detectable signal in the fluorescent channel, even at strong light intensities and high gain (see top row in panel in Fig 4.2). When CFDA-SE was used to stain pristine soil, the results showed clear punctate foci representing live soil bacteria (see second row Fig 4.2).

Fig 4.2. Autoclaved and pristine environmental samples stained using CFDA-SE. The autoclaved soil exhibited no fluorescent foci. In contrast, pristine soil exhibited clear fluorescent punctate foci.

4.2.2 Testing the scope of CFDA-SE across the bacterial kingdom While microscopy analysis suggested that CFDA-SE selectively stained live bacteria, the scope of the stain has never been fully explored; i.e. it was unknown if CFDA-SE could detect a multitude of bacterial phyla across the bacterial kingdom. To test whether this was the case, a collection of bacterial strains was stained with CFDA-SE (see Table 4.1).

Table 4.1. Strains tested for CFDA-SE staining efficiency. Species with brackets signify novel species with the percentage corresponding to sequence similarity of closest relative. Species (sequence similarity) Abbreviation Phylum (Class) Caulobacter cresentus C. cresentus (Alpha)proteobacteria Bordetella bronchiseptica B. bronchiseptica (Beta)proteobacteria

Myxococcus xanthus M. xanthus (Delta)proteobacteria Escherichia coli E. coli (Gamma)proteobacteria Amycolatopsis orientalis A. orientalis Actinobacteria Streptomyces venezuelae S. venezuelae Actinobacteria Terriglobus roseus T. roseus Acidobacteria Bryobacter aggregatus B. aggregatus Acidobacteria

Gemmata obscuriglobus G. obscuriglobus Planctomycetes Luteolibacter arcticus L. arcticus Verrucomicrobiaceae

Gemmatimonas aurantiaca G. aurantiaca Gemmatimonadetes Flavobacterium branchiarum (98.16 %) F. branchiarum Bacteroidetes Pedobacter duraquae (98.00 %) P. duraquae Bacteroidetes

Together the represented phyla and subphyla in Table 4.1 constitute > 85 % of the total population of bacteria found in environmental samples [61, 149]. The strains in Table 4.1 were stained using CFDA-SE. The results showed that while some strains stained well (e.g. B. subtilis, see Fig 4.3 B), most bacteria (especially Proteobacteria) exhibited sporadic staining (e.g. M. smegmatis – see Fig 4.3 C). Finally, even when strains stained well, they seemed to be actively or passively secreting CFDA-SE, leading to increased background fluorescence (e.g. S. venezuelae – see Fig 4.3 D). It is worth noting that the increased background fluorescence could be mitigated through increasing the rounds of spinning and washing the samples.

Fig 4.3. Composites of stained autoclaved soil and pure cultures stained with CFDA-SE. A autoclaved soil B B. subtilis C M. smegmatis. D S. venezuelae).

Overall, the combination of sporadic staining and secretion of cleaved CFDA-SE was deemed unsatisfactory for later quantification. Analysing the results, it was postulated that when staining failed to take place, it was a result of CFDA-SE failing to penetrate the membrane. Furthermore, given the short period of incubation (20 min), it was deemed likely that cleaved CFDA-SE secretion was a result of active pumping. To test these hypotheses, the following detergents and chelating agents were used to increase membrane permeability: Triton, Tween 80, Tween 20, EDTA and Brij- 35. It was found that adding 0.25 M of EDTA to samples increased permeability and stopped the secretion of cleaved CFDA-SE. Going forward, when comparing the results of staining samples that have been suspended in a solution of water with samples that have been resuspended in 0.25 M of EDTA, the two protocols will be referred to as CFDA-SE/water and CFDA-SE/EDTA, respectively.

The combined effect of using the CFDA-SE/EDTA protocol led to significant increases in fluorescence across the tested phyla. Furthermore, the secretion of cleaved CFDA-SE seized to the extent that background fluorescence was no longer optically visible in the sampled microscope images. To illustrate the effect, S. venezuelae and M. smegmatis stained with CFDA-SE/water and CFDA-SE/EDTA were compared (see Fig 4.4.). While S. venezuelae stained with CFDA-SE/water (see Fig 4.4, first row) had low intensity and secreted a considerable amount of cleaved CFDA-SE, S. venezuelae stained with CFDA-SE/EDTA (see Fig 4.4, second row) had a higher fluorescent intensity and secreted almost no cleaved CFDA-SE into the nearby environment. When looking at M. smegmatis, staining with CFDA-SE/EDTA (see Fig 4.4, last row) improved the coverage, leading to a close to 100 % cell coverage.

Fig 4.4. Panel of S. venezuelae and M. smegmatis stained with CFDA-SE with water or EDTA.

When applying CFDA-SE/EDTA protocol to a variety of strains, the staining efficiency increased substantially compared to the CFDA-SE/water protocol. There were, however, still problems associated with staining Proteobacteria. Of the tested model species of Proteobacteria, M. xanthus (Deltaproteobacteria) stained well and C. crescent (Alphaproteobacteria - see Fig 4.5, first row, left image) and E. coli (Gammaproteobacteria - see Fig 4.5, first row, right image) stained with 80-100 % efficiency, albeit with low intensity. The fluorescent images, however, indicated that B.

60 bronchiseptica (Betaproteobacteria) stained with CFDA-SE/EDTA had a 30-50 % coverage ( see Fig 4.5, first row, centre image). The remaining model species stained with CFDA-SE/EDTA exhibited ≈ 100 % efficiencies. The staining intensity, however, varied markedly between strains with, G. obscuriglobus (see Fig 4.5, third row, centre image and right image) staining so brightly that larger cell densities led to autofluorescence from the surrounding media. Overall the fluorescent microscopy images showed that applying a combination of CFDA-SE and EDTA allowed for the detection of > 90 % of all cells for model species representing > 85 % of cells found in environmental samples.

Fig 4.5. Composites of fluorescent and light images of pure cultures stained with CFDA-SE with EDTA. The results show that it is possible to identify > 90 % of phyla representing 85 % of the bacterial kingdom. The brackets signify that the strain used is a novel strain discovered from soil isolates and the percentage attached represents the sequence similarity to the closest identified relative.

Having established that CFDA-SE/EDTA stains a high proportion of soil species, it was possible to apply the CFDA-SE/EDTA protocol to pristine soil. The resulting pictures illustrated that the

62 technique allows for selective staining of soil bacteria, with little background fluorescence (see Fig 4.6). The pictures also showed that the shaking protocol broke up soil into smaller clay particles but failed to prevent multiple cells from inhabiting many of the larger soil particles. Microcolonies have been speculated to consist of a limited number of cells from the same species and groups of bacteria in close proximity could therefore be argued to be pure isolates. Particles with multiple geographically distinct cells could, however, be observed when using a 200 μm filtration step, suggesting that these particles were not inhabited by only one species. (see Fig 4.6 bottom row).

Fig 4.6. Autoclaved and pristine soil stained with CFDA-SE/EDTA and filtered using a 200 μm filter.

When placing the finding of mixed colonies on larger particles into the thesis’ framework of developing a high-throughput platform for antibiotic discovery; mixed cultures would complicate

63 the process of downstream identification of potential antibiotic producers. The time of (manually) shaking soil samples was therefore increased from 30 sec to 2 min to decrease the average size of soil particles and thereby decrease the number of soil particles containing multiple species. The soil samples were then filtered through 5 μm and 35 μm filters to eliminate the remaining larger fragments. When comparing the three filtration levels of 200 μm, 35 μm, 5 μm, there was a clear visual trend of a loss of cells/ml of solution when smaller filters were applied (see Fig 4.7). At 5 μm almost all biodiversity was lost, as < 90 % of all pictures had no bacteria. The loss of biodiversity between 35 μm and 200 μm was less drastic and all images from the 35 μm filter samples included ≥ 1 bacterium. After 35 μm filtration, soil debris was, however, so small that when bacteria were found they grew in pure colonies on soil debris in almost all cases; 29/30 particles found that contained bacteria had only one bacterium adhered to them. 35 μm filtration therefore seemed to be the best compromise, as it retained a high level of biodiversity, yet ensured relatively pure cultures.

Fig 4.7. Pristine soil stained using CFDA-SE/EDTA and filtered using 35 μm and 5 μm filters. Microscope images of 35 μm and 5 μm. Solutions filtered with 35 μm filters showed considerably more biodiversity than solutions filtered with 5 μm filters.

4.2.3 Quantification of soil bacteria using flow cytometry Another advantage associated with the use of ≤ 35 μm filters was that such filtration allowed for soil samples to be analysed using standard flow cytometry. This in turn provided an avenue for quantification of the bacterial population. To quantify bacterial concentrations, it was important to first establish the maximum fluorescence of soil particles that were not inhabited by live bacteria. This allowed for the designation of any particle with a higher fluorescence as a CFU. To find the maximum fluorescence of inanimate soil particles, autoclaved soil was stained using the CFDA-SE/EDTA protocol developed above and the fluorescence level was determined on an BD LSR

Fortessa using forward scatter and the fluorescent 530/30 nm channel (see Fig 4.8 A). Using autoclaved soil, it was possible to establish a threshold level of fluorescence, which only resulted in a very low false positive rate (average false positive rate = 2.5/250,000 per event, with the associated standard deviation = 0.4/250,000). Bacteria could then be identified by applying the gate from the autoclaved sample (see Fig 4.8 B). This allowed for a quantification of CFUs in the solution. The quantity of solution could in turn be measured by adding a known number of counting beads, which could be distinguished from the total population by using fluorescent channels not affected by the CFDA-SE stain (see e.g. Fig 4.8 C).

Fig 4.8. Isolating soil bacteria using flow cytometry. A Autoclaved soil. B Unprocessed soil. In A and B The blue line represents the gating of soil bacteria. The horizontal axes represent forward scatter (height). The vertical axes represent a green excitation laser (487 nm) and captured emission at 530 ± 15 nm (height). C Isolation of Trucount beads in B using two distinct fluorescent channels. The blue box on the right represents the gating of Trucount beads and the blue box on the left represents the rest of the debris in the solution, which exhibit lower fluorescence in the 780_60 YG-H channel. The horizontal axis represents a yellow green excitation laser (561 nm) with emission at 780 ± 30 nm (height) and the vertical axis represents a violet excitation laser (405 nm) with emission at 710 ± 20 nm (height).

4.2.4 Measuring bacterial population of soil dilutions filtered with 35 μm and 5 μm filters The developed CFDA/SE quantification protocol was applied to soil samples at 35 μm and 5 μm (see corresponding flow cytometry examples in Fig 4.9). The experiment was run for three different moist soil samples taken from three locations in Hyde Park London. Using 35 μm filters, the CFU count for the average soil solution was estimated at 4.03 × 107 CFUs/ml, which estimates the soil population at 4.03 × 108 CFUs/g of soil. This was in accordance with the prior literature estimates [59, 134, 150]. The inter-sample standard deviation was 9.70 × 106 CFUs/ml. The three samples were also run using a 5 μm filter and the average CFU count was 3.65 × 106, with a

66 standard deviation of 2.22 × 106 CFUs/ ml. Each sample therefore on average lost ≥ 90 % of the estimated CFUs by switching from a 35 μm filter to a 5 μm filter. The results indicated that there was a significant advantage associated with using the 35 μm filter.

Fig 4.9. Isolating soil bacteria from environmental samples using CFDA-SE/EDTA and 35 μm and 5 μm filters. A Soil filtered at 5 μm. B Soil filtered at 35 μm. The blue line in A-B represents the gating of soil bacteria. The horizontal axes represent forward scatter (height). The vertical axes represent a green excitation laser (487 nm) and captured emission at 530 ± 15 nm.

4.2.5 Further controls To test the accuracy of the developed flow cytometry method, a several additional experiments were carried out. The uncultivability of soil dwelling species meant that comparing the CFU count of the CFDA-SE/EDTA quantification protocol and standard CFU counts using petri dishes was not possible using environmental samples, as uncultivable bacteria would not grow on standard petri dishes. Instead proxy species had to be used.

4.2.6 Quantifying proxy species in soil Autoclaved soil stained with CFDA-SE/EDTA, and CFDA-SE/water were compared. It was found that there was a small but observable increase in background when using EDTA (see Fig 4.11, left column). There was therefore a need to apply different gates for CFDA-SE/water and CFDA- SE/EDTA. Having established the gate of CFDA-SE/water and CFDA-SE/EDTA, the proxy species S. venezuelae (S. venezuelae) and Pseudomonas putida (P. putida) were grown in autoclaved soil for 3-5 days. S. venezuelae and P. putida samples were subsequently divided into three aliquots, with one aliquot being stained using CFDA-SE/water, a second aliquot being stained with CFDA-

SE/EDTA, and the third aliquot being plated out on appropriate growth media of the proxy strains (examples of the flow cytometry results can be found in Fig 4.10). All experiments were carried out in triplicate.

Fig 4.10. Comparing the efficiency of CFDA-SE/water and CFDA-SE/EDTA when staining S. venezuelae and P. putida using flow cytometry. The horizontal axes represent forward scatter (height). The vertical axes represent a green excitation laser and captured emission at 530 ± 15 nm.

The three quantification protocols (CFDA-SE/Water, CFDA-SE/EDTA, and plating) showed convergence when estimating the population of S. venezuelae, but CFDA-SE/water significantly underestimated the P1H2 population at the P < 0.01 significance level (see Table 4.2).

Table 4.2. Population estimates using CFDA-SE/water, CFDA-SE/EDTA, and plating

Species (mean/standard deviation) CFDA-SE/water CFDA-SE/EDTA Plating S. venezuelae mean 6.97 × 106 6.92 × 106 6.73 × 106 S. venezuelae standard deviation 1.90 × 105 3.99 × 105 9.45 × 105 P. putida mean 8.50 × 105 1.86 × 107 1.92 × 107 P. putida standard deviation 5.99 × 105 1.05 × 106 1.68 × 106

These results confirmed the earlier microscope observations that S. venezuelae stained well both with CFDA-SE/water and CFDA-SE/EDTA, whereas P. putida only stained well when using CFDA- SE/EDTA. The combined results indicated that CFDA-SE/EDTA quantification could accurately detect and quantify two very different proxy species that had been cultured in soil, suggesting that the protocol should have high accuracy in quantifying the bacterial populations in natural soil.

4.2.7 Sensitivity and selectivity of the CFDA-SE/EDTA quantification protocol The rate of sensitivity (rate of false negativities) and specificity (rate of false positives), i.e. how often soil bacteria were falsely identified as soil debris and how often soil particles were falsely identified as bacteria respectively, was explored by using S. venezuelae. Microscopy and flow cytometry evidence presented above suggested that CFDA-SE/water and CFDA-SE/EDTA quantification protocols could be equivocated when staining S. venezuelae. Additionally, CFUs in S. venezuelae also seemed resistant to mechanical stress, which was likely a result of their hyphae chains, as that meant that each CFU included several viable cells. Mechanical damage could therefore kill one cell without compromising the viability of a CFU. This meant that CFDA-SE/water estimates could be verified as real colonies through single cell sorting. S. venezuelae was therefore incubated in soil for 5 days to allow it to adhere to soil particles. S. venezuelae was subsequently stained using CFDA-SE/water before being sorted by an BDFACs Aria III cell sorter. Given that earlier experiments had been carried out on a different flow cytometer, the threshold gate for distinguishing between soil debris and bacteria had to be redrawn. Accordingly, soil was autoclaved and diluted in water before being stained with CFDA-SE. The stained solution was then run through the sorter to set a threshold gate (see Fig 4.11 A). Having redrawn the threshold gate, S. venezuelae could be separated from soil debris and identified CFUs and soil debris could be single sorted into separate 96-well plates (see Fig 4.11 B), which had been filled with 2.0 % agar MYM growth media. 3 separate grams were taken and stained before being single sorted into a

96-well plate. 288/292 of the putative S. venezuelae cells exhibited growth, meaning that they had been correctly identified. 0/292 identified soil debris exhibited growth, again showing that they had been correctly identified as inanimate soil particles. The results indicated that flow cytometry could efficiently distinguish between CFDA-SE stained S. venezuelae and soil debris with very few false positives or negatives.

Fig 4.11. Sorting S. venezuelae using an BDFACs Aria III cell sorter. A Autoclaved soil was stained using CFDA-SE/water to set a threshold gate. B Soil inoculated with S. venezuelae was stained with CFDA-SE/water and S. venezuelae and soil particles were sorted into separate 96-well plates. The horizontal axes represent forward scatter (height). The vertical axes represent a green excitation laser (487 nm) and captured emission at 530 ± 15 nm.

4.3 Discussion The difficulty of staining a broad range of bacterial phyla, while leaving complex organic material untouched, has prevented scientists from publishing a rigid protocol for quantifying soil bacteria in soil solution. As stated earlier, CTC staining has gained prominence but suffers from long incubation periods [145, 146]. Other staining protocols have shown an ability to stain cells in environmental samples, but have not been applied to a broad range of bacterial phyla [145]. Fuller et al. [133] made steps towards addressing such shortcomings with the publication of their CFDA protocol. In their experiment they, however, only stained a variety of Gammaproteobacteria, Deltaproteobacteria, and Firmicutes [133]. The sub-phyla and phyla represented in their samples constitute less than 10 % of bacteria in global environmental samples [61]. This meant that while CFDA-SE has been established as a stain with a high specificity towards live bacteria of certain phyla and sub-phyla, further investigations had to be carried to establish whether this specificity

70 covered a broader spectrum of all bacterial phyla. This was done in this chapter and the results showed that without the addition the EDTA, the CFDA staining protocol stains only a limited number of phyla. With the addition of EDTA the protocol stains a very high percentage of soil bacteria and therefore constitutes another important step towards the accurate quantification of soil bacteria.

The developed CFDA-SE/EDTA quantification protocol has certain limitations. Firstly, the protocol fails to stain a minority of bacteria, leading to an error in sample estimates. A second error is the minority of filtered soil debris that contain more than one species. Finally, and potentially most significantly, there is an indication that a loss of biodiversity occurs when applying a 35 μm filter compared to a 200 μm filter. Although this effect might be overestimated as larger fragments ceteris paribus leads to higher sample concentration in specific locations and pictures were chosen based on the presence of fragments. It is therefore almost certain that there was some selection bias; which led to larger pore size filters having inflated bacterial counts.

The loss of biodiversity associated with smaller filter sizes could be addressed by combining CFDA- SE/EDTA staining with a dislodging technique that is more efficient at either breaking down soil debris to smaller components or dislodge bacteria from soil particles. Alternatively, a larger nozzle size when using flow cytometry would allow for a larger filter size and thereby reduce the loss of biodiversity. Until a modified protocol has been developed, the above protocol, however, provides a highly accurate estimate of the bacterial population size in a given soil solution and can therefore be used to load an appropriate number of soil bacteria into the in situ screening platform developed in the following chapters.

Finally, early indications (not included in this thesis) suggest that the protocol would be even more efficient in aquatic environments. This is mainly because there are fewer particles of > 35 μm, meaning that less biodiversity is lost due to filtration. The developed protocol could consequently be used as a universal quantification protocol for a variety of environmental bacterial experiments.

Chapter 5. ISSA platform design

5.1 Introduction To screen uncultivable bacteria for antibiotic production, the concept of in situ cultivation [93, 116, 117] and the Waksman platform [30] had to be merged in order to create a novel high throughput platform that tested for antibiotic secretion of soil bacteria in situ. This chapter describes the development of the physical ISSA platform and the development of the corresponding protocols to allow for the loading, analysis, and harvesting of a high number of producer strains. This incorporated four areas of research that all had to be successfully developed and tested for the platform to be operational:

i) Physical platform development. ii) Loading protocol development. iii) Testing the experimental robustness during incubation. iv) Harvesting protocol development.

This chapter will discuss each in turn.

5.1.1 Design considerations for the physical ISSA platform A variety of considerations had to be incorporated into the design process of the physical platform. Firstly, classical in situ cultivation platforms suspend bacteria in semi-isolation that allow nutrient exchange but ensure pure colony formation [93, 116, 117]. The ISSA platform therefore had to incorporate such a state into the basic design. A second consideration was the focus on high throughput. To achieve high throughput the ISSA platform had to facilitate parallel loading, incubation, readout, and harvesting of a high number of experiments simultaneously. A third key design aim was to minimise compartment volume. This was because a majority of soil bacteria have been shown to have fixed colony sizes of < 100 cells [88, 92]. Given the concentration of 107- 109 bacterial cells per gram of soil [59], bacteria could be expected to be conditioned to be fully operational in spaces composed of only a few nanolitres (nl). Given that the maximum colony size and optimal colony performance could be expected at all compartment sizes > 1 nl, it followed that increases in compartment size would not lead to further antibiotic production. As antibiotic concentration is a function of production/volume, it was therefore highly likely that a small compartment size increased the chance of a produced antibiotic reaching concentrations that

72 impact luminescent output of the reporter strain. Using a small compartment size was therefore postulated to potentially increase platform sensitivity to antibiotic production.

