The Design and Application of a Molecular Profiling Strategy to Identify Polymicrobial Acute Sepsis Infections

University of Calgary PRISM: University of Calgary's Digital Repository

Graduate Studies The Vault: Electronic Theses and Dissertations

2015-08-14 The Design and Application of a Molecular Profiling Strategy to Identify Polymicrobial Acute Sepsis Infections

Faria Crowder, Monica

Faria Crowder, M. (2015). The Design and Application of a Molecular Profiling Strategy to Identify Polymicrobial Acute Sepsis Infections (Unpublished doctoral thesis). University of Calgary, Calgary, AB. doi:10.11575/PRISM/28047 http://hdl.handle.net/11023/2388 doctoral thesis

University of Calgary graduate students retain copyright ownership and moral rights for their thesis. You may use this material in any way that is permitted by the Copyright Act or through licensing that has been assigned to the document. For uses that are not allowable under copyright legislation or licensing, you are required to seek permission. Downloaded from PRISM: https://prism.ucalgary.ca UNIVERSITY OF CALGARY

The Design and Application of a Molecular Profiling Strategy to Identify Polymicrobial Acute

Sepsis Infections

Monica Martins Pereira Faria-Crowder

A THESIS

SUBMITTED TO THE FACULTY OF GRADUATE STUDIES

IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE

DEGREE OF DOCTOR OF PHILOSOPY

GRADUATE PROGRAM IN MICROBIOLOGY, IMMUNOLOGY, AND INFECTIOUS

DISEASES

CALGARY, ALBERTA

AUGUST 2015

Sepsis is a term used to describe an array of clinical presentations ranging from mild body dysfunction to multiple organ failure. These clinical signs result from a systemic inflammatory response to microbes or microbial products present in sterile sites such as blood.

Current clinical diagnostics rely on culture techniques to identify systemic infections. However, culture lacks sensitivity and a positive result is only obtained in 40% of cases thereby limiting our knowledge of sepsis microbiology.

This doctoral study described the development of methods for direct detection of bacteria or bacterial products in blood. A method of lysing host cells and a bacterial DNA extraction protocol was developed and evaluated on mock bacterial communities spiked into whole blood.

The results indicated that viable bacteria could be recovered down to 10 CFU/ml using this method. Paired-end Illumina sequencing of the 16S rRNA gene also indicated that the bacterial

DNA extraction method enabled recovery of bacterial DNA from spiked blood.

This method demonstrated improved detection of systemic bacterial infections involving bacteria as well as their products in three cohorts of clinically septic patients. Application of the paired-end Illumina 16S rRNA sequencing to saponin treated blood from intensive care unit

(ICU) and emergency department (ED) patients indicated there were bacterial DNA profiles present in whole blood. These patterns were examined alongside the patient’s clinical data and indicated common molecular profiling patterns were associated with the primary source of infection. Polymicrobial DNA was present in the blood samples with the taxonomic profiles suggesting commensal microbiota were implicated in addition to a principal pathogen. Bacterial

DNA from Streptococcus and Staphylococcus were abundant in patients that died in the ICU.

ii Overall this study identified common bacterial DNA patterns in the blood of septic patients which were associated with the patients’ primary source of infection, implicated the commensal microbiota in systemic infection and suggested that sepsis infection may not always involve persistent bacterial bloodstream infections. Rather, this study concluded that bacterial products or viable organisms are likely cyclically present and cleared from the bloodstream resulting in a robust inflammatory response.

iii

Acknowledgements

I would like to acknowledge my supervisors, Dr. Michael Surette and Dr. John Conly, for their tireless efforts on this thesis research. Dr. Michael Surette has provided extensive support on the scientific development of this project and the technical aspects of this project. Further, his insight into microbial populations and the dynamics of the host-pathogen interactions at the interface of infection have provided me with an excellent knowledge base during my doctoral training. Dr. John Conly has provided invaluable clinical insight into this project. He has helped me interpret clinical data and understand the implications of this doctoral research in a clinical environment. Without Dr. Conly’s support on this project the interpretation of this data would have been limited and many of the important conclusions would not have been made.

I would like to acknowledge my supervisory committee members Dr. Michael Parkins and Dr. Paul Kubes for their support during my doctoral work. I would like to acknowledge the many members of the Alberta Sepsis Network who facilitated the sample collection and clinical data collection for this study. In particular, Joseé Wong was instrumental in enabling the sample collection for this study. Dean Yergens assisted in the amalgamation of the immense amount of clinical data for this study. I would also like to acknowledge the research nurses and assistants who identified and enrolled patients with the ICU and ED at Foothills Medical Centre and the

Alberta Children’s Hospital including Janice Hammond, Linda Knox, Christine Skinner, Dori-

Ann Martin, Claudia Maki, Stacey Ruddell, and Dan Lane. Lastly, I would like to acknowledge

Dr. Brent Winston for allowing us to use the CCEPTR resources and Dr. Chip Doig for allowing me to contribute to the Alberta Sepsis Network team as well as providing funding for this doctoral project. I would also like to acknowledge Derrice Knight for her efforts in organizing

iv the Alberta Sepsis Network research days, annual meetings, and facilitating the knowledge transfer within the group.

I would like to acknowledge members of the Surette laboratory for their support throughout the years and their assistance in the analysis of the data set. In particular, Dr. Jennifer

Stearns provided extensive support on the analysis and bioinformatics for this project. I would also like to thank Dr. Christopher Sibley for teaching me many of the techniques used in this project and for instilling his knowledge of bacterial identification on me.

Lastly, I want to acknowledge my family for all their support during this doctoral project.

They kept me grounded and provided the strength I needed to overcome many of the challenges faced during my project.

v Dedication

This thesis is dedicated to my husband, Ray Crowder who provided a support base for me throughout my PhD program. He has given me the light I needed in darker times and challenges me daily to be a better person. For that I am forever grateful.

vi Table of Contents

Abstract ...... ii Acknowledgements ...... iv Dedication ...... vi Table of Contents ...... vii List of Tables ...... xi List of Figures and Illustrations ...... xii List of Symbols, Abbreviations, and Nomenclature ...... xiv

CHAPTER ONE: INTRODUCTION ...... 1 1.1 Sepsis ...... 1 1.1.1 Definition of Sepsis ...... 1 1.1.1.1 Sepsis vs. Bacteremia ...... 2 1.1.2 Clinical Significance and Epidemiology of Sepsis ...... 2 1.2 Sepsis Infections ...... 4 1.2.1 Underlying Sepsis Infections ...... 4 1.2.2 Infectious Agents in Sepsis ...... 5 1.2.2.1 Bacteria ...... 5 1.2.2.2 Fungi ...... 7 1.2.2.3 Viruses ...... 8 1.2.3 Microbial Translocation into the Bloodstream ...... 8 1.2.4 Inflammatory Response in Sepsis ...... 10 1.3 Sepsis Management and Therapy ...... 13 1.3.1 Antibiotic Therapy ...... 13 1.3.2 Supportive Therapy in Sepsis ...... 16 1.3.3 Immunosuppressive/Immunomodulatory Therapy in Sepsis ...... 17 1.4 Clinical Identification of Sepsis ...... 19 1.4.1 Physiological Identification of Sepsis ...... 19 1.4.1.1 Immunological Bio-Markers ...... 20 1.4.2 Blood Culture Diagnostics ...... 22 1.4.2.1 Advantages of Blood Culture ...... 23 1.4.2.2 Limitations of Blood Culture ...... 23 1.5 Molecular Approaches to Sepsis Diagnostics ...... 25 1.5.1 Proteomics in Sepsis Diagnostics ...... 25 1.5.2 Molecular Approaches using Blood Culture ...... 29 1.5.3 Blood Culture Independent Molecular Approaches ...... 31 1.5.4 Application of Next-Generation Sequencing to Sepsis ...... 35 1.6 Rationale for the Study ...... 36 1.7 Hypothesis and Objectives ...... 36 1.7.1 Research Objectives ...... 37

CHAPTER TWO: MATERIALS AND METHODS ...... 38 2.1 Bacterial Strains and Culture Conditions ...... 38 2.1.1 Bacterial Strains and Plate Pools ...... 38 2.1.2 Culture Conditions ...... 38 2.1.2.1 Liquid Media ...... 38

vii 2.1.2.2 Solid Media ...... 40 2.2 Patient Cohorts and Ethics ...... 42 2.2.1 Ethical Approval ...... 42 2.2.2 Enrolment Criteria ...... 43 2.2.2.1 Adult ICU Septic Patients ...... 43 2.2.2.2 Non-Septic ICU Patients ...... 43 2.2.2.3 Adult ED Patients ...... 44 2.2.2.4 Paediatric ED Patients ...... 44 2.2.2.5 Adult Healthy Control Individuals ...... 44 2.2.3 Patient Admissions and Clinical Diagnostic Data ...... 45 2.3 Collection of Human Blood and Tissue Samples ...... 45 2.3.1 Collection of Human Blood ...... 45 2.3.2 Collection of Biologic Tissue and Fluids from ICU patients ...... 47 2.4 Processing of Clinical Samples ...... 47 2.4.1 Saponin Treatment of Blood ...... 47 2.4.2 Quantitative Culture ...... 48 2.5 Partial 16S rRNA gene PCR and Sequencing ...... 49 2.5.1 Colony Preparation ...... 49 2.5.2 PCR Mix and Cycle Conditions ...... 49 2.5.3 Agarose Gel Electrophoresis ...... 50 2.5.4 PCR Product Sequencing ...... 50 2.6 DNA Extraction from Clinical Samples ...... 51 2.6.1 In-house DNA Extraction Method ...... 51 2.6.2 Validation of DNA Extraction Protocol ...... 52 2.6.2.1 Comparison of Extraction With or Without Lytic Enzymes ...... 52 2.6.2.2 Comparison of Dynabeads to Zymo DNA Clean & Concentrator-25 ...53 2.7 Terminal-Restriction Fragment Length Polymorphism ...... 54 2.7.1 PCR Reaction and Cycle Conditions ...... 54 2.7.2 PCR Purification ...... 54 2.7.3 CfoI Digestion and Purification ...... 55 2.7.4 Fragment Analysis ...... 55 2.7.5 T-RF Determination for Clinical Isolates ...... 55 2.8 Fluorescent Activation Cell Sorting (FACS) ...... 56 2.9 Paired-end Illumina 16S rRNA Sequencing ...... 56 2.9.1 PCR Primers and Conditions ...... 56 2.9.2 Sample Preparation ...... 57 2.9.3 Illumina Sequencing ...... 58 2.9.4 Sequence Processing and Taxonomic Identification ...... 58 2.9.5 Analysis ...... 58 2.9.5.1 Taxonomic Identification and Filtering ...... 58 2.9.5.2 Diversity Measures ...... 59 2.10 Mock Community Analysis ...... 61

CHAPTER THREE: METHOD DEVELOPMENT AND OPTIMIZATION ...... 63 3.1 Introduction ...... 63 3.2 Method Development ...... 64 3.2.1 Blood Collection and Anticoagulant Selection ...... 64 viii 3.2.2 Saponin Blood Treatment ...... 64 3.2.2.1 Addition of Saponin Improved Microbial Recovery ...... 64 3.2.2.2 Saponin Did Not Impede Microbial Growth ...... 65 3.2.3 Enzymatic Lysis and Column Purification ...... 70 3.2.4 Purification and Concentration of DNA ...... 73 3.2.5 Saponin Treatment of Blood Spiked with Mock Bacterial Communities ...... 78 3.2.5.2 Summary of Culture-Dependent Results ...... 84 3.2.5.3 Culture-Independent Results ...... 88 3.2.6 Case Studies ...... 97 3.2.6.1 Filtering OTU Table ...... 98 3.2.6.2 Removal of Low Abundance OTUs ...... 99 3.2.6.3 Case-Studies ...... 99 3.3 Discussion ...... 104

CHAPTER FOUR: BACTERIAL DNA PROFILES FROM INTENSIVE CARE UNIT PATIENTS ...... 115 4.1 Introduction ...... 115 4.2 Results ...... 116 4.2.1 Sample Cohorts ...... 116 4.2.2 Bacterial DNA Patterns in Septic Blood ...... 116 4.2.2.1 Sequence Filtering ...... 117 4.2.2.2 OTU Diversity of Streptococcus and Staphylococcus ...... 118 4.2.2.3 Bacterial DNA patterns in ICU SB samples ...... 120 4.2.2.4 Blood Culture Data and Bacterial DNA Profiles ...... 136 4.2.2.5 Characteristics of Patients in the SB Clusters ...... 138 4.2.2.6 Septic ICU Patient Blood Samples that did not Cluster ...... 138 4.2.3 Bacterial DNA Profiles of Blood from Healthy Adults ...... 140 4.2.3.1 OTU Diversity in Healthy Blood Samples ...... 143 4.2.3.2 The HB Samples Clustered Separately from the SB Samples ...... 149 4.2.3.3 Clustering of SB and HB Samples ...... 152 4.2.4 Changes to Bacterial DNA Patterns during ICU Admission ...... 155 4.2.4.1 ASN464 Case Study ...... 155 4.2.4.2 ASN465 Case Study ...... 160 4.2.4.3 ASN469 Case Study ...... 161 4.2.4.4 ASN475 Case Study ...... 162 4.2.5 Septic ICU Patient’s Primary Infection Samples ...... 163 4.2.5.1 In-depth Culturing of Sepsis Primary Infections ...... 163 4.2.5.2 Correlation Between Primary Infection and Whole Blood Molecular Profiles ...... 166 4.2.5.3 Taxonomic Diversity in Primary Infection Samples ...... 171 4.2.6 Discussion ...... 174

CHAPTER FIVE: BACTERIAL DNA PROFILES FROM EMERGENCY DEPARTMENT PATIENTS ...... 190 5.1 Introduction ...... 190 5.2 Results ...... 191 5.2.1 Sample Cohorts ...... 191

ix 5.2.2 Culture Results from Saponin Treated Whole Blood ...... 191 5.2.3 Molecular Profiling of Emergency Blood Samples ...... 194 5.2.4 Molecular Profiling of Children’s Blood Samples ...... 202 5.2.1 Comparison of CB and EB Patients to Healthy Controls ...... 209 5.2.2 Comparison of Bacterial DNA Patterns between ED and ICU Whole Blood Samples ...... 212 5.2.3 Discussion ...... 227

CHAPTER SIX: DISCUSSION ...... 236 6.1 Overview of Findings ...... 236 6.1.1.1 Method Development ...... 236 6.1.1.2 Application to Clinical Samples ...... 238 6.2 Challenges and Limitations ...... 242 6.3 Future Directions ...... 249 6.4 Clinical Implications ...... 252 6.5 Conclusions ...... 258

x List of Tables

Table 2.1: Bacterial strains and plate pools used in this study...... 39

Table 2.2:Solid-media used for this study...... 41

Table 2.3:Patient cohorts used in the study...... 46

Table 2.4: OTUs identified as contamination and removed from OTU table...... 60

Table 3.1: Percent recovery and limits of detection for synthetic community bacteria...... 87

Table 3.2: Taxonomic identification and relative OTU abundance of synthetic community organisms...... 92

Table 3.3: OTU representative sequence taxonomic identification and alignments for synthetic community samples...... 96

Table 4.1: Breakdown of OTU taxonomic identities for Streptococcus and Staphylococcus abundant in SB samples...... 119

Table 4.2: SB clusters, clinical microbiology, and OTU identification...... 125

Table 4.3: Admission data for the SB patients in Groups 1-3...... 130

Table 4.4: Septic ICU patient admissions and outcome data for the three groups of SB...... 139

Table 4.5: Antibiotics administered from day 1-28 for SB case studies...... 159

Table 4.6: Culture results from sepsis patient’s primary infections...... 164

Table 4.7: Comparison of the diversity with each sample from a patient’s primary infection and SB sample...... 173

Table 5.1: Agar-based culture results from saponin treated blood compared to blood culture. . 192

Table 5.2: OTUs identified in the EB whole blood samples...... 197

Table 5.3: Clinical data and OTU abundance for FED patients in the EB cohort...... 199

Table 5.4: OTU abundance for the CB cohort samples...... 203

Table 5.5: Clinical data and OTU abundance for AER-G2 patients in the CB cohort...... 207

Table 5.6: Clinical data for all the whole blood clinical samples clustered based on weighted UniFrac UPGMA...... 216

xi List of Figures and Illustrations

Figure 1.1: Molecular approaches that are currently being assessed for application in sepsis diagnostics ...... 26

Figure 3.1: Preliminary experiment on ASN087 samples indicated the addition of saponin to whole blood improved recovery of microbial DNA ...... 66

Figure 3.2: Saponin did not impact the viability of bacteria...... 68

Figure 3.3: FACS sorting of whole blood spiked with ASN087 culture pools and treated with 0.85% saponin indicated that microbial cells were still intact...... 71

Figure 3.4: Enzymatic digestion requirement for bacterial DNA extraction of whole blood...... 74

Figure 3.5: Zymo column based purification recovered more T-RFs when compared to Dyna magnetic bead based purification ...... 76

Figure 3.6: Limit of detection for synthetic communities of bacteria spiked into whole blood. .. 80

Figure 3.7: The percent recovery of bacteria from a mixed community spiked into whole blood...... 82

Figure 3.8: Recovery of individual isolates spiked into whole blood...... 85

Figure 3.9: Illumina 16S rRNA sequencing of DNA from Synthetic Communities ...... 89

Figure 3.10: OTU abundance of 16S rRNA Illumina sequenced DNA from synthetic communities ...... 93

Figure 3.11: Bacterial DNA profiles of case studies from septic ICU patients ...... 100

Figure 4.1: Septic whole blood samples cluster into three groups based on their taxonomic bacterial DNA profiles ...... 121

Figure 4.2: PCoA of SB samples that had low sequencing depth indicate they cluster mainly with the Group 2 samples ...... 141

Figure 4.3: The bacteria DNA profiles of healthy blood ...... 144

Figure 4.4: The bacterial DNA profiles of healthy adults blood were distinct from adult septic patient’s blood ...... 150

Figure 4.5: HB Samples cluster separately from Group 1 and Group 3 SB samples but are found within Group 2 SB samples ...... 153

Figure 4.6: The bacterial DNA profile in blood changes during a patient's stay in the ICU ...... 156

xii Figure 4.7: Molecular profiling of primary infection samples and whole blood from septic ICU patients ...... 168

Figure 5.1: The taxonomic profile of adult ED patients indicated polymicrobial DNA was present in whole blood ...... 195

Figure 5.2: Taxonomic profiles of CB samples ...... 204

Figure 5.3: The whole blood samples from CB and EB patients cluster separately from HB samples ...... 210

Figure 5.4: The bacterial DNA profiles of ICU and ED samples clustered together and separately from HB samples ...... 213

xiii List of Symbols, Abbreviations, and Nomenclature

Symbol Definition AF Abscess fluid ANS Autonomic nervous system APACHE Acute Physiology and Chronic Health Evaluation ASN Alberta sepsis network BAL Bronchoalveolar lavage BHI Brain-heart infusion BHIco Brain-heart infusion with colistin and oxolinic acid BLAST Basic Local Alignment Search Tool bp Base pair CB Pediatric patient blood CBA Colombia blood agar CCEPTR Critical Care Epidemiologic and Biologic Tissue Bank Resource CFU Colony forming unit Choc Chocolate agar CNA Colombia CNA agar CoNS Coagulase-negative Staphylococcus CRP C-reactive protein CT Chest tube fluid DNA Deoxyribonucleic acid EB Emergency patient blood ED Emergency department ELISA Enzyme-linked immunoabsorbent assay ESI Electrospray ionization ETT Endotracheal tube fluid FAA Fastidious anaerobe agar FACS Fluorescent activated cell sorting FISH Fluorescent in-situ hybridization GM-CSF Granulocyte macrophage colony stimulating factor GPAC Gram-positive anaerobic cocci HB Healthy adult blood HMP Human microbiome project HOMD Human Oral Microbiome Database IC ICU control patient blood ICU Intensive care unit

xiv IFN Interferon IL Interleukin iNKT Invariant natural killer T-cells IQR Inter-quartile range JP Jackson-Pratt fluid LC-SF Light-cycler SepsiFast LOS Length of stay LPS Lipopolysaccharide MAC MacConkey agar MALDI-TOF Matrix-assisted laser desorption-ionization-time-of-flight McKay McKay agar ml Milliliter MRSA Methicillin-resistant Staphylococcus aureus MS Mass spectrometry MSA Mannitol-salt agar NCBI National Center for Biotechnology Information NETs Neutrophil extracellular traps NLRs (NODs)-like receptors NODs Nucleotide-binding oligomerization domain NTC Negative template control OTU Operational taxonomic unit PAMPs Pathogen-associated molecular patterns PBS Phosphate-buffered saline PCoA Principal coordinates analysis PCR Polymerase chain reaction PF Peritoneal fluid pmol Picomolar PNA Peptide nucleic acid PRRs Pattern recognition receptors QIIME Quantitative Insights Into Microbial Ecology rcf Relative centrifugal force RIGs RNA-binding RNA helicases RNA Ribonucleic acid rRNA Ribosomal RNA RT-PCR Real-time PCR SB Sepsis patient blood SC Synthetic community SI Sepsis patient infection

xv SIRS Systemic inflammatory response SMG Streptococcus milleri group SN ICU neurological trauma patient blood SOFA Sequential organ failure assessment SP Sputum sp. Species SSC Surviving sepsis campaign T-RF Terminal restriction fragment TF Tissue factor Th Helper T-cell TLR Toll-like receptor TNF Tumor necrosis factor TREM Triggering receptor expressed on myeloid cells TRFLP Terminal restriction fragment length polymorphism TSA Tryptic soy agar TSY Tryptic soy yeast TTSS Type-III secretion system UPGMA Unweighted Pair Group Method with Arithmetic Mean UR Urine VRE Vancomycin-resistant Enterococcus α Alpha β Beta γ Gamma μl Microliter

xvi

Chapter One: Introduction

1.1 Sepsis

1.1.1 Definition of Sepsis

The word “sepsis” originates from a Greek word meaning decay or putrefaction and was first documented in the poetry of Homer [1]. In 1990, the public awareness of sepsis was heightened following the January 19th release of the Morbidity and Mortality Weekly Report, published by the Centers for Disease Control and Prevention, which indicated sepsis rates were rising and accounted for a significant portion of annual health-care expenditures in the USA [2].

A consensus definition for sepsis was derived by the American College of Chest Physicians and the Society of Critical Care Medicine in 1992 [3]. The resulting definition of sepsis was that of a systemic inflammatory response syndrome (SIRS) in which pathogenic or potentially pathogenic microorganisms invaded normally sterile tissues, fluids or body cavities [3]. SIRS itself can be septic or non-septic and is present when patients have one or more of four clinical criteria: body temperature >38°C or <36°C, heart rate >90 beats per minute, hyperventilation with >20 breaths per minute, and a white blood cell count >12,000 cells/µl or <4000 cells/µl [2,4]. Sepsis exists as a continuum of clinical severity including sepsis, severe sepsis and septic shock. Sepsis is defined as SIRS criteria plus evidence of infection or suspected infection. Severe sepsis is defined as sepsis plus evidence of organ dysfunction. Septic shock is defined as severe sepsis plus evidence of acute circulatory failure or persistent hypotension that cannot be explained by any other causes [2]. Along each stage in the sepsis continuum there is greater need for clinical intervention as the concomitant risk for mortality increases [2]. From a microbiologic

perspective, sepsis is the result of interactions between microorganisms and/or their products with host immune factors that are released in response to their presence [5].

1.1.1.1 Sepsis vs. Bacteremia

Often the terms ‘sepsis’ and ‘bacteremia’ are used interchangeably. However, bacteremia may occur without the systemic inflammation characteristic of sepsis. Bacteremia can be categorized as transient, intermittent or persistent. The transient form of bacteremia lasts from minutes to hours and is generally associated with medical procedures involving anatomical sites colonized with normal microbiota [5]. These cases are often non-problematic and result after dental work, endoscopy/colonoscopy, or catheterization [5]. Intermittent bacteremia is seen in the cases of closed-space infections such as abscesses or focal point infections such as pneumonia. These infections are cyclical and lead to recurring episodes of bacteremia, due to the repeated entrance and clearance of the same organisms in the bloodstream. Persistent low-grade bacteremia is often a result of an intravascular focus of infection such as those seen in infective endocarditis or vascular graft infections [5]. These infections can be harder to detect due to a lower bacterial load. Failure to treat any of these forms of bacteremia or the presence of co- morbidities in a patient can result in the progression from bacteremia to sepsis.

1.1.2 Clinical Significance and Epidemiology of Sepsis

Sepsis is a global problem accounting for approximately 2% of all hospitalizations in developed countries and accounts for 6-50% of intensive care unit (ICU) admissions [6,7].

Mortality rates have ranged from 30% in developed countries to up to 80% in underdeveloped

countries [2]. Those who do survive sepsis are likely to have permanent organ damage, cognitive impairment, and physical disability [4].

In a 1995 study, Rangel-Frausto and colleagues examined the incidence of sepsis in 3

ICUs in the USA over a 9-month period [8]. This study, considered one of the most comprehensive studies on the clinical significance of early stage sepsis, revealed that 68% of

ICU patients had at least two SIRs criteria during their time in the ICU [8]. Of these patients,

26% developed culture-positive sepsis, 18% developed severe sepsis and 4% developed septic shock [8]. Patients with microbiology confirmed sepsis were at a higher risk of transitioning into severe sepsis, with 64% of all sepsis patients transitioning within 1 day of diagnosis.

Despite modern technologies and advances in health care, sepsis rates continue to climb and have more than doubled in the last ten years [7]. This increase is due to an aging population, the growing number of immunocompromised individuals, increasing use of invasive procedures and increasing numbers of antibiotic resistant organisms [9,10]. The costs associated with sepsis care are staggering with an estimated $14.6 billion spent on hospital care for septic patients per annum in the USA with each individual patient costing the medical system approximately

$22,000 [11]. In Canada, the Canadian Institute for Health Information revealed that from 2008-

2009 more than 30,500 patients were hospitalized with sepsis with a mortality rate over 30% and that it is one of the leading causes of in-hospital mortality accounting for 10.9% of hospital deaths from 2008-2009 [10]. Although sepsis can occur at any age, a risk factor is age, as individuals over 65 and infants less than 1 year of age are predisposed to higher risk. In Canada,

60.6% of all sepsis hospitalizations occurred in adults above 60 years of age with a median age of 66 years [10]. Neonates are at risk for developing severe sepsis due their prematurity and low

birth-weight [4]. Furthermore, in patients diagnosed with sepsis, 44.5% presented with pre- existing co-morbidities with diabetes and cancer being the most frequently encountered [10].

1.1.2.1.1 Epidemiology of Sepsis in Southern Alberta

Almost a decade ago, a prospective cohort study was done to look at the epidemiology of sepsis in the Calgary region [12]. Of all the patients included in the study, 5% had culture confirmed bloodstream infections at initial presentation [12]. The median age of patients was 63 years and the sepsis rate was higher in males over females. The median ICU stay was 2.1 days with an ICU mortality rate of 17% [12]. In a parallel study examining bacteremia in the Calgary region, the majority of the cases (80%) of severe bloodstream infections were considered community acquired [13].

1.2 Sepsis Infections

1.2.1 Underlying Sepsis Infections

The presence of several types of underlying infections can ultimately lead to sepsis. The infections themselves may be community associated or nosocomial with previous studies indicating roughly equal rates [14,15]. However, there is a rise in the cases of community acquired bloodstream infections with an increasing incidence of multi-drug resistant Gram- negative bacilli and methicillin-resistant Staphylococcus aureus (MRSA) [16].

The majority of bloodstream infections in sepsis often arise from sources of primary infections including pneumonia, endocarditis, catheter infections, intra-abdominal infections, urogenital infections, and surgical wound infections [4]. Out of these, respiratory infections are

responsible for the majority of sepsis cases with pneumonia-associated sepsis having the highest mortality rate [10,11,17]. Intra-abdominal sepsis results from an infection penetrating into the sterile peritoneum, for example, when organ perforation from trauma allows microorganisms to enter a normally sterile site [4]. These infections can be generalized as in the case of peritonitis or localized as in the case of an intra-abdominal abscess. In urogenital sepsis, fecal microbiota or catheter-associated microbiota will migrate and colonize the urethra [13,14]. Endocarditis is defined as inflammation of the inner-layer of the heart. Infective endocarditis involves bacteria attaching to a valve surface and forming a vegetative body composed of cells, platelets, and fibrin [4]. Organisms shed from the infection site can enter the bloodstream, which can result in sepsis [4].

1.2.2 Infectious Agents in Sepsis

1.2.2.1 Bacteria

Although any microbial agent can be implicated in sepsis, over 80% of bloodstream infections are attributed to bacteria [15,18-21]. The most commonly isolated bacteria from blood cultures are coagulase-negative Staphylococci (CoNS), S. aureus, Enterococcus species,

Escherichia coli, and Pseudomonas aeruginosa [18]. An international study on the prevalence of infections in the ICU indicated that 62.2% of blood culture culture-confirmed infections were

Gram-negative, 46.8% were Gram-positive, and 4.5% were anaerobes [22].

In terms of Gram-negative infections, the most common organisms isolated are

Pseudomonas species, E. coli, Klebsiella species, Enterobacter species, and Acinetobacter species [18,22,23]. These Gram-negative bacteria are often ubiquitous and can survive in

hospital water supplies (P. aeruginosa) or in desiccated environments even after disinfection (A. baumannii) [18]. Many of these organisms are multi-drug resistant with up to 60% of the Gram- negative isolates recovered from blood culture identified as resistant to three or more antibiotics

[18]. In Calgary, E. coli has been identified as a common cause of bloodstream infections with fluoroquinolone resistant E. coli implicated in these infections [24]. The extended-spectrum beta- lactamase and carbapenemase producing Enterobacteriaceae are a global problem [18] as is the emergence of pan-resistant Acinetobacter species [18,25].

Gram-positive organisms implicated in sepsis mostly fall into three genera: the streptococci, the staphylococci and the enterococci. For the staphylococci, MRSA and methicillin-sensitive S. aureus infections are common [18]. However, there has been an increased incidence of CoNS nosocomial sepsis infections [26]. In clinical culture, the streptococci are often distinguished by their Lancefield groupings despite the reduced accuracy of these groupings based on 16S ribosomal RNA (rRNA) genotyping [27]. Group A streptococci such as S. pyogenes are beta-haemolytic and implicated in severe invasive diseases including sepsis [27]. S. agalactiae is one of the most common causes of neonatal sepsis [27]. Other streptococci are also implicated in several sepsis infections including the Streptococcus milleri group (SMG) consisting of S. anginosus, S. constellatus, and S. intermedius [27]. The alpha- haemolytic S. pneumoniae is one of the principal pathogens associated with culture confirmed community-acquired pneumonia and pneumonia-related sepsis [18]. Enterococcus species are now one of the leading causes of nosocomial infection and the emergence of vancomycin- resistant Enterococcus (VRE) has become a global concern [28].

Anaerobic bacteria often account for a minority of the culture-confirmed bloodstream infections yet they have been associated with an increased risk of mortality [29]. The Gram-

positive anaerobic cocci (GPAC) bacteria account for 25-30% of anaerobes recovered from clinical infections including sepsis [30]. The GPAC includes Anaerococcus species, Finegoldia magna, Micromonas species, Peptoniphilus species, Parvimonas species, Peptococcus species and Peptostreptococcus species [30,31]. Other anaerobes implicated in clinical infections, not just sepsis, are Prevotella species, Bacteroides species, Clostridium species, Veillonella species,

Fusobacterium species, and Porphyromonas species [32]. Recently, an evaluation of the blood- culture confirmed anaerobic bacteremia from 2000-2008 in the Calgary Health Region indicated the overall incidence rate was 8.7 per 100,000 per year [29]. The most commonly isolated anaerobic bacteria were Bacteroides fragilis (39%), Clostridium species (12%), and

Peptostreptococcus species (10%) [29]. This study also identified 22.5% of the culture confirmed anaerobic bloodstream infections were polymicrobial with mixed anaerobic and aerobic bacteria [29].

1.2.2.2 Fungi

Fungal infections in sepsis are the second most common after bacterial infections and account for approximately 10-15% of bloodstream infections [18]. The most common cause is

Candida species but other fungal organisms including Saccharomyces, Aspergillus, and

Cryptococcus have also been implicated [33]. Fungal infections are mostly seen in immunosuppressed/immunocompromised patients such as those with human immunodeficiency virus or cancer [33,34]. The rate of fungal sepsis has been on the rise and is predicted to be a result of the increased use of immunosuppressive agents, broad-spectrum antibiotics, and more aggressive surgical procedures [18,34].

1.2.2.3 Viruses

The role of viruses in sepsis is often overlooked. In the 2008 consensus statement, viral infections were rarely mentioned and many surveys of sepsis fail to mention viruses [35,36].

This may be due to poor diagnostics for viral sepsis limiting their detection, however, newer techniques have recently been developed [35]. Viral sepsis infections are more prevalent in paediatric ICUs and in neonatal ICUs [35]. In neonates, viruses associated with sepsis include herpes simplex virus, enterovirus, and parechovirus [35]. In adults and the elderly, viral infections are often found in associated with bacterial infections [35].

1.2.3 Microbial Translocation into the Bloodstream

In order to cause a systemic infection, organisms must be able to translocate from the infecting site into the vasculature. The majority of the data on microbial translocation comes from bacteria. Specifically, the pathogenic E. coli, Shigella species, Listeria species, Salmonella species, and Yersinia species are known to interact with epithelial cells in the gastrointestinal tract to initiate bacterial translocation [37-40]. Some of the mechanisms used for translocation involve cell-adhesion proteins as bacterial receptors and the type III secretion system (TTSS)

[38]. The TTSS has a needle-like probe that injects bacterial effectors directly into host cells.

Some of these effectors can modify the cytoskeleton in order to allow bacterial entry into the epithelial cells [41].

Inflammation and other host processes also increase the risk of bacterial translocation [42].

Physiologically, the mechanisms of bacterial translocation are diverse and complex. Decreased

blood flow caused by disseminated coagulation can impair the mucosal barrier allowing for bacteria to translocate across the endothelial layer into mesenteric lymph nodes with subsequent translocation to organs and into the bloodstream possible [43]. During inflammation, production of nitric oxide disrupts tight junctions making mucosal linings hyperpermeable thereby permitting paracellular bacterial translocation [44]. Further, expression of inflammatory cytokines also increases the leakage of microbial products across epithelial and endothelial cells

[45].

Cells, such as enterocytes, express toll-like receptors (TLRs) that are important for sensing bacteria and activating host defense mechanisms [46]. However, some bacteria have developed methods to employ TLRs for phagocytosis and subsequent translocation across mucosal barriers

[41]. Other bacteria are able to use lipid rafts to cross epithelial layers [41]. In airways, pneumococci are able to adhere to epithelial cells and endothelial cells using cell-specific mechanisms for internalization and invasion [47]. These factors are not well understood but interactions with immunoglobulin receptors and platelet activating receptors may be potential mechanisms for entry into the bloodstream [47]. The group B streptococci can also invade endothelial cells using endocytic pathways [48]. Bacterial production of exotoxins such as elastase and protease, which break down host epithelial cells, can also increase permeability of epithelial layers thereby enabling translocation [49,50]. Once bacteria can pass the mucosal and epithelial barriers, the uptake of bacteria by sub-mucosal macrophages can also result in the dissemination of these organisms [41]. In addition, oral bacteria are believed to enter the bloodstream via the gingival sulcus [48].

1.2.4 Inflammatory Response in Sepsis

The complexity involved in the host responses to bacteremia and sepsis goes beyond the scope of this project. For more details, there are many excellent reviews on the subject [46,51-

53]. This section will provide a very superficial overview of the recognition of pathogens and the downstream responses involved.

Initial recognition of invading microbes involves pattern recognition receptors (PRRs) that recognize pathogen-associated molecular patterns (PAMPs) [54]. These cell surface and intracellular receptors include TLRs, RNA-binding RNA helicases (RIGs), and nucleotide- binding oligomerization domain (NODs)-like receptors (NLRs) [55]. In the case of bacteria, cell surface components as well as DNA are the key triggers of the host immune response that are recognized mainly through TLRs [46]. The lipidA portion of Gram-negative bacterial lipopolysaccharide (LPS) is the principal PAMP in Gram-negative organisms. Gram-positive organisms lack LPS but have lipoteichoic acids that are embedded in the thick cell membrane and contain lipoproteins that act as PAMPs [54]. Other PAMPs include flagellin and bacterial super-antigens [54]. Intracellular receptors such as NOD2 recognize the muramyl dipeptide from peptidoglycan [54]. Bacterial DNA is also recognized as a PAMP by TLR9 [54].

Following the PRRs binding to their ligand, there is a signal transduction cascade that results in the transcription of various genes involved in inflammation and immunity. Activation of NLRs via PAMPs results in the formation of a multi-protein complex called the

“inflammasome” [51]. This contributes to the production of pro-inflammatory cytokines and antimicrobial peptides. The triggering receptor expressed on myeloid cells (TREM) found on blood neutrophils and monocytes also magnifies the TLR and inflammasome mediated responses

[51]. Ultimately these signal transduction cascades direct the transcription of pro-inflammatory cytokines including tumour necrosis factor (TNF)-α, interleukin (IL)-1, IL-6, IL-12, and the chemokine IL-8 [54,56]. Neutrophils play a crucial role in the initial response to bacteremia and sepsis infections. These cells migrate rapidly from the blood to the site of infection in response to chemokines such as IL-8 [51]. Neutrophils will recognize and phagocytose foreign microbes through PRRs, antibody presentation and complement receptors that bind to complement and antibody coated organisms [51]. Alternatively, neutrophils can also release their cellular contents and DNA to form neutrophil extracellular traps (NETs), which serve as a physical barrier to trap and block the dissemination of microbes [57]. In addition, the DNA bound proteins in NETs have shown in-vitro and in-vivo killing of bacteria, protozoans, and yeast [58-60].

During sepsis, the adaptive immune response shifts from a helper T-cell(Th)1-cell response to a Th2-cell response [52]. The effects of this are drastic in that the Th1 cells secrete pro-inflammatory cytokines such as IL-12 and interferon (IFN)-γ whereas Th2 cells secrete cytokines such as IL-10 and IL-4 which dampens the inflammatory response [52]. This can result in a prolonged immunosuppressive state, which increases the patient’s risk for acquiring nosocomial infections [61]. This later immunosuppressive state is problematic as over 70% of deaths in sepsis occur after 3 days of ICU intervention and can occur weeks after the initial diagnosis [56].

Many harmful clinical outcomes in sepsis patients can be attributed to the downstream effects of the inflammatory responses, mainly due to the activation of the complement cascade and its subsequent activation of the coagulation pathway. Activation of the complement cascade occurs early in sepsis and is associated with the harmful production of excessive C5a, a potent chemoattractant that triggers the expression and release of proinflammatory proteins [56]. Late-

onset neutrophil dysfunction also results from C5a signalling leading to defective bacterial clearance and potential for lethal secondary bacteremia [52]. C5a also mediates apoptosis of lymphocytes and adrenal medulla cells that further accentuates the immunosuppression. Lastly, the development of septic cardiomyopathy is a direct result of high levels of C5a [52].

Coagulation can lead to further harmful effects once initiated through the complement system. Indeed, the release of C5a is also implicated in the activation of tissue-factor (TF), which activates the extrinsic coagulation pathway [56]. In the early stages, thrombin activation results in hypercoagulability from the intra- and extravascular fibrin production [52]. Thrombin also activates several pro-inflammatory cascades including those involved in TNF, IL-1β, and IL-6 production, which further stimulate coagulation pathways [52]. The pro-inflammatory cytokines also serve to up-regulate the production of TF on mononuclear cells that results in systemic activation of coagulation as these cells circulate in the bloodstream [52]. Pro-inflammatory cytokines also result in the inhibition of fibrinolysis as the pro-inflammatory environment results in elevated levels of fibrinolysis inhibitors such as thrombin-activated fibrinolysis inhibitor [52].

Lastly, the consumption thrombin regulators also contribute the development of intravascular coagulation [52]. The clinical result of this hypercoagulation is disseminated deep-vein thrombosis in sepsis patients [52]. The accumulation of fibrin in the microvasculature is associated with defects in microcirculation, which can ultimately lead to organ dysfunction [52].

During inflammation, neurotransmitters released by the autonomic nervous system

(ANS) mediate interactions between the immune system and the nervous system [52]. The sympathetic ANS mediates the adrenergic pro-inflammatory pathway through the release of catecholamines from the adrenal medulla, sympathetic neurons, phagocytic cells, and lymphocytes [52]. Catecholamines, including dopamine, epinephrine, and norepinephrine, bind

to adrenergic receptors on immune cells, which ultimately results in cytokine release [52]. This further aggravates the adverse effects of the pro-inflammatory response in sepsis. During septic shock, cardiovascular equilibrium is lost as the production and release of catecholamines becomes insufficient, likely due to the apoptosis of medullary cells [52]. Further complications result from the well-documented interactions of bacteria and catecholamines. Work done in 1992 by Lyte and colleagues revealed that catecholamines had direct interactions with Gram-negative bacteria that resulted in enhanced bacterial growth [62]. Since then studies on other enteric bacteria have shown an increase in growth in the presence of catecholamines [63]. The mechanism by which bacteria sense and uptake catecholamines is unclear and a true adrenergic receptor has yet to be identified [63]. Not only does bacterial growth increase but also the binding to the intestinal mucosa is enhanced thereby increasing the risk of translocation of gut bacteria into the bloodstream and subsequent bacteremia [52]. This is facilitated by the loss of barrier function due to the pro-inflammatory mediated epithelial damage and weakening of tight junctions [52].

1.3 Sepsis Management and Therapy

1.3.1 Antibiotic Therapy

Antimicrobial therapy is often administered within the first hour of a patient presenting with the physiological signs of sepsis. In order to help guide a physician on the antimicrobial regimen to administer, the identification of the primary infection site and a Gram-strain of specimens from the site is critical [64]. The Surviving Sepsis Campaign provides general guidelines for the selection and administration of antimicrobials in sepsis [65-67]. A continuous

intravenous infusion of antibiotics is often administered with therapy recommended for 7-10 days [66].

Initial antimicrobial therapy must cover a broad range of Gram-positive, Gram-negative, anaerobic bacteria and occasionally fungi. Combinations of β-lactams and aminoglycosides were often used in sepsis as they display synergistic interactions [64]. This is believed to be a result of

β-lactam mediated membrane damage to bacterial cell walls thereby increasing the uptake of aminoglycosides into the bacteria [68]. However, the emergence of broad-spectrum antibiotics and highly bactericidal antibiotics has reduced the need for aminoglycoside and β-lactam combination therapy [69]. These antibiotics include the extended-spectrum penicillins, third or fourth generation cephalosporins, and the carbapenems. Piperacillin is a broad-spectrum β- lactam antibiotic that is commonly used in combination with the β-lactamase inhibitor tazobactam [70]. The use of piperacillin/tazobactam in sepsis patients is preferred for its broad- spectrum activity and limited adverse side effects [70]. Common cephalosporins used in sepsis include ceftazidime and ceftriaxone [64]. These are often combined with metronidazole, which has antibacterial activity against anaerobic organisms [69].

The use of aminoglycosides, gentamicin or tobramycin, may be warranted for Gram- negative infections. These antibiotics function by inhibition of bacterial protein synthesis [68].

However, aminoglycosides have nephrotoxicity as a potential side effect, which is a concern in sepsis patients [64].

The fluoroquinolones, such as ciprofloxacin, have been shown to have similar efficacy to

β-lactam antibiotics and are predominantly used in treating Gram-negative sepsis [69]. However, the newer class of respiratory fluoroquinolones, such as levofloxacin, have been shown to have good broad-spectrum activity and are recommended for respiratory sepsis infections [71].

Fluoroquinolones induce bacterial cell death through the introduction of DNA double-stranded breaks by targeting the bacterial type II DNA topoisomerases, DNA gyrase and topoisomerase

IV [68]. Macrolides, such as azithromycin, are another class of protein inhibitor antibiotics that may also be used in the setting of pneumonia-related sepsis with studies suggesting better outcomes when macrolide therapy was used [72].

There has been an increase in the presence β-lactam resistant Gram-positive organisms, which has lead to the increased use of glycopeptide antibiotics, mainly vancomycin [15].

Glycopeptides function through inhibition of cell-wall synthesis through binding of peptidoglycan units thereby blocking transglycosylase and transpeptidase activity [68]. However, due to their low permeability, glycopeptides are only effective against Gram-positive organisms

[68]. Patients that have hypersensitivity to β-lactams may be treated with linezolid which functions by inhibition of the 50S ribosome by blocking the initiation of translation [68].

Linezolid has activity against MRSA and VRE and is often used in culture confirmed cases of such infections [69].

In the case of highly antibiotic resistant bacterial infections, the use of tigecycline or colistin may be warranted [73]. Tigecycline is a recently developed antibiotic that can block protein synthesis by binding to the 30S ribosomal subunit thereby inhibiting entry of the transfer-

RNA. It is active against many Gram-positive, Gram-negative, and anaerobic bacteria.

Specifically, it is used as a last resort antibiotic against antibiotic resistant organisms including

MRSA, VRE, and A. baumannii [74]. Colistin is a polymyxin antibiotic that is specifically active towards Gram-negative bacteria by virtue of its interaction with the LPS in the outer membrane.

Colistin functions similar to that of a detergent and solubilizes the membrane leading to cell death [75]. It can be used as a last resort treatment for multidrug resistant P. aeruginosa, A.

baumannii, and Enterobacter species but has significant side effects with both nephrotoxicity and neurotoxicity [75].

1.3.2 Supportive Therapy in Sepsis

The “Surviving sepsis campaign” (SSC) provides universally applicable guidelines for the clinical management of sepsis [67,76]. This guideline is extensive and is updated every four years. Briefly, the SSC outlines guidelines for fluid therapy, catecholamine therapy, vasopressor therapy, steroid therapy, and the use of inotropic agents such as dobutamine [36]. Catecholamine therapy is used to restore systemic circulation and prevent organ damage due to systemic hypotension in sepsis [66]. Vasopressor therapy is used in sepsis when there is persistent hypotension that does not respond to fluid therapy [66]. Corticoid steroid therapy in sepsis is heavily debated but can be used in cases of septic shock to restore systemic blood pressure [77].

Lastly, inotropic agents are often used to improve cardiac output but their use requires a careful evaluation of the patient’s oxygen saturation and cardiac function [78].

Mechanical ventilation is often used for supportive therapy in sepsis patients and the SSC outlines several points to consider with mechanical ventilation including monitoring airway pressure, patient elevation, gradual weaning off mechanical ventilation, and a conservative fluid strategy to prevent lung oedema or weight gain in patients [76]. Glucose control is also an important supportive therapy for sepsis patients with intravenous insulin administered after stabilization in the ICU [36]. In severe sepsis patients, that have acute renal failure, the use of continuous renal replacement therapy is often implemented [36]. Patients with severe sepsis may also receive heparin in order to prevent deep vein thrombosis associated with extensive

coagulation in sepsis [36]. Overall, extensive supportive therapy is often required for sepsis patients in the ICU with the SSC setting international guidelines to help ensure universal implementation of therapy.

There is also Early Goal-Directed Therapy (EGDT), which was introduced as a method of treatment for patients prior to ICU admission [79]. EGDT is initiated in patients presenting with suspected sepsis in the ED or in-patient wards. Therapeutic intervention involves the stabilization of patients based on outcome measures (temperature, heart rate, urine output, and blood pressure) prior to admission to the ICU or hospital ward. The EGDT model also includes the administration of antibiotics within the first hour of suspected sepsis [80]. Overall, implementation of EGDT has resulted in significant reduction in morbidity, mortality, vasopressor use, and health-care costs when compared to the standard-therapy [80].

1.3.3 Immunosuppressive/Immunomodulatory Therapy in Sepsis

Given the many harmful effects of the inflammatory response in sepsis, much of the development in immunomodulatory therapy in sepsis traditionally focused on suppressing the proinflammatory responses observed. However, in over 30 clinical trials, anti-cytokine and anti- inflammatory drugs failed to show a benefit in sepsis patients with some drugs having adverse effects and reduction in survival rates [56]. Recent work has highlighted that septic patients can have markedly different inflammatory responses in sepsis. Cohorts of patients may have a rapid production of both pro-inflammatory and anti-inflammatory cytokines co-currently whereas other patients may have a predominance of anti-inflammatory cytokines or globally depressed cytokine production [56].

Based on this, newer approaches in sepsis immunotherapy are focusing on augmenting the immune response rather than targeting a specific cytokine or chemokine. Some therapies that show promise include the use of granulocyte macrophage colony stimulating factor (GM-CSF) in patients who have entered the immunosuppression phase to prevent the incidence of nosocomial infections and restore pro-inflammatory cytokine production [56]. The use of IL-7 is also being explored in sepsis therapy. IL-7 is able to induce the proliferation of naïve and memory T-cells thus replenishing lymphocytes that are diminished during sepsis [56]. In addition, IL-7 can also increase the ability of T-cells to become activated, block the sepsis-induced apoptosis, improve the T-cell recruitment to the sight of infection and increase the T-cell receptor diversity [56].

Combination therapy of IL-7 with IL-15 is also being explored [56]. IFN-γ, a strong monocyte and macrophage activator, therapy has also shown success in small trials. IFN-γ therapy has been approved in fungal infections and has been evaluated in patients with persistent Staphylococcus sepsis infection with success [56].

Lastly, there have been several studies exploring the role of IL-17 in sepsis. Although IL-

17 was originally identified in activated T-cells, the release of this cytokine has been observed in many innate immune cells including neutrophils, macrophages, and invariant natural killer T- cells (iNKT, thought to serve as a branch between the innate and adaptive immune responses)

[81]. Studies in mice have demonstrated that neutralization or blockade of the IL-17A isoform was protective in that it reduced the levels of bacteremia, the levels of circulating pro- inflammatory cytokines/chemokines, and the levels of mortality were reduced [81]. Currently, development of the human monoclonal anti-IL-17 antibodies is underway with clinical trials anticipated [81]. Blockade of IL-17 may prevent the harmful downstream effects of the inflammatory response in sepsis, but it cannot prevent infection.

1.4 Clinical Identification of Sepsis

In order to identify sepsis, clinicians rely on a combination of clinical presentations, biological markers of inflammation, and blood culture identification of an infecting organism.

1.4.1 Physiological Identification of Sepsis

Clinical presentation of SIRs criteria is often suggestive of sepsis but lack specificity as up to 93% of patients in the ICU can present with up to two SIRs criteria during their stay [4].

One of the most widely used severity-of-disease classification systems is the Acute Physiology and Chronic Health Evaluation (APACHE) score. This system was developed by Knaus and colleagues in the late 1970s and was later modified to the APACHE II score in 1985 [82]. The scoring takes into account patients underlying chronic illnesses as well as the acute physiological trouble that has brought them into the ICU [82,83]. Each patient is assessed within 24 hours of

ICU or ED admission and given an integer score between 0-71. This score comes from 12 routine physiological measurements including age, rectal temperature, mean arterial pressure, arterial pH, heart rate, respiratory rate, serum sodium levels, serum potassium levels, serum creatinine levels, hematocrit, white blood cell count, and Glasgow coma score [82]. The score can be reassessed several times during a patient’s stay in ICU and changes in the score can indicate either improvement or clinical decline thereby allowing change of care as needed [83].

This score may indicate severity of illness when patients are admitted to ICU and is used in combination with other parameters to gauge treatment.

The sequential organ failure assessment (SOFA) score is used to quantify the degree of organ dysfunction in a patient. The score is based on six different organ systems; respiratory, hematologic, cardiovascular, coagulation, neurological, and renal [84]. For each system there is a score from 0 (normal) to 4 (severe failure) with the overall maximum score of 24. Mortality correlates well with SOFA scores where a score between 9-12 correlates to 25% mortality in the

ICU, 13-16 with 50% mortality in ICU, 17-20 with 75% mortality in ICU and a score above 20 correlates with 100% mortality in ICU [84,85].