5.1.2 Loading protocols considerations A primary objective was to ensure homogenous and reproducible loading across the high throughput array of both reporter and producer strains, to facilitate intra-array and inter-array comparisons. Furthermore, when creating an antibiotic screening platform, it was important to incorporate the function of antibiotics in nature into the design considerations. If antibiotics were assumed to provide a competitive advantage by deterring growth of other species in an area, there were three hypotheses on how this could theoretically be achieved [151]:

i) Antibiotics allow invasion of habitats already inhabited by competing species. ii) Antibiotics inhibit growth of competitors when the producer strain and the competitor are at comparable densities in the same habitat. iii) Antibiotics prevent the invasion of established habitats by competitors and allow for only marginal gains in inhabited territory.

If the correct hypothesis was A or B, antibiotic screening could be achieved by developing a loading procedure that loaded a mixed population of producer and reporter strains in a single compartment. If C was correct, spatial or temporal isolation of the producer strain and reporter strain was needed to ensure antibiotic production by the producer strains. A literature review suggested little consensus surrounding the function of antibiotics in nature [152-154]. For a broader discussion of the function of antibiotics see the Introduction in Chapter 7. Given the lack of consensus, a primary objective for the loading protocol was to create a platform that allowed flexibility in the spatial/temporal relationship between the antibiotic producer strain and reporter strain, so that scenario A, B, and C could all be explored. A secondary objective was to optimise the ability to place a defined number of bacteria in each compartment. Stochastic loading, at best, follows a Poisson distribution. Equation 5.1 gives an estimate of the likelihood of k bacterial cells to be present in a sub-chamber, if liquid volumes containing an average of λ cells are dispensed.

Equation 5.1: Poisson distribution equation for dispensing cells Poisson distribution: P(k) = e-λλK/k!

Where P(k) = Probability of k cells dispensed into each sub-chamber, and λ = average cells in liquid dispensed.

A 1:1 ratio of reporter strains to soil strains therefore cannot be achieved consistently across an array using stochastic loading. Instead the highest ratio of compartments that will have one cell per compartment is achieved when the average cells/compartment = 1. When dispensing liquid that on average contains one bacterial cell into a compartment, only 36.8 % of compartments will contain exactly one soil species cell, 26.8 % of compartments can be expected to contain zero soil species cells, and 36.4 % of chambers will have two or more soil bacterial cells in each chamber. The probability of having exactly one cell in a compartment when dispensing an average of λ cells and the cumulative probability of at least one cell per chamber when dispensing an average of λ cells into a sub-chamber can be seen in Table 5.1 – assuming equal distribution of cells in the liquid and 100 % loading accuracy.

Table 5.1. Poisson distribution values for λ = (1-5) λ 1 2 3 4 5 P(k=1) 0.3679 0.2706 0.1493 0.0732 0.0337 P(k≥1) 0.6321 0.8647 0.9502 0.9816 0.9932

Alternatively, a 1:1 ratio of soil bacteria to reporter strains can theoretically be achieved, if bacteria are separated from the soil environment and single sorted into compartments ( given no dispensing error).

5.1.3 Incubation and harvesting considerations The primary concern during incubation and loading was the risk of compartments drying out [115]. This meant that chosen incubation locations had to have more water potential than the compartments in the ISSA throughout the experiment. During harvesting, it was essential that compartments were not exposed to the external environment for prolonged periods. A second primary concern was the risk of compartments becoming contaminated either from the outside environment or through cross-compartment contamination.

5.2 Results

5.2.1 Platform design The fundamental physical platform design format was chosen to be in accordance with 384-well plates (see Fig 5.2 left) to ensure ease of downstream processing. The platform design incorporates a central array plate. The array has a defined number of discrete compartments, with each compartment representing individual assays when carrying out ISSA experiments. Each compartments can encompass a producer strain and a reporter encapsulated in a solidifying agent that would fill the compartment. Semi-permeable membranes (Whatman® Nuclepore™ Track- Etched Membranes (L × W 8 in. × 10 in., pore size 0.03 μm) allow the array to be sealed from contamination from the exterior. Two protective outer plates that have identical design to the inner array are located on either side to protect the integrity of the membrane from outside rocks/debris. The identical design ensures that there are channels between the semi-isolated inner compartments and the external environment, which facilitates metabolite and nutrient exchange between the array and the external environment [108]. Outer protective clamps fix the five components in place (See Fig 5.1 right for full list of components).

Fig 5.1. Design of the ISSA prototype. The left illustration represents a top view of the standard 384-compartment ISSA design – numbers are in mm. The right illustration represents a side view of the ISSA platform and includes the individual components.

While the default design mirrors a 384-well plate format, the platform design was chosen to allow for different array designs. To change array number and diameter of compartments, the array and the surrounding protective plates will have to be machined to be identical, but all other components can be reused. The material choice was key to unlocking more flexibility in the platform design. After testing a variety of materials including acetal, stainless steel, and titanium, 75 it was found that grade 2 titanium afforded the best choice of construction material. Titanium has high corrosion resistance [155], which was important given the variety of pH value of different soil environments [156]. Furthermore, titanium is considered biologically inert [157], and it was found that grade 2 titanium remained rigid at thicknesses down to 0.5 mm. Finally, grade 2 titanium allowed for precise laser cutting. Consequently, 0.5 mm grade 2 titanium plates could be cut accurately with compartment diameters of ≥ 0.2 mm (see Fig 5.2).

Fig 5.2. 0.5 mm grade 2 titanium sheet laser cut with decreasing compartment diameter (ø) in a standard 384-well plate format. The compartment diameter starts at 0.8 mm (left) and is reduced by 0.1 mm every 3 columns. The smallest compartments (furthest three columns on the right) have 0.1 mm ø.

5.2.2 Developing a loading protocol Having established titanium as the plate and outer frame material, the number of compartments explored ranged between 384 – 5180-compartments per array (see Fig 5.3). For proof of principle experiments, compartment sizes were chosen to be 0.6 mm (141 nl volume per compartment), as this format was found to be the smallest possible for a ready visual confirmation of the integrity of the plug downstream. Later antibiotic discovery assays used arrays with diameters of 0.3 mm (31 nl volume per compartment), as 0.3 mm was found to be the smallest diameter that allowed for manual harvesting. At 0.3 mm, it was possible to laser cut 5180-compartments into a single

76 array while retaining the structural integrity of the titanium sheets. Higher numbers of compartments were found to decrease the structural integrity of the titanium and resulted in warped arrays.

Fig 5.3. Different designs of the ISSA platform.

5.2.2.1 Developing a loading protocol – single cell sorting A variety of methods were tested to optimise the dispensing protocols. These methods included single cell dispensing using the BDFACs Aria III flow cytometry cell sorter and stochastic loading via sound wave liquid dispensing using the ECHO 555, loading via dipping, loading via mechanical dispersion, and loading via desiccator-induced vacuum. The BDFACs Aria III cell sorter has single cell sorting capabilities and could therefore theoretically reach a higher control of the number of cells dispensed into each compartment, if loading errors could be eliminated. The accuracy was consequently tested to assess whether the sorter provided an optimal loading platform. This was achieved by filling 768 compartments in purpose cut 1536-compartment acetal plates with Malachite green. Malachite green reacts to phosphate by changing colour from yellow to green [158]. Standard sheath fluid is composed of a mix of materials, which includes PBS [131]. Given the setup of the experiment, this meant that accurately dispensed droplets from the sorter changed the colour of malachite green from gold to green. The purpose cut 1536-well plates were created with diameters of 1.75 mm and 1.00 mm. The results showed that at 1.75 mm ø, the sorter failed to hit > 20 % of all wells (see Fig 5.4 A). When dispensing sheath fluid into 1.00 mm compartments, the efficiency fell to < 50 % (see Fig 5.4 B). Loading the ISSA platform using a BDFACs Aria III cell sorter was therefore found unsuited for our purposes, as the diameter of the design of the array compartments had been established at 0.3 - 0.6 mm ø (see Section 5.2.1).

Fig 5.4. Colorimetric accuracy test of the FACS ARIA dispensing capabilities. The wells were filled with Malachite green. The BDFACs Aria III cell sorter dispensed PBS. Accurate dispensing = green well, a miss leads = golden well. A PBS was dispensed into 1.75mm diameter wells. B PBS was dispensed into 1.00mm diameter wells.

5.2.2.2 Developing a loading protocol – stochastic loading The following tested methods were all stochastic loading methods. The Echo 555 liquid dispenser uses sound waves to dispense 2.5 nl droplets to a designated area (see Fig 5.5A). Initial experiments confirmed that the ECHO 555 could accurately dispense molten agar (the maximum concentration used was 2.0 % agar). To test the accuracy of the ECHO 555, 6 x 2.5 nl droplets of Malachite green were dispensed into 0.8 mm diameter holes in a 384-hole chip. A dried filter paper saturated with PBS was placed behind the chip. This meant that accurately dispensed Malachite green would clear the holes and hit the filter paper to form green spots (see Fig 5.5 A). With uniform compartments of 0.8 mm ø, the ECHO 555 dispensed Malachite green into 382/384 holes (99.48 %) (see Fig 5.5 B). The same test was carried out on a custom-made plate with decreasing well sizes ranging from 0.8-0.1 mm ø (see Fig 5.3 C). In this test the ECHO 555 showed 100 % accuracy when dispensing into holes with ≥ 0.6mm ø. The accuracy dropped to < 50 % at 0.5 mm ø, and < 25 % at 0.4 mm ø. Using the ECHO 555 to load the array therefore limited the minimum compartment diameter to 0.6 mm ø. Furthermore, the ECHO 555 used more than one hour to fill all compartments in a 384-compartment array, leading to non-symmetric evaporation inside the compartments of the array during the loading phase. This was deemed unacceptable as a prime aim was to ensure homogenous loading.

Fig 5.5. Colorimetric accuracy test of the ECHO 555 dispensing capabilities. A Malachite green is dispensed into a 384-hole plate using an ECHO 555. The plate protects a dried filter paper that has been saturated with PBS. If the ECHO 555 is accurate, dispensed drops will clear the plate and interact with the filter paper behind to create green spots. B Malachite green is dispensed through 0.8mm diameter holes. C Malachite green was dispensed through a plate with decreasing hole sizes ranging from 0.8 mm ø - 0.1 mm ø. the hole diameter decreases by 0.1 mm every 3 columns from left to right. The insert panel shows a zoom of four holes in the respective custom-made plates.

Given the problems of the tested mechanical loading platforms, emphasis was shifted towards three manual loading protocols: loading via dipping, loading via mechanical disturbance, and vacuum-assisted loading. When carrying out proof of principle loading protocols, the developed luxCDABE reporter strain GI2 [126] was mixed in 2xTY media with 5×108 - 109 GI2 cells/ml. The temperature was kept at 41 °C, as GC2 cultures were found to struggle at higher temperatures. It was found that ≥ 1.0 % agar solidifies at disruptive rates at normal laboratory temperatures. 3.0 % Ultra-low Gelling Temperature Agarose (ULGTA) was therefore used as a gelling agent. It was found that 3.0 % ULGTA solidifies, but slowly, at room temperatures of 20-30°C. Fast execution at normal laboratory temperatures was consequently found to be feasible for the loading protocols discussed below. Alternatively raising the room temperatures to > 30°C increased the time of ULGTA solidification to an extent where it was no longer a factor during subsequent experiments. Given the above reasons, final proof of principle experiments were carried out using 3.0 % ULGTA 2xTY solution inoculated with 5×108 - 109 GI2 cells/ml with a starting solution temperature of 41°C and the room temperature raised to > 30 °C.

Dipping to inoculate was well established in the literature [108, 115]. Dipping was therefore the first protocol that was assessed. The setup for the dipping protocol consisted of filling a 500 ml beaker with the proof of principle solution (see Fig 5.6 left). The array was subsequently dipped in the beaker, while the bacterial solution was still liquid. After loading, the array consistently showed non-homogenous loading (see Fig 5.6 bottom left), which was deemed unsatisfactory. The cause of the non-homogenous loading was determined by eye to be air bubbles trapped in the array compartments. The trapped air bubbles had not been reported in the literature and were most likely caused by the small array compartment size. The aim of the developed mechanical disruption protocol was to dislodge these air bubbles via manual mechanical disruption, which would lead to homogenous loading. To allow for loading from one side; the bottom outer, the lower protective plate, the membrane, the array plate, and the top frame were all assembled in that order. In this setup the open side of the array faced upwards, which allowed agar to poured on top of the array. The entire complex will be referred to as the “Mechanical Disruption Loading Complex”. Proof of principle solution (30 ml) was placed on top of the array and was held in by the barriers of the upper frame (height 3 mm). Small brushes were then used to physically dislodge the air bubbles (See Fig 5.6 centre). It was found that mechanical disruption increased loading efficiency compared to dipping, which led to loading efficiencies of between 70 % - 100 % (see Fig 5.6 bottom centre), which was again deemed insufficient. Vacuum-assisted loading was developed as an alternative way to dislodge trapped air bubbles (see Fig 5.6 right). The loading platform was modified so that the top frame was replaced by a 6 mm height frame. An increased volume of proof of principle media (50 ml) was then poured on top the loading platform. A vacuum was then applied, which led to an expansion of the volume of the air bubbles. When the volume of the air bubbles became significantly greater than the volume of the compartments, the air bubbles were forced out and drifted up through the liquid solution. The increased height of the top frame and the increased liquid allowed space for the air bubbles to “clear” from the underlying compartments. When air bubbles cleared, the proof of principle solution replaced the air, ensuring uniform stochastic loading. After vacuum-assisted loading, the loading platform frame was removed and excess ULGTA solution was scraped off the array. A semi-permeable upper membrane was then attached to the array. The standard size upper protective layer, and outer were subsequently placed on top, before the platform was screwed together. Overall when executed correctly the protocol led to ≈ 100 % loading efficiency, with highly homogenous loading (see Fig 5.6 bottom right). Furthermore,

80 the protocol facilitated loading of plates with a variety of array designs, which allowed for flexibility of the design of the array.

Fig 5.6. Manual loading protocols and efficiencies. The efficiency of dipping, mechanical disruption, and vacuum-assisted loading was assessed by measuring the luminescence of the GC2 reporter strain in the array after loading.

5.2.3 Sequential loading To enable spatial or temporal separation of the producer and reporter strain, it was essential to develop a protocol for sequential loading. The vacuum protocol was therefore developed further to incorporate two inoculation steps (see Fig 5.7 A). To assess the ability of sequential loading, the media composition was changed to 3.0 % ULGTA SMS, as this was the planned media for in situ incubation. The array was then loaded with no reporter strains using the vacuum-assisted loading protocol described above. After scraping off excess solution, the loading frame was replaced with a sterile loading frame. The filled array was subsequently covered in 30 ml of 3.0 % ULGTA SMS reporter strain solution and left for 45 min at 37 °C to ensure that each compartment had been inoculated with the reporter strain. At this temperature the ULGTA did not solidify, which allowed reporter strains to colonise every compartment in the platform. The array was then sealed, and the fully assembled ISSA was covered with filter paper saturated with SMS media and left to incubate overnight at 28°C. The resulting sub-compartments in the array created physical

81 separation between the soil bacteria and reporter strain in each compartment, but enabled metabolite exchange (see Fig 5.7 bottom right).

Fig 5.7. Sequential loading protocol. 50 ml of 3.0 % ULGTA SMS is mixed with an appropriate dilution of soil bacteria and poured on top of a loading platform. The loading platform is subsequently placed in a vacuum until the air bubbles have been replaced by media inside the ISSA compartments. The excess media is scraped off and a solution of 30 ml containing the reporter strain is placed on top of the array for 45 min. The ISSA is sealed and incubated for 10 h. The outcome leads to physical distinct layers of producer strain and reporter strain, with the micron thin upper layer restricting the maximum population of the reporter strain.

The developed sequential loading protocol led to consistent homogenous loading across the array at day 1 (see Fig 5.8), thereby facilitating downstream ISSA screenings.

Fig 5.8. Luminescence exhibited by reporter strains after the sequential loading protocol. The image shows that the sequential loading protocol leads to homogenous loading of compartments even in the 5180-compartment format.

5.2.4 Developing a harvesting protocol The low volume of the ISSA compartments (31 nl – 131 nl) resulted in the array compartments drying out in less than 10 min when the compartments could equilibrate with the outside environment. It was envisioned that ISSA platforms could have several hits in one array. It was therefore deemed essential to find a method for harvesting a high number of compartments; either through quick mass unloading or by extending the time before the array dried out, thereby increasing the time for harvesting. For harvesting protocol optimisation experiments, the 384- compartment ISSA format was used, as this allowed for plugs to be unloaded into standard sterile 384-well plates. The condition of the plugs could then be examined more closely. Furthermore, it was found that compartments in the 5180-compartment ISSA format dried out slower than in the 384-compartment format, presumably due to their lower surface area/volume ratio. Protocols designed for the 384-compartment ISSA arrays could therefore be applied to other formats.

This thesis focused on developing harvesting protocols using centrifugation and mechanical disruption. Arrays loaded with the luminescent GC2 strain (5×108 - 109 cells/ml) were used as proof-of principle bioreporters when developing both protocols. Harvesting success was estimated through detection of luminescent plugs from the compartments. The centrifugation protocol was based around quick unloading (< 10 min). When harvesting, the array (0.6 mm ø) was

83 placed in between upper protective plate and a lower funnel plate with identical designs but 0.7 mm ø. The three plates were placed on top of the 384-well plate filled with 1.0 % agar SMS (see Fig 5.9 A). 10 ml of sterile water was subsequently poured onto a purpose-cut filter paper and placed on top of the upper protective layer. The complex was clamped together using autoclave tape. Centrifugation with 1792 G was applied for 10 min. It was found that the funnel plate prevented compartment plugs from travelling across the array plate and instead forced loose plugs downwards into the corresponding well in the 384-well plate. The centrifugation harvesting method allowed for <85 % harvesting efficiency in proof of principle experiments, where the microarray was newly loaded. The harvesting efficiency, however, fell to between 5.0 - 15 %, when harvesting arrays that had been incubated for more than 2 weeks in situ. The loss in loading efficiency was potentially due to partial drying out of the plugs inside the array during incubation. Given the low harvesting efficiency after in situ incubation, an alternative protocol was sought. Mechanical disruption harvesting relied on creating a closed system, which extended the period before the array compartments dried out after the ISSA chip had been retrieved and opened. Each compartment could then be selectively unloaded over a longer period. The protocol started by assembling a harvesting complex consisting of a 384-well plate filled with solid 1.0 % agar SMS, a lower protective plate (to prevent cross-contamination of the 384-well plate from the array plate), the array plate, and an upper protective plate (see Fig 5.9 B). A waterproof sterile membrane was applied to the top of the protective plate and the complex was clamped shut using autoclave tape. The impermeable membrane ensured that each individual compartment became part of a closed system, which consisted of the underlying 1.0 % SMS agar, the array compartment plug, and a limited dead space of air. The closed system increased the period before the array plugs dried out from < 10 min to > 3 h. Each plug was harvested by removing local strips of the impermeable membrane using sterile scalpels, before emptying the underlying compartments with purpose designed “harvesting pins”. The harvesting pins fitted inside each compartment but had a 90° angle kink that ensured that the pin could not penetrate the underlying SMS agar suspension in the 384-well plate. Instead compartment plugs attached to the top of the 1.0 % SMS in the wells and could subsequently be retrieved and analysed as seen fit. A new sterile pin was needed for each targeted compartment to ensure sterility. The protocol consistently took about one hour to carry out when harvesting a full 384-array plate. The protocol was therefore inside the maximum time allowed before the plugs had dried out. It was found that using the mechanical disruption protocol resulted in harvesting efficiency of > 98%.

Fig 5.9. Harvesting protocols. A Order of the assembled components needed for the developed centrifugation harvesting protocol. B Order of components required for the developed mechanical disruption harvesting protocols.

5.2.5 Sterility experiments To ensure the integrity of the experiments, it was important to fully prevent penetration of bacteria from the outside, as this would remove the control of the number of soil bacteria in the experiment. To test the integrity of the seal of the ISSA platform, in vitro experiments were set up by placing fully assembled blank ISSA platforms in 2xTY GC2 reporter strain inoculated media for up to one week. The ISSA array was then retrieved and tested for luminescence. It was found that applying a sealing agent (Radio Spares Pro Transparent Silicone Sealant Paste), between the periphery of both sides of the array plate to seal the array plate and the membrane, was essential to ensure sterility. Without the sealing agent, the sterility of the seal was broken consistently within 24-48 h. When arrays were sealed with silicone, all inner compartments (6 x 384- compartments) remained blank after a one-week incubation. The in vivo experiments were set up by placing loaded blank arrays in the soil for up to three weeks. 100 colonies were subsequently harvested and incubated in 384-well plates. Experiments were carried out in triplicate. It was found that all sampled compartments (3 x 100) remained non-contaminated from the external environment after three weeks of incubation in vivo. It was, however, found that when applying a sealing agent, the first line of compartments closest to the glued periphery could be compromised by glue residues filling up the compartment and pushing out the growth media. This led to a higher rate of false positives, as luminescent strains died out when growing on silicone. The first line of rows and columns from the outside were consequently eliminated from all further analysis.

5.2.6 Crosstalk experiments To test for intra-array communication, specific levees were designed and 3D-printed. The levees allowed for separate loading of two sections of a 384-compartment ISSA. To test for crosstalk four

85 columns of an ISSA array were filled with the P1H2 ( a P. putida luminescent strain developed in Chapter 6) and the ISSA platform was incubated in soil for two weeks. In the original setup with compartments isolated purely through pressure from the protective plates, it was found that some crosstalk occurred. To eliminate crosstalk, a silicone mesh was applied to the inner side of the protective plates to create local seals and thereby minimise crosstalk. This eliminated crosstalk when the semi-permeable membranes were flush with the array (see Fig 5.10 A). When small air bubbles between the array and the semi-permeable membranes were allowed, significant cross talk occurred (see Fig 5.10 B). The results shown in Fig 5.10 led to changes to the protocol to ensure that the membranes were flush with the array. To minimise crosstalk in subsequent experiments, the membrane attached to the bottom side of the array in the loading platform was autoclaved with the platform, which ensured a flush membrane. After loading the array and placing the top membrane on the array, this membrane was straightened by applying 2 ml of SMS media on top and smoothing the membrane manually.

Fig 5.10. Crosstalk experiments. The experiments partially loaded luminescent strains in the four columns furthest to the left. The array was scanned for crosstalk by detecting luminescence pre- incubation and post-incubation. A Successful partial loading of the ISSA platform. B Unsuccessful

86 partial loading of ISSA platform. When crosstalk occurred, air bubbles between the array plate and the semi-permeable membrane could be seen by eye.

5.2.7 In vivo experiments to establish the cultivation ability of the ISSA platform Having established the full flow of quantifying soil bacteria (see Chapter 4) and loading, incubating, and harvesting the bacteria (Chapter 5), it was possible to test the ability of the ISSA platform to domesticate soil bacteria. To do this ISSA platforms were incubated in situ, but without the reporter strain. To increase the margin of error for quantification, three separate ISSAs were loaded in each experiment. Using the CFDA-SE/water protocol developed in Chapter 4, the three ISSAs were loaded with an estimated average of 0.1, 1.0, and 10.0 estimated soil bacteria per compartment, respectively. The ISSAs were harvested after 2-3 weeks of in vivo incubation and each plug was grown on individual plates. Domestication rates in ISSAs with 1.0 bacteria per plate were found to be around 15-20 %, which was similar to previously reported numbers in the literature [115].

5.2.8 Building a library of uncultivable bacteria Individual plugs from the ISSA platforms filled with 1.0 bacterium per compartment were selected for further analysis. These plugs were plated out on standard LB-agar plates and SMS plates and grown at room temperature. Individual growth was tracked over a period of 120 days. 48 bacterial isolates showed growth on SMS agar plates, versus 44 on LB agar plates (with perfect overlap between the two populations). The 48 isolates were sequenced using 16S rRNA sequencing. It was found that the primer set 27F (5'- AGAGTTTGATCMTGGCTCAG-3') and 907R (5'- CCGTCAATTCMTTTGAGTTT-3') had universal coverage of grown colonies. The sequence results included 22 species (see Table 5.2) representing three phyla and five classes of bacteria (see Fig 5.11). The five classes of bacteria were Gammaproteobacteria, Betaproteobacteria, Bacilli, Sphingobacteria, and Flavobacterium. Two of the 22 species identified potentially represented new species with Pedobacter duraquae (98.00 %) and Flavobacterium branchiarum (98.16 %) the nearest relatives of the isolated bacteria. Repeats were relatively infrequent, and 75-80 % of arrays exhibited no growth, leading to the conclusion that bacteria could not travel easily between compartments. The results also indicate that while cultivability increased markedly, the increase

87 in domestication was focused around increases in domestication of certain classes of bacterial phyla.

Fig 5.11. Cultivated phyla from ISSA platform when run without reporter strain.

Table 5.2. Species designation of domesticated strains using the ISSA platform without reporter strains

Sequence similarity in Phyla Class Nearest related species %

Proteobacteria Gammaproteobacteria Pseudomonas turukhanskensis 99.61

Proteobacteria Gammaproteobacteria Pseudomonas turukhanskensis 99.53

Proteobacteria Gammaproteobacteria Pseudomonas turukhanskensis 99.51

Proteobacteria Gammaproteobacteria Pseudomonas turukhanskensis 99.15

Proteobacteria Gammaproteobacteria Pseudomonas turukhanskensis 99.1

Proteobacteria Gammaproteobacteria Pseudomonas turukhanskensis 99.56

Proteobacteria Gammaproteobacteria Pseudomonas turukhanskensis 99.24

Proteobacteria Gammaproteobacteria Pseudomonas turukhanskensis 99.49

Proteobacteria Gammaproteobacteria Pseudomonas turukhanskensis 99.72

Proteobacteria Gammaproteobacteria Pseudomonas helmanticensis 99.65

Proteobacteria Gammaproteobacteria Pseudomonas helmanticensis 99.53

Proteobacteria Gammaproteobacteria Pseudomonas helmanticensis 99.72

Proteobacteria Gammaproteobacteria Pseudomonas helmanticensis 99.76

Proteobacteria Gammaproteobacteria Pseudomonas extremaustralis 99.53

Proteobacteria Gammaproteobacteria Pseudomonas extremaustralis 99.78

Proteobacteria Gammaproteobacteria Pseudomonas extremaustralis 99.7

Proteobacteria Gammaproteobacteria Pseudomonas simiae 99.55

Proteobacteria Gammaproteobacteria Pseudomonas simiae 99.62

Proteobacteria Gammaproteobacteria Pseudomonas simiae 99.21

Proteobacteria Gammaproteobacteria Pseudomonas baetica 99.32

Proteobacteria Gammaproteobacteria Pseudomonas baetica 98.8

Proteobacteria Gammaproteobacteria Pseudomonas canadensis 100

Proteobacteria Gammaproteobacteria Pseudomonas kilonensis 98.89

Proteobacteria Gammaproteobacteria Pseudomonas paralactis 99.93

Proteobacteria Gammaproteobacteria Pseudomonas costantinii 99.4

Proteobacteria Gammaproteobacteria Pseudomonas brenneri 99.76

Proteobacteria Gammaproteobacteria Rugamonas rubra 100

Proteobacteria Gammaproteobacteria Rugamonas rubra 99.74

Proteobacteria Gammaproteobacteria Aeromonas aquatica 100

Proteobacteria Gammaproteobacteria Buttiauxella ferragutiae 100

Proteobacteria Gammaproteobacteria Rhizobacter dauci 100

Proteobacteria Gammaproteobacteria Acinetobacter bohemicus 99.74

Proteobacteria Betaproteobacteria Janthinobacterium svalbardensis 99.88

Proteobacteria Betaproteobacteria Janthinobacterium svalbardensis 99.76

Proteobacteria Betaproteobacteria Janthinobacterium svalbardensis 99.87

Proteobacteria Betaproteobacteria Janthinobacterium svalbardensis 99.88

Proteobacteria Betaproteobacteria Janthinobacterium 100

Proteobacteria Betaproteobacteria Janthinobacterium 99.65

Proteobacteria Betaproteobacteria Janthinobacterium 99.88

Proteobacteria Betaproteobacteria Janthinobacterium 99.53

Proteobacteria Betaproteobacteria Janthinobacterium lividum 100

Proteobacteria Betaproteobacteria Janthinobacterium lividum 100 Firmicutes Bacilli Bacillus weihenstephanensis 99.81 Firmicutes Bacilli Bacillus weihenstephanensis 99.54 Firmicutes Bacilli Bacillus wiedmannii 100

Bacteroidetes Sphingobacteria Pedobacter duraquae 98 Bacteroidetes Flavobacteria Flavobacterium branchiarum 98.16

5.3 Discussion The physical design was designed largely in accordance with previously designed in situ cultivation platforms [108]. The chapter successfully developed the necessary components for loading, incubation, and harvesting of a known number of soil organisms. The developed ISSA platform was shown to lead to uniform loading of either producer strains or producer strains and a reporter strain, with little crosstalk between compartments. The shown capacity of the ISSA platform to mix or separate the producer strain and the reporter strain allowed for further exploration of antibiotic production in later chapters. Furthermore, the developed ISSA platform was used to build a library

89 of uncultivable bacteria. The overall domestication rates found (15-20 %) were consistent with prior cultivation platforms [115]. The analysis of cultivated bacteria furthermore showed that the ISSA platform could domesticate and identify novel species. This corresponded well with phylogenetic data from other in situ platforms. Ben Dov et al. [117] noted that they struggled to transfer their cultivation rates in situ to domesticated species that showed growth in the laboratory. The breakdown by Nichols et al. [108] similarly showed the Ichip could domesticate only limited diversity in cultivated phyla (Proteobacteria or Firmicutes). The lack of phylogenetic diversity of the strains identified by the ISSA therefore confirms existing barriers to the cultivation of more fastidious phyla already noted in the literature.

In this project the platform was, however, used exclusively as a single-round platform, using the growth media SMS, for period of 2-3 weeks. Identified variations that could boost cultivation rates have, however, been identified in the literature. Cultivation rates could potentially be increased through multiple rounds of domestication [115]. Additionally, shifts in growth media in ISSA compartments could potentially also provide access to different phyla, as this has been shown to be the case when bacteria are isolated in vitro [103]. Finally, longer in situ incubation periods should facilitate an increase in domestication rates, as most soil bacteria have been shown to be slow growers [88, 105, 106].

Chapter 6. Development of luminescent reporter strains

6.1 Introduction Developing robust and luminescent bioreporter strains was critical for the ISSA platform, as it is the switching off of the bioreporter signal that indicates antibiotic discovery. It was therefore essential to incorporate a high intensity and stable bioreporter that switched off when the bioreporter strain was killed/inhibited.

6.1.1 Environmental bacterial bioreporters Environmental bioreporters have been used for a variety of reasons, including the detection of general environmental toxicity [159-161], specific organic pollutants [162, 163], and specific heavy metal pollutants [164, 165]. By targeting particular genetic mechanisms, genetic engineering can be used to build cell circuits with controllable on/off signal switches tuned to a particular chemical, chemical class, or toxicological effector [166]. Bioreporter signals include bacterial luciferase, firefly luciferase, Green Fluorescent Protein (GFP), and β-Galactosidase (lacZ) [166]. The luxCDABE operon has gained prominence as a biosensor signal [166]. One advantage associated with the luxCDABE operon, is that it depends on metabolic energy to emit luminescence. Consequently, biostatic or dead cells cannot emit luminescence [167]. luxCDABE therefore provides a direct indicator of the metabolic state of the reporter strain population. Additionally, the luxCDABE operon needs no further substrate addition to the media and the signal is easily detected using luminescence detectors [125]. The luxCDABE operon has also been synthetically optimized to be integrated and expressed in a wide variety of bacterial hosts, including both Gram-positive and Gram-negative strains [168].

6.1.2 Bioreporter strain selection The choice of reporter strain was a crucial component in the overall development of this project. Antibiotic resistance is a phenomenon that encompasses a broad range of bacteria across the bacterial kingdom. There were accordingly a multitude of bacterial targets that could be incorporated into the ISSA platform as reporter strains. To increase both the impact and chance of success of the project, the reporter strain development had to be focused on a select group of bacterial strains. The following factors were considered when choosing strains for reporter strain development: clinical relevance, genetic tractability, abilities to survive in soil, and human and environmental safety. Clinical relevance was based on the newly published WHO report [26]. The

91 report lists three threats as critical: carbapenem-resistant Acinetobacter baumannii (A. baumannii), carbapenem-resistant Pseudomonas aeruginosa (P. aeruginosa), and carbapenem- resistant (3rd generation cephalosporin resistant) Enterobacteriaceae. The Enterobacteriaceae family includes a further three targets on the WHO list [26]. Enterobacteriaceae was therefore deemed of clear clinical relevance. The Enterobacteriaceae family has several member-species, which are well studied and genetically tractable - most noteworthy E. coli. A variety of strains of E. coli have been documented to survive in soil over prolonged periods of time [169], including the clinically relevant strain O157:H7 [126]. The high genetic tractability of E. coli has meant that many pathogenic strains have been rendered harmless through genetic engineering. When e.g. verotoxigenic genes are removed from O157:H7 [169], it poses limited to no risk for human health [126, 169]. P. aeruginosa is genetically tractable using many of the same standard laboratory techniques used for E. coli. Furthermore, the Pseudomonas genus encompasses several soil- dwelling species, including P. putida and Pseudomonas fluorescens (P. fluorescens). As soil bacteria, both P. putida and P. fluorescens can survive in low nutrient environments for extensive periods of time [170]. A. baumannii is an opportunistic nosocomial pathogen almost exclusively associated with hospital infections [171]. Pathogenesis includes bloodstream infections with associated mortality rates ranging between 8-35 % dependent on type of infection and strain [172]. A. baumannii has a relative recent history of pathogenesis, and molecular tools are consequently, at present, under development [173]. Given the clear clinical relevance and genetic tractability E. coli and Pseudomonas spp. were consequently chosen as ISSA platform bioreporters.

When developing the bioreporter strains, the primary considerations were luminescent intensity and luminescent stability. Luminescent intensity was important, as the small size of ISSA compartments meant that only a limited number of bioreporters could be loaded into each compartment. Luminescent stability was deemed essential to limit downstream false positives. The rate of soil bacteria that produce antibiotics has been estimated to be 1:10,000 [118]. A false positive rate much higher than 1:10,000 would consequently make downstream identification of antibiotic-producing organisms difficult. Chromosomal integration was therefore chosen. Weitz et al. [125] have developed a promotor-less luxCDABE operon transposon (pUT mini-Tn5 luxCDABE – see Fig 6.1 A), which causes random integration in both Pseudomonas spp. [125] and E. coli spp. [126] through conjugation Ritchie et al. (2003) developed chromosomally integrated luxCDABE biosensor strains (lux-integrants) of a non-toxigenic strain of E. coli O157:H7 [126], which remained

92 viable and luminescent over time in challenging conditions see (Fig 6.1 B). In a survival study, starved populations of E. coli O157:H7 Tn5 luxCDABE remained viable and luminescent after being placed in a natural environment for > 4 weeks - albeit with a lower cell population and lower luminescence per cell [126]. The pUT mini-Tn5 luxCDABE therefore provided a tested solution for stable luminesce integration in the desired bioreporter genera.

Fig 6.1. E. coli bioreporter development. A pUT mini-Tn5 luxCDABE harbouring the luxCDABE cassette flanked by a kanamycin resistance gene. Figure taken from J. H. Weitz et al (2001) [125]. B Survival of chromosomally integrated E. coli O157:H7 strain in artificial groundwater at 15°C. (▪) represent the mean of three replicates ± standard deviation as assessed by culturable cell counts (○) represents luminescence (pRLU) per cell. Data represent the mean of three replicates ± standard deviation. Figure and description modified from Ritchie et al. [126].

A second focus was to develop Gram-positive bioreporter strains for in vitro verification of broad- spectrum activity of identified producer strains. Here Mycobacterium smegmatis (M. smegmatis) and Bacillus subtilis (B. subtilis) were chosen as two diverse strains representing Actinomycetes and Firmicutes that were genetically tractable and safe to use in category 1 safety conditions. M. smegmatis is a soil organism that is often used as a safe model organism for M. tuberculosis [174- 176]. The increased prevalence of multi-resistant strains of M. tuberculosis has meant that discovering novel antibiotics targeting this strain has taken on new importance [177]. B. subtilis is part of the Firmicutes phylum, which contains several antibiotic-resistant threats including methicillin-resistant and vancomycin-resistant Staphylococcus aureus and penicillin-non- susceptible Streptococcus pneumoniae [26]. LuxCDABE operons have been developed for both M. smegmatis and B. subtilis. For M. smegmatis, Andrea et al. [129] developed the

93 plasmid pMV306hsp+LuxG13 (Addgene plasmid # 26161 – see Fig 6.2 A), which harbours the full lux operon but does not contain an integrase, and the pBS-Int (Addgene plasmid # 50000 – see Fig 6.2 B), which is a suicide plasmid that carries the integrase gene but no attachment site. Transformation via electroporation of M. smegmatis with the pMV306hsp+LuxG13 and pBS-Int plasmids has been shown to lead to site-specific integration of the luxCDABE operon. The integration was shown to be retained in > 99 % cells after four weeks [129]. For B. subtilis, Radeck et al. [128] developed the plasmid pBS3Clux (Addgene plasmid # 55172 - see Fig 6.2 C ). The pBS3Clux is a promoterless luxCDABE operon with restriction sites upstream to allow for easy integration of suitable promotors [128]. The plasmid leads to site-specific integration and variance in the luminescence profile is therefore caused mainly by the choice of promoter. The promotor plasmids Pveg (Addgene plasmid # 55172 - see Fig 6.2 D) and PlepA (Addgene plasmid # 55175) were specifically designed for integration, but other operons can be designed synthetically by adding appropriate upstream and downstream restriction sites [128].

Fig 6.2. Plasmid maps for integration of luxCDABE operons in M. smegmatis and B. subtilis. A plasmid maps for M. smegmatis. B Plasmid maps for B. subtilis. All plasmid maps modified from the originals found on www.addgene.org.

6.2 Results

6.2.1 Testing prior published strains

The O157:H7 Lux strains (RA1, RA2, GC1, And GC2 published in Ritchie et al. [126]) were kindly donated by Ken Forbes . The strains are non-pathogenic E. coli O157:H7 and had been documented

95 to survive in soil for >4 weeks [126, 169]. These published strains were tested for their suitedness as bioreporters in the ISSA platform. GC1, GC2, RA1, and RA2 were loaded into 384-compartment ISSA array according to the vacuum-assisted loading protocol developed in Chapter 5. The initial results showed homogenous distribution (see Fig 6.3 left). After two weeks of incubation in soil, the luminescence had reduced to an extent where it was difficult to detect in a majority of wells (see Fig 6.3 right). The plates were subsequently unsealed, and the integrity of the plugs were successfully verified through confirmation by eye.

Fig 6.3. Luminescence of GC2 bioreporter before and after incubation in soil. To increase visual detection, the brightness was increased and the contrast decreased in the day 14 image.

6.2.2 Protocol for development of lux integrants The high false positive rate exhibited in Fig 6.3 combined with the low expected rate of antibiotic producers (1:10,000) meant that a more robust luminescent strain had to be developed for the ISSA platform to be operational. The previously developed pUT mini-Tn5 luxCDABE transferase [125] was used to develop P. putida and P. fluorescent reporter strains according to the protocol published in Weitz et al. [125]. The protocol relied on growing the donor strain containing the transferase and the designated receiver strain overnight. The two strains were then mated for 24- 48 h on LB (no antibiotics). The biofilm produced by the mating was subsequently plated out on kanamycin plates (50 mg/mL) to select reporter strains with luxCDABE operons that had been correctly integrated upstream of an endogenous promoter. After applying selective pressure using the kanamycin marker present on the pUT mini-Tn5 luxCDABE plasmid, colonies were screened for luminescence. The protocol resulted in approximately 100-200 lux integrants per experiment (see Fig 6.4 left). To increase the number of luxCDABE integrants (lux integrants) in each mating and thereby the chance of an advantageous integration, ampicillin was incorporated into the

96 conjugation stage. The addition of ampicillin placed a selective pressure on the donor strain to retain the tn5 transferase and was therefore theorised to increase the active period of conjugation. Both P. putida and P. fluorescens were found to be naturally resistant to ampicillin, which meant that they could be mated on ampicillin plates without further modifications. The conjugation step was set to 48 h to allow maximum time for integrations to take place. Finally, the quantity of mixed media (that could generate biofilm) was increased from 100 µl to 500 µl and the petri dishes used for conjugation were increased from 90-mm ø plates to 150-mm ø plates. The generated biofilm was diluted and plated out on 30 150-mm ø kanamycin plates. An estimated 30,000 CFUs exhibited luminescence per experiment. The changes in the protocol therefore resulted in a 150-300 fold increase in the rate of viable lux integrants. Given the large number of lux integrants, a screening step was incorporated into the protocol, so that the most luminescent strains were selected for further analysis. To test the increase in efficiency gained by the screening process 96 colonies were chosen randomly and 96 colonies were chosen after having pre-screened for luminescence. The results showed a marked increase in average and peak luminescence when incorporating a screening step (see Fig 6.4 B).