1.4.1.1 Immunological Bio-Markers

Early on in sepsis, there is a characteristic up or down regulation of genes for at least 39 proinflammatory cytokines, chemokines and growth factors [61]. Many of these are used in combination as markers of sepsis including IL-6, C-reactive protein (CRP), and procalcitonin

[4,61]. IL-6 is induced by TNF-α and, due to its long half-life, it can be measured accurately in blood. High levels of IL-6 have been observed in septic shock [5]. However, IL-6 levels can also be elevated in non-infectious SIRs. As such, IL-6 levels may be used to predict severity and clinical outcomes but cannot exclusively indicate sepsis as the cause of inflammation [5]. CRP is an acute-phase protein made in hepatocytes and alveolar macrophages in response to cytokines such as IL-6. CRP regulates bacterial opsonisation, phagocytosis, and complement [5]. CRP production is rapid in response to inflammation but it is also short-lived in the blood. As such, timing is crucial to ensure proper measurement of CRP levels. Procalcitonin has been identified as a marker of severe bacterial infections and can be used to distinguish patients with sepsis from those with SIRs [86]. Procalcitonin levels rise significantly in response to bacterial infections

but not in non-infectious or viral inflammatory processes [86]. The Procalcitonin and Survival

Study assessed whether the use of procalcitonin improved the detection sepsis and improved therapy in critical care patients. Despite the initial prospects for procalcitonin use as an indicator of bacterial sepsis infections, the results indicated that when procalcitonin was used, patients ended up having a longer ICU admission and greater exposure to broad-spectrum antibiotics thereby compromising the renal function in these patients [87]. Although there is still controversy surrounding its use as a sole biomarker for sepsis, procalcitonin has been approved in conjunction with other clinical diagnostics for risk assessment of patients and monitoring antibiotic success [88].

Another potential bio-marker is the TREM-1 receptor which, when bound to a ligand, triggers the release of cytokines but it is not up regulated in non-infectious SIRs [5]. In a 2004 study, Bollaret and colleagues demonstrated that TREM-1 was more sensitive as well as specific over CRP and procalcitonin as a biomarker for sepsis infections [89]. More recently, soluble

TREM-1 showed better predictive value for poorer outcomes in ED patients when compared to

CRP and procalcitonin [90].

Overall, the majority of the cytokines, chemokines and growth factors up-regulated in sepsis are more indicative of inflammation and do not confirm that there is an infection present.

CRP and IL-6 have a role in revealing the extent of inflammation but are not definitive enough for clinical use. Procalcitonin and TREM-1 have potential for a sepsis biomarker yet the unfavourable outcomes in the procalcitonin trial have suggested care needs to be taken in using one marker for sepsis. In addition, none of these biomarkers provide any information on the infectious agent thereby limiting their use in directing sepsis therapy.

Overall, the clinical manifestations of sepsis are not definitive and there are no sepsis- specific biomarkers that can specifically identify the infecting organism. Thus, only a confirmed microbial infection remains the true definitive identification for sepsis.

1.4.2 Blood Culture Diagnostics

In order to definitively determine if a patient has sepsis, there must be proof of infection.

In addition, identification of the infectious agent is crucial for the proper administration of antimicrobial therapy. Blood cultures to detect bacteremia are considered the “gold standard” and play an important role in sepsis diagnostics alongside other clinical infection cultures.

The principal of blood culture is simple and requires very little manipulation of samples.

Usually, two sets of blood culture bottles are taken per adult patient and optimally 20-30ml of blood for each set. For paediatric patients the volume per bottle is usually 4ml.

The main manufacturer of blood-culture bottles in North America is Becton Dickson

(BD) diagnostics with their patented BACTECTM Plus culture media [91] and Bactec 9240 instrument. This media has non-ionic resins, that are not publically disclosed by BD Diagnostics, that absorb a range of antimicrobials while promoting the growth of fungal and bacterial organisms [91]. The bottles are incubated aerobically and anaerobically while they are continuously monitored for growth for up to 5 days in an automated instrument. Growth is detected by gas production that will trigger an alarm indicating positive results. At this stage laboratory personnel will identify the cultivated organism(s) using clinical and laboratory standards institute guidelines [92]. First, aliquots from the positive blood culture bottles are plated onto Trypticase soy agar supplemented with 5% sheep’s blood, chocolate agar, and

MacConkey agar. Gram-staining is also done at this time. Sub-culture is required in order to obtain enough inoculum to prepare suspensions for biochemical testing and for antimicrobial susceptibility testing [93]. These suspensions are then loaded into automated systems such as the

Vitek-2 system (bioMérieux) and the Phoenix system (BD Diagnostics) for rapid identification and antimicrobial susceptibility testing.

1.4.2.1 Advantages of Blood Culture

The principal advantage of blood culture is that it allows for recovery of viable organisms that can be evaluated for antimicrobial susceptibility [5]. This is a critical advantage of this system since improper antibiotic administration is a risk factor for mortality in patients with life- threatening infections [5]. Further, this system is automated and easily standardized providing ease of use and universal application.

1.4.2.2 Limitations of Blood Culture

Blood volume significantly affects the rate of microbial isolation from blood culture in both adult and paediatric patients [5]. In essence, the more blood that is used as an inoculum, the greater the rate of isolation from blood culture. In an adult patient, this may not be a significant concern. However, obtaining sufficient blood volume is problematic in paediatric patients, especially in neonates. Indeed, a 2007 study by Curtis and colleagues reported 2.1% of blood culture samples were positive and indicated that the major contributing factor this low positive rate was insufficient blood volume [94]. Another problem with blood culture is that the results are severely impacted by time delays from the time of inoculation to loading the instrument.

Efficiency decreased and the false-negative rate increases dramatically when bottles are held at room temperature for more than 12 hours [95]. Blood culture is also not effective for slow- growing and difficult to culture organisms including Mycoplasma species [96]. Furthermore, organisms considered non-cultivable by standard culture methods (i.e., Rickettsia species) are also under-represented by blood culture [96]. Lastly, blood-culture is severely limited by the turnaround time for a positive signal. Usually a positive-signal is present after 24-48hr of incubation [14]. However, it takes additional time and labour to identify the organism and determine antimicrobial susceptibility. Further, blood culture relies on traditional culturing techniques. With the emergence of molecular sequencing technologies and the knowledge of the microbiome, it has become increasingly apparent that traditional cultivation techniques can only recover a small fraction of the bacteria that are associated with disease [61,97-101]. As such, even when blood culture is positive, the results are not necessarily representative of the sepsis infection, rather they represent the organisms that are able to grow easily in blood culture media.

As part of a recent Canadian clinical trial, the clinical blood culture microbiological data was available for 778 patients admitted to ICU with septic shock at 27 centers in Canada, the

USA, and Australia [61]. The study reported that only 26% of the blood cultures drawn from patients identified an infecting organism [61]. The median delay between a patient meeting septic shock criteria and obtaining a preliminary microbiology report was 30 hours [61]. In this initial report, the organism was only identified at the Gram-stain level and required an additional

24 to 48 hours before an actual organism was identified and antimicrobial resistance was known

[61]. The downstream effects of the delays in blood culture and low sensitivity can be fatal since the inappropriate administration or delay of antibiotics can increase the risk of mortality by 8%

per hour [5,61]. These issues have resulted in many attempts to improve the diagnosis of bloodstream infections.

1.5 Molecular Approaches to Sepsis Diagnostics

Due to the limitations of blood culture and the need for rapid diagnostics in sepsis, there has been significant work on developing novel approaches to sepsis diagnostics. These are often broken down into the blood culture-dependent approaches and blood culture-independent approaches. The blood-culture dependent approaches are aimed at reducing the time to identify an organism, provide a more definitive identification, and identify clinically important antimicrobial resistant pathogens (i.e., MRSA and VRE). The blood culture-independent approaches are aimed at circumventing the use of blood culture, providing rapid identification of pathogens, and identifying markers of antibiotic resistance [102]. Figure 1.1 outlines the current technology development in sepsis diagnostics, which will be discussed below.

1.5.1 Proteomics in Sepsis Diagnostics

Proteomic technology is being implemented for microbial identification from blood culture bottles. Specifically, matrix-assisted laser desorption-ionization time-of-flight mass spectrometry (MALDI-TOF MS) has been used to identify bacteria or fungi by examining their proteomic profiles [5]. The advantages of using MALDI-TOF MS are the rapid turnaround time, low sample volume requirements, and reasonable reagent costs [103]. Essentially, this method uses the peptide mass to charge ratio (m/z) values from MS peaks as an indication of molecular masses present in the sample [103].

Figure 1.1: Molecular approaches that are currently being assessed for application in sepsis diagnostics. Adapted from [102].

This generates pathogenomic profile patterns that can be used to identify certain species or certain genes within a species including those coding for antibiotic resistance machinery or virulence factors [5]. In terms of bacterial identification, MALDI-TOF MS measures the highly abundant ribosomal proteins that would be present in all living cells [103]. By using abundant and essential proteins, the MS protein signatures are not greatly influenced by variability in environmental or growth-phase conditions [103]. The cost per identification is estimated at $0.50

US dollars however it can go up to $10.00 for isolates that require PCR and 16S rRNA identification to resolve any discrepancies [103]. In terms of accuracy, the MALDI-TOF MS system had high overall accuracy with high-confidence identification for 94.4% of isolates tested in one study [103].

One limitation of MALDI-MS is that it requires mass databases against which to compare experimental data. As such, the data is only as powerful as the database is and can only be compared against known masses. Indeed, a recent study revealed poor identification rates for isolates belonging to the Streptococcus genus (41%), Gram-negative anaerobes (17%), and the

Gram-positive anaerobes (57%) [103]. Further, for 69% of the isolates identified there was no result available using MALDI-TOF MS. Further development of databases would be needed to ensure coverage for all possible microbes. Currently, application of MALDI-TOF MS directly to biological samples is not yet possible [5]. This is due to the requirement of a high concentration of bacterial cells for the generation of mass-spectral profiles since it does not have the analytical sensitivity to detect less than 104 -105 CFU/ml [92]. In bloodstream infections, attempts to quantity microbial loads have been limited but are presumed to be low due to the high percentage

of failed blood cultures [51]. As such, it is unlikely this method will be applicable for whole blood sampling.

Recently, the use of electrospray ionization MS (ESI-MS) coupled with PCR has been proposed for microbial diagnostics from blood culture [92]. This method allows for genotypic characterization of bacteria directly from positive blood culture bottles using PCR amplification with broad-range primers that are specific for groups of microbes [92]. Essentially, this method involves the extraction of DNA from positive blood culture bottles, PCR amplification of various gene targets, and then analysis of the amplicon using ESI-MS to determine the (m/z) ratio [92].

This generates a genetic fingerprint for the sample that is then compared to a database of known

DNA compositions to generate a microbial identification. When compared to conventional blood culture, the PCR/ESI-MS was able to identify organisms that were not isolated in blood culture

[92]. In terms of microbial identification, there was 100% concordance at the Gram-stain level,

98.72% at the genus level, and 96.58% at the species level. PCR/ ESI-MS was also able to distinguish S. pneumoniae in a mixed infection with other streptococci [92]. One concern with this study was the high recovery rate of Saccharomyces cerevisiae in 32% of the samples. The study deemed this as contamination from the blood culture bottles, as PCR/ESI-MS from an un- inoculated blood culture bottle came back positive for the organism [92]. Overall, the PCR/ESI-

MS method shows potential as a diagnostic tool in sepsis as it combines the sensitivity of PCR with the rapid nature of MS [5]. This method may be promising in the future if instrumentation and reagent costs drop, automation is possible for interpretation of spectral data, and if DNA is obtained from whole blood rather than blood culture bottles.

1.5.2 Molecular Approaches using Blood Culture

Methods that use blood culture may be advantageous in order to decrease the amount of time between a positive result and identification of the growing pathogen. They may also be useful in cases of culture-negative results due to fastidious organisms that fail to grow in blood culture [5]. The use fluorescent in-situ hybridization (FISH) in microbial identification involves using probes that consist of a peptide nucleic acid (PNA) and a fluorophore. These PNA probes are mimics of DNA that have a peptide backbone instead of a sugar phosphate backbone. Based on the probe sequence, they are able to bind to species-specific targets on the 16S rRNA of living bacteria [104]. The commercially available AdvanDx PNA-FISH kits (AdvanDx) target the rRNA genes are available for limited number of bacteria including S. aureus or CoNS, E. faecalis or other selected enterococci, E. coli, and P. aeruginosa [5]. PNA-FISH can be done in a short time (approximately 3 hours) and is easy to perform [5,104]. Overall sensitivities and specificity for the PNA probes were high at 99% and 100%, respectively [5]. In comparison to conventional diagnostics, the use of PNA-FISH reduced the median time to final microbiology report from 3.7 days to 1.1 days and could identify up to 95% of the bacterial or fungal organisms found in blood culture [102]. Although the use of PNA-FISH is easy to implement, it is still costly and only a limited panel of probes are approved for clinical use [92]. Further, there are not many species level probes available limiting the detection to the genus level for many bacteria [102].

The use of 16S rRNA based microbial identification kits for positive blood cultures have been developed. The MicroSeq 500 kit from Applied Biosystems uses primers to sequence the first 500bp of the16S rRNA gene [105]. Sequence data is then analyzed using MicroSeq software

and the MicroSeq database consisting of full-length 16S rRNA sequences from 1400 different bacteria [105]. However, this method requires DNA from colonies growing on a plate resulting in at least one round of sub-culturing from positive blood culture and cannot be used for negative blood culture bottles.

Pyrosequencing has also been used to identify organisms from blood culture bottles without the need for sub-culturing and biochemical testing for microbial identification.

Pyrosequencing accurately classified 98.8% of organisms when compared to conventional methods for blood culture bottles positive for one organism [105]. The limitations were in generating sequence data for highly similar streptococci isolates and S. pyogenes isolates. With respect to mixed infections, pyrosequencing only identified 68% of the organisms present in positive blood culture bottles [105]. One limitation with direct sequencing from blood culture bottles was the presence of background sequences as determined by sequencing inoculated

BACTEC bottles [105]. Indeed, commercial blood culture medium is sterile but it is not free from microbial DNA that may be a result of bacterial growth in the media prior to sterilization

[106]. As such, in order to validate the pyrosequencing results, an evaluation of each lot of blood culture bottles would be needed before their use in clinical studies [105]. This would be labour intensive and likely limit the applicability of this method in a clinical diagnostic laboratory.

Multiplex PCR has also been employed on blood culture bottles. The Hyplex

BloodScreen assay combines multiplex PCR with a enzyme-linked immunoabsorbent assay

(ELISA)-like hybridization to detect several bacterial species [5]. This assay is available for detection of resistance markers as well including the vancomycin resistance van genes and methicillin-resistance mecA gene [5]. The turnaround time for this assay ranges between 3-4 hours and it has shown high sensitivity ranging from 96.6-100% for identifying different species

[5]. However, this assay only identifies a few Gram-positive and Gram-negative organisms, 10 in total, which limit its use in sepsis diagnostics.

Recently, microarray-based assays have also been used for identifying blood culture positive organisms. The Prove-it Sepsis (Mobidiag) kit was the first sepsis specific microarray assay developed [5]. The array has targets for more bacterial species, 60 in total, than the multiplex assay and also has probes for mecA resistance gene as well as 13 fungi [5]. The turnaround time is similar to the PNA or the Hyplex systems and the sensitivity is between 94-

95% [5]. This microarray system is promising as it has the potential to be applied directly to whole blood samples by making improvements to the DNA amplification step which would allow for lower starting titers of bacterial DNA [5].

1.5.3 Blood Culture Independent Molecular Approaches

The vast majority of assays being used for direct sampling from blood involve PCR- based technology and DNA extraction directly from blood. There are several considerations that must be taken into account when using molecular methods to ensure proper interpretation of the data. Once a sample is obtained, the DNA must be extracted and purified for subsequent use.

There are several commercially available kits for DNA extraction. Some popular extraction methods include the Qiagen QIAamp blood kit or the Qiagen tissue kit (Qiagen) [92,106-108], variations of the phenol-chloroform extraction [106,109], the EasyMAG bacterial DNA isolation kit (BioMereieux), the MolYsis complete kit (Molzym), the Looxster kit (SIRS-Lab), and the

SeptiFast kit (Roche diagnostics) [110]. Most of these kits involve some type of cell-wall disruption to allow DNA release from cells, followed by enzymatic lysis of proteins and RNA,

DNA separation using an enrichment column, then purification of the DNA, and final elution of

DNA into a sterile fluid. Variations of this general protocol have been employed by different companies in order to increase DNA yield. For example, some kits will separate host DNA from microbial DNA including the MolYsis kit and the Looxster kit. The MolYsis kit has a human

DNA depletion system where there is selective lysis of blood cells using chaotropic buffers

[110]. This is followed by degradation of the DNA using chemically resistant DNases [110]. The application of the MolYsis technology has been shown to increase the sensitivity of bacterial

DNA isolation from human blood by 100-fold [111]. The Looxster kit involves complete cell lysis using bead beating on whole blood, DNA extraction and purification, and then separation of human and pathogen DNA using specific binding of pathogen DNA to an enrichment column matrix [110]. Overall, the use of pre-analytical prokaryotic DNA enrichment can increase the ratio of bacterial to human DNA from 26% to 74% [111]. The EasyMAG bacterial DNA isolation system uses a NucliSens lysis tube that comes with the kit to lyse whole blood [110].

The lysed blood mixture is then mixed with magnetic silica beads and the rest of the extraction process is carried out in the automated EasyMAG device [110].

Currently, molecular approaches to microbial diagnostics in sepsis have employed either a pathogen-specific assay, or a broad-range assay [5]. Pathogen-specific assays have little overall benefit in sepsis diagnostics due to the large number of potential pathogens involved in bloodstream infections and the opportunity for polymicrobial bloodstream infections [5].

Broad-range assays have the greatest clinical applicability for bloodstream infection diagnostics as they do not rely on culture as well as their ability to directly detect any non- cultivable or cultivable pathogens [5]. PCR or quantitative real time PCR (RT-PCR) is done to amplify universal bacterial genes that can have variable regions allowing for genus and/or

species identification. Commercially available kits have been developed for molecular diagnostics of blood. The SepsiTest (Molzym) kit uses RT-PCR amplification of the 16S rRNA genes in bacteria and fungi followed by sequence analysis of amplicons [112]. This method also employs a selective degradation of human DNA, using the MolYsis technology, thereby improving the lower end detection limit of the assay [5]. SepsiTest is able to detect over 300 different pathogens and has a detection limit of 20-40 CFU/ml for S. aureus [5,108]. The reported turnaround time for this method is 8-12 hours; 4 hours for DNA extraction and PCR plus 4-8 hours for sequence analysis [5,112]. A benefit of this system is the PCR procedure used will detect only viable microbes thereby eliminating any concerns about whether or not PCR signals correspond to living pathogens [112]. Further, the SepsiTest showed no PCR inhibition or false-positive results when evaluated by Kuhn and colleagues [112]. In comparison to blood culture, the diagnostic sensitivity and specificity were reported to be 87% and 85.8%, respectively [113]. The SepsiTest kit is promising as it has been shown to overcome some of the concerns associated with molecular diagnostics; limited performance due to PCR inhibitors in blood and reagent-borne contamination [112]. There is potential for this to be used a rapid diagnostic if the sequencing time can be minimized.

The VYOO system uses a multiplex PCR to detect 35 common sepsis bacterial pathogens and 6 fungal pathogens [5]. It can also detect genes associated with antibiotic resistance such as the mecA, vanA, vanB, vanC, and blaSHV [5]. Similar to the SepsiTest kit, the VYOO method also employs a selective human DNA extraction method using the Looxster affinity chromatography method [110,111]. The PCR amplified products are run on an agarose gel and the resulting banding pattern is used for pathogen identification [5].The VYOO method does not require any post-amplification sequence analysis which helps minimize the risk of any contamination

associated with post-amplification processing. As well, the lower limit of detection for the

VYOO assay is reported, by the company, as 3-10 CFU/ml [5]. Evaluation of the VYOO kit has been limited, but one study using the VYOO kit to study ICU sepsis cases reported detection of a pathogen in 36% of suspected sepsis cases [114]. This was an improvement over blood culture that only detected 12% of bloodstream infections and in terms of fungemia, there was an 11% detection rate whereas blood culture failed to detect any fungal bloodstream infections [114].

The light-cycler SepsiFast (LC-SF) is another multiplex-PCR based assay that targets 25 of the most prevalent bacterial and fungal pathogens associated with sepsis [115]. This method consists of three parts: mechanical lysis and purification of DNA from whole blood, RT-PCR amplification using broad-range primers in three parallel reactions (one for Gram-positives, one for Gram-negatives, and one for fungal organisms), detection with specific probes, and automated identification of species [115]. The turnover time from blood draw to results reported to clinicians ranged from 7-15 hours [115]. Clinical studies using LC-SF have shown mixed results with the overall sensitivity and specificities remaining low. A study from France compared LC-SF to blood culture and reported the sensitivity of LC-SF to be 78% and the specificity as 99% [115]. A Danish study reported that 52% of all causative pathogens were identified using LC-SF in comparison to 37% identified using blood culture [116]. Another study reported better identification from blood culture when compared to LC-SF but the authors gave no explanation for the poor results [117]. Overall, the LC-SF system is limited in its clinical applicability due to the relatively small panel of organisms included in the multiplex-PCR, the high signal-to-noise ratio issue, and relatively high costs ranging from $200-300 per test [5].

Commercially available molecular diagnostic kits are still costly, lack automation, and many are not approved for use in North America [103]. Many of the methods proposed also

involve very high skill sets to ensure proper use of equipment, such as mass spectrometers, or proper analysis of results for microarrays, multiplex PCR, and MS data. PCR based methods are also challenging due to the difficulty in isolating DNA from biological samples that consist of natural inhibitors and abundant eukaryotic DNA [103]. As such, extensive protocols are needed to eliminate PCR inhibitors, reduce the eukaryotic DNA load and also lyse prokaryotic cells in order to obtain DNA [103]. Lastly, there is no gold standard for PCR based methods in terms of

DNA extraction, PCR amplification targets, and analysis targets [96]. This makes comparison between studies difficult and leads to issues in reproducing results. As such, new strategies need to be explored to address the current limitations in molecular diagnostics.

1.5.4 Application of Next-Generation Sequencing to Sepsis

In recent years there has been a strong focus on studying microbe-human interactions in health and disease. The Human Microbiome Project (HMP) was developed in 2007 as an extension of the human genome project [97]. Since its initiation, the HMP has resulted in development of numerous reference genomes, helped develop next-generation sequencing strategies, and expanded ribosomal sequence databases. Recent studies have shifted the previous understanding that a single pathogen causes infectious disease to the concept of polymicrobial communities involved in disease [100,101,118-122]. Currently, sepsis is most often considered as a single microbial infection with rare cases of polymicrobial sepsis [22,29]. However, these results are based on blood culture confirmed infections, which currently represents a minority of sepsis cases. As such, the use of next-generation sequencing technology and cultivation- independent approaches may increase the detection of polymicrobial infections in sepsis.

1.6 Rationale for the Study

Although blood cultures can play an important role for examining susceptibility patterns in cultivated organisms, it lacks the sensitivity necessary to provide consistent results, with successful diagnosis of microbiologically confirmed bloodstream infections occurring 20-30% of the time in ICU patients with nosocomial bloodstream infections or sepsis [61,123]. Moreover, inherent limitations to the processes of traditional culture cannot be overcome by technical modifications [5,61]. As a result, there is a poor understanding of microbial infections in sepsis and it is imperative that newer methods be explored to provide more accurate diagnostics for sepsis bloodstream infections.

1.7 Hypothesis and Objectives

Sepsis is the result of an inflammatory process in response to microorganisms and/or their products present in normally sterile sites such as blood. Current culture based diagnostics fail to identify the organisms in the majority of bloodstream infections, leading to an underrepresentation of the diversity present. I hypothesize that developing methods for the direct detection of bacteria from blood samples, using culture and/or molecular approaches, will improve the detection of bacterial infections in the bloodstream as well as increase the detection of polymicrobial infections. This study focused on the development of methods to isolate microbes from blood without using blood culture, the application of next-generation sequencing to identify polymicrobial bacterial DNA in blood, and to correlate the bacterial DNA

profiles identified in blood with clinical data to identify common microbial patterns in sepsis.

The specific objectives are outlined below.

1.7.1 Research Objectives

1. Develop novel methodologies to isolate microbes and microbial DNA from blood

without conventional blood culture cultivation.

2. Apply these methodologies to clinical samples obtained from adult and paediatric

sepsis patients admitted to the intensive care unit and the emergency

department.

3. Compare the bloodstream bacterial DNA profiles to the primary infection site

bacterial DNA profiles, when a clearly defined primary infection was present, to

examine the correlation between them.

4. Compare the results obtained from the novel methodologies to readily available

clinical data and diagnostic laboratory microbiology results.

Chapter Two: Materials and Methods

2.1 Bacterial Strains and Culture Conditions

2.1.1 Bacterial Strains and Plate Pools

The bacterial strains and bacterial plate pools used in this study are listed in Table 2.1.

Frozen cultures of the strains were maintained at -80°C prior to use. The frozen stocks were prepared from heavy growth of bacteria cultivated on solid media that were transferred into

1.5ml of sterile 10% skim milk (BD Difco, Mississauga, ON) using a sterile cotton swab. For plate pools, the entire plate of bacterial growth was mixed with 1.5ml of 10% (w/v) skim milk using a sterile inoculating loop and the entire mixture was frozen.

2.1.2 Culture Conditions

2.1.2.1 Liquid Media

Colonies of bacteria were suspended in brain-heart infusion (BHI) broth (BD Difco,

Mississauga, ON) or phosphate-buffered saline (PBS). BHI was prepared by dissolving 37g of

BHI powder in one litre of deionized water. PBS was prepared using PBS tablets (Sigma-

Aldrich, Oakville, ON) with each tablet dissolved in 200ml of deionized water. All liquid media was autoclaved prior to use to ensure sterility.

Table 2.1: Bacterial Strains and Plate pools used in this study.

Bacterial Strains recovered from ASN087 BAL Fluid 16S Identification (HOMD) T-RF Size (bp) CFU/ml 6 Neisseria flava 209 10 4 Neisseria flavescens 370 10 4 Capnocytophaga sp 86,584 10 4 Streptococcus anginosus 584 10 5 Streptococcus constellatus 584 10 3 Staphylococcus epidermidis 234 10 2 Capnocytophaga sp 86,584 10 3 Streptococcus mitis 574 10 6 Bifidobacteriaceae [G-1] sp 202 10 4 Neisseria elongata 370 10 5 Prevotella melaninogenica 99 10 4 Lachnospiraceae [G-1] sp 108 10 Fusobacterium necrophorum 194 5 10 Plate Pools ASN087-CBA ASN087-TSY ASN087-FAA Strains used in Synthetic Community Organism Source Growth Condition Enterobacter hormaechei Peritoneal Fluid 5% CO2 Escherichia coli Urine 5% CO2 Fusobacterium necrophorum BAL Anaerobic

Micrococcus luteus Whole blood 5% CO2 Neisseria flava BAL 5% CO2 Prevotella melaninogenica BAL Anaerobic Staphylococcus aureus Whole blood 5% CO2 Staphylococcus epidermidis Whole blood 5% CO2 Streptococcus intermedius Abscess fluid 5% CO2 Streptococcus pneumoniae Chest tube aspirate 5% CO2

2.1.2.2 Solid Media

Solid media used in this study included Colombia blood agar (CBA), Colombia CNA agar (CNA), Mannitol Salt agar (MSA), MacConkey Agar (MAC), Trypticase Soy agar (TSA) supplemented with 3% yeast extract (TSY), BHI agar supplemented with colistin sulphate

(10µg/ml) and oxolinic acid (5µg/ml) (BHIco), all from BD Difco, Mississauga, ON. Fastidious anaerobic agar (FAA, Neogen Acumedia, Lansing, MI), Chocolate agar (Dalynn Biologics,

Calgary, AB), and McKay agar [124] were also used. Preparation of CBA, CNA, MSA, MAC and FAA was done following the manufacturer’s guidelines. For TSY, TSA was used following manufacturers guidelines and 3g of yeast extract (BD Difco, Mississauga, ON) was added for every 1L of media prepared. Media was autoclaved and cooled to 50°C prior to pouring 20ml per sterile 100 x 20 mm petri plate (VWR, Edmonton, AB). Difibrinated sheep’s blood (Dalynn

Biologics, Canada) was added to a 5% final volume for CBA, CNA, and FAA after the media was autoclaved and cooled to 50°C. For BHIco, the colistin was prepared in deionized water and the oxolinic acid was prepared in 0.1M sodium hydroxide. Both additives were sterilized using a

0.2µM syringe filter (VWR, Edmonton, AB) and added after the media was autoclaved and cooled to 50°C. Chocolate agar plates were purchased from Dalynn Biologics, Calgary, AB.

McKay agar was made following previously published guidelines [109,124]. Solid media was allowed to cool at room temperature for 24 hours prior to long-term storage at 4°C. The role of each media type is outlined in Table 2.2.

Table 2.2:Solid-media used for this study.

Media Name Abbreviation Selective Ingredients Distinguishable Properties Colombia Blood CBA 5% defibrinated Distinguish between α, β, and γ Agar1 sheep’s hemolysis Colombia CNA CNA 5% defibrinated Distinguish between α, β, and γ Agar1 sheep’s blood hemolysis

Colistin and nalidixic Inhibit growth of Gram-negative acid bacteria Mannitol Salt Agar1 MSA 7.5% sodium chloride Inhibits most bacteria except staphylococci Mannitol and phenol Separates mannitol fermenting S. red aureus from other staphylococci MacConkey Agar1 MAC Crystal violet and bile Inhibit growth of Gram-positive salts bacteria Distinguish between lactose and non-lactose fermenters Tryptic Soy Yeast TSY 3% Yeast Extract Enhance bacterial growth Agar2 Brain Heart BHIco Colistin and oxolinic Supress the growth of Gram- Infusion acid negative bacteria with colistin and oxolinic acid Agar3 Fastidious FAA Sodium Oxygen scavengers Anaerobe Agar4 bicarbonate/pyruvate L-Cysteine and L- Growth stimulant for anaerobes Arginine Haemin and Vitamin Growth stimulant for Bacteroides K species Sodium succinate Growth stimulant for Prevotella species Chocolate Agar1 Choc 2% Haemoglobin Enables growth of Haemophilus species IsoVitaleX™ Enhances growth of Neisseria Enrichment species McKay Agar2,5 McKay Colistin and oxolinic Supress growth of Gram- acid negative bacteria Sulfadiazine Select Streptococcus milleri group bacteria over other streptococci 1[125] 2[109] 3[100] 4[126] 5[124]

2.2 Patient Cohorts and Ethics

2.2.1 Ethical Approval

Samples collected for this study were done under the umbrella of the Critical Care

Epidemiologic and Biologic Tissue Bank Resource (CCEPTR) for Infection and Inflammation in the ICU. Approval for CCEPTR was granted by the Conjoint Health Research Ethics Board of the University of Calgary with the Ethics ID for the study E-22236 on April 7, 2009. Human samples were obtained in concordance with the guidelines of the Tri-Council policy statement.

Written informed consent was collected from all subjects or their surrogate decision maker. The original ethics approval included patients admitted to the intensive care unit (ICU) of Calgary and Edmonton hospitals. Amendments to the original submission were made to include patients in the emergency department (ED) at the Foothills Medical Center in Calgary (Alberta, Canada) who met specific criteria for eligibility for the study. Paediatric samples were obtained from patients enrolled in the ED at the Alberta Children’s Hospital in Calgary (Alberta, Canada).

In addition to blood, patients were required to indicate if other samples could be collected through CCEPTR. These samples included urine, expectorated lung secretions (sputum), secretions from a breathing tube, catheter or device initially inserted for diagnosis, monitoring or treatment, biopsy samples that were sent for lab analysis (with Calgary Lab Services approval), and the fluid collected during dialysis. The CCEPTR ethics protocol included the collection of samples from healthy controls. Whole blood was collected from healthy adults who consented to be a part of the study. A request for fresh whole blood and tissue samples was submitted to the

CCEPTR coordinators at the initiation of the study. All unused blood and tissue samples were

discarded as per CCEPTR guidelines. No genetic testing was performed on the biological samples.

2.2.2 Enrolment Criteria

2.2.2.1 Adult ICU Septic Patients

Patients 18 years of age or older admitted to the ICU of the Foothills Medical Center who met the published criteria for SIRS and clinical suspicion or confirmation of infection within the first 24 hours of ICU admission or within the first 24 hours of a newly acquired infection were eligible for the study [36,127]. SIRS criteria include; body temperature > 38°C or < 36°C, heart rate > 90/min, evidence of hyperventilation by respiratory rate > 20/min or PaCO2 < 32 mmHg, and white blood cell count > 12,000 cells/μl [127]. Exclusion criteria included patients in which clinical care was withdrawn.

2.2.2.2 Non-Septic ICU Patients

Briefly, ICU controls were considered as patients with 2 or less SIRS criteria and not suspected of having an infection. These patients were admitted to the ICU for various reasons including: post-elective surgery that did not involve a mucous membrane including spinal surgery, cranial surgery, and coronary artery bypass graft surgery; major trauma excluding head injuries with a musculoskeletal injury, intraabdominal injury or pulmonary contusion; brain injury including traumatic brain injury, intracerebral haemorrhage, subarachnoid haemorrhage, or ischemic stroke. Patients with suppressed immune systems, including those with steroid or

biological prescriptions within the previous 4 weeks, haematological malignancies, pancreatitis within the last 24 hours, or burns covering more than 10% of the body surface were excluded from the control group.

2.2.2.3 Adult ED Patients

For ED patients the enrolment criteria included patients over 18 years of age, enrolment within the first 24 hours of admission to the ED, 2 or more SIRS criteria and clinical suspicion or confirmation of infection.

2.2.2.4 Paediatric ED Patients

Patients enrolled at the Alberta Children’s Hospital, Calgary, Alberta were included in the study if they met the following criteria; under the age of 18, greater than two SIRS criteria present, clinical suspicion or confirmation of infection, and antibiotic treatment ordered for the suspected or confirmed infection.

2.2.2.5 Adult Healthy Control Individuals

Control individuals were adults above 30 years of age who worked in a health care setting but at the time collected had no major autoimmune disorders, no symptoms of illness, were not on anti-inflammatories for the prior 7 days, and were otherwise healthy (no colds/fever/chills/etc.) in the prior 7 days with a normal respiratory rate and temperature measured as supportive data.

2.2.3 Patient Admissions and Clinical Diagnostic Data

Clinical diagnostic laboratory data was collected during patient enrolment in the study.

Data was considered relevant to the sample if collected within a 24 hour period prior to or after enrolment in the study or within 24 hours of the sample being collected for subsequent days in the ICU. Admissions data was obtained from the Alberta Sepsis Network (ASN) secure server.

Clinical data included the APACHEII score [82] and the sepsis-related organ failure assessment

(SOFA) score [84]. Each patient was identified only by a unique identifier based on the site in which the sample was obtained, the type of sample collected, and the order in which the patients were enrolled (Table 2.3). Clinical laboratory results were collected from Calgary Lab Services and included all diagnostic cultures done that were relevant to the patient’s clinical presentation as well as all pharmacy related data for therapy administered.

2.3 Collection of Human Blood and Tissue Samples

Under the CCEPTR guidelines, trained and licensed nurses or phlebotomists collected whole blood and biological samples.

2.3.1 Collection of Human Blood

No more than 30ml of blood was collected from a given patient on a given day. Based on the CCEPTR requirements for blood, a maximum of 4ml of blood and 2ml of blood were collected from adult and paediatric patients, respectively.

Table 2.3:Patient cohorts used in the study.

Cohort Abbreviation Type of Sample Number of Samples

Paediatric patients enrolled in the CB Whole Blood 28 ED at Alberta Children’s Hospital

Patients enrolled in the ED at EB Whole Blood 52 Foothills Medical Center

Healthy adults enrolled at HB Whole Blood 12 Foothills Medical Centre

Adult ICU control patients IC Whole Blood 5 enrolled at Foothills Medical Centre

Adult ICU septic patients SB Whole Blood 175 enrolled at Foothills Medical Centre

Adult ICU septic patients SI Primary infection 23 enrolled at Foothills Medical Centre

Adult ICU neurological trauma SN Whole Blood 15 patients enrolled at Foothills Medical Centre

Whole blood was collected once on days 1, 7, 14, and 28 for patients enrolled in the ICU.

Patients enrolled in the ED had blood collected once on the day of ED admission. When possible, blood was collected from an existing arterial line, central line, or venous line in sterile

K2EDTA spray coated vacutainers (BD Diagnostics, Mississauga, ON). If an existing line could not be used then a licensed nurse or phlebotomist (through Calgary Laboratory Services) did a peripheral blood draw directly into the vacutainers.

2.3.2 Collection of Biologic Tissue and Fluids from ICU patients

Additional consent to use samples other than blood was required prior to collection.

Other biological samples were only collected if they had been taken as part of the patient’s routine care. Biologic samples included bronchoalveolar lavage (BAL) fluid, endotracheal tube

(ETT) fluid, chest tube (CT) fluid, peritoneal drainage (PF) fluid, abscess drainage fluid (AF),

Jackson-Pratt (JP) drainage fluid (from surgical wounds or infected tissue), and urine (UR) collected in sterile containers.

2.4 Processing of Clinical Samples

2.4.1 Saponin Treatment of Blood

Whole blood was treated with 0.85% saponin (Sigma-Aldrich, Oakville, ON) final volume at room temperature for 15 minutes to lyse red and white blood cells. Following treatment, the blood was centrifuged at 20,800 rcf for 15 minutes to remove lysed cells in the supernatant. The

supernatant was removed and the remaining cells were washed 1-3x with 1ml sterile

DNase/RNase free double distilled water (Life Technologies, Burlington, ON, Burlington, ON).

After each wash the cells were spun at 20,800 rcf for 15 minutes and the supernatant was removed. Cells were suspended in 500µl sterile PBS for storage prior to DNA extraction or in

500µl BHI broth for cultivation.

2.4.2 Quantitative Culture

Biological fluid samples were first sheared to an even consistency using a 1ml tuberculin slip-tip syringe (BD Diagnostics, Mississauga, ON). Ten-fold serial dilutions were prepared as needed in BHI media. The dilutions were plated using 100µl of sample per solid media type. To ensure equal distribution across each agar plate, sterile 3mm borosilicate glass balls (VWR,

Edmonton, AB) were used with 10-15 balls/plate added and the plates were agitated to help distribute the liquid. The glass balls were aseptically removed prior to incubation of the solid media. The plates were incubated aerobically at 37°C with 5% CO2 for 48-72 hours. For anaerobic cultivation, plates were incubated in a Ruskinn Concept 400 (Ruskinn, Bridgend, UK) anaerobic chamber for 72-96 hours. Recovered colonies were identified morphologically and similar colonies from each plate were counted to estimate the colony forming unit (CFU)/ml of each isolate. One to two representative colonies of each morphotype were sub-cultured onto the same solid media type they were initially recovered from. In order to ensure the isolate growth was pure, the sub-culturing was done two subsequent times for a total of 3 re-purifications. Once growth was deemed pure, the resulting growth was collected as outlined in section 2.1.1. In

addition, a portion of the colony growth was used for partial 16S rRNA PCR as outlined in

Section 2.5.

2.5 Partial 16S rRNA gene PCR and Sequencing

2.5.1 Colony Preparation

In order to identify recovered colonies, one-two colonies from the final re-purification step were mixed with sterile 10% chelex (Bio-Rad, Mississauga, ON) in a 1.5ml boil-proof tube

(Axygen, Tewksbury, MA). The tubes were boiled in a water bath set on a hot plate for 15 minutes to enable cell lysis. The cellular debris was then pelleted by centrifugation at 20,800 rcf for 10 minutes with an Eppendorf 5430R bench-top centrifuge (Eppendorf, Mississauga, ON).

Samples were kept on ice to ensure the pellet stayed consistent prior to use.

2.5.2 PCR Mix and Cycle Conditions

The universal 16S rRNA DNA primers 8F (5'AGAGTTTGATCCTGGCTCAG3') and 926

R (5'CCGTCAATTCCTTTRAGTTT3') were used. Primers were synthesized at the University of Calgary, Core DNA Services laboratory (University of Calgary). Primers were reconstituted in

DNase/RNase free deionized water (Life Technologies, Burlington, ON) to a concentration of

100pmol/µl and then further diluted to 10pmol/µl prior to use. Thin-walled 0.2ml PCR tubes

(Axygen, Tewksbury, MA) were used and reactions were performed in a Mastercycler Gradient thermal cycler (Eppendorf, Mississauga, ON). Recombinant Taq DNA polymerase (Life

Technologies, Burlington, ON) was used. Each PCR reaction consisted of 1X PCR Buffer (Life

Technologies, Burlington, ON), 1.5mM MgCl2 (Life Technologies, Burlington, ON), 0.8mM dNTPs (Life Technologies, Burlington, ON), 1pmol/µl of each forward and reverse primer, 1U of Taq polymerase, 3-5µl of supernatant from the chelex mixture, and sterile DNase/RNase free deionized water to a total volume of 50µl. PCR cycle conditions were initial denaturation at 94˚C for 1 minute, 32 cycles of denaturation at 94˚C for 1min, annealing for 1 minute at 57˚C and extension at 72˚C for 1min followed with a final extension at 72˚C for 10 minutes.

2.5.3 Agarose Gel Electrophoresis

PCR products were separated and visualized using agarose gel electrophoresis. Ultrapure agarose (Life Technologies, Burlington, ON) was used at a 2% concentration. Samples were mixed with a 6X DNA loading dye (0.25% bromophenol blue and 40% sucrose in deionized water) with 5µl of sample used per reaction. The gels were submerged in 1X E-buffer in a Gel

Electrophoresis unit (Bio-Rad, Mississauga, ON) and samples mixed with 6X DNA loading dye were added to the wells. DNA was separated at 110V and a 1Kb+ DNA ladder (Life

Technologies, Burlington, ON) was added to the gel to ensure the product size was accurate

(900bp). Gels were stained in 1% ethidium-bromide for 15-30 minutes, rinsed in water, and visualized under UV transillumination using a ChemiDoc™ XRS+ System (Bio-Rad,

Mississauga, ON).

2.5.4 PCR Product Sequencing

PCR products were purified and sequenced unidirectionally from the 8F primer at the

Beckman Coulter Genomics (Danvers, USA) facility. The sequences were trimmed using DNA

Sequence Scanner Software v1.0 (Applied Biosystems, Burlington, ON) and then aligned to curated ribosomal sequence databases including HOMD (human oral microbiome database, www.homd.org), and Greengenes (www.greengenes.lbl.gov/) to determine the taxonomic identification of the bacteria.

2.6 DNA Extraction from Clinical Samples

2.6.1 In-house DNA Extraction Method

To each 500µl sample of saponin treated blood (Section 2.4.1) or sheared infection fluid sample (Section 2.4.1), 50µl of Lysozyme (100mg/ml, Sigma-Aldrich, Oakville, ON), 20µl of

Mutanolysin (10U/µl, Sigma-Aldrich, Oakville, ON), and 20µl of RNase A (10mg/ml, Life

Technologies, Burlington, ON) were added and incubated overnight at 37°C. Following this,

50µl of 25% sodium dodecyl sulphate (SDS, Sigma-Aldrich, Oakville, ON), 50µl of 20mg/ml proteinase K (Invitrogen, Life Technologies, Burlington, ON), and 100µl 5M NaCl were added.

The mixture was incubated at 65°C for 1-2 hours. Cellular debris was pelleted by centrifugation at 20,800 rcf for 10 minutes. The supernatant was then treated with one standard phenol- chloroform-isoamyl (25:24:1, Life Technologies, Burlington, ON) alcohol extraction. The DNA in the aqueous layer was transferred to a Zymo DNA Clean & Concentrator™-25 (Zymo

Research, Irvine, CA) column containing 200µl of ChIP DNA Binding Buffer (Zymo Research,

Irvine, CA). The column was spun for 1 minute at 20,800 rcf and the flow-through was discarded. Wash buffer was added at 500µl twice to the column with a 1 minute, 20,800 rcf centrifugation and discard of flow-through in between each wash. A final 1 minute, 20,800 rcf

centrifugation step was done to ensure the column was completely dry and free of any ethanol carry-over from the wash buffer. Pre-warmed DNase/RNase free deionized and UV irradiated water was used to elute the DNA with 50µl added per column. DNA was quantified using a

Nanodrop 2000c Spectrophotometer.

2.6.2 Validation of DNA Extraction Protocol

In order to validate the DNA extraction protocol in section 2.6.1, whole blood from a healthy donor was collected in K2EDTA vacutainers and spiked with a mixed bacterial community recovered on CBA/TSY/FAA from a septic pneumonia BAL fluid sample from an

ICU patient (Table 2.1). This plate pool represented bacteria recovered at or greater than 103

CFU/ml. The species richness and diversity made this an ideal sample for evaluation of the DNA extraction method. All bacteria recovered on the plate pools were stored as outlined in Section

2.1.1.

2.6.2.1 Comparison of Extraction With or Without Lytic Enzymes

(100mg/ml), or 50µl of Lysozyme (100mg/ml) and 20µl of Mutanolysin (10U/µl). The remainder of the protocol followed Section 2.6.1.

2.6.2.2 Comparison of Dynabeads to Zymo DNA Clean & Concentrator-25

Aliquots of 500µl of blood were spiked with 50µl of the bacterial plate pool and subjected to a 0.85% saponin treatment as described in section 2.4.1. DNA was extracted from the resulting blood pellet as outlined in section 2.6.1 up to the end of the phenol-chloroform- isoamyl alcohol extraction step. At this stage, two methods to purify and concentrate the DNA were assessed; the DNA binding column (Zymo DNA Clean & Concentrator™-25, Zymo) system and the DNA binding magnetic beads (Dynabeads DNA DIRECT, Life Technologies,

Burlington, ON) method. Zymo columns were used as outlined in section 2.6.1. For Dynabeads, the aqueous layer from the phenol-chloroform-isoamyl alcohol extraction was transferred to a sterile 1.5ml tube and 200µl of resuspended Dynabeads were added. The mixture was allowed to sit for 5 minutes at room temperature. The tubes were then transferred to the Dynal MPC magnet

(Life Technologies, Burlington, ON) and the beads were separated from the lysate for 2 minutes.

The lysate was pipetted out and the tubes were removed from the magnet. The Dynabeads complex was washed twice with 200µl of 1X wash buffer. The supernatant was removed each time by placing the tubes in the Dynal MPC magnet and pipetting the supernatant off. The

Dynabeads complex was then eluted in 40ul of resuspension buffer. The DNA was then eluted from the Dynabeads by heating the Dynabeads complex for 5 minutes at 65°C, then transferring the tube to the Dynal MPC magnet and pipetting the supernatant, containing the DNA, into a sterile tube.

2.7 Terminal-Restriction Fragment Length Polymorphism

2.7.1 PCR Reaction and Cycle Conditions

TRFLP was done following published protocols [109]. PCR amplification was done using the 8F (5'AGAGTTTGATCCTGGCTCAG3') and 926R (5'CCGTCAATTCCTTTRAGTT

T3') primers. The 8F primer was fluorescently tagged with 6-FAM at the 5’ end (Applied

Biosystems, Burlington, ON). PCR conditions were the same as those outlined in section 2.5.2 except the DNA used was from a DNA extraction as outlined in Section 2.6.1. For these reactions, 20ng of DNA was used. PCR was done with in triplicate and the products were pooled prior to digestion.

2.7.2 PCR Purification

PCR products were purified using the Zymo DNA Clean & Concentrator™-5 kit (Zymo

Research, Irvine, CA) following manufacturer’s guidelines. Briefly, two volumes of the DNA

Binding Buffer were added to the PCR mixture. The whole mixture was then transferred to the

Zymo-Spin™ Column in a collection tube. The mixture was centrifuged for 1 minute at 20,800 rcf and the flow-through was discarded. The Wash Buffer was then added at a 200µl volume twice with a 1 minute, 20,800 rcf centrifugation step following each addition. The column was transferred to a sterile 1.5ml tube and 20µl of sterile DNase/RNase free deionized water was added to the column then spun for 1 minute at 20,800 rcf. The final eluent was kept and stored at

-20°C until use.

2.7.3 CfoI Digestion and Purification

The purified PCR products were digested overnight with 20 units of cfoI (10U, Sigma-

Aldrich, Oakville, ON). In each 20µl reaction there was 2µl of 10x Restriction Endonuclease

Buffer SL (Sigma-Aldrich, Oakville, ON), 2µl of cfoI, 50-200ng of PCR product, and

DNase/RNase free UV sterilized deionized water. The reaction was vortexed briefly then spun down for 6 seconds to ensure all the reaction was at the bottom of the tube. The incubation was done overnight in a 37°C water bath. Following digestion, the products were purified as outlined in section 2.7.2 and eluted into 10µl DNase/RNase free UV sterilized deionized water.

2.7.4 Fragment Analysis

Fragment analysis was done at the UCDNA core facility (University of Calgary).

Approximately 5ng of purified digested product was injected into an ABI 3730 Genetic Analyzer

(Applied Biosystems, Burlington, ON) with the LIZ1200 size standard (Applied Biosystems,

Burlington, ON). Analysis was done using GeneMapper 4.0 software (Applied Biosystems,

Burlington, ON) and the percent of total peak area for each fragment size was calculated [109].

A minimum of 50 relative fluorescent units was set as a threshold and fragments of 50 base pairs or less as well as fragments above 800 base pairs were excluded from the analysis.

2.7.5 T-RF Determination for Clinical Isolates

In order to predict which T-RFs corresponded to clinically relevant bacteria, purified isolates of bacteria recovered from in-depth cultivation and taxonomically identified using partial

16S PCR (Section 2.5) were analyzed using TRFLP. T-RFs were assigned based on the fragment analysis of the 16S rRNA amplified from these purified colonies [109]. A threshold of 200 relative fluorescent units was set as a minimum to ensure partial-digested products were not included. Samples were tested in duplicate to ensure the T-RF size was consistent. Samples with multiple T-RFs were examined in triplicate to ensure the presence of more than one cut site was valid.

2.8 Fluorescent Activation Cell Sorting (FACS)

FACS flow cytommetry was performed on a whole blood samples treated with saponin as outlined in section 2. 4.1. Briefly, 1µl of FM® 1-43 (Life Technologies, Burlington, ON) dye and 1µl of SYTO 9 (0.34 mM) dye (Life Technologies, Burlington, ON) were incubated with

250 µl the blood pellet resuspended in PBS for 5 minutes at room temperature. FACS sorting was done on the BD FACSAriaII (BD Diagnostics, Mississauga, ON) at the Nicole Perkins

Microbial Communities Core Labs in the Snyder Institute for Chronic Diseases, Calgary,

Canada. Sorted cells were eluted into PBS and stored at -20°C until use.

2.9 Paired-end Illumina 16S rRNA Sequencing

2.9.1 PCR Primers and Conditions

Partial 16S rRNA gene PCR amplification was done following previously published guidelines [128,129] with some modifications. Briefly, 4 pmol of each primer, 200µl of each dNTP, 1.5mM of Mg2Cl2, and 1 U of recombinant Taq polymerase (Life Technologies,

Burlington, ON) were used in a 50l reaction. To maximize PCR amplification, the 50l reaction was split into 3 reactions of equal volume. PCR amplification of the V3 region was done using the primers 341F (5’CCTACGGGAGGCAGCAG3’) and 518R (5’ATTACCGCGGCTGC

TGG3’). Primer modifications included the addition of Illumina multiplexing, bridge amplification and sequencing regions [128]. Reverse primers were bar-coded to allow multiple processing of samples. For whole blood samples, 200-300ng of DNA was PCR amplified whereas 50-100ng of DNA was used for infection samples. The PCR cycle consisted of an initial denaturation at 94˚C for 2 minutes followed by 30 cycles of denaturation at 94˚C for 30 seconds, annealing at 50˚C for 30 seconds, extension at 72˚C for 30 seconds followed by a final extension at 72˚C for 10 minutes. The entire PCR reaction was run on a 2% agarose gel. PCR products were excised from the gel and purified using the QIAquick gel extraction kit (Qiagen, Toronto,

ON) following manufacturer’s guidelines with the optional isopropanol wash step included.