Fig 6.4. Developing lux integrants. A Comparison of the original and optimised protocol for generating luminescent reporter strains. Left: total number of luminescent integrants when using protocol developed in Weitz et al. [125]. Right: luminescent readout of 1 of 30 plates when using the protocol described in Section 6.2.2. B Comparing the luminescence of 96 luminescent integrants when picking lux integrates with or without screening them first for luminescence. Luminescence was quantified using a BMG LABTECH CLARIOstar.

After 9 h growth on kanamycin LB agar plates at 37 °C, 292 colonies were selected by strength of luminescence emission for further analysis. The 292 lux integrants were incubated for one week in LB agar (no antibiotics) at 30°C and tested for luminescence periodically. The 17 lux-integrants that were deemed to have the highest and most stable luminescence were purified. Four colonies

97 from each purified reporter strain were incubated in 200 µL 2.0 % agar 2xTY media (with an initial cell population of 2 × 106) and 200 µL 2.0 % SMS agar (with an initial cell population of 2 × 103) in 96-well plates for five weeks (see Fig 6.5 A for an overview of the lux integrant selection protocol). This meant that high nutrient and high initial population data could be compared to low nutrient and low population data. 17 data points were taken over the five weeks and the resulting time course data showed that all replicates of both the here developed strains P1H2 (P. putida) and FH1 (P. fluorescens) had substantially higher luminescent output in all conditions, and at all points after day 0 (see Fig 6.5 B). Furthermore P1H2 and FH1 stayed luminescent for significantly longer than the previously developed strains RA1, RA2, GC1, GC2 [126] and lux P. putida and lux P. fluorescens [125]. Furthermore, P1H2 and FH1 both exhibited significant detectable luminescence for all five weeks. Five other developed strains exhibited similar results to P1H2 and FH1. P1H2 exhibited the highest average luminescence and the smallest variance in the 2.0 % agar SMS and was consequently chosen as the primary reporter strain for ISSA incubation as 2.0 % agar SMS most closely resembled in situ conditions.

Fig 6.5. Protocol for the selection of integrants with high and stable luminescence. A After integration the colonies were grown overnight on 15 cm ø petri dishes with kanamycin (50 mg/ml).The highest luminescent emitters were isolated in individual 96-well plate wells and grown in LB at 37˚C for one week. The top performers across the week were then isolated and each integrants was tested in solid media (SMS and LB) for 5 weeks. B Comparison of luminescence emission over time of the newly developed reporter strains P1H2 (P. putida) and FH1 (P. flourescens) with the donated strains GC2 and P. flourescens luxCDABE (here abbreviated as FLU).

In situ incubation with the developed reporter strain P1H2 was carried out for three weeks in the 384-compartment ISSA format, leading to 100 % compartments detected post-incubation. The experiment was then repeated in the 5180-compartment ISSA format, using sequential loading. The sequential loading in the 5180-compartment ISSA resulted in > 99.99 % coverage (see Fig 6.6). The luminescence emission of P1H2 was consequently deemed sufficiently intense and robust to be used in in vitro and in situ ISSA screens going forward.

Fig 6.6. Luminescence of P1H2 before and after in situ incubation. The experiment was set up in accordance with the sequential loading protocol and the ISSA array was incubated in situ for 20 days. Luminescence was easily detected even after day 21 showing that P1H2 remains luminescence for extended period during in situ incubation.

6.2.3 Development of secondary reporter strain for in vitro experiments A second aim was to develop luminescence reporter strains for the detection of broad-spectrum antibiotic producers among producer strains already identified via the ISSA-platform (secondary reporter strains). GC2 provided an E. coli reporter strain that was sufficiently stable to be used in standard laboratory conditions. To extend the phylogenetic range of secondary reporter strains, a luxCDABE cassette was integrated into the strain M. smegmatis mc2155 according to the protocol described in Andreu et al. [129]. The resulting luminescent strain was labelled M. smegmatis G13 Lux. The strains B. subtilis 168 and Mini Bac were chosen for B. subtilis reporter strain development. B. subtilis 168 has a long history of being used for genetic modification [178], whereas the Mini Bac strain has been developed for use in synthetic biology work and has a minimal genome, is non-sporulating, and is easy to transform [127]. B. subtilis 168 and Mini Bac luxCDABE integrants with the promoters Pveg and PlepA were therefore developed using the protocol described in Radeck et al. [128]. After successful integration of the reporter strains 100- 1000 cells were mixed into 200 µl of 2.0 % agar 2xTY media and incubated for 14 days at 28°C. The results showed that M. smegmatis G13 Lux and Mini Bac Pveg Lux both exhibited significant

100 luminescence for up to 7 days. Luminescence emission was compared to wells containing no reporter strains using a 99.9 % confidence interval (one sided t-test. h0-hypothesis: no emitted luminescence). The luminescent profiles exhibited by M. smegmatis G13 Lux and Mini Bac Pveg Lux allowed for these developed reporter strains to be used for classical in vitro experiments for antibacterial compound production, which in most experiments consist of 5-7 day coincubation with producer strains in classical incubation medias [154, 179], or spotting of producer strain extract on reporter strains grown in high nutrient media overnight [118, 125].

Fig 6.7. Luminescence over time of Mini Bac Pveg Lux and M. smegmatis G13 Lux.

6.3 Discussion This chapter optimizated the previously developed Weitz protocol [125] to enable a 300-fold higher integration of developed luxCDABE strains. The resulting P1H2 (P. putida) and FH1 (P. fluorescens) showed consistent and high luminescence over five weeks of in vitro incubation, even at low initial cell concentrations (100 cells) and in low nutrient media. P1H2 further showed high luminescence and stability when incubated in soil for 3 weeks > 99,9 % wells could be detected after 20 days of in situ incubation. The developed strains exhibited higher and more constant luminescence over a variety of conditions than previously published strains [126] and could therefore provide value to other researchers in need of strong and stable long-term bioreporter strains that can survive in challenging environments. The consistent and strong luminescence exhibited by P1H2 (and potentially FH1) therefore facilitates the detection of antibacterial producers in the ISSA format, if antibiotic production can switch off the luminescence. The developed secondary reporter strains M. smegmatis G13 Lux and Mini Bac Pveg Lux exhibited

101 sufficient luminescence to detect luminescence in high nutrient media for up to 7 days. This was also found to be the case for the developed luxCDABE E. coli strains donated by Weitz. As most classical antibiotic experiments have found that producer strains have reached peak production in less than 7 days, the consistency achieved in these reporter strains should facilitates downstream screening of identified antibacterial producers for activity against other pathogens.

Overall the developed strains provide sufficient reporter strains for primary identification (Pseudomonas putida) and secondary identification (M. smegmatis, B. subtilis, and E. coli) of antibacterial production in the ISSA platform. An operational ISSA platform is, however, theoretically able to screen soil bacteria for antibacterial compound production against any pathogen. Building sufficiently strong bioreporters for other known pathogens would therefore lead to a more flexible platform. The obvious next primary bioreporter to be built would be an E. coli strain as Enterobacteriaceae represent the third most urgent threat [26]. For this the optimised protocol for lux integration developed in section 6.2.2 would work. It is also worth noting that the strains we developed using random integration were significantly brighter and more stable than strains that had designated integration sites and therefore relied on optimised high expression promoters. Incorporating a random-integration component into the lux integration process might therefore be preferable to synthetic promoters, if future researchers decide to develop lux bioreporters for antibiotic screening using less well-established model strains for e.g. the most urgent threat A. baumannii [26].

102

Chapter 7. Understanding antibiotic secretion

7.1 Introduction The BACCU platform relies on the ability of one soil species to produce sufficient antibiotics to kill/inhibit the adjacent reporter strain. The aim of this chapter is to optimise the physical relationship between the producer strain and reporter strain, so that the producer strain has the highest possible chance of killing/inhibiting the reporter strain.

When high producing strains are grown in pure cultures and in high nutrient conditions that have been optimised for antibiotic production, they can produce antibiotic concentrations many times higher than the minimum inhibitory concentration (MIC) [180]. This could lead to the assumption that antibiotics are a powerful weapon that bacteria can use freely to fight off competitors. Natural conditions are, however, characterized by complex species interactions and low nutrient conditions [88]. The combination of low nutrients and high interspecies interaction in natural environments has been argued to render antibiotics ineffective, to the extent that antibiotics has been proposed by some researchers to be a mode of communications rather than a weapon to fight off competitors [153, 181]. If this hypothesis is true, it would be hard to elicit a sufficient antibiotic response during in situ incubation to knock out the reporter strain, as such quantities would not be needed to communicate.

There is, however, increasing evidence that when antibiotics are secreted in natural conditions, it is often to deter competitors [151, 154, 182, 183]. In vitro coincubation methods can be conditioned to mimic natural conditions, as the producer strain must produce antibiotics in real time, while in resource and/or territory competition with one or more reporter strains. Using nutrient poor agar plates Garbeva et al. [184] showed that P. fluorescens Pf0-1 could secrete bactericidal compounds in sufficiently high concentrations to inhibit growth of susceptible species when colonies were grown on two sides of an agar plate [184]. Similarly, Abrudan et al. [154] explored antibiotic production within the Streptomyces genus. They found that Streptomyces species upregulate antibiotic production as a response to competitors [154]. Based on their findings, it was determined that most antibiotic secretion by Streptomyces strains is used as a weapon to fight off specific competitors [154]. There is therefore evidence to suggest that antibiotics, at least in certain cases are used as a weapon during resource competition.

Assuming that antibiotics can be used as a weapon, there a three potential ways that the secreting

103 strains could use antibiotics to gain a competitive advantage when in competition [151].

1) Antibiotics allow invasion of habitats inhabited by competing species. 2) Antibiotics inhibit growth of competitors when the producer strain and the competitor are at comparable densities in the same habitat. 3) Antibiotics prevent the invasion of established habitats by competitors.

These hypotheses have an impact on the ideal physical relationship between the producer strain and the reporter strain in the ISSA format. If hypotheses 1-2 are true, the producer strain and reporter strain can be mixed in close proximity. If hypothesis 3 is true, the producer strain and reporter strain would require spatial separation for antibiotic production to be effective. Wiener [151] used Streptomyces griseus (a streptomycin producer) as a producer strain and resistant and non-resistant B. subtilis as reporter strains to investigate the above hypotheses [151]. It was found that S. griseus could only use antibiotics to gain an advantage when defending established territory. This would mean that spatial isolation would increase the sensitivity of the ISSA platform to antibiotic secretion as the producer strain would then have an established territory to defend.

To explore this hypothesis, proof of concept experiments had to be set up. Mixing the producer strain and reporter strain in liquid media prior to inoculation in agar allowed for the exploration of antibiotic production when the producer strain and reporter strain are in proximity. Mixing the producer strain with the agar media and allowing it to solidify before the reporter strain was added on top allowed for physical separation through layering.

To test the effectiveness of mixing/layering, model reporter strains and model producer strains had to be identified. The luxCDABE strains developed in Chapter 6 were developed to be reporter strains in such experiments. Strains that had been documented to have high production of potent broad-spectrum antibiotics were identified in the literature were selected as producer strains. The selected strains were all Streptomyces species and included: Streptomyces aureofaciens (S. aureofaciens), Streptomyces venezuelae (S. venezuelae), Streptomyces cattleya (S. cattleya), and Streptomyces fradiae (S. fradiae). Streptomyces is a genus of filamentous bacteria that produce most of the antibiotics used today [185]. S. aureofaciens is a producer of several tetracyclines, including tetracycline [186] and has been shown to yield 2150 μg/ml tetracycline in batch reactions in synthetic media [179]. S. venezuelae is an often-used model strain in

104

Streptomyces research [187]. S. venezuelae is associated with shorter mycelia than most other Streptomyces strains and is the strain most heavily associated with chloramphenicol (CM) production [188]. Chloramphenicol is a broad-spectrum antibiotic affecting both Gram-positive bacteria and Gram-negative bacteria [188]. The production of chloramphenicol in S. venezuelae has been noted to be low or sporadic in the literature [189, 190]. S. venezuelae has also been reported to produce jadomycin B (up to 50 μg/ml) when exposed to heat shock or ethanol shock [191]. S. cattleya produces thienamycin, cephamycin C, and penicillin N [192]. Thienamycin has been described as the most potent of the naturally produced antibiotics [193]. Due to thienamycin being highly resistant to bacterial β-lactamases, it shows an extraordinary broad spectrum of activity [194], including against multi-resistant Pseudomonas strains [192]. Finally, S. fradiae produces neomycin [179], tylosin [22], and fosfomycin [195]. Neomycin is a broad-spectrum antibiotic [179]. Production of neomycin in S. fradiae has been documented at > 170 μg/ml in batch reactions [196].

7.2 Results

7.2.1 Comparing S. aureofaciens and S. venezuelae as producer strains To carry out antibiotic production assays, it was essential to be able to quantify and handle a known number of dispensed producer cells. S. venezuelae was characterized by short hyphae chains when grown in liquid media. The short hyphae chains made it easier to quantify cell numbers compared to the other three model strains selected, which all exhibited more extreme hyphae growth. The inability of S. venezuelae to produce high quantities of chloramphenicol reported in the literature, however, questioned the ability of S. venezuelae to create consistent clearance zones in the ISSA format. The ability of S. venezuelae to clear reporter strains was therefore compared to the established high antibiotic producer S. aureofaciens by carrying out in vitro co-incubation experiments [5]. S. venezuelae and S. aureofaciens were grown in standard MYM media. S. venezuelae and S. aureofaciens were then spun down and serially diluted onto six solid medias that had been identified as antibiotic production medias in the literature (MYM, GYM, YM, ISP4, YEMES, and GI production medium). GI production media was chosen as it has been use for optimised chloramphenicol production in S. venezuelae [197]. Each serial dilution had three replicates. The serial dilutions were grown for five days at 28°C to allow for antibiotic production. The petri dishes with 50-500 distinct colonies were selected and a 2.5 mL thin layer of LB 2.0 % agar mixed with the reporter strain GC2 was placed on top of the producer strain. It was observed

105 that S. venezuelae could not produce consistent clearance zones, even in GI production media (Fig 7.1 left panel). S. venezuelae often only created local clearance zones (clearance was only observed close to clusters of S. venezuelae colonies). The rest of the petri dish exhibited “milky” areas, where the GC2 reporter strain was able to grow unimpaired. When carrying out similar experiments, S. aureofaciens was able to consistently produce global clearance zones, as no trace of the reporter strain could be detected in any area of the petri dish (see Fig 7.1 right panel), indicating that it was a more consistent antibiotic producer. Overall S. aureofaciens produced global clearance zones on 14 of 18 petri dishes, with S. venezuelae only producing global clearance zones on 2 of 18 petri dishes.

Fig 7.1. Clearance zones caused by antibiotic secretion by S. venezuelae and S. aureofaciens after 5 days in YM media.

7.2.2 Protocol development for reduction of hyphae by Streptomyces species grown in liquid media S. aureofaciens grew in large hyphae when cultured in liquid media. A method was consequently developed to break up cell clusters and thereby increase the CFUs in the media. S. aureofaciens was grown in 10 ml of GYM media that contained 10-15 2 mm diameter glass beads. The glass beads created mechanical sheering of the mycelia that led to substantially smaller colony sizes. The cells were then passaged to GYM media mixed with five different detergents and glass beads (Tween20, Tween80, Brij-35, Triton, and SDS). Three detergents (Tween20, Tween80, and Brij-35) showed promising results in breaking down mycelia, while still allowing cell growth. The ability of these detergents to break down mycelia was quantified after filtration with 35 μm filters and 5 μm

106 filters (see Table 7.1 for resulting OD600 figures). It was hypothesized that the solution with the highest concentration of small hyphae chains (approximately 100mers for the 30 μm filter, and

5mers for the 5 μm filter) would have the highest associated OD600 after filtration, as small hyphae chains could pass the 35 μm and 5 μm and filters. Visual observation using light microscopy confirmed this hypothesis.

Table 7.1. OD600 of S. aureofaciens using different detergents Tween20 (1.0 %) Tween80 (1.0 %) Brij-35 (0.5%) No detergent No filter 2.05 2.05 2.05 2.05 30μm filters 1.65 1.55 2.05 0.25 5μm filters 0.07 0.05 0.4 0.01

As can be seen in Table 7.1 the detergents had no impact on OD600 prior to filtration, suggesting that growth was not deterred. Table 7.1 also suggests that Brij-35 was the best performer, as cultures grown in Brij-35 exhibited the highest OD600 post-filtration at both 30-μm filters and 5-μm filters. The final mycelia-reducing growth protocol consisted of growing S. aureofaciens for two- three days in 10 ml MYM media with 10-15 glass beads (the switch from GYM TO MYM was due to observed increases in antibiotic production by the producer strains – data not shown). The culture was then passaged (1/30 dilution) to 10 ml of MYM media with 10-15 glass beads containing 0.125-0.5 % Brij-35. After another two days the culture could be filtered with 5 μm filters to remove remaining hyphae. The protocol allowed for a dramatic reduction of hyphae (see Fig 7.2 A-C). Additionally, the protocol did not affect the ability of S. aureofaciens to secret antibiotics (see Fig 7.2 D-E). Finally, the protocol could also be used to break up the hyphae when growing S. cattleya and S. fradiae in liquid culture.

107

Fig 7.2. S. aureofaciens grown in various conditions. (A-C) light microscopy pictures 100x. A S. aureofaciens in GYM media with glass beads. B S. aureofaciens cultured with GYM media with 0.5 % Brij-35 and glass beads. C S. aureofaciens cultured with 0.5 % Brij-35 and glass beads and filtered with 5μm filters. (D-F) Coculturing of S. aureofaciens and GC2 in various conditions. Cut outs from S. aureofaciens were grown 5 days in MYM 2.0 % agar media. A sphere was cut out and placed in the centre of a petri dish. GC2 mixed into the LB agar media was then poured on top and around the S. aureofaciens cut-outs. The dark haloes exhibited in D and E represent clearance zones due to antibiotic secretion. D unfiltered S. aureofaciens on Tet opt synthetic media (5 days) E S. aureofaciens on GYM media (5 days) after having been cultured with Brij-35 and 2mm glass beads and filtered with 5 mm filters. F S. aureofaciens on GYM media after insufficient time (3 days) for S. aureofaciens to produce sufficient antibiotics to prevent GC2 from growing uninhibited.

7.2.3 Exploring antibiotic production in different setups when the reporter strain and the producer strain are introduced simultaneously Cocultivation assays were set up in standard 96-well plates to explore the abilities of the designated producer strains to kill/inhibit the reporter strains in different media. Approximately 2 × 105 CFUs from the producer strains (S. fradiae, S. venezuelae, S. aureofaciens, and S. cattleya) were mixed with 20-200 CFUs of the four developed reporter strains (P1H2 (P. putida), GC2 (E. coli), Mini Bac Pveg lux (B. subtilis), and M. smegmatis G13 Lux) in 225 μl volumes of six solid medias (2xTY, R2A, YEMES, MYM, MYM + CaCo3, and ISP4). A matrix was set up to test all combinations of producer strains, reporter strains, and media in a 96-well plate. Four plates were set up, with each 96-well plate having a unique reporter strain mixed into every chamber. On each

108 plate, every two rows had a separate producer strain and every two columns had separate growth media. This meant that all tested conditions were replicated four times (in local 2 x 2 matrices). Finally, a “blank” plate was set up where each reporter strain was grown in pure culture in all six growth medias. On the blank plate, every two rows had a separate reporter strain and every two columns had separate growth media. The final matrix was therefore composed of the “cocultivation” 4 x 4 x 6 x 4 matrix and the “blank” 4 x 4 x 6 matrix, and consequently consisted of a total of 480 wells. The experiment was run in a standard incubator at 28°C, but a 37°C heat shock was carried out for the first hour to boost antibiotic production. It was found that wells started to dry out after six days and the experiment was terminated. Only P1H2 and Mini Bac Pveg lux consistently emitted luminescent in pure culture in all growth medias after day 6. The results showed that Mini Bac Pveg lux was generally more sensitive than P1H2 to antibiotic production (see Fig 7.3). This was later confirmed by spotting known concentration of antibiotics onto agar plates with P1H2 and Mini Bac Pveg Lux (data not shown). The matrix experiment also showed that S. aureofaciens produce sufficient antibiotics in a variety of conditions to eliminate any signal from various reporter strains. S. aureofaciens turned off luminescent emission (established as luminescence emission of < 10 % of the comparable blank wells) in all four replicates in 23/24 tested conditions after 3 days. The matrix therefore provided strong evidence that antibiotic production had caused marked changes in luminescence of reporter strains, as S. aureofaciens (a known high producer) was more efficient in switching off reporter strains than S. venezuelae (a known low producer) across a variety of medias and because Mini Bac Pveg Lux, which is more sensitive to antibiotic production, was turned off more often than the less sensitive P1H2.

Fig 7.3. Testing the ability of model producer strains to kill/inhibit the reporter strains P1H2 and Mini Bac Pveg Lux. The luminescence readout was taken after 3 days of mixed cucultivation of

109 producer and reporter strains. Numbers denote different solid medias (2.0 % ΜLGTA): 1 = 2xTY, 2 = R2A, 3 = YEME, 4 = MYM+ CaCO3, 5 = MYM, 6 = ISP4.

7.2.4 Exploring the effect of layered and mixed setups on platform sensitivity to antibiotic production The experiments discussed in Section 7.2.3 showed that cocultivation can indeed lead to sufficient antibiotic production by producer strains to kill and or inhibit reporter strains. To minimise the rate of false positives and false negatives of the ISSA platform, it was, however, deemed important to explore which platform setup yielded the most robust results when using model organisms. Supported by the data from Wiener et al. [151], it was hypothesised that spatial separation of the producer strain and reporter strain could yield increases in sensitivity to antibiotic production. To explore this hypothesis, two different set-ups were explored in 96-well plates (see Fig 7.4 A for concept illustration). In the first setup, a known number of producer strains were pipetted into the wells and allowed to solidify, before a known number of reporter strains were placed on top (described as a “layered” setup). In the second setup, a known quantity of producer strains and reporter strains were mixed and placed in a well (described as a “mixed” setup). The producer strain S. aureofaciens was serially diluted with CFU counts of 104, 103, 102, 10 CFUs and 0 CFUs (Blank). The reporter strain Mini Bac Pveg Lux was incubated in two concentrations: 105 CFUs and 102 CFUs. Mini Bac Pveg Lux was either “mixed” into the wells or “layered” on top. This resulted in 5 x 2 x 2 matrix, which explored the antibiotic sensitivity of three variables (the initial concentration of producer strain, the initial concentration of the reporter strain, and the mixed/layered setup). All 20 possible conditions of the matrix were set up, with each condition replicated four times. After setting up the experiment, The matrix was heat shocked at 37°C for one hour and subsequently incubated at 28°C in a humidity chamber to prevent the wells from drying out. All blank wells could be detected at all points during the 14 days for both layered and mixed wells, which showed that the reporter strain Mini Bac Pveg Lux remained stable in MYM media (See Fig 7.4 B for readouts at day 1, day 7 and day 14).

110

Fig 7.4. Platform sensitivity to antibiotic production in layered and mixed setups. A The hypothesis tested was that layering could lead to increased antibiotic production sensitivity. The layered setup separated the producer strain and reporter strain by constricting the reporter strain to a small at one side of the well (here the top). The mixed format did not separate producer strains and reporter strains. The setups had identical CFU counts for both producer and reporter strain. B Detected luminescence for the two setups over time.

7.2.5 Identifying significant variables using multivariable regression To test whether the difference in luminescence associated with each variable was statistically significant, the experiment was replicated and the luminescence of each well over time was integrated and quantified. A multivariate regression was run using the average luminescence of wells of the day 14 combined data from the two 80-well experiments. Two wells from the replicate experiment had to be excluded due to dispensing errors, leaving a combined 158 data points. Through central limit theorem, the data could be analysed using multivariable regression [198, 199]. The luminescence of each well was integrated, and average value was used. The luminescence of each well was then standardised, so that different plates (original/replicate) and setups (mixed/layered) could be compared. To standardise the wells, each well was divided with

111 the average value of the blank wells from the same setup (layered/mixed) and plate (original/replicate). By dividing values with the average blank values, values under 1,000 meant a decrease in luminescence compared to average blanks and therefore potentially suggested that the reporter strain had been inhibited by antibiotic production. The explanatory variables incorporated into the multivariable regression were:

• Mixed/layered (taking values 0 or 1 respectively), • Low/high initial concentration of the reporter strain (taking values 0 and 1 respectively). • Log(value) of the initial concentration of producer strains (values ranged from 1-4).

And the null-hypothesis was that none of these variables explained the variation in luminescence. The multivariate regression can be found in Table 7.2. The blank values were not included in the regression as they had already been included in the standardisation of other wells. The total number of observations was therefore 126. The multivariable regression explained 29.9 % (R2 = 0.299) of the variance in standardised luminescence. The mixed/layered variable and producer strain variable were both found to be highly significant. The layered well-format led to a decrease of approximately 38 % in expected standardised luminescence compared to the mixed-well format. Furthermore, 10 fold increases in the initial concentration of producer strains resulted in a decrease of approximately 27 % in expected standardised luminescence. Finally, the initial concentration of reporter strains was found to have only a minor impact. A 1000 fold increase in initial reporter strain concentration resulted in an increase in expected standardised luminescence of around 23 %. The P-value of the reporter strain concentration was, however, below the threshold of P < 0.01, meaning that the reporter strain variable could not be confirmed as significant. Finally, the intercept value of 1.471 deviated significantly from the expected blank value of 1.000 with a significance of P < 0.001, suggesting that the model failed to accurately predict blank values, which raised questions of the overall model fit.

112

Table 7.2. Multivariable regression of standardised luminescence from layered & mixed setup

Regression Statistics Multiple R 0.546 R Squared 0.299 Adjusted R Squared 0.281 Standard Error 0.581 Observations 126

ANOVA Df SS MS F Significance F Regression 3 17.512 5.837 17.308 1.992E-09 Residual 122 41.145 0.337 Total 125 58.657

Coefficients Standard Error t Stat P-value Lower 95% Intercept 1.471 0.145 10.113 7.829E-18 1.18327218 Layered/mixed -0.382 0.104 -3.688 3.383E-04 -0.5866945 Producer strain -0.265 0.046 -5.731 7.342E-08 -0.3569343 Reporter strain -0.225 0.104 -2.176 0.031 -0.4301782

To increase the accuracy of the model, the data was separated into two regressions; one for layered data and one for mixed data. When running a regression exclusively on data from the layered wells (see Table 7.3), the R2 value of the regression increased to 0.433. Again, the initial concentration of producer strain was highly significant and a 10x increase in the producer strain population led to an estimated reduction of luminescence of about 26 %. The initial concentration of reporter strains was statistically insignificant and could therefore be excluded as a variable from the model. Furthermore, the intercept of 0.972 fell within one standard deviation of 1.000, suggesting that the model could accurately predict blank values. Overall the model suggested that in the layered setup, wells with 10 CFUs would have a lower expected luminescence than blank wells. The regression therefore suggested that the layered setup was very sensitive to antibiotic production.

113

Table 7.3. Multivariable regression of standardised luminescence using only data from the layered setup

Regression Statistics Multiple R 0.658 R Squared 0.433 Adjusted R Squared 0.414 Standard Error 0.339 Observations 64

ANOVA Df SS MS F Significance F Regression 2 5.343 2.672 23.287 3.058E-08 Residual 61 6.998 0.115 Total 63 12.342

Coefficients Standard Error t Stat P-value Lower 95% Intercept 0.972 0.112 8.678 3.024E-12 0.748 Producer strain -0.258 0.038 -6.818 4.765E-09 -0.334 Reporter strain 0.026 0.085 -0.309 0.758 -0.195

Conversely when running a multivariable regression using only the mixed well data (see Table 7.4), the data looked less promising. Firstly, the R2 value was 0.229, which was much lower than R2 value for the layered model. The intercept value was estimated at 1.615, which was found to be significantly different from 1.000 at the 95 %-confidence interval, suggesting that the model overestimated the luminescence of blank wells. The high intercept value combined with the producer strain coefficient suggested that the mixed setup needed at least 103 initial producer strain CFUs to detect antibiotic production after day 14. Finally, the reporter strain was statistically significant at the P < 0.05 level, meaning that the mixed setup looked to be more sensitive to the initial concentration of reporter strains than the layered setup.

114

Table 7.4. Multivariable regression of standardised luminescence using only data from the mixed setup

Regression Statistics Multiple R 0.478 R Squared 0.229 Adjusted R Squared 0.203 Standard Error 0.737 Observations 62

ANOVA Df SS MS F Significance F Regression 2 9.521 4.760 8.754 4.69E-04 Residual 59 32.084 0.544 Total 61 41.605

Coefficients Standard Error t Stat P-value Lower 95% Intercept 1.615 0.244 6.607 1.25E-08 1.12573537 Producer strain -0.269 0.084 -3.211 0.002 -0.4369606 Reporter strain -0.485 0.187 -2.586 0.012 -0.8599899

The overall finding of the multivariable regression was that the initial concentration of the producer had a strong and significant effect on luminescence, with a higher initial concentration leading to lower luminescence. Furthermore, the layered vs mixed setup was found to have a strong and significant effect on luminescence with the layered approach being more sensitive to antibiotic production and producing luminescence variations that were less erratic. Finally, the initial concentration of reporter strains seems to be significant only for the mixed setup, suggesting that physical separation in layered format insulates the producer strain from effects associated with high initial concentration of the reporter strain.

7.2.6 Further analysis of the multivariable regression results To analyse the findings of the multivariable regressions, the standardised average luminescence of each initial concentration of producer strains for layered and mixed wells was plotted over the 14-day period of the experiment (see Fig 7.5 A). In layered wells (see Fig 7.5 A left), the two highest initial concentrations of producer strains (10,000 and 1,000 CFUs) dropped to standardised luminescence values of < 5.0 % after day 4 and remained below 10 % for the remaining period. The wells with a producer strain CFU of 100 exhibited a steady decline in luminescence and reached < 5.0 % standardised luminescence at day 14. Finally, the layer with initial concentration of producer

115 strains of 10 CFUs showed only a marginal decline. This confirms the finding for the multivariable regression, which estimated that the layered format could detect concentrations of down to 10 CFUs. It, however, also shows that time is an important contributor to sensitivity, as lower concentrations of reporter strains take longer to have an effect on the reporter strain luminescence.

The trends exhibited in mixed setups were markedly different. In the mixed setup, wells with the three highest initial concentrations of producer strains exhibited decreasing luminescence for the first four days (see Fig 7.5 A right). After day 4, wells with initial producer strain concentrations of 100 CFUs started to rebound. After day 7, wells with initial producer strain concentrations of 1000 CFUs started to rebound. The quantitative effect of this rebounding effect was increased by some wells reaching luminescence intensities, which were 2-3 times higher than the average blank wells. These high luminescence wells meant that the three layers containing the lowest level of producer strains exhibited average luminescence values that were similar or higher than the blank layer. In total, standardised luminescence in the layered format exhibited a continued decline over the 14- day period, exhibiting only 23 % standardised luminescence at day 14 (see Fig 7.4 D blue line). Conversely, the mixed wells exhibited the lowest standardised luminescence of 60 % luminescence after day 7, before standardised luminescence rebounded to 83 % at day 14 (see Fig 7.4 D red line). The layered setup therefore again looked superior to the mixed setup.

116

Fig 7.5. Standardised luminescence over time for layered and mixed setups. A Average standardised luminescence of initial producer strain concentrations over time in layered (left) and mixed (right) setups. B Average standardised luminescence over time in layered and mixed setups.

It is worth noting that wells that rebounded after having switched off were restreaked after experiments and were found to be tolerant to tetracycline. This increases the evidence that changes in standardised luminescence were caused at least partially by antibiotic production.

7.3. Discussion The developed protocol to decrease the size of mycelia chains for Streptomyces made it possible to explore antibiotic production in a variety of cocultivation conditions. This was done by inoculating the processed Streptomyces strains with reporter strains developed in Chapter 6 in chambers in standard 96-well plates. It was found that model antibiotic producer strains could kill/inhibit model reporter strains in a variety of medias. Using S. aureofaciens as the producer strain and Mini Bac Pveg Lux as the reporter strain, it was possible to explore the effect of the initial concentration of producer strains, the initial concentration of the reporter strains, and the physical relationship between reporter strains/producer strains using multivariable regression. The data suggested that the largest impact on expected standardised luminescence came from the

117 initial concentration of producer strains, with higher numbers likely resulting in higher production of antibiotics and therefore a higher concentration of antibiotics in each well. The physical separation of the producer strain and reporter strain also showed high statistical significance, which added evidence to the hypothesis of Wiener [151] that antibiotic producers have an advantage when established in an area. The multivariable regression was run using only one type of reporter strain, producer strain, and media. It would therefore add power to the statistical findings, if experiments with other conditions were explored in a similar format.

The above findings had numerous implications on the ISSA platform setup in Chapter 8. Firstly, the finding that the initial concentration of producer strains was important had design implication for the ISSA. While the ISSA platform does not allow for increases in initial concentrations of uncultivable bacteria, compartment sizes could be decreased, leading to higher antibiotic concentration given a fixed quantity of antibiotic production. The results also suggested that physical separation resulted in higher sensitivity, meaning that layering was determined as the setup for later ISSA platforms. Finally, the initial concentration of reporter strains seemed to have a negligible impact on sensitivity in the layered format, meaning that the initial concentration of the reporter strain did not have to be tightly regulated and the ISSA setup could instead focus on having high initial luminescent signal strength.

When looking at how these experiments complement research on the function of antibiotic production in situ, Thomashow et al. [182] showed that Pseudomonas fluorescens 2-79 and P. aureofaciens 30-84 produced the antibiotic phenazine-1-carboxylic acid in the rhizosphere in natural soil. Follow up experiments pointed to the antibiotic production being, at least in some cases, a means to protect wheat roots against specific pathogens by preventing the pathogens from accessing the rhizosphere [182, 183]. This supported the conclusion that antibiotics can be used as weapons in situ and most likely to defend colonized high value areas in direct response to invading species. In vitro experiments from Abrudan et al. [154] and Garbeva et al. [200] and Wiener [151] have already supported these finding. The experiments in section 6.2 adds further in vitro experimental evidence that suggests that this hypothesis could be correct, as antibiotic producers were able to kill/inhibit competing species across a variety of high nutrient conditions when established in a territory, thereby proving the ability of producer strains to use antibiotics as a means to assert dominance in a high nutrient territory.

118

Chapter 8. In vitro and in situ screening of antibiotic production in soil bacteria using the developed ISSA platform

8.1 Introduction The aim of this chapter was to run preliminary tests of the ability of the developed ISSA platform to detect and isolate soil bacteria that produce antibacterial compounds. This section therefore drew from the components developed in previous chapters. The quantification method developed in Chapter 4 had shown that it was possible to retrieve, isolate, and quantify soil bacteria (potential producer strains) from natural conditions. Chapter 5 had developed a physical ISSA design that ranged from a 384-comparment format with 0.6 mm ø to 5180-compartment format with an associated 0.3 mm ø. Chapter 5 also developed loading protocols that could be used for consistent loading of the producer strains and the reporter strains. The reporter strain P1H2 developed in Chapter 6 had been shown to emit luminescence consistently and over long periods even in low nutrient conditions. The components described above facilitated the construction of a prototypical ISSA platform that could be tested for its ability to detect antibacterial compounds. Chapter 7 optimised the relationship of these components in relation to antibiotic sensitivity and selectivity. The results of Chapter 7 led to the hypothesis that minimising compartment volume increased the sensitivity to antibiotic production and that physical separation resulted in higher selectivity and sensitivity to antibiotic production. The 5180-compartment format was therefore used as the array format as the compartment in this format had the smallest ø. Furthermore, the sequential loading protocol was used, as it ensured physical separation of the reporter and producer strains. Once the ISSA platform had been loaded, it could either be used to screen for antibiotics in vitro (in laboratory conditions) or in situ (in the natural environment of the isolated soil bacteria). Using the ISSA platform in vitro meant that nutrient conditions could be controlled, and that luminescence could be screened daily. In vitro ISSA platform screening, however, also meant that the soil bacteria were not screened in their native environment. A boost in cultivation could therefore not be assumed. During in vitro experiments, the ISSA platform was therefore assumed to screen cultivable bacteria for antibacterial production in a high throughput format. Conversely, during in situ cultivation the ISSA platform screened bacteria in their native environment. Nutrient control and luminescent detections were therefore limited to the start and end of the experiment in this setting. The in situ location of the ISSA platform, however, meant that the cultivation increase from 1.0 % to 15-20 % reported in Chapter 5 could be expected and furthermore that environmental triggers could lead to the identification of antibacterial production by strains that

119 have not been previously associated with antibacterial production. In situ ISSA screening therefore provided the highest chance of finding novel antibacterial producing strains, and therefore in the long run the highest chance of finding novel antibiotics.

8.2 Results

8.2.1 In vitro screening To test the sensitivity of the ISSA platform, experiments were carried out in laboratory conditions (in vitro), so that luminescence could be measured daily. To discover potential producer strains, soil was retrieved from various location including (Hyde Park (UK), Silwood Park (UK) and Contencie (France)). 5180-compartment arrays with 0.3 mm diameter were loaded with 3.0 % ULGTA 1/10 SMS solution, which was mixed with soil bacterial populations that resulted in an average of either 0.0 or 1.0 soil bacteria per compartment. The reporter strain P1H2 was subsequently loaded as a separate layer on top, using the developed sequential loading protocol. The platforms were incubated overnight at 28°C and scanned after 10 h for the first data point (day 1) and daily afterwards until day 7. In between scans, the ISSA platform was incubated in a humidity chamber and wrapped in purpose cut damp paper covering both sides to prevent compartments drying out over time. During measurements approximately 3 mm layer of 1/10 strength SMS solution was placed on top of the plates to firstly protect against compartments drying out and secondly to resupply wells with an influx of nutrients. To test the readout sensitivity before and after the nutrient addition, one half of the array were measured 5 min after the nutrient addition (labelled “t(5min)”). The second half of the array was measured 20 min after nutrient addition (labelled “t(20 min)”). The 15 min lag between t(5 min) and t(20 min) allowed time for the reporter strain populations at t(20 min) to increase their metabolic activity in response to the increased nutrient content. Looking at the readout at day 1 and day 7 for t(5 min) and t(20 min) there was a clear increase in luminescence when data was collected at t(20 min) both at day 1 and day 7 (see Fig 8.1 A). The effect could be assessed by quantifying the compartment luminescence of the t(5 min) readouts and t(20 min). At day 7, t(5 min) yielded an average of 17,628 AU, which was lower than t(20 min), which averaged 30,872 AU. Histograms of t(5 min) and t(20 min) showed that especially t(20 min) seemed to deviate visually from the classic bell curve associated with a normal distribution (see Fig 8.1 B). The skew and kurtosis of t(5 min) and t(20 min) were, however, found to be within the boundary of ±2.0, established in the literature

120

[201-203]. For t(5min) skew = 0.04 and kurtosis = -0.97. For t(20 min) skew = -0.56 and kurtosis = - 0.44. The distributions were therefore assumed to be normally distributed. Looking at the variance of the distributions the standard deviations of t(5 min) and t(20 min) were 7,181 AU and 8,375 AU, respectively. The combination of a relatively low mean and still high standard deviation meant that at t(5 min), 38/2,080 observations were observed to be between 0-5,000 AU and 141/2,080 observations between 5,000-7,500 RLU. At t(20 min) the higher mean meant that 0/2080 exhibited luminescent values between 0-5000 AU and 1/2080 observation exhibited values between 5000- 7500 RLU. This meant that at t(20 min) there was a dynamic range within which antibacterial compound producers could be detected without being mistaken for false positives.

Fig 8.1. Blank in vitro luminescent array results. A Luminescence at day 1 and day 7 of t(5 min) array and t(20 min) array. B Histogram of t(5 min) array and t(20 min) min array at day 7. The histograms show a clear shift towards higher luminescence at t(20 min) leaving room for tail end observations.

Using the same setup as above, a solution with an average of 1.0 soil bacterium per compartment (b/c) was loaded into the ISSA array. Half of the array was again measured at t(5 min) and the second half of the array at t(20 min). The increase in time after nutrient addition, again yielded an associated increase in luminescence. At day 7, t(5 min) observations had an average of 16,264 AU and a standard deviation of 5,217 AU. In comparison, t(20 min) observations had an average of 121

37,119 AU with a standard deviation of 7,510 AU (see Fig 8.2 A for luminescence images at t(5 min) and t(20 min)). It was visually clear that several compartments showed significantly lower luminescence than the surrounding compartments. When analysing the histogram, the distributions looked to have stabilised around a normal distribution, meaning that standard deviations could be applied (The skew and kurtosis were both outside the range of 2.0, but this was assumed to be caused by the high number of outliers). Running a one-sided t-test for significance, 12 compartments exhibited luminescence at a level that was significantly lower than the overall distribution at the P < 0.00001 significance level. At this significance level, 1:100,000 compartments could expect to yield a false positive. The compartments exhibiting low luminescence were often connected (see Fig 8.2 A), suggesting that the responsible producer strains either produced sufficient quantities of antibiotics to kill/inhibit reporter strains in surrounding compartments or were able to move between the array plate and the overlying membrane and colonize the adjacent compartments.