2.9.2 Sample Preparation

Purified samples were sent to the McMaster DNA Sequencing Facility (McMaster

University, ON) and prepared as described by Whelan et al., 2014 [129]. The samples were first tested on an Agilent BioAnalyzer High Sensitivity DNA chip (Agilent Technologies,

Mississauga, ON) then quantified using RT-PCR with the Illumina PhiX control library as a standard (Illumina, San Diego, CA). RT-PCR was done using the

SYBR fast 2x qPCR mastermix (KAPABiosystems, Wilmington, MA) and the primers P5 5’ AA

TGATACGGCGACCACCGA-3’ and P7 5’CAAGCAGAAGACGGCATACGA-3’ which bind to the flowcell regions of the adaptor sequences.

2.9.3 Illumina Sequencing

Following the RT-PCR quantification, the 16S rRNA gene V3 region pools were loaded into the Illumina cassette with the PhiX control DNA in a 9:1 ratio. Sequencing was done on the

Illumina MiSeq instrument (Illumina, San Diego, CA). The sequencing was done in the forward and reverse direction with 250 base pair read lengths. The completed run was demultiplexed with

Illumina’s Casava software (version 1.8.2).

2.9.4 Sequence Processing and Taxonomic Identification

The sequencing data was processed in the Surette lab with custom, in-house Perl scripts

[129]. The sequences were trimmed using Cutadapt [130] and paired-end sequences were aligned using PANDAseq [131]. AbundantOTU+ [132] was used to pick operational taxonomic units

(OTUs) at a clustering threshold of 97% sequence similarity. The taxonomic identification was assigned using the Ribosomal Database Project classifier [133] using the Greengenes reference database, February 4th 2011 release [134] as a training set with QIIME defaults (0.8 minimum confidence cut-off). Tables with the taxonomic identification of each OTU and the relative DNA abundance were derived from this data.

2.9.5 Analysis

2.9.5.1 Taxonomic Identification and Filtering

Taxonomic summaries and subsequent analysis were done using QIIME version 1.7.0

[135]. Following the removal of the “noRoot” OTU, representing non-bacterial and likely human 58

DNA amplification, and singletons, the samples with fewer than 500 reads were also removed from the final analysis. Healthy controls and reagent controls were also used to determine which

OTUs represented DNA contamination. In addition, Burkholderia OTUs were removed due to the presence of confirmed Burkholderia contamination in the water bath used for incubation at

37°C during the DNA extraction protocol. The OTUs identified and removed from the final OTU table used for analysis are listed in Table 2.4. The representative sequence for each OTU was also aligned to 16S rRNA sequences using the HOMD database (www.homd.org) and to the

National Center for Biotechnology-Basic Local Alignment Search Tool (NCBI-BLAST, http://blast.ncbi.nlm.nih.gov/Blast.cgi) megablast nucleotide search tool. Sequences that aligned with greater than 97% sequence identity were considered for a higher order prediction of the sequences represented by the OTU.

2.9.5.2 Diversity Measures

β-diversity was used as a measure to examine differences between samples. Both weighted and un-weighted UniFrac distance and clustering of the samples was done and visualised using principal coordinate analysis (PCoA) [136,137]. PCoA plots were visualized using KiNG version

2.21 visualization software [138]. Composite unweighted pair group method with arithmetic mean (UPGMA) phylogeny was used to generate phenetic trees based on hierarchal clustering.

Table 2.4: OTUs identified as contamination and removed from OTU table.

Taxonomic Level OTU Taxonomic Identification Genus Burkholderia Genus Bradyrhizobium Genus Cupravidus Genus Variovorax Genus Aquabacterium Genus Sphingobium Genus Rhodobacter Genus Sphingomonas Genus Rhodanobacter Genus Thermus Genus Meiothermus Genus Janthinobacterium Genus Methylobacterium Genus Mesorhizobium Genus Acidocella Genus Chryseobacterium Genus Acidovorax Genus Delftia Genus Novosphingobium Genus Akkermansia Genus Paucibacter Order Rhizobiales Order Rhodobacterales Order Rhodospirillales Order Sphingomonadales Order Burkholderiales Order Oceanospirillales

In order to examine how robust the UPGMA and PCoA clustering was, jackknife β- diversity was done on evenly re-sampled OTU tables using both unweighted and weighted

UniFrac distance [136]. α-diversity measures were used to examine the diversity within a given sample. Several diversity measures were available through the QIIME pipeline but Shannon- index and observed species were used. Shannon index was used to measure the richness within each sample whereas observed species indicated total OTU number in a sample [139]. In order to normalize the α-diversity across samples with varying numbers of reads and ensure evenness, rarefaction was done at varying sequence depth and α-diversity was examined in the rarefied

OTU tables [139].

2.10 Mock Community Analysis

Ten bacterial strains recovered from sepsis blood or source infections (Table 2.1) were used to generate a mixed community. Colonies from each sample were resuspended in sterile M9 salts + 0.1% cysteine to a McFarland 2 unit. Viability in saponin was assessed by incubating 10-1 through to 10-8 dilutions of each isolate with 0.85% saponin for 1 hour at room temperature. CFU were determined prior to and after treatment. Communities of 100µl of each strain diluted at a

10-1, 10-3, or 10-5 dilution were mixed together resulting in the three communities 104 – 107

CFU/ml (SC1), 102 – 105 CFU/ml (SC3), and 101 – 103 CFU/ml (SC5) of each strain. Human whole blood was collected from healthy adults who consented to have their blood used for these experiments in K2EDTA vacutainers and 20µl of each synthetic community was added to 500µl of whole blood. Saponin was added at a 0.85% final volume and incubated for 30 minutes at room temperature. Three treatments with saponin were performed; addition of 0.85% saponin

with no further washes, addition of saponin at 0.85% with the addition of a 1ml sterile double distilled water wash, or addition of saponin at 0.85% with two washes with 1ml sterile double distilled water.

All recovered cells were resuspended in BHI and plated in duplicate on MSA, CNA, CBA,

MAC, TSY, and FAA using 100µl per plate prior to and after addition to blood. In addition,

10µl of each strain alone were diluted and spiked into 250µl of 0.85% saponin treated whole blood, incubated for 30 minutes at room temperature and plated on their respective media type

(Table 2.1). DNA was also extracted from all communities for TRFLP and paired-end 16S rRNA

Illumina profiling.

Chapter Three: Method Development and Optimization

3.1 Introduction

There is a need for more rapid and comprehensive diagnostics for bloodstream infections in the management of sepsis. In a recent Canadian study, the reported overall positivity of clinical diagnostic blood culture was only 26% [140]. Further, the time required for bacterial growth and identification is between 24-72 hours [5]. This results in a large time delay between collection of the sample and identification of an etiological agent. Due to their short turnaround time and ability to directly detect any non-cultivable or cultivable bacteria, molecular diagnostics using broad-range assays and primers targeting variable regions in the 16S rRNA or 18S/23S rRNA gene have the greatest clinical applicability for sepsis diagnostics [5,92,96].

Most sepsis infections are considered to be a result of a Gram-positive, Gram-negative, or anaerobic organism [22], and reports of sepsis patients with polymicrobial infections is often low. However, a recent Calgary study investigating anaerobic bloodstream infections reported

26% of assessed diagnostic blood-culture samples were polymicrobial [29]. Similarly, a population-based study on bloodstream infections in ICU patients done 10 years prior reported

12% of positive diagnostic blood cultures were polymicrobial [13]. Despite this, many of the molecular methods utilized in the diagnosis of sepsis focus on detecting 1 to 2 infecting organisms, and the concept of a microbiota as the cause for sepsis remains relatively unexplored.

In order to better understand bloodstream infections, a method was developed to better recover microbial cells from blood specimens and to optimize DNA extraction for molecular profiling.

This involved using the molecular community profiling methods of TRFLP and paired-end

Illumina 16S rRNA sequencing on mock bacterial communities spiked into whole blood in order to assess this methods ability to identify polymicrobial infections.

3.2 Method Development

3.2.1 Blood Collection and Anticoagulant Selection

The first step in developing this methodology was to determine how whole blood should be collected from patients. Currently, in clinical settings, blood collection tubes (vacutainers) containing anti-coagulants such as heparin, sodium citrate, sodium polyanetholesulfonate (SPS), acid citrate dextrose (ACD), and ethylene-diamine-tetra acetic acid (EDTA). Most molecular studies employing whole blood or plasma have used EDTA with success [96,107,141,142] and as such, it was decided that K2EDTA spray coated tubes (BD Vacutainer®, BD Diagnostics,

Mississauga, ON) would be used for studies outlined in this thesis. Other anti-coagulants have been shown to be bactericidal [143,144], impair the PCR reaction [106,142,145,146], or reduce

PCR efficiency [142].

3.2.2 Saponin Blood Treatment

3.2.2.1 Addition of Saponin Improved Microbial Recovery

To test whether saponin pre-treatment could improve microbial recovery, we used the compound on a whole blood sample collected fresh from an ICU patient, ASN-KS, who tested positive for Staphylococcus epidermidis in blood cultures collected by Calgary Laboratory

Services one-day prior. Each millilitre of whole blood was treated with saponin as outlined in 64

Section 2.4.1 and subsequently cultured as outlined in Section 2.4.2. In parallel, 100µl of untreated whole blood was plated directly on MSA and CBA. Following incubation, no growth was present for the untreated whole blood on either MSA or CBA, while the saponin treated blood had growth present on all media types. Partial 16S rRNA sequencing indicated that the colonies were all S. epidermidis, which matched the diagnostic laboratory results for this patient.

In another preliminary experiment, the whole blood sample from ASN087 was treated with saponin prior to DNA extraction. In parallel, DNA from the BAL sample was also extracted. Both samples were analyzed using TRFLP. The TRFLP profiles indicated polymicrobial DNA was present in the whole blood and BAL samples. Further, T-RFs were common to both samples suggesting that the airway microbiota contributed to the bacteremia in this patient (Figure 3.1). These results suggested that the saponin treatment could be applied to whole blood samples to improve recovery of viable bacteria (ASN-KS) as well as permit molecular profiling (ASN087) and was therefore systematically evaluated for this study.

3.2.2.2 Saponin Did Not Impede Microbial Growth

In order to ensure that saponin did not impact microbial viability, bacterial strains recovered from sepsis infections (Table 2.1) were co-incubated with 0.85% saponin as outlined in Section 2.10. The CFU/ml after the incubation with saponin showed no significant changes in microbial growth when compared to CFU/ml values immediately prior to the addition of saponin

(Figure 3.2). Only F. necrophorum had a 2-log drop in CFU/ml following treatment with saponin

(Figure 3.2). However the p-value was insignificant (p=0.2375, two-tailed t-test).

Figure 3.1: Preliminary experiment on ASN087 samples indicated the addition of saponin to whole blood improved recovery of microbial DNA. There were 14 distinct T-RFs in the

BAL sample and 16 in the blood. T-RFs in red were seen in the sputum and in the blood suggesting they translocated into the blood from the lungs. Based on TRFLP of pure isolates cultured from the BAL of this patient, the T-RFs 209 & 370, 235 and 584 were predicated to represent Neisseria sp., Staphylococcus sp., and Streptococcus milleri group (SMG), respectively

(Table 2.1).

Figure 3.2: Saponin did not impact the viability of bacteria. The average CFU/ml and the standard deviation were plotted for each sample prior to and after the incubation with 0.85% saponin for 1 hour at room temperature either with ambient air or in an anaerobic chamber for F. necrophorum and P. melaninogenica. Each experiment was done in triplicate with samples recorded in duplicate. There were no statistically significant differences found in the viable cell count of the bacteria (two-tailed Students t-test, p > 0.05% significance).

To determine whether bacterial cellular integrity was maintained after saponin treatment, bacterial cells from the ASN087 BAL fluid bacterial community recovered on TSY and FAA

(Table 2.1) were spiked into whole blood and treated with saponin. The resulting pellet was then labeled with the lipophilic fluorescent dye FM® 1-43 and Syto-9, a membrane permeable nucleic acid dye. The FM® 1-43 dye was used to detect intact bacterial membranes [147] and the Syto-9 was used to detect double-stranded DNA within the cells [148]. Double-stained cells were identified and enumerated using a two-color fluorescence coincidence and gating recovery based on size to exclude cells outside of the microbial cell range. There were 22 T-RFs recovered in the

ASN087 TSY culture pool (Figure 3.3A). Following treatment with saponin, there were 4 T-RFs that did not stain positive for both FM1-43 and Syto-9 (Figure 3.3A). Furthermore, our data identified 3 T-RFs that were not recovered from TRFLP done on the DNA from the ASN087

TSY culture pool (Figure 3.3A). There were 27 T-RFs identified in DNA from the ASN 087

FAA culture pool (Figure 3.3B). Of these, 11 were also present in the FACS population as well as 9 T-RFs not present in the ASN087 FAA culture pool (Figure 3.3B). Based on these data, we approximated a 77% recovery of intact and viable bacterial species from the TSY culture pool and a 41% recovery from the FAA culture pool.

3.2.3 Enzymatic Lysis and Column Purification

Since the causative agent(s) in bloodstream infections can be a range of Gram-positive or

Gram-Negative bacteria, proper and accurate identification of the infecting pathogens requires a robust DNA extraction method applicable to any pathogen. Currently, there is no “gold- standard” for DNA extraction from clinical samples much less blood.

Figure 3.3: FACS sorting of whole blood spiked with ASN087 culture pools and treated with 0.85% saponin indicated that microbial cells were still intact. The DNA extracted from

SYTO-9 and FM1-43 double-labeled cells recovered (blue bars) was compared to DNA extracted from culture pools (black bars) indicated that cells had intact plasma membranes containing DNA. For the TSY culture pool, 22 T-RFs were identified in the culture pool prior to treatment (A). Of these, 17 were recovered after FACS. For the FAA culture pool, 27 T-RFs were identified prior to treatment and 10 were recovered following FACS (B).

A 100.00%

r 10.00%

T 1.00%

0.10% 56 57 60 76 77 81 86 89 90 105 110 116 122 130 144 160 166 167 174 203 208 209 234 370 583 B

100.00%

e r 10.00%

1.00%

0.10%

2 9 1 7 7 8 1 6 9 6 7 9 5 0 2 6 8 5 7 9 6 7 9 3 6 3 9 1 0 6 9

4 4 4 4

5 5 6 6 7 7 8 8 8 9 9 9 0 1 1 1 1 2 3 3 8 9 0 1 4 6 6 7 8 8 8

5 7 6 8

1 1 1 1

1 1 1 1 1 1 1 2 2 2 5 5 5 5 5 5

1 3 5 Size (bp)

Since commercially available kits have not been sufficiently validated for molecular studies, our research group developed our own comprehensive DNA extraction protocol that was evaluated for use in this study and has been successfully applied to recent molecular profiling studies

[100,149]. The first step in evaluating the DNA extraction method was to examine if the enzymatic digestion step was warranted. TRFLP analysis of the DNA recovered from the

ASN087 CBA plate pools following treatment without lysozyme or mutanolysin had 7 T-RF peaks recovered (Figure 3.4). The addition of lysozyme increased the recovery to 10 T-RFs

(Figure 3.4). Following digestion with lysozyme and mutanolysin, there was recovery of several lower abundant peaks and a total of 24 T-RFs recovered (Figure 3.4). Based on this data, the

DNA extraction method developed for use with whole blood samples included a cell-wall digestion step including both lysozyme and mutanolysin.

3.2.4 Purification and Concentration of DNA

In order to ensure that the DNA extracted from a clinical sample could be used for downstream PCR, blood spike experiments were completed to evaluate two independent DNA purification kits (Section 2.6.2.2). When the two methods were compared using the sample

ASN087 CBA culture pool spiked into whole blood, there were more T-RFs recovered using the

Zymo columns (24) compared to the Dynabeads (19; Figure 3.5). When comparing both methods, the most highly abundant T-RFs were recovered using both methods (Figure 3.5). Out of the 19 T-RFs recovered using Dynabeads, 16 were also recovered with Zymo columns (Figure

3.5). Furthermore, the T-RFs recovered with Zymo columns alone were mainly low abundant T-

RFs representing less than 1% of the total peak area (Figure 3.5).

Figure 3.4: Enzymatic digestion requirement for bacterial DNA extraction of whole blood.

Whole blood collected in a K2EDTA spray-coated vacutainers was spiked with a CBA plate pool from ASN087 BAL sample. For each assessment, 500µl of blood were spiked with 50µl of the culture pool and subjected to a 0.85% saponin treatment as described in section 2.4.1. DNA was extracted from the resulting blood pellet in one of three ways; with no enzymatic lysis (black bars), 50µl of lysozyme (100mg/ml) (gray bars), or 50µl of lysozyme (100mg/ml) and 20µl of mutanolysin (10U/µl) (blue bars). The remainder of the DNA extraction followed the procedure outline in section 2.6.1. TRFLP was performed on the purified DNA as outlined in section 2.7 and the total percent peak area for each T-RF identified was plotted. More T-RFs were present when enzymatic digestion was done with lysozyme (10, grey) and lysozyme plus mutanolysin

(24, blue) identified, respectively, when compared to no enzymatic treatment with only 7 T-RFs present. The greatest number of T-RFs recovered was when the sample was digested with lysozyme and mutanolysin (blue) indicating the use of both cell-wall digesting enzymes was preferable.

100.00%

No Enzyme

Lysozyme

Mutanlosyin+Lysozyme

10.00%

1.00%

0.10% 52 54 60 79 80 87 89 116 118 130 160 166 167 174 189 203 206 207 209 210 213 231 233 234 241 364 370 384 584 T-RF Size (bp)

Figure 3.5: Zymo column based purification recovered more T-RFs when compared to

Dyna magnetic bead based purification. TRFLP of DNA extracted from blood spiked with

ASN087 BAL culture pools purified and concentrated with DNA binding column (Zymo DNA

Clean & Concentrator™-25, Zymo Research, Irvine, CA) system and the DNA binding magnetic beads (Dynabeads DNA DIRECT, Life Technologies, Burlington, ON) method following manufacturers guidelines. The T-RFs recovered using Dynabeads are shown in blue whereas those recovered using Zymo columns are in black. There were 19 T-RFs recovered using the

Dynabeads whereas there were 24 recovered using Zymo column based purification.

100.00%

10.00%

a Zymo

Dyna

1.00%

0.10% 52 54 60 65 79 80 87 89 101 116 118 130 160 166 167 174 189 203 209 213 218 234 241 364 370 384 584 T-RF Size (bp)

Based on our data, the use of DNA columns improved the ability to recover DNA from less abundant organisms in a polymicrobial DNA sample (Figure 3.5).

3.2.5 Saponin Treatment of Blood Spiked with Mock Bacterial Communities

In order to assess if the saponin method could recover total microbial abundance, a mock community of organisms was spiked into whole blood. The rate of recovery as well as the limits of detection were systematically assessed. In addition, the DNA recovered from the mock community following treatment was used for paired-end Illumina 16S rRNA sequencing. This allowed for evaluation of the method cultivation-dependent and cultivation-independent approaches. In order to determine the limits of detection, the synthetic communities consisted of bacteria at three concentration ranges: 104 – 107 CFU/ml (SC1), 102 – 105 CFU/ml (SC3), and

101 – 103 CFU/ml (SC5). A panel of Gram-positive, Gram-negative, and anaerobic bacteria recovered from sepsis infections were used (Table 2.1). Each step in the saponin treatment protocol was evaluated to see if there was a loss of diversity during the processing of blood

(Section 2.10).

3.2.5.1.1 Limit of Detection

As per the saponin blood-treatment protocol, the CFU/ml of each organism was experimentally determined by plating dilutions of the synthetic communities prior to and after the addition to whole blood. In SC1, 104 – 105 CFU/ml was added for all isolates. As determined by counting colonies recovered on agar plates, we were able to recover >105 CFU/ml for 6 of the

10 organisms and between 104 – 105 CFU/ml for both E. coli and S. pneumoniae (Figure 3.6A).

The anaerobic bacteria F. necrophorum could not be recovered from SC1 after any of the treatments whereas P. melaninogenica was detectable up until the final wash step (Figure 3.6A).

For the SC3 sample in blood, the CFU/ml ranged from 102 – 103 prior to treatment and above 103

CFU/ml recovered after the saponin treatment (Figure 3.6B). Following the first wash, bacteria were recovered at or above 103 CFU/ml (Figure 3.6B). N. flava could not be identified after one wash whereas E. coli was not recovered after one wash but was identified following the second wash (Figure 3.6B). After the wash steps, F. necrophorum and P. melaninogenica could not be recovered (Figure 3.6B). The CFU/ml of SC5 in blood prior to the addition of saponin ranged from 1 to 15 (Figure 3.6C). After the saponin treatment, 7 organisms were recovered (Figure

3.6C). This dropped to 6 organisms after the first wash while 7 were detected after two washes

(Figure 3.6C). For E. hormaechei, a one-log increase in growth to over 103 CFU/ml was recovered after both washes (Figure 3.6C). Similar increases in growth were seen for S. aureus,

S. epidermidis, and S. intermedius (Figure 3.6C) suggesting an enrichment of these organisms in the saponin-treated blood pellet. Both anaerobes failed to be detected after any treatment.

3.2.5.1.2 Percent Recovery

The addition of saponin alone resulted in an 80% recovery rate for 9 out of the 10 organisms in SC1 and SC3 (Figure 3.7A). However only 4 out of 10 organisms achieved this level in SC5 (Figure 3.7A). Saponin treatment followed by a single wash had highly variable results with 80% recovery achieved for 8 of the 10 organisms in SC1, 6 out of 10 organisms for

SC3 and 1 out 10 organisms for SC5 (Figure 3.7B).

Figure 3.6: Limit of detection for synthetic communities of bacteria spiked into whole blood. The CFU/ml of each bacterium in the community was determined prior to blood spiking

(solid black bars). The CFU/ml of bacteria recovered after each step in the saponin-blood treatment protocol was determined; addition of 0.85% saponin with no further washes (solid grey bars), addition of saponin at 0.85% with the addition of a 1ml sterile double distilled water wash

(diagonal lined bars), or addition of saponin at 0.85% with two washes with 1ml sterile double distilled water (solid white bars). The limit of detection at each step in the process was evaluated as the lowest CFU/ml added in which the organism could be subsequently recovered from SC1

(A), SC3 (B) and SC5 (C). All recovered cells were resuspended in BHI and plated in duplicate on MSA, CNA, MAC, TSY, and FAA using 100µl per plate prior to and after addition to blood.

The mean CFU/ml of each organism from two independent experiments is plotted.

Figure 3.7: The percent recovery of bacteria from a mixed community spiked into whole blood. Whole blood was inoculated with each community and saponin was added at a 0.85% final volume and incubated for 30 minutes at room temperature. The CFU/ml of bacteria recovered after each step in the saponin-blood treatment protocol was determined; addition of

0.85% saponin with no further washes (A), addition of saponin at 0.85% with the addition of a

1ml sterile double distilled water wash (B), or addition of saponin at 0.85% with two washes with 1ml sterile double distilled water (C). The percent recovery was determined based on the percentage of the initial inoculum that was recovered after each treatment. The effect of concentration on the percent recovery was explored by comparing the recovery of bacteria from

SC1 (black bas), SC3 (grey bars), and SC5 (diagonal striped bars). All recovered cells were resuspended in BHI and plated in duplicate on MSA, CNA, MAC, TSY, and FAA using 100µl per plate prior to and after addition to blood. The mean CFU/ml of each organism from two independent experiments was used for the calculations. Percentages were determined by dividing the recovered CFU/ml over the initial CFU/ml.

After two washes, 5 of the 10 organisms could be recovered from all SCs regardless of concentration. One organism, M. luteus, exhibited a significant decrease in recovery from the

SC1-3 samples to the SC5 sample (Figure 3.7C). A concentration dependent loss of diversity for

M. luteus, N. flava and S. epidermidis was observed (Figure 3.7A). By the end of the saponin blood-treatment protocol there was an enrichment of 5 of the aerobic bacteria and a complete loss of anaerobic bacteria in the community (Figure 3.7C). Overall, our data suggests that the inclusion of saponin alone could result in the recovery of all organisms in the microbial community.

The recovery of each bacterium from the synthetic community was assessed independently to ensure the mock community results were accurate. Recovery of each organism was >80% using saponin alone with the notable exception of N. flava (Figure 3.8). Anaerobic organisms, F. necrophorum and P. melaninogenica, were recovered under all treatment conditions when cultivated alone and under anaerobic conditions (Figure 3.8).

3.2.5.2 Summary of Culture-Dependent Results

A summary of the results for the culture-dependent assays is shown in Table 3.1. Overall, the majority of the isolates (8 out of 10) were recovered after two washes in all three SCs (Table

3.1). Anaerobic organisms were difficult to recover from mixed community samples and were not detected after two washes (Table 3.1). Since all washes were completed under atmospheric oxygen conditions it was not surprising that these organisms were not cultivated after this treatment.

Figure 3.8: Recovery of individual isolates spiked into whole blood. Each organism was diluted and 10µl was spiked into 250µl of 0.85% saponin treated whole blood, incubated for 30 minutes at room temperature and plated on their respective media type (Table 2.1). The CFU/ml of each bacterium was determined prior to blood spiking (black bars). The CFU/ml of bacteria recovered after each step in the saponin-blood treatment protocol was determined; addition of

0.85% saponin with no further washes (grey bars), addition of saponin at 0.85% with the addition of a 1ml sterile double distilled water wash (white bars), or addition of saponin at 0.85% with two washes with 1ml sterile double distilled water (diagonal lines). All recovered cells were resuspended in BHI and plated in duplicate on MSA, CNA, MAC, TSY, or FAA using 100µl per plate. The data plotted represents pooled data from 3 independent experiments with the mean

CFU/ml and standard deviation plotted.

Table 3.1: Percent recovery and limits of detection for synthetic community bacteria.

Percent Recovery (%) Limit of Detection (CFU/ml) Saponin Alone Saponin + 1 Wash Saponin + 2 Washes Saponin Saponin Saponin+ Alone +1 Wash 2 Washes Organism SC1 SC3 SC5 SC1 SC3 SC5 SC1 SC3 SC5 E. coli 82.8 100 73.7 24.8 0 0 32.3 51.7 15.5 102 104 103 E. hormaechei 100 100 100 100 100 100 100 100 100 10 10 10 N. flava 100 100 23.4 100 100 75.6 100 100 37.8 105 104 104 M. luteus 100 68.7 8.0 100 6.9 0 20.1 0 0 102 10 10 S. epidermidis 100 100 100 100 100 39.6 100 100 100 10 10 10 S. aureus 100 100 39.3 100 100 31.7 100 100 100 10 10 10 S. intermedius 100 100 100 100 100 19.9 100 100 100 10 10 10 S. pneumoniae 100 100 0 96.9 100 0 100 100 100 102 102 10 F. necrophorum* 0 100 0 0 0 0 0 0 0 102 103 104 P. melaninogenica* 100 100 0 100 0 0 0 0 0 102 102 102 *Based on recovery alone (Figure 3.8) not in SC

When each of these organisms were spiked into blood alone we were able to recover viable cells, however, there was a sequential reduction in the CFU/ml recovered as the organisms were exposed to oxygen during the centrifugations and hypotonic washes (Figure 3.8).

Surprisingly, E. coli and N. flava had the poorest recovery in the mixed communities and, compared to the other organisms, higher inoculums were required to recover these organisms

(Figure 3.8). The limit of detection ranged from 101 CFU/ml to 104 CFU/ml. For 6 of the 10 bacteria, the lowest limit of detection was obtained after the full saponin blood-treatment protocol (Table 3.1). In these experiments, E. hormaechei, S. aureus and S. epidermidis were over-represented in the recovered communities whereas E. coli and N. flava were under- represented. Since all these organisms grew in the presence of saponin (Figure 3.2) these results suggested that there were differing abilities to grow in mixed communities and in blood.

3.2.5.3 Culture-Independent Results

Bacterial DNA sequences from each community prior to treatment and after blood spiking experiments was assessed. Unmapped reads resulting from primer cross-reactivity to human DNA [150,151] were removed. These reads were represented taxonomically as “noRoot” as they did not correlate to bacterial DNA in the ribosomal database and had alignments corresponding to human chromosomal DNA. The proportion of these “noRoot” signals varied in each sample (Figure 3.9). As the bacterial concentration decreased from SC1 to SC5, the proportion of “noRoot” DNA increased (Figure 3.9).

Figure 3.9: Illumina 16S rRNA sequencing of DNA from Synthetic Communities. Paired- end Illumina 16S rRNA sequencing was done on DNA recovered from each sample. Taxonomic summaries for the synthetic community samples after each step in the saponin blood-treatment protocol were compared. Each bar represents the total DNA sequenced for the sample and the relative abundance of “noRoot” DNA (grey), other OTUs (diagonal line) which represent OTUs that could not be correlated to the bacteria present in the synthetic community, and the SC organisms (black). The table indicated the relative proportion of each class of DNA in the sample. The data represented the taxonomic profiles from two independent experiments with each sample PCR amplified in triplicate.

100% 90%

e 80%

n a 70%

b 60%

A 50%

N Other OTUs

e 40%

v "noRoot" DNA

a l 30% SC Organisms

e R 20% 10% 0% SC1 SC1 SC1 SC1 SC3 SC3 SC3 SC3 SC5 SC5 SC5 SC5 Sap 1Wash Full Sap 1Wash Full Sap 1Wash Full Treatment

The lowest proportion of “noRoot” DNA was seen in the samples treated with saponin alone as compared to those when saponin and hypotonic washes were implemented (Figure 3.9). This suggested the hypotonic washes increased the abundance of human DNA. Despite this, the molecular profiles indicated that the saponin-only treated communities had poor representation of the synthetic community organisms as compared to those with the full saponin blood- treatment across all three SC dilutions (Table 3.2).

The taxonomic summaries, identified to the genus level when possible, resulted in a total of 203 OTUs identified. As this was a gross over-estimation of the diversity, OTUs were filtered to remove OTUs representing less than 0.1% relative DNA abundance. This value was chosen as it still enabled the recovery of DNA from all mock community organisms and represented the majority of bacterial DNA in each sample while filtering out low abundance DNA. This resulted in 35 OTUs identified that represented over 99% of the relative DNA abundance for all treatments, except in SC5 where only 97.9% of the DNA sequenced was represented (Figure

3.10A). OTUs corresponding to the synthetic community organisms included

Enterobacteriaceae representing E. hormaechei, Staphylococcus representing S. aureus and S. epidermidis, Escherichia representing E. coli, Streptococcus representing S. pneumoniae and S. intermedius, Neisseria representing N. flava, Actinomycetales representing M. luteus

Fusobacterium representing F. necrophorum, and Prevotella representing P. melaninogenica

(Table 3.2). In order to confirm that the OTUs recovered were from the synthetic community organisms, the representative sequence from the top 20 OTUs present in all the SC samples were aligned to the HOMD ribosomal sequence database and to NCBI as per Section 2.9.5.1 (Table

3.2.). Each bacterium in the synthetic community was present in the top 20 OTUs with the exception of S. epidermidis, which was the 53rd most abundant OTU (Table 3.3).

Table 3.2: Taxonomic identification and relative OTU abundance of synthetic community organisms.

OTU Taxonomic ID SC1 SC1 SC1 SC1 SC3 SC3 SC3 SC3 SC5 SC5 SC5 SC5 Sap 1Wash Full Sap 1Wash Full Sap 1Wash Full

Enterobacteriaceae 32.22% 43.58% 54.37% 54.78% 36.15% 29.76% 39.65% 37.26% 24.60% 37.40% 1.94% 53.01% (E. hormaechei) Staphylococcus (S. aureus, S. 35.59% 26.11% 9.76% 9.40% 40.03% 42.14% 19.45% 16.69% 13.03% 36.95% 94.93% 6.54% epidermidis) Escherichia 10.82% 0.01% 0.33% 0.26% 0.90% 5.36% 7.56% 13.30% 19.67% 0.03% 0.70% 1.00% (E. coli) Streptococcus (S. pneumoniae, S. 0.30% 2.98% 0.17% 0.24% 0.00% 2.56% 0.80% 2.08% 2.72% 1.27% 0.16% 0.19% intermedius) Neisseria 0.13% 0.00% 0.06% 0.10% 0.00% 0.24% 1.20% 0.36% 1.83% 0.00% 0.03% 0.01% (N. flava) Actinomycetales 0.01% 0.01% 0.02% 0.12% 0.00% 0.55% 0.28% 1.09% 0.95% 0.00% 0.12% 0.07% (M. luteus) Fusobacterium 0.12% 0.00% 0.05% 0.13% 0.00% 0.29% 0.75% 0.40% 0.40% 0.00% 0.02% 0.00% (F. necrophorum) Prevotella 0.25% 0.00% 0.01% 0.14% 0.00% 0.02% 0.33% 0.44% 0.31% 0.01% 0.01% 0.02% (P. melaninogenica) Other OTUs 20.56% 27.32% 35.23% 34.84% 22.90% 19.08% 29.98% 28.39% 36.47% 24.34% 2.09% 39.15%

Figure 3.10: OTU abundance of 16S rRNA Illumina sequenced DNA from synthetic communities. Taxonomic summaries for the synthetic community samples after each step in the saponin blood-treatment protocol were compared. Each bar represents the total DNA sequenced for the sample and the relative abundance of each OTU in the molecular profile. All OTUs that could not be correlated to the bacteria present in the synthetic community were combined into

“Other OTUs” (A), which accounted for 20-40% of the total OTU abundance. These OTUS were also removed from the taxonomic summaries to resolve the OTU abundance for each synthetic community organism (B). Each bar represents the combined results from two separate experiments with each PCR sample amplified in triplicate.

A 100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0% SC1 SC1 SC1 SC1 SC3 SC3 SC3 SC3 SC5 SC5 SC5 SC5 Sap 1Wash Full Sap 1Wash Full Sap 1Wash Full

Enterobacteriaceae Staphylococcus Escherichia Streptococcus Neisseria Actinomycetales Fusobacterium Prevotella Other OTUs

B 100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0% SC1 SC1 SC1 SC1 SC3 SC3 SC3 SC3 SC5 SC5 SC5 SC5 Sap 1Wash Full Sap 1Wash Full Sap 1Wash Full

Enterobacteriaceae Staphylococcus Escherichia Streptococcus Neisseria Actinomycetales Fusobacterium Prevotella

Table 3.3: OTU representative sequence taxonomic identification and alignments for synthetic community samples.

OTU ID NCBI ID HOMD ID SUM Enterobacteriaceae Enterobacter sakazakii Enterobacter hormaechei 37.26% Staphylococcus Staphylococcus aureus Staphylococcus aureus 29.86% Klebsiella Klebsiella pneumoniae Klebsiella pneumonia 24.96% Escherichia Escherichia coli Escherichia coli 2.85% Enterobacteriaceae Enterobacter and Klebsiella Enterobacter hormaechei 2.18% Streptococcus Streptococcus pneumoniae Streptococcus pneumoniae /oralis/mitis 0.62% Streptococcus Streptococcus intermedius Streptococcus intermedius 0.48% Rhodanobacter Rhodanobacter thiooxydans No Significant Hits 0.19% Neisseria Neisseria mucosa Neisseria flava/sicca/mucosa/oralis 0.12% Actinomycetales Sanguibacter keddieii Micrococcus luteus 0.12% Gammaproteobacteria Shewanella algae/halitosis No Significant Hits 0.12% Fusobacterium Fusobacterium necrophorum Fusobacterium necrophorum 0.09% Staphylococcus Staphylococcus aureus Staphylococcus aureus 0.09% Xanthomonadaceae Rhodanobacter thiooxydans No Significant Hits 0.07% Enterobacteriaceae Enterobacter hormaechei Enterobacter hormaechei 0.06% Prevotella Prevotella melaninogenica Prevotella melaninogenica 0.06% Bacillus Bacillus bataviensis No Significant Hits 0.06% ClostridialesFamilyXI.IncertaeSedis Parvimonas micra Parvimonas micra 0.06% Pseudomonas Pseudomonas pseudoalcaligenes Pseudomonas pseudoalcaligenes 0.042% Staphylococcus Staphylococcus epidermidis Staphylococcus epidermidis 0.003%

Both SC1 and SC5 had DNA from all representative organisms whereas SC3 had less than 0.01% relative DNA abundance of the Streptococcus, Neisseria, Actinomycetales,

Prevotella, and Fusobacterium OTUs (Figure 3.10B, Table 3.2). Despite this, OTUs representing all the synthetic community bacteria were present after the full saponin blood-treatment protocol.

The exception was Fusobacterium DNA that could not be detected from SC5 when treated with saponin or with saponin plus two washes (Figure 3.10B, Table 3.2). The results from Table 3.3 suggested that the Staphylococcus recovered in the community mainly represented S. aureus.

Overall, these results indicated we could apply paired-end Illumina profiling to bacterial

DNA recovered from whole blood subjected to the saponin blood-treatment protocol. The best recovery of microbial diversity was seen with the full treatment, despite the increase in erroneous human DNA amplification (Figure 3.10). There were two unexpected findings from these results, namely the abundance of OTUs that had a taxonomic identity of Klebsiella as well as the marked underrepresentation of S. epidermidis (Table 3.3), which was cultivated easily from saponin- treated blood (Figure 3.6). Based on these findings, the full saponin blood-treatment method was examined in three case studies to assess the application to human sepsis blood samples.

3.2.6 Case Studies

In order to validate the method, clinical samples from adult ICU patients with sepsis were collected and evaluated using culture-dependent and culture-independent approaches. Whole blood samples were treated with saponin and hypotonic washes (Section 2.4.1). Primary infection samples were cultured (Section 2.4.2). Taxonomic summaries were filtered to remove the contaminating sequences listed in Table 2.4. A total of 329,171 sequences were recovered

and represented by 266 OTUs. The taxonomic summary combined OTUs at the same taxonomic level resulting in 118 taxa represented.

3.2.6.1 Filtering OTU Table

3.2.6.1.1 Removal of Pseudomonas OTUs

Initial taxa summaries revealed a high proportion of Pseudomonas OTUs in the blood samples. It was difficult to determine if these OTUs were representative of infection or contamination. There were 4 Pseudomonas OTUs and 3 Pseudomonadaceae OTUs present in these samples. The Pseudomonas taxa represented 38% of the OTU abundance in ASN165 blood, 35% in ASN167 blood, and 12% in ASN328 blood. All Pseudomonas OTUs were present at negligible levels in the primary infection samples. The sequences were aligned to the HOMD

Database to look at the possible species level identification. The most abundant Pseudomonas

OTU had 99.4% sequence similarity to P. pseudoalcaligenes, whereas the second most abundant

OTU had a 98.1% sequence similarity to P. fluorescens. The last two OTUs had less than 97% matches to Pseudomonas species and P. fluorescens. There was only one Pseudomonadaceae

OTU present in multiple samples and it had a 95.6% sequence similarity to P. otitidis as well as

P. aeruginosa. Since these OTUs were highly abundant in all blood samples but not in the infection samples, it suggested a common contaminant was present during the processing of these blood samples and they were removed from the final analysis. The Pseudomonadaceae

OTUs were removed as the sequence similarity to P. aeruginosa was below the 97% similarity level.

3.2.6.2 Removal of Low Abundance OTUs

Following the filtering of Pseudomonas, Pseudomonadaceae, and the OTUs listed in

Table 2.4, the bacterial DNA profile consisted of 260 OTUs. For clarity, only OTUs that represented more than 0.1% of the relative DNA abundance in the sample were plotted. This resulted in 48 OTUs representing 29 taxons.

3.2.6.3 Case-Studies

The first case involved a septic pneumonia patient, ASN165. This individual was a 73- year-old male patient admitted with a bacterial pneumonia, an APACHEII score of 22, and a

SOFA score of 13. This patient spent 46 days in ICU and 102 days in hospital. Chest tube (CT) aspirate fluid was collected on Day 1 and 3 and whole blood was collected on Day 3 of the patient’s ICU stay. Culturing of the primary infection sample was done for the CT fluid and recovered only S. pneumoniae at 105 CFU/ml (Figure 3.11). Illumina sequencing of the 16S rRNA V3 region resulted in over 155,000 reads for Day 1 and Day 3 CT fluid. The genera

Streptococcus represented 99.99% OTU abundance on Day 1 and Day 3. Day 3 whole blood was processed with saponin and cultured (Figure 3.11). Partial 16S rRNA sequencing of recovered isolates indicated two Streptococcus species (S. mitis/oralis/pneumoniae and S. vestibularis) and

Actinomycetes species were present at less than 10 CFU/ml (Figure 3.11). Molecular profiling of

Day 3 blood resulted in 431 reads representing 13 OTUs. In the taxonomic summary, the

Gammaproteobacteria DNA represented 49.8% of the relative abundance followed by the

Enterobacteriaceae DNA at 18.0%.

Figure 3.11: Bacterial DNA profiles of case studies from septic ICU patients. Taxonomic summaries of the bacterial DNA extracted from the primary infection sample and saponin treated whole blood from each case study patient. Each bar represents the combined results from two separate experiments with each PCR sample amplified in triplicate. The size of the bar indicated the relative DNA abundance with the taxonomic identification labeled for each major group. The table indicated the comparison of the clinical diagnostic culture results to our culture results.

100

Admitting Bacterial pneumonia Liver abscess Drug overdose Diagnosis Diagnostic Blood Negative Negative Negative Culture Diagnostic Fluid S. pneumoniae Streptococcus milleri Group S. aureus and S. pneumoniae Culture Primary S. Pneumoniae (CT) S. anginosus and S. intermedius S. aureus, S. intermedius, S. Infection and S. mitis/oralis/pneumoniae, S. (AF) constellatus , L. fermentum, L. Saponin Culture vestibularis, and M. luteus (Blood) salivarius,L. oris, L. vaginalis, Actinomycetes sp. (Blood) and L. rhamnosus (ETT)

Lactobacillus Fusobacterium

Veillonella Rothia Streptococcus Streptococcus Gemella Propionibacterium Enterobacteriaceae Enterobacteriaceae Prevotella Bacteroides

Prevotella

Streptococcus

Gammaproteo Gammaproteo bacteria Staphylococcus bacteria

Gammaproteo bacteria

Streptococcus Streptococcus Streptococcus

101

The Streptococcus DNA represented 7.9% of the relative DNA abundance whereas the

Actinomycetales represented 2.6% (Figure 3.11). Staphylococcus DNA was present at 4.3%

(Figure 3.11). Remaining OTUs represented less that 3% of the relative DNA abundance included Proteobacteria, Caulobacter, Propionibacterium, Lachnospiraceae, and

Microbacterium (Figure 3.11). Clinical diagnostic culture results were also obtained for this patient. Blood culture results were negative on Day 1 and Day 3, whereas sputum culture indicated growth of S. pneumoniae on Day 1. The only OTU in common between the ASN165 blood and CT fluid was Streptococcus. Overall, the molecular profiling of the blood and CT fluid correlated with both the in-depth culture and diagnostic laboratory cultures (Figure 3.11).

The second case study, ASN167, was a 37-year-old patient admitted to ICU with a pyogenic liver abscess related sepsis. This patient had an admitting APACHEII of 15, a SOFA score of 4, and spent 1 day in ICU followed by 23 days in hospital. Abscess drainage fluid (AF) and whole blood were obtained. In-depth culture of the abscess fluid indicated two members of the Streptococcus milleri group (SMG), S. anginosus and S. intermedius, were present at 105

CFU/ml (Figure 3.11). Clinical lab culture of the AF also indicated heavy SMG growth. The bacterial DNA profile of the AF resulted in 146,583 reads with the Streptococcus OTU representing 99.99% of the total OTU abundance (Figure 3.11). Saponin treated whole blood was cultured and M. luteus was recovered at less than 10 CFU/ml. Molecular profiling of saponin treated blood from ASN167 was done for Day 1 of ICU admission (Figure 3.11). There were

4434 reads that matched to 17 OTUs. In the taxonomic summary, Gammaproteobacteria DNA was the most abundant at 43.7%, followed by the Streptococcus DNA at 22.1% and the

Enterobacteriaceae DNA at 14.2% (Figure 3.11). Remaining OTUs present at 4% or less of total

OTU abundance were; Bacillus, Proteobacteria, Caulobacter, Propionibacterium,

102

Granulicatella, Gemella, Rothia, Streptophyta, Actinomycetales, and Microbacterium (Figure

3.11). Blood culture results for ASN167 were positive for SMG three days prior to ICU admission. Upon ICU admission, blood culture results were negative after 72 hours of cultivation. The only OTU found in both the blood and AF was Streptococcus. The results from the synthetic community analysis revealed that M. luteus could only be identified at the order

Actinomycetales level. As such, the Actinomycetales OTU likely correlated with the in-depth culture results in which M. luteus was recovered (Figure 3.11). These results also suggested that

Streptococcus from the abscess infection was present in the bloodstream. Similar to the above results, a good correlation between molecular profiling, the clinical diagnostic laboratory data, and in-depth culture data was observed for both the AF and whole blood samples.

The final case study was ASN328, a 26-year-old male patient admitted to ICU with a drug overdose. The patient had an admitting APACHEII of 22, SOFA score of 10, spent 4 days in ICU, and 7 days in hospital. In-depth culture of the ETT fluid indicated the patient had S. aureus at 103 CFU/ml, S. intermedius and S. constellatus at 102 CFU/ml, as well as five species of Lactobacillus (L. fermentum, salivarius, oris, vaginalis, rhamnosus) at 102 CFU/ml (Figure

3.11). However, the diagnostic lab report indicated S. aureus present moderately and heavy S. pneumoniae. Molecular profiling of the ETT fluid resulted in 21,481 reads representing 36

OTUs. Representing 61% of the taxonomic profile was Streptococcus. The Prevotella group represented 8.7% of the relative DNA abundance followed by Gemella at 7.4%, Veillonella at

4.6%, Fusobacterium at 2.6%, Staphylococcus at 2.5%, Lactobacillus at 2.2%, Rothia at 1.8%, and Granulicatella at 2.0%. Molecular profiling of whole blood from ASN328 resulted in 3003 reads that fell into 26 taxonomically grouped OTUs. The Staphylococcus taxonomic group was the most abundant at 29.1%. This was followed by the Gammaproteobacteria DNA present at

103

13.1%, Lachnospiraceae at 11.9%, Bacteroides at 7.8%, Streptococcus at 7.6%, and Prevotella

DNA at 4.0%. The remaining OTUs represented less than 2% of the relative DNA abundance but included several of the OTUs also identified in the ETT fluid sample (Figure 3.11). Diagnostic blood culture obtained from ASN328 on the same day was negative. There were several OTUs present in both the blood and ETT fluid including those representing Staphylococcus,

Streptococcus, Rothia, Gemella, Prevotella, Fusobacterium, and Lactobacillus (Figure 3.11). In this case study, the in-depth culture indicated greater bacterial diversity was present in the ETT sample when compared to the clinical laboratory culture. The molecular profiling also suggested that Staphylococcus bacteria or bacterial products were present in the bloodstream yet the clinical diagnostic results remained negative.

Collectively, the in-depth culture and molecular profiling provided greater insight into these patient’s bacterial infections when compared to the clinical diagnostic results.

3.3 Discussion

Many molecular profiling studies chose their DNA extraction methods without an obvious rationale or reported validation [152]. The extraction of microbial DNA from blood can be difficult since the ratio of human-to-microbial DNA may be orders of magnitude apart.

Further, the inherent issues in recovering DNA from Gram-positive or fungal organisms must also be considered [153]. To date, the use of molecular diagnostics in sepsis have been met with modest success and extraction protocols from commercially available kits lack sufficient DNA extraction protocols [92,107,108,110,115-117,154,155]. Differences in the structure of bacterial cell walls impact the efficiency of cell lysis leading to distortion of the microbial community

104

predicted composition [152]. Furthermore, the majority of these kits use indirect methods to lyse bacterial cells, such as heat or bead beating. While this may improve lysis, it also shears DNA, therefore, negatively impacting PCR efficiency and increases the formation of chimeric products

[152]. For these reasons, mechanical disruption was not used in the DNA extraction method.

Many pathogens implicated in sepsis are difficult to lyse organisms, including Gram- positives such as S. pneumoniae, S. pyogenes, and S. aureus [156]. Despite this, few DNA extraction methods use enzymatic lysis (whereby lysozyme is incorporated) prior to purification of DNA with a kit-based method [108]. However, our data demonstrated that lysozyme alone was insufficient to recover DNA from all bacteria in a mixed community (Figure 3.4). Other groups have employed methods to selectively remove eukaryotic DNA prior to extraction of prokaryotic DNA [110,112,113]. Although this helped improve PCR, they involved difficult separation protocols and resulted in decreased DNA yields. As such, these methods were not included for this study.

Interactions with blood cells and serum proteins can reduce recovery of bacteria leading to reduced bacterial DNA yield from blood [5]. Furthermore, whole blood contains many naturally occurring substituents that can impair PCR reactivity including heme, leukocyte DNA, immunoglobulin G found in plasma, haemoglobin found in red blood cells and lactoferrin found in white blood cells [157]. As such, the development of our method required the removal of human blood cells and PCR inhibitors present in whole blood.

105

Since commercially available kits have not been sufficiently validated for molecular studies, a comprehensive DNA extraction protocol incorporating enzymatic lysis was developed.

Hen egg-white lysozyme catalyzes the hydrolysis of 1,4-β-linkages in the peptidoglycan of bacterial cells [152]. However, differences in cell-wall structure between bacterial species leads to varying susceptibility to lysozyme digestion [152]. For example, bacteria with O-acetylated peptidoglycan including the bacterial pathogens N. gonorrhoeae, P. mirabilis and S. aureus are resistant to lysozyme but susceptible to mutanolysin [158,159]. Mutanolysin is a muralytic enzyme that acts on the peptide to glycoside bonds in peptidoglycan [159]. It has been shown to also have activity against harder to lyse Streptococcus, Lactobacillus, and Actinomyces species

[160]. Figure 3.4 demonstrated that recovery of DNA from a polymicrobial sample was optimal when lysozyme and mutanolysin were used together. The results were comparable to previous studies where a lytic enzyme cocktail of containing lysozyme, mutanolysin and lysostaphin (a

Staphylococcal specific pentaglycin cleaving enzyme) consistently lysed cells of different species more effectively than lysozyme alone [152]. The addition of lysostaphin was deemed unnecessary given the results of the mock community since Staphylococcus DNA was recovered at high abundance under all conditions (Figure 3.10).

Part of the method development involved determining the best way to remove potential

PCR inhibitors including proteins and RNases. The addition of Proteinase K and RNase A was essential to allow for PCR amplification of purified DNA (data not shown). A phenol- chloroform-isoamyl alcohol extraction was used to separate DNA from lipids and degraded cellular components. Furthermore, phenol-chloroform-isoamyl alcohol extraction increased the concentration of DNA recovered when compared to other DNA extraction methods [152]. This helped eliminate any PCR inhibitors present in the preparations by separating the DNA from

106

other organic matter [106]. Lastly, a column-based purification step (Zymo columns) was incorporated to concentrate and purify the DNA (Figure 3.5).

In this study, a method of lysing blood cells prior to DNA extraction was evaluated in order to improve the recovery of bacterial cells and DNA from small volumes of blood. The use of detergents in sepsis diagnostics has been evaluated for blood culture media since the 1990s when it was shown to improve the recovery of fungal organisms, coagulase-negative

Staphylococcus, and Pseudomonas species while reducing the incidence of false-positive results

[161]. Saponin, a plant metabolite, interacts with cellular membrane components including phospholipids and sterols thereby resulting in the lysis of both red and white blood cells

[93,162]. Saponin at low concentrations (1% w/v) does not impact microorganism growth [162] and recent studies have used saponin at concentrations (1-5% w/v) to increase the number of bacteria present in blood culture positive samples prior to assessment with MALDI-TOF MS

[163,164]. In our studies, saponin was added at a final concentration of 0.85% directly with 1-

2ml of whole blood. In accordance with previous studies, there was no significant loss of bacterial growth across a panel of bacteria ranging from highly ubiquitous organisms (E. coli, S. epidermidis, S. aureus), fastidious organisms (S. intermedius, S. pneumoniae), and anaerobic organisms (P. melaninogenica and F. necrophorum, Figure 3.2). There was a two-log decrease in

CFU/ml for F. necrophorum but it was not significant with a p-value of 0.238 (Student’s unpaired t-test). The drop in CFU/ml was likely a result of manipulations to the organism outside of the anaerobic chamber and exposure to atmospheric oxygen. Studies have shown that the related oral cavity organism F. nucleatum is killed by a 6-hour exposure to oxygen [165]. As such, it is not unexpected that even limited exposure to oxygen would have impacted the recovery of F. necrophorum. In contrast, the Prevotella species have been shown to have

107

adaptive aerotolerance [166], which could account for the consistent recovery of P. melaninogenica even after atmospheric oxygen exposure (Figure 3.2).

Since the growth of bacteria was not inhibited with saponin, its addition to whole blood was assessed to determine how it would impact recovery of microorganisms in synthetic communities. Following saponin treatment and hypotonic washes of whole blood, FACS analysis using a lipid membrane dye and DNA dye suggested recovered bacterial cells were intact as the majority of cells, gated by size to ensure they were microbial, stained positive for both dyes (Figure 3.3). TRFLP results for the pre-treated and FACS sorted cells demonstrated that PCR inhibitors present in the original sample were sufficiently diluted out and did not impair

PCR amplification. Overall, the DNA extraction method preserved the microbial diversity with better recovery seen in the aerobic TSY plate pool as compared to the anaerobic FAA plate pool

(Figure 3.3). This could be a result of oxygen exposure during the blood treatment, cell labeling, and FACS.

There are limitations to these results that restricted our ability to assess microbial recovery. Namely, TRFLP data only provided a semi-quantitative assessment and the culture pools were highly concentrated. Further, the FACS data did not definitively indicate that microbial cells were intact or alive. In order to determine if our method could recover total microbial abundance from a defined community, mock communities representative of organisms typically recovered from sepsis infections were used (Table 2.1). The applicability of the saponin blood-treatment method was assessed in a cultivation-dependent and -independent approach. The mock community was spiked into blood collected from healthy donors in vacutainers to parallel the collection from actual septic patients enrolled in the study. The percent recovery of cultured organisms was highly variable in the mock community with over-representation of E.

108

hormaechei, S. aureus and S. epidermidis observed across all conditions (Figure 3.7). Minor variability in recovery of bacteria when blood was treated with saponin alone compared to saponin and hypotonic washes was observed (Figure 3.6). Not surprisingly, higher initial inoculum’s resulted in improved recovery (Figure 3.6). The anaerobic organisms were difficult to recover from the mixed community and not detected after two washes (Figure 3.7). Since washes were performed in atmospheric oxygen conditions, it was expected that these organisms would not be cultivable following this procedure. Fastidious S. intermedius and S. pneumoniae were recovered under all concentrations when treated with saponin alone as well as saponin with two hypotonic washes (Figure 3.6).

Surprisingly, E. coli and N. flava had the poorest recovery in the mixed communities, as higher inoculums were required to recover these organisms compared to the other bacteria (Table

3.1). Other culture-independent studies have reported under-representation of Gram-negatives, including E. coli, with the reasons for this unknown [152]. Since these organisms grew in saponin alone (Figure 3.2) and the recovery improved following hypotonic washes (Figure 3.8), it was speculated that the inhibition of growth was due to factors in the blood or interactions with other microbes in the community. Indeed, human blood cells and plasma have many bacteriostatic and bactericidal effects [167]. The use of SPS in clinical culture media is used to circumvent this effect [167]. In this study, no SPS was added to samples as it is also a strong

PCR inhibitor [106] and SPS-treated samples could not be PCR amplified for molecular profiling

(data not shown). Since the recovery of N. flava improved with each subsequent hypotonic wash, the reduced recovery in saponin alone was likely due to strong bacteriostatic effects of whole blood on these organisms (Figure 3.6). Previous studies have demonstrated that E. coli growth decreases by 20% following drastic changes in osmolarity [168]. Therefore, the reduction in

109

CFU/ml of E. coli following the washes with sterile distilled water could be partially attributed to autolysis. In addition, studies examining the role of the oral microbiota in resistance to E. coli colonization in mice revealed that viable counts of E. coli decreased in a synthetic community including Staphylococcus and Streptococcus species [169]. This inhibition was partially attributed to the ability of the Streptococcus species to produce hydrogen peroxide [169].

Although this study was done with murine species, it suggested that a reason for the poor recovery of E. coli could be partially attributed to antagonist interactions with the Streptococcus and Staphylococcus in the mock community. It is unknown if there were other possible antagonist interactions occurring the mock communities but the potential for inter-microbial interactions in skewing the microbial composition in the recovered communities was plausible.

The molecular profiling of the mock communities spiked into blood was done to examine whether DNA profiles accurately represented the microbial community. Molecular profiling of the mock communities indicated primer cross-reactivity with human DNA (Figure 3.9). Studies in our lab have demonstrated that these sequences match to human chromosomal DNA (data not shown) and non-specific amplification of human DNA with universal 16S primers has been well documented [151,170,171]. Not unexpectedly, the proportion of amplified human DNA increased as the concentration of bacteria decreased from SC1 to SC5 (Figure 3.9). Nevertheless, these sequences were easily removed from the taxonomic profile thereby permitting analysis of the bacterial components of the DNA profiles.

The over-estimation of diversity in next-generation sequencing studies has been well documented [172]. In particular, the use of heuristic methods for OTU clustering overestimates the number of groups. In the mock community samples, there were 203 OTUs that grossly overestimated the diversity in these mock communities consisting of 8 bacterial genera. For this

110

study, some of the diversity likely resulted from the highly optimized DNA extraction protocols and sensitivity of the PCR. With these methods optimized to recovery all microbial DNA, there is also a risk of amplifying any environmental DNA present during the blood collection, the

DNA extraction procedure, and the PCR set-up. Similar studies have also reported recovery of

DNA not correlated to the mock community, and found higher rates of contamination in samples with lower bacterial DNA [173]. This was also observed in the mock communities with the SC5 samples having the greatest abundance of OTUs that could not be correlated to the mock community. In such studies, sequences that represented less than 1% of the total DNA diversity were excluded from the final analysis. This helped justify the filtering of sequences representing less than 0.1% of the relative DNA abundance in our samples. This reduced the OTU number from 210 to 35 and permitted the detection of DNA from the mock community organisms.

Although there was still over-estimation of diversity, the proportion of OTUs that could not be taxonomically assigned to the mock community organisms was quite low accounting for at most

20% of the DNA in SC5 (Figure 3.9). These OTUs likely represent contaminant DNA that were minor components of the taxonomic profile yet when the template DNA became limiting, their relative abundance was increased.

Despite the amplification of non-bacterial DNA, detection of contaminant DNA sequences, and OTU derived over-estimation of diversity, the recovery of the representative DNA in the mock community was optimal after the full saponin blood-treatment protocol (Table 3.2, Figure

3.10). All of the mock community organisms were identified but at varying levels of sequence resolution ranging from M. luteus identified as Actinomycetales, the E. hormaechei identified as

Enterobacteriaceae, and the remaining identified at the general level (Table 3.3). Taken together, the composition of the bacterial DNA profiles paralleled the culture-based composition fairly

111

well as the DNA composition was domination by Staphylococcus and Enterobacteriaceae reflecting the enrichment of these organisms in the mock communities following the blood spiking experiments (Figure 3.7).

The application of the methodology developed to three ICU case studies demonstrated the strength of molecular profiling analysis when compared to clinical diagnostics. In the case of

ASN165, the combination of the molecular profiling and clinical data indicated that this patient had S. pneumoniae pneumonia with this pathogen implicated in the bloodstream infection

(Figure 3.11). This method provided more conclusive evidence for a S. pneumoniae infection as compared to the clinical diagnostic laboratory blood culture that failed to recover any bacteria from two blood culture bottles inoculated with 10-20 ml of blood whereas two Streptococcus species were recovered from 1ml of saponin treated blood and Streptococcus DNA was detected in the blood.

Patient ASN167 had a hepatic abscess-associated sepsis yet their clinical data suggested that the patient no longer had an active bacteremia (Figure 3.11). However, the molecular profiling data suggested Streptococcus bacteria or bacterial products were still present in the bloodstream (Figure 3.11). This would not be unexpected as abscess infections provide a reservoir of bacteria that can continuously be shed into the bloodstream [5]. Taken together, the molecular profiling correlated with the clinical context of the patient and provided additional support that the patient still had a Streptococcus bloodstream infection despite a negative blood culture.

For ASN328 the molecular profiling data and in-depth culture data suggested a more complex infection aetiology when compared to the clinical diagnostic results. The presence of several Lactobacillus species, Prevotella, and Fusobacterium in the ETT fluid from this patient

112

suggested that there was an aspiration event resulting in upper airway microbiota translocating to the lower airways in addition to the Streptococcus and Staphylococcus infection (Figure 3.11). L. rhamnosus, L. salivarius, L. fermentum and L. oris are all common lactobacilli found in the oral cavity microbiota [174,175]. L. vaginalis is rarely cultured in oral samples but has been detected using molecular profiling methods in saliva and fecal human samples [174]. Molecular profiling also detected Rothia, Prevotella, Granulicatella, Veillonella, Gemella, and Fusobacterium OTUs that have also been recovered in studies on chronic airway infections [100,109]. The blood molecular profile for ASN328 shared several OTUs with the ETT fluid (Figure 3.11) suggesting that there was possible translocation of these bacteria or their DNA from the lungs into the bloodstream despite the negative blood culture results.

Blood culture based assessments indicate that the incidence of polymicrobial sepsis is low ranging from 10-20% [13,29]. Despite this, polymicrobial DNA profiles were identified in all three patients using a non-targeted molecular profiling method. In addition, some of the OTUs identified in the blood molecular profile were the same genus as bacteria cultivated from either the primary infection sample or the blood sample (Figure 3.11). The Gammaproteobacteria OTU recovered in all case study blood samples suggested this OTU could represent a common contaminant in the whole blood samples (Figure 3.11). In addition, there was several low abundance OTUs recovered. Based on the mock community results, suggesting that up to 35% of

OTUs were not representative of the initial microbiota, these low abundance OTUs may represent an inflation of diversity using this method on whole blood where the concentration of recovered bacteria were low (Figure 3.11). Staphylococcus DNA was also identified in the whole blood molecular profiles (Figure 3.11). In the ASN165 and ASN167 samples the levels were quite low and this OTU could not be correlated to the clinical presentation of these patients.

113

However, in ASN328 this OTU could be correlated to the clinical presentation since this patient presented with S. aureus in the lungs and likely represented a role for this organism in the bloodstream infection. In summary, the molecular profiling results provided additional data to support the presence of bloodstream infections in these patients however the results were analyzed cautiously with the clinical data used to support the conclusions.

In summary, a method for the lysis of host cells in blood was developed and assessed on mock communities and clinical samples. This method allowed for the recovery of intact viable microbial cells from blood. A DNA extraction method and paired-end Illumina 16S rRNA sequencing method were assessed on mock communities spiked into whole blood. These methods enabled the recovery of microbial diversity that paralleled the culture dependent results.

These methods also demonstrated successful cultivation and DNA amplification from clinical

ICU samples with the data providing additional insight into each patient’s clinical presentation of sepsis.

114

Chapter Four: Bacterial DNA Profiles from Intensive Care Unit Patients

4.1 Introduction

Success using traditional culture-based approaches in clinical diagnostic microbiology is limited by the highly selective culture conditions used (i.e., blood culture). Targeted approaches to microbiology in which comprehensive, non-selective, culture conditions are used have been successful [100] but have not been implemented into routine clinical diagnostics due to costs and labour requirements. As such, it is widely stated that most of the microbiome is not readily cultured [98], yet attempts to robustly culture the human microbiota are limited. Blood culture diagnostics show limited positivity, which has limited our understanding of sepsis bloodstream infections [5]. As a result of these limitations, the use of a culture-independent approaches have been proposed.

Next-generation sequencing technology for the taxonomic identification of bacteria has become the logical extension of 16S rRNA gene sequencing in clinical microbiology. To the best of our knowledge, there have been no molecular profiling-based evaluations of invasive infections associated with sepsis or critically ill patients [61]. In order to determine the propensity for polymicrobial infections in septic ICU patients, whole blood samples were obtained, subjected to saponin mediated blood cell lysis, DNA extraction, partial 16S rRNA

PCR, and paired-end 16S rRNA Illumina sequencing. These DNA-based molecular profiles were examined to determine how the bacterial DNA recovered from whole blood could provide insight into the bloodstream infections, their correlations with primary infections, and their temporal changes during ICU treatment.

115

4.2 Results

4.2.1 Sample Cohorts

Whole blood and infection samples were collected from ICU patients (Table 2.3). The whole blood samples were classified as the SB cohort with each patient having a unique ASN identifier. The first control group included patients in the ICU who were admitted for post- operative recovery or were suffering from major trauma (classified as IC samples). The neurological trauma ICU patients (classified as the SN cohort) were considered a separate control group as these patients all had a brain-related event including trauma, haemorrhage or ischemic stroke but were not septic. However, this group would have a higher risk of bacteremia based on the results of Wong et al. demonstrating an immunosuppressive effect following brain injury [176]. Blood was also collected from 12 healthy adults (classified as HB samples) as the final control. These patients/adults were chosen since they would represent the potential for contaminating DNA from the blood collection process including skin-associated bacteria or bacterial DNA present in the sterile vacutainers as well as the potential for bacteremia from opportunistic organisms found associated with health care environments [177].

4.2.2 Bacterial DNA Patterns in Septic Blood

In order to determine if the optimized DNA extraction from whole blood would improve the detection of bacterial sepsis infections, samples were collected from 116 ICU patients admitted to the Foothills Medical Center with a clinical suspicion or diagnosis of sepsis. Within

24 hours of admission, 4ml of whole blood was collected in a K2EDTA vacutainer. Following

116

saponin blood-treatment on 1-2ml of blood and DNA extraction, the 16S V3 rRNA sequence was amplified and sequenced using the Illumina MiSeq. Remaining blood was stored for later use or treated with saponin and cultured as outlined in section 2.4.2.

4.2.2.1 Sequence Filtering

Following sequence processing outlined in Section 2.9.4 and 2.9.5, the OTUs that were detected less than 10 times in the population were excluded resulting in 460,386 sequences representing 355 OTUs. There was a large range in the number of sequences per sample with a minimum number of sequences per sample of one and a maximum of 166,596. There were 141 taxonomically distinct groups clustered at the genus level when possible. Despite this, there were many OTUs that were only given a taxonomic identification at the family level (i.e.:

Enterobacteriaceae) or the phyla level (i.e., Proteobacteria) or the class level (i.e.,

Gammaproteobacteria) reflecting the maximum level in which the RDP Classifier would, with confidence, identify the OTU [133]. In terms of relative DNA abundance in the SB samples, the top OTU was the genus Streptococcus, followed by the genus Staphylococcus, the family

Enterobacteriaceae, and the genus Escherichia. In order to examine the phylogenetic relationships in the SB, samples with at least 500 sequences were compared using weighted

UniFrac and jackknife resampling of equally resampled OTU tables was done in which a subset of the samples within each group were repeatedly resampled to ensure clustering was consistent

[137]. Due to low sequencing depth, 62 patient samples were not clustered and will be discussed in Section 4.2.2.6.

117

4.2.2.2 OTU Diversity of Streptococcus and Staphylococcus

Since the Staphylococcus and Streptococcus taxonomic groups were represented by several highly abundant OTUs, this suggested that more than one species were present in the SB cohort of samples. With the paired-end Illumina reads, the 160bp representative sequence for each OTU was aligned to the NCBI using BLAST to further identify the OTUs identified in all

116 patients (Table 4.1). Multiple alignments were recovered and the species identifications were based on the alignments that had the closest match to the representative sequence. However, it is important to note that these alignments are only suggestive of the species that might be represented by these OTUs. The Streptococcus OTU that matched to the S. intermedius/anginosus group was the most prevalent of the Streptococcus OTUs followed by the

S. pneumoniae/oralis/mitis group (Table 4.1). Other Streptococcus OTUs were far less common but included alignments to clinically relevant organisms such as the S. dysgalactiae/agalactiae group and S. pyogenes. S. sinensis has rarely been identified in human infections but had been found associated with infective endocarditis [178]. The last two Streptococcus OTUs were present at scant levels and the representative sequence matched to non-human associated species

[27].

The Staphylococcus OTU that aligned to S. aureus was by far the most abundant in the SB samples (Table 4.1). The other Staphylococcus OTUs represented less common but still clinically significant species. Interestingly, the second most abundant Staphylococcus OTU aligned to S. sciuri, which was originally considered to be an animal-associated species of

Staphylococcus [179,180].

118

Table 4.1: Breakdown of OTU taxonomic identities for Streptococcus and Staphylococcus abundant in SB samples.

OTU ID OTU Abundance Representative Sequence ID % # Match Streptococcus 8 33.50% Streptococcus intermedius/anginosus 100 Streptococcus 5 10.59% Streptococcus 100 pneumoniae/oralis/mitis Streptococcus 40 2.33% Streptococcus salivarius/vestibularis 100 Streptococcus 175 1.58% Streptococcus 99 dysgalactiae/agalactiae Streptococcus 127 0.50% Streptococcus sinensis 100 Streptococcus 25 0.34% Streptococcus pyogenes 100 Streptococcus 302 0.06% Streptococcus pluranimalium 99 Streptococcus 700 0.06% Streptococcus uberis/porcinus 99 Staphylococcus 2 96.61% Staphylococcus aureus 100 Staphylococcus 44 2.52% Staphylococcus sciuri 99 Staphylococcus 82 0.48% Staphylococcus intermedius 99 Staphylococcus 144 0.23% Staphylococcus saprophyticus 99 Staphylococcus 304 0.09% Staphylococcus epidermidis 100 Staphylococcus 119 0.03% Staphylococcus aureus 97 Staphylococcus 434 0.02% Staphylococcus 100 pseudointermedius/chromogenes Staphylococcus 876 0.01% Staphylococcus aureus 97

119

However, there have been several recent reports of S. sciuri in hospital environments and associated with urinary tract infections [179,180]. The presence of a S. saprophyticus OTU

(Table 4.1) was not surprising as this species is commonly associated with community-acquired urinary tract infections [181] as well as sepsis [182]. There have also been reports of S. pseudointermedius and S. intermedius infections in humans in a hospital setting thereby making the presence of OTUs aligned to these species noteworthy [183,184].

4.2.2.3 Bacterial DNA patterns in ICU SB samples

The taxonomic profiles of the 54 patient samples with a sequencing depth above 500 were clustered into three main groups (Figure 4.1A). The jackknife support values were relatively low for the clustering groups, which may reflect the low sequencing depth (Figure

4.1B). The proportion of non-bacterial but likely human DNA (referred to as the “noRoot” OTU) in the SB ranged from 99.98% to 0.007% with the average being 92.4% (Table 4.2). For samples with a genus level identification, the representative OTU sequence was examined using BLAST to identify a suggested species level identification (Table 4.2). The clinical data for the patients listed in Table 4.3.

The Group 1 samples were distinguished by the abundance of Streptococcus DNA in the blood. Group 1 samples were divided into two clades; samples with 65% or higher relative abundance of Streptococcus and samples with less than 65% but greater than 30% Streptococcus

DNA (Figure 4.1B).

120

Figure 4.1: Septic whole blood samples cluster into three groups based on their taxonomic bacterial DNA profiles. Taxonomic profiles of whole blood samples with 500 or more sequences and clustered using weighted UniFrac (54 patients). A composite unweighted pair group method with arithmetic mean (UPGMA) tree of all the samples was generated with the profiles ordered based on their placement in the UPGMA tree (A). Three groups of SB samples were clearly identified. Group 1 was defined by the abundance of Streptococcus in the profile,

Group 2 by the abundance of Gram-negative OTUs, and Group 3 by the abundance of

Staphylococcus. The jackknife support values were low for each cluster group (B). The average taxonomic profile for each cluster group shows the breakdown of the bacterial DNA distribution in each taxonomic cluster group (C).

121

Streptococcus Staphylococcus Enterobacteriaceae Escherichia Gammaproteobacteria Pseudomonas Finegoldia Proteobacteria Xanthomonadaceae Klebsiella Bacillaceae Moraxella Legionella Lachnospiraceae Prevotella Fusobacterium Bacteroides Serratia Neisseria Anaerococcus

122 B

ASN477.1 0.9 ASN474.1 1 ASN476.1 0.9 ASN461.SAP ASN466.1 1 0.7 ASN348.1 0.4 ASN464.1

1 ASN424 ASN434 0.8 ASN408 1 0.5 0.5 ASN454 ASN418 1 ASN409 0.5 ASN451 0.6 ASN440 0.3 ASN436 0.5 ASN458 ASN479.1 ASN415 1 ASN343.1 1 1 ASN339.1 0.3 ASN340.1 0.4 ASN444 0.8 ASN432 0.4 ASN381SAP 0.3 1 ASN371 0.3 1 ASM379 ASN420 1 ASN473.1 ASN328 0.7 0.6 ASN297 1 1 ASN292 0.3 0.5 ASN300 ASN338 ASN363 0.2 1 ASN315 1 0.9 ASN429 0.8 ASN438 ASN475.1 0.9 ASN168 1 ASN167 ASN294 ASN368 0.4 1 0.8 ASN376 0.8 ASN357 0.7 ASN366 ASN463.1 1 ASN465.1 0.6 ASN470.1 0.4 0.4 ASN469.1 ASN452 0.8 1 ASN349 0.8 ASN350 ASN455

0.04

123

C Pseudomonas Moraxella Klebsiella Xanthomonadaceae Proteobacteria Escherichia Moraxella Gammaproteobacteria Escherichia Escherichia Klebsiella Pseudomonas Clostridium Clostridium Gammaproteobacteria Proteobacteria Enterobacteriaceae Enterococcus Streptococcus Escherichia Enterobacteriaceae Staphylococcus Serratia Gammaproteobacteria Finegoldia Staphylococcus Streptococcus Lachnospiraceae Enterobacteriaceae Streptococcus Streptococcus Streptococcus Staphylococcus Staphylococcus

Staphylococcus

Bacillaceae Staphylococcus Staphylococcus

Bacillaceae Prevotella Bacillaceae Staphylococcus Staphylococcus Prevotella Bacillaceae

124

Table 4.2: SB clusters, clinical microbiology, and OTU identification.

Sample Blood Culture Other Culture* “noRoot” Top RepSeqID OTU % OTU(s) Group 1 A ASN455 Group G Group G Streptococcus 98.8 175 Streptococcus Streptococcus dysgalactiae/agalactiae ASN350 Negative Negative 97.2 5 Streptococcus pneumoniae/oralis/mitis ASN349 Negative Streptococcus intermedius 99.4 5 Streptococcus pneumoniae/oralis/mitis ASN452 Negative Not Done 98.9 8 Streptococcus intermedius/anginosus ASN469 Campylobacter Enterococcus 92.8 8 Streptococcus ureolyticus, intermedius/anginosus Fusobacterium species ASN470 VRE VRE 74.8 8,5 Streptococcus intermedius/anginosus, Streptococcus pneumoniae/oralis/mitis ASN465 Negative VRE 95.4 8 Streptococcus intermedius/anginosus ASN463 Negative Not Done 97.4 8 Streptococcus intermedius/anginosus B ASN366 Negative Streptococcus anginosus, 91.9 2,5 Staphylococcus aureus, Prevotella species, CoNS Streptococcus pneumoniae/oralis/mitis ASN357 Negative VRE 97.8 5 Streptococcus 125

pneumoniae/oralis/mitis ASN376 Not Done Not Done 90.1 5,2 Streptococcus pneumoniae/oralis/mitis, Staphylococcus aureus ASN368 Negative Klebsiella pneumoniae, 94.8 5,2 Streptococcus Haemophilus parainfluenzae, pneumoniae/oralis/mitis, Prevotella species Staphylococcus aureus ASN294 Staphylococcus aureus Staphylococcus aureus 94.9 5 Streptococcus pneumoniae/oralis/mitis Group 2 2AI ASN167 Negative SMG 94.1 32 Gammaproteobacteria ASN168 Not Done Fungal 96.9 32 Gammaproteobacteria ASN475 Pseudomonas Pseudomonas aeruginosa 98.0 15,32, Proteobacteria, aeruginosa 48 Gammaproteobacteria, Pseudomonas sp. ASN438 Negative Legionella pneumophila 98.4 3,4 Enterobacter sp., Klebsiella sp. ASN429 Negative Legionella pneumophila 87.1 3 Enterobacter sp. ASN315 Negative Not Done 83.3 3 Enterobacter sp. ASN363 Staphylococcus Staphylococcus aureus 0.72 11 Serratia marcescens aureus, Gram-negative bacilli 2AII ASN338 Not Done Not Done 96.8 2,101 Staphylococcus aureus, Anaerococcus sp. ASN300 Negative Fungal 92.5 32 Gammaproteobacteria ASN292 Negative Not Done 91.0 6,76,12 Bacillus sp., 5 Lachnospiraceae, Bacillus sp.

126

ASN297 Fungal Fungal 92.1 40,15, Streptococcus sp., 125, Proteobacteria, Bacillus sp. ASN328 Negative Staphylococcus aureus, 94.9 2 Staphylococcus aureus Streptococcus pneumoniae ASN473 Pseudomonas Not Done 92.5 15 Proteobacteria aeruginosa ASN420 Negative Not Done 97.5 2,59 Staphylococcus aureus, Proteobacteria 2AIII ASN379 Negative Not Done 77.5 2 Staphylococcus aureus ASN371 CoNS Not Done 93.8 2,5,13 Staphylococcus aureus, Streptococcus pneumoniae/oralis/mitis, Escherichia coli ASN381 Negative Legionella pneumophila 94.0 13,3,2 Escherichia coli, Enterobacter sp., Staphylococcus aureus ASN432 Negative Group A Streptococcus 95.2 8,2,3 Streptococcus intermedius/anginosus, Staphylococcus aureus, Enterobacter sp. ASN444 Negative Not Done 99.2 8,3 Streptococcus intermedius/anginosus, Enterobacter sp. ASN340 Group C Not Done 99.3 2,13 Staphylococcus aureus, Streptococcus Escherichia coli ASN339 Negative Not Done 98.6 2,3 Staphylococcus aureus, Enterobacter sp. ASN343 Negative Fungal 99.7 2,15 Staphylococcus aureus, Proteobacteria

127

2AIV ASN415 Negative Not Done 93.3 97 Prevotella melaninogenica 2B ASN479 Negative Finegoldia magna 99.2 81 Finegoldia magna Group 3 A ASN458 MRSA MRSA 84.2 2 Staphylococcus aureus ASN436 Negative VRE 92.5 2 Staphylococcus aureus ASN440 Enterococcus faecium Not Done 97.2 2 Staphylococcus aureus ASN451 Not Done Not Done 99.5 2 Staphylococcus aureus ASN409 Not Done Not Done 98.6 2 Staphylococcus aureus ASN418 Not Done Not Done 87.1 2 Staphylococcus aureus ASN454 Negative Not Done 98.4 2 Staphylococcus aureus ASN408 Not Done Not Done 80.2 2 Staphylococcus aureus ASN434 Pseudomonas Not Done 8.3 2 Staphylococcus aureus aeruginosa ASN424 Bifidobacterium Not Done 88.8 2 Staphylococcus aureus species ASN464 Not Done Not Done 76.2 2,8,5 Staphylococcus aureus, Streptococcus intermedius/anginosus, Streptococcus pneumoniae/oralis/mitis ASN348 Negative Micrococcus species, 98.2 2,5 Staphylococcus aureus, Streptococcus viridians group Streptococcus pneumoniae/oralis/mitis ASN466 Negative CoNS, Coryneform bacilli, 92.8 2,8,13 Staphylococcus aureus, Candida parapsilosis Streptococcus intermedius/anginosus, Escherichia coli 128

ASN461 Fusobacterium Fusobacterium 90.8 2 Staphylococcus aureus necrophorum B ASN476 Escherichia coli Not Done 80.2 6,2,19 Bacillus sp., Staphylococcus aureus, Lysinibacillus sp. ASN474 Negative Not Done 32.6 6,2,17 Bacillus sp., Staphylococcus aureus, Moraxella sp. ASN477 Negative Not Done 96.6 6,2,32 Bacillus sp., Staphylococcus aureus, Gammaproteobacteria *Primary infection culture,VRE/MRSA Swab, toxin detected, or PCR positive

129

Table 4.3: Admissions data for the SB patients in Groups 1-3.

Sample Age Gender Admitted Admitting Diagnosis Admitting Max SOFA ICU From APACHE Outcome II Group 1 A ASN455 47 F In-patient Sepsis-Unknown 29 17 Dead ASN350 44 M Other Bacterial pneumonia 19 18 Dead ASN349 60 F In-patient Intracranial abscess 26 8 Alive ASN452 76 M ED Bacterial pneumonia 13 5 Alive ASN469 66 M OR Surgery for cellulitis 27 10 Alive ASN470 69 M In-patient Sepsis-Gastrointestinal 34 17 Dead ASN465 70 F In-patient Cardiac arrest, post-kidney transplant 28 12 Alive ASN463 70 F In-patient Congestive heart failure 28 7 Alive B ASN366 53 M OR- Tonsil or pharyngeal infection 19 10 Alive Emergency ASN357 80 F Other Septic shock 28 8 Alive ASN376 25 M OR- Haemothorax or haemopneumothorax 20 7 Alive Emergency ASN368 60 M OR- Leaking biliary anastamosis 11 7 Alive Emergency ASN294 63 F Other Self poisoning with sedatives or 10 8 Alive hypnotics Group 2 2AI

130

ASN167 37 F ED Hepatic abscess 15 5 Alive ASN168 52 F In-patient Bacterial pneumonia 31 11 Alive ASN475 77 F In-patient Gastrointestinal abscess 18 7 Alive ASN438 57 F ED Pneumonia-Other 21 6 Alive ASN429 71 F In-patient Respiratory cause 16 10 Alive ASN315 29 M OR- Necrotizing fasciitis and septic shock 7 5 Alive Emergency ASN363 33 M ED Septic shock 16 12 Alive 2AII ASN338 25 M OR- Traumatic rupture or laceration of liver 30 15 Alive Emergency ASN300 69 F OR- Small bowel infarction 24 16 Alive Emergency ASN292 63 M OR- Septic shock 34 16 Alive Emergency ASN297 58 F OR- Oesophageal or gastro-oesophageal 24 10 Alive Emergency tumour ASN328 26 M ED Self poisoning with narcotics 22 11 Alive ASN473 77 M ED Bacterial pneumonia 32 11 Dead ASN420 56 F ED Bacterial pneumonia 10 6 Alive 2AIII ASN379 74 M In-patient Pneumonia 28 12 Alive ASN371 58 M Other Bleeding duodenal ulcer 21 1 Alive ASN381 53 F OR- Necrotizing fasciitis and bacterial 16 4 Alive Emergency pneumonia ASN432 57 F ED Bacterial pneumonia 20 8 Alive ASN444 77 F ED Emphysema/bronchitis 24 4 Alive ASN340 67 M ED Cutaneous cellulitis 15 12 Alive ASN339 38 F In-patient Intracranial abscess 23 11 Alive

131

ASN343 56 M In-patient Inhalation pneumonitis (gastrointestinal 27 8 Alive contents) 2AIV ASN415 76 F ED Pneumonia-Other 30 9 Alive 2B ASN479 57 M ED Septic arthritis 19 11 Alive Group 3 A ASN458 63 M OR- Bacterial pneumonia and 40 20 Dead Emergency cardiovascular surgery ASN436 73 M ED Sepsis-Gastrointestinal 23 8 Alive ASN440 65 F In-patient Congestive heart failure and 28 12 Alive emphysema/bronchitis ASN451 79 F OR- Surgery for gastrointestinal 27 14 Alive Emergency perforation/rupture ASN409 53 M In-patient Respiratory cause 22 10 Alive ASN418 46 F ED Surgery for (resection) gastrointestinal 26 10 Dead vascular ischemia, ASN454 78 M ED Upper gastrointestinal bleeding ND* 4 Dead ASN408 45 M In-patient Surgery for abdomen-trauma 15 5 Alive ASN434 51 M In-patient Sepsis-Unknown 26 12 Alive ASN424 68 M In-patient Surgery for (resection) gastrointestinal 14 12 Dead vascular ischemia, ASN464 62 M OR- Surgery for gastrointestinal 26 12 Alive Emergency perforation/rupture ASN348 77 F OR- Septic shock 27 12 Alive Emergency ASN466 70 F In-patient Surgery for 31 9 Alive cholecystectomy/cholangitis (gallbladder removal)

132

ASN461 24 F ED Bacterial pneumonia 19 12 Alive B ASN476 66 M In-patient Septic arthritis 31 9 Alive ASN474 42 F In-patient Sepsis-Pulmonary 29 19 Dead ASN477 54 M ED Bacterial pneumonia 17 9 Alive *No Data

133

Group 1 patients also had Staphylococcus DNA present in their blood representing 6-25% of the relative DNA abundance (Figure 4.1). Using the representative sequence alignments to predict the Streptococcus species, four of the patients had the Streptococcus pneumoniae/mitis/oralis as the principal OTU, four had the Streptococcus intermedius/anginosus as the principal OTU, one had the Streptococcus dysgalactiae/agalactiae OTU, three had similar abundance of the

Streptococcus pneumoniae/oralis/mitis and Staphylococcus aureus OTUs, and one patient had similar abundance of Streptococcus intermedius/anginosus and Streptococcus pneumoniae/oralis/mitis OTUs (Table 4.1). Group 1A consisted of patients in which there was one principal taxonomic group in the bacterial DNA profile whereas Group 1B had patients in which two taxonomic groups were present at similar levels (Table 4.2).

Group 2 SB samples had the greatest diversity in terms of taxonomic representation

(Figure 4.1A). A unifying trend for Group 2 was low abundance of the Streptococcus and

Staphylococcus DNA with Group 2A having the lowest abundance of Streptococcus (Figure

4.1C). Group 2A was further subdivided into groups I, II, III, and IV based on their abundance of

DNA from Gram-negative bacteria (Figure 4.1A). In Group 2AI, the majority of the DNA diversity was represented by the Gammaproteobacteria, Proteobacteria, and Pseudomonas taxonomic groups in the first clade whereas Group 2AII were represented by the

Enterobacteriaceae and Klebsiella DNA (Figure 4.1C). Within Group 2AI, there was one SB sample in which the Serratia taxon represented 100% of the relative DNA abundance (Figure

4.1A). There was only one Serratia OTU present in the SB samples and the representative sequence had several alignments with 100% matches to the Serratia marcescens 16S rRNA gene

(Table 4.2). There was also one patient in Group 2AI, ASN438, in which Legionella DNA represented 25% of the relative DNA abundance. This was the only ICU patient where

134

Legionella DNA was recovered and there was only one Legionella OTU (Figure 4.1A). The representative sequence for the Legionella OTU was aligned to the NCBI database and had several alignments with 100% match to Legionella pneumophila subsp. Pneumophila (Table

4.2). Interestingly, this patient was admitted during the Legionella outbreak of November-

December 2012 in the Calgary region (J. Conly, personal communication). The Group 2AII samples had greater taxonomic diversity and low amounts of Enterobacteriaceae DNA (Figure

4.1C). The principal OTUs identified in the Group 2AII patients had sequence identities matching Bacillus, Gammaproteobacteria, Lachnospiraceae, Xanthomonadaceae, and

Staphylococcus (Table 4.2). The Group 2AIII isolates had a mix of OTUs representing both

Gram-positive and Gram-negative bacteria in equal proportions. Within the first clade of group

2AIII the samples were distinguished by a relatively even split between the Gram-negative taxonomic groups for Fusobacterium, Neisseria, Escherichia, Klebsiella, and

Enterobacteriaceae and the Gram-positive taxonomic groups for Streptococcus and

Staphylococcus (Figure 4.1A).

In the second Group 2AIII clade, the prevalence of Anaerococcus representing 10-12% of the DNA abundance was a distinguishing factor (Figure 4.1A). In this clade, 30-40% of the taxonomic diversity was represented by Streptococcus and Staphylococcus with Gram-negative taxa representing the remaining DNA diversity (Figure 4.1C). The principal taxonomic groups found in each patient were often a combination of Streptococcus/Staphylococcus with

Enterobacteriaceae/Escherichia (Table 4.2). Group 2AIV consisted of one patient sample in which the abundance of the Prevotella DNA at 30% separated in from the other Group 2 samples

(Figure 4.1C). Group 2B was also represented by a single SB sample in which the Finegoldia

DNA represented 76% of the OTU abundance (Figure 4.1C). There was only one Finegoldia

135

OTU, which had a 100% alignment to Finegoldia magna (Table 4.2). The final group in the SB samples was distinguished by the Staphylococcus DNA abundance. In the first Group 3A clade, the Staphylococcus represented 37-75% of the bacterial DNA (Figure 4.1A). Within Group 3A,

11 of the patients had one principal OTU that had alignments to S. aureus and the remaining three patients had a mix of principal OTUs that had alignments to S. aureus, Streptococcus

intermedius/anginosus, Streptococcus pneumoniae/oralis/mitis, and E. coli (Table 4.2). The second Group 3B clade was distinguished by the Bacillaceae and Moraxella DNA representing

25-41% and 5-14% of the molecular profiles (Figure 4.1C, Table 4.2). This was also the only group in which Clostridium and Enterococcus represented a detectable level of the bacteria DNA taxonomic profiles (Figure 4.1C).

Overall the bacterial DNA patterns identified in the whole blood of septic patients fell into three groups: (1) patients in which Streptococcus DNA was the most abundant, (2) those in which the Gram-negative and the GPAC DNA were abundant, and (3) those in which the

Staphylococcus and Bacillaceae DNA were abundant (Figure 4.1). Each septic patient had a bacterial DNA profile in blood that represented multiple bacterial taxonomic groups with the exception of ASN363 where the Serratia OTU represented the majority of the bacterial DNA.

4.2.2.4 Blood Culture Data and Bacterial DNA Profiles

The blood culture results for the SB patients were compared to the molecular profiles

(Table 4.3). Of the 54 patients clustered, blood culture results were obtained for 46 patients.

Only 14 of the 46 patients (30%) had a positive blood culture result and of those, only 5 had a cultured organism that matched the principal taxonomic group in the cluster (Table 4.3). There

136

were some samples in which good concordance between the bacterial DNA profile and clinical microbiology data was present. For example, the first patient in Group 1, ASN455, had

Streptococcus DNA representing over 75% of the bacterial DNA and the clinical diagnostic blood culture results reported Group G Streptococcus (Table 4.2). The representative sequence

ID for the Streptococcus OTU in this patient also matched to S. agalactiae/dysgalactiae, which are members of the Group G Streptococcus [27]. The bacterial DNA profile profile for ASN363, in which S. marcescens represented the majority of the bacterial DNA, was more informative than the clinical diagnostic blood culture results, which could only identify the bacteria to a

Gram-stain level, Gram-negative bacilli. In addition, S. aureus was also recovered in the blood culture but was not a major component of the bacterial DNA profile for ASN363 (Figure 4.1B).

There were also samples in which the clinical data but not the clinical diagnostic blood culture results supported the bacterial DNA profile. Patient ASN438 had Legionella DNA present in their blood yet were negative for clinical diagnostic blood culture (Table 4.2).

However, the lung fluid culture results for this patient indicated there was a L. pneumophila infection (Table 4.2). A similar pattern was seen for ASN479 in which clinical diagnostic blood culture was negative, yet F. magna was cultured from the patient’s synovial fluid and Finegoldia

DNA was also present in the blood (Table 4.2). As such, the bacterial DNA profiles for these patients suggested that bacteria below the level of blood culture detection or bacterial products were present in the bloodstream that corresponded to the patient’s primary infection despite being negative using clinical diagnostic blood culture (Table 4.2).

137

4.2.2.5 Characteristics of Patients in the SB Clusters

The metadata for each septic ICU patient was available and it included the admissions data including the patients’ age, sex, APACHE II score, SOFA score, the ICU length of stay

(LOS), and outcome (Table 4.3, Table 4.4). Both Groups 1 and 2 were approximately distributed equally between male to female patients, whereas Group 3 had two-thirds male patients (Table 4.4). The median age was the lowest for Group 2 at 57 years whereas for Group

1 and Group 3 the median age was 63 (Table 4.4). The median APACHE II score, a measure of disease severity [82], was higher for Groups 1 and 3 at 26 whereas Group 2 had a median

APACHE II of 21.5 (Table 4.4). The SOFA score, a measure of organ failure [84], was the highest on Day 1 for the Group 3 patients followed by Group 2 and Group 1 at 7. The same trend was seen for the Max SOFA as well with Group 3 at 12, Group 2 at 10 and Group 1 at 8

(Table 4.4). The ICU LOS was comparable for all three groups with a median stay of 5-6 days.

With respect to ICU mortality, only 1 patient in Group 2 (Group 2AII) died whereas Groups 1 and 3 had 23.1% and 15.4% mortality rates, respectively (Table 4.4). Overall, the patients in the Groups 1 and 3 clusters had the worst clinical presentations of sepsis with higher APACHE

II and SOFA scores resulting in the death of 23.1% and 15.4% of the patients (Table 4.4).

4.2.2.6 Septic ICU Patient Blood Samples that did not Cluster

Of the 116 septic ICU patients enrolled in the study, 62 were not used in the SB ordination due to low sequencing depth (less than 500 sequences).

138

Table 4.4: Septic ICU patient admissions and outcome data for the three groups of SB.

Group 1 Group 2 Group 3 N=13 N=24 N=13 Sex Female 46.2% 54.2% 38.5% Male 53.8% 45.8% 61.5% Age (years) Median 63 57 63 IQR (53-70) (52.75-69.5) (53-68) APACHE II Median 26 21.5 26 IQR (19-28) (16-28) (22-29) SOFA-Day 1 Median 7 7 10 IQR (7-10) (4-10) (9-12) SOFA-Max Median 8 10 12 IQR (7-12) (6.5-12) (9-12) ICU LOS (days) Median 5 6 6 IQR (4-7) (4-10.25) (4-16) ICU Outcome Alive 76.9% 95.8% 84.6% Dead 23.1% 4.2% 15.4%

139

The literature states that certain thresholds of sequencing depth are required to accurately describe the α-diversity (within a sample) or β-diversity (between samples) patterns [185]. In order to see if these samples would fall into the same cluster groups that were defined with good sequence depth samples, the β-diversity of all samples was compared (Figure 4.2).

Principal coordinates analysis (PCoA) was used to visualize where the low-sequence depth samples clustered in relation with the SB samples clustered in Figure 4.1. This was done using weighted UniFrac and using the cluster groups identified in Figure 4.1A to color-code the high- sequence depth samples. Overall, the PCoA plot demonstrated that when the SB samples with low sequencing depth were plotted alongside the clustered SB samples, the ordination of the samples did not change (Figure 4.2). As such, this indicated that despite using lower sequencing depth than suggested [139] and having lower jackknife support values that commonly supported [137], the ordination of the SB samples remained the same indicating that the analysis was still robust. Further, the data suggested that similar bacterial DNA profiles might also be present in the lower sequencing depth samples (Figure 4.2). Despite this, they were not subsequently analyzed as the statistical support for their grouping was weak.

4.2.3 Bacterial DNA Profiles of Blood from Healthy Adults

After determining that the SB blood samples had bacterial DNA patterns that clustered into three groups, the next step was to examine healthy blood (HB) samples. The rationale was to determine if there was bacterial DNA in HB or determine if there was a source of contamination in the processing of whole blood and subsequent Illumina sequencing.

140

Figure 4.2: PCoA of SB samples that had low sequencing depth indicate they cluster mainly with the Group 2 samples. Principal coordinates analysis, based on weighted UniFrac was done for all Day 1 SB samples from the ICU (n=116). Of these samples, 54 were used to distinguish SB into three clusters; Group 1A (orange) and Group 1B (green); Group 2AI

(purple), Group 2AII (yellow), Group 2AIII (light blue), Group 2AIV (turquoise), and Group

2B (pink); and Group 3A (grey), and Group 3B (brown). The remaining 62 samples (dark blue) were overlapped with the cluster groups. Circles were added to visualize the area in the PCoA plot that each cluster group isolates occupied. Majority of the low sequence depth samples had bacterial DNA profile profiles similar to those in SB Group 2 with a limited number showing similarity to Group 1 (n=11) or Group 3 (n=8).

141

1B 3B

2AII 2AIII 2B 3A

2AI 2A IV

142

In order to parallel the sepsis population cohort, whole blood from 12 healthy adults was subjected to the same DNA extraction and sequencing protocols. The age range was 28 to 73 years of age and the median age was 46. Of these 12 HB samples, one failed to amplify in the initial PCR and one had less than 50 sequences amplified and was thereby removed from the analysis. The remaining 10 samples had a minimum sequencing depth of 772 sequences per sample and a maximum of 33,133 sequences per sample with a median of 6174 sequences per sample. A total of 285,067 sequences were identified representing 519 OTUs representing 105 taxonomic groups. The healthy control samples had a range of non-bacterial likely human

DNA (“noRoot” OTU) amplification ranging from 98.6% to 7.0% (Figure 4.3A). This was comparable to the range in SB of 99.98% to 0.007% (Table 4.2).

In addition, PCR was performed on all the reagents and buffers used in the saponin blood-treatment and the DNA extraction. Of these, only PBS gave a positive signal and was included in the analysis as well as two negative template controls (NTC).

4.2.3.1 OTU Diversity in Healthy Blood Samples

Interestingly, all the healthy individuals had similar taxonomic profiles (Figure 4.3C).

In all samples, the top OTUs were Enterobacteriaceae, Staphylococcus, and Escherichia

(Figure 4.3B) representing 54.7% up to 96.3% of the OTU diversity in each sample. Lower levels of Streptococcus DNA were also present and ranged from 0.32% to 4.17% (Figure

4.3B). Majority of the remaining OTUs did not represent human-associated bacteria.

Rhodanobacter species have been identified in soil samples and sewage sludge [186].

143

Figure 4.3: The bacteria DNA profiles of healthy blood. Whole blood was collected from 12 adult donors that worked in a health care setting but were healthy at the time of sampling. Two negative template controls (NTC) and sterile PBS were included for comparison. Taxonomic summaries for the blood samples were compared. Each bar represents the total DNA sequenced for the sample and the relative abundance of each OTU identified. The proportion of

“noRoot” non-bacterial DNA (black) and bacterial DNA (grey) varied in each HB sample (A).

The bacterial DNA profile profiles of the HB samples were similar to each other but distinct from the NTCs and PBS samples (B). When the HB samples were analyzed with the clinical samples, the bacterial DNA profile profiles including a Klebsiella OTU that was not identified in the samples when they were analyzed alone (C). The letter in front of each taxonomic group indicates the level of taxonomic depth with p__ representing phyla, f __representing family, o__ representing order, and g__ representing genus.

144

A 100%

90%

80%

70%

60%

50% Bacterial DNA noRoot 40%

30%

20%

10%

0% HB1 HB2 HB3 HB4 HB5 HB6 HB7 HB8 HB9 HB10

145 B

100% f__Xanthomonadaceae

p__Proteobacteria 90%

o__Sphingomonadales

80% g__Janthinobacterium

f__Lachnospiraceae 70% g__Streptococcus

g__Brevibacillus 60% g__Sphingomonas

50% g__Rhodanobacter

g__Cupriavidus 40% g__Lactobacillus

g__Schlegelella 30% g__Methylobacterium

20% g__Escherichia

g__Staphylococcus

10% f__Bradyrhizobiaceae

f__Enterobacteriaceae 0%

1 2 3 4 5 6 7 8 9 0 O T S B B B B B B B B B 1 2 N B H H H H H H H H H B H O P H C G E N

146 C

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0% HB1 HB2 HB3 HB4 HB5 HB6 HB7 HB8 HB9 HB10

f__Enterobacteriaceae g__Staphylococcus g__Klebsiella g__Escherichia p__Proteobacteria g__Brevibacillus g__Streptococcus f__Xanthomonadaceae g__Pseudomonas o__Actinomycetales f__Lachnospiraceae g__Lactobacillus g__Lactococcus o__Streptophyta g__Paenibacillus g__Prevotella f__Bacillaceae c__Gammaproteobacteria f__ClostridialesFamilyXI.IncertaeSedis c__Betaproteobacteria Other

147

Schlegelella species are moderate thermophiles and are recovered from sludge or hot springs

[187]. Cupravidus has been identified as contamination in another 16S next-generation sequencing study and is also an industrial-associated organism [122].

The NTC and PBS samples had bacterial DNA profiles that were almost identical to each other yet distinct from the HB samples. The predominant OTU in these samples was

Bradyrhizobiaceae representing 60-70% of the relative DNA abundance (Figure 4.3C). This

OTU was also present in the HB samples but at very low levels (Figure 4.3C).

Janthinobacterium DNA was present as well, which is a major constituent of the human skin microbiota [188]. The remaining OTUs were not representative of human-associated bacteria.

This included Methylobacterium that is recovered from various environmental samples [189] as well as Sphingomonas that are ubiquitous organisms found in many environmental samples [190] and had previously been recovered as a contaminant during this study. This OTU was subsequently filtered from the samples (Section 2.9.5.1). These data indicated that the bacterial

DNA present in the NTC/PBS samples likely reflected the contamination from the laboratory environment as well as contamination from skin, yet all of these OTUs were minor components of the HB bacterial DNA profiles.

Interestingly, when the HB samples were analyzed with the SB clinical samples, there was a Klebsiella OTU identified (Figure 4.3C). In chapter three, the Enterobacteriaceae OTU from the synthetic communities matched to E. hormaechei as well as Klebsiella (Table 3.3).

Since there is a high degree of 16S rRNA similarity between Klebsiella and Enterobacter species

[191,192], there may have been the generation of a consensus sequence using AbundantOTU+ that the RDP classifier could taxonomically identify to the genus Klebsiella rather than

148

Enterobacteriaceae when the DNA from clinical samples were included in the downstream analysis and OTU generation.

4.2.3.2 The HB Samples Clustered Separately from the SB Samples

The HB bacterial DNA profiles were compared to the SB samples in order to determine if the bacterial DNA patterns present in HB samples were identified in SB samples. To examine differences in the bacterial DNA composition, phylogenetics using weighted unifrac metrics were used to examine changes in the relative taxon abundance between the HB and SB samples

[136,137,193]. PCoA was used to visualize the phylogenetic relationships in the samples and indicated that the HB samples clustered separately from the SB samples (Figure 4.4A). Jackknife resampling of 10 tables subsampled to a depth of 500 was done to ensure clustering was consistent with the aggregate results shown as confidence ellipsoids [137]. The HB samples still clustered separately from the SB samples in this analysis as there was no overlap in the ellipsoids surrounding each sample (Figure 4.4B). In order to determine if the clustering of the SB and HB samples was statistically significant, permutational analysis of variance (PERMANOVA) was done. This test was chosen, as it assumes no distribution and allows for the comparison of categorical factors such as sample type [139]. The PERMANOVA analyses indicated that the SB and HB groups were significantly different from each other (p = 0.001) thereby supporting the clustering seen in the PCoA.

149

Figure 4.4: The bacterial DNA profiles of healthy adults blood were distinct from adult septic patient’s blood. Principal coordinates analysis (PCoA), based on weighted UniFrac, indicated the healthy blood samples (red) clustered separately from septic blood samples (green)

(A). Since variation in the number of sequences per sample can impact diversity matrices [137], jackknife resampling using the same number of sequences per sample was done with weighted

UniFrac distances calculated using 500 sequences per sample. The clustering of the HB (green) from the SB (yellow) samples was conserved over repeated sampling with the aggregate results shown as confidence ellipsoids around each sample point (B).