122

Fig 8.2. 1.0 bacterium in vitro luminescent array results. A Luminescence detection images at day 1 and day 7 of left array at t(5 min) right array at t(20 min). B Histogram of t(5 min) array and t(20 min) min array at day 7. The histograms show a clear shift towards higher luminescence at t(20 min) and tail end observations. C Visualisation of select tail end observations at day 1 and day 7.

Three hits identified in Fig 8.2 (Hit A, Hit D and Hit E) were compared with a negative control sample of 136 compartments over time to analyse when antibacterial compound secretion could be established (see Fig 8.3 A for luminescence of the selected compartments from day 1 - day 7). The data suggested that the onset of antibacterial compound secretion by antibiotic producers varied from compartment to compartment. Hit A initially exhibited lower luminesce than Hit D and Hit E and could be seen to be significant at the P < 0.0001 level at day 4 (see Fig 8.3 B). Hit E, while starting at a higher luminescence than Hit A, exhibited statistically significant decreases in luminescence at the P < 0.0001 level at day 4 (see Fig 8.3 B). Finally, the decrease in luminescence of Hit D only became statistically significant at day 6 (see Fig 8.3 B). Furthermore, after day 3 there was a decrease in the ratio of standard deviation/average luminescence (see Fig 8.3 C). The

123 decrease in this ratio meant that outliers became easier to detect. Consequently, at day 7, all three hits were significant at the P<0.00001 level. There therefore seemed to be value in running experiments over prolonged periods of time (minimum 4 days), as longer incubation times increased the level of hits and, furthermore, allowed a higher certainty when identifying these hits.

Fig 8.3. Analysis of select tail end observations at day 0 and day 7. A Visualisation of Hit A, Hit D and Hit E in comparison to surrounding wells at day 1 - day 7. B Luminescence of Hit A, Hit D, and Hit E compared to a sample of negative controls between day 1 – day 7. Error bars represent confidence intervals at the P < 0.0001 significance level. C Value of error at the P < 0.0001 significance level/average luminescence.

8.2.2 Harvesting of in vitro hits Hit A – Hit F from the in vitro ISSA experiment (see Fig 8.2 A) were harvested using the harvesting protocol developed in Chapter 5. Each harvested plug was then resuspended in 100 µl of R2A and plated out on R2A petri dishes, before being incubated for 48 h at room temperature. To eliminate any residual reporter strains, the petri dishes were analysed using luminescence detection (see Fig 8.4). Colonies showing no luminescence were further purified before being identified using 16S rRNA sequencing and then screened for antibiotic production.

124

Fig 8.4. Purification of producer strains. Plates are first photographed and screened for luminescence. By combining the two images, it is possible to deduct which colonies are producer strains by excluding colonies that emit luminescence.

Using the developed harvesting protocol, it was possible to identify colonies that grew isolated from the residual reporter strains. In total three rounds of in situ ISSA screening were carried out. After harvesting, the three rounds yielded a total of 12 identified hits (see Table 8.1 for phylogenetic analysis). The identified hits included the known producers Brevibacillus parabrevis (B. parabrevis), Janthinobacterium lividum (J. lividum), Bacillus cereus (B. cereus), and Stenotrophomonas maltophilia (S. maltophilia). B. parabrevis is the host organism associated with the production of the bactericidal antibiotic tyrothricin (composed of tyrocidine and gramicidin) [204]. J. lividum is the host organism associated with janthinocins A, B, and C, which are broad- spectrum antibacterial compounds with comparable efficacies to vancomycin [205]. B. cereus is associated with the production of the broad-spectrum antibacterial compound Zwittermicin A [206, 207]. S. maltophilia has been associated with the antifungal maltophilia [208], but has not been associated with the production of an antibacterial compound. The ISSA platform therefore consistently identified known antimicrobial producers, which added strong support to the claim that the ISSA platform selectively detects antibacterial compound producers. The experiments also

125 isolated the strains Massilia varians and Microbacterium oxidans, which had never previously been associated with antibiotic production.

Table 8.1. Harvested hits from in vitro ISSA platform screenings Round Strain Sequence Similarity in %

Round 1 3 x Stenotrophomonas maltophilia 99.23-100 Janthinobacterium lividum 100

Round 2 4 x Stenotrophomonas maltophilia 100

Massilia varians 99.54 Round 3 Microbacterium oxidans (Hit E) 100 Brevibacillus parabrevis (Hit A) 100 Bacillus cereus (Hit C) 100

8.2.3 Confirming antibacterial compound production by isolated hits To confirm that the strains in Table 8.1 produced antibiotics, the strains were plated out on four growth medias (R2A, SMS, R4 and 2xTY) on large petri dishes (150mm) for 5 – 14 days. 30 mm diameter circles were periodically cut out from the plates and placed on a new plate. The cut outs were then placed in competition with P1H2 or Mini Bac Pveg Lux by pouring a layer of LB agar with the designated reporter strain in media around the cut out from the producer strain before incubating the new cocultivation assay overnight. M. varians and S. maltophilia showed clear evidence of antibiotic production against Mini Bac Pveg Lux (see Fig 8.5), but not P1H2.

Fig 8.5. Clearance zones by M. varians and S. maltophilia. Overlay of 5 day colonies of the respective producer strains with the reporter strain Mini Bac Pveg Lux created visually detectable clearance zones for S. maltophilia and M. varians, thereby confirming that the two strains produce antibacterial compounds. The larger clearance zone associated with M. varians suggests a higher quantity or potency of the antibacterial compound against b. subtilis.

126

The examination of B. parabrevis showed clear evidence of antibiotic production against both P1H2 and Mini Bac Pveg Lux (see Fig 8.6). The experimental results suggested that prolonged periods of incubation at 5°C degrees increased the visual clearance zones. As stated earlier, P1H2 had been shown to be stable over longer periods of time, even at low temperatures. it was therefore possible to compare luminescent output with visual clearance zones. The results of P. parabrevis (see Fig 8.6) provided preliminary evidence that the two methods complemented each other, as the clearance zones shown in standard images largely corresponded to the clearance zones exhibited in the luminescence detection images. Finally, the secondary screening of P. parabrevis provided an example of upregulation of metabolic activity at the edge between areas with < MIC and > MIC of the secreted antibacterial compounds (see Fig 8.6 lower panel picture two from the left), suggesting that a metabolic upregulation happened when reporter strains were under stress.

Fig 8.6. Antibiotic clearance of P1H2 and Mini Bac Pveg Lux caused by P. parabrevis. The clearance zones were detected via normal photography (top) and via luminescence detection imaging (bottom).

8.2.4 In situ screening Given the ability of ISSA platform to identify antibacterial producers in vitro, in situ experiments were deemed feasible. In situ incubation meant that, if functional, increased domestication and domestic triggers should facilitate the detection and domestication of novel antibacterial compound producers. Two ISSA arrays were loaded with 0 (blank) or 1.0 soil bacteria per

127 compartment respectively. After overnight incubation the platforms were placed in the native soil of the isolated soil bacteria for 20 days and analysed after retrieval (see Fig 8.7 A bottom row). The bottom left corner of the 1.0 bacterium per compartment array seemed to have an edge effect on the bottom two quadrants in the first column from the left (see Fig 8.7 A red box). To eliminate the possibility of false positives these two quadrants were therefore also excluded, leaving 4148 compartments for further analysis. All compartments showed some luminescence, but with optically visible variation. Some compartments also showed visibly lower luminescence than surrounding compartments (see Fig 8.7 A green box). To see whether these compartments exhibited statistically significant decreases in luminescence, each compartment was integrated, and average values calculated. When analysing the corresponding histograms of the images taken at day 20 (see Fig 8.7 B) both the blank array and the 1.0 bacterium per compartment array seemed normally distributed and fell within the ±2 range in both skew and kurtosis. The blank array distribution exhibited skew = 0.448 and kurtosis = 0.170. The 1.0 bacterium per compartment array distribution exhibited skew = -0.339 and kurtosis = -0.036. The blank array exhibited a mean of 24,956 AU and a standard deviation of 5,448 AU. The 1.0 bacterium per compartment array exhibited a mean luminescence of 28,408 AU and a standard deviation of 6,288 AU. No compartment in either array fell below the P < 0.0001 significance level and it was therefore not possible to say that any compartment showed significance.

128

Fig 8.7. In vivo luminescent array results for blank array and 1.0 bacterium per compartment array. A Luminescence detection images at day 1 and day 20 of blank and 1.0 bacterium per compartment. B Histograms of blank array and 1 bacterium per compartment array.

8.2.2.1 Harvesting wells at the P < 0.001 significance level Four strains that showed significance at the P < 0.001 significance level were, however, harvested. Of these only one strain could be purified. This strain was identified as Bacillus mycoides (B. mycoides), which had never been associated with antibiotic production prior to these experiments. To test the ability of B. mycoides to produce antibacterial compounds, the strain was grown in three growth medias (LB, 2xTY and MYM) for 5-14 days. Periodically cut outs would be placed on new petri dishes and a layer of LB-agar with either Mini Bac Pveg Lux and P1H2 was then poured on top. The cultures were then grown overnight. The cocultivation assays showed antibiotic production against B. subtilis in both LB and 2xTY media, confirming the antibiotic capabilities of B. mycoides (see Fig 8.8 left and centre images). B. mycoides produced extreme hyphae when grown in solid media (see Fig 8.8 right image). Hyphae is characteristic of many antibiotic producers and might therefore further strengthen the argument that B. subtilis is an antibiotic producer.

129

Fig 8.8. Clearance zones of Mini Bac Pveg Lux caused by B. mycoides.

8.3 Discussion The experimental evidence showed that the ISSA platform was consistently able to detect antibacterial producing strains. In four rounds of ISSA platform experiments (3 x in vitro and 1 x in situ) the ISSA platform was able to detect six strains. Of these six identified hits, four strains have subsequently been confirmed as antibacterial compound producers in standard laboratory experiments (S. maltophilia, M. varians, B. parabrevis, and B. mycoides). Of the two strains that have not yet been confirmed as antibacterial producers, A. lividans is a known antibiotic producer and M. oxidans is an actinomycete (the most prominent phyla for antibiotic production). There is therefore a high probability that both strains will be able to produce antibacterial compounds given the right stimulus. It is also worth noting that S. maltophilia has so far been discovered numerous times in two different assays deriving from distinct physical locations, which again suggests that ISSA platform hits are reproducible and non-random. In terms of platform optimisation, there seems to be evidence to suggest that reactivation through nutrient addition (and potentially temperature increase when run in situ) leads to higher sensitivity to antibacterial compounds. Furthermore, it seems that the reporter strain populations stabilize over time in compartments where no antibacterial compounds are produced. Additionally, strong producer strains seem to increase their dominance over their compartment over time leading to lower luminescence. The ISSA platform therefore seems to be increasingly sensitive over time. The evidence therefore suggests that longer incubation periods increase the ability of the ISSA platform to detect antibacterial production, at least in vitro. The optimisation observations are, however, provisional and more rounds of ISSA platform screenings will have to be carried out before the platform is fully optimised. One final observation is that the Mini Bac Pveg Lux (B. subtilis) had

130 been observed to be associated with a lower MIC and be resistant to fewer antibiotics than P1H2 (P. putida). Furthermore, it is well documented that antibacterial compounds exhibit a broad spectrum of action and that antimicrobial producers often produce more than one antibacterial compound [209]. Nevertheless, it was surprising to see that of hits identified when screening against P1H2, more were able to inhibit Mini Bac Pveg lux than P1H2 in follow up experiments. More hits are necessary before this finding can be confirmed, but for now it adds evidence to the notion that finding effective solutions is harder when combatting Gram-negative pathogens than when combatting Gram-positive pathogens.

131

Chapter 9. Final summary & future perspectives

The spread of multi-resistant bacteria has meant that bacterial pathogens are projected to overtake cancer as the biggest threat to human health by 2050 [24]. Despite the looming health crisis, the established pharmaceutical industry has to date failed to find novel solutions, and no new antibiotic class has been discovered for 40 years [37]. New innovations are therefore needed.

When looking at classical antibiotic discovery, most antibiotics were found by screening the 1.0 % of bacteria that are readily cultivable [31, 33]. These bacteria were often screened in pure cultures and in high nutrient conditions [31, 32]. Antibiotic production is, however, tightly regulated and it has been shown that secondary metabolite secretion is boosted in cocultivation and when returning strains to natural conditions [43]. This project set out to build a platform that could screen the total terrestrial bacterial kingdom biome in situ. The ISSA platform introduced in this thesis merged the concept of the Waksman platform of cocultivation [30] with novel in situ bacterial cultivation platforms [108] to produce a hybrid high-throughput in situ screening (ISSA) platform.

To build such a platform, it was essential to detect and quantify soil bacteria so that they could be dispensed in known quantities into ISSA platform compartments. The CFDA-SE/EDTA protocol developed here was shown to selectively stain > 90 % of live bacteria representing phyla that make up > 85 % of bacteria living in the environment. The CFDA-SE/EDTA protocol therefore provided a staining technique that could be used for the ISSA platform. Proof of principle studies using the developed reporter strain Mini Bac Pveg Lux and the producer strain S. aureofaciens showed that small compartment size and physical separation of the reporter strain and producer strain both increased the sensitivity of the ISSA platform. To increase the chances of detecting antibacterial compound producers, the ISSA platform was therefore designed to minimise compartment volume (final volume was 31 nl) and increase compartment numbers per array (final numbers were 4080 compartment per array when excluding the outer rows/columns). This outcome was facilitated by using biologically inert titanium, and by introducing a vacuum-assisted loading protocol that pushed out air bubbles thereby insuring uniform loading. Finally, by optimising the pUT mini-Tn5 luxCDABE protocol, it was possible to develop reporter strains (including P1H2) that emitted strong and stable luminescence over long period of time, even in low nutrient conditions.

132

The development of the above components allowed for the ISSA platform to be operational both in vitro and in situ. More experiments must be carried out using in situ incubation, but provisional results are promising. Firstly, when analysing the experimental data of the ISSA platform, the integrated average values of luminescence seem to be normally distributed, meaning that results can be analysed using standard statistics. Secondly, 11 of the 12 hits, that have been isolated and identified, have been verified as antimicrobial producers either in downstream experiments or through a literature search. This suggests that the platform has high specificity in identifying antibacterial compound producers.

11 of 12 identified hits have come from in vitro screening. This means that the ISSA platform has been confirmed as a high throughput in vitro screening platform. Carrying out in situ screening, the ISSA platform has identified and verified B. mycoides as a producer of antibacterial compounds. Multiple ISSA platform screening rounds are currently taking place and will hopefully confirm the ability of the ISSA-platform to detect more antibacterial compounds producing bacteria in situ, thereby confirming the ISSA platform as the first in situ high-throughput antibiotic discovery platform.

If the ability of the ISSA platform to detect antibacterial compound producers is confirmed, several steps could be taken to optimise and build on the platform. While the CFDA-SE/EDTA protocol efficiently identifies and quantifies soil bacteria, it is at present limited by several bacterial species adhering to large particles. These particles at present therefore must be excluded, leading to loss of biodiversity. Developing, a novel shaking technique that allows for a higher level of breakup of soil particles without killing the adhered bacteria would therefore increase the biodiversity that can be screened using the ISSA platform. Furthermore, at present, the harvesting protocol is inefficient, as the protocol struggles to separate the producer strain from reporter strain when they are grown on agar plates after harvesting. Incorporating the newly developed synthetic kill switches into the reporters strains would be an elegant way to gain control of the harvesting protocol [210].

Furthermore, the flexibility of the ISSA platform allows for future development of novel reporter strains. At present the ISSA platform targets Pseudomonas pathogens, but it could screen for

133 antibiotic producers targeting different clinically relevant bacterial pathogens, if suitable bioreporters were developed. Finally, identified hits must be analysed and the responsible compound identified. This would most likely require an interdisciplinary approach, involving molecular biologists, bioinformaticians, and organic chemist. The project therefore requires a concerted effort of a multitude of partners to reach the stated aim of discovering novel antibiotics. Given the varied threat posed by antibiotic resistance [6, 26] and the lack of novel viable antibiotic compounds [31, 36], the overall promise of the results presented in this thesis suggests that such an effort would be justified. It is therefore my hope that the work presented in this thesis will generate the required enthusiasm to organise such collaborations.

134

Chapter 10. Bibliography

1. Bentley, R. and J. Bennett, What is an antibiotic? Revisited. Advances in applied microbiology, 2003. 52: p. 303-331. 2. Waksman, S.A., What is an antibiotic or an antibiotic substance? Mycologia, 1947. 39(5): p. 565-569. 3. Blair, J.M., et al., Molecular mechanisms of antibiotic resistance. Nature Reviews Microbiology, 2015. 13(1): p. 42. 4. Davies, J. and D. Davies, Origins and evolution of antibiotic resistance. Microbiology and molecular biology reviews, 2010. 74(3): p. 417-433. 5. Hughes, D. and A. Karlén, Discovery and preclinical development of new antibiotics. Upsala journal of medical sciences, 2014. 119(2): p. 162-169. 6. Prevention, C.f.D.C.a., Control of infectious diseases. MMWR. Morbidity and mortality weekly report, 1999. 48(29): p. 621. 7. Grove, R.D. and A.M. Hetzel, Vital Statistics Rates in the United States, 1940-1960: By Robert D. Grove and Alice M. Hetzel. 1968: US Government Printing Office. 8. Lederberg, J., Infectious history. Science, 2000. 288(5464): p. 287-293. 9. Schofield, R., D. Reher, and A. Bideau, The decline of mortality in Europe. 1991. 10. Smith, R. and J. Coast, The true cost of antimicrobial resistance. Bmj, 2013. 346: p. f1493. 11. Ventola, C.L., The antibiotic resistance crisis: part 1: causes and threats. Pharmacy and Therapeutics, 2015. 40(4): p. 277. 12. Aloush, V., et al., Multidrug-resistant Pseudomonas aeruginosa: risk factors and clinical impact. Antimicrobial agents and chemotherapy, 2006. 50(1): p. 43-48. 13. Cars, O., et al., Meeting the challenge of antibiotic resistance. BMJ: British Medical Journal (Online), 2008. 337. 14. Pizzo, P.A., Fever in immunocompromised patients. New England Journal of Medicine, 1999. 341(12): p. 893-900. 15. Anaissie, E., Opportunistic mycoses in the immunocompromised host: experience at a cancer center and review. Clinical Infectious Diseases, 1992. 14(Supplement_1): p. S43- S53. 16. Berendonk, T.U., et al., Tackling antibiotic resistance: the environmental framework. Nature Reviews Microbiology, 2015. 13(5): p. 310. 17. Baroud, á., et al., Underlying mechanisms of carbapenem resistance in extended-spectrum β-lactamase-producing Klebsiella pneumoniae and Escherichia coli isolates at a tertiary care centre in Lebanon: role of OXA-48 and NDM-1 carbapenemases. International journal of antimicrobial agents, 2013. 41(1): p. 75-79. 18. Warner, D.M., W.M. Shafer, and A.E. Jerse, Clinically relevant mutations that cause derepression of the Neisseria gonorrhoeae MtrC‐MtrD‐MtrE Efflux pump system confer different levels of antimicrobial resistance and in vivo fitness. Molecular microbiology, 2008. 70(2): p. 462-478. 19. Dhanji, H., et al., Dissemination of pCT-like IncK plasmids harboring CTX-M-14 extended- spectrum β-lactamase among clinical Escherichia coli isolates in the United Kingdom. Antimicrobial agents and chemotherapy, 2012. 56(6): p. 3376-3377. 20. Dahmen, S., et al., Characterization of extended-spectrum beta-lactamase (ESBL)-carrying plasmids and clones of Enterobacteriaceae causing cattle mastitis in France. Veterinary microbiology, 2013. 162(2-4): p. 793-799.