150

PC2 (23%)

PC3 (7.5%) PC1 (33%)

151

4.2.3.3 Clustering of SB and HB Samples

In order to visualize the phylogenetic differences between the HB and SB samples, the taxonomic profiles of the UPGMA tree generated using weighted UniFrac and jackknife values were calculated for 10 tables subsampled to a depth of 500 were compared (Figure 4.5). The HB samples all clustered together in the UPGMA tree with no SB samples found on the same branch

(Figure 4.5). In terms of the SB clusters, the addition of the HB samples did not impact the tree structure as the distribution of all three SB clusters was preserved (Figure 4.5). The Group 1A and 1B isolates all still clustered into the same distribution (Figure 4.5). The Group 3A and 3B isolates were similarly unaffected by the clustering with HB. For Group 2, all of the clusters remained intact in terms of the distribution of samples and the branching within the tree. The

Group 2AI, 2AII, 2AIV, and 2B groups were unaffected by the HB samples. The Group 2AIII cluster was disrupted by the HB samples (Figure 4.5). The HB samples were similar to the Group

2AIII samples due to the abundance of the Staphylococcus, Klebsiella, Escherichia, and

Enterobacteriaceae DNA (Figure 4.5). However, the difference between the Group 2AIII samples and HB samples was the presence of bacterial DNA representing Fusobacterium,

Neisseria, and Anaerococcus in the Group 2AIII samples (Figure 4.5). Interestingly, the patients in Group 2AIII were all clinical diagnostic blood culture negative with the one exception of a patient, ASN340, that was cultured with Group C Streptococcus (Table 4.2). Taken together, these data suggested that the bacterial DNA recovered in the HB samples likely resulted from the exquisite sensitivity of this technique that had been developed to amplify low-abundance DNA sequences.

152

Figure 4.5: HB Samples cluster separately from Group 1 and Group 3 SB samples but are found within Group 2 SB samples. A composite unweighted pair group method with arithmetic mean (UPGMA) phylogeny of all the samples was generated with from the jackknife, weighted

UniFrac β-diversity comparison of the SB and HB. The SB samples were labeled based on the cluster groups identified in Figure 12. The SB cluster group’s 1A, 1B, 2AI, 2AII, 2AIV, 3A, and

3B remained intact when clustered with HB. The HB cluster divided the Group 2AIII cluster. In terms of the SB clusters, the addition of the HB samples did not impact the tree structure as the distribution of all three SB clusters was preserved.

153

Streptococcus Staphylococcus Enterobacteriaceae Escherichia Gammaproteobacteria Pseudomonas Proteobacteria Fusobacterium Xanthomonadaceae Klebsiella Bacillaceae Moraxella Legionella Prevotella Finegoldia Fusobacterium

154

The uniformity between the HB samples combined with the limited OTU similarity between these samples and the NTC/PBS controls (Figure 4.3) suggested that the DNA present in these samples was likely obtained during the sample collection procedure, all done with venipuncture, as the DNA profiles could not be solely attributed to PCR-based contamination.

4.2.4 Changes to Bacterial DNA Patterns during ICU Admission

ICU patients with sepsis are on aggressive intravenous antibiotic therapy [36], which is expected to impact the patient’s microbiota (commensals and pathogens) and subsequently was predicted to change the bacterial molecular profiles in the bloodstream. In order to determine if the bacterial DNA patterns within a septic patient changed during ICU stay, time course samples of patients admitted to the ICU were collected for a subset of the SB patients on days 1, 7, 14, and 28. The taxonomic profiles on each day were compared for a patient. There were four patients with a day 1,7,14, and 28-blood sample (Figure 4.6).

4.2.4.1 ASN464 Case Study

ASN464 was a 62-year-old male admitted with a gastrointestinal rupture resulting in sepsis. Antibiotic therapy administered on days 1-7 was metronidazole (Table 4.5), which would provide coverage for predicted anaerobic gastric microbiota that have leaked into the abdominal cavity following the rupture [194].

155

Figure 4.6: The bacterial DNA profile in blood changes during a patients stay in the ICU.

Taxonomic bacterial DNA profiles were summarized for whole blood samples collected on Day

1, 7, 14 and 28 of the patients ICU stay. The relative abundance of each bacterial taxa are colour coded by genus when possible, by family for Enterobacteriaceae and Xanthomonadaceae, class

Gammaproteobacteria, and the phyla Proteobacteria. Each panel (A=ASN464,B=ASN465,

C=ASN469, D=ASN475) represented the taxa distribution for each ICU patient.

156

A B 100% 100%

90% 90%

80% 80%

70% 70%

60% 60%

50% 50%

40% 40%

30% 30%

20% 20%

10% 10%

0% 0% ASN464.1 ASN464.7 ASN464.14 ASN464.28 ASN465.1 ASN465.7 ASN465.14 ASN465.28 C D 100% 100%

90% 90%

80% 80%

70% 70%

60% 60%

50% 50%

40% 40%

30% 30%

20% 20%

10% 10%

0% 0% ASN469.1 ASN469.7 ASN469.14 ASN469.28 ASN475.1 ASN475.7 ASN475.14 ASN475.28 Streptococcus Staphylococcus Enterobacteriaceae Escherichia Gammaproteobacteria Pseudomonas Enterococcus Proteobacteria Xanthomonadaceae Klebsiella Bacillaceae Moraxella Enterococcus Brevibacillus

157

No clinical diagnostic blood culture data was reported on day 1 but on day 14, the clinical diagnostic blood culture was positive for CoNS. The bacterial DNA profile for ASN464 indicated Staphylococcus DNA was present in the blood with a steady reduction in the

Staphylococcus DNA abundance from day 1 to 28 (Figure 4.6A). Vancomycin therapy was administered from days 13 to 21 and day 28 onwards (Table 4.5), which likely contributed to the decline of the Staphylococcus. Conversely, the Streptococcus DNA relative abundance increased from day 1 to 14 then decreased on day 28 (Figure 4.6A). This was surprising as the patient was also receiving piperacillin/tazobactam on days 13-16, and 22-onwards which should have good broad-spectrum coverage [195] (Table 4.5). The relative abundance of Gram-negative

DNA including Enterobacteriaceae, Escherichia, and Gammaproteobacteria was the highest on day 14. On day 28 the bacterial DNA taxonomic profile had several groups identified that were not present on days 1, 7 or 14 including Enterococcus, Lactococcus, Moraxella, Fusobacterium,

Pasturella, Clostridiales, and Bacillaceae (Figure 4.6A). These data suggested a relative increase in bacterial DNA diversity over time, which was consistent with a decrease in the DNA that could be associated to potential pathogens including Staphylococcus and Streptococcus. Further, on day 28 the bacterial DNA profile had several OTUs identified that were not present on days 1,

7 or 14 including the genera Enterococcus, Lactococcus, Moraxella, Fusobacterium, Pasturella,

Clostridiales, and Bacillaceae family (Figure 4.6A). Interestingly, these organisms are associated with gastrointestinal ruptures [196]. This patient was admitted for such an infection yet by day

28 the presence of such OTUs in the molecular profile would not be anticipated if therapy was successful. These results suggested that perhaps there was still leakage of gastrointestinal contents into the bloodstream even on day 28.

158

Table 4.5: Antibiotics administered from day 1-28 for SB case studies.

Patient Antibiotic Time Range (day 1-28) ASN464 Ceftriaxone 1 Levofloxacin 11-13 Metronidazole 1-7 Piperacillin/tazobactam 13-16, 22-28+ Vancomycin 13-21, 28+ ASN465 Metronidazole 27-28+ Piperacillin/tazobactam 1-2, 8-12 Sulfamethoxazole/trimethoprim 1-28+ Vancomycin 1-6, 17, 20 Linezolid 5-12 ASN469 Clindamycin 1-5 Metronidazole 5-28+ Piperacillin/tazobactam 1-28+ ASN475 Ciprofloxacin 2-13 Piperacillin/tazobactam 1-28+ Vancomycin 2-6, 13-25

159

4.2.4.2 ASN465 Case Study

ASN465 was a 70-year-old female patient admitted with cardiac arrest in addition to a urinary tract related sepsis. Clinical diagnostic blood culture was negative, however, the urine culture was positive for VRE on day 1 yet these results were only received 5 days following admission to ICU. As such, the patient was on vancomycin therapy from days 1-6, 17, and 20 with linezolid therapy only introduced on day 5 to treat the VRE infection (Table 4.5). A positive blood culture was obtained on day 7 in which P. aeruginosa was recovered. However, the molecular profiling indicated that Pseudomonas DNA was present on all days and was the most abundant on day 14 (Figure 4.6B). As such, this patient may have had a Pseudomonas bloodstream infection, which was undetected using conventional clinical diagnostics until 7 days into the patients ICU admission. There was no apparent change in the antibiotic therapy to reflect the positive blood culture as the patient was on sulfamethoxazole/trimethoprim from days 1 to 28 and piperacillin/tazobactam on days 1 to 2 and 8 to 12 (Table 4.5). ASN465 also had a consistent reduction in Streptococcus DNA from day 1 to 14 in which the relative abundance dropped from

58% to 4%. However, from day 14 to 28 the relative abundance of Streptococcus DNA increased back to 16% despite the patient being on two broad-spectrum antibiotics (Figure 4.6B). The abundance of Staphylococcus stayed fairly consistent from days 1 to 28 as it represented between

10-15% of the relative DNA abundance despite vancomycin treatment (Figure 4.6B). The prevalence of the Pseudomonas DNA on day 14 despite the patient’s antibiotic therapy suggested there was a possible source of Pseudomonas that resulted in persistent detection of this

DNA in the bloodstream. Possible sources of Pseudomonas ICU infections include ventilators and catheters [15,197]. The molecular profile data suggested that such a cycle of shedding and clearance in the bloodstream could have occurred in this patient since the Pseudomonas OTU

160

was intermittently recovered in the blood. In summary, there was once again an observable increase in bacterial DNA diversity overtime in this patient that appeared to correlate with the decrease in Streptococcus and Pseudomonas DNA.

4.2.4.3 ASN469 Case Study

ASN469 was a 60-year-old male patient admitted with sepsis from a cellulitis infection.

The clinical diagnostic blood culture was positive on day 1 for Fusobacterium and

Campylobacter ureolyticus. Antibiotic therapy included clindamycin on days 1 to 5, piperacillin/tazobactam from days 1 to 28, and metronidazole from days 5 to 28 (Table 4.5).

Culture on the infected tissue indicated both Enterococcus and E. coli were present. The day 1 blood molecular profile did not indicate there was Campylobacter DNA present at detectable levels and Fusobacterium represented only 0.33% of the relative DNA abundance suggesting the antibiotic treatment resolved the bloodstream infection (Figure 4.6C). Interestingly,

Streptococcus, Escherichia, and Enterobacteriaceae DNA were all prevalent in the day 1 sample

(Figure 4.6C). This suggested a possible translocation of E. coli from the cellulitis infection into the bloodstream on day 1. In contrast, Enterococcus DNA had the highest relative abundance on days 7-28 perhaps suggesting this organism, also identified in the cellulitis infection, translocated into the bloodstream but at a later stage in the infection (Figure 4.6C). Interestingly, there did not appear to be an increase in bacterial DNA diversity as the infection was treated. This was consistent with the predicted shedding of organisms or their products from the cellulitis infection into the bloodstream at later stages.

161

4.2.4.4 ASN475 Case Study

ASN475 was a 77-year-old female patient with a gastrointestinal abscess infection with

P. aeruginosa cultured from the blood and abscess drainage fluid on day 1. This patient was treated with piperacillin/tazobactam from days 1 to 28, ciprofloxacin from days 2 to 13, and vancomycin from days 2 to 6 and 13 to 25 (Table 4.5). On day 1 the Pseudomonas OTU was detected in the bloodstream molecular profile, which correlated with the blood culture results

(Figure 4.6D). On days 7 and 14 the abundance of the Staphylococcus and family Bacillaceae

OTUs increased whereas the Pseudomonas OTU was virtually undetected (Figure 4.6D). This suggested the antibiotic therapy was effective at resolving the P. aeruginosa infection whereas

Staphylococcus DNA was abundant despite the addition of vancomycin therapy on days 2-6

(Figure 4.6D). Interestingly, on day 28, the Pseudomonas OTU was once again detected. This suggested there was a temporary resolution of the bloodstream infection attributed to P. aeruginosa and possible re-infection in the bloodstream based on the changes in the

Pseudomonas DNA relative abundance in the blood during the ICU admission.

Overall, the phylogenetic relationships of the molecular profiles over each subsequent day were distinct suggesting the bacterial DNA profile changed in each patient during their ICU treatment. In each case study, shifts in relative DNA abundance may have been partially attributed to their antibiotic treatment. Nevertheless, despite aggressive broad-spectrum antibiotic therapy, bacterial DNA was still present in these patients up to 28 days post-admission with some taxonomic groups correlating to clinical diagnostic blood culture results.

162

4.2.5 Septic ICU Patient’s Primary Infection Samples

Sepsis bloodstream infections are often acquired as a secondary infection from another primary insult including respiratory tract infections, intra-abdominal infections, urinary tract infections, and closed-space abscess infections [4,198]. As such, in order to examine how the primary infection site bacterial DNA profiles compared to the sample patients SB bacterial DNA profile, primary infection fluid samples were obtained from a subset of septic ICU patients.

These samples were processed with the same DNA extraction method, the 16S rRNA V3 region was PCR amplified, and then sequenced using paired-end Illumina sequencing. The molecular profile from the patient’s primary infection sample was compared to SB sample obtained at the same time point and processed the same way. In addition, culturing of the primary infection sample was done for a subset of samples using agar-plate based methods and the resulting colonies were identified using partial 16S rRNA PCR (Table 4.6).

4.2.5.1 In-depth Culturing of Sepsis Primary Infections

The results of the culturing were compared to the clinical diagnostic laboratory culture results from a sample obtained from the same site within ± 24 hours of our sample collection

(Table 4.6). There were 21 primary infection samples collected from septic ICU patients. The airway samples were obtained from bronchoalveolar lavage fluid (BAL), sputum (SP), and endotracheal tube fluid (ETT). Intra-abdominal samples were obtained from chest tube (CT) fluid, peritoneal fluid (PF) and Jackson-Pratt fluid (JP).

163

Table 4.6: Culture results from sepsis patient’s primary infections.

Patient Admission Sample Type Primary Infection Clinical Laboratory Diagnosis Culture* Culture 165CT Bacterial Chest tube fluid Streptococcus pneumoniae Streptococcus pneumoniae pneumonia 167AF Hepatic Abscess Fluid Streptococcus intermedius, Streptococcus milleri group abscess Streptococcus anginosus 168BAL Bacterial Bronchoalveolar No Growth Candida pneumonia lavage fluid 186BAL Pulmonary Bronchoalveolar Staphylococcus Coagulase-negative contusion lavage fluid epidermidis, Streptococcus Staphylococcus salivarius 201JP Traumatic Jackson-Pratt No Growth Not done rupture or drainage fluid laceration of liver 207JP Duodenal Jackson-Pratt Escherichia coli, Candida perforation drainage fluid Streptococcus salivarius, Staphylococcus epidermidis 290CT Empyema Chest tube fluid No Growth Not done 328ETT Self-poisoning Endotracheal Staphylococcus aureus, Staphylococcus aureus, with narcotics tube fluid Lactobacillus salivarius, Streptococcus pneumoniae Lactobacillus rhamnosus, Lactobacillus fermentum, Lactobacillus oris, Lactobacillus vaginalis, Streptococcus intermedius, Streptococcus constellatus 330PF Septic shock Peritoneal fluid Streptococcus anginosus Not done 332SP Bacterial Sputum Not done Not done pneumonia 344PF Large bowel Peritoneal fluid Enterobacter Citrobacter sp., perforation or cancerogenus, Escherichia Streptococcus milleri rupture coli group, Pseudomonas aeruginosa, Group B Streptococcus, Streptococcus viridans group, Bacteroides fragilis 348JP Septic shock Jackson-Pratt Not done Micrococcus sp., drainage fluid Streptococcus viridans group 350BAL Bacterial Bronchoalveolar No Growth Not done pneumonia lavage fluid 400AF Bacterial Abscess Fluid Escherichia coli Not done pneumonia

164

403PF Bacterial Peritoneal fluid Not done Staphylococcus aureus pneumonia 404UR Renal infection Urine Not done Not done 407UR Urinary tract Urine Not done Pseudomonas aeruginosa infection 408PF Abdominal Peritoneal fluid Not done Not done trauma 417JP Gastrointestinal Jackson-Pratt Not done Aeromonas sp., Klebsiella bleeding drainage fluid pneumoniae 432SP Bacterial Sputum Streptococcus pyogenes Group A Streptococcus pneumonia 438PF Pneumonia Peritoneal fluid Not done Not done *As outlined in Section 2.4.2

165

Urine (UR) represented urinary tract infection samples and abscess fluid (AF) was obtained from abscess infections. Of the 21 samples, in-depth culture was done on 14. The in-depth culture was able to recover bacteria from the primary infection sample in 10 of the 14 samples

(Table 4.6). Overall, the in-depth culture had good concordance with the clinical laboratory culture with the bacteria reported from the clinical culture also recovered in the in-depth culture

(Table 4.6). The exception was ASN344PF where the in-depth culture had no similarity to the clinical diagnostic lab culture (Table 4.6). The in-depth culture often recovered more bacterial species than what was reported in the clinical diagnostic lab reports (Table 4.6). In addition, the identification of the bacteria using 16S rRNA sequencing (Section 2.5) gave a species identification of the organism recovered for ASN167AF, ASN186BAL, and ASN432SP whereas the clinical diagnostic culture only indicated the bacterial group (Table 4.6). Taken together, the in-depth culture of the primary infection site fluid samples provided a species level taxonomic identification of bacteria recovered in the infection, provided information on the polymicrobial infections present, and provided information that could be used for the interpretation of the bacterial DNA profiles recovered from these samples.

4.2.5.2 Correlation Between Primary Infection and Whole Blood Molecular Profiles

In the previous chapter, three case studies (ASN167, 168, and 328) were highlighted as part of the methodology development (Figure 3.11). In all three cases, there was at least one taxonomic group identified in the primary infection site that correlated with the patient’s clinical presentation and was also present in the whole blood molecular profile (Figure 3.11). This suggested that the bacterial DNA profile in the primary site of infection accounted for a portion of the bacterial DNA recovered in the whole blood samples. In order to determine if this was

166

common to all sepsis patients in this study, the molecular profiles for whole blood and primary infection samples for 17 additional patients were compared. Of these, five samples (ASN168,

186, 201, 207, and 290) had very low sequencing depths limiting our confidence in molecular profiles and were subsequently removed from the analysis.

The patients who had lower airway infections including ASN165, ASN328, ASN332,

ASN350, and ASN432 had Streptococcus DNA present in the airway sample molecular profile as well as in the whole blood molecular profile (Figure 4.7). Further, for the majority of these patients, the in-depth culturing also recovered a Streptococcus species (Table 4.6) supporting the presence of the DNA in the molecular profile. As such, patients with respiratory infections seemed to be more likely to have Streptococcus bacteria or their products present in the bloodstream.

The patients with a JP or PF sample suggested there was an intra-abdominal infection resulting from trauma to the gastrointestinal tract. In these patients, the molecular profiling of the primary infection indicated gastric microbiota including Fusobacterium, Bacteroides, Prevotella,

Peptostreptococcus, Enterobacteriaceae, Escherichia, Klebsiella, and Anaerococcus were likely present in the peritoneal space (Figure 4.7). In addition, Staphylococcus DNA was often present in the JP and PF samples (Figure 4.7). Although not considered as part of the enteric microbiota,

S. aureus have been documented post-operative gastric wound infections including duodenal perforation [196,199]. Overall, there was good correlation between the PF/JP fluid molecular profiles and clinical diagnostic laboratory cultures for ASN330, ASN348, ASN344, ASN408, and ASN438 (Figure 4.7). Despite this, the bloodstream molecular profile suggested that these organisms were not always present in the bloodstream infection (Figure 4.7).

167

Figure 4.7: Molecular profiling of primary infection samples and whole blood from septic

ICU patients. Taxonomic bacterial DNA profiles were summarized for whole blood samples and the patients corresponding primary infection sample. Primary infection samples included intra-abdominal samples from peritoneal drainage fluid (PF), abscess drainage fluid (AF), and

Jackson-Pratt drainage fluid (JP). Lower respiratory infection samples included chest tube fluid

(CT), bronchoalveolar lavage fluid (BAL), expectorated sputum (SP), and endotracheal tube fluid (ETT). Urogenital tract infections were represented by urine samples (UR).

The relative abundance of each bacterial taxon was colour coded by genus when possible; by family for Enterobacteriaceae, Xanthomonadaceae, and Lachnospiraceae; by class for Gammaproteobacteria; and by phyla for Proteobacteria.

168

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

3 3 F 7 T 8 F 0 P 2 F 1 P 1 L 0 F 3 4 7 F 8 P 7 P 2 F 8 . . 6 2 P 3 S 3 P . J . 5 P 0 R 0 R 0 P 0 J 1 S 3 P 3 T 5 A 1 T 3 0 3 2 3 4 4 8 8 A 3 3 4 U 4 U 4 8 4 . 4 . 4 . 4 C 6 7 E 3 3 4 4 4 4 B 0 4 7 0 7 2 8 5 1 6 N 8 N 3 N 3 N 3 3 3 3 0 N 4 N 0 N 0 N 4 N 41 N 3 N 3 N 6 N 1 S 2 S S S N N N 5 S S 4 S 4 S S S 4 S 4 S 1 S N A 3 A N A N A N S S S 3 A N A N A N A N A N A N A N A N A S N S S S A A N S S S S S S S S A S A A A A S A A A A A A A A A A Streptococcus Bacteroides Staphylococcus Lachnospiraceae Enterobacteriaceae Veillonella Prevotella Escherichia Fusobacterium Gammaproteobacteria Lactococcus Pseudomonas Enterococcus Propionibacterium Proteobacteria Neisseria Xanthomonadaceae Aeromonas Klebsiella Actinomycetales Anaerococcus Bacillaceae Finegoldia Legionella

169

Indeed, for ASN330, 348, and 344 the only common taxonomic group between the PF and whole blood was Streptococcus (Figure 4.7). The clinical diagnostic culture/in-depth culture of the

PF/JP fluid did recover Streptococcus species in these patients (Table 4.5). However, for both

ASN408 and ASN438, the Streptococcus, Staphylococcus, Escherichia, Enterobacteriaceae, and

Klebsiella OTUs were present in both the PF and whole blood molecular profiles.

There were also instances where the primary infection JP/PF sample did not correlate with the whole blood based on molecular profiling. ASN403 was unique in that the molecular profile of the PF indicated Staphylococcus DNA was dominant (Figure 4.7). The clinical culture for this patient recovered S. aureus in the peritoneal fluid, which correlated to the molecular profiling data (Table 4.6, Figure 4.7). Despite this, the Staphylococcus OTU was not highly abundant in the bloodstream molecular profile whereas the Streptococcus OTU represented 71% of the DNA abundance (Figure 4.7). However, this patient had a large-bowel perforation or rupture, which often results in translocation of gut flora into the peritoneum [200]. In this type of intra-abdominal sepsis, there is an observed simplification of the infection from an abundance of enteric Gram-negative bacilli, Bacteroides, Streptococcus, and other gastric bacteria to the formation of intra-abdominal abscesses dominated by few anaerobic species such as Bacteroides and Fusobacterium [200]. As such, the variation in the in-depth culture and clinical laboratory culture may have been a result in a delay between the collection time for the initial clinical diagnostic laboratory culture and the sample collection for the in-depth culture. For ASN417 the molecular profile of the PF correlated to the clinical laboratory culture since the Klebsiella and

Aeromonas OTUs represented majority of the DNA abundance and the clinical culture recovered both Aeromonas species and K. pneumoniae (Table 4.6, Figure 4.7). Despite this correlation, the bloodstream molecular profile indicated Staphylococcus DNA represented 64% of the DNA

170

abundance whereas the PF bacteria were undetected (Figure 4.7). Taken together, these results suggested that the patients an intra-abdominal infection had gastric microbiota present in the

PF/JP fluid and although they were more likely to have bacterial DNA from such organisms present in the bloodstream, not all organisms could be identified in the bloodstream molecular profile.

The urinary tract associated sepsis infections (ASN404 and ASN407), had little correlation between the bloodstream molecular profile and the whole blood molecular profile

(Figure 4.7). For ASN404, Escherichia represented 89% of the DNA abundance in the urine sample yet only represented 5.7% of the DNA abundance in the whole blood profile (Figure 4.7).

For ASN407, several OTUs representing anaerobic bacteria including Fusobacterium,

Bacteroides, Prevotella, Peptoniphilus, Anaerococcus, and Finegoldia were detected in the urine molecular profile (Figure 4.7) whereas the clinical laboratory cultured P. aeruginosa (Table 4.6).

The only OTU shared between the urine and bloodstream samples was Streptococcus, which was a minor component of the molecular profile for both samples (Figure 4.7). As such, the strongest correlations between a primary infection and whole blood sample were in patients with respiratory infections whereas patients with intra-abdominal infections had some correlations and those with urinary tract infections having no correlations between the primary site of infection and the bloodstream infection using molecular profiling.

4.2.5.3 Taxonomic Diversity in Primary Infection Samples

In the comparison of the primary infections and whole blood samples from each patient, it appeared that there were differences in the amount taxonomic diversity within each sample.

171

Indeed, the primary infection samples appeared to have fewer taxonomic groups than the corresponding blood sample from the same patient (Figure 4.7). In order to examine if these differences between the patient’s whole blood and the primary infection samples were significant, the α-diversity of the samples were compared. The Shannon Diversity index was used to measure the community structure within each sample in term of the richness and evenness. The observed species estimate used the number of OTUs in the sample to measure the within sample diversity.

For most patients, both the observed species and Shannon index indicated that there was a statistically significant difference in the α-diversity between the whole blood and primary infection sample obtained from a patient (Table 4.7). Patients ASN320, ASN330, and ASN350 were the exceptions in which one α-diversity measure had a statistically significant different whereas the other did not (Table 4.7). The samples in which the observed species was not statistically significant but the Shannon index was indicated that the difference in the α-diversity was due to differences evenness in the bacterial DNA profile. Conversely, when the Shannon index showed a statistically significant difference but the observed species did not, it suggested that the number of OTUs were similar between the samples but that the distribution of OTUs was different. Overall, the α-diversity of the primary infection sample was statistically lower than that of the corresponding blood sample obtained from the same patient (Table 4.7). These data indicated that primary infection samples often had reduced bacterial diversity when compared to blood samples.

172

Table 4.7: Comparison of the diversity with each sample from a patient’s primary infection and SB sample.

Sample Observed species Shannon diversity Significantly Different* Primary Whole Primary Whole Observed Species Shannon infection blood infection blood ASN165 1 21 0.01 3.27 Yes (P < 0.0001) Yes (P < 0.0001) ASN167 1 19 0 3.10 Yes (P < 0.0001) Yes (P < 0.0001) ASN328 21 34 3.35 4.16 Yes (P < 0.0001) 0.48 ASN330 31 33 4.18 4.24 0.06 Yes (P < 0.0001) ASN332 3 19 0.57 3.41 Yes (P < 0.0001) Yes (P < 0.0001) ASN344 27 21 3.86 3.06 Yes (P < 0.0001) Yes (P < 0.0001) ASN348 16 12 2.47 2.18 Yes (P < 0.0001) Yes (P=0.045) ASN350 21 20 3.6 2.83 0.3 Yes (P < 0.0001) ASN403 3 13 0.11 1.70 Yes (P < 0.0001) Yes (P < 0.0001) ASN408 27 13 3.78 1.67 Yes (P < 0.0001) Yes (P < 0.0001) ASN432 3 25 0.14 3.36 Yes (P < 0.0001) Yes (P < 0.0001) ASN438 24 18 3.27 3.02 Yes (P < 0.0001) Yes (P=0.007) *Student’s t-test based on a rarefied OTU table using the rarefaction of 108 sequences repeated 10 times

173

4.2.6 Discussion

This study sought to determine how the bacterial DNA recovered from whole blood could be used to further our understanding of sepsis bloodstream infections. This approach involved the application of paired-end Illumina 16S rRNA sequencing to taxonomically identify bacterial

DNA recovered from septic ICU patients’ blood samples and primary infection samples as well as blood from healthy adults.

The bacterial DNA amplified from HB samples was not unexpected since PCR is sensitive and the universal 16S primers allow for the amplification of any DNA from a bacterial source

[201]. Contamination from reagents and the environment (laboratory and hospital) is a common problem in PCR using universal bacterial gene probes. In addition, the 16S primers have documented cross-reactivity to human DNA, which accounted for the amplification of the non- bacterial “noRoot” OTU. As such, when applying this technology to blood, where the ratio of human DNA compared to bacterial DNA was quite high, there was an abundance of amplified non-16S human DNA. In addition, there was recovery of bacterial DNA from the HB samples that was represented predominantly by OTUs for Enterobacteriaceae, Escherichia, and

Staphylococcus (Figure 4.3). These OTUs were not recovered from the NTS or PBS control samples suggesting PCR reagents were likely not the source of this bacterial DNA (Figure 4.3).

The DNA present in these samples was interpreted as a mixture skin contamination and environment contamination as a result of the way in which the blood samples were collected.

These were peripheral blood draws into vacutainers. Since this was not done in a hospital setting, the same procedures used for blood culture to minimize skin contamination were not employed

[202]. Further, studies have indicated that antisepsis at the skin cannot completely prevent

174

contamination since 20% of the skin bacteria are located in deep layers of the skin or in structures that the surface antisepsis do not penetrate [202]. The skin was clean using a sterile isopropyl alcohol swab. These swabs are effective at killing bacteria but do not degrade bacterial

DNA [203]. As such, DNA from skin microbiota may have been introduced during the venipuncture of the skin. This could account for the abundance of Staphylococcus DNA recovered as the Staphylococcus are considered as part of the skin microbiota [204]. Although the Enterobacteriaceae are not commonly associated with the skin microbiota, a study assessing the hands of health care workers recovered Enterobacter species as well as other Gram-negative bacteria, not taxonomically identified, in addition to Staphylococcus species from individuals in which there was documented skin damage [205]. This skin damage was associated with the irritant action of repeated hand washing and the use of gloves by health care workers [205]. As such, since our samples were collected from health care workers by health care workers, it would not be unreasonable to attribute the Enterobacteriaceae and Escherichia DNA as skin contamination as well. Overall, the DNA representing human-associated taxonomic groups was likely obtained from the skin microbiota during the venipuncture.

The non-human associated OTUs could be from the vacutainers themselves. Indeed, numerous studies applying molecular profiling to diagnostic samples had indicated bacterial

DNA contamination was present in commercially produced products [173,206]. The BD

Diagnostics website indicated that their vacutainers are sterilized using gamma radiation which would kill bacteria but not destroy bacterial DNA [207]. As such, any remnant bacterial DNA present in these samples during their manufacturing may have contributed to the bacterial DNA profiles observed in the HB samples. Further, the BD Vacutainer® Luer-Lok™ Access Device

(BD Diagnostics) used to fill the vacutainers, were not kept in sterile packaging prior to use. As

175

such, the surface of the vacutainer puncturing device was exposed to environmental sources of bacterial DNA and likely contributed to the bacterial DNA recovered in the healthy samples.

Since many of the bacteria represented by these OTU were associated with industrial sources they may be remnants from the preparation of these vacutainers.

In addition, the ten HB samples had almost identical bacterial DNA profiles, which would also suggest the contaminant source was equally present in all patients (Figure 4.3). As well, the HB samples were significantly different from the SB samples in phenetic diversity based on weighted UniFrac (PERMANOVA, p = 0.001) (Figure 4.5). Indeed, these results also highlighted that the DNA extraction protocols were optimized to such a level that bacterial DNA likely from minute contaminants was also amplified. Attempts to quantitate the bacterial load in the HB and SB samples, using RT-PCR, were unsuccessful due to the cross-reactivity of the 16S primers to human DNA in these samples (data not shown). As such, the total abundance of the

HB bacterial DNA was unknown. Nevertheless, knowledge of this type of contamination in the whole blood samples indicated that caution was needed for the interpretation of the

Staphylococcus, Enterobacteriaceae, and Escherichia OTUs present in the SB sample’s molecular profiles. These OTUs were not interpreted as significant unless there was clinical evidence to support the presence of bacterial DNA represented by these OTUs in the patients.

Following the analysis of the HB samples, the bacterial DNA profile of whole blood from

54 septic ICU patients were identified and found to cluster into three groups on a UPGMA tree

(Figure 4.1A). The Group 1 samples were enriched with Streptococcus whereas Group 3 was enriched with Staphylococcus DNA (Figure 4.1B). The Group 2 samples were separated into 5 sub-groups in which the abundance of Gram-negative OTUs characterized the DNA profile

(Figure 4.1B). Of these 5 sub-groups, the Group 2AIII sample clustering was disrupted when the

176

HB samples were included in the phylogenetic analysis (Figure 4.5). These samples were distinguished from the HB by the prevalence of certain OTUs including Fusobacterium,

Neisseria, and Anaerococcus in the last three patients (Figure 4.5). As such, Group 2AIII samples had limited differences in their molecular profiles from the HB samples limiting the interpretation of their DNA profile. These patients perhaps had lower bacterial DNA abundance in the sample thereby increasing the relative abundance of the contaminants in the taxonomic profile. The prevalence of Gram-positive infections in sepsis is reported at 40% [208]. The

Group 1 and Group 3 clusters accounted for 48% of the patients thereby in-line with the published data [13,29,208]. It is important to note that the SB molecular profiles had bacterial

DNA identified from both Gram-positive and Gram-negative bacteria with the exception of

ASN363 (Table 4.2). As such, although the Group 1 and Group 3 samples implicated a Gram- positive organism in the systemic infection, the data also suggested there was the potential for polymicrobial infections based on the prevalence of DNA representing multiple bacteria whereas the clinical diagnostic blood culture only recovered organisms that were able to grow in blood culture media.

The clinical diagnostic blood culture results were positive for 30% of the patients examined in the SB cohort (Table 4.2) whereas all SB samples had molecular bacterial DNA profiles in blood that were distinct from the HB samples (Figure 4.1). Although the presence of bacterial DNA in the blood did not signify the presence of viable organisms, it did suggest that the blood cultures were under-representing the presence of bloodstream infections in this cohort.

Further, within each group there were instances in which the whole blood molecular profile correlated with the patients’ clinical presentation and their diagnostic culture results from blood or other sources. Indeed, for the Group 1 and Group 3 samples, these correlations combined with

177

the predominance of one taxonomic group is the bacterial DNA profile suggested that the patients with diagnostic blood culture negative samples likely were representative of systemic infections with Streptococcus (Group 1) or Staphylococcus (Group 3) bacteria or bacterial products from these organisms.

In addition, there were cases in which a cultivated organism was not recovered but the bacterial DNA profile indicated there was one organism representing 90% of the molecular profile (Figure 4.1). In these cases, the molecular profiling data could be used in context with the clinical suspicion of infection to help direct antimicrobial therapy. Indeed, for ASN479 there was an excellent correlation between the whole blood molecular profile in which Finegoldia was the principal OTU recovered and the patient’s clinical presentation. This patient had septic arthritis with F. magna infections having been documented in such patients despite being very rare [30].

Further, the recovery of F. magna in the patient’s drainage fluid combined with the molecular profile data supported a clinical suspicion of a F. magna bloodstream infection. As such, antibiotic therapy targeting this organism would have been predicted as beneficial for the patient.

Not all the SB samples had a sequencing depth that allowed for good interpretation of β- diversity [209]. In the blood samples the amount of bacterial DNA template was low as compared to the host template resulting in the high relative abundance of the “noRoot” OTU

(Table 4.2). Following the removal of the “noRoot” reads, the samples often had a low number of remaining sequences. A reasonable cut-off was needed to ensure that differences in the taxonomic structure of samples could be identified. The strength of UniFrac β-diversity to identify meaningful patterns in various datasets has been well documented [137]. Indeed, even in small sample size simulations (50 sequences) the UniFrac values could be used to discriminate between samples [137]. However, when the expected similarity in microbial communities among

178

different samples is anticipated to be high, more sequencing reads are required to identify relationships [137]. It is also known that between 500-1000 reads/sample is sufficient, but not ideal, to distinguish differences in phylogenetic composition between two samples using β- diversity. As such, although the communities were not sampled well at a depth of 500 reads, reflected in the low jackknife support values, this depth still enabled the separation of SB samples based on their community composition. This level was selected as it permitted evaluation of more of the SB samples with the knowledge that the interpretation of the profiles required caution in the absence of good clinical data. Indeed, even at this level the clustering only accounted for 54 of the 116 patients. In order to determine if the remaining patients had similar bacterial DNA profile profiles, PCoA was done with no sampling depth set. In this analysis, majority of the low sequence depth samples appeared to align with the three SB clusters with 12 outliers detected (Figure 4.2). This would suggest that the majority of whole blood samples had similar molecular profiles as were observed in the UPGMA clustering (Figure 4.1) despite lower sequencing depth. In summary, the sequencing depth of 500 sequences was used for the majority of this study and clinical data was used to corroborate the results obtained from the molecular profiles and support the UPGMA cluster groups observed.

For most samples used in molecular profiling studies, the proportion of non-bacterial

“noRoot” DNA is small or negligible (J. Stearns, personal communication). However, as the proportion of bacterial DNA decreased to a low proportion of the total DNA (i.e., samples in which majority of the DNA is host derived), the PCR amplification of human DNA can represent the majority of the sequenced DNA. Indeed, this was a challenge with the DNA extractions from whole blood in this study as the “noRoot” OTU accounted for a large proportion of the total

DNA in the blood spike experiments (Figure 3.9) and in the clinical whole blood samples (Table

179

4.2). The mock community analysis suggested that as the level of bacteremia decreased the proportion of the “noRoot” OTU increased (Figure 3.9) thereby indicating that the issue was associated with the ratio of bacterial to human DNA. In the ICU samples, the amount of

“noRoot” was lower in the primary infection samples with the average being 50%. Culture results for these samples indicated they did have higher bacterial load (Table 4.6) further indicating that the ratio of human to bacterial DNA influenced the abundance of the “noRoot”

OTU. Overall, the abundance of “noRoot” DNA in the whole blood samples indicated that the ratio of host DNA to bacterial DNA was high. Since this was a ratio-based issue, the use of larger blood volumes was not predicted to circumvent these limitations. Nevertheless, the removal of these DNA sequences from the taxonomic profile enabled the analysis of the remaining, low proportion, bacterial DNA in the SB samples.

In order to interpret the molecular profiles, the clinical data for the patients was essential.

In terms of clinical severity, the proportion of patients who died in the ICU was highest when they had a bacterial DNA profile that clustered in Groups 1 or 3 (Table 4.3). Within these cluster groups the predominant OTUs were Streptococcus (Group 1) or Staphylococcus (Group 3,

Figure 4.1B). Within each group, there was at least one patient in which there was a culture confirmed Streptococcus or Staphylococcus infection where the patient died (Table 4.2, Table

4.3). As such, this data suggested there was a greater risk for mortality in patients that had a molecular profile in which the Streptococcus or Staphylococcus DNA represented a large portion of the taxonomic profile and there was clinical data to support an infection with viable bacteria.

There was an interesting trend observed in which there was a gender preference for male patients in the Group 3 cluster (Table 4.4). Gender differences have been documented in sepsis with studies in humans and mice demonstrating females have a better outcome in polymicrobial sepsis

180

due to the detrimental immunological effects of the male hormone, testosterone [210]. Based on these observations, a population-based study that examined gender differences in ICU patients indicated fewer women were admitted to the ICU or were diagnosed with sepsis/septic shock

[210]. Despite this, the SB cohort had an overall even ratio of men to woman with only Group 3 exhibiting a gender preference. In terms of mortality, of the 8 patients that died, 5 were males with only 1 female that died in the Group 3 cohort and 1 in the Group 1 cohort (Table 4.3).

Although the cohort is small, the gender differences combined with the mortality rates suggested that Staphylococcus infections might be more detrimental in males with sepsis. The reasons for this are unknown but may warrant further investigation.

Within Group 2, the majority of the SB samples had taxonomic profiles in which DNA representing several genera of bacteria were present without one highly prevalent taxon. (Figure

4.1A) In addition, the bacteria represented by these DNA profiles were considered microbiota commonly identified in both the respiratory and gastrointestinal tract [211]. These patients also presented with clinical indications of either gastrointestinal infections/trauma or respiratory infections (Table 4.3). With these types of infections and their inflammatory response, reduced mucosal barrier function may be implicated. It is well known that the majority of mucosal barriers, excluding the skin, are leaky in order to accommodate fluid exchange [212,213]. The junctional complexes in between epithelial cells play an important role in regulating the mucosal barrier function and the tight junctions are responsible for sealing the space between cells [212].

During inflammatory processes, such as those in response to bacterial infection, cytokines IFNγ and TNF have been linked to increased paracellular transport of large molecules across tight junctions [214,215]. Although whole bacteria are too large to be transported across tight junctions, there is evidence that bacterial products (i.e., LPS) can be transported when barrier

181

function is reduced [213,216]. Taken together, these data suggested that these samples represented a subset of sepsis in which increased mucosal permeability might be implicated in the release of microbial products into the bloodstream, where they are sensed as PAMPS, thereby eliciting a systemic inflammatory response. Consequentially, the DNA recovered in these SB samples represented the diversity of bacterial products released.

The molecular profiling of whole blood demonstrated that almost every patient had

Streptococcus DNA present at detectable levels in the bloodstream (Figure 4.1A). This was not attributed to contamination of the samples since the HB and NTC controls had little to no

Streptococcus DNA recovered (Figure 4.3). This observation suggested there may be a larger role for Streptococcus species in bloodstream infections than appreciated using blood culture based microbiology. In the absence of culture-confirmed infections, empiric therapy focused on

Streptococcus infections may be warranted.

The taxonomic profiles indicated that multiple several highly abundant OTUs represented the Streptococcus, Staphylococcus, and Enterobacteriaceae taxons indicating the DNA representing these genera were likely from various species or groups. The paired-end sequencing allowed a prediction of the species level diversity present for the Streptococcus and

Staphylococcus but caution was required as these results were based on sequence alignments whereas the RDP classifier provided the best depth of identification in which there was statistically significant confidence [133]. Table 4.1 indicated that the principal predicted

Streptococcus species found in whole blood were the S. intermedius/anginosus (representing the

SMG) and S. pneumoniae/oralis/mitis groups at 33.5% and 10.59% respectively. Interestingly, the literature indicated that S. pneumoniae is usually the third most common organism recovered from clinical diagnostic blood culture positive bloodstream infections [217]. Specifically, a

182

Calgary-based studying indicated 89% of culture-positive bloodstream infections were a result of

S. pneumoniae [218] whereas the SMG represented a small minority of the Streptococcus bloodstream infections [13]. It is important to note that these studies reflected culture confirmed bloodstream infections whereas this data only reflected the propensity of bacterial DNA in whole blood limiting the interpretation. Despite these limitations, the abundance of these OTUs may suggest some trends for the potential source of sepsis bloodstream infections. For example, the separation of S. pneumoniae/oralis/mitis by partial 16S rRNA sequencing is not possible [27].

However, the likely source of this DNA would be the respiratory tract since both S. mitis and S. oralis are major constituents of the upper respiratory tract and oral microbiota [219,220]. As such, these data suggest a greater involvement of upper airway microbiota in sepsis. This study suggested a greater role for the SMG in acute bloodstream infections than previously anticipated based on clinical diagnostic blood culture results from population-based studies on bacteremia in the Calgary region [13,29,218]. This would be consistent with the prevalence of SMG in invasive disease [101,221] whereas the underrepresentation of SMG in bacteremia may reflect the difficulty in cultivating as well as identifying the SMG bacteria using conventional diagnostic laboratory cultivation approaches [124]. The role for SMG in human infections is increasingly detected using targeted culturing and culture-independent approaches

[100,101,109,124,222].

There is high 16S rRNA similarity among the Enterobacteriaceae and a given organism may have multiple 16S rRNA genes with different sequences to due the high intragenomic heterogeneity [223]. As such, it was difficult to resolve the Enterobacteriaceae OTUs present in the SB samples (Table 4.1). Despite this, the abundance of various Enterobacteriaceae OTUs in the whole blood molecular profiles suggested there were diverse organisms from the

183

Enterobacteriaceae prevalent in sepsis infections. In Calgary, the Enterobacteriaceae accounted for 27% of the clinical diagnostic blood culture confirmed bloodstream infections [13]. A limitation to our analysis was the amount of Enterobacteriaceae identified in the HB samples

(Figure 4.3). Therefore, the prevalence of this OTU in SB samples was only considered significant if there were clinical data supporting their prevalence. For example, patents ASN438 and 429 in Group 2AI in which this OTU was highly abundant (Figure 4.1A) were diagnosed with respiratory infections with L. pneumophila in the airways (Table 4.2) and did not have clinical data to support an Enterobacteriaceae infection (Table 4.3). Further, this patient was part of the Legionella outbreak in November-December of 2012 in which eight individuals in the

Calgary health region all had respiratory infections with the rare L. pneumophila subtype

Knoxville strain (J. Conly, personal communication). Indeed, ASN438 had the Legionella OTU present in the whole blood molecular profile representing 30% of DNA abundance whereas blood culture repeatedly failed to indicate a systemic L. pneumophila infection (Figure 4.1A). As such, this patient was likely suffering from a Legionella-associated sepsis not an

Enterobacteriaceae-associated sepsis. In contrast, patient ASN371 in Group 2AIII had a bleeding duodenal ulcer which suggested gastric microbiota were implicated in the sepsis infection [196]. In this case, the presence of Enterobacteriaceae as well as other OTUs representative of gastric microbiota could be aligned to the clinical presentation suggesting a clinical correlation. As such, the proper interpretation for Enterobacteriaceae DNA in the whole blood profile required clinical data to support it.

For Staphylococcus, the OTU with sequence alignment to S. aureus represented 96.61% of the Staphylococcus OTUs present in SB (Table 4.1). Further the Group 3 SB samples, which were phylogenetically distinct from the HB samples, suggested that approximately one third of

184

the SB samples had Staphylococcus DNA at levels above the background levels of

Staphylococcus recovered in the HB samples (Figure 4.4). Most reports on clinical diagnostic blood culture confirmed bloodstream infections indicate S. aureus as the second most common organism [217]. As such, the abundance of Staphylococcus DNA in a cohort of SB samples was not unexpected given the prevalence of this organism in bloodstream infections. The clinical data for the Group 3 patients indicated that the majority (10 of 17) of the patients had sepsis following a documented surgical event as they were admitted directly from the OR (Table 4.3). The remaining infections were respiratory (5 of 17) and one with septic arthritis. S. aureus infection have been documented in all of these clinical presentation [224] suggesting the molecular profile data correlated well with the clinical data. Further, this suggested that an infection with

Staphylococcus might be common following surgery, which has been documented in other literature reports, but is not a source of infection commonly associated with Staphylococcus

[196,225]. Lastly, one patient in this group had a clinically confirmed MRSA infection further supporting the validity of the molecular profiling data. As such, the prevalence of the

Staphylococcus OTU in this cohort was likely reflective of Staphylococcus associated systemic infections with viable organisms or bacterial products in these patients.

The paired-end sequencing suggested there were other Staphylococcus infections beyond

S. aureus (Table 4.1). Most population-based assessments cluster the CoNS bloodstream infections together since clinical laboratories don’t distinguish these organisms beyond this level

[13,226]. However, the second most abundant Staphylococcus OTU present in the SB samples had a match to S. sciuri, followed by S. intermedius, S. saprophyticus, and S. epidermidis (Table

4.1). There is increasing evidence that S. sciuri is an emerging pathogen and should be monitored for its prevalence in nosocomial infections [179,180]. S. intermedius is commonly isolated from

185

dogs but has been associated with human infections due to canine exposure and, in limited cases, without canine exposure [184]. S. saprophyticus is commonly associated with urinary tract infections [227]. Taken together, the molecular profiling data suggested that there might be a larger role for diverse CoNS in sepsis then currently appreciated using clinical diagnostic blood culture.

The molecular profiling also suggested some interesting cases of DNA-emia that were not expected based on the patient’s blood culture results. This included ASN363 in which the

Serratia OTU represented 100% of the relative DNA abundance. The blood culture results indicated there were Gram-negative bacilli that could have potentially been S. marcescens. The blood culture results also indicated S. aureus was cultured (Table 4.2). A review of the antibiotic therapy administered to this patient indicated that antimicrobial therapy included vancomycin as well as piperacillin/tazobactam. In this case, the vancomycin would have likely been effective against the S. aureus based on the sensitivity data reported by the clinical diagnostic laboratory

[36] and the piperacillin/tazobactam should have been effective against the S. marcescens based on the literature [228]. However, since the Gram-negative bacilli failed to grow, no antimicrobial susceptibility testing was done. As such, it is unknown if the antibiotic therapy was sufficient in this patient or not. S. marcescens is known to be ubiquitous in the environment and is not known to be difficult to culture [229]. As such, the reason for the failed detection of this organism in blood culture was unknown but could be attributed to an effect of previously administered antibiotics. Nevertheless, complimenting the blood culture data with the bacterial DNA profiling data could have resulted in a better diagnosis for this patient.

The SB samples also had an abundance of anaerobic OTUs present in the whole blood bacteria DNA profiles. Recently, a Calgary based study indicated that the percentage of blood

186

culture-confirmed anaerobic bloodstream infections was 22.5% [29]. This study also indicated that patients with clinical diagnostic blood culture confirmed anaerobic bloodstream infections seemed to have poorer clinical outcomes [29]. Since over half of the SB samples had OTUs identified to anaerobic organisms it was worth exploring the source of this DNA in the blood samples. Among the anaerobic OTUs recovered were Prevotella, Fusobacterium,

Lachnospiraceae, Anaerococcus, Finegoldia, Bacteroides, and Clostridium (Figure 4.1A).

Prevotella and Fusobacterium species are both members of the oral microbiota [230] whereas

Prevotella, Lachnospiraceae, Bacteroides and Clostridium species are part of the gastrointestinal microbiota [231]. Anaerococcus and Finegoldia are both part of the GPAC, which are found in the normal microbiota of skin as well as mucosal surfaces of the oral, upper respiratory, and gastrointestinal tracts [30]. The Group 2AII patients had OTUs for Prevotella, Bacteroides,

Anaerococcus, and Lachnospiraceae present (Figure 4.1). The combination of these OTUs suggested that the DNA came from the gastrointestinal tract. Interestingly, majority of the Group

2AII patients were admitted from the OR following gastrointestinal trauma (Table 4.3). In contrast, the Group 2AIII and 2AIV patients were mainly admitted with airway infections, skin infection, or abscess infections (Table 4.3). The molecular profile from these patients had anaerobic OTUs representing Fusobacterium, Prevotella, and Anaerococcus (Figure 4.1A).

Among these, was the one 2AIV SB sample from ASN415 in which the Prevotella OTU represented 30% of the relative DNA abundance (Figure 4.1A). In addition, other OTUs from upper-respiratory tract bacteria such as Neisseria suggested the anaerobic DNA was derived from the respiratory tract. ASN461 was a patient with a bacterial pneumonia (Table 4.3). The

ASN461 molecular profile was found in Group 3A due to the abundance of Staphylococcus

DNA in the molecular profile (Figure 4.1A). However, the Fusobacterium OTU represented

187

30% of the relative DNA abundance. F. necrophorum was cultivated from the blood culture and pleural fluid obtained from this patient (Table 4.2). Taken together, the anaerobic bacterial DNA in the whole blood was likely derived from constituents of either the respiratory tract in patients with a respiratory/skin infection or the gastrointestinal tract in patients with an intra-abdominal infection with possible impairment in mucosal barrier function permitting the release of their products into the bloodstream.