135

21. Cottell, J.L., M.A. Webber, and L.J. Piddock, Persistence of transferable extended- spectrum-β-lactamase resistance in the absence of antibiotic pressure. Antimicrobial agents and chemotherapy, 2012. 56(9): p. 4703-4706. 22. Group, E.E.J.W., The bacterial challenge: time to react. A call to narrow the gap between multidrug-resistant bacteria in the EU and development of new antibacterial agents. Stockholm: European Centre for Disease Prevention and Control, 2009. 23. Prevention, C.f.D.C.a., Antibiotic resistance threats in the United States, 2013. 2013: Centres for Disease Control and Prevention, US Department of Health and Human Services. 24. Resistance, R.o.A., Tackling drug-resistant infections globally: final report and recommendations. 2016: Review on Antimicrobial Resistance. 25. Willyard, C., The drug-resistant bacteria that pose the greatest health threats. Nature News, 2017. 543(7643): p. 15. 26. Organization, W.H., Global priority list of antibiotic-resistant bacteria to guide research, discovery, and development of new antibiotics. Geneva: World Health Organization, 2017. 27. Liu, Y.-Y., et al., Emergence of plasmid-mediated colistin resistance mechanism MCR-1 in animals and human beings in China: a microbiological and molecular biological study. The Lancet infectious diseases, 2016. 16(2): p. 161-168. 28. Williams, K., The introduction of ‘chemotherapy’using arsphenamine–the first magic bullet. Journal of the Royal Society of Medicine, 2009. 102(8): p. 343-348. 29. Bennett, J.W. and K.-T. Chung, Alexander Fleming and the discovery of penicillin. 2001. 30. Waksman, S.A. and H.B. Woodruff, The soil as a source of microorganisms antagonistic to disease-producing bacteria. Journal of Bacteriology, 1940. 40(4): p. 581. 31. Payne, D.J., et al., Drugs for bad bugs: confronting the challenges of antibacterial discovery. Nature reviews Drug discovery, 2007. 6(1): p. 29. 32. Katz, L. and R.H. Baltz, Natural product discovery: past, present, and future. Journal of industrial microbiology & biotechnology, 2016. 43(2-3): p. 155-176. 33. Lewis, K., Antibiotics: Recover the lost art of drug discovery. Nature, 2012. 485(7399): p. 439-440. 34. Kahlmeter, G., Defining antibiotic resistance-towards international harmonization. Upsala journal of medical sciences, 2014. 119(2): p. 78-86. 35. Butler, M.S. and A.D. Buss, Natural products—the future scaffolds for novel antibiotics? Biochemical pharmacology, 2006. 71(7): p. 919-929. 36. Lewis, K., Platforms for antibiotic discovery. Nature reviews Drug discovery, 2013. 12(5): p. 371-387. 37. Silver, L.L., Challenges of antibacterial discovery. Clinical microbiology reviews, 2011. 24(1): p. 71-109. 38. Kohanski, M.A., D.J. Dwyer, and J.J. Collins, How antibiotics kill bacteria: from targets to networks. Nature Reviews Microbiology, 2010. 8(6): p. 423. 39. Chan, P.F., et al., Novel antibacterials: a genomics approach to drug discovery. Current Drug Targets-Infectious Disorders, 2002. 2(4): p. 291-308. 40. Lipinski, C.A., et al., Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Advanced drug delivery reviews, 1997. 23(1-3): p. 3-25. 41. Harvey, A.L., Natural products in drug discovery. Drug discovery today, 2008. 13(19-20): p. 894-901.

136

42. Pettit, R.K., Mixed fermentation for natural product drug discovery. Applied microbiology and biotechnology, 2009. 83(1): p. 19-25. 43. Seyedsayamdost, M.R., High-throughput platform for the discovery of elicitors of silent bacterial gene clusters. Proceedings of the National Academy of Sciences, 2014: p. 201400019. 44. Council, N.R., Treating infectious diseases in a microbial world: report of two workshops on novel antimicrobial therapeutics. 2006: National Academies Press. 45. Ochi, K. and T. Hosaka, New strategies for drug discovery: activation of silent or weakly expressed microbial gene clusters. Applied microbiology and biotechnology, 2013. 97(1): p. 87-98. 46. Wu, C., Y.H. Choi, and G.P. van Wezel, Metabolic profiling as a tool for prioritizing antimicrobial compounds. Journal of industrial microbiology & biotechnology, 2016. 43(2- 3): p. 299-312. 47. Baranasic, D., et al., Draft genome sequence of Streptomyces rapamycinicus strain NRRL 5491, the producer of the immunosuppressant rapamycin. Genome announcements, 2013. 1(4): p. e00581-13. 48. Bister, B., et al., Abyssomicin C—A Polycyclic Antibiotic from a Marine Verrucosispora Strain as an Inhibitor of the p‐Aminobenzoic Acid/Tetrahydrofolate Biosynthesis Pathway. Angewandte Chemie International Edition, 2004. 43(19): p. 2574-2576. 49. Amann, J., Die direkte Zählung der Wasserbakterien mittels des Ultramikroskops. Centralbl. Bakteriol, 1911. 29: p. 381-384. 50. Jacob, W.F., M. Santer, and A.E. Dahlberg, A single base change in the Shine-Dalgarno region of 16S rRNA of Escherichia coli affects translation of many proteins. Proceedings of the National Academy of Sciences, 1987. 84(14): p. 4757-4761. 51. Klappenbach, J.A., J.M. Dunbar, and T.M. Schmidt, rRNA operon copy number reflects ecological strategies of bacteria. Applied and environmental microbiology, 2000. 66(4): p. 1328-1333. 52. Yang, B., Y. Wang, and P.-Y. Qian, Sensitivity and correlation of hypervariable regions in 16S rRNA genes in phylogenetic analysis. BMC bioinformatics, 2016. 17(1): p. 135. 53. Lane, D.J., et al., Rapid determination of 16S ribosomal RNA sequences for phylogenetic analyses. Proceedings of the National Academy of Sciences, 1985. 82(20): p. 6955-6959. 54. Woese, C.R., Bacterial evolution. Microbiological reviews, 1987. 51(2): p. 221. 55. Yarza, P., et al., Uniting the classification of cultured and uncultured bacteria and archaea using 16S rRNA gene sequences. Nat Rev Micro, 2014. 12(9): p. 635-645. 56. Stackebrandt, E., Taxonomic parameters revisited: tarnished gold standards. Microbiol. Today, 2006. 6: p. 152-155. 57. Lagos, L., et al., Current overview on the study of bacteria in the rhizosphere by modern molecular techniques: a mini‒review. Journal of soil science and plant nutrition, 2015. 15(2): p. 504-523. 58. Smets, W., et al., A method for simultaneous measurement of soil bacterial abundances and community composition via 16S rRNA gene sequencing. Soil Biology and Biochemistry, 2016. 96: p. 145-151. 59. Torsvik, V., L. Øvreås, and T.F. Thingstad, Prokaryotic diversity--magnitude, dynamics, and controlling factors. Science, 2002. 296(5570): p. 1064-1066. 60. Curtis, T.P. and W.T. Sloan, Exploring microbial diversity--A vast below. Science, 2005. 309(5739): p. 1331-1333.

137

61. Delgado-Baquerizo, M., et al., A global atlas of the dominant bacteria found in soil. Science, 2018. 359(6373): p. 320-325. 62. Nemergut, D.R., et al., Global patterns in the biogeography of bacterial taxa. Environmental microbiology, 2011. 13(1): p. 135-144. 63. Hug, L.A., et al., A new view of the tree of life. Nature microbiology, 2016. 1: p. 16048. 64. Schulz, F., et al., Towards a balanced view of the bacterial tree of life. Microbiome, 2017. 5(1): p. 140. 65. Lauber, C.L., et al., Pyrosequencing-based assessment of soil pH as a predictor of soil bacterial community structure at the continental scale. Applied and environmental microbiology, 2009. 75(15): p. 5111-5120. 66. Rousk, J., et al., Soil bacterial and fungal communities across a pH gradient in an arable soil. The ISME journal, 2010. 4(10): p. 1340. 67. O'brien, S.L., et al., Spatial scale drives patterns in soil bacterial diversity. Environmental microbiology, 2016. 18(6): p. 2039-2051. 68. Fierer, N. and R.B. Jackson, The diversity and biogeography of soil bacterial communities. Proceedings of the National Academy of Sciences of the United States of America, 2006. 103(3): p. 626-631. 69. Fierer, N., et al., Cross-biome metagenomic analyses of soil microbial communities and their functional attributes. Proceedings of the National Academy of Sciences, 2012. 109(52): p. 21390-21395. 70. Noronha, M.F., et al., Taxonomic and functional patterns across soil microbial communities of global biomes. Science of The Total Environment, 2017. 609: p. 1064- 1074. 71. Teeling, H. and F.O. Glöckner, Current opportunities and challenges in microbial metagenome analysis—a bioinformatic perspective. Briefings in bioinformatics, 2012: p. bbs039. 72. Overmann, J., B. Abt, and J. Sikorski, Present and future of culturing bacteria. Annual review of microbiology, 2017. 71: p. 711-730. 73. Olson, N.D., et al., Metagenomic assembly through the lens of validation: recent advances in assessing and improving the quality of genomes assembled from metagenomes. Briefings in bioinformatics, 2017. 74. Nagarajan, N. and M. Pop, Parametric complexity of sequence assembly: theory and applications to next generation sequencing. Journal of computational biology, 2009. 16(7): p. 897-908. 75. Treangen, T.J., et al., Genesis, effects and fates of repeats in prokaryotic genomes. FEMS microbiology reviews, 2009. 33(3): p. 539-571. 76. Delmont, T.O., et al., Reconstructing rare soil microbial genomes using in situ enrichments and metagenomics. Frontiers in microbiology, 2015. 6: p. 358. 77. Howe, A.C., et al., Tackling soil diversity with the assembly of large, complex metagenomes. Proceedings of the National Academy of Sciences, 2014: p. 201402564. 78. Albertsen, M., et al., Genome sequences of rare, uncultured bacteria obtained by differential coverage binning of multiple metagenomes. Nature biotechnology, 2013. 31(6): p. 533. 79. Rodrigue, S., et al., Whole genome amplification and de novo assembly of single bacterial cells. PloS one, 2009. 4(9): p. e6864. 80. Lasken, R.S., Genomic sequencing of uncultured microorganisms from single cells. Nature Reviews Microbiology, 2012. 10(9): p. 631.

138

81. Spang, A., et al., Complex archaea that bridge the gap between prokaryotes and eukaryotes. Nature, 2015. 521(7551): p. 173-179. 82. Hover, B.M., et al., Culture-independent discovery of the malacidins as calcium-dependent antibiotics with activity against multidrug-resistant Gram-positive pathogens. Nature microbiology, 2018. 3(4): p. 415. 83. Gabor, E.M., W.B. Alkema, and D.B. Janssen, Quantifying the accessibility of the metagenome by random expression cloning techniques. Environmental Microbiology, 2004. 6(9): p. 879-886. 84. Vester, J.K., M.A. Glaring, and P. Stougaard, Improved cultivation and metagenomics as new tools for bioprospecting in cold environments. Extremophiles, 2015. 19(1): p. 17-29. 85. Leis, B., A. Angelov, and W. Liebl, Screening and expression of genes from metagenomes, in Advances in applied microbiology. 2013, Elsevier. p. 1-68. 86. Liebl, W., et al., Alternative hosts for functional (meta) genome analysis. Applied microbiology and biotechnology, 2014. 98(19): p. 8099-8109. 87. Vartoukian, S., R. Palmer, and W. Wade, Cultivation of a Synergistetes strain representing a previously uncultivated lineage. Environmental microbiology, 2010. 12(4): p. 916-928. 88. Ferrari, B.C., S.J. Binnerup, and M. Gillings, Microcolony cultivation on a soil substrate membrane system selects for previously uncultured soil bacteria. Applied and Environmental Microbiology, 2005. 71(12): p. 8714-8720. 89. Fierer, N., M.A. Bradford, and R.B. Jackson, Toward an ecological classification of soil bacteria. Ecology, 2007. 88(6): p. 1354-1364. 90. Koch, A.L., Microbial physiology and ecology of slow growth. Microbiology and Molecular Biology Reviews, 1997. 61(3): p. 305-318. 91. Ruoff, K.L., Nutritionally variant streptococci. Clinical microbiology reviews, 1991. 4(2): p. 184-190. 92. Winding, A., S.J. Binnerup, and J. Sørensen, Viability of indigenous soil bacteria assayed by respiratory activity and growth. Applied and Environmental Microbiology, 1994. 60(8): p. 2869-2875. 93. Kaeberlein, T., K. Lewis, and S.S. Epstein, Isolating" uncultivable" microorganisms in pure culture in a simulated natural environment. Science, 2002. 296(5570): p. 1127-1129. 94. Watve, M., et al., The'K'selected oligophilic bacteria: A key to uncultured diversity? Current science, 2000: p. 1535-1542. 95. Rappé, M.S., et al., Cultivation of the ubiquitous SAR11 marine bacterioplankton clade. Nature, 2002. 418(6898): p. 630-633. 96. Jousset, A., Ecological and evolutive implications of bacterial defences against predators. Environmental microbiology, 2012. 14(8): p. 1830-1843. 97. Hahn, M.W. and M.G. Höfle, Flagellate predation on a bacterial model community: interplay of size-selective grazing, specific bacterial cell size, and bacterial community composition. Applied and Environmental Microbiology, 1999. 65(11): p. 4863-4872. 98. Philippot, L., et al., The ecological coherence of high bacterial taxonomic ranks. Nature Reviews Microbiology, 2010. 8(7): p. 523-529. 99. Nichols, D., et al., Short peptide induces an “uncultivable” microorganism to grow in vitro. Applied and environmental microbiology, 2008. 74(15): p. 4889-4897. 100. Barcina, I., et al., Survival strategy of Escherichia coli and Enterococcus faecalis in illuminated fresh and marine systems. Journal of Applied Bacteriology, 1990. 68(2): p. 189-198.

139

101. Mukamolova, G.V., et al., A bacterial cytokine. Proceedings of the National Academy of Sciences, 1998. 95(15): p. 8916-8921. 102. Alain, K. and J. Querellou, Cultivating the uncultured: limits, advances and future challenges. Extremophiles, 2009. 13(4): p. 583-594. 103. Köpke, B., et al., Microbial diversity in coastal subsurface sediments: a cultivation approach using various electron acceptors and substrate gradients. Applied and environmental microbiology, 2005. 71(12): p. 7819-7830. 104. Epstein, S., The phenomenon of microbial uncultivability. Current opinion in microbiology, 2013. 16(5): p. 636-642. 105. Davis, K.E., S.J. Joseph, and P.H. Janssen, Effects of growth medium, inoculum size, and incubation time on culturability and isolation of soil bacteria. Applied and environmental microbiology, 2005. 71(2): p. 826-834. 106. Song, J., H.-M. Oh, and J.-C. Cho, Improved culturability of SAR11 strains in dilution-to- extinction culturing from the East Sea, West Pacific Ocean. FEMS microbiology letters, 2009. 295(2): p. 141-147. 107. Uphoff, H., et al., The microbial diversity in picoplankton enrichment cultures: a molecular screening of marine isolates. FEMS Microbiology Ecology, 2001. 35(3): p. 249-258. 108. Nichols, D., et al., Use of ichip for high-throughput in situ cultivation of “uncultivable” microbial species. Applied and environmental microbiology, 2010. 76(8): p. 2445-2450. 109. Zengler, K., et al., Cultivating the uncultured. Proceedings of the National Academy of Sciences, 2002. 99(24): p. 15681-15686. 110. Bruns, A., H. Cypionka, and J. Overmann, Cyclic AMP and acyl homoserine lactones increase the cultivation efficiency of heterotrophic bacteria from the central Baltic Sea. Applied and Environmental Microbiology, 2002. 68(8): p. 3978-3987. 111. Lynd, L.R., et al., Microbial cellulose utilization: fundamentals and biotechnology. Microbiology and molecular biology reviews, 2002. 66(3): p. 506-577. 112. D'Onofrio, A., et al., Siderophores from neighboring organisms promote the growth of uncultured bacteria. Chemistry & biology, 2010. 17(3): p. 254-264. 113. Stewart, E.J., Growing unculturable bacteria. Journal of bacteriology, 2012. 194(16): p. 4151-4160. 114. Ferrari, B.C. and M.R. Gillings, Cultivation of fastidious bacteria by viability staining and micromanipulation in a soil substrate membrane system. Applied and environmental microbiology, 2009. 75(10): p. 3352-3354. 115. Berdy, B., et al., In situ cultivation of previously uncultivable microorganisms using the ichip. Nature protocols, 2017. 12(10): p. 2232. 116. Aoi, Y., et al., Hollow-fiber membrane chamber as a device for in situ environmental cultivation. Applied and environmental microbiology, 2009. 75(11): p. 3826-3833. 117. Ben-Dov, E., E. Kramarsky-Winter, and A. Kushmaro, An in situ method for cultivating microorganisms using a double encapsulation technique. FEMS microbiology ecology, 2009. 68(3): p. 363-371. 118. Ling, L.L., et al., A new antibiotic kills pathogens without detectable resistance. Nature, 2015. 517(7535): p. 455-459. 119. Piddock, L.J., Teixobactin, the first of a new class of antibiotics discovered by iChip technology? Journal of Antimicrobial Chemotherapy, 2015. 70(10): p. 2679-2680. 120. Gavrish, E., et al., Lassomycin, a ribosomally synthesized cyclic peptide, kills Mycobacterium tuberculosis by targeting the ATP-dependent protease ClpC1P1P2. Chemistry & biology, 2014. 21(4): p. 509-518.

140

121. Peoples, A.J., et al., Neocitreamicins I and II, novel antibiotics with activity against methicillin-resistant Staphylococcus aureus and vancomycin-resistant Enterococci. The Journal of antibiotics, 2008. 61(7): p. 457. 122. Zong, Y., et al., Developing Equipotent Teixobactin Analogues against Drug-Resistant Bacteria and Discovering a Hydrophobic Interaction between Lipid II and Teixobactin. Journal of medicinal chemistry, 2018. 61(8): p. 3409-3421. 123. Mello, L.D. and L.T. Kubota, Review of the use of biosensors as analytical tools in the food and drink industries. Food chemistry, 2002. 77(2): p. 237-256. 124. Su, L., et al., Microbial biosensors: a review. Biosensors and Bioelectronics, 2011. 26(5): p. 1788-1799. 125. Weitz, H.J., et al., Construction of a modified mini-Tn5 luxCDABE transposon for the development of bacterial biosensors for ecotoxicity testing. FEMS microbiology letters, 2001. 197(2): p. 159-165. 126. Ritchie, J.M., et al., A stable bioluminescent construct of Escherichia coli O157: H7 for hazard assessments of long-term survival in the environment. Applied and environmental microbiology, 2003. 69(6): p. 3359-3367. 127. Rahmer, R., K. Morabbi Heravi, and J. Altenbuchner, Construction of a super-competent Bacillus subtilis 168 using the PmtlA-comKS inducible cassette. Frontiers in microbiology, 2015. 6: p. 1431. 128. Radeck, J., et al., The Bacillus BioBrick Box: generation and evaluation of essential genetic building blocks for standardized work with Bacillus subtilis. Journal of biological engineering, 2013. 7(1): p. 29. 129. Andreu, N., et al., Rapid in vivo assessment of drug efficacy against Mycobacterium tuberculosis using an improved firefly luciferase. Journal of Antimicrobial Chemotherapy, 2013. 68(9): p. 2118-2127. 130. Schindelin, J., et al., Fiji: an open-source platform for biological-image analysis. Nature methods, 2012. 9(7): p. 676. 131. Givan, A.L., Principles of flow cytometry: an overview. Methods in cell biology, 2001. 63: p. 19-50. 132. Harwood, C.R., Bacillus subtilis and its relatives: molecular biological and industrial workhorses. Trends in biotechnology, 1992. 10: p. 247-256. 133. Fuller, M.E., et al., Development of a vital fluorescent staining method for monitoring bacterial transport in subsurface environments. Applied and environmental microbiology, 2000. 66(10): p. 4486-4496. 134. Lindahl, V. and L.R. Bakken, Evaluation of methods for extraction of bacteria from soil. FEMS Microbiology Ecology, 1995. 16(2): p. 135-142. 135. Ramsay, A.J., Extraction of bacteria from soil: efficiency of shaking or ultrasonication as indicated by direct counts and autoradiography. Soil biology and biochemistry, 1984. 16(5): p. 475-481. 136. Herron, P.R. and E.M. Wellington, New method for extraction of streptomycete spores from soil and application to the study of lysogeny in sterile amended and nonsterile soil. Applied and Environmental Microbiology, 1990. 56(5): p. 1406-1412. 137. Macdonald, R., Sampling soil microfloras: dispersion of soil by ion exchange and extraction of specific microorganisms from suspension by elutriation. Soil Biology and Biochemistry, 1986. 18(4): p. 399-406.