Time course sampling indicated that there were changes to the bloodstream bacterial DNA profile that may have resulted from the antibiotic therapy administered to these patients. There were some OTUs that were maintained and changed in relative abundance whereas some OTUs were lost and others gained (Figure 4.6A-D). Overall, there was evidence suggesting an increase in bacterial DNA diversity as the abundance of the potential pathogens DNA was decreased.

This suggested that low abundant DNA sequences constituted a larger proportion of the relative

DNA abundance, as the sepsis infection was resolved. There were also instances in which the later day samples suggested another infection was acquired in the patient including those commonly associated with ventilator and catheter usage in the ICU [232,233].

The in-depth culture of the primary infection samples had good concordance with the clinical laboratory culture with the bacteria reported from the clinical culture also recovered in the in-depth culture (Table 4.6). The blood and primary infection samples often shared at least one taxonomic group suggesting that the primary infection site provided the source for the bloodstream infection (Figure 4.7). However, not all DNA from the primary infection site was detected in the whole blood molecular profiles. This could suggest that only some organisms were capable of translocation or persistence in the bloodstream. Interestingly, the main genera of bacteria shared between the whole blood and primary samples were Streptococcus and

188

Staphylococcus, suggesting these organisms disseminated into the blood from the primary infection site. Indeed, Streptococcus and Staphylococcus species have developed mechanisms to evade complement-mediated clearance in blood through the production of complement binding proteins and complement degrading proteases [234]. As such, the development of these immune evasion strategies may explain why the SB samples were enriched with DNA representing either

Streptococcus or Staphylococcus.

The prevalence of polymicrobial DNA in whole blood from the SB patients suggested there was a greater propensity for polymicrobial infections in sepsis than appreciated using cultivation-dependent approaches. Further, the time course sampling suggested that the bloodstream infections were not constant. These data also demonstrated the strength of the molecular profiling data when it was evaluated alongside the patient’s admissions data and, to a limited extent, their culture data. Lastly, the presence of bacterial DNA in the blood without evidence of microbial translocation into the vasculature should not be dismissed. Indeed, there were many samples that had bacterial DNA patterns that were not associated with contamination but could not be confirmed with clinical diagnostic blood culture or other cultures. In addition, compromised mucosal barrier function may also be implicated in the systemic release of bacterial products from microbiota-associated organisms, which can also result in sepsis. Taken together, the identification bacterial products including DNA may be more insightful than culture for detecting both viable pathogen-associated sepsis and bacterial product induced sepsis.

189

Chapter Five: Bacterial DNA Profiles from Emergency Department Patients

5.1 Introduction

Approximately one-third of septic patients are admitted from the ED [235]. Once a patient presents with sepsis, the risk of mortality increases with each hour delay in the administration of appropriate antibiotic therapy [80]. As such, EGDT has been implemented in most emergency departments [79] in which antibiotic treatment should commence within one hour and no later than six hours from triage [36,79]. Although the implementation of EGDT has demonstrated success in the reduction in mortality in sepsis patients, there is a risk in giving patients antibiotic therapy without confirmed infection. Antibiotic overuse is a problem in sepsis and the emergence of multidrug resistant organisms can result from extensive antibiotic therapy

[88]. Furthermore, there is a risk of acquiring nosocomial fungal infections after broad-spectrum antibiotic therapy [34]. Since blood culture is currently the principal strategy to identify a bloodstream infection and the results are only available after at least 24 hours, the implementation of a molecular method to ensure the proper identification of a bloodstream infection when patients are admitted to the ED with suspected sepsis would be beneficial.

The saponin blood treatment method to lyse blood cells followed by a comprehensive

DNA extraction and molecular profiling using paired-end Illumina sequencing protocol indicated polymicrobial DNA was present in septic ICU patients with common bacterial DNA patterns identified that could be correlated to sepsis bloodstream infections. Therefore, a logical extension was to examine if this method could be used to identify bacterial DNA patterns in patients admitted to the ED with suspected sepsis and determine if they were the similar to those

190

observed in the ICU patients. In addition, culturing from whole blood following the saponin treatment was assessed in this cohort.

5.2 Results

5.2.1 Sample Cohorts

Whole blood was collected from ED patients at the Foothills Medical Center and Alberta

Children’s Hospital both in Calgary, Alberta, Canada. Patients who were enrolled at Foothills

Medical Center were identified with the FED prefix and their patient number. There were 52 patients enrolled in this emergency blood (EB) cohort. Patients at the Alberta Children’s Hospital were identified with the prefix AER.G2. There were 28 patients enrolled in this children’s blood

(CB) cohort. No primary infection samples were obtained from the ED patients.

5.2.2 Culture Results from Saponin Treated Whole Blood

Since the saponin blood treatment protocol indicated intact microbial cells could be cultivated from whole blood, fresh samples were processed and cultured when possible following the method outlined in Section 2.4.2. Table 5.1 shows a summary of the culture results from SB and EB samples. Culture was attempted on 44 SB samples and 28 EB samples. For each sample, the starting volume of blood was 1ml. Of these, culture results were obtained for 6 SB samples (13.6%) and 11 EB samples (39.3%). The success rate for culture from EB samples was

3 times that of SB samples. Four samples had multiple isolates present in the blood (Table 5.1).

The recovery of M. luteus was common across the samples (Table 5.1).

191

Table 5.1: Agar-based culture results from saponin treated blood compared to blood culture.

Sample Blood Culture Sample 16S rRNA Sequence Match* CFU/ml Type Result ASN165 SB Streptococcus sp. 6 Negative Streptococcus vestibularis 6 Actinomyces sp. 2 ASN167 SB Micrococcus luteus 9 Negative ASN290 SB Bacillus fusiformis 2 Negative Micrococcus luteus 1 ASN298 SB Acinetobacter radioresistens lawn Not Done ASN320 SB Bacillus subtilis 1 Negative Group A ASN322 SB Micrococcus luteus 1 Streptococcus FED008 EB Staphylococcus epidermidis 1 Negative FED009 EB Moraxella olsoensis 1 Negative FED012 EB Corynebacterium mucifaciens 1 Not Done Staphylococcus epidermidis 24 Staphylococcus warneri 24 Staphylococcus FED014 EB Staphylococcus aureus 2 aureus FED013 EB Propionibacterium acnes 1 Negative FED016 EB Pantoea sp. 1 Negative FED023 EB Acinetobacter baumannii 1 Escherichia coli Micrococcus luteus 1 Staphylococcus epidermidis 1 FED024 EB Proteus mirabilis lawn Proteus mirabilis Staphylococcus warneri 6 FED027 EB Micrococcus luteus 1 Negative FED028 EB Micrococcus luteus 1 Negative Serratia FED036 EB Serratia marcescens 30 marcescens Bacillus clausii 1 Corynebacterium mucifaciens 4 Micrococcus luteus 1 *Methodology described in Section 2.5

192

Micrococcus species are often considered as blood culture contaminants [202]. There was also recovery of other common skin-associated contaminants including S. epidermidis, P. acnes, C. mucifaciens, and Bacillus species (Table 5.1) [202].

For the EB patients, FED009 and FED016 had M. olsoensis and Pantoea sp. recovered whereas the blood culture results were negative (Table 5.1). For FED014, FED024, and FED036 the organism cultivated matched what was recovered in the clinical diagnostic blood culture

(Table 5.1). For FED023, A. baumannii, a problematic multi-drug resistant organism implicated in nosocomial bacteremia [75,236], was recovered whereas blood culture indicated an E. coli infection (Table 5.1).

Most of the organisms recovered in the SB ICU patient samples were likely skin contaminants with the exception of patients ASN165 and ASN298 in which Streptococcus sp. and A. radioresistens were recovered respectively (Table 5.1). The clinical presentation of

ASN165 was discussed at length in Chapter Three (Figure 3.11). The recovery of A. radioresistens in an ICU patient was interesting since incidences of Acinetobacter sp. (not A. baumannii) infections are rare [236]. However, the incidence of A. radioresistens in clinical isolates has been increasing [25]. Lastly, the CFU/ml recovered from the whole blood ranged from 1 to 30.

Overall, there was limited success in culturing directly from whole blood from SB samples and modest success with EB samples. Culturing from SB samples mainly represented organisms classified as blood culture contaminants [202] present at less than 10 CFU/ml.

However, for EB samples there were a few cases where the cultivated organisms were also present in blood culture results thereby demonstrating that viable bacteria could be recovered from 1ml of whole blood using our method. In addition, the recovery of A. radioresistens (Table

193

5.1) suggested the presence of these emerging pathogens in the southern Alberta population.

Lastly, the culturing data also provided an approximation of the microbial load (1-30CFU/ml) in cases of viable bacteria sepsis.

5.2.3 Molecular Profiling of Emergency Blood Samples

The adult patients in the EB cohort had a median age of 53 years with the youngest patient at 19 years and the oldest at 89 years. There were 30 male patients (58%) and 22 female patients (42%). The average number of SIRS criteria met by the patients was 3 with the minimum being 2 and maximum was 4. Five of the patients were admitted to ICU including

FED019, FED029, FED034, FED036, and FED049. There was a range in sequencing depth from a minimum of 9 sequences per sample to a maximum of 3124 sequences per sample. Of the 52

EB samples, only 12 had over 500 sequences per sample and were used for the analysis (Figure

5.1). The OTUs recovered are listed in Table 5.2. The principal OTUs identified in the EB cohort were Serratia, Staphylococcus, Gammaproteobacteria, Enterobacteriaceae, Proteobacteria,

Streptococcus, Lactococcus, Xanthomonadaceae, Bacillaceae, Anaerococcus, Pseudomonas,

Propionibacterium, Actinomycetales, Brevibacillus, and Moraxella (Table 5.2). There were three

Streptococcus OTUs prevalent in these samples. Based on the representative sequence alignments these OTUs were predicted to represent the S. pneumoniae/oralis/mitis group, the

SMG, and the S. salivarius group (Table 5.2).The two Pseudomonas OTUs may have represented environmental species or could represent the clinically relevant P. aeruginosa (Table

5.2).

194

Figure 5.1: The taxonomic profile of adult ED patients indicated polymicrobial DNA was present in whole blood. The taxonomic summaries of whole blood samples with 500 or more sequences were examined using weighted UniFrac. The UPGMA tree of all the samples was generated with the profiles ordered based on their placement in the tree with jackknife support values indicated. Two groups of EB samples were clearly identified. Group 1 was defined by the abundance of Staphylococcus DNA in the profile, Group 2A by the abundance of Gram-negative

OTUs Escherichia, Klebsiella, Enterobacteriaceae, and Serratia, Group 2B by the abundance of Streptococcus, Lactococcus and Gram-negative OTUs .

195

196

Table 5.2: OTUs identified in the EB whole blood samples.

OTU ID (predicted group/species) OTU # Abundance Serratia 11 12.72% Staphylococcus 2 8.28% Gammaproteobacteria 32 7.88% Escherichia (E. coli) 13 5.64% Proteobacteria 15 5.56% Streptococcus (pneumoniae/oralis/mitis) 5 4.26% Streptococcus (SMG) 8 4.14% Streptococcus (vestibularis/salivarius) 40 3.64% Enterobacteriaceae 3 3.46% Lactococcus 67 3.41% Xanthomonadaceae 53 3.33% Bacillaceae 6 3.15% Anaerococcus 101 2.89% Pseudomonas 48 2.27% Propionibacterium 106 2.23% Klebsiella 4 1.61% Actinomycetales 178 1.31% Pseudomonas 68 1.25% Pseudomonadaceae 87 1.06% Brevibacillus 39 0.96% Moraxella 17 0.93%

197

However, given the distribution of the Pseudomonas taxa in almost all the samples and lack of clinical data to support Pseudomonas infections in the samples with this OTU, the abundance of this DNA was attributed to environmental contamination. Similarly, the Gammaproteobacteria in these samples was also considered a contaminant due to its distribution across majority of the samples (Figure 5.1).

Using β-diversity measures, the EB samples in which the sequencing depth was above

500 sequences fell into two phylogenetic clusters with high jackknife support values. There were three EB samples in Group 1 (FED7, FED31, and FED56) in which the abundance of the

Staphylococcus OTU unified the cluster (Figure 5.1). FED7 had a catheter-related infection in which S. aureus was cultured from the blood and urine (Table 5.3). The other two samples were blood culture negative but FED31 had an endovascular infection whereas FED56 had a soft- tissue infection (Table 5.3). The Group 2 samples split into two branches (Figure 5.1). The

Group 2A samples consisted of FED36, which was distinct from the other EB samples due to its unique bacterial DNA profile with the abundance of the Serratia OTU (Figure 5.1). The other

Group 2A samples (FED14, FED42, FED39, and FED44) had polymicrobial DNA recovered with the Gram-negative OTUs Escherichia, Enterobacteriaceae, Klebsiella or Escherichia dominant (Figure 5.1). The Group 2B samples (FED15, FED4, FED57, and FED34) had polymicrobial DNA with Streptococcus, low levels of Staphylococcus, and a mix of other low abundance OTUs (Figure 5.1). All of the EB Group 2 samples had Streptococcus and

Lactococcus OTUs present (Figure 5.1). The Escherichia OTU was the most abundant in the molecular profile for the Group 2A samples (Figure 5.1).

198

Table 5.3: Clinical data and OTU abundance for FED patients in the EB cohort.

Patient Age Gender SIRS Primary Focus of Blood Culture Other Culture* Top OTU (s) (1-4) Infection Group 1 FED31 34 M 2 Endovascular Negative Clostridium Anaerococcus, difficile Staphylococcus FED56 76 M 4 Skin or soft tissue Unknown Bacillaceae FED7 32 M 4 Catheter related Staphylococcus aureus Staphylococcus Staphylococcus aureus Group 2A FED36 52 M 4 Endovascular Serratia marcescens MRSA Serratia

FED14 43 F 3 Skin or soft tissue Staphylococcus aureus Staphylococcus Streptococcus, aureus Gammaproteobacteria

FED42 51 M 2 Lung Streptococcus Escherichia, pneumoniae Gammaproteobacteria

FED39 65 F 3 Unknown Group B Streptococcus Escherichia, Streptococcus FED44 44 F 2 Lung Streptococcus Enterobacteriaceae, pneumoniae Klebsiella Group 2B FED15 67 M 2 Bone/Joint Negative Streptococcus, Bacillus

FED4 46 F 4 Urinary Tract Escherichia coli Escherichia coli Gammaproteobacteria

199

FED57 44 F 3 Lung Unknown Streptococcus, Actinomycetales FED34 49 M 3 Lung Streptococcus Lactococcus, pneumoniae Streptococcus *Primary infection culture , VRE/MRSA Swab, toxin detected, or PCR positive

200

There was only one Escherichia OTU identified which likely represented E. coli based on the representative sequence alignment to the NCBI database (Table 5.2). In Group 1, FED31 had the

Anaerococcus OTU representing 30% of the relative DNA abundance.

When the EB bacterial DNA profile results were compared to blood culture results and the clinically predicted source of infection, there was little correlation seen (Table 5.3). There were blood culture results available for 35 of the EB patients and of these 14 had a positive blood culture result (40%). The blood culture organisms could be correlated to the bacterial DNA profile for two EB samples FED7 and FED36 (Table 5.3). FED7 had S. aureus recovered from blood culture and the Staphylococcus OTU represented 60% of the bacterial DNA profile (Figure

5.1, Table 5.3). FED36 was a patient in whom S. marcescens was cultured by blood culture and in-depth culture (Table 5.3). In addition, the bacterial DNA profile for FED36 indicated 97% of the DNA abundance also correlated to the Serratia OTU (Figure 5.1). As such, FED7 and

FED36 represented patients in whom there was an active bloodstream infection with a viable organism that was detected by both culture and molecular profiling techniques.

Although there was limited clinical data available for the EB patients, correlations were observed between the primary site of infection and the bacterial DNA pattern. The Group 1 EB patients all had a different primary source of infection whereas the patients with lung infections were all found in Group 2 (Table 5.3). Interestingly, three of these patients with a lung source of infection had S. pneumoniae recovered from the clinical diagnostic blood culture (Table 5.3). In examining the bacterial DNA profile in these patients, the Streptococcus OTU was present but did not represent the most abundant OTU (Figure 5.1). Lastly, there were also two Group 2 EB samples in which the bacterial DNA pattern did not correlate with the primary source of infection data or the clinical diagnostic laboratory blood and infection samples. The first sample

201

was from FED14 in which the patient presented with a soft tissue infection with S. aureus cultured from blood and the tissue sample (Table 5.3), however, the bacterial DNA profile indicated only a low level of Staphylococcus DNA was recovered (Figure 5.1). The second patient was FED4 who presented with a urinary tract infection and had E. coli recovered from both blood and urine (Table 5.3) whereas the Escherichia OTU represented only 1% of the relative DNA abundance (Figure 5.1). Overall, the cases in which the clinical data, diagnostic blood culture and molecular DNA patterns were in agreement were limited for these samples.

5.2.4 Molecular Profiling of Children’s Blood Samples

Clinical data was available for 24 of the paediatric, deemed CB samples, patients. The median age was 48 months with children from ages 1 month and to 15 years included in the cohort. There were 11 male (46%) and 13 female (54%) patients. The paired-end Illumina sequencing resulted in a range of sequencing depth from 14 sequences per sample to 14,860 sequences per sample. The most abundant taxonomic group in the CB cohort was the

Staphylococcus OTU representing 41.5% of the total DNA abundance (Table 5.4). Following this, the principal OTUs in the CB cohort were Bacillaceae, Streptococcus, Escherichia,

Moraxella, Enterobacteriaceae, Planococcaceae, Gammaproteobacteria, Clostridium,

Brevibacterium, Enterococcus, Proteobacteria, and Klebsiella (Table 5.4).

There were 9 CB samples in which there was at least 500 sequences per sample (Figure

5.2). Using β-diversity measures, 9 of the CB samples in which the sequencing depth was at least

500 sequences fell into two phylogenetic clusters with high jackknife support values (Figure

5.2).

202

Table 5.4: OTU abundance for the CB cohort samples.

OTU ID OTU # Abundance Staphylococcus 2 41.46% Bacillaceae 6 12.19% Streptococcus (SMG) 8 7.00% Escherichia (E. coli) 13 3.59% Moraxella 17 3.59% Streptococcus (pneumoniae/oralis/mitis) 5 3.05% Enterobacteriaceae 3 2.70% Planococcaceae 19 2.54% Gammaproteobacteria 32 2.40% Clostridium 28 1.89% Brevibacterium 55 1.75% Enterococcus 30 1.59% Proteobacteria 15 1.45% Klebsiella 4 1.10%

203

Figure 5.2: Taxonomic profiles of CB samples. The taxonomic summaries of whole blood samples with 500 or more sequences were examined using weighted UniFrac. The UPGMA tree of all the samples was generated with the profiles ordered based on their placement in the tree with jackknife support values indicated. Two groups of CB samples were identified. Group 1A was defined by the abundance of Streptococcus in the profile, Group 1B by the abundance of

Streptococcus and Enterobacteriaceae, Group 2A by the abundance of Staphylococcus, and

Group 2B by the abundance of Staphylococcus, Bacillaceae, Moraxella, Enterococcus, and

Clostridium OTUs.

204

205

The Group 1A cluster consisted of CB samples with Streptococcus and Staphylococcus taxons representing the majority of the bacterial DNA profile whereas Group 1B had

Enterobacteriaceae as the most abundant taxon followed by Streptococcus then Staphylococcus

(Figure 5.2). The predicted Streptococcus group was determined by aligning the OTU representative sequence to the NCBI database. For AERG2.106 and AERG2.113, the

Streptococcus taxa were predicted to represent Streptococcus intermedius/anginosus and

Streptococcus pneumoniae/oralis/mitis whereas for ASNG2.102 all three Streptococcus OTUs were represented in the taxa summary (Table 5.5). The Group 2A samples consisted of CB patients in which Staphylococcus represented the majority of the bacterial DNA profile (Figure

5.2). The Group 2B samples consisted of three CB patients where the bacterial DNA profiles consisted of Bacillaceae and Staphylococcus at equal abundance and all three profiles having

Moraxella, Clostridium, and Enterococcus taxons representing 10-3% of the relative DNA abundance (Figure 5.2).

The clinical data obtained for the CB cohort samples was limited. With respect to blood culture data, all of these samples were blood culture negative except AERG2.205 in which Gram- positive cocci resembling Staphylococcus were recovered (Table 5.5). This could be correlated with the bacteria DNA profile in this patient in which Staphylococcus represented one of the most abundant OTUs (Figure 5.2). There was no clear correlation between the patients primary focus of infection and the bacterial DNA pattern recovered from whole blood.

Nevertheless, polymicrobial DNA profiles were present in CB whole blood samples despite the negative clinical diagnostic blood culture data.

206

Table 5.5: Clinical data and OTU abundance for AER-G2 patients in the CB cohort.

Patient Age Gender SIRS Primary Focus of Admitted Blood Culture Other OTU(s) (1-4) Infection to Culture* Hospital Group 1A AERG2.106 2 F 2 Pneumonia No Negative Streptococcus, Escherichia, Staphylococcus AERG2.102 4 F 3 Peritonitis/Appendicitis Yes Negative Escherichia Staphylococcus, coli Streptococcus AERG2.113 4 F 3 Meningitis No Negative Staphylococcus, Streptococcus Group 1B AERG1.106 2 F 2 Pneumonia No Negative Enterobacteriaceae, Streptococcus Group 2A AERG2.043 7 F 2 Peritonitis/Appendicitis Yes Negative Staphylococcus

AERG2.076 10 F 2 Duplicate Cyst No Negative Staphylococcus

Group 2B AERG2.205 2 M 2 Pneumonia No Gram-positive Bacillaceae, cocci Staphylococcus, resembling Moraxella, Staphylococcus Enterococcus, Clostridium

207

AERG2.235 No No No No data No data Negative Bacillaceae, data data data Staphylococcus, Moraxella, Enterococcus, Clostridium AERG2.198 3 F 4 No data No Negative Bacillaceae, Staphylococcus, Moraxella, Enterococcus, Clostridium *Primary infection culture, VRE/MRSA Swab, toxin detected, or PCR positive

208

5.2.1 Comparison of CB and EB Patients to Healthy Controls

The blood samples from the adults and paediatric ED were clustered with the HB samples in order to examine if the bacterial DNA patterns present in these cohorts were distinct from healthy adults. The phylogenetic clustering in the UPGMA tree was split into two main branches

(Figure 5.3). The first branch consisted of the majority of the CB samples with all the CB-1A and CB Group 2A and 2B samples all in this branch. The EB Group 1 samples were all present in this branch with one EB Group 2B sample as well (Figure 5.3). In the second branch of the

UPGMA tree consisted of the HB samples that remained clustered together as well as the EB

Group 2A and 2B samples as well as one CB Group 1B sample (Figure 5.3). Overall, the

UPGMA tree indicated that there were two bacterial DNA patterns present in these samples that were distinct from the healthy control bacterial DNA patterns. The first bacterial DNA pattern was characterized by the relative abundance of Gram-positive OTUs including Staphylococcus,

Streptococcus, Lactococcus, Anaerococcus, and Bacillaceae OTUs (Figure 5.3). The second bacterial DNA pattern was characterized by a prevalence of Gram-negative OTUs including

Gammaproteobacteria, Proteobacteria, Enterobacteriaceae, Escherichia, and Klebsiella in addition to the Streptococcus OTUs. Further, the CB and EB cohort samples were all intermixed indicating phylogenetic similarities were present between adult and paediatric ED patients

(Figure 5.3).

Although there was limited clinical data, the EB Group 1, CB Group 1A and CB Group 2 patients all clustered on the same branch in Figure 5.3 were admitted mainly due to soft tissue infections or intra-abdominal infections (Table 5.3, Table 5.5).

209

Figure 5.3: The whole blood samples from CB and EB patients cluster separately from HB samples. Taxonomic bacterial DNA profiles were summarized for all whole blood samples with

500 or more sequences. A composite unweighted pair group method with arithmetic mean

(UPGMA) phylogeny of all the samples was generated with the profiles ordered based on their placement in the UPGMA tree and clustered using weighted UniFrac with a sampling depth of

500 sequences (B). The EB and CB samples were labeled by their cluster groups identified in

Figure 20A and Figure 21A. The HB samples clustered separately from both the CB and EB samples. The CB and EB samples clustered together in one of two branches, the first consisting of samples with an abundance of Staphylococcus or Staphylococcus and Bacillaceae and the other consisting of samples with a low abundance of Streptococcus and several Gram-negative

OTUs including Gammaproteobacteria, Proteobacteria, Enterobacteriaceae, Escherichia,

Klebsiella, and Serratia.

210

Streptococcus Staphylococcus Enterobacteriaceae Escherichia Gammaproteobacteria Pseudomonas Enterococcus Proteobacteria Xanthomonadaceae Klebsiella Moraxella Serratia Prevotella Propionibacterium Fusobacterium Neisseria Lactococcus Actinomycetales Anaerococcus Bacillaceae Bacillus

211

In contrast, the majority of the EB Group 2 patients and the one CB Group 1 patient had lung infections or pneumonia (Table 5.3, Table 5.5). Taken together, both paediatric and adult patient with suspected sepsis in the ED shared bacterial DNA patterns in their blood that could be associated to the primary infection source.

5.2.2 Comparison of Bacterial DNA Patterns between ED and ICU Whole Blood Samples

Since patients enrolled in the ED were all suspected of having sepsis, it was important to compare their bloodstream bacterial profiles to those patients who were enrolled in the ICU with sepsis. The ICU samples were labeled based on their cluster group (Figure 4.1). Similarly, their cluster groups from Figure 5.1 and Figure 5.2 distinguished the EB and CB samples. This was done to determine which bacterial DNA patterns overlapped and to determine if they could be linked to their infection characteristics. To ensure consistency in the comparison between samples of varying sequencing depth, the OTU table was rarefied to 500 sequences per sample and the UPGMA tree was generated using weighted UniFrac (Figure 5.4). The clustering of all the whole blood samples together indicated that the bacterial DNA patterns present in the ICU patients were similar to those recovered in the EB and CB patients. There were 5 branches in the

UPGMA tree, each of which consisting of similar bacterial DNA patterns. The clinical characteristics of the samples in each branch were examined (Table 5.6).

The first branch of the UPGMA tree consisted of only SB samples that had the

Streptococcus OTU as the most abundant in their bacterial DNA profile (Figure 5.4). These patients all had good correlation between their clinical data, clinical diagnostic cultures and the bacterial DNA profile recovered (Table 5.6). As such, these patients potentially had a bloodstream infection in which Streptococcus species were implicated.

212

Figure 5.4: The bacterial DNA profiles of ICU and ED samples clustered together and separately from HB samples. Taxonomic bacterial DNA profiles were summarized for all whole blood samples with 500 or more sequences. A composite unweighted pair group method with arithmetic mean (UPGMA) phylogeny of all the samples was generated with the profiles ordered based on their placement in the phylogenetic tree and clustered using weighted UniFrac.

Their cluster groups identified in Figure 4.1, Figure 5.1, and Figure 5.2 labelled the SB, EB, and

CB samples. The samples clustered into 5 branches on the phylogenetic tree, which are labeled above each group with lines showing the separation between the branch groups.

213

1 2 3A 3B 3C 4A 4B 4C 4D 4E 5

214

Streptococcus Staphylococcus Lachnospiraceae Enterobacteriaceae Veillonella Prevotella Escherichia Fusobacterium Gammaproteobacteria Lactococcus Pseudomonas Enterococcus Propionibacterium Proteobacteria Neisseria Xanthomonadaceae Serratia Klebsiella Actinomycetales Anaerococcus Bacillaceae Finegoldia Moraxella Capnocytophaga Legionella Bacteroides Haemophilus Streptophyta Clostridium

215

Table 5.6: Clinical data for all the whole blood clinical samples clustered based on weighted UniFrac UPGMA.

Patient Age Original Primary focus Diagnostic Blood Diagnostic Primary Source (for Correlation Cluster of infection Culture Infection Culture1 primary (with Group infection bacterial sample) DNA profile) Branch 1 ASN455 47 SB-1A Urinary tract Group G Streptococcus Group G Catheter Yes Streptococcus ASN350 44 SB-1A Lung Negative Negative Yes ASN349 60 SB-1A Intracranial Negative Streptococcus Yes abscess intermedius ASN452 76 SB-1A Lung Negative Not done Yes Branch 2 ASN479 57 SB-2B Bone/Joint Negative Finegoldia magna Drainage Yes fluid Branch 3A ASN379 74 SB-2AIII Lung Negative Not done Likely ASN371 58 SB-2AIII Intra-abdominal CoNS Not done Unclear ASNC21 48 HB None Not done Not done ASNC18 28 HB None Not done Not done ASNC19 31 HB None Not done Not done ASNC17 58 HB None Not done Not done ASNC20 37 HB None Not done Not done ASNC23 40 HB None Not done Not done ASNC14 46 HB None Not done Not done ASNC22 73 HB None Not done Not done

216

ASNC13 37 HB None Not done Not done ASNC12 58 HB None Not done Not done FED39 65 EB-2A Unknown Group B Streptococcus Not done No FED44 44 EB-2A Lung Streptococcus Not done No pneumoniae ASN381 53 SB-2AIII Skin/Soft tissue Negative Legionella No Branch 3B ASN366 53 SB-1B Upper Negative Streptococcus Wound Yes respiratory anginosus, culture Prevotella, CoNS ASN357 80 SB-1B Intra-abdominal Negative VRE VRE/MRSA Likely swab (rectal or nasal) ASN376 25 SB-1B Lung Not done Not done Likely ASN368 60 SB-1B Intra-abdominal Negative Klebsiella Pleural fluid Yes pneumoniae, Haemophilus parainfluenzae, Prevotella FED34 49 EB-2B Endovascular Streptococcus Gram-positive cocci Lung fluid Yes pneumoniae ASN465 70 SB-1A Heart Negative VRE VRE/MRSA Likely swab (rectal or nasal) ASN463 70 SB-1A Heart Negative Not done Likely ASN468 19 SN Splenectomy Not done Not done Likely AERG2.106 2 CB-1A Lung Negative Not done Likely ASN469 66 SB-1A Skin/Soft tissue Campylobacter Not done No ureolyticus, Fusobacterium ASN467 49 SN Intracranial Negative Not done Likely aneurysm

217

ASN470 69 SB-1A Gastrointestinal VRE VRE VRE/MRSA No swab (rectal or nasal) ASN294 63 SB-1B Lung Staphylococcus aureus Staphylococcus Lung fluid Yes aureus ASN300 69 SB-2AII Gastrointestinal Negative Fungal Lung fluid Likely FED15 67 EB-2B Bone/Joint Negative Not done Unclear FED4 46 EB-2B Urinary tract Escherichia coli Escherichia coli Urine No FED57 44 EB-2B Lung Unknown Unknown Likely ASN432 57 SB-2AIII Lung Negative Group A Lung fluid, Yes Streptococcus Pleural Fluid AERG1.106 2 CB-1B Lung Likely ASN444 77 SB-2AIII Lung Negative Not done Yes Branch 3C ASN475 77 SB-2AI Gastrointestinal Pseudomonas Pseudomonas Abscess fluid Yes aeruginosa aeruginosa FED14 43 EB-2A Skin/Soft tissue Staphylococcus aureus Not done Yes ASN167 37 SB-2AI Hepatic abscess Negative SMG Abscess fluid Yes ASN168 52 SB-2AI Lung Not done Fungal Lung fluid Yes FED42 51 EB-2A Lung Streptococcus Not done Yes pneumoniae FED36 52 EB-2A Endovascular Serratia marcescens Not done Yes ASN363 33 SB-2AI Intra-abdominal Staphylococcus Staphylococcus Catheter, Yes aureus, Gram-negative aureus, Wound bacilli Enterococcus ASN429 71 SB-2AI Lung Negative Legionella Lung fluid No pneumophila ASN315 29 SB-2AI Skin/Soft tissue Negative Not done Likely ASN438 57 SB-2AI Lung Negative Legionella Lung fluid Yes pneumophila Branch 4A ASN461 24 SB-3A Lung Fusobacterium Fusobacterium Pleural fluid Yes

218

necrophorum Branch 4B ASN418 46 SB-3A Intra-abdominal Not done Not done Likely ASN451 79 SB-3A Gastrointestinal Not done Not done Likely ASN440 65 SB-3A Lung Enterococcus faecium Not done No ASN458 63 SB-3A Lung MRSA MRSA Lung fluid Yes ASN409 53 SB-3A Lung Not done Not done Likely ASN436 73 SB-3A Gastrointestinal Negative VRE VRE/MRSA No swab (rectal or nasal) ASN424 68 SB-3A Gastrointestinal Bifidobacterium Not done No AERG2.76 10 CB-2A Cyst Negative Not done Likely AERG2.43 7 CB-2A Intra-abdominal Negative Not done Likely ASN454 78 SB-3A Gastrointestinal Negative Not done Likely ASN408 45 SB-3A Intra-abdominal Not done Not done Likely ASN434 51 SB-3A Catheter-related Pseudomonas Not done No aeruginosa ASN466 70 SB-3A Intra-abdominal Negative CoNS, Coryneform Abdominal Yes bacilli fluid AERG2.113 4 CB-1A Meningitis Negative Not done Likely AERG2.102 4 CB-1A Intra-abdominal Negative Not done Likely ASN462 25 SN Head/Chest Negative Not done* Lung fluid Yes Trauma ASN348 77 SB-3A Intra-abdominal Negative Micrococcus, Wound Yes Streptococcus culture viridians group ASN464 62 SB-3A Intra-abdominal Not done Not done Yes FED7 32 EB-1 Catheter-related Staphylococcus aureus Not done Yes Branch 4C AERG2.235 - CB-2B Unknown Unknown Unknown Unclear ASN476 66 SB-3B Urinary tract Escherichia coli Not done No ASN487 65 SN Unknown Negative+ Not done Likely 219

ASN474 42 SB-3B Lung Negative Not done Likely AERG2.198 3 CB-2B Unknown Negative Not done Unclear AERG2.205 2 CB-2B Lung Gram-positive cocci Not done Yes resembling Staphylococcus ASN477 54 SB-3B Lung Negative Not done Likely FED56 76 EB-1 Skin/Soft tissue Unknown Unknown Likely Branch 4D FED31 34 EB-1 Endovascular Negative Not done Likely ASN340 67 SB-2AIII Skin/Soft tissue Group C Streptococcus Not done Yes ASN339 38 SB-2AIII Intracranial Negative Not done Likely abscess ASN343 56 SB-2AIII Lung Negative Fungal Lung fluid Likely Branch 4E ASN328 26 SB-2AII Lung Negative Staphylococcus Lung fluid Yes aureus, Streptococcus pneumoniae ASN292 63 SB-2AII Intra-abdominal Negative Not done Yes ASN297 58 SB-2AII Lung Fungal Fungal Lung fluid Likely ASN338 25 SB-2AII Lung Not done Not done Likely ASN473 77 SB-2AII Lung Pseudomonas Not done No aeruginosa ASN420 56 SB-2AII Lung Negative Not done Likely Branch 5 ASN415 76 SB- Lung Negative Not done Likely 2AIV 1Culture result, VRE/MRSA swab result, toxin detected, or PCR positive *Positive on Day 14 for S. aureus and Haemophilus influenzae +Positive on Day 14 for CoNS

220

The second branch of the UPGMA tree contained the only SB sample in which F. magna was implicated in the bloodstream infection due to it’s recovery from the patient’s drainage fluid and the abundance of the Finegoldia OTU in the bacterial DNA profile (Table 5.6, Figure 5.4).

The third branch of the UPGMA tree was sub-divided into three groups based on the sub- divisions within the branch (Figure 5.4). Within the first group, Branch 3A, were some clinical samples as well as all of the HB samples. As previously observed, the HB samples were all found within the same cluster group, which was distinct from all of the ICU and ED patient whole blood samples (Figure 5.4)

Some of the bacterial DNA profiles from SB Group 2AIII patients (ANS479, ASN371, and ASN381) and EB Group2A samples (FED39 and FED44) clustered in the same branch as the HB samples. It was possible to correlate the clinical data for ASN379 (evidence of a lung infection) with the bacterial DNA profile indicating Streptococcus, Fusobacterium, Prevotella, and Neisseria DNA was abundant (Table 5.6, Figure 5.4). All of these OTUs were representative of upper airway commensal bacteria [129] which are often recovered in the lungs during lower airway infections [100]. Based on this, it was plausible that DNA from such organisms would be recovered in the blood of this patient. However, for the remaining samples found within the same branch as the HB samples, there was not enough clinical data to support their bacterial DNA profile in blood (Table 5.6). As such, the bacterial DNA patterns in these patients were considered as uncorrelated to the patient’s infection.

The second group in branch 3, Branch 3B, consisted of samples in which the bacterial

DNA profile was characterized by the prevalence of Streptococcus with Lactococcus, Prevotella,

Fusobacterium, and Neisseria OTUs also present (Figure 5.4). The patients with Lactococcus recovered with Streptococcus were all adult ICU and ED patients. The majority of the patients in

221

this group had infections in the lungs followed by intra-abdominal/gastrointestinal related infections (Table 5.6). For these infections there was evidence to suggest a correlation between the bacterial DNA pattern and the patient’s clinical diagnostic blood culture or primary infection culture (Table 5.6). Interestingly, the 3B branch had the most endovascular or heart related infection samples (Table 5.6). Endovascular infections, in the context of sepsis, likely represented infective endocarditis. As such, the clustering of patients with these infections suggested that the bacterial DNA profile in which Streptococcus was predominant were often associated with infective endocarditis associated bloodstream infections. Streptococcus species are among the most common organisms recovered in infective endocarditis [27], which supported the bacterial DNA patterns identified in these patients. In contrast, there were also cases in which the bacterial DNA patterns did not correlate with the patient’s clinical presentation. These were often associated with patients that had a urinary tract infection (FED4) or were positive for VRE based on a nasal or rectal swab (ASN357, ASN465, and ASN470)

(Table 5.6). For ASN469 the patient had a skin/soft tissue infection and had Fusobacterium recovered in the diagnostic blood culture. Despite this, the bacterial DNA pattern did not indicated the Fusobacterium OTU was abundant (Figure 5.4). Overall, the bacterial DNA patterns identified in the branch 3B group had good correlation to the clinical data with the exception of the patient with a urinary tract infection and the patient with a skin/soft tissue infection. Further, the bacterial DNA patterns in blood suggested that VRE identified from rectal swabs were not present in the bloodstream.

The final group in branch three (3C) consisted of patients with bacterial DNA profiles that had low abundance of Streptococcus with higher relative abundance of

Gammaproteobacteria, Proteobacteria, or Enterobacteriaceae (Figure 5.4). In chapter three, the

222

abundance of the Gammaproteobacteria OTU in the case studies was attributed to contamination since the OTU could not be correlated to the patient’s clinical data (Figure 3.11). Further,

Gammaproteobacteria or Proteobacteria OTUs were too high of an order to be correlated to genus or species level bacteria identified in the clinical cultures. The Enterobacteriaceae OTU was also highly abundant in the HB samples, which suggested that without clinical data to support an infection with Enterobacteriaceae, these OTUs were also likely contaminants. Taken together, in the group 3C patients, there was no evidence to support infections with

Gammaproteobacteria, Proteobacteria, or Enterobacteriaceae whereas other OTUs in the bacterial DNA profiles could be correlated to the clinical data. For example, ASN475 had the

Gammaproteobacteria and Proteobacteria OTUs prevalent in their whole blood bacterial DNA profile. However, the Pseudomonas OTU (Figure 5.4) could be correlated to the patient’s clinical presentation since P. aeruginosa was recovered by the clinical diagnostic culture from the gastrointestinal abscess (Table 5.6). Similarly, ASN438 had the Enterobacteriaceae OTU at the highest relative DNA abundance (Figure 5.4) but had L. pneumophila recovered from the clinical diagnostic lung fluid culture (Table 5.6). This suggested that the Legionella OTU present in the bacterial DNA profile (Figure 5.4) was implicated in systemic infection, not the

Enterobacteriaceae. As such, it appeared that the branch 3C samples represented cases in which the bacterial DNA profile had higher levels of contaminant DNA namely Gammaproteobacteria,

Proteobacteria, and Enterobacteriaceae that could not be associated to the patient’s clinical data.

The exceptions were ASN363 and FED36, which were patients that had the Serratia OTU in the blood bacterial DNA profile (Figure 5.4). In both patients, this OTU was correlated to the diagnostic blood culture (Table 5.6) indicating these patients likely were bacteremic with viable

S. marcescens.

223

The fourth branch of the UPGMA tree was split into 5 subgroups (Figure 5.4). However, the unifying feature between these sub-groups was the Staphylococcus taxon. The first group

(4A) had one sample from ASN461 in which there was good correlation between the bacterial

DNA profile with the Fusobacterium OTU in the prolife (Figure 5.4) likely representing the presence of F. necrophorum. Indeed, this organism was recovered from the clinical diagnostic blood culture and was also identified in the pleural fluid culture (Table 5.6). The bacterial DNA profile also suggested that, in addition to F. necrophorum, that this patient may have had polymicrobial bloodstream infection with Staphylococcus implicated based on the abundance of this OTU in the whole blood bacterial DNA profile (Figure 5.4)

The 4B branch samples consisted of patients in which the Staphylococcus OTU represented the majority of the relative bacterial DNA abundance (Figure 5.4). In terms of clinical correlations, there were only four samples in which the bacterial DNA profile did not support the patient’s clinical diagnostic cultures (Table 5.6). Interestingly, as observed in the branch 3B samples, these included patients in which VRE was identified in the rectal swab

(ASN436) or when Enterococcus was identified in the diagnostic blood culture (ASN440) (Table

5.6). The other two patients were ones in which Bifidobacterium were identified in the blood culture (ASN424) and P. aeruginosa was identified in the blood culture (ASN434). As such, for these patients it was difficult to align the abundance of Staphylococcus in the whole blood bacterial DNA profile to their clinical presentation. However, for the remainder of the branch 4B samples the abundance of the Staphylococcus in the whole blood bacterial DNA profile was well aligned with the patient’s source of infection or with the clinical diagnostic culture results indicated Staphylococcus in the patients blood or primary infection sample. There were twelve patients in the 4B branch that had intra-abdominal or gastrointestinal related infections (Table

224

5.6). The presence Staphylococcus in such infections has been documented and was discussed in chapter four [196,237]. There were also instances of lung infections in the branch 4B samples with one of these (ASN458) having MRSA recovered from both the diagnostic blood culture and lung fluid culture and was also discussed in chapter four (Table 5.6).

The branch 4C samples had polymicrobial bacterial DNA profiles in which Bacillaceae was the most prevalent OTU (Figure 5.4). Further these patients all had Staphylococcus,

Clostridium, Moraxella, and Enterococcus OTUs represented in the bacterial DNA profiles

(Figure 5.4). Staphylococcus, Moraxella, and Enterococcus have been identified, as major constituents in molecular profiling of the upper airways of children [238] but are less abundant in adults [129]. Interestingly, the 4C branch group contained three pediatric CB samples with one of them identified as having a lung infection (Table 5.6). The adult patients mostly had lung infections as well (Table 5.6). This suggested that the bacterial DNA pattern present in these samples was likely derived from bacteria present in the upper airways that may be implicated in the lung infection and potentially disseminated into to the bloodstream from the lungs.

The 4D branch samples were unique in that they had the Anaerococcus OTU representing

10-30% of the relative DNA abundance in the bacterial DNA profile (Figure 5.4). Anaerococcus is GPAC, which have been implicated, in human infections of the mouth, skin/soft tissue, bone/joints and the upper respiratory tract [31]. Within the 4D group, the patients had a range of primary infections including endovascular, skin/soft tissue, abscess, and lung infections (Table

5.6) thereby supporting the range of diversity associated with GPAC infections. The clinical data for the patients in this group was limited but for ASN340 there was a likely correlation between the whole blood bacterial DNA profile and the clinical presentation of this patient. ASN340 had a skin/soft tissue infection in which Group C Streptococcus was recovered from the clinical

225

diagnostic blood culture (Table 5.6). The Streptococcus OTU was also present in the whole blood bacterial DNA profile as were the OTUs for anaerobic bacteria including Anaerococcus and Prevotella (Figure 5.4). These data suggested the possibility of a bloodstream infection likely resulting from a Streptococcus skin infection. In addition, there was the possibility of anaerobic bacteria dissemination in the blood. As such, the prevalence of Anaerococcus in this branch suggested that the GPAC might be implicated in septic bloodstream infections in these patients despite their diagnostic blood culture results being negative.

The 4E branch also consisted of patients with bacterial DNA profiles in which there were anaerobic OTUs present including Anaerococcus, Prevotella, Lachnospiraceae, and Bacteroides

(Figure 5.4). The prevalence of Prevotella and Lachnospiraceae OTUs in the molecular profiling of upper airway of adults has been documented [129]. Indeed, the majority of the patients in this cluster had lung-associated infections (Table 5.6). ASN328 was evaluated in Chapter Three and this patient had excellent correlation between the in-depth culture performed in this study of the

ETT fluid to the diagnostic clinical blood and lung fluid cultures (Figure 3.11). As such, patients in this cluster might be reflective of aspiration-associated airway infections in which bacteria from the upper respiratory or gastrointestinal tract microbiota migrated to the lungs. This could explain the prevalence of upper airway associated OTUs in the whole blood bacterial DNA profiles for these patients and suggested that such organisms could be implicated in systemic infection.

The final branch in the UPGMA tree consisted on one patient sample from ASN415 in which the Prevotella OTU represented 29% of the relative DNA abundance (Figure 5.4). This patient had a lung infection but there was no clinical diagnostic data for comparison. Despite this, the instances of upper airway anaerobic OTUs identified in the bloodstream bacterial DNA

226

profiles were numerous in the branch 4C-E samples (Figure 5.4). As such, the Prevotella OTU in the whole blood of this patient suggested a possible dissemination of a Prevotella species that likely originated from the upper airway microbiota.

Overall, the ICU SB clustering remained consistent for the Group 1 and Group 3 samples with many of the EB, CB, and SN samples clustering in these groups (Figure 5.4). The SB Group

2 samples did not cluster along the original clustering pattern found in Figure 4.1. Taken together, there were common bacterial DNA patterns in both ED and ICU samples which reflected a predominance of Streptococcus or Staphylococcus. Additionally, analysis of the primary infection site suggested that other OTUs prevalent in the bloodstream bacterial DNA profiles were a result of commensal microbiota that were representative of either the upper respiratory tract or the gastrointestinal tract suggesting a greater role for such organisms in the pathogenesis of sepsis.

5.2.3 Discussion

Early identification of bloodstream infections in suspected sepsis patients admitted to the

ED would allow for more expedient administration of appropriate antibiotics and potentially reduce the risk of severe sepsis or septic shock. The application of a molecular profiling method on whole blood could circumvent the limitations of a culture-based diagnostic methods with regards to time delays and relatively low sensitivity [61]. The use of paired-end Illumina 16S rRNA sequencing on DNA isolated from whole blood, pre-treated with saponin to lyse blood cells, allowed for the identification of polymicrobial bacterial DNA profiles in a subset of adult and paediatric patients admitted to the ED with suspected sepsis (Figure 5.1 and Figure 5.2)

227

while culturing of whole blood recovered viable bacterial cells from adult ED patients (Table

5.1). Lastly, the analysis of all the clinical samples obtained on day 1 of admission to the ICU and ED indicated common bacterial DNA patterns were present in whole blood (Figure 5.4).

Although the culturing of saponin treated whole blood had limited success, these data helped provide information on the concentrations of circulating bacteria in blood. This level of quantification is rarely reported since blood culture results only reflect bacterial concentrations following broth enrichment [51]. The range in CFU/ml was between 1-30 colonies (Table 5.1).

Although these values only reflect what was recovered in a very small blood sample (1ml), it did suggest that concentrations of bacteria in blood were quite low. Further, despite the low success rate from ICU samples, the recovery of a viable organism in almost 40% of the EB samples suggested that cultivation directly from blood following the saponin method was still improved when compared to diagnostic blood culture in which the success rate is often below 20% [61].

As such, the use of saponin directly on blood from ED patients could recover viable bacteria in less time and with a lower starting inoculum when compared to diagnostic blood culture.

From the 52 FED samples included in the EB cohort, only 12 were included in the analysis (23%), which was lower than anticipated. For the CB cohort, 9 of the 28 patient samples

(32%) were included in the analysis. Given the limited number of samples in both the EB and

CB cohort, the clustering patterns present in the UPGMA tree could only be analyzed to a limited extent. Despite this, there were two phylogenetic clusters identified in the EB samples in which the abundance of Staphylococcus and Streptococcus DNA delineated the two groups

(Figure 5.1). None of the EB samples clustered with the HB samples indicating that the bacterial

DNA present in the EB samples were not solely derived from contamination that resulted in the

HB profiles (Figure 5.3). Within Group 1A there was one EB sample in which the abundance of

228

the Staphylococcus OTU correlated with the clinical diagnostic blood culture (Figure 5.1, Table

5.3). As such, this data suggested that the other samples within this cluster might have had

Staphylococcus associated infections that were not identified using the clinical diagnostic blood culture. For the Group 2 EB samples, all the patients presented with a lung infection with three patients having S. pneumoniae recovered from the clinical diagnostic blood culture (Table 5.3).

Although the relative abundance of the Streptococcus DNA was low, it aligned with the diagnostic clinical cultures and clinical presentation of respiratory infections (Table 5.3). With

S. pneumoniae being one of the most common pathogens associated with community acquired pneumoniae [47,218], the Group 2 patients likely all had Streptococcus associated infections in which either a viable organism in case of the blood culture confirmed S. pneumoniae infection or bacterial products from a respiratory Streptococcus infection disseminated into the bloodstream.

Overall, very few EB samples were successfully evaluated. However, the ones that could be evaluated demonstrated that there were correlations between the bacterial DNA patterns and the clinical data suggesting modest success in applying this method to EB samples. Interestingly, only five of the EB patients were subsequently admitted to the ICU with FED34 and FED36 included in the EB analysis. The remainders were either admitted to an in-patient ward or the outpatient clinic. As such, although there was suspicion of sepsis, the vast majority of these patients did not progress to the point of critical care intervention. Taken together, this suggested that many of the EB patients may not have been septic and that our enrolment criteria were possibly not stringent enough to ensure the proper samples were collected. It also indicated the difficulty in identifying septic patients early in the ED.

The paediatric CB samples clustered into two groups on a UPGMA tree that showed similarities to the SB Group 1 and Group 3 distributions (Figure 4.1, Figure 5.2). The CB isolates

229

also clustered separately from the HB samples (Figure 5.3). There was limited data for clinical correlations since most of these samples were diagnostic blood culture negative (Table 5.5). The lack of clinical diagnostic blood cultures for comparison challenged our ability to determine if the bacterial DNA patterns present fit the clinical context of paediatric sepsis for these patients.

However, an epidemiological study of paediatric sepsis in the United States indicated that the most common infecting organism in diagnostic blood culture confirmed bloodstream infections was Staphylococcus followed by Streptococcus [239]. Since the majority of the CB samples had

Staphylococcus or Streptococcus as an abundant OTU in the bacterial DNA profile (Figure 5.2), this suggested a role for these organisms or their products in the CB patient’s bloodstream infections. The use of molecular diagnostics in paediatric settings has already been explored extensively due to the volume limitations impact on blood culture efficacy in these settings

[108,240-242]. Given the success of the paired-end Illumina sequencing in the CB cohort and increased need for better identification of bloodstream infections in paediatric settings, the implementation of this method in a paediatric setting would likely be beneficial.

The assembly of all the clinical samples and healthy controls indicated that the bacterial

DNA patterns observed in the ICU patients were also present in the EB and CB cohorts. When the clinical data was examined for each branch group it became increasingly apparent that the bacterial DNA patterns were influenced by the primary source of infection. The branches in which the Streptococcus OTU was abundant were associated mainly with lung infections, abscess infections and endovascular infections (Table 5.6). Prevalence of Streptococcus species is well supported in all these infection sites [27,243]. Indeed, there were several cases in which the diagnostic blood culture or primary infection sample recovered Streptococcus from patients within the branch 1 and 3B groups (Table 5.6). Taken together, patients that present with lung

230

infections, abscess infections, and endovascular infections that have Streptococcus DNA recovered in whole blood likely represented patients in whom there was a Streptococcus infection or Streptococcus products in the bloodstream.

Interestingly, the clusters with the abundant Staphylococcus OTU were associated primarily with intra-abdominal/gastrointestinal infections followed by lung infections then skin/soft-tissue infections (Table 5.6). Staphylococcus species are known to cause respiratory infections [244] and skin/soft-tissue infections [245]; however, they are not often associated with gastrointestinal infections [197]. Despite this, studies have implicated S. aureus infections in cases of intra-abdominal or gastrointestinal trauma [196]. In our results, there were many cases in which the Staphylococcus DNA could be correlated to the clinical data, indicating that these patients were likely presenting with a Staphylococcus associated systemic infection (Table 5.6) with a greater role for Staphylococcus infections associated with intra-abdominal or gastrointestinal infections observed.

The abundance of the Moraxella OTU was also unique to the 4C branch (Table 5.5). The prevalence of this OTU in the CB samples was not unexpected since paediatric patients often have Moraxella in childhood otitis media infections [121,246]. Recently, an analysis of the microbiota of the upper respiratory tract indicated an abundance of Moraxella within the nasopharyngeal microbiota of children [238]. The results from the CB cohort indicated that

Moraxella might also play an important role in systemic infections within paediatric patients

(Figure 5.4). These infections may be a result of a disseminated infection following otitis media or as a result of upper airway microbiota or their products present in the bloodstream due to reduced mucosal barrier function. For the adult samples in the 4C group, the abundance of

Moraxella was not anticipated since infections with Moraxella are not as common in adults

231

[121]. However, a recent study on the upper airway microbiota in the elderly indicated that

Moraxella was part of the oropharynx and anterior nares microbiome in both adults and elderly

[129]. The age of the adult patients in the 4C group was 42-76 years. As such, the abundance of the Moraxella OTU in these patients suggested that upper respiratory tract microbiota or their products might be implicated in these systemic infections.

The abundance of the Bacillaceae OTU in the 4C branch was surprising as it was always identified in conjunction with the Staphylococcus OTU (Figure 5.4). Co-infections with

Staphylococcus and Bacillus species are not considered as a well documented human polymicrobial infection [247]. Although the Bacillus OTU could be a result of contamination, its lack of abundance in the HB samples (Figure 5.4) would suggest that it is likely not a result of any manipulation of the blood, DNA extraction regents, PCR contamination or Illumina processing. A recent infection control study examined the microbial composition of air in South

African hospitals [248]. In this study, Bacillus were identified as part of the air microbiota and the authors cautioned that this organism could be of concern in immunosuppressed populations

[248]. As such, it could be possible that Bacillus species are present within hospital air supplies in Calgary, however, their role as a clinical pathogen remains unclear.

Interestingly, four of the ICU neurological trauma control (SN) patients had bacterial

DNA patterns that clustered with the septic ICU and ED patients thereby suggesting these patients potentially had bacteremia. The European Prevalence of Infection in Intensive Care study indicated that 45% of patients in European ICUs were considered infected, with almost half of these infections acquired in the ICU [22]. Indeed, the ICU has the greatest use of invasive procedures and therapies that increase the risk of possible exposure to bacteria during admission

[22]. As such, it would not be unexpected that these SN patients were bacteremic but likely non-

232

septic. Two of the SN patients clustered with the 3B patients. ASN468 was admitted to ICU following the surgical removal of their spleen (Table 5.6). This patient was not considered to be septic however they were at a greater risk of acquiring a nosocomial infection, as individuals without a spleen are susceptible to infections with encapsulated bacteria [51]. Given the fact that some Streptococcus are encapsulated bacteria that avoid complement mediated killing [27], the risk of this patient acquiring a nosocomial Streptococcus infection, as suggested by the bacterial

DNA profile, was not unreasonable. The second SN patient, ASN468, was admitted with an intracranial aneurysm (Table 5.6). Recently, a link between bacterial mediated inflammation and the rupturing of intracranial aneurysms was evaluated [249]. This study reported that bacterial

DNA could be recovered from over 50% of the ruptured aneurysm samples assessed [249]. This group established that oral microbiota were found in the intracranial aneurysm samples with S. mitis group bacteria being the most common [249]. In addition, animal studies have revealed that there is an immunosuppressive inflammatory response in mice after the induction of a stroke

[176]. In these studies, disseminated bacterial infections were identified in mice within 8 hours following the stroke event [176]. The results from such studies combined with the presence of the Streptococcus OTU in the whole blood bacterial DNA profile (Figure 5.4) for this patient suggested that a Streptococcus bacteria or bacterial products bacteremia was plausible.

There were also two SN samples within the branch 4B (ASN462) and 4C (ASN487) groups on the UPGMA tree (Figure 5.4) suggesting a Staphylococcus bacteremia was present.

There was evidence that both of these SN patients acquired nosocomial Staphylococcus infections based on their positive clinical diagnostic blood culture results on day 14 (Table 5.6).

The presence of Staphylococcus infection in brain trauma patients has been well documented

[250-253]. As such, these patients may not have presented with a bloodstream infection on the

233

first day of ICU admission, based on clinical diagnostic blood culture, but the bloodstream bacterial DNA profile suggested they likely had Staphylococcus bacteria or bacterial products present. Collectively, these results indicated the difficulty in trying to identify non-infected ICU controls. Indeed, although these patients may not have had outward indications of infection using clinical diagnostic culturing methods, the clinical presentations combined with the bacterial

DNA profiles present in their blood suggested bacteremia was present at the time of sample collection.

There were also instances in which the bacterial DNA profile in whole blood could not be correlated to the patient’s clinical presentation. This was often the case for Enterococcus infections or urinary tract infections with E. coli or P. aeruginosa (Table 5.6). The reasons for this were unknown but may reflect the differences in immune evasion ability between the organisms or viability in blood. Other molecular profiling studies have indicated an underrepresentation of E. coli but for reasons unknown [152]. Indeed, there was poor recovery of E. coli in the mock community analysis (Figure 3.7), which was thought to possibly be a result of autolysis when the samples were treated with hypotonic washes [168]. The inability to detect P. aeruginosa may have been a result of the rapid clearance of these organisms by liver Kupffer cells [51]. Both blood culture and PCR based methods have no documented difficulties in recovering Enterococcus from blood [254]. In addition, Enterococcus DNA was recovered in

ASN469 during the time course sampling (Figure 4.6). As such, the lack of recovery in the whole blood samples might also be attributed to rapid immune clearance of Enterococcus and not due to a limitation in the methodology. In addition, patients often had a positive VRE rectal or nasal swab, which indicated carriage (Table 5.6). Despite this, these patients rarely had VRE

234

present in the blood suggesting that although there was carriage of the organism, this did not imply it would be implicated in the bloodstream infection.

In summary, the evaluation of the molecular profiling strategy on samples from patients presenting with sepsis-like symptoms in the ED at Foothills medical centre and the Alberta

Children’s Hospital indicated that only a minority of these patients could be analyzed using the criteria established in chapter four (minimum sequencing depth of 500). For the patients in whom the criteria could be applied, the bacterial DNA patterns identified in their blood was supported by their clinical data for the majority of the EB samples or by the epidemiological patterns present in paediatric patients for the CB samples. Once all the clinical samples were analyzed together on a UPGMA tree, the bacterial DNA patterns present in the patient cohorts were all clustered separately from the HB samples. In addition, the different branch groups were analyzed alongside the patient’s clinical data demonstrating trends between the source of infection and the bacterial DNA patterns. Taken together, a greater role for the commensal microbiota was suggested and Streptococcus as well as Staphylococcus species (viable bacteria or their products) were predicted to be major pathogens in septic bloodstream infections.

235

Chapter Six: Discussion

6.1 Overview of Findings

The primary focus of this thesis research was to develop methods for the direct detection of bacteria in whole blood using both culture and molecular approaches to improve the identification of bacteria in the blood and primary infection site of sepsis patients. The use of molecular profiling was predicted to improve the detection of polymicrobial infections. Lastly, through the examination of several patient cohorts and control groups, the molecular profiling data was hypothesized to enable the identification of common microbial patterns in the blood of sepsis patients with correlations to their clinical presentation. This was achieved through the development of a blood treatment protocol, DNA extraction protocol and molecular profiling strategies.

6.1.1.1 Method Development

Since whole blood is known to contain PCR inhibitors, the first objective in this study involved a strategy to separate microbial cells from whole blood [157]. The blood lysing detergent, saponin, was chosen for its documented ability to lyse host cells [255,256] without compromising microbial growth [93,161-164]. Results from ASN087 indicated that the use of saponin enabled the successful recovery and PCR amplification of bacterial DNA from whole blood (Figure 3.1) while FACS sorting suggested intact bacterial cells were present following the saponin blood-treatment (Figure 3.3). Following these results, the saponin blood-treatment protocol and DNA extraction method were evaluated using whole blood spiked with diverse

236

microbial constituents (Table 2.1). A panel of bacterial isolates recovered from sepsis bloodstream and primary infections (Table 2.1) repeatedly showed no statistically significant decrease in growth after a one-hour incubation with saponin (Figure 3.2). Additionally, mock communities generated with the same bacterial isolates were spiked into whole blood from healthy donors. The results indicated that the saponin lysis and hypotonic washes did not inhibit the recovery of bacteria from a mock community spiked into whole blood (Figure 3.6, Figure

3.7).

Data from the DNA extraction method validation indicated that enzymatic lysis with both lysozyme and mutanolysin resulted in the best recovery of microbial DNA diversity when compared to lysozyme alone or no-enzymatic lysis (Figure 3.4). The method validation also compared a magnetic bead based purification method to a column method. The recovery from the column purification process resulted in more consistent and representative microbial DNA following blood-spiking experiments (Figure 3.5).

The paired-end Illumina profiling of DNA recovered from the defined synthetic community (SC) of bacteria spiked into healthy blood indicated that the total microbial diversity present in the sample was recovered (Figure 3.10). The recovery of the mock community organisms was optimal when the saponin blood-treatment was followed up with two washes using DNase/RNase free distilled water (Figure 3.10). It was also evident that the PCR amplification paralleled the culture results (Figure 3.7) with an enrichment of the

Enterobacteriaceae OTU and Staphylococcus OTU (Figure 3.10). Overall, the evaluation of the

DNA extraction method was successful. This method has been applied successfully to other microbial community studies [100,101,129,238]. The blood spiking experiments demonstrated

237

that this method could be successfully applied to whole blood samples following the addition of a saponin blood cell lysis treatment.

The molecular profiling of HB samples indicated there was the potential for background levels of bacterial DNA present in blood (Figure 4.3). Despite this, these bacterial DNA patterns were distinct from the clinical samples (Figure 4.4) including the ICU controls (Figure 5.4). This

DNA was attributed to a combination of contamination from the skin during venipuncture and contamination from the blood collection system.

6.1.1.2 Application to Clinical Samples

For the last three objectives, clinical samples from Foothills Medical Centre and the

Alberta’s Children Hospital, Calgary, Alberta, Canada were collected and processed (Table 2.3).

In order to compare the molecular profiles, the OTU tables were assessed using α-diversity and

β-diversity measures. Jackknife support values were calculated for 10 tables subsampled to a depth of 500 sequences. The bacterial DNA profiles from 54 adult septic ICU patients clustered into three phylogenetic groups on a UPGMA tree (Figure 4.1A). These data suggested there were common bacterial DNA patterns present in the whole blood of sepsis patients.

In order to determine what these patterns signified, the clinical data for the ICU patients was examined. Of the many clinical variables measured (age, sex, mortality, SIRS criteria,

APACHE II score, SOFA score, length of stay, source of infection, and admitting diagnosis), associations with predicted infection source (based on the admissions diagnosis), mortality, and gender were observed (Table 4.3). The patients with a Group 1 bacterial DNA profile in the whole blood represented commensal microbiota from the upper respiratory tract or the skin in

238

addition to Streptococcus as the predicted pathogen (Table 4.3, Figure 4.1). The patients with a

Group 2 bacterial DNA profile presented mainly with a gastrointestinal infection or trauma suggesting that they developed sepsis mainly as a result of released gastrointestinal microbiota

(Table 4.3, Figure 4.1). However, the patients within Group 2 with abscess infections or airway infections had bacterial DNA profiles that correlated to upper airway and skin associated microbiota (Table 4.3, Figure 4.1). Lastly, emergency surgical interventions were often associated with a Group 3 bacterial DNA profile with Staphylococcus as the predicted pathogen

(Table 4.3, Figure 4.1). Overall, the bacterial DNA profiles correlated to the patient’s source of infection, with the Group 3 bacterial DNA profile being more common in males than females, and patients with a Group 1 or Group 3 bacterial DNA profile having a greater risk of mortality in the ICU.

When the bloodstream bacterial DNA profiles were compared to the primary infection

DNA profiles (obtained from the same patient), the majority of patients shared at least one OTU

(Figure 4.7). Despite this, not all OTUs identified in the primary infection molecular profile were present in whole blood. This suggested that the bacterial DNA patterns present in blood were not entirely derived from the primary infection. Indeed, the inflammatory response to infection indirectly increases the risk of bacterial entry into the bloodstream through its effects on tight junction integrity and the catecholamine-mediated increase in enteric microbial growth [52].

However, once bacteria enter the bloodstream they must be able to evade the intravascular immune system comprised of neutrophils, monocytes, iNKT cells, and Kupffer cells [257].

Bacteria have evolved numerous strategies to evade the intravascular immune system. For example, bacteria that produce DNases, including S. pneumoniae, are able to escape NET- mediated killing in the vasculature [258,259]. Bacteria including S. aureus and E. coli are able to

239

bind to the vascular endothelium and this binding is enhanced due to the expression certain receptors during inflammation [260,261]. Further, successful intravascular pathogens also produce various exotoxins and proteases that can block recruitment of immune cells or the detection of bacteria [262]. Interestingly, both Staphylococcus and Streptococcus have developed many of the immune evasion strategies outline above, which may explain the propensity for their

DNA recovered in blood (Figure 4.1). As such, when the bacterial DNA profiles from blood were compared to the primary infection site, the shared taxonomic groups likely indicated the organisms that had successfully translocated and could persist in the vasculature. Alternatively, in patients where no correlation was seen between the infection sample and blood, the inflammation induced by the primary infection may have reduced mucosal integrity thereby permitting the leakage of bacterial products from the gastrointestinal tract or respiratory tract that are usually associated with the healthy microbiota. Although bloodstream infections are often associated with a primary infection, up to 10% of bloodstream infections cannot be attributed to any known source [198]. As such, a primary infection may not be a prerequisite for sepsis whereas the inflammatory response to infection may increase the risk of bacteremia developing from the reduced mucosal integrity and leakage of bacterial PAMPS into blood that can subsequently enhance the inflammatory response.

Time course sampling of a subset of ICU patients suggested that the bloodstream molecular profile shifted during the ICU stay and that bacterial DNA could be detected up to 28 days after ICU admission (Figure 4.6). Shifts in the bacterial DNA patterns also suggested that bacteremia was dynamic such that bacteria, or their products, may not be consistently present within the bloodstream. Indeed, given the intravascular immune system and the hepatic clearance of intravascular bacteria mediated by Kupffer cells, a cycle of infection and clearance is

240

predicted in blood [257]. These transient episodes of bacteremia could result in an accumulation of inflammation and a SIRS. Moreover, the high failure rate of diagnostic blood culture may be partially attributed to the intermittent nature of the presence of bacteria in blood.

The UPGMA clustering of all the clinical samples from ICU and ED patients showed that there were two principal bacterial DNA patterns; one in which Streptococcus was dominant and one in which Staphylococcus was dominant. Indeed, out of 79 samples, 24 samples were found in branches 1 and 3B indicating 30% of the clinical bacterial DNA patterns were dominated by

Streptococcus, whereas 32 samples were found in branches 4A-4D indicating 41% of the bacterial DNA profiles were dominated by Staphylococcus (Figure 5.4). In total, this represented

70% of the clinical samples. For the remaining samples in which there was recovery of mixture of DNA, Groups 3C and 4E (Figure 5.4), these may reflect instances in which there was an increased gastrointestinal or airway mucosal permeability possibly due to an increase in inflammatory cytokines [212]. This was predicted to increase the release of bacterial products, many of which are known PAMPs, into the bloodstream that could initiate an inflammatory response and sepsis. In such instances, the presence of several taxonomic DNA groups from many bacteria would be expected. The remaining 3 samples represented unique cases of systemic infections with Finegoldia or Serratia in which the molecular DNA profiling correlated well with the clinical data and diagnostic blood culture data. Lastly, there were 13 samples in which the bacterial DNA profile could not be correlated with the patient’s clinical presentation.

Collectively these findings would suggest that in 16.5% of the cases, the molecular profiling data could not be used to imply a systemic infection was present. Overall, the bacterial DNA profiling of whole blood samples from adult and paediatric patients was correlated to a predicted

241

bloodstream infection with either a viable organism or bacterial products in 75% of the samples analyzed in this study.

6.2 Challenges and Limitations

With respect to the culture done in this study, there were some limitations to this method include the limited success with recovering anaerobic bacteria due to manipulations done under mainly atmospheric oxygen until the cells were returned to the anaerobic chamber (Figure 3.7).

This could be circumvented in the future by performing all steps in the saponin blood-treatment in an anaerobic chamber. Another limitation in the cultivation-dependent approach was the low recovery of E. coli and N. flava in in the blood spiking experiments (Figure 3.8). This was possibly due to potential bacteriostatic and bactericidal effects of whole blood [167]. There was also the enrichment for certain organisms, E. hormaechei and S. aureus, in the cultivation- dependent approach that was not anticipated. Although our approach was mainly molecular- based, these findings revealed the diversity in microbial viability and recovery in-vitro. Even without a broth based enrichment step, which is the premise of blood culture, there was still unequal recovery of the mock community over repeated experiments despite efforts to control the initial concentrations.

There were some limitations to the DNA extraction method validation in this study. The first was that the lysostaphin enzyme was never assessed. Initially, since the recovery of the

Staphylococcus OTU in the paired-end Illumina was abundant, the use of this enzyme did not seem to be warranted. The ease of recovering Staphylococcus DNA was also supported by other studies comparing DNA extraction methods on mock communities using pyrosequencing [173].

242

This study also used S. aureus and S. epidermidis in their mock communities but did not delineate if the Staphylococcus OTU was representative of both organisms. Using the paired-end

Illumina approach it was possible to examine if the Staphylococcus OTUs had alignments to S. aureus or other CoNS. Interestingly, there was limited recovery of the OTU with an alignment to

S. epidermidis with our method (Table 3.3, Table 4.1). This could have been a result of several factors including PCR bias for S. aureus, selection of the OTU consensus sequences, or lack of S. epidermidis lysis. In order to evaluate if the addition of a Staphylococcus specific enzyme would improve the recovery of S. epidermidis, the mock communities could be assessed with lysostaphin added as well and subsequently sequenced to examine if there is better recovery of S. epidermidis DNA. To assess if there was PCR bias for amplification of S. aureus DNA over S. epidermidis DNA, mock communities without S. aureus could be sequenced using paired-end

Illumina and the recovery of Staphylococcus DNA could be assessed. Indeed, if there were poor recovery of Staphylococcus OTUs it would suggest that there was insufficient lysis of S. epidermidis resulting in no DNA template. However, if there was Staphylococcus DNA recovery than this would suggest there was a PCR bias for the S. aureus DNA in the amplification of the

16S rRNA gene prior to the paired-end Illumina sequencing. Most studies have focused only on

S. aureus when assessing DNA extraction protocols [26,156]. Consequently, it may be important to evaluate the method to ensure the recovery of S. epidermidis DNA, as this is becoming an increasingly important organism in nosocomial sepsis infections [26].

The blood spiking experiments with the mock communities indicated that the V3 primers used in this study had cross-reactivity to non-bacterial, likely human, DNA. This “noRoot” OTU was attributed to the well-documented erroneous amplification of human DNA in clinical samples with universal 16S rRNA gene primers. This issue has been reported since the early

243

days of PCR [170] and is still problematic in contemporary 16S rRNA gene studies

[151,170,171]. In this study, the abundance of “noRoot” DNA often represented a large portion of the amplified sequences in whole blood. This was unique to our study and likely reflected the low ratio of bacterial to host DNA in these samples. It is difficult to know the exact concentration of bacteria in bloodstream infections since the blood culture results only indicate the CFU/ml of bacteria after a broth-enrichment. However, in the limited number of samples where culture from saponin treated whole blood was successful, the CFU/ml were between 1 to

30 (Table 5.1) suggesting the concentration of bacteria would be low in the clinical samples. The saponin blood-treatment method was used to help lyse blood cells and minimize the presence of extracellular human DNA from these lysed cells in the samples by sequentially removing the supernatant, the results indicated that not all human DNA was removed. This may have been a result of cellular debris pelleting with the bacterial cells while the lysed blood cells were present in the supernatant. Another possibility would be the recovery of membrane-limited extracellular vesicles that are found in many human body fluids including blood serum [263]. There are three types of these membrane-limited extracellular vesicles of which microvesicles were the most likely to co-purify bacterial cells in the saponin-treated blood. This is due to their similarity in size to bacteria, their isolation using differential centrifugation at 18,000-20,000 rcf, and their best-known sources are red blood cells, platelets, and endothelial cells [263]. The ability of saponin to lyse these membrane-limited extracellular vesicles is unknown. As such despite the removal of blood cells, some human DNA still remained in these preparations and when the bacterial template was low, this DNA contributed to a significant proportion of the taxonomic profiles.

244

The McMaster sequencing facility was able to remove the “noRoot” OTU from the OTU table using custom Perl scripts, which permitted downstream analysis of the bacterial sequences

[238]. Despite this, for future studies, the use of a different 16S rRNA gene primer set may be warranted. There are nine hypervariable regions in the 16S rRNA gene [223]. Of these, the V6 and V3 regions are often selected for pyrosequencing or Illumina based profiling studies [264].

Both the V3 and V6 regions have conserved flanking regions and an optimal length for the currently used next-generation platforms [264,265]. However, for paired-end Illumina, the V3 region was selected as it had better taxonomic resolution and longer length when compared to the V6 region [128,265]. The V3 region primers have 99% accuracy at the genus level whereas the V6 have 97% accuracy [265]. Despite this reduced accuracy, the use of V6 primers rRNA primers could be assessed to see if they have reduced cross-reactivity to human DNA. Another primer set that could be assessed are 16S rRNA primers for amplification of the V8 and V9 region [266]. These primers were used in the mock community assessment outlined above and on clinical BAL samples with the authors reporting no cross-reactivity to human DNA [173]. As such, it may be beneficial to examine these alternative primer sets for use on blood samples to improve the primer specificity for the bacterial DNA. This would come at a potential loss in taxonomic resolution, which would need to be examined as well. Indeed, since it is possible to filter out the “noRoot” OTU, this could be tolerated if the alternative approach diminished the sensitivity of the sequencing.

Whole blood from healthy adults was collected as these samples represented the baseline for the molecular profiling of blood and they should, theoretically, be sterile samples. As such, it was predicted that the HB samples would not have bacterial DNA present or that any DNA present would represent laboratory-based contamination. Given the universal nature of the 16S

245

rRNA gene primers, any DNA present in whole blood or during the PCR processing would be amplified and present in the taxonomic profile for the HB samples. Several steps were taken to reduce the amount of laboratory derived DNA contamination including the UV-irradiation of the

PCR buffers (not the Taq polymerase), the DNase/RNase double-distilled water, the 5% (w/v) saponin solution, the PBS, the PCR reaction vials, and the PCR preparation tubes. In addition, all surfaces and gloves were cleaned with a bleach-based solution prior to use. All pipette tips were aerosol-barrier tips certified to be free of all DNA/RNA contamination. These steps seemed to reduce the impact of laboratory-derived contamination since the HB samples had limited taxonomic overlap in their bacterial DNA profiles with the PBS/NTC samples (Figure 4.4).

A concern with the HB profiles was that the OTUs recovered were also identified in the clinical samples (Figure 4.3) and did not reflect common sources of contamination DNA such as

Escherichia from the Taq polymerase or Stenotrophomonas from the environment

[173,267,268]. As such, there was the possibility that the DNA recovered could be partially attributed to bacteria present in the blood of healthy adults. There is little information in the literature examining the abundance of bacteria in the blood of healthy individuals. However, one study has previously visualized pleomorphic bacteria, in healthy whole blood samples [269].

Using PCR and FISH techniques, this group also recovered bacterial DNA in clinically healthy individuals with cloned PCR fragments generating a sequence identified as a Stenotrophomonas species [269]. The Stenotrophomonas OTU was not present in the HB samples from this study.

Stenotrophomonas has previously been recovered in other molecular profiling studies, although, was dismissed as environmental contamination [173]. Another study detected bacterial DNA in whole blood collected in EDTA vacutainers [270]. Using 16S sequence clone-libraries, the study indicated Rimerella, Bacillus, Microbacterium, Propionibacterium, Acidovorax,

246

Stenotrophomonas, and Pseudomonas DNA was recovered [270]. Once again, these organisms were not identified as major components of the HB bacterial DNA profiles (Figure 4.3). Lastly, none of the clinical samples clustered with the HB samples (Figure 5.4) further suggesting the bacterial DNA signature in HB represented low level contamination that were not observed in patients with bacterial infections. Indeed, since the taxonomic profiles are based on relative abundance, the presence of DNA associated with the sepsis infection would lower the relative abundance of any contaminating DNA. Consequently, there was limited evidence to support the

HB DNA profiles representing bacteria present in healthy blood. Rather, it was more plausible that insufficient antisepsis during the venipuncture combined with the likely presence of DNA in the collection vials resulted in the amplification of minute levels of DNA present during the procedure. Indeed, phlebotomy may account for the human-associated bacterial DNA present in the HB samples (i.e., Staphylococcus, Escherichia, Enterobacteriaceae) as contamination in venipuncture due to skin associated bacteria recovered from injection needles [271,272] and the inability of antisepsis to eliminate all bacteria or their DNA [273] have been documented. Taken together, the bacterial DNA profiles in the HB were derived from a combination of contaminant

DNA from the phlebotomy and possible contamination from the blood vacutainers or their components (i.e., Luer-Lok™ Access Device). Indeed, when developing a robust DNA extraction protocol and implementing a highly sensitive PCR-based detection protocol, there are challenges associated with background noise. Despite these challenges, the protocol demonstrated it could recover bacterial DNA from small volumes of whole blood and that the bacterial DNA profiles recovered from septic patients were phylogenetically distinct from those present in our healthy controls.

247

Another limitation was that the bacterial DNA profiles reflected relative DNA abundance and not total DNA abundance. This meant that no conclusions could be made in regards to the quantity of bacterial DNA in these samples. Indeed, it was unknown if the quantity of bacterial

DNA in HB samples was different from SB samples. Attempts to quantify the bacterial load using RT-PCR were unsuccessful due to PCR amplification of the human DNA in the samples.

As such, the bacterial DNA profiles could indicate the taxonomic diversity in each sample with

OTU abundance reflecting ratios of each OTU in the sample not the bacterial load. Further to this, the sequencing read cut-offs used in the evaluation of molecular patterns (500 reads per sample) may have resulted in a bias towards more common taxa and a loss of rare taxa. Due to the low number of reads in many of the samples, these cut-offs were essential to ensure there was no over-estimation of diversity in light of higher levels of contamination observed in the HB samples. Nevertheless, the principal focus of this study was to develop a molecular approach towards identifying the major organisms of clinical importance that may not always be identified using clinical diagnostic blood culture. In this capacity the study was successful and evaluation of rare taxa could be evaluated as a future objective.

Based on these limitations, it was essential that each sample be evaluated within the clinical context. Many reviews of molecular profiling strategies have highlighted the importance of analyzing molecular data in conjunction with other clinical measures of severity (i.e.,

APACHAE II, SOFA scores), markers of infection (i.e., Procalcitonin), and markers of inflammation (i.e., IL-6, IL-10) [5,102,111,274]. Indeed, when the bacterial DNA patterns were aligned with the clinical data it was apparent that meaningful patterns were observed in the data despite the limitations to this method.

248

6.3 Future Directions

Based on some of the limitations outlined above, the future experiments could be those envisioned to better optimize this methodology. As mentioned above, this would include the assessment of lysostaphin in the enzymatic cocktail used in the DNA extraction as well as evaluating different primer sets to reduce the cross-reactivity to human DNA. Lastly, it would be beneficial to evaluate our methodology alongside some of the currently used DNA extraction methods developed for sepsis such as the MolYsis kit and Looxster kit. This would allow for the comparison of methods to determine if the elimination of human DNA is beneficial or not. Our method did not add in a human DNA degradation step whereas the MolYsis and Looxster systems both integrate this step into their protocols.

Since this study indicated there were common bacterial DNA patterns recovered in whole blood, it would be beneficial to survey more patients to see if these patterns are consistent. In the next phase of sampling better control patients would need to be identified as the ICU control patients were at risk for bacteremia thereby limiting our interpretations of their molecular DNA profiles. Sampling of in-ward patients, not ICU patients, may be a better control group. Another control group would be patients attending the outpatient clinic receiving intravenous antibiotic treatment for an infection. Although samples were collected from this group, there were not enough available for analysis for this thesis work. As such, expanding the sampling from these patients would be beneficial to see if there is a difference in the bacterial DNA patterns of patients who are predicted to have non-systemic infections and are healthy enough to not be admitted. It would also be interesting to obtain blood samples from patients following dental procedures ranging from basic tooth cleaning to more invasive dental surgeries. Since the oral

249

microbiota is implicated in transient bacteremia [48], the bacterial DNA patterns from such samples may indicate if there are differences in bacterial DNA patterns associated with low risk bacteremia and those in sepsis.

The results in this thesis suggested that bacterial DNA in whole blood could be correlated to the patient’s clinical presentation. As such, it would be beneficial to expand this technique to measure bacterial load based on the abundance of bacterial genes in blood. Although there was limited success in using RT-PCR to measure bacterial load in these samples, further attempts at optimizing this methodology are warranted. Perhaps the use of different primer sets, not 16S primers, could be explored. There has been success in measuring DNAemia using non-16S genes to quantify and monitor bacterial infections [51]. For example, the use of lytA, a S. pneumoniae specific autolysin gene, was used to monitor the risk of culture-confirmed bacteremia in patients with community-acquired pneumoniae [114]. The study concluded that there was a correlation between elevated gene copies of lytA in the bloodstream and a diagnostic clinical blood culture confirmed infection with S. pneumoniae [114]. Similar findings have been confirmed in two subsequent studies using gene copies as markers for S. pneumoniae infections [275,276]. Further, the use of mecA (conferring methicillin resistance) gene copies to measure a patients response to anti-MRSA antimicrobial therapy showed an inverse correlation between the number of gene copies of mecA and successful antibiotic therapy in cases of clinical diagnostic blood culture confirmed MRSA bacteremia [277]. A similar study was done in which the levels of the oxa-51 gene (conferring resistance to carbapenems) were measured in survivors and non-survivors of A. baumannii diagnostic blood culture confirmed bacteremia [278]. This study demonstrated that increased copies of the oxa-51 gene correlated with poor clinical outcome in the patients surveyed [278]. As such, measuring bacterial DNA in the bloodstream has provided valuable

250

clinical insight and will likely continue to develop into a viable diagnostic tool. The limitation of these studies was a dependence on culture to determine which gene targets should be used.

However, if the paired-end Illumina 16S rRNA gene profiling in blood was used to first identify which bacterial DNA was present in blood, the dependence on culture would be mitigated.

Following taxonomic profiling, the use a targeted approach to could be used to quantitate the microbial load using genes specific to the clinically relevant pathogens. This could also be expanded to identify genes that encode antibiotic resistance mechanisms. Taken together, a two- step molecular approach to identifying bloodstream infections could be developed through the combination of bacterial gene profiling and RT-PCR amplification of targeted genes of interest.

In addition, exploration of other bacterial products as well as host products in blood should be examined. Collaborators with the ASN project have examined the metabolomic profiles from the SB patients included in this study. In addition, there is luminex-based data that measured the activity of inflammatory mediators in the blood of these SB patients. Aligning the bacterial DNA profiles with the bacterial/host metabolomic profiles, and host inflammatory mediator profiles would provide a robust analysis of the role of bacterial products and the host responses in sepsis bloodstream infections.

Stratification of septic patients based on bacterial DNA and other bacterial products (e.g.

LPS, peptidoglycan) might also be useful in guiding therapy that would focus on reducing mucosal permeability. Indeed, for the patients in whom leakage of microbial products, including bacterial DNA, were predicted to cause the sepsis inflammatory response, therapy to improve mucosal barrier function may be complementary to antibiotic therapy. For example, elevated levels of myosin light chain kinase have been implicated in the TNF-induced reduced barrier function in epithelial and endothelial cells [212]. Therefore, an alternative sepsis therapeutic

251

could be drugs targeted at reducing myosin light chain kinase activity or production. In addition, for patients with a primary pathogen and host-microbiota associated bacteria implicated in systemic infection, therapy with both antibiotics and anti-myosin light chain kinase may have better therapeutic outcomes in reducing the pathogen load as well as mitigating the translocation of microbiota or bacterial products.

6.4 Clinical Implications

This study used a custom-developed DNA extraction protocol that has been evaluated in numerous studies within our research group [129,238]. This method has consistently provided more robust and consistent results when compared to commercially developed kits (M. Surette, personal communication). Many of the reviews of current molecular diagnostics highlight the need for a ‘gold-standard’ in DNA extraction from clinical samples [5,102,222]. Since our method has been successfully evaluated in clinical infections and in healthy human samples, it may serve as a benchmark for the development of a standardized DNA extraction method.

Further, with the addition of the saponin blood pre-treatment, this method could be developed as a ‘gold-standard’ for sepsis molecular diagnostics. Nevertheless, some of the limitations addressed with this method and automation of the method would need to be explored prior to any integration into a clinical diagnostic laboratory.

Despite the challenges outlined in the interpretation of the molecular profiling data, this study still provided novel insights into sepsis infections. In particular, this study demonstrated the validity of using a blood-cell lysing detergent, saponin, in the initial treatment of whole blood. Saponin provided an alternative to the current approach, addition of SPS to clinical media

252

or blood vacutainers, to mitigate the bactericidal and bacteriostatic [167] effects of blood. SPS is a potent PCR inhibitor and co-purifies with DNA [106] making its removal from DNA preparations difficult and laborious. Indeed, this study showed that saponin did not impact bacterial viability and could be used directly on whole blood to permit the recovery of intact bacterial cells from mock communities. In the limited clinical assessment done using a culture- dependent approach, this study also recovered bacterial cells from small volumes (1-2ml) of whole blood whereas blood culture requires 10-fold higher volumes to ensure efficacy. The use of saponin did not impact the ability to PCR amplify DNA recovered from whole blood. As such, this study indicated a novel use of saponin for the pre-treatment of whole blood prior to a culture-dependent or culture-independent assessment of whole blood. The potential for the use of saponin in this context would be easily implemented in clinical diagnostic labs as saponin is already used as a constituent of blood culture media [162,164]. It has already proven advantageous in studies using MALDI-TOF MS [93,163] and could continue to be developed as a ‘gold-standard’ for whole blood collection. An extension of this idea would be the development and usage of a saponin-coated blood collection vacutainer, which is currently not commercially available. This would allow for the immediate lysis of whole blood cells upon collection. Following this, a clinical diagnostic laboratory could easily centrifuge these tubes to separate bacteria from lysed blood cells to use for downstream molecular and culture diagnostics.

This study further demonstrated that small volumes of whole blood could be used for both the culture-dependent and culture-independent approaches. This would be advantageous over the current system in which up to 30-40ml of blood is collected per sampling of an adult patient [5]. In majority of ICU patients, blood cultures are taken several times during a patient’s admission, often without added clinical utility [279]. This repeated sampling increases the costs

253

associated with a patients hospital care [280] but also increases the patients discomfort from repeated blood draws. In a paediatric setting, a major hurdle in sepsis diagnostics is the blood volume requirements [94]. The development of saponin-coated vacutainers would be highly advantageous in the paediatric setting by reducing the volume of blood needed. Indeed, it would provide peace of mind to parents who may feel uncomfortable with the volumes of blood currently required for blood culture. Taken together, the development of a saponin-coated vacutainer to collect 4-6ml of whole blood from adults and children may provide an alternative or supplement to clinical diagnostic blood culture.

There is scepticism about the application of culture-independent approaches in sepsis with several reviews of PCR based diagnostics indicating that DNAemia may not be synonymous with a blood-stream infection [5,96,111,274]. DNA may be present in the bloodstream as a by-product of infection or could result from dead cells or be a result of contamination during processing of blood samples [96]. Indeed, even though both of these concerns were challenges experienced in this study, the evaluation of the clinical data enabled meaningful interpretations of the bacterial DNA patterns identified. Further, our taxonomic profiling can be combined with more targeted approaches to measure bacterial specific genes and approximate microbial load in blood. In addition, there is no definitive data indicating that the presence of bacteria in the bloodstream must remain constant in sepsis nor is it absolutely required for a patient to be considered septic [3]. Indeed, the high failure rate of blood culture would suggest that circulating bacteria are not always present [51]. Moreover, it is well known that bacterial products alone can elicit strong inflammatory responses. The Gram-negative LPS is known as a potent bacterial PAMP and the presence of LPS in the blood can result in sepsis [53].

Bacterial DNA, itself, is also a PAMP and is known to activate macrophages [53]. As such, the

254

results from this study suggested that bacterial DNA patterns in the blood should not be dismissed when there are clinical indications of SIRS, but no evidence of a viable organism in blood, since bacterial products alone can initiate an inflammatory response.

This study highlighted that the paired-end Illumina 16S rRNA sequencing approach was able to identify polymicrobial bacterial DNA profiles in septic patients. These infections were identified in whole blood and in primary infection samples from ICU patients as well as adult and paediatric ED patients. Although there were challenges to the interpretation of the data, there were several important findings that could have important clinical implications. The first was that the majority of the bloodstream bacterial profiles represented polymicrobial DNA. The results also indicated that there was a probable role for commensal bacteria in sepsis as DNA representing microbiota-associated organisms was often present in the bacterial DNA profiles from blood. The third was that majority of sepsis whole blood samples clustered into two groups indicating there were common bacterial DNA patterns present with Streptococcus and

Staphylococcus predicated to play a much larger role in the bloodstream infection when compared alongside the clinical diagnostic blood culture results from these patients. Taken together, these results could be used to tailor antimicrobial therapy for bloodstream infections to target these organisms. In addition, some non-traditional pathogens were also detected in the septic patients including S. marcescens, F. magna and other GPAC which may warrant monitoring as the rate of GPAC infections have been increasing [30]. Overall, these results provide some novel insights into sepsis infections, which were not observed using the clinical diagnostic blood culture techniques.

Although only a limited number of case studies were described, the temporal changes in the bloodstream bacterial molecular profiling provided some interesting insight. In each profile

255

at least one OTU showed a steady decrease in relative abundance from days 1-14 (Figure 4.6).

This suggested that there was a detectable response to antibiotic treatment, which could be tracked using molecular profiling. There were also instances in which the molecular profile suggested bacteremia might not have resolved completely as there was bacterial DNA corresponding to the clinical blood cultured organism detected several days prior (Figure 4.6).

As such, the use of bacterial molecular profiling in blood may provide a unique approach to monitoring responses to antimicrobial therapy and could be used to examine the risk of nosocomial infections in sepsis patients. In addition, other non-septic ICU patients are at risk of acquiring nosocomial infections and may also benefit from molecular profiling if their ICU treatment extends beyond a few days.

Currently, the paired-end Illumina approach outlined in this study would not be ideal for a clinical diagnostic laboratory. The time between receiving a blood sample to the final sequencing result would take three days; one day to process the sample, extract DNA, run the

PCR, one day to run the Illumina sequencing, and one day for the analysis. Another limitation is that the species level identification is still questionable given the short read-length generated.

Since many of the genera identified in sepsis are also part of the human microbiota, it would be essential to obtain a definitive species identification if this method were to supersede blood culture diagnostics. However, the MiSeq Illumina platform has the greatest promise for clinical applicability. It has a single-day turn around time for each run, it can handle 2167 samples using all bar-coded primers, it has good taxonomic depth with paired-end reads, a cost of $800-1500 per run and it generates significant sequence results with 1.5Gb of data from 5 million reads

[128,281]. In addition, the continually decreasing per-base sequencing costs and the increasing ability to multiplex thereby increase throughput are also advantageous [122,281]. This approach

256

is also less labour intensive and time consuming thereby making it more cost-effective. It is a more objective identification tool that is not affected by phenotypic variation or technologist bias, which would decrease laboratory error rates [122]. As such, as the read lengths continue to get longer and computers continue to evolve to handle these large datasets, the time to run and analyze a sample will be significantly decreased. Further, with the introduction of a saponin- coated vacutainer and automation of the DNA extraction procedure, it would be plausible that a microbiota profile could be available for a patient within 24 hours of admission. Combined with targeted gene amplification to examine microbial load or presence of antibiotic resistance genes, this method could very well be developed such that culture based diagnostic would no longer be required from blood. Indeed, there is a growing body of literature demonstrating the high failure rates and contamination issues [19,51,273] associated with clinical diagnostic blood culture. In contrast, many studies evaluating molecular approaches are showing success [102,282-284]. This puts into question the utility of clinical diagnostic blood culture and the suggested over- utilization of blood culture may warrant review [285]. Although each clinical diagnostic blood culture is relatively inexpensive, repeated blood draws can cause patient discomfort often without added benefit to patient care [286]. With the cost of a septic patients care already being high [11], the over utilization of clinical diagnostic blood cultures should be considered. It may be of greater clinical value to invest the funds towards implementation of molecular diagnostics that have demonstrated promising results.

257

6.5 Conclusions

The paired-end Illumina 16S rRNA community profiling of bacteria has been successfully applied to human clinical samples to provide a more robust evaluation of polymicrobial infections [128,238,281,287]. To our knowledge, this was the first study in which paired-end Illumina 16S rRNA gene community profiling was done on whole blood. The results of this study indicate that a saponin blood pre-treatment lysis steps combined with the paired-end

Illumina sequencing enabled molecular-profiling of small volumes of whole blood. This method was successfully validated in both adult ICU and ED cohorts as well as paediatric ED patients.

Taken together, the variation in success of culturing compared to the uniformity of the molecular profiling suggested that the recovery of viable bacteria in blood is not required for a patient to be septic. Indeed, there were numerous cases in which the molecular profiling of whole blood correlated with the clinical presentation of the patient in absence of a culture confirmed organism. It is important to note that the bacterial DNA profiles implicated bacteria such as

Staphylococcus and Streptococcus in the majority of patients. Members of these genera are not considered as traditionally difficult bacteria to cultivate. This indicates that the discrepancy between cultivation negative yet bacterial DNA positive blood samples could be attributed to an alternative concept of sepsis bloodstream infections: a cycle of transient acquisition and clearance of the bacteria whereas the presence of bacterial products remains consistent. In addition, some cases of sepsis may be driven by the presence of bacterial products not viable organisms in the blood, likely resulting from increased mucosal permeability, with further experiments needed to confirm this hypothesis. As such, the detection of bacterial products including DNA, RNA, proteins, and metabolites, may provide more insight into systemic sepsis

258

infections rather than attempts to recover viable organisms from a host system that has evolved to rapidly clear such infections from the tissue and vasculature.

259

References

1. Funk DJ, Parrillo JE, Kumar A (2009) Sepsis and septic shock: a history. Crit Care Clin 25: 83-101, viii. 2. Guillermo Ortiz-Ruiz MAP, Eugen Faist, and Carmelo Duenas Castell, editor (2006) Sepsis. Second ed. New York: Springer. 3. Bone RC, Balk RA, Cerra FB, Dellinger RP, Fein AM, et al. (1992) Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. American College of Chest Physicians/Society of Critical Care Medicine. Chest 101: 1644-1655. 4. Adrie J-MCaC, editor (2009) Sepsis and Non-infectious Systemic Inflammation. Weinheim: WILEY-VCH. 5. Mancini N, Carletti S, Ghidoli N, Cichero P, Burioni R, et al. (2010) The era of molecular and other non-culture-based methods in diagnosis of sepsis. Clin Microbiol Rev 23: 235-251. 6. Martin GS (2012) Sepsis, severe sepsis and septic shock: changes in incidence, pathogens and outcomes. Expert Rev Anti Infect Ther 10: 701-706. 7. Faix JD (2013) Biomarkers of sepsis. Crit Rev Clin Lab Sci 50: 23-36. 8. Rangel-Frausto MS, Pittet D, Costigan M, Hwang T, Davis CS, et al. (1995) The natural history of the systemic inflammatory response syndrome (SIRS). A prospective study. JAMA 273: 117-123. 9. Restrepo MI, Mortensen EM, Waterer GW, Wunderink RG, Coalson JJ, et al. (2009) Impact of macrolide therapy on mortality for patients with severe sepsis due to pneumonia. Eur Respir J 33: 153-159. 10. Husak L, Marcuzzi A, Herring J, Wen E, Yin L, et al. (2010) National analysis of sepsis hospitalizations and factors contributing to sepsis in-hospital mortality in Canada. Healthc Q 13 Spec No: 35-41. 11. Angus DC, Wax RS (2001) Epidemiology of sepsis: an update. Crit Care Med 29: S109-116. 12. Laupland KB, Davies HD, Church DL, Louie TJ, Dool JS, et al. (2004) Bloodstream infection-associated sepsis and septic shock in critically ill adults: a population-based study. Infection 32: 59-64. 13. Laupland KB, Gregson DB, Zygun DA, Doig CJ, Mortis G, et al. (2004) Severe bloodstream infections: a population-based assessment. Crit Care Med 32: 992-997. 14. Beekmann SE, Diekema DJ, Chapin KC, Doern GV (2003) Effects of rapid detection of bloodstream infections on length of hospitalization and hospital charges. J Clin Microbiol 41: 3119-3125. 15. Diekema DJ, Beekmann SE, Chapin KC, Morel KA, Munson E, et al. (2003) Epidemiology and outcome of nosocomial and community-onset bloodstream infection. J Clin Microbiol 41: 3655-3660. 16. Lim CJ, Cheng AC, Kong DC, Peleg AY (2014) Community-onset bloodstream infection with multidrug-resistant organisms: a matched case-control study. BMC Infect Dis 14: 126. 17. Martin GS, Bernard GR (2001) Airway and lung in sepsis. Intensive Care Med 27 Suppl 1: S63-79.

260

18. Sydnor ER, Perl TM (2011) Hospital epidemiology and infection control in acute-care settings. Clin Microbiol Rev 24: 141-173. 19. Brun-Buisson C, Doyon F, Carlet J (1996) Bacteremia and severe sepsis in adults: a multicenter prospective survey in ICUs and wards of 24 hospitals. French Bacteremia- Sepsis Study Group. Am J Respir Crit Care Med 154: 617-624. 20. Pittet D, Thievent B, Wenzel RP, Li N, Auckenthaler R, et al. (1996) Bedside prediction of mortality from bacteremic sepsis. A dynamic analysis of ICU patients. Am J Respir Crit Care Med 153: 684-693. 21. Valles J (1997) [Bacteremias in intensive care]. Enferm Infecc Microbiol Clin 15 Suppl 3: 8- 13. 22. Vincent JL, Rello J, Marshall J, Silva E, Anzueto A, et al. (2009) International study of the prevalence and outcomes of infection in intensive care units. JAMA 302: 2323-2329. 23. Bone RC (1993) Gram-negative sepsis: a dilemma of modern medicine. Clin Microbiol Rev 6: 57-68. 24. Peirano G, Pitout JD (2014) Fluoroquinolone-Resistant Escherichia coli Sequence Type 131 Isolates Causing Bloodstream Infections in a Canadian Region with a Centralized Laboratory System: Rapid Emergence of the H30-Rx Sublineage. Antimicrob Agents Chemother 58: 2699-2703. 25. Turton JF, Shah J, Ozongwu C, Pike R (2010) Incidence of Acinetobacter Species Other than A. baumannii among Clinical Isolates of Acinetobacter: Evidence for Emerging Species. Journal of Clinical Microbiology 48: 1445-1449. 26. Loonen AJ, Jansz AR, Kreeftenberg H, Bruggeman CA, Wolffs PF, et al. (2011) Acceleration of the direct identification of Staphylococcus aureus versus coagulase- negative staphylococci from blood culture material: a comparison of six bacterial DNA extraction methods. Eur J Clin Microbiol Infect Dis 30: 337-342. 27. Facklam R (2002) What happened to the streptococci: overview of taxonomic and nomenclature changes. Clin Microbiol Rev 15: 613-630. 28. Tendolkar PM, Baghdayan AS, Shankar N (2003) Pathogenic enterococci: new developments in the 21st century. Cell Mol Life Sci 60: 2622-2636. 29. Ngo JT, Parkins MD, Gregson DB, Pitout JD, Ross T, et al. (2013) Population-based assessment of the incidence, risk factors, and outcomes of anaerobic bloodstream infections. Infection 41: 41-48. 30. Murphy EC, Frick IM (2013) Gram-positive anaerobic cocci--commensals and opportunistic pathogens. FEMS Microbiol Rev 37: 520-553. 31. Song Y, Liu C, McTeague M, Finegold SM (2003) 16S ribosomal DNA sequence-based analysis of clinically significant gram-positive anaerobic cocci. J Clin Microbiol 41: 1363-1369. 32. Mikamo H (2011) Preface. Guidelines for diagnosis and treatment of anaerobic infections. J Infect Chemother 17 Suppl 1: 2. 33. Hardak E, Yigla M, Avivi I, Fruchter O, Sprecher H, et al. (2009) Impact of PCR-based diagnosis of invasive pulmonary aspergillosis on clinical outcome. Bone Marrow Transplant 44: 595-599. 34. Enoch DA, Ludlam HA, Brown NM (2006) Invasive fungal infections: a review of epidemiology and management options. J Med Microbiol 55: 809-818.

261

35. Eley ACAaB (2009) Viral sepsis in the pediatric intensive care unit. Journal of Pediatric Infectious Diseases 4: 161-172. 36. Dellinger RP, Levy MM, Carlet JM, Bion J, Parker MM, et al. (2008) Surviving Sepsis Campaign: international guidelines for management of severe sepsis and septic shock: 2008. Crit Care Med 36: 296-327. 37. DeVinney R, Gauthier A, Abe A, Finlay BB (1999) Enteropathogenic Escherichia coli: a pathogen that inserts its own receptor into host cells. Cellular and Molecular Life Sciences 55: 961-976. 38. Cossart P, Pizarro-Cerda J, Lecuit M (2003) Invasion of mammalian cells by Listeria monocytogenes: functional mimicry to subvert cellular functions. Trends Cell Biol 13: 23-31. 39. Galan JE, Pace J, Hayman MJ (1992) Involvement of the epidermal growth factor receptor in the invasion of cultured mammalian cells by Salmonella typhimurium. Nature 357: 588- 589. 40. VanNhieu GT, BenZeev A, Sansonetti PJ (1997) Modulation of bacterial entry into epithelial cells by association between vinculin and the Shigella IpaA invasin. Embo Journal 16: 2717-2729. 41. Sherman MP (2010) New concepts of microbial translocation in the neonatal intestine: mechanisms and prevention. Clin Perinatol 37: 565-579. 42. Alexander JW, Boyce ST, Babcock GF, Gianotti L, Peck MD, et al. (1990) The process of microbial translocation. Ann Surg 212: 496-510; discussion 511-492. 43. Yao YM, Redl H, Bahrami S, Schlag G (1998) The inflammatory basis of trauma/shock- associated multiple organ failure. Inflammation Research 47: 201-210. 44. Salzman AL, Menconi MJ, Unno N, Ezzell RM, Casey DM, et al. (1995) Nitric-Oxide Dilates Tight Junctions and Depletes Atp in Cultured Caco-2bbe Intestinal Epithelial Monolayers. American Journal of Physiology-Gastrointestinal and Liver Physiology 268: G361-G373. 45. Bruewer M, Luegering A, Kucharzik T, Parkos CA, Madara JL, et al. (2003) Proinflammatory cytokines disrupt epithelial barrier function by apoptosis-independent mechanisms. Journal of Immunology 171: 6164-6172. 46. Mogensen TH (2009) Pathogen recognition and inflammatory signaling in innate immune defenses. Clin Microbiol Rev 22: 240-273, Table of Contents. 47. Kadioglu A, Brewin H, Hartel T, Brittan JL, Klein M, et al. (2009) Pneumococcal protein PavA is important for nasopharyngeal carriage and development of sepsis. Mol Oral Microbiol 25: 50-60. 48. Deshpande RG, Khan MB, Genco CA (1998) Invasion of aortic and heart endothelial cells by Porphyromonas gingivalis. Infect Immun 66: 5337-5343. 49. Park PW, Pier GB, Preston MJ, Goldberger O, Fitzgerald ML, et al. (2000) Syndecan-1 shedding is enhanced by LasA, a secreted virulence factor of Pseudomonas aeruginosa. J Biol Chem 275: 3057-3064. 50. Winstanley C, Langille MG, Fothergill JL, Kukavica-Ibrulj I, Paradis-Bleau C, et al. (2009) Newly introduced genomic prophage islands are critical determinants of in vivo competitiveness in the Liverpool Epidemic Strain of Pseudomonas aeruginosa. Genome Res 19: 12-23.