141

138. Nunan, N., et al., Quantification of the in situ distribution of soil bacteria by large-scale imaging of thin sections of undisturbed soil. FEMS Microbiology Ecology, 2001. 37(1): p. 67-77. 139. Hobbie, J.E., R.J. Daley, and S. Jasper, Use of nuclepore filters for counting bacteria by fluorescence microscopy. Applied and environmental microbiology, 1977. 33(5): p. 1225- 1228. 140. Yu, W., et al., Optimal staining and sample storage time for direct microscopic enumeration of total and active bacteria in soil with two fluorescent dyes. Applied and Environmental Microbiology, 1995. 61(9): p. 3367-3372. 141. Williams, S., et al., Distinguishing between living and nonliving bacteria: evaluation of the vital stain propidium iodide and its combined use with molecular probes in aquatic samples. Journal of Microbiological Methods, 1998. 32(3): p. 225-236. 142. Mayr, C., A. Winding, and N. Hendriksen, Community level physiological profile of soil bacteria unaffected by extraction method. Journal of Microbiological Methods, 1999. 36(1): p. 29-33. 143. Créach, V., et al., Direct estimate of active bacteria: CTC use and limitations. Journal of microbiological methods, 2003. 52(1): p. 19-28. 144. Ullrich, S., et al., Toxic effects on bacterial metabolism of the redox dye 5-cyano-2, 3- ditolyl tetrazolium chloride. Applied and environmental microbiology, 1996. 62(12): p. 4587-4593. 145. Boulos, L., et al., LIVE/DEAD® BacLight™: application of a new rapid staining method for direct enumeration of viable and total bacteria in drinking water. Journal of microbiological Methods, 1999. 37(1): p. 77-86. 146. Bartosch, S., et al., CTC staining and counting of actively respiring bacteria in natural stone using confocal laser scanning microscopy. Journal of microbiological methods, 2003. 52(1): p. 75-84. 147. Fuller, M.E., et al., Field-scale evaluation of CFDA/SE staining coupled with multiple detection methods for assessing the transport of bacteria in situ. FEMS Microbiology Ecology, 2001. 37(1): p. 55-66. 148. Grégori, G., et al., Resolution of viable and membrane-compromised bacteria in freshwater and marine waters based on analytical flow cytometry and nucleic acid double staining. Applied and Environmental Microbiology, 2001. 67(10): p. 4662-4670. 149. Karimi, B., et al., Biogeography of soil bacteria and archaea across France. Science advances, 2018. 4(7): p. eaat1808. 150. Dunbar, J., et al., Empirical and theoretical bacterial diversity in four Arizona soils. Applied and environmental microbiology, 2002. 68(6): p. 3035-3045. 151. Wiener, P., Experimental studies on the ecological role of antibiotic production in bacteria. Evolutionary Ecology, 1996. 10(4): p. 405-421. 152. Cornforth, D.M. and K.R. Foster, Competition sensing: the social side of bacterial stress responses. Nature Reviews Microbiology, 2013. 11(4): p. 285-293. 153. Yim, G., H.H. Wang, and J.D. FRS, Antibiotics as signalling molecules. Philosophical Transactions of the Royal Society of London B: Biological Sciences, 2007. 362(1483): p. 1195-1200. 154. Abrudan, M.I., et al., Socially mediated induction and suppression of antibiosis during bacterial coexistence. Proceedings of the National Academy of Sciences, 2015. 112(35): p. 11054-11059.

142

155. Bermúdez, M.-D., et al., Erosion–corrosion of stainless steels, titanium, tantalum and zirconium. Wear, 2005. 258(1-4): p. 693-700. 156. Killham, K., Soil ecology. 1994: Cambridge University Press. 157. Oldani, C. and A. Dominguez, Titanium as a Biomaterial for Implants, in Recent advances in arthroplasty. 2012, InTech. 158. Penney, D., et al., Analysis and testing of biological stains--the Biological Stain Commission Procedures. Biotechnic & histochemistry, 2002. 77(5-6): p. 237-275. 159. Ivask, A., et al., Bioavailability of Cd in 110 polluted topsoils to recombinant bioluminescent sensor bacteria: effect of soil particulate matter. Journal of soils and sediments, 2011. 11(2): p. 231-237. 160. Vollmer, A.C., et al., Detection of DNA damage by use of Escherichia coli carrying recA':: lux, uvrA':: lux, or alkA':: lux reporter plasmids. Applied and environmental microbiology, 1997. 63(7): p. 2566-2571. 161. Rattray, E., et al., Luminescence-based nonextractive technique for in situ detection of Escherichia coli in soil. Applied and environmental microbiology, 1990. 56(11): p. 3368- 3374. 162. Applegate, B., et al., Pseudomonas putida B2: a tod-lux bioluminescent reporter for toluene and trichloroethylene co-metabolism. Journal of Industrial Microbiology and Biotechnology, 1997. 18(1): p. 4-9. 163. Abd-El-Haleem, D., et al., A luxCDABE-based bioluminescent bioreporter for the detection of phenol. Journal of Industrial Microbiology and Biotechnology, 2002. 29(5): p. 233-237. 164. Bondarenko, O., et al., Bioavailability of Cd, Zn and Hg in soil to nine recombinant luminescent metal sensor bacteria. Sensors, 2008. 8(11): p. 6899-6923. 165. Hynninen, A. and M. Virta, Whole-cell bioreporters for the detection of bioavailable metals, in Whole Cell Sensing System II. 2009, Springer. p. 31-63. 166. Xu, T., et al., Genetically modified whole-cell bioreporters for environmental assessment. Ecological indicators, 2013. 28: p. 125-141. 167. Gregor, C., et al., Strongly enhanced bacterial bioluminescence with the ilux operon for single-cell imaging. Proceedings of the National Academy of Sciences, 2018: p. 201715946. 168. Craney, A., et al., A synthetic luxCDABE gene cluster optimized for expression in high-GC bacteria. Nucleic acids research, 2007. 35(6): p. e46. 169. van Elsas, J.D., et al., Survival of Escherichia coli in the environment: fundamental and public health aspects. The ISME journal, 2011. 5(2): p. 173-183. 170. Compeau, G., et al., Survival of rifampin-resistant mutants of Pseudomonas fluorescens and Pseudomonas putida in soil systems. Applied and Environmental Microbiology, 1988. 54(10): p. 2432-2438. 171. Antunes, L., P. Visca, and K.J. Towner, Acinetobacter baumannii: evolution of a global pathogen. Pathogens and disease, 2014. 71(3): p. 292-301. 172. Falagas, M.E. and P.I. Rafailidis, Attributable mortality of Acinetobacter baumannii: no longer a controversial issue. Critical care, 2007. 11(3): p. 134. 173. Cerqueira, G.M. and A.Y. Peleg, Insights into Acinetobacter baumannii pathogenicity. IUBMB life, 2011. 63(12): p. 1055-1060. 174. Powell, T.R., et al., Protocol development to test the activity of the antimicrobial peptide human defensin 2 against Mycobacterium smegmatis. 2017. 175. Liu, H. and A.M. Deutschbauer, Rapidly moving new bacteria to model-organism status. Current opinion in biotechnology, 2018. 51: p. 116-122.

143

176. Gengenbacher, M. and S.H. Kaufmann, Mycobacterium tuberculosis: success through dormancy. FEMS microbiology reviews, 2012. 36(3): p. 514-532. 177. Manson, A.L., et al., Genomic analysis of globally diverse Mycobacterium tuberculosis strains provides insights into the emergence and spread of multidrug resistance. Nature genetics, 2017. 49(3): p. 395. 178. Anagnostopoulos, C. and J. Spizizen, Requirements for transformation in Bacillus subtilis. Journal of bacteriology, 1961. 81(5): p. 741. 179. Darken, M.A., et al., Production of tetracycline by Streptomyces aureofaciens in synthetic media. Applied microbiology, 1960. 8(1): p. 46-51. 180. Waksman, S.A. and H.A. Lechevalier, Neomycin, a new antibiotic active against streptomycin-resistant bacteria, including tuberculosis organisms. Science, 1949. 109(2830): p. 305-307. 181. Davies, J., Are antibiotics naturally antibiotics? Journal of Industrial Microbiology and Biotechnology, 2006. 33(7): p. 496-499. 182. Thomashow, L.S., et al., Production of the antibiotic phenazine-1-carboxylic acid by fluorescent Pseudomonas species in the rhizosphere of wheat. Applied and Environmental Microbiology, 1990. 56(4): p. 908-912. 183. Raaijmakers, J.M., D.M. Weller, and L.S. Thomashow, Frequency of antibiotic-producing Pseudomonas spp. in natural environments. Applied and Environmental Microbiology, 1997. 63(3): p. 881-887. 184. Garbeva, P. and W. De Boer, Inter-specific interactions between carbon-limited soil bacteria affect behavior and gene expression. Microbial ecology, 2009. 58(1): p. 36-46. 185. Van der Meij, A., et al., Chemical ecology of antibiotic production by actinomycetes. FEMS microbiology reviews, 2017. 41(3): p. 392-416. 186. Minieri, P.P., et al., A new broad spectrum antibiotic product of the tetracycline group. Antibiotics annual, 1953. 1954: p. 81-87. 187. Flärdh, K. and M.J. Buttner, Streptomyces morphogenetics: dissecting differentiation in a filamentous bacterium. Nature Reviews Microbiology, 2009. 7(1): p. 36. 188. Vining, L. and D. Westlake, Chloramphenicol: Properties, biosynthesis, and fermentation. Drugs and the Pharmaceutical Sciences[DRUGS PHARM. SCI.]. 1984., 1984. 189. Ahmed, Z. and L. Vining, Evidence for a chromosomal location of the genes coding for chloramphenicol production in Streptomyces venezuelae. Journal of bacteriology, 1983. 154(1): p. 239-244. 190. Vitayakritsirikul, V., et al., Improvement of chloramphenicol production in Streptomyces venezuelae ATCC 10712 by overexpression of the aroB and aroK genes catalysing steps in the shikimate pathway. Antonie van Leeuwenhoek, 2016. 109(3): p. 379-388. 191. Doull, J.L., et al., Conditions for the production of jadomycin B byStreptomyces venezuelae ISP5230: Effects of heat shock, ethanol treatment and phage infection. Journal of industrial microbiology, 1994. 13(2): p. 120-125. 192. Kahan, J., et al., Thienamycin, a new β-lactam antibiotic I. Discovery, taxonomy, isolation and physical properties. The Journal of antibiotics, 1979. 32(1): p. 1-12. 193. Coulton, S. and E. Hunt, 3 Recent Advances in the Chemistry and Biology of Carbapenem Antibiotics, in Progress in medicinal chemistry. 1996, Elsevier. p. 99-145. 194. Núñez, L.E., et al., The biosynthetic gene cluster for the β-lactam carbapenem thienamycin in Streptomyces cattleya. Chemistry & biology, 2003. 10(4): p. 301-311. 195. Hendlin, D., et al., Phosphonomycin, a new antibiotic produced by strains of Streptomyces. Science, 1969. 166(3901): p. 122-123.

144

196. Majumdar, M. and S. Majumdar, Effects of minerals on neomycin production by Streptomyces fradiae. Applied microbiology, 1965. 13(2): p. 190-193. 197. Sekurova, O.N., et al., Activation of chloramphenicol biosynthesis in Streptomyces venezuelae ATCC 10712 by ethanol shock: insights from the promoter fusion studies. Microbial cell factories, 2016. 15(1): p. 85. 198. Johnson, R.A. and D.W. Wichern, Multivariate analysis. Encyclopedia of Statistical Sciences, 2004. 8. 199. Norman, G., Likert scales, levels of measurement and the “laws” of statistics. Advances in health sciences education, 2010. 15(5): p. 625-632. 200. Garbeva, P., et al., Transcriptional and antagonistic responses of Pseudomonas fluorescens Pf0-1 to phylogenetically different bacterial competitors. The ISME journal, 2011. 5(6): p. 973-985. 201. George, D. and P. Mallery, SPSS for Windows Step by Step: A Simple Guide and Reference 18.0 Update. 2010. 202. Trochim, W.M. and J.P. Donnelly, Research methods knowledge base. 2001. 203. Gravetter, F.J. and L.B. Wallnau, Essentials of Statistics for the Behavioral Sciences (PSY 200 (300) Quantitative Methods in Psychology). 2010: Boston: Cengage Learning. 204. Dubos, R.J. and R.D. Hotchkiss, The production of bactericidal substances by aerobic sporulating bacilli. Journal of Experimental Medicine, 1941. 73(5): p. 629-640. 205. O'SULLIVAN, J., et al., Janthinocins A, B and C, novel peptide lactone antibiotics produced by Janthinobacterium lividum. The Journal of antibiotics, 1990. 43(8): p. 913-919. 206. Kevany, B.M., D.A. Rasko, and M.G. Thomas, Characterization of the complete zwittermicin A biosynthesis gene cluster from Bacillus cereus. Applied and environmental microbiology, 2009. 75(4): p. 1144-1155. 207. Silo-Suh, L.A., et al., Biological activities of two fungistatic antibiotics produced by Bacillus cereus UW85. Applied and environmental microbiology, 1994. 60(6): p. 2023-2030. 208. Jakobi, M., et al., Maltophilin: A new antifiingal compound produced by Stenotrophomonas maltophilia R3089. The Journal of antibiotics, 1996. 49(11): p. 1101- 1104. 209. Challis, G.L. and D.A. Hopwood, Synergy and contingency as driving forces for the evolution of multiple secondary metabolite production by Streptomyces species. Proceedings of the National Academy of Sciences, 2003. 100(suppl 2): p. 14555-14561. 210. Chan, C.T., et al., 'Deadman'and'Passcode'microbial kill switches for bacterial containment. Nature chemical biology, 2016. 12(2): p. 82.

145

Appendix I

The following macro allows the user to analyse 4x4 matrixes and quantify the integrated luminescence of compartments.

A1.1 Pre-processing image macro. run("Scale...", "x=5 y=5 interpolation=Bilinear average create"); run("Grays");

//get file info getDimensions(width, height, channels, slices, frames);

//set canvas size canvas=floor(sqrt((width*width)+(height*height)));

//draw line and work out angle setTool("line"); title = "Get info"; msg = "draw line for new angle\n then select OK to continue"; waitForUser(title, msg); getLine(x1, y1, x2, y2, lineWidth); op=x1-x2; ad=y1-y2; an=atan2(ad, op); an=180-(an*180/PI); if (an>270){ an=an-360; } else if (an>90){ an=an-180; }

//change canvas size and rotate //size="width=" + canvas + " height=" + canvas + " position=Center zero"; //run("Canvas Size...", size); rot="angle=" + an + " interpolation=Bilinear"; run("Rotate... ", rot); setTool("rectangle");

A1.2. Create grid macro

//initialize run("Set Measurements...", "area mean min center centroid display redirect=None decimal=3");

146 run("Set Scale...", "distance=0 known=0 pixel=1 unit=pixel"); run("Clear Results"); roiManager("Reset"); Well=newArray("A","B","C","D","E","F","G","H","I","J","K","L","M","N","O","P","Q","R","S","T"); stnum=newArray("1","12"); Curr=0; fn=getTitle(); getDimensions(ImageWidth, ImageHeight, ImageChannels, ImageSlices, ImageFrames);

//options Dialog.create("Overlay options"); Dialog.addMessage("All measurements are in pixels"); Dialog.addNumber("Size of ROI:", 15); Dialog.addNumber("Spacing within the Pattern in X and Y:", 19); Dialog.addMessage("Pattern options"); Dialog.addNumber("Pattern Start Label number - Column:", 1); Dialog.addNumber("Pattern Start Label number - Row:", 1); Dialog.addNumber("Numberof Patterns Across:", 11); Dialog.addNumber("Number of Patterns Down:",15); Dialog.addMessage("Pattern spacing"); //Dialog.addRadioButtonGroup("Pattern Start number:", stnum, 1, 1, 1); Dialog.addNumber("Pattern Spacing in X:", 116); Dialog.addNumber("Pattern Spacing in Y:", 112); Dialog.addNumber("Initial ROI position in X and Y:", 30); Dialog.show();

//responses diam=Dialog.getNumber(); dotSpacing=Dialog.getNumber(); stX=Dialog.getNumber(); stY=Dialog.getNumber(); across=Dialog.getNumber(); down=Dialog.getNumber(); //st=Dialog.getNumber(); //st=Dialog.getRadioButton(); groupSpacingX=Dialog.getNumber(); groupspacingY=Dialog.getNumber(); RoiPos=Dialog.getNumber();

//initial roi makeRectangle(RoiPos, RoiPos, diam, diam); //makeOval(RoiSize, RoiSize, diam, diam); waitForUser("","Move roi to top left square"); run("Measure");

147 x=round(getResultString("X", 0))-diam/2; y=round(getResultString("Y", 0))-diam/2;

//start of loop to create roi patterns //rows down for (groupRow=0;groupRow

//columns across for (groupCol=0;groupCol

148

} //end of loop roiManager("Show All"); run("Clear Results");

A1.3 Quantify luminescence macro

//initialize waitForUser("","Select corrected image"); fn=getTitle(); getDimensions(ImageWidth, ImageHeight, ImageChannels, ImageSlices, ImageFrames); run("Clear Results"); roiManager("Deselect"); roiManager("measure");

149

Appendix II – Terms for using previously published work

AII.1. Oxford University Press

This Agreement between Imperial College London -- Daniel Blicher Holst Hansen ("You") and Oxford University Press ("Oxford University Press") consists of your license details and the terms and conditions provided by Oxford University Press and Copyright Clearance Center. License Number 4494300601097 License date Dec 22, 2018 Licensed content Oxford University Press publisher Licensed content FEMS Microbiology Letters publication Construction of a modified mini-Tn5 luxCDABE transposon for the Licensed content title development of bacterial biosensors for ecotoxicity testing Licensed content Weitz, Hedda J.; Ritchie, Jenny M. author Licensed content date Apr 1, 2001 Type of Use Thesis/Dissertation Institution name Title of your work Student Publisher of your Imperial College London work Expected publication Dec 2018 date Permissions cost 0.00 USD Value added tax 0.00 USD Total 0.00 USD Title Student Institution name Imperial College London Expected Dec 2018 presentation date Portions 2C Imperial College London 60 bramber road

Requestor Location london, w14 9pb United Kingdom Attn: Imperial College London Publisher Tax ID GB125506730 Billing Type Invoice

150

Imperial College London 60 bramber road

Billing Address

london, United Kingdom w14 9pb Attn: Imperial College London Total 0.00 USD Terms and Conditions

STANDARD TERMS AND CONDITIONS FOR REPRODUCTION OF MATERIAL FROM AN OXFORD UNIVERSITY PRESS JOURNAL

1. Use of the material is restricted to the type of use specified in your order details.

2. This permission covers the use of the material in the English language in the following territory: world. If you have requested additional permission to translate this material, the terms and conditions of this reuse will be set out in clause 12.

3. This permission is limited to the particular use authorized in (1) above and does not allow you to sanction its use elsewhere in any other format other than specified above, nor does it apply to quotations, images, artistic works etc that have been reproduced from other sources which may be part of the material to be used.

4. No alteration, omission or addition is made to the material without our written consent. Permission must be re-cleared with Oxford University Press if/when you decide to reprint.

5. The following credit line appears wherever the material is used: author, title, journal, year, volume, issue number, pagination, by permission of Oxford University Press or the sponsoring society if the journal is a society journal. Where a journal is being published on behalf of a learned society, the details of that society must be included in the credit line.

6. For the reproduction of a full article from an Oxford University Press journal for whatever purpose, the corresponding author of the material concerned should be informed of the proposed use. Contact details for the corresponding authors of all Oxford University Press journal contact can be found alongside either the abstract or full text of the article concerned, accessible from www.oxfordjournals.org Should there be a problem clearing these rights, please contact [email protected]

7. If the credit line or acknowledgement in our publication indicates that any of the figures, images or photos was reproduced, drawn or modified from an earlier source it will be necessary for you to clear this permission with the original publisher as well. If this permission has not been obtained, please note that this material cannot be included in your publication/photocopies.

8. While you may exercise the rights licensed immediately upon issuance of the license at the end of the licensing process for the transaction, provided that you have disclosed

151 complete and accurate details of your proposed use, no license is finally effective unless and until full payment is received from you (either by Oxford University Press or by Copyright Clearance Center (CCC)) as provided in CCC's Billing and Payment terms and conditions. If full payment is not received on a timely basis, then any license preliminarily granted shall be deemed automatically revoked and shall be void as if never granted. Further, in the event that you breach any of these terms and conditions or any of CCC's Billing and Payment terms and conditions, the license is automatically revoked and shall be void as if never granted. Use of materials as described in a revoked license, as well as any use of the materials beyond the scope of an unrevoked license, may constitute copyright infringement and Oxford University Press reserves the right to take any and all action to protect its copyright in the materials.

9. This license is personal to you and may not be sublicensed, assigned or transferred by you to any other person without Oxford University Press’s written permission.

10. Oxford University Press reserves all rights not specifically granted in the combination of (i) the license details provided by you and accepted in the course of this licensing transaction, (ii) these terms and conditions and (iii) CCC’s Billing and Payment terms and conditions.

11. You hereby indemnify and agree to hold harmless Oxford University Press and CCC, and their respective officers, directors, employs and agents, from and against any and all claims arising out of your use of the licensed material other than as specifically authorized pursuant to this license.

12. Other Terms and Conditions:

AII.2. American society for Microbiology ASM authorizes an advanced degree candidate to republish the requested material in his/her doctoral thesis or dissertation. If your thesis, or dissertation, is to be published commercially, then you must reapply for permission.

152