262

51. Christaki E, Giamarellos-Bourboulis EJ (2014) The complex pathogenesis of bacteremia: from antimicrobial clearance mechanisms to the genetic background of the host. Virulence 5: 57-65. 52. Rittirsch D, Flierl MA, Ward PA (2008) Harmful molecular mechanisms in sepsis. Nat Rev Immunol 8: 776-787. 53. Van Amersfoort ES, Van Berkel TJ, Kuiper J (2003) Receptors, mediators, and mechanisms involved in bacterial sepsis and septic shock. Clin Microbiol Rev 16: 379-414. 54. Sriskandan S, Altmann DM (2008) The immunology of sepsis. J Pathol 214: 211-223. 55. Rosenthal ANaKS (2010) Cytokine Storms Systemic Disasters of Infectious Diseases. Infectious Diseases in Clinical Practice 18: 188-192. 56. Hotchkiss RS, Monneret G, Payen D (2013) Immunosuppression in sepsis: a novel understanding of the disorder and a new therapeutic approach. Lancet Infect Dis 13: 260- 268. 57. Brinkmann V, Reichard U, Goosmann C, Fauler B, Uhlemann Y, et al. (2004) Neutrophil extracellular traps kill bacteria. Science 303: 1532-1535. 58. Guimaraes-Costa AB, Nascimento MT, Froment GS, Soares RP, Morgado FN, et al. (2009) Leishmania amazonensis promastigotes induce and are killed by neutrophil extracellular traps. Proc Natl Acad Sci U S A 106: 6748-6753. 59. Ma AC, Kubes P (2008) Platelets, neutrophils, and neutrophil extracellular traps (NETs) in sepsis. J Thromb Haemost 6: 415-420. 60. Urban CF, Reichard U, Brinkmann V, Zychlinsky A (2006) Neutrophil extracellular traps capture and kill Candida albicans yeast and hyphal forms. Cell Microbiol 8: 668-676. 61. Boyd JH, Russell JA, Fjell CD (2014) The meta-genome of sepsis: host genetics, pathogens and the acute immune response. J Innate Immun 6: 272-283. 62. Lyte M (1992) The role of catecholamines in gram-negative sepsis. Med Hypotheses 37: 255-258. 63. Lyte M, Vulchanova L, Brown DR (2011) Stress at the intestinal surface: catecholamines and mucosa-bacteria interactions. Cell Tissue Res 343: 23-32. 64. Bochud PY, Glauser MP, Calandra T (2001) Antibiotics in sepsis. Intensive Care Med 27 Suppl 1: S33-48. 65. Dellinger RP, Levy MM, Carlet JM, Bion J, Parker MM, et al. (2008) Surviving Sepsis Campaign: international guidelines for management of severe sepsis and septic shock: 2008. Intensive Care Med 34: 17-60. 66. Dellinger RP, Levy MM, Rhodes A, Annane D, Gerlach H, et al. (2013) Surviving sepsis campaign: international guidelines for management of severe sepsis and septic shock: 2012. Crit Care Med 41: 580-637. 67. Schorr CA, Dellinger RP (2014) The Surviving Sepsis Campaign: past, present and future. Trends Mol Med 20: 192-194. 68. Kohanski MA, Dwyer DJ, Collins JJ (2010) How antibiotics kill bacteria: from targets to networks. Nat Rev Microbiol 8: 423-435. 69. Bochud PY, Bonten M, Marchetti O, Calandra T (2004) Antimicrobial therapy for patients with severe sepsis and septic shock: an evidence-based review. Critical Care Medicine 32: S495-512.

263

70. Roberts JA, Roberts MS, Robertson TA, Dalley AJ, Lipman J (2009) Piperacillin penetration into tissue of critically ill patients with sepsis--bolus versus continuous administration? Crit Care Med 37: 926-933. 71. Wispelwey B, Schafer KR (2010) Fluoroquinolones in the management of community- acquired pneumonia in primary care. Expert Rev Anti Infect Ther 8: 1259-1271. 72. Restrepo MI, Mortensen EM, Waterer GW, Wunderink RG, Coalson JJ, et al. (2009) Impact of macrolide therapy on mortality for patients with severe sepsis due to pneumonia. European Respiratory Journal 33: 153-159. 73. Micek ST, Welch EC, Khan J, Pervez M, Doherty JA, et al. (2010) Empiric combination antibiotic therapy is associated with improved outcome against sepsis due to Gram- negative bacteria: a retrospective analysis. Antimicrob Agents Chemother 54: 1742-1748. 74. Greer ND (2006) Tigecycline (Tygacil): the first in the glycylcycline class of antibiotics. Proc (Bayl Univ Med Cent) 19: 155-161. 75. Levin AS, Barone AA, Penco J, Santos MV, Marinho IS, et al. (1999) Intravenous colistin as therapy for nosocomial infections caused by multidrug-resistant Pseudomonas aeruginosa and Acinetobacter baumannii. Clin Infect Dis 28: 1008-1011. 76. Dellinger RP, Levy MM, Rhodes A, Annane D, Gerlach H, et al. (2013) Surviving Sepsis Campaign: international guidelines for management of severe sepsis and septic shock, 2012. Intensive Care Med 39: 165-228. 77. Patel GP, Balk RA (2012) Systemic steroids in severe sepsis and septic shock. Am J Respir Crit Care Med 185: 133-139. 78. Kumar A, Schupp E, Bunnell E, Ali A, Milcarek B, et al. (2008) Cardiovascular response to dobutamine stress predicts outcome in severe sepsis and septic shock. Crit Care 12: R35. 79. Rivers E, Nguyen B, Havstad S, Ressler J, Muzzin A, et al. (2001) Early goal-directed therapy in the treatment of severe sepsis and septic shock. N Engl J Med 345: 1368-1377. 80. Otero RM, Nguyen HB, Huang DT, Gaieski DF, Goyal M, et al. (2006) Early goal-directed therapy in severe sepsis and septic shock revisited: concepts, controversies, and contemporary findings. Chest 130: 1579-1595. 81. Bosmann M, Ward PA (2012) Therapeutic potential of targeting IL-17 and IL-23 in sepsis. Clin Transl Med 1: 4. 82. Knaus WA, Draper EA, Wagner DP, Zimmerman JE (1985) APACHE II: a severity of disease classification system. Crit Care Med 13: 818-829. 83. Wheeler MM (2009) APACHE: an evaluation. Crit Care Nurs Q 32: 46-48. 84. Vincent JL, Moreno R, Takala J, Willatts S, De Mendonca A, et al. (1996) The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. On behalf of the Working Group on Sepsis-Related Problems of the European Society of Intensive Care Medicine. Intensive Care Med 22: 707-710. 85. Zygun D, Berthiaume L, Laupland K, Kortbeek J, Doig C (2006) SOFA is superior to MOD score for the determination of non-neurologic organ dysfunction in patients with severe traumatic brain injury: a cohort study. Crit Care 10: R115. 86. Jensen JU, Heslet L, Jensen TH, Espersen K, Steffensen P, et al. (2006) Procalcitonin increase in early identification of critically ill patients at high risk of mortality. Crit Care Med 34: 2596-2602.

264

87. Jensen JU, Hein L, Lundgren B, Bestle MH, Mohr T, et al. (2012) Kidney failure related to broad-spectrum antibiotics in critically ill patients: secondary end point results from a 1200 patient randomised trial. BMJ Open 2: e000635. 88. Schuetz P, Chiappa V, Briel M, Greenwald JL (2011) Procalcitonin algorithms for antibiotic therapy decisions: a systematic review of randomized controlled trials and recommendations for clinical algorithms. Arch Intern Med 171: 1322-1331. 89. Gibot S, Kolopp-Sarda MN, Bene MC, Cravoisy A, Levy B, et al. (2004) Plasma level of a triggering receptor expressed on myeloid cells-1: its diagnostic accuracy in patients with suspected sepsis. Ann Intern Med 141: 9-15. 90. Jeong SJ, Song YG, Kim CO, Kim HW, Ku NS, et al. (2012) Measurement of plasma sTREM-1 in patients with severe sepsis receiving early goal-directed therapy and evaluation of its usefulness. Shock 37: 574-578. 91. Mach C. R. CLS, Stewart L., and Pfeltz R. F. (2006) Effectiveness of the BACTEC™ Blood Culture Resin-based Antimicrobial Removal System with Fluoroquinolones, Antifungals, Daptomycin, Tigecycline and Polymyxin B. As presented at the Interscience Conference on Antimicrobial Agents and Chemotherapy (ICAAC), 2006. 92. Kaleta EJ, Clark AE, Johnson DR, Gamage DC, Wysocki VH, et al. (2011) Use of PCR coupled with electrospray ionization mass spectrometry for rapid identification of bacterial and yeast bloodstream pathogens from blood culture bottles. J Clin Microbiol 49: 345-353. 93. Lupetti A, Barnini S, Castagna B, Capria AL, Nibbering PH (2010) Rapid identification and antimicrobial susceptibility profiling of Gram-positive cocci in blood cultures with the Vitek 2 system. Eur J Clin Microbiol Infect Dis 29: 89-95. 94. Connell TG, Rele M, Cowley D, Buttery JP, Curtis N (2007) How reliable is a negative blood culture result? Volume of blood submitted for culture in routine practice in a children's hospital. Pediatrics 119: 891-896. 95. Sautter RL, Bills AR, Lang DL, Ruschell G, Heiter BJ, et al. (2006) Effects of delayed-entry conditions on the recovery and detection of microorganisms from BacT/ALERT and BACTEC blood culture bottles. J Clin Microbiol 44: 1245-1249. 96. Fenollar F, Raoult D (2007) Molecular diagnosis of bloodstream infections caused by non- cultivable bacteria. Int J Antimicrob Agents 30 Suppl 1: S7-15. 97. Turnbaugh PJ, Ley RE, Hamady M, Fraser-Liggett CM, Knight R, et al. (2007) The human microbiome project. Nature 449: 804-810. 98. Peterson J, Garges S, Giovanni M, McInnes P, Wang L, et al. (2009) The NIH Human Microbiome Project. Genome Research 19: 2317-2323. 99. Tuttle MS, Mostow E, Mukherjee P, Hu FZ, Melton-Kreft R, et al. (2011) Characterization of bacterial communities in venous insufficiency wounds by use of conventional culture and molecular diagnostic methods. J Clin Microbiol 49: 3812-3819. 100. Sibley CD, Grinwis ME, Field TR, Eshaghurshan CS, Faria MM, et al. (2011) Culture enriched molecular profiling of the cystic fibrosis airway microbiome. PLoS One 6: e22702. 101. Sibley CD, Church DL, Surette MG, Dowd SE, Parkins MD (2012) Pyrosequencing reveals the complex polymicrobial nature of invasive pyogenic infections: microbial constituents of empyema, liver abscess, and intracerebral abscess. Eur J Clin Microbiol Infect Dis.

265

102. Liesenfeld O, Lehman L, Hunfeld KP, Kost G (2014) Molecular diagnosis of sepsis: New aspects and recent developments. Eur J Microbiol Immunol (Bp) 4: 1-25. 103. Cherkaoui A, Hibbs J, Emonet S, Tangomo M, Girard M, et al. (2010) Comparison of two matrix-assisted laser desorption ionization-time of flight mass spectrometry methods with conventional phenotypic identification for routine identification of bacteria to the species level. J Clin Microbiol 48: 1169-1175. 104. Forrest GN, Roghmann MC, Toombs LS, Johnson JK, Weekes E, et al. (2008) Peptide nucleic acid fluorescent in situ hybridization for hospital-acquired enterococcal bacteremia: delivering earlier effective antimicrobial therapy. Antimicrob Agents Chemother 52: 3558-3563. 105. Jordan JA, Jones-Laughner J, Durso MB (2009) Utility of pyrosequencing in identifying bacteria directly from positive blood culture bottles. J Clin Microbiol 47: 368-372. 106. Fredricks DN, Relman DA (1998) Improved amplification of microbial DNA from blood cultures by removal of the PCR inhibitor sodium polyanetholesulfonate. J Clin Microbiol 36: 2810-2816. 107. Schabereiter-Gurtner C, Nehr M, Apfalter P, Makristathis A, Rotter ML, et al. (2008) Evaluation of a protocol for molecular broad-range diagnosis of culture-negative bacterial infections in clinical routine diagnosis. J Appl Microbiol 104: 1228-1237. 108. Reier-Nilsen T, Farstad T, Nakstad B, Lauvrak V, Steinbakk M (2009) Comparison of broad range 16S rDNA PCR and conventional blood culture for diagnosis of sepsis in the newborn: a case control study. BMC Pediatr 9: 5. 109. Sibley CD, Parkins MD, Rabin HR, Duan K, Norgaard JC, et al. (2008) A polymicrobial perspective of pulmonary infections exposes an enigmatic pathogen in cystic fibrosis patients. Proc Natl Acad Sci U S A 105: 15070-15075. 110. Wiesinger-Mayr H, Jordana-Lluch E, Martro E, Schoenthaler S, Noehammer C (2011) Establishment of a semi-automated pathogen DNA isolation from whole blood and comparison with commercially available kits. J Microbiol Methods 85: 206-213. 111. Pletz MW, Wellinghausen N, Welte T (2011) Will polymerase chain reaction (PCR)-based diagnostics improve outcome in septic patients? A clinical view. Intensive Care Med 37: 1069-1076. 112. Kuhn C, Disque C, Muhl H, Orszag P, Stiesch M, et al. (2011) Evaluation of commercial universal rRNA gene PCR plus sequencing tests for identification of bacteria and fungi associated with infectious endocarditis. J Clin Microbiol 49: 2919-2923. 113. Wellinghausen N, Kochem AJ, Disque C, Muhl H, Gebert S, et al. (2009) Diagnosis of bacteremia in whole-blood samples by use of a commercial universal 16S rRNA gene- based PCR and sequence analysis. J Clin Microbiol 47: 2759-2765. 114. Rello J, Lisboa T, Lujan M, Gallego M, Kee C, et al. (2009) Severity of pneumococcal pneumonia associated with genomic bacterial load. Chest 136: 832-840. 115. Wallet F, Nseir S, Baumann L, Herwegh S, Sendid B, et al. (2010) Preliminary clinical study using a multiplex real-time PCR test for the detection of bacterial and fungal DNA directly in blood. Clin Microbiol Infect 16: 774-779. 116. Westh H, Lisby G, Breysse F, Boddinghaus B, Chomarat M, et al. (2009) Multiplex real- time PCR and blood culture for identification of bloodstream pathogens in patients with suspected sepsis. Clin Microbiol Infect 15: 544-551.

266

117. Tsalik EL, Jones D, Nicholson B, Waring L, Liesenfeld O, et al. (2010) Multiplex PCR to diagnose bloodstream infections in patients admitted from the emergency department with sepsis. J Clin Microbiol 48: 26-33. 118. Zhang L, Gowardman J, Morrison M, Krause L, Playford EG, et al. (2014) Molecular investigation of bacterial communities on intravascular catheters: no longer just Staphylococcus. Eur J Clin Microbiol Infect Dis. 119. Dowd SE, Wolcott RD, Sun Y, McKeehan T, Smith E, et al. (2008) Polymicrobial nature of chronic diabetic foot ulcer biofilm infections determined using bacterial tag encoded FLX amplicon pyrosequencing (bTEFAP). PLoS One 3: e3326. 120. Al Masalma M, Lonjon M, Richet H, Dufour H, Roche PH, et al. (2012) Metagenomic analysis of brain abscesses identifies specific bacterial associations. Clin Infect Dis 54: 202-210. 121. Peters BM, Jabra-Rizk MA, O'May GA, Costerton JW, Shirtliff ME (2012) Polymicrobial interactions: impact on pathogenesis and human disease. Clin Microbiol Rev 25: 193- 213. 122. Salipante SJ, Sengupta DJ, Rosenthal C, Costa G, Spangler J, et al. (2013) Rapid 16S rRNA next-generation sequencing of polymicrobial clinical samples for diagnosis of complex bacterial infections. PLoS One 8: e65226. 123. Hugonnet S, Sax H, Eggimann P, Chevrolet JC, Pittet D (2004) Nosocomial bloodstream infection and clinical sepsis. Emerg Infect Dis 10: 76-81. 124. Sibley CD, Grinwis ME, Field TR, Parkins MD, Norgaard JC, et al. (2010) McKay agar enables routine quantification of the 'Streptococcus milleri' group in cystic fibrosis patients. J Med Microbiol 59: 534-540. 125. Difco (2009) Manual of Microbiolgical Culture Media-Second Edition; Mary Jo Zimbro BS, MT (ASCP), David A. Power PD, Sharon M. Miller BS, MT (ASCP), George E. Wilson M, B.S., MT (ASCP), Julie A. Johnson BA, editors. Sparks, Maryland, USA: Becton, Dickinson and Company. 126. Brazier JS, Goldstein EJ, Citron DM, Ostovari MI (1990) Fastidious anaerobe agar compared with Wilkins-Chalgren agar, brain heart infusion agar, and brucella agar for susceptibility testing of Fusobacterium species. Antimicrob Agents Chemother 34: 2280- 2282. 127. Levy MM, Fink MP, Marshall JC, Abraham E, Angus D, et al. (2003) 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Conference. Crit Care Med 31: 1250-1256. 128. Bartram AK, Lynch MD, Stearns JC, Moreno-Hagelsieb G, Neufeld JD (2011) Generation of multimillion-sequence 16S rRNA gene libraries from complex microbial communities by assembling paired-end illumina reads. Appl Environ Microbiol 77: 3846-3852. 129. Whelan FJ, Verschoor, C.P., Stearns, J.C., Rossi, L., Luinstra, K., Loeb, M., Smieja, M., Johnstone, J., Surette, M.G., and Bowdish, D.M.E (2014) The loss of topography in the microbial communities of the upper respiratory tract in the elderly. Annals of the American Thoracic Society In Submission. 130. Martin M (2011) Cutadapt removes adapter sequences from high-throughput seqeuncing reads. EMBnet J 17: 10. 131. Masella AP, Bartram AK, Truszkowski JM, Brown DG, Neufeld JD (2012) PANDAseq: paired-end assembler for illumina sequences. BMC Bioinformatics 13: 31.

267

132. Ye Y (2010) Identification and quantification of abundant species from pyrosequences of 16S rRNA by consensus alignment. 2010 IEEE International Conference on Bioinformatics and Medicine BIBM: 153-157. 133. Wang Q, Garrity GM, Tiedje JM, Cole JR (2007) Naive Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy. Appl Environ Microbiol 73: 5261-5267. 134. DeSantis TZ, Hugenholtz P, Larsen N, Rojas M, Brodie EL, et al. (2006) Greengenes, a chimera-checked 16S rRNA gene database and workbench compatible with ARB. Appl Environ Microbiol 72: 5069-5072. 135. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, et al. (2010) QIIME allows analysis of high-throughput community sequencing data. Nat Methods 7: 335-336. 136. Lozupone C, Hamady M, Knight R (2006) UniFrac--an online tool for comparing microbial community diversity in a phylogenetic context. BMC Bioinformatics 7: 371. 137. Lozupone C, Lladser ME, Knights D, Stombaugh J, Knight R (2011) UniFrac: an effective distance metric for microbial community comparison. ISME J 5: 169-172. 138. Chen VB, Davis IW, Richardson DC (2009) KING (Kinemage, Next Generation): a versatile interactive molecular and scientific visualization program. Protein Sci 18: 2403- 2409. 139. Navas-Molina JA, Peralta-Sanchez JM, Gonzalez A, McMurdie PJ, Vazquez-Baeza Y, et al. (2013) Advancing Our Understanding of the Human Microbiome Using QIIME. Microbial Metagenomics, Metatranscriptomics, and Metaproteomics 531: 371-444. 140. Russell JA, Walley KR, Singer J, Gordon AC, Hebert PC, et al. (2008) Vasopressin versus norepinephrine infusion in patients with septic shock. New England Journal of Medicine 358: 877-887. 141. Lam NYL, Rainer TH, Chiu RWK, Lo YMD (2004) EDTA is a better anticoagulant than heparin or citrate for delayed blood processing for plasma DNA analysis. Clinical Chemistry 50: 256-257. 142. Garcia ME, Blanco JL, Caballero J, Gargallo-Viola D (2002) Anticoagulants interfere with PCR used to diagnose invasive aspergillosis. Journal of Clinical Microbiology 40: 1567- 1568. 143. Rosett W, Hodges GR (1980) Antimicrobial activity of heparin. J Clin Microbiol 11: 30-34. 144. Mermel LA, Stolz SM, Maki DG (1993) Surface antimicrobial activity of heparin-bonded and antiseptic-impregnated vascular catheters. J Infect Dis 167: 920-924. 145. Beutler E, Gelbart T, Kuhl W (1990) Interference of heparin with the polymerase chain reaction. Biotechniques 9: 166. 146. Bai X, Fischer S, Keshavjee S, Liu M (2000) Heparin interference with reverse transcriptase polymerase chain reaction of RNA extracted from lungs after ischemia-reperfusion. Transpl Int 13: 146-150. 147. de Albuquerque JP, Keim CN, Lins U (2010) Comparative analysis of Beggiatoa from hypersaline and marine environments. Micron 41: 507-517. 148. Monis PT, Giglio S, Saint CP (2005) Comparison of SYTO9 and SYBR Green I for real- time polymerase chain reaction and investigation of the effect of dye concentration on amplification and DNA melting curve analysis. Anal Biochem 340: 24-34.

268

149. Whelan FJ, Verschoor CP, Stearns JC, Rossi L, Luinstra K, et al. (2014) The Loss of Topography in the Microbial Communities of the Upper Respiratory Tract in the Elderly. Ann Am Thorac Soc. 150. De Vlaminck I, Khush KK, Strehl C, Kohli B, Luikart H, et al. (2013) Temporal response of the human virome to immunosuppression and antiviral therapy. Cell 155: 1178-1187. 151. Kommedal O, Simmon K, Karaca D, Langeland N, Wiker HG (2012) Dual priming oligonucleotides for broad-range amplification of the bacterial 16S rRNA gene directly from human clinical specimens. J Clin Microbiol 50: 1289-1294. 152. Yuan S, Cohen DB, Ravel J, Abdo Z, Forney LJ (2012) Evaluation of methods for the extraction and purification of DNA from the human microbiome. PLoS One 7: e33865. 153. de Boer R, Peters R, Gierveld S, Schuurman T, Kooistra-Smid M, et al. (2010) Improved detection of microbial DNA after bead-beating before DNA isolation. J Microbiol Methods 80: 209-211. 154. Vondracek M, Sartipy U, Aufwerber E, Julander I, Lindblom D, et al. (2011) 16S rDNA sequencing of valve tissue improves microbiological diagnosis in surgically treated patients with infective endocarditis. J Infect 62: 472-478. 155. Palka-Santini M, Cleven BE, Eichinger L, Kronke M, Krut O (2009) Large scale multiplex PCR improves pathogen detection by DNA microarrays. BMC Microbiol 9: 1. 156. Rantakokko-Jalava K, Jalava J (2002) Optimal DNA isolation method for detection of bacteria in clinical specimens by broad-range PCR. J Clin Microbiol 40: 4211-4217. 157. Abu al-Soud W, Radstrom P (2001) Purification and characterization of PCR-inhibitory components in blood cells. Journal of Clinical Microbiology 39: 485-493. 158. Clarke AJ, Dupont C (1992) O-acetylated peptidoglycan: its occurrence, pathobiological significance, and biosynthesis. Can J Microbiol 38: 85-91. 159. Shiba T, Harada S, Sugawara H, Naitow H, Kai Y, et al. (2000) Crystallization and preliminary X-ray analysis of a bacterial lysozyme produced by Streptomyces globisporus. Acta Crystallogr D Biol Crystallogr 56: 1462-1463. 160. Yokogawa K, Kawata S, Nishimura S, Ikeda Y, Yoshimura Y (1974) Mutanolysin, bacteriolytic agent for cariogenic Streptococci: partial purification and properties. Antimicrob Agents Chemother 6: 156-165. 161. Murray PR, Spizzo AW, Niles AC (1991) Clinical comparison of the recoveries of bloodstream pathogens in Septi-Chek brain heart infusion broth with saponin, Septi-Chek tryptic soy broth, and the isolator lysis-centrifugation system. J Clin Microbiol 29: 901- 905. 162. van Doorne H, van der Tuuk Adriani WP, van de Ven LI, Bosch EH, de Natris T, et al. (1998) Saponin, an inhibitory agent of carbon dioxide production by white cells: its use in the microbiologic examination of blood components in an automated bacterial culture system. Transfusion 38: 1090-1096. 163. Meex C, Neuville F, Descy J, Huynen P, Hayette MP, et al. (2012) Direct identification of bacteria from BacT/ALERT anaerobic positive blood cultures by MALDI-TOF MS: MALDI Sepsityper kit versus an in-house saponin method for bacterial extraction. J Med Microbiol 61: 1511-1516. 164. Lupetti A, Barnini S, Morici P, Ghelardi E, Nibbering PH, et al. (2013) Saponin promotes rapid identification and antimicrobial susceptibility profiling of Gram-positive and Gram-

269

negative bacteria in blood cultures with the Vitek 2 system. Eur J Clin Microbiol Infect Dis 32: 493-502. 165. Carlsson J, Frolander F, Sundquist G (1977) Oxygen tolerance of anaerobic bacteria isolated from necrotic dental pulps. Acta Odontol Scand 35: 139-145. 166. Silva VL, Carvalho MA, Nicoli JR, Farias LM (2003) Aerotolerance of human clinical isolates of Prevotella spp. J Appl Microbiol 94: 701-707. 167. Palarasah Y, Skjoedt MO, Vitved L, Andersen TE, Skjoedt K, et al. (2010) Sodium polyanethole sulfonate as an inhibitor of activation of complement function in blood culture systems. J Clin Microbiol 48: 908-914. 168. Leduc M, van Heijenoort J (1980) Autolysis of Escherichia coli. J Bacteriol 142: 52-59. 169. He X, McLean JS, Guo L, Lux R, Shi W (2014) The social structure of microbial community involved in colonization resistance. ISME J 8: 564-574. 170. Edwards U, Rogall T, Blocker H, Emde M, Bottger EC (1989) Isolation and direct complete nucleotide determination of entire genes. Characterization of a gene coding for 16S ribosomal RNA. Nucleic Acids Res 17: 7843-7853. 171. Bosshard PP, Zbinden R, Altwegg M (2002) Turicibacter sanguinis gen. nov., sp. nov., a novel anaerobic, Gram-positive bacterium. Int J Syst Evol Microbiol 52: 1263-1266. 172. Sun DL, Jiang X, Wu QL, Zhou NY (2013) Intragenomic heterogeneity of 16S rRNA genes causes overestimation of prokaryotic diversity. Appl Environ Microbiol 79: 5962-5969. 173. Willner D, Daly J, Whiley D, Grimwood K, Wainwright CE, et al. (2012) Comparison of DNA extraction methods for microbial community profiling with an application to pediatric bronchoalveolar lavage samples. PLoS One 7: e34605. 174. Dal Bello F, Hertel C (2006) Oral cavity as natural reservoir for intestinal lactobacilli. Syst Appl Microbiol 29: 69-76. 175. Ahrne S, Nobaek S, Jeppsson B, Adlerberth I, Wold AE, et al. (1998) The normal Lactobacillus flora of healthy human rectal and oral mucosa. J Appl Microbiol 85: 88-94. 176. Wong CH, Jenne CN, Lee WY, Leger C, Kubes P (2011) Functional innervation of hepatic iNKT cells is immunosuppressive following stroke. Science 334: 101-105. 177. Doyle JS, Buising KL, Thursky KA, Worth LJ, Richards MJ (2011) Epidemiology of infections acquired in intensive care units. Semin Respir Crit Care Med 32: 115-138. 178. Woo PCY, Tam DMW, Leung KW, Lau SKP, Teng JLL, et al. (2002) Streptococcus sinensis sp nov., a novel species isolated from a patient with infective endocarditis. Journal of Clinical Microbiology 40: 805-810. 179. Dakic I, Morrison D, Vukovic D, Savic B, Shittu A, et al. (2005) Isolation and molecular characterization of Staphylococcus sciuri in the hospital environment. J Clin Microbiol 43: 2782-2785. 180. Stepanovic S, Jezek P, Vukovic D, Dakic I, Petras P (2003) Isolation of members of the Staphylococcus sciuri group from urine and their relationship to urinary tract infections. Journal of Clinical Microbiology 41: 5262-5264. 181. Jordan PA, Iravani A, Richard GA, Baer H (1980) Urinary tract infection caused by Staphylococcus saprophyticus. J Infect Dis 142: 510-515. 182. Nataro JP, St Geme JW, 3rd (1988) Septicemia caused by Staphylococcus saprophyticus without associated urinary tract infection. Pediatr Infect Dis J 7: 601-602.

270

183. Van Hoovels L, Vankeerberghen A, Boel A, Van Vaerenbergh K, De Beenhouwer H (2006) First case of Staphylococcus pseudintermedius infection in a human. J Clin Microbiol 44: 4609-4612. 184. Pottumarthy S, Schapiro JM, Prentice JL, Houze YB, Swanzy SR, et al. (2004) Clinical isolates of Staphylococcus intermedius masquerading as methicillin-resistant Staphylococcus aureus. Journal of Clinical Microbiology 42: 5881-5884. 185. Lundin D, Severin I, Logue JB, Ostman O, Andersson AF, et al. (2012) Which sequencing depth is sufficient to describe patterns in bacterial alpha- and beta-diversity? Environmental Microbiology Reports 4: 367-372. 186. Kim YS, Kim SJ, Anandham R, Weon HY, Kwon SW (2013) Rhodanobacter umsongensis sp. nov., isolated from a Korean ginseng field. J Microbiol 51: 258-261. 187. Chou YJ, Sheu SY, Sheu DS, Wang JT, Chen WM (2006) Schlegelella aquatica sp. nov., a novel thermophilic bacterium isolated from a hot spring. Int J Syst Evol Microbiol 56: 2793-2797. 188. Grice EA, Kong HH, Renaud G, Young AC, Program NCS, et al. (2008) A diversity profile of the human skin microbiota. Genome Res 18: 1043-1050. 189. Lai CC, Cheng A, Liu WL, Tan CK, Huang YT, et al. (2011) Infections caused by unusual Methylobacterium species. J Clin Microbiol 49: 3329-3331. 190. Yim MS, Yau YC, Matlow A, So JS, Zou J, et al. (2010) A novel selective growth medium- PCR assay to isolate and detect Sphingomonas in environmental samples. J Microbiol Methods 82: 19-27. 191. Paradis S, Boissinot M, Paquette N, Belanger SD, Martel EA, et al. (2005) Phylogeny of the Enterobacteriaceae based on genes encoding elongation factor Tu and F-ATPase beta- subunit. Int J Syst Evol Microbiol 55: 2013-2025. 192. Brenner DJ, Fanning GR, Knutson JKL, Steigerwalt AG, Krichevsky MI (1984) Attempts to Classify Herbicola Group-Enterobacter-Agglomerans Strains by Deoxyribonucleic- Acid Hybridization and Phenotypic Tests. International Journal of Systematic Bacteriology 34: 45-55. 193. Hamady M, Knight R (2009) Microbial community profiling for human microbiome projects: Tools, techniques, and challenges. Genome Res 19: 1141-1152. 194. Brazier J, Chmelar D, Dubreuil L, Feierl G, Hedberg M, et al. (2008) European surveillance study on antimicrobial susceptibility of Gram-positive anaerobic cocci. International Journal of Antimicrobial Agents 31: 316-320. 195. Wong G, Brinkman A, Benefield RJ, Carlier M, De Waele JJ, et al. (2014) An international, multicentre survey of beta-lactam antibiotic therapeutic drug monitoring practice in intensive care units. J Antimicrob Chemother 69: 1416-1423. 196. Boey J, Wong J, Ong GB (1982) Bacteria and septic complications in patients with perforated duodenal ulcers. Am J Surg 143: 635-639. 197. Custovic A, Smajlovic J, Hadzic S, Ahmetagic S, Tihic N, et al. (2014) Epidemiological surveillance of bacterial nosocomial infections in the surgical intensive care unit. Mater Sociomed 26: 7-11. 198. Mayr FB, Yende S, Angus DC (2014) Epidemiology of severe sepsis. Virulence 5: 4-11. 199. Sartelli M, Catena F, Ansaloni L, Moore E, Malangoni M, et al. (2013) Complicated intra- abdominal infections in a worldwide context: an observational prospective study (CIAOW Study). World J Emerg Surg 8: 1.

271

200. Onderdonk AB, Weinstein WM, Sullivan NM, Bartlett JG, Gorbach SL (1974) Experimental intra-abdominal abscesses in rats: quantitative bacteriology of infected animals. Infect Immun 10: 1256-1259. 201. Corless CE, Guiver M, Borrow R, Edwards-Jones V, Kaczmarski EB, et al. (2000) Contamination and sensitivity issues with a real-time universal 16S rRNA PCR. J Clin Microbiol 38: 1747-1752. 202. Calfee DP, Farr BM (2002) Comparison of four antiseptic preparations for skin in the prevention of contamination of percutaneously drawn blood cultures: a randomized trial. J Clin Microbiol 40: 1660-1665. 203. McDonnell G, Russell AD (1999) Antiseptics and disinfectants: activity, action, and resistance. Clin Microbiol Rev 12: 147-179. 204. Rosenthal M, Goldberg D, Aiello A, Larson E, Foxman B (2011) Skin microbiota: microbial community structure and its potential association with health and disease. Infect Genet Evol 11: 839-848. 205. Rocha LA, Borges LFDE, Gontijo PP (2009) Changes in hands microbiota associated with skin damage because of hand hygiene procedures on the health care workers. American Journal of Infection Control 37: 155-159. 206. Segal LN, Alekseyenko AV, Clemente JC, Kulkarni R, Wu B, et al. (2013) Enrichment of lung microbiome with supraglottic taxa is associated with increased pulmonary inflammation. Microbiome 1: 19. 207. Trampuz A, Piper KE, Steckelberg JM, Patel R (2006) Effect of gamma irradiation on viability and DNA of Staphylococcus epidermidis and Escherichia coli. J Med Microbiol 55: 1271-1275. 208. Vincent JL, Sakr Y, Sprung CL, Ranieri VM, Reinhart K, et al. (2006) Sepsis in European intensive care units: results of the SOAP study. Crit Care Med 34: 344-353. 209. Hughes JB, Hellmann JJ, Ricketts TH, Bohannan BJM (2001) Counting the uncountable: Statistical approaches to estimating microbial diversity. Applied and Environmental Microbiology 67: 4399-4406. 210. Wichmann MW, Inthorn D, Andress HJ, Schildberg FW (2000) Incidence and mortality of severe sepsis in surgical intensive care patients: the influence of patient gender on disease process and outcome. Intensive Care Med 26: 167-172. 211. Cox MJ, Cookson WO, Moffatt MF (2013) Sequencing the human microbiome in health and disease. Hum Mol Genet 22: R88-94. 212. Turner JR (2009) Intestinal mucosal barrier function in health and disease. Nat Rev Immunol 9: 799-809. 213. Van Itallie CM, Holmes J, Bridges A, Gookin JL, Coccaro MR, et al. (2008) The density of small tight junction pores varies among cell types and is increased by expression of claudin-2. J Cell Sci 121: 298-305. 214. Madara JL, Stafford J (1989) Interferon-gamma directly affects barrier function of cultured intestinal epithelial monolayers. J Clin Invest 83: 724-727. 215. Taylor CT, Dzus AL, Colgan SP (1998) Autocrine regulation of epithelial permeability by hypoxia: role for polarized release of tumor necrosis factor alpha. Gastroenterology 114: 657-668.

272

216. Watson CJ, Hoare CJ, Garrod DR, Carlson GL, Warhurst G (2005) Interferon-gamma selectively increases epithelial permeability to large molecules by activating different populations of paracellular pores. J Cell Sci 118: 5221-5230. 217. Laupland KB (2013) Incidence of bloodstream infection: a review of population-based studies. Clin Microbiol Infect 19: 492-500. 218. Kellner JD, Vanderkooi OG, MacDonald J, Church DL, Tyrrell GJ, et al. (2009) Changing Epidemiology of Invasive Pneumococcal Disease in Canada, 1998-2007: Update from the Calgary-Area Streptococcus pneumoniae Research (CASPER) Study. Clinical Infectious Diseases 49: 205-212. 219. Pflughoeft KJ, Versalovic J (2012) Human microbiome in health and disease. Annu Rev Pathol 7: 99-122. 220. Cho I, Blaser MJ (2012) The human microbiome: at the interface of health and disease. Nat Rev Genet 13: 260-270. 221. Ahmed RA, Marrie TJ, Huang JQ (2006) Thoracic empyema in patients with community- acquired pneumonia. Am J Med 119: 877-883. 222. Sibley CD, Peirano G, Church DL (2012) Molecular methods for pathogen and microbial community detection and characterization: current and potential application in diagnostic microbiology. Infect Genet Evol 12: 505-521. 223. Chakravorty S, Helb D, Burday M, Connell N, Alland D (2007) A detailed analysis of 16S ribosomal RNA gene segments for the diagnosis of pathogenic bacteria. J Microbiol Methods 69: 330-339. 224. Lowy FD (1998) Staphylococcus aureus infections. N Engl J Med 339: 520-532. 225. Kim HK, Missiakas D, Schneewind O (2014) Mouse models for infectious diseases caused by Staphylococcus aureus. J Immunol Methods. 226. Gherardi G, Angeletti S, Panitti M, Pompilio A, Di Bonaventura G, et al. (2011) Comparative evaluation of the Vitek-2 Compact and Phoenix systems for rapid identification and antibiotic susceptibility testing directly from blood cultures of Gram- negative and Gram-positive isolates. Diagn Microbiol Infect Dis. 227. Raz R, Colodner R, Kunin CM (2005) Who are you--Staphylococcus saprophyticus? Clin Infect Dis 40: 896-898. 228. Traub WH (2000) Antibiotic susceptibility of Serratia marcescens and Serratia liquefaciens. Chemotherapy 46: 315-321. 229. Blossom D, Noble-Wang J, Su J, Pur S, Chemaly R, et al. (2009) Multistate outbreak of Serratia marcescens bloodstream infections caused by contamination of prefilled heparin and isotonic sodium chloride solution syringes. Arch Intern Med 169: 1705-1711. 230. Han YW, Shi WY, Huang GTJ, Haake SK, Park NH, et al. (2000) Interactions between periodontal bacteria and human oral epithelial cells: Fusobacterium nucleatum adheres to and invades epithelial cells. Infection and Immunity 68: 3140-3146. 231. Ramakrishna BS (2013) Role of the gut microbiota in human nutrition and metabolism. Journal of Gastroenterology and Hepatology 28: 9-17. 232. Barbier F, Andremont A, Wolff M, Bouadma L (2013) Hospital-acquired pneumonia and ventilator-associated pneumonia: recent advances in epidemiology and management. Curr Opin Pulm Med 19: 216-228. 233. Aly NY, Al-Mousa HH, Al Asar el SM (2008) Nosocomial infections in a medical-surgical intensive care unit. Med Princ Pract 17: 373-377.

273

234. Zipfel PF, Wurzner R, Skerka C (2007) Complement evasion of pathogens: common strategies are shared by diverse organisms. Mol Immunol 44: 3850-3857. 235. Shapiro N, Howell MD, Bates DW, Angus DC, Ngo L, et al. (2006) The association of sepsis syndrome and organ dysfunction with mortality in emergency department patients with suspected infection. Ann Emerg Med 48: 583-590, 590 e581. 236. Visca P, Petrucca A, De Mori P, Festa A, Boumis E, et al. (2001) Community-acquired Acinetobacter radioresistens bacteremia in an HIV-positive patient. Emerging Infectious Diseases 7: 1032-1035. 237. Seifert H (2009) The clinical importance of microbiological findings in the diagnosis and management of bloodstream infections. Clin Infect Dis 48 Suppl 4: S238-245. 238. Stearns JC, Davidson, C. J, McKeon, S., Whelan, F. J., Shah, M.E., Scryners, A.B., Bowdish, D. M. E., Kellner, J. D., and Surette, M.G. (2014) Maturation of Bacterial Communities in the Upper Respiratory Tract ISME J In Submission. 239. Watson RS, Carcillo JA, Linde-Zwirble WT, Clermont G, Lidicker J, et al. (2003) The epidemiology of severe sepsis in children in the United States. American Journal of Respiratory and Critical Care Medicine 167: 695-701. 240. Shang SQ, Chen GX, Wu YD, Du LZ, Zhao ZY (2005) Rapid diagnosis of bacterial sepsis with PCR amplification and microarray hybridization in 16S rRNA gene. Pediatric Research 58: 143-148. 241. Lucignano B, Ranno S, Liesenfeld O, Pizzorno B, Putignani L, et al. (2011) Multiplex PCR Allows Rapid and Accurate Diagnosis of Bloodstream Infections in Newborns and Children with Suspected Sepsis. Journal of Clinical Microbiology 49: 2252-2258. 242. Laforgia N, Coppola B, Carbone R, Grassi A, Mautone A, et al. (1997) Rapid detection of neonatal sepsis using polymerase chain reaction. Acta Paediatrica 86: 1097-1099. 243. Jean-Louis Vincent JC, Steven M. Opal, editor (2002) The Sepsis Text. Norwell, MA: Kluwer Academic Publishers. 244. Nagaoka K, Yanagihara K, Harada Y, Yamada K, Migiyama Y, et al. (2014) Predictors of the pathogenicity of methicillin-resistant Staphylococcus aureus nosocomial pneumonia. Respirology 19: 556-562. 245. White NR, Fowler LL (2014) Retroperitoneal and Cutaneous Necrotizing Fasciitis Secondary to Necrotizing Pancreatitis. J Emerg Med. 246. Dunne EM, Smith-Vaughan HC, Robins-Browne RM, Mulholland EK, Satzke C (2013) Nasopharyngeal microbial interactions in the era of pneumococcal conjugate vaccination. Vaccine 31: 2333-2342. 247. Brogden KA, Guthmiller JM, Taylor CE (2005) Human polymicrobial infections. Lancet 365: 253-255. 248. Setlhare G, Malebo N, Shale K, Lues R (2014) Identification of airborne microbiota in selected areas in a health-care setting in South Africa. BMC Microbiol 14: 100. 249. Pyysalo MJ, Pyysalo LM, Pessi T, Karhunen PJ, Ohman JE (2013) The connection between ruptured cerebral aneurysms and odontogenic bacteria. J Neurol Neurosurg Psychiatry 84: 1214-1218. 250. Helling TS, Evans LL, Fowler DL, Hays LV, Kennedy FR (1988) Infectious complications in patients with severe head injury. J Trauma 28: 1575-1577.

274

251. Bronchard R, Albaladejo P, Brezac G, Geffroy A, Seince PF, et al. (2004) Early onset pneumonia: risk factors and consequences in head trauma patients. Anesthesiology 100: 234-239. 252. Campbell W, Hendrix E, Schwalbe R, Fattom A, Edelman R (1999) Head-injured patients who are nasal carriers of Staphylococcus aureus are at high risk for Staphylococcus aureus pneumonia. Crit Care Med 27: 798-801. 253. Ewig S, Torres A, El-Ebiary M, Fabregas N, Hernandez C, et al. (1999) Bacterial colonization patterns in mechanically ventilated patients with traumatic and medical head injury. Incidence, risk factors, and association with ventilator-associated pneumonia. Am J Respir Crit Care Med 159: 188-198. 254. Honarm H, Falah Ghavidel M, Nikokar I, Rahbar Taromsari M (2012) Evaluation of a PCR assay to detect Enterococcus faecalis in blood and determine glycopeptides resistance genes: van a and van B. Iran J Med Sci 37: 194-199. 255. Dourmashkin RR, Dougherty RM, Harris RJ (1962) Electron microscopic observations on Rous sarcoma virus and cell membranes. Nature 194: 1116-1119. 256. Bangham AD, Horne RW, Glauert AM, Dingle JT, Lucy JA (1962) Action of saponin on biological cell membranes. Nature 196: 952-955. 257. Hickey MJ, Kubes P (2009) Intravascular immunity: the host-pathogen encounter in blood vessels. Nat Rev Immunol 9: 364-375. 258. Wartha F, Beiter K, Albiger B, Fernebro J, Zychlinsky A, et al. (2007) Capsule and D- alanylated lipoteichoic acids protect Streptococcus pneumoniae against neutrophil extracellular traps. Cell Microbiol 9: 1162-1171. 259. Sumby P, Barbian KD, Gardner DJ, Whitney AR, Welty DM, et al. (2005) Extracellular deoxyribonuclease made by group A Streptococcus assists pathogenesis by enhancing evasion of the innate immune response. Proc Natl Acad Sci U S A 102: 1679-1684. 260. Marshall BT, Long M, Piper JW, Yago T, McEver RP, et al. (2003) Direct observation of catch bonds involving cell-adhesion molecules. Nature 423: 190-193. 261. Laschke MW, Kerdudou S, Herrmann M, Menger MD (2005) Intravital fluorescence microscopy: a novel tool for the study of the interaction of Staphylococcus aureus with the microvascular endothelium in vivo. J Infect Dis 191: 435-443. 262. Finlay BB, McFadden G (2006) Anti-immunology: evasion of the host immune system by bacterial and viral pathogens. Cell 124: 767-782. 263. Gyorgy B, Szabo TG, Pasztoi M, Pal Z, Misjak P, et al. (2011) Membrane vesicles, current state-of-the-art: emerging role of extracellular vesicles. Cellular and Molecular Life Sciences 68: 2667-2688. 264. Mao DP, Zhou Q, Chen CY, Quan ZX (2012) Coverage evaluation of universal bacterial primers using the metagenomic datasets. BMC Microbiol 12: 66. 265. Huse SM, Dethlefsen L, Huber JA, Mark Welch D, Relman DA, et al. (2008) Exploring microbial diversity and taxonomy using SSU rRNA hypervariable tag sequencing. PLoS Genet 4: e1000255. 266. Kunin V, Engelbrektson A, Ochman H, Hugenholtz P (2010) Wrinkles in the rare biosphere: pyrosequencing errors can lead to artificial inflation of diversity estimates. Environ Microbiol 12: 118-123.

275

267. Champlot S, Berthelot C, Pruvost M, Bennett EA, Grange T, et al. (2010) An efficient multistrategy DNA decontamination procedure of PCR reagents for hypersensitive PCR applications. PLoS One 5. 268. Segal LN, Alekseyenko AV, Clemente JC, Kulkarni R, Wu B, et al. (2013) Enrichment of lung microbiome with supraglottic taxa is associated with increased pulmonary inflammation. Microbiome 1. 269. McLaughlin RW, Vali H, Lau PCK, Palfree RGE, De Ciccio A, et al. (2002) Are there naturally occurring pleomorphic bacteria in the blood of healthy humans? Journal of Clinical Microbiology 40: 4771-4775. 270. Nikkari S, McLaughlin IJ, Bi WL, Dodge DE, Relman DA (2001) Does blood of healthy subjects contain bacterial ribosomal DNA? Journal of Clinical Microbiology 39: 1956- 1959. 271. Gibson T, Norris W (1958) Skin fragments removed by injection needles. Lancet 2: 983- 985. 272. Brecher ME, Hay SN (2005) Bacterial contamination of blood components. Clin Microbiol Rev 18: 195-204. 273. Weinstein MP (2003) Blood culture contamination: persisting problems and partial progress. J Clin Microbiol 41: 2275-2278. 274. Bauer M, Reinhart K (2010) Molecular diagnostics of sepsis--where are we today? Int J Med Microbiol 300: 411-413. 275. Peters RP, de Boer RF, Schuurman T, Gierveld S, Kooistra-Smid M, et al. (2009) Streptococcus pneumoniae DNA load in blood as a marker of infection in patients with community-acquired pneumonia. J Clin Microbiol 47: 3308-3312. 276. Werno AM, Anderson TP, Murdoch DR (2012) Association between pneumococcal load and disease severity in adults with pneumonia. J Med Microbiol 61: 1129-1135. 277. Ho YC, Chang SC, Lin SR, Wang WK (2009) High levels of mecA DNA detected by a quantitative real-time PCR assay are associated with mortality in patients with methicillin-resistant Staphylococcus aureus bacteremia. J Clin Microbiol 47: 1443-1451. 278. Chuang YC, Chang SC, Wang WK (2010) High and increasing Oxa-51 DNA load predict mortality in Acinetobacter baumannii bacteremia: implication for pathogenesis and evaluation of therapy. PLoS One 5: e14133. 279. Ehrenstein BP, Jarry T, Linde HJ, Scholmerich J, Gluck T (2005) Low rate of clinical consequences derived from results of blood cultures obtained in an internal medicine emergency department. Infection 33: 314-319. 280. Hall KK, Lyman JA (2006) Updated review of blood culture contamination. Clinical Microbiology Reviews 19: 788-+. 281. Caporaso JG, Lauber CL, Walters WA, Berg-Lyons D, Huntley J, et al. (2012) Ultra-high- throughput microbial community analysis on the Illumina HiSeq and MiSeq platforms. ISME J 6: 1621-1624. 282. Burdino E, Ruggiero T, Allice T, Milia MG, Gregori G, et al. (2014) Combination of conventional blood cultures and the SeptiFast molecular test in patients with suspected sepsis for the identification of bloodstream pathogens. Diagn Microbiol Infect Dis. 283. Avolio M, Diamante P, Modolo ML, De Rosa R, Stano P, et al. (2014) Direct Molecular Detection of Pathogens in Blood as Specific Rule-In Diagnostic Biomarker in Patients

276

With Presumed Sepsis - Our Experience on a Heterogeneous Cohort of Patients With Signs of Infective SIRS. Shock. 284. Ziegler I, Josefson P, Olcen P, Molling P, Stralin K (2014) Quantitative data from the SeptiFast real-time PCR is associated with disease severity in patients with sepsis. BMC Infect Dis 14: 155. 285. Parikh K, Davis AB, Pavuluri P (2014) Do we need this blood culture? Hosp Pediatr 4: 78- 84. 286. Braun L, Riedel AA, Cooper LM (2004) Severe sepsis in managed care: analysis of incidence, one-year mortality, and associated costs of care. J Manag Care Pharm 10: 521- 530. 287. Maughan H, Wang PW, Diaz Caballero J, Fung P, Gong Y, et al. (2012) Analysis of the cystic fibrosis lung microbiota via serial Illumina sequencing of bacterial 16S rRNA hypervariable regions. PLoS One 7: e45791.

